(Duderstadt, Louis-J. - Hamilton) Nuclear Reactor Analysis (1976)

Nuclear
Reactor
Analysis
James J. Duderstadt
Louis J. Hamilton
. Department of Nuclear Engineering

The University of Michigan
Ann Arbor, Michigan
JOHN WILEY & SONS, Inc.

New York / London / Sydney / Toronto
Errata
Duderstadt & Hamilton (1976 ed.)
Page Error Correction
𝐷𝑖+1 + 𝐷𝑖 1 𝐷𝑖+1 + 𝐷𝑖 1
Pg 80, eq 5-122 𝑎𝑖,𝑖+1 = − ( ) 𝑎𝑖,𝑖+1 = − ( )
Δ𝑖 Δ𝑖 + Δ𝑖+1 Δ𝑖+1 Δ𝑖 + Δ𝑖+1
𝑎𝑛,𝑛+1 𝑎𝑛,𝑛+1
Pg 184, eq 5-137
𝑎𝑛,𝑛 + 𝑎𝑛,𝑛−1 𝐴𝑛−1 𝑎𝑛,𝑛 − 𝑎𝑛,𝑛−1 𝐴𝑛−1
Pg 244, after eq 6-43 𝑠0 𝑠1
Pg 244, line 14 |𝑠𝑜 | |𝑠1 |
ΣaP ⁄Σ𝑎 − ΣaP ⁄Σ𝑎

Pg 568, eq. 15-8 Δρ = − ⋯ = Δρ = − ⋯ =
1 + 𝐿2 𝐵𝑔2 1 + 𝐿2 𝐵𝑔2
𝛾𝑃 𝛾𝑃
Pg 576, eq 15-31 Δ𝜌 = − ≅ 0.00463 Δ𝜌 = − ≅ −0.00463
𝜈 𝜈
Pg 607, Σ𝑎 for Cu 0.0326 0.326

Copyright C 1976 by John Wiley & Sons, Inc.
All rights reserved. Published simultaneously in Canada.
No part of this book may be reproduced by any means,
nor transmitted, nor translated into a machine language
without th~ written permission of the publisher.
Library of C:O,,gnn Catologi,lg in hblicatitR, Doto:

Duderstadt, James J 1942-
N uclear reactor inalysis.
Includes l>ibliographical references and indc;x.
I. Nuclear reactors--Design and construction.
2. Nuclear engineering. 3~ Nuclear fission.
I. Hamilton, Louis J., 1941- joint author.
II. Title.
TK9202.D77 621.48'32 75-20389
ISBN 0-471-22363-8
Printed in the. United States of America
10 9 8 7 6 5 4 3 2 I
Nuclear. Reactor Analysis
BOARD OF ADVISORS, ENGINEERING
A. H-S. Ang Civil Engineering-Systems

University of Illinois and Probability
Donald S. Berry Transportation Engineering
Northwestern University
James Gere Civil Engineering and
Stanford University Applied Mec~cs
J. Stuart Hunter Engineering Statistics
Princeton University
T. William lAmbe Civil Engineering-Soil
R. V. Whitman Mechanics
Massachusetts Institute of
Technology
Perry L. McCarty Environmental Engineering
Stanford University
Don T. Phillips Industrial Engineering
Purdue University
Dale F. Rudd Chemical Engineering
University of Wisconsin
Robert F. Steidel Jr. Mechanical Engineering
University of California-
Berkeley
Richard N. White Civil Engineering-Structures
Cornell University
To Anne and Jacqueline
vii
Preface
The maturation of the nuclear power industry, accompanied by a redirection in

emphasis from research and development to the large-scale installation of nuclear
power systems, has induced a corresponding change in the growing demand for
well-trained nuclear engineers. Earlier laboratory needs for research-oriented Ph.
D. reactor physicists have been largely replaced by industrial requirements for
more broadly trained nuclear engineers at the B. S. and M. S. levels who are
capable of designing, constructing, and operating large nuclear power systems.
Most universities are rapidly reorienting their own nuclear engineering programs in
response to these changes. ·
Of central importance in such programs are those introductory courses in
nuclear reactor analysis that first introduce the nuclear engineering student to the
basis scientific principles of nuclear fission chain reactions and lay a foundation for
the subsequent application of these principles to the nuclear design and analysis of
reactor cores. Although several excellent texts have been written on :the subject of
nuclear reactor theory, we have found both the.material and orientation of existing
treatments somewhat outdated for today's nuclear engineering student. For ex-
ample, the availability of large fast digital computers has had a very strong
influence on the analytical techniques used in modem nuclear reactor design. In
most cases such modern methods of reactor analysis bear little resemblance to the
precomputer techniques of earlier years. And yet most existing texts on nuclear
reactor theory dwell quite heavily on these outdated techniques, stressing analytical
methods to the near exclusion of numerical techniques and digital computation.
Furthermore most introductory texts on nuclear reactor theory present a rather
narrow view of nuclear reactor analysis by concentrating only on the behavior of
the neutron population in the reactor core. However the neutronics analysis of a
reactor core cannot be divorced from other nonnuclear aspects of core analysis
such as thermal-hydrauli~s, structural design, or economic considerations. In any,
ix
X
practical design study, the interplay between these various facets of reactor analysis
must be taken into account, and this should be reflected in modern nuclear
_engineering education programs.
We have attempted to write a reactor analysis text more tailored to the needs of
the modern nuclear engineering student. In particular, we have tried to introduce
the student to the fundamental principles governing nuclear fission chain reactions
in a manner that renders the transition to practical nuclear reactor design methods
most natural. This goal has led to a very considerable emphasis on numerical
methods suitable for digital computation. We have also stressed throughout this.
development the very close interplay between the nuclear analysis of a reactor core
and those nonnuclear aspects of core analysis, such as thermal-hydraulics or
materials studies, which play a major role in determining a reactor design. Finally,
we have included illustrations of the various concepts that we develop by consider-
ing a number of more practical problems arising in the nuclear design of various
types of power reactors.
The text has been organized into four parts. In Part I we present a relatively
elementary and qualitative discussion of the basic concepts involved in nuclear
fission chain reactions, incluq.ing a brief review of the relevant nuclear physics and
a survey of modern power reactors. In Part 2 we develop in some detail a
particularly simple and useful model of nuclear reactor behavior by assuming that
the neutrons sustaining the fission chain reaction diffuse from point to point in the
reactor in such a way that their energy and direction oi motion can be ignored
(one-speed diffusion theory). In Part 3 we generalize this model to develop the
primary tool of nuclear reactor analysis, multigroup diffusion theory. In Part 4 we
illustrate these methods of nuclear reactor analysis by considering several impor-
tant applications in nuclear reactor design.
We include a wide variety of problems to further illustrate these concepts, since
we have learned by past experience that such problems are essential for and
adequate understanding of nuclear reactor theory. The degree of difficulty spanned
by the problems is enormous, ranging from simple formula substitution to prob-
lems requiring extensive outside reading by the student. Since most universities
have at their disposal large time-sharing computer systems, we have not hesitated
to include problems that require rudimentary programming experience as well as
access to digital computers. Also, since more and more nuclear engineering
programs have access to libraries of the more common reactor design computer·
codes, we also include problems utilizing such codes. It is our hope that the volume
and variety of problems are sufficient to provide the instructor with the opportun-
ity to select those problems most appropriate to this particular needs.
The same broad scope also characterizes the material included in the text. We
have attempted to prepare a text suitable for a wide variety of students, including
not only senior-year B. S. or first-year M. S. nuclear engineering students, but for
students from other disciplines as well, such as electrical and mechanical engineer-
ing, physics, and chemistry, who desire an exposure to the principles underlying
nuclear reactor design and operation. Hence the text has been written with the
intent of providing suitable material for a student with only a modest background
in modern physics and applied mathematics, such as would be included in the
curriculum of most undergraduate engineering or science students. To this end,
much of the early material is presented in an essentially self-contained fashion.
xi
However it is our intent that this text should also serve as a useful reference in
more advanced courses or for practicing engineers, and. therefore we include more
advanced material when appropriate (particularly in later chapters). In such cases
we provide numerous references to supplement our treatment. Certainly the en-
tirety of the material presented would be overwhelming for a one- or even two-term
course. (We have distributed the material among three terms at Michigan.) Rather
we have sought to provide a text sufficiently flexible for a wide variety of
. applications. Hence we do not apologize for the scope and occasionally more
rugged terrain covered by the text, since the instructor can always choose a less
demanding route by selecting an appropriate subset of this material.
The units employed in this text are the International System of Units (SI), their
derivatives, and several non-SI units (such as the electronvolt or the barn) which
are recognized by the International Organization for Standardization for use in
special fields. Unfortunately, the vast majority of nuclear engineering literature
published in the United States prior to 1975 makes use of British units. To assist
the reader in coordinating this literature with the SI units used in this text, we have
included brief tables of the appropriate conversion factors in Appendix I.
As with any text at this level, very little of the material presented has originated
with the authoi:-s, but rather has been accumulated and assimilated from an
enormous variety of sources, some published, many unpublished. We generally
attempt to present material in the fashion we have found most successful from our
own teaching experience, frequently sacrificing originality for effectiveness of
presentation. Throughout the text we attempt to acknowledge the sources of our
material.
However, we would particularly like to acknowledge the impact made upon this
work by several of our associates. In presentation, we have chosen to utilize the
very appealing pedagogical approach of P. F. Zweifel by introducing as much of
reactor analysis as possible within the one-speed diffusion approximation before
continuing to discuss neutron energy-dependence. Our attempts to relate the basic
concepts of nuclear reactor theory to practical reactor analysis have relied heavily
upon numerous discussions, lecture materials, admonitions, and advice of .Harvey
Graves, Jr., to whom we are particularly grateful. We would also like to acknowl-
edge the assistance of a. group of truly exceptional former students (and now
practicing nuclear engineers) including Thomas Craig, David Chapin, Lawrence
Emmons, Robert Grossman, Ronald Fleming, Robert McCredy, William Martin,
Pltjlip Meyer, Sidney Karin, Russell Mosteller, William G. Price, Jr., Robert
Steinke, and Paul J. Turinsky, as well as scores of other students who have
suffered, sweated, and occasionally cursed their way through the many sets of
lecture notes which led to this text.
It is particularly important to acknowledge the considerabl~ assistance provided
by other staff members at Michigan including A. Ziya Akcasu; David Bach, John
Carpenter, Chihiro Kikuchi, Glenn Knoll, John Lee, Robert Martin, Richard K.
Osborn, Fred Shure, and George C. Summerfield. Also we should acknowledge
that much of the motivation and inspiration for this effort originated at Caltech
with Harold Lurie and Noel Corngold.aiid at Berkeley with Virgil Schrock. But,
above all, we would like to. thank William Kerr, without whose continued en,-
couragement and support this work would have never been completed.
We also wish to express our gratitude to Miss Pam Hale for her Herculean
xii
efforts (and cryptographic abilities) m helping prepare the vanous drafts and
manuscripts which led to this text.
James J. Duderstadt
Louis J. Hamilton
ANN ARBOR, MICHIGAN

FEBRUARY 1975
Contents
PARTl
Introductory Concepts
of Nuclear Reactor
Analysis
CHAPTER 1 An Introduction to Nuclear Power Generation 3
I.VI Nuclear Fission Reactors 4

II. Role of the Nuclear Engineer 6
III. Scope of the Text 7
CHAPTER 2 The Nuclear Physics of Fission Chain Reactions 10
I. Nuclear Reactions 12
II. Nuclear Fission 54
CHAPTER 3 Fission Chain Reactions and Nuclear Reactors-

an Introduction 74
I. The MultipJication Factor and Nuclear Criticality 74

II. An Introd"4ction to Nuclear Power Reactors 88
III. Nuclear Reactor Design 96
xiii
xiv
PART2
The One-Speed
Diffusion Model
of a Nuclear Reactor
CHAPTER 4 Neutron Transport 103
I. Introductory Concepts 105

II. The Neutron Transport Equation 111
III. Direct Numerical Solution of the Transport Equation 117
IV. The Diffusion Approximation 124
CHAPTER 5 The One-Speed Diffusion Theory Model 149
I. The One-Speed Diffusion Equation 150

II. Neutron Diffusion in Nonmultiplying Equation 157
III. The One-Speed Diffusion Model of a Nuclear Reactor 196
IV. Reactor Criticality Calculations 214
V. Perturbation Theory 219
CHAPTER 6 Nuclear Reactor Kinetics 233
I. The Point Reactor Kinetics Model 235

II. Solution of the Point Reactor Kinetics Equations 241
III. Reactivity Feedback and Reactor Dynamics 257
IV. Experimental Determination of Reactor Kinetic Parameters 268
V. Spatial Effects in Reactor Kinetics 276
PART3
The Multigroup
Diffusion Method
CHAPTER 7 Multigroup Diffusion Theory 285
I. A Heuristic Derivation of the Multigroup Diffusion Equations 286

II. Derivation of the Multigroup Equations from Energy-Dependent
Diffusion Theory 288
III. Simple Applications of the Multigroup Diffusion Model 295
IV. Numerical Solution of the Multigroup Diffusion Equations 301
xv
V. Multigroup Perturbation Theory 308

VI. Some Concluding Remarks 311
CHAPTER 8 Fast Spectrum Calculations· and Fast Group

Constants 315
I. Neutron Slowing Down in an Infinite Medium 317

II. Resonance Absorption (Infinite Medium) 332
III. Neutron Slowing Down in Finite Media 347
IV. Fast Spectrum Calculations and Fast Group Constants 358
- CHAPTER 9 Thermal S ~ Calculations and Thermal

Group Constants · 375
I. General Features of Thermal Neutron Spectra 377

U. Approximate Models of Neutron Thermalization 384
III. General Calculations of Thermal Neutron Spectra 392
CHAPTER 10 Cell Calculations for Heterogeneous Core Lattic~ 398
I. Lattice Effects in Nuclear Reactor Analysis 399

II. Heterogeneous Effects in Thermal Neutron Physics 413
Ill. Heterogeneous Effects in Fast Neutron Physics 427
IV. Some Concluding Remarks in the Multigroup Diffusion Methods 439
PART4
An Introduction.-
to Nuclear Reactor
Core Design
CHAPTER 11 General Aspects of Nuclear Reactor Core Design 447
I. Nuclear Core Analysis 448

II. Other Areas of Reactor Core Analysis 454
III. Reactor Calculation Models 460
CHAPTER 12 Thermal-hydraulic Analysis of Nuclear Reactor Cores 467
I. Introduction 467
II. Power Generation in Nuclear Reactor Cores 473
xvi
III. Radial Heat Conduction in Reactor Fuel Elements 475

IV. Forced Convection Heat Transfer in Single-Phase Coolants 483
V. Boiling Heat Transfer in Nuclear Reactor Cores 489
VI. Hydrodynamic Core Analysis 498
VII. Thermal-hydraulic Core Analysis 502
CHAPTER 13 The Calculation of Core Power Distributions 515
I. Static Multigroup Diffusion Calculations 516

II. Interaction of Thermal-hydraulics, Neutronics, and Fuel Depletion 519
III. Parameterization of Few-Group Constants 522
IV. Flux Synthesis Methods 525
V. Local Power-Peaking Effects 529
CHAPTER 14 Reactivity Control 537
I. Introduction 537
II. Movable Control Rods 540
III. Burnable Poisons 551
IV. Chemical Shim 554
V. Inherent Reactivity Effects 556
CHAPTER 15 Analysis of Core Composition Changes 566
I. Fission Product Poisoning 567

II. Fuel Depletion Analysis 580
III. Nuclear Fuel Management 589
xvU
APPENDICES
· A. Some Useful Nuclear Data 685
I. Miscellaneous. Physical Constants 605

II. Some Useful Conversion Factors 605
III. 2200 m/sec Cross Sections for Naturally Oc~urring Elements 606
IV. 2200 m/sec Cross Sections of Special Interest 610
B. Some Useful Mathematical Formulas 611
C. Step Functions, Delta Functions, and Other Exotic Beasts 613
I. Introduction 613
II. Properties of the Dirac 6-Function 614
A. Alternative Representations 614
B. Properties 614
C. Derivatives 615
D. Some Properties of Special Functions 616
E. Some Assorted Facts on Linear Operators . 621
I. Scalar Products 621

II. Linear Operators 622
III. Linear Vector Spaces 622
IV. Properties of Operators 623
V. Differential Operators 624
F. An Introduction to Matrices and Matrix Algebra 626
I. Some Definitions 626

II. Matrix Algebra 628
G. An Introduction to Laplace Transforms 629
I. Motivation 629
II. "Cookbook" Laplace Transforms 631
H. Typical Nuclear Power Reactor Data 6.14
I. Units Utilized in Text 6.16
Index 6.19
Nuclear Reactor Analysis
1
Introductory Concepts
. of Nuclear Reactor
Analysis
1
An Introduction to
Nuclear Power
Generation
It has been more than three decades since the first nuclear reactor achieved a
critical fission chain reaction beneath the old Stagg Field football stadium at the
University ·of Chicago. Since that time an extensive worldwide effort has been
directed toward nuclear reactor research and development in an attempt to harness
the enormous energy contained within the atomic nucleus for the peaceful genera-
tion of power. Nuclear reactors have evolved from an embryonic research tool into
the mammoth electrical generating units that drive hundreds of central-station
power plants around the world today. The recent shortage of fossil fuels has made
it quite apparent that nuclear fission reactors will play a dominant role in meeting
man's energy requirements for decades to come.
For some time electrical utilities have been ordering and installing nuclear plants
in p_reference to fossil-fueled units. Such plants are truly enormous in size, typically
generating over 1000 MWe (megawatts-electric) of electrical power (enough to
supply the electrical power needs of a city of 400,000 people) and costing more
than half a billion dollars. It is anticipated that some 1000 nuclear power plants
will be installed in the United States alone by the year 2000 with an electrical
generating capacity of about 1,000,000 MWe and a capital investment of more than
$600 billion, 1 with this pattern being repeated throughout the world. The motiva-
tion for such a staggering commitment to nuclear power involves a number of
factors that include not only the very significant economic and operational advan-
tages exhibited by nuclear plants over conventional sources of power, but their
substantially lower environmental impact and vastly larger fuel resources as wen. 2- 7
The dominant role played by nuclear fission reactors in the generation of
.electrical power can be expected to continue well into the next century. Until at
least A.D. 2000, nuclear power will represent the only viable alternative to
3
4 / INTRODUCTORY CONCEPTS OF NUCLEAR POWER REACTOR ANALYSIS
fossil-fueled plants for most nations. 2•3 The very rapid increase in fossil-fuel costs
that has accompanied their dwindling reserves has led to a pronounced cost
advantage for nuclear plants which is expected to widen even further during the
next few decades.4•6 Of course, there are other longer-range alternatives involving
advanced technology such as solar power, geothermal power, and controlled
thermonuclear fusion. However the massive practical implementation of such
alternatives, if proven feasible, could probably not occur until after the tum of the
century since experience has shown that it takes several decades to shift the energy
industry from one type of fuel to another,2 due both to the long operating lifetime
of existing power machinery and the long lead times needed to redirect manufac-
turing capability. Hence nuclear fission power will probably be the dominant new
source of electrical power during the productive lifetimes of the present generation
of engineering students.
I. NUCLEAR FISSION REACTORS
The term nuclear reactor will be used in this text to refer to devices in which
controlled nuclear fission chain reactions can be maintained. (This restricted
definition may offend that segment of the nuclear community involved in nuclear
fusion research, but since even a prototype nuclear fusion reactor seems several
years down the road, no confusion should result.) In such a device, neutrons are
used to induce nuclear fission reactions in heavy nuclei. These nuclei fission into
lighter nuclei (fission products), accompanied by the release of energy (some 200
MeV per event) plus several additional neutrons. These fission neutrons can then be
utilized to induce still further fission reactions, thereby inducing a chain of fission
events. In a very narrow sense then, a nuclear reactor is simply a sufficiently large
mass of appropriately fissile material (e.g., 235 U or 239 Pu) in which such a controlled
fission chain reaction can be sustained. Indeed a small sphere of 235 U metal slightly
over 8 cm in radius could support such a chain reaction and hence would be
classified as a nuclear reactor.
However a modem power reactor is a considerably more complex beast. It must
not only contain a lattice of very carefully refined and fabricated nuclear fuel, but
must as well provide for cooling this fuel during the course of the chain reaction as
fission energy is released, while maintaining the fuel in a very precise geometrical
arrangement with appropriate structural materials. Furthermore some mechanism
must be provided to control the chain reaction, shield the surroundings of the
reactor from the intense nuclear radiation generated during the fission reactions,
and provide for replacing nuclear fuel assemblies when the fission chain reaction
has depleted their concentration of fissile nuclei. If the reactor is to produce power
in a useful fashion, it must also be designed to operate both economically and
safely. Such engineering constraints render the actual nuclear configuration quite
complex indeed (as a quick glance ahead to the illustrations in Chapter 3 will
indicate).
Nuclear reactors have been used for over 30 years in a variety of applications.
They are particularly valuable tools for nuclear research since they produce
copious amounts of nuclear radiation, primarily in the form of neutrons and
gamma rays. Such radiation can be used to probe the microscopic structure and
dynamics of matter (neutron or gamma spectroscopy).
INTRODUCTION TO NUCLEAR POWER GENERATION / 5
The radiation produced by reactors can also be used to transmute nuclei into
artificial isotopes that can then be used, for example, as radioactive tracers in
industrial or medical applications. Reactors can use the same scheme to produce
nuclear fuel from nonfissile materials. For example, 238 U can be irradiated by
neutrons in a reactor and transmuted into the nuclear fuel 239 Pu. This is the process
utilized to "breed" fuel in the fast breeder reactors currently being developed for
commercial application in the next decade.
Small, compact reactors have been used for propulsion in submarines, ships,
aircraft, and rocket vehicles. Indeed the present generation of light water reactors
used in nuclear power plants are little more than the very big younger brothers of
the propulsion reactors used in nuclear submarines. Reactors can also be utilized as
small, compact sources of long-term power, such as in remote polar research
stations or in orbiting satellites.
Yet by far the most significant application of nuclear fission reactors is in large,
central station power plants. A nuclear power plant is actually very similar to a
fossil-fueled power plant, except that it replaces the coal or oil-fired boiler by a
nuclear reactor, which generates heat by sustaining a fission chain reaction in a
suitable lattice of fuel material. Of course, there are some dramatic differences
between a nuclear reactor and, say, a coal-fired boiler. However the useful quantity
produced by each is high temperature, high pressure steam that can then be used to
run turbogenerators and produce electricity. At the center of a modern nuclear
plant is the nuclear steam supply system (NSSS), composed of the nuclear reactor,
its associated coolant piping and pumps, and the heat exchangers ("steam genera-
tors") in which water is turned into steam. (A further glance at the illustrations in
Chapter 3 will provide the reader with some idea of these components.) The
remainder of the power plant is rather conventional.
Yet we must not let the apparent similarities between nuclear· and fossil-fueled
power plants overshadow the very significant differences between the two systems.
For example, in a nuclear plant sufficient fuel must be inserted into the reactor
core to allow operation for very long periods of time (typically one year). The
nuclear fuel cycle itself is extremely complex, involving fuel refining, fabrication,
reprocessing after utilization in the reactor, and eventually the disposal of radioac-
tive fuel wastes. The safety aspects of nuclear plants are also quite different, since
one must be concerned with avoiding possible radiological hazards. Furthermore
the licensing required by a nuclear plant before construction or operation demands
a level of sophisticated analysis totally alien to fossil-fueled plant design.
Therefore even though the NSSS contributes only a relatively modest fraction of
the total capital cost of a nuclear power plant (presently about 20%), it is of central
concern since it not only dictates the detailed design of the remainder of the plant,
but also the procedures required in plant construction and operation. Furthermore
it is the low fuel costs of the NSSS that are responsible for the economic
advantages presently enjoyed by nuclear power generation.
The principal component of the NSSS is, of course, the nuclear reactor itself. A
rather wide variety of nuclear reactors are in operation today or have been
proposed for future development. Reactor types can be characterized by a number
of features. One usually distinguishes between those reactors whose chain reactions
are maintained by neutrons with characteristic energies comparable to the energy
of thermal vibration of the atoms comprising the reactor core (thermal reactors)
and reactors in which the average neutron energy is more characteristic of the
much higher energy neutrons released in a nuclear fission reaction (fast reactors).
Yet another common distinction refers to the type of coolant used in the reactor.
In the United States, and indeed throughout the world, the most popular of the
present generation of reactors, the light water reactor (LWR) uses ordinary water
as a coolant. Such reactors operate at very high pressures (approximately 70-150
bar) in order to achieve high operating temperatures while maintaining the water in
its liquid phase. If the water is allowed to boil in the core, the reactor is referred to
as a boiling water reactor (BWR), while if the system pressure is kept sufficiently
high to prevent bulk boiling (155 bar), the reactor is known as a pressurized water
reactor (PWR). Such reactors have benefited from a well-developed technology and
performance experience achieved in the nuclear submarine program.
A very similar type of reactor uses heavy water (D20) either under high pressure
as a primary coolant or simply to facilitate the fission chain reaction. This
particular concept has certain nuclear advantages that allow it to utilize low-
enrichment uranium fuels (including natural uranium). It is being developed in
Canada in the CANDU series of power reactors and in the United Kingdom as
steam generating heavy water reactors (SGHWR).
Power reactors can also utilize gases as coolants. For example, the early MAG-
NOX reactors developed in the United Kingdom used low-pressure CO2 as a
coolant. A particularly attractive recent design is the high-temperature gas-cooled
reactor (HTGR) manufactured in the United States which uses high-pressure
helium. Related gas-cooled reactors include the pebble-bed concept and the ad-
vanced gas cooled reactors (AGR) under development in Germany and the United
Kingdom, respectively. _
All of the above reactor types can be classified as thermal reactors since their
fission chain reactions are maintained by low-energy neutrons. Such reactors
comprise most of the world's nuclear generating capacity today, and of these the
LWR is most common. It is generally agreed that the LWR will continue to
dominate the nuclear power industry until well into the 1980s, although its market
may tend to be eroded somewhat by the successful development of the HTGR
or advanced heavy water reactors.
However as we will see in the next chapter, there is strong incentive to develop a
fast reactor which will breed new fuel while producing power, thereby greatly -
reducing nuclear fuel costs. Such fast breeder reactors may be cooled by either
liquid metals [the liquid metal-cooled fast breeder reactor (LMFBR)] or by helium
[the gas-cooled fast breeder reactor (GCFR)]. Although fast breeder reactors are
not expected to make an appreciable impact on the nuclear power generation
market until after 1990, their development is actively being pursued throughout the
world today.
Numerous other types of reactors have been proposed and studied-some even
involving such exotic concepts as liquid or gaseous fuels. Although much of the
analysis presented in this text is applicable to such reactors, our dominant concern
is with the solid-fuel reactors cooled by either water, sodium, or helium, since these
will comprise the vast majority of the power reactors installed during the next
several decades.
II. ROLE OF THE NUCLEAR ENGINEER

The nuclear engineer will play a very central role in the development and
application of nuclear energy since he is uniquely characterized by his ability to
assist in both the nuclear design of fission reactors and their integration into large
power systems. In the early days of the reactor industry a nuclear engineer was
usually regarded as a Ph. D.-level reactor physicist primarily concerned with
nuclear reactor core research and design. Today, however, nuclear engineers are
needed not only by research laboratories and reactor manufacturers to develop and
design nuclear reactors, but also by the electrical utilities who buy and operate the
nuclear power plants, and by the engineering companies who build the power
plants and service them during their operating lifetimes.
Hence an understanding of core physics is not sufficient for today's nuclear
engineer. He must also learn how to interface his specialized knowledge of nuclear
reactor theory with the myriad of other engineering demands· made upon a nuclear
power reactor and with a variety of other disciplines, including mechanical,
electrical, and civil engineering, metallurgy, and even economics (and politics), just
as specialists of these other disciplines must learn to interact with nuclear en-
gineers. In this se~se, he must recognize that the nuclear analysis of a reactor is
only one facet to be considered in nuclear power engineering. To study and master
it outside of the context of these other disciplines would be highly inadvisable. In
the same sense, those electrical, mechanical, or structural engineers who find
themselves involved in various aspects of nuclear power station design (as ever
increasing numbers are) will also find some knowledge of nuclear reactor theory
useful in the understanding of nuclear components and interfacing with nuclear
design.
Future nuclear engineers must face and solve complex problems such as those
involved in nuclear reactor safety, environmental impact assessment, nuclear power
plant reliability, and the nuclear fuel cycle, which span an enormous range of
disciplines. They must always be concerned with the economic design, construc-
tion, and operation of nuclear plants consistent with safety and environmental
constraints. An increasing number of nuclear engineers will find themselves con-
cerned with activities such as quality assurance and component standardization as
the nuclear industry continues to grow and mature, and of course all of these
problems must be confronted and handled in the public arena.
III. SCOPE OF THE TEXT
Our goal in this text is to develop in detail the underlying theory of nuclear
fission reactors in a manner accessible to both prospective nuclear engineering
students and those engineers from other disciplines who wish to gain some
exposure to nuclear reactor engineering. In every instance we attempt to begin with
the fundamental scientific principles governing nuclear fission chain reactions and
then carry these fundamental concepts through to the level of realistic engineering
applications in nuclear reactor design. During this development we continually
stress the interplay between the nuclear analysis of a reactor core and the parallel
nonnuclear design considerations that must accompany it in any realistic nuclear
reactor analysis.
We must admit a certain preoccupation with nuclear power reactors simply
because most nuclear engineers will find themselves involved in the nuclear power
industry. This will be particularly apparent in the examples we have chosen to
discuss and the problems we have emphasized. However since our concern is
always with emphasis of fundamental concepts over specific applications, most of
the topics we develop have a much broader range of validity and would apply
equally well to the analysis of other types of nuclear reactors. And although our
principal target is the prospective nuclear engineer, we would hope that engineers
from other disciplines would also find this text useful as an introduction to the
concepts involved in nuclear reactor analysis.
The present text develops in four progressive stages. Part 1 presents a very brief
introduction to those concepts from nuclear physics relevant to nuclear fission
reactors. These topics include not only a consideration of the nuclear fission
process itself, but also a consideration of the various ways in which neutrons, which
act as the carrier of the chain reaction, interact with nuclei in the reactor core. We
next consider from a qualitative viewpoint the general concepts involved in
studying nuclear chain reactions. Part 1 ends with an overview of nuclear reactor.
engineering, including a consideration of the various types of modern nuclear
reactors, their principai components, and a qualitative discussion of nuclear reactor
design.
Parts 2-4 are intended to develop the fundamental scientific principles underly-
ing nuclear reactor analysis and to apply these principles for derivation of the most
common analytical tools used in contemporary reactor design. By way of illustra-
ticn, these tools are then applied to analyze several of the more common and
significant problems facing nuclear engineers.
Part 2 develops the mathematical theory of neutron transport in a reactor. It
begins with the most general description based on the neutron transport equation
and briefly (and very qualitatively) reviews the standard approximations to this
equation. After this brief discussion, we turn quickly to the development of the
simplest nontrivial model of a nuclear fission reactor, that based upon one-speed
neutron diffusion theory. This model is used to analyze both the steady state and
time-dependent behavior of nuclear reactors, since although the model has very
limited validity in practical reactor analysis, it does illustrate most of the concepts
as well as the calculational techniques used in actual reactor design.
In Part 3, we develop the principal tool of modern nuclear reactor design, the
multigroup diffusion model. Particular attention is devoted to the calculation of the
multigroup constants appearing in these equations, as well as to the practical
numerical solution of the equations themselves.
In Part 4, we attempt to give an overview of the methods used in nuclear reactor
core design. In particular, we consider the application of the concepts and tools
developed in the earlier sections to a variety of problems faced by the nuclear
engineer, including criticality calculations, the determination of core power distri-
butions and thermal-hydraulics analysis, bumup and control studies, and fuel-
loading requirements. While certainly incomplete, we do feel that the problems we
have chosen to examine are representative of those encountered in nuclear reactor
design and serve to illustrate the concepts developed in the earlier chapters of the
text.
REFERENCES
l. The Nuclear Industry, USAEC Report WASH-1174-73, 1973. The USAEC publishes an
annual report on the status of private nuclear industry within the United States. These
reports provide a very comprehensive survey of the growth of the nuclear power industry.
2. Chauncey Starr, Sci. Amer. 225, 39 (1971).
3. Fourth International Conference on the Peaceful Uses of Atomic Energy (International
Atomic Energy Agency, Geneva, 1972), Vol. 1.
4. A study of base-load alternatives for the Northeast Utilities System, Report to the Board
of Trustees of Northeast Utilities, Arthur D. Little, Inc. (1973).
5. Nuclear Fuel Resources and Requirements, USAEC Report WASH-1243 (1973); Nucl.
Ind. 21 (2) (1974); Nucl. News 17, (5) (1974).
6. David J. Rose, Science 184, 351 (1974).
7. An Assessment of Accident Risks in U. S. Commercial Nuclear Power Plants, USAEC
Report WASH-1400 (1974).
Most current information concerning the nuclear power industry appears in a number of
journal publications. Although the number of journals appearing in the field of nuclear
science and engineering is quite voluminous, a brief list of journals of more general interest
would include:
Nuclear Engineering International (Europressatom): A British journal distinguished by its
elaborate, multicolored diagrams of nuclear power plants.
Nuclear Industry (Atomic Industrial Forum): A monthly news magazine written more from
the viewpoint of ~e consumer of nuclear power products-namely, the electrical utilities.
Nuclear News (American Nuclear Society): A monthly news magazine published by the
American Nuclear Society, the principal technical organization of nuclear engineering in
the United States.
Nuclear Safety (Office of Information Services, U. S. Atomic Energy Commission): A
journal highlighting recent developments in the field of nuclear reactor safety.
Nuclear Science and Engineering (American Nuclear Society): The principal technical
research journal of nuclear engineering.
Other more research-oriented journals in the field of nuclear science and engineering
include:
Annals of Nuclear Science and Engineering (formerly Journal of Nuclear Energy) (Pergamon,
New York).
Journal of Nuclear Science and Technology (Atomic Energy Society of Japan).
Nuclear Engineering and Design (North-Holland, Amsterdam).
Nuclear Technology (American Nuclear Society).
Soviet Atomic Energy (Consultant's Bureau).
References on nuclear reactor theory include:
Bell, G. I., and Glasstone, S., Nuclear Reactor Theory, Reinhold, New York (1970).
Henry, Allan F., Nuclear Reactor Analysis, M.I.T. Press, Cambridge, Mass. (1975).
Glasstone, S. and Edlund, M. C., The Elements of Nuclear Reactor Theory, Reinhold, New
York ( 1952).
Glasstone, S. and Sesonske, A., Nuclear Reactor Engineering, 2nd Edition, Van Nostrand,
Princeton, N.J. (1975).
Lamarsh, J. R., Introduction to Nuclear Reactor Theory, Addison-Wesley, Reading, Mass.
(1966).
Lamarsh, John R., Introduction to Nuclear Engineering, Addison-Wesley, Reading, Mass.
(1975).
Meghreblian, R. V. and Holmes, D. K., Reactor Analysis, McGraw-Hill, New York (1960).
Murray, Raymond L., Nuclear Energy, Pergamon Press, New York (1975).
Weinberg, A. M. and Wigner, E. P., The Physical Theory of Neutron Chain Reactors, Chicago
U. P. (1958).
Zweifel, P. F., Reactor Physics, McGraw-Hill, New York (1973).
2
The Nuclear Physics of
Fission Chain Reactions
The primary objective in the design and operation of a nuclear reactor is the
utilization of the energy or radiation released by a controlled chain reaction of
nuclear fission events maintained within the reactor core. Such fission reactions
occur when a heavy atomic nucleus such as 235 U splits or fissions into two lighter
nuclei with an attendant release of both energy and radiation. Yet just how are
such fission reactions ·induced in a reactor? The .rate at which such naturally
occurring heavy nuclei will fission on their own (spontaneous fission) is very slow. It
should also be apparent that one cannot simply smash two nuclei together to
induce such a reaction, since the large electrical charges of heavy nuclei would lead
to a very strong repulsion. A more attractive idea is to slam a neutral particle
(which doesn't feel the nuclear charge) into a big, "overweight" nucleus and hope
that this splits it. An ideal candidate for the incident particle is the neutron. Indeed
experiments have shown that certain nuclei have an enormous appetite for
neutrons, but after swallowing them suffer a case of violent indigestion that results
in their fission. As an example of such a reaction, consider a neutron incident upon
a 235 U nucleus:
Neutron+ 235 U-""7'Jission products+ more neutrons + energy.
The products of such a reaction (e.g., lighter nuclei, neutrons, and gammas) emerge
with very large kinetic energy (some 200 MeV) which is then converted into heat as
they slow down by banging into neighboring atoms in the reactor fuel. It is this
heat energy that one utilizes to produce steam and eventually electrical power in a
nuclear power plant.
Yet just as significantly the fission reaction kicks loose a few neutrons that may
then go on to induce more fission reactions. Hence we can use the neutrons to
THE NUCLEAR PHYSICS OF FISSION CHAIN REACTIONS / 11
propagate a chain of fission reactions. In this sense, then, the neutron plays the role
of the chain carrier while the fission reactions supply the desired energy.
However there are other possible nuclear reactions a neutron can undergo that
do not lead to fission and hence are unproductive in nature. Indeed since there are
usually two or three neutrons emitted in each fission reaction, it should be apparent
that if each neutron resulted in another fission, the chain reaction would quickly
grow without bound. One such parasitic reaction involves the capture of the
neutron by a nucleus which then emits a gamma ray rather than fissioning.
Another possible reaction involves the neutron simply bouncing or scattering off of
a nucleus. After several such scattering reactions, the neutron might eventually leak
out of the uranium core of the reactor. Such processes remove neutrons from the
reactor and tend to inhibit the chain reaction.
Therefore one o( the primary tasks of the nuclear engineer is to follow the
neutron "economy" in a nuclear reactor in order to monitor and control the
behavior of the fission chain reaction. That is, he must learn how to design the
reactor so that there is a balance between the production of neutrons in fission
reactions and the loss of neutrons due to capture or leakage. The study of such
processes is known as either nuclear reactor theory, nuclear reactor physics or
sometimes simply as neutronics. It is essentially the subject of this text.
However the achievement of a stable chain of fission reactions is only a part of
the responsibility of the nuclear engineer. In addition he must learn how to extract
and use the energy liberated in these fission reactions. This task involves the
subjects of heat transfer, fluid flow, structural and materials analysis, and power
systems analysis and interacts strongly with the nuclear analysis of a reactor core.
It is discussed in the latter chapters of the text.
We first turn our attention to a development of the fundamental concepts
involved in predicting the distribution of neutrons in a nuclear reactor in order to
understand and design a fission chain reaction system. We need to consider
essentially two different subjects: (a) the determination of the probabilities of
occurrence of various neutron-nuclear reactions and (b) the derivation and solu-
tion of an equation that uses these probabilities to determine the neutron density
and fission reaction rate in a nuclear reactor core.
The above discussion clearly indicates the importance of being able to determine
the rate at which various types of neutron-nuclear reactions occur within the
reactor. However it is important to keep in mind that there are enormous numbers
of neutrons (typically 108 per cm3) and even larger numbers of nuclei (1<>2 2 per cm3)
in the reactor core. Hence we really need concern ourselves only with the average
behavior of the neutrons and nuclei in the reactor in a statistical sense. That is, we
wish to calculate the probabilities that various types of neutron-nuclear interactions
will occur. These reaction probabilities are expressed in terms of parameters called
nuclear cross sections.
These cross sections represent the fundamental data utilized by the nuclear
engineer in his analysis of a nuclear reactor, much in the same way that thermal or
structural data are used by the mechanical engineer or circuit device parameters
are used by the electrical engineer. Hence some familiarity with. the physics
underlying the determination and behavior of such cross sections is necessary for
effective nuclear reactor analysis.
In this chapter we will review those aspects of nuclear physics that are particu-
larly relevant to the study of fission chain reactions. It should be stressed that this
presentation is not intended to be complete. Indeed we would anticipate that most

engineering students will have had some exposure to modem atomic and nuclear
physics in earlier courses. (Students who find most of the material in this chapter to
be totally alien territory would be well advised to consult one of the several
excellent standard references on introductory nuclear physics 1--4 containing sub-
stantially more thorough discussions of these topics.) Unfortunately, however, most
conventional treatments of these subjects do not place sufficient emphasis on the
study of nuclear reactions in general, or neuti:on-nuclear reactions in particular, for
our purposes (although there are several notable exceptions5-6).
We will begin with a brief introduction to spontaneous nuclear radioactive decay
as an example of a nuclear reaction. We then consider nuclear collision reactions
and introduce the concept of a nuclear cross section. Here we will devote particular
attention to a qualitative discussion of cross sections characterizing neutron-
nucleus reactions. Our final topic will be that of the nuclear fission reaction itself
and the radiation emanating from such reactions.
I. NUCLEAR REACTIONS
There are essentially two types of nuclear reactions of importance in the

study of nuclear reactors: (a) spontaneous disintegrations of nuclei and (b) reac-
tions resulting from the collision between nuclei and/ or nuclear particles. An
example of the first type of reaction would be the radioactive decay of fission
products, since these are frequently unstable. Such disintegration reactions depend
only on the properties of an individual nucleus. The neutron-nucleus collision
events involved in the fission chain reaction are an example of the second type of
reaction. These collision reactions depend not only on the properties of the
colliding particles, for example, the neutron and the nucleus, but also the relative
velocity with which they strike one another.
Before diving off into a discussion of nuclear reactions, let us first introduce
some notation. We will denote the number of protons in an atomic nucleus by Z
(the atomic number), the number of neutrons by N, and the total number of
nucleons (protons plus neutrons) by A (the mass number). A specific nucleus will be
denoted by a symbol such as 1X, where X is the chemical symbol for the atom of
interest. For example, ~H, 1~C, and ~~U are notations for three such nuclei. We will
refer to various species of nuclei as nuc/ides. Nuclei characterized by the same
atomic number Z but different mass numbers A are referred to as isotopes (e.g.,
233u , 234u , 23su , and 23su) . s·mce th e nuc1eus 1s · al system, 1t
• a quantum mech amc ·
92 92 92 92
may be found in any of a number of possible energy states. The general notation
1X refers to a nuclear ground state, while an asterisk is used to denote a nucleus in
an excited state, 1X*. Long lived excited states of nuclei are referred to as nuclear
isomers or isomeric states and are denoted by a superscript m (e.g., 11;m1n).
A. Radioactive Decay
Certain nuclei are unstable in the sense that they may spontaneously undergo
a transformation into a different nuclide, usually accompanied by the emission of
energetic particles. Such a spontaneous nuclear transformation is referred to as
radioactive decay. The three most common types of radioactive decay found in
naturally occurring nuclides include alpha decay, in which the nucleus emits a
helium nucleus ~He; beta decay, which corresponds to the conversion of a neutron
in the nucleus into a proton, generally accompanied by the emission of an electron
and a neutrino; and gamma decay, the transition of a nucleus from one excited
state to a lower excited state with the accompanying emission of a photon.
However other types of radioactive decay are possible in a nuclear reactor since
many unstable nuclides are produced in fission which do not occur in nature. For
example, certain nuclei such as ~iKr may decay by emitting a neutron. (We will
later find that this particular type of decay process is extremely important for
reactor operation.)
The fundamental law describing radioactive decay is based on the experimental
observation that the probability that a nucleus will decay in a given time interval is
essentially constant, independent of the age of the nucleus or its environment,
dependent only on the type of the nucleus itself. Hence the time rate of change of
the number of original nuclei of a given type must be proportional to the number
of nuclei present at that time. Let us call the proportionality constant A. Then if
N(t) is the number of original nuclei left at time t, we find
dN
--=AN(t).
dt · (2-1)
Here A is referred to as the radioactive decay constant characteristic of the nucleus
and has units of inverse time. If we initially have N0 nuclei present, then at any
later time t the number of nuclei present will be given by an exponential law:
(2-2)
The rate at which nuclei are decaying is given by
(2-3)
From this time behavior, it is apparent that the probability that a given nucleus will
decay in a time interval t to t + dt is just
p(t)dt =Ae-AJ dt. (2-4)

Since radioactive decay is a statistical phenomenon, we cannot predict with any
certainty precisely when a given nucleus will decay. However we can calculate the
mean lifetime i of the nucleus before decay using our expression for p(t) from Eq.
(2-4)
- roo roo -Al I
t= Jo dttp(t)=A Jo dtte =I. (2-5)
Hence on the average a given nucleus will decay after a time 1/A.
A closely related quantity is the length of time necessary for half of the original
number of nuclei present to decay away. Such a time T112 is referred to as
radioactive half-life for the nucleus and can be calculated from its definition by
noting
(2-6)
or
(2-7)
It is common practice to tabulate such radioactive half-lives of various unstable

nuclei in preference to their mean lifetime i or decay constant A.. 8
Yet another definition of some importance is that of the activity characterizing a
sample of radioactive material. · This quantity is simply the total number of
disintegrations occurring per second AN (t). Activities are usually measured in units
of curies, where one curie (Ci) is defined to be that quantity of radioactive nuclei
for which the number of disintegrations per second is 3.70 x 1010• (This is roughly
the activity of 1 g of radium.)
Actually it is more common and far more useful to regard the dependent
variable N ( t) as the atomic number density ( # / cm3) of the nuclide of interest
rather than the total number of nuclei present in the sample. We will adhere to this
practice in our subsequent discussion.
Most radioactive decay processes are somewhat more complicated than those
described by Eq. (2-1). For example, the decaying nuclide may itself be produced
by some type of source, say, R (t) nuclei/ cm3 • sec. Then the nuclide balance
equation becomes
dN
th= -AN (t) + R (t). (2-8)
We can also write similar equations describing several nuclides, each of which
decays into another. Consider, by way of example, the radioactive decay chain:
Then the appropriate equations describing the number of nuclides of each type
present are
dNx
dt= -AxNx+Rx,
dNy
dt = -A.yNy+AxNx+ Ry, (2-9)
dN2
dt = -A.2 N 2 +AyNy+R2 ,
where Rx(t) is the production term for the X-nuclide, and so on. Since this is just a
system of linear first-order differential equations with constant coefficients, it can
easily be solved using standard techniques, and hence we will defer further
discussion to the problems at the end of the chapter.
Very similar considerations also hold for the transition of nuclei between
different excited states. Such states represent the quantum levels available to the
nucleus. We can again characterize the probability that the nucleus will "decay"
out of one excited state into a lower state by a decay constant A, and once again
also develop the concept of a mean lifetime for the excited state i. A useful related
THE NUCLEAR PHYSICS OF FISSION CHAIN REACTIONS / IS
concept here is the uncertainty or width r of the energy level characterizing the
excited state. This width is related to the mean lifetime of the state by the
Heisenberg uncertainty principle:
(2-10)
Hence the width of the state can be expressed in terms of its decay constant by
(2-11)
B. Nuclear Collision Reactions

The study of nuclear collision reactions can be formulated in a manner very
similar to that used to describe chemical reactions. Indeed, the familar notation for
a chemical reaction
a+b~c+d (2-12)
is frequently adopted to describe nuclear reactions. However since in nuclear

reactions, one particle is usually considered to be a projectile while the other
particle is taken as a target, one sometimes uses the more detailed notation
a (b, c) d C
Target J ~ Projectile
o-+-e
b a
As an example, the reaction
ln+235u~236U+-v
0 92 92 "
would be written as
23su(
92 n, Y)236U
92 ·
The general class of such reactions would be simply denoted as (n, y) reactions.
Nuclear reactions are generally accompanied by either the absorption or emis-
sion of energy. One can calculate the energy released by (or required for) a given
nuclear reaction by using the important result from the theory of relativity:
E=mc 2 (2-13)
'
where c is the speed of light and m is the mass converted into energy in a reaction.
The appropriate quantity to use for the variable m that appears in this formula is
the mass difference between the interacting particles before and after the collision.
16 / INTRODUCTORY CONCEPTS OF NUCLEAR. POWER. REACTOR ANALYSIS
For the reaction a(b,c)d we would calculate the reaction energy as
· (2-14)
If Q > 0, then we say the reaction is exothermic, which corresponds to a release of

energy in the reaction. If Q < 0, then the reaction is said to be endothermic, and
energy must be supplied to the colliding nuclei in order to stimulate the reaction to
occur. Obviously, nuclear fission is an example of an exothermic reaction.
There are a wide variety of possible nuclear reactions. The reactions of most
interest in the analysis of a nuclear fission reactor involve interactions between
neutrons and nuclei and include
Nuclear fission (n,fission):
(2-15)
Radiative capture (n, y):
(2-16)
Scattering (n,n) or (n,n'):
bn+1x➔in+1X [elastic scattering ( n, n)]

➔An+(1X)* (inelastic scattering ( n, n')] (2-17)
➔bn+1X+y [inelastic scattering (n,n')].
We have already discussed the nuclear fission reaction. In radiative capture the
incident neutron is absorbed by the target nucleus to form a new nuclide of mass
number A + 1. As we will see later, this "compound" nucleus is formed in an
excited state. In a radiative capture reaction, it will eventually decay to its ground
state by emitting a high-energy photon, that is, a gamma ray. An alternative type of
capture reaction of some importance in reactor control is the (n,a) reaction which
occurs in 1~B, for example.
The third reaction of importance is scattering. In this reaction the neutron simply
scatters off of the nucleus (n,n), although in some cases, it may first combine with
the nucleus to form a compound nucleus for a short time before being reemitted
and will frequently leave the nucleus in an excited state from which it later decays
by gamma emission.
The importance of the fission reaction to nuclear reactor operation is obvious.
Both radiative capture and scattering are also extremely important since they
influence the neutron economy and hence the chain reaction. We will concentrate
specifically on neutron-nuclear reactions as we turn to a more quantitative treat-
ment of nuclear reactions of importance in fission chain reactions.
1. MICROSCOPIC CROSS SECTIONS
The probability that a neutron-nuclear reaction will occur is characterized by

a quantity called a nuclear cross section. Let us first define this quantity operation-
ally by considering a beam of neutrons, all traveling with the same speed and
direction, which is incident normally upon and uniformly across the face of a target
of material. If the target is sufficiently thin (say, one atomic layer thick), then no
nuclei in the target will be shielded by other nuclei from the incident neutron beam
(see Figure 2-1). In this case we would expect that the rate of neutron-nuclear
reactions in the target will be proportional to both the incident neutron beam
intensity J (in units of number of neutrons/cm2·sec) and the number of target
atoms per unit area NA (# /cm2). If we call the constant of proportionality a, we
can write the rate at which reactions occur per unit area on the target as
Rate= R = a I NA
(2-18)
[ # ] [cm2 ]
cm2·sec [ cm: sec ] [ c:2 ] .
We have indicated the units of each of these quantities since they imply that the
proportionality factor a must have the units of an area.
If the incident neutrons and target nuclei could be visualized as classical
particles, a would quite naturally correspond to the cross sectional area presented
by each of the target nuclei to the beam. Hence a is known as the microscopic cross
section characterizing the probability of a neutron-nuclear reaction for the nucleus.
We might continue to think of a as the effective cross sectional area presented by
the nucleus to the beam of incident neutrons. Since the nuclear radius is roughly
10- 12 cm, the geometrical cross sectional area of the nucleus is roughly 10- 24. cm2.
Hence we might expect that nuclear cross sections are of the order of 10- 24 cm2. In
fact microscopic cross sections are usually measured in units of this size called
barns (b). However this geometrical interpretation of a nuclear cross section can
frequently be misleading since a can be much larger (or smaller) than the geometri-
cal cross section of the nucleus due to resonance effects which, in turn, are a
consequence of the quantum mechanical nature of the neutron and the nucleus.
For example, the absorption cross section of •~iXe for slow neutrons is almost one
million times larger than its geometrical cross section.
I neutrons/cm 2 • sec
FIGURE 2-1. A monoenergetlc neutron beam lnddent normally upon a thin target.
We can give a slightly more formal definition of the microscopic cross section by
rearranging Eq. (2-18) to write
Number of reactions/nucleus/sec (R/ NA)

u= = (2-19)
Number of incident neutrons/cm2 /sec I
In this sense, then, if the target has a total cross sectional area &. , all of which is
uniformly exposed to the incident beam, then
u Probability per nucleus that a neutron

(2-20)
€£ = in the beam will interact with it
Thus far we have been discussing the concept of a nuclear cross section in a
rather abstract sense without actually specifying the type of reaction we have in
mind. Actually such cross sections can be used to characterize any type of nuclear
reaction. We can define a microscopic cross section for each type of neutron-
nuclear reaction and each type of nuclide. For example, the appropriate cross
sections characterizing the three types of reactions we discussed earlier, fission,
radiative capture, and scattering, are denoted by ur, uy, and u5, -respectively. We can
also assign separate cross sections to characterize elastic scattering ue in which the
target nucleus remains in its ground state, and inelastic scattering uin in which the
target nucleus is left in an excited state. Since cross sections are related to
probabilities of various types of reactions, it is apparent that
In a similar sense we can define the absorption cross section characterizing those
events in which a nucleus absorbs a neutron. There are a number of possible types
of absorption reactions including fission, radiative capture, (n,a) reactions, and so
on. (Actually one could argue that fission is not really an absorption reaction since
several neutrons are created in the fission reaction. It has become customary,
however, to treat fission as an absorption event and then add back in the fission
neutrons released in the reaction at another point, as we will see later.) Finally, we
can introduce the concept of the total cross section ut characterizing the probability
that any type of neutron-nuclear reaction will occur. Obviously
A schematic diagram9 of the heirarchy of cross sections along with their conven-
tional notation is shown in Figure 2-2. Notice that in general one would define the
absorption cross section to characterize any event other than scattering
In a similar fashion, one occasionally defines a nonelastic cross section as any event
other than elastic scattering
Thus far we have defined the concept of a microscopic cross section by

considering· a beam of neutrons of identical speeds incident normally upon the
a, (total)
I
a, (scattering)
a. (absorption)
ae 0 in
(elastic (inelastic
scattering) scattering)
cr1 (fission) o.., (radiative

capture)
FIGURE 2-2. Neutron cnm section heirarchy.
surface of a target. However it is certainly conceivable that such cross sections will
vary, depending on the incident neutron speed (or energy) and direction. Indeed if
the microscopic cross section for various incident neutron energies is measured, a
very strong energy dependence of the cross section is found. The dependence of
neutron cross sections on the incident beam angle is usually much weaker and can
almost always be ignored in nuclear reactor applications. We will return later to
consider in further detail the dependence of cross sections on incident neutron
energy. First, however, it is useful to develop a quantity closely related to the
microscopic cross section, that of the macroscopic cross section.
2. MACROSCOPIC CROSS SECTIONS

Thus far we have considered a beam of neutrons incident upon a very thin
target. This was done to insure that each nucleus in the target would be exposed to
the same beam intensity. If the target were thicker, the nuclei deeper within the
target would tend to be shielded from the incident beam by the nuclei nearer the
surface since interactions remove neutrons from the beam. To account for such
finite thickness effects, let us now consider a neutron beam incident upon the
surface of a target of arbitrary thickness as indicated schematically in Figure 2-3.
We will derive an equation for the "virgin" beam intensity J(x) at any point x in
the target. By virgin beam we are referr1ng to that portion of the neutrons in the
beam that have not interacted with target nuclei. Consider a differential thickness
of target between x and x + dx. Then since dx is infinitesimally thin, we know that
the results from our study of thin targets can be used to calculate the rate at which
neutrons suffer interactions in dx per cm2 • If we recognize that the number of
target nuclei per cm2 in dx is given by dN A= N dx, where N is the number density
of nuclei in the target, then the total reaction rate per unit area in dx is just
(2-21)
20 / INTRODUCTORY CONCEFfS OF NUCLEAR POWER REACTOR ANALYSIS
I I X
x"'O x x+dx
FIGURE 2-3. A monoenergetic neutron beam Incident nonnally on a thick target.
Notice that, consistent with our prescription that any type of interaction will
deflower an incident neutron, we have utilized the total microscopic cross section at
in computing dR.
We can now equate this reaction rate to the decrease in beam intensity between
x and x+dx
-dl(x)= -[l(x+dx)-l(x)]=a/Ndx. (2-22)
Dividing by dx we find a differential equation for the beam intensity I (x)
dl
dx = -Na/(x). (2-23)
· If we solve this equation subject to an incident beam intensity of 10 at x =0, we

find an exponential attenuation of the incident beam of the form
(2-24)
The product of the atomic number density N and the microscopic cross section
at- that appears in the exponential term arises so frequently in nuclear reactor
studies that it has become customary to denote it by a special symbol:
(2-25)
One refers to :It as the total macroscopic cross section characterizing the target
material. The term "macroscopic" arises from the recognition that :It characterizes
the probability of neutron interaction in a macroscopic chunk of material (the
target), whereas the microscopic cross section characterizes the probability of
•interaction with only a single nucleus.
It should be noted that :It is not really a "cross section" at all, however, since its
units are inverse length. A more appropriate interpretation can be achieved by
reexamining Eq. (2-22) and noting that the fractional change in beam intensity
occurring over a distance dx is just given by
(2-26)
Hence, it is natural to interpret ~t as the probability per unit path length traveled
that the neutron will undergo a reaction with a nucleus in the sample. In this sense
then
exp( - ~tx) = probability that a neutron moves a distance dx without

any interaction;
~, exp ( - ~,x) dx = probability that a neutron has its first interaction in dx
= p(x)dx.
With this interaction probability, we can calculate the average distance a neutron
travels before interacting with a nucleus in the sample
(2-27)
It is customary to refer to this distance as the neutron mean free path since it
essentially measures the average distance a neutron is likely to stream freely before
colliding with a nucleus.
The reader has probably noticed the similarity of this analysis to our earlier
treatment of radioactive decay. The spatial attentuation of a neutron beam passing
through a sample of material and the temporal decay of a sample of radioactive
nuclei are similar types of statistical phenomenon in which the probability of an
event occurring that removes a neutron or nucleus from the original sample
depends only on the number of neutrons or nuclei present at the position or time of
interest. It should be stressed that both the mean free path and the mean lifetime
for decay are very much average quantities. There will be statistical fluctuations
about these mean values.
If we recall that ~t is the probability per unit path length that a neutron will
undergo a reaction, while the neutron speed v is the distance traveled by the
neutron in a unit time, then evidently
v~1 = [ cm ][cm- 1]=[sec- 1]= Frequency with which
sec reactions occur. (2-28)
This quantity is usually referred to as the collision frequency for the neutron in the
v~tr
sample. Its reciprocal, [ 1, is therefore interpretable as the mean time between
neutron reactions.
Thus far our discussion has been restricted to total macroscopic cross sections
that characterize the probability that a neutron will undergo any type of reaction.,
We can generalize this concept by formally defining the macroscopic cross section
for any specific reaction as just the microscopic cross section for the reaction of
interest multiplied by the number density N characterizing the material of interest.
For example, the macroscopic fission cross section would be defined as
(2-29)
In a similar fashion we can define
ll / INTRODUCTORY CONCEPTS OF NUCLEAR POWER REACTOR ANALYSIS
Notice also that
(2-31)
It should be stressed that while one can formally define such macroscopic cross
sections for specific reactions, our earlier discussion of neutron penetration into a
thick target applies only to the total macroscopic cross section l':t. We could not
extend this discussion, for example, to the calculation of the probability of neutron
penetration to a depth x prior to absorption by merely replacing l':t in Eq. (2-24) by
l':a, since it may be possible for the neutron to undergo a number of scattering
reactions before finally suffering an absorption reaction. We can calculate these
specific reaction probabilities only after a more complete consideration of neutron
transport in materials (Chapters 4 and 5).
The concept of a macroscopic cross section can also be generalized to homo-
geneous mixtures of different nuclides. For example, if we have a homogeneous
mixture of three different species of nuclide, X, Y, and Z, with respective atomic
number densities N x, Ny, and N z, then the total macroscopic cross section
characterizing the mixture is given by
(2-32)
where a{ is the microscopic total cross section for nuclide X, and so on. It should
be noted that such a prescription for determining the macroscopic cross section for
a mixture arises quite naturally from our interpretation of such cross sections as
probabilities of reactions.
As we mentioned earlier, all neutron-nuclear reaction cross sections (fission,
radiative capture, scattering, etc.) depend to some degree on the energy of the
incident neutron. If we denote the neutron energy by E, we acknowledge this
dependence by including a functional dependence on E in the microscopic cross
section o(E) and hence by inference also in the macroscopic cross section l':(E).
However the macroscopic cross section can depend on additional variables as
well. For example, suppose that the target material does not have a uniform
composition. Then the number density N will depend on the position r in the
sample, and hence the macroscopic cross sections themselves will be space-
dependent. In a similar manner, the number densities might depend on time-
suppose, for example, that' the nuclide of interest was unstable such that its number
density was decaying as a function of time. Therefore in the most general case we
would write
l':(r,E, t) = N (r,t)o(E) (2-33)
to indicate the explicit dependence of the macroscopic cross section on neutron

energy E, position r, and time t.
In summary then, nuclear cross sections can be used to characterize the probabil-
ity of various types of neutron-nuclear reactions occurring. They obviously will be
a very basic ingredient in any study of fission chain reactions. The determination
of such cross sections is the task of the nuclear physicist and involves both
experimental measurement and theoretical calculations. The enormous amount of
cross section information required for nuclear reactor analysis is gathered by
numerous nuclear research centers throughout the world. These cross section data
are compiled, evaluated, and then organized into data sets to be used by nuclear
engineers. We will return in a later section to discuss in further detail such nuclear
cross section data sets. For convenience, however, we have included in Appendix A
a table of some of the more important cross sections characteristic of "thermal"
neutrons-that is, of neutrons whose energies are comparable to the thermal energy
of atoms in a reactor core at room temperature, E==0.025 eV-which serves to
illustrate typical orders of magnitudes of these quantities.
EXAMPLE: As a specific illustration, let us calculate the mean free path for a
thermal neutron in graphite. According to the table in Appendix A, carbon is
nearly a pure scatterer with microscopic cross sections o5 =4.8 band 0 8 ==4.0x 10- 3
b. If we assume a mass density of 1.60 g/ cm3, we can calculate the atomic
number ·density in graphite as N = .0803 X 1024 cm- 3 • Consequently the macro-
scopic scattering and absorption cross sections are
and the total cross section is
~t=0.385cm- 1-- ~s·

The mean free path is therefore
1
>..= ~=2.6cm.
t
Notice how small the absorption cross section in graphite is compared to its
scattering cross section. Indeed since ~s/~a = 1.2 X 103, one can infer that a thermal
neutron in graphite will make some 1200 scattering collisions on the average before
being absorbed. This very low absorption cross section makes graphite an ideal
material for nuclear reactor applications.
C. Characteristics of Neutron-Nuclear Cross Sections

Before considering in detail the various types of neutron-nuclear reactions
significant in nuclear reactor analysis, it is useful to give a brief discussion of some
of the relevant physics underlying the behavior of these cross sections. There are
two aspects involved in the analysis of neutron cross sections:
(a) the kinematics of two-particle collisions and
(b) the dynamics of nuclear reactions.
The kinematics of two-body collisions, that is, the application of the laws of
conservation of momentum and energy to such collisions, should be very familiar
to the reader from introductory courses in mechanics or modern physics. However
consistent with our attempt to make this presentation as self-contained as possible,
and being well aware of the short half-life of this type of information in the fast
core memory of most students, we will review such kinematic calculations in
Section 2-1-D.
The dynamics of nuclear reactions is concerned with the fundamental physical
mechanisms involved in such collision events. The two mechanisms of most interest
in nuclear reactor applications are those of potential scattering, in which the
neutron merely bounces off of the force field of the nucleus without actually
penetrating the J).Uclear surface, and compound nucleus formation, in which the
incident neutron is actually absorbed by the nucleus to form a new nucleus of mass
number A + 1 which then decays by emitting a gamma, a neutron, or perhaps
fissioning.
We will give a brief discussion of both of these interaction mechanisms, and then
turn our attention to a specific survey of the various types of cross section behavior
encountered in the more common materials utilized in nuclear reactors.
1. MECHANISMS OF NEUTRON-NUCLEAR INTERACTION

The simplest type of nuclear reaction occurring in a nuclear reactor is
potential scattering, in which the neutron scatters elastically off the nuclear poten•
tial without ever penetrating the nucleus itself. This type of collision event is very
similar to that which would occur between two hard spheres (e.g., billiard balls),
and the cross section for such a reaction is essentially just the geometrical cross
section of the nucleus. Potential scattering cross sections are characterized by a
rather flat energy dependence from about I eV up to the MeV range.
Another common type of reaction encountered in a nuclear reactor is that in
which the incident neutron is first absorbed by the nucleus 1,X to create a new
compound nucleus Ai1 X. This compound nucleus subsequently decays by emitting
an energetic particle. Compound nucleus formation occurs in many neutron-
nuclear reactions of interest to the reactor engineer, including fission, radiative
capture, and certain types of scattering.
That such a mechanism must be involved in these reactions can be inferred from
the relatively long times (at least on a nuclear time scale) for such events to occur.
Whereas it would take a slow neutron (traveling at 1()5 cm/sec) some 10- 17 sec to
cross the nucleus, neutron-nuclear reactions such as fission occur on a time scale
of some 10- 14 sec-or some 1000 transit times. Hence the incident neutron must
first be absorbed by the original nucleus and rattle around a bit, distributing both
its kinetic energy and the additional binding energy supplied by the added neutron
to the other nucleons in the nucleus, before the compound nucleus finally decays.
The long lifetime of the compound nucleus implies that the disintegration process
is essentially independent of the original mode of formation; that is, the compound
nucleus lasts long enough to "forget" most of the characteristics of the incident
neutron (such as which direction it came from).
The formation of a compound nucleus actually corresponds to a so•called
resonance reaction, in which the incident neutron energy matches one of the energy
levels in the compound nucleus. To be more specific, consider a neutron incident
upon a nuclide 1,X:
0-----
1n
0
As we will see later when we develop the topic of collision kinematics, the energy
availabl~ for such a reaction is the center of mass (CM) energy Ec=(M/m+M)E,
where m is the neutron mass, Mis the nuclear mass, and Eis the neutron kinetic
energy in the laboratory system. The actual energy of the excited level of the
THE NUCLEAR PHYSICS OF FISSION CHAIN REACTIONS / 2S
compound nucleus is much higher due to the additional binding energy of the
added neutron Eb. If Ec + Eb is very close to a nuclear energy level of the
compound nucleus Ai•x, one expects that the probability for compound nucleus
formation will be much larger than if Ec + Eb does not "match" this energy level.
Hence we expect that the cross sections for such compound nuclear reactions will
exhibit sharp peaks or resonances at those neutron energies E for which this energy
matching occurs. By way of illustration, we have shown the resonance structure in
the low-energy cross section behavior of a number of nuclei in Figure 2-4.
1,000,000
100,000
10,000
1000
100
0.001 0.01 0.1 10.0

Energy (eVI
FIGURE 2-4. Low-energy cross section behavior of several important nuclides. 12

The second stage of a compound nucleus process is the decay of the compound
nucleus. This decay may occur in a variety of ways, as indicated schematically
below:
/" An +1,X Resonance elastic scattering

.,,;;:,, An+ (1,X)* Inelastic scattering
1n+AX~('4+ 1X)* (2-34)
0 z z ~A~ 1X + y Radiative capture
~ Aix
Z
+AZ x +2-3 01n
1
2
2
Fission
Since we have argued that the final disintegration mode is essentially independent
of the neutron absorption process which creates the compound nucleus, we might
expect that the cross section energy dependence for compound nucleus reactions
will exhibit certain similarities. These similarities will become apparent as we
consider in more detail the specific neutron-nuclear reactions of importance in
nuclear fission reactors.
2. A QUALITATIVE DISCUSSION OF NEUTRON CROSS SECTIONS

(a.) RADIATIVE CAPTURE
Radiative capture reactions are quite significant for reactor analysis since they
remove neutrons from the chain reaction. Such reactions proceed via compound
nucleus formation in which the incident neutron is first absorbed to form the
compound nucleus of mass number A+ 1, and then this nucleus subsequently
decays by emitting a cascade of high-energy gammas. For this reason, the func-
tional dependence of the capture cross section on the neutron kinetic energy E
exhibits a resonance behavior at those energies at which the CM energy Ec plus the
neutron binding energy Eb match an energy level of the compound nucleus. We
have shown this resonance situation schematically for neutron capture in 238 U in
Figure 2-5. In particular, we have indicated the energy level diagram for the
compound nucleus 239U for one of the low lying resonances at E=6.67 eV. This
resonance has a very narrow width of 0.027 eV and an extremely high peak (several
thousand barns). The subsequent radiative decay of the compound 239 U nucleus is
usually a cascade process, with the new nucleus jumping down through a number
of energy levels accompanied by the emission of several gamma rays. Since the
excited levels are typically in the MeV range, the total energy of the emitted
gammas will be quite large.
For resonances (i.e., energy levels) which are spaced widely apart, it is possible to
describe the energy dependence of the absorption cross section by a very simple
expression known as the Breit-Wigner single-level resonance formula: 1-4
r ( E )1;2 1
o..,(Ec) = oo ry Eo --2' (2-35)
C l+y
Here E 0 is the energy at which the resonance occurs (that is, the energy Ec at which
Ee+ Eb matches the energy level of the compound nucleus), r is the so-called total
line width of the resonance that essentially characterizes the width of the energy
level and the full width at half-maximum (FWHM) of the resonance, while r.., is
E
100
2??2222222222?722222>2
namttld?ll?lllll/4
10
Incident
·neutr.on
---,t.----.,.--
I
_____ zzzz:lZZZ~ZZZZZZZZZZ
. N~utron 6.67 eV
mc2 + M
238
::ne-t1c...iti...en-er_g_y_ _.!__ _____ ___ ___
2:~u 10 100 a..,(E) (l:>I
"f Cascade
-----------
2:~u
FIGURE 2-S. An energy-level diagram of the capture resonance In U 238 at E=6.61 eV.
the radiative line width essentially characterizing the probability that the compound
nucleus will decay via gamma emission. Here a 0 is the value of the total cross
section ai(E) at the resonance energy E 0 and can be written in terms of the reduced
neutron wavelength ~o at E 0 as
(2-36)
while g = (2J + I) /2(2/ + l) is a statistical spin factor given in terms of the nuclear
spin / and total spin J. r n is the neutron line width and varies in energy as
(2-37)
We have sketched the Breit-Wigner resonance shape versus the CM kinetic energy
Ee in Figure 2-6. It should be noted that since resonance absorption is primarily of·
importance in heavy nuclei, one can usually approximate Ec--E. For low energies
E«E0 , the cross section behaves as essentially 1/ £ 112 or 1/v. For large energies
E-:t> E0 , the cross section drops off quite rapidly as E - 512 • It is also important to
note that such absorption cross sections are largest at low energies. (A list of
several of the lower lying resonances of 238 U is given in Table 8-2.) The energy
levels in heavy nuclei become relatively more closely spaced at higher energies.
Indeed for energies above roughly 1 keV in heavy nuclei such as 238 U, the
absorption resonances become so closely spaced that they cannot be resolved by
experimental measurements. The treatment of neutron absorption in such unre-
~olved resonances is a very difficult but important task in nuclear reactor analysis.
0 max
I
I
I
I
I
I
r+-►i
I .I
I I
I
I,
I
I
I
Eo
FIGURE 2-6. A single-level capture resonance.
(b.) NUCLEAR FISSION
Nuclear fission is yet another reaction that proceeds by compound nucleus

formation. Once again a compound nucleus is first formed by neutron absorption.
This compound nucleus then decays by fissioning into two lighter nuclei. Since the
mode of disintegration is relatively independent of the formation mechanism, we
might again expect the cross sections for nuclear fission. to exhibit a resonance
structure very similar to that characterizing radiative capture. This is certainly the
case for fission cross sections characterizing nuclei such as 233 U, 235 U, and 239Pu.
However the fission cross sections for other heavy nuclei such as 232Th and 238U
exhibit a somewhat different structure in that they are essentially zero until the
incident neutron energy exceeds a threshold of roughly an MeV. To understand
this latter behavior, we must examine in more detail the· fis-sion process itself, a task
we shall defer until Section 2-11.
(c.) SCATTERING
i.) Inelastic scattering

In an inelastic scattering reaction, the incident neutron is first absorbed by the
nucleus to form a compound nucleus. This nucleus then subsequently decays by
reemitting a neutron. However the final nucleus is left in an excited state. Such
reactions usually occur only for relatively high neutron energies, say above 10 keV,
since the neutron kinetic energy must exceed a certain threshold energy in order to
excite the first excited state of the compound nucleus. We have depicted this
reaction schematically in Figure 2-7.
Since much of the kinetic energy of the incident neutron may be converted into
the energy of excitation of the target nucleus, it is possible for a neutron to lose a
large amount of energy in an inelastic scattering reaction. It should be noted in
particular that kinetic energy is not conserved in such an inelastic reaction.
ii.) Elastic resonance scattering
A very similar compound nucleus reaction involves first the absorption of the
incident neutron, followed by the reemission of the neutron with the target nucleus
Neutron in
'Y Emission
-----------------------1--------------.►
~+1C2 - - - - - - - -
FIGURE 2-7. A schematic of the energy-level diagram characterizing inelastic scattering.
returning to its ground state (see Figure 2-8). In contrast to inelastic scattering,
kinetic energy is conserved in elastic events. As one finds with most compound
nucleus reactions, there is again a resonance_ behavior in the corresponding scatter-
ing cross section. In this case, however, the cross section energy dependence is
somewhat different from that observed in radiative capture or fission. The cross
section may actually decrease before rising to a resonance maximum.
This structure arises because the process of resonance elastic scattering may
interfere (in a quantum mechanical sense) with potential scattering. The
appropriate modification of the Breit-Wigner formula to account for scattering
interference is
(2-38)
resonance interference potential

scattering scattering scattering
where R is the nuclear radius (given approximately by R,_ l.25 x 10- 13A 1/ 3 cm).
iii.) Potential scattering
The simplest type of nuclear interaction is potential scattering, in which the
incident neutron scatters elastically off of the nuclear potential without penetrating
the nuclear surface. Such "billiard-ball" collisions are characterized by an essen-
F,
Neutron in
------~
Neutron
out
mc 2 + M A.c 2
AX o, (E)
z
------------ A+ 1
zX
FIGURE 2-8. The energy-level diagram for resonance elastic scattering.
tially energy-independent cross section op which is of the order of magnitude of the

geometric cross section of the nucleus for intermediate energies, 47TR 2 •
(d) TOTAL NEUTRON CROSS SECTIONS
Recall that we defined the total cross section for neutron-nuclear reactions as
the sum of the cross sections for each type of reaction
(2-39)
Such cross sections can be measured by performing transmission experiments such

as those we used to operationally define the concept of a cross section. That is, one
shoots a monoenergetic beam of neutrons of energy E at a thin target and then
measures the fraction of the incident beam that penetrates the sample.
It is of interest to see if we can understand the energy dependence of total cross
sections in terms of the various different reactions and reaction mechanisms we
have discussed earlier. To be specific, let us consider the total cross section of a
common reactor material, graphite (1~C), which is shown over the energy range
from 10- 3 eV to lOMeV in Figure 2-9. One can distinguish essentially five different
regions of cross section behavior for this material. For very low energies, the cross
section behaves as E- 112 • At about 10- 3 eV the cross section becomes very jagged
and irregular. These spikes in the cross section smooth out by an energy of 10- 1 ev,
and the cross section from this energy up to 105 eV is essentially constant. Above
105 eV the cross section begins to show a detailed structure once again, until it
eventually drops off above 107 eV.
10
.c
c' 5
0 0
0
·;;
r.>
i
Ill
...0r.>
.;
.....
0 1.0
I-
,0-4 10-3 10-2 10-1

Neutron energy, eV
10
5
0 0
.c
c'
0
·;;
r.>
i
Ill 1.0
e
r.>
.;
0 0.5
I-
o.1...__.,____._........,.........,__._...._ _.__...._..........~ ~ - - - ~ - ~ ~ ~ - - - - - - - ~ - - - -
104 10s 106 107
Neutron energy, eV
FIGURE 2-9. The total scattering cross section of 12C.
In an effort to explain each of these different types of behavior, let us begin with
the smooth cross section variation for intermediate energies (region Q)). In this
energy range, the cross sections of most nuclei are dominated by potential scatter-
ing (except for heavy nuclei in which both resonance scattering and absorption
may be dominant effects because of low-lying energy states). The cross section
magnitude is essentially just the geometric area presented to the neutron by the
nucleus (--5b) and depends very little on neutron energy.
The jagged behavior in the MeV region@ corresponds to resonance reaction
mechanisms since now the incident neutron energy is comparable to that of the
lowest energy levels in the compound nucleus 1lC. Since the nuclear energy levels
lie closer to the ground state for heavier nuclei, we would expect that such
resonance structure would appear at progressively lower energies for the heavier
nuclides. For example, we have already noted that a pronounced resonance occurs
in 238 U at 6.67 eV.
The falloff in the cross section at very high energies (region (})) is easily
understandable in terms of the neutron wavelength
>.. = !!_ = __h_ = _2_.8_6_x_t_o_-_9 cm. (2-40)

P \!2mE YE(eV)
32 / INTRODlTCTORY CONCEPTS OF NUCLEAR POWER REACTOR ANALYSIS
Since this wavelength decreases with increasing neutron energy, it is apparent that
for sufficiently high energies, the probability of neutron interaction with the
nucleus will similarly decrease. Such very high energy behavior has little relevance
to nuclear reactor analysis, since neutrons in fission chain reactions rarely exceed
lOMeV in energy.
The irregular, jagged behavior at low energies in region@ is also a wavelength
effect. For sufficiently small energies, the wavelength becomes comparable to the
interatomic spacing, and the neutron interacts not with a single nucleus but rather
with an aggregate of nuclei. If the material has a regular structure (such as the
crystalline structure of graphite), the neutron will be diffracted, just as X-rays are
diffracted when passing through a crystal. This is accompanied by a sensitive
energy dependence as the neutron wavelength becomes comparable to multiples of
the spacing between various crystal lattice planes. For sufficiently small energies,
the wavelength becomes so large that diffraction becomes impossible, and the cross
section becomes smoothly varying again.
There are two other important effects which influence the neutron cross section
behavior at very low energies (regions (D and @). If the neutron energy is less
than the chemical binding energy of atoms in the sample (- I eV), the neu trori. will
no longer be interacting with a free nucleus, but rather with an aggregate of bound
nuclei. It can interact and excite the internal modes of the sample, such as crystal
lattice vibrations or molecular rotations. For these low energies, the thermal motion
of the nuclei also becomes very important. If we recall that these thermal motions
are essentially characterized by the sample temperature T and that the atoms in
thermal equilibrium at this temperature are characterized by a thermal energy } kT,
where k is the Boltzmann constant, k=8.6173x 10- 5 eV /°K, then for neutron
energies comparable to this thermal energy (at room temperature, E = 0.025 eV), the
motion of the nuclei must be considered. We will return in the next section to
discuss the modifications this requires in the neutron cross section. Suffice it to say
at this point that such considerations imply that for small neutron energies, the
cross section will behave as 1/ E 112 •
Similar features are found in most neutron cross sections of interest in reactor
analysis, although the particular energy regions in which such behavior arises will
vary from nuclide to nuclide. For example, we have summarized the characteristics
of nuclei of different mass numbers in Table 2-1. 10
We will return later from time to time to discuss more specific features of
neutron cross section behavior which are of particular significance for nuclear
reactor applications.
3. NUCLEAR DATA SETS

The key ingredient in all reactor calculations is a knowledge of the various
relevant neutron-nuclear cross sections. The complicated dependence of such cross
sections on neutron energy and angle of. incidence, combined with the large
number of isotopes involved in nuclear reactor analysis implies that neutron cross
section data can be quite massive. Such data have been accumulated over the past
few decades by both experimental measurements and theoretical calculations. The
tabulation of these data has grown from a single volume, BNL-325 11 (the so-called
"barn book") in the 1950s to a six-volume set 12 in the 1960s to the point where
nuclear data are most conveniently kept on magnetic tape or in computer memory.
(Although it should be mentioned that a new third edition 13 of BNL-325 has
recently been issued.)
TABLE 2-1 A Summary of Croa Section Behavior for Nuclei of Various Mass Number10
Slow Epithermal Fast

neutrons neutrons neutrons
E<leV leV<E<.IMeV .IMeV<E <20MeV
Separated resonances
Light
nuclei Potential scattering
A<25
~Resonance scattering, (n,2n), (n,p)
Separated Overlapping Continuum

resonances resonances resonances
Intermediate
nuclei Resonance scattering, radiative capture
25<A<80
Potential scattering Inelastic scattering
Separated Overlapping Continuum

resonances resonances resonances
Heavy
nuclei Radiative capture
A>80
Inelastic scattering, (n,2n)
Basic nuclear data appear in a wide variety of forms. For example, raw cross
section data are usually provided by a variety of experiments, each char-
acterized by a different degree of accuracy. Multiple sets of experimental data may
exist giving different values for the same cross sections. Data may be provided by
different theoretical calculations of varying accuracy. There are also numerous
gaps where no cross section measurement or theory is available or applicable. The
enormous volume of such varied nuclear data would overwhelm the nuclear
engineer in his efforts to extract those cross sections of relevance to his particular
needs.
Hence a number of years ago it was decided to consolidate and standardize all of
the cross section information i:rito one data set. To this end the Evaluated Nuclear
Data File (ENDF) 14 was established to consolidate, organize, and present these
data in a form convenient for nuclear applications. The ENDF system contains
both neutron and photon cross section data along with data-processing computer
programs which can manipulate the data into the most convenient form for the
user. These data are stored in three. computer library systems:
(1) CSISRS (Cross Section Information Storage and Retrieval System): This
data set contains essentially unevaluated raw data from experimental
measurements.
(2) ENDF / A: This data set contains both complete and incomplete sets of
nuclear data as soon as they become available. For each isotope there
34 / INTRODUCTORY CONCEPfS OF NUCLEAR POWER REACTOR ANALYSIS
may be, more than one data set for a particular reaction, or there may be
none at all for certain reactions of interest.
(3) ENDF /B: The ENDF /B data set contain~ only complete, evaluated sets
of nuclear data presented in a form that can be most conveniently utilized
by the nuclear designer. Data for approximately 80 isotopes are included
for all significant neutron-induced reactions in the energy range 10- 5 eV
to 20MeV. In particular, cross section data are provided for the reactions
(n, y), (n,fission), (n,p), (n,a), (n,n), (n,n'), (n,2p) and (n,2n), as well as
for the differential scattering cross sections (described in Section 2-1-D).
_ENDF /B is therefore regarded as the standard source of nuclear data for use in
nuclear reactor analysis in the United States. (There are comparable data sets in
Europe and the Soviet Union.) The ENDF /B data set is continually being
reevaluated and updated as new cross section measurements become available.
Revised versions of the data set are issued at one- or two-year intervals.
The effort involved in preparing the ENDF /B set is enormous. It not only
involves collecting and organizing the massive amount of nuclear data available
from an enormous variety of sources, but evaluating these data, checking them for
consistency, filling in the gaps in the data using existing theories (or educated
guesses), testing the data against experimental measurements, and arranging the
data in a convenient form for use.
The actual data contained in the ENDF /B set are usually not contained in
tabular form, but rather in the form of numerous fitting parameters that can be
assembled by a processing code into a fully evaluated set of cross section data for
any material of interest. For example, the set contains resonance parameters such
as level widths, resonance energies, and cross sections at resonance rather than a
tabulated set of cross section data for various energies (which would require a
prohibitively large storage). The resonance cross section can then be generated
using these parameters in a Breit-Wigner type resonance formula.
We will return to discuss further details of how such data sets are manipulated
into the forms useful for reactor calculations when we have developed a more
complete understanding of the various methods used in nuclear reactor analysis. At
the end of this chapter we have provided a list of the most useful sources of current
neutron cross section data. 15
D. Some Generalizations of the Concept of the Cross Section

I. DIFFERENTIAL SCATTERING CROSS SECTIONS
Neutron cross sections provide a quantitative measure of the probability that
various types of neutron-nuclear reactions will occur. For example, we have
introduced o.(E) to characterize the probability that a neutron with kinetic energy
E incident upon a nucleus will be absorbed. Similar cross sections have been
introduced to describe reactions such as scattering. ·
It is frequently useful to introduce a generalization of the concept of a neutron
cross section characterizing the scattering reaction. In such reactions the incident
neutron will usually experience a change in both direction of motion and energy in
the scattering event. (Just imagine a billiard-ball collision.) The microscopic scatter-
ing cross section will describe the probability that such a scattering collision occurs.
However it provides no information about the change in neutron direction or
energy that occurs in such a collision. This latter information is very important in
certain types of reactor studies. To characterize it, we must introduce the concept
of the differential scattering cross section.
Before we develop these generalizations, let us offer both an explanation and an
apology for the mathematical detail that follows. It has been our intent to keep the
level of mathematics in these introductory chapters as low as possible. Unfor-
tunately, however, to describe cross sections characterizing changes in neutron
energy and direction, one must utilize a bit of vector notation. Frequently students
tend to become somewhat intimidated by the notation customarily used to describe
the treatment of the direction of neutron motion. We wish to reassure the reader
that regardless of appearance, the actual level of mathematical analysis we will use
is very rudimentary and essentially includes only vector algebra with an added
dash of volume integration.
First we must introduce variables that characterize the motion of the incident
neutron. The natural choice would be the neutron velocity v. Then the cross section
we wish to define would describe the probability that a neutron incident with a
velocity v would be scattered by a nucleus to a new velocity v'.
However in reactor analysis it will be more convenient to describe the neutron
motion with slightly different variables. We will essentially decompose the neutron
velocity vector into two components, one variable characterizing the neutron speed
and a second variable for the neutron direction of motion. We use the kinetic
energy of the neutron E = ½mv 2 instead of the neutron speed itself. Then to specify
the direction of neutron motion, we introduce a unit vector Sl in the direction of the
neutron velocity vector v:
(2-41)
where we have chosen to represent this direction unit vector in spherical velocity-
v,
ev Vy
I
I
¢ ....... I
'-..J
FIGURE 2-10. The neutron direction unit vector O in spherical coordinates.

space coordinates (O,<t>) (see Figure 2-10). Notice that to describe the incident
neutron velocity we now specify both its energy E and its direction 0.
At this point it is convenient to consider· how one integrates over these variables
(as we shall have cause to do later in this section). Suppose we wished to integrate
over all possible neutron velocities. This integration could then be performed in
either Cartesian or spherical velocity coordinates:
However we have defined the unit vector in the direction of the velocity vector v as
0. Hence we can identify the angular portion of the integration in Eq. (2-42) as just
the integration over this direction:
r .. r 2.,,. r.,,.
1,/0
4.,,.
=Jn0 d</> Jn0 sinfJdO. (2-43)
In this sense we see that the differential dO of the unit vector ncorresponds to a
differential solid angle
dO=sinOdOdct,. (2-44)
One final modification is useful here. We will usually choose to work with the
neutron energy E rather than the neutron speed. Hence rather than choosing to
integi:ate functions of v over all neutron velocities, we will integrate functions of E
and Oover all possible neutron energies and directions:
(2-45)
So much for mathematical preliminaries. We will now proceed to introduce the

concept of a cross section that characterizes the probability that a neutron is
scattered from an initial energy E and direction of motion O to a final energy E'
and direction of motion O'. To make life simple, we will first do this for the
situation in which we are only interested in the change in neutron energy in
scattering. Imagine a beam of neutrons of incident intensity I, all of energy E,
incident upon a thin target of surface atomic density NA· Then the rate/cm2 at
which neutrons will be scattered from their original energy E to a final energy E' in
the range E' to E' + dE' is proportional to the beam intensity J, the target surface
density NA• and the differential range dE' of final energies. _ We will define the
microscopic differential scattering cross section as(E➔ E') as the appropriate propor-
tionality parameter
(2-46)
Hence we find that as(E➔ E') characterizes the probability that a scattering
collision changes the original neutron energy from E to E' in dE'. It is important to
notice that this differential scattering cross section is a "distribution" in the sense
that it is associated with a certain range of final energies, E' to E' + dE'. Hence its
dimensions are cm2 /eV.
There is a very simple relationship between the differential scattering cross
section os(E➔ E') and our earlier definition of the µricroscopic scattering cross
section os(E). If we recognize that the latter quantity is just related to the
probability that a neutron of energy E will suffer a scattering collision, regardless
of the final energy E' to which it is scattered, then it is apparent that os(E) is just
the integral of the differential scattering cross section os(E➔ E') over all final
energies E'
{2-47)
Of course this relationship explains the origin of the term "differential."

It should be mentioned that occasionally one encounters a somewhat differ-
ent notation for the differential scattering cross section which may be written as
do/ dE. We find the convention of denoting differential cross sections by a multiple
variable argument [such as (E➔ E')] to be more convenient for our purposes, and
hence will use this notation throughout.
We can similarly introduce the concept of a differential scattering cross section
describing the probability that a neutron scatters from an incident direction !l to a
final direction !l' in a very similar manner:
Once again, os(!l➔!l') is related to our earlier microscopic scattering cross section
by an integration over all final directions
(2-48)
Two comments are useful here. First, it should be mentioned that the dependence
of the scattering cross section os(!l) on the incident neutron direction is usually
ignored. Indeed very few microscopic scattering cross sections in reactor applica-
tions depend on the incident neutron direction because the nuclei in any macro-
scopic sample are usually randomly oriented, and thus any directional dependence
averages out when averaged over all possible nuclear orientations. (One could
imagine a sample in which most of the nuclei could be aligned-e.g., a ferromag-
netic material-but such situations can safely be ignored in reactor analysis.)
In this case, however, even though the differential scattering cross section
os(!l➔O') will not depend on the incident neutron direction, it will depend on the
change in neutron direction. This is most conveniently expressed in terms of a
functional dependence on the angle through which the incident neutron is scattered
-the so-called scattering angle fJ, or more conveniently, the cosine of this scattering
38 / INTRODUCTORY CONCEYfS OF NUCLEAR POWER REACTOR ANALYSIS
I
I
I
I I
I I
I I
1(/J/
I f..:.t
I I
0
,/
/I
....... ...... /I
......
/I
-......._ fl
. . . . . . . . . . z......
.......
FIGURE 2.11. Definition of the scattering angle 9.
angle, µ.0 =cos8, which can be conveniently expressed as the dot product between
the unit direction vectors µ 0 = 0 ·O'. (Again, the dependence on azimuthal angle cf>
does not arise for materials in which the m,1clei have a random orientation.) One
occasionally denotes this functional dependence by writing
(2-49)
We will continue to use the somewhat more formal notation by writing os(O➔ D'),
even though we know that this differential cross section usually depends only on
JLo·
Thus far we have developed the concept of differential scattering cross sections
that characterize the probability of scattering from one energy to another or one
direction to another. We can combine these concepts by defining a double differen-
tial scattering cross section that characterizes scattering from an incident energy E,
direction O to a final energy E' in dE' and O' in dO'
Again, alternative notations are occasionally used such as os<E,O➔ E',0') or

2 A
d 0 8 /dEdO.
We can again relate the double differential scattering cross section to the
differential scattering cross section or the scattering cross section by integratton
over energy or angle:
o 8 (E➔ E')= Ldil'os(E➔ E',D➔il'),

4w
(2-50)
or
or
(2-52)
The concept of a differential scattering cross section can also be applied to

macroscopic cross sections by merely multiplying by the atomic number density N:
l:s( E~E', n~Sl')- Nas( E~E', a~n')

l:s( E~E') = N<1s( E ~E') (2-53)
l:s(!i~!i') = Nas(!i~fr)
Such differential cross sections are quite important in nuclear reactor analysis
since they determine the manner in which neutrons move about in a reactor core,
as well as the rate at which they leak out of the reactor. The measurement or
calculation 0f such cross sections can become quite involved, and the amount of
data necessary to adequately represent differential cross sections is usually rather
voluminous. Such data are contained in evaluated cross section files such as
ENDF /B, as well as in cross section compilations such as BNL-400. 16
Although the calculation of such differential scattering cross sections is usually
formidable, there is one instance of considerable importance to nuclear reactor
analysis in which such cross sections can be calculated in a straightforward fashion
merely by using the laws of conservation of energy and momentum. This is the
situation in which neutrons scatter elastically from stationary nuclei. To prepare
the way for the calculation of such cross sections, let us first decompose the
differential scattering cross section into two factors:
(2-54)
If we recall our earlier definition of the scattering cross section, as(E), then it
becomes apparent that we can identify
Probability that a neutron scattering with

p (E~E')dE' = m1tia
... 1 energy E wi·11 emerge wit. h a new energy
E' in the interval E' to E' + dE'.
We can explicitly calculate this quantity for the situation in which neutrons of
moderate energies (E < 1MeV) scatter elastically via potential scattering from
stationary nuclei of low mass number A.
2. KINEMATICS OF NEUTRON SCATTERING FROM

STATIONARY NUCLEI
The kinematics of any two-body collision process is simplified very con-
siderably when analyzed within the center-of-mass (CM) coordinate frame. We
have sketched the collision event before and after the collision in both the LAB
and CM coordinate frames in Figure 2-12. Here, lower-case notation corresponds
Target' Scattering
Neutron COM nucleus angle - LAB
system
0 ... ~ ® ----- -----
Ill VL VCM M
LAB - Before LAB After
Scattering
angle - CM
system
CM - Before CM - After
FIGURE 2-12. Definition of collision coordinates in LAB and center-of-mass (CM) systems.
to the neutron and upper-case notation to the nucleus. The subscripts L and C refer
to LAB or CM frames, respectively.
The velocity of the CM frame is defined by
(2-55)
where we have assumed that the initial nucleus velocity VL is zero and noted that
the nucleus-neutron mass ratio M / m is essentially just the nuclear mass number A.
If we note that the neutron and nucleus velocities in the CM frame are given by
(2-56)
V . 1
c=-vcM=- A+l Vv
then it is apparent that the total momentum in the CM frame is zero, as it must be.
We can relate the total kinetic energies in the LAB and CM frames by
computing
0
LAB: EL=½mvt+½~,
CM: Ee= ½mv~+ ½MVi= ½µv'f_. (2-57)

where we have introduced the reduced mass µ, = mM/ ( m + M). Hence we find the
important relation between the energy in the CM and the LAB frames as
(2-58)
In particular it should be noted that the total energy in the CM system is always
less than that in the LAB system. The energy difference is taken up by the center of
mass motion itself.
Using conservation of momentum and energy, it is easy for one to demonstrate
that the magnitudes of the CM velocities do not change in the collision:
' A
Ve= Ve= A + l vL,
(2-59)
' - Ve-
Ve- - --vL
1 .
A +1 '
only their velocity vectors are rotated through the CM scattering angle Oc. This fact
allows one to relate the scattering angles in the LAB and CM frames. Consider the
vector diagram in Figure 2-13 illustrating the velocities and scattering angles in
these two frames.
If we note from this diagram that
V1, sin ()L = Ve sin()e,
(2-60)
then we can relate the scattering angles in the CM and LAB frames by
VcsinOc sinOe
tan()L = , = - 1- - - - (2-61)
VeM + vecosOe ()
A +cos e
This relationship is particularly useful since cross sections are usually calculated •
in the CM frame, but are measured and used in the LAB frame. If we denote the
differential scattering cross sections characterizing scattet;ing through angles ()L and
Oc in the LAB and CM frames, aL(OJ and aeM(Oc) respectively, we can use
(2-62)
FIGURE 2-13. Relation between the scattering angles in the LAB and CM frames.
to relate the LAB and CM differential scattering cross sections by
(2-63)
Returning to our vector diagram in Figure 2-13 and using the law of cosines, we
can find
v'2+ v2 - v'2
O )=
cos ( 180 0 - c C CM L
(2-64)
2vcvcM
but using Eqs. (2-55) and (2-59), we can rewrite this as
1/2mvt E' A 2 + 1 +2A cos8c

(2-65)
1/2mvt =E = (A+ 1)2
It is useful to introduce a parameter related to the nuclear mass number A
A -1 ) 2
a= ( A+l . (2-66)
Then we can rewrite the final neutron energy after collision, Et- E ', in terms of the
incident neutron energy, Ei=E as
(2-67)
Let us study this very important relation in more detail. First notice that it
implies that the energy transfer from the neutron to the nucleus is directly related
to the scattering angle in the CM frame. For example, if Oc = 0, then the neutron
would lose no energy (Et= E). This corresponds, of course, to no collision at all (a
"miss"). The maximum energy loss occurs in a backscattering collision in which
Be= 180°. In this case, Et= aEi. Hence the maximum energy that a neutron can
lose in an elastic scattering collision with a stationary nucleus is (1- a)Ei. For
example, in scattering collisions with hydrogen nuclei (A = 1), the neutron could
conceivably lose all of its energy, while in a collision with a heavy nucleus such as
238 U it could lose at most 2% of its incident energy.
In summary, then, we have discovered two very important facts. First, a neutron
cannot gain energy in an elastic collision with a stationary nucleus (Et is always
less than EJ Second, the neutron cannot emerge from an elastic scattering collision
with an energy Er less than aEi.
We can now go one step further and actually calculate the scattering probability
distribution, P(E➔ E'), or in our present notation, P(Ei➔ E,), for the case of
elastic scattering from stationary nuclei. First we note from our preceding discus-
sion that P(Ei➔ Er) must vanish if the final energy Er does not fall within the range
aEi <Er< Ei. To calculate P(Ei➔ Et) in this range, we will utilize the relationship
Eq. (2-67).
If we recall that there is a one-to-one relationship between the neutron energy

transfer and the scattering angle, we can infer that there must similarly be a
relationship between the probability of the neutron experiencing a given energy
transfer Ei➔ Er and the probability that it will be scattered through a given
scattering angle 9c. Yet the probability of scattering through an angle 9c into d9c
about 9c is just given by
(2-68)
Hence we can equate
(2-69)
If we now differentiate Eq. (2-67)
(2-70)
and substitute this into Eq. (2-69), we find the very important result
4'1TacM(9c)
P(Er-~Er)= (1-a)Eias' (2-71)
0, otherwise
To complete the determination of P(Er-~Er), we still need to know the differen-

tial scattering cross section in the CM frame. This knowledge must come from both
a consideration of quantum mechanics and the detailed nuclear physics of the
interaction. Fortunately, however, we can avoid such considerations since the CM
potential scattering cross sections characterizing neutrons of interest in reactor
applications (with energies E< lOMeV) do not depend on 9c for light nuclei (say
A ~ 12). That is, the scattering in the CM frame is isotropic such that
(2-72)
Such behavior is known as "s-wave" scattering (a term which arises from quantum
mechanics) and is the most common form of elastic scattering in nuclear reactors.
For heavier nuclei, there will tend to be some mild angular dependence of acM(9c),
but since elastic scattering from such nuclei does not contribute appreciably to
neutron energy loss in most nuclear reactor types, we will confine our attention
here to s-wave elastic scattering from stationary nuclei. Then, using Eq. (2-72) in
Eq. (2-71), we find that the scattering probability distribution for elastic s-wave
scattering from stationary nuclei takes the form
1
P (Ei➔ Er) = (1- a)Ei ' (2-73)
0, otherwise
44 / INTRODUCTORY CONCEPTS OF. NUCLEAR POWER REACTOR ANALYSIS
Notice in particular that the probability of scattering from an energy Ei to a final

energy Er is independent of the final energy Er.
Using this probability distribution, we can compute the average final energy of a
neutron suffering an elastic scattering collision as
(2-74)
Hence the average energy loss in such collisions is
(2-75)
For example, neutrons suffering scattering collisions in hydrogen (a=O) will lose
on the average half of their original energy in each collision. By way of contrast, in
a scattering collision with a 238 U nucleus, they will lose on the average less than 1%
of their original energy.
We can now write the differential scattering cross section characterizing elastic
(s-wave) scattering from stationary nuclei by substituting Eq. (2-73) into Eq. (2-54)
oiEJ
as<Ei➔ Er)= (1-a)Ei' (2-76)
0, otherwise.
This is about as far as we can go in determining the explicit form of os(Ei➔ Er)
since the elastic scattering cross section as(E) itself depends on the details of the
nuclear potential and is generally obtainable only by measurement. Fortunately
a8(E) characterizing potential scattering is only weakly dependent on energy and
can frequently be taken as constant over a wide range of neutron energies.
Such elastic scattering plays a very important role in nuclear reactor behavior
since it tends to slow the fast fission neutrons down to thermal energies. However
inelastic scattering processes are also important, particularly in fast reactors where
neutron moderation by light isotopes is minimized. Since kinetic energy is not a
conserved quantity in an inelastic scattering collision (the nucleus is left in an
excited state), we can no longer use simple kinematical arguments to determine
P(Ei➔ Er) for such processes. Rather one must rely on measurements of the
differential scattering cross section or on nuclear models. Although such models
are useful for qualitative estimates of neutron scattering, most detailed reactor
studies simply use the measured cross section in essentially tabular form for the
various different energy transfer combinations Ei➔ Er.
One final comment should be made concerning the "other half" of a scattering
event, namely the nuclear recoil. Although this is of little concern to the neutron
economist, it is of very considerable concern to the reactor designer since the recoil
energy of a nucleus suffering a collision with a fast neutron will be sufficient to rip
it completely out of its crystalline lattice. To be more specific, the average recoil
energy of a nucleus suffering an elastic scattering collision with a neutron is just
½(1- a)Ei [recall Eq. (2-75)). For fast neutrons, this recoil energy will be in the keV
to low MeV range. Hence the recoiling nucleus will not only be torn out of its own
lattice position by the collision, but will possess sufficient recoil energy to dislocate
other nuclei in the lattice, leading to significant radiation damage to the material.
This is an extremely important process in materials exposed to the high radiation
environment of a nuclear reactor core and must be taken into account in nuclear
reactor design. We will return to consider it in more detail in Chapter 11.
3. EFFECTS OF NUCLEAR MOTION

Thus far we have studied neutron cross section behavior under the assump-
tion that the target nuclei are at rest; but of course the nuclei are always in a state of
thermal motion. Fortunately the speeds characterizing nuclear motions are often
very much less than those of the neutrons, and for many purposes this nuclear
motion can be neglected entirely.
There are two instances in which such thermal motion of the nuclei must be
taken into account, however. If the neutron speeds are comparable to the nuclear
speeds, then of course one can no longer treat the nuclei as stationary. This will
occur when the neutron energy becomes comparable to the thermal energy of the
nuclei, that is V:::'. Vth=(kT/ M) 112 or E::::'.kT. Such a comparison indicates that for
neutron energies less than roughly I eV, one must specifically account for the
thermal motion of the nuclei. As we mentioned earlier, such low-energy neutrons
are referred to as thermal neutrons.
There is also a situation in which the effects of nuclear motion must be taken
into account even when the neutron speed is much larger than that of the nuclei.
This arises when considering processes in which the cross sections exhibit sharp
resonances. Since the width of these resonances may be quite narrow, much less
than I eV in most cases of interest, even the modest speeds of nuclear thermal
motion can significantly affect the energy dependence of the neutron cross section
in the vicinity of the resonance. Such a phenomenon is known as the Doppler effect,
since it is closely akin to the familiar frequency shift that accompanies variations in
relative motions between source and receiver in sound propagation.
In this section we will first examine the effects of nuclear motion on neutron
cross section behavior for both thermal neutrons and cross section resonances. We
will then briefly discuss the modifications that occur in the differential scattering
cross section when the thermal motion of the nuclei is taken into account.
(a) THERMALLY A VERA GED INTERACTION RATES
Let us begin by considering an interaction between a neutron of velocity v

and a nucleus moving at a velocity V. If the atomic number density of such nuclei
is N, then the interaction frequency for such reactions is given by
lv-Vla(lv-Vl)N, (2-77)
where we have noted that it is the relative speed, lv-Vj, that occurs in this
interaction frequency.
Of course not all of the nuclei in the target sample will be moving with the same
velocity V. In general they will have a distribution of velocities, and in fact, the
interaction frequency we are really interested in is the average of Eq. (2-77) over
these various nuclear velocities. Suppose we define the distribution of nuclear
46 / INTRODUCTORY CONCEPTS OF NUCLEAR POWER REACTOR ANALYSIS.
velocities
~(V)d3V= Number of target atoms/cm3 with velocities (2-78)

Vin d 3 V about V.
Then the probability of a neutron interaction per neutron per second with a
nucleus of any velocity is obtained by averaging Eq. (2-77) over 0L(V)
(2-79)
Notice that we have essentially used this expression to define an averaged cross
section o(v) depending only on the neutron speed v. That is, we can write the
appropriately averaged cross section characterizing neutrons moving with a speed v
through a sample of nuclei with a velocity distribution 0L (V) as
(2-80)
It is important to remember that this is the cross section that would be measured in
an experiment (such as the transmission experiment we described earlier). No
experiment looks directly at the true neutron-nuclear reaction cross section, but
rather measures an average of a(jv - VI) over the distribution of nuclear velocities.
The same is true for applications of such cross sections to nuclear reactor analysis.
Since any reactor is at a finite temperature, the nuclei comprising its core will be in
thermal motion, and hence one must take care to always use cross sections that
have been appropriately averaged over these nuclear velocities. Since this distribu-
tion function depends on the temperature characterizing the material of interest,
the "thermally averaged" cross sections will similarly depend on temperature.
It is useful to consid~r the application of this result to two particularly simple
examples of cross section behavior. We mentioned earlier that many nuclear cross
sections behave essentially as 1/ v (e.g., below a capture resonance). If nuclear
motion is to be included, we would express such behavior as
(2-81)
If we substitute this form into Eq. (2-80) and note that ,the nuclear velocity
distribution function is normalized such that
(2-82)
then we find the averaged cross section becomes just
o(v,T)=-Y. (2-83)
V
Hence in this instance the observed cross section depends on neutron speed v in
exactly the same way as the true cross section depends on relative speeds.
Furthermore the observed cross section is independent of temperature.
A second example of interest is that in which the cross section o(lv - VI) is a very
slowly varying function of relative speed. Since the nuclear velocity distribution is
rather sharply peaked around V = V th• we can approximate Eq. (2-80) by neglecting
V in the relative speed in the int~grand for neutron speeds v» V th· In this case, we
again find that
o(v,T)~o(v) (2-84)
such that the true cross section and the observed cross section are again the same,
and the observed cross section is temperature-independent. This behavior is
frequently exhibited by the scattering cross section for many reactor materials.
It is also of interest to examine the behavior of the averaged cross section for
small neutron speeds v« V th· Then we can replace Iv- VI by Vin the integrand to
find
o(v,T) ~
v-,.O V
-NI fd 3 VVa(V)q)"[(V,T). (2-85)
Since the integral is now just a constant, independent of the neutron speed v, we
find that the behavior of o(v, T) as the neutron speed v becomes very small is just
o( v, T)-- C as v~O. (2-86)

V
This explains the low-energy behavior we observed earlier in the total cross section
for graphite. Of course this result can be explained physically by simply recogniz-
ing that for small neutron velocities, the neutron appears to be essentially
stationary to the more rapidly moving nuclei. Hence the collision rate ceases to
depend on the neutron speed and depends only on the nuclear speed distribution
(i.e., the temperature of the scattering material).
Such averages over the thermal velocity distribution of the nuclei in a material
must always be performed in measuring or utilizing cross sections characterizing
thermal neutrons. However such an average is also extremely important in deter-
mining the correct effective cross sections to use when describing resonance
behavior. ·
Life could become exceedingly complex indeed if a detailed estimate of the
nuclear velocity distribution function 0t (V) were required, since this depends on
the complicated microscopic dynamics of atoms in the reactor (e.g., atomic
vibrations in crystalline lattices or atomic motions in liquids). Fortunately it is
sufficient for most purposes to represent the nuclear velocity distribution by the
Maxwell-Boltzmann distribution characterizing an ideal gas in thermal equilibrium
at a temperature T:
0L(V)=NM(V,T)=N ( 21rkT
M )3/2 exp(-MV /2kT).
2 (2-87)
Then one can write the thermally averaged cross section as
o(v, T)= ¾f d 3 Vlv-Vla(jv-Vl)M(V, T). (2-88)
W,e will make use of this particular average in the next section.
48 / INTRODUCTORY CONCEPfS OF NUCLEAR POWER REACTOR ANALYSIS
(b) THE DOPPLER EFFECT ON CROSS SECTION RESONANCE BEHAVIOR
We have seen that the actual or true cross section depends on the relative
speed between the neutron and the target nucleus. However since the nuclei
themselves are in thermal motion, this relative speed may be either greater or less
than the neutron speed. This difference in relative speeds gives rise to a "Doppler
shift" effect in resonance cross section behavior.
We can take account of this effect by merely substituting the Breit-Wigner
resonance cross section formulas into our expressions for the thermally averaged
cross sections in Eqs. (2-80) or (2-88). For the purposes of this calculation it is
usually adequate to assume that the nuclear velocities are described by a Maxwell-
Boltzmann distribution M (V).
Now recall that the single-level Breit-Wigner formula for a capture resonance
gives the cross section
(2-89)
in terms of the center of mass energy Ee, Our task is to express Ee in terms of the
nuclear velocity vector V and then perform the integration over nuclear velocities
indicated in the averaging formula Eq.(2-88). There 1s really nothing particularly
complicated about this task, except for a bit of vector algebra and the fact that the
integral that arises cannot be explicitly performed (rather, it must be performed
numerically or tabulated).
Since this manipulation is not particularly enlightening, we will only outline the
major steps here and refer the interested reader to more exhaustive treatments that
exist in numerous places throughout the literature. 18- 21 We begin by noting that for
the case of a Maxwell-Boltzmann distribution of target nuclei, one can partially
perform20 the integration over nuclear velocity V to rewrite Eq. (2-88) as
(2-90)
where vr =Iv-VI while vth = (kT / m) 112• If we substitute the Breit-Wigner formula
(2-3j)) for_ ay( v,) into this integral, we find an exact expression for the averaged
cr_efss section
(2-91)
Here we have defined y=2(Ee-Eo)/f where we recall that the CM energy is

Ee=½ µ.v;. However this is where we get stuck, because unfortunately this integral
cannot be evaluated analytically. Fortunately it is rather easy to compute the
integral numerically. However before discussing such calculations, it is useful to
cast Eq. (2-91) into a slightly different form by defining the variables
x-2(E-E0)/r and f =I' /I' 0 (2-92)
where r O is the so-called Doppler width of the resonance
(2-93)
Then one can write
(2-94)
where
i'(r,x)=½f 00
-2E/r
1:
y
2 [exp[- (v-:,)']-exp[- (v+~r)
2vth 2vth
2
]]· (2-95)
In practice, the i'(r,x) would be calculated for each value of f and x of interest
using straightforward numerical integration techniques. To better understand the
implications of such calculations, we have sketched the thermally averaged capture
cross section as determined by Eq. (2-94) in Figure 2-14.
In particular, we have sketched the dependence of this cross section on energy
for several different temperatures T. It should first be noted that as the temperature
T increases, the resonance broadens, while its peak magnitude decreases. For this
reason, one frequently refers to resonance cross sections that have been averaged
over the distribution of nuclear velocities as "Doppler-broadened" cross sections.
It should be stressed that we still have not introduced any additional assump-
tions or approximations into this derivation of the Doppler-broadened resonance
cross section form Eq. (2-94). And in several modern computer codes, Doppler-
Eo E
FIGURE 2-14. Doppler-broadening of a resonance with Increasing temperature.

broadened cross sections are calculated directly from this expression. 19 However
there is an alternative approximate expression for the Doppler-broadened Breit-
Wigner resonance cross section first derived many years ago by Bethe and
Placzek 18 which is more commonly found in textbooks. If one introduces the
following approximations into '1'(t,x):
(a) Neglect the second exponential in the integrand of '1'(t,x).
(b) Replace v' -v; by (v' 2 - v';)/2v'-which is equivalent to approximating
(2-96)
(c) Extend the lower limit of integration toy= - oo;

then one finds
(2-97)
while
(2-98)
The Bethe-Placzek cross section Eq. (2-98) is used very frequently in reactor
calculations and for convenience we have tabulated l[;{r,x) in Table 2-2. The
approximations used to obtain the Bethe-Placzek form break down for high-
temperature target distributions and for low energy resonances. For example, a
comparison for the 0.296 eV fission resonance of 239Pu at T= 2000°C shows a
difference in broadening of 50% between the exact and approximl:,l.te formula-
tions.22
However since it is still traditional to utilize the Bethe-Placzek form in analyzing
resonance behavior, we will include a brief analytical study of its behavior along
with that of the exact expression. First notice that for low temperatures T ➔o we
can see from Eqs. (2-92) and (2-93) that r ➔ oo. Hence the integrand of both '1'(tx)
and l[;(tx) will vanish except in the neighborhood of y,_,x. We can therefore
replace y by x in the denominator of the integrands to write
Hence for low temperatures we arrive at the usual Breit-Wigner form
(2-100)
which is a comforting, although certainly expected, result.

THE NUCLEAR PHYSICS OF FISSION CHAIN REACTIONS I 51
TABLE 2-2 Tabulations of the Doppler Broadening Functionst
The if; Function
r 0 0.5 2 4 6 8 lO 20 40
0.05 0.04309 0.04308 0.04306 0.04298 0.04267 0.04216 0.04145 0.04055 0.03380 0.01639
0.10 0.08384 0.08379 0.08364 0.08305 0.08073 0.07700 0.07208 0.06623 0.03291 0.00262
0.15 0.12239 0.12223 0.12176 0.11989 0.11268 0.10165 0.08805 0.07328 0.01695 0.00080
0.20 0.15889 0.15854 0.15748 0.15331 0.13777 0.11540 0.09027 0.06614 0.00713 0.00070
0.25 0.19347 0.19281 0.19086 0.18324 0.15584 0.11934 0.08277 0.05253 0.00394 0.00067
0.30 0.22624 0.22516 0.22197 0.20968 0.16729 0.11571 0.07042 0.03880 0.00314 0.00065
0.35 0.25731 0.25569 0.25091 0.23271 0.17288 0.10713 0.05724 0.02815 0.00289 0.00064
0.40 0.28679 0.28450 0.27776 0.25245 0.17359 0.09604 0.04566 0.02109 0.00277 0.00064
0.45 0.31477 0.31168 0.30261 0.26909 0.17052 0.08439 0.03670 0.01687 0.00270 0.00064
0.50 0.34135 0.33733 0.32557 0.28286 0.16469 0.07346 0.03025 0.01446 0.00266 0.00063
The x Function
r 0 0.5 2 4 6 8 10 20 40
0.05 0 0.00120 0.00239 0.00478 0.00951 0.01415 0.01865 0.02297 0.04076 0.05221
0.10 0 0.00458 0.00915 0.01821 0.03573 0.05192 0.06626 0.07833 0.10132 0.05957
0.15 0 0.00986 0.01968 0.03894 0.07470 0.10460 0.12690 0.14096 0.12219 0.05341
0.20 0 0.01680 0.03344 0.06567 0.12219 0.16295 0.18538 0.19091 0.11754 0.05170
0.25 0 0.02515 0.04994 0.09714 0.17413 0.21909 0.23168 0.22043 0.11052 0.05103
0.30 0 0.03470 0.06873 0.13219 0.22694 0.26757 0.26227 0.23199 0.10650 0.05069
0.35 0 0.04529 0.08940 0.16976 0.27773 0.30564 0.27850 0.23236 0.10437 0.05049
0.40 0 0.05674 0.11160 0.20890 0.32442 0.33286 0.28419 0.22782 0.10316 0.05037
0.45 0 0.06890 0.13498 0.24880 0.36563 0.35033 0.28351 0.22223 0.10238 0.05028
0.50 0 0.08165 0.15927 0.28875 0.40075 0.35998 0.27979 0.21729 0.10185 0.05022
tT. D. Beynon and I. S. Grant, Nucl. Sci. Eng. 17, 547 (1963).
At the other extreme, the high temperature limit T-HX) implies that r➔ o. In this
case we find
a (E
Y ,
T) ➔ o
rY ( Eo )1;2 t-
or -E -
-
2-vii
1 00
dy
---exp
I+ Y2
[- -
;2x2
4
- l
-oo
(2-101)
which is a Gaussian shape characterized by the Doppler width r O rather than the
"natural" line width r. Hence as the temperature increases, the resonance broadens
out from its natural width to eventually approach a width that depends on the
temperature as r½. · .
One other observation is important. If the Bethe-Placzek form is examined in

greater detail, it becomes apparent that regardless of the temperature, the area
under the resonance remains constant. We can demonstrate this by simply integrat-
ing over the resonance
r
Jr
-
dEo.,,(E, T) = o0 y
r.,, Lr dx'IJ;(r,x). (2-102)
resonance resonance
However since the contribution to the integral from the "wings" of the resonance
(far-off resonance) are so small, we can approximately extend the range of
integration to ± oo so that we can explicitly perform the integral to find
JdEa.,,(E,T)r;;;;r.o yr.,, Joo-ool+y

0 -dy- Joo dx exp[ -¼(x-y)2f 2] =o r.,,-,,,2 ,
2
-oo
0 (2-103)
a result which is temperature-independent.

It should be pointed out that such behavior is a consequence of the Bethe-
Placzek approximation. More generally, the area under the Doppler-broadened
resonance given by the exact expression Eq. (2-94) will in fact change with
temperature. 22 However for the temperature range and resonances of interest in
most reactor applications, this change is relatively small, and one can essentially
assume that the area under the resonance is relatively insensitive to temperature
changes. This fact will prove of some importance in our later study of reactor
behavior, since it will imply that an increase in the temperature of the absorbing
material will increase the rate at which neutrons are absorbed in the resonance
range of the cross section.
Thus far we have only examined the effect of thermal nuclear motions on a
capture resonance. However we could of course have also substituted in our
expression for scattering cross sections in the vicinity of a resonance, Eq. (2-38),
into the expression for thermally averaged cross sections Eq. (2-88) and labored
through some algebra to again arrive at Doppler-broadened cross sections which
again involve integrals that cannot be performed explicitly. We will avoid the
agony of such manipulation and only state the result of such labors here:
(2-104)
where we have defined another tabulated function 21 x(f,x) (see Table 2-2) to
characterize the interference term:
X =-f-loo
x(r, )- , ~ dy
yexp[ -¼f2(x-y)2]
1+y 2 (2-105)
V7T
-oo
We will later utilize these expressions to study the absorption of neutrons in ·

nuclear reactors. Although we have consistently denoted the thermally averaged
cross sections with an over-bar, for instance, as(E, T), we will usually omit this
notation in future discussions and simply regard all resonance cross sections as
having been Doppler-broadened.
(c) DIFFERENTIAL SCATTERING CROSS SECTIONS WITH UPSCATTERING
We will now discuss the modifications necessary in our study of the differen-
tial scattering cross section characterizing elastic potential scattering when effects
of nuclear motion must be taken into account. Recall that in our earlier discussion
we found that the neutron could not gain energy in an elastic collision with a
stationary nucleus. It can only lose energy in such a collision. It is customary to
refer to such a process as "downscattering/' since the neutron will scatter down in
energy.
We can make this more explicit by considering a particularly simple example in
which an incident neutron scatters elastically from a stationary hydrogen nucleus
(a proton). Then the scattering probability distribution is just
(2-106)
That is, the scattering probability is independent of the final energy Er and
vanishes for Er> Ei (corresponding to upscattering in energy).
Let us now consider the situation in which the neutron suffers elastic scattering
collisions in a hydrogen gas at finite temperature T in which the nuclei are in
motion with a Maxwell-Boltzmann velocity distribution M (V, T). It is necessary to
repeat our earlier consideration of two-body kinematics to include the motion of
the target nucleus. One would then have to average the cross section characterizing
such scattering over the nuclear velocity distribution M (V, T), just as we did in the
previous sections. Since these tasks are rather tedious, we will simply note that the
results of such calculations23 are that for such a proton gas at temperature T, the
scattering probability is given by
(2-107)
where the error function is defined by
2 rx
erfx== --Jn dte- 1 •
2
(2-108)
\{:ii 0
We have plotted P(Ei➔ Er) for several incident neutron energies Ei in Figure 2-15.
_ It should first be noted that unlike the situation in which the hydrogen nuclei were
initially at rest, the scattering probability now depends on the final energy Ef'
Furthermore this probability is not zero for Er> Ei for. a finite temperature gas,
hence implying that it is possible for the neutron to gain energy in a scattering
collision. Such "upscattering" events are significant for incident neutron energies
up'to about lOkT. Above this energy, the scattering probabilhy begins to resemble
that characterizing stationary nuclei (i.e., a T=O hydrogen gas).
0 0.5 1.0 1.5 2.0
FIGURE 2-15. The scattering probability distribution P(Er-+Er) cbaracterizing neutron scatter-
ing from a proton gas at temperature T.
One could repeat this calculation for a free gas of arbitrary mass number A, but
the expressions for the scattering probability and differential scattering cross
section become quite complicated. Since there are also other effects important in
low-energy neutron scattering such as chemical binding and diffraction, we will
defer a more detailed discussion of thermal neutn;:m cross section behavior until
Chapter 9. However even this brief discussion should indicate the importance of
accounting for the thermal motion of the target nuclei in measuring or using
low-energy neutron cross sections.
II. NUCLEAR FISSION
A. Fission Physics
The binding energy per nucleon in atomic nuclei reaches a maximum of
8.7MeV for nuclei mass numbers of about 50 (see Figure 2-16). Hence it is possible
to produce more tightly bound nuclei and thereby release energy by either fusing
together lighter nuclei (nuclear fusion) or inducing a heavy nucleus into fissioning
into two nuclei of intermediate mass number (nuclear fission). The observed
stability of heavy nuclei against spontaneous fission is due to the short-range
nuclear forces within the nucleus giving rise to a potential energy barrier that must
be overcome before the nucleus will fission. The size of this fission barrier is
typically 6-9MeV in most heavy nuclei of interest. Hence to induce nuclear fission,
one must add a sufficient amount of energy to the heavy nucleus to overcome this
fission barrier.
This can be done in a variety of ways. One could simply slam an energetic
particle (with kinetic energy greater than the fission barrier) into the nucleus. An
example of such a reaction would be photofission, in which a high-energy gamma
strikes a heavy nucleus, thereby inducing fission. An alternative scheme would be
to let the heavy nucleus capture a neutron. Then the binding energy of the added
neutron itself might be sufficient to overcome the fission barrier and induce fission.
This later process can in fact occur in certain heavy nuclei such as 233 U, 235 U,
THE NUCLEAR PHYSICS OF FISSION CHAIN REACTIONS /_ SS
...:g_
~ 4
...
"'C:
O,·
-~ 3
-0
C:
ii:i
2
20 40 60 80 100 120 140 160 180 200 220 240

Mass number
FIGURE 2-16. Binding energy per nucleon versus mass number.
239 Pu, and 241 Pu. Such nuclides that can be induced to fission with neutrons of
essentially zero kinetic energy (or of more relevance to nuclear reactor applications,
thermal neutrons having very small kinetic energies, at least compared to nuclear
energies) are referred to as fissile nuclides. We will see later that such fissile
nuclides represent the principal fuels used in fission chain-reacting systems.
With most heavy nuclides, the additional binding energy provided by a captured
neutron is not sufficient to push the heavy nucleus over the fission barrier.
Frequently, however, one can add a dash of extra energy to the neutron, for
instance by giving it a kinetic energy of an MeV or so, and this is sufficient to lift
the nucleus the rest of the way over the barrier to cause fission. Nuclides that can
be fissioned with such "fast" neutrons are referred to as fissionable. Examples are
232Th, 238 U, and 240Pu (as well as fissile nuclei such as 235 U). Although such
fissionable nuclides do play an important role as nuclear fuels they are unable to
sustain by themselves a stable fission chain reaction and hence must always be
used in combination with a fissile nuclide such as 235 U or 239 Pu.
There is also a small possibility that certain heavy nuclei will fission spon-
taneously via the barrier penetration mechanism familiar from quantum
mechanics. However the probability for such an event is quite low in most nuclides
of interest as nuclear fuels. For example, the half-life for spontaneous fission in
238 U is some 6.5 x 10 15 years. However even this very slow spontaneous fission rate
can be of importance in nuclear systems, since even a few neutrons can be rapidly
multiplied to appreciable numbers in a growing chain reaction.
B. Fission Cross Sections

We noted earlier that nuclear fission is a process that proceeds via compound
nucleus formation, much as does radiative capture. Hence it is not surprising that
fission cross sections show considerable resonance structure. In Figure 2-l 7 we
N - ~coco 'II' N
-
u,
0
C
.2
Ci
'i!iOR
-
"b
&i
I
ii: 13
C ~
g !
w
:::,
a,
C
-
"'o
-
LJ...W.c>LaJuulu.ul.u..Lo.L....o-'--L._.....wwaww,1w.iliu..u.~........i.o1.1.1..iwww1u,ili.u._,__i_,_L.L...W 10
FIGURE 2-17. Fission cross sections of 233 U, 235U, and 239Pu.
S6
C"I
...
C')
...
0
...
0 ...0
"o ...
LD
...
0
C
0
·.:.u C')
0
II
C i
·i .~ I
i! u
....,.
C
0
.
~
QI
C
:::, w
C
No
...0
0
0
N
l.u..~ll,ul.,J,wJwili_w_w_wj_u...,_Dlmi,wtl,wJw.a.u....L._LL..>_J,j,olulwiw,Lwluu_l LU..LJ..LW-'-'"""' IQ
FIGURE 2-17. ( Continued)

57
"0
It)
-
0
) ~
' '<I"
0
( ~
\
} ~
1
'
:::: '
~
C:
0 -M
·.;:;
I.I
- "'
0
C:
0 .
'ti e
u:
lR
Cl)
I.I
l >
>
e>
C GI
0 C:
~ w
~
"'C N
0
r 0
0
-
N
11111 :.Jw..,...-~Ld~.......,-=clu_c.uow~<>llWW.u1.u..L.L->1..wu1w.1.w,~
..
10
FIGURE 2-17. (Continued)

58
have shown the fission cross sections characterizing the principal fissile nuclides,
233 U, 235 U,
and 239Pu, taken from ENDF /B-IV.
The indicated cross section behavior is very similar to that of radiative capture
cross sections. However this would be expected since we have seen that compound
nucleus formation via neutron absorption is essentially independent of the mode of
compound nucleus disintegration or decay, for example, via fission or gamma
emission. It is particularly important to note that the fission cross section is over
two orders of magnitude larger for low-energy or thermal neutrons than for
high-energy fast neutrons (above I keV). The thermal neutron fission cross sections
are indeed enormous for these fissile isotopes, ranging up to thousands of barns in
magnitude. Such behavior will prove of very considerable importance in our later
studies of nuclear reactors.
We have also indicated the fission cross sections characterizing the principal
fissionable nuclides of interest, 232Tb, 238 U, and 240Pu (see Figure 2-18). This
cross section behavior is somewhat different than that characterizing fissile nuc-
lides since fissionable nuclides can only be fissioned by sufficiently high-energy
neutrons. This implies that their fission cross sections will have a threshold energy,
below which the cross section drops to zero. Even above this threshold energy
(roughly I MeV), the fission cross sections are quite low, being less than two barns.
When a neutron is absorbed by a fissile isotope such as 235 U, it may induce that
isotope to fission. Yet it is also possible that the compound nucleus formed by the
neutron absorption, 236 U*, might simply decay to its ground state by gamma
emission. The relative balance between the probability of fission. and radiative
capture is an extremely important factor in nuclear reactor applications. We
characterize this balance by the capture-to-fission ratio, defined by
(2-109)
3 r-----r---r---...-----,,-----r----.---..-----,----r----,
242pu - - - - - - -
t:>
2 4 6 8 10
Neutron energy, Mev
FIGURE 2-18. Fission cross sections of principal fissionable isotopes. 12

This ratio depends not only on the isotope of interest, but as well on the incident
neutron energy E. It is plotted in Figure 2-19 for the three primary fissile nuclides.
It can be seen that most neutron absorption in such isotopes leads to fission events
(with the exception of a small range of a > 1 for 235 U). It should be noticed that a
decreases quite appreciably above 0.1 MeV. This latter fact will prove to be of
considerable importance when we discuss the concept of a fast breeder reactor.
C. Nuclear Fission Reactions

A typical nuclear fission reaction such as
· bn+ 2~~U➔(~~U*) ➔fission reaction products (2-110)
spews out a variety of reaction products, including the fissioned nuclei or fission
products and several neutrons as well as numerous gammas, betas, and neutrinos.
And of course it also releases a very considerable amount of energy. Indeed a
glance at the binding energy per nucleon before and after the fission reaction from
Figure 2-16 suggests that energy on the order of 200MeV will be released in each
fission reaction.
The fission fragment nuclei produced by the fission reaction are both highly
charged and highly energetic. They slow down via collisions with adjacent atoms,
losing energy and charge (picking up electrons) in the process. This is in fact the
principal mechanism by which the fission energy eventually appears as heat
generated in the fuel material. However these fission products are usually quite
unstable as well, being somewhat neutron-rich, and will subsequently decay,
usually via beta emission. The energy released in such radioactive decay reactions
can amount to as much as 4-5% of the total energy released in the fission reaction.
Since such "decay heat" will appear with an appreciable time delay corresponding
to the half-lives of the various nuclei involved, it can lead to difficulties unless
properly anticipated in fission reactor design.
0
0
o- Q] U-235
t- & U-233
a: X PU-239
0::
I ORTA IN CLOS LY-SPRC D
z RESDN 1NCE REG I ON SU Pf RESSE □ l
oo
1--1 •
(f) ....
(f) ·•""\Is,,,..
><•x•"',; Ii
"'
-~~
LL )C X xx>00t
0
I
""' ....... ,
---
t--- "' """- ,"' "' ~ il:o,![J,:i,.
o
w-
0::
:::J
1
• --
~- J
."' " -.
~4-A....- .
,,_,-ii'.
'lie,
.
I-
Q_
•,
'
~ .,.
a: xe
~
<
LJ
X 1:!.Jl!lr- •~
-
-~ \.
l!I '°'1',
0
'10- 3 10- 2 10-I 10° 10+ 1 10+ 2 10+ 3 lO+ll 10+5 10+ 6 10+ 7
NEUTRON ENERGY, EV
FIGURE 2-19. Variation of o: with energy for 233 U, 235 U, and 239Pu.
Yet just as significant as the energy released in the fission reaction is the fact that
several neutrons are also produced in the reaction. These neutrons can be used to
propagate a fission chain reaction. Most of these fission neutrons appear essentially
instantaneously (within 10- 14 sec) of the fission event. These neutrons are referred
to as prompt. However a very few neutrons (less than 1%) appear with an
appreciable time delay from the subsequent decay of radioactive fission products.
Although only a very small fraction of the fission neutrons are delayed, these
delayed neutrons are vital for the effective control of the fission chain reaction.
The total number of neutrons (both prompt and delayed) released in a fission
reaction will vary. However in most nuclear applications we only need concern
ourselves with the average number of neutrons released per fission, which we
denote by P. This quantity will depend on both the nuclear isotope involved and
the incident neutron energy, generally tending to increase with increasing neutron
energy. We have shown P(E) as a function of energy for the principal fissile
isotopes in Figure 2-20.
The neutrons produced in the fission reaction emerge with a distribution of
energies, with the average fission neutron energy being roughly 2MeV. As with
other fission parameters, this distribution will depend on the nuclear isotope
involved, to a lesser degree on the incident neutron energy, and will differ for
prompt and delayed neutrons. To characterize this variation in fission neutron
energy, it is convenient to define the fission neutron energy spectrum, or more
simply, the fission spectrum, x(E), defined as
x( E) dE - Average number of fission neutrons emitted with

energy E in E to E + dE per fission neutron. (2-111)
v49= 2.874 + 0.138 E

v25 = 2.432 + 0.066 E(0 .,;; E .,;; 1)
= 2.349 + 0.15E (E > 1)
3 v 23 = 2.482 + 0.075 E (0 .,;; E ..;; 1 )
= 2.412 + 0.136 E IE> 1)
2 ,__......___.___._......___.__,....__........__.__,....__......___,,___,....__........__.__,....___ _ _ _ __
0 5 10 15
E (MeV)
FIGURE 2-20. Average neutron number per fJSSion,, as a function of energy24

2 3 4 5
E (MeV)
FIGURE 2-21. Flmon spectrum for thermal neutron Induced fission in nsu.
A typical prompt neutron fission spectrum is shown in Figure 2-21. One can also
represent the fission spectrum by a simple empirical expression; for example, the
prompt neutron fission spectrum of 235 U is given by
x( E) = 0.453 e- 1.o36 E sinh V2.29 E . (2-112)
This proves useful in performing simple estimates of fission chain reaction be-
havior. However more elaborate studies would simply use the fission spectra
tabulated in a nuclear data set such as ENDF /B.
Because of the importance of delayed fission neutrons to nuclear reactor control,
it is useful to introduce a few related concepts useful for their description. By way
of example, consider a typical fission product decay scheme leading to the emission
of a delayed neutron as sketched in Figure 2-22. It should be noted in particular
that the decay sequence leading to the delayed neutron emission is first the beta
decay of 87Br to 87Kr*, followed by the subsequent decay of 87 Kr* to 86Kr via
neutron emission. The effective time delay of this process is controlled by the
beta-decay-in this case, the half-life is some 55 sec. We refer to the fission
fragment whose beta-decay yields a daughter nucleus which subsequently decays
via delayed neutron emission as a delayed neutron precursor. Of course a very large
number (at least 45) of different delayed neutron precursor isotopes will be
produced in a fission chain reaction. It has been customary (and found to be
adequate) in reactor analysis to group these precursors into six classes
characterized by approximate half-lives of 55, 22, 6, 2, 0.5, and 0.2-sec, respectively ..
Each precursor group will contain a number of different isotopes. For example,
while the 55-sec precursor group is due almost entirely to one precursor, 87 Br, there
are at least two major contributors, 88 Br and 1371, to the 22-sec group. The
composition of the remaining groups are considerably more complex.
87 Br 55s
Neutron
emission
FIGURE 2-22. Decay of the 87 Br delayed neutron precursor.
Since the relative isotopic yield per fission will vary for different fuel isotopes,
the detailed characteristics of the precursor groups will similarly be isotope-
dependent. To this end, let us define:
"A;= Decay constant ( ,8-decay) of ith precursor group

,B; = Fraction of all fission neutrons (both prompt and delayed) emitted per
fission that appear from ith precursor group
,8 = ~ ;,B; = Total fraction of fission neutrons which are delayed.
In Table 2-3 we have listed the half-lives, relative yield fractions ,8;/ ,8, and total
delayed neutron yields vd = v,B, for the precursor groups characterizing the principle
fissionable isotopes recommended by ENDF /B-IV25 . Although dependent on the
fuel isotope, these data do not depend sensitively on the incident neutron energy
below about 4 MeV and hence can be used for either thermal or fast reactor
analysis.
The energy spectrum of delayed fission neutrons is considerably lower than that
of prompt fission neutrons and again depends on both the delayed neutron group
and fissioning isotope. We have given a rough composite delayed neutron fission
spectrum in Figure 2-23 along with typical measured spectra for the 55-sec and
22-sec groups of 235 U. More detailed spectrum data can be found in the review
article of Cox. 25
Actually there are additional processes that can contribute delayed neutrons to
the chain reaction. Photoneutron reactions (y,n) are particularly important in
reactors containing appreciable amounts of deuterium or beryllium. The decay
times of these processes are even longer than those characterizing delayed fission
neutrons (ranging up to 125 m). 25 However these photoneutrons can usually be
accounted for in reactor analysis by simply including one or more additional
groups of delayed neutrons such as those tabulated above. Furthermore one can
64 / lNTRODUCTORY CONCEPTS OF NUCLEAR POWER REACTOR ANALYSIS
TABLE 2-3 Delayed Neutron Yield and Half-Life Data25
232Th: vd = 0.0527 ± 0.0040 n / f 233U: v d = 0.0074 ± 0.0004

Group T1 tsec) Relative Yield Group T1 (sec) Relative Yield
2 2
1 56.030 0.034 1 55.110 0.086

2 20.750 0.150 2 20.740 0.274
3 5.740 0.155 3 5.300 0.227
4 2.160 0.446 4 2.290 0.317
5 0.571 0.172 5 0.546 0.073
6 0.211 0.043 6 0.221 0.023
235U: r,d=0.01668±0.00070 n/f 238U: vd=0.0460±0.0025 n/J

Group Ti (sec) Relative Yield Group T 1 (sec) Relative Yield
2 2
1 54.51 0.038 ± 0.004 I 52.38 , 0.013

2 21.84 0.213±0.007 2 21.58 0.137
I
3 6.00 0.188 ± 0.024 3 5.00 0.162
4 2.23 0.407 ± 0.010 4 1.93 0.388
5 0.496 0.128±0.012 5 0.493 0.225
6 0.179 0.026 ± 0.004 6 0.172 0.075
2l9Pu: r,d=0.00645±0.00040 n/j 2'!0J>u: vd=0.0090±0.0009 n/f

Group Ti (sec) Relative Yield Group T1 (sec) Relative Yield
l l
1 53.15 0.038 1 53.56 0.028

2 22.29 0.280 2 22.14 0.273
3 5.19 0.216 3 5.14 0.192
4 2.09 0.328 4 2.08 0.350
5 0.549 0.103 5 0.511 0.128
6 0.216 0.035 6 0.172 0.029
241 Pu: pd =0.0157±0.0015 n/ f

Group Ti (sec) Relative Yield
l
1 54.0 0.010
2 23.2 0.229
3 5.6 0.173
4 1.97 0.390
5 0.43 0.182
6 0.2 0.016
THE NUCLEAR PHYSICS OF FISSION CHAIN REACTIONS / 6S
1.0
I
I
\ \
I \ '\
I \ '\
\ \\
\.' \
\'
',,
' '-~~1 ',
-- --
(235U)
1.0 2.0
Neutron energy E (MeV)
FIGURE 2-23. Composite delayed neutron spectrum.
also find delayed neutrons generated by neutron-absorption processes such as

certain (n,p) reactions, for example, 110(n,p) 17N, since 17N decays by neutron
emission with a half-life of 4.165 sec.
The energy released in a nuclear fission reaction is distributed among a variety
of reaction products. We have classified these reaction products as to both range
and emission time, and have indicated the approximate percentage of the fission
energy (some 200 MeV) carried by each:
TABLE 2-4 Energy Release in Nuclear Fission
Reaction Product Energy(%) Range Time Delay
Kinetic energy of
fission fragments 80 <.01 cm instantaneous
Fast neutrons 3 10-100 cm instantaneous
Fission gamma energy 4 100cm instantaneous
Fission product p decay 4 short delayed
Neutrinos 5 nonrecoverable delayed
Nonfission reactions
due to neutron
capture 4 100cm delayed
The majority of the fission energy appears as the kinetic energy of the fission
fragments and is deposited essentially at the point of fission in the nuclear fuel.
Note, however, that some of the fission energy appears as kinetic energy of
neutrons (3%) and gammas (4%) with relatively long ranges. This energy will be
distributed over the core of the reactor and adjacent material such as shielding. In
Figure 2-24 we have noted the types of emergent radiation.27• 28
Furthermore it should be noted that some 4% of the fission energy appears in the
form of heat generated by the decay of radioactive fission products. If the nuclear
reactor were to be suddenly shut down, this decay heat would continue to be
produced and would have to be removed; otherwise the reactor core temperature
would rise dramatically, causing fuel element melting and failure. The removal of
such decay heat is one of the most serious problems in reactor safety studies.
Notice also that a sizable amount of energy (as much as 20MeV per fission) may
be liberated by the high-energy gammas produced in radiative capture (n, y)
reactions.
It is customary to use an effective energy release per fission in determining the
portion of the total energy of fission that can be recovered by a coolant and hence
contributes to the thermal power output of the reactor. Although this energy will
vary somewhat with the type of reactor and the detailed core composition, it is
typically of the order of 192MeV. (A more detailed tabulation of useful energy
release per fission has been calculated using atomic mass data by James 25 and is
given in Table 2-5.) Of this 192MeV, some 168MeV appears as fission fragment
energy, while 7 MeV appears as beta energy. These short-range contributions
deposit their energy in the nuclear fuel. If we also take into account the energy
deposited in the fuel ( ,_, 7%) due to fast neutrons and gammas, we find that some
,c,
fission -y's
~ic
scattering -y's
~
neutrons
•
F.1ss1on
. •/ / "
fragment
~:~~:,(' ~
••-->• ~i,,~
Fission
• ➔
Neutron
••---,),•* Capture -y's
Radiative
8J slowing capture
~ down
Radioactive'-:., -::::- --s,..
decay of ~
fission
fragment .....
--~~ Bremsstrahlung
---~ ~
~
Decay {3's
Decay~
FIGURE 2-24. Types of radiation produced in a fission chain reaction.

97% of the recoverable fission energy is deposited directly in the fuel material. The
remainder is deposited in the coolant or structural materials by neutrons and
gamma radiation, with less than 1% typically being deposited in shielding due to
gamma radiation. Actually as we will see in Chapter 12, the energy deposited in
other regions of the reactor is usually reassigned to the fuel in order to simplify the
thermal analysis of the reactor.
TABLE 2-S25 Effective Energy Released in Fission
The effective energy released in and following fission of the principal fissile isotopes by thermal
neutrons are:
233 U: 190.0±0.5 MeV /f
235 u: 192.9±0.5 MeV /f
239 Pu: 198.5±0.8 MeV /f
241 Pu: 200.3±0.8 MeV /f
The effective energy released following fission of the major fissionable isotopes by 235 U fission spectrum
neutrons are:
232Th: 184.2±0.9 MeV /f
234U: 188.9± 1.0 MeV /f
236U: 191.4±0.9 MeV /f
23au: 193.9±0.8 MeV /f
237Np: 193.6± 1.0 MeV /f
238Pu: 196.9±0.8 MeV /f
240Pu: 196.9± 1.0 MeV /f
242Pu: 200.0± 1.9 MeV /f
These values include all contributions except from neutrinos and very long-lived fission products.
D. Fission Fuels
Our previous discussion has indicated that there are a number of possibilities
available for fueling a fission chain-reacting system. In particular, we have noted
that the principal nuclidest of concern in nuclear reactor applications are:
Fissile nuclides: 233 U, 235 U, 239 Pu, 241 Pu
while those susceptible to fast neutron fission are:
Fissionable nuclides: 232Tb, 238 U, 240 Pu, 242 Pu
Because of both the energy threshold that neutrons must exceed in order to induce
fission in fissionable nuclides and the relatively large value of a characterizing such
nuclides, only the first class of nuclides are capable of sustaining a fission chain
reaction. Of these isotopes, only 235 U is found in nature-and then, only as 0. 711 %
of natural uranium (which is composed primarily of 238 U). Although reactors can
be fashioned out of natural uranium with even this low concentration of 235 U if one
t A bit of conventional notation26 for such nuclides remains as debris from the secrecy of the
atomic weapons program during World War II. Two-digit code numbers are used to identify
each isotope where the first digit is the atomic number minus 90 and the second digit is the last
digit of the mass number. Hence 235 U is denoted as "25," 238 U as "28," 239Pu as "49," and so on.
is sufficiently clever, most present-day reactor types are fueled with uranium in
which the percentage of 235 U has been increased or enriched above its natural
value. As we will see later, such uranium enrichment is an extremely complicated
and expensive process.
There is yet another way to obtain fissile isotopes, however. It is found that when
certain nuclides absorb neutrons, they then undergo a sequence of radioactive
disintegrations that eventually result in the formation of a fissile isotope. The two
most important examples of such neutron transmutation reactions are:
23sU(n,y)239U I!;, 239Np f!.;, 239Pu

23min 2.3d
Isotopes that can be transmitted into fissile nuclides via neutron capture are
referred to as fertile. The fertile isotopes of most interest are 238 U and 232Th, which
are in abundant supply throughout the world.
Yet where does one find the neutrons necessary for this process? In a nuclear
reactor. Indeed since most present-day reactors are fueled with low-enrichment
uranium that may contain as high as 98% 238 U, such transmutation processes will
occur quite naturally as the fertile nuclei capture excess neutrons from the fission
chain reaction. The key parameter in such processes is the number of neutrons
produced in each fission reaction per neutron absorbed in the fuel nuclei. (Here we
must remember that not all neutron absorptions in the fuel lead to fission-some
result in radiative capture.) We will define
'I} =Averagenumber of neutrons produced per neutron

absorbed in fuel.
For a fuel composed of a single fissile isotope, we can write
(2-113)
Most fuels, however, contain a mixture of isotopes. In this case, we would use the
macroscopic fission and absorption cross sections characterizing each isotope to
write
L P/2.{
j
11= (2-114)
L~{
j
The dependence of this very important quantity on energy E is shown for the
four principal fissile isotopes in Figure 2-25. It should be noted that 71(£) is
generally of the order of 2 for low-energy neutrons, but increases with energy
above 0.1 MeV as the capture-to-fission ratio a falls off. If we are to attempt to
utilize the neutrons "left over" from the chain reaction to convert fertile isotopes
into fissile material, it is apparent that we require 11(E) to be at least greater than 1,
since one neutron per fission is needed to sustain the chain reaction. Of course, a
o,____..____....,____--'-___._____._______.___.____...___ ____,
10-2 10-1 10 1 10 2 103 10 4 ,as ,oe
Neutron energy (eV)
FIGURE 2-25. Variation of 71 with energy for 233 U, 235 U, 239 Pu and 241 Pu.
certain fraction of the fission neutrons will be absorbed in nonfuel materials, and
others will leak out of the reactor and be lost to the chain reaction. Nevertheless it
is apparent that 11(E) is sufficiently greater than unity to enable appreciable
conversion using any of these isotopes.
Indeed it might even be possible to produce more fissile material than one
depletes in maintaining the fission chain reaction. For this to occur, one would
have to operate with fissile isotopes and neutron energies for which 71(£) was
greater than two, since one neutron would be needed to maintain the chain
reaction, while one neutron would be used to produce a new fissile nucleus to
replace the one destroyed in the fission reaction. Any excess over this (and over the
number of neutrons lost to the chain reaction via nonproductive capture or
leakage) could then be used to produce or breed new fissile material.
It is apparent from Figure 2-25 that the most favorable situation for accomplish-
ing this would involve relatively fast neutrons in the 0.1-1 MeV range. The most
suitable fuel would be 239Pu. Such is the motivation behind the development of the
70 / INTRODUCTORY CONCEYI'S OF NUCLEAR POWER REACTOR ANALYSIS
fast breeder reactor which operates with a chain reaction in a 239Pu/ 238 U fuel
mixture maintained by fast neutrons in order to achieve this large value of 17.
However if we recall the energy dependence of the fission cross section itself, it is
apparent that it is more difficult to use fast neutrons to sustain the chain reaction,
since the cross sections for fast fission are some two orders of magnitude smaller
than those characterizing thermal neutrons. This suggests that it might be easier to
achieve a sustained chain reaction using slow neutrons, since then the probability
of fission is appreciably larger. Yet we must remember that the neutrons produced
in the fission reaction are quite energetic with average energies in the MeV range.
Hence in order to take advantage of the large fission cross sections for slow
neutrons, one must slow down the fast fission neutrons to thermal energies
( < 1eV). As we will see in the next chapter, this can be accomplished rather easily
by using elastic scattering collisions.
REFERENCES
1. I. Kaplan, Nuclear Physics, 2nd Edition, Addison-Wesley, Reading, Mass. (1963).

2. 0. Oldenberg and N. C. Rasmussen, Modern Physics for Engineers, McGraw-Hill, New
York (1966).
3. M. G. Bowler, Nuclear Physics, Pergamon, New York (1973).
4. A.H. Foderaro, The Elements of Neutron Interaction Theory, M. I. T. Press, Cambridge
(1971).
5. A. M. Weinberg and E. P. Wigner, The Physical Theory of Neutron Chain Reactors,
University of Chicago Press (1958).
6. J. R. Lamarsh, Introduction to Nuclear Reactor Theory, Addison-Wesley, Reading, Mass.
(1966).
7. A. deShalit and H. Feshbach, Theoretical Nuclear Physics, Wiley, New York (1974).
8. C. M. Lederer, J. M. Hollander, and I. Perlman, Table of Isotopes, 6th Edition, Wiley,
New York (1968).
9. H. Goldstein, in Fast Neutron Physics, Vol. II ed. J. B. Marion and J. L. Fowler,
Interscience, New York (1963), p. 2227.
10. K. H. Beckurts and K. Wirtz, Neutron Physics, Springer, Berlin (1964).
11. D. J. Hughes, et al., Neutron Cross Sections, BNL-325 (1955).
12. D. J. Hughes, et al., Neutron Cross Sections, 2nd Edition and Supplements, BNL-325
(1958, 1960, 1964, 1965).
13. Neutron Cross Sections, 3rd Edition, BNL-325 (1975).
14. H. C. Honeck, ENDF /B-Specifications for an Evaluated Nuclear Data File for
Reactor Applications, BNL-50066 (1966); P. B. Heming, Development of ENDF /B
System, in Proceedings of 3rd Conference on Neutron Cross Sections and Technology,
USAEC CONF-71031 (1971), pp. 348-353.
15. A partial list of sources of neutron cross section data is given in the following references:
CINDA 74 (and succeeding years), An Index to the Literature of Microscopic Neutron
Data, International Atomic Energy Agency, Vienna (1974).
R. J. Howerton, S. T. Perkins, C. P. Altamirano, and K. L. Hill, A Bibliography of the
Experimental Data of Neutron-Induced Reactions, UCRL-50400, Vol. 2 (1970).
16. D. I. Garber, L. G. Stromberg, M. D. Goldberg, D. E. Cullen, and V. M. May, Angular
Distributions in Neutron Induced Reactions, BNL-400, 3rd Edition (1970).
17. D. Okrent, R. Avery, and H. H. Hummel, Proceedings of the International Conference
on Peaceful Application of Atomic Energy, Vol. 5 (1955), p. 357.
18. H. A. Bethe and G. Placzek, Phys. Rev. 51, 450 (1937).
19. W. G. Pettus, Nucl. Sci. Eng. 31, 168 (1968).
20. R. L. Murray, Nucl. Sci. Eng. 26, 362 (1966).

21. L. Dresner, Resonance Absorption in Nuclear Reactors, Pergamon, (1960).
22. D. E. Cullen, Nucl. Sci. Eng. 54, 498 (1974); D. G. Lindstrom, U. of Michigan, M.S.
Thesis ( 1960).
23. E. P. Wigner and J. E. Wilkins, Jr., AECD-2275 (1944); M. M. R. Williams, The Slowing
Down and Thermalization of Neutrons, North-Holland, Amsterdam (1966), p. 75.
24. G. R. Keepin, Physics of Nuclear Kinetics, Addison-Wesley, Reading, Mass. (1965).
25. S. A. Cox, USAEC Document ANL/NDM-5 (1974); L. Tomlinson, UKAEA AERE-
R6993 (1972); R. J. Tuttle, Nucl. Sci. Eng. 56, 37 (1975).
26. M. F. James, J. Nucl. Energy 23, 517 (1969); 25, 513 (1971).
27. S. Glasstone and M. C. Edlund, The Elements of Nuclear Reactor Theory, Van Nostrand,
Princeton, N. J. (1952).
28. N. M. Schaeffer, Reactor Shielding for Nuclear Engineers, USAEC Document TID-25951
(1973).
PROBLEMS
2-1. What target isotope must be used for forming the compound nucleus t1Na when the
incident projectile is: (a) a neutron, (b) a proton, or (c) an alpha particle?
2-2. A very important type of radioactive decay process in nuclear reactors is one in which
fission products decay by neutron emission since such processes strongly influence the
time behavior of the fission chain reaction. The slowest such decay process in most
reactors is one characterized by a decay constant of 0.0126 sec- 1• Assuming that such
a process controls the rate at which one can decrease the power level of a reactor,
calculate the titre necessary to decrease the reactor power level from 3800 MW
(thermal) to 10 MW (thermal).
2-3. Consider an initially pure sample of radioactive material whose successive decay
products are themselves radioactive with differing half-lives. Write the isotopic rate
equations characterizing the concentration of various isotopes in the sample if the
decay chain is of the form 1-2-3_4_ ... N. Solve this set of equations in a stepwise
fashion for the isotopic concentrations N 1(t},Ni(t}, ... In particular, determine the
long-time composition of the sample if the half-life of one of the isotopes is very much
longer than those characterizing other isotopes in the chain.
2-4. A fission product of very considerable importance in thermal reactor operation is
135Xe, which has an enormous thermal absorption cross section of 2 x 106 b. This
nuclide can be produced either directly as a fission product or by beta decay of 1351, as
indicated by the radioactive chains below:
p
p 135Cs
_,,.?f
1351 ~ 13sxe
~ 1J6Xe
i
fission
i
fission
+n
Write the rate equations describing the concentration of 1351 and 135Xe in a nuclear
reactor. Then assuming a constant production rate of these isotopes from fission and
transmutation rate by neutron capture, determine the steady-state or saturated con-
centration of 135Xe.
2-5. The measured line width of the gamma-ray resonance of 57 Fe is 3.4X 10-s eV.
Determine the lifetime of this excited state.
2-6. Boron is a common material used to shield against thermal neutrons. Estimate the
thickness of boron required to attenuate an incident thermal neutron beam to 0.1 % of
its intensity. (Use the thermal cross section data in Appendix A.)
2-7. Suppose we consider a beam of neutrons incident upon a thin target with an intensity
of 1012 neutrons/cm2-sec. Suppose further that the total cross section for the nuclei in
this target is 4 b. Using this information, determine how long one would have to wait,
on the average, for a given nucleus in the target to suffer a neutron interaction.
2-8. A free neutron is unstable against beta decay with a half-life of 11.1 :r: Determine the
relative probability that a neutron will undergo beta-decay before being absorbed in
an infinite medium. Estimate this probability for a thermal neutron in H10.
2-9. Determine the number of scattering collisions a thermal neutron will experience on the
average before being absorbed in H 20, D 20, 238 U, and cadmium, respectively.
2-10. How many mean free paths thick must a shield be designed in order to attenuate an
incident neutron beam by a factor of 1000?
2-11. Using the data from BNL-325, compute the mean free paths of neutrons with the
following energies in the specified materials: (a) 14MeV neutrons in air, water, and
uranium (characteristic of thermonuclear fusion neutrons), (b) 1MeV neutrons in air,
water? and uranium (fast breeder reactor neutrons), and (c) 0.05 eV neutrons in air,
water, and uranium (thermal reactor neutrons).
2-12. Determine the kinetic energy at which the wavelength of a neutron is comparable to:
(a) the diameter of a nucleus, (b) an atomic diameter, (c) the interatomic spacing in
graphite, and (d) the diameter of a nuclear reactor core. (Only rough estimates are
required.)
2-13. Suppose that the total cross section of rhodium has been measured and the following
values have been obtained for the resonance parameters of a well-isolated resonance at
E 0 = 1.26 eV: 0 0 =5000 b, f=0.156 eV, and o5 =5.5 b. Plot the value of the total cross
section for values of the energy between 0.2 and 40 eV. Calculate the thermal
absorption cross section and compare this with the measured value of 156 b. (Assume
that resonance scattering can be neglected.)
2-14. At higher energies, the differential elastic scattering cross section in the CM system
exhibits anisotropy (so-called "p-wave" scattering) of the form
Plot the scattering probability P(Er-~E1) against final energies E, for this more general
cross section behavior for the three cases a > 0, a= 0, and a< 0. Give a physical
interpretation of your sketches.
2-15. Using the Maxwell-Boltzmann distribution M(V, T), calculate the most probable
energy of the nuclei characterized by such a distribution. Also calculate the average
thermal energy of these nuclei.
2-16. The partial widths of the first resonance in 236 U at 5.49 eV are rY = .029 eV and
r n = .0018 eV. Plot the Doppler-broadened capture cross section at the temperature of
0°K, 20°C, and I000°C. [Use the tabulated 1/;(f,x) function.]
2-17. Show that the total area under a Doppler-broadened resonance is essentially inde-
pendent of temperature.
2-18. Using the differential scattering cross section characterizing a proton gas at tempera-
ture T, compute the corresponding macroscopic scattering cross section 'l:..(E). In
particular, determine the behavior of this cross section for low energies E.
2-19. A neutron is absorbed in a 235 U nucleus at t=O. Describe a probable life history of the
resulting 236 U and its successors on the assumption that it undergoes fission. Give
order of magnitude estimates of characteristic times at which various events occur.
Describe the various particles injected into the system as a result of this fission.
2-20. Determine the fission-rate density necessary to produce a therm.al power density of
400 kW /liter (typical of a fast breeder reactor core). Assume that the principal fissile
isotope is 239Pu.
2-21. An indium foil is counted at 5:00 p.m. Tuesday and found to yield 346,573 CPM in a
counter with a 50% efficiency for the 54-min In-I 16m activity. What is the probability
that none of these radioactive In-116m nuclei will remain in the foil at 2:00 p.m.
Thursday, the same week? (Note: ln(l - x)- - x,x« I) [Victims working this problem
can thank Dr. Ronald Fleming.]
2-22. Compute and plot the parameter 11 for uranium enriched in 235 U as a function of its
enrichment (atom percent 235 U) at thermal neutron energies.
3
Fission Chain
Reactions and
Nuclear Reactors
--an Introduction
In order to sustain a stable fission chain reaction and thereby achieve a constant
production rate of fission energy, one must design a nuclear reactor in such a way
that the rates of neutron absorption and leakage are balanced by the rate of fission
neutron production. In this chapter we will develop a very simple model of nuclear
reactor behavior based on such a neutron balance principle in order to introduce a
number of the concepts involved in studying fission chain reactions. General
aspects of the design and operation of nuclear fission reactors can then be
understood in terms of this model, and the principal components of such systems
and their functions can be discussed (although, of course, the analysis and design
of these components will require more elaborate models developed in later chapters
of this text). We will be able to introduce and compare the various major types of
nuclear reactors being utilized for electrical power generation throughout the world
today. Finally the simple discussion in this chapter will allow us to outline the
principal design functions of the nuclear engineer in order to lay an appropriate
foundation for our further development of the more sophisticated methods re-
quired in modern nuclear reactor analysis.
I. THE MULTIPLICATION FACTOR AND NUCLEAR

CRITICALITY '
A. The Multiplication Factor

In Chapter 2 we indicated that an essential idea involved in tapping the
energy released in nuclear fission was to use the fission neutrons from one fission
reaction to induce yet another reaction. In this way one could propagate a chain of
74
FISSION CHAIN REACTIONS AND NUCLEAR REACTORS-AN INTRODUCTION / 75
such reactions by using the neutron as a chain carrier. It should be apparent that if
we wish to maintain a stable or steady-state chain reaction, that is, one that does
not grow or decay away with time, we must arrange things so that precisely one
neutron from each fission will induce another fission event. The remaining fission
neutrons will then either be absorbed in capture reactions or will leak out from the
system. We must design the nuclear reactor to achieve this very delicate balance
between fission reactions and neutron capture and leakage, as we indicate schema-
tically in Figure 3-1.
We can express this requirement in mathematical form. Since the neutrons play
the central role in maintaining the fission chain reaction, let us focus our attention
on them for a moment. A given neutron will be "'born" in a fission event and will
then usually scatter about the reactor until it meets its eventual "death" in either an
absorption reaction or by leaking out of the reactor. Certain numbers of these
neutrons will be absorbed by fissile or fissionable nuclei and induce further fission,
thereby leading to the birth of new fission neutrons, that is, to a new "generation"
of fission neutrons. Suppose that we could somehow measure the number of
neutrons in two successive fission neutron generations. We would then define the
ratio of these numbers as the multiplication factor k characterizing the chain
reaction
. 1. .
k = Mu1tip Number of neutrons in one generation
1catlon factor= . . . .
Number of neutrons m preceding generation
Actually since the number of fission neutrons in any generation is proportional to

the number of fission events spawning that generation (recall that each fission
reaction releases, on the average, v fission neutrons), we could have just as easily
defined k using the number of fission events in each generation. However since we
are primarily concerned with monitoring the number of neutrons present in the
reactor in order to study the chain reaction, we will find it more convenient to use
the definition above.
Now notice that if k- 1, the number of neutrons in any two consecutive fission
generations will be the same, and hence the chain reaction will be time-
independent. We refer to a system characterized by k = 1 as being critical.
i
r
Fission-fragmen~
2-3 Fission4'
neutrons
•
nucleus ~ ~
23su :--1!1,t 0
@ ~. f--200 MeV of energy
- Incident
neutron
~~
@
--..___ Radiative
~pture
!
• ~e-y
! kage from
system
FIGURE 3.1. A simple schematic of a fission chain reaction.

