the physics of low dimensional
TRANSCRIPT
-
7/24/2019 the Physics of Low dimensional
1/457
-
7/24/2019 the Physics of Low dimensional
2/457
THE PHYSICS OF
LOW-DIMENSIONAL
SEMICONDUCTORS
-
7/24/2019 the Physics of Low dimensional
3/457
TF-IE PI-IYSICS OF
LOW-DIMENSIONAL
SEMICONDUCTORS
AN INTRODUCTION
JOHN H. DAVIES
Glasgow University
CAMBRIDGE
UNIVERSITY PRESS
-
7/24/2019 the Physics of Low dimensional
4/457
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid , Cape Tow n, Singapore, So Paulo
Cam bridge University Press
40 West
20th
Street, New York,
N Y
10011-4211,
USA
www.cambridge.org
Information on
this title:
www.cambridge.org/9780521481489
0Cam bridge University Press 1998
This book 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 perm ission of Cambridge University Press.
First published
1998
Reprinted 1999, 2000, 2004, 2005
Printed in the United States of Am erica
A catalog record for this
publication
is
available from
the British Library.
Library
of
Congress
Cataloging in
Publication
Data
Davies, J. H.
(John H.)
The
physics
of
low-dimensional semiconductors
:
an introduction
/
John H. Dav ies.
p. cm.
Includes bibliographical
references
and index.
ISBN 0-521-48148-1 (hc) . ISBN
0-521-48491-X (pbk.)
1. Low-dimen sional semiconductors. I . T i tle .
QC611.8.L68039 1997
537.6'221 dc21
7-88
ISBN-13
978-0-521-48148-9
hardback
ISBN-10 0-521-48148-1 hardback
ISBN-13
978-0-521-48491-6 paperback
ISBN-10 0-521-48491-X
paperback
Cam bridge University Press has no
responsibility
for
the persistence
or accuracy
of URLs
for
external or
third-party Internet W eb sites referred to
in this book
and does not guarantee that any
content on
such
Web
sites is, or
will remain, accurate
or appropriate.
-
7/24/2019 the Physics of Low dimensional
5/457
-
7/24/2019 the Physics of Low dimensional
6/457
To
Christine
-
7/24/2019 the Physics of Low dimensional
7/457
-
7/24/2019 the Physics of Low dimensional
8/457
CONTENTS
Preface
age
xiii
Introduction
v
1 OUNDATIONS
1.1
Wave Mechanics
and the Schr6dinger
Equation
1.2
Free
Particles
1.3
Bo und Par t ic les: Quantum
Wel l
1.4
Charge and
Current Densities
1.5
Operators and
Measurement
3
1.6 Mathematical Properties
of Eigenstates
0
1.7
Counting
States
2
1.8
Filling
States: The Occupation Function
0
Further Reading
0
Exercises
1
2
LECTRONS AND
PHONONS
IN
CRYSTALS
5
2.1
Band Structure in One Dimension
5
2.2
Motion of
Electrons in Bands
0
2.3 Density of States
4
2.4
Band Structure in
T w o and Three
Dimensions
5
2.5
Crystal Structure of the
Common Semiconductors
7
2.6 Band Structure of the
Common Semiconductors
1
2.7 Optical Measurement
of Band
Gaps
9
2.8
Phonons 0
Further Reading
6
Exercises
6
vil
-
7/24/2019 the Physics of Low dimensional
9/457
viii
ONTENTS
3
ETEROSTRUCTURES
0
3.1 General Properties of Heterostructures 0
3.2
Growth
of Heterostructures 2
3.3
Band Engineering
5
3.4
Layered St ructures:
Quantum Wells and
Barriers
8
3.5 Doped Heterostructures
2
3.6
Strained Layers
6
3.7 Si l icon Germanium Heterostructures 00
3.8 W i r e s and Dots 02
3.9
Optical Confinement
05
3.10 Effective-Mass Approximation
07
3 .11
Effect ive-Mass Theory in Heterostructures 11
Further Reading
14
Exercises
14
4
UANTUM
WELLS AND LOW-DIMENSIONAL SYSTEMS
18
4.1
Infinitely
Deep Square
W e l l
18
4.2
Square
Wel l
of
Finite Depth
19
4 .3 Parabol ic W el l
25
4.4
Triangular Well 28
4 .5 Low-Dimensional Systems
30
4 .6
Occupation of Subbands 33
4.7
Two- and
Three-Dimensional Potential Wel l s
35
4.8
Further Confinement
B e y o n d T wo
Dimensions
40
4.9 Quantum Wells in Heterostructures
42
Further Reading
46
Exercises 46
5
UNNELLING TRANSPORT
50
5 .1 Potential Step 50
5 .2 T-Matrices 53
5.3
More on T-Matrices
58
5 .4
Current and Conductance
62
5 .5 Resonant Tunnelling
67
-
7/24/2019 the Physics of Low dimensional
10/457
CONTENTS
x
5 .6
Superlattices
and
Minibands
77
5 .7 Coherent
Transport
with Many Channels
83
5 .8
Tunnelling
in
Heterostructures
95
5.9 What Has Been Brushed Under
the Carpet?
99
Further Reading
00
Exercises 01
ELECTRIC
AND
MAGNETIC
FIELDS
06
6.1 The Schrtidinger Equation with Electric and
Magnetic Fields
06
6 .2
Uniform Electric
Field 08
6 .3 Conductivity and Resistivity Tensors
16
6 .4 Uniform Magnetic
Field
19
6 .5 Magnetic
Field in a
Nar ro w
Channel
33
6 .6
The Quantum H all Effect
38
Further
Reading
45
Exercises
46
7
PPROXIMATE METHODS
49
7.1
The
Matrix
Formulation of Qu antum
Mechanics 49
7 .2
Time-Independent
Perturbation Theory 52
7.3 k p
Theory
61
7 .4 W K B
Theory
63
7 .5
Variational Method
70
7 .6
Degenerate
Perturbation
Theory
73
7 .7
Band S tructure:
Tight Binding
75
7 .8 Band S tructure:
Nearly Free Electrons 80
Further
Reading
84
Exercises
84
8 CATTERING
RATES: THE GOLDEN RULE
90
8.1
Golden Rule for
Static Potentials
90
8 .2
Impurity
Scat tering
95
83
Golden Rule for Oscillating Potentials
01
8 .4
Phonon Scat tering
02
-
7/24/2019 the Physics of Low dimensional
11/457
CONTENTS
8.5
Optical Absorption
08
8.6 Interband Absorption 13
8.7
Absorption in a Quantum
Well
16
8.8 Diagrams and the
Self-Energy 21
Further Reading
24
Exercises
24
9
HE
TWO-DIMENSIONAL ELECTRON GAS
29
9.1
Band Diagram of M odulation-Doped Layers 29
9.2 Beyond the
Simplest Model
36
9.3
Electronic Structure of a 2DEG 42
9.4
Screening by
an
Electron Gas
49
9.5
Scattering by Remote Impurities
56
9.6
Other Scattering Mechanisms
62
Further Reading
65
Exercises
66
10
OPTICAL PROPERTIES OF
QUANTUM
WELLS
71
10.1
General Theory
71
10.2 Valence-Band Structure: The Kane M odel
77
10.3 Bands in a Quantum Well
84
10.4 Interband Transitions in a Quantum
Well
87
10.5 Intersubband
Transitions in a Quantum Well
93
10.6 Optical Gain and Lasers
95
10.7
Excitons
97
Further Reading
06
Exercises
06
A TABLE
OF PHYSICAL C ONSTANTS
09
A2
PROPERTIES
OF IMPORTANT SEMICONDUCTORS 10
A3
PROPERTIES
OF
GaAsAlAs
ALLOYS
AT
ROOM TEMPERATURE
12
A4
HERMITE S
EQUATION: HARMONIC OSCILLATOR
13
-
7/24/2019 the Physics of Low dimensional
12/457
CONTENTS
i
AS
AIRY
FUNCTIONS: TRIANGULAR WE LL
15
AB
KRAMERS KRONIG RELATIONS AND
RESPONSE FUNCTIONS
17
A6.1
Derivat ion
of the Kram ersKronig Rela t ions 17
A6.2
Model Response Functions
19
Bibliography
23
Index
27
-
7/24/2019 the Physics of Low dimensional
13/457
-
7/24/2019 the Physics of Low dimensional
14/457
PREFACE
I joined the Department of Electronics and
Electrical
Engineering at
Glasgow Uni-
versity
som e ten years ago. M y research was performed in a group w orking
on
advanced sem iconducting devices for both electronic
and optical applications.
