Essential Quantum Physics

This book provides a first course on quantum mechanics and describes
simple applications to physical phenomena that are of immediate and

everyday interest.
The first five chapters introduce the fundamentals of quantum mech-

anics and are followed by a revision quiz with which readers may test
their understanding. The remaining chapters describe applications,
including the theory of spin and its application to magnetic resonance
imaging, physics of lasers, molecular binding, simple properties of
crystalline solids arising from their band structure, and the operation of
junction transistors.
Ideal either as a course text or a self-study text, the book contains

nearly 100 exercises and hints to their solution. It will be valuable for
undergraduate students at 2nd or 3rd year in physics, electronic
engineering, mathematics, chemistry.
Downloaded from Cambridge Books Online by IP 150.244.221.208 on Thu Feb 13 17:09:15 GMT 2014.
http://ebooks.cambridge.org/ebook.jsf?bid=CBO9781139171243
Cambridge Books Online © Cambridge University Press, 2014
Essential Quantum Physics
Essential Quantum Physics
Peter Landshoff
University of Cambridge
Allen Metherell
University of Central Florida
Gareth Rees
University of Cambridge
CAMBRIDGE
UNIVERSITY PRESS
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore,
Sao Paulo, Delhi, Dubai, Tokyo
Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America by Cambridge University Press, New York
www. Cambridge. org

Information on this title: www.cambridge.org/9780521629935
© P. V. Landshoff, A. Metherell and W. G. Rees 1997
This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 1997

Reprinted 1999,2001
A catalogue recordfor this publication is available from the British Library
Library of Congress Cataloguing in Publication data

Landshoff, Peter, 1937-
Simple quantum physics. - 2nd ed. / Peter Landshoff, Allen Metherell, Gareth Rees.
p. cm.
Includes index.
ISBN 0 521 62011 2 (hb). - ISBN 0 521 62993 4 (pb)
1. Quantum theory. I. Metherell, Allen, 1937- . II. Rees, Gareth, 1959- . III. Title.
QC174.12.L37 1998
530.12-dc21 97-25151 CIP
ISBN 978-0-521-62011-6 Hardback

ISBN 978-0-521-62993-5 Paperback
Transferred to digital printing 2010
Cambridge University Press has no responsibility for the persistence or

accuracy of URLs for external or third-party internet websites referred to in
this publication, and does not guarantee that any content on such websites is,
or will remain, accurate or appropriate. Information regarding prices, travel
timetables and other factual information given in this work are correct at
the time of first printing but Cambridge University Press does not guarantee
the accuracy of such information thereafter.
Contents
Constants of quantum physics x
1 Preliminaries 1
Atoms; Photons; Wave nature of matter; Problems
2 The Schrodinger equation 8

Wave functions and operators; Example: the one-dimensional
potential well; Probability interpretation and normalisation;
Beams of particles; Continuity conditions; Problems
3 Special solutions 19
Particle in a box; The one-dimensional square well; The lin-
ear harmonic oscillator; The tunnel effect; The delta-function
potential; The WKB approximation; Alpha decay; Problems
4 The superposition principle 36

Linear operators; Wave packets; Ehrenfest's theorem; Her-
mitian operators; Operators and observables; Commutators;
Problems
5 The hydrogen atom 49

Good quantum numbers; Orbital angular momentum; Spher-
ically symmetric potentials; The hydrogen atom; Many-
electron atoms; Two-body systems; The deuteron; Problems
Revision quiz 62
6 The hydrogen molecule 63

The ionised hydrogen molecule; Other molecules; Problems.
7 Introduction to perturbation theory 70

Time-independent perturbation theory; Time-dependent per-
turbation theory; Transition probability; Energy uncertainty
principle; Sudden change in the Hamiltonian; Example: decay
of tritium; Problems
vn
Vlll
8 Spin 82
Two kinds of angular momentum; Spin | ; The electro-
magnetic interaction; The Zeeman effect; Spin precession;
Problems
9 Masers and lasers 93

Radiative transitions; Resonant absorption and stimulated
emission; Electric dipole transitions; The ammonia molecule;
The ammonia maser; Population inversion; The laser; Holog-
raphy; Problems
10 Band structure of crystals 112

Electrons in crystals; Band structure; Number of levels in a
band; Band overlap; Simple consequences of band structure;
Problems
11 Electron motion in crystals 126

Electron velocity; Motion in an external electric field; Electric
current; Effective mass and holes; Thermal excitation; Pair
excitation in intrinsic semiconductors; Problems
12 Transistors 141
Impurities; n- and p-type semiconductors; Impurities and
crystal colour; Semiconductor junction; The diode; The junc-
tion transistor; Two simple circuits; Problems
Appendices
A Power-series solutions 158
B The delta function and Fourier transforms 162
C Orbital-angular-momentum operators 173
D Electrodynamics 175
E Bloch waves 178
Hints for the problems 180
Index 201
Preface
This book is intended as a first course on quantum mechanics and its

applications. It is designed to be a first course rather than a complete
one, and it is based on lectures given to mathematics and physics students
in Cambridge. The book should be suitable also for engineering students.
The first five chapters deal with basic quantum mechanics, and are fol-
lowed by a revision quiz to test the reader's understanding of them. The
remaining chapters concentrate on applications. In most courses on quan-
tum mechanics, the first application is to scattering problems. While
recognising the importance of scattering theory, we have chosen rather
to describe the application of quantum mechanics to physical phenomena
that are of more everyday interest. These include molecular binding, the
physics of masers and lasers, simple properties of crystalline solids aris-
ing from their electronic band structure, and the operation of junction
transistors.
A few problems are included at the end of each chapter. We urge the
student to work through all of these, as they form an integral part of the
course. Some hints on their solution may be found at the end of the book.
A previous edition of this book was published under the title Simple
Quantum Physics in 1979. In this new edition, the main change is the
addition of a chapter on the theory of spin, and its application to magnetic
resonance imaging. We have also described the WKB approximation
and its application to a-decay, and have made a number of other minor
changes.
Christ's College, Peter Landshoff

Cambridge Allen Metherell
January 1997 Gareth Rees
IX
http://dx.doi.org/10.1017/CBO9781139171243.001
Constants of quantum physics
Dirac's constant h = h/2iT = 1.05 x 10" 3 4 J s

Charge of electron -e = -1.60 x 10" 19 C
Fine-structure constant e2IA-KSQTIC = 1/137
Speed of light c = 3.00 x 108 m s" 1
Mass of electron rae = 9.11 x 10" 3 1 kg = 0.51 MeV/c 2
Mass of proton rap = 1.67 x 10" 2 7 kg = 938 MeV/c 2
Electron volt 1 eV = 1.60 x 10" 1 9 J
Boltzmann's constant kB = 1.38 x 10" 2 3 J K" 1
1
Preliminaries
Atoms
An atom consists of a positively charged nucleus, together with a number
of negatively charged electrons. Inside the nucleus there are protons, each
of which carries positive charge e, and neutrons, which have no charge.
So the charge on the nucleus is Ze, where Z, the atomic number, is the
number of protons. The charge on each electron is —e, so that when the
atom has Z electrons it is electrically neutral. If some of the electrons are
stripped off, the atom then has net positive charge; it has been ionised.
The electrons are held in the atom by the electrostatic attraction between
each electron and the nucleus. There is also an attraction because of the
gravitational force, but this is about 10~40 times less strong, and so may
be neglected. The protons and neutrons are held together in the nucleus
by a different type of force, the nuclear force. The nuclear force is much
stronger than the electrical force, and its attraction more than counteracts
the electrostatic repulsion between pairs of protons. The nuclear force
does not affect electrons. It is a very short-range force, so that it keeps
the neutrons and protons very close together; the diameter of a nucleus is
of the order of 10~15 m. By contrast, the diameter of the whole atom is
about 10~10 m, so that for many purposes one can think of the nucleus
as a point charge. The mass of the proton or neutron is some 2000 times
that of the electron, so nearly all the mass of the atom is in the nucleus.
It is natural to think of the electrons as being in orbit round the nucleus,
figure 1.1, just as the planets are in orbit round the sun. The electrostatic
force that keeps the electrons in their orbits is an inverse-square-law force,
just as is the gravitational force that keeps the planets in orbit, so that the
two systems would seem to obey precisely similar equations. However,
there is a serious difficulty. When a particle moves in a curved orbit its
velocity vector is continuously changing: the particle is being accelerated
towards the centre of its orbit. According to classical electrodynamics,
when a charged particle is accelerated it inevitably radiates energy (this
is the basic principle of radio transmission). So according to classical
physics the electron would continuously lose energy and its orbit would
form a spiral which would gradually collapse into the nucleus.
1 Preliminaries
Figure 1.1. Classical picture of negatively charged electrons in orbit round the posi-
tively charged nucleus of an atom.
The reason that this does not happen is that very small systems, such as
atoms, do not obey classical mechanics. To describe an atom one has to
use quantum mechanics. In quantum mechanics, as opposed to classical
mechanics, one cannot arbitrarily choose a value for the energy of the
orbiting particle and then find an orbit corresponding to that energy; only
certain discrete values of the energy are allowed. When the electron is in
its lowest allowed energy level, it cannot radiate any more energy, and so
the total collapse of the atom is not possible.
One can also use quantum mechanics to describe the solar system. Just
as for the electrons, the allowed energy levels of the planets are discrete.
If a planet in orbit is given an impulse, its energy is allowed to change
only to that of one of the other allowed discrete levels. However, the
separation between these levels is so small that this is not a very real
restriction, and classical mechanics is perfectly adequate to describe the
system. The effects of quantum mechanics are generally only important
for submicroscopic systems.
The chemistry of an atom is determined by the charge on its nucleus.
Thus atoms whose nuclei differ only in the number of neutrons that they
contain have similar chemical properties; they are said to be isotopes
of the same element. For example, the atom of the common form of
hydrogen contains just a single proton, that is, Z = 1; but hydrogen has
a stable isotope, called deuterium, whose nucleus consists of one proton
and one neutron. Atoms can be bound together to form molecules (see
chapter 6), and different isotopes of the same element do this in the same
way. Ordinary water H2O consists of molecules containing two hydrogen
nuclei and one oxygen nucleus, while 'heavy' water D2O has deuterium
nuclei instead of the ordinary hydrogen. The chemical properties of heavy
Photons
water are exactly the same as those of ordinary water, but there are some
differences in its physical properties. In particular, it is denser because of
the extra neutrons.
Photons
In a metal, the atoms are effectively anchored to fixed sites by the electro-
static forces due to all the other atoms. The outermost orbital electrons of
the atoms are almost free, and move through the metal when an electric
field is applied (see chapter 11). If the metal is bombarded with light,
some of the electrons can actually escape from the surface of the metal
and can be detected as an electric current. This is the photoelectric ef-
fect. The number of electrons that escape in a given time rises with the
intensity of the beam of light, but the energy with which they escape does
not depend on the beam intensity. Rather it depends on the colour or
frequency v of the light. The kinetic energy T with which the electrons
escape is found to obey the equation
hu = T + W. (1.1)
Here h is Planck's constant,
h = 6.626 x 10" 34 J s,
and W is the energy that must be given to the electron to enable it to

overcome the electrostatic attraction of the metal. (The value of W varies,
according to the state within the metal from which the electron is ejected.
For a given metal, there is a definite minimum value Wo, called the work
function of the metal.)
These results are explained by the hypothesis that a beam of light can
be thought of as a collection of particles, called photons. The number of
photons is proportional to the intensity of the light, and the energy E of
each photon is proportional to the frequency,
E = hv. (1.2)
The electron is ejected from the metal when one of the photons collides
with it and is absorbed by it, so giving up all its energy to the electron.
The number of electrons ejected rises as the intensity of the light is in-
creased because there are then more incident photons, and so there is a
greater chance of a photon being absorbed.
4 1 Preliminaries
Photons move with the speed of light, so their kinematics must be de-
scribed by the laws of special relativity. The energy of a particle whose
speed is v and whose rest mass is m is
E = mc2/{l-v2/c2)ll\ (1.3a)
so that when v = c the energy can be finite only if m = 0; that is, photons
have zero mass. In terms of the momentum p of the particle, (1.3a) reads
E = c(m2c2+p2)1'2, (1.36)
so that for a photon

E = cp. (1.4)
If a beam of light is shone normally on a perfect conductor it is reflected,
that is, the momentum of each photon is reversed. This must occur by
some sort of force being exerted on the photons, and the conductor must
experience and equal and opposite force. This is a realisation of the clas-
sical idea of radiation pressure.
Equations (1.2) and (1.4) are tested in the Compton effect. When photons
collide with free electrons or protons, not bound into a solid, they cannot
be absorbed because it can be shown (see problem 1.2) that this would
violate conservation of energy and momentum. (In the photoelectric effect
some of the energy, W, is absorbed by the other particles in the metal.)
However, a free particle can scatter the photon, so changing its energy
and therefore its frequency; at the same time the particle recoils. The
kinematics of the process can be worked out using (1.2), (1.4) and the
relativistic energy-momentum conservation laws (see problem 1.4), and
the results are found to agree with experiment.
Equation (1.2) also helps to explain atomic spectra. We have said that,
according to the results of quantum mechanics, the allowed energy levels
of the electrons in atoms are discrete. If a beam of light is shone on a
collection of atoms, the photons can be absorbed by the atoms if, and
only if, their energy is equal to the difference between the energies of
two electron levels. The absorption of the photon then excites the atom,
sending the electron from the lower to the higher level. (The number of
photons that can be absorbed in this way of course depends on how many
atoms happen to be in the lower of the two states to start with.) Thus only
photons with certain discrete frequencies are absorbed. Conversely, an
atom in an excited state can decay by emitting a photon; the frequency of
the photon depends on the difference between the initial and final energies:
hv = E2 -Ei.
Wave nature of matter
Slits Screen
Figure 1.2. The double-slit experiment. There is darkness at points on the screen such
that the path difference between rays that pass through the two slits is (n -f ^
The energy levels of an atom (or molecule) depend on what element it

is, so that the spectrum of frequencies absorbed and emitted provides a
useful way of identifying substances.

Although light can be thought of as a collection of photons, it also has
wave-like properties. For example, a coherent beam of light is diffracted
when it is shone through a pair of closely separated slits: if a screen
is placed at large distance behind the slits, a pattern of light and dark
fringes appears on it. The spacing of these fringes is calculated from the
wavelength A of the light. See figure 1.2. Dark fringes appear at points on
the screen such that their distances from the two slits differ by (n + |)A,
where n is an integer. Then the light received from the two slits is exactly
out of phase; the two components cancel.
So quantum mechanics gives light a dual nature. In some respects it
behaves like a collection of particles, in others like a wave. The same
is true for electrons and other particles; quantum mechanics associates
waves with every kind of particle.
The necessity for this is illustrated by the phenomenon of electron diffrac-
tion. If a beam of electrons is passed through a crystal, it is diffracted.
If a fluorescent screen is set up behind the crystal, a diffraction pattern
appears on the screen. The regularly spaced atoms in the crystal cause
the diffraction. The pattern can be explained by associating with the elec-
trons a wave of wavelength A, which changes with the momentum of the
6 1 Preliminaries
electrons according to de Broglie's relation
A - h/p. (1.5)
The waves may be assigned a characteristic frequency v, which may be

chosen so as to be related to the electron energy E by the relation E =
hv as for photons (see (1.2)). However, in the case of electrons, the
frequency v is not directly measurable and there is some arbitrariness in
its definition; for example, E may or may not include the rest-mass energy
of the electron. The part played in the theory by the electron frequency
will become apparent in the next chapter.
The de Broglie relation (1.5) applies also to photons. This follows if we
make the assumption that the quantum-mechanical waves that describe
photons have the same frequency v and wavelength A as the corresponding
classical electromagnetic waves. Because the classical waves have speed
c, this implies that
\v = c (1.6)
and combining this relation with (1.2) and (1.4) gives (1.5).
Classical electromagnetic waves are associated with a very large number
of photons (see problem 1.1). The waves of quantum mechanics may
describe either a collection of particles or a single particle. It is important
to understand that quantum-mechanical waves are more abstract than
classical waves. Consider an experiment where light is diffracted through
a pair of slits, or where electrons are diffracted through a crystal. Suppose
that only one photon or electron is allowed to come into the experiment. In
this case we cannot predict with certainty what will be the angle 9 through
which the photon or electron is diffracted. However, if the experiment is
repeated many times, we find a probability distribution for the angle 9 that
has the same shape as the variation of intensity with 9 in an experiment
where there is a continuous beam of photons or electrons.
This suggests that the association of a quantum-mechanical wave with a
photon, or with any other kind of particle, is somehow statistical. We
explore this in the following chapters. As will become clear, according
to quantum theory one can never predict with certainty what will be the
result of a particular experiment: the best that can be done is to calculate
the probability that the experiment will have a given result, or one can
calculate the average value of an observable quantity if the experiment
that measures it is repeated many times.
Problems
1.1 A radio transmitter operates on a wavelength of 100 m at a power of
1 kW. How many photons does it emit per second?
1.2 Using energy-momentum conservation, show that an electron that is
not in a bound state cannot absorb a photon.
1.3 A particle has mass 1 kg. How long does it take to move through a
distance of 1 m if its de Broglie wavelength, (1.5), is comparable with
the wavelength of visible light? What is the corresponding answer if
the particle is an electron?
1.4 A photon of momentum p, and therefore of wavelength /i/p, scatters
on an electron that is initially at rest. Using relativistic kinemat-
ics, deduce from the conservation of energy and momentum that as
the result of the scattering the wavelength of the photon changes by
(h/mc)(l — cos#), where 9 is the angle through which it scatters and
m is its rest mass. (This scattering process is known as Compton
scattering, and the quantity h/mc is the Compton wavelength of the
electron.)
1.5 Associated with the electron there is an antiparticle, the positron,
which has equal mass and equal, but opposite, charge.
A positron impinges on an electron which is at rest. They annihilate
into two photons. Show that the sum of the wavelengths of the two
photons is Ao(l — cos#), where 6 is the angle between their directions
of motion and XQ is the Compton wavelength of the electron.
2
The Schrodinger equation
The basic equation of quantum mechanics is known as the Schrodinger

equation. It is not possible to prove that the equation must be true, just
as we cannot prove that Newton's laws, the basis of classical mechanics,
must be true. All that can be done is to work out the consequences
of the equation in different physical contexts, and to compare them as
exhaustively as possible with experiment. In this chapter we begin by
showing how the experimental facts that we have already described make
the truth of the equation plausible.
The Schrodinger equation describes non-relativistic particles, whose en-
ergy E and momentum p are related by
E=p2/2m. (2.1)
Non-relativistic kinematics can be used so long as the energy E is not

comparable with, or larger than, the rest-mass energy me2. Relativistic
quantum mechanics is much more difficult than the non-relativistic theory,
and will not concern us in this book. In particular, we shall not describe
the quantum mechanics of the photon; for this particle m = 0, and so
relativistic mechanics must always be used.
Wave functions and operators

We explained in chapter 1 that with a particle of energy E and momentum
p we somehow associate a wave of frequency v — E/h and wavelength
A = h/p. Instead of using v and A, it is convenient to introduce the
angular frequency
and the wave vector fc, whose direction is in the direction of wave propa-
gation and whose magnitude is
k = 2TT/A.
We also work with

h = h/2n
Wave functions and operators
(pronounced '/i-cross') instead of h. Then we have
p = hk
E = hu. (2.2)
For a free particle, which is not interacting with any other particle or with
a potential, p and E are constant. Hence we expect such a particle to be
described by a wave for which k and u are constant:
tt(r,<)= tfei(*'r-wt). (2.3)
Here r denotes the position vector and TV is a constant. In many physical

applications, we write a wave in the form of a complex exponential, as is
done here, and understand that only the real part is physically meaningful.
In quantum mechanics, however, it will turn out that both the real and
the imaginary parts of \I> are needed. The plane wave (2.3) is the simplest
example of a wave function: it describes a free particle. Instead of (2.3),
we could use JVe""1^"1""^; the choice of sign is a matter of convention.
In order to guess how to arrive at wave functions that describe particles
that are not free, we perform a simple manipulation on (2.3). Differentia-
tion with respect to a component Xj (j = 1, 2 or 3) of r simply multiplies
\I/ by i times the corresponding component kj of k. So, with (2.2), we
have
)V=PjV (j = 1,2,3) (2.4a)
or, in more concise vector notation,
(-iftV)tf = p * . (2.46)
Equation (2.4a) says that if we apply the differential operator —ihd/dxj

to #, the result is to multiply \I> by the number pj. We say that # is
an eigenfunction of the operator (—ihd/dxj) with eigenvalue pj. (More
properly, since the differential equations (2.4) do not determine ^ until we
have imposed suitable boundary and continuity conditions on \I>, we should
at this stage specify these conditions. However, we defer the discussion
of this until the end of this chapter.) In this way, an experimentally
observable quantity, the momentum p of a particle, is associated with
a differential operator — iftV. We generalise this to other experimentally
observable quantities by introducing three basic assumptions:
10 2 The Schrodinger equation
(a) States of a system are described by wave functions # .

(b) Observable quantities are associated with operators.
(c) When the value of an observable Q is known to be g, the system is
in a state whose wave function is an eigenfunction of the operator
Q corresponding to Q, with eigenvalue q. That is,
qV. (2.5)
We shall elaborate on these assumptions in chapter 4.

Consider the particular example of the observable that is the energy of
a particle. If it is a free particle, this is just p2/2m. So one would
expect the corresponding differential operator to be obtained by replacing
p in this expression by its corresponding operator —i^V, and so we have
(-iftV) 2 /2ra, that is, — Ti2V2 /2m. If the particle moves in a potential
V(r), this operator becomesf
H = (-h2/2m)V2 + V(r) (2.6)
corresponding to the classical expression p2/2m + V(r). H is called the

Hamiltonian operator. To find the possible energy levels E of the particle,
we must find the eigenvalues E of H, that is, we must solve the equation
(2.7a)
(subject to appropriate boundary conditions). This equation
[(-h2/2m) V2 + 7(r)]tt(r, t) = £ * ( r , t) (2.7b)
is the time-independent Schrodinger equation.

Notice that if the particle is not a free particle, so that 7 ^ 0 , the plane
wave (2.3) is not a solution of the Schrodinger equation; the wave function
ty will be more complicated.
The time-independent Schrodinger equation applies when the particle is in
a state such that its energy takes a definite value E. In a general situation
this may not be the case, for example when V depends explicitly on the
time t as well as on the position r of the particle. In order to guess what
f Although H is an operator, we shall follow the usual convention and not write the
operator symbol " over it.
Example: the one-dimensional potential well 11
is the generalisation of the Schrodinger equation that describes this, we

return again to the plane-wave solution (2.3) appropriate to a free particle.
Making use now also of the second relation in (2.2), we find that the plane
wave satisfies
The obvious generalisation of this to the case where V ^ 0 is
[(-h2/2m) V2 + V]*(r, t) = \h 8V/dt (2.8a)
or
HV = ihdV/dt. (2.8b)
This is the time-dependent Schrodinger equation.
This is the equation that is valid in all circumstances whether or not the
particle is free and whether or not it is known to be in a state of definite
energy. In the particular case where the particle is known to be in a state
of definite energy E, the time-independent equation (2.7) is valid also.
When both equations hold,
This means that the time dependence of \P(r,4) is then trival:
y(r,t) = $(r)e-iEt/n. (2.9)
The function ^ ( r ) is called the time-independent wave function and it

satisfies the time-independent Schrodinger equation
Hip = E\j). (2.7c)
We repeat that a solution of the form (2.9) is applicable only when the
energy takes a definite value. For reasons that will be explained below, a
solution of this type is known as a stationary-state solution. Comparing
(2.9) with (2.3), we see that in the special case of a free particle
if>(r)=Neik'r (2.10a)
where, from (2.7),

E = h2k2/2m. (2.106)
Example: the one-dimensional potential well

It turns out that for a particle in a bound state, for example an elec-
tron bound in an atom, the values of E that are allowed by the time-
independent Schrodinger equation are discrete. That is, the energy levels
of a bound state are quantised.
The simplest mathematical model of a bound-state situation is as follows.
Suppose that the wave function ip depends on only one coordinate, x say,
and let the potential be the infinite square well
0 0<x<a ( 2 n )
oo otherwise.
As only one of the three coordinates is explicitly involved, we call this a

one-dimensional model. To find the states of definite energy E we solve
the time-independent Schrodinger equation, which here reduces to
[{-h2/2m)(d2/dx2) + V]t/>{x) = Et/){x). (2.12)
This equation is to be valid for all x. If E is finite, we deduce that in

the region outside 0 < x < a, where V is infinite, the wave function ip(x)
must vanish. Otherwise the term VI/J in the Schrodinger equation (2.12)
would be infinite in this region, which is not possible because neither of
the other two terms in the equation would be similarly infinite, and so the
equation would not be satisfied. In the region 0 < x < a, where V = 0,
the Schrodinger equation reduces to
IL / £ lib] (1/ I Ju / ""~~ 1 j (// I %XJ I
and its general solution for ^ is a linear combination of smkx and coskx,
where E — Ti2k2/2m.\ As we shall see at the end of this chapter, ip(x)
has to be continuous; in particular, it has to be continuous at x = 0 and
x — a. Hence, since it vanishes for x < 0 and for x > a, we must impose
the conditions tp(O) = ip(a) — 0. The condition at x = 0 picks out the
sin&x solution for -0, and that at x — a imposes a restriction on the
allowed values of k:
k = nn/a n = 1,2,....
f Throughout we shall use primes to denote differentiation with respect to a space vari-
able and dots to denote differentiation with respect to a time variable.
Probability interpretation and normalisation 13
We could also allow negative integers n, but this would not give any
additional wave functions; the allowed wave functions are
Mx) = {Cn sm(nnx/a) 0 < x< a }

^ nV ' \ 0 elsewhere. v J
where the Cn are normalising constants (we explain below how these are
chosen). The corresponding allowed energy values are
En = n2h2n2/2ma2 n = 1,2,.... (2.136)
That is, the allowed energy values for the bound system are discrete.
Probability interpretation and normalisation

As we have said in chapter 1, the wave function \& associated with a parti-
cle in quantum mechanics does not describe a concrete wave; it is not like
the waves of classical physics. Indeed, it was not until after it had been
discovered that the Schrodinger equation gives the right energy levels of a
system that the correct interpretation of \I> was suggested. We explained
that the wave function has a statistical interpretation, and the basic as-
sumption is that |\I/(r,£)| 2 is a probability density. That is, |\&(r,£)| 2 d 3 r
is the probability that a measurement of the position of the particle de-
scribed by \I> will give a result in the infinitesimal volume element d 3 r in
the neighbourhood of the point r. In order for this interpretation to be
consistent, \I/ must be arranged to satisfy the normalisation condition
(2.14)
/ •
Here we have used the symbol / d 3 r as a shorthand for JJJ dx dy dz, and
the integration is supposed to extend over all space. So (2.14) corresponds
to saying that there is unit probability that a measurement of the position
of the particle finds that the particle is actually somewhere.
If we have found a solution # to the Schrodinger equation that is not
correctly normalised so as to satisfy (2.14), we can usually get one that
is by simply multiplying by a suitable constant. An exception is the
special case of the plane wave (2.3), for which the normalising integral
(2.14) diverges for all values of the multiplying constant N. The reason
for this is that strictly the plane wave does not correspond to a physical
situation. It has |\I>|2 = \N\2, independent of r, so that it gives equal
probability of finding the particle anywhere throughout space. In practice,
one knows that the particle is confined within some volume £)o> f° r example
within a given building, and that therefore the wave function vanishes
outside fio- No problems arise for finite f2o> &s then the normalisation
integral converges, and by dividing ^ by a suitable finite constant we can
make it converge to 1, as is required in (2.14). The plane-wave solution
(2.3) is best thought of as a mathematical limit, where QQ —>• oo, of a
physically allowed solution. Apart from the normalisation problem, which
can in fact be circumvented by methods that we shall not discuss here, the
mathematics of the f2o —> oo situation is rather simpler than that of the
proper wave functions, and if the volume QQ is large (for example when
fi0 represents the volume of a typical experimental apparatus, which is
almost always very large when measured on a quantum-mechanical scale),
the numerical results are changed very little by taking the limit QQ —> oo.
Suppose that the particle is confined to a finite volume £2o> so that its
wave function * vanishes outside that volume. Then if \l/(r, t) is correctly
normalised at a given time i, it remains so. For, if we integrate over any
volume fi,
d
* Jn Jn
= Jn[ d r q3
3
(*** + * * * ) . (2.15a)
To evaluate this, we use the Schrodinger equation and its complex conju-
gate:
[(-h2/2m)V2
[(-h2/2m)V2 +
Thus
—
at 2miJa '
= -77—. \ d3r V • O T V * - * V * * ) . (2.156)
Probability interpretation and normalisation 15
Figure 2.1. The integration volume for (2.15), and its bounding surface S.
The divergence theoremf can now be used to convert this to an integral

over the surface S that encloses the volume of integration (see figure 2.1):
=- [ [dS-j
Js
where
j = (ft/2rai)(\I>*V\I/ - * V * * ) . (2.15c)
This relation is valid for any volume Q. For the normalisation integral
(2.14), we have to integrate over all space. However, since we are assuming
that \I> vanishes outside some volume f2O5 it is sufficient to integrate over
any volume that contains fi0- Because the surface S is then outside the
volume $~20? ^ , and therefore j , vanishes all over it, and so the integral
(2.15b) vanishes.
What we have said so far about normalisation applies to any wave function
\I/(r, t). When the wave function corresponds to a state of definite energy,
f The divergence theorem states that for any suitably well-behaved vector U
= f dS u.
Js
Here the vector dS is a vector whose magnitude is equal to an element dS of
the area of the surface S bounding the volume Q, and its direction is that of the
outward normal to S. See, for example, H Jeffreys and B S Jeffreys, Methods of
Mathematical Physics, (Cambridge University Press)
its time dependence factorises out trivially, as in (2.9). In this case
,*)| 2 = h/>(r)|2
so that the probability density at each point r is independent of time.
This is the reason that we call a state of definite energy a stationary state.
Beams of particles
So far, we have considered a single particle, such as an electron. Exper-
iments often involve beams of particles, for example a beam of electrons
from an accelerator fired at an atom. If the beam is not too intense, that
is, the particle density is not too high, each electron interacts with the
atom independently of the presence of the other electrons, and interactions
between the electrons are negligible. This means that we may describe the
problem in terms of the same Schrodinger equation as describes a single
electron moving in the potential produced by the atom.
Suppose that when there is just one electron the potential produced by the
atom results in a wave function \&(r, t). In the case of an electron beam,
the same wave function describes the density of particles at the point r at
time t. However, it is convenient then to normalise \I> not by the integral
(2.14) appropriate for a single-particle problem, but instead such that the
integral Jd 3 r|\I/(r, i)\2 is the total number of particles at time t. Then
|#(r,£)| 2 = p(r,t). (2.16)
Here p(r,i) is the average density that would be measured by setting up

the same state many times and repeatedly measuring the particle density.
This average value is called the expectation value of the particle density.
In addition to the expectation value of the particle density, p, we may
define j , the expectation value of the particle flux. This is such that, if we
take any closed surface 5,
A q • _ the expectation value of the number of particles

per unit time that cross the surfaces'.
A positive value for this integral would correspond to a depletion of the

total number of particles within S:
(2.17)
Continuity conditions 17
With # now normalised according to (2.16), this is just the relation

(2.15c). That is, the appropriate expression for j is
j A ( . (2.18)
2rai
In most cases, the potential and the rate at which the beam delivers par-
ticles to the target vary very slowly, if at all, on the atomic timescale.
Hence there is effectively a stationary-state situation, with the wave func-
tion factorising as in (2.9) and p and j independent of t.
Continuity conditions
The Schrodinger equation (2.8) is supposed to be valid at all points r.
This means that if the potential V is well behaved, the wave function and
its space derivative must both be continuous functions of r. For if this
were not so, the second derivative V 2 ^ would be infinite at the points of
discontinuity, and this cannot be because there cannot be just one term
in the Schodinger equation that has an infinity. The infinity would have
to be balanced by another infinite term and there is no such term unless
V has an infinity.
The physical consequences of this are not surprising: both p(r,t) and
j(r,t) have no discontinuous changes as r varies. That is, no particles
are being created or destroyed at any point r.
Problems
2.1 How are the stationary-state solutions of the Schrodinger equation
changed if a constant is added to the potential V(r)? Show that
this change has no observable consequences. (This means that, as in
classical mechanics, the point at which V vanishes may be chosen ar-
bitrarily. By convention, V is often defined so as to vanish at infinity.)
2.2 For each of the following, estimate the difference between the speeds
of the particle when it is in the first excited state and when it is in the
lowest-energy state: (i) an electron confined by an infinite square-well
potential whose width is roughly equal to the radius of an atom (about
2Q-io m j . ^ a t e n n i s b a n confined by an infinite square-well potential
whose width is equal to the width of a tennis court. What do your
answers tell you about the mechanics of microscopic and macroscopic
systems?
2.3 Determine the constant Cn in the wave function (2.13a) for the infinite
square well, so as to satisfy the normalisation condition (2.14). Show
that the probability interpretation of the wave function leads to the

average values
(x) = \a, ((x - (x))2) = (a 2 /12)(l - 6/n27r2).
Show that as n —> oo these average values approach the values that
are obtained from classical mechanics.
2.4 Show that for a one-dimensional system in a stationary state the par-
ticle flux (see (2.18)) is independent of both t and x. Does this result
have a simple generalisation to the three-dimensional case?
2.5 A system consists of two particles, so that its stationary-state wave
functions VKri? r2) are functions of the coordinates of each. What is
the obvious probability interpretation of ip?
Write down the time-independent Schrodinger equation. In the general
case, the potential will be a function F(ri, r2). What is the structure of
the function V in the special case where the particles interact only with
an external field of force, and not with each other? Show that in this
case the stationary-state wave functions have the structure ^(ri, r2) =
)V;2(^2)- What equations do the functions î and V>2 obey?
3
Special solutions
Particle in a box
Consider the stationary states of a particle confined within the rectangular
box-shaped volume
0 <x < a, 0 < y < 6, 0 < z < c. (3.1)
Suppose that within the box the potential is zero and outside it is infi-
nite. Within the box, the particle moves freely and the time-independent
Schrodinger equation is
2ra (ITS
~t~ dy2 +dzin
\dx2 + ITS 2
Elementary solutions to this equation are
sin sin sin

W\r) = kix k2y k3z
cos cos cos
where
E = (h2/2m)(k21+kl + k23). (3.3)
As we explained for the example of the one-dimensional infinite square-
well potential in chapter 2, because the potential is infinite at all points
outside the box, the wave function rp(r) must vanish for r outside the
box. To make ij)(r) continuous, it must also vanish on the faces of the
box. Hence (3.3) must vanish on each of the six planes
x = 0, y = 0, z = 0,
x = a, y = 6, z = c.
The first three conditions pick out the sin functions, rather than the cos
functions. The last three conditions impose restrictions on the allowed
values of &i, &2 and k3:
. qn rn sir
ki = — , k2 = — , k3 = —,
a b c
19
20 3 Special solutions
where q, r and s are integers. Thus the allowed wave functions are
1/2
sm
Vqrs ( r ) = — - — sin —^- sin , (3.4)
v }
\abcj a b c
and the allowed values of the energy are
hn (q2 r2 s2
^ 2m \ a2 b2 cz
Notice that all these allowed values of Eqrs are positive. In (3.4) we
have included the normalisation factor (8/abc)1/2, so as to satisfy the
normalisation condition
the integration being over the volume of the box, or equivalently, since
i\) — 0 outside the box, over all space.
The significant feature of these results is that the allowed values Eqrs
of the energy are discrete, because of the condition that 9, r and 5 are
integers. Notice that we may as well confine them to being non-negative
integers, since changing the sign of any one of them merely changes the
sign of the wave function, a change that has no physical significance.
These are the stationary-state solutions. The corresponding time-vary ing
wave functions are, from (2.9),
*qrs(r,t) = ^grs(r)e-iE<>r*t/h. (3.6)
In general, if we imagine that we somehow introduce the particle into the

box at time t = 0, it is unlikely that we will succeed in putting it into a
stationary state. By superposing the stationary-state solutions (3.6),
c
qrs®qrs{rj), (3.7)
q,r,s=l
where the cqrs are constants, we obtain more general wave functions \I>,
which can be verified to satisfy the time-dependent Schrodinger equation
(2.8) and also have built into them the correct boundary condition that
they vanish on the faces of the box. Imagine we know that we have
introduced the particle into the box in such a way that its wave function
The one-dimensional square well 21
(a) (b)
Figure 3.1. (a) The one-dimensional square-well potential. (6) The Coulomb potential
produced by a charged atomic nucleus.
at t — 0 is *(r,0) = f(r). (In practice, it would be very hard to determine

/(r).) Then
x 1/2 oo
. qirx . my .
= (— 1 V sin sin —-— sin
\abcj ^
This is just a triple Fourier series for / ( r ) , which can be inverted by stan-
dard methods to give the constants cqrs. If f(r) is known, the complete
wave function (3.7), for all t > 0, is thus determined in the form of an
infinite series.
The one-dimensional square well

We now consider a one-dimensional problem, where both the potential
and the wave function are functions of one of the three coordinates only,
say x. We investigate the bound states corresponding to the square-well
potential
0 \x\ > ±6, (3 g)
V =
-U <h
with U positive. This potential represents a very crude one-dimensional
model of the potential experienced by an electron in an atom, due to the
nucleus (see figure 3.1).
With the potential (3.8), the time-independent Schrodinger equation (2.7)
becomes
{-h2/2m)(d2ip/dx2) = Ex/} \x\
(-h2/2m)(d2ip/dx2) -Uip = \b. (3.9)
The solution to the first equation is a linear combination of elkx and e~lkx,
where E = h2k2/2m. However, for a bound state the particle should be
more or less confined to values of x in the vicinity of the potential, and so
we should impose the boundary condition \ij)\ -> 0 as x -+ ±oo. Hence we
reject the case E > 0, since when k is real the solutions do not satisfy this
requirement. To satisfy the boundary condition, we must have E < 0. If
we write, with a and (3 > 0,
E = -h2a2/2m, E + U = h2p2/2m, (3.10)
the required solutions to the first equation in (3.9) are
where A and B are constants. (We have rejected a term eax for x > | 6 ,
and a term e~aa; for x < —16, so as to make |^| —> 0 as x —> ±oo). The
solution to the second equation in (3.9) is
i/>(x) = Csinfix + Dcospx, (3.116)
where C and D are constants.

