QM2
QM2
on
Quantum Mechanics
with Examples of Solved Problems
Ennio Gozzi
gozzi@ts.infn.it
La meccanica quantistica?
Non `e altro che materia
che dorme e sogna
Fabia Gardinazzi
Poetessa
(19592010)
TABLE OF CONTENTS
Page
Introduction xi
Chapter 1: Schrodinger equation. Wronskian theorem. 1
1.1 The time-dependent Schr odinger equation: general properties . . . . . 1
1.2 Solution of the time-dependent Schr odinger equation . . . . . . . . . 2
1.2.1 Properties of the solution . . . . . . . . . . . . . . . . . . . . 3
1.3 Discrete energy levels . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 Qualitative analysis . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 Quantitative analysis and Wronskian theorem . . . . . . . . . 7
1.4 The Dirac delta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Harmonic oscillator and Hermite polynomials . . . . . . . . . . . . . 15
1.6 Schr odinger equation for multidimensional problems . . . . . . . . . . 19
1.7 Central potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.8 Problems and Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Chapter 2: An Introduction to Hilbert Spaces for Physicists 37
2.1 Basic denitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.1.1 Linear space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.1.2 Scalar product . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.1.3 Sequences and convergence . . . . . . . . . . . . . . . . . . . . 39
2.1.4 Orthonormal systems . . . . . . . . . . . . . . . . . . . . . . . 41
2.2 Examples of Hilbert spaces . . . . . . . . . . . . . . . . . . . . . . . . 43
2.2.1 The space l
2
() . . . . . . . . . . . . . . . . . . . . . . . . . 43
vi Table of Contents
2.2.2 The space /
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.3 Operators in Hilbert spaces . . . . . . . . . . . . . . . . . . . . . . . 45
2.3.1 Projection operators . . . . . . . . . . . . . . . . . . . . . . . 48
2.4 The eigenvalue problem . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.5 Functions of one operator . . . . . . . . . . . . . . . . . . . . . . . . 53
2.6 Commutativity and compatibility . . . . . . . . . . . . . . . . . . . . 54
2.7 Complete set of commuting operators . . . . . . . . . . . . . . . . . . 56
2.8 Dirac notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.9 Spectral decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.10 Tensor Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.11 Problems and Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Chapter 3: Postulates of Quantum Mechanics 69
3.1 Diracs correspondence rules . . . . . . . . . . . . . . . . . . . . . . . 73
3.2 More on the postulates of quantum mechanics . . . . . . . . . . . . . 75
3.2.1 Expectation values . . . . . . . . . . . . . . . . . . . . . . . . 75
3.2.2 Compatibility of observables and measurements. . . . . . . . . 76
3.2.3 Incompatible observables . . . . . . . . . . . . . . . . . . . . . 77
3.3 Heisenberg uncertainty principle . . . . . . . . . . . . . . . . . . . . . 79
3.4 Position and momentum representations . . . . . . . . . . . . . . . . 81
3.5 Solution of the time-dependent Schr odinger equation . . . . . . . . . 88
3.6 Density Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.6.1 Time Evolution of the Density Matrix . . . . . . . . . . . . . 95
3.7 Problems and Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Chapter 4: Harmonic Oscillator in Operatorial Form. 109
4.1 Harmonic Oscillator in Operatorial Form. . . . . . . . . . . . . . . . . 109
Chapter 5: Angular momentum quantization 115
5.1 Spectrum of the angular momentum operators . . . . . . . . . . . . . 117
5.2 Spherical Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Table of Contents vii
5.3 Addition of angular momenta . . . . . . . . . . . . . . . . . . . . . . 130
5.4 Clebsh-Gordon Coecients . . . . . . . . . . . . . . . . . . . . . . . . 135
5.5 Problems and Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 136
Chapter 6: Symmetries in Quantum Mechanics 143
6.1 Rotations, symmetries, unitary transformations and degeneracy . . . 143
6.2 Symmetries, generators and degeneracy . . . . . . . . . . . . . . . . . 147
6.3 Brief introduction to Lie groups. . . . . . . . . . . . . . . . . . . . . . 149
6.4 Scalar and Tensor Operators . . . . . . . . . . . . . . . . . . . . . . . 151
6.5 Wigner-Eckart Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 153
Chapter 7: Spin 159
7.1 Quantum description of particles carrying spin . . . . . . . . . . . . . 161
7.2 Pauli equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
7.3 Singlet and triplet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.4 Explanation of the Stern-Gerlach experiment . . . . . . . . . . . . . . 171
Chapter 8: Identical particles 173
8.1 General considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.2 Pauli exclusion principle . . . . . . . . . . . . . . . . . . . . . . . . . 176
8.3 Exchange interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
8.4 Slaters determinant . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
8.5 Entangled states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Chapter 9: Hydrogen atom 181
9.1 Degeneracy of the energy levels . . . . . . . . . . . . . . . . . . . . . 189
9.2 Form of the eigenfunctions . . . . . . . . . . . . . . . . . . . . . . . . 189
9.3 Zeeman eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
9.4 Landau Levels Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 193
9.5 Review Problems and Solutions . . . . . . . . . . . . . . . . . . . . . 197
viii Table of Contents
Chapter 10: Time-independent perturbation theory 217
10.1 Non-degenerate case . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
10.1.1 First order corrections . . . . . . . . . . . . . . . . . . . . . . 221
10.1.2 Second order corrections . . . . . . . . . . . . . . . . . . . . . 221
10.1.3 Correction to the states . . . . . . . . . . . . . . . . . . . . . 222
10.1.4 Projection operators in perturbation theory . . . . . . . . . . 224
10.1.5 Summary of formulas for rst-order perturbation theory . . . 225
10.2 Perturbation theory: degenerate case . . . . . . . . . . . . . . . . . . 226
10.3 Problems and Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 232
Chapter 11: Schrodinger, Heisenberg and interaction pictures 235
11.1 Schr odinger and Heisenberg pictures . . . . . . . . . . . . . . . . . . 235
11.1.1 Time evolution of
O
H
(t) . . . . . . . . . . . . . . . . . . . . . 237
11.1.2 Time evolution of the average values. . . . . . . . . . . . . . . 239
11.1.3 Constants of motion . . . . . . . . . . . . . . . . . . . . . . . 240
11.1.4 Evolution equation of q and p and Ehrenfest equation . . . . . 241
11.2 Interaction picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
11.2.1 Evolution of the operators in the interaction picture . . . . . . 244
Chapter 12: Time-dependent perturbation theory 245
12.1 General derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
12.1.1 Structure of f(t,
b,a
) . . . . . . . . . . . . . . . . . . . . . . . 253
12.1.2 Transition to nearby states . . . . . . . . . . . . . . . . . . . . 253
12.1.3 Transition probability per unit of time and Fermis golden rule 256
Chapter 13: Path integral approach to quantum mechanics 259
13.1 General derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
13.2 Perturbation theory via path integrals . . . . . . . . . . . . . . . . . 269
Table of Contents ix
Chapter 14: Semiclassical (WKB) quantum mechanics 273
14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
14.2 Hamilton-Jacobi formalism . . . . . . . . . . . . . . . . . . . . . . . . 275
14.3 WKB method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
14.3.1 General considerations . . . . . . . . . . . . . . . . . . . . . . 278
14.3.2 WKB solutions and their validity . . . . . . . . . . . . . . . . 279
14.3.3 Connection formulae . . . . . . . . . . . . . . . . . . . . . . . 282
14.4 decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Chapter 15: INTRODUCTION TO SCATTERING 287
15.1 introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
Chapter 16: The Schrodinger Field and Second Quantization 289
16.1 The many body problem. . . . . . . . . . . . . . . . . . . . . . . . 294
16.2 Second Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
16.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
INTRODUCTION
These notes on Quantum Mechanics (QM) are not original but are based on material
I took and adapted from many books: Landau, Messiah, Sakurai, Cohen-Tannoudji,
etc. and in large part from the handwritten notes of my colleague GianCarlo Ghirardi
who taught this course for many many years before retiring. Dierently from many
other books, these notes contain 34 problems with their detailed worked-out solutions.
I will present in class at least 10 extra problems with solutions of which I will provide
copy of my handwritten notes. Several other problems are left as homeworks and I
advise you to do them because they may end up being part of the nal written exam.
All these problems are not original but taken from various sources: Landau, Lim,
DEmilio and Picasso, et al. and some, very original ones, were generously provided
to me by GianCarlo Ghirardi.
My personal advice to the students of this course is to study rst these notes (or
even better the original handwritten ones of Ghirardi) and only later on, if needed,
to turn to the books quoted above. This advice is given only in order to speed up the
preparation of the exam and avoid getting lost in the details and notation of many
dierent books, notation that I have tried to make uniform in these notes.
A second advice is to work out many more of the 50 problems presented and solved
in these notes. The solution of problems is what a physicist should learn to do in
every course and later on in his professional life. It is the only manner to really master
the theoretical aspects presented in class or learned from the book.
A third very important advice (but also very personal ) is to waste no time, at
xii Introduction
this stage of your preparation, in wondering about the very strange behaviour of
quantum mechanics. This strange behaviour ranges from interference eects, to the
tunnel eect, to the Schr odinger cat paradox, to the non-locality intrinsic in QM and
many other things. This strange behaviour is a consequence of a set of postulates
which are given at the beginning and that we will spell out in details. Once these
postulates are accepted, the strange behaviour of QM comes as a consequence. So
somehow I like to look at quantum mechanics as a mathematical theory with its own
postulates and consequences (theorems) and one should not wonder too much if the
consequences are rather strange. After all, the experiments conrm this strange
behaviour. Remember that in physics only experiments are the ultimate test of a
theory and QM is the best tested theory around. The strange behavior is only due
to the fact that we, as human beings, are macroscopic objects and so we can develop
an intuition only about macroscopic phenomena and not about microscopic ones like
the things that happen, for example, in an atom and that are described in a strange
way by QM. I would like sometime to be transformed into an electron for just a day,
and I would like that not only because in that manner I would loose weight, but also
because for one day I would experience a dierent world and develop the intuition to
understand QM. Remember anyhow that the role of theoretical physics is not to
understand phenomena but to develop the equations which describe these physical
phenomena and that can predict new ones. In this respect the strange theory which
is QM has been an extremely successful one. So as an electron I would only be able to
develop an intuition for these strange phenomena but I would not be able to explain
them.
As I said I like to see QM as a mathematical theory with its own postulates and
theorems and corollaries of theorems and so on. In a mathematical theory the only
things that we can do are the following:
Introduction xiii
1. to check that the postulates are not in contradiction with each other (and some-
times contradictions are subtle);
2. to check if the postulates are redundant, so that we may be able to nd a smaller
set which is sucient to explain all of QM phenomena;
3. to check if we can replace one postulate with a new one and still obtain a con-
sistent theory.This is somehow analog to what has been done in non-Euclidean
geometries where the fth postulate of Euclid has been replaced by a new one
and all the others left the same. The new theory was consistent, had no con-
tradiction but it described a new world (a curved one instead of the at one
described by the Euclidean geometry).This same thing could happen in QM
where, with some a new postulate replacing an old one, we may obtain a con-
sistent theory but which describes a dierent world;
4. to check if one or two of the postulates could be unied in a new postulate
with the same physical consequences we have tested so far and the potential to
explain new phenomena. This is the road followed, for example, by those who
modied the Schr odinger equation (for a review see ref [1]). With the postulate
of a new and dierent time evolution, they could replace both the postulate of
the Schr odinger time evolution and the one of collapse of the wave function (at
least for measurement of the position operator);
5. to check if the postulates of QM could be proved as theorems of a more basic
theory than QM.
These are the things that we, as theoreticians, should do instead of wondering about
the strange behaviour of QM. As experimentalists instead we should look for new
phenomena which QM could (or could not) explain.
xiv Introduction
A deep unsolved problem that I personally feel is still open is the transition from
the microscopic to the macroscopic world and the interaction between the two. The
chair on which you are seated is made of hundred of trillions of billions of atoms and
so it should be possible in principle to describe it using QM because atoms behave
according to QM, but instead we are unable to describe the chair using QM. So
we use an eective theory which is classical mechanics (CM). I call it eective
because somehow it manages to describes hundred of trillions of billions of atoms in a
single stroke. People usually think that the transition from QM to CM is achieved by
sending the Planck constant to zero. This is not correct, not only because the Planck
constant has dimension but also because in this semiclassical limit we do not get
exactly CM but CM plus some phases which spoils the classical behaviour. The
same happens when, instead of the semiclassical limit, we take the limit of large N
(N being the number of atoms). So the problem of how to get CM from QM is still
open. Also the problem of how a macroscopic object (subject to the rules of CM)
interacts with a microscopic one (subject to the rules of QM) is still open and it is at
the basis of the so called measurement problem and of all the modern mesoscopic
physics phenomena discovered over the last 20 years. Attempts to correctly achieve
the transition from QM to CM have been done by many people. The two I know
better are summarized in ref.[2] and [3].
The solution of the problem of the transition and interaction between QM and CM
may shed some light not only on several of the strange features of QM and on
mesoscopic physics phenomena but also on CM itself. As we saw before, CM is only
an eective theory and not a fundamental one like QM. It could (in principle) be
obtained from QM by using some procedure and limit (but not the ones mentioned
above) like for example the one of considering as basic variables blocks of phase space
of dimension and not points in phase-space. This may overthrow concepts like
Introduction xv
the one of degrees of freedom and similar ones which are considered well-established
and not questionable concepts of CM. In this procedure it will for sure emerge that
CM is not only an eective theory but also an approximate one . This may
lead to the conclusion that CM is not the correct theory to describe all macroscopic
phenomena. Actually there are some macroscopic phenomena that we are unable to
explain and which were discovered over the last 30 years. They are those astronomical
phenomena which need the introduction of theoretical concepts like dark matter
and dark energy. May it be that instead of all these dark things of which we have
not found any experimental evidence, the way out is that CM may have to be modied
(or improved in his approximate status) at those large scales? These improvements
can come from a more rigorous and less approximate derivation of CM from QM. All
this gives you an idea of the problems that we may be able to solve starting from a
better understanding of QM.
In this book you will not nd any of these more advanced topics or others like
quantum relativistic equations, modications to the Schr odinger equation, entangle-
ment, non-locality, quantum information, teleportation etc. which are left for more
advanced courses taught by my colleagues at the University of Trieste.
These notes are in this L
A
T
E
X and Web form thanks to the hard work of Alessandro
Candolini and the nancial support of the Consorzio per lo sviluppo della sica
of Trieste University. I wish to warmly thank both Alessandro and the Consorzio
expecially his president prof. Ghirardi and its director prof. Mardirossian. I wish
also to thank in advance those colleagues and students who, reading these notes, will
indicate to me any misprints and conceptual or computational errors which for sure
are present in these notes.
These notes are dedicated to the memory of my dear friends Fabia and Paolo.
xvi
Ennio Gozzi
Trieste, September , 2012
Bibliography
[1] A. Bassi and G. C. Ghirardi, Phys. Rept. 379 (2003) 257.
[2] G. C. Ghirardi, R. Grassi and F. Benatti, Found. Of Phys. 25 (1995) 5.
A. Bassi, G. C. Ghirardi, Phys. Lett. A 275, (2000) 373.
[3] A. A. Abrikosov (jr), E. Gozzi, D. Mauro, Ann. of Phys. 317 (2005) 24.
Chapter 1
SCHR
2
2m
2
(x, t)
x
2
+ V (x)(x, t) = i
(x, t)
t
, (1.1)
where V (x) is the potential acting on the particle.
The main properties of this equation are
if (x, t) is a solution, then also all the functions of the form
(x, t) = e
i
(x, t) (1.2)
are solutions of the same equation, provided that the phase does not depend
on x, t. This follows from the fact that Eqn. (1.1) is linear in . The phase can
anyhow depend on such external parameters as a constant electric or magnetic
eld.
The set of all solutions of the form (1.2) is referred to as ray. We shall take
into account only one element for each ray, but it is important that, once you
have chosen the phase, you do not change the phase when you sum two states.
2 Chapter 1. Schr odinger equation. Wronskian theorem.
Example 1.1. Let us suppose to choose the phase for the state
1
and the
phase for another state
2
, thus the sum of the two states is
=
1
e
i
+
2
e
i
.
If now we change the phase of
1
for an amount and the phase of
2
for an
amount , the sum would become
=
1
e
i
e
i
+
2
e
i
e
i
Clearly, and
do not belong to the same ray, since in general
,= e
i
,
where is a phase. Thus, the sum of two states is a state while the sum of two
rays may not be a ray,
Thus, you must sum states and not rays.
1.2 Solution of the time-dependent Schr odinger equation
We now turn to the problem of how to solve the time-dependent Schr odinger equa-
tion (1.1). Let us seek a solution of the form
(x, t) = (x)(t) . (1.3)
Inserting (1.3) in Eqn. (1.1) and dividing by (x, t) we obtain
1
(x)
_
2
2m
d
2
(x)
dx
2
+ V (x)(x)
_
= i
1
(t)
d(t)
dt
. (1.4)
i
C
i
i
(x)
i
(t).
1.2. Solution of the time-dependent Schr odinger equation 3
The left-hand side of Eqn. (1.4) depends only on x while the right-hand side depends
only on t. The only way out is that both sides be equal to a constant, say E:
i
1
(t)
d(t)
dt
= E , (1.5)
1
(x)
_
2
2m
d
2
(x)
dx
2
+ V (x)(x)
_
= E . (1.6)
The solution of Eqn. (1.5) is (up to a constant)
(t) = e
iEt/
. (1.7)
and Eqn. (1.6) can be written as:
2
2m
d
2
(x)
dx
2
+ V (x)(x) = E(x) . (1.8)
Eqn. (1.8) is the so-called stationary (or time-independent) Schr odinger equation.
Few remarks:
In the above derivation, special care must be taken in dealing with the zeros of
(x) and (t), however it is possible to prove that they do not lead to problems.
Total, and not partial, derivatives have been used in writing Eqn. (1.4) since
and depend only on x and t respectively.
1.2.1 Properties of the solution
1. Since (x) is related to the probability density (x) of nding the particle in x
via the relation [Born interpretation of the (x)]
(x) = [(x)[
2
, (1.9)
we must require that be normalized:
_
+
(x) dx = 1 =
_
+
[(x)[
2
dx ,
thus (x) must be square-integrable: L
2
.
4 Chapter 1. Schr odinger equation. Wronskian theorem.
2. Moreover (x) must be nite for all x, otherwise if (x) were innite at some
point x
0
then also the probability density would become innite at that point
x
0
.
3. (x) must be single-valued because from it we will derive observable results;
4. (x) and its derivative will be chosen to be continuos at least away from sin-
gularities of the potential. This choice is compatible with the physical in-
terpretation of the wave-function and its derivative. Compatible means that
the continuity of the wave-function shall imply the continuity of the probability
density which is a reasonable physical condition. Analogously for the derivative
of the wave-function:its continuity is compatible with the continuity of the
probability current dened below. A more mathematical argument is the follow-
ing based on the fact that the derivability of a function implies the continuity
of the same. In the Schroedinger equation we have the second derivative of the
wave-function and, if this does exists, it implies that the rst derivative exists
and is a continuos function. This last in turn, as it is a derivative, it implies that
the wave-function itself is continuos. So this is the reason why the wave-function
and its derivative must be continuos, at least for regular potential.
Before concluding this section, we should recall the continuity equation which in the
one-dimensional case reads
(x, t)
t
=
J(x, t)
x
,
where the probability density (x, t) and the probability current J(x, t) are given re-
spectively by
(x, t) = [(x, t)[
2
= (x, t)
(x, t) ,
J(x, t) =
i
2m
_
(x, t)
(x, t)
x
+ (x, t)
(x, t)
x
_
.
1.3. Discrete energy levels 5
V (x)
x
E
x
Fig. 1.1. The potential V (x) used in the qualitative analysis of the stationary Schr odinger
equation and the energy E chosen for the analysis.
1.3 Discrete energy levels
We would like to understand in this section if the Schr odinger equation (1.8) can
have discrete energy solutions like several experimental evidences seem to indicate for
atoms and other systems.
1.3.1 Qualitative analysis
The stationary Schr odinger equation (1.8) can be rewritten as
d
2
(x)
dx
2
=
2m
2
[V (x) E] (x) . (1.10)
Since it is a second-order linear dierential equation, the values of (x) and its rst-
order derivative at some initial point x = a must be provided in order to solve the
equation, i.e.:
(a) =
a
,
d(x)
dx
x=a
=
a
.
Let us consider a potential V (x) like the one plotted in Fig. 1.1. Between x
and
x
< a < x
x=a
< 0 ,
6 Chapter 1. Schr odinger equation. Wronskian theorem.
(x)
x
x
a
Fig. 1.2. Three attempts to nd an acceptable solution to the time-independent
Schr odinger equation.
and (x) remains concave downwards up to x
the wave function goes into a region where V (x) E > 0, thus it becomes concave
upwards. In this case, the wave function diverges towards +, as shown in Fig. 1.2
(solid line).
Another possibility is to decrease the slope of the wave function after x
. In this
way intersects the x-axis and it becomes negative. Beyond the point of intersection
we have
d
2
(x)
dx
2
< 0 ,
thus the wave function becomes concave downwards and again it diverges, this time
towards , see Fig. 1.2 (dashed line).
A third possibility occurs when the slope of the wave function after x
is such that
decreases very slowly so that it remains concave upwards, it does not intersect the
x-axis and tends to zero asymptotically. This solution has a behavior at least at +
that allows it to be normalizable. See Fig. 1.2.
1.3. Discrete energy levels 7
1.3.2 Quantitative analysis and Wronskian theorem
Let us write the stationary Schr odinger equation
d
2
(x)
dx
2
=
2m
2
[V (x) E](x)
in the following way
d
2
z(x)
dx
2
+ [
E U(x)]z(x) = 0 , (1.11)
where
E
2m
2
E , U(x)
2m
2
V (x) , z(x) (x) .
Let us take U(x) continuous and bounded from below.
Given two arbitrary solutions z
1
(x) and z
2
(x) of Eqn. (1.11) corresponding to two
dierent energies, we can dene the following object, which is called the Wronskian:
W(z
1
, z
2
) z
1
z
2
z
2
z
1
, (1.12)
where z
1
= dz
1
/ dx and z
2
= dz
2
/ dx. If, at a certain point x = a, the Wron-
skian (1.12) is zero, then
z
1
(a)z
2
(a) z
2
(a)z
1
(a) = 0 ,
that is,
z
2
(a)
z
2
(a)
=
z
1
(a)
z
1
(a)
,
which means that the logarithmic derivatives
1
(x) + F
1
(x)z
1
(x) = 0 , (1.13a)
z
2
(x) + F
2
(x)z
2
(x) = 0 , (1.13b)
b
a
=
_
b
a
[F
1
(x) F
2
(x)] z
1
(x)z
2
(x) dx . (1.14)
Proof. If we multiply Eqn. (1.13a) by z
2
and Eqn. (1.13b) by z
1
and take the dierence
of the two, we obtain
z
2
z
1
z
1
z
2
. .
W
(z
1
,z
2
)
+(F
1
F
2
)z
1
z
2
= 0 . (1.15)
The rst term is, modulo a sign, nothing but the derivative of the Wronskian W(z
1
, z
2
)
with respect to x, as you can see by direct dierentiation of Eqn. (1.12):
W
(z
1
, z
2
) = z
1
z
2
z
2
z
1
.
Integrating Eqn. (1.15) in x between a and b yields just Eqn. (1.14).
Corollary 1.1.1. If Eqs. (1.13) correspond to the stationary Schr odinger equa-
tion (1.11) for two dierent energies
E
1
and
E
2
respectively, then F
1
(x) =
E
1
U(x),
F
2
(x) =
E
2
U(x) and Eqn. (1.14) becomes
W(z
1
, z
2
)
b
a
= (
E
1
E
2
)
_
b
a
z
1
(x)z
2
(x) dx . (1.16)
Corollary 1.1.2. If the two solutions z
1
and z
2
are associated with the same energy
E
1
=
E
2
, then we get from Eqn. (1.16)
W(z
1
, z
2
)
x=a
= W(z
1
, z
2
)
x=b
,
so
W(z
1
, z
2
) = const ,
i.e., the Wronskian is independent of x since the interval (a, b) is arbitrary.
Corollary 1.1.3. If z(x,
E) is a solution of Eqn. (1.11) and if its logarithmic deriva-
tive, i.e.:
F(x,
E) =
z
(x,
E)
z(x,
E)
1.3. Discrete energy levels 9
at the point x = a has a xed value f
a
which is independent of
E, i.e.:
F(a,
E) = f
a
, for all
E ,
then F(x,
E) is a monotonic function of
E, increasing if x < a and decreasing if x > a.
This is because the derivative of F with respect to
E is given by (see proof below)
F(x,
E)
E
=
1
z
2
(x,
E)
_
x
a
z
2
(,
E) d . (1.17)
and from Eqn. (1.17) we see that the derivative is negative if x > a and positive if
x < a.
Proof. Let us consider the following boundary conditions:
z(a,
E) = z
a
, z
(a,
E) = z
a
,
and replace
E with
E +
E. There will be two nearby solutions z and z + z which
correspond to the values
E and
E +
E and satisfy the same boundary conditions.
According to Corollary 1.1.1, at rst order in we obtain
W(z, z + z)
b
a
=
E
_
b
a
z
2
(x) dx . (1.18)
At the point x = a, the Wronskian is zero: W(z, z+z)[
x=a
= 0. This is so because the
two solutions satisfy the same boundary conditions, since z
a
and z
a
are independent
of
E. For all other values of x in the interval (a, b) we have
W(z, z + z) = zz
z = z
2
_
z
z
_
= z
2
F .
Thus, Eqn. (1.18) becomes
z
2
F
x=b
=
E
_
b
a
z
2
(x) dx ,
that is,
F
x=b
=
1
z
2
(b)
_
b
a
z
2
(x) dx .
from which it follows that F is a monotonic function of
E, increasing or decreasing
for b < a or b > a respectively.
10 Chapter 1. Schr odinger equation. Wronskian theorem.
U(x)
x
U
(+)
U
()
Fig. 1.3. Potential U(x) and its asymptotic behavior at .
Let us now take a potential U(x) such as the one plotted in Fig. 1.3, with U
()
>
U
(+)
.
First case. We consider
E > U
()
. In this case,
E U(x) is positive at the two
extrema and constant, Eqn. (1.11) becomes at +
z
+ [
E U
(+)
]
. .
k
z = 0 ,
that is,
z
= kz , k
E U
(+)
> 0 ,
and the asymptotic behavior of z at + is
z e
i
kx
for x +. (1.19a)
At , Eqn. (1.11) becomes
z
+ [
E U
()
]
. .
k
z = 0 ,
that is,
z
kz ,
k
E U
()
> 0 ,
and the solution is
z e
i
kx
for x . (1.19b)
1.3. Discrete energy levels 11
Solutions (1.19) are bounded and oscillating and they can be tted continuously with
each other. So there is no constraint on the energy. Every value of energy is allowed
and the spectrum is continuous.
Remark. Oscillating solutions are allowed, even if strictly speaking they are not
truly normalizable. They are normalizable in a generalized sense (theory of distribu-
tions). In fact, the scalar product of two plane waves e
ikx
and e
ik
x
is
_
+
e
ikx
e
ik
x
dx (k k
) .
On the RHS we do not have the usual Kronecker delta but the Dirac delta (a distribu-
tion) of which we will give details later on. The space of square-integrable functions L
2
must be enlarged in order to include also those functions which are square-integrable
in the distributional sense. If we call this space
L
2
, it is possible to prove that L
2
is
dense in
L
2
, that is to say, each element in
L
2
has a nearby element in L
2
.
Second case. Let us now consider U
()
>
E > U
(+)
. In this case,
EU is negative
for x so Eqn. (1.11) becomes
z
= kz , k
E U
()
< 0 ,
and the asymptotic behavior of z at is given by
z e
|k|x
for x , (1.20a)
which tends to zero for x and therefore can be accepted. At +, in Eqn. (1.11)
we have instead
z
= k
z , k
E U
(+)
> 0 ,
and the solution is
z e
i
x
for x +, (1.20b)
which is an oscillating function. Again it is possible to match the two solutions (1.20)
and therefore also in this case there are no constrains on the energy and the spectrum
is continuous.
12 Chapter 1. Schr odinger equation. Wronskian theorem.
Third case. We consider U
()
> U
(+)
>
E. Eqn. (1.11) becomes at both extrema
z
= k
z , k
E U
< 0 ,
and the asymptotic solutions are
z e
|k
|x
for x , (1.21)
In this case, the solutions are bounded and damped at both extrema. Now we have to
t them in order to guarantee the continuity of z(x) and of its rst-order derivative.
Let us consider the logarithmic derivatives of the two solutions, and indicate them
with F
+
(x,
E) and F
(x,
E). They will be equal to each other at a certain point
x = a. This is so because both z and z
and z
+
converge to the same solution z, both z
and z
+
must satisfy the same boundary conditions in a. Now F
+
is valid for x > a and
according to Corollary 1.1.3 it is a monotonic decreasing function of
E, while F
is
valid for x < a and is a monotonic increasing function of
E. Now F
+
and F
must be
equal in x = a and this can happen only for some particular values
E
n
of
E. This is
the proof that in the present case only certain values of
E are allowed. These values
are xed and isolated and the spectrum is discrete.
1.4 The Dirac delta
In this section we will provide some introductory details on the Dirac delta that we
mentioned before. Let us recall the main properties of the Diracs -function.
A function is a map between, for example, the space of real numbers R and R, or
between the space of complex numbers C and C, or between C and R. A functional
is a map between a space of functions and the space of real or complex numbers:
T[f(x)] R .
1.4. The Dirac delta 13
Example 1.2 (functionals). The action
S =
_
t
1
t
0
/(q(t), q(t), t) dt
associates to a function q(t) the real number S obtained by evaluating the
integral above.
The Diracs delta function (x) is dened via the relation
_
+
(x x
0
)f(x) dx = f(x
0
) , (1.22)
that is, the Dirac delta associates to the function f(x) the number f(x
0
).