It
soon became
apparent that advances in physics
and
technology had left
a gap be-
hind them in the education
of
postgraduate students. These students
came
from a
wide range of backgrounds, both in physics and engineering; some had received
extensive instruction in quantum me chanics
and
solid state physics, w hereas others
had only
the
smattering of semiconductor physics needed to explain the operation
of
classical
transistors. Their projects were equally
diverse,
ranging from
quantum
dots and electro-optic mo dulators to Bloch oscillators and ultrafast
field-effect
tran-
sistors. Some excellent reviews were available, but
most started at
a level beyond
many
of the
students. The
same was true of the
proceedings
of
several summ er
schools. I therefore initiated a lecture course
with
John
Barker on
nanoelectronics
that instantly attracted
an
enthusiastic audience. The course
was given
for several
years
and evolved into this book.
It was difficult to keep
the length of the lecture course manageable, and a book
faces the same problem. The applications of heterostructures and
low-dimensional
semiconductors continue to grow steadily, in
both physics and engineering.
Should
one display the myriad ways in which the properties
of
heterostructures
can be
harnessed,
or
concentrate
on
their physical foundations? T here seeme d to be a
broad gap in the literature, between
a
textbook
on quantum
mechanics
and solid
state physics illustrated with semiconductors,
and an analysis of the devices that can
be made.
I have aimed towards the textbook,
a fortunate decision as there are now
some excellent books describing the applications. The experience of
teaching at a
coup le of sum mer schools also convinced me that a more
introductory treatment
would b e useful,
one
that concentrated
on the basic
physics.
This book addresses
that need.
Acknowledgements
Several colleagues contributed to
the course out of
which this
book developed.
John
Barker, Andrew
Long, and Clivia Sotomayor-Torres
shared the
lecturing at various
times
and
helped to shape
the syllabus. Several students and
postdoctoral research
-
7/24/2019 the Physics of Low dimensional
15/457
xlv
REFACE
assistants encouraged
m e
to continue the course and learn some topics that were
new to me. I would particularly like to thank A ndrew Jen nings, Michael and Frances
Laughton, Alistair Meney, and John Nixon. It is also
a
pleasure to thank An drew
Long and
my w ife
for
their helpful comments
on the
manuscript.
M any colleagues
have
kindly provided data that I have been allowed to replot
in
a
convenient w ay to illustrate
the text. I
am
very grateful
for their help, particularly
to those who generously supplied unpublished measurem ents and
calculations,
and
to Mike Burt ,
who also gave
advice on effective-mass theory.
It has taken a long time to com plete this
book.
I don 't imagine that I am the first
author who has sadly und erestimated
the effort required to turn
a pile
of lecture
notes
into
a
coherent man uscript. Most of the work has been done in evenings, between
reading bedtime stories to my daug hters and feeling
exhaustion setting in.
A s most
parents
with youn g children w ill appreciate, this interval is
short and
frequently non-
existent. I
am
very grateful to my family
for
their forbearance and encouragement.
I would also like to thank the publishers
for
their tolerance,
as they might well have
despaired of ever receiving
a finished manu script.
The final
proofreading was carried
out at the Center for Quantized Electronic Structures (QUEST) in the University of
California at
Santa Barbara. It is a
pleasure to ack nowledg e their hospitality
as
well
as the financial support of QUEST and the Leverhulme Trust during this period.
I wo uld like to finish with a quotation from the preface by E R eif to his book,
Fundamentals of statistical
and
thermal physics. It must reflect many authors' feel-
ings
as
their
books
approach
publication.
It has been said that 'an author nev er finishes a book, he merely
abandons
it'. I have
com e to appreciate vividly
the truth of this
statement
and
dread to see
the
day w hen, looking at
the manuscript
in print, I am sure to realize that so many things could have
been done
better
and explained
more clearly. If I
abandon the book nevertheless,
it is in the m odest hope that it may be useful to others despite its
shortcomings.
John Davies
Milngavie, September 1996
-
7/24/2019 the Physics of Low dimensional
16/457
INTRODUCTION
Low-dimensional systems have revolutionized sem iconductor physics. They rely
on
the technology of heterostructures, where the composition of a semiconductor can
be changed
on the
scale
of a nanometre. For
example,
a sandwich of
GaAs between
two layers
of A l , Ga
,As
acts like an
elementary quantum well. The
energy levels
are
widely separated if the well is narrow, and all electrons may be trapped
in the
lowest level. Motion
parallel to
the
layers is not affected, however, so
the
electrons
remain free in
those directions. The result is
a two-dimensional electron
gas, and
holes can be trapped in the same way.
Optical measurements provide
direct
evidence for the
low-dimensional behaviour
of electrons and holes
in a quantum well. The
density of
states
changes from a
smooth parabola
in three
dimensions to
a
staircase in a
two-dimensional system.
This
is seen clearly
in
optical
absorption, and the
step at
the
bottom
of the
density
of
states enhances
the
optical properties. This is put to practical
use in
quantum-well
lasers, whose threshold current is lower than that
of a three-dimen sional device.
Further
assistance
from technology is needed to harness low-dimensional systems
for transport.
Electrons and
holes
must
be introduced by doping, but the carriers
leave charged impurities behind, which limit their mean
free path. The solution to
this problem is
modulation doping, where carriers are
removed
in space from
the
impurities that
have provided them. This
has raised
the mean
free
path of
electrons
in a
two-dimensional electron gas to around 0.1 mm
at low tem perature. It is now
possible to fabricate
structures
inside which electrons are
coherent
and must
be
treated as
w aves rather than particles. Observations of
interference attest to
the
success
of this approach. Again, there
are
practical
applications such as field-effect
transistors in
direct-broadcast
satellite receivers.
A s these examples
show,
complicated technology underpins experiments
on low-
dimensional systems. In con trast ,
it turns out
that most
of the
physics can be under-
stood with relatively straightforward
concepts. The
aim
of
this
book
is to explain
the
physics that underlies
the
behaviour of
most low-dimensional systems
in
semicon-
ductors, considering both
transport and
optical properties.
The
me thods described,
such
as perturbation theory,
are standard but have
immediate application the
quantum-confined
Stark effect , for
example, is both a
straightforward illustration of
perturbation theory and the basis
of a
practical
electro-optic modulator.
The most
XV
-
7/24/2019 the Physics of Low dimensional
17/457
xvi
NTRODUCTION
advanced technique used is Fermi's golden rule, which marks a traditional dividing
line between 'elementary' and 'advanced' quantum mechanics.
The
disadvantage
of
this approach is that it is impossible to describe
more than a
tiny
fraction of the applications of the basic
theory. M any topics
of
current
research,
such as the chaotic behaviour of
electrons in microcavities or the
optical properties
of self-organized quantum dots, have to be om itted. Fortunately there are several
surveys of the applications of
low-dimensional semiconductors,
and
also more ad-
vanced theoretical descriptions. This book provides the foundations on
which they
are built.
Outline
Chapters 1 and 2 provide
the
foundations
of quantum mechanics and solid state
physics
on
which
the
rest
of the book
builds.
The
presentation is intended only
as a
refresher course, and there
are
some
suggestions for textbooks
if
much of
the
ma terial is unfamiliar.
A survey
of heterostructures is given in Chapter 3, and
Chapter
4
covers the basic theory of low-dimensional systems. This entails the
solution of simple
quantum-well problems, and an appreciation of how trapping
in
such
a well makes a three-dimensional electron behave as though it is only tw o-
dimen sional. Chapter 5
is devoted to tunnelling, with applications
such
as resonant-
tunnelling
diodes.
Electric and magnetic fields provide important probes of
many
systems
and are
discussed
in
Chapter
6. Perhaps
the m ost drama tic result is
the
quantum Hall
effect
in a
magnetic
field,
which is found to be of value as a standard
of resistance w hile continuing to tax our und erstanding. Chapter
7
contains a range
of approximate m ethods used to treat systems
in a
steady state. These have wide
application,
notably to band structure.