Obviously if we have been fortunate enough to have chosen just that reactor
configuration and composition so that the reactor is critical with k = I, then the
number of neutrons in the reactor will always remain the same.
By a similar argument, we can conclude that if k < 1, the number of neutrons
decreases from generation to generation, and hence the chain reaction dies out. We
then refer to the system as being subcritical. Finally, if k > 1, then the chain
reaction grows without bound as the number of neutrons in each successive
generation is larger. Such a system is said to be supercritical.
In summary:
k l supercritical.
Hence the primary objective of the nuclear engineer is to design the nuclear reactor
so that it is critical. One possible approach would be to choose a particular reactor
material composition and configuration, then calculate k for this choice, and if k is
not unity (and of course it usually won't be on the first try), readjust the reactor
design until the criticality condition, k = 1, is achieved.
Actually life is a bit more complicated than this. Some method has to be
provided by which the neutron population can be built up to appreciable levels in
the core (about 109 neutrons/cm3) to yield the required power generation. In
principle this could be done by merely inserting a source of neutrons into a critical
assembly. Then any source neutrons appearing in the reactor would tend to induce
fission reactions, thereby producing fission neutrons, which would have their
progeny maintained by the chain reaction. However most neutron sources are
sufficiently weak that it would take a very long time to build up an appreciable
neutron population in a reactor using this method. Instead one can simply make k
temporarily greater than unity so that the reactor is supercritical, say by withdraw-
ing some absorbing material to alter the balance between fission and absorption.
The neutron population in the reactor will then grow. Once the desired neutron
population has been reached, the reactor can be returned to critical, for example,
by reinserting the absorbing material. A very similar procedure can be used to
lower the neutron population in the reactor. The reactor is taken subcritical until
the desired neutron population is reached and then restored to critical once again.
Altering the multiplication factor k characterizing a reactor in this way is known as
nuclear reactor control. It is a very important aspect of nuclear reactor analysis.
It should now be apparent that the multiplication factor k plays an extremely
important role in determining nuclear reactor behavior. The calculation of the
multiplication factor k characterizing a given reactor configuration and composi-
tion is one of the primary objectives of nuclear reactor analysis, and much of our
attention in this text will be devoted to developing various procedures for perform-
ing this calculation.
The definition of the multiplication constant k in terms of successive fission
neutron generations is sometimes known as the '"life-cycle" point of view 1 because
of its similarity to biological population growth. This definition is a bit awkward,
however, since it is usually rather difficult to determine the neutron generation
time. For example, some neutrons may induce fission immediately after their birth
in a fission reaction. Others may first slow down to thermal energies before
inducing fission. Some neutrons may not induce fission reactions at all, but will
instead be absorbed in nonproductive capture or leak out of the system.
A somewhat more practical definition of the multiplication factor k can be given
in terms of a neutron balance relation by defining
k = Rate of neutron production in reactor = P ( t) (3-1)

Rate of neutron loss (absorption plus capture) in reactor L(t)
Here we have explicitly noted that the production and loss rates may change with
time (e.g., due to fuel consumption).
We will find the "neutron balance" definition of multiplication a somewhat more
useful concept, since it is consistent with the approach that we will use to develop
more elaborate models of nuclear reactor behavior in later chapters. In particular,
we can then define th~ neutron lifetime, /, in an unambiguous fashion as
(3-2)
where N ( t) is the total neutron population in the reactor at a time t. This latter
approach is also particularly convenient for studying the time behavior of the
neutron population in a reactor.
B. Simple Kinetics of Chain Reactions

Imagine that we could somehow count the number of neutrons N ( t) in a
nuclear reactor at a time t. Then obviously the time rate of change of N (t) is given
by
dd'N = Production rate - Loss rate= P ( t)- L( t). (3-3)

t .
However if we use our definition of the multiplication factor k as given by the

neutron balance relation, we can write
dt L( t)
l
-dN = [ -P(t) -1 L(t)=(k-l)L(t). (3-4)
To proceed further we can use our definition of the neutron lifetime I to write
dN (k-1)
dt= I N(t). (3-5)
If we assume that both k and / are time-independent (of course they will not be in
general), then we can solve this simple ordinary differential equation for the
neutron population at any time t, assuming that there are initially N 0 neutrons in
the reactor at time t = 0, to find
(3-6)
In particular, note that this very simple model of nuclear reactor kinetics agrees with
our earlier definition of reactor criticality in terms of k {see Figure 3-2). Yet this
model also tells us that the growth or decay of the neutron population in a reactor
obeys an exponential growth law. Such exponential growth is quite commonly
found in the study of population dynamics. Indeed the study of the '"neutron"
population in a reactor core is mathematically rather similar to the study of
biological populations, and hence the terminology of the latter field is frequently
adopted in reactor physics (e.g., generation, birth, life, death, virgin, daughter).
We will later find that the power level of a nuclear reactor is essentially
proportional to its neutron population. Hence we can also regard the time behavior
of the reactor power level as being exponential with a time constant or reactor
period T given by
I
T= k-1. (3-7)
In particular it should be noted that as the multiplication factor k approaches

unity, the reactor period T approaches infinity which corresponds to a time-
independent neutron population or reactor power level.
However suppose that k is not equal to unity. Then how rapidly might we expect
the power level of the reactor to change? Suppose, for the sake of illustration, we
increased k to make the reactor ever so slightly supercritical by an amount of
k = 1.001. Since the neutron lifetime in a typical power reactor is about 10- 4 sec,
we find this corresponds to a reactor period of T = 0.1 sec. Hence in one second the
power level of the reactor will increase by a factor e 10 =22,000. Thus it appears that
the reactor will respond very rapidly to changes in the multiplication factor. In fact,
if a power reactor did indeed respond this rapidly, then it would be difficult to
N(t)
N(O)
t
FIGURE 3-2. Time behavior of the number of neutrons in a reactor.
control the reactor power level, for a 0.1 % change in the multiplication factor is
rather common. Fortunately we have omitted something from this simple model
which tends to greatly increase the neutron lifetime / and hence T, thereby slowing
down the reactor time response. This is the effect of delayed neutrons on the chain
reaction. However this is a tale for another time, so we will leave our study of
reactor kinetics with the promise of returning later to patch up this model in order
to provide a more optimistic picture of nuclear reactor time behavior.
C. A Formal Calculation of k: The Four-Factor Formula

Let us now turn our attention to the calculation of the multiplication factor
for, say, a pile of uranium that one wishes to make into a nuclear reactor. We
probably should add some coolant to remove fission heat and perhaps some
structural material to hold the core together. However we will assume that we can
treat these materials as intimately and homogeneously mixed so that the composi-
tion of the reactor is uniform.
Now to calculate k we must determine the possible fate of neutrons in a given
fission generation. Fortunately this is rather easy to do since there are only two
possible alternative destinies available to the neutron. First it might leak out of the
reactor and be lost to the chain reaction. If it does not leak out, then it must
eventually be absorbed.t This absorption may correspond to a nonproductive
capture event in either the fuel or other materials, or the absorption may induce a
fission reaction, in which case a new fission neutron generation is produced. We
can represent these destinies schematically, as shown below:
Leak out of system Absorbed in junk

~
L
/ Radiative capture
Fission or
neutron p:--..._ . or
~~ m
Absorbed in system
P., At.o,bed ia fuel
P1 Fission
- - - - - - - - - - r, new fission neutrons ◄E---------.....J
To make this more formal, suppose we define the probabilities for each of these
possible events as follows:
p = Probability that neutron will not leak out of system before

NL absorption
p = Conditional probability that if neutron is absorbed, it
AF- will be absorbed in the fuel
p = Conditional probability that if neutron is absorbed in
r- fuel, it will induce a fission reaction.
tor course, yet a third alternative would be a decay of the neutron into a proton, electron, and
neutrino, but since the half-life for decay of a free neutron is 11.7 minutes, and the typical
neutron lifetime I in the reactor is less than 10- 3 sec, we can safely ignore this alternative.
These latter two conditional probabilities are easily calculated. The conditional
probability for absorption in the fuel P AF can be expressed simply as the ratio of
the macroscopic absorption cross sections for the fuel ~r,and for the fuel plus the
rest of the material in the core ~a· (We will usually indicate with a superscript the
material to which we are referring. The absence of the superscript will imply that
the macroscopic cross section is the total for all of the materials in the system.)
Thus we can write
(3-8)
It should be kept in mind that this expression has been introduced only fqr the
situation in which the reactor has a uniform composition. Unfortunately for the
reactor analyst all modern reactors have nonuniform compositions varying from
point to point (e.g., due to fuel elements, coolant channels, support structure). In
this more general case one can still use Eq. (3-8) if the macroscopic cross sections
Ia are regarded as spatial averages over the reactor. It should also be noted that we
have not yet specified the neutron energy at which these cross sections are to be
evaluated. Again, we will later find that the cross sections appearing in Eq. (3-8)
must be appropriately averaged over energy, just as they are over space.
It is customary in reactor terminology to refer to this probability as the thermal
utilization of the reactor and denote it by P AF= J. This term arose in the early
analysis of thermal reactors in which essentially all fissions in the fuel were induced
by thermal neutrons. In this case the cross sections in f would be evaluated at
thermal neutron energies and would represent the effectiveness of the fuel in
competing with other materials in the reactor for the absorption of thermal
neutrons, that is, the effectiveness with which the reactor utilized the thermal
neutrons in the fuel. The expression in Eq. (3-8) actually applies to any type of
reactor. However we will fall in line with convention and refer to it as the thermal
utilization and denote it by "f."
The conditional probability for inducing a fission reaction in the fuel can also be
expressed in terms of cross sections. In this case we simply take the ratio of the
fission cross section to that of the absorption cross section (due to both fission and
radiative capture) in the fuel material:
(3-9)
We are now ready to utilize these probabilities to determine the multiplication

factor k. The general scheme is to play a game of "follow the neutron." Suppose we
start with N 1 neutrons present in the reactor in a given fission generation. Then
with the help of the above probabilities and our diagram, we can compute the
number of neutrons in the next generation as:
(3-10)
or
(3-11)
where we have recalled that 'IJ = v( or Io~) is the number of fission neutrons
produced per absorption in the fuel. We can now use our definition of the
multiplication factor k as being the ratio of the number of neutrons in two
successive fission generations to write
(3-12)
The nonleakage probability PNL appearing in Eq. (3-12) and characterizing neutron
leakage from the core is much more difficult to compute. It will require more
elaborate mathematics (and in a realistic calculation, the use of a digital computer},
and hence we will defer a discussion of it until later.
As a momentary detour, however, suppose our reactor were of infinite extent.
Then since no neutrons could leak out, we immediately conclude that we must set
the nonleakage probability PNL = 1. The corresponding multiplication factor is then
known as the infinite medium multiplication factor and denoted by
(3-13)
Now of course no reactor is of infinite size. Nevertheless kt;t;) is a useful parameter

in reactor analysis since it essentially characterizes the multiplication properties of
the material in the reactor as distinct from the geometry of the reactor core. Of
course since PNL < 1 more generally for a finite reactor from which some neutron
leakage can occur, we must have k 00 > 1 in order to have any chance of achieving a
critical chain reaction.
There are a couple of important modifications that must be introduced into this
simple development in order to understand how the present generation of so-called
"thermal" reactors works. We must account for the fact that the neutrons in a
nuclear reactor have a distribution of energies. As we saw in Chapter 2, the fission
neutrons are born at very high energies in the MeV range. However the fission
cross section is largest at very low energies-indeed, at those energies correspond-
ing to neutrons in thermal equilibrium with the reactor core at a temperature T,
e.g., for T=300°C, E= kT=0.05 eV. Hence it is obviously to our advantage to try
to slow down, or in the language of reactor physics, "moderate," the fast fission
neutrons to take advantage of the fact that slow neutrons are more likely to induce
fission reactions. This can be accomplished rather easily, simply by letting the fast
neutrons collide with light nuclei, thereby losing some of their kinetic energy in
elastic scattering collisions. The lighter the nucleus involved, the more kinetic
energy per collision will be lost on the average by the neutron and hence the more
effective the slowing down or moderation. In fact the best nucleus to use is
hydrogen, which is fortunately quite commonly available in the form H 20. Hence
if we just let the fast neutrons rattle around in water for a bit, they will quickly slow
down to the desired thermal energy. In this sense, we refer to water as a neutron
moderator. Numerous other materials can be used as moderators in nuclear
reactors, and we will discuss these in greater detail later.
The presence of such neutron moderation in a reactor suggests several modifica-
tions to our earlier calculation of the multiplication factor k. Suppose we first
modify our diagram of the various possible neutron destinies to take into account
neutron energy as shown in Figure 3-3. Now since most fissions will be induced by
thermal neutrons, we will regard f and r, as being evaluated at thermal neutron
r---~--------------
1 V
MeV fission
neutrons
I
I
+
I
Slowing down
I via scattering
I collisions with
I light nuclei
I
l
I
! (Resonance
1I absorption
or
leakage)
I
I
I
I
I
I
A
I
I
I
I
I
I Sub-eV thermal
neutrons
I (leakage or --
I parasitic capture)
L-----<----------''----
fission
FIGURE 3-3. Processes characterizing a neutron generation in a thermal reactor.
energies. For example, f would now refer to the ratio of thermal neutron absorp-
tions in the fuel to total thermal neutron absorptions and thereby become more
deserving of its designation as the "thermal" utilization. Similarly, Y/ is now
identified as the average number of fission neutrons produced per absorption of a
thermal neutron in the fuel.
Then to account for processes that occur while the neutron is slowing down to
thermal energies, we will introduce two new quantities. We first define a factor that
takes account of the fact that, although most fissions will be induced in fissile
material by thermal neutrons, some fissions will be induced in both fissile and
fissionable material by fast neutrons. Hence we will scale up our earlier expression
for k by a fast fission factor c
Total number of fission neutrons (from both fast and thermal fission)
E=---------------------------- (3-14)
Number of fission neutrons from thermal fissions
The fast fission factor € is usually quite close to unity in a thermal reactor with
typical values ranging between € = 1.03 and £ = 1. 15.
The second factor we will introduce will characterize the possibility that the
neutron might be absorbed while slowing down from fission to thermal energies.
Since most absorptions occurring during the slowing down process correspond to
resonance capture in heavy nuclei such as 238 U, we refer to this factor as the
resonance escape probability p:
Fraction of fission neutrons that manage to

p = slow down from fission to thermal energies (3-13)
without being absorbed.
Finally it is useful to modify our definition of the nonleakage probability to take

account of the fact that there will be two distinct phases of neutron leakage that
will require two rather different types of analysis in our later work. First the
neutron may leak out while slowing down. Indeed since the neutron mean free path
is relatively large for high energies, such fast neutron leakage may be quite
appreciable. A second leakage process may occur after the neutron has managed to
slow down to thermal energies. After slowing down, the neutron may continue to
scatter and eventually leak out before it has had an opportunity to be absorbed. To
take account of these two processes, we will break up our earlier nonleakage
probability as follows:
(3-16)
where
= Probability that fast neutron will not leak out
P FNL-
(fast nonleakage)
= Probability that thermal neutron will not leak out

P TNL-
(thermal nonleakage ).
If we now insert these new definitions into our earlier expressions (3-12) and
(3-13), we find that the infinite medium multiplication factor becomes
(3-17)
This is known as the four-factor formula. Moreover one now writes
(3-18)
which is known, surprisingly enough, as the six-factor formula.
EXAMPLE: To more vividly illustrate these ideas, we can list the values of each
of the factors in the six-factor formula for a typical thermal reactor: 11 = l.65,
f=0.71, £= 1.02,p=0.87, PFNL =0.97, PTNL =0.99::::>k 00 = 1.04::::>k= 1.00.
Hence provided we can calculate each of these factors, our criticality condition
k =1 can then be easily checked (in the above example we fudged up the
parameters a bit to yield a critical system). Of course, the calculation of these
factors is quite difficu~t in general. Indeed one cannot really separate the various
84 / INTRODUCTORY CONCEPTS QF NUCLEAR POWER REACTOR ANALYSIS
conditional probabilities as was done in these formulas. Instead alternative schemes

based on iterative numerical methods must be used in practice to arrive at a
criticality condition.
Nevertheless the four-factor and six-factor formulas are quite useful because
they provide insight into the various mechanisms involved in nuclear fission chain
reactions and on rare occasions may actually be of use in making crude estimates
in nuclear design. They are also useful in illustrating the trends of parameter
variation in a core design.
For example, although r, and E are essentially fixed once the fuel has been
chosen, the thermal utilization f and resonance escape probability p can be varied
considerably by changing the ratio of fuel density to moderator density. All of
these parameters can be varied by using a heterogeneous lattice of fuel elements
surrounded by moderator rather than a uniform, homogeneous mixture of fuel and
moderator.
One can also vary the nonleakage probabilities by simply making the reactor
core larger, or surrounding the reactor by a material with large scattering cross
section so that some of the neutrons leaking out will be scattered back into the
reactor. Actually when leakage is changed, there will be some change in the
parameters in the four-factor formula as well since these are actually averages over
the various neutron energies. in the reactor, and this distribution of energies will
vary with the amount of leakage. Such considerations have given rise to a
somewhat different notation for the multiplication factor characterizing a finite
system which is occasionally referred to as the effective multiplication factor and
denoted by kerr
(3-19)
There are other prescriptions for defining the multiplication factor. In particular we
will introduce one of these schemes later when we consider the analytical treatment
of the neutron energy dependence in more detail. However for now we will
continue to regard the multiplication factor as the ratio between either the number
of neutrons in two successive fission generations or the neutron production and
loss rates in the reactor.
We can use the six-factor formula for the multiplication factor to gain a bit more
insight into the goals of reactor design and operation. There are several ways to
adjust k in the initial design of the reactor. One could first regard the size of the
reactor as the design variable. Since the ratio of surface area to volume decreases as
the reactor geometry is enlarged, one can control the relative importance of the
leakage factors by adjusting the reactor size. For a given core composition (with krxi
greater than 1, of course) there will be a certain critical size at which k-1. An
alternative way to achieve the same reduction in leakage is to surround the reactor
with a scattering material that acts as a neutron reflector. Most thermal reactor
cores are so large that leakage represents a rather small loss mechanism (typically
about 3% of the neutrons leak out from the core in large thermal reactors).
Usually the core size and geometry for a power reactor are dictated by thermal
considerations, for instance, the size of the core necessary to produce a given power
output while being provided with sufficient cooling so that the temperature of the
reactor materials will not become excessively high. The primary design variable at
the disposal of the nuclear engineer is the core composition. In particular he can
vary the composition (enrichment) and shape of the fuel, the ratio of fuel to
FISSION·CHAIN REACTIONS AND NUCLEAR REACTORS-AN INTRODUCTION / 85
moderator density, the type of moderator, coolant, and structural materials used, or
the manner in which reactor multiplication is controlled. One would refer to the
amount of fuel required to achieve a critical chain reaction as the critical mass of
fuel.
In reality, however, a nuclear reactor is always loaded with much more fuel than
is required merely to achieve k = 1. For example the LWR is typically loaded with
sufficent fuel to achieve a multiplication of about k = 1.25. This extra multiplication
is required for several reasons. First if the reactor is to operate at power for a
period of time, one must provide enough excess fuel to compensate for those fuel
nuclei destroyed in fission reactions during the power production. Since most
contemporary reactors are run roughly one year between refueling, a sizable
amount of excess fuel is needed to compensate for fuel burnup. A second
motivation arises from the fact that the multiplication of a reactor tends to
decrease as the reactor power level and temperature increase from ambient levels to
operating levels. Additional multiplication is needed to compensate for this effect.
Finally one must include enough extra multiplication to allow for reactor power
level changes. For example, we have seen that if we wish to increase the reactor
power level, we must temporarily adjust k to a value slightly greater than I so that
the reactor is supercritical. The reactor can then be returned to critical when the
desired power level has been reached.
Of course when this excess multiplication is not being used, some mechanism has
to be provided to cancel it out to achieve reactor criticality. This is the function of
reactor control mechanisms. Such control is usually achieved by introducing into
the reactor core materials characterized by large absorption cross sections. They
will then tend to eat up the excess neutrons produced in the chain reaction. In
terms of our six-factor formula such absorbing materials lower the value of the
thermal utilization J, since they compete with the fuel for neutron absorption. A
variety of types of reactor control are used in power reactors. For example, the
neutron absorber might be fabricated into rods which can then be inserted into or
withdrawn from the reactor at will to vary multiplication. Sometimes the absorber
is fabricated directly into the fuel itself. Or it may be dissolveu in the reactor
coolant. When such control absorbers are used to hold down the excess multiplica-
tion introduced to compensate for fuel bumup, one refers to them as shim control.
They may also be used to force the reactor subcritical in the case of an emergency;
then they are known as scram control. Finally they may just be used to regulate the
power level of the reactor; then they are referred to as maneuvering control
elements.
The ease with which such control elements can control the fission chain reaction
will depend on how rapidly the reactor responds to variations in multiplication.
Since fuel burnup occurs over very long periods of time (typically weeks or
months), a rapid response of shim control is not required-which is fortunate,
because rather large amounts of multiplication must be manipulated (typically
changes of 10-20% in k). The normal power variations in the reactor are due to
much smaller changes in multiplication ( <0.1%) and are characterized by essen-
tially the reactor period T which in turn is proportional to the neutron lifetime /.
However we saw earlier that the lifetime of prompt fission neutrons was quite
short, typically about 10- 4 sec. The effective neutron lifetime is greatly increased
by the presence of delayed neutrons, however. We recall that about 0.7% of the
neutrons produced in fission are delayed anywhere from 0.6 to 80 sec since they
arise from fission product radioactive decay. Hence the effective neutron lifetime is
actually the average of the prompt neutron lifetime and the average decay time of
these delayed neutrons, properly weighted, of course, by their relative yield frac-
tions. When this is taken into account, one finds that the effective neutron lifetime
is almost two orders of magnitude longer, terr,_, 10- 1 seconds. Hence a multiplica-
tion of 0.1 % would now correspond to a reactor period of T- IO seconds, well
within the control capability of a reactor control system.
D. Conversion and Breeding
If we recall our earlier expression for the multiplication factor k in Eq. (3-12),
it is evident that since the thermal utilization f and the nonleakage probability PNL
are both less than I, we require r, to be substantially greater than 1 if a critical
fission chain reaction is to be possible. Fortunately as we can see from Figure 2-25,
this condition is not only satisfied, but in fact for many energies one finds that
11 > 2. Hence we in fact appear to have an extra neutron. This "bonus" neutron can
be put to good use if we recall that certain fertile isotopes can be transmuted into
fissile material via neutron capture. In particular, 238U can be transmuted into
239 Pu, while 232Tb can be transmuted into 233 U. Hence if we load the core of a
reactor with such fertile material, we can use the extra neutron to produce a new
fissile fuel material. This process is frequently referred to as conversion, and nuclear
reactors whose principal job is to produce 239Pu or 233 U are known as converter
reactors.
Actually all modem power reactors are converter reactors in a sense, although
this is not their primary function, since they contain substantial amounts of 238 U
which will be transmuted into 239Pu via neutron capture during normal operation.
For example, a LWR will contain a fuel mixture of roughly 3% 235 U and 97% 238 U
in a freshly loaded core. After a standard operating cycle (usually one year), this
core will contain roughly 1% 235 U and 1% 239 Pu which can then be separated out of
the spent fuel and refabricated into fresh fuel elements for reloading (so-called
"plutonium recycling").
These considerations suggest that it might in fact be possible to fuel a reactor
with 239 Pu and 238 U and then produce directly the fuel (239Pu) needed for future
operation. Indeed it might even be possible to produce more 239Pu than is
burned-that is, to "breed" new fuel. This is the essential idea behind the concept
of a breeder reactor.
To discuss this concept in more detail, it is useful to define the conversion ratio
Average rate of fissile atom production

CR- Average rate off'ISSI'le atom consumpt10n
. .
This quantity is also referred to as the breeding ratio (BR) if it is greater than one.
If we have conversion then, consuming N atoms of fuel during reactor operation
will yield CR·N atoms of the new fissile isotopes. For example, most modern
LWRs are characterized by a conversion ratio of CR a;; 0.6. By way of contrast,
HTGRs are characterized by somewhat higher conversion ratios CRa;;Q,8 and
hence are sometimes referred to as advanced converter reactors.
For breeding to occur we require that the conversion ratio be greater than unity,
CR - BR> 1. Of course for this to happen we must have 11 > 2 since slightly more
than one fission neutron is needed to maintain the chain reaction (some neutrons
FISSION CHAIN REACTIONS AND NUCLEAR REACTORS-AN INTRODUCTION / ff"/
will leak out or be absorbed in parasitic capture) while one neutron will be needed
to replace the consumed fissile nucleus by converting a fertile into a fissile nucleus.
If we return to Figure 2-25 we can see that the only attractive breeding cycle for
low-energy (i.e., thermal) neutrons would involve 233 U, that is, the 232Th/ 233 U
process. To breed using 238 U / 239Pu requires that we use fast neutrons with energies
greater than 100 keV. And of course this is the motivation behind the development
of the fast breeder reactor.
At this point, it is useful to digress a bit and discuss the average energy of the
neutrons sustaining the chain reaction in various types of nuclear reactors. As we
have seen, the energies of neutrons in a reactor span an enormous range, from 10
MeV (usually the maximum energy of fission neutrons) down to as low as 10- 3 eV
after having suffered a number of scattering collisions with nuclei and slowing
down. Furthermore the neutron cross sections depend sensitively on the neutron
energy. As the examples in Chapter 2 indicated, the general trend is for cross
sections to decrease with increasing energies. This feature is particularly true of
absorption cross sections such as capture or fission.
The fact that the fission cross section or is largest at low energies implies that it is
easiest to maintain a fission chain reaction using slow neutrons. Hence early
nuclear reactors used low mass number materials such as water or graphite to slow
down or moderate the fast fission neutrons. Such moderating materials slow the
neutrons down to energies comparable to the thermal energies of the nuclei in the
reactor core. Reactors characterized by an average neutron energy comparable to
such thermal energies are referred to as thermal reactors. Such reactors require the
minimum amount of fissile material for fueling and are the simplest reactor types
to build and operate. Most nuclear power plants in this country and abroad utilize
thermal reactors.
However we have also seen that there is a very definite advantage in keeping the
neutron energy high, since the number of neutrons emitted per neutron absorbed in
the fuel 11 is largest for fast neutrons. Hence one can use the "extra" neutrons
available in a fission chain reaction maintained by fast neutrons to convert or
breed new fuel. However since or is smaller, one also needs much more fuel to
sustain the chain reaction. Furthermore to keep the neutron energy high, one wants
to utilize only high mass-number materials in the core to keep neutron slowing
down to a minimum. Such reactors characterized by average neutron energies
above 100 keV are known as fast reactors. It is felt by many that fast reactors will
eventually replace the current generation of thermal power reactors because of
their ability to breed fuel.
To make some of these ideas a bit more precise, we have compared the
important nuclear parameters v, T/, and or for typical nuclear fuels at energies
characterizing both thermal and fast reactors. (To be more precise, these quantities
have been calculated by averaging the energy-dependent nuciear parameters v(E),
11(£), and o,(E) over the neutron energy distributions found in typical LWRs and
LMFBRs.) One should first note that the fission cross sections in fast reactors are
some two orders of magnitude lower than those in thermal reactors. Hence even
though fast reactors exhibit considerably higher conversion ratios (typically, CR
= BR,_, 1.2-1.5) due to a larger value of T/, their fissile inventory requirements may
run as much as 30-40 times those required by thermal reactors just to maintain a
critical chain reaction with fast neutrons. This table also indicates that while 235 U
will yield a slightly higher conversion ratio than 239Pu in thermal reactors, the use
of 239Pu does exhibit a sizable advantage in fast reactors since the capture-to-fission
ratio a 49 falls off quite markedly for large neutron energies. In Table 3-1 we have
only indicated the nuclear fission properties of 238 U in fast reactors, since this
isotope is fissionable and hence contributes only a modest fraction of the fissions
occurring in thermal reactors (--2-5%) in contrast to its rather large contribution in
fast reactors (--20%).
TABLE 3-1 Energy-Averaged Fission Parameters Chancterlziog

Thermal and Fast Reactors
Thermal Reactor Fast Reactor

(LWR) (LMFBR)
23SU 239Pu 23su 239Pu 23su
, 2.4 2.9 2.6 3.1 2.6

fj 2.0 1.9 2.1 2.6 0.3
IJr(b) 280 790 1.9 1.8 0.06
We have mentioned only a few of the considerations involved in comparing the

nuclear behavior of thermal versus fast reactors. As we develop more sophisticated
methods of reactor analysis, we will return frequently to contrast the application of
these methods for thermal and fast systems.
II. AN INTRODUCTION TO NUCLEAR POWER REACTORS
A. Nuclear Power Plants

The schematic illustration4 of a typical large nuclear power plant appears in
Figure 3-4. It is apparent from this illustration that the nuclear reactor itself is only
one of a great many components in such a plant. Actually aside from the nuclear
reactor and its associated coolant system, such power plants are remarkably similar
to large fossil-fuel fired plants. Only the source of the heat energy differs, that is,
nuclear fission versus chemical combustion. Hence most of the components of
large central-station power plants are common to both nuclear and fossil units.
A very crude diagram of the major components of an electrical power plant is
given in Figure 3-5. As we have sketched it in this diagram, the steam supply
system could be either a fossil-fuel fired boiler or a nuclear reactor and its
associated coolant loops.
All of the current large power plants operate on a steam cycle (a so-called
Rankine cycle) in which the heat generated by combustion or nuclear fission is
used to convert water into high-pressure, high-temperature steam. This steam is
then allowed to expand against the blades of a turbine. In this way the latent energy
of the steam is converted into the mechanical work of turning the turbine shaft.
This shaft is connected to a large electrical generator that converts the mechanical
turbine energy into electrical energy which can then be distributed to an electrical
power grid. The low-pressure steam leaving the turbine must then be recondensed
in a steam condensor into water so that it can be pumped back to the steam supply
system to complete the cycle. The condensor requires large quantities of ambient
I
Steam generators~
N2N?N2W1Sl2NZfS AA7V'vl\,
~
"TT
II
!!
c3
Turbine,-..J.11 I 111 [&fk~~.Eiff~P~Lli.J
- I ll tl : :~.:F-.~ ~ 5 E :
-1:~~~#,:$.~Jj;.~\
·C: ·.; '.\.~/
·.:.
. ..;....:.;.:..:a=::;.2=·:::!::·. _~:\.:;:;... ~r.~.J{'}:...-•.... :.-, .". ~ :-/

,~-i~~ :J:=·:·...
JJ.• • .•·. o·•· .':-:'.,..,.-_J -'"""-"-''·' ·~:- ~·;,·::·~:;; i~~~

.. !-===!}-..Jf .... . . "'.,..._-__ _J
-.·.-_;'. .6.. i.'~-:/•.:_;. . ..

Containment building
Fuel-handling
building
Turbogenerator building
I FIGURE 3-4. A schematic diagram of a nuclear power plant.4

r,:::::.==============-== Steam
Lo,
Fossil-fired boiler
•t
-----.. ,--:=========---== Feedwater
, ____ J
BWR
Steam Primary
loop
(He)
Feedwater
PWR '-===========---== Feedwater

HTGR
Steam
t
Intermediate
Feedwater
loop (Na)
Primary loop (Na)
Steam_.
Multiple-stage turbine Generator
Condenser
Feedwater pump
Transformer
Feedwater
Feedwater heaters
Balance of plant
FIGUREJ-S. Comparison of different steam-supply systems.
90
FISSION CHAIN REACTIONS AND NUCLEAR REACTORS- AN INTRODUCTION / 91
temperature cooling water which is usually obtained from artificial cooling ponds or
cooling towers.
This is of course a very oversimplified description of the major components of a
power plant, but it does serve to illustrate that these components are quite similar
for both nuclear and fossil-fueled stations. Actually as far as the steam cycle itself
is concerned, the primary difference between the two types of plant is that the
fossil-fueled boiler supplies slightly higher temperature, higher pressure steam,
thereby reducing the design requirements on the turbine (although it should be
mentioned that more advanced reactor types such as the HTGR supply system
steam at conditions quite comparable to those of modem fossil-fueled units). Of
course there are numerous other differences in the various subsystems of the plants,
as well as in their operation. However the major features of the plants, aside from
their steam supply system,_ are quite similar.
B. The Nuclear Steam Supply System (NSSS)
The NSSS consists essentially of three major components: (a) a nuclear

reactor supplying the fission heat energy, (b) several primary coolant loops and
primary coolant pumps that circulate a coolant through the nuclear reactor to
extract the fission heat energy, and (c) heat exchangers or steam generators that use
the heated primary coolant to turn feedwater into steam. Several very simplified
diagrams of NSSS components are given in Figure 3-5.
A variety of possible coolants can be used in the primary loops of the NSSS.
Indeed nuclear reactor types are usually characterized by the type of coolant they
use, such as LWRs or gas-cooled reactors. There are also a variety of possible
NSSS configurations. For example, one may actually produce the steam in the
reactor core itself. Or one may use a single-phase primary coolant such as water or
helium to transfer the fission heat energy to a heat exchanger where it is used to
produce steam (see Figure 3-5). In the liquid metal-cooled NSSS, an intermediate
coolant loop must be utilized to isolate the steam generator from the very high
induced radioactivity of the primary coolant loop passing through the reactor.
The most·common coolant used in power reactors today is ordinary water, which
serves as both coolant and moderating material in the reactor. There are two major
types of LWR: pressurized water reactors (PWR) and boiling water reactors
(BWR). In a PWR the primary coolant is water maintained under very high
pressure (,_, 155 bar) to allow high coolant temperatures without steam formation
within the reactor. The heat transported out of the reactor core by the primary
coolant is then transferred to a secondary loop containing the '"working fluid" by a
steam generator. Such systems typically contain from two to four primary coolant
loops and associated steam generators.
In a BWR, the primary coolant water is maintained at a sufficiently low pressure
(,_, 70 bar) for appreciable boiling and steam formation to occur within the reactor
core itself. In this sense the reactor itself serves as the steam generator, thereby
eliminating the need for a secondary loop and heat exchanger. In both the PWR
and BWR, the nuclear reactor itself and the primary coolant are contained in a
large steel pressure vessel designed to accomodate the high coolant pressures and
temperatures. In a PWR, this pressure vessel must be fabricated with thick steel
walls to contain the very high primary coolant pressures. By way of contrast, the
BWR pressure vessel need not be so thick, but must be much larger to contain both
92 / INTRODUCTORY CONCEPI'S OF NUCLEAR POWER REACTOR ANALYSIS
the nuclear reactor and steam moisture-separating equipment.

A very closely related class of reactors utilizes D20 as moderator and either D20
or H 20 as primary coolant. The most common type of such heavy water reactors,
the CANDU-PHW, utilizes the pressure tube concept in which each coolant
channel in the reactor is designed to accommodate the primary system pressure
which is again kept high to prevent boiling. As with a PWR, the primary coolant
thermal energy is transferred via a steam generator to a secondary loop containing
light water as the working fluid. More recently, heavy water pressure tube reactors
have been designed (e.g., the CANDU-BLW or SGHWR) which produce H 20
steam directly in the core similar to a BWR. ·
Yet another type of reactor uses gas coolants. Although CO2 has been used as
the coolant in the MAGNOX class of natural uranium fueled, graphite moderated
reactors for many years in the United Kingdom, most present interest is directed at
HTGRs using helium under high pressure to cool a reactor fueled with enriched
uranium and moderated by graphite. The helium coolant is then passed through
steam generators to transfer the thermal energy on to a secondary loop. containing
water as a working fluid. It should also be mentioned that such HTGRs have the
potential of being combined with gas turbines (rather than steam turbines), thereby
eliminating the steam cycle altogether.
Gas coolants have also been proposed for use in fast breeder reactors (GCFRs).
Because of the very high power densities required by such reactors, extremely high
coolant flow rates would be required. Nevertheless the rather large breeding ratios
(BR- 1.5) achievable in the GCFR make it appear a very promising alternative to
other fast reactor designs.
The final class of nuclear reactors utilizes liquid metals such as sodium as a
primary coolant. Although sodium could be used in thermal reactors if alternative
moderation were provided, its primary advantages occur in fast breeder reactors
which require a primary coolant with low moderating properties and excellent
heat-transfer characteristics. The LMFBR NSSS actually uses two sodium loops. A
primary sodium loop is used to remove fission heat from the reactor; this coolant is
then passed through an intermediate heat exchanger in which it transfers its heat
energy to a secondary sodium loop, which, in turn, carries heat to a steam
generator. The intermediate loop isolates the steam generator from the radioactiv-
ity induced in the primary sodium coolant.
The NSSS of a modem nuclear power plant is completely contained within a
reactor containment structure designed to prevent the release of radioactivity to the
environment in the even of a gross failure of the reactor coolant system. This
nuclear island within the plant is usually fabricated out of steel-lined concrete and
contains not only the reactor itself (and its associated pressure vessel), but also the
primary coolant system including the primary pumps, steam generators, piping,
and auxiliary systems. A glance back at the schematic diagram of a nuclear power
plant in Figure 3-4 can quickly identify the containment building and the NSSS
contained within it.
C. The Nuclear Reactor Core

At the heart of the NSSS is the nuclear reactor. Far from being just a
relatively simple "pile" of fuel and moderator a la Fermi, a modem power reactor
is an enormously complicated system designed to operate under the most severe
conditions of temperature, pressure, and intense radiation. To introduce the general
FIGURE 3-6. A large pressurized water reactor.4
components of a typical power reactor, we will consider the specific example of a

modern large PWR as illustrated in Figure 3-6. 4 The reactor proper consists of a
core containing the fuel, coolant channels, structural components, control elements,
and instrumentation systems. In this particular example, the core is a cylindrically
shaped lattice roughly 350 cm in diameter by 370 cm in height consisting of long
fuel assemblies or bundles. These assemblies consist of a large number of long,
narrow fuel rods or fuel elements, which are metallic tubes containing the nuclear
fuel in the form of ceramic pellets. Individual fuel elements and assemblies for such
a PWR are shown in Figure 3-7.
In the reactor core one induces and maintains the nuclear fission reactions that
produce the desired heat. The core itself is enclosed in a much larger container, a
reactor pressure vessel, designed to withstand the enormous pressures of the coolant
(up to 155 bar) as well as to isolate the reactor core from the remainder of the
NSSS.
Most of our attention in this text will be directed at the nuclear analysis of the
reactor core itself, since this is the principal responsibility of the nuclear engineer.
However as we have mentioned earlier, we will try to develop this nuclear analysis
within the context of other considerations influencing the design of the nuclear
steam supply system.
Before we conclude this brief introduction to nuclear reactors, however, it is
useful to make a short list of the various components of nuclear reactor systems as
well as to introduce some of the standard nomenclature used in nuclear reactor
engineering.
Expansion spring
I nsu lator-wafer~_.-•J11
Fuel element Fuel assembly FIGURE 3-7. Fuel element; fuel assembly.
(I) Fuel: Any fissionable material. This can be either fissile material such as
233 U, 235 U, 239Pu, or 241 Pu or fissionable material such as 232Th, 238 U, or
240 Pu. Most modern power reactors utilize this fuel in a ceramic form-
either as an oxide such as UO 2, a carbide such as UC, or a nitride, UN.

(2) Fuel element: The smallest sealed unit of fuel. In an LWR or LMFBR
the fuel element is a metal tube containing ceramic pellets of fuel (such
as UOJ. (See Figure 3-7.) In an HTGR the fuel element can be regarded
as either a tiny (300 µ.m diameter) particle of uranium carbide coated
with pyrolytic graphite layers, or as a cylindrical fuel pin composed of
these fuel particles bound together with a graphite binder.
(3) Fuel assembly or bundle: The smallest unit combining fuel elements into
an assembly. For example, in a LWR the fuel assembly is composed of
several hundred fuel elements fastened together at top and bottom with
coolant nozzle plates and with several spring clip assemblies along the
length of the fuel (see Figure 3-7). In an HTGR the fuel assembly is a
hexagonal block of graphite with holes into which the cylindrical fuel
pins are inserted. Fuel is usually loaded into a reactor core or replaced
one fuel assembly at a time. A typical power reactor core will contain
hundreds of such fuel assemblies.
FISSION CHAIN REACTIONS AND NUCLEAR REACTORS-AN INTRODUCTION / 9S
(4) Moderator: Material of low mass number which is inserted into the
reactor to slow down or moderate neutrons via scattering collisions.
Typical moderators include light water, heavy water, graphite, and
beryllium.
(5) Coolant: A fluid which circulates through the reactor removing fission
heat. The coolant can be either liquid, such as water or sodium, or
gaseous, such as helium or carbon dioxide. It may also serve a dual role
as both coolant and moderator, such as in the LWR.
(6) Coolant channel: One of the many channels through which coolant flows
in the fuel lattice. This may be an actual cylindrical channel in the fuel
assembly, as in the HTGR, or an equivalent channel associated with a
single fuel rod, as in a LWR.
(7) Structure: The geometry and integrity of the reactor core is maintained
by structural elements such as support plates, spacer grids, or the
metallic tubes used to clad the fuel in some reactor designs. The
structural materials may also serve a dual role by moderating neutrons
such as the graphite in an HTGR.
(8) Control elements: Absorbing material inserted into the reactor to control
core multiplication. Although most commonly regarded as movable rods
of absorber, control elements may also consist of fixed absorbers or
absorbing materials dissolved in the coolant. Common absorbing
materials include boron, cadmium, gadolinium, and hafnium.
(9) Reactor core: The total array of fuel, moderator, and control elements.
(10) Reactor blanket: In a breeder or high conversion reactor the core is
usually surrounded by a blanket of fertile material that more effectively
utilizes the neutrons leaking out of the core.
(11) Reflector: A material characterized by a low absorption cross section
used to surround the core in order to reflect or scatter leaking neutrons
back into the core.
(12) Shielding: The reactor is an intense source of radiation. Not only must
operating personnel and the public be shielded from this radiation, but
reactor components must as well be protected. Hence absorbing material
is introduced to attenuate both neutron and gamma radiation. Thermal
shielding is used to attenuate the emergent core radiation to levels that
do not result in significant heat generation and hence damage in reactor
components. Biological shielding reduces the radiation still further to
acceptable levels for operating personnel.
(13) Support structure: The support plates that serve to maintain the core
geometry.
(14) Reactor pressure vessel: The high pressure containment for reactor and
associated primary coolant system.
It is also useful to introduce at this point several quantities which are used to
describe reactor performance. The units in which these quantities are usually
expressed are denoted in brackets.
(1) Reactor thermal power [MWt]: The total heat produced in the reactor
core.
(2) Plant electrical output [MWe]: Net electrical power generated by the
plant.
Plant electrical output

(3) Net plant efficiency [%]: R th .
eactor erma1 power
. Total energy generated over time period
(4) Plant capacity factor [%}: (Plant ratmg
. ) X ( time
. ) .
Average plant electrical power level

(5) Plant load factor[%]: p ak .
e power 1eve1
Integrated electrical energy output capacity
(6) Plant availability factor [%]: - - - - - - - - - - - - - - - -
Total rated energy capacity for period
Reactor thermal power

(7) Core power density [kW /liter]: T 1 ·
ota core vo1ume
(8) Linear power density [kW/mJ: Thermal heat generated per unit length of
coolant channel.
Reactor thermal power
(9) Specific power [kW /kg]: Total mass of fissionable material·
(10) Fuel loading [kg]: Total mass of fissionable material.
. Mass of fissile material
(11) Fuel enrichment [%]: M
ass off1ss1.1e and f ertt.1e materia
. 1.
(12) Fuel burnup [Megawatt-days/metric ton uranium==MWD/TU]:

Energy generated in fuel during core residence
Total mass of fuel
Fuel burnup
(13) Fuel residence time:
(Specific power) X ( capacity factor) ·
These are the more common terms used in characterizing nuclear plant perfor-
mance. We will introduce other more specific concepts and terminology later as we
develop the more detailed theory of nuclear reactor behavior.
III. NUCLEAR REACTOR DESIGN
A. General Design Functions of the Nuclear Engineer

The design of a large nuclear power plant is an enormously complex task and
involves the coordination of a remarkably diverse range of disciplines. Each major
component of the plant requires a separate and distinct design analysis and is
usually the responsibility of a specific engineering design team. For example, the
design of the reactor pressure vessel or steam generators is usually performed by
the reactor supplier, while the turbogenerator and switchgear design is the re-
sponsibility of the electrical equipment manufacturer. The coordination among
these different design projects is extremely important, however, since the designs
frequently interact to a very high degree.
The primary responsibility for the nuclear design of the reactor core rests with
the nuclear engineer. This design must be accomplished within numerous con-
straints imposed on the reactor operation. The nuclear analysis and design of a
reactor core is highly dependent on other areas of core design, including thermal-
hydraulic design, structural analysis, economic performance, and so on. The
criteria for a design effort are quite varied, encompassing considerations of per-
formance, reliability, economics, and safety. These criteria are frequently contra-
dictory in nature, and hence require optimization.
The complete nuclear design of a given core configuration is performed many
times, initially to survey design parameters, identify constraints, then to refine the
design while interacting with other facets of the plant design, and finally, to
establish a reference design that. provides a calculational base against which
optimization calculations can be compared.5
This design process is very similar to that utilized in other fields of engineering.
One first must attempt to define the various design constraints that include
considerations of system performance in terms of both system reliability and
economic performance and safety criteria. Next a preliminary design is proposed,
drawing on available information such as plants already in operation, experimental
mockups, and frequently, old-fashioned intuition. Such a design includes a set of
specifications involving quantities such as fuel enrichment, coolant flow rates and
temperatures, core configurations, reload patterns, and so on. A detailed analysis
of this preliminary design is then performed in order to evaluate its predicted
performance and ascertain whether it conforms to the constraints imposed on the
system. For example, one would want to calculate the core power and temperature
distribution, the pressure drop of the coolant as it passes through the core,
coolant-flow conditions, and the fuel lifetime. When possible, these calculations are
compared against experiments in order to validate the computational models used.
A detailed evaluation of the preliminary design will then lead to more detailed
designs and analyses as one attempts to optimize the tradeoff between system
performance and design constraints. As a final design is approached, one attempts
to define detailed system specifications.
The above procedures emphasize the importance of adequate models of a
nuclear reactor in order to carry out the required parameter and optimization
studies. These models must be realistic since nuclear reactors are fa1 too expensive
to be built without detailed and accurate design information. Unfortunately any
calculation sufficiently realistic to be of use in reactor design is far too complex to
be carried out by hand. Hence the digital computer plays a very key role in nuclear
reactor design. 6
A key task of the nuclear reactor engineer is to develop models of nuclear
reactors that can then be analyzed on the computer. Such models result in large
computer programs or "codes" which can then be used by other nuclear engineers
in reactor design. Most of our emphasis in this text is on learning how to synthesize
such approximate models of reactor behavior and then cast them in a form suitable
for reactor design. In the language of nuclear engineering, then, this text should be
regarded as a primer on nuclear methods development.
A word of caution should be inserted here, however. In the early days of reactor
development it was hoped that one would eventually be able to accurately model
nuclear reactor behavior utilizing only fundamental principles and measured
nuclear data. However over the past three decades of reactor development ex-
perience it has become apparent that the accuracy of nuclear data and computa-
tionally feasible analytical methods are simply not sufficient to allow this. 6 Instead
nuclear reactor analysis has relied heavily on that very basic ingredient utilized in
most other areas of engineering design known as the "enlightened fudge." That is,
most nuclear analysis methods or computer codes contain empirical parameters
that have been adjusted or cali~rated by comparing the predictions of the methods
with actual experimental measurements. While such empirical input is usually very
successful in yielding accurate nuclear design information with a minimum amount
of effort, the novice nuclear engineer should approach any existing nuclear analyti-
cal method or computer code with a high degree of skepticism, since methods
calibrated to work well for one range of parameters may fail miserably when
applied to new situations in which only limited experience is available.
The intimate relation between computers and reactor design cannot be over-
stressed. It is almost impossible for the present-day nuclear engineer to function
without a reasonable background in computer techniques (both in programing and
numerical analysis). Nevertheless the increasingly heavy reliance of the nuclear
reactor industry on elaborate computational models of reactor performance makes
it even more imperative that the nuclear engineer possess a very thorough
background in the fundamental physical and mathematical concepts underlying
these models, as well as a healthy dose of skepticism when he attempts to utilize
their predictions in reactor analysis.
B. Some Concluding Remarks

In these last three chapters we have attempted to introduce several simple but
important concepts involved in nuclear fission chain reactions. We have also
provided a brief overview of nuclear reactor systems and the function of the
nuclear engineer in the design of such systems. With this background we now turn
our attention to a development of the theory underlying the nuclear analysis of
fission reactors. We have seen that the neutron plays the central role as the chain
carrier perpetuating the chain reaction. The key problem of reactor theory, then, is
to determine the distribution of neutrons in a reactor core. This will not only allow
one to study the chain reaction process itself, but, as we will later find, since the
neutron density is proportional to the rate at which fission reactions occur and
hence proportional to the core power density, the neutron density is also the key to
the subsequent thermal and mechanical analysis of the reactor. As we have
mentioned earlier, there are essentially two aspects to this problem.
One must study the interaction of neutrons with matter-specifically, with the
nuclei that make up the matter. This amounts to either experimental or theoretical
determination of the probabilities that various neutron-nuclear interactions will
occur-that is, a determination of the appropriate neutron-nuclear cross sections.
This, however, is not the principal concern of the theory we will develop in this text
but is more properly the domain of the nuclear physicist. Hence we tend to take
microscopic cross section data as given (in a form to be discussed later), and turn
our attention instead to the manner in which these data are utilized in nuclear
reactor analysis.
Of comparable importance is the study of the transport or diffusion of neutrons
within a nuclear reactor core as they stream around inside the core, suffering
collisions with nuclei, occasionally being absorbed, inducing fission reactions, or
leaking out through the surface of the core. It is this latter study that will allow us
to develop models for calculating the distribution of neutrons within the reactor
core.
Our theoretical approach is to begin with an essentially exact description of the

neutron density in the reactor based on the so-called neutron transport equation.
This equation, while relatively easy to derive, is extremely difficult to solve, and
hence we will be concerned with developing various approximations to it that lend
themselves more readily to practical application. We begin our actual study of
nuclear reactor theory by using the simplest such approximation, that in which the
neutron energy dependence is neglected by assuming all neutrons to be
characterized as having a single speed, and describing their transport from point to
point as a simple diffusion process. This very simple model suffices to develop most
of the concepts, as well as to illustrate most of the practical computational
techniques used in more detailed reactor analysis.
We next develop a more sophisticated model of the neutron density behavior
based on breaking up the range of neutron energies into intervals or "groups" and
then describing the diffusion of neutrons in each of these groups separately,
accounting for the transfer of neutrons between groups caused by scattering. Such
multigroup diffusion models are the principal tools used in modern reactor analysis,
and we consider them in some detail.
In the final section of the book we illustrate these models by applying them to
analyze several typical problems encountered in nuclear reactor design. In particu-
lar, we explore the relation between such nuclear analysis methods and the other
types of analysis required in nuclear reactor core design.
REFERENCES
I. A. M. Weinberg and E. P. Wigner, The Physical Theory of Neutron Chain Reactors, The
2. S. Glasstone and A. Sesonske, Nuclear Reactor Engineering, 2nd Ed., Van Nostrand,
Princeton, N.J. (1975).
3. M. M. El-Wakil, Nuclear Energy Conversion, Intext, Scranton (1971).
4. There are a number of other useful sources on nuclear power systems. Each of the major
suppliers of NSSS prepare detailed systems descriptions. For example, a very informative
reference is
Systems Summary of a Westinghouse Pressurized Water Reactor Nuclear Power Plant,
Westinghouse Electric Corporation (1971).
The most detailed descriptions of modem nuclear power plants can be found in the
multivolume set of Preliminary Safety Analysis Reports (PSARs) or Final Safety Analysis
Reports (FSARs) prepared for each nuclear plant. Of particular interest are the standard
safety analysis reports prepared for each of the major NSSS types. For example:
Babcock and Wilcox Standard Nuclear Steam System, B-SAR-241 (1974).
BWR/6 Standard Safety Analysis Report, General Electric Company (1973).
CESSAR, Combustion Engineering Standard Safety Analysis Report, System 80,
Combustion Engineering (1973).
GASSAR 6, General Atomic Standard Safety Analysis Report, GA-Al3200 (1975).
Clinch River Breeder Reactor Plant, Reference Design Report, Westinghouse Electric
Corporation, 1974; PSAR (1975).
RESAR-3, Reference Safety Analysis Report, Westinghouse Nuclear Energy Systems
(1973).
5. A. Sesonske, Nuclear Power Plant Design Analysis, USAEC Document TID-26241 (1973).
6. J. Chernick, Reactor Technol. 13, 368 (1971).
PROBLEMS
3-1 What is the maximum value of the multiplication factor that can be achieved in any
conceivable reactor design?
3-2 Using the alternative definition of the multiplication factor based on the concept of
neutron balance, repeat the derivation of the six-factor formula.
3-3 A spherical reactor composed of 235 U metal is operating in a critical steady state.
Discuss what probably happens to the multiplication of the reactor and why, if the
system is modified in the following ways (treat each modification separately, not
cumulatively): (a) the reactor is rapidly compressed to one-half its original volume, (b)
a large, fat reactor operator accidentally sits on the reactor, squashing it into an
ellipsoidal shape, (c) a thick sheet of cadmium is wrapped around the outside of the
reactor, (d) the reactor is suddenly immersed in a large container of water, (e) a source
of neutrons is placed near the reactor, (f) another identical reactor is placed a short
distance from the original reactor, and (g) one simply leaves the reactor alone for a
period of time.
3-4 One defines the doubling time for a breeder reactor as the amount of time required for
the original fissile loading of the reactor to double. Find an expression for the doubling
time td in terms of: (a) the original fissile loading Mp, (b) the power level of the reactor
P= wrFc where Fr is the fission rate occurring in the reactor core, (c) and the breeding
ratio BR.
3-5 A detailed comparison of typical power reactor core parameters is given in Appendix
H. Choose one of the reactor types in this Appendix and perform the following
calculations: (a) verify that the average linear power density, power density, and
specific power given in the table are consistent with the core volume, thermal power
rating, and fuel loading, (b) determine the discharge fuel burnup when the capacity
factor of the nuclear unit is 80% and the fuel residence time is three years, (c)
determine a range for core height and core diameter and sketch a core cross-section for
one array of assemblies using the tabulated core data.
3-6 Calculate and plot k«i as a function of enrichment from 0.7% 235U to 100% 235U. Use
the thermal cross section data of Appendix A and assume p = E == 1.
3-7 Derive a relationship between the waste heat rejected from a plant of a given output
and the thermal efficiency of the plant. (Several years ago such waste heat was referred
to as "thermal pollution." In a countermove, several of the more optimistic spokesmen
for the nuclear power industry coined the phrase "thermal enrichment.") Using this
expression, estimate the waste heat rejected by: (a) a modem fossil-fuel plant, (b) a
LWR plant, (c) a HTGR plant, (d) an LMFBR plant, and (e) a fusion reactor plant,
assuming that all of these plants are rated at 1000 MWe. Treat the efficiency of the
plant as that for an ideal (Carnot) heat engine.
3-8 Consid~r an infinitely large homogeneous mixture of iJsu and a moderating material.
Determine the ratio of fuel-to-moderator density that will render this system critical for
the following moderators: (a) graphite, (b) beryllium, (c) water (H20), and (d) heavy
water (D20). Use the thermal cross section data given in Appendix A.
3-9 Modify the simple description of the time behavior of the neutron population in a
reactor given by Eq. (3-5) to account for the presence of a source in the reactor
producing S0 neutrons per second. In particular, determine the time behavior of the
neutron population for each of the three cases: k < 1, k = l, and k > l.
2
The One-Speed
Diffusion Model
ofa
Nuclear Reactor
4
Neutron Transport
We now turn our attention to the central problem of nuclear reactor theory, the
determination of the distribution of neutrons in the reactor. For it is the neutron
distribution that determines the rate at which various nuclear reactions occur
within the reactor. Furthermore by studying the behavior of the neutron population
we will be able to infer the stability of the fission chain reaction. To determine the
distribution of neutrons in the reactor we must investigate the process of neutron
transport, that is, the motion of the neutrons as they stream about the reactor core,
frequently scattering off of atomic nuclei and eventually either being absorbed or
leaking out of the reactor. Most reactor studies treat the neutron motion as a
diffusion process. In effect one assumes that neutrons tend to diffuse from regions
of high neutron density to low neutron density, much as heat diffuses from regions
of high to low temperature, or even more analogously, as one gas of molecules
(corresponding to the neutrons) would diffuse through another (the nuclei) to
reduce spatial variations in concentration.
Unfortunately, however, while the treatment of thermal conduction and gaseous
diffusion as diffusion processes is usually found to be quite accurate, the treat-
ment of neutron transport as a diffusion process has only limited validity. The
reason for this failure is easily understood when it is noted that in most diffusion
processes the diffusing particles are characterized by very frequent collisions that
give rise to very irregular, almost random, zigzag trajectories. However, we have
seen that the cross section for neutron-nuclear collisions is quite small (about 10- 24
cm2). Hence neutrons tend to stream relatively large distances between interactions
(recall that the mean free path characterizing fast neutrons is typically on the order
of centimeters). Furthermore, the dimensions characterizing changes in reactor core
composition are usually comparable to a neutron mfp (e.g., a reactor fuel pin is
typically about 1 cm in diameter).
103
104 / THE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
Hence one frequently requires a more accurate description of neutron transport

that takes into account the relatively long neutron mfp and neutron streaming.
Such a description has been borrowed from the kinetic theory of rarefied gases
(which are also characterized by long mfp)-more precisely, the kinetic theory of
gas mixtures. The fundamental equation describing dilute gases was first proposed
more than one century ago by Boltzmann, and even today the Boltzmann equation
remains the principal tool of the gas dynamicist. 1 Its counterpart for the neutron
"gas," the so-called neutron transport equation, is far younger (less than 40 years
old}, far simpler (e.g., it is a linear equation in contrast to the Boltzmann equation,
which is nonlinear), but usually strikes far more terror in the hearts of fledgling
nuclear engineers who are intimidated by its frightening reputation within the
nuclear reactor community. Neutron transport theory has come to be associated
with a hideous plethora of impenetrable mathematics, unwieldy formulas, and
(eventually) the expenditure of enormous amounts of money on computer number-
crunching.
This is most unfortunate because the neutron transport equation is much simpler
to derive (requiring only the concept of neutron conservation plus a bit of
vector calculus) and to understand than the neutron diffusion equation that we
shall utilize in most of our development of reactor analysis. It is also a far more
fundamental and exact description of the neutron population in a reactor-indeed,
it is the fundamental cornerstone on which all of the various approximate methods
used in nuclear reactor analysis are based.
It does have one major drawback, however. It is usually very difficult to solve
the transport equation for any but the simplest modeled problems (and even these
require an inordinate amount of analytical work). However that is quite all right,
since it is not our intent to attack the transport equation head on. Rather the job of
the reactor analyst is to develop suitable (i.e., calculationally feasible and accurate)
approximations to it. Usually, however, only by comparing these various approxi-
mate theories to the transport equation from which they originated can one really
assess their range of validity.
There is another reason for including an introduction to the neutron transport
equation in even an elementary discussion of nuclear reactor analysis. Although
neutron diffusion theory is usually found adequate for reactor applications, it owes
its accuracy to various schemes that have been developed to "patch it up" using
results from more accurate transport equation solutions. For example, we will find
that the neutron diffusion equation is quite invalid near the boundary of a reactor,
or near a highly absorbing material such as a fuel rod or a control element.
Nevertheless we can continue to use diffusion theory to describe the reactor
provided w~ fudge it a bit by inserting so-called "transport corrections" into the
boundary conditions accompanying the diffusion equation.
So hopefully we have made a case for our inclusion of a very introductory
discussion of neutron transport theory within an elementary text. We would
caution the reader not to be intimidated by the notation or the apparent
strangeness of the equation we will develop. He should find it rather easy to
understand the derivation and interpretation of this equation.
Furthermore the effort he expends in understanding the material in this chapter
will provide him with a much deeper and more thorough understanding of the
approximate methods we will develop in later chapters. ·
With this strong note of encouragement, let us now add a qualification. We have
attempted to develop these later approximations in a manner independent of this
NEUTRON TRANSPORT / 105
chapter since we recognize that many nonnuclear engineers reading this text may
not really need (or care) to understand the limitations and ranges of validity of
nuclear analysis methods. Therefore if the reader is faint of heart in confronting the
transport equation, and strong in faith in being able to accept the rather heuristic
arguments necessary to develop these approximate theories (e.g., the neutron
diffusion equation) without recourse to the transport equation, or perhaps just
disinterested, he can proceed immediately to the development of neutron diffusion
theory in the next chapter.
The more formal discussion in this chapter will also serve to introduce the
standard numerical approximation schemes used to analyze the neutron behavior
in nuclear reactors. As in other areas of physical analysis, we will find that the
usual maxim applies: that the "brute force" numerical approach (i.e., discretize
everything in sight and slap in on a computer) is conceptually the simplest
approach to understand and computationally the most expensive calculation to
perform. The more elegant approximate methods require far less computational
effort but far more in the way of mental gymnastics in order to understand the
significance and reliability of their predictions.
I. INTRODUCTORY CONCEPTS
A. Neutron Density and Flux

Our ultimate goal is to determine the distribution of neutrons in a nuclear
reactor core. This requires accounting for the neutron motion about the core and
neutron interactions with nuclei in the core. We will begin by defining the neutron
density N (r, t) at any point r in the reactor core by
N(r,t)d 3r=expected number of neutrons in

·d 3r about rat a time t. (4-1)
The word "expected" has been inserted into this definition to indicate that this will
be a statistical theory in which only mean or average values are calculated. (The
actual neutron density one would obtain from a series of measurements would
fluctuate about this mean value, of course.) The neutron density N (r, t) is of
interest because it allows us to calculate the rate at which nuclear reactions are
occurring at any point in the reactor. To understand this, let us suppose for
convenience that all the neutrons in the reactor have the same speed v. Now recall
that on~ can express the frequency with which a neutron will experience a given
neutron-nuclear reaction in terms of the macroscopic cross section characterizing
that reaction~ and the neutron speed v as
v~ =interaction frequency. (4-2)
Hence we can define the reaction-rate density F(r,t) at any point in the system by
merely multiplying the neutron density N (r, t) by the interaction frequency v~:
expected rate at which
F(r,t)d 3r=v~N(r,t)d 3r=interactions are occurring (4-3)
in d 3r about rat time t.
X FIGURE 4-1. The neutron density N (r, t).
For example, if we consider a thermal neutron density of N = 108 cm- 3 in a

graphite medium then using the total cross section tabulated in Appendix A of
~, = 0.385 cm - I and a corresponding neutron speed of 2.2 x 1()5 cm/ sec, we would
find a reaction rate density of 8.47 x 1012 reactions/ cm3/ sec. In this particular
case, most of these reactions would consist of scattering collisions.
These concepts can easily be extended to the case in which the neutron density is
different for various neutron energies E by defining
expected number of neutrons in d 3r

N (r,E,t)d 3rdE =about r, energies in dE about E, at (4-4)
time t.
Notice that this "density" is defined with respect to both space and energy. One
can also generalize the concept of reaction rate density to include energy depen-
dence as
F(r,E,t)d 3rdE= v~(E)N (r,E,t)d 3rdE. (4-5)
The product vN(r,t) arising in Eqs. (4-3) and (4-5) occurs very frequently in
reactor theory, and therefore it is given a special name:
,P(r, t) = vN (r, t) = neutron flux [cm- 2 • sec- 1]. (4-6)
Although it will certainly prove convenient to work with ,p(r, t) rather than N (r, t)
(since then one does not have to worry about including the neutron speed v in the
reaction rate densities), the tradition in nuclear engineering of referring to this
quantity as the neutron "flux" is very misleading. For ,p(r, t) is not at all like the
fluxes encountered in electromagnetic theory or heat conduction, since these latter
fluxes are -vector quantities, whereas ,p(r, t) is a scalar quantity. Actually the
"neutron current" J(r, t), which we shall introduce momentarily, corresponds more
closely to the conventional interpretation of a "flux." To avoid unnecessary
confusion over this unfortunate convention, the student would probably do best at
this point to think of the neutron flux as simply a convenient mathematical variable
(speed x density) to use in computing reaction rates:
F(r, E, t) = ~(E )cp(r,E, t). (4-7)
A bit later we will introduce a physical interpretation of the neutron flux.
B. Angular Densities and Currents

The significance of the neutron density N (r, t) or flux cp(r, t) in determining
nuclear reaction rates leads us to search for an equation that describes these
quantities. Unfortunately there is no exact equation that is satisfied by N (r, t) or
cp(r, t)-only approximate equations. To understand why, we must generalize the
concept of the neutron density somewhat.
First let us determine just which variables characterize the state of an individual
neutron. Certainly these include the neutron position r, energy E (or speed
v=(2E/m) 112), and the time tat which the neutron is observed. Yet notice that to
specify the state of the neutron, we must also give its direction of motion
characterized by the unit vector O= v/ lvl. (Actually one could worry about
specifying other variables such as the neutron spin; but for reactor calculations, the
variables r, E, 0, and t provide a sufficient description of the state of the neutron.)
Let us now generalize the concept of density by defining the angular neutron
density that depends on all of these variables
expected number of neutrons in

n(r E O t)d3rdEdO=d 3r about r, energy d~ about E, (4-8)
' ' ' moving in direction O in solid
angle dO at time t.
[The term "angular" arises from the fact that n(r, E, 0, t) depends on the velocity
spherical coordinate angles () and cp specifying the neutron direction O (see Figure
4-2).] This is the most general neutron density function we need to define since it
happens that one can derive an essentially exact equation, the neutron transport
equation, for the angular neutron density· n(r,E,0, t).
However before deriving· this equation, it is useful to introduce several other
definitions. We will first define the angular neutron flux in a manner similar to that
in which we earlier defined the neutron flux, simply by multiplying the angular
density by the neutron speed v:
<p(r,E,0, t) = vn(r,E, 0, t). (4-9)
A related concept is the angular current density, defined by
j(r,E,0,t)-vOn(r,E,0,t) = Oqi(r,E,O,t). (4-10)
Notice that since Ois a unit vector, the angular flux is actually nothing more than
the magnitude of the angular current density
Iii= ID Iqi= cp. (4-11)

IA
n
FIGURE 4-2. 1be position and direction variables characterizing a neutron.
The angular current density has a useful physical interpretation. Consider a small
area dA at a point r. [Here we will use the convention that dA=e8 dA where es is
the unit vector normal to the surface.) In particular note that if we consider a small
area dA at a point r, then
expected number of neutrons passing
j(r,E,0,t)·dAdEdO= through an area dA per un!t timeAwith (4-12)
energy E in dB, direction O in dO
at time t.
We can also define an angular interaction rate
J(r,E,0,t) = v~(r,E)n(r,E,0,t) =~(r,E)cp(r,E,0,t). (4-13)

All of these angle-dependent quantities can be related to our earlier definitions in
Section 4-1-A by simply integrating over the angular variables. For example:
N(r,E,t)= 1 dOn(r,E,0,t),
4'1T
(4-14)
____,
____,
____,
____, FIGURE 4-3. Neutrons incident on a dffferential
element of area dA.
or
N(r,t)= f 00dEN(r,E,t) = l 00dE i dlln(r,E,Sl,t)

0 0 4,r
(4-15)
Sometimes quantities such as N (r, t) and </>(r, t) which do not depend on Sl are
referred to as scalar or total densities and fluxes, to distinguish them from
n(r,E,Sl,t) and cp(r,E,Sl,t). We find this nomenclature cumbersome and will avoid
it in our development.
Notice that if the angular density is independent of Sl (i.e., it is isotropic) then we
find that Eq. (4-14) demands the presence of a 4w normalization factor in the
angular density
A 1
n(r,E,0,t)= 4'iTN(r,E,t). (4-16)
More generally, however, n(r,E,Sl,t) will have a directional dependence-

particularly if we are near a boundary or a source of neutrons, as a little
geometrical reasoning applied to Figure 4-4 should indicate.
In a similar fashion, we find
q,(r,E,t)= i dllcp(r,E,Sl,t),
4'1T
(4-17)
and
q,(r,t)= f 00dEq,(r,E,t) = f 00dEi

0 0 4'1T
dllcp(r,E,0,t) . (4-18)
Finally, we can define the neutron current density J(r, E, t) in terms of the angular
current density j(r,E,Sl,t) as
J(r,E,t)= i4,r
dOj(r,E,0,t), (4-19)
and
J(r,t)= loo dEJ(r,E,t) = loo dE i
0 0 4'1T
dOj(r,E,Sl,t) (4-20)
FIGURE 4-4. Anisotropies in the angular density n(r,E,'2,t) near a boundary or

a neutron source.
110 / TIIE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
Notice that J(r, t) is actually what would be referred to as the "flux" in other fields
of physics, since if we have a small area dA at a position r, then
J(r,t)·dA= net rate at which neutrons pass (4-21)

through a surface area dA.
The units of both J(r,t) and <j>{r,t) are identical [cm- 2 •sec- 1]. However J is a
vector quantity that characterizes the net rate at which neutrons pass through a
surface oriented in a given direction, whereas </> simply characterizes the total rate
at which neutrons pass through a unit area, regardless of orientation. Such an
interpretation would suggest that J is a more convenient quantity for describing
neutron leakage or flow (e.g., through the surface of the reactor core), while </> is
more suitable for characterizing neutron reaction rates in which the total number of
neutron interactions in a sample (e.g., a small foil) is of interest. Although the
angular flux and current density are very simply related, we will find that there is
no simple analogous relationship between J and <j>. These concepts may appear a
bit confusing at first, but they will become more familiar after we have illustrated
their application in both our further theoretical development and the problems at
the end of the· chapter.
A closely related concept is that of the partial current densities, J ::t (r, t) which
correspond to the total rates at which neutrons flow through a unit area from left
to right (J +) or right to left (J _). If we recall our earlier definition of j{r, E, Sl, t),
then it becomes apparent that
J +(r,t) = ( 00 dE
- lo
r
J2'1T±
dSles·j(r,E,O,t),
(4-22)
where 2'17'::t is merely a convenient notation to indicate that the angular integration
is performed only over directions with components along the surface normal (2'11'+)
or in the opposite direction (2'11'-). For example, if we choose to define the polar
coordinates that specify O along the normal to the surface, then in the integration
for J +•</>would range from Oto 2'17', while (J would range only from Oto '1T /2.
It is evident from this definition that
(4-23)
Hence J is sometimes referred to as the net current density, since it can be

constructed as the sum of the partial current densities.
I /J(r.t)
~dA _3--;,..
1_ (r,t)
FIGURE 4-5. Partial and total current densities.

II. THE NEUTRON TRANSPORT EQUATION

We will now derive an exact equation for the angular neutron density in a
system by simply balancing the various mechanisms by which neutrons can be
gained or lost from an arbitrary volume V within the system. That is, we will
consider mechanisms that will change the number of neutrons in this volume that
are characterized by a specific energy E and are traveling in a specific direction 0.
It is convenient to use a bit of vector calculus here, but hopefully this will not
obscure the simple physics behind this equation (which is just the mathematical
expression of a ..count-the-neutrons" game).
To this end, consider any old arbitrary volume V. The number of neutrons in V
with energy E in dE and traveling in a direction O in dO within this volume is just
(Since n(r,E,O,t), is a "density" in E and O space, we must multiply it by dE and

dO in order to get a number.) The time rate of change of this number, then, is
given by a balance relation
it [f/(r,E,ll,t)d r]
3 dEdO=gain in V-loss from V. (4-24)
If we assume that the arbitrary volume V is chosen not to depend on time, we can
bring the time differentiation inside the spatial integration
(4-25)
We will now classify the various ways that neutrons can appear or disappear from
V, and then we will try to write mathematical expressions for each of these
mechanisms in terms of the angular density n(r,E,O,t).
Gain mechanisms:
(DAny neutron sources in V (e.g., fissions).
a) Neutrons streaming into V through the surface S.
® Neutrons of different E', O' suffering a scattering collision in V that changes
E', O' into the E,O of interest.
V:
s
FIGURE 4-6. An arbitrary volume V with surface area S.

Loss mechanisms:
@Neutrons leaking out through the surface S.
G) Neutrons in V suffering a collision. (It is obvious that an absorption interac-
tion removes a neutron from V; and since by definition a scattering collision
changes E, fa and since we are only keeping track of neutrons in V with this
specific energy and direction, a scattering collision also amounts to a loss of
neutrons.)
We can now write a mathematical expression for each of these contributions. We
will work progressively from the easiest to the more difficult:
© Source terms: If we define
" ,.. rate of source neutrons appearing
s(r,E,D,t)d 3rdEdD-ind 3raboutr dEaboutE and (4-26)
A A ' ' '
dD about 0
then obviously
(4-27)
[This term was really easy-we only needed to define a source density,
s(r, E, 0, t).]
G) Loss due to collisions in V: The rate at which neutrons suffer collisions at a
point r is
ft (r,E,0, t) == v~r(r, E)n(r, E, 0, t). (4-28)
Hence integrating this collision rate over the volume V, we find
(4-29)
0) Gain due to neutrons scattering into dE about E, dO about O from other

energies E' and directions Sl': If we recall from Chapter 2 that the probability
of scattering from E', Sl' to E, 0 is given in terms of the double-differential
scattering cross section, then the rate at which neutrons scatter from E',0' to
E,0 is
(4-30)
However we must consider contributions from any E',O'. Hence
This is known as the inscattering term since it characterizes neutrons scattering

from other energies or directions ihto dE dO. · ·
Q)@Leakage into or from the volume V: We will combine these terms together and
calculate the net leakage through the surface S. If we use the concept of the
angular current density j(r, E, O,t), we can write the rate at which neutrons of
E, 0 leak out of a piece of the surface, dS, as
j(r,E, D,t)·dS= vlln(r,E,ll, t)·dS. (4-32)
Hence the leakage contribution over the entire surface area S is
(4-33)
We can rewrite this in terms of a volume integral if we use Gauss's theorem
l dS · A(r) = fv d 3r V · A(r), (4-34)
to find
[fsdS·vlln(r,E,ll,t) ]dEdll= [fvd 3rV·vlln(r,E,ll,t) ]dEdll
= [fvd 3rvll·Vn(r,E,ll,t) ]dEdll. (4-35)
Here we have noted that
V·v!l=v!l·V (4-36)
since D does not depend on r.

If we now combine all of these terms such that
rate of change of number = (D +a)+ 0) _ @- G), (4-37)

of neutrons in V
then we find
.(d 3r[ ~~ +vll·Vn+v}-.:,n(r,E,ll,t)
- fo 00
dE 1 L.,,dll'v'}-.:s(E'➔E,ll'➔!l)n (r,E',O',t)-s(r,~,n,t) ]dEd!l=O. (4-38)
However we now apply the fact that the volume V was quite arbitrarily chosen.
Hence the only way for the integral to vanish for any V is for its integrand to be
identically zero-that is,
fany V
d 3 rf(r)=0 ~J(r)=O. (4-39)
Hence we arrive at a balance relation
T
an + vO·Vn + v}-.:,n(r,E,O,t)
A A
U( l
= i dO' LdE' v'}-.: 5 (E'➔ E,D 1 ➔D)n(r,E',O',t)+s(r,E,O,t).

4.,, 0
(4-40)
This is known as the neutron transport equation. Several general features of the
equation should be noted: First, it is a linear equation in the unknown dependent
variable n(r,E,0,t) with seven independent variables (r=x,y,z;E;D=O,tp;t).
Since it contains both derivatives in space and time as well as integrals over angle
and energy, it is known as an "integrodifferential" equation.
However the presence of the derivatives suggest that we must also specify
appropriate initial and boundary conditions for the angular density. Since only a
single time derivative appears in the equation, we can simply choose the initial
condition to be the specification of the initial value of the angular density for all
positions, energies, and directions:
Initial condition: n(r,E, 0, 0) = n0(r,E, 0), all r,E, 0. (4-41)
The boundary conditions will depend on the particular problem of interest.

Suppose for purposes of illustration, that we consider the reactor to be surrounded
by an infinite vacuum so that if a neutron leaks out, it can never be scattered back
into the system. Actually to be more precise, we should also assume that the system
geometry is characterized by a nonreentrant surface such that a neutron streaming
out through the surface will never reenter the surface at another point (see Figure
4-7). Then our appropriate boundary condition would simply express the fact that
there can be no neutrons entering the system from the outside. That is, we require
the angular neutron density on the surface to vanish for all inward directions
(4-42)
where r5 denotes a point on the surface S. There are. other possible boundary
conditions, but we will discuss these later.
It is convenient to rewrite the neutron transport equation along with its initial
and boundary conditions in terms of the angular flux
-VI -a
O<p
t
A
+!l·V<p+~,(r,E)<p(r,E,!1,t)
"
= r dD' ( 00 dE'};s(E'~E,O'~O)<p(r,E',O',t) +s (r,E,D,t), (4-43)

J4w lo
Initial condition: <p(r,E,0,0)=<p0(r,E,0), (4-44)
Boundary condition: <p(r5 ,E,0,t)=O if Sl·t\<0,

(4-45)
all r5 on S.
Reentrant surface Nonreentrant surface

FIGURE 4-7. Examples of reentrant and nonreentrant surfaces.
NEUTRON TRANSPORT / US
FIGURE 4-8. A free surface boundary.
EXAMPLE: Suppose we try to make this equation a little bit less abstract by
applying it to the special case in which there is plane symmetry, that is, where the
neutron flux depends only on a single spatial coordinate, say, x (as shown in Figure
4-9). Then the directional derivative Sl · V reduces to
(4-46)
For convenience, we will choose our angular coordinate system with its polar
coordinate axis in the x-direction. Then Ox= cos 0. The assumption of plane
symmetry also implies that there is no dependence on the azimuthal angle cp. Hence
,.
n
FIGURE 4-9. Coordinates characterizing plane

Symmetry.
the one-dimensional form of the transport equation becomes
1 aq, acp
- ~ +cos8~ +"2.,cp(x,E,8,t)
V vt uX
= lo" d8' sin8' fo 00

dE' "2.s(E' ➔E,8' ➔fJ)q,(x,E',fJ', t)+ s(x,E,8, t). (4-47)
A final modification of the angular variable is useful. It is customary to rewrite this

equation in terms of a new variable, µ=cos8. Note that as 8 ranges between O and
'TT, µ ranges from 1 to - 1. Hence the usual form of the one-dimensional transport
equation is written as
1 acp acp
--a
V t
+µ.-
0X +"2.,cp(x,E,µ,t)
= f +l
dµ' Loo dE'"2.s(E' ➔ E,µ' ➔ µ)cp(x,E',µ',t)+s(x,E,µ,t). (4-48)
-I 0
We can easily generalize Eq. (4-43) to include nuclear fission by including a

component in the source term to account for fission neutrons. The rate at which
neutrons with energy E' and direction O' induce fission events is just "2.,(E')
cp(r,E',0',t). If 11(E') is the average number of fission neutrons produced by a
fission induced by a neutron of energy E', then the total rate at which fission
neutrons are born at a position r is just
i dO' Loo dE'11(E')"2.,(E')cp(r,E',O',t) .

4.,, 0
(4-49)
These fission neutrons will have an energy distribution given by the fission
spectrum x(E). If we assume that they are emitted isotropically, then the fission
source term we should include in the transport equation is just
,.. X( E)
s,(r,E,0,t)= ,-4-
i ,. i oo
dO'
,.
dE'v(E')"2.t(E')cp(r,E',O',t). (4-50)
'TT 4.,, 0
Actually we should qualify this argument a bit by admitting that we have assumed
all of the fission neutrons to appear instantaneously at the time of fission. Hence s,
is actually the source term corresponding to prompt fission neutrons. We will
develop the modifications necessary for delayed fission neutrons in Chapter 6.
The neutron transport equation provides an essentially exact description of the
neutron distribution within the reactor (at least, provided one is supplied with
appropriate cross section information). Its solution would yield the angular flux
cp{r,E,Sl, t) containing essentially all the information (actually considerably more)
we require concerning the nuclear behavior of the reactor. All we have to do is
solve this equation.
Yet notice that: (a) the neutron transport equation has seven independent
variables: x, y, z, 8, q,, E, t, (b) the dependence of the macroscopic cross sections
on position r is extremely complicated because of the complex, nonuniform
structure of most reactor cores, and (c) as we have seen in Chapter 2, the cross
section dependence on energy is also extremely complicated including resonance
structure, threshold effects, and so on. These considerations would immediately

suggest that any attempt to solve the transport equation for a realistic system will
involve heavy use of digital computers. Unfortunately no computer is sufficiently
large enough (yet) to solve this equation in the general form in which we have
derived it. This recognition implies that the major task of nuclear reactor analysis
will be to introduce suitable approximations to the neutron transport equation
which will allow us to solve it (at least on a computer), and yet still preserve enough
reality that we can obtain a useful description of the nuclear reactor.
We will begin our discussion of such approximations to the transport equation
by first briefly outlining the various "brute-force" procedures that can be used to
reduce this equation to a discretized form more suitable for digital computation.
We then will develop in a consistent fashion the principal approximations to the
transport equation (such as neutron diffusion theory) used in nuclear reactor
analysis.
III. DIRECT NUMERICAL SOLUTION OF THE TRANSPORT

EQUATION
Let us consider how one would attempt to solve the neutron transport
equation directly using the aid of a digital computer. First we must recognize that
digital computers are terrible at calculus (e.g., handling derivatives or integrals or
such). Their real talent is in solving very large systems of algebraic equations.
Hence our first task is to convert the transport equation into a system of algebraic
equations more suitable for a digital computer. This is accomplished by
"discretizing" each of the variables in the transport equation, that is, by replacing
functions of continuous variables by a discrete set of values at a discrete set of
points. The derivatives and integrals appearing in the transport equation must also
be replaced by a corresponding discrete representation. In this way one arrives at a
set of algebraic equations for the discrete representation of the dependent variable
(in our case, the neutron flux cp).
The discretization of the transport equation-or, indeed, any such differential
equation-can be accomplished by using either discrete ordinates methods or
function expansions.1 •8 To illustrate these approaches, suppose we have an equation
for a function, say f(x), which contains derivatives and integrals:
df dx
F ( J(x), dx, d2j
2 , ••• ,
f dx J(x ), ...
I I )
=0. (4-51)
In the discrete ordinate approach, one begins by representing the unknown f(x)
only by its values at a discrete set of points X; of the independent variable x. That
is, one first discretizes the domain of variation of x into a mesh of discrete points,
each of which is labeled by a subscript i. Then we replace j(x) by its value at each
of these meshpoints
f(x) ➔J(x;)-J;, i= 1, ... ,N. (4-52)
(Of course, these values J; are still unknown.) Notice that what we have actually
done is to replace a Junction f (x) by a column vector f
(4-53)
118 / 1HE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
In this sense then. the system of algebraic equations we will arrive at for the
unknown components of J can be written as a matrix equation.
We must next replace the various operations in the original equation by their
discretized counterparts. For example, we would represent derivatives by finite
difference formulas such as
(4-54)
Integrals would be represented as sums or numerical quadrature formulas such as
f dxf(x)= ~ wJ(x;)= ~ wJ;,

a
b N
i=l
N
i-1
(4-55)
where the w; are known as the quadrature weights. A thorough description of such
procedures can be found in any elementary textbook on numerical analysis, 7' 8
although frequently it is more useful to derive such numerical approximations
directly for the specific equation under investigation (as we will have occasion to
do in Chapter 5).
Such procedures lead eventually to a set of coupled algebraic equations for the
components J; that can be solved on a digital computer. Frequently these discrete
values of the unknown f(x) provide an adequate representation. However occa-
sionally one wishes to reconstruct the original unknown J(x) for all values of x
from the discrete values inf. Then one must interpolate between the point values!;
at X;, for example by using-polynomials. (See Figure 4-10.)
An alternative way to arrive at a discrete representation of an equation is to
write the unknown function as an expansion in a finite number of known functions
(frequently polynomials). If we call these expansion functions p 1(x), then we would
write
N
f(x)~ ~ J,pt(x). (4-56)
/= 1
Hence once again we find that the functionf(x) is represented by a vector
(4-57)
although in this case, the components of the vector are just the unknown expansion
coefficients!,. Notice that if we can determine these expansion coefficients, then we
can easily reconstruct the unknown function f(x) by merely using Eq. (4-56).
Interpolation is not required as it is with the discrete ordinates approach.
EXAMPLE: When the dependent variable ranges between -1 and + 1, a very

convenient choice of expansion functions are the Legendre polynomials:
(4-58)
This choice of expansion functions is frequently used to represent the angular

dependence of the neutron flux in one-dimensional problems in which µ=cos 8 is
f(x) Actual function
•
•
•
• • • Discrete representation
•
I
Interpolated representation
FIGURE 4-10. Discrete ordinate representation of a function.
the natural independent variable:

N
rp(x,µ,E,t) = L rp (x,E, t)P
1 1( µ). (4-59)
/= 1
(Here we might recall Eq. (4-48) as an example in which such an expansion would
prove suitable.)
There are a variety of techniques one can now use to obtain a set of algebraic
equations for the expansion coefficients from the original equation for f(x). For
example, it is frequently possible to substitute the expansion Eq. (4-56) into the
original equation, multiply by each of the expansion functions p 1(x), integrate over
the independent variable x, and then use various properties of the p1(x) (such as the
property of orthogonality, which we will discuss later) to arrive at a set of algebraic
120 / THE ONE.SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
equations for the f,. One can also use more elaborate schemes such as the calculus
of variations or so-called weighted residual methods to arrive at the set of algebraic
equations. Since we will make only very limited use of such function expansions in
our elementary development of numerical analysis methods in this text, we will
refer the interested reader to several detailed descriptions9• 10 of these techniques for
more information.
A. Discretization of the Angular Dependence

Let us first consider how these techniques can be used to discretize the
direction variable Sl in the neutron transport equation. In the discrete ordinate
approach, 11 we would first represent the independent variable by a discrete set of
directions or rays On,n = 1, ... ,N.
We then represent functions of Sl by only their values at each of these mesh
directions:
(4-60)
In such a discrete ordinate treatment of angle, the integral over Sl becomes a

summation:
(4-61)
where the wn are appropriately chosen quadrature weights for the particular
numerical integration scheme used to handle the angular integrals. In this scheme,
the transport equation reduces to a coupled set of N equations of the form:
1 acpn
--a
A
+On·Vcpn+~,cpn(r,E,t)
V t
(4-62)
where n= I, ... ,N while

CJln(r, E, t)-(()(r,E, On, t). (4-63)
This set of equations is commonly referred to as the SN equations 11 • 15 after early

one-dimensional treatments utilizing trapezoidal quadrature corresponding to treat-
ing the angular variation of the flux as N straight line segments.
One can also use functional expansions as an alternative scheme _to discretize the
angular variables. In the general case, this corresponds to expanding the angular
dependence of the flux in a finite series of the spherical harmonics3 Y,m(Sl)
= Y1m(fJ,cp) familiar from quantum mechanics: 12• 13 -
N +I
cp(r,E,Sl,t) = ~ ~ Cf!tm(r,E,t) Y1m (Sl). (4-64)
I-Om=-/
Then by substituting this expansion into the original transport equation, multiply-
ing by spherical harmonics of different order, Yl'm'(Sl), and integrating over the
angular variables, one can use orthogonality to obtain a coupled set of equations
for the expansion coefficients <J>,m(r, E, t). Since this set of equations is rather
complicated when written out for general geometries, we will refer the interested
reader to other sources for the general form of the equations. 3•4
In one -dimension, an expansion in spherical harmonics corresponds to an
expansion in Legendre polynomials, P1(µ.), where µ.=cosfJ:
(4-65)
In this case the general form of the equations for the expansion coefficients is
somewhat simpler and can be written as
1 a<J>, (I+ 1) a<p/+ 1 1 a<J>, _ 1

~at+ (21+ 1) ~ + (2/+ 1) ~ +}:t<p,(x,E,t)
= (<» dE'}:s (E'~E)<pi(x,E', t) + s1 (x,µ.,E), (4-66)

lo '
where one defines the angular components of the differential scattering cross
section as
(4-67)
This set of equations is known, naturally enough, as the PN equations.3•4

In the particular case in which the expansion in spherical harmonics is truncated
after two terms, that is N = 1, the expansion for the angular flux takes the form:
~ I 3
cp(r, E, n, t):::: 4'77 'Poo(r,E, t) + 4'1T [ 'P1Ar,E,t)Qx + <J>1y(r, £, t)Qy + 'P1Ar,E, t)Qz].
(4-68)
We will find in the next section that this P I approximation to the angular flux is
very closely related to neutron diffusion theory.
Hence we find that both discrete ordinate and functional expansion methods can
be used to discretize the angular variables, giving rise to the SN or PN equations
respectively. It might be mentioned that although the P 1 equations are used very
frequently in nuclear reactor analysis (indeed, we shall have occasion to use them
several times in our ensuing development), the higher order PN equations are rarely
used in practical calculations. Rather one usually relies on the discrete ordinate
approach if a more detailed treatment of the neutron directional dependence is
required.
B. Treatment of the Energy Variable

Very similar techniques can be used to discretize the energy variable E. Here,
however, the discrete ordinate approach is far more common than function
expansions. This can be easily understood when it is recognized that for the latter
technique to be effective, the expansion functions must bear some resemblance to
the actual functional dependence of the neutron flux on the independent variable.
The dependence of the angular flux on the neutron direction Ois usually rather
weak, hence a set of general functions such as the spherical harmonics will provide
an adequate description.
However, the neutron energy E spans an enormous range from 10- 3 eV up to
10 eV. The dependence of the neutron distribution on energy is determined by
7
quite different processes in different regions of energy. For example, at high

energies the neutron energy dependence is dominated by the fission spectrum. At
intermediate energies, neutron slowing down and resonance absorption are the
dominant processes, while at low energies, neutron thermalization is important. To
expand the neutron flux in a set of functions that adequately describe all of these
processes is clearly hopeless. Indeed such function expansions are capable of
describing neutron energy behavior only for a restricted range of neutron energies
or a specific reactor type. (An example of such an expansion known as energy or
spectrum synthesis is given in Chapter 13.)
One must be careful even when applying the discrete ordinate approach to the
energy variable. The difficulties involved become quite apparent when the very
detailed dependence of the neutron cross sections on energy is recalled. It clearly
would be unthinkable simply to consider these cross sections tabulated at several
discrete points as an adequate representation of this detailed structure.
Instead one first breaks up the neutron energy range into intervals or so-called
energy groups:
Eg E
(Don't worry about the fact that these subscripts appear to run in the wrong
direction-towards decreasing energy. It turns out that this is a more convenient
labeling since neutrons tend to slow down in energy.) The neutron transport
equation (or diffusion equation) is then integrated over each energy group in order
to define appropriate average values of the various cross sections characterizing
each group. For example, one would define the absorption cross section
characterizing a group g as
L.
f. Eg-ldEL-a( E )<p( E)
=-Es_ _ _ _ _ __
Dg - f. Eg-ldE<p( E) (4-69)
EB
Of course these are only formal· definitions of the cross sections characterizing the
group, since the flux itself appears in their definition. However they do form the
basis for the practical calculation of these quantities since one can insert approxi-
mations to the flux in order to calculate the group cross sections L."-. If we apply
this scheme to the SN equations derived in the previous section, we find the
coupled set of equations for the group fluxes q;!:
1 aq;; 0 V g L. g -
Vat+ n· <f>n + t,/Pn -
""' ~ L.g'-+g
L.I Wn, L.I
g' + g
sn, ...n <f>n' Sn, (4-70)
g 11' g'
n=l, ... ,N g=l, ... ,G.

Such sets of equations are known as the mu/tigroup equations (in this case, the
multigroup SN equations) and play a very important role in nuclear reactor analysis.
We will return in Chapter 7 to discuss their derivation in the case in which the
angular flux is treated within the diffusion approximation.
C. Treatment of Space and Time
The final step is to discretize the space and time variables. Although this can
be done using function expansions (more specifically, expansions in the spatial or
temporal modes of the system), such modal expansions are rarely used in general
studies (although they are occasionally used in certain types of spec;ific calculations
in which the modes can be easily calculated). Instead one utilizes a direct discrete
ordinate treatment in most cases. First the spatial variables r=(x,y,z) are decom-
posed into an appropriate spatial mesh. The various derivative terms are then
replaced by finite difference equations defined on this mesh.
Finally the time variable is broken into discrete time steps, say t 0, t 1, t 2, ••• and the
corresponding time derivatives are replaced by suitable difference formulas. The
detailed mesh structure and difference formulas one utilizes depends on the type of
problem one is investigating and will not be discussed here, since we will consider
such topics in much greater detail in later chapters.
D. Solution of the Discretized Equations

After using one scheme or another to discretize all of the variables in the
neutron transport equation, we are left with a large system of algebraic equations
for the components of the discretized representation of the flux, for example,
cp(x;,YpZk,Eg, nn,tJ. Such algebraic equations can then be solved using routine
numerical methods on a digital computer.
Unfortunately such a calculation becomes an immense undertaking if only a
"brute force" discretization of the transport equation is applied. For example,
consider "typical" mesh sizes of 100 X 100 x 100 space points, 10 energy groups,
and 10 angle points. Then for each time step we wish to calculate, we must solve
108 simultaneous algebraic equations-a rather formidable task, even on a modern
computer.
Hence the nuclear engineer cannot blindly depend on the computer to solve his
problem, but rather he must first rely on physical insight (rather than brute force
mathematics) to reduce the transport equation to more manageable form. He may
be able to eliminate one or more of the independent variables in the transport
equation. For example, one is usually interested in time-independent problems so
that the time variable can be ignored. Frequently the reactor geometry can be
modeled by a one- or two-dimensional calculation. The most useful class of
approximations to the transport equation eliminate the angular variable 0. The
most popular (and useful) scheme for accomplishing this is to introduce the
neutron diffusion approximation. Since neutron diffusion theory will constitute our
primary tool in our study of nuclear reactor behavior, just as it does for the nuclear
reactor industry, we will take some care in our derivation of the neutron diffusion
equation.
IV. THE DIFFUSION APPROXIMATION
A. The Neutron Continuity Equation

For most reactor calculations, the details of the angular dependence of the
flux are not necessary, and we really only need to calculate the angle-integrated
flux:
<t>(r,E,t)= l4w
dOcp(r,E,ll,t), (4-71)
in order to calculate nuclear reaction rates and hence study the chain reaction (e.g.,
by calculating the multiplication factor k).
Surely we can formulate an equation for cp(r, E, t) by simply integrating the
transport equation over angle. Let's try and see what happens. That is, we will
integrate each term of the transport equation (4-43) over the direction variable 0:
(4-72)
We can simplify each of these terms somewhat by the straightforward manipula-

tions indicated below:
CD = f dol at = l 1-at r
4'1T V
aq)
V )4'1T
dOcp I acp
= Vat'
(4-73)
(4-74)
G) = i4w
dOs(r,E,0,t) -S(r,E, t). (4-75)
Here we have used our earlier expression Eq. (4-16) for the neutron flux cp in terms
of the angular flux cp and also simply defined a source term S (r, E, t). To evaluate
the inscattering term © we first recall that ~sCE'~E,0'~0) usually depends only
on the scattering angle cosine µ.0 =fr· 0. This implies that
(4-76)
where ~ s(£' ~£) is just the "single" differential scattering cross section defined in
A A
Eq. (2-46). Hence we can interchange the order of integrations over O and 0' to
write:
(4-77)
So far everything is straightforward; but unfortunately the last term @ cannot be

evaluated in terms of cp(r, E, t). In fact we find that @ must be evaluated in terms
of the neutron current J by using Eq. (4-19):
@ = Ldll
4'17
ll·Vcp=V· L dllllcp=V·J(r,E,t).
4'17
(4-78)
If we now rewrite Eq. (4-72), it, talces the form
_!_ aacp + V ·J(r,E,t) + ~t(r,E)tp(r,E,t) = ( 00

dE'~s(E' ➔E)tp(r, E', t) + S (r,E,t).
V t Jo
(4-79)
This is known as the neutron continuity equation, since it is just the mathematical
statement of neutron balance.
It is important to note that this equation contains two unknowns, cp(r, E, t) and
J(r,E, t), unlike the neutron transport equation, which only contained one un-
known, the angular flux cp(r,E, 0, t). Hence by removing the angular dependence
we have in the process introduced another unknown, J(r, E, t), and hence we now
have an insoluble problem (i.e., one equation in two unknowns). The moral of this
story is that you don't get something for nothing-that is, merely integrating out
the angular dependence doesn't remove the complexities of the angular variation. It
just shifts them by demanding that one obtain yet another equation relating
cp(r, E, t) and J(r, E, t).
It is impossible to express J(r, E, t) in terms of cp(r, E, t) in a general and exact
manner. This is more apparent if we recall the definitions Eqs. (4-17) and (4-19):
cp(r, E, t)- i
4'17
dllcp(r, E, n, t), (4-17)
J(r,E,t)= Ldllllcp(r,E,ll,t).
4'17
(4-19)
It is obvious that these two quantities are entirely different functions, although they
can both be expressed in terms of an angular integral of the angular flux
cp(r, E, U, t). Hence there is no reason why one would expect these functions to be
simply related.
Undaunted by our failure to find a simple equation for cp(r,E,t), suppose we shift
our attention instead to developing an equation for the current density J(r,E,t). By
comparing the definitions in Eqs. (4-17) and (4-19) above, we are tempted to try
multiplying the transport equation by D and then integrating once again over
angle. Actually since the direction variable D is a vector,
n= ex sine coscp+ ey sine sin<t> + ez cos 0,
'---v-' '--v-- --.- ( 4-80)
ax ~ ~
we should multiply the transport equation by each component separately and
integrate. For example, the Qx component would yield
Each of these terms can be simplified in a manner similar to that used in deriving
the neutron continuity equation [Eq. (4-79)]:
(D = -1 -a
v at
i ,.
4'1T
dO Qxcp = -
v
1 a1x
-
at , (4-82)
Q) = ~t ( dQ Ox<p = "2.tJx, (4-83)

J4'1T
~ = ( dD nxs(r,E,D,r)-S1 (r,E,t). (4-84)
)4'/T X
Now to handle the inscattering term @), write
Next we do something a bit sneaky. Since D is a unit vector, we can write

D' -Q' = 1. We will insert this
into Eq. (4-85) so that we can rewrite it as:
Now we recall again that "2.s(E' ➔ E,D' ➔ D) depends only on the cosine of the
scattering angle µ.0 = O' ·0. Thus we can write
We will define
l4'1T
i
dD 0):2:"2. 8 (E'➔ E,O' •Q) = ½ dDD ·O'~s(E'➔ E,D' ·O)
4'1T
so that @) finally becomes
@= f 00<JE 1 °'2. 8 (E'➔ E) ( d!l'O:cp(r,E',Sl',t)= ( 00 dE 1 "i. 8 (E 1 ➔ E)Jx(r,E',t).

lo 1 J4w Jo 1
(4-89)
We are almost finished. Thus far each of our terms has been expressed in terms of
the current density Jx, except for the source term S1x which is a known term. But
wait! We still haven't considered the streaming term:
(4-90)
A quick glance at the integral term confirms our fears; once again the streaming
term has kicked out yet another new unknown. To see this more clearly, we can
combine these results along with similar results for QY and Qz to write Eq. (4-81) as
an equation for the current density J:
I aJ
--a +V·
1 dll,. OOcp(r,E,O,t)+"i.tJ(r,E,t)
.. .
V t 4w
(4-91)
However just as with the neutron continuity equation [Eq. (4-79)], we find that
integrating over Sl yields one equation but two unknowns, J(r, E, t) and
Il(r,E,t)= 1 dOOOcp(r,E,0,t).
4w
(4-92)
[Here we have taken the luxury of using a symbolic notation of writing two vectors
together, Sl!l. If this bothers you, just interpret this as a convenient notation for
taking each of the various combinations of components QxQx,QxQy,•••,QzQz
separately to construct a quantity with nine components, II xx• IIxy• ... , IIzz• Such
quantities are referred to as tensors (or, in this case, dyadics), but we won't need to
get so formal here.] It should be evident that we can get a new equation for
Il{r, E, t) by multiplying the transport equation by 00 and integrating, but this new
equation will contain yet another unknown,
1dO OOO<p(r,E,O,t).
4w
Hence all we are doing by multiplying by [Ot and integrating is generating an

infinite set of coupled equations (which we can't solve). [Incidently, it should be
apparent that the culprit is the "streaming" or "leakage" term O·V cp which
contains a factor of O and hence generates the new unknowns in each equation.]
The only way to cut off this chain of equations is to introduce an approximation.
We shall do this by assuming that the angular flux is only weakly dependent on
angle, and in so doing, we will generate the neutron diffusion equation. However it
is useful to first discuss several simplifications that can frequently be introduced
into the neutron transport equation.
B. Common Simplifications to the Neutron Transport Equation
1. THE ONE-SPEED APPROXIMATION

It is frequently convenient to suppress the neutron energy dependence by
assuming that one can characterize the neutrons by a single energy or speed. We
will find in Chapter 7 that if one chooses the appropriate effective cross sections,
such a representation will in fact frequently yield a reasonable description of the
reactor. However for now we will introduce the one-speed approximation in a rather
artificial manner by simply assuming that the neutron energy does not change in a
scattering collision. This can be inserted into the transport equation [Eq. (4-43)] in
a rather convenient manner by simply assuming a differential scattering cross
section of the form
(4-93)
where 13 (E' - E) is the Dirac 6-function defined by the property
f dx'f(x')/3 (x- x') =f(x) (4-94)
for any sufficiently well-behaved function f(x). [See Appendix C for a more
detailed discussion of animals such as the 6-function.] Using this definition, the
inscattering term in Eq. (4-43) becomes
(4-95)
Since all of the terms in the• transport equation are now evaluated at the same
energy, we may as well eliminate the explicit dependence on energy to write the
one-speed neutron transport equation as
This equation is still far too complicated to solve (even using brute force numerical
techniques) in realistic geometries. So we'll introduce yet another simplification.
2. ISOTROPIC SOURCES AND SCATIERING

One major simplification that can be introduced into the transport equation
arises when one assumes both isotropic neutron sources
... 1
s(r,O,t) = 4'17 S (r,t) (4-97)
and isotropic scattering (in the LAB system)
(4-98)
The assumption of isotropic neutron sources is usually not too restrictive since
most sources such as fission are indeed essentially isotropic. Unfortunately
although neutron scattering is usually isotropic in the CM system, it is far from
isotropic in the LAB system, particularly for low mass number scatterers such as
hydrogen. Undeterred by such physical considerations, we will assume for the
moment that isotropic scattering is present. Then the one-speed transport equation
simplifies still further to
1 aq; " " l:s

-T+O•Vq;+~1q;(r,O,t)=-
4
l dOq;(r,O,t)+
", ,. , S (r, t)
4'1T . (4-99)
V ~ '1T ~
However even this equation is extremely difficult to solve in general.
3. OTHER SIMPLIFICATIONS
Thus far we have mutilated the energy and angular dependence of the
transport equation in the interest of mathematical expediency-and still have not
arrived at anything we can hope to solve (at least analytically). So in frustration we
now turn our attention to the remaining time and spatial variables. First we will
completely eliminate the time variable by agreeing to consider only steady-state
transport problems. Then Eq. (4-99) simplifies to
"
O·Vcp+~ 1cp(r,O)= -4
,.. ~s l dO'q;(r,O')+
" ,. - 4(r)- . S
(4-100)
'1T 411 '1T
Next, we will assume that the system under study has uniform composition such
that the cross sections do not depend on position. Finally we will simplify the
system geometry, for example, by considering only planar or spherical symmetry.
In the case of planar symmetry we arrive at a rather simple-looking equation
µ.
O(J)
ax +l:1qi(x,µ.)=
l; 5
T
f +1
-1
, , S ( X)
dµ. cp(x,µ. )+-2-. (4-101)
This equation can actually be solvedanaJytically 6 -but only with rather sophisti-
cated mathematical techniques beyond the scope of this text. So even after a
number of rather questionable approximations, one arrives at an equation that can
still only be solved with great difficulty.
After this rather pessimistic glance at the difficulties involved in solving the
transport equation, let us remark that there is one very important class of transport
problems that can be solved exactly with only a minimal expenditure of effort-
those involving neutron transport in a purely absorbing medium.
4. NEUTRON TRANSPORT IN A PURELY ABSORBING MEDIUM
Frequently we are interested in neutron transport in a medium in which
scattering can be ignored. This might occur in a vacuum, for example (or more
realistically, a gas-filled region of a reactor). Or it might apply in a very highly
absorbing medium such as a fuel element or a control rod. In these cases, the
transport equation becomes
(4-102)
We have omitted the time dependence here since it is rarely of relevance when such
transport problems are of interest.
This equation can be solved exactly for any source distribution since it can be
converted into a simple first-order differential equation. Consider first the case of
neutron transport in a vacuum in which ~a=O:
0 ·V<p(r,E, 0) = s(r,E, 0). (4-103)

However O·V is just the directional derivative in the direction 0. If we define a
variable R that measures distance along this direction (see Figure 4-11) then
A a
o-v~-- (4-104)
aR'
where R is measured in the - 0 direction, and we find
O(f) A
oR = -s(r,E,0). (4-105)
If we now integrate with respect to this variable, we find
(4-106)
Notice that this expression simply equates the neutron angular flux at position r in
direction O to the total number of source neutrons emitted in this O (obtained by
integrating back along - 0).
EXAMPLE: Consider an isotropic point source located at the origin (for con-
venience, we will suppress the energy dependence for this example). The math-
s (r, E,_0_)..-rr,rr;,,,.
FIGURE 4-11. Neutron transport In a vacuum.

ematical representation for such a source is just
(4-107)
where 8 (r- r') is just the three-dimensional version of the Dirac 8-function (see
Appendix C) defined by
f d r'f(r')8 (r-r') = f(r).

3 (4-108)
Hence we find
,.
q:i(r, 0) = S0
ioo dR 8(r-RO) (4-109)
0 4 7T •
This still looks a bit strange. [Actually it can be shown that q:i(r, Sl) vanishes unless
one is looking along the direction r, as one would expect from Figure 4-12.]
Suppose we compute instead the neutron flux itself:
( " ,.. So (oo dR )A( d08(r-RO).

" ,.
<J>(r)= J,. dOq:i(r,0)=
4w
47T JnO 4w
(4-110)
If we multiply the integrand by R 2 / R 2 , we can identify a volume integration (in

spherical coordinates) over the dummy vector R:
S
<J>(r)=~
4w
1 1 ,. 0
00
R 2 dR
4w
8(r-R)
dO---=~
R2
S
4w
f 8(r-R)
d 3R - - - .
R2
(4-111)
Yet using the definition of 8 (r- r') given by Eq. (4-108), we find that the flux
resulting from an isotropic point source at the origin is just
(4-112)
that is, <P falls off with distance as 1/ r 2 because of the ever increasing surface area
(4wr 2) over which the S0 source neutrons/sec must be isotropically distributed.
I
I
I
I
I FIGURE 4-12. A point source emitting neutrons
I isotropically at the origin of an infinite medium.
Now return to Eq. (4-102), which characterizes transport in a purely absorbing

medium. Suppose this medium is homogeneous, that is, ~a is not a function of
position. Then we can use an integrating factor exp[~8 r·&l] to rewrite Eq. (4-102) as
(4-113)
However we can integrate this equation just as we did Eq. (4-103) to find
or canceling out the integrating factor from both sides
(4-115)
This solution again has a very plausible interpretation when it is recognized that
exp(- ~aR) is just the attenuation that would occur between the source point and
the observation point r. Notice that this immediately reduces to our vacuum result
for the case in which ~a= 0.
We can also obtain an exact solution for the situation in which ~a depends on
position. We need only use a slightly more complicated attenuation factor
(4-116)
where a{r,r') is known as the optical thickness or optical depth of the media and is
defined by
(4-117)
Note that a is essentially a measure of the effective absorption between points r'
and r.
EXAMPLE: Consider once again our point source, only this time assume that it
is imbedded at the origin of an infinitely large medium characterized by a uniform
absorption cross section ~a• Then repeating our earlier analysis using Eq. (4-115)
yields
(4-118)
which is similar to our vacuum result, with the exception of an additional attenua-
tion factor exp( - l:ar) due to the absorption.
This very important result can be easily generalized to the situation in which the
source is located at an arbitrary point r' by merely shifting the coordinate system
origin to find ·
(4-119)
•Field point
FIGURE 4-13. Coordinates characterizing a disturbed source s(r,E,0).
We can finally use this result to synthesize the neutron flux resulting from an
arbitrary distribution of isotopic sources, S (r), in an infinite absorbing medium as
q,(r) = f d 3r'
exp(-~ lr-r'I)
a
4?Tlr-r'l2
S (r'). (4-120)
C. The One-Speed Diffusion Equation

We now turn our attention toward the development of an approximate
description of neutron transport more amenable to calculation than the neutron
transport equation itself. To make life simple, we will first work within the
one-speed approximation represented by Eq. (4-96). Let us first note the explicit
forms taken by the neutron conservation equation and the corresponding equation
for the current density Jin the one-speed case:
1 aq,
-V ~
ut
+ V · J + ~t'l>(r, t) = ~ (/>(r, t) + S (r, t),
8 (4-121)
(4-122)
Here we have noted explicitly the simplifications that occur in the inscattering term
when the one-speed approximation is introduced. More specifically,
(4-123)
and
(4-124)
But
(4-125)
where we have defined the average scattering angle cosine µ.0 as
(4-126)
As an aside, it should be noted that one can easily calculate 'fi.o for the case of
elastic scattering from stationary nuclei whens-wave scattering is present. For then
we know that in the CM system, ocM(Oc)= o8 /4'1T. Hence if we use ocM(Oc)dOc
= oL(OJdDL, we find
(4-127)
But recall
l+AcosOc
COS 0L = -r----_-_-_:::-..::;-..:::-.::::.::::.::::.:::.:::::: (4-128)
VA 2 +2A cos0c+ 1
If we substitute this into Eq. (4-127) and perform the integration, we find the very
simple result
(4-129)
Now that we have justified the forms of Eqs. (4-121) and (4-122) let us consider
how we might eliminate the annoying appearance of the third unknown,
a,
f dO OOcp(r, t). We will accomplish this by assuming that the angular flux is only
weakly dependent on angle. To be more specific, we will expand the angular flux in
angle as
cp(r, D, t) = (l)o(r, t) + (1)1x(r, t)Ox + cply(r, t)Oy + q:,lz(r, t)Oz + · · · , (4-130)
and neglect all terms of higher than linear order in 0. Actually a slightly different
notation for tlie unknown functions cp0 , q:,1x, cp 1y, q:, 1z is useful. Write Eq. (4-130) as
A 1 3
cp(r,O,t) =:: 477 ct>+ 477 [JXOX +Jyny +JZOZ]
1 3 A
~ 4 '1T cf>(r, t) + 477 J(r, t)· O. (4-131)

Notice that we have labeled the unknown expansion coefficients as the flux and
current. That this notation is perfectly consistent can be seen by noting from Eq.
(4-131) that
dO+ -4 J(r,t)· 0A
liA3 0
i
41t
dOcp(r,O,t)=<f>(r,t)-
A A
4 '1T 77 4'11'
O=cf>(r,t),
'IT
(4-132)
and
4w
""" =<J>(r,t)-
i dOOcp(r,O,t) 4li "" + -43[ Jx (r,t) L""
QxO
'TT
dOO
4w 'TT 4w
dO
However one can easily demonstrate (see Problem 4-14) that the integral of the
product of any two components of Sl gives
i=j
i,j=x,y,z. (4-134)
i=/= j
Hence
i4w
dO Slcp(r,Sl,t)=J(r,t). (4-135)
Of course Eqs. (4-132) and (4-135) are identical to our original definitions of the
flux and current earlier in Eqs. (4-17) and (4-19).
We will now use the approximate form of the angular flux in Eq. (4-131) to
evaluate the second term in Eq. (4-122):
V ·£'17 dO SlSlcp(r, Sl,t) =V · £'17 dll nn[ 4~ </>+ i'TT J ·Sl l (4-136)
Next note that the integral of the product of any odd number of components of Sl
vanishes by symmetry:
1 dll
4w
Q1 gmQ 11 =0
X y Z
if l,m, or n is odd. (4-137)
If we use both Eqs. (4-134) and Eq. (4-137) we can evaluate

r A A A
V 'j,1 dO OOcp = 3 Vcf>(r, t).

1
(4-138)
4w
Hence by assuming that the angular flux depends only weakly on angle-more
specifically, that the angular flux is only linearly anisotropic-we have managed to
express the third unknown appearing in Eq. (4-122) in terms of the neutron flux
<J>(r, t). We have now achieved our goal of obtaining a closed set of two equations
for two unknowns, c/>(r,t) and J(r,t):
1 act>
~ Tt + V ·J + ~a(r)cp(r, t) = S (r, t), (4-139)
1 aJ
v Tt + 31 Vct>+~ 1r(r)J(r,t)=S 1(r,t). (4-140)
We have noted here that

(4-141)
and defined the macroscopic transport cross section
(4-142)
(We will comment on this definition in a moment.) These two equations are known
in nuclear reactor analysis as the P 1 equations (in the one-speed approximation)
since the approximation of linearly anisotropic angular dependence in Eq. (4-131)
in one-dimensional plane geometry is equivalent to expanding the angular flux in
Legendre polynomials in µ. = cos 9 and retaining only the / = 0 and / = 1 terms
(4-143)
hence the name P 1 approximation. Notice that this could be easily generalized to
obtain the PN approximation.
In principle we could now use the P 1 equations to describe the distribution of
neutrons in a nuclear reactor. However it is customary to introduce two more
approximations in order to simplify these equations even further. First we will
assume that the neutron source term s(r, Sl, t) is isotropic. This implies, of course,
that the source term S 1(r, t) vanishes in the equation for the current density. As we
mentioned earlier, this approximation is usually of reasonable validity in nuclear
reactor studies.
As our second approximation, we will assume that we can neglect the time
derivative v- 1 aJ/at in comparison with the remaining terms in Eq. (4-140). This
would imply, for example, that
(4-144)
that is, that the rate of time variation of the current density is much slower than the
collision frequency vl': 1• Since vl': 1 is typically of order 1<>5 sec- 1 or larger, only an
extremely rapid time variation of the current would invalidate this assumption. We
will later find that such rapid changes are very rarely encountered in reactor
dynamics. Hence we are justified in rewriting Eq. (4-140) as
(4-145)
We can solve Eq. (4-145) for the neutron current density in terms of the neutron
flux
1
J(r, t) = - Jl':tr(r) Vcf>(r, t). (4-146)
If we define the neutron diffusion coefficient D by
(4-147)
then we can rewrite Eq. (4-146) as
J(r, t) = - D (r)Vcf>(r, t). (4-148)
Hence we have found that in certain situations the neutron current density is
proportional to the spatial gradient of the flux. This very important relation arises
quite frequently in other areas of physics where it is known as Fick's law. It is also
occasionally referred to as the diffusion approximation.
Before we consider the physical implications of this relationship, let us use it to
simplify the P 1 equations. If we substitute this into Eq. (4-139) we find
1 at/>
- -0 -V •D (r)Vcp+ ~a(r)cp(r, t) = S (r, t). (4-149)
V t
This very important equation is known as the one-speed neutron diffusion equation,
and it will play an extremely significant role in our further studies of nuclear
reactors. We will discuss its solution in considerable detail in Chapter 5, and we
will use it as the basis of a very simple but very useful model of nuclear reactor
behavior.
Let us now return to consider the diffusion approximation [Eq. (4-148)] in more
detail. Notice that it implies that a spatial variation in the neutron flux (or density)
will give rise to a current of neutrons flowing from regions of high to low density.
Physically this is understandable since the collision rate in high neutron density
regions will be higher with the corresponding tendency for neutrons to scatter more
frequently away toward lower densities. The rate at which such diffusion occurs
<Ji(y)
J(y)
y
FIGURE 4-14. A schematic representation of Flck's law.
depends on the diffusion coefficient which, in turn, is inversely proportional to the

transport cross section ~tr·
It is convenient to introduce the concept of a transport mean free path
(4-150)
The transport mfp can be regarded as a corrected mfp accounting for anisotropies
in the scattering collision process. Since µ.0 is almost always positive-that is, biased
in the direction of forward scattering-the transport mfp Atr will always be
somewhat larger than the actual mfp, A= (~J- 1• This essentially accounts for the
fact that neutrons experiencing forward scattering tend to be transported somewhat
further in a sequence of collisions than those being isotropically (or backward)
scattered.
Hence since D = Atrf 3, we see that diffusion is also enhanced in a material with
pronounced forward scattering (e.g., hydrogen), although this of course also de-
pends on the magnitude of the macroscopic cross sections.
It is important to keep in mind that the diffusion approximation is actually a
consequence of four different approximations: (a) the angular flux can be
adequately represented by orJy a linearly anisotropic angular dependence [Eq.
(4-131)], (b) the one-speed approximation, (c) isotropic sources, and (d) the neutron
current density changes slowly on a time scale compared to the mean collision time
[Eq. (4-144)). Actually only the first of these approximations is really crucial. The
remaining approximations can be relaxed provided we are willing to work with the
P 1 equations rather than the neutron diffusion equations (as one frequently is).
It is natural to ask when the ang'ijlar flux is sufficiently weakly dependent on
angle so that the diffusion approximation is valid. More detailed studies of the
transport equation itself indicate that the assumption of weak angular dependence
is violated in ·the following cases: (a) near boundaries or where material properties
change dramatically from point to point over distances comparable to a mean free
path, (b) near localized sources, and (c) in strongly absorbing media. In fact strong
angular dependence can be associated with neutron fluxes having a strong spatial
variation. Usually if one is over several mean free paths from any sources or
boundaries in a weakly absorbing medium, the flux is slowly varying in space, and
diffusion theory is valid.
D. The Energy-Dependent Diffusion Equation

Let us now try to repeat this analysis for the case in which the neutron energy
dependence is retained. Again we will approximate
A 1 3 A
cp(r,E,0,t)a;; 4 '1T cJ>(r,E,t)+ 4 '1T J(r,E,t)·O, (4-151)
so that we can evaluate
V ·l
4,r
dO OO(l)(r,E,O,t)a;;½vcJ>(r,E,t). (4-152)
If we now use this in the pair of Eqs. (4-79) and (4-91), we arrive immediately at
the energy-dependent P1 equations
!v : +V•J+~i(r,E)<f>(r,E,t)== ( 00dE'~s(E'➔E)<f>(r,E',t)+S(r,E,t), (4-153)

vt Jo
!v a/+
vt
-31 V<t>+~t(r,E)J(r,E,t)== Jf
0
00
dE'~s (E'➔E)J(r,E',t)+S 1 (r,E,t).
1
(4-154)
Continuing our analogy with the derivation of the one-speed diffusion equation, we
will again assume: (a) isotropic source S 1 _Q (b) IJl- 1oIJI/ ot«v~tCr,E) so that Eq.
(4-154) can be rewritten as
However we can now see a problem that appears to prevent a straightforward

generalization of Fick's law to include energy dependence. For we cannot bring
J(r,E', t) out of the scattering integral since it depends on the integration variable
E'.
Of course, if we were allowed to assume isotropic scattering in the LAB system,
then ~s/E' ➔ E)=O and we could find
(4-156)
But the assumption of isotropic scattering is far too gross for most reactor
calculations.
We could proceed formally by merely defining an energy-dependent diffusion
coefficient
(4-157)
which would automatically yield
J(r,E,t)= -D (r,E)V<f>(r,E,t). (4-158)
Of course, this approach is highly artificial because as defined in Eq. (4-157)

D (r, E) still depends on J(r, E, t).
One common procedure for avoiding this difficulty is to neglect the anisotropic
contribution to energy transfer in a scattering collision by setting
(4-159)
so that
(4-160)
Then we find a natural generalization of the diffusion coefficient:
(4-161)
Actually none of these derivations of an energy-dependent diffusion equation are

particularly satisfying because we have ignored the fact that in neutron transport
processes, spatial transport, directional changes, and energy changes are intimately
mixed. We can only provide a more satisfactory derivation of Eq. (4-158) after we
have discussed comparable approximations characterizing neutron energy transfer
in scattering collisions. Such a derivation must await our discussion of neutron
slowing down in Chapter 8.
For the present, we will simply assume that the generalization of Fick's law to
include energy-dependence is given by Eq. (4-158) with an energy dependent
diffusion coefficient as defined by Eq. (4-161). If we now substitute this into Eq.
(4-153), we arrive at the energy-dependent diffusion equation
1 a</>
- T -V ·D (r,E)Vc/>+~i(r,E)</>(r,E,t)
V ut
(4-162)
This equation plays a very important role in nuclear reactor analysis since it is
frequently taken as the starting point for the derivation of the multigroup diffusion
equations. These latter equations represent the fundamental tool used in modern
nuclear reactor analysis.
E. Diffusion Theory Boundary Conditions

Since the neutron diffusion equation has derivatives in both space and time, it
is apparent that one must assign suitable boundary and initial conditions to
complete the specification of any particular problem. Since the diffusion equation
itself is only an approximation to the more exact transport equation, we might
suspect that we can use the transport theory boundary conditions as a guide in our
development of appropriate diffusion boundary conditions. It will suffice to con-
sider this development within the one-speed approximation.
Recall that the transport theory boundary conditions we discussed earlier were:
Initial condition: <p(r, 0, 0) = CJ>o(r, 0), (4-163)
Boundary condition: <p(rs,O,t)=O for O·es<O, all rs on s. (4-164)
We can obtain the appropriate initial condition for the diffusion equation by
merely integrating the transport condition over angle to obtain:
Initial condition: <f>(r, 0) = </>0(r) (4-165)
The boundary conditions are a bit harder to come by. Actually we will require
several types of boundary condition, depending on the particular physical problem
of interest. We will group these boundary conditions into one of several classes:
1. GENERAL MATHEMATICAL CONDITIONS ON THE FLUX
Although strictly speaking they are not boundary conditions, we should first
mention those mathematical properties that the function cp(r, t) must exhibit in
order to represent a physically realizable neutron flux. For example, cp(r, t) must be
a real function. Furthermore since both the neutron speed v and density N cannot
be negative, we must require that cp(r, t) be greater than or equal to zero. In most
cases we can also require that cp(r, t) be bounded. However we should add here that
one occasionally encounters pathological models of physical neutron sources that
cause cp(r, t) to diverge. An example would be the familiar point source. Another
situation is the so-called Milne problem, in which one studies the behavior of the
flux near a vacuum boundary fed by a source of infinite magnitude located at
infinity. There will also be certain symmetry conditions that we can place on cp(r, t)
resulting from geometrical considerations, such as plane or axial symmetry. These
conditions will become more understandable as we consider specific examples later
in Chapter 5.
2. BOUNDARIES AT INTERFACES
Consider next an interface between two regions of differing cross sections.
Now clearly the correct transport boundary condition is that
(4-166)
where 'Pi is the angular flux in region 1, while rp 2 is the angular flux in region 2.
This condition ensures conservation of neutrons across the boundary and can
easily be derived directly from the transport equation.
Unfortunately we cannot satisfy this boundary condition exactly using diffusion
theory. At best, we can only ensure that angular moments of Eq. (4-166) are
satisfied. And since diffusion theory yields only the first two moments of the
angular flux, cp(r, t) and J(r, t), at best we can demand
(4-167)
FIGURE 4-15. An interface boundary.

142 / mE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
and
(4-168)
Hence the interface diffusion theory boundary conditions are simply those corre-
sponding to continuity of flux and current density across the interface:
cf, 1(rs, t) =q,i(rs, t)
- D 1Vct, 1(rs, t) = - D 2 V q,i{r8 , t). (4-169)
We will occasionally find it mathematically expedient to imagine an infinitestimally

thin source of neutrons S at an interface boundary. Then the interface boundary
conditions are modified to read:
cf, 1(r8 , t) =<J>i{r5 , t)
es ·Ji(rs, t)-es ·J1(rs, t) = S ( t), (4-170)
where es is the unit normal to the surface.

3. VACUUM BOUNDARIES
Recall that our transport theory boundary condition was merely a mathemati-
cal statement that there could be no incoming neutrons at a free or vacuum
boundary
for !l·dS<O,
(4-171)
all r5 on S.
FIGURE 4-16. A vacuum boundary.

Once again diffusion theory will only be able to approximate this boundary
condition. Notice in particular that the boundary condition is given only over half
of the range of solid angle (corresponding to incoming neutrons). Hence suppose
we again seek to satisfy the transport boundary condition in an "integral" sense by
demanding
(4-172)
where we have recognized that this integral condition is equivalent to demanding

that the inwardly directed partial current J _ vanish on the boundary.
Unfortunately diffusion theory is capable of only approximating even this
integral condition, since it cannot yield the exact form for J ±. Indeed if we use the
P 1 approximation [(4-131)] for the angular flux, we find that the partial current
densities J ± are approximated in diffusion theory by
(4-173)
Hence our diffusion theory approximation to the transport boundary condition

Eq. (4-171) is just
(4-174)
For convenience consider this boundary condition applied to a one-dimensional

geometry with the boundary at x = xs
or
I
=-- (4-175)
2D
Notice that this relation implies that if we "extrapolated" the flux linearly beyond
the boundary, it would vanish at a point
(4-176)
For this reason, one frequently replaces the vacuum boundary condition
(4-177)
by the slightly simpler condition
(4-178)
where X5 is referred to as the "extrapolated" boundary. More advanced transport

theory calculations of the extrapolated boundary indicate that one should choose
(4-179)
//>diffusion theory
',
'
__ ~ ' - , - · <Pexact · - · -
'-
',,;
' - , Linear extrapolation
XS
~
Xs
X
,___ _ _ zo----;:,>1
FIGURE 4-17. The behavior of approximate and exact representations of the neutron flux
near a vacuum boundary.
where the "extrapolation length" z0 for plane geometries is given by
(4-180)
More complicated extrapolation length formulas can be derived for curved

boundaries, 14 although they are rarely necessary since Eq. (4- 180) suffices unless
the radius of curvature of the boundary is comparable to a mean free path.
It should be remembered that the true flux does not vanish even outside the
boundary. The diffusion theory flux is a poor representation of the true flux near
the boundary (as we saw earlier, diffusion theory is not valid near a boundary).
The boundary conditions we have derived are intended to yield the proper flux
only in the interior of the reactor, that is, several mean free paths away from the
reactor boundary.
These boundary conditions complete our description of neutron transport within

the diffusion approximation. The neutron diffusion equation will play a very
fundamental role in our development of nuclear reactor analysis methods. We will
begin our study of nuclear reactor behavior using one-speed diffusion theory, since
while this description has only limited quantitative validity, it does allow us to
illustrate rather easily the principal concepts of nuclear reactor theory, as well as to
develop the mathematical techniques used in more sophisticated models. With this
background, we will then develop and apply the principal tool of the nuclear
reactor designer, multigroup diffusion theory.
REFERENCES
1. W. G. Vincenti and C. H. Kruger, Jr., Introduction to Physical Gas Dynamics, Wiley,

New York (1965); K. Huang, Statistical Mechanics, Wiley, New York (1963).
2. P. F. Zweifel, Reactor Physics, McGraw-Hill, New York (1973).
4. G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand, Princeton, N. J.
(1970).
5. K. M. Case, F. de Hoffmann, and G. Placzek, Introduction to the Theory of Neutron
Diffusion, Vol. I, Los Alamos Scientific Laboratory Report (1953).
6. K. M. Case and P. F. Zweifel, Linear Transport Theory, Addison-Wesley, Reading,
Mass., (1967); B. Davison, Neutron Transport Theory, Oxford University Press, (1958).
7. E. Isaacson and H. B. Keller, Analysis of Numerical Methods, Wiley, New York (1966).
8. B. Carnahan, H. A. Luther, and J. 0. Wilkes, Applied Numerical Methods, Wiley, New
York (1969).
9. W. M. Stacey, Jr., Modal Approximations: Theory and an Application to Reactor Physics,
M.I.T. Press, Cambridge (1967).
IO. M. Becker, The Principles and Applications of Variational Methods, M.I.T. Press,
Cambridge (1964).
11. G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand, Princeton, N. J.,
(1970), pp. 214-249.
12. L. I. Schiff, Quantum Mechanics, 3rd Edition, McGraw-Hill, New York (1968).
13. A. Messiah, Quantum Mechanics, Vol. I., North-Holland, Amsterdam (1958).
14. Reactor Physics Constants, USAEC Document ANL-5800, 2nd Edition (1963).
15. K. D. Lathrop, Reactor Technol. 15, 107 (1972).
PROBLEMS
4-1 We have defined the angular neutron density n(r,E,0,t) in terms of the neutron
energy E and the direction of motion 0, but one could as well define an angular
density that depends instead on the neutron velocity v, n(r, v, t). Calculate the
relationship between these two dependent variables.
4-2 Two thermal neutron beams are injected from opposite directions into a thin sample
of 235 U. At a given point in the sample, the beam intensities are l0 12 neutrons/cm2 •
sec from the left and 2X 10 12 neutrons/cm2 ·sec from the right. Compute: (a) the
neutron flux and current density at this point and (b) the fission reaction rate density
at this point.
4-3 Suppose that the angular neutron density is given by
no
n(r,O)= 4 '11' (1-cosO),
where O is the angle between O and the z-axis. If A is the area perpendicular to the
z-axis, then what is the number of neutrons passing through the area A per second:
(a) per unit solid angle at an angle of 45° with the z-axis, (b) from the negative z to
the positive z direction, (c) net, and (d) total?
4-4 In a spherical thermal reactor of radius R, it is found that the angular neutron flux
can be roughly described by
cf>o { E ) sin( 'TTY/ R )
q,(r,E,0)= 4 '11'Eexp - kT r .
Compute the total number of neutrons in the reactor.

4-5 Demonstrate that in an isotropic flux, the partial current density in any direction is
given by J + -<f>/4.
4-6 One of the principal assumptions made in the derivation of the neutron transport
equation is that there are no external forces acting on the neutrons, that is, that the
neutrons stream freely between collisions. Suppose we were to relax this assumption
and consider such a force acting on the neutrons (say, a constant gravitational force
in the z-direction). Then we might expect additional terms to appear that involve this
force. Derive this more general transport equation.
4-7 Explain briefly whether or not the transport equation as we have derived it adequately
describes the spatial and angular distribution of neutrons: (a) in a flux of the order of
100/cm2 ·sec, (b) scattering in a single crystal, (c) passing through a thin pure
absorber, and (d) originating within a very intense nuclear explosion in which the
explosion debris are moving outward with very high velocities.
4-8 Develop the particular form of the transport equation in spherical and cylindrical
geometries. To simplify this calculation, utilize the one-speed form of the transport
equation (4-100) in which time dependence has been ignored.
4-9 An interesting model of neutron transport involves transport in a one-dimensional
rod. That is, the neutrons can move only to the left [say, described by an angular
density n_(x,E,t)] or to the right [n+(x,E,t)]. One need only consider forward
scattering events described by 'l:.:(E'-E) or backward scattering events described by
'l:. 8-(E'-E). Perform the following: (a) derive the transport equations for n+ and n_,
(b) make the one-speed approximation in this set of equations, that is,
'l:.t(E) ='l:.a(E)+'l:.:(E) + 'l:.;(E) ='l:.a +'l:.: + 'l:.;,
'l:.f(E'-E)='l:.:8(E'-E),
and (c) describe the boundary and initial conditions necessary to complete the
specification of the problem if the rod is characterized by a length L.
4-10 Consider the following differential equation:
d2j
- 2 +x3/(x)=2x(4-x),
dx
where f(x) is defined on the interval O,;;;; x,;;;; 4. Discretize this equation by first
breaking up the independent variable range into four segments of equal length. Next
use a finite-difference approximation to the d3// dx 2 term to rewrite the differential
equation as a set of algebraic equations for the discretized unknown f(x;)= J;. For
convenience, assume boundary conditions such that /(0)=0= /(4). Solve this set of
equations for the J; and then plot this solution against x using straight-line interpola-
tion.
4-11 Consider the steady-state one-speed transport equation assuming isotropic scattering
and sources in a one-dimensional plane geometry
Expand the solution to this equation in the first two Legendre polynomials
cp(x, µ.).;;: cp0 (x)½P0 ( µ,) + cp 1(x)½ Pi ( µ,),
Po(µ.)= 1, P1( µ,) = µ,.
Substitute this expansion into the transport equation, multiply by Po(µ,) and P 1( µ.)
respectively, and integrate over µ. to obtain a set of equations for the unknown
expansion coefficients cp0(x) and cp 1(x). (These are just the P 1 equations.)
4-12 Use Simpson's rule to write a numerical quadrature formula for the angular integral
f ~}dµ.q,(x,µ.) for N equal mesh intervals.
4-13 Develop the multigroup form of the transport equation as follows: First break the
energy range 0..;; E..;; lO MeV into G intervals or groups. Now integrate each of the
terms in the transport equation, Eq. (4-43), over the energies in a given group, say
E 8 ..;; E..;; E 8 _ 1• (Remember that the group indexing runs backwards such that 0= EG
< EG-1 < · · · E8 < E8 _ 1 < · · · E 1 < E 0 = lO MeV.) Now by defining the group fluxes as
the integral of the flux over each group, and the cross sections characterizing each
group as in Eq. (4-69), determine the set of G equations representing the transport
equation.
4-14 By writing out the components of the direction unit vector n in polar coordinates,
demonstrate explicitly that
i=j
i¥=}
4-15 Demonstrate that f 4 .,.dOO~O;'~=O if l, m, or n is odd.

4-16 Verify Eq. (4-138) using the identity in Problem 4-15.
4-17 Explicitly demonstrate by integration that jjo=2/3A for elastic scattering from
stationary nuclei when such scattering is assumed to be isotropic in the CM system.
4-18 Consider an isotropic point source emitting S 0 monoenergetic neutrons per second in
an infinite medium. Assume that the medium is characterized by an absorption cross
section l:a, but only by negligible scattering. Determine the rate at which neutron
absorptions occur per unit volume at any point in the medium.
4-19 Compute the neutron flux resulting from a plane source emitting neutrons isotropi-
cally at the origin of an infinite absorbing medium. Hint: Just represent the plane
source as a superposition of point sources.
4-20 In a laser-induced thermonuclear fusion reaction, a tiny pellet is imploded to super
high densities such that it ignites in a thermonuclear burn. In such a reaction some
10 17 14 MeV neutrons will be emitted essentially instantaneously (within 10- 11 sec).
Compute the neutron flux at a distance of 1 m from the reaction as a function of time,
assuming that the chamber in which the reaction occurs is evacuated.
4-23 Compute the thermal neutron diffusion coefficients characterizing water, graphite,
and natural uranium. Then compute the extrapolation length z 0 characterizing these
materials.
4-24 Assuming·that the diffusion approximation [Eq. (4-131)) is valid, compute the partial
current densities in the z direction defined by Eq. (4-22). Use the one-speed
approximation.
4-25 Consider the time-dependent one-speed P 1 equations assuming isotropic sources and
plane symmetry. Eliminate the current density J(x,t) to obtain one equation for the
neutron flux ,p(x, t). Compare this equation with the one-speed neutron diffusion
equation and indicate what differences you might expect in solutions to the two
equations.
4-26 Try to construct solutions to the one-speed transport equation in an infinite sourceless
medium:
by seeking solutions of the form qJ(x,µ.)=x(µ.) exp(-x/11) where,, and x(µ,) are to be
determined.
4-27 Consider the time-independent one-speed transport equation uader the assumption of
isotropic sources and scattering
By regarding the right-hand side of this equation as an effective source, use the result
[Eq. (4-120)] to derive an integral equation for the neutron flux q>(r).
5
The One-Speed
Diffusion Theory Model
In this chapter we will develop the one-speed diffusion model of neutron

transport. This model plays an extremely important role in reactor theory since it is
sufficiently simple to allow detailed calculations and also sufficiently realistic to
allow us to study many of the more important concepts arising in nuclear reactor
analysis. Of course any model characterizing all of the neutrons in a reactor by a
single speed (or energy) and treating their transport from point to point as a
diffusion process cannot be expected to yield accurate quantitative estimates.
Nevertheless if the cross sections appearing in this theory are properly chosen, one
can use the one-speed diffusion model to make preliminary design estimates.
Moreover many of the mathematical techniques we will use to solve and analyze
this model are in fact identical to those applied to the more sophisticated models
(e.g., multigroup diffusion theory) used in modern nuclear reactor design.
The rigorous mathematical derivation of the one-speed diffusion model from the
neutron transport equation has been given in Chapter 4. In this chapter we will give
a more heuristic physical derivation of the one-speed diffusion equation and then
apply this equation to study nuclear reactor behavior. The solution of the diffusion
equation draws upon many familiar topics from a field that has become known as
mathematical physics, including methods for solving boundary value problems
involving both ordinary and partial differential equations. 1- 3 For most problems of
practical interest in nuclear reactor studies, one must employ methods from
numerical analysis as well to allow the solution of the diffusion equation on a
high-speed computer. We would expect that many of these topics (e.g., separation
of variables methods in the solution of partial differential equations or Gaussian
elimination for solving systems of algebraic equations) are already quite familiar to
the advanced undergraduate. However we will continue our effort to make this
149
presentation as self-contained as possible by briefly reviewing such methods when

they arise (although numerous references will also be provided). Throughout this
development we would again caution the reader to avoid letting the smokescreen
generated by these various mathematical techniques obscure the fundamental
physical concepts governing the behavior of the neutron population in the reactor.
I. THE ONE-SPEED DIFFUSION EQUATION
A. Derivation of the Diffusion Equation

We will suppress the neutron energy dependence by assuming that all of the
neutrons can be characterized by a single kinetic energy. Such a one-speed (or
one-group) approximation greatly simplifies the mathematical study of nuclear
reactor behavior. Of course such an approximation is also highly suspect, particu-
larly in light of the fact that neutron energies typically encountered in a reactor
span a range from 10- 3 to 107 eV, and neutron cross sections depend sensitively on
energy over most of this range. We will later be able to show, surprisingly enough,
that if one regards the one-speed approximation as an average description and
chooses the appropriate average cross sections, then in fact the one-speed model
can actually be used to obtain a quantitative description of a nuclear reactor.
As yet a further justification for our exhaustive study of the one-speed approxi-
mation, we would remark that most energy-dependent theories (e.g., multigroup
diffusion theory) are solved by performing a sequence of one-speed calculations for
each successive energy group. Hence the methods we develop for analyzing our
one-speed model will later be extended directly to more sophisticated descriptions.
We will characterize the neutron distribution in the reactor by the neutron
density N (r, t) which gives the number of neutrons per unit volume at a position r
at time t. Actually we will find it more convenient to work with the neutron flux,
t/>(r, t) = vN (r, t), since then we can compute the rate at which various types of
neutron-nuclear reactions occur per unit volume by merely multiplying the flux by
the corresponding macroscopic cross section. For example, the rate at which fission
reactions occur per cm3 at a point r would be given by ~r(r)cf>(r,t).
We will derive an equation for the neutron flux by merely writing down a
mathematical statement of the fact that the time rate of change of the number of
neutrons in a given volume must be simply the difference between the rate at which
neutrons are produced in the volume and the rate at which they are lost from the
volume due to absorption or leakage.
dS
FIGURE 5-1. A control volume for monitoring neutron

balance.
THE ONE-SPEED DIFFUSION THEORY MODEL / 151
To be more precise, consider an arbitrary volume V of surface area S located

anywhere within the reactor. We will examine this "control" volume carefully to
determine how the neutron population within it changes. Evidently the total
number of neutrons in V at a time t can be obtained by simply integrating over the
volume
(5-1)
Hence the tim,e rate of change of the number of neutrons in V must be just
-d [
dt
Jd3
v
r-cp(r,t
1
v
>] = J
v
d 1
3r -
v
-acf>
at
= Production in V - absorption in V
- net leakage from V. (5-2)
We can easily write down mathematical expressions for the gain and loss terms. If
we define a neutron source density S (r, t), then
Production in V= fvd 3rS(r,t). (5-3)
Since the absorption rate density at any point in V is just ~a(r) cf>(r, t), it is· obvious
that the total rate of neutron loss due to absorption in V is just
Absorption in V = J:,d r ~a(r) cf>(r, t).

3 (5-4)
The term describing neutron leakage out of or into Vis a bit more difficult. If
J(r, t) is the neutron current density, then the net rate at which neutrons pass out
through a small surface element dS at position r5 is J(rs, t) ·dS. Hence the total net
leakage through the surface of V is just
Net leakage from V= fsdS·J(r,t). (5-5)
Now we could combine all of these terms back into Eq. (5-2) as they stand, but first
it is convenient to convert Eq. (5-5) into a volume integral similar to the other
terms. The common way to convert such surface integrals into volume integrals is
to use Gauss's theorem to write
(5-6)
If we now substitute each of these mathematical expressions into Eq. (5-2), we find
l V
d 3r [ -1 -acf> - S + ~ cf>+ V · J ] = 0.
V at a
(5-7)
IS2 / THE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
But recall that we chose our volume .V to be arbitrary. That is, Eq. (5-7) must hold
for any volume V that we would care to examine. However the only way this can
occur is if the integrand itself were to vanish. Hence we find we must require
I o<f>
-V -ot = -V·J-~a<t>+S. (5-8)
Of course this equation is still quite exact (aside from the one-speed approxima-
tion), but it is also quite formal since it contains two unknowns, <[>(r, t) and J(r, t).
To proceed further, we need yet another relationship between these two variables.
Unfortunately there is no exact relationship between <[>(r, t) and J(r, t). One must
resort to an approximate relation. Now it is well known from other fields of physics
such as gaseous diffusion that the current density is approximately proportional to
the negative spatial gradient of the density, or in our case, the flux. That is,
particles will tend to flow from regions of high to low density at a rate proportional
to the negative density gradient. Stated mathematically, one finds that
J(r, t):::::: - D (r)V<f>(r, t), (5-9)
where the constant of proportionality D (r) is known as the diffusion coefficient,

while Eq. (5-9) is referred to as Fick's law. Of course to postulate such a
relationship between current density and flux implies nothing about its range of
validity. Indeed we do not even know what the diffusion coefficient D is. This
situation is very common in macroscopic descriptions of physics, and relationships
such as Eq. (5-9) are usually referred to as "transport laws" (not to be confused
with the transport equation-we are talking the language of the physicist now)
while the proportionality coefficients are known as "transport coefficients." Ex-
amples are numerous and include Fourier's law of thermal conduction (thermal
conductivity), Stokes' law of viscosity (shear viscosity), Ohm's law (electrical resis-
tivity), to name only a few. In all cases, one is forced to go to a microscopic
description in order to evaluate the transport coefficient and examine the range of
validity of the transport law.
However this is of course exactly what we did by deriving the diffusion approxi-
mation [Eq. (5-9)J from the neutron transport equation in Chapter 4. There we
found that
(5-10)
where 'fi.o is the average cosine of the scattering angle in a neutron scattering
collision. Furthermore, we found that Eq. (5-9) was valid provided it was used to
describe the neutron flux several mfp away from the boundaries or isolated
sources, the medium was only weakly absorbing, and provided the neutron current
was changing slowly on a time scale comparable to the mean time between
neutron-nuclei collisions. It is important to keep these limitations in mind as we
apply this approximation in nuclear reactor analysis.
Henceforth we will accept the diffusion approximation [Eq. (5-9)] as providing a
valid expression for the neutron current density in terms of the neutron flux. If we
substitute Eq. (5-9) into Eq. (5-8), we arrive at the one-speed neutron diffusion
equation
1 ocp
-;;- at= V· D (r}Vrp- ~a(r)cp(r, t) + S (r, t). (5-11)
This equation will form the basis of much of our further development of nuclear
reactor theory.
B. Initial and Boundary Conditions

We must augment this equation with suitable initial and boundary conditions.
Although these conditions have been developed in a more rigorous fashion from
neutron transport theory in Chapter 4, we will remotivate them here using plausible
physical arguments.
The appropriate initial condition involves specifying the neutron flux c/>(r, t) for
all positions r at the initial time, say t = 0:
Initial condition: cp(r, 0) = c/>0 (r), all r. (5-12)
The boundary conditions are a bit more complicated and depend on the type of
physical system we are studying. The principal types of boundary conditions we
will utilize include the following:
1. VACUUM BOUNDARY
At the outside boundary of a reactor, one would like to construct a boundary

condition corresponding to the fact that no neutrons can enter the reactor through
this surface from outside. Implicit in this fact is the assumption that the reactor is
surrounded by an infinitely large vacuous region. Of course no reactor is
surrounded by a vacuum, but rather by air, concrete, and a host of other materials.
It is frequently convenient to assume that the reflection of neutrons back into the
reactor from such materials is negligible so that nonreentrant boundary conditions
apply.
There is only one problem; diffusion theory is incapable of exactly representing
a nonreentrant boundary condition. The closest one can come would be to demand
that the inwardly directed partial current
(5-13)
vanish on the boundary (clearly an approximation, since this expression for J _

already is approximate). Actually we really shouldn't worry much about a consis-
tent free surface boundary condition within the diffusion approximation, for we
have already indicated that diffusion theory is not valid near the boundary anyway.
It can only be expected to hold several mfp inside the boundary. (
Hence what we should really look for is a "fudged-up" boundary condition
which, although it may have little physical relevance at the boundary, does in fact
yield the correct neutron flux deep within the reactor where diffusion theory is
valid. More detailed transport theory studies indicate that the proper boundary
condition to choose is one in which the diffusion theory flux <[>(r, t) vanishes at a
distance
(5-14)
outside of the actual boundary of the· reactor. This extrapolated boundary is usually
denoted symbolically with a tilde. For instance, if the physical surface is at r5, then
the flux will be taken to vanish on the extrapolated boundary <[>(is, t) = 0.
A side comment here is appropriate. For most reactor materials, Atr is quite
small, usually being of the order of several centimeters or less. When it is
recognized that most reactor cores are quite large (several meters in diameter), then
it is understandable that one frequently ignores the extrapolation length z0 and
simply assumes the flux vanishes on the true boundary.
Furthermore few realistic reactor geometries are surrounded by a free surface.
Rather they are surrounded by coolant flow channels or plen1:1ms, structural
materials, thermal shields or such. Hence while the concept of a free surface is
useful pedagogically for painting a picture of an idealized reactor geometry
surrounded by a vacuum of infinite extent, it is rarely employed in modern nuclear
reactor analysis.
2. INTERFACES (MATERIAL DISCONTINUITIES)

The structure of a nuclear reactor core is extremely complex, containing
regions of fuel, structural material, coolant, control elements, and so on. Hence
while one rarely encounters situations in which the material cross sections l':(r)
depend continuously on position, one frequently is faced with what might be
termed "sectionally uniform" cross sections that change abruptly across an inter-
face separating regions of differing material composition. Our usual procedure in
treating such discontinuities in material properties will be to solve the diffusion
equation in each region separately and then attempt to match these solutions at the
interface using appropriate boundary conditions.
Once again the diffusion equation is not strictly valid within several mfp of the
interface. However in this case one can argue that conservation· of neutron
transport across the interface demands continuity of botli the normal component of
the neutron current density J(r, t) and the neutron flux cp(r, t).
This condition is occasionally modified to account for the physically fictitious
but mathematically expedient convenience of including an infinitesimally thin
neutron absorber or source at the interface. Then while the neutron flux is still
continuous across the interface, the normal component of the current experiences a
JUmp:
es. [J(r:)-J(r;)] = s, (5-15)

where e5 is the interface surface normal, while S would represent a source term (if
positive) or an absorption term (if negative). (See Figure 4-15.)
3. OTHER TYPES OF BOUNDARY CONDITIONS

It is frequently convenient to impose other types of boundary conditions
upon the neutron flux. For example, we know the flux must be nonnegative, real,
and finite. Actually in our mathematical modeling of neutron sources we will
THE ONE-SPEED DIFFUSION THEORY MODEL / 15S
occasionally encounter a situation in which the neutron flux becomes singular at a

localized source (e.g., a point source). However since such sources are mathemati-
cal idealizations, this singular behavior doesn't bother us, and in general we will
demand that ·the flux be finite away from such sources. This condition is particu-
larly useful in geometries in which certain dimensions are infinite.
We will also occasionally be able to use symmetry properties to discard physi-
cally irrelevant solutions of the diffusion equation. For example, in one-
dimensional slab geometries, we can choose the coordinate origin at the centerline
of the slab, and then use symmetry to eliminate solutions with odd parity [i.e.,
4>(- x)- -4>(x)].
Other types of boundary conditions are encountered in practice. A very common
problem in reactor calculations involves the determination of the flux in a small
subregion of the reactor fuel lattice, a so-called unit cell repeated throughout the
lattice. For example, such a cell might contain a single fuel rod surrounded by
coolant (a Jue( cell) or several fuel assemblies along with a control element (a
control cell). Since these unit cells are repeated in a regular fashion throughout the
core lattice, one can argue that there should be no net transfer of neutrons between
cells, that is, that the neutron current density J(r) vanish on the cell boundaries.
This is an example of a boundary condition on the current. In such cell calcula-
tions it is also frequently necessary to obtain a diffusion theory solution in the
vicinity of a strong absorber (e.g., a fuel rod or control element). The appropriate
boundary condition at the interface between the diffusing medium and the ab-
sorber is handled much like that characterizing a free surface. That is, one uses a
transport-corrected boundary condition on the flux or current to yield the proper
diffusion theory solution at a distance of several mfp from the interface. We will
Fuel element
.£c. Fuel element
0 00000
0
0 00 0
0
0 ~~.,!/Jf'p 0
0
0
a.......,,..,~ule,_,....,.,,Q
0 0 0
0
e=,3 E"'° 0 0
0
0
0>'"""'11[\~~
$fl!"\\\ 0 gg
0
0 il ,~\ '\, 0
0
0 0
0 0 G
Fuel cell Control cell

.FIGURE S-2. Typical unit fuel and control cells.
give explicit examples of such boundary conditions when we consider cell calcula-
tions in Chapter 10.
We should reemphasize that the application of neutron diffusion theory in
reactor analysis is successful in large part because the diffusion equation and its
boundary conditions have been modified using more exact transport theory correc-
tions. For example, D = [3(~t - ii~s)r 1 contains a correction to account for
anisotropic scattering. Furthermore the boundary conditions on the flux or current
at a free surface or adjacent to a highly absorbing region contain transport
corrections to yield the proper neutron flux deeper within the diffusing region.
Such transport corrections frequently yield diffusion theory estimates that are far
more accurate than one would normally expect, especially when we recall the
rather strong approximations required to derive the neutron diffusion equation.
C. A Summary of the One-Speed Diffusion Model

To summarize then, the model we will initially use to describe the neutron
population in a nuclear reactor consists of the neutron diffusion equation:
1 ac[>
- T -V ·D (r)Vcf> + ~a(r)cf>(r, t) = S (r, t) (5-16)
V ut
along with suitable initial conditions:
cf>(r,O) =cf>0 (r), all r (5-17)
and boundary conditions:

(a) Free surface: c[>(rs, t) = 0 [or J _(r5, t) = O]
(b) Interface: cf> and normal component of J continuous across interface, (5-18)
(c) 0 < c[>(r, t) < oo (except in the neighborhood of localized sources).
Here the diffusion coefficient D = Atr/ 3 = [3(~t - µ0~J]- 1 while the extrapolation
length characterizing a free surface boundary condition is z 0 =0.1104 Atr•
The neutron diffusion equation [Eq. (5-16)] may be classified as an example of a
linear partial differential equation of the parabolic type. 4 This type of equation has
been thoroughly studied by mathematicians and physicists alike for years, since it
also describes processes such as heat condition, gas diffusion, and even a wave
function (notice, if we stick an "i" in front of the time derivative, we have
essentially just the Schrodinger equation familiar from quantum mechanics). As we
proceed to apply this equation to nuclear reactor analysis, we will review several of
the more popular schemes available for solving such equations.
In many cases we will deal with situations for which the medium in which the
neutrons are diffusing is uniform or homogeneous such that D and ~a do not
depend on position. Then the one-speed diffusion equation simplifies to
1 acp
- -a - D V2cp + ~acp(r, t) = S (r, t). (5-19)
V t
The explicit form taken by this equation will depend on the specific coordinate
system in which we choose to express the spatial variable r. For convenience, we
have included the explicit forms taken by the Laplacian operator, V2 , in the more
common coordinate systems in Appendix B. 5
THE ONE-SPEED DIFFUSION 1HEORY MODEL / 157
We will frequently consider situations in which the flux is not a function of time.
Then Eq. (5-19) becomes
(5-20)
This equation is known as the Helmholtz equation and is also a very familiar beast
in mathematical physics. It is useful to divide by - D to rewrite Eq. (5-20) as
2 I S (r)
V <f>(r)- -<1>(r) = - - (5-21)
L2 D
where we have defined the neutron diffusion length L
(5-22)
We will later find that L is essentially a measure of how far the neutrons will
diffuse from a source before they are absorbed.
We now tum our attention to the application of this model to some important
problems in nuclear reactor theory. We will first study neutron diffusion in
"nonmultiplying" media-that is, media containing no fissile material. Then we
will tum to the study of the neutron flux in fissile material and begin our
investigatj_on of nuclear reactor core physics.
II. NEUTRON DIFFUSION IN NONMULTIPLYING MEDIA
We will first apply Eq. (5-20) to study the diffusion of neutrons from a
steady-state source in a nonmultiplying medium. All of the mathematical tech-
niques we will use are standard methods which arise in the solution of ordinary or
partial differential equation boundary value problems and are discussed in any text
on mathematical physics or applied mathematics. 1- 3
A. Elementary Solutions of the Diffusion Equation
I. PLANE SOURCE IN AN INFINITE MEDIUM

Perhaps the simplest problem in neutron diffusion theory is that of an
infinitely wide plane source located at the origin of an infinite, homogeneous
medium. The source is assumed to be emitting neutrons isotropically at a rate of S0
neutrons/ cm2 • sec. Since both the source plane and the medium are of infinite
extent, the neutron flux <f>(r) ➔<f>(x) can only be a function of the distance x from
the source plane. Hence the diffusion equation [Eq. (5-21)] reduces to the one-
dimensional form
(5-23)
where we have mathematically modeled the source by a Dirac 8-function: Hence

we just have an inhomogeneous ordinary differential equation to solve with a
FIGURE 5-3. A plane source of neutrons in a infinite

x medium.
slightly weird source. Note that if we restrict x=;FO, the source term disappears from
Eq. (5-23):
d 2cf> 1
- - -cf>(x) = 0 X=;FO. (5-24)
dx 2 L2 '
Our approach will be to solve this homogeneous equation for x =;F0, and then use a
boundary condition at x = 0 to "fix up" these solutions.
We could obtain this boundary condition directly by integrating Eq. (5-23) from
x=O- E: to x=O+ E: across the source plane and then taking the limit as E:~0 to
find
(5-25)
[See Problem 5-2]. However we might also merely note that this is just a special
case of the more general interface boundary condition [Eq. (5-15)]. If we use the
symmetry of the geometry to assert thai Jx(O+)= -Jx(O-)=J(O), then our
boundary condition at the source plane becomes just
So
lim J(x)=-. (5-26)
x-o+ 2
This source boundary condition makes sense physically, since it merely says that
the net neutron current at the origin on either side must be just half of the total
source strength.
We are not through with boundary conditions yet. Since we have a second-order
derivative, d 2/ dx 2, we need another boundary condition. We will use the boundary
condition of finite flux as x~oo.
Hence the mathematical problem to be solved is
d 2cf> 1
- - -rp(x)=O, x>O,
dx2 L2
with boundary conditions:
d</> So
(a) lim - D - = -
x-o+ dx 2 (5-27)
(b) lim cf>(x) < oo.
x-oo
We will then use symmetry to infer the solution for x <0.