Eqn. (1.22) suggests that it is possible to represent the Dirac delta distribution
as the limit of a suitable sequence of functions. For example, if we choose x
0
= 0 we
have that the action of the Dirac delta in Eqn. (1.22), namely
_
+
(x)f(x) dx = f(0) ,
is the same as the one achieved by the following operation:
(x) lim
l
l
e
l
2
x
2
,
which must be understood as performed under an integral:
_
+
(x)f(x) dx = lim
l
_
+
e
l
2
x
2
f(x) dx .
(Prove the latter equation as an homework.)
Note that the set of all functions f
l
(x) = l/
exp(l
2
x
2
) as we send l to is
innite. These functions are Gaussians and the area below them is equal to one. (See
Fig. 1.4.)
14 Chapter 1. Schr odinger equation. Wronskian theorem.
0
1
2
3
4
5
f
l
(
x
)
=
l
l
2
x
2
3 2 1 0 1 2 3
x
f
1
(x)
f
2
(x)
f
5
(x)
f
10
(x)
0
1
2
3
f
l
(
x
)
=
l
s
i
n
(
l
x
)
l
x
3 2 1 0 1 2 3
x
f
1
(x)
f
2
(x)
f
5
(x)
f
10
(x)
Fig. 1.4. Representations of the Dirac delta (x).
Other representations of the Dirac delta are given by
1
l
sin
2
lx
x
2
,
1
l
1
x
2
+
1
l
2
,
1
sin lx
x
.
(See Fig. 1.4.)
The Dirac delta (xy) is somehow the generalization to the continuous case of
the Kronecker delta
i
j
, the latter being dened for discrete indexes i, j as
i
j
=
_
_
1 if i = j
0 if i ,= j
By denition, Kronecker delta acts on a vector f
i
as
f
i
=
i
j
f
j
.
Now, if the indexes i, j become continuous, the various f
j
become functions f
x
or
f(x) and the sum becomes an integral over x, that is
f(y) =
_
(x y)f(x) dx .
1.5. Harmonic oscillator and Hermite polynomials 15
1.5 Harmonic oscillator and Hermite polynomials
Find the spectrum and the wave functions of the stationary Schr odinger equation for
an harmonic oscillator with potential V (x) =
1
2
kx
2
.
The stationary Schr odinger equation is
2
2m
d
2
(x)
dx
2
+
1
2
kx
2
(x) = E(x) .
First, let us change variables :
x ,
_
mk
2
_
1/4
.
The Schr odinger equation in the new variable reads
d
2
(x())
d
2
+ (
2
) = 0 , (1.23)
where
2E
_
m
k
_
1/2
=
2E
c
,
c
=
_
k
m
_
1/2
. (1.24)
It is easy to prove that for + the function
() =
n
e
2
/2
satises Eqn. (1.23) with = 0. To prove this fact, let us calculate the rst and
second order derivatives of ():
d()
d
= n
n1
e
2
/2
n+1
e
2
/2
,
d
2
()
d
2
= [n(n 1)
n2
n
n
(n + 1)
n
+
n+2
] e
2
/2
.
For , the second-order derivative behaves like
d
2
()
d
2
n+2
e
2
/2
,
and for = 0 Eqn. (1.23) is asymptotically fulllled, as you can verify by direct
substition:
d
2
()
d
2
2
()
n+2
e
2
/2
n
e
2
/2
= 0 .
16 Chapter 1. Schr odinger equation. Wronskian theorem.
Now, we try to nd a complete (i.e., not only asymptotic) solution of Eqn. (1.23)
of the form
() = H() e
2
/2
, (1.25)
where H() is a polynomial in . Inserting Eqn. (1.25) into Eqn. (1.23) yields the
following equation for the polynomial H():
d
2
H()
d
2
2
dH()
d
+ ( 1)H() = 0 . (1.26)
Now, let us make the following ansatz for H():
H() =
s
_
a
0
+ a
1
+ a
2
2
+ . . .
, (1.27)
with a
0
,= 0 and s 0. By substituting into Eqn. (1.26) and by equating to zero all
the coecients of the various powers of , we get
s(s 1)a
0
= 0 ,
(s + 1)sa
1
= 0 ,
(s + 2)(s + 1)a
2
(2s + 1 )a
0
= 0 ,
(s + 3)(s + 2)a
3
(2s + 3 )a
1
= 0 ,
.
.
.
(s + + 2)(s + + 1)a
+2
(2s + 2 + 1 )a
= 0 ,
(1.28)
where is an integer number. Since a
0
,= 0, we get from the rst equation that
s = 0 or s = 1 .
The second equation gives
s = 0 or a
1
= 0 ,
or both of them. The third equation gives a
2
as a function of a
0
and the fourth
equation gives a
3
as a function of a
1
. In general, from = 3, the th equation gives
a
1
as a function of a
3
.
1.5. Harmonic oscillator and Hermite polynomials 17
According to problems presented at the end of this chapter, the wave functions
(x) for one dimensional system cannot be degenerate for bound states and moreover
since the potential is even V (x) = V (x) they will be either even or odd, that
is,
(x) = (x) .
Now, let us consider separately the two cases s = 0 and s = 1 that solve the rst
equation in (1.28)
s = 0 : The rst equation in (1.28) is satised and also the second is fullled without
having to choose a
1
= 0. However, in this way the parity would not be dened,
since we would have both the zeroth power with a
0
and the rst power with a
1
.
Instead, if we choose a
1
= 0, the second equation in (1.28) is still satised and
we get the even polynomials.
s = 1 : In this case, we get the odd polynomials since
s
[a
0
+ a
1
+ . . .] = [a
0
+ a
1
+ . . .] ,
and a
1
must be zero in order to satisfy the second equation in (1.28).
In (1.28) there will be a nite or an innite number of terms depending on how we
choose s and . If the series did not stop, then the ratio of the coecients a
+2
/a
2s + 2 + 1
(s + + 2)(s + + 1)
2
, (1.29)
but these coecients are just the same of the expansion of
n
e
2
2
, (1.30)
this means that () = H
n
() e
2
/2
would become
n
e
2
2
2
/2
=
n
e
3/2
2
which is
not normalizable.
18 Chapter 1. Schr odinger equation. Wronskian theorem.
Let us check that the coecients of the series expansion of (1.30) goes as in
Eqn. (1.29). The th term of the series expansion of (1.30) is
n
(
2
2)
m
1
m!
and the ( 1)th is
n
(
2
2)
m1
1
(m1)!
,
thus the coecients are 2
/! and 2
1
/( 1)! respectively and their ratio is 2/ as
in Eqn. (1.29).
Thus, in order to have normalizable wave functions the series (1.28) must stop, that
is we must have polynomials with a nite number of terms. From Eqs. (1.28) we see
that this happens if
= 2s + 2 + 1 .
In fact, in this way a
+2
is zero and the following coecients obtained from a
+2
will
be zero too. Now, let us call s + = n, which is an integer number. The above
relation becomes
= 2n + 1 .
By replacing with E via Eqn. (1.24) we get
E
n
=
_
n +
1
2
_
c
. (1.31)
These are the discrete energy levels. The lowest energy occurs for n = 0 and it is
given by
E
0
=
1
2
c
,
which is called the zero-point energy.
Since is nite, H
n
will be polynomials and not series. In particular,
H
0
() = 1 ,
H
1
() = 2 ,
H
2
() = 4
2
2 , . . .
1.6. Schr odinger equation for multidimensional problems 19
These polynomials satisfy the equation
H
n
() 2H
n
() + 2nH
n
() = 0 .
and they are referred to as Hermite polynomials.
We want to stress the fact that also in the case of the harmonic oscillator the
discreteness of the energy spectrum comes from the normalization requirement of the
wave functions.
1.6 Schr odinger equation for multidimensional problems
Until now, we have dealt exclusively with one-dimensional problems. Now, we turn
to the two- and three-dimensional ones. We will focus on those problems where
the potential is separable, which in the three-dimensional case and using Cartesian
coordinates means
V (x, y, z) = V
1
(x) + V
2
(y) + V
3
(z) . (1.32)
It may happen that a potential is not separable in Cartesian coordinates, but it
is separable in another set of coordinates, for example in cylindrical or spherical
coordinates. This means that using for example spherical coordinates, we have
V (r, , ) = V
1
(r) + V
2
() + V
3
() .
In which set of coordinates the potential is separable depends on the symmetries of
the system. For the time being, we will consider potentials which are separable in
Cartesian coordinates. In this case, the stationary Schr odinger equation reads
_
2
2m
_
2
x
2
+
2
y
2
+
2
z
2
_
+ V
1
(x) + V
2
(y) + V
3
(z)
_
(x, y, z) = E(x, y, z) .
We look for solutions of the form
(x, y, z) =
1
(x)
2
(y)
3
(z) .
i
(x)
j
(y)
k
(z).
20 Chapter 1. Schr odinger equation. Wronskian theorem.
Inserting this into the Schr odinger equation we obtain
2
2m
d
2
1
(x)
dx
2
2
(y)
3
(z) + V
1
(x)
1
(x)
2
(y)
3
(z) +
2
2m
d
2
2
(y)
dy
2
1
(x)
3
(z) + V
2
(y)
1
(x)
2
(y)
3
(z) +
2
2m
d
2
3
(z)
dz
2
1
(x)
2
(y) + V
3
(z)
1
(x)
2
(y)
3
(z) = E
1
(x)
2
(y)
3
(z) .
by dividing both sides by
1
(x)
2
(y)
3
(z) we get
_
2
2m
d
2
1
(x)
dx
2
+ V
1
(x)
1
(x)
_
1
1
(x)
+
_
2
2m
d
2
2
(y)
dy
2
+ V
2
(y)
2
(y)
_
1
2
(y)
+
_
2
2m
d
2
3
(z)
dz
2
+ V
3
(z)
3
(z)
_
1
3
(z)
= E .
Note that each of the three terms on the left-hand side is a function of a dierent
variable: the rst is a function of x, the second of y and the third of z. In order to
have the sum equal to a constant, namely E, each term must be a constant:
_
2
2m
d
2
1
(x)
dx
2
+ V
1
(x)
1
(x)
_
1
1
(x)
= E
1
,
_
2
2m
d
2
2
(y)
dy
2
+ V
2
(y)
2
(y)
_
1
2
(y)
= E
2
,
_
2
2m
d
2
3
(z)
dz
2
+ V
3
(z)
3
(z)
_
1
3
(z)
= E
3
,
with E
1
+ E
2
+ E
3
= E. The initial problem of solving the three-dimensional
Schr odinger equation is thus reduced to solving the following three Schr odinger equa-
tions in one dimension:
2
2m
d
2
1
(x)
dx
2
+ V
1
(x)
1
(x) = E
1
1
(x) ,
2
2m
d
2
2
(y)
dy
2
+ V
2
(y)
2
(y) = E
2
2
(y) ,
2
2m
d
2
3
(z)
dz
2
+ V
3
(z)
3
(z) = E
3
3
(z) .
1.6. Schr odinger equation for multidimensional problems 21
The three solutions are associated with energy eigenvalues E
1
, E
2
and E
3
such that
E
1
+ E
2
+ E
3
= E .
It may happen that by adding three dierent eigenvalues E
1
, E
2
and E
3
we obtain
the same energy E:
E
1
+ E
2
+ E
3
= E .
The three wave functions corresponding to the eigenvalues E
1
, E
2
and E
3
which
we shall denote with
1
(x),
2
(y) and
3
(z) in general will be dierent from those
associated with E
1
, E
2
and E
3
say,
1
(x),
2
(y) and
3
(z). This means that the
global wave functions
E
(x, y, z) =
1
(x)
2
(y)
3
(z)
and
E
(x, y, z) =
1
(x)
2
(y)
3
(z)
have the same energy E but they are two dierent functions. In this case, we speak
of degeneracy, i.e., there are dierent wave functions associated to the same energy.
Now, we want to give an example of a potential which is separable in spherical
coordinates but not in Cartesian coordinates. We dene the spherical coordinates in
the usual way:
x = r sin cos , 0 < r < , (1.33a)
y = r sin sin , 0 , (1.33b)
z = r cos , 0 < 2 . (1.33c)
Let us consider the following potential
V (r, , ) = kr
2
+ l
2
tan + s
2
sin
2
. (1.34)
Of course, it is separable in spherical coordinates, since it can be written in the form
V (r, , ) = V
1
(r) + V
2
() + V
3
() ,
22 Chapter 1. Schr odinger equation. Wronskian theorem.
where
V
1
(r) = kr
2
, V
2
() = l
2
tan , V
3
() = s
2
sin
2
.
By inverting
V (x, y, z) = k(x
2
+ y
2
+ z
2
) + l
2
_
x
2
+ y
2
z
+ s
2
y
2
x
2
+ y
2
,
which is not separable as a sum
V
1
(x) +
V
2
(y) +
V
3
(x).
1.7 Central potentials
With the word central potentials we mean those potentials which, once they are
written in polar coordinates, depend only on r, that is
V (x, y, z) = U(r) .
It is not dicult to prove that the Laplacian in spherical coordinates reads
2
x
2
+
2
y
2
+
2
z
2
=
1
r
2
r
_
r
2
r
_
+
1
r
2
sin
_
sin
_
+
1
r
2
sin
2
2
.
It should be noted that the transformation from the Cartesian coordinates to the
spherical coordinates is not one-to-one along the z-axis (in fact, can be both 0 or
and can take any value). The Laplacian reects this fact via the presence of singu-
larities in = 0, = and r = 0. Thus, the solution of the stationary Schr odinger
equation can have spurious singularities in those points due to the singularities of the
Laplacian in spherical coordinates.
Using the expression for the Laplacian in spherical coordinates, the stationary
Schr odinger equation for a central potential U(r) becomes
2
2m
_
1
r
2
r
_
r
2
r
_
+
1
r
2
sin
_
sin
_
+
1
r
2
1
sin
2
2
_
(r, , ) + U(r)
(r, , ) = E
(r, , ) . (1.35)
(r, , ) = R(r)Y (, ) .
Inserting this into Eqn. (1.35) and multiplying by r
2
/(RY ), we get:
1
R(r)
d
dr
_
r
2
dR(r)
dr
_
+
2m
2
[E U(r)]r
2
=
_
1
Y sin
_
sin
Y
_
+
1
Y sin
2
2
Y
2
_
.
Note that the left-hand side depends only on r while the right-hand side is a function
only of and . For the two sides to be equal to each other it must be that they
both equal a constant, which we indicate with . This leads to:
1
R
d
dr
_
r
2
dR
dr
_
+
2m
2
[E U(r)]r
2
= , (1.36a)
1
Y sin
_
sin
Y
_
+
1
Y sin
2
2
Y
2
= . (1.36b)
Eqn. (1.36b) is basically related to the quantization of angular momentum and it
will be treated in chapter 5. Instead, the solution of Eqn. (1.36a) can be already
worked out at this point. Let us rewrite R(r) = (r)/r, where (r) is a new function.
Inserting this parametrization of R(r) into Eqn. (1.36a) we get
2
2m
d
2
(r)
dr
2
+
_
U(r) +
2
2mr
2
_
(r) = E(r) . (1.37)
This is nothing but the Schr odinger equation for a particle moving in a eective
potential given by the sum of the original potential U(r) plus the centrifugal term
2
2mr
2
. We shall see in Chapter 9 that the centrifugal potential is linked to the angular
momentum and it is non-zero only when the angular momentum does not vanish. In
fact, it is possible to prove that this term represents the rotational kinetic energy.
In the case of the hydrogen atom, U(r) is given by
U(r) =
e
2
r
,
24 Chapter 1. Schr odinger equation. Wronskian theorem.
or, if we consider one electron spinning around a nucleus of charge Ze, it is
U(r) =
Ze
2
r
.
The radial equation (1.37) becomes
2
2m
1
r
2
d
dr
_
r
2
dR
dr
_
Ze
2
r
R +
2
2mr
2
R = ER ,
where m is not the mass of electron, but its reduced mass
m =
m
e
m
N
m
e
+ m
N
,
where m
e
is the actual mass of the electron and m
N
is the mass of the nucleus. Since
m
e
m
N
, we have m m
e
.
Let us introduce the following new variables
r ,
2
8m[E[
2
, =
Ze
2
_
m
2 [E[
_
1/2
.
The radial equation above is then transformed into the following one:
1
2
d
d
_
2
dR(r())
d
_
+
_
1
4
2
_
R = 0 .
This is a dierential equation which is well-known, and we shall solve it in chapter 9
after having obtained the value of in chapter 5 via the quantization of the angular
momentum.
1.8 Problems and Solutions
Problem 1.1. Show that the following wave function
(x, t) = Aexp
_
Cm
2
x
2
i
2
_
C
m
t
_
,
is a solution of the Schr odinger equation (1.1) for an harmonic oscillator with
potential
V (x) =
1
2
Cx
2
.
1.8. Problems and Solutions 25
Calculate the probability density associated with this wave function and compare
the result with the classical probability of nding the particle in a interval dx.
Solution. First, we must evaluate the derivatives of :
(x, t)
t
=
i
2
_
C
m
(x, t) ,
(x, t)
x
=
Cm
x(x, t) ,
2
(x, t)
x
2
=
Cm
(x, t) +
Cm
2
x
2
(x, t) .
Replacing these expressions in the Schr odinger equation (1.1) we get
2
2m
Cm
(x, t)
2
2m
Cm
2
x
2
(x, t) +
1
2
Cx
2
(x, t) = i
_
i
2
_
_
C
m
(x, t) ,
from which it follows
2
_
C
m
(x, t)
1
2
Cx
2
(x, t) +
1
2
Cx
2
(x, t) =
2
_
C
m
(x, t) ,
which is identically satised.
Accordingly to the Born interpretation of the wave function, the quantum probability
density (x, t) of nding the particle in x at the time t is given by
(x, t) = [(x, t)[
2
= A
2
exp
_
Cm
x
2
_
,
which does not depend on t; this is a common feature of all the wave functions corresponding
to xed values of the energy. (x) is plotted in Fig. 1.5(b).
Let us now evaluate the classical probability P(x) dx of nding the particle in a small
interval dx centered at x. This probability will be proportional to the time taken by the
particle to cross dx, thus it is inversely proportional to the velocity of the particle in x:
P(x) =
B
2
v(x)
,
where B is a constant. At a given energy E we have
E =
1
2
mv
2
+
1
2
Cx
2
,
26 Chapter 1. Schr odinger equation. Wronskian theorem.
V (x)
x
2E
C
+
2E
C
0
E
(a)
|(x)|
2
x
2E
C
+
2E
C
0
(b)
P(x)
x
0
2E
C
+
2E
C
(c)
Fig. 1.5. Comparison between (c) classical and (b) quantum probability densities for an
harmonic oscillator whose potential is shown in (a).
1.8. Problems and Solutions 27
thus
v(x) =
_
2
m
_
E
1
2
Cx
2
,
and
P(x) =
B
2
_
2
m
_
E
1
2
Cx
2
.
P(x) is plotted in Fig. 1.5(c).
Note that the probability of nding the particle at the origin x = 0 is minimum in
the classical case, while it is maximum in the quantum case. The classical particle has its
maximum velocity at x = 0 and therefore goes rapidly through the origin and the probability
of nding the particle there is minimum. In the quantum case instead the situation is just
the opposite and the particle is more likely to be found at the origin. Note however that
there is a non-null quantum probability of nding the particle also outside the extrema of
the oscillation range. The dierence between the classical and quantum case decreases at
higher energies. At low energies, quantum eects becomes dominant.
Problem 1.2. Show that in a one-dimensional problem the energy spectrum of
bound states is always non-degenerate. By degeneracy we mean that there are
two or more linearly independent solutions associated to the same energy value.
Solution. Let us prove this fact ab absurdo and suppose the opposite is true. Let
1
(x)
and
2
(x) be two linearly independent solutions corresponding to the same energy E:
1
(x) +
2m
2
[E V (x)]
1
(x) = 0 ,
2
(x) +
2m
2
[E V (x)]
2
(x) = 0 .
Dividing the former by
1
and the latter by
2
(and staying away from their zeros) we get
1
(x)
1
(x)
=
2m
2
[E +V (x)] ,
2
(x)
2
(x)
=
2m
2
[E +V (x)] ,
which implies that
1
(x)
1
(x)
=
2
(x)
2
(x)
=
2m
2
[E +V (x)] ,
28 Chapter 1. Schr odinger equation. Wronskian theorem.
and from the rst equality it follows that
1
= (
2
)
1
)
= 0 .
By integrating the previous equation we get
1
= const.
This equation must hold for all x and since the wave function of a bound state must be
zero at , the constant appearing on the right-hand side of this equation must also be zero,
which implies
1
=
2
.
By integrating, we get
log
1
= log
2
+ log C ,
which leads to
1
= C
2
.
Therefore, the two states are linearly dependent. This in contradiction with the hypothesis
and it implies that there is no degeneracy.
Problem 1.3. Show that the rst derivatives of the stationary wave functions
are continuous even if the potential V (x) has a nite discontinuity.
Solution. Suppose the discontinuity is at the point x
0
. Let us consider an interval (x
0
, x
0
+ ) and replace V (x) in such interval with a potential V
1
(x) linearly interpolating
between V (x
0
) and V (x
0
+ ). In this way, we obtain a continuous potential whose
solutions satisfy the stationary Schr odinger equation
d
2
1
(x)
dx
2
+
2m
2
[E V
1
(x)]
1
(x) = 0 .
1.8. Problems and Solutions 29
By integrating this equation in the interval (x
0
, x
0
+) we get
_
x
0
+
x
0
d
2
1
(x)
dx
2
dx =
2m
2
_
x
0
+
x
0
[V
1
(x) E]
1
(x) dx ,
from which it follows
d
1
(x)
dx
x=x
0
+
d
1
(x)
dx
x=x
0
=
2m
2
_
x
0
+
x
0
[V
1
(x) E]
1
(x) dx . (1.38)
In the limit 0, we have V
1
(x) V (x) and
1
(x) (x). Since the integrand on the
right-hand side is nite, it remains nite also in the limit 0. Let us recall now the
mean-value theorem, which states that
_
b
a
f(x) dx =
f[b a] ,
where
f is the mean value of the function f(x) over the interval (a, b). If
f remains nite
and (b a) 0 then also the integral of f(x) must vanish. This is exactly what happens
in the case of Eqn. (1.38), since the integrand is nite. Thus,
lim
0
_
d(x)
dx
x=x
0
+
d(x)
dx
x=x
0
_
= 0 ,
that is,
d(x)
dx
x=x
0
0
=
d(x)
dx
x=x
0
+0
,
and therefore the derivatives are continuous.
Remark. If the potential V (x) of the previous problem had an discontinuity, then
it would have been impossible to apply the mean value theorem. In such cases, the
derivative
could have a nite gap. We will see an example in the next problems.
Problem 1.4. The potential energy of a system is
U(x) =
U(x) + (x x
0
) ,
where
U(x) is a bounded function. Which is the behavior of the solution of the
Schr odinger equation (x) and its derivative near the point x
0
?
30 Chapter 1. Schr odinger equation. Wronskian theorem.
Solution. The Schr odinger equation reads
2
2m
d
2
(x)
dx
2
+ [
U(x) +(x x
0
)](x) = E(x) . (1.39)
(x) must be continuous because of its probabilistic meaning. Let us investigate if also the
derivative remain continuous. By integrating Eqn. (1.39) between x
0
and x
0
+ with
> 0 and by sending 0 we obtain
2
2m
_
x
0
+
x
0
d
2
(x)
dx
2
dx =
_
x
0
+
x
0
U(x) +(x x
0
) E](x) dx ,
from which it follows
2
2m
_
d(x)
dx
x=x
0
+
d(x)
dx
x=x
0
_
= (x
0
)
_
x
0
+
x
0
U(x) E](x) dx .
If (x) is continuous and
U(x) is bounded we can use the mean-value theorem, so in the
limit 0 the integral on the right-hand side of the previous equation vanishes and we get
lim
0
2
2m
_
d(x)
dx
x=x
0
+
d(x)
dx
x=x
0
_
= (x
0
) ,
that is,
d(x)
dx
x=x
0
+0
d(x)
dx
x=x
0
0
=
2m
2
(x
0
) . (1.40)
This means that the derivative is discontinuous at the point x
0
. Remember the result proved
in Problem (1.3): for potentials with a nite gap the derivative of (x) must be continuous.
In this case however the Dirac delta has in x
0
an innite jump, so the result of Problem (1.3)
cannot hold and in fact the derivative is discontinuous.
Problem 1.5. Find the energy levels and the wave functions of the bound states
of a particle in a potential U(x) = (x), with > 0.
Solution. The Schr odinger equation is
2
2m
d
2
(x)
dx
2
(x)(x) = E(x) .
1.8. Problems and Solutions 31
For x ,= 0, this is nothing but the usual equation of a free particle:
2
2m
d
2
(x)
dx
2
= E(x) .
The solutions of the equation above are easily found to be exponentials, however here we
consider bound states so we must require that the solutions go to zero at , thus the only
allowed solutions take the form
(x) =
_
_
Ae
kx
if x > 0
Be
kx
if x < 0
,
where k is related to E. The relation between k and E can be found by inserting the solution
(x) into the Schr odinger equation of the free particle; we get
2
2m
A(k
2
) e
kx
= EAe
kx
,
from which it follows
E =
2
k
2
2m
.
E is negative (bound state) and k has not yet been determined. Now, let us determine A
and B. By using Eqn. (1.40), the jump in the derivative of (x) is given by
d(x)
dx
x=x
0
+0
d(x)
dx
x=x
0
0
=
2m()
2
(x
0
) .
[Note that in problem 1.4 we have (x x
0
) while here we have (x).] Hence,
A(k) e
k0
kBe
k0
=
2m
2
A .
In the right-side of the latter equation we could have used B as well. In fact, because of the
continuity condition, we would get A = B. Therefore,
A(k) kA =
2m
2
A ,
that is,
k =
m
2
.
32 Chapter 1. Schr odinger equation. Wronskian theorem.
This implies that there is only one value of k which corresponds to a bound state of our
problem, its energy is given by
E =
2
2m
k
2
=
2
2m
m
2
4
=
m
2
2
2
.
A is found by requiring (x) to be normalized:
_
[(x)[
2
dx = 1. The normalized wave
function corresponding to the only bound state is given by (x) =
k e
k|x|
. (Prove the
latter result as an homework.)
Problem 1.6. Prove that under an even potential V (x) = V (x) the eigen-
functions solution of the stationary Schr odinger equation for bound states are
even or odd, but in any case they always have well-dened parity.
Solution. Let us start with
_
2
2m
d
2
dx
2
+V (x)
_
(x) = E(x) .
Now, we prove that also (x) is a solution of the same equation with the same value of
energy E. Replacing x with x in the equation above yields
_
2
2m
d
2
dx
2
+V (x)
_
(x) = E(x) ,
and exploiting the fact that V (x) = V (x) we get
_
2
2m
d
2
dx
2
+V (x)
_
(x) = E(x) ,
This proves that both (x) and (x) are solutions of the Schr odinger equation with the
same energy E.
Now, suppose (x) and (x) are bound states belonging to the discrete spectrum. From
Problem (1.2) we already know that in this case there cannot be degeneracy, hence (x)
and (x) must be linearly dependent:
(x) = C(x) .
1.8. Problems and Solutions 33
Normalizing
_
+
[(x)[
2
dx = 1 ,
we get
[C[
2
_
+
[(x)[
2
dx = 1 ,
and thus [C[
2
= 1. From this it follows that (apart from a phase) C = 1 and therefore
(x) = (x), that is the wave functions are even or odd. This result holds for the
discrete spectrum. For the continuous part of the spectrum there can be degeneracy and so
it is not always true that (x) = C(x) and the energy eigenfunctions may have parity
that it is not well-dened.
Problem 1.7. A two-dimensional harmonic oscillator has Hamiltonian given by
H =
p
2
x
2M
+
1
2
kx
2
+
p
2
y
2M
+
1
2
ky
2
. (1.41)
Study the spectrum and calculate the degeneracy.
Solution. The two-dimensional stationary Schr odinger equation reads in this case
_
2
2M
2
x
2
2
2M
2
y
2
+
1
2
kx
2
+
1
2
ky
2
_
(x, y) = E(x, y) . (1.42)
Since the potential is separable, we can look for solutions of the form
(x, y) =
1
(x)
2
(y) .
So Eqn. (1.42) becomes equivalent to the following two equations:
_
2
2M
d
2
dx
2
+
1
2
kx
2
_
1
(x) = E
1
1
(x) ,
_
2
2M
d
2
dy
2
+
1
2
ky
2
_
2
(y) = E
2
2
(y) ,
34 Chapter 1. Schr odinger equation. Wronskian theorem.
with E
1
+ E
2
= E. E
1
and E
2
are given by the solution of the one-dimensional harmonic
oscillator, that is,
E
1
=
_
n +
1
2
_
c
, E
2
=
_
m+
1
2
_
c
,
where
c
= (k/M)
1/2
and n, m are (non-negative) integer numbers labeling the eigenvalues.
The total energy is
E = E
1
+E
2
= (n +m+ 1)
c
= (N + 1)
c
,
where N = n + m is any non-negative integer number. The corresponding wave functions
are
N
(x, y) =
1,n
(x)
2,m
(y) ,
where we have denoted with
1,n
(x) the eigenfunctions corresponding to the eigenvalue
E
1,n
= (n + 1/2)
c
and with
2,m
(y) the eigenfunctions with eigenvalue E
2,m
= (m +
1/2)
c
, N = n +m. It is clear that there are many dierent values of n and m such that
n + m = N with N xed and whose corresponding wave functions have the same energy.
For example, we can choose n
= n 1 and m
+m
,
and in the same way we can choose n
= n 2 and m
= m + 2. Associated to these
dierent n, m there are dierent total wave functions (x, y):
N
(x, y) =
1,n1
(x)
2,m+1
(y) ,
N
(x, y) =
1,n2
(x)
2,m+2
(y) .
Therefore, there is degeneracy. Consider N = 0. In this case, the only allowed combination
of n, m is
n = 0 , m = 0 ,
since N must be non-negative, so there is only one solution in this case, namely
0
(x, y) =
1,0
(x)
2,0
(y) .
