ELSEVIER Journal of Nuclear Materials 216 (1994) 97-123

Theory and experimental background on dimensional changes

in irradiated alloys *
L.K. Mansur
Metals and Ceramics Division, Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, TN 37831, USA

Abstract

When a material is irradiated with energetic neutrons or charged particles, a complex sequence of reactions takes
place. Structure, composition, and properties are altered over an extremely wide scale, spanning atomic defects,
meso-scale microstructures and macroscopic properties. Particularly interesting are radiation-induced dimensional
changes, Such changes, which can occur in engineered components of fission and fusion reactors on a scale of
meters, are driven by lattice defects at the subnanometer level. Because of their technological importance and their
high scientific challenge, the dimensional changes termed radiation-induced swelling and creep have elicited
sustained intensive research by basic and applied materials scientists for many years. The present paper is intended
as a brief tutorial on salient features of this work. The presentation is divided into three parts. Background is first
sketched emphasizing experimentally observed features and applications. Next, the theoretical framework and
specific models that have been developed to understand radiation-induced swelling and creep in isotropic materials
are described. Lastly, selected experiments designed and/or interpreted in terms of theory are highlighted to
illustrate the current state of understanding of the physical bases of these phenomena.

1. Introduction sional instability. Dimensional changes include swelling,

shrinkage, creep and growth. The first two are volume
Materials irradiated with energetic particles re- changes and the last two are volume-conserving shape
spond by undergoing changes in structure and proper- alterations. Several processes may occur together.
ties. Irradiating particles transfer energy to the mate- Swelling can occur by the formation of cavities or of
rial primarily by displacement of atoms from their less dense phases or, to a smaller extent, by the reten-
original sites and by ionization and electronic excita- tion of atomic lattice vacancies. Shrinkage involves the
tion. In general, all processes and properties of the formation of more dense phases or the actual loss of
material potentially are affected. These include, for material. Such loss can be caused by surface processes
example, internal microstructures and phase distribu- like sputtering, or by volume processes leading to the
tions, dimensions, mechanical properties, electrical, and release of atoms and molecules that diffuse out of the
magnetic properties, and corrosion. material. The latter processes are not important in
Particularly intriguing, because they may occur on structural alloys or ceramics but can be prominent in
an ordinary human scale, are changes in dimensions. irradiated polymers, for example. Irradiation creep is a
Gross changes in size and shape often occur as a result stress-directed shape change that occurs under irradia-
of irradiation; these phenomena are termed dimen- tion in addition to ordinary thermal creep. Similarly,
irradiation growth is a shape change caused by irradia-
tion in anisotropic materials. It does not require the
application of a stress.
* Research supported by the Division of Materials Sciences, Bombarding particles that produce such changes
U.S. DOE under Contract DE-ACO5-840R214CKl with Martin include neutrons, heavy ions, light ions, electrons and
Marietta Energy Systems, Inc. y-rays. Depending upon their energies, all of the above

OO22-3115/94/$07.00 Q 1994 Elsevier Science B.V. All rights reserved

SSDI 0022-3115(94)00321-E
98 L.K. Mansur/Journal of Nuclear Materials 216 (1994) 97-123

particles can displace atoms in a material by elastic developed to understand the dimensional changes, are
collisions, by other types of nuclear reactions, or, in the described. This section includes necessary background
case of y-rays, by the production of reaction electrons. on microstructures and mechanisms. The following
The primary knock-on atoms (pkas), ions and electrons part, Section 4, describes direct comparisons of the
also can cause extensive ionization and electronic exci- theory with special experiments designed to probe
tation. Generally, neutrons and heavy ions in the MeV mechanisms. Such comparisons have been key to devel-
range impart so much energy to the pka that a dis- oping the present high level of understanding of di-
placement cascade is produced consisting of highly mensional changes under irradiation.
localized production of interstitials and vacancies asso- The present paper is intended to be tutorial in
ciated with a single initiating event. The processes and nature, for readers interested in the above topics. By
reaction pathways by which the displacement and elec- reference, it incorporates an extended book chapter by
tronic energy are dissipated determine the structure the author [l], that more fully develops the theory and
and property changes exhibited by the material. the experimental background, and contains a more
In the present paper we concentrate on swelling complete set of references.
and creep in metals and alloys. Related phenomena in
a range of these and other materials must be excluded
because of length limitations. The presentation is given 2. Experimental background
in three major subdivisions. First, in Section 2, basic
experimental background and facilities are described. 2.1. Dimensional changes
This includes a sketch of the historical origins of re-
search on swelling and creep. Next, in Section 3, major Swelling was discovered in 1967 during electron
elements of the kinetic theory of radiation effects, microscopy investigations of a stainless steel irradiated

Fig. 1. Photograph of 316 stainless steel rods before and after irradiation at 533°C to a fluence of 1.5 X 10z3 n/m* in the EBR-11
reactor.
 L.K. Mansur/Jounuzl of Nuclear Materials 216 (1994) 97-123 99

in a fast reactor [21. Cavities were found in specimens

irradiated to doses greater than 1O26n/m* at tempera-
tures in the range 400-610°C. The swelling was in
excess of 1% in some cases. In their paper, Cawthome
and Fulton showed that the cavities could not be
helium-supported bubbles, because in the fast reactor
too few helium atoms were produced by the (n, a)
reactions. However, they suggested that the helium did
serve in getting swelling started by providing a nucle-
ation site for the condensation of excess radiation
produced vacancies. In our later development of the
theory, we will see that helium is indeed crucial in
controlling swelling. Fig. 1 is a photograph of a highly
0 50 100 150 200 250 300
irradiated stainless steel specimen showing extensive
swelling. We now know that swelling can reach many TIME (h)
tens of percent in alloys not designed to be swelling Fig. 2. In-reactor deformation as a function of time for
resistant. uranium at 220°C irradiated in a flw of 6 x 1016 n/m2 s. Solid
Irradiation creep was discovered in 1955 in the curves indicate when stress and irradiation were applied si-
fissile metal uranium [3]. In 1959, irradiation creep was multaneously. Dashed sections indicate where reactor was not
reported in a stainless steel [4]. Fig. 2 from Refs. [3,5] operating. After Ref. [5].
shows the essential behavior characterizing irradiation
creep. When irradiated under stress, the material shows
significant deformation. However, when either the 2.2. Facilities and applications
stress was removed and irradiation continued, or the
irradiation is interrupted while the applied stress is In the nuclear technologies based on fission and
maintained, no creep was seen. fusion reactors, interest in dimensional stability centers

Fig. 3. Artist’s rendition of the Experimental Breeder Reactor-II (EBR-II). Shown are the reactor vessel, internals, and reactor
core. EBR-II is located at Idaho Falls, Idaho.
100 L.K. Mansur/Journal of Nuclear Materials 216 (1994) 97-123

Fig. 4. Artist’s conception of the International Thermonuclear Experimental Reactor.

Fig. 5. Artist’s rendition of the multiple ion laboratory at Oak Ridge, TN. The three beam lines are configured so that the ion
beams can be impinged on a target simultaneously.
 L.K Manmr/Joumal of Nuclear Materials 216 (1994) 97-123 101

mainly on structural metallic alloys. Thus, although

dimensional change occurs in many types of materials,
including fissile metals and polymers as outlined above,
as well as in ceramics and semiconductors, for exam-
ple, we focus on structural alloys. In particular, Fe-
Ni-Cr base austenitic alloys have received most atten-
tion because of their prominent application in fission
and fusion reactor systems.
Of the fission reactor systems, the liquid metal
cooled fast breeder reactor occupied much of the inter-
est in past years. However, gas-cooled fast and thermal
reactors and water-cooled reactors also suffer some
problems from radiation-induced dimensional instabil-
ity. Fig. 3 shows the EBR-II, an experimental liquid-
metal-cooled fast flux reactor, operating at Argonne
National Laboratory-West in Idaho. The cladding of
the fuel elements and the core structural components
are exposed to high fluxes of fast neutrons.
Fusion reactors, which are now in the research,
design and technology development stages, will have a
large mass of structural material exposed to fast neu-
trons, with a significant proportion of the neutrons at Percent Swelling 0s F”nctlonr Of Tempemlwe mid FluenCe

an energy of 14.1 MeV. Fig. 4 shows an artists concep-

Fig. 6. Three-dimensional trend plot based on experimental
tion of the current design of the International Ther- data showing swelling as a function of temperature and neu-
monuclear Experimental Reactor (ITER). Fusion reac- tron fluence in a fast reactor. After Ref. [ll].
tor technology is being developed throughout the world,
and a significant fraction of the effort is devoted to
materials characterization and design for radiation en- after high fluences of fast neutrons 1111. The signifi-
vironments. Since fusion reactors have not yet come cance of high swelling levels in applications is that: (1)
into existence, a variety of experimental techniques, engineered shapes and dimensions can become grossly
including irradiations in fission reactors, is being ap- unstable; (2) non-uniform swelling that results from the
plied to learn about the potential behavior of materials sensitivity of swelling to temperature and dose rate, for
in fusion reactors. example, can lead to bowing of components and to
Another way to obtain radiation effects information buildup of stresses; and (3) the lifetimes of components
for both basic mechanistic investigations and for tech- may be severely limited if operational problems are to
nology applications is through the use of accelerators. be avoided.
Fluxes of ions, including self-ions of structural materi-
als (such as Fe and Ni), typically employed at energies
of several MeV or higher, can produce radiation dam-
age and can induce qualitatively similar phenomena, “RI] RELIEVED BY IRRADIATION CREEP
such as swelling and creep, as do neutron fluxes [6,7].
Fig. 5 shows the triple ion facility at ORNL. The three
accelerators can accelerate a wide variety of ions to C
: 150
energies up to 5 MeV. Beams can be applied singly or
simultaneously to a specimen. The facility has been a %
workhorse for radiation effects research [8]. Accelera- p 100
tors such as the facility illustrated in Fig. 5, have been iii
used by researchers to make major contributions to the 50
fundamental understanding of radiation effects in ma-
terials. For example, recent conferences on fusion re-
01
actor materials contain major blocks of papers from 0 5 10 15
the international community on the applications of DOSE (dpa)
accelerators [9,10]. Fig. 7. Calculated stress relaxation as a function of dose in a
The importance of dimensional instability to tech- tightened bolt irradiated in a fast reactor. Stress relaxation
nological applications can be illustrated by Figs. 1, 6, rate based on measured irradiation creep rates, after Ref.
and 7. Fig. 6 shows the swelling levels that can result 1121.
102 L.K. Mansur/Joumal of Nuclear Materials 216 (1994) 97-123

Irradiation creep can help accommodate stresses these property changes are shown. The striped ellipti-
induced by differential swelling. If it is too large, cal images are interstitial dislocation loops formed by
however, irradiation creep can lead to buckling of the planar aggregation of interstitials. The light images
components. Another significant consequence of irra- are cavities, a new ‘phase’ consisting of empty space,
diation creep is that it can rapidly undo engineered introduced by irradiation. Cavity volume in aggregate
stress distributions. Thus, for example, clamps or other represents the macroscopic swelling of the crystal; the
devices designed to maintain a stress under normal empty atomic lattice sites in the cavities account for an
conditions would be quickly rendered ineffective under equivalent enlargement of the crystal by the addition of
irradiation. Fig. 7 illustrates this point. A tightened new atomic lattice sites. The darker images are second
bolt irradiated under the typical conditions of a fast phase precipitates, in general of different structure and
reactor would loosen its hold after a small dose [12]. composition than the matrix. Tangled linear images in
the micrograph are dislocation lines.
2.3. Structure and composition The dislocations are major participants in the pro-
cesses of swelling and creep. An extra lattice site
Wholesale development of structure and composi- created by the attachment of an interstitial to a dislo-
tion on near-atomic and larger scales, often referred to cation, corresponding to an empty site in a cavity,
as ‘microstructure’ and ‘microcomposition’, is responsi- causes climb of the dislocation by one lattice site.
ble for swelling, creep and other property changes. Fig. Irradiation creep can occur by the stress-directed glide
8 is a transmission electron micrograph of a 300 series of dislocations released from obstacles by climb result-
stainless steel irradiated at about 500°C to a dose of 10 ing from the absorption of point defects. This is termed
dpa [13]. Many of the extended defects that underlie climb-enabled glide creep. Another type of creep, climb