We now impose the boundary conditions that i/> and d^/dx are continuous
at x — ±§&.'
Ae~*ab = Csin \(3b + Dcos |/36,

= Cficos \$b - Dp sin |/36,
= Cpcos |/36 |
By eliminating A, B, C, J? from these equations, we obtain a relation

between P and a. We can manipulate the equations as they stand, but
the algebra is somewhat reduced if we first realise that the possible wave
functions for the system fall into two classes (see problem 3.7):
(a) Even-parity wave functions, which are symmetric about the origin,
ip(x) = ip{-x). For these, from (3.11), A = B and C = 0, and we
find that
a = /?tan|/?6. (3.12a)
The linear harmonic oscillator 23
(6) Odd-parity wave functions, which are antisymmetric about the origin,
il>(x) = ~^{-x). For these, A = -B and D = 0, and
a = -0 cot \(3b. (3.126)
The reason for the existence of these two classes of wave function is that
the potential is symmetric about x = 0.
If we combine either (3.12a) or (3.126) with (3.10), we obtain a discrete
set of allowed values for the energy E. Each of these lies in the range
0 > E > - [ / ; that is, the bound states correspond to energies lying
within the well. The equations (3.10) and (3.12) can only be solved by
numerical methods, but we can get some information about the solutions
by graphical methods. If we eliminate a between (3.10) and (3.12), we
find that either
tan 2 |/?6 = — 1 + 2mU/Ti201 (for even-parity solutions),

or
cot |/36 = - 1 + 2mU/h2/32
2
(for odd-parity solutions). (3.13)
We have sketched the two sides of these equations, against /?, in figure 3.2.
In each case the two curves intersect in just a finite number of pointsf,
so resulting in a finite set of allowed values for E. The number of such
allowed values depends on the magnitudes of U and 6. One can show
that if U < 0, there are no bound-state solutions. This is not unexpected:
the potential is then repulsive instead of attractive, and so has no bound
states.
The linear harmonic oscillator

We now consider another one-dimensional problem, for which the potential
is
V(x) = \mu2x2. (3.14)
In classical dynamics, this would correspond to an attractive force whose
magnitude is proportional to the distance from the origin, with the simple
harmonic solution x = smut or cosut. In classical physics most oscillat-
ing systems are simple harmonic, at least if the amplitude of oscillation is
sufficiently small. For this reason the simple harmonic oscillator plays a
f Because we have squared equations (3.12) we have to take care to accept only those
points of intersection for which tan ^fib or — cot \fib is positive.
Figure 3.2. Solution to (3.13).
central role in classical physics, and the same is true in quantum physics.
So the mathematics that follows, with suitable adaptations, applies also,
for example, to the quantisation of the electromagnetic field, since accord-
ing to classical electrodynamics the electromagnetic field in a light wave
oscillates harmonically at each point of space. Thus, just as we shall find
that in quantum mechanics the energy of a simple linear oscillator with the
potential (3.14) can only take certain discrete values, so also the energy
in an electromagnetic field appears in discrete quanta. These quanta are
what we know as photons.
The time-independent Schrodinger equation for the potential (3.14) is
{-h2/2m){d2^/dx2) + (\mu2x2 - = 0. (3.15)
To this is applied the boundary condition that ip —> 0 as x —> ±oo, as in

the previous problem. This has the consequence that E must take one of
the values
En = (n+ \)hu, n = 0,1,2,.... (3.16)
For other values of E the equation has no solutions obeying the boundary
condition.
The tunnel effect 25
We give an outline proof of this in appendix A. We show there that the

corresponding wave functions are
tl>n(x)=e-m»x2'MHn(x), (3.17)
where Hn(x) is a polynomial of degree n in x (it is known as the Hermite

polynomial). If n is even, Hn(x) contains only even powers of a;, so that
i[;n(—x) = ipn(x); if n is odd, Hn(x) contains only odd powers, so that
<0n(-x) = —ij)n^x). As in the case of the square well, the potential is
symmetric about x = 0, and so we again find that the solutions have
definite parity.
Notice that, according to (3.16), the lowest possible energy of the oscillator
is EQ = \Tiu. This is known as zero-point energy. In classical mechanics
the lowest possible energy is, of course, zero. It corresponds to the particle
being at rest at the origin, but in quantum mechanics this is not allowed
by the uncertainty principle, which we explain in the next chapter.
The time-dependent wave function corresponding to (3.17) is
In general, if the oscillator is somehow set in motion, it will not be in a

stationary state. We may write general solutions to the time-dependent
Schrodinger equation, satisfying the appropriate boundary conditions, by
superposing stationary-state solutions:
x,t), (3.18)
n=0
where the constants cn may take any values. For arbitrary choices of
these constants, the probability density at each point x oscillates in time
with period 2TT/U;, though the oscillation is not simple harmonic. Notice
that although each \I/n is a stationary-state wave function and so satisfies
the time-independent Schrodinger equation, this is not true of the wave
function \I> that we get by superposing them on each other. This wave
function satisfies only the time-dependent Schrodinger equation; it does
not correspond to a stationary state.
The tunnel effect

An important phenomenon of quantum mechanics, which is not allowed
by the laws of classical physics and which is the basis of some of the ap-
plications in the later chapters of this book, is the tunnel effect. Consider,
for simplicity, a one-dimensional potential, similar to the square well but
opposite in sign:
T/ _ // 0 x < 0 and x > a,

V
~\U 0<z<a,
with U positive. In classical mechanics a particle approaching the origin

from x < 0 would see this as a barrier which it could surmount only if
its energy were greater than U. However, in quantum mechanics a beam
of particles will be partially transmitted through this barrier even if the
energy of each particle is less than [/; that is, some of the particles pass
on to x = +oo, while others are reflected back towards x = — oo. In fact,
this happens also when E > U.
Suppose that the incoming beam of particles has energy E and has the
time-independent wave function
&n = Aeikx, (3.20)
where, as we discussed in chapter 2, the constant A depends on the density

of particles in the beam. For x < 0 the Schrodinger equation is
(-h2/2m)(d2^/dx2) = Elf), (3.21)
so that if V>in is to satisfy it, h2k2/2m = E. In the region of the barrier,

0 < x < a, the Schrodinger equation is
(-h2/2m)(d2ilj/dx2) + U^ = E%l>, (3.22)
so that there the wave function is
BeKX + Ce~"x (3.23)
where h2K2/2m = U — E, so that if E < [/, n is real.

There are two boundary conditions at each edge of the barrier: ip and
dil)/dx are continuous at x — 0 and at x = a. Applying these boundary
conditions gives us four equations, but so far we only have two unknowns,
B and C, k and K being determined once we have chosen E. In order to
The tunnel effect 27
be able to satisfy the boundary conditions we must introduce two more

unknowns, so we need a transmitted wave in x > a,
, (3.24)
and a reflected wave in x < 0,
where T and R are constants. The complete wave function is then (3.24)
in x > a, (3.23) in 0 < x < a and
*P(x) = Aeikx + Re~ikx (3.25)
in x < 0.
Applying the boundary conditions at x = 0 and x — a now gives us
A + R = 5 + C,
a
= «(Be Ka - Ce~Ka). (3.26)
We may solve these equations, to determine i?, T, J5 and C in terms of

the incoming amplitude A. In particular, we find that
T = ^hA ( 3 27)
eika[e«a(k + IK)2 + e~â(n + ik)2)' K }
so that there is indeed a non-zero transmitted wave. Notice that there

is a solution for any positive value of E; this is to be contrasted with
bound-state problems, where only certain discrete values of E turn out to
be allowed.
If one solves also for i?, one finds that
|i?|2 + \T\2 = \A\2. (3.28)
This equation is the condition that every particle is either reflected or

transmitted, and none is lost. To see this, we use the particle flux (2.18);
in this one-dimensional problem it is
In a stationary state, where the particle density does not vary with time,
this must be independent of rr, so that no particles disappear. (This
particular simplicity of the consequences of the particle-conservation re-
quirement is peculiar to one-dimensional problems; see problem 2.4.) In
x < 0, (3.25) gives
j = (hk/m)(\A\2-\R\2), (3.30a)
and in x > a (3.24) gives
j = (hk/m)\T\2. (3.306)
Putting (3.30a) and (3.306) equal to each other, we retrieve (3.28). Notice
that, according to (3.24), in x > a the particle density is \ip\2 = \T\2.
The factor hk/m in (3.306) is the speed of the particles (see the relation
between wave number and momentum in (2.2)), so that (3.306) says that,
as we expect, the particle flux is equal to the particle density times the
speed. The form (3.30a), applicable in x < 0, is more complicated because
some of the particles are moving in one direction, the rest in the other. In
a more general situation, where the particles do not have a single speed,
the form of j does not have such a simple interpretation.
The delta-function potential

Our discussion of the tunnel effect, and of bound states in one dimen-
sion, was in terms of the simplest possible potential, the square barrier or
square well. These form prototypes for repulsive or attractive potentials
in one dimension. In many respects the solutions for other, more realistic,
potentials in one, or indeed in three, dimensions have similar proper-
ties. However, the solutions for even these simple potentials involve a fair
amount of algebra, so it is often simpler to consider a mathematical limit
that is easier to work with.
Consider the square-barrier potential (3.19) and let a -> 0, with at the
same time U -> oo, in such a way that all remains fixed and equal to A,
say. The resulting potential is just
V(x) = XS(x). (3.31)
(For a description of the delta function, see appendix B.) To find the sta-
tionary states, we have to solve the time-independent Schrodinger equation
(3.32)
The WKB approximation 29
This equation is to be valid for all x. Prom it, we conclude that i/;(x)
is continuous even at x = 0. For suppose that ij)(x) were discontinuous
at x = 0. Then V'O^) would contain a delta-function part (or a worse
singularity if the discontinuity of i\) is infinite) and so ip" would contain the
derivative of S(x) (or worse). This cannot be, because such a singularity
in \j)" is not balanced by a similar singularity in any other term of the
Schrodinger equation (3.32). However, ip'(x) must be discontinuous at
x = 0, so that ij)"{x) contains a delta-function singularity that balances
that of the second term of the Schrodinger equation.
Take the wave function in x < 0 to be an incoming plane wave together
with a reflected wave, just as in (3.25):
%l>(x) = Aeikx + Re~ikx. (3.33a)
Let the transmitted wave be as in (3.24):
i/j(x)=Teikx x>0. (3.336)
In order that the wave function be continuous at x — 0,
A + R = T. (3.34)
f
To obtain the correct discontinuity of ip (x) at x = 0, integrate the
Schrodinger equation (3.32) from x — — e to x = +e and then let e —> 0.
The various terms of the Schrodinger equation give
A /
J -£-e
p£
E I dxip(x) - 2eEi/>(0) -> 0,

J —£
where disc ip denotes the discontinuity of tpf at x = 0, that is, the value of
1
ip'(x) at x just greater than zero minus its value at x just less than zero.
Hence
(-h2/2m) disc V' + A^(0) = 0 (3.35)
and, from the wave function (3.33),
{-n2/2m)[ikT - ik(A - R)] + XT = 0.
With (3.34), we thus find that
R = m\A/(ih2k - m\),
T = ih2kA/{ih2k - mX). (3.36)
It is straightforward to check that the particle-conservation condition
(3.28) is satisfied.
The WKB approximation

We have seen how to calculate the tunnelling probability for a square
barrier. How can we adapt this analysis to irregularly-shaped barriers?
In general, we will not be able to find an exact solution and will have to
rely on approximate methods.
Consider first the transition amplitude (3.27) for a particle of energy E
incident on a square-barrier potential of height U. Suppose that the width
a of the barrier is so small that na » 1, where as before
K2 = 2m{U-E)/h2. (3.37)
Then the intensity transmission coefficient \T/A\2 (the ratio of the trans-
mited intensity to the incident intensity) is approximately
(Jf^J>2"a (3.38a)
and, if Ka is large enough,
ln|T/A| 2 « -2na (3.386)
Because our approximations have led to a result for ln|T/A| 2 that is lin-
ear in Ka, if we follow the first potential barrier by another one for which
the corresponding parameters are K! and a', then ln|T/A| 2 will become
modified by the addition to it of —2n'a'. A smooth, arbitrarily shaped
potential barrier V(x) can be treated as a series of adjacent rectangular
barriers, each of infinitesimal width Ax, as shown in figure 3.3. In the re-
gion xi < x < x2 we may use (3.37) and (3.38b) to write the contribution
to ln|T/^4|2 from one of these infinitesimal barriers:
- 2 [2m(V(x) - E)/n2] 2 Ax (3.39a)
and the transmission coefficient for the whole barrier is approximately
dx[2m(V(x)-E)/h2]^\ (3.396)
This is the WKB formula, which was derived more rigorously by Wentzel,
Kramers and Brillouin.
Alpha decay 31
V(x)
\
\
\ \
Figure 3.3. A potential barrier of any shape approximated by a series of rectangular

barriers.
Alpha decay
We use the WKB approximation to derive a very simple model of the
a-decay of a radioactive nucleus. In this decay, the nucleus emits an a-
particle, which is composed of two neutrons and two protons, so that its
mass M is about 4 times that of the proton and its charge is 2e. The a-
particle is held within the nucleus by a strong attractive force, the nuclear
force. This force has finite range, so that if the a-particle escapes from
it by tunnelling, it is subjected outside the nucleus only to the weaker
Coulomb force.
Consider first the wave function of an a-particle bound within the nucleus
by the nuclear force; we include the effect of the Coulomb force later.
Model the potential corresponding to the nuclear force by a square well,
and simplify further by pretending that the potential is one-dimensional
and that the a-particle moves only in the ^-direction. So the nuclear-force
potential is
' -U 0 < x < i?,
V(x) = (3.40a)
oo otherwise.
A stationary-state wave function corresponding to positive energy E that
satisfies the Schrodinger equation for this potential in the well 0 < x < R
and vanishes at x = 0 and x = R, as is required when the potential is
infinite outside the well, is
i>{x) = A(eiKX - e~iKX) = 2L4 sin KX (3.41a)
with
E + U = h2K2/2M such that sin«i? = (3.416)
V(x)
E
0
-U
Figure 3.4. One-dimensional model potential for a-decay.
We now include the Coulomb force, so that for x > R the potential be-
comes
2Ze2
V(x) = (3.406)
where Ze is the charge of the 'daughter' nucleus that remains after the
a-particle has escaped. We leave the potential unaltered for x < i?, so
that it is as shown in figure 3.4.
With classical mechanics, the a-particle cannot escape if its energy E is
less than Vb, the maximum height V(R) of the barrier, but in quantum
mechanics it can tunnel through the barrier. Because the potential is no
longer infinite for x > i?, we no longer require the wave function to vanish
at x — R, and so the wave function (3.41) will become modified. The
coefficients of e1KX and e~1KX will no longer have equal magnitude, and
they will even become time-dependent so that we should use the time-
dependent Schrodinger equation. However, for energies E much less than
Vb the tunnelling probability is small and the wave function in 0 < x < R
is changed only a little by the addition of the Coulomb tail to the potential,
so it is a reasonable approximation to keep the form (3.41); that is, we
can assume that A varies very slowly with t. For very large positive x the
potential is extremely small, so that the wave function there is to a good
approximation Telkx with h2k2/2M = E, as in (3.24). The current j then
Alpha decay 33
is (hk/M)\T\2, as in (3.30b). The WKB approximation (3.39b) gives
where r = 2Ze2/4neoE is the distance at which the Coulomb potential is

equal to E. The integration can easily be done by making the change of
variable x — r 1 / 2 sin#. If the a-particle energy E is small compared with
the height of the barrier, so that r > R, the integral is approximately
^Trr1/2. So we find that the current at large x is
If we normalise the wave function suitably, the probability that the a-

particle is within the nucleus is
rR dx\iJ>(x)\ 2
= 2S\A\2R, (3.43)
=s ,
Jo
where S is the cross-sectional area of the nucleus in our one-dimensional
model. As we have said, A varies very slowly with t\ in fact P obeys the
continuity condition (2.17):
where here j is the current at the surface of the nucleus, that is at x = R

(because the potential is taken to be infinite for x < 0 there is no current
at x — 0). As we have explained, the current at x = R is equal to that
at very large x, because all the particles that leave the nucleus escape to
infinity. So from (3.42), (3.43) and (3.44) we have
dP P
-* - "7 <
with
/M\2 i 7,P2 / M\W
(3.456)
Integration of (3.45a) gives P = Poe~t/T, where Po is the value of t at

time t — 0. An exponential decay law such as this is typical for unstable
systems in quantum mechanics. After time r approximately half of the
a-particles have escaped from the nucleus; we call r the half-life for an
a-particle to remain in the nucleus once it has been formed. To obtain
the half-life of the nucleus itself, we must divide by the probability that
two neutrons and two protons within the nucleus come together to form
an a-particle.
Although this model is quite crude, it well describes the observed strong
dependence of the half-life of a-particle emitters on the a-particle energy.
Problems
3.1 Sketch the wave functions for the one-dimensional square-well potential
(3.8).
It is found that the average distance of the particle from the centre of
the well can be rather greater than the width of the well. How is this
possible, in terms of your sketches?
3.2 A particle of mass m is bound by a one-dimensional potential well of
depth U and width b. Show that the condition that there is just one
bound state is that
3.3 A one-dimensional harmonic oscillator carries charge e and is placed

in a uniform electric field £, so that the potential becomes
V(x) = \mw2x2 — e£x.
Show that each energy level is reduced by e2£2 /2mu2. What are the
new wave functions? [Hint: replace x by a new variable.]
3.4 By consulting appendix A, calculate the n = 0 and n = 2 wave func-
tions of the linear harmonic oscillator (leave them unnormalised).
Verify that they are orthogonal, that is,
£ dxipo(x)ijj2{x) = 0.
3.5 A beam of particles is incident on (i) a very deep square-well potential,

(ii) a narrow barrier (ka < 1) whose height is equal to twice the kinetic
Alpha decay 35
energy of each particle. Show that in case (i) the particles are nearly
all reflected, while in case (ii) they are nearly all transmitted.
Notice how different these results are from those which obtain in the
corresponding situations in classical mechanics.
3.6 Find the energy level of the bound state for the potential V(x) =
—fiS(x). Verify that this agrees with that for the square-well potential
(3.8) in the limit U -» oo, with Ub fixed and equal to JJL.
3.7 A potential V(x) is such that V(x) = V(—x) and each bound-state
energy level corresponds to only one independent wave function. Show
that each of these wave functions has either even parity or odd parity.
[Hint: the first step in the proof is to show that if ip(x) is a solution of
the Schrodinger equation, so also is ij){—#), with the same value of E.]
3.8 The nuclei IQQFm and 928U have approximately the same radius and
both are a-particle emitters. The a-particle energies are 7.4 MeV and
4.2 MeV, respectively. The half-life of ôFm is \ hour; assuming that
the probability of the formation of an a-particle is the same, what is
the half-life of jf
4
The superposition principle
Linear operators
In the examples of chapter 3, we took linear combinations of solutions to
the Schrodinger equation and so constructed further solutions: see (3.7)
and (3.18). The reason that this works is that ihd/dt and the Hamiltonian
operator if, which appear in the Schrodinger equation, are linear, and,
furthermore, the boundary and continuity conditions on the wave function
^ are linear.
By linear boundary or continuity conditions, we mean that if ^ i and \I/2
are two functions that satisfy them, then so does the function (ciî +
c2\I/2), where C\ and c2 are any complex numbers. This is evidently the
case for wave functions, which we normally require to vanish outside some
region of space, or at infinity.
By a linear operator Q, we mean that
(4.1)
An example of an operator that is not linear is the operator S that squares

the wave function:
In chapter 2 we said that one of the basic assumptions of quantum me-

chanics is that observables correspond to operators. This assumption is
usually supplemented by the further requirement that the operators be
linear, and in this chapter we explore some of the consequences of this.
Notice that the operators we have met so far, — iftV corresponding to
momentum and H corresponding to energy, are linear.
The consequence of linearity is that if we superpose wave functions that
describe possible states of a system, we get another wave function that
describes a possible state. It is often useful, as in our examples of (3.7)
and (3.18), to superpose stationary-state solutions. A basic mathematical
property that can be proved is that these stationary-state wave functions
36
Wave packets 37
are complete. This means that any allowable # ( r , i ) , that is, any wave
function that satisfies the time-dependent Schrodinger equation and the
appropriate boundary conditions for the system, can be written as a sum
of stationary-state wave functions
Wave packets
As an example, consider a free particle moving in the ^-direction. The
stationary states are the plane waves
corresponding to a definite value p of the momentum and a definite value

p2/2m of the energy. N(p) is a constant. As the squared modulus of this
wave function is independent of re, there is equal probability of finding the
particle at each point x. To construct a wave function corresponding to
a particle that is localised in space, we must take a superposition of the
stationary states (4.3). These are labelled by the momentum p, which is
here a continuous variable. This is because, as we explained in chapter
2, the plane wave (4.3) describes a mathematical limit where the volume
Q of the system has been allowed to become infinite. For finite f2 the
superposition (4.2) would be a sum over discrete values of p (see (3.7)).
As Q, —> oo this sum becomes an integral:
- [dpN{p)eipx/h-ip2t/2mh. (4.4)
Here N may take a different value for each value of p: it is now a function
of p. The resulting wave function is not a stationary state; if we wish
to localise the particle in space, it does not have definite momentum or
energy. Rather, (4.4) expresses the wave function as a superposition of
components having different values of momentum p and energy p2/2m.
Setting t = 0, we have
= [dpN(p)eipx'h. (4.5)
38 4 The superposition principle
That is, N(p) is the Fourier transform of $(x,0) (see appendix B). By
inverting the transform, we may express N(p) in terms of \I>(:r,0):
N(p) = ^jdx^{x^)e-'^lh. (4.6)
Hence if we know the wave function \fr at t = 0, we can calculate N(p).

Inserting the result into (4.4), we can then find the wave function # at
other times t.
As a simple example, suppose that we take a Gaussian:
*(ar, 0) = No e-(*-*o)2/a2+iPox/hm (4.7)
Here NQ is a normalisation constant, chosen so that Jda;|\I>(a;,0)|2 = 1,

which requires that
N0 = (ia a *)- 1 / 4 -
The wave function (4.7) is called a wave packet: the probability density
|^(x,0)| 2 is localised around the point x = x$. If we prepare the state a
large number of times and each time measure the position of the particle
at t = 0, because |^| 2 is the probability density the average value will be
(x) = Jdxx\V(x,0)\2=x0. (4.8)
The results of the separate measurements will be scattered around this

average value. A measure of the width of their distribution is the quantity
Ax, defined by
(Ax)2= [dx(x-(x))2\y(x,0)\2. (4.9)
Ax is called the width of the wave packet. With (4.8) we find that, at
t = 0, Ax = \a.
If we insert (4.7) into (4.6), we obtain an integral whose evaluation is
discussed in appendix B. It gives
2 2 2
N(p) =
EhrenfesVs theorem 39
This has the same functional form as #(£,0), a property that is peculiar
to the Gaussian. One can check that J dp\(/)(p)\2 = 1, where
d>(p) = [l/(2nh)1/2]N(p).
The interpretation is that, just as |\I>|2 is the probability density in position

space, so \(f)\2 is the probability density in momentum space.
That is, if we measure the momentum, the probability of obtaining a
result in the interval (p, p + dp) is just \(j){p)\2dp. If we prepare the state
many times and measure the momentum each time, the average value so
obtained will be
(p) = Jdpp\ct)(p)\2=po, (4.11)
and the distribution of values obtained will have a width Ap defined by
(Ap)2 = Jdp(p-(p))2\ct>(p)\2. (4.12)
With (4.7), we find that Ap = h/a. Hence at t = 0
AxAp= \U.
There is a general principle, the Heisenberg uncertainty principle, which

says that for any wave function, Ax and Ap, defined by (4.9) and (4.12),
satisfy
AxAp> \Tt. (4.13)
Ax and Ap are the uncertainties in the position and momentum associated

with the wave packet. This means that we cannot know the position and
momentum together to better than to a certain degree of accuracy. The
allowed degree of accuracy depends on the shape of the wave packet; in this
sense a Gaussian is the best possible shape in that it gives Ax Ap = \h
rather than greater than \7i. Notice that in the case of a plane wave, the
momentum takes a definite value, so Ap = 0. However, since |\I>|2 has the
same value for all x, in this case the integral (4.9) diverges and Ax = 00.
Ehrenfest's theorem
Having obtained N(p) in (4.10), we may now use (4.4) to calculate *(x, t)
for t / 0 (see problem 4.2). The result is that ^(x,t) remains Gaussian
in shape at subsequent times t. The average position (x) of the particle
moves with velocity po/m, and the width Aa: of the wave packet increases
with increasing time. The reason that the wave packet spreads out as t
increases is that the different momentum components of the wave function
correspond to different velocities p/m.
There is a theorem, due to Ehrenfest, that states that whatever the shape
of the wave packet, and for a general potential V,
(d/dt)(r) = (p)/m (4.14a)
and
(d/dt)(p) = <-W>, (4.146)
where
= /*d3rW|«f
(see problem 4.4). For a sufficiently narrow wave packet, ( W ) is almost

equal to the value of W at r = (r). Then the two equations (4.14)
show that the wave packet as a whole moves like a classical particle. This
does not mean that it describes a classical situation because, as we have
seen, the momentum and position of the particle cannot be simultane-
ously known with absolute certainty. Crudely speaking, the quantum-
mechanical description is very close to the classical description if the
widths Ar, Ap of the wave packet in position and momentum are small.
This statement is not precise because we have not specified a length L
and a momentum P with which we must compare Ar and Ap in order to
decide whether they are indeed small. L and P will be parameters some-
how typical of the system under investigation. Because we know from the
uncertainty principle that ArAp > | / i , it is necessary that
if a classical description is to be a good approximation.
Hermitian operators 41
Hermitian operators
We have said that it is assumed in quantum mechanics that to each ob-
servable Q there corresponds a linear operator Q. We now introduce an
additional assumption about this operator: that it is Hermitian.
Let ^ i and ^2 be any two of the allowed wave functions for the system
under discussion. We write
(4.15)
In this definition the integration is, as usual, over all space. The operator
Q is said to be Hermitian if, for all choices of \&i and #25
*, (4.16a)
that is, if
fd3r^l(Q^2) = /d3r(Q#i)*#2. (4.166)
As an example, the Hamiltonian operator H is Hermitian. To show this,

we use the fact that V = V*, since the potential V is real, so that
Just as in (2.15), we may transform this to an integral over the surface

enclosing the volume of integration and conclude that for a closed system,
where the wave functions vanish outside some volume £2o5 the integral van-
ishes. As we explained in chapter 2, an unbounded system, corresponding
to the mathematical limit fi0 ->• 00, is a little more subtle, and we shall
not discuss it.
An important property of a Hermitian operator is that its eigenvalues are
real, as we now show. For if Qty = g\I>, where q is one of the eigenvalues,
then
(\I/, Qty) = q{*&, \fr), (4-17)
where
42 Jf. The superposition principle
Take the complex conjugate of (4.17) and use (4.16a) for the case \&i =
Hence, since (\P, \P) > 0, q = q*. Because, as we explained in chapter 2,

the eigenvalues of the operator Q corresponding to an observable Q are
possible values obtained in measurements of Q, it is natural to impose
the restriction that Q be Hermitian so as to ensure that these values are
real. We shall argue below that whenever we measure <2, the result is one
of the eigenvalues of Q.
Another important property of a Hermitian operator is that two eigen-
functions \&i and ^ corresponding to different eigenvalues qi and #2,
are orthogonal. By this we mean that
To show this, since by definition Qî = qiî and Q\&2 = #2^2, we write
the chain of equations
q2 (
Here we have used (4.16a) and the fact that qi is real. So if q\ ^ q2->
we have the required result. If qi = q2, we cannot conclude that \I>i
and \&2 are orthogonal. However, in this case any linear combination
(ciî +02^2) is also an eigenfunction of Q, with the same eigenvalue, and
we can choose two linear combinations that are orthogonal (see problem
4.7). If we normalise each eigenfunction, so that it satisfies ( ^ , ^ ) = 1,
we then have an orthonormal set:
(4.18)
where 8rs = 1 if r = s and 8rs = 0 if r ^ s.

A basic mathematical assumption, which is not altogether straightforward
to check in practice, is that the set is complete. By this we mean that any
allowable wave function \I> of the system can be expanded as a linear
combination of the eigenfunctions \&r of Q:
#r. (4.19)
Operators and observables 43
If we want \I> to be correctly normalised, so that (#, *) = 1, the coefficients

cr will satisfy
X>| 2 = 1. (4-20)
To verify this, use the orthonormality condition (4.18). From this condi-
tion we find also that
(*.,*) = c , . (4.21)
Operators and observables

We explained in chapter 2 that it is assumed that when the value of an
observable Q is known to be g, the system is in a state whose wave function
is an eigenfunction of the corresponding operator Q, with eigenvalue q. We
say that the system is then in an eigenstate of Q. This assumption implies
that any measurement of Q will leave the system in an eigenstate of Q.
This follows if we assume that quantum physics is not so capricious that,
if with suitable care we make the same measurement of Q twice, in rapid
succession, we expect to get different answers. The two measurements
must be close enough in time so that we can be sure that no outside
influence has disturbed the system, and that the natural time development
of the system has not resulted in a change in the value of Q. After the first
measurement, we know what will be the result of the second measurement
and so, according to the assumption, the first measurement must have left
the system in an eigenstate of Q and the result of the measurement is the
corresponding eigenvalue q.
In general, the system will not be in an eigenstate of Q before the first
measurement. So the first measurement must have disturbed the system,
so as to throw it into an eigenstate of Q.
Suppose that, before any measurement of Q is made, the system is in a
state whose wave function is \I>. The result of measuring Q is one of the
eigenvalues of Q but, unless ^ happens to be an eigenstate of Q, we cannot
predict which eigenvalue it will be, that is, which eigenstate the system will
jump into because of our act of measurement. However, we can calculate
the probability of obtaining a particular eigenvalue. One would guess that
this probability is related to the coefficients cr in the expansion (4.19)
of \I> in terms of the eigenfunctions \I>r. The most natural assumption is
that the probability Pr of throwing the system into the eigenstate * r by
making the measurement, and so obtaining the eigenvalue qr as the value
of Q, is
P r = |c r | 2 = | ( * r , * ) | 2 . (4.22)
Then, because of the normalisation condition (4.20),
(4-23)
That is, the probability of getting some result for the measurement is 1.
Let us now recapitulate our basic assumptions about observables:
(a) To each observable Q there corresponds a linear, Hermitian operator

Q.
(b) If the value of Q is known to be g, the system is in an eigenstate of Q,
with eigenvalue q.
(c) The result of measuring Q can be any of the eigenvalues of Q. When
the system is in the state \I/, the probability that a measurement of Q
will give the eigenvalue qr is Pr — |(\t r , \£)| 2 .
If we somehow prepare the same state ^ many times, and on each occasion
measure Q at the same relative instant in time, the average answer will be
If we use the definition (4.19) of the c r , and the orthonormality condition

(4.18), we find that
J] < (4.24)
This is called the expectation value of Q in the state \I/. It is often written
(Q). Earlier in this chapter we have already encountered two particular
examples, the expectation values of position and momentum:
(r)= /d3r**(r#)= /d3rr|*|2,
(p) = /d 3 r#*(p#) = f d3ry*(-ihVV). (4.25)
The operator f, corresponding to position, simply multiplies any wave

function on which it acts by r . It may be verified that the expression
(4.25) for (p) is equivalent to the definition (4.11) in terms of (f)(p).
Commutators 45
Commutators
Suppose that we measure an observable Q and then another observable
R. The first measurement will throw the system into an eigenstate of
Q. In general, this will not be an eigenstate of R, so that the second
measurement will disturb the system again. Thus, by measuring i?, we
lose the information we have about Q.
This is another manifestation of the uncertainty principle, which we have
already encountered for the particular case of the observables r and p.
In general, one cannot know the exact values of two observables Q and R
simultaneously, because generally their eigenstates are not the same.
However, some pairs of observables may simultaneously be measured ex-
actly. It can be shown that two operators Q and R have a complete set of
common eigenstates if, and only if, their commutator
[Q,R] = QR-RQ (4.26)
vanishes. That is, if we operate on any wave function first with Q and
then with R, the result is the same as when we apply the operators in the
reverse order. We then say that the operators Q and R commute. So the
condition that we may simultaneously know the values of two observables
is that the corresponding operators commute.
The generalisation of the uncertainty principle (4.13) for Ax Apx to an
arbitrary pair of observables can be shown to be (see problem 4.15)
(4.27)
On the right of this inequality there appears the modulus of the expectation
value of the commutator of the corresponding operators, in the state for
which the measurement is being made. The commutator is, of course,
itself an operator. For the particular case of components of position and
momentum, it turns out that the operator is a simple number:
[xi.pj] = ih6ij (4.28)
where i and j label the different components, and <Jy = 1 if i — j and 0 if

i 7^ j . To check this, introduce an arbitrary wave function \I> and let the
commutator operate on it:
Thus the two sides of the relation (4.28) are equal when applied to any
wave function #, which is the same thing as saying that (4.28) is a true
identity. Because the commutator (4.28) vanishes when i ^ j , we can
measure simultaneously the exact values of a given component of r and a
different component of p. When we take the same components of r and p,
the commutator (4.28) is ift, and so the general uncertainty relation (4.27)
then agrees with (4.13): we cannot measure the exact values together.
The theory of measurement in quantum mechanics is a deep, and even
controversial, subject; our discussion has necessarily been brief.
Problems
4.1 Calculate (p) and Ap for the Gaussian wave packet (4.7) at t = 0 by
taking expectation values of the appropriate operators.
4.2 A Gaussian wave packet is given at t = 0 by (4.7). Calculate 3/(z,£)
from (4.10) and (4.4) far enough to verify that
(x) — x0 + pot/m,
(Ax) = \a2 + h2t2/m2a2.
2
4.3 Show that, for any one-dimensional wave function, the values of (p) and
(p2) calculated from the momentum-space probability density |^(p)|2
are equal to the values obtained from the expectation values of the cor-
responding operators using the coordinate-space wave function tp(x).
4.4 Prove Ehrenfest's theorem, equations (4.14), for the one-dimensional
case where the wave function depends on x and t only.
4.5 Show that both (x) and (p) are zero for a one-dimensional harmonic os-
cillator and use the uncertainty principle to deduce that the minimum
energy must be greater than, or equal to, \hw.
4.6 A dart of mass 1 kg is dropped from a height of 1 m, the intention being
to hit a given point on the ground below. The uncertainty principle
imposes a limitation on the accuracy that can be achieved; show that
this is only of the order of a tenth of a nuclear diameter. [Neglect
the uncertainties in position and momentum in the vertical direction;
these produce second-order effects on the accuracy.]
4.7 The wave functions ipi(x) and -02(x) are normalised, and satisfy
(/0ij'02) = c. Find a normalised linear combination of them that is
orthogonal to ipi.
4.8 Show that the momentum operator — i W is Hermitian for a bounded
system.
Commutators 47
4.9 Use the divergence theorem (page 15 footnote) to show that in a

bounded system the expectation value of the one-particle kinetic-
energy operator is positive.
By considering the expectation value of the Hamiltonian, deduce that
for a particle moving in a potential well of arbitrary shape the lowest
bound state has an energy level that is always higher than the bottom
of the well.
4.10 Q is a Hermitian operator and the wave function * is such that
Qfy = q\&^ where q is a number. $ is any other wave function. Show
that
4.11 Show that, for any linear operators A,B and (7,
[A, [B, C]] + [S, [C, A]] + [C, [i, B]] = 0.

(The second relation is called the Jacobi identity.)
4.12 A quantum system has a complete orthonormal set of energy eigen-
states V>n, with different eigenvalues En. The operator Q corresponds
to an observable and is such that
Wi=fe Q^2 = û Q*Pn=0 forn>3.