1.8. Problems and Solutions 35
The energy of this level is E
0
=
c
and there is not degeneracy. Let us consider now
N = 1. In this case, there are two allowed combinations of n, m, namely
n = 1 , m = 0 and n = 0 , m = 1 .
The two corresponding total wave functions are
1
(x, y) =
1,1
(x)
2,0
(y) ,
1
(x, y) =
1,0
(x)
2,1
(y) .
Problem 1.8. Calculate the spectrum and the eigenfunctions of a two dimen-
sional innite well potential of sides a and b:
V (x, y) =
_
_
0 if 0 < x < a , 0 < y < b ;
+ otherwise .
Solution. The Schr odinger equation in the range 0 < x < a and 0 < y < b is
2
2m
_
2
x
2
+
2
y
2
_
(x, y) = E(x, y) ,
and since we have innite wells, so that the particle cannot escape, the boundary conditions
are
(0, y) = (a, y) = 0 , (x, 0) = (x, b) = 0 .
The potential is separable, hence we seek solutions of the form
(x, y) =
1
(x)
2
(y) .
Following the general procedure outlined in this section, it is easy to prove that the functions
36 Chapter 1. Schr odinger equation. Wronskian theorem.
1
(x) and
2
(x) must satisfy the following equations:
2
2m
d
2
1
(x)
dx
2
= E
1
1
(x) , (1.43a)
2
2m
d
2
2
(y)
dy
2
= E
2
2
(y) , (1.43b)
(1.43c)
with E
1
+E
2
= E. Using the following boundary conditions
1
(0) =
1
(a) = 0 ,
2
(0) =
2
(b) = 0 ,
the boundary conditions for (x, y) are automatically satised. Therefore, the solutions of
Eqs. (1.43) are those of two one-dimensional innite wells, one in the x-direction and the
other in the y-direction:
_
_
_
1,n
(x) =
_
2
a
sin
nx
a
E
(n)
1
=
2
2m
_
n
a
_
2
,
_
_
_
2,m
(y) =
_
2
b
sin
my
b
E
(m)
2
=
2
2m
_
m
b
_
2
,
where n, m are integer numbers, n, m > 0. Thus, the total eigenfunctions of the two-
dimensional innite well potential are given by
n,m
(x, y) =
_
2
a
_
2
b
sin
nx
a
sin
my
b
,
and the corresponding energies are
E
n,m
=
2
2m
_
_
a
_
2
n
2
+
_
b
_
2
m
2
_
.
Chapter 2
AN INTRODUCTION TO HILBERT SPACES FOR
PHYSICISTS
2.1 Basic denitions
Denition 2.1 (Hilbert space). An Hilbert space is a linear space endowed with a
scalar product and such that it is complete and separable.
Let us now clarify the meaning of the various terms.
2.1.1 Linear space
A set S is a linear space over a numerical eld N if
1. for all pair of elements , of S it is possible to assign one and only one
well-dened element in S, called the sum of and and denoted with
= + ;
2. for all pair a N and S it is possible to assign one and only one element
in S called product of by a and denoted with a.
The above operations must satisfy the following properties:
+ = + , a(b) = (ab) ,
( + ) + = + ( + ) , 1 = ,
a( + ) = a + a , 0 = 0 = ,
(a + b) = a + b ,
38 Chapter 2. An Introduction to Hilbert Spaces for Physicists
where S is referred to as the null element of S. has the following properties:
S , + = ,
and it is unique. Moreover,
a = .
Denition 2.2 (linear independence). The elements
1
,
2
, . . . ,
n
S are said to
be linearly independent if the equation
a
1
1
+ a
2
2
+ . . . + a
n
n
=
implies a
1
= a
2
= . . . = a
n
= 0.
Denition 2.3 (dimension). A linear space S is said to have nite dimension n if
it is always possible to nd n linearly independent elements in S but it is impossible
to nd n + 1 linearly independent elements. If there are arbitrarily many linearly
independent vectors the linear space is said to be innite-dimensional.
2.1.2 Scalar product
Denition 2.4. A scalar product is a mapping S S into C, denoted with (, ),
satisfying the following properties:
1. (, a
1
+ b
2
) = a(,
1
) + b(,
2
);
2. (, ) = (, )
;
3. (, ) 0 and (, ) = 0 if and only if = .
From these properties, one can easily prove that
(a
1
+ b
2
, ) = a
(
1
, ) + b
(
2
, ) .
2.1. Basic denitions 39
The norm of is dened as
||
_
(, ) .
The distance between two elements , S is given by
d
,
| | ,
and it is possible to prove that
1. || 0;
2. |a| = [a[ ||;
3. | + | || +|| (triangle inequality).
Theorem 2.1 (Schwarz inequality).
[(, )[ || || .
Denition 2.5 (Orthogonality). Two vectors and in an Hilbert space are said
to be orthogonal if (, ) = 0.
2.1.3 Sequences and convergence
In this section, we collect some basic denitions.
Denition 2.6. A sequence
1
,
2
, . . . ,
n
, . . . of elements of S is said to converge
to an element if the numerical sequence |
1
| , . . . , |
n
| , . . . converges to
zero.
Denition 2.7. A set is called closed if it contains all its limit points.
Denition 2.8. A set U is said to be dense in S if U together with all its limit
points is equal to S.
40 Chapter 2. An Introduction to Hilbert Spaces for Physicists
Theorem 2.2 (Cauchy convergence criterion). A sequence
1
,
2
, . . . ,
n
. . . satises
the Cauchy criterion if for each > 0, there exists an integer N such that |
m
n
|
for all n, m N. This is a necessary condition for the convergence of a sequence .
Denition 2.9 (completeness). A set S is said to be complete if all Cauchy se-
quences in S do converge to an element of S.
Denition 2.10 (separability). A set S is said to be separable if for any S and
> 0 there exists a
n
, of a countable sequence, such that |
n
| < . (Here, n
depends on both and .)
For example, the real line R is a separable set since the rational numbers are dense
in R and there are countably many rational numbers.
Example 2.1. Consider the space (
0
(0, 1) of all continuous functions f(t), 0
t 1, with the norm given by
|f|
2
=
_
1
0
[f(t)[
2
dt .
We shall prove that this space is not complete.
Let us consider the sequence of functions f
n
(t) dened as
f
n
(t) =
_
_
0 if 0 t
1
2
1
n
1 if
1
2
+
1
n
t 1
linear in t if
1
2
1
n
< t <
1
2
+
1
n
.
It is easy to prove that the discontinuous function
f(t) =
_
_
0 if 0 t
1
2
1 if
1
2
< t 1
satises the relation
lim
n
_
1
0
[f
n
(t) f(t)[
2
dt = 0 .
2.1. Basic denitions 41
Thus, the sequence f
n
(t) being convergent is also a Cauchy sequence, but it does
not converge to a continuous functions, that is, to an element in (
0
(0, 1), therefore
(
0
(0, 1) is not complete and cannot be an Hilbert space.
Before considering specic examples of Hilbert spaces, we need some other deni-
tion.
2.1.4 Orthonormal systems
Denition 2.11. A sequence
n
S is said to be orthonormal if
(
m
,
n
) =
m,n
,
for any two elements
n
and
m
in the sequence.
Denition 2.12. An orthonormal set
n
S is said to be a complete orthonor-
mal system if there exists no ,= orthogonal to every element of the sequence,
i.e.,
(,
k
) = 0 k = .
For example, in the three-dimensional Euclidean space the set of unit vectors along
the x- and y-directions is not a complete orthonormal set, even if it is an orthonormal
one.
Denition 2.13. Let U be a set of elements of an Hilbert space. The set of all
the linear combinations of the vectors in U is again a linear space, called the linear
variety spanned by U. If we add all its limit points, we obtain a closed linear variety
that contains U, which we shall denote with [U]: this is the closed linear variety
spanned by U.
Theorem 2.3. In an n-dimensional Hilbert space, every orthonormal set has a num-
ber of elements less than or equal to n. It is a complete orthonormal system only if
it has n elements.
42 Chapter 2. An Introduction to Hilbert Spaces for Physicists
Theorem 2.4. In a innite-dimensional Hilbert space, every orthonormal set has a
nite or countably innite number of elements. If it is a complete orthonormal system
then it must have countably innite many elements.
It should be emphasized that the converse is not generally true: it may happen that
a orthonormal set with an innite number of elements is not a complete orthonormal
system. In fact , it is sucient to remove one element from a complete set in order
to get a non-complete one which has innite elements.
Let
k
be an orthonormal set. Given an element , we dene the coecients
a
k
= (
k
, ) ,
and build the series
k
a
k
k
= .
This series is always convergent in an Hilbert space, moreover is orthogonal to
each
k
and if the set is complete this means that = .
Theorem 2.5. Let
n
be an orthonormal set. It is complete if and only if one of
the following three conditions is satised:
1. for any S,
=
k
a
k
k
,
where a
k
= (
k
, );
2. the closed linear variety spanned by
k
is equal to S;
3. for any pair of elements , S,
(, ) =
k
(,
k
)(
k
, ) .
2.2. Examples of Hilbert spaces 43
2.2 Examples of Hilbert spaces
2.2.1 The space l
2
()
l
2
() is the linear space having as elements all the sequences of complex numbers
a
k
such that
n
k=1
[a
k
[
2
converges as n +.
The basic operations in l
2
() are dened as follows. Let a = a
k
, b = b
k
and
c = c
k
in l
2
() and , C (i.e. they belong to the complex numbers).
1. We dene
a = a
k
, a + b = a
k
+ b
k
.
2. The scalar product is dened as
(a, b) =
k
a
k
b
k
.
Note that it satises all the properties to be actually a scalar product over l
2
()
since
(a, b) = (b, a)
,
(a, b + c) = (a, b) + (a, c) ,
and (a, a) is a positive real number and it is zero only if a = (0, 0, 0, . . .).
It is possible to prove that l
2
() is complete and separable.
Theorem 2.6. Every innite-dimensional Hilbert space is isomorphic to l
2
().
2.2.2 The space /
2
/
2
is the space of the square-integrable complex-valued functions of k real variables
(q
1
, . . . , q
k
), i.e.,
_
+
_
+
[(q
1
, . . . , q
k
)[
2
dq
1
dq
k
< .
44 Chapter 2. An Introduction to Hilbert Spaces for Physicists
The basic operations in /
2
are dened as follows. Given any two functions f and
g in /
2
and a complex number we dene f to be the function f(q
1
, . . . , q
k
) and
f + g to be the function f(q
1
, . . . , q
k
) + g(q
1
, . . . , q
k
). It is possible to prove that f
and f +g dened in such way actually belong to /
2
. The case for f is trivial, let us
consider the case of f + g.
We have to prove that
_
_
[f + g[
2
dq
1
dq
k
< . For simplicity, we shall
denote the integration element as dq = dq
1
dq
k
. We have:
_
[f + g[
2
dq =
_
[f[
2
dq +
_
[g[
2
dq +
_
2f
g dq .
Since
2f
g 2 [f[ [g[ ,
it follows
[f + g[
2
[f[
2
+[g[
2
+ 2 [f[ [g[ = ([f[ +[g[)
2
.
Moreover, as
([f[ [g[)
2
0 ,
it also true that
[f[
2
+[g[
2
2 [f[ [g[ .
Hence,
[f + g[
2
[f[
2
+[g[
2
+[f[
2
+[g[
2
= 2 [f[
2
+ 2 [g[
2
,
and since both
_
[f[
2
dq and
_
[g[
2
dq are < also
_
[f + g[
2
dq < .
The /
2
scalar product is dened as
(f, g) =
_
f
(q
1
, . . . , q
k
)g(q
1
, . . . , q
k
) dq
1
. . . dq
k
The properties of the scalar product are all satised except the one which says that
(f, f) = 0 implies f = 0. In fact, (f, f) = 0 means
_
[f[
2
dq = 0 and this does
not imply that f = 0, it is sucient that [f[ , = 0 over a set of zero measure (dq).
2.3. Operators in Hilbert spaces 45
Therefore, we consider equivalent two functions which dier on a subset having zero
measure. The null element of our Hilbert space is given by the (equivalent class of)
the functions that are zero almost everywhere.
Theorem 2.7. /
2
is innite-dimensional.
Proof. Let O
1
, . . . , O
n
be n non-overlapping subsets having Lebesgue measure greater
than zero, but nite. Let f
1
(q
1
, . . . , q
k
) take value 1 in O
1
and zero elsewhere. Since
_
[f
1
[
2
dq is equal to the measure of O
1
, which for hypotesis is nite, f
1
belongs to
/
2
. In the same way we can dene the functions f
2
, . . . , f
n
and these n functions
are linearly independent. Since the construction holds for an arbitrary large n, this
means that for all n it is possible to specify n linearly independent functions in /
2
,
thus /
2
is innite dimensional.
von Neumann in his book Mathematical Foundations of Quantum Mechanics has
shown that /
2
is complete and separable.
2.3 Operators in Hilbert spaces
We start with a more general setting: we consider operators between generic spaces,
that is, mappings from a set R into another set R
=
T ,
where R and
and
T
2
: the sum of these two vectors is a well-dened vector, since we are in
a linear space. It is easy to see that the domain of
T
3
is D
T
3
= D
T
1
D
T
2
.
Inverse of an operator. If the operator
T maps dierent vectors into dierent vec-
tors,
=
T ,
it is possible to dene the inverse
T
1
of
T as
=
T
1
.
Note that D
T
1 = R
T
and R
T
1 = D
T
.
Additive operators. The additive operators are those for which the following prop-
erty holds:
T( + ) =
T +
T ,
for all , D
T
.
Homogeneous operators. if for all a C and for all D
T
we have
T(a) = a
T ,
then the operator is said to be homogeneous. If instead
T(a) = a
T ,
then the operator is said to be anti-homogeneous.
Linear operators. The linear operators are those which are both additive and ho-
mogeneous. Those which are instead additive and anti-homogeneous are called
anti-linear.
2.3. Operators in Hilbert spaces 47
Theorem 2.8. The domain and the range of a linear (or anti-linear) operator are
linear spaces.
Theorem 2.9. Necessary and sucient condition for a linear (or anti-linear) operator
T to be invertible is that
T = implies = .
We introduce the notion of continuity for an operator.
Denition 2.14. An operator
T is said to be continuous at
0
S if for each > 0
there exists a
0
|
it follows that |
T
T
0
| < .
Related to the notion of continuity is the notion of boundeness.
Denition 2.15. An operator
T is said to be bounded in a subset U S if there
exists a real number R such that for all U
|
T| < R|| .
Denition 2.16. Two operators, indicated by
T and
T
T, ) = (,
T
) , (2.1)
for all , .
Eqn. (2.1) implies also that (prove the following as an homework ):
(,
T) = (
, ) .
You can prove also that
(
=
T , (a
T)
= a
, (
T
1
T
2
)
=
T
1
.
(Do this last one as an homework.)
48 Chapter 2. An Introduction to Hilbert Spaces for Physicists
Denition 2.17. If
T =
T
U
U
=
U
U = 11 ,
As a consequence of the previous denition, it follows
=
U
1
.
Thus, unitary operators are invertible.
Theorem 2.10. An operator is unitary if and only if it is dened everywhere and
norm-preserving.
Let us see only one part of the proof, namely, if an operator is unitary then it
preserves the norms:
(
U,
U) = (,
U
U) = (, ) .
Theorem 2.11. If
T is Hermitian and
U is unitary, then also
U
T
U
1
is Her-
mitian.
2.3.1 Projection operators
Let N be any subset of some Hilbert space H and consider the set of all the elements
of H which are orthogonal to all the elements of N . This is a closed linear subset
which we shall denote with H N .
Theorem 2.12. Let N be a closed linear subset of an Hilbert space H . Then each
H can be decomposed in one and only one way as the sum = + of two
vectors N and H N . is called the projection of in N .
Denition 2.19. Let N be a closed linear subset in an Hilbert space H . The
operator
P
N
(dened everywhere in H ) which assigns to each H its projection
2.4. The eigenvalue problem 49
in N (as dened by the previous theorem) is called the projection operator of N .
Using the notation employed in the statement of the previous theorem,
P
N
= .
Theorem 2.13. It is possible to prove that
P
N
is linear, Hermitian and idempotent
(that is,
P
2
N
=
P
N
).
Theorem 2.14. An operator
P dened everywhere is a projection operator if and
only if it is linear, Hermitian and idempotent. The closed linear subset N on which
P|
2
= (
P, ) , |
P| || , |
P
P| = |
P( )| | | ,
and moreover
P
H N
= 11
P
N
.
Homework 2.1. Let
P and
Q be projector operators on the two linear varietys M
and N , respectively. Prove that
P
Q is a projector operator if and only if
P and
Q
commutes among themselves and show that in that case
P
Q project on the variety
M N .
Homework 2.2. If
P
1
,
P
2
,
P
3
, . . . ,
P
N
are projector operators, prove that
P
1
+
P
2
+
P
3
+ . . . +
P
N
is a projector operator if and only if
P
i
P
j
=
i,j
P
i
.
2.4 The eigenvalue problem
A key role in quantum mechanics is played by the eigenvalue problem, that means to
solve the equation
H
r
= h
r
r
, (2.2)
50 Chapter 2. An Introduction to Hilbert Spaces for Physicists
where
H is a given linear
operator, h
r
is a real or complex number known as eigen-
value and
r
is any non-zero
r
,
r
) = (
r
, h
r
r
) ,
that is,
h
r
(
r
,
r
) = h
r
(
r
,
r
) .
Therefore, h
r
= h
r
, which means that h
r
is real.
The eigenvalue problem can be stated without problem also for non-Hermitian operators, however
in the following we are mainly concerned with Hermitian ones, since this is the case relevant for
quantum mechanics.
Of course, Eqn. (2.2) would hold trivially for any real or complex number h
r
if
r
was the null
vector of the Hilbert space. Moreover, if
r
is any solution of Eqn. (2.2) it is easily shown that
r
is a solution too, for any real or complex number . So, it is sucient to consider normalized
solutions: |
r
| = 1. However, for the continuous spectrum (see later) normalization must be
interpreted in the sense of distributions.
2.4. The eigenvalue problem 51
Let us consider two dierent eigenvectors
r
and
s
corresponding to dierent
eigenvalues h
r
and h
s
, respectively: we have
(
H
r
,
s
) = (
r
,
H
s
) ,
(h
r
r
,
s
) = (
r
, h
s
s
) ,
h
r
(
r
,
s
) = h
s
(
r
,
s
) . (2.3)
From Eqn. (2.3) and using the fact that h
r
is real, we get
(h
r
h
s
)(
r
,
s
) = 0 .
Since h
r
,= h
s
, we obtain
(
r
,
s
) = 0 ,
that is,
r
and
s
are orthogonal and thus also linear independent.
In the case of continuous spectrum, the Hilbert space must be enlarged to include
improper eigenstates which formally satisfy Eqn. (2.2) but that strictly speaking do
not belong to the Hilbert space and that are normalized in the generalized sense of the
Dirac-delta that we mentioned in the previous chapter. We have already encountered
such situation discussing plane wave function solutions of the free particle Hamilto-
nian. The reader should realize at this point that the stationary Schr odinger equation
is nothing but an eigenvalue problem for the Hamiltonian dierential operator in the
Hilbert space /
2
of the square-integrable wave functions.
We now introduce the concept of degeneracy of an eigenvalue. When two or more
linearly independent eigenvectors
r
,
r
, . . . correspond to the same eigenvalue h
r
,
H
r
= h
r
r
and
H
r
= h
r
r
,
we say that h
r
is a degenerate eigenvalue. Any linear combination of
r
,
r
, . . . is an
52 Chapter 2. An Introduction to Hilbert Spaces for Physicists
eigenvector with the same eigenvalue h
r
:
H(
r
+
r
) =
H
r
+
H
r
= h
r
r
+ h
r
r
= h
r
(
r
+
r
) .
That is, for any given eigenvalue h
r
all linear independent eigenvectors correspond-
ing to h
r
form a (closed) linear eigenvariety. For any eigenvariety one can extract
two or more (depending on the dimension of the eigenvariety) mutually orthogonal
eigenvectors that span the whole variety.
The set of all discrete and continuous eigenvectors,
k
,
, is said to be complete
if for any H
=
k
(
k
, )
k
+
_
d(
, )
, (2.4)
and for any pair , H
(, ) =
k
(,
k
)(
k
, ) +
_
d(,
)(
, ) . (2.5)
(Here, H denotes the Hilbert space.)
Hermitian operators whose spectrum is complete in the sense specied above are
called hypermaximal operator. In the following, we shall assume that all the Her-
mitian operators are hypermaximal.
Remark. There are eigenvalues and eigenvectors also for non-Hermitian operators.
Theorem 2.17. If
i
is eigenvector of the operator
A with eigenvalue a
i
:
A
i
= a
i
i
,
then
A
n
i
= a
n
i
i
,
for every positive integer n.
2.5. Functions of one operator 53
Theorem 2.18. If
A is a linear Hermitian operator, the equation
A
n
= 0 is equiv-
alent to
A = 0.
Theorem 2.19. Let
F(
H) =
H
n
+ a
1
H
n1
+ . . . + a
n
= 0
be the lowest-degree equation satised by the linear Hermitian operator
H. Then
the roots of F(x) = 0 are never degenerate;
the roots of F(x) = 0 are the only eigenvalues of
H;
the set of eigenvectors of
H is complete.
The equation F(
H) = 0 is called the secular equation.
2.5 Functions of one operator
We will specify here what we mean by a function F(
H) of an operator
H.
Let us suppose that
H is a hyper-maximal Hermitian operator acting on an Hilbert
space H . We dene F(
H) in terms of functions of the eigenvalues of
H, namely,
F(
H) is the operator which has the same eigenvectors
r
of
H and F(h
r
) are the
corresponding eigenvalues, i.e.,
H
r
= h
r
r
and F(
H)
r
= F(h
r
)
r
.
It is not even necessary for F to be dened over all x, it is sucient that F is dened
for the set of points x = h
r
.
If
H is hypermaximal, F(
H) is well-dened for all vectors : from Eqn. (2.4), i.e.,
=
r
+
_
d ,
54 Chapter 2. An Introduction to Hilbert Spaces for Physicists
it follows
F(
H) =
r
F(h
r
)
r
+
_
F(h
d .
For this state to be well-dened, it must belong to the Hilbert space. For example, if
the Hilbert space is H = /
2
, one must have
[F(h
r
)
r
[
2
+
_
[
F(h
)[
2
d < .
The adjoint of F(
H) is indicated by F
(
H) and it is such that
(, F(
H)) = (F
(
H), ) .
It is easy to prove that
F
(
H)
r
= F
(h
r
)
r
.
2.6 Commutativity and compatibility
If is eigenvector of both
A and
B, then [
A,
B] = 0. In fact, from
A = a and
B = b
one gets
[
A,
B] =
A
B
B
A =
A(b)
B(a)
= b
A a
B = ba ab = 0 .
Theorem 2.20. Two hyper-maximal Hermitian operators share a complete set of
eigenvectors if and only if they commute.
Proof. Let us suppose that there is no degeneracy. Suppose [
A,
B] = 0. Let
bn
A
a
= a
a
.
2.6. Commutativity and compatibility 55
Let us expand
a
on the eigenvectors
bn
:
a
=
bn
c
bn
a
bn
. (2.6)
Since
A
a
= a
a
_
Aa
_
a
= 0 ,
it follows from Eqn. (2.6) that
bn
(
Aa)c
bn
a
bn
= 0 . (2.7)
Now, we shall prove that the state
(
Aa)c
bn
a
bn
is an eigenvector of
B with eigenvalue b
n
:
B
=
B
_
(
Aa)c
bn
a
bn
_
= (
Aa)
Bc
bn
a
bn
= (
Aa)c
bn
a
b
n
bn
= b
n
_
(
A a)c
bn
a
bn
_
= b
n
A
_
c
bn
a
bn
_
= a
_
c
bn
a
bn
_
,
that is, c
bn
a
bn
are eigenvectors of
A with eigenvalue a, but we already know they are
also eigenvectors of
B with eigenvalue b
n
, therefore the various c
bn
a
bn
are simultane-
ously eigenvectors of both
A and
B. Eqn. (2.6) is the expansion of one eigenstate of
A on the eigenstates of
B. Since the set of eigenstates of
A is complete, any H
can be written as
=
an
c
an
an
but the eigenstates
an
can be expanded in terms of the eigenstates of
B, therefore
we can conclude that can be expanded in common eigenstates of
A and
B.
For a proof of the converse, let
n
be the complete set of common eigenstates of
A and
B. Clearly,
[
A,
B]
n
= 0 , (2.8)
because we have
[
A,
B]
n
= (
A
B
B
A)
n
=
A(b
n
n
)
B(a
n
n
)
= b
n
a
n
n
a
n
b
n
n
= 0 .
Now, let be any vector in the Hilbert space. We can expand over the basis of
n
=
n
c
n
n
,
From this formula and using the previous result (2.8) we nd immediately [
A,
B] = 0,
thus
A and
B commute because is any element of the Hilbert space.
2.7 Complete set of commuting operators
The main result of the last section, namely the theorem on simultaneous diagonal-
lizability of two commuting Hermitian operators, has been obtained assuming no
degeneracy.
Now, let us suppose there are two linearly independent eigenstates
r
and
r
cor-
responding to the same (degenerate) eigenvalue a
r
of
A. Clearly, the eigenvalue a
r
is not sucient to uniquely characterize its eigenvectors. Let
B be an hypermaximal
Hermitian operator commuting with
A. In the eigenvariety associated with the eigen-
value a
r
, we look for eigenvectors of
B. Such eigenstates are linear combinations of
2.7. Complete set of commuting operators 57
r
and
r
, and we shall denote them with
r
and
r
. Suppose that they correspond
to two dierent eigenvalues b
r
and
b
r
of
B, respectively. Then, we can say that the
two eigenstates
r
and
r
are uniquely specied by a pair of dierent eigenvalues:
r
(a
r
, b
r
) ,
r
(a
r
,
b
r
) .
If this is actually the case, the degeneracy is removed. Otherwise, we can iterate the
procedure and look for a third operator
C which commutes both with
A and
B. We
look for linear combinations of
r
and
r
, say
r
and
r
, that are also eigenstates
of
C. In general, we repeat the procedure until all degeneracies have been lifted. If
the eigenvalues of
C associated to
r
and
r
are dierent, say, c
r
, c
r
, then the triples
(a
r
, b
r
, c
r
) and (a
r
, b
r
, c
r
) are dierent and characterize uniquely
r
and
r
. If this is
the case,
A,
B and
C are said to form a complete set of commuting operators.
Denition 2.20. A set of mutually commuting hypermaximal operators is said to
be complete when specifying the eigenvalues with respect to all the operators, we
can determine uniquely a common eigenvector.
Theorem 2.21. Let
A,
B,
C, . . . be a complete set of commuting hypermaximal op-
erators. Any other hyper-maximal operator
F which commutes with all the operators
in this set must be a function of
A,
B,
C, . . .
Denition 2.21. A subspace S in the domain of an operator
A is called invariant
if by applying
A to any of the elements in S we get a new state which belongs to S
too. Therefore, the image of an invariant subspace S is again within S.
Denition 2.22. A set of operators is said to be irreducible if they have no common
invariant subspace.
Theorem 2.22. If a set of operators
A,
B. . . is irreducible, then any operator which
commutes with all of them is a multiple of the identity.
58 Chapter 2. An Introduction to Hilbert Spaces for Physicists
2.8 Dirac notation
We summarize briey how to translate the notation used so far in the Dirac notation.
[ = ket. They represent vectors in the Hilbert space.
[ = bra. They represent the space of linear mappings ([) that associate
to every state [ a corresponding complex number given by the scalar product
(, ).
scalar product: (, ) [;
eigenvalue problem:
H
r
= h
r
r
H[
r
= h
r
[
r
;
superposition of states:
[ = [
1
+ [
2
, [ =
_
2
1
f() [ d ;
Theorem. The object
O [u v[ acts as a linear operator.
Proof. Let us apply
O to a state [:
O[ = [u v[ .
Now, v[ is nothing but a number that multiplies the state [u, thus
O[
has produced a state. Let us now prove how it acts on a linear superposition of
states:
O[[ + [] = [u v[ [[ + []
= [u v[ + [u v[
=
O[ +
O[ ,
so this proves that
O = [u v[ is a linear operator.
2.9. Spectral decomposition 59
Dual space: Given a ket [, the corresponding bra [ is dened as that linear
map for which
[ = ||
2
.
Operators: Let us suppose that acting with an operator
A on a state [ we
get the state [
A[. The scalar product of this state with another state
[ is
[
= [
_
A[
_
;
to indicate this scalar product we introduce the notation [
A.
Adjoint: The adjoint
A
of an operator
A is dened by
[ = [
A ,
where we have used the notation previously introduced. From this relation, it
is clear that
A acts on the space of kets, while
A
=
A[ ,
then
[
A
[ .
2.9 Spectral decomposition
Theorem 2.23. For a hypermaximal Hermitian operator
H having spectrum h
n
and
corresponding eigenstates [h
n
the following spectral decomposition formula holds:
H =
n=1
h
n
[h
n
h
n
[ . (2.9)
In other words,
H is the sum of the projector operators [h
n
h
n
[ multiplied by the
corresponding eigenvalues h
n
.
60 Chapter 2. An Introduction to Hilbert Spaces for Physicists
For operators having also a continuous spectrum [
h
, Eqn. (2.9) must be general-
ized as
H =
n=1
h
n
[h
n
h
n
[ +
_
h
2
h
1
h
[
h
h
[ d
h
. (2.10)
Proof. We restrict ourselves to the case (2.9) and we assume no degeneracy.
Since
H is hyper-maximal, its eigenstates form a complete basis for the Hilbert
space. Therefore, any state [ in the Hilbert space can be expanded in the following
way
[ =
n
c
n
[h
n
. (2.11)
The operator identity (2.9) must be understood in the sense that if we apply the
operator
H to some state [ the result
H[ is just the same state we would obtain
applying the right-hand side of Eqn. (2.9), i.e.:
H[ =
_
n
h
n
[h
n
h
n
[
_
[ . (2.12)
Let us prove this formula. By inserting the expansion formula (2.11) in the left-hand
side of Eqn. (2.12) we get
H[ =
H
_
n
c
n
[h
n
_
=
n
c
n
H[h
n
=
n
c
n
h
n
[h
n
,
while the right-hand side of Eqn. (2.12) is nothing but
_
n
h
n
[h
n
h
n
[
_
[ =
n
h
n
[h
n
h
n
[ .