Fig. 8. Transmission electron micrograph of a 300 series stainless steel irradiated at 500°C to a dose of 10 dpa. After Ref. [13].
 L.R Mansur/Joumal of Nuclear Materials 216 (1994) 97-123 103

creep, can occur as a result of the different rates of

climb for dislocations with Burgers vectors at different
orientations to the applied stress. A taxonomy of these
TLC ==v
and several other mechanisms of irradiation creep is NEUTRON
10'eV
CASCADE CItEM-ION l(1” I
given in Ref. [14]. 4 ,

Precipitates, examples of which can be seen in Fig. UNSTABLE htATRIx 10” 6 DISPLAcaD PRIMAaY lo%V
1 *
8, may affect swelling and creep in several ways. These INIzRsTlTlAL DIFFUSION 10' I DISPLKED SECONDARIES 10’ eV
1 I
ways include what can be termed direct, indirect and
VACANCY DIIWJSION 1 I UNSI-ABE h5Allux 1 CV
mediated effects [15]. An example of a direct effect 1 &
would be the attachment of a cavity to a precipitate. MIcROSt’R”cIuRAL EVOLUTION 1~7’
I THERMAL DIFFUSION kT

During void growth, the cavity attached to the precipi-

tate will grow more rapidly than one in the matrix Fig. 9. Characteristic time and energy scales pertaining to the
fates of point defects and to the microstructural changes for
because the precipitate-matrix interface collects point
which they are responsible. After Ref. [ll.
defects for the cavity [16]. An example of an indirect
effect on swelling or creep would be the absorption of
point defects at an unattached precipitate, in turn to a sink is 10e6 s. Vacancy diffusion takes longer,
affecting the fluxes of point defects to cavities and typically between 10m3 to 10’ s at elevated tempera-
dislocations. Physical changes in the matrix as a result tures. Evolution of a microstructure of extended de-
of precipitation by the depletion of solutes, for exam- fects like dislocation loops, cavities and precipitates,
ple, can be characterized as a mediated effect. The driven by persistent long-term point defect fluxes, takes
altered matrix can affect swelling and creep in turn several orders of magnitude longer. The time and
through changes in point defect trapping, changes in energy regimes prior to the thermal diffusion of point
diffusivities, and changes in other properties. defects has been well-described in other papers in
Precipitation can be induced, enhanced or sup- these proceedings [20-241. The present paper empha-
pressed by irradiation. For example, precipitates that sizes the kinetics of point defect clustering and mi-
are ‘wrong’ for the bulk matrix composition under crostructural evolution; our starting point is thermal
normal thermal equilibrium conditions are often found point defect diffusion.
in irradiated materials. They may be formed because of In the kinetic theory, the central theme is to accu-
the development of local micro-alloys that are pro- rately account for the reactions of point defects with
duced by irradiation near point defect sinks, such as each other and with solutes, clusters and extended
cavities, dislocations, dislocation loops, grain bound- point defect sinks in the matrix. In the simplest embod-
aries, surfaces as well as other precipitates. Such local iment of the theory, the point defect production pro-
alloys can exhibit enrichments or depletions in certain cesses are assumed to yield a given generation rate per
elements by orders of magnitude relative to the bulk unit time, distributed uniformly throughout the mate-
composition. This phenomenon of solute segregation rial. In effect, all of the point defect production physics
under irradiation has received a great deal of attention is ultimately subsumed in a source term in homoge-
in radiation effects research [17-191. neous rate equations. This overall generation rate is
certainly the most important characteristic of the point
defect production. Other aspects have been incorpo-
3. Theory rated into the kinetic theory as well. These include the
spatially non-homogeneous production in cascades [25]
3. I. Framework and the direct in-cascade formation of small point
defect clusters [26]. Non-homogeneous production can
In beginning a discussion of the kinetics of radiation affect processes that are sensitive to concentration
effects, it is useful to draw a perspective in terms of fluctuations. Clustering can lead to different fractions
time and energy scales. Fig. 9 summarizes important of migrating and clustered defects for vacancies and
time and energy regimes [l]. The left-hand side shows interstitials, and the clusters themselves can act as
markers in the time evolution sequence of microstruc- sources and sinks for point defects.
ture. The right-hand side shows significant stages in Fig. 10 presents a flow diagram summarizing the
the degradation of energy of a fast neutron interacting fates of defects, starting at the top with the creation of
with the material. the initial displacements. At the bottom, the defects
Cascades are created in time of order lo-l3 s. The will have followed one of two broad branches, either
affected matrix is unstable for about one hundred leading to changes in properties or to a null effect on
times longer. Short term annealing and related pro- many properties. Between the top and bottom of the
cesses take place subsequently. A typical time interval diagram are shown several paths and processes that
between the creation of an interstitial and its diffusion the defects may follow. Therefore, Fig. 10 is essentially
104 L.K. Mansur /Journal of Nuclear Materials 216 (1994) 97-123

DISPLACEMENTS swelling and irradiation creep built on this foundation.

1 The present description is based on the comprehensive
AVAILABLE DEFECTS
treatment of Ref. [l].
DIFFUSION OF VACANCIES
1 AND INTERSTITIALS A general approach to clustering begins with the
discrete master equations

%=Erv(k;j)C,- &v(j;k)Cj+Gj-L,. (1)

k k
FORYAilON OF ABSORPTION AT SiLK
EXTENDED DEFECTS EXIST,NG SlNKS RECOMBINATION Here Cj is the concentration of clusters of type j. The
transition rate w(k; j) is the rate per unit concentra-
tion of cluster type k of transitions to a cluster of type
j. The summations denote all possible reactions leading
to a cluster of type j and all possible reactions remov-
ing a cluster of type j, respectively. The term Gj is the
RECWSINATION
rate at which clusters of type j are formed directly,
excluding the diffusive encounter reactions incorpo-
rated in the first summation. This includes direct for-
OAhGE DAMAbE IS
ACCUMULATES ANNIHILATED mation in cascades as well as thermal emission of
+ defects from sinks. The term Lj is the loss rate of
HARDENING,
NO CHANGES
IN PROPERTIES
clusters of type j, such as by diffusion to sinks, includ-
CREEP,.
ing large cavities and dislocation lines, that are not
SOLUTE REDISTRISUTION
counted as clusters in the second summation, as well as
Fig. 10. Flow diagram summarizing several main conceptual by direct destruction in cascades, for example. To solve
elements in the evolution of radiation effects from the initial these equations, knowledge of many mechanisms is
displacement interaction. This diagram also serves as a
required including details of diffusion of each species,
thumbnail sketch of main aspects treated in the kinetic theory
interaction parameters and point defect generation
of radiation effects. After Ref. Ill.
characteristics.
A direct approach to solving Eq. (1) is by numerical
methods without further mathematical reduction of the
a thumbnail sketch of the kinetic theory of radiation
equations. This is sometimes termed the hierarchy-of-
effects. It summarizes the sequences and broad rela-
equations method. In practice, this approach is limited
tionships of processes that must be accounted for.
to a relatively small number of equations that can be
solved in reasonable times. Large clusters and long
3.2. Clustering rate equations
irradiation times cannot be treated. However, this ap-
proach provides a detailed analysis of the very early
The kinetic theory of point defect clustering is built
stages of cluster formation. This approach is limited by
upon continuum descriptions. In this approach, the
the necessity for one equation for each type of cluster.
actual physical system is replaced by an effective
In a high-dose irradiation, large cavities and disloca-
medium, where all processes occur continuously in
tion loops would require discrete cluster equations
time and space. The medium is constructed so that
running into hundreds of millions. To reduce this diffi-
each infinitesimal point is assigned both defect genera-
culty, grouping schemes can be devised. Groups cen-
tion and loss (by recombination and by absorption at
tered about a certain size are ascribed the characteris-
sinks) character, so that, in a volume-averaged sense,
tics of one type of cluster.
the effective medium generates and loses the same
By requiring the clusters to grow and shrink only by
number of defects as the actual system operating with
the addition or loss of single point defects, we can
spatially discrete sinks and temporally discrete events.
simplify Eq. (1) to
Thus, for example, discrete sinks such as voids, disloca-
tion loops and point defect clusters are ‘smeared out’ acj(t>
in the continuum. Initially, this approach was em- - = w( j - 1; j)Cj_I(t) + w( j + 1; j)Cj+,(t)
at
ployed, as it has been in chemistry and related disci-
plines, as a common sense and intuitively appealing -[w(j; j+l)+w(j; j-l)]Cj(t), j>2.
method [27]. Later, the approach was substantiated by
(2)
rigorous mathematical construction [28].
In the present section we formulate relationships This specialization of Eq. (1) excludes the possibility of
for point defect concentrations, clustering, and absorp- migration of clusters to sinks. Such loss processes can
tion. Later sections cover more specific models for be reinstated after further simplifications, as in Ref.
 L.K Mansur/Joumal of Nuclear Materials 216 (1994) 97-123 10.5

1261. The equations for j = 1 have special importance Here L? is atomic volume, rc is cavity radius, the D’s
and will be described subsequently. If we imagine j to are diffusion coefficients, the Zc’s are cavity capture
be a continuous variable and use a Taylor series expan- efficiencies, the C’s are physical point defect concen-
sion in Eq. (2) to relate all functions to their values at trations and C”(r,) is the thermal concentration of
size j, we simplify the description to a continuum point defects at a cavity of radius rc. The subscripts i
diffusional approximation in size space, known as a and v denote interstitials and vacancies, respectively.
Fokker-Planck equation, The concentrations are obtained by solving the con-
tinuity equations
acj(r)
-=-$Cj(t)[w(j: j+l)-w(i;i-I)]} Wv
at kTVU, + G, - RC,Ci - K,C, = ~

+~~{Cj(t)[W(j;i-l)+W(j;i+l)l}. (6)
and
(3)
DiCi ac,
The first term in Eq. (3) describes a ‘drift’ in size V DiVCi + - kT VUi + Gi- RC,Ci - KiCi= at
space driven by the excess condensation of one point
defect type over the other, in analogy with the physical (7)
drift of diffusing species in real space due to elastic or
Here G denotes vacancy or interstitial generation by
coulombic forces, for example. The second term de-
displacement and thermal emission from sinks. R is
scribes a ‘diffusion’ in size space, analogous to diffu-
the recombination coefficient given by rlrr,(Di + D,),
sion of matter or energy in real space. The first term
where rr, is the radius of the recombination volume.
ensures that a large cluster will grow inexorably in a
The remaining terms on the left sides describe losses to
radiation field; the second term accounts for the fact
sinks. These are modeled in two ways. The terms
that two different clusters introduced at the same time
involving spatial derivatives describe leakage and drift
at the same size may differ in size at a later time due to
to a sink modeled as discrete. The symbol U is interac-
random encounters with point defects.
tion energy with the discrete sink, and the K, is the
Cavity nucleation theory and cavity growth theory
reaction rate constant that describes the loss rates per
start with Eqs. (2) or (3), as described in Ref. 111.Both
unit point defect concentration of type (Y to sinks
theories are developed in some detail there. Later
modeled as distributed throughout the continuum. Of
discussion in the present paper is based on cavity
these sinks there may be II types such as dislocations,
growth theory and, therefore, its origin is sketched
cavities, dislocation loops, and so on, and K,,i = C,
below. The simplest form of the theory assumes that
Kc,i where K:,i is the vacancy or interstitial loss rate to
there are only point defects and point defect sinks,
the particular sink of type n. Here K:,i is defined as
such as dislocations, cavities, grain boundaries and
the product of a sink strength S~,i, and the appropriate
precipitates, for example. The clusters such as cavities
diffusion coefficient Dv,i. It should also be pointed out
and dislocation loops are characterized by the dimen-
that here and in the remainder of this paper, diffusion
sion j, indicating the number of contained point de-
is considered to be isotropic. The form of the sink
fects or the dimension r, indicating the cluster radius.
strength is discussed in the next section.
In Eq. (3) only the first term in curly braces is
The right sides of Eqs. (6) and (7) express changes
retained in growth theory. Rewriting in terms of the
in defect concentration with time. However, except at
variable r gives
low temperatures, the relaxation times for changes in
the concentrations are generally much shorter than
at ar ’ (4) characteristic times for changes in K, and K, by
microstructural evolution. When the right sides of Eqs.
where now r replaces j and N(r) dr replaces Cj as the (6) and (7) are set to zero (steady-state conditions), and
number density of clusters at sizes between r and when we treat all sinks according to continuum de-
r + dr. The important term in describing the cluster scriptions, Eqs. (6) and (7) become algebraic equations
growth rate is i,, the time derivative of the cavity that are easily solved for C, so that
radius. Cluster growth theory centers on physical
mechanisms affecting i,. The growth rate of a cavity of c = [KiKv+R(Gi-Gv)I
radius rc is determined by the net flux of vacancy ” 2R.K
volume per unit cavity area per unit time

[(
4RG,KiK,
dr,_fi
- ; [ Z,CD,C, - - r,)] . x If (8)
dt
ZtDiCi Z;D$Z( (5) [KiK,+R(Gi-G,)]’
106 L.K. Mansur/Journal of Nuclear Materials216 (1994) 97-123

and I I I I I I I
T=275% Cdm)
c,= [KiKv+R(Gv-Gi)I ,n-6~__,_ C,(6) /
I
2RKi

4RGiKiK,
(9)
[K,K,+R(G,-GJ]~
More will be said about non-steady state conditions
later.