Find a complete orthonormal set of eigenfunctions of Q.
The observable is measured and is found to have the value +1. The
system is left undisturbed and after a time t the observable is measured
again. Calculate the probability that the value +1 is found again.
4.13 The operator Q does not depend explicitly on t. Show that for any
solution \& of the time-dependent Schrodinger equation
4.14 Show that for a free particle
_d 2 _ (pxx + xpx) d2 2 _ 2(p|)
4.15 We define the Hermitian conjugate FT of an operator F such that,

for any pair of allowed wave functions of the quantum system under
discussion, ( * i , F t * 2 ) = (*2,F#i)*. Show that: i
(i) if F is Hermitian, then FT = F,
(ii) (F + iG)t =Ft-i(?t,
(iii) (FG)t = GtFt,
(iv) if $ = F * then ($, *) = (
The operators Q and .R correspond to observables and A is any real
number. Show that the expectation value of
(Q + i\R){Q-i\R)
in any state * is real and > 0. Hence prove that
and so deduce the general uncertainty principle (4.27).
5
The hydrogen atom
Good quantum numbers

In general, if we measure an observable Q at two different times, we shall
obtain different eigenvalues of Q. This is because the system will have
undergone a natural time development, governed by the time-dependent
Schrodinger equation. (In addition, the system may have been disturbed
from outside, for example by a measurement of some other observable.)
The only states whose natural time development is trivial are the station-
ary states (2.9), for which the particle density |\fr|2 is constant in time at
each point r. The stationary states are the eigenstates of the Hamiltonian
operator H. If we measure the observable Q, we throw the system into an
eigenstate of the corresponding operator Q. If this state is also an eigen-
state of H, its time development is trivial and a subsequent measurement
of Q will give the same result as the first measurement. According to the
discussion given at the end of the last chapter, in order to ensure that the
eigenstates of Q and H are the same, we must have the condition
[Q,H] = 0. (5.1)
In these circumstances we say that the observable Q is a good quantum

number. Its value can be used to label a stationary state, because it does
not change with time. (See also problem 4.13.)
Orbital angular momentum

An observable that is often useful for labelling stationary states is the
orbital angular momentum. In classical mechanics the orbital angular
momentum of a particle about the origin of coordinates is r A p. In
quantum mechanics the corresponding operator is usually called TiL, and
it is given by
(5.2)
An important property of the operator L is that its square is closely related

to the operator V2 which appears in the Hamiltonian. The calculation of
49
50 5 The hydrogen atom
Figure 5.1. Spherical polar coordinates.
I? is described in appendix C and the result is that

2
d i d
(5.3)
sin 0
where 6 and (j) are polar and azimuthal angles of spherical polar coordi-
nates (figure 5.1). In spherical polar coordinates,!
r.2
V
1 (5.4)
=-
so that the part of V 2 that involves the angles 9 and </> is just —L2/r2.
If we now define the polar axis 9 = 0 of the spherical polar coordinate
system to lie along the ^-direction, it can be shown from the definition
(5.2) of L that its z-component takes the simple form
Lz = —i d/d4>. (5.5)
f For the calculation of V in spherical polar coordinates, see any textbook on vector
calculus, for example H Jeffreys and B S Jeffreys, Methods of Mathematical Physics,
(Cambridge University Press).
Orbital angular momentum 51
According to (5.3), L2 does not involve <j> explicitly, only d/d(/). This
means that L2 commutes with Lz, that is, [L2,LZ] = 0, so that the two
operators have simultaneous eigenfunctions. It is usual to denote these by
Y/m(0,</>), where the labels I and m are defined in terms of the eigenvalues
of L2 and Lz as follows:
L2Ylm=l(l + l)Ylm. (5.6)
That is, m is the eigenvalue of Lz but, for a reason that will become ap-
parent shortly (see (5.9) below), the eigenvalue of L2 is written as 1(1 + 1).
The first equation in (5.6), together with (5.5), tells us that Y/m varies
with <j> simply as eim<f>. That is,
where P[n(9) is some function of 9 only. Using the second equation of

(5.6), together with (5.3), we find that
1 d
sin0d0
The solutions of this equation, which is known as Legendre's equation,

have been well studied. For most values of / and m the solutions are
infinite for 0 = 0 or TT, and so they do not then yield a physically accept-
able wave function. It is found that in order to avoid infinities, and also
to ensure that any linear combination of the wave functions Yim gives a
probability density that is single valued (the probability density should
not change when <p is increased by 2?r, since values of <f> differing by 2TT
correspond to the same point r ) , we must choose / to be a non-negative
integer and restrict m to integers from —I to /. That is, we have the
eigenvalues
1(1 + 1) for i 2 , where I = 0 , 1 , 2 , . . . ,

m for L z , where m = — /, — I + 1 , . . . , I — 1, or / (5.9)
and then one of the two solutions of the differential equation (5.8) is free
from infinities.
Strictly, a given eigenvalue 1(1 + 1) for L2 corresponds to the square of
the angular momentum being equal to U2l(l + 1). However, the usual
terminology is to call the value assigned to / the orbital angular momentum
of the state. The value of m is called the magnetic quantum number, the
reason for this is that effects resulting from the quantisation of the z-
component of the angular momentum are observed when atoms radiate
photons in a magnetic field which points in the ^-direction.
Notice that [L^Lj] ^ 0 for i =fi j . Hence two different components of L
do not have simultaneous eigenfunctions, and we cannot simultaneously
assign definite values to the corresponding observables. However, there
is nothing special about the z-direction. Let La be the component of
L in any other direction. Then it is easy to show that La commutes
with L2. This can be done by direct calculation or by realising that any
three orthogonal directions can be chosen for the coordinate axes, so that
this new direction could equally well have been called the z-direction and
the result follows because we already know that Lz commutes with L2.
Instead of the simultaneous eigenfunctions Yim of Lz and Z 2 , one can
work with the analogous simultaneous eigenfunctions of La and Z 2 , and
the possible eigenvalues of La are m', where m' = —/,—/ + 1 , . . . , / — 1 or
I
This means that it is possible to know simultaneously the exact values of
both the orbital angular momentum I and of any one of its components.
It is not possible, however, to have certain knowledge simultaneously of
more than one component of the angular momentum. (The only exception
is when there is no orbital angular momentum, that is, / = 0. Then all
components of the orbital angular momentum are zero.) Suppose that
a measurement of the square of the orbital angular momentum gives the
value h2l(l + 1), and that a measurement of the ^-component gives the
value mh, with m an integer between —/ and +1. If we now measure some
other component, the value of / will remain unchanged but the system
will have been disturbed in such a way that the information about Lz is
lost. Instead, we will have determined a value m' for this new component,
where m! also is some integer between —/ and +/. If we know the value of
I and the original value of m, we can predict the probability of obtaining
any of the possible values ra', but we shall not describe this calculation
here.
There is in quantum mechanics another type of angular momentum, apart
from orbital angular momentum. This is an intrinsic angular momentum,
not associated with the motion of a particle, and it is given the name
spin. In quantum theory even a point particle may have spin. It turns
out that in the case of spin angular momentum the restriction on the
quantum number / is a little less severe: in addition to possibly being
integral, it may also be half-integral. In particular, for the case of electrons
Spherically symmetric potentials 53
and protons the spin angular momentum has / = | ; we say that these
particles have spin | , and this is always true for all electrons and all
protons. The corresponding possible values for the magnetic quantum
number associated with the spin are m = ± | . (Photons, on the other
hand, have spin 1.) In addition to their spin, the particles may carry
orbital angular momentum, as we have already described. We discuss
spin further in chapter 8.
Spherically symmetric potentials

The properties of the orbital angular momentum are of particular im-
portance in the analysis of the motion of a particle in any potential V
that is spherically symmetric. If V is spherically symmetric about the
origin, it depends only on the spherical polar coordinate r, and not on
the angles 6 and fa so that V = V(r). Then [V,L2] = 0, because L2
involves only the derivatives d/d6 and d/d(j), but not d/dr (see (5.3)).
Similarly, [V, Lz] = 0. The operators L2 and Lz also commute with the
kinetic-energy operator, because this operator is also spherically symmet-
ric: p2/2M — (p2 + py +p2z)/2M involves each of the three coordinate
directions equally. (We have written the particle mass as M, to avoid
confusion with the magnetic quantum number m.)
Hence for the case of a spherically symmetric potential
[H, L2] = 0,
[H,Lz] = 0 (5.10)
and we can find simultaneous eigenfunctions of H, L2 and Lz. That is,

we can find stationary states where L2 and Lz take definite values, and
we can use these values to label the stationary states. This result is not
true in the case of a potential V(r) that is not spherically symmetric, and
that depends on the direction of r as well as on its magnitude.
Making use of (5.4), we have to solve the time-independent Schrodinger
equation
& {l& h) (5.11)

We look for solutions i/j(r) that are simultaneous eigenfunctions of L2 and
Lz, so that their angular dependence is
1>(r) = R{r)Ylm(d,it>), (5.12)
where Y\m is defined by (5.6) and R(r) is some function that remains to
be found. Inserting this structure (5.12) into (5.11), and replacing L2 by
its eigenvalue /(/ + 1), we find that R(r) satisfies the differential equation
Because m does not appear in this equation, the eigenvalues E obtained

from the equation will depend only on / and not on m. That is, for each
value of / the states with the (21 + 1) possible different values of m given
in (5.9) all have the same energy. We say that these states are (2/ + 1)-
fold degenerate. The reason for this degeneracy is that, for a spherically
symmetric potential, no direction in space is physically different from
any other. For a rotationally symmetric Hamiltonian, the energy cannot
depend on the magnitude of the angular-momentum component in any
particular direction.
In the differential equation (5.13) for R(r), the term involving 1(1 + 1) is
simply added to V(r). Thus it has the effect of an additional potential,
which is repulsive. It is known as the centrifugal potential, by analogy
with classical mechanics. Usually we expect to find that the ground state
of a system (that is, the state of lowest energy, with the tightest binding)
corresponds to / = 0, so that the repulsive centrifugal potential is then
absent. We shall see below in (5.17) that this is indeed the case for the
hydrogen atom, though for some choices of the potential V(r) the ground
state can correspond to a larger value of /.
The hydrogen atom

The hydrogen atom consists of an electron of charge — e bound to a proton
of charge +e. Suppose that the centre of mass of the atom is at rest. Since
the proton is nearly 2000 times heavier than the electron, the proton is
then almost at rest and we may as a good approximation treat it as if it
were. We explain how to relax this approximation later.
The approximation has reduced the problem to that of a single body, the
electron, moving in the Coulomb field of the fixed proton. This field
corresponds to the potential
V(r) = -e2/47T£0r, (5.14)
where we have taken the proton to be at rest at the origin. We insert this
potential into (5.13) and look for a solution for the electron wave function
The hydrogen atom 55
that corresponds to a bound state. For a bound-state solution, we impose

the condition that R(r) -> 0 as r —> 00, so that the electron is localised
somewhere near the proton. In a bound state, the electron does not have
enough energy to escape to infinity. Because the potential (5.14) is zero
at infinity, we therefore expect that the energy E will be negative, just as
in the case of the one-dimensional square well (chapter 3). So we write
E = -ftV/2Af. (5.15)
Then (5.13) becomes
We show in appendix A that this has solutions such that R(r) —> 0 at
infinity and the behaviour at r = 0 is suitable if, and only if, Me2/AKSQKTI
is a positive integer n, and / is restricted to integers from 0 to (n — 1).
That is,
Me 4 1
E
& ^ " = 1,2,3,...,
Z = 0,l,2,...,(n-l). (5.17)
The integer n is known as the principal quantum number. The ground

state corresponds to n = 1 and therefore I = 0.
It is interesting that the result (5.17) for E was first obtained by Bohr,
using what is now known as 'old-fashioned' quantum theory. His work
appeared some twelve years before the foundations of modern quantum
theory were laid by Heisenberg, Schrodinger and Dirac. Bohr's theory
turned out to be too simple and restrictive in its applications, but it
does happen to give the correct result for the bound-state energies of the
hydrogen atom. Bohr postulated that the electron moves round the proton
in a classical circular orbit, and imposed a quantisation condition which
requires that the orbital angular momentum be an integral multiple of h.
The ground state of the atom corresponds to n = 1 and I = 0 in (5.17).
When / = 0, we must have m = 0, and from (5.7) and (5.8) one can
see that Foo(#50) is a constant (remember that Ybo is the solution that is
everywhere finite). We show in appendix A that the ground-state wave
function is
56 5 T/ie hydrogen atom
where
. (5.18)
It happens that ao is the radius of the ground-state orbit of the electron
in Bohr's model; it is known as the Bohr radius. In the proper quantum
theory the expectation value of the radius of the atom in its ground state
i S
o /
err |\I/f r = 4n drr6 3— = | a 0 .

Jo ™o
Notice that in addition to the (21 + l)-fold degeneracy with respect to ra,
which is found for any spherically symmetric potential, we find in (5.17)
an additional degeneracy. For each value of n, we have the same value
of E for the n different values of /. This degeneracy is peculiar to the
Coulomb potential, and so it is known as accidental degeneracy.
Many-electron a t o m s
Consider now an atom of atomic number Z > 1, so that the charge on
the nucleus is Ze and there are Z electrons when the atom is electrically
neutral. Each electron interacts with the nucleus through a potential
—Ze2/A-KE^r so that if we neglect the interactions between the electrons
the possible electron energy levels are obtained from (5.17) by making the
replacement e2 —> Ze2. The neglect of the electron-electron interaction
turns out to be a fair approximation if Z is not too large.
As we have explained, the electron has spin | . A basic principle of quan-
tum mechanics, the Pauli exclusion principle, states that no two identical
particles of spin | may occupy the same quantum-mechanical state. Cor-
responding to each wave function R(r)Yim(6', </>) there are two different
electron states, because the ^-component of the electron spin can be ei-
ther + | or — | . Hence in the ground state of the atom, two electrons
occupy the lowest-energy level (5.17), that is, n = 1, Z = 0. In the next
level, n = 2, there can be two electrons with / = 0, and six with / = 1,
because then m can take any of the three values - 1 , 0 , + 1 . We fill in the
levels in this way, starting from the lowest, until we have allocated all the
electrons, and so obtain the configuration of the ground state of the atom,
(For an atom with very many electrons, the discussion is more compli-
cated, because the energy of the interactions between pairs of electrons
cannot then be neglected.)
Two-body systems 57
Two-body systems
For the hydrogen atom at rest, we assumed that it is a good approxima-
tion to regard the proton as being at rest. We then wrote the Schrodinger
equation for the wave function ip{r) for the electron. If we do not make
this approximation, which reduces the problem to the single-body case,
we must introduce a wave function ip{ri,r2) that depends on the coordi-
nate ri of the electron and also on the coordinate r 2 of the proton (see
problem 2.5).
Generally, for a system of two interacting particles the potential will be a
function of the separation between the particles. Hence the Hamiltonian
operator takes the form
H = -(ft 2 /2Mi)V? _ (h2/2M2)V22 + V(rx - r 2 ), (5.19)
where V\ operates only on the coordinate r\ and V 2 operates only on the

coordinate r 2 . In (5.19) the first two terms are the separate kinetic-energy
operators for the two particles. It is convenient to change to centre-of-
mass variables, as one would do in solving a two-body problem in classical
mechanics:
so that r is the relative separation of the two particles and JR is the position
of their centre of mass. In terms of these variables, (5.19) may be reduced
with a little algebra to the form
H = -{h2/2M)V2R - (h2/2»)V2r + V(r), (5.21)
where
M = M1+M2, n = M1M2/{M1 + M2). (5.22)
The quantity \x is known as the reduced mass of the system.
Because V does not depend on R, the time-independent Schrodinger equa-
tion
H<ip{rur2) =Et
has solutions of the form
where
E + E' = Etot. (5.23)
The second equation in (5.23) says that the centre of mass R of the system
has a motion equivalent to that of a single particle of mass M moving
freely, in zero potential. If the system is at rest, E1 = 0. The first
equation states that the relative motion of the two particles is the same as
the motion of a single particle of mass \i moving in the potential V. This
result is precisely that obtained in classical mechanics.
Hence in expression (5.17) for the energy levels of the hydrogen atom, the
mass M of the electron should be replaced by the reduced mass \i of the
system. However, because the mass of the electron is much smaller (by a
factor of about 2000) than that of the proton, /z « M.
When the result (5.17), modified in this way, is compared with experi-
ment, there is good agreement. As we explained in chapter 1, the com-
parison is made by examining the emission or absorption spectrum of the
atom. The main discrepancy between theory and experiment may be ac-
counted for by including in the Hamiltonian magnetic interaction terms
in addition to the electrostatic Coulomb potential. The most important
of these terms arises because the electron has both orbital and spin an-
gular momentum. It can be shown from relativistic quantum mechanics
that a charged particle of spin | , such as the electron, necessarily carries
an intrinsic magnetic dipole moment, and the magnitude of this dipole
moment can be calculated with high accuracy. The orbital motion of the
electron generates a magnetic field; in rather crude terms this is because a
charged particle moving in a closed orbit is equivalent to a current loop.
The energy of interaction between this magnetic field and the electron's
dipole moment must be included in the Hamiltonian. As we explain in
chapter 8, it leads to a small correction to the energy levels (5.17). For
a given value of n, the states with different values of / are then not ex-
actly degenerate. There are other effects of a relativistic origin, and when
all corrections have been made the calculations agree with experiment to
better than one part in 106.
The deuteron 59
The deuteron
The nucleus of the isotope of hydrogen known as deuterium is called the
deuteron. It is a bound state of a neutron and a proton. As their masses
M are almost equal, the reduced mass of the system is /i « | M .
According to (5.23), the relative motion of the proton and the neutron is
described by the Schrodinger equation
\-V]ip{r) =E^{r). (5.24)
Suppose that V depends only on \r\ — r, and treat this as in (5.11), where
now L is associated with the relative orbital angular momentum of the two
particles. Replace L2 by its eigenvalue, as in (5.13). We explained that
we expect to find that the ground state of the system, with the tightest
binding, corresponds to / = 0, so that the repulsive centrifugal potential
is absent.
As a simple model for the potential, take the spherical well
—U for r < a
^0 for r > a.
Then for / = 0
(d2/dr2)ri?(r) = ( ^ l ) . J" r >a (5.26)

v v ;
' ' [ -KzrR(r) for r < a, v }
where
E = -h2k2/2^ E + U = h2K2/2fi. (5.27)
Hence we have the solution
\ie:k" f Vr
? >a (5.28)
v
for r < a, '
where A and B are constants. Here we have imposed the constraint that
R(r) —> 0 as r -» oo, so that we indeed have a bound state. We have
also required that R(r) does not diverge at the origin: if it possessed a
singularity of the form r " 1 , it would not, in fact, satisfy the Schrodinger
equation at the point r = 0. At r = a, we want both R(r) and Rf(r) to
be continuous. This gives the equation
k = -K cot Ka (5.29)
which, together with (5.27), corresponds to an equation for E with discrete

solutions.
It turns out that when realistic values are taken for U and a, the deuteron
is very weakly bound and the ground state is the only bound state. The
wave function goes to zero very slowly as r —> oo, and (r) « 2a. That
is, most of the time the neutron and proton are separated by a distance
greater than the range a of the potential. The possibility of this being
so, and of the system nevertheless being a bound state, is a peculiarity of
quantum mechanics.
Problems
5.1 How does the function Y/m (#,</>) change if the coordinate axes are
rotated through an angle a about the z-direction?
5.2 The quantity e2/Airsohc is called the fine-structure constant. Verify
that it is dimensionless; it is approximately equal to 1/137. Calculate
also the Bohr radius, given in (5.18).
5.3 Calculate the energy levels of the hydrogen atom according to Bohr's
theory as follows. Assume that the electron is in circular orbit and cal-
culate its orbital angular momentum TiL in classical mechanics. Then
require L to be an integer. Verify that this leads to the same energy
levels as the correct procedure based on Schrodinger's equation.
Calculate the velocity of the electron in the lowest-energy Bohr orbit
and determine whether the use of non-relativistic kinematics is justi-
fied.
5.4 Find the most probable distance from the nucleus of an electron in the
ground state (5.18) of the hydrogen atom.
5.5 A hydrogen atom has electron wave function
iP{r,O,(f)) = Cre~Kr ^
By explicitly applying to it the operators Lz, L2 and H, determine the

values of the quantum numbers /,ra and n and express n in terms of
the Bohr radius a$.
5.6 Derive an equation that determines the energies of the bound states of
orbital angular momentum / = 0 of a particle moving in the attractive
potential
V{r) = (-Ti2U/2m) S(r - a),
where U and a are constant. [See the discussion of delta-function
potentials in chapter 3.]
The deuteron 61
5.7 The alkali atoms have an electronic structure that is approximately

hydrogen-like, in that their chemical properties and spectral lines are
determined essentially by a single electron. As a model for the poten-
tial in which this electron moves, take
V(r) = (-e2/47TSOr){l + b/r).
Following the procedure described in appendix A, show that the energy

levels are
Enl = (-e2/87re0a0)[n-D(l)}-2,
where
5.8 In a one-dimensional problem, two particles, each of mass m, interact

through the potential ^muj2(xi — x2)2, where x\ and x2 are their
position coordinates. Find the energy levels of the system when its
centre of mass is at rest.
5.9 The binding energy of the deuteron is approximately 2 MeV and its
radius 2 x 10~ 15 m. Estimate the depth U of the potential in the
spherical-well model, assuming that it is much greater than the binding
energy.
Revision quiz
This quiz is designed to check that you have followed and understood the
basic features of quantum theory.
1 What are the diameters of an atom and of its nucleus?
2 What do we learn from the photoelectric effect?
3 What is the wavelength associated with an electron, and how do we
know this?
4 What are the operators corresponding to the momentum and energy of
a particle?
5 Write down the time-dependent Schrodinger equation. How and under
what circumstances can one derive the time-independent Schrodinger
equation from it?
6 What is the probability interpretation of the wave function?
7 What is the expression for the particle flux j in terms of the wave
function?
8 What are the continuity conditions that are usually satisfied by the
wave function, and how are they derived?
Why are the conditions different for the problem of a particle in a box
and for a particle moving in a delta-function potential? What are the
conditions in these cases?
9 What is the tunnel effect?
10 How does one use stationary-state solutions to express general solutions
of the time-dependent Schrodinger equation? What mathematical as-
sumption is implied?
11 What is the Heisenberg uncertainty principle?
Under what conditions can one simultaneously know the values of two
observables?
12 What is meant by a Hermitian operator? What properties do its eigen-
values and eigenfunctions have r
13 What are the basic assumptions about observables?
14 What is meant by the expectation value of an observable, and how is
it calculated?
15 Why is the orbital-angular-momentum operator useful in problems
where the potential is spherically symmetric?
16 Describe the energy levels of the hydrogen atom. What are their de-
generacies?
17 What is the Pauli exclusion principle?
18 How does one deal with a two-body system?
62
6
The hydrogen molecule
In chapter 5 we saw how in quantum mechanics electrons are bound to

nuclei so as to form atoms. We now give a rather abbreviated account of
how atoms bind together to form molecules. There is more than one type
of molecular binding. We shall confine our discussion to the type known
as covalent binding. The possibility of this type of binding relies on an
effect that is peculiar to quantum mechanics, the tunnel effect, which we
have already encountered in chapter 3.
The ionised hydrogen molecule

The simplest molecule is the ionised hydrogen molecule, which consists
of two protons and one electron. The Coulomb force between the two
protons tends to push them apart; we investigate how the presence of the
electron overcomes this repulsion and holds the molecule together.
An exact calculation is difficult, but we can discuss the general features
of the bonding by making suitable approximations. As the protons are
much heavier than the electron, we may neglect their motion compared
with that of the electron, and so regard them as fixed. We show that the
expectation value of the energy, considered as a function of the proton
separation i?, has a minimum for a certain value of i?, so that there is a
stable equilibrium configuration.
Suppose first that R is so large that in the vicinity of each of the protons
the Coulomb field of the other is completely negligible. Then there are
two sets of stationary states: the electron is bound to one or other of the
two protons with a wave function corresponding to the ordinary hydrogen
atom. In particular, the lowest-energy level of the system is EQ, where
EQ is the energy of the ground state of the hydrogen atom. There are two
different states corresponding to this energy. In one of them the electron
is described by the wave function c/)(r — ri), where r\ is the coordinate
vector of one of the protons and <j> is the ground-state wave function (5.18)
of the hydrogen atom:
' (6.1)
63
64 6 The hydrogen molecule
In the other state, the wave function is <f>(r — r 2 ), where r2 is the coordinate
vector of the other proton. Notice that these wave functions are real.
Now reduce R to a value such that the Coulomb field of one proton is
beginning to be felt in the vicinity of the other. In classical physics the
configuration of the system would be little altered, because it would re-
quire energy from outside to tear the electron from one proton and attach
it to the other. But in quantum mechanics the tunnel effect does allow
this transition to occur without any external energy being supplied. This
means that </>(r — r\) and <fi(r — r2) are no longer stationary-state wave
functions. The system eventually settles down into a rather different sta-
tionary state, which we now discuss.
The stationary-state wave functions are the eigenfunctions of the Hamil-
tonian
2
P2 P2 P2
Lv
2m r v
4TT£O \r —
2
h (6 2)
where m is the mass of the electron. Here, the first term is the kinetic-
energy operator for the electron — remember that we are neglecting the
kinetic energies of the protons. We write the stationary-state wave function
for the ground state of the molecule as a superposition of <f>(r — r±) and
(p(r- r 2 ):
(6.3a)
where
fa = 0 ( r - n) 1 = 1,2. (6.36)
Then |ci| 2 is the probability of finding that the electron is in the con-
figuration with wave function </>i, that is, in orbit round the first proton.
Similarly, |c 2 | 2 is the probability offindingthat it is in orbit round the sec-
ond proton. The expansion (6.3) represents an approximation; it should
contain also terms corresponding to configurations in which the electron
is bound to one of the protons with a wave function associated with one of
the excited states of the hydrogen atom. However, these terms are impor-
tant only when R is taken to be so small that the extra Coulomb energy
of repulsion between the protons is readily able to excite the electron into
one of these states.
The wave function ip(r) has to satisfy the time-independent Schrodinger
equation H^ — EI/J. Insert in this equation the expansion (6.3) for %j).
Then pre-multiply the equation by <f>i and integrate over all space, and
The ionised hydrogen molecule 65
repeat this with 4>2:
n-E) = c2(K12E-H12),
E)
where
Hij = (faHfa) ij = 1 or 2,
and we have used the normalisation (<f>i,<f>i) = (</>25</>2) = 1 implied by

(6.1). Because of the symmetry of the problem, Hu — H22 and H\2 —
H2i- This means that eliminating c\ and c2 from the equations (6.4) gives
/ TIT J?}2 ( J{ fT TT \ 2
which leads to the two solutions
n 12
(i) E = E+= , withci=c2,
(ii) E = E- = Hn J*12, with d = -c2. (6.6)

1 — K\2
In both these solutions, |ci| 2 = |C212, that is, there is equal probability of
finding the electron in orbit round either proton. This result shows how
dramatic are the consequences of the tunnel effect; the situation is very
different from that in classical mechanics. Crudely speaking, the electron
continually tunnels back and forth from the potential well surrounding one
proton to that surrounding the other.
We must now attempt a rough calculation of Hu and Hi2. The first two
terms of the Hamiltonian H in (6.2) are nothing but the hydrogen-atom
Hamiltonian of which cf>i is the ground-state eigenfunction, so
= Eo - / d6r—r-±— + -—-. (6.7)

A v }
—~ ' \r - r2\ 4ire0R
Because of the definition in (6.3) of 0i, we may write the integral in (6.7)
as
where we have written r' = r — r\ and R = r2 — r\. We see from (6.1)

that </>(r') is very small when \r'\ » a0. So if R » a0, then \r'\ < R
throughout that part of the integration from which the integral receives
most of its contribution. Hence we may use the expansion
l/[( r ' _ fi)2]l/2 = [r/2 _ 2f ./ .R + fl2]-l/2
= (1/i?) [1 + ( r ' • i?/i? 2 ) + . . - ] . (6.9)
When we insert this expansion into the integral, the first term integrates
to give —e2/AneoR- The second term, being odd in each of the three
coordinates of r', integrates to give zero. Hence for large i?, (6.7) reduces
to
i f n « E o. (6.10a)
This approximation is correct up to and including terms of order ao/i? 2 ,
where the factor CLQ appears here because (pi is appreciable only in the
region where r' « oo- That is, the correction is of order OQ
On the other hand, when R —> 0 the integral (6.8) evidently remains
finite. The last term in the Hamiltonian (6.7) diverges as R -> 0, and so
dominates for small R. Hence for small R we have
(6.106)
We consider H\2 similarly:
H12 = Ai2ô - ~A / d r- r+
4T£ J \ \r-n
r \
The integral here is —V(R), where
(6.12)
This integral does not diverge at r' = 0, because the volume element d 3 r '
contains a factor r'2 when it is written in spherical polar coordinates. Nev-
ertheless, relatively small values of r' are most important in the integral,
and so when R is large
Hence for large R we may neglect the last term in (6.11) compared with
the others, and
(6.13a)
Other molecules 67
Figure 6.1. E± plotted against R.
The integral (6.12) is evidently finite at R = 0, so that for small R the

last term in (6.11) is dominant and
#12 « #12 e2/4<jre0R. (6.136)
Finally, from its definition in (6.5), #12 is small for large i?, because when
(pi is large 02 is nearly zero, and vice versa. If we use this, and put (6.10)
and (6.13) together, the energies E± in (6.6) become
£0 =F V(R) R large,
E± (6.14)
R small.
By inspection of (6.12) we see that V(R) is positive, and it decreases in

magnitude as R is increased, becoming small for large R. Hence the plots
of E± against R must have the shapes drawn in figure 6.1. Because of
(6.14), E+ and E- are both very large and positive for small R. At large
i?, E+ approaches its asymptotic value Eo from below. Therefore E+
must have a minimum at some value of R\ that is, there is a configuration
of stable equilibrium. The corresponding value of the energy E+ is less
than EQ, which is the minimum energy of a separate proton and hydrogen
atom, so that the configuration is a bound state.
Other molecules
A similar type of covalent binding provides most of the binding in the neu-
tral hydrogen molecule, in most organic molecules, and in some inorganic
molecules. (However, most inorganic molecules are bound mainly by a
different mechanism, known as ionic binding.) When only one electron
is involved, the covalent bond can be strong only between similar atoms:
the symmetry properties Hn = #22 and Hu = #21 were of crucial im-

portance in our analysis. (See problem 6.3.) However, two electrons can
form a strong covalent bond between two atoms A and B even when these
are different. The theory is similar, but with <j>i now the wave function for
the state where the first electron is attached to atom A and the second to
atom 5 , and 02 the wave function for the corresponding state where the
electrons are interchanged. It is evident that in this case the equalities
Hn = H22 and H12 = #21 do hold.
Problems
6.1 As a one-dimensional model for the ionised hydrogen molecule, sup-
pose that the potential near an atom is a delta function, so that for
the molecule
V(x) = (-h2U/2m)[8(x -R) + 5(x + R)].
Show that the wave function of the electron has the form
,, v (A{eKX±e~KX) 0<x<R
and find equations that give the bound-state energy as a function of

R. [You may find it helpful to refer to the discussion of delta-function
potentials in chapter 3.]
6.2 The muon is a particle whose properties are closely similar to those
of the electron, except that it is some 200 times heavier (and it is un-
stable, with a lifetime of about 10~6 s). If it is used to bind together
two nuclei, how does the size of the resulting molecule compare with
that of the electron-bound molecule?
[We explained at the end of chapter 5 that the radius of the deuteron,
the nucleus of deuterium, is quite large. When a muon binds two
deuterons together the distance between the deuterons is sufficiently
small for there to be a substantial overlap between their wave func-
tions, and so there is a good chance that the two nuclei will fuse and
form a helium nucleus before the muon decays.]
6.3 Suppose that, in (6.4), Hu ^ H22 and Hi2 ^ H2i- Let the energies
of the ground states of the two separate atoms be respectively EQI
and JE 02 , with EOi < #02- Show that for large R the lowest-energy
eigenvalue E differs from EOi by an amount that is not linear in small
quantities, as in (6.14), but quadratic. (The consequence is that a
single electron does not give strong covalent binding between unlike
atoms.)
Other molecules 69
6.4 Three atomic nuclei are fixed at the corners of an equilateral triangle
and an electron is introduced into the system. If this electron were
bound to the atom i in isolation, it would have wave function <^.
Obtain expressions analogous to (6.6) for the energy levels.
7
Introduction to perturbation theory
In this chapter we describe the approximation method for problem solving

known as perturbation theory. There are two types of perturbation theory,
time-independent perturbation theory, used to find the stationary states
of a system, and time-dependent perturbation theory, used to calculate
certain quantities called transition probabilities.
Time-independent perturbation theory

In chapter 5 we described a model of the deuteron in which the proton and
neutron are bound together by the simple spherical-well potential given
in (5.25). This is certainly not the true potential, but it is the model
that most easily leads to a solution of the time-independent Schrodinger
equation. The true potential will also be an attractive potential, and so the
simple form (5.25) is an approximation to it. Thus the question arises:
suppose that we are able to find the eigenvalues of a Hamiltonian Ho,
can we now calculate, at least approximately, the eigenvalues of another
Hamiltonian
H = H0 + H1 (7.1)
that is not too different from Ho?

Let 0j(r), j — 1, 2, 3 , . . . be the various eigenfunctions of HO, correspond-
ing to eigenvalues Ej. Assume that there is no degeneracy, that is, that
the Ej are all distinct. We shall also suppose that the (j)j are orthonormal,
meaning that they meet the condition
(<f>j,<l>k) = 6jk, (7.2)
and that they form a complete set, as discussed in chapter 4. This allows
the eigenfunctions ^ of H to be expanded as linear combinations of the
_ y&, (7.3)
3
with the Cij constant.
70
Time-independent perturbation theory 71
We consider the case where Hi is small. By this we mean that corre-

sponding to a given eigenfunction fa of HQ with eigenvalue E^ there is
an eigenfunction ipi of H that is not too different from </>;, and that corre-
sponds to an eigenvalue Ei + AEi that is not too different from E{. More
precisely, we expect that our approximations will be good when AEi is
small compared with the difference between E{ and the next eigenvalue of
#o, and that all the coefficients cy, except for c#, are such that |cy | < 1.
Then if we normalise fa so that (fa, fa) — 1, which implies that
= 1
' (7-4)
cu will differ from 1 by a quadratically small quantity.