To solve this equation, we note the general solution
cp( x) = A exp( - 1)+ B exp ( 1). (5-28)
Applying the boundary conditions, we find
(b)===}B=O
(a)=} lim -D (- A exp(- x )) =AD= So

X-+0 L L L 2
or
Hence our solution is
SoL
cp(x)= 2D exp - L '
( x) x>O, (5-29)
and by symme~y we can infer
x<O. (5-30)
Hence the neutron flux falls off exponentially as one moves away from the source
plane with a characteristic decay length of L. As one might expect, the larger L
(i.e., the smaller ~a), the less the neutron flux is attenuated as we move into the
medium. Notice also that the magnitude of the flux is proportional to the source.
That is, doubling the source strength will double the neutron flux <J>(x) at any
position x, but this should have been anticipated since the neutron diffusion
equation is linear and hence the principle of superposition holds.
The reader should be reminded that, while this solution may provide a reason-
able description of the neutron flux in a medium (provided it is not too highly
absorbing), it is certainly not valid within several mfp of the source plane itself.
Indeed more accurate transport theory studies6•7 indicate that the neutron flux does
not look at all exponential near the source. In fact there are additional components
to the solution. Fortunately if absorption is not too strong, these transport theory
"transients" rapidly diminish as one moves several mfp away from the source
plane, and the simple exponential behavior predicted by diffusion theory is found,
with one mild modification. Transport theory predicts that the "relaxation length"
characterizing the exponential decay is not L = (D /~J•l 2, but rather is given by
the root of the transcendental expression
(5-31)
Fortunately if ~a «~t (as it must be for diffusion theory to be valid), one can
' Transport theory
Diffusion t
FIGURE 5-4. A comparison between the diffusion theory and transport theory solutions of the
plane source problem.
expand Eq. (5-31) to find
L= [1+I(~a)+o(~a)
I
V3~ ~a 5 ~t ~t
2
]. (5-32)
. tr
For example, in graphite ~t = .383 cm- 1 while ~a= .00032. Hence ~a/~t = 8.3 X
10- 4, which implies that the correction to L given by higher terms in this expansion
(that is, by transport theory) is only about 0.03% in this material.
2. POINT SOURCE IN AN INFINITE MEDIUM

As a variation on this theme, let us repeat this calculation for the case of an
isotropic point source emitting S0 neutrons/ sec at the origin of an infinite medium.
Since the source is isotropic, there can be no dependence of the neutron flux on
angle. Hence the diffusion equation in spherical coordinates reduces to
1 d 2 d<f> 1
- - r ---<t>(r)=0 r>0. (5-33)
,2dr dr L2 '
We will use our previous problem as ·a guide, and seek solutions such that the
boundary conditions are
(a) lim 4'1Tr 2J (r) = S0,

r-+0
(b) lim <f>(r) < oo.

T-+C()
One can readily verify that the fundamental solutions to Eq. (5-33) are of the form
, - 1 exp(±r/ L);
hence we are led to seek
exp(-r/ L) exp(r/ L)
cJ>(r)==A - - - - + B - - - . (5-34)
r r
Applying the boundary conditions, we find that (b) implies that we choose B =0,
while (a) implies that A= S0/ 4'1TD. Hence the solution is
S0 exp(- r / L)
cf>(r)= 4'1TrD · (5-35)
An interesting application of this result is to calculate the mean-square distance

to absorption in a nonmultiplying medium. Note the number of neutrons absorbed
between r and r + dr is just
(5-36)
and thus the probability that the neutron is absorbed in dr is just
p(r)dr= { 2 exp(-~ )dr. (5-37)
We can then calculate
(5-38)
Hence the neutron-diffusion length L has the interesting physical interpretation as

being 1/\.'6 of the root mean square (rms) distance to absorption
(5-39)
That is, L measures the distance to which the neutron will diffuse (on the average)
away from the source before it is absorbed. It should be stressed that we have
calculated the rms distance from the source to the point of absorption, not the total
path length traveled by the neutron. This path length will be very much longer
since the neutron suffers a great many scattering collisions before it is finally
absorbed. For example, in graphite the thermal diffusion length is 59 cm. Hence
therms distance to absorption from a point source is ((r 2)) 112 = \.'6 L= 144 cm. If
we recall that the mfp characterizing thermal neutrons in graphite is 2.5 cm and
also recall that the average number of scattering collisions suffered by the neutron
before absorption is 1500, then it is apparent that the average path length or track
length traveled by the neutron is about 3700 cm, considerably larger than \.'6 L.
One can actually use Eq. (5-39) to define the neutron diffusion length in
situations in which diffusion theory would not apply (e.g., strongly anisotropic
scattering or large absorption). Then one would first determine the flux 4>(r)
resulting from an isotropic point source using transport theory (or whatever
description is relevant) and then calculate L by using
(5-40)
where the integral is taken over all space.

The cylindrical geometry problem of a line source at the origin of an infinite
medium can be worked out in a very similar way. However to do so here would
"deprive" the reader of an opportunity to try his own hand at such diffusion theory
problems. Hence we have left the line source as an exercise in the problem set at
the end of the chapter. We will instead turn our attention to problems in finite
geometries.
3. FINITE SLAB GEOMETRIES

Let us now modify our isotropic plane source by assuming that it is imbedded
at the center of a slab of nonmultiplying material of width a surrounded on both
sides by a vacuum (see Figure 5-5). We will set up this problem in a manner very
similar to that for infinite plane geometry, except that we will add vacuum
boundary conditions on either end of the slab.
d 2<t> 1
-dx2 - -<t>(x)=O
L2 ,
(a)
(5-41)
(b)
Here we have replaced the boundary condition at infinity by the vacuum boundary
condition-in this case, using an extrapolated boundary o/2= a/2+ z 0 • If we again
seek a general solution of the form of Eq. (5-28), then applying the boundary
condition (b) implies
<1>{ ~ )=O=A exp{- 2 ~J+ Bexp{ 2~ )=>B= -A exp{- f ).
X = -a/2 x=O X
FIGURE 5-5. A plane source at the origin of a finite slab.

Then boundary condition (a) implies

1
A D ( 1 + exp ( - ~-))-
= 2SL .
Our final solution is therefore
x ( - ( a- x) ) . [ ( a- 2lxl) ]
SL exp( - L )-exp L SL smh 2L
if>(x) = 2D (
l+exp - -
a) = 2D (
cosh -
a) · <542)
L 2L
This solution is sketched in Figure 5-5. It looks somewhat similar to the infinite
medium result [Eq. (5-29)], except for dropping off more rapidly near the
boundaries due to neutron leakage. We should caution the reader once again that
the solution is not valid within several mfp of the vacuum boundary Gust as it is
not valid near the source plane at the origin).
A variant on the above problem involves replacing the vacuum by a material of
different composition than the slab itself (as sketched in Figure 5-6). The general
procedure for attacking such multiregion problems is to seek solutions of the
diffusion equation characterizing each region, and then match these solutions using
interface boundary conditions. Once again we can use geometrical symmetry to
allow us to restrict our attention to the range O< x < oo.
In region G) we will seek a solution of
d2cf>1 1
- - -cf> (x)=O (5-43)
dx2 L21 i '
where L 1 = YD /'.Ea,
1 is the diffusion length characterizing region Q). Similarly in
region@
a
2 <x<oo.
FIGURE S-6. A multiregion or reflected slab.

164 / 1HE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
Now we will need several boundary conditions. We can use our earlier conditions
[Eq. (5-27)]
(a)
(b) X ➔ OO.
In addition we will use the interface conditions
(d) or
Using our earlier work as a guide, we can seek general solutions
cp 1(x)=A 1 cosh { +B 1 sinh { in region (D

l l
cf>i(x)=A 2 exp(- { 2 )+B2 exp({2 ) inregion@
and apply the boundary conditions to find (after a bit of algebra)
+ D2 L 1 sinh ( 2~ 1 )
SLi D 1L 2.cosh ( 2~ 1 )
)+D L sinh( 2~J,

A1=---------------- (5-44)
2D1 D 2 L 1 cosh( 2 ~ 2 1 2
We have sketched the form of this solution in Figure 5-6. Several features of this
solution are of some interest. Note that while the neutron flux is continuous across
the interface, the derivative of the flux is not. This later discontinuity is, of course,
a consequence of the fact that the diffusion coefficients in the two regions differ;
hence to obtain continuity of current J, we must allow a jump in dcp/ dx across the
interface.
We have compared the solution for this problem with that obtained earlier for
the slab surrounded by a vacuum. It should be noted that the flux in the central
region falls off somewhat more slowly when the vacuum is replaced by a diffusing
material. This can be readily understood by noting that the material surrounding
the slab will tend to scatter neutrons back into the slab that would have otherwise
been lost to the vacuum. Such materials used to reduce neutron leakage are known
as reflectors. Any material with a large scattering cross section and low absorption
cross section would make a suitable neutron reflector. For example, the water
channels surrounding LWR cores act as reflectors. In the HTGR, graphite blocks
are added to the top and bottom of the reactor core to serve as neutron reflectors.
A concept very closely related to that of neutron reflectors is the reflection
coefficient or albedo, defined as the ratio between the current out of the reflecting
region to the current into the reflecting region:
(5-45)
(
To make this concept more precise, suppose we want to attach a reflecting slab of
thickness a to a reactor core (or perhaps a medium with a neutron source in it such
as the slab geometry we have just considered). If we are given a current density
Jin= J + entering the reflecting slab surface, which we locate at x = 0 (see Figure
5-7) for convenience, then we can solve the diffusion equation characterizing the
reflecting region:
d2cf> I
- - -cf>(x)=O O<x<a (5-46}
dx2 L2 '
subject to boundary conditions J +(O)=Jin' c/>(a)=O. We can then solve for the flux
cf>(x) in the reflecting slab and use this solution to compute the albedo a as (see
Problem 5-11):
(5-47)
It is of interest to plot the albedo for the slab reflector versus slab thickness as
shown in Figure 5-8. For thin reflectors, very few of the neutrons are reflected and
hence the albedo is small. As the reflector becomes very thick, the albedo
J_(O)
-------
X
x=a FIGURE 5-7. The albedo problem.
1.0
1 - 2D/L
1 + 2DIL
FIGURE 5-8. The albedo for a finite slab plotted

a versus slab thickness.
approaches an asymptotic limit dependent only upon the material properties D and
L:
a ➔ a 00 = (5-48)
l+ 2D.
L
For example, the albedos characterizing infinitely thick reflectors of graphite, H 20,
and D 20 are 0.93, 0.82, and 0.97 respectively.
The albedo can be used to replace the detailed solution in the reflecting region
by an equivalent boundary condition at the edge of the core using Eq. (5-45). That
is, if we use our earlier definitions of the partial current densities J ±, then the
effective boundary condition on the flux in the region x < 0 is just
(5-49)
Used in this manner, the albedo becomes a very useful device for obtaining
boundary conditions for reactor core calculations.
One can continue this game of solving the diffusion equation in various one-
dimensional geometries indefinitely. As we mentioned earlier, it is simply an
exercise in ordinary differential equations. A variety of two- and three-dimensional
problems can also be studied. However since these latter problems involve partial
differential equations, we will defer their treatment until after we have introduced
the separation of variables approach for solving such equations in Section 5-111-C.
If we really want to get masochistic, we can remove some of the symmetry in our
earlier problems so that the original partial differential equation (5-20) would have
to be solved directly-for example, a point source set off-center in a sphere.
However there is very little in the way of new physics to be learned from such
exercises. Hence we will bypass further examples in favor of moving directly to
more general problems. In particular, we will study how our previous solutions for
plane and point sources can be used to determine the neutron flux resulting from
an arbitrary distribution of neutron sources.
4. GENERAL DIFFUSION PROBLEMS

Recall that the neutron flux resulting from an isotropic point source of
strength S0 located at the origin of an infinite medium was found to be
(5-50)
Suppose this source was located at the point r' instead. Then the flux could be
found by a simple coordinate translation as
(
exp ( - lr-r'I)
L
<t>(r)=So 4wDlr-r'I . (5-51)
Next suppose we have several point sources at positions r~, each of strength S;.
Then we can use the fact that the diffusion equation is linear to invoke the
principle of superposition and write
S;exp ( -
lr-r;I)
L
<t>(r)= ~l
4wDlr-r~I · (5-52)
• •
•
•
Several point sources
Distributed source FIGURE 5-9. Superposition of several point sources.

Finally suppose we have an arbitrary distribution of sources characterized by a

source density S (r). Then the flux resulting from this distributed source is just
exp ( - lr-r'I)
,t,(r)= f d'r' 4,rDlr:r'I S(r'). (5-53)
This is frequently rewritten as
(5-54)
where
exp(- lr~r'I}
(5-55)
4wDlr-r'I
is known as the point diffusion kernel for an infinite medium. [The expression kerne/ 8
is a mathematical term used to denote a function of several variables (including the
variables of integration) in an integral of the form
f dx' K(x,x')f(x'). (5-56)
Here K(x,x') would be the known kernel, whilef(x) might either be known [as the
source density S (r)] or unknown as the flux q> in the inscattering term of the
transport equation
(5-57)
where ~s(E'➔ E,O' ➔D) is known as the scattering kernel.]

As a second example of such kernels, consider the flux resulting from a plane
source at the origin
If this had been located at x', then the flux at a position x would be
(5-59)
i:.u•[
Hence in general, for S ➔ S ( x') dx', we find
~(x)= 2~ exp(- lx~x'I) ]s(x')

= J_: dx'GP 1(x,x')S(x'), (5-60)
where we have identified the plane source diffusion kernel for an infinite medium
(5-61)
Once again we will allow the reader the privilege of deriving the line source
diffusion kernel. Actually both the plane source and line source kernels could have
been superimposed from the point source diffusion kernel.
Notice that these kernels all depend only on the difference in spatial coordinates,
that is, r- r' or x - x'. Such displacement kernels arise because we have assumed an
infinite, homogeneous medium. For such infinite geometries we can use these
kernels to compute the neutron flux arising from any source distribution.
Unfortunately most geometries of interest are not uniform and are certainly not
infinite. Hence we must generalize this discussion to determine the diffusion
kernels characterizing other geometries and boundary conditions. Before attempt-
ing this generalization it is useful to step back a moment and try to obtain a general
mathematical perspective of just what problem it is that we really wish to solve.
Actually all we are doing is trying to solve an inhomogeneous differential
equation of the form
Mcp(x) = S (x), (5-62)
where M is a differential operator such as
d2
M=---.
I (5-63)
dx2 L2
(We will leave it as understood that one must also apply suitable boundary
conditions.) There are a variety of techniques available to solve such inhomo-
geneous problems. The approach we have been using thus far is known as the
Green's function method: 9
(a) GREEN'S FUNCTION METHODS
In this technique we first construct the solution to
(5-64)
Then if we call cf>g(x) = G (x,x') the "Green's function" for the operator M, we find
that the general solution to Eq. (5-64) is just
f
cp(x) = dx' G (x,x')J(x'). (5-65)
[Proof:
Mcp= f dx' MG (x,x')f(x') = f(x)] (5-66)
8 (x- x')
EXAMPLE: Consider
Mcp=(!!!__ __
l )cf>(x)=- S(x) (5-67)
dx 2 L 2 D
for an infinite medium, - oo < x < + oo. Now define G by
d2G 1 «5(x-x')
- - -G(x x')== - - - - . (5-68)
dx 2 L2 ' D
However we have just solved this for
, , L (
G(x,x )== GPi(x,x )== 2 D exp -
Ix-L x'I) . (5-69)
Hence the plane source kernel is just the Green's function for this operator, and we
find
q,(x) == J_: dx' GP1(x,x')S (x'). (5-70)
It should not take much contemplation to convince the reader that this is just the
scheme we have been using to construct infinite medium solutions by first deter-
mining the appropriate diffusion kernel or Green's function. The construction of
the Green's function characterizing a finite geometry is not much more difficult,
although one can no longer simply solve for the flux resulting from a source
conveniently located at the origin, and then perform a coordinate translation to
arrive at the Green's function. This latter complexity should be easily understood
when it is recognized that shifting the source "off-center" in a finite geometry will
destroy the symmetry of the problem. Hence one no longer finds a simple
displacement kernel form for the Green's function. We will provide an alternative
scheme for constructing these kernels later in this section.
(b) VARIATION OF CONSTANTS
Perhaps a more familiar scheme for the solution of inhomogeneous differen-

tial equations is variation of constants, 10 in which one first determines the linearly
independent solutions q>~im of the homogeneous equation
(5-71)
and a particular solution of the inhomogeneous equation

(5-72)
Since none of these solutions are required to satisfy the boundary conditions
pertaining to the specific problem of interest, they are usually rather straightfor-
ward to find. Then one seeks the general solution to the problem as
(5-73)
and applies the boundary conditions to determine the unknown coefficients of the
homogeneous solutions, A 1 and A 2 •
EXAMPLE: Consider a uniform source S(x)== S0 in an infinite medium. Then
(5-74)
with boundary conditions such that cf>(x) < oo as X ➔ ± oo. The solutions to the
homogeneous equation are exp(± x/ L), and the particular solution is obviously
cf>(x) = S 0 L 2 / D. Hence we will seek a general solution
(5-75)
Then applying the boundary conditions implies that both A and B must be zero,
and hence the solution is
(5-76)
Usually for the method of variation of constants to be of use in more general

problems, one must be able to guess cf>par.(x) "by inspection." Hence for most
problems, the Green's function technique or the method we will describe next is
more convenient.
(c) EIGENFUNCTION EXPANSION METHODS.
One of the most powerful methods available for solving boundary value
problems is to seek the solution as an expansion in the set of normal modes or
eigenfunctions characterizing the geometry of interest. Rather than beginning with a
general description of this very important scheme, we will introduce it by consider-
ing a specific example.
EXAMPLE: We will attempt to determine the neutron flux resulting from an
arbitrary distributed source in a finite slab of width a. That is, we wish to solve
a a
--<x"-
2 2'
(5-77)
subject to the vacuum boundary conditions:
(a) <t>(1)=o
(b) ct>( - ~) =0.
Since we have taken the source S(x) to be arbitrary, we cannot assume symmetry
to restrict our attention to the range O< x < ii /2.
Our approach to solving this problem may at first seem a bit irrelevant. We begin
by considering a homogeneous problem very similar to Eq. (5- 77):
(5-78)
Here B 2 is just an arbitrary parameter-at least for the moment. Let us now solve
this associated homogeneous problem by noting the general solution
if; ( x) =A 1 cos Bx+ A 2 sin Bx. (5-79)

Our boundary conditions require
.i,(+
-r - 2ii)-A
- i cos (Bii)+A
T - sm . (Bii)
T . 2 (5-80)
Adding and subtracting these equations, we find that we must simultaneously

reqmre
and
(5-81)
How do we achieve this? Certainly we cannot set A 1 and A 2 equal to zero since
then we would have the "trivial" solution ip(x)=0. Instead we must choose the
parameter B such that these conditions are satisfied. Of course there are many
values of B for which this will occur. For example, if we choose A 2 = 0, then any
n= 1,3,5, ... (5-82)
will give rise to a solution

n'1TX
o/n ( X ) =An cos --- , n = 1,3, 5, ... (5-83)
a
which obviously satisfies both the differential equation [Eq. (5- 78)] and the
boundary conditions. Alternatively, we could have chosen A 1 =0, in which case
n=2,4,6, ... (5-84)
yields solutions
n =2,4,6, ... (5-85)
Hence our homogeneous problem can be solved only for certain values of the
parameter B. One refers to the values of B 2 for which nontrivial solutions exist to
the homogeneous problem as eigenvalues:
Eigenvalues: Bn2-- ( na'1T )2, n = 1,2, ... (5-86)
The corresponding solutions are referred to as the eigenfunctions of the problem:
n'1TX) ,
Ancos ( --- n= 1,3,5, ...
Eigenfunctions: o/n ( x) = { . { n'1TX
a ) (5-87)
A nSlil --- , n =2,4,6, ...
a
The reader may have already encountered eigenvalue problems in a somewhat

different form:
(5-88)
where ,fl,. is the eigenfunction corresponding to the eigenvalue >..,.. However by

comparing this form with Eq. (5-78), one can easily identify H = d 2 / dx 2 , A-+- B 2 ,
and \[;,.-+,f!,.(x).
Notice that in the example Eq. (5-78) we actually find two types of eigenfunc-
tions: the cosine functions corresponding to odd n and symmetric about the origin,
and the sine functions corresponding to even n and antisymmetric about the origin.
Had we restricted ourselves to symmetric sources S ( x) = S ( - x ), we could have
eliminated the antisymmetric solutions [Eq. (5-85)] from further consideration. We
have sketched the first few eigenfunctions for the slab geometry in Figure 5-10.
In acoustics these eigenfunctions would be identified as the normal modes or
natural harmonics of the system, and this terminology is frequently carried over to
reactor analysis. Notice that the A,. are still undetermined and are, in fact,
arbitrary. These can be chosen in a number of ways, but for now we will just set
A,.= 1 for convenience.
So now this auxiliary problem has given us an infinite set of solutions ,f!,.(x).
What good are they? Well, they have a couple of very useful properties. First notice
that the product of any two of these functions will vanish when integrated over the
l~
slab unless the functions are identical:
0, if m=pn
_ ~ dx,[!m( x),[!,.( x) = ~, if m=n.
(5-89)
2
This property is known as orthogonality and proves to be of very considerable
usefulness, as we will see in a moment.
The second property of the eigenfunctions i/ln(x) is that they form a complete set
in the mathematical sense that any reasonably "well-behaved" 1 function f (x) can
be represented as a linear combination of the l[;,.(x):
00
J(x) = ~ en ,f!,.(x). (5-90}

n-l
FIGURE 5-10. The eigenfunctions for a slab geometry.

Of course, such a representation is only of formal interest unless some scheme is

available for determining the expansion coefficients en, but this is where orthogon-
ality comes in handy. Multiply Eq. (5-90) by 'Pm(x) and integrate over x to find
a a a
2 00 2 2
f_ ~dxl/Jm(x)j(x) = ~ en f_ ~dxl/Jm(x)l/Jn(x) = cm f_ ~dxlf;!(x)
2 n= l 2 2
or
a
2
cm=-= f 2 .
11 dxj(x)l/Jm(x). (5-91)
a --
2
Hence given any function f(x), we can evaluate the appropriate expansion
coefficients cm by a simple integration. It should be mentioned that for the specific
example of a finite slab we have been considering, an eigenfunction expansion is
simply a Fourier series expansion and is probably already quite familar to most
students. 1 However, the properties of orthogonality and completeness characteriz-
ing such trigonometric functions also hold for much more general eigenfunc-
tion expansions.
With this background, we are finally ready to return to solve our original
boundary value problem. As we mentioned, the essential idea is to seek the solution
as an expansion in the eigenfunctions 'Pn(x):
00
</>(x) = ~ en 'Pn(x). (5-92)

n=l
We will also expand the source term in a similar fashion

00
S(x)= ~ snl/Jn(x). (5-93)

n=l
Notice that since S (x) is known, we can use orthogonality to determine the source
expansion coefficients [as in Eq. (5-91)] as
a
2 2
sn= a f
-~dxS(x)l/ln(x) (5-94)
2
Of course since </>(x) is unknown, we cannot determine the en in a similar fashion,

but that is just what we can use the original equation [Eq. (5-77)] to accomplish. If
we substitute Eqs. (5-92) and (5-93) into Eq. (5-77), we find
(5-95)
However using Eq. (5-78), we can eliminate d2l/ln/ dx 2 to find

00
~ en( B;+-¼
n~I L
)l/Jn= ~ f Sn'Pn·
n=I
(5-96)
11IE ONE-SPEED DIFFUSION THEORY MODEL / 175
Thus we are left with one equation for an infinite number of unknowns, the en.
Fortunately orthogonality once again comes to our rescue. Multiply by V'm(x),
integrate over x, and use orthogonality to find
sn/D
C n= = (5-97)
B2+ _l_
n L2
Thus we find the flux for any source distribution as
(5-98)
where Vin (x) = cos( mrx / a), n odd and Vin (x) = sin( nwx / ii), n even.
We can rewrite· this in a bit more familiar form if we substitute Eq. (5-94) into
Eq. (5-98) and rearrange things a bit:
11
2
= f_ ~ dx'Gp1(x,x')S (x'). (5-99)
2
Hence we have found an explicit representation of the plane diffusion kernel or

Green's function for a finite slab as an eigenfunction expansion
G ( ')= __±._ ~ V'n(x)'l/ln(x')

pl X,X -~ "-I 2 2 • (5-100)
a a n=l 1 + L Bn
Note in particular that the Green's function for a finite geometry is no longer a
displacement kernel, that is, a function only of x - x', as it was for infinite
geometries. ·
This intimate relationship between Green's functions and eigenfunction ex-
pansions is actually a very general result. Suppose we consider the diffusion
equation characterizing any homogeneous geometry:
(5-101)
subject to the usual vacuum boundary conditions on the extrapolated boundary:
Boundary conditions: <f>(r5) = 0, r5 on surface.

As before, we first construct the eigenfunctions as the nontriv.ial solutions to the
associated homogeneous problem
(5-102)
Boundary condition: Vin (is)= 0.
One can demonstrate that these eigenfunctions are orthogonal in the sense that
(5-103)
Since Eq. (5-102) is homogeneous, any solution 'PnCr) may be multiplied by a

constant. We will scale each 'Pn so that the eigenfunctions are normalized such that
(5-104)
[Our earlier slab eigenfunctions 'Pn(x) would be normalized if we multiplied them

by (2/5) 112 .] The set of orthogonal and normalized eigenfunction:., {'PnCr)} is said to
be orthonormal. It can also be shown to be complete. Hence we can expand
q,(r) = ~ cn'PnCr),
n
(5-105)
S (r) = ~ sn'Pn(r),
n
where
If we substitute these expansions into Eq. (5-101), we can use orthogonality as

before to solve for
(5-106)
Hence we find
(5-107)
where
(5-108)
as a general result for any geometry (although the eigenfunctions or spatial modes
'Pn(r) may be very hard to construct in practice).
B. Numerical Methods for Solving the Neutron Diffusion Equation
l. INTRODUCTION
Thus far we have confined our attention to neutron diffusion in homogeneous

(or perhaps regionwise homogeneous) media since in this case the one-speed
diffusion equation could be solved analytically. However in any realistic reactor
calculation the heterogeneous nature of the core must be taken into account. One
not only must consider nonuniformities corresponding to fuel pellets, cladding
material, moderator, coolant, control elements, but spatial variations in fuel and
coolant densities due to nonuniform core power densities and temperature distribu-
tions as well. Such complexities immediately force one to discard analytical
methods in favor of a direct numerical solution of the diffusion equation. In fact
• even when an analytical solution of the diffusion equation is possible, it is
frequently more convenient to bypass this in favor of a numerical solution,
particularly when the analytical solution may involve numerous functions that have
to be evaluated numerically in any event, or when parameter studies are required
that may involve a great many such solutions.
The general procedure is to rewrite the differential diffusion equation in finite
difference form and then solve the resulting system of difference equations on a
digital computer. It is perhaps easiest to illustrate this approach by a very simple
example (sufficiently simple, in fact, to enable analytical solution). Suppose we
wish to solve
(5-109)
subject to the boundary conditions characterizing a finite slab of width a:
cf>(O) = cf>( a)= 0.
(For convenience we will ignore the extrapolation length.)

We first discretize the spatial variable x by choosing a set of N + 1 discrete points
equally spaced a distance a= a/ N apart (for convenience).
4
~
--------~~--1---1----~,}--1----
X;-1 Xi Xj +1 ~1
We now want to rewrite Eq. (5-109) at each of these discrete points x;, but to do so
we need an approximation for d 2cf>/ dx. 2• Suppose we Taylor expand cf> at X;± 1 in
terms of its value at the point x;:
(5-110)
If we add these expressions, we find
(5-111)
to within order .!i2 • Hence if a is chosen sufficiently small, this three-point central
difference formula should be a reasonable approximation to the value of d 2 f/>/ dx 2 at

the point X;.
If we now use this difference formula to write Eq. (5-109) at any mesh point x;,
we find
(5-112)
where again we have defined S; - S ( x;). We can rearrange this difference equation
to rewrite it as
(5-113)
or
a.l, 1-
. I.i..
'f'1- I+ aI,
. l'f6l 't'z I= s.l
..i.. + a.l, I·+ i.i.·+ (5-114)
i=l, ... ,N-1
Hence we now have reduced Eq. (5-109) to a set of N-1 algebraic equations for
N+ 1 unknowns (f/>0,r/> 1,f/>2, ... ,cpN). If we add on the boundary conditions at either
end, say ff>o = 0, cpN = 0, we can now imagine solving (or imagine the computer
solving) this set of algebraic equations. In this particular case, the system of
algebraic equations can be solved directly using Gaussian elimination.11 More
generally one must use iterative methods to solve the finite-difference equations.
This very simple example illustrates the two essential tasks involved in the
numerical solution of the diffusion equation: (a) derivation of the corresponding
difference equations and (b) formulation of a suitable algorithm for solving these
equations on a digital computer. The methods used will vary from problem to
problem. For example, whereas a direct solution of the difference equations [e.g.,
Eq. (5-114)] is possible for one-dimensional problems, iterative methods are re-
quired for two- and three-dimensional problems. Furthermore one generally desires
to work with nonuniform meshes in reactor calculations, to account for the fact
that the neutron flux may vary much more rapidly in certain regions than in others.
In this section we illustrate several of the techniques that are commonly applied
in reactor analysis to the derivation and solution of difference equations. However
just as in our earlier study of analytical techniques, we do not intend this
discussion to be a detailed discussion of numerical methods in nuclear reactor
calculations. Instead we refer the interested reader to the extensive literature on
this topic. 12- 16
2. DERIVATION OF DIFFERENCE EQUATIONS FOR

ONE-DIMENSIONAL DIFFUSION PROBLEMS
We will now consider the more general form of the one-dimensional diffusion
equation in plane geometry
d df/>
- dx D (x) dx + ~a(x)cp(x) = S (x) (5-115)
subject to boundary or interface conditions that we will leave arbitrary for the
moment. Actually we should ·remark here that this form of the diffusion equation is
even a bit too general for most reactor applications. One rarely encounters reactor
configurations in which the composition varies in a continuous way from point to
point [i.e., D (x) and ~a(x)]. Rather the system properties are assumed to be
essentially uniform in various subregions of the reactor core (or can be suitably
represented by spatially averaged or "homogenized" properties within each subre-
gion). Hence the far more common situation is one in which the diffusion equation
[Eq. (5-109)] with constant D1 and ~aJ must be solveq. in a number of regionsj. We
will develop the difference equations for the general diffusion equation [Eq.
(5-1 h>)J:' however, since they are not really any more difficult to derive or solve,
and in certain cases they are useful in avoiding technical difficulties arising in less
general approaches (such as the handling of region interfaces).
As in our simple example we begin by setting up our discrete spatial mesh as
shown below, although we will now allow for nonuniform mesh spacing.
.::i., .:'.1.2 A, .::1..-+1
I ' I ' \/
Xo
I
x,
\
X2
I H I
X;-1
I
xi
'I
x,+1
~~
XN-1 XN
There are a variety of schemes that can be used to generate a difference equation
representation of Eq. (5-115) on this mesh. We have already considered a simple
problem in which a Taylor series expansion was used to derive a central difference
formula for d 2cp/ dx 2 • A more common scheme is to integrate the original differen-
tial equation over an arbitrary mesh interval, and then to suitably approximate
these integrals (after an occasional integration by parts) using simple mean values
or difference formulas. By way of illustration, suppose we integrate Eq. (5-115)
over a mesh interval xi - .1.J2 < x <xi+ .1.i+ 1/2 surrounding the mesh point X;,
• • .
Let us choose the simplest scheme to approximate the integrals by expressing them
as the value of the integrand evaluated at the meshpoint X; times the integration
interval. For example,
(5-116)
(5-117)
The derivative term requires a bit more work. First write
i '
x•+ A;+1
1
I::.;
x---
2
2
dx
d
dx- D ( x)-
dx
dcp
=D
dcp
( x)-1
x,+ -2-
t::.,
dx x/- -
A1+1
2
(5-118)
To handle dcp/ dx, we can use a simple two-point difference formula [which can be
derived by subtracting Eqs. (5-110)1:
X· + Ai+1
dcp
dx
I ~ c/>;+ l -cf>;
Al+t == A
' 2
x-+-
I 2
L.i;+l
X· - Ai
I 2
• •
Furthermore we will use a centered average for D:
(5-119)
Then we find that Eq. (5-118) can be written as
1 X;
I
+-
!::.;+I
t::.. 2
x-- _'_
2
d
d -D(
xd
X
x d
drk D. · l . -
)-'1"-'.::'.~
X
a c/>, - l
i
( D. ·+1
_,,,_ D ..
ai+ l + ......!..:!..=.
ai
J) ,1,,,+-
D ..·'-,1,,_
'1",
1+I
ai+ 1 '1", + 1•
(5-120)
If we now combine Eqs. (5-116), (5-117), and (5-120), we arnve at a set of
difference equations very similar to our earlier results
a.1,1-
. 1,1,,·
't"t- l + a.l,to/1
_,1,,_ + a.t,I·+ 1,1,,·+
Y1
l = s.,
i
(5-121)
where
(5-122)
Hence once again we have arrived at a set of N-1 three-point difference equations
for the N + I unknown discretized fluxes, cp0 ,cf> 1, ... ,c/>N· In the particular case in
which the mesh size Ll; is constant and the coefficients D (x) and ~a<x) do not
depend on x, we return to our earlier results [Eq. (5-113)] derived via a Taylor
series expansion. Actually we should note that the coefficients aiJ depend only on a
single subscript, j. However double subscripting is useful, for we will rewrite these
algebraic equations as a matrix equation.
Our final task is to append to these equations two additional equations taking
into account the boundary conditions. Of course we could simply use the vacuum
extrapolated boundary conditions, </>o = 0, </>N = 0 as before (taking care to place the
mesh points x 0, and xN on these extrapolated boundaries). More general boundary
conditions (such as nonreentrant current) can be developed by taking the final two
difference equations in the set as
and
(5-123)
Such sets of three-point difference equations are characteristic of one-

dimensional diffusion problems (indeed of any ordinary differential equation of
second order). The coefficients aiJ will depend upon the scheme used to derive the
difference equations. Fortunately if the mesh spacing Ll is small, these differences
will be insignificant in actual calculations. Since the spatial variation of the flux is
essentially characterized by the diffusion length L, one generally chooses a mesh
spacing Ll less than L.
Similar three-point difference equations will also arise in curvilinear geometries
with one-dimensional symmetry. For convenience we will assume regionwise uni-
form properties. Then in cylindrical coordinates, the diffusion equation becomes
d2</> 1 d<j> ]
-D -
[ dr2
+ -
rdr
- +~<j>(r)=S(r),
a
(5-124)
while in spherical coordinates, we find

2
-D - </>
d -d<j> ] +~ a<J>(r)=S(r).
+-2rdr (5-125)
[ dr2
Hence we can derive difference equations corresponding to these geometries, using

either of the earlier techniques, to find for uniform mesh spacing 17
(5-126)
where now
(5-127)
where c=0, 1,2 for plane, cylindrical, or spherical geometries, respectively. To

complete the specification of these difference equations, we will consider the case
of vacuum boundary conditions. For reference we first consider the slab geometry
with uniform mesh spacing ll:
--------L---------'e.-1
Boundary conditions: </> 0= </>N = 0,
(5-128)
S0 = SN=0.
Notice if we were to apply Eq. (5-126) to the case i = 0 (assuming <f> _ 1 = 0), then this
boundary condition would imply </>, = 0. However this is inconsistent with the i = 1
equation. Hence we must ignore the i = 0 (and i = N) cases in applying the
difference equation, and use the boundary conditions to eliminate the unknowns </>o
and </>N from the i = 1 and i = N - 1 equations. For example, the i = 1 equation is
0
a,,/o+ a11</>1 + a12<1>2= S1. (5-129)
Now consider the case of cylindrical geometry. Our boundary conditions are
now somewhat different. At rN we still have the usual vacuum boundary condition
</>N = 0. At the origin, we use symmetry to imply
l i
R
d</> I = 0 ==> </>o = </>1, S0= S 1. (5-130)

• • • dr r=O
ro r, r; rN
Again we ignore the cases i=O and i=N. At i= 1, we find that a 10 =0 and hence
(5-131)
The i = N - 1 equation is simplified by </>N = 0. Very similar considerations hold for

the case of spherical geometry.
Thus we have now derived the general form of the difference equations
characterizing one-dimensional neutron diffusion in plane geometry, and the
difference equations characterizing regionwise homogeneous cylindrical and spheri-
cal geometries. (The extension to nonuniform media and mesh intervals for
cylindrical and spherical geometries will be left as an exercise for the reader.) Our
next task is the determination of a suitable prescription or algorithm for solving this
system of algebraic equations.
3. SOLUTION OF THREE-POINT DIFFERENCE EQUATIONS

Suppose we have developed an appropriate set of difference equations similar
to Eq. (5-121). We must now solve for the discretized fluxes cf>;, To be more explicit,
let's first write out these equations in detail
a11cf>1 + a12cf>2 =S1

a21cf>1 + a22cf>2 + az3'P3 =S2
a32cf>2 + a33cf,3 + a34cf,4 =S3 (5-132)
aN _ l,N-2cf>N-2 + aN-1,N- 1cf>N- I =SN-I
It is useful to rewrite these equations in matrix form
or
(5-133)
where A is an (N - 1) X (N -1) matrix, and cf> and S. are (N - 1) dimension column

vectors~(For a short review of matrix algebra, the reader should refer to Appendix
F.) In particular the reader should note that the finite difference diffusion matrix A
is tridiagonal. This feature arises only for such one-dimensional geometries in which
three-point difference equations arise. In two- and three-dimensional geometries,
more complicated matrix structures are encountered.
Notice that the solution of this set of algebraic equations is tantamount to
inverting the matrix d, to find
(5-134)
Such tridiagonal matrices can be inverted directly using Gaussian elimination (the
"forward elimination-backward substitution" method). 18 The general scheme is to
subtract a;,;- 1 / a;- i,;-J times the (i- l)th row from the (i)th row to eliminate the
a;;- J element in the (i)th row. At each step, the (i)th row is then divided by its
diagonal element, and the procedure is continued. For example, we have indicated
schematically the "forward elimination'' on the first two equations in Eq. (5-132):
0 a12 a12
a12 1 0 0
a22 a(J
'v
- all
a22 a23 - (a _
22
all
a21a12)
a
11
a23
a12
1 0 1 A1 0
all
- 0 1
(a _
22
a23
a21a12)
a
11
- 1 A2 (5-135)
such that we eventually arive at a system of equations of the form
A1 0 0
1 A2 0
0 1 (5-136)
cl f)
0
where
A = an,n+l
n
an,n + an,n-1 An-1 ,
(5-137)
We can now substitute back up the matrix to find
(5- 138)
and so on.
Thus Gaussian elimination cons1stmg of forward elimination and backward
substitution can be used to directly solve the difference equations [Eqs. (5-132)].
This scheme is particularly important since it frequently appears as an integral part
of the iterative methods used in two- and three-dimensional diffusion problems.
For this reason it is useful to formalize Gaussian elimination a bit by noting that
what we have in fact accomplished by forward elimination is the factorization of
the matrix A into a product of a lower Cb) and upper CID triangular matrix: 19
0 0 A1 0
(a22-a21A1) 0 1 A2
A=
G32 (a33- a32A2) 0 1
k u (5-139)
Hence our sequence of steps in Gaussian elimination begins with
(5-140)
First we perform a forward elimination sweep to construct and invert b.
(5-141)
followed by a backward substitution to invert U and solve for
(5-142)
As an aside we should observe that while such methods for solving systems of
linear algebraic equations are most easily understood and analyzed (mathemati-
cally) in matrix notation, they are most easily programmed when written as a
simple algorithm such as Eq. (5-138). For example, one could simply construct a
loop to generate and store all An and an using Eq. (5-137) and then evaluate all cf>n
using Eq. (5-138).
This algorithm for solving such sets of three-term equations (i.e., inverting
tridiagonal matrices) is easily programmed and executed on a digital computer.
The algorithm would also formally work for solving difference equations
characterizing two- or three-dimensional diffusion problems, however it then en-
counters some severe computing limitations. To visualize this more clearly, we will
now briefly comment on the numerical solution of multidimensional diffusion
equations.
4. DERIVATION OF
MULTIDIMENSIONAL DIFFERENCE EQUATIONS
Most detailed neutron diffusion calculations characterizing nuclear reactors
require either two- or three-dimensional treatments. Such details are particularly
important in studying power profiles in large reactors subject to nonuniform fuel
loading and depletion. Hence we now must consider the numerical solution of the
more general diffusion equation
-V · D (r)Vcf, + La{r)cp(r) = S (r). (5-143)
Once again the geometry of interest is discretized into a mesh of cells such as the
rectangular grids illustrated in Figure 5-11. Perhaps the most general way to derive
FIGURE 5-11, Rectangular two- and three-dimensional grids
difference equations for the mesh is to integrate the diffusion equation [Eq. (5-143)]
over the spatial volume of a given mesh cell, using this to define the spatially
averaged cell properties. In general one can write20
(5-144)
(5-145)
(5-146)
(5-147)
Here the sum is taken over the adjacent mesh point neighbors j = I, ... , J where
J = 2, 4, or 6 in 1-, 2-, or 3-dimensional Cartesian geometries, while
J
L;= ~ liJ, (5-148)
J=I
where the mesh coupling coefficients liJ are determined by the particular mesh
geometry and finite-difference scheme one chooses. For example, in Cartesian
coordinates using essentially the approximation schemes represented by Eqs. (5-
116) and (5-120), one would find
(5-149)
where we define
~-- = dis_tance between mesh (5-150)

1J pomts i andj.
The difference equations representing Eq. (5-143) then take the form
(5-151)
where i runs over all of the mesh points.

EXAMPLE: As an example, let's derive the difference equations characterizing

the two-dimensional diffusion equations for a uniform medium with a uniform
mesh in rectangular geometry
(5-152)
Our mesh will be defined such that the mesh points are denoted by x 0 , x 1, ••• , X; .•. ,
xN; Yo, y 1, ... , Yp ... , YM with mesh spacing .:ix and .:iy respectively. The most
direct manner in which to derive the difference equations is to use a central
difference formula to approximate
. .+ c/>·+
c/>·1 - I ,] - 2cplJ l I ,J.
(5-153)
1
'=
X;,Y·
'-.,.1
yj
'"
y1
"
"
"
"
FIGURE S-12. The two-dimensional spatial mesh
(We will defer to an exercise at the end of the chapter the more general derivation
for nonuniform media using mesh cell integration.) Substituting these expressions
into Eq. (5-152) evaluated at the mesh point (x;,Y) we find
+ ~ +2D ( - -
1 + -1- )]
( .:ix )2 (iiy )2
.. =S..
c/>,,1
(5-154)
[ a 1J
i=l, ... ,N-1

J=l, ... ,M-1.
As before, we can use boundary conditions to specify c/>oJ,c/>N,J'c/>;,o, and c/>;,M·

5. ITERATIVE SOLUTION OF
MULTIDIMENSIONAL DIFFERENCE EQUATIONS
We now turn our attention to the solution of these difference equations. Our
first task is to cast the set of equations into matrix form. This requires first
assigning a single index to each mesh point. For example, in a two-dimensional
mesh we could label the mesh points as
(i,J)-+k = i + (j-1 )(N-1).
For a two-dimensional problem, the matrix structure takes the following form-in
this case, for a five-by-four mesh point array
(5-155)
Notice that the tridiagonal form we encountered in the one-dimensional case has
now been augmented by two additional side-band diagonals-as we might have
expected, since on closer examination we find the two-dimensional case yields a
five-point difference equation.
Similarly, assigning a single index to each mesh point of a three-dimensional
problem yields the matrix structure which corresponds to a seven-point difference
equation.
Now let's consider how we might solve such systems-that is, invert such
matrices. Since Gaussian elimination can be applied to any matrix (formally, at
least), we might first consider applying this technique to obtain a direct inversion.
Recall that for a one-dimensional diffusion equation, the forward elimination
sweep reduced the original tridiagonal matrix to a form with only two diagonals,
while the backward substitution step completed the matrix inversion
Backward
substitution (5-157)
After a bit of examination, it becomes apparent that when a similar forward sweep
is conducted on the five-diagonal matrix characterizing two-dimensional problems,
the result is to fill in all of the zero entries between the main and outer diagonal.
(5-158)
This implies that one will require considerably more computer memory to allow
such a direct inversion of the matrix. Such a direct algorithm is also rather
complicated to program and leads to problems resulting from computer round-off
error. For these reasons, it is far more efficient to use an iterative procedure to
invert such matrices when N is large, since such schemes attempt to preserve the
sparse structure of the original matrix in their operations.
Let's illustrate the basic idea with a simple example: Suppose we wish to invert a
matrix d,-that is, we wish to solve
--
Acp=S.
-
(5-159)
We first compose A into its diagonal and off-diagonal elements
A
= D
=
B
(5-160)
Now !2 can be easily inverted:
(5-161)
Hence suppose we use Eq. (5-160) to rewrite Eq. (5-159) first as
(5-162)
and then invert D to find
</> = D - '[ B </> ]+ D - Is. (5-163)
Now is where the iterative philosophy comes in. Suppose we guess </> on the
right-hand side-call the guess <1><0>-and then use it to calculate a new guess, </>(I)
as
(5-164)
We can continue this iteration, calculating the m + 1 guess as
</> (m + I)= D - I B </> (m) + D - I S . (5-165)
Hopefully, then, as m becomes large, we converge to the true solution
</> (m) - </> • (5-166)
Hence the general idea behind such iterative schemes is to generate improved
guesses or iterates </>(m) by solving the original system of equations in an approxi-
mate, but efficient-:- manner. We continue such an iterative process until two
successive iterates </>(m) and </>(m+ 1> are sufficiently close together, at which point
the iteration is stopped and </>(m+ 1> is regarded as the solution. Notice that
throughout the iterative process-:- we maintain the sparse structure of the original
five-diagonal matrix A, thereby significantly reducing storage and calculational
requirements.
The particular scheme we have presented is known as the Jacobi-Richardson or
Point-Jacobi method, and although it is a very simple scheme, it has the drawback
that it converges very slowly. The reader might very roughly think of the conver-
gence rate of such iterative processes as being determined by how big a chunk of
the original matrix he is willing to invert on each iteration. (More precisely, the
convergence rate is determined by the size of the matrix norm of n- 1B.) In the
Point-Jacobi method, only a relatively small bit of the matrix, its main diagonal, is
inverted on each step and hence we might expect convergence to be slow. (As an
extreme example, one bites off a much bigger chunk in Gaussian elimination-the
whole matrix A-and hence only a single iteration is needed.) One can accelerate
this convergence in several ways. First, one could attempt to invert a bigger chunk
of A on each iteration. It is also possible to use information about the next flux
iterate during an iterative step. Finally, one can extrapolate from earlier flux
iterates in order to more rapidly approach the true solution.
To understand how to improve the Jacobi iterative scheme, let's write it out
explicitly in terms of the algebraic system
+ a12</>Jm) + a13</>1m) + ... + a

IN N
</> (m) =S,
a </> (m} + a </>(m+ I) + a 23 </>(m) + ... + a </>(m) =S2

21 I 22 2 3 2N N
....____
solving only for these ,
a </> (m) + aN2</>(m) + aN3 </>(m) + ... + a </>(m +I) =SN

NI I 2 3 NN N
(5-167)
Hence we can solve for the m + 1 flux iterate immediately as
<t>[m+O= : ..
11
[s;- ± 1= I
au<f>}m>], i= 1,2, ... ,N. (5-168)
9'= i
It should be noted here that the Jacobi scheme does not use all of the available
information during each iteration. For example, if the equations are solved in
sequence from i = I to i = N, as they would be on a computer, then the solution of
the first equation yields <1>t+ 1>; but to find <1>t+ 1>using the second equation, <t>fm>
is used rather than the improved estimate <t>fm+ 1>. Similarly, solving the third
equation for <1>t+ 1>makes use of <1>t> and <1>t> rather than <t>fm+ 1> and <1>t+ 1> which
are known. If these estimates are used as soon as they are generated, a more
efficient iterative scheme known as the Gauss-Seidel or successive relaxation method
is obtained. In this case, the system of equations in each iteration is solved as
a 11 </> (m
I
+I) + a 12</>(m)
2
+ a 13 </>(m)
3
+ ... + a IN </>(m)
N -
_ SI
a 21 </> (m
I
+I)+ a 22</>(m
2
+I) + a23</>(m)
3
+ ... + a 2N</>(m)
N
_
-
S2
(5-169)
and the solution is
[
S- -
l
i~l
""'
)=I_;/
a ..<f>(m + I) -
lj j
f au<f>5m>] .
j=i+1_j
From solution of From previous (5-170)
previous equations (mth) iteration
in current (m + I)
iteration
This can be rewritten in matrix form by decomposing A into the sum of an upper
and lower triangular matrix:
A = L
(■l
(5-171)
=
Here k contains elements of the main diagonal and below it, while U contains
elements above the main diagonal. Now we write Eq. (5-159) as
(5-172)
The Gauss-Seidel scheme described above amounts to inverting b by forward

elimination, stepping row by row down the matrix. Hence our iterative scheme is
(5-173)
The fact that the Gauss-Seidel method utilizes the latest iterate elements of cf,(m+ J)
when solving successive equations yields a factor of two better in error reduction
per iteration than the Jacobi method.
It is possible to accelerate the convergence of the iteration scheme even further
by introducing an acceleration parameter to extrapolate the iterative flux estimate.
This procedure, known as the successive overrelaxation (SOR) method, can be
illustrated by considering how one utilizes the cf,(m) iteration to determine the cf,(m+ I)
estimate. The first step in the calculation of cf,P,, + J) is to compute the Gauss-Seidel
estimate, which we will label as cp/m+ 1/ 2) for convenience.
cp/m+ ½) = _!_ [ S.-

a;;
I
i~I
-"'
j=/
a .. ,1,.(m+ I)
IJ'f'J -f
j=i+/
aiJcf,t)]
(5-174)
Gau~Seidel From SOR in From SOR in
estimate current iteration previous iteration
[(m + l)st] [mth]
Now cp/m+ I) is calculated as a linear combination of cf,; (m+ ½land the previous SOR
iterate
(5-175)
Here the extrapolation or acceleration parameter w ranges between 1 and 2. Of

course for w = I we return to the Gauss-Seidel method in which no extrapolation is
used. The iterative algorithm for each element can then be written as
,1,.~m+l)=~[s.- i~la ..,1,.(m+I)_

'f'I a.. I -"' IJ'f' J
~
-"'
a._,1,.(m)]+(l-w),1,.~m)
y'f' j 'f'I •
(5-176)
II j=I j=i+I
True </l·
solution '
SOR</l 1m1
'
:revious ,p.lm--ll
iterate '
0 w
20
FIGURE 5-13. Flux extrapolation in the SOR method
We have sketched in Figure 5-13 how the flux extrapolation can enhance conver-
gence to the true solution cf,.
The optimum value of w-giving the maximum rate of convergence can be related
to characteristics of the original matrix d. In certain cases one can achieve a
convergence rate as much as two orders of magnitude larger than the Jacobi
method. It should be noted, however, that the estimate used for w can strongly
affect the convergence rate of this method, and it frequently must be determined
by experience.
Very similar methods can be applied to three-dimensional diffusion problems. In
this case the diffusion matrix A has seven diagonal elements as indicated below
= (5-177)
Again iterative methods are utilized in which the outer diagonal elements are
handled in a manner similar to those used in two-dimensional problems. However
there is some reduction in iterative convergence rates due to a loss of procedure
implicitness caused by the additional diagonal elements.
Such iterative algorithms for the solution of the finite difference equations
characterizing two- or three-dimensional diffusion problems are frequently referred
to as inner iterations. This terminology arises from the fact that in nuclei>r reactor
criticality calculations, the solution of the diffusion equation is itself imbedded in
yet another iterative scheme-the so-called outer or source iterations-necessary to
handle the presence of a fission term. We will study this latter scheme in Section
5-IV.
6. NODAL METHODS
There are many instances in nuclear reactor analysis in which one requires a
full three-dimensional calculation of the neutron flux, for example, in core fuel
depletion or control rod ejection studies. Although a direct numerical solution of
the diffusion equation can be performed on a modern digital computer, it is
extremely expensive to do so, particularly when a series of such calculations would
be required for a parameter study. We desire a scheme for determining the
three-dimensional core flux distribution that avoids the large storage and execution
time requirements of a direct finite difference treatment of the diffusion equation.
Such a scheme is provided by so-called nodal methods. 22 - 23 The general idea is to
decompose the reactor core into relatively large subregions or node cells in which
the material composition and flux are assumed uniform (or at least treated in an
average sense). One then attempts to determine the coupling coefficients
characterizing node cell to node cell leakage and then to determine the node cell
fluxes themselves.
To develop this approach in more detail, consider the neutron diffusion equation
in its general time-independent form given by Eq. (5-143). Now we know that we
can formally write the solution to this equation as
<t,(r)= f d r'G(r,r')S(r'),
3
(5-178)
where G (r, r') is the diffusion kernel or Green's function for the particular geometry
of interest that satisfies
-V · D (r)VG (r,r') + La(r)G (r, r') = S (r- r'). (5-179)
Notice, in particular, that G (r, r') can be interpreted physically as the flux resulting
at a position r from a unit point source at r'.
Of course we usually cannot construct G (r, r')-if we could, we would have
already solved our problem. But suppose we ignore this annoyance for the moment.
We will instead introduce the principal aspect of nodal methods by dividing the
reactor core (or, more typically, one quadrant or octant of the core, since some
symmetry is usually present) into N node cells as shown schematically in Figure
Node cell n
Node cell 11'
FIGURE 5-14. Nodal cell division of a reactor core

5-14. We now integrate Eq. (5-178) over the volume Vn of the nth node cell
(5-180)
If we define the spatial averages over the nodal cells,
(5-181)
(5-182)
(5-183)
then we can rewrite Eq. (5-180) as

N
c/>n = L Knn'sn'' (5-184)
n'= I
or in matrix form
(5-185)
where cf> and S are N-dimension column vectors and K is an N X N matrix.

Notice that the matrix elements Knn' can be interpreted as the probability of a
neutron born in cell n' diffusing to cell n. Hence K is referred to as the nodal
transfer matrix while Knn' are know as the nodal coupling coefficients.
Thus if we know K_, we can easily determine the flux resulting from a given
source S by a single matrix multiplication. But of course we don't know K since we
don't k;:ow G (r, r'). The key to such nodal methods therefore lies in ou;-ability to
approximate or guess the coupling coefficients Knn'·
Of course from a formal point of view, if the number of nodal cells N is large,
the nodal method becomes equivalent to the finite difference scheme and hence
loses any calculational advantages. The real power of the nodal approach is
realized only when the number of node cells N is small, since then the cells are
large enough that they become coupled via neutron diffusion only to nearby
cells-that is, the transfer matrix K is sparse (i.e., it has many zero elements).
However choosing large node cells places the burden of the calculational effort on
an estimate of the coupling coefficients Knn'•
The determination of these coefficients is usually accomplished in a most
empirical fashion (a nice way of saying they are fudged). Typically the Knn' are
determined by assuming a flat nodal source and uniform composition in each cell
and allowing neutron transfer to only the six nearest neighbor cells. The transfer
coefficients are represented as linear combinations of several simple trial functions
(e.g., from one-dimensional slab geometry calculations). The blending coefficients
in this representation are then determined by comparison with more accurate finite
difference benchmark calculations and lots of experience, fiddling, and fudging.
If the transfer coefficients are properly chosen, then such nodal methods can be
extremely useful in generating three-dimensional flux distributions when only
limited accuracy is required. Unfortunately such empirical schemes for choosing
the Knn' are quite problem-sensitive and require a good deal of experience on the
part of the reactor analyst.
We will leave numerical methods for solving diffusion equations until later when
we must generalize these methods to account for fission processes and energy-
dependence. The above discussion has been an admittedly curse description of
numerical methods for solving differential equations. There is a vast literature on
this subject that provides the details of the methods we have so briefly outlined. 12- 16
And perhaps the most valid argument for presenting only a brief sketch of such
numerical methods lies in the recognition that these topics are of such vital
importance to the practicing nuclear engineer, that he almost certainly will have
had or will take further courses on numerical analysis in any event.
III. ONE-SPEED DIFFUSION THEORY

OF A NUCLEAR REACTOR
A. Introduction
Thus far we have studied the diffusion of neutrons in nonmultiplying media
as described by one-speed diffusion theory. We now wish to apply this theory to
the study of nuclear reactors in which fissile material is present. Hence we must
determine how to include nuclear fission in the one-speed diffusion equation (5-16).
To this end, let us first recall the sequence of events involved in a fission chain
reaction. To be specific, we first consider the processes occurring in thermal
nuclear reactors (see Figure 5-15). Fission neutrons are born at high energies in the
MeV range. It is possible that such fast neutrons induce fission in either fissile
(235 U or 239 Pu) or fissionable isotopes (238 U). It is far more likely that the fast
fission neutrons will be moderated to lower energies by elastic scattering collisions
with light moderator nuclei (e.g., :H or 1~C). As the fission neutrons are slowed
down, they pass through energies comparable to the absorption resonances in
heavy nuclei such as 238 U and hence experience an appreciable probability of being
absorbed. They may also leak out of the reactor core during this slowing down
process. In a thermal reactor, however, over 85-90% of the neutrons will manage to
slow down to thermal energies. They will then diffuse about the reactor core until
they either leak from the core or are absorbed. If they are absorbed in the fuel,
then they may induce a new fission, thereby repeating the cycle.
The processes involved in fast reactors are somewhat similar (see Figure 5-16). In
such reactors an effort is made to prevent the fission neutrons from slowing down
before they will have had a chance to induce fission. Low mass number material is
avoided to reduce the energy loss via elastic scattering. However some moderation
will occur, due both to elastic scattering from materials such as oxygen (remember,
most fast reactor fuels are oxides) as well as inelastic scattering from materials such
107
/
106 ~ /
Fissior0 Scatter ®
FissionX
10 5
104 -
-
j I
Moderation
E
103 -
!!
Resonance
Conversion
absorption
10 1
10- 1 Thermal
diffusion _ _.,_Absorption~ Fission
10--2
FIGURE S-15. A schematic of the various processes involved in a thermal reactor
as sodium. In particular there will be a tendency for some of the neutrons to slow
down to the energy range in which appreciable resonance absorption may occur,
although most of the fission reactions will be induced by neutrons with energies
above this range. Hence neutron moderation, leakage, and resonance absorption all
play an important role in fast reactor physics, just as they do with thermal reactors.
It should be evident that the various processes occurring during this sequence are
strongly energy dependent. We will apply our one-speed diffusion model to study
such processes, however. Our motivation is partly pedagogical since this model is
by far the simplest description of nuclear reactor behavior and allows us to
introduce many concepts of nuclear reactor analysis in the simplest possible
framework. However as we have noted earlier, the one-speed diffusion model can
also provide a very useful qualitative description of certain reactor types (notably
very thermal or very fast reactors) provided one uses the correct values for the
cross sections (:i:a, :i:r, :i:1,) which appear in the model.
B. The Fission Source Term

We now direct our attention toward determining a way to include fission in
the one-speed diffusion equation. We will assume that diffusion, absorption, and
fission all occur at the same energy. Then a term to represent fissions can easily be
derived by noting that if ~r<P(r, t) is the fission reaction rate density, then the rate at
which fission neutrons appear in the reactor-that is, the "fission source"-is given
by
(5-186)
10 7
@) ........
,os Scatter - Leakage
Fission'
E ,os !!
Resonance
absorption
- Fission_!
- Breeding
104
103
FIGURE 5-16. Neutron processes involved in a fast reactor
If this is the only source of neutrons in the reactor,t then the appropriate diffusion
equation becomes
-V1 -a
a<1>
t
-v •DV</> + La</>(r, t) = IILr</>(r, t). (5-187)
Note here that we can identify the various components of the macroscopic
absorption cross section which appear in Eq. (5-187) as:
L = Lmoderator + Lstructure + Lcoolant + Lfuel

a a a a a
and
(5-188)
C. The Time-Dependent "Slab" Reactor
l. GENERAL SOLUTION
We will begin our study of nuclear reactor behavior as described by the

one-speed diffusion equation by considering a uniform slab of fissile material
characterized by cross sections La, L 1r, and Lr. This unrealistic appearing "slab
reactor" is chosen to introduce many of the concepts of nuclear reactor analysis,
since its one dimensional geometry greatly facilitates the detailed solution of the
one speed diffusion equation.:t: The appropriate mathematical description of the
t Actually we should hedge here a bit. Equation (5-187) actually represents only "prompt"
neutrons, that is, those born instantaneously in fission. The "delayed" neutrons arising from
fission product decay require a slightly different treatment. We will defer this modification until
Chapter 6.
i In this sense it is somewhat akin to the "vibrating string" or "simple harmonic oscillator"
problems in physics that also get beaten to death since they contain most of the interesting
physics-and yet are easy to solve.
a X =Q a
2 2 FIGURE 5-17. The slab reactor
neutron flux in such a reactor is

2
1 a<j, a </>
- T - D- 2
+La<f>(x,t) = VLr<f>(x,t), (5-189)
V ut dX
with initial condition: </>( x, 0) = <j, 0 ( x) = <j,0 ( - x) (symmetric),
and boundary conditions: <1>(; ,t) = <1>( - ; , t) = 0.

Notice that we have assumed that our initial flux is symmetric. We will find later
that such an assumption will imply similar symmetry for all times, <j,(x, t) =
</>( - x, t). This will simplify our manipulations somewhat.
Unlike our earlier studies of time-independent neutron diffusion, we are now
faced with a partial differential equation to solve. There are a number of ways to
attack such equations, but perhaps the simplest is to use separation of variables 1- 3
by seeking a solution of the form
</>( x, t) = If' (x) T ( t). (5-190)
If we substitute this form into Eq. (5-189) and divide by it,(x)T(t), we find
(5-191)
Here we have noted that since we have a function only of x set equal to a function
only of t, both terms must in fact be equal to a constant. We have named this
constant - A. However A is as yet unknown.
Hence the separation of variables given by Eq. (5-190) has reduced the original
partial differential equation in two variables to two ordinary differential equations:
dT
dt = -AT(t),
d1it, A
D -2 +(vLr-La)it,(x)= - -1/;(x). (5-192)
dx v
We can easily solve the time-dependent equation
(5-193)
where T(O) is an initial value which must be determined later. To solve the
space-dependent equation, we must tack on the boundary conditions:
dx2
~
D d2i/; + ( + vLr- La)iti(x)=O,
v
Boundary condition: i;i( ~) = i;i( - ~) = 0. (5-194)
Here ;\ is still to be determined. However we recall the eigenvalue problem
(5-195)
has symmetric solutions (we are only interested in symmetric solutions since cpo(x)
is symmmetric):
eigenfunctions: ilin(x) =cosBnx

(5-196)
eigenvalues: n= 1,3,5, ...
If we identify Eq. (5-194) as the same problem, it is apparent that we must choose
n= 1,3,5, ... (5-197)
These values of ;\n are known as the time eigenvalues of the equation, since they
characterize the time decay in Eq. (5-193). The general solution to Eq. (5-189) must
therefore be of the form
(5-198)
This solution automatically satisfies the boundary conditions. To determine the An,
we use the initial condition to write
Initial condition: cp(x,0)=cp0 (x)= ~Ancos mr_x. (5-199)

n a
odd
Using orthogonality, we find

i1
2
An= a f- 2
()
'!_dX</>o X cos
mrx
T. (5-200)
2
Thus we have found that the flux (for any symmetric initial distribution) can be
represented as a superposition of modes, each mode weighted by an exponential
factor:
lJ
~ [ ~ f_
2
</>(x,t)= !!._dx' <1>o(x')cosBnx']exp( -Ant)cosBnx, (5-201)
2
odd
where the time eigenvalues An are given by
(5-202)
Before we proceed to examine this solution in more detail, it is useful to make a

few comments about the separation of variables approach. First notice that the
separation parameter that arose was in fact identified as an eigenvalue. Thus
separation of variables is essentially equivalent to an eigenfunction expansion.
Indeed if we had the foresight to expand the spatial dependence of the flux in the
eigenfunctions for the slab (using symmetry to restrict this expansion to odd n),
</>( x, t) = "" mrx ,

~ Tn (t) cos --- (5-203)
n a
odd
where we have noted that the expansion coefficients now must be time-dependent,
then we could have immediately arrived at an equation for the Tn
(5-204)
by substituting Eq. (5-203) into the original equation (5-189) and using the
orthogonality property of the eigenfunctions. This alternative approach is
frequently useful when encountering problems in which sources are present, be-
cause the separation of variables approach we have presented applies only to
homogeneous equations.
Finally, note that although we initially sought solutions If/( x) T ( t) which were
separable in x and t, these solutions were eventually superimposed to yield a
nonseparable function of space and time [cf. Eq. (5-201)]. Hence separation of
variables does certainly not imply a separable solution. Interestingly enough,
however, there is one very important situation in which such separability will occur,
that involving the behavior of the neutron flux for very long times.
2. LONG TIME BEHAVIOR

Notice that one can order Bf< B] < · · · B; = (mr / a)2 · · · . Hence the time
eigenvalues must similarly be ordered such that A1 <A 3 <A 5 < A7 < · · · . This means
that the modes corresponding to larger n decay out more rapidly in time. If we wait
long enough, then only the fundamental mode remains:
as t-HXJ. (5-205)
This implies that regardless of the initial shape of <f> 0(x) the flux will decay into the
X
FIGURE 5-18. Time decay of higher order spatial
modes in the slab reactor
fundamental mode shape. Of course, we have implicitly assumed that A I will not be
zero. The coefficient of the fundamental mode is just
2 ;:,
- wx '
A 1 =-::-
a f 2_dx'</>0 (x')cos-_-.
a
2
a
(5-206)
Since </>0 (x) must be nonnegative in the slab to represent a physically realizable
flux, then it is apparent that A 1 > 0.
Actually for sufficiently large Lr, - An may be positive corresponding to an
exponentially growing flux. However the same argument will hold since -,\ > -;\ 3
> · · · . Hence regardless of whether the flux grows or decays, it will eventually
approach a "persistent" or fundamental cosine distribution.
It is customary to refer to the value of B;
characterizing this mode as
2
Bf= ( ; ) _ B82 = geometric buckling. (5-207)
This nomenclature is used since B} is a measure of the curvature of the mode

shape
(5-208)
Since there will be a larger current density J and hence leakage induced by a mode
with larger curvature or buckling, we might expect that the mode with least
curvature will persist in time the longest.
3. THE CRITICALITY CONDITION

Let us now see what is required to make the flux distribution in the reactor
time-independent-that is, to make the fission chain reaction steady-state. We will
define this situation to be that of reactor criticality:
Criticality=when a time-independent neutron flux can be sus-

tained in the reactor (in the absence of sources
other than fission).
Notice that we have qualified this definition by specifically demanding that the flux
be time-independent in the absence of a source. As we have seen in Chapter 3 (and
will see later in the problem set at the end of this chapter), a source present in a
critical system will give rise to an increase in flux that is linear in time.
If we write out the general solution for the flux
00
<f>(x,t)=A 1 exp(-;\ 1 t)cosB 1x+ ~ Anexp(-;\nt)cosBnx, (5-209)

n=3
n odd
it is evident that the requirement for a time-independent flux is just that the
fundamental time eigenvalue vanish
(5-210)
since then the higher modes will have negative ;\n and decay out in time, leaving
just
(5-211)
If we rewrite this "criticality condition" using the notation Bf= Bf, then we find
we must require
(5-212)
It has become customary to refer to

VLr- La . .
D - B~= material buckling (5-213)
since it depends only on the material composition of the reactor core (whereas Bf
depends only on the core geometry). Hence our criticality condition can be written
very concisely as
(material composition) B~ = Bf (core geometry). (5-214)
Thus to achieve a critical reactor, we must either adjust the size (Bf) or the core
composition (B~) such that B~ = B;. We also note
B~ > Bf =;,;\ 1 < 0 =;, supercritical,

Bm2 = Bg2 =;, ;\ I = 0 =;, critical, (5-215)
B~ < Bf=;, ;\ 1 > 0 =;, subcritical.
In particular notice that by increasing the core size we decrease Bf, while by
increasing the concentration of fissile material we increase Lr and hence B~. Both
of these modifications would therefore tend to enhance core multiplication.
Yet recall that in Chapter 3 we expressed the criticality condition in terms of the
multiplication factor k. We can make the connection between these two criteria if
we write the time eigenvalue as
(5-216)
Now recall that ( vL)- 1 is the mean lifetime for a given neutron-nuclear reaction to
occur. Hence ( vLa)- 1 must be just the mean lifetime of a neutron to absorption
(ignoring leakage-that is, in an infinite medium). Furthermor.e, we can identify
l'Lr VLFuel LFuel

- = - - r_ _a _ _ 1JJ-k (5-217)
L .._,Fuel L - - oo·
a .;,a a
Now the only remaining task is to identify (1 + L 2B;)- 1• Recall that the rate of
neutron leakage is given by
rate s v
f.
Leakage= dS ·J = d 3rV ·J = - f. f.vd r D'i1
3 2
<f,, (5-218)
where we have used both Gauss's theorem and the diffusion approximation. Hence
we can write
Rate of neutron absorption

----,-----,-----~----=----------
Rate of neutron absorption plus leakage
However we can identify this ratio as just
1
Nonleakage probability= PNL = (5-220)
1 + L 2Bg2
Therefore we can interpret
1 Bn = p (-1-) =
( VLa)( 1 + L 2 NL VLa
1 _ neutron lifetime
in a finite reactor,
(5-221)
since we have just reduced the lifetime to absorption in an infinite medium to take
account of neutron leakage. If we now combine Eqs. (5-217) and (5-220), we find
that the multiplication factor k for this model becomes just
(5-222)
Thus we can identify our fundamental time eigenvalue as just the inverse of the
reactor period
k-1 1
-A,=-1-= r· (5-223)
If we also recall from Eq. (5-202) that
(5-224)
then it is apparent that the various forms of the criticality condition are indeed
equivalent:
(5-225)
In particular notice that by using PNL = (1 + L 2B}-)- 1 we can avoid the analysis of
the initial value problem and proceed directly to the criticality condition
(5-226)
We will return later to consider how these results can be applied to reactor
criticality studies, but first we will extend them to more general reactor geometries.
D. The Criticality Condition for More General Bare Geometries

Note that the only quantity characteristic of the reactor size or geometry that
appears ink or PNL is the geometric buckling, B;.
For the case of a slab reactor of
width a we found B}-=(Tr/a,)2. We might suspect that for more general geometries
we need only replace this by the geometric buckling characterizing the specific
geometry under consideration. This suspicion is in fact easily verified, but only for
so-called "bare" geometries in which the reactor composition is uniform. For the
more complicated multiregion geometries, such as reactors composed of a core
surrounded by a reflecting material, one can no longer derive simple expressions
for PNL or k in terms of the reactor geometry and composition.
Consider, then, a bare reactor of uniform composition surrounded by a free
nonreentrant surface characterized by vacuum boundary conditions. If the reactor
is critical then the neutron flux must satisfy the steady-state diffusion equation
(5-227)
subject to the boundary condition </>(is)= 0 for is on the extrapolated surface. Of

course in general there will be no solution to this equation unless we have
happened to hit on just the right combination of composition and system size.
To see this more clearly, divide Eq. (5-227) by - D so that it can be written as
v2<1>+( V~r;~a )<t>(r)=O, (5-228)
boundary condition: </>(is)= 0.

Sometimes Eq. (5-228) is written in a somewhat different form as
v2</> + ( k:~ 1 )<t>(r) = o, (5-229)
boundary condition: </>(is)= 0.

Now notice that this equation is identical to that which generates the spatial
eigenfunctions for this geometry
(5-230)
boundary condition:
We know that this latter equation has nontrivial solutions lfn(r) only for certain
values of the parameter B 2, the eigenvalues B;. Hence by comparing Eqs. (5-228)
and (5-230) we find that the steady-state diffusion equation for the flux cp(r) will
only have nontrivial solutions when the core composition is such that ( v Lr - La)/ D
is equal to an eigenvalue B;, and then the flux cp(r) will be given by the
corresponding eigenfunction lfn(r).
However since there are an infinite number of possible eigenvalues B;, we might
be tempted to think that there are an infinite number of values (PLr- La)/ D B;,
for which the reactor is critical. However it should be recalled that in the case of a
slab geometry, only the lowest eigenvalue Bf= ('fl/ a)2 _ B; had a corresponding
eigenfunction i/; 1(x) = cos1Tx / a that was everywhere positive. The eigenfunctions or
spatial modes lfn(x) corresponding to higher eigenvalues oscillated about zero. This
same feature also characterizes the eigenfunctions lfn(r) of more general geometries.
Only the eigenfunction if; 1(r) corresponding to the smallest eigenvalue Bf is
everywhere nonnegative. Since the neutron flux can never be negative, it is
apparent that the only solution to the eigenvalue problem Eq. (5-230) physically
relevant is that corresponding to the smallest eigenvalue, Bf- B;. Hence for the
reactor to be critical we require
(5-231)
just as for the slab.

Thus we can continue to use PNL =(I+ L 2B;)- 1 and k=(vLr/LJ(l + L 2B;)- 1
for more general bare geometries provided we identify the geometric buckling B;
as the smallest eigenvalue Bf of the Helmhotz equation
(5-232)
subject to the boundary conditions that cp(r) vanish on the extrapolated boundary
of the reactor. The corresponding critical flux distribution cp(r) is then given by the
fundamental eigenfunction if; 1(r), which is everywhere nonnegative.
It should be pointed out that although the Helmhotz equation (5-232) will
provide us with the flux shape in a critical reactor, it will tell us nothing about the
magnitude of the flux. Since it is a homogeneous equation, if cp(r) is a solution, then
any multiple of cf>(r) is also a solution. Of course the magnitude of the flux was
determined for us by the initial condition cp 0(r) when we studied reactor criticality
by solving the full time-dependent diffusion equation (5-187). However this latter
approach is far too cumbersome to use in practice.
Instead we merely note that a critical reactor can operate at any flux level-at
least mathematically. (Of course one must provide for adequate core cooling,
shielding, etc., but these factors are extraneous to our present model of the reactor
so we won't worry about them here.) Hence we will merely assume that the
magnitude of the flux is determined by the desired thermal power output of the
core. If the usable energy produced per fission event is wr, then the thermal energy
deposited in the core per unit volume per second is just given in terms of the fission
reaction rate density as
(5-233)
This is just the local thermal power density at position r in the core. Hence the total
power generated by the core is just the integral of the power density over the core
volume
(5-234)
This relation can be used to determine the magnitude of the flux in terms of the
core thermal power level.
Thus we now have developed a rather simple scheme to study the criticality of a
nuclear reactor-at least a bare, uniform reactor. The only mathematical effort
required is the solution of the Helmhotz equation characterizing the geometry of
interest for the geometric buckling Bf (the fundamental eigenvalue Bf) and the
critical flux shape <f>(r) [the fundamental spatial eigenfunction l{, 1(r)]. To illustrate
how these quantities are determined, we will consider a simple yet very important
example:
EXAMPLE: A Right Circular Cylindrical Core

The most common reactor core shape is that of a right circular cylinder of height
H and radius R. (Actually a sphere would be the more optimum geometry from the
aspect of minimizing neutron leakage, but spheres are very inconvenient geometries
to pass coolants through.) The appropriate form of the Helmholtz equation is then
1 a a</> a2<1>
--r-+-+ B 2</>(r,z)=O, (5-235)
r ar ar az 2
subject to boundary conditions
Since this is a homogeneous partial differential equation, we can seek its solution
using separation of variables
</>(r,z) = '!R,(r) '.:l(z).
Then if we substitute this form into Eq. (5-235), we arrive at two ordinary
I
H
l FIGURE 5-19. Finite cylindrical reactor core

differential equations
1 d d6A 2 -
-;: dr' dr +a 6A(r)=0, 6A(R)=0, (5-236a)
~z<f; +;\2?l(z)=0, ?l( ± f )=o, (5-236b)
where the separation constants a 2 and ;\ 2 are constrained by the relationship

B 2 = a 2 + ;\ 2 . Each of these equations represents a separate eigenvalue problem that
one can use to determine a and ;\ (and hence B 2). The eigenfunctions and
eigenvalues of the axial equation are well known to us
?lnC z) = cos ( n1rz)

iI ' n=l,3, ... (5-237)
To construct the eigenfunctions of the radial Eq. (5-236a), we first identify its
general solution in terms of zeroth order Bessel functions (see Appendix D)
(5-238)
Since Yo(ar) ➔ oo as r➔ 0, we must set C=0. Applying our boundary condition at

r= R, we find
(5-239)
where Pn are the zeros of J 0 • In particular, the smallest such zero is Po= 2.405 ...
(kind of like 1r to a Bessel function). Hence we find the eigenfunctions and
eigenvalues generated by the radial equation (5-236a) are just
n=0, 1, ... (5-240)
Therefore, consistent with our prescription of seeking the smallest value of B 2 as

our geometric buckling, we find
B;= ( PoR )l + (iI7T )2 ' (5-241)
corresponding to a spatial flux shape
<f,(r,z)=AJ0 ( Por)
R cos ( 1rz)
iI . (5-242)
Since this is the geometry most frequently encountered in reactor design, it is

useful to calculate the normalization factor A in terms of the core power level P by
noting
(R
P= fvd rwcLc<f>(r)=wcLc21rA Jo drrJ 0
3 ( P0 r)
R
f:_il dz cos
(iI
1rz)
=
WcLcA4 VJ 1(P 0 )
7TPo .
2
(5-243)
Thus we find
(5-244)
It should perhaps be mentioned that since reactor cores are fabricated from either
square- or hexagonally-shaped fuel assemblies, one can only approximate such
cylindrical geometries. However for most purposes one can assume the reactor core
is essentially a right circular cylinder.
One can proceed in a very similar manner to analyze other bare core geometries.
For convenience, we have tabulated the geometric buckling and critical flux profile
in other common geometries in Table 5-1.
TABLE S-1 Geometric Bucldings and Critical Flux Profiles Characterizing Some
Common Core Geometries
Geometric Buckling B; Flux profile
Slab
• (g)2 'TTX
COS-;;-
a
Infinite
Cylinder
m (~f lo cor)
R
-
Sphere
• 1C
Cir r-1 •
sm ('1RTr)
r)
Rectangular
Parallelepiped ~~ 1 (ir +(; r+(fr cos( 'TT; ) cos( cos( w; )
I-a~)
Finite
Cylinder
ra1l (~f +(;r 10 cor)
R cos ('TTZ)
ii
The reader should not be deceived into believing that such criticality calculations
are always so straightforward. For we must remember that the expression we have
derived for the nonleakage probability PNL= ( 1 + L 2Bg2)- 1 holds only for uniform,
bare reactor geometries (i.e., single-region cores). As we will find later, it is no
longer possible to derive simple expressions for PNL or k in terms of the reactor
geometry and composition for multiregion (e.g., reflected) reactors.
These results can be used to determine the core geometry or composition that
210 THE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
will yield a critical reactor. For example, if the material composition is specified,
one can compute the material buckling B~ in terms of the macroscopic cross
sections using Eq. (5-213). Then by using the criticality condition B~ = B; (along
with Table 5-1), one can infer the core dimensions that will yield a critical system.
In the more usual situation the nuclear designer will be given B; (rather than B~)
since the core dimensions are determined by limitations on core thermal perfor-
mance (and not nuclear considerations). That is, the core must be built a
sufficiently large size to avoid excessively high temperatures for a desired power
output. The nuclear designer must then determine the fuel concentration or loading
(i.e., B~) that not only will result in a critical system, but will also allow the core to
operate at a rated power for a given time period.
EXAMPLE: As a specific example we will study the one-speed diffusion model

of a bare, homogeneous cylindrical reactor with material composition representa-
tive of that of a modern PWR such as described in Appendix H. We will use
"homogenized" number densities corresponding to a PWR core operating at full
power conditions and containing a concentration of 2210 ppm of natural boron (as
boric acid) dissolved in the water coolant for control purposes. The fuel is taken as
UO 2 enriched to 2.78% 235 U. In Table 5-2 we have listed the number densities and
microscopic one-speed cross sections for this core. (Here the cross sections are
actually averages over the neutron energy distribution in such a reactor core.)
TABLES-2 Number Density and Microscopic Cross Sections
Material N(l/b-cm) a,r(b) a.(b) ai(b) V
H 2.748 X 10- 2 0.650 0.294 0 0

0 2.757X 10- 2 0.260 l.78X 10- 4 0 0
Zr 3.694X 10- 3 0.787 0.190 0 0
Fe l.710X 10- 3 0.554 2.33 0 0
235U l.909x 10- 4 1.62 484.0 312.0 2.43
23su 6.592X 10- 3 1.06 2.11 0.638 2.84
13sxe 0.000 1.21 2.36 X 1()6 0 0
JOB 1.001 X 10- 5 0.877 3.41 X 10- 3 0 0
This data can be used to calculate the macroscopic cross sections tabulated in
Table 5-3. Here we have also included the relative absorption rates in each material
which serve as a measure of neutron balance within the core.
TABLES-3 Macroscopic Cross Sections
Material ~,,(cm- 1) ~.(cm- 1) v~i(cm- 1) Relative Absorption

H l.79X 10- 2 8.08X 10- 3 0 0.053
0 7.16 X 10- 3 4.90X 10- 6 0 0
Zr 2.91 X 10- 3 7.01 X 10- 4 0 0.005
Fe 9.46X 10- 4 3.99X 10- 3 0 0.026
235U 3.08x 10- 4 9.24X 10- 2 0.145 0.602
23su 6.95X 10- 3 l.39X 10- 2 l.20X 10- 2 0.091
JOB 8.77x 10- 6 3.41 X 10- 2 0 0.223
3.62X 10- 2 0.1532 0.1570 1.000
It should be noted that the transport cross sections used in this example have
been artifically adjusted (reduced by almost an order of magnitude) to take some
account of fast neutron leakage which would normally not be described by a
one-speed model.
We can use these cross sections to calculate a number of important parameters
characterizing the PWR core:
Diffusion coefficient: D = 9.21 cm

Infinite multiplication constant: koc, = v~r/~a = 1.025
Material buckling: B~=(v~r-~a)/ D=4.13X 10- 4 cm- 2
Extrapolation distance: z 0 =0.7L\r= 19.6cm
Leakage fraction for a critical core: 1- PNL =0.025 (5-245)
Next we will compute the critical core dimensions. If we assume that the core
height is fixed at 370 cm by thermal considerations, then we can determine the
radius at which a core with such a composition will be critical. First calculate the
axial buckling B}
(5-246)
Then using Eq. (5-241), we can determine the radial buckling B,2
(5-247)
Hence we can solve for the critical radius as
R= Vo) -z0 = 108 cm.

~ ( Bz (5-248)
(It should be noted that this is somewhat smaller than the radius of 180 cm for a
typical PWR core. This illustrates the limitation of such a one-speed model for
obtaining quantitative estimates in reactor analysis.)
E. Reflected Reactor Geometries

To illustrate the complications that arise with multiregion core geometries, we
will return to our slab reactor and add a reflector of nonmultiplying material of
thickness b to either side (see Figure 5-20). For the purposes of this analysis we will
characterize the reactor core by superscript "C" and the reflector by "R." Rather
than repeat our earlier analysis of the initial value problem for this geometry, we
will proceed directly to examine the time-independent diffusion equations that
must be satisfied by the fundamental mode flux shape. As in our earlier analysis of
the nonmultiplying reflected slab, we will seek a solution in each region of the
x=O a _.!! +b
2 2
FIGURE 5-20. Reflected slab reactor
reactor and then use interface conditions to match these solutions. Hence we must
solve
Core:
(5-249)
a a ~
Reflector: -<x<-+b
2 2 '
subject to the set of boundary conditions:
(a) </>C( i) =<J>R( i)

(b) JC(!)=JR(!)
(c) </>R( i + E) = 0.
Note that we have used the reactor symmetry to narrow our attention to the range
of positive x. As we noted earlier in Section 5-111-D this problem will have no
solution unless we choose the proper combination of core composition and size.
We would anticipate that a criticality condition relating these core characteristics
would emerge in the course of our analysis.
The general approach, as always, is to determine the general solutions in the core
and reflector and then use the boundary conditions to determine the unknown
coefficients. In the core the general solution will be
(5-250)
where we have utilized the symmetry of the core to discard the sine term. Here the
material buckling characterizing the core is defined by
(5-251)
while in the reflector we will seek a solution satisfying the vacuum boundary
condition (c)
(5-252)
where the reflector diffusion length LR=(DR/~~)½. We now apply the interface
boundary conditions to find
T b ),
Accos ( Bea) =ARsinh ( LR
(5-253)
Dividing these expressions, we can cancel Ac and AR to find
C C Bma
C ) - DR ( b~ )
D Bm tan ( - - - LR coth LR . (5-254)
2
Notice that this equation represents a relation between reactor composition

(Dc,B;,DR,LR) and size (a,b) that must be satisfied if a solution to the steady-
state diffusion equations (5-249) is to exist. Hence this is just the reactor criticality
condition for this particular geometry. Admittedly, it doesn't look anything like our
earlier condition, B~ =Bf, that characterized a bare reactor. In fact the criticality
condition for a reflected reactor is transcendental-one cannot obtain an explicit
solution for the critical size or composition. Instead, either numerical or graphical
techniques must be used. The latter technique is more useful for our present
discussion. Rewrite Eq. (5-254) as
(5-255)
If we plot the LHS against (B;a/2), we can then determine the solution of this
transcendental equation graphically by noting where it intersects the value of the
RHS, as shown in Figure 5-21. [Actually since there will be many such intersec-
tions, we are only interested in the lowest value of (B;a/2).] From this graph we
notice that the critical value of B; must be such that
B;a 7T
-2< -
2
or (5-256)
in contrast to the bare (unreflected) core in which B; 2

= ( 7T / ii)2. Hence we see that
the width a required for criticality is somewhat smaller when a reflector is added,
but we would expect this since a reflector is added primarily to reduce neutron
leakage.
It is conventional to define the difference between bare and reflected core
dimensions as the reflector savings 8:
8=a (bare)-a (reflected). (5-257)

RHS
1T I
21
I
I
I FIGURE 5-21. Graphical solution for
I reflected reactor criticality
For example, the reflector savings for our slab reactor can be written as
(5-258)
For a thick reflector b »LR this simplifies to
(5-259)
which is essentially a measure of the maximum reflector savmgs that can be

realized.
Reflectors serve another function besides reducing neutron leakage. They tend to
flatten the flux and hence the power distribution in the reactor core. Unfortunately
one-speed diffusion theory is not adequate to describe this effect, which results in a
peaking of the thermal flux in the reflector region (see Figure 5-19), so we must
defer a further discussion of reflected cores until we have developed multigroup
diffusion theory.
IV. REACTOR CRITICALITY CALCULATIONS
A. Introduction
Let us now turn to the very important topic of determining the composition
or size of a reactor that will yield criticality. It should be apparent after the last
example that in most practical reactor designs one cannot simply determine the
geometric buckling for a core geometry and then use B~ = Bf to arrive at critical-
ity.
One "brute force" procedure would be to determine the lowest time eigenvalue
of .- ( i 911
µc 1;.....-15 <.., ' ;\
-V ·D (r)Vr/> + (~a - P~r)rf>(r) = -rj>(r), (5-260)
V
and then keep adjusting things until ;\ = 0. However this is rather awkward, and
moreover would tend to introduce errors in an unnatural manner when we

generalize our analysis to include energy dependence.
Instead suppose we write our diffusion equation as
-V · DV</> + ~a<f>(r) = P~r</>(r), (5-261)

boundary condition: <f>(r.) =0
(which, of course, is the steady-state equation we solved analytically in the earlier

simple examples). Unfortunately this equation has no solution in general-unless
we just happen to hit on the exact combination of core composition and geometry
such that the reactor is critical (a highly unlikely possibility on a computer).
What we can do is introduce an arbitrary parameter "k" into this equation as:
(5-262)
Then for some value of k, we assert that this equation will always have a solution.
The idea is to pick a core size and composition and solve the above equation while
determining k. If k should happen to be unity, we have chosen the critical size and
composition. If k=fa l, however, we must choose a new size and composition and
repeat the calculation. As one might expect, k turns out to indeed be the multipli-
cation factor we defined earlier in Chapter 3, as we will demonstrate later.
We could give a formal mathematical proof that Eq. (5-262), or its generaliza-
tions will always have a solution for some k, but it is more convenient to simply
argue physically that since varying k will vary the effective fuel concentration
NF-NF/ k, one can always achieve a critical system by making k sufficiently
small.
Sometimes a slightly different formulation is used in which one pretends that P,
the number of neutrons emitted per fission, is in fact variable. (Of course it isn't,
but it is a useful device to regard it as adjustable for the moment.) Now physically
we know that there must be some value of P, call it Pc, that will yield a nontrivial
solution to
(5-263)
regardless of what composition or geometry we have chosen. Hence the idea is to

determine this Pc, then readjust composition and geometry until we have forced
actual = P •
Pc--'> P (5-264)
If we compare this approach to our earlier scheme in which we calculate k, it is

evident that
(5-265)
From a mathematical point of view, each of these approaches introduces a new

parameter into the steady-state diffusion equation, either k or Pc, which can then
be regarded as an eigenvalue in a subsequent analysis. Once this eigenvalue has
been calculated, one can return and readjust composition and geometry in an effort
to force this eigenvalue to a desired value (e.g., k--'>l or Pc--'>P). Hence the criticality
calculation is converted into a sequence of eigenvalue problems for the criticality

eigenvalue (sometimes also called the multiplication eigenvalue).
Of course, in general there will be a set of criticality eigenvalues kn correspond-
ing to the eigenvalue problem represented by Eq. (5-262). For example, our earlier
analysis for the bare slab reactor indicated the existence of the set of eigenvalues:
kn=(PLr/L.)(l+L 2B;)- 1 where B;=(mr/ii)2, n=l,2 .... Only the largest such
eigenvalue (in this case, k 1) will correspond to an everywhere-nonnegative flux
distribution cp(r) and hence to a critical reactor configuration. We will refer to the
largest criticality eigenvalue k 1 as the effective multiplication factor and denote it by
keff· As we will see below, keff can be identified as the multiplication factor k for the
reactor core defined earlier in terms of fission neutron generations in Chapter 3.
B. Numerical Criticality Searches

We have seen in Section 5-111 how one can obtain a criticality condition for a
bare, uniform reactor. Let us now see how the criticality search is conducted in
practical reactor calculations in which numerical methods must be used to solve the
one-speed diffusion equation. To simplify our manipulations, let us first rewrite the
criticality eigenvalue problem (5-262) in operator notation as
1
Mcp= k Fcp, (5-266)
where we identify
M 0
- -V •D (r)V + La(r)
0 0 Destruction operator
(leakage plus absorption)
F0 PLc(r) 0 = Production operator
(fission)
We will leave the boundary conditions on cp(r) as understood.

Of course in any numerical solution, finite-difference methods will lead to a
representation of the neutron diffusion equation (5-266) as a matrix eigenvalue
problem for the eigenvalue k- 1• The solution of such eigenvalue problems can be
accomplished using a common technique from numerical analysis known as the
power method. We will introduce this scheme using physical arguments.
First notice that if we assumed that the "fission source" term S - F<t, on the RHS
of Eq. (5-266) was known, then the remaining part of the equation would be
effectively just the diffusion equation for the neutron flux resulting from this source
in a nonmultiplying medium. We presumably already know how to solve this
problem (cf. Section 5-11), but we do not really know the fission source Fcp since it
involves cp itself. Hence we will do the next best thing and try to guess it by making
an initial estimate of
(5-267)
We next solve for the flux ,p< 1l resulting from this source estimate:
(5-268)
using our earlier procedures. With this solution, we can now explicitly calculate the
fission source resulting from this flux <1>< 1> as
(5-269)
.This can then be taken as a new estimate of the fission source and used to generate
a new flux, </> <2>, and so on-provided we can also generate improved estimates of
k. That is, we can iteratively solve for an improved source estimate s<n+I) from an
earlier estimate S (n) by solving
(5-270)
for </>(n+ 1> and then computing
(5-271)
However we also need a prescription for generating improved estimates of k<n>.

This prescription can be obtained by returning to our original eigenvalue
problem (5-266). As n becomes large, we would anticipate that (if our fission source
iteration scheme really works), </>(n+ 1> will converge to the true eigenfunction <f>(r)
that satisfies Eq. (5-266). That is, for large n
(5-272)
The convergence of </>(n) to <f>(r) and k(n) to k can be proven mathematically. It can
also be motivated physically by recognizing that if indeed we have adjusted k such
that a steady-state or self-sustaining flux profile were possible, then regardless of
the initial fission source estimate, successive fission neutron generations will
eventually fall into this distribution.
Now for finite n, it is highly unlikely that </>(n+I) and k(n+I) will satisfy Eq.
(5-266) exactly. Nevertheless if we integrate Eq. (5-272) over all space, we should be
able to obtain a reasonable estimate for k<n + 1> as
fd3r F</>(n+ I)
k<n+o _ _ _ _ _ __
(5-273)
f d3rM<f>(n+I)
However F</>(n+ 1>is just the (n + l)st estimate of the fission source, while we can use
Eq. (5-270) to write M</>(n+I) in terms of the nth estimate of this source to find
(5-274)
We can now use this relationship to compute a new guess of k<n+I) from </>(n+l) and
k<n>.
We should note that this prescription is quite consistent with our earlier in-
terpretation of k as the multiplication factor-that is, the ratio of the number of
neutrons in two consecutive fission generations-if we note that a factor of k(n)
must be inserted in the denominator since [k(n)r Is (n) is in fact the effective fission
source that generates S (n + I)_
We can now use Eqs. (5-270), (5-271), and (5-274) as the basis of an iterative
algorithm to determine both k and cf>. For large n, we expect that cf> will converge to
the fundamental eigenfunction of Eq. (5-266) corresponding to the largest eigen-
value keff to which k(n) converges. (Recall we have agreed to denote this largest
eigenvalue by keff·) In practice one continues this iteration until the error in k
and/ or S decreases below some specified amount:
s<n)_ s<n-1) I
and/or max <)
Sn < E2 • (5-275)
r
I
Notice that by scaling the source term appearing in the diffusion equation (5-266)
by a factor of 1/ k(n) in each iteration, we will prevent the rapid growth or decrease
of successive source iterates (causing possible overflow or underflow) in the event
that a number of iterations are required when k is not close to unity. That is,
dividing the source term by k(n) removes the dependence of the flux iterate cp<n+ I)
on n [at least as cp<n+ I) approaches the true solution].
This iterative scheme to determine the effective multiplication factor keff and the
corresponding flux cp(r) is known as the power iteration or source iteration method.
The iterations themselves are known as outer or source iterations.
In addition to such outer or source iterations, one will also be required to
perform inner iterations to solve the diffusion problem
(5-276)
when two- or three-dimensional calculations are necessary. The general strategy

then takes the form sketched in Figure 5-22.
C. Source Extrapolation
Needless to say, there is strong incentive to perform as few iterations as
possible in converging to the desired accuracy. For that reason, one usually
attempts to accelerate the source iteration convergence by extrapolating ahead to a
new source guess. This is accomplished by introducing an extrapolation parameter
(much as is used in relaxation methods). For example, in a one-parameter extrapo-
lation, one would use as the source definition ·
(5-277)
A two-parameter extrapolation takes the form
The extrapolation parameters a and /3 range between O and 1, and can be chosen
by using methods based on Chebyshev polynomial interpolation. 13
Guess core geometry

and composition
--
Guess initial fission

source s<0>and k{O)
''
Inner
iterations
-
/
M</>(n+l) = _I_s(n)
k(n)
--
Critic ality
s{n+l) = F</>(n+l) sea rch
k(n+l) =
f d3rs(n+ll(r)
_l_f
k<n>
d3rs<n!(r)
Outer
iterations
,
'
Convergence test
lk(n)_k(n-1) I <t & I s<n!_s<n-1)1

<t2"
-No- '
k(n) 1 s<n!
I
Yes
Yes
FIGURE 5-22. Calculation strategy for reactor criticality calculation
V. PERTURBATION THEORY
It is frequently of interest to compute the change in core multiplication

caused by a small change in the core geometry or composition. Fortunately if this
change or "perturbation" is sufficiently small, one does not have to repeat the
original criticality calculation, but instead can use well-known techniques of
perturbation theory to express the corresponding change in multiplication in terms
of the fluxes characterizing the unperturbed core.
By way of example, consider a very simple one-speed diffusion model of a bare,
homogeneous reactor in which the criticality relation is
PLr/La
k = -2 -2 = I. (5-279)
1 +L B
Now suppose we were to uniformly modify or perturb the absorption cross section
to a new value
(5-280)
where we will assume that the perturbation 8 ~a is small-that is,
(5-281)
Then the value of k' corresponding to the perturbed core can be written as
(5-282)
where we have expanded k' in 8 ~a/~a and have neglected all terms of higher than
first order in the perturbation ( 8 ~J. This has allowed us to express the perturbed
multiplication factor k' in terms of the unperturbed multiplication k and the
perturbation 8~a·
These general features appear in applications of perturbation theory to more
general problems in nuclear reactor analysis in which the perturbations may be
localized or in which the multigroup diffusion equations are used as the basic
model of the core behavior. Although the general ideas are essentially as simple as
those in the example above, it is necessary to introduce a few mathematical
preliminaries. (For more details, the reader is referred to Appendix E.)
We will describe the multiplication of the core by the criticality eigenvalue
problem [Eq. (5-266)]:
(5-283)
where we will leave it as understood that the solution of this equation, <f,(r), must
satisfy appropriate boundary conditions such as <f,(r,) = 0 on the surface of the core.
Now suppose we define the inner product (f,g) between any two functions f(r)
and g(r) as
(5-284)
where f"(r) denotes the complex conjugate of J(r), and Vis the core volume.
We can now use this inner product to define the operator Mt adjoint to the
operator Mas that operator Mt for which
(M'.f,g)=(f,Mg) (5-285)
for every J(r) and g(r) satisfying the boundary conditions J(i,) = 0= g (i.).
We can use this definition to explicitly construct the adjoint of an operator.
Consider for example the operator F = P~f° which simply corresponds to multiply-
0
ing a function by v~r(r). If we write
(f,Fg)= fvd 3rf"v~cg= fvd 3r(v~cf)*g
= (v~cf,g) = (F1/,g), (5-286)
where we have merely shuffled ~r(r) around in the integral (noting that ~r is real)
to identify
(5-287)
Notice that in this case, pt and Fare in fact identical. We refer to such operators
as being self-adjoint.
For a more complicated example, consider the spatial derivatives in the diffusion
operator M: o,
I
v-
3
(f, V·DVg) = fvd r.f"'V·!!_~~- (5-288)
b
Now if we use the vector identity
V·ab=aV·b+b·Va, (5-289)
I
we can rewrite this as \ ' j) \l J . vf V

(f, V·DVg)= ( d 3rV,[rf)iJ~- ( d 3r[Vfj•(DV···g·J. > (5-290)
Jv 1:: VJ Jv I' ~,:;.-
Ct kJ {IL" IJ
1
Using Gauss's law, we can convert the first term into an integral over the surface:
(5-291)
However since we require that f and g vanish on the surface, this term vanishes. If
we repeat this procedure we find we can rewrite
(f,V·DVg)= Jd r[V·DVJ]*g=(V·DVJ,g).
V
3
(5-292)
Hence we find that
(5-293)
Thus we have again encountered a self-adjoint operator.

From these examples, it is apparent that the operator M = -V · DV 0 + ~a is
O O
also self-adjoint, Mt= M. We will continue to distinguish between the adjoint and
direct operators Mt and M however, since for more general multigroup diffusion
calculations, M will not be self-adjoint (as we will find in Chapter 7). We will
define the adjoint flux cf> t as the corresponding solution of
(5-294)
(Although again we keep in mind that Mt= M and Ft= F implies that cf> t = cf> for
the one-speed diffusion model of a reactor.)
To understand the application of these concepts, let us go back to the criticality
equation (5-283). Now suppose we were to perturb the macroscopic absorption
.
cross sectiqn, say by adding a localized absorber, to a new value
,
(5-295)
We will assume that this perturbation 8~/r) is small and attempt to calculate the
corresponding change in k as governed by the perturbed criticality problem
M'cf>'= ~,Fcf>'. (5-296)
Note here that the perturbation in the core absorption appears as a perturbation
8M in the diffusion operator
M'=M+8M, (5-297)
To calculate the change in k, first take the scalar product of Eq. (5-296) with the
adjoint flux cf> t characterizing the unperturbed core, that is, satisfying Eq. (5-294),
(cpt,Mcp')+(cpt,8Mcf>')= t (cf>t,Fcf>'). (5-298)
Now using the definition Eq. (5-285) of the adjoint operator, we find
(5-299)
Hence we find
_l _ _!_) = (cf> t, 8Mcp')

( k' k (5-300)
(cf> t,Fcp') .
We could now calculate 8k=k'-k. However it is far-more convenient to define

the core reactivity
k-1 (5-301)
p=-k-,
which essentially measures the deviation of the core multiplication from unity.
Then since the perturbation in reactivity is just
llp = p' _ P = k';, 1 _ k ~I =( i _~' ), (5-302)

we can use Eq. (5-300) to find the change in reactivity due to the addition of an
absorption cross section 82.a<r)
(c/> t, 8 2-ac/>')
Ap=-----. (5-303)
(c/> t, Fe/>')
As it stands, this expression is still quite exact, but also still quite formal since it
involves the perturbed flux, cf>', which we usually don't know (and usually don't
want to calculate).
However this is where the idea of "perturbation theory" comes in. For if the
perturbation 82.a is small, then presumably the corresponding perturbation in the
flux <></>=cf>' - cf> is similarly small. Hence we can write
(c/> t, 8 2.ac/>) (c/> t, 8 2.a<>c/>) (c/> t, 8 2.ac/>)(c/> t, F8cf>)

Ap= - - - - - - - - - - + - - - - - - - + ... (5-304)
(c/>t,Fc/>) (cf>t,Fc/>) (c/>t,Fc/>{
Then neglecting second and higher order quantities in the perturbation-that is,
using first order perturbation theory-we find
(c/> t, 8 2.ac/>)
ti.p~----. (5-305)
(c/>t,Fc/>)
Since the one-speed diffusion operator is self-adjoint, we know cf> t = cf>, and hence
we find
fv d rcf>(r)82.a(r)cf>(r)
3
ti.p~ - -------- (5-306)

fv d rcf>(r)v2.r(r)cf>(r)
3
It should be noted that all of this analysis was exact until we neglected second-
order terms in Eq. (5-304). Thus, we have calculated a first-order estimate of the
reactivity change Ap due to introducing a localized absorber 82.a<r) in terms of the
unperturbed flux distribution.
EXAMPLE: Consider a bare slab reactor characterized by one-group constants

D, 2.a, and v2.r, We will perturb this reactor by imagining that an additional
absorber is uniformly inserted in the region O< x < h. One might consider this to be
a model of a bank of control rods inserted to a depth h in the core. Of course to
allow the application of perturbation theory, we must assume this absorption to be
relatively small.
Hence our perturbation is
0<x<h
(5-307)
h<x<a.
If we note that the unperturbed flux in this reactor is

. 'TTX
cf, ( x ) = cf,0 sm
- , (5-308)
a
then we compute the reactivity change due to an insertion of the absorber to a

depth has:
,/,t
i1p(h) = - O a
l8 }:.acf,2 dx
= - 2 8}:.a [ 'TTh - _! sin 2'1Th ] (5-309)
P }:.r lo cf,2 dx 'TT P }:.r 2a 4 a
It is customary to refer to the reactivity change due to such an absorber as the

"worth" of the absorber. (This concept will be defined more precisely in Chapter
14.) Hence the reactivity worth can be sketched for various insertion depths h as
shown in Figure 5-23.
It is also of interest to compute the "differential worth" defined as
(5-310)
Note that the differential worth is at a maximum when the edge of the absorbing
region (e.g., the tip of the control rods) is in the region of largest flux in the center
of the core (see Figure 5-23). Such an analysis, while certainly of interest in
illustrating general trends, is of limited usefulness in detailed control studies
because of the highly absorbing nature of most control elements. Such elements
very strongly perturb the flux in their vicinity, hence invalidating the use of
perturbation theory. We will consider alternative methods required for computing
control rod worth in Chapter 14.
One can obtain more general expressions for the reactivity change induced by
perturbations in the core parameters. For example, if we were to simultaneously
perturb
D'=D+8D, (5-311)
the corresponding reactivity change then would be
fvd3r[ ( v8 }:.r - 8 }:.a)cf,2 - 8D Iv' ct,12]

i1p=------------- (5-312)
fv d3rp}:.~2
The adjoint flux cf, \r) has a rather interesting physical interpretation. Suppose we
imagine an absorber inserted into the reactor core at a point r0 such that
(5-313)
Here a is the effective strength of the absorber. (If we were to imagine that the
8-function was, in fact, a mathematical idealization of an absorber of volume
VA• then a= }:.:VA·) Now strictly speaking, perturbation theory should not be valid
for such a singular perturbation, but we will dismiss such concerns with a wave of
THE ONE-SPEED DIFFUSION TIIEORY MODEL / 225
x=O h a
a FIGURE S-23. Relative and differential control

2 rod bank worth
the hand and use our earlier result to find the corresponding reactivity change as
fvd 3rcp \r)8 ~a(r)cf>(r)

l1p= - - - - - - - - -
fv d 3rcp t(r)v~r(r)cf>(r)
= - ~ cf> t(ro)cf>(ro), (5-314)
where we have denoted the denominator by a constant CS (since it is independent

of the perturbation). If we recognize that cxcp(r0) is just the absorption rate at r0 , we
then find
(5-315)
is simply proportional to the change in reactivity per neutron absorbed at r0 per

second. In this sense, then, the adjoint flux cf, t(r) is a measure of how effective an
absorber inserted at a position r is in changing the reactivity of the core. Evidently
if cf, t(r) is large at r, the core multiplication will be quite sensitive to the absorption
of neutrons at that point. Hence cf> t(r) is sometimes referred to as the neutron
importance or the importance function.
We can see this from a somewhat different perspective if we consider the flux
induced in a subcritical reactor by an arbitrary source S (r) as governed by
(5-316)
Consider the adjoint problem
(5-317)
(Of course for one-speed diffusion theory, M = Mt is self-adjoint, but we will retain
the generality for a bit.) Notice that we have allowed the source st(r) appearing in
the adjoint equation to differ from that in Eq. (5-316).
Now suppose we multiply Eq. (5-316) by </> t and integrate over r, then multiply
Eq. (5-317) by </> and integrate, and then subtract these two results to find
(</> t, (M - F)</>)-((Mt - Ft)</> t,<1>) = (</> t,s )- (st,</>). (5-318)
However by the definition of the adjoint Mt - Ft, the LHS is zero. Hence we find:
(5-319)
Since this must hold for any choice of S (r) and S\r), we will use it to our
advantage by specifying S (r) as a unit point source at r0 :
(5-320)
and st(r) as the cross section }:ir) characterizing an imagined detector placed in
the core. Then we find
(5-321)
Hence in this instance the adjoint flux is simply the response of a detector in the
core to a unit point source inserted at a position r0 . Once again we find that</> \r0)
is a measure of the importance of a neutron event (in this case, the production,
rather than the absorption, of a neutron) at a point r0 in contributing to the
response of a detector with cross section }:ir) (as opposed to reactivity).
For the simple one-speed diffusion model we have been studying, the adjoint
flux </> t(r) is identical to the flux itself. Hence perturbations affecting the creation or
destruction of neutrons will have the most pronounced affect in those regions in
which the flux is largest. This is not true, however, for the more general multigroup
diffusion model, as we will see in Chapter 7.
REFERENCES
1. J. W. Dettman, Mathematical Methods in Physics and Engineering, McGraw-Hill, New

York (1969).
2. B. Friedman, Principles and Techniques of Applied Mathematics, Wiley, New York (1956).
3. G. Arfken, Mathematical Methods for Physicists, 2nd Edition, Academic, New York
(1970).
4. P. M. Morse and H. Feshbach, Methods of Theoretical Physics, Vol. I, McGraw-Hill,
New York (1953), p. 173.
5. Ibid., pp. 115-117.
6. B. Davison, Neutron Transport Theory, Oxford U. P. (1958), pp. 51-55.
7. K. M. Case and P. F. Zweifel, Linear Transport Theory, Addison-Wesley, Reading,

Mass. (1967).
8. G. Arfken, Mathematical Methods for Physicists, 2nd Edition, Academic, New York
(1970), pp. 733-739.
9. Ibid., pp. 748-768.
10. E. A. Coddington, Ordinary Differential Equations, Prentice-Hall, Englewood Cliffs, N.
J. (1961), p. 67.
York (1969).
12. M. Clark, Jr., and K. F. Hansen, Numerical Methods of Reactor Analysis, Academic,
New York (1964).
13. E. Wachspress, Iterative Solution of Elliptic Systems, Prentice-Hall, Englewood Cliffs, N.
J. (1966).
14. R. S. Varga, Matrix Iterative Analysis, Prentice-Hall, Englewood Cliffs, N. J. (1962).
✓ 15. H. Greenspan, et al., Computing Methods in Reactor Physics, Gordon and Breach, New
York (1968).
16. G. E. Forsythe and C. B. Moler, Computer Solution of Linear Algebraic Systems,
Prentice-Hall, Englewood Cliffs, N. J. (1967).
17. T. Craig, SLODOG, A Modified One-group, One-dimensional Diffusion Code, Univer-
sity of Michigan Nuclear Engineering Report (1968).
J. (1966), p. 22.
19. Ibid., p. 24, 25.
20. R. G. Steinke, A Review of Direct and Iterative Strategies for Solving Multi-dimensional
Finite Difference Problems, University of Michigan Nuclear Engineering Report (1971).
21. W. R. Cadwell, et al., PDQ-An IBM-704 Code to Solve the Two-dimensional Few-
group Neutron Diffusion Equations, WAPD-TM-70 (1957) and related reports, WAPD-
TM-179, WAPD-TM-230, WAPD-TM-364, WAPD-TM-678.
22. D. L. Delp, et al., FLARE, A Three Dimensional Boiling Water Reactor Simulator,
General Electric Company Report, GEAP 4598 (1964).
23. E.G. Adensam, et al., Computer Methods for Utility Reactor Physics Analysis, Reactor
and Fuel Processing Technology, Vol. 12, No. 2 (1969).
J. (1966). p. 83.
25. R. Froehlich, in Mathematical Models and Computational Techniques for Analysis of
Nuclear Systems, USAEC CONF-730414-P2 (1973), p. VII-1.
PROBLEMS
5-1 Compare the derivation of the one-speed neutron diffusion equation with that for the
equation of thermal conduction, taking care to point out the assumptions and
approximations used in each case. Refer to any text on heat transfer such as those
listed at the end of Chapter 12.
5-2 By considering a plane source or absorber of neutrons located at the origin of an
infinite medium, derive. the interface condition Eq. (5-15) on the neutron current
density by modeling the source term in the one-dimensional diffusion equation as
S8 (x) and then integrating this equation over an infinitesimal region about origin.
5-3 Compute the rms distance (<x 2)) 112 a neutron will travel from a plane source to
absorption using one-speed diffusion theory. Compare this result with the rms
distance to absorption in a strongly absorbing medium (in which neutron scattering
can be neglected). In particular, plot the rms distance to absorption in water in which
boron has been dissolved against the boron concentration to determine whether the
diffusion theory result for ((x 2)) 112 ever approaches the result characterizing a purely
absorbing medium. (Use the thermal cross section data in Appendix A.)
5-4 It is possible to derive an expression for the relaxation or diffusion length L from the
one-speed transport equation characterizing a homogeneous medium
(we have assumed isotropic scattering for convenience). Seek a solution of the form
cp(x,µ,)=x(µ,)exp(-x/L) to this equation in order to eliminate the x-dependence.
The resulting homogeneous equation for x( µ,) can be reduced to an algebraic
equation for L by eliminating f ~ ldµ, x( µ, ). By following this procedure, derive a
transcendental equation for the diffusion length L:
5-5 Using the assumption that ~a«~., expand L as a power series in ~a/~v substitute
this expansion into the equation above, and evaluate the coefficients of the expansion
in order to derive Eq. (5-32) and obtain the transport corrections to the diffusion
length L = (D /~a) 112 •
5-6 Determine the neutron flux in a sphere of nonmultiplying material of radius R if an
isotropic point source of strength S 0 neutrons per second is placed at the center of the
sphere. Assume the sphere is surrounded by a vacuum.
5-7 The Milne problem: Imagine a diffusing medium in the half space x > 0 with a source
of infinite magnitude at infinity such that the boundary condition on the flux is that
q,(x)~S0 exp(x/ L) as X ➔ OO. Perform the following calculations:
(a) Using one-speed diffusion theory and the boundary condition of zero reentrant
current, determine the flux in the medium.
(b) Repeat the solution of this problem using the extrapolated boundary concept.
(c) Determine the conditions under which these two boundary conditions might be
expected to yield similar results.
5-8 Consider a slab of nonmultiplying material with a plane source at its origin emitting
S 0 neutrons/cm2 ·sec. By solving this problem first with the condition of zero-
reentrant current and then extrapolated boundaries, compare the absorption rate in
the slab predicted by these two approaches. Also calculate the rate at which neutrons
leak from the slab in each case.
5-9 Consider a thermal neutron incident on a slab-shaped shield of concrete I m in
thickness, and determine the probability that: (a) the neutron will pass through the
shield without a collision, (b) it will ultimately diffuse through the shield, and (c) it
will be reflected back from the shield. (For convenience, treat the concrete as if it had
the composition of 10% H 20, 50% calcium, and 40% silicon.)
5-10 Consider an infinite nonmultiplying medium containing a uniformly distributed
neutron source. If one inserts an infinitesimally thin sheet of absorber at the origin,
determine the neutron flux throughout the medium.
5-11 Derive the expression given by Eq. (5-47) for the albedo characterizing a slab of
material of thickness a. In particular plot this albedo for a slab of water for various
thicknesses. (Use thermal cross section data.) Comment on the behavior of the albedo
as given by Eq. (5-47) for both very thin and very thick slabs.
5-12 One defines the blackness coefficient characterizing a region as
where a denotes the surface of the region. Yet another useful parameter characterizing
interfaces is the ratio of the current density J to the flux at the interface
Determine the relation between these parameters and the albedo, assuming that
diffusion theory can be used to describe the material adjacent to the region of interest.
(It should be remarked that one frequently uses these concepts to characterize very
highly absorbing regions such as fuel elements or control rods in which diffusion
theory will usually not be valid in the highly absorbing region.)
5-13 In reactor analysis it is frequently of interest to determine the neutron flux in a
so-called unit fuel cell of the reactor, that is, a fuel element surrounded by a
moderator. As a model of such a cell characterizing a cylindrical fuel element,
consider a fuel pin of radius a surrounded by a moderator of thickness b. For reasons
that will become more apparent in Chapter 10, one assumes that the fission neutrons
that slow down to thermal energies appear as a source uniformly distributed over the
moderator-but not directly in the fuel. Furthermore it is assumed that there is no net
transfer of neutrons from cell to cell-that is, the neutron current vanishes on the
boundary of the cell (although the neutron flux will not vanish there).
Determine the neutron flux in this cell geometry. In particular, determine the
thermal utilization characterizing the cell by computing the fraction of those neutrons
slowing down that is absorbed in the fuel.
5-14 Consider a one-dimensional slab model of a fuel cell in which the center region
consists of the fuel, and the outer regions consist of a moderating material in which
neutrons slow down to yield an effective uniformly distributed source of thermal
neutrons S0 neutrons/ cm3 • sec. Determine the neutron flux in this cell. In particular,
compute the so-called self-shielding factor fs defined as the ratio between the average
flux in the fuel to the average flux in the cell.
5-15 Consider two isotropic point sources located a distance a apart in an infinite
nonmultiplying medium. Determine the neutron flux and current density at any point
in a plane midway between the two sources.
5-16 Determine the infinite medium Green's functions or diffusion kernels characterizing
cylindrical and spherical geometries.
5-17 By representing a plane source as a superposition of isotropic point sources, construct
the plane source kernel Gpi(x,x') by using the point source kernel GPi(r,r').
5-18 Obtain an expression for the plane source diffusion kernel characterizing a finite slab
of width a by solving for the neutron flux resulting from a unit plane source at a
position x' in the slab. This can be most easily accomplished by seeking a separate
solution on either side of the source plane which satisfies the vacuum boundary
conditions at either end, and then matching these solutions at the source plane using
. the interface condition of continuity of flux and a discontinuity in the current density
given by the source strength.
5-19 Consider a neutron source emitting a monodirectional beam of neutrons into an
infinite medium. Using one-speed diffusion theory, calculate the neutron flux in the
medium. For convenience, locate your coordinate system with its origin at the source
and align the x-axis along the source beam. Since the source is highly anisotropic, you
cannot apply diffusion theory directly. Rather, compute the distribution of first
scattering collisions of the source neutrons along the x-axis, and then assume that
each of these collisions acts in effect as an isotropic point source of neutrons for the
subsequent diffusion theory analysis (assuming that such scattering is isotropic).
5-20 Use the method of variation of constants to determine the flux in a finite slab that
contains a uniformly distributed neutron source.
5-21 Construct the spatial eigenfunctions of the Helmholtz equation in spherical geometry.
5-22 Construct the spatial eigenfunctions of the Helmholtz equation in infinite cylindrical
geometry.
5-23 Construct the spatial eigenfunctions of the Helmholtz equation characterizing a
parallelepiped geometry.
5-24 Demonstrate explicitly that the eigenfunctions for the slab geometry are indeed
orthogonal.
5-25 Consider a slab of nonmultiplying material containing a uniformly distributed
neutron source. Determine the neutron flux in the slab: (a) by directly solving the
diffusion equation and (b) using eigenfunction expansions. Demonstrate that these
solutions are indeed equivalent.
5-26 Determine the neutron flux in a long parallelepiped column of nonmultiplying
material caused by a uniform source distributed across the face of the column. In
particular determine the spatial behavior of the flux far from the source plane.
Describe how one might measure the neutron diffusion length in the column by
studying the spatial behavior of the flux.
5-27 The objective of this problem is to write a computer code to calculate the flux in a
uniform nonmultiplying slab containing an arbitrary source distribution. Perform the
following steps:
(a) Derive the finite difference form of the appropriate one-speed diffusion equation
using vacuum boundary conditions. Use 30 mesh points and equal mesh spacing.
(b) Write the equations in matrix form, A cf>=~-
(c) Derive the steps necessary to solve this equation using the simultaneous relaxation
method. (Gaussian elimination would be better, but less instructive.)
(d) Write the necessary computer program. Input should consist of slab thickness, D,
~a, and S;. Output should include a tabulation of X;, cf>;, and S;. Pay particular
attention to your convergence criterion and initial flux guess cf>/0> for the inner
iterations of the SR method.
(e) "Check out" your code by solving at least two problems with known analytical
solutions (e.g., cosine source or uniform source). Show how the inner iterations of
your program converge by plotting cf>?> for various values of n.
5-28 Repeat the analysis of the time-dependent slab reactor given in Section 5-3, but with
the addition of a plane source of neutrons of strength S 0 located at the origin of the
slab. In particular determine the long time behavior of the reactor when it is critical.
5-29 Determine the geometric buckling B;and critical flux profile in the following bare
reactor geometries: (a) sphere, (b) infinite cylinder, and (c) parallelepiped.
, 5-30 We can define the power-peaking factor for a given reactor core as the ratio between
the maximum power density and the average power density in the core. Recognizing
that the power density is proportional to the neutron flux in the one-speed approxima-
tion, compute the power-peaking factor for three common geometries: (a) sphere, (b)
cube, and (c) finite cylinder.
5-31 A homogeneous one-speed bare reactor has a cylindrical configuration. Determine:
(a) the radius and length of the reactor as functions of the buckling so that the volume
of the critical reactor, and hence its mass, is a minimum and (b) the minimum volume
as a function of buckling.
Jezebel is a bare, fast, spherically shaped critical reactor constructed of pure 239 Pu
metal (density 15.4 g/cm3). Calculate the critical radius and critical mass of the
reactor using the one-group data: P = 2.98, ac = 1.85 b, ay = 0.26 b, and a 1r = 6.8 b.
5-33 There has been considerable interest in the possibility of super heavy nuclei with mass
numbers A > 300. Such nuclei would be characterized by large values of P ( ~6-10).
Using the results of Problem 5-32, study the effect of varying v on the critical radius
and mass of a bare sphere of such material.
5-34 Two infinite slabs, each of thickness a in the x direction, are separated by an inner
region of thickness 2a and are bounded by vacuum on their outer surfaces. The slab
material is of composition to give k 00 = 1.2 and thermal diffusion length of 50 cm.
,..
Determine the thickness a for criticality when: (a) inner region is vacuum and (b)
inner region is a medium with k 00 = 1 and same D and L as the outer slabs.
5-35 A bare spherical reactor is made with 235 U uniformly dispersed in graphite (p= 1.7)
with an atomic ratio Ne/ N 25 = Hf. For the cross section values given below, calculate
the critical size and mass of the reactor according to one-group diffusion theory. If the
reactor is modified by placing a cavity (vacuum) of half the total radius in its center,
find the critical size for this case. Recalculate the critical radius if the center void is
filled instead by a perfect absorber. Use as data: a~=4.3 b, a;=0.003 b, a'; 5 = 105 b,
al5 = 584 b, and D = 0.9 cm.
A bare spherical reactor is to be constructed of a homogeneous mixture of D 20 and
235
U. The composition is such that for every uranium atom there are 2000 heavy water
molecules (i.e., N 020 / N 25 = 2000). Calculate: (a) the critical radius of the reactor
using one-speed diffusion theory (Data: 17 25 =2.06, D 0 20 =0.87 cm,.
};?20 = 3.3 X 10-
•
5
5
cm-1, a;>2°=0.001 b, and a~ =678 b.) and (b) the mean number of scattermg
collisions made by a neutron during its lifetime in this reactor.
5-37 There is strong motivation to obtain as flat a power distribution as possible in a
reactor core. One manner in which this may be accomplished is to load a reactor with
a nonuniform fuel enrichment. To model such a scheme, consider a bare, critical slab
reactor as described by one-speed diffusion theory. Determine the fuel distribution
Np(x) which will yield a flat power distribution P(x)=wr};r(x)<[>(x)=constant. For
convenience, assume that fuel only absorbs neutrons and that it does not significantly
scatter them. Also assume that all other materials in the core are uniformly distri-
buted.
5-38 A one-dimensional slab reactor system consists of three regions: vacuum for x < O; a
multiplying core for O< x < a; and an infinite nonmultiplying reflector for x > a.
Calculate the core thickness a that will yield a critical system.
Suppose a cannot be made large enough to achieve criticality. Then determine the
flux at all points when an external source S0 is uniformly distributed throughout the
reflector region x > a.
5-39 Consider a bare slab reactor with material composition such that };a= 0.066 cm - 1,
D = 0.90, and v};r = 0.070 cm - 1• Modify the one-speed diffusion computer code
developed in Problem 5-27 so that you can calculate the width a that will yield
criticality.
One possible procedure is to guess an initial slab width a and then perform a source
iteration calculation to determine k,11. To simplify the calculation, choose S <0l(x) = 1.0
= constant. The integrated fission source that appears in the estimate of k<n) given by
Eq. (5-274) can be performed using simple trapezoidal quadrature
where dx is the mesh spacing. Again use 30 mesh points and require a convergence
criterion on k(n) of
After each criticality calculation, readjust the slab width a and recalculate k,11. After
several such calculations, plot k,11 against a to determine the critical slab width a*.
Compare this with the analytical expression for a*.
5-40 Investigate the convergence of the inner and outer iterations in the one-speed
diffusion code developed in Problem 5-39 for the following modifications:
(a) Use the outer iteration source S (nl(x) as the initial estimate in the inner iterations.
(b) Determine the sensitivity of both inner and outer iterations to the convergence
criteria.
(c) Attempt to accelerate the outer iterations by source extrapolation.
5-41 Verify the general first-order perturbation theory result Eq. (5-312).
5-42 Why is the adjoint system introduced in developing the perturbation equations?
Illustrate your answer with an example showing that only the use of the adjoint
system will yield the desired result.
5-43 Describe a reasonable experimental procedure by which one could measure the
variation of neutron importance within a reactor core.
5-44 Calculate the error in the critical mass of a bare homogeneous spherical reactor due to
a 1% error ink. Assume that only the core size would be adjusted to give criticality.
5-45 Calculate the relative worth of a control rod bank inserted axially into a cylindrical
reactor core.
5-46 Consider a critical bare slab reactor of thickness a which is composed of a homo-
geneous mixture of fuel and moderator. Estimate the reactivity change if a thickness
8x of the fuel-moderator mixture at a position x is replaced by pure moderator. What
8x at a distance from the centerline of x = 0.4 a is required to give the same reactivity
change as a perturbation thickness 8x 0 at the center of the slab?
5-47 Variational methods can be used in a manner very similar to perturbation theory to
estimate the multiplication of a given core configuration using only crude guesses of
the flux in the core. For example, a useful variational expression for the multiplication
of a core described by a one-speed diffusion theory is
f d rcp(r)Mcf,(r)
3
k-1=------
f d rcp(r)Fcf,(r)
3
Compare the accuracy of such a scheme for a slab of width a where the estimates of
the flux or "trial functions" cf,(x) are taken as simple quadratic polynomials that
vanish on the boundaries of the slab [i.e., cf,( x) = 1 - (2x / a)2].
6
Nuclear Reactor Kinetics
For a nuclear reactor to operate at a constant power level, the rate or neutron
production via fission reactions should be exactly balanced by neutron loss via
absorption and leakage. Any deviation from this balance condition will result in a
time-dependence of the neutron population and hence the power level of the'
reactor. This may occur for a number of reasons. For example, the reactor operator
might desire to change the reactor power level by temporarily altering core
multiplication via control rod adjustment. Or there may be longer term changes in
core multiplication due to fuel depletion and isotopic buildup. More dramatic
changes in multiplication might be caused by unforeseen accident situations, such
as the failure of a primary coolant pump or a blocked coolant flow channel or the
accidental ejection of a control rod.
lt is important that one be able to predict the time behavior of the neutron
population in a reactor core induced by changes in reactor multiplication. Such a
topic is known as nuclear reactor kinetics. However, we should recognize that the
core multiplication is never completely under the control of the reactor operator.
Indeed since multiplication will depend on the core composition, it will also
depend on other variables not directly accessible to control such as the fuel
temperature or coolant density distribution throughout the reactor, but these
variables depend, in turn, on the reactor power level and hence the neutron flux
itself. The study of the time-dependence of the related processes involved in
determining the core multiplication as a function of the power level of the reactor
. is known as nuclear reactor dynamics and usually involves a detailed modeling of
the entire nuclear steam supply system. Although we briefly discuss several of the
more important "feedback" mechanisms involved in determining core multiplica-
tion later in this chapter, our dominant concern is with predicting the time
behavior of the neutron flux in the reactor for a given change in multiplication.
233
234 / fflE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
The principal applications of such an analysis are not only to the study of
operating transients in reactors, but also to the prediction of the consequences of
accidents involving changes in core multiplication, and to the interpretation of
experimental techniques measuring reactor parameters by inducing time-dependent
changes in the neutron flux. One can roughly distinguish between two different
types of analysis depending on the time scale characterizing changes in the neutron
population or core properties. For example, one is interested in relatively short-
term changes possibly ranging from fractions of a second up to minutes in length
when analyzing normal changes in reactor power level (e.g., startup or shutdown)
or in an accident analysis. By way of contrast, changes in core composition due to
fuel burnup or isotope buildup usually occur over periods of days or months.
Needless to say, the analysis required for each class of time behavior is quite
different.
The reader will recall that we have already considered a particularly simple
example of nuclear reactor kinetics when we discussed the time behavior of the
neutron flux in a slab reactor model in Chapter 5. Although this earlier analysis
was useful for deriving the condition for reactor criticality, it is not valid for an
accurate description of nuclear reactor kinetics, since it assumed that all fission
neutrons appeared promptly at the instant of fission. As we demonstrate in the next
section, it is essential that one explicitly account for the time delay associated with
delayed neutrons in describing nuclear reactor kinetics.
Fortunately the one-speed diffusion model we have been using to study reactor
criticality is also capable of describing qualitatively the time behavior of a nuclear
reactor, provided we include the effects of delayed neutrons. Indeed such a model
is frequently too detailed for practical implementation in reactor kinetics analysis
due to excessive computation requirements, particularly when the effects of
phenomena such as temperature feedback are included. For that reason, we will
begin our study of nuclear reactor kinetics by reducing the one-speed diffusion
model still further with the assumption that the spatial dependence of the flux in
the reactor can be described by a single spatial mode shape (the fundamental
mode). Under this assumption, we can remove the spatial dependence of the
diffusion model and arrive at a description involving only ordinary differential
equations in time. This model is sometimes known as the point reactor kinetics
model, although this is somewhat of a misnomer since the model does not really
treat the reactor as a point but rather merely assumes that the spatial flux shape
does not change with time. Although we will rely heavily on the point reactor
kinetics model in our study of nuclear reactor time behavior, we will indicate its
generalizations to include a nontrivial spatial dependence as well as feedback
mechanisms.
There are two other aspects of nuclear reactor kinetics that we will not touch
upon in this chapter. The first topic involves the study of nuclear reactor control
and includes not only the analysis of the various schemes used to adjust core
multiplication but those mechanisms by which changes in the core power level can
affect multiplication as well. The second topic involves the study of long-term
changes in the core power distribution due to fuel depletion and the buildup of
highly absorbing fission products in the reactor core.
However the study of both of these subjects involves only a steady-state analysis
of the neutron flux in the reactor-or, at most, a sequence of steady-state criticality
calculations. Only the time-dependence of the slowly varying changes in core
composition such as those due to fuel depletion must be explicitly considered.
Therefore such topics can be more appropriately developed in later chapters.
NUCLEAR REACTOR KINETICS / 235
I. THE POINT REACTOR KINETICS MODEL
A. The Importance of Delayed Neutrons in Reactor Kinetics

In our earlier treatment of the simple bare slab reactor (Section 5-111-C), we
found that the neutron flux in such a system could be written as a superposition of
spatial modes (or eigenfunctions) characteristic of the reactor geometry, each
weighted with an exponentially varying time-dependence:
</>(r, t) = ~ An exp( -Ant) lfn(r). (6-1)

n
Here the spatial eigenfunctions were determined as the solution to the eigenvalue
problem [Eq. (5-232)]:
(6-2)
while the time eigenvalues An were given by
(6-3)
These eigenvalues are ordered as - A1 > - A2 > · · · . Hence for long times the flux
approaches an asymptotic form
where we identify
I=[ v};a( 1 + L2B}) ]-

v}:t/}:a k
1
=mean lifetime of neutron in reactor, -·---·-------
/'} )
(6-5) '
I
k = - - 2- -
l+L Bj
=
l+L Bg2
"'i =multiplication factor.
It would be natural to inquire as to just how long one would have to wait until such
asymptotic behavior sets in. We can determine this rather easily by assuming that •
the reactor is operating in a critical state such that A1 = 0, and then estimating An. If
we recall that for a slab B;= n 2 (1r /a)2, then
Now in a typical thermal reactor, a~300 cm, v~3X 105 cm/sec, and D~l cm.
Hence the higher order An are of the order of 100---1000 sec - 1, which implies that
the higher order spatial modes die out very rapidly indeed.
We can utilize this fact to bypass our earlier separation of variables solution of
the one-speed diffusion equation [Eq. (5-187)] by merely assuming a space-time
separable flux of the form
cp(r, t) = vn ( t )iJ;i(r), (6-7)
where i/; 1(r) is the fundamental mode or eigenfunction of the Helmholtz equation,
Eq. (6-2). If we substitute this form into the one-speed diffusion equation, we find
that n ( t) satisfies
!; =( k~ 1 )n(t). (6-8)
In this sense, n(t) can be interpreted as the total number of neutrons in the reactor
at time t [if we normalize Jd3rifii(r)= l]. Actually since the normalization of n(t) is
arbitrary, we could also scale the dependent variable n(t) to represent the total
instantaneous power P ( t) being generated in the reactor core at any particular
time. That is, we could let
(6-9)
where wr is the usable energy released per fission event. Since the reactor power
level is usually a more convenient variable to monitor, we will frequently express
the reactor time-dependence in terms of P ( t).
Equation (6-8) represents a somewhat simplified form of a more general set of
equations, which we shall derive in a moment, that are commonly referred to as the
point reactor kinetics equations. This term arises since we have separated out the
spatial dependence by assuming a time-independent spatial flux shape 1/ii(r). In this
sense we have derived a "lumped parameter" description of the reactor in which
the neutron flux time behavior is of the form of the product of a shape factor i/; 1(r)
and a time-dependent amplitude factor n(t),
(6-10)
characterized by a time constant
T_ k ~ 1 = reactor period. (6-11)
This model is of course identical to that developed in our qualitative discussion of

chain-reaction kinetics in Section 3-1-B except that one-speed diffusion theory has
now given us an explicit expression for the neutron lifetime I [Eq. (6-5)].
However there is, of course, something very important missing from this model.
For by assuming a fission term in Eq. (5-186) of the form P'1:r<f,(r,t), we have
implied that all fission neutrons appear promptly at the time of fission. But we
know that a very small fraction ( /3~0.7%) of such neutrons are emitted with
appreciable time delay (recall the discussion of Section 2-11). Although these
delayed neutrons are only of minor significance in steady-state critical reactors,
they are extremely important for reactor time behavior. For if we recall that the
prompt neutron lifetime / is typically of the order 10- 4 sec in a thermal reactor
(10- 7 sec in a fast reactor), then it is apparent that the reactor period predicted by
this model would be far too small for effective reactor control.
We can give a crude estimate of the influence of delayed neutrons on the reactor
time behavior by noting that the effective lifetime of such neutrons is given by their
prompt neutron lifetime I plus the additional delay time \- 1 characterizing the
/3-decay of their precursor. If we weight both prompt and delayed neutrons by their
respective yield fractions, we can estimate an average neutron lifetime <t)
characterizing all fission neutrons as
<l)=(l-/3)/+ ~/3;[~; +1]. (6-12)
Using the delayed neutron data given in Table 2-3, we find that this average
lifetime is typically <l)~0.l sec, which is considerably longer than the prompt
neutron lifetime l~ 10- 6 - 10- 4 sec. Hence delayed neutrons substantially increase
the time constant of a reactor so that effective control is possible.
This fact suggests a related idea; suppose we consider a reactor that is very
slightly subcritical when only prompt neutrons are considered. Suppose further that
the fraction /3 of delayed neutrons provides just enough extra multiplication to
achieve criticality. This fraction will, in fact, control the criticality-and hence the
time constant. However if k- 1 ~ /3, the reactor will be critical (or supercritical) on
prompt neutrons alone, and the reactor period should become very short, since the
delayed neutrons are not needed to sustain the chain reaction. Obviously we should
design a reactor such that this situation will never occur.
B. Derivation of the Point Reactor Kinetics Equations
In actual fact, one cannot proceed so heuristically. We must first set up a set
of equations describing the time dependence of the delayed neutrons. To this end,
we must define the precursor atomic number density:
C; (r, t)d 3r= expected number of "fictitious" precursors of
ith kind in d 3r about r that always decay by (6-13)
emitting a delayed neutron.
Note that C; (r, t) is only some fraction of the true precursor isotope concentration,
since only a fraction of the ith isotope nuclei eventually decays by delayed neutron
emission. For example, the 87 Br precursor described in Section 2-11 is characterized
by a fictitious precursor concentration
C8 r(r,t)=(.029) (.7) Br(r,t). (6-14)
One can immediately write down a balance relation for these precursor concentra-
tions by referring to our earlier discussion of radioactive decay to identify
Number of precursors]
decaying in =\C;(r,t)d 3r, (6-15)
[ 3
l
d r /sec
Number of precursors
being produced = /3;v~rc/>(r, t)d 3r. (6-16)
[
in d 3r /sec
[This latter relation assumes the precursors don't migrate or diffuse before decay-
ing.] Hence our precursor concentration balance equation is just
(6-17)
Now our old friend, the one-speed diffusion equation, can still be used to describe
the flux-provided we treat the delayed neutron contribution to the fission source
explicitly by writing it as
6
Sr{r,t)=(l-,B)r~rc/>(r,t)+ L \C;(r,t). (6-18)
i= I
Hence our system of equations describing the neutron flux in a reactor including
delayed neutrons is
i= 1, ... ,6. (6-19)
We will apply these once again to the asymptotic situation in which both the flux
and precursor concentrations can be written as separable functions of space and
time
cp(r, t) = vn ( t )1h (r ),
(6-20)
C;(r,t)= C;(t)1Jii(r),
where t/; 1(r) is the fundamental mode of Eq. (6-2).If we substitute these forms into
Eq. (6-19) and use our expressions for the prompt neutron lifetime I and multiplica-
tion factor k for a bare, uniform reactor as given by Eq. (6-5), we arrive at a set of
ordinary differential equations for n(t) and C;(t):
(6-21)
i= 1, ... ,6.
These are known as the point reactor kinetics equations and represent a generaliza-
tion of Eq. (6-8) to include the effects of delayed neutrons. One frequently rewrites
these equations in a somewhat different form by introducing two definitions. First
we must define the mean neutron generation time:
A= J_ = mean generation time between birth of neutron

(6-22)
k and subsequent absorption inducing fission.
If k~ I, then A is essentially just the prompt neutron lifetime /. Next we define a

very important quantity known as the reactivity, which essentially measures the
deviation of core multiplication from its critical value k = I,
k(t)-1
p(t)- k(t) =reactivity. (6-23)
Notice here that we have explicitly indicated that k and hence p may be functions
of time. We will comment more on this in a moment.
These definitions allow us to rewrite Eqs. (6-21) in perhaps their most conven-
tional form
dn [ p(t)-/3]
dt = A n(t)+ ;;1
6
A.;C;(t),
(6-24)
dC; /3;
dt = An(t)-;\;C;(t), i= 1, ... ,6.
Hence we now have a set of seven coupled ordinary differential equations in

time that describe both the time-dependence of the neutron population in the
reactor and the decay of the delayed neutron precursors. Unfortunately the
solution of this system of equations is not as straightforward as it might first appear
for several reasons. First the reactivity p(t) is usually a function of time and in fact
frequently depends on the neutron population n(t) itself. Hence the equations will
generally be nonlinear. Furthermore the time constants characterizing the nuclear
processes represented by the equations range all the way from A~l0- 6-10- 4 sec to
the lifetime of the longest lived precursor, usually about 80 sec. These widely
different time scales complicate even a direct numerical solution of Eqs. (6-24)
since the time step size allowed in most standard numerical schemes (e.g., Runge-
Kutta or predictor-corrector) is primarily controlled by the smallest time constant
-in our case, the prompt neutron lifetime. Since reactor dynamics usually occur
on a time scale characterized by the delayed neutrons, such direct approaches tend
to be quite inefficient and more sophisticated methods are frequently required.
C. Limitations of the Point Reactor Kinetics Model

A number of questionable assumptions have entered into the derivation of
these equations, such as the one-speed diffusion approximation, and a time-
independent spatial shape. Fortunately the point reactor kinetics equations can be
derived in a much more general fashion in which such assumptions are not
necessary. 1 Such derivations usually proceed from the transport equation itself and
are also usually very formal. They lead, however, to the set [Eq. (6-24)] in which
only the definitions of /3, A, and p are changed. Hence provided one uses the more
general expressions for these parameters, the point reactor kinetics equations can
be regarded as having a much broader domain of validity.
One major modification that must be introduced into the equations is to take
some account of energy-dependent effects. These arise primarily because the
delayed neutrons appear with somewhat lower energies than do the prompt fission
neutrons (recall Figures 2-21 and 2-23). Hence in a thermal reactor, they do not
have to slow down quite so far and therefore are characterized by a somewhat
higher probability of inducing thermal fission (by as much as 20%). Of course, this
may work in just the opposite direction in fast reactors, since delayed neutrons
usually appear with energies below the fast fission threshold. We can take account
of these effects by characterizing each delayed neutron group by a slightly different
fast nonleakage probability and a resonance escape probability, say P~NL and p ;_
Then we would merely modify the definition of the parameters appearing in the
point reactor kinetics equations as
(6-25)
(6-26)
(6-27)
6
/3-"JJ= L A-
i= I
(6-28)
In large thermal power reactors, A is typically several percent larger than the
physical value /3; (although in small research reactors, fi; may be as much as
20-30% larger than /3;). It should also be kept in mind that the delayed neutron
yield fractions A must be calculated as suitable weighted averages over the relevant
mixture of fuel isotopes. In fact one will find that in most thermal spectrum
reactors these averaged delayed neutron fractions will decrease with core life since
the most common bred materials, 239 Pu and 233 U, have somewhat lower delayed
neutron yields than 235 U.
In 239 Pu/ 238 U-fueled fast reactors, the situation becomes much more compli-
cated since there are as many as six fissioning isotopes, each characterized by six
different precursor groups. To handle 36 delayed neutron precursor concentration
equations would be very complicated indeed. Fortunately the recent trend toward
direct experimental measurement of isotopic yields as opposed to precursor group
data should allow one to eventually solve directly for the isotopic concentrations of
the major precursor isotopes (e.g., 87 Br, 88 Br, and 1371) and only lump the minor
isotopes into effective precursor groups. This would then eliminate the problem of
isotopic averaging for several fissionable isotopes, and decouple the data used in
the kinetics calculation more effectively from the particular reactor core composi-
tion.
Perhaps the most serious approximation involved in the point reactor kinetics
equations involves the assum1)tion that the flux can be adequately represented by a
single, time-independent spatial mode 1h(r). It should first be noted that this shape
function is actually not the fundamental mode characterizing a critical system, but
rather the fundamental mode characterizing the reactor core that has been sub-
jected to a reactivity change away from critical. Nevertheless it is common to
utilize a shape function 1h(r) characterizing a critical core configuration if the
reactor is close to a critical state or on a truly asymptotic period.
When the changes in core composition are sufficiently slow, as in fuel depletion
or fission-product poisoning studies, one can perform an instantaneous steady-state
criticality calculation of the shape function 1h(r), even though this shape will slowly
change with time. Such a scheme is known as the adiabatic approximation. 2
More elaborate procedures exist for including a time variation in the shape
function. 3 However for rapidly varying transients in which spatial effects are
important, one is usually forced to solve the time-dependent neutron diffusion

equation directly (at considerable expense). We will return to discuss such spatially
dependent kinetics problems later in this chapter.
Of course one could proceed in a much more empirical fashion by noting that
the point reactor kinetics equations hold for more general situations than we have
considered, but then simply to postulate that the correct values for the parameters
/3, A, and p are available (perhaps from experimental measurement). In this sense,
all detailed considerations of the flux spatial shape are avoided (or, rather, swept
under the carpet).
Let us now briefly examine the reactivity p(t) = [k ( t)- 1]/ k(t) appearing in the
point reactor kinetics equations. We know that the multiplication factor k, and
hence p, depends on the size and composition of the reactor. In our specific
one-speed diffusion model of a bare reactor core, we have found an explicit form
for k [Eq. (6-5)]. Hence by changing the size or composition-say by inserting or
withdrawing a rod of absorbing material or adjusting a poison concentration-we
can change p and hence control the reactor. In this sense, p will in general be a
function of time partly under the control of the reactor operator.
However for any reactor operating at power, p will also depend on the flux itself
due to several factors. First, the power level will influence the temperature of the
components of the reactor core. However the atomic concentrations of materials in
the core depend sensitively on their temperature. As the temperature changes, they
may contract or expand or change phase. This in turn will cause a change in the
macroscopic cross sections-and hence in the reactivity. Furthermore temperature
changes may directly affect the microscopic cross sections (e.g., via the Doppler
effect). Finally, the atomic concentration of materials in the core will vary as
fission products are produced or fuel nuclei are fissioned and depleted. This will
also strongly influence the reactivity.
Such processes whereby the reactor operating conditions will affect the criticality
of the core are known as feedback effects and play an extremely important role in
reactor operation. Stated mathematically, such feedback effects imply that the
reactivity must be regarded as a nonlinear function of the power level p[n(t), t].
Hence the point reactor kinetics equations are actually a coupled set of nonlinear
ordinary differential equations that are extremely difficult to analyze analytically,
with the exception of certain very simplified model cases.
Although we will eventually discuss the physics of several of the more prominent
feedback mechanisms, we will initially limit our study to those cases in which p(t)
is a specified function of time (and hence the point reactor kinetics equations are
linear). Such a situation is commonly referred to as the zero-power point reactor
model, since it ignores the feedback that would occur due to variations in the
reactor power level. Although limited in this sense, this model does reveal quite a
bit about the time behavior of the neutron population in a reactor.
II. SOLUTION OF THE POINT REACTOR KINETICS

EQUATIONS
A. Solution With One Effective Delayed Group

We will first apply the point reactor kinetics equations to the situation in
which the reactivity is a prescribed function of time. In fact we will begin with the
r
I
very simple situation in which we imagine a reactor operating at some given power
level P O prior to a time, say t = 0, at which point the reactivity is changed to a
nonzero value p0 • Rather than attempting to solve the full set of point reactor
kinetics equations for this situation, we will first consider the case in which all
delayed neutrons are represented by one effective delayed group, characterized by
a yield fraction
(6-29)
and an averaged decay constant
(6-30)
The point reactor kinetics equations for this simplified case become
: = [ Po~ ,8 ]P(t)+>.C(t),
(6-31)
~; = f P(t)->.C(t), n,o.
Note that we have chosen to work with the instantaneous reactor power level P(t)
as our dependent variable. The precursor concentration is then slightly modified to
C = cnew = wf~fcold' Now prior to t = 0, the reactor is operating at a steady-state
power level P 0 • Hence we find that for t < 0 we must require
(6-32)
This relationship yields the appropriate initial conditions that accompany Eq.
(6-31):
(6-33)
This simple initial value problem can be solved in a variety of ways (the easiest
being by Laplace transform~ as discussed in Appendix G). We will use a more
pedestrian approach by merely seeking exponential solutions of the form
(6-34)
where P, C, and s are to be determined. Then if we substitute these forms into Eq.
(6-31), we find the algebraic equations
sP= (
p0 -/3) P+>.C,
A
(6-35)
/3
sC= A P->.C.
This set of homogeneous equations has a solution if and only if
or
As 2 + (AA+ {3-p 0 )s- PoA=0. (6-36)
Hence we have arrived at a characteristic equation for the parameter s. In

particular we find that there are two possible values for s
(6-37)
and hence our general solutions will be of the form
P ( t) = P 1 exps 1t + P 2 exps 2 t,
(6-38)
C(t)= C 1 exps 1t+ C2 exps 2 t.
To determine the· unknown coefficients we can apply both the initial conditions
and the equations (6-35) to obtain four algebraic equations for the four unknowns
P 1, P 2, C 1, and C2 • Since these results are still rather cumbersome, we will examine
the solution in the case in which (/3-p 0 +AA)2»4AAp 0 (or AA//3«1) and IPol«/3.
Then the two roots are approximately given by
I
II (6-39)
f
while the power P(t) becomes
P ( t) =::::. P O[ ( /3 _/3 Po ) exp ( /3Apo

_ Po ) t - (
/3 Po E.
_ Po ) exp - / /3 -A Po ) t ] . (6-40)
We have sketched this solution in Figure 6-1 for both positive and negative
reactivity insertion of an amount IPol = 0.0025 into a reactor characterized by
{3=0.0075, A=0.08 sec- 1, and A= 10- 3 sec. In particular, it should be noted that
after a very rapid initial transient (s 2- 1~.2 sec), the reactor time response becomes
exponential with a period of T=s,- 1~25 sec. This time constant characterizing the
more slowly varying asymptotic behavior is occasionally referred to as the stable
reactor period.
B. The Inhour Equation

The reciprocal time constants s 1 and s 2 characterizing the single delayed
group model were given as the roots of a quadratic equation (6-36) in terms of p 0 ,
/3, A, and A. We can generalize this result to the situation in which there are several
groups of delayed neutrons by first rewriting Eq. (6-36) in a slightly different form.
Po> O
1.5
1.0
P(t)
Pa Po<. 0
0.5
0.1 0.3 0.5 0. 7 0.9 1.1

t(sec)
FIGURE 6-1. Reactor power level behavior following a positive step reactivity insertion
Using our definitions of p and A in Eqs. (6-22) and (6-23), we can rewrite Eq.
(6-36) as
Po
=_!]__+_I(£)
sl+ I sl+ I s+;.\ .
( 6-41)
~- - \ ~-· I
:/ti' _,
This equation will determine the decay constants s for any constant reactivity p0•
To generalize to six delayed groups, we write
s/ 1 ~ s/3;
Po= sl+ I + sf+ I ,L. s+;.\ =p(s). (6-42)
i= I I
This equation, which expresses the decay constants sJ as the roots of a seventh-
order polynomial, is known in reactor theory as the inhour equation. [This
terminology arises from the very early attempts to express reactivity in units of
"inverse hours" or inhours, which were defined as the amount of reactivity required
to make the reactor period equal to one hour. Reactivity is more commonly
expressed today either in decimals or percentages or pcm (per cent mille = 10- 5) of
!::.k/ k. We will later introduce another unit called the dollar, in which pis measured
in units of the delayed neutron fraction /3.]
The roots of Eq. (6-42) are most conveniently studied using graphical techniques.
In Figure 6-2 we have plotted the right-hand side of Eq. (6-42) for various values of
s. The intersection of these curves with the line p = p0 yields the seven decay
constants sJ characterizing the time-behavior
7
P(t)= ~ PJexpsi. (6-43)
)=I
The root lying farthest to the right, s0 , is identified as the reciprocal reactor period,
s 1 = r- 1 • It is apparent from Figure 6-2 that only this root will ever be positive.
The remaining roots si' j > I, can then be identified as transients that die out
rapidly after a reactivity p0 is inserted into the reactor.
p(s)
)_
_·__
1 _ : \ : I !
I- -L - I I I 1.0
I I I I I
-----i---_J___ I·-+- -----

I I I I I I
11 I
II
FIGURE 6-2. A graphical determination of the roots to the inhour equation
It should be noted that the range of p0 is bounded by unity,
k-1
- oo < P = -k- < 1. (6-44)
In the limiting cases we find
p0 =0~s 1 =0 (critical)
Po- 1 ~ s 1-oo ( supercritical)
( subcritical)
This last limit is particularly interesting because it implies that no matter how much
negative reactivity we introduce, we cannot shut the reactor down any faster than
on a period T= 1/A 1 determined by the longest-lived delayed neutron precursor. In
235
U fueled thermal reactors, A1 1~80 sec.
Several other limiting cases are of particular interest:
(1) Small reactivity insertions (p0 «/3): Then we can assume that the magni-
tude of s0 is small such that
Hence in the inhour equation (6-42) we can neglect s 1 in comparison with

I - 1 and \ to write
6
p0 =:::.s 1l+s 1 ~ /3;/\.
i= 1
Thus the reactor period is given by
(6-45)
However this is of course identical to the result suggested by our earlier

heuristic argument. Hence for small reactivities the reactor period is
determined essentially by the average neutron lifetime (/) including
delayed neutrons.
(2) Large positive reactivity insertions (p 0 » {3): In the opposite extreme of
large reactivity insertion, we can assume s 1»,\ to write the inhour
equation (6-42) as
s1 1-1 6 si+/31-1
Po~ - - - + --- L f3 = - - - - (6-46)
s1 + l -1 s1+1-1 i=I ' s1+1-1
Hence we can solve for the reactor period as
(6-47)
which is just the result we would have obtained by totally ignoring

delayed neutrons. Thus for large positive reactivity insertions, the reactor
response is determined essentially by the prompt neutron lifetime /.
(3) p0 = /3: This is essentially the "break point" between a reactor kinetic
response controlled by delayed neutrons and that governed by prompt
neutrons alone. If we refer to the point reactor kinetics equations (6-24), it
is apparent that for the reactor to be critical on prompt neutrons alone,
we would require p0 = {3. For p0 < /3 the delayed neutron contribution is
needed for reactor criticality, and hence the time response of the reactor
will be determined to a large extent by the time delay characterizing the
precursor /3-decay. Some common terminology is to refer to the range
0 < p < /3 as delayed critical, while p;;;, /3 is referred to as prompt critical or
prompt supercritical. Obviously it is very important to avoid this latter
situation in reactor operation, since the reactor time response would be
very rapid if prompt criticality were exceeded.
Actually the transition between these two classes of time response is not quite so
abrupt. In Figure 6-3 4 we have shown the .asymptotic reactor period T= s 1- 1
plotted versus p 0 for positive reactivity insertions in 233 U, 235 U, and 239 Pu-fueled
cores. The rather dramatic decrease in the reactor period in the neighborhood of
prompt critical p 0 = /3 should be noted.
The signifiGance of the prompt critical condition has led to the custom of
measuring reactivity in units of {3. More precisely, a p = /3 is referred to as one
dollar ($) of reactivity. For example, a reactivity of $.40 would correspond to
235
p = 0.4/3 = 0.0028 D..k / k = 280 pcm for a U-fueled thermal reactor.
C. The Inverse Method5

There are very few problems for which it is possible to obtain an exact
solution for P(t) given a specific p(t). Actually it is frequently more appropriate to
invert the problem by determining that p(t) which will yield a desired P(t)
behavior, since this is more in line with the philosophy of reactor control.
NUCLEAR REACTOR KINETICS I 247
10
~
.2l
0
"C
...>
·;;
·;;
10-1
.,""'
er:
10-2
10-s 10-5 10-4 10- 3 10-2 10- 1 10 10 2 103
Asymptotic period (T), sec
10
~
.2l
0
"C
...>
·;;
·;; 10- 1
.,""'
er:
10-2
10-s 10-5 10-4 10-3 10-2 10-1 10 102 103
10
~
.2l
0
"C
...>
·;;
·;; 10-1
.,""'
er:
10-2
10-s 10-5 10-4 10-3 10-2 10-1 10 102 103
FIGURE 6-3. Reactor period versus reactivity for fissile isotope delayed-neutron data 4
In order to solve for p(t) in terms of P ( t), let us derive yet another form of the
point reactor kinetics equation. Suppose we begin by formally solving the equations
for the precursor concentrations in terms of P (t):
(6-48)
where we have implicitly assumed that C;(t) exp;.\J-O as t - - oo and then let
T = t- t' to obtain the second integral. [Note that if we wish to take explicit account
of initial conditions on C;( t), say at a time t = t0 , then we would merely separate
out the contribution in the first form of the integral in Eq. (6-48) from t = - oo to
t= t0 to represent C;(t0).] We will now substitute this into the first of the point
reactor kinetics equations to write
(6-49)
If we define the delayed neutron kernel D ( 'T)
(6-50)
(noting that D ( 'T) d'T is just the probability that a delayed neutron will be emitted in
a time d'T following a fission event at 'T = 0) then we arrive at an "integro-
differential" form of the point reactor kinetics equations
(6-51)
We can now rearrange this equation to yield p(t) in terms of P(t):
(6-52)
This particular relationship is important for two reasons: (a) it can be used in
principle to determine the time-dependence of the applied reactivity required to
yield a specific power variation-that is, to program the control rod motion and (b)
the interpretation of the measured power responses in transient analyses of reactiv-
ity changes can be used to provide information about the feedback mechanism in
the reactor. Several specific applications of this relationship are of considerable
interest:
1. PERIODIC POWER VARIATION

Suppose we seek the reactivity that will yield a sinusoidally varying power of
the form
(6-53)
(Note that we must require P 0 > P 1 since a negative power would offend our
physical intuition.) Inserting this expression into Eq. (6-52) and performing a few
manipulations (left for the reader's enjoyment to the exercises at the end of the
chapter), we arrive at
P1 sin(wt-<1>)
p(t)= -p IY(iw)I----, (6-54)
o P, .
l+ psmwt
0
where
(6-55)
and Y ( iw) 1s a function which looks suspiciously like a chunk of the inhour
equation
Y(iw)=iw A+
[
L 1w+
6. /3.'A.
i= I I
l
. (6-56)
(It will cross our path again.) Notice in particular that the reactivity insertion that
gives rise to a purely sinusoidal power variation is periodic-but not sinusoidal (at
least for large power variations). One can show, in fact, that p(t) has a negative
bias. This fact proves of some importance in reactor oscillator experiments in
which an absorbing material is oscillated within the core and the core power
response is then measured in an effort to infer system parameters (e.g., /3 and A).
2. REACTIVITY AFTER A POSITIVE POWER TRANSIENT

As a second application of Eq. (6-52), suppose we determine the reactivity
present following a power transient in which the reactor power increases from its
initial value P O at t = 0, and then decreases back to P O at a later time t0
(6-57)
Now we know that for a positive transient, P (t0 - T) > P (t0) = P 0 and hence the
integral is always positive. Furthermore it is apparent that at t = t0, the slope
dP / dtl, 0 ..; 0. Thus we can infer that the reactivity following a positive power
excursion must, in fact, be negative. That is, the reactor must be taken subcritical
to return the power to its original level.
3. RAMP REACTIVITY INSERTION

Suppose the reactor power level is found to increase very rapidly from an
initial level P 0 as P 0 exp(at 2) for t>O. Then using Eq. (6-52) we can find that for
long times t,
p(t)~ {3+2Aat. (6-58)
Hence p( t) approaches a linear function of time-that is, a ramp insertion of

reactivity above prompt critical. We can turn this calculation around to infer that
the response of the reactor power P ( t) to a ramp insertion p( t) ~ yt should behave
as exp[ yt 2 /2 A] for long times. This conclusion has particular relevance for certain
classes of fast reactor accident models in which fuel melting is assumed to lead to
core reassembly in a prompt supercritical configuration with a ramp reactivity
insertion. Obviously the power increase resulting from such an occurrence would
be very rapid indeed.
~Approximate Solutions
As we have seen, exact solutions of the point reactor kinetics equations are
known for only a few special reactivity insertions. Hence we now turn our attention
to approximate schemes for solving these equations in the absence of feedback.
l. CONSTANT DELAYED NEUTRON PRODUCTION RATE

APPROXIMATION
In certain problems, such as when the reactor is shut down by rapid insertion
of the safety rods, we are interested in the response of the reactor power to a given
reactivity insertion in short time intervals following a time t 0 • During such short
time intervals we can ignore the change in the rate of production of delayed
neutrons, replacing C;(t) by C;(O). Hence in this approximation, the point reactor
kinetics equation becomes
dP [ P( t) - /3 ]
dt = A P(t)+ Q(t), (6-59)
where the effective source of delayed neutrons Q ( t) is now presumed known and
given by
6
Q(t)= L \C;(O). (6-60)
i= I
However we now just have a first-order inhomogeneous differential equation that

we can easily integrate for any given p(t).
A rather interesting application of this result is to the case of a reactor scram in
which control rods are rapidly inserted into the reactor core to shut the reactor
down in an emergency situation such as the loss of primary coolant flow. Since the
rod insertion takes a finite time, we cannot really treat the reactor scram as a step
reactivity change (as in Section 6-11-1). A more reasonable model is to assume a
negative ramp insertion, that is, p(t)= - yt. If we substitute this into Eq. (6-59) and
solve this equation subject to the initial condition imposed by steady-state power
operation prior to the scram,
(6-61)
then we find that the power level decreases as
lfo
2
P(t)=Po{exp- l(y~ +/31)+
1
dt'exp- l[fu-1') -/3(t-t')]}·
2
(6-62)
After the scram rods have been fully inserted, the negative reactivity becomes
constant, and one can determine P ( t) for subsequent time using our earlier
solutions of the point reactor kinetics equation for constant p(t) = - p0 •
2. THE PROMPT JUMP APPROXIMATION

In our earlier study of the response of the reactor power P (t) to a step change
in reactivity, we found that there was initially a very rapid transient behavior on a
time-scale characteristic of the prompt neutron lifetime, followed by a more slowly
varying response governed by delayed neutron behavior. Since this transient is so
rapid, a very useful approximation to make for systems below prompt critical is
NUCLEAR REACTOR KINETICS / 2S1
one in which the prompt neutron lifetime is essentially taken to be zero such that
the power level jumps instantaneously to its asymptotic behavior (recall Figure
6-1 ). This so-called prompt jump approximation is effected by merely neglecting the
time derivative dP / dt in the point reactor kinetics equation
6
O=[p(t)-,B]P(t)+A ~ \C;(t)
i= I
(6-63)
dC. ,B. .
-i= ~ P(t)-\C;(t).
Since the delayed neutron production cannot respond immediately to a step change
in reactivity, this model predicts that a reactivity jump from p 1 to p2 causes an
instantaneous change in reactor power from P 1 to P 2 as given by
(6-64)
r
The prompt jump approximation is frequently used in numerical studies of the i .

i '
point reactor kinetics equations since it eliminates the very short time scale due to
A which plagues finite difference methods. However it is also of use in simplifying
analytical estimates based on the one-effective delayed group model, since in this
case the precursor concentration C ( t) can be eliminated to find a simple first-order
differential equation for the power P ( t)
dP [ dp
[p(t)-,8] dt+ dt +>.p(t) ] P(t)=O. (6-65)
If p(t) is given, we can again solve for P ( t) as
(6-66)
where
-It [
A ( t) -
0
d-r
p(-r)+;\p(-r) J
,B - p( T) . (6-67)
For example, for a ramp insertion p(t) = -y,Bt, the prompt jump approximation
implies
\
I
p ( t) = p (0) e->.t [ 1 - yt]-( I+~). (6-68)
The prompt jump approximation is frequently found to yield an adequate
description of reactor kinetic behavior. Numerical solutions of the point reactor
kinetics equations have demonstrated the approximation is valid to within about
1% up to reactivity insertions of p 0 = $0.50. 6
2S2 / THE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
3. SMALL AMPLITUDE APPROXIMATION (LINEARIZATION)
Suppose that we assume that small reactivity variations will produce only
small changes in the reactor power from its equilibrium value P 0 • We already know
that this assumption is not true for a critical reactor since even a slight positive step
in reactivity gives rise to an exponentially increasing power response that eventu-
ally grows beyond any bound. However the assumption will still be valid if we
consider only short times following the step insertion. Furthermore for certain
classes of reactivity changes such as a periodic reactivity insertion with an
appropriate negative bias, the resulting power variations remain small for all times.
In this case, the point reactor kinetics equations reduce from a set of linear ODEs
with variable coefficients to a set of linear ODEs with constant coefficients.
Consider again the integrodifferential form of the point reactor kinetics equa-
tions
dP [ p(t) - /3 ] /3 ( oo
dt= A P(t)+ A Jo drD(r)P(t-r). (6-69)
If we now let p(t) denote the power variations about a reference level P 0
P(t)=P0 +p(t), (6-70)
then substitution of Eq. (6-70) into Eq. (6-69) and use of the identity
(6-71)
yields (after a bit of manipulation)
dp p(t) p(t)p(t) /3 (oo

dt = -r,Po+ A + A Jo drD(r)[p(t-r)-p(t)]. (6-72)
Thus far our analysis has been exact, merely cons1stmg of a bit of algebraic
manipulation. We now introduce our key approximation by assuming that p(t) and
p(t) are sufficiently small that we can neglect the second-order term p(t) p(t) to
write
dp p(t) /3 (oo
dt = A P0 + Alo drD(r)[p(t-r)-p(t)]. (6-73)
This approximation is sometimes (incorrectly) referred to as the linearization

approximation. Actually both the original point reactor kinetics equation as well as
Eq. (6-72) are already linear, provided we are given p(t). All we have done is to
remove p(t) as a time-dependent coefficient in the equation and replace it by a
time.-dependent inhomogeneous term p(t)P0 / A.
Equation (6-73) forms the basis of nuclear reactor stability analysis, since it
relates the response of the reactor power to a small change in reactivity. At this
point one can essentially turn the mathematical analysis of this equation over to
the systems (or control) engineer who will then proceed to hack away at it using the
powerful tools of linear systems analysis.7 It is not our intent here to venture very
far into this topic.
However with the recognition that many readers of this text (particularly
electrical engineers) will already be acquainted with methods of linear systems
analysis, we do feel it useful to provide a very short discussion of this topic in an
effort to bridge the gap to reactor kinetics. 8•9 (Other readers may wish to bypass the
remaining discussion of this section.)
The key tool used in linear systems analysis is the Laplace transform (defined
and discussed in Appendix G). We will define the Laplace transform j(s) of a
function f(t) by
(6-74)
Then we can easily Laplace transform the "linearized" reactor kinetics equation
(6-73) and solve for
ft (s) = P0 Z (s)p(s) (6-75)
where we have defined the zero power transfer function Z (s)
(6-76)
Hence to determine the behavior of the reactor power, we need only study the
poles of Z (s) and p(s). However there is a great deal more we can do by employing
the very powerful methods of linear systems analysis. In particular we can study
the stability of the reactor when it is operating at power (i.e., when we introduce
feedback).
The concept of a transfer function is much more general. The response (or
output) of any physical system to a signal (or input) applied to it can be expressed
in terms of such transfer functions. More precisely, we define
. Laplace transform of response

Transfer function= L f f.
ap1ace trans orm o mput
= Z (s). (6-77)
This can be conveniently represented by a "block diagram."
Input I
(cause) _ _ ____,,,:,. Z(s) 1 - - - - - - - + ) Output
(effect)
In our case, the input is the reactivity p( t), while the output is the fractional power
change p(t)/ P 0 . Of course other choices are possible, such as the inlet coolant
temperature and pressure.
The transfer function defined for a linear system in this way is sometimes called
the "open-loop" transfer function, since we have assumed that the output (power
level) does not affect the input (reactivity) in any way. Later we will consider the
case in which we allow feedback effects:
Z(s) f---r---Output
Feedback
(That is, we "close" the feedback loop.)