An other example is
th e W K B
method, w hich
can be used to find
the
energies of allowed states in a quantum well or to estimate
the rate of
tunnelling through
a
barrier. Fermi's golden
rule is derived
in
Chapter
8
and used to calculate
the
scattering
of electrons by im purities
and phonons. Optical
absorption is another
major application. The final
chapters are
devoted to
the
two
principal
low-dimensional systems.
The
two-dim ensional electron gas
in
Chapter
9
is used prim arily for its transport properties, whereas the optical properties
of a
quantum well
in
Chapter 10 find employment
in
devices such as semiconductor
lasers.
The book is pitched at
the
level
of beginning postgraduate and advanced under-
graduate students. All
the basic techniques in this
book
were
in
my undergraduate
physics
courses,
although few
of the applications
had been invented It is assumed
that
the
reader has had a
glimpse
of quantum mechanics and
solid state physics,
but
only at
the
level covered
in
many courses
quaintly nam ed 'mode rn physics'.
The
first two chapters will be rather hard going
if
this is not
the case . My
experience has
been that students with a degree in
physics find the basic theory familiar, but the
applications
improve their understanding imm ensely. The background of students
from electrical
engineering varies
widely,
but the lecture course evolved to address
-
7/24/2019 the Physics of Low dimensional
18/457
INTRODUCTION
vii
this,
and the
level
of the book
should be suitable
for
them too.
The theoretical level
has deliberately been k ept low and should not provide any impediment.
Exercises
Each chapter has around twenty exercises. Their difficulty
varies
considerably; some
are trivial, whereas others require numerical
solution.
I did almost all
the
calculations
for
this
book with an obsolete spreadsheet (Trapeze, dating from 1988), so the
numerical aspects are n ot serious. How ever, a symbolic-manipulation program such
as
M aple, Mathematica,
or the l ike w ould m ake these tasks easier. Occasionally
some integrals
or special functions appear
and references for
these are given
in the
bibliography.
Units
I
have used SI units throughou t this
book,
with the exception of the
electron volt
(eV), which is far too conven ient to
abandon. The main problems
for
users
of
CGS
units
are in
equations concerning electric
and
magnetic
fields. Removing
a factor
of 47r
c
o from equations
for
electrostatics
in SI
units should give the
corresponding
result in CGS units.
The
magnetic
field or flux
density B
is measured
in tesla (T) in
SI
units,
and formulas should be divided by
the
velocity of light c wherever B or the
vector potential A appears to give their form in
CGS units. Finally, I
use nanometres
rather than Angstrom units for lengths; 1 nm 10
A .
Notation for
vectors
It is often necessary to distinguish betwe en two- and
three-dime nsional vectors in
low-dimensional systems, particularly
for position and
wave vectors. I have tried
to follow a consistent notation
throughout this
book, the penalty being that some
familiar
formulas look
slightly odd.
Most
low-dimensional
structures are
grown
in
layers
and the z-axis is taken
as
the direction of growth,
normal to
the
layers. Vectors in the xy-plane, parallel to
the layers, are
denoted with lower-case letters. Thus
the position in the plane
is
r =
(x,
y).
Upp er-case letters
are
used
for the
corresponding three-dimensional
vector,
so
for position
R = (r, z =
(x, y, z).
Similarly, wave vectors
are
written
as
K = (k, k
z ) = (kx , k b ., k z
).
The
only o ther quantity that needs to be d istingu ished
in this way is
Q, which is used
for the
wave vector
in
scattering.
This notation
requires some familiar results to be w ritten w ith upper-case let-
ters
for
consistency.
The
energy of free electrons in three
dimensions
is E 0
(K)
h2K2/2ino,
for exam ple. I hope that
the
clarity offered b y
consistent usage
offsets
this sm all disadvantage.
References
and
further reading
It is not appropriate
in a
textbook to give detailed references to
original
papers. In-
stead there is a bibliography of more
advanced
books and articles from review jour-
nals.
In one or
two
cases
I
have referred to
original
papers when I w as unable to find
-
7/24/2019 the Physics of Low dimensional
19/457
xviii
NTRODUCTION
an appropriate review. There are also several summ er schools
on
low-dimensional
semiconductors that publish their proceedings; naturally I
have
provided a reference
to one that I edited.
There
are
several
books
that develop
the
material
in
this
book
further. A mon g
them, Bastard (1988) gives a lucid account of the electronic structure of heterostruc-
tures, including
a
thorough description of the Kane model , with their electronic
and
optical properties. W eisbuch and
Vinter's book (1991) is
an
enlargement of
an earlier review
in
Willardson and
Beer (19664 They describe
the applications
of
heterostructures
as
well
as
their physics, with a particularly good
section on
quantum-well lasers. Their list of references exceeds 600 entries, which gives some
idea of the activity in
this
field. Finally, K elly's book
(1995)
is
notable for the
breadth
of its coverage.
H e
describes the technology of fabrication and an
enormous range
of applications of
heterostructures
in
physics
and engineering.
Just
a
glance at
the
topics covered
in
this survey leaves
one with
no
doub t that this
field
will advance
vigorously into the
next century.
-
7/24/2019 the Physics of Low dimensional
20/457
FOUNDATIONS
This book is about low-dimensional semiconductors, structures in
which electrons
behave as though they
are free
to move in only two
or
fewer free dimensions. Most
of
these
structures are
really
heterostructures,
meaning that they
comprise more
than one
kind
of material. Before we c an investigate the
properties
of a heterostruc-
ture,
we need to understand
the
behaviour of electrons in a uniform semiconduc tor.
This in turn rests on the foundations of quantum mechanics, statistical mechan-
ics, and the band theory of crystalline solids. The
first two chapters
of this
book
provide a review of these foundations. U nfortunately i t is impossible
to provide
a full tutorial within the space available, so the reader should consu lt
one of the
books suggested at the end of the appropriate chapter
if
much
of the
material is
unfamiliar.
This
first chapter covers
quantum
mechanics
and
statistical physics. Some topics,
such
as the
theory
of angular mom entum, are
not included althoug h they
are vital
to
a thorough course on quantum mechanics.
The
historical background, treated
at length
in
most textbooks
on quantum
m echanics, is also om itted. There is little
attempt to justify quantum
mechanics, although the rest
of the book could be said to
provide
support because we
are able to explain numerous experimental observations
using the basic theory developed in this chapter.
1.1
Wave Mechanics and
the
Schrbdinger
Equation
Consider the motion of a single particle, such
as an
electron, moving in one dimen-
sion
for
simplicity. Elementary classical mechanics is based on the concept of a
point
particle, whose position x and momentum p (or velocity
v =
plm) appear in
the
equations of motion. These quan tities
are
given directly by New ton's laws,
or
can be calculated from the Lagrangian or Hamiltonian functions in more
advanced
formulations.
W ave mechanics is
an elementary
formulation of quantum theory that,
as
its
name implies, is centred on a wave function
t u
i (x, t). Quantities such
as position
and
momentum
are
not given directly,
but must be deduced from W . Instead of
-
7/24/2019 the Physics of Low dimensional
21/457
1 . FOUNDATIONS
New ton's laws we
have a
wave equation that governs the evolution
of
tli(x, t). In
one dimension this takes the form
h 2
a
2
p(x,
t)
(x)tli (x , t) = iht P (x, t),
1.1)
2m
ax-
t
which is
the time-dependent Schrddinger equation. W e shall see
a partial
justifica-
tion of its form a little later. This equation describes a particle m oving in a region
of
varying po tential energy r(x);
forces do
not
enter directly.
The
potential energy
could
arise from an
electric
field expressed as a scalar potential, but the inclusion
of a
magnetic field is more
complicated and will be deferred to Chapter 6.
A
useful simplification is obtained by looking
for separable
solutions
where
the
dependence on
x
and t is decoupled, ) = ilf(x)T (t).
I shall
use capital
for the
t ime-dependent w ave function
and the
lower-case letter *'
for the
time-
independent function. Substituting the product into the time-dependent Schrtidinger
equation (1.1) and dividing by
T
gives
1 dT (t)
2
d
2
*(x)
T
ih
dt 2m
dx2
V(x)*(x)1
.
1.2)
Now
the left-hand side is a function of
t
only, whereas the right-hand side is a
function of
x
only, so
the
separation has succeeded.
This
makes sense only
if
both
sides
are equal to
a constant, E,
say.