Our goal would be met if we were able to prove that the expansion coecients in
Eqn. (2.11) are actually given by c
n
= h
n
[. This result on c
n
can be easily derived
directly by means of Eqn. (2.11):
h
i
[ =
n
c
n
h
i
[h
n
. .
n,i
= c
i
,
hence c
n
= h
n
[. This completes the proof.
2.10. Tensor Product 61
Here is a summary of some important identities encountered in this chapter. First
of all, remember the important Eqn. (2.4), i.e.:
=
k
(
k
, )
k
+
_
d(
, )
.
We can write it using the Dirac notation as follows:
[ =
k
[ [
k
+
_
d
[ [
,
or
[ =
k
[
k
k
[ +
_
d [
[ .
This is possible because
k
[ is a number. Since the above expression must hold
for every state [ in the Hilbert space, it is usually written as an operator identity
by dropping [ in every term:
11 =
k
[
k
k
[ +
_
d [
[ , (2.13)
where 11 is the identity operator. Eqn. (2.13) is referred to as decomposition or
resolution of the identity. It means that applying the left-hand side to a state [
we get the same result as if we apply the right-hand side to the same state.
Let us now turn to the spectral decomposition, Eqn. (2.10):
H =
n=1
h
n
[h
n
h
n
[ +
_
h
2
h
1
d
h
[
h
h
[
h
. (2.14)
It is then easy to generalize this to a generic function F(
H) of the operator
H as :
F(
H) =
n=1
F(h
n
) [h
n
h
n
[ +
_
h
2
h
1
F(
h
) [
h
h
[ d
h
. (2.15)
2.10 Tensor Product
The tensor product, usually indicated with the symbol , is an operation which
can be applied to any couple of vector-spaces even innite-dimensional ones like the
62 Chapter 2. An Introduction to Hilbert Spaces for Physicists
Hilbert spaces. Before giving more mathematical details let us illustrate the case
with a physical example. Let us consider a system made of two particles 1 and 2
and take for each particle a particular wave function: (r
1
) and
(r
2
). The product
(r
1
)
(r
2
) represents a wave function of the combined system. Of course it is not the
most general wave function for the combined system. This will be a (r
1
, r
2
) which
is not in general factorized into two parts like the wave function above. It will be
in general a linear combination of factorized wave functions . In fact let us look at
(r
1
, r
2
) as a wave function of particle 1, then it can be expanded on a basis of wave
functions of particle 1 as:
(r
1
, r
2
) =
C
i
i
(r
1
). (2.16)
Clearly the coecients C
i
must be functions of r
2
, i.e.:
(r
1
, r
2
) =
C
i
(r
2
)
i
(r
1
)
and we can interpret the C
i
(r
2
) as wave functions of r
2
multiplied by some coecients
a
i
like:
C
i
(r
2
) = a
i
i
(r
2
)
and the relation (2.16) becomes:
(r
1
, r
2
) =
i
a
i
i
(r
2
)
i
(r
1
). (2.17)
So we can say that the space of wave functions of the combined system is formed
by the linear combinations of products of wave functions of particle 1 and of wave
functions of particle 2. If the Hilbert space of particle 1 is indicated with H
(1)
and
the one of particle 2 with H
(2)
, the Hilbert space of the combined system of the two
particles ,whose wave functions we have built in the example above, is dened as the
tensor product of the two separate Hilbert spaces and indicated as:
H H
(1)
H
(2)
(2.18)
The states in Eqn. (2.17) can be indicated also in the following way:
2.10. Tensor Product 63
(r
1
, r
2
) =
a
i
i
(r
2
)
i
(r
1
)
The observables associated to particle 1 will be indicated with the symbol
O
(1)
and
those associated to particle 2 as
O
(2)
. If
O
(1)
acts on (r
1
) as follows:
O
(1)
(r
1
) =
(r
1
)
then on the tensor product it will act as follows:
O
(1)
((r
1
)
(r
2
)) =
(r
1
)
(r
2
). (2.19)
Similarly an observable
O
(2)
acting on particle 2 as :
O
(2)
(r
2
) =
(r
2
)
will act on the tensor product space as
O
(2)
((r
1
)
(r
2
)) = (r
1
)
(r
2
) (2.20)
An equivalent manner to indicate this procedure is the following: an operator
O
(1)
acting on the Hilbert space H
1
get extended, in the tensor product space H
H
1
H
2
, into the following operator:
O
(1)
=
O
(1)
11
where 11 is the identity operator acting on the Hilbert space of particle 2. The
notation above summarize the operation indicated in Eqn. (2.19) where the state
(r
2
) is left intact.
The same procedure we follow for an operator
O
(2)
acting on the states of particle
2. In the tensor product space this operator get extended as follows:
O
(2)
=11
O
(2)
64 Chapter 2. An Introduction to Hilbert Spaces for Physicists
In general the operators acting in the Hilbert space tensor product of the two, will
have the form
O(r
1
, r
2
)
i
A
i
(r
1
)
B
i
(r
2
)
and will act on a generic state belonging to the tensor product space:
(r
1
, r
2
) =
C
j
j
(r
1
)
j
(r
2
)
in the following manner:
O(r
1
, r
2
)
(r
1
, r
2
) =
i,j
i
C
j
A
i
(r
1
)
j
(r
1
)
B
i
(r
2
)
j
(r
2
) (2.21)
One last remark: as r
1
and r
2
refer to dierent particle, they commute. The same
happens for the operator
A
i
(r
1
) and
B
i
(r
2
)
2.11 Problems and Solutions
Problem 2.1. Prove that
([u v[)
= [v u[ .
Solution. Let
O [u v[. We know that
[
=
O[ [
O
[ .
Let us now investigate if it possible from
([u v[)
[ [
, (2.22)
to prove that:
([v u[) [ = [
. (2.23)
2.11. Problems and Solutions 65
We know that Eqn. (2.22) is equivalent to
[ = [ ([u v[)
= [u v[ ,
i.e.,
[ = a v[ with a = [u. Now, taking the dual we get (see the proof afterwards)
[
= a
[v , (2.24)
i.e.,
[
= ([u)
[v = u[ [v .
u[ is a number, and we can equivalently put it in the following form:
[
= |
|
2
,
if
[ = v[ a ,
then we get
= | [v |
2
|a|
2
.
This holds if [
= a
= aa
v[v = |a|
2
| [v |
2
.
Problem 2.2. Prove that the operator
= [ [
with [ = 1 is a projection operator.
66 Chapter 2. An Introduction to Hilbert Spaces for Physicists
Solution. It is necessary to prove the following facts.
1.
P
is Hermitian, i.e.,
P
=
P
.
3.
P
is idempotent, i.e.,
P
2
=
P
. We have
P
2
= ([ [) ([ [) = [ [
. .
1
[ = [ [ =
P
.
Problem 2.3. Prove that, in the Hilbert space /
2
of the square-integrable func-
tions on x, the operator i/x is Hermitian.
Solution. /
2
is endowed with the scalar product dened for any two square-integrable
functions f(x) and g(x) of one variable x as
f[g =
_
+
(x)g(x) dx .
An operator
O is Hermitian if
O
=
O where
O
f[g = f[
Og ,
therefore for an Hermitian operator
O we have
Of[g = f[
Og .
Let us consider
O = i/x. The above relation becomes
_
+
_
i
f
x
_
g(x) dx
?
=
_
+
(x)i
g
x
dx .
We prove that this equality actually holds by integrating by parts the integral on the left-
hand side. We have
i
_
+
x
g(x) dx = i f
(x)g(x)[
+
. .
0
+
_
+
(x)i
g
x
dx
2.11. Problems and Solutions 67
The surface term vanishes since both f
A[ = a
1
[
B[ = b
3
[
C [ = c
5
[
.
.
.
M[ = m
15
[
where the numbers 1, 3, 5, . . . are just to say that we have more than one eigenvalue and
we select one particular eigenvalue among all of them. Since the set is complete the
correspondence between the eigenstate [ and the set of numbers (a
1
, b
3
, c
5
, . . . , m
15
)
is one-to-one, we can use the set of eigenvalues to label the state [ , in particular we
shall use the notation:
[a
1
, b
3
, c
5
, . . . , m
15
i
c
i
[o
i
.
Let us suppose that both [ and the states [o
i
are normalized and that the
eigenvalues o
i
are not degenerate. Then, the probability of nding o
i
in a mea-
surement of
O is
P
o
i
= [c
i
[
2
. (3.1)
This procedure is possible if
O is a complete set of commuting operators by
itself. Otherwise, there is degeneracy . In this case we must look for Hermitian
operators dierent than
O which commute each other and with
O and that form
a complete set of commuting operators. The expansion formula for [ becomes
in this case
[ =
i,r
c
i,r
[o
i
, r ,
72 Chapter 3. Postulates of Quantum Mechanics
where the index r labels the degeneracy. In this case the probability of nding
o
i
in a measurement of
O becomes
P
o
i
=
r
[c
i,r
[
2
. (3.2)
Postulate 6 (Measurement postulate). Let us suppose we prepare a system in the
state [ and perform a measurement of an observable
O. Let us also suppose
that we get the eigenvalue o
i
. If immediately after we perform another mea-
surement of
O we get o
i
again with probability 1. This means that immediately
after the rst measurement the original state [ has become [o
i
:
[
measurement of
O
[o
i
measurement of
O
[o
i
.
In other words, the state [
measurement of
O
[o
i,t
.
It should be noted this postulate implies there exist in quantum mechanics two dif-
ferent ways in which states transform: the (reversible) time evolution given by the
time-dependent Schr odinger equation and the (irreversible) wave packet reduction
associated to any measurements process.
3.1. Diracs correspondence rules 73
Let us now put forward a technical remark regarding the measurement postulate
in the case in which the measured observable
O does not form a complete set of
commuting Hermitian observables by itself and degeneracy occurs. Let o
i
be the
outcome of a measurement of
O. A whole set of eigenstates [o
i
, r are now associated
to the eigenvalue o
i
, being r the index which labels the degeneracy. The question
is: in which of the various [o
i
, r should the state collapse immediately after the
measurement? We know the various [o
i
, r span an eigenvariety. The state in which
[ collapses is the projection of [ on this eigenvariety, and we shall denote this
state with [
o
i
. By denoting with
P
i
the projection operator on the eigenvariety
spanned by [o
i
, r, the measurement process can be represented schematically in the
following way:
[
measurement of
O
with outcome o
i
[
o
i
=
P
i
[
|
P
i
[ |
.
3.1 Diracs correspondence rules
These rules, to some extent, allow us to build most of the observables in quantum
mechanics starting from their classical analogous.
Postulate 7. 1. In classical mechanics, observables are functions of q and p, for
example O
1
(q, p) and O
2
(q, p). The corresponding quantum observables
are obtained
1
i
[
O
1
( q, p),
O
2
( q, p)] .
Excluding certain ambiguities in the ordering, see remark 2 on the next page.
74 Chapter 3. Postulates of Quantum Mechanics
For example,
q, p
P.B.
= 1
1
i
[ q, p] = 1 .
So, we have
[ q, p] = i . (3.3)
These are called Heisenberg commutation rules.
Now, let us discuss the limits of this correspondence principle.
1. There exist quantum observables having no classical analogous, such as spin,
isospin, and so. In this case the correspondence principle does not work.
2. Since q and p do not commute at the quantum level, ambiguities may arise due
to the fact that more than one quantum Hermitian operator can be associated
to the same classical observable by changing the ordering in which the various q
and p terms are considered. To clarify this point, let us consider for example the
observable O
1
= pq. We can write this expression equivalently at the classical
level as O
1
= qp or O
1
= (qp + pq)/2. According to the correspondence rule,
the associated quantum observables are
qp q p =
O
1
qp + pq
2
q p + p q
2
=
O
1
pq p q =
O
1
It is easy to prove that
O
1
,=
O
1
,=
O
1
. For example,
O
1
= q p = p q + [ q, p] =
O
1
+ i .
This means that, in constructing the quantum observables using the correspon-
dence principle starting from the classical observables, we must be aware of the
3.2. More on the postulates of quantum mechanics 75
fact that there exist quantum observables with no classical analogous and, for
those having classical analogous, we must prescribe the ordering of q and p.
The latter ambiguity is referred to as ordering ambiguity.
3.2 More on the postulates of quantum mechanics
There is a further postulate called spin-statistics which we will present in chapter 8.
In this section we shall comment on the postulates of quantum mechanics introduced
in the previous sections and we shall try to understand them more deeply.
3.2.1 Expectation values
We have seen from postulate 3 that the outcomes of a measurement are probabilistic
and not deterministic: one can only calculate the probability that by measuring an
observable
O the outcome will be an eigenvalue o
i
. Accordingly to the postulate 3
the probability is given by the following rule: if the system is in the state [, rst of
all you have to expand [ on the basis of the eigenstates of
O (resolving degeneracy
if any)
[ =
i
c
i
[o
i
, c
i
= o
i
[ ,
then the probability of nding o
i
performing a measurement of
O on the state [ is
P
i
= [c
i
[
2
.
(Do not confuse this P
i
with the projector operator
P
i
introduced in the last section.)
Thus, one can only speak about mean value
O of an observable
O on a quantum
state [:
O =
i
P
i
o
i
=
i
[c
i
[
2
o
i
.
76 Chapter 3. Postulates of Quantum Mechanics
It is not dicult to prove that the mean value can also be obtained using the following
expression :
O = [
O[ . (3.4)
In fact, as the [o
i
are a complete set, we can write
i
[o
i
o
i
[ = 1 ,
and using this we get:
O = [
O[ = [
O
_
i
[o
i
o
i
[
_
[ =
i
[
O[o
i
o
i
[ =
i
[o
i
[o
i
o
i
[
=
i
o
i
[o
i
o
i
[ =
i
o
i
[ o
i
[ [
2
=
i
o
i
[c
i
[
2
.
3.2.2 Compatibility of observables and measurements.
Let us consider two commuting hyper-maximal Hermitian operators
A and
B:
[
A,
B] = 0 .
In literature it is usually found the statement that these two observables can be
measured simultaneously, and are said to be mutually compatible. This means the
following thing: let us suppose we measure
A and nd the eigenvalue a
1
. Immediately
after, say at the time /2 where is a strictly positive real number as small as we
want, we measure
B and we nd say b
2
. At a time /2 after the measurement of
B
we perform a measurement of
A again and we nd a
1
with probability 1 because of
postulate 3. The reason why we consider small times is that we do not want the
state to change too much during the time evolution.
To understand how this concept of compatibility of two observables is linked
to the postulates of quantum mechanics, we consider now the case in which the
The formula is also correct in the case of degeneracy and when observables have continuous
spectrum.
3.2. More on the postulates of quantum mechanics 77
eigenvalue a
1
above shows degeneracy and thus there exists a whole eigenvariety of
eigenstates associated to a
1
. After the measurement of
A, the state [ of the system
immediately before the measurement of
B collapses into [
a
1
, i.e. the projection of
[ on the eigenvariety associated to a
1
. For the postulates of quantum mechanics, if
after a small time /2 we measured
A again we would nd again a
1
. After a further
identical time interval let us perform a measurement of
B. Since
A and
B commutes,
from the theorems on theory of Hilbert spaces we know
A and
B share at least one
complete set of eigenstates. This means there must be at least one eigenstate of
B in
the eigenvariety associated to a
1
. Let us suppose that there are two such eigenstates.
If we nd b
2
in the measurement of
B, [
a
1
will collapse in the state [a
1
, b
2
. It is
now clear that if, after a time /2, we measure
A again, since the system is in the
state [a
1
, b
2
we will obtain a
1
again.
Therefore, we can say that a measurement of
B has not changed the result of a
measurement of
A: we get a
1
at the time t = 0 and at the time we get a
1
again. It
is said also that a measurement of
B does not disturb
A. That is, we can measure
simultaneously
A and
B. This is the meaning of compatibility of two observables.
3.2.3 Incompatible observables
We know that two non-commuting observables cannot share a complete set of eigen-
states. They can have a set of eigenstates in common, but this set cannot be complete.
For example, q and p do not commute: [ q, p] = i, and it is easy to show that
these two observables have no eigenstate in common. In fact, if there were one such
common eigenstate, say [v, then we would get
[ q, p] [v = q p [v p q [v = qp
v
[v pq
v
[v
= p
v
q [v q
v
p [v = p
v
q
v
[v q
v
p
v
[v = 0 ,
while from [ q, p] = i we would get [ q, p] [v = i [v , = 0.
78 Chapter 3. Postulates of Quantum Mechanics
Two non-commuting operators
A and
B, [
A,
B] =
C ,= 0, can have in general one
eigenstate in common but it is easy to prove that this eigenstate must be an eigenstate
of
C with eigenvalue zero. In the case of q and p, i does not admit zero as eigenvalue.
Let us return to our discussion about measurements of incompatible observables.
Let us expand the state [q
0
(in which our system has collapsed after a measurement
of q) on the basis of the eigenstates of p:
[q
0
=
p
C(q
0
, p) [p . (3.5)
Here the summation notation has been used, however the reader should keep in mind
that p has a continuous spectrum and the summation must be replaced actually
by an integration over the continuous variable p. Since q and p have no common
eigenstates, C(q
0
, p) are always dierent from 1 for any p. After a time /2 we
perform a measurement of p and let us suppose to nd p
0
. After a second amount of
time /2 we measure again q. The outcome is not q
0
with probability 1 as in case of
the previous section when the two measured observables were compatible. What we
have to do is to expand [p
0
on the eigenstates of the position operator
[p
0
=
C(q, p
0
) [q
where
C(q, p
0
) are dierent from 1 for any q. So in general there is a non-zero
probability P
q
= [
C(q, p
0
)[
2
of nding a value q, and thus the outcome of the second
measurement of q is not necessarily q
0
.
The conclusion is that in the case of two non-commuting observables, the measure-
ment of one observable disturbs the second observable. Consequently, the mea-
surement postulate must be restated in the following way: after a small time the
system is still in q
0
if and only if no measurement process has been done in between
of a quantum observable which does not commute with the rst.
3.3. Heisenberg uncertainty principle 79
3.3 Heisenberg uncertainty principle
Besides the probabilistic aspects of measurements, there are also correlations among
them if they satisfy proper commutation relations. We will analyze here the most
famous case. Let
Aand
B be two non-commuting hyper-maximal Hermitian operators
such that their commutator satisfy the relation
[
A,
B] = i
C , (3.6)
where
C is an Hermitian operator. Moreover we require that the states obtained by
applying these operators to the original states belong to the same original Hilbert
space.
Given the average value of the observable
A with respect to the state [
A = [
A[
and the mean square
A =
_
[
A
A]
2
A
B
1
2
[
C [ . (3.7)
In particular, in the case in which
A = q and
B = p we have
C = and the previous
relation becomes
q p
1
2
.
This is Heisenberg uncertainty principle.
We now give a proof of Eqn. (3.7). We start by evaluating (
A)
2
. Let us diagonalize
A:
A[a
k
= a
k
[a
k
.
Being
A hyper-maximal, we can expand any [ on the basis of the eigenstates of
A
according to the usual expansion formula:
[ =
k
c
k
[a
k
, c
k
= a
k
[ .
80 Chapter 3. Postulates of Quantum Mechanics
By inserting the latter in the expression for (
A)
2
we get
(
A)
2
= [[
A
A]
2
[
=
j,k
c
k
c
j
a
k
[[
A
A]
2
[a
j
j,k
c
k
c
j
_
a
k
A
__
a
j
A
_
a
k
[a
j
. .
j
k
=
k
[c
k
[
2
_
a
k
A
_
2
.
Since [c
k
[
2
is nothing but the probability of nding a
k
in a measurement of
A, the
last formula is nothing but the usual formula of the standard deviation.
By multiplying
_
A
_
2
by
_
B
_
2
we get
(
A)
2
(
B)
2
= [[
A
A]
2
[ [[
B
B]
2
[
= |(
A
A) [ |
2
|(
B
B) [ |
2
(3.8)
since both
A and
B are Hermitian and since
A and
A
A
A ,
B
B
B .
Then Eqn. (3.8) can be rewritten as
(
A)
2
(
B)
2
= |
A[ |
2
|
B[ |
2
.
Using the Schwarz inequality on the vectors
A[ and
B[ we get
_
_
_
_
A
_
[
_
_
_
_
_
_
_
B
_
[
_
_
_
A
__
B
_
,
from which it follows
A
B
A
__
B
_
, (3.9)
The modulus of a complex number is always bigger than the modulus of its imaginary
part, thus on the right-hand side of Eqn. (3.9) we can take the imaginary part and
3.4. Position and momentum representations 81
perform the following steps:
A
B
1
2i
_
[(
A
A)(
B
B)[ [(
A
A)(
B
B)[
1
2i
_
[(
A
A)(
B
B)[ [(
B
B)(
A
A)[
_
1
2i
[[(
A
A), (
B
B)][
=
1
2
[
C [ .
A
B = 0 only if [[
A,
B][ = 0, that is when [
C [ = 0.
3.4 Position and momentum representations
Let us start from the commutation relation
[ q, p] = i , (3.10)
and ask ourselves if there exists a complete set of eigenstates common to both oper-
ators:
_
_
_
q [q, p = q [q, p
p [q, p = p [q, p
If there were such [q, p, it would be possible to expand any state [ as
[ =
_
dq dp C(q, p) [q, p ,
Now, applying the commutatator [ q, p] to [ using the expansion formula above
would yield
[ q, p] [ =
_
qp C(q, p) [q, p dq dp
_
pq C(q, p) [q, p dq dp = 0 ,
while we know by Eqn. (3.10) that
[ q, p] [ = i [ , = 0 .
82 Chapter 3. Postulates of Quantum Mechanics
Thus, as a consequence of the commutation relation (3.10) there cannot be such
common eigenstate [q, p.
Another question: is it possible to represent q and p operators by means of nite-
dimensional matrices? If it were possible then, by taking the trace of Eqn. (3.10), we
would get :
Tr[ q, p] = Tr( q p) Tr( p q) = Tr( q p) Tr( q p) = 0 ,
where the two main properties of the trace, namely linearity and cycliccity, were used.
But using the right-hand side of Eqn. (3.10) we get directly
Tr[ q, p] = Tr i = iN ,
being N the dimension of the matrix. For the commutation relation (3.10) to hold,
it is thus not possible to represent q and p by means of nite-dimensional matrices.
Now, since q is Hermitian under the usual scalar product (that is, the /
2
scalar
product [ =
_
[ q[ =
_
q(q) q
[q dq .
The left-hand side above is simply
q
[ q[ = q
[ = q
(q
) ,
and this should be equal to
_
q(q) q
[q must behave
like a Diracs delta function (q q
), so that
_
q(q) q
[q
. .
(qq
)
dq =
_
q(q)(q q
) dq = q
(q
) .
Hence, the eigenstates [q belong to a generalized Hilbert space and are normalized
in the distributional sense of the Diracs delta function.
Let us turn to the momentum operator p. It is often said that commutation rela-
tion (3.10) implies p can be realized as
p = i
d
dq
.
In fact, if we apply the commutation relation (3.10) to (q) and we use this represen-
tation for p as a dierential operator, Eqn. (3.10) is actually fullled. Let us check
this fact:
[ q, p](q) = i
_
q
d(q)
dq
dq(q)
dq
_
= i(q) .
Total derivative was used having in mind the one-dimensional case; in multi-dimensional case
partial derivatives must be used in writing the representation above.
84 Chapter 3. Postulates of Quantum Mechanics
Let us now work out the matrix elements of p in the position representation. First,
the matrix elements of the commutator [ q, p] are evaluated between any two position
eigenstates [q and q
[:
q
[[ q, p][q = q
[ q p[q q
[ p q[q = (q
q) q
[ p[q .
Next, Eqn. (3.10) is used to get
q
[[ q, p][q = q
[i[q = i q
[q = i(q q
) .
By matching the two previous expressions we are led to
(q
q) q
[ p[q = i(q q
) . (3.11)
In other words, q
as a delta function (q q
), that is
_
(q
q) q
[ p[q dq
= i
_
(q q
) dq
.
The latter equation is fullled if
q
[ p[q = i
d
dq
(q q
) .
In fact, integrating by parts yields
_
(q
q)
_
i
d
dq
_
(q q
) dq
= i
_ _
d
dq
(q
q)
_
(q q
) dq
+
(q
q)(i)(q q
=
_
i(q q
) dq
.
Strictly speaking instead of p = i d/ dq one should more properly say that the
representation of p in the basis of position eigenstates [q is given by the matrix
elements
q
[ p[q = i
d
dq
(q q
) .
3.4. Position and momentum representations 85
From this equation it is easy to nd how p acts on a state [ in the position repre-
sentation:
q
[ p[ = q
[ p
_
[q q[ dq
=
_
(i)
_
d
dq
(q q
)
_
(q) dq
= i
d
dq
_
(q q
)(q) dq
= i
d(q
)
dq
.
This claries the meaning of the statement that p acts on (q) as i d/ dq.
Momentum representation
Instead of diagonalizing q, we now turn to the problem of diagonalizing p, which
we shall prove in an exercise to be Hermitian and thus to have a complete basis of
eigenstates:
p [p
= p
[p
,
_
[p
[ dp
= 11 ,
where 11 denotes as usual the identity operator.
We are interested in how the momentum eigenstates [p are represented in the basis
of position eigenstates [q, that is, what are the matrix elements q[p. We have
q
[ p[p
= p
[p
,
and by inserting the resolution of the identity formula
_
[q q[ dq = 1 (completeness
of position eigenstates) in the left-hand side above we get
q
[ p[p
=
_
q
[ p[q q[p
dq = p
[p
.
By using the representation of p derived previously, the equation above can be rewrit-
ten as
i
_
d
dq
(q q
) q[p
dq = p
[p
.
86 Chapter 3. Postulates of Quantum Mechanics
The derivative with respect to q
(q q
) q[p
dq = i
d
dq
_
(q q
) q[p
dq = i
d
dq
[p
,
yielding the following dierential equation for the matrix elements q
[p
:
i
d
dq
[p
= p
[p
whose solution is
q
[p
= C(p
) e
i
,
where C(p
[q = (q q
) .
Because of the completeness of momentum eigenstates we have
q
[q =
_
q
[p p[q dp =
_
[C(p)[
2
e
i
(q
q)p
dp
and this should be equal to (q
) =
1
2
_
e
ip(q
q)
dp .
we get
[C(p)[ =
1
2
,
and thus
q[p =
1
2
e
i
qp
.
Homework 3.1. Prove that p
[p
= (p
).
Let us now check that
_
[p p[ dp = 11 . (3.12)
3.4. Position and momentum representations 87
Let us sandwich the above relation between q
[p p[q = q
[11[q = (q q
) .
The left-hand side can be rewritten as
_
dp
1
2
e
i
p
1
2
e
qp
=
1
2
_
dp e
i
p(q
q)
=
1
2
_
d
_
p
_
e
i(
p
)
= (q q
) ,
and this proves Eqn. (3.12).
Fourier transform
It is possible to prove that the position and the momentum representations of a given
abstract state [ are one the Fourier transformed of the other respectively.
By using the completeness of the momentum eigenstates we can expand [ in the
usual way
[ =
_
[p p[ dp ,
where p[ can be viewed as a function of p, the so-called momentum-space wave
function
(p) p[ .
The position-space wave function had already been dened as
(q) = q[ .
By using the resolution of the identity and by applying the formula for the matrix
elements q[p found in the previous section, we get
(q) = q[ =
_
q[p p[ dp
=
1
2
_
e
i
qp
(p) dp .
88 Chapter 3. Postulates of Quantum Mechanics
Setting = 1 yields
(q) =
1
2
_
e
iqp
(p) dp ,
which is exactly the Fourier transform. Of course, if is not set to one, the correct
formula is the one given before.
Let us now look for the momentum representation of the position operator q. We
note that in the momentum representation p is represented as a multiplicative oper-
ator, i.e.:
p [p = p [p .
Let us consider the commutation relation (3.10) as usual. The matrix element of the
commutator between two momentum eigenstates p[ and [p
is
p[[ q, p][p
= p[ q p[p
p[ p q[p
= (p
p) p[ q[p
.
Because of the commutator relation, this is also equal to
p[[ q, p][p
= p[i[p
= i p[p
= i(p
p) ,
thus
p[ q[p
= i
d
dp
(p
p) ,
which means that the operator q acts as +i
d
dp
in the momentum representation.
3.5 Time-dependent Schr odinger equation: formal solution and time-
evolution operator
The time-dependent Schr odinger equation is (formally)
i
d [(t)
dt
=
H[(t) . (3.13)
3.5. Solution of the time-dependent Schr odinger equation 89
The formal solution is given by
[(t) = e
Ht
[(0) . (3.14)
The operator-valued exponential function is understood here in the sense we have
used to dene a function of an operator, namely, by its action on the eigenstates of
the original operator:
e
Ht
[E
n
= e
Ent
. .
This is the exponential of a number
[E
n
Using this in (3.14) we immediately get (3.13). Another denition of the exponential
of an operator is the following one:
e
Ht
=
n=0
_
i
Ht
_
n
1
n!
.
Note that this is not a Taylor expandion but a converging series for all value of t.
Using this expression we can now prove that the solution (3.14) formally satises the
time-dependent Schr odinger equation. In fact, inserting the (3.14) into the left-hand
side of Eqn. (3.13) yields
i
d
dt
_
n=0
_
i
Ht
_
n
1
n!
_
[(0) = i
n=1
1
n!
_
i
Ht
_
n1
n
_
i
H
_
[(0)
= i
n=1
1
(n 1)!
_
i
Ht
_
n1
_
i
H
_
[(0)
=
n=1
i
(n 1)!
_
t
_
n1
(i
H)
n
[(0)
=
n=0
i
n!
_
t
_
n
(i
H)
n+1
[(0) , (3.15)
This holds for time-independent Hamiltonian operators. Hamiltonian operators which depend
explicitly on time may occur, for example, when dealing with time-varying electric or magnetic
elds. We shall return on this when we shall discuss the interaction picture and the time-dependent
perturbation theory.