It should be noted that the concept of a transfer function allows one to use the
principle of superposition to write the total output resulting from several inputs as
the sum of the individual output from each of the inputs. In particular since
e{ 8 ( t)} = l, the response to a 8-function reactivity input is just the inverse Laplace
transform of the transfer function itself.
(6-78)
Here, '.£: ( t) is the so-called unit impulse response, that is, the power response to a
unit impulse reactivity insertion. The s1 are the roots of the inhour equation, while
we have noted that for a critical system, s0 = 0.
One can now use the convolution theorem (Appendix G) to reexpress Eq. (6-75)
in the "time domain" as
(6-79)
Note that '.£: (t- r) is just the Green's function for the linearized point reactor
kinetics equation. Hence it is not surprising that '.£: (t) plays an extremely important
role in the study of nuclear reactor kinetics. In particular one can infer the stability
properties of a linear system by studying '.£: ( t).
One defines such a linear system to be stable if its response to any bounded input
is also bounded. It is possible to show (see Problem 6-27) that a necessary and
sufficient condition for stability is for
(6-80)
In this regard it is interesting to note that since the zero power reactor response
function '.£:(t)-A- 1 as t-oo, the integral of '.£:(t) over all time is not bounded.
This then implies that a critical reactor without feedback is unstable with respect to
bounded reactivity inputs. However as we will see later, there are always sufficient
negative reactivity feedback mechanisms present in reactors to render them stable
under almost any conceivable operating conditions.
As an example of the use of the reactor transfer function, let us determine the
response of the reactor to a sinusoidal reactivity input
p(t) = 8psinwt. ( 6-81)
Then if we note that the Laplace transform of this input is just
_ 8pw
p(s)= 2 2' (6-82)
s +w
we can easily invert Eq. (6-75) to find
(6-83)
where cj>(w) = arg[Z (iw)] is known as the phase shift of the transfer function, while
IZ(iw)I is known as the system gain. The first term in Eq. (6-83) arises from the
poles of p(s) on the imaginary axis at s = ± iw, while the remaining terms are due to
the poles of Z (s). These latter poles are just the roots si of the inhour equation
Y (s) = 0. Of course for the critical system we are considering, s 7 < s6 < · · · < s 1 = 0.
Hence as t ➔ oo, only the oscillating terms and the s 1 = 0 term remain, and we find
the asymptotic behavior of the power oscillations as
p(t) 8p
Po =8plZ(iw)lsin(wt+cj>)+ wA. (6-84)
Notice in particular that there is a shift in the average power level of P0 8p/wA.
If a reactor operating at a steady-state power level is subjected to a sinusoidal
perturbation in reactivity, the power will oscillate with the source frequency, but
with a phase shift cj>(w)=arg{Z} (actually a phase lag) and an amplitude propor-
tional to G(w)=IZ(iw)I, Hence we can obtain the values of Z(s) on the imaginary
axis in the complex s-plane by measuring experimentally the amplitude and relative
phase of power oscillations as functions of frequency, induced by a sinusoidal
reactivity insertion with constant amplitude. This is the basis of the zero-power
pile-oscillator experiment. We have sketched the gain and phase shift of the zero
power transfer function for a 235 U / 238 U-fueled system in Figure 6-4.
F. Some General Comments on Applications of the Point Reactor

/ Kinetics Equations ·
'------
The point reactor kinetics equations are commonly applied to analyze most
short-term transients in nuclear reactor behavior. The popularity of the point
reactor kinetics model is not due necessarily to its validity (which is limited), but
rather the fact that alternative methods that also treat the spatial dependence of the
flux are usually prohibitively expensive to use in the large variety of problems
typically encountered in a core design effort.
2S6 / THE ONE-SPEED DIFFUSION MODEL OF A NUCLEAR REACTOR
8
4
3
'-
0
t.:,
0 -4
ai
~ -8
0
N
-12
-16
0.01 0.1 10 100 1000

w
-20°
]"
6, -4CY'
~
"' -6CY'
"" -80°
-100°
0.01 0.1 10 100

w
FIGURE 6-4. Gain and phase shift of the zero-power reactor transfer function
Most experimental measurements on reactors involving a reactivity change can

be analyzed via the point reactor kinetics equations. Indeed since such experiments
are usually performed at zero power, it is frequently not necessary to append to the
point reactor kinetics equations suitable models of reactivity feedback. Although
we will briefly review a variety of experimental techniques commonly used to
measure reactor kinetics parameters later in this chapter, we would merely mention
here as an example experiments in which the reactivity worth of control rods are
measured by withdrawing a rod and then determining the resulting asymptotic
reactor period. The inhour equation can then be used to relate this period to the
worth of the control rod.
Normal operating transients in nuclear power reactors are usually very easy to
analyze, since they are typically rather slow. This is because the rate of change of
reactor power level in a large plant is limited not by neutronic considerations, but
rather by the rate of temperature change allowable in nonnuclear plant com-
ponents. For such slow transients (typically a 5% change in power per minute or
less for maneuvering) reactivity insertions are kept far below prompt critical, and
hence the prompt jump approximation, in which the mean generation time A is set
equal to zero, is adequate. Furthermore on such a time scale the spatial power
shape usually remains in a fundamental mode and hence the point reactor kinetics
equations are valid. The solution of the point reactor kinetics equations for such
slow transients does necessitate consideration of reactivity feedback, however.
Frequently one finds the point reactor kinetics equations augmented with transient
models of the remainder of the plant suffices for on-line system performance
studies in the plant itself (using hybrid digital-analog process computers).
The most stringent application of the point reactor kinetics equations is in the
analysis of hypothetical reactor accident situations in which reactivity insertions
may be quite large (several $) and power transients very rapid. Indeed one
frequently encounters situations in which such equations are not valid. Typically
one uses them to provide a parameter study of the effects of various input data
such as reactivity insertions on reactor behavior under the postulated accident
conditions. Occasionally spatially dependent kinetics calculations are performed to
check the predictions of the point reactor model. Obviously an effort is always
made to generate conservative predictions.
For most severe transients in thermal reactors, such as a loss of coolant accident
or the ejection of a control rod, the usual transient thermal-hydraulics models are
adequate to supplement the core neutronics calculations. However in fast systems
very severe transients may lead to fuel melting and slumping with an appreciable
change in core configuration. Indeed such core rearrangements may produce large
positive reactivities, leading to very large energy release, followed by core disas-
sembly (that is, a small nuclear explosion). Such severe accidents, although quite
remote in probability, must nevertheless be studied in reactor analysis. They
require not only models of the core neutronics (e.g., the point reactor kinetics
equations) and thermal-hydraulics, but as well models of the physical motion of the
core due to melting and disassembly. Needless to say, such analyses can become
very complicated indeed.
III. REACTIVITY FEEDBACK AND REACTOR DYNAMICS
A. Mathematical Description of Feedback

Thus far we have assumed that the reactivity p(t) appearing in the point
reactor kinetics equations is a given function of time. However we know, in fact,
that it depends on the neutron flux (or power level) itself. This dependence arises
because the reactivity depends on macroscopic cross sections, which themselves
involve the atomic number densities of materials in the core:
~(r, t) = N (r, t)a(r, t). (6-85)
Now it is easily understandable how the atomic density N (r, t) can depend on the
reactor power level, since: (a) material densities depend on temperature T, which in
turn depends on the power distribution and hence the flux and (b) the concentra-
tions of ~ain nuclei is constantly changing due to neutron interactions (buildup
of poison or burnup of fuel). However it should also be noted that we have
explicitly written the microscopic cross sections as explicit functions of r and t.
This must be done since the cross sections that appear in our one-speed diffusion
model are actually averages of the true energy-dependent microscopic cross sec-
tions over an energy spectrum characterizing the neutrons in the reactor core. And
this neutron energy spectrum will itself depend on the temperature distribution in
the core, and hence the reactor power level.
Such reactivity variation with temperature is the principal feedback mechanism
determining the inherent stability of a nuclear reactor with respect to short-term
fluctuations in power level. Evidently our first task in constructing a model of
temperature feedback is to determine the temperature distribution in the core. Of
course, one could begin by writing the fundamental equations of heat and mass
transport characterizing the reactor. That is, one could write the equations of
thermal conduction and convection using the fission energy deposition as the
source of heat generation. The motion of the coolant through the core would then
be described by the equations of hydrodynamics. As we will find in our detailed
development of thermal-hydraulic core analysis in Chapter 12, this fundamental
approach results in a formidable set of nonlinear partial differential equations
unless further simplifications are introduced.
Hence it has become customary to avoid the complexity of a direct sqlution of
the full set of thermal-hydraulic equations by replacing the spatially dependent
description by a "lumped parameter" model similar in philosophy to the point
reactor kinetics model. The reactor core is typically characterized by several
average temperatures, such as an average fuel temperature, moderator temperature,
and coolant temperature. One then models a simple dependence of the reactivity
on these temperature variables.
Of course one must append to the original point reactor kinetics model equations
describing the changes of these core temperatures produced by changes in the
reactor power level. Typically such calculations are based on a model of a single
average fuel element and its associated coolant channel, and describes the conduc-
tion of fission heat out of the fuel element and into the coolant, and then its
subsequent convection out of the reactor core. We will consider such models in
some detail in Chapter 12.
Our more immediate concern in this chapter, however, is to study how the results
of such thermal-hydraulic models, namely the core component average tempera-
tures such as TF (fuel) and TM (moderator or coolant), can be used in suitable
models of reactivity feedback. To this end let us return to consider our point
reactor kinetics model. We will write the reactivity p(t) as a sum of two contribu-
tions
(6-86)
The 8p notation signifies that the reactivity is measured with respect to the
equilibrium power level P0 [for which p=O]. Furthermore, 8Pext(t) represents the
"externally" controlled reactivity insertion such as by adjusting a control rod.
8pr[P] denotes the change in reactivity corresponding to inherent feedback
mechanisms. This latter component is written generally as a functional of the
reactor power level.
When the reactor is operating at a steady-state power level P 0 , then there will be
a certain feedback reactivity Pr[ P 0]. Since in almost all cases this will correspond to
a negative reactivity, it is customary to refer to Pr[P0 ] as the power defect in
reactivity. To sustain the criticality of the system, we must supply a counteracting
external reactivity p 0 (such as by withdrawing control rods) such that
(6-87)
In this sense then, we define our incremental reactivities as
8pextC t) = PextC t)- Po,

8pr[ P] = Pr[ P] - Pr[ Pol· (6-88)
It is also useful to recall our definition of the incremental power
(6-89)
From Eq. (6-88) we note that 8pr[p=0]=8pr[P0]=0.

We can now sketch the block diagram characterizing a reactor with feedback as
shown in Figure 6-5. In particular, it should be recalled that we have already
analyzed the "black box" describing the reactor kinetics, since this is just given by
the point reactor kinetics equation without feedback, which can be most con-
veniently written in the form Eq. (6-72). This equation can be regarded as
determiningp(t) for any p(t). Of course we commonly analyze this equation under
one of a variety of approximations, for example: (a) ignoring delayed neutrons
( D = 0) for very large reactivity insertions, (b) prompt jump approximation, and (c)
linearization for small reactivity insertions, in which case we can solve Eq. (6-73) in
terms of the open loop or zero-power transfer function to find
( 6-90)
We therefore turn our attention to the study of the black box describing the
feedback functional 8pr[ p ].
B. Models of Temperature Feedback
1. TEMPERATURE COEFFICIENT OF REACTIVITY
Of most concern in the study of short-term reactivity feedback is the effect of

the core temperature T on the multiplication of the core. One usually expresses this
in terms of a temperature coefficient of reactivity a.T defined as
( 6-91)
p(t)
-
OPext p
p(t): p[p]
Po Incremental
External Net power
L reactivity reactivity
Neutron
opt kinetics
opi[pl
Feedback
mechanisms
FIGURE 6-5. Closed-loop block diagram

This should more properly be referred to as an isothermal temperature coefficient,

since it assumes that the core can be characterized by a single uniform temperature
. T.
Such a simple model illustrates quite clearly the effect of temperature feedback
on reactor stability. For if a reactor were to possess a positive a.r, then an increase
in temperature would produce an increase in p, hence the power would increase,
causing a further increase in temperature, and so on. In this sense, the reactor
would be unstable with respect to temperature or power variations. The more
desirable situation is that in which a.T is negative, since then an increase in
temperature will cause a decrease in p, hence a decrease in reactor power and
temperature which tends to stabilize the reactor power level.
The temperature coefficient of reactivity depends on a great many processes
occurring within the core. Since the details of these processes depend on the
specific reactor type under consideration, we will defer a discussion of just how
such temperature coefficients are evaluated to Chapter 14.
However we should remark at this point that the concept of such an isothermal
temperature coefficient has a very limited range of usefulness in reactor analysis. In
a heterogeneous reactor, most of the fission energy release is confined to the fuel
elements. This energy m~st then be transferred by thermal conduction through the
fuel element, the clad, and then into the coolant before it can be withdrawn from
the core. Hence one has a very nonuniform temperature distribution. For example,
fuel centerline temperatures may range as high as 2700°C, while coolant tempera-
tures are as low as 250°C.
Hence one usually introduces several such temperature coefficients of reactivity,
each of which characterizes the reactivity feedback due to a variation in the
effective or average temperatures of each major component of the core, such as
fuel, moderator, and structure
(6-92)
However one must also keep in mind that the various thermal processes in the core
are characterized by widely different time behaviors. The fuel temperature re-
sponds relatively rapidly to any power level changes. However it takes an appreci-
able amount of time to transfer this energy to the coolant, and hence its tempera-
ture response is much slower. For this reason, it is convenient to divide up the
temperature coefficient of reactivity into prompt and delayed components.
Effects that depend on the instantaneous state of the fuel-for instance, re-
sonance absorption (Doppler effect) or thermal distortion of fuel elements-may
be regarded as prompt, while effects that depend primarily on the moderator or
coolant-neutron energy spectrum and thermal expansion of moderator material-
are delayed. Prompt feedback mechanisms are of particular importance in reactor
safety studies, since they play a very important role in limiting any reactor
transients that may occur.
For these reasons, the isothermal temperature coefficient of reactivity is not
really a very useful quantity. A more useful quantity is the change in reactivity
caused by a change in reactor power. We will consider this in detail in the next
section.
There is one instance in which the concept of an isothermal temperature
coefficient is of some use, however. When the reactor is at zero power, there is no
fission energy being released in the fuel, and hence the entire reactor core can
essentially be characterized by a single temperature. As the temperature of the core
is increased, say by heating the primary coolant, there will be a decrease in core
multiplication due to temperature feedback. One defines the so-called temperature
defect of reactivity, .:ipTD• as the change in reactivity that occurs in taking the
reactor core from the fuel-loading temperature (i.e., the ambient temperature) to
the zero power operating temperature
_ ( T zero power ap
iiPTn- J ar dT. (6-93)
Tambient
7
The magnitude of .:ipTD is primarily determined by the coolant temperature

coefficient of reactivity, and depends sensitively on the moderator-to-fuel ratio and
soluble poison concentration. In a LWR .:ipT0 ~.02-.04(.:ik/k).
2. POWER COEFFICIENTS OF REACTIVITY

A far more useful parameter characterizing feedback is the power coefficient
of reactivity defined by
(6-94)
where the 'I'; are again the effective temperatures associated with each core
component. Such a parameter takes account of the temperature differences occur-
ring in a reactor while it is operating at power. The reactivity due to power
feedback can then be written as
(6-95)
Obviously if a reactor is to be inherently stable against power-level fluctuations, it

must be designed with ap< 0.
A closely related quantity is the power defect, which is defined to be the change
in reactivity taking place between zero power and full power
-lpfullpower ap
lipp 0 = aP dP. (6-96)
0
The power defect can be quite sizable. For example, in the LWR the power defect
is typically of the order .:ipp0 -.0l - .03(.:ik/ k).
Thus far we have only discussed how the reactivity depends on temperature. The
remaining problem of how the temperature depends on the power level is strongly
relate~ the reactor type and involves a thermal-hydraulics analysis of the core.
For slow power changes, one can use the steady-state analysis developed in
Chapter 12 to determine the core temperatures for a given power level and hence
the power coefficient of reactivity in terms of the temperature coefficients of the
various components of the core.
Such a steady-state thermal analysis will no longer be valid for more rapid power
transients. For example, the thermal time constant of the fuel is frequently as large
as 10 seconds. For power transients on shorter time scales the fuel temperature
coefficient of reactivity (principally determined by the Doppler effect) will be the
dominant factor in determining the power coefficient of reactivity.
Although a realistic description of temperature feedback involves a transient
thermal-hydraulics model of the reactor core such as those we will develop in
Chapter 12, several alternative lumped parameter models of the reactor tempera-
ture dependence are occasionally used for qualitative studies of reactor dynamics.
One common model assumes a single effective coolant temperature Tc, and
models the fuel temperature TF by Newton's law of cooling
(6-97)
where K and Y are thermal constants characterizing the core. At the opposite
extreme would be an adiabatic model, in which the heat loss is assumed to be
negligible (such as in a very rapid transient)
dTF
dt = KP ( t). (6-98)
Still another model assumes constant power removal such that

dTF
dt = K(P- P0 ). (6-99)
All of these models can be considered as special cases of a general linear feedback
functional:
(6-100)
A little inspection should convince you that the feedback kernel h(t) in each of
these cases takes the form:
Newton's law of cooling: h(t) = aK e-yr

Adiabatic model: h(t)=aK[P0 =0] (6-101)
Constant power removal: h(t)= aK.
If more reliable design information is required, then one is usually forced to go

to a multigroup diffusion calculation coupled with a detailed thermal-hydraulic
analysis of the core. One can then calculate the multiplication for several different
power levels, and hence determine directly the dependence of reactivity on power
level. The corresponding power and temperature coefficients are frequently pre-
calculated and stored in tabular form for use in determining the reactivity feedback
to use in point reactor kinetics models.
C. The Transfer Function of a Reactor with Feedback
1. CLOSED-LOOP TRANSFER FUNCTION
Let us now return to consider the dynamic behavior of a reactor with

feedback. In particular we will consider the effects of feedback on the reactor
response to an external reactivity insertion. First we will consider the power

oscillations resulting from a periodic reactivity insertion of small amplitude, since
this in effect measures the reactivity-to-power transfer function.
A typical series of reactivity power transfer function gain measurements 10• 11 is
shown in Fig. 6-6. In particular notice how different the transfer function at power
is from the zero power transfer function. The .marked resonance behavior in the
vicinity of .03 cps at P 0 = 550 kW in the first plot is apparent. Such behavior is due
to the presence of feedback. As the power level increases, the resonance peak
becomes narrower and higher. As we shall see, this implies that for sufficiently
large powers, the reactor is unstable.
Let us go back and develop a mathematical expression for the transfer function
characterizing a reactor with feedback. 1 We will restrict ourselves to small power
variations about the equilibrium level P O so that the feedback can be adequately
represented as a linear functional similar to Eq. (6-100)
(6-102)
Note that with this sign convention the negative feedback necessary for reactor
2000
1000
800
600
] 400
(.'.)
........
'
2 kW''
200
100 ' - - - - - - - - ' - - - - ' - - - - - - - - - ' - - - - '

0.01 0.05 0.1 0.5
Frequency (cps)
+12
+8
3
(.'.) 0
0
Cl
.2 -4
0
N
-8
-12
-16
0.01 0.1 1 10 100
w, rad/sec
FIGURE 6-6. Examples of measured closed-loop transfer functions 10• 11

stability will imply that h( T) < 0. It should also be kept in mind that the feedback
kernel h(t) actually depends on the equilibrium power level P 0 •
If we substitute Eq. (6-102) into Eq. (6-72), we find
00
dp
dt = AI [ 8pext(t)+ Jo( dTh(T)p(t-T) ] [Po+p(t)]
/3 roo
+ AJo dTD(T)[p(t-T)-p(t)]. (6-103)
Of course this equation is still nonlinear. We will consider only small power
variations p(t)«P0 such that we can linearize Eq. (6-103) to write
Now, as before, we will assume the reactor is operating at a steady state power
level P 0 prior to t=0. Then by Laplace transforming Eq. (6-104) we find
or
(6-105)
where Z (s) is the usual zero power transfer function given by Eq. (6-76) while
H (s) = e{h(t)} is the feedback transfer function. We have further defined the
reactivity-to-power or closed-loop transfer function L(s)
Z (s)
L(s)- 1- P 0 H (s)Z (s) (6-106)
Notice that as P0 -o, L(s)-z (s), the zero power transfer function. This notation is
consistent with our earlier block diagram that has been relabeled in Figure 6-7. Just
p(s)
-
0 Pext (s) Po
Z(s)
-
op, (s)
H(s)
FIGURE 6-7. Closed-loop Transfer function

as in the zero power case, the inverse of L(s), /(t)- 1

e-
{L(s)}, is the Green's
function for the point-reactor kinetics equation with feedback
(6-107)
Let us examine L(s) in a bit more detail. Unlike Z (s), L(s) is analytic at s = 0
with a value
(6-108)
Hence the long time response to a positive step reactivity insertion of magnitude
8pext-8p 0 is just
1
p( t) = P 0 8p 0 fo dT I ( T)-P 0 8p 0 L(O). (6-109)
This implies that the reactor approaches a new equilibrium power level
P (t-oo) = P 0 [ 1 + 8p 0 L(O)]. (6-110)
This is in sharp contrast to the zero power (i.e., no feedback) reactor whose power
level grew exponentially for long times. The equilibrium state occurs when the
power level reaches a value such that the feedback reactivity just compensates for
the step reactivity insertion
(6-111)
2. RESPONSE TO A SINUSOIDAL REACTIVITY INSERTION

Let us once again consider a sinusoidal reactivity variation
(6-112)
Then ,ihe long time response is given in analogy to the zero feedback case [Eq.
(6-84)]by
p(t)
~ = IL(iw)I sin (wt +<j> ), </>( w) = arg { L ( i w) }. (6-113)
0
Now if we recall the form for L(s) given by Eq. (6-106), we can see that a
resonance in the gain G(w)= IL(iw)I will occur when
1- P0 H (iw)Z (iw) = 1 + P0 IH(iw)Z (iw)lexpi0-O, (6-114)
where
0(w)=arg{ -H(iw)Z(iw)}. (6-115)

However for the resonance condition [Eq. (6-122)] to be satisfied, we require that
both of the conditions
0 (w) = 180° and P0 IH (iw)Z (iw)I = 1 (6-116)
be simultaneously satisfied. These two conditions will determine a critical power

level P crit and a resonance frequency wcrit at which a true resonance situation will
arise. As long as the reactor power P 0 is kept below Pcrit' only a finite resonance
peak such as those illustrated in Figure 6-6 will occur. At this peak, the gain IL(iw)I
will assume a value
(6-117)
In the light of this discussion we can more readily understand the transfer function
gain measurements given in Figure 6-6.
In almost all cases of reactor design, the critical power level P crit above which the
reactor is susceptible to such instabilities is always quite far above actual or design
operating levels. Nevertheless it is important to be able to anticipate such inherent
instabilities in the design so that effective countermeasures can be taken (such as
by modification of the original design or by the addition of stabilizing feedback
control). Such considerations arise in reactor stability analysis.
3. LINEAR STABILITY ANALYSIS

The stability of a reactor with feedback can be investigated by examining the
singularities of the closed-loop transfer function
L(s)= Z(s) (6-118)

l-P0 H(s)Z(s)
in the complex s-plane. First notice that since Z (s) appears both in the numerator
and denominator, its poles sj [that is, the roots of the inhour equation Y (sj)
= Z -1(s) = OJ "cancel." Hence the poles of L(s) are simply the zeros of
l-P0 H(s)Z(s)=O. (6-119)
Now suppose that Eq. (6-119) were to have a simple root at s = s 0 [i.e., a pole of
L(s)]. Then when we invert the Laplace transform, this pole will contribute a term
inp(t) of the form exp(s0 t). Hence if s0 is in the right half s-plane (RHP), thenp(t)
will grow exponentially in time. This would imply an unstable response to an
applied reactivity perturbation (within the linear approximation, of course). If the
root s0 lies in the LHP, then terms of the form exp(s 0 t) will decay in time. Hence to
study reactor stability, it is obviously important to determine if any of the poles of
L(s) [i.e., zeros of Eq. (6-119)] lie in the RHP.
Instabilities may arise for sufficiently large power levels P 0 even with negative
reactivity feedback, as the sequence of diagrams in Figure 6-8 indicate. For PO=0,
we can identify the poles of L(s) as just those of Z (s) [i.e., the inhour equation
roots]. Then, for the case in which <Xp=H(O)<O, increasing P 0 will provide more
negative feedback and hence shift the pole s 1 to the left and into the LHP.
However the other poles will also shift. In particular, the pole s2 will shift to the
right. For·some sufficiently large power level P0 , the poles s 1 and s2 will coalesce,
forming a complex conjugate pair that moves off of the real axis and into the
complex plane. For still larger P 0 , this pair moves back to the right, until for some
critical power level P0 = Pcrit the poles move into the RHP and the reactor becomes
unstable. When the poles cross the imaginary axis, we observe a "resonance" in the
tfansfer function gain JL(iw)J as we have already noted. Hence the problem of the
linear stability of our equilibrium state reduces to the problem of determining the
sign of the real parts of the poles of L(s). In particular, we have found that a
reactor is linearly "strictly stable" when the poles of L(s) all have negative real
parts. The response of the power in the critical case, when any of these poles have
real parts equal to zero, is not correctly described by linear analysis and depends
on nonlinearities of the point reactor kinetics equation.
s-plane s-plane * s-plane \ s-plane
P0 = 0
s3 s2
l
P0 = P1> 0
s1
P 0 = P2 > P~
-¾-X-------X-X-X-+--- -X X-+---
s3 s2 s1
P0 = Pc,it/
\
/}II;
\
'")f
I
FIGURE 6-8. Shifting of poles of the closed-loop transfer fnnction with increasing reactor power
level P 0
r-
\
Ther~ are numerous tricks for determining the signs of these poles. These include
methods 12 such as those utilizing Nyquist diagrams, root-locus plots, and Routh-
Hurwitz criteria. However such subjects are more properly the concern of linear
system analysis or control theory, so we will simply refer the interested reader to
the references listed at the end of the chapter. 1• 8• 9
4. NONLINEAR POINT REACTOR KINETICS

Thus far our study of the point reactor kinetics equations with feedback has
been restricted to situations in which the reactivity changes and corresponding
power changes are sufficiently small that these equations can be linearized. In
particular our study of the stability of the reactor has been restricted to the
consideration of stability in the small-that is, to the study of perturbations and
responses sufficiently small for a linear analysis. However for larger perturbations
nonlinear effects must be taken into account. In these cases, the resulting con-
clusions about stability may be quite different.
For example, we have seen that the linearized point reactor kinetics equation
predicts that the reactor will be unstable if the power exceeds some critical value
Pcrit•However even though the reactor is linearly unstable, it may be stable in the
nonlinear description. Hence it is of some interest to study the significance of linear
stability theory within the more general framework of nonlinear stability theory.
It should be remarked, however, that there are many different approaches to
nonlinear point reactor kinetics-none of which is completely satisfactory.
Furthermore such results as do exist provide only sufficient (as opposed to neces-
sary) conditions for stability. Sufficiency conditions are usually much too
restrictive for practical application. Furthermore, the very limited validity of the
models investigated by nonlinear stability methods usually discount their value for
practical reactor design. And of course there is also the more pragmatic philosophy
adopted in most nuclear reactor design favoring an ultraconservative design that
guarantees linear stability under all conceivable operating conditions, hence obviat-
ing the need for a nonlinear stability analysis.
5. SOME FINAL COMMENTS ON REACTOR STABILITY ANALYSIS

It might seem that an integral part of any reactor safety analysis would be an
investigation of the stability of the reactor design. In particular such an analysis
should consider the possibility that the reactor might become unstable under some
feasible combination of operating conditions. In practice, however, such stability
studies do not play near as significant a role in reactor design as one might expect.
Reactor instabilities usually can occur only if one of the temperature feedback
coefficients happens to be positive over some range of operating conditions.
However it is usually quite easy to design a power reactor so that it will always be
characterized by large, negative temperature coefficients (as we will see in Chapter
14). Indeed all power reactor designs tend to be ultraconservative in this respect (as
they do with respect to all safety considerations). Hence there has been relatively
little motivation to perform detailed stability analyses of reactor core designs. And
in those instances in which methods of stability analysis have been applied to
practical reactor designs, investigations have usually been limited to the linear
domain.
IV. EXPERIMENTAL DETERMINATION OF REACTOR

KINETIC PARAMETERS AND REACTIVITY
From our earlier development of the point reactor kinetics equations, we have
found that the three most important parameters characterizing the kinetic behavior
of a nuclear reactor are the reactivity p, the prompt generation time A, and the
effective delayed neutron fraction /3. A variety of experimental techniques have
been developed to measure both these as well as dynamic (i.e., feedback)
characteristics of the reactor. Such methods can be classified as either static or
dynamic measurement techniques.
A. Static Techniques for Reactivity Determination

l. NEUTRON MULTIPLICATION MEASUREMENT (RECIPROCAL
MULTIPLICATION METHOD)
Perhaps the most common measurement is the so-called critical loading
experiment or reciprocal multiplication method, in which the steady-state neutron
flux resulting from a source in a subcritical assembly is measured as fuel is added

to the assembly. If the reactor is characterized by a multiplication factor k, then
one can crudely think of the amplification of the original source neutrons by the
assembly as given by
'DL = S0 + kS0 + k 2S 0 + · · ·
=(l-k)- 1S 0 =MS0 , (6-120)
where S 0 is the rate at which source neutrons are emitted in the reactor. The usual
procedure for a safe approach to delayed critical in core loading consists of
plotting M- 1 (or the reciprocal neutron counting rate) as a function of some
parameter that controls reactivity (e.g., fuel mass) and then extrapolating this M- 1
plot to zero to determine the critical loading (as sketched in Figure 6-9). During a
stepwise approach to delayed critical by the reciprocal multiplication method, the
neutron level following each addition of reactivity must be allowed to stabilize in
order to obtain an accurate indication of the asymptotic multiplication before
proceeding with the next reactivity addition. As one approaches delayed critical,
the buildup of the precursor concentrations can become bothersome.
\X
\ Data points
\ X
\ X
~
' '\
X
a, X
>
C
'', Estimate of
I ', critical mass
~ - - - -Fuel
-- --x--
mass
-
FIGURE 6-9. Critical loading measurements
2. FUEL SUBSTITUTION TECHNIQUES4
One can determine reactivity by uniformly substituting a i:, ,,son for a small
fraction of the fuel. The poison is usually chosen such that it has the same
scattering and absorption properties as the fuel (or as closely as possible). Then
first-order perturbation theory yields a reactivity change due to the substituted
poison of
(6-121)
This substitution method can be used for control-rod calibration by balancing the
effect of a given rod movement by adding a proper amount of distributed poison.
B. Dynamic Techniques for Reactivity Measurements
1. ASYMPTOTIC PERIOD MEASUREMENTS
Perhaps the simplest type of kinetic measurement one can perform is to make
a small perturbation in the core composition of a critical reactor, and then to
measure the stable or asymptotic period of the resultant reactor transient. Using
the inhour equation, one can then infer the reactivity "worth" of the perturbation
from a measurement of the asymptotic period. It should be noted that the period
method for all practical purposes applies only to positive periods, since negative
periods are dominated by the longest delayed neutron precursor decay and hence
provide very low sensitivity to negative reactivity.
2. ROD DROP METHOD

Let us consider a reactor operating at some equilibrium power level P 0 when
it is suddenly shut down by the introduction of a negative reactivity - 8p (the "rod
drop"). From our earlier studies of the point reactor kinetics equations we know
that after a few prompt neutron lifetimes, the reactor power level drops to a lower
level P 1, determined by the amount of reactivity insertion and remains at this
"quasistatic" level until it is ultimately decreased by delayed neutron precursor
decay. We can use our earlier result [Eq. (6-64)] from the prompt jump approxima-
tion to write
(6-122)
Hence we can solve for the reactivity insertion in dollars in terms of the power
levels P O and P I as
(6-123)
To determine P 1, one need only extrapolate back the asymptotic behavior follow-
ing the rod drop to t = 0.
3. SOURCE JERK METHOD

A very similar technique can be used to measure the multiplication of a
subcritical assembly. Suppose we consider such a subcritical system maintained at
a power level P 0 by a neutron source of strength S0 • One can express the
equilibrium power level P 0 using the point reactor kinetics equation as
(6-124)
where 8p is the degree of subcriticality of the system. Suppose the source is now
jerked out of the core. Once again the prompt jump approximation can be used to
express the lower "quasistatic" power level P 1 as
(6-125)
Using Eqs. (6-124) and (6-125), along with the equilibrium forms 'of the point-
reactor kinetics equation prior to the source jerk, one can find
(6-126)
Hence the dollars subcritical of the assembly is given very simply by
(6-127)
The source jerk method is very similar to the rod drop method in concept. However
it is somewhat simpler to perform since it requires only the removal of a small mass
of material (the source) in contrast to the rapid release of one or more control rods.
4. ROD OSCILLATOR METHOD

If one oscillates a control rod in a sinusoidal fashion in a critical reactor,
there will be a corresponding oscillation in the reactor power. By measuring the
gain and magnitude of the phase shift characterizing the power oscillations, one
can measure the reactor transfer function, either at zero power or operating power:
~ = L(iw)Sp. (6-128)
0
For low powers, L(iw)~Z (iw). In this case, the high-frequency behavior of the
transfer function
1
Z(iw)~-- (6-129)
~
wA
can be used t~measure either the prompt generation time A or the reactivity worth
of the oscillating rod,
SP
Sp=wAP. (6-130)
0
Such oscillator measurements can also be used to determine the stability

characteristics of the reactor.
5. PULSED NEUTRON METHODS

By measuring the transient behavior of the neutron population following a
burst of neutrons injected into an assembly, one can measure a variety of impor-
tant parameters. Suppose we first consider a pulse of neutrons injected at t = 0 into
a nonmultiplying assembly. Then according to one-speed diffusion theory, the flux
satisfies
(6-131)
subject to the initial condition applying to the pulsed source, <f>(r, 0) = <f> 0(r). For
long times, the flux in the assembly will approach the asymptotic form
(6-132)
where i/; 1(r) is the fundamental mode of the assembly geometry. Hence the
asymptotic behavior of the flux is governed by the decay constant
(6-133)
If one measures a 0 for various assembly sizes, then plotting a 0 against Bf will yield
VLa (the intercept at Bj=O) and vD (the slope). (See Figure 6-10.) Actually, there
are higher order terms in Bf due to both transport and energy-dependent correc-
tions
(6-134)
which add a curvature to the ao(B;) plot.

In multiplying assemblies if one assumes that both the pulse injection and
measurement are performed on a time scale short compared to delayed neutron
lifetimes, then one can use
(6-135)
to find decay constants of the form

a 0 = v(La - PLr) + vDB;, (6-136)
B2
g
FIGURE 6-10. The fundamental decay constant versus geometric buckling in a pulsed neutron
experiment
In this case, the a 0 versusB;curve appears as that shown in Figure 6-10.

One can use this technique to directly measure reactivity. For a rapid pulse, the
time behavior of the flux or power will be determined by prompt neutron kinetics
such as
I dP p-[3
a = -
o Pdt
- = -
A.- (6-137)
Hence a 0 can be used to infer p, provided /3 and A are known. If the core is
maintained at delayed critical, then p = 0 and the pulsed neutron method measures
a0 =-A/f3. (6-138)
C. Noise Analysis in Nuclear Reactors 13, 14

Thus far we have been analyzing the reactor as essentially a deterministic
system. Yet we know that all of the dynamic variables describing the reactor
(power level, flux, temperature, etc.) actually fluctuate in a statistical fashion about
some mean value. We are unable to predict with certainty the future values of these
variables but rather can only specify the probability that they will assume a certain
value.
The statistical nature of the neutron diffusion process has been stressed re-
peatedly. Moreover the concept of a cross section, and, indeed, even the quantum
mechanical description of the neutron interactions, must be interpreted in a
probabilistic sense. There are numerous other sources of such statistical fluctua-
tions or noise in a reactor, such as in fluid flow, coolant boiling, and mechanical
vibration. In fact, statistics enters even into the measurement of the reactor state
due to detector noise.
At low power le~s the statistical fluctuations associated with the fundamental
nuclear processes occurring in the core will be dominant. However at higher power
levels, the reactor noise will be predominantly due to disturbances of a nonnuclear
nature (e.g., coolant flow). Regardless of its origin, reactor noise is important in
reactor dynamics for at least two reasons. First, it interferes with the precision with
which a desired quantity can be measured in a reactor; that is, one must extract the
signal of interest out of the background noise. Second, however, since the noise
originates from various processes occurring in the reactor, it can actually serve as a
source of information about the system. The latter of these properties is of most
interest to us here. We will study how noise analysis can be used to measure the
transfer function of the reactor.
1. CROSS CORRELATION METHODS
Let us begin by introducing two definitions. We will define the autocorrelation

of a function x(t) by
1
cpx)T)= lim lT
T-H~)
JT x(t)x(t+T)dt,
-T
(6-139)
and the cross correlation of two functions, x(t) and y(t) by
1
cpxy(T)= lim lT JT x(t)y(t+T)dt. (6-140)
T-'>oo -T
If the functions x(t) and y(t) are periodic, then the limit process can be omitted
provided Tis chosen as the period. In general, however, x(t) and y(t) will not be
periodic. In fact we will later consider them to be random variables in the sense that
only their probability distribution can be specified.
Now we will set up the cross correlation between reactivity and power (using a
notation in which the limit process T-HX! is to be understood)
(6-141)
If we substitute in our expression for p(t)/ P 0 from Eq. (6-107) (after extending the
time domain to - oo < t' < t), we find
cppP ( 7) = }T iT/PextC t- 7) [ lo 00
du8pext(t- u)l ( u)] dt

= i 00
1
l(u)[ 2 T f_T/Pext(t-7)8pext(t-u)dt ]du
(6-142)
where we have identified the autocorrelation function of reactivity
(6-143)
As a final step, we take the Fourier transform of Eq. (6-142), defined as
(6-144)
to find
(6-145)
but g {/ (u)} is just the closed loop transfer function L(iw). Hence we find that we
can write
(6-146)
Here cppP = g {cppP} is referred to as the cross-spectral density, while if;PP- g {cppp} is
referred to as the reactivity (or input) spectral density.
To apply this result in a practical measurement, one varies the reactivity of the
reactor in a random or pseudorandom fashion and then measures the correspond-
ing variations in the reactor power or flux. In this sense, the measurement is very
similar to that involved in a rod oscillator experiment. To construct the time-
correlation function, one measures both 8pexlt) and p(t) at the time t and at
various delayed times 1-1 + ~7, t + 2~7, .... Using these measurements, one can
then construct both the time-correlation functions cp(t) and their Fourier trans-
forms numerically. For example, if measurements are taken at delay intervals of
~T, then the cross-spectral density is given by
\if { cppP( T)} ~ ~ cppp(n~T) exp(iwn~T)~T. (6-147)

n
Then by taking the ratio of \if { cppP} and \if { cpPP}, one can obtain both the gain and
phase of the closed-loop transfer function L(iw).This experiment becomes particu-
larly simple if one can use a so-called "white" noise source as the reactivity input.
Such sources are characterized by a spectral density that is constant in frequency,
that is, \if { cp PP} =constant= A. Then
\if{ cppP} =AL(iw). (6-148)
Hence by merely measuring the cross correlation, one can determine both the
amplitude and phase of the transfer function.
This experiment serves, then, as an alternative to a reactor oscillator transfer-
function measurement. Unlike the latter, it does not suffer from background noise
(rather taking advantage of such random fluctuations), and hence does not require
nearly so large an input signal. However both of these experiments suffer from the
fact that one must perturb the reactor by introducing an externally controlled
reactivity signal in order to perform the measurement. It is possible to bypass this
difficulty and measure the amplitude of L(iw) directly from the inherent noise
naturally present in the reactor.
2. AUTOCORRELATION MEASUREMENTS
Consider the a\uocorrelation of the fluctuations in the reactor power
cppp ()
__ l
T - 2T
f T
p(t)p(t+T)
Po Po dt. (6-149)
-T
If we substitute in Eq. (6-107) for p(t)/ P 0 and again take a Fourier transform in
time, we find
\if{ cppp} = L( - iw)L(iw) \if{ cpPP}, (6-150)
or
. )12= \'f{cppp}
IL( lW W:-{ ) . (6-151)
;J- cpppJ
Hence the square of the magnitude (i.e., the gain) of the transfer function can be
determined as the ratio of the power and reactivity spectral densities.
Although we can easily measure the power spectral density, the reactivity
spectral density characterizing inherent reactivity noise is not experimentally ac-
cessible. If one has reason to believe that such fluctuations are truly random in
nature, then \if { cp PP}= constant, and the measurement of the power autocorrelation
function by itself will yield the amplitude of the transfer function (although phase
information will have been lost).
Such a measurement of IL(iw)I is extremely simple. It has the advantage that it

does not perturb the reactor. Indeed it can be used in a real time mode to obtain ari
instantaneous measurement of the reactor transfer function. Its only significant
disadvantage is its limited accuracy.
Such noise measurements have been used to measure a large variety of nuclear
reactor parameters, and have proven to constitute a valuable diagnostic technique
in determining reactor kinetic behavior.
V. SPATIAL EFFECTS IN REACTOR KINETICS
Thus far our analysis of time-dependent nuclear reactor behavior has been
based on the point reactor kinetics model in which the neutron flux was assumed to
be the product of a time-independent spatial shape factor 1h(r) and a time-
dependent amplitude factor P (t) [cf. Eq. (6-7)]. However this assumption will
obviously be invalid in many cases of interest. For example, one of the most
important safety questions concerning reactor analysis involves the reactor kinetic
behavior following the postulated ejection of the control rod with highest reactivity
worth. Such a strongly localized perturbation in the core composition would
certainly cause a considerable deviation from the spatial shape factor 1h(r) and
would invalidate the point reactor kinetics model.
Furthermore most modern power reactor cores are quite large from a neutronic
point of view, being as much as 200 diffusion lengths in diameter. Hence the
neutronic behavior in such cores tends to be quite loosely coupled from point to
point. This means that a change in the flux or power density (or core multiplica-
tion) at one point in the core will not be felt at other points until after an
appreciable time delay. For example, one frequently finds rather significant "tilts"
in the spatial power distribution across the core due to nonuniform coolant
temperature or fission product buildup. The phenomenon of such loose spatial
coupling can be seen rather vividly by measuring the reactor transfer function, say,
under zero power conditions, at several points in a large power reactor core. That
is, one inserts a sinusoidal or pseudorandom reactivity variation at one point r in
.the core (say, by moving a control rod in and out of the core), and then measures
the amplitude and phase of the resulting power or flux oscillations by locating a
neutron detector at various other points r' in the core. If the reactor were truly
described by the point reactor kinetics equation, then the transfer function Z (iw;
r, r') measured at various points r would be the same. And for low frequencies, one
does indeed find that Z (iw; r, r') is essentially independent. However for
higher frequencies in large reactor cores, one finds that actual measurements will
yield different results, depending on where the oscillator and detector are placed.
Such measurements reveal that the transfer function is actually spatially dependent
for higher frequencies, because of the time it takes to propagate a disturbance in
the neutron flux from one point to another in the core. This implies a breakdown in
the point reactor kinetics model that characterizes every point in the reactor by the
same time-dependence P ( t).
As a general rule, one finds that the point reactor kinetics equations are
incapable of predicting the detailed behavior of reactor transients initiated by rapid
local changes in reactivity. More precisely, if the neutron flux changes rapidly on a
time scale of the order of the effective neutron lifetime (1), the point reactor
kinetics equation should be regarded as suspect.
In these instances, one must take explicit account of the spatial dependence of
the neutron flux. In fact one is frequently forced to perform a brute force
numerical solution of the time-dependent neutron diffusion equation and precursor
equations. Unfortunately in most cases in which such spatial effects are significant,
one cannot rely on the one-speed approximation to provide an adequate descrip-
tion of the neutron energy-dependence. Hence a direct numerical study of nuclear
reactor kinetics 15 usually involves the solution of the multigroup, time-dependent
diffusion equations-at a considerable computational expense.
For such finite difference solutions of the multigroup diffusion equations to be
feasible, it is necessary to employ a variety of numerical techniques to accelerate
the computation. These calculations are usually coupled with transient thermal-
hydraulic calculations and can become quite involved. In general they are only
used as a last resort to describe situations in which detailed spatial behavior is of
paramount importance in a rapid transient (such as the rod-ejection accident
analysis).
A variety of less direct methods exist. These can generally be classified as nodal,
quasistatic, or modal techniques. The nodal approach 16 is very similar to that
discussed in Section 5-11 in that the core is divided into a number of regions or
nodes. As in he static case, the primary difficulty in this approach is determining
1
the paramet~ that couple the flux at various nodal points. This is usually
accomplished by a variety of approximate schemes adjusted by numerous empirical
fits to either experimental or benchmark calculation data.
So-called quasistatic methods 17 essentially work within the framework of the
point reactor model, performing periodic static spatial calculations to obtain the
shape functions necessary for evaluating the point kinetics parameters p, /3, and A.
The practical utilization of this method has been rather limited to date.
Perhaps the most popular alternative 18 to direct finite difference schemes in-
volves expanding the flux in a finite series of known spatial functions or modes,
f/r), and then obtaining a set of equations for the time-dependent expansion
coefficients cp/t):
N
cj>(r, t) = L cpi t)fir). (6-152)
j=I
Of course the ideal scheme would be to choose f/r) as the spatial eigenfunctions of
the perturbed reactor. Unfortunately one usually does not know these modes.
Instead it is common to expand in an alternative finite set of spatial functions and
then use either weighted residual or variational methods (recall Section 5-III-C) to
determine the appropriate set of modal equations for the 'Pp Such schemes are
referred to as synthesis methods and will be discussed in more detail in Chapter 13.
Although such spatially dependent kinetics problems are of importance in the
very rapid transients arising in postulated accident analysis, they also arise in the
study of much longer term fuel depletion and fission product buildup phenomena.
In these latter cases, however, one can treat the neutron flux in a quasi steady-state
manner by performing a sequence of criticality calculations as the core composi-
tion changes with time. We will devote considerable attention to the analysis of
such fuel depletion or burnup problems in Chapter 15.
REFERENCES
l. A. Z. Akcasu, G. S. Lellouche, and L. M. Shotkin, Mathematical Methods in Nuclear

Reactor Dynamics, Academic, New York (1971).
2. R. Scalletar, in Proceedings of Conference on Neutron Dynamics and Control, D. L.
Hetrick and L. E. Weaver Eds., USAEC CONF-650413 (1966).
3. W. M. Stacey, Jr., Space-Time Nuclear Reactor Kinetics, Academic, New York (1969).
4. G. R. Keepin, Physics of Nuclear Kinetics, Addison-Wesley, Reading, Mass. (1965).
5. A. Z. Akcasu, G. S. Lellouche, and L. M. Shotkin, Mathematical Methods in Nuclear
Reactor Dynamics, Academic, New York (1971), p. 91.
6. R. Goldstein and L. M. Shotkin, Nucl. Sci. Eng. 38, 94 (1969).
7. P. H. Hammond, Feedback Theory and its Applications, English Universities Press,
London (1958); J. L. Bower and P. M. Schultheiss, Introduction to the Design of
Servomechanisms, Wiley, New York (1958).
8. D. L. Hetrick, Dynamics of Nuclear Reactors, University of Chicago Press (1971).
9. L. E. Weaver, Reactor Dynamics and Control, Elsevier, New York (1968).
10. F. W. Thalgott, et al., Proceedings of Second U. N. Conf. on Peaceful Uses of Atomic
Energy, Vol. 12, 242 (1958).
11. J. A. Deshong, Jr., and W. C. Lipinski, USAEC Document ANL-5850 (1958).
12. A. Z. Akcasu, G. S. Lellouche, and L. M. Shotkin, Mathematical Methods in Nuclear
Reactor Dynamics, Academic, New York (1971), Chapter 6.
13. J. A. Thie, Reactor Noise, Rowman and Littlefield, Totowa N.J. (1963).
(1970), pp. 511-514.
15. D. R. Ferguson and K. F. Hansen, Nucl. Sci. Eng. 51, 189 (1973).
16. W. M. Stacey, Jr., Space-Time Nuclear Kinetics, Academic, New York (1969), Section
1.4.
17. K. 0. Ott, Nucl. Sci. Eng. 26, 563 (1966).
18. J. B. Yasinsky and A. F. Henry, Nucl. Sci. Eng. 22, 171 (1965).
PROBLEMS
6-1 At time t=O, a decaying point source emitting s0 e->-1 neutrons per second is placed
at the center of a homogeneous bare spherical reactor which is being maintained in a
subcritical state. Determine the time-dependence of the neutron flux that would be
measured by a detector placed outside of the reactor. Use one-speed diffusion theory
and ignore delayed neutrons.
6-2 Consider a spherical assembly operating at a critical steady-state power level at time
t = 0. At this instance a neutron burst is suddenly inserted at the center of the reactor.
Derive an expression for the length of time before the reactor flux will once again
assume its fundamental mode shape to within 10%. Plot this time versus assembly
radius R for a thermal assembly moderated with H 20. Ignore delayed neutrons.
6-3 (a) What is the maximum possible reactivity insertion capable in a 235 U fueled
reactor? (Express your answer in $.)
(b) What is the maximum possible reactivity of a 235 U-fueled reactor having a
nonleakage probability of 0.6 for fission neutrons and 0.7 for delayed neutrons?
6-4 Estimate the prompt neutron lifetime in a large thermal reactor fueled with 235 U and
moderated with: (a) H 20, (b) graphite, and (c) D 20.
6-5 Calculate the effective neutron lifetime (1) for a thermal reactor fueled with either
233
U or 235 U and a fast reactor fueled with 239 Pu.
6-6 Give a physical discussion of the difference between the prompt neutron lifetime and
mean neutron generation time A for both an infinite and finite reactor. (Refer to
Keepin, 4 p. 166, for a more thorough discussion.)
6-7 In our development of an equation describing the delayed neutron precursor con-
centration, we assumed that these precursors did not migrate from the point of fission.
This is a reasonable assumption in solid-fueled reactors. Consider, however, a reactor
core in which the fuel is in gaseous form. Then the precursors may diffuse from the
point of fission before decaying. If the diffusion coefficient characterizing a given
precursor migration is Dp develop a generalization of the precursor concentration
equations to account for this process.
6-8 Derive an expression for the effective delayed neutron yield fractions '/I; characterizing
a mixture of several fissile isotopes.
6-9 Rederive the inhour equation for the case of a fast reactor fueled with 239 Pu and 238 U
treating delayed neutrons from both isotopes explicitly.
6-10 Repeat the solution of the point reactor kinetics equations with one effective delayed
group for a constant reactivity insertion Po using Laplace transform methods (see
Appendix G).
6-11 Demonstrate that for small reactivity insertions and ( {3- Po+ >.A)2»4A>.p0 the reactor
power P(t) is given by the approximate form [Eq. (6-40)].
6-12 Estimate the reactor period induced by a positive reactivity insertion of I $ in an
infinite 235 U-fueled thermal reactor moderated with water and a fast 239 Pu-fueled
reactor.
6-13 Provt({hat for six groups of delayed neutrons, the inhour equation has seven roots s;
of which six have negative values.
6-14 Consider a two-stage delayed neutron production process as sketched below:
Fission product A ! B! C + delayed neutron.
Assuming that B is the only delayed neutron precursor of interest, write the point
reactor kinetics equations including both rate equations for A and B. By then deriving
the corresponding inhour equation, demonstrate that such a two-stage decay process
can actually yield a negative effective delayed neutron fraction.
6-15 Initially a reactor is operating at a steady-state power level P0 • Using the point reactor
kinetics equation with one equivalent group of delayed neutrons, determine the stable
reactor period T for a positive step reactivity insertion of I $. Use the fact that
>.A//3« 1 to simplify your answer. Calculate a numerical value for T in the case in
which >-=0.l sec- 1 and A=0.OOlsec- 1•
6-16 According to the point kinetics equations with one equivalent group of delayed
neutrons, how long should a steady source of S 0 neutrons per second be left on in
order to raise the steady state reactor power level from P 1 to P 2? Assume >.A//3«1.
Roughly plot P (t) against t.
6-17 Using the point kinetics equations with one equivalent group of delayed neutrons,
derive an expression for p(t) such that P ( t) = P O + at.
6-18 Use the one-speed, one equivalent delayed neutron group form of the point reactor
kinetics equation to analyze the following situation: A reactor operator wishes to
reduce the power level of a reactor from P 1 to P2 • He therefore takes the reactor
subcritical to a reactivity p0 , where Po is a negative number, and after a time T, he
restores the reactivity to zero, whereupon the reactor levels out to P2 • In terms of P 1,
P 2 , p0 , and the other constants of the system, how long must he choose the time T to
be? (Assume >.A//3« I for convenience.)
6-19 Repeat the derivation of the integrodifferential form [Eq. (6-51)] of the point reactor
kinetics equation for the case in which P(t) and C/t) are specified at an initial time,
chosen at t = 0.
'
6-20 Demonstrate that for a periodic power variation P(t)=P0 +P 1 sinwt, one requires a
periodic reactivity insertion of the form given by Eq. (6-54). Demonstrate that this
reactivity insertion has a negative bias in the sense that its average over a period is
negative.
6-21 Compute the time-dependence of the reactivity insertion for a positive power excur-
sion of very short duration compared to delayed neutron lifetimes. Express this result
in terms of the total energy released in the excursion.
6-22 By considering the point reactor kinetics equation with one effective delayed group,
determine the time-dependence of the reactor power for a ramp reactivity insertion of
p(t) = - yf3t. Compare this with the result predicted by the prompt jump approxima-
tion, Eq. (6-68).
6-23 Demonstrate that following a reactivity change from p 1 to p2, the power levels before
and after the reactivity step are given by P 2 / P 1 = ( /3- p 1) / /3- p2) in the prompt jump
approximation.
6-24 Provide the details omitted in the derivation of the alternative form of the point-
reactor kinetics equation given by Eq. (6-72).
6-25 In reactor kinetics problems in which reactivity is inserted at a very rapid rate,
important changes occur in the neutron density while the delayed neutron emission
changes only slightly. Assuming under this approximation that for short times after
reactivity insertion Eq. (6-59) is applicable, determine the reactor power level P(t) if
the time-dependence of reactivity is p( t) = p0 exp (t / T ).
6-26 Explicitly perform the Laplace transform inversion of the zero power transfer func-
tion Z (s) to obtain the impulse response function '.£: ( t).
6-27 Prove that a necessary and sufficient condition for stability of a linear system is that
fo 00
I'.£: (t)I dt < oo. [Hint: To prove sufficiency, show that this condition implies that
the reactor power will be bounded for any bounded reactivity input. To prove
necessity, consider the specific reactivity input p( - t) = '.Z(t)/1 '.Z (t)I and show that for
this bounded input, the output p(t) is unbounded if the above condition does not
hold.]
6-28 Calculate the amplitude and phase angle for the zero-power transfer function and
plot these as functions of frequency. Assume one group of delayed neutrons with
iJ=.0065, ;\=0.08 sec- 1, and A= 10- 4 sec.
6-29 Write and run a simple computer program that solves the point reactor kinetics
equations numerically. Assume only one effective delayed group of neutrons, and use
either a Runge-Kutta or predictor-corrector scheme to integrate the equations. In
particular, study the time-step magnitude necessary to handle step reactivity insertions
of p0 =0.2 $, 1 $, and 5 $.
6-30 Set up the point reactor kinetics equations, including temperature feedback but no
delayed neutrons. Assume that the power removal rate is proportional to the devia-
tions of the temperature fodl some steady-state operating condition. Linearize the
equations, solve them, and discuss the solutions. Assume step reactivity input at t =0.
6-31 A reactor operating at steady state with constant heat removal rate experiences a step
insertion of positive reactivity of 3 $.
(a) Assuming a prompt negative temperature coefficient of reactivity, a heat removal
rate during the excursion that remains constant at the steady-state value, and a
constant delayed neutron contribution, give an expression for the maximum
temperature reached during the excursion. (Make any necessary assumptions and
state what they are.)
(b) Approximately what is the power level when this maximum temperature is
reached?
6-32 In a prompt critical reactor excursion a large amount of reactivity (measured above
prompt critical) Po is instantaneously inserted in an equilibrium reactor at t = 0.
Assume that: (a) the effect of delayed neutrons is negligible on the time scales under
consideration and (b) the reactor shuts itself down by thermal expansion of the core
in such a way that negative reactivity is "added" proportional to the total heat energy
generated up to time t, that is, p = Po - y f ~dt' P ( t'). Find the power level P ( t) where t
is measured from the time of reactivity insertion and in units of the prompt neutron
lifetime. [This is known as the Fuchs-Hansen model of a reactor excursion.]
6-33 Determine the total energy generated in the excursion modeled in Problem 6-32 in
terms of the reactivity insertion, delayed neutron fraction, and reactivity feedback
coefficient y. Also determine the peak power generated in the excursion. Discuss the
implications of these results for reactor safety.
6-34 A homogeneous fuel-moderator mixture is assembled as an unreflected critical reac-
tor. No cooling is provided. Suddenly a control rod is removed, bringing k to l.02.
Estimate the maximum temperature that results. The thermal conductivity is high
enough that the core can be treated isothermally. As data, use: ar= -2.5X 10- 4 /°C,
/3=0.0075, and A= 10- 5 sec.
6-35 An interesting alternative to the pulsed neutron experiment is a modification of the
diffusion length experiment in which an oscillating source of thermal neutrons is
placed against one end of a long column of the material to be studied, and then the
oscillating component of the resulting neutron flux in the column is measured at
various positions. Determine expressions for the attenuation and phase shift (relative
to the source) of the flux as functions of frequency w for a nonmultiplying column as
described by one-speed diffusion theory. [This is known as the neutron wave experi-
ment.]
6-36 perive the expression obtained for the system gain in terms of the autocorrelation
functions in Eq. (6-151).
3
The Multigroup
Diffusion Method
-------
..
7
Multigroup Diffusion Theory
Thus far we have based our study of nuclear reactor theory on a particularly
simple model of neutron transport, one-speed diffusion theory. This model
certainly suffices to introduce most of the important concepts of reactor analysis as
well as many of the computational methods used in modern reactor design. It can
even be used on occasion to provide useful qualitative infor.mation such as in
preliminary survey design studies. However for most of the problems encountered
in practical nuclear reactor design the one-speed diffusion model is simply not
adequate.
Two very significant assumptions were made in deriving the one-speed diffusion
model. We first assumed that the angular flux was only weakly dependent on angle
(linearly anisotropic, in fact) so that the diffusion approximation was valid. Usually
this assumption is reasonably well satisfied in large power reactors provided we
take care to modify the analysis a bit in the vicinity of strong absorbers, interfaces,
and boundaries to account for transport effects.
The principal deficiency of the model is the assumption that all of the neutrons
can be characterized by only a single speed or energy. As we have seen, the
neutrons in a reactor have energies spanning the range from lOMeV down to less
than 0.01 eV-some nine orders of magnitude. Furthermore, we have noted that
neutron-nuclear cross sections depend rather sensitively on the incident neutron
energy. Hence it is not surprising that practical reactor calculaE·s will require a
more realistic treatment of the neutron energy dependence. (In eed. 'it is surprising
that the one-speed diffusion equation works at all. Its success epends on a very
• judicious choice of the one-speed cross sections that appear in the equation.)
We will now allow the neutron flux to depend on energy, but rather than treat
the neutron energy variable E as a continuous variable, we will immediately
285
286 / THE MULTIGROUP DIFFUSION METHOD
discretize it into energy intervals or groups. That is, we will break the neutron
energy range into G energy groups, as shown schematically below:
Group g
~
E
~ - - + - -.....~·~--...,l1---~l----1H,-....+----+-----1
Ea-1
Notice that we are using a backward indexing scheme, corresponding physically to

the fact that the neutron usually loses energy during its lifetime (and mathemati-
cally to the fact that one always solves the discretized equations starting at high
energies and working successively to lower energies).
As in our earlier discrete ordinates approach, it would be possible to discretize
cf,(r, E, t) by considering it only to be defined at each energy mesh point £ 8 •
However it is more convenient to define the discretized fluxes instead to be the
integrals of cf,(r, E, t) over the energies of each group, such that the multigroup fluxes
cf,g(r, t), represent the total flux of all neutrons with energies E in the group
E8 < E < E8 _ 1. Then our task is to determine equations for cf,8 (r, t). We will find
that these equations take the form of a set of diffusion equations describing the
neutrons in each energy group. The equations are coupled to one another since
neutrons may experience changes in energy and hence pass from group to group.
For example fission neutrons will usually be born in the highest energy groups and
then cascade downward in energy from group to group as they are moderated by
scattering collisions.
Recalling the rather detailed dependence of neutron cross sections on neutron
energy E, one might expect that a great many such energy groups would be
necessary to adequately describe a nuclear reactor. Surprisingly enough, however,
most nuclear reactor calculations achieve sufficient accuracy using only few-group
diffusion descriptions. The ability to describe a reactor adequately with a relatively
small number of energy groups is not simply fortuitious, but rather is a con-
sequence of a careful choice of the energy-averaged cross sections that characterize
the neutrons in each group. Since the calculation of such group-averaged cross
sections or multigroup constants is the key to the successful implementation of
multigroup diffusion theory, we will consider it in some detail in Chapters 8 and 9.
In this chapter, we will first give a heuristic derivation of the multigroup
diffusion equations based on the concept of neutron balance. We will then rederive
these equations in a more rigorous (and useful) fashion and illustrate their applica-
tion using several important examples. Finally we will discuss the common calcula-
tional strategies useful for solving this set of coupled partial differential equations.
I. A HEURISTIC DERIVATION OF THE MULTIGROUP

DIFFUSION EQUATIONS
Perhaps the most straightforward manner in which to arrive at the form of

the multigroup diffusion equations is to apply the concept of neutron balance to a
given energy group by balancing the ways in which neutrons can enter or leave this
MULTIGROUP DIFFUSION THEORY / 287
group. Consider then a typical energy group g:
Group g E ")r
y---..... ~
After a bit of reflection, it should be apparent that such a balance would read as
follows:
Time rate of source

change of change duel [ abso_rption l neutrons
neutrons in
=- ~ - m + appearing
[ leakage
l l
group g group g in group g
neutrons neutrons
_ scattering + sca_ttering .
(7-1)
out of
[ group mto
g group g
It should be noted that we have taken explicit account of the fact that a scattering
collision can change the neutron energy and hence either remove it from the group
g, or if it is initially in another group g', scatter it to an energy in the group g. We
will characterize the probability for scattering a neutron from a group g' to the
group g by something akin to the differential scattering cross section "'i:..(Eg'-Eg) (a
so-called group-transfer cross section), "'i:.sg'g• Note that the' cross section characteriz-
ing the probability that a neutron will scatter out of the group g is then given by
G
Lsg = ~ Lsgg'· (7-2)
g'= I
We will similarly define an absorption cross section characterizing the group g, "'i:.ag•
and a source term Sg gjving the rate at which source neutrons appear in group g.
Finally we will define a diffusion coefficient Dg so that the leakage from group g
can be written within the diffusion approximation as V · DgVcf>g• If we combine all of
these terms, we find a mathematical representation of the balance relations [Eq.
(7-1)]
If we separate out that component of the source due to fissions, then we can write
G
sg = Xg ~ vg,"'i:.tg•c/>g• + st\ (7-4)
g'= I
where Xg is the probability that a fission neutron will be born with an energy in
group g, while Lrg' is the fission cross section characterizing a group g' and vg, is the
average number of fission neutrons released in a fission reaction iduced by a
neutron in group g'.
Hence we now have a set of G coupled diffusion equations for the G unknown
group fluxes cf>g(r, t). It shouldn't take much imagination to see that many of the
same techniques that we used for the one-speed (or in our present terminology,
one-group) model will also hold for the G-group system. In fact, we will see later in
this chapter that the specific structure of the multigroup equations makes their
solution extremely simple in most cases-provided one can handle the diffusion
equation characterizing each individual group.
The more serious problem concerns just how one determines the group constants
that appear in these equations:
(7-15)
We have only given a very vague definition of these constants in our heuristic
derivation of the multigroup equations. Hence before we can concern ourselves
with just how these equations are to be solved, we must go back and give a more
careful derivation of these equations in an effort to obtain a more explicit and
useful definition of the group constants.
II. DERIVATION OF THE MULTIGROUP EQUATIONS FROM

ENERGY-DEPENDENT DIFFUSION THEORY
Perhaps the most satisfying manner in which to derive the multigroup
diffusion equations characterizing the average behavior of neutrons in each energy
group is to integrate (i.e., average) the equation for the energy-dependent neutron
flux, cf,(r, E, t), over a given group, Eg < E < Eg- 1 • We will assume that this flux can
be adequately described by the energy-dependent diffusion equation:
+ x( E) fo dE' v( E')Lr( E')cf,(r, E', t)

00
+ Sext(r, E, t). (7-6)
Notice that we have inserted the explicit form for the fission source developed
earlier in Eq. (4-50).
We might mention that other energy-dependent equations could be used as a
starting point for the development of the multigroup diffusion equations. 1•2 For
example, we could have first developed the multigroup form of the transport
equation and then introduced the diffusion approximation for each group. We will
consider an alternative approach based upon the P I equations in Chapter 8.
However all of these approaches yield very similar forms for the multigroup
diffusion equations, with only some minor variations in the expressions for the
group-averaged cross sections.
We will begin by eliminating the energy variable in the energy-dependent
diffusion equation by integrating Eq. (7-6) over the gth energy group characterized
by energies Eg < E < Eg- I:
We will proceed further by making some formal definitions. First define the
neutron flux in group g as
Eg-1
cJ>/r,t)= f. dEcJ>(r,E,t). (7-8)

Eg
Next define the total cross section for group g as

} Eg-1
L 1g=-f. dELi{E)cJ>(r,E,t), (7-9)

c/>g Eg
the diffusion coefficient for group g as
(7-10)
and the neutron speed characterizing group g as
} }f
- =-
Eg-1 }
dE -cJ>(r,E,t). (7-11)
Vg c/>g E
g
V
The scattering term requires a bit more work. If we break up the integral over E' to
write
(7-12)
then it becomes evident that we want to define the group-transfer cross section as
(7-13)
A very similar procedure is followed for the fission term by writing

and then defining the fission cross section for group g' as
(7-15)
while defining
Eg-1
xg= J,Eg
dEx(E). (7-16)
If we now use these purely formal definitions to rewrite Eq. (7-7), we arrive directly
at the multigroup diffusion equations:
g = 1, 2, ... , G. , (7-17)
Several comments concerning these equations are necessary. The multigroup
diffusion equations (7-17) are still quite exact (within the diffusion approximation,
that is), but they are also quite formal in the sense that the group constants are as
yet undetermined. While it is true that our derivation has yielded explicit expres-
sions for these group constants, it is apparent that in order to calculate them we
would need to know the flux </>(r, E, t), and this is just the function we were trying to
calculate in the first place by discretizing the energy-dependent diffusion equation
(7-6). Hence it seems as if our development has been a bit circular.
In fact the multigroup "constants" as we have defined them still depend on space
and time. They will be rigorously constant only in the case in which the neutron
flux is of the separable form
</>(r, E, t) = 1/; (r, t )cp ( E ) , (7-18)

in which case they reduce to group averages over the neutron flux energy spectrum
cp(E). Unfortunately the flux in a nuclear reactor is usually not separable in energy,
and in general we will find that the group constants do indeed depend on space
and time (although in a manner somewhat more subtle than the above discussion
might indicate).
However perhaps you have already guessed the game we will play now, for we
will, in fact, attempt to guess or approximate the intragroup fluxes,
</>(r, E, t) ~ </>approxCr, E, t ), (7-19)
in our calculation of the group constants, e.g.,
(7-20)
as averages over these approximate intragroup fluxes.

In the next two chapters we will develop a number of schemes for approximating
the flux <[>(r, E, t) within a group-usually by first neglecting its spatial and time
dependence. The accuracy required of this flux estimate is primarily dictated by the
group structure itself. Of course for an extremely fine group structure, the cross
sections and hence the flux would tend to be smoothly varying over a given group,
and hence even a very crude approximation to the flux would be sufficient. For
example, in Chapter 8 we will demonstrate that in a thermal reactor the flux
behaves very roughly as<[>(£)~ 1/ E for energies between l eV and 105 eV. Hence
this functional form could be inserted into our definitions to calculate the group
constants characterizing groups in this energy range. However when we remember
the very complicated dependence of the cross sections on energy (in particular,
their resonance structure), it becomes apparent that the groups would have to be
very finely divided indeed for such a crude approximation to the intragroup fluxes
to yield meaningful results.
In actual practice, one usually works with from two to 20 groups in reactor
calculations. Such few group calculations can only be effective with reasonably
accurate estimates of the group constants (and hence the intragroup fluxes). The
most common approach is to actually perform two multigroup calculations. In the
first of these calculations, the spatial and time dependence is ignored (or very
crudely approximated), and a very finely structured multigroup calculation is
performed to calculate the intragroup fluxes (usually relying on various models of
neutron slowing down and thermalization). The group constants for this fine
spectrum calculation are frequently taken to be just the tabulated cross section data
averaged (with, for example, a l / E weighting) over each of the fine groups.
These intragroup fluxes are then used to calculate the group constants for a
coarse group calculation (including spatial dependence):
.-~
Simple geometry Multigroup Detailed geometry
fine spectrum constants - coarse group
calculation calculation calculation
t Iteration (perhaps) -
'
I
This scheme of first calculating a neutron flux spectrum and then collapsing the
cross section data over this spectrum to generate few group constants is the most
common method in use today. It should be noted that in a calculation of this type,
the spectrum calculation (and hence the few-group constants) will depend sensi-
tively on the particular reactor being analyzed and its operating conditions (e.g.,
fuel loading, isotopic composition, temperature, and coolant conditions). In fact
such group constants will have to be calculated in each region of the reactor in
which the core composition varies appreciably, for example, because of variations
in fuel enrichment or moderator density. Furthermore these group constants will
have to be recalculated whenever these regionwise core properties change, such as
during fuel burnup or reactivity adjustment via the movement of control rods.
Hence spectrum calculations and the generation of few-group constants must be
performed repeatedly in reactor design.
It is usually necessary to take some account of the spatial dependence of the flux
in the generation of group constants. This is frequently done by performing the
calculation of the· neutron spectrum for a typical cell of the reactor core lattice,
using a variety of approximate techniques to account for flux variation throughout
the cell. The flux spectrum cp(E) that one then uses to generate group constants
will be a spatial average of the true flux <j>(r, E) over the cell. Such cell calculations
are essential for an adequate description of thermal reactors in which the neutron
mfp is relatively short, and we will devote considerable attention to them in
Chapter 10.
The above discussion illustrates a very important feature of nuclear reactor
analysis: the separation of the treatment of energy- and space-dependence in
multigroup diffusion theory. That is, one first utilizes a rather crude description of
the flux spatial dependence (say, fon only a given cell of the reactor lattice) to
generate a detailed representation of the energy spectrum cp(E) suitable for the
generation of accurate multigroup constants. One then uses these group constants
in a few-group diffusion equation analysis of the reactor. In this latter calculation,
the energy dependence of the flux is treated rather coarsely (only a few energy
groups), while more emphasis is directed toward adequately describing the spatial
variation via the diffusion equation. We will find that this procedure of separating
the treatment of the various independent variables r, E, and t arises very frequently
in nuclear reactor analysis. If performed properly, it will allow a rather detailed
study of the reactor for only a modest amount of computational effort.
How many groups are necessary for a reactor calculation? This will depend on
the problem one is considering. For example, in very crude survey calculations of
thermal reactors, two groups (one to characterize fast neutrons and the other to
characterize thermal neutrons) may be sufficient. Most LWR calculations are
performed using a four-group diffusion model3 (three fast groups, one thermal
group) while gas-cooled reactor analysis typically uses seven to nine groups. 4 For
fine detail, one may have to go as high as 20 groups (this is particularly true in
fast-reactor calculations) 5, and in spectrum calculations, the number of groups can
range as high as 1000 (so-called microgroup structure), which is almost as detailed
as the tabulated cross section data itself. 6
Let us look now in a bit more detail at the structure of the multigroup diffusion
equation. In particular, consider the scattering term in the equation. Recall that in
our study of the kinematics of neutron scattering collisions in Chapter 2, we noted
that if the incident neutron energy E was substantially greater than the thermal
energy of the target nuclei (typically less than 0.1 eV), the neutron could never gain
energy in a scattering collision. Such "fast" neutrons will only slow down in a
scattering collision. Hence in these fast groups, we can set
~sg'g = 0, for g' > g. (7-21)
Since most few-group diffusion calculations utilize only one thermal group to
describe the neutrons with E < 1eV (assuming that neutrons cannot scatter up out
of the thermal group), we can generally simplify the scattering term to write
G g-1
L ~sg'g</>g' = L }:,sg'g</>8 , + }:,sgg</>g• (7-22)
g'= I g'= I
Here we have taken care to separate out the in-group scattering term }:. 888 which
characterizes the probability that a neutron can suffer a scattering collision and
lose sufficiently little energy that it will still remain within the group. It is
customary to transfer this term to the left-hand side of the multigroup equation
(7-17) and to define a removal cross section,
}:_Rg = }:_tg - }:_sgg, (7-23)
which characterizes the probability that a neutron will be removed from the group
g by a collision. Note that the removal cross section is sometimes defined such that
it does not contain absorption Lag· We will use the above definition in our
development, however. We will see later that the neglect of upscattering (that is, the
assumption that the neutron can never gain or scatter up in energy in a collision)
greatly simplifies the solution of the multigroup diffusion equations.
One frequently achieves an additional simplification of the multigroup equations
by choosing the group spacing such that neutrons will only scatter to the next
lowest group-that is, such that
G
~ Lsg'g</>g' = Lsg- J,g</>g-1 + Lsgg</>g• (7-24)
g'= I
In this case, one refers to the multigroup equations as being directly coupled. If we
recall from Section 2-11-D that a neutron of energy E cannot scatter to an energy
below a.E in a single elastic scattering collision, then it is apparent that to achieve
direct coupling we should choose our group spacing such that E8 _ 1/ E8 > 1/ a.. For
heavier moderators (e.g., 12C), this is easy to do. Unfortunately in hydrogeneous
moderators a.H = 0, and hence direct coupling cannot be strictly achieved. However,
if one chooses E8 _ 1/ E8 > 150, then the probability of the neutron "skipping" the
next lowest group in a scattering collision with hydrogen is less than 1%, and hence
direct coupling is effectively achieved. We might mention as well that one typically
chooses a group structure such that the ratio, E8 _ 1/ E8 , is kept constant from group
to group. The motivation for this choice will become apparent in Chapter 8.
We will most frequently be concerned with situations in which both the time-
dependence and the presence of an external source can be ignored (e.g., criticality
---E,_3 --.--E,_3
----,-- E,-2 _.....,.._Eg-2
I
-...,....-+- E, _, E g-1
E, Eg E,
I I
I I
'I'
I
*I
I I
E,+ 1 Eg + 1 EI+ 1
I
I
I
I
I I
E,+2 E,+2 E,+2
No Directly Non directly
upscattering coupled coupled
FIGURE7-1. Alternative types of multigroup coupling

calculations). In this case the multigroup equations can be written as
(7-25)
The structure of these equations can be seen more clearly in matrix form
-V·D1V+~R, 0 0
-L s,2 -V•D2V+~R2 0
-Ls13 -Ls,, -V•D3V+~R3
P1X1~r, P2X1~r2
P1X2Lr 1 P2X2~r2
(7-26)
where we have inserted the usual criticality eigenvalue k. Notice in particular that
the neglect of upscattering has led to a lower triangular form for the "diffusion"
matrix M. The fission matrix Eis full, however, since fission neutrons induced by
a neutron absorption in a lower group will appear distributed among the higher
energy groups.
In the case of directly coupled groups, M becomes a simple bidiagonal matrix of
the form -
(7-27)
By way of contrast, if one chooses to assign several groups to the thermal energy
range in which appreciable upscattering occurs, there will be a full submatrix
within M corresponding to ~sg'g for g' or gin the thermal range:
~=(~)
t Thermal groups
(7-28)
We will return in a later section to discuss a general strategy for solving such
systems of diffusion equations. Before doing so, however, it is useful to consider
several simple applications of the multigroup diffusion equations.
III. SIMPLE APPLICATIONS OF THE MULTIGROUP

DIFFUSION MODEL
A. One-Group Diffusion Theory
First suppose we set up the "one-group" diffusion equation by defining
E 0 = oo and E 1 =0. Then if we note that
(7-29)
- - - - - £0 = oo
and 1 group {
- - - - - £1 =0
(7-30)
we find that the multigroup equations yield an old friend, the one-speed diffusion
equation
(7-31)
Of course it should be stressed that this equation is still of only formal significance
until we provide some prescription for calculating the group constants [that is, the
intragroup flux-which, in this case, is cp(r, E, t)]. Nevertheless it is comforting to
know that if we chose the group constants properly, even one-speed diffusion
theory could give an accurate description of nuclear reactor behavior.
B. Two-Group Diffusion Theory

A more enlightening application involves the case of two energy groups, one
chosen to characterize fast neutrons and a second for thermal neutrons. The cutoff
energy for the thermal group is chosen sufficiently high such that upscattering out
of the thermal group can be ignored. This corresponds to an energy of 0.5-1.0eV in
water moderated reactors, but may range as high as 3eV in high-temperature
gas-cooled reactors. If we choose the group structure as shown below:
Thermal group Fast group
Ea = 10 MeV
then we can identify:
c/> 1(r,t)= J.
E1
Eo
dEcp(r,E,t)=fast flux, (7-32)
E1 •
cpi(r,t)= J.
E2
dEcp(r,E,t)= thermal flux. (7-33)
We can simplify the group constants for this model somewhat. Consider first the
fission spectrum. Since essentially all fission neutrons are born in the fast group
(recall the fission spectrum x(E) illustrated in Figure 2-21), we can write
(7-34)
Hence the fission source will only appear in the fast group equation:
(fast), (7-35)
(thermal). (7-36)
We can proceed to calculate the scattering and removal cross section. First since
there is no slowing down out of the thermal group,
f. £ 1~leV
E2=0
dE"'i:..(E'-E) = "'i:..(E'), (7-37)
Hence we find
(7-38)
Thus the removal cross section for the thermal group is just
(7-39)
as we might have expected. The remainder of the group constants are defined as
before in the previous section. In practice they would be calculated by first
performing a fine spectrum calculation for the group of interest, and then averag-
ing the appropriate cross section data over this spectrum to obtain the group
constants. For example, a fast spectrum calculation would be performed to calculate
the fast group constants v 1, "'i:.f 1, "'i:.R 1, "'i:. 812 , and D 1 (as described in Chapter 8) while a
thermal spectrum calculation would be performed to calculate the thermal group
constants v2 , "'i:.r2 , D 2 , and "'i:.a2 (as described in Chapter 9).
We will consider the application of two-group diffusion theory to a reactor
criticality calculation. 7 Then we can set both the time derivatives and the external
source terms equal to zero to write the two-group diffusion equation as
1
- V ·D1Vc/>1 + LR,c/>1 = k [ V1Lr 1c/>1 + P2Lr2c/>2],
-v. D2Vcf>2 + La,lf>2 = Ls,,c/>1 · (7-40)
Notice that we have inserted a multiplication factor (1/ k) in front of the fission
source term since we are eventually going to be performing a criticality search.
Also notice that while the source terms in the fast group correspond to fission
neutrons, the source term in the thermal group is due only to slowing down from
the fast group.
As a specific illustration, we will apply the two-group diffusion equations, Eq.
(7-40), to analyze the criticality of a bare, uniform reactor assuming that both fast
and thermal fluxes can be characterized by the same spatial shape i/;(r):
(7-41)
We have omitted the subscript g from the geometric buckling B;= B 2

so as not to
confuse it with the group index g. Then if we substitute
(7-42)
into Eq. (7-40), we find the algebraic equations
( D 1B 2+ LR, - k- 1v 1Lr)c/> 1 - k- 1v2Lr2c/> 2= 0,
- Ls12c/>1 + ( D2B 2+ La,)cf>2 = 0. (7-43)
However this algebraic system has a solution if and only if
(7-44)
We can now solve for the value of the multiplication factor k, which will yield a
nontrivial solution of the two-group equations:
(7-45)
It is of interest for us to see if we can relate this expression to our earlier

expressions for k-notably the six-factor formula. First notice that the first term in
Eq. (7-45) represents neutron multiplication due to fissions occurring in the fast
group, whereas the second term represents multiplication due to thermal fission.
Since we expect the thermal fission contribution to be dominant in those situations
in which such a two-group analysis makes sense, let us first examine
L 51 /LR, vi(Lr/La,)
=-------- (7-46)
(1 + Lf B 2 ) (1 + L1B 2 )
From our earlier discussion in Section 5-111-D, it is evident that
(7-47)
are just the fast and thermal nonleakage probabilities. Notice that the diffusion
length L 1 characterizing the fast group is defined somewhat differently as
(7-48)
but this is consistent with our earlier definition of the diffusion length, since both
La and L512 act to remove neutrons from the fast group. The only unidentified term
is the ratio L 5 12 /LR I . However for a homogeneous reactor we know that this ratio is
just:
Rate at which neutrons slow

down to thermal group
(7-49)
Rate at which neutrons are
removed from fast group
which we can identify as the resonance escape probability p characterizing slowing

down from group 1 to group 2. Hence
(7-50)
In a very similar manner we can identify the fast multiplication factor as
(7-51)
where 1) 1 = v 1LiI /LaFl and we have defined a "fast utilization factor" J1 = L!'I /LR I in
analogy to the thermal utilization f2 •
To complete our identification with the usual six-factor formula, we evidently
must identify the fast fission factor t: as just
(7-52)
Then we find
(7-53)
which is the usual six-factor formula. Of course, this definition of £ is somewhat

awkward, since it depends upon quantities such as the thermal utilization, the
thermal nonleakage probability, and the resonance escape probability.
The identification of the various components of the six-factor formula in terms
of two group-constants is not unique, and alternative schemes can be found in the
literature. 1• 7 Such arbitrariness is indicative of the limited usefulness of the six-
factor formula for more realistic reactor analysis. In fact the two-group expression
fork [Eq. (7-45)] is far more appropriate for the description of most thermal power
reactor types which are characterized by a somewhat harder neutron energy
spectrum than the natural uranium, graphite-moderated reactors which motivated
the development of the six-factor formula.
The two-group diffusion model can be used to demonstrate a number of the
various applications of the multigroup formalism. For example, one frequently
wishes to generate the group constants for a few-group calculation using the
neutron spectrum generated by a many-group calculation. Such a procedure is
known as group collapsing, since it expresses few-group constants in terms of
many-group constants.
To illustrate this, we can derive expressions for the one-group constants in terms
of two-group constants. For example,
~R/P1 + La,'P2 - ~s 12</>1

=-------- (7-54)
</>1 + </>2
or using Eq. (7-43) to eliminate <f, 2 in terms of </> 1:
(7-55)
The remaining one-group constants can be given as
D= D1</>1+D2</>2 = (D2B2+"};a)D1+L.,2D2
(7-56)
</>1 +</>2 D2B2+"};a2 +Ls,2 '
P1Lr,</>1 + P2Lr2</>2 ( D2B 2 + "};aJP1Lr, + P2Ls,2"};f2

PLr=------=----------- (7-57)
</>1 + </>2 D2B 2+ La2 + Ls,2
(These relations can be generalized to the case of collapsing from G groups, but we
will leave this development to the problem set at the end of the chapter.)
C. Modified One-Group Diffusion Theory

In the analysis of large thermal reactors it is sometimes possible to simplify
the two-group diffusion equations even further by ignoring thermal leakage. If we
note that in a large LWR, D 2 ~0.5 cm, B 2 ~ 10- 4 cm- 2, while ~a2 :=::0.l cm- 1, then
we find
(7-58)
Hence the neglect of - D 2 v' 2</> 2 in the second of Eqs. (7-40) may occasionally be
valid for the investigation of problems in which thermal neutron diffusion does not
have a significant reactivity effect.
If we now solve the simplified equation for the thermal group flux
(7-59)
and substitute this into the fast-group equation, we find the modified one-group
diffusion model
(7-60)
or rearranging for the case of a homogeneous reactor
(7-61)
where
(7-62)
where we recall Lf = D 1/~R,· This model is sometimes referred to as one- and

one-half-group diffusion theory, since it is midway between the one- and two-group
schemes.
A very useful (and common) modification of this scheme which takes some
account of both fast and thermal leakage within a one-group treatment is obtained
by replacing Lf in Eq. (7-61) by the so-called migration area M2 defined as
(7-63)
We will show in Chapter 9 that M 2 is essentially just 1/6 the mean square distance
traveled by a neutron from its birth in fission to its eventual demise via thermal
absorption.
We have developed these very simple few-group models to demonstrate how
multigroup techniques can be used to evaluate many of the quantities of interest in

reactor analysis. (Still further examples are included as exercises at the end of this
chapter.) We could continue our discussion by considering more elaborate multi-
group diffusion models, such as the four-group model customarily used in LWR
analysis, or the ultrafine group structure required in fast reactor studies. However
we will defer such topics until we have studied schemes for the generation of
multigroup constants.
Thus far our illustrations of multigroup diffusion calculations have been ex-
tremely simple. They usually ignored spatial dependence (or at best, assumed only
a fundamental mode dependence). In any realistic calculation, one must take into
account the inhomogeneous nature of the reactor core by actually solving the
multigroup diffusion equations in detail. Hence we now turn our attention to the
strategy for solving these more general equations.
IV. NUMERICAL SOLUTION OF THE MULTIGROUP

DIFFUSION EQUATIONS
A. Successive Solution of the Multigroup Equations

We now consider a strategy for solving the multigroup diffusion equations on
digital computers. Suppose we begin by writing these equations (7-26) out in detail
as
1
-V·D1Vc/>1 +LR 1c/>1 = kx1s
1
-V 'D2Vc/>2 + LR,c/>2 = k x2S + Ls12c/>1
(7-64)
Notice that here we have assumed that there is no upscattering and also defined the
fission source as
G
S(r)= ~ vg,Lt,,c/>g,(r). (7-65)
g'=I
It is very important to note that the spatial dependence of the fission source is
identical in each group diffusion equation.
Now the essential scheme is just as before. We begin by guessing a fission source,
S (r) and a multiplication eigenvalue k:
(7-66)
Next, we calculate the flux in the first group:
(7-67)
Having obtained this flux, we can then proceed to the diffusion equation
characterizing the next lowest energy group
(7-68)
and solve this for </>i 1>(r) since the right-hand side is now known. We can continue
on in this fashion to determine all of the group fluxes:
(7-69)
Having done so, we can then calculate a new fission source

G
s<l)(r) = °"
.::., vg' L f,,</>O>(r)
g' , (7-70)
g'= I
and a new value of k
f d rs<l)(r)3
ko> = - - - - - - - (7-71)
_l_ f d3r s<o>(r) ·
k(O)
We can then proceed to perform each source iteration by solving down the
multigroup equations toward increasingly lower energies. This scheme of solving
successively the equations in the direction of lower energies is enabled by the
assumption that there is no upscattering. This implies that the flux in the higher
energy groups always determines the source term in the lower energy groups. In
effect, we are merely inverting a lower triangular matrix (as the matrix formulation
in the previous section made apparent).
If one chooses a multigroup structure in which more than one group is assigned
to the thermal energy range in which appreciable upscattering can occur, then
such a successive groupwise solution of the multigroup diffusion equations is no
longer possible. One must solve the equations characterizing the thermal group
simultaneously. If the number of such fully coupled groups in which both upscatter-
ing as well as downscattering occurs is small (e.g., in HTGR calculations,4 it is
common to use three to four thermal groups), a direct simultaneous solution (i.e.,
matrix inversion) can be accomplished. However if the number of thermal groups
is large, as it may be in thermal spectrum calculations, then iterative solution
schemes will be necessary (similar to the inner iterations used in multidimensional
diffusion calculations).
A great deal is known about the mathematical nature of such multigroup
diffusion eigenvalue problems. 8•9 Under rather weak restrictions on the group
fluxes and their boundary conditions, one can show that there will always exist a
maximum eigenvalue keff that is real and positive. The corresponding eigenfunction
is unique and nonnegative everywh~re within the reactor. These features are
reassuring, because we would anticipate that the largest eigenvalue will characterize
the multiplication of the system, and the corresponding eigenfunction will describe
the flux distribution within the core (which cannot be negative). One can also
demonstrate that the above source iteration will converge to this "positive
dominant" eigenvalue keff and the corresponding eigenfunction.
Such formal considerations are interesting in their own right, as well as being
useful in the investigation of algorithms devised for the solution of the multigroup
diffusion equations. In actual practice, however, one must also discretize the spatial
dependence in order to solve the group-diffusion equations. That is, one chooses a
spatial mesh and finite difference scheme, just as we did in Chapter 5, and then
discretizes the diffusion equations for each group. We now turn to a brief
discussion of several more practical aspects of the solution of such equations.
B. Strategies for Solving the Finite-Differenced Multigroup Diffusion

Equations
If we recall the general discussion of finite difference representations of the
neutron diffusion equations given in Section 5-11-B, it is apparent that the general
structure of the finite-differenced multigroup diffusion equations takes the form: 11
(Note here that we will occasionally write the group index g as a superscript in
order to avoid confusing it with the spatial mesh indices i and j.) Notice that in
addition to the coupling to different energy group fluxes at a given mesh point due
to the fission source and scattering, the finite difference equation is coupled as well
to the flux at adjacent spatial mesh points because of the effect of spatial diffusion.
If we denote the number of spatial mesh points by N and the number of groups
by G, then Eq. (7-72) represents a set of G X N simultaneous linear algebraic
eq"ll;ations. One usually normalizes the flux at one energy group-space mesh point
(since the overall normalization of the flux is arbitrary in a criticality calculation).
Hence we have G X N equations available to determine the G X N - I fluxes and
the multiplication eigenvalue kerr· As. before, it is convenient to rewrite this set of
equations as a matrix eigenvalue problem
1
Mcp=-Fcp.
=- k=-
(7-26)
EXAMPLE: Consider four energy groups with a five by four two-dimensional