Then
the left-hand side gives
T (t)
c x
exp
Et
h
xp(icot),
1.3)
where E = hco. This
is
s imple harmonic
variation in time. W e have no
choice
about the
com plex exponential function: i t cannot be replaced by
a real sine or
cosine, nor is exp(+icot)
acceptable. The
form of the time-dependent Schrtidinger
equation requires
exp(icot)
and this convent ion
is follow ed throughout quantum
mec hanics. It contrasts with other areas
of physics, where exp(icot) may be used
for the
dependence
of oscillations on time, or engineering,
where exp(+
jcot)
is
usual. Un fortunately this choice of sign has a far-reaching
influence and crops up in
unexpected places,
such
as the
sign of the imaginary
part of the complex dielectric
function E
r (W )
(Section
8.5.1).
The spatial part of the
separated Schrticlinger equation becom es
V(x)i/(x) =
2m dx 2
This
is
the
time-independent Schr6dinger equation.
The equation takes the same
form
in
three
dimensions
with
a-/ax-
replaced by V2 a2/ax2+a2/ay2+a2/az2.
Thus the solutions of the
time-dependent Schr6dinger equation take the form
iEt
(x, t)
= (x)
exp
1.5)
h2 d
2 lif
(1.4)
-
7/24/2019 the Physics of Low dimensional
22/457
1.2
FREE
PARTICLES
Later we shall justify the identification of the
separation
constant
E as the
energy
of the particle. Thus
the solutions of the
time-independent
Schriidinger
equation
describe states of the particle with
a definite energy, known
as
stationary states.
Again, this term will be justified
later, but first we shall look at some simple but
important solutions of the
Schriidinger
equation.
1.2FreeParticles
The simplest example is a particle
(an
electron,
say) in free space, so
V(x) = 0
everywhere.
The
time-independent
Schrtidinger
equation is
h
2 d2
0
.
E*(x).
1.6)
2m dx
2
This is the standard (Helmholtz) wave equation, and is s imple
enough that we
can guess the possible solutions. One choice is to
use
complex exponential waves,
(x) = exp(-Fikx) or
exp(ikx).
Alternatively we could use real trigonometric
functions and
write
1r
(x)
= sin
kx
or cos
kx. It turns
out that
the
choice
of real
or complex functions has important
consequences.
Substitution shows
that any of
these functions is
a solution
with
h
2
k
2
E=
2m
(1.7)
Classically
the kinetic energy can be written as E = p
2
/2m so we deduce that the
momentum
p = hk.
Com bining this with
the relation between energy
and
frequency
yields
the two
central relations of old
quantum theory:
E = hco = hv Einstein),
1.8)
h
p = hk =
3
-
de Brogl ie) .
1.9)
D ividing the energy by
h gives the dispersion relation
between frequency
and wave
num ber to be co
= (h12m)k
2
.
This
is nonlinear, which means that the
velocity of
particle waves is
a
function of
their frequency
and must be defined carefully.
The
two
standard definitions are as
fol lows:
(1.10)
phase velocity)
h
- - =
Th
k 2m 2m
dco hk
(group velocity)
g =
dk
h
m
where v e l is the classical velocity. The w ave packet sketched
in Figure 1.1
shows the
s ignif icance of these two v elocit ies. The
w avelets ins ide the packet m ove along at
the
-
7/24/2019 the Physics of Low dimensional
23/457
1.
FOUNDATIONS
FIGURE 1.1. A
wave packet, showing
the envelope that moves at v g
while
the wavelets inside
move at v p
h.
phase
velocity v p
h
while
the
envelope moves at
the
group velocity
v g
.
If
this packet
represents
a
particle such as an electron, we are usually interested in the behaviour
of the w ave function as a
whole rather than its internal mo tion. The group velocity
is then
the
appropriate
one and
it is
a relief
that this agrees with
the
classical result.
Even if
we
use a wave pack et to represent
an electron, it is still spread out
over
space rather than localized at a point as in classical mechanics. This is inevitable in
a picture based
on waves, and
mean s that we cannot give
the location of a
particle
precisely. W e'll look at this further in Section
1.5.3.
Finally, we found that
E
2k2/2m5
w hich m eans that E >
0
if k is a
real number. The wave num ber becomes imaginary, k
K , if
E 0.
This
contrasts with classical mec hanics,
where the
state of
lowest energy has
the
particle sitting still, anywhere
on the
bottom
of the well, with
no
kinetic energy and
E = 0.
Such behaviour would violate
the
uncertainty principle
in quantum me chanics, described
in Section 1.5,
so even
the
lowest state has
posit ive
zero-point
energy.
The wave functions
in the square
well
have an important
symmetry property.
Those w ith n
odd are
even functions of x
about the cen tre of the well ,
whereas those
with n even are odd functions
of x
(i t would be neater if the numbering of
states
started with
0
rather than
1). This symm etry holds for
any potential well where
V(x)
is
an
even function
of
x.
Sym m etry properties like this
are important in
providing
selection rules
for
many processes, such
as optical
absorption. This
is a very simple
examp le; group theory is needed to describe
more com plicated sym metries, such
as
those of a crystal.
-
7/24/2019 the Physics of Low dimensional
25/457
(a)
Al GaAs GaAs
A lG a A s
barrier
ell arrier
6
.
FOUNDATIONS
Now that we have
calculated
the
energy levels
in a quantum w ell, an obvious
question is how do we measure them experimentally. Optical methods provide
the
most direct techniques, so w e shall next take a look
at optical
absorption in a quantum
well.
1.3.1
OPTICAL ABSO RPTION IN
A
QUANTUM WELL
The quantum
well looks like
an artificial model, which is at
home in a textbook but
has little
application in the
real world. Although
an
infinitely deep
well cannot be
made,
it is simple nowadays to grow
structures that
are close to ideal finite wells.
The energy levels in a finite well will be calculated in Section 4.2, but in practice
the
infinitely
deep
well is often used
as an approximation
because its results
are
so
simple.
A heterostructure consisting of a
thin
sandwich of GaAs between thick layers
of
AlGaAs provides a simple quantum w ell,
shown
in Figure 1.3(a). (`AlGaAs'
really means an alloy such as A10.3Ga0.7As, but the
abbreviation is
universal.) To
justify this, we need to anticipate some concepts
that will be explained more fully
later
on.
First,
look at
the
behaviour of
electrons. Free
electrons have
energy
so(k)
=
h2k2/2m0.
Electrons
in a
semicond uctor l ive
in the
conduction
band,
which changes
their energy
in
two w ays. First, energy
must
be measured f rom
the bottom of the band
FIGURE 1.3.
Optical absorption in a quantum wel l formed by
a layer of GaAs surrounded by
AlGaAs.
(a)
Potential well
in conduction and valence band,
showing two b ound states
in
each;
the
energy gap
of GaAs is really much larger than this diagram implies.
(b)
Transitions between states
in the quantum well produce absorption
lines between the band gaps of the GaAs w ell
and AlGaAs
barrier.
-
7/24/2019 the Physics of Low dimensional
26/457
1.3
BOUND PARTICLES: QUANTUM WELL
at
E,
rather than from
zero. Second,
electrons behave
as
though their
mass
is
mom
e,
where the effective mass m
e
0.067 in
GaAs. Thus s(k) = Ee +h2 k 2
12m o t t l
e
. The
sandwich acts like
a quantum well because
E,
is higher in AlGaAs
than
in
GaAs,
and the
difference A E provides
the
barrier that
confines the
electrons. Typically
A E
0.2-0.3
eV,
which is not large.
How ever, we shall approximate it
as
infinite
to find
the
energy levels in a well
of
width a. Ad apting equation (1.12)
shows
that
the
energy of the boun d states, labelled w ith n e
,
is
2m
o m e
a 2
W e could m easure these energy levels by shining light on the
sample and
de-
termining which frequencies were absorbed.
A photon
is absorbed by exciting
an
electron from
a
lower level to
a
higher
one, and the
energy
of the photon ma tches the
difference
in electronic energy levels. W e m ight therefore hope to see absorption at
a
frequency given by
hw =
Ee2 Eel,
for
example. Unfortunately this is
a difficult
experiment and a different technique
is usually used.
Semiconductors
have energy levels
in
other bands. The most
important of
these is
the
valence band which
l ies below
the conduction band. The top of this is at
E
v
and
the band curves downwards as a function
of
k, giving
s(k) =E, h 2 k 2 /2momh,
which contains another
effec tive mass m h (mh =
0.5 in
GaAs). The conduction and
valence bands are separated by an
energy called the
band gap
given by E
g = E,
E.