3.3. Irradiation uariable shifts

Before reviewing more detailed models of swelling,

irradiation creep and local point defect concentration
fluctuations, we highlight the fact that a less model-de- ,o-14 ,u12 ,o-tO ,o-6 ,o-6
DOSE(dpa10-4 10.2 1 102
pendent analysis leads to the powerful concept of
irradiation variable shifts [30]. The idea is that when Fig. 11. Calculated dose dependence of vacancy and intersti-
one irradiation variable, like temperature, dose or dose tial concentrations for three different levels of microstructural
rate is changed, a shift in other variables can be sink strength. After Ref. [29].
prescribed to preserve an integral quantity such as the
total number of point defects absorbed by sinks during
an irradiation. Such shifts are calculable. The value of upon the sink strength. Three cases are shown in Fig.
such an analysis is that it permits better design of 11. At high sink strengths, Ci levels off at 7i to its
experiments and better comparison of results from steady-state concentration, while C, builds up until
different experiments. T~=K;',at which time it levels off to a steady-state
When Eqs. (6) and (7) are solved in their time-de- value. For the physically unachievable case of zero sink
pendent forms, the results obtained are shown in Fig. strength, C, and Ci build up equally until ra =
I1 [29]. Here, for simplicity we have not modeled any (RG)-'I', at which time they level off to C, = Ci =
sinks as discrete, so that the first terms involving spa- (G/R)‘j2. In th e intermediate case, where there is
tial derivatives are eliminated., At first, point defect some sink strength, the concentrations first stabilize at
concentrations build up equally for a time TV,of order (G/R)'12after ~a. After more time passes, the first
K;',or for a time ra of order (RG)-'/2, whichever is interstitials begin to arrive at sinks and the point defect
shorter, as determined by the point defect sink strength concentrations begin to diverge. The absorption rates
and damage rate. During this time the concentrations of defects must be equal in the steady state. The
are given by Gt. The behavior at longer times depends condition to insure this after T, = K;l is &D,C, =

Table 1
Concentrations and losses of point defects - limiting cases
Regime CV =i NS NR

Initial transient K,q2A2 Rq3A3

(7 < Tir TR) VA 9A
2G 3G
Steady state
Recombination dominant

(no sinks or ~a < 7 < 7,) (G/R)“2 (G/RF

(7 > 7”) (G/R)“Z(ZiDi/ZvDv)“2 (G/R)“2(Z,Dv/ZiDi)1’2

Steady state
Sinks dominant
G G RGqA
(7 > 7,) 118
E Ki KiK,
 L.K. h4ansur/Joumal of Nuclear Materials 216 (1994) 97-123 107

giDiCi (where thermal emission can be ignored), where

Zi and z, are the weighted average point defect
capture efficiencies. Since, typically, Di s=- D, this re-
quires that C, > Ci. Prior to T, the concentrations 900

exhibit transient behavior; after 7, they are in steady 800

state.
700
Fig. 11 displays limiting cases. In the figure A 1 eV
denotes dpa, defined as KtO, where K is the displace- 600
ment rate per unit volume, and correspondingly, A, = 1.5eV
500
Kr,O. The concentrations of point defects in each II
I I I I I I I I
limiting case can be derived. Examples are given above 400 1 ’
for times after 7, (Eqs. (8) and (9)). 12 5 10 20 50 100 200 500 1000
Ratio of Dose Rates
The number of defects per unit volume that have
recombined up to time T is Fig. 12. The temperature shift above the baseline temperature
of 473 K required at constant dose in order to maintain the
NR = RIrCiC, dt. same point defect absorption at sinks as a function of dose
0 rate, in units of initial dose rate. Results shown for three
different vacancy migration energies. After Ref. [30].
To be specific, consider the case where the material is
in steady state, T > T, and the concentrations are con-
trolled by recombination; Eqs. (8) and (9) give C, =
(G/R)1’2(~iDi/~~D,)1’z and Ci = (G/R)‘/* (z, For example, in the steady-state recombination
Dv/i?;Di) ‘I2 . Substituting these in Eq. (10) gives NR = dominated regime we show in Ref. [29] that for N, to
CT. Similarly, we may obtain the number of defects be invariant at fixed temperature, dose 1 and dose rate
lost to sinks per unit volume up to time 7 as 1 must have the relationship to dose 2 and dose rate 2
as follows:
Nsj = j7KjCj dt , (11)
0
(13)
where the subscript j denotes v or i. Again, consider
the recombination dominated steady state, 7 > 7,. In where A = dose = Kr = Gr/f, where f is the fraction
this case we find by substitution of the C, and Ci given of defects remaining after in-cascade processes reduce
above, Eq. (111, the number of displacements.
Similarly, for a given change in dose rate [30], we
K! find the shift in temperature required at constant dose
NW=
(RG)
(12) to keep N, invariant in the steady-state recombination
dominated regime
The corresponding fractions of defects lost to recombi-
nation or to sinks, fR and f,, respectively can be (kTf/E?) ln(G2/Gl)
obtained by dividing NR or N, above by KT. Table 1 T2 - ” = 1 - (kT,/EF) ln(G,/G,) ’ (14)
shows C,, Ci, N,, NR, f,, and fR for a variety of
limiting conditions. where k is Boltzmann’s constant, T is temperature and
Imposing invariance requirements on N, or Na is a E,” is the vacancy migration energy. Finally, for a given
powerful method that enables us to derive important change in dose at the same dose rate, we find the shift
relationships for variable shifts. For example, consider in temperature required to maintain N, invariant in
the question, “For a given change in dose rate, what the steady-state recombination dominated regime to be
change in dose (at the same temperature) is required to given by
maintain the number of defects absorbed at sinks, N,, ( - 2kT?/E?)ln(A2/Al)
invariant?” Additional questions in the same spirit are
T2- ” = 1 + (2kT,/EF) ln(A,/A,) . (15)
obtained when we interchange temperature and dose
rate or when we interchange dose and temperature in As an example, the temperature shift to keep N,
the above question. The experimental relevance of the invariant for changing dose rate at constant dose, Eq.
relations derived from answering these questions stems (141, is shown in Fig. 12 for a range of vacancy migra-
from the fact that N,, the number of defects absorbed tion energies.
at sinks, i.e., not recombining, can be used in some Another type of temperature shift can be derived,
cases as a rough measure of the radiation-induced requiring not N,, but the net flux of vacancies over
changes in microstructures or properties that have ac- interstitials to a particular type of sink (cavities) to be
cumulated. invariant. This net flux is relevant to swelling rate. The
108 L.K Mansur/Journal of Nuclear Materials 216 (1994) 97-123

temperature shift derived in the same way and for the tion capture radius for the point defect. The subscripts,
recombination dominated steady-state case to keep v or i are used to denote the different radii for vacan-
swelling rate invariant is, cies and interstitials. Generally, Zd > Z,” because rdi >
rdv. When cavities or other sinks are included, Eq. (18)
(kTf/( JY + 2.@))ln(G2/Gl)
PI takes on a more complex form that includes not only
T2 - T1 = 1 - (kT,/( EF + 2E,f)) In(G,/G,) ’ dislocation parameters but parameters of all other
where Et denotes the vacancy formation energy. The sinks in the system. Results for both random and
derivation of temperature shifts for swelling is de- periodic arrays are given elsewhere [31].
scribed more fully in Ref. [30]. The treatment in Ref. The sink strength of a grain boundary is obtained in
[30] for swelling is general in the sense that equations a similar way
are derived for temperature shift where point defect 60
sink strengths are different under conditions 1 and 2 @= _ S1’2d -C 1,
d2 ’
and where point defect loss can be dominated by
recombination or by sinks. Eq. (16) is highly simplified. and
Similar generalizations for Eqs. (13)-(15) are straight- 6S”2
forward. S@’= -, S”2d > 1, (21)
d
3.4. Sink strengths and sink efficiencies where d is the grain diameter. Note that when other
The sink strengths in Eqs. (8) and (9) determine the sinks are important, Eq. (21), the strengths of these
point defect concentrations. In these equations, K = other sinks are an inherent part of the grain boundary
DS, where S is the sink strength, i.e., the sink strength sink strength.
is that quantity which, when taken in a product with
3.5. Swelling
the volume-averaged point defect concentration and
the point defect diffusion coefficient, gives the loss rate When the expressions (8) and (9) for the point
to all sinks of that specific type. The sink strength is defect concentrations are substituted into Eq. (51, we
obtained by solving the spatial diffusion problem for obtain
the type of sink in question, such as a cavity, disloca-
tion, or grain boundary. This solution shows that the dr,_R
- ,(zfz: -Z,dZF)F,
point defect loss rate is given by a simple integral dt
property of the sinks times the bulk point defect con- where F is a function of total dislocation and cavity
centration. The theory of sink strengths has been dis- sink strengths and of their ratio, of dose rate, and of
cussed extensively in Ref. [31]. point defect recombination [l]. The quantity ZdZt -
A derivation of the sink strength of a cavity is given Z,dZF is the bias. It determines the capacity for swelling;
in Ref. [l], where results for other sinks are also it must be positive in order for any swelling to occur. It
quoted. The sink strength of a cavity is obtained as expresses the partitioning of vacancies to cavities and
wr,/D interstitials to dislocations. Values of this quantity nec-
SC= 4ar,( 1 + S112rC) (17)
1 + S’12rC + wr,/D ’ essary to explain observed swelling range from about
0.01 to 1 depending on the material and on values
where S is the total sink strength of all types of sinks.
assigned to other parameters, particularly defect gener-
The first part (4prC) is a simple geometric term, the
ation rate and vacancy migration energy. About an
second part (1 + S112rJ is often designated as a multi-
order of magnitude of this two orders of magnitude
ple sink correction or interactive term, and the third
range is trivial. Some researchers assume that the
part is the sink capture efficiency, designated as Zc.
defect production rate is the same as the damage
The latter measures how good an absorber the sink is.
production rate (i.e., K = G, for example). Thus, a
Alternatively, the second (interactive) and third terms
small bias will produce the correct void growth, since
can be lumped together as a generalized capture effi-
G and bias, ZdZyC- Z,dZF, occur only as a product.
ciency. Here w is the point defect transfer velocity at
Other researchers use a more realistic value of G -C K.
the cavity surface.
In this case the bias must be larger, but the product of
The result for the straight dislocation is
G and bias is unaffected. These considerations are
Sd = ZdL) (18) discussed fully in Ref. [l].
where Eq. (22) can be written in two limiting forms:
2?r
Zd= 1’2 Qf/“Q:/‘( ZfZ,c - Z,“??)
(19)
ln(rL/rd)
(1 + Qi)l’*( 1 + QV)1’2
Here L is the dislocation density, r,=(TL)-l/* is a
measure of dislocation spacing and r, is the disloca- (recombination dominated) (23)
 L.L Mansur/Journal of Nuclear Materials 216 (1994) 97-123 109