spatial mesh. Then each matrix has (4 X 5 X 4)2 = 6400 elements and the flux vector
cf, has 80 elements. One typically allows fission neutrons to appear only in the
uppermost energy group, Xi= 1, Xg = 0, g > 1. Furthermore, we will assume directly
coupled groups such that ~sg'g = 0 if g' =fag- 1. We can explicitly exhibit the
structure of the matrix eigenvalue problem in this case as shown in Figure (7-2). 11
f
~~~
~~
~)~ I
JI~;,220
~~ X
1
~-~'- • - =T 1 •
~ ~
:~~~
~~
~.~
"'- '\.2 "'-2 2 '\.2 2 2 "'-2 '\.2 I

-~1---2)-(- Di,i-5 }{- Di,i-1 }-( ~2 Di,i-5 D;,;-1 D;,,+1 D,,i+s } { - Di,i+l \ { D;,,+5 ){ 2)
( ...
' ('1; ,-s>
2
('1; i-1>
2 ..I/
'
+ ('1; i-s> 2 + ('1; i-1>
2 + ('1; ,+1> 2 + ('1; ,+s> 2 ('1,· ,·+1>
2r\ - ('1,· ,+s> 2 cf>,
"'- , '\. , '\. , , "'- , , , "'- , "'- I
FIGURE 7-2. The matrix structure of a four-group, 5x4 spatial mesh point criticality calculation
Let us now review the general iterative strategy for solving this eigenvalue
problem.
(D One first makes an initial guess of the source vector s_<0) and the multiplication
eigenvalue k<0).
a) At this point one proceeds to solve the inhomogeneous matrix equation
M ,i,. (n+ o = _I_ s<n) (7-73)
=::r. k<n) -
for the next flux iterate, <t,<n+ I)_ This solution involves a number of substeps:
2-1 One solves the inhomogeneous diffusion equation characterizing each of
the energy groups g
(7-74)
by solving first for the highest energy group, g = 1 (noting R 0 -0, <f>bn)-0),
and then using <t>\n+ I) to solve for <t>~n+ 1>, and so on, solving successively
down the groups":- -
EXAMPLE: If we return for a moment to our earlier four-group, five-by-four

spatial mesh problem, we find that each matrix in the group inhomogeneous
equation [Eq. (7-74)) has (5 x4)2=400 elements while the flux and source vectors
have 20 elements. Note that the fission source ~n) vanishes for all but the highest
energy group since Xg = 0, g > I. Furthermore the highest energy group obviously
has no slowing down source. The matrix form of the group diffusion equation [Eq.
(7-74)) is shown in Figure 7-3 for this example. 11
2-2 Of course, solving even the inhomogeneous diffusion equation [Eq. (7-74))
for a single group is no trivial matter. For multidimensional problems,
iterative techniques will be necessary such as those discussed in Chapter 5
(e.g., SOR). Such inner iterations usually take the previous flux estimate
<t,<n-l)
_g as their first guess in solving Eq. (7-74). It should be mentioned that
a variety of schemes have been proposed (and utilized) for coupling such
inner iterations to the outer (source) iterations to accelerate convergence.
Q) Having obtained the flux estimate <t,<n+ I), one can now determine the next
multiplication eigenvalue estimate. A variety of weighting schemes can be used
to determine the ratio of fission source estimates. One such scheme can be
derived by taking the scalar product of the fission source vector E <t,<n+I) with
the equation defining the iterative scheme
M ,i,.(n+I)= _1_ F ,i,.(n) (7-75)

-Y k(n) = Y '
and then noting that presumably
M 1,s> ""'
2
+
11
D 7,8
(l>1,s>
2
+
11
D 7,12 }-( _
(Ll.7,12>
2
"" "" 11
D 7,8 )--(
(l>1,s>
2 Df 12 2)--(<J>~(n)
(Ll.1,12) I
)< ( I V "- g-l➔g V"-1<t,f(n))
O/ \ i.7 f I
""' ""
FIGURE 7-3. The matrix structure of the group-diffusion calculation
"" "" "" I if g >I
@ At this point one tests the source iteration for convergence, such as by
comparing
(7-78)
or a pointwise criterion
S (n+l)_ S (n)
& &
max (7-79)
S (n+ I)
g;
(or both). If the changes in k<n> or the elements of §_<0 > or p_<0 > are sufficiently
small, one assumes that convergence has been achieved, and the iterative
procedure is ended. If not, then a new fission source is calculated and the
iteration continues.
@ Usually the fission source §_(n) used in the next iteration is chosen via an
extrapolation scheme (cf. Section 5-IV-C) to accelerate convergence of the
source iterations, for example,
(7-80)
The positive nature of the multigroup diffusion operators implies similar proper-
ties for the matrices resulting from finite-differencing these equations. Hence many
of the formal conclusions concerning convergence to the positive dominant ei-
genvalue and eigenfunction can also be shown to hold for the finite-differenced
multigroup diffusion equations. The theory of such numerical procedures has been
placed on firm ground by Varga 10 and Wachpress, 8 and the interested reader
is referred to their treatises for more detail.
Thus far we have only concerned ourselves with the usual criticality calculation
for kerr- Actually a variety of different types of criticality searches are typically
conducted in an effect to achieve a desired core multiplication (which is frequently
not equal to 1.0). For example, two such schemes are: (a) core size searches
(usually the size of a given core region is varied, say by varying the position of one
boundary, until the desired value of kerr is achieved) and (b) composition search
(the atomic density of a specific fuel isotope, or several such isotopes, is varied,
perhaps only in a given subregion of the core).
The multigroup diffusion equations can also be used for a variety of other
applications. For example, one can use these equations to determine the neutron
flux maintained in a subcritical assembly by a steady-state source (by retaining the
source term in the multigroup diffusion equations). They can also be used to
calculate the time eigenvalues a occurring in solutions of the form eat by inserting
an effective absorption term (a/ vg)c/>g into the equations, and then solving for a as
a matrix eigenvalue problem. Such "a-searches" are of use in studying reactor
kinetic behavior.
V. MULTIGROUP PERTURBATION THEORY
Before we conclude this general discussion of the multigroup diffusion

equations, it is useful to redevelop the procedure of perturbation theory as it
applies to the multigroup equations as written in matrix form:
(7-81)
In order to determine the change in reactivity induced in the core by a small

perturbation, we must generalize somewhat our concept of inner products and
adjointness to account for the matrix nature of the multigroup eigenvalue problem
[Eq. (7-81)]. To this end, define the inner product between two G-dimensional
vectors [(r) and £(r) as
(7-82)
We can now use this inner product to construct the adjoint of the operators Mand
F:
(7-83)
Since the adjoint of a matrix is obtained by first taking the transpose of the matrix
and then complex-conjugating each of its elements, it is evident that
(7-84)
and similarly
' ')
... (7-85)
Note in particular that Mt=t=-M and E_t=t=-f-that is, the multigroup criticality
problem is not self-adjointHence we find </>T=/=-</>.
We can use our earlier expressions for the reactivity change corresponding to
perturbations in the core composition, if we recognize that such perturbations will
now have a matrix character-for example,
M'= M+ 8M, (7-86)
and
E'= E+ 8F. (7-87)

For example, one could imagine a perturbation in the absorption cross section
characterizing the second group as being represented by
8M= (7-88)
The corresponding reactivity change is then given by
tlp= (±t. [ §E- Mt]±) . (7-89)

(cJ>t.fcJ>)
To make these ideas more precise, let us consider the particularly simple example
of two-group diffusion theory as described by Eq. (7-40). In matrix form, these
equations become
The adjoint equations are
(-V·D~V+"'i:.R,
(7-91)
Mt
Note that Mt=/:- M and F =/:- E, hence c/> t =/:- cj>.

Suppose --;.e consider the re-;:ctivity changeinduced by perturbing the thermal
absorption cross section by an amount 8"'2:.a,· Then
8F=0'
~
(7-92)
Hence we can compute
(7-93)
to find
(7-94)
In analogy with our earlier one-speed calculation, suppose we set
(7-95)
Then
(7-96)
Thus we find
fractional change in
c6.p
<t>!(r0 ) = - ( ) = reactivity per neutron (7-97)
a<f>i ro absorbed per unit time.
In particular, if </>i is large at r0, then the change in p introduced by a thermal

absorber at r0 will be large. That is, <t>!(r) measures the "importance" of the point r
with respect to reactivity changes induced by perturbing the thermal absorption at
that point.
For the more general multigroup problem, <t>J(r) can be identified as the neutron
importance function for group g, since <t>J(r) is proportional to the gain or loss in
reactivity of a reactor due to the insertion or removal of one neutron per second in
the group g at point r.
In general, one finds that the multigroup neutron importance (or adjoint fluxes)
differs substantially from the multigroup fluxes. This is shown for a two-group
calculation for a reflected slab geometry core in Figure 7-4.
t------ Reflector
Core: U 235 + H 2 O
N25 - 1
NW - 300
Reflector: H 2 0
Two-group calculation
Infinite slab core
Infinite reflector
0.5 1------
I/J2
v-, '
\
\ '\ ' '-,
"
Critical ' - ,
'-......
'--..._..._
half-thicknes's'., ..._ ---- __
0 5
= 5.14 cm ,..._
0 .....__ _ _ _ _ _.....__ _ _ _ _ _...J...._ _ _
10
Distance from ct_, cm
--- - - - ------
----:::::-::::::-:~--
__:====-=-=-------J'----,;►
15 20
FIGURE 7-4. The two-group fluxes and adjoint fluxes for a reflected slab reactor
VI. SOME CONCLUDING REMARKS
We have now developed the multigroup diffusion equations and outlined a

strategy for solving them. The major uncertainty at this point concerns just how
one determines the group constants. These latter calculations rely on clever guesses
or approximations for the intragroup fluxes, and more specifically on one's ability
to determine the neutron energy spectrum characterizing fast and thermal
neutrons. This is the subject to which we will next turn our attention as we study
the slowing down and thermalization of neutrons.
REFERENCES
l. A. M. Weinberg and E. P. Wigner, The Physical Theory of Neutron Chain Reactors,

University of Chicago Press, (1958), p. 509.
(1970), p. 181.
3. R. L. Hellens, The Physics of PWR Reactors, in New Developments in Reactor Physics
and Shielding, USAEC CONF-720901 (1972).
4. M. H. Merrill, Nuclear Design Methods and Experimental Data in Use at Gulf General
Atomic, GULF-GA-Al2652 (GA-LTR-2), 1973.
5. R. Avery, Review of FBR Physics, in New Developments in Reactor Physics and
Shielding, USAEC CONF-720901 (1972).
6. B. J. Toppel, A. L. Rago, and D. M. O'Shea, MC2 , A Code to Calculate Multigroup
Cross Sections, ANL-7318, Argonne National Laboratory, 1967.
8. E. L. Wachspress, Iterative Solution of Elliptic Systems and Applications to the Neutron
Diffusion Equation of Reactor Physics, Prentice-Hall, Englewood Cliffs, N. J. (1966).
9. G. J. Habetler and M. A. Martino, Proceedings of Symposium on Applied Mathematics,
Vol. XI, American Mathematical Society (1961), p. 127.
10. R. S. Varga, Matrix Iterative Analysis, Prentice-Hall, Englewood Cliffs, N. J. (1962).
11. R. G. Steinke, A Review of Direct and Iterative Strategies for Solving Multi-
Dimensional Finite Difference Problems, University of Michigan Nuclear Engineering
Report (1971).
(1966).
PROBLEMS
7-1 Estimate the fast-group constants characterizing H 20 if the fast group is taken from
E 1 = 1eV to E 0 = 10 Me V and the neutron energy spectrum over this group is taken as
</>(E)~l/ E.
7-2 Estimate the minimum group spacing that will yield directly coupled multigroup
equations for 12C, 2D, 9 Be, and 22Na.
7-3 What percentage of the neutrons slowing down in hydrogen will tend to skip groups if
the group structure is chosen such that E8 _ 1/ E8 = 100?
7-4 Write out the detailed form of the multigroup diffusion equations, M<t>=k- 1f.<t>, for a
four-group model in which: (a) there is direct coupling, (b) the fission sour~ exists
only in the upper two groups, and (c) only the lowest group contains thermal
neutrons.
7-5 Repeat Problem 7-4 for the case in which the lowest two groups correspond to
thermal neutrons (which will be characterized by appreciable upscattering).
7-6 Calculate the critical size and mass of a bare sphere of pure 235 U metal using the
group constants characterizing groups 1 and 2 in Table 7-1.
TABLE 7-1 ANL Four-Group Microscopic Cross Sections (in barns)t
Lower
g Energy u Fission H 20
EL [lnE0 / Ed Spectrum 0-1r 0-y Og-g+ I
I 1.353 MeV 2 0.575 3.08 0 2.81

2 9.12 keV 7 0.425 10.52 0 4.04
3 0.4 eV l7.D3 0 16.55 0.035 4.14
4 0 - 0 68.6 0.57 0
g u23s u23s
p at, p or ay
()"tr or aY "inR 0-eR 0-inR aeR
I 4.7 2.65 1.3 0.1 1.4 0 4.7 2.65 0.53 0.04 2.1 0
2 7.0 2.55 1.4 0.3 0 0 7.0 - 0 0.18 0 0
3 51.0 2.5 23.0 18.0 0 0.01 11.0 - 0 0.8 0 0.01
4 .597.0 2.5 490 97 0 0 13.0 - 0 2.4 0 0
tFrom Reactor Physics Constants, ANL-5800 (1963). [Here, ainR and aeR refer
to the inelastic and elastic scattering removal cross sections for the group
similar to the definition given in Eq. (7-23).]
7-7 Compare the critical radius of a 235 U sphere as given by one-group, two-group, and
modified one-group (1-1/2 group) diffusion models. Again use the data from groups 1
and 2 of Table 7-1.
7-8 Determine the thermal flux due to an isotropic point source emitting S 0 fast neutrons
per second in an infinite moderating medium. Use two-group diffusion theory. In
particular, discuss the solution to this problem for the case in which L 1 > L 2 and
L1<L2.
7-9 Consider a plane source emitting S 0 fast neutrons/cm2 sec at the center of a
subcritical slab of thickness a. Determine the two-group fluxes established in the
assembly.
7-10 We have listed in Table 7-2 typical group constants characterizing a PWR (in both
two-group and four-group forms). Calculate the multiplication factor of a PWR core
of height 370 cm and diameter 340 cm using two-group diffusion theory.
TABLE 7-2 Few-Group Diffusion Theory Constants for a Typical PWR Reactor Core
Two-Group Four-Group
Group
Constant I of 2 2 of 2 I of 4 3 of 4 3 of 4 4 of 4
P~r .008476 .18514 .009572 .001193 .01768 .18514

~r .003320 .07537 .003378 .0004850 .006970 .07527
~a .01207 .1210 .004946 .002840 .03053 .1210
D 1.2627 .3543 2.1623 1.0867 .6318 .3543
~R .02619 .1210 .08795 .06124 .09506 .1210
7-11 Using the group constants of Table 7-2, calculate each of the terms in the six-factor
formula for the core described in Problem 7-10.
7-12 The neutron "age" -r is defined as 1/6 of the mean-square distance a fast neutron will
travel before it slows down or is absorbed. Derive an expression for the age -r in terms
of two-group constants. In particular, compute the age for the group constants
characterizing a LWR in Table 7-2.
7-13 Recall that we defined the conversion ratio CR characterizing a reactor as the ratio of
the production rate of fissile nuclei to the destruction rate of fissile nuclei. Derive an
expression for the conversion ratio in a slightly enriched uranium-fueled reactor (such
as a LWR) at the beginning of core life (i.e., such that the plutonium density is zero).
7-14 One important method of controlling reactivity in a PWR is to dissolve a poison such
as boron in the coolant. If we assume that boron only affects the thermal absorption,
~a , derive an estimate of the critical boron concentration that will render k 00 = 1. Use
2
this expression to estimate the boron concentration necessary to render the reactor
described in Problem 7-10 critical. Take a:, 0:
~o, 2 =2207 b.
7-15 Calculate the one-group constants corresponding to collapsing the two-group set
given in Table 7-2.
7-16 Derive a general group collapsing expression for N-group constants in terms of
N X M-group constants (e.g., 2-group in terms of 4-group ).
7-17 Calculate the two-group constants corresponding to collapsing the four-group con-
stants in Table 7-2 by assigning groups 1, 2, and 3 to group 1 and group 4 to group 4,
respectively. Compare these with the two-group constants listed in Table 7-2.
7-18 Using two-group theory, determine the critical core width of a slab reactor with core
composition similar to that of a PWR and surrounded by an infinite water reflector.
Use the two-group constants of Table 7-2 supplemented with group constants
characterizing the water reflector: D 1 =1.13 cm, D 2 =0.16 cm, ~R,=0.0494 cm- 1
~ ~. 12, ~a2 = 0.0197 cm- 1, ~a,= .0004 cm- 1 •
7-19 A beam of Jin fast neutrons/cm2 /sec is incident on the plane face of an infinite
homogeneous nonmultiplying half-space. Use two-group diffusion theory and partial
current boundary conditions to calculate J out, the partial current of thermal neutrons
going back toward the fast source.
7-20 Consider a thin slab of fuel surrounded by an infinite moderator. Suppose that the
slab is thin enough that it is essentially transparent to nonthermal neutrons. However
because of its large thermal absorption cross section, all thermal neutrons striking the
surface of the slab are absorbed. Suppose further that every thermal neutron absorbed
in the fuel results in the net production of 'II fission neutrons. Using the two-group
diffusion theory model, develop an equation for the effective multiplication factor for
this system.
7-21 Assuming that two-group diffusion theory with one group of delayed fission neutrons
represents a valid description of the reactor, write the three coupled differential
equations for the time and spatial dependence of the neutron precursor fragment
concentration C(r,t) and the fast and slow group fluxes cf, 1(r,t) and cf,i{r,t) respec-
tively. The average speed, diffusion constants, and removal (by all types of collision)
cross sections in the fast and slow groups are v 1, D 1, ~R, and v 2 , D 2 , and ~R2
respectively. Fission can be produced only by slow group neutrons, and the cross
section is ~r2, There are " neutrons emitted per fission, all into the fast group, with a
fraction f3 delayed. The decay constant of the neutron precursor is ;\, and the cross
section for transfer of neutrons from the fast group to the slow group is ~., 2 • By
assuming that the spatial shapes of cf, 1, cf,2 and C are that of the fundamental mode
with geometric buckling B 2 , find an equation for the transient and stable periods (i.e.,
the inhour equation). How many transient terms are there?
7-22 Write out the explicit form of the matrix equations Mcf,=k- 1f.p_ and d, 82s=
§g + B, 8 _ i'.&- i for a four-group, 4 x 4 X 3 three-dimensional spatial mesh problem in
which the lowest two groups are both taken in the thermal range in which significant
upscatter can occur.
7-23 Write a simple two-group, one-dimensional diffusion code (similar to the one-group
code written in Problem 5-27). Then treating the LWR described in Problem 7-10 as a
slab geometry, calculate the width necessary for reactor criticality. ,
7-24 Using ANL-7411, list and contrast various multigroup diffusion codes. In particular,
compare their group structure, inner-outer iteration strategy, source-extrapolation
methods, criticality search options, and estimated running times.
7-25 Calculate cp 1,cp2 and the adjoint fluxes, c/>j,c/>i for a bare, spherical reactor of radius R.
7-26 Compute and sketch both the two-group fluxes and adjoint fluxes for the reflected
slab reactor described in Problem 7-18.
7-27 A critical system consisting of a slab surrounded on both sides by infinite reflectors is
to be described by two-group diffusion theory. The center third of the core suddenly
has its thermal group absorption cross section increased slightly. Find the change in
core multiplication by perturbation theory. Then to demonstrate that the work saved
over directly calculating the multiplication factor for the perturbed system is very
great, carry out the latter calculation until as many pages are filled as were required to
perform the perturbation calculation.
8
Fast Spectrum Calculations
and
Fast Group Constants
The principal tool of nuclear reactor analysis is multigroup diffusion theory. In

Chapter 7 we developed the general form of the multigroup diffusion equations and
prescribed a strategy for their solution. However these equations contained various
parameters known as group constants formally defined as averages over the
energy-dependent intragroup fluxes cj>(r, E) which must be determined before these
equations assume more than a formal significance. The determination of suitable
approximations for the intragroup fluxes, that is, the neutron energy spectrum, is
the key to the generation of group constants that will yield an accurate few-group
description of nuclear reactor behavior.
Hence our goal in this chapter and Chapter 9 is to develop methods for
generating few-group constants by averaging fundamental microscopic cross sec-
tion data over suitable approximations to the neutron energy spectrum. Of course
there are two aspects to this problem, since both the energy and the spatial
dependence of the intragroup fluxes must be estimated. Because of the extremely
complicated dependence of microscopic neutron cross sections on energy, it is
necessary to provide a rather careful treatment of the energy dependence of the
intragroup flux. Of course if the number of groups is very large ( > 50), one can
frequently get by with rather crude estimates of this energy dependence. For
example, for energies above 1 MeV, one might approximate cj>(E) by the fission
spectrum x(E); for intermediate energies, 1 eV < E < 1 MeV, we will see that cp(E)
behaves very roughly as 1/ E; while for energies below 1 eV, we might model cj>(E)
by a Maxwell-Boltzmann distribution characterizing neutrons in thermal equi-
librium with the reactor core material at a temperature T.
Unfortunately, for the generation of few-group constants, one requires a far
better treatment of the energy dependence of the intragroup flux. In fact one is
315
usually required to perform a detailed multigroup calculation of the neutron energy

spectrum using a fine-group structure (with a crude treatment of spatial depen-
dence) which can then be used to average or "collapse" fundamental cross section
data to generate few-group constants.
The methods used to generate the neutron energy spectrum vary, depending on
the range of neutron energies of interest. For example, at high energies the
dominant process is neutron slowing down via both elastic and inelastic scattering.
At intermediate energies, resonance absorption becomes quite important. At low
energies, upscattering becomes appreciable as the neutrons tend to approach
thermal equilibrium with the nuclei comprising the reactor core. Hence it is
customary to divide the range of neutron energies into three different regions, each
characterized by these different types of interaction, as indicated below:
Upscattering Elastic scattering Elastic scattering

from stationary, (anisotropic in
Chemical free nuclei (iso- CM, p-wave)
binding tropic in CM,
s-wave) Inelastic scattering
Diffraction
No upscattering No upscattering
resonance absorp- resonance absorption
tion (unresolved resonances)
(resolved reson- fission sources
ances)
_____ ____
\.0
.....,,..
1 eV
) ""------..__,,..______10 eV
.)
5 107 eV
Neutron Neutron moderation Fast fission

Thermalization or slowing down
In this chapter we will be concerned with the calculation of the neutron energy
spectrum characterizing fast neutrons. Hence our study of the calculation of fast
neutron spectra and the generation of fast group constants will be dominated by a
development of the theory of neutron slowing down and resonance absorption. In
Chapter 9 we then consider the calculation of thermal neutron spectra and thermal
group constants involving the development of the theory of neutron thermalization.
One must also account for the spatial dependence of the intragroup fluxes, as
well as the spatial variation of the reactor core composition. Of course a detailed
treatment of this spatial dependence in the generation of multigroup constants
would be impractical-indeed the primary motivation for the generation of few-
group constants is to yield a sufficiently coarse treatment of neutron energy
dependence to allow such a detailed spatial calculation. Instead we must introduce
a very approximate treatment of the spatial dependence of the intragroup fluxes.
Of course the most drastic such approximation is simply to ignore the spatial
dependence altogether in the calculation of the neutron spectrum. That is, one
effectively assumes that the intragroup fluxes can be calculated by assuming that
neutron slowing down and thermalization occur in an infinite medium in which
there is no spatial dependence. We will begin our study of both fast and thermal
spectra for this situation because of its simplicity.
FAST SPECTRUM CALCULATIONS AND FAST GROUP CONSTANTS / 317
In most cases, however, one cannot so cavalierly ignore the spatial dependence
of the intragroup fluxes. Hence we will modify our study of infinite medium
spectrum calculations to include a very simple treatment in which the intragroup
flux spatial and energy dependence are assumed to be separable, and this spatial
dependence is then characterized by a single parameter (an effective value of the
geometric buckling Bf).
However even this extended treatment is inadequate for many situations in
which the detailed geometry of the reactor lattice must be taken into account. This
is particularly true for an adequate treatment of both resonance and thermal
neutron absorption. The few-group constants must be modified to account for the
variation of both the neutron flux and the material composition in a small
subregion or cell of the reactor core. In effect one must spatially average the
few-group constants over a unit cell of the core lattice (much as the microscopic
group constants have been averaged over energy to generate the few-group con-
stants) before they can be used in a few-group diffusion calculation. We will
consider such cell-averaging techniques in Chapter 10.
Therefore the generation of appropriate few-group constants usually involves a
sequence of both energy and spatial averages of the microscopic cross section data
over approximate estimates of the neutron energy spectrum and spatial dependence
characterizing the neutron flux in a given subregion of the reactor of interest. As
such, these group constants are dependent on the reactor design of interest, and
also the region of the reactor core under investigation, its present operating
conditions (e.g., coolant densities, temperatures), as well as its past history (e.g.,
fuel depletion, fission product buildup). The generation of multigroup constants is
an extremely important (and usually time consuming) facet of nuclear design since
it is the key to the successful implementation of the multigroup diffusion technique.
I. NEUTRON SLOWING DOWN IN AN INFINITE MEDIUM
A. Introduction
Our initial goal is the calculation of the energy dependence of the neutron
flux cf,(r,E) in the fast energy range, E>l eV, in which upscattering can be ignored.
To simplify this calculation, we will first consider neutron slowing down from
sources uniformly distributed throughout an infinite, uniform medium. 1•2 •4 In this
case, all spatial dependence disappears, and the neutron continuity equation, Eq.
(4-79), reduces to an equation for the neutron energy spectrum cf,(r, E)-cf,( E) of the
form
(8-1)
That is, the neutron transport equation simplifies rather dramatically to an integral
equation in the single variable E, which we will refer to as the infinite-medium
spectrum equation.
Of course the neglect of all spatial dependence may seem very drastic, and we
will later study schemes for reincluding it in our analysis. In most large reactors,
however, leakage is a relatively minor effect in comparison to neutron energy
variation. Furthermore even spatial effects due to core heterogeneities are usually
of secondary importance, particularly in the fast neutron energy range. Hence we

are certainly justified in focusing our initial attention on the energy dependence of
the neutron flux.
Before we can proceed further, we must provide more information concerning
the energy dependence of the neutron cross sections "2,i(E) and "2,.(E' ~E). Consis-
tent with our present concern with fast neutron spectra, we will assume that the
scattering nuclei are initially at rest and free to recoil. That is, we will consider only
fast neutron energies much greater than the thermal energy of the nuclei E » kT so
that upscattering in collisions can be ignored.
We will further restrict our initial investigation by considering only the process
of elastic scattering that is isotropic in the center of mass system. In this case, our
earlier study of two-body collision kinematics has indicated that the differential
scattering cross section is of the form
E <E'<E
a (8-2)
otherwise,
where a= (A - 1/ A + 1)2. This form is sufficiently simple to enable us to obtain

analytical solutions to the neutron slowing down equation
in many cases of interest, even without specifying the detailed forms of the cross
sections "2,.(£) and "2,a(E). Of course it will be necessary to use more general
numerical methods in solving Eq. (8-3) for most practical situations. However those
analytical results we will develop are of considerable use in formulating efficient
numerical schemes, as well as in checking their accuracy.
By restricting our attention to s-wave elastic scattering, we have limited the
applicability of our initial analysis to neutron energies below I MeV in light
moderators, and below 0.1 MeV in heavier materials,3 since the average scattering
angle cosine µ,0 in the CM system is given roughly by µ,0 =0.07 A 213£[MeV]. More
specifically, in hydrogen, s-wave scattering is dominant below several MeV. By way
of contrast, for heavier nuclei such as 160, p-wave scattering becomes significant
above roughly 50 ke V. We will justify our restricted study of only s-wave scattering
by noting that in thermal reactors, most neutron moderation is due to light
elements such as lH, fD, or 1~C.
For neutron energies above 0.1 MeV, inelastic scattering from heavy nuclei
becomes important and cannot be ignored. Unfortunately the complexities of
inelastic processes severely limits the extent to which one can use analytical
methods to investigate neutron slowing down. Hence our consideration of inelastic
scattering will be deferred until our discussion of the numerical solution of the
neutron slowing down equations.
We will begin our study of Eq. (8-3) for the particularly simple case of neutron
slowing down in an infinite medium of hydrogen. This study is of particular
importance since hydrogen is a very common moderator, and moreover this
analysis will serve to introduce a number of concepts which prove useful in the
study of more general neutron slowing down problems.
We will then turn to the study of neutron slowing down via elastic scattering
from nonhydrogeneous moderators. The reader should be cautioned that while the
detailed analysis of neutron slowing down in this section may look lengthy and
complicated, the actual results of this analysis are extremely simple. We will find
that the neutron flux will usually behave essentially as </>(E)~ 1/L.(E)E, provided
one is not near energies characterizing a neutron source or a strong cross section
resonance. However, as is, characteristic of so much of physics, the analysis
necessary to derive and justify this result is considerably more complex than the
result itself.
B. Neutron Moderation in Hydrogen
1. SLOWING DOWN IN THE ABSENCE OF ABSORPTION
We will first examine the very simple case in which all absorption is ignored.
Actually such an assumption is not too drastic in hydrogen, since a!!/ a!"'~.014.
However we are more generally concerned with neutron slowing down in a
hydrogeneous moderator containing in addition a heavy mass absorber (such as
238
U). Even in this case, most neutron absorption occurs in fairly narrow energy
ranges corresponding to absorption resonances, and for energies far from these
resonances, the neglect of absorption will yield the correct quantitative behavior of
the neutron spectrum <f>(E).
For hydrogen we set A= 1 and hence aH =0. Then in the absence of absorption,
Eq. (8-3) becomes
(8-4)
We will begin by determining the neutron energy spectrum </>(E) resulting from a
monoenergetic source of strength S0 emitting neutrons of energy E 0 • It proves
convenient to introduce a change of dependent variable in Eq. (8-4) by identifying
the collision rate density F(E) = L.(E)</>(E), so that we can rewrite the slowing
down equation (8-4) as
(Eo F(E')
F(E)= )E dE' E' +S0 8(E-E0 ). (8-5)
Here we have set the upper limit of integration at E 0 since no neutrons can achieve
energies higher than the source energy if only downscattering is possible.
We could solve this problem as we did the plane source diffusion problem of
Chapter 5, by restricting our attention to energies E < E 0 and using the source as a
boundary condition as E-E0 • However it is more convenient in this case to solve
Eq. (8-5) directly. First note that because of the singular source, the solution F(E)
must contain a term proportional to 8 ( E - E 0). Hence we will seek a solution of the
form
F(E) = Fc(E) +C8 (E- E 0 ), (8-6)

where C is an undetermined constant and Fc(E) is the nonsingular component of

F(E). Substituting this form into Eq. (8-5), one finds
(8-7)
However the singular contributions must be equal. Hence we demand C = S0• Then
we find that Fc(E) satisfies the remaining nonsingular equation
(8-8)
Notice something rather interesting here. Fc(E) satisfies an equation in which the
source term is just that corresponding to source neutrons making a single collision
at the source energy £ 0-that is, the effective source presented by "once-collided"
neutrons
(8-9)
Thus Fc(E) can be interpreted as the collision density due to neutrons that have
suffered at least one collision (hence the subscript "c"). In a similar sense, we now
identify the singular component of F(E), S 0 8 (E- E 0), as simply the contribution
due to uncollided source neutrons (present only at the source energy £ 0).
We now must solve the remaining equation (8-8) for Fc(E). We can solve this
integral equation by first differentiating it to convert it into an ordinary differential
equation (refer to Appendix B):
dFc = _.!!_ [
dE dE
J:EodE , Fe(E'E')
E
l _.!!_ ( So )
+ dE E 0 (8-10)
or
(8-11)
The general solution to Eq. (8-11) is of the form Fc(E)= C/ E. To determine the
constant C, note that we can infer an initial condition at E = E 0 directly from the
integral equation (8-8):
(8-12)
Hence we must choose C = S0 to find the total solution
(8-13)
or
(8-14)
It should be observed that if the scattering cross section is only weakly depen-
dent on energy (which it usually is), then below the source energy £ 0 the flux
assumes a 1/ E form. This functional form could in fact be used as a crude
estimate of the flux in generating group constants characterizing neutron slowing
down, but we will proceed to develop more sophisticated models.
A very similar analysis can be used to construct the solution corresponding to a
distributed source S (£). In this case, one finds (see Problem 8-2) that
cp( E ) =
1
( )
Joo dE 1
S (E ) +
1 S(E)
( ) . (8-15)
Ls E E E Ls E
Once again we note that below the source energies, the flux behaves essentially as
1/ E.
2. SOME USEFUL DEFINITIONS
(a) THE SLOWING DOWN DENSITY
It is useful to introduce several new definitions at this point in order to

facilitate our later analysis. We first want to define the neutron slowing down density
q(E) describing the rate at which neutrons slow down past a given energy E. To be
more general, we define
q(r,E)d3r= number of neutrons slowing down past energy (8-16)

E per sec in d 3r about r.
If we recall that the differential scattering cross section describes the probability
II 11
that a neutron will scatter from an initial energy E' to a final energy E in dE ,
then we can write
Rate at which neutrons

that suffer collision = [ ( ELs(E'-E 11 )cp(r,E')dE ]dE'.
11
(8-17)
at energy E' in dE' J~0
slow down past E
Hence the total slowing down density resulting from all initial energies E' > E is
given by
(8-18)
We can apply this definition to calculate the slowing down density in our infinite,
nonabsorbing, hydrogeneous moderator as
q(E)=
l i E
dE"
[ 00
dE'
~ (E')cp(E')
s E'
l
0
- tdE"[f dff,' (Fe(E')+S 8(E'-E

0 0 ))]
=E [
(
)E
00
dE'
Fc(E')
E' +
S0l
Eo =EFc(E)=S0. (8-19)
Hence for this simple problem, the slowing density is constant and equal to the
source. This is of course understandable, since in the absence of absorption and
leakage, all source neutrons must eventually slow down below the energy E.
(b) THE NEUTRON LETHARGY
The energy range spanned by neutron slowing down is extremely large,

ranging from 107 eV down to 1 eV. Furthermore we have found that in elastic
scattering the neutron tends to lose a fraction of its incident energy rather than a
fixed amount of energy. These considerations suggest that it would be more
convenient to use as an independent variable, the logarithm of the neutron energy
E. To this end, we define the neutron lethargy
Eo
u=ln-. (8-20)
E
Here the energy E 0 is chosen to be the maximum energy that neutrons can achieve
in the problem. It is usually set either at the source energy for a monoenergetic
source problem, or chosen as 10 MeV for a reactor calculation.
Notice that as a neutron's energy E decreases, its lethargy u increases. That is, as
a fast neutron loses energy via scattering collisions, it moves more slowly, becom-
ing more "lethargic." The neutron lethargy is such a convenient variable that it is
customary to perform fast spectrum calculations in terms of u rather than E. Hence
we must convert all of our earlier equations over to this new independent variable.
To accomplish this, we first compute the relationships between differentials
E )( - Eo) - dE
du= ( Eo ~ dE=~- (8-21)
Then, for example, we can calculate the collision density in terms of lethargy by
writing
F(u)du= - F(E)dE
or
F(u) = EF(E).
(The minus sign appears because lethargy increases as energy decreases.) If we

apply this to our earlier example of slowing down in an infinite, nonabsorbing
medium of hydrogen, we find for this case, the collision rate density is a constant in
the lethargy variable
F(u)= EF(E)= S0 • (8-23)
In more general cases, we will find that the collision density F is usually a much
more slowly varying function of u than E and hence is easier to approximate in the
lethargy variable.
As a second example, we can compute the elastic scattering probability function
P ( E '-E) in lethargy by noting
(8-24)
Thus
1 ( E ) eu'-u
(8-25)
=(1-a) E' =(1-a)'
Also
E< E' < E/ a~u-ln(l/ a)< u' < u. (8-26)
Hence we find
P(u'-u)={ (~~-;)' u-ln(l/a)<u'<u (8-27)

0, otherwise.
We can now rewrite our slowing down equation for hydrogen in terms of the
lethargy variable as
(8-28)
Now notice something rather interesting; from Eqs. (8-18) and (8-27), it is apparent
the slowing down density in the lethargy variable is given by
(8-29)
Now suppose we differentiate this expression with respect to u
(8-30)
Hence we can reidentify q(u) from Eq. (8-29) to write
dq
du + q(u) = ~.(u)cp(u). (8-31)
These equations relating the slowing down density to the flux are peculiar to
slowing down in hydrogen, but we will later find that they also hold when
absorption and spatial dependence are included. They will prove to be of consider-
able use in numerical studies of neutron slowing down.
As a final exercise involving the lethargy variable, let us compute the average
lethargy gain (corresponding to the average logarithmic energy loss) of a neutron in
a collision with a nucleus of arbitrary mass number:
(8-32)
or
2
a (A - 1) ( A + I )
g= 1 + 1 _ a Ina= 1- 2 A ln A_ 1 . (8-33)
By way of example, the mean lethargy gain per collision, g, is tabulated for several
moderators of interest in Table 8-1.
TABLE 8-1 Slowing Down Parameters of Typical Moderators
Number of
collisions
from 2 MeV
Moderator A a ~ p(g/cm3] to l eV ~~.[cm- 1] ~~,/~a
H l 0 1 gas 14
D 2 .111 .725 gas 20
H 2O .920 1.0 16 1.35 71
D 2O .509 1.1 29 0.176 5670
He 4 .360 .425 gas 43 1.6 X 10- 5 83
Be 9 .640 .209 1.85 69 0.158 143
C 12 .716 .158 1.60 91 0.060 192
238U 238 .983 .008 19.l 1730 0.003 .0092
An interesting application of this quantity is to compute the average number of

collisions necessary to thermalize a fission neutron, that is, to slow it down from 2
MeV to 1.0 eV. This is given by
ln 2X 106
<#)= 1.0 14.5
g =-g- (8-34)
and is compared for several moderators of interest in Table 8-1. We can see rather
dramatically how much more effective low mass-number nuclei are at moderating
fast neutrons.
(c) MODERATING POWER AND MODERATING RATIO
The above analysis has indicated that the number of scattering collisions
necessary to slow a neutron to thermal energies is inversely proportional to f
Certainly the better moderators will be characterized by large values of f However
they must also be characterized by large scattering cross sections ~. since otherwise
the probability of a scattering collision occurring will be too small. Hence a more
appropriate measure of the moderating or slowing down power of a material is the
product~~.- To this end, one defines:
Moderating power=~~ •. (8-35)
However, even this parameter is not sufficient in itself to describe the effec-
tiveness of a material for neutron moderation, because obviously one also wishes
the moderator to be a weak absorber of neutrons. Thus it is customary to choose as
a figure of merit the moderating ratio, defined as
. . ~~.
M o deratmg rat10=y. (8-36)
a
The moderating power and moderating ratio are given for several materials of
interest in Table 8-1. From this comparison it is evident that D 2O is the superior
moderator. Indeed the moderating ratio of D 2O is sufficiently large that it can be
used to construct a reactor fueled with natural uranium. Unfortunately it is also
very expensive. Hence most thermal reactor designs choose to use somewhat
poorer, but cheaper, moderators such as light water or graphite, although this
usually requires the use of enriched uranium fuels.
3. SLOWING DOWN WITH ABSORPTION
.We will now include an absorption term in our infinite medium slowing down
equation. Actually the absorption cross section of hydrogen is negligible. The
physical situation we want to describe is that of a strongly absorbing isotope mixed
in hydrogen (such as 238 U). However this isotope can also scatter, and since its
mass number is not unity, it would invalidate our analysis. For simplicity, there-
fore, we will assume the absorber to be "infinitely massive" so that it does not slow
down neutrons-it only absorbs them. We will also ignore inelastic scattering.
Then the appropriate slowing down equation (again with a monoenergetic source at
energy E 0) is just
(8-37)
As in our earlier analysis, we seek the solution for the total collision density
F(E) = ~i(E)c[>(E) as the sum of a collided and uncollided contribution, similar to
Eq. (8-6). Then we find that the collided contribution Fc(E) must satisfy
dE'"i..(E')F/E') "i,.(£ S
l
Eo )
0
F(E)= -----+-- -0 (8-38)
c
E
E'"i, t (E') "i, t (EO) E O .
Note that again the effective source term in this equation corresponds to once-
collided source neutrons, suitably scaled down to account for neutron absorption
at the source energy £ 0 •
We again solve this equation by first differentiating to find
(8-39)
We can easily integrate this ordinary differential equation using the initial condi-
tion at E = E 0 implied by Eq. (8-38) to find
(8-40)
Notice in particular that if we were to set La_ 0 we would return to our earlier
solution Fc(E)= S0 / E. We can compute the slowing down density as before to
find
(8-41)
If we now note that
S0 = rate at which source neutrons are emitted at energy £ 0
q(E)=rate at which neutrons slow down past E,
then we can identify
q ( E) = probability that a source neutron is not absorbed while

(8-42)
So - slowing down from £ 0 to E.
Since most absorption in the slowing down energy range is due to resonances in the
absorption cross section, it is natural to identify the resonance escape probability to
energy E in hydrogen as just
(8-43)
To proceed further we need explicit forms for "i,a(E) and "i,i(E). We will develop
these later in this chapter when we return to consider resonance absorption. First,
however, we will generalize these results to the case of A > 1.
C. Neutron Moderation in Media with A >l

1. SLOWING DOWN WITHOUT ABSORPTION
We now turn our attention to the more general case in which neutrons slow
down via elastic scattering in an infinite medium of arbitrary mass number A > 1.
Once again we first consider the situation in which absorption is ignored. Then the
slowing down equation for the collided flux resulting from a monoenergetic source
at energy E0 is just
E
E<E0 . (8-44)
If we once again write this equation in terms of the collision density Fc(E) and
differentiate with respect to E, we arrive at
dFC F/E)
=---- (8-45)
dE E (1-a)E'
(1-a)-
a
The appearance of the first term corresponding to the upper limit of integration is
what fouls us up, since this term depends not on E but rather on E / a. This
equation is related to a differential-difference equation 20• 21 and requires an alterna-
tive method of solution. Its solution is complicated because the collision density is
in fact discontinuous (or possesses discontinuous derivatives) at energies anE0
(corresponding to the minimum energy a neutron can slow down to inn collisions).
These discontinuous "transients" in the collision density will eventually smooth out
to an asymptotic solution far below the source energy of the form Fc(E)~l/ E
(indeed, one can readily verify that such a function is a solution of Eq. (8-44) when
the source term is not present). Fortunately this asymptotic behavior is sufficient
for most slowing down studies, since one is usually concerned with either hydro-
gen-dominated slowing down (such as in a LWR) in which the transients do not
arise or with the behavior of the neutron flux far below the fission source energies.
Only for intermediate mass number moderators such as deuterium do the
transients cause difficulties.
However for completeness we will sketch the detailed solution 1 of the slowing
down equation for arbitrary mass number A. It is most convenient to recast Eq.
(8-44) in terms of lethargy. However we must be careful in our study of this
equation since a neutron can only gain a maximum lethargy of ln(l / a) in a
collision. Hence the source term will disappear from the equation for lethargies
u>ln(l/a). We first consider only the lethargy range 0<u<ln(l/a), in which
case Eq. (8-44) becomes
0< u <In_!__ (8-46)

a
We have denoted the solution in this range as cp 0 (u). Notice that to gain a lethargy
below u=ln(l/a), the neutron would have to experience at least two collisions.
Hence we refer to O< u < ln(l/ a) as the "first collision interval". We will introduce
the collision density to write
1
O<u<ln-. (8-47)
a
Now differentiate with respect to u to find
dFo- ( - a ) F(u)=O.
- (8-48)
du 1-a 0
The corresponding boundary condition is found, as usual, by examining Eq. (8-47)

as u-o to find
(8-49)
Hence we can readily solve for
(8-50)
[Note that in terms of energy variables
Fo(E)= (1-a)E
So ( Eo)
E l
J:a · (8-51)
To handle higher order collision intervals-say between ( n - 1) ln ( 1/a) < u

<(n+ l)ln(l/a) one must split up the range of integration to write
Fn (u) = -1- Ju 1
du' Fn (u')e" -u ' - - - -(n - l)ln _!_
a
1- a nln -
a
- - - - n ln _!_
a
I
nln-
1
+ -1-
a
f.
u-Jn -
a
I U
d , Fn-1 ( U ') e u' - u .
- - - - (n+l)ln_!_
a
a
(8-52)
Again we differentiate to find
(8-53)
We can integrate this to write
Fn ( u) = Fn ( n ln l /a) exp [ ( l ~ a )(u- n ln ~ )]

- l~a l;n _!_ du'
a
Fn- I ( u' - ln 1/ a)exp [ ( l ~a)( u - u')]. (8-54)
Since we know F 0(u), we can use this expression to generate higher order Fn(u).
For example
F,(u)-s,(
1
~~';)exp[ (1 ~.)u] - s, (;:, (u-ln !)exp[ (1~ a }u]
(8-55)
Note that there is a discontinuity between F 0(u) and F 1(u) at u=lnl/a:
F 0 (ln 1/ a)- F 1(ln 1/ a)= a/(1- a). (8-56)
In a similar manner, one can generate the higher order collision densities Fn(u).
These functions (so-called "Placzek functions") are illustrated in Figure 8-1. 5
Notice in particular how these functions appear to be smoothing out as the
neutrons suffer increasingly more collisions. In fact for u»lnl/a, we can use an
asymptotic solution to equation (8-46) in which the source term is neglected
F c (u') u'-u
J
u
,
Fe( u) = du ( l _ a) e . (8-57)
1
u-ln-
a
It is evident that F/u) =constant= C is a solution to this equation. However it is
2.6
2.4 A=4 f= 2.35
2.2
--
; 2.0
1.6
A=2
f1 = 1.38
1.4
1.2
A=1
1.0 ~ - - - - ~ - - - - ~ - - - - ~ - - - -
0 1 u
In J_
C( 2 In a 3 In J_
C(
FIGURE8-1. The collision density F(u) for A =2 and 4.

not evident just what relationship this solution has to Fn(u) for large n. Further-
more, this homogeneous equation provides no prescription for determining C.
Perhaps the most direct way to generate the asymptotic solution is to solve Eq.
(8-47) using Laplace-transform methods. 6 We will adopt a somewhat more pe-
destrian approach by noting that the slowing down density in the lethargy variable
is given by
(8-58)
If we substitute in our asymptotic form Fc(u)= C, we find
C u u' + In _I_ a
q(u)=-
_ f. du'eu'f adu"e-u"=c[I+-
_ lna]=Cf (8-59)
1 au-In-1 u 1 a
a
However in a nonabsorbing medium, we know that q(u) must be constant and

equal to the source rate S0 (since there is nothing to prevent the neutrons from
slowing down). Hence we must choose C = S0 / g, or
So
Fc(u)= T' (8-60)
or
So
</>( u) = f~\( u) . (8-61)
We can also write this asymptotic solution in terms of the energy variable as
(8-62)
Hence asymptotically at least the flux characterizing neutron slowing down in a

nonabsorbing medium of mass number A is very similar to that found in hydrogen,
with the exception of an additional factor, the mean lethargy loss per collision, f
As we have noted, the primary qualitative difference in the more general solution is
the occurrence of irregular (discontinuous) behavior near energies at which source
neutrons appear or strong absorption occurs. As one moves down in energy (up in
lethargy) these transients damp out, leaving the smooth asymptotic behavior given
by Eq. (8-62). The rapidity with which this asymptotic behavior is reached
decreases with the mass number of the moderating material (although such be-
havior is always present in hydrogen).
More frequently one is concerned with neutron slowing down in a mixture of
nuclides. Although we will treat this situation in some detail in later sections, we
will note here that our earlier analysis can be easily generalized to a mixture of
nuclides by writing
(8-63)
where L 8 (E) is the total scattering cross section and ~ is the average lethargy gain
per collision
N N
~- L ~;L~;)(E)/ L "i,~;)(E). (8-64)
i= I i= I
It should be noted that~(£) is in general energy (or lethargy)-dependent because of

the energy dependence of the scattering cross section.
2. SLOWING DOWN WITH ABSORPTION

We could now introduce absorption (again assuming an infinitely massive
absorber) into Eq. (8-44):
(8-65)
or in terms of the collision density,
(~ '[ L (E')] F(E')

8
F(E)= )E dE Li(E') (1-a)E' +S(E). (8-66)
The presence of the factor [L.(E)/Li(E)] makes it very difficult to make much
progress toward an analytical solution to this equation, with the exception of very
special cross section behaviors and elaborate approximation techniques. 1• 2 Rather
than discuss such techniques here, we will recognize that one is usually most
interested in the effects of resonance absorption, and so we will proceed directly to
a discussion of the various methods used to calculate the resonance escape
probability. We will return later to discuss approximate treatments of slowing
down when we consider as well how one treats spatial effects in neutron slowing
down.
D. Inelastic Scattering
Thus far we have concerned ourselves, with neutron moderation via elastic
collisions in which a neutron merely bounces off of a nucleus in billiard-ball
fashion, losing some energy in the process. However for higher energy neutrons
( > 50 keV), inelastic scattering processes are possible in which an appreciable
fraction of the incident neutron energy goes into exciting the nucleus into a higher
nuclear quantum state. Such scattering is extremely important in heavy mass nuclei
in which slowing down by elastic scattering is negligible. For example, if we
consider a l MeV neutron incident on a 238 U nucleus, the average energy lost in an
elastic scattering collision would be (1- a )E0 /2 = 0.0085 MeV. By way of contrast,
the average energy lost in an inelastic scattering collision is about 0.6 MeV.
Of course for such inelastic scattering processes to occur, the incident neutron
energy must be above the threshold corresponding to the lowest excited state of the
target nucleus. For light nuclei, this threshold is quite high (4.4 MeV in 12C and 6.1
MeV in 160, for example). However in heavier nuclei the inelastic scattering
thresholds are much lower (450 keV in 23 Na and 45 keV in 238 U), and inelastic
processes become quite important.
Since inelastic scattering is much more significant for fast neutrons in high
mass-number materials, it might be expected to be of considerable consequence in
fast reactors. Indeed inelastic scattering in materials such as sodium is the
dominant slowing down mechanism in fast reactors and must be carefully
accounted for in any realistic reactor analysis.
Unfortunately the details of the inelastic 'scattering cross sections are quite
complicated, and analytical investigations such as those we have applied to elastic
scattering are restricted to very simple nuclear models (e.g., the Weisskopf evapora-
tion model 2 or few level models 2°). Instead, one must usually proceed in practice
with a brute force fine-structure multigroup calculation. That is, one writes the
slowing down equation in multigroup form and solves it directly for many groups
in order to obtain the intragroup fluxes for calculating the few-group constants.
Such fine spectrum calculations may require as many as a thousand or more
microgroups to adequately treat the details of inelastic scattering. Furthermore
because of the rather large energy loss experienced by a neutron in an inelastic
scattering event, several energy groups are usually coupled together (indirect group
coupling). Later in this chapter we will consider in detail how inelastic scattering is
treated in the generation of fast group constants.
II. RESONANCE ABSORPTION (INFINITE MEDIUM)
A. Introduction
As neutrons slow down from fission energies in a nuclear reactor, they
experience an appreciable probability of being absorbed in the numerous sharp
capture resonances which characterize heavy nuclei such as 238 U or 232 Th. Such
resonance absorption of neutrons is an extremely important phenomenon in
nuclear reactors. It not only affects reactor multiplication, but fuel burnup and
breeding performance and reactor control characteristics as well.
One can distinguish several different types of resonance absorption in nuclear
reactors. Of paramount importance in thermal reactors is absorption in the well-
resolved low-energy resonances of fuel materials such as 238 U or 232 Th. The analysis
of neutron absorption in such well-separated resonances is straightforward, and we
will consider it in detail in this section.
Above several keV in fertile materials (and as low as 50 eV in fissile isotopes),
one finds that the resonance structure becomes so finely detailed that resonances
can no longer be individually resolved. The treatment of resonance absorption in
the region of such unresolved resonances is considerably more complicated, since it
requires the use of nuclear models to describe the resonance structure. Although
this subject is of some importance in fast reactor analysis, its complexity induces us
to refer the interested reader to more advanced studies of resonance absorption for
a detailed discussion. 7
An accurate treatment of resonance absorption is essential to reactor criticality
calculations, since this is one of the primary neutron loss mechanisms in both
thermal and fast reactors. Such a process enters into a multigroup diffusion
calculation through the fast multigroup absorption cross sections ~ag' and the
accurate estimate of these group constants will be one of our principal concerns. In
general, there will be a depression in the neutron flux at those energies in the
vicinity of a strong resonance, and the analysis of such flux depressions is of
considerable importance in determining multigroup constants.
Since resonance absorption in fertile material (2 38 U, 232Th) can lead to the
production of fissile material (2 39 Pu, 233 U), an accurate analysis of resonance
absorption is also of importance in predicting conversion or breeding ratios. Indeed
small inaccuracies in the treatment of resonance absorption can propagate sizable
errors in the estimate of both fuel depletion and fertile-to-fissile conversion8•
Resonance absorption is also extremely important in determining the kinetic
behavior of the reactor. The amount of such absorption depends sensitively on the
fuel temperature through the Doppler broadening mechanism. In fact the dominant
reactivity feedback mechanism of significance in very rapid power transients is
usually that due to the Doppler effect.
A rigorous treament of resonance absorption would attempt to determine the
neutron flux <f,(r, E) in a fuel lattice cell either by solving the transport equation
directly or using Monte Carlo techniques. Although such calculations are occasion-
ally performed, they are far too expensive for use in routine design analysis.
Instead one usually removes the spatial and angular dependence to arrive at an
equation for the energy-dependent neutron flux similar to the slowing down
equation (8-3). A direct numerical solution of this equation can be, and is,
commonly performed. However the large number of calculations typically required
in a reactor design preclude the extensive use of such direct methods in favor of
simpler approximate solutions of the slowing down equation to determine the
neutron flux in the vicinity of a resonance. Such approximate methods are usually
analytical in nature, although various aspects of the solution may require numerical
evaluation. When used for appropriate problems, they provide reasonable accuracy
with considerably less calculational effort than the more rigorous methods.
In this section we will attempt to introduce some of the more elementary
concepts and approximations useful in the study of resonance absorption. We will
confine our discussion to resonance absorption in an infinite medium (consistent
with our earlier study of neutron slowing down) and develop the principal
approximations useful in the calculation of the resonance escape probability and
fast group constants. However the results of such infinite medium calculations are
of only limited utility, since spatial variations of the neutron flux in a fuel lattice
can strongly influence resonance absorption. We defer the study of spatially
dependent effects until we consider cell calculation techniques in Chapter 10.
Our general approach is to consider neutron slowing down in an infinite,
homogeneous mixture of a heavy isotope characterized by absorption and scatter-
ing resonances and a moderator material having a constant scattering cross section
(at least over the resonance) and a negligible absorption cross section. In particular
we will study the solution of the neutron slowing down equation in the vicinity of a
well-isolated resonance.
To illustrate the essential ideas involved, we will begin by considering the
situation in which the moderating material is hydrogen and scattering from the
absorber nuclei is neglected. This case is particularly simple, since the neutron
slowing down equation can be solved exactly when the mass number of the
scatterer is unity.
B. Resonance Absorption in Hydrogen Plus An Infinitely Massive

Absorber
We will begin by studying resonance absorption due to an infinitely massive
absorber distributed uniformly through an infinite medium of hydrogen. The
infinite absorber mass implies that all neutron slowing down due to elastic
scattering will be due to hydrogen. We will also ignore inelastic scattering pro-
cesses. We have already solved the neutron slowing down equation (8-37) for this
case in Section 8-11-A to find the flux
(8-67)
where the resonance escape probability to energy E, p(E), is given by
(8-68)
We will apply this exact expression to calculate the resonance escape probability
for a single resonance at energy E 0 . For convenience we will ignore scattering from
the infinitely massive absorber and absorption in hydrogen, by assuming N Ho-:i
»NA at, o-;1« o-;. If we denote the resonance escape probability for this particular
resonance asp, then we find
(8-69)
The notation f Eo indicates that the integral is to be performed over energies in the
neighborhood of the resonance.
To proceed further, we must introduce an explicit form for the capture cross
section o-;(E). We will use the Doppler-broadened Breit-Wigner cross section
developed in Section 2-1-D:
A ) 1;2 E-E)
f=f ( 4E0 kT ' x=2 ( r O • (8-70)
We will first evaluate p in the limit in which the absorber concentration is

sufficiently dilute that the scattering from hydrogen is dominant even at the
resonance energy, that is, NAo-0«NHo-:i. This is known as the infinite dilution
approximation, and it is equivalent to assuming that the absorber concentration is
so dilute that the absorption resonance does not perturb the slowing down form of
the flux. Using this approximation in Eq. (8-69), we find that
(8-71)
To proceed further, we note that the major contribution to the integral comes from
those energies close to resonance, E '~ Ea. Hence we can extract 1/ E '~ 1/ Ea to
find
(8-72)
where the limits of integration have been extended to infinity since only a minor
contribution comes from the wings of the resonance. Hence in the infinite dilution
limit
(8-73)
Several features of this result are of interest. First note that the resonance escape
probability increases with increasing moderator density (or cross section), since
then the neutrons slow down through the resonance more rapidly. This effect is
particularly important in water moderated reactors because it leads to large
negative power coefficients of reactivity. That is, increasing power and hence
moderator temperature will decrease moderator density (usually via steam forma-
tion). Hence the resonance escape probability will decrea-se, thereby decreasing
reactivity.
Next it should be noted that the resonance is more effective in absorbing
neutrons if the resonance energy Ea is lower. This is easily understood when it is
recalled that the asymptotic collision density behavior is as F (E)~ I/ E. Hence
lower energy neutrons will experience more collisions with absorber nuclei and
therefore a higher probability of being absorbed. For this reason, the most
significant resonance absorption in thermal reactors occurs in the lower lying
resonances of fertile materials such as the 6.67 eV resonance in 238 U. The resonance
parameters characterizing several of the lower lying resonances of 238 U which are
significant in thermal reactor design are listed in Table 8-2.
238
TABLE 8-2: Low-Lying Resonance Data for U
Eo fn(eV) fy(eV) a 0 (b) rp(eV) ½(I - aA)E0(eV)
6.67 .00152 .026 2.16 X 105 1.26 .055

20.90 .0087 .025 3.19 X JO" 1.95 .174
36.80 .032 .025 3.98 X 104 3.65 .306
66.54 .026 .022 2.14 X 104 2.26 .554
102.47 mo .026 1.86 X 104 3.98 .850
116.85 .030 .022 1.30 X 104 1.32 .966
165.27 .0032 .018 2.41 X 103 0.98 1.37
208.46 .053 .022 8.86 X 103 2.63 1.73
Finally we should note that although a;(E) is temperature-dependent through

the Doppler-broadening mechanism, the infinite dilution limit p 00 has no tempera-
ture dependence. However this might have been anticipated since we have already
noted that in this approximation the neutron flux is not perturbed by the res-
onance.
Therefore we turn to the more general case of finite dilution. Unfortunately we
are now hampered in our ability to obtain a closed form evaluation of the relevant
integrals. If we substitute the form Eq. (8-70) for a;-(E) into Eq. (8-69) and extend
the limits of integration, we find that the general result for the resonance escape
probability
(8-74)
must be written in terms of a tabulated function 9 J (t/3),
J (f ,/3) =Joo dx 1/;(tx)

/3+ 1/;(f,x)
(8-75)
0
where
(8-76)
10 2 _.._
~
....
........
5 '-.1'
~
2
""'t---I"' ~
:'----
I "'-,_
1' 1'.I =
CJ
o.t
I
~ ' "~1
~~
= 0.2
10 1 = o.5 ,," ' r\ '

'
"' '
5
i"
'
= .
"' .,., "l's.'\
\ I\
l'lc \
'
' "\'
\'\_'\ ~
'~
2
'~
--..,.
5
10- 1 '
4 6 8 10 12 14 16 18
j (where /3 = 2i X 10- 5 )
FIGURE 8-2. The function J (t,/3) [L. Dresner, Resonance Absorption in Nuclear Reactors,
Pergamon, New York (1960)].
It is apparent from this form that the finite dilution resonance integral is indeed
temperature-dependent. In Figure 8-2 we have shown the function J (t /3) plotted
versus /3 for various values of the temperature parameter t, From this plot it is
apparent that as temperature T increases, f decreases, implying that J (t/3)
increases and hence that the resonance escape probability p decreases. That is,
resonance absorption increases with increasing temperature. This is the so-called
Doppler effect referred to earlier in our discussion of temperature coefficients of
reactivity.
TABLE 8-3 The Function J(f,P) for f=0.1-1.0 and P=zi X 10-st
j J(K,/3)
1=0.1 1=0.2 1=0.3 1=0.4 1=0.5 1=0.6 1=0.7 1=0.8 1=0.9 1= 1.0
0 4.979(2*) 4.970(2) 4.969(2) 4.968(2) 4.968(2) 4.968(2) 4.967(2) 4.967(2) 4.967(2) 4.967(2)
I 3.532 3.517 3.514 3.513 3.513 3.513 3.513 3.513 3.513 3.513
2 2.514 2.491 2.487 2.485 2.485 2.484 2.484 2.484 2.484 2.484
3 1.801 1.767 1.761 1.759 1.758 1.757 1.757 1.757 1.757 1.757
4 1.307 1.257 1.248 1.245 1.244 1.243 1.243 1.243 1.242 1.242
5 9.667(1) 8.993( I) 8.872(1) 8.831(1) 8.812(1) 8.802( I) 8.796(1) 8.792(1) 8.790(1) 8.788(1)
6 7.355 6.501 6.335 6.278 6.252 6.238 6.230 6.225 6.221 6.218
7 5.773 4.777 4.562 4.485 4.450 4.430 4.419 4.412 4.407 4.403
8 4.647 3.589 3.328 3.230 3.183 3.158 3.143 3.133 3.126 3.121
9 3.781 2.759 2.471 2.354 2.297 2.265 2.245 2.232 2.223 2.217
10 3.045 2.153 1.867 1.741 1.675 1.638 1.614 1.598 1.587 1.579
II 2.367 1.676 1.423 1.301 1.235 1.194 1.168 1.151 1.138 1.129
12 1.730 1.268 1.074 9.718(0) 9.119(0) 8.739(0) 8.484(0) 8.304(0) 8.174(0) 8.077(0)
13 l.164~ 9.081(0) 7.815(0) 7.087 6.629 6.322 6.107 5.950 5.833 5.744
14 7.172(0) 6.014 5.342 4.914 4.624 4.419 4.268 4.154 4.066 3.997
15 4.088 3.658 3.371 3.169 3.022 2.911 2.826 2.759 2.706 2.663
16 2.204 2.067 1.966 1.889 1.829 1.781 1.743 1.712 1.687 1.666
17 1.148 1.109 1.078 1.053 1.033 1.016 1.002 9.904(-1) 9.805(- I) 9.722(-1)
18 5.862(-1) 5.757(-1) 5.671(- I) 5.599( - I) 5.539( - I) 5.488( -1) 5.445(-1) 5.408 5.376 5.348
19 2.963 2.936 2.913 2.894 2.877 2.863 2.851 2.840 2.831 2.823
20 1.490 1.483 1.477 1.472 1.468 1.464 1.461 1.458 1.455 1.453
21 7.468(-2) 7.452(-2) 7.437(-2) 7.424(-2) 7.413(-2) 7.403(-2) 7.395(-2) 7.388(-2) 7.381(-2) 7.375(-2)
22 3.739 3.735 3.732 3.728 3.726 3.723 3.721 3.719 3.718 3.716
23 1.871 1.870 1.869 1.868 1.868 1.867 1.867 1.866 1.866 1.865
24 9.358(-3) 9.356(-3) 9.355(-3) 9.352(-3) 9.350( -3) 9.349( -3) 9.348(-3) 9.346(-3) 9.345(-3) 9.344(-3)
25 4.680 4.680 4.679 4.679 4.678 4.678 4.678 4.677 4.677 4.677
26 2.340 2.340 2.340 2.340 2.340 2.340 2.340 2.340 2.340 2.340
27 1.170 1.170 1.170 1.170 1.170 1.170 1.170 1.170 1.170 1.170
28 5.851(-4) 5.851(-4) 5.851(-4) 5.851(-4) 5.851(-4) 5.851(-4) 5.851(-4) 5.851(-4) 5.851(-4) 5.851(-4)
29 2.925 2.926 2.926 2.926 2.926 2.926 2.926 2.926 2.926 2.926
30 1.463 1.463 1.463 1.463 1.463 1.463 1.463 1.463 1.463 1.463
31 7.314(-5) 7.314(-5) 7.315(-5) 7.315(-5) 7.315(-5) 7.315(-5) 7.314(-5) 7.314(-5) 7.314(-5) 7.314(-5)
tL, Dresner, Resonance Absorption in Nuclear Reactors. Pergamon, New York (1960).
*Numbers in parentheses are powers of 10, which multiply the entry next to which they stand and all
unmarked entries below it.
To understand this behavior, recall that our earlier analysis indicated that
increased temperature causes a resonance to broaden. However since the area
under the resonance is essentially (almost) temperature-independent, the resonance
peak drops with temperature. However the broadened resonance increases the
energy range over which absorption occurs. This effect outweighs the slight
lowering of the resonance peak and gives rise to an enhanced absorption with
increasing temperature.
This effect arises because of the phenomenon of self-shielding. We will show later
that the neutron flux is depressed (see Figure 8-3) for those energies in the
neighborhood of the resonance. This effect is known as energy self-shielding since
the strong absorption of the resonance tends to shield the absorber nuclei from
neutrons with energy E~E0 (the flux depression). (We will later study a related
phenomenon known as spatial self-shielding.) As temperature increases, the res-
onance peak decreases, thereby decreasing self-shielding and the flux depression,
and increasing resonance absorption [i.e., the energy-integrated reaction rate
La{E)</>(E)].
'
- __
J2S ................
,,--
-.......
0:J
.......
.......
.......
Ea E
FIGURE 8-3. Flux depression in the neighborhood of a resonance.
This latter interpretation also explains why the infinite dilution limit shows no
temperature dependence, since in this case there is no flux depression (the absorber
concentration is too low to perturb the flux) and hence no self-shielding.
C. Resonance Integrals
We now turn our attention to the more general case of neutron slowing down
in an infinite, homogeneous mixture of absorber and moderator. It is possible to
cast the resonance escape probability characterizing this problem into a form very
similar to that for hydrogeneous moderators, Eq. (8-68). If we imagine the neutrons
being produced by a source of strength S 0 at energy E 0 , then the total absorption
rate experienced by neutrons while slowing down from energies E 0 to E is just
Hence the resonance escape probability is just
_ S0 - neutron absorption rate _ _ 1

p(E)- S -1 -S
J. Eo , , ,
dE ~a(E )<f,(E ). (8-77)
O o E
In a very similar manner we can define the resonance escape probability P; for
any single, well-isolated resonance located at an energy E;. To do so, we will
assume that all absorption occurs in a number of well-separated resonances lying
below the source energy E 0 . If the individual resonances are sufficiently widely
separated, then the flux occurring just above a given resonance at energy E; will
have assumed the asymptotic form,
s<i)
eff
<f>(E)~<f>asym(E) = -_- ' (8-78)
f2'..sE
where Sey/ is the original source strength S0 reduced by the resonance-escape
probabilities for all higher energy resonances-that is,
s(i)_
eff - s0 IIP}' (8-79)
j
(See Figure 8-4.) Hence we can identify the absorption probability for the res-
onance at E; as just
Absorption probability= ----b f. dE ~~( E )<t>( E ), (8-80)

Seff E;
FIGURE 8-4. Definition of effective source for a well-separated resonance.

and thus the resonance escape probability for this particular resonance becomes
just
(8-81)
To write this in a more useful form, suppose we normalize the flux cp(E) so that
far above the resonance it behaves as
I
cp(E)~- for E»E;, (8-82)
E
Then the resonance escape probability for the resonance is just
[ ct>( E )~ 1/ EJ. (8-83)
This is still not sufficiently convenient for our purposes, since to compute the
total resonance escape probability for a series of resonances, we would have to
multiply the corresponding P; together as
Ptotal = II P;· (8-84)
To facilitate this calculation, we first note that in general the absorption probability
1- P; is quite small. Hence we can use the expansion
exp [ - ( I - p;)] ~ I - ( I - p;) + .. · ~ P; (8-85)

to rewrite
(8-86)
where we have defined the effective resonance integral for the ith resonance as
[ct>(E)~ i; ]· (8-87)
Then to compute the total resonance escape probability, we need only add the
resonance integrals for each resonance
(8-88)
The resonance integral I; is a very useful quantity for characterizing a resonance.

We first should note that I; has the dimensions of a microscopic cross section
(barns), if we remember the normalization condition, </>( E)~ I/ E for E » E;, In
effect, I; represents the average absorption cross section characterizing the res-
onance, averaged over the flux within the resonance. We will find that I; is largely
independent of the properties of the moderator.
The resonance integral I; has yet another extremely important practical applica-
tion since it can be used directly in the generation of multigroup constants. To see
this, first recall the definition of the multigroup absorption cross section for a given
group g:
(8-89)
Now since the resonances are quite narrow, over most of the group energy range
the flux behaves asymptotically as </>(E)~ I/ E such that
(8-90)
Furthermore essentially all neutron absorption will occur in those resonances

contained in group g. Hence using our earlier definition of the effective resonance
integral, we find
iEg iEg
L =-----=--- (8-91)
ag ln(Eg_ 1/Eg) !lug
where i E g is written to indicate that only those resonances contained in the group
g contribute to Lag' and !lug represents the lethargy width of group g. Hence we
can generate multigroup constants characterizing absorption directly by using
resonance integrals. (We will later provide an alternative prescription using P;-)
Therefore "all" we need to do is figure out a way to calculate or measure the
resonance integrals themselves. We will now turn our attention toward several
approximate schemes for doing this.
D. Approximate Calculations of Resonance Integrals

We will now consider in more detail neutron slowing down through a single
resonance characterizing an absorber (denoted by "A") distributed uniformly
throughout a moderator ("M"). We will assume that the neutron flux has achieved
its asymptotic behavior </>(E)~ I/ E before reaching the resonance energy. We will
furthermore assume that the moderator has a constant scattering cross section and
a negligible absorption cross section.
The appropriate form of the neutron slowing down equation characterizing this
situation is
(8-92)
It is possible to solve this equation numerically. 10 However we will develop several

simple approximate solutions based on assumptions concerning the width of the
resonance.
Let us begin by noting several features of the cross sections that appear in this
equation. We can effectively ignore the absorption cross section of the moderator
over the range of the resonance to write
(8-93)
where we have further noted that scattering from the moderator is characterized by
the potential scattering cross section ~;;1 which is essentially constant over the
narrow range of the resonance.
The resonance absorber is characterized by a more complicated cross section
dependence
(8-94)
where we have separated the resonance and potential scattering components of the
cross section. We can now characterize the effective width of the resonance by
defining the practical width r P as the energy range over which the resonance cross
section exceeds the nonresonance component (which is essentially just ~~ + ~;;1
= ~P). To estimate rP, we can use the Breit-Wigner formula to compare
(8-95)
or
(8-96)
Typically for low-lying resonances in 238 U, Lo/Lµ~l0 3 • Thus rP~30fS'..7 eV

which is much larger than the actual width of the resonance.
Nevertheless for all resonances of interest in reactor analysis, the practical width
is much smaller than the average energy lost in a collision with a moderator
nucleus
(8-97)
This means that most of the range of integration, E < E' < E / aM, in the moderator
integral in Eq. (8-92) will be sufficiently far from the resonance that the flux
assumes its asymptotic form, </>(E)~ l / E, and hence we can replace the flux in the
integral with this asymptotic form with negligible error:

E
~ a: dE' (-1 )= L~
l E
aM dE 'L~c/>( E ')
(1-aM)E'
L~ (
(1-aM) JE E' E' E ·
(8-98)
If we substitute this into Eq. (8-92), we find that our slowing down equation
becomes
(8-99)
The major approximation we will make in our solution of this equation will involve
our treatment of the absorber scattering integral. This approximation will again be
based on the relative width of the resonance, but now compared to the energy loss
suffered in a collision with an absorber nucleus.
l. NARROW RESONANCE (NR) APPROXIMATION
We will first assume that the practical width r P of the resonance is small
compared to the average energy loss suffered in a collision with the absorber
nucleus:
(8-100)
Then we can approximate the absorber scattering integral as we did that for the
moderator by replacing cp(E) in the integrand by its asymptotic form
(8-101)
In the second integral we have noted that the absorber scattering cross section over
most of this range is essentially just that for potential scattering L:.
Using these results in Eq. (8-99), we find
(8-102)
Hence the narrow resonance approximation to the flux is
(8-103)
NR approximation as
I
NR_ A -
- fE/EaY ( E )c/>NR ( E) -
i dE
£
A
ay ( E)
[ l
We can calculate the corresponding expression for the resonance integral in the
LM + LA
s
Li{ E)
p
• (8-104)
Eo
This integral can now be computed using our Doppler-broadened resonance

shapes, just as it was for hydrogen. Since the average energy loss per collision
(l - a.A)E0/2 increases with increasing energy £ 0, we might expect the NR
approximation to be valid for higher energy resonances. For lower energy res-
onances, the energy loss in collisions with absorber nuclei will no longer be larger
than the practical width of the resonance r P and we must then consider an
alternative approximation.
2. NARROW RESONANCE INFINITE MASS ABSORBER (NRIM) OR WIDE

RESONANCE (WR) APPROXIMATION
We now consider the other limit in which the practical width of the resonance
is very wide compared to the energy loss suffered in a collision with an absorber
nucleus
(8-105)
(although we continue to treat the resonance width as narrow compared to

moderator energy loss f:.E IM). This is occasionally known as the wide resonance
case, although more frequently one obtains the corresponding approximation to the
absorber scattering integral by assuming the absorber to be infinitely massive such
that neutrons suffer no energy loss in a collision with the absorber. That is, we take
the limit as A-oo or a.A =(A- I/ A+ 1)2-1 to find
Then our slowing down equation [Eq. (8-99)] becomes

LM
L 1(E)</>(E)=Lt(E)</>(E)+ -j-. (8-107)
We can then solve for the flux in the NRIM approximation as
(8-108)
and the corresponding resonance integral as
(8-109)
Notice that the NR and NRIM approximations differ only in the way in which
scattering from the absorber nuclei is treated. If we note that usually the absorber
density is sufficiently low that L~»L~ then in fact the only difference between the
two forms for I lies in the subtraction of the absorber scattering cross section
· Lt(E)from the total cross section in the denominator of the NRIM expression for
/.
If the absorber concentration is very small, Li{E)~L~, and we pass to the

infinite dilution form of the resonance integral
(8-110)
By comparing Eqs. (8-104) and (8-109) with (8-110), one finds that loci is larger
than either 1NR or JNRIM_ This is due to the self-shielding effect. If we examine
either of the approximate forms, Eq. (8-103) or (8-108), for the flux <f,(E) within the
resonance, we note that the flux decreases as L~(E) increases-that is, flux
depression occurs. This flux depression lowers the effective absorption in the
resonance and hence / is similarly decreased. We will find later in Chapter 10 that
spatial self-shielding due to fuel lumping has the same effect on resonance absorp-
tion since it also reduces /.
One can go further and introduce the Doppler-broadened Breit-Wigner line
shapes to calculate these resonance integrals in terms of resonance parameters. It is
customary (although not always correct 11 ) to neglect interference scattering such
that our cross sections become
(8-111)
Then inserting these into our expressions for the resonance integrals yields
(8-112)
and
(8-113)
where the definition of the J(t/3) is given in Eq. (8-75).

Here f3 is the parameter essentially characterizing the dilution of the absorber,
while !; characterizes the temperature. As f3 or f3' becomes large, both the NR and
NRIM forms approach the infinite dilution limit
(8-114)
As in the case of hydrogeneous moderators, we find that there is no temperature-

dependence in this limit. If we recognize that the temperature range of most
interest in reactor applications corresponds to 0.1 < !; < 1.0, then we find that unless
1.6 x 10- 4 < f3 < 2.6, there will be no temperature dependence of/ (and hence of p ).
For the range L~/NA~l0-103 b per absorber atom encountered in thermal
reactor design, one can estimate the derivative of / with respect to temperature as
(8-115)
Hence in this parameter range, Doppler broadening will cause a significant de-
crease in p with increasing temperature.
The NR approximation is commonly used for all but the lowest lying resonances
for which the NRIM approximation is more satisfactory. In Table 8-4 we have
given examples 12 of such calculations for several resonances in 238 U. These are
compared with direct numerical calculations of the resonance integral. Although
neither approximation seems to yield satisfactory results for the resonances at
intermediate energy, 13 when large numbers of resonances are accounted for, the
errors tend to average out. For convenience, we have also summarized the results
of these various schemes for handling resonance absorption in Table 8-5.
TABLE 8-4: Resonance Absorption Calculations12
Resonance (1-p;) (1-p;) (1-p;)

Energy E 0 (eV) NR NRIA Exact
6.67 0.2376 (21 %) 0.1998 (1.8%) 0.1963
21.0 0.07455 ( + 10.4%) 0.07059 ( + 4.5%) 0.06755
36.9 0.04739 (- 18.6%) 0.06110 ( + 5.0%) 0.05820
81.3 0.010163 ( + 5.9%) 0.008109 ( -15.5%) 0.009596
90. 0.001114 ( +0.9%) 0.000998 ( - 9.6%) 0.001104
117.5 0.009035 ( - 1.5%) 0.009500 ( + 3.6%) 0.009170
192. 0.005086 ( - 28.6%) 0.01228 (73.0%) 0.007119
212. 0.004440 ( - 11.6%) 0.007689 (+53.1%) 0.005021
TABLE 8-5: Resonance Integrals for Single Resonances
Hydrogen+ Infinite Mass NR NRIM

Absorption Approximation Approximation
Homogeneous
Infinite dilution
o0 rY'IT o0 ry'IT o0 rY'IT
Zero temperature ]=-- /=-- ]=--
2£0 2£0 2£0
o0 ry'IT oory o0 rY'IT
Homogeneous /=-- I=- 'IT
/=--
2E0a 2 0
£ Yac-b 2 /4 2E0a
NAot
Finite dilution ll=l+--M-
NMo,
NAoOry NAoo ~ NAoofy
Zero temperature a 2 =1 + b = - - oA/ X a 2 =1+
- NHo~r NMo~ P 'IT - NMo~r
NAoo
C=a+--M
NMo,
~Mr (~M+~A) r ~M
Homogeneous I= NS E J(t/3) I= s P -2.J(t/3) I=-' S_J(t,/3')
A 0 NA Eo NA Eo
Finite dilution
Finite temperature J(t/3)= 10

00
f(tx)dx
/3+f(tx)
/3
~M+~A
s p
NAoo
/3'
NMo~r
NAoOry
NHo:ir
/3-N A or
O y
We will return later in Chapter 10 to discuss the modifications required in these

results in order to account for heterogeneities (i.e., fuel lumping). Now, however,
we will continue with our study of neutron slowing down in an effort to take
account of spatial effects (i.e., finite media).
III. NEUTRON SLOWING DOWN IN FINITE MEDIA
A. The Lethargy-Dependent P 1 Equations

We now turn our attention to the determination of fast neutron spectra when
the spatial dependence of the neutron flux must be considered. Since the effects of
anisotropic scattering are usually significant, we will begin our study with the P 1
equations (4-153) and (4-154) rather than the energy-dependent diffusion equation.
This will allow a far more consistent treatment of the spatial and energy variables
than that provided by energy-dependent diffusion theory. It will be sufficient to
consider the steady-state form of these equations in only a single spatial variable x
since our intent is to develop a rather simple treatment of the spatial dependence
that we can use in the generation of fast group constants. It will prove convenient
to rewrite the P I equations in the lethargy variable u as
(8-116)
where ~~0 ( u' -u) and ~~ 1( u' -u) are the isotropic and linearly anisotropic com-
ponents of the differential cross section characterizing elastic scattering from
isotopic species "i." (Unless otherwise indicated, we will always use the cross
section notation ~s to refer only to elastic scattering in this section.) It should be
noticed that we have restricted these equations to the description of neutron
slowing down by truncating the upper limit of integration at u' = u (that is, ignoring
upscattering).
For the moment, we will include the effects of inelastic scattering and fission in
the source term
N N
S0(x,u)=~
i=l
lu du'~ln(u'-u)<f>(x,u')+ ~ x;(u) 1 du'v;~}(u')<f>(x,u')+ Sext(x,u),
O i=l O
00
(8-117)
and focus our attention on approximate treatments of slowing down via elastic
scattering. Although inelastic scattering can play a very important role in neutron
slowing down (particularly in fast reactors), it is not nearly as susceptible to
approximate treatments as elastic scattering and usually requires a direct mul-
tigroup treatment, discussion of which will be deferred until later in this section. It
should also be noted that we have explicitly ignored any anisotropic components of
inelastic scattering or fission by setting the source term S 1(x,u)=O in the second P 1
equation (8-116). Our neglect of this anisotropic source component is motivated by

recalling that both of these reactions involve compound nucleus formation in heavy
nuclei and hence should be adequately described as processes that are isotropic in
the LAB system.
To proceed further, we will now introduce the slowing down densities
characterizing elastic scattering from each isotopic species by rewriting Eq. (8-18)
in terms of the lethargy variable as
(8-118)
We have inserted the subscript "O" to distinguish qb(x, u) from a closely related
quantity defined in terms of the current density J (x, u):
(8-119)
We can make these definitions a bit more explicit by recalling that the elastic
scattering processes of greatest concern in nuclear reactor analysis involve s-wave
scattering, for which we can write
1 1 '
u-n-<u<u (8-120)
Ci;
otherwise.
One can similarly show that for s-wave scattering,
u - In_!_ < u' < u

Ci;
otherwise.
(8-121)
Thus we can use these cross sections to write the slowing down densities in more
f
explicit form. For example,
I
~i(
,,;. u ') e u'-u"
J
u u'+ln-
a-
qb(x,u)= du' 'du" •(1-a;) cf,(x,u')
1
u-ln- u
a;
(8-122)
To implement these definitions, we will differentiate the slowing down densities

with respect to u:
(8-123)
or noting that
J du"L~/u-u")=L~(u),
u
00
(8-124)
we find
(8-125)
In a very similar manner we can find
(8-126)
where we have noted that
J du" L~J u-u") = µ L~( u).

u
00
01
(8-127)
We can now use Eqs. (8-125) and (8-126) to replace the elastic scattering terms in
the P 1 equations (8-116) with the derivatives of the elastic slowing down densities:
(8-128)
Here we have defined the nonelastic cross section

N
Lne(u)-Li{u)-L:'(u)=Li{u)- L L~(u) (8-129)
i= I
and the transport cross section
(8-130)
The P 1 equations in the form (8-128) will serve as our point of departure for the
study of neutron slowing down in finite media. We would stress that the only
essential approximation used in deriving these equations involves the assumption
that the angular neutron flux (f!(X,JJ,, u) can be adequately represented by only a
linearly anisotropic dependence on the angular variable µ,-that is, the P 1
approximation.
By introducing the slowing down densities, we have eliminated the integral terms
characterizing elastic scattering. However in the process we have of course intro-
duced additional dependent variables, qh(x, u) and q\(x, u). Although these
variables are formally defined in terms of the flux and current by Eqs.(8-118) and
(8-119), it will prove convenient to seek alternative approximate equations for the
slowing down densities that can be used to augment the P 1 equations (8-128).
B. Approximate Treatments of Neutron Slowing Down-Continuous

Slowing Down Theory
In most schemes used to generate fast neutron spectra, elastic scattering is
treated using approximate methods that are usually referred to collectively as
continuous slowing down theory. In such methods, an approximate differential
equation is developed for each slowing down density by expanding the collision
densities L~(u') <f>(x,u') and L~(u')J(x,u') which appear in the integrals of the exact
definitions [Eqs. (8-118) and (8-119)] of the slowing down densities as Taylor series
expansions about u' = u and retaining only one or two terms of these expansions.
This in effect replaces the original integral equation by a set of coupled first-order
differential equations in the lethargy variable.
The motivation behind such an approximate treatment might be questioned
when it is recognized that a direct solution of the multigroup representation of the
original fast spectrum equations could be performed on any modern digital
computer-particularly since the absence of upscattering would allow a direct
successive solution of the multigroup equations for successively higher lethargies.
However to generate such a fine-group representation, it is necessary to calcu-
late fine-group constants by making a simple estimate of the intragroup flux,
say <f>(u)~constant. We have noted, however, that the scattering cross sections
L~(u' ➔ u) characterizing elastic scattering vanish discontinuously outside of the
collision interval, u - In 1/a;< u' < u. Such a stepwise behavior is difficult to
approximate accurately by a multigroup representation. Hence it is to our advan-
tage to replace the scattering integrals by smoothly varying terms more susceptible
to a multigroup representation by seeking approximate equations for the slowing
down densities, qb(x,u) and q\(x,u).
1. SLOWING DOWN IN HYDROGEN (A= 1)
We will begin by examining the particularly simple case of slowing down

from hydrogen in which A = 1 and hence a= 0. This case is a particularly
appropriate starting point since it can be treated exactly. If we set a= 0 in Eqs.
(8-118) and (8-119), we find
ql;( x, u) = fou du' L~( u')eu'-u<P( x, u'),

(8-131)
qf(x,u)= j fou du'L~(u')e (u'-u)/ J(x,u').
3 2
However of more relevance is the fact that by differentiating these expressions, we

can identify two simple differential equations,
0
:! +ql;(x,u)=L~(u)<[>(x,u),
(8-132)
oqf 3 H 2 H
a;:;+ 2 q1 (x,u)= 3 L 8
(u)J(x,u),
which can be used to supplement the P 1 slowing down equations (8-128). It should
be stressed that while these equations are exact, they only apply to the case of
hydrogen.
2. THE AGE APPROXIMATION

For the more general case of nonhydrogeneous moderators, one can not
obtain such differential equations exactly characterizing the slowing down densi-
ties. However essentially all treatments of elastic scattering attempt to obtain
approximate differential equations for qb(x,u) and q\(x,u) in the form
aqb .
;.\0
I
~
uu + qo(x,u) = /30 </>(x,u),
I
(8-133)
To accomplish this one first notes that in many cases the elastic scattering collision
density L~(u)<f>(u) is a slowly varying function of lethargy. For example, for zero
absorption in an infinite medium far below source energies, we know that the
asymptotic form of the collision density is in fact constant, L~( u)<f>(u) = S0 / ~;- Thus
if the system of interest is not too strongly absorbing, and if the collision interval is
not too large-that is, if A is large-we are tempted to expand the collision density
appearing in the integrals of Eqs. (8-118) and (8-119) as Taylor expansions about
u' = u and retain only low-order terms:
L~(u')</>(x,u')'.:::'.L~(u)<f>(x,u)+(u'- u) :u [L~(u)<f>(x,u) ],
(8-134)
L~(u')J(x,u')'.:::'.L~(u)J(x,u)+(u'-u) :u [L~(u)J(x,u)].
If we substitute the first of these expansions into Eq. (8-118), we find
(8-135)
where we have identified
e u,_ u - a; l a;ln _!_

a;
J
u [
~; = 1
du' 1 -a; = 1 - 1 - a., ' (8-136)
u-ln-
a;
(1n 1-)
2
u
eu u-a.l,_
a; a;
du' [ 1 - a; ' ( u' - u) = 2( 1 - a;) (8-137)

a;=
I I
u-ln-
a;
It is common to truncate Eq. (8-135) after only one term to write
(8-138)
This result is known as the age approximation or sometimes as continuous slowing

down theory. The former name will only become apparent in Section 8-IV-C. The
latter name arises because the neglect of higher order terms in the expansion of the
collision density effectively implies that neutrons lose only an infinitesimal amount
of energy in each collision.
To proceed further, we will assume that we can rather arbitrarily set
(8-139)
that is, we will neglect anisotropic energy exchange. Then the second of equations
[Eq. (8-128)] can be solved for
(8-140)
which yields just Fick's law. This can then be substituted into the first equation
(8-128) to find
where we have used the age approximation for all isotopes in the system. This is
known as the age-diffusion equation. It is sometimes also referred to as inconsistent
P1 theory since we have not treated the slowing density qi(x, u) in the same fashion
as we treated qh(x,u).
The age-diffusion equation is primarily of historical importance in reactor
analysis since it is restricted to slowing down in heavy mass moderators such as
graphite and would certainly not apply to hydrogeneous moderators: It is of
interest simply because it can be solved exactly in certain special cases (as we will
see in Section 8- IV-C).
We can partially alleviate the restriction to heavy mass moderators by using the
exact equations (8-132) to describe hydrogen moderation and then use the age
approximation (inconsistent P 1) only to describe nonhydrogeneous species (A> 1).
This scheme, which is one of the common methods used in computing fast spectra
in LWRs, is known as the Selengut-Goertzel method.
It is a relatively simple matter to remove the inconsistency present in the age
approximation [i.e., Eq. (8-139)] by substituting Eq. (8-134) into the definition of
qi(x,u) given by Eq.(8-119) to find
qi(x,u);;;;; t;~~(u)J (x,u) (8-142)

where
-A-I
Y;= ; · (8-143)
If Eqs. (8-138) and Eq. (8-142) are both used in the P 1 equations, one arrives at the
consistent P1 approximation.
One other modification occasionally encountered in the treatment of neutron

slowing down in weakly absorbing systems involves retaining both terms in the
expansion (8-135) to write
qo(x,u)~~L.(u)<f>(x,u)+a Ju [L.(u)</>(x,u)]
:::'.~L.(u)</>(x,u)+ I Ju [qo(x,u)] (8-144)
where we have reintroduced Eq. (8-138) into the second term to achieve a
consistent second-order expansion. Now in an infinite, sourceless medium, Eq.
(8-128) can be written as
(8-145)
If we substitute Eq. (8-144) into Eq. (8-145), we can solve for
(8-146)
However we can approximate a/~~ - ~ (see Problem 8-29) to write
(8-147)
This form of the age approximation is frequently used instead of our earlier form
given by Eq. (8-138). For weak absorption, both expressions yield very similar
results.
3. THE GRUELING-GOERTZEL APPROXIMATION

Thus far we have provided prescriptions for handling neutron slowing down
in either very heavy (age theory) or hydrogeneous (exact) moderators. However
neither of these schemes is adequate for light, nonhydrogeneous materials-with
deuterium 2 D being a prime villian. Hence we are motivated to construct an
alternative to the age approximation for nonhydrogeneous media.
To accomplish this, we will use the Taylor expansion of the collision densities to
write approximate equations for qh(x,u) and q\(x,u) in the forms Eq. (8-133). To
illustrate how this may be accomplished, suppose we use the expansion Eq. (8-135)
to calculate
To cast this into the form of Eq. (8-133), we can ignore terms in a 2<1>/au 2 as being
of higher order and choose Ao; so that terms in a</>/ au cancel. This yields
(8-149)
A very similar calculation will yield the second of equations (8-133) with the choice
Augmenting the set of P 1 equations (8-128) with Eq. (8-133) yields what is known
as the Greuling-Goertzel approximation. Since higher order terms in the collision
density expansion have been retained in this scheme, the Greuling-Goertzel
approximation will provide a far more accurate treatment of intermediate mass
moderators such as 2 D or 4 Be than that provided by age theory. This approxima-
tion also has the property that it reduces to the exact equations describing
hydrogen in the limit A- I.
4. SUMMARY OF THE P 1 SLOWING DOWN EQUATIONS

It is convenient to summarize these various approximate treatments of
neutron slowing down by writing the P 1 equations as
(8-151)
where the slowing down densities are given by the equations
aqf;'- H H
a;; +q0 (x,u)=~s cp(x,u),
2 aqr 4
3 ai; +qr(x,u)= g~~J(x,u), (exact treatment of H)
(8-152)
Note we have separated the treatment of hydrogenous (H) from nonhydrogeneous

(NH) slowing down. The various approximations we have discussed can then be
characterized by a particular choice of the coefficients Ao;, ;\Ji, f3o;, and /3li. We
have summarized these choices in Table 8-6 below: ·
TABLE 8-6: Choice of Parameters for Continuous Slowing Down Models
Ao; Ali /3o; /31;

Age approximation
Selengut-Goertzel 0 0 t~~i) 0
Consistent age
approximation 0 0 t~li) ri~\i>
Greuling-Goertzel
(s-wave scattering) Eq. (8-149) Eq. (8-150) ~;~(i) t~(i)
These equations can be rather easily solved on a digital computer provided the
spatial dependence can be simplified.
5. TREATMENT OF THE SPATIAL DEPENDENCE IN THE P 1 SLOWING

DOWN EQUATIONS
Recall that we are going to use the solution of the P 1 slowing down equations
to generate fast group constants. Hence we only need a crude method to account
for gross leakage effects. In particular we will assume that the lethargy and spatial
dependence of the flux are separable and furthermore characterize the spatial
dependence of each of the variables by a simple buckling mode. For example, the
flux would be written as
cf,(x, u) =cf,(u) exp (iBx), (8-153)
with similar expressions for current, source, and slowing down densities. That is, we
approximate the spatial dependence by a single Fourier mode, exp(iBx). (This
turns out to be somewhat more convenient than choosing a slab geometry mode
such as cos Bx since it results in different spatial variation for the flux and the
current when the real part of the solutions is taken.) Here the parameter B will
characterize the leakage in each region of the core in which the neutron spectrum
cf,(u) is to be calculated, and must be chosen from other considerations. For
example, if we were considering a bare, homogeneous reactor we know the flux
would indeed be separable, and B would be determined by
(8-154)
More generally, the flux will not be separable in rand u, but B 2 can be regarded as
characterizing the average core leakage. In the event that an accurate estimate is
required, one could first estimate B 2, calculate group constants, use these in a
multigroup diffusion calculation of the flux, and then use this flux to calculate a
new B 2 ~('v 2cp/cf,). This improved estimate of B 2 could then be used to regenerate
the group constants. Fortunately multigroup constants are rarely sufficiently sensi-
tive to the choice of B 2 to require such an iterative approach.
Assuming for the moment that we have specified B 2, we can then substitute the
separable forms Eq. (8-153) into the P 1 equations to arrive at

dqH dqNH;
iBJ ( u) + '2.ne( u )<t,( u) = - d~ - L ;u + S0 ( u ),
i
(8-155)
augmented by the set
(8-156)
and
Hence we now have arrived at a set of simple, first-order differential equations that
we can integrate numerically. We should recall here that inelastic scattering and
fission have been included in the source term as
N N
So( u) = ~
i=I
lu du' '2.\n( u' -u)<t,( u') + ~ X;( u) l
O i=I O
00
du' P;'i:.~( u')<t,( u') + Sext ( u).
(8-157)
6. AN ALTERNATIVE TREATMENT OF SPATIAL DEPENDENCE-THE

B 1 METHOD 15
Thus far we have considered approximating the spatial dependence of the
solutions to the P 1 slowing down equations by a single mode of the form exp(iBx).
There is an alternative scheme that bypasses the P 1 equations entirely, and for
completeness we will sketch its derivation here. To do so, we must return to the
steady-state transport equation, written in its one-dimensional form, Eq. (4-48), as
acp
µ,-a + '2.i( u)cp( x, µ,, u) =
f +I (u
dµ,' Jn du''i:..( u' -u, µ,' -µ, )cp( x, µ,', u') + S ( x, µ,, u ).
X -I 0
(8-158)
Now assume an angular flux and source spatial dependence of the form similar to
that in Eq. (8-153):
cp( x, µ,, u) = cp( µ,, u)eiBx,
S(u) .
S (x,µ,,u) = - -e•Bx_ (8-159)
2
(Note that we have assumed an isotropic source consistent with our earlier
assumptions concerning fission and inelastic scattering.) Then we find
+I lu du'Ls(u'-u,µ,'-µ,)cp(µ,',u')+ -S(u)- .
[Lt(u)-iBµ,]cp(µ,,u)=
f
-1
dµ,'
0 2
(8-160)
At this point, if we multiplied by 1 or µ, and integrated over µ, and then expanded

cp= 1/2<j>(u)+3/2µ,J(u), we would rederive the P 1 equations. Instead suppose we
first divide by (Lt - iBµ,) and furthermore assume only linearly anisotropic scatter-
ing to write:
(8-161)
Now multiply by 1 and µ, and integrate over µ, to find the coupled pair of
equations:
A ( u) [ {u
</>( u) = Li{ u) Jo du' Lso( u' --'?U )</>( u') + S0 ( u)
l
(8-162)
After some rearranging, one can rewrite these equations as
(8-163)
where
B 2 t -I ( B )
(
~ )
an ~ 4(B) (8-164)
y( u) = [B _1( B ) ] ~ l + 15 Lt .
3 --tan -
Lt Lt
Upon rewriting the scattering integrals in terms of slowing down densities, one can
recognize that the B 1 equations above are very similar to the P1 equations, except
for the appearance of the factory. For bare slab geometries, the B 1 equations are
found to yield slightly more accurate spectra. In fact, if only linearly anisotropic
scattering is present, the B 1 equations provide an exact treatment of angular effects.
However for more complicated geometries, as well as for large thermal reactor
cores in which leakage is relatively small, both methods yield very similar results.
The more recent tendency has been to generate group constants using the B 1
equations. However because of their simplicity (and mathematical similarity), we
will continue our discussion using the P 1 equations for the calculation of fast
neutron spectra.
IV. FAST SPECTRUM CALCULATIONS AND FAST GROUP

CONSTANTS
We now turn our attention toward the application of the P 1 and B 1 equations
to the calculation of fast neutron spectra suitable for the generation of fast group
constants to be used in few-group diffusion codes. We will study three different
schemes that can be used to generate fast group constants. The first such scheme
will be the numerical solution of the P 1 (or B 1) equations by discretizing these
equations on a fine group mesh (typically 50-100 groups). This scheme was
originally developed for the analysis of LWRs, and when applied to the generation
of group constants for hydrogeneous systems, it is known as the MUFT tech-
nique.16 For many years it and its various modifications such as the GAM codes 17
have been the mainstay of thermal reactor design. For this reason, and also
because the techniques used in the MUFT-GAM codes to solve the P 1 (or B 1)
equations are common to other fast spectrum codes, we will consider this scheme
in some detail.
We will also consider the calculation of fast reactor spectra in which inelastic
scattering is of primary importance and for which a direct solution of the P 1
equations over an ultrafine group structure is required. Then finally for com-
pleteness (and historical sentiment), we will briefly develop the analytical age-
diffusion model (although such a model has very little relevance to modern day
reactor calculations).
A. MUFT-GAM TYPE FAST SPECTRUM CALCULATIONS
Let us begin by considering the P 1 equations applied to describe neutron

slowing down in a mixture of hydrogen plus a single heavier isotope of mass
number A. (The modifications of the method we will describe to handle more
general mixtures of moderators is straightforward.) To be specific, we will utilize
the age approximation to describe slowing down from this heavy isotope coupled
with an exact treatment of hydrogen-that is, we will use the Selengut-Goertzel
description. It should be stressed that the procedure we will outline would apply
equally well to the other approximate treatments of slowing down that we have
discussed (e.g., the consistent P 1 approximation or the Goertzel-Greuling method).
The P 1 equations characterizing a single spatial mode can then be written as
iBJ+[Ls+LR+L-
a a m
dqfj
du
dqt
]<t>=----dU
+
lu du'Lt(u'-u)<t>(u')+x,
0
dqf! H H
du + qo = Ls </>, (8-165)
2 dqr H 4 H
3du+qi=9"'2.sJ,
qt=~AL~</>.
Here we have found it convenient to separate the removal cross section into three
parts:
"'2.: = "smooth" absorption cross section,
L~ = "resonance" absorption cross section,
00
Lin=£ du' Lt( u-u') = inelastic scattering cross section.
Remember that removal due to elastic scattering has already been accounted for in
the terms involving the slowing down density.
To solve these equations, we will now discretize them into multigroup form. We
will choose a fine group structure in the lethargy variables as shown in Figure 8-5.
We will label these fine groups with a subscript n to avoid confusion with the
coarse group structure denoted earlier by g.
The set of equations (8-165) will now be averaged over the fine group structure
in a manner very similar to the multigroup equations developed in Chapter 7. For
example, the group averages of dependent variables such as </>(u), J (u), or q(u) are
defined as:
(8-166)
Simple difference expressions can be used for derivatives:
(8-167)
The fine group constants characterizing smoothly varying (nonresonant) cross

sections are determined by simple averaging:
(8-168)
or
~A _
,;,n'n = D.u1 }
D.u '
f Un d
u
f Un• d ,~A( , )
u ,;,in u -u . (8-169)
n n Un-1 Un'-l
u = 0 10 MeV
..., C:
"'0
"'·-~
lL
;;:
1 MeV
C:
;;:
0
"O 0.1 MeV
5 .;
Cl
C:
-~
C: ...,0 0
0 0
·;:; "'e vi ...,a,
.!!! un-1
::, C. "'
0 co ::,
...,0 10 keV
co0 .;
.,E 2
Un }
~n+l .6.un Cl ·.;; "'
·.;; 0
..., > > "'e
E .,:
.;::. coC: coC:
"' ...,
...,2
0
a,
e "' lL
"'
a: a:"' ~
"'
.:l0 (.9
C.
...,"' 2
s:
...J
I-
I
...J
Cf) 1 keV
<(
t;
...."'
"' 10 C. e e ...J
::, ...,::, ...,::, e

-8 2 0
2
C:
0
·;:; 0
...,::,
e Cl
I t; C. 2 0
...,::,
0
0
;;: C.
::,
0
"'
..0
t;
C.
::,
.s"' 100 eV
...,2 I-
2Cl a,"' 2Cl C.
::,
"' 0
2
C.
::, !::, C: I Cl
2 0 "'0C: C:
a,
>
I
.
C:
Cl lL a, a,
I "'a, Cf) a,
C:
a: x 10 eV
i.i: ui
15
1 eV
co co
E §-
:;; 0
.,: ~
E
a, 0.1 eV
.,:
I- Cl
I-
20 0.02 eV
u
FIGURE 8-5. Typical few-group structm·es.
The group constants representing resonance absorption in the group can either be
calculated in terms of the resonance integrals characterizing resonances in the
group:
(8-170)
or in terms of the resonance escape probability Pn for the group by noting:
f u. du}:~( u)</>( u) = (1- Pn)( q~_, + qt._J

Un-1
(8-171)
Hence the multigroup form of the slowing down equations (8-165) can be written
as:
iBJi1un +L:n<Pnilun +Lin"<Pnilun + (1-pn)[ q~_, + q~_,]
n-1
= Xnilun + ilun ~ L~n<f>n,.lun' -{ q~ - qf/;._ 1)- {q~ - q~_J
n'= I
3iB <Pnilun + Lt,n Jn .l Un= - ( q HIn - q 'n-1

H )
'
{ q~ + q~_,)ilun/2 + { q~ - q~_,) = L~q>n.lun,

(8-172)
( qe + qe_ ,)ilun/2 + ¾(qe - qe_,) = ¾L~ Jnilun,
{q~ + q~_,)ilun/2 = (~ALtt <Pnilun, n=I, ... ,N
We have now arrived at a set of SN algebraic equations for the unknowns Jn, <Pn,
A To so1ve t h ese equations,
q0Hn, q H1n, andqon· · we b " by notmg
egm · q H = q H = q Ao=
00 10 00
(since no neutrons can slow down into the uppermost group). We can then solve
for J I and q> 1 using the P I equations (8-172) for n = I. Having determined these, we
then solve the remaining equations (8-172) for q0H,I qf,I and q0A.I We can proceed in
this manner, solving successively for the variables characterizing increasingly
higher lethargy groups (i.e., increasingly lower energy groups).
Once we have obtained <Pn, we can then use it to generate few-group constants.
For example, if we utilized a 54-group MUFT calculation to generate one-fast-
group constants, we would find
(8-173)
or
DI= --54
_ _ __ (8-174)
~ <Pnilun
n=I
Sometimes an alternative scheme is used to calculate the group-diffusion

coefficients. This procedure comes from requiring that Dg be defined such that for
a group of lethargy width .lu
J duJ(x,u)= -ng:Jx J du<t>(x,u).

A~ A~
(8-175)
If we use our single buckling mode forms for J (x, u) and <P(x, u), we find that Dg is
then defined by
54
~ Jn.lun
J n= I
Dg= iB 54 (8-176)
~ <Pnilun
n=I
In summary then, MUFT-GAM-type schemes generate a fast neutron spectrum

and calculate few-group constants by using the P 1 (or B 1) slowing down equations,
approximating the spatial dependence by a single spatial mode characterized by an
"equivalent bare core buckling" B 2, treating elastic scattering by a continuous
slowing down model (an exact treatment for hydrogen and either the age
approximation or the Goertzel-Greuling model for materials with mass number
A > 1), and handling inelastic scattering using a multigroup transfer matrix. We
have not dwelt in detail on the treatment of resonance absorption, because such
calculations will usually correct the infinite medium resonance escape probabilities
we considered earlier for fuel lumping (heterogeneous) effects. We will return to
discuss this in some detail in Chapter 10.
Most typically the MUFT scheme is used to generate either one-fast-group or
three-fast-group constants for LWR calculations with the group structure as shown
in Figure 8-5 (we have also indicated the thermal group, which must be handled by
methods to be discussed in the next chapter).
An alternative to the MUFT fast spectrum procedure is that utilized in the GAM
code and its offspring, in which a direct fine-group solution of the slowing down
equation (in the B 1 or P 1 approximation is solved). Typically of the order of 100
lethargy groups are used, varying in width Au from 0.1 to 0.25. In most respects
(aside from its direct multigroup treatment of elastic scattering), GAM spectrum
codes are very similar to MUFT codes and are commonly used to generate fast
group constants for HTGRs.
B. Fast Group Constant Generation for Fast Reactor Calculations

The generation of few-group constants for fast reactor applications involves
much more in the way of direct or brute force techniques than comparable thermal
reactor calculations. To a large degree this is because of the relatively limited
experience in fast reactor design and hence the desire for as accurate a calculation
as possible. It also arises, however, because of the increased complexity of the
variety of nuclear processes that influence a fast reactor spectrum-particularly the
effects of resonance absorption and inelastic scattering. In a thermal reactor, the
strong elastic scattering present in the system and the fact that most fissions are
induced by thermal neutrons will tend to wash out many of these details. However
in a fast reactor essentially all fissions are induced by neutrons with energies in just
those ranges in which inelastic scattering and resonance absorption are most
important.
At the present time, the most general class of codes for generating few-group
constants for fast reactor calculations includes the MC2 code and its variants. 18
This code begins with evaluated cross section data (ENDF /B) and proceeds to
perform a sequence of group averages from ultrafine groups with Au~ 1/ 120, to
fine groups with Au~ 1/ 4, and finally to broad groups with Au~ 1/2 to 1. In the
more advanced versions of this code, a direct multigroup spectrum calculation is
performed at higher energies, whereas a continuous slowing down calculation is
used to handle elastic scattering at lower energies. The continuous slowing down
calculation uses an improved Goertzel-Grueling approximation similar to that we
discussed in the previous section. Spatial dependence is handled in either the P 1 or
B I approximation with again a single spatial mode characterizing leakage with an
effective buckling B 2 • A number of rather sophisticated techniques are used to
handle resonance absorption-particularly for the overlapping or unresolved res-
onance regions [the isolated resonances are most typically handled by the narrow
resonance (NR) approximation].
As we have mentioned, codes such as MC2 are much too elaborate for daily use
in reactor design and are intended more for the evaluation of calculational
techniques and comparison with critical experiments. As more experience is
acquired in the design of cores for commercial fast reactors, less elaborate
(although sufficiently accurate) techniques will evolve in a manner similar to the
MUFT-GAM schemes used in thermal reactor design.
We have noted that in thermal reactors the multigroup constants are sensitive
functions of core composition and core operating conditions and hence must be
recalculated many times in a core design. In fast reactors the expense of a direct
calculation of the multigroup constants coupled with a somewhat lower sensitivity
to core environment has motivated the development of sets of universal micro-
scopic multigroup constants which can be used for any fast reactor core design.
The most well-known of such sets of fast group constants include the YOM set, 22
the Hansen-Roach or LASL 16-group set, 23 and the Bondarenko or ABN set. 24
The effects of core environment on the multigroup constants are then taken into
account by using so-called "j-factors"25 (sometimes also referred to as Bondarenko
self-shielding factors) which are defined in such a way that if aeff is the group
constant characterizing the particular core of interest, while aav is the universal
group constant, then
Here, aav would essentially correspond to the infinite dilution cross section (similar
to the infinite dilution resonance integral). The set of universal microscopic group
constants is typically calculated using a very simple intragroup flux-for example,
a 1/ E collision density. The j-factor then takes into account the detailed in-
tragroup flux behavior due to resonance self-shielding, for example. The j-factors
are usually tabulated as functions of concentration and temperature and included
with the cross section sets.
C. Age-Diffusion Theory
For historical completeness, let us return to consider the age-diffusion equa-
tion in a bit more detail. Recall that this equation takes the form
oq
- D ( u) v' 2cj>(r, u) + ~a( u)cj>(r, u) = - au+ S (r, u) (8-177)
where
q(r, u) =:: ~~.( u)c/>(r, u). (8-178)
Now we know that this equation is only valid for neutron slowing down in
moderators with reasonably large mass numbers (such as graphite). Nevertheless,
age-diffusion theory has been thoroughly studied in the past because one can
obtain an explicit solution of this equation in certain simplified cases.
For convenience, we will first consider the solution of the age-diffusion equation
in the absence of a source. If we then substitute Eq. (8-178) into (8-177), we find
(8-179)
Let us introduce an integrating factor to eliminate the absorption term by defining

a new dependent variable, q(r,u):
(8-180)
Then if we substitute this form into Eq. (8-179), we find
(8-181)
To solve this equation, we next make a change of independent variable by defining
(8-182)
and note
a a-r a D ( u) a
au au
= a-r = tL.( u) a:;:- ·
(8-183)
If we now use this new independent variable in Eq. (8-181), we find that q(r,-r)
satisfies
(8-184)
Notice that the form of this equation is very similar to that of the time-dependent
diffusion equation. For this reason, -r is known as the Fermi age in analogy to the
time variable t even though its units are cm2 rather than sec.
By noting the similarity between the age equation (8-184) and the time-
dependent diffusion equation, we can adapt our entire earlier analysis to solve this
equation. We won't do this here, however, because age theory is primarily of
historical interest and is only of limited utility in present-day reactor design. We
will, however, generalize our results to include absorption by noting
q(r,-r)=q(r,-r)exp -
A [ f0
U , La( u') ] -
A
du lL.(u') =q(r,-r)p(u), (8-185)
where we have identified the exponential factor as just the resonance escape
probability p(u) to lethargy u.
To illustrate these ideas more clearly, let us consider the example of an infinite
planar source emitting neutrons of energy E 0 at the center of a nonabsorbing slab
of width a (as we have seen, absorption can be easily included) [see Figure 8-6]. We
will define the lethargy variable with respect to the source energy E 0 so that u = 0
corresponds to the source energy. If we also note
(8-186)
FIGURE 8-6. The slowing down density q(x,r),resulting from a monoenergetic plane source at
the origin of a slab.
then we can write the appropriate form of the age equation as
--=-
ax2 a-r (8-187)
subject to: (a) initial condition: q(x,0) = S8 (x)

(b) boundary conditions: q( - ti/2,-r) = 0= q( ti/2,-r)
To solve this, we will try an eigenfunction expansion (noting by symmetry that only
even eigenfunctions need be used)
q(x,-r)= Lfn(-r)cosBnx, (8-188)

n
If we substitute this into the age equation
(8-189)
and use orthogonality we find
(8-190)
The solution to this equation is just
(8-191)
Hence the general solution to Eq. 8-187 is
q(x,T)= ~Anexp(-B;T)cosBnx. (8-192)

n
Now using the initial condition at T = 0, we require
q(x,0)= ~AncosBnx= SS(x). (8-193)

n
In the usual manner we can now multiply by cosBmx, integrate over x, and use
orthogonality to evaluate all An to find the solution
(8-194)
Just as the fundamental mode (n = 1) will dominate the solution of the time-
dependent diffusion equation, we find that for large values of T, that is, for neutron
energies far below the source energy E 0 , the solution of the age-diffusion equation
approaches
(8-195)
This is a very useful expression, since it essentially describes the spatial distribution
of neutrons slowing down to thermal energies in a bare, ~omogeneous assembly
from fission sources. Since there is no absorption in our modeled problem, we can
interpret the decrease in q(x,T) with increasing age as being due to neutron leakage
while slowing down.
Note in particular that if all of the source neutrons had initially been distributed
in the fundamental mode cosBgx shape (which would correspond, for example, to
the appearance of fission neutrons in a bare, critical slab reactor), then the loss of
neutrons during slowing down would be described by the exp( - B;T) factor. That
is, we can identify the fast nonleakage probability PFNL introduced earlier (Chapter
3) in the six-factor formula for k as just
(8-196)
where T 1h is the age corresponding to thermal energies (usually taken as ~ 1 eV):
(8-197)
Of course this expression for PFNL is valid only in the age approximation as applied
to a bare, uniform reactor.
The age-diffusion equation can also be solved in several other standard
geometries. For example, the slowing down density resulting from a plane source in
an infinite, nonabsorbing medium is
(8-198)
while that resulting from a point source in such a medium is
2
Sexp( ~; )
q (r r) = - - - - - (8-199)
pt ' (47rr)3/2
This last result can be used to demonstrate that
2
r= ¼<r )= 1/6 the average (crow-flight distance)2
from the point where a neutron enters
a system with £ 0 to the point at which
it slows down to an ager. (8-200)
One usually calculates or measures the age-to-thermal energies r 1h or more com-

monly, the age-to-indium resonance at 1.45 eV, ,,.IN (since this is the most common
experimental measurement). In Table 8-7, we have listed the age r 1h for several
common moderators, along with other parameters characterizing thermal neutron
behavior in these moderators.
TABLE 8-7: Diffusion Parameters for Some Common Moderators
Moderator Density (g/ cm3) D(cm) ~.(cm- 1) L(cm) -r(cm2) M(cm)

H 2O 1.00 0.16 0.0197 2.85 26 5.84
D 2O 1.10 0.87 2.9x 10- 5 170 131 170
Be 1.85 0.50 I.Ox 10- 3 21 102 23
Graphite 1.60 0.84 2.4x 10- 4 59 368 62
Diameter Diameter
Reactor Type (L)(cm) -r(cm) (M)(cm) (M) (L)
PWR 1.8 40 6.6 56 190
BWR 2.2 50 7.3 50 180
HTGR 12.0 300 21 40 63
LMFBR 5.0 35
GCFR 6.6 35
If we recall that the thermal nonleakage probability is PTNL =(1 +L 2B;)- 1, and
note further that for most larger power reactors,
2 2
B;
is sufficiently small that both
L B; and rBg are much less than one, then we can write the total nonleakage
probability as
(8-201)
where we have defined the neutron migration area M 2 as
(8-202)
The migration length M can be interpreted as 1/ V6 of the rms distance a neutron

travels from its appearance as a fast fission neutron to its capture as a thermal
neutron (see Problem 8-42). We have listed the migration lengths for typical
moderators in Table 8-7. The migration length proves useful as a scale with which
to characterize the size of a reactor core. For example, a modern LWR is
typically~50 Min diameter, whereas a comparable HTGR core is about 40 M. In
this sense, even though the physical size of the HTGR is two to three times that of
a L WR, its "neutronic" size is actually somewhat smaller. Both reactor types are
certainly far more loosely coupled in a neutronic sense than fast reactor cores ( ~30
L) and hence are more susceptible to spatial power transients.
D. Some Additional Comments on Fast Spectrum Calculations

As we have seen, most schemes for generating fast group constants first
perform a fine multigroup solution of the neutron slowing down equations (in the
P 1 or B 1 approximation) to obtain the fast neutron spectrum, and then average
microscopic cross section data over this spectrum to obtain the desired few-group
constants. Microscopic cross section data enter this calculation in two different
ways. First, the microscopic cross sections available from nuclear data files such as
ENDF /B must be converted into fine multigroup constants in order to allow the
calculation of the fast neutron spectrum. Then these data must be averaged over
this spectrum to generate the few group constants themselves.
Such basic nuclear data are usually given in the form of pointwise data along
with various interpolation specifications in a data set such as ENDF /B. Hence the
first step in a fast spectrum calculation is to prepare a library of fast fine-group
constants by averaging these basic cross section data over the groups to be used in
the fast spectrum calculation. Usually either a 1/ E or a fission spectrum is used as
a weighting function. These fast libraries are typically supplied as part of a given
fast spectrum code, since their generation from a basic nuclear data set can be
rather expensive. A schematic of this procedure is shown in Figure 8-7.
A rather important aspect of such cross section libraries involves the preparation
of resonance cross section data. In the resolved resonance region, the techniques
we have discussed (e.g., the NR or NRIM methods) can be used to calculate the
resonance integrals characterizing the resonances in each of the fine groups.
However in the region of unresolved resonances more elaborate schemes must be
used. Although a rather crude treatment of the unresolved resonance region is
usually sufficient in thermal reactor calculations, in a fast reactor as much as 40%
of the resonance absorption can occur in unresolved resonances, and hence a
detailed calculation is required.
THERMOS
MUFT-
GAM [:] Monte
Carlo
FIGURE 8-7. Generation of cross section libraries from ENDF /8.
In most fast spectrum codes, one has the option of specifying the group structure
characterizing the few-group constants that he wishes to generate. In general one
wishes to use the smallest number of coarse groups consistent with the desired
accuracy in the corresponding multigroup diffusion calculation. In fact, the
necessary group structure will usually depend on the type of calculation of interest.
For example, the calculation of integral quantities such as the core multiplication
factor can usually be accomplished with fewer groups than an estimate of the
power-peaking in a fuel assembly.
The lower cutoff of the fast spectrum is usually chosen high enough so that
upscattering out of the thermal groups can be neglected. In a LWR, this cutoff is
typically chosen at around 0.6 eV. However in an HTGR, it must be chosen
somewhat higher (typically 2.38 eV) because the higher moderator temperature in
such reactors causes appreciably larger upscattering. The remaining group structure
is usually chosen such that different physical phenomena tend to be isolated within
a given group. For example, in a three-fast-group calculation, the highest energy
group is usually chosen to contain all of the fission spectrum, the middle group is
characterized by elastic slowing down, and most of the resonance absorption is
confined to the lowest energy group. (See Figure 8-5).
A fast spectrum calculation not only supplies fast few-group constants and the
fast neutron spectrum, but also will provide the rate at which neutrons slow down
into the thermal energy range. This latter information is needed to complete the
determination of few-group constants by generating a thermal neutron spectrum
and then averaging cross section data over this spectrum to obtain thermal group
constants.
REFERENCES
l. J. H. Ferziger and P. F. Zweifel, The Theory of Neutron Slowing Down in Nuclear

Reactors, M.I.T. Press, Cambridge (1966).
2. M. M. R. Williams, The Slowing Down and Thermalization of Neutrons, North-Holland,

Amsterdam ( 1966).
3. Ibid., Chapter VIII.
4. A. Radowsky, (Ed.), Naval Reactors Physics Handbook, Vol. I, Selected Techniques,
USAEC Report (1964), Chapter 2.
University of Chicago Press (1958), p. 290.
6. J. H. Ferziger and P. F. Zweifel, The Theory of Neutron Slowing Down in Nuclear
Reactors, M.I.T. Press, Cambridge (1966) pp. 73-78; R. E. Marshak, Rev. Mod. Phys. 19,
185 (1947).
7. G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand, Princeton, N.J.,
(1970), pp. 439-443.
8. P. E. Meyer and P. F. Zweifel, Trans. Am. Nucl. Soc. 13, 305 (1970).
9. M. E. Rose, W. Miranker, P. Leak, and G. Rabinowitz, A Table of the Integral it,(x,t),
BNL-257 (T-40), (1953); WAPD-SR-506 (1954); K. K. Seth and R.H. Tabony, USAEC
Document TID-21304 (1964).
10. L. W. Nordheim, Nucl. Sci. Eng. 12, 457 (1962).
11. Y. lshiguro, Nucl. Sci. Eng. 49, 526 (1972); R. Goldstein, Nucl. Sci. Eng. 49, 526 (1972).
12. K. T. Spinney, BNL-433 (1960).
13. P. E. Meyer, Approximate Methods for the Calculation of Resonance Absorption,
University of Michigan Ph.D. Dissertation, 1974.
14. G. Goertzel and E. Greuling, Nucl. Sci. Eng. 7, 69 (1960).
15. H. Hurwitz and P. F. Zweifel, J. Appl. Phys. 26, 923 (1956).
16. H. Bohl, Jr., E. M. Gelbard, and G. H. Ryan, MUFT-4, A Fast Neutron Spectrum
Code, Report WAPD-TM-22 (1957).
17. G.D. Joanou, E. J. Leshan, and J. S. Dudek, GAM-1 a Consistent P 1 Multigroup Code
for the Calculation of Fast Neutron Spectra and Multigroup Constants, General Atomic
Report GA-1850 (1961).
18. B. J. Toppel, A. L. Rago, and D. M. O'Shea, MC2 A Code to Calculate Multigroup
Cross Sections, ANL-7318 ( 1967).
19. J. R. Lamarsh, An Introduction to Nuclear Reactor Theory, Addison-Wesley, Reading,
Mass. (1966).
20. M. Lineberry and N. Corngold, Nucl. Sci. Eng. 53, 153 (1974).
21. R. Bellman and K. Cooke, Differential-Difference Equations, Academic, New York
(1962).
22. S. Yiftah, D. Okrent, and P. A. Moldauer, Fast Reactor Cross Sections, Pergamon, New
York (1960).
23. I. I. Bondarenko, et. al., Group Constants for Nuclear Reactor Calculations, Consultants
Bureau, New York (1964).
24. Reactor Physics Constants, 2nd Ed., USAEC Document ANL-5800 (1963), pp. 568-579.
25. M. Segev, Nucl. Sci. Eng. 56, 72 (1975).
PROBLEMS
8-1 The lower cutoff energy Ec for the fast region is usually chosen such that there is
negligible upscattering above this energy. If the moderator is modeled as a free proton
gas [cf. Eq. (2-107)], compute the cutoff energy Ec such that less than 0.1% of the
thermal region neutrons will be upscattered above Ec. Plot this cutoff energy versus
moderator temperature T.
8-2 Determine the neutron flux </>(E) resulting from an arbitrary source in an infinite
hydrogeneous medium by: (a) solving the infinite medium slowing down equation
with this general source term, and then (b) using the solution obtained for a
monoenergetic source as a Green's function for the more general problem.
8-3 (a) Derive an expression for the neutron balance in a nuclear reactor in terms of the
scalar flux <j>(r, E, t), and the net current J(r, E, t ).

(b) Derive an expression for the neutron slowing down density q(r, E, t) in terms of
<j>(r, E, t).
(c) Prove that in a purely hydrogeneous medium and in the steady state one may
write
where S represents the sources. You may use the lethargy variable u rather than
the energy E if you prefer. Throughout, assume E»kT.
8-4 Transform the infinite medium slowing down equation (assuming only elastic scatter-
ing) from the energy variable to the lethargy variable.
8-5 Repeat the derivation of an expression for the neutron slowing down density in an
infinite moderating medium by first calculating P (E' > E), the probability that a
neutron with energy E' scatters below an energy E in a collision. Then integrate the
product of this quantity with the collision density L. 5 (£')</>(E') over all E'> E to find
q(E).
8-6 Show that the neutron continuity equation can be written quite generally in terms of
the slowing down density q(r, E, t) as
8-7 Solve the coupled slowing down equations for an infinite hydrogeneous medium
dq dq
La(u)<j>(u)= - du +S(u); du +q(u)=L.(u)<j>(u) for a monoenergetic source at leth-
argy u=0.
8-8 Calculate the mean lethargy loss per collision ~, the moderating power, and the
moderating ratio for H 2 O and D 2O at energies of 1 eV and 100 eV. Estimate the
necessary cross section data from BNL-325.
8-9 Consider neutrons slowing down in an infinite medium with no absorption.
(a) By equating the number of collisions in a lethargy interval du with the number of
collisions in a time dt, obtain an expression for the time required for neutrons to
slow down to thermal energies.
(b) Calculate the time required for fission neutrons to slow down to thermal in
graphite and water. Assume L; = 0.385 cm_, while L.~ 20 = 0.528 cm - 1 ( constant
in energy).
2
8-10 Demonstrate that for large mass numbers, ~-A+ .
213
8-11 Show that the average increase in lethargy ~ in an elastic (s-wave) collision is one-half
the maximum lethargy increase.
8-12 Calculate the multigroup transfer scattering elements Lsg'g for hydrogen under the
assumption that L 5(u) and the intragroup flux <j>(u) can be taken as constant. Assume
the groups are equally spaced with a lethargy width l:lu.
8-13 Demonstrate that the multigroup transfer scattering elements characterizing elastic
scattering for directly coupled groups of lethargy width !:lug are given by Lsg---+g+ 1
= Lsg~g/ !:lug (Hint: Balance the scattering rates into and out of each group.)
8-14 As an example of a nondirectly coupled multigroup calculation, consider a scheme m
which ail groups are of the same lethargy width l:lu and in which neutrons are able to
skip at most one group. Assume that only s-wave elastic scattering need be considered
and that the scattering cross section L 5(u) can be treated as lethargy-independent.
Calculate the elements of the scattering transfer cross section Lsg'g for all g' and g,
assuming a constant intragroup flux <j>(u)=<j>. Also verify that the sum of these
transfer cross sections is indeed equal to the group-averaged scattering cross section
Lsg·
8-15 Determine the neutron flux cp(E) resulting from a monoenergetic source at energy E0
in an infinite hydrogeneous medium if the scattering cross section is taken to be
constant and the absorption cross section is inversely proportional to the neutron
speed (a "1/v absorber").
8-16 Derive the expression for the energy spectrum of the flux in the first collision interval
of a moderator of mass A which has an arbitrarily -varying scattering cross section
L 8 (E) and an absorption cross section La{E). Assume that the medium is infinite and
homogeneous, that q0 neutrons are introduced per sec per unit volume at an energy
E 0 , and that La{E0)=0.
8-17 Solve the infinite medium slowing down equation characterizing slowing down from a
monoenergetic source in a medium of mass number A in the absence of absorption by
using Laplace transforms. In particular, determine the asymptotic behavior of the
solution for large lethargy u.
8-18 Calculate the resonance escape probability in hydrogeneous media for finite dilution
but at zero temperature. Show explicitly that this expression reduces to the infinite
dilution result in the limit as NA a0/ NHa:i-o.
8-19 An infinite homogeneous medium consists of a single isotope that scatters neutrons
isotropically in the CM system with an energy-independent scattering cross section L 8 •
Assume that the macroscopic absorption cross section is zero for energies E > E I and
that La= constant for E < E 1• If a source emits S0 neutrons per cm3 per sec at some
energy E 0 where aE0 < E 1 < E 0 , find the collision density F(E) of neutrons in this
collision interval.
8-20 Consider the slowing down of neutrons in an infinite, homogeneous mixture of
hydrogen (number density NH and scattering cross section o:1) and an infinite mass
absorber (number density NA and capture cross section oy). We will model a
resonance of the absorber centered about an energy E I by assuming that oY = constant
= oyo/ aT for E 1 - bT < E < E 1 + bT and zero elsewhere, where a, b, and <Iyo are
constants and Tis the absolute temperature. The resonance energy E 1 is well below
the neutron source energy and 2bT (resonance width)«E 1• The scattering cross
section of hydrogen o:1 can be assumed constant in the vicinity of the resonance.
(a) Calculate the resonance escape probability p 1 and show that p 1 decreases as the
temperature T increases.
(b) Show that for the case of infinite dilution (NAoy0 «NHo:1) the resonance escape
probability p 1 is independent of temperature.
8-21 Consider a medium in which the average lethargy gain ~ and the scattering cross
section L 8 are independent of energy. A common expression for p(E) is
p(E)=exp [ - JE
(Eo La(E')dE'
~[L.+L/E')]E'
l .
When, if ever, is this expression (a) exact, (b) a good approximation, and (c) a poor
approximation?
8-22 Assume that the resonance region consists of two widely spaced resonances at E 1 and
E 2, such that E 1 - E 2»aE 1• Within !::,.E 1 and /::,.£2, La/L,= oo, and everywhere else
La/L, = 0. What is the overall resonance escape probability if:
(a) E 1 -aE 1»/::,.E 1, and E 2 -aE2»/::,.E2;
(b) E 1 -aE 1 =/::,.E 1, and E 2 -aE2=/::,.E2.
8-23 Compute the resonance escape probabilities for a NH/ Nu= 1.0 mixture of 238 U and
hydrogen at temperatures 0°K and 500°C for the 6.67 eV, 21 eV, and 208 eV
resonances of 238 U as given in Table 8-2.
8-24 (a) Develop an accurate equation for the resonance escape probability for an infinite
homogeneous mixture of nonabsorbing moderator of mass A and nonscattering
absorber whose absorption cr9ss section is infinite over the range £ 1 to £ 2 and
zero at all other energies. The resonance region is several orders of magnitude
below the source energy.
(b) What is the resonance escape probability when £ 2 = aE 1?
8-25 Calculate the escape probabilities for T= 300°K, 600°K, and 1200°K for both the
6.67 eV and 208 eV resonances in 238 U, assuming that ~;-1 /NA= IO and 1000 b per
238
U atom. Use the NRIM approximation for the lower resonance and the NR
approximation for the higher resonance. (Ignore interference scattering terms.)
8-26 Estimate the probability that a 2 MeV neutron emitted in an infinite medium of water
will not be absorbed while slowing down to 1 eV.
8-27 Demonstrate explicitly that for any capture resonance, a1 / aT > 0 and hence ap / aT
< 0. (Use either the NR or NRIM expression for the effective resonance integral.)
8-28 Derive the equations (8-132) for the slowing down densities qft-(x,u) and qr(x,u) in
hydrogen.
8-29 Show that a/ g~ - g by plotting both parameters as functions of a for 0 <a< I.
8-30 Explicitly derive the expressions for ;\0 , ;\ 1, /3 0 , and /3 1 in the Goertzel-Greuling
approximation.
8-31 Derive the B 1 equations by explicitly calculating the first two angular moments of the
transport equation as described in Section 8-III-B and then rearranging these results
into the form of Eq. (8-163).
8-32 Using the group-averaging schemes discussed in Section 8-IV, develop in detail the
fine multigroup slowing down equations (8-172) from the set (8-165).
8-33 Determine the following quantities in terms of the fine-group MUFT representation:
(a) source rate, (b) slowing down rate to thermal, (c) leakage rate, and (d) absorption
rate. Then calculate the slowing down probability, leakage probability, and absorp-
tion probability. Verify that these latter probabilities all add to unity.
8-34 Consider an infinite slab of thickness L composed of pure atomic hydrogen with a
number density NH. The slab contains an external source of the form S (x, u)
=x(u)exp(iBx) where x(u) is the 235 U fission spectrum and x=O denotes the
midplane of the slab.
Write a computer program that will solve the P 1 equations for this problem. Your
program should tabulate <f>(u), J(u), q0(u), and q 1(u); it should also calculate the fast
leakage and absorption probabilities and the age to thermal. Run your program for
several slab thicknesses and compare the results.
Use the MUFT 54 group scheme and a pointwise representation of the fission
spectrum as given by Eq. (2-112), and use BNL-325 to obtain a pointwise representa-
tion of ~~(u).
8-35 Determine the slowing down spectrum of fission neutrons slowing down in an infinite
medium of water using a MUFT-type fast spectrum code. Repeat this calculation for
an infinite medium of D 20.
8-36 Determine the slowing down density established by a monoenergetic plane source at
the origin of an infinite moderating medium as given by age-diffusion theory.
8-37 Determine the slowing down density resulting from a point source in an infinite
moderating medium (using age-diffusion theory).
8-38 Sources of monoenergetic fast neutrons are distributed in a moderating slab as
S(x)=S0 cos('1Tx/a), where a is the width of the slab. Using age-diffusion theory,
determine the slowing density in the slab. Then calculate the average probability that
a source neutron leaks out of the slab while slowing down.
8-39 Age theory for hydrogen fails principally because the collision density may not be
slowly varying over the limits of the scattering integral. Why not?
8-40 Consider a "zero-temperature" reactor, namely, a reactor in which all nuclei are at
rest and thus slowing down theory is valid for all lethargies. Assume that the reactor is
bare with a geometric buckling B;, and that all fission neutrons are born with zero
lethargy. Then using age-diffusion theory, derive the criticality relation for this reactor
in the form:
00
v(u)~c(u)
1=
1
0
du ( ) p(u)exp[ - B;T(u) ]-
~~tu
8-41 Estimate the fast and thermal nonleakage probabilities for each of the reactor types
listed in Table 8-7. Use core sizes from Appendix H.
8-42 Demonstrate that M 2 = L 2 + T is indeed 1/ 6 the mean-square distance a neutron will
travel from its origin in fission to its eventual absorption as a thermal neutron.
8-43 Consider an effectively infinite medium in which a monoenergetic pulse of fast
neutrons is instantaneously injected at time t = 0 uniformly throughout the medium.
Using age theory, determine the time-dependent slowing down density q(u,t).
9
Thermal Spectrum Calculations
and
Thermal Group Constants
We now turn our attention to the generation of multigroup constants characteriz-

ing low-energy neutrons. Once again the general approach will be to develop
methods for determining the detailed energy dependence or spectrum of such
neutrons in those situations in which the spatial dependence can be ignored (or
treated in a crude manner) and then to average microscopic cross section data over
this energy spectrum.
A detailed investigation of the neutron energy spectrum below several eV
becomes quite involved due to the complicated nature of the neutron scattering
process. For at such low energies we can no longer ignore the thermal motion of
the nuclei,. as we did in the study of neutron slowing down. The energy of such
slow or thermal neutrons is comparable as well to the binding energy of the atoms
in molecular or crystalline materials, and hence the neutron will tend to interact
with an aggregate of atoms rather than with a single nucleus. These features greatly
complicate the determination of cross sections characterizing the scattering of
thermal neutrons, and such cross sections will exhibit a rather involved dependence
on both neutron energy and scattering angle. Moreover thermal neutron cross
sections depend, in a detailed manner, on the temperature and physical state (i.e.,
solid, liquid, or gas) of the scattering medium, unlike fast neutron cross sections,
which depend primarily on the nuclear species. The complicated nature of such
cross sections renders the determination of thermal neutron spectra rather in-
volved, and one usually must resort to direct numerical methods.
The study of thermal neutron behavior is customarily referred to as neutron
thermalization. 1-4 Actually, however, the subject of neutron thermalization can be
classified into two separate problems: (a) the calculation of cross sections
characterizing thermal neutron scattering in various materials and (b) the use of
these cross sections in the determination of the energy spectrum characterizing
375
low-energy neutrons (e.g., for use in determining thermal group constants for
few-group diffusion calculations). These subjects are quite interrelated since those
features characteristic of thermal neutron cross sections may have a rather signifi-
cant influence on the determination of thermal spectra.
For example, the fact that the kinetic energy of a low-energy neutron is
comparable to the thermal energy of atomic motion means that microscopic
thermal neutron cross sections actually must be regarded as averages over the
thermal distribution of nuclear speeds and hence are temperature-dependent, as
we saw in Chapter 2. Furthermore, it will be possible for the neutron to gain
energy-that is, to upscatter-in a scattering collision with a moving nucleus. This
will complicate the numerical solution of the fine-structure multigroup equations
used to determine thermal spectra.
Of similar importance is the fact that the energy of thermal neutrons is compar-
able to the chemical binding energy of the scattering nuclei (e.g., in a molecule or a
crystal lattice). Hence the nucleus will no longer recoil freely, and thus binding will
become significant in determining the energy and angle change of a neutron in a
collision. In fact one can show that the scattering cross section of a bound nucleus
is somewhat larger than that of a free nucleus by an amount that is roughly
2
abound= (l + 1/ A)2afree' where A is the mass number of the scattering nucleus. For
example, in hydrogeneous materials for which A= 1, we would find that the actual
low-energy scattering cross section is some four times larger than the free-atom
scattering cross section afree·
Of related significance are inelastic scattering processes in which the internal
states of the scattering system (e.g., molecular vibration and rotation or crystal
lattice vibration) are excited by neutron scattering collisions. (Such thermal in-
elastic scattering processes should not be confused with nuclear inelastic scattering
in which the nucleus itself is excited into a higher quantum state. The latter process
is of little concern for the low energies characterizing thermal neutrons.) Inelastic
scattering gives rise to a complicated cross section dependence on energy and
angle.
For very low energies, the neutron wavelength is comparable to the interatornic
spacing of the scattering material. Hence the neutron wave function experiences
diffraction effects Gust as light is diffracted). However, we should add that while
such coherent interference effects can be quite important in determining the very
low energy behavior of neutron scattering cross sections, they are rarely of
importance in nuclear reactor behavior.
Needless to say, such considerations greatly complicate the subject of neutron
thermalization. The determination of thermal neutron cross sections is extremely
complicated and can be regarded as essentially a subfield of statistical mechanics
and solid and liquid-state physics. Since the measurement and calculation of such
cross sections can infer information about the microscopic structure and dynamics
of materials, thermal neutron scattering (or thermal neutron spectroscopy) has
received an exhaustive treatment in the scientific literature, and we will not attempt
to duplicate that treatment in this text.4--6
Fortunately most of the complicated details of thermal neutron cross section
behavior are of secondary concern in nuclear reactor analysis. Indeed in most large
thermal power reactors the neutron energy spectrum is sufficiently well
"thermalized" that rather crude models of the neutron scattering process are
sufficient for the generation of thermal group constants. For example, in LWR
THERMAL SPECTRUM CALCULATIONS AND THERMAL GROUP CONSTANTS / 377
calculations, the core can frequently be modeled as an ideal gas of protons (totally
ignoring binding and diffraction effects) seeded with a uniformly distributed
absorber corresponding to the fuel. Such a model is particularly useful for generat-
ing the thermal group constants to be used in survey or parametric studies (e.g.,
core lifetime studies). Of course, more detailed studies of such large, heterogeneous
reactor cores characterized by strong temperature gradients require more elaborate
models of the scattering material. This is particularly true for the analysis of
reactors utilizing solid moderators, such as the HTGR, which uses graphite as a
moderator. However, in all cases the models of the physical and chemical structure
of the scattering material used to calculate thermal neutron cross sections for
nuclear reactor analysis are very crude indeed when compared to the sophisticated
theories and measurements that exist for the interpretation of thermal neutron
scattering data.
Hence our concern in this chapter will be to first illustrate some of the simple
ideas involved in neutron thermalization and then to develop those tools forming
the basis for thermal spectra calculations and used to generate thermal group
constants.
I. GENERAL FEATURES OF THERMAL NEUTRON SPECTRA
A. Thermal Equilibrium
Although the detailed form of the cross section characterizing thermal
neutron scattering is extremely complicated, depending as it does on the tempera-
ture and physical structure of the scattering material, there are several simple
general features of such cross section behavior with important implications for
thermal neutron spectra. Suppose we once again consider a situation in which we
imagine sources of neutrons distributed uniformly throughout an infinite medium
such that spatial and time dependence can be ignored. Then we recall that the
neutron continuity equation (4-79) simplies to
Notice that we have allowed the range of integration to extend to E=O, since we
now wish to analyze the low energy behavior of the neutron flux.
The first useful property of thermal neutron cross sections-and indeed, all
differential scattering cross sechons-is essentially just a definition:
(9-2)
However even this familiar property tells us something interesting. For if we

integrate Eq. (9-1) over all energy and use the relation Eq. (9-2), we find the
balance relation
(9-3)
That is, for a steady-state flux </>(£) to be possible in an infinite medium, we

require that the rate at which source neutrons appear be just equal to the rate at
which neutrons are absorbed (since there is no leakage). If we recognize that in
thermalization problems all source neutrons will appear as fission neutrons slowing
down into the thermal energy range, and the absence of upscattering above the
thermal cutoff energy Ee implies that Eq. (9-1) holds with the upper limit truncated
at Ee, then one can rewrite Eq. (9-3) as (see Problems 9-2 and 9-3)
rate at which
:otal rate of absorption= [EcdE"2:. (E) (E)= (E )= neutro_ns slow (9-4)
m thermal range/cm 3 )
0
a </> q c down mto thermal
. range/cm 3•
The differential scattering cross section or scattering kernel "2:..(E'-E) can be

shown to possess another very important property:
(9-5)
where M (E) is the Maxwell-Boltzmann distribution function characterizing the

energies of the particles of an ideal gas at temperature T:
M (E)= 27T I VE exp(- kEr)· (9-6)

(7rkT)3 2
Eq. (9-5) is known as the principle of detailed balance, and it must be satisfied by
any neutron cross section characterizing neutron scattering from a system of nuclei
in thermal equilibrium at a temperature T (regardless of their structure or detailed
dynamics). This property is essentially a consequence of the laws of statistical
mechanics characterizing the scattering material. 7 It implies an extremely impor-
tant consequence for the neutron energy spectrum in a reactor, as we will now
demonstrate.
Suppose we consider the special case in which we set the absorption and source
terms in our infinite medium equation, Eq. (9-1), to zero:
(9-7)
Then we claim that the solution to this equation, regardless of the detailed form of
the scattering cross section, must be just the neutron flux characterizing neutrons in
thermal equilibrium at the same temperature T as the scattering medium:
(9-8)
where n0 is the neutron number density in the medium. This result is a consequence
of the principal of detailed balance, for we can use Eq. (9-5) to demonstrate that
<l>M is indeed a solution of Eq. (9-7):
"2:..(E)cpM(E) = fo 00
dE' L 5 (E'-E )v'n 0 M (E')
= Ia 00
dE' "2:. 5 ( E-E')vn 0 M ( E) = "2:. 5 ( E )</>M( E ). (9-9)
Hence the principle of detailed balance ensures that the equilibrium spectrum of
the neutrons (in the absence of absorption, sources, leakage, etc.) will be a
"Maxwellian" characterized by the moderator temperature T-that is, that the
neutrons will eventually come into thermal equilibrium with the moderator nuclei.
In this sense, the neutrons behave as a very dilute gas that will gradually come into
thermal equilibrium-namely, "thermalize"-with the system through which it is
diffusing. From the properties of the Maxwell-Boltzmann distribution function we
know that the most probable neutron energy and speed in such a situation are then
given in terms of the system temperature T as:
Most probable energy= ET= kT= 8.62x 10- 5T(eV)
Most probable speed= vT =

,y----;;;-
/2kT = 1.28 X 10 ,/-
4
v T (cm/sec). (9-10)
For example, at T=293°K, ET=.025 eV while vT=2.2X 105 cm/sec. (Although at

typical reactor core operating temperatures, T = 590°K corresponds to ET= .051 eV
and vT=3.l x 105 cm/sec.)
The detailed balance property only guarantees that the neutrons will be in
thermal equilibrium with the medium if there are no mechanisms present that tend
to introduce nonequilibrium behavior. Of course, in a nuclear reactor core even for
a very thermal reactor the neutron distribution will never be precisely in thermal
equilibrium because of one of the following effects: (a) presence of absorption, (b)
presence of sources, (c) leakage of neutrons, or (d) time dependence.
These effects all act to perturb the neutron distribution away from thermal
equilibrium. In fact, neutron thermalization can physically be regarded as a
competition process between the moderator atoms attempting to "thermalize" the
neutron distribution into thermal equilibrium with them, and those effects tending
to perturb or distort the thermal neutron spectrum from this equilibrium distribu-
tion.
B. Nonequilibrium Thermal Spectra

The presence of absorption or leakage or a slowing down source can act to
distort the thermal spectrum, that is, the solution cf>(E) to Eq. (9-1), from a
Maxwellian flux c/>M(E). We will consider later procedures for calculating thermal
spectra to any desired accuracy by solving Eq. (9-1) directly in Section 9-111.
However it is useful to give a somewhat more qualitative discussion of nonequi-
librium neutron spectra at this point to lay the groundwork for this subsequent
discussion.
Consider first the addition of an absorption term to Eq. (9-7). One then finds
that the thermal neutron spectrum will be slightly shifted to higher energies, almost
as if its temperature were effectively increased by the addition of absorption (see
Figure 9-1). This "absorption heating" of the spectrum can be easily understood,
however, when it is recalled that most absorption cross sections behave essentially
as 1/ E 112 for low energies. Hence the lower energy neutrons in the spectrum will
tend to be preferentially depleted by absorption, and this leads to an effective shift
(at least in the normalized spectrum) to higher energies.
Exactly the opposite effect occurs when one accounts for neutron leakage. This
can be modeled by adding a term of the form D (E)B; to the total cross section in
(a) Absorption
1.0
r. - --- Maxwellian
flux
rp(E)
0.1
\
\
0.01
E 0.001 0.01 0.1 10
E(eV)
(b) Leakage
1.0
vD(E) rp(E)
0.1
0.01
E 0.001 0.01 0.1 10
E(eV)
(c) Slowing down source

1.0
rp(E)
0.1
(d) Thermal resonance
1.0
rp(E)
0.1
0.1
E(eV)
FIGURE 9-1. Effects of nonequilibrium perturbations on thermal neutron flux spectra.
380
Eq. (9-1). Note that this leakage term appears as an effective absorption. Now since
vD (E) = v[3~ 1,{E)]- 1 tends to increase with increasing energy, one finds that
higher energy neutrons will tend to leak more rapidly from the system. This causes
the equilibrium spectrum to shift to lower energies-that is, to a "diffusion
cooling" effect.
Obviously the presence of a source term in Eq. (9-1) will also perturb the
spectrum. For example, one is most commonly concerned with a source
corresponding to neutron slowing down from higher energies. In this case one
expects the neutron spectrum to behave as 1/ E for energies above several eV.
Several other effects act to perturb the thermal spectrum away from equilibrium.
For example, certain nuclides such as 135Xe, 235 U, or 239 Pu, have low-lying res-
onances in the thermal range. The neutron flux will be depressed in the vicinity of
these resonances, as shown in Figure 9-1. Such low-lying resonance effects are
particularly important toward the end of core life in reactors fueled with low-
enrichment uranium, since the appreciable inventory of the plutonium produced
can cause a strong distortion of the thermal spectrum.
C. Effective Neutron Temperature Models

We have seen that the addition of absorption (or leakage) to a system tends to
shift the equilibrium neutron spectrum to higher (or lower) energies, much as
would occur if one were to change the temperature characterizing the Maxwell-
Boltzmann distribution. For that reason, early studies of neutron thermalization
represented the neutron spectrum by a Maxwellian with an effective temperature
Tn=i=- T:s,9
Since absorption was the dominant loss mechanism in large thermal reactor cores,
the effective neutron temperature Tn was modeled as depending only on the relative
absorption in the core
(9-12)
where Twas the moderator temperature and A was a dimensionless coefficient that
had to be determined empirically by performing experimental spectrum measure-
ments on the assembly of interest (typically, A~l.2-1.8). It should be noted that
the parameter r that appears in this expression is just the inverse of the moderating
ratio that measures the effective competition between absorption and scattering in
determining the thermal spectrum.
The simple effective temperature model given by Eq. (9-11) is incorrect at high
energies since it fails to yield the "1 / E" behavior caused by a slowing down
source. This can be easily corrected by adding to the model a slowing down term
(9-13)
where t:.(E / kT) is a "joining function" characterizing the transition between the
Maxwellian and the slowing down spectrum (see Figure 9-2) while>-.. is a normaliza-
tion factor given by
(9-14)
The effective neutron temperature model played an important role in early
reactor analysis and still is useful in obtaining a qualitative understanding of
thermal reactor spectra. It is interesting to note that if indeed the neutron flux
could be characterized by a Maxwellian cf,M(E, Tn) at a temperature T 0 , then the
total flux characterizing the thermal group would be just
(9-15)
where we have identified the most probable neutron speed vT = \/ 2kT0 / m . Hence
we find that 'PT depends on the effective neutron temperature T 0 •
An interesting application of this result is to the calculation of the reaction rates
characterizing thermal neutrons. First recall that the effective cross sections which
must be used would be averaged over a distribution of nuclear speeds at some
temperature T. That is, the effective reaction rate depends both on the core
temperature T and the neutron temperature T 0 :
(9-16)
Now most thermal absorption cross sections behave as 1/ v in the thermal neutron
range. In Chapter 2, we found that when averaged over the distribution of nuclear
speeds, the effective absorption cross section characterizing a 1/ v absorber was
independent of temperature and depended on neutron speed v as
Lo L (E )v
La( E, T) = ___!_ = a o o (9-17)
V V
1.0
0.6
0.4
0.2
5 10 15 20
E
kT
FIGURE 9-2. The joining function, t::,.(E / kT).
where L~= v 0 La(E0 ) is just the absorption frequency evaluated at any reference
speed v 0 • Hence the absorption rate characterizing thermal neutrons is just
(Ee
Fa= Jo dEvLa(E,T)n 0 M(E,Tn)=La(E0 )v0 n0
= La ( Eo)cJ,o, (9-18)
where cj,0 is the neutron flux evaluated at the reference speed v 0 .

Since most neutron cross sections are measured in materials at the ambient
temperature T= 20.46°C, it has become customary to tabulate thermal neutron
cross section data at the corresponding speed v 0 =(2kT/m) 112 =2200 m/sec or
energy E 0 =kT=0.025 eV. That is, one tabulates La{E0). The reference flux cj, 0 is
then referred to as the 2200 meter-per-second flux. If we want to relate these
quantities to the thermal flux cf,T characterizing a different temperature Tn, we first
note
T ) 1;2
cf,T = ,J-; ( T: cJ,o, (9-19)
We can also express the thermal group constant characterizing a 1/ v absorber in a

Maxwellian flux in terms of La{E0) as
(9-20)
Note that one must take care to distinguish between the 2200 m/sec cross section
La{E0) and the thermally-averaged cross section at ambient temperature, Lath =
('1T/4) 112 L/E0), which differ by some 12%.
One can modify these results to calculate the Maxwellian-averaged thermal
group constants of non-1 / v absorbers in this form by inserting a non-1 / v factor
gaCTn) to write
(9-21)
Such non-1 / v factors have been tabulated for a number of isotopes of interest. 10
However, since we will make little use of the effective temperature model in our
study of thermal neutron spectra, we will avoid a further discussion of such
non-1 / v corrections. ·
There are several important reasons for abandoning the effective neutron tem-
perature model in favor of a more sophisticated treatment of thermal neutron
spectra. First, this rather simple model requires a good deal of empirical guess work
in determining the absorption heating coefficient A and the transition function
t:.(E / kT). Furthermore one finds in practice that this model is inadequate when
there is either strong absorption (i.e., f~0.1) or when the spectrum is appreciably
influenced by resonance absorption in the thermal range. Since r for a typical
LWR ranges from a value of 0.2 for beginning of core life (with zero void fraction
coolant) to as high as 0.6 at the end of core life (with high-quality steam
conditions), it is apparent that the effective temperature model has little relevance
to this class of reactors. In addition, the presence of resonance absorbers in the
thermal range (e.g., 239 Pu) that can build up to appreciable levels near the end of
core life also act to invalidate this model.
D. An Overview of Techniques for Calculating Thermal Spectra
Most schemes for calculating thermal neutron spectra directly solve the
infinite medium spectrum equation for the thermal energy range
(9-22)
Here S (E) is usually taken as a slowing down source, and the cutoff energy Ee is
taken sufficiently high that no appreciable upscattering out of the thermal range
occurs.
If the cross sections are known, then Eq. (9-22) can be written in a fine-structure
multigroup form and solved directly. There are several difficulties with such a brute
force solution, however. First, the pointwise representation of the detailed cross
section data in a nuclear data file such as ENDF /B is quite extensive and can
require large amounts of data handling (particularly for the scattering kernels). In
fact, it is sometimes preferable to directly generate the thermal cross section data
from suitable theoretical models of the scattering material for the particular
problem of interest. Furthermore the fact that appreciable upscattering occurs in
the thermal range means that the multigroup representation of Eq. (9-22) involves
full scattering matrices, rather than the lower triangular structures that appear
when treating neutron slowing down. This implies that the multigroup thermaliza-
tion equations must be solved simultaneously rather than successively from upper
to lower energy groups, as were the multigroup slowing down equations. Such a
solution usually requires iterative techniques (similar to those used in multidimen-
sional diffusion calculations).
However in certain instances one can model the scattering material in such a
way so as to reduce the integral equation (9-22) to a simple second-order differen-
tial equation, which is considerably easier to solve. One very popular such model
represents the moderator as an ideal gas of unbound hydrogen atoms (or protons)
which is in thermal equilibrium at a given temperature T. Because the proton gas
model is extremely useful for modeling the thermal neutron behavior in LWRs, we
will consider its development in some detail in this chapter.
II. APPROXIMATE MODELS OF NEUTRON

THERMALIZATION
A. The Proton Gas (Wigner-Wilkins) Model 11 • 12

Perhaps the simplest description of neutron thermalization yielding results of
sufficient accuracy for use in reactor design is that which models the reactor core
as a proton gas (mass number A= I) in thermal equilibrium at a temperature T.
Such a model obviously ignores both chemical binding and diffraction. However it
does describe upscattering and has been proven remarkably successful for generat-
ing thermal spectra useful in L WR design.
To be more specific, we will consider the solution of the infinite medium
spectrum equation (9-22) with the source term omitted. We will include the
presence of a source by demanding that <f,(E) be properly normalized for large

energies to yield the proper slowing down behavior. We will examine this equation
for the case of a free proton gas under the assumptions:
(1) The microscopic scattering cross section a~ is independent of the relative
velocity vr between neutron and proton.
(2) The absorption cross section behaves as aa( vr) = y / vr. (This assumption
can be relaxed in practice.)
(3) The proton gas is in thermal equilibrium at a temperature T. In this case
we can use two-body kinematics to calculate the thermally averaged
differential scattering cross section (cf. Section 2-I-D) as
I ••
erfyl;, E'>E
(9-23)
exp [
(E'-E)] ..
kT erfVkf ,
u, E'<E
In Eq. (9-23) we have defined the free atom scattering cross section of hydrogen,
~~ = N Ha~. We can calculate the macroscopic scattering cross section by
integrating ~.(E'-E) over all final neutron energies to find
(9-24)
1
where we have defined
(9-25)
( rl
We noted the form of the differential scattering cross section for a proton gas in
Figure 2-15. In particular, we found that it approached the usual slowing down
form
E'>E
0, E'<E (9-26)
for large E, E'»kT. It should also be noted from Eq. (9-23) that ~.(E'-E) has a
discontinuous first derivative at E= E'. In fact, the structure of the scattering
kernel "i:. 8(£'-E) is very similar to that we encountered for the Green's functions
of second-order differential equations (cf. Section 5-11). That is, Eq. (9-22) is of the
form
a ( x )</>( x) = Jbdx' G ( x, x')</>( x')
a
(9-27)
where G (x,x') is the Green's function of a differential operator

d 2G dG
LG (x,x') = a(x)- + b(x)-d + c(x)G (x,x') = 8 (x- x'). (9-28)
dx 2 x
Hence if one can determine L, then by operating with L on Eq. (9-27) as
La ( x )</>( x) = f dx' f,.,G ( x, x')</>( x') = </>( x ),

a
b (9-29)
one can convert the original integral equation into a second-order ordinary
differential equation:
d2 d
a(x)-2 (a.</>)+b(x)-d (a</>)+ [ c(x)a.(x)-1 ]</>(x)=O. (9-30)
dx x
This procedure was utilized by Wigner and1 Wilkins 11 to convert Eq. (9-22) into a
differential equation for if;=</>(E)/[EM(E)]2:
-ix { ix P{x) [V(x)+f]l{;(x)} + { W(x)[V(x)+f]- ~ }l{;(x)=O (9-31)
where
P(x)= e-x
2
+ y:;; xerf(x), (9-32)
2-a(kT) NAy
f= H = H'
Lrr vrNHas
Eq. (9-31) is known naturally enough as the Wigner-Wilkins equation. To it one

appends the boundary conditions that the flux variable vanish as x-o [E-O] and
be normalized to the slowing down source for x» I [E » kT].
The Wigner-Wilkins equation has been studied analytically in exhaustive detail
using a wide variety of approximation techniques-none of which really concern us
here. 1 Of much more interest is the fact that such second-order ODEs can be very
easily solved on a digital computer. To facilitate this solution, one first introduces a
change of dependent variable to recast Eq. (9-31) into a nonlinear first-order
differential equation (a Ricatti equation) more suitable for numerical integration:
dJ S(E) [ 1-2P(E)] P(E) / c

dE = 2kT + 2E J(E)- 2E 12 (E), (9-33)
where
while
(9-35)
Here we have modified this derivation to include a non-1 / v absorption and

modeled leakage with a D (E)B 2 term where
(9-36)
Although the equation for J ( E) looks rather complicated 1t 1s in fact quite

simple to solve numerically. Thermal spectrum codes based on solutions of the
Wigner-Wilkins equation are occasionally referred to as SOFOCATE-type
methods 12 ( after the early thermal spectrum codes used for L WR calculations).
In most of these codes, Eq. (9-33) is first solved for J (E) numerically using
Milne's predictor-corrector method. 13 To get this method started, an asymptotic
solution for E « kT is used. Once J ( E) is known, cj,( E) is then determined from Eq.
(9-34) using a numerical integration (e.g., trapezoidal rule). Typically on the order
of 50-60 mesh points are found to yield suffficient accuracy over the interval
0 < E < 1 eV. (It is amusing to note that Eq. (9-33) is integrated from lower to
higher energies, in contrast to fast spectrum calculations which always proceed
downward in energy-or upwards in lethargy, the more convenient variable for
slowing down calculations.)
The Wigner-Wilkins equation can be applied to mixtures of scattering isotopes
by simply replacing the free atom cross section by an effective scattering cross
section characterizing the mixture:
L~- L ~JN/1{=iL 8•
(9-37)
j
In fact sometimes (although very infrequently) the Wigner-Wilkins equation is

applied by this modification to the analysis of nonhydrogeneous media in the hope
that the actual spectrum will not be too sensitive to the details of the scattering
process.
In Figure (9-3) we have illustrated the shape of the Wigner-Wilkins spectrum for
a 1/ v absorber, a slowing down source, and a thermal resonance (recall Figure
9-1). It should be noted that this model yields not only the correct solutions for
La-O[ci,-ci,M] and E«kT[cj,-1/ E], but also accounts for both absorption
"heating" effects as well as flux depression in the vicinity of a thermal resonance.
,-- .........
X
::,
;;::
QJ
10
/,
/
,/-
/, /
---- \ ''\axwellian
>
·;::;
""
ai
,/-
a:
0.001 0.01 0.1

Neutron energy eV
FIGURE 9-3. Calculated thermal flux shapes.
One can now generate thermal group constants by simply performing

numerically the averages of the micro·scopic cross section data over these computed
thermal spectra.
It is very important to take into account the spatial heterogeneities present in a
reactor core when generating thermal group constants because the thermal
neutron mfp is frequently small compared to the distances characterizing the core
lattice structure. Fortunately there are procedures available by which one can make
simple modifications to infinite medium spectrum calculations to account for these
spatial effects, thereby avoiding a detailed multigroup calculation of neutron
diffusion. We will return to discuss these methods in more detail in Chapter 10.
It should be noted that the proton gas model effectively converts an integral
equation (9-22) into an ordinary differential equation. Such a result is of some
significance when considering numerical solutions, since these differential
equations are considerably easier to solve than the full matrix equations that arise
from discretizing the integral equation directly. In particular it allows us to
calculate the numerical solution at one energy mesh point directly from the
previous one or two mesh points, so that only a small computer memory is required
and iterative methods are avoided.
There are other models of neutron thermalization that exploit the techniques of
converting the integral equation (9-22) into a differential equation more suitable for
machine calculations. We will next examine a model that can be utilized for survey
estimates of the thermal spectrum when the moderator mass number A is large
(e.g., 12c).
B. The Heavy Gas ModeI 14

Obviously the proton gas model is of limited utility in analyzing neutron
thermalization in nonhydrogeneous moderators. Alternative models are needed for
heavier mass moderators such as graphite, which is used in gas-cooled reactors.
One of the simplest thermalization models is based on expanding the cross section
for a free gas of arbitrary mass number A in powers of 1/A, hence arriving at an
approximation which should be valid for large mass numbers.
The differential scattering cross section for a free gas can be derived in a
straightforward, if somewhat cumbersome, consideration of two-body kinematics
(similar to that of Section 2-1-D). This cross section can be written as
+erf( 0¼ -p¼ ') +erf(0V€ +pW) }, E' §; E, (9-38)
where
(A+ 1) (A-1) E
0=---, f=- (9-39)
2\/A p= 2\/A ' kT
If we now expand in inverse mass number A - 1 and retain only lowest order terms,
we arrive at the heavy gas model of the scattering kernel

1
{ (E+ E') }
}:.(E'--'>E)=}:rr(ff,)2 8(E-E')+ A [8'(E-E')+kT8"(E-E')]
(9-40)
Here 8' and 8" are the first and second derivatives of the Dirac 8-function and are
defined in Appendix C. The corresponding total scattering cross section for this
model is
(9-41)
If we now substitute this rather singular scattering kernel back into our integral
equation and note that for large A, g~2/ A, such that
00
2
( dE'}:.(E'--'>E)<j>(E')=g}:rr [kTEd <j> +Edd<j> +<j>] +}:.(E)<j>, (9-42)
Jo dE 2 E
then we find that Eq. (9-))Lbecomes
(9-43)
This simple second-order differential equation is known as the heavy gas equation.
In many ways the heavy gas approximation is similar to the age approximation we
developed to describe neutron slowing down in heavy mass moderators, and in the
high-energy limit the heavy gas model just reduces to age theory.
Using the usual techniques from the theory of second-order ordinary differential
equations, one can laboriously generate asymptotic solutions to this equation for
different absorption cross section energy dependences-for example, }:a(E)
==constant= }:a, or }:a( E) = y / v. It can also be integrated directly to generate
thermal neutron spectra for moderators such as graphite. For example, the numeri-
cal solution of the heavy gas equation (assuming 1/ v absorption) is shown in
Figure 9-4 for several different values of the parameter r = ~a( kT) / g~rr-
It should be noted that as the moderator mass becomes increasingly heavier (i.e.,
g smaller), the moderator is increasingly less capable of slowing down neutrons and
hence the spectrum hardens. In this model, then, a heavier gas is equivalent to the
addition of absorption.
Unfortunately the heavy gas (or even free gas) model is far less effective at
predicting thermal spectra in graphite moderated reactors (such as the HTGR)
than is the proton gas model for the LWR 15 because of the significance of chemical
binding effects. This has led to a rather limited use of this model in practice.
C. Synthetic Scattering Kernel Models

In our two previous models of neutron thermalization we found that the
scattering kernel }:.(£' --'>E) was of such a form that we could transform the
10 3
r ~ 2.s
10 2
1.25
10
0.25
10-1
0.025
,o-2 '---'---'---'---'---'---'----
0 2 4 6
-J!r
FIGURE 9-4 The neutron spectrum predicted by the heavy gas model.
integral form of the infinite medium spectrum equation (9-1) into a differential
equation that was far better suited to machine computation.
It is natural to seek other models of neutron thermalization which also lead to
differential equations. One such model is the generalized heavy gas or primary
model 16 proposed by Horowitz in which the scattering terms in Eq. (9-1) are
modeled by a differential equation very similar to the heavy gas model (9-42):
fo dE'2..(E'-E)q,(E')-2:.(E)q,(E)~~2:rr ~ { f(E)[ EkT ::

00
+ (E- kT)q,]}
(9-44)
Here f(E) is an arbitrary function that must be determined either by fitting to

experimental spectrum measurements or by a fit to an integral of the scattering
kernel itself 2:.(E' -E). The primary model satisfies the detailed balance condition
and includes some accounting for chemical binding effects. Moreover the storage
requirements and machine time required to solve the differential equation
generated by Eq. (9-44) are quite small when compared to the machine labor
involved in solving the integral equation (9-1) directly.
Unfortunately the primary model due to Horowitz fails to yield satisfactory
results when strong absorption is present-particularly in the vicinity of thermal
resonances. To circumvent this, Cadilhac 17 developed a slightly more elaborate
approximation known as the secondary model containing two (rather than one) free
functions. One first writes the scattering kernel as the product of these two
arbitrary functions u(E) and v(E) such that
E>E'
M(E)"'2. (E~E')==-{ u(E)v(E'), (9-45)
s u(E')v(E), E<E'.
(Note again the discontinuous derivative at E' = E.) Then one finds that the
infinite medium spectrum equation (9-1) can be written as the coupled set of
differential equations
dq
"'2.a(E)</>(E)= dE + S (E),
d [ </>( E ) ] . d [ dq ] (9-46)
dE M(E) =J(E)q(E)- dE k(E) dE '
wherej(E) and k(E) are given in terms of the "free" functions, u(E) and v(E), as
j(E)= v)E) d~ [v(E)k(E)], v 1(E) IaE dE'v(E'),
k(E) = [ M (E)"'2..(E) r 1. (9-47)
---Exact
- - - - - Primary model
------Secondary model
......_
~
·c:,
t
; 2
~
-e
......_
!::!
-€).
5 6 7 8
Jlr
FIGURE 9-5. Comparison of the primary and secondary synthetic kernel model predictions for
the flux in the neighborhood of the 240 Pu resonance (moderator is graphite).
This model is particularly interesting since for proper choices of u(E) and v(E) it
will reduce to several of our earlier models including the proton gas model, the
heavy gas (or primary) model, the Fermi age model, or the Goertzel-Greuling
model. Hence the secondary model is evidently capable of bracketing the ther-
malization properties of the true scattering kernel ~.(E' -£).
One can choose the free functions u(E) and v(E) in any of a number of ways.
For example, one can choose these functions by demanding that the first two
energy moments of the true scattering kernel ~.(£'-£) be preserved by the
approximate model [Eq. (9-45)]. The spectra predicted by the secondary model are
frequently in excellent agreement with the results of a brute force numerical
solution of Eq. (9-1) using the true scattering kernel~.(£'-£). This is true even in
the vicinity of thermal resonances, as illustrated in Figure 9-5.
Although the secondary model is usually not used for detailed thermal spectrum
studies, it is frequently used to generate the initial guesses of the spectrum used in
more elaborate iterative calculations necessary to solve the infinite medium
spectrum equation (9-1) directly.
III. GENERAL CALCULATIONS OF THERMAL NEUTRON

SPECTRA
A. Generation of Thermal Cross Section Data

As we have mentioned, thermal neutron cross sections are complicated by a
sensitive dependence on the temperature and chemical state of the scattering
material. Although such cross section information is tabulated and stored in
nuclear data sets such as ENDF /B, it is useful to discuss briefly how these basic
data are calculated. It is customary to express the differential scattering cross
section a.(E'-E,Il'-O) in terms of a quantity, S(a,/3) 1-<>· 15 known as the scatter-
ing law for the material of interest:
(9-48)
Here, ab is the bound atom cross section, while a and /3 are related to the neutron
momentum and energy exchange in a scattering collision:
a=(E'+E-2VFE O·O')/kT,
/3 -(E'- E)/ kT. (9-49)
S (a, /3) depends in a very complicated manner on the detailed structure and
dynamics of the scattering material. Most commonly, however, one introduces
several horrifyingly brutal approximations (at least to the solid-state physicist,
although not to the nuclear engineer) in order to allow the calculation of S ( a, /3).
First one accounts for only incoherent scattering (i.e., no diffraction) such that
S(a,/3) is replaced by the incoherent part of the scattering law, S.(a,/3). Then this
quantity is calculated in the so-called Gaussian approximation 18 in which
2 -f
w ('r)=
00
-oo
d/3 p( f3) [ ( /3) ( /3)]
fisinh(/3/ 2) cosh 2 -cos 2 . (9-50)
The entire dependence of the neutron scattering on the motion of the moderator
atoms is contained in the function p(/3) or equivalently, p(w), w= kT/3/n. For a
harmonic solid, one would identify p(w) as the phonon frequency distribution.
All of the essential physics enters into modeling the form of the function p(w). In
liquid moderators such as water, one uses models that account for the vibrations,
hindered rotations, and translations of the water molecule (this gives rise to the
so-called Nelkin scattering kernel 1-6 used for water or the St. John-Brown 1-6 kernel
for D 20). For polycrystalline materials such as graphite, the details of lattice
vibrations must be taken into account. (The usual model used to describe graphite
is known as the Park's scattering kernel 1-6.) It is necessary to use these more
elaborate models of the microscopic dynamics of the moderator in performing
detailed reactor studies.
The generation of thermal cross section data usually begins with the calculation
of the scattering law S(o:,/3) for a given phonon frequency distribution p(w) using a
computer program such as the GASKET code. 19 A supplementary calculation is
then needed to generate the actual scattering kernel "2-.(E' ~E, O' ~o) from the
scattering law (e.g., the FLANGE code). 20 These data can then either be used to
prepare a restricted thermal cross section library, or transferred into a complete
nuclear data set such as ENDF /B.
B. Thermal Spectrum Calculations

For accurate determination of thermal neutron spectra and the generation of
thermal group constants, one must perform a direct solution of the infinite medium
spectrum equation (9-1) by writing it in a discrete multigroup form
Ltn </>n = ~ Lsn'n </>n• + Sn (9-51)

n'= I
where n is the fine group index. Since the thermal energy range is rather narrow
(0,;;; EZ l eV), and typically from 50 to 100 energy groups are used, one can
usually just use input cross section data on a pointwise basis (evaluated at the
center of a group, for example).
Actually Eq. (9-1) is a bit too simplified, since one usually adds in a crude spatial
treatment using the P 1 or B 1 method in a manner similar to that used to calculate
fast spectra. Then the thermal leakage is determined by specifying an equivalent
geometric buckling B;.
Such infinite medium spectrum calculations must also be corrected for core
heterogeneities. In the next chapter we will develop several prescriptions for
modifying the thermal group constants to account for flux nonuniformities which
arise in a fuel lattice cell.
C. Coupling between Fast and Thermal Spectrum Calculations

In a rigorous determination of the fast and thermal neutron energy spectrum,
the fast and thermal energy regions will be coupled together. Downscattering from
higher energies provides the source term in the thermal range, while thermal
neutron induced fission reactions will give rise to the fission source in the fast
neutron region. In addition there may be some mild influence on the lower energy
behavior of the fast spectrum due to upscattering of thermal neutrons.
Fortunately the coupling between fast and thermal spectrum regions is
sufficiently weak that the calculation of fast and thermal neutron spectra are
usually performed separately. However many fast spectrum codes will generate the
P O and P I slowing down sources to be inserted into thermal spectrum calculations.
Usually all downscattered neutrons from the fast region are assumed to go into the
highest energy thermal group, except for hydrogeneous moderators in which the
slowing down neutrons are distributed over all thermal groups. If upscattering out
of the thermal region is allowed, these neutrons are usually assumed to go into the
lowest energy fast group.
REFERENCES

Amsterdam ( 1966).
2. G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand, New York (1970),
Chapter 7.
3. K. H. Beckurts and K. Wirtz, Neutron Physics, Springer, Berlin (1964).
4. Neutron Thermalization and Reactor Spectra, International Atomic Energy Agency,
Vienna (1968), Vols. 1-11.
5. I. I. Gurevich and L. V. Tarasov, Low Energy Neutron Physics, Wiley, New York (1968).