Ag ain there is a quantum well because
E, is at a different level in the
GaA s well
and AlGaAs barriers.
The energies
of the bound states
are
h
2 7
2
n 2
Ehnh
vGaAs
1.14)
2m0mha
2
Everything is 'upside down' in the valence band,
as
shown
in Figure 1.3(a).
The valence band
is com pletely
full, and the conduction band
com pletely empty,
in a pure semiconductor at
zero
tempe rature. Optical absorption must therefore
lift
an
electron from
the valence band
into
the conduction band. In a
bulk sample
of
GaAs this can occ ur provided that
hw > EgG a A s ,
the band gap of
GaAs. Similarly we
need
hco > E
AlGaAs
in AlGaAs. This
process leaves behind
an
empty state or hole
in the valence band,
so
a
subscript
h
is used to identify parameters
of the valence
band.
Now
look at the quantum well. Although the well is of
GaAs, absorption
cannot
start at
hw =
E
gG a A s because
the
states in the well are quantized. The lowest en-
ergy at which
absorption can occur is given by the
difference
in energy
Eel Eh]
between
the lowest state in the
well
in the conduction band and the
lowest state
in the
well
in the valence band.
A bsorption can occur at higher energies by us-
ing other states, and we shall see later that the strongest transitions occur between
corresponding states in the
two bands, so
set n e
=n
= n.
Therefore strong
h 2
7 2
n
2
Eene
cGaAs
(1.13)
-
7/24/2019 the Physics of Low dimensional
27/457
h
2 z 2 n 2
1
1
1
=
E
G a A s
g m0a2m
Mh
(1.15)
1 .
FOUNDATIONS
absorption occurs at the frequencies given by
ntO
n
=
(
Een
_
Ehn
=E
GaAs
h2z2 n
2
2mom
e
a 2
(E
v
GaAs
h2
7 .
t
2
n 2
2momha
2 )
The
energies look like those
in a quantum well where
the effective mass
is men,
given by 1 /
Meh =
1/M
e 1/Mh.
This is called the optical effective mass; almost
every process has its own
effective
mass
If the
wells really were infinitely
deep,
there would be
an
infinite series
of
l ines
with frequencies given by equation
(1.15).
The barriers in the
semiconductor
are
finite,
and absorption occurs in the
AlGaAs barriers
for all frequencies where
n
o )
> EAlGaAs.
The resulting spectrum is shown in Figure
1.3(b), assuming that
there
are
two bound states
in
both
the conduction and valence
bands. No absorp-
tion is
possible for hco < E
G aA
s,
and there is
a
continuous band of absorption
f
or
h c o
FA1GaAs.
Between these two frequencies lie
two discrete lines produced
'g
by transitions between states in the quantum well,
with energies given by equa-
tion (1.15). The width of
the well can be inferred from the energy of
these lines
if
the effective m asses are known.
This
is
a routine check to see that layers
have
been
grow n co rrectly .
In
practice a slightly different experiment is usually performed, called photo-
luminescence
(PL).
Light with hco > E
g
AlG
a
A s is shone o n the sample, which excites
many electrons from
the valence
to
the conduction band
everywhere. Some of
these electrons become trapped
in the quantum well, and the same thing happens
to the
holes in the valence band.
It is then
possible for an
electron to fall from
the conduction band
into
a hole in the valence band and release the difference
in
energy
as
light. This luminescence
is
the reverse
process to
absorption and
can
occur at
the
same energies. Only
the
lowest levels
are
usually seen, so the PL
spectrum should show a
line at
ha)]
.
A n
example of a photoluminescence spectrum
is shown in Figure
1.4. The
sample has four wells of different widths, each of
which contributes
a peak to
the PL. A
detailed analysis of this spectrum is left
as
an exercise.
Unfortunately the true picture is slightly
more
complicated.
One problem is that
the valence band is not as simple as we have assumed .
A
better model is to
assume
that there are two varieties of
holes, heavy and
light. Thus two
sets of spectral
lines should be seen, although
the
heavy holes
are
much more
prominent.
A
further
complication is that electrons
and
holes bind together to form
excitons,
analogous
to hydrogen atoms,
and this
modifies the
energies slightly.
This problem will be
addressed in Section 10.7.
-
7/24/2019 the Physics of Low dimensional
28/457
12
10
8 -
6
4 -
2 -
c
o
n
s
1
Ait
(a)
1.4
CHARGE AND
CURRENT DENSITIES
9 nm 6 nm 4 nm 2 nm
650
00 50
00
wavelength / n m
FIGURE 1.4.
Photoluminescence as a function of wavelength for a sample with four quantum
wells of different widths, whose conduction and valence bands are shown on the right. The barriers
between the w ells
are mu ch thicker than drawn.
[Data
kindly supplied by Prof. E. L. Hu, Un iversity
of California at Santa B arbara.]
1.4 Charge
and
Current Densities
The Schrbdinger equation yields a wave function x l i ( x , t ) , and we should now con-
sider how to deduc e quantities
of interest from it.
First we would like to know the
location of the
particle. This follows from
the
squared modulus
of the
wave function:
1W (x ,
01
2
OC
probability density
of
finding the
particle at x. 1.16)
This
does n ot strictly im ply that the particle itself is spread out:
it should be inter-
preted
as a
statement about our know ledge
of the
particle. H owever, this
distinction
will not be important for the topics covered in this book,
and we can
put the relation
in a more physically transparent
form by using
the charge density. If the
particle
has
charge
q,
this becom es
qlw (x
,
t ) 1
2
a
charge
density of particle,
1.17)
or
q 1W ( x , t)1 2 dx c x charge in a region d x around x.
1.18)
If the
electron is bound w ithin some volum e,
we know that
the total charge
enclosed
in that volume must be
q The
propo rtionality can then be turned into an
equality:
q 1
1 P
(x,
0 1
2 =
p
x) =
charge
density
of particle. 1.19)
-
7/24/2019 the Physics of Low dimensional
29/457
1 .
FOUNDATIONS
A n integral over
the total volume must
recover the total charge:
f
p(x)dx
qlkli(x,t)1
2
dx = q.
1.20)
Removing the charge q
from this equation gives
f
T(x, t)1
2
dx = 1.
1.21)
This is the standard condition for normalizing the wave function,
and
means that
t ) 1
2 is
the probability density of finding the particle.
Not all wave functions can be normalized in this way. T he free electron is an
obvious example because
the integral over all space would
diverge. In this
case
one
can talk only
about
relative
probabilities.
In
practice, we can get around this
difficulty by starting with
the
electron
in a large but
finite
box, for which there
are no
problems, and
letting
the volume of the box go to infinity at
the end of the
calculation. We shall do this in Section 1.7
to calculate
the density of
states.
Normalization gives physical
dimensions to the wave function . Take the
infinitely
deep well as an
example . The
wave functions O
n
(x) =
A
n
sin(turx/a),
and the
condition for normalization is
1 = f
an7441
sin
2
dX =
0
a
2
Thus
the normalized wave function,
if A ,
is taken to be real, is
2
u r x
On(x)=
in
a
(1.22)
(1.23)
Normalization has given
the
wave function
dimensions of (length) -1 / 2 in one
di-
mension. This
is often useful
as a check.
A plane wave such
as
Ok(x)
=
A e
i
"
n an
infinite
volume
can be normalized
in a
slightly different
way.
Here
the
densi ty I
( / ) k ( x ) 1 2
=
I A
2
,
which can be
set
to
a
given density
of
particles.
Now that we have a charge density, there should be a
current density
J (or
just current in one dimension)
associated with
i t .
These must obey the continuity
equation
a J
(1.24)
a x t
to ensure
the conservation of charge and
part icles. In
three
dimensions
8
Jlax
becomes
div
J .
To const ruct
a
current den sity, start w ith
the
time-dependent
Schrbdinger
equation
h 2 8
a
W
(x,
t)
+ v(x ,
t) tp( x, t) = ih tp(x, t).
2m ax 2
t
(1.25)
-
7/24/2019 the Physics of Low dimensional
30/457
1 .4 CHARGE AND CURRENT DENSITIES
1
M ultiply both sides
on the
left by
the
complex conjugate of the wave function,
x l i * :
h2 92
2m
k W ' x2
+ kIJ*V = IP 1.26)
at
Obtain
a second equation by going back to
the
Schr6dinger equation, taking its
complex conjugate,
and multiplying from the left by
his gives
h
2
2
LP xlirkr =
1.27)
2m ax
2
t
Now subtract
(1.27)
from
(1.26).