DOSE DEPENDENCE
RELATlVE RATES
OF SWELLlNG
OF EVOLUTlON
IS DETERMINED
OF THE VARIOUS
BY THE
MODES OF
(23) and (24) are multiplied by 4ar,fN, so that the left
POINT DEFECT LOSS AND GENERATION side becomes I’; the void volume fraction, V= $rr2Nc.
DOSE DEPENDENCE OF SWELLING CAN BE PREDICTED BY THE
THEORY. FOR EXAMPLE, FOR RECOMBINATION MINOR AND
Large classes of experimental observations can be ex-
DISLOCATION DENSITY SENSlBLY CONSTANT pressed as simple phenomenological relations [l]. At
the early stages every vacancy in a cavity may be
matched by an interstitial in a dislocation loop. In this
case L a V’/*. At higher doses dislocation loops give
way to a dislocation network whose density is main-
tained constant by competing loop formation and seg-
ment annihilation processes. In this case, L = constant.
DOSE
These results are inserted into Eqs. (23) and (24).
The results show that swelling may be expressed as
proportional to (dose)“, where n may take on values
EXPERlMENTALLY OBSERVED BEHAVIOR OF SEVERAL TYPES IS
CONSISTENT WlTH THIS PICTURE. KEY PARAMETER IS 0. from 3/5 to 3 [l]. In general, during an extended
irradiation, several regimes of characteristic exponents
will be traversed. Fig. 13 is a schematic depiction of the
predicted dose dependence of swelling. These theoreti-
cal predictions for dose dependence of swelling are
generally in agreement with results observed experi-
OBSERVED FOR FAST REACTOR OBSERVED FOR MATERIALS.
IRRADIATION OF STAINLESS WlTH HlGH CAVITY DENSITY, mentally, as discussed in section 4.3.
STEEL, LOW HELIUM, TYPICAL HIGH HELIUM, LOW
SINK STRENGTHS DISLOCATION DENSITY
The temperature dependence of swelling implied by
Eq. (5) can be obtained by computation. Ref. [33], for
Fig. 13. Schematic behavior of predicted dose dependence of
example, shows calculated temperature dependences
swelling. When the ratio of dislocation to cavity sink strength
compared with experimental data. Swelling peaks at
is much greater than unity, the swelling rate is low and the
dose exponent is high; when the ratio is reversed, the swelling intermediate temperatures and is negligible at much
rate is low and the dose exponent is low. After Ref. [l]. lower and higher temperatures. The swelling is low at
low temperatures because the vacancy is practically
immobile. Its concentration in the matrix therefore
and builds up, and more vacancies and interstitials are lost
OGQ,Q,( ZfZ; - Z,dZf ) by mutual recombination. At high temperatures an-
dr,
-= other process diminishes swelling. Cavities emit ther-
dt r,ZdZ,dL(l + Q,)(l + Q,)
mal vacancies rapidly at high temperatures, and this
(sink dominated). (24) counterbalances the net vacancy influx driven by irradi-
ation. At intermediate temperatures swelling is maxi-
In these equations Qi,” denotes the ratio of dislocation
mized - both thermal emission and mutual recombina-
to cavity sink strength, Eq. (23) applies where most
tion are less important, and the excess flow of vacan-
point defects are lost by bulk recombination, and Eq.
cies to cavities is maximized. ‘High’ and ‘low’ tempera-
(24) applies where most are absorbed at sinks ‘. These
tures are dependent on the properties of the material
are the two extremes in modes of point defect loss. As
and in particular on how the overall sink strength
can be seen, in the former case the cavity growth rate
changes with temperature. In some materials, over a
depends parabolically on the dose rate; in the latter
certain temperature range, lowering the temperature
case the cavity growth rate depends linearly on the
increases the sink strength so that recombination is
dose rate. Here, Q is defined as
reduced at the lower temperature.
Z&L The location of the peak swelling temperature de-
Qi,v = pends on dose rate, sink strength, and the predominant
4ar,N,ZF,, ’
mode of defect loss. When dose rate is increased, more
where L is the dislocation density and N, is the cavity point defects are created, but their diffusion velocities
density. remain fixed. To remove defects at the higher rate in
To obtain a swelling-versus-dose prediction, the dose the steady state requires the point defect concentra-
dependence of dislocation density must be known. Eqs. tions to be higher. This results in more recombination.
However, by increasing temperature, the same relative
ratio of recombination rate to absorption rate at sinks
r Of course, most point defects absorbed at sinks recom- may be restored. By requiring this ratio to be invariant,
bine there because of their absorption in nearly equal num- we can obtain a temperature shift for swelling as a
bers. Only a small excess accumulates at each sink dictated by function of dose rate, following along the lines of
the bias. Section 3.3 above.
 110 L.K. Mansur/Journal of Nuclear Materials 216 (1994) 97-123

onset of swelling. The insoluble rare gas helium is one

of primary concern. It is produced at levels of tens of
parts per million by (n, IX) transmutation reactions in
fast reactors. In future fusion reactors helium is ex-
pected to be produced to levels of thousands of parts
per million. Gases may affect swelling through several
reaction pathways. These include cavity pressurization,
induced changes in microstructural sink strength, in-
duced changes in microcomposition, and alterations in
atomic transport.
One of the most important effects is cavity pressur-
ization. Gas pressure suppresses thermal vacancy emis-
sion from cavities, as can be seen from the expression
for C:(r,), the thermal vacancy concentration near a
I I I I
cavity of radius rc,
CAVITY RADIUS

Fig. 14. Cavity growth rate versus cavity radius for the critical
number of gas atoms, n:, and both lower and higher numbers (27)
of contained gas atoms.
Here y is the surface energy and P is the gas pressure
A general equation for the temperature shift of the in the cavity. When Eq. (27) is substituted into Eq. (51,
swelling rate has been derived [30], the resulting cavity growth rate versus radius is as
T2 - Tl shown in Fig. 14. For a given number of contained gas
= [ kTf/(Et’ + nEJ)] [W%/Gd + In ~1 atoms, there are two roots, one at a stable radius, r,S,
1 - [ kT,/(EF + no,‘)] [ln(G,/G,) + ln M] ’ and one at a critical radius r,“. The cavity must achieve
a size beyond r,” to grow by bias-driven growth. The
(26) point on the figure denoted rf is the critical radius
for which Eq. (16) is a special case. The number n is 1 with no gas. Gas makes achieving critical radius easier.
or 2 depending on whether loss is dominated by sinks With more contained gas r,” is reduced. At the same
or by recombination, respectively. The quantity M con- time more gas brings rf closer to r,‘. As more gas is
tains microstructural information about the material in added, r,’ is decreased and r,” is increased up to a
conditions 1 and 2. The appropriate expressions for M critical number of gas atoms, nz, where r,” and r,” meet
are contained in Ref. [30]. at r,*. These quantities are also shown in Fig. 14. If any
more gas is added, a critical radius no longer exists,
3.6. Critical cavity radius
and bias-driven cavity growth is inevitable. There are
The effects of gaseous impurities on swelling are of thus two qualitatively different paths to achieve bias-
particular interest in the theory because they affect the driven growth. The cavity may depart from its stable

2.19 XVT7m 2.15 X1V7m

0 2.4 X W7 0 2.4X 14' 0 2.4 X 10-'

At (8)
I .A I A I I
V

0 10 20 30 40 50 60 70 60 90 loo
5.2 X 1O-3 2.9 x lo-2
TIME (sl tw

Fig. 15. Point defect concentrations calculated using the cascade diffusion theory at an arbitrary reference point for a typical
neutron irradiation of nickel at 500°C and 10V6 dpa/s: left, vacancy concentration; right, interstitial concentration. After Ref. [25].
 L.K. Mansur/Joumal of Nuclear Materials 216 (1994) 97-123 111

size by stochastic fluctuations, on occasion exceeding

r,“. Alternatively, the cavity may accumulate more than
the critical number of gas atoms, whereby bias-driven
growth is insured; no fluctuations are required. These
two possibilities merge as gas is added. The critical
quantities have been found to be extremely useful in
understanding a number of aspects of swelling. More
will be said about comparisons with experiments in
Section 4.

3.7. Cascade diffusion theory

Thus far our theoretical description has been in

terms of continuum reaction rate theory, wherein both
point defect generation and loss are approximated as
continuous in time and uniform throughout space. In
reality, however, point defects are generated in cas-
cades. An approach that introduces point defects in
discrete cascades, yet maintains the advantages of con-
tinuum reaction rate theory is termed cascade diffusion
Fig. 16. Integral (a) and differential (b) importance functions
theory [25]. Cascade diffusion theory describes sporadi- for point defect concentrations. Calculated using the cascade
cally fluctuating concentrations at any point in the diffusion theory. After Ref. [25].
material because of the superposition of contributions
from random cascades occurring throughout. These
fluctuations can be crucial in understanding certain cascades at large distances, while spikes are con-
processes in an irradiated material, such as cascade-in- tributed by relatively few nearby cascades and by ar-
duced creep, to be discussed in a later section. The rival coincidences of defects from several different
continuum rate theory described above can be derived relatively nearby cascades. For the conditions consid-
through approximations and limits of the cascade diffu- ered [25], it is found that - lo3 cascades in various
sion theory. stages of dispersion make non-negligible contributions
A cascade is introduced as a delta function. The to the concentration. Because of more rapid diffusion,
cascade diffuses in the lossy medium, where the solu- the interstitial profile is different. At any time there is
tion for point defect concentration is described with at most one cascade making a non-negligible contribu-
appropriate boundary conditions by the equation tion to the concentration at a point. The interstitial
ac(P, t)
concentration profile, Fig. 15, consists of extremely
=D(v* - s)C(p, t) + G(p, t). (28) large spikes whose decay time is less than a microsec-
at
ond separated by a few milliseconds. For most of the
The solution is time there is negligible interstitial concentration.
Interesting information on where in space the point
exp(-SD), (29) defects come from can be obtained from the theory.
Fig. 16 shows differential and integral importance
where p is the distance from the cascade center, t is functions versus distance from the point of reference.
the time since its introduction and S is the same sink These results show how important each shell of volume
strength described earlier. This solution is applicable corresponding to a radial distance increment is in
far from a discrete sink. However, the presence of a contributing to the concentration. For example, the
sink modeled as discrete, such as a cavity or disloca- most important infinitesimal radial increment is at one
tion, changes the solution. The solutions in the pres- absorption length, defined as the inverse square root of
ence of these two types of discrete sinks have also been the sink strength (S-‘/‘I. We note also that the region
derived [25]. within three absorption lengths contributes more than
Fig. 15 shows the point defect concentrations at an 80% of the total average defect concentration observed
arbitrary point calculated for a typical neutron irradia- at the reference point.
tion of nickel at 500°C and 10e6 dpa/s. On the left is An application of the statistical side of the above
the vacancy and on the right is the interstitial concen- analysis is the investigation of cascade coincidence
tration. The vacancy concentration has a more or less probabilities. For example, within a small volume, cas-
steady background on which are superimposed large cades occur at an average rate defining an average
spikes. The background is contributed by numerous time interval for n cascades to occur (average n-fold
 112 L.K Mansur/Journal of Nuclear Materials216 (1994) 97-123