On inserting the expansion (7.3) into the time-independent Schrodinger
equation Hfa = (Ei + AEi) fa, we obtain
(Ho + Hx) Y, Ciî(r) = (Ei + AEi) £ ^ ^ ( r ) . (7.5)
To simplify this equation, use the definition HQ(/)J = Ejfy of the eigen-
functions (f)j. Also, neglect terms that are quadratic in small quantities;
that is, omit those terms where either Hi or AEi is multiplied by cy with
i ^ j , and replace ca by 1. Then
+ X > ; - Edarfjir) = AJE^(r). (7.6)
Now pre-multiply by <^(r) (remember that Hi may contain differential

operators), integrate over all space and use the orthonormality condition
(7.2). This gives
= (cf>uHi<j)i). (7.7)
That is, the shift AEi in the level Ei resulting from the addition of the
perturbation Hi to the original Hamiltonian HQ, is just the expectation
value of Hi calculated from the original wave function fa.
Usually this is as far as we need go in the calculation: the level shift
is the quantity of most interest. However, if we wish also to calculate the
perturbed wave function ^ , we instead pre-multiply (7.6) by </>£(r), with
k ^ i, and again integrate over all space. We then find that
cik = (<j>k,Hifa)/(Ei -Ek) k± i. (7.8)
72 7 Introduction to perturbation theory
By inserting this result into (7.3), together with ca = 1, we have an

expression for ipi(r).
Our calculations are correct to first order in small quantities. It is possible
to extend the calculation to give expressions for the higher-order terms,
but we shall not do this here.
Notice the importance of our original assumption that the level E{ is
non-degenerate; otherwise the expression (7.8) that we have obtained for
Cik would be singular for some A;, and our assumption that all the c^
with k ^ i are small would be inconsistent. If we are interested in the
perturbation AE{ to a particular level E{, our calculation is valid so long
as that level is non-degenerate; it does not matter if some of the other
levels happen to be degenerate. It is possible also to deal with the case
where the level of interest is degenerate. For two degenerate states i and
k such that the numerator in (7.8) vanishes, there is no problem, and the
previous analysis applies without modification. We shall not pursue the
case where the numerator does not vanish.
Example
If we want to apply perturbation theory to calculate the eigenvalues of a
Hamiltonian H, the first step is to write it in the form (7.1). This division
of H into two parts Ho and Hi may be a purely mathematical procedure;
the only requirements are that we can find the eigenstates of HQ, and that
Hi is small in the sense that we have described. Usually, however, the way
in which we choose to divide H will have an obvious physical significance.
As a simple example, consider again the problem of a particle in a rect-
angular box, as described at the beginning of chapter 3. Suppose that
the particle does not now move freely within the box, but rather that it
carries a charge e and is acted on by a uniform electric field £. Take the
field to be parallel to the rr-axis, so that the potential takes the form
V = -e£x. (7.9)
If either e or £ is small, so that this potential is weak, an obvious choice

is to identify Hi with V, so that Ho is the kinetic-energy operator whose
eigenfunctions are given in (3.4), namely
. / . / 8 \ . qnx . my . snz ,_ .
4>qrs(r) = -T- I s i n - — s i n — - sin , (7.10)
\abcj a b c
Time-dependent perturbation theory 73
with q, r, s integers. The corresponding unperturbed energy levels are,

from (3.5),
2 2 2
\
In order to be able to apply the perturbation theory developed above, we

must evidently confine ourselves to the case where the lengths a, 6, c of
the sides of the box are unequal, so as to ensure that the level Eqrs, for
which the shift AEqrs is being calculated, is non-degenerate.
From (7.7), the energy levels in the presence of the electric field are
Eqrs + AEqrs, where
AEqrs = = -e£ / dxdydzxc/)2qrs = -\e£a. (7.12)

•/box
In this simple problem, the perturbation has turned out to be the same
for every level.
Time-dependent perturbation theory

We now investigate the behaviour of a system under the influence of a
perturbation that varies with time:
H = H0 + H1(t). (7.13)
We suppose that, as before, #0 does not vary with t. We again use

the stationary-state wave functions of Ho to make an expansion of the
wave function \t(r,£) of the system. Because we are interested in the
time dependence of \I>, we must now include in the stationary-state wave
functions of Ho their time variation:
*j(r,t) = <j>j(r)e-iEJt'\ (7.14)
where the <j>j(r) are the same eigenfunctions as in (7.3). Because H varies
with t, it does not have stationary-state wave functions; we must solve the
time-dependent Schrodinger equation
H^(r,t) = i/i*(r,*). (7.15)
Suppose that as t -> -oo, Hi(t) -4 0, so that initially H coincides with

Ho. Thus initially the system may be in a stationary state of Ho. We shall
suppose that it is initially in the stationary state i and we shall denote by
î the wave function that coincides with <3>; at t = — oo.
When the perturbation Hi (t) begins to switch on, \I>; will start to become
different from $;. We use the expansion
r,t). (7.16)
Notice that the coefficients a^ are functions of t, because the $j are not
stationary-state wave functions for the perturbed system. At t — — oo,
an = 1, and all the other a^ are zero. Inserting this expansion into the
Schrodinger equation (7.15), and using the equation Ho$j(r,t) = ih$j
which defines $ j , we find that
yî^-. (7.17)
On pre-multiplying this equation by $£, and integrating over all space,

we obtain, with the help of the orthonormality conditions ($&, $j) = 5k j ,
ihdik =
Notice that in (7.17) and (7.18) we have had to be careful about the order
in which we have written the wave functions $ relative to Hi, because H\
may contain the derivative operator d/dr. However, we do not allow Hi
to contain the operator d/dt, so that H does not operate on the coefficient
functions a^(t); these simply multiply H.
So far, we have made no approximations: (7.18) is exactly equivalent
to the time-dependent Schrodinger equation. We now suppose that Hi
is small. This means that we expect the coefficient functions a^-(t) to
depart rather little from their original values, at least if the time t is not
too large. We shall suppose that the value of t is such that au(t) is still
close to 1, and that aij(t) for i ^ j is still small. We then retain in (7.18)
only those terms that are first order in small quantities, which means that
on the right-hand side we keep only the term in the sum for which j — i
and in it we replace an by 1:
ihaik(t) = ($*,#!$;) = (&,tfi&) e ' ^ - W * . (7.19)
Integrating with respect to t, we have
aik {t)=8ik-1- f dt'(4>k,H1tf)<j>i)eitE'>-B*'"i, (7.20)

where the 5ik is the integration constant that takes account of the initial
conditions at t — — oo.
Transition probability 75
Transition probability
We have supposed that initially, at t = —00, the Hamiltonian is just i?o,
and that the system is then in the stationary state $i of Ho- Suppose that
after some time T the perturbation Hi(t) vanishes, so that the Hamiltonian
is again Ho. If we now measure the energy of the system, the system will
be thrown into a stationary state of Ho and will remain in this state. The
probability that the measurement results in the value Ek, and so throws
the system into the stationary state <!>£, is, according to the discussion
given in chapter 4 and to (7.16), just |a^| 2 . Provided that this probability
is small, so that our approximations are valid, we may calculate it from
(7.20):
m 2
\«ik(t)\2 = To
(7.21)
This, then, is the probability that our measurement finds that the per-
turbation Hi has induced a transition from the state $; to the state $fc.
Notice that a^ involves the term (fa, Hi fa). Frequently one can show
without detailed calculation that this term vanishes for all but a few final
states fa, regardless of the value of t'. That is, without detailed calcula-
tion one can find those final states fa that are readily accessible from the
initial state fa. Those final states fa for which (7.21) vanishes may still
be possible final states, but with a transition probability that is non-zero
only at second or higher order in the perturbation H\. Such transitions
are known as forbidden transitions; usually (though not always) their
corresponding transition probabilities are very small.
Example
As an example, consider again the problem of a charged particle in a box,
and let the perturbation again be caused by a uniform electric field in the
^-direction, as given in (7.9). Suppose that this perturbation is switched
on at time t = 0, and then switched off again at time T. We assume that
for t < 0 the system is in the stationary state (f)qrs. The probability that a
measurement at some time t > T will find that there has been a transition
to the stationary state 4>QRS is
2
2
(1/h ) I dt'(<f>QRS, -
Jo
(7.22)
where
I(UJ) = / dt'eiuJt'
Jo
u = (EQRS-Eqrs)/h. (7.23)
The term ((f)QRs,x<j>qrs) is readily evaluated, using the explicit form of

the wave functions given in (7.10):
{(/>QRS,x<f>qr8)= / dxdydz(j)QRSx(j)qrs
Jbox
0 Q - q even (Q ^ q)
X | 8Q/[2(Q2 2)2] Q _ q
(7.24)
So first-order perturbation theory allows only those transitions for which

R = r , 5 = 5 and Q differs from # by an odd integer. Alternatively, the
system may remain in its initial state.
Energy uncertainty principle

Prom (7.23) we find that
|J(o,)| 2 = (4/UJ2) sin 2 \uT. (7.25)
A plot of this function against UJ is shown in figure 7.1. The height of the
central peak is T 2 , while that of the peaks next to it on either side is some
twenty times less. Hence the transitions that occur most readily are those
to levels for which UJ lies within the central peak. This peak is centred on
UJ — 0 and it has half-width 2TT/T. SO the levels to which transitions are
most likely are those whose energy EQRS differs from the original energy
Eqrs by an amount huj of order h/T.
The shorter the duration T of the perturbation, the greater will be the
spread in energy of this band of preferred final levels. This somewhat
surprising result is peculiar to quantum mechanics, and is a manifestation
of the energy uncertainty principle. We explained in chapter 4 that making
a measurement on a system is liable to induce a perturbation, and, in
fact, this principle also governs the accuracy with which it is possible to
measure the energy of a system: it can be shown that if a limited time
T is available for making the measurement, then the energy cannot be
Sudden change in the Hamiltonian 77
-2TT/T 2TT/T
Figure 7.1. Plot of |/(CJ)|2 against w (see (7.25)).
determined better than to within an amount of order h/T. In particular,

if a system is unstable and so has a finite lifetime r, no measurement
can determine its energy to within an accuracy better than about Tt/r.
The energy uncertainty principle is important also for understanding how
the tunnel effect is possible: it allows a particle momentarily to acquire
enough energy to surmount a potential barrier.
Sudden change in the Hamiltonian

Up to now, we have considered perturbations H\[t) that vanish for very
large t, so that the Hamiltonian ultimately returns to its original form
HQ. Consider now the case where the Hamiltonian is initially equal to
the time-independent operator i? 0 , as before, but then at time t = t0 the
Hamiltonian instantaneously changes, so that for all t > to it is equal to
the time-independent operator H'Q. We suppose that we are able to solve
the time-independent Schrodinger equation for both Hamiltonians, and
that we know that the eigenvalues of Ho are Ej, j = 1,2,... and those of
Hf0 are E'^ j = 1,2,... Let the system initially be in a stationary state
of Ho, with energy E{. We calculate the probability that a measurement

of the energy at some time t later than t0 will result in the value J5J-, and
so throw the system into the corresponding stationary state of HfQ. In the
previous work we calculated the first term in a power-series expansion for
the probability; here we are able to give an exact expression.
Let the stationary states of Ho have the orthonormal wave functions
<frj(r, t) as in (7.14). Let those of H'o have the orthonormal wave functions
»;(r,t) = *;(r)e-"3«/\ (7.26)
where
Note that here the prime does not denote differentiation, and the functions
$'• may be completely different from the $j. We do not need to assume
that the difference between HQ and H'o is small. For t < to, the wave
function \I/(r,t) of the system is, by assumption, equal to $*. For t > to
we expand it as a linear combination of the <fr ^:
($i{r,t) t<t0}
* ( r , t ) = < Ebj&jfat) t>t0. (7.27)
I 3
For t > to the <J>^- are stationary-state wave functions of the Hamiltonian,
and therefore the coefficients bj are constant. The wave function (7.27)
will automatically satisfy the Schrodinger equation
HV = ihd$/dt (7.28)
for both t < to and t > t0. At t = t 0? * must be continuous. For if it were
not, d^jdt would have a singularity proportional to S(t — to) at t = to
(see appendix B), and the Schrodinger equation (7.28) would not then be
satisfied at t = t0 since H merely has a finite discontinuity there. The
continuity condition at t — t0 gives
^ ( r , t o ) = *t(r,«o). (7.29)
3
Multiplying this equation by $'fc* (r, t 0 ), integrating over all space and using
the orthonormality conditions in (7.26), we obtain
(7.30)
Hence the probability that a measurement of the energy at some time
t > to yields the value E'k is
M 2 = l(&,*)| 2 . (7.31)
Example: decay of tritium 79
Example: decay of tritium

Tritium is the isotope of hydrogen whose nucleus consists of one proton
and two neutrons. When a neutron is not bound into a nucleus it is
unstable: it undergoes the decay
neutron ->• proton + electron + neutrino. (7.32)
(The neutrino is a particle that has zero charge and also zero mass, so
that like the photon it moves with velocity c.) Consider this decay in the
frame of reference in which the neutron is at rest. The reason that the
decay is possible is that the rest mass mn of the neutron is greater than the
sum of the rest masses of the proton and the electron, so that the initial-
state energy mnc2 is sufficient both to create the final-state particles and
to give them some kinetic energy. However, when the neutron is bound
in a nucleus, the decay may not be possible. Consider, for example, the
deuteron. If its neutron were to decay, the nucleus would have to break
up, because there is no bound state of two protons. However, this break-
up is not allowed by energy conservation: the binding of the deuteron is
strong enough to ensure that the rest mass of the deuteron is less than the
sum of the rest masses of two protons and an electron.
In the case of tritium, energy conservation does allow one of its neutrons
to decay as in (7.32). This is because there is a bound state of two protons
and a neutron, namely the isotope 3 He of helium. The decay is
tritium —>> 3 He + electron + neutrino. (7.33)
Further, the nucleus 3 He is bound rather more tightly than the original
tritium nucleus, so that quite a lot of kinetic energy is available in the
final state. The electron is so energetic that it rapidly escapes, rather
than being trapped in an orbit corresponding to one of the atomic states
of the helium atom.
Suppose now that, before its decay, the tritium nucleus was accompanied
by an orbiting electron in the atomic ground state, so that the tritium
atom was neutral. When the decay (7.33) occurs, the charge on the nu-
cleus suddenly changes from e to 2e, so that the stationary atomic states
correspond to new electron wave functions. Because the 3 He has only one
orbiting electron, it is not neutral but positively ionised. If we measure
the energy of the orbiting electron after the decay of the nucleus, we do not
necessarily find that the orbiting electron is in the new ground state. The
probability that we do find this to be so is calculated from (7.31) by using
the original ground-state wave function for fc and the new ground-state
wave function for (f)'k.
According to (5.18) and the subsequent work in chapter 5,
^ ( ) a o ) - 3 / 2 e - 2 r / a 0 ) (7.34)
where the Bohr radius an is calculated in terms of the appropriate reduced

mass and is approximately the same in both wave functions. Thus,
= (8/ôg) 47rr 2 dre- 3r/ao

= 512/729. (7.35)
So the probability that the helium ion is found to be in its ground state is
a little less than 3/4.
If the helium atom is not in its ground state, and if the electron has not
escaped from the atom, it will be in an excited state. In this case it
will subsequently decay back into the ground state, emitting one or more
photons in the process (see chapter 9). Hence tritium gas provides a useful
source of illumination that needs no external supply of power.
Problems
7.1 A hydrogen atom is placed in a weak electric field 5. Show that to
first order in £ the ground-state energy is unchanged.
7.2 Calculate the first-order perturbation to the ground-state energy of a
linear harmonic oscillator due to an additional potential ex2. Verify
that your result agrees with the exact answer to first order in e.
7.3 A quantum system is capable of existing in only two independent
states 0i and </>2, which are eigenstates of the Hamiltonian Ho with
eigenvalues E\ and E2, respectively. The system is modified by the
addition of a term H± to the Hamiltonian. Find expressions for the
new eigenvalues E[ and E'2\ (i) by exact calculation, (ii) in lowest-
order perturbation theory. Verify that the perturbation-theory result
is an approximation to the exact result.
7.4 A particle of mass m and charge e is contained within a cubical box
of side a. Initially the particle is in the stationary state of energy
Example: decay of tritium 81
3h2/K2/2ma2. At time t = — oc a uniform electric field is switched

on parallel to one of the edges of the cube. Obtain the expression
in lowest-order perturbation theory for the probability of eventually
finding the particle in a state of energy 3/i 2 7r 2 /raa 2 , assuming that the
electric field varies with time as £ e~f /T .
7.5 A charged particle is scattered by an electron bound to a hydro-
gen atom. The problem is considered in terms of a crude one-
dimensional model in which the electron is described by a harmonic-
oscillator wave function and the charged particle induces the pertur-
bation Hi — X8(x — vi). Show that the probability that the elec-
tron is excited from the ground state n = 0 to the n — 1 state is
7r\2h6/2vsm8uj6.
7.6 A particle is in the ground state in a one-dimensional box (given by
the infinite square-well potential (2.11)). What is the probability that
it remains in the ground state when the wall separation is suddenly
doubled?
8
Spin
Two kinds of angular momentum

In chapter 5 we explained that, just as in classical mechanics, quantum
theory introduces an orbital angular momentum vector TiL — r x p asso-
ciated with the motion of a particle. We explained that there is another
type of angular momentum, not associated with the motion of a particle,
which is given the name spin. These two types of angular momentum oc-
cur too in classical mechanics; for example, the earth has orbital angular
momentum because of its motion round the sun, with also spin angular
momentum because it spins about an axis. In classical mechanics the
spin angular momentum depends on the radius of the spinning body, and
vanishes in the limit of zero radius. But in quantum theory even a point
particle may have spin angular momentum.
We have already mentioned in chapter 5 that the fact that electrons have
spin influences how the Pauli exclusion principle controls the energy levels
of multi-electron atoms. We explain in this chapter some of the many other
experimental reasons why we know that electrons, protons, photons and
other particles have spin angular momentum. First, however, we outline
how quantum theory allows even a point particle to have spin. For the
case of orbital angular momentum, the eigenvalues of L2 and Lz are given
in (5.9):
1(1 + 1) for X2, where / = 0,1, 2 , . . . ,

m for L z , where m = —/, — I + 1 , . . . , I — 1, or /. (8.1)
These allowed values for / and m may be deduced by explicitly solving

the differential equation (5.8) for the simultaneous eigenfunctions of L2
and Lz. However, there is another method, which instead uses what is
called the algebra of angular momentum. One starts with the commuta-
tion relations (4.28) between the components of the position and momen-
tum operators, and derives from them commutation relations between the
components of L:
[Lx,Ly] = iLz, [Ly,Lz] = \LX, [LZ,LX] = iLy. (8.2)
82
Spin \ 83
By simple algebraic manipulation, without solving any differential equa-

tion, one may show from these commutation relations that, for given /,
the allowed values of m are just those shown in the second line of (8.1).
This requires then that — I and / are separated by a non-negative integer,
so that 21 is a non-negative integer. The values of / given in the first line
of (8.1) satisfy this condition, but it also allows instead
/ = § , § , §, .... (8.3a)
These new solutions for / are consistent with the algebra of angular mo-
mentum but not with the explicit differential equation (5.8) that is associ-
ated with orbital angular momentum, so we identify them with a different
kind of angular momentum. As it is not associated with the motion of the
particle, rather it is intrinsic to it; we call it spin.
It is convenient to keep the symbol UL for orbital angular momentum, and
use instead US when we are referring to spin. We denote the eigenvalues
of S2 by 5(5 + 1). The angular-momentum algebra requires 25 to be a
non-negative integer:
5 = 0, | , 1 , | , 2 , § , . . . . (8.36)
Experiment finds that a given type of fundamental particles always has
the same spin 5. Thus, electrons all have spin 5 = | , as do protons and
neutrons. Photons have spin 5 = 1. Some particles have zero spin, for
example 7r-mesons (these are produced when protons are smashed together
at high energy).
Generally, then, a particle will have both orbital angular momentum and
spin. Its total angular momentum TiJ is the vector sum of the two, and
we write
J = L + S. (8.4)
The rule for calculating the eigenvalue of Jz for a simultaneous eigenstate
ij) of Lz and Sz is simple: if Lztp = mip and Sztp = sz%l) then Jzip = (m +
sz)ij). However, the rules (which follow from the algebra) for calculating
the allowed values j(j + l) for the eigenvalues of J2 are more complicated.
If the eigenvalue of L2 is 1(1 + 1) and that of S2 is 5(5 + 1), then the allowed
values for j are
\l - 5|, |/ - s\ + 1, . . . (/ + 5 - 1), or (I + 5). (8.5)
This is something like classical mechanics, where the minimum magnitude
of the sum of two vectors is again the difference between their separate
magnitudes, and the maximum is again their sum, but in quantum me-
chanics the details are more complicated and we shall not elaborate on
them.
84 8 Spin
Spin \
The special case of spin \ is particularly important, because it applies
to the principal constituents of matter: the electron, proton and neutron.
For spin 5, and only for spin | , we represent the algebra in a very simple
particular concrete form. First, it is useful to write for this case
5 = \a. (8.6)
The operator a is known as the Pauli spin matrix. Since we are describing
the case where S2 has eigenvalue 5(5 + 1), a2 has the single eigenvalue
3. Since we only use the operator a when it is applied to states of spin | ,
whenever we use it the operator 5 2 takes the value 3, and so effectively
a2 = 3. (8.7)
Further, because when we apply it to a spin-| state, Sz has two possible

eigenvalues, ± | , oz has possible eigenvalues ±1 and o2z therefore has only
one possible eigenvalue: effectively
a2z = 1. (8.8a)
There is nothing special about the ^-direction: we may choose the z-axis
to be in any direction without changing the physics. So the square of the
component of a in any direction must be 1; in particular
a\ = 1, a2y = 1. (8.86)
To derive further relations, we now use the algebra satisfied by the com-
ponents of a, which we obtain from (8.2) by replacing L with \a\
[crx,ay] = 2iaz, [<Jy,<rz] = 2iax, [(Jz,crx] - 2iay. (8.9)
Multiply the last of these relations by crx, multiplying each term first on
the left and then on the right:
(Jza2x — crxazax = 2iayax. (8.10a)
We use (8.86) to set <J\ — \ and then add these relations, so obtaining
axay + ayax = 0 . (8.106)
The electromagnetic interaction 85
With the first commutation relation of (8.9), this gives

axay = \az. (8.8c)
In a similar way we may show also that
GyGz = \GX, GZGX = \Gy. (8.8d)
While everything involving spin \ can be calculated by using the six simple
relations (8.8), it is often useful to cast them into a more concrete form.
We can do this by associating a 2 x 2 matrix with each component of a:
/on x
{ 8 U a )
'
This set of matrices is called a "representation" of the components of a.
Although we have written = signs in (8.11a), more properly we should
say "is represented by". Each matrix is a Hermitian matrix, because
the operator it represents is a Hermitian operator. The representation
(8.11a) is not unique: we could have used larger matrices. It is not even
unique if we restrict ourselves to 2 x 2 matrices. It has the feature that
the matrix that represents GZ is chosen to be diagonal. This means that
its eigenstates have a particularly simple representation. We usually label
these eigenstates by t and | . They are represented by
It is easy to see that
!)(i)-(i)- (S!)(?)~(!
so that u^ is the eigenstate of GZ with eigenvalue +1, and u± is its eigen-
state with eigenvalue —1.
To obtain a general spin-| stationary state, we take a linear combination
of u^ and u^. The coefficients that multiply them will usually be functions
of r. So the state has a representation of the form
We call u a "two-component spinor". We choose to normalise it so that
Jd3ruUr)u(r) = Jd3r (h/>i(r)|2 + |^(r)| 2 ) = 1. (8.14)

The obvious interpretation of \ipi(r)\2 is that it is the probability density
that a measurement of GZ will give the result +1, and |V;2(^)|2 is the
probability density that it will give the result - 1 .
86 8 Spin
The electromagnetic interaction

We saw in chapter 5 that the bound states of a spherically-symmetric
potential, such as the Coulomb potential which binds the electron in the
hydrogen atom, have energies that are independent of the eigenvalue m of
the ^-component of the orbital angular momentum. As we shall show, this
degeneracy is broken when the atom is placed in a magnetic field. One of
the direct pieces of evidence that the electron has spin \ comes from the
structure of this degeneracy breaking.
To study this, we must first add a term to the Hamiltonian to take account
of the interaction of the electron with the external electromagnetic field.
It is found that the correct way to do this is to follow the same procedure
as in classical physics. In classical mechanics, a Hamiltonian H is used to
give the equations of motion of a system by eliminating p in the equations
_ dp__8H
(8 5)
dt - dp' d* " dr'
These equations are known as Hamilton's equations. For example, for a

particle moving in a potential V(r), where the Hamiltonian is
V(r), (8.16)
Hamilton's equations give
r = p/M, p = - VV,
and together these equations give, correctly, mr = — VV. Suppose that

the particle is an electron, so that it has charge —e, and that we switch
on an external electromagnetic field. This field may be described by a
scalar potential </>(r,t) and a vector potential j l ( r , i ) , so that its electric
and magnetic vectors are given by
A, B = VAA. (8.17)
(See appendix D.) The Hamiltonian then changes from Ho to
H = V + (p + eA)2/2M - ecj). (8.18)
In order to verify this, one shows that Hamilton's equations (8.15), to-
gether with (8.17), now give
Mr = -VV -e£ -er AB. (8.19)
The Zeeman effect 87
We do this in appendix D. In (8.19), the last two terms are just the correct
Lorentz force for the action of the electromagnetic field on the charged
electron.
In quantum mechanics, the Hamiltonian (8.18) becomes an operator, be-
cause p must be replaced by the operator p = —i/iV. The vector potential
A is not fully determined from the definitions (8.17), and it is convenient
to impose on it the further gauge condition V • A = 0 (see appendix D).
This has the consequence that, when we multiply out the term (p + eA)2
in (8.16), the order in which we write p and A in their dot product does
not matter. For if \I> is any wave function,
ihA • V * = -ih(V • A)V = 0.
We can therefore write (p + eA)2 = p2 + 2ep • A + e2A2. In most

applications, the external electromagnetic field is small, in the sense that
it is much smaller than the internal Coulomb field that binds the electron
in the atom. Then the term ep • A is rather larger than e 2 A 2 , which can
therefore be neglected. So then (8.16) and (8.17) give
H = HQ + i?i,
Hi = (e/M)p - A-e<l> = {e/M)A • p - ecf). (8.20)
The Zeeman effect

Suppose now that a hydrogen atom is placed in a uniform external mag-
netic field B pointing in the z-direction. A vector potential that satisfies
the gauge condition V • A = 0 and B — V A A is
A = \rf\B. (8.21a)
If there is no external electric field, 0 = 0, and so (8.20) gives
# i = (e/2M)r A B.p = (e/2M)B.(r A p)

= {eh/2M)BLz. (8.216)
This commutes with the original spherically-symmetric hydrogen-atom

Hamiltonian Ho. The eigenstates (5.12) of Ho which we found in chapter
5 corresponded to definite values m for the operator L z , so in these states
Hi takes the value
AE = (eh/2M)Bm. (8.21c)
88 8 Spin
This has to be added to the original energy levels (5.17), which were
independent of m, so placing the atom in a magnetic field removes some
of the degeneracy.
This affects the spectrum of photon frequencies seen when the atom makes
transitions between different levels. For example, consider a transition
from the first excited state of the hydrogen, n = 2, to its ground state,
n = 1. Here n is the principal quantum number: see (5.17). The ground
state is necessarily / = 0 and so m = 0; it is not affected by the external
magnetic field. The n = 2 level can be I = 0, which again is unaffected;
but it can also be / = 1, for which m can take any of the values 1,0 or
— 1. The m = ±1 states have their energies shifted by ±(e/2M)B, so that
the n — 2 level is split into three levels that are close together (because
in practice B is "small" in the sense we have already explained). So we
expect that the n = 2 to n = 1 transition should give three lines in the
spectrum whose frequencies are close to one another.
This is not what is observed experimentally. The discrepancy is explained
by the fact that the electron has spin, and therefore it can have an intrinsic
magnetic dipole moment. A spinless particle cannot have an intrinsic
magnetic dipole moment, because there is no direction intrinsic to the
particle along which it can point. However, for a particle that has spin
vector 5, there can be an intrinsic magnetic dipole moment pointing in
the direction of S. Its magnitude may be found from experiment, but in
the case of the electron it may also be calculated from a relativistic wave
equation for the electron which was derived by Dirac. The electron's
intrinsic magnetic dipole moment vector is found to be — (eh/M)S, so
that in the presence of the external magnetic field B pointing in the z-
direction there is an additional term
{eh/M)S • B = (eh/2M)Baz. (8.22)
in Hi. Before we included this term, the stationary-state wave functions

4>nim{r) could be in any spin state, that is they could be multiplied by
any spinor whose components are independent of r, because the Hamil-
tonian was independent of a. When we add in the term (8.22), we obtain
eigenfunctions of the new Hamiltonian by multiplying VWi(^) by a two-
component spinor u whose components do not depend on r, but which are
eigenfunctions of oz. There are two of these, u^ and u± given in (8.116),
corresponding to eigenvalues ±1 for az. So every energy level we had
before is split into two, with additional energy ±(e/i/2M), depending on
whether Sz has eigenvalue ±1. This applies even to the / = 0 states, in
particular to the n = 1 ground state.
Spin precession 89
There should be another term in Hi, though it is relatively unimportant

unless the external field B is very weak. This extra term is called a spin-
orbit interaction. Just as (8.22) is the interaction of the intrinsic magnetic
dipole moment with the field, so we may think of (8.216) as the interaction
with the field of a magnetic dipole moment —(eh/2M)L. This is familiar
from classical electromagnetism, where a charged particle in orbit behaves
effectively like a loop of electric current, and a loop of electric current has
an associated magnetic dipole moment. So we expect Hi to include a
term that represents the interaction between the two magnetic moments;
its precise form may be calculated from Dirac's relativistic equation. Ev-
idently it does not depend on B, so when B is large enough this extra
term is less important than the ones we have described. The proton in
the nucleus also has a magnetic dipole moment, which interacts with that
of the electron; again this is relatively unimportant if B is not too weak.
Spin precession
In (8.13) we introduced a two-component spinor u(r) to describe a station-
ary state of a spin-1 particle. It satisfies the time-independent Schrodinger
equation
Hu(r) = Eu{r). (8.23a)
The Hamiltonian H is represented by a 2 x 2 matrix; for example it might
contain a term involving az as in (8.22). If we are interested instead in a
state that is not stationary, we must use the time-dependent Schrodinger
equation and allow the spinor to depend on t as well as r;
Hu(r, t) = ih—u(r, t). (8.236)
A simple example is a proton which, before any magnetic field is switched

on is in a bound state corresponding to some spin-independent Hamilto-
nian Ho. Suppose that its normalised time-dependent wave function is
${r,t) and that its spin points in the ^-direction, so that it is described
by the spinor
( (8.24a)
with
H0V(r,t) = ih—V. (8.246)
Suppose that at time t = 0 a, magnetic field B is switched on pointing in
the rc-direction. The Hamiltonian then becomes
H = Ho - UJSX = Ho - \wox, (8.25a)
90 8 Spin
where
u) = 2hvB, (8.256)
with /i the magnetic moment of the proton. The effect of adding a spin-
dependent term to the Hamiltonian is to make the spin state of the proton
change with t. We look for a solution to the time-dependent Schrodinger
equation (8.236) of the form
(8.26)
Substitute this into (8.236) and use (8.246) and the representation (8.11a)
for ax:
0 -±7i(
fa 0 JU2(t)J -""dt\u2{t)
or
= ik
diUl^
ih^-u2(t). (8.276)
dt
If we introduce
= U1(t)+U2(t),
= Ul(t)-u2(t). (8.28a)
the equations (8.276) become simple:
ift-rr = ~^0JU, i^-r~ = §ti>V. (8.286)
Their solution is
= Uoe-*iwt, V(t) = V0e5'Mt. (8.28c)
We have said that the spin state at t = 0 is such that t*i(0) = 1 and
« 2 (0) - 0, so that Uo = Vo = 1. So, from (8.26), (8.28a) and (8.28c),
«(r,0 = • ( , .
Problems 91
Because the original wave function \&(r, t) was supposed to be normalised

to 1, and because cos2 ^cut + sin2 \ut = 1, the spinor u(r,t) is normalised
to 1 also: it satisfies the normalisation condition (8.14) for all t. The
probability that a measurement of Sz gives + | at time t is cos2 \ut, and
the probability that it gives - \ is sin2 \wt. The expectation values of the
components of S at time t are
(5) = fd3rJ(r,t) \au(r,t). (8.30a)
Using the representation (8.11) for the components of a we find

(5) = (0,-sinu;£,cosu;£). (8.306)
What we have found is that, if initially we know that the spin points
in the +z direction, at later times there is non-zero probability that a
measurement will find that it points in the — z-direction. The expectation
values of the components of the spin vector 5 vary as if, in classical terms,
the spin vector steadily rotates in the yz plane with angular frequency u:
this is called spin precession.
Because the magnetic moment of the proton points along its spin vector,
when its spin precesses the direction of its magnetic moment vector also
oscillates. Just as in classical physics, this causes the emission of electro-
magnetic radiation with angular frequency u. Conversely, if electromag-
netic radiation of angular frequency u is shone on the precessing proton,
there is a relatively high probability that it is absorbed. An important
application is to nuclear magnetic resonance imaging, used in medicine
and chemistry. A specimen is subjected to a magnetic field, so that the
spins of the protons within it precess. Electromagnetic radiation is shone
on the specimen and measuring the frequencies that are absorbed by dif-
ferent parts of it gives information about the magnetic moments of the
nuclei concerned and hence about its structure.
Problems
8.1 Derive the commutation relations (8.2).
8.2 What are the eigenvalues of ax? With the representation (8.11a) find
representations for its corresponding eigenstates.
8.3 With the definitions S± = (Sx ± iSy), show that the spinors given in
(8.116) obey S-U^ = u± and S+u± = u^. Evaluate also S-U± and
S
92 8 Spin
By writing S2 = S2 + §(S+ S_ + 5-54.), verify that both u t and

satisfy
2
8.4 The operators S^1) and S^ are the spin vectors of two electrons and
the components of S^ commute with those of S^2\ The total spin
vector is the operator S — S^ + S(2\ What are the eigenvalues of
Szl What does (8.5) give as the eigenvalues of S2 ?
It is clear that the two-particle spinors u^ and u^ are both eigen-
spinors of Sz with eigenvalue 0; however, they are not eigenspinors
of S2. Show that the superpositions x^ = ^(uti ^ Ît) a r e e ig e n "
2
spinors of S , and find the eigenvalues. [Hint: use the results of the
previous question.]
8.5 A particle of spin | and magnetic moment | TIJAG is at rest with its spin
pointing along a magnetic field B. A small additional magnetic field B'
is switched on, such that the angle between the directions of B and B'
is 45°. Use perturbation theory to calculate the new stationary-state
energy correct to order B' and compare with the exact answer.
8.6 An electron is in a state in which the y-component of its spin is known
to be + | . The x-component of its spin is now measured; calculate the
probabilities of obtaining each of the possible results. What can be
said about a further measurent of the y-component of its spin?
9
Masers and lasers
Radiative transitions
When a system changes its energy as the result of the emission or the
absorption of a photon, it is said to undergo radiative transition. There
are three kinds of radiative transition. In the presence of an electromag-
netic field the system can absorb a photon, so that its energy is raised to
a higher level. If the system is initially in a state other than the ground
state, it can emit a photon and so shed energy. This can happen with-
out the presence of any external electromagnetic field, in which case a
spontaneous emission is said to occur. On the other hand, if the excited
system is placed in an electromagnetic field that varies in time with the
appropriate frequency, the probability that it will emit a photon can be
greatly enhanced. In this case the emission process is known as stimu-
lated emission. Stimulated emission is the basis of maser and laser action,
which we describe in this chapter.
First, we construct a simple model that allows us to investigate how the
internal structure of a quantum system changes as the result of a radia-
tive transition. Although the model is very simple, it will allow us to
understand the main features of the proper, more exact treatment. We
shall deal mainly with absorption and stimulated emission, the two types
of radiative transition where there is an external electromagnetic field.
Resonant absorption and stimulated emission

We consider the interaction between an electron bound to an atom and
an electromagnetic wave. In the absence of the electromagnetic field, the
electron wave function satisfies the Schrodinger equation with Hamiltonian
HQ. Let the stationary-state wave functions of Ho be $^ = (j)k^~iEkt^h,
corresponding to energies E^. In reality, only the ground state is exactly
a stationary state, because the higher states can spontaneously emit ra-
diation and so change their configuration. By taking all the levels E^ to
be stationary states, we are making an approximation in which the possi-
bility of spontaneous emission is neglected. Our calculation will consider
only the stimulated emission (and absorption) induced by the external
93
94 9 Masers and lasers
electromagnetic field.
We shall take this field to be the plane wave given by
0 = 0, A = Ao sin(K • r - cut). (9.1)
Then, according to (8.17), the electric and magnetic vectors are
S = UJAQ COS(K • r — ut)

B = K A Ao cos{K • r - ut). (9.2)
In order to satisfy Maxwell's equations (which are listed in appendix D),

we must impose the constraints.
u = cK, K'A0 = 0. (9.3)
The second of these equations also makes A satisfy the gauge condition
V • A = 0. We shall choose the directions of the coordinate axes such
that the direction of propagation, parallel to the wave vector K, is in the
^-direction, and such that Ao is in the ^-direction. Then, from (8.20)
and (9.1),
Hi — (eAo/m)px sin(if z — u)t)

= (eA0/2mi)px[e[iKz-^ - e-K**-"*)]. (9.4)
We suppose that the field is weak enough to allow us to use the time-
dependent perturbation theory developed in the last chapter. Imagine
that the field is switched on at time t = 0 and is switched off again at
some time T. We suppose that the electron was initially in the eigenstate
<!>i of Ho, and calculate the probability that a measurement, after the field
has been switched off, will find that a transition has occurred to some
other eigenstate $ / of Ho. This transition probability (see (7.20) and
(7.21)) is the squared modulus of
10
}, (9.5)
JO
where
ufi = (Ef - Ei)/h,

(9.6)
Resonant absorption and stimulated emission 95
If Ef > Ei, the transition must have occurred through the absorption of
energy from the electromagnetic field, and the more complete treatment in
which the field is quantised shows that, in fact, the electron has absorbed
a single photon from the field. If Ef < E^ the electron has emitted a
single photon. In either case, we see that the transition probability is
proportional to the square of the intensity of the field. In the case of
absorption, it is easy to understand intuitively that the probability should
increase with the field intensity, but the result is perhaps surprising for
the case of emission: the presence of the field stimulates the electron to
emit a photon.
The integral (9.5) gives
aif = - ^ ^ - . (9.7)
a Ufi — UJ ujfi + UJ
Suppose that the angular frequency u of the external field is chosen such
that hu is close to the difference between the energies E{ and Ef of the
initial and final states. Then the denominator of one of the two terms in
(9.7) is small, and so u) has been tuned to give a resonance in the transition
probability. The term with the small denominator then dominates over the
other term, which may be neglected in comparison with it. In the case
of absorption, where Ef > E^ so that cj/i > 0, the first term becomes
dominant and the transition probability is
K / l
h2 {ufi-u)* • (9 8)
'
In the case of emission, where Ufi < 0, the second term in (9.7) leads to
an analogous expression.
We have already encountered a function very similar to (9.8): see (7.25)
and figure 7.1. Unless u is within a distance of about 2TT/T from a;/*, the
transition probability is very small. In practice, Ufi is often in the optical-
frequency range, so that when T is equal to one second, the transition
occurs readily only if u differs from Ufi by less than one part in 1014.
(However, this estimate ignores the fact that excited levels of atoms have a
finite lifetime and so, because of the energy uncertainty principle which we
described in chapter 7, their energy is not precisely determined. Allowing
for this typically increases the frequency spread to about one part in 105.)
Let us consider what happens when u;, the frequency of the incident
radiation, is equal to CJ/J, the frequency corresponding to the energy
difference between the initial and final states. Then (9.8) becomes
\aif\2 = lu^^T2/^2. Recall that \aif\2 is the probability that a transi-
tion has taken place by time T. In macroscopic systems, the transition
rate is, however, observed to be constant, that is \a>if\2 is linear in T.
We can reconcile these facts if we consider not a transition to a single
final state / , but rather to a narrow range of states. The probability P
of a transition from state i into any state whose energy is in the range
{Ef,Ef + AEf) is given by
Here g(Ef) is the density of states at energy Ef, that is the number of
states in the range (Ef,Ef + AEf) is g(Ef)AEf. The integration in (9.9)
is over the energy interval AEf. However, as we have already remarked,
for large T the transition probability (9.8) is extremely small unless Ufi
is almost equal to a;, so that it is an excellent approximation to replace
g(Ef) and | ^ | 2 with their values at E = (Ei + TIUJ) and to extend the
integration to ±oo. Making the change of variable a = |(cj/i — o;)T, (9.9)
can then be rewritten as
(9.10)
Dividing by T, we obtain the transition rate
I\_+/ = — luji \2g(Ei + fiu)). (9-11)
This is known as Fermi's golden rule.
Electric dipole transitions

In order to evaluate the transition probability (9.8), it is necessary to
calculate u^K Consider a hydrogen atom for definiteness, and take the
origin at the nucleus of the atom. Then if ^ and c/>f are electron wave
functions corresponding to low-lying atomic levels, they are small unless
Kr «: 1 (see problem 9.1). Hence it is a good approximation to replace
the exponential in the expression (9.6) for uffi by unity:
(9.12a)
The ammonia molecule 97
Now, according to the commutation relations (4.28),

[x,px]=ih, [x,py] = [x,pz] = 0. (9.13)
Pre-multiply and post-multiply the first of these relations in turn by px;
this gives two relations which when added together yield
[x,px] = 2ihpx.
Similarly, the other two relations in (9.13) give
Since Ho = (p2x + p* +p2z)/2m + V(r), and V(r) trivially commutes with

x, we thus have
px = (m/ih)[x,H0] = {m/ih){xH0 - Hox). (9.14).
Inserting this relation into (9.12a), and using the fact that fc and </>/ are
eigenfunctions of #o, we obtain
u% = [Ao(Ei - Ef)/2h]{(j>f, -excfn). (9.126)
Now, —ex is the rr-component of the electric dipole moment of the atom.
For this reason, the transitions corresponding to the approximation we
have made, that e±lKz « 1, are known as electric dipole transitions. If
we had not replaced the exponential in (9.6) by unity, we should have
obtained additional terms corresponding to electric quadrupole and higher
multipole transitions, and also terms corresponding to magnetic dipole
and multipole moments. The magnetic terms may be understood from
the fact that, crudely speaking, the electron in orbit round the nucleus is
equivalent to an electric current loop, which gives the atom a magnetic
moment that interacts with the electromagnetic field.
The transition probabilities are greatest for the electric dipole transitions.
However, for many pairs of levels E{ and Ej the quantity (</>/,— exfc)
vanishes: compare the result (7.24) for the case of a particle in a box. This
leads to certain selection rules: radiative transitions readily occur only
between certain pairs of atomic levels. It may be shown that spontaneoues
emission is subject to exactly the same selection rules. In consequence,
the corresponding lines in the emission or absorption spectrum of the atom
are much more pronounced than those that are associated with transitions
that can occur only via a magnetic or higher electric multipole mechanism.
(a) (b)
Figure 9.1. Two configurations of the ammonia molecule (neither is a stationary state).
S is the spin vector and /X is the electric dipole moment.
The ammonia molecule