6. D. E. Parks, M. S. Nelkin, N. F. Wikner, and J. R. Beyster, Slow Neutron Scattering and
Thermalization with Reactor Applications, Benjamin, New York (1970).
7. A. C. Zemach and R. L. Glauber, Phys. Rev. 101, 118 (1956).
8. R. F. Coveyou, R.R. Bate, and R. K. Osborn, ORNL-1958 (1955).
9. E. R. Cohen, Nucl. Sci. Eng. 2, 227 (1957).
10. C. H. Westcott, Effective cross section values for well-moderated thermal reactor
spectra, AECL-1101, 1962 (3rd Edition).
11. E. P. Wigner and J. E. Wilkins, AECD-2275 (1944); M. M. R. Williams, The Slowing
Down and Thermalization of Neutrons, North-Holland, Amsterdam (1966), p. 75.
12. H. Amster and R. Suarez, The calculation of thermal constants averaged over a Wigner
Wilkins flux spectrum; SOFOCATE, WAPD-TM-39 (1957).
York (1969) p. 386.
14. J. E. Wilkins, USAEC Document CP-2481 (1944); Ann. Math. 49, 189 (1948).
15. J. R. Beyster, N. Corngold, H. C. Honeck, G. D. Joanou and D. E. Parks, P /258 in
Third U. N. Conference on Peaceful Uses of Atomic Energy (1964).
Amsterdam (1966), p. 273.
17. M. Cadilhac, J. Horowitz, J. L. Soule, and 0. Tretiakoff, Proceedings BNL Conference
on Thermalization 2, 1962; M. Cadilhac, et. al., Third U. N. Conference on Peaceful Uses
of Atomic Energy (1964); M. M. R. Williams, The Slowing Down and Thermalization of

Neutrons, North-Holland, Amsterdam (1966), p. 292.
18. P. A. Egelstaff and P. Schofield, Nucl. Sci. Eng. 12, 260 (1962).
19. J. U. Koppel, J. R. Triplett and Y. D. Naliboff, GASKET, a unified code for thermal
neutron scattering, General Atomic Report GA-7417 (1966).
20. H. C. Honeck and D. R. Finch, FLANGII (Version 71-1), a code to process thermal
neutron data from an ENDF /B tape, Savannah River Laboratory Report DP-1278
(1971).
PROBLEMS
9-1 Show that the scattering cross section arr of a free nucleus is related to that
characterizing a perfectly bound nucleus ab by ab= (1 + A - 1)2arr. (Hint: First dem-
onstrate that ab= 4?T dab

dQ I = 4'17 da1r
dQ I . Then use the expression for da1r/ dQ
cos(/= 1 cosfJ= 1
for s-wave scattering.)
9-2. Generalize the definition of the slowing down density q(r, E) given by Eq. (8-18) to
account for upscattering as well as downscattering.
9-3 Using the definition of the slowing down density for energies E > Ee in the infinite-
medium spectrum equation (9-1), derive the balance condition Eq. (9-4).
9-4 Check by explicit calculation whether the detailed balance condition holds for the
scattering kernels characterizing: (a) a free proton gas, (b) the heavy gas model, and
(c) a free gas of arbitrary mass number A.
9-5 Give three reasons why the neutron energy distribution in a thermal homogeneous
reactor is not Maxwellian. What specific physical effects cause deviations at high
energies and what physical effects give rise to deviations at low energies?
9-6 Calculate the average energy and velocity of neutrons in a Maxwellian distribution at
a temperature T 0 •
9-7 Plot the behavior of Ecp(E) against E for the effective temperature model given by Eq.
(9-13). In particular show that the peak of Ecp(E) occurs at E~2kT0 •
9-8 Suppose that a 1/ v absorber is suddely inserted uniformly into an infinite medium
containing a well-established Maxwellian distribution of neutrons.
(a) Show that the total density of neutrons in the medium steadily decreases in time,
but that the neutron energy distribution, and hence the effective neutron tem-
perature, does not change.
(b) Discuss qualitatively the change in the energy distribution for a non-I/ v ab-
sorber.
[Hint: Solve the time-dependent infinite medium spectrum equation as an initial value
/
problem subject to the initial condition, cp(E, t = 0) = cf>M(E).]
✓ 9-9 A water-moderated thermal reactor operating at 300°C has a computed average
thermal flux of 2 X 10 14 neutrons/ cm 2-sec. Compute the absorption rate density in the
water in this reactor. [Take PH,o=0.72 g/cm 3 (at 155 bar).]
Demonstrate with equations how to measure the neutron temperature in the center of
a reactor core by simultaneous activation of manganese (1 / v neutron absorber) and
lutetium (.142 eV resonance absorber) foils in the core.
9-11 Sketch the dependence of the thermal group constants characterizing a 1/ v absorber
on r = "2.a/ g"2.. if the thermal spectrum is characterized by a Maxwellian flux
cf>M(E, T 0 ) with an effective neutron temperature T 0 = T(I + AI').
9-12 Demonstrate that the requirement expressed by equation (9-3) demands that we
choose the "joining function" normalization as 'A.= \Iii cf>TI' /2(1- f) in the effective
temperature model, if we assume a slowing down source into the thermal group of
q(Ec)=f'£ 5'A..
9-13 Verify the expression given for ~,(£) characterizing a free proton gas by explicitly
integrating the form given in Eq. (9-23) for ~,(E' ➔ E).
9-14 The second energy moment of the scattering kernel,
1
E/a.(E' ➔E)
00 00
M 2 ==.-- 2 ( dE' ( dE M(E')(E'-
(kT) Jo Jo
is a measure of the mean squared energy exchanged between the neutron and the
scattering atoms. Demonstrate that for a free proton gas M 2 = 2 V2 orr.
9-15 Derive the Wigner-Wilkins equation (9-31) from the infinite medium spectrum
equation (9-22) characterizing a proton gas. [Refer to E. P. Wigner and J. E. Wilkins,
AECD 2275 (1944) or M. Williams, Thermalization and Slowing Down of Neutrons,
North Holland, Amsterdam (1967), p. 77 for assistance.]
9-16 Using the variable transformation defined by Eq. (9-34), derive the Ricatti equation
(9-33) from the Wigner-Wilkins equation.
v 9-17 Write a computer program that integrates the nonlinear differential equation (9-33)
characterizing a free proton gas. Use either a predictor-corrector or Runge-Kutta
scheme with an energy mesh size of LlE=0.001 eV. Determine the flux cf>(£) from
J ( E) by performing the integration indicated in Eq. (9-34) using trapezoidal quadra-
ture. For convenience, choose a 1/v absorption cross section ~a(E)=~~/V E. Allow
~~, ~f:, B2 , and T to be read in as inputs to the code.
9-18 Use the computer code written in Problem (9-17) to determine the spectrum in water
at room temperature. Compare this spectrum with that predicted by the effective
temperature model.
9-19 Run a SOFOCATE-type thermal spectrum code to generate the thermal neutron
spectrum for a typical PWR and BWR core using the parameters tabulated in
Appendix H. Use exit coolant conditions.
9-20 Repeat the spectrum calculation in Problem 9-19 for the PWR with boron added to
the water moderator in concentrations of 500, 1000, and 1500 ppm (parts-per-million).
9-21 Derive the heavy gas equation from the infinite medium spectrum equation (9-22) by
expanding cf>(£') in the scattering integral in a Taylor series about E, and then
truncating all terms of order higher than (E- E')2. It will be helpful to note that for
A»I,
fo dEa,(E'➔ E)(E'-E)"""
00
3 or,(E'-2kT),
fo dEa,(E' ➔ E)(E'-E)2""" ~ orrE'kT.

00
9-22 Demonstrate that for E»kT, the solution to the heavy gas equation is just cf>(E)~C/
~~rrE.
9-23 Determine the choice of the functions u(E) and v(E) of the secondary model that will
yield the proton gas model and the heavy gas model.
9-24 A very useful approximation to the thermal neutron scattering kernel ~,(E ' ➔ E) is
the so-called synthetic or simple degenerate kernel (SDK) model
where
[See N. Corngold, P. Michael, and W. Wollman, Nucl. Sci. Eng. 15, 13 (1963).]
Demonstrate that this kernel preserves the correct total scattering cross section ~,(£)
and also satisfies the principle of detailed balance.
9-25 Solve the infinite medium spectrum equation for the flux resulting from an arbitrary
source S (E) using the SDK model.
9-26 The SDK model can be used to study the time-dependent thermalization of a neutron
pulse injected into a moderator at an energy £ 0 • To this end, consider the initial value
problem
.!. aac/> + ~1(E )cf,(£, t) = ( OO dE'~.(E' ➔E )cf,(E', t)

V t Jo
subject to the initial condition
Assuming an absorption ~a(E) = ~~/ v and the SDK model, solve this problem for
cf,(E, t). (Hint: Use a Laplace transform in time.)
9-27 Age-diffusion theory can be used to provide the effective slowing down source for the
neutron diffusion equation characterizing the thermal group. In this manner, deter-
mine the distribution of thermal neutrons resulting from an isotropic point source of
monoenergetic fast neutrons located at the origin. In particular, discuss the asymp-
totic behavior of your solution far from the source for VT ~ L.
9-28 Describe possible experimental techniques for measuring: (a) the neutron age to
thermal and (b) the thermal neutron diffusion length. Assume that the only available
neutron source is a fast ( ~2 MeV) source. Justify your discussion by simple calcula-
tions. Then discuss how your experiments would differ for graphite and for light
water.
9-29 Neutrons of lethargy zero are produced uniformly throughout an infinite medium.
Assume that they then slow down by elastic collisions until they reach Eth and there
enter a one-speed diffusion process. The medium is characterized by the macroscopic
cross sections ~ .. ~a, ~r, and ~ 1, which are all independent of neutron energy for all
energies. Find the fraction of fissions caused by the thermal neutrons. Assume that
the age approximation is valid, that is, q(u)=~~ 1cf>(u).
9-30 Demonstrate that when age-diffusion theory is used to describe neutron slowing
down, and one-speed diffusion theory is used to describe thermal neutron diffusion,
then the multiplication factor for a bare, homogeneous reactor is
9-31 Compute and plot the critical mass and the fast and slow nonleakage probabilities for
a spherical assembly of 235 U and moderator as a function of the radius of the
assembly for the following moderators: (a) H 20, (b) D 20, (c) Be, and (d) graphite.
Determine the minimum critical mass in each case. (Use the parameters listed in
Table 8-7.)
10
Cell Calculations
for
Heterogeneous Core Lattices
Thus far we have restricted ourselves to the study of reactor cores in which fuel,
moderator, coolant, and structural materials were assumed to be intimately and
homogeneously mixed. However, nuclear reactor cores are of course constructed in
a highly heterogeneous configuration to facilitate thermal design (coolant channels,
heat-transfer surfaces), mechanical design (structural integrity, fuel fabrication and
handling), and reactivity control (control rods, burnable poisons, instrumentation).
For example, the reader should recall the rather detailed structure of the typical
PWR core illustrated in Figures 3-6 and 3-7. Such heterogeneities in the reactor
fuel array or lattice must be taken into account in nuclear design since they will
cause a local spatial variation in the neutron flux which may strongly influence
core multiplication.
The degree to which core lattice effects must be taken into account in reactor
design depends on the characteristic dimensions of the lattice structure, for ex-
ample, the diameter of a fuel pin or the spacing between fuel elements, compared
to the mean free path of neutrons in the core. For example, in the LWR the
thermal neutron mean free path is typically on the order of one cm, comparable to
the fuel pin diameter. Hence the flux distribution in the fuel might be expected to
be quite different from that in the moderator or coolant channel, thereby necessi-
tating a detailed treatment of the heterogeneity. By way of contrast, the much
longer mean free path characteristic of the neutrons in a fast reactor (typically tens
of centimeters) allows a much grosser treatment of lattice effects.
Of course, a detailed treatment of the core lattice on a scale sufficiently fine to
account for the spatial variation of the flux in the neighborhood of a given fuel
element is clearly out of the question, since it would require an unmanageably large
array of mesh points in a multigroup diffusion calculation (typical LWRs have over
50,000 fuel elements). Indeed due to their strongly absorbing nature, fuel and
398
CELL CALCULATIONS FOR HETEROGENEOUS CORE LATIICES / 399
control elements in the core frequently require a more accurate treatment of

neutron transport than that provided by diffusion theory. Hence one must adopt a
more piecemeal approach by seeking to selectively "homogenize" the analysis of
the core, usually by providing prescriptions for including lattice effects into existing
infinite medium neutron energy-spectrum calculations or by calculating few-group
constants that have been spatially averaged over the finer details of the flux
distribution in the lattice.
Of course, the type of treatment one chooses will depend on the purpose of the
calculation. For example, one can contrast a hand calculation based on the
six-factor formula suitable for a crude survey estimate with an extremely detailed
transport calculation that might be used in a comparison with a critical experiment
or perhaps as a benchmark for the testing of other calculational schemes. Our
concern in this chapter is with more routine design calculations that must be
performed very frequently and hence place a premium on calculational ease. For
such schemes to yield sufficient accuracy, one is forced to rely on frequent cross
calibration with experiment-that is, to accept a certain amount of empiricism
(fudging) in the technique. And of course as in most fast yet accurate methods, one
relies heavily on a cancellation of errors.
For example, in a LWR core study, one might attempt to generate group
constants for a few-group diffusion analysis using a SOFOCATE-MUFT scheme
similar to that discussed in Chapters 8 and 9. Yet of course these procedures
perform calculations for a homogeneous medium in which only gross leakage
effects are accounted for via the P 1 or B 1 approximation. Hence our objective here
is to provide a prescription for modifying these homogeneous results to account for
heterogeneous lattice effects. The general approach is to divide the periodic array
of the reactor lattice into a number of identical unit cells and then calculate the
effective group constants characterizing one such cell. The detailed analysis of a
unit cell must account for the strong spatial variation of the neutron energy
spectrum within the cell. The cell is homogenized by averaging the group constants
characterizing materials in the cell over the spatial flux distribution within the cell
to produce cell-averaged or so-called self-shielded group constants that can then be
used in a multigroup diffusion analysis of the entire core.
In this chapter we develop several procedures applicable for performing such cell
calculations and thereby generating cell-averaged group constants. The highly
absorbing nature of the fuel region in a reactor lattice cell will necessitate a
transport theory description of the neutron flux, based on either so-called collision
probability methods or a direct solution of the transport equation itself (or perhaps
a Monte Carlo calculation). Since the trend in recent years has been toward more
detailed treatments of the heterogeneous effects in core lattices, we will include a
brief discussion of both analytical and direct numerical techniques for studying
neutron transport within lattice cells.
I. LATTICE EFFECTS IN ,NUCLEAR REACTOR ANALYSIS
A. A Qualitative Discussion of Heterogeneous Effects on Core

Mui ti plication in Thermal Reactors
Before we launch into a detailed study of how lattice effects are included in
reactor core analysis methods, it is useful to give a simple qualitative discussion of
the effects of fuel lumping on core multiplication in thermal reactors, since this
reveals most of the relevant physics. To this end we will simply examine how each
of the various terms in the six-factor formula are modified in passing from a
homogeneous reactor core, in which the fuel and moderator are intimately mixed,
to a heterogeneous lattice, in which the fuel is lumped separately from the
moderator. This discussion actually has a rather interesting historical significance,
since without fuel lumping it would have been impossible to achieve a critical
assembly using natural uranium and graphite in Fermi's "pile" at the University of
Chicago in 1942. More specifically, for a natural uranium system, 1/ = 1.33 while
t = 1.05. If one studies j and p for various homogeneous mixtures of natural
uranium and graphite, then, at best one finds jp = 0.59. 1 Hence, for a homogeneous
reactor core composed of these materials, k 00 < (1.33) (1.05) (0.59) = 0.85. Obviously
such a system could never be made critical.
Fermi and Szilard2 noted that if the fuel were lumped into a heterogeneous
lattice, then the resonance escape probability p would increase dramatically. This
occurs because neutrons that are slowed down to resonance energies in the
moderator are primarily absorbed in the outer regions of the fuel element-hence
leading to a depression in the neutron flux within the fuel at resonance energies
(see Figure 10-1). That is, the outer layers of the fuel tend to shield its interior from
resonance energy neutrons, thereby decreasing the net resonance absorption and
Moderator
</> (r, E 1 )
. - - - - - q,(r, E0 )
. - - - - - q,(r, E,h)
FIGURE 10-1. Flux behavior in the neighborhood of a fuel pin.

CELL CALCULATIONS FOR HETEROGENEOUS CORE LATTICES / 401
hence increasing the resonance escape probability p. This "self-shielding" effect is

sufficiently strong that k 00 increases to a value of 1.08 in a graphite-natural
uranium lattice. (It should be noted that this spatial self-shielding is in many ways
analogous to the energy self-shielding we encountered in our discussion of the
Doppler effect in resonance absorption. Both effects tend to decrease resonance
absorption, thereby increasing the resonance escape probability.)
There are other effects due to fuel lumping, however. On the positive side, the
fast fission factor f will increase somewhat in a heterogeneous assembly because
the probability of a fast neutron suffering a collision with a fuel nucleus while its
energy is still above the fast fission threshold will increase. On the negative side, the
thermal utilization J will decrease somewhat because the thermal flux tends to be
depressed in the fuel, hence yielding less absorption in the fuel at thermal energies
(again due to self-shielding). Since thermal absorption in the fuel (in contrast to
epithermal resonance absorption) usually leads to fission, the net result is a
decrease in core multiplication due to J. Fortunately this decrease is far outweighed
by the increase in p.
To examine these effects in somewhat more detail, we will now consider the
influence of fuel lumping on each term in the six-factor formula. Then in later
sections of this chapter, we will turn to the more practical problem of just how such
effects are included in core neutronics analysis.
First recall that 17 depends only on the macroscopic cross sections characterizing
the fuel
(10-1)
(where the latter expression holds for a mixture of fuel isotopes). Hence one would
not expect fuel lumping to appreciably affect this ratio. In actuality, however, the
cross sections that appear in 1J are group constants characterizing the thermal
energy group. These are dependent, of course, on the thermal neutron energy
spectrum, and this spectrum depends, in turn, on the fuel-moderator lattice
configuration. Hence there will be a slight modification in 17 when going to a
heterogeneous lattice. This change is usually ignored in less sophisticated thermal
spectrum codes that include heterogeneities via thermal disadvantage factors. More
elaborate cell calculations accounting for space-energy effects within the cell will
include this correction automatically.
In our earlier treatment of homogeneous systems we defined the thermal utiliza-
tion J as the ratio of the rate of thermal neutron absorption in the fuel to the total
rate of thermal neutron absorption in all materials. This definition can be applied
as well to a heterogeneous core by writing
Jd r L~ (r)<f>(r)
3
f=----=--------,------ (10-2)
Jd r L~ (r)<f>(r) + Jd r L~(r)<f>(r) ·
3 3
Here we are considering the core to be made up of only two types of material, fuel,
denoted by the superscript "F", and moderator, denoted by "M". (The extension to
more than two regions will be given later.) Since the core is made up of a number
of identical fuel cells, we can consider the average in Eq. (10-2) as being taken only
over the volume Veen of one such cell. Now if we recognize that the macroscopic
cross sections ~~(r) and ~~(r) are actually constant over the volume VF of the fuel
and VM of the moderator respectively, and vanish elsewhere, we can limit the range
of integration to each of these regions and pull the cross sections out of the
integrals to write
(10-3)
Next, suppose we define the spatially averaged flux in each region as
- - l ( 3
</>p=v ), d r<f>(r), (10-4)
F VF
Then we can rewrite the thermal utilization fin terms of these averages as
(10-5)
or, dividing both numerator and denominator by ~F V p,
( 10-6)
where we have defined the thermal disadvantage factor r as the ratio of the average
flux in the moderator to that in the fuel
(10-7)
This terminology arises because the thermal flux tends to be depressed in the highly
r
absorbing fuel region, leading to a value of > l. Hence since the average flux is
somewhat higher in the moderator than in the fuel, the fuel nuclei are at a relative
r-------------7
: Moderator M :
I I
I i'///JJ.UJ> I
I I
I I
I I
I I
I I
I I
I I
L ___________ J FIGURE 10-2 A two-region unit fuel cell.
disadvantage in competing with moderator nuclei for the capture of thermal

neutrons.
The depression of the thermal flux in the fuel is again a consequence of the
self-shielding effect. That is, neutrons born in fission events in the fuel will tend to
thermalize in the moderator and then must eventually diffuse back into the fuel to
induce a further fission. However the highly absorbing nuclei near the surface of
the fuel pin tend to absorb the thermal neutrons diffusing back in from the
moderator and hence in effect shield the fuel nuclei in the interior of the pin. This
leads to the observed flux depression.
We can compare this more general definition of thermal utilization f with our
earlier expression for a homogeneous system
LFhorn
horn a
! = -L-F-ho_rn_+_L_M_h_o_rn ' ( 10-8)
a a
if we consider the homogeneous system to consist of unit cells of the same volume
Veen= VF+ V M as our heterogeneous cell, but with the fuel and moderator now
spread uniformly over the cell. Hence we would now find the fuel and moderator
number densities in the homogeneous cell as
NV N M VM
N horn = ______!'.._____ horn=
N M • ( 10-9)
F V cell ' Vcell
If we now note that the macroscopic cross sections for the homogeneous cell are
LFhorn=Nhorn aF LMhorn=Nhorn aM
a F a' a M a,
we find from Eq · (10-9)
(10-10)
Comparing this with our more general definition for a heterogeneous system in Eq.
(10-6), which we will now refer to as fhet, we note that in general fhet < !horn, since
t ;;,. 1 (as the flux depression in the fuel would imply). Therefore lumping the fuel
into a heterogeneous lattice will actually lower thermal utilization, thereby decreas-
ing core multiplication.
One can generalize the concept of thermal utilization even further to account for
a multiregion fuel cell. Consider, for example, a three-region fuel cell composed of
fuel, clad, and moderator material. Then in analogy _with our two-region example,
we would write
(10-11)
If we now divide through by the average flux in the fuel, we find that the thermal
utilization f involves two thermal disadvantage factors, one for the fuel to modera-
tor, ~M/~F• and one for the fuel to clad, ~c/~F·
As yet we have said nothing concerning just how one calculates these thermal
disadvantage factors. Of course, it might be argued that the concept of thermal
utilization within the context of the six-factor formula has very limited utility aside
from crude survey estimates. We will see later, however, that the disadvantage
factor t can be used to spatially average thermal group constants over unit fuel
cells, and hence plays an extremely important role in reactor design. Hence we will
devote a considerable amount of attention toward its calculation in the next
section. We will examine both approximate ("quick and dirty") ways to estimate t
as well as more elaborate schemes based on transport theory and collision-
probability methods.
Since the fast neutron mfp is typically large compared to the scale of lattice
dimensions, one usually need only consider the thermal nonleakage probability
PTNL· Lattice effects enter primarily through a change in the thermal group
constants, which accounts for spatial flux variation within a cell. In particular one
can show that the diffusion length characterizing a heterogeneous core is approxi-
mately just that characterizing a pure moderating medium decreased by the
thermal utilization L 2 :::::: L~ (1- J) (see Problem 10-3). However since leakage from
a large th~rmal reactor core is a relatively minor effect, we will avoid an explicit
discussion of heterogeneous effects on PTNL since it will have a small effect on core
multiplication (and will be accounted for by using cell-averaged group constants in
any event).
Perhaps the most significant effect due to heterogeneous arrangement of fuel in a
thermal reactor is a significant increase in the resonance escape probability p. This
modification occurs as a consequence of two phenomena. First there is a geometri-
cal effect arising because the physical separation of the fuel and the moderator will
allow some neutrons to slow down without ever encountering the fuel. This effect is
of secondary importance to the phenomenon of self-shielding, however.
To understand this second effect more clearly, consider the diagram of the
spatial dependence of the neutron flux at several different energies characterizing a
resonance given in Figure 10-1. Of course at fission energies we might expect the
flux to peak in the fuel since the fission sources are confined to the fuel. However
once we have dropped in energy much below the fission energy, we will begin to
see a flux depression in the fuel. This arises because the fission neutrons must
escape the fuel pin into the moderator in order to be appreciably moderated.
(Nuclear inelastic scattering from fuel isotopes as well as elastic scattering from
light isotopes such as oxygen admixed into the fuel cause some moderation, but
this is a secondary effect in thermal reactors.) Hence the moderator presents
effectively a volumetric source of neutrons appearing at the lower energies. These
neutrons must then either downscatter to even lower energies or diffuse into the
fuel where they are absorbed.
The fuel presents a very highly absorbing medium to the neutrons of resonance
energy diffusing in from the moderator. This absorption is sufficiently strong that
many of the neutrons incident on the fuel are absorbed in the outer layers of the
fuel pin. Hence the fuel nuclei in the pin interior see a somewhat depressed flux
due to the effective shielding presented by the fuel nuclei near the fuel pin surface.
Such self-shielding is present to a certain degree at all energies below fission
energies. However it becomes much more pronounced when the fuel absorption
cross section is large-such as at a resonance in the absorption cross section or in
the thermal energy range.
This effect is quite pronounced. For example, the resonance integral
characteristic of natural uranium uniformly mixed with moderating material is
about 280 b; lumping the uranium, we can reduce the resonance integral to a value
of 9 b-a 30-fold reduction. We will see later that one can usually write the
effective resonance integral in the form
l=C 1 +c2 ( ~) (10-12)
where AF is the surface area of the fuel lump and MF is its mass (proportional to its
volume). As (Ap/ Mp) dec'reases-corresponding to more highly heterogeneous

lattice configurations-the resonance integral decreases.
To treat the effects of heterogeneities in resonance absorption requires the use of
several concepts from transport theory. Since this subject is of considerable
importance in thermal reactor design, we will discuss it in some detail in Section
10-111.
The fast fission factor «: is also increased somewhat by going to a heterogeneous
lattice. To understand why, one need only recall that the energy of a neutron must
be above a certain threshold in order to induce a fast fission reaction in a
fissionable isotope such as 238 U. By lumping the fuel, one effectively increases the
probability that a high-energy fission neutron will encounter a fuel nucleus before
it is slowed down below the fast fission energy threshold, either by elastic scattering
collisions with moderator nuclei or inelastic scattering from fuel nuclei.
In summary, then, lumping the fuel into a heterogeneous lattice can significantly
increase k 00 for natural and slightly enriched (< 5%) uranium cores. The dominant
effect is contained in the behavior of the thermal utilization f and the resonance
escape probability p. To illustrate these trends, we have indicated in Figure 10-3 the
variation of each of the factors in the four-factor formula with heterogeneity (in
this case measured by the fuel pin pitch/ diameter ratio) for a typical PWR lattice.
2.0....---------.-----------,
~--.... --------.!!!!;'________ _
1.5 \
''
J
' ' ',-
........
.... ,
........
--------
-::::.... €
---- ----
-----
----- ---
f
-"---
p .,,,.,- - -
--.:..--,
////
0.5
/
/
0.0 ..___ _ _ _ _ _......:;.i..-_ _ _ _ ___.

1.0 1.5 2.0
Pitch/diameter
FIGURE 10-3. Effect of fuel lumping on k 00 •
B. Core Homogenization
To be more specific, let us outline one possible approach to the treatment of
core lattice effects. We begin by noting that reactor cores have a regular or periodic
lattice structure in which one subelement or so-called unit cell is repeated
throughout the core. For example, a fuel subassembly or group of fuel subassemb-
lies such as those sketched in Figure 10-4 could be regarded as a unit cell. On a
more detailed scale, a given fuel element and adjacent coolant channel might be
chosen as the unit cell (Figure 10-5). In fact, for reactor types using coated particle
fuels, one is frequently required to consider a fuel microsphere or "grain" as a unit
cell.
Of course the reactor lattice structure is not precisely regular because of control
or instrumentation devices, nonuniform fuel loadings and coolant densities, core
boundaries, and so on. These lattice irregularities are usually handled within the
global multigroup diffusion calculation of the core flux distribution. For the
purposes of our present analysis, we will ignore these gross lattice irregularities and
assume that the core can be represented as an infinite array of identical lattice
cells.
The essential scheme is then to perform a detailed calculation of the flux
distribution in a given unit cell of the lattice-usually assuming that there is zero
net neutron current across the boundary of the cell (using arguments based on the
symmetry of the lattice). The various multigroup cross sections characterizing
materials in the cell are then spatially averaged over the cell, using the flux
distribution as a weighting function. In this way one can characterize the cell by
effective group constants accounting for the inhomogeneous flux distribution in the
cell. This scheme essentially replaces the actual unit cell by an equivalent homo-
geneous unit cell characterized by these effective cross sections.
For example, one usually begins by considering a typical fuel cell-namely, fuel
plus clad plus coolant. The fuel cell is first reduced to an equivalent cell of simpler
geometry to expedite calculations (see Figure 10-6a) (taking care to preserve
volume fractions).
Our primary interest is usually concerned with the generation of effective fast
and thermal group constants for such a cell. In the fast range, the heterogeneities
enter primarily as modifications to the resonance escape probability and fast
fission cross sections. Hence it is usually sufficient to simply perform the usual
infinite medium fast spectrum calculation, taking care, however, to account for
heterogeneous effects in resonance absorption and fast fission via techniques that
will be discussed later in this chapter.
The much shorter mean free path characterizing thermal neutrons necessitates a
somewhat more detailed treatment of heterogeneities in determining the thermal
flux spectrum in a fuel cell. For less detailed core calculations, one can frequently
get by with simply modifying the results of an infinite medium thermal spectrum
calculation (e.g., SOFOCATE) to account for the variations in average flux in the
fuel, ;pp, the moderator, ;pM, and the clad, ;pc- Of primary concern here is the
calculation of the so-called thermal disadvantage factors, ;pM;;pF and ;pc/;pp, which
enter into modifying the infinite medium spectrum results. More detailed calcula-
tions of thermal spectra usually involve a direct solution of the neutron transport
equation characterizing the cell. Both schemes will be discussed in Section 10-11.
Such fuel cell calculations are customarily performed under the assumption of
zero net neutron leakage between cells (which decouples the fuel cells from one
another). Actually such single fuel rod spectrum calculations are of questionable
Fuel assembly with

rod-cluster control
0
00
O O 0
00000000
Fuel assembly without

rod-cluster control
PWR assemblies
Fue I pin\' Coolant channel HTGR assemblies
00~000 0 0000000
000000 0 0000000
0000000 · 0000000
0000000 0000000
0000000 0000000
0000000 0000000
0000000 0000000
Co ntrol
' I bl ade
I
-
"
BWR assemblies
FIGURE 10-4. Typical fuel assemblies.
validity when rod neighbors include water holes, poison shims, control rods, or
Pu-loaded fuel pins, since one then needs to account for cell-to-cell leakage.3
For coated particle fuels, such as those utilized in the HTGR, additional
correction factors must be used to account for the heterogeneity represented by the
microscopic grain structure of the fuel in the calculation of cell-averaged thermal
spectra and group constants. 4
The next step in the analysis of the core is to consider a typical fuel assembly or
grouping of fuel assemblies, including control or shim elements (Figure 10-6b). The
few-group constants calculated for the fuel cell can be used to describe most of the
assembly, with the exception of control material which requires rather specialized
/Coolam ""'°"'' --=
0 0 0 0
Equivalent
cells
0
0 0 0
~Fuel
0 0 0 ~ ) J:
Square lattice Hexagonal lattice
FIGURE 10-S. Typical unit fuel assemblies.
(a) Fuel-cell homogenization

0
000000
oeooeo ••••
ooeeoo,
oo••~
oeooeo
000000
--~ --+---+--------r- '
(b) Fuel-assembly homogenization
(c) Core homogenization

FIGURE 10-6. Reactor-core homog_enization.
408
techniques. Usually a detailed multigroup two-dimensional diffusion or transport

code is used to determine the flux in• such an assembly, and then once again these
fluxes are used to generate assembly averaged group constants.
The final step is to use either these assembly-averaged group constants (or, in
very detailed calculations, the original fuel cell group constants) to determine the
flux and power distribution over the entire core. The symmetry of the core will
frequently allow one to consider only one quadrant in detail. It may occasionally
be desirable to homogenize the core still further to facilitate gross survey calcula-
tions (as in Figure 10-6c).
C. Cell-Averaging Techniques
The goal of most of our analysis in this chapter will be to calculate effective
group constants that have been spatially averaged over the flux distribution in a
lattice cell and therefore can be used to characterize the cell in subsequent
multigroup diffusion calculations in which the cell structure is ignored. To discuss
this subject of cell averaging in more detail, let us consider the calculation of the
cell-averaged group constant characterizing a two-region cell in which material
composition is uniform in each region. We will again denote these regions by "M"
and "F" for convenience (refer to Figure 10-2.)
The cell-averaged group constant characterizing a general cross section would
then be defined as
(10-13)
Such cell-averaged group constants are frequently referred to as self-shielded or

simply shielded group constants, since averaging over the spatial flux distribution in
the cell accounts for the flux depression caused by self-shielding. We can use the
fact that the cross section L(r, E) is constant within each region to rewrite <Lg\en
as
where we have defined the cell-averaged intragroup fluxes as
- - l ( 3
</>p(E)=y ), d r</>(r,E). (10-15)
F Vp
Now if in fact the neutron flux in the cell were separable in space and energy such
that
</>(r, E) = </>(r)i/;( E ), (10-16)
that is, if both regions of the cell were characterized by the same intragroup
spectrum, then one could collapse Eq. (10-14) to
( 10-17)
where the regionwise group constants are defined as
(10-18)
while the spatial flux averages are defined as
- I
<l>M = V ),( 3
d r<f>(r), etc., (10-19)
M VM
and the cell-disadvantage factor is defined as before,
(10-20)
Of course the neutron flux in a typical lattice cell is rarely separable in space and
energy. Nevertheless in many cell homogenization schemes such separability is
assumed in order to separate the determination of the neutron spectrum if;(E) to be
used as the weighting function in energy averaging from the calculation of the
disadvantage factor f to be used as the weighting parameter in spatial cell-
averaging. Such procedures begin by homogenizing fuel, coolant, clad and so on
over the fuel cell volume and then calculating the neutron energy spectrum if;(E)
for this homogenized cell using an infinite-medium spectrum code. Next this
spectrum is used to calculate group constants for each region of the cell [e.g., ~:
and ~~]. An auxiliary one-speed spatial calculation of the cell disadvantage factor
f is performed, usually at an average energy characterizing the group of interest,
and finally the regionwise group constants ~~ and ~: are combined using f to
obtain the cell-averaged group constant <~g\en· This scheme is indicated schemati-
cally in Figure 10-7.
A minor modification of this scheme would be to calculate an energy-dependent
disadvantage factor r(E) by simply allowing the one-speed cross sections appear-
ing in formulas for f to be energy-dependent. 5 Then one can interchange the spatial
and energy-averaging to first calculate the spatial average
(10-21)
I~g• I~B
t:g. Ii,g. Vg I~g
Direct calculation of cp(r,E) (THERM¢S)

Simultaneous space-energy average
NF
u! (E)
u[, (E)
v(E) u/ (E)
<I.(E)>cell
< It,(E) >c.11
<v!E>It(E)>,.11
FIGURE 10-7. Various schemes for computing cell-averaged group constants.
and then the energy average over i/;(E)
(10-22)
This latter scheme has also been sketched in Figure 10-7. It should be noted that in
most cases the use of energy-dependent disadvantage factors f (£) for cell averag-
ing yields results that are very comparable to those using the somewhat simpler
disadvantage factor f simply calculated at the mean energy of the group.
The most serious limitation of both of the above schemes is the assumption of
space-energy separability. One procedure for overcoming this is to first perform a
spectrum calculation for a homogeneous mixture of cell constituents. 6 The few-
group constants generated from this homogeneous spectrum calculation are then
used in a few-group one-dimensional spatial calculation of the flux in the cell. The
spatial flux distribution obtained in this calculation is then used to determine the
appropriate disadvantage or shielding factors. These factors can then be used to
adjust the fine-group cross section data used in the spectrum calculation. The
spectrum obtained from these shielded or cell-averaged fine group constants will
usually be much closer to the average spectrum in the cell, and averaging the
shielded fine-group constants over this spectrum will usually provide adequate
cell-averaged cross sections.
In certain cases it may be necessary to perform a detailed calculation of the
spatially dependent spectrum cp(r, E) in the cell and calculate the cell-averaged
group constants directly from the rigorous definition given by Eq. (10-13). Because
this latter procedure is becoming increasingly common in providing cell-averaged
group constants for accurate multigroup diffusion estimates of core parameters, we

will consider one such direct approach based on collision-probability methods in
Section 10-11-C.
It is important to stress once again that the effect of fine-scale lattice
heterogeneities enter reactor calculations through group constants that are adjusted
to account for the spatial variation of the neutron flux within the lattice. Most
frequently such group constants are modified by using disadvantage factors for the
lattice cell of interest. Hence the primary effort involved in accounting for lattice
effects involves the calculation of these disadvantage factors. As we will find,
diffusion theory is usually inadequate for such calculations, and transport methods
must be used.
We perhaps should mention that there are other prescriptions that prove very
useful for defining cell-averaged absorption cross sections in the instance that the
central region ("F") is very highly absorbing (as it would be for either a fuel pin or
a control rod). Such schemes usually proceed by demanding that the rate of
absorption calculated with the effective cell-averaged cross section be identical to
the true absorption rate in the region F.
For example, if L!ff is the effective cell absorption cross section, while -;phom is the
average flux in the cell which has been homogenized, then obviously we require
(10-23)
(Here we have neglected the absorption of the region M.) Hence the effective
absorption cross section for the cell is given by
(10-24)
where we have defined the self-shielding factor
-
</>F
is=---, (10-25)
</>horn
which accounts for the flux depression in region F. Therefore to calculate L!rr we
need only scale the cross section for the fuel region by ( VF/ Vceu)f.-that is,
decrease L: by the volume fraction occupied by the fuel, as well as by the flux
depression or self-shielding factor fs. Since -;phom = (VF/ Vceu)-;pF + ( VMl Veen) -;pM
we can write
(Veen/VF) (Vcen/ VF)

= = (10-26)
1 + (VM/ VF)(-;pMj-;pF) 1 +(VM/ VF)t
[Note that we can then identify Eq. (10-24) as equivalent to our earlier result for
the case in which LZ1~0.]
Yet a third scheme frequently applied to situations in which the region F is very
highly absorbing (e.g., a control rod) is to equate:
Absorption rate = V Leff~ = S J = rate at which neutrons (l0- 2?)

in cell cell a horn F F enter the region F '
where SF is the surface area of region F, while J F is the average neutron current
density into the absorbing region at this surface. Then
eff_ SF JF
}: ---- (10-28)
a V -
cell <Phom
Once again we are faced with determining a factor characterizing the spatial
variation of the flux in the cell, JF/~hom·
Hence all of these homogenization schemes require a detailed study of the spatial
variation of the neutron flux in the cell. We will separate our study of the spatial
transport of neutrons in a lattice cell into two parts. First we will concern ourselves
with thermal neutrons and then later with fast neutrons. The physics of each of
these problems is a bit different, since the dominant lattice effects on thermal
neutron behavior enter as a decrease in the thermal utilization, while the dominant
fast neutron effects are to enhance the resonance escape probability and the fast
fission factor.
II. HETEROGENEOUS EFFECTS IN THERMAL NEUTRON

PHYSICS
A. Thermal Utilization, Disadvantage Factors, and Cell-Averaged

Thermal Group Constants
As we have seen, the core of a nuclear reactor is comprised of thousands of
individual fuel cells, each characterized by the fuel element itself, usually some
cladding material (separated from the fuel element by a gap), and an adjacent
moderator that may also serve as a coolant (see Figure 10-5). We have also noted
that it would be prohibitively expensive to perform a detailed few-group diffusion
calculation taking into account the detailed configuration and composition of each
fuel cell-indeed, the very high absorption present in the fuel element renders the
use of the diffusion approximation itself highly questionable.
Instead the approach is to perform a detailed calculation of the flux in only one
typical fuel cell and then to use this flux to spatially average the thermal group
constants over the cell. Let us illustrate this by an example: Consider a typical
square lattice fuel cell with a cylindrical fuel pin. One first replaces this geometry
by an equivalent unit cell to simplify the calculations, taking care to preserve the
same volume of fuel VF and moderator VM·
We will now calculate the thermal group constants for this cell using the simplest
of the cell-averaging techniques described in the previous section. That is, we first
perform a thermal spectrum calculation for an infinite medium of identical com-
position as the cell, however, with the number densities of fuel and moderator (and
clad) being uniformly distributed over the cell. Then using this homogeneous cell
spectrum 1¥hom(E), we compute the thermal group constants characterizing each
cell region, for example, L~, L;;.1, L~, Li,, and PLi- (Here we have left the thermal
group index as understood.) If we can then determine the disadvantage factors
characterizing the cell r, we can compute the cell-averaged thermal group constants
(for a two-region cell) as
~;, + L;;1( V Ml VF){

<Lir\en = 1 + ( V Ml VF){ '
(10-29)
We could also calculate the thermal utilization for the cell as
(10-30)
Thus we once again find that the key to including heterogeneities in the generation
of thermal group constants rests on our ability to estimate the spatial dependence
of the thermal flux in the cell-that is, to determine the thermal disadvantage
factor r,
For the large natural uranium-graphite moderated reactors of interest during the
early years of the nuclear energy program, one could utilize one-speed diffusion
theory to calculate r, However in the more highly enriched and tightly packed core
lattices utilized in today's modern power reactors, diffusion theory estimates are
quite poor. Hence we will describe two alternative schemes useful for determining
the thermal disadvantage factor, both of which are based on transport theory. The
first method is an analytic scheme first proposed by Amouyal, Benoist, and
Horowitz,7 and referred to as the ABH method. It relies on the concept of an
escape or collision probability characterizing neutron transport in the cell. The
second scheme we shall discuss actually solves the energy-dependent neutron
transport equation directly for the equivalent cell of interest to generate the thermal
flux and perform spatial averaging over the cell. This scheme, known as THER-
MOS8, discretizes the integral form of the transport equation, thereby reducing it to
a system of algebraic equations solvable on a computer.
There are still other methods for calculating spatially dependent thermal neutron
spectra, ranging all the way from the assumption of a fundamental spatial mode
(such as was used in the BN or PN methods discussed in Chapter 8) to direct SN
solutions of the transport equation or Monte Carlo calculations. The choice of the
method will depend on both the detail required in the design and computer
capability (and allowable expense). In recent years, the trend has been toward
more careful treatment of the spatial detail of the lattice and more precise transport
descriptions, with increasing use of Monte Carlo techniques. 9
B. Escape Probability Methods in Cell Calculations

We will first investigate an analytical approach for the determination of cell
disadvantage factors that utilizes the concept of the probability that a neutron born
in one region will escape the region before suffering a collision. The use of such
"escape" probabilities is quite common in studying neutron transport in highly
absorbing media (e.g., fuel). 10-- 12 We will outline the specific application of such
concepts in a scheme, developed by Amouyal, Benoist, and Horowitz (the ABH
method).
The ABH method combines aspects of transport theory, collision probability
methods, and diffusion theory to calculate the thermal disadvantage factor
characteristic of a unit fuel cell. More specifically, diffusion theory is used to
describe the flux in the moderator, although a transport correction is introduced
into the boundary condition at the moderator-fuel interface. Neutron transport in
the fuel is described by multiple collision escape probabilities. All of these calcula-
tions are performed in the one-speed approximation and yield !; in terms of the
one-group constants characterizing materials in the cell.
Before we begin our description of the method, let us be a bit more precise in our
definition of a unit fuel cell. We will make four assumptions concerning the
treatment of the cell:
(1) There are no neutrons slowing to thermal energies in the fuel region. This
assumption is quite reasonable in thermal power reactor lattices, since
moderation in the fuel occurs only via inelastic scattering or elastic
scattering from admixed materials such as oxygen or carbon, both of
which yield inconsequential moderation when compared to the surround-
ing moderating region.
(2) We can treat the spatial distribution of neutrons slowing down within the
moderator region as uniform. To motivate this assumption, notice that in
most cases the distance required to slow a fission neutron to thermal
energies (\,/(yr ) is quite large compared to the lattice spacing or pitch.
For example, in a LWR therms distance to slow down is roughly 15 cm,
compared to a typical lattice pitch of 1-2 cm. Hence the neutrons slowing
down within any moderating region come from a large number of the
surrounding fuel pins. This tends to yield a more uniformly distributed
slowing down source.
(3) We will finally assume that there is no net flow of neutrons between the
cells. Surely this requirement would be true for an infinite lattice of
identical fuel cells. It would be expected to break down only near the core
boundaries or near to control elements or fuel pins of nonuniform
enrichment or composition (e.g., Pu-loaded fuel pins, which are
characterized by strong absorption resonances in the ~.5 eV range).
(4) We will assume that a one-speed treatment of the neutron flux in the cell
is sufficient. Actually, a one-speed calculation ignores the fact that the
flux in the cell is frequently inseparable in space and energy, but provided
the proper thermal group constants are used in the one-speed treatment,
this latter correction can frequently be ignored.
We will now use this model to calculate the thermal disadvantage factor f
characterizing the cell. Actually the ABH method for calculating f proceeds
somewhat indirectly by first calculating the thermal utilization f for the cell, and
then using this quantity and Eq. (10-6) to infer the disadvantage factor f as
(10-31)
The key quantity involved in calculating the thermal utilization is the probability
that a neutron appearing uniformly and isotropically in the moderator (i.e., slowing
down source) will be absorbed in the fuel region. The calculation of such prob-
abilities is far from trivial, particularly if one of the cell regions is sufficiently
highly absorbing to invalidate diffusion theory. However approximate schemes can
be used to calculate these probabilities that utilize the concepts of escape and
collision probabilities.
1. ESCAPE PROBABILITIES
To be more specific, we will consider a two-region cell such as that illustrated

in Figure 10-8. Then we define the absorption probabilities characterizing the two
regions of the cell as
P = probability that a neutron appearing uniformly and isotropically

pp- in region F will eventually be absorbed in region F,
(10-32)
p = probability that a neutron, appearing uniformly and isotropically
FM in region F will eventually be absorbed in region M,
and similarly for PMM and PMF· Notice that since the cell has only two regions
which incorporate all space, we must have
Ppp+PFM= 1
(10-33)
PMM+PMF= 1.
FIGURE 10-8. The absorption probabilities characterizing a two-region cell.

The calculation of these probabilities is complicated by the fact that they must
account for scattering processes within each region as well as instances in which a
neutron may scatter back and forth between the regions several times before finally
being absorbed.
Although these absorption probabilities may be very difficult to calculate in
practice, they do exhibit several simple and useful properties. The absorption
probabilities characterizing cross-region transfer obey a reciprocity relation of the
form:
(10-34)
This very useful relation can be proven (see Problem 10-10) by noting that the
Green's function characterizing the cell is symmetric such that G (r, r') = G (r', r)
(whether described by one-speed diffusion or transport theory 11 ).
This feature of absorption probabilities is particularly useful in the calculation of
the thermal utilization f, for by definition f is just the probability that a thermal
neutron is absorbed by the fuel. That is, f is just PMF• since we have assumed that
all thermal neutrons first appear as a slowing down source uniformly distributed in
the moderator. We can therefore use Eq. (10-34) to write
(10-35)
Thus we are now faced with calculating the absorption probability for the fuel
region to moderator region P FM· This quantity is somewhat easier to approximate,
since the fuel is usually sufficiently highly absorbing that neutrons tend to make
very few collisions in the fuel region before being absorbed.
So how do we calculate PFM? We begin by breaking it up into two factors
PFM=PF/3M (10-36)
where
p = probability that a neutron born uniformly and isotropically

F in the fuel escapes from the fuel before being absorbed,
(10-37)
/3 = conditional probability that the neutron, having escaped
M- from the fuel, will then be absorbed in the moderator.
We will refer to PF as the escape probability characterizing the fuel region, since it
gives the probability that a neutron eventually escapes the fuel (possibly after a
number of scattering collisions) without being absorbed. In a similar manner we
can identify /3M as a measure of the blackness of the moderator region, since it
measures the probability that a neutron entering this region will be absorbed. Our
goal then is to calculate the escape probability PF for the fuel region and the
conditional "absorption" probability /3M for the moderator.
A word of caution at this point is advisable. There are a variety of different
definitions of "escape probabilities" and "collision probabilities" floating around in
the reactor physics literature. Since this concept was originally introduced to
describe transport in purely absorbing media in which any collision within the fuel
resulted in neutron absorption, 10 the escape probability is occasionally defined as

the probability that a neutron escapes the region without suffering any kind of a
collision. We will denote this latter concept as the first-flight escape probability,
however, since it characterizes the neutron streaming out of the fuel without
interacting.
2. FIRST-FLIGHT ESCAPE PROBABILITIES AND COLLISION

PROBABILITIES
Since the fuel is usually very highly absorbing, one can frequently calculate
the escape probability PF in terms of a much simpler quantity Pp 0 , the first-flight
escape probability defined by
probability that a neutron originating uniformly and iso-

p FO = tropically in the fuel will make its next collision in the
moderator (that is, will escape the fuel without making
a collision). (10-38)
The calculation of PFo is simply a geometrical problem, since one need only
determine the probability that a neutron will stream out of the fuel region before it
collides with anything. The first-flight escape probabilities PFo have been calcu-
lated10 and tabulated 13 for most of the common fuel geometries and we will avoid
wading through the solid geometry necessary to repeat these calculations here. It is
useful to note several specific cases, however. For a very small fuel region we
evidently must have
small
PFo ~ 1 (10-39)
fuel
lump
(since the probability of leaking out before suffering a collision must be almost
FIGURE 10-9. The first-flight escape probabilities for a two-

region cell.
1.0
0.9
0.8
0.7
0 0.6
~
0.5
0.4
0.3
0.2
0.1
2 3 4 5 6 7 8 9 10
FIGURE 10-10.
~.
First-flight escape probability Pp0 10•
unity). For a large fuel region such that the fuel dimensions are large compared to
a mfp, one finds 10
large S
p ~ F (10-40)
FO fuel 4V ~F
lump F t
where SF is the surface area of the fuel region while ~r

is the total cross section of
the fuel. The collision probability for intermediate size regions (size comparable to
a mfp) is more complicated. We have included a plot of PFo for several common
geometries in Figure 10-10.
A closely related concept is the collision probability characterizing a region
probability that a neutron originating uniformly and iso-

p Fe= tropically in the fuel will make first collision in the ( 10-41)
fuel before escaping.
From this definition it is apparent that
Ppc= 1-PFo· (10-42)
We will find it more convenient to work with the first-flight escape probability Ppo•
In the special case of a purely absorbing or "black" fuel lump, the escape
probability PF is just equal to the first-flight escape probability Pp 0 , since any
collision results in absorption. More generally, however, the neutron may scatter
several times before being absorbed or escapi,ng. Hence we should really write PF
in terms of "multiple collision" escape probabilities

00
Pp=Ppo+Pp1+Pp2+··· = ~ PFn• (10-43)

n=O
where the PFn are generalizations of the first flight escape probability Ppo which
describe the probability that the neutron escapes from the fuel lump after
scattering n times within the fuel. We can calculate these quantities in terms of Ppo
if we assume that the distribution of first, second, and further collisions is uniform.
For example, -: . /·I
j?dt~
~-- f
(10-44)
and so on. Thus under the assumption of spatially uniform collision densities, we
find the escape probability is given by
"' .J__
l. }";;
P F =PFO 1+(1-PFO)-+(1-P )
2( -}:; )2 +··· -[~;~-;;-]
Cl· s v,::-v
}:F FO }:F
[
_J_ ---.,_~,[-., J t
1
t "LI: f r-0
. "(, --,- LF(l-P)-
-~l = 1+ _a FO - , (10-46)
'ff:-Ls+f>r-vTr · Ee; -1/J!=ol.s lf

where we have summed the geometric series. Actually in the ABH method this (
result is improved somewhat by accounting for a nonuniform distribution of first
collisions. 7 In this case PF is given for a cylindrical fuel pin of radius a by
where the coefficients o: and /3 are given for cylindrical fuel in Figure 10-11. [Note
that if o:=/3=0, this is identical to Eq. (10-46).]
3. THE ABH METHOD

Our final task is to calculate /3M, the probability that a neutron escaping the
fuel will be absorbed in the moderator. In the ABH method, this is accomplished
by combining Eqs. (10-35) and (10-36) to write
(10-48)
oc:E~
FIGURE 10-11. The ABH parameters oc and {3. 1
To calculate /3M, we now approximate PF by using its value for a large, purely
absorbing fuel lump
(10-49)
Then
(10-50)
Now all we need do is calculate PMF' the transfer probability characterizing source
neutrons born in the moderator escaping to the fuel where they are then absorbed.
Ah, but isn't PMF just the thermal utilization, that we were trying to calculate
originally? Yes, that is true. However the key idea in the ABH method is to note
that one can use a relatively crude scheme to calculate PMF in obtaining /3M,
provided the other quantities used in computing fare treated adequately. That is,
one can rewrite Eq. (10-35) as
(10-51)
and then note that since the second term on the right-hand side is small, relatively
crude approximations of it will suffice. If we rewrite Eq. (10-51) usingf-PMF and
Eq. (10-49), we find
(10-52)
Hence we need only estimate PMF· This can be done to sufficient accuracy using
diffusion theory with a transport-corrected boundary condition at the fuel-
FIGURE 10-12. Fuel cell geometry for calculating PMF·
moderator interface and assuming PMF~PM. That is, we solve

subject to boundary conditions:
(a) dq,M I =0 (10-53)

dr r=b
(b) _l_ dq,M I =

1
-
<l>M dr d
r=a
for the geometry indicated in Figure 10-12. Here we have used a transport-
boundary condition at r= a, involving a parameter d which is given in Figure 10-13
in terms of the fuel radius and the transport mean free path. We need only solve
this diffusion problem, and then use the fact that
PM= 21raDM dq,M I (10-54)

q0 VM dr a
in order to determine PM. Avoiding the details, we will only give the final result
(10-55)
where L?.t. = DM/~~' and E is a lattice function,1 which for cylindrical geometries
takes the form
1 t )K
0( 1( -k )+K t )1 (-k)
0( 1
(10-56)
/1( 1M )K1( :M )- K1( 1M )11( :M)
1 .4 ,----,-----.-----r---r----r-----r----r---~--.---~
">:Si,J' , .2
2lC:
l9
"'
'o Probable shape of dependence for a
C:
0
cylinder of finite radius
-~ 1.0
0 Large black cylinder
C.
E
x.,
~
IllC:
:.J 0.8
0.6 =---'-----:----'---~---'---_j_--...J._--..J__--J...._-__J
0 2 3 4 5
Radius of black cylinder_B__
A"
FIGURE 10-13. The transport boundary correction d.13
Hence our final expression becomes
(10-57)
Similar results for a three-region cell (including clad) can be obtained. 5 • 14

The ABH method actually gives remarkably good results. For a range of
fuel-to-moderator ratios, the values of ~M/~F calculated by the ABH method are
within 0.5% of those calculated by a transport theory analysis. In Figure 10-14 we
have compared the ABH method results for a cylindrical pin with the results of a
more accurate transport calculation (THERMOS).
It should be noted that the disadvantage factors calculated by the ABH method
depend on various macroscopic cross sections. The most common scheme is to
consider these cross sections as being evaluated at the mean energy of the
Maxwellian spectrum characterizing the moderator temperature. However it is
possible to also regard the disadvantage factors as depending implicitly on energy
via the energy-dependence of the cross sections that appear in them. That is, one
can calculate a disadvantage factor for each energy point utilized in the thermal
spectrum calculation. These energy-dependent disadvantage factors can then be
utilized to determine cell-averaged group constants, as we indicated in Section
10-1-3.
l 1
/ 0 · t) 1 t O, 1/ 10 1 , J ,
/ 1 ,
1
1.8
..,,
1.6
NH/Nu= 4.0
j·
...0
.....,
(.)
CIJ
....,.,C
C)
1.4
~/0
"O
i5
>
I)!
1.2
4:1/□/ '
po
□__.,.,---
□~
1.0
0.25 0.4 0.55 0.7 0.85 1.0
Fuel Radius (cm)
1.35
I I I I
a = 0.4915 cm
A-;::::-
1.28 - c.~+
,::::::::.+ ..,--
..,,
....,0
/+ X
(.)
1.21 - +ye. / 0-- -
7~
LL
CIJ
C)
~
.,
C ~o
> 1.14 - -
.,
"O
/o/x
i5"'
X
1.07
I -----□--
~
----□
1.0 ---□I I I I
0 2 3 4 5
NH/Nu
+ Thermos (Nelkin Scattering Kernel)

c. A8H (using Nelkin Kernel)
O A8H (using proton gas Kernel)
□ Diffusion Theory (using proton gas Kernel)
X BN L Experimental Values
FIGURE 10-14. Comparison of Disadvantage Factor Calculations for Cylindrical Fuel Pins of
1.3% Enrichment Fuel, H 20 Moderator.
424
C. Integral Transport Methods for Thermal Cell Calculations

(THERMOS)
Thus far we have described methods for including lattice heterogeneities in
thermal spectrum calculations that essentially separate the treatment of the spatial
and energy dependence of the neutron flux. That is, the spatial behavior of the flux
in the unit cell is treated in the one-speed approximation (such as in the ABH
method), and then the thermal disadvantage factors calculated from this spatial
analysis are used to generate cell-averaged or "self-shielded" thermal group con-
stants using the results of an infinite homogeneous medium thermal spectrum code.
Such a scheme is characterized by a minimal calculational effort while still
producing results of sufficient accuracy for many reactor design calculations.
However occasionally a more detailed analysis of the spatial dependence of the
thermal spectrum in a fuel cell is required, for which the assumption of space-
energy separability is inadequate. For such calculations it is usually also necessary
to take into account the more detailed nature of thermal neutron scattering in the
moderator using the scattering kernels generated by more elaborate methods (e.g.,
GASKET) than those such as the Wigner-Wilkins scheme.
Such an analysis usually requires a detailed solution of the energy-dependent
transport equation characterizing the cell. Perhaps the most popular of such
transport methods is the THERMOS code developed by Honeck, 8 which uses an
integral form of the transport equation to calculate the spatially dependent thermal
neutron spectrum in a cell characterized by a one-dimensional symmetry (usually
cylindrical). The feasibility of the approach used in THERMOS relies heavily on
the assumption of isotropic scattering.
To sketch the method, let us first recall the form of the energy-dependent
transport equation, assuming isotropic sources and scattering:
O·V<p+~r(r,E)q:i= -
I lE l dO'~.(r,E'-E)q:i(r,E',O')+ S(r,E) ,
cdE'
A A
4 wo 4.,,. 4w
( 10-58)
where we have noted that for such thermal spectrum problems, one is usually
interested in energies E below some cutoff energy Ec (typically of the order of 1
eV). The source term actually represents neutrons slowing down below Ec from
higher energies:
(10-59)
Now notice that the transport Eq. (10-58) contains only a single derivative in
space. Hence one can use standard integrating factor techniques (refer back to
Section 4-IV-C) to integrate this equation over space. After a subsequent integra-
tion over angle, one arrives at the so-called "integral form" of the neutron
transport equation
where
T(r,r',E)
exp[- t-•ld, }:,(r-,P[ ,E) l (10-61)
4'1Tlr-r'l2
Note here that T(r,r',E) is, in fact, the uncollided flux at r of energy E from a unit
point source at r' of the same energy E. It is sometimes known as the transport or
first-flight kernel, and we will later see that it is closely related to an escape
probability.
This equation is now solved numerically for a unit fuel cell, assuming zero net
neutron current across cell boundaries and a uniformly distributed slowing down
source only in the moderator. The integral transport equation is first written in
multigroup form
<1>g(r) = f 3
d r' T(r,r',Eg )[ g~
1
L.,.• (r')</>g,(r') + Sg (r')],
1..;g..;G. (10-62)
To handle the spatial variable, the cell is divided into N subregions. One then
assumes that <1>g(r) is spatially independent within each subregion such that Eq.
(10-62) can be written as
(10-63)
where
(10-64)
Notice here that T;.n can be interpreted as the transfer or coupling coefficient
characterizing neutron transport between subregions n and m. In fact, if we denote
by P;.n the probability that a neutron of energy Eg appearing in region m will suffer
its next collision in region n, then the escape probabilities P!,n are related to the
coupling coefficients T!,n by:
g Vn
Pmn= y~,, Tgmn'
m
'S;'n
(10-65)
The THERMOS method first calculates the coupling coefficients T!m for the cell
of interest, usually by numerically performing the integration indicated in Eq.
(10-61). Then the N x G multigroup, spatially discretized equations are solved using
standard iterative techniques (e.g., overrelaxation). As one might expect, the
calculation of the T!m is very time-consuming. Furthermore since we are actually
solving a discretized integral equation in space (as opposed to a differential
equation), it is not surprising to find that the matrices involved are full (that
restricts one in practice to using a small number of subregions).

THERMOS works best for tight, highly absorbing lattices with rapidly spatially
varying properties. Once having obtained the spatially dependent flux in the cell
cf>(r,E), one can directly calculate the cell-averaged or self-shielded cross sections as
f d 3r N; (r)a(i)( E )cf>(r, E)
<~(;)( E )) cell= _vc_e1_1_ _ _ _ _ _ __
(10-66)
f Vcell
d 3rct,(r,E)
Of course one can also calculate thermal disadvantage factors,
(10-67)
The problem with codes such as THERMOS which attempt to calculate in some
detail the spatial dependence of the flux in a unit fuel cell is one of cost relative to
less sophisticated schemes such as the ABH method. Typical running times for
such detailed transport codes are several orders of magnitude longer than those for
schemes based on relatively simple estimates of the disadvantage factors. Hence
the usual procedure is to calculate the fine structure within fuel cells using
THERMOS only in detailed design studies. These results are then used to
homogenize or self-shield group constants for a few group, two-dimensional
diffusion or transport calculation performed on a fuel subassembly or group of
subassemblies. The procedure is usually augmented by a few experimentally
obtained corrections.
III. HETEROGENEOUS EFFECTS IN FAST NEUTRON

PHYSICS
A. Resonance Escape Probabilities in Lumped Fuels
l. THE SLOWING DOWN EQUATIONS FOR A TWO-REGION CELL

In our introductory discussion of lattice effects on core multiplication, it was
stressed that fuel lumping can cause rather dramatic changes in resonance
absorption due to self-shielding effects. Indeed the effective resonance integrals for
the fuel can be decreased from their homogeneous values by as much as an order
of magnitude. Hence it is essential that we discuss schemes for calculating the
resonance integrals characterizing heterogeneous lattices. It is evident that such
schemes must account for the spatial dependence of the flux in the fuel cell. Once
again the method we will use is based on the concept of collision or escape
probabilities-that is, the probability that a neutron originating in one region will
make its next collision in another region.
For convenience, we will consider the fuel cell to be composed of only two
species, fuel and moderator. (The extension to multiple isotopes or moderator
admixed into the fuel is given in many of the standard references 15- 17 on this
subject.) Our first task is to write a balance equation describing the neutron flux in
the cell. Recall that the balance equation describing neutron slowing down in an
infinite, homogeneous medium is
We will develop a generalization of this equation to account for a heterogeneous

fuel cell by defining the energy-dependent first-flight escape probabilities:
P Fo( E) = probability that a neutron of energy E originating in the fuel wiH make
its next collision in the moderator (that is, will escape the fuel without
suffering a collision),
PM 0 (E) = probability that a neutron of energy E originating in the moderator will
make its next collision in the fuel.
We can now use these first flight escape probabilities to develop a generalization of
Eq. (10-68). For suppose we interpret cf,F(E) and cf,M(E) as the volume-averaged
flux in the fuel and moderator, respectively. Then for example,
(10-70)
represents the average rate at which neutrons slow down to an energy E in the
moderator. Hence by multiplying this expression by PM 0 (E) we can compute the
rate at which neutrons of energy E are transferred from the moderator into the
fuel. In a similar fashion, one can calculate the rate at which neutrons slowing
down to energy E in the fuel suffer their next collision in the fuel as
(10-71)
The sum of these contributions must equal the total rate at which collisions are
occurring in the fuel; hence we arrive at the balance relation for the fuel
(10-72)
CELL CALCULATIONS FOR HETEROGENEOUS CORE LA1TICES / 429
We can write a similar balance relation for the moderator region:
(10-73)
These two equations represent the generalization of the slowing down equation
(10-68) to a heterogeneous two-region cell. As they stand, these coupled integral
equations are exact. However they are only formal until the escape probabilities
Pp 0 (E) and PM 0 (E) have been specified (in much the same sense that the
multigroup equations were also exact but of only a formal significance until the
multigroup constants were determined). In practice, these escape probabilities are
usually computed only approximately by assuming that the source neutrons in each
region l;lppear uniformly and isotropically. (This is referred to as the flat source
approximation.)
These escape probabilities in effect allow us to separate the treatment of the
spatial and energy variables in the study of neutron slowing down in the cell. The
first-flight escape probabilities Pp 0 (E) and PM 0 (E) will be calculated by consider-
ing spatial neutron transport at a given energy E. Once determined, these escape
probabilities will be inserted into the slowing down equations involving only the
energy variable. First we will proceed to apply Eqs. (l 0-72) and (l 0- 73) to the study
of resonance absorption in lattices, and defer until later the calculation of the
escape probabilities.
It is possible to decouple these equations by making the narrow resonance
approximation for the moderator
(10-74)
so that we can replace the average flux in the moderator by its asymptotic form
(10-75)
[Here we have again normalized the flux far above the resonance by removing the
factor l}:. in q,(E)~l/l}:.E, where for a heterogeneous cell, l}:;.=(lF~~VF
+lM}:~VM)/ Veen· Then in this region, '1>M(E)~'/>p(E)~l/ E.] If we now use this
in the second term on the RHS of Eq. (10-72) corresponding to slowing down in
the moderator, we find that it becomes
E
f.
"M '2:~( E ')'1>M( E ') }:~
VMPM 0 (E) dE' (l-a )E' = VMPM 0 (E)E. ( 10-76)
E M
Hence by substituting this NR form into Eq. (10-72) we will have eliminated the
appearance of the moderator flux in the equation for the fuel region, thereby
decoupling this equation from the moderator region balance equation.
It is useful to make one further manipulation before inserting the NR
approximation for the moderator into Eq. (10-72). One can demonstrate that the
first-flight escape probabilities, 17 when calculated assuming a flat source in each of
the fuel and moderator regions, must satisfy a reciprocity theorem similar to Eq.
(10-34):
(10-77)
If we furthermore assume that absorption is negligible in the moderator

[L;'1(E)~L~(E)], we can rewrite Eq. (10-76) as
(10-78)
Hence our slowing down equation for the fuel region becomes
LF(E),I,. (E)=[l-P (E)]

t 't"F FO
i E
°'F
dE'
LF(E')cf, (E')
s F
(1- ap)E' +
Once we know Pp 0 (E) we can solve this equation for cf,p(E)-either using
p
FO
(E)LF(E)
£
t
.
(10-79)
analytical approximations or direct numerical methods-and thereby calculate the

effective resonance integral for the resonance:
(10-80)
(where once again we should keep in mind the normalization that cj,(E)~l/ E far
above the resonance).
2. APPROXIMATE CALCULATIONS OF THE RESONANCE INTEGRALS

We will now introduce the standard NR and NRIM approximations into our
calculation of the resonance integral. First consider the narrow resonance
approximation (NR) in which we can approximate
E
( °'F L'; ( E ')cf,p( E ')
' L;
(10-81)
JE dE (1-ap)E' ~E.
If we substitute this into Eq. (10-79), we can immediately solve for the flux in the
fuel as
(10-82)
Hence the NR approximation to the resonance integral becomes just
( 10-83)
where we have cancelled out the fuel number density NF.

It is customary to express the first-flight escape probability Pp0 (E) in terms of a
"pseudo" cross section ae known as the escape cross section, which characterizes
neutron removal from the fuel lump via leakage to the moderator:
(10-84)
That is, if ae is proportional to the probability of a neutron leaking out of the fuel
before suffering its next collision, then since ae + ar characterizes neutron loss
either via absorption or collisions with fuel nuclei, this physical interpretation of ae
is consistent with our earlier interpretation of the escape probability Pp0 (E). Of
course, in order to determine ae we must first calculate Pp0 (E) and then use Eq.
(10-84). However ae is particularly useful since it tends to be only weakly depen-
dent on energy (much like a potential scattering cross section).
If we insert Eq. (10-84) into the resonance integral and note
(10-85)
we find
( 10-86)
But recall that our earlier result for a homogeneous system was
(10-87)
Hence it is apparent that we can obtain the heterogeneous result [Eq. (10-86)) from
the homogeneous resonance integral [Eq. (10-87)) by simply replacing
( 10-88)
These results give rise to what are commonly referred to as the equivalence
relationships for resonance absorption: (a) heterogeneous lattices with the same
values of ae will have the same resonance integral, regardless of the surrounding
moderator and (b) a heterogeneous lattice with a given value of ae will have the
same resonance integral as a homogeneous reactor with~~/ NF-ae.
cf~
NJ( i ,,,
A very similar result can be obtained for the NRIM approximation. In this case,
one finds
(10-89)
Of course before proceeding further we must still determine how to calculate the
escape cross section ae or equivalently, the first-flight escape probability PF0 (E).
Recall that we had already assumed that PF0 (E) and PM0 (E) could be calculated
using the flat source approximation. Although this approximation might be
reasonable far from the resonance energy, it is certainly not strictly valid at this
energy since the flux in the fuel element is quite strongly varying due to
self-shielding. However such an approximation is found to be adequate in most
applications of interest.
Hence we are now faced with calculating PF0 (E), the probability that a neutron
born uniformly and isotropically in the fuel makes its next collision in the
moderator. Fortunately, the resonance integral is not overly sensitive to the
detailed behavior of PF0 (E). In fact, it is usually sufficient to introduce a
particularly simple approximation for PF0 (E) first suggested by Wigner. 18 First
recall that we know the limiting behavior of PF0 (E) for both small and large fuel
lumps. For small fuel lumps, obviously
as (10-90)
For large fuel lumps, we can effectively use the black lump result
s
p (E) F as (10-91)
FO ➔ 4V ~F
F t
With these limits in mind, Wigner chose a simple interpolation formula between
the limits (known as the Wigner rational approximation)
(10-92)
This rather crude approximation is usually valid to within 10% in most lattices of
interest. If we now compare Eq. (10-92) with our earlier definition of ae given by
Eq. (10-84), we can identify
Rational
(Je (10-93)
approximation
Here we have noted that 4 VF/ SF can be shown to be geometrically equal to the
average length of a chord drawn across the fuel region. 10• 19 (See Problem 10-15.)
Hence the only place that the fuel geometry enters is into the escape cross section
oe. As we have seen, we can now adapt our earlier homogeneous resonance integral
results to heterogeneous lattices by merely replacing~~ /NF-ae~Sp/4 VFNF.
To proceed further, notice that we can rewrite Eq. (10-83) as two terms
(10-94)
where the first term is identical to the NR approximation for an infinite medium
composed only of fuel, while the second term depends on geometry through PFo·
From Eq. (10-92) it is apparent that Pp0 ~Sp/ VFNF~(Sp/ Mp) in the rational
approximation, where MF is the mass of the fuel lump. Hence Eq. (10-94) suggests
that we can write the resonance integral in the form
(10-95)
As we will see in a moment, this relationship 1s born out by experimental

measurements.
If we now substitute in the Doppler-broadened resonance forms
(10-96)
(where we have again neglected interference scattering), we can write Eq. (10-94) as
(10-97)
where we recall
if;(tx)
J
oo
J(t/3)= o dx if;(tx)+/3' ( 10-98)
while
00
( Pp 0 (x)if;2(s,x)
L(t,t/3)= Jo dx if;(s,x)+ /3 , (10-99)
where
(10-100)
We have encountered the J(s,/3) function before. The L(t,t/3) function has been
tabulated for common lattice geometries. 20
A very similar result can be obtained for the NRIM approximation
(10-101)
where
( 10-102)
Since the NRIM approximation results in ignoring scattering from the absorber, we
have chosen to generalize our result somewhat by including a moderator admixed
into the fuel lump 15 • 16 (e.g., oxygen in UO 2 fuels) denoted by a potential scattering
cross section ~:m-
3. ROD SHADOWING AND THE DANCOFF CORRECTION
Thus far we have treated resonance absorption in a fuel lump as if the unit
fuel cell were truly isolated from other fuel cells. However, if the fuel rods in a
lattice are in fact separated by a moderator that is not many mean free paths thick,
it is possible for neutrons with energies in the resonance region to pass from one
fuel lump to another. This invalidates our earlier calculation based on collision
probabilities, for in that calculation we assumed that the escape probability from
the fuel PFo implied that on escaping from the fuel, the neutron would suffer its
next collision in the moderator. However if other fuel lumps are nearby, this next
collision might also occur in the fuel. Hence we should try to calculate a modified
escape probability P; 0 that a source neutron born in the fuel suffers its next
collision in the moderator-even though there may be adjacent fuel elements. Then
we can use our earlier analysis, merely replacing PFo by P; 0 .
We shall not develop a detailed derivation 21 • 22 of a modified collision probability
P;0 including rod shadowing effects here, but merely indicate that the result of
such a calculation is
( 10-103)
where <R )F is the average chord length characterizing the fuel lump [<R )F = 2R
for an infinitely long cylinder] while C is a tabulated parameter known as the
Dancoff-Ginsberg factor 22 that depends on the fuel geometry and cross section.
Actually this merely corresponds to increasing the average chord length in the fuel
FIGURE 10-15. Rod-shadowing effects.

lump by a factor (I - c)- 1-that is,
1-C
(10-104)
(R)F (R)F
(no shadowing) (shadowing)
Hence the Dancoff-Ginsberg factor simply decreases the effective leakage from the
lump (taking into account the possibility that the neutron might eventually be
absorbed in a neighboring fuel lump).
One can now merely use PJ 0 in place of PFo in the calculations of the resonance
integrals for the lattice of interest. The effect of such a correction is to decrease the
surface area of a single fuel element by a factor of 1 - C. This fact is of particular
use if one of the various correlations for the resonance integral is to be used. We
have given a short table of the Dancoff factor C in Table 10-1.
TABLE 10-1 Dancoff Corrections C for Two Parallel Circular Cylinders2 2
(radius/~,)
Pitch/~, 0 0.25 0.50 1.0 1.5 2.0
2.0 .182 .170 .160 .144 .132 .123

2.25 .155 .132 .114 .086 .067 .0522
2.5 .136 .107 .0849 .0550 .0364 .0245
2.75 .122 .088 .0649 .0362 .0208 .0122
3.0 .111 .074 .0505 .0244 .0122 .0062
4.0 .081 .040 .0205 .0057 .0016 .0005
5.0 .065 .024 .0091 .0015 .0003
6.0 .054 .015 .0043 .0004
7.0 .046 .0094 .0021 .0001
8.0 .040 .006 .0011
9.0 .036 .004 .0006
10.0 .032 .0028 .0003
Notice that decreasing the surface area of a fuel lump will reduce the
corresponding resonance integral from its value for an isolated fuel element. Hence
one finds that the Dancoff correction for neighboring fuel elements corresponds
effectively to a correction for the shadowing of one fuel element by another.
4. EMPIRICAL CORRELATIONS FOR RESONANCE INTEGRALS

A number of experimental measurements of resonance integrals have been
performed that parameterize the data as empirical correlations of the form of Eq.
(10-95). For example, the total resonance integral (i.e., over all resonances) in U0 2
at a temperature of 20°C can be written either23 as
[These forms can actually be derived from the expressions (10-97) and (10-101) in
certain limits.] Both of these expressions are found to work equally well over the
range of SF/ MF characteristic of U0 2 fuel lattices.
Experimental correlations have also been determined to characterize the

temperature-dependence of the resonance integral. For example, in the temperature
range from 20-600°C one can use23
(10-105)
where ,B~.006-.008, depending on fuel type and rod radius. A very convenient
correlation has been proposed by Strawbridge and Barry5 to characterize both
lattice parameters and temperature effects:
/
28
= 2.16x + 2.56 + [ .0279x - .0537] VT
where
}:: (1- C) ]1;2 (10-106)

x= [ NPFo+ (R) N '
28 F 28
while C is the Dancoff-Ginsberg factor and (R )F is the mean chord length of the
fuel pin. Typical results from this correlation are shown for several different fuel
temperatures in Figure 10-16.
26
• Hellstrand's correlation for U02 rods
+ Hellstrand's correlation for U-metal rods
24 - Metal-oxide correlation
22
20
e 1a
co
N
.....
16
14
12
10
8
3 4 5 6 7 8 9
X
FIGURE 10-16. Comparison of metal-oxide resonance integral correlation with Hellstrand's

correlations for isolated rods. 5
Although such correlations cannot be used directly in reactor calculations (since

the core lattice differs considerably from the pure U0 2 lattice described by the
correlation), they do prove extremely useful in normalizing resonance integral
calculations within fast spectrum codes, as we will see in Section 10-111-C.
B. Modifications in the Treatment of Fast Fission

Although the number of fast fission events occurring in a thermal reactor is
not large, such reactions can be quite significant since they can provide a sizable
fraction of the excess reactivity of a core. It is particularly important to take some
account of the enhancement of the probability of a fast fission reaction by fuel
lumping. This will cause the fast fission factor f to be increased from its value for a
homogeneous reactor core.
The calculation of heterogeneous modifications to the fast fission factor is
considered in detail in several of the standard references. 24•25 We will focus our
attention instead on the more practical question of just how heterogeneous effects
are included in the treatment of fast fission in conventional fast spectrum codes.
The rate at which fast fission reactions occur depends on the magnitude of the
fast neutron flux in the fuel region. From our earlier discussion, we expect this fast
flux (say, in the energy range .5-10 MeV) to be somewhat larger in the fuel than in
the moderator, since once neutrons have entered the moderator, they are rapidly
slowed down below the fast fission energy threshold. Hence we might expect that a
homogeneous fast spectrum calculation will tend to underpredict the amount of
fast fission occurring, since it will use the average flux characterizing a homo-
geneous system.
To correct this, one can follow a procedure very similar to that used in our
treatment of thermal fission by defining a fast utilization factor (similar to the
thermal utilization factor) which is the ratio of the neutron removal rate in the fuel
to the total rate of fast neutron removal. Notice that we include in this definition
all mechanisms for neutron removal, including neutron absorption, elastic and
inelastic scattering. Then it is evident in analogy to out treatment of the thermal
utilization that
(10-107)
Here LR is the removal cross section for fast neutrons. We will now introduce a
flux fast advantage factor, defined as
(10-108)
(Notice that at fission neutron energies, the fuel is now at an advantage in

competing with the moderator for neutron absorption). Using this definition and
our expression for fp, we can solve for the fast advantage factor in terms of the fast
utilization factor as
(10-109)
Once we have determined f F• we can calculate new group constants characterizing

the fast range that include the effect of fast flux enhancement due to fuel lumping.
Of course the success of this technique is contingent on our ability to calculate

JF' One could always use brute force methods based on Monte Carlo calculations,
but it is far more common to use alternative techniques 26 based on successive
collision probabilities very similar to those used in the ABH method. This latter
approach is particularly appropriate since the inclusion of lattice effects in fast
fission is not nearly so critical as in thermal group constant generation. Indeed
errors in the treatment of fast fission rarely lead to an error of over 0.1% in core
multiplication. Hence rather crude methods are usually sufficient for accounting
for lattice heterogeneities in fast fission for thermal reactor design. We will avoid a
detailed discussion of the application of the techniques develope<;i in Section
10-11-B and instead refer the interested reader to standard sources such as
Strawbridge and Barry.26
C. The Inclusion of Heterogeneous Effects in Fast Spectrum Codes

As we have seen, the most significant effects of a heterogeneous fuel ar-
rangement on the calculation of fast group constants enter into the treatment of
resonance absorption, although modifications are frequently also included in the
treatment of fast fission.
The dominant heterogeneous effect in resonance absorption in thermal reactors
is that of self-shielding, which substantially reduces the value of the resonance
integral. One can include this effect in the calculation of resonance integrals by
using collision probability methods. For crude treatments, one can simply use the
rational approximation to calculate the effective escape cross section for the lattice
geometry of interest, and then use ~~/NA ➔ ae in the usual homogeneous calcula-
tions. More accurate calculations based on a more realistic estimate of the escape
probability PFo require an evaluation of both the J(t/3) and L(t,t/3) factors.
Rod-shadowing effects are accounted for by using the Dancoff factor in the
calculation of /.
The resonance integrals for each fine-spectrum group can then be calculated and
used to calculate either the effective absorption cross section for the group
(10-110)
or the resonance escape probability for the group,
(10-111)
Of course to perform a detailed calculation of the Doppler-broadened resonance

integral for each resolved resonance would be prohibitively expensive. Hence one
customarily only accounts for Doppler-broadening and heterogeneous effects in the
more significant resonances. For example, in a slightly enriched uranium-fueled
thermal reactor, one might only perform a detailed calculation for the lowest
energy resonances of 238 U and 240Pu. The remaining resonances would then be
characterized by their zero-temperature, homogeneous resonance integrals.
The resonance escape probabilities for each group are adjusted (fudged) so that
the total resonance integral agrees with an empirical correlation (although the
amount of resonance absorption assigned to each microgroup in the fast spectrum
code will vary with the details of the calculation). For example, one scheme might
be to multiply every resonance escape probability p 0 for 238 U by the same fudge
factor L, and then vary L until / 28 when calculated assuming zero absorption for
all other elements agrees with the Strawbridge and Barry correlation5 [this is known
as the "w*-search," where w* = (1 - p 28 ) / p 28 ].
In fast reactor analysis, the small spacing between fuel elements and the large
neutron mfp allows one to largely ignore heterogeneous effects in the calculation of
resonance integrals. Furthermore for the high-energy resonances of importance to
fast reactor behavior,27 the NR approximation is usually valid. Other complications
arise, however, such as unresolved resonances and p-wave resonances, somewhat
complicating the analysis of resonance absorption in fast systems.
The cross sections characterizing the fast microgroups can be scaled to account
for fast flux peaking, which tends to enhance fast fission. Such an effect can be
included in both thermal and fast reactor analysis.
With these adjusted resonance integrals and fast cross sections, one can now
continue on to calculate the fast neutron energy spectrum and generate the fast
few-group constants. Hence as in the case of thermal neutron physics, one finds
that the effects due to lattice heterogeneities enter the treatment of fast neutron
physics via a simple modification of fast-group constants, which accounts for the
spatial flux variation in the lattice cell (i.e., spatial self-shielding).
IV. SOME CONCLUDING REMARKS ON THE MULTIGROUP

DIFFUSION METHOD
In the past four chapters we have developed the principal tool of nuclear
reactor analysis, multigroup diffusion theory. The multigroup diffusion equations
are capable of yielding the neutron flux in a nuclear reactor core to an accuracy
sufficient for most reactor design problems, provided adequate care is taken in the
determination of the multigroup constants entering these equations.
Indeed we have found that the solution of the multigroup diffusion equations is
rather straightforward. Most of our effort was expended in developing suitable
prescriptions for calculating the group constants, for these group constants had to
account for the rather complicated energy dependence of the intragroup fluxes
used in averaging basic cross section data, as well as for spatial neutron transport
effects arising in core lattices.
Hence a multigroup diffusion analysis of a reactor core really consists of two
principal tasks: (a) generation of the multigroup constants, and (b) solution of the
multigroup diffusion equations proper.
It should be noted that various input data are necessary in order to determine
the group constants. For example, one obviously requires basic cross section data
characterizing the various isotopes appearing in the reactor. This is usually provided
in the form of fast and thermal cross section libraries which give these data as
fine-group constants (just the basic cross section data averaged over each of the
fine groups to be used in the fast and thermal spectrum calculations).
Next one requires information concerning the composition and geometry of the
core. Since the core composition will vary throughout the core (e.g., in various
zones of differing fuel enrichment or moderator density), one will usually be
required to generate different group constants for each region of the core in which
the gross core composition is significantly different. On a finer scale, one must
provide the dimensions of the lattice unit cell for which the group constants are to
be generated. Finally one must input the temperatures of each of the major
components of the cell (e.g., fuel temperature).
The basic dimensions of the lattice geometry can easily be provided at the
beginning of the core analysis. However the necessary input describing the cell
composition and temperature must first be calculated, since these will depend on
considerations such as fuel depletion and isotope buildup, control insertion, and
the thermal-hydraulic behavior of the core. As we will see later, this information is
usually supplied by other computational modules in a reactor core analysis model
(or computer code).
Now that we have developed the basic model used to describe the neutronic
behavior of a nuclear reactor, it is appropriate that we turn to a discussion of how
this model is applied in nuclear reactor analysis. The application of multigroup
diffusion theory to reactor analysis will depend to some degree on the reactor type
and the detailed information desired. In the remaining chapters of this text, we will
concern ourselves with various examples of how such a model is applied to nuclear
reactor analysis, and in particular, to how such applications interact with other
facets of reactor core analysis and design.
REFERENCES
1. S. Glasstone and M. C. Edlund, The Elements of Nuclear Reactor Theory, Van Nostrand,
Princeton, N.J. (1952), Chapter 9.
2. H. D. Smyth, Atomic Energy for Military Purposes, Princeton University Press, Princeton,
N.J. (1945).
3. R. L. Hellens, The physics of PWR reactors, in New Developments in Reactor Physics
and Shielding, CONF-720901, (1972) Vol. I, p. 3.
4. M. H. Merrill, Nuclear design methods and experimental data in use at Gulf General
Atomic, Gulf-GA-Al2652 (1973).
5. L. W. Strawbridge and R. F. Barry, Nucl. Sci. Eng. 23, 58 (1965).
Atomic, Gulf-GA-Al2652 (1973) Chapter 3.
7. A. Amouyal, P. Benoist, and J. Horowitz, J. Nucl. Energy 6, 79 (1957); J. R. Lamarsh,
Introduction to Nuclear Reactor Theory, Addison-Wesley, Reading, Mass. (1966), pp.
382-389.
8. H. C. Honeck, THERMOS, a thermalization transport code for reactor lattice calcula-
tions, BNL-5826 (1961); Nucl. Sci. Eng. 8, 193 (1960).
9. R. L. Crowther, Physics of measurements of BWR reactors and comparison with theory,
in New Developments in Reactor Physics and Shielding, CONF-72090 I, Vol. I. p. 114.
10. K. M. Case, F. de Hoffmann, and G. Placzek, Introduction to the Theory of Neutron
Diffusion, Los Alamos Scientific Laboratory Report (1953).
12. G. C. Pomraning, Transport methods for the calculation of spatially dependent thermal
spectra, in Reactor Physics in the Resonance and Thermal Regions, A. J. Goodjohn and G.
C. Pomraning (Eds.), M. I. T. Press, Cambridge (1966), p. 207; D. R. Askew, Proceed-
ings of Conference on Numerical Reactor Calculations, IAEA, Vienna (1972), pp.
185-196.
13. Reactor Physics Constants, USAEC Document ANL-5800 (2nd Edition) (1963).
14. A. Amouyal, P. Benoist, and J. Horowitz, J. Nucl. Eng. 23, 58 (1965).
15. L. Dresner, Resonance Absorption in Nuclear Reactors, Pergamon, New York (1960).
(1966), 390-401.
17. J. H. Ferziger and P. F. Zweifel, The Theory of Neutron Slowing Down in Nuclear
.
Reactors, M. I. T. Press, Cambridge (1966) .
18. E. P. Wigner, et al., J. Appl. Phys. 2, 257 (1955).

19. G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand, Princeton, N.J.
(l 970), Chapter 8.
20. F. T. Adler and L. W. Nordheim, Tables for the computation of resonance integrals,
GA-377 (1958).
21. G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand, Princeton, N.J.
(1970), p. 123.
22. Reactor Physics Constants, USAEC Document ANL-5800 (2nd Edition) (1963), p. 280.
23. E. Hellstrand, P. Blomberg and S. Horner, Nucl. Sci. Eng. 8, 497 (1960); E. Hellstrand,
J. Appl. Phys., 28, 1493 (1957).
(1966), pp. 402-408.
25. R. V. Meghreblian and D. K. Holmes, Reactor Analysis, McGraw-Hill, New York
(l 960), pp. 692-698.
26. L. W. Strawbridge and R. F. Barry, Nucl. Sci. Eng. 23, 58 (1965); R. L. Hellens and H.
C. Honeck, IAEA Tech. Rep. Ser. No. 12, 27 (1962).
27. J. Chernick, Reactor Techn. 13, 368 (1971).
28. P. F. Zweifel, Reactor Physics, McGraw-Hill, New York (1973), p. 166.
PROBLEMS
10-1 Determine the ratio of neutron mfp to lattice dimension (e.g., fuel-pin diameter or
lattice pitch) in LWR, HTGR, and LMFBR cores for both fast and thermal
neutrons. Also compare the ratio of core size to migration length in these reactors.
10-2 Derive an expression for the thermal utilization of a three-region lattice cell (includ-
ing fuel, clad, and moderator) in terms of thermal disadvantage factors. Also derive
an expression for the cell-averaged group constants characterizing this cell in terms
of the appropriate disadvantage factors.
10-3 Show that the diffusion length characterizing a heterogeneous lattice cell can be
written approximately as L 2 ~Li(l-f), where f is the thermal utilization for the
cell. Does the nonleakage probability PTNL characterizing a bare, uniform core
increase or decrease when going from a homogeneous to a heterogeneous lattice?
10-4 Calculate the disadvantage factor f characterizing a two-region slab geometry
consisting of a fuel region of width 2a, surrounded by moderating regions of width b.
Use one-speed diffusion theory. (See Figure 10-17.)
i . . E I E - - - - - - - - - - Fuel cell ----------,),-.ii

I
I
I
i
'..IE--b~
I
Moderator Moderator I
i,E-----X----'►
FIGURE 10-17. A two-region slab fuel cell.

10-5 Calculate the self-shielding factor for the lattice cell of Figure 10-17. (Assume that
one-speed diffusion theory is valid.)
10-6 Calculate the effective absorption cross section characterizing the cell of Figure
10-17 using the definition given by Eq. (10-28). Compare this with the cell-averaged
absorption cross section calculated using the disadvantage factor or the self-shielding
factor.
10-7 Derive the following formulas for the radius of the equivalent unit cell in terms of
the lattice pitch p (i.e., the distance between fuel pin centerlines in the lattice):
(a) r=0.564 p, square lattice
(b) r=0.525 p, hexagonal lattice (see Figure 12-14).
10-8 Prove that the Green's function characterizing a lattice cell has the property
G(r,r')= G(r',r) when described by one-speed diffusion theory.
10-9 Repeat Problem 10-8 for the case when G (r, r') is described by one-speed, isotropic
scattering transport theory.
10-10 Using the symmetry of the Green's function G(r,r') for the cell, prove the reciprocity
relation:
10-11 Explain the physical difference between the absorption probability PPM• the escape
probability Pp, and the first-flight escape probability Ppo•
10-12 Calculate the absorption probability PPM for the simple slab cell of Figure 10-17.
Use one-speed diffusion theory.
10-13 Calculate the escape probability Pp for the slab cell of Figure 10-17 using one-speed
diffusion theory.
10-14 Calculate the first-flight collision probability Ppo for the cell of Figure 10-17.
10-15 Demonstrate that the average chord length characterizing a non-reentrant geometry
volume is given by
<R)=4V/S.
(The reader might find it useful to refer to Case, de Hoffmann, and Placzek 10 or
Meghreblian and Holmes 25 or Bell and Glasstone. 21 )
10-16 Demonstrate that the first-flight escape probability for a non-reentrant volume can
be written generally as
where Rmax and Rmin are the maximum and minimum chord lengths of the volume
and cp(R) is the chord length distribution function such that cp(R)dR is the
probability that a chord length lies between R and R + dR. 10• 21 • 25
10-17 Using the result of Problem 10-16 demonstrate that the first-flight escape probability
characterizing a large lump is given by Pp0 = S/4V~f.
10-18 Derive an expression for the cell-averaged macroscopic absorption cross section
<~a)cen characterizing a three-region fuel cell in terms of thermal disadvantage
factors. Describe qualitatively how you might calculate these disadvantage factors. 5
10-19 Find the escape probability for a slab of half-thickness a and scattering and
absorption cross sections ~. and ~a using one-speed diffusion theory.
10-20 Verify the ABH expression for [I/f- I] as given by Eq. (10-57) by solving the
diffusion problem [Eq. (10-53)] to obtain PMP·
10-21 Using the ABH method, compute the thermal utilization for one of the PWR core
designs given in Appendix H assuming 2.8% enrichment fuel and a water density of
Ptt,o=0.71 g/cm3• Also determine the thermal disadvantage factor for this lattice.
(Use the thermal cross section data in Appendix A.)
10-22 Derive the integral form of the transport equation (10-60). [Hint: Follow the analysis
of Section 4-IV-B.]
10-23 Prove the identity Eq. (10-65) relating the transfer coefficients Tmn and the first-
flight escape probabilities P m--.n·
10-24 Prove that the first-flight escape probabilities satisfy the reciprocity relation:
10-24 Derive the expression (10-97) for the Doppler-broadened NR resonance integral.
10-25 Derive the expression (10-101) for the NRIM form of the resonance integral.
10-26 An improved rational approximation28 is given by
2
10+ 5(R)pLi + 5( (R)pL;)
Ppo= - - - - - - - - - - -2 - - - - -3.
10+ 14(R)pLi + 5( (R)pL;) + 5( (R)pL;)
Graphically compare this to the Wigner rational approximation for (R)pLi ranging
from 0.1 to 10.0.
The following study (Problems 10-27-32) of a water-moderated reactor lattice
will require the use of a MUFT-type fast-spectrum code and a SOFOCATE-type
thermal spectrum code. The objective of these problems is to investigate the effects
of enrichment, moderator-to-fuel ratio (NM/ Np), moderator temperature, and fuel
temperature on the infinite lattice multiplication characteristics of a typical PWR
core design. Use the core data for one of the PWR designs given in Appendix H.
10-27 Determine volume fractions of H 20, U02, and Zr.
10-28 Calculate the thermal spectrum and spectrum-averaged 2-group constants for the
reactor at zero power and operating temperature. Plot and compare with a Max-
wellian distribution in pure water at 300°C.
10-29 Calculate the fast spectrum and spectrum-averaged group constants for the same
conditions.
10-30 Determine the following parameters for the uncontrolled core: kwp,p 28 ,f,j,Y/,T,L2,
and CR (conversion ratio).
10-31 Estimate the concentration of boron in the coolant necessary for k 00 = I. Recalculate
the thermal spectrum and compare with the spectrum for the zero boron concentra-
tion lattice.
10-37 Reevaluate the parameters in Problem 10-30 for the lattice containing the critical
boron concentration.
4
An Introduction to
Nuclear Reactor Core
Design
11
General Aspects of Nuclear Reactor Core
Design
The primary responsibility for the nuclear design of a reactor core rests with the
nuclear engineer. He must determine that set of system parameters which will yield
safe, reliable, and economical reactor operation at the rated power level over the
desired core lifetime. The principal tools used by the nuclear engineer in this task
consist of a number of models of neutron behavior in the reactor that are
implemented by a multiplicity of computer programs or codes used to simulate the
nuclear behavior of the reactor core. The nuclear analysis of the core cannot be
performed in an independent manner (as we have developed it thus far in this text),
but rather it must interact strongly with other aspects of core design, including
thermal-hydraulic analysis of core cooling, structural analysis of core components,
economic performance, and so on. Such "nonnuclear" design considerations will
place a number of constraints on the nuclear design of the core. For example, the
core size and geometry will usually be determined by thermal considerations, since
one must insure that core power densities are sufficiently low to prevent tempera-
tures in the core from becoming excessively high. As yet another example, the
length of time that a fuel element can be utilized in a reactor core is usually
determined not by the depletion of fissile material, but rather by the ability of the
fuel element to withstand the radiation damage and thermal and mechanical
stresses experienced in the reactor core environment.
The nuclear analysis of a given core configuration must be performed many
times during the core design. Rather crude models are used initially to identify the
major constraints that will be placed on the design. Usually in these preliminary
design studies, one draws heavily on past design experience. Such studies are used
to identify the range over which system parameters can be varied while still
conforming to the constraints placed on core performance. Throughout this analy-
447
448 / AN INTRODUCTION TO NUCLEAR REACTOR CORE DESIGN
sis, the nuclear engineer will interact strongly with other design efforts, particularly
with the thermal-hydraulic analysis.
As more information is accumulated, the design effort advances to increasingly
detailed studies in an effort to narrow in on a reference design with which one can
perform tradeoff studies in order to determine the best choice of system parameters
to achieve the optimum core performance consistent with design constraints.
Naturally during these latter stages of the design process the analytical models used
to predict core behavior become more detailed (and much more expensive to
apply).
It is convenient to separate the various aspects of nuclear reactor core design
into several general areas including nuclear design, thermal-hydraulic design,
structural and materials design, and, of course, economic and reliability (safety)
evaluations of core performance. Although our primary concern is with the nuclear
analysis and design of the core, it is important to present this topic within the more
general framework of nuclear reactor core design to illustrate the strong interplay
between the various aspects of the design procedure.
Of course, the specific details of core design will depend strongly on the reactor
type under consideration (as well as past experience of the designer). More specific
information concerning core design can be obtained from a number of sources,
including several advanced monographs. 1- 5 Although the specific details of many
aspects of core design are regarded as proprietary information by reactor manu-
facturers, many of the general features of such design methods are available in a
number of reports. An additional source of information is contained in the safety
analysis reports (PSAR's or FSAR's) required of every nuclear power plant
licensed in the United States. 6
Our discussion here is of a much more introductory nature, with the primary
intent of providing an appropriate atmosphere for illustrating how the fundamental
concepts and models developed in the earlier portions of this text are applied in
practice to nuclear reactor design. We first give a brief overview of the specific
design problems faced by the nuclear engineer, then briefly consider how these
problems are constrained by other facets of the nuclear core design, and finally,
provide an overview of reactor calculational models usually implemented in the
form of computer code packages capable of performing a variety of calculations of
interest in reactor design.
I. NUCLEAR CORE ANALYSIS
A. Design Functions of the Nuclear Engineer

Since our primary concern is with the nuclear analysis of a reactor core, we
will begin with a brief discussion of the principal design activities of the nuclear
engineer, which can be conveniently grouped into three general areas:
1. DETERMINATION OF CORE CRITICALITY AND POWER

DISTRIBUTIONS
The calculation of the core multiplication and flux or power distribution is, of
course, the most common type of analysis performed in nuclear core studies. [This
explains our preoccupation with this subject during the earlier chapters of this text.]
The nuclear engineer is concerned with the determination of the core power
GENERAL ASPECTS OF NUCLEAR REACTOR CORE DESIGN / 449
distribution since it is of central importance to both the corresponding thermal

analysis and fuel depletion studies of the core. For example, one would like to
design a core that will result in a flat radial and axial power distribution
throughout core life and also provide sufficient reactivity to yield adequate fuel
burnups while maintaining adequate reactor control. The calculation of core power
distributions will depend sensitively on parameters such as core enrichment,
moderator-to-fuel ratio, core geometry, the location and types of reactivity control,
and fuel element design. The core power density will also depend on both space
and time (because of fuel burnup and isotope production over core life).
We will later find that the parameters of greatest interest to the core thermal
designer are the ratios of the peak-to-average power densities in the core (so-called
"hot channel" or "power peaking" factors) which, along with the axial core power
profile, allow the determination of whether the thermal limitations on core per-
formance will be exceeded by a given core design. We will also find that there is a
strong feedback from the thermal core analysis. since the core temperature will
strongly affect coolant density and resonance absorption, which, in turn, will affect
reactivity.
The nuclear designer must also determine the fuel loading which will guarantee
reactor criticality over the desired core lifetime. This requires compensating for fuel
depletion as well as reactivity effects due to both temperature feedback and fission
product buildup. At this point various details of the fuel design enter, such as the
moderator-to-fuel ratio, the fuel element dimensions and configuration, and en-
richment.
2. REACTIVITY AND CONTROL ANALYSIS

An analysis must be performed to determine the amount of negative reactiv-
ity or control required to compensate for the excess reactivity contained in the
initial fuel loading as well as to allow for flexible and safe reactor operation. One
must allocate this reactivity among several different control mechanisms, including
movable control rods, soluble neutron poisons in the coolant ("chemical shim"),
and neutron poisons that burn out over core life ("burnable poisons" or
"mechanical shim"). It is important to study the interaction of such control
elements with the nuclear behavior of the core, both in static and dynamic
situations. Such calculations are necessary to perform the detailed design of
individual control elements, as well as control rod patterns and withdrawal and
insertion sequences during reactor operation (rod "programming").
One must also study the inherent reactivity changes that will occur in the core
with power and temperature changes by calculating the various reactivity feedback
coefficients that determine the short-time kinetic behavior of the core. Of particular
concern are the reactivity coefficients characterizing coolant or moderator density
and temperature changes, and the temperature coefficient of reactivity for the fuel
(primarily determined by the Doppler effect).
There will also be longer term reactivity effects such as those due to the buildup
of fission product poisons which must be studied.
3. DEPLETION ANALYSIS
During reactor operation the fuel composition will change as fissile isotopes
are consumed and fission products are produced. The nuclear designer must
monitor these processes over core life in an effort to ascertain fuel composition and
reactivity as a function of energy removal. This requires studying the depletion and
production chains for the principal isotopes (e.g., 235 U-238 U or 233 U-232Th) coupled
with the equations determining the neutron flux in the core. The calculation of the
core multiplication and power distribution must be made many times over the
operating lifetime of the core as the core composition changes. The study of the
interaction of the core power distribution with the time-dependent production or
depletion of nuclei in the core is known as depletion or burnup analysis. It is perhaps
the most time-consuming and expensive aspect of nuclear reactor analysis-and
also perhaps the most important (aside from reactor safety analysis), since it will
indicate the economic performance of the nuclear reactor. Depletion analysis is
closely related to the topic of nuclear fuel management in which one tries to
optimize the fuel loading, arrangement, and reloading in order to achieve the most
economical power generation within the design constraints (e.g., safety margins)
placed on reactor operation.
From this discussion it is apparent that the responsibilities of the nuclear
designer are quite varied and numerous. He must establish the limitations on and
determine the values of the fuel-to-moderator volume ratio, the fuel rod diameter,
and the fuel element arrangement. The fuel loading requirement must then be
determined, taking into account temperature and power reactivity defects, fuel
depletion, and fission product buildup. Control requirements must next be estab-
lished, including reactivity requirements, control rod geometry and patterns, and
the possible use of chemical shim and burnable poisons. The fuel distribution and
refueling arrangement must then be studied in order to achieve economic fuel
management within the constraints imposed by safe reactor operation. The kinetic
characteristics of the reactor must be determined for safety analyses of core
operation, including the various reactivity coefficients that arise, fission product
poisoning, and the analyses of various hypothetical accidents.
All of these design functions rely on the basic theory of nuclear chain reactions
that we have developed in the preceding chapters. However because of the
accuracy and detail of the information required for actual reactor design, rather
sophisticated applications of this theory are necessitated, which, in turn, require the
extensive use of modern digital computers.
B. Models of Core Nuclear Behavior

As we have mentioned, modern nuclear reactor design depends heavily on
various mathematical models of the nuclear fission chain reaction applied to core
analysis using digital computers. The computer programs or "codes" that represent
these mathematical simulations of the reactor core are generally quite complex and
are frequently the result of many years of extensive development and testing at the
various nuclear laboratories in this country and abroad. 8 More recently, such codes
have become the subject of various proprietary restrictions, and while their general
features are usually common knowledge, the details of the codes used in present-
day reactor design are classified as proprietary information by the reactor manu-
facturers. Hence our discussion here must be of a general nature. One can generally
group such codes into one of four different types.
1. CROSS SECTION LIBRARY PROCESSING CODES
We noted in Chapter 2 that the basic source of nuclear cross section data for
nuclear design in this country is the ENDF /B file, which contains data compiled
and evaluated from all known cross section information. The amount of such data
requires that it be stored on magnetic tape and manipulated using complex data
handling codes. 7 Such codes not only select the cross section data of interest and
prepare them in a form suitable for input to reactor design codes, but also
interpolate existing data to fill any gaps which may exist, as well as apply various
theoretical models (such as the optical model of nuclear structure) to generate cross
section data in those regimes in which no existing data exist. These library codes
also generate differential scattering cross sections, resonance integrals, and thermal
energy scattering kernels.
The differential scattering cross sections for elastic scattering are usually
generated as a sequence of terms in a Legendre polynomial expansion. By way of
contrast, inelastic scattering and ( n, 2n) processes are usually assumed to be
isotropic in the laboratory system and calculated using available data on the
appropriate nuclear states or a theoretical model of the nucleus.
One of the more difficult aspects of cross section generation concerns the
treatment of resonance cross sections. Because such resonances are usually quite
large in magnitude and yet quite narrow compared to even the energy intervals
characterizing the cross section data, it is necessary to make some attempt to
account for flux depression in the resonances in order to include effective res-
onance integrals in the cross section data set. In the regime in which the
resonances are isolated and measured, this can be accomplished using the standard
techniques (e.g., the NR or NRIM approximations). Various nuclear models can be
used to generate resonance parameters for the energy range in which the res-
onances are still isolated but not measured. The most difficult area to treat is the
energy range in which the resonances are not on.ly unresolved but overlap so
appreciably that they cannot be considered independently. It is also usually
necessary to account for heterogeneous effects (see Chapter 10), temperature
effects on the resonance structure, and overlap of resonances of different materials.
2. MULTIGROUP CONSTANT (MGC) GENERATION CODES

As we have seen, the energy range spanned by the neutrons as they are born
in fission and slow down to eventual capture or leakage at thermal energies (at
least in thermal reactors) is enormous, covering some eight orders of magnitude.
Since the cross sections themselves depend sensitively on energy, it is apparent that
one must proceed rather carefully in generating few-group constants for use in
multigroup diffusion calculations.
One usually proceeds in two steps. The energy range of interest is first divided up
into a very fine multigroup structure, and the cross section data supplied by a
library code are simply averaged over these groups (for example, in the slowing
down range one might use a l / E spectrum). Appropriate approximations to the
effective resonance integrals of interest are also included in this set of "fine-group
constants." These fine-group constants then serve as the microscopic cross section
data used in fast and thermal spectrum codes that perform an approximate
calculation of the neutron energy-dependence for the nuclear assembly of interest,
and then average or collapse the fine-group constants into few-group constants
over these approximate spectra. It should be noted that whereas fine-group con-
stants are usually evaluated without reference to the detailed system under con-
sideration, the spectrum generation codes generate few group MGC for the specific
system of interest. These MGC are then used in static and kinetic design analysis.
Of course, the trick enabling one to perform a detailed study of the neutron
energy spectrum is the temporary neglect of the detailed spatial dependence of the
neutron flux. It is customary to assume a simple single mode flux shape-that is, to
assume a buckling that characterizes the region of the core under consideration.
The angular flux dependence is simplified using either the P I or B I approximation
(although sometimes higher order BN methods may be used).
However the success of MGC generation codes frequently depends on how these
codes are corrected for the rather strong spatial dependence of the flux that occurs
in the vicinity of the fuel rod or control rod. We have already examined several
methods for correcting cross section group constants for these heterogeneous
effects, such as disadvantage factors and collision probabilities. In the fast region,
it is usually sufficient to merely correct the calculation of the resonance integrals
for self-shielding effects using escape probabilities calculated for the cell geometry
of interest. In the thermal range, one may use either approximate techniques such
as the ABH method to determine disadvantage factors, or more elaborate cell-
calculation techniques such as THERMOS9 to directly calculate the self-shielded
cross sections.
We have seen that these MGC are usually then spatially averaged over the fuel
cell or perhaps a fuel assembly in order to generate "homogenized" cross sections
most consistent with the spatial mesh to be used in the few-group multigroup
diffusion calculation.
We have already referred to the various common schemes to generate few-group
constants for LWRs-namely, MUFT-GAM- 10 type calculations for the fast
spectrum, and SOFOCATE 11 or TEMPEST- 12type calculations for the thermal
spectrum, including perhaps a THERMOS calculation for cell analysis. The cor-
responding types of code for HTGRs include GAM 13 or MICROX 14 for the
fast-group constants and GATHER5 for thermal-group constants. The group
structure necessary for fast reactor analysis must be considerably more detailed,
and ultrafine group structures such as those employed in MC2 are used to generate
MGC for fast assembly analysis.
3. STATIC DESIGN CODES

Static design codes are used to obtain the global spatial dependence of the
neutron flux throughout the reactor core. This information is required for accurate
predictions of the fuel loading, power distributions, temperature dependence of
reactivity, excess reactivity, shutdown margins, shielding requirements, and other
quantities. These codes are usually few-group diffusion or transport codes that
utilize the MGC generated by the spectrum codes discussed in the previous section.
One customarily uses such codes to determine the multiplication factor (eigen-
value) and flux distribution in the system of interest. Usually multidimensional
diffusion codes are sufficient for this task. However occasionally transport codes
are necessary to determine cell-correction factors in the vicinity of strong absorbers
or voids.
Indeed before a diffusion theory calculation can be performed the heterogen-
eities in the core must be homogenized. Although MGC generation codes aug-
mented with prescriptions such as those provided by the ABH method usually
provide sufficient accuracy for this purpose, it is occasionally necessary to use a
transport code such as ANISN 15 or TWOTRAN 16 • With such corrections, mul-
tigroup diffusion codes are usually adequate for a wide class of problems, including
the determination of the over-all flux and power distribution, the effects of fuel
zoning, reactivity predictions, and so on.
Such codes can also be used to calculate temperature and power coefficients of
reactivity. This is usually performed by merely calculating the core multiplication
for several different core temperatures or power levels (using different MGC for
each temperature, of course).
4. TIME-DEPENDENT DESIGN CODES
The fourth class of codes attempt to explicitly treat the time dependence of
the nuclear behavior of the core. Of course by including this additional variable,
one is usually forced to utilize a coarser description of the remaining variables-for
example, space and energy. Such codes can themselves be broken into three
separate classifications:
(a.) DEPLETION CODES
In a depletion calculation, one is concerned with monitoring the core compo-

sition and flux distribution in the reactor over the core life. In particular one must
take into account the time- and space-dependence of fissile material depletion, the
transmutation of fertile to fissile material, fission product buildup, and the ad-
justment of control elements to maintain core criticality. Depletion codes involve
the solution of the rate equations describing isotope concentrations, as well as the
neutron balance equations describing core criticality. Because of the many
variables involved, such as power density, reload frequency, excess reactivity
requirements, and varying reload patterns, even the simplest depletion problem
may require solutions at many points in time.
Because of their low computer cost and general flexibility, extensive use is made
of zero-dimensional depletion codes for survey calculations. However, more de-
tailed analyses will require the use of one- or two-dimensional depletion codes in
order to account for the spatial flux variation and control-rod and reload patterns.
(b.) FUEL CYCLE ANALYSIS
One of the major applications of fuel depletion codes is to the analysis of fuel
cycle strategy-that is, a determination of the optimum fuel reloading strategy for
minimizing the fuel cycle costs. Such estimates of initial core composition and
reload fuel composition must take into account the requirements for heat removal,
coolant pumping power, peak fuel temperatures, and so on, and hence are in-
variably a compromise between that core composition entailing the lowest fuel
cycle cost and that composition resulting in acceptable system temperatures.
(c.) REACTOR KINETICS ANALYSIS
In order to analyze the transient response of the reactor to both normal

operating conditions and postulated accident situations, one must utilize codes
based on the nuclear reactor kinetics equations discussed in Chapter 6. Such
calculations differ dramatically from depletion calculations in two respects: (a) the
time scale of reactor transient behavior is usually of the order of minutes or less,
whereas depletion calculations are concerned with core behavior over times as long
as several years and (b) in kinetics calculations, the reactor is usually subject to a
nonzero net reactivity (due to control, operating conditions, etc.). In a depletion
calculation, one always assumes that control is adjusted to hold the core critical at
all times. Hence, in a depletion calculation, one is actually only concerned with the
steady-state equations describing the neutronics.
Reactor kinetics analysis plays an important role in core design, since it is

necessary to assess safety margins, to select and place control and safety instru-
mentation, to design the control and protective systems, and to determine the need
and effectiveness of engineered safeguards.
Of course at the heart of such codes is a numerical solution of the reactor
kinetics equations describing the neutron flux time behavior in the core. However
one must also account for coolant flow and heat transfer in order to determine
feedback reactivity. In addition, there is strong coupling to other dynamic pro-
cesses in the NSSS, such as the primary coolant loop, steam generator, coolant
pumps, and feedwater system. Hence the study of reactor dynamics can become
quite complicated.
As input to such codes one requires not only basic nuclear data, such as cross
sections, temperature coefficients, and so on, but thermal-hydraulic information as
well. These latter quantities are frequently determined using static design codes.
The simplest level of description would be the use of point-reactor kinetics to
describe the core neutronics. However, in many instances the spatial dependence of
the flux is important, and then either time-dependent multigroup diffusion codes or
hybrid schemes such as synthesis must be utilized for an adequate description.
II. OTHER AREAS OF REACTOR CORE ANALYSIS

We have outlined the various types of calculation required in the nuclear
analysis and design of a reactor core. These include the calculation of core
criticality and power distributions, the determination of reactivity coefficients, fuel
loading requirements and core arrangement, reactivity control calculations, fuel
depletion studies, and reactor safety analysis. There are numerous constraints
imposed on such nuclear design. Of course the core composition and configuration
must be chosen such that sufficient excess reactivity is available for power genera-
tion over a reasonable time period (usually on the time scale of years). The
reactivity control must be capable of ensuring the safe and reliable operation of the
reactor over core life. And these requirements must be met while at the same time
minimizing the economic cost of the power generation. These constraints are
frequently in conflict with one another. For example/there is incentive to operate
the reactor at the largest possible power density consistent with maintaining fuel
and coolant temperatures below limits set by safety considerations. Yet such a high
power density frequently does not lead to the optimum economic power generation
over the entire core life, since the flux and hence power distribution will shift as the
fuel is depleted.
The nuclear analysis and design of the core cannot be decoupled from other
considerations such as the thermal behavior of the core or the behavior of the
various materials that comprise the core. This can perhaps be understood more
clearly by briefly describing the other types of analysis that arise in reactor core
design.
A. Thermal Core Analysis

The energy released in nuclear fission appears as kinetic energy of fission
reaction products and eventually as heat generated in the reactor fuel elements.
This heat must be removed from the reactor core and used to generate electrical
power. The study of the sequence of processes involved in the transport and
utilization of fission heat energy is most properly the concern of the mechanical
engineer. However since it has such a significant bearing on the nuclear design of
the core, we will devote a considerable portion of Chapter 12 to a summary of the
thermal analysis of nuclear reactor cores.
The primary objectives of thermal core design include achieving a high power
density (to minimize core size), a high specific power (to minimize fuel inventory),
and high coolant exit temperatures (to maximize thermodynamic efficiency). How-
ever these objectives are subject to several important constraints. For example, one
must always ensure that the core temperatures remain below the melting points of
core components (particularly for the fuel and the clad). There are also frequently
limits on heat transfer rate between the fuel element and the coolant, since if this
heat transfer rate becomes too large, film boiling of liquid coolants may occur
which will result in a rapid rise in clad temperatures. Hydraulic considerations also
enter. For example, the coolant pressure drop across the core must be kept low to
minimize pumping requirements as well as hydraulic stresses on core components.
Such thermal-hydraulic constraints must be studied over core life, since as the
power distribution in the core changes due to fuel burnup or core reloading, the
temperature distribution will similarly change. Furthermore since the cross sections
governing the neutronics of the core are strongly temperature- and den-
sity-dependent, there will be a strong coupling between the thermal-hydraulic and
neutronic behavior of the reactor core. We will see in Chapter 12 that this coupling
has a strong influence on the manner in which reactor criticality calculations must
be performed.
B. Mechanical Analysis of Reactor Cores

It is of vital importance to choose materials and design core components
capable of withstanding the environment of intense radiation, high pressure, and
high temperature of a nuclear reactor core. Of central concern is the structural
design of fuel elements subject to rather severe mechanical stresses. For example,
as the fuel produces energy via fission reactions, it will also produce fission product
gases (roughly 27 atoms of such gases will be produced per 100 atoms fissioned). 17
These gases will diffuse through the fuel pin and into the fuel-clad gap, where they
will accumulate, producing rather sizable pressures on the inner clad surface.
Typically an attempt is made to keep these pressures below the design coolant
pressures. However the internal fuel element gas pressure must not become too
low, or else the clad may fail due to the outside coolant pressure (for example,
should a gap appear between two fuel pellets). The fission events occurring in the
fuel will also cause considerable swelling of the fuel pellets, which places additional
internal stresses on the clad. There are other stresses on the fuel element, such as
those from the external coolant pressure. The enormous thermal gradients across
the fuel pin and clad, as well as the temperature changes (thermal cycling)
accompanying changes in reactor power level, render the mechanical design 18 of
the fuel element even more difficult, as do a number of other considerations
involving radiation damage. These factors are discussed below.
The mechanical design of other core components must also be considered in core
design. For example, the various internal structure required to support the fuel,
flow baffles, and control rod assemblies must be designed to withstand the intense
core environment. And of course the reactor pressure vessel itself represents a
formidable mechanical design problem, since it must withstand extremely high
pressures and radiation intensity over the operating lifetime of the reactor (-30
years). One must keep in mind throughout the fabrication costs and maintenance
of the mechanical assemblies in the core (with particular attention to movement of
control rod assemblies and refueling operations).
C. Materials Problems in Reactor Core Design

As we mentioned above, the environment of a nuclear reactor core is
characterized by very high pressures, large thermal gradients, and intense nuclear
radiation, which place very stringent demands on the materials comprising core
components. The behavior of materials subjected to large mechanical and thermal
stresses is not a new area of investigation (except possibly to the extremes
encountered in modern reactor design), and a great deal is known about the
fabrication of high-strength alloys and ceramics which can withstand these stresses.
However the nuclear reactor core subjects materials to an additional demand, that
posed by the intensive bombardment by nuclear radiation to which materials in the
core are exposed. Over a period of time, such radiation can dramatically alter the
properties of these materials. 19 For example metals will become brittle and swell
and corrosion is enhanced. The reactor engineer must be very careful to anticipate
such radiation damage in his design, as well as to adopt a sufficiently conservative
design to compensate for the lack of experience available with the behavior of
materials under very high irradiation fluences.
These effects of radiation on reactor materials have become particularly impor-
tant in today's maturing nuclear reactor industry. In order to achieve the lowest
possible power costs, nuclear fuel elements must be used in a reactor as long as
possible (i.e., high burnup). Actually the principal limitation on the amount of
burnup is not the loss of 235 U or 239 Pu nuclei through fission, but rather the
attendant radiation damage to the fuel and the cladding material, which would
lead to fuel element failure if the fuel is left in the core too long. 20 Hence whereas
nuclear reactors were limited by nuclear considerations during the 1950s, and by
thermal design during the 1960s, today's modern yower reactors are primarily
limited by the radiation damage which can be withstood by reactor materials.
This situation becomes even more difficult in fast-breeder reactors because of the
intensity of fast neutron radiation. It has been found that stainless steel (a major
structural component of the fast breeder) swells rather dramatically (several per-
cent) when irradiated over long periods of time by fast neutrons.2 1 Such swelling
must be accounted for in the mechanical design of the fast reactor core (with a
considerable associated economic penalty).
We have seen that there are several types of high-energy radiation present in a
nuclear reactor core. Most of these result from the nuclear fission reaction itself,
although lesser amounts arise from associated reactions such as radiative neutron
capture. Of course, most fission energy is carried by the massive fission-fragment
nuclei. These cause catastrophic damage to the adjacent fuel material, but because
of their large electrical charge, the range of the fission fragments is extremely short
( < 20µm); and hence this damage is localized in the fuel within the immediate
vicinity of the fission event. A potentially more serious type of radiation is that due
to fast neutrons. Because of their neutrality, neutrons have rather long ranges (as
~
much as several meters). Hence they can damage material located anywhere in the
reactor core. Gamma radiation is also characterized by large ranges, but is of
secondary importance to fast neutron damage.
The actual effect of radiation on a material depends sensitively on the type of
material, the type of radiation, and the conditions during the time of irradiation
(such as temperature). However some general observations can be made. For
example, for irradiated metals, hardness, tensile strength, and impact resistance
increase, while ductility decreases (corresponding to an increase in brittleness).
The significance of radiation damage in reactor core design becomes particularly
apparent when one examines nuclear fuel performance in the LWR. As we will see,
the low cost of nuclear fuel (relative to fossil fuel) is the principal factor that leads
to the economic advantages of nuclear power. That such costs are realizable is due
in no small measure to the significant advances made in nuclear fuel design and
performance since the mid-1960s. Nuclear fuel elements must be designed subject
to several criteria22 intended to guarantee the fuel performance up to the lifetime
limit: (a) the fuel temperature at the hottest point always must be below the
melting point, (b) any fuel displacement that might influence core temperature
distribution and multiplication must be minimized, (c) the cladding must remain
leak-proof, and (d) the outer geometry of the fuel pin (length, diameter, straight-
ness) is subject to very small tolerances.
As we have seen, nuclear fuel elements are subjected to intensive irradiation.
They are furthermore subject to extremely large temperature variations. This is
caused to a large degree by the rather poor ability of the principal types of nuclear
fuel, uranium oxide or carbide, to conduct heat. Such temperature variations place
enormous thermal stresses on the fuel elements and interact strongly with the
changes in the fuel induced by irradiation.
The principal radiation effects that must be accounted for in fuel element design
include: 22 (a) fuel creep and swelling, fission gas release, pore migration, chemical
changes, and change in radiation and axial fuel density profile, (b) cladding
mechanical properties, swelling by void formation, and corrosion, (c) fuel pin radial
heat transfer and temperature distribution, mechanical and chemical interaction
between fuel and cladding, swelling and bowing of the pin, and (d) fuel bundle
changes in component geometry and interaction of fuel pins and spacers.
Such considerations have led to a number of modifications in fuel element
designs for modern power reactors. Current designs provide for increased void
volume to accommodate fuel swelling and fission gas release associated with the
higher burnups used in present-day power reactors. The excellent corrosion resis-
tance of zirconium and stainless steel alloys has led to their almost exclusive use as
a cladding material. As the nuclear power industry obtains more operating ex-
perience with fuel design and behavior under long-term irradiation, it will be more
able to develop advanced designs capable of very high burnups and power
densities.
The commercial success of the LMFBR will also be critically dependent on the
attainment of low fuel cycle costs and therefore on the ability to run fuel elements
to very high burnups. Only rather recently has any appreciable experience in the
behavior of materials in high fast neutron flux environments been available.
Perhaps the most dramatic effect thus far observed occurs in structural materials
such as stainless steel. After long periods of irradiation by fast neutrons, the steel is
observed to swell. Closer examination indicates the presence of small voids in the
irradiated material. Of course such swelling and void formation are highly unde-
sirable in a reactor core in which mechanical and structural tolerances must be
kept very refined over the lifetime of the core (up to 30 years).
The voids are caused by fast neutrons that rip through the crystal lattice,
knocking atoms out of their lattice positions. These vacancies tend to migrate
together to form voids and hence induce the swelling. By raising the steel to high
temperatures the voids can be annealed out. Unfortunately at the anticipated
operating temperatures of the LMFBR such swelling can be quite pronounced and
must be accounted for in core design.
D. Economic Analysis23- 25
The justification for nuclear power plants must reside in their economic
advantages over more conventional sources of electrical power. The cost of
electrical power can be broken down into a number of factors, including the capital
cost of constructing the plant, the annual cost of operating and maintaining the
plant, and the annual costs for fuel. The capital investment required for the
construction of nuclear power plants is usually greater than that required for
conventional power plants. Furthermore, operating and maintenance costs account
for only a small fraction of the total cost of producing electricity. Hence the
primary advantage enjoyed by nuclear power is in the lower cost of its fuel.
It is important to recognize that nuclear fuels are totally different from fos-
sil fuels, both in their processing and utilization, as well as in their costs. There
are a large number of sophisticated and expensive processing operations required
by the fuel before it is inserted into the reactor core. It is then "burned" in the
reactor for several years before being removed. Even after several years of use in a
reactor, the fuel possesses a sizable concentration of fissile material. Hence it must
be removed from the core, reprocessed, and refabricated into new fuel elements.
The byproduct waste from the reprocessed fuel is highly radioactive and must be
disposed of with considerable care.
Those operations involved in the extraction, preparation, utilization, reprocessing
and disposing of nuclear fuels are referred to as the nuclearjuel cycle. Such a cycle
extends over a period of several years, and th,e costs associated with the nuclear
1
fuel cycle must be monitored over this period of time. In this sense, nuclear fuel
costs are much different than fossil-fuel costs, since a number of charges other than
direct materials costs are involved which may either lead or lag utilization of the
fuel material by several years. The primary costs associated with the nuclear fuel
cycle include the following:
(1) Costs of net isotope consumption that result from the conversion of
uranium and plutonium into fission products and from the reduction in
the 235 U-enrichment of the uranium remaining in the spent fuel. These
costs are associated with the exploration, mining, and enrichment of the
uranium.
(2) Processing costs, such as those incurred in uranium purification, conver-
sion, fuel fabrication, and reprocessing.
(3) Financing costs associated with the large working-capital requirements of
the nuclear fuel cycle.
The management of the various activities involved in obtaining, irradiating, and
disposing of fuel materials is referred to as nuclear Juel management and is a
principal conce&. of nuclear engineering. 26 Such activities must be performed
GENERAL ASPECTS OF NUCLEAR REACTOR CORE DESIGN / 4S9
subject to several very important constraints. Of course one desires to minimize

electrical generation costs, but must also ensure that the safety of the reactor is not
compromised. For instance, the fuel temperature must always be kept below
melting points and the control margin must be maintained within safe limits.
Needless to say, the complexity of accounting for nuclear fuel costs and capital
plant investment requires a rather sophisticated economics analysis of the plant
design. Nuclear power costs will depend on parameters that vary widely, depending
on the location of the plant, the type of the reactor, and even the time at which the
economic study is performed. Such considerations make it apparent that each
reactor manufacturer and utility have access to a technical group capable of
predicting the behavior of the nuclear fuel, performing an economic analysis of
such fuel utilization and analyzing the total power system requirements involving
the plant (including other conventional and nuclear plants in the system).
E. Safety and Regulatory Considerations27- 29

Of course, all reactor designs are subjected to extremely thorough studies to
ensure that they are compatible with existing safety and regulatory standards. For
example, the response of the reactor design to reactivity insertions resulting from
severe disturbances that could arise only under the most extreme circumstances
must be determined.
The principal safety concern inherent in nuclear reactor operation does not
involve the possibility of a nuclear explosion. Such an event is quite impossible in
thermal reactors, and it requires an agile mind to concoct a sufficiently fantastic
scenario to initiate such an explosion even in a fast reactor. Rather, the problem is
the large inventory of radioactive fission products that accumulate in the reactor
fuel. As long as these fission products remain in the fuel, they represent no hazard
to either plant personnel or nearby population. Yet should they be released and
transported to populated areas, substantial dangers could arise.
Hence nuclear reactors must be designed such that under no credible-or even
incredible-operating situation could such radioactive material be released from
the core. To achieve this guarantee, not only must the reactor core and coolant
system be carefully designed against every conceivable accident situation, but
auxiliary systems must be incorporated into the core as well-so-called engineered
safeguards-to ensure that fission products are contained in any foreseeable acci-
dent.
The subject of nuclear reactor safety is exceedingly complex, enmeshed in a
labyrinth of complex technical, regulatory, political, philosophical, and even emo-
tional issues. We will avoid a detailed discussion of these topics in this section,
choosing instead to merely illustrate several of the considerations that arise in
reactor safety.
As we have mentioned, the primary concern in nuclear reactor safety analysis is
that the large fission product inventory produced in the reactor core is kept intact
and not released in any conceivable accident situation. There are several barriers to
such fission product release. The primary barrier is the metal clad of the fuel itself,
which isolates the fuel pellets from the coolant. Should this clad be ruptured, the
pressure vessel containing the reactor core and the coolant will then assume the
role of a secondary barrier to fission product release. In the catastrophic event of
failure of the primary coolant piping or pressure vessel, the containment structure
itself provides yet a third barrier to fission product release.
It is the task of the nuclear reactor designer to dream up possible accident
situations in which one or more of these containment barriers might be breached,

and then to design the nuclear reactor system so that if such situations did arise,
there would still be no danger to the public. There are essentially three lines of
defense against such accident.
First, of course, is the careful attention to design, component fabrication,
construction, and plant operation. Such monitoring is referred to as quality
assurance or safety assurance and is of central concern in nuclear power generation.
Secondly there are separate safety systems designed to take protective action in the
case of abnormal reactor behavior. Examples would include the scram control
systems designed to shut the reactor down if abnormal operating conditions are
encountered. Finally, nuclear plants are equipped with various engineered
safeguards systems to protect against the consequences of highly unlikely but
possible catastrophic accidents.
All nuclear reactors possess such engineered safeguards systems to prevent
fission product release in the event of an accident. It is extremely important to
thoroughly analyze the behavior of the reactor system in the event of an accident
situation, not only to determine the appropriate engineered safeguards systems, but
also to study the performance and interaction of these systems under postulated
accident conditions. However since the particular type of safeguards required
depend sensitively on the reactor type under consideration, we will avoid any
detailed discussion of nuclear reactor safety systems here.
III. REACTOR CALCULATION MODELS
As we have repeatedly stressed throughout this text, the complexities of

modern nuclear reactor design require that the various theories of nuclear reactor
behavior we have been studying be implemented in a form suitable for digital
computer calculations. A variety of such computer programs or codes have been
developed over the years to assist the engineer in all facets of nuclear reactor
analysis.
Frequently, a number of such codes are grouped together into code packages
which can be applied to a variety of design tasks. Since our primary concern is with
the nuclear analysis of a reactor core, we will confine our attention to a general
discussion of those code packages which usually comprise the calculational model
for describing the nuclear behavior of the core. 30--32
However, we should bear in mind that the nuclear behavior of the core
frequently serves as the input to other aspects of core design. For example, the
determination of the neutron flux in the core implies the core power distribution
which is necessary for thermal-hydraulic analysis, fuel depletion studies, and the
optimization of plant economic parameters:
/ Thermal-hydraulics
Neutron flux - - - - Power density ~ Fuel depletion
Economics
There will, of course, be considerable feedback from these latter design functions
back to the nuclear analysis. For example, the thermal-hydraulic calculations will
'
determine both core temperatures and coolant densities essential to the generation
of macroscopic group constants. The depletion calculation will determine the
isotopic composition of the fuel throughout core life.
Although the details of such nuclear code packages or computer models will vary
from one design group to another (or for one application or another), the general
features are usually very similar. Of course all aspects of the neutronic analysis of
the reactor can be traced back to the neutron transport equation, but as we have
repeatedly emphasized, the direct solution of this equation is usually quite intract-
ible. Hence numerous approximations are usually required in order to develop the
mathematical models that serve as the basis of reactor design codes. Typically,
these codes suppress certain independent variables in order to allow a detailed
analysis of the process of interest. For example, multigroup constant generation
codes usually suppress spatial dependence-either by assuming an infinite medium
or an effective buckling mode-in order to facilitate a detailed treatment of the
neutron energy. In a similar sense, static design codes employ a rather coarse
multigroup structure in order to allow a detailed study of the spatial dependence of
the neutron flux. In time-dependent codes one frequently ignores both spatial and
energy dependence.
The proper utilization of such nuclear reactor codes requires not only a thorough
knowledge of the various approximations that have entered into the development
of the code, but a good deal of common sense, experience, and just plain
old-fashioned good luck as well.
In Figure 11-1 we have illustrated the principal components (code modules) of a
computational model (code package) suitable for the nuclear analysis of a reactor
core. For purposes of reference we have indicated on this diagram the chapter of
this text in which the underlying theory of each module has been developed.
However, for completeness, we will review the function of each module here.
The basic structure of the model consists of a module to generate macroscopic
group constants, followed by a module to utilize these group constants in a
determination of the core multiplication, flux, and power distribution. The power
distribution can then be used as the input to a thermal-hydraulics module which
solves the equations of heat transfer and fluid flow to determine the temperature
and coolant density distribution in the core. Of course, this latter information is
required for the generation of macroscopic group constants; hence feedback to the
earlier MGC module (and possible iteration) will be necessary.
At this point one has presumably calculated the core multiplication keff and
power distribution, but in general the core will not be critical (keff=I= 1). Hence a
control adjustment module is necessary to calculate the degree of control element
withdrawal or insertion to return the core to a critical state (again via an iterative
process).
When a critical core configuration has been achieved, one passes to a depletion
module that uses the resulting flux distribution as input for a solution of the rate
equations describing fuel isotope depletion, fertile isotope transmutation, and
fission product buildup for a given period of reactor operation. After this depletion
step, the new isotopic composition of the fuel is returned to the group-constant
module, and the entire calculation is repeated (since the core is now no longer
critical).
After a series of depletion steps, the fissile inventory drops sufficiently low that
the control system can no longer return the core to criticality. At this point, the
Chapter in which types

of analysis implemented
in module are discussed:
Macroscopic group-constant generation module
Parameterized Effective
Fast and thermal
spectrum codes group-constant control cross
tables sections
Flux-power-reactivity module
,(static multigroup diffusion codes)
k011 , ,P(r) , P(r)
@ Thermal----,hydraulic module
Control-adjustment module
Depletion module
Economics module
Optimum mills/kWhr power costs
FIGURE 11-1. Computational model of a nuclear reactor.
useful operating life of the core is deemed to have ended, and one passes on to an
economics module that utilizes the total energy produced over the core lifetime and
the amount of fissile material consumption (and production) to compute the fuel
costs of the operating cycle.
Various different types of input data are needed for this sequence of calculations,
including the geometry of the reactor core, lattice, and control system, the initial
core loading (composition, etc.), inlet coolant conditions, as well as fundamental
cross section data.
A somewhat more detailed description of each of the code modules follows.
A. Macroscopic Cross Section Module

One of the most important aspects of calculations based on the multigroup
treatment of the neutron energy dependence (as is essentially all nuclear design) is
the generation of few-group constants. As we have seen, such group constants are
formed by averaging the microscopic cross section information obtained from
evaluated data sets such as ENDF /B over not only a neutron energy spectrum
hopefully characterizing that in the reactor core of interest, but averaging these
cross sections as well over the detailed spatial dependence of the flux in a unit fuel
cell in order to account for heterogeneous effects (such as self-shielding). As input,

such modules require the number densities characterizing the materials in the fuel
cell to be analyzed. The coolant density and average moderator, structure, and fuel
temperatures are provided from a thermal-hydraulic module, while the fuel density
(as well as fission product densities) are provided by a depletion module. The
densities of other materials such as structure are provided by input specifications.
The energy-averaged, self-shielded macroscopic group constants are then deter-
mined either by performing fast and thermal spectrum calculations for the fuel cell
of the specified composition, or by computing a small number of descriptive
parameters that can then be used to obtain the macroscopic group constants from
suitable tables constructed for the reactor core of interest. (The business of group
constant parameterization is a very useful and important topic in depletion calcula-
tions because of the large number of times such group constants must be generated
over core life. We will return to consider it in greater detail later in Chapter 13.)
Such group constants must be generated for each region of the core in which the
composition is different. One must also account for the contributions of fission
products such as xenon and samarium and for control absorption which might be
present in the cell of interest. As we will see in Chapter 14, the effect of control
elements or burnable or soluble poisons in the cell are usually included by
generating effective absorption cross sections characterizing the control absorber
and then adding these to the absorption group constant characterizing the cell.
These effective control cross sections can also frequently be parameterized at
considerable computational savings.
B. Flux-Power-Reactivity Module
At the heart of the neutronics code package is the module that solves the
multigroup diffusion equations to determine the flux and power distribution in the
core as well as the core multiplication. We have already examined this particular
module in some detail in Chapters 5 and 7. In particular, we have studied two
different approaches to this calculation: (a) that based on the conventional finite-
difference solution of· the multigroup diffusion equations and (b) that based on
so-called nodal methods which decompose the core into "node cells" and then
calculate only the average power or flux for each of these cells. The latter scheme is
particularly useful when information is required about the three-dimensional flux
distribution in the core, since a direct finite difference solution of the multigroup
diffusion equations is extremely expensive.
C. Thermal-hydraulic Module
The flux-power module provides the power distribution required by the
thermal-hydraulic module which then calculates core temperatures and heat fluxes
in order to make certain that the thermal limitations placed on core performance
are not exceeded. The core temperatures and coolant density profiles are returned
to the macroscopic cross section module. As we will see in more detail in Chapter
13, some iteration between the thermal-hydraulic and nuclear core analysis is
usually necessary.
Since the thermal-hydraulic analysis of a reactor core plays such an important
role in nuclear reactor design, we will describe in some detail the various models
and concepts used in this module in Chapter 12.
D. Control-Adjustment Module
The control-adjustment module takes the core multiplication calculated by
the flux-power module and estimates the amount of control insertion or withdrawal
necessary to return the reactor to a critical state. In determining how to adjust the
control reactivity (e.g., movable rods or chemical shim), the control module may be
guided by a prespecified control management scheme. The control appears in the
nuclear analysis as effective control cross section group constants that are added to
the absorption group constants characterizing those core regions in which a control
element is present. We will develop the necessary theory for calculating these
effective cross sections in Chapter 14 (although we might remark that the basic idea
is very similar to that involved in calculating fuel-cell averaged group constants to
permit the analysis of a heterogeneous reactor lattice). The control module then
returns the effective control group constants for the new control pattern to the
macroscopic cross section module for the next criticality calculation.
E. Depletion Module
The depletion module solves the rate equations describing the isotopic
changes in core composition during reactor operation. Such equations must
account for fuel burnup, the conversion of fertile to fissile material, and the
buildup of nonsaturating fission products. The input to the depletion module is the
flux calculated by the flux-power-reactivity module, while the number densities
calculated by the depletion module are returned as input to the macroscopic cross
section module. We will discuss such depletion calculations in some detail in
Chapter 15.
F. Economics Module
Such code packages will frequently also contain a module that uses informa-
tion from the depletion analysis to determine the cost of the power generation (i.e.,
fuel costs) for the period. Such modules must account for the various costs
associated with the nuclear fuel cycle for the core under study. We will not discuss
such fuel economics calculations here, but will instead refer the interested reader to
a number of sources of additional information on this topic. 23- 25
G. Concluding Remarks
Of course, this rather brief description of the general features of computer
code models used to analyze nuclear reactor core behavior does not really do
justice to the complex interactions existing among the various modules or with
other aspects of the core design. Other nonnuclear aspects of the design such as
mechanical design or materials limitations on fuel element performance usually
enter the nuclear analysis model as constraints placed on core performance. For
example, the extent to which a fuel element can suffer radiation damage without
experiencing an appreciable probability of failure will usually be represented by a
limit on fuel burnup. Clad strength requirements may appear in the calculation as a
limitation on clad surface temperature (or clad surface heat flux).
Furthermore we have not described in any detail the enormous variety of design
problems to which such a model can be applied. Some of these will be illustrated in
later chapters, when we try to "fill in the gaps" to complete our description of the
models used in nuclear reactor analysis.
Finally we should stress once again that such mathematical or computer models
of an extremely complex system such as a nuclear reactor "work" primarily
because past experience (rather than elegant mathematical analysis) has indicated
the validity of the approximations used in developing the model (although some-
times only because of a fortunate cancellation of errors), and, equally important,
because such models have been "fine-tuned" by empirical adjustments of a number
of parameters appearing in the model. For this reason we should stress once again
that a given model (code package) will require considerable insight, ingenuity, and
a dose of good fortune on the part of the user if it is to be extended into areas of
application that depart considerably from the expetience on which the model was
constructed.
REFERENCES
l. A. Sesonske, Nuclear Power Plant Design Analysis, USAEC TID-26241 (1973).

2. H. W. Graves, Jr., Nuclear Reactor Design, University of Michigan Lecture Notes
(unpublished) (1969).
3. A. Radkowsky (Ed.) Naval Reactors Physics Handbook, Vol. I, USAEC (1964).
5. M. H. Merrill, Nuclear Design Methods and Experimental Data in use at Gulf General
6. Standard Safety Analysis Reports, General Electric, Westinghouse, Combustion Engine-
ering, Babcock and Wilcox, and Gulf General Atomics.
7. H. C. Honeck, ENDF /B: Specifications for an Evaluated Nuclear Data File for
Reactor Applications, USAEC Report BNL-50066, 1966; revised by S. Pearlstein (1967).
8. J. Chernick, Reactor Tech. 13, 368 (1971); Reactor Physics Constants, USAEC Document
ANL-5800, 2nd Edition (1963).
9. H. C. Honeck, THERMOS, BNL-5826 (1961); Nucl. Sci. Eng. 8, 193 (1960).
10. H. Bohl, Jr., E. M. Gelbard, and G. H. Ryan, MUFT-4, WAPD-TM-22 (1957).
11. H. Amster and R. Suarez, SOFOCATE, WAPD-TM-39 (1957).
12. R.H. Shudde and J. Dyer, TEMPEST, NAA Program 3W-354 (1960).
13. G.D. Joanou, E. J. Leshan, and J. S. Dudek, GAM-1, GA-1850 (1961).
14. P. Walti and P. Koch, MICROX-A Two-Region Flux Spectrum Code for the Efficient
Calculation of Group Cross Sections, Gulf-GA-Al0827 (1972).
15. W. W. Engle, Jr., A Users Manual for ANISN, A One Dimensional Discrete Ordinates
Transport Code with Anisotropic Scattering, USAEC Report K- I 693 (1967).
16. K. D. Lathrop, TWOTRAN, A Fortran Program for Two-Dimensional Transport,
GA-8747 (1968).
17. W. B. Lewis, Nucl. Appl. 2, 171 (1966).
18. N. J. Palladino, Mechanical Design of Components for Rea·ctor Systems, in The
Technology of Nuclear Reactor Safety, T. J. Thompson and J. G. Beckerley (Eds.), M.I.T.
Press, Cambridge (1973), Vol. II.
19. R. B. Holden, Ceramic Fuel Elements, Gordon and Breach, New York (1966).
20. P. G. Shewmon, in Education and Research in the Nuclear Fuel Cycle, D. M. Elliot and
L. E. Weaver (Eds.) Oklahoma University Press, Norman (1970).
21. C. M. Cox and F. J. Homan, Nucl. Appl. Tech. 9, 317 (1970); P.R. Huebotter, Reactor
Techn. 15, 156 (1972).
22. A. Sesonske, Nuclear Power Plant Design Analysis,USAEC TID-26241 (1973), Chapter 7.
23. D. M. Elliott and L. E. Weaver (Eds.) The Nuclear Fuel Cycle, Oklahoma University
Press, Norman (1972).
24. A. Sesonske, Nuclear Power Plant Design Analysis, USAEC TID-26241 (1973), Chapters
2 and 3.
25. M. M. El-Wakil, Nuclear Energy Conversion, Intext, Scranton (1971), Chapter 17.
26. D. M. Elliott and L. E. Weaver (Eds.) The Nuclear Fuel Cycle, Oklahoma University
Press, Norman (1972).
27. The Technology of Nuclear Reactor Safety, T. J. Thompson and J. G. Beckerley (Eds.),
28. An Assessment of Accident Risks in U. S. Commercial Nuclear Power Plants, USAEC
Document WASH-1400 (1974).
29. A. Sesonske, Nuclear Power Plant Design Analysis, USAEC TID-26241 (1973), Chapter
6.
30. B. J. Toppel, The Argonne Reaction Computation (ARC) System, ANL-7332 (1967).
31. H. C. Honeck, et. al., JOSHUA-Reactor Physics Computational System, in Proceed-
ings of Conference on the Effective Use of Computers in the Nuclear Industry,
Knoxville (1969), pp. 324-336.
32. Mathematical Models and Computational Techniques for Analysis of Nuclear Systems,
USAEC Document CONF-730414, Vols. I and II (1973).
Several universities have developed simplified reactor design code packages

which are far better suited for instructional use than the production codes utilized
by the nuclear reactor industry. Two particularly flexible and comprehensive code
packages are:
Nuclear Engineering Computer Modules, Department of Nuclear Engineering,

Virginia Polytechnic Institute and State University, Blacksburg, Va. (1974)
UMNE Computer Code Library, Department of Nuclear Engineering, The Univer-
sity of Michigan, Ann Arbor, Mi. (1971)
12
Thermal-Hydraulic Analysis of Nuclear
Reactor Cores
I. INTRODUCTION
A. The Relationship of Thermal Core Analysis to Nuclear Reactor

Design
A nuclear power reactor is designed to produce heat that can then be used to
generate electrical energy, usually by way of an associated steam thermal cycle. -
Once we have acknowledged the fact that the primary function of the reactor is
really just that of a rather exotic heat source for turning water into steam, it
becomes apparent that the determination of core temperature distributions and
heat transfer rates, namely, the thermal analysis of the reactor core, must play a
very important role in reactor design. Indeed the design of a reactor core depends
as much on thermal as nuclear considerations. For one must design the core in
such a way that it can produce the desired thermal power without exceeding
temperature limitations on core components that might lead to fuel failure and the
release of radioactive material into the coolant.
Such thermal limitations constitute the primary factor in determining core size.
For as we have seen, a critical mass of fissile material can theoretically operat - b
any power level if sufficient cooling can be provided. Hence one first determines
the core power density that can be accommodated by the intended primary cooling
system, and then determines the reactor core size necessary to meet the desired
reactor thermal power output at this power density. The efficient cooling of the
core also plays a very important role in the detailed fuel element design. For
example, the desire for large contact area between fuel and coolant to enhance heat
transfer will influence the choice of core volume ratios (e.g., coolant-to-fuel).
Furthermore the hydrodynamic behavior of the coolant as it flows through the core
467
will play a role in determining the core lattice design. Since nuclear power reactors
operate at far higher power densities than those characterizing fossil-fueled steam
generators, a core design to optimize thermal performance is obviously of consider-
able importance.
After determining the basic fuel element geometry and core volume from
thermal-hydraulic considerations, one then performs a nuclear analysis of the core
to determine the fissile fuel concentration or loading necessary to allow this core to
operate at rated power over the desired core lifetime. (Of course the required fuel
loading is usually considerably larger than that which would be determined by
purely nuclear considerations.) In this sense, then, the thermal analysis of the core
usually determines the gross features of the core geometry prior to the actual
nuclear core analysis.
The nuclear analysis of a reactor core is rather intimately related to its thermal
analysis in other ways. We have seen that nuclear cross sections that determine
core multiplication depend sensitively on temperature (e.g., the Doppler effect).
Furthermore the material composition of the core depends on thermal considera-
tions because of the density changes accompanying the addition of heat energy to
the coolant (e.g., expansion or vapor formation). On the other hand, since this heat
energy is generated by the fission reactions induced by the neutron flux in the
reactor, the temperature distribution in the core will depend sensitively on its
neutronic behavior.
Thermal considerations also dominate the safety analysis of the reactor design.
Indeed one of the most critical aspects of modern power reactor designs is that of
providing auxiliary cooling systems that can continue to remove heat from the
reactor core should the primary coolant system fail.
A nuclear designer will frequently require information concerning the thermal
behavior of the reactor core. Hence he must be prepared to either interface with the
mechanical engineers responsible for thermal core analysis or undertake this
analysis himself.
Hence, we feel it advisable to include a rather brief overview of thermal core
analysis in this text. Our intent is not to provide a complete development of this
topic, but rather to indicate those areas in which thermal analysis will interact with
nuclear design and hence drift into the domain of nuclear engineering.
Our presentation will essentially follow the path traced out by thermal energy in
the reactor, from its origin as fission heat energy in the fuel until it is finally
removed from the reactor core by the coolant.
Fission-product ------► Conduction Transfer across

energy in fuel through fuel fuel-clad gas gap
Forced convection Transfer from

Conduction across
in coolant out of clad surface
clad
core to coolant
Steam generator -----'► Turbine Condenser
As we have seen, the energy released by the nuclear fission reactions appears
THERMAL-HYDRAULIC ANALYSIS OF NUCLEAR REACTOR CORES / 469
primarily as kinetic energy of the various fission reaction products. The bulk of this
fission product energy is rapidly deposited as heat in the fuel material very close to
the location of the fission event. This heat is then transported via thermal
conduction across the fuel element, across the gap separating the fuel from the
clad, and then across the clad to the clad surface. It is then transferred from the
clad surface to the coolant by forced convection. The mass motion of the coolant
then carries the thermal energy up and out of the reactor core, either as sensible
heat (i.e., coolant temperature rise) or latent heat (i.e., thermally induced phase
change-boiling). (See Figure 12-1.)
The direct concern of the nuclear core designer usually ceases at this point,
although the power-plant engineer must continue to follow the heat energy as it is
used to convert feedwater into steam, then as the steam is used to drive a
turbogenerator, and then finally recondensed into feedwater. Actually the re-
mainder of the thermal cycle will be very intimately coupled to the reactor core
analysis since it will determine variables such as primary system pressure, inlet
coolant temperature, and desired coolant exit conditions.
~( tr
tt l1
1 t
...
C
(1)
E tt t
t
(1)
<ii
<ii
:,
t
LL
FIGURE 12-1. Reactor core cooling.
B. Objectives and Limitations of Thermal Design 1- 5

There are a number of obvious objectives for the thermal performance of a
reactor core. For example, one wishes to achieve as high a core power density as
possible since this will reduce the core volume required for a desired thermal
output. Furthermore high fuel specific powers (power generated per unit mass of
fuel) are desired in order to minimize the required fuel inventory. One also desires
high coolant exit temperatures, since this not only leads to higher thermodynamic
efficiencies, but also to the production of higher temperature steam that will reduce
the demands placed on the turbine.
Unfortunately one's ability to achieve these objectives is restricted by a number
of limitations imposed by the thermal behavior of the core. It is useful to discuss
briefly several of the more significant limitations on core thermal performance
before we consider the thermal analysis of a nuclear reactor core.
l. FUEL TEMPERATURE LIMITATIONS

One of the principal limitations on allowable power densities in a reactor core
arises from the desire to prevent the fuel temperature at any point in the core from
exceeding its melting point. The melting point for the ceramic fuels used in most
power reactors is extremely high, being roughly 2800°C for U0 2 and 2500°C for
UC. 6 Nevertheless the power densities achieved in modern power reactors are
sufficiently high that centerline fuel melting (maximum fuel temperatures normally
occur at the fuel element centerline) represents an important limitation in reactor
core design. It should be noted that such temperature limits are sensitive to fuel
irradiation and tend to decrease with increasing fuel exposure. For example, the
melting point of U0 2 fuel is commonly taken as 2800°C when first irradiated, and
then reduced by 32°C for each 10,000 MWD/MTU burnup. 7
There are other types of fuel thermal limitations. For example, in metallic fuels
one must hold fuel temperatures below those points at which phase changes in the
fuel occur (e.g., T = 667 ° C in uranium metal). Furthermore, the low thermal
conductivity of ceramic fuels leads to high temperature gradients that can cause
fuel cracking and swelling.
Fuel elements subjected to temperatures sufficiently high to induce centerline
melting will experience a significantly higher probability of failure. (By "failure"
here we are referring to a loss in the functional behavior of the fuel element caused
by a change in its physical properties.) Hence, although nothing catastrophic will
happen if a few fuel elements in the core temporarily exceed centerline melting
temperatures, considerable care is taken such that these limits will not be exceeded
during normal core operating conditions.
We will find later that the maximum temperature achieved in the fuel is
relatively insensitive to the diameter of the fuel rod and depends primarily on the
thermal conductivity of the fuel and the linear power density in the core. Therefore
one usually expresses the thermal limits imposed by fuel melting in terms of a
limitation on the maximum achievable linear power density in the core. For
example, in U0 2 -fueled reactors this limitation is about 660 W / cm. Present power
reactors of this type are usually designed to operate with a maximum linear power
density of 460-500 W / cm under normal operating conditions.
2. CLAD THERMAL LIMITATIONS

Although the fuel element cladding in a nuclear reactor core will rarely
experience temperatures approaching its melting point (unless critical heat flux or
burnout conditions are exceeded), the clad thermal behavior does place limitations
on the core power density. The clad serves as a containment barrier to prevent
radioactive fission products produced in the fuel from being released into the
coolant (and subsequently being swept out of the reactor core). The cladding is
subjected to rather severe stresses, both from the high-pressure coolant surrounding
the fuel element, and by fission gas pressure and fuel swelling inside the fuel
element. The buildup of fission gas pressure in the fuel pellet-clad gap and the
cracking and swelling of the fuel pellet can both cause excessive cladding stresses
and strains if not limited. Such phenomena not only depend on fuel irradiation, but
also on thermal history (e.g., temperature variations due to reactor power level
changes). Furthermore the clad is subjected to high thermal stresses due to the
enormous thermal gradients across the clad thickness. Since clad strength is a
sensitive function of temperature and thermal history, one must limit core thermal
performance to avoid clad failure. (Such limitations constitute one of the motiva-
tions for using coated particle fuels such as in the HTGR in which the clad is
actually a ceramic coating of a tiny fuel particle.)
A more critical thermal limitation is frequently placed on the heat flux that can
be transferred from the clad to the coolant in liquid-cooled reactors (e.g., PWRs or
BWRs). Above certain heat flux magnitudes, the heat transfer to the coolant will
become unstable as a film of vapor forms to cover the fuel element surface. At this
point the clad temperature will increase dramatically (several hundred degrees)
leading to clad failure. This thermal limitation, known as the critical heat flux, is of
primary concern in water-cooled reactor cores in which the coolant temperature is
allowed to approach the boiling point.
3. COOLANT TEMPERATURE LIMITATIONS

In many reactor types, one wishes to limit the coolant temperature in order to
achieve suitable coolant behavior. For example, in a PWR one desires to keep the
bulk coolant temperature below its saturation temperature so that no bulk boiling
occurs. Similarly in a LMFBR there is motivation to keep the sodium temperature
sufficiently low that the heat transfer capability of the coolant is maximized.
There are also limitations on the allowable temperature rise of the coolant as it
passes through the core. On one hand, one would like this temperature rise to be
large enough to allow efficient heat transfer in the steam generator; but there is
also strong motivation to minimize this temperature rise in order to lessen thermal
shocks in the reactor core (such as during a reactor scram). The temperature rise in
water-moderated reactors is kept quite small-for example, in a PWR, the coolant
typically rises from an entrance temperature of 293°C to a exit temperature of
315°C-a temperature increase of only 22°C. In an LMFBR, this temperature rise
is somewhat larger, amounting to some 140°C, while in a HTGR, the helium
temperature rises by as much as 500°C. These coolant parameters are summarized
for several reactor types in Appendix H.