90 Chapter 3. Postulates of Quantum Mechanics
while introducing the solution into the right-hand side of the equation yields
n=0
_
i
Ht
_
n
1
n!
[(0) =
n=0
_
t
_
n
i
n!
(i
H)
n+1
[(0) ,
which is equal to (3.15).
Now we turn to the problem of how the formal solution (3.14) can be practically
used to nd [(t) once [(0) is known. Since the Hamiltonian
H is an Hermitian
operator, its eigenstates [E
n
,
H[E
n
= E
n
[E
n
,
form a complete orthornormal set and can be used to expand any given [(0) ac-
cording to the formula
[(0) =
n
c
n
[E
n
, c
n
= E
n
[(0) .
Now, we apply Eqn. (3.14):
[(t) = e
Ht
n
c
n
[E
n
n
c
n
e
Ht
[E
n
n
c
n
e
Ent
[E
n
n
c
n
(t) [E
n
where
c
n
(t) = c
n
e
Ent
= E
n
[(0) e
Ent
.
So, the probability of nding E
n
by performing an energy measurement at the time
t is given by
P
En
(t) = [c
n
(t)[
2
.
However, it is easily seen that
[c
n
(t)[
2
= c
n
(t)c
n
(t) = c
n
e
+
i
Ent
e
Ent
. .
=1
c
n
= [c
n
[
2
.
3.6. Density Matrix 91
This means the probability above does not change with time.
Another important concept is the so-called time-evolution operator
U(t) dened by
U(t) = e
Ht
.
Since
H is Hermitian,
U(t) is unitary:
(t) = e
i
Ht
=
U
1
(t) .
In fact,
(t)
U(t) = e
i
Ht
e
Ht
= 11 ,
being 11 the identity operator.
3.6 Density Matrix
One of the postulate of Quantum Mechanics tells us that to prepare a system we need
a complete set of commuting observables. The string of eigenvalues, each associated
to a dierent observables, will uniquely characterize the state of the system. What
about if we do not have a complete set of observables or if we have many many
particles and we cannot perform measurements on all of them because they are too
many. In both cases we will have some probabilities p
1
, p
2
, p
m
that the particle
is in one the normalized states 1[ , 2[ m[ . If we now ask which is the average
value of an observable
O , as a result we will get:
O =
i
p
i
i[
O[i (3.16)
This statistical mixture can be obtained by introducing what is called the density
operator or matrix, dened as:
i
p
i
[i i[ (3.17)
92 Chapter 3. Postulates of Quantum Mechanics
where the i[ are normalized states but not necessarily orthogonal to each other, and
the p
i
are probabilities that sum up to one, so:
p
i
0 ,
i
p
i
= 1
The average value dened in Eqn.(3.16) can then be written as
O = Tr(
O) (3.18)
This is easy to prove, in fact from
Tr(
O) =
i
p
i
Tr([i i[
O) (3.19)
in order to get (3.18) we have just to show that:
Tr([i i[
O) = i[
O[i
This can be proved easily considering that [i i[
P
i
is a projection operator, so
P
2
i
=
P
i
. If we then use the cyclic property of the trace we have:
Tr [i i[
O = Tr
P
2
i
O = Tr
P
i
P
i
The last expression can be written as
Tr [i i[
O[i i[ = i[
O[i Tr [i i[ = i[
O[i Tr
P
i
.
We know that
Tr
P
i
= Tr [i i[ = i[i = 1
The last step comes from the fact that the trace operation is the sum of the diagonal
elements of an operator, so:
Tr [i i[ =
j
j[i i[j = i[i i[i = i[i
3.6. Density Matrix 93
where we have choosen a basis of orthogonal vectors j[ of which i[ is just one of
them. We could choose this basis because the trace in invariant under a unitary
change of basis:
Tr
O = TrU
OU
= TrU
U
O = TrU
1
U
O = Tr
O
Another property of the density operator can be obtained if we choose in Eqn.(3.18)
O = 11. We get:
Tr = 1
The density matrix will help us not only in getting average values but also in
getting probability distributions. For example if
P
D
is the projector operator upon
the subspace spanned by the eigenvectors of
O located in a certain domain D of the
spectrum of
O, then the probability T
D
that the measurement of
O belongs to D is
T
D
=
m
p
m
m[
P
D
[m = Tr(
P
D
) (3.20)
In particular, using Eqn.(3.20), the probability of nding the system in the quantum
state represented by the Hilbert state element [ is given by
T
= Tr( [ [) = [ [
So, given , we can calculate all average values and statistical distributions of mea-
surements of observables.
In general one needs to introduce a density matrix when the system cannot be
specied by a single state [. If this instead this were the case, then would be:
= [ [
and, if was normalized, we would have
2
= (3.21)
94 Chapter 3. Postulates of Quantum Mechanics
These kinds of density matrices are called pure. In this case we could use the state
itself [ instead of the density matrix.
In general anyhow the is not pure and does not have the form above but the
following one:
=
p
i
[i i[ (3.22)
and as a consequence
2
,= .
In this case we say that the density matrix is a mixed one because it gives a statistical
mixture which cannot be specied by a single state.
We can say that the density matrix is a unifying formalism to treat both systems
that can be specied by a single state and those which are statistical mixture of many
states.
Several properties of the density matrix are of relevance and we will list them here.
From the fact that the p
i
are positive probabilities which sum up to one: p
i
0
and
i
p
i
= 1, we can easily derive that p
2
i
1 and from this we can prove that:
2
1
Other properties of are the following: rst it is hermitian (because the p
i
are real)
= ,
second, because the probabilities sum up to one and the states are normalized, its
trace is one:
Tr = 1
and third it is a positive semidenite operator:
[ [ 0 (3.23)
3.6. Density Matrix 95
This last property can be easily proved using Eqn. (3.22). In fact using that
expression we have [ [ =
i
p
i
[ < [i > [
2
and from the fact that the p
i
0
and [ < [i > [
2
0 we get Eqn.(3.23).
Let us now attack the problem of whether a physical system is specied by one and
only one density matrix or not. For the states we know that there is not a one to
one correspondence: any state up to a global phase specify the same physical system.
This ambiguity on the states is not transferred to the density matrix, in fact :
=
i
p
i
[i i[
i
p
i
e
i
i
[i i[ e
i
i
=
i
p
i
[i i[ .
Anyhow the has a lot of other freedoms. It is possible, for example, to prove that
the set of states [i and [m generate the same density matrix if (and only if) they
are transformed into each other by a unitary matrix U
i,m
:
[i =
m
U
i,m
[m
The number of states [i and [m must be the same and normalized. We will leave
the proof of this theorem as an homework. Basically it will turn out that:
=
i
p
i
[i i[ =
m
q
m
[m m[
where the q
m
are dened by:
p
i
[i =
m
U
im
q
m
[m
3.6.1 Time Evolution of the Density Matrix
As the density matrix is built out of kets and bras, the time evolution is obtained by
making the time-evolution of these components.
Suppose
96 Chapter 3. Postulates of Quantum Mechanics
(0) =
p
i
[i i[
then
(t) =
p
i
[i, t i, t[ =
i
p
i
U(t) [i i[
(t) =
U(t) (0)
(t) (3.24)
We have seen previously in this chapter the expression of the operator of time
evolution
U(t) which is:
U(t) = e
i
H
t
and so Eqn.(3.24) becomes:
(t) = e
i
H
t
(0)e
i
H
t
.
Taking the derivative with respect to t of the expression above we get the equation
of motion of :
t
=
i
[ ,
H]
which is called von Neumann equation and is the quantum analogue of the Liouville
equation.
3.7 Problems and Solutions
Problem 3.1. By using the derivation of the uncertainty relation presented in
this chapter , look for the states of minimal uncertainty which satisfy
q p =
1
2
. (3.25)
3.7. Problems and Solutions 97
Solution. The equality holds when the following two conditions, encountered in the vari-
ous steps of the Heisenberg uncertainty theorem, are both satised:
1. the Schwarz inequality becomes an equality. This is true if and only if the two vectors
in the inequality are proportional to each other:
[
A
A] [ = k[
B B] [ ,
with some constant k.
2. the complex number in the right-hand side of Eqn. (3.9) is purely imaginary:
[(
A
A)(
B
B)[ = i ,
with a real constant.
Thus,
k
[(
B
B)(
B
B)[
. .
|(
B
B)||
2
= i .
This means k must be purely imaginary: k = i, with real. This is due to the fact that the
expectation value on the left-hand side above is real: in fact, it is nothing but the square of
the norm |(
B
B) [ |. So we have
_
A
A
_
[ = i
_
B
_
[ .
Now, we choose
A = q and
B = p = i
d
dq
and we image that [ is already realized on q
as a wave function (q) in the /
2
space. The previous equation becomes
[ q q] (q) = i
_
i
d
dq
p
_
(q) ,
that is,
d(q)
dq
+ [ q i p] (q) q(q) = 0 ,
from which it follows
d
=
q +i p q
dq ,
98 Chapter 3. Postulates of Quantum Mechanics
which can be integrated to yield
log =
1
_
q
2
2
+ (i p q)q
_
+ const.
Finally, by taking the exponential of both sides we get
(q) = Aexp
_
q
2
2
q
q +
i p
q
_
= Aexp
_
(q q)
2
2
+i
p
q
q
2
2
_
.
A can be found by normalizing the wave function. Of course, should be less than zero in
order to be normalizable. The result is
(q) =
4
2
[[
exp
_
(q q)
2
2 [[
+i
p
q
_
. (3.26)
Problem 3.2. Determine the momentum probability distribution of the energy
eigenstates of an harmonic oscillator whose mass and frequency are m and
respectively.
Solution. The Hamiltonian is
H =
p
2
2m
+
1
2
m
2
x
2
, (3.27)
and the energy eigenfunctions in the position representation are
n
(x) =
_
m
_1
2
1
2
n/2
n!
e
mx
2
2
H
n
_
x
_
m
_
, (3.28)
where
H
n
() = (1)
n
e
2 d
n
e
2
d
n
are the Hermite polynomials.
3.7. Problems and Solutions 99
Of course, we could Fourier transform
n
(x) to obtain the momentum-space wave function
(p) whose square modulus is nothing but the desired probability density function. However,
it seems easier to write
H in the momentum representation:
H
p
=
p
2
2m
2
2
m
2
d
2
dp
2
.
The eigenvalue problem for the operator
H
p
reads
_
2
2
m
2
d
2
dp
2
+
p
2
2m
_
n
(p) = E
n
n
(p) ,
which can be written as
d
2
n
(p)
dp
2
+
2
2
m
2
_
E
n
p
2
2m
_
n
(p) = 0 . (3.29)
The position-space stationary Schr odinger equation had the form
d
2
n
(x)
dx
2
+
2m
2
_
E
n
1
2
m
2
x
2
_
n
(x) = 0 . (3.30)
By performing the replacement
p
m
with x
_
m
Eqn. (3.29) formally becomes equivalent to Eqn. (3.30) whose solution are Eqn. (3.28). By
the same substitution into Eqn. (3.28), and properly normalizing, we get:
n
(p) =
1
2
n
n!
m
e
p
2
2m
H
n
_
p
m
_
.
Problem 3.3. Let
H =
p
2
2m
+ V ( q)
be the Hamiltonian of a one-dimensional quantum system having discrete spec-
trum with eigenstates [n and eigenvalues E
n
. By using the commutation relation
100 Chapter 3. Postulates of Quantum Mechanics
Eqn. (3.10) between position coordinate q and momentum p, prove the following
relation:
n
(E
n
E
0
) [n[ q[0[
2
=
2
2m
,
where [0 and E
0
are the ground state and its corresponding eigenvalue respec-
tively.
Solution. By direct evaluation we have
[
H, q] =
1
2m
[ p
2
, q] =
i
m
p ,
hence
_
[
H, q], q
_
=
i
m
[ p, q] =
2
m
,
from which it follows
_
n
_
[
H, q], q
_
n
_
=
2
m
. (3.31)
The left-hand side in the latter expression is also equal to
_
n
_
[
H, q], q
_
n
_
= n[
H q
2
2 q
H q + q
2
H[n
= 2E
n
n[ q
2
[n 2 n[ q
H q[n
=
m
2E
n
n[ q[m m[ q[n
m,l
2 n[ q[m m[
H[l l[ q[n
=
m
2E
n
[n[ q[m[
2
m,l
2 n[ q[m E
m
l,m
l[ q[n
=
m
_
2E
n
[n[ q[m[
2
2E
m
[n[ q[m[
2
_
= 2
m
(E
n
E
m
) [n[ q[m[
2
.
By taking n = 0 and comparing with the right-hand side of Eqn. (3.31) we get
m
(E
m
E
0
) [m[ q[0[
2
=
2
2m
.
3.7. Problems and Solutions 101
Problem 3.4. Let
f be an Hermitian operator having a discrete spectrum with
N dierent eigenvalues. Prove that
f
N
can be expressed as a linear combination
of the operators 1,
f, . . . ,
f
N1
.
Solution. We dene the following operator:
G
N
i=1
_
f f
i
_
= (
f f
1
)(
f f
2
) (
f f
N
) .
Let
f be hypermaximal, so that its eigenstates [f
i
form a complete orthonormal basis and
any state [ can be expanded as
[ =
k
a
k
[f
k
, a
k
= f
k
[ .
Now, let us apply
G to [:
G[ =
_
(
f f
1
)(
f f
2
) (
f f
N
)
_
_
N
k=1
a
k
[f
k
_
.
We note that
G[ =
_
(
f f
1
)(
f f
2
) (
f f
N
)
_
_
a
N
[f
N
+
N1
k=1
a
k
[f
k
_
=
_
(
f f
1
)(
f f
2
) (
f f
N
)
_
a
N
[f
N
+
_
(
f f
1
)(
f f
2
) (
f f
N
)
_
_
N1
k=1
a
k
[f
k
_
=
_
(
f f
1
)(
f f
2
) (
f f
N
)
_
_
N1
k=1
a
k
[f
k
_
.
This is so because
(
f f
N
)a
N
[f
N
= a
N
f [f
N
a
N
f
N
[f
N
= 0.
102 Chapter 3. Postulates of Quantum Mechanics
By considering the next terms in the sum, we get by induction
G[ = 0 ,
that is,
G = 0 (
G is the null operator). Thus,
G = 0 =
f
N
i=1
f
i
f
N1
+
1
2
i,j
f
i
f
j
f
N2
+ + (1)
N
N
i=1
f
i
,
and as a consequence
f
N
=
N
i=1
f
i
f
N1
+
1
2
i,k
f
i
f
k
f
N2
+ + (1)
N
N
i=1
f
i
.
Problem 3.5. A particle of mass m moves in a innite potential well of width
a. At the time t = 0 the particle was prepared in the state given by the wave
function
(x, t = 0) = Asin
3
x
a
. (3.32)
The questions are:
1. Find the wave function (x, t) at any later time t > 0.
2. Find the time T at which the particle turns back to its initial state.
Solution. We recall the energy levels of the one-dimensional innite potential well are
E
n
=
2
2
(n + 1)
2
2ma
2
, n = 0, 1, 2, . . .
The corresponding normalized wave eigenfunctions are
n
(x) = x[E
n
=
_
2
a
sin
(n + 1)x
a
.
3.7. Problems and Solutions 103
As we have just seen, the state [(t) at any later time is found by applying the unitary
time-evolution operator
U(t) = exp(i
Ht/) to the initial state [(0). For this proce-
dure to work properly we have to determine the expansion coecients c
n
of [(0) on the
Hamiltonian eigenstates [E
n
whose wave function x[E
n
is given above:
[(0) =
n
c
n
[E
n
, c
n
= E
n
[(0) .
Inserting the identity
_
[x x[ dx = 1 yields
c
n
= E
n
[(0) =
_
E
n
[x
. .
n
(x)
x[(0)
. .
(x,t=0)
dx .
The integral involved in the calculation of c
n
must in principle be solved for each n:
c
n
=
_
2
a
A
_
a
0
sin
(n + 1)x
a
sin
3
x
a
dx .
There is however a faster way to get the correct result, that is, we can manipulate directly
the initial wave function Eqn. (3.32) in order to identify the components. We can write
(x, t = 0) = Asin
3
x
a
= Asin
x
a
_
1 cos
2
x
a
_
= Asin
x
a
A
1
2
_
1 + cos
2x
a
_
sin
x
a
=
A
2
sin
x
a
A
2
sin
x
a
cos
2x
a
=
A
2
sin
x
a
A
4
_
sin
3x
a
sin
x
a
_
=
3
4
Asin
x
a
1
4
Asin
3x
a
The trigonometric identity
cos p sin q =
1
2
[sin(p +q) sin(p q)]
has been used in the last steps above. In this way we get the initial state as a sum of x[E
0
Ht
(x, t = 0)
= e
Ht
n=0
c
n
n
(x) c
n
=
_
a
0
n
(x)(x, t = 0) dx ,
=
n=0
c
n
e
Ent
n
(x)
= c
0
e
E
0
t
0
(x) +c
2
e
E
2
t
2
(x) ;
where, c
0
=
3
4
A
_
a
2
and c
2
=
1
4
A
_
a
2
. The nal result is
(x, t) =
A
4
_
2
a
_
3 sin
x
a
exp
_
i
2
t
2ma
2
_
sin
3x
a
exp
_
9i
2
t
2ma
2
__
,
which can also be written as
(x, t) =
A
4
_
2
a
exp
_
i
2
t
2ma
2
__
3 sin
x
a
exp
_
i
4
2
t
ma
2
_
sin
3x
a
_
.
For the wave function (x, T) at a later time T > 0 to be equal to (x, t = 0) the
following condition must hold:
exp
_
i
2
T
2ma
2
_
= exp
_
4i
2
T
ma
2
_
= 1 ,
that is
2
T
2ma
2
= n(2) , 4
2
T
ma
2
= 8n(2) ,
for some integer n. By taking n = 1 (we nd the smallest period T)
2
4T
min
ma
2
= 16 ,
from which it follows
T
min
=
16ma
2
4
2
=
4ma
2
.
3.7. Problems and Solutions 105
Problem 3.6. The state of a free particle at initial time t = 0 is given by the
following Gaussian wave function:
(x, t = 0) = Aexp
_
x
2
2a
2
+
imv
0
x
_
.
Calculate the wave function (x, t) at the time t and also the expectation value
and the mean square deviation of the position operator x, that is, (t)[ x[(t)
and (t)[ x
2
[(t) respectively.
Solution. We need to recall the following Gaussian integrals:
_
0
exp
_
ax
2
_
dx =
1
2
_
a
,
_
+
exp
_
ax
2
+bx
_
dx =
_
a
exp
b
2
4a
,
_
0
exp
_
ax
2
_
x
2k
dx =
1 3 (2k 1)
2
k+1
_
a
2k+1
,
_
0
exp
_
ax
2
_
x
2k+1
dx =
k!
2a
k+1
,
_
+
dx
1
_
+
dx
N
exp
_
a
ij
x
j
x
i
_
=
N
2
[det A
ij
]
1/2
.
The Hamiltonian of a free particle is
H =
p
2
2m
,
its eigenstates are
H[p =
p
2
2m
[p ,
thus
H does not form a complete set since we still have the degeneracy [p. For this
reason, we shall use instead the eigenstates of p:
p [p = p [p .
We expand the initial state [(0) on the basis of the momentum eigenstates:
[(0) =
_
C(p) [p dp , C(p) = p[(0) .
106 Chapter 3. Postulates of Quantum Mechanics
It should be noted that we know how the momentum eigenstates do evolve in time under
the free Hamiltonian. C(p
) is given by
p
[(0) =
_
p
[x x[(0) dx
=
_
1
2
e
ip
(x, 0) dx
=
A
2
_
exp
_
x
2
2a
2
+i
mv
0
x
i
p
_
dx .
The latter is a Gaussian integral, which can be evaluated exactly to yield:
C(p) = aAexp
_
1
2
(mv
0
a pa)
2
_
.
By applying the time-evolution operator, we get
(x, t) = e
Ht
(x, t = 0)
= e
Ht
_
C(p)
p
(x)
. .
x|p
dp
=
aA
2
_
exp
_
ip
2
t
2m
+i
px
a
2
2
2
(mv
0
p)
2
_
dp
= A
_
1 +
it
ma
2
_
1
2
exp
_
_
ma
2
2
(x v
0
t) +
3
x
2
t +ia
4
m
2
v
0
(2x v
0
t)
_
2m(a
4
2
+
t
2
2
m
2
)
_
1
_
A is determined by normalizing the wave function:
_
[(x, t)[
2
dx = 1. The reason for
which we require (x, t) to be normalized is that we are interested in nding mean values
and mean square deviations, thus the wave function must be normalized. We have
[(x, t)[
2
=
A
2
_
1 +
2
t
2
m
2
a
4
exp
_
_
(x v
0
t)
2
a
2
_
1 +
t
2
2
m
2
a
4
_
_
_
.
By imposing
_
[(x, t)[
2
dx = 1 one nds
A
2
= (a
2
)
1
2
.
The expectation value of the position operator x on the evolved state [(t) is
(t)[ x[(t) =
_
x[(x, t)[
2
dx .
3.7. Problems and Solutions 107
Again, this is nothing but a Gaussian integral which can be calculate explicitly. The result is
v
0
t. It should be noted at this point that the quantum expectation value obey the classical
laws of motion. This is a more general result which we shall encounter again later: the
so-called Ehrenfest theorem.
The mean square deviation of the position operator x on the evolved state [(t) can be
calculated in the same manner using the formulas for the Gaussian integrals and the result
is
a
2
2
_
1 +
t
2
2
m
2
a
4
_
.
Note that at t = 0 the mean square deviation is a/
H =
1
2m
p
2
+
1
2
m
2
q
2
. (4.1)
Let us recall that the position and momentum operators satisfy the commutation
relation
[ q, p] = i .
To simplify the notations, let us introduce the following operators:
and
Q
_
m
_1
2
q ,
P
_
1
m
_1
2
p ,
so that
H =
1
2
_
P
2
+
Q
2
_
,
with
[
Q,
P] = i .
110 Chapter 4. Harmonic Oscillator in Operatorial Form.
Let us also dene the following other operators:
a
1
2
2
_
Q+ i
P
_
,
a
1
2
2
_
Qi
P
_
.
A direct computation shows that :
[ a, a
] = 1
and
H =
1
2
_
a a
+ a
a
_
.
By introducing the operator
N a
a ,
we can rewrite the operator
H as
H =
N +
1
2
. (4.2)
It is easy to work out the commutation rules of a and a
with
N:
N a = a
_
N 1
_
, (4.3)
N a
= a
_
N + 1
_
. (4.4)
We are now in a position to prove the following theorem.
THEOREM:
If
N [ = [ ,
i.e., if [ is an eigenvector of
N with corresponding eigenvalue , then:
a) 0;
b) if = 0, then a [ = 0;
c) if ,= 0, then a [ is a non-zero vector whose norm is given by [ and which
4.1. Harmonic Oscillator in Operatorial Form. 111
is an eigenvector of
N with eigenvalue 1; a
N [ = [ ,
with [ > 0. So,
[
N[ = [ a
a[ = | a [|
2
= [ . (4.5)
and
_
_
a
[
_
_
2
= [ a a
[ = [
N + 1[ = ( + 1) [ . (4.6)
Now, by denition all the vectors in the Hilbert space have non-negative norms and
the only vector having zero norm is the null vector. From Eqn. (4.5) we get that,
since | a [ |
2
is the norm of a state, it is a positive number as well as [, so we
have 0. This proves proposition a).
If = 0, then | a [ | = 0, and consequently a [ = 0, which proves proposition b).
If ,= 0, then a [ is a vector with non-zero norm given in (4.5). Using Eqn. (4.3)
we get
N a [ = a
_
N 1
_
[ = ( 1) a [ .
This proves part of point c). From (4.6) we get the norm of a
[. Since 0 the
minimum value of this norm is 1 [ when = 0, and even this is a norm dierent
from zero. Now let us take into account Eqn. (4.4):
N a
[ = a
_
N + 1
_
[ = ( + 1) a
[ .
This relation tells us that a
[ is eigenvector of
N with eigenvalue + 1. This
completes the proof of the theorem. Via repeated applications of the operator a on
112 Chapter 4. Harmonic Oscillator in Operatorial Form.
the eigenvector [ we get that the states :
a [ , a
2
[ , . . . , a
p
[ ,
are eigenvectors of
N with eigenvalues
1 , 2 , . . . , p .
From proposition a) of the previous theorem, these eigenvalues must be positive or
zero, so there must exists a certain positive integer p for which we have p = 0,
i.e., = p. Therefore, we get the important result that the generic eigenvalue is
zero or a positive integer because p is integer. If p were not zero but the smallest
non-integer positive number , then we would always be able to construct the state
a [ p which would be an eigenvector of
N with eigenvalue ( p1) , whose norm
would be proportional to p 1 and negative, but this cannot happen in a Hilbert
space. This proves that the eigenvalues are given by all the (non-negative) integers.
From this and Eqn.(4.2) we get that, calling by n a positive integer, the energy
levels of the original Hamiltonian are:
E
n
= (n +
1
2
)
Now,
N is a Hermitian operator, so its eigenvectors [0, [1, . . . , [n, . . . , make up a
basis for the Hilbert space. There are no degeneracies and they make up a complete
set. From the relation
[ a
a[ = | a [|
2
= [ ,
we get that a [ is normalized if = 1 and [ = 1. Then, from
[ a a
[ = [
N + 1[ = ( + 1) [ ,
we get that
_
_
a
[1
_
_
2
= 2 1[1 = 2
4.1. Harmonic Oscillator in Operatorial Form. 113
so
_
_
a
[1
_
_
=
2 .
But we already know that a
[1 =
2 [2 ,
and in general we shall have
a
[n = (n + 1)
1/2
[n + 1 .
Analogously, we can determine the normalization coecient in the relation a [n
[n 1. If [n is normalized then
n[ a
a[n = | a [n|
2
= nn[n = n ,
and so
| a [n| =
n , a [0 = 0 .
We can conclude that the vectors
[n = (n!)
1
2
a
n
[0 ,
are eigenstates of
N:
N [n = n[n ,
and they make an orthonormal basis:
n[n
=
n,n
.
114 Chapter 4. Harmonic Oscillator in Operatorial Form.
To check the orthogonality, let us consider for example 1[2:
1[2 =
1
2
0[ a a
[0
=
1
2
0[
_
a
a + 1
_
a
[0
=
1
2
0[ a
a a
[0 +
1
2
0[ a
[0
=
1
2
0[ a
a a
[0
=
1
2
0[ a
_
1 a
a
_
[0
=
1
2
0[ a
[0
1
2
0[ a
a[0 = 0 .
The same can be easily proved for all states labelled by dierent integers.
The fact the states are labelled by positive integer agrees with the result we got
in chapter 1) where we solved explicitly the Schroedinger equation associated to the
harmonic oscillator and did not use operator methods like we did here. The wave-
functions we obtained in chapter 1) are nothing else and the x-representation of the
states [n,i.e:
osc.
n
(x) = x[n
Chapter 5
ANGULAR MOMENTUM QUANTIZATION
In classical mechanics the angular momentum of a particle is dened as
L = x p ,
where the symbol indicates the vector product. In components, we get
L
x
= yp
z
zp
y
, L
y
= zp
x
xp
z
, L
z
= xp
y
yp
x
.
By using the correspondence rules, the three components of the angular momentum
operator in quantum mechanics are:
L
x
= y p
z
z p
y
= i
_
y
z
z
y
_
,
L
y
= z p
x
x p
z
= i
_
z
x
x
z
_
,
L
z
= x p
y
y p
x
= i
_
x
y
y
x
_
.
We know that observables in quantum mechanics must be Hermitian operators. Are
L
x
,
L
y
and
L
z
actually Hermitian? It is easy to check that in fact they are.
By using the fundamental commutation rules among position and momentum ob-
servables, namely,
[ x
i
, p
j
] = i
i,j
,
one can immediately calculate the corresponding commutation rules among the vari-
116 Chapter 5. Angular momentum quantization
ous components of the angular momentum operator, for example
[
L
x
,
L
y
] = [ y p
z
z p
y
, z p
x
x p
z
]
= [ y p
z
, z p
x
x p
z
] [ z p
y
, z p
x
x p
z
]
= y[ p
z
, z] p
x
[ z p
y
, ( x p
z
)]
= i y p
x
+ x[ z, p
z
] p
y
= i ( x p
y
y p
x
)
= i
L
z
.
The other commutators are:
[
L
z
,
L
x
] = i
L
y
, [
L
y
,
L
z
] = i
L
x
.
There is a compact notation to write these formulas, namely,
[
L
i
,
L
j
] = i
ijk
L
k
, (5.1)
where
ijk
is the totally-antisymmetric tensor whose value is
_
_
1 if even permutation of : x, y, z
0 if two indices are repeated
1 otherwise
Note that this result is the one obtained, via the correspondence principle, from the
classical case where the Poisson brackets between the components of the angular
momentum satisfy the relation
L
i
, L
j
=
ijk
L
k
. (5.2)
Homework 5.1. A) Check Eqn. (5.2). B) Find the commutator between
L
i
and q
j
,
where the latter are the position variables.
5.1. Spectrum of the angular momentum operators 117
5.1 Spectrum of the angular momentum operators
In this section we look for the spectrum of the angular momentum operators. We
shall see that the spectrum of the angular momentum is discrete in contrast with the
classical case. All the results of this section actually relies only on the commutation
rules (5.1) and thus they apply to every triplet of operators whose algebra satises
Eqn. (5.1). To emphasize this fact, we shall replace the notation
L
x
,
L
y
and
L
z
with
the more general
J
i
where i stands for x, y or z, or more generally i = 1, 2, 3. The
commutation relations read
[
J
i
,
J
j
] = i
ijk
J
k
. (5.3)
The angular momentum is not the only operator satisfying this algebra: also spin,
isotopic spin, etc. do obey the same commutation rules.
It should be noted at this point that the three operators
J
i
do not commute with
each other. Therefore, we cannot diagonalize two components simultaneously. But it
is easy to check that the following two operators do commute:
J
z
and
J
2
=
J
2
x
+
J
2
y
+
J
2
z
.