The ammonia molecule is a bound state of an atom of nitrogen together
with three hydrogen atoms (figure 9.1). Thus it is a very complicated
system: there are four nuclei and, when the molecule has no net charge,
ten electrons. A full calculation of the energy levels of this fourteen-body
system is much too difficult to undertake, but the system has an important
property which we can understand by making an analysis similar to that
given in chapter 6, where we described the ionised hydrogen molecule.
Consider the ammonia molecule in a coordinate frame in which its centre
of mass is at rest. To specify fully the states of motion of the four-
teen constituent particles, a large number of parameters would be needed.
However, only one of these need concern us here, namely the total angu-
lar momentum. It turns out that in the ground state the total angular
momentum of the molecule is non-zero: it has non-zero spin arising from
the sum of the orbital-angular-momentum and spin vectors of its separate
constituent particles. In the ground state, the spin vector 5 is normal to
the plane defined by the three hydrogen nuclei (that is, the expectation
values of the spin components in directions parallel to the plane are zero).
The direction in which S points distinguishes the two sides of this plane,
so that the two configurations shown in figure 9.1, in which the nitrogen
nucleus is positioned on opposite sides of the plane, correspond to differ-
ent quantum-mechanical states. If the molecule had no spin, figure 9.1(6)
would be obtained from figure 9.1 (a) simply by turning the molecule over,
and so would not correspond to a different internal state.
The ammonia maser 99
Let the wave functions fa and $2 describe configurations of the molecule

corresponding to figure 9.1 (a) and figure 9.1(6), respectively. These wave
functions differ only in that they make the expectation value of the coordi-
nate of the nitrogen nucleus appear on different sides of the plane defined
by the expectation values of the coordinates of the hydrogen nuclei. Be-
cause of the symmetry of the system,
#11 = # 2 2 and H12 = H2i, (9.15)
where H^ = (
As in the discussion of the hydrogen molecule in chapter 6, the wave func-
tions (j>i and (fi2 do not describe stationary states. Their significance is
merely that they represent configurations for which (9.15) applies. Be-
cause of the tunnel effect, transitions occur between these states, and as
in chapter 6 the stationary states are
V>± = ( 0 i ± f c W 2 (9.16)
(see (6.3) and (6.6); the l / \ / 2 is just a normalisation factor). The cor-
responding energies E± are to be calculated from a very complicated
four teen-particle Hamiltonian H and, unlike in chapter 6, the wave func-
tions fa and <f>2 are functions of many coordinates. The calculation is not
possible in practice.
Without making a detailed calculation, it may however be shown that the
ground state of the molecule corresponds to the wave function ?/>+. It is
found experimentally that the level separation
E_ - £ + = hu0 (9.17)
corresponds to a frequency of 24000 MHz. This is the frequency of the

photons that are emitted or absorbed in a transition between the two lev-
els. The corresponding wavelength, 12.5 mm, is in the microwave range.
This level separation is very much less than the energy needed to excite
the state of next-lowest energy. Thus, although the molecule has very
many other stationary states, corresponding to, for example, rotational
excitations of the whole molecule or excitations of its separate atoms, the
molecule is unlikely to be found in one of these states unless the tem-
perature is so high that there are frequent and violent collisions between
molecules. If this were not the case, an analysis based on an expansion
of the wave function similar to (6.3), in terms of </>i and (j)2 only, would
not have been a good approximation.
The ammonia maser

The word maser is an acronym for microwave amplification by stimulated
emission of radiation. The first maser was constructed in 1954, using
ammonia gas.
At normal temperatures, very nearly equal numbers of molecules of the
gas are found in each of the two lowest levels E+ and E-. The first
stage in the operation of the maser is to separate, at least partially, those
molecules in the lower level E+ from those in the higher level E-. The
component enriched with molecules in the higher level is then squirted
through a resonant cavity, tuned to angular frequency u;o, so that some of
the molecules emit photons of energy TILOQ and emerge from the cavity in
the lower level E+. The emission process is stimulated emission, which
we described in the last chapter, so that the number of photons emitted
into the cavity by the gas increases if the intensity of the radiation already
present in the cavity is increased. Hence the maser acts as an amplifier
of any external radiation of frequency OJQ that is fed into the cavity. (The
maser can, to some extent, be tuned; that is, the frequency LOQ at which
it operates can be varied. To achieve this, one has to vary the energy gap
(E- —£?+), for example by placing the maser in a magnetic field.)
We now consider in some detail the two stages in the operation of the
maser.
The separation
The separation is achieved by passing the gas through a static electric field
S. This field interacts with the electric dipole moment /JL of the molecule.
This dipole moment exists because the molecule is not spherically sym-
metrical. In fact, the electrons of the molecule tend to be found closer
to the nitrogen nucleus than to the plane containing the hydrogen nuclei,
so that the two states fa and fa have equal and opposite dipole moment,
normal to the plane of the hydrogen nuclei, as shown in figure 9.1. Thus,
if /i 0 is ^ e expectation value of the dipole moment in the state fa,
H22 = Eo + vo'£, (9.18)
where Eo is the value of Hn or H22 in the absence of the electric field.

We shall suppose that £ is small compared with the internal field of the
molecule; it would need a very strong external field £ for this not to be
the case. This means that small changes in the geometry of the molecule
The ammonia maser 101
due to the applied field £ can be neglected. Hence not only the dipole
moment /io can be taken to be unchanged by £, but also H12 and #21,
which, as in (6.13), depend mainly on the overlap if 12 = (</>i, ^2) between
the wave functions <f>i and fa.
In practice, K\2 < 1, so that in the expressions (6.6) for E± we may
neglect Ku compared with 1. According to (9.17), we define the difference
between E+ and E- to be huo, so that (6.6) gives
H12 = -±huj0. (9.19)
If we now use the values (9.18) for Hn and #22 in equations (6.4), elim-
inating c\ and C2 from them we find that in the presence of the electric
field £ the energies E± are changed to
(9.20)
Thus, whatever the direction of £ relative to that of /LIQ, the energy of

the upper level E- has been increased a little by the presence of the
electric field, and the energy of the lower level E+ has been decreased.
The stationary-state wave functions also will be changed a little from the
ij)± of (9.16); however, since in practice \£\ is small on the atomic scale,
this change is small.
Suppose now that the magnitude of £ varies with position. In practice,
£ will still be constant on the atomic scale of length, so that we need not
take account of its variation in the quantum-mechanical Hamiltonian, and
it is sufficient to account for it classically. That is, the two components
of the gas will experience forces
-VE'± = ±(l/huo)V(tio • Sf (9.21)
in opposite directions. So if the gas is squirted through a region where

£ varies across the jet, the two components are deflected in opposite
directions (figure 9.2). The component that contains the molecules with
wave function ?/>-, in the upper of the two energy levels, is made to pass
into the resonant cavity.
Output
Input
Ammonia
jet Separator
Collimator Resonant
cavity
Figure 9.2. Schematic diagram of the ammonia maser.
Stimulated emission in a radiation cavity

Consider two parallel conducting planes, z = 0 and z = a, with empty
space between. A solution to Maxwell's equations (listed in appendix D)
in the space between the planes is
£ = (£Q sin Kz cos art, 0, 0),

B = (0, (£0/c) cos Kz smut, 0). (9.22)
where
Because the planes z = 0 and z = a are conductors, we must have

£x — £y — 0 o n them. This requires that
K = riir/a, n integer. (9.23)
The need to satisfy the boundary conditions at z — 0 and z — a has led

us to consider a standing wave (9.22), instead of a travelling wave as in
(9.2). However, the analysis of the interaction of the radiation (9.22) with
the ammonia molecule can be carried through in a way closely similar to
that described earlier.
Suppose that a jet of ammonia gas enters the cavity with each of the
molecules initially in the ^_ state. The presence of an electromagnetic
field (9.22) will stimulate the molecules to emit photons and so make
transitions downwards in energy into the i/>+ state. However, unless u
is tuned so as to be very close to U)Q, the probability of such transitions
occurring will be very small. Because of (9.23) and the relation UJ = cK,
this tuning may be achieved by suitable choice of the distance a between
Population inversion
the conducting walls of the cavity. When u is tuned in this way, the
emission occurs fairly readily, and the photon that is emitted adds to
the intensity of the electromagnetic field (9.22), so that the probability of
further molecules undergoing stimulated emission is then enhanced. Each
transition further reinforces the field and so increases the probability of
yet more transitions occurring, so that the field builds up rapidly. That
is, the original electromagnetic field (9.22) rapidly becomes amplified.
The standing wave (9.22) corresponds to a collection of photons that
bounce back and forth between the two conducting planes. The momen-
tum vector of each photon is normal to the planes, and when a photon hits
one of the planes its momentum is reversed. What our simple discussion
does not show is that each extra photon that results from the stimulated
emission actually has the same wave vector, and therefore from (2.2) the
same momentum, as the other photons that are already present in the
cavity. It therefore exactly reinforces the existing field. To show this,
one must consider in more detail the equivalence between electromagnetic
fields and systems of photons, which is beyond the scope of this book.
The speed with which the ammonia gas passes through the cavity deter-
mines the time that the molecules spend in the resonant system. This is
an important factor in the design of the maser. The reason for this is
that once a molecule has undergone stimulated emission and so made a
downward transition to the ground state ?/>+, until it has left the cavity it
is now liable to absorb radiation and so make a transition back to the state
V>-. If this is allowed to occur, it decreases the intensity of the field in the
cavity instead of amplifying it. This problem is avoided by ensuring that
the molecules which have dropped into the ground state escape through
the exit hole before an appreciable probability of reabsorption occurs.
Population inversion
Because a molecule in the ground state -0+ is liable to make a transition
that absorbs energy from the field instead of adding to its intensity, the
initial separation of the gas into two components is a vital stage in the
operation of the maser. If an unseparated, equal mixture of the two com-
ponents ?/>+ and ip- enters the cavity, the net effect is zero: there are
as many upward transitions that absorb photons as there are downward
transitions that emit them.
Even a small quantity of gas contains a very large number of molecules
— about 1025 kg" 1 . The molecules continually collide with one another
and so exchange energy. It is completely impracticable to try to follow the
effect of the collisions on the state of each individual molecule. Instead, one
resorts to statistical methods, and it turns out that this is quite sufficient
to give an accurate description of the macroscopic properties of the gas.
There is little interest in determining at any instant the state of a given
molecule; what matters is the configuration of the system as a whole. That
is, one needs to know how many molecules at any instant may be expected
to be in a given quantum-mechanical state, rather than which particular
molecules are in that state.
The molecules continually collide with each other and so change their
states. The collisions become more frequent and more violent as the tem-
perature of the gas increases, since by definition the average kinetic energy
of the molecules is proportional to the temperature (measured on the ab-
solute, or Kelvin, scale). A remarkable result of statistical physics is that
it predicts the average distribution of the states of the molecules without
knowing anything in detail about what happens in the collisions. The
demonstration of this may be found in any book on statistical physics;
here we just quote the results.
Consider any system of particles that is in dynamical equilibrium, so
that in particular no heat flows into or out of the system. Suppose that
each particle may be found in any one of a number of different states.
Let these have energies £?», i — 1,2,..., with corresponding degeneracies
gi (so that the level E{ is associated with gi independent states). Then
according to classical statistical mechanics, the average number U{ of par-
ticles expected to be in the ith level at any given instant is given by the
Maxwell-Boltzmann distribution:
m = Cgie~*i/KBl. (9.24)
Here T is the absolute temperature and ks is a universal constant, known

as Boltzmann's constant. The constant C may be calculated in terms of
the total number N = ]T^ n^ of particles:
C =N V>e-*'/*Br . (9.25)
Evidently C varies with the temperature T.

The Maxwell-Boltzmann distribution is a formula of classical mechan-
ics. In certain circumstances it is important to take account of quantum-
mechanical effects, but at room temperatures (9.24) adequately describes
the distribution of states of the molecules in the gas. Because the energy
The laser 105
difference between the levels E+ and £L of ammonia is small, we deduce

from (9.24) that in a quantity of gas that is in dynamical equilibrium
the numbers of molecules in these two levels are very nearly equal, the
population of the lower level E+ being only slightly larger than that of
the higher level E-. The separation process achieves a population inver-
sion in the component of the gas that is injected into the cavity. In this
component the majority of the molecules are in the higher level, and the
gas is not in dynamical equilibrium. If it is left to return to equilibrium,
intermolecular collisions ensure a rapid return to the Maxwell-Boltzmann
distribution.
From (9.24) we may see also the truth of our earlier assertion that, unless
the temperature T is high, the populations of the higher-energy levels are
rather smaller than those of the lowest-lying levels E+ and E-.
The laser
The principles underlying the operation of the laser are much the same as
those of the maser, the basic difference being that the photons produced
correspond to the optical range of frequencies. In lasers, and nowadays
also in masers, the material that emits the photons is kept stationary
within the radiation cavity, instead of being squirted through the cavity. If
the amplifying material is solid, the parallel mirrors that form the radiation
cavity may be the end faces of the material itself. A resonant cavity
of this form was used in the first successful laser, whose operation was
announced in 1960. This laser had as its active medium a cylinder of
synthetic ruby, which is an aluminium oxide crystal containing a small
percentage of chromium. The end faces of the ruby cylinder were rendered
accurately parallel to each other and then both were silvered, one fully
and the other partially. Population inversion was achieved by a process
known as optical pumping, which uses a helical flashtube wrapped around
the ruby cylinder. The flashtube emits white light, that is, light with a
wide range of frequencies.
The active ingredient of the ruby is the chromium, whose (simplified)
energy-level diagram is shown in figure 9.3. When the flashtube is fired,
the ruby crystal is flooded with photons, most of which dissipate their en-
ergy as heat. However, some in the blue and green ranges of the spectrum
are absorbed by the chromium and cause transitions from the ground state
to one of the two energy bands shown in the figure. As we have explained,
the relative probabilities of transitions between different levels of a sys-
tem are governed by selection rules. In the case of the energy bands it
Energy
Non-radiative
transitions
Resonant Metastable state

absorption
of green and Stimulated
emission
blue light
Ground state
Figure 9.3. Energy levels in a ruby crystal.
turns out that the preferred decay is to a metastable state, also indicated
in figure 9.3. (This decay is a non-radiative transition: the energy is not
given up by the emission of a photon, but rather it appears in the form of
phonons, which are quanta of vibration of the other atoms in the crystal
lattice about their mean positions.) So if the flashtube is fired with high
power, a large number of chromium ions are driven into the metastable
state: a large population inversion occurs.
The metastable state decays to the ground state by spontaneous emission
of photons. (Its lifetime is rather long on the atomic scale: a few millisec-
onds, compared with the 10~8 s typical of spontaneous transitions when
dipole transitions are allowed.) Most of the photons emitted by sponta-
neous emission pass out of the crystal and are lost; but just a few move in
the right direction to be reflected back and forth between the two silvered
end faces. In this way, a standing-wave electromagnetic field of the form
(9.22) begins to be set up. This field stimulates the emission of further
photons, so that the intensity of the field (9.22) builds up very rapidly as
the metastable state depopulates. The partially silvered end of the crystal
allows the leakage of an intense pulse of laser light. This is the output of
the pulsed ruby laser.
Laser technology has developed very rapidly since 1960, and there are
now many types of laser, for example the semiconductor lasers described
in chapter 12.
The light emitted by a laser has two important properties. First, it can
be very intense: the power per unit area of cross-section in a laser beam
can be at least 109 times that which may be obtained from a conventional
Holography 107
Figure 9.4. Light reaching a point P directly from a coherent source S and reflected
from a point object O.
light source. Secondly, laser light is not only highly monochromatic, but
also highly coherent. In a conventional light source, the atoms decay in
an uncorrelated fashion and so the phases of the electromagnetic wave
packets associated with the different photons are random. In a laser, the
stimulated emission is precisely in phase with the electromagnetic field
that causes it. There are some variations in phase of the field as a whole,
mainly caused by fluctuations in the distance between the mirrors resulting
from thermal and mechanical disturbances, but these are extremely small.
In a laser beam, the phase difference between the light at points along the
beam separated by as much as 100 km is likely to be unchanging with
time.
Holography
The high intensity of laser light has obvious applications, such as to cutting
and welding in engineering or medicine. It also makes possible the study
of fundamental phenomena, such as non-linear optics, arising from the
changes in the properties of materials when they are subjected to very
strong electromagnetic fields. An important application of the high degree
of coherence of laser light is to holography. We digress briefly from our
main subject of quantum physics in order to describe this technique of
three-dimensional photography.
Imagine first a coherent, monochromatic point source of light S at a dis-
tance R from a point object O (figure 9.4). Light that comes from S in the
direction of O will be scattered by O, and hence O behaves effectively as a
second point source. The two sources S and O will be mutually coherent;
that is, the phase difference between the light from them will be constant
in time at any point P in the space surrounding them. Hence, an inter-
ference pattern is established: if the two components of the light reaching
P are exactly out of phase, then there is darkness at P, while if they are
in phase, then P is a point of maximum brightness. The condition that
(a)
Photographic
plate
(b)
T
Figure 9.5. (a) Interference fringes corresponding to figure 9.4, with a photographic
plate on which a hologram is recorded. (6) Reconstruction of a virtual image O1 and
a real image O", using a light source Sf.
there be destructive interference at P, that is, darkness, is
K{rx -R-r2) + a = (2m + 1)TT, (9.26a)
where
= 0,±l,±2,.
Here K is the wave number of the light (so that its wavelength is 2TT/K),
v\ and r 2 are the distances of P from S and O as in figure 9.4, and a is a
possible phase change induced by the reflection of the light from S on the
object O. Similarly, the condition that there be constructive interference
Holography 109
Mirror
Photographic
Laser plate
Mirror
Figure 9.6. Set-up used to produce a hologram of a three-dimensional object,
at P is
K(rx - R - r2) + a = 2ra'7r, (9.266)
where
The points P that satisfy either of the two conditions (9.26) lie on a sys-
tem of hyperboloids of revolution having O and S as foci, as is shown
in figure 9.5(a). In this figure we show also the cross-section of a pho-
tographic plate that has been placed so as to intersect the interference
pattern. This plate records a hologram, which is a thin section of the
interference pattern. Development of the plate causes silver grains to be
deposited in the emulsion, with a high density at points P correspond-
ing to constructive interference, and low density in between at points of
destructive interference.
To view the holographic image, one illuminates the hologram with a source
S", having the same position relative to it as the original source S and emit-
ting light of the same colour. Unlike 5, the source S' does not have to be
coherent. The light from S' that falls on the hologram is diffracted by the
light and dark fringes recorded on it. Diffraction occurs in two directions,
one such that the rays emerging on the other side of the hologram appear
to have originated from the point O' corresponding to the position of the
original object O, and the other such that the rays converge on the point
O" that is the reflection of O' in the photographic plate (see figure 9.56).
Thus there is a virtual image at O' and a real image at O". To view the
virtual image, one looks at the plate from behind but focuses the eye on O'
because this is the point from which the rays appear to have diverged; this
is why holography produces a three-dimensional effect. The real image is
made apparent by placing a screen behind the plate, passing through the
point O", or by placing the eye beyond the point O" and focussing it on
that point (see figure 9.5(6)).
In practice, because a suitably coherent point source of light S is not, in
fact, available, a laser beam is used to produce the hologram. The beam is
spread out by a lens and split into two parts. One part is reflected directly
onto the photographic plate, and the other is reflected onto the plate by
the object. The set-up needed to produce a hologram of an extended
three-dimensional object is shown in figure 9.6. The interference fringes
now have a much more complicated geometry, but the principles involved
are the same as for a point object.
Problems
9.1 The wave number K corresponds to the radiation associated with
transitions between the two lowest energy levels of the hydrogen atom
(see (5.17)). Verify that the ground-state wave function (5.18) is small
for all values of r that do not satisfy Kr <C 1.
9.2 Assuming that the states fa and fa of the ammonia molecule are eigen-
states of the electric-dipole-moment operator, show that the stationary
states ip± have zero dipole moment.
9.3 For a black body, the power emitted per unit area in the wavelength
interval from A to A + dA is
(2nhc2/\s)[d\/(ehc'Xk*T - 1)].
A helium-neon laser operating at a wavelength of 633 nm with an

output power of 1 mW has a linewidth of 1 kHz and produces a beam
of cross-sectional area 10~6 m2. What would be the temperature of
the black body that emitted photons at the same rate, in the same
frequency interval and from the same area?
9.4 The beam from a laser is usually extremely well collimated. By ap-
plying the uncertainty principle to estimate the transverse momentum
of the photons, show that a helium-neon laser beam which initially
has diameter 1 mm can be collimated to within an angle of about
3 x 10~3 degrees.
How small would the filament of an ordinary household torch have to
be to achieve the same collimation?
Holography 111
9.5 A gas of non-interacting particles of mass m is contained in

a cube of side L and obeys the Maxwell-Boltzmann distri-
bution. Write down the formula for the probability that a
given particle is in the state corresponding to wave function
(8/L 3 ) 1 / 2 sin(q7Tx/L) sin(r7ry/L) s'm(s7r z/L). Show that as L —> oo the
probability that the ^-component of the velocity of a given particle lies
between u and u + du is
9.6 In the previous question, a membrane of area A is embedded in one

wall of the cube. It is permeable to all particles of the gas whose
normal component of velocity is greater than u$. Show that the rate
at which particles escape from the box is
(NA/L3){kBT/2<irm)1/2 e-
m
9.7 In quantum statistical mechanics for a system of electrons, the

Maxwell-Boltzmann distribution is replaced by the Fermi-Dirac dis-
tribution
where the Fermi energy ( is independent of E{. What form does this
distribution take at T = 0, and why?
Notice that when E{ » £ the two distributions are approximately the
same.
10
Band structure of crystals
Electrons in crystals
A crystal consists of a collection of atoms arranged in a regular array,
the spacing between atoms being of the same order of magnitude as the
dimensions of the atoms. Each atom is more or less anchored to one point,
called its site in the lattice, by the electrostatic forces produced by all the
other atoms. We shall not find it necessary here to discuss the details of
how this comes about; nor shall we consider the various patterns in which
the atoms can be arranged. It will be sufficient to remember the essential
feature that the structure of the crystal is periodic in space.
We have seen in chapter 5 that the energy of an electron bound to an atom
is restricted to certain discrete values. Imagine that we can assemble a
crystal of identical atoms whose spacing L can be altered at will. If L is
large enough, the motion of an electron in one of the atoms will be affected
to a negligible extent by the electrons and nuclei of the other atoms. Each
atom then behaves as if it were isolated, with its electrons in discrete
bound states. In figure 10.1 (a) we have drawn a schematic diagram of
the potential V(r) in which an electron moves in this situation. Suppose
that the spacing L is now reduced (figure 10.1(6)). The potential V(r)
in the neighbourhood of a given atom is now affected by the presence
of the nuclei and electrons of the other atoms, particularly those that are
closest. An exact calculation of this situation is a complicated many-body
problem and is not tractable. To a reasonable approximation, however,
the motion of a given individual electron in a crystal can be treated as a
single-particle problem, with the potential determined by the electric field
produced by all the other particles in the crystal.
We shall show in this chapter that, as L is reduced, each original discrete
atomic energy level spreads out into a band of closely spaced levels (fig-
ure 10.2). These bands are separated by energy gaps that are forbidden
to the electrons. The band structure is all-important in determining the
properties of electrons in crystals.
In a crystal, an electron is not bound to any one particular atom, but
rather to the crystal as a whole. If it is not to escape from the crystal,
112
Band structure 113
V(r)
(a)
V(r)
(b)
Figure 10.1. Schematic diagram of the potential in a crystal: (a) when the atoms are
widely separated; (b) when they are closer together.
its energy must be negative (where the zero of energy is defined to be the
potential of the free space outside the crystal). If the crystal temperature
is high enough, an electron can gain sufficient energy by collisions with
the vibrating nuclei to escape from the crystal; this is called thermionic
emission. However, we shall confine the discussion to the electrons that
remain bound in the crystal. An electron in one of the low-lying bands,
such that its energy is less than F max (see figure 10.1(6)), passes from
one atom to another by tunnelling through the potential barriers between
the nuclei. We shall find that there are also energy bands between Vmax
and 0; electrons in these bands are readily transmitted over the potential
barriers.
To explore the band structure, we consider mainly a one-dimensional
model. This provides a good guide to electron behaviour in real crys-
tals. Certain aspects, however, require a discussion of electron motion in
more than one dimension; this is necessary, for example, in order to under-
stand why divalent metals such as calcium and magnesium are reasonably
good electrical conductors.
114 10 Band structure of crystals
E
A
Figure 10.2. The band structure that develops when the atomic separation L is de-
creased.
Band structure
We do not need to know in detail how the potential varies with position
in the crystal. The most important feature is its periodicity. To have
a readily solvable problem, we consider an electron in a one-dimensional
periodic square-wave potential (figure 10.3). This model is sufficiently
realistic for many purposes. For convenience, we define the zero of the
potential to coincide with the top of the wells, rather than with the value
of the potential in free space outside the crystal.
Consider first an electron with energy E such that —U<E<0. The
Schrodinger equation is
in the wells,
between the wells, (10.1)
Band structure 115
V(x)
-U
Figure 10.3. One-dimensional periodic square-well potential.
or
„ _ r a2i> between the wells, (10.2)
in the wells,
where
a2 = -2mE/hz, (32 = (2m/hz){E + U).
The procedure for solving (10.2) is suggested by a theorem due to Bloch,
which is derived in appendix E. One looks for solutions of the form
(10.3)
where k is real and Uk{x) is periodic, with the same period L as the
period of the crystal lattice. Bloch's theorem states that solutions of the
type (10.3) form a complete set, at least in the limit where the total length
of the crystal is taken to be infinite. Evidently (10.3) represents a wave of
wavelength 2?r/A;, whose amplitude Uk{x) varies across each atomic site,
but is the same from one atom to the next. The electron belongs not to
one atom, but has equal chance of being found in the vicinity of any of
them.
The solution to each of the two differential equations (10.2) consists of
simple exponentials. We write
a - L < x < 0, (10.4)

\CeQX+£»e-QI 0 < x < a.
Then from (10.3)
Aei({3-k)x
a - L < x < 0, (10.5)
•{ Ce(a-ik)x
0 < x < a,
and because we require uk(x) to be periodic, (10.5) determines uk{x), and

so also il>(x), everywhere. In particular, for a < x < L we replace x by
(x — L) in the first equation of (10.5):
At both edges of each well we require if) and rj)1 to be continuous, or equiv-
alently u and u' to be continuous. The periodicity of Uk{x) guarantees
that if these conditions are satisfied at the edges of just one well, then
they are satisfied for every well. This, of course, is the reason for looking
for a solution of the Bloch type (10.3). Imposing the conditions at x = 0
on the two functions in (10.5) we have
A + B =C + D,
i{p - k)A - i(/3 + k)B = {a - ik)C - (a + ik)D. (10.7)
Prom the conditions applied at x = a to the second function in (10.5) and

to (10.6),
i(P - k)A e-W~k)b - i(p + k)B el

= (a - ik)Ce^-[k)a -(a + ik)Deâ+ik)a, (10.8)
where
b = L-a. (10.9)
Equations (10.7) and (10.8) are four homogeneous equations in four un-
knowns, A, B, C and D. In order that they have a non-trivial solution the
determinant of their coefficients must vanish. When this 4 x 4 determinant
is expanded it results in the condition
[(a 2 - /32)/2a/3]smhaa sin/3b + cosh aa cos pb = cos kL. (10.10)
Although we have derived this equation for the case where E lies in the
range -U < E < 0, it is valid also when E > 0; then a 2 , defined in (10.2),
is negative and in (10.10) we write a = \al and so obtain
[-{a12 + p2)/2a'P]s'ma'a sin pb + cos a'a cos pb = coskL. (10.11)
According to (10.2), when a,L and U are fixed, the quantities a (or a1)
and p are functions of E, and so the left-hand sides of (10.10) and (10.11)
Number of levels in a band 117
Gap Gap
Figure 10.4. Plot of the function f(E) defined in (10.12).
correspond to a complicated function of E. We write these equations

together as
f{E) = cos kL (10.12)
and in figure 10.4 we plot the function f(E) against E for a typical set of
values of a,L and U. Because k is real, |cosA;I/| < 1. Thus those values
of E for which \f(E)\ > 1 are not accessible; that is, the allowed values of
E fall into bands determined by the condition |/(i£)| < 1, with forbidden
energy gaps between the bands.
Notice that, according to figure 10.4, the bands corresponding to the
lowest-lying levels are narrowest: the levels associated with the electrons
that are most tightly bound to the atoms are affected least by the presence
of the other atoms. Notice also that the band structure persists even for
positive values of E, that is, above the top of the potential wells. The
energy gaps decrease in width with increasing E, but they can still be of
appreciable width even when E is so large that the electron nearly has
enough energy to escape from the crystal into the surrounding free space.
As we shall describe below, when the level structure is calculated in more
than one dimension, an alternative possibility is that neighbouring energy
bands overlap, and this is found to be the case in certain types of crystal.
Number of levels in a band

Except for the smallest microcrystals, the properties of the crystal and
the disposition of the allowed energy levels do not depend critically on the
boundary conditions applied at its surfaces. The simplest choice of bound-

ary condition is to require that the electrons be strictly confined within
the crystal, that is ip = 0 on the faces, or in the linear model ip = 0 at each
end of the chain of atoms. However, this has two disadvantages. First,
the boundary conditions can then not be satisfied by the simple Bloch
wave functions (10.3); one must instead take combinations of Bloch waves
to achieve a standing wave (a travelling wave of the type (10.3) moving
in the positive-:*; direction is reflected at the crystal face, so generating
another travelling wave moving in the negative-a; direction). Secondly, a
standing-wave solution is often unsatisfactory because transport of energy
or charge is far more conveniently expressed in terms of travelling waves.
Since the crystal surface has little effect on the bulk properties of the crys-
tal, one may as well choose boundary conditions that more or less eliminate
it. This is achieved with periodic boundary conditions. The wave function
ip is required to take the same value at corresponding points on opposite
faces of the crystal, so that an electron is not reflected at the faces but
leaves the crystal and simultaneously reenters it at the corresponding point
on the opposite face. Both types of boundary condition result in the same
number of levels within each band, but they give very slightly different
energies to corresponding levels. For all but the smallest microcrystals
the levels within a band are so closely spaced that they almost form a
continuum and so this difference has no consequence.
In the linear model the periodic boundary condition is equivalent to bend-
ing the chain of atoms round to form a closed loop. Choosing the coordi-
nates such that the ends of the chain are x = 0 and x = JVX, we require
^(0) = ip(NL). From (10.3) we see that this requires that eikNL = 1, so
that the possible values of k are discrete:
k = 2nn/NL, n = 0, ± 1 , ± 2 , . . . . (10.13)
Furthermore, we find from (10.3), (10.5), (10.6), (10.10) and (10.11) that
both if) and the corresponding energy E are unchanged if k is either in-
creased or decreased by an integral multiple of 2n/L. Thus we may,
without losing any of the states, confine k to any chosen interval of length
2n/L, for example
-n/L <k< n/L. (10.14)
As k takes each of the allowed values (10.13) that fall within this interval,
we pass through every one of the allowed levels in each band. Thus each
band has just N allowed levels, where N is the total number of atoms in
the crystal.
Band overlap 119
Figure 10.5. One-dimensional model for a crystal having two atoms per unit cell.
This result is not unexpected. When the atoms are very far apart each
band consists of just a single level, the energy level of the solitary atom.
However, the single level corresponds to N different states, because the
electron can be attached to any one of the N different atoms. As the
atoms are brought together, the AT-fold degenerate levels spread out into
bands. This is a generalisation of the two-atom problem that we discussed
in chapter 6.
The property that the number of levels in each band is equal to the number
of atoms in the crystal is special to the simplest crystal structure. In many
cases the crystal lattice consists not of a regularly spaced array of single
atoms, but rather of an array of unit cells that each consist of more than
one atom. The model of figure 10.3 represents a crystal in which the unit
cell consists of a single atom; in figure 10.5 we show a one-dimensional
potential that can be used as a model for a crystal in which the unit cell
consists of two atoms. Generally, the number of levels in each energy band
is equal to the number of unit cells in the crystal.
Band overlap
Equations such as (10.10) and (10.11), which relate k and i£, are called
dispersion relations. We consider now the solution to these equations,
which expresses E as a function of A;. It is not possible to obtain an
explicit expression for this solution; it is necessary to use either graphical
or numerical methods. We discuss the general features of the solution.
A given value of k corresponds to many values of E. The number of
values for which E < 0, corresponding to (10.10), depends on the depth
and width of the atomic potential wells, but, for E > 0, (10.11) always
gives an infinite number of E values for a given k. Because (10.10) and
(10.11) involve k only through cos kL the solutions for E are symmetric
about the origin in k. According to figure 10.4, the edges of the allowed
energy bands occur when cos kL = ± 1 , and if the range of definition of
k is chosen to be (10.14) this corresponds to either k = 0 or k = ±n/L.
Prom these considerations, we deduce that the dispersion curves have the
form drawn in figure 10.6.
k
-n/L n/L
Figure 10.6. Dispersion curves corresponding to figure 10.4.
In the one-dimensional model, gaps always occur between the bands, re-
gardless of how high up in energy the bands are. However, in the real
world of three dimensions the bands may overlap. In three dimensions the
Bloch wave function (10.3) becomes
tl>(r)=eik'ruk(r), (10.15)
so that A; is a vector. The dispersion curve becomes a three-dimensional
surface in four-dimensional (£7, k) space. The properties of the crystal
may be different in different directions, so that if we draw sections of this
surface corresponding to different directions for k the resulting curves
need not be the same. Each such section will have an appearance similar
to figure 10.6, with gaps between the bands, but the positions of these
gaps and their widths may vary with the direction chosen for k. As this
direction is varied, it may be that the lowest position of the top of a given
gap between two bands is lower in energy than the highest position of the
bottom of that gap (figure 10.7). If this happens, the bands overlap and
there is no gap between them: it is possible to get from one to the other by
continuously varying k. This occurs in divalent metals and semimetals,
and has important consequences.
Simple consequences of band structure 121
Figure 10.7. Band overlap.
Simple consequences of band structure

The band structure of a material has a profound effect on its physical
properties, in particular on its electrical conductivity and its optical prop-
erties.
The electrical conductivity takes a wider range of values for different ma-
terials than does any other physical property: at room temperature, the
ratio of the conductivity of the best conductor to that of the best insula-
tor is about 1028. It is conventional to classify a solid as an insulator, a
semiconductor, a semimetal or a metal, partly according to how good a
conductor it is.
We explained in chapter 5 that the Pauli exclusion principle allows at most
one electron to be in each quantum-mechanical state. Since the electron
has two independent spin configurations (the ^-component of its spin has
eigenvalues ± | ) , the N levels of a band can accommodate at most 27V
electrons. At the absolute zero of temperature the electrons will be found
in states of as low an energy as possible; that is, the levels are filled up,
with two electrons in each, starting from the lowest level of the lowest
band.
Suppose that the crystal consists of one type of atom only, each having
charge Ze on its nucleus. Although most elements crystallise into forms
containing one atom per unit cell, a few of interest to us here form crystals
with two atoms per unit cell. In order to keep the discussion general, let
the number of atoms per unit cell be m. Then, if the crystal contains
N unit cells, it has ZmN electrons when it is electrically neutral. We
have said that the maximum number of electrons in any one band is 2N.
Hence at the absolute zero of temperature, T = 0, there are three basically
different ways in which the bands can be populated, distinguished by
whether the number Zm of electrons per unit cell is even or odd, and if it
is even by whether or not there is band overlap. We consider these cases
in turn.
Insulators and semiconductors

In these crystals Zm is even and no band overlap occurs. At T = 0 a
number of bands are completely filled with electrons and the remainder
are empty. The topmost filled band is called the valence band, and the
next band above it, which is empty, is called the conduction band. If the
crystal is placed in a strong electric field, some of the electrons in the
valence band may be excited into the conduction band. However, with an
electric field of ordinary magnitude the electrons do not acquire enough
energy to jump the energy gap between the bands. As all the valence-band
states are already filled, the quantum state of the crystal cannot change
when such a field is applied. Therefore the field does not cause a net
electric current to flow.
At temperature T > 0 some of the valence-band electrons are thermally
excited into the conduction band. In the next chapter we show that the
number of electrons that appear in the conduction band varies rapidly
with the width A of the energy gap and with the temperature T. In an
insulator, A is so large that the electron concentration in the conduction
band, and therefore also the electric current, are negligibly small even
up to the melting temperature of the solid. In a semiconductor, A is
small enough for there to be a significant electron concentration in the
conduction band; this increases as T increases.
In the next chapter we explain how this enables the semiconductor to
carry an electric current. The electrons that are thermally excited into
the conduction band leave unfilled states or holes in the levels they pre-
viously occupied in the valence band (see figure 10.8). If an electric field
is applied, the electrons in the conduction band can fairly easily change
their states by going to nearby unfilled levels in the conduction band; the
quantum state of the crystal changes in such a way that there is a net
drift of electrons. Another contribution to this net drift comes from some
of the valence-band electrons, which can change their states by occupying
the holes in the valence band. Hence the semiconductor carries a current.
Such a semiconductor is known as an intrinsic semiconductor. In chap-
ter 12 we shall explain that the conductivity of an insulator or an intrinsic
Conduction
band
Valence
band
Figure 10.8. Excitation of electrons from the top of the valence band of a semiconductor
into the bottom of the conduction band.
semiconductor may be increased by the presence in the crystal of a small

concentration of certain impurity atoms; the material is then said to be a
doped semiconductor.
Semimetals and divalent metals

Arsenic, antimony and bismuth each have an odd number of electrons
per atom, but crystallise into a structure having two atoms per unit cell.
Since the total number of electrons per unit cell is even, they would be
insulators were it not for the fact that their conduction and valence bands
overlap to a small extent (we recall figure 10.7). This means that even at
T = 0 there are some electrons in the lowest levels of the conduction band,
leaving unfilled those states in the valence band that have higher energy.
These crystals are therefore fairly good conductors; they are known as
semimetals. A similar situation occurs in the divalent metals, of which
calcium and strontium are examples. These have an even number of
electrons per atom and crystallise into a structure with one atom per unit
cell, so that again each has an even number of electrons per unit cell.
Metals
The atoms of the noble metals (copper, silver etc.) and of the alkali
metals (sodium, potassium etc.) have an odd number of electrons and one
atom per unit cell. Hence their valence band is completely filled and their
conduction band half full. The high concentration of conduction-band
electrons ensures that metals are good electrical conductors.
For insulators and semiconductors, the magnitude A of the energy gap be-
tween the valence and conduction bands has an important influence also
on the colour of the pure crystal. Consider, for example, diamond, which
is one of the crystal forms of carbon. This has A = 5.6 eV. A photon
in the visible frequency range has energy in the range 1.7 to 3.5 eV, less
than A, so that an electron in the valence band cannot absorb such a pho-
ton. There are not enough conduction-band electrons to give appreciable
absorption, so all visible light passes through the crystal. Diamond is
therefore colourless. Sulphur, on the other hand, has A = 2.4 eV, so that
its valence-band electrons can absorb photons of frequency u) such that
TIUJ > 2.4 eV and be knocked into the conduction band. Thus the blue
end of the spectrum is absorbed, and the crystal is yellow in appearance.
These remarks apply to pure crystals; as we indicate in chapter 12, the
introduction of impurities can have an important effect.
Problems
10.1 In the free-electron model of a metal the potential is taken to be
constant, independent of position in the crystal. What form do the
Bloch waves take in the one-dimensional version of the model? Sketch
the dispersion curves, analogous to those of figure 10.6. (A more
realistic model is the nearly-free-electron model, which starts from the
free-electron model and then uses perturbation theory to introduce a
small additional potential around each atomic site.)
10.2 A crystal cube has N atoms arrayed in a simple cubic lattice and
its volume V is large. How does (10.14) generalise to this three-
dimensional case? How many allowed fc-vectors are there per unit
volume of fe-space?
If there are Z electrons per atom, what is the largest value of \k\ for
the states occupied at zero temperature?
10.3 (i) As a model for a one-dimensional crystal, take the potential as
Investigate the level structure, using a Bloch wave function.