It should be noted that while such temperature limitations set upper limits on the
temperatures achieved in the core, there are also lower limits imposed on the inlet
coolant temperature due to the available condensor cooling. The size of a reactor
core will be determined primarily by the amount of heat transfer area necessary to
transfer the desired thermal power utilizing the given temperature differences.
4. OTHER THERMAL LIMITATIONS

There are other types of thermal limitations placed on coolant behavior that
have a more nuclear origin. We have seen that the core multiplication depends
sensitively on the coolant density. For example, in a LWR the coolant serves as a
moderator. Decreasing the coolant density will reduce moderation and hence
reactivity. By way of contrast, in a LMFBR, decreasing coolant density will cause a
hardening of the neutron energy spectrum, which enhances reactivity. Since the
coolant density depends sensitively on temperature, there is frequently a very
strong coupling between the thermal and nuclear behavior of a reactor. It is
essential to account for such coupling, both in static reactor design and reactor
transient studies. Indeed stability considerations will frequently place restrictions
on the allowable temperatures or heat fluxes achieved in the core.
In summary then, the principal limitations on core thermal performance include
the avoidance of significant centerline fuel melting, maintaining the heat flux below
the maximum value allowed by coolant conditions (burnout), and limiting the
stresses due to fission gas release, fuel swelling, and thermal gradients on the clad.
These limitations place constraints on the surface heat flux, the linear power
density, and the volumetric power density in the fuel element. Usually one
determines the maximum heat flux allowed by burnout considerations, and then
adjusts the fuel rod diameter to achieve an optimum compromise between linear
power density (i.e., fuel centerline temperature) and power density (fuel inventory,
core size). 4
It is apparent that the first goal in thermal analysis is to determine the
temperature distribution throughout the core for a given fission power distribution
and coolant inlet condition, since one must ensure that the maximum temperatures
achieved in the core do not exceed any of the thermal limitations. In particular, one
would want to determine the maximum allowable overpower at which the reactor
could operate and still be within the thermal design constraints. It should be kept
in mind that the fission power density is highly nonuniform. Hence the temperature
distribution is similarly nonuniform-and, will depend on time, since the fission
power distribution will shift over core life as the fuel burns nonuniformly. The
usual procedure is to estimate the set of local operating conditions representing the
most extreme case, (e.g., highest temperatures or heat fluxes), and then to design
the core so as to guarantee satisfactory performance for this "hot spot," 1- 3 although
it should be recognized that this may, in fact, yield an overconservative design.
The thermal analysis begins by determining the volumetric source of fission heat
throughout the core. Since this is proportional to the fission rate density, one
requires an initial estimate of the neutron flux distribution in the core. For the
purpose of most thermal analysis, it is assumed that this fission heat source is
confined to the fuel elements.
Since in the steady state this fission heat energy must be removed by the coolant,
one can use a simple energy balance to calculate the coolant temperature as it
passes up through the core using the inlet coolant temperature (presumed known).
Then one can work backward to determine the temperature rise across the clad,
gap, and fuel element to the fuel centerline using the equations of thermal
conduction (Fourier's law) and forced convection (Newton's law of cooling). In
simple analyses, one performs such a calculation only for an average coolant
channel and a hot coolant channel chosen to represent the worst possible case. In
more realistic analyses, one must also take into account flow conditions and the
intermixing of flow between coolant channels.
One is also occasionally interested in the transient thermal behavior of a reactor
when subjected to power level or coolant flow rate changes. Such transient studies
are of very considerable importance in reactor safety studies and in the modeling
of nuclear reactor dynamics.
In this chapter we will sketch the basic concepts and procedures involved in the
thermal analysis of a nuclear reactor core. Although we will usually try to relate
these concepts and methods to well-known laws of thermodynamics, we will give
very few derivations. We do not intend this chapter to be a thorough development
of heat transfer and fluid flow, but rather a brief description of how standard
concepts from these disciplines are applied to nuclear reactor analysis. We have
attempted to include sufficient explanation and discussion of these methods to
allow t.qe reader with only rudimentary exposure to heat transfer and fluid flow to
still benefit from the chapter. For more detailed treatments of many of these topics,
we must refer the reader to one of several standard references. 8- 10
II. POWER GENERATION IN NUCLEAR REACTOR CORES
The energy released in a nuclear fission reaction is distributed among a

variety of reaction products characterized by different ranges and time delays. In
thermal design, the energy deposition distributed over the coolant and structural
materials is frequently reassigned to the fuel in order to simplify the thermal
analysis of the core. We can determine the volumetric fission heat source in the
core q"'(r) by multiplying the fission reaction rate density for each isotope by wfi),
the recoverable energy released per fission event, to find
( 12-1)
Of course, since the flux and number density of the fuel vary across the reactor
core, there will be a corresponding variation in the fission heat source.t
The simplest model of fission heat distribution would correspond to a bare,
homogeneous core. In particular, we should recall the one-group flux distribution
for several typical geometries given in Table 5-1. Of course, the geometry of most
interest will be the cylindrical core, for which we have sketched in Figure 12-2 the
radial and axial flux profiles. We have also indicated the modifications to these
profiles resulting from the addition of reflectors (since most power reactor designs
are reflected, either by coolant bypass flow, additional moderating material, or
perhaps a blanket of fertile material).
To account for heterogeneities in the core lattice, one can use the gross flux
profiles to determine the flux in each of the fuel elements of the core. Usually the
fuel element cross section is so small compared to the core size that we can assume
q'"(r) to be constant across the width of the fuel rod for many calculations. For
example, we would take the fission heat source in a rod at a radius 'r in a bare,
twe will utilize the more or less conventional notation of mechanical engineers when referring to
heat densities in which the number of primes on q indicates the dimensionality of the density.
Hence q'" represents a volumetric heat source in units of W /cm3 ; q" represents a heat flux
(W /cm2); while q' represents a linear heat density (W /cm) typically used to describe the heat
generated per unit length of a fuel rod.
q,(r, 0) I
I
I
1,.-,
-'I '\
• ~ Reflected
-+-----~R~_.___z ~
',\
H __ _ 1'___ _
2 I
I
+----++----+------+--- ~ q,(O.z) .
FIGURE 12-2. Flux distribution in a cylindrical core.
uniform cylindrical core to be
(12-2)
(where, for convenience, we have assumed only one fuel isotope and used a
one-group representation). Here R and iI are to be regarded as effective core
dimensions that include extrapolation lengths as well as an adjustment to account
for a reflected core (i.e., reflector savings).
Notice that while we can frequently neglect the radial variations in the fission
heat generation in the fuel rod, we cannot ignore the axial variation. In fact, if we
remember that there is an appreciable depression of the thermal flux in the fuel rod
due to self-shielding, then it is apparent that there may even be situations in which
we may want to account for radial variation of the fission heat generated in the rod
(such as in the calculation of the Doppler coefficient of reactivity).
There are numerous other factors that will perturb the power distribution of
the reactor core. For example, the reactor is usually not load~d with fuel of
uniform enrichment. At the beginning of core life, higher enrichment fuel is
commonly loaded toward the edge of the core_ in order to flatten the power
distribution. During core operation, the fuel will burn nonuniformly, leading to
power nonuniformities that must be accounted for in thermal core analysis.
Furthermore, subsequent refueling of the core will not be uniform and will lead to
further variation in the power distribution.
Control rods will also have a major effect on the axial and radial flux profiles.
There will also be local peaking of the flux, for instance in water gaps in LWRs or
near reflectors. All of these variations in the flux and power distributions must be
predicted by the nuclear engineer and included in a subsequent thermal analysis of

the core.
From the above discussion, it is apparent that the fission power density will vary
considerably over the core. Such variations occur not only from the usual flux
variations in a bare, homogeneous reactor core, but also due to the heterogeneous
nature of the core, the nonuniformities of fuel burnup and fuel loading, the
presence of coolant gaps and control elements, and small variations due to
manufacturing tolerances of core components. These will cause corresponding
variations in the temperature distribution in the core. A common technique for
accounting for these variations is to estimate a set of local operating conditions for
the "worst case" ("worst" in the sense of most closely approaching thermal
limitations), and to evaluate the reactor performance for this set of conditions.
Associated with such local conditions are so-called "hot channel" factors relating
this extreme case to the average core behavior.
One most commonly defines the hot channel of the core as that coolant channel
where the core heat flux and enthalpy rise is a maximum. Conditions in the hot
channel are defined by several ratios of local conditions to average conditions, and
these ratios, termed the hot channel factors or power peaking factors, are considered
in some detail in Section 12-VII.
The major results of the nuclear calculations on a core can frequently be relayed
to the thermal designer in terms of nuclear hot-channel factors. These factors,
together with an estimate of the axial power distribution shape, can serve as the
basis for the initial plant thermal design, although there will be later interaction
between the thermal and nuclear designs.
Once we are given an axial flux or power distribution and the coolant inlet
conditions (e.g., temperature and pressure), we can comp11te the conditions of the
coolant as it passes up the coolant channel through the core of the reactor. Such a
calculation involves only a simple energy balance equating the heat flux passing
into the coolant from the fuel element to the change in temperature (or, more
generally, the enthalpy) of the coolant as it absorbs this heat.
The more difficult analysis concerns the determination of the radial temperature
rise from the coolant to the clad surface, then across the clad and the gap to the
fuel itself. Such a calculation involves a study of heat conduction through the fuel
element and forced convection heat transfer from the clad surface to the coolant.
This analysis is frequently complicated by a phase change of the coolant as it
passes up the channel. Finally, we are usually also concerned with the
hydrodynamic flow conditions of the coolant, such as the pressure drop along the
channel, flow mixing, and other characteristics of the coolant.
We will consider each of these topics in turn, first developing the relations
between the temperature drop across the fuel, gap, and clad using the thermal
conduction equation; then studying the transfer of heat from the clad surface to
the coolant via forced convection considering both single phase and two-phase
coolants; and finally giving a brief outline of the hydraulic analysis of the core.
III. RADIAL HEAT CONDUCTION IN REACTOR FUEL

ELEMENTS
A. The Equation of Heat Conduction
Let us begin by quickly reviewing the topic of heat transfer in solids via
thermal conduction. 8- 10 If we perform a simple energy balance for an arbitrary
volume in the solid, we arrive at an equation expressing the time rate of energy
change as the difference between energy addition due to distributed heat sources
q"'(r, t) and energy loss due to heat transport:
a
at(pcT)= q"'(r,t)-V·q"(r,t), ( 12-3)
where T (r, t) is the local temperature of the solid, p(r, t) and c are its density and
specific heat, respectively, and q"(r, t) is the heat flux vector expressing the rate at
which heat flows across a surface. This energy conservation equation is derived in a
manner identical to that used in Chapter 5 to derive the neutron diffusion equation.
And as in that case, the balance equation contains two unknowns, T (r, t) and
q" (r, t). We will now introduce an approximation that is the direct analog to the
diffusion approximation used in neutron transport by assuming that the heat flux
vector q" (r, t) is proportional to the temperature gradient
q"(r,t)= -kVT(r,t). (12-4)
This relationship, known as Fourier's law of thermal conduction, is a very good

approximation for most materials. The coefficient of proportionality k is referred
to as the thermal conductivity. [Of course Eq. (12-4) is the analog to Fick's law.]
When we substitute Eq. (12-4) into the energy conservation equation, we arrive at
the equation of thermal conduction:
a
at(pcT)-V·kVT= q'"(r, t). (12-5)
We need both initial and boundary conditions to complete the mathematical

description of heat conduction. These conditions can be derived in a general
fashion using energy balance ideas, but we will use physical arguments to more
directly motivate the conditions appropriate for the specific problems we will
consider.
We will generally consider the case of time-independent heat transport, for
which Eq. (12-5) simplifies to
-V ·kVT (r) = q'"(r). (12-6)
It should be noted here that the thermal conductivity k appearing in this

equation is temperature-dependent in most materials. Although this tem-
perature-dependence is usually ignored in studying thermal conduction in materials
such as the gap or the clad in which a relatively small temperature drop occurs, it
must be considered in treating thermal conduction through the fuel, since in this
case we may have a temperature variation of several thousand degrees.
B. Heat Transfer in Cylindrical Fuel Elements

Most modern power reactor cores are composed of cylindrical fuel elements
that contain ceramic fuel pellets in metallic tubes (i.e., cladding) (although the
HTGR is a notable exception). Our goal will be to calculate the temperature drop
from the centerline of the fuel, where the maximum temperature normally occurs,
to the surface of the clad in terms of the various physical properties of the fuel
element. For the purposes of this calculation, we can regard the clad surface
temperature as a reference temperature, since it will be determined in terms of the

coolant temperature and eventually in terms of the coolant inlet conditions and the
reactor power level. To simplify this calculation, we will make several useful
assumptions:
(1) We will assume that we can neglect thermal conduction in the axial
direction. This is a valid assumption since the temperature gradient across
the fuel element in the radial direction is several orders of magnitude
larger than that in the axial direction. Axial heat transfer will occur
outside the fuel element via forced convection in the coolant.
(2) The fission energy will be assumed to appear as a uniform heat source
distributed through the fuel. Actually there is some variation in this heat
source since it is proportional to the flux in the fuel, and we have seen in
Chapter 10 that there is an appreciable flux depression in the fuel due to
self-shielding effects. However neglect of this spatial variation 1s
frequently adequate in thermal core analysis.
(3) Only steady-state heat transfer in the core will be considered.
We will now apply the equation of heat conduction to study the radial heat
transfer in each region of the reactor fuel element (see Figure 12-3).
I
I
I I
I Coolant I
IL _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _J
FIGURE 12-3. Radial fuel element geometry.
1. FUEL
The heat conduction equation for a cylindrical fuel pin with a uniform
volumetric heat source q"' in which axial heat conduction can be ignored takes the
form
( 12-7)
where we have used axial symmetry to omit the angular dependence.

If kF were constant, we could solve this equation immediately for T(r). However
we must recognize that kF will depend strongly on the fuel temperature. Hence we
will follow a somewhat less direct path by first integrating Eq. (12- 7) from the fuel
centerline at r = 0 to a radius r
(12-8)
Now divide by r and integrate once again, this time from r = 0 to the fuel pin
surfacer= rp,
(12-9)
where we have denoted the fuel centerline and surface temperatures as T'-L = T (0)
and TF = T(rp), respectively. If we now define an average thermal conductivity for
the fuel
(12-10)
we can rewrite Eq. (12-9) to give the temperature drop across the fuel as
(12-11)
In the situation where kF can be assumed constant, one can actually determine the
functional form of T(r) by integrating Eq. (12-7) directly to find a parabolic
temperature profile
(12-12)
Let us define the linear power density of the fuel element q' as the power
generated per unit length
q ,_ 2q"' ,
=7Trp (12-13)
Then Eq. (12-11) can be rewritten in terms of q' as
q'
~ Tl FUEL= -=--- (12-14)
477 kp
In particular we should note that the fuel temperature drop is not dependent on the
fuel radius. This temperature drop depends only on the linear power density q' and
the thermal conductivity of the fuel. This explains why linear power density is a
useful core parameter.
The fuel thermal conductivity is quite low for ceramics such as U02, for which
kF is typically 0.020-0.030 W /cm°K.'1 [Although it should be noted that kF will
change somewhat with fuel exposure.] This low thermal conductivity leads to rather
large temperature differences across the fuel element. For example, a linear power
density of 500 W / cm in a typical LWR would lead to a temperature difference of
1400°C.
Since the fuel centerline temperature Tt is limited by the fuel melting point to
about 2800°C for U02, it becomes apparent that one must limit the linear power
density in order to avoid fuel centerline melting. For U02 fuel elements, the limit
on linear power density is roughly 660 W / cm. Most power reactors operate such
that the maximum linear power density is 420---460 W / cm.
It should also be noted that the weak dependence of the temperature drop on the
fuel radius suggests that, at least from a thermal standpoint, one can achieve the
same linear power density without exceeding core temperature limitations by using
smaller fuel rods. Since the use of smaller rods can reduce fuel inventory, this
feature can have an important bearing on the economics of core design.
It is possible to repeat our analysis of the heat transfer in the fuel for the
situation in which we explicitly account for the flux depression. Such an analysis
indicates that the assumption of a uniform heat source q'" actually implies a
somewhat larger LiTIFuEL than that corresponding to a distributed heat source.
Therefore our calculations above are somewhat on the conservative side.
2. GAP
In reactor types in which the ceramic fuel pellets are encased in a metallic
tube or clad, there will be a small clearance or gap between the fuel and the clad
several thousands •of a cm in thickness which is usually filled with some inert gas
such as helium. Although the thickness of the gap is quite small, the low thermal
conductivities of gases will cause a rather large temperature drop across the gap. If
the gap spacing were uniform, we could simply solve the equation of thermal
conduction for the gas in the gap to find the temperature drop across the gap.
More specifically since there is no heat production in the gap, we would solve
l d dT
- -k r-=0 (12-15)
rdradr ·
If we assume that the thermal conductivity of the gas ka is constant, and also note
that at the fuel surface we must satisfy continuity of heat flux emerging from the
fuel by requiring
I r /fl
-ka dT\ =q"=-q-=~ (12-16)

dr r=rp 2'1Trp 2 '
then we can integrate Eq. (12-15) to find the temperature drop across the gap as
(12-17)
where ta is the gap thickness and Tc is the inside surface temperature of the clad.
Since ta is usually very small ( ~ .005 cm), we can expand the log term to write
q rF t a
llt q I ( ta )
LiTlaAP:::::-2- ka = 21Trp ka . (12-18)
After a period of operation, the gap will contain a mixture of the original fill gas
and fission product gases such as Xe and Kr. Hence the thermal conductivity ka
will change over core life. Typically kHe=0.002 W /cm°K, while kxe+Kr=0.0001
W /cm°K. 12 Hence a linear power density of 500 W /cm and a gap thickness of
ta= .005 cm would give rise to a temperature difference across the gap of 300°C or
more.
480 / AN INTRODUCTION TO NUCLEAR REACTOR-CORE DESIGN
There is another much more significant burnup effect, however. As fission events
occur in the fuel, the fuel pellets will swell and crack and will actually come into
contact with the clad in many places (see Figure 12-4). Such a phenomenon is very
difficult to treat analytically, and hence one usually defines an effective coefficient
of gap heat transfer ha such that the temperature drop across the gap is
q"
~TlaAP= h. (12-19)
a
The empirical coefficient ha will be a function of the mean gap thickness, gas
conductivity, the contact pressure between the fuel pellet and the clad, surface
roughness, clad material, and, of course, burnup. It typically ranges in value
between 0.5 to 1. I W / cm 2 ° K, depending on the material type, the fuel exposure,
and the contact pressure between the clad and the pellet. 13
If we recognize that the heat flux across the gap in the steady state must be just
that amount of heat produced in the fuel divided by the surface area of the fuel
q"'(7rr 2H) "' ,

q" = F = q rF = _q_
(12-20)
27TrpH 2 27Trp '
then we can calculate the temperature drop across the gap as
q"' rF q'
~TlaAP= ~ = 27Tr h (12-21)
a Fa
Prior to
irradiation
After
irradiation
FIGURE 12-4. The effect of irradiation on the fuel-

clad gap.
For a linear power density of 500 W /cm this temperature difference can range
from 140° to 280°C, depending on the value chosen for h 0 .
3. CLAD
If we assume that the clad thermal conductivity kc is constant and that there
is no heat production in the clad, we can solve
_!_ .!!.__ rk dT =0 (12-22)

r dr cd,
to find the temperature drop across the clad as
q'"r F2 rF + t C
t::.T ICLAD= Tc-Ts= -2k ln--- (12-23)
c rF
where Ts is the oater clad surface temperature and t c is the clad thickness. (We
have ignored the gap thickness t0 in this calculation.) In practice the clad thickness
le is quite small compared to the fuel rod diameter, and hence we are once again
justified in expanding the log to write
q"' rF t C q' tC
t::.TlcLAD~-2- k = 27Tr k (12-24)
C F C
for the temperature drop across the clad. Clad materials are usually chosen to have
large thermal conductivity (e.g., kzr = 0.11 W /cm° K). This fact, when combined
with the small clad thickness le, leads to relatively small temperature drops across
the clad. For example, at q' = 500 W / cm, a clad of thickness tc = .053 cm would
experience a temperature drop of 78°C.
4. CLAD SURF ACE TO COOLANT

Heat transfer from the clad surface to the coolant is described by Newton's
law of cooling
(12-25)
where h. is the convective heat-transfer coefficient. This coefficient will depend

upon the properties and flow conditions of the coolant, and its determination will
be the subject of the next section of this chapter. For now, however, we will assume
that it is known, and proceed to calculate the temperature drop to the coolant using
Eq. (12-25) for q" to find
(12-26)
As we will later find, heat transfer from the clad surface to the coolant is an
extremely efficient process and values of h8 range as high as 4.5 W / cm 2 ° K. In a
LWR, t::.TlcooL is typically only about 10--20°C.
5. TOTAL TEMPERATURE DROP ACROSS THE FUEL ELEMENT

We can sum all of these separate temperature drops from the fuel centerline
to the coolant to write
(12-27)
Now, in general, the coolant temperature is essentially fixed by the remaining

thermal components of the plant (such as the steam generator and condensor).
Furthermore, regardless of where we are along the axis of the core, the coolant
temperature varies comparatively little (e.g., in a LWR, the coolant rises from 293°
to 315°C, while in a LMFBR it rises from 370° to 510°C). The fuel centerline
temperature is constrained to remain below the melting point of the. fuel (e.g.,
2800°C). Hence the allowable temperature drop across the fuel element is limited
by the fuel melting point; there will be corresponding limits on the allowable linear
power density. This maximum allowable linear power density essentially deter-
mines the size of the reactor core for a desired power level. If _we rewrite Eq.
(12-27) as
(12-28)
we can see that there is strong motivation to maximize the heat transfer coefficients
h0 , h8 , kc, and kF in order to maximize the allowable linear power density and
hence minimize the required core size.
In Table 12-1 we have listed several thermal properties of the common fuel, clad,
and coolant materials used in nuclear reactor cores. 3 By way of illustration, we
have used these data to calculate the temperature drops across a LWR fuel element
generating a linear power density of q' = 500 W / cm. The temperature profile across
the fuel element is shown in Figure 12-5.
TABLE 12-1: Some Physical Properties of Interest in Fuel-Rod Design
Linear Thermal
Thermal Expansion Heat
Temperature Density Conductivity Coefficient Capacity
Material oc g/cm3 W/cm°K m/m°K J/kg
UO2 -Unirradiated (95%p,1l 540-2700 10.4 .024 l.75x 10- 5 221
UC-Unirradiated (92%pT) 540-1400 12.6 .130 l.08X 10- 5 140
Stainless steel (Type 304) 340 8.0 .163 l.73X 10- 5 325
Zircaloy IV 340 6.44 .107 6.lOX 10- 6 183
Water at 155 bar 300 0.72 .004 2930
Sodium 540 0.81 .542 698
Helium 340 .002
Fission gas mixture (Xe+ Kr) 340 l.3X 10- 4
t P.rTheoretical density
2000°c
1000°C
0°C
t
t t
FIGURE 12-5. Temperature distribution in a cylindrical fuel pin.
IV. FORCED CONVECTION HEAT TRANSFER IN SINGLE-

PHASE COOLANTS
In the preceding section we treated the transfer of heat from the clad surface
to the coolant using Newton's law of cooling
(12-29)
where h. is the coefficient of convective heat transfer. The magnitude of h. varies

greatly for different types of coolants and flow conditions. For example, h. in water
can range from .05 to as high as 4.5 W / cm2 ° K (if boiling occurs), while for a gas
such as helium, h 8 is only .005-.05 W /cm2 °K. We will study how h 8 is determined
for die case of forced convection heat transfer-namely, that situation in which
heat is transported away from a surface by mass motion of a coolant adjacent to
the surface maintained by an externally induced flow (e.g., coolant pumps), in
contrast to natural convection, in which the density variation in a temperature
gradient induces a flow (e.g., in a swimming-pool reactor).
A. Hydraulic Flow in Channels

The determination of the convective heat transfer coefficient h. for various
coolant and flow conditions characteristic of nuclear power reactors is intimately
related to the subject of hydraulics, that is, the study of incompressible flow in
channels. 14 We will avoid a detailed discussion of hydraulics and application of
hydrodynamics equations to describe forced convection in reactor cores in favor of
a brief survey of the principal results of significance for reactor analysis. Since the
coolant flow in most power reactors is highly turbulent, the quantitative analysis of
thermal-hydraulics in reactor cores has a highly empirical flavor in any event.
484 / AN INTRODUCTION TO NUCLEAR REACTOR-CORE DESIGN
The pressure drop experienced by an incompressible fluid flowing in a channel

of circular cross section can be written as
llp=2(L/ D )pu;J, (12-30)
where L is the length of the channel, D is its diameter, p is the fluid density, and u,
is its mean flow velocity along the channel. An empirical factor f known as the
Fanning friction factort, has been inserted to account for the effect of turbulent
flow. In general this factor will depend both on the roughness of the channel wall
and the Reynold's number characterizing the flow which is defined by
(12-31)
where µ, is the shear viscosity of the fluid. For turbulent flow, the friction factor f
must be measured experimentally as a function of both Re and channel surface
roughness and is customarily plotted against these parameters on a chart known as
a Moody diagram (see Figure 12-6). On this chart the transition from laminar to
turbulent flow occurs at a critical Reynolds number of Re*= 2100.
roughness E
1.0
0.09
Complete turbulence, rough pipes
0.08
0.07 0.05
0.04
0.06
0.03
~
C
·a 0.05 0.02
<t
•
s;:;
0.04
0.015
0.01
0.008 ~1<::i
J 1~~
0.006 i!l-
0.03
0.004
"'
C
-,1<::i ,:::
O">
:,
0.025
0.002 e
·;; "'>
0.001
0.0008 ,; "'
I
o.02r;t:-:j~-~:;;;;:~,:;;:;;-:~t~~:====t::5~===~====~
0.003-0.03 0.0006 er
Concrete 0.001-0.01 0.0004
0.015 Wood stave 0.0006-0.003
Cast iron 0.00085 0.0002
Galvanized iron 0.0005 0.0001
Asphalted cast iron 0.0004
Commercial steel or 0.000,05
O.Ol wrought iron 0.00015 .......
0.009 Drawn tubing 0.000005
0.000,01
0.0086 1Q3 10• 1Q5 106 10 7 10 8
0.000,005
Reynolds number, DiizP
µ 0.000,001
FIGURE 12-6. The Moody diagram for the Fanning friction factor. [L. F. Moody, Trans.
ASME 66, 671 (1944)].
For laminar flow (Re< 2100), one can explicitly solve the equations of hydro-
dynamics characterizing incompressible flow in a circular pipe (so-called Hagen-
Poiseuille flow) to obtain the friction factor as
f= 16/Re, Re<2100 (laminar flow). (12-32)
For turbulent flow (Re > 2100) one must use empirical correlations for the friction
tNote that one occasionally encounters an alternative factor known as the Darcy-Weisbach
friction factor. The relationship between the two friction factors isfFanning=(l/4)foarcey-Weisbach·
factor. In particular for turbulent flow in smooth pipes (such as arise in reactor
applications), one customarily uses the Blasius formula 10
J=0.0791 Re- 0 ·25 , 2100<Re< 105 (turbulent flow). ( 12-33)
Frequently it becomes necessary to extend these results to flow in channels of

non circular cross section. For this purpose, one replaces D by an equivalent
hydraulic diameter Dh defined by
(12-34)
where S is the flow area and Z is the wetted perimeter of the flow.
As an example, suppose we consider flow through a bundle of circular fuel
elements of diameter d and square pitch p. Then one defines the equivalent
hydraulic diameter by locating a unit cell and calculating S and Z for this cell as
S = p 2 - 1rd 2 / 4, Z = 1rd (see Figure 12-7). Hence for this geometry, we find
(12-35)
The hydraulic diameter for triangular lattice geometries is given in Figure 12-7.
~ p-------j
0001 0001
0 0---0 0---000
0
6(3" chaaae,
Flow chaaaeo
Square lattice Triangular lattice
Dh =d [ ~ ( ~)2 - ~ Dh = d[
2
f (1)2- 1]
FIGURE 12-7. Typical coolant channels.
In summary then, one can use Eq. (12-30) and corresponding expressions for f
and Dh to calculate the pressure drop for either laminar or turbulent flow of a fluid
in a channel, such as for a coolant passing through a reactor core. We will utilize
these results later when we discuss the hydraulic analysis of coolant flow in a
reactor core. However we will first turn to an investigation of heat transfer from a
surface to a moving fluid.
B. Forced Convection Heat Transfer

We have indicated that coolant flow in nuclear reactor cores is highly
turbulent. This turbulence tends to greatly enhance heat transfer from the clad
surface to the coolant over that which would be present if the flow were laminar.
However the complexity of fluid turbulence once again forces us to rely on various
486 / AN INTRODUCilON TO NUCLEAR REACTOR CORE DESIGN
empirical correlations for calculating quantities such as the convective heat transfer
coefficient h•.
The convective heat transfer coefficient h. is usually expressed in terms of the
thermal conductivity of the fluid k, the hydraulic diameter of the channel Dh, and a
dimensionless parameter Nu
(12-36)
where Nu is known as the Nusselt number which characterizes both the physical
properties of the fluid and the dynamical characteristics of its flow. For a flow that
is laminar with a constant wall heat flux passing into the fluid, one can solve the
equations of hydrodynamics8- 10 to calculate Nu= 48 / 11 = 4.364. For turbulent
flow, one must again rely on empirical correlations to determine Nu (in the same
fashion as we used the Blasius formula to determine the friction factor f).
Fortunately one can use dimensional analysis to find that the Nusselt number can
at most be a function of two other dimensionless parameters, the Reynolds number
Re which characterizes the conditions of the flow, and a new parameter, the
Prandtl number Pr,
Pr=µc/k, (12-37)
which characterizes the physical properties of the coolant fluid.

Several of the more common correlations for the Nusselt number (and hence the
convective heat transfer coefficient, hJ are tabulated in Table 12-2 for various
ranges of Prandtl numbers. The most common such correlations used in reactor
analysis are the Dittus-Boelter correlation, which characterizes fluids with Pr~ 1
such as water or helium
(12-38)
and the Lyon-Martinelli correlation, which characterizes liquid metals subject to a

constant-wall heat flux:
Nu=6.3+0.030 (RePr)°-8• (12-39)
TABLE 12-2: Common Forced Convection Correlations15
Pr< 0.1 (liquid metals)
Nu=6.3+0.03 (Re Pr) 0·8 (constant heat flux)
Nu=4.8+0.03 (Re Pr)°- 8 (constant temperature)
0.5<Pr< 1.0
Nu= 0.022 Pr0 ·6 Re0·8 (constant heat flux)
Nu=0.021 Pr0 ·6 Re0·8 (constant temperature)
1.0< Pr<20 (water and light liquids)
Nu=0.0155 Pr°·6 Reo.s3
Pr > 20 (oils and other viscous liquids)
Nu=0.0118 Pr0·3 Re0 ·9
C. Choice of Reactor Coolants

With the preceding discussion of the analysis of fluid flow and heat transfer
serving as a background, let us now turn briefly to the considerations involved in
choosing a reactor coolant. From a nuclear standpoint, one desires coolants with
low neutron absorption cross sections, low induced radioactivity, good radiation
stability, and, depending on whether one is concerned with thermal or fast reactors,
strong or weak moderating properties. And, of course, one desires high heat
transfer coefficients, low pumping losses, and low cost.
Clearly, two of the characteristics of greatest concern are the magnitude of the
heat transfer coefficient h. and the size of the pumping power required to pump the
coolant through the primary coolant loop (reactor core, piping, heat exchangers,
etc.). The value of the heat transfer coefficient depends on many factors, such as
the geometrical shape of the channel, the flow rate, the heat flux, the temperature,
as well as the physical properties of the coolant itself. The pumping power can be
estimated in terms of the pressure drop t::.p across the core as
pumping power=t::.pAJiz, (12-40)
where Ac is the cross-sectional flow area and uz is the mean flow velocity.
In order to evaluate the suitability of various candidates for reactor coolants, we
have compared a number of thermal and hydraulic parameters characterizing
typical coolants in Table 12-3.
TABLE 12-3: Physical Properties of Some Typical Coolants
Tinlet !:,. Tcore Pressure t:,.Pcore u, Density h.

Coolant [OC] (OC] [bar] (bar] [cm/sec] [g/cm 3] [W /cm2 °K]
H 20 300 20 100 2 400 .70 2.8-4.S
He 500 400 40 16,000 .0024 .005-.05
Na 400 ISO 10 10 400 .85 2.2-5.S
Clearly in terms of these quantities, liquid metals would make the best coolants.
However there are of course, a number of nuclear advantages in coolants such as
water that often outweigh the heat transfer capability of liquid metals (e.g.,
moderating power). One should also note here the very high flow rates required in
order for a gas coolant to achieve a heat-transfer capability comparable to that of a
liquid. Such flow rates require rather large pumping powers. Indeed in an HTGR
roughly 5-10% of the power output is required just to pump the coolant through
the core. As another example, the height of an LMFBR core is to a large degree
determined by the allowable pressure drop across the core ( ~ 10 bar) which can be
handled by the primary pumps.
We will return later to compare the more detailed thermal-hydraulic analysis of
each of the principal reactor types. For now, however, we will proceed to discuss
the determination of axial temperature distributions in a reactor core cooled by a
single-phase coolant.
D. Axial Temperature Distributions

Consider now the determination of the axial temperature distribution in a
channel containing a single-phase coolant. For purposes of illustration, we will first
assume that the axial power distribution is that characterizing a bare, homogeneous
cylindrical core and given by Eq. (12-1). For more general axial power distribu-
tions, one will usually have to resort to numerical methods such as those discussed
later in this chapter.
We will consider a coolant channel in the reactor to be defined as the flow area
associated with one fuel element. As Figure 12-7 indicates, such channels are rarely
circular in cross-section. However by using an equivalent hydraulic diameter, we
can apply our earlier analysis.
To compute the temperature of the coolant as it passes up the coolant channel,
we simply perform an energy balance by equating the heat energy gained by the
coolant to that produced in the fuel and passed through the channel walls over a
small distance dz
(12-41)
where w is the mass flow rate and the linear power density is just
q' ( z) = qo cos( 7TZ I ii) (12-42)
where qo = wrJ.wrLf/>of (2.405 rr/ R ).
It should be noted that by restricting ourselves to a single-phase coolant we have

assumed that all the heat is absorbed as sensible heat and that no coolant phase
change is allowed to occur in the channel. In the more general case, we would use
the specific enthalpy h (or energy content per unit mass of coolant) to write the
energy balance [Eq. (12-41)] as simply
wdh = q'(z)dz. (12-43)
We can now integrate Eq. (12-41) up the channel from the coolant inlet at
z = - H /2 to find
wcP ),( Tn dT= q0,

T,nlet
f
-H/2
z wz' ) ,
dz , cos ( ---
H
(12-44)
or
'ii
Tn(z)- Tinlet= ~
7TCPW
[sin( w_z) +sin( w1!_) ]·
H 2H
(12-45)
In particular, the exit coolant temperature is given by
(12-46)
[For example, in a LWR, (ii - H)/ H ~20 cm/366 cm~.06.] We have now
evaluated the coolant temperature distribution at any point z in the channel:
(12-47)
Hence using our earlier work on thermal conduction in radial fuel elements, we can
work back to calculate the temperatures in the clad and fuel by simply adding on
the relevant temperature differences, say, as given by Eq. (12-27). We have plotted
the axial temperature profile of the various fuel element temperatures in Figure
12-8. It should be noted that while the power density in a bare, uniform core peaks
at the midplane (z = 0), the fuel element temperatures peak somewhat above this
point for the simple cosine axial power distribution.
H z
2
FIGURE 12-8. Axial temperature distributions for a cylindrical fuel element.
It is important to keep in mind the limited validity of the above calculations. In

general, the axial power profile in a reactor core is not of the form given by Eq.
(12-42) due to fuel and moderator density variations. Furthermore, we have
assumed in our analysis that the thermal parameters such as h 8 and cP are constant
when in fact they may vary considerably as one passes up the coolant channel.
Finally, we have ignored the possibility of coolant phase change which can very
strongly modify these temperature profiles.
V. BOILING HEAT TRANSFER IN NUCLEAR REACTORS
A. Introduction
Thus far we have restricted our analysis of heat transfer between the clad
surface and the coolant to situations in which there is no phase change. However
many types of power reactors are designed to operate so that there will, in fact, be
localized boiling at the clad surface, since boiling heat transfer is usually the most
efficient heat transfer mechanism. Such localized boiling does not necessarily lead
to bulk boiling of the coolant, however. For example, in a PWR, the clad surface
temperature is allowed to exceed the saturation temperature of the coolant.
However although small vapor bubbles will form on the clad surface, they quickly
collapse as they leave this surface. This phenomenon, known as subcooled boiling, is
an extremely efficient heat transfer mechanism.
By way of contrast, in a BWR the bulk coolant temperature is allowed to exceed
the saturation temperature. Bulk or saturated boiling occurs and results in appreci-
able vapor (steam) formation within the core.
The primary form of boiling heat transfer used in nuclear reactors is so-called
nucleate boiling, in which many small bubbles form around small nucleation points
on the clad surface and then break away into the coolant. It is desirable to operate
in the nucleate boiling regime since this is the most efficient form of heat transfer.
Indeed the efficiency of heat transfer in the nucleate boiling regime is so high that
the clad surface temperature will rise no more than 3-6°C above the fluid
saturation temperature in a LWR.
There are disadvantages accompanying nucleate boiling heat transfer, however.
The pressure drop associated with this mechanism is somewhat larger than that
characterizing single-phase heat transfer. But even more serious is the fact that heat
transfer instabilities may set in. One must take care that the heat flux does not
become sufficiently large that the small bubbles formed in nucleate boiling coalesce
into a vapor film that covers the surface, since then the heat transfer efficiency
drops dramatically, and the clad surface temperature will rise by several hundred
degrees. This situation is known as the departure from nucleate boiling (DNB) and
will be discussed in greater detail later.
In this section we will briefly discuss several more general aspects of boiling heat
transfer, and then we will introduce the various methods used to analyze this
phenomenon in thermal core analysis.
B. Boiling Heat Transfer

Suppose we first consider a heating element submerged in a pool of stationary
fluid. A sketch of the heat flux transferred from the heating surface to the fluid
versus the temperature difference between the surface and the bulk fluid tempera-
ture is shown in Figure 12-9. 15 From this sketch it is evident that one can
distinguish several different regions of heat transfer.
0-a: There is little liquid superheat; heat transfer is by natural convec-
tion.
a-b: A few bubbles are formed but collapse after leaving the surface
(although agitation will increase heat transfer).
a-c: The number of bubbles formed increases rapidly-this is the nucleate
boiling regime.
c: "Burnout" point or DNB is reached.
c-d: Bubbles become so numerous that they begin to coalesce and clump
near the heating surface (the vapor covering the surface acts as a
heat insulator)-this is known as partial nucleate boiling.
d-e: A continuous blanket of vapor forms over the heating surface-film
boiling regime.
e-f: Thermal radiation from the surface comes into play-film and
radiation regime.
It should particularly be noted that if the heat flux q" is the independent variable,
then increasing q" beyond the DNB point will result in a large increase in the
surface temperature for a given coolant temperature (see Figure 12-9.) This
phenomenon is known as burnout and must be avoided in reactor operation.
Single-phase Nucleate Partial Film

6 convection boiling film boiling
10
boiling
I
-➔ i--➔---➔-
,as
I
I
I
I
104
103 ,___ _ _ _...__ ____.__ ___.__ _ ____.__._ _......__ __.___ _ _ __..__ ____,
10- 1 10 10 2 103 104

Ts - T,, [oC]
FIGURE 12-9. Heat flux versus temperature difference for pool-boiling heat transfer.
These concepts also apply in a flowing channel. Suppose we consider a channel

with a uniform wall heat flux q;an· Then one will find temperature profiles up the
channel as shown in Figure 12-10. Once again we can distinguish several regions of
heat transfer:
.. ] 6
T
CD 0 0
.___ _ _ _ ___.__ ____.__ _ _ _ _ _ _ _ _ _ _ _ z
FIGURE 12-10. Boiling-channel temperature profiles.
(1) Region (D: Single-phase convection.

In this region, we can use the analysis we developed in the previous
section to determine the convective heat-transfer coefficient hs
arising in Newton's law of cooling
(12-48)
For example, for turbulent flow one customarily would use the
Dittus-Boelter equation (12-38) to determine Nu and hence h 5 • For
such single-phase forced convection, h. is relatively constant for

fixed values of coolant velocity, fuel geometry, and coolant proper-
ties. Since the coolant flow is outside and parallel to the fuel
elements, one must use the equivalent hydraulic diameter in this
expression.
(2) Region@: Subcooled boiling: Ts> T.a1but Tn < Tsat·
Once the clad surface temperature reaches the saturation tempera-
ture of the coolant, the temperature drop is no longer linear in the
heat flux. In this region, the heat transfer process is extremely
efficient, and the temperature difference is quite small compared to
the other temperature differences in the core. In LWRs it is cus-
tomary to use an empirical correlation developed by Jens and
Lottes 16 in order to calculate the temperature difference between
the clad surface and the coolant:
( 12-49)
where p is the pressure in units of bar and q" is the heat flux in
units of MW /m 2 • This correlation is valid for pressures, tempera-
tures, and coolant flow rates characteristic of both PWRs and
BWRs. For example, in a PWR at 155 bar with a heat flux of
1.5 X 106 W /m 2 , one would find a temperature difference of only
3°C. There will be very little change in the temperature difference
with heat flux in this boiling regime. Once the clad surface tem-
perature reaches the saturation temperature of the coolant, it tends
to remain near that level unless the heat flux reaches the DNB
value.
One can use these relationships to determine the position zscB at
which subcooled boiling begins. We equate Tn = Tsat in Eq. (12-49):
(12-50)
Then noting also that
(12-51)
we can find
q"
Tnlzsce = Ts - h
(12-52)
and then solve this for the position zscB·

(3) Region Q): Saturated or bulk boiling: T n = Tsat·
Eventually sufficient heat is transferred to the coolant that it
reaches its saturation temperature and begins bulk boiling. Sub-
sequent heat addition will cause further boiling, and hence the
coolant temperature will remain essentially constant and equal to
Tsat up the remainder of the channel.
C. Two-Phase Flow
When the coolant is a two-phase mixture of liquid and vapor, the coolant
flow pattern can become quite complicated. One can distinguish several types of
two-phase flow, as shown in Figure 12-11. It is useful to introduce some definitions
characterizing the vapor content of a coolant at rest:
. . mass of vapor in mixture

Static quahty=x.= . ,
total mass of mixture
volume of vapor in mixture
Void fraction= a (12-53)
total volume of liquid-vapor mixture ·
We can compute these quantities in terms of the cross-sectional areas in the

channels occupied by liquid (A 1) and vapor (Ag). The void fraction is given by
( 12-54)
cross-section
Unstable slug flow
0
0 "C
0
0 ~0 C: ""
0 :..=
0
"
.0 0 ·-
Subcooled o :, .0
(/)
boiling flow
FIGURE 12-11. Flow patterns in a heated channel.
while the static quality is
( 12-55)
For example, the void fraction is given for water at various qualities and pressures
in Figure 12-12.
1.0 p = 1 bar
0.9
0.8
0.7
i:j 0.6
C
0
·;:;
u 0.5
.....e
'O
'i5 0.4
>
0.3
0.2
0.1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Quality, x
FIGURE 12-12. Void fraction versus quality for bulk boiling.
It is also useful to express the specific enthalpy h in terms of quality. First recall
that for a single-phase fluid
(12-56)
If we denote the heat of vaporization of a liquid as hrg, then the enthalpy of a

saturated vapor at Tsat is just
(12-57)
If we have a mixture of liquid and vapor, then the enthalpy of the mixture is just
(12-58)
We can actually turn this around to compute the quality in terms of the enthalpy as
(12-59)
(Note Eq. (12-59) can be used to define a negative quality.) Typical values of
enthalpy for various temperatures and pressures are given in Table 12-4 for
illustration. 17
THERMAL-HYDRAULIC ANALYSIS OF NUCLEAR REACTOR CORES I 495
TABLE 12-4 Thermodynamic Data on Steam
P(bar) r ••,1oq hr[MJ/kg] h18[MJ/kg] cP Xs P1/P 8 a

1 100 .419 2.26 4.186 .I 1602 .995
70 285 1.26 1.51 5.44 .1 20.6 .7
155 345 1.68 .96 7.95 .1 5.9 .4
218 373 2.10 0 ~62 .1 .1
When the two-phase mixture is flowing, we must be more careful in our

definitions since the vapor has a tendency to slip past the liquid because of its
bouyancy. We then must define the quality in terms of a parameter S known as the
slip ratio which relates the different flow velocities:
Slip ratio= S = vg/ vi' (12-60)

Then we can define a flow quality
mass flow rate of vapor
x= total mass flow rate
=---
wg + w1
(12-61)
or
I
x=------- (12-62)
I+~( 1-a)_!_
Pg a S
One can invert this to write the void fraction a as
1
a=-------
1+ _Pg (-1_-_x )s
(12-63)
Pt X
This latter equation is sometimes known as the "S - a - x" relationship.

D. Heat Addition to Boiling Flow
Consider now a fluid flowing through a channel adjacent to a fuel element
generating a linear power density q'(z). Then we can write the energy balance for
the fluid as
wdh = q'(z)dz. (12-64)
If we integrate this in the usual manner, we can find the enthalpy at any point in
the coolant channel as
h(z)=hinlet+ -1
w
izq'(z)dz.
0
(12-65)
Now suppose that bulk boiling begins at a point z 88 up the channel; that is, up to
z88 the coolant temperature rises to Tsat· Above z 88 , there is no further temperature
change,· but instead vapor formation occurs. We can determine z 88 from Eq.
(12-65) by solving
(12-66)
for z88 . Now for z > z88 , the heat addition will cause a phase change. Then one
finds from Eq. (12-58) that the enthalpy can be expressed in terms of the flow
quality as
(12-67)
If we combine Eqs. (12-66) and (12-67), we find that the quality at any point
z > z 88 in the channel can be written as
I
x(z)= -h- fz q'(z)dz, (12-68)
W fg zaa
Once we know x(z), then we can determine the void fraction a(z) provided we
know the slip ratio S (z). In general, S has a vulue of about 1 near z 88 and tends to
increase from this value to a maximum of about 3 as one passes up the channel.
A knowledge of the void fraction is essential to the nuclear analysis of the core,
since a determines the coolant density and hence the macroscopic cross section of
the coolant. We will later find that there is a very strong coupling between the
thermal core behavior which determines a(z) and the axial flux profile in reactor
cores in which boiling occurs. As we mentioned earlier, the void fraction is
extremely important in the calculation of the void coefficient of reactivity. There is
very strong coupling between the thermal core behavior which determines a(z) and
the axial flux profile in boiling water reactors.
E. Flow Boiling Crisis or DNB

Our earlier discussion of boiling heat transfer indicated that if the wall heat
flux exceeded a certain magnitude, there would be an unstable transition from
nucleate boiling to film boiling with a corresponding increase in wall surface
temperature. This "boiling crisis," in which the clad surface temperature rises
dramatically, corresponds to a sudden drop in the heat transfer coefficient h. due
to a change in the boiling mechanism from nucleate to film boiling. The boiling
crisis is known by a number of names, including burnout, critical heat flux, and
DNB. Such a phenomenon in a flowing coolant is more complicated than a
pool-boiling crisis due to the added effects of forced convention and bubble
clouding that tend to shield the heat transfer surface. Flow instabilities can also
complicate the situation. For example, a sudden drop in pressure can increase the
local superheat and perhaps lead to DNB.
One should be a bit more careful in terminology here, since for low flow rates,
one may not have nucleate boiling prior to exceeding the critical heat flux, and
hence strictly speaking DNB does not occur. Furthermore the heat flux may be
sufficiently low so that the subsequent temperature rise of the clad is still below
melting or failure limits. (This might occur at high exit qualities in a BWR in which
the wall heat flux was only of the order of 3 x 105 W /m 2.) However, in most power
reactors, the heat flux approaches 3 X 106 W /m2 , and exceeding the critical heat
flux will almost certainly lead to clad failure.
The detailed mechanism involved in such a flow crisis is closely related to the
flow pattern. There are basically two types of phenomenon of interest in nuclear
reactor studies (see Figure 12-13):
(l) Subcooled or low-quality DNB: Here DNB can be caused by detachment
of the bubble boundary layer. This is the type of DNB that one would
design against in a PWR.
(2) Burnout in high-quality region: In this phenomenon, burnout occurs in a
liquid film that covers the heat transfer surface. Such a process is of
concern in BWRs. Burnout in the high-quality region will occur when the
liquid film "dries out." This process is slow compared to subcooled DNB,
since the high-velocity vapor core provides fairly good single-phase con-
vection heat transfer.
.
, •
0
0 Burnout , Liquid
(J. go{) film
t!) O O 0
0
.i> 0
0
I
-0
,
~o q" e0 i
~ ...
en
00 0 u ., .,
0.-
00 0
0 ~ 0 .!!!
0. -0 ~
g-o
00 "'J: -0
0
>.~
0 3:
0
•
•
Subcooled or low-quality DNB
FIGURE 12-13. Types of boiling-flow crisis.

1 0
Burnout in high-quality regions
It is extremely important to be able to determine the critical heat flux at which

DNB will occur. Unfortunately this heat flux depends on a large number of factors,
such as channel shape, surface condition, physical properties of the coolant, and
flow conditions. Various empirical correlations based on experimental data have
been developed for those ranges of conditions encountered in modern power
reactors. However because of the complexity (and rapid obsolescence) of such
correlations, we will refer the reader elsewhere for more information. 1• 5• 18 • 19
The core designer must use such correlations to ensure that the critical heat flux
is not exceeded during core operation. One usually defines the minimum ratio of
the critical heat flux to the heat flux achieved in the core as the DNB ratio (DNBR)
(see Figure 12-14),
q" ( )
DNBR= DNB z . ( 12-69)
q" ( z)
One must design the reactor core so as to prevent the DNBR from dropping below
z
FIGURE 12-14. Critical heat flux in a uniformly loaded core.
a chosen value under a high heat flux transient condition for the most adverse set
of mechanical and coolant conditions. For example, in PWRs a common constraint
is to maintain DNBR > 1.3 in the event of a 115% overpower condition.7
F. Axial Temperature Distributions

Thus far we have confined our discussion of axial temperature distributions
to channels containing a single-phase coolant. Let us now consider how such
temperature distributions would be altered when coolant boiling is present. First
consider the "hot" channel of a PWR in which nucleate boiling occurs. The
coolant enters the channel with a temperature T < Tsat· After passing some distance
up the channel, the clad surface temperature reaches Tsat (although the bulk
coolant temperature Tn remains below Tsat throughout the channel), and nucleate
boiling heat transfer begins. Such heat transfer is so efficient that the clad surface
temperature remains essentially constant the rest of the way up the channel (see
Figure 12-15a).3
Consider now a coolant channel in a BWR. Once again the coolant enters below
saturation temperature, and the clad surface temperature will eventually reach Tsat
before the bulk coolant temperature. However further up the channel the coolant
reaches saturation, and additional heat added to the channel increases its quality
(see Figure 12-15b)3. Both the coolant and clad surface temperatures will remain
essentially constant after boiling begins.
We should again insert a disclaimer concerning these diagrams which have been
sketched for a simple cosine power profile. As we have mentioned, the coolant
density variations will affect core multiplication and hence the shape of the power
profile. As we will see later, in LWRs the fact that the coolant density is higher
near inlet tends to shift the axial power peak toward the bottom (inlet) of the core.
VI. HYDRODYNAMIC CORE ANALYSIS
The thermal performance of a reactor core will depend strongly on coolant

flow behavior. For example, higher coolant flow rates will lead to better heat
Temperature profiles in a PWR
Nonboiling
length Boiling length
T sat
FIGURE 12-15. Temperature profiles in a BWR.
transfer coefficients and higher critical heat flux limits. However higher flow rates
will also result in larger pressure drops across the core, hence larger required
pumping powers and hydrodynamic stresses on core components.
A very important aspect of nuclear reactor core analysis involves the determina-
tion of the coolant flow distribution and pressure drop across the core. Since the
flow in a reactor core is highly turbulent, we are forced to employ once again
various empirical correlations in order to analyze the coolant hydrodynamic
behavior.
A. Single-Phase Coolant Pressure Drops

One can identify several mechanisms that will cause a pressure drop along the
coolant channel: (a) friction losses from the fuel rod bundle (channel friction), (b)
friction losses from the spacer grids used to maintain fuel assembly geometry, (c)
friction losses at the core inlet and exit flow plenum (contraction or expansion),
and (d) elevation. We will study each of these components in turn.
Consider first the pressure drop due to channel friction. Here we can use the
result Eq. (12-30) involving the friction factor
~Plrriction = 2( ih )p 2f, Uz (12-70)
where L is the total channel length and Dh is its hydraulic diameter. In most cases
one can calculate fusing the Blasius formula, Eq. (12-33).
Another source of pressure drop important in most reactor types is caused by the
spacer grids that bind the fuel elements into a fuel bundle (see Figure 3-7). There is
an additional pressure loss associated with the coolant inlet and exit. Friction losses
due to spacer grids or inlet or exit geometry are classified as form losses, since they
correspond to a change in the coolant momentum due to a change in the coolant
channel geometry. One usually writes
(12-71)
where K; is a form friction factor due to a change in the coolant channel cross-
sectional area of the ith type. K; is quite complicated for spacer grids and must be
determined experimentally for each spacer grid design. The typical KToTAL range is
0.6-1.2.
The final component is the pressure drop due to coolant elevation as it passes up
the core
H/2
~Pl elevation= J
-H/2
P( Z) g dz. (12-72)
For purposes of illustration, we have tabulated typical values of each of these

pressure drop components for several reactor types in Table 12-5. We have also
included in this table the pressure drop across the entire pressure vessel (which
essentially determines the pumping power requirements in the primary loop).
TABLE 12-5: Typical core pressure drops
Reactor
Type ilprriction /1pform /1pgravity ti.pace 11Pcore /1pvessel
PWR .40 bar .61 .16 1.45 3.40

BWR .40 .61 .16 .05 1.49 3.40
HTGR .134 .03 .16
LMFBR 5.4 .81 .30 6.51 9.5
B. Boiling Channel Pressure Drops

The pressure drop occurring in a boiling channel is somewhat different
because of the phenomena involved in two phase flow. There is first an additional
pressure drop due to the acceleration which accompanies the coolant-vapor forma-
THERMAL- HYDRAULIC ANALYSIS OF NUCLEAR REACTOR CORES / 501
tion and expansion 1•3
D.p lace= Pt Uz 2
[
(I -x)2
I
-a
x2 P
+ -a ~
p
-
8
l
1 · (12-73)
The hydrostatic pressure drop must also be computed somewhat differently as
H/2
D.p Ielevation=
J -H/2
[ ap8 + ( I - O'. )Pi] g dz• (12-74)
Finally for two-phase flow a somewhat different friction and form pressure drop
must be used. It is customary to modify the friction pressure drop by inserting a
parameter RMN to write
(12-75)
where RMN is known as the Martinelli-Nelson friction multiplier, 20 which is a

tabulated function of pressure and flow quality. In a BWR, RMN ranges from 1.0
(x=0) to as high as 4 (x=20%).
The form losses can be written as
/j. I
.P2-<1>rorm
= K p -Uz
g
2
2[ (I - X)2
I -a +
!.._2 !!!_
a Pg
l
, ( 12-76)
where Kg is the appropriate form factor for two-phase flow.

It should be stressed that there is a large degree of uncertainty in these methods
used for estimating coolant hydrodynamic behavior. The predictions of such
analyses must usually be confirmed by experimental testing before they are applied
in making detailed core design decisions.
C. Multichannel Analysis
Our earlier single-channel analysis has not accounted for phenomena such as
flow mixing between channels, which is caused by both turbulence and pressure
gradients. Such mixing can lead to energy transfer between channels as well as
pressure losses.
One could characterize coupling between channels by writing a generalized
energy balance equation for a channel m
~ ~ ~
(axial mass (cross flow (mixing ( 12-77)

flow rate) rate/length) coefficient)
Here the cross channel coupling coefficients wn---->m and E must be evaluated from
empirical correlations based on experiment.
One can also have flow redistribution effects in the inlet plenum causing a
nonuniform inlet velocity to each channel. Such flow redistribution could, in fact,
lead to instabilities, and must be accounted for in more detailed core hydraulic
analysis. 21
VII. THERMAL-HYDRAULIC CORE ANALYSIS
A. Hot Channel Factors

One of the principal goals of the reactor designer is to ensure that none of the
thermal limitations on the core behavior are exceeded. Thus far we have discussed
two such limitations. First one must design the core so that the fuel centerline
temperature does not exceed the fuel melting point. This limitation is usually
expressed as a restriction on the linear power density
q'(r) < q;,,ax[ ~660 W / cm in U02 fuel elements J. (12-78)
Yet another limitation arises from requiring that the clad surface heat flux always
remains below its DNB limit:
q"(r) < qDNB[ ~300 W /cm 2 in LWRs ]. (12-79)
There are usually other thermal limitations placed on core performance. For
example, thermal and fission gas stresses on the clad can limit power generation.
Furthermore, thermal-hydraulic performance is also frequently limited by nuclear
stability considerations (particularly in BWRs that experience an appreciable
variation in moderator density throughout the core).
Certainly one approach to thermal-hydraulic analysis of the core would be to first
perform a detailed three-dimensional calculation of the core power distribution,
taking into account the effects of fuel burnup, fission product buildup, control
distributions, and moderator density variations over core life. This information
could then be used to determine the coolant flow and temperature distribution
throughout the core. However, such a calculation is usually prohibitively expensive
for routine design applications.
A more common approach is to investigate how closely the "hot channel" in the
core approaches the operating limitations. Then if one can ensure that the thermal
conditions of this channel remain below the core limitations, the remaining
channels in the core will presumably fall within design limitations. One usually
defines the hot channel in the core as that coolant channel in which the core heat
flux and enthalpy rise is maximum. Associated with this channel are various "hot
channel" or "hot spot" factors relating the performance of this channel to the
average behavior of the core.
More precisely, we will define the hot assembly in a reactor core as that fuel
assembly having the maximum power output. The hot spot in the core is the point
of maximum heat flux or linear power density, while the hot channel is defined to
be that coolant channel in which the hot spot occurs, or along which the maximum
coolant enthalpy increase occurs.
We first will introduce the concept of nuclear hot channel factors, which take
into account the variation of the neutron flux and fuel distribution within the core.
First, we define the radial nuclear hot channel factor
N average heat flux of the hot channel

F =---------------
R average heat flux of the channels in core
H/2
J q"(rttc)dz
=--------
-H/2 for I channels
(12-80)
/ (pins).
1 "'JH/2 q"(r;)dz
-"-'
1 i=l -H/2
One can similarly define the axial nuclear hot channel factor:
pN _ maximum heat flux of hot channel

2
- average heat flux of hot channel
m:,xq"(rttc)
=-------- (12-8 I)
1 JH/2 II
H q (rttc)dz
-H/2
The total nuclear hot channel factor or nuclear heat flux factor is then just
pN = maximum heat flux in the core = pNpN ( 12-82)

q average heat flux in the core R z·
By way of example, consider a homogeneous, bare cylindrical core described by

a power distribution given by Eq. (12-1). We can then calculate
1 0 (0) f 2
H/ cos(-;;
-H/2
)dz
F%= =2.32, (12-83)
(R1 ( 2•405r)
},, 0 R 27Trdr f H/2
cos ( 'lTZ)
H dz
0 -H/2
and
(12-84)
This implies an overall nuclear hot-channel factor of
F:' = (2.32)(1.57) = 3.638. (12-85)
This is actually quite conservative. A zon:e-loaded PWR22 will typically have a

nuclear hot channel factor of = 2.6. F:'
The nuclear heat flux hot channel factor is defined assuming nominal fuel pellet
and rod parameters. However, there will be local variations in enrichment, fuel-
pellet density and diameter, surface area of fuel rod and eccentricity of the
fuel-clad gap, and so on due to manufacturing tolerances. The more general heat
flux hot channel factor or total power peaking factor Fq is defined as the maximum
heat flux in the hot channel divided by the average heat flux in the core, allowing
for these manufacturing tolerances. 1- 3 One relates Fq to F;' by defining an
engineering heat flux hot-channel factor F/
F
FE= _ q (12-86)
q FN
q
that specifically accounts for manufacturing tolerances. Typically FqE is close to

unity (in a modern PWR, FqE ~ 1.03).
One can next define an enthalpy-rise hot channel factor
maximum coolant enthalpy rise

Fa = - - - - - - - - - (12-87)
H - average coolant enthalpy rise ·
This factor is a function of both variations in the power distribution and coolant
flow. For example, some 3-10% of the coolant flow bypasses the fuel assemblies,
due to leaks past mechanical seals or the presence of other core components. 1- 3
There will be nonuniformities in the coolant flow from channel to channel,
although these are reduced somewhat by the lower inlet plenum. Flow mixing
between coolant channels can also occur, as well as flow redistribution in which
pressure variations due to coolant expansion or variations in nucleate boiling cause
flow nonuniformities.
By way of example, we can calculate the enthalpy hot channel factor taking into
account only the power variation in a bare, homogeneous cylindrical core:
llh(r) = l_f
W
H/2 q'(r,z)dz=
-H/2
2HA
7TW
lo( 2.405r
R
), (12-88)
or
-6.h =- 1
7TR2
lR
0
HA
6.h(r)27Trdr= .862-
7TW
or
6.hlmax
Fl!.H = -_- = 2.32. (12-89)
6.h
The third hot channel factor occasionally introduced is that characterizing the
clad surface temperature
(12-90)
Actually this factor is usually of little interest in LWRs in which nucleate boiling
keeps the clad surface temperature quite close to the coolant temperature. How-
ever, it can be of more significance in HTGRs and LMFBRs in which the
temperature jump from the coolant channel surface to the coolant is much larger.
The concept of such hot channel factors greatly simplifies thermal core analysis,
since it provides a mechanism for accounting for nonuniformities in a crude
THERMAL- HYDRAULIC ANALYSIS OF NUCLEAR REACTOR CORES / 505
single-channel analysis of the reactor core. However the use of hot channel factors
has only a limited validity, and will fail when a more precise knowledge of core
temperature and flow distributions is required. Indeed one might almost regard the
popularity of such an approach as evidence of the crude state of predictive theories
of thermal-hydraulic core behavior that must depend on numerous empirical
correlations and experimental measurements to achieve adequate results. (In com-
parison to methods used in thermal-hydraulic core analysis, the reader would
probably agree that the methods we have developed for the nuclear analysis of a
reactor core look very rigorous and elegant indeed!)
B. Determination of Reactor Core Size

It is instructive to use some of these ideas to examine how the thermal design
influences the determination of the reactor core size. Let us break this discussion
into several steps:
l One first determines the various hot channel factors characterizing the
core, using both computer codes describing the core neutronics and ther-
mal-hydraulic behavior, as well as experience from earlier designs.
2 Next one determines the maximum acceptable power level in the hot
channel, taking into account both steady-state and transient operating
conditions. This is usually expressed in terms of a limit on linear power
density
(12-91)
Hence the average linear power density is given in terms of this limiting
value and the nuclear hot channel factor F:'-
3 The fuel-to-moderator volume ratio in the core and the fuel element radius
'FE are largely determined by nuclear considerations. For a square lattice
(Figure 12-7) one notes that
P 2-w,2FE
=---- (12-92)
w,2
F
Here of course the fuel-element radius must take into account the gap and
clad thickness
(12-93)
One usually specifies the lattice geometry in terms of the pitch to diameter
ratio (p /2rFJ· Typically this ranges between 1.2-1.5 for both LWRs and
LMFBRs.
4 The thermal power output of the core is given by
(12-94)
where N R is the total number of fuel rods in the core, L is the active fuel
rod length, and C 1 is the fraction of the core power generated in the fuel
( ~95-97%). If one desires a fuel-assembly design with NA rods per as-

sembly (~200), then the number of assemblies required by the core
thermal limitations is just
MWtC 1F:'
# assemblies= - - - - - (12-95)
q~AX LNA •
In a similar manner, if we note that the average core power density

q'J..'vE=q~vE/p 2, we can compute the required core diameter Dcore by
equating
2
MW = ,,, 1rD core
2 ) L= q'AVE ( 1r D core ) L
t qAVE ( 4 2 4 . (12-96)
p
Hence the required core diameter is just
2 112
= [ 4p MWt F:' ] (12-97)
Dcore , ·
1rqMAX
Of course there are other thermal-hydraulic constraints on core design, such as

the minimum allowable DNBR and maximum core pressure drop which we have
not included in this simple analysis. It should be noted that in this analysis the
allowable power densities and hence the core size are limited by the hot channel
factor. It is obviously to one's advantage to obtain a F:'
as close to unity as
possible. As we will see later, the power profile can be flattened considerably using
zonal fuel loading and control element insertion.
It should be noted that hot channel factors such as F:' and Ft>.H that account for
nonuniformities in the core power profile and flow distributions can be augmented
by engineering hot channel factors, FqE' which account for variations in flow and
power due to effects such as component fabrication tolerances, fuel swelling and
cracking during burnup, as well as uncertainties in thermal-hydraulic nuclear
analysis. In a sense, these hot channel factors can be regarded as representing the
conservatism in design.
A very major improvement in core thermal performance can be achieved by a
more accurate estimate of these factors so as to eliminate excessive design con-
servatism. For example, the enthalpy rise hot channel factor Ft>.H was revised
downward from 3.36 to 1.7 during PWR development during the 1960s based on
both operating and design experience. 1 This allowed a considerable improvement
in thermal performance. In a similar sense, the thermal design of advanced reactor
cores such as the LMFBR tend to use considerably more conservative estimates of
hot channel factors, and their designed performance usually falls well within
thermal constraints.
We should mention at this point that the thermal-hydraulic analysis of the core is
actually quite closely coupled to the nuclear analysis. For example, the neutron
flux and hence power distribution over core 'life is necessary to calculate the
nuclear hot channel factors. Furthermore the core nuclear design usually provides
the moderator-to-fuel ratio and fuel pin diameter required by the thermal analysis.
We have also noted that the thermal - hydraulic analysis must supply core
densities and temperatures back to the nuclear analysis. For example, the coolant
void distribution will strongly affect the power profile. Furthermore changes in
core temperatures will cause reactivity feedback (e.g., the change in fuel tempera-
ture that leads to reactivity changes via the Doppler effect). Needless to say, such
"neutronic-thermal-hydraulic" coupling is a very important facet of nuclear reac-
tor analysis, and we will return to consider it in Chapter 13.
C. Thermal-Hydraulic Design Codes

A variety of computer codes have been developed to analyze the thermal-
hydraulic behavior of nuclear reactor cores. 23 In order to illustrate such models, we
will consider only a very simple scheme designed to perform a single channel
thermal-hydraulic analysis for the hot or average channel of a light water
moderated core. In such a model, the hydrodynamic and thermodynamic equations
describing one-dimensional axial flow are discretized and solved explicitly step-
wise up the coolant channel for an arbitrary axial power profile. The model we will
consider accounts for three flow regimes: (a) single-phase flow, (b) subcooled
boiling two-phase flow, and (c) bulk boiling two-phase flow. Such a model is
capable of predicting axial coolant, clad, and fuel temperature profiles, void
fraction, pressure drop, the critical heat flux qf;N 8, and the DNB ratio for the core.
Using the nuclear hot channel factors, one can also calculate the average thermal
performance of the core.
We will divide the coolant channel into segments of length !:i.z; as shown in
Figure 12-16. We will then consider the relevant equations describing each of the
flow regions.
1. SINGLE-PHASE FLOW (hr< hr(T,aJ or z < z 5 c 8 ):
If we discretize the energy balance equation (12-43) for a given channel

segment, we find
(12-98)
Hence we can solve for the enthalpy at the point z;+ 1 in terms of that at z; as
(12-99)
Remember that the power profile q; is supplied as input to the model (perhaps from
a nuclear criticality calculation).
Next one calculates the pressure drop up the segment as the sum of contributions
Pi+I =p; - !:,.p{ !:i.p? (12-100)

.._.,,...., .._.,,....,
friction hydrostatic
Zsce
FIGURE 12-16. Discretization of coolant channel.
If we recall our earlier hydrodynamic analysis, it is apparent that
(12-101)
b.pform= ~ K;pui/2. ( 12-102)

i
Once we have determined the coolant enthalpy and pressure at zi+ 1, we can go to
thermodynamic tables (stored in computer memory or tape) to find the correspond-
ing coolant density and temperature
(12-103)
We can use these to compute quantities such as Re;+ 1,Nu;+i,Pr;+i· Next, we work
inward to determine the fuel element temperatures at z; + 1• For instance, the clad
surface temperature is
(12-104)
where hs;+, (Nu;+" Rei+ 1) is usually determined from the Dittus-Boelter correlation
(12-38) for single-phase flow. The fuel centerline temperature is then
(12-105)
Finally the fluid velocity is determined by using mass flow conservation
( 12-106)
to find
(12-107)
2. TWO-PHASE FLOW (h > hr(TsaJ or z>zscB)

The energy balance relation giving the enthalpy at each point in the channel
remains unchanged. The pressure drop is modified to account for two-phase flow
effects. In particular, the Martinelli-Nelson correlation is used for fl.pf
( 12-108)
where RMN; is a tabulated function of P; and X;· We also add in
( 12-109)
where
(12-110)
This will allow us to determine h; + 1 and P; + 1• Then using thermodynamic tables,

we can generate
h;+J} ==;> hf;+i'hg;+i'Xi+i' Ti+I = TsatCP;+1) (12-111)

Pi+I Pt;+1'Pg;+1·
To determine the clad wall temperature, the Jens-Lottes correlation is used
(12-112)
and then the fuel centerline temperature can be determined using Eq. (12-105)
If the slip ratio were known, one could then use the S - a - x relationship to
determine the void fraction. However it is common practice instead to use a
correlation (again due to Martinelli and Nelson) to determine a;+ 1 as a function of
X; + 1 and P; + 1• Then the S - a - x relationship is used to determine the slip ratio
(12-113)
If we note that
wg= VgP~g= VgPgaAc= XW, (12-114)
we can calculate
(12-115)
and then use the slip ratio to find
V1i+I = V g;+t /S.1+ l' (12-116)
Finally, one can use these values of the flow parameters to compute the critical
heat flux qf'>NB at each point in the channel using an appropriate empirical
correlation. The DNBR can similarly be calculated since the linear power density
q' and hence the heat flux q" is known.
Thus the strategy for analyzing the channel is very simple. One merely takes the
coolant inlet conditions and the power profile q'(z) and marches up the channel,
calculating the relevant hydrodynamic and thermodynamic parameters for the next
channel mesh-point using conservation equations and tabulated thermodynamic
data and correlations. The actual mechanics of such single-channel codes is quite
simple (although thermodynamic data parameterization and storage may be more
complicated).
D. Some Additional Comments

Our illustration of thermal-hydraulic analysis in the previous section was
specifically concerned with water-cooled reactors. However this model could also
be used to analyze LMFBRs with only a minor modification in the correlations
used to generate the coolant properties, since the fuel element lattice structure of a
LMFBR is remarkably similar to that of a LWR. Indeed the thermal-hydraulic
analysis of a liquid metal-cooled core is even a bit simpler since the coolant
temperature is always kept well below the saturation point to permit only single-
phase flow to occur. Since LMFBR fuel elements are frequently fabricated with a
sodium bond between the fuel and the clad, a slightly different gap conductance
would be used.
A very similar type of analysis would apply to the gas-cooled fast reactor
(GCFR) that uses helium to cool a fuel lattice almost identical to that of the
LMFBR. Again, the coolant is single phase (gas). However there is a slight
complication since high-speed gas flow can be accompanied by compressible flow
phenomena (flow choking, shock waves, etc.). At the level of our simple single-
channel analysis, these effects can frequently be accommodated by a minor
adjustment of correlations. However in more elaborate calculations (e.g., system
depressurization), such effects would require a different type of analysis.
The HTGR can also be described by a similar single-channel analysis (with the
appropriate adjustments for compressible flow), with one major difference. In the
LWR, LMFBR, and GCFR, the coolant flows about the cylindrical fuel elements
in a bundle. However the coolant in an HTGR flows through channels in a
graphite block containing the fuel (see Figure 10-4). Although the description of
axial flow through a circular channel is simple enough, the analysis of the radial
heat conduction from the fuel pins through the graphite block to the coolant
channel wall is a complicated two-dimensional heat conduction problem. 24 Hence
the calculation of the temperature distribution in HTGR fuel assemblies usually
requires a direct two-dimensional numerical solution of the thermal conduction
equation (a task beyond our present patience in this book, although it is obviously
very similar to a two-dimensional neutron diffusion calculation).
REFERENCES
j"l. L. S. Tong and J. Weisman, Thermal Analysis of Pressurized Water Reactors, American
Nuclear Society, Hinsdale, Ill. (1970); L. S. Tong, Nucl. Eng. Design 6, 301 (1967).
2. M. M. El-Wakil, Nuclear Heat Transport, Intext, Scranton (1971).
3. H. W. Graves, Jr., Nuclear Reactor Design, University of Michigan Lecture Notes
(unpublished), 1969.
4. A. Sesonske, Nuclear Power Plant Design Analysis, USAEC TID-26241 (1973).
5. H. Fenech and W. M. Rohsenow, Heat transfer, in The Technology of Nuclear Reactor
Safety, T. J. Thompson and J. G. Beckerley (Eds.), M. I. T. Press, Cambridge (1973),
Vol. II.
6. J. A. Christensen, R. J. Allio, and A. Biancheria, WCAP-6065 (1965).
7. Reference Safety Analysis Report (RESAR-3), Vol. II, Part 4.4, Westinghouse Nuclear
Energy Systems (1973).
8. W. M. Rohsenow and H. Choi, Heat, Mass, and Momentum Transfer, Prentice-Hall,
Englewood Cliffs, N.J. (1961).
9. F. Kreith, Principles of Heat Transfer, 3rd Ed., Intext, Scranton (1973).
10. R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena, Wiley, New York
(1960).
11. L. S. Tong and J. Weisman, Thermal Analysis of Pressurized Water Reactors, American
12. See references contained in L. S. Tong and J. Weisman, Thermal Analysis of Pressurized
Water Reactors, American Nuclear Society, Hinsdale, Ill. (1970); J. Weisman and R. W.
Bowring, Nucl. Sci. Eng. 57, 255 (1975).
13. R. A. Dean, Thermal Contact Conductance, M. S. Thesis, University of Pittsburgh
(1963).
14. L. D. Landau and E. M. Lifshitz, Fluid Mechanics, Pergamon, New York (1959).
15. M. M. El-Wakil, Nuclear Heat Transport, Intext, Scranton (1971), p. 299.
/ 16. W. H. Jens and P. A. Lottes, Analysis of heat transfer, burnout, pressure drop, and
density data for high pressure water, USAEC Report ANL-4627 (1961).
17. J. H. Keenan and F. G. Keyes, Thermodynamic Properties of Steam, Wiley, New York
(1937).
18. E. Janssen and S. Levy, Burnout limit curves for boiling water reactors, General Electric
Company Report APED 3892 (1962).
/ 19. J. Weisman and R. W. Bowring, Nucl. Sci. Eng. 57, 255 (1975).
20. R. C. Martinelli and D. B. Nelson, Prediction of pressure drops during forced circula-
tion boiling of water, Trans. ASME 70 (1948).
21. Mathematical Models and Computational Techniques for Analysis of Nuclear Systems,
USAEC Document CONF-730414-P2, Vol. I, Session III (1973).
22. L. S. Tong and J. Weisman, Thermal Analysis of Pressurized Water Reactors, American
23. D. S. Rowe, COBRA-II: A Digital Computer Program for Thermal-Hydraulic Sub-
channel Analysis of Rod-Bundle Nuclear Fuel Elements, BNWL-1229 (1970); BNWL-
1695 (1973); J. Weisman and R. W. Bowring, Nucl. Sci. Eng. 57, 255 (1975).
24. Preliminary Safety Analysis Report, Summit Power Station, Delmarva Power, Vol. 3,
Chapter 4 (1973), Section 4.4.2.2.
PROBLEMS
12-1 Compare the average thermal power densities ( q', q", q'") of each of the major
reactor types (e.g., PWR, BWR, HTGR, LMFBR, GCFR, ... ) along with that of the
boiler of a fossil-fuel-fired plant. Use data from PSAR's for actual plants, if possible.
12-2 Derive the equation of heat conduction by considering an energy balance for an
arbitrary volume (in analogy to the derivation of the neutron diffusion equation).
12-3 Determine the temperature profile in plate-type fuel elements composed of fuel of
thickness 2rF sandwiched between a clad of thickness le, Assume a gap thickness tG.
12-4 Derive an expression for the average fuel temperature in a cylindrical fuel pellet as a
function of the surface and centerline temperature. Assume a constant heat source
and thermal conductivity.
12-5 Determine the temperature distribution in an infinitely long cylinder, assuming an
internal source given by q"'(r) = q0 J0(r / L), where the surface of the cylinder is at
temperature T •. Assume that the thermal conductivity depends on temperature
through the relationship k( T) = k 0[ l + a(T- T.)]. Sketch the temperature profile for
the constant a positive, zero, and negative.
12-6 The flux depression in a cylindrical fuel pellet can be modeled by assuming that the
radial dependence of q"'(r) is of the form q'"(r) = qii"(l + ar 2). Derive an expression
for the temperature rise in the fuel in terms of the linear power density q0. In
particular compare this temperature drop with that resulting from a uniform heat
flux for a 12% flux depression (i.e.,ari = 0.12) for the case in which q0=500 W / cm
and kF= .0245 W /cm °K.
12-7 Calculate the equivalent heat transfer coefficient, hG, characterizing radial gas gaps
between pellet and clad of 0.010 and 0.005 cm. Assume rF»tG and constant gas
properties with temperature for both helium and fission product gas.
12-8 Consider two cylindrical fuel rods of radii a and 2a respectively which are to operate
under the limitation that the maximum center-to-outer surface temperature
difference is ~ T. Assuming uniform heat generation in the rods, which rod would be
able to supply more heat?
12-9 It is frequently of interest to determine the temperature distribution in shielding
material being heated by incident radiation (e.g., photons or neutrons). Such a
calculation can easily be performed in analogy to our study of thermal conduction in
fuel elements, provided one uses a distributed heat source. Consider a slab shield of
thickness L with a radiation flux </>o incident upon one face. If the radiation intensity
is assumed to be attenuated as <[>0 exp(- ~ax), determine the temperature distribution
across the shield. Assume that we maintain the surfaces at x = 0 and x = L at
specified temperatures T 0 and TL.
12-10 A 8 cm-thick flat-plate iron shield is subjected at one face to 3.0 MeV /photon y
radiation of uniform 10 13 flux. The absorption cross section for 3 MeV ys in iron is
0.282 cm - 1• If the temperatures of both sides of the plate were to remain equal, find:
(a) the temperature distribution in the plate, (b) the difference between maximum
and surface temperatures, and (c) the surface cooling on each side of the plate
necessary to maintain the above conditions.
12-11 Using the fuel element data for the BWR/6 core design given in Appendix H,
estimate the fuel centerline temperature that occurs at the point of maximum heat
flux if it is assumed that the coolant temperature at this point is 290°C with a
convective heat transfer coefficient of h8 = 5.0 W / cm2 °K.
12-12 Repeat the calculation of Problem 12-11 for the LMFBR core data assuming a
sodium temperature of 500 °C and h.=4.0 W /cm2 °K.
12-13 Consider water at I bar, 20°C flowing through a smooth, circular pipe of 2 cm I.D.
and length 3m. If the flow velocity is 3 m/sec, determine the pressure drop along the
pipe and the pumping power necessary to maintain the flow in this pipe.
12-14 Compute the equivalent hydraulic diameter Dh for a hexagonal lattice such as that
sketched in Fig. 12-7.
12-15 A cylindrical LMFBR core contains N fuel elements of length L. Sodium coolant
flows through the core in the axial direction with a total mass flow rate w that is
uniformly divided among the coolant channels in the core. It is found that this
reactor core has a positive sodium void coefficient of reactivity. To correct this, a
new design is proposed in which the cross-sectional area of the cylindrical core is
doubled (i.e., twice as many fuel elements), while the core length is halved, thereby
increasing neutron leakage and achieving a negative void coefficient. If the total
coolant flow rate does not change in the new core design, estimate the pumping
power required relative to that for the original core. (Assume turbulent flow.)
12-16 A reactor coolant loop may be hydrodynamically unstable if operated with a portion
of the loop in the transition region between laminar and turbulent flow. Because the
flow resistance changes when transition occurs from laminar to turbulent flow (and
vice versa), conditions are favorable for flow oscillation. Consider a loop consisting
of constant diameter piping and a constant pressure pump (t:..p across pump
independent of flow rate).
(a) Describe how this type of instability can occur in such a simplified loop.
(b) If the transitions take place at a Reynolds number of approximately 2500,
determine the ratio of the maximum to minimum flow rates in the pipe (i.e.,
Wmax/wmiJ·
12-17 Compare the heat transfer coefficients and the pumping power per 100 m length of 2
cm I.D. smooth-drawn tubing for the following coolants: (a) air at 10 bar, 100
m/sec, and 200°C, (b) He at the same conditions, (c) water at 10 m/sec, 200°C, 1
bar. and (d) sodium at the same conditions.
12-18 A reactor is cooled by a gas that flows turbulently through its core. Its temperature
rise is small compared to its absolute temperature, and the pressure drop is small
compared to the absolute pressure. Suppose the coolant pressure is now doubled.
The reactor is to require no more pumping power than before. The temperature rise
of the coolant is to be held constant. No core modifications are to be made.
Determine: (a) the maximum possible increase in reactor power and (b) assuming
that this power increase is made, the extent of temperature rise across the gas film as
the heat transfer increases.
12-19 Determine the axial clad surface and fuel centerline temperature distribution present
in a bare, uniform cylindrical core. In particular determine those positions at which
the maximum fuel centerline and clad surface temperatures occur.
12-20 As we have seen, one of the primary limitations on thermal performance is the
avoidance of fuel centerline melting. However it is obviously to our advantage to run
the fuel at all points in the reactor as close to this limit as possible (consistent with
safety requirements).
(a) Determine the axial power profile that would yield a uniform axial fuel tempera-
ture in the core.
(b) Determine the axial fuel loading distribution that would yield this power profile
(use one-speed diffusion theory).
(c) Compare the feasibility of obtaining such profiles for a HTGR and a PWR.
12-21 Consider a channel of length L in which there is a bottom-peaked power profile of
the form
q'(z) = q/i{1r / ii)(ii- z) sin[('1T / ii )(ii- z)].
Determine: (a) the maximum/average heat flux in the channel, (b) q'(z) in terms of
the average channel heat flux, and (c) the enthalpy rise at any position in the
channel.
12-22 Consider an LMFBR core with design parameters as given in Appendix H. If the
linear power density in the core is assumed to be a uniform 300 W / cm up the core
and the flow area per fuel rod is 1 cm2 , calculate: (a) the flow rate per rod, (b) the
coolant transit time through the core, and (c) the core power density.
12-23 Water flows in a circular channel 3 m long by 1 cm I.D. The system pressure is 70
bar. The water enters the channel with average velocity 3 m/sec and temperature
260°C. The wall heat flux has a cosine distribution (no extrapolation lengths) with a
peak value of 200 W /cm2 • Determine: (a) axial coolant temperature distribution, (b)
axial quality distribution, (c) axial void fraction distribution, assuming a constant
slip ratio of 2, (d) acceleration pressure drop in the channel, (e) friction pressure
drop in the channel, and (f) graph T 0 , X, a, and p versus z. Assume thermodynamic
equilibrium exists at each axial position (i.e., neglect subcooled boiling).
12-24 For Problem 12-23 determine the axial position where subcooled boiling starts. What
is the coolant temperature at this position? The wall temperature?
12-25 Consider a PWR operating in such a fashion that nucleate boiling occurs at the point
of maximum power density. How might you expect the following changes to affect
both the maximum fuel centerline temperature and the DNBR, assuming a constant
average linear power density: (a) 10% reduction of coolant flow in the hot channel,
(b) modification of the core analysis to account for nonuniform power generation in
the fuel pellets, (c) radial fuel migration and densification producing a 4% void
volume along the fuel pellet axis (d) the axial core power profile shifting from a
centered to a bottom-peaked distribution (such that both the peak power and
integrated power remain the same), and (e) the axial core power profile shifting to a
top-peaked distribution. (See Tong and Weisman 1 for typical DNB correlations.)
12-26 The outlet quality of an average coolant channel of a BWR is 14.6%. The system
pressure is 70 bar, and the inlet temperature is 220°C [h(220°C, 70 bar)= 1.23
MJ/kg]. Determine: (a) the exit void fraction for a slip ratio of 2.0, (b) the average
heat generation (W / cm) of the fuel rods if the mass flow rate per fuel rod is 10,000
kg/hr and the core height is 375 cm, and (c) the nonboiling length for the
bottom-peaked power profile of Problem 12-21.
12-27 Using the description provided in Section 12-VII-C, write a computer program
capable of analyzing the hot channel of a LWR (either PWR or BWR).
The following problems will require the use of a simple single-channel thermal -
hydraulics code suitable for a LWR.
12-28 Use the thermal-hydraulics code to evaluate the performance of any one of the
PWR designs given in Appendix H. In particular: (a) compare a bottom-peaked and
centered power profile from the standpoint of: (i) peak linear power density, (ii) peak
fuel centerline temperature, and (iii) DNB ratio; (b) roughly plot the following for
the hot channel of the bottom peaked profile: (i) coolant temperature, (ii) clad
surface temperature, (iii) heat flux, and (iv) DNB heat flux; and (c) evaluate the
effect of a simultaneous 10% reduction in flow and 10% increase in power level on
DNB ratio and centerline fuel temperature.
12-29 Use the thermal-hydraulics code to evaluate the performance of the BWR-6 design
in Appendix H. Use the bottom-peaked power profile and determine: (a) coolant
temperature, fuel temperature, void fraction, and flow quality profiles; and (b) the
recirculation ratio (mass ratio of recirculated coolant to coolant leaving the vessel as
steam) and the reactor vessel inlet temperature (i.e., feedwater temperature).
13
The Calculation
of
Core Power Distributions
The central component of most computational models of nuclear reactors

consists of a group of calculational modules that analyze the static neutronic
behavior in the reactor core. As we have repeatedly emphasized, these static
calculations are usually based on a solution of the multigroup diffusion equations
for the multiplication eigenvalue keff and the multigroup fluxes 'f>g(r) characterizing
a given core configuration. One can then construct the corresponding core power
distribution by calculating the local fission rate as in Eq. ( 12-1 ).
Such static calculations not only involve what we have referred to as a flux-
power-reactivity module that solves the multigroup diffusion equations to generate
the core flux and power, but also a module that generates the necessary macro-
scopic group constants and a thermal-hydraulics module analyzing heat transfer
and fluid flow in the reactor core to determine core temperatures and coolant
densities. Hence our consideration of the calculation of core power distributions
cannot be confined to the flux-power module alone, but must instead consider this
coupled group of three calculational components (see Figure 13-1).
We are of course already very familiar with most of the features of these
components. For example, Chapters 8, 9, and 10 dealt with the generation of
multigroup constants. In Chapters 5 and 7 we discussed in some detail various
methods for solving the time-independent multigroup diffusion equations. In
Chapter 12 we considered the various features of models of the thermal-hydraulic
behavior of a nuclear reactor core. Our goal in this chapter will be somewhat
different (and considerably briefer) since we now wish to examine several of the
more practical details of the calculation of core power distributions and explore the
various ways in which such calculations interact with other aspects of the core
design.
515
Macroscopic group constant generation

...,...._
/or~
Parameterized Effective
Fast and thermal group-constant
spectrum codes control cross
tables sections
',
I Flux-power-reactivity module
I
k.11, rt,(r), P(r)
,,
TM, Tf, NM
Thermalhydraulics module
I I
FIGURE 13-1. Modules involved in the calculation of core power distributions.
It is certainly not surprising that the calculation of core multiplication plays a

very important role in nuclear reactor analysis. Of course such calculations are
necessary in order to determine the fuel loading or control system adjustment
necessary to achieve reactor criticality. However the calculation of the multiplica-
tion eigenvalue keff for a sequence of different core configurations can also
determine quantities such as control element worths, reactivity feedback
coefficients (e.g., by calculating keff for the core at various different temperatures),
and reactivity changes accompanying various different fuel loading patterns.
The static reactor flux and power distributions are also of considerable impor-
tance in reactor design. In particular, we have found that the power distribution is
essential for the subsequent thermal-hydraulic analysis of the core. For example,
the nuclear hot channel factors and axial power profile determine how closely the
core performance approaches thermal design limitations. Furthermore the neutron
flux distribution is necessary for the determination of fuel burnup and isotope
buildup. However since the thermal-hydraulic core behavior as well as its fuel
depletion will affect both core composition and microscopic cross sections (i.e.,
through temperature effects), the coupling between these various components of a
reactor model becomes very important. For example, the reactor designer would
like to utilize the information provided by such core analysis to achieve a flat
power profile (desirable from both a thermal as well as a fuel burnup viewpoint)
while providing sufficient excess reactivity to yield long core lifetimes and hence
high fuel exposures, but within the safety constraints imposed on core thermal and
nuclear (i.e., control) performance.
I. STATIC MULTIGROUP DIFFUSION CALCULATIONS
Perhaps the central problem in reactor analysis is the determination of the

spatial flux and power distribution in a reactor core under steady-state operating
conditions. There is, however, a considerable variation in the degree of accuracy
and spatial detail of the power distribution required in different facets of reactor
analysis and design. For example, in determining the optimum core thermal
THE CALCULATION OF CORE POWER DISTRIBUTIONS / S17
performance one will require as accurate an estimate of the power distribution as

possible in order to ensure that thermal constraints are not exceeded. In· contrast,
considerably less detail is required in the comparison of various fuel loading
schemes involved in fuel cycle management. These two examples illustrate the
contrast between absolute estimates of core power distributions required for
ensuring that performance limitations are not exceeded, and relative estimates
required in optimization or tradeoff studies.
The calculation of the gl9bal power distribution is most commonly accomplished
by solving the few-group diffusion equations
g-1 l G
- V· Dg (r)V</>g(r) + LR,(r)</>g(r) = L L••·• </>g,(r) + k Xg L rg,Lr,.</>g-(r). ( 13-1)
g'=l g'=I
These equations are solved by using finite-difference methods to discretize the

spatial variable and then, following the usual inner-outer iteration strategy, solving
for the criticality eigenvalue keff and the corresponding multigroup flux <1>g(r). To be
more precise, the reactor core is broken up into a spatial grid or mesh-for
example, with M mesh cells. Then the multigroup diffusion equation is integrated
over a typical cell, and standard sum and difference formulas are used to represent
the terms in the equation. Hence the multigroup diffusion equations are replaced
by an M X G set of algebraic equations-that is, an (M X G) dimension matrix-
eigenvalue problem
(13-2)
The elements of the matrices M and f are the few-group constants supplied by a
macroscopic cross section module. The matrix eigenvalue problem can then be
solved by standard power-iteration methods (usually accelerated by source ex-
trapolation).
The complexity of such calculations depends sensitively on the nature of the
design information required. During the early stages of the design process it is
common to use very fast running one-dimensional survey codes. For core lifetime
studies in which the power and flux distributions must be calculated at many
time-steps during core life, one commonly uses one-dimensional diffusion codes
(such as in the FEVER' depletion code) or even zero-dimensional descriptions
(such as in the LEOPARD 2 code) for preliminary survey studies.
Most detailed design calculations of the core power distribution must be per-
formed with two- and three-dimensional diffusion models, such as the PDQ 3 series
of codes developed at the Naval Reactor Laboratories. An accurate description of
the flux and power distribution in the reactor core using the finite-differenced
multigroup diffusion equations requires that the mesh spacing be at least compar-
able (or less than) the minimum neutron diffusion length in the core ( ~0:5 cm in a
LWR, ~4-5 cm in an HTGR), even after core homogenization such as that
described in Chapter IO has been performed. For example, it is common in LWR
analysis to assign one mesh point per fuel pin4 in a two-dimensional radial
calculation of the core power distribution (and even then some discrepancies arise
in certain types of burnup studies).
Fortunately many thermal reactor cores are characterized by a rather weak

coupling between the transverse and axial flux distributions. Hence one can
frequently iterate back and forth between a two-dimensional transverse and a
one-dimensional axial multigroup diffusion calculation to generate a three-
dimensional "map" of the core power distribution, thereby avoiding the expense of
a full three-dimensional treatment.
Although the use of such two-dimensional plus one-dimensional models of the
core power distribution have been quite commonly used in reactor design, there is
increasing tendency to go to direct three-dimensional diffusion calculations when
detailed evaluations of a specific core design are required. For the large, loosely
coupled cores characteristic of modern power reactors, such three-dimensional
studies require rather fine mesh structures and hence considerable computer
storage and calculational running time expenses are incurred (even for a core
subregion such as an octant). This difficulty is aggravated when there is a strong
coupling between the thermal-hydraulic and neutronic core behavior, as there is,
for example, in a BWR or when core fuel depletion is of concern. It has been found
that accurate multigroup diffusion calculations of the global power distribution
typically require between four (LWR) and nine (HTGR) energy groups for thermal
reactor analysis.
In fast reactor cores, the neutron mean free path is quite large. Hence fast
reactors are more susceptible to a homogeneous analysis (except for a small
fraction of neutrons with energies less than several keV). On the other hand, it is
much more important to treat the energy-dependence accurately in fast reactor
analysis. Hence one usually relies on many groups (~20) in a diffusion calculation
for a few-region model of the core in fast reactor analysis.
In Table 13-1 27 we have summarized the computational requirements for a
multigroup diffusion calculation of the core power distribution in various reactor
types.
There has been strong motivation to develop alternatives to the standard
finite-difference treatment of the multigroup diffusion equations. We have already
examined one such scheme based on a nodal representation 5 of the neutron flux in
the reactor core. Although such three-dimensional coarse nodal methods have
customarily utilized a one-group description of the neutron flux (such as in the
FLARE code6), more recent extensions to multigroup nodal schemes have been
developed. 7
Another alternative approach is to use so-called finite element techniques 8 to
calculate the neutron flux at each of a number of ultracoarse mesh points. Such
finite-element methods have been used for some time in performing calculations in
mechanical stress analysis, and appear to have considerable (although not fully
realized) potential for multidimensional neutron diffusion calculations.
Other coarse mesh techniques have been developed that utilize a mesh structure
of the order of one mesh point per fuel assembly. However in order to ensure the
stability of such techniques, it is usually necessary to alternate between coarse and
fine mesh calculations in performing the source iterations required in a criticality
calculation. This latter scheme, known as coarse mesh rebalancing 9 , also appears to
have considerable potential for accelerating multigroup diffusion calculations.
There is yet another alternative to a direct multidimensional diffusion calculation
that has proven remarkably successful in reactor analysis, the so-called flux
synthesis 10 method, in which a multidimensional flux is "synthesized" from a
combination of one- or two-dimensional calculations. Because of the importance of
THE CALCULATION OF CORE POWER DISTRIBUTIONS / 519
TABLE 13-1 Meshpoint Requirements for MGD Analysis of Various Reactor Types
Core Volume Core-Averaged Characteristic Diameter of Reactor

Reactor (m3) Power Density Diffusion Length in Diffusion Lengths
Typet (W /cm 3) (cm)
PWR 40 75.0 1.8 190

BWR 60 50.0 2.2 178
HTGR 430 7.0 12.0 63
LMFBR 5.7 530.0 5.0 36
GCFR 10.7 280.0 6.6 33
Necessary Number of Number of Groups Necessary Number of

Mesh Points (One per (Thermal Groups) Group-Space Meshpoints
Reactor Diffusion Length) (One per Diffusion Length)
Typet 2-D 3-D 1-D 2-D 3-D
PWR 36,100 6,859,000 4 (1) 760 144,400 27,436,000

BWR 31,666 5,635,000 4 (I) 712 126,664 22,540,000
HTGR 3,947 247,969 7 (4) 441 27,629 1,735,783
LMFBR 1,270 45,270 10 (0) 360 12,700 452,700
GCFR 1,115 37,253 10 (0) 330 11,150 372,530
tsased on a 3000 MWt core. [R. Froehlich, in USAEC CONF-730414-P2 (1973), p. VII-46-49]
this method in nuclear reactor design, we will consider it in some detail in Section
13-IV.
We should remark that most multigroup diffusion codes usually have the
capability of computing not only the neutron fluxes <J>g(r) but also their adjoints
<1>J(r). These quantities are extremely useful in making perturbation theory esti-
mates of reactivity changes due to changes in core composition or configuration.
Although perturbation theory is capable of only limited accuracy, it can provide
useful initial guesses for more detailed diffusion calculations. And since one can
frequently calculate the adjoint fluxes with only slightly more effort than that
involved in simply calculating the fluxes by themselves, many design codes have
been developed with perturbation theory options.
II. INTERACTION OF THERMAL-HYDRAULICS,

NEUTRONICS, AND FUEL DEPLETION
We have noted that the power distribution found in a reactor will deviate
considerably from that predicted for a bare, uniform cylindrical core. Such varia-
tions arise due to the presence of blankets or reflectors in most core designs,
nonuniform fuel loading (or burnup), and nonuniform coolant densities.
Many nuclear reactors are characterized by a rather strong coupling between
coolant density and neutronic behavior. As an example, consider first LWRs in
which the water coolant also serves as a moderator. Since most L WR cores are
"undermoderated," a local decrease in water density will cause a decrease in
moderation and hence a decrease in local power density. As the coolant passes up
through the core, it absorbs heat from the fuel elements and eventually will initiate
either subcooled (nucleate) or bulk boiling. Hence there will be an increase in

coolant void fraction-that is, a decrease in coolant density-as it passes up
through the core. Since the coolant density is maximum at the bottom core inlet,
one would therefore expect a bottom peaked power profile (as sketched in Figure
13-2).
A similar thermal-hydraulic-neutronic coupling can arise in liquid metal-cooled
fast breeder reactors. As we will find later, the liquid metal coolant (sodium) will
cause appreciable moderation or softening of the neutron energy spectrum in a fast
reactor. This will lead to a decrease in reactivity, since r, for a fast reactor increases
with average neutron energy. Hence a local decrease in coolant density can lead to
an increase in local reactivity and thereby an increase in the local power density
(note that this effect is in the opposite direction of coolant void reactivity effects in
LWRs). Although weaker than the coupling that arises in LWRs, this effect must
nevertheless be accounted for in core calculations.
The coupling between coolant properties and core neutronic behavior is very
much weaker in gas-cooled reactors such as the HTGR, because coolant phase
30%
X -I
0
> TI
.t::
co:::, 20% S,
C'
n
0
::0 ;;;
;:
...
C
0"' 10%
0
u
0 50 100 150 200 250 300 350

z (cm)
.;;
.!:::
C
:::,
Q) 0.6
>
·;::;
"'
-.;
..=.
~ 0.4
><
~
LL
0.2
200 250 300 350

z (cm)
FIGURE 13-2. Bottom Dux-peaking in a BWR.
change cannot occur, and also because the coolant does not provide appreciable
moderation in the core due to its very low density.
To illustrate the interactions between thermal-hydraulic and neutronic cal-
culations, let us examine the calculation of steady-state power distributions in
LWRs. One usually begins such a calculation by supplying the thermal-hydraulics
module with an initial estimate of the hot-channel factors and the axial flux profile
in the core. This module will then calculate the core temperature distribution and
coolant void fractions. (Actually, the thermal-hydraulics module may only calcu-
late the conditions characterizing the hot channel and an average channel, but this
thermal information is frequently sufficient for further core design purposes.) The
core temperatures and coolant density variation are then returned to the macro-
scopic group constant generation module where they are used in the generation of
group constants characterizing the core. In particular the average fuel temperature
is used to determine the correct Doppler-broadened resonance integrals, while the
moderator temperature will influence the thermal group constants. The coolant
density is used in the generation of both microscopic (i.e., spectrum-averaged) and
macroscopic group constants.
With this macroscopic group constant information, one can now use the flux-
power-reactivity module to calculate the core power distribution and nuclear hot
channel factors, which can then be used as an input to the thermal-hydraulics
module for a second improved estimate of core temperatures and void distribu-
tions. Many code packages will automatically perform such an iteration back and
forth between thermal-hydraulic and flux-power calculations until the calculated
power distribution converges to its true shape. Such an iterative approach is
extremely important in BWRs since the void fraction in a boiling channel varies
quite dramatically from inlet to exit. In Figure 13-1, we have indicated the basic
calculational procedure involved in performing coupled thermal-hydraulic-
neutronic calculations, and in Figure 13-3, we have sketched typical coolant
density variation and power distributions in a BWR. We have noted as well the
typical convergence of these profiles as one iterates between thermal-hydraulic and
neutronic calculations.
Such coupling effects can be troublesome, not only in the mathematical analysis
of the reactor, but also in the actual reactor core design since they can lead to
power peaking away from the optimally flat shape desired for the thermal design.
This can be counteracted to a degree by control rod insertion. For example, if one
inserts control rods into those regions in which power peaking is expected to occur,
the negative reactivity of the control rod absorption should cancel the enhanced
reactivity due to coolant density changes and eliminate the peaking. Such a scheme
is commonly used in BWRs that use bottom-inserted control rods to inhibit the
tendency toward bottom-peaked axial flux profiles.
These coupling effects are greatly complicated by nonuniform fuel depletion,
however. The fission rate and hence the rate of fuel burnup is proportional to the
neutron flux. Hence the fuel will deplete more rapidly in those assemblies exposed
to higher fluxes, thereby leading to decreased local reactivity and hence, a modified
flux distribution. As we will find in our discussion of core depletion calculations in
Chapter 15, such nonuniform burnup effects on the flux and power profiles can
become quite complicated when interactions with coolant density variations and
control rod patterns are taken into account.
Initial guess
Intermediate iteration
0.5 1.0
z
H
1.0
P(z)
Pmax
0.5
0.5 1.0
z
H
FIGURE 13-3. Convergence of neutronics-tbermal-bydraulics iteration.
III. PARAMETERIZATION OF FEW-GROUP CONSTANTS
One of the most important and complex aspects of nuclear reactor analysis is
the generation of suitable thermal and fast few-group constants characterizing
absorption, scattering, removal, and fission cross sections and diffusion coefficients
for each core region of interest. A variety of input data are necessary for this
calculation. In particular, the fuel and moderator temperatures supplied by the
thermal-hydraulics module are used in the generation of Doppler-broadened res-
onance integrals and thermal spectra. The moderator density is also supplied for
the core region of interest by the thermal-hydraulics module. The composition of .
the fuel is provided either as direct input or by the depletion module. Using this
information along with basic microscopic cross section libraries, one can then
generate the fast and thermal spectra and hence compute the macroscopic few-
group constants, using appropriate cell-averaging techniques to account for self-
shielding. The absorption group constants are usually modified by adding an
effective cross section characterizing any control elements in the region of interest.
(These latter effective control cross sections are supplied by the control adjustment
module.) The complete set of group constants for all core regions of interest are
then passed along to the flux-power reactivity module.
The group constants necessary for a multigroup diffusion calculation must be
generated hundreds or perhaps even thousands of times in a core lifetime study.
Not only must they be generated for each region (e.g., node cell) of the reactor core
characterized by different composition, but also regenerated for each of these
regions whenever the composition changes (e.g., via fuel burnup or moderator
density changes). If detailed fast and thermal spectra had to be recalculated for
each change in composition, the computing costs for group constant generation
would be formidable.
In practice it is found that few-group constants frequently depend on relatively
few parameters involving core temperatures and material densities. Hence it is far
more efficient to construct tables of the values of the few-group constants for
several values of these parameters, and then to use interpolation schemes to
evaluate the group constants when necessary. 11 To provide an illustration of how
such parameterization might work, we will give a very simple example of how one
might generate group constants for L WRs within the MUFT-SOFOCATE
approach 12 •
First recall that thermal group constants were defined as averages of microscopic
cross sections a{ (E) ( characterizing reactions of type x in material j) over the
thermal neutron energy spectrum <f,(E). In the SOFOCATE approach the thermal
spectrum is obtained by solving the Wigner-Wilkins equation (9-13) characterizing
a free proton gas. It should be noted that the only parameters that enter this
equation are the temperature T and the ratio of absorption to scattering r
= "2-/kT)/fi:. •. Hence we might expect that the microscopic thermal group con-
stants generated by a SOFOCATE-type code will depend only on the parameters r
and T. This dependence is in fact rather smoothly varying, as shown, for example,
in Figures 13-4 and 13-5, which show the dependence of a} 5 on r and T. In fact
one can show, using the effective neutron temperature model (see Problem 9-10),
that 1/v absorbers will yield group constants that behave as aa~C,/(1 + C2f) 112•
The Wigner-Wilkins equation can also be applied for non- I/ v absorption, such
as that due to resonances in the thermal range. However the above two parameters
will not suffice in this more general situation. This is particularly relevant to
high-burnup cores, which usually contain an appreciable amount of 239 Pu, which is
characterized by a strong thermal resonance. One can always introduce additional
parameters to account for the presence of 239 Pu such as
(13-3)
The dependence of group constants on this parameter is essentially linear.

The above discussion suggests that it should be possible to construct tables of the
dependence of thermal group constants on parameters such as T, r, and if!. A
macroscopic cross section module would then use data from thermal-hydraulic and
depletion modules to compute these parameters and then use the precalculated
group constant tables, along with appropriate interpolation schemes, to calculate
the required thermal group constants.
For fast group constants, we should first recall that MUFT-type calculations
solve the exact slowing down equation for hydrogen coupled with the age or
Goertzel-Greuling approximation for mass numbers A > 1 to generate the fast
spectrum. Inelastic scattering is treated by a direct microgroup approach, while
resonance absorption is typically treated within the NR or NRIM approximations
(with appropriate heterogeneous corrections).
Perhaps the key parameter of use in characterizing fast-group constants is the
moderator-to-fuel density ratio in a homogenized lattice cell. (One occasionally
370
360
-;;;
...
C:
24
,/T-°K½
FIGURE 13-4. The dependence of o~5 on the moderator temperature.
360
-;;;
E 350
.,
B
\~ocu 340
330
0.10 0.14 0.18 0.22 0.26 0.30 0.34

r
FIGURE 13-5. The dependence of o~5 on r, moderator temperature=303°C.
also uses the ratio of moderator volume to fuel volume in a fuel cell, which gives
essentially the same results.) Such a parameter characterizes the effectiveness of
neutron moderation in the core. We have plotted the dependence of various fast
group constants on this ratio in Figure 13-6. Such a parameterization works quite
well, provided the moderator densities are not too low.
Unfortunately no truly satisfactory parameter exists for characterizing resonance
integrals (aside from the fuel temperature). If such parameterizations are desired, it
8
- .-------
a28 X 102
1'1
- 6 ---- -- --a.~ - - -
"'
E '"2
"'
.c
0 _25
,,,-- - - - - -
"'
C '"2
.g 5
u
5l /
"'
"'0 I
u 4
2 L-----'------'------'---~~---:'
0 0.5 1.0 1.5 2.0 2.5
Water density/uranium dioxide density
FIGURE 13-6 Group 1 averaged cross sections (solid line); Group 2 averaged cross sections
(dashed lines).
is usually necessary to rely on various empirical correlations for the resonance

integrals, similar to those discussed in Section 10-111.
The above discussion has been intended as a very simple illustration of group
constant parameterization. A variety of other parameters might be used, such as
those characterizing burnup or control rod insertion. The specific parameterization
scheme one chooses (if any) will depend on the accuracy desired from the
calculation and the design problem of interest.
IV. FLUX SYNTHESIS METHODS
A. Spatial Synthesis
Spatial flux synthesis is a technique whereby one- or two-dimensional dif-
fusion calculations are combined to yield a representation of a three-dimensional
flux distribution. There are a variety of techniques, both heuristic and formal, for
blending such lower dimensional calculations together. For purposes of illustration,
let us first consider a simple example. 13
Suppose we wish to analyze a cylindrical reactor core that has been partitioned
into both radial and axial zones (denoted by indices / and m respectively) (see
Figure 13-7) in whl~h the core composition is presumably different. If the reactor
core were of uniform composition (which it is not necessarily assumed to be), the
flux would be separable in the axial and radial coordinates, and in particular would
lm
1
---/--+- FIGURE 13-7. Axial and radial zoning.
have the form (for a given energy group)
(13-4)
Of course the flux is more generally not of this separable form, but the idea of
spatial synthesis is to represent it as such and then attempt to determine an
appropriate radial shape factor c/>t(r) in each axial region m. Here the axial
buckling Bz is presumed known (or guessed).
If we substitute this separable flux shape into the multigroup diffusion equations,
we arrive at a one-dimensional equation for the radial flux profile
g-1 G
= L ~:;,.c/>t+ !xg L vg,~~c/>t, (13-5)

g'-1 g'=I
which is characterized by an effective total cross section
(13-6)
Notice that here DtB/ represents removal of neutrons due to axial diffusion. We
can now solve this one-dimensional problem to determine the radial flux profile
c/>t, and from this calculate the radial buckling for each region (/,m) as
Lf
g
r1
r1-1
'v2</>t (r)dr
B2
r,,m
=------- ( 13-7)
An analogous approach can be taken to determine the axial profile. That is, one
solves the axial one-dimensional diffusion equation assuming an effective cross
section
( 13-8)
where
B,2(
I
z) = B/Im for z in region m. (13-9)
Having determined this axial profile, one can then calculate an axial buckling
B}Im for each region. This can be used to readjust the effective cross section in the
radial multigroup diffusion equations and hence allow one to determine a new
radial profile. By iterating back and forth between one-dimensional radial and
axial profiles, one can eventually converge on a three-dimensional solution in each
region of the form
(13-10)
To the extent that the flux in each region can be approximated as a separable
function of r and z, this scheme will yield adequate results. Although this proce-
dure is actually a very primitive form of a flux-synthesis method, it is more
commonly known as a buckling-iteration method. 13 It can be generalized by using a
two-dimensional radial calculation and a one-dimensional axial calculation.
However a far more useful generalization is to choose a more elaborate repre-
sentation or "synthesis" of the flux. Suppose we represent the flux in the core as a
superposition of separable terms:
N
c/>g(x,y,z)= ~ c/>g,,(x,y)1[1g,,(z). (13-11)
n=I
The radial shape functions c/>g. (x,y) are usually constructed by performing two-
dimensional static flux calculations for x-y slices at a small number (N) of axial
locations. To determine the axial "blending" coefficients 'Pg,, (z) one can use a
variety of techniques. For example, a common scheme is to use weighted residual
methods, 14 in which one requires that a weighted integral of the diffusion equation
over the x-y plane vanishes
J= l, ... ,N, (13-12)
for N different weighting functions w/x,y). Here Mand Fare the usual diffusion
and fission operators (we have avoided using a multigroup notation for simplicity).
It should be noted that relations such as Eq. (13-12) will lead to a set of N-coupled
one-dimensional diffusion equations for the axial coefficients "1n(z). A more satisfy-
ing approach is to use a variational principle to derive these one-dimensional
problems, which can be regarded effectively as just the average of the three-
dimensional diffusion equation over the x-y plane (weighted by the transverse flux
shapes c/>n(x,y)).
All such schemes lead to very similar one-dimensional problems for the 'I/lg. (z).
Notice that as we have written Eq. (13-11), the flux synthesis applies to the entire
core. Such a scheme is known as single-channel synthesis, 10• 15 • 16 since the axial
coefficients if;8• (z) are the same across the core.
Yet in our earlier example we saw that it was useful to divide the core into both
radial and axial regions. Hence we are led to generalize the expansion Eq. (13-11)
such that we use different radial expansion functions in different radial regions
("channels").
N
cf,; (x,y,z) = ~ if;;.( z )c/>:. ( x,y ). (13-13)
n=I
Such multichannel synthesis provides a great deal more flexibility in the choice of
expansion functions. 17
Single-channel synthesis methods are commonly used to synthesize three-
dimensional fluxes from the results of both two- and one-dimensional calculations.
The expansion functions are usually chosen, using physical insight, such that they
bracket the expected flux behavior. They can be generated by performing static
criticality calculations in one- or two-dimensional form for various transverse slices
of the core. Such techniques have proven quite successful in many problems. (It
might be noted that many synthesis codes synthesize the reactor power density
rather than the flux itself, 16 for it is found that this gives a more accurate estimate
of the three-dimensional core power distribution, which is of course usually the
information one seeks.)
Experience with multichannel synthesis is far more limited. However in those
cases in which the flux is highly nonseparable or in which it is difficult to choose
the appropriate expansion functions, multichannel synthesis methods should prove
superior.
B. Spectral Synthesis
The general synthesis process of building up a complicated solution out of
simpler, but not elementary, component parts has received a great deal of attention
in other aspects of core analysis as well. There are, for example, problems in which
significant details of the flux distribution can be predicted in advance, and in these
situations the full calculation of the flux using standard multigroup finite-difference
equations generates (at great expense) large amounts of redundant information.
For example, an accurate calculation of the flux distribution in a fast reactor using
the finite-difference multigroup equations would require the use of a great many
groups (20-30) to treat the energy-dependence at each spatial point, when in fact it
is known that the energy spectrum shifts fairly smoothly in space from one typical
mode to another. The actual "information content" consists of these modal spectra
and their relative strengths at each point, so effort is wasted in the finding of the
multigroup solution. Synthesis techniques can be used to greatly simplify such
calculations.
Thus far we have confined our attention to synthesis in the spatial variables, but
the energy variable is also amenable to such a treatment. 18• 19•20 In the sense that the
multigroup approximation can be interpreted as a synthesis method, synthesis of
the energy-dependence of the flux has been used for a long time. In the derivation
of the multigroup equations it is implicitly assumed that the fine structure of the
energy-dependence is fairly constant over larger spatial regions, and that it is only
necessary to compute scale factors to be applied to precalculated spectra for each
disjoint energy group. This proves very successful in applications to thermal
spectrum reactors, where indeed the dominant energy effect is the coupling of the
neutron "birth" region through the resonance region to the thermal region, and
usually only a few energy groups are sufficient. However things are not so simple in
fast reactor analysis, where many groups would be required to adequately describe
the energy-dependence.
Proper multigroup analysis of a fast reactor requires calculations in 20-30 groups
primarily because of the effects of the resonances in the fast region. The spatial
variations of the flux, however, are fairly smooth because the long mean free paths
of fast neutrons make fine structural detail "invisible." This effect compensates
somewhat, but not entirely, for the greater number of energy variables because the
number of spatial variables can be reduced, but it does not alter the fact that the
spectrum is everywhere in transition, so that many material regions should be used.
Using 20 groups makes a two-dimensional diffusion code expensive to run and
detailed three-dimensional analyses almost impossible.
Spectral or energy synthesis involving the use of overlapping, rather than disjoint,
energy expansion functions seems to hold out the promise of solving the space and
energy fast flux problem without doing all of the (diffusion theory) work. Spectral
synthesis methods are implemented by expanding the flux as a linear combination
of known functions of energy
</>(r,E)= Li/lm(r)x,,,(E). (13-14)

m
Just as in spatial synthesis methods, weighted residual or variational techniques can

be used to obtain a set of coupled diffusion equations for the space-dependent
combining functions 1/lm(r). These equations are essentially equivalent in structure
to the usual multigroup diffusion equations, except the functions 1/lm(r) refer to the
portion of the neutron flux characterized by a given energy spectrum x,,,(E) rather
than a restricted range of energies Eg < E < Eg_ 1. Hence spectral synthesis is
sometimes referred to as the overlapping multigroup technique.
Numerous studies have demonstrated that spectral synthesis methods do indeed
provide an accurate alternative to multigroup diffusion calculations. 18 • 19 When
combined with suitable numerical algorithms,2° the spectral synthesis method can
also yield very considerable savings in computation time over conventional mul-
tigroup techniques. Hence these methods are currently under active investigation
and application, particularly to problems in fast reactor analysis.
V. LOCAL POWER-PEAKING EFFECTS
In any heterogeneous core configuration there will be local variations in the

neutron flux or power distribution that must be taken into account in reactor core
design. We have already examined how the spatial variation of the flux in the
vicinity of a fuel pin is treated by using collision probability methods to compute
cell-averaged group constants. We will employ very similar techniques to analyze
control elements in Chapter 14.
However there are other local spatial flux variations occurring due to variations
in core composition and playing an important role in reactor design. Of major
concern is the local power-peaking that occurs at the boundary between fuel and
moderator regions. For example, in L WRs one is concerned with the power-
peaking that occurs in water channels between fuel assemblies or the channels from
which control rods have been withdrawn. Since we have found that the moderator
acts as an effective source of thermal neutrons (due to slowing down), we might
expect that the fuel elements adjacent to such water channels will see a larger
thermal flux and hence experience a higher power density. Since the local flux near
the channel may be considerable higher than the average flux in the region, one
must take care that constraints on core power densities (such as the critical heat
flux limitations) are not exceeded. To this end, one would like to compute the local
power-peaking factor for the channel that gives the peak-to-average flux.
Similar effects can occur at the boundaries between zones of different fuel
loading, or between the reactor core and reflector regions (such as in HTGR cores
characterized by axial fuel-loading zones). Although the detailed calculation of
such local power-peaking factors generally involves multidimensional, multigroup
diffusion calculations for the core region of interest, we can illustrate most of the
physical ideas involved with a very simple modeled calculation. 21 •22
We will consider a one-dimensional slab geometry composed of two adjacent
regions characterized by different composition (Figure 13-8). For example, one
region might represent a water channel and the other region, a fuel assembly (in
which fuel elements and coolant channels have been appropriately homogenized).
For point of reference, we will assume that region l is characterized by a lower
absorption than region 2 so that we can confine our attention to determining the
power-peaking that occurs in region 2 due to the presence of region 1. For
convenience, we will use a one-speed diffusion model:
(13-15)
where S 1 and S 2 represent spatially uniform slowing down sources within each
region. Consistent with our interpretation of this modeled problem as representing
a unit cell of the core lattice, we will use zero-current boundary conditions at x = 0
0 0
0 a b
FIGURE 13-8. Two-region slab-geometry model of power peaking.
and x = b, as well as the interface conditions at x = a:

11(0)=0, Jz(b)=0,
(13-16)
<{> 1(a)= </>z{ a), J 1(a)=Jz(a).
We can easily solve Eq. (13-15) for the fluxes <{> 1(x) and <{>i(x) in each region.
Actually we are not interested in the flux itself, but rather in the power-peaking
factor FPP characterizing region 2, which is defined as the peak-to-average power
ratio
F = </>max = _ _~_f2_<1>_z(_a_)__
pp </>ave
b-a
f
~f2 b
<{>z(x)dx
'
(13-17)
where we have noted here that the peak power will occur at the interface between
the two regions. If we solve for <{>z(x) and then substitute into Eq. (13-17), we find
the general result
(13-18)
where
A2=---------------- (13-19)
b- a ) D2L1 a . ( b- a)
cosh ( ~ + D L cothz:smh ~
2 I 2 I 2
Actually this result is a bit too complicated for our purposes. Instead we will
assume that region 1 (e.g., the water channel) is narrow such that a«L 1, while
region 2 (the fuel) is wide such that b-a» L 2 • Then one can reduce Eq. (13-18) to
the simpler approximate form
F
PP
= - - -aS1
-- (13-20)
S2+ (b-a)
This simple result reveals several interesting aspects of local power-peaking. First,
the peaking will increase as the width of the moderating region a increases, or as
the diffusion length L 2 in the fuel decreases (e.g., by using higher enrichment fuels).
We can decrease the power-peaking by either decreasing the slowing down source
S I in the channel (e.g., by using control rod followers characterized by weak
moderating properties), increasing the slowing down source in the fuel region, or
decreasing the fuel region width b-a.
There are other possible schemes for reducing such local power peaking that are
not described by this model. For example, it is common practice in BWRs to use
fuel elements of somewhat lower enrichment near fuel assembly boundaries. One
can also use distributed burnable poisions to flatten out the power-peaking.
The power-peaking in water channels in a LWR is mitigated by some extent by
thermal-hydraulic coupling effects. This is particularly true in BWRs in which the
very large negative void coefficient of reactivity due to boiling causes an inherent
self-flattening of local power peaks. 23 •26 We have sketched the power-peaking in a
water channel adjacent to a BWR fuel assembly for several different coolant
densities in Figure 13-9.
Yet another effect which gives rise to power-peaking is fuel densification, which
can occur in metal clad fuel elements. 24 It has been found that ceramic fuel pellets
can shrink both axially and radially when irradiated in a reactor core. This
shrinkage gives rise to gaps in the fuel-pellet column within the fuel element.
Because of decreased neutron absorption in the gap, there will be power-peaking in
fuel elements adjacent to the gap.
The primary power peaking effect in HTGRs arises at boundaries between zones
of differing fuel loading25 (see Figure 13-10). Although the power-peaking factor is
usually close to unity (typically FPP~ l. l), such peaking effects must be accounted
for in HTGR core design.
0000000 00 000
0000000 00 ·000
0000000 00 000
0000000 0 000
0000000 000
0000000 000
0000000 ·000
12.4 cm
} I
\____,--I I
Water ,,_,--- Zr wrapper
ct.
20 Fuel + moderator
Water density:
16
0 0.762 g/cm 3 FPP = 1.330
@ 0.467 g/cm 3 F 00 = 1.343
@) 0.291 g/cm 3 FPP = 1.352

12
Flux (arbitrary units)
4
'
0.95 cm i I
0 - ..~ i-.._•_-0.-20_c_m_ _ _ _ _ _ 6.7 c m - - - - - - - - - - -
FIGURE 13-9. Power-peaking in a water channel adjacent to a BWR fuel assembly.

1.4
i
j i
0
o. 1.2
e~ ,·
~ i
j 1.0 \:::;:
e
0
~
g 0.8
-~
Axial power/time
.!:
... 0.6
~ - Beginning of cycle
0 ---- End of first year
c..
0.4 -·- End of fuel life
Fraction of core height, z/zmax
FIGURE 13-10. Power-peaking due to axial zones of different fuel loading in an HTGR. 25
REFERENCES
1. F. W. Todt and L. J. Todt, FEVER/Ml-A One-Dimensional Depletion Program for

Reactor Fuel Cycle Analysis, Gulf General Atomic Report GA-9780 (1969).
2. R. F. Barry, LEOPARD-a spectrum-dependent non-spatial depletion code for the
IBM-7094, USAEC Report WCAP-6058 (1964).
3. W. R. Cadwell, PDQ-7-a program for the solution of the neutron diffusion equations
in two dimensions, USAEC Report WAPD-TM-678 (1967).
4. R. L. Hellens, The physics of PWR reactors, New Developments in Reactor Physics and
Shielding, CONF-720901, (1972), Vol. I., p. 3.
5. E. G. Adensam, et al., Reactor and Fuel Proc. Tech. 12, (2) (1969).
6. D. L. Delp, D. L. Fischer, J. M. Harriman, and M. J. Stedwell, FLARE, a three-
dimensional boiling water reactor simulator, General Electric Report GEAP-4598
(1964).
7. R. G. Steinke, A coarse nodal method for solving the neutron diffusion equation,
University of Michigan Ph.D. Dissertation (1972); H. W. Graves, Jr., Evaluation of
power distribution in large reactors using a two-group nodal method, University of
Michigan Ph. D. Dissertation (1973).
8. C. F. Kang and K. F. Hansen, Trans. Am. Nucl. Soc., 14, 199 (1971).
9. S. Nakamura, Nucl. Sci. Eng. 43, 116 (1971).
10. S. Kaplan, Nucl. Sci. Eng. 13, 22 (1962).
11. M. R. Bucker and H. C. Honeck, Mathematical Models and Computational Techniques
for Analysis of Nuclear Systems, USAEC Report CONF-730414-Pl (1973), p. IX-42.
12. S. Karin, P. E. Meyer, and R. G. Steinke, An automated modular light water reactor
calculation code package, Proceedings of Conference on New Developments in Reactor
Mathematics and Applications, CONF 710302 (Vol. 2) (1971) p. 983.
13. A. Radkowsky (Ed.), Naval Reactors Physics Handbook, Vol. I, Selected Basic Tech-
niques, USAEC Report TID-7030 (1964) p. 656.
14. W. M. Stacey, Jr., Modal Approximations: Theory and An Application to Reactor Physics,
15. J. B. Yasinsky and S. Kaplan, Nucl. Sci. Eng. 28, 426 (1967); S. Kaplan, Nucl. Sci. Eng.
13, 22 (1962).
16. R. D. Traylor, V. Malakhof, and S. Leighton, SCANAL, a single-channel synthesis
depletion code with triangular mesh in the horizontal plane, Gulf General Atomic
Report GA-9423 (1969).
17. E. L. Wachspress and M. Becker, Variational multichannel synthesis with discontinuous
trial functions, USAEC Report KAPL-3095 (1965).
18. P. G. Lorenzini and A. H. Robinson, Nucl. Sci. Eng. 44, 27 (1971).
19. R. J. Neuhold, Nucl. Sci. Eng. 43, 74 (1971).
20. W. G. Price, Jr. and J. J. Duderstadt, Nucl. Sci. Eng. 55, 98 (1974).
21. H. W. Graves, Nuclear Reactor Design, University of Michigan Nuclear Engineering
Department Lecture Notes (unpublished) (1969).
22. R. W. Deutsch, Nucleonics 16, 95 (1958).
23. R. L. Crowther, Physics measurements of BWR reactors and comparison with theory,
New Developments in Reactor Physics and Shielding, USAEC CONF-720901 (1972), p.
114.
24. Technical session on fuel densification, Trans. Am. Nucl. Soc. 8, 120 (1974).
25. GASSAR 6, General Atomic Standard Safety Analysis Report, GA-Al3200 (1975).
26. N. McFarlane, Nucl. Appl. Tech. 9, 634 (1970); E. Fuller, Nucl. Appl. Tech. 9, 622 (1970).
27. R. Froehlich, in Mathematical Models and Computational Techniques for Analysis of
Nuclear Systems, USAEC Document CONF-730414-P2 (1973), p. VII-I.
28. 0. J. Marlowe, et al., WANDA, A One-Dimensional Few Group Diffusion Code,
WAPD-TM-28 (1956), WAPD-TM-241 (1960); H.P. Flatt and D. C. Baller, AIM-5, A
Multigroup One-Dimensional Diffusion Code, NAA-SR-4694 (1960).
PROBLEMS
13-1 Numerous factors in a nuclear reactor core design depend on the moderator-to-fuel
volume ratio VM/ VF in a fuel cell. Suppose we imagine decreasing this ratio in a
LWR, by decreasing lattice pitch while keeping fuel rod radius fixed. Indicate
qualitatively how and why each of the parameters below might be affected by such a
change in the reactor core if the core is initially undermoderated.
(a) Nuclear factors: 71, f, p, t:, PFNu PTNu and kw
(b) Thermal-hydraulic factors: Core power density, coolant flow rate, core pressure
drop, fuel and clad temperature, and DNBR.
(c) Safety factors: Temperature coefficient of reactivity and response to a loss of
coolant accident.
(d) Economic factors: Fuel inventory, core size, and conversion ratio.
13-2 Using first-order perturbation theory, estimate the sign of the change in the thermal
group constant ol5 due to the presence of 239 Pu. For convenience, model the thermal
resonance of 239 Pu as a Dirac 8-function such that
and, in particular, examine the cases of E 0 ~ Er= kT.

13-3 Demonstrate that for small I/;, the thermal group constants o depend linearly on iJ;.
13-4 Provide a physical explanation for the dependence of the fast-group constants on the
moderator-to-fuel ratio shown in Figure 13-6. (Remember that these results were
generated using a MUFT scheme.)
13-5 The following problem will require the use of a very simple code package which
includes: (a) a simple group constant generation code (e.g., SOFOCATE-MUFT),
(b) a one-dimensional few-group diffusion code (e.g., WANDA28 or AIM), and (c)
a single-channel thermal-hydraulics code capable of analyzing a boiling water
channel in a BWR. This problem demonstrates a number of topics in Chapter 13,
including thermalhydraulic coupling and group constant parameterization. Consider
the BWR/6 core design given in Appendix H. Idealize the reactor core as an infinite
bare slab (Figure 13-11). Using a simple LWR code package, determine both keff and
the axial power distribution for the reactor. In your solution to the problem,
carefully outline your calculational procedure as well as the reasoning that led to the
procedure. Hints are as follow:
(a) The purpose of this problem is to demonstrate the strong coupling between the
neutronic calculations and thermal-hydraulic calculations in reactor design.
(b) You will have a limited amount of computer funds, so give careful thought as to
how you can best spend the money to get maximum accuracy in your solution.
(c) Do not treat this as a criticality search problem! Both composition and geometry
are fixed.
Procedure:
(i) Run the thermal-hydraulics code to obtain representative water densities,
fuel temperatures, and moderator temperatures. Use a bottom-peaked
profile since k'$J (and power) are expected to be greater in a nonboiling
region of the core.
(ii) Run the fast and thermal spectrum codes for high, low, and average water
density, using average fuel and moderator temperature. Determine
~{ (Ptt20). Actually ~{ is a function of True! as well. Since there is not a
direct correlation between True! and Ptt 2o, one should plot ~{ versus both
True! and Ptt 2o, However to minimize computing costs, the same fuel
temperature True! can be used for all ~{. Plot ~{ versus Ptt2o so that the
spectrum codes need not be run for other water densities. (Use the graph to
obtain ~{ for other Ptt 2o,)
(iii) Divide the core into suitable regions and find the average density in each
region from the thermal-hydraulics code. From these densities determine
all ~{ for each region as input to the diffusion code.
(iv) From the multigroup diffusion calculation, obtain keff and the power
profile for input to the thermal-hydraulics calculation.
(v) Since neutronics and thermal-hydraulics are coupled, iterate between the
MGD and T-H codes until convergence of keff and the power profile
occurs.
Fuel rods
370 cm
FIGURE 13-11. Slab model of axial BWR power calculation.

536 / AN INTRODUCflON TO NUCLEAR REACTOR CORE DESIGN
13-6 Use a one-dimensional code to perform a buckling iteration between axial and radial
calculations on the BWR of Problem 13-5 similar to that described in Section 13-IV.
For convenience, assume that the moderator density profile is fixed, obviating the
necessity of performing a thermal-hydraulic calculation.
13-7 Using the weighted residual method, develop the general form of the equations
describing the axial weighting coefficients in a single-channel synthesis scheme. For
convenience, treat only the case of one-group diffusion theory.
13-8 Choose the energy-expansion functions in the spectral synthesis method to be
disjoint step functions
I, Em< E<, Em-I

Xm ( E) = 0, otherwise,
m=l, ... ,M.
Then using the weighted residual method, verify that this choice indeed leads to the
usual multigroup diffusion equations. For convenience, choose the weight functions
wm(E) to be the x,,,(E) proper. (This is known as the Galerkin weighting scheme.)
13-9 Explicitly derive the form given for the power-peaking factor FPP in Eq. (13-20).
B-10 Repeat the derivation of the power-peaking factor for the slab geometry of Figure
13-8 by equating the current leaving the moderator region to the rate of thermal
neutron production in this region.
13- ll Derive the power-peaking factor characterizing two adjacent fuel regions of differing
enrichment.
13-12 In the BWR described in Problem 13-5, water slots are provided between fuel
assemblies to accomodate the cruciform shaped control rods (see Figure 13-9). When
the control rods are withdrawn from the water channels, power-peaking can occur in
neighboring fuel rods. We will model the geometry by the one-dimensional repre-
sentation shown in Figure 13-9.
(a) Use a one-dimensional, one thermal group diffusion model to determine: (i)
shape of the flux in the unit cell, and (ii) the power-peaking factor FPP for the
unit cell.
(b) What is the trend in FPP if: (i) the water channel thickness is increased, (ii) the
fuel enrichment is increased, and (iii) stainless steel cladding is substituted for
zircaloy cladding.
Justify your answers.
14
Reactivity Control
I. INTRODUCTION
Nuclear reactors must be initially loaded with a significantly larger amount of

fuel than that required merely to achieve criticality, since the intrinsic multiplica-
tion of the core will change during core operation due to processes such as fuel
burnup and fission product production. Sufficient excess reactivity must also be
provided to compensate for negative reactivity feedback effects such as those
represented by the temperature and power defects of reactivity. In contrast to the
core geometry and volume, which are primarily determined by thermal considera-
tions (i.e., achieving the required core thermal power output utilizing a power
density consistent with thermal design limitations), the fuel loading or enrichment
will be determined by the desire to build into the core sufficient excess reactivity to
allow full power operation for a predetermined period of time.
To compensate for this excess reactivity, it is necessary to introduce an amount
of negative reactivity into the core which one can adjust or control at will. This
control reactivity can be used both to compensate for the excess reactivity nec-
essary for long term core operation and also to adjust the power level of the
reactor in order to bring the core to power, follow load demands, and shut the core
down. The control reactivity is most often present in the form of strong neutron
absorbers that can be inserted into or withdrawn from the core (although movable
core reflector elements or fuel assemblies and coolant flow have also occasionally
been used for reactivity control). The determination of the control reactivity
requirements and the apportionment of control reactivity among various types of
control elements is a very important aspect of nuclear reactor core design. In this
introductory section we will classify the control requirements of nuclear reactor
537
cores, as well as the various control elements commonly used in power reactors. It
is first useful to introduce several definitions characterizing reactivity control:
(1) Excess reactivity Pex: The core reactivity present with all control elements
withdrawn from the core. Pex will be a function of both time (due to fuel
burnup and isotope production) and temperature (due to reactivity feed-
back). Larger values of Pex will generally imply longer core lifetimes, but
at the expense of larger control requirements and poorer neutron
economy (since with more control reactivity in the core, there will be
more neutron absorption).
(2) Shutdown margin Psm= The negative reactivity of the core present when all
control elements have been fully inserted to achieve minimum core
multiplication. Again the shutdown margin Psm is a function of time and
temperature. For example, the shutdown margin for a "cold," "clean"
core, that is, a core at ambient temperatures and with a fresh fuel
composition in which no depletion or fission product buildup has oc-
curred, will be quite different from the shutdown margin characterizing a
core that has been operating at power for some time. Typically shutdown
margins are chosen such that the core multiplication is below critical (e.g.,
k = 0.99) even with the most reactive control rod stuck in the full "out"
position (the so-called stuck-rod criterion). 1
The shutdown margin not only characterizes the core multiplication in its
shutdown state, but is also related to the rate at which the reactor power
level may be reduced in an emergency shutdown or "scram." In particular
we recall that the prompt jump approximation implies that the fractional
power level decrease achieved immediately after control insertion is given
approximately by:
Power before control insertion /3 - Po /3

Power after control insertion
= /3 pI
= p + /3 · ( 14-1)
- sm
For example, a shutdown margin of Psm = .030 would imply that the
reactor power level in a 235 U-fueled thermal reactor ( /3 = .00655) would
drop to 18% of its initial value immediately following a scram.
(3) Total control element worth Lip: The difference between the excess reactiv-
ity and the minimum reactivity when all control elements are fully
inserted. That is,
Lip= Pex + Psm' (14-2)
The precise definition of reactivity worth is occasionally given a slightly

different characterization, depending on whether the control element of
interest is intended for power-level changes or for longer term fuel burnup
compensation.2 We will utilize the simple definition implied by Eq. (14-2)
above.
(4) Control element worth Lip;: One can also define the reactivity worth of an
individual control element as the reactivity change induced in the core by
the full insertion of the element. In Table 14-1, we have listed typical
values of the excess reactivity requirements, shutdown margins, and
REACTIVITY CONTROL / 539
control element worths for the principal reactor types. We will return later
in this chapter to discuss how this reactivity control is allocated among
various control elements. It should be noted that the amount of excess
reactivity and hence control required for thermal reactors is considerably
larger than that required for fast reactors (in which fuel breeding com-
pensates for fuel depletion).
TABLE 14-1: Reactivity Requirements in Major Reactor Types
Reactivity (t:i.k/ k) BWR PWR HTGR LMFBR
Excess reactivity Pex of

clean core:
at 68 °F 0.25 0.293 0.128 0.050
at op. temp 0.248 0.037
ateq. Xe, Sm 0.181 0.073
Total worth of control, l::i.p 0.29 0.32 0.210 0.074
Control-rod worth 0.17 0.07 0.16 0.074
Burnable poison worth 0.12 0.08 0.10
Chemical shim worth 0.17
Shutdown margin P,m
Cold and clean 0.04 0.03 0.082 0.024
Hot and eq. Xe, Sm 0.14 0.137 0.037
One can distinguish several different types of control requirements:

(1) Scram control: The reactor control system must be capable of shutting the
reactor down under any credible operating conditions. Elements used for
such scram control purposes must be capable of inserting negative reac-
tivity very rapidly and must operate with an extremely high degree of
reliability.
(2) Power regulation: Certain control elements are designed to compensate
for small reactivity transients caused by changes in load demand, core
temperature, and for power-level maneuvering.
(3) Shim control: Shim control elements are designed to cover the excess
reactivity necessary to compensate for long-term fuel depletion and
fission product buildup, as well as to shape the power distribution in the
core in order to obtain better thermal performance and more uniform fuel
burnup. Although the reactivity worth of such elements must of necessity
be quite large, shim control adjustments are made very gradually over
long time periods.
There are several schemes used for introducing control absorption into a nuclear
reactor core. One common method is to insert movable rods of absorbing material
into the core. Such movable control elements not only can be used to adjust the core
power, but because of their rapid response can also be used for scramming the
reactor, as well as for shim and power shaping. Fixed absorbing materials are
sometimes fabricated into the core, with the intent that such absorption will
gradually burn out along with the fuel. These purnable poisons are useful for
extending the initial core lifetime of reactors. A third very popular control
mechanism in LWRs involves dissolving a poison such as boric acid in the coolant
itself. Such a soluble poison provides a very uniformly distributed shim control
which minimizes spatial power profile perturbations.
S40 / AN INTRODUCTION TO NUCLEAR REACTOR CORE DESIGN
A large power reactor will contain many control elements distributed in a

pattern about the core in an effort to achieve efficient reactivity control within
flux-peaking limitations, as well as to facilitate power shaping. The grouping of
such control rods, the determination of the control rod withdrawal sequence, and
the coordination of movable control elements with burnable and soluble control
poisons all fall within the realm of what is known as control management. Since the
reactivity worth of control elements will change, due to both the depletion of
control absorber nuclei and changes in the core power distribution, control
management analysis must continually monitor the worth of the control system
components throughout core life in order to maintain both an adequate shutdown
margin and sufficient operating maneuverability.
II. MOVABLE CONTROL RODS
A. General Considerations Involved in Control Rod Design

Movable control rods are used for a variety of purposes-to control reactivity
for power maneuvering, shim control, or reactor scram. In the early days of nuclear
reactor development, control elements were typically cylindrical in shape. However
in today's tightly packed fuel lattices, control elements are typically cruciform-
shaped blades or clusters of narrow control rods, although larger cylindrical rods
continue to be utilized in gas-cooled reactors. We have included illustrations of
typical movable control rod assemblies in Figures 14-1, 14-2, and 14-3.
I,
0
,,,,
I
• O'
0
lo
"'
0
•
I " • Blade
I ",,
Q
I
•I
" t
I
0
•
FIGURE 14-1. BWR control rod assembly. 15

assembly
Top
nozzle
Grid
assembly
FIGURE 14-2. PWR rod cluster control assembly. 16
Steel cable to
control rod drive
Absorber
container ;
Rod ~ / .
Control
fuel
assembly
Shock
absorber
FIGURE 14-3. HGTR control rod and control fuel assembly.2 1
541
A number of factors are involved in the design of such control elements. 3 For
example, one would desire to achieve the required reactivity worth with a minimum
perturbation on the flux distribution in the core. Obviously if the control rod is too
highly absorbing (too "black"), there will be a very pronounced flux depression in
its vicinity. Hence one generally attempts to design rods such that their thickness is
not much over several mean absorption lengths. Such considerations place a
premium on high surface-to-volume ratios for control elements (such as those
characterizing cruciform blades).
In thermal reactor control, most of the neutron absorption in control elements
(some 60%-80%) occurs in the thermal energy range. However there will be an
epithermal absorption effect due to resonance absorption. In an ideal design, a
mixture of absorbing isotopes would be used to provide absorption over a range of
neutron energies (e.g., the Ag-In-Cd alloy used in PWR control elements).
There are a number of other considerations involved in designing a control
element. For example, one desires a composition and design having a long lifetime
against absorber depletion. This requires a large number of absorbing atoms per
unit volume. The depletion of absorber material is a prime factor in determining
control rod thickness. The control element must also be designed with considera-
tion given to mechanical performance. For example, the intense radiation en-
vironment in which the rod must function will cause high radiation damage.
Furthermore the control element must exhibit good resistance to corrosion, as well
as being able to tolerate the high operating temperatures in the core. Such
temperatures are augmented by the heat generated within the element due to
neutron capture. Finally the fabrication costs of the element will be an important
factor in determining the design.
B. The Calculation of Control Rod Worth

The calculation of the reactivity worth of individual control elements and
groups of such control elements and the effects of such elements on the power
distribution in the core of a nuclear reactor constitute a very important facet of
nuclear reactor design. A variety of methods have been developed to calculate
control rod worth.
One might be tempted to utilize perturbation theory to estimate the effect of a
localized control absorber on core multiplication. Unfortunately, however, the
strong absorption characterizing a control element causes a severe local distortion
of the flux, particularly in those fuel assemblies adjacent to the rod. The control
element is characterized by strong self-shielding, much as fuel pins are. Hence
perturbation theory is of limited use in control studies, unless the control element is
characterized by relatively weak absorption. (Although the control absorption is
usually too strong to allow meaningful perturbation theory estimates in thermal
reactors, the decrease of absorption cross sections and increase in mean free paths
with increasing neutron energy allows some justification for use of perturbation
methods in fast reactor control studies. And, as we saw earlier in Chapter 5, even
though perturbation theory usually does not provide meaningful estimates of
absolute control rod worths, it does in fact prove useful in studying the relative
worth of various control element patterns and positions.)
A more common approach involves the use of one-speed diffusion theory
treatments that utilize transport-corrected boundary conditions at the surface of
the control rod. The literature4 abounds with diffusion theory studies of cylindrical
rods inserted into bare, homogeneous cylindrical reactor cores-either on or off
core axis. However such analyses are of little relevance to control studies of large
power reactors with relatively tight fuel lattices in which large numbers of elements
are distributed about the core in an effort to achieve a more uniform power
distribution (see Figure 14-4).
~l s ~
-0- 1 +o- lo- l lo-

s t, X X X lo- l s
X -o-1 ¼+ 14- to- X
9 X ,_ X -- X 0
__
,
9 9 14- 1 ¼+ 14' 1-4 • 9
T 1-" X - l:I
9 -:::l> -t IH H,: 14- l HJ ± ~
s :)1 X X x,. X ,_ s
-!
II
l
,~l X
± ~l
™ 14- !4l-- RC-- X - II
,~
I Q_
_9- ,~ ..2.. tt ttl fti HJ ~ ..2..

- 11"" X ,_ X X .2..
s
X
lo-
-H
X ™X l4l X t" ,0.
X
-I s
lo- -!ol lo- 1 I<> FIGURE 14-4. Location of control assemblies in a
lo- l s +o-
PWR core. 19
We will instead approach the calculation of control rod worths in a manner very
similar to that utilized for the inclusion of fuel lattice heterogeneities in the
generation of thermal group constants. We will use the fact that modern power
reactors contain large numbers of control elements which tend to be reasonably
uniformly distributed across the core allowing division of the core into cells, each
containing one control element. We will then perform a detailed analysis of one
such control cell, similar to the analysis we presented earlier for a fuel cell, in an
effort to determine an effective absorption cross section for the control element
which, when multiplied by the average flux in the control cell, yields the correct
absorption rate in the element. Such effective cross sections can then be included in
the multigroup diffusion theory analysis of the overall core. Obviously such a
scheme is very similar to the generation of "self-shielded" thermal group constants
discussed in Chapter 10.
As in Chapter 10, our approach will be to isolate a local region in the vicinity of
the control element from the remainder of the core and to analyze the neutron flux
in the region in detail. Typical control cells for several reactor types are sketched in
Figure 14-5 by way of illustration. As in our earlier analysis, we will decouple the
control cell from the rest of the core by demanding that the net neutron current
across the cell boundary is zero. The influence of adjacent cells can be represented
by relaxing this boundary condition to allow nonzero flux gradients across the
boundary.
Our approach will be to use diffusion theory with transport-corrected boundary
conditions on the control element surface in order to determine the flux in the
control cell. 5 We can then use this flux to determine an effective cross section ~~ff
characterizing the control element by requiring that
(14-3)
0000000 0000000
0000000 0000000
0000000 0000000
0000000 0000000
0000000 0000000
0000000 0000000
0000000 , 0000000
A
)
I •
' ,
i
-Fuel element and
channel assembly
Control blade j
Note:
Special corrected corner rods
have different enrichment
Fuel assembly than standard rods
~
with rod-cluster (1) Intermediate enrichment
(2) Low enrichmer\t
control
BWR
PWR
Reserve shutdown hole
Coolant hole
Control-rod channel
Burnable poison hole
HTGR
FIGURE 14-5. Control cells for several reactor types.
where ;j, is the volume-averaged flux in the cell, Vcen is the control cell volume, and
Sc the control element surface area, respectively, while Jc is the average neutron
current at the surface of the control element. This usually reduces to a two-
dimensional geometry (such as those in Figure 14-5), in which case we can write
pc Jc
L eff_
-- - (14-4)
c Acell 4' '
where Pc is the exposed control element perimeter and Acen is the cross-sectional
area of the control cell.
Hence our task is obviously to determine the ratio J j~ since this will yield ~~rr.
The highly absorbing nature of the control element invalidates the direct applica-
tion of diffusion theory. Instead, one uses a diffusion theory calculation outside the
element, collision probability methods within it, and then couples these at the
control element boundary using "blackness theory" 7 or effective transport-
corrected boundary conditions (much as was done in the ABH method).
The advantage of this approach is that it yields a cell parameter Jc/~ that is
relatively insensitive to the material composition surrounding the control element.
Hence the detailed calculation of Jj~ need be performed only once and can be
reapplied regardless of the change in the cell composition (due to fuel depletion,
moderator density changes, or fission product buildup).
To indicate more clearly just what is involved in such a calculation, let us
consider the explicit example of a cruciform-shaped control blade inserted into a
homogeneous cell. 6 We will not attack this two-dimensional problem directly, but
rather will replace it by an equivalent one-dimensional problem that preserves the
same ratio of control rod surface to fuel volume (see Figure 14-6). Such a scheme
should give a reasonable estimate if the span / of the control blade is substantially
greater than the diffusion length L characterizing the surrounding medium. The
surface-to-volume ratio of the rod is
=------ (14-5)
Vcell 4m 2 - 8 t/ + 4t 2 •
Our equivalent one-dimensional problem will be as shown in Figure 14-6 with a

surface-to-volume ratio of 1/ a; hence we will demand
a= (m 2 - 2tl + t 2 )/2!. ( 14-6)
We will now use one-speed diffusion theory to determine the flux in the fuel
( 14-7)
12m Fuel
+
moderato\
Fuel
+
moderato~ Control
2t
,..,__---2/---,>,\
FIGURE 14-6. A one-dimensional model of a cruciform control blade cell.

where we assume a uniformly distributed source of neutrons S0 slowing down in

the fuel-moderator mixture. The boundary conditions corresponding to the cell are:
J(O)=O
(14-8)
J (a)/cp(a)= o:.
Notice that we have introduced a parameter o: characterizing the ratio of the

current to the flux at the surface of the control rod. Since the control rod is a
strong absorber, o: must be determined from a separate transport calculation-
much as one determines the extrapolation length characterizing a free plane
surface, z0 =.7104 ;\tr· Indeed, by comparing Eq. (14-8) with Eq. (4-175), one can
identify an effective extrapolation length as d= D / a.
It is convenient to rewrite our expression for the effective control cross section as
(14-9)
We will solve the above diffusion equation in the fuel to determine the ratio of the
surface flux <l>c to the average flux ~- This formulation is particularly enlightening
because it stresses that the diffusion equation is only being used to describe the flux
outside of the control element.We can easily solve this problem to find
</>(x)=- 1-
So
~a [
o:cosh L
a X a
o: cosh L + ~aL sinh L
l
. (14-10)
Then we can compute the maximum-to-average flux ratio as
cp(a)
- (14-11)
</>
If we note that Pc/ A cell= 1/ a for our one-dimensional problem, we can find
~
~eff = a (14-12)
c ( ~a
a - + -I coth - a) - I
o: L L
In particular, notice that the control rod worth (which is proportional to ~~ff)
increases as the ratio of control rod surface to fuel volume increases (i.e., a
decreases) or as fuel absorption decreases, as we might have expected. Rod worth
also increases with increasing fuel diffusion length, since more of the fuel volume is
then sensitive to flux perturbations introduced by the rod.
The calculation of the transport correction parameter o: involves the use of
transport theory or collision probability methods. If the control element were a
perfectly black absorbing slab, then a transport calculation7 indicates that o: =0.47.
For gray elements, the result is more complicated and depends on the geometry of
the control element. For a slab element of width t, such an analysis yields
(14-13)
where En(z) is the exponential integral
(14-14)
For more complicated element geometries, such as sandwich or pin structures,

more complex expressions for a must be used. 8•9 In a similar sense, on~ can
calculate 'Pc/;p for more complicated cell geometries.
A number of other considerations must be included in a more accurate de-
termination of control rod worth. For example, most control blades have structural
materials in their tips that protrude into high flux regions. The effective cross
section of these regions must usually be calculated separately. Furthermore one
finds a flux depression at the corner of the blade, with a corresponding flux-
peaking at the opposite corner of the cell. The effective control cross section must
be corrected for this effect.
C. Control Rod Effects on Core Power Distributions

The primary function of movable control elements in a reactor core is to
control core reactivity. However we have seen that the strongly absorbing nature of
control elements will cause major perturbations in the neutron flux in the vicinity
of the control element. Thus far we have confined our treatment to a consideration
of such perturbations on a local scale, such as in a control cell. However as the
control elements are inserted or withdrawn, they will of course also strongly affect
the overall flux distribution and hence the power distribution of the reactor core. It
is extremely important to predict the interaction of control elements on core power
distributions in order to ensure that the power density limitations demanded by
thermal considerations such as those discussed in Chapter 12 are not exceeded.
Furthermore, control rod patterns can actually be used to shape the power
distribution into a more favorable profile for core thermal performance and fuel
burnup.
In determining the effects of control rod insertion patterns on core power
distributions, one must be careful to include the many interactions that will affect
the power profile. For example, suppose we consider the variation in the axial
power profile of a core controlled by rod insertion from the top. At the beginning
of core life, the rods must be inserted to compensate for the large excess reactivity
of the fresh core loading. The region of the core in which the rods have been
inserted will experience decreased multiplication and hence relatively lower fluxes.
Hence since the overall reactor power will be kept constant, there will be a
flux-peaking near the bottom of the core. As the fuel in the core is depleted, the
control rods will be gradually withdrawn to compensate for the reduced reactivity
of the fuel. During the early stages of core life, the fuel near the bottom of the core
will have been exposed to relatively higher fluxes and hence experience larger
burnups. Thus as the control rods are withdrawn, the flux peak will shift toward
the top of the reactor core. Such interactions between rods and fuel burnup can be
minimized by using more uniformly distributed control such as burnable poisons

and chemical shim.
Beginning of core life (BOL)
End of core life (EOL)

FIGURE 14-7. Interaction between control rod insertion and flux profile.
Another area of concern is the interaction between control rod positions and the
buildup of fission product poisons. The accumulation of fission products
characterized by large absorption cross sections, such as 135Xe, can lead to spatial
power oscillations in the core (as we will see in more detail when we study the topic
of fission product poisoning in Chapter 15). In certain instances the incorrect
management of control rod patterns can lead to instability of these spatial oscilla-
tions.
Yet a third type of interaction is found in BWR cores in which the large water
density toward the bottom of the core leads to a bottom-peaked flux profile. By
inserting control rods from the core bottom, one can compensate for the enhanced
reactivity introduced by the higher moderator density and hence achieve a
somewhat more uniform axial flux profile.
Of course there are some disadvantages in the use of control elements for
power-profile shaping. For example, the achievement of favorable radial power
REACTIVTIY CONTROL / 549
profiles is sometimes accomplished only at the expense of satisfactory axial

profiles. Furthermore the achievement of optimal fuel burnup may be partially
offset by the poor neutron economy of a core with high control insertion. These
factors must be considered in devising an optimal control management scheme.
To more explicitly study the effects of rod insertion on core power distributions,
let us consider a simple model of the interaction of a bank of control rods with an
axial flux profile. 10 We will model the reactor as a cylinder and treat the "rodded"
and "unrodded" regions of the core as homogeneous (see Figure 14-8). Hence the
insertion or withdrawal of the control rod bank is represented by a shift of the
boundary between the two regions (a so-called "windowshade" model).
z
::c:
rh
No control
L
FIGURE 14-8.
q,(z)
Insertion of a uniform rod bank into a bare uniform cylindrical core.
To simplify this analysis, we will assume that the flux is separable in rand z such
that
(14-15)
Then the axial flux profile 2:(z) satisfies
(14-16)
where the effective axial buckling Bz; is given by

2_koo;-1 (Po)2
B ---- - ( 14-17)
z, L2 R
and the i subscript has been included to distinguish between core compositions
(k 00 ;) and hence flux profiles (2'.;) in the unrodded (region 1) and rodded (region 2)
core regions.
We will seek solutions in each of these regions which vanish at the top and
bottom of the core,
(14-18)
(where we have assumed the control bank absorption is sufficiently large that Bz~ is
negative) and require the usual interface boundary conditions:
(14-19)
Using these interface conditions to match the solutions at z = h, we find that the
criticality condition becomes
(14-20)
(assuming D 1~D 2) while the relationship between the axial flux amplitudes is
A2 sinBz,h
(14-21)
A1 = sinh[ Bz (H-h))
2
.
The axial flux profile has been sketched for several different rod-bank insertion
depths h in Figure 14-8. As expected, this result indicates that rod insertion can be
used to shift and shape the peak of the axial flux profile. Hence many core designs
use partial-length control rods" containing absorbing material only over a limited
fraction of their length in order to more effectively shape the axial profiles.
D. Control-Adjustment Calculations
In our description of a general nuclear analysis code package, we indicated
that a very important component was the control adjustment module, which adjusted
the amount of control in the core to return the reactor to a critical state following a
fuel depletion step. In this sens~, the control adjustment module simulates the
operation of an actual reactor in which shim control is slowly withdrawn to
compensate for the loss of excess reactivity due to fuel burnup.
In many reactor models, one has the option of either adjusting·the concentration
of a dissolved absorbing isotope in the coolant (i.e., chemical shim) or adjusting the
positions of movable control rods using a prescribed insertion-removal sequence
scheme. Chemical shim adjustment can be modeled by merely adding a prescribed
amount of absorption cross section L; uniformly across the core. The treatment of
movable solid control elements is more complex, as we have seen, and requires the
determination of an effective control cross section for each control cell in which the
rod position has changed. Such effective control cross sections will depend not only
on the rod insertion depth, but also on the coolant density and fuel composition of
each cell. Since the frequent calculation of such cross sections can be rather
expensive, it is customary to tabulate effective control cross sections versus coolant
density and cell burnup so that a control adjustment step only requires a table
reference and interpolation (much as is done with multigroup constants).
Since the amount of control necessary to return the core to critical is not known
a priori, one must usually proceed iteratively by guessing a control adjustment,
computing keff' and then readjusting control if keff is not sufficiently close to unity.
The convergence rate of such an iterative control adjustment procedure depends on
one's ability to estimate the required control change. Such a procedure can be
greatly accelerated by using perturbation theory to estimate the reactivity worth of
a given control adjustment. Such perturbation estimates work quite well for
adjustments of a uniformly distributed control absorber, such as chemical shim.
However the limited validity of perturbation theory for strongly localized ab-
sorbers, such as movable control rods, frequently requires more elaborate treat-
ments in which the local perturbation on the neutron flux in the vicinity of the
control element is taken into account. 12
III. BURNABLE POISONS
The lifetime of a given core fuel loading, that is, the period during which the
core has sufficient excess reactivity to permit startup and full power operation, is
generally determined by the amount of fuel initially loaded into the reactor core.
(Although it should be mentioned that a major factor in determining fuel element
lifetime in today's modern power reactors is the amount of radiation damage a fuel
element can withstand before experiencing an appreciable probability of failure.)
Of course the amount of fuel loaded into the reactor will depend in part on the
excess reactivity that can be conveniently compensated for by the reactor control
elements.
In order to increase the allowable initial core fuel loading, it is common to load
into the core materials characterized by high neutron absorption cross sections
(poisons) that compensate for such excess reactivity during the early stages of core
life. 13- 15 Such absorbers are chosen such that they "burn out" (i.e., are transmuted
by neutron capture into isotopes with low capture cross sections) somewhat faster
than fuel burnup, so that later in core life they contribute negligible negative
reactivity. Hence these burnable poisons can nearly match the time behavior of the
excess fuel reactivity as it decreases over core life, thereby allowing larger initial
fuel inventories without a corresponding increase in control requirements.
Burnable poisons thus possess a number of advantages. They increase core
lifetime without any decrease in control safety, reduce the amount of mechanical
control required, and if distributed in a proper fashion, can also improve core
power distributions, for example, by suppressing reactivity in high flux regions,
such as near coolant channels.
Such burnable poisons are usually fabricated into the initial core loading as
either fixed control blades or curtains or mixed into certain fuel pins. For example,
in present BWR designs, 15 gadolinium-loaded fuel pins with an initial reactivity
worth of 0.12 t::.k / k are loaded into the core until an equilibrium fuel cycle is
achieved. In PWRs, borosilicate glass tubes are placed in the core in the initial core
loading. 16 The locations of such burnable poisons are shown in typical fuel
assembly control cells in Figure 14-5.
From this discussion, several desirable characteristics of burnable poisons are
apparent. Obviously they should be characterized by absorption cross sections
somewhat higher than those of the fuel, since then they will burn out more rapidly
SS2 / AN INTRODUCTION TO NUCLEAR REACTOR CORE DESIGN
than the fuel, leaving minimal poison residue at the end of the fuel cycle.
Furthermore the isotopes formed by neutron capture in the poison should have low
absorption cross sections. Finally, the burnable poison, as well as its surrounding
clad or structural material, should not affect the structural integrity of the core
(such as by swelling).
In this section we will examine two approaches for loading burnable poisons into
a reactor core. The first merely assumes a uniform distribution of the poison
material among the fuel. As we will see by a simple analysis, such a homogeneous
scheme encounters difficulties in matching control requirements. Hence a more
desirable scheme is to load the poisons in a heterogeneous fashion to take
advantage of self-shielding effects that can provi-de a closer match between poison
burnout and fuel depletion.
A. Uniform Distribution of Burnable Poisons

For convenience, consider an infinite medium of homogeneously mixed fuel
and poison. Then the rate equations describing the burnup of fuel and poison
nuclei are just
dNF F
dt = -aa<f>(t)Np(t),
(14-22)
dNp
dt = -11~</>(t)Np(t).
We can integrate each of these equations to find
Np(t)= Np(O)exp[ - a:<l>(t)]

(14-23)
Np(t)=Np(O)exp[-a~<l>(t)]
where <l>(t) is the time-integrated neutron flux or neutron fluence
(14-24)
Now the core multiplication can be estimated as
(14-25)
We can sketch the form of k(t) for various choices of poison cross section as shown
in Figure 14-9.
Initially the change in reactivity will be high due to poison burnup. This results
in a reactivity mismatch. The mismatch will be less severe for a poison with a small
a~, but then there will be a poison residue that will shorten core life. Thus there is a
certain fuel loading and corresponding core life beyond which the reactivity rise
k(t)
a P
•1
FIGURE 14-9. Core multiplication versus time for various uniform burnable poison loadings.
due to the relatively rapid depletion of the poison will be greater than that which
can be handled by the available mechanical control. Ideally one would like to
employ a poison of lower cross section early in core life (to reduce reactivity
mismatch) and then switch to a poison of large cross section late in core life (to
minimize poison residue). One can effectively achieve this by locating the poison in
a region characterized by a different flux than that seen by the fuel.
B. Heterogeneous Distributions of Burnable Poisons

The general idea underlying the use of heterogeneously lumped burnable
poisons is to take advantage of the strong self-shielding that will occur in such