In fact,
[
J
z
,
J
2
] = [
J
z
,
J
2
x
+
J
2
y
+
J
2
z
]
= [
J
z
,
J
2
x
] + [
J
z
,
J
2
y
] + [J
z
,
J
2
z
]
. .
=0
= [
J
z
,
J
x
]
J
x
+
J
x
[
J
z
,
J
x
] + [
J
z
,
J
y
]
J
y
+
J
y
[
J
z
,
J
y
]
= i
_
J
y
J
x
+
J
x
J
y
J
x
J
y
J
y
J
x
_
= 0 .
In the same way it is also possible to prove that
[
J
x
,
J
2
] = 0 , [
J
y
,
J
2
] = 0 .
Since
J
z
and
J
2
are hermitian and commute, they can be diagonalized simultaneously
118 Chapter 5. Angular momentum quantization
and their common eigenstates can be chosen to be orthogonal:
_
_
_
J
2
[, m =
2
[, m
J
z
[, m = m[, m
(5.4)
The eigenvalues
2
and m have been written in this form in order for and m to
be dimensionless, since
J
2
and
J
z
have dimensions of
2
and respectively. [, m
denotes the (normalized) common eigenstate. For the time being, no restriction is
required for and m; of course, they must be real numbers since
J
2
and
J
z
are
Hermitian and thus their eigenvalues are real numbers.
The next step is to introduce the two auxiliary operators
J
(+)
J
x
+ i
J
y
,
J
()
J
x
i
J
y
.
These two operators are one the Hermitian conjugate of the other:
(+)
=
J
()
,
J
()
=
J
(+)
.
Moreover, it is easy to prove that the following relations hold:
[
J
z
,
J
(+)
] =
J
(+)
, (5.5a)
[
J
z
,
J
()
] =
J
()
, (5.5b)
[
J
(+)
,
J
()
] = 2
J
z
, (5.5c)
as well as the following ones
J
(+)
J
()
=
J
()
J
()
=
J
2
J
z
(
J
z
) , (5.6a)
J
()
J
(+)
=
J
(+)
J
(+)
=
J
2
J
z
(
J
z
+) . (5.6b)
Homework 5.2. Check Eqs. (5.5)(5.6).
Let us now apply Eqn. (5.6a) on a given eigenstate [, m satisfying the eigenvalue
condition (5.4):
J
(+)
J
()
[, m = (
J
2
J
2
z
+
J
z
) [, m =
2
[ m(m1)] [, m . (5.7a)
5.1. Spectrum of the angular momentum operators 119
Analogously, applying Eqn. (5.6b) on [, m yields:
J
()
J
(+)
[, m = (
J
2
J
2
z
J
z
) [, m =
2
[ m(m + 1)] [, m . (5.7b)
Applying instead Eqs. (5.5) on [, m yields:
J
z
J
(+)
[, m = (
J
(+)
J
z
+
J
(+)
) [, m = (m+ 1)
J
(+)
[, m , (5.8a)
J
z
J
()
[, m = (
J
()
J
z
J
()
) [, m = (m1)
J
()
[, m . (5.8b)
Together, these relations tell us that if [, m is eigenstate of
J
z
with eigenvalue m,
then
J
(+)
[, m and
J
()
[, m are eigenstates of
J
z
too with eigenvalues (m + 1)
and (m 1) respectively. However, in deriving this conclusion the following fact
should be stressed. Let us consider the norms of
J
(+)
[, m and
J
()
[, m; these can
be calculated directly by taking the scalar product of Eqs. (5.7) above with , m[.
Assuming as usual that [, m is correctly normalized, the result is
_
_
_
J
(+)
[, m
_
_
_
2
=
2
[ m(m+ 1)] , (5.9a)
_
_
_
J
()
[, m
_
_
_
2
=
2
[ m(m1)] . (5.9b)
Homework 5.3. Prove Eqs. (5.9).
Clearly, the terms on the right-hand sides of Eqs. (5.9) above must be non-negative
real numbers. In fact, by denition the norm of a vector is a non-negative real number,
moreover here we are considering the square of a norm. This imposes a constraint on
the values that and m can take. To explain this point, let us generalize Eqs. (5.7)-
(5.9) to the case in which
J
(+)
or
J
()
are applied multiple times on [, m. By
induction, one can prove that the following relations hold:
J
z
_
J
p
(+)
[, m
_
= (m+ p)
_
J
p
(+)
[, m
_
, (5.10a)
J
z
_
J
q
()
[, m
_
= (mq)
_
J
q
()
[, m
_
, (5.10b)
120 Chapter 5. Angular momentum quantization
where p, q = 1, 2, . . . are the number of times the operators
J
(+)
and
J
()
respectively
have been applied on [, m. Each time
J
(+)
(respectively,
J
()
) is applied on [, m
the corresponding eigenvalue of
J
z
increases or decreases respectively of one unit.
Let us now x a particular value of m, say m
0
. By applying many times the
operators
J
(+)
and
J
()
on [, m
0
one can generate the whole series of eigenstates of
the type [, m:
. . . ,
J
2
()
[, m
0
,
J
()
[, m
0
, [, m
0
,
J
(+)
[, m
0
,
J
2
(+)
[, m
0
, . . .
Of course, in general neither
J
p
(+)
[, m
0
nor
J
q
()
[, m
0
is guaranteed to be correctly
normalized even assuming that [, m
0
is normalized. In fact, we have seen that
the contrary happens in the case p = 1, see Eqs. (5.9). The associated normalized
eigenstates are indicated as:
. . . , [, m
0
2 , [, m
0
1 , [, m
0
, [, m
0
+ 1 , [, m
0
+ 2 , . . .
and the corresponding eigenvalues for
J
z
are given by
. . . , (m
0
2), (m
0
1), m
0
, (m
0
+ 1), (m
0
+ 2), . . .
The chain must stop in both directions. Let us now explain why. Eqs. (5.7) can be
generalized as follows:
J
(+)
J
()
_
J
q
()
[, m
0
_
=
2
[ (m
0
q)(m
0
q 1)]
_
J
q
()
[, m
0
_
,
J
()
J
(+)
_
J
p
(+)
[, m
0
_
=
2
[ (m
0
+ p)(m
0
+ p + 1)]
_
J
p
(+)
[, m
0
_
,
and Eqs. (5.9) become
_
_
_
J
(+)
[, m
0
+ p
_
_
_
2
= , m
0
+ p[
J
()
J
(+)
[, m
0
+ p
=
2
[ (m
0
+ p)(m
0
+ p + 1)] , (5.11a)
_
_
_
J
()
[, m
0
q
_
_
_
2
= , mq[
J
(+)
J
()
[, m
0
q
=
2
[ (m
0
q)(m
0
q 1)] , (5.11b)
5.1. Spectrum of the angular momentum operators 121
or, stated in another way,
_
_
_
J
p+1
(+)
[, m
0
_
_
_
2
=
2
[ (m
0
+ p)(m
0
+ p + 1)]
_
_
_
J
p
(+)
[, m
0
_
_
_
2
,
_
_
_
J
q+1
()
[, m
0
_
_
_
2
=
2
[ (m
0
q)(m
0
q 1)]
_
_
_
J
q
()
[, m
0
_
_
_
2
.
Homework 5.4. Verify Eqs. (5.10)(5.11).
Since the (square of) the norm of a vector in a Hilbert space must be (by denition)
a non-negative real quantity, both quantities appearing on the right-hand side of
Eqs. (5.11)
(m
0
+ p)(m
0
+ p + 1) and (m
0
q)(m
0
q 1) (5.12)
must be non-negative. Remember that p and q are integer because they indicates the
number of times we apply the
J
(+)
and
J
()
. Moreover, both p and q can be made
arbitrary large since we can apply
J
(+)
or
J
()
on [, m
0
as many times as we want,
and therefore for any given and m
0
it is always possible to choose two positive
integer numbers p and q large enough to let these quantities in (5.12) to become
negative. The only way out is to admit the existence of two (non-negative) integer
numbers q
0
and p
0
such that the chain stops, that is,
J
p
0
(+)
[, m
0
, = 0 ,
J
p
0
+1
(+)
[, m
0
= 0 ,
J
q
0
()
[, m
0
, = 0 ,
J
q
0
+1
()
[, m
0
= 0 .
This implies that
J
()
J
(+)
_
J
p
0
(+)
[, m
0
_
=
J
()
_
J
p
0
+1
(+)
[, m
0
_
= 0 ,
J
(+)
J
()
_
J
q
0
()
[, m
0
_
=
J
(+)
_
J
q
0
+1
()
[, m
0
_
= 0 ,
from which it follows that at some point the expression in (5.12) must become zero,
i.e.:
= (m
0
+ p
0
)(m
0
+ p
0
+ 1) , (5.13a)
= (m
0
q
0
)(m
0
q
0
1) . (5.13b)
122 Chapter 5. Angular momentum quantization
As the left-hand sides are equal, we get that:
(m
0
+ p
0
)(m
0
+ p
0
+ 1) = (m
0
q
0
)(m
0
q
0
1) .
Performing the calculation, this relation leads to
m
2
0
+ m
0
p
0
+ m
0
p
0
+ p
2
0
+ m
0
+ p
0
= m
2
0
+ q
2
0
2m
0
q
0
m
0
+ q
0
,
and thus
m
0
=
q
0
p
0
2
,
where q
0
and p
0
are by construction two non-negative integer numbers. From here we
get that m
0
must be either integer or half-integer. Now, if you dene
j
q
0
+ p
0
2
,
we get from one of the Eqs. (5.13) that
= j(j + 1) .
This means the eigenvalues of
J
2
are of the form j(j + 1)
2
, where j = (q
0
+ p
0
)/2
is an integer or half-integer: j = 0, 1/2, 1, 3/2, . . . Note also that for a given j the
smallest eigenvalue of
J
z
is given by m
min
, where
m
min
= m
0
q
0
=
_
q
0
p
0
2
_
q
0
= j ,
and analogously the largest eigenvalue of
J
z
is m
max
, where
m
max
= m
0
+ p
0
=
_
q
0
p
0
2
_
+ p
0
= j .
Therefore, the eigenvalues of
J
z
are of the form m where
m = j, j + 1, j + 2, . . . , j 2, j 1, j
. .
(2j+1) terms.
.
The fact that m and j can take only this discrete set of values is known as angular
momentum quantization.
5.1. Spectrum of the angular momentum operators 123
From now on we shall denote the eigenstates of
J
2
and
J
z
with [j, m:
J
2
[j, m =
2
j(j + 1) [j, m ,
J
z
[j, m = m[j, m .
Once j is xed (do remember that j must be integer or half-integer) the set of all
2j +1 eigenstates [j, m , where m = j, j +1, . . . , j 1, j, makes up what is called
a multiplet with xed j. We can move within the multiplet leaving unchanged the
eigenvalue of
J
2
and changing only the eigenvalue of
J
z
one unit at a time (in units of
) by using the operators
J
(+)
and
J
()
. For example, we can start from [j, j (with
the highest eigenvalue m = j of
J
z
) and move downwards by applying recursively
J
()
until the eigenstate [j, j is reached. Basically , we can construct the general state
[j, m as
[j, m = c
()
j,m
J
jm
()
[j, j .
The coecient c
()
j,m
accounts for the normalization. We can also start from [j, j
and move upwards through successive applications of
J
(+)
:
[j, m = c
(+)
j,m
J
j+m
(+)
[j, j .
Let us evaluate the coecients c
()
j,m
and c
(+)
j,m
. The simplest case is
J
()
[j, m = c
()
j,m
[j, m1 ,
J
(+)
[j, m = c
(+)
j,m
[j, m + 1 .
We have already calculated the norm of
J
(+)
[j, m and
J
()
[j, m. By using Eqs. (5.9),
and making the choice of having real and positive coecients, we nd
c
()
j,m
=
_
(j + m)(j m + 1) ,
c
(+)
j,m
=
_
(j m)(j + m + 1) .
where we have made the choice for the coecients c
()
j,m
to be real and positive. This
choice will be crucial to determine the relative phases among states of the same
124 Chapter 5. Angular momentum quantization
multiplets. Thus,
J
()
[j, m =
_
(j + m)(j m + 1) [j, m1 , (5.14a)
J
(+)
[j, m =
_
(j m)(j + m + 1) [j, m+ 1 . (5.14b)
By iterating the same procedure, we get the general formulas
[j, m =
1
jm
1
_
(2j)!
(j + m)!
(j m)!
J
jm
()
[j, j , (5.15a)
[j, m =
1
j+m
1
_
(2j)!
(j m)!
(j + m)!
J
j+m
(+)
[j, j . (5.15b)
Homework 5.5. Prove Eqs. (5.15).
From Eqs. (5.14) we can derive the matrix elements
j, m
J
()
[j, m =
_
(j + m)(j m + 1)
m
,m1
, (5.16a)
j, m
J
(+)
[j, m =
_
(j m)(j + m + 1)
m
,m+1
. (5.16b)
So the matrix elements of
J
()
and
J
(+)
are dierent from zero only for the elements
next to the main diagonal. From the matrix elements of
J
()
and
J
(+)
it is possible
to calculate also the matrix elements of
J
x
and
J
y
since
J
x
=
1
2
(
J
(+)
+
J
()
) ,
J
y
=
1
2i
(
J
(+)
J
()
) .
The matrix elements of
J
z
can be easily calculated, because
J
z
is diagonal in the
representation of its eigenstates:
j, m
J
z
[j, m = m
m,m
.
The explicit matrix representations of
J
x
,
J
y
and
J
z
are for j = 1/2:
J
x
=
2
_
_
0 1
1 0
_
_
,
J
y
=
2
_
_
0 i
i 0
_
_
,
J
z
=
2
_
_
1 0
0 1
_
_
;
5.2. Spherical Harmonics 125
for j = 1:
J
x
=
2
_
_
_
_
_
0 1 0
1 0 1
0 1 0
_
_
_
_
_
,
J
y
=
2
_
_
_
_
_
0 i 0
i 0 i
0 i 0
_
_
_
_
_
,
J
z
=
_
_
_
_
_
1 0 0
0 0 0
0 0 1
_
_
_
_
_
;
for j = 3/2:
J
x
=
2
_
_
_
_
_
_
_
_
0
3 0 0
3 0 2 0
0 2 0
3
0 0
3 0
_
_
_
_
_
_
_
_
,
J
y
=
2
_
_
_
_
_
_
_
_
0 i
3 0 0
i
3 0 2i 0
0 2i 0 i
3
0 0 i
3 0
_
_
_
_
_
_
_
_
,
and
J
z
=
2
_
_
_
_
_
_
_
_
3 0 0 0
0 1 0 0
0 0 1 0
0 0 0 3
_
_
_
_
_
_
_
_
.
5.2 Spherical Harmonics
We now come back to the specic case of orbital angular momentum. We shall employ
the notation
L instead of
J when referring to the orbital angular momentum, and j
will be replaced by l. In the last section we have said that j is integer or half-integer.
As we shall discuss in this section, l and therefore m must be integer numbers for the
orbital angular momentum.
The rst step is to look for a representation of the abstract eigenstates [l, m of
L
2
and
L
z
. This is somewhat analogous to the case in which the eigenstates of the
Hamiltonian
H, say [E, were represented on the basis of the position eigenstates:
x[E =
E
(x). In that case,
E
(x) were the eigenfunctions of the Schr odinger oper-
ator in the position representation, and were found to be square integrable functions
of the spatial coordinate x. Here we shall do the same thing with the eigenstates
126 Chapter 5. Angular momentum quantization
x
y
z
(x;y;z)
# r
Fig. 5.1. Polar coordinates.
[l, m but the basis is not that of the Cartesian coordinates [x, y, z but that of the
polar coordinates [r, , :
r, , [l, m = Y
l,m
(, ) .
As we shall see, Y
l,m
are functions not depending explicitly on r and are usually
indicated with the symbol Y
l,m
; they are referred to as spherical harmonics. The
next goal is to calculate these functions.
First of all, we recall the usual relation between Cartesian and polar coordinates
(see Fig. 5.1)
x = r cos sin , (5.17a)
y = r sin sin , (5.17b)
z = r cos . (5.17c)
The three components of the angular momentum can also be rewritten in terms of
polar coordinates. For example, the z-component of the angular momentum reads
L
z
= i
_
x
y
y
x
_
;
5.2. Spherical Harmonics 127
x and y can be replaced directly by using Eqs. (5.17) while the derivatives /x and
/y can be rewritten in polar coordinates by using the chain rule of dierentiation:
x
=
x
+
x
+
r
x
r
,
and analogously for the other derivatives. In this way one nds
L
x
= i
_
sin
+ cot cos
_
,
L
y
= i
_
cos
+ cot sin
_
,
L
z
= i
.
Homework 5.6. Derive the expressions above for the components of the orbital
angular momentum in polar coordinates.
In the representation of [r, , the eigenvalue problem
L
z
[l, m = m [l, m be-
comes
i
. .
Lz
Y
l,m
(, ) = mY
l,m
(, ) . (5.18)
Homework 5.7. Try to derive Eqn. (5.18). Bring in mind that the resolution of the
identity formula in polar coordinates reads
1 =
_
[, , [ sin dd .
The latter relation follows from the form of the integration volume
_
dxdy dz =
_
r
2
dr
_
sin dd .
The solution of the eigenvalue problem (5.18) is given by
Y
l,m
(, ) =
e
im
m
l
() , (5.19)
where
m
l
() is a function to be determined which depends only on while the whole
dependence on is included in the factor e
im
. We have put a factor 1/
2 as we
did for the normalization of plane wave functions.
128 Chapter 5. Angular momentum quantization
In order to determine the function
m
l
() we need to express the operators
L
(+)
and
L
()
in polar coordinates:
L
()
=
L
x
i
L
y
= e
i
_
+ i cot
_
. (5.20)
We recall from the previous section that
L
(+)
[l, l = 0 ,
which now becomes
L
(+)
Y
l,l
(, ) = 0 ,
and using Eqn. (5.19)
_
d
d
l
cos
sin
_
l
l
() = 0 ,
we get
d
d
_
1
sin
l
l
l
()
_
= 0 .
Its solution is given by
l
l
() = k sin
l
.
The value of k is xed by requiring
l
l
() to be normalized to 1:
_
0
[
l
l
()[
2
sin d = 1 .
The result is
l
l
() = (1)
l
_
(2l + 1)!
2
1
2
l
l!
sin
l
.
By repeatedly using Eqn. (5.20) we can obtain a general expression for
m
l
():
m
l
() =
1
lm
1
2l!
(l + m)!
(l m)!
L
lm
()
l
l
() .
The nal result is
Y
l,m
(, ) = (1)
l
_
2l + 1
4
(l + m)!
(l m)!
e
im
sin
m
d
lm
d(cos )
lm
(sin )
2l
. (5.21)
5.2. Spherical Harmonics 129
Remark: sometimes Y
l,m
(, ) are denoted as Y
m
l
(, ). The spherical harmonics
can be written in a more compact way in terms of associated Legendre functions
P
m
l
(cos ).
The abstract states [l, m form a complete orthonormal basis as they are eigenstates
of the Hermitian operators
L
2
and
L
z
L
2
[l, m =
2
l(l + 1) [l, m
L
z
[l, m = m[l, m .
Thus also their representation (spherical harmonics) makes up a basis on which we
can expand every function of and , i.e.:
(, ) =
l,m
c
l,m
Y
l,m
(, ) .
For a given l we know that
l, m[l, m
=
m,m
,
and this relation turns into
_
2
0
d
_
0
d sin Y
l,m
(, )Y
l,m
(, ) =
m,m
or more generally
_
2
0
d
_
0
d sin Y
l,m
(, )Y
l
,m
(, ) =
m,m
l,l
.
Let us now consider in some detail the case of a plane wave exp(i k r) in three
dimensions and nd its expansion on spherical harmonics. As k r is a scalar product,
it is invariant under rotations and there is no loss of generality if we restrict ourselves
to the case in which the vector k lies along the z-axis:
e
ikr
= e
ikr cos
.
Expanding on spherical harmonics yields an expression of the form
e
ikr cos
=
l,m
c
l,m
(r)Y
l,m
(, ) .
130 Chapter 5. Angular momentum quantization
In general, the expansion coecients c
l,m
depend on r. In the case of plane waves,
the left-hand side in the previous equation does not depend on , that implies that
the expansion can be rewritten in terms of
l,m
() only, that is,
e
ikr cos
=
l,m
c
l,m
(r)
l,m
() .
The coecients of the expansion depend on l only and can be written in terms of the
so-called spherical Bessel functions J
l
:
c
l,m
(r) J
l
(kr) .
As exp(ikr cos ) depends on via cos , the
l,m
() which depend on via cos are
only those having m = 0, therefore
e
ikr cos
=
l
c
l
(kr)
l,0
() , c
l
(kr) J
l
(kr) .
This is the spherical harmonics expansion of the plane wave.
5.3 Addition of angular momenta
Let us now consider two systems with given angular momenta and we want to nd
the angular momentum of the composite system. The results of this section are not
restricted to the case of orbital angular momentum
L
i
but to any operator
J
i
satisfying
the algebra (5.3).
It should be stressed that here we are not dealing with the problem of having to
sum two numbers, instead we have to sum two operators. For the i-components this
reads
J
(T)
i
=
J
(1)
i
+
J
(2)
i
.
Hereafter T denotes the total system (1 + 2) and the labels (1) and (2) indicates the
two systems.
5.3. Addition of angular momenta 131
Being an operator,
J
(T)
has its own spectrum of eigenvalues. What we shall prove
in this section is that the possible eigenvalues of (
J
(T)
)
2
are given by one among the
set of values
[j
(1)
j
(2)
[ , . . . , (j
(1)
+ j
(2)
) ,
where j
(1)
and j
(2)
are the eigenvalues of (
J
(1)
)
2
and (
J
(2)
)
2
respectively, that is
(
J
(1)
)
2
[ = j
(1)
(j
(1)
+ 1)
2
[ .
and an analogous relation holds for (
J
(2)
)
2
and (
J
(T)
)
2
.
(1) and (2) refer to two dierent systems, so the angular momentum components
of the rst system do commute with each component of the second system:
[
J
(1)
i
,
J
(2)
j
] = 0 .
From this relation it is easy to prove that the following four operators commute with
each other:
(
J
(1)
)
2
, (
J
(2)
)
2
,
J
(1)
z
,
J
(2)
z
, (5.22)
and the same is true also for the following set of four operators
(
J
(1)
)
2
, (
J
(2)
)
2
, (
J
(T)
)
2
,
J
(T)
z
. (5.23)
Homework 5.8. Prove that the operators (5.23) commute with each other.
Let us diagonalize simultaneously the set (5.22). The eigenvalue problem reads
(
J
(1)
)
2
[j
(1)
, j
(2)
, m
(1)
, m
(2)
=
2
j
(1)
(j
(1)
+ 1) [j
(1)
, j
(2)
, m
(1)
, m
(2)
,
(
J
(2)
)
2
[j
(1)
, j
(2)
, m
(1)
, m
(2)
=
2
j
(2)
(j
(2)
+ 1) [j
(1)
, j
(2)
, m
(1)
, m
(2)
,
J
(1)
z
[j
(1)
, j
(2)
, m
(1)
, m
(2)
= m
(1)
[j
(1)
, j
(2)
, m
(1)
, m
(2)
,
J
(2)
z
[j
(1)
, j
(2)
, m
(1)
, m
(2)
= m
(2)
[j
(1)
, j
(2)
, m
(1)
, m
(2)
.
132 Chapter 5. Angular momentum quantization
If instead we diagonalize the second set of operators, namely the set (5.23), we get
(
J
(1)
)
2
[j
(1)
, j
(2)
, J
(T)
, M
(T)
=
2
j
(1)
(j
(1)
+ 1) [j
(1)
, j
(2)
, J
(T)
, M
(T)
,
(
J
(2)
)
2
[j
(1)
, j
(2)
, J
(T)
, M
(T)
=
2
j
(2)
(j
(2)
+ 1) [j
(1)
, j
(2)
, J
(T)
, M
(T)
,
(
J
(T)
)
2
[j
(1)
, j
(2)
, J
(T)
, M
(T)
=
2
J
(T)
(J
(T)
+ 1) [j
(1)
, j
(2)
, J
(T)
, M
(T)
,
J
(T)
z
[j
(1)
, j
(2)
, J
(T)
, M
(T)
= M
(T)
[j
(1)
, j
(2)
, J
(T)
, M
(T)
J
(T)
z
are made up of
J
(T)
i
.
In order to simplify the notation we will denote the common eigenstates of (
J
(1)
)
2
and (
J
(2)
)
2
at xed j
(1)
and j
(2)
with [m
(1)
, m
(2)
, i.e. the states [j
(1)
, j
(2)
, m
(1)
, m
(2)
will be denoted by [m
(1)
, m
(2)
once j
(1)
and j
(2)
have been xed. Similarly , the states
[j
(1)
, j
(2)
, J
(T)
, M
(T)
will be indicated by [J
(T)
, M
(T)
.
Let us try to write [J
(T)
, M
(T)
in the basis of [m
(1)
, m
(2)
. In this way we shall
obtain the possible values of J
(T)
and M
(T)
once j
(1)
and j
(2)
are xed. The solution
of this problem is based on the following two observations, namely
1. Every vector [m
(1)
, m
(2)
is an eigenstate of
J
(T)
z
with corresponding eigenvalue
M
(T)
= m
(1)
+ m
(2)
,
that is
J
(T)
z
[m
(1)
, m
(2)
= (m
(1)
+ m
(2)
) [m
(1)
, m
(2)
.
This is because
J
(T)
z
=
J
(1)
z
+
J
(2)
z
.
5.3. Addition of angular momenta 133
2. For every value of J
(T)
there are various M
(T)
and therefore there is a certain
number, say N(J
(T)
), of vectors in the space of [m
(1)
, m
(2)
. Every set contains
all the eigenvalues M
(T)
from J
(T)
to J
(T)
.
We shall instead denote with n(M
(T)
) the number of states [m
(1)
, m
(2)
which
have the same M
(T)
.
It is easy to prove that the relation among n(M
(T)
) and N(J
(T)
) is
n(M
(T)
) =
J
(T)
|M
(T)
|
N(J
(T)
) , (5.24)
because if J
(T)
is larger than M
(T)
then surely it has some (M
)
(T)
which is equal to
M
(T)
, (M
)
(T)
being in the range from J
(T)
to J
(T)
.
From Eqn. (5.24) it follows that
N(J
(T)
) = n(J
(T)
) n(J
(T)
+ 1) . (5.25)
In fact, n(J
(T)
) and n(J
(T)
+ 1) dier only for one term, namely that of J
(T)
.
From Eqn. (5.25) one sees that in order to determine N(J
(T)
) it is sucient to
determine n(M
(T)
). Now, n(M
(T)
) is the number of pairs (m
(1)
, m
(2)
) such that
M
(T)
= m
(1)
+ m
(2)
.
To nd this number one can employ a diagram like the one plotted in Fig. 5.2 for the
special case j
(1)
= 7/2 and j
(2)
= 2. n(M
(T)
) is the number of pairs (m
(1)
, m
(2)
) in
that plot that are situated along the line of equation M
(T)
= m
(1)
+m
(2)
. If j
(1)
> j
(2)
we get from Fig. 5.2
n(M
(T)
) =
_
_
0 if [M
(T)
[ > j
(1)
+ j
(2)
j
(1)
+ j
(2)
+ 1 M
(T)
if j
(1)
+ j
(2)
[M
(T)
[ [j
(1)
j
(2)
[
2j
(2)
+ 1 if [j
(1)
j
(2)
[ M
(T)
0
If we put these numbers into Eqn. (5.25) we obtain
N(J
(T)
) = 1
134 Chapter 5. Angular momentum quantization
m
(1)
m
(2)
j
(1)
j
(2)
j
(1)
+ j
(2)
Fig. 5.2. Addition of angular momenta. The bullets display all the pairs (m
(1)
, m
(2)
)
for j
(1)
= 7/2 and j
(2)
= 2. Remember that the only allowed values of m
(1)
and
m
(2)
given j
(1)
and j
(2)
are m
(1,2)
= j
(1,2)
, j
(1,2)
+ 1, . . . , j
(1,2)
1, j
(1,2)
thus m
(1)
=
7/2, 5/2, . . . , 5/2, 7/2 and m
(2)
= 2, 1, 0, 1, 2. The lines correspond to M
(T)
=
m
(1)
+m
(2)
. For example, the red line in gure is the line of the pairs (m
(1)
, m
(2)
) such that
M
(T)
= j
(1)
j
(2)
+ 3. To count the number of pairs n(M
(T)
) you have simply to count
how many pairs are situated along the line corresponding to M
(T)
.
for
J
(T)
= (j
(1)
+ j
(2)
), (j
(1)
+ j
(2)
1), . . . , [j
(1)
j
(2)
[ ,
otherwise for all the other eigenvalues we get N(J
(T)
) = 0.
Remark. These are the eigenstates at j
(1)
and j
(2)
xed.
From the previous considerations we obtain the following :
Theorem 5.1. In the (2j
(1)
+ 1)(2j
(2)
+ 1)-dimensional space spanned by
[j
(1)
, j
(2)
, m
(1)
, m
(2)
with j
(1)
and j
(2)
xed, the possible values of the total angular momentum J
(T)
are
J
(T)
= (j
(1)
+ j
(2)
), (j
(1)
+ j
(2)
1), . . . , [j
(1)
j
(2)
[ .
and to each of the possible values of J
(T)
it is associated one and only one multiplet
of eigenstates [J
(T)
, M
(T)
of the total angular momentum.
5.4. Clebsh-Gordon Coecients 135
5.4 Clebsh-Gordon Coecients
In the last section dealing with the sum of angular momenta, we have made use of
two basis: the [j
1
, j
2
, J, M and [j
1
, j
2
, m
1
, m
2
. We did not insert any other label
into these kets because we made the hypothesis that
j
2
1
,
j
2
2
,
J
2
,
M form a complete
set of observables and the same for the operators associated to the other ket. If this
were not so, we would have had to introduce a further label, which we will indicate
with the symbol , to indicate the eigenvalues of a set of operators to be added to
the previous ones to make them a complete set . The states would be indicated as:
[, j
1
, j
2
, J, M , and [, j
1
, j
2
, m
1
, m
2
.