The
terms w ith
the potential cancel provided that
V(x, t)
is real. The
two terms
on the
right-hand side add and are clearly the
derivative
of a product,
so the
difference becom es
h2
( *
O
2
v
IT 1
2
2m x
2
x
2
t
To simplify the left-hand
side, use the rule for the
derivative
of a product:
(1.28)
( .a
ax(
j
ax qj )
kr\ /tp
O x
()T*W2 2 T
ta
(1.29)
W hen this is applied to (1.28),
the
products of single derivatives cancel and
it reduces
to
h2
tp*
w
v*)
4JF .
2m ax
x
t
(1.30)
Finally, moving the factor of ih
to
the left and
multiplying throughout by
q to turn
the probab ility densities into charge densities gives
O
r h
q
(w*
* V 1 =
q
1 2
) =
ax
L2im
x
x 1
t
t
(1.31)
Com paring this w ith
the
continuity equation
(1.24) shows
that
the current density
is given by
J(x,t) = hq
(kr
tp
p .
1.32)
2im
x
x
In three dimensions
the
derivative atP/ax becomes the gradient
V
1 1 .
The dependence
on
t ime vanishes from both p
and J
for a stationary state be-
cause exp(icot) cancels between
k I J and k l i * . This partly explains
the
origin of the
term 'stationary', althoug h the states may still carry a
current (constant
in time)
so it
is slightly m isleading. H owever,
a
stationary state where
lk
(x) is purely real carries
no c urrent. This
applies to
a
particle in a box and to bound states
in
general.
A
su-
perposition
of
bound states is needed to generate a current. This feature emphasizes
that the
wave function in quantum
mechanics is
in general
a 'genuine' com plex
quantity.
This contrasts with
the
complex notation widely used for oscillations in
-
7/24/2019 the Physics of Low dimensional
31/457
12
.
FOUNDATIONS
systems ranging from electric circuits to balls
and
springs. Here the
response is real
and the
comp lex form is used only for
convenience.
As an exam ple, consider tli(x,
t) =
A
exp[i(kx cot)],
which describes
a plane
wave m oving in the
+x
-direction. Its charge density
p = qIAl
2
,
uniformly over all
space, and the
current
J = q(hk1m)
IAl
2 . Now hklm = plm = v so
J =
pu,
which is the expected result (like
'J = nev').
The
Schrtidinger
equation is linear, so further wave func tions can be c onstructed
by superposing basic solutions. For example,
kIJ(x,
t) = [ A + exp(ikx) +
xp(ikx)] exp(icot)
1.33)
describes a superpo sition of
waves
travelling in opposite directions. The
quantum-
mec hanical ex pression for the
current gives
the
expected result
hqk
=(IA+I IA-1 2
).
1.34)
There is an interesting result for tw o counter-propagating
decaying
waves,
k V (x , t) = [B+ exp(Kx) + B_ exp(K x)] exp(ic.ot). 1.35)
Neither com ponent would
carry a
current by itself because it is real,
but the
super-
position gives
J =
hqic
(B B *
B* B_)
2hqK
Im(BB*).
(1.36)
The
wave
must contain components decaying in
both
directions, with a phase
dif-
ference between them , for a current to flow. This effect is shown
in Figure 1.5 for
a
wave hitting a barrier. We shall see in
Chapter 5 that an oscillating wave turns
into a decaying one
inside a
high barrier. If the
barrier is infinitely
long, it contains
a single
decaying wave
and
there is no net
current. A
finite barrier,
on the
other
hand, transmits
a
(small) current
and must
contain two counter-propagating decay-
ing waves.
The
returning wave (exp
Kx) from the far end of the
barrier carries
the
information
that the
barrier is finite and that a current flows.
(a)
FIGURE 1.5.
Current carried by counter-propagating decaying waves.
(a) An
infinitely thick
barrier contains a single
decaying exponential that carries
no current. (b)
A
finite barrier contains
both growing and
decaying
exponentials
and passes
current.
(The wave function is complex,
so
the
f igure
is only
a rough guide.)
-
7/24/2019 the Physics of Low dimensional
32/457
1.5
OPERATORS
AND
MEASUREMENT
3
1 .5
Operators
and Measurement
It is now tim e to return to the theory of quantum mechanics in a
little
more
depth,
and
to see how physical quantities can be deduced from
the
wave function.
1.5.1
OPERATORS
It is
a
postulate
of quantum mechanics that observable quantities can be represented
by op erators that act on the
wav e function (although it is a
further postulate that
the
wave function itself is
not
observable). Operators will be denoted w ith
a
hat or
circumflex.
The position, mom entum, and total energy can be represented by the
following operators
on
kli(x,
t):
x
x,
1.37)
a
p ih, 1.38)
ax
E
ih
.
a t
(1.39)
A n important feature is that
the
momentum
/ 3
appears as a
spatial
derivative.
More
com plicated operators can be constructed from these compon ents.
For ex-
ample, the
Ham iltonian function
H
= p2 /2m V(x) gives the total
energy of a
classical particle
in the type of
system that we
have
studied, where energy is con-
served. This becomes a Hamiltonian operator
fi in quantum mechanics
and is given
by
2
a 2
= H (2 , )3) =
-
v(x).
2m ax
2
(1.40)
Equating the effect
of
this operator w ith that
of the energy operator gives 1
7
:141 =
EkIJ ,
or
[
2
h
a
W
(x, t). 1.41)
x)
x, t) =
2m ax2
t
W e
are
back to
the
time-depend ent Schr6dinger equation (1.1).
The
t ime-independen t Schriidinger equation can now be w ritten concisely
as
i f * (x) = (x),
where E is a number, not an operator.
This resembles
a matrix
eigenvalue e quation: there is an operator
acting on the
wave function
on one side,
and a constant
mu ltiplying it
on the
other.
The
ideas of eigenvectors and eigenvalues
work
in much the same way
for
differential operators
as for matrices, and
similar
terminology is used. Here 1t is called
an eigenfunction or eigenstate, and
E is
the
corresponding eigenvalue.
This
will be develop ed further
in Section 1.6.
-
7/24/2019 the Physics of Low dimensional
33/457
14
.
FOUNDATIONS
The current density can be rewritten
in
terms of the momentum operator, giving
J(x,t) = 1 -
141)
(
1 -1
41)
2
This show s that
the
current is related to
the velocity
p/ in. A more
elaborate expres-
sion is needed
in a
magnetic field, which complicates the relation between velocity
and
momentum.
This
will be considered in
Chapter
6 .
N o w look
at the
effect of the momentum operator on a wav e funct ion.
A plane
wave
iii(x) =
A exp(ikx) gives
fr = in (Ae i k x) hkAe i k x = (hk4f.
1.43)
dx
Again this has reduced to
an
eigenvalue
equation. We interpret this
as
meaning that
the momentum has
a definite value
p =
fl k,
a result which we inferred earlier by
analogy with classical mechanics.
A
further postulate
of quantum
mechanics states that
the only
possible values
of a
physical observable are the eigenvalues of its corresponding operator. If the
wave function is
an
eigenfunction
of
this operator,
as in the case of the plane wave
and momentum, the observable has
a
definite
value. In general this is not
the case.
Consider the
effect o f the
momentumoperator
on a
particle
in a box:
d
ihnz
A ,
zx
(1),(x)
ih
dx
A
n
si
nzx
n
os
1.44)
a
These wave functions
are
not eigenfunctions
of f9,
and
therefore
do
not have a
definite
value of
momentum. Measurements of
momentum would yield a range of
values which we could characterize
in
te rms of an average value (zero
here) and
a
spread. Taking another derivative
shows
that O n (x)
is
an
eigenfunction of /3
2 .
It
therefore has
a
definite value of kinetic energy, whose operator
t = 13
2
12m.
Similar issues arise
when we measure the position of a
particle, which we shall
consider nex t .
1.5.2
EXPECTATION VALUES
Suppose
that we
are
given ( x , t) for some particle. Two simple quantities that
we might wish to know
are the average position of the
particle and how well it
is localized
about
that position. Note
that we cannot say that
the particle is at a
particular
point ,
unl ike
in classical mechanics, because we
are
using
a
picture based
on
waves .