poisons. Early in core life, the self-shielding will reduce the effective absorption
cross section of the poison, thereby minimizing reactivity mismatch. As the poison
depletes, however, the effective cross section of the poison relative to the fuel will
increase, due to self-shielding, which is the desired behavior in order to maintain
reactivity as high as possible near the end of core life.
To be a bit more precise, reconsider our equation describing the time rate of
change of the poison concentration
(14-26)
Here we have inserted a self-shielding factor J.(t) defined such that
average flux in poison

Js( t) = average flux in fuel · (14-27)
We have explicitly indicated that J.(t) will depend on time because of poison
depletion, which will reduce the self-shielding as the poison burns out (see Figure
14-10). Initially the poison burnout is relatively slow due to strong self-shielding.
However as poison depletion proceeds, self-shielding becomes less significant until
eventually f.(t)-1. Hence by using various self-shielded burnable poisons (e.g.,
boron or gadolinium), one can minimize the reactivity mismatch early in core life
and still avoid a large residue of unburned poison that would shorten core lifetime.
Burnable poisons will also affect the reactivity coefficients of the reactor. Since
most candidates for burnable poisons are 1/ v absorbers, the microscopic absorp-
tion cross section of the poison will be larger in a cold core. Hence a lumped
poison element will be more highly self-shielded and therefore less effective in a
cold core than in a hot core. This enhances the negative behavior of the tempera-
ture coefficient of reactivity.
Change in flux depression as poison burns out
'l::
eff
(t)
1.0 ---------------------
f.(t)
FIGURE 14-10. Effect of variable self-shielding on poison burn'hut. 18
IV. CHEMICAL SHIM
In light water reactors, boric acid is frequently dissolved in the coolant (to
concentrations of ~2000 ppm) to act as shim control. 17• 18 Such soluble poisons or
chemical shim have several advantages. Since the poison distribution is uniform
and independent of the amount of reactivity being controlled, the fuel loading can
REACTIVITY CONTROL / SSS
be more easily distributed to yield a uniform power distribution, such as by

zone-loading patterns. Chemical shim reduces the mechanical control rod re-
quirements quite considerably. Since such rods are expensive and occupy a sizable
fraction of the core volume, the elimination of mechanical control where possible is
desirable.
In Table 14-1, the relative mechanical and chemical control requirements of a
typical PWR are indicated. We have also sketched in Figure 14-11 the boron
concentration required to compensate for excess reactivity over core life in a
typical reactor design (comparing control requirements both with and without a
burnable poison).
2000---......--""T"------r---.----~----,-------r---r----,
\
\
\ Note: Hot, full power, rods out
i...
\.- .............
~
1600
..............
C
0 ..........
-~ 1200 .............
.,
C
',, ,/Without burnable poison
g
8 ',./'
......
i 800
.........
...... ......
......
°B .........
:€
u 400 With burnable poisons ',,,
......
......
Core average burnup (MWD/MTU)

FIGURE 14-11. Boron concentration versus first cycle burnup with and without burnable poison
rods. 11
A suitable soluble poison must be an isotope characterized by a large neutron

absorption section which is soluble in the coolant. It should be of a noncorrosive
nature and relatively stable so it will not adhere to core components. Boric acid
possesses these requirements when used in LWRs.
Chemical shim does have several disadvantages, however. Since the rate at which
one can inject or withdraw appreciable amounts of poison from the coolant is quite
small, the reactivity insertion rates are correspondingly small (with maximum
insertion rates of ~3 x 10- 5 /sec). Hence chemical shim is only of use to compen-
sate for relatively slow reactivity changes such as those due to fuel burnup or
conversion, fission product poisoning, and moderator temperature change (tem-
perature defect).
Chemical shim can have a major effect on the moderator void coefficient of
reactivity. We will find that in LWRs this void coefficient is usually quite negative,
since a decrease in coolant density leads to a decrease in moderation and hence
reactivity. However if a soluble poison is present, a decrease in coolant density will
also lead to a decrease in poison concentration-a positive reactivity effect. The
desire for a negative void coefficient will frequently limit the amount of chemical
shim allowed (typically to less than Ap~0.20 Ak/ k). We have shown this effect for
a modern PWR core design in Figure 14-12.
...-....
-
20
::!;1~
~o
~1~
"' ..,-"' 0
.,
C:
·.:;
:;::
.....,
8
~ -20
..,::,
.,e
a.
..,.,
E
..,ts -40
.,e
'C
0
::iE
-60
0 100 200 300

Moderator temperature (° C)
FIGURE 14-12. The effect of chemical shim on the moderator temperature coefficient in a
PWR. 11
There are other possible disadvantages, such as the possibility of sudden reactiv-
ity surges due to chemical shim that has been entrained in the core and suddenly
breaks loose to escape. Chemical shim can also interact in an unfavorable way with
fission product transients (xenon and samarium) toward the end of core life, as we
will indicate in Chapter 15. Nevertheless the advantages of chemical shim make it a
very attractive alternative to mechanical control in those reactor designs in which it
can be utilized.
V. INHERENT REACTIVITY EFFECTS
To complete our discussion of reactivity control, it is appropriate to consider

some of the more common mechanisms that give rise to short-term reactivity
changes in a nuclear reactor core. Perhaps the most important such mechanism is
the variation of reactivity with core temperature. As we saw earlier in Chapter 6,
this variation is usually expressed in terms of a temperature coefficient of reactivity
defined as the derivative of the core reactivity with respect to temperature
(14-28)
Here we have noted that more generally we must take into account the temperature
variation in the reactor by computing separate temperature coefficients characteriz-
REACTIVITY CONTROL / S57
ing the fuel, coolant, moderator, and so on. It should also be noted that these
coefficients are themselves functions of temperature.
Actually the more useful quantity is the power coefficient of reactivity ap that
describes the variation in reactivity due to a change in reactor power. This latter
parameter is directly related to the inherent stability of the reactor, and must be
kept negative for all conceivable operating conditions for a given reactor design.
However, ap is dependent on a very large number of different phenomena,
including not only core neutronic behavior, but thermal-hydraulic phenomena as
well. Thus in practice it is found to be far more convenient to analyze instead the
temperature coefficients of reactivity that characterize the various components of
the core. Of course, in principle, the power coefficient of reactivity could then be
determined from the a¥> as
(14-29)
provided one can perform a thermal-hydraulic analysis of the core to determine

the change in core temperatures ½ resulting from a change in power level.
Temperature variations in a reactor core will affect core multiplication, both
because of the resulting density changes in core components (due to expansion or
phase change) that change macroscopic cross sections, and because of a change in
the thermal motion of core nuclei which changes microscopic cross sections (e.g.,
the Doppler effect). Although many introductory treatments attempt to study this
phenomenon by calculating the isothermal temperature coefficient characterizing
each component of the six-factor formula, 23 we will examine the temperature
coefficients associated with each major component of the reactor core. In practice
it is sufficient to consider the reactivity coefficients characterizing only the fuel,
moderator (if any), coolant, and structure.
The two dominant temperature effects in most reactors are the change in
resonance absorption (Doppler effect) due to fuel temperature changes, and the
change in the neutron energy spectrum due to changing moderator or coolant
density (due to temperature, pressure, or void fraction changes). One must also
occasionally be concerned with core structure effects such as differential expansion
and rod bowing. Fortunately in most cases of practical interest, these physical
phenomena are sufficiently separable to allow an independent analysis of each
reactivity coefficient.
For example, suppose we assume that the dominant contributions to the tem-
perature coefficient of reactivity are due to Doppler-broadening of fuel resonances
and the change in the spectrum due to moderator density changes. If we note that
3p 1 3k 1 dk
aT = 3T = k 23 T ;;:,: k dT ' (14-30)
then we can write
(14-31)
Now changes in fuel temperature Tp do not effect the shape of the thermal neutron
energy spectrum. In reactor cores in which solid moderators are used (e.g.,
HTGRs), a change in moderator temperature will only affect the tlJ.ermal spectrum
and not the absorption rate in the resonance energy range, hence each temperature
coefficient can be analyzed separately.24 In liquid-moderated reactors (e.g., LWRs),
a change in moderator temperature will change moderator density significantly,
thereby influencing slowing down and hence resonance absorption. Even so, it is
customary to analyze af and ar separately, even when this interaction is present.
In this spirit, we will now examine each of the major reactivity feedback
tp.echanisms of concern in nuclear reactor analysis. We will then consider the more
practical problem of just how the corresponding reactivity coefficient for each
mechanism is calculated.
A. Fuel Temperature Coefficient of Reactivity

First we should recall from our study of resonance absorption that an
increase in fuel temperature will lead to Doppler-broadening of resonances with a
corresponding decrease in self-shielding, hence an increase in resonance absorption.
We can estimate the effect of such Doppler broadening on reactivity by recalling
that the resonance escape probability can be written in terms of the effective
resonance integral / as
(14-32)
Hence we can calculate the Doppler coefficient of reactivity in terms of the

resonance integral as
F 1 dk I dp I di
aT - k dT - - dT - Inp I dT . (14-33)
F p F F
We mentioned in Chapter 10 that a very useful correlation25 for the temperature-

dependence of the resonance integral was
I (T) = I (300°K)[ 1 + /3 (YT - V300°K ) l (14-34)
If we use this in Eq. (14-33), we find
(14-35)
Here /3 is a constant that depends only on fuel composition and geometry. For
example, for 238 UO 2 ,
(14-36)
232
while for ThO 2
(14-37)
In most thermal reactor types, the Doppler effect arises primarily from capture
resonances in fertile material in the fuel. For example, in slightly enriched uranium
cores, 238 U is the principal contributor early in core life, although 240 Pu becomes
important later in core life. In an HTGR, 232Th is of primary concern. The
contribution of other heavy metal isotopes such as 235 U or 239 Pu should be included
in estimates of Doppler coefficients, but their contribution is usually quite small in
low fissile-density cores.
This situation is modified somewhat in fast reactors, 26 since the neutron energy
spectrum spans the resonance regions of both fertile and fissile materials. Hence
increasing temperature will increase both La</> and Lt<f>. There may be a net increase
in reactivity. Since the Doppler effect is a fast feedback mechanism, and since the
prompt neutron lifetime of a fast core is so short, the sign of the Doppler
coefficient is extremely important.
The Doppler coefficient will tend to be more positive in a core with a hard
spectrum (such as in a metal-fueled core). To soften the spectrum, one uses oxide
fuels. Here the oxygen will degrade the spectrum to yield a negative Doppler
coefficient. Indeed several fast reactor core designs have actually considered add-
ing small amounts of moderator (such as BeO) to the core to soften the spectrum
and enhance the negative Doppler coefficient (although at considerable economic
penalty).
Although the reactivity change caused by the Doppler effect is frequently small
compared to other reactivity changes during normal operation, it becomes very
important for the postulated power excursions of concern in reactor safety analysis.
In such an excursion, rapid fuel temperature changes will lead to strong Doppler
feedback, which becomes the principal mechanism acting to terminate the power
excursion.
Since the Doppler coefficient depends sensitively on fuel composition, it will
change due to fuel burnup and hence must be monitored over core life. For
example, in a slightly enriched core, the Doppler coefficient becomes more negative
as a function of core life due to buildup of 240 Pu.
B. Coolant or Moderator Reactivity Coefficients
1. SOLID MOD ERATORS

In solid moderator reactor cores, such as the graphite-moderated HTGR, the
primary reactivity effect is due to a hardening of the thermal neutron spectrum
with increasing temperature, which causes a change in the thermal group constants.
The magnitude and sign of the associated moderator 1emperature coefficient
depends on a number of complicated interactions, which leads to a rather compli-
cated dependence on temperature (see Figure 14-13).
2. LIQUID-MODERATED REACTORS
The moderator coefficient of reactivity is a measure of the change in reactiv-
ity due to a change in moderator properties such as temperature, density, pressure,
or void fraction. When the moderator is a liquid also functioning as a coolant, such
as in a LWR, these reactivity coefficients may be quite sizable. We will consider
the reactivity coefficient associated with each property of the moderator.
(a) MODERATOR TEMPERATURE COEFFICIENTS
An increase in moderator temperature, keeping density constant, will lead to

a hardened neutron spectrum, resulting in increased resonance absorption in
3 ,------,----,------.----r----,r------,----,---,----r----,
2 Moderator
~
0 0
----
E
(.)
.3- -1
-----
~It
-
~1"""
...,
·c:;
C
a,
-2
-3
fa,
0 -4
(.)
~
...,::, -5
~
a,
C.
E a, -6
f-
-7
-8
-9
0 300 600 900 1200 1500 1800 2100 2400 2700
Temperature (' K)
FIGURE 14-13. Temperature-dependence of temperature coefficients of reactivity for a large

HTGR. 21
low-lying resonances in isotopes such as 238 U, 240 Pu, or 232Th. The hardened
spectrum will also cause an increase in the capture-to-fission ratio in 235 U and
239
Pu, hence a decrease in 11· Both of these effects render the moderator tempera-
ture coefficient more negative.
(b) MODERATOR DENSITY CHANGES
The dominant reactivity effect in water-moderated reactors arises from

changes in moderator density, due either to thermal expansion or void formation
(steam production). The principal effect is usually the loss of moderation that
accompanies a decrease in moderator density and causes a corresponding increase
in resonance absorption. 27 To see this more clearly, let us return to our expression
Eq. (14-32) for the resonance escape probability
(14-38)
where we have specifically identified the moderator density. Hence the void or
density coefficient of reactivity is just
M_ 1 dk _ ( 1 ) dNM
uT - k dTM -ln p dTM · (14-39)
Since dNM/ dTM will be negative and may be quite large, particularly if the coolant
is near saturation temperature, the void coefficient of reactivity can also be large
(see Figure 14-14).
-20.---,---r-----r--~-,------,----r-----r---,---,
-60
-100
-;;;
cJ
0
>
?f.
-140
Eu
..s-
~I~
~1 .... -180
>
<!j
-220
-260
-300_~-:-'::---::-:--:::--:'=----::'::-----::'::---=1-c--=-1,--.J...____,J
0 10 20 30 40 50 60 70 80 90 100
Voids%
FIGURE 14-14. Void coefficient of reactivity for a large BWR. 20
In most cases, the void coefficient for a water-moderated core will be large and
negative. However if appreciable chemical shim is used, this reactivity coefficient
becomes more positive since then a decrease in coolant density will cause a
decrease in poison concentration and hence macroscopic absorption cross section.
In order to achieve a negative moderator coefficient, one must limit the concentra-
tion of soluble poison.
(c) MODERATOR PRESSURE COEFFICIENT
The moderator pressure coefficient relates the change in moderator density

that results from a coolant pressure change, to a change in reactivity. Since most
reactor cores are operated at high pressure, this is a relatively small effect. For
example, a change of 4 bar in a PWR has approximately the same reactivity effect
as does a half-degree change in moderator temperature. 28
(d) COOLANT REACTIVITY COEFFICIENTS IN LMFBRS
In thermal reactors, the coolant void coefficient is negative, but in a large fast
core, the sodium void coefficient av may in fact be positive. That is, an increasing
temperature will increase the void fraction, hence increasing reactivity and thereby
temperature. There are two competing effects which determine the sign of av: (a)
spectrum effects, where the coolant degrades the spectrum, lowering the effective
ij; hence any loss of sodium hardens the spectrum, implying av> 0 and (b) leakage
effects, where loss of sodium near the edge of the core will increase leakage, hence
implying av< 0. These effects compete in such a way that the void coefficient is
positive in the center of the core and negative near the edge. One finds that this
effect becomes more pronounced with larger cores. One can design around this
effect by increasing leakage using a "spoiled" core geometry (e.g., pancake
geometries) but this will incur penalties in large fuel inventory and poor core
breeding ratios.
C. Core Expansion Effects

An increase in core temperature can cause an expansion of core structural
components with an attendant change in core configuration and hence reactivity. If
the core were to expand uniformly, the dominant effects would be an increase in
core size and coolant volume fraction. The latter would usually be the largest effect
and would tend to decrease reactivity in a fast core (in which such expansion
effects are most important).
Perhaps a more significant effect would be due to differential expansion leading
to fuel rod bowing. Since the bowing would usually be in the direction of a thermal
gradient, the fuel rods would bow in toward the center of the core, reducing
coolant volume fraction and increasing reactivity. Such bowing effects led to
instabilities in some of the early LMFBR core designs. 29 They can usually be
eliminated by careful design of the fuel element support structure.
A somewhat different expansion effect that occurs in metal clad fuel reactor
cores is differential fuel-clad expansion, which can lead to increased fuel-clad
contact pressure and hence increased gap thermal conductance. The net result is a
lowering of fuel temperatures and hence an increase in reactivity (although this is a
second-order effect).
D. Calculation of Reactivity Coefficients

Although the simple expressions we have developed for several of the tem-
perature coefficients characterizing a reactor core are useful for a qualitative
understanding of reactivity feedback effects, a practical calculation usually involves
performing a sequence of static criticality calculations for different fuel or mod-
erator temperatures, and then determining the appropriate reactivity coefficients
directly from a table of keff as a function of TF and TM or whatever other variable
is of interest. Such calculations may be greatly expedited if the relevant physics
determining the coefficient of interest can be isolated. For example, if the primary
effect is known to be dependent on a change in the thermal neutron energy
spectrum (such as the moderator temperature coefficient characterizing a HTGR),
then an infinite medium spectrum calculation may be used to determine the
reactivity coefficient.
There are a great many details involved in the accurate determination of
reactivity coefficients. For example, one must determine just what temperatures to
use in characterizing the fuel or moderator. Typically one uses a flux-weighted
temperature average in determining the Doppler coefficient of a fuel pin to take
into account the flux depression that results from self-shielding. Perhaps the most
complete discussion of such calculations is given in the Safety Analysis Reports
covering each reactor type, 15 • 16• 21 • 22 and the reader should refer to these reports for
further information.
For purposes of illustration, we have tabulated the principal reactivity
coefficients characterizing several of the more common reactor types in Table 14-2.
TABLE 14-2: Temperature Coefficients
BWR PWR HTGR LMFBR

Fuel-temperature coefficient
Doppler (pcm/°K) -4 to -1 -4 to -1 -7 -0.6 to -2.5
Isothermal temperature coefficient
Coolant void (pcm/%void) -200 to -100 0 0 -12to +20
Moderator (pcmj°K) -50 to -8 -50 to -8 +1.0
Expansion (pcm/°K) ~0 ~0 ~0 -.92
Temperature defect (%Ilk/ k) 2.0-3.0 2.0-3.0 0.7 0.5
Power defect (%Ilk/ k) 1.5-2.5 1.5-2.5 4.0 0.8
Xe worth (%Ilk/ k) 2.6 2.6 3.3 0.0
Sm worth (%Ilk/ k) 0.7 0.7 0.5 0.0
REFERENCES
1. A. Sesonske, Nuclear Power Plant Design Analysis, USAEC Document TID-26241

(1973), Chapter 6.
(1966), pp. 499-503.
3. H. W. Graves, Jr., Nuclear Reactor Design, University of Michigan Nuclear Engineering
4. J. R. Lamarsh, Introduction to Nuclear Reactor Theory Addison-Wesley, Reading, Mass.
(1966), pp. 509-517; R. V. Meghreblian and D. K. Holmes, Reactor Analysis, McGraw-
Hill, New York (1960). Ch. 11.
5. A. Radkowsky (Ed.), Naval Reactors Physics Handbook, Vol. I, Selected Basic Tech-
niques, USAEC Report TID-7030 (1964).
(1966), pp. 518-519; H. W. Graves, Jr., Nuclear Reactor Design, University of Michigan
Nuclear Engineering Department Lecture Notes (unpublished) (1969); A. F. Henry,
USAEC TID-7532, Part I (1957).
7. C. W. Maynard, in Naval Reactor Physics Handbook, Vol. 1, A. Radkowsky (Ed.),
USAEC Report TID-7030 (1964), pp. 409-447.
8. M. Michelini, Nucl. Sci. Eng. 42, 162 (1970).
10. H. W. Graves, Jr., Nuclear Control Management, University of Michigan Nuclear
Engineering Department Lecture Notes (unpublished) (1969).
I 1. RESAR, PWR Standard Safety Analysis Report, Westinghouse (1973) Chapter 4.
12. D. C. Wade, Trans. Am. Nucl. Soc. 11, 314 (1968).
13. A. Radkowsky (Ed.), Naval Reactors Physics Handbook, Vol. I, Selected Basic Techni-
ques, USAEC Report TID-7030 (1964), p. 800.
14. H. W. Graves, Jr., Nuclear Reactor Design, University of Michigan Nuclear Engineering
15. BWR/6 Standard ·Safety Analysis Report, General Electric (1973).
16. Reference Safety Analysis Report, Westinghouse Nuclear Energy Systems (1973).
17. P. Cohen and H. W. Graves, Jr., Nucleonics 22, 75 (1964).
18. H. W. Graves, Jr., Nuclear Control Management, University of Michigan Nuclear
Engineering Department Lecture Notes (unpublished) (1969).
19. CESSAR, Standard Safety Analysis Report, System 80, Combustion Engineering (1973).
20. BWR/6 Standard Safety Analysis Report, General Electric (1973).
21. Preliminary Safety Analysis Report, Summit Power Station, Delmarva Power (1973).
22. Preliminary Safety Analysis Report, Clinch River LMFBR Demonstration Plant, Project
Management Corporation (1975).

(1966), pp. 451--466.
25. E. Hellstrand, P. Blomberg, and S. Horner, Nucl. Sci. Eng. 8, 497 (1960).
26. D. Okrent and H. Hummel, Temperature Coefficients of Fast Reactors, Am. Nucl. Soc.,
Hindsdale, Ill. (1968).
(1966) p. 458; H. W. Graves, Jr., Nuclear Reactor Design, University of Michigan
Nuclear Engineering Department Lecture Notes (unpublished) (1969).
28. Reference Safety Analysis Report, Westinghouse Nuclear Energy Systems, Vol. 4 (1973).
29. T. J. Thompson and J. G. Beckerley, The Technology of Nuclear Reactor Safety, Vol. I,
PROBLEMS
14-1 Calculate the effective control cross section ~~ff characterizing a circular control cell
geometry in which the control rod is of radius a and the cell is of radius R. The
transport parameter a for this geometry is given in ANL-58009 as
;\= _I =0. 7 I04+ 0.2524 + 0.0949 + ...

3a ~;a (~;a/
14-2 Demonstrate that for a perfectly black absorbing slab, the transport correction
parameter is a= 0.47.
14-3 Determine the effective control cross section ~~ff for the slab geometry control cell
described in Section 14-11-B if the control rod is sufficiently weakly absorbing that
diffusion theory is valid. (That is, calculate a using diffusion theory in the control
rod.)
14-4 Compare a for a black rod (a=0.47) and the diffusion theory description (Problem
14-3) with the transport theory result given by Eq. (14-13) for various values of ~aa.
14-5 Suppose the shim rods of a power reactor are ganged together in such a way that
they can be treated as a uniformly distributed absorption cross section inserted into
a cylindrical core to a depth h. Using one-group perturbation theory, determine and
sketch the relative distance of insertion of the rods as a function of time over the
core life.
14-6 The measured thermal flux at the surface of the scram rods of a low power research
reactor is approximately 3 X 1012 neutrons/ cm2 /sec. The rods are stainless steel with
2% (by weight) boron. How long will it take to produce 90% burnup of the boron at
the rod surface. (oa for naturally occurring boron at 0.025 eV is 796 barns).
14-7 Repeat the calculation of the effect of rod-bank insertion on the axial power
distribution [i.e., resolve Eq. (14-16)] for partial length control rods of active length/,
l<H.
14-8 Show that in a critical reactor, a small change in soluble poison concentration
produces a change in reactivity that is approximately equal to the fractional change
in poison concentration multiplied by the fraction of neutrons absorbed in the
poison.
14-9 A reactor is to operate at constant power. The reactor is initially loaded with excess
fuel plus a burnable poison. Neglecting fission product poisoning, determine: (a) the
poison concentration as a function of time and (b) the reactivity as a function of
time if it is zero at startup.
14-10 A reactor that is highly enriched in 235 U and contains a burnable poison is required
to operate under the conditions that the excess reactivity never falls below p 1 and
never rises above p2 • Assume that a homogeneous model can be used to describe the
reactor. Develop expressions for both the initial poisoning ratio and the maximum
fluence for a core life t1•
14-11 From the attached two-group constants for a typical low-power research reactor find
the total worth of the shim rod using a two-group diffusion code. Run the code for
four intermediate rod positions and plot the rod worth versus depth of insertion.
Using perturbation theory, calculate the rod worth versus position, and compare
with these results. Also plot the peak/ average power ratio as a function of insertion
depth and explain your results. (Model the core as a slab of width 50 cm and treat as
a "windowshade" problem.)
The two-group constants for the core without the shim rod are:
Group D P~f ~f ~.,2 ~a

1 1.44440 0.001865 0.000763 0.028743 0.001905
2 0.24712 0.0780 0.0321 0.052200
When the shim rod is inserted, ~a2 = 0.053606; all other group constants remain
unchanged.
14-12 Suppose that it were possible to choose a burnable poison whose effective micro-
scopic cross section, o%(t) were a controllable function of time (e.g., by careful design
of self-shielding burnout). What time-dependence would be required to yield a
uniform excess reactivity over core life?
14-13 Calculate the isothermal temperature coefficient of reactivity for the thermal and fast
nonleakage probabilities aT(PTNd and aT(PFNd, assuming that the primary tem-
perature effect is core expansion. Assume a thermal expansion coefficient of f3 for
the core.
14-14 Consider a LWR fueled with slightly enriched uranium.
(a) Indicate how, for small void fractions, to calculate approximately the void
coefficient of reactivity, i.e., the fractional change in multiplication per fractional
change in void.
(b) In PWRs the water will frequently contain boron in solution. Indicate the effect
this is likely to have on the void coefficient and explain.
(c) What is the effect of water temperature (at a given pressure) on the results of (a)
and (b) above.
14-15 If a large BWR has a zero moderator density coefficient at operating temperature
(Ptt 20 =0.8 g/cm 3) and zero power, what will the sign of the coefficient be: (a) at full
power with 10% void fraction and (b) at refueling temperature (pH 2o= I g/cm3)?
Give reasons for your answers.
15
Analysis of Core Composition Changes
The atomic densities of various isotopes in a reactor core are continually

changing due to nuclear processes such as fission, neutron capture, and radioactive
decay. For example, fission events will reduce the concentration of fissile isotopes
such as 233 U, 235 U, or 239 Pu. Neutron capture will reduce fertile nuclide concentra-
tions while building up the concentrations of transmutation products such as 239 Pu.
In addition we must remember that fission reactions will produce a large variety of
fission product nuclides, most of which are radioactive and will subsequently decay
into other isotopic species.
It is extremely important to monitor the isotopic composition in the core during
reactor operation, since changes in core composition can affect core multiplication
as well as flux and power distributions. Certainly the prediction of fuel depletion
and conversion in the core is essential for the determination of fuel loading
requirements. Furthermore certain of the fission product nuclei (or their progeny
following radioactive decay) are characterized by extremely large absorption cross
sections and hence may significantly affect the reactivity of the core. For example,
135
Xe has an enormous thermal neutron cross section of oa = 2.7 X 106 barns. A very
small amount of such fission product "poisons" may significantly affect the
multiplication of the reactor, since as they accumulate they will absorb neutrons
from the chain reaction.
The analysis of core composition changes is complicated by the fact that the time
and spatial variation in isotopic composition will depend on the neutron flux
distribution-which itself depends on core composition. Fortunately changes in core
composition occur relatively slowly (on time scales of hours, days, or even months)
so that the reactor core can always be kept in a critical state by control element
adjustments. This means that although a time-dependent analysis of the rate
566
ANALYSIS OF CORE COMPOSIDON CHANGES / 567
equations describing changes in isotopic number density is necessary, the neutronic

behavior can be studied by performing a sequence of instantaneous static criticality
calculations for each core composition encountered. Hence although the analysis of
core composition changes does involve time-dependent phenomena, the neutronic
part of the analysis involves only static calculations of core flux distributions and
hence differs considerably from the treatment of nuclear reactor kinetics of
Chapter 6.
Our intent in this chapter is to examine the procedures useful for analyzing such
time-dependent changes in core composition. We will first examine short-term
composition changes due to fission product buildup that affect the operation of the
reactor. In particular we will apply these methods to an analysis of the very
important problem of xenon and samarium transients in thermal power reactor
cores.
We will then continue to consider the longer term process of fuel depletion and
develop methods for core depletion analysis. This will allow us to examine the
variables determining core lifetime and fuel loading requirements. Finally, we will
illustrate these methods by considering the very important problem of in-core fuel
management.
I. FISSION PRODUCT POISONING

Certain fission products possess extremely large thermal neutron absorption
cross sections. Of particular concern are 135Xe and 149Sm, whose absorption cross
sections 1 are shown in Figure 2-4. The significance of these extremely strongly
absorbing fission products is compounded by their relatively large fission yields.
The buildup of such fission product poisons can appreciably affect core multiplica-
tion and hence reactor operation. It should be noted, however, that since these
absorption cross sections fall off quite rapidly for neutron energies above 1 eV,
fission product poisoning is primarily of concern in thermal reactors (and only of
secondary importance in fast reactor operation).
In order to illustrate the importance of this phenomenon, let us make a very
simple estimate of the reactivity change induced in a reactor core by the addition
of a fission product poison. To first order, the effects of such poisons enter through
the thermal utilization f. If we imagine introducing a poison characterized by a
macroscopic cross section ~!' uniformly throughout a homogeneous reactor core,
then we can write the thermal utilization characterizing the "poisoned" core as
( 15-1)
Hence we can compute the reactivity change from a critical reactor in which ~!' = 0
as
(15-2)
Now if we recall our earlier definition of reactivity as given by the six-factor

formula
k-1
p=-k-, (15-3)
and note that
( 15-4)
will also change as ~: changes, then we can compute
f' - f PTNL - PTNL

6.p= -f' + P' (15-5)
TNL
If we use Eqs. (15-1) and (15-4), we find
( 15-6)
and
( 15-7)
Hence
( 15-8)
Here we have noted that the second term in brackets is usually quite small
( < 10- 3) and can be ignored. In fact for many large power reactor cores, leakage is
sufficiently small that L 2Bg2 « 1 and to first order, the reactivity change due to such
a poison is just the fraction of the total macroscopic absorption cross section due to
the poison. (Here we must keep in mind the crude nature of this estimate which has
totally ignored any spatial variation in the poison concentration-a variation that
we will later find is rather important.)
Now recall that the macroscopic cross section for the poison is given by
~: = Npa:, where Np is the number density of the poison while a: is its thermal
absorption cross section. Hence in order to estimate the reactivity change due to
fission product poisoning, we must calculate the number density of the poisoning
isotope Np at any time t.
The procedure is simple-at least in concept. To determine Np(t), we must solve
the rate equations describing the various production and decay processes that can
affect the poison concentration. We will illustrate this procedure with two simple
(but nevertheless important) examples involving the buildup of 135Xe and 149Sm.
A. Xenon Fission Product Poisoning

135
Xe is the most significant fission product because of its enormous thermal
neutron absorption cross section and its relatively large fission yield. Actually 135Xe
ANALYSIS OF CORE COMPOSffiON CHANGES / 569
can be produced not only directly as a fission product but may also result from the
/3-decay of 135 Te. A portion of the production decay scheme of the A= 135 chain is
shown in Figure 15-1. Fortunately this rather complicated decay scheme can be
considerably simplified by assuming that the decay of 135Te to 1351 is instantaneous
(i.e., that 135 1 is produced directly as a fission product). Furthermore we will also
ignore the short-lived isomeric state 135mXe, and assume that all 135 1 nuclei will
decay directly to the ground state 135 Xe. Hence the effective decay scheme we will
study is shown in Figure 15-2.
Thus, we now must write isotopic rate equations similar to those discussed in
Chapter 2 to describe the atomic densities of 135 1 and 135 Xe. Let us denote the
atomic number densities of these two isotopes as / (r, t) and X (r, t) respectively.
Furthermore let y 1 and Yx denote the effective fraction of the fission products
which are 1351 and 135 Xe, while A1 and Ax are the /3-decay constants for these two
isotopes. Using these parameters, we can then write the coupled rate equations
describing the simplified decay scheme illustrated in Figure 15-2 as
Iodine
of = Y1Lrc/>(r, t) - A,(r, t). ( 15-9)

'dt direct iodine
from fission decay
Xenon
(15-10)
direct iodine xenon xenon
from fission decay decay absorption
Here the macroscopic cross sections and neutron flux are to be interpreted as
one-group averages. Notice that we have included a loss term in the xenon balance
equation to account for the fact that capture will deplete the 135 Xe concentration
(as well as the neutrons, of course). Strictly speaking, we should have included a
similar term in the iodine equation, but the absorption cross section of 1351 is
sufficiently low that such a term would only be significant for extremely high flux
levels (c/> > 10 16 cm - 2 sec- 1). The values of the yields and decay constants are
tabulated in Table 15-1 2• 3 •
IT
15.3 min
13-
1.7 19.2 6.58hr 9.17 hr
sec ~ - - ~ sec ~~-~ (91%) ~~-~
a!</>
FIGURE 15-1. A portion of the decay scheme for A= 135.
~-----136Xe
a/¢
FIGURE 15-2. A simplified decay scheme for 135 Xe.
TABLE 15-1: Fission Product Yields and Decay Constants
Fission Decay
Product Yields 233u 235u 239pu 24Ipu constants
Y1(%) 4.884 6.386 6.100 7.694 A1=0.1035 hr-I
Yx(%) 1.363 0.228 1.087 0.255 Ax= 0.0753 hr- I
Yp(%) 0.66 1.13 1.9 Ap=0.0128 hr-I
Before we can proceed further to solve these equations, we need information

about the flux in the reactor </>(r, t). Suppose that we were to merely assume that the
flux behavior was a known function of space and time. Then we could immediately
integrate the iodine equation to find
(15-11)
and then substitute this solution into the xenon equation and integrate it to
determine the 135Xe concentration as a function of space and time:
!
X (,, t) ~ X (,, 0) + [ dt'[ A, I(,, t') + Yx:i:,(,, t')~(,, I') ]exp[ f dt" [Ax+•:~(,, t") i]]
Xexp[-£ dt 1
11
[Ax+a~<f>(r,t")] l (15-12)
Of course, for any complicated flux behavior, these solutions will require numerical
integrations, and until we specify the form of <f>(r, t), this formal solution is of little
use in understanding the general behavior of the xenon concentration in the
reactor. Hence we will examine the time-behavior of the xenon concentration for
several particularly simple examples of flux behavior.
l. STARTUP OF A CLEAN CORE REACTOR
Suppose that we suddenly bring the reactor to a steady-state flux level </>o at
time t = 0. We will assume that prior to this time, the reactor core has been in a
ANALYSIS OF CORE COMPOSITION CHANGES / 571
shutdown state with zero fission product poison concentrations (i.e., a "clean"
core). We need only substitute a time-independent flux </>(r, t) = </> 0(r) into our
general solutions, Eq. (15-11) and (15-12), and use the initial conditions /(r,0)=0
= X (r, 0) to find the time behavior of the fission product concentrations following
startup:
(15-13)
and
(15-14)
The qualitative time behavior of these concentrations following a clean startup is

shown in Figure 15-3. In particular, it should be noted that these concentrations
eventually level off at equilibrium levels for long times following startup:
(15-15)
That is, the concentrations of these fission product poisons in a reactor operating at
constant flux will eventually saturate at those equilibrium values for which the
production of poisons from fission is just balanced by the decay and neutron
capture losses of the poisons. These equilibrium concentrations could also have
been determined directly from the rate equations (15-9) and (15-10) themselves by
merely setting dX / dt = 0 = di/ dt.
The equilibrium concentration X 00 is of interest because we can substitute it into
Eq. (15-8) to determine the negative reactivity which the xenon will contribute
under steady-state power conditions.
(15-16)
In a critical reactor in which leakage can be ignored, k = (v"2.r/'i:.a)PE = 1 implies

that "2.r/"2.a=(vpE)- 1 ~1/v and therefore
(15-17)
,,
q,(r, t)
q,o-----------.
-~---------......-----t
I(t)
X(t)
x~ -------~-------f
- - " - - - - - - - - - . . _ __ _ _ _ _ t
to
135 135 Xe
FIGURE lS-3. Qualitative behavior of 1 and concentrations following a cold, clean
startup and then shutdown.
Notice this reactivity will depend on the flux level cf, 0• However for large flux levels
characteristic of those in power reactors,
"A.
cf,0 » ~ =0.756X 10 13 cm- 2 sec- 1, (15-18)
(Ja
and the xenon reactivity will approach a maximum value of
(15-19)
This amounts to 6.p=::. - .026 for both 233 U and 235 U-fueled reactors, a rather sizable
reactivity. This, of course, must be accounted for in determining excess reactivity
and control system requirements.
2. REACTOR SHUTDOWN
Suppose now that after operating the reactor for a long period of time at a
constant flux level cf,0 , we suddenly shut the reactor down. An examination of the
rate equations (15-9) and (15-10) reveal several consequences of setting the flux cf,
suddenly equal to zero. First the removal of 135Xe due to neutron capture ceases,
ANALYSIS OF CORE COMPOSmON CHANGES / 573
leaving 135Xe decay as the sole removal mechanism. However the production of
135
Xe via decay of 135 1 will continue. Since the half-life of 1351 is shorter than that of
135
Xe, the 135Xe concentration may initially build up before decaying out. To study
this more explicitly, suppose we solve the rate equations (15-9) and (15-10)
characterizing the shutdown reactor:
aJ
at = - >-.if (r, t),
(15-20)
ax =A1J(r,t)->--xX(r,t)
at
subject to the initial conditions that at the time of shutdown (t = 0), the poison
concentrations have attained their equilibrium values J 00 and X 00 • The iodine
equation is easy to solve
(15-21)
One can now insert this into the xenon equation and integrate to find
If we use the explicit forms for X 00 and J 00 in Eq. (15-8), we can find the negative
reactivity introduced by the 135Xe buildup as
Some typical values are shown plotted in Figure 15-4. The maximum value of
negative reactivity depends quite sensitively on the flux level prior to shutdown. In
fact unless
(15-24)
no buildup of xenon following shutdown will occur. However since this ratio is
quite small (4X10 11 and 3X10 12 cm- 2 sec- 1 in 235 U- and 233 U-fueled reactors,
respectively), a xenon transient buildup will occur following shutdown in most
power reactors.
We can calculate the time at which the maximum negative reactivity occurs as
(15-25)
where we have also noted the value assumed by tmax in the limit of large flux levels
0.6
0.5
0.4
0.3
0.2
0.1
30 40 50
t [hr l
FIGURE 15-4. Negative reactivity due to 135 Xe buildup following shutdown at t = 0.
cf,0 • In particular, one should note that for a period of about 30 hours after reactor
shutdown, a rather sizable reserve of positive reactivity would be needed to restart
the reactor. This, of course, is a major consideration in scheduling shutdown for
short periods of time.
In large power reactors the xenon transients following shutdown are mitigated to
a degree by several factors. 5 First the neutron fluxes in large power reactors are
somewhat lower than in research or propulsion reactors (typically cf, 0 ;:S 10 14 cm- 2
sec- 1). Furthermore the xenon buildup can be minimized by avoiding an abrupt
shutdown and instead programming the shutdown in a gradual manner to burn out
some of the xenon while shutting down. Finally, the existence of a power defect of
reactivity means that the reactor can usually be brought to a low power level
following shutdown and xenon buildup in order to burn out some of the xenon
before increasing the flux to full power levels.
3. XENON TRANSIENTS FOLLOWING POWER LEVEL CHANGES

Let us calculate the xenon transient following a change in reactor flux level
from cf,0 to cf, 1• If we assume that this change occurs at t = 0, we can solve the
isotopic rate equations (15-9) and (15-10) as an initial value problem with ct,- np 1 0
and / (0) = / 00 , X (0) = X 00 • In particular, we find that following the power-level

change
(15-26)
and
(15-27)
As we have sketched in Figure 15-5, a flux level increase is followed by a xenon

concentration decrease and vice versa. These xenon transients following flux level
changes are characterized by times of the order of eight hours. Such xenon
transients induce reactivity transients, which although slow enough to be compen-
sated for by shim control, can nevertheless be troublesome.
Fot example, suppose we consider a PWR operating at some nominal power
level toward the end of core life. Since the fuel will be relatively highly depleted,
,,
 (t)
</Jo
I(t)
X(t)
I
I I
1------------1
I X=, I
I I
t0 t1
135
FIGURE 15-5. Variation in Xe concentration following power level changes.
the amount of chemical shim present will be small. However since the rate at which
reactivity can be changed is proportional to the soluble poison concentration, the
reactivity response dp/ dt will also be small. Now suppose the reactor power level is
suddenly reduced. Then, as we have just seen, the xenon concentration will begin
to build up. One would then desire to decrease the chemical shim concentration to
compensate for the negative reactivity introduced by the xenon transient, and there
may be insufficient chemical shim present to allow this. One encounters a similar
situation following a power increase, since now chemical shim must be rapidly
inserted into the coolant to compensate for xenon burnup. Such considerations can
place strong limitations on the allowable chemical shim concentrations in the core
design.
B. Samarium Fission Product Poisoning

A very similar analysis can be given for 149Sm, which is also characterized by
large fission yields (see Table 15-1) and absorption cross sections (a!=58,500 b).
The appropriate decay scheme is
Fission ~ 149
Nd
2.0 hr 54 hr
Again, it is consistent to neglect the 149Nd and assume fission yields 149Pm directly.
The corresponding rate equations are then
Promethium:
(15-28)
Samarium:
aa~ = ApP (r, t)- a!<f>(r, t)S (r, t). (15-29)
The equilibrium concentrations can again be obtained by setting aP /at= 0

= as/at to find
(15-30)
The corresponding negative reactivity due to saturated samarium poisoning is then
'YP
~p:::: - - p ~ 0.00463. (15-31)
We can again study the shutdown behavior using P oci and Soci as initial conditions.
Notice that since 149Sm is stable (unlike 135Xe), it can only be removed by neutron
capture. Hence after shutdown it will build up to a steady level
S(t)=Soci+Poci(l-exp-Apt) - Soci+Poci. (15-32)

(--Ht)
ANALYSIS OF CORE COMPOSmON CHANGES / S77
0.05
0.04
0.03
--"-
-:::.
o<1a
-;; [ 1 + ~s (1-exp-Apt) l. (15-33)
Typical plots of 6.p(t) are given for a 235 U-fueled reactor in Figure 15-6. The much
longer time scale characterizing 149Sm buildup should be noted. Indeed it takes
roughly one month for the 149Sm concentration to approach saturation levels in a
core.
The smaller absorption cross section and yield fraction and longer precursor
half-life characterizing samarium imply that it will lead to far less severe reactivity
effects than 135Xe. Nevertheless, 149Sm buildup must be accounted for in nuclear
analysis.
C. Other Fission Product Poisons

There are a great many fission product nuclides other than 135 Xe and 149Sm
that have appreciable neutron absorption cross sections. However the cross sec-
tions of these other isotopes are not sufficiently large in most cases for their
concentration to be depleted by neutron capture. Hence these fission products do
not saturate but rather tend to accumulate in a core indefinitely. Furthermore the
fission yields of these isotopes are not as large as those of 135Xe or 149Sm. It has
become customary6 to account for these nonsaturating fission products in critical-
ity calculations by lumping them together into several effective groups
characterized by effective cross sections and yields. (For one-speed thermal calcu-
lations, one frequently assumes that each fission produces a permanent poison
a;
nucleus with effective absorption cross section ~50 b.) We will return later in
this chapter to discuss how such considerations enter core lifetime and burnup
studies.
Since the absorption cross sections of all nuclides drop off rapidly with increas-
ing neutron energy, neutron capture will not appreciably affect fission-product
concentrations in fast reactors. Hence all fission products can be lumped into
nonsaturating groups characterized by effective absorption cross sections in fast
reactor analysis.
D. Spatial Effects in Fission Product Poisoning
1. A PERTURBATION THEORY ESTIMATE OF REACTIVITY WORTH
In our earlier estimate of the reactivity change caused by fission product

poisoning, we merely computed the changes in the thermal utilization as the poison
concentration built up. However such an estimate is actually only valid for a
uniform concentration of poison in an infinite core. More generally, we know that
the poison concentration will depend on the spatial variation of the flux in a finite
core, and hence our estimate of reactivity worth of the poison is only of limited
validity. We can improve this estimate by using perturbation theory.
Let us compute the reactivity due to the equilibrium xenon concentration in a
core operating with a steady-state flux cf> 0(r). Then we know
(15-34)
If we treat this as a perturbation in the absorption cross section, 8~a ➔~~(r), then
we can use the first-order perturbation result given by Eq. (5-306) to find the
reactivity change due to the xenon as
(15-35)
This expression can be evaluated numerically to compute the equilibrium reactivity

worth of the xenon poisoning in a core. Note that for large power levels, cf> 0 »"J\x/
a:, this reduces to just our earlier approximate result for an infinite medium
(15-36)
For low flux levels,

reactor)
c/>0 «Ax/ a:, in an uniform core (e.g., a low-power research
6-p ➔ (15-37)
</>o«Ax/ a';
For example, in a slab, one finds that in this limit
(15-38)
where 6.p 00 is the reactivity change characterizing an infinite medium model. That
is, the reactivity worth is some 33% larger when the spatial variation of the flux is
taken into account. In a low-power reactor core, this spatial weighting factor is
more typically between 2.5 and 3.0 (see Problem 15-12).5
2. XENON-INDUCED POWER OSCILLATIONS IN LARGE POWER

REACTORs1- 10
Thus far we have ignored the effect of the reactivity introduced by the fission
product poisons on the neutronic behavior of the core. The interaction between
xenon fission product buildup and the changes in the neutron flux distribution that
accompanies local changes in reactivity can lead to spatial oscillations m the power
distribution in a large thermal reactor core.
To explain this phenomenon, let us first consider a point-reactor model in which
the reactor has been operating at a steady-state flux level for a period of time. Then
our earlier study of xenon fission product poisoning has indicated that there will be
a saturation in the 135Xe concentration resulting from the balance of 135Xe
production (via direct fission and 1351 decay) and loss due to decay to 135Cs and
transmutation (via neutron absorption) to 136Xe. (Recall Figure 15-2.) Now suppose
a small perturbation increase in the flux occurs. Then 135Xe will transmute more
rapidly to 136Xe (instantaneously), depleting the 135 Xe concentration, hence de-
creasing the absorption and increasing the reactivity and the flux. However the
increased flux transmutes even more 135 Xe, and hence the initial flux perturbation
grows with time (unstable). Such an instability can only exist for power levels
higher than a certain threshold value. For 235 U-fueled reactors, this threshold is
3 X IO" neutrons/ cm2 • sec. Below this threshold the stabilizing effect of the direct
xenon yield from fission is more important than the destabilizing effect of the
xenon decaying from 135 1.
Actually this type of instability is relatively unimportant in practical reactor
operations, since it is easily controlled by normal control rod movement. A much
more serious spatial xenon instability can arise, however, which requires a more
complex control rod program. To understand this, consider the very simple model
illustrated in Figure 15-7 consisting of two coupled xenon-unstable point reactors,
separated by a region of nonmultiplying material. Suppose further that there is a
control system keeping the total power of all three core regions constant (although
the flux or power in an individual region is not constant).
A slight increase in the power level on one side of the core will give rise to the
unstable xenon process described for a point reactor. Since the control system
keeps the total power constant, the flux on the other side decreases. This process
continues with an increasingly steeper tilt in the flux resulting. Two effects will
FIGURE 15-7. A simple model of xenon-induced power

oscillations.
limit this tilt: (a) burnup of most of the xenon on the high flux side and (b) the
steep flux tilt which creates a flux gradient and hence a current that carries all the
excess neutrons being produced to the other side. The flux will remain tilted for
several hours. Eventually the high flux side will have created an 1351 concentration
much greater than that originally present. Since the decay half-life of 135 1 is 6.7
hours, more xenon will be created after this delay period. Similarly on the low side
less xenon will be created. This reverses the flux tilt eventually and produces a
side-to-side oscillation with a period of 15-30 hours. Thus the xenon process tends
to be self-limiting and produces the effect of a moving "hot spot" to the reactor
operator.
The oscillation of the power distribution in a reactor can cause changes in the
power-peaking factor Ft However this can be easily accounted for in design to
ensure that thermal performance constraints are not exceeded. Such xenon-induced
power oscillations are more of an operational than a safety problem.
Although such oscillations can be controlled by control rod motion, there is
strong motivation to design the reactor core in such a way that xenon oscillations
are minimized, since the load-following requirements of a power reactor operating
on a utility power grid imply that such control adjustments may incur considerable
economic penalty. In general, negative feedback mechanisms such as moderator
void formation or the Doppler effect will tend to suppress xenon-induced power
oscillations. For example, the very large negative void coefficient in a BWR will
usually suppress the oscillations within one or two cycles.
The stability of the spatial power distribution with respect to xenon-induced
oscillations will decrease with increasing core size or decreasing neutron migration
length. As a rule of thumb, xenon oscillations will be a problem if the reactor core
is over 30 migration lengths in size. 10 Since this is the case with most large power
reactor cores, xenon oscillations can represent .a serious design and control con-
sideration. Strong reactivity feedback, such as the moderator void coefficient in a
BWR, will help to suppress such oscillations, however. In addition, partial length
control rods can be used to damp out power oscillations by suitable rod program-
ming.
II. FUEL DEPLETION ANALYSIS
A. Introduction
Fuel depletion analysis is concerned with predicting the long-term changes in
reactor fuel composition caused by exposure to neutron flux during reactor
operation. Such changes have an important bearing on the operating life of a
reactor, as well as on its stability and control. One must first ensure that the shift in
the core power distribution that accompanies fuel burnup does not result in the
exceeding of core thermal limitations. Sufficient excess reactivity must be provided
in the fresh core loading to achieve the desired fuel exposure (consistent with safety
limitations). And, of course, a detailed analysis of core composition is necessary in
order to optimize fuel exposure to achieve minimum power costs as well as to
determine the value of discharged fuel. Since the fuel costs over the operating
lifetime of the reactor can exceed those of the capital cost of the plant itself, the
incentive for accurate analysis of fuel depletion is quite high.
A variety of nuclear processes must be monitored during a depletion study.
These include, of course, the consumption of fissile nuclides (fuel burnup ). How-
ever one must also account for the conversion of fertile isotopes into fissile isotopes
and the production of numerous fission products. Finally, one must monitor the
reactivity balance to ensure core criticality (usually by determining the change in
reactivity over a period of core operation and then adjusting control to compensate
for this reactivity change).
A complete burnup calculation would involve the solution of the coupled
reaction rate equations describing such processes for the hundreds of different
types of isotopes in a reactor core. In practice, one usually introduces approxima-
tions that greatly simplify these calculations. For example, the only fission products
that are usually treated explicitly are those with large capture cross sections and
fission yields, namely, 135Xe and 149Sm. The remaining fission products are lumped
into one or several groups, each characterized by an effective cross section.
Furthermore any nuclide with a short half-life is omitted from a burnup calcula-
tion.
The specific isotopes monitored depend on the particular fuel used in the core.
The isotopes of interest in both uranium- and thorium-fueled reactors are indicated
in Figure 15-8. In addition to fission products and the other isotopes produced
during fuel irradiation, one must monitor concentrations of burnable poisons such
as boron or gadolinium.
The general procedure is to solve the time-dependent rate equations describing
isotopic concentrations together with the static multigroup diffusion equations
describing the spatial flux profile in a reactor core that is kept critical by control
adjustments while its core composition is changing due to fuel burnup and isotope
production. As we indicated earlier, this analysis can become quite involved since
one also may have to account for changes in core thermal-hydraulic behavior that
accompany changes in the power distribution, as well as implement a prescribed
control management scheme to maintain a critical core operating at rated power
over core life. Hence before we consider in detail how such practical depletion
studies are performed, we will first consider several extremely simple examples to
illustrate the general ideas involved in such studies.
Uranium-fueled cores:
Arn241
f3 l(l3y)
Pu239(n, Y )Pu240(n, y) Pu24l(n, Y )Pu242
/3 l(2.4d)
Np231 Np239
/3 l(6.7d) /3 I (23 min)

u2Js( n, y) u236( n, y) u231 u2Js(n, y) u239
Thorium-fueled cores:
u233(n, y) u234
/3 l(27d) /3 !(6.7hr)
Pa23J(n, y) Pa234
/3 !(22rnin)
Th232(n, y) Th233
FIGURE 15-8. Isotope chains of interest in fuel depletion analysis.
582 / AN INTRODUCilON TO NUCLEAR REACTOR CORE DESIGN
B. Some Simple Examples of Fuel Depletion
l. THE DEPLETION OF A SINGLE ISOTOPE
Suppose a reactor were fueled with only a single isotope, characterized by an

atomic number density Np(r, t) and one-group absorption cross section
the fuel burnup would be described by the simple rate equation
a:.
Then
(15-39)
Now if the flux were known in advance, we could easily solve this equation to find
(15-40)
or in terms of the neutron fluence, (r, t), defined by Eq. (14-24),
(15-41)
Of course, this is still a formal solution since the flux <f>(r, t) depends on the fuel
density N p(r, t) through the equations describing reactor criticality (i.e., the mul-
tigroup diffusion equations).
We can use this result to illustrate the two standard approximations used in more
elaborate depletion studies:
(a) CONSTANT FLUX APPROXIMATION:
Assume that over the time interval of interest, the flux can be treated as
constant
</>(r, t) = </>0 (r). (15-42)
Then we can solve for
(15-43)
(b) CONSTANT POWER APPROXIMATION:
One assumes instead that the reactor power is constant
(15-44)
where wa is the energy released per neutron absorption in the fuel ( wa= ( af / a:)wr)-
Then one can integrate
(15-45)
to find
(15-46)
The isotopic depletion resulting from the assumptions of constant flux or

constant power are quite different, as shown in Figure 15-9. However for short time
intervals, these solutions are similar since for small t, Eq. (15-43) becomes
Np(r, t)::::: N p(r, 0) [ 1 - <J!q,0(r)t]

P0 (r)
=Np(r,0)- - - t . (15-47)
wa
This is significant because while we cannot rely on the assumption of constant

power or flux over times as long as a core lifetime, we can use these approxima-
tions over very short time steps, since both yield the same result as a short-time
Taylor expansion of the isotopic density
aNF I
Np(r,t)=Np(r,O)+tat t= +···
0
P 0 (r)
:::::Np(r,0)- - - t . ( 15-48)
wa
That is, these two types of approximation will be used to characterize the flux or
power over a time step or "depletion step" in a finite-difference solution of the
time-dependent isotopic rate equations.
t
FIGURE 15-9. Depletion of a single fuel isotope in the constant flux or power approximation.
2. FUEL DEPLETION IN A SIMPLE REACTOR MODEL 11

Of course, we cannot decouple fuel depletion from reactivity effects if the fuel
concentrations change appreciably. In general, one must consider both the
neutronic behavior and control adjustment of the core as fuel depletes over core
life. It is of interest to illustrate the ideas involved in the solution of the isotopic
depletion equations and control adjustment by considering a simple model in
which the neutronic analysis is bypassed by assuming an infinite one-speed reactor

characterized by a multiplication k = r,J. That is, we assume a simple model of an
infinite, homogeneous, thermal reactor with a single fissile isotope (e.g., 235 U) and
no fertile material. We will represent the control rods used to balance the excess
reactivity of the initial core loading by a uniformly distributed absorption cross
section Lc(t). The criticality condition for this simple model is just
(15-49)
We can solve this for the required control
(15-50)
Now as the core life proceeds, the fuel cross section L!'\t) decreases, while the
poison concentration increases. Hence Lc(t) will drop to zero eventually at some
time t1, corresponding to the end of core life. It is possible to calculate t1, provided
the power history of the reactor is known.
For simplicity, we will suppose that the reactor is operated at constant power
over the core life. This implies
(15-51)
where L!'(O)cp(O) characterizes the initial core loading. After a time t,
Np( t) = N p(O) - L!' (O)cp(O)t
= Np(O)[ 1- a!'cp(O)t ], (15-52)

or
(15-53)
If we substitute this into (15-51), we can solve for the flux
cp(O)
cp(t)---- (15-54)
1 - a!'cp(O) t
(Note how cp(t) must increase as the fuel is depleted in order to yield a constant
power production.)
We can now use this flux history to compute the fission product poison
concentrations. Recall from our earlier work that the equilibrium fission product
concentrations are given by:
Lx( t) = ( Y1 + Yx)LrCf> = ( Y1 + Yx)Lr(O)cp(O)

(15-55)
a Ax/ a;+ cp Ax/ a;+ cp( t)
for xenon, and
(15-56)
for samarium. [Note here that Lr(t)=L~(t)/(l+a).]

We must also worry about non-saturating or "permanent" poisons
(15-57)
Substituting Eqns. (15-53) to (15-57) into our expression for the control cross
section, Eq. (15-50), we find
( y1+ Yx)Lr(0)cf>(0)
Lc(t)= (1J- l)L~(0)[ 1-a;cf>(0)t ]-L~-
Ax/ a:+c/>(t)
- ypLr(0) [ 1 - a;cf>(0) t] - appLr(0)cf>(0)t. (15-58)
If we now substitute Eq. (15-54) for cp(t), and then set Lc(t1)=0, we can solve for
the core lifetime t1 as
Ax
for c/>( t)« x, (15-59)
(Ja
or
'IJPex(l +a) - ( Y1 + Yx + Yp) Ax

//~------------------ for cf>( t)» - , (15-60)
[ ('IJ-1)(1 + a)a;-(y 1+ Yx+ Yp)a; + app]c/>(0) ax
a
where Pex = k 00 (0)- 1/ k 00 (0) is the initial excess reactivity. Hence as we might
expect, the core lifetime is essentially proportional to the excess reactivity loading
Pex· Indeed we must choose an initial excess reactivity sufficiently large to com-
pensate for fission product buildup,
Y1 + Yx + Yp Y1 + Yx + Yp
Pex> 1J(l+a) = (15-61)
V
before criticality can be achieved. And of course the core lifetime 1s inversely
proportional to the operating power or flux level cp(0).
C. Calculational Methods of Depletion Analysis 12• 13

We have noted that the analysis of fuel depletion involves a variety of nuclear
processes. These can be roughly classified into calculations involving: (a) solution
of the isotopic depletion equations, (solving for the nuclei densities as functions of
time and position; this requires a knowledge of the neutron flux) and (b) solution
of the static multigroup diffusion equations for the neutron flux (this task will
usually also involve including thermal-hydraulic coupling, using as input the cross
sections characterizing the depleting isotopes, and adjustment of the control
representation to achieve core criticality). It is customary to decouple these calcula-
tions such that the depletion equations are solved over time intervals in which the
flux or power is assumed to be constant. At the end of each time interval, the
depleted densities are used to calculate new group constants, and the multigroup
diffusion equations are solved to determine a new flux shape for the next time
interval.
1. ISOTOPIC DEPLETION EQUATIONS
The reaction rate equations describing the number densities of nuclei in the
core can be derived using simple balance ideas. If NA (r, t) is the number density
characterizing a nuclide of type A, then the general rate equation characterizing a
production-type decay scheme as shown below
~ A
C
/
takes the form
(15-62)
Here
A.ANA loss due to radioactive decay of A,
[ ; CJ~$g ]NA loss due to neutron capture by A,
>--aNa gain due to decay of B to A,
[ ; a~$g ]Ne gain due to transmutation of C to

A via neutron capture.
The fluxes and microscopic cross sections appearing in these equations are mul-
tigroup averages and must be generated by suitable group-constant generation
codes and multigroup diffusion calculations.
These equations are nonlinear and inhomogeneous because the fluxes and
microscopic cross sections not only vary in space and time, but depend on the
densities of the depleting isotopes as well. Such equations must be written for each
species of interest. In a detailed study of fuel depletion in a modern power reactor,
some 15-24 heavy nuclides and 25-50 fission products are required for the
analysis. 12 When one adds to this the fact that the differential equations describing
isotopic depletion contain both long and short time constants (half-lives or large
absorption cross sections) that complicate their numerical solution, it is apparent
. that fuel depletion calculations can become quite involved.
2. GENERATION OF GROUP CONSTANTS

A very important facet of any depletion calculation is the generation of the
necessary group constants. These are required not only for the multigroup diffusion
equations, but also for the isotopic rate equations. Our earlier discussion of a
typical macroscopic cross section module illustrated several aspects of such calcula-
tions. We recall that these group constants consisted of two factors: (a) the number
densities supplied from the depletion module (or perhaps the thermal-hydraulic
module or input) and (b) the microscopic group constants generated either by fast
and thermal spectrum codes or by parameterization (where one of the parameters
may be the burnup (MWD/MTU) or fluence):
~x (r, t) = L N1-(r, t)a{ (r, t). (15-63)

• j •
The microscopic cross sections will change in time as the fuel depletes, since the
neutron spectrum in the fuel cell will change, as will the effect of self-shielding.
Many cross sections are sufficiently slowly varying with fuel depletion that they
need only be recalculated at intervals of several time steps and can be interpolated
at intermediate times. More rapidly varying cross sections which exhibit strong
depletion-dependence of self-shielding factors or spectrum will require adjustment
at shorter time intervals.
3. FLUX-POWER CALCULATIONS
One must use these few-group constants in the multigroup diffusion equations
to determine the flux and power distribution in the core. Interaction with a control
adjustment module is usually necessary in order to adjust core multiplication back
to critical after a depletion time step. Such flux-power calculations may range from
zero- to three-dimensional descriptions, and have already been described in detail
in Chapter 13.
4. SOLUTION OF THE DEPLETION EQUATIONS

The direct solution of the coupled equations describing the number densities
of core materials and the multigroup diffusion equations describing the neutron
flux is impractical. Instead one must separate these calculations. The coupling
between the depletion and neutronics calculations can be handled in one of several
ways:
(a) SPATIAL TREATMENT:
The multigroup diffusion equations must be solved many times as depletion

proceeds. Such solutions become very expensive if detailed spatial treatments are
required. One can distinguish between two different classes of studies:
(i) Macroscopic burnup physics:
In macroscopic studies one is concerned with determining the global depletion
behavior of the reactor core. In particular, such studies seek to determine the
long-term behavior of reactivity variation (and the core lifetime), the core power
and burnup distributions, and total core fuel material inventories.
(ii) Microscopic burnup physics:
In microscopic studies one is more concerned with an accurate calculation of
reaction rates and fuel composition as functions of burnup at a point or in a cell
representative of a given region of the core. That is, microscopic studies are usually
concerned with the depletion analysis of a lattice unit cell. Such microscopic
calculations are used to provide the homogenized or effective cross sections needed
for the macroscopic global depletion studies. 14
S88 / AN INTRODUCTION TO NUCLEAR REACTOR CORE DESIGN
(b) TIME TREATMENT:
The treatment of the time-dependence of the depletion and the changes in the
flux distributions is usually handled by separating the depletion calculation from
the neutronic calculation. To illustrate this approach, suppose that one knows the
number densities in the clean, fresh core at time t = 0. Then the macroscopic group
constants can be generated, and the multigroup fluxes can be calculated (although
a control adjustment will be required to achieve core criticality).
Next the depletion equations are solved for time O< t < D.t by assuming that over
this time step either:
Constant flux for O< t < D.t:
Constant power density for O< t < D.t:
~r
g
(r, t)</>g(r, t) = L. N.J (r, t)a{g </>(r, t) = ~rg (r, 0)</>(r, 0). (15-64)
J
(One usually applies the constant flux approximation over the depletion step, since
the assumption of constant power density is difficult to apply for more than one
fissile isotope.) Under these assumptions, the isotopic depletion equations (written
here in matrix notation for convenience)
dN
dt = A N(t)+ F(t) (15-65)
can be solved (in the constant flux approximation for purposes of illustration) as
N(t+D.t)=(exp Al::i.t) N(t)+ A _'(exp AD.t-1) F(t). (15-66)
Of course one must choose the depletion-time step size D.t such that the variation of
the flux over this time step is negligible. The solution of the depletion equations for
a given time step can also be obtained by representing the system as a linear
combination of linear decay chains and then using the analytic solutions for a
given linear chain. 15
Having determined the number densities N; at the later time t = D.t, one can now
generate new multigroup constants (either by using parameterized cross sections or
repeating spectrum calculations), and then use these to calculate a new multigroup
flux distribution. Once again, control adjustment will be required to achieve core
criticality, as will flux normalization to maintain the required core thermal power
output.
This scheme can then be continued for another time step, D.t < t < 21::i. t (which
may have a variable size), and so on until the excess reactivity of the core drops
sufficiently low that core life is terminated.
A variety of depletion codes have been used in the analysis of core burnup.
These range from zero-dimensional codes such as LEOPARD 16 used for survey
calculations (such as those to determine the approximate fuel enrichment required
to yield a desired burnup) to two- and three-dimensional codes such as PDQ-
HARMONY17 or CITATION, 18 which are used in final design calculations to give
detailed nuclei densities and power distributions useful for the study of control
management.
III. NUCLEAR FUEL MANAGEMENT
We have stressed that the primary source of the economic advantages

enjoyed by nuclear power over conventional power sources lies in the substantially
lower fuel costs of nuclear power stations. An extremely important facet of nuclear
engineering involves determining the parameters of both the initial reactor core as
well as subsequent reload cores to minimize fuel costs. This topic is known as
in-core fuel management, and it is perhaps the primary motivation behind depletion
studies of fuel burnup. It is important to at least indicate the various considerations
that arise in this particular aspect of reactor core design. However such a discus-
sion is most appropriately imbedded within a brief overview of the nuclear fuel
cycle in order to more properly understand the significance of each of the design
variables that must be optimized by the nuclear engineer. 19-21
A. An Overview of the Nuclear Fuel Cycle

The various steps in obtaining, processing, using, and disposing of nuclear
fuel are shown schematically in Figure 15-10 for a uranium fuel cycle21 (thorium
fuel cycles are very similar since they also must use a fissile-fuel feed involving 235 U
or 233U):
(l) Mining: Both underground and open-pit techniques are used to mine
uranium ore in a manner similar to that used in other low-grade ore
mining. The ore obtained from mines in the United States averages only
about 0.25% uranium oxide (U30 8).
(2) Concentration: Milling is necessary to extract uranium oxide from the
raw ore. The ore is pulverized and leached with sulfuric acid to dissolve
the U 30 8. One then uses solvent extraction or ion exchange to recover
the dissolved U 30 8. Calcination (roasting) is then performed to produce
yellow cake, a crude oxide containing some 70--90% U 30 8. Yellow cake is
the form of uranium most commonly traded in commodity markets.
Further calcination and solvent extraction is used to refine the yellow
cake to essentially pure U0 3.
(3) Conversion: Hydrogenation is first used to convert U0 3 to U0 2. Then
reacting U0 2 with hydrogen fluoride produces UF 4, which can then be
converted into a gas, uranium hexafluoride, UF 6, by adding fluorine salt.
Such a gaseous form is most suitable for isotope separation and enrich-
ment. Uranium hexafluoride is chosen because it is a gas at a lower
temperature than any other uranium compound.
(4) Enrichment: Essentially all power reactors (with the exception of certain
heavy-water reactors or the early gas-cooled, graphite-moderated reac-
tors) utilize enriched uranium, that is, uranium with higher than the
natural 0.7% concentration of 235 U (at least in their initial core loading).
The enrichment of uranium is a very difficult and expensive process
since it involves separating two isotopes, 235 U and 238 U, with very little
mass difference and essentially no chemical difference. A variety of
separation techniques have been used or proposed:
(a) Electromagnetic separation: Huge mass spectrometers were used
for isotope separation during the Manhattan project, but this
scheme was rapidly replaced by gaseous diffusion techniques.
(b) Gaseous diffusion: Since the diffusion coefficient for a gas to
0 Mine
0 Mill
Depleted
Uranium tails
Plutonium Blend Fabrication ®
Spent fuel
assemblies assemblies
Valuable
isotopes
® Waste
Energy
FIGURE 15-10. Fuel cycle flow diagram.21
pass through a porous membrane is inversely proportional to the

square-root of its mass, one can pass the UF 6 gas through
thousands of porous barriers to separate the UF 6 (2 35 U) from the
UF6 (238U).
(c) Ultracentrifuges: This simply involves use of very high-speed
centrifuges26 or nozzle devices to separate the two isotopic forms
of UF 6 •
(d) Laser excitation27 : Of more recent consideration are schemes
that use high-powered lasers to selectively excite 235 U in uranium
vapor or UF 6 by tuning the laser wavelength to select the isotopic
mass shift in the electronic or vibrational energy levels of the
mixtures, thereby selectively exciting (i.e., "tagging") one isotopic
species in preference to another. A number of techniques then

can be used to segregate the excited species including photoioni-
zation followed by electrical separation or chemical separation.
The gaseous diffusion method has been the primary technique used for
uranium enrichment for the past 30 years. Recently, however, develop-
ments in ultracentrifuge design have indicated that these latter methods
may, in fact, be superior to gaseous diffusion techniques. Most recently,
advances in the development of high-powered tunable lasers has given
rise to a large scale research effort in laser separation techniques.
(5) Fabrication: Following enrichment, the UF 6 is chemically converted into
the form to be used in the fuel element. This is usually a ceramic such as
U02 or UC. The resulting ceramic powder is then compacted (and
sintered) into small pellets, which are then loaded into metallic tubes
(the clad) or into the fuel-moderator matrix (as in the HTGR).
(6) Fuel burnup in the reactor core: The fuel assemblies are loaded into the
reactor core for power production. These assemblies are typically
irradiated in the core for a period of several years. The fuel burnup or
lifetime can be limited by either reactivity considerations (i.e., the
multiplication of the core drops too low for further power production) or
by limitations on material stability under high radiation fluences. The
latter limitation usually determines the lifetime of fuel in today's high-
burnup designed cores.
(7) Spent fuel storage and decay: After being irradiated in the core, the fuel is
intensely radioactive due to fission product buildup. The used or spent
fuel is removed from the core and stored in water pools for several
months to allow the short-lived fission products to decay out.
(8) Reprocessing: The spent fuel is then shipped to reprocessing facilities to
reclaim unused uranium (which can then be recycled back as UF 6 for
conversion or reenrichment) and plutonium.
(9) Waste disposal: The radioactive waste products remaining after repro-
cessing are then converted into either liquid or solid forms for storage
and are shipped to various depositories for burial (and surveillance).
The above outline is a description of only the uranium-plutonium fuel cycle used
in most modern power reactors. However other possible fuel cycles have been
considered for future reactors, such as the thorium- 233 U cycle proposed for HTGR
operation. These latter fuel cycles are quite similar to the uranium fuel cycle we
have considered above, since they usually employ 235 0 as fissile feed in addition to
recycled 233 U.
The management of the various activities involved in obtaining, irradiating, and
disposing of fuel materials is referred to as nuclear fuel management and is a
principal concern of nuclear engineering. Such activities must be performed subject
to a number of important constraints. Of course, one desires to minimize electrical
generation costs, but must also ensure that the safety of the reactor is not
compromised-for example, the fuel temperature must always be kept below
melting and the control margin must be maintained within safe limits.
The management of the processes involved in the nuclear fuel cycle can be
classified into three types of activity: (a) head-end fuel management (mining,
conversion, enrichment, and fuel fabrication), (b) in-core fuel management
(evaluation of reactivity and control requirements, power distribution analysis, and
core capability evaluation), and (c) tail-end fuel management (fuel storage, ship-
ping, reprocessing, and waste disposal). We will consider only in-core fuel
management, since it most directly involves nuclear reactor analysis. However it is
first useful to consider in greater detail the various fuel cycles of interest in power
reactor operation.
B. Power Reactor Fuel Cycles
1. LWR FUEL CYCLES

Fuel requirements for a typical LWR of the 1000 MWe class are shown in
Figure 15-11 along with a LWR fuel cycle diagram in Figure 15-12.21 We recall
that such reactors are fueled with enriched uranium (from 2-3 w / o-weight per-
cent). The spent fuel is discharged from the core with a 235 U concentration of
roughly .8%. The conversion of 238 U results in a plutonium concentration of about
the same magnitude. This plutonium is recovered by fuel reprocessing.
The recovered plutonium can then be recycled back into the L WR fuel cycle, as
shown in Figure 15-13. This reduces the requirements for uranium feed by
approximately 33%. There are several technical problems involved in plutonium
recycling, 22 including the complications of fabricating plutonium (a very toxic and
radioactive material) and matching the nuclear performance of a mixed oxide fuel
(Pu0 2 and UO 2) with enriched uranium fuel in a core. Such problems do not seem
formidable, however, and the large inventory of plutonium that is beginning to
accumulate from the existing generation of large power reactors (coupled with the
lagging commercial development of the fast breeder reactor, which would provide
an alternative market for this plutonium) provide strong incentives for implement-
ing plutonium recycle in thermal power reactors.
2. HTGR FUEL CYCLES
The fuel cycle of the HTGR is quite different from that of light water
reactors since it is based on a 232Th-233 U conversion cycle.23 As Figure 15-14
indicates, the initial loading of the HTGR core contains highly enriched uranium
( ~93% 235 U) in the form of uranium carbide along with thorium oxide or carbide
as the fertile material. The thorium is then converted to 233 U, which is recovered
and recycled back (eventually) to be mixed into reload fuel. The motivation behind
using such a fuel cycle is primarily due to nuclear considerations. The number of
neutrons per thermal neutron absorbed in 233 U , 'T/ 23 , is some 10% larger than that
characterizing 235 U. Hence it is possible to achieve much higher conversion ratios
using 233 U (indeed, such "advanced converter" reactors are capable of operating
with a conversion ratio as high as 0.8 compared with 0.6 for an LWR). The higher
core temperature of the HTGR favors the use of 233 U over 239 Pu as a reload fuel,
since 'T/ 49 tends to decrease with increasing core temperature.
As we have seen, the fuel design of the HTGR is much different from that of
either the LWR or the LMFBR, in order to facilitate high temperature operation.
The fuel composition is of small pellets of UC or ThO 2 coated with graphite and
loaded into graphite blocks. Such a fuel design leads to much higher fabrication
costs, however, and hence the fuel elements of an HTGR must be run to much
higher burnups (some 80,000 MWD /MTU as compared to 30,000 MWD /MTU
for a LWR or 100,000 MWD /MTU for the LMFBR).
The higher conversion ratio of the HTGR leads to a more efficient utilization of
Ore U30s UF6
I
85,500 tons/yr
:> I
162
:>
Mine Milling , =G
•~-l
Enriched
u - 150 :>
1
l•.I.Ir ~l f ❖i
:::1.
J
Fuel UF6
< 36 < 53 I
o.,,:. "'"'":
tails storage (not
Power reactor Fabrication Enriching UF6 required for reactor
but must be stored
safely. has value for
future breeder reactor blanket)
OIIWL
High-level
Spent fuel solid waste
36 7
Reprocessing Federal
repository
Low-level
wastes -~~rnU'FJLtt<JiZU'.S~ .·..' :

50
~
FIGURE 15-11. Uranium requirements for a 1000 MWe PWR. [The Nuclear Industry, VSAEC Report WASH-1174-73 (1973)].
~
Recovered
plutonium, 792 g/d
3.3% Spent
Fuel u235 fuel Fuel Fission
PWR reprocessing i-----,► products
preparation UO 2
Enriched U F6
Natural
uranium Natural
Recovered
concentrates
470 kg/d
Uranium
conversion
UF6
0.711%
'>---------' UFs, 0.8% u2Js
u235
....__ _ _ _ _ _ _ _~Depleted UF
6
0.2% U235
FIGURE 15-12. LWR fuel cycle (no Pu recycle). 21
Recovered plutonium, recycled
2.1% Spent
Fuel u235 fuel Fuel Fission
preparation PWR i---.....,.► products
UCi+ reprocessing
PuO2
Enriched UF6
Natural
uranium Natural Recovered
concentrates Uranium UF6 UF6 , 0.5% LJ235
conversion >-E-------'
300 kg/d 0.711%
u235
...._________ --,► Depleted UF

6
o.2% u23s
FIGURE 15-13. LWR fuel cycle (Pu recycle). 21
Recovered uranium, recycled
Thorium uc.i Spent Thorium

concentrates Fuel + ThC 2 fuel Fuel +
preparation HTGR reprocessing fission
products
Enriched UF6 , 93% U 235
Natural
uranium
Uranium
concentrates Natural
refining +
conversion UF6
....__ _ _ _ _ _ _ _........,,► Depleted UF

6
0.2% u235
FIGURE 15-14. HTGR fuel cycle.21
594
ANALYSIS OF CORE COMPOSffiON CHANGES / 595
uranium. In fact, an HTGR requires only about half the fuel feed U 3O 8 of an LWR
of similar rating. 21
3. LMFBR FUEL CYCLES

The fuel cycle of the fast breeder reactor differs from that of the L WR and
HTGR in that the larger conversion ratio allows a net gain in the amount of fissile
material produced during core operation. It is estimated that the LMFBR can
utilize some 60-70% of the energy available in natural uranium (taking account of
the transmutation of 238 U to 239 Pu). The more efficient utilization of available fuel
resources is the primary justification for the LMFBR, although there are side
benefits such as higher coolant temperatures (and hence lower waste heat release)
than for the LWR.
A typical fuel cycle for the LMFBR is shown in Figure 15-15, in which it is
recognized that most of the transmutation of 238 U into 239 Pu will occur in the
blanket, while the fuel burnup and power production will occur in the core. The
LMFBR can use as fertile material either natural uranium or depleted uranium
from the tails stream of the isotope-separation plants.
Net
product
Recovered plutonium, recycled plutonium
~ 1000 g/d
+
~ 15% Pu
Core PuO2 Reactor
preparation Spent
core
fuel
Natural
or
depleted I
uranium
4-8 kg/d Neutrons iii! Fuel
reprocessing
Fission
products
Blanket Reactor
preparation blanket
Irradiated
.___ _ _-----J blanket u
U + Pu
Recovered uranium, recycled
FIGURE 15-15. LMFBR fuel cycle.21
4. INTERACTION AMONG FUEL CYCLES

It should be recognized that there will be a very substantial interaction among
these various fuel cycles in a mature nuclear power industry. For example, the
substantial amounts of plutonium produced by LWRs can be either directly
recycled or used to fuel the first generation of fast breeder reactors. Similarly the
excess plutonium produced by the LMFBR can be used to fuel either future
LMFBRs or fed back into the LWR fuel stream.
There will also be strong interaction with the fuel cycle of advanced converter
reactors such as the HTGR, which can be used to greatly stretch out the usefulness
of available fuel, without, however, having to carry the high fissile fuel inventory of
true fast breeders (since the relatively dilute fuel of the HTGRs can be run at
around twice the fissile rating typical of fast breeder reactors). 24
These interactions will determine to a very large degree the cost of the nuclear
fuel cycle, since they not only influence the cost of the fresh fuel loading, but also
determine the value of the spent fuel (i.e., the value of the plutonium produced in
the core). Such interactions are quite complicated and no doubt will lead to even
greater uncertainty in the future fuel costs of nuclear power reactors. It is apparent
from this discussion, however, that the real incentive for developing new types of
reactors such as the LMFBR is based almost entirely on expected improvements in
the economics of their fuel cycles, rather than changes in their capital costs or
operating costs.
TABLE 15-2 Enrichment Requirement of Various Reactor Types

(Thousands of kg SWU (or MT SWU) and short tons of U 3O 8 for 1000 MWe Power
Plants at 0.25% Tails Concentration)
BWR PWR HTGR LMFBR

First core requirement
U-235 Enrichment, % 2.0 2.6 93
MTSWU 229 233 347 24ot
STUp8 580 500 420 (70)
Annual reload requirements
at 80% plant capacity factor
U-235 Enrichment, % 2.7 3.2 93
Early years-no recycle
MTSWU 101 109 133 12ot
ST U 30 8 170 180 160 (35)
Later years-full recycle
MTSWU 62 73 82 0
STU 0 110 110 JOO (1-2)
tswu equivalent of fissile Pu required: I kg Pu fissile"" 100 kg SWU
C. In-Core Fuel Management
I. INTRODUCTION
(a) GENERAL FUEL MANAGEMENT CONSIDERATIONS 20
Perhaps the most important responsibility of the nuclear engineer concerned

with the nuclear fuel cycle is the management of the fuel in the core-including the
selection of refueling schedules, fuel loading patterns, and the prediction (and
measurement) of fuel burnup and isotope buildup, in an effort to satisfy both
nuclear and economic constraints. In this section we will consider only a few of the
many facets of in-core fuel management.
It is during fuel irradiation that many of the debit aspects of fuel economics
occur (e.g., in the depletion of fissile material such as 235 U). However the in-core
portion of the fuel cycle is also where all of the credit aspects such as plutonium
production occur.
The nuclear engineer would like to achieve minimum fuel costs consistent with
nuclear and thermal constraints on the core performance. To this end, he must
determine the optimum distribution of fuel enrichment for both the initial and
reload cores. The refueling period and reload strategy must be specified, along with
the control management of both movable rods and shim (soluble and burnable
poisons). Hence the nuclear engineer has considerable flexibility (and considerable
responsibility) in managing the incore utilization of the fuel.
(b) CONSTRAINTS
The task of determining the optimum arrangement of fuel assemblies and

loading is extremely complex, involving the core geometry and composition,
neutronic-thermal-hydraulic coupling, reactivity control, and operational require-
ments. In a typical core design, the core size, fuel assembly design, control devices,
and coolant conditions will be fixed before the detailed in-core fuel management
scheme is determined. The variables at the nuclear engineer's disposal are then the
enrichment of fresh fuel assemblies, the frequency of fuel reloading, the reloading
pattern, and the associated control management program. A number of constraints
on in-core fuel management must be considered.
We noted earlier that the two principal thermal limitations on core performance
involved the requirements that the fuel centerline temperatures remain below
melting point, and that the surface heat flux not exceed its DNB limit. Such
requirements place limits on the local power density (in the case of the fuel
centerline temperatures) and also on the integral of the power density or coolant
enthalpy up each coolant channel (in the case of the critical heat flux limit). Such
limitations must be kept in mind when determining a fuel arrangement, since the
core must be capable of operating at its rated power level throughout the reactor
cycle without exceeding these thermal limitations. Since the core power distribution
will change over core life as fuel depletion occurs, one must be able to predict the
local power density throughout the cycle. This is an extremely expensive and
time-consuming aspect of fuel management.
The reactor control system must always be capable of controlling the excess
reactivity loaded into the core to compensate for fuel depletion. Although this
excess reactivity is usually larger in the initial core loading (since all of the fuel is
fresh), one can use burnable poisons and/ or chemical shim to adequately control
it. In later reload cores, there are fewer alternatives available, since, for example,
burnable poisons cannot be easily inserted into irradiated fuel assemblies. Such
considerations impose serious upper limits on the allowable fuel burnup of a core.
The reactor should always be capable of responding to load-demand changes.
Since the control rod worth is dependent on the flux profile, and since this flux
profile will change over core life, one must ensure that adequate reactivity response
is available for such power maneuvering.
Fuel burnup is also limited to a considerable degree by the amount of radiation
damage the fuel can withstand without experiencing appreciable probability of
failure. For example, fuel irradiation can induce swelling of the fuel, which places a
strain on the cladding. Furthermore, the buildup of gaseous fission products within
the element can cause a stress on the clad, which leads to a plastic deformation of
the fuel rod at high temperature (high-temperature creep). The enormous thermal
gradients across the fuel pin and cladding, combined with the variations in
temperature which occur during reactor startup and shutdown (thermal cycling)
can also lead to clad fracture in the intense radiation environment of the core. Such
fuel failure limitations on fuel burnup are particularly significant when very high
burnups are required, such as in the LMFBR.
There are frequently external requirements placed on the scheduling of core
refueling. For example, one wishes to avoid a reactor shutdown for refueling during
periods of peak power demand (e.g., in the middle of the summer or winter
months). In general, more frequent refueling leads to lower fuel inventory re-
quirements (less excess reactivity is required for shorter core lifetimes), but at the
expense of more reactor down time and enhanced thermal cycling.
2. REFUELING SCHEDULES
Rarely does one replace the entire core of a reactor in a refueling operation;
only a fraction of the core is replaced with fresh fuel at any one time. The time
period between such refuelings will be referred to as a reactor cycle, and in any
partial refueling scheme, the complete fuel cycle consists of a number of such
reactor cycles.
In general, the fuel inventory required for a given energy production decreases as
the frequency of partial refueling increases. However, there is a tradeoff here, since
more frequent core refueling can lead to increased shutdown time and hence
increased costs due to power outage. For some time now, most reactors have been
refueled on a yearly basis (although it should be noted that the pressure-tube
design of the CANDU or SGHWR type reactors allows continuous or very rapid
refueling). However recently alternative refueling cycles have been proposed.
For example, the development of rapid refueling systems for PWRs make
possible refueling times of three to four days (compared to present down times of
several weeks). 28 Such rapid refueling allows semiannual refueling at six-month
intervals, with considerable savings in fuel inventory (estimated at some 10% of the
annual fuel cycle costs). The tendency for BWRs has been to move in the direction
toward longer reactor cycles, with refueling now proposed on an 18-month basis.
This refueling frequency appears economically attractive to the BWR because of
the ease with which its fuel lifetime can be stretched by adjusting coolant void
fractions. 29
Annual refueling schedules of LWRs tend to replace 1/4 to 1/3 of the core
during each refueling operation. However, shifting to a semiannual schedule will
allow the replacement of only 1/ 6 of the core at a time. The higher conversion
ratio of the HTGR allows a refueling scheme in which only 1/6 of the core is
replaced each year. This longer fuel residence time is facilitated by the coated
particle fuel design for the HTGR which is capable of much higher burnups.
3. REFUELING PATTERNS
We now turn our attention to a consideration of how the fresh fuel is
distributed in the reactor core during refueling, and just which spent fuel elements
are withdrawn (assuming that partial refueling schemes will be used). Suppose we
attempted to load the core with a uniform enrichment. Then, the flux peaking at
the center of the core would lead to higher burnup of these fuel elements with a
considerable flattening of the core power distribution toward the end of core life
(see Figure 15-16). Unfortunately, such a uniform arrangement leads to a larger
power-peaking factor F"!: early in core life. It also results in a very low burnup of
fuel elements inserted toward the edge of the core.
ANALYSIS OF CORE COMPOSITION CHANGES / S99
P(r) i----
Beginning of core life
R r
End of core life
P(r)I-----
R r
FIGURE 1S-16. Shift in power distribution due to nonuniform burnup for batch refueling.
Hence a nonuniform core loading is clearly desirable-in both the initial core
and subsequent replacement cores. There are two common refueling schemes. 13•20
(a) ZONAL LOADING (IN-OUT CYCLING):
In this scheme, one loads unirradiated fuel in a zone on the periphery of the
core. The irradiated fuel is shuffled in toward the inner zone, while the fuel in the
central zone is withdrawn from the core. The motivation in this scheme is to make
use of the inherent reduction in reactivity which accompanies fuel depletion as a
power-flattening mechanism. By way of example, a three-cycle zonal pattern is
shown in Figure 15-17. Such a pattern can even be implemented on the initial core
loading by using fuel assemblies with varying enrichments.
Zonal loading does have disadvantages, however. In large cores subjected to high
burnups, appreciable distortions in the flux distribution can arise which could lead
to large power-peaking factors.
(b) SCATTER (ROUNDELAY) LOADING:
An alternative scheme loads fuel in a scatter or random pattern to achieve a

more uniform fuel distribution. One still will find some power-flattening since the
average burnup in the central zones of the core is more pronounced than near the
Zonal loading Combination scatter-zonal loading
FIGURE 15-17. Alternative core loading schemes.
core periphery. The power distribution characterizing a scatter-loaded core has a

somewhat flattened form of the distribution characterizing a uniformly loaded
core. There is a fine-scale ripple in the local power density, however.
Scatter-loading has the additional advantage that no reshuffling of irradiated
fuel is required.
In practice, most refueling schemes involve some combination of zonal and
scatter loading techniques. For example, many PWRs are loaded in such a fashion
that fresh fuel is loaded into the outer zone, while the fuel irradiated in the outer
zone is scatter loaded into the center zones.
The detailed analysis and optimization of in-core fuel arrangement is a very
complicated process. One must take into account core geometry, control manage-
ment schemes, the coupling between core neutronics and thermal-hydraulic per-
formance, and the constraints placed on core performance by operational re-
quirements. Although this aspect of fuel management is still frequently performed
by trial and error, using liberal dashes of physical intuition, there is now more of a
tendency to attempt to implement more direct optimization methods 12 (e.g., linear
programming) for in-core fuel management.
However, a very big headache with any elaborate fuel management scheme is the
flexibility frequently required of a refueling program. For example, an unexpected
alternation of a refueling pattern may be required to remove a fuel assembly
leaking fission products which is discovered during a normal core reloading
operation. Hence a capability is required to allow the rapid evaluation of a new
loading pattern to ensure that various thermal-hydraulic, mechanical, and nuclear
limitations are not exceeded during the subsequent core operation.
REFERENCES
1. H. M. Summer, The neutron cross section of 135 Xe, AEEW-Rl 16 (1962).

2. R. Fleming, Xenon transients and their measurement, University of Michigan Nuclear
Engineering Department Report (unpublished)(1973).
3. M. E. Meek and B. F. Rider, Compilation of fission product yields, NEDO-2154 (1972).

(1966).
5. H. W. Graves, Jr., Nuclear Reactor Control Management, University of Michigan
Nuclear Engineering Department Lecture Notes (unpublished) (1969).
7. W. M. Stacey, Reactor Tech. 13, 252 (1970).
(1970), pp. 555-562.
9. J. Canosa and H. Brooks, Nucl. Sci. Eng. 26, 237 (1966).
10. D. Randall and D. St: John, Nucleonics 16, 82 (1958); Nucl. Sci. Eng. 14, 204 (1962).
(1966) pp. 481-484.
12. R. L. Crowther, Methods of fuel management analysis-an overview, in Mathematical
Methods and Models in Reactor Calculations, CONF-730414-Pl (1973) p. 11-1.
13. A. Sesonske, Nuclear Power Plant Design Analysis, USAEC TID-26241 (1973).
14. D. L. Delp, et al., FLARE, GEAP-4598 (1964); R. Goldstein, et al., Trans. Am. Nucl.
Soc. 10, 300 (1967).
15. T. R. England, CINDER, a one-point depletion and fission product program, W APD-
TM-334 (1962).
16. R. F. Barry, LEOPARD, a spectrum dependent non-spatial depletion code for the
IBM-7094, WCAP-3741 (1963).
17. R. J. Breen, 0. J. Marlowe, and C. J. Pfeifer, HARMONY: system for nuclear reactor
depletion computation, WAPD-TM-478 (1965).
18. T. B. Fowler and D. R. Vondy, nuclear reactor core analysis code: CITATION,
ORNL-TM-2496, Rev. 2 (1971).
19. D. M. Elliot and L. E. Weaver (Ed.), Education and Research in the Nuclear Fuel Cycle,
Oklahoma U. P., Norman (1970).
20. H. W. Graves, Nuclear Fuel Management, University of Michigan Nuclear Engineering
21. E. A. Mason, Overall view of the nuclear fuel cycle, in Education and Research in the
Nuclear Fuel Cycle, D. M. Elliot and L. E. Weaver (Ed.), Oklahoma U. P., Norman
(1970).
22. J. Haley, in Education and Research in the Nuclear Fuel Cycle, D. M. Elliot and L. E.
Weaver (Ed.), Oklahoma U. P., Norman (1970); V. 0. Uotinen, et al., Nucl. Tech. 18,
129 (1973); R. Astley, et al., Nucl. News 14, 29 (1971).
23. P.R. Kasten, in Education and Research in the Nuclear Fuel Cycle, D. M. Elliot and L. E.
Weaver (Ed.), Oklahoma U. P., Norman (1970); USAEC Report WASH-1097 (1969).
24. P. Fortescue, A reactor strategy: FBR's and HTGR's, Nucl. News 15, 36 (1972).
25. G. Gyorey, private communication (1974).
26. D. Olander, Technical basis of the gas centrifuge, Advances in Nuclear Science and
Technology, Vol. 6, (1972).
27. B. Snavely, Separation of uranium isotopes by laser photochemistry, USAEC Report
UCRL-75725 (1974).
28. N. H. Andrews, C. A. Olmstead, H. G. Houser, Nucl. News 16, 71 (1973).
29. E. D. Fuller, A. L. Phillips, and F. E. Wicks, Nucl. News 16, 71 (1973).
PROBLEMS
15-1 A thermal power reactor is shut down after an extended period of high-power
operation. What happens to: (a) the power output of the reactor (you may ignore
the short-term transients if you wish), (b) the average reactor temperature, and (c)
the ability of the reactor to be restarted.
15-2 Solve the isotopic rate equations to determine the buildup in xenon concentration
following startup of a clean core to a steady-state flux level </>o·
15-3 Determine the ratio of atomic number densities of equilibrium 135Xe and 235 U as a
function of the steady-state thermal flux level of a reactor.
15-4 Consider a subcritical assembly in which a neutron flux of </>o = 106 cm - 2 - sec - 1 is
maintained by a source. How long must this source be left on before the xenon and
samarium concentrations reach 90% of their saturation value, respectively?
15-5 Demonstrate that there will be no buildup of xenon following shutdown unless the
flux prior to shutdown
Explain this result physically.

15-6 Determine the time following shutdown at which the maximum xenon concentration
occurs.
15-7 Consider a large power reactor operating at a thermal flux level of </>o = 10 14 cm - 2
sec - 1 with a power defect of dpp0 = 0.04 and an excess reactivity (in the cold, clean
state) of Pex = 0.15. Determine the length of time following a scram one has available
to restart the reactor before xenon buildup will be sufficiently large to outweigh the
excess reactivity available. How long will this deadtime period last during which the
reactor cannot be restarted?
15-8 Determine the behavior of the samarium concentration in a reactor following a flux
level change </>o-</> 1•
15-9 Compare the effects pf 135Xe and 149Sm buildup in thermal reactors fueled with 233 U,
235
U, and 239 Pu, respectively.
15-10 Plot the trajectory of iodine and xenon buildup in the phase-plane with coordinates
I (t) and X (t) following a clean core reactor startup and then a shutdown some time
later.
15-11 A nuclear reactor is subcritical and operating at very low power level. Assume that
no automatic control devices are operative. Reactivity is inserted (as by control rod
withdrawal) continuously and infinitely slowly so that any short-lived transients
induced come to rest. Making any further assumptions you require, discuss what
happens as the reactivity increases over a wide range.
15-12 Using first-order perturbation theory, estimate the reactivity worth of equilibrium
buildup in a bare cylindrical core in which a low flux level </>o«>..x/ a~ is maintained.
15-13 Consider a slightly enriched uranium-fueled reactor operating at a constant flux level
</>o· Determine the concentrations of 235 U, 239 Pu, and 241 Pu as functions of time.
(State carefully your assumptions concerning the particular nuclide chains you
choose to consider.)
15-14 Determine the maximum possible burn up (MWD /MTU) achievable in a reactor
with a given enrichment and conversion ratio (assume that CR< l). Then compare
this theoretical limit with the actual design burnup goals for each of the major
reactor types (PWR, BWR, HTGR, CANDU, SGHWR) and comment on this
comparison.
15-15 A high-temperature gas-cooled reactor is initially fueled with a mixture of 235 U and
232
Th. As the reactor is operated, the converted isotope 233 U accumulates as the 235 U
is consumed. Develop the isotopic concentration equations for 235 U, 233 Pa, and 233 U.
Solve these equations, assuming that the flux and the thorium concentration remain
constant.
15-16 It has been proposed to fuel a batch-loaded reactor with transuranic nuclide X,
which undergoes both fission and capture. Nonfission capture in X results essentially
instantaneously in nuclide Y, which is also fissionable. Nuclide Y is also subject to
radioactive decay and further capture reactions.
(a) If the reactor is just critical with an X concentration of X 0 , what is the
condition that a non-zero reactivity lifetime be attainable with this loading and
no initial excess reactivity?
(b) Assuming that X and Y are in secular equilibrium near the end of the reactivity
lifetime of the fuel charge, what burnup fraction can be attained?
Neither nuclide undergoes resonance absorption. Neglect neutron absorption in
nuclides Wand Z. Assume constant flux operation. The nuclear parameters of X, Y,
and the fission products are to be assumed known.
®
...~
0 ___ 1_+_a_x_•--► ©~~@
~-"-•"
I, :.:, l 1 aY
1 + ay •
Fission Fission
15-17 Solve the differential equations for buildup of 240Pu and 241 Pu in terms of 4>
= f &'i>(t')dt'. Assume constant cross sections and zero initial concentration of all
plutonium isotopes.
15-18 Give a qualitative discussion of the proposed fuel cycle for the LMFBR. In
particular, discuss: (a) the interaction of the LMFBR fuel cycle with the LWR fuel
cycle, (b) the relative advantages and disadvantages of different proposed fast
reactor fuels (e.g., oxides versus carbides versus metal alloys), and (c) the economic
advantages (if any) of the LMFBR.
15-19 The molten salt-breeder reactor (MSBR) is a fluid-fueled reactor designed to operate
as a thermal breeder on pure 232Th feed. The most significant reactions that occur in
the core are:
(n, y) J.. 13 (n,J)

232
Th ~ 233 Pa ~ 233 U ~ Fission products.
Steady-state operation at zero reactivity is maintained by continuous fission product

removal at a rate proportional to the concentration.
Rate of removal of fission _ CN

products from core (atoms/sec) - FP
where N FP is the concentration of fission products in the core in atoms/ cm 3 and C is

a constant. The heavy elements are not removed. The volume of the core is V and
the average flux is 'Po· The reactor is sufficiently well moderated to justify neglecting
all fast effects. When fueled with pure 235 U (but no thorium) the critical concentra-
tion is NJ5• Determine: (a) the minimum value of constant C and (b) the power
generated by the reactor for C > Cmin· Assume all nuclear parameters are known
and neglect xenon and samarium poisoning.
15-20 A bare cylindrical reactor, containing an initial loading of 100 kg of 235 U (no 238 U)
operates in batch irradiation. Because of radiation damage to the fuel elements, the
maximum local depletion of 235 U due to neutron absorption is limited to 60%.
Estimate the maximum thermal energy that can be produced from each fuel loading.
What assumptions are involved? Use oy/i1r=0.18.
15-21 A nuclear rocket may be required to operate for a period as short as a few minutes,
but during this period it is desirable to obtain power at a constant and maximum
rate. To accomplish this with an initially clean reactor core, it would be necessary to
program the fission rate so that the total instantaneous power from prompt and
delayed (fission product) energy sources is a constant. The time-dependence of the
delayed sources can be determined from the Way-Wigner empirical correlation,
which states that the rate of liberation of decay heat at a time t after a single fission
varies as t- 1.2. This correlation applies for all times greater than 10 sec after the
fission. A fraction J of the total recoverable energy from a single fission follows this
decay law, and all other energy sources are assumed to be prompt. Assuming that
the reactor power can be brought quickly (instantaneously) to the desired power P at
the time of startup, how should the fission rate F(t) be programmed with respect to
time so that P remains constant?
15-22 An important isotope for power production in space is 238 Pu.
(a) Suggest a practical scheme whereby 238 Pu can be produced with the aid of a
reactor.
(b) Write down, but do not solve, the differential equations governing the con-
centration of 238 Pu and each of its precursors. Neglect resonance absorption
effects.
Appendix A
Some Useful Nuclear Data
I. MISCELLANEOUS PHYSICAL CONSTANTS

6.022045 x 1Q23 mol - 1
Avogadro's number, NA
Boltzmann constant, k 1.380662x 10- 23 J K- 1
0.861735X 10- 4 eV K- 1
Electron rest mass, m. 9.109534x 10- 31 kg
0.5110034 MeV
Elementary charge, e 1.6021892 X 10- 19C
Gas constant, R 8.31441 Jmo1- 1 K- 1
Neutron rest mass, mn l.6749544X 10- 27 kg
939.5731 MeV
Planck's constant, h 6.626176X 10- 34 J Hz- 1
Proton rest mass, mp 1.6726485 X 10- 27 kg
938.2796 MeV
Speed of light, c 2.99792458 X 108 m sec- 1
II. SOME USEFUL CONVERSION FACTORS

1 eV l.6021892x 10- 19 J
1 MeV 106 eV
lamu 1.6605655 X 10- 27 kg
931.5016 MeV
lW 1 joule/sec
1d 86400 sec
1 mean y 365.25 d
8766 h
3.156X 107 sec
1 curie 3.7000X 10 10 disintegrations/sec
1 °K 8.617065 X 10- 5 eV
605
I III. 2200m/sec CROSS SECTIONS FOR NATURALLY OCCURRING ELEMENTS
[From Reactor Physics Constants, ANL-5800 (1963)].
Nuclei Microscopic Macroscopic

Atomic per Cross Section, Cross Section,
Element or Unit b cm- 1
Atomic
No.
H
or
Compound
Mol.
Wt.
1.008
Density,
g/cm3
8.9t
Vol.
x10-24
5_3t
1- iio ~ (Ja (JS (JI ~. ~. ~t
0.3386 1.000 0.33 38 38 1.7t 0.002 0.002

H 2O 18.016 I o.0335t 0.676 0.948 0.66 103 103 0.022 3.45 3.45
D 2O 20.030 1.10 0.033It 0.884 0.570 0.001 13.6 13.6 3_3t 0.449 0.449
2 He 4.003 11.st 2.6t 0.8334 0.425 0.007 0.8 0.807 o.02t 2.1t 2.1t
3 Li 6.940 0.534 0.0463 0.9047 0.268 71 1.4 72.4 3.29 0.065 3.35
4 Be 9.013 1.85 0.1236 0.9259 0.209 0.010 7.0 7.01 124* 0.865 0.865
BeO 25.02 3.025 0.0728: 0.939 0.173 0.010 6.8 6.8 73t 0.501 0.501
5 B 10.82 2.45 0.1364 0.9394 0.171 755 4 759 103 0.346 104
6 C 12.01 I 1.60 0.0803 0.9444 0.158 0.004 4.8 4.80 32t 0.385 0.385
7 N 14.008 0.0013 5_3t 0.9524 0.136 1.88 10 11.9 9_9t sot 60t
8 0 16.000 0.0014 5_3t 0.9583 0.120 20t 4.2 4.2 0.000 21t 21t
9 F 19.00 0.0017 5_3t 0.9649 0.102 0.001 3.9 3.90 o.01t 20t 20t
10 Ne 20.183 0.0009 2.6t 0.9667 0.0968<2.8 2.4 5.2 7_3t 6.2t 13.st
11 Na 22.991 0.971 0.0254 0.9710 0.0845 0.525 4 4.53 0.013 0.102 0.115
12 Mg 24.32 1.74 0.0431 0.9722 0.0811 0.069 3.6 3.67 0.003 0.155 0.158
13 Al 26.98 2.699 0.0602 0.9754 0.0723 0.241 1.4 1.64 0,0!5 0.084 0.099
14 Si 28.09 2.42 0.0522 0.9762 0.0698 0.16 1.7 1.86 0.008 0.089 0.097
15 p 30.975 1.82 0.0354 0.9785 0.0632 0.20 5 5.20 0.007 0.177 0.184
16 s 32.066 2.07 0.0389 0.9792 0.0612 0.52 1.1 1.62 0.020 0.043 0.063
17 Cl 35.457 0.0032 5.3t 0.9810 0.0561 33.8 16 49.8 0.002 sot 0.003
18 A 39.944 0.0018 2.6t 0.9833 0.0492 0.66 1.5 2.16 1.7t 5.6t
3.9
19 K 39.100 0.87 0.0134 0.9829 0.0504 2.07 1.5 3.57 0.Q28 0.020 0.048
20 Ca 40.08 1.55 0.0233 0.9833 0.0492 0.44 3.0 3.44 0.010 0.070 0.080
21 Sc 44.96 2.5 0.0335 0.9852 0.0438 24 24 48 0.804 0.804 1.61
22 Ti 47.90 4.5 0.0566 0.9861 0.0411 5.8 4 9.8 0.328 0.226 0.555
J
23 V 50.95 5.96 0.0704 0.9869 0.0387 5 5 10.0 0.352 0.352 0.704
24 Cr 52.01 7.1 0.0822 0.9872 0.0385 3.1 3 6.1 0.255 0.247 0.501
25 Mn 54.94 7.2 0.0789 0.9878 0.0359 13.2 2.3 15.5 1.04 0.181 1.22
26 Fe 55.85 7.86 0.0848 0.9881 0.0353 2.62 11 13.6 0.222 0.933 1.15
27 Co 58.94 8.9 0.0910 0.9887 0.0335 38 7 45 3.46 0.637 4.10
28 Ni 58.71 8.90 0.0913 0.9887 0.0335 4.6 17.5 22.1 0.420 1.60 2.02
29 Cu 63.54 8.94 0.0848 0.9896 0,0309 3.85 7.2 11.05 0.0326 0.611 0.937
30 Zn 65.38 7.14 0.0658 0.9897 0.0304 I.JO 3.6 4.70 0.072 0.237 0.309
31 Ga 69.72 5.91 0.0511 0.9925 0.0283 2.80 4 6.80 0.143 0.204 0.347
32 Ge 72.60 5.36 0.0445 0.9909 0.0271 2.45 3 5.45 0.109 0.134 0.243
33 As 74.91 5.73 0.0461 0.9911 0.0264 4.3 6 10.3 0.198 0.277 0.475
34 Se 78.96 4.8 0.0366 0.9916 0.0251 12.3 11 23.3 0.450 0.403 0.853
35 Br 79.916 3.12 0.0235 0.9917 0.0247 6.7 6 12.7 0.157 0.141 0.298
36 Kr 83.80 0.0037 2.6t 0.9921 0.0236 31 7.2 38.2 8 It 19t 99t
37 Rb 85.48 1.53 0.0108 0.9922 0.0233 0.73 12 12.7 0.008 0.130 0.138
38 Sr 87.63 2.54 0.0175 0.9925 0.0226 1.21 10 11.2 0.021 0.175 0.195
39 Yt 88.92 5.51 0.0373 0.9925 0.0223 1.313 4.3 4.3 0.049 0.112 0.160
40 Zr 91.22 6.4 0.0423 0.9927 0.0218 0.185 8 8.2 0.008 0.338 0.347
41 Nb 92.91 8.4 0.0545 0.9928 0.0214 1.16 5 6.16 0.063 0.273 0.336
42 Mo 95.95 10.2 0.0640 0.9931 0.0207 2.70 7 9.70 0.173 0.448 0.621
43 Tc 98.0 0.9932 0.0203 22
44 Ru IOI.I 12.2 0.0727 0.9934 0.0197 2.56 6 8.56 0.186 0.436 0.622
45 Rh 102.91 12.5 0.0732 0.9935 0.0193 149 5 154 10.9 0.366 11.3
46 Pd 106.4 12.16 0.0689 0.9937 0.0187 8 3.6 11.6 0.551 0.248 0.799
47 Ag 107.88 10.5 0.0586 0.9938 0.0184 63 6 69 3.69 0.352 4.04
48 Cd 112.41 8.65 0.0464 0.9940 0.0178 2450 7 2457 114 0.325 114
49 In 114.82 7.28 0.0382 0.9942 0.0173 191 2.2 193 7.30 0.084 7.37
50 Sn 118.70 6.5 0.0330 0.9944 0.0167 0.625 4 4.6 0.021 0.132 0.152
51 Sb 121.76 6.69 0.0331 0.9945 0.0163 5.7 4.3 10.0 0.189 0.142 0.331
52 Te 127.61 6.24 0.0295 0.9948 0.0155 4.7 5 9.7 0.139 0.148 0.286
53 I 126.91 4.93 0.0234 0.9948 0.0157 7.0 3.6 10.6 0.164 0.084 0.248
54 Xe 131.30 0.0059 2.7t 0.9949 0.0152 35 4.3 39.3 95t 12t 0.001
55 Cs 132.91 1.873 0.0085 0.9950 0.0150 28 20 48 0.238 0.170 0.408
g 56 Ba 137.36 3.5 0.0154 0.9951 0.0145 1.2 8 9.2 0.018 0.123 0.142
-..J
g III. ( Continued)
QO
z
57
Element
La 138.92
A p
6.19
N
0.0268
1- P:o
0.9952
f
0.0143
a.
8.9
o,
15
o,
24
~.
0.239
~,
0.403
~.
0.642
58 Ce 140.13 6.78 0.0292 0.9952 0.0142 0.73 9 9.7 0.021 0.283
0.263
59 Pr 140.92 6.78 0.0290 0.9953 0.0141 11.3 4 15.3 0.328 0.116 0.444
60 Nd 144.27 6.95 0.0290 0.9954 0.0138 46 16 62 1.33 1.79
0.464
61 Pm 145.0 0.9954 0.0137 60
62 Sm 150.35 7.7 0.0309 0.9956 0.0133 5600 5 5605 173 0.155 173
Sm2 0 3 348.70 7.43 o.0128t 0.974 0.076 16,500 22.6 16,500 211 0.289 211
63 Eu 152.0 5.22 0.0207 0.9956 0.0131 4300 8 4308 89.0 0.166 89.2
Eu2O 3 352.00 7.42 o.0121t 0.978 0.063 8740 30.2 8770 111 0.383 111
64 Gd. 167.26 7.95 0.0305 0.9958 0.0127 46,000 1403
65 Tb 158.93 8.33 0.0316 0.9958 0.0125 46 1.45
66 Dy 162.51 8.56 0.0317 0.9959 0.0122 950 100 1050 30.1 3.17 33.3
Dy2O3 372.92 7.81 0.0126t 0.993 0.019 2200 214 2414 27.7 2.7 30.4
67 Ho 164.94 8.76 0.0320 0.9960 0.0121 65 2.08
68 Er 167.27 9.16 0.0330 0.9960 0.0119 173 15 188 5.71 0.495 6.20
69 Tm 168.94 9.35 0.0333 0.9961 0.0118 127 7 134 4.23 0.233 4.46
70 Yh 173.04 7.01 0.0244 0.9961 0.0115 37 12 49 0.903 0.293 1.20
71 Lu 174.99 9.74 0.0335 0.9962 0.0114 112 3.75
72 Hf 178.5 13.3 0.0449 0.9963 0.0112 105 8 113 4.71 0.0359 5.07
73 Ta 180.95 16.6 0.0553 0.9963 0.0110 21 5 26 1.16 0.277 1.44
74 w 183.86 19.3 0.0632 0.9964 0.0108 19.2 5 24.2 1.21 0.316 1.53
75 Re 186.22 20.53 0.0664 0.9964 0.0107 86 14 100 5.71 0.930 6.64
76 OS 190.2 22.48 0.0712 0.9965 0.0105 15.3 11 26.3 1.09 0.783 1.87
77 Ir 192.2 22.42 0.0703 0.9965 0.0104 440 30.9
78 Pt 195.09 21.37 0.0660 0.9966 0.0102 8.8 10 18.8 0.581 0.660 1.24
79 Au 197.0 19.32 0.0591 0.9966 0.0101 98.8 9.3 107.3 5.79 0.550 6.34
80 Hg 200.61 13.55 0.0407 0.9967 0.0099 380 20 400 15.5 0.814 16.3
81 Ti 204.39 11.85 0.0349 0.9967 0.0098 3.4 14 17.4 0.119 0.489 0.607
82 Pb 207.21 11.35 0.0330 0.9968 0.0096 0.170 11 11.2 0.006 0.363 0.369
83 Bi 209.0 9.747 0.0281 0.9968 0.0095 0.034 9 9 0.001 0.253 0.256
84 Po 210.0 9.24 0.0265 0.9968 0.0095
85 At 211.0 0.9968 0.0094
86 Rn 222.0 0.0097 2.6t 0.9970 0.0090 0.7
87 Fr 223.0 0.9980 0.0089
88 Ra 226.05 5 0.0133 0.9971 0.0088 20 0.266
89 Ac 227.0 0.9971 0.0088 510
90 Th 232.05 11.3 0.0293 0.9971 0.0086 7.56 12.6 20.2 0.222 0.369 0.592
91 Pa 231.0 15.4 0.0402 0.9971 0.0086 200 8.04
92 u 238.07 18.9 0.04783 0.9972 0.0084 7.68 8.3 16.0 0.367 0.397 0.765
U02 270.07 10 o.0223i 0.9887 0.036 7.6 16.7 24.3 0.169 0.372 0.542
93 Np 237.0 0.9972 0.0084 170
94 Pu 239.0 19.74 0.0498 0.9972 0.0083 1026 9.6 1036 51.1 0.478 51.6
95 Am 242.0 0.9973 0.0082 8.000
t Value has been multiplied by 105 •

t Molecules/cm 3 •
I
IV. 2200 m/sec CROSS SECTIONS OF SPECIAL INTEREST:
JOB: oa=3837b
11B: oa=0.005
13sXe: oa=2.7X 106
233U: <Iy =49 or= 524
23su: (Jy = 101 o1 =577
23su: oy=2.73
239Pu:. <Iy =274 or=74l
240Pu: <Iy =286 a 1 =0.03
241Pu: (Jy =425 o1 =950
242Pu: <Iy =30 or<0.2
610
Appendix B
Some Useful Mathematical
Formulas
(l) Solution of First-Order Linear Differential Equations:
df
dx +a(x)f(x)=g(x) (B-1)
(B-2)
(2) Differentiation of a Definite Integral:
d ra(x)dx'F(x,x')=F(x,a)dda -F(x,b)ddb + (a(x)dx'aF~x,x') (B-3)

dX jb(x) X X Jb(x) X
(3) Representation of Laplacian V2 in Various Coordinate Systems:

(a) Cartesian:
(B-4)
611
612 / SOME USEFUL MATHEMATICAL FORMULAS
(b) Cylindrical:
2 1 a a 1 a2 a2
V =--r-+--+-
r ar ar r2 ao2 az2
(B-5)
(c) Spherical: z
(B-6)
(4) Gauss' Divergence Theorem:
f/ rV·A= ldSe,A
3
(B-7)
where e. is the unit vector normal to the surface element dS.

(5) Green's Theorem:
f d 3rt'<f,·t'tf,= f dSq,e,Vtf,- f d 3rq,V2tf, (B-8)
f d r(<f,V2t/,-tf,V <f,)= f dSe.-(q,Vtf,-tf,V<f,)

3 2
(B-9)
(6) Taylor Series Expansion:
(B-10)
(7) Fourier Series Expansion:
(B-11)
where
(B-12)
Appendix C
Step Functions, Delta Functions,
and Other Exotic Beasts
I. INTRODUCTION
Consider the discontinuous function 0(x) defined by the properties
0(x)= { ~ (C-1)
:,,
X
0(x) is the unit "step function" introduced by Heaviside in his development of operational
calculus (now known as integral transform analysis). One can perform numerous operations
on 0(x). In particular it can be integrated to yield the ramp function
71(x)=fx dx'E>(x')= { O, (C-2)

-oo x,
Let's try something a bit more unusual by taking the derivative of 0(x). Clearly this is
ridiculous, because this derivative, call it B(x), is undefined at x=0 because 0(x) is
discontinuous at this point:
8(x)=0'(x)= lim [ E>(x+t:)-E>(x)] ={ 0, x,;=0 (C-3)

,-o t: oo, x=0.
Nevertheless Dirac, Heaviside, and others have made very good use of this strange
"function." To be more specific, the Dirac 8-function, 8 (x), has the properties
8(x-x 0 )= { O, (C-4)
oo,
613
614 / STEP FUNCilONS, DELTA FUNCilONS, AND OTHER EXOTIC BEASTS
In a sense, it resembles a generalization of the Kronecker 8-function
8mn-
-{ 0, m,t=n
l, m=n.
The most useful property of the Dirac 8-function occurs when it is integrated along with a
well-behaved function, say f(x):
f dxf(x)8 (x- x 0) = J(x 0 ). (C-5)
This property not only is very interesting, but extremely useful in mathematical physics.
Unfortunately the proof of this property-and, indeed, all of the theory of such generalized
functions-requires a rather potent dose of mathematics. [Such generalized functions are
really not functions at all, but rather a class of linear functionals 1 called "distributions"
defined on some set of suitable test functions (which are "infinitely differentiable with
compact support").]
Fortunately one does not need all of this high-powered mathematics in order to use
8-functions. Only a knowledge of their properties is necessary.
II. PROPERTIES OF THE DIRAC 8-FUNCTION
A. Alternative Representations
(C-6)
(C-7)
B. Properties
8(x)=8(-x), (C-8)
l
8(ax)= laf8(x), a,t=O, (C-9)
8[g(x)]= ~ lg}xn)l8(x-xn), [g(xn)=O,g'(xn),t=OJ (C-10)
x8(x)=O, (C-1 l)
J(x)8 (x-a) = J(a)8 (x- a), (C-12)
f 8(x-y)8(y-a)dy= 8(x-a), (C-13)
8(x)= _l
2'1T
Joo
-oo
dkeikx. (C-14)
Actually these properties only make sense when inserted in an integral. For example,
property (C-8) really should be interpreted as
f dxf(x)8(x) = f dxf(x)8 ( -x) =J(O). (C-15)

STEP FUNCTIONS, DELTA FUNCTIONS, AND OTIIER EXOTIC BEASTS / 615
C. Derivatives
One can differentiate a 8-function as many times as one wishes. The mth derivative is
defined by
(C-16)
One can show
8[ml(x)=(-l)m8[m)(-x), (C-17)
f 8[ml(x-y )Bl•l(y- a)dy= 8[m+•l(x- a), (C-18)
(C-19)
Perhaps of more direct use is the application of these properties to the first derivative
J_: 8'(x)J(x)dx= - f'(0), (C-20)
8'(x) = - 8'( - x), (C-21)
f 8'(x-y)8 (y- a)dy = 8'(x- a), (C-22)
x8'(x)= -8(x). (C-23)
One can generalize the concept of a 8-function to several dimensions. For example, we
would define the three-dimensional 8-function by
f d r'8(r'-r)J(r')=J(r).
3 (C-24)
Note that we could write this in Cartesian coordinates as
8 (r-r') = 8 (x- x')8 (y-y')8 (z - z'). (C-25)
Such multidimensional 8-functions are of very considerable use in vector calculus.

More detailed discussions of the Dirac 8-function and its relatives are found in the
following references:
(1) J. W. Dettman, Mathematical Methods in Physics and Engineering, 2nd Edition,
McGraw-Hill, New York (1969).
(2) M. J. Lighthill, Fourier Analysis and Generalized Functions, Cambridge U. P.
(1959).
(3) A. Messiah, Quantum Mechanics, Vol. I, Wiley, New York (1965), pp. 468-470.
Appendix D
Some Properties of Special
Functions
( l) Legendre Functions:
(a) Defining equation:
(l- x 2 )j"-2xf' + l(l+ l)J=0, /=integer. (D-1)
(b) Representation:
(D-2)
(c) Properties:
(D-4)
(d) Recurrence relations:
P/+ 1(x)-xP/ (x)=(l+ l)P1 (x) (D-5)

(I+ l)P1+ 1(x)- (21 + l)xP1 (x) + IP1_ 1(x) = 0. (D-6)
(2) Associated Legendre Polynomials:

2
2
(l-x )j"-2xf'+[t(l+l)- m ]J=O. (D-7)
l-x 2
616
SOME PROPERTIES OF SPECIAL FUNCTIONS / 617
(b) Representation:
(D-8)
(c) Spherical harmonics:
(D-9)
(d) Properties:
(D-10)
(D-11)
(3) Bessel Functions:

x3/" + xf' +(x 2 - n 2 )f =0 (D-12)
(b) Solution: Jn(x), Bessel function of first kind

Yn(x), Bessel function of second kind
(c) Representation:
oo k
""" ( - l) (x )n+2k J. (-x) cos(mr )- J -n (x)
J.(x)= ~ f(k+ l)f(k+n+ l) I ' Yn(x)= .
smmr
k=O
(D-13)
(d) Hankel functions:
HJ 1l(x) =J. (x) + i Yn (x) (D-13)
HPl(x)=Jn (x)-i Y. (x) (D-14)
(4) Modified Bessel Functions:

(D-15)
(b) Solution: 1.(x), modified Bessel function of first kind

K.(x), modified Bessel function of second kind
(c) Representation:
In (x) = i-nJn (ix)= i"Jn ( - ix)
Kn (x) = I in+ 1HJ 1l(ix) = I i-n- 1HJ2l( - ix) (D-16)

618 / SOME PROPERTIES OF SPECIAL FUNCTIONS
2.4 K 0 (x)
2.0
1.6
1.2
0.8
0.4
] 0 (x)
FIGURED-I. The Bessel functions of zero order.
(5) Useful Expansions of Bessel Functions for small x:
x2 x4 x6
Jo(x)=l--+----+ ··· (D-17)
4 64 2304
X x3 XS
J,(x)=2-T6+ 384 -··· (D-18)
Yo(x)= ¾ [ ( y+ln I )J (x)+ ~

0
2
+ • •.], y=0.577216 (D-19)
Y1(x)=¾[(r+lnj)1 1(x)-f-¾+···] (D-20)
x2 x4 x6
/o(x)=l+-+-+--+ · · · (D-21)
4 64 2304
X x3 XS
I 1(x)= - + - + - + · · · (D-22)
2 16 384
(D-23)
(D-24)
(a) Asymptotic expansions for large x:
ex ( l+-+
Io(x)=-- 1 ... ) (D-25)
Vfix 8x
(D-26)
K 0(x)=· ~
V2x
1
e-x(1- - + · · ·)
8x
(D-27)
(D-28)
SOME PROPERTIES OF SPECIAL FUNCTIONS / 619
(b) Recurrence relations:
(D-29)
(D-30)
(D-31)
(D-32)
(D-33)
(c) Integrals:
(D-34)
(D-35)
(6) Gamma Function:

(a) Definition:
(D-36)
(b) Properties:
f(z+l)=zf(z)
f(l/2)=\,,'1";, f(O)=oo, f(l)=l, ... ,f(n)=(n-1)!. (D-37)
(7) Error Function:

(a) Definition:
2 rx
erf(x)= - - ) dte- 1
, (D-38)
\,,'1"; 00
(b) Complementary error function:
erfc(x)= 1-erf(x) = -2- Joo dte- 12 (D-39)

\,,'1"; X
(8) Exponential Integrals:

(a) Definition:
oo -xt oo dt-e-= Joo dt~

J J
-xt -t
En(x)= dt½-, E 1(x)= (D-40)
I t I f x t
(b} Properties:
(D-41)
E~(x)= -En-i(x) (D-42)

1
En(x)= - - [e-x-xEn-i(x)], n>l (D-43)
n- 1
(D-44)
620 / SOME PROPERTIES OF SPECIAL FUNCTIONS
REFERENCES
1. M. Abramowitz and I. Stegun (Eds.), Handbook of Mathematica Functions, Dover, New
York (1965).
2. H. Margenau and G. M. Murphy, The Mathematics of Physics and Chemistry, 2nd Ed.,
Vol. I, Van Nostrand, Princeton, N.J. (1956).
3. I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products, 4th Ed.,
Academic Press, New York (1965).
4. P. M. Morse and H. Feshbach, Methods of Theoretical Physics, Vol. I and II, McGraw-
Hill, New York (1953).
Appendix E
Some Assorted Facts
on Linear Operators
I. INNER PRODUCTS
We define the inner product or scalar product (f,g) of two functions f (x) and g(x) as
(f,g)= f dxw(x)f *(x)g(x),

a
b (E-1)
[here w(x) is frequently taken as 1] wheref*(x) denotes the complex conjugate of f(x). This
inner product has the properties
(a) (f,g) = ( g,f)*'
(b) (af1(x)+ bfi(x), g)= a*U1,g)+ b*U2,g),
(c) (j,f) >0 if f(x)'i"'O,
[Note: Sometimes one defines a real inner product
(f,g)= fba dxw(x)f(x)g(x).

Such a form satisfies properties (a) and (b), but property (c) is satisfied only for real
functions. Hence if one is concerned with complex valued functions, then it is necessary to
use the more general form (E-1).]
We define the norm of a functionf(x) as
llfll = [(f,!)] 112 • (E-2)
The norm is a real, positive number [by property (c) above] and characterizes the
"magnitude" of the function f(x).
Two functions, f (x) and g(x), are said to be orthogonal if their inner product vanishes,
(f,g)=O. (E-3)
621
622 / SOME ASSORTED FACTS ON LINEAR OPERATORS
II. LINEAR OPERATORS
An operator refers to a mathematical operation by which we convert a function J(x)

into another function g(x)-that is,
AJ(x)=g(x). (E-4)
(In the more colorful language of functional analysis, an operator is "a mapping of a linear
vector space into a linear vector space.")
EXAMPLES:
d df
Differential operator: A O
= dx, AJ = dx •
Integral operator: A O
= lb dx' k(x,x') 0 , Af = lb dx' k(x,x')f(x').
a a
Unit operator: A 0 ::l 0 , Af=f(x).
Null operator: A =0°,
0 Of=O.
Displacement operator: A 0 =exp(-a ix 0 ), Af=f(x-a).
Ao=(o)2, Af=[f(x)]2.
If A(af+bg)=aAJ+bAg, then we refer to A as a linear operator. Note that the first five
operators above are linear, while the sixth operator is a nonlinear operator.
Such operators can frequently be manipulated formally via ordinary algebra, much as one
manipulates numbers or functions. However one must be very careful. In particular note
that operators are generally not commutative
ABJ(x),t= BAJ(x). (E-5)
There is one other animal quite similar to an operator, that occasionally arises in
mathematics. One refers to this mathematical operation, which converts a function into a
scalar, as a functional %
'!f'{J(x)} = a =number (E-6)
(or, if you prefer, "a mapping of a vector space into a scalar field").
EXAMPLE: A definite integral
'!f'{J(x)}= fbdxf(x)=number. (E-7)

a
In this sense, if we fix f(x), then we can regard the inner product of f(x) with g(x) as a
functional of g(x), that is,
%{ g(x)} = (J,g) =number. (E-8)
III. LINEAR VECTOR SPACES
We have now defined the concept of an operator. Of course, such a definition is

incomplete until we specify the types of functions we are going to let it act upon. For
example, it makes no sense to let a differential operator d/ dx act on a discontinuous
function, since df/ dx is not defined at the point of discontinuity.
SOME ASSORTED FACTS ON LINEAR OPERATORS / 623
Hence in practice one studies an operator A by first specifying the class or type of
function on which he will let A operate. Such classes of functions usually possess the
properties of a linear vector space, that is, a set of functions closed under the operations of
addition and of multiplication by a scalar and containing a zero element.
EXAMPLES
c 0[a,b]: The set of all continuous functionsf(x) defined on the interval xE[a,b].
C[a,b]: The set of all functions defined on xE[a,b] with n continuous de~ivatives.
L 2[a,b]: The set of all functionsf(x) defined on xE[a,b] such that
(This is an example of a "Hilbert space".)

Lp[a,b]: The set of all functionsf(x) defined on xE[a,b] such that
Jbdxlf(x)IP < oo.

a
(This is an example of a "Banach space".) Note that for studying a differential

operator of order n, we usually choose a corresponding space cn- 1• For an
integral operator, the LP spaces are usually chosen.
The particular linear vector space or class of functions for which an operator A is defined
is referred to as its domain 6D. Usually, 6D will be a subspace of a larger vector space V (that
is, the functions belonging to the class 6D will also belong to a more general class V). For
instance, A= d/ dx is defined on the domain of all continuous functions 6D = c 0• This class,
however, is actually contained in the more general class of piecewise continuous functions P
(i.e., 6D C P).
IV. PROPERTIES OF OPERATORS

Consider now a linear operator A defined on a domain of functions 6D . Then we refer
to A as a bounded operator in this domain if
IIAfll (E-9)
m<oo for allj(x) E 6D.
If A is bounded on 6D, then we can define its norm (an operator norm) as the largest of these
values
IIAJII
IIA 11 = ~a~ 7f7f. (E-10)
Let us go one step further and associate an inner product (f,g) with any two functions J,g
contained in the domain 6D . Then we can define a positive definite operator as one for which
(j,Af)>O for allf'T"O, (E-11)
Using the inner product (f,g), we can now introduce the very important concept of the
operator At adjoint to A. We define the adjoint At of an operator A as that operator At for
which
(A 1/,g) = (j,Ag) for allj,gE 6D. (E-12)

624 / SOME ASSORTED FACTS ON LINEAR OPERATORS
Actually it may occur in practice that At and A are defined over different domains-say 6))t
and 6D . Then this definition of the adjoint operator must be defined by
(A 1/,g) = (!,Ag) for all gE 6D, f E 6))t. (E-13)

Should
(E-14)
then we refer to the operator A as self-adjoint. Self-adjoint operators are very popular in
mathematics because they are easy to study, particularly with regard to their associated
eigenvalue problem
{E-15)
For instance, for self-adjoint operators the eigenvalues A are all real. Furthermore one can
make statements regarding the eigenfunctions 1/;;,..(x), such as that they form a complete set.
One reason for the preoccupation of mathematicians with self-adjoint operators concerns
their role in quantum mechanics (all quantum mechanical observables correspond to
self-adjoint operators).
V. DIFFERENTIAL OPERATORS
Let's make a few more observations about the particular case of differential operators
L, defined on some domain 6D. For such operators, we can always construct the adjoint L
using integration by parts:
EXAMPLE: L=d/dx, 6D =C[a,b]
(!,Lg)= f bdxf*(x) dxdg =J*(b)g(b)-J*(a)g(a)- fb dx dj*

a
dx g(x)
a
=J*(b)g(b)-J*(a)g(a)+(- ix ,g). (E-16)
Hence provided J(J,g)=f*(b)g(b)- f*(a)g(a)=O, we could identify the adjoint of Las
(E-17)
How do we force J (J,g) to vanish? (Here J (J,g) is sometimes referred to as the conjunct of
the operator L.) For a given problem, we will be given certain boundary conditions on g(x)
-say g(a)=c 1, g(b)=c2 • Hence by an appropriate choice of boundary conditions to be
satisfied by f(x) [in this case f (a)= c!, f (b) = ct] we can compel J (J,g) to vanish. In general
the boundary conditions we are required to place on f (x) will not be the same as those we
place on g(x)--that is, in the more abstract language developed earlier, the domain of
definition of the adjoint operator Lt, f E 6Dt, will generally not be the same as the domain on
which L is defined.
A couple of definitions are useful here. If
L=Lt, (E-18)
then we refer to the operator L as being formally self-adjoint. If the domains of definition are
also identical
and 6j) = 6Dt, (E-19)
SOME ASSORTED FACTS ON LINEAR OPERATORS / 625
then we refer to the operator L plus the corresponding boundary conditions as being
self-adjoint.
There are numerous fascinating problems and concepts that arise in the study of
differential operators (which, of course, play the central role in most of physics). However it
is most appropriate to refer the interested reader to one of the standard sources for more
information:
(1) B. Friedmann, Principles and Techniques of Applied Mathematics, Wiley, New
York, (1956).
(2) R. Courant and D. Hilbert, Methods of Mathematical Physics, Vol. I, Wiley-
Interscience, New York (1956).
(3) J. W. Dettman, Mathematical Methods in Physics and Engineering, 2nd Edition,
McGraw-Hill, New York (1969).
Appendix F
An Introduction to Matrices
and Matrix Algebra
I. SOME DEFINITIONS
One defines a matrix of order (m X n) to be a rectangular array of m rows and n

columns
all a12 a13 aln
a21
A= (F-1)
a!i
aml amn
The matrix elements a!i will be identified by subscripts denoting their row i and column}. If
the matrix has the same number of rows as columns, it is said to be a square matrix; for
example,
(F-2)
A diagonal matrix has nonzero elements only along its main diagonal:
(F-3)
A tridiagonal matrix would have nonzero elements only along its central three diagonals:
A= (F-4)
626
AN INTRODUCTION TO MATRICES AND MATRIX ALGF8RA / 627
The unit matrix is the diagonal matrix with elements au= I, i = J:
I 0 0
0 1 0
I= 0 0 1 (F-5)
For two matrices to be equal, each of their matrix elements must be equal:
(F-6)
The transpose of a matrix is obtained by interchanging its rows and columns:
[ =AT] ij =[A]
= .. jl
(F-7)
or
T
(F-8)
The determinant of a matrix is formed by taking the determinant of the elements of the
matrix:
det d = 1 d I = (F-9)
Of course, the determinant of a matrix is a scalar-that is, just a number.

One defines the cofactor of a square matrix for an element au by deleting the ith row and
}th column, calculating the determinant of the remaining array, and multiplying by (- 1y+1 :
a 11 a 12 a 14
a31 a32 a34

(cof d,) 23 =cof =(- l);+J a41 a42 a44 · (F-10)
We can construct the adjoint or Hermitean conjugate of a matrix by complex-conjugating

each of its elements and then transposing as
or (F-11)
For example,
If the determinant of a matrix vanishes, det(4) = 0, then the matrix d. is said to be singular.
If det(d,)7"'0, the matrix is said to nonsingular.- -
628 / AN INTRODUCTION TO MATRICES AND MATRIX ALGEBRA
II. MATRIX ALGEBRA

Two matrices of the same order may be added by adding their corresponding elements
(the same holds for subtraction):
.. ·i- (F-12)
In order for matrix multiplication to be possible, the number of columns of the first matrix
must equal the number of rows of the second matrix. One then calculates the matrix
elements of £=d,;!1, as
n
cii = L a;kbkj, (F-13)
k=I
or more explicitly
a11 b12 b13 a 11 b 11 + a 12 b21 + · · ·

a2, h22 b23
= a2,h12+a22h22+··· . (F-l 4)
d: 11= a3, bn b33
Notice that matrix multiplication is not commutative-that is, 4· !l.'i=l!: 4 in general.

A very important matrix concept is the inverse of a square matrix, d, =-,,
which is defined
by the relation
(F-15)
The inverse can be calculated as
(F-16)
For example, consider
Then
1.dl=3,
while
-1)
2 .
Hence
Notice that if a matrix is singular, that is, det(d) = 0, then it has no inverse.
Appendix G
An Introduction to Laplace Transforms
I. MOTIVATION
Differential equations play a central role in the description of most scientific

phenomena. Moreover, in many cases these phenomena can be approximately described by
a particularly simple type of differential equation-namely, those with constant coefficients.
In this Appendix we will try to develop one of the most powerful tools for solving such
equations: the application of integral transforms, and more specifically, the use of Laplace
transforms to solve differential equations.
The analogy between the use of transform methods to solve differential equations and the
use of logarithms to simplify arithmetic operations is quite striking. Suppose we wish to
multiply two complicated numbers a and b together. Then an easy way to do this is to use
logarithms
a---~loga,
a X b ~ log a+ log b ~ elloga + logbl =a X b
That is, by first taking logs we have simplified the original problem, reducing it to a simple
sum.
This is essentially the idea behind integral transform techniques. Suppose we symbolically
represent the transform operation on a function as
J(t)-j (s).
Then the idea is to transform the differential equation of interest
df ~ _ - _
dt + •. • ~~ sf(s) + ••• ~ f(t)
r-::=-J _
629
630 / AN INTRODUCTION TO LAPLACE TRANSFORMS
In this manner, the integral transform can be used to convert this differential equation into a
simpler problem (frequently an algebraic equation) that can then be solved rather easily for
the transformed solution. We then must somehow "invert" the transform to obtain the
actual solution of interest.
EXAMPLE: Consider the very simple ordinary differential equation (familiar from
prompt neutron reactor kinetics)
~~ -( ~ )n(t)=O, n(0) = n0 • (G-1)
Now define the Laplace transform of n(t) as
(G-2)
To transform the ordinary differential equation (G-1), multiply bye- st and integrate overt
or using integration by parts
sfi (s)- n(0)- (p/ A)fi (s) = 0,
but this is now just an algebraic equation which can be easily solved for
~ no
n(s)= s-(p/A) (G-3)
We must now "invert" fi(s) to find
n(t)= e- 1{fi(s)}. (G-4)
By noting that
we find
(G-5)
EXAMPLE: Integral transforms can also be applied to the solution of partial differential
equations. Consider, for example, the initial value problem for a nonmultiplying slab in
one-speed diffusion theory
Initial condition: cf,(x,0)=cf, 0 (x)

(G-6)
Boundary condition: cf,(0, t) = cf,(/, t) = 0. 0
AN INTRODUCTION TO LAPLACE TRANSFORMS / 631
Define the Laplace transform of cp(x, t) with respect to t by
(G-7)
Now multiplying (G-6) by e- st and integrating over all times t, we find the transformed
partial differential equation becomes
2
1 - d-~ -
-[scp(x,s)-cp(x,O)] =D - ~a</>(x,s).
v dx 2
Since the boundary conditions also depend on time, we must transform them to find:
~(O,s) =~(/,s) = 0.
Hence if we regards only as a parameter, the application of Laplace transforms has reduced
our original partial differential equation (G-6) to an inhomogeneous ordinary differential
equation in x
r
s)-
dcp - ( ~a+- cp(x,s)=</>o(x)
D-2
dx v
Boundary condition: (G-8)
We can now solve this in any of the standard ways (e.g., eigenfunction expansions or
Green's functions) to find ~(x,s), and then invert to find
cp(x,t)= e- 1 {~(x,s)}. (G-9)
Hence as should be apparent from these simple examples, Laplace transforms can be used
to greatly simplify the solution of differential equations by: (a) transforming the original
differential equation, (b) solving the transformed equation (which is now presumably a
simpler equation such as an algebraic equation or ordinary differential equation) for the
transformed solution, and (c) finally inverting the transformed solution to obtain the desired
solution of the original equation. It is usually a straightforward task to complete the first two
steps. The final step, that of inversion, can frequently be accomplished in a "cookbook"
fashion by merely looking up the inverse in a table of Laplace transforms that some other
fellow has had to work out. The general theory of how to perform such inversions from
scratch is important, however, since the inverses of many of the functions one encounters in
practice are not tabulated. However since it is heavily steeped in the theory of functions of a
complex variable, we will avoid a detailed discussion of Laplace transform inversion via
contour integration here and simply refer the reader to one of several standard texts. 1- 3
II. "COOKBOOK" LAPLACE TRANSFORMS

We will now set up the recipes for solving differential equations with Laplace
transforms. First we must determine just what types of equations we can consider:
(a) This can be any linear differential equation (ordinary or partial) in which the variable to
be transformed runs from O to oo. (Such as an initial value problem in time or a
half-space problem in space.)
(b) We will further restrict ourselves to the study of differential equations with constant
coefficients (i.e., the coefficients in the equation do not depend on the variable to which
632 / AN INTRODUCTION TO LAPLACE TRANSFORMS
we are applying the transform). This restriction can sometimes be relaxed; however we
will not consider the more general problem of differential equations with variable
coefficients here.
We will define the Laplace transform of a functionf(t) by
(G-10)
There are of course some restrictions on the type of functionf(t) and the ranges of values of
s for which this integral will be properly defined, but let's not worry about details at this
stage of the game.
The general scheme for transforming the differential equation we are interested in solving
is the same as before-namely, multiply by e- st and integrate over all t, using liberal
integration by parts. One then solves the resulting transformed equation and attempts to
invert the solution.
To facilitate in the preparation of a table of Laplace transforms (a cookbook), one merely
takes the transforms of as many different functions as possible. Several useful transforms of
general functions are: 4•5
Derivatives: e { df}
dt = sf- (s)- J(O). (G-11)
Recall that we obtained this by integration by parts. Further integration by parts yields
e{ :;t} =s'f (s)-sn-1(0)-sn-3/'(0)- .. • -pn-ll(O). (G-12)
Integration: e{ lo'dt'f(t')} = ¾J (s) (G-13)
00 00
Proof: e{lo'dt'J(t')} = lo dt e- st lo'dt'J(t') = - e:" lo' dt'J(t')I; + ¾lo dte-''f(t)
=.!J(s).
s
Differentiation bys: e{tf(t)}= - d] (G-14)

ds
Proof: d]
ds
= (
)0
00
dtf(t) !!_ (e- st ) =
ds
-1
o
00
dte- st [tf (t)].
(G-15)
Complex translation: !:{ e 'J(t)} =J (s-a)
0
dte 01 e-''f(t)= Joroo dte-<•- 0 >'j(t)=J- (s-a).

Proof:
1 0
00
Real translation: e{J(t- a)E>(t- a)}= e-asJ (s) (G-16)
where 0(t) is the step function,
E>(t)= { 6
AN INTRODUCTION TO LAPLACE TRANSFORMS / 633
Several examples of more specific transform pairs are:
f(t) j (s)= e{f(t)}

1
s+a
sin wt
s2+w2
s
coswt
s2+w2
1
8(t)
s
8 (t) 1
Several other very useful relations4 •5 are
Convolution theorem: (G-17)
(This result is useful for relating the inverse of the product of two transformed functions.)
Initial value theorem: limf(t)= limsj(s) (G-18)

1-+0 S-+00
Final value theorem: Jim f(t)= limsj (s) (G-19)

t---+OO S--+O
There are a number of reasonably complete tables of such transform pairs. 4• 5 After
obtaining the transformed solution, one can then turn to such tables in an effort to locate
the desired inverse. However in many cases it will be necessary to proceed with a direct
inversion calculation.
REFERENCES
1. P. M. Morse and H. Feshbach, Methods of Theoretical Physics, Vol. 1, McGraw Hill, New
York (1953), Chapter 4.
2. W. Kaplan, Operational Methods for Linear Systems, Addison-Wesley, Reading, Mass.
(1962).
3. H. S. Carslaw and J.C. Jaegar, Operational Methods in Applied Mathematics, Dover, New
York ( 1948).
4. P. A. McCollum and B. F. Brown, Laplace Transform Tables and Theorems, Holt,
Rinehart, and Winston, New York (1965).
5. F. E. Nixon, Handbook of Laplace Transforms, Prentice-Hall, Englewood Cliffs, N. J.
(1960).
e
~
Appendix H
Typical Nuclear Power Reactor Data
PWR PWR PWR CANDU
General Data (W) (B&W) (CE) BWR/6 HTGR LMFBR GCFR PHW
Thermal output (MW t) 3411 3600 3800 3579 3000 2410 2530 1612
Electrical output (MWe) 1150 1200 1300 1200 1170 1000 1000 500
Efficiency(%) 33.7 33.3 34.2 33.5 39.0 39.0 39.5 31.0
Fuel type UO2 UO2 UO2 UO2 UC, ThO2 PuO 2, UO 2 PuO 2 , UO2 UO2
Coolant H 20 H 20 H 20 H 20 He Na He D 20
Structural material Zircaloy Zircaloy Zircaloy Zircaloy-2 graphite 316SS 316SS Zircaloy-4
Moderator H 20 H 20 H 20 H 20 graphite D 2O
Core Data
Active height (cm) 366 363 381 376 634 91 148 410
Equivalent active
diameter (cm) 337 352 363 366 844 222 270 680
Height/ diameter 1.09 1.03 1.05 1.03 0.75 0.41 0.55 0.60
Active core volume (1) 32,800 37,600 40,000 63,910 354,000 6300 8510 130,000
Average core power
density (kW /1) 104 95.7 95 56.0 8.4 380 297 12.4
Fuel weight (kg) 90,200 94,900 103,000 138,vOO 39,000 19,000 28,000 80,000
Specific power (kW /kgU) 37.8 37.9 36.9 25.9 77 126 90 20.4
Burnup (MWD /MTU) 33,000 33,000 33,000 27,500 98,000 100,000 100,000 10,000
C.R. 0.5 0.5 0.5 0.5 0.7 1.3 1.5 0.45
Fuel Assemblies square square square canned hexagonal hexagonal hexagonal pressure
Type bundles bundles bundles -square graphite canned canned tube
bundles prisms bundles bundles bundles
Number of
Assemblies 193 205 241 732 3944 394 347 473
Fuel-element array 17X 17 17X 17 16X 16 8x8 132 pins hex hex pressure
tubes
Assembly dimension (cm)21.4X21.4 21.7 X21.7 20.3 x 20.3 14X 14 35X79 12x 12 17X 17 8x50
Assembly pitch (cm) 21.5 21.8 20.7 30.5 36.1 12.4 17.5 27.9
Number of fuel
elements/ assembly 264 264 236 63 132 217 225 28
Total number of
fuel locations 50,952 54,120 56,876 46,116 35,496 85,464 77,031 13,244
Fuel Element Data

Type Clad rod Clad rod Clad rod Clad rod Graphite UC, Wire-wrap Vented
ThO 2 rod clad rod clad rod Clad rod
Fuel-element pitch (cm) 1.25 1.27 1.28 1.62 .725 1.14 1.65
Fuel element 0. D. (cm) .94 .96 .97 1.25 1.56 .579 .805 1.52
Pitch/ diameter 1.32 1.32 1.33 1.30 1.25 1.41 1.08
Clad thickness (cm) .0572 .0597 .0635 .0864 .038 .0295 .038
Fuel-pellet diameter (cm) .819 .823 .825 1.056 1.56 .66 .739 1.44
Pellet-clad gap (cm) .0082 .010 .0089 .008 .012 .012
Fuel enrichment 2.1/2.6/3.1 2.91 1.9 /2.4 /2.9 2.2-2.7 93.5 10-15 10-15 nat U
Thermal Hydraulic Data

System pressure (bar) 155 155 155 72 50 14 86 89
Coolant flow (IG6 kg/hr) 62 68 72 47 5 50 10 23.9
Average linear power
density (W / cm) 178 178 175 206 78.7 295 217 200
Maximum linear power
density (W /cm) 426 483 410 440 229 492 390 528
Average heat flux
(W /cm2) 68.5 64 65 50.3 20.4 105 93 50
Maximum heat flux
(W /cm2) 183 168 173 111.5 58.3 237 168 115
Minimum DNBR 1.3 1.4 1.3 1.9
Inlet temperature (°C) 300° 300° 296° 269° 337° 380° 332° 249°
~ Outlet temperature (°C) 332° 333° 328° 286° 755° 552° 642° 293°
UI
Maximum fuel
temperature (°C) 1788° 2021 ° 1882° 1829° 1410° 2000° 2200° 1500°
Appendix I
Units Utilized in l~ext
BASE SI UNITS
length meter m
mass kilogram kg
time second s
thermodynamic temperature kelvin k
amount of substance mole mol
electric current ampere A
DERIVED SI UNITS
Derived SI
Unit SI Unit British Unit
Length
centimeter cm 10- 2 m 0.3937 in.
Volume
cubic centimeter cm3 10-6m3 0.06102 in. 3
liters 1 10-3m3 0.03532 ft 3
Velocity
centimeter/ second cm/s 10- 2 m•s- 1 0.03281 ft/s
Mass
gram g 10- 3 kg 2.205 X 10- 3 lb
tonne t(MT) 103 kg 22051b
Density
gram/cubic centimeter g/cm 3 103 kg/m 3 62.42 lb/ft3
Force
newton N 1 kg·m·s- 2 0.2248 lbf
Heat and energy
joule J 1 N·m 9.478X 10- 4 BTU
Power
watt w 1 J-s- 1 3.412 BTU /hr
636
Derived SI
Unit SI Unit British Unit
Pressure
bar bar 105 N·m- 2 14.50 lbs/in2 (psi)
Energy density
joule/ cubic centimeter J/cm 3 10- 6 J·m- 3 26.83 BTU /ft 3
Power density
watt/cm (linear) W/cm ID2W·m- 1 0.0305 kW /ft
watt/cm2(energy flux) W/cm2 104 W·m- 2 0.930 kW /ft 2
watt/cm 3 W/cm 3 106 W·m- 3 28.35 kW /ft 3
Temperature
degree Celsius oc °K+273.15 °F= l.8°C+32
Thermal units
thermal conductivity W/m·K 6.933 BTU /h · ft- °F
heat transfer coefficient W/m 2 ·K 0.1761 BTU/h·ft2-°F
specific heat capacity J/kg·K 2.388x 10- 4 BTU/lbm-°F
Electric charge
coulomb C A·s
NON-SI UNITS*
Electron volt (energy) eV l.60219 X 10- 19J
Atomic mass unit (mass) u l.66053 X 10- 27 kg
Bar (pressure) bar 105 N/m2
Barn(cross section) b 10 -24 cm2= 10 -2s m2
*Recognized by the International Organization for Standardization
637
Index
ABH Method, 414, 420-424 Alpha decay, 13

ABN cross section set, 363 Amplitude factor, 236
Absorption cross section, 18 Angular current density, 107-108
effective control, 543-54 7 density, 107
macroscopic, 21 flux, 107
microscopic, 18 ANISN, 452,465
Absorption heating, 379 Annular flow, 493
Absorption probability, 416-417 Asymptotic period measurement, 270
Accident, design basis, 459 Asymptotoic reactor kinetics, see Reactor
loss of coolant, 459 kinetics
Activity, radioactive, 14 Atomic mass number, 12
Adiabatic approximation, 240 Atomic number, 12
temperature feedback model, 262 Atomic number density, 14, 606-609
Adjoint equation, 222 Autocorrelation, 273-276
flux,222,224 Availability, plant, 96
matrix, 627 Avogadro's number, 605
operator, 220, 623
multigroup, 308 B1 Method, 356-358
one-speed diffusion, 221 Barn, 17
Advanced converter reactor, 86 Bessel equation, 617
Advanced gas cooled reactor (AGR), 6 function, 617-619, 208
Age, approximation, 351-353 Beta decay, 13
definition, 364 and delayed neutrons, 62-63
to indium resonance, 367 Bethe-Placzek Doppler treatment, 50-52
physical significance, 367 Binding Energy, 54-55
to thermal, 367 Blackness coefficient, 228, 41 7
Age-diffusion theory, 352, 363-368 Blanket, 95, 595
criticality condition, 374, 397 Blasius formula, 485
derivation, 351 Block diagram, 25 3
equation, 352, 363 BNL-325 (barn book), 32, 70
fast nonleakage probability, 366 Boiling, 489-498
Age theory, see Age-diffusion theory Boiling Water Reactor (BWR), 6, 91,498,521
Albedo, 165-166 Boltzmann equation, 104; see also Neutron
Algorithm, 178, 183 transport equation
Alpha, 59-60; see also Capture-to-fission ratio Bondarenko self-shielding factor, 363
639
640 INDEX
Boron, use in control, 551,554 see also specific codes

Bound atom cross section, 376 Concentration, 589
Boundary conditions, age equation, 365 Condensor, 88
derivation, 114, 140-144 Consistent P 1 method, 352
diffusion equation, 140-144, 153-156 Constant delayed neutron production rate
extrapolated, 143, 154 approx., 250
transport equation, 114 Containment structure, 92
Breeding, 69, 86-87 Continuity equation, 125; see also Neutron
ratio, 86 continuity equation
Breit-Wigner formula, 26-28 Continuous slowing down theory, 350-355
for capture, 26 Control adjustment module, 464, 550
for s-wave scattering, 29 Control cell, 155, 542, 544
Buckling, axial, 211 Control elements, 85, 95, 539-542
definition of, 202 Control management, 540
geometric, 202 Control rods, 540-542
interpretation, 202 clusters, 541
material, 203. cruciform, 540, 545
radial, 211 followers, 5 31
table, 209 grey, 542
Buckling iteration method, 527 management, 540
Burnable poison, 551-554 partial length, 550
Burnout, 490, 496-498 shutdown margin, 538
Burnup, 450, 580-588; see also Depletion stuck rod criterion, 538
worth, 224, 538, 542
Capacity factor, plant, 96 worth calculation, 542-547
Capture, see Radiative capture Conversion, 86, 589
Capture cross section, 16, 18 Conversion ratio, 86
Capture-to-fission ratio, 59-60 Converter reactors, 86
Carbide fuels, 94, 470-471 advanced, 86
Cell-averaged group constants, 409-413 Convolution theorem, 633
Center-of-mass (CM) system, 40 Coolant, 91, 95, 486-487
Chain carrier, 11, 75 channel, 95,487
Chain reaction, 11, 75-86 loops, 91
Chebyshev polynomial extrapolation, 218 primary, 91
Chemical shim, 554-556 secondary (intermediate), 92
CINDER, 601 temperature rise, 471, 487-489, 498
CITATION, 588,601 voiding, 493-495
Oad, 95 Core, 95
swelling, 457 expansion effects, 562
temperature drop, 481 homogenization, 406-409
Coarse group calculation, 291 power density, 96, 515-519
Coarse mesh rebalancing, 518 power distributions, 473, 515-519
Code for heavy isotopes, 67 see also Nuclear reactor core
Codes, 450-454, 460-465; see also Computer Critical heat flux, 4 71, 496
codes loading experiment, 268
Collision frequency, 21, 105 mass, 85
intervals, 3 28 size, 84
kinematics, 39-45 Criticality, 75-76, 202
number to thermal, 324 condition, 76, 202
probability, 419 eigenvalue, 216
rate density, 105 factor k, 75, 215; see also Multiplication
Complete set, 173, 176 factor
Compound nucleus, 24-26 search, 214-219, 301-307
cross section for formation, 24 Cross flow mixing, 501,504
formation mechanisms, 24 Cross section, absorption, 18
Computer code, 450-454, 460-465 data file, 32-34
modules, 461-462 differential, 34-39
packages, 461 double-differential, 3 8
INDEX 641
elastic scattering, 18, 28 equation, 137, 40, 153; see also Neutron
escape, 431 diffusion equation
evaluated, 34 kernel, defined, 16 8
fission, 28, 55-59 eigenfunction expansion, 175
heirarchy, 19 plane, 168
inelastic scattering, 18, 28 point, 168
kernel, 168 length, 157, 161
library handling codes, 34, 450-451 interpretation, 161
macroscopic, 20, 19-23 transport corrections, 15 9
microscopic, 17, 16-19 theory, 133, 149-226
of mixtures, 22 energy dependent, 138-140
multigroup, 288 multigroup, 285-311
nonelastic, 18, 349 one-speed, 133, 149-226
radiative capture, 26-27 validity of, 138, 152
removal, 292 Dirac delta function, 613-615
thermal, 383 Direct group coupling, 293
total, 18, 30-32 Direction of neutron motion, 35
transfer, 287, 289 Disadvantage factor, 402, 413-425, 427
transport, 136 Discrete ordinates, 11 7, 120; see also SN
2200 m/sec, 383, 606-609 method
Cross spectral density, 274 Displacement kernel, 169
CSISRS, 33 Dittus-Boelter correlation, 486
Curie, 14 DNB, 490, 496-498
Current, see Neutron current density DNB Ratio, 497
Dollar, 246
D2 0 moderated reactors, 6, 92 Doppler broadening, 49, 337
Dancoff-Ginsberg factor, 434-435 coefficients, table, 563
Darcey-Weisbach friction factor, 484 effect, 45, 48, 337
Decay chain, 14 and resonance escape, 337
Decay constant, radioactive, 13 and temperature coefficients, 345, 558-559,
Delayed critical, 246 563
Delayed neutrons, 61-65 width, 49
energy spectrum, 65 Doubling time, 100
groups, 63 Downscattering, 292,318
kernel, 248
miscellaneous data, tables, 64 Economic analysis, 458
precursors, 6 2, 23 7 module, 464
role in reactor kinetics, 79, 85, 23 7 Effective control cross sections, 543-54 7
Delta function, see Dirac Delta function multiplication factor keff, 84, 216
Density of elements, table, 606-609 neutron lifetime, 237
Departure from nuclear boiling (DNB), 490, neutron temperature model, 381-384
496-498 resonance integral, 338, 430
Depletion analysis, 449, 580-588 Eigenfunctions, 1 72
module, 464 expansions in, 171-176
single fuel isotope, 5 82-5 83 Eigenvalues, 172
Design, see specific type of criticality, 216
Detailed balance, 3 78 time, 200
Difference Equations, 118, 178-193, 517 Elastic scattering, 18
one-dimensional, 178-185 angular distributions, 43
two and three dimensional, 185-193 Breit-Wigner formula, 29
Differential scattering cross section, defined, change in neutron energy, 43-44
36 chemical binding effects, 32, 376
elastic, 36 in CM system, 40-43
Differential worth, 224 compound nucleus, 26, 28
Diffusion, 137 with Doppler broadening, 52
Diffusion approximation, 137, 139, 152 kinematics, 39-45
coefficient, 136, 139, 152 low energy, 32, 376
cooling,381 p-wave, 72
642 INDEX
resonant, 28-29, 52 Fast utilization factor, 437

s-wave, 43 Feedback effects, 241, 257, 556-563
Electrical generator, 88 mathematical description, 257-268
Emergency core cooling, 45 9 Fermi age, see Age
En functions, 547,619 Fermi age theory, see Age-diffusion theory
ENDF/A, 33 Fertile isotope, 68
ENDF/B, 34 FEVER, 517,533
Endothermic reactions, 16 f-factor, 363
Energy groups, 122, 286, 529 Fick's law, derivation, 133-139, 352
Energy levels, 14 physical interpretation, 137
Energy released in fission, 66-6 7 validity, 137
Energy spectrum, see Neutron energy spectrum Fine spectrum calculation, 291, 315, 362
Energy synthesis, 528-529 Finite difference methods, 118, 176-196
Engineered safeguards, 459 Finite element methods, 518
Enrichment, 68, 96, 589-591 Fissile nuclei, defined, 55
Enthalpy, 494 Fission, 54
Equation of continuity, 124-127 barrier, 54-55
Equivalence relations, 431 chain reaction, 11, 75
Equivalent hydraulic diameter, 485 cross section, 28, 55-60
Error functions, 619 energy release in, 65
Escape cross section, 431 fragments, 60, 66
Escape probability, defined, 417 fuels, 67
first-flight, 418-420 gamma rays, 66
reciprocity relation, 417,430 heat generation, 60,473
Wigner rational approximation, 432 mechanics of, 54
Eta (17), 68 neutrons, 4, 61
and breeding, 69, 86 products, 4, 60, 66
variation with energy, 69 decay of, 60, 66
Evaluated Nuclear Data File (ENDF), 33 gas release, 455
ENDF/A, 33 hazards of, 459
ENDF/B, 34 poisons, 567-580
Excess reactivity, 538-539 yields, 570
Excited nuclear states, 14 Q-value, 16
decay of, 15 recoverable energy, 65
mean life, 15 source term, 197, 216, 238
width, 15 spectrum, 62, 65
Exothermic reactions, 16 spontaneous, 55
Expansion in known functions, 118-120 threshold energy, 54
Exponential integral function, 547, 619 Fissionable, 5 5, 6 7
Extrapolation distance, defined, 144, 154 FLANGE, 393
for control rods, 546, 564 FLARE, 518,533
at free surface, 144, 154 Flat power, 506, 598
for fuel rods, 422, 423 Flat source approximation, 429
from transport theory, 144 Fluence, 552
Flux, see Neutron flux
Fanning friction factor, 485 Flux-power-reactivity module, 463, 515-519
Fast advantage factor, 437 Flux synthesis, 525-529
Fast breeder reactor, 6, 87, 92, 595; see also buckling iteration, 5 27
Fast reactor multichannel, 528
Fast fission factor, 82, 437 single channel, 527
Fast group constants, 315-369 spectral (energy), 528-529
Fast nonleakage probability, 83, 366 Flux time, 552; see also Fluence
Fast reactor, 87, 92, 595 Flux weighted cell averages, 409-413
feed material, 595 Forced convection, 483-489
fuel cycle, 595 pressure drop, 484,499
gas cooled, 6, 92 single phase heat transfer, 485
·sodium cooled, 6, 92 two-phase heat transfer, 490
Fast spectrum calculations, 315-369 Form losses, 500
INDEX 643
friction factor, 500 Group constants, see Multigroup constants

Four-factor formula, 83 Group diffusion method, see Multigroup diffu-
Fourier series, 174, 612 sion
Fourier's law, 152,476 Group fluxes, 122
Free atom cross section, 385 Group transfer cross section, 287, 289
Free gas scattering kernel, 3 8 8
Free path, see Mean free path Hagen-Poiseuille flow, 484
Free surface, 114, 142, 153; see also Vacuum Half-life, radioactive, 21
boundary Harmonics of flux, 173, 201
Friction factor, 484 Heat conduction, 475-483
Fuchs-Hansen model, 280 equation of, 476
Fuel, 94 Heat generation, 473-475
assembly, 94 Heavy gas model, 388-389
bundle, 94 Heisenberg Uncertainty Principle, 15
burnup and depletion, 96, 580-589 Hellstrand correlation, 435
cell, 155, 406 Helmholtz equation, 157, 206
clad gap, 479-481 Heterogeneous reactors, 398-440
cladding, 95,481 equivalent cell, 406, 408, 415
cycle, 458, 592-596 lattice, 398
densification, 532 lattice pitch, 442, 485
elements, 94 multiplication factor, 399,405
enrichment, 96, 589-591 resonance escape probability, 404,427
fabrication, 591 thermal utilization, 401, 413
failure, 4 70 High temperature gas cooled reactor (HTGR),
loading, 96 6,92,592
management, 589-600 Homogeneous reactors, 205
residence time, 96 criticality relation, 206
substitution techniques, 269 power distribution, 207, 209
temperature coefficient, 557-559 Homogenization techniques, 406-413
Functional, 622 Hot assembly, 502
Fundamental mode, 201, 235 Hot channel, 502
Fusion, 54 Hot channel factors, 502-505
clad surface, 503
Gain, 255 engineering, 504
GAM, 362 enthalpy rise, 504
Gamma decay, 13 heat flux, 503
Gamma function, 619 nuclear, 502
(,y,n) reaction, 63 axial, 503
Gap conductance, 480 radial, 503
Gas cooled reactors, 6, 92, 592; see also HTGR Hot spot, 502
Gaseous diffusion, 589 HTGR, see High temperature gas cooled
Gaseous fission products, 455, 4 79 reactor
GASKET, 393 Hydraulic diameter, 485
GATHER,452 Hydrodynamic core analysis, 498-502
Gauss' theorem, 151,612 Hydrogen, collision density in, 320
Gaussian approximation, 393 neutron moderation by, 319-326
Gaussian elimination, 183-185 neutron thermalization by, 384-388
Gauss-Seidel (successive relaxation) method, resonance absorption, 334-338
191
Generation time, see Mean generation time Importance function, 226, 310; see also Adjoint
Generalized heavy gas model, 390 Flux
Goertzel-Greuling approximation, 353-354 Incoherent approximation, 392
Graphite, 23, 30, 108, 160, 166 Inconsistent P 1 theory, 351
Green's functions, defined, 169 In-core fuel management, 596-600
finite medium, 175 constraints, 597
infinite medium, 168 refueling patterns, 598-600
Green's theorem, 612 refueling schedules, 598
Group collapsing, 291, 299 Inelastic scattering, 18, 28, 331
644 INDEX
angular distribution, 34 7 Lattice effects, 398-440

compound nucleus, 24 function, 422
energy loss in, 28, 331 pitch, 442
threshold energy, 331 see also Heterogeneous reactors
Infinite medium multiplication factor, 81, 83 Leakage,81
spectrum equation, 317 fast, 83
Infinitely dilute resonance integral, 334, 345 thermal, 83
lngroup scattering, 292 in transport equation, 112
Inhour, 244 Legendre polynomials, 616-61 7
Inhour equation, 244 LEOPARD, 517,533
Inner iteration, 193, 218 Lethargy, 322
Inner product, 220, 621 average increase (t), 324
In-out fuel loading, 599 table, 324
Inscattering term, 112 Level width, see Line width
Integral transport theory, 425-427 Lifetime, of delayed neutrons, 237
Intensity, of neutron beam, 17 in fast and thermal reactors, 236
Interaction area, 18 of prompt neutrons, 77, 204, 235
Interface conditions, 141-142, 154 of reactor transfer function, 272
Intermediate coolant loops, 92 Light water reactor (LWR), see PWR or BWR
Intragroup fluxes, 290 Line width, Doppler, 49
Inverse method, 246-249 natural, 51
Isomeric state, 12 neutron, 27
Isomers, 12 partial, 26
Isothermal temperature coefficient, 260 practical, 34 2
Isotopes, 12 radiative, 27
Isotopic depletion equations, 5 86 and resonances, 26
Isotropic source, 128, 132 total, 26
sea ttering, 12 8 Linear operators, 169
Iterated fission sources, 217 Linear power density, 96,473,478
Iterative matrix inversion, 188-192 Linear stability analysis, 266
acceleration methods, 192 Linear vector spaces, 622-623
Gauss-Seidel (successive relaxation), 191 Linearized reactor kinetics, 25 2-25 5, 264
Jacobi-Richardson (point Jacobi), 190 Linearly anisotropic flux, 134, 135, 138
Successive overrelaxation (SOR), 192 scattering, 134
Liquid metal cooled fast breeder reactor
J(f,(j) function, 336 (LMFBR), 6, 92, 595
Jens-Lottes correlation, 492 Load factor, 96
Jump condition, 142, 154 Loss of coolant accident, 459
LU decomposition, 185
k, 75, 79; see also Multiplication factor Lumped parameter approach, 236, 262
k.,,,81,83 Lyon-Martinelli correlation, 486
keff, 84, 216
Kernel, diffusion, 168 Macroscopic cross section, 20
eigenfunction expansion, 175, 176 control, 543-54 7
scattering, 168 module, 462
Kinetics, see Reactor kinetics Maneuvering control, 85, 539
Kronecker delta function, 614 Martinelli-Nelson friction multiplier, 501
Material buckling, 203
L factors, 439 Materials problems in core design, 456-458
L (t,f,(j) function, 433 Matrix, 626-628
Lab system, 40 adjoint, 627
Laplace transform, 629-633 inversion, 628
convolution theorem, 633 multiplication, 628
tables, 633 transpose, 627
Laser isotope separation, 590 Maxwell-Boltzmann distribution, 4 7, 378
LASL 16-group ... , group cross section set, Maxwellian distribution, 47,378
360,363 Maxwellian flux, 378
Lattice, 398 MC 2 (Multigroup Constant Codes), 362
INDEX 645
Mean chord length, 433, 442 Net plant efficiency, 96, 100
Mean free path, 21 Neutrinos, in fission, 65, 66
transport, 138 Neutron age, 364
Mean generation time, 238 angular current density, 107
Mean life, radioactive, 13 angularflux, 107
Mechanical core analysis, 455-456 attenuation, 20
Megawatts, electrical (MWe), 95 capture, 16, 18, 26
thermal (MWt), 95 continuity equation, 124-127
Mesh point, 178 cross sections, 16-34
Microscopic cross section, 17 current density, 109
2200 m/s values, tables, 606-609 decay, 79
Migration area, 300, 368 delayed, 61-62
Milne problem, 141, 228 density, 105
Minumum critical mass, 85 diffusion, 103, 133, 149-226
Moderating power, 325 diffusion equation, 133, 153
ratio, 325 boundary conditions, 140, 153
Moderator, 81, 95 derifttion, 133,151
table of properties, 324 energy-dependent, 138
temperature coefficient, 559 by epfunctions, 1 71
types, 95 with free surface, 16 2
Moderator-to-fuel ratio, 524, 534 multigroup, 290
Modified one-group method, 300-301 for multiregion system, 163
Monte-Carlo method, 333,414 numerical solution, 176
Moody diagram, 484 one-speed, 133,149
MUFT, 358-362 for planar source, 157
Multichannel synthesis, 528 for point source, 160
Multigroup constants, 288, 290 solutions of, distributed sources, 167
generation codes, 451-452 direction vector, 35
parameterization, 522-525 energy spectrum, 290
Multigroup diffusion method, 285-311 flux, 106
derivation, 286-295 energy-dependent, 106
diffusion equations, 287, 290 isotropic, 109
fluxes, 286, 289 and reactor power, 206, 207
group collapsing, 291, 299 therm~ 383
intragroup fluxes, 290 2200 m/sec, 383
modified one-group method, 300-301 generation time, 238
one-group method, 295 importance, 225
perturbation theory, 308 interactions of, 12-26
SN equations, 120, 122 leakage, 79
transport methods, 122 lethargy, 322
two-group method, 295-299 · lifetime, 77, 237
Multiplication factor, 75 low-energy, 376
definition, 75 mass, 605
effective, 84,216 moderation, 317-332; see also Slowing Down
four-factor formula, 83 nuclear reactions, 15-23
infinite medium, 81 prompt, 61
life-cycle viewpoint, 76 lifetime, 79
neutron balance viewpoint, 77 partial current densities, 110, 143
and reactor criticality, 76 pulsed, 272
six-factor formula, 83 scattering, see Elastic, Inelastic scattering
two group form, 297 slow, stt Thermal neutrons
slowingdown, 317-332
Narrow resonance approximation (NR), 343 slowing down equation, 318
Narrow resonance, infinite mass approximation, spectroscopy, 376
344 thermal,45
Natural harmonics, 173 thermalization, 375
Natural uranium-graphite reactor, 400 transport eguation, 113, 114
Nelkin kernel, 393 wavelength, 27, 31
646 INDEX
waves, 281 design, 96

width, 27 design data, 634-635
Neutron Spectrum, 290 fast, 6, 87, 92
effective temperature, 381 GCFR, 6, 96, 636-637
fast, 316, 315-369 heterogeneous, 398-440
thermal, 316, 375-394 homogeneous, 205
Neutron transport equation, 103-144 HTGR, 6, 96,592, 636-637
analytical solution, 129 lifetime, 551, 585
approximation of, 124 LMFBR, 6, 92,595, 636-637
derivation, 111 LWR, 6, 91,592
energy-dependent, 113 MAGNOX, 6, 92
numerical solution, 117-123 noise, 273
one-speed, 128 physics, 11
in purely-absorbing media, 129-133 prompt critical, 246
Neutronics, 11 prompt subcritical, 246
Newton's law of cooling, 262, 481 PWR, 6, 91, 636-637
Nodal methods, 194-196 reflected, 211
cells, 194 SGHWR,6, 92
coupling coefficients, 195 slab, 198 -
in kinetics, 277 temperature coefficient, 259, 556
transfer probabilities, 19 5 theory, 11
Noise, reactor, 273-276 thermal, 5, 81
Nonelastic cross section, 273, 349 transfer function, 253, 264
Nonleakage probability, 79, 83 see also specific reactor types
fast, 83 Nuclear Recoil, 44
thermal, 83 Nuclear steam supply system (NSSS), 5, 91
Nonlinear point reactor kinetics, 267-268 Nucleate boiling, 490
Non-1/v absorption, 383 Nucleon, 12
Non-1/v factors (g-factors), 383 Nucleus, binding energy, 54
Nonreentrant surface, 114,153 compound, 24
Norm, function, 621 excited states, 14, 24
· matrix, 190 radius, 29
oper!ltor, 623 Nuclide, 12
Normal modes, 173; see also Eigenfunctions Numerical methods, 117-123, 176-196
Nu (v), 61 quadrature, 118
artificial, 215 Nusselt number, 486
energy dependence, 62
Nuclear constants, see Cross sections One dimensional diffusion codes, 17 6-185
Nuclear Data Sets, 32-34 One group method, 295; see also One speed
Nuclear fuel management, 589-600 approximation
Nuclear methods development, 97 One speed approximation, derivation, 128,
Nuclear motion, 45-54 150,295
Nuclear power peaking factors, see Hot modified, 300
channel factors One speed diffusion equation, 137, 149-226
Nuclear reactions, 15; see also Cross sections derivation, 150, 133-138
collision, 15 Operator, 169, 622
kinematics, 39 adjoint, 220, 623
nomenclature, 15 Laplacian, 156, 611-612
Nuclear Reactor, 4 multigroup diffusion, 303
blanket, 95 perturbation, 219
breeder, 87, 92,595 self-adjoint, 221, 624
BWR, 6, 91, 636-637 Optical thickness or depth, 132
CANDU, CANDU-PHW, CANDU-BLW, 92, Orthogonality, 173,621
636-637 Orthonormal, 176
control, 76 Outer iteration, 218
converter, 86 extrapolation, 218
core, 85, 95 multigroup, 305
criticality of, 76 one-speed, 218 ·
INDEX 647
Overlapping group technique, 529 local, 502, 529-532

Overrelaxation, 192 Power regulation, 539
Oxide fuels, 94 Practical width, 342
Prandtl number, 486
P1 approximation, 121, 136 Precursor, delayed neutron, 62, 237
P1 equations, 121, 135-136 balance equation, 238
lethargy dependent, 34 7 concentration, 237
slowing down equations, 349 Pressure vessel, 91, 9 5
PN method, 121 Pressurized Water Reactor (PWR), 6, 91
equations, 121 Primary model, 390
Parabolic partial differential equation, 156 Probability distribution function, in neutron
Parks' kernel, 393 transport, 21
Partial current, 110,143,153; see Neutron in radioactivity, 13
partial current density scattering, 39
Particular solution, 170 Prompt critical, 246
Pcm (per cent mille), 244 Prompt generation time, see Mean generation
PDQ, 513, 533 time
PDQ-HARMONY, 588,601 Prompt juror approximation, 25 0-251
Period, 78, 244 Prompt neutrons, 61
Periodic power variation, 248 average energy, 61
Permanent fission-product poisons, 577 lifetime, 77, 204
Perturbation theory, applications of, 219 most probable energy, 61
first-order, 223 source term, 116
multigroup, 308 spectrum, 61
one-group, 219-226 Prompt subcritical reactor, 246
two-group, 309 Prompt temperature coefficient, 260, 559
Phase shift, 255 Proton gas model, 384-388; see also Wigner-
Photofission, 54 Wilkins model
Photoneutron reactions, 63 Pulsed neutron measurements, 272-273
Physical constants, table, 605 Pumping power, 487
Pitch, 442, 485 p-wave scattering, 72
Placzek function, 329
Plane symmetry, 115 Q-value, 16
Plant availability factor, 96 Quadrature weights, 118
Plant capacity factor, 96 Quality assurance, 7, 460
Plant electrical output (MWe), 95 Quality, flow, 495
Plant load factor, 96 static, 493
Plant thermal output (MWt), 95 Quasistatic method, 277
Plutonium, 55, 67
Point Reactor kinetics model, 235-241 Radiation damage, 456
equations,236,238,239 Radiative capture, 16, 18, 26
integrodifferential form, 248 Breit-Wigner formula, 26
limitations, 239 with Doppler broadening, 48
Point source, age theory, 367 see also Resonance escape probability
diffusion theory, 160 Radiation from a reactor, 66
in purely absorbing media, 130-132 Radioactivity, 12-15
Poisons, 551, 567 activity, 14
burnable, 551-554 chain decay, 14
see also Fission product poisons decay constant, 13
Potential scattering, 23, 29 half-life, 13
Power, see Nuclear reactor power mean-life, 13
Power coefficient of reactivity, 261, 55 7 Rational approximation, 432
Power defect, 258, 261 Reaction rate density, 105
Power distribution, 515-533 Reactivity, 222,239, 537-563
calculation of, 515-532 analysis, 449, 537-563 _
shaping, 521,547 and control rod worth, 538
Power method, 216, 218 excess reactivity, 5 3 8, 5 85
Power peaking factor, 230,475,502 feedback, 257-268, 556-563
648 INDEX
and fission product poisoning, 567 mation, 344

and fuel burnup, 580 numerical computations, 342 l
inherent, 556 temperature coefficient of, 345, 556
I
inventory, 538; see also Excess reactivity temperature dependence of, 337
and perturbation, 222 widely spaced resonances, 344
power coefficient of, 261, 557 Resonance integral, 338
ramp change in, 249, 251 effective, 340
spectral density, 2 74 empirical formulas for, 435
step change in, 24 2 equivalence theorems, 431
temperature coefficient of, 259-262, 556-563 for heterogeneous reactol"s, 427-437
Reactor, see Nuclear reactor for homogeneous reactors, 338
Reactor cycle, 598 physical significance, 341
Reactor design, 96, 447-465 table, 346
design data, 634-635 temperature coefficient of, 345, 558
Reactor dynamics, 233, 257-268 used to calculate group constants, 341
Reactor kinetics, 78, 233-277 Reynold's number, 484
feedback effects, 257-268, 556-563 Ricatti equation, 386
Fuchs-Hansen model, 280 Rod bowing, 562
large negative reactivities, 246 Rod drop method, 270
period, 78, 244 Rod oscillator experiment, 271
point reactor model, 235-241 Rod shadowing, 434
prompt critical, 246 Rod worth, 538, 542-547
prompt jump approximation, 250-251 Roundelay fuel loading, 599
ramp reactivity insertion, 249, 251
scram, 250 S-a-x relationship, 495
and small reactivities, 245 Safety, 459
stable period, 243 St. John-Brown kernel, 393
transfer function, 25 3, 264 Samarium, 576-577
Reactor period, 78, 244 at constant power, 576
and control rod worth, 270 decay constant, 570
and reactivity, curve, 24 7 equilibrium, 576
stable, 243 fission yield, 570
Reactor Power, 206,473, 515-533 after shutdown, 577
Reciprocal multiplication method, 268 Scalar flux, 109
Reciprocity theorem, 417,430 Scalar product, 621; see also Inner product
Recoil, nuclear, 44 Scattering, 16, 18
Reduced mass, 41 angle of, 37, 43
Reduced wavelength, 27 angle cosine, 3 8
Reentrant surface, 114 coherent, 376
Reflected reactor, 211 cross section, 28
Reflection coefficient, 165 elastic resonance, 28
Reflector, 95, 165 interference, 29
Reflector savings, 213 kernel, 168
Refueling, 598 law, 392
Removal cross section, 292 potential, 29
Resonance, 24-26 p-wave, 72
data for 2 38 U, table, 335 resonance, 29
Doppler broadening, 49 s-wave,43
and level width, 26 upscattering, 5 3
practical width, 342 see also Elastic; Inelastic scattering
Resonance absorption, 26,49, 332-347,427- Scatter-loading, 599
437 Scram, 539
infinite dilution approximation, 334 control, 85, 539
in heterogeneous reactors, 427-437 Secondary model, 391-392
in homogeneousreactors, 332-347 Selengut-Goertzel method, 352
in hydrogeneous media, 334-338 Self-adjoint operator, 221, 624
narrow resonance approximation, 343 Self-shielded group constants, 363, 409
narrow resonance-infinite mass approxi- Self-shielding, 229, 338
INDEX 649
energy, 338 Temperature coefficient of reactivity, 259,

factor, 229,363,412,553 556-563
spatial, 404 coolant, 559
Separation of variables, 199 fuel, 558
Shape factor, 236 isothermal, 260
Shim control, 85, 539 moderator, 559
Shutdown (of reactor), 245,250,538 prompt, 260
and fission product buildup, 572 and reactor stability, 260
Shutdown margin, 538 Temperature defect, 261
Simple degenerate kernel (SDK), 396 TEMPEST, 452
Six-factor formula, 83 Thermal analysis, see Thermal-hydraulic
Slip ratio, 495 analysis
SN method, 120-121 Thermal conduction, 4 75-483
equations, 120 conductivity, 476
Slowing down, 317-332; see Moderation in fuel elements, 558
Slowing-down density, 321 Thermal core analysis, 467-511
with absorption, 325 Thermal disadvantage factor, 402
anisotropic component, 348 Thermal group constants, 275, 316
and flux, 321 Thermal-hydraulics analysis, 46 7-511
in hydrogen, 319 design codes, 507
SOFOCATE, 387 limitations, 4 70
Solid angle, 218 module, 463
Source extrapolation, 218 neutronics interactions, 519-5 21
Source iteration method, 216-219, 301-307 objectives, 469
Source jerk method, 271 Thermal limitations, 469-4 73
Spatial modes, 176 clad, 470
Specific enthalpy, 494 coolant, 4 71
Specific power, 96 fuel, 470
Spectral synthesis, 528-529 Thermal neutrons, average energy, 379
Spectrum calculations, see Neutron energy cross sections of, 383, 375-394
spectrum diffusion of, 379
Spherical coordinates, 35, 612 flux,383
Spherical harmonics, 120 interaction rates, 382
Spontaneous fission, 10, 55 migration area, 368
Stability, 254 most probable energy, 379
in the small, 267 nonleakage probability, 368
Steam generators, 5, 91 spectra, 316, 375-394
Step function, 613 speed of, 379
Strawbridge-Barry correlation, 436 Thermal power (MWt), 95,473,505
Streaming term, 127 Thermal reactors, 5, 81
Structure, 95 Thermal spectrum calculations, 316, 375-394
Stuck Rod Criterion, 538 nonequilibrium, 379
Subcritical assembly, 76 thermal equilibrium, 378
Successive overrelaxation, 192 Thermal transport path, 468
Supercritical reactor, 76 Thermal utilization, 80, 82
S-wave scattering, 43 by ABH method, 420
Swelling in fuel, 455,480 in heterogeneous reactor, 403
in stainless steel clad, 456 Thermalization, see Neutron thermalization
Synthesis methods, 277, 525-5 29; see also Thermally averaged interaction rates, 45, 382
Flux synthesis Thermodynamic efficiency, 96, 100
Synthetic scattering kernels, 389, 396 THERMOS, 425
Three-point difference equations, 180-185
Taylor series expansion, 351,612 Tilt in power, 579
Temperature, of clad, 481 Time behavior of reactor, see Reactor kinetics
of coolant, 487 Time eigenvalues, 200, 235
of fuel, 477 Total cross section, 18, 30
of moderator, 559 Total flux, 109
of neutrons, 381 Transfer function, 253, 264
650 INDEX
closed loop, 264 Variation of constants, 170

with feedback, 264 Variational me·,i1ods, 120, 232
zero-power, 253 Void coefficient of reactivity, 560-561
Transport cross section, 136, 349 Void fraction, 493
Transport equation, 111-117; see also
Neutron transport equation w-search, 439
Transport kernel, 4 26 Wavelength, neutron, 31
Transport laws, 152 reduced, 27
Transport mean free path, 138 Way-Wigner correlation, 604
and diffusion coefficient, 136 Weighted residual methods, 120, 527
and extrapolation distance, 144 Width, see Line width
Transport theory, 103-144 Wigner rational approximation, 432
derivation of transport equation, 111-117 Windowshade model, 549
Fick'slaw, 137,139,152 Wigner-Wilkins equation, 386
one-speed transport equation, 128
Trial functions, 23 2 Xenon, 568-580
Tridiagonal matrix, 183,626 at constant power, 571
Turbine, 88 cross section, 1 7, 25, 566
Two-body collision kinematics, 39-45 decay constant, 570
Two-group method, 295-299 equilibrium, 571
criticality condition, 297 fission yield, .570
group constants, 296 oscillations, 578
and perturbation theory, 309 transients, 574
operator, 309 after shutdown, 573
2200 meter/sec cross sections, 383, 606-610 Xi(O, 324
for mixtures, 331
Ultracentrifuges, 590 table, 324
Unit cell, 155,229,406
Unit impulse response, 254 Yellow cake, 589
Unstable, 254, 266 YOM cross section set, 363
Upscattering, 53
Uranium, 55-70, 589 Zero-dimension calculation, 517, 588
enrichment, 68, 96, 589-591 Zero-power point reactor model, 241-257
hexaflouride, 589 Zero-power transfer function, 253
oxide, 94, 591 Zonal loading (in-out cycling), 599

(Duderstadt, Louis-J. - Hamilton) Nuclear Reactor Analysis (1976)

Uploaded by

Copyright:

Available Formats

(Duderstadt, Louis-J. - Hamilton) Nuclear Reactor Analysis (1976)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Duderstadt, Louis-J. - Hamilton) Nuclear Reactor Analysis (1976)

Uploaded by

Copyright:

Available Formats

Nuclear

. Department of Nuclear Engineering

JOHN WILEY & SONS, Inc.

Duderstadt & Hamilton (1976 ed.)

Page Error Correction

Pg 244, after eq 6-43 𝑠0 𝑠1

Pg 244, line 14 |𝑠𝑜 | |𝑠1 |

ΣaP ⁄Σ𝑎 − ΣaP ⁄Σ𝑎

Pg 607, Σ𝑎 for Cu 0.0326 0.326

Library of C:O,,gnn Catologi,lg in hblicatitR, Doto:

A. H-S. Ang Civil Engineering-Systems

The maturation of the nuclear power industry, accompanied by a redirection in

ANN ARBOR, MICHIGAN

CHAPTER 1 An Introduction to Nuclear Power Generation 3

I.VI Nuclear Fission Reactors 4

CHAPTER 2 The Nuclear Physics of Fission Chain Reactions 10

CHAPTER 3 Fission Chain Reactions and Nuclear Reactors-

I. The MultipJication Factor and Nuclear Criticality 74

CHAPTER 4 Neutron Transport 103

I. Introductory Concepts 105

CHAPTER 5 The One-Speed Diffusion Theory Model 149

I. The One-Speed Diffusion Equation 150

CHAPTER 6 Nuclear Reactor Kinetics 233

I. The Point Reactor Kinetics Model 235

CHAPTER 7 Multigroup Diffusion Theory 285

I. A Heuristic Derivation of the Multigroup Diffusion Equations 286

V. Multigroup Perturbation Theory 308

CHAPTER 8 Fast Spectrum Calculations· and Fast Group

I. Neutron Slowing Down in an Infinite Medium 317

- CHAPTER 9 Thermal S ~ Calculations and Thermal

I. General Features of Thermal Neutron Spectra 377

CHAPTER 10 Cell Calculations for Heterogeneous Core Lattic~ 398

I. Lattice Effects in Nuclear Reactor Analysis 399

CHAPTER 11 General Aspects of Nuclear Reactor Core Design 447

I. Nuclear Core Analysis 448

CHAPTER 12 Thermal-hydraulic Analysis of Nuclear Reactor Cores 467

III. Radial Heat Conduction in Reactor Fuel Elements 475

CHAPTER 13 The Calculation of Core Power Distributions 515

I. Static Multigroup Diffusion Calculations 516

CHAPTER 14 Reactivity Control 537

CHAPTER 15 Analysis of Core Composition Changes 566

I. Fission Product Poisoning 567

· A. Some Useful Nuclear Data 685

I. Miscellaneous. Physical Constants 605

B. Some Useful Mathematical Formulas 611

C. Step Functions, Delta Functions, and Other Exotic Beasts 613

D. Some Properties of Special Functions 616

E. Some Assorted Facts on Linear Operators . 621

I. Scalar Products 621

F. An Introduction to Matrices and Matrix Algebra 626

I. Some Definitions 626

G. An Introduction to Laplace Transforms 629

H. Typical Nuclear Power Reactor Data 6.14

I. Units Utilized in Text 6.16

I. NUCLEAR FISSION REACTORS

II. ROLE OF THE NUCLEAR ENGINEER

III. SCOPE OF THE TEXT

References on nuclear reactor theory include:

presentation is not intended to be complete. Indeed we would anticipate that most