(ii) Explain how to repeat the calculation for the potential
which is a model for a crystal having two types of atom.

10.4 The tight-binding model for a crystal is as follows. When the electron
is bound to a solitary atom, it has energy Eo and wave function <f>,
and the binding is tight. In the crystal, the atoms are not very close
together. The wave function
clearly satisfies the Bloch condition (E.10) and one can show that it
approximately describes a stationary state. By considering the expec-
tation value of the Hamiltonian if, show that
£?« Eo+ 2cos kl / dx</>*(x)Vo(x)<l)(x-L),
where VQ(X) is the potential that would correspond to a solitary atom

at the origin. (Notice that ip(x), as given, is not normalised.)
Deduce that each atomic level Eo gives a single band of energy levels
in the crystal.
10.5 Show that the number of states in a band is unaffected by whether
periodic boundary conditions are applied, or instead the electrons are
strictly confined within the crystal.
11
Electron motion in crystals
Electron velocity
In the previous chapter we showed that in a perfectly regular crystal the
stationary-state electron wave functions are the Bloch waves
*k(r,t)=uk(r)Sk-r-E>t'h\ (11.1)
This expression is the same as (10.15), except that we have added the
factor e~lEkt/h so as to write the time-dependent wave function. The
values allowed for the components of the vector k are discrete, but if the
crystal is large enough these discrete values are so close together that for
many purposes k may be treated as a continuous variable. As we know
from the discussion in chapter 4, apart from questions of degeneracy the
wave functions corresponding to different values of k are orthogonal, and
we may choose the normalisation
= 6kk,. (11.2a)
Here, the integration is over the volume of the crystal and dkk> is zero
unless the vectors k and k' are the same, in which case it is equal to 1.
From (11.1) and (11.2a),
= <W- (11.26)
The Bloch wave function (11.1) satisfies the Schrodinger equation
if*fc = ih4>k, (11.3)
with
where V is the crystal potential. In terms of uk, these equations read
h(k)uk = Ekuk, (11.4)
126
Electron velocity 127
where
The Bloch wave function (11.1) is not an eigenfunction of the electron mo-
mentum operator p = —i/iV. However, we may calculate the expectation
value of the momentum:
(P) = (#*, -ihv*k) = (I*, (p + hk)uh). (11.5)
In order to evaluate this, we first write (11.4) with k replaced by some

other allowed vector k' = k + 6k. If we approach the continuum limit, Sk
may be chosen to be essentially in any direction and also arbitrarily small
in magnitude. If we neglect quadratic and higher powers of Sk, (11.4)
gives
h(k)uk> + 6k • (dh(k)/dk)uk = Ekuk> + 6k • (dEk/dk)uk. (11.6)
We now pre-multiply this equation by u*k and apply the integration / d3r.
The first term on the left-hand side becomes
(uk,h(k)uk>) = Ek(uk,uk>),
where we have again used (11.4). This vanishes because of (11.26). The
first term on the right-hand side of (11.6) similarly gives zero contribution.
The other terms in (11.6) give
where again (11.26) has been used to simplify the right-hand side. Because
the direction of 6k is arbitrary, (11.7) remains true if we cancel the factor
5k on each side. Prom the definition of the operator h(k) in (11.4),
dh(k)/dk = (h/m)(p + hk). (11.8)
Thus (11.5), (11.7) and (11.8) give
(p) = (m/h)(dEk/dk) (11.9a)
or
v = (l/h)(dEk/dk), (11.96)
where v = (p)/m is the expectation value of the velocity of the electron.
128 11 Electron motion in crystals
Ek against k dEjdk against k d2Ek/dk2against k
Figure 11.1. Plots of Ek, dEk/dk and d2Ek/dk2 against k for the three lowest bands
of figure 10.6.
In classical mechanics, the direction and magnitude of the electron velocity

would continually be changing, because of collisions between the electron
and the atoms of the crystal. The result (11.96) shows that in quantum
mechanics v is constant in both direction and magnitude. This result is
valid only for a perfectly regular crystal lattice.
In the one-dimensional model described in the previous chapter, the graphs
of Ek plotted against k are as drawn in figure 10.6. We reproduce these
in figure 11.1, together with the graphs of dEk/dk and d2Ek/dk2, for the
three lowest bands. Notice that the maximum value of v = h~1dEk/dk
increases as we go up to higher bands. This is because the width of the
bands increases with increasing band number, while the range of definition
of k (given in (10.14)) remains fixed.
Motion in an external electric field

The Bloch wave functions (11.1) describe stationary states, corresponding
to no variation with time of the probability density at each point in the
crystal. If we wish to consider the motion of an electron in the presence
Motion in an external electric field 129
of an external electric field, it is more appropriate to use a localised wave

packet. This is a superposition of stationary-state wave functions:
,t), (11.10)
where the function N(k) determines the shape and spread of the wave
packet, and /d 3 fc means integration over each of the components of the
vector k. Because each Bloch wave (11.1) satisfies the time-dependent
Schrodinger equation (11.3), so also will any superposition (11.10). How-
ever, (11.10) is not a stationary state.
We shall suppose that N(k) is sharply peaked at some value of k and
that it falls rapidly to zero outside the immediate vicinity of this value.
This means that each component of the wave packet corresponds to ap-
proximately the same value v = dEk/dk of the expectation value of the
electron velocity. We explained in chapter 4 that a narrow spread in k
corresponds to a wide spread in r. If we had, instead, chosen a wave
packet whose spread in r was less than the dimension of a unit cell of the
crystal, the expectation value of the electron velocity would have varied
as the wave packet moved across each unit cell. This variation may be
traced back to the variation over the unit cell of the factor uk(r) in the
Bloch wave function (11.1). However, for a wave packet that spans many
unit cells, the variation of Uk{r) is averaged out and the time variation of
v is negligible.
That is the situation in the absence of an external electric field. In prob-
lem 4.4 we proved Ehrenfest's theorem (4.14), that a wave packet moves
like a classical particle. Hence in an electric field £ the vector k about
which N(k) is peaked changes with time in such a way that the rate of
change of the electron energy is equal to the rate of work done by the field:
dEk/dt = -eSv. (11.11a)
However, dEk/dt = k • dEk/dk = Tik • v, and so
k = -(e/h)£. (11.116)
This result holds provided that the necessary empty electron states exist
so as to allow k to change.
When £ is constant, k increases uniformly with time in the direction
opposite to 5. At whatever point of the dispersion curve k starts, it
moves in this direction until it reaches the limit of the range of definition
-n/L n/L
Figure 11.2. Motion of the representative point in the presence of an electric field £.
When the point reaches G, it reappears at Gf.
of the vector k. In the one-dimensional model, this is the point G in

figure 11.2, corresponding to k = n/L. Because the states with k =
n/L and k = — n/L coincide, the representative point now reappears at
the point G' in figure 11.2 and then it again moves across towards G.
Prom figure 11.1 we see that this means that the electron velocity v =
h~1dEj£/dk oscillates back and forth as the cycle is repeated. This means
that in real space the electron oscillates back and forth also.
We have assumed that no band overlap occurs, as is true in the one-
dimensional model. When there is band overlap, even a weak field may
induce the electron to make a transition from one band to another, and
then the motion is more complicated.
Electric current
In a full band, all the allowed representative points along the correspond-
ing dispersion curve are occupied and the exclusion principle therefore
forbids the distribution of these points to change when a field is applied.
For each electron with velocity v there is another with velocity — v, and
so the total current due to all the electrons in a full band vanishes.
But suppose that the conduction band is half full and is separated from
the band above it by a gap. Let the dispersion curve for the conduction
band have the shape drawn in figure 11.2; if, instead, it is the other way
up like the middle band in figure 11.1, the discussion must be modified
accordingly. The representative points for the electrons, which are initially
in the lower half of the band, will be strung out along the (v, k) curve like
a chain of beads. When there is no electric field the points are distributed
symmetrically about k = 0, as in figure 11.3(a), so that the total current
Electric current 131
\v
.r\
I
(a) (b)
i
1 1
r\ 1
1
j 1
k
1 % 1
1 1
Figure 11.3. The velocity of conduction band electrons (a) when there is no electric
field; (6), (c) and (d) with an electric field 8, at successive times.
is zero. When a field is applied towards the left, the representative points
begin to move to the right, as shown in figure 11.3(6). The vector sum
of the velocities of the collection of electrons is now non-zero, and so a
net electric current flows in the direction of the electric field (remember
that the electrons have negative charge). When the electrons reach the
right-most edge of the range of definition of k they reappear at the left,
figure 11.3(c), until in figure 11.3(rf) the sum of the velocities has changed
sign. An electric current nowflowsin the reverse direction. Subsequently,
it changes direction again, and so on.
This rather surprising result, that the direction of the electric current
oscillates back and forth with time, does not apply if the conduction band
overlaps the next higher band, and if all the higher bands similarly overlap
each other. Then the electric current increases steadily, remaining in the
direction of the applied field.
In either case, the steady state described by Ohm's law is not found; the
crystal has no electrical resistance. This is because the discussion has
assumed that the crystal lattice is perfectly regular. In practice this is
never exactly true. Even at the absolute zero of temperature the nuclei
are not completely anchored at their lattice sites, but oscillate with zero-
point energy (see the discussion of the harmonic oscillator in chapter 3).
In addition, the exact periodicity of the lattice is disturbed, in nearly all
crystals, by the presence of imperfections such as geometrical dislocations
and concentrations of impurity atoms.
Imperfections and impurities apart, the main source of resistance in a
solid is the thermal vibration of the atoms about their lattice sites.
Because the displaced atoms exert forces on one another, their vibrations
are correlated. This correlation allows an analysis in terms of lattice
waves; these travel through the crystal with speeds of the order of the
speed of sound. Just as the energy of an electromagnetic wave is quantised,
so is the energy of a lattice wave; the quantum of energy is called a phonon.
Phonons behave in many respects like particles, as do photons.
As the electrons move through the crystal, they repeatedly collide with
phonons. As the temperature T increases, the collisions become more
frequent and more violent. When an electric field is applied to the crys-
tal, the representative points illustrated in figure 11.3(a) begin to move
according to (11.11), but this motion is opposed by the phonon scatter-
ing. The scattering with phonons tends to prevent the electrons from
increasing their energy and so from rising into the uppermost parts of the
conduction band. A steady state is rapidly reached: the distribution of
representative points becomes static, displaced asymmetrically from its
original position as in figure 11.3(6). A steady electric current then flows.
Effective mass and holes

We have seen that when a force is applied to an electron in a perfectly
regular crystal, the electron may sometimes move in the direction opposite
to the force. We now show that, as a result of the presence of the crystal
lattice, the electron moves as if it had a mass m* different from its real
mass m. This effective mass m* varies with the energy Ek, and in a part
of each band it is negative.
For simplicity, consider the one-dimensional case. Prom (11.96),
dt dt [h dk ) ~h Ok* k- (1L12)
If we now use (11.116), this gives
m*dv/dt = -eS (11.13)
Effective mass and holes 133
with
1 1 d2Ek
(11.14)
m* h2 dk2
The relation (11.13) is just Newton's law for the motion of the electron,
but with an effective mass m*.
We have plotted d2Ek/dk2 for the one-dimensional model in figure 11.1;
the effective mass m* is proportional to (d2Ek/dk2)~1. For each of the
bands, m* is negative near the top of the band. This means that the
acceleration of an electron near the top of a band is opposite to the applied
force. Suppose that a particular band is nearly full, with a few levels near
the top of the band unfilled. This occurs in a semiconductor, where at
zero temperature the valence band is filled and the conduction band is
empty, but at non-zero temperature a few of the electrons from the top
of the valence band are thermally excited into the conduction band. It
is convenient to regard the unfilled states as fictitious particles, called
holes. If the levels were filled with electrons, these would have negative
charge and negative effective mass. The holes correspond to an absence
of these electrons and so they have positive charge and positive mass.
The dynamics of the charge carriers in the nearly filled band may then be
described either in terms of the large number of electrons in the band or
in terms of the small number of holes.
We illustrate this by considering the case in which a single electron is
missing from an otherwise full band. When an electric field is applied,
the total electric current contributed by the band is
u (11.15a)
where the summation extends over all the states i except for the unfilled
state j . However, when the band is full there is no net electric current,
because for each state corresponding to velocity v there is another corre-
sponding to velocity — v. Hence
0= -
where the summation extends over the whole band, and so (11.15a) may
also be written as
J = evj. (11.156)
This is just the current that would be contributed by a solitary positively
charged carrier with velocity Vj equal to that which an electron occupying
the level j would have had. Similarly, the rate of change of momentum of
all the electrons in the band is
dP x—s *dv{ /i 1 1^? \
(1L16a)
X>'dT
However, when the band is filled, the exclusion principle forbids its overall
quantum state from changing, so that then the rate of change of its total
momentum vanishes:
Hence (11.16a) may also be written as
dP/dt = (-m^dvj/dt (11.166)
So the rate of change of momentum also may be calculated by ignoring

the electrons, and assigning to the unfilled states j fictitious particles of
effective mass (—m*) and velocity Vj.
Thermal excitation
In a metal or semimetal at zero temperature, all states in the conduction
band with energies up to some maximum value (o are filled, and the rest
are empty. At temperatures T > 0 the distribution of occupied states
does not end abruptly at £o- As a result of thermal excitation, electrons
are found with energies greater than (o; the distribution tails off smoothly
beyond energy (Q. In a semiconductor or insulator at T = 0, there are
no electrons in the conduction band and no holes in the valence band.
At T > 0 some electrons are excited into the conduction band and leave
behind holes in the valence band.
To study the effects of temperature it is rarely necessary to consider the
details of the collision processes that cause electron excitation. In a solid
in thermal equilibrium, the electrons are distributed in the available en-
ergy levels in accordance with the predictions of statistical mechanics.
Although the electrons continually change levels as a result of collisions,
statistical mechanics gives their expected average distribution among the
levels independently of the details of the collisions. Because the effects of
the exclusion principle are important, the classical Maxwell-Boltzmann
distribution (9.24) is not applicable; instead, the appropriate distribution
is the Fermi-Dirac distribution. If the level E{ has degeneracy Qi (with
Thermal excitation 135
a minimum value gi = 2 when spin is taken into account), the average

number of electrons expected in that level is
Ui =
(^
Here, the parameter ( is called the Fermi energy or, more correctly, the
chemical potential. It is independent of energy, but it varies with the tem-
perature T. It is determined in terms of the total number N of electrons
by the implicit equation
This usually has the effect that £ varies as log &BT, SO that it varies
rather slowly with T. In practice, the variation between T = 0 and room
temperature is almost negligible.
The Fermi-Dirac distribution (11.17) is derived in any textbook on sta-
tistical mechanics. It has the property that at zero temperature
9i Ei < Co,
where £0 is the value of £ at T = 0. For temperatures T > 0 the distribu-

tion has the shape drawn in figure 11.4; in this figure we have supposed
that all the degeneracies gi have the same value g. For large values of E{,
such that (E{ — () » k^T, the exponential in the denominator of (11.17)
dominates and
E / k T
(11.20)
where C = e^^kBT. So for energies for above the Fermi energy we again
have the classical Maxwell-Boltzmann distribution (9.24). This is because
the average number n; of electrons in each level is then very small, so that
the exclusion principle has little effect.
In any bound quantum system of finite size the allowed values of Ei are
discrete, but for a large system they are usually so close together that it
is an excellent approximation to treat Ei as a continuous variable. Let
p(E) &E be the number of states in the energy range E to E + dE. The ex-
pected average number of electrons in this range is then, as the continuum
limit of (11.17),
p(E)dE
i. (11.21)
Figure 11.4. The Fermi-Dirac distribution at temperatures T > 0, for the case in
which all the levels have the same degeneracy g.
The equation that determines C,is the analogue of (11.18):

p(E) dE
N =
J (11.22)
e(E-O/kBT
The average number density n(E) given by (11.21) is the product of p(E)
and a factor that has the shape drawn in figure 11.4 (with g set equal to 1).
In the energy bands, p(E) is a function that depends on the nature of the
crystal, whereas in the gaps p(E) = 0. Unless the temperature T is very
high, the Fermi level £ for a metal or semimetal lies in the conduction
band, while for an insulator or a semiconductor it lies somewhere in
the energy gap between the valence band and the conduction band. We
demonstrate this last result. In an insulator or an intrinsic semiconductor,
there are just enough electrons to fill all the levels up to the top of the
valence band:
N
-L dEp(E),
where Ev is the energy at the top of the valence band. However, from
(11.23)
(11.22) evaluated at T = 0,
/•Co
JV= / dEp(E). (11.24)
J — OO
That is, at T = 0 all the levels up to the Fermi energy ( 0 are filled, and
all those above it are empty. For (11.23) and (11.24) to be consistent,
Pair excitation in intrinsic semiconductors 137
p(E) must vanish for values of E between Ev and (o? so that ( 0 must be
in the energy gap. As T is increased ( changes slowly, but remains in the
energy gap until T becomes much greater than room temperature.
Pair excitation in intrinsic semiconductors

There are two obvious ways in which an electron in an intrinsic semicon-
ductor can be knocked from the valence band into the conduction band,
so producing an electron-hole pair of current carriers. The first way
is to bombard the crystal with photons whose frequency u is such that
Tiuj > A, where A is the magnitude of the energy gap between the valence
band and the conduction band. The rate of pair production increases with
the intensity of the light, and so when a steady electric field is applied the
electric current also increases with the light intensity. This is the principle
of the photoelectric cell.
Secondly, when T > 0 pair excitation occurs because of collisions between
electrons and phonons. The number of pairs so excited can be calculated
from the Fermi-Dirac distribution (11.21). Let the number of electrons
excited into the conduction band, or into even higher bands, be Ne. Be-
cause the exponential in the denominator of the Fermi-Dirac distribution
grows rapidly as E increases, most of these excited electrons will actually
be near the bottom of the conduction band, and most of the holes that
they leave behind will be near the top of the valence band. Hence, from
what we have said earlier in this chapter, both the excited electrons and
the holes will nearly all have positive effective mass ra*.
If again Ev is the energy at the top of the valence band, and A is the
energy gap, the Fermi-Dirac distribution (11.21) gives the total number
of electron-hole pairs as
On the other hand, the total number of electrons in the valence band and
in lower bands is
This must be equal to (N - ATe), where the total number N of electrons

is given by (11.23). Thus
P(E)
Figure 11.5. Model in which the state-density function p(E) has the same shape at the
top of the valence band as at the bottom of the conduction band.
-L dE-
p(E)
(11.27)
This is the equation that determines

For purposes of illustration, consider the special case where the level den-
sity p(E) has the same shape near the bottom of the conduction band as at
the top of the valence band (figure 11.5). To a reasonable approximation,
this is true in silicon, germanium and many other semiconductors. Then
near the energy gap
p(E)=p(Ev (11.28)
Most of the contribution to the integral on the left-hand side of (11.27)

comes from near the bottom end of the integration, and most of the con-
tribution to the integral on the right-hand side comes from the top end,
so we do not need to know about p(E) elsewhere. One may readily check
that (11.27) is satisfied by
= Ev (11.29)
So in this special case £ is independent of T and lies exactly in the middle

of the energy gap.
Pair excitation in intrinsic semiconductors 139
Inserting (11.29) into (11.25) and evaluating it for the case k&T < A, so
that the exponential in the denominator dominates, we have
-L Ev+A
(11.30)
Since most of the contribution to this integral arises from values of E near
the lower end of the integration, we see that approximately
Ne oc e" A / 2fcBT . (11.31)
At room temperature, k^T « ^j e V. For silicon or germanium, the en-

ergy gap A is about 1 eV, so that the exponential is very small indeed
at room temperature. However, the total number N of electrons is very
large, so that the exponential is multiplied by a very large number, and at
room temperature the conductivity is fairly good. However, the exponen-
tial decreases rapidly as T decreases, and at around 0°C the conductivity
of silicon or germanium is only a seventh of its value at room tempera-
ture (remember that T = 0 corresponds to — 273° C). Electronic devices
that contain semiconductors frequently incorporate feedback loops in their
circuits in order to compensate for changes of temperature.
Problems
11.1 Consider the free-electron model for a crystal (problem 10.1), defining
the constant potential to be zero inside the crystal.
(i) In the case of a one-dimensional lattice, show that
p(E) = {NL/Kh)(m/2E)ll2.
(ii) In the case of a simple square lattice of area A, show that
p(E) = (Am/nh2).
(iii) In the case of a simple cubic lattice of volume V, show that
p(E) =
11.2 An atomic monolayer is formed by evaporating a metal onto a flat

substrate. As a model for the behaviour of the electrons in the metal
film, suppose that the potential V(r) vanishes for 0 < z < a, where a
is the atomic diameter, and is infinite otherwise. Hence the electrons
travel freely in the film. How does the density of states p(E) vary with
El
11.3 In the free-electron model of a metal, the electrons behave as a set
of independent particles obeying the Fermi-Dirac distribution. An
electron having momentum component normal to the metal surface
greater than a critical value (2mVb)1/2 can escape from the surface;
this is called thermionic emission. In practice Vb ^> ( + k&T. If a
collecting electrode is placed near the metal surface and is maintained
at a sufficiently positive potential relative to the metal, it will collect
all the electrons that are emitted. Show that the current per unit area
emkBT c(C
11.4 As a model of a semiconductor, suppose that there are N electron

states all having the same energy —A, and that these are all filled at
zero temperature. (That is, the valence band has zero width.) The
conduction band is a continuum of states of energy E > 0, with density
given by
p(E) =
with A a constant. Show that the number of electrons in the conduction
band at temperature T is
where, if T is so small that e c//eBT < 1 and that n < TV,
12
Transistors
Impurities
We have so far described the properties of intrinsic semiconductors, in
which the crystal lattice is perfectly regular. However, the electrical prop-
erties of semiconductors, unlike metals or semimetals, are drastically af-
fected by the addition of small traces of impurity atoms. Observable
effects occur with impurity concentrations as low as a few parts in 108,
and increasing the impurity concentration to one part in 105 can increase
the conductivity by as much as a factor of 103 at room temperature and
1012 at liquid-helium temperatures. The semiconductor is said to be doped
with impurity atoms.
In order to study the effect of doping, we first consider a crystal consisting
of a periodic array of one type of atom, except that just one of the atoms
has been replaced by an atom of a different type. As a one-dimensional
model of this situation, we take the same infinite chain of square wells as
in chapter 10, but with one of the wells having a different depth, Ui say
(figure 12.1).
We recall that for the perfectly regular crystal, the stationary state solu-
tions are Bloch waves of the form (10.3). These have the property that
when the required continuity conditions on the wave function are satisfied
at the two edges of one of the potential wells, they are automatically sat-
isfied also at the edges of all the other wells. For this reason, we might
guess that in the case of the potential of figure 12.1, the stationary-state
wave function is again composed of Bloch waves in the parts of the crystal
where the lattice is perfectly regular, but that it has a different form in
the neighbourhood of the impurity.
The Bloch wave elkxuk(x), with k real, represents a wave travelling from
left to right. If such a wave is incident on the impurity, part of it is re-
flected, and the remaining part is transmitted across the impurity (com-
pare (3.24) and (3.25)). Hence the total wave function to the left of the
impurity atom is
eikxuk{x) +Re~ikxu-k{x), (12.1)
141
142 12 Transistors
*— L
u
1
Figure 12.1. One-dimensional model for a crystal with a single impurity atom,
and to the right of the impurity atom it is
Teikxuk(x). (12.2)
Here, R and T are constants. Inside the potential well that represents the
impurity atom, the solution to the time-independent Schrodinger equation
is
JLJ e "T u e yiz.o)
where
The boundary conditions at the edges of the wells corresponding to the

host atoms, that is, at the edges of the wells other than the impurity,
determine E in terms of k. The relationship is exactly as in chapter 10:
there is the same band structure as if the impurity were not there. The
constants i?, T, B and C are determined by the four conditions that tj)
and ij)1 be continuous at both edges of the impurity well.
There is also a different type of solution, in which the electron is more or
less localised in the vicinity of the impurity atom. In this solution, the
wave function still satisfies the Bloch condition (E.10), except near to the
impurity, but k is pure imaginary. Take the origin x — 0 to be in the
middle of the impurity well. To the left of the impurity the wave function
is
eKXvi(x) K > 0, (12.4)
and to the right of the impurity it is
(12.5)
where the functions vi(x) and V2(x) are both periodic. Because of the
exponential factors, the wave function goes to zero as x -» ±oo, so the
electron is trapped in the neighbourhood of the impurity.
n- and p-type semiconductors 143
By making vi(x) and v2(x) periodic, we again ensure that when we satisfy
the required continuity conditions at the edges of any single well on either
side of the impurity, they are automatically satisfied at the edges of the
well corresponding to all the host atoms. For a given K, the functions v\
and V2 are determined in a way precisely similar to the determination of
Uk(x) in chapter 10. Each contains a multiplicative constant that remains
to be fixed and that determines its normalisation. Since the complete
potential is symmetric about x = 0, the solutions fall into two classes. One
class has positive parity and the other negative parity, ip(x) = ±if;(—x).
For definiteness, consider the positive-parity solutions (the negative-parity
ones may be discussed similarly). For these v\{x) = +V2(—x), so that the
pair of functions v\ and v2 now contains one multiplicative constant that
remains to be fixed, A say. Inside the impurity well, the positive-parity
solution is
B(ehx + e''nx), (12.6)
where 7 is defined in (12.3). Imposing a normalisation condition on the

overall wave function, for example (ipîp) = 1, determines the constant B
in terms of A. There are two more continuity conditions to be satisfied:
that \j) and ip' are continuous at one edge of the impurity well (the cor-
responding conditions at the other edge are then automatically satisfied,
because of the even-parity condition ip(x) = ifj(-x)). Since there are only
two unknowns, A and E (which determines 7 and «), we expect a discrete
set of solutions.
Thus, in addition to the usual band structure associated with the host
atoms, there are discrete trapped-electron levels resulting from the impu-
rity. This result continues to apply when further impurity atoms are sub-
stituted for host atoms, provided that their concentration remains so low
that they are far enough apart for there to be little overlap of the trapped-
electron wave functions. Each impurity level has degeneracy equal to twice
the number of impurity atoms (the factor of 2 corresponding to the two
possible electron spin states). However, if their concentration becomes
greater than about one part in 103, the impurity levels begin to spread
out into bands, in the same way as the levels of the host atoms spread into
bands when a large number of host atoms are brought together to form a
crystal.
n- and p-type semiconductors

Suppose that a crystal of silicon is doped with a few parts per million of
phosphorus. Silicon has a valence of 4, but phosphorus has a valence of
144 12 Transistors
5. So at each lattice site where a phosphorus atom replaces a silicon atom

there are five valence electrons, whereas only four are needed to form the
covalent bonds that anchor the atom to its nearest-neighbour host atoms
(figure 12.2(a)). The extra electron either occupies one of the degenerate
discrete impurity levels, in which it is trapped near an impurity atom, or
it is found in the conduction band. It turns out that the lowest impurity
level lies just below the bottom of the conduction band (figure 12.2(6)).
So at zero temperature the fifth valence electron is in this level, because
it is the lowest level available to it. However, when T > 0 the probability
is high that the electron is thermally excited into the conduction band.
In terms of the Fermi-Dirac distribution, this is because the density of
states p(E) is much greater in the conduction band, where there are very
many levels close together, than in the impurity level.
The effect of doping silicon with phosphorus is therefore to provide an
excess of electrons in the conduction band, that is an excess of negative
carriers which can contribute to an electric current. A material doped in
this way is called an n-type semiconductor; it is said to be doped with
donor atoms.
Alternatively, the silicon crystal may be doped with atoms that have a
valence of 3, such as boron atoms. The lowest impurity level is then found
to be just above the top of the valence band. The impurity atoms each
bring with them one electron less than is needed to saturate the nearest-
neighbour covalent bonds at the sites that they occupy. At temperatures
T > 0 electrons are thermally excited from the valence band into the
degenerate impurity level, being trapped near the impurity atoms and
so providing the missing bonds. They leave behind holes in the valence
band. These are mobile and are positive carriers which can contribute to
an electric current. The material is known as a p-type semiconductor; it
is said to be doped with acceptor atoms.
Impurities and crystal colour

The presence of iitipurity atoms can have an important effect on the colour
of a crystal. For example, in pure aluminium oxide the energy gap between
the valence band and the conduction band is so large that visible light is
not absorbed; the crystal is transparent. If a small quantity of chromium is
added, say | % , there are two relatively narrow bands of impurity levels.
These are situated in the energy gap, in such a position that photons
corresponding to green and yellow light are absorbed. So now the crystal
appears red; in fact, it is ruby. If instead titanium is added, the position
Semiconductor junction 145
Lattice site
carries charge
Si (a)
Fifth valence
electron
Conduction band
Impurity
level
(b)
Valence band
Figure 12.2. (o) A silicon crystal doped with the donor impurity phosphorus, with (b)
the impurity level just below the conduction band.
of the impurity levels is such that red light is absorbed. The crystal now
appears blue; it is sapphire.
Semiconductor junction
Consider two pieces of a given semiconductor, one doped with donor atoms
and the other with acceptor atoms. Suppose that each piece has a face
that is accurately plane, and imagine bringing the two pieces into contact
at their plane faces. This forms a pn-junction. (In practice, rather than
146 12 Transistors
pressing together two separate pieces of crystal, the junction is manufac-

tured from a single piece of host crystal by varying the doping in different
parts of it as the crystal is grown. This produces a transition region be-
tween the p-part and n-part that is typically about 1 fim in width. In our
discussion, we imagine that the transition is abrupt.)
At the instant when the two pieces are brought together the p-type mate-
rial has an excess of holes in the valence band, compared with the n-type
material. Similarly, the n-type material has an excess of electrons in the
conduction band, compared with the p-type material. See figure 12.3(a).
After the contact is made, it is energetically favourable for some of the
excess electrons in the conduction band of the n-type material to cross to
the p-type material and annihilate some of the holes there. As they do so,
a net negative charge is built up in the p-type material, and a net positive
charge in the n-type material. Thus an electrostatic potential is set up,
and this eventually stops the flow of further electrons across the junction.
The shape of this electrostatic potential 4> is drawn in figure 12.3(c). It
causes an additional energy — e<j> to be superimposed on the level struc-
ture, so that when dynamical equilibrium is reached the band structure
becomes as drawn in figure 12.3(6). Typically, the region where the elec-
trostatic potential <f> varies appreciably, so that there is an appreciable
electric field — V</>, is of width 10 fim.
Suppose that heat is applied to the junction, so that extra electron-hole
pairs are created there by thermal excitation. The electric field — V</>
drives the electrons towards the n-type material and the holes towards the
p-type material. So if the two sides of the crystal are joined to an external
circuit, the effect of the heat is to drive a current through the crystal from
the n-type side to the p-type side, and round the circuit.
In an exactly similar way, a current is generated if light is shone on the
junction, so that the absorbed photons create electron-hole pairs. This is
the photovoltaic effect, which is the basis of the solar cell. Its operation
may be reversed: if, instead, current is driven through the junction by an
external battery, electrons and holes recombine in the junction region, so
producing photons. This is the light-emitting diode. If two faces of the
crystal, perpendicular to the junction plane, are polished flat and made
parallel to each other, the device operates as a semiconductor laser.
The diode
We now explain how the pn-junction acts as a diode rectifier: when a bat-
tery is connected such that the p-side is at a higher electrostatic potential
The diode 147
Conduction
band
(a)
Valence o o o o
band ° °o°o
P-type n-type
\
\
(b)
\
\
<t>0
(c)
p-type n-type
Figure 12.3. The pn-junction: (a) the band structure at the instant the two pieces are
brought together and (b) in dynamical equilibrium. The difference between (a) and
(6) is caused by the electrostatic potential <j> drawn in (c).
than the n-side, current flows readily, but only a limited current flows
when the polarity of the battery is reversed.
Suppose first that no external electrostatic potential difference is applied.
In chapter 11 we explained that in an intrinsic semiconductor the Fermi
energy ( lies somewhere in the energy gap. Its position is shifted when
the crystal is doped, and the Fermi energies of each of the two pieces, p
148 12 Transistors
and n, are further shifted when the pieces are pressed together to form the
junction. However, statistical mechanics tells us that any system of elec-
trons in dynamical equilibrium has a unique Fermi energy. So dynamical
equilibrium for the pn-j unction is reached when the Fermi energies ( of
the p-part and n-part coincide. We do not need to know the precise final
value of the common Fermi energy, but in practice it lies somewhere in
the overlap between the energy gaps on the two sides of the junction, as
drawn in figure 12.3(6).
According to the Fermi-Dirac distribution (11.21), the number of electrons
on the n-side that are above the bottom E% of the conduction band on the
p-side is
where pn(E) is the density of states on the n-side. Some of these electrons
will be moving in the right direction to cross over to the p-side, and some
of these will actually reach the p-side conduction band before collisions
reduce their energy to below E%. Hence there is a diffusion current of
electrons from the n-side to the p-side (that is, an electric current from
the p-side to the n-side):
/•OO
Jo oc / dEfr(E)e-(E-Mk»T. (12.8)
JEI
Similarly, some of the electrons in the conduction band on the p-side dif-
fuse into the n-side. In order that there be dynamical equilibrium, the
corresponding current must be equal and opposite to Jo- Exactly similar
considerations apply to the hole carriers.
Suppose now that an external electrostatic potential difference V is ap-
plied across the junction, by connecting the two sides of the crystal to a
battery. We assume that all this potential difference appears across the
junction region; if this is not a good approximation, the Ohmic potential
drop IR in the rest of the crystal may be allowed for once the current I
is known. With the extra driving potential V, the crystal is no longer in
dynamical equilibrium; there is no longer a single Fermi energy through-
out the crystal. However, on each side far from the junction there is a
situation close to equilibrium, not much affected by the additional current
flow caused by V. This is provided that V is not too large, so that the
drift velocity superimposed on the electron motion, corresponding to the
additional current, is small. Then outside the junction region each piece
The diode 149
EP+eV-
EP+eV
n (a)
I I
(b)
p n 1-
V
Figure 12.4. (a) The band structure of figure 12.3(6) when there is an additional
potential difference V whose sign is indicated in (b).
of crystal may, to a good approximation, be described by a Fermi-Dirac

distribution, with the Fermi energy in the same place relative to the bands
as it was before V was applied. (See figure 12.4, and compare it with fig-
ure 12.3(6) — we have chosen to define the origin E = 0 of the energy
scale such that its position relative to the n-type bands has not changed,
and have drawn the figure for the case where the sign of V is such that
the p-type bands move upwards in energy.) In the narrow transition re-
gion at the junction, the Fermi level is not defined: the additional electric
field corresponding to V is not small, and the situation there is far from
dynamical equilibrium.
Because the p-type side has the same Fermi-Dirac distribution as before
(in spite of the shift of origin of £7), the number of electrons in the conduc-
tion band on that side is the same as before. Thus the diffusion current
arising from electrons passing from the p-side to the n-side is still Jo given
by (12.8). However, the diffusion current in the other direction is changed.
Now in the n-type region only those electrons above the level E =
150 12 Transistors
-*» V
Figure 12.5. Plot of current / against potential difference V for a semiconductor diode.
The signs of V and / are defined in figure 12.4(6).
have sufficient energy to pass over to the p-type conduction band, and so
they carry a diffusion current proportional to
dEpn (12.9)
Assuming that the density of states pn(E) varies only slowly with E, this
is approximately
Hence the total electric current that passes through the crystal from the
n-side to the p-side is
(12.10)
This is the contribution from electron carriers; the hole carriers contribute
similarly.
/ is plotted against V in figure 12.5. When V is positive, / < /Q, but for
negative V, the current increases exponentially with the magnitude of V
(until the approximations made in the derivation are no longer valid).
The semiconductor diode can be made to operate as a laser (see chap-
ter 9). The first semiconductor laser was made in 1962, using stimulated
emissions associated with transitions from the conduction to the valence
band as the source of laser radiation. The pn-junction is forward-biased
(the p-type side is at a higher electrostatic potential than the n-type side),
providing the necessary pumping, and the end-faces of the diode, perpen-
dicular to the plane of the junction, are polished to give high reflectivity.
The diode 151
Conduction
band
(a)
Valence
band
n-type p-type n-type
Emitter Base Collector

I I I
R
(b)
4-'
Emitter Base Collector
Conduction
band
(c)
Valence
band
n-type p-type n-type
Figure 12.6. The npn-junction transistor (a) with no external potential, (c) with
external potentials applied as in (6).
152 12 Transistors
Semiconductor lasers can be made very small — the size of a pin head
— and have wide application, for example in compact disc players and in
fibre-optic communications.
The junction transistor

The junction triode transistor consists of a thin slice of p-type crystal
sandwiched between two pieces of n-type crystal. Alternatively it may be
a pnp-sandwich, but for definiteness we consider the npn case.
When no external potentials are applied, the disposition of the energy
bands at each of the two junctions is an in figure 12.3(6). We draw this
in figure 12.6(a), for the case where the two pieces of n-type material
are doped with equal impurity concentrations. Suppose now that the
device is connected to a pair of batteries as in figure 12.6(6). The p-type
slice (called the base) is at slightly lower electrostatic potential than the
left-hand n-type piece (the emitter), and the right-hand n-type piece (the
collector) is at rather higher potential.
The donor impurity concentration in the n-type material is made much
greater than the acceptor impurity concentration in the p-type material, so
that almost all the electric current is carried by electrons rather than holes.
The potential across the emitter/base circuit causes a flow of electrons
through the crystal from the emitter to the base, as in the case of the
diode. The base is made very thin, less than 5 /im across, so that most
of these electrons diffuse across it before collisions stop or reverse their
motion. The potential applied to the collector lowers the level of the
bottom of the conduction band in the collector region by a large amount.
Hence when the electrons arrive in the collector region from the base-
region conduction band, their energy is high enough for them to be able
to suffer many collisions, and cascade down the levels in the conduction
band, before they are stopped. So almost all of them reach the collector
terminal. That is, almost all the current of electrons that enters the emitter
emerges from the collector:
/c~/e- (12.11)
The emitter current / e is controlled by the magnitude of the potential dif-
ference between the emitter and the base, as for the diode. This potential
difference may be regarded as the input to the circuit; it may vary with
time. The output of the circuit is the potential difference ICR across the
resistor shown in figure 12.6(6). This can be much greater than the input,
if R is large — though R must not be too large, for then the collector
potential would be lowered so much that the device would not operate. A
The junction transistor 153
Gate—^
Insulator
— n
c ' ^ Drain
Source
Figure 12.7. A field-effect transistor.
power gain of 104 is typical.