interval). However, for any time interval, even one Herring-Nabarro creep is a mechanism that oper-
much shorter than the average one, there is a finite ates by the transport of thermal vacancies between
probability that n cascades will occur. When the vol- oriented grain boundaries in the absence of irradiation
ume is of radius of the order of the mean point defect [34]. A similar process operates between oriented dislo-
absorption length, S-i/‘, and the time interval is of cations. We may term this process stress-induced pre-
the order of the mean point defect diffusion time, the ferred emission (SIPE). In addition, however, stress
probability of multiple cascade occurrence may be of also changes the capture efficiencies of oriented dislo-
physical interest. Ref. 1251calculates and illustrates the cations for the point defects that are generated during
magnitude of n-fold intervals. Although the fraction of irradiation. The largest effect is on the absorption of
n-fold intervals for times of very short duration is self-interstitials. The greater capture efficiency of dis-
small, it nevertheless may be significant when it is locations whose Burgers vectors are more nearly aligned
realized that typically 102i-10” cascades/s m3 occur with the stress axis can then lead to creep. The varia-
during irradiation. Thus, highly unlikely events may be tion in capture efficiency of a dislocation with orienta-
responsible for part of the microstructural develop- tion can arise because of the inhomogeneity interac-
ment observed in experiments. tion, caused by a change in elastic polarizability of the
As pointed out in Ref. 1251,very often (high proba- point defects [35-371. Recently, it has been suggested
bility) within intervals on the order of the vacancy that the capture efficiency can be changed even more
lifetime, and occasionally (low probability) within inter- strongly by a stress-induced anisotropy in point defect
vals on the order of the interstitial lifetime at elevated diffusion, termed elastodiffusion [38]. It arises because
temperatures, multiple cascades may occur in a volume the material is actually built on a discrete atomic
whose linear dimension is on the order of the point lattice (as opposed to a continuum), which results in a
defect absorption length. The coincidence of multiple change in symmetry of the diffusion field in the pres-
cascades within these physically meaningful volumes ence of an applied stress. Originally, the specific effect
and time intervals ensures intense interaction of de- of the inhomogeneity interaction on creep was termed
fects from different cascades, possibly resulting in dif- stress-induced preferred absorption (SIPA). However,
ferent microstructures than in cases of no coincidence. we now include all the stress-induced interactions that
Although such events are relatively rare, enough of result in a change of dislocation capture efficiencies
them can occur over large volumes and long times to under this acronym.
be important in the macroscopic manifestations of ra- The climb-enabled glide mechanisms of irradiation
diation effects. creep can operate whenever climb occurs. This in-
cludes the processes described above, but also includes
3.8. Irradiation creep by dislocation climb and glide swelling; as cavities grow, dislocations absorb excess
interstitials and climb. Climb-enabled glide caused by
Much of the theoretical work on irradiation creep swelling, and by SIPA, termed preferred absorption
has been in terms of steady-state point defect concen- glide (PAG), have been described in a unified formula-
trations. More recently we have understood that large tion in Ref. [39]. These climb and the climb-enabled
contributions to creep can occur as a result of several glide processes have been most highly developed on a
types of transients in point defect populations. In the theoretical level and in application to experimental
present section the basic elements of irradiation creep data. A number of other mechanisms also have been
mechanisms are described, with emphasis on behavior proposed. For a review of the wide variety of mecha-
in the steady state. In the following section the more nisms that have received attention, see Ref. [14].
recent theoretical work on transient mechanisms is The relationships of irradiation creep to swelling
outlined. were exposed in Ref. [40]. For the specific mechanism
Creep can take place by two distinct processes, of swelling-driven creep, the relationship is direct and
atomic absorption (climb of dislocations) and relative simply expressed. However, even for processes where
motion of crystal planes (glide of dislocations). In the there is no direct cause and effect relationship between
former, atoms are transferred from planes more nearly swelling and creep, there is a lock-step relation, ex-
parallel to those more nearly perpendicular to the pressible in simple mathematical form, for each creep
stress direction. In the latter, planes inclined to the process. Thus, even where one does not cause the
stress direction slip, atom row by atom row, with re- other, definite relationships can be extracted because
spect to one another through the glide of dislocations. both swelling and creep are driven by the same point
The latter process generally also is coupled with atomic defect concentrations.
transport (climb-enabled glide), since gliding disloca- Referring to Fig. 11 we see that the time interval
tions become pinned at obstacles and planar slip stops before 7, can be described as a transient regime, and
unless they are released by atomic absorption or emis- the time after T, can be referred to as the (quasi-l
sion. steady-state regime (A, = Kflr, is shown on the figure.)
 LX Mansur/Joumal of Nuclear Materials 216 (1994) 97-123 113

The interval rV can range from milliseconds to unlimit- tions of density Lj, whose Burgers vectors are aligned
edly large times, depending on the material, the sink in the j direction,
strength and especially the temperature. After T, the Ii = a( Z”jD,C, - Z,d’D,C, + Z,“‘D,C, ) L’.
point defect concentrations change only gradually in
response to the relatively slow evolution of microstruc- Here 0 is the atomic volume, Zdpi is the ca:t
ture. Hence, the term quasi-steady-state rather than efficikncy of dislocations in orientation j, the D are
steady-state. As mentioned above, much of the theoret- diffusion coefficients and the C are concentrations,
ical work on irradiation creep has been for this quasi- with the identity of the defect denoted by the corre-
steady-state. This averaged picture is an appropriate sponding subscript. The symbol CT is the thermal
foundation for the point defect concentrations neces- equilibrium concentration of vacancies at a dislocation
sary for the calculation of fluctuation-insensitive defor- of orientation j. For a tensile stress D aligned with the
mation processes. These include processes that depend l-axis,
only on cumulative point defect currents: swelling; Ctl = Ct exp(cTR/kT), (34)
creep by SIPA, and creep by climb-enabled glide driven
cd2 = cd3 = ce
by swelling and/or SIPA. ” ” “7 (35)

By considering the local point defect concentration where Ct is the thermal equilibrium vacancy concen-
fluctuations as revealed by cascade diffusion theory, tration in the bulk. Conservation of atoms requires that
Fig. 15, we discovered a new mechanism of irradiation this current be related to the climb velocity of disloca-
creep that we termed cascade-induced creep 1411.Simi- tions by
larly considering transients in the average picture prior
vi = I’/bL’. (36)
to time T,, Fig. 11, two additional transient irradiation
creep mechanisms are apparent, startup-induced tran- From [39], the magnitude of the average climb velocity
sient interstitial creep [42], and glide-induced transient is
vacancy creep [43]. These will be outlined in Section
3.9. First, however, we sketch the formulation for creep u= ; (( Z;‘D,C, - Z,d’D,C, + Z,“‘D,C,“’ 1
by dislocation climb and climb-enabled glide.
Creep rate by climb-enabled glide may be expressed + 21Zf2DiCi - Z,d2DvC, + Z,“‘D$,“’ I), (37)
as assuming Lj = L/3 for j = 1, 2, 3, where L is the total
dislocation density. Eq. (37) contains contributions from
icr = uf3 (30)
swelling, SIPA and SIPE.
where v is the average dislocation climb velocity and f If there were no SIPA, i.e., Zdl = Zd2 = Zd3 = Zd,
is a function that translates climb to glide. In the then Eq. (37) reduces to
dislocation bowing model [44],
u = ; [ ZfD,C, - Z:Dv( C, - cf)] , (38)
f = (PzyE, (31)
where c: is the sink-weighted average of the C given
where E = u/E is the elastic deflection caused by a in Eqs. (34) and (35). When inserted in Eq. (30), Eq.
stress u in a material with Young’s modulus E. The (38) produces the swelling-driven creep discussed in
process modeled is the climb of a dislocation over an Ref. [40], with SIPE superimposed when ci is non-
obstacle, the subsequent glide, and pinning at another negligible.
obstacle. The released segments again bow out under For no swelling, but SIPA and SIPE processes oper-
the applied stress after repinning, leading to strain. ating, a second specialized form of v in Eq. (37) may
When climb again releases the segments, the strain E be derived. In the steady state,
represented by a bowed dislocation is converted to an
increment of permanent deformation. ( ZflDiCi - Z,d’D,C, + Z,d’D,C,d’)
Another model for f is an obstacle model, where = - 2( Zd2DiCi - Z,d2D,C, + Z,d2D,C,d2). (39)
precipitates, dislocation loops or dislocation segments,
for example, may act as barriers to glide [45]. In this This reduces Eq. (37) to
case
u= ; ; 1Z:‘DiCi - Z,d’Dv( C, - Cv”‘) I.
f =adbL/h, (32)
Similarly, in the absence of irradiation (no swelling and
where a is a numerical constant of order unity, d is the no SIPA), SIPE may occur, and a third specialized
obstacle spacing and h is the obstacle height. The form of Eq. (37) is obtained
symbol b denotes the magnitude of the Burgers vector.
The velocity v is obtained directly from the current
(41)
of excess interstitial volume per unit volume to disloca-
114 LX Mansur/Journal of Nuclear Materials216 (1994) 97-123

a form of Herring-Nabarro creep by exchange of va-

cancies among dislocations. A similar analysis for creep
by climb only, considering SIPA and SIPE processes,
gives [39]

E, = $L{ [ AZ”D,C, - AZ$‘&]

+D&t[Z$ exp(ufi/kT) -Z$‘]]. (42)

Here the AZd are defined as (Zd’ - Zd2), and it is
assumed that Zd2 = Zd3, for both types of defect.
More specialized work has been devoted to deter-
mining the various Zd” in relation to the stress axis. 01 I I I I I
0 20 40 60 00 100
The inhomogeneity interaction [35-371, stress-induced
diffusional anisotropy [38], and the presence of other r(nm)

nearby dislocations have been examined. The results Fig. 17. Vacancy concentration profiles after a dislocation has
are input into the more global description outlined arrived at an obstacle. The labels (a)-(d) correspond to nor-
above in the form of the Zd” and the AZ. The analysis malized times between 0 and 3000 nm-* respectively, curve
has usually been restricted to the steady state. In terms (e) corresponds to the profile at infinite time. After Ref. [43].
of Fig. 11, the concentrations far to the right are
assumed to apply, i.e., the condition
sion profiles at the new location 2. The time to achieve
Z”D,(C, - cv’) = z,D,C, (43) steady state, T, = K;’ = (D,S,)-’ is essentially the
time it takes a vacancy created at a mean position in
is imposed. Here the .?! and c,” are, respectively, the
the material to reach a sink. Using reasonable sink
sink-weighted average capture efficiency and thermal
strength parameters, T, can range from 3 x lo-’ to
vacancy concentration at sinks.
2 X 10” s for temperatures from 873 to 273 K, respec-
tively. Yet, when a stress, (T, is applied to an unpinned
3.9. Irradiation creep caused by transients dislocation, the glide velocity is given by

Transient changes in average point defect concen- u ,=G[l-exp(-Bt/m)], (44)

trations may be important under certain conditions.
The theory in Section 3.8 before Eq. (39) is general in where B is a drag coefficient, t is time, m is the
that it applies to all regions of Fig. 11. In addition, the effective mass per unit length and b is the Burgers
fluctuations illustrated in Fig. 15, which are ignored by vector [43]. The dislocation reaches its terminal veloc-
definition in the uniform concentration approximation ity ub/B after a time of the order of m/B. For
approach of Section 3.8, may also have significant reasonable materials parameters the terminal velocity
consequences to irradiation creep. Irradiation creep by of 60 m/s at a stress of 20 MPa is reached after the
transient processes is covered in the present section. dislocation travels only 6b. If the barriers are taken as
a distance Lm112 apart, the glide time is about 1 x
3.9.1. Glide-induced creep by transient vacancy absorp- lo-” s for reasonable dislocation densities. This is
tion negligible compared to T,.
In modeling climb-enabled glide creep, it has gener- The excess number of defects absorbed during the
ally been assumed that steady-state diffusion profiles transient up to r, after a glide event is determined by
are maintained in the vicinity of a dislocation before, the difference between the averaged uniform steady-
during and after a glide event. On closer examination, state concentration of point defects and the later
however, it is found that the glide between obstacles is steady-state diffusion profile about the dislocation.
so rapid that the steady-state profile that may be
established near a dislocation during climb is not main-
tained during glide [43]. The dislocation arrives at its 2
new position on another obstacle so rapidly that the There is also a transient interstitial flux. However, this
event is like creating a dislocation suddenly in the transient is of much shorter duration than that of the vacancy,
and the magnitude of interstitial current absorbed is orders of
uniform point defect concentration field that existed
magnitude less than for the corresponding vacancy transient.
before the dislocation was introduced. This is dictated by the steady-state point defect levels prior to
During the transient that follows, the dislocation the arrival of the dislocation, which are approximately in the
absorbs an unbalanced excess vacancy flux in the pro- ratio Di /D, for the vacancy and interstitial transient contri-
cess of establishing new steady-state point defect diffu- butions.
 L.K Mansur/.Joumal of Nuclear Materials 216 (1994) 97-123