In order to dene the states [, j
1
, j
2
, J, M unambiguously we will assume them
to have norm one and we will adopt an appropriate convention for its phase. In the
Hilbert space H (, j
1
, j
2
) the states [, j
1
, j
2
, J, M and [, j
1
, j
2
, m
1
, m
2
form two
orthonormal basis and we can pass from one to the other by a unitary transformation:
[, j
1
, j
2
, J, M =
m
1
,m
2
[, j
1
, j
2
, m
1
, m
2
, j
1
, j
2
, m
1
, m
2
[, j
1
, j
2
, m
1
, m
2
.
It is possible to prove that the coecients of the transformation, i.e:
, j
1
, j
2
, m
1
, m
2
[, j
1
, j
2
, m
1
, m
2
, J
, M
= M
,
J,J
M,M
and
, J, M[
J
, J
, M
=
_
J(J 1) MM
,
J,J
M,M
1
.
The same will happen for the matrices representing
j
1
and
j
2
. As a consequence also
the components of the eigenvectors of
J
2
and
J
z
, i.e , j
1
, j
2
, m
1
, m
2
[, j
1
.j
2
, J, M
136 Chapter 5. Angular momentum quantization
, will be independent of . So we can say that these transition coecients have
a pure geometrical interpretation depending only on the angular momenta and their
orientation and not upon the physical nature of the dynamical variables labelled by
. These coecients are called Clebsh-Gordon (C.G) coecients and will be written
without the label , as j
1
, j
2
, m
1
, m
2
[j
1
.j
2
, J, M. Taking account of this fact the
expansion of the state [, j
1
, j
2
, J, M can then be written as::
[, j
1
, j
2
, J, M =
m
1
,m
2
[, j
1
, j
2
, m
1
, m
2
j
1
, j
2
, m
1
, m
2
[j
1
.j
2
, J, M .
In order to completely determine the C.G. coecients, we should x the phases of
the vectors , j
1
, j
2
, J, M[. For a xed J the relative phases of the 2J associated
states have been xed via the choice of taking real and positive the coecients of
Eqn.(5-14a,b). Once this is done the , j
1
, j
2
, J, M[ have only the freedom of a phase
which can change with J. We x this phase by requiring that, whatever
is J, the
components of , j
1
, j
2
, J, J[ along , j
1
, j
2
, J, J j
1
[ be real and positive i.e:
j
1
, j
2
, J, J j
1
[JJ 0.
Many properties of the C.G coecients follow directly from their denition and
the requirements above for the phases. To get all of the C.G coecients we have just
to use the recursion relations (5-14-a,b) and the orthonormality relation between the
states. It is then easy to nd several symmetry properties of the C.G coecients which
greatly facilitates their calculation and tabulation as was done by Racah. For more
details the reader should study sect.27 of chapter XIII and appendix C of Messiah.
5.5 Problems and Solutions
J can go from [j
1
j
2
[ to j
1
+j
2
5.5. Problems and Solutions 137
Problem 5.1. Let us consider two weakly interacting systems whose states are
specied by the quantum numbers
(l
(1)
, m
(1)
) , (l
(2)
, m
(2)
)
of the angular momentum and of its projection along the z-axis. Calculate the
possible values of the total angular momentum
L
(T)
and the mean values of
L
(T)
i
and (
L
(T)
)
2
on the corresponding eigenstates.
Solution. Note that here not only l
(1)
and l
(2)
but also m
(1)
and m
(2)
are xed. It is
possible to prove that
max
_
[l
(1)
l
(2)
[ , [m
(1)
+m
(2)
[
_
L
(T)
l
(1)
+l
(2)
. (5.26)
The upper limit comes obviously from the theory of addition of angular momenta developed
in the previous section. In fact, we have seen that
[l
(1)
l
(2)
[ L
(T)
l
(1)
+l
(2)
.
The lower limit however is dierent. Why? We know that
L
(T)
M
(T)
L
(T)
and M
(T)
= m
(1)
+m
(2)
. Thus in our present case if [m
(1)
+m
(2)
[ is higher than [l
(1)
l
(2)
[,
we must stop at [m
(1)
+m
(2)
[.
Of course, since 1 and 2 are two dierent systems, the angular momentum components
of the rst system do commute with each component of the second one:
[
L
(1)
i
,
L
(2)
j
] = 0 .
It is also possible to prove that
L
(1)
x
=
L
(2)
x
= 0 =
L
(1)
y
=
L
(2)
y
.
138 Chapter 5. Angular momentum quantization
We have
L
(T)
x
=
L
(1)
x
+
L
(2)
x
= 0 ,
L
(T)
y
=
L
(1)
y
+
L
(2)
y
= 0 ,
L
(T)
z
=
L
(1)
z
+
L
(2)
z
= (m
(1)
+m
(2)
). .
(Prove this as homework.)
Let us now consider (
L
(T)
)
2
:
(
L
(T)
)
2
=
_
L
(1)
+
L
(2)
_
2
= (
L
(1)
)
2
+ (
L
(2)
)
2
+ 2
L
(1)
L
(2)
.
We know that
(
L
(1)
)
2
=
2
l
(1)
(l
(1)
+ 1) , (
L
(2)
)
2
=
2
l
(2)
(l
(2)
+ 1) .
We have only to calculate
2
L
(1)
L
(2)
= 2
L
(1)
x
L
(2)
x
+ 2
L
(1)
y
L
(2)
y
+ 2
L
(1)
z
L
(2)
z
.
Remember that the states on which we are calculating the mean values are
[l
(1)
, m
(1)
, l
(2)
, m
(2)
= [l
(1)
, m
(1)
[l
(2)
, m
(2)
,
where the symbol stay for tensor product and an operative denition has been given in
chapter 3, thus for example:
L
(1)
x
L
(2)
x
=
_
l
(1)
, m
(1)
[ l
(2)
, m
(2)
[
_
L
(1)
x
L
(2)
x
_
[l
(2)
, m
(2)
[l
(1)
, m
(1)
_
= l
(1)
, m
(1)
[
L
(1)
x
[l
(1)
, m
(1)
l
(2)
, m
(2)
[
L
(2)
x
l
(2)
, m
(2)
=
L
(1)
x
L
(2)
x
= 0 .
In the same way
L
(1)
y
L
(2)
y
=
L
(1)
y
L
(2)
y
= 0 ,
L
(1)
z
L
(2)
z
=
L
(1)
z
L
(2)
z
=
2
m
(1)
m
(2)
.
Thus,
(
L
(T)
)
2
=
2
_
l
(1)
(l
(1)
+ 1) +l
(2)
(l
(2)
+ 1) + 2m
(1)
m
(2)
_
. (5.27)
5.5. Problems and Solutions 139
Problem 5.2. In the same setting of the previous problem, calculate the proba-
bilities of the various values of the angular momentum, assuming m
(1)
= l
(1)
and
m
(2)
= l
(2)
1.
Solution. By looking at Eqn.(5.26), the range of possible values of
L
(T)
is:
max
_
[l
(1)
l
(2)
[ , [m
(1)
+m
(2)
[
_
L
(T)
l
(1)
+l
(2)
.
In the special case in which m
(1)
= l
(1)
and m
(2)
= l
(2)
1,
max
_
[l
(1)
l
(2)
[ , [m
(1)
+m
(2)
[
_
= max
_
[l
(1)
l
(2)
[ , [l
(1)
+l
(2)
1[
_
and so the range becomes
l
(1)
+l
(2)
1 L
(T)
l
(1)
+l
(2)
.
This relation tells that there are only two possible values of L
(T)
, namely
l
(1)
+l
(2)
, l
(1)
+l
(2)
1 .
If we denote by P(L
(T)
) the probability of nding L
(T)
in a measurement, then
P(l
(1)
+l
(2)
) +P(l
(1)
+l
(2)
1) = 1 .
and the mean value (
L
(T)
)
2
can be written (in units of )
(
L
(T)
)
2
=
L
(T)
L
(T)
(L
(T)
+ 1)P(L
(T)
)
= (l
(1)
+l
(2)
)(l
(1)
+l
(2)
+ 1)P(l
(1)
+l
(2)
) +
(l
(1)
+l
(2)
1)(l
(1)
+l
(2)
)
_
1 P(l
(1)
+l
(2)
)
_
. .
=P(l
(1)
+l
(2)
1)
. (5.28)
From Eqn. (5.27) with m
(1)
= l
(1)
and m
(2)
= l
(2)
1 we get
(
L
(T)
)
2
= l
(1)
(l
(1)
+ 1) +l
(2)
(l
(2)
+ 1) + 2l
(1)
(l
(2)
1) ,
140 Chapter 5. Angular momentum quantization
comparison between this expression and Eqn. (5.28) yields
P(l
(1)
+l
(2)
) =
l
(2)
l
(1)
+l
(2)
, P(l
(1)
+l
(2)
1) =
l
(1)
l
(1)
+l
(2)
.
Problem 5.3. Prove that, starting from the commutation relation
[
L
i
,
O] = 0 ,
between the angular momentum operators
L
i
and an Hermitian operator
O, it
is possible to deduce that the matrix elements
L, M[
O[L, M
L
z
O
O
L
z
= 0 .
By applying it to the eigenstate [M we get
L
z
O[M =
O
L
z
[M = M
O[M ,
and this tells us that
O[M is eigenstate of
L
z
with eigenvalue M. Thus,
M
OM M
[M =
M
,M
,
which is the rst part of what we have to prove.
Let us now prove the second part. Since every
L
i
commutes with
O,
L
(+)
commutes with
O too:
L
(+)
O =
O
L
(+)
,
5.5. Problems and Solutions 141
and by taking the matrix elements of
L
(+)
O between M + 1[ and [M we get
M + 1[
L
(+)
O[M = M + 1[
L
(+)
[M .
We know that
L
(+)
acts on [M in the following way:
L
(+)
[M =
_
(L M)(L +M + 1) [M + 1 .
It is also easy to prove that
M + 1[
L
(+)
=
_
(L M)(L +M + 1) M[ .
Hence,
_
(L M)(L +M + 1) M[
O[M = M + 1[
O[M + 1
_
(L M)(L +M + 1)
which implies
M[
O[M = M + 1[
O[M + 1 ,
This proves that the expectation values of
O are independent of M.
Problem 5.4. Find the eigenstates of the total angular momentum of two sys-
tems whose angular momenta are L
(1)
= 2 and L
(2)
= 1. How many states are
there? How many of them have M
(T)
= 0?
Solution. For the rules of addition of angular momenta, L
(T)
can take only the following
values:
[L
(1)
L
(2)
[ , . . . , L
(1)
+L
(2)
.
In our case, L
(1)
= 2, L
(2)
= 1 thus [L
(1)
L
(2)
[ = 1 and L
(1)
+L
(2)
= 3, therefore
L
(T)
=
_
_
3 or
2 or
1
142 Chapter 5. Angular momentum quantization
Associated to each value of L
(T)
are the corresponding values of M
(T)
, for example for
L
(T)
= 3 the possible values of M
(T)
are
M
(T)
= 3, 2, 1, 0, 1, 2, 3 .
Therefore, the states [L
(T)
, M
(T)
with L
(T)
= 3 are
[3, 3 , [3, 2 , [3, 1 , [3, 0 , [3, 1 , [3, 2 , [3, 3 .
There are seven states [L
(T)
, M
(T)
with L
(T)
= 3. For L
(T)
= 2 the possible states are
[2, 2 , [2, 1 , [2, 0 , [2, 1 , [2, 2 ,
there are ve such states. Finally, there are three states with L
(T)
= 1, namely
[1, 1 , [1, 0 , [1, 1 .
There is a total of 7 + 5 + 3 = 15 states, three of which have M
(T)
= 0: [3, 0, [2, 0 and
[1, 0.
Chapter 6
SYMMETRIES IN QUANTUM MECHANICS
6.1 Rotations, symmetries, unitary transformations and degeneracy
In quantum mechanics the basic quantities are the probabilities:
P = [ .
If now we change the states [ via a unitary transformation
U:
U [ = [
,
it is easy to prove that the probabilities are left unchanged. In fact,
P
= [
U[ = [
U
1
U[ = [ = P .
Therefore, the probabilities do not change under unitary transformations of states.
What happens to the mean values of an observable? Let us consider the expectation
value of an observable
O on a state [, it is given by
O = [
O[ .
Now, let [ be transformed according to an unitary transformation
U: [
=
U [.
In order the mean value to be the same,
O must be transformed according to
=
U
O
.
In fact, in this way one gets
= [
U
O
U[ = [
O[ .
144 Chapter 6. Symmetries in Quantum Mechanics
So by performing a unitary transformation we have to change both the states and the
operators.
Somehow unitary transformations play in quantum mechanics the role that in clas-
sical mechanics is played by canonical transformations.
Let us now turn back to rotations. In classical mechanics, angular momentum is
the generator of the rotations, in the sense that any observable O transforms under
a rotation of an angle around the z-axis according to the following rule:
O
= O + L
z
, O
P.B.
, (6.1)
as long as can be considered small enough.(P.B) stays for Poisson Brackets. If we
now look at innitesimal time translations, these are generated by
H and accordingly
we have
O
= O + t H, O
P.B.
.
By using the correspondence rule, we may infer the transformation in quantum me-
chanics:
=
O + t
[
H,
O]
i
.
We also know that for nite transformations the time-evolution operator is given by
U
t
= e
Ht
. (6.2)
In the same way, for the innitesimal rotation given by Eqn. (6.1) the quantum
analogous becomes:
=
O +
[
L
z
,
O]
i
. (6.3)
For a nite rotation we should get the analogous of Eqn. (6.2), namely
= e
Lz
.
This means that under nite rotations the states should transform according to
[
=
U
[ ,
6.1. Rotations, symmetries, unitary transformations and degeneracy 145
while observables
O transform in the following way:
=
U
= e
Lz
Oe
i
Lz
.
Let us now see if from this, for small angles, we get (6.3). Let us start from
= e
Lz
Oe
i
Lz
,
by expanding the exponentials we get
=
_
1
i
L
z
+ . . .
_
O
_
1 +
i
L
z
+ . . .
_
=
O +
i
[
O,
L
z
] +O(
2
)
=
O +
i
[
L
z
,
O] +O(
2
)
which is nothing but Eqn. (6.3).
If [
L
z
,
O] = 0, then
O
=
O, i.e., the observable
O is invariant under rotations.
Let us now try to derive explictely that
U
= e
i
Lz/
generates a nite rotation
around the z-axis.
For an innitesimal rotation we have:
[
=
U
[ = e
Lz
[ .
Taking the derivative of [
d
=
i
L
z
[
.
If we perform explicitly a traslation of an innitesimal angle on then (r, , )
goes into
(r, , + ), hence
(r, , + ) = (r, , ) +
d
d
(r, , )
= (r, , )
i
L
z
(r, , ) + . . .
146 Chapter 6. Symmetries in Quantum Mechanics
and since is small we can write
[
= e
Lz
[ .
To get a nite rotation, we have to multiply many innitesimal rotations:
e
Lz
e
Lz
(6.4)
By applying the Baker-Hausdor formula
e
A
e
B
= e
H(
A,
B)
,
where
1(
A,
B) =
A+
B +
1
2
[
A,
B] + . . .
and the dots denote commutators of commutators, we get from Eqn. (6.4)
e
(
1
+
2
+...)
Lz
, (6.5)
because the various
L
z
commute among themselves. Eqn. (6.5) means that even for
nite rotations the generator is given by
U() = e
i
Lz/
that is what we wanted to
prove.
So far we have dealt with rotations around the z-axis. Whats about generic rota-
tions? It is well-known from analytic mechanics that any rotation can be written as a
rotation of some angle around a direction specied by a unit vector n. Thus, three
angles are needed in order to specify a generic rotation: the two that specify n and
the rotation angle . The unitary operators associated to such rotations are given by
U(, n) = e
L
, (6.6)
where n
L = n
x
L
x
+ n
y
L
y
+ n
z
L
z
. Eqn. (6.6) reduces to the previous formula for
the case of rotations about the z-axis.
Remark: The statement we made at the beginning of this section that mean values
are left unchanged under unitary transformations, is always true except in the case
6.2. Symmetries, generators and degeneracy 147
of the time evolution generated by e
i
Ht/
. Under this transformation, mean values
change in time. We will see later that in this case states and observables must be
treated on a dierent ground (Schr odinger and Heisenberg picture).
6.2 Symmetries, generators and degeneracy
In classical mechanics, a symmetry is a transformation which leaves the equation of
motion invariant. It is easy to prove that at the Hamiltonian level this implies the
existence of a quantity Q(q, p) having null Poisson brackets with the Hamiltonian:
Q(q, p), H
P.B.
= 0 .
If Q(q, p) does not depend explicitly on time, then
dQ(q, p)
dt
=
Q
q
q
t
+
Q
p
p
t
=
Q
q
H
p
Q
p
H
q
= Q, H
P.B.
.
If Q, H
P.B.
= 0 then dQ/ dt = 0, i.e., Q is a conserved quantity (Noether theorem).
By using the corrispondence principle,
,
P.B.
1
i
[, ] ,
we get
Q, H
P.B.
1
i
[
Q,
H] ,
and from Q, H
P.B.
= 0 we obtain [
Q,
H] = 0. This means that at the quantum
level a symmetry is signaled by the presence of a charge
Q which commutes with the
Hamiltonian.
(x) ,
leads to the conservation of charge. The second name comes from the fact that the
transformation
q = q, Q
P.B.
,
p = p, Q
P.B.
,
(6.7)
148 Chapter 6. Symmetries in Quantum Mechanics
generates a symmetry of the equation of motion. Here, is an innitesimal parameter,
When does not depends on time, we speak of global symmetry, otherwise we speak
of local or gauge transformation.
Homework 6.1. Prove that if Q, H
P.B.
= 0 then the equation of motion are left
unchanged in form, i.e., Q generates a symmetry.
From the correspondence principle, it may seem that if a charge is conserved at
the classical level then it is conserved also at quantum level. Actually, this is not
always the case, at least when considering systems having innitely many degrees of
freedom, such as elds. In such cases quantum eects may break the symmetry. The
two most studied examples of such eects are the spontaneous symmetry breaking
and the anomalies, the former being related for example to the Higgs mechanism
while the latter is responsible for the
0
2 decay and it also implies that there
must be as many quarks as leptons.
Let us now turn to investigate what are the rst consequences of the presence of
a symmetry in quantum mechanics. The rst consequence is degeneracy. Let us
suppose we know the eigenstates of the Hamiltonian
H:
H[E
n
= E
n
[E
n
.
For simplicity, we are assuming the eigenstates to be discrete. We have learned that
if
H has a symmetry, then there must be a charge
Q which commutes with it:
[
Q,
H] = 0 .
Now, it is easy to prove that the state
[
E
n
Q[E
n
is again an eigenstate of
H with the same eigenvalue of [E
n
, namely E
n
, thus we
6.3. Brief introduction to Lie groups. 149
have degeneracy. The proof goes as follows. From [
Q,
H] = 0 we get
[
Q,
H] [E
n
= 0
Q
H [E
n
H
Q[E
n
= 0
QE
n
[E
n
H[
E
n
= 0
E
n
Q[E
n
H[
E
n
= 0
E
n
[
E
n
H[
E
n
= 0 ,
from which it follows
H[
E
n
= E
n
[
E
n
.
If now we consider the unitary operator
U associated to
Q, i.e.,
U() = e
Q
,
where is the parameter of the symmetry, we can construct a whole family of states
U() [E
n
= [
E
n
, ,
and they are all degenerate with [E
n
, i.e.:
H[
E
n
, =
H e
Q
[E
n
= e
Q
H[E
n
= e
Q
E
n
[E
n
= E
n
e
Q
[E
n
= E
n
[
E
n
, .
Remark: not every degeneracy is due to symmetries. There are also accidental de-
generacies.
6.3 Brief introduction to Lie groups.
The set of unitary operators
U(, n) = e
L
which implement the generic rotations
on the Hilbert space form a structure which in mathematics is called a group. A group
( is a set of elements g ( on which an internal operation is dened (let us indicated
it with ) such that:
g g
= g
(
150 Chapter 6. Symmetries in Quantum Mechanics
Moreover in ( there must exist an identity element, e, such that:
e g = g e = g
and nally, given any element g there must exist in ( the inverse of g, indicated by
g
1
such that:
g g
1
= g
1
g = e.
Some groups are a set on which the concept of continuity can be dened and this,
together with few other properties, make what are called the Lie groups. For these
it is always possible to nd what are called the generators
O
i
of the group. They are
such that every element U of the group can be written as :
U = e
i
i
i
O
i
where the
i
are a set of continuos parameters which are as many as the group gen-
erators
O
i
. The group structure of U induces the following algebra on the generators:
[
O
i
,
O
j
] = C
k
i,j
O
k
where [, ] is the commutator and the C
k
i,j
are constants called structure constants. The
above commutation relation denes what is called a Lie-algebra among the generators.
For the rotation group the analogue of this algebra are the commutators among the
angular momentum operators. Given an abstract Lie algebra we can build the Lie-
group but then we have to decide on which space it acts. The explicit realization of
the algebra and the group elements on this space is called a representation of the
group and of the algebra. For example for the rotation group and algebra we have
seen in the previous chapter that the spaces on which it can act are vector spaces of
dimension 2J + 1 with J integer or half-integer. So we can say that each value of J
labels a representation.
This is just a brief introduction to the concept of groups. More will be studied in
advanced courses next year.
6.4. Scalar and Tensor Operators 151
6.4 Scalar and Tensor Operators
From now on we shall concentrate con the rotation group.
There are observables which are invariant under rotation like for example:
R
2
x
2
+ y
2
+ z
2
These operators are called scalars and we will indicate them with
S. From the invari-
ance of
S under U = e
J
, i.e:
S = U
SU
we get that
S commutes with the generators
J :
[
S,
J] = 0. (6.8)
In an exercise in chapter 5) we proved that, from the commutation relation (6.8),
we can derive that:
J, M[
S [J
, M
=
J,J
M,M
S
(J)
(6.9)
where S
(J)
is a quantity which depends on J but not on M. Eqn.(6.9) tells us also
that the state
S [J
, M
, M
), i.e
S does not
change the quantum numbers of the state to which it is applied. We can generalize
the concept of scalar operator to vector and tensor operators. They are dened by
the manner they transform under rotation. A vector operator
V is dened as:
U
V
i
U
=
3
J=1
R
ij
V
J
where R
ij
are the same three by three rotation matrix we would obtain if we apply
a rotation to (x, y, z). If we take two vector operators
V and
W, we can build the
following object with 9 components:
V
i
W
j
(6.10)
152 Chapter 6. Symmetries in Quantum Mechanics
plus all their linear combinations. Their transformation properties would be
(
V
i
W
j
)
= R
ik
R
js
(
V
k
W
s
).
This is a particular tensor. We see that it is an object with multi-index and each of
them transform linearly, in our case with the matrix R. Among the tensors we will
pay particular attention to the irreducible ones. These are dened as follows: let us
suppose the space of our tensors has dimension N. If there is a subset of this space
of dimension M < N such that any rotation brings every element of this subset into
elements of the same subset, then we say that the tensor space is reducible, in the
opposite case we say that it is irreducible. When it is reducible we can represents our
tensor in the subspace M. For example the tensor we have built in Eqn.(6.10) is not
irreducible. Its 9-dimensional space contains irreducible subspaces of dimension 1,3
and 5. They are respectively the space of the scalar product:
V
i
W
i
that under
rotation is left unchanged and it is a 1-dimensional space, the 3-dimensional space
associated to the the vector product:
V
W =
ijk
V
j
W
k
which is a vector and so
under rotation goes into a vector, and the 5-dimensional space associated to J = 2.
The dimensions of all these spaces is always 2J +1 with J integer or half-integer. In
general, it is easy to prove that, if the (2J +1) operators
T
(J)
M
with J M J are
the standard components of an irreducible J-th order tensor operator
T
(J)
, then they
transform under rotation according to the law:
U
T
(J)
M
U
1
=
T
(J)
M
R
(J)
M
M
(6.11)
where R
(J)
M
M
are the matrices of rotation for the states in the J representation.
This means that these matrix elements are the same as those which enter the trans-
formation:
U [J, M =
[J, M
R
(J)
M
M
(6.12)
Using Eqn.(6.12) and (6.11) it is easy to prove that the operators
T
(J)
M
have the
6.5. Wigner-Eckart Theorem 153
following commutation relations with the generators
J
,
J
z
[
J
,
T
(J)
M
] =
_
J(J + 1) M(M 1)
T
(J)
M1
(6.13)
and
[
J
z
,
T
(J)
M
] = M
T
(J)
M
(6.14)
6.5 Wigner-Eckart Theorem
What we shall do in this section is to give the general form of the matrix elements
of irreducible tensor operators. Somehow this is the analogue of what we did in
Eqn. (6.9) for scalar operators. For convenience of notation in this section we will
indicate the irreducible tensors with the notation:
T
(K)
q
which means we are in the
(K) representation and take the q element. Let us now indicate with
J
, K, M
, q[JM (6.15)
the Clebsh-Gordon coecients which in the full notation we would write as:
J
, K, M
, q[J
, K, J, M (6.16)
where J
J
2
, J
z
are not a complete set of operators and that, in order to specify uniquely the
states, we need some extra quantum numbers which we will indicate with . It is
then possible to prove the following relation:
, J, M[
T
(K)
q
[
, J
.M
=
1
_
(2J + 1)
, J[ [T
(K)
[ [
, J
, K, M
, q[J, M
(6.17)
The quantity that we have simbolically indicated with , J[ [T
(K)
[ [
, J
is called
reduced matrix elements and it is a number which depends only on (J, J
, ,
, K)
but not on (M, M
(6.18)
and
[J J
[ K (J + J
) (6.19)
This is so because on the R.H.S of Eqn. (6.17) we have a Clebsh-Gordon coecient
which is dierent from zero only if the relations (6.18) and (6.19) are satised. Let
us apply all of this to a vector-operator
V . It is called vector operator because, as
we explained in previous section, under the rotation generators
J it transform like a
vector which means it has the following commutation relations with
J:
[
J
x
,
V
x
] = 0, [
J
x
,
V
y
] = i
V
z
, [
J
x
,
V
z
] = i
V
y
(6.20)
and cyclically with the other components of
J. We can also build the
V
and
J
and
derive from (6.20) the following set of commutation relations:
[
J
x
,
V
] =
V
z
, [
J
y
,
V
] = i
V
z
, [J
z
, V
] =
(6.21)
and
[
J
+
,
V
+
] = 0, [
J
+
,
V
] = 2
V
z
, [
J
,
V
+
] = 2
V
z
, [
J
,
V
] = 0 (6.22)
Let us now calculate the matrix elements of
V
z
between J, M[ and [J
, M
, i.e.:
J
[
V
z
[J, M (6.23)
We will see that a lot of these elements will be zero. For example it is easy to prove
that if M ,= M
the matrix elements above are zero. The proof goes as follows: we
know from the commutation relations above that [
V
z
,
J
z
] = 0 and if we sandwich it
6.5. Wigner-Eckart Theorem 155
among the states we get: J
, M
[ [
V
z
,
J
z
] [J, M = 0. Applying
J
z
on the states on
the left and on the right we get:
J
, M
[ [
V
z
,
J
z
] [J, M = (M M
) J
[
V
z
[J, M = 0
Now if we choose M ,= M
[
V
z
[J, M = 0.
Another easy thing to prove is that the matrix elements: J, M[ V
[J, M
are dierent
from zero if and only if M
] =
[J, M
=
1
JM[ [
J
z
,
V
] [J, M
) J, M[
V
[J, M
= 0
As a consequence, if M ,= M
[J, M
= 0
We will now prove the important result that the matrix elements of
V are propor-
tional to the matrix elements of
J. Let us remember from (6.22) that
J
+
and
V
+
commutes so their matrix elements between any state are zero. Let us in particular
sandwich it between J, M + 2[ and [J, M and we get:
J, M + 2[
J
+
V
+
[J, M = J, M + 2[
V
+
J
+
[J, M .
Let us now insert a completeness
[J
, M
[
V
+
[J, M = (J) J, M
[
J
+
[JM (6.25)
It is easy to prove that this proportionality can be extended to all the components
of
V with the same proportinality factor (J). Let us prove this for the z-component.
We start from the commutation relation:
[
J
,
V
+
] = 2
V
z
(6.26)
and let us sandwich both sides between J, M[ and [J, M. Acting with the
J
which
is inside (6.26) on the bra and the ket on which we have sandwiched the above
commutator, we obtain for the LHS of (6.26):
_
J(J + 1) M(M + 1) J, M + 1[
V
+
[J, M
_
J(J + 1) M(M 1) J.M[
V
+
[J, M 1
This last expression, once we use (6.25) in it, becomes:
_
J(J + 1) M(M + 1) J, M + 1[
J
+
[J, M
_
J(J + 1) M(M 1) J, M[
J
+
[J, M 1
6.5. Wigner-Eckart Theorem 157
This expression can be easily worked out and we get that it is equal to 2M
which can be written as 2(J) J, M[
J
z
[J, M . So, going now back to (6.26), we can
conclude that:
JM[
V
z
[JM = (J) J, M[
J
z
[J, M
which is what we wanted to prove. In a similar manner we can prove the same
proportionality for all the other components of
V . This is basically the Wigner-
Eckart theorem for vector operators.
Chapter 7
SPIN
In the quantization of angular momentum, or better, of the operators
J
i
satisfying
the same algebra of angular momentum operators, namely
[
J
i
,
J
j
] = i
ijk
J
k
,
we proved that we can get only integer or half-integer values. Now, we are interested in
trying to check if the half-integer cases can be associated to some physical phenomena.
Of course, half-integer values cannot be related to orbital angular momentum. In fact,
suppose we perform a rotation about z on some wave function (). We know the
unitary operator corresponding to this rotation is given by
U() = e
Lz
,
being the angle of rotation. Now, suppose we select () as an eigenstate of
L
z
with eigenvalue /2, i.e.,
L
z
1
2
() =
1
2
1
2
() ,
and let us performe a rotation on 1
2
of an angle of 2:
U(2)1
2
()
. .