The pnp-transistor works similarly: the signs of all the potentials and
currents are reversed, and holes are the principal carriers.
By doping a single piece of host crystal, many junction transistors can
be placed together on a single chip of crystal. With suitable doping tech-
niques, such an integrated circuit may also contain resistors and capac-
itors (though not inductors), but as these cannot be manufactured so
conveniently circuits are designed in such a way as to replace them with
transistors as far as possible. We have seen that the essential physics of
the transistor is in the junction region, which is very thin; the remainder
of the material has no active function. Hence a very large number of tran-
sistors can be accommodated in a single integrated circuit, several million
on a crystal chip the size of a finger nail.
Largely for constructional reasons, integrated circuits also commonly use
a different type of transistor, the field-effect transistor. These can also
have the advantage of much lower power consumption. There are various
types offield-effecttransistor; an example is the metal-oxide-silicon field-
effect transistor (MOSFET), shown in figure 12.7. A wafer of insulator is
sandwiched between a layer of metal, called the gate, and a piece of doped
semiconductor. This system replaces the base of a junction transistor. If
the semiconductor is p-type, it is placed between two pieces of n-type
crystal; these are called the source and the drain.
The drain is raised to a positive potential relative to the source. This
causes little flow of current since, as we have seen, under these conditions
electrons cannot readily cross the p-type region. However, the situation
changes if the gate is raised to a positive potential. Although the gate is
electrically insulated from the p-type material, this causes an electric field
to be set up near the upper surface of the p-type material. Consequently,
the conduction and valence bands are lowered in energy in this region (fig-
154 12 Transistors
Gate /Insulator
Conduction
band
Depth below surface

of p-type material
Figure 12.8. Lowering of conduction and valence bands near the surface of the p-type
material when a positive potential is applied to the gate.
ure 12.8), and the Fermi level £ appears in the conduction band. According
to the Fermi-Dirac distribution, there is then a substantial population of
electrons in the conduction band in this region, even though in the bulk of
the p-type material holes are the majority carriers. This inversion layer
at the surface allows electrons to flow from the source across to the drain.
The larger the gate voltage, the greater the electron concentration in the
inversion layer and the greater the current.
Two simple circuits

The simplest application of the diode is in the demodulator circuit, which
is one of the first stages of a radio receiver. The radio-frequency carrier
wave, modulated in amplitude by the lower-frequency audio signal, is ap-
plied between the p-type terminal of the diode and earth. If the other
side of the diode is connected to earth through a resistor, current flows
through the diode and the resistor when the input to the diode is positive.
That is, the output potential across the resistor corresponds to the input
with its lower half chopped off. This is shown in figure 12.9(a); in the
circuit symbol that denotes the diode, the arrow indicates the direction in
which current can pass freely. If the resistor is now replaced by a capaci-
tor (figure 12.9(6)), the radio-frequency component is removed, since the
impedence of the capacitor is low for high frequencies. The output is now
Two simple circuits 155
Diode
t -M-
o
t
Input Output
o
L
Earth (a)
Diode
T -H- T t
Input Output
J T
Earth (b)
Figure 12.9. A radio-frequency carrier wave, amplitude modulated with an audio-

frequency signal, applied to a diode circuit (a) chops off the negative components of
the input, (b) also removes the high-frequency component.
just the audio-frequency signal, which is amplified in further stages of the

radio receiver.
Notice that the rectifier is needed. To see this, suppose that the input
is the very simple amplitude-modulated wave cos pt cos ut where u is the
radio frequency and p is the audio frequency. However,
cosptcosut = 5 [cos (a; + p)t + COS(CJ — p)t]. (12.12)
Hence, the input is the sum of two high-frequency components, and so the
capacitor would short circuit the whole signal if the rectifier were not there.
The signal passed by the rectifier is F(t) = cosptcosut6{cosptcosut),
where 9{x) = 1 for x > 0 and 0(x) = 0 for x < 0. If this function
is represented as a Fourier integral, it is found to have a low-frequency
component corresponding to the audio signal. It is not easy to calculate
the Fourier transform, but a function that has much the same shape as
F(t) is G(t) = cos2 \pt cos2 \ut. Like F(t), G(t) is non-negative, and
both F(t) and G(t) oscillate in a similar fashion. However,
156 12 Transistors
Figure 12.10. The flip-flop circuit.
G(t) = \(COSpt+1)(COS Ut+ 1)

= ^(cosptcosut + cosut + cospt + 1). (12.13)
As we have seen in (12.12), the first term in this expansion consists of
high-frequency components only, as does the second term; these are short-
circuited by the capacitor. So the output from the circuit is ^(cospt + 1),
which is the audio signal superimposed on a constant potential.
We have already described the use of the junction transistor as an am-
plifier. This is an example of an analogue circuit: the output varies
continuously with the input. Many electronic instruments, in particular
calculators, rather make use of digital circuits. In a digital circuit, a given
circuit element operates in only two recognisable states, for example car-
rying a significant current and carrying little or no current. An example
Two simple circuits 157
of such a circuit is the bistable circuit or flip-flop. This is drawn in fig-

ure 12.10, where two npn-junction transistors are used. Suppose that
transistor 1 is conducting. Then the point A is at rather lower potential
than the battery terminal E. Hence the point C, and so also the base of
transistor 2, is at a comparatively low potential. Thus transistor 2 is not
conducting, and so the point B is at a potential somewhat higher than
point A. Hence also the point D, and thus the base of transistor 1, is at a
comparatively high potential. This is as it should be if the transistor 1 is
conducting, so the situation is stable. Suppose now that a negative signal
is applied to the base of transistor 1, so that it stops conducting. Then
the potential of point A rises, and therefore also that of point C and of
the base of transistor 2. Thus transistor 2 now conducts; this is the other
stable situation. If either transistor 2 is switched off by a negative signal
applied to its base, or transistor 1 is switched on by a positive signal
applied to its base, the circuit returns to the first stable configuration.
Problems
12.1 In a one-dimensional crystal an impurity atom replaces the atom at
the origin. Investigate the energy-level structure, taking as a model
for the potential
12.2 The tunnel diode is a junction between p- and n-type materials which
are both very heavily doped, so that the impurity levels are spread out
into bands. In the n-type material the Fermi level lies in the conduction
band, and in the p-type material it lies in the valence band. The
transition region at the junction is made so narrow that tunnelling
readily takes place across the interface.
Sketch the energy bands near the interface (a) when no external po-
tential is applied, (6) when the n-type material is connected to a small
negative potential V relative to the n-type, and (c) when it is connected
to a fairly large negative potential.
Sketch the graph of the current I through the junction against the
potential V. (The correct curve has a region of negative 'resistance'
dl/dV.)
Appendix A
Power-series solutions
In this appendix we describe briefly the procedure for solving differential

equations by means of power-series expansions. This procedure is used for
finding the stationary-state wave functions of the linear harmonic oscillator
(chapter 3) and of the hydrogen atom (chapter 5).
The linear harmonic oscillator

The Schrodinger equation to be solved is given in (3.15):
{-h2/2m)(d2tf>/dx2) + (\mu2x2 - E)%l> = 0. (A.I)
We are interested in solutions having the property I/J —>• 0 as x -> dboo, so
we begin by examining the differential equation (A.I) at very large values
of x. For any given eigenvalue £?, when x is sufficiently large E is much
less than ^muj2x2 and so the term — Ety in the equation is then much less
important than the other two terms. This suggests that it may be useful
to write
tl>(x)=e~mu'x2'2hH(x), (A.2)
because the exponential factor has the property that for large x
The exponential e+mujx l2h has the same property, but if we used this
instead, H(x) would have to obey very much more stringent conditions
at x — ±oo in order to satisfy the required boundary conditions for -0-
Substitute (A.2) into (A.I):
H" - (2mujx/h)Hf + [2m(E - \hu)/h2]H = 0. (A.3)
Look for a power-series solution:

oo
Y,asXs. (A.4)
s=0
158
The hydrogen atom 159
We substitute this into the left-hand side of the differential equation (A.3)
and equate to zero the coefficient of xs:
2m \ ( ) /
[ \
For any choice of ao and ai, this relation determines as for all s > 1. There
are two parameters ao and a\ that can be chosen arbitrarily, because the
differential equation (A.3) is second-order.
When s is large, (A.5) gives
2mcj
(A.6)
(22 fi S
This means that, for any fixed value of #, the ratio of successive terms in
the series (A.4) tends to zero as s -> 00, and therefore the series converges.
Consider now the series expansion of the function AemuJX /h xB, where A
and B are constants. Its coefficients also have the property (A.6) for large
s. If we make the plausible assumption, which can be justified, that the
form of H(x) at large x is controlled by the high powers in the series, we
conclude that for large x
Inserting this behaviour into expression (A.2) for -0, we see that, whatever
may be the values of A and J5, we do not have the desired behaviour for
ip at infinity.
The only way to avoid this is to arrange that the series expansion of H(x)
terminates at some value of s, so that H(x) is a polynomial. From (A.5),
we see that this will happen if and only if:
either a\ = 0, so that all odd powers of x are absent, and for some even
value 5 = 2n, E— \TIUJ = 2nTujo, so that the series terminates at 5 = 2n,
or ao = 0, so that all even powers of x are absent, and for some odd
value s = 2n + 1, E - ±hu = (2n + l)hw, so that the series terminates
at s = 2n + 1.
Together, these two possibilities correspond to the results quoted in (3.16)
and (3.17).
160 Power-series solutions
The hydrogen atom

For the hydrogen atom, the differential equation to be solved is given in
(5.16):
We again want a solution R(r) that goes to zero as r -> oo.

The procedure for obtaining the solution is similar to that above. How-
ever, in the previous example we knew that we had obtained both the
independent solutions of the second-order differential equation (A.3) by
making the expansion (A.4), because the two parameters ao and a\ could
be chosen freely. If we proceed here in exactly the same way, we obtain
only one solution. In order to reveal both of the two independent solutions,
we must take a slightly more general form of series expansion; otherwise,
the procedure is just the same.
When r is large, the term At2 dominates over the other two terms in the
square bracket in (A.7) . If we neglect the other two terms, the resulting
differential equation for rR(r) has solutions e ± K r . By analogy with (A.2),
we therefore write
rR(r) = e-Krf(r). (A.8)
Substituting this into the differential equation (A.7), we find
Now make the modified series expansion
5=0
and substitute this expansion into the left-hand side of the differential
equation (A.9). The lowest resulting power of r that we find is ra~2, and
on equating the coefficient of this to zero we find
ao[a(a-l)~ 1(1 + 1)] = 0. (A.ll)
This is known as the indicial equation. It tells us that if we want to choose

the constant a such that a0 ^ 0, so that the expansion (A. 10) begins with
the power r°", then either a = — I or a — I + 1.
The hydrogen atom
If we make the choice a = — /, the resulting wave function ip(r) diverges

like r~l~l at the origin. This means that then ip(r) does not satisfy the
Schrodinger equation at r = 0, even though it does satisfy the equation
at all other points r . Because the Schrodinger equation is to be satisfied
at every point r , we must reject this solution.
We therefore set a = / + 1. We now equate to zero the coefficient of
For any choice of ao, this determines all the other coefficients as. When s
is large,
as+i/as ~ 2K/S. (A.13)
Hence, arguing as before, we conclude that for large r
f(r)~Ae2KrrB,
which is not an acceptable behaviour. So again the series must terminate.

From (A. 12), we see that [(Me2/A-KSQKTI2) — a] must be a non-negative
integer JV, which is equivalent to the result (5.17), with N — n — I — 1.
(Recall that we already know, from the analysis of orbital angular mo-
mentum in chapter 5, that / must be a non-negative integer.)
For the ground-state solution, we have / = N = 0 and f(r) = ao- So the
wave function is just (5.18), with the multiplicative constant chosen to
normalise it correctly.
Appendix B
The delta function and Fourier transforms
In this appendix we present a heuristic discussion of the properties of

the delta function 5(x), and we show how by using the delta function one
may readily and simply understand the properties of Fourier integrals.
Our presentation will be sufficient for practical purposes; a mathemati-
cally rigorous treatment may be found in pages 15 to 29 of the book by
Lighthill.f
The delta function

The delta function is an example of a generalised function or distribu-
tion. The simplest way to understand its properties is to use a limiting
procedure. Define the function
A£(x) = (l/eV5r) e~x /£

(e > 0). (B.I)
For each value of the constant parameter e, this function satisfies
a:) = l. (B.2)
When plotted against #, as in figure B.I, the function A£(x) has a peak
at the origin. This peak has height 1/ey/n and width of order e (exactly
how we define the width does not matter), so that if e is allowed to become
very small the peak becomes very tall and very narrow. Outside the peak,
the function becomes extremely small. This means that if we define an
integral
poo
Af] = / dxA£(x)f(x), (B.3)
J -c
then for a wide class of functions f(x) the value of the integral when e
is very small depends almost entirely on the values that f(x) takes very
M J Lighthill, Introduction to Fourier Analysis and Generalised Functions (Cambridge

University Press, 1958).
162
The delta function 163
Figure B.I. Plot of the function A£(x) against x.
close to the origin, and ultimately
lim/ e [/] = y dx Ae(x)/(0) = HO). (B.4)
Here we have used (B.2).

A crude definition of the delta function is
S(x) = lim A£(x). (B.5)
The mathematically rigorous definition is arrived at via (B.3) and (B.4):
dxS(x)f(x)=f(0) (B.6)
for a suitable class of 'test functions' f(x). The test functions are ordi-
nary functions. Different types of generalised function or distribution are
defined by using different classes of test function, for example tempered
distributions are defined using functions f(x) that, together with all their
derivatives, are bounded by some fixed power of x at x = ±00.
By making a simple change of variable in (B.6), one obtains
dx 6(x - a)f(x) — f(a). {B.I)
164 The delta function and Fourier transforms
Figure B.2. Plot of A'€(x) against x.
This may be compared with the property of the symbol
Hence 6(x — a) is a generalisation of 8{j to the case where its discrete

suffixes become continuous variables.
From the way in which we have arrived at 8(x)1 it follows immediately
that S(x) = 0 except at x = 0. At the origin, 5(x) is positive infinite, in
such a way that for any positive a and b,
dxS(x) — 1. (B.8)
J —<
This is why it is not a function in the ordinary sense.

One may define also the derivative 5'(x) of 8(x), and higher derivatives.
When e is small, the derivative of A£(x) has two peaks close to the origin,
one positive and one negative, as drawn in figure B.2. As £ —> 0, each of
these peaks becomes very narrow and very tall, and the two peaks each
approach very close to the origin. Now, an integration by parts gives
/ ]!!'0O- / dxA£(x)f(x). (B.9)

J— J—
The delta function 165
1—
Figure B.3. Plot of @£(x) against x.
Because of the explicit definition (B.I) of Ae(x), the first term on the
right-hand side vanishes unless f(x) explodes violently at infinity. So, by
letting e —> 0, we arrive at the definition of 8'{x)\
pCO pOO
I dx8'(x)f(x) = - dxS(x)f'(x) = -f'(0). (B.10)

J-co J-co
We have used (B.6) for the case where f(x) is replaced by f'(x). Higher
derivatives of S(x) may be defined similarly. Because the rth derivative
of A£(x) changes sign r times very near the origin when e is very small,
successive derivatives of 5(x) are more and more singular at the origin.
Consider now the indefinite integral of Ae(x),
S£(x)= f dyA£(y). (B.ll)

J — oo
This is plotted against x in figure B.3. As e -> 0, the step in this function
becomes progressively steeper, until finally the function changes abruptly
from 0 to 1 at the origin. In fact
dy6(y)=6(x), (B.12)
where
6(x) =
0 x<0.
Notice that the value of an integral such as (B.6) is unchanged if we change

the value of its integrand by a finite amount at not more than a finite
number of points. Hence two generalised functions that differ by a finite
amount at a finite number of points are equal in the sense of distribution
theory. This means, in particular, that it does not matter what value we
assign to 6{x) at x = 0; any finite value will serve, but 0(0) = \ is perhaps
the most natural definition.
If we differentiate (B.12) with respect to x, we obtain
d6/dx = S(x). (B.13)
Thus distribution theory allows us to define the derivative of a function

that has a finite discontinuity. Consider the function F(x), defined to be
equal to the ordinary differentiable function f(x) for x < a, and equal to
the ordinary differentiable function g(x) for x > a (see figure B.4). We
have
F(x) = /Or) + [g(x) - f(x)]0(x - a), (B.14)
and so, applying the usual rule for differentiation of a product of functions,
F'(x) = f'(x) + [g'(x) - f'(x)]6(x - a) + [g(a) - f(a)}S(x - a). (B.15)
In the last term, we obtain initially [g(x) — f(x)]5(x — a), but we may
replace x by a inside the square bracket because the delta function vanishes
unless x — a.
Fourier transforms
In our discussion of the delta function, we used a particular function
Ae(ai), defined by the Gaussian function (B.I), which has a peak that
becomes progressively taller and narrower as e —> 0. There is nothing
special about this particular choice of A£(x); almost any function having
this property will be just as good, provided it is chosen so as to satisfy
the integral condition (B.2) for each value of e. Another example is the
function
Consider now the integral
Fourier transforms 167
f(x)
Figure B.4. Plot of F(x) aganist x.
This is simple to evaluate by elementary means: it is just
/ d k e ikx+ek
+
J-
1
IX + £ IX — 6
= 2TTA£(X). (B.17)
To proceed further in a rigorous fashion, we should multiply both sides

by a test function f(x), integrate from —oo to oo with respect to ±, and
then let e -> 0. However, in the same informal sense that we understand
the delta function to be defined by (B.5), we write
£W
lim = 2TT6(X). (B.18)
It is usual even to write
(B.19)
The integral here does not converge in the usual sense. It is understood
that, as in (B.18), a damping factor e~£^ is to be included in the inte-
grand so as to make the integral converge, and after the integral has been
calculated e is allowed to tend to zero.
Consider now the function $(fc), defined by the integral
[ (B.20)
V27T J-
for some suitable function <j>(y). If (j)(y) -> 0 sufficiently rapidly as y ->
±00, this integral converges in the usual sense. Otherwise it is necessary
to include in the integrand a damping factor e~£^ and let e —> 0 after the
integral has been calculated. If <j>(y) is bounded by some power of y as
y -> ±00, this procedure will give a convergent result. We now calculate
1
4
Again, a damping factor e~e\h\ may be needed in order to make this
integral converge. We insert the representation (B.20) for $(k) and obtain
-£= (°° dke-'lkx$(k) = ~ f

27r
V2TT J-00 7-
where we have used (B.19) and then (B.7).

Putting (B.20) and (B.21) together, and changing the dummy variable y
of integration in (B.20) to x, we have
dxeikx(f)(x)
(B.22)
lkx
(f)(x) = -4= I dke-' $(k).
This is the Fourier inversion theorem: if one of the relations holds, so does
the other. We say that the functions $ and <j> are Fourier transforms of
each other.
It may be that both the integrals in (B.22) converge without needing any
additional damping factors. This is the case of classical Fourier trans-
forms. Most of the textbooks consider only classical Fourier transforms;
in this case it may be shown that both $(k) and (p(x) are ordinary func-
tions. The possibility of including damping factors allows the very pow-
erful extension of Fourier transform theory to include the possibility of $
Fourier transforms
or <j> being a distribution rather than an ordinary function. In the manip-

ulation in (B.21) we have implicitly interchanged the order of the k- and
y-integration, and when damping factors are needed we have interchanged
doing the integration with taking the e -¥ 0 limit. It is possible to show
that if either $ or 0, together with all its derivatives, is bounded by some
power at infinity, then these interchanges are justified. Further, it may
be shown that if either one of the functions $ or <fi has this property at
infinity, then so does the other.
Notice the factor 1 / ^ T T in front of each of the integrals in (B.22). Often,
one chooses to define the Fourier transforms instead by writing 1/2TT in
front of one of the two integrals; then no factor appears in front of the
other. However, we shall use the definition (B.22).
Simple examples of Fourier transforms
(i) S(x) <—> -=. (B.23a)
For, because of (B.6),
1 f°
\/2W-oo V2^
The inverse relation is obtained from (B.19), by replacing the dummy

variable k of integration by —A;.
(ii) e-*'/a» <_> « e-'*V4. (B.236)

v2
This is an example of a classical Fourier transform: no damping factors

are needed. We have
2
o o-a fc2/4 r
^2^ J_
To perform the integration here, make the change of variable (x - \\ka2) — y

and use the standard integral
dye-y la = a v ^
(Strictly, after the change of integration variable the y-integration is no

longer along the real axis in the complex y plane. However, the integra-
tion may be restored to the real axis by using Cauchy's contour-integral
theorem.)
(iii) For any pair of Fourier transforms </>(#),
(B.23c)
This is proved immediately, by differentiating the second relation of (B.22)

repeatedly with respect to x.
(iv) 0(x) <—> lim -4= ( - ), (B.23d)
where 6(x) is the step function defined in (B.12). For, putting in the
necessary damping factor,
--L f dxeikx9(x) = lim - - L I dxeikx~£X,
and this integration may be performed by elementary means. Notice that

because the result is singular at k = 0, it is important to retain the ie in
the answer. To illustrate this, we show how to verify the inverse relation.
We have to evaluate
1 f°° ., i / 1 \
lim dfce
. -/*= / ^ = TT7Z ' ( B - 24 )
The integrand is analytic throughout the complex k plane, except for a

pole at k = -\e. As e —> 0, this pole approaches the origin, but the term
Fourier transforms 171
/ \
/ \
/ \
/ \
I /
\ ^Pole /
\ /
\
Figure B.5. Integration contours for the integral (B.24).
ie tells us that it approaches the real axis, along which lies the contour of
integration, from below rather than from above. See figure B.5. Suppose
now that x < 0. Then the exponential factor in the integral (B.24) vanishes
as k —> oo in the upper half of the complex k plane sufficiently rapidly
that the integral taken over an infinite semicircle in the upper half-plane
would be zero. Hence we may add this contour to the original integration
contour along the real axis, without changing the value of the integral.
The integration contour is now closed and, because the pole is in the
lower half-plane, the integral vanishes by Cauchy's theorem. When x > 0,
we instead add on to the original integration an integration over an infinite
semicircle in the lower half-plane. The integral is then evaluated by taking
the residue at the pole, which now lies within the contour.
(v) Let </>i(x) and fafe) be the Fourier transforms of and
respectively. Then
4= f dk> (B.23e)
This is known as the convolution theorem. The proof is as follows:
J_
dk2 5(k - h - k2)
and this integral is equivalent to the integral in (B.23e).

Finally, we prove Parseval's theorem. This is that, for any pair of trans-
forms (j)(x) and $(&),
POO /*OO
/ dx\</)(x)\2= / dk\${k)\2. (B.25)

J— oo J — oo
For the left-hand integral is

/»OO
/ dx<f)*(x)<t>{x)
J — oo
/•o
= /
J—
Appendix C
Orbital-angular-momentum operators
The orbital-angular-momentum operator is defined in (5.2):

hL = r A(-ifiV), (C.I)
so that
Lx = -i(yd/dz - zd/dy),
Ly = —\(zd/8x - xd/dz),
Lz = -i(xd/dy-yd/dx). (C.2)
In spherical polar coordinates (see figure 5.1)
X = p COS (/>,
y = psin</>,
z = pcot9, (C.3)
where p = r sin#.
To calculate Lz in spherical polar coordinates, we need
^ + +
dx dx dp dx 89 dx dcf)
and the similar expression for d/dy. Here the partial derivative 8/dp
is calculated keeping 9 and (j) constant, 8/86 is calculated with p and <£
constant, and d/dcj) is calculated with p and # constant. On the other
hand, each partial derivative with respect to x is calculated with y and z
constant. Prom (C.3),
dx = cos (f)dp — ps'mcf) d(/>,
dy = sin <f> dp + p cos <f> d</>,
dz = cot 9 dp - p cosec2(9 d0. (C.5)
Solving these three simultaneous equations for dp, d<j> and d0 we have
dp = cos (/> dx + sin (/> dy,
pdcj) — — sin 0 dx + cos 0 dy,
pd9 = sin 0 cos 0(cos 0 dx + sin 0 dy) — sin2 0 dz. (C.6)
173
174 Orbital-angular-momentum operators
To calculate the partial derivatives with respect to x, we set dy = dz = 0,

and so have
— = cos</>, 7— = - sin6cos0cos </>, — = —sine/). (C.7)

ox ox p ox p
We insert these in (C.4), and calculate d/dy similarly. Using these expres-
sions, with x and y from (C.3), we calculate Lz from (C.2). The result is
given in (5.5). The other two components of L may be calculated in the
same way, so that altogether
Lx = i(sin (j)d/de + cot 9 cos (j)d/d(j)),

Ly — —i(cos (j> d/dO — cot 6 sin <j> d/dcj)),
Lz = -id/d<f>. (C.8)
It is then straightforward to evaluate
L = LXLX + LyLy + LZLZ)
with the result given in (5.3).
Appendix D
Electrodynamics
The electric field £ and the magnetic field B obey four equations known
as Maxwell's equations. Maxwell wrote down these equations before either
special relativity theory or quantum mechanics were known, but neither
of these two developments led to any need to change the equations. They
are
V • £ = 0,
V • B = 0,
V A£ = -B,
where the dots, as usual, denote partial derivatives with respect to t. We

have written the equations for fields propagating in empty space. In the
presence of electric charges or of a magnetically or electrically polarisable
medium, the equations must be modified. The constants So and no are
related to the speed of light c by
c2 = 1/eoMo. (D-2)
A general theorem of vector calculus allows one to deduce from the second
of Maxwell's equation that, at least locally, B may be expressed in terms
of a vector potential A:
B = VAA. (D.3)
By calculating V • B , it is straightforward to verify that the second equa-
tion of (D.I) is indeed satisfied if B is expressed in this way, but the
converse result is also valid.
If we insert (D.3) into the third of Maxwell's equations, we find that
V A (£ + A) = 0. (D.4)
Another general theorem of vector calculus then allows us to deduce that

£ + A can be written as the gradient of a scalar potential (f>:
(D.5)
175
176 Electrodynamics
It is again straightforward to verify that if S is written in this way then

it automatically satisfies (D.4), and again the converse is also true.
If we take some arbitrary scalar function f(r,t) and change A and <j>
together in the following way,
A -> A + V / ,
then the fields B and £, given by (D.3) and (D.5), remain unaltered. The
fields B and S are physically measurable quantities, but the functions A
and (j) are not directly measurable. Hence the transformation (D.6), which
is called a gauge transformation, can have no physical consequences. This
freedom allows one to impose on (p and A certain constraints known
as gauge conditions, and one example of a possible gauge condition is
V • A = 0, which is the condition introduced in chapter 8.
In (8.18) we wrote the Hamiltonian for an electron of charge — e in inter-
action with an electromagnetic field:
H = V + (p + eA)2/2m - ecp. (D.7)
In classical physics, the Hamiltonian leads to equations of motion through

the elimination of p from Hamilton's equations (8.15), which are
dr_dH dp_ dH
Here, the operator d/dr is usually written as V, and d/dp denotes the
corresponding differentiation with respect to p. The differentiation d/dr
is carried out with p kept constant, and r is kept constant for d/dp. If
we apply Hamilton's equations to the Hamiltonian (D.7), we obtain
r = (p + eA)/m,
d e
at VV m {[(p + eA) • V)A + (p + eA) A (V A A)} + eV</>.
(D.9)
To write the second of these equations, we have used the vector identity
Vu2 = 2(u • V)u + 2u A (V A u).
The first equation in (D.9) gives
(d/dt)p = mr-e(d/dt)A
= mr -eA- e(r • V)A. (D.10)
Electrodynamics 177
Here, we have used the fact that A generally varies with both t and r,
so that its total derivative dA/dt is not simply equal to dA/dt = A,
but rather dA/dt — A + (r • V)A. We reduce the right-hand side of the
second equation of (D.9) using the first equation of (D.9) and also (D.3)
and (D.5), and so obtain the equation of motion
mr = - W - eS - er A B. (D.ll)
Appendix E
Bloch waves
The Hamiltonian that describes the motion of an electron in a one-

dimensional model of a crystal is unchanged when the electron coordinate
is displaced through a distance L:
H(x + L) = H(x). (E.I)
We define the translation operator D such that for any function f(x)
Df(x) = f(x + L). (E.2)

This definition applies only to functions / , not to operators such as H.
However, Hf is a function, and so we have
DH(x)f{x) = H(x + L)f(x + L)= H(x)Df(x),
where we have used (E.I) and (E.2). Hence, for any function f(x),
[D, H{x)]f(x) = (DH(x) - H(x)D)f(x) = 0. (E.3)
If an operator gives zero when it is applied to any function /(#), this is

equivalent to saying that the operator vanishes. Thus (E.3) gives
[D,H] = 0. (E.4)
As we explained at the end of chapter 4, this implies that the operators D

and H have a complete set of common eigenstates; that is, the stationary
states can be chosen to be eigenstates of D. Let %j) be any one of these,
so that
Dt/){x) = &l)(x), (E.5)
where c is the eigenvalue of D and is a constant. Because the operator
D does not correspond to an observable, it need not be Hermitian, and c
can be complex.
Because of the definition (E.2) of JD, (E.5) gives
). (E.6)
178
Block waves 179
The properties (E.I) and (E.6) are valid for all pairs of points (x,x + L)
that lie within the crystal. Suppose that the crystal extends from x = —oo
to x = +00, so that (E.6) implies that
/•OO 1 /*O
,tl,)= d^|V(^)|2 = 0 2 / (E.7)

J-00 \c\ J-
By making a change of integration variable from x to (x + L) we also have

the identity
/•OO /«OO
/ dx\ip(x)\2= dx\ip{x + L)\2. (E.8)

J — oo «/— oo
Together, (E.7) and (E.8) give |c| 2 = 1, so that c = eia with a real.
Without loss of generality, we may set a = kL:
c = eifcL, (E.9)
where k is real. According to (E.6) and (E.7), each time that we displace
the coordinate x through distance L the wave function is multiplied by
elkL. So for any positive or negative integer m
<iP{x + ml) = eikrnLip(x). (E.10)
This relation is known as the Block condition. If we choose to write the

wave function in the form (10.3), namely
xl>(x)=eikxuk(x), (E.ll)
it is straightforward to see that the Bloch condition requires that

Uk(x + mL) = Uk{x), so that Uk(x) has the periodicity L of the lattice.
In order to derive the result that k is real, we needed (E.6) to be true for
all values of x. This is not so when the crystal is taken to be of finite
length. Then there are also solutions with k complex. In the case of such
a solution, write k = k\ + i&2, with ki and &2 real. The wave function
(E.ll) then contains a factor e~k2X. Depending on the sign of &2, this
is largest at one or other of the two ends of the crystal, and it decreases
exponentially with increasing distance from that end. Unless the crystal is
very small, the wave function is negligibly small over most of its interior.
For this reason, the solution is called a surface state. It turns out that
surface states make a negligible contribution to the bulk properties of all
but the very smallest microcrystals, and we have ignored them in this
book.
Hints for the problems
Chapter 1
1.1 5 x 1029 photons per second.
1.2 Work in the frame in which the electron is initially at rest, so that its
energy is me2. Let the momentum of the photon be p; according to
(1.4), its energy then is cp. So if the photon is absorbed, the electron
afterwards has momentum p and energy E = me2 + cp. This means
that E2 > m2c4 + c2p2, in violation of (1.36).
1.3 The wavelength of visible light is of the order of 500 nm, so for a mass
of 1 kg the answer is about 3 x 1020 years, and for an electron about
10"3 s.
1.4 Suppose that after the scattering the photon has momentum p'. Then
according to (1.4) its energy has changed from cp to cp1. Applying
(1.36) to the electron after the scattering gives
me2 + cp- cp1 = c[m2c2 + (p - p')2]1'2.
Squaring both sides, using p • p' = pp' cos 0, and dividing through by
2mc3ppf/h gives the answer.
1.5 The calculation is similar to problem 1.4.
Chapter 2
2.1 According to (2.76), adding a constant VQ to V(r) is equivalent to
instead subtracting the same constant from E. Hence, according to
(2.9), the stationary-state solutions become multiplied by eiVot/h. This
leaves unchanged the probability density |\I/|2 for all r and all t.
2.2 The allowed values of \mv2 are given by (2.136) so the allowed values
of v are nhn/ma. We need the difference between the values of v for
n = 2 and n = 1. For the electron, this is about 3 x 106 m s" 1 , and for
the tennis ball it is about 10" 33 m s" 1 . In the case of the tennis ball,
the answer is negligibly small: the velocity is effectively a continuous
variable and quantum mechanics need not be used to describe the
motion.
180
Hints for the problems i
2.3 Cn = (2/a) 1 / 2 (times an arbitrary complex number of modulus 1).
(x)= I" x\ipn(x)\2 dx

Jo
({x-(x))2) = (x2)-(x)2-, (x2) = [x2\^n(x)\2dx.