Such extra vacancies cause a climb in the vacancy

direction. If this climb is enough to cause the disloca-
tion to climb over the obstacle and glide to another
obstacle, an increment of creep is produced that would
not have occurred if the steady-state profiles were
maintained during the glide event. This process has
been analyzed through a model where a cylinder of
material, corresponding to the dislocation capture ge-
ometry, is suddenly removed from a medium contain-
ing a uniform point defect concentration. The time-de-
pendent concentration is obtained from the diffusion
equation describing continuity in the volume surround-
ing the cylinder, with the empty cylinder surface as a -.-.- o.sqr
point defect sink. Fig. 17 shows results at several times
I
-- o.osdpa
I
.25%owPcA
mwwa,e~
I
I
I
after the dislocation arrives at a new obstacle. lPO 200 400 600
TEMPEmnJRE cc)
The position of a dislocation during the transient of
Fig. 17 is shown in Fig. 18. The time scale is expressed Fig. 19. Theoretically predicted creep strain (solid curve)
in terms of log(D). The climb in the positive direction accounting for startup indirect transient interstitial creep.
is due to the extra vacancies absorbed during the Points are experimental data and dashed curves are calcu-
lated results. After Ref. [48].
transient up to establishing the steady-state diffusion
profile. The negative climb at much longer times re-
tant realization is that during the time interval r,, all
sults from the persistent absorption of a small net
dislocations receive large excess fluxes of interstitials
excess of interstitials in the steady state as a result of
[46-481. If T, exceeds the duration of the experiment,
an assumed SIPA bias. For given obstacles and disloca-
then dislocations always receive excess interstitial
tion positions on the obstacle, these results are used to
fluxes. This persistent unbalanced interstitial flux is the
obtain dislocation escape frequencies. Irradiation creep
basis of a recently recognized mechanism of irradiation
rates can then be calculated [43].
creep that we termed startup-induced transient inter-
3.9.2. Startup-induced creep by transient interstitial ab- stitial creep [48].
sorption
From Fig. 11, the origin of this mechanism can be
Above, if the temperatures were too low, the va- understood. Only after rv do the concentrations obey
cancy profile near the dislocation would never adjust to the relation .?!”D,(C, - Cz) = ziDiCi. After this time
the presence of the dislocation during the entire dura- there is at most only the small excess trickle of one
tion of the experiment. However, T, after startup mea- type of point defect to dislocations or asymmetric par-
sures the time taken to establish the overall averaged titioning of the two types of point defects among dislo-
uniform steady-state vacancy concentration in the ma- cations, caused by swelling and SIPA processes, re-
terial after the initiation of the irradiation. An impor- spectively. Prior to 7, the above relation does not hold
and the interstitial concentration Ci, is much higher
than it is at steady state.

=7----l The transient interstitial creep that occurs in the

time interval T, in a climb-enabled glide model is
calculated by using the general time-dependent creep
relations, Eqs. (30) and (37), in conjunction with the
time-dependent point defect concentrations obtained
by solving Eqs. (6) and (7). Fig. 19 shows the calculated
creep rate as a function of temperature, including
transient interstitial creep [48]. Typical irradiation and
materials parameters applicable to stainless steels were
used in the calculation. It can be seen that the tran-
sient contribution dominates creep at temperatures
below about 200°C. At high temperatures, the transient
contribution is negligible and creep is dictated entirely
100 102 10' IO' 100 10'0 by the steady-state processes.
4nm2)
Fig. 18. Climb position as a function of time (both normal- 3.9.3. Cascade-induced creep
ized) after a dislocation has arrived at an obstacle. After Ref. In the transient creep mechanisms discussed above,
[431. the new creep processes arise from additional contri-
116 L.K Mansur/Journal of Nuclear Materials216 (1994) 97-123

butions to climb-enabled glide that result from point stability has been to develop and compare theory with
defect absorption. The additional point defect absorp- experimental results. In general, two types of compar-
tion is calculated under a continuum reaction rate isons can be distinguished. The first is based on the
theory where point defect production is modeled as fact that a great deal of experimental data has been
taking place at its average rate at every point in space obtained to characterize swelling and to a lesser extent
and continuously in time. In an average sense this is irradiation creep, in terms of dependences on parame-
correct. However, as the discussion of Section 3.7 ex- ters such as dose, temperature and material, for exam-
plains, point defect production in cascades leads to ple. Such data is often collected on a purely empirical
local fluctuations in point defect concentrations and level, without any deliberate comparisons with theory
fluxes. The existence of these fluctuations can lead to in mind. After the fact, such data can sometimes be
an additional mechanism of irradiation creep, termed collated and compared with theoretical predictions.
cascade-induced creep [41]. Such comparisons are of some use for ascertaining the
In response to the cascade-induced fluctuations in correctness of predicted broad trends and overall be-
point defect fluxes, a dislocation segment makes climb havior. However, because these experiments were gen-
excursions. During an excursion there is a probability erally not designed with forethought as critical tests of
that the segment will reach an unpinning position on the theory, the comparisons are of a low level of
the obstacle and be freed to glide. The calculation of applicability that is not commensurate with the enor-
this unpinning frequency and the subsequent creep mous amounts of effort represented by the collection
deformation has been carried out [41]. It has been of experimental data. This is natural, since much of the
found that the unpinning frequency can be substantial experimental work was directed to the practical devel-
when compared to that caused by swelling and by opment of nuclear power and not at physical research.
stress-induced preferred absorption. For obstacles of Another type of comparison involves experiments
height 40b (where b is a lattice dimension), for exam- that have been designed with the objective of providing
ple, it was found that cascade-induced creep gave the critical tests of mechanistic understanding. Below, we
largest contribution for the parameter values em- review comparisons of each type. The intent is to give
ployed. However, for increasing obstacle size the rela- the flavor of how theory and experiment have played
tive contribution of cascade induced creep diminishes, mutually supporting roles. The examples shown are
and for obstacles twice this height, swelling driven only a few of the many that are available and no claim
creep becomes the dominant mechanism. is made that this section is exhaustive or representative
in its coverage.

4. Comparisons of theory with experiments 4.1. Temperature shift of swelling with dose rate

One of the continuing long-term themes of the In Sections 3.3 and 3.5, expressions were derived for
more fundamentally oriented research on dimensional temperature shift of total point defect absorption and
net point defect absorption with dose rate. In Ref. [49],
Packan, Farrell and Stiegler compared experimentally
Ki’ observed temperature shifts with the theoretically pre-
I I I I I I I
1 dicted values. Both their own data and data gathered
Ni from the literature were included in the comparison.
Fig. 20, from Ref. [49], shows the observed swelling
and its dose rate-temperature shift in nickel. The line
represents both the Packan et al. experimental results
and the theoretically calculated temperature shift for
those conditions. It can be seen that the theoretical
predictions track the experimental data reasonably well.

4.2. Critical radius and critical number of gas atoms

The existence of a critical radius or critical number

I I I I I I I of gas atoms for swelling was discovered on a purely
468 theoretical basis. It has led to several important experi-
300 400 500 600 700
PEAK SWELLING TEMPERATURE 1%) mental demonstrations. Swelling is now much better
Fig. 20. Temperature shift of peak swelling as a function of understood, and this understanding has been used to
dose rate. Points are experimental data and line is theoreti- uncover principles to avoid or control swelling. Perhaps
cally predicted result. After Refs. [30,49]. the most obvious consequence of the existence of a
 LX Mansur/Joumal of Nuclear Materials 216 (1994) 97-123 117

critical radius is the theoretical prediction of bimodal

cavity size distributions in the presence of helium. In
particular, cavities containing less than n: gas atoms
should remain stationary below a critical radius, but
those containing more than n: should grow. This
knowledge has been used in a number of experiments
to obtain a measure of r,*, the minimum critical radius
or maximum stable radius, under given irradiation con-
ditions. In turn the direct experimental measurement
of r,* permits the extraction of other parameters, such
as bias, important in swelling calculations. In addition,
comparisons of experimentally determined critical
quantities can reveal large differences in the swelling
potential of materials. Fig. 21 shows a measured bi-
modal cavity size distribution in ion-irradiated Fe-1OCr
alloy [50]. Specimens were implanted with 300 appm
helium and then irradiated to 30 dpa. A clear bimodal
0 20 40 60
is observed, with a separation at a cavity diameter of
CAVITY DIAMETER lnm)
about 5 nm <r,* = 2.5 mn). Similar bimodals in the
Fig. 22. Cavity size distributions showing (upper) unimodal
presence of gas have been observed repeatedly in ex- cavity population in 15Ni alloy and (lower) bimodal cavity size
perimental work [51]. distribution in 35Ni alloy after 40 dpa irradiation with 4 MeV
Knowledge of the existence of a critical radius has Ni ions following a helium injection and anneal. After Ref.
also been valuable in tailoring swelling resistant mate- [511.
rials. In particular, because of the above analysis we
have realized that swelling can be avoided even up to
high doses and high contained gas contents, by (1) ity of the critical radius with composition [52]. It has
creating a material with a large critical radius/critical helped explain the large differences in swelling be-
number of gas atoms under experimental conditions of tween Fe-Ni-Cr alloys of different Ni contents. Higher
interest or, (2) dispersing gas so widely that the gas nickel alloys have been found to swell less than lower
accumulation rate in any cavity, leading up to the nickel alloys, largely by increasing the dose to the onset
achievement of critical number of gas atoms in that of swelling [53,54]. We hypothesized that the increase
cavity, is extremely slow. in dose could be due to an increase in critical radius/
Experimental demonstrations of these two ap- critical number of gas atoms required for swelling [52]
proaches have been carried out. The first, correspond- in the high nickel alloys. An experiment was designed
ing to approach (1) above, shows the extreme variabil- to test the hypothesis, viz., a sequenced helium injec-
tion, anneal, and irradiation. The injection of 400 appm
He and anneal at 675°C produced similar bubble size
40 , distributions in 15 and 35 Ni alloys. This step created
I I I I I I I
bubble size distributions that spanned the critical radii
35
c,=3.0xlo*2m-3
in both alloys. On irradiation with the Ni ions, these
size distributions would be expected to separate into
3
1 30 tn d,=2.3 nm
two parts, with the upper cutoff radius of the smaller
size group giving a measure of r,* for each alloy.
Fig. 22 shows measured size distributions after irra-
diation with Ni ions to 40 dpa. In the low nickel alloy,
k no small cavities were observed. A size distribution of
15 -
8 Cc= 6.7 x 1020m-3 large cavities at a density of 5.3 x 10” mm3 and cen-
tered at 35-40 nm was observed. In the high nickel
alloy, the initial distribution separated, with the large
cavities centered at about 35 nm and the small cavities
showing an upper cutoff of about 10 nm diameter.
-0 5 10 15 20 25 30 35 These results were interpreted to show that the critical
CAVITY DIAMETER (nm) radii in the two alloys were different by at least a factor
Fig. 21. Cavity size distribution showing bimodal population in of 5, in agreement with the above hypothesis. The
Fe-1OCr implanted with 300 appm helium annealed and absence of a small-size group of cavities in the low
irradiated to 30 dpa with 4 MeV Fe ions. After Ref. [50]. nickel alloy was interpreted to mean that the small-size
118 L.K Mansur /Journal of Nuclear Materials 216 (1994) 97-123