W e know tha t
the probability density for finding a
particle is
P(x,t) cx
W
(x,
t)1
2 ,
and
can
use
this in the standard formula for
finding a m e a n value. This gives
( 1.42)
(x(t)) = f x P(x,t)dx,
1.45)
-
7/24/2019 the Physics of Low dimensional
34/457
1.5 OPERATORS AND MEASUREMENT
5
where
angle
brackets
( )
are
used to denote exp ectation
values. For
normalized wave
functions this becomes
(
x(t))
=
f x141 (x , 01
2
dx = f W*(x
To answer
the question of how well
the particle is localized,
the standard
deviation
A x defined by
(Ax )2 =
(
x
2)
x
)2 ,
where (x
2
) is the expectation
value of x
2
, given in the
(x
2
) = f x 2 P(x) dx = f
T*(x, t)
Take the
lowest state of a particle in a box as an
example.
2
r X
dx
f
(X) = s i n
2
, t) x 41(x , t) dx .
a common
same way by
x 2 ) dx .
Then
a
2'
1 \
(1.46)
measure is
(1.47)
(1.48)
(1.49)
(1.50)
(1.5 1 )
a
which is obvious from sym me try,
and
2
TX
2
x 2
)
=
f x
2
s in
2 dx
a
a
Thus
3
1
7 2 )
18 a
1
A x
--
12 7 2
The
particle is most likely to be foun d in the middle
of the
well, but
with considerable
spread around this (which increases for higher states).
The same questions can be asked about
the
momentum
of the
particle
and
can be
answered
in the same way using
the momentum operator.
The general expression
for the expectation value
(q)
of some physically observable quantity
q
is
(q) = f W*(x, t) W(x, t) dx ,
1.52)
where is
the
corresponding operator.
For example, the average value of the mo-
mentum is given by
in
w x ,
x
8x
This can be ex tended to quantities such
as
( p
2
)
and
A p
as
was done
for
x.
The results of these expressions are physical quantities, and must
therefore be
mathematically real numbers. This
requires that physical quantities be rep resented
by Hermitian operators. Such operators
have real
eigenvalues, which guarantees
( p) = f x , t)
x , t)dx =
(1.53)
-
7/24/2019 the Physics of Low dimensional
35/457
16
.
FOUNDATIONS
that measurements
on the
wave function will yield real values. Their properties are
reviewed briefly
in Section 1.6. Non-H ermitian operators are important in other
applications, notably
as
creation
and annihilation operators
in field
theory, but will
not be used
in
this
book.
Expectation
values of stationary states
are constant in time,
because their depen-
dence
on t ime cancels between
k I J and W". For example, (x) is constant for
any
stationary state, so the
particle appears to be 'stationary'. A superposition of states
is required for the
particle to 'move' in the sense that (x) varies with t ime.
Going
back to the one-dimensional well again, we can construct
a
moving w ave function
from the first two states,
t
= 0) = 2414)1(x) + A202(x)
A s
this wave function evolves
in time, the average position
becomes
a
2a,41,42
(82 e i)t i
.
(x(t))
=
os
2
7 1 . 2
(1.54)
(1.55)
The
particle oscillates back
and
forth in the
well at angular frequency
(82 1)
h
given by
the
difference
in energy
of the
two levels.
The
analysis is left
as an exercise.
1.5.3
MOTION OF A
WAVE PACKET
Elementary classical mechanics rests on the concept of point
particles, whose po-
sition and mo men tum can be specified precisely,
but
this is not tenab le in
wave
mechanics.
The
natural
analogue
is a
wave packet like that in Figure 1.1, a wave
that is restricted to
a finite region by
an
envelope. This
also provides another illumi-
nating example of expectation
values. Start with
a plain carrier wave
exp(ip
o
x/h),
and m odulate it with
a
Gaussian envelope at
t
= 0:
1
x
)
(x x0)
2
1
(x ,
t
= 0) =
xp
xp
(27d
2
)
1 /
4
d 2
(1.56)
The probability density of this is
a
normalized Gaussian function with mean xo and
standard deviation
d:
[
(x xo)
2
IT(x, t =
0
) 1
2
=
(27rd2)1/2
ex
2d
2
(1.57)
It is clear from this that (x) = x o
and Ax =
d at t
= 0.
W e can make
the wave
packet
as
localized as
we d esire by choosing
an
appropriate
value of d.
The carrier
has defini te m omentum po but
we
have
had to m ix many waves
together to get the wave p acket, so there is now a range of momenta in the
wave
function. There are
two ways of
extracting
(p)
and Ap. One
is to
use the definitions
-
7/24/2019 the Physics of Low dimensional
36/457
1.5
OPERATORS
AND MEASUREMENT 7
of the
expectation
values
like equation
(1.53)
given earlier. The
other way is to
write the
wave function
as a
function
of mo mentum rather than
position.
Since
we kno w that a plane wave exp(ipx/h) has definite m omentum p, the distribution
of
mo menta w ithin W is given by resolving i t into
plane
waves
-
just a Fo urier
transform. Thus
the wave function
in
mom entum space (I)
(p, t)
is related to that
in
real space by
(+_px\ dp
t) = f
I ) (p, t) exp
h 12.7T
'
dx
(I)(p,
t) = f
)
exp
h )
N
/27rh
(1.58)
(1.59)
The factors of
,,,[2
-
th
ensure that
s z l )
has the
same normalization
as
W .
Taking
the Fourier transform
of
equation (1.56)
for the
Gaussian wave packet
gives
1
-i (p - po)xo]
exp
[
( Po) 2 1
(1)(p,
t
= 0) =
[27 (h 12d)2]1/4
exp
(h/2d)
2
(1.60)
whose p robability density is
1
(
2
P PO)
4(p, t
= 0)1 =
xp [
1.61)
[27 (h/2d)
2 ]I/2
(h/2d)
2
j
This is another norm alized Gaussian with m ean (p) = p
from the carrier and
standard
deviation Ap =
h I2d .
A n important
result comes from the
product of the standard deviations in
space
and
momentum:
h
Ax Ap =
d
d
= 1.62)
Thus
the
better we localize
the
particle to fix its position in
real space,
the more
waves we need
and the wider
the
spread in mom entum becomes. This
is
the
fa-
mous Heisenberg uncertainty principle: we cannot m easure both
the position and
momentum of a particle to arbitrary precision. It contrasts with
the
classical picture
where both x
and
p could be know n precisely. Gaussian wave p ackets happen to
give the minimum uncertainty, an d in
general
the result is
Ax Ap > 1.63)
As one example, a plane wave
exp(ipo x
h) definitely has mom entum po
, so
Ap = 0,
but it is spread even ly over all space giving
Ax =
Do.
The uncertainty principle also forces the lowest state
in a quantum
well to have
non zero kinetic energy, unlike classical mechanics where it wo uld be
still. The
momentum
p
wo uld be know n exactly (zero) if the particle were at rest,
giving
-
7/24/2019 the Physics of Low dimensional
37/457
1 B
. FOUNDATIONS
Ap = 0,
while
A x
is finite because w e know that
the
particle is somewhere in the
well . Thus
Ax Ap = 0, which is not allowed. The only w ay around this is for the
particle to
have a
zero-point energy in the lowest state so that both Ax and A p are
non-zero.
This
can be used to estimate the zero-point energy. C onsider, for
example,
an
infinitely
deep
potential well of width
a. W e know that
the
particle is in the well,
so roughly A x a/4.
This
me ans that Ap h/2Ax =
2h/a.
Taking the kinetic
energy as
(Ap)
2 /2m
gives
an estimate
of (h
2 /2m)(2/a) 2 for the
energy of the
ground state. The exact
result has
7 r
instead of
2.
This
explains
the
dependence
of
the
energy levels on the
width a:
making
the well narrower reduces the spread
of
the
particle in
real space
and
therefore increases its
range of momenta and hence
the
energy. This
principle can be extended to estimate
the zero-point
energy
in any
well by including
the
mean potential energy.
Returning to
the
Gaussian wav e packet, we foun d that it has
minimum uncertainty
(in the sense
of the
product
Ax Ap)
at t = 0, but this changes as it evolves in
t ime.
W e know that a plane wave
exp(i px Ih) evolves
in
time like exp(i tot) with
hco =
p
2 /2m.