Jo
Classically, the particle is equally likely to be found anywhere within

the well and so
1 fa 1 fa
(x) = - / xdx = | a , {x2) = - x2dx = \a2
a
a Jo Jo
and so ((x — (x))2) = jâ2.
2.4 In one dimension,
"" ' ' 2mi V dx dx
In a stationary state, &(x,t) = ip(x)e~lEt/h, and therefore
where if)1 = di/j/dx. This is manifestly independent of t. To show that

it is independent of x, calculate dj/dx and use the time-independent
Schrodinger equation to show that it vanishes.
In three dimensions both p and j are independent of t. Applying the
divergence theorem to (2.17) gives V • j = 0 . This does not imply
that j is independent of r.
2.5 Normalise ij) such that Jd 3 rid 3 r2 |^( r i 5 r2)\2 = 1, where the integra-
tions over both r\ and V2 extend over all space. Then the integrand is
the probability of finding the first particle in the volume element d 3 ri
and the second in d3r2. The Schrodinger equation is
where V? is the V2 operator constructed from r» (i = 1,2).
182 Hints for the problems
If the particles do not interact with each other, V = Ui(ri) +

Then there are separable solutions i/;(ru r2) = ^ i ( r i ) ( )
2rrii
and E = Ex + E2.
Chapter 3
3.1 According to (3.11), within the well i/> oscillates sinusoidally, while
outside the well it goes exponentially to zero. At the two edges of the
well, both ij) and dip/dx are continuous. From (3.10), the lowest levels
correspond to the smallest allowed values for ft and so, from (3.116),
their wave functions have the smallest number of oscillations.
If the values of U and b are such that in a given state the value of a
is very small, the exponential fall-off of the wave function beyond the
edges of the well is very slow, and then there is a good chance that the
particle will be found outside the well.
3.2 There must always be at least one even-parity solution of (3.13) —

see figure 3.2. As U decreases, the dashed curve moves downwards
and the odd-parity solution disappears when its intersection with the
cot 2 |/36 curve is where the latter touches the (3 axis, that is when it
is at (3 = TT/6.
3.3 With the changes of variable x' = x — eE/muP1 and E' = E +
e 2 £ 2 /2raa; 2 , the new Schrodinger equation becomes just the same
as the old one (3.15). So the new wave functions are just ipn(xf),
where ^n is the function in (3.17), and the new energy levels are
2 2 2
\
3.4 Prom (3.17), the wave functions are Hermite polynomials Hn(x) multi-
plied by an exponential factor. Prom appendix A, HQ(X) is a constant
and (see (A.5)) H2(x) = a o (l - 2mux2/h) with a 0 a constant. To

verify the orthogonality, use / ^ dxe~Xx = (TT/A) 1 / 2 and
j°° dxx2e~Xx2 =-± (^ dxe~Xx\
3.5 Use (3.27) and the definitions
h2k2/2m = E, ft V/2ra = U-E
to show that in case (i) \T\2 « : \A\2, and in case (ii) \T\2 « \A\2.
Notice that (3.27) is valid also when U is negative, as in (i); then n is
imaginary.
3.6 For x 7^ 0, the Schrodinger equation is i/>" = a2tp, where
E = -h2a2/2m. We choose the solutions Ae~ax in x > 0 and Beax
in # < 0, so that ip —> 0 as x —> ±oo as required for a bound state.
This requirement also picks out the case E < 0, so that a is real,
rather than E > 0. At x = 0, I/J must be continuous, so that A — B.
The other condition at x — 0 is (3.35), with A replaced by —/i. This
gives a = m/i/h2, and so E = —m/j,2/2h2. When [/ —>> oo, (3.10) gives
/? -> oo such that [/ - h2f32/2m. Hence if C/6 = /i, \fib - mn/h2f3.
Thus in (3.12a) the right-hand side becomes mfi/h2, but in (3.126) it
diverges and gives no finite value for a.
3.7 The first step is achieved simply by replacing x by —x throughout
the Schrodinger equation (always valid for any equation!) and using
V(x) = V(—x). Because each value of E is supposed to correspond to
only one independent wave function, it must be that ^{—x) = Cfi/j(x))
for some constant C. Replacing x by —x in this equation, we see that
ip(x) = Cip(-x). Combining these two equations gives C2 = 1, so
C = ±l.
3.8 The mass of the a-particle is about 4 times that of the proton, that is
M « 3750 MeV/c 2 , and e2/he0 = 4TTC/137. SO (3.456) gives for nuclei
of the same radius r oc E~ 1 / 2 exp(4.0Z J E~ 1 / 2 ), where the a-particle
E is in MeV. This gives r\j/rFm w 5 x 1013 and hence T\J W 3 x 109
years. The experimental value is 4.5 x 109 years, so the simple model
gives a surprisingly accurate result.
Chapter 4
4.1 /»OO
= / dx \&*(—ihfy1).
J—oo
= UP - (P))2) = (p2) - (p)2

/•OO
r
I O II / v
J — oo
4.2 #(z,t) = Nie-(x-<x))2MAx)2eL~, where
JVi = (a2 + 2iM/m)'1/2{ir/2a2)'1/A
and e1- is a phase factor that varies with both x and t.

4.3 The operator corresponding to px is —ihd/dx, so (px) = — ih J dx ip*if)'.
Write tp(x) as the Fourier transform of 0(p) and differentiate it to write
^'(o;) in terms of the Fourier transform oi pcf>(p). Hence show that
(px} = J dp \(f)(p)\2. A similar approach may be used for the expecta-
tion value of any power of px.
4.4
— (x) = — dx ty*xV = / dx(4/*x^ + **x*).
at at y J
From the time-dependent Schrodinger equation, this is
f
Integrating the first term by parts twice, and using the fact that for
a wave packet the wave function ^ vanishes at x = ±00, gives the
result. The second part is done similarly.
4.5 (x) and (p) are zero because of the symmetry about the origin.
Because (p) and (x) are zero, this is
(Ap)2/2m+±muj2(Ax)2.
So, from the uncertainty relation (4.13),
(E) > h2/8m(Ax)2 +
The minimum value of the right-hand side is when (Ax)2 = h/2mu,

giving (E) > \fiuj.
4.6 The maximum accuracy is achieved by attempting to drop the dart
from directly above the target point, but the uncertainty principle
implies that in practice the point of release is displaced through a
horizontal distance Ax and the dart has a horizontal component of
momentum Ap (in the same direction) at the time of release, where
AxAp > \Tt. Since the time to drop a height I is (21/g)1/2, the dart
hits the ground at distance
R = Ax + (2l/g)1/2Ap/m > Ax + (2l/g)1/2h/2mAx.
The minimum occurs for Ax = (lh2/2m2g)llA, soR> (8lh2/m2g)1/A.

4.7
4.8 The identity (4.16) is proved by integration by parts.
ft2 f 3 h2 f
= -— d rV-(rV$) + - - /
2m J 2m J
The divergence theorem transforms the first integral to an integral
over a surface that encloses the system so that \I> vanishes over
this surface; this integral therefore vanishes. The second integral is
(h2/2m) J d 3 r | V#| 2 , which is positive. It follows that in each bound
state
(V) = Jd3r V\*\2 > Jd3r

where Vo is the minimum value of V.
4.10 This follows immediately from (4.16) and the fact that q must be
real, because Q is Hermitian.
4.12 By taking linear combinations of the given wave functions, and ap-
plying Q to them, one finds that the complete set of eigenfunctions
is
fa = (fa -
The first of these has eigenvalue +1, so the system is left in this state
after the first measurement. At a later time t the wave function is
lElt/h
e~lElt/h
= (</>!e~' elE2t/h)/V2.
I/J2 2 e-'lE2t/h
++ I/J
The required probability is
Cos[(E1-E2)t/h}}.
4.13 This follows immediately from the time-dependent Schrodinger equa-

tion.
4.14 For a free particle, H = (p2x + p2 + pi)/2m. The only part of this
which does not commute with x2 is the first term, so ihd(x2)/dt =
([# 2 ,p 2 ] )/2m. The commutator may be simplified as in problem 4.11.
To differentiate again with respect to i, evaluate the commutator of
the result for d(x2)/dt with H/ih.
4.15 (i) is just the definition (4.16); (ii), (iii) and (iv) follow immediately
from the definition of the Tsymbol given in the question.
The expectation value in the state * is (*, F^F^), where F = Q-iXR
and so, because Q and R correspond to observables and therefore are
Hermitian, F t = Q + iXR. From (iv), (V,F^FV) = ( $ , $ ) , which
from its definition / d 3 r | $ ( r , t ) | 2 is real and > 0. Multiplying out
the terms in pip,
($,$) = (Q2) -i\([Q,R}) - X2(R2).
This expression has its minimum when A = — |i( [Q,R] )/(JR2),

whence the required inequality follows. The uncertainty principle
(4.27) is obtained from this inequality by making the replacements
Q Q ( Q ) R R ( )
Chapter 5
5.1 Under the rotation, 0-^6 and <f> -» <j> — a. So, according to (5.7), YJm
is multiplied by e~ imQ .
5.2 e2/47re0r2 is a force, so dim (e2/47re0) = ML 3 T~ 2 . The momentum
operator is (-iftV), so dim (h) = ML 2 T" 1 .
a0 = (47rsohc/e2)(h/mc) « 5 x 10~ n m.
5.3 The force towards the centre of the orbit is mv2/a = e2/4neoa2. So
TiL = mva = (mae2/A-KSQ)1!2, = rih say. The corresponding energy
is
\mv2 - e2/4ne0a = -e2/8n£0a = me4/32TT2
as in (5.17). When n = 1, the velocity is e2 /A-KSQU = c/137, so that
v <ti c and non-relativistic kinematics are justified.
5.4 When the wave function is spherically symmetric, the probability
that the distance is between r and r + dr is P(r) dr, with P(r) =
Anr2 \ip(r)\2. The most probable distance is where P(r) is largest,
which is at r = 2CLQ.
5.5 Applying L2 in (5.3) to ip multiplies it by 2, so that according to (5.6)
1 = 1. Also (5.6) shows that m = - 1 . (5.13) and (5.14) yield
\Mr 2M Aneor-EJ
so that /^ = l/2ao, where ao is given in (5.18), and £7 is as given in

(5.17) with n = 2.
5.6 When / = 0, the wave function depends on r only. According to (5.13),
it satisfies
[rR(r)]" + ^[rR(r)] = -U6(r - a) [rR(r)].
For r 7^ a, the solutions to this equation are e±KT where

n — (2mE/h)1^2. Since we want a bound-state solution, where R -> 0
as r —> oo, and since, for the reason explained in the text, we require
R to be non-infinite at r = 0, we take
«r r>a,
Kr
-eKr) r<a.
Atr = a,R must be continuous, so that Ae~Ka = B(e~Ka—eK'a). Also,

from the differential equation, disc [rR(r)]f = —U[rR(r)]
r=a
(compare with (3.35)). So
-KAe~Ka + KB(e-Ka + e«a) = -UAe~Ka.
Eliminating A and B gives the equation that determines the energies

of the bound states: /^(coth/â — 1) = U.
5.7 An extra term —me2b/27reoh2r2 appears in the square brackets in

(A.7) and (A.9); the substitution (A.8) is the same. The square
bracket in the indicial equation (A. 11) acquires the extra term
—me2b/2ireofi2 = — 26/ao, where ao is the Bohr radius given in (5.18).
The required root of the indicial equation is a = I + 1 — D(l). The
denominator of the right-hand side of (A. 12) has an extra term, but
the numerator is still the same and so again me2/47reoK,h2 — a must
be an integer N = n — I — 1.
5.8 By analogy with (5.20), write r = x\ — X2 and R = $x\ +£2). Then,
with /ip{xi^X2) — / 0i(r) ^ ( i ? ) , when the centre of mass is at rest we
have
m dr2
and ip2(R) = 0 (see (5.23)). However, the first of these equations
is obtained from equation, (3.15), for the linear harmonic oscillator
by making the changes m —> \m and u —> y/2cu. So the energy
eigenvalues are E = (n + $THJO\/2.
5.9 Equations (5.27) and (5.29) give
1
cot Ka = — — = —
K
which « 0 if \E\ < U and so Ka w TT/2. This gives U « 26 MeV.
Chapter 6
6.1 For x 7^ ±i?, the Schrodinger equation is if)" = K2ip, where
E — — TI2K2/2m. The potential is symmetric about x = 0 and so
(see problem 3.6) the wave functions have definite parity: either
tp(-x) = ij)(x) or tl>(—x) = -ip(x).
With such wave functions, imposing the appropriate continuity and

discontinuity conditions at x = ±R guarantees that they are satisfied
also at x = —R. At x = ±i?, ip is continuous and disc if>' = —Uip.
Hence Be~«R = A(eKR ± e~KR) and
KR
- nA{eKR T e~KR) = -UB e~KR.
Eliminating A and B gives U - n = K tanh KR or K coth KR.
6.2 The equilibrium value of R corresponds to the minimum of the integral

V(R) given in (6.12). According to (6.1), the only length scale in the
problem is that set by ao, so the required value of R is proportional
to ao. From (5.18), the muon-bound molecule is therefore some 200
times smaller.
6.3 As in (6.10a), for large R we have Hn « EOi and H22 ~ E02. Elimi-
nate c\ and c2 between the two equations (6.4). The resulting equation
for E has one solution near E\ and the other near E2. To examine
the solution near E\, write E = Ei + S where S is assumed small, and
insert this in the equation. This gives that (E2 — E{)5 is quadratic in
small quantities.
6.4 With the obvious notation, the symmetry of the problem results in
•Hii — H22 = i?33 and H\2 = H2\ = H2% = H$2 = H\$ = H%\.
Write (j) — ci<f>\ + c2cf)2 + c^fo and obtain the equations correspond-
ing to (6.4). Eliminate c\,c2 and C3 by making the determinant of
their coefficients vanish. The resulting equation for E has two de-
generate solutions E = (Hn — Hi2)/(l — Ki2) and a third solution
E = (Hn+2H12)/(l + 2K12).
Chapter 7
7.1 Take the z-axis in the direction of £. In (7.1), take Ho as the Hamil-
tonian before the field is introduced and Hi = e£z. With the ground-
state wave function (6.1), the energy shift (7.7) is
AE=(na30)-1 f
= (2/a30) r ° d r r 2 e - 2 r / a o / d(cos0) eSr cos 9.

Jo J-i
The cos# integration gives zero, so there is no first-order energy shift.

7.2 Prom appendix A, H0(x) is a constant, so the normalised unperturbed
ground-state wave function is Vo(#) = (nh/mu)'1ê~mujx2l2h. The
energy shift to first order is AJ5 = J dx ex2ifj% = eTi/2mu). The per-
turbed Hamiltonian \sp2/2m+\mJ2x2, where J2 = u2+2e/m. The
exact value of the new ground-state energy is \Tiu' = \hw + eTi/2mu)
+ higher-order terms.
7.3 (i) Let V be an eigenfunction of the Hamiltonian H0 + Hi, and write it
as the linear combination cifa +c2cf)2. Prom the Schrodinger equation,
- Ei)fa
To this equation, apply in turn the operations Jd3r<f>l and /d 3 r(/> 2 ,
using the orthonormality of fa and fa (see (4.18)). Eliminating c\ and
c2 then gives
(fa, Hifa)+ E (fa, Hi fa)

((/>2, Hi fa) (fa^Hifa) + E2 — E
This is a quadratic equation for E, with two solutions. Expanding the
square root in the solutions gives, to first order, the same result as
perturbation theory.
7.4 According to (3.4) and (3.5), the state of energy 3h27r2/2ma2 has wave
function
fan — /a) sin(7n//a) si
The perturbation —e£(t)x can induce a transition to the state whose

wave function is
^211 = (8/a3)1/2sin(27rx/a) sin(7ry/a) sin(7r^/a),
but not to either of the other two states of energy 3h27r2/ma2. The
transition probability is calculated from (7.22), with
t'-t'2/T2
-/-,
(see (B.236)).
7.5 The normalised wave functions are ipo(x) = (a/7r)1/4e~2aa;2 and
•0i (a;) = (27r) 1 /2( a / 7r )3/4 ;re -iax 2 ) w h e r e a _ muJ/fl T h e required
probability is
1°
J —c
f
2 2 2
= (2a2v2\2/h2) dtte -a v t +iut
J —a
The integral is
(see (B.236)).
7.6 The probability is given by (7.31), with fa = (2/a)1/2 sm{nx/a) and
0'fc = (l/a) 1 / 2 sin(7rx/2a). The integration is from 0 to a, and the
answer is 32/9TT2.
Chapter 8
8.1
[Lx,Ly] = -^[VPz ~ ZPy, ZPx ~ Wz]
a
Z]px+X[z,pz]py}
8.2 Solve
-o C
C
1 0) \u2
giving U2 = cui and u\ = CU2- So the eigenvalues of ax are c = ±1, the
same as those of a z . The corresponding eigenstates are represented
by the normalised spinors
0
.3 With 5 = | a , the matrix representations (8.11a) give
s+ s
-{o oj' --{i o
and when these matrices multiply the spinors given in (8.11b) the
results given for S-u^ and S+u± follow. Also S-u^ = S+u^- = 0.
.4 The possible results of measuring Sz and Sz are t t , J 4 ? t i a n d
It, for which Sz has eigenvalues +1,-1,0 and 0 respectively. If the
eigenvalues of S2 are s(s + 1), then according to (8.5) the allowed
values of s axe | | — | | and ( | + | ) .
Write S2 = S2Z + \{S+S- + 5-5+) with 5± = 5 ^ + S^. Use
= 0, etc
to derive 5 2 x+ = 2x + (so that 5 = 1) and S2x~ = 0 (so that 5 = 0).

8.5 Choose the z-axis to be along £?, and the a:-axis so that B' is in the
xz-plane. Then the particle is initially described by the spinor
••(!
and the addition to the Hamiltonian caused by switching on B' is
HUB'
The energy shift is u^Hiu = —/jihB'/2y/2. The initial energy was

| , while the exact final energy is
l
2 2 2
-\\& \B + B'\ = -±nhB ((1 + B'/BV2) + (B'I By/2) ) ,
and the square root may be expanded in powers of B' jB by Taylor's

theorem.
8.6 Initially the spin state is represented by a spinor that satisfies ayu = u.
With the representation for cry
where the factor l/\/2 normalises the spinor to unity. The possible
results of the measurement are the eigenvalues of |cr x , namely ± | .
The corresponding spinors u± obey ayu± =± u± and are
The two probabilities are calculated from (4.22) and are each equal
to | . A further measurement of the y-component of the spin will sim-
ilarly give one of the two results ± | , each with probability | .
(Of course, usually the probabilities will not be equal; we have ob-
tained this result because we have chosen directions that are orthogo-
nal to each other.)
Chapter 9
9.1 Equation (5.17) gives Tiu = 3me4/327r2h3SQ. For a photon, K = u/c.
So, with (5.18), Ka0 = 3e2/8ne0hc = (3/2)(l/137). Hence if it is not
the case that Kr <^ 1, then r/ao » 1 and the exponential in the wave
function (5.18) is very small.
9.2 The stationary states ip± are not eigenstates of the electric-dipole-
moment operator /i, but we may calculate the expectation of JJL in
these states. Because the eigenstates </>i and </>2 correspond to different
eigenvalues ±/i 0 , they are orthogonal. Hence, using (9.16),
|(0i>£<£i) + |(02, £<fo) = 0,

and similarly for the -0- -state.
9.3 The answer is 4.6 x 109 K.
9.4 The uncertainty relation (4.13) implies that when the width of the
beam is known to be A, the photons can have a transverse momentum
component of at least h/2A. The total momentum of each photon
is calculated from the de Broglie relation (1.5), and is 2-KTI/X. SO
6 > (h/2A)(27rh/\) = A/4?rA radians.
In the case of the torch, 0 = \wj'/, where w is the width of the filament
and / is the focal length of the reflecting mirror. With / = 10 mm,
w has to be 10~6 m to achieve the same value of 9. It would not be
possible to construct a filament of this size with a reasonable light
output.
9.5 The average number of particles in the given state is
nqrs = Ce-E«rs/kBT,
where C = N(EQtR9se-E**s'k»T)-1 and from (3.5),
Eqrs = (h27r2/2mL2)(q2 + r2 + s2).
(The fact that the states in this problem are degenerate makes no
difference here; the degeneracy enters only if we calculate the number
of particles in a level rather than in a state, as in (9.24).) The average
fraction of particles that are in the state is nqrs/N, which is also the
probability that a given particle is in the state.
The velocity of the particle is v = p/m = hk/m = (hir/mL)(q,r,s).
As L -> oo, the allowed values of EQRS become arbitrarily close, and
r r i 2 i"1
C ~ TV UmL/hn)3 / d3ve"mv lk»T\
= N UmL/h7r)3(27rkBT/m)3/2j .
In this limit, the number of allowed sets of values of (g, r, s) correspond-
ing to the volume element d 3 i; of velocity space ~ {mL/fnr)3d3v. If
d3v is infinitesimal, so that all the corresponding states have essen-
tially the same energy, the average number of particles in these states
is C(mL/h7r)3d3ve~^mv2/kBT. Integrate over vy and vz to obtain the
number having vx in the desired interval.
Notice that the final answer is in fact valid for a large box of any
shape.
9.6 Take the x-direction normal to the membrane, and use the result of the
previous problem. A particle whose rr-component of velocity is u will
reach the wall containing the membrane during the time interval At if
it is within a distance uAt of the wall, and a fraction A/L2 of such
particles will hit the membrane region of the wall. Hence the number
of particles in time At that pass through the membrane is
N f du(m/2nkBT)^2e'mu^2kBT(uAt/L)(A/L2).
JuO
9.7 At T = 0, ni = 0 for E{ > ( and rc* = g{ for E{ < (. This is because

at T = 0 the energy of the system is minimised, but the exclusion
principle does not allow more than one electron per state.
Chapter 10
10.1 The Bloch waves here have Uk(x) in (10.3) equal to a multiple of
e2lrnx/L, where r is any integer, and V + E = h2(k + 2rn/L)2/2m.
The dispersion curves are shown in the figure below. In this model,
there are no gaps between the bands, but these appear when the
model is modified so as to become the nearly-free-electron model.
r=-3
r=-2
10.2 The three components of k satisfy k{ — 2n;7r/F 1 / 3 , where the

Ui are integers. Hence there are V/(2TT)3 allowed vectors per unit
volume of fc-space. Taking account of the two spin states of each
electron, the number of electron states having |fc| < kF is therefore
(|7rA;p)Vy47r3. At T = 0, the NZ electrons occupy all the states up
to kF = (3TT 2 JVZ/TO 1 / 3 -
10.3 (i) For 0 < x < L, the potential is zero and so I/J and Uk have the
forms shown in the second equations of (10.4) and (10.5), respectively.
For — L < x < 0, Uk is obtained from the second equation of (10.5)
by replacing x by (x + L). Impose the conditions that, at x = 0, if) is
continuous and disc xp' = —Uip (0) - see (3.35). This gives
cosh ah — (U/2a) sinhaL = cos kL.
(ii) In 0 < x < L, take rp and uk as before, and in L < x < 2L take
similar forms but with different coefficients. Impose the conditions
on ip and disc rp' at x = L. Because V(x) now has period 2L, so
does uk(x). Use this to obtain uk(x) in -L < x < 0 from uk(x) in
L < x < 2L, and impose the conditions on ip and disc ipf at x = 0.
Eliminate the four coefficients from the four conditions.
10.4
where
and where
n)L
f dx 0*(:r - nL)0(a; - ml),
m,n
n)L
fdx<l>*(x-nL)H<l>(x-mL).
However, from the definition of <^,
H<t>(x — mL) = Eo(f>(x — mL) + V^ Vo(x — rL)(j)(x — m l ) .
The first term in (?/>, -H"^) contributes £Jo to £", and the second con-
tributes
[dx(/)*(x-nL)V0(x-rL)<f)(x-mL)
J2 eik(m-n)L f dx(f)*(x - nL) <f>{x - mL)

m,n J
Because the binding is tight and the atoms are not very close together,
the main contribution to the denominator comes from those terms
having m — n, so that the denominator is approximately equal to TV,
the number of atoms. Likewise, the main contribution to the numerator
comes from those terms having r = n = m ± 1. There are N pairs of
such terms and together they give the desired result.
10.5 In the case of periodic boundary conditions, the allowed values of
k are given by (10.13), and are used to form Bloch waves ipk(x) =
eikxuk{x). In the other case, take the ends of the crystal to be x = 0

and x = NL. Just as for the free particle in a box the wave functions
are sine functions, so here they will be combinations of ipk a n d ^-fc-
Define the normalisation of Uk and u-k such that they are equal to 1
at x = 0; because of their periodicity they are then also equal to 1 at
x — NL. The required wave functions are then ij)k{x) — ?/>_*; (#); this
combination is chosen so as automatically to vanish at x = 0, and it
will also vanish at x = NL if k = nn/NL with n = 1,2,3,.... We
here restrict k to positive values, because the corresponding negative
values would again give the same wave functions, apart from a sign,
and so do not correspond to additional quantum-mechanical states.
In the interval (10.14), the number of allowed values of k is thus the
same as in (10.13).
Chapter 11
11.1 (i) Prom (10.13), there is one allowed wave vector in the interval
2ir/NL of fc-space. Taking into account the two spin states, there are
therefore NL/n states per unit interval of &-space. In the free-electron
model, E = h2k2/2m and dE/dk = h2k/m, and so
p(E) = (NL/7r)/{h2k/m) = (NL/nh){m/2E)1/2.
(ii) This time there is one allowed wave vector in the area 4TT2 /A of
fc-space, so that there are Aj2ix2 states per unit area. For |fc| between
k and (k + dk) there are therefore (yl/27r2)27r/jdfc states having all the
same energy. So p(E) = (Ak/n)/(h2k/m).
(iii) There are V/4TT3 states per unit volume. For |fc| between k and
(k+dk) there are (V/A7r3)A7tk2dk states of the same energy. So p(E) =
(Vk2/7r2)/(h2k/m).
11.2
p(E)/(Am/nJi2)
J I I II
E/(nznz/2ma2)
The stationary-state wave functions are e^klX+k2y>) s'mq(7rx/a), with

E = h2(k2 + k2)/2m + h2q27T2/2ma2. The allowed values of k\ and
k2 are as in part (ii) of the previous problem, while q = 1,2,3, —
Take first q = 1. Then J5 > h2n2/2ma2 and the contribution to p(i?)
is Am/nTi2 as in part (ii) above. For q = 2, there is a similar, ad-
ditional contribution; this begins at E = 4h2n2/2ma2. And so on.
11.3 From part (iii) of problem 11.1, for unit volume of the metal
p(E) = (2m3E/n4h6)1/2. The Fermi-Dirac distribution (11.21) gives
the number of electrons n(E) in the unit volume per unit E. Since
E = p2/2m, the number per unit p is obtained by multiplying this by
dE/dp = p/m and is
{p2 dp/n2h3) ê(p2/2m-O/kBT + ^ -1
The momentum distribution is spherically symmetric in the free-

electron model, so the number of electrons per unit volume of p-space
is obtained by dividing by Anp2. Only electrons having pi/2m > Vo
can escape, and for these p2/2m — ( » &BT, SO that the exponential
in the denominator dominates. The current is therefore
/ dpx / dpy / dpz

J(2mVo)1/2 J-oo J-oo
e e</B 2 2 2
!2± e-p x/2mkBT e-p y/2mkBT e-p z/2mkBT
3 3
m 4 ^
11.4 Near the valence band, p(E) = NS(E + A). The number of electrons
in the valence band is obtained by inserting this into the Fermi-Dirac
distribution (11.21) and integrating around E = — A:
" --N-U.
e-(A+C)/kBT
This results in the given expression for n. But also
AVEdE e-(E-O/kBT
Equating the two expressions for n gives e^/fciaT + e ( A + 2 < ^/ f c B T =

(2iV/^l)(A;Br)- 3 / 2 7r- 1 / 2 . When e - ^ f c B T <C 1, the first term is neg-
ligible and the result follows.
Chapter 12
12.1 The function Uk{x) determined in problem 10.3 may be used to con-
struct the solution given in (12.1) and (12.2). The constants R and T
are determined by making I/J continuous at x = 0, and disc if)' = —U21P
there. The same level structure obtains as in problem 10.3.
Consider now the trapped-electron levels. In 0 < x < L, where
the potential vanishes, we have if)(x) — Ceax + De~ax, where E =
—h2a2/2m. Writing rj> as in (12.4), and making Vi(x) periodic, we find
that in L < x < 2L we have i/>{x) = Cea^x~L^KX a L KX
+De- ^~ ^ .
Imposing the conditions at x — L, that ip is continuous and disc ipf =
—C/iV>, gives two equations. With the even-parity solution, one of the
two corresponding conditions is satisfied automatically. The other,
together with the two previous equations, gives on eliminating C, D
and ekL
(C/1C/2 + \Ul ~ 2a 2 ) t a n h a L = 2aUx.
There is no negative-parity solution.
12.2
Conduction
band
Valence
band
p-type n-type p-type n-type

(a) (b)
In the case (a) where there is no external potential, the numbers of elec-
trons passing across the interface in each direction are equal. Because
the valence band of the p-type material overlaps the conduction band of
the n-type region, some of the electron flow is the result of tunnelling
between the two; the remainder is due to electrons passing directly
from one conduction band to the other, as in the normal junction diode.
When (b) the n-type material is connected to a negative potential, so
that its energy levels are shifted upwards relative to those in the p-type
region, the flow of electrons from p to n is reduced. But the tunnelling
of electrons in the other direction is greatly increased, because of the
large number of empty electron states in the valence band of the n-

region. The current is then greater than for a normal diode, in which
the tunnelling does not occur. When the potential difference is in-
creased to such a value that overlap of the p-type valence band with the
n-type conduction band no longer occurs, the tunnelling ceases and the
situation is the same as for the normal junction diode (see figure 12.5).
Normal
diode *V
Tunnel
diode
Index
acceptor atom 144 of electrons with crystal lattice 113,

alkali 123 127, 132, 137
alpha decay 31 of molecules 99
angular momentum colour of crystal 123, 144
addition 83 commutator 45
algebra 82 completeness 37, 42, 45, 115
orbital 49, 82, 173 Compton effect 4, 7
orbital 52-3 conduction band 122, 144
spin 52, 82 conductivity 121, 139, 141
antiparticle 7 continuity conditions 9, 17, 19, 22, 26,
atom 36,59
acceptor and donor 144 for delta-function potential 29
alkali 61 convolution theorem 171
diameter of 1 crystal
hydrogen 54 boundary conditions 118, 125
many-electron 56 colour 123, 144
atomic number 1 lattice 112, 115, 119
atomic spectra 4 resistance 131
band overlap 117, 119-20, 123, 130, de Broglie's relation 6, 9

131 degeneracy 54
band structure 112, 114 accidental 56
consequences 121 delta function 162
number of levels 118-9 demodulator 154
of npn-junction 152 density of particles 13, 16
of pn junction 146, 149 deuterium 2
beam of particles 16, 26 deuteron 59, 68, 79
black-body radiation 110 diffraction 5
Bloch's theorem 115, 179 diode 146, 150, 155
Bohr atom 55 light-emitting 146
Bohr radius 56 tunnel 157
bound state 12, 21, 28, 55, 59 dipole
box potential 19 electric 96
with electric field 72, 75 magnetic 58, 88-9, 97
transitions 96
chemical potential 135 Dirac equation 88
coherence 107 dispersion relation 119-20, 129-30
collisions divalent metals 113, 120, 123
201
202 Index
divergence theorem 15 Hamilton's equations 86, 176

doping 123, 141, 143-5 Hamiltonian 10, 41
classical 86
effective mass 132 for two-body system 57
Ehrenfest's theorem 40, 46, 129 sudden change 77
eigenfunction 9 harmonic oscillator 23, 46, 158
eigenstate 43-4 heavy water 2
common 45 Hermite polynomial 25, 159
eigenvalue 9, 41, 43 Hermitian operator 41, 44
electric current 130, 144, 146-8, 150, holes 122, 133, 137
152, 154 holography 107
electric field, motion of electron in 128 hydrogen atom 54-5, 160
electromagnetic field 175 in magnetic field 87
electromagnetic interaction 86, 176 molecule 63
electron 1, 53
collision with crystal lattice 113 impurities 132, 141, 144, 152, 157
collisions with crystal lattice 127, insulator 121-3, 134, 136
132, 137 integrated circuit 153
diffraction of 5 inversion layer 154
momentum in crystal 127, 134 isotope 2
motion in electric field 128
spin 83, 86, 88 Jacobi identity 47
trapped 142 junction transistor 152
velocity in crystal 127, 129-30 junction, semiconductor 145
energy gap 112, 117, 122-23
laser
exclusion principle 56, 82, 121, 130,
semiconductor 146, 150
134
Legendre's equation 51
expectation value 16, 44
light-emitting diode 146
Lorentz force 87
Fermi energy 111, 135
in diode 147-48, 154 magnetic quantum number 52
in transistor 153 maser 100
Fermi-Dirac distribution 134, 137, Maxwell-Boltzmann distribution 104,
140, 144 135
Fermi-Dirac distribution 111 measurements 38, 42-3, 45-6, 77
field-effect transistor 153 metals 120, 123-4, 134, 136
fine-structure constant 60 molecule 67
flip-flop 157 ammonia 98
flux of particles 16, 27 hydrogen 63
forbidden transitions 75 momentum operator 9, 46
Fourier transform 38 MOSFET 153
free-electron model 124, 139
neutrino 79
gauge condition 87, 176 neutron 1, 2, 59, 79, 83-4
golden rule 96 normalisation condition 13
good quantum number 49 npn transistor 152
Index 203
nuclear force 1 spherically-syymetric 53

nucleus 1 power-series solutions for differential
equations 158
observables: basic assumptions 10, 41 principal quantum number 55
operator probability
energy 10 of result of measurement 43
Hermitian 41 of transition 75
Hermitian conjugate 48 probability density 13
linear 36 in momentum space 39
momentum 9 in plane wave 13
orbital-angular-momentum 173 in stationary state 16
translation 178 proton 1, 53, 83, 89
optical properties 121
optical pumping 105 quantum mechanics
op tics,non-linear 107 basic assumptions 10, 13, 36, 41-2
orbital angular momentum 49, 52-3, classical limit 2
82, 173 statistical nature 6
^-component 50, 52 quantum number
orthogonality 42 good 49
orthonormal set 42 magnetic 52
orbital-angular-momentum 52
pair excitation 137, 146
parity 22, 25, 35 radiation pressure 4
Parseval's theorem 172 reduced mass 57
particle density 13 resistance of crystal 131
Pauli resonance imaging 91
exclusion principle 56, 82, 121, 130, resonant absorption 91, 95
134 ruby 105, 144
matrix 84
perturbation theory Schrodinger equation
time-dependent 73 time-dependent 11
time-independent 70 time-independent 10
phonon 106, 132, 137 selection rules 97, 106
photoelectric cell 137 semiconductor 122-3, 133-4, 136, 137,
photoelectric effect 3 140-1
photon 3, 6, 24, 52, 83, 93, 95, 99, 102, doped 123, 141
105 intrinsic 122, 136-7, 147
photovoltaic effect 146 n-type and p-type 144
plane wave 11, 37 semiconductor junction 145
pnp transistor 152 semimetals 120-1, 123, 134
population inversion 105-6 special relativity 4
positron 7 spectrum 88
potential spherical polar coordinates 50
barrier 26, 28, 30, 77 spin 52, 56, 82, 98
delta-function 60 z-component 56, 84
periodic 114 Pauli matrix 84
spherical well 59, 61 precession 89, 91
204 Index
spin-orbit interaction 58-9 tritium, decay of 79

spinor 85 tunnel diode 157
spontaneous emission 93, 106 tunnel effect 26, 28, 30, 32, 63, 65, 77,
square barrier 26, 30 99, 113
square well 21, 31, 34
infinite 12, 17, 19 uncertainty principle 25, 39, 45
stationary state 11, 49 for energy 76
statistical mechanics 104, 111, 134 generalised 45, 48
stimulated emission 93, 95, 100, 102 unstable system 77
superposition 20
of stationary states 36 velocity of electron in crystal 127,129,
130
thermionic emission 113, 140
wave function
tight-binding model 124
transistor Bloch 116,118,120,124,126,128-9,
field effect 153 179
npn 152 wave packet 37, 38, 40, 129
pnp 152 wave vector 8
transition WKB formula 30
dipole 96 work function 3
radiative 88, 93 Zeeman effect 87
transition probability 75 zero-point energy 25, 132
trapped electron 142
triode 152

Essential Quantum Physics

Uploaded by

Copyright:

Available Formats

Essential Quantum Physics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Essential Quantum Physics

Uploaded by

Copyright:

Available Formats

This book provides a first course on quantum mechanics and describes

simple applications to physical phenomena that are of immediate and

The first five chapters introduce the fundamentals of quantum mech-

Ideal either as a course text or a self-study text, the book contains

Cambridge University Press

www. Cambridge. org

© P. V. Landshoff, A. Metherell and W. G. Rees 1997

This publication is in copyright. Subject to statutory exception

First published 1997

A catalogue recordfor this publication is available from the British Library

Library of Congress Cataloguing in Publication data

ISBN 978-0-521-62011-6 Hardback

Transferred to digital printing 2010

Cambridge University Press has no responsibility for the persistence or

Constants of quantum physics x

2 The Schrodinger equation 8

4 The superposition principle 36

5 The hydrogen atom 49

6 The hydrogen molecule 63

7 Introduction to perturbation theory 70

9 Masers and lasers 93

10 Band structure of crystals 112

11 Electron motion in crystals 126

Hints for the problems 180

This book is intended as a first course on quantum mechanics and its

Christ's College, Peter Landshoff

Dirac's constant h = h/2iT = 1.05 x 10" 3 4 J s

Here h is Planck's constant,

and W is the energy that must be given to the electron to enable it to

so that for a photon

The energy levels of an atom (or molecule) depend on what element it

Wave nature of matter

electrons according to de Broglie's relation

The waves may be assigned a characteristic frequency v, which may be

The basic equation of quantum mechanics is known as the Schrodinger

Non-relativistic kinematics can be used so long as the energy E is not

Wave functions and operators

We also work with

(pronounced '/i-cross') instead of h. Then we have

tt(r,<)= tfei(*'r-wt). (2.3)

Here r denotes the position vector and TV is a constant. In many physical

or, in more concise vector notation,

Equation (2.4a) says that if we apply the differential operator —ihd/dxj

(a) States of a system are described by wave functions # .

We shall elaborate on these assumptions in chapter 4.

H = (-h2/2m)V2 + V(r) (2.6)

corresponding to the classical expression p2/2m + V(r). H is called the

(subject to appropriate boundary conditions). This equation

[(-h2/2m) V2 + 7(r)]tt(r, t) = £ * ( r , t) (2.7b)

is the time-independent Schrodinger equation.

is the generalisation of the Schrodinger equation that describes this, we

The obvious generalisation of this to the case where V ^ 0 is

[(-h2/2m) V2 + V]*(r, t) = \h 8V/dt (2.8a)

This means that the time dependence of \P(r,4) is then trival:

y(r,t) = $(r)e-iEt/n. (2.9)

The function ^ ( r ) is called the time-independent wave function and it

Hip = E\j). (2.7c)

where, from (2.7),

Example: the one-dimensional potential well

As only one of the three coordinates is explicitly involved, we call this a

qrs(r,t) = ^grs(r)e-iE<>rt/h. (3.6)