Fe-13Cr-15Ni Ternary (P,Si,Ti,C)-Modified

0.4 dpa/0.2 appm He/675C 109 dpaI2000 appm He/675C

Fig. 23. Cold worked Fe-Ni-Cr alloy modified with Si, Ti and P, irradiated simultaneously with 4 MeV Ni ions and 200-400 keV
He ions to (a) 0.4 dpa at 0.2 appm and (b) 109 dpa at 2000 appm He. Insets show phosphide precipitates with many cavities at the
precipitate-matrix interfaces. After Ref. [%I.

group existed below the TEM resolution limit for cavi- of cavities collecting helium, the gas buildup rate in
ties, about 1 nm diameter. each cavity was reduced and the onset of swelling was
Another study, corresponding to approach (2) above, delayed. Calculations accompanying the analysis of Fig.
shows how the achievement of the critical radius/ 23 showed that several thousand appm He can be
critical number of gas atoms can be delayed by tailor-
ing the microstructures [55,56]. The approach was to
slow the accumulation rate of gas atoms in each indi-
vidual cavity in order to delay the onset of swelling.
The alloys were irradiated with dual beams of Ni and
He ions. One alloy was the pure ternary Fe-lSNi-13Cr. 0
z .7
Another had a similar base composition, but was al- IL
p .6
loyed with 0.8 Si, 0.2 Ti, 0.04 C and 0.05 P. z
Swelling was observed early in the pure ternary, at I?
= .5

0.4 dpa and 0.2 appm helium. In the other alloy,

swelling was not observed at doses of 109 dpa and 2000
appm helium. Swelling suppression was attributed to
the formation of an extremely high density of needle-
shaped M,P phosphide precipitates resulting in three
v
orders of magnitude more cavities in the complex alloy ,001 .Ol .l 1 10 100 1000
compared to the simple alloy. TEM, Fig. 23, showed Q. Ratio of Dislocationto CavitySink Strengths
the formation of this high number of cavities at precip- Fig. 24. Theoretical prediction for swelling rate versus Q,
itate-matrix interfaces. Thus, by increasing the number normalized to its maximum at Q = 1.
 L.K. Mansur/Journal of Nuclear Materials216 (1994) 97-123 119

0.0
accommodated when the cavity density is > lo** rnm3. I 1
AUSTENITIC ALLOYS
I I I I

Such swelling suppression results have been observed l .kMmmlelal. oSpmQueelal.(a) 0LeasndMmsur
r&Q!abynal. 0 wmmw9landal~. c COllacmce _
in a number of experiments where extreme gas disper- 0.5 -
~P&lMMdF~l ASmMelal.
v Mazlru 0 hnlka et al.
sion has been achieved.
p 0.4 - HIGH SWELLING
LOWL - _ HIGHL
4.3. Cavity growth rate

As can be seen from Eqs. (.5)-(9), virtually all mate-

rials and irradiation parameters affect cavity growth
rate through their influences on the concentrations C,
and Ci. The relative strengths of cavities and disloca-
tions, in particular, have important and characteristic
effects. In general, Eq. (5) or (22) is integrated to
lo-3 lo.2 lo” 1oO 10’ lo2 103 10’
obtain the dose dependence of swelling and its depen- SINK STRENGTH RATIO (0)
dence on the ratio of dislocation to cavity sink strengths. 0.12 I I I I I I
This ratio is denoted by Q and is given by FERRITIC/MARTENSITIC ALLOYS

0.10 . Gellee
(a) 0 Halon at-d Mansur

. Gdba (b) A YHekmdKlwh

n Smmte(al. v Fandl
. Sullwph et al. 0 Mariasz (d)
II Little and SlDw

The swelling rate, v= 4rrrzN,+,, is also strongly de- LOWL - _ HIGHL

pendent on this ratio. For example, when sinks are the HIGH Nc LOW N,

dominant mode of point defect loss, it is easy to show

TYPE S
[l] that
Q
va (1 +Q)”
Fig. 24 shows the calculated behavior of swelling rate
versus Q normalized to its maximum value. Under the
lo.3 lo.2 10’ roO 10' 102 103 104
SINK STRENGTH RATIO (0)
above simplified assumptions, the curve is symmetric Fig. 25. Experimentally observed swelling rates as a function
about Q = 1, in the sense that Q =x and Q = l/x of Q for both austenitic (upper) and ferritic/martensitic
result in the same swelling rate for x B 1. That is, if (lower) alloys. Points are swelling rates calculated from
dislocations are the dominant sink by a certain ratio, swelling data and curves are drawn to aid the eye. After Ref.
the swelling rate is expected to be essentially the same [56], where details of references are provided.
as when the cavities are the dominant sink by the same
ratio.
Fig. 13 shows schematically the resulting swelling small a swelling scale and too early in dose to show up
versus dose. The top curve is the general behavior on a figure. In this case only the constant swelling rate
obtained. Three regimes of dose dependence can be with dose, n = 1, or the decreasing swelling rate with
recognized. In each regime the dose dependence ap- dose, n < 1, behavior may be reported, corresponding
proximately follows Va (dose)“. At low doses, disloca- to the lower middle and right plots. However, in the
tions typically are the dominant sink, Q X- 1, and n > 1. lower left curve, the experimental data collection cov-
At intermediate doses Q = 1, and in this case n = 1. At ers only the condition where Q B 1, n > 1 is observed.
high doses cavities typically can become the dominant The other behaviors (n = 1 or n < 1) would be ex-
sink, Q s 1, for this case n < 1. This latter case em- pected to occur at doses higher than those reported.
bodies the situation where most point defects of either A large array of cavity swelling data has been shown
type are absorbed at cavities, essentially resulting in to be consistent with the above predictions. Some of
efficient annihilation of interstitials and vacancies at these data are shown in Fig. 25 for stainless steels and
the sink with little net absorption. Therefore, swelling other alloys, and for diverse irradiation conditions cov-
cannot increase rapidly with dose after the condition ering both ion and neutron damage [57]. The expected
Q 4 1 is achieved. trends are reflected in the data, i.e., for either Q >> 1
The lower part of Fig. 13 shows individual regimes or Q +Z 1, the swelling was small. For Q = 1, the
of the upper curve, corresponding to types of behavior largest swelling rates were observed. It should be em-
often observed experimentally. Reported experimental phasized that correlation of a given Q value with
data may exhibit only one (or two) regimes. For exam- swelling in this way corresponds to theoretical predic-
ple, the entire period of Q B 1, n > 1 may occur at too tions in limiting cases. Where the microstructure
120 L.K. Mansur/Journal of Nuclear Materials 216 (1994) 97-123

evolves such that swelling does not occur predomi-

nantly in a limiting regime, then a simple limiting case
analysis, corresponding to a snapshot for one set of
conditions, would not be applicable. In practice, then,
a
comparing the Q value of the microstructure at the
highest dose with the measured swelling may not corre- ;
0

spond to an expected limiting behavior if much of the 4

swelling did not occur under that condition. Neverthe- 4
less, theoretical predictions can always be compared
with experiments with the additional effort of numeri-
cal integrations.
4.4. Precipitate-assisted cavity growth + 2 3

rC”O
4 5

Influences of precipitates on cavity swelling were

discussed in terms of direct, indirect and mediated Fig. 26. Calculated curves for precipitate-enhanced cavity
effects in Ref. [15]. As described there, the scope of growth for three ratios of precipitate radius, rp, to initial
precipitate effects is extremely broad. Above, in Sec- cavity radius, rO. The symbols rc and rW denote the matrix
tion 4.2, we described an effect of a finely dispersed cavity radius and the precipitate-attached cavity radius, re-
second phase. The high interfacial area effectively dis- spectively. The cross-hatched area indicates the experimental
perses helium into many bubbles and thereby delays data. After Ref. [58].
the achievement of the critical number of gas atoms in
a cavity. This delay in turn postpones the onset of
swelling. Another example of a direct effect on swelling
is the collection of point defects at the precipitate
matrix interface and the consequent acceleration of the This ratio results from the fact that the precipitate-
growth rate of an attached cavity. The effect was cavity pair represents a larger sink than the cavity in
predicted on theoretical grounds [16] and confirmed by the matrix.
experiments [58]. For cavities of the order of or less than the attached
In the theoretical work it was assumed that the precipitate size, Eq. (48) predicts that the enhance-
particle-matrix interface acts as a collector for point ment in growth rate is significant. The equation pre-
defects. This interface was taken also to provide a path dicts that in the time required for a matrix cavity to
for the diffusion of point defects to the attached cavity. increase to several times its initial size, a precipitate-at-
This collector behavior augments the absorption of tached cavity grows to a size several times greater than
point defects at the cavity-matrix interface. The ratio the matrix cavity. This model suggests that for a given
of the growth rate of a cavity attached to a particle to precipitate number density (after the critical radius has
the growth rate of a cavity in the matrix was expressed been achieved - see Section 4.2), swelling is dictated
in terms of three multiplicative functions: by precipitate size. Some after-the-fact comparisons of
drq - rcrb neutron irradiation-induced swelling were made and
The, were consistent with the above predictions [57]. How-
dr, rep
ever, a much more convincing critical experiment was
where rC,, is the radius of a cavity attached to a also carried out [58].
precipitate and r, is the radius of a cavity in the Fig. 26 shows the integrated effect of point defect
matrix. The quantity rb is the effective radius of the collection at the particle interface on cavity size. It is
cavity-precipitate pair [16]. The b in Eq. (47) is a bias predicted, for example, that in the dose interval re-
modification function, which involves the products and quired for the matrix cavity to growth by a factor of 3,
differences of the capture efficiencies and densities of the precipitate-attached cavity grows by a factor of _ 6
all sinks in the material. Under certain conditions, this for an initial precipitate to cavity radius ratio of 15.
term may be taken as unity. The term e is a sink These predictions were tested using ion irradiation
strength modification function. When the growth rates experiments.
of attached cavities and matrix cavities are being com- A stainless steel alloy was first irradiated with 4
pared in the same microstructure this function also MeV ions to a peak dose of 70 dpa at 657°C. This
reduces to unity and the ratio of growth rates of produced coarse G-phase particles. Irradiation was next
attached and unattached cavities is given by carried out with a-particles to a total helium concen-
, A ..\1/2
dr (rc”p+ri) ’ tration of 400 appm. This resulted in helium bubbles
cp= on the dislocations and at the interfaces between the
I, .
dr, rc:, G-phase particles and the matrix. The average size of
 L.K. Mamur /Journal of Nuclear Materials 216 (1994) 97-123 121