This
applies to each Fourier component
of the
wave pack et, so the
wave function
in Fourier space for t >
0
is
1
szti(p, t) =
[27 (h12d)21114
exp
[-i(P-
h
13)x exp
(h/2d)2
P P())2
1 exp (
2h
iP2t).
m
(1.64)
W e
must
transform this back to real space to find
(x ,
t). A little rearrangement
gives
(
p
o (x po t 12m)1
x, t)
xp [
[27 (h12,d)
2
]
1 1 4
x
f
exp
[
i po)(x
x o pot/m)1
h
x exp
(p
p
0
) 2
ht
p
L 4(h/2d) 2
m d 2
LI I23
Th
(1.65)
The prefactor gives
a ca rrier
wave with momentum po, moving at the phase velocity
v
p
h = p0/2m. Inside
the
integral,
the first exponential
shows
that
the
wave packet
is now centred
on x o
+
p
o t I
m and
therefore m oves at
the
group velocity v p =
po/m.
The second
exponen tial, which controls the width of the wave packet, is also
modified. Evaluation
of the
integral
shows
that
ht
Ax(t)
1.66)
2,n2d)
-
7/24/2019 the Physics of Low dimensional
38/457
1.5
OPERATORS AND MEASUREMENT
The pulse spreads
out in space
as
it propagates. The mom entum remains unchanged
if there are no forces acting o n the
particle, so
the product
A x
zip grows
and
our
information abou t the
particle deteriorates in time. This
is the
typical effect
of
dispersion as
seen,
for
example,
in communications.
Dispersion arises because a wave packet necessarily co ntains
a range of mom enta,
each of which propagates at a different velocity causing the wave p acket to spread.
Eventually this overwhelms the initial
width.
The range of velocities is (Ap)Im
so at
large times we expect Ax (Ap)t I
m = ht 12md,
in agreement with equa-
tion
(1.66).
A short pulse
contains
a wider
range of m omen ta than a longer pulse
and will eventually become longer.
1.5.4
FURTHER PROPER TIES
OF
OPERATORS
The uncertainty relation can be traced back to properties
of the operators involved.
W e are trying to measure both the position and momentum of the particle described
by the wave packet. A problem arises because of the
order of these operations.
Suppose we first measure
the
mom entum, then the position. The operators acting
on the wave function are
iptp = x ( ih
k p inx
aT
ax
x
The op posite
order gives
= ( i
w =
in (c a
(1.67)
(1.68)
The
last line follows from
the
derivative
of the produc t. Clearly the results
are
different
and the order
of the
operations is significant. Subtracting
the
two gives
i
1.69)
Since this equation holds
for any T,
we can w rite it
for the
operators alone
as
[2,
h.
1.70)
The no tation ri , p 1
is called a
commutator.
Two operators are
said to
commute
if
[A, B] =
0 since the order
of their operation is unimportant. It is
possible to
measure two physical quantities simultaneously to arbitrary accuracy only
if
their
operators commute. Clearly this does n ot apply to
x
and
p,
and their accuracy is
limited by
the
uncertainty p rinciple.
Similar relations apply to other coordinates
and their corresponding m om enta
such
as [j), fry] =
in.
On the other hand
[ 5 ' ,
fix
] = 0,
so these quantit ies can be
measured simultaneously. Some further examples are given
in the problems.
-
7/24/2019 the Physics of Low dimensional
39/457
20
. FOUNDATIONS
A quantity w hose operator commutes with the
Ham iltonian is called a constant of
the motion because its value
does not change
with
time. For example, [P, = 0
for a free
particle so i ts m omentum remains constant. These constants usually
arise
from some symm etry
of the
system, translational
invariance in
this
case.
W e
have
seen that
the
order of
operators such
as
and p
is important and
that
they cannot be reordered like numbers. The
sam e is true
of matrices, and
we shall
see later that operators can be represen ted by
matrices instead
of the
differential
operators used here. Further, the choice of operators depends on the way in which
the
wave fun ction is represented. W e derived the wave function of a wave packet in
mom entum space before and could use the corresponding operators
a
= ih.
1.71)
ap
These obey the
same
commutation relation
] =
ih
as the
earlier forms
in x
and are
therefore
an equally valid choice.
1.6Mathematica Properties
of
Eigenstates
This is
a
brief
section on formal properties of
eigenstates,
w hich w ill be needed
later
in the construction of perturbation
theory. Further details can be found
in a
book on mathematical methods
for
physics such as
Mathews and
Walker (1970).
W e
have already seen that the
wave functions
in the
infinitely deep square
well
can be norm alized. Assume that w e are dealing with
a finite system, so we c an
ignore the
problems posed by plane waves
and the
like.
Let the eigenstates (wave
functions) of the
Ham iltonian be
(/), (x)
with corresponding
eigenvalues (energies)
E n
,
and norm alize each state such that
f
14),(x)1
2
dx =
1 .
1.72)
The range of
integration covers
the
region w ithin which the particle can m ove,
0
-
7/24/2019 the Physics of Low dimensional
40/457
-
7/24/2019 the Physics of Low dimensional
41/457
1 . FOUNDATIONS
W e know that each eigenstate
evolves
in
time like O n (x) exp(i E n tlh), and therefore
the
given state evolves
as
n
kI J (x, t) =
n
ck, (x)
exp
Et)
h
1.81)
n=1
This is how we followed
the
evolution of the wave packet
in Section
1.5.3.
1.7
Counting
States
A complete
description of a
system requires the energies
and wave functions
of
all
its states. Clearly this is
an impossible
task
for
anything
but the
simplest systems,
and
most
of the information would
in any
case be unwanted.
For many applications
the
density
of
states
N (E) is adequate.
The definition is that
N (E) BE is the number
of states of the system w hose energies
lie in the range E
to
E 3E.
Clearly this
tel ls u s nothing about the wave functions at all, just
the distribution of energies. W e
shall first calculate
the
density of states of a one-dimensional system before looking
at more
general results.
1.7.1
O NE DI M E NSI O N
A n imm ediate problem, as we saw
in the
previous section, is that
the wave functions
exp(ikx ) cannot be normalized in the
usual way
if the particles travel through all
space. The simplest way around this problem is to put the particles in a
finite box
of
length
L, and se t L
oc
at the end of the calculation. Hav ing put the particles
in a box, we need to choose boundary condit ions. Two are
comm only used.
(i )
Fixed
or box
boundary conditions, in
which the
wave function vanishes at
the
boundary:
f (0) = f(L) =
0.
1.82)
(ii) Periodic or Bornvon Karman
boundary
conditions, in which we
imagine
repeating
the system p eriodically with the same w ave function
in
each sys-
tem. The
wave function at
x =
L must match
sm oothly to that at
x = 0,
which requires
(0) =
(L),
a l f r
(1.83)
a x
x=0
x
x=L
Fixed boundary
conditions are
obviously
the same as the
particle
in a box studied
earlier.
The
energy levels are given by
E0(k) =
h
2
k
2 /2/12 and the allowed values of
k
are
J r
in
L
m = 1,
2, 3, . . . .
1.84)
-
7/24/2019 the Physics of Low dimensional
42/457
1.7
COUNTING STATES
3
The wave functions are standing waves w ith this choice. They carry no
current, and
the
allowed values of k are all positive.
Periodic boundary
conditions require
a different choice
of
k.
W e can
use trav-
elling exponential waves rather than
sine waves, and
they
must
obey
exp(ikL) =
exp(i k0) = 1 = exp(27ni). This
also satisfies
the condition on the gradient, and
the
norm alized states
are O
n
(X)
= L - 1 1 2
exp(i
k
n
x).
The allowed
values of
k are
27n
=
= 0, 1,2,
(1.85)
These
are
twice as far apart
as
with fixed bound ary
conditions, but both signs
of k
are
permitted and there are two degenerate states at each energy level (excep t for
k =
0),
with opposite signs of k and velocity.
This
raises the
following
crucial question: does the density
of
states, which we
are
trying to
calculate, depend on
which boundary conditions
we choose to apply to our
artificial
box?
Fortunately it can be shown that
the
result is insensitive to bo undary
cond itions as
L co. It is usually
more appropriate to treat
free electrons
as
travelling
rather than
standing
waves, so periodic boundary conditions are
generally
used.
To turn these allowed
values of k and
E
into a density of
states,
plot the allowe d
values of k along a line
as in Figure
1.6. This
is a simple
one-dimensional