the bubbles at the particles and elsewhere were simi- types - after the fact juxtapositions of theory with
lar. Further irradiation was then carried out with 4 relevant experimental data, and comparisons of theory
MeV Ni ions to a peak dose of 20 dpa. During this with specially designed controlled-variable critical ex-
irradiation, growth of both the dislocation-nucleated periments performed as direct tests of the theory.
and the particle-nucleated cavities occurred. The pre- The kinetic theory of radiation effects spans a vast
cipitate-attached cavities grew at a substantially faster range of phenomena that ultimately translate the ini-
rate than the matrix cavities. The dislocation-nucleated tial interactions of irradiating particles with matrix
cavities increased their radii by factors of 2-3. During atoms to macroscopically measurable property changes.
the same time interval, the precipitate-attached cavi- In its most general form, all transport and reaction
ties grew by factors of 4-8. The initial precipitate to processes of point and extended defects are included
cavity size ratio was in the range 15-30. The experi- in principle. Particular aspects reviewed here include
mental data are represented by the cross-hatched area the framework of clustering and the concept of irradia-
in Fig. 26, superimposed on the calculated curves. tion variable shifts based on point defect absorption
These results confirmed the predictions of the precipi- invariance. Cascade diffusion theory is outlined and
tate point defect collector model. contrasted with standard homogeneous rate theory.
More specific models for swelling including bias, sink
4.5. Startup-induced transient interstitial creep strengths, critical radius and the effects of sink strength
ratios are described. Aspects of creep by climb-enabled
Above, in Section 3.9, we described three mecha- glide of dislocations are covered. Transient point de-
nisms of irradiation creep driven by transients in the fect processes are emphasized, since it has been real-
point defect concentrations. The mechanism of ized more recently that such transients may contribute
startup-induced transient interstitial creep, which oper- more dimensional change than steady-state processes
ates at relatively low temperatures, consistently ex- under some circumstances.
plains observations of high creep rates in recent neu- The theory based on defect reactions has estab-
tron irradiation experiments [42,48]. lished a basic understanding of swelling. Alloys and
Using the time-dependent solutions of Eqs. (6) and special experiments based on theoretical concepts have
(7) together with derived expressions that translate the been successful at demonstrating swelling resistance
associated climb into glide [39], Stoller et al. [48] calcu- and other predictions. Swelling resistant alloys can now
lated the creep by this process. Fig. 19 shows the be designed with confidence for technological applica-
results. Below about 200°C most of the creep is calcu- tions. Irradiation creep is understood to a reasonable
lated to be from transient interstitial absorption with a level, and calculations for creep are in rough agree-
minor part contributed by steady-state processes. At ment with experiments.
temperatures much above 2OO”C,most of the creep is Challenging prospects for future work can perhaps
from steady-state processes. The points on Fig. 19 best be framed as questions. For example, in other
represent the experimental measurements of irradia- papers in this volume, discrete atomic descriptions are
tion creep in the ORR at about 8 dpa and at the used to calculate point defect production. Continuum
temperatures indicated. The theoretical calculations reaction rate theory, as described herein, makes use of
show reasonable agreement with the experiment and the results of such calculations as relatively simple
are consistent with the idea that startup-induced tran- source terms for point defects, for in-cascade recombi-
sient interstitial creep is the main contributor to irradi- nation, and for in-cascade clustering. The level of the
ation creep in neutron-irradiated stainless steels below two descriptions is completely different. What is the
about 200°C. optimum mixture or fusion of the continuum and dis-
crete approaches to achieve the best description of
structure, composition and property changes? Another
5. Summary and conclusions worthy question is how best to take advantage of the
extensive theoretical work on radiation-induced
A tutorial is presented that covers dimensional swelling and creep to advance the mechanistic under-
changes in structural alloys caused by irradiation. standing of other related phenomena that have not
Background, including experimentally observed phe- been treated so thoroughly, such as hardening and
nomena as well as facilities and applications, is first embrittlement. As outlined in this paper, the two most
sketched. Dimensional changes are emphasized since powerful tools for understanding how irradiation leads
historically they have received most attention. This is to property changes are theoretical modeling and the
followed by a more or less self-contained description of thorough-experimental characterization of microstruc-
the kinetic theory of radiation-induced swelling and ture. How can we further improve the combined appli-
creep. Lastly, several comparisons of the theory with cation of these tools to attack new problems in radia-
experiments are described. The comparisons are of two tion effects?
122 L.K Man.w/Joumal of Nuclear Materials 216 (1994) 97-123

References [25] L.K. Mansur, A.D. Brailsford and W.A. Coghlan, Acta
Metall. 33 (19851 1407.
[l] L.K. Mansur, in Kinetics of Nonhomogeneous Processes, [26] C.H. Woo, B.N. Singh, as in Ref. 7; C.H. Woo and B.N.
ed. G.R. Freeman (Wiley, New York, 1987) pp. 377-463; Singh, Philos. Mag. 65A (1992) 889.
L.K. Mansur, Nucl. Technol. 40 (19781 5. [27] AC. Damask, G.J. Dienes, Point Defects in Metals
[2] C. Cawthorne and E.J. Fulton, Nature 216 (1967) 575. (Gordon & Breach, New York, 1963).
[3] ST. Konobeevsky, N.F. Pravdyuk and V.I. Kutaitsev, [28] A.D. Brailsford, J. Nucl. Mater. 60 (19761 257.
Proc. on Peaceful Uses of Atomic Energy, Geneva, 1955, [29] R.E. Stoller and L.K. Mansur, Proc. Int. Conf. on Radia-
vol. 7 (United Nations, New York, 19561 p. 433. tion Materials Science, Alushta, USSR, May 1990, vol. 1
[4] J.W. Joseph, Jr., USAEC Report DP-369, E.I. DuPont de (Kharkov Institute of Physics and Technology) pp. 52-67.
Nemours and Co. (1959). [30] L.K. Mansur, J. Nucl. Mater. 78 (1978) 156; 206 (1993)
[5] AS. Zaimovsky, G.Y. Sergeev, V.V. Titova, B.M. Levit- 306.
sky and Y.N. Sokursky, Proc. 2nd Conf. on the Peaceful [31] A.D. Brailsford and R. Bullough, Philos. Trans. Roy.
Uses of Atomic Energy, vol. 5 (United Nations, Geneva, Sot. London 302 (1981187.
19581 p. 566. [32] A.D. Brailsford and R. Bullough, Philos. Trans. Roy.
[6] K. Farrell, P.J. Maziasz, E.H. Lee and L.K. Mansur, Sot. London 302 (1981).
Radiat. Eff. 78 (1983) 277. [33] R.E. Staller and G.R. Odette, Proc. 13th Int. Symp. on
[7] C. Abromeit, in these Proceedings J. Nucl. Mater. 216 Radiation Induced Changes in Microstructure, part I,
(19941 78. ASTM-STP 955, eds. F.A. Gamer, N.H. Packan, and
[8] L.K. Mansur, E.H. Lee, M.B. Lewis and S.J. Zinkle, AS. Kumar (American Society for Testing and Materials,
Applications of multiple-ion irradiations to metals, ce- Philadelphia, 19871 pp. 371-392.
ramics, and polymers, Proc. Int. Conf. on Evolution in [34] F.R.N. Nabarro, Report on the Conf. on the Strength of
Beam Applications, November 1991, Takasaki, Japan, ed. Solids (Physical Society, London, 1947) p. 75.
by Japan Atomic Energy Research Institute, published by C. Herring, J. Appl. Phys. 21 (1950) 437.
Radiation Application Development Association. [35] P.T. Heald and M.V. Speight, Philos. Mag. 29 (19741
[9] R. Klueh, R.E. Stoller and D.S. Gelles, eds., Proc. 5th 1075.
Int. Conf. on Fusion Reactor Materials, J. Nucl. Mater. [36] W.G. Wolfer and M. Ashkin, J. Appl. Phys. 47 (1976)
191-194 (1992). 791.
[lo] D.G. Rickerby, H. Stamm, K. Ehrlich, M. Victoria, eds., [37] R. Bullough and J.R. Willis, Philos. Mag. 31 (19751 855.
Proc. 6th Int. Conf. on Fusion Reactor Materials, J. [38] C.H. Woo, J. Nucl. Mater. 120 (1984) 55.
Nucl. Mater. 212-215 (1994). [39] L.K. Mansur, Philos. Mag. A39 (1979) 497.
[ll] J.O. Stiegler, Oak Ridge National Laboratory, as quoted [40] L.K. Mansur and W.A. Coghlan, in Point Defects and
in Ref. 111. Defect Interactions in Metals, eds. J.I. Takamura, M.
[12] M.L. Grossbeck and L.K. Mansur, Mater. Sci. Forum Doyama and M. Kiritani (University of Tokyo Press,
97-99 (1992) 407. 19821 p, 907.
[13] J.O. Stiegler and L.K. Mansur, Ann. Rev. Mater. Sci. 9 [41] L.K. Mansur, W.A. Coghlan, T.C. Reiley and W.G.
(1979) 405. Wolfer, J. Nucl. Mater. 103&104 (19811 1257.
[14] J.R. Matthews and M.W. Finnis, J. Nucl. Mater. 159 [42] M.L. Grossbeck, L.K. Mansur and M.P. Tanaka, Proc.
(1988) 257. 14th Int. Symp. on Effects of Radiation on Materials, vol.
[15] L.K. Mansur, M.R. Hayns and E.H. Lee, Proc. Conf. on II, ASTM-STP 1046, eds. N.H. Packan, R.E. Staller and
Phase Stability During Irradiation, Pittsburgh, PA, Octo- AS. Kumar (American Society for Testing and Materials,
ber 5-9, 1980, eds. J.R. Holland, L.K. Mansur and D.I. Philadelphia, 19901 p. 537.
Potter (The Metallurgical Society of the American Insti- [43] A.D. Brailsford, W.A. Coghlan and L.K. Mansur, Proc.
tute of Mining, Metallurgical and Petroleum Engineers, 15th Int. Symp. on Effects of Radiation on Materials,
New York, 1980) p. 359. ASTM-SIP 1125, eds. R.E. Stoller, AS. Kumar and D.S.
1161 L.K. Mansur, Philos. Mag. A44 (19811 867. Gelles (American Society for Testing and Materials,
[17] E.A. Kenik, in Ref. [lo]. Philadelphia, 19921 p. 540.
[18] L.E. Rehn, P.R. Okamoto, D.I. Potter and H. Wieder- [44] J.H. Gittus, Philos. Mag. 25 (1972) 345.
sich, J. Nucl. Mater. 74 (19781 242. [45] D.J. Michel, P.L. Hendrick and A.G. Pieper, J. Nucl.
[19] E.P. Simonen, L.A. Charlot and S.M. Bruemmer, Proc. Mater. 75 (1978) 1.
15th Int. Symp. on Effects of Radiation Materials, [46] S.R. McEwen and V. Fidleris, Philos. Mag. 31 (19751
ASTM-STP 1125, eds. R.E. Stoller, A.S. Kumar and D.S. 1149.
Gelles (American Society of Testing and Materials, [47] E.P. Simonen, J. Nucl. Mater. 90 (1980) 282.
Philadelphia, 1992) pp. 621-632. [48] R.E. Stoller, M.L. Grossbeck and L.K. Mansur, Proc.
[20] M.T. Robinson, ibid. Ref. [7]. 15th Int. Symp. on Effects of Radiation on Materials, vol.
1211 L.R. Greenwood, ibid. Ref. [7]. II, ASTM-SIP 1125, eds. R.E. Stoller, A.S. Kumar and
[22] R.S. Averback, ibid. Ref. [7]. D.S. Gelles (American Society for Testing and Materials,
(231 J. Adams, ibid. Ref. [7]. Philadelphia, 1992) p. 517.
[24] D.J. Bacon, ibid. Ref. [7].
 L.K. Mansur/Joumal of Nuclear Materials 216 (1994) 97-123 123

[49] N.H. Packan, K. Farrell and J.O. Stiegler, J. Nucl. Mater. ness and N.L. Peterson (American Society of Metals,
78 (1978) 143. Metals Park, OH, 1976) p. 227.
[SO] L.L. Horton and L.K. Mansur, Proc. 12th Int. Symp. on [54] F.A. Garner and W.G. Wolfer, J. Nucl. Mater. 122&123
Effects of Radiation on Materials, ASTM-SIP 870, eds. (1984) 201.
F.A. Garner and J.S. Perrin (American Society for Test- [55] E.H. Lee and L.K. Mansur, J. Nucl. Mater. 141-143
ing and Materials, Philadelphia, 1985) pp. 344-362. (1986) 695.
[51] L.K. Mansur, E.H. Lee, P.J. Maziasz and A.F. Rowcliffe, [56] E.H. Lee and L.K. Mansur, Philos. Mag. A61 (1990) 733.
J. Nucl. Mater. 141-143 (1986) 633. [57] E.H. Lee and L.K. Mansur, Metall. Trans. A21 (1990)
[52] E.H. Lee and L.K. Mansur, Philos. Mag. A52 (1985) 493. 1021.
[53] W.G. Johnston, T. Lauritzen, J.H. Rosolowski and A.M. [58] E.H. Lee, A.F. Rowcliffe and L.K. Mansur, J. Nucl.
Turkalo, in Radiation Damage in Metals, eds. S.D. Hark- Mater. 103&104 (1981) 1475.