1
2
(+2)
= e
Lz2
1
2
()
= e
2
1
2
1
2
()
= 1
2
() .
We expect to come back to the same state instead this does not happen. There are
two possible ways of overcome this diculty: either 1
2
() cannot be an eigenstate
160 Chapter 7. Spin
Fig. 7.1. Illustration of the Stern-Gerlach experiment.
of
L
z
or
L
z
is not the operator of rotation. Both are absurd conclusions. Thus there
must be a dierent interpretation of the half-integer eigenvalues.
In 1922 Stern and Gerlach performed a well-known experiment in which a beam
of atoms having only one valence electron was sent through a strongly inhomogeneus
magnetic eld like in Fig. 7.1. What did Stern and Gerlach nd? The beam splitted
into two beams!
Such results had been already explained within the framework of Bohr model by
Uhlenbeck and Goudsmit by assuming that the electron carries an intrinsic angular
momentum, called spin, whose components can take only the values /2 and to
which it corresponds an intrinsic magnetic moment
2
e
m
0
c
.
A quantity which is used to measure magnetic moments is the so-called Bohr
magneton
0
, dened by
0
e
2m
0
c
. (7.1)
When a charged particle (with charge e) rotates with (orbital) angular momentum
m then its magnetic moment due to the orbital angular momentum is m
0
= . We
will return on this issue when we will talk about the Zeeman eect.
If we measure the angular momentum in units of and the magnetic moment in
units of the Bohr magneton then the ratio
/
0
L
z
/
= g ,
is called gyromagnetic ratio and is equal to 1. On the contrary, for the spin it
7.1. Quantum description of particles carrying spin 161
comes out that this ratio is 2. This is called anomalous gyromagnetic ratio and it
is explained by the Dirac relativistic equation.
The interaction Hamiltonian between a magnetic moment and a magnetic eld is
H
I
= B . (7.2)
Thus, if the magnetic moment is generated by a rotation of a charge, the Hamiltonian
is
H
I
= B ,
while in the case in which the magnetic moment is due to spin we get
H
I
= 2 B .
In describing the spin we used at the beginning the expression intrinsic angular
momentum but this is misleading. The electron seems to be a point-like particle
at the present energy (7 TeV) and so the picture of something rotating on itself is
not appropriate (at least at the present energy). What is more appropriate to say is
that the electrons and many other particles behave as small magnets endowed with a
magnetic moment which at the quantum level becomes a set of three operators with
the same algebra of the one of the angular momentum. But this does not imply that
this quantum number is a sort of angular momentum.
Other quantum observables that you will encounter in particle physics, like the
isotopic spin etc., have the same algebra as angular momentum but have nothing to
do with it.
7.1 Quantum description of particles carrying spin
Since spin is not a space-time property of a particle, how can we realize the wave
function of a particle with spin?
We have seen that the algebra associated to spin is
[ s
i
, s
j
] = i
ijk
s
k
, (7.3)
162 Chapter 7. Spin
s
z
can only take the values /2 the number of states in the multiplet is 2l + 1 = 2.
This means the matrices s
i
are two-by-two matrices. Two-by-two matrices satisfying
the algebra (7.3) have been worked out in section 5.1 and take the form
s
i
=
2
i
, (7.4)
where
i
are the so-called Pauli matrices whose form is
x
=
_
_
0 1
1 0
_
_
,
y
=
_
_
0 i
i 0
_
_
,
z
=
_
_
1 0
0 1
_
_
.
Their form can be obtained via simple arguments as follows. s
z
must have eigenvalues
/2, therefore
z
must have eigenvalues 1 as it is clear from Eqn. (7.4). Since z
is an arbitrary direction, and we could choose the x and y directions instead of z,
then also
x
and
y
must have eigenvalues 1 . If this is so, their squares have only
eigenvalues equal to 1 and thus they must be the identity matrix:
2
x
=
2
y
=
2
z
= 11 =
_
_
1 0
0 1
_
_
. (7.5)
Now, consider the commutation rule
y
z
z
y
= 2i
x
.
We multiply this expression by
z
from the right and we make use of Eqn. (7.5):
y
z
y
z
= 2i
x
z
.
If we multiply instead from the left we get
z
y
z
y
= 2i
z
x
.
By summing the two previous expressions, it follows
2i (
x
z
+
z
x
) = 0 ,
7.1. Quantum description of particles carrying spin 163
i.e.,
x
,
z
= 0 ,
where , denotes the anti-commutator. In the same way it is possible to prove that
in general:
i
,
j
=
i,j
11 .
Using this formula and the commuation rules
[
i
,
j
] = 2i
ijk
k
,
it is not dicult to prove that
r
s
=
r,s
11 + i
rsk
k
,
where, in the last term on the R.H.S, summation over k is understood.
Now, we choose the representation in which
z
is diagonal and we keep in mind
that its eigenvalues are 1. Thus,
z
=
_
_
1 0
0 1
_
_
s
z
=
_
_
2
0
0
2
_
_
.
In order to obtain
x
in this representation, we write
x
in the form
x
=
_
_
a b
c d
_
_
.
From the anti-commutation rules we must have:
x
z
=
z
x
,
so, using
z
= (
1 0
0 1
), we get a = d = 0 and nally, from
2
x
= 11, we obtain bc = 1.
In fact,
_
_
a b
c d
_
_
_
_
1 0
0 1
_
_
=
_
_
a b
c d
_
_
=
_
_
1 0
0 1
_
_
_
_
a b
c d
_
_
=
_
_
a b
c d
_
_
,
164 Chapter 7. Spin
from which it follows a = d = 0, and from
2
x
= 11 we get
_
_
0 b
c 0
_
_
_
_
0 b
c 0
_
_
=
_
_
bc 0
0 bc
_
_
=
_
_
1 0
0 1
_
_
bc = 1 .
Since these matrices must be Hermitian, c
+
(x, y, z, t)
(x, y, z, t)
_
_
.
Sometimes, another notation is used to denote this state, namely, (x, y, z, t; w),
where w is the spin quantum number. Note that can be written as
=
_
_
+
0
_
_
+
_
_
0
_
_
,
where
z
_
_
+
0
_
_
=
_
_
1 0
0 1
_
_
_
_
+
0
_
_
=
_
_
+
0
_
_
,
7.2. Pauli equation 165
i.e.,
_
+
0
_
is eigenstate of
z
with eigenvalue +1. In the same way one proves that
_
0
_
is eigenstate of
z
with eigenvalue 1:
z
_
_
0
_
_
=
_
_
1 0
0 1
_
_
_
_
0
_
_
=
_
_
0
_
_
.
So,
+
(x, y, z) is a state of the electron such that we have a probability 1 of nd-
ing +/2 in a measurement of s
z
. Therefore, [
+
(x, y, z)[
2
is the probability den-
sity of nding the particle in (x, y, z) with component +/2 of spin along z, while
[
(x, y, z)[
2
is the probability density to localize the particle at (x, y, z) and with
/2 component of the spin along z.
The Hamiltonian of a particle with charge e and intrinsic magnetic moment
within an electromagnetic eld generated by the scalar potential V (x, y, z) and the
vector potential A(x, y, z) is given by
H =
1
2m
_
p
e
c
A
_
2
eV
. .
H
0
+B
s
, (7.6a)
where
s
=
0
and so we have already taken account of the anomalous gyromagnetic
ratio. H
0
is found by applying the minimal coupling prescription. H can be rewritten
in terms of Pauli matrices as follows:
H = H
0
+
0
B , (7.6b)
where
0
is the Bohr magneton.
The Schr odinger equation for system with spin is called Pauli equation:
i
t
=
H , (7.7)
where =
_
_
, with the normalization condition:
_
V
_
[
+
[
2
+[
[
2
_
dV = 1 .
166 Chapter 7. Spin
Eqn. (7.7), rewritten in components, reads
H
0
_
_
_
_
+
0
B
_
_
_
_
= i
t
_
_
_
_
.
Moreover we have
B = B
x
x
+ B
y
y
+ B
z
z
=
_
_
B
z
B
x
iB
y
B
x
+ iB
y
B
z
_
_
,
and by inserting this in the Pauli equation we get
H
0
+
+
0
B
z
+
+
0
(B
x
iB
y
)
= i
+
t
,
H
0
+
0
(B
x
+ iB
y
)
+
0
B
z
= i
t
.
Let us assume that the magnetic eld is uniform and lets try to separate the spatial
part of the wave function from its spin part. We write
=
0
(x, y, z)
s
,
s
=
_
_
a
b
_
_
,
where
0
(x, y, z) are solutions of the Schr odinger equation associated to
H
0
. If we
solve the stationary Pauli equation using the ansatz
u
n
(x, y, z; w) = u
0
n
(x, y, z)
s
,
where u
0
n
are eigenfunctions of H
0
, the stationary Pauli equation splits into
H
0
u
0
n
(x, y, z) = E
0
n
u
0
n
(x, y, z) , (7.8a)
0
B
s
= E
s
s
, (7.8b)
and the total energy is
E
n
= E
0
n
+ E
s
.
By inserting the expression for
s
= (
a
b
) Eqn. (7.8b) becomes
0
B
z
a +
0
(B
x
iB
y
)b = E
s
a , (7.9a)
0
(B
x
+ iB
y
)a
0
B
z
b = E
s
b . (7.9b)
7.2. Pauli equation 167
The solutions a, b will be dierent from zero if the following determinant vanishes:
det
0
B
z
E
s
0
(B
x
iB
y
)
0
(B
x
+ iB
y
)
0
B
z
E
= 0 ,
from which it follows
E
s
=
0
[B[ , [B[ =
_
B
2
x
+ B
2
y
+ B
2
z
.
By substituting these values into Eqs. (7.9) we nd, for E
s
= +
0
B, the eigenfunc-
tions:
1
=
_
_
a
1
b
1
_
_
,
where
a
1
=
[B[ + B
z
2 [B[
e
i
, b
1
=
[B[ B
z
2 [B[
e
i
_
+arctan
By
Bx
_
,
and is an arbitary real phase, while for E
s
=
0
B we get
2
=
_
_
a
2
b
2
_
_
,
where
a
2
=
[B[ B
z
2 [B[
e
i
, b
2
=
[B[ + B
z
2 [B[
e
i
_
+arctan
By
Bx
_
.
In particular, if B = (0, 0, [B[), i.e.the magnetic eld is along z, the solutions are
1
_
_
1
0
_
_
,
2
_
_
0
1
_
_
.
The corresponding energies will be dierent: E
n
= E
n
0
[B[.
The total angular momentum of the system, if besides spin we have also the orbital
angular momentum, is one of the two following values: j = l
1
2
. More properly,
j characterizes the total magnetic moment, since spin is not actually an angular
momentum as we explained at the beginning.
Spin appears not only in the Stern-Gerlach experiment, but also it plays a role in
the explanation of the anomalous Zeeman eect, which will be presented after having
discussed the Hydrogen atom.
168 Chapter 7. Spin
7.3 Singlet and triplet
Consider two spin-1/2 particles. According to the rules of the sum of operators like
angular momentum, we know that the sum can take only the values 1 and 0. Let us
now look for the wave functions
s,sz
of the total spin and its z-component.
The form of the wave functions
(1,0)
and
(0,0)
can be sought according to one of
the two following arguments:
1. Directly from the equation for the eigenfunctions of s
2
(square of total spin);
2. or by using the shift operators s
.
We use method 1. As homework solve the exercise using the method 2.
The z-component for s
tot
= 1 can be one of the following three values:
s
z
= 1, 0, 1 ,
while in the case s
tot
= 0 the only possibility is s
z
= 0. Thus we have four states:
s,sz
=
1,1
,
1,0
,
1,1
,
0,0
.
Let us now construct the explicit form of these wave functions. The single-particle
wave functions are
_
_
1
0
_
_
(1)
and
_
_
0
1
_
_
(1)
,
and
_
_
1
0
_
_
(2)
and
_
_
0
1
_
_
(2)
.
The wave-function
1,1
must then be of the form
1,1
=
_
_
1
0
_
_
(1)
_
_
1
0
_
_
(2)
,
7.3. Singlet and triplet 169
where with we indicate the tensor product. The total spin operator is
S
(tot)
z
=
S
(1)
z
+
S
(2)
z
the operator
S
(1)
z
will act only on the rst part of the tensor product above while
S
(2)
z
will act only on the second part, and therefore, by putting = 1, we get:
S
(tot)
z
1,1
=
_
S
(1)
z
+
S
(2)
z
_
_
_
_
_
1
0
_
_
(1)
_
_
1
0
_
_
(2)
_
_
=
_
_
1
2
_
_
1
0
_
_
(1)
_
_
1
0
_
_
(2)
+
1
2
_
_
1
0
_
_
(1)
_
_
1
0
_
_
(2)
_
_
= 1
_
_
_
_
1
0
_
_
(1)
_
_
1
0
_
_
(2)
_
_
= 1
1,1
.
Since s
tot
can only be zero or one, a wave function with s
z
= 1, like the one above,
can only belong to s
tot
= 1.
In the same way on can easily construct the eigenstate
1,1
, which is
1,1
=
_
_
0
1
_
_
(1)
_
_
0
1
_
_
(2)
.
Let us now turn to the problem of constructing the state
0,0
. As it has a total s
z
= 0,
by following the previous rules it is easy to see that such a state will be of the form:
0,0
=
_
_
1
0
_
_
(1)
_
_
0
1
_
_
(2)
+
_
_
0
1
_
_
(1)
_
_
1
0
_
_
(2)
. (7.10)
The coecients and will be determined imposing that
0,0
satises also s
(tot)
= 0,
i.e.,
_
S
(tot)
_
2
0,0
= s(s + 1)
0,0
= 0 ,
170 Chapter 7. Spin
from which it follows
0,0
[
_
S
(tot)
_
2
[
0,0
= 0 ,
and thus
_
_
_
S
(tot)
x
0,0
_
_
_
2
+
_
_
_
S
(tot)
y
0,0
_
_
_
2
+
_
_
_
S
(tot)
z
0,0
_
_
_
2
= 0 .
Since the three terms on the left-hand side are norms, the only way the previous
expression will be fullled is to require that each of the three terms be individually
equal to zero. For instance, let us evaluate
S
x
0,0
=
1
2
_
(1)
x
+
(2)
x
_
_
1
0
_
_
(1)
_
_
0
1
_
_
(2)
+
_
_
0
1
_
_
(1)
_
_
1
0
_
_
(2)
_
_
.
By using
x
= (
0 1
1 0
) whose action is
_
_
0 1
1 0
_
_
_
_
1
0
_
_
=
_
_
0
1
_
_
,
_
_
0 1
1 0
_
_
_
_
0
1
_
_
=
_
_
1
0
_
_
,
we get
S
x
0,0
=
1
2
( + )
_
_
_
_
1
0
_
_
(1)
_
_
1
0
_
_
(2)
+
_
_
0
1
_
_
(1)
_
_
0
1
_
_
(2)
_
_
.
For this to be zero we must have = . So, the normalized state
0,0
is given by
0,0
=
1
2
_
_
_
_
1
0
_
_
(1)
_
_
0
1
_
_
(2)
_
_
0
1
_
_
(1)
_
_
1
0
_
_
(2)
_
_
.
Finally, let us construct
1,0
. Also
1,0
will be of the form (7.10) but it must
be orthogonal to
0,0
since, having dierent s, they belong to dierent eigenvalues
of (
S
(tot)
)
2
and we know that eigenstates belonging to dierent eigenvalues of an
Hermitian operator are orthogonal. The only normalized state of the form (7.10) and
orthogonal to
0,0
is
1,0
=
1
2
_
_
_
_
1
0
_
_
(1)
_
_
0
1
_
_
(2)
+
_
_
0
1
_
_
(1)
_
_
1
0
_
_
(2)
_
_
.
7.4. Explanation of the Stern-Gerlach experiment 171
0,0
is referred to as singlet and it is an anti-symmetric state with respect to an
interchange of the two particles, while
1,0
is referred to as triplet and it is symmetric
under the interchange of the two particles. They are called this way because s
(tot)
= 1
has three states while s
(tot)
= 0 has just one.
7.4 Explanation of the Stern-Gerlach experiment
An uniform magnetic eld B produces a torque on a magnetic dipole
q
, given by
=
q
B ,
but it does not produce any force. In order to get a force on a magnetic dipole a
non-uniform magnetic eld is needed. This is analogous to the case of an electric
dipole: if the electric eld is uniform, both charges q and q feel exactly the same
force (but opposite in direction) and so the net force acting on the dipole is zero. If
instead the eld is non-uniform one of the two charges , being located in a dierent
position than the other, will experience a force dierent than the one acting on the
other and the net force on the dipole will be non-vanishing. In the case of an electric
dipole the force is
F = d
z
E
z
,
where d
z
= ez is the electric dipole moment along z. (For simplicity, we are assuming
the electric eld to vary only in the z-direction.)
A similar force also occurs in the case of a magnetic dipole in a non-uniform mag-
netic eld:
F =
z
B
z
.
In the Stern-Gerlach experiment, electrons have a magnetic moment due to spin, and
this magnetic moment can only take two values
0
because, taking account of the
anomalous gyromagnetic ratio, we have:
z
=
0
z
.
172 Chapter 7. Spin
The force acting on the electrons will then be
F
z
= +
0
B
z
or F
z
=
0
B
z
,
according to the components of the spin we are considering. As the forces have oppo-
site directions, they will split the beam into two components. This is the explanation
of the Stern-Gerlach experiment.
Chapter 8
IDENTICAL PARTICLES
8.1 General considerations
In this chapter we will deal with systems having two or more identical particles.
In classical mechanics two particles are said to be identical if the Hamiltonian is
invariant under the interchange of the corresponding coordinates of the two particles
in the phase space, i.e.,
H(
(1)
,
(2)
) = H(
(2)
,
(1)
) .
Since in classical mechanics we can speak about trajectories (in phase space), two
identical particles can be distinguished when the initial points of their trajectories are
dierent. In fact, the trajectories (in the phase space) of two particles whose starting
points are dierent will never cross, so there is no ambiguity at all in identifying the
trajectory of one particle and the one of the other .
Things are dierent in quantum mechanics. Indeed, in quantum mechanics we are
not dealing with points and trajectories in phase space but with states in an Hilbert
space, and even if we prepare the initial states of the two particles at t = 0 localized
one very far away from the other, time-evolution will spread the spatial distribution
of such states and at some time in the future there will be some overlapping of the
two wave functions. In the region of overlapping we will not be able to say if the wave
function belongs to particle 1 or 2. So, the problem of identical particles must be
studied in a more complete way in quantum mechanics. We will see in this chapter
that such considerations will lead us to new interesting results.
174 Chapter 8. Identical particles
Following the general rules of quantum mechanics, let us prepare the system of two
particles by identifying a complete set of commuting operators. Let us denote the
operators of this set in the following way:
O
1
(
1
,
2
) ,
O
2
(
1
,
2
) ,
O
3
(
1
,
2
) , . . . ,
O
n
(
1
,
2
) , (8.1)
where
1
and
2
label variables corresponding to the particle 1 and 2, respectively.
Two particles are called identical if the set (8.1) is invariant under the exchange of
1
and
2
, i.e.,
O
1
(
1
,
2
) =
O
1
(
2
,
1
) ,
O
2
(
1
,
2
) =
O
2
(
2
,
1
) ,
O
3
(
1
,
2
) =
O
3
(
2
,
1
) ,
.
.
.
O
n
(
1
,
2
) =
O
n
(
2
,
1
) .
(8.2)
As a consequence, if the state of the two particles is prepared in the state (
1
,
2
)
then the state prepared using the set of operators with
1
and
2
interchanged will be
(
2
,
1
) and these two wave functions must be equivalent, i.e., they must be equal
up to a phase:
(
1
,
2
) = e
i
(
2
,
1
) . (8.3)
If we interchange again the two particles we get
(
1
,
2
) = e
i
e
i
(
1
,
2
) ,
and thus e
2i
= 1, so e
i
= 1. From this it follows
(
1
,
2
) = (
2
,
1
) , (8.4)
which means the wave functions corresponding to two identical particles must be
either symmetric or anti-symmetric under the exchange of the two particles.
Let us introduce an operator
P which exchanges the two particle, i.e which acts in
the following way on the wave function of two identical particles:
P(
1
,
2
) = (
2
,
1
) .
8.1. General considerations 175
It is clear that
P
2
= I ,
thus the eigenvalues of
P are 1. It is also easy to prove that
P =
P
=
P
1
(work
it out as an homework). Let
O be any observable, then the action of
P on
O will be
O
P =
P
1
O
P .
Clearly,
O(
1
,
2
)
P =
O(
2
,
1
) ,
and, since for identical particles
O(
1
,
2
) =
O(
2
,
1
), we get
O(
1
,
2
)
P =
O(
1
,
2
) ,
which is equivalent (after having multiplied by
P on the left) to:
O(
1
,
2
)
P =
P
O(
1
,
2
) ,
i.e.,
P commutes with the whole complete set of observables. Therefore, we have
two possibilities: either
P is a function of
O
1
,
O
2
, . . . ,
O
n
or the set is not complete.
For identical particles, the latter is what happens: for identical particles,
P must be
added to the initial set of operators in order to get a complete set.
When preparing the initial state, being
P now part of the complete set of observ-
ables, it must be diagonalized together with the other observables of the set. Since
its eigenvalues are 1, we immediately get symmetric or anti-symmetric states. But
which states do we have to select?
The answer to the previous question is provided by quantum eld theory through
a well-known theorem called spin-statistics theorem. Since we have not yet studied
quantum eld theory, we take the content of that theorem as a new postulate for QM:
176 Chapter 8. Identical particles
Spin-statistics postulate: Identical particles of half-integer spin must have
anti-symmetric wave functions (fermions). Identical particles of integer spin must
have symmetric wave functions (bosons).
The basic ingredients of quantum eld theory that enable us to prove such result
are:
1. Lorentz invariance;
2. locality of elds;
3. existence of a ground state for
H.
What about a system made by one boson and one fermion? The spin-statistics pos-
tulate does not apply to this case since the two particles are no longer identical (one
of them is a boson, the other is a fermion).
8.2 Pauli exclusion principle
As a consequence of the spin-statistics postulate, two spin one-half particles never hap-
pen to be in the same state. In fact, if we denote with
Q
1
and
Q
2
the single-particle
states of the two particles, where Q
1
and Q
2
are the quantum numbers corresponding
to the single-particle wave functions of the rst and the second particle respectively,
the anti-symmetric wave function of the total system of the two particles is
(
1
,
2
) =
1
2
[
Q
1
(
1
)
Q
2
(
2
)
Q
1
(
2
)
Q
2
(
1
)] . (8.5)
If the states of the two particles were the same, i.e., if Q
1
= Q
2
(same quantum
numbers), we would get
(
1
,
2
) =
1
2
[
Q
1
(
1
)
Q
1
(
2
)
Q
1
(
2
)
Q
1
(
1
)] = 0 ,
8.3. Exchange interaction 177
therefore the probability to nd the two particles in the same state vanishes. This
result is known as Pauli exclusion principle, and it explains various things con-
cerning atomic systems. For example, for each atomic level we can have no more
than two electrons, one having s
z
=
1
2
and the other with s
z
=
1
2
, where s
z
is the
spin z-component. If we had a third electron, this would have to occupy the next
atomic level. It is as if the third electron felt a repulsive force from the core of the
previous two electrons. The Paulis exclusion principle is responsible for much of the
periodic table of elements. Another consequence of the Pauli exclusion principle is
the so-called exchange energy, discussed in the next section.
8.3 Exchange interaction
Even if no spin-dependent potential were included into the Schr odinger equation of
two identical fermions, the eect of spin would appear in the energy of some levels
via the Paulis exclusion principle. Indeed, such eect is measurable.
Let us consider two spin-half particles interacting through a central potential U([r
1
r
2
[), where r
1
denotes the vector position of the rst particle and r
2
that of the
second one. The total wave-function (in both spin and spatial variables) must be
anti-symmetric, because of the spin-statistics postulate. For s
(tot)
= 0 (singlet) we
already know, from the previous chapter, that the spin part of the wave-function is
anti-symmetric so the spatial (i.e., orbital) wave-function must be symmetric:
s=0
=
1
2
[
1
(r
1
)
2
(r
2
) +
1
(r
2
)
2
(r
1
)] , (8.6)
while for s = 1 the spin part is symmetric and so the orbital part must be anti-
symmetric:
s=1
=
1
2
[
1
(r
1
)
2
(r
2
)
1
(r
2
)
2
(r
1
)] . (8.7)
178 Chapter 8. Identical particles
Now, let us calculate the mean value of the potential energy of these two states:
_
[
s=0
U(r
1
r
2
)
s=0
] d
3
r
1
d
3
r
2
= A+ J ,
_
[
s=1
U(r
1
r
2
)
s=1
] d
3
r
1
d
3
r
2
= AJ ,
where
A =
_
U [
1
(r
1
)[
2
[
2
(r
2
)[
2
d
3
r
1
d
3
r
2
,
and
J =
_
U
1
(r
1
)
1
(r
2
)
2
(r
2
)
2
(r
1
) d
3
r
1
d
3
r
2
.
The latter integral is referred to as exchange energy.
The previous result shows that the singlet and triplet states have dierent (mean)
potential energies. This eect is only due to the spin-statistics postulate, and it is an
eect which has been measured.
8.4 Slaters determinant
If we are dealing with N identical spin one-half particles instead of only two, how
can we construct the most general anti-symmetric wave-function? The answer is the
following:
(
1
,
2
, . . . ,
N
) =
1
N!
Q
1
(
1
)
Q
1
(
2
)
Q
1
(
N
)
Q
2
(
1
)
Q
2
(
2
)
Q
2
(
N
)
.
.
.
.
.
.
.
.
.
.
.
.
Q
N
(
1
)
Q
N
(
2
)
Q
N
(
N
)
, (8.8)
where we must nd at least N dierent combinations Q
1
, Q
2
, . . . .Q
N
of single-particle
quantum numbers in order to have N dierent single-particle wave-functions. If we
cannot nd such N dierent quantum numbers, (
1
,
2
, . . . ,
N
) is obviously zero.
The wave-function above is known as Slater determinant.
8.5. Entangled states 179
8.5 Entangled states
Let us suppose we have two particles 1 and 2 and the corresponding Hilbert spaces
1
1
and 1
2
. The states of the system of the two particles belong to the tensor product
of the two one-particle Hilbert spaces:
1
(tot)
= 1
1
1
2
.
Some wave-functions belonging to this Hilbert space will be of the form:
(tot)
(1, 2) =
1
2
but not all of them. In fact in the Hilbert space 1
(tot)
we can also take linear
combinations of such wave-functions such as:
(tot)
(1, 2) =
1
2
+
2
.
In general this is not be the tensor product of two states belonging to the Hilbert
spaces 1
1
and 1
2
, i.e., in general
(tot)
(1, 2) =
1
2
+
2
,= (1) (2) .
If it is not the product , then the state of the two particles is referred to as an entangled
state.
Looking at the state above, we see that we are not able to answer the question
in which state is one of the two particles: particle 1 could be in
1
or in
1
and
analogously for particle 2. The entangled states are one of the most peculiar feature
of Quantum Mechanics, as Schr odinger pointed out in 1935. They are at the origin
of the many non-local features of Quantum Mechanics.
Chapter 9
HYDROGEN ATOM
We have already encountered the hydrogen atom studying central potentials in
section 1.7. In that occasion, we have approached the Schr odinger equation by using
the method of the separation of variables. The result was that the wave function in
spherical coordinates could be written in the form
2
[E U(r)] r
2
= , (9.2a)
1
Y sin
_
sin
Y
_
+
1
Y sin
2
2
Y
2
= , (9.2b)
where is some arbitrary constant. The latter equation above now looks familiar,
in fact by writing the angular momentum operator
L
2
=
L
2
x
+
L
2
y
+
L
2
z
in spherical
coordinates we get
L
2
=
2
_
1
sin
_
sin
_
+
1
sin
2
2
_
, (9.3)
and Eqn. (9.2b) can be rewritten as
1
sin
_
sin
Y (, )
_
+
1
sin
2
2
Y (, )
2
= Y (, ) ,
i.e., as
L
2
Y (, ) =
2
Y (, ) .
182 Chapter 9. Hydrogen atom
This means that Y (, ) are nothing more that the eigenfunctions of
L
2
and we
already know from the general theory of angular momentum that their eigenvalues
are of the form l(l + 1), therefore
= l(l + 1) ,
and the functions Y (, ) are the spherical harmonics Y
lm
(, ).
Inserting the value above for , Eqn. (9.2a) becomes
1
R(r)
d
dr
_
r
2
dR(r)
dr
_
+
2m
2
[E U(r)] r
2
= l(l + 1) .
If we parametrize R(r) as
R(r) =
(r)
r
,
the (r) satises the equation:
2
2m
d
2
(r)
dr
2
+
_
U(r) +
l(l + 1)
2
2mr
2
_
(r) = E(r) . (9.4)
Note that this looks like a one-dimensional Schr odinger equation in the coordinate r
for a potential given by U(r) plus a centrifugal term
l(l+1)
2
2mr
2
.
The original Hamiltonian of our system, having a potential which depends only on
r, is invariant under rotations and thus commutes with both
L
2
and
L
z
. It is therefore
possible to diagonalize simultaneously each of these three operators,
H,
L
2
,
L
z
:
_
H = E ,
L
2
= [l(l + 1)]
2
,
L
z
= m .
(9.5)
can then be labelled by the three numbers: (E, l, m). The last two equations in (9.5)
are those having as solution the spherical harmonics Y
lm
(, ).
Let us now turn back to the radial equation, i.e., Eqn. (9.4). The wave function
normalization condition reads
_
+
0
drr
2
_
0
dsin
_
2
0
d
(r, , )
2
< .
183
which, once the wave function is written in the factorized form
(r, , ) = R(r)Y (, ) ,
and taking into account the normalization of the spherical harmonics, yields
_
0
drr
2
[R(r)[
2
< .
Let us remember that we wrote R(r) =
(r)
r
, so from the above relation we get
_
0
dr ((r))
2
< ,
and we also gather that (r) must go to zero as r 0 at least as fast as r:
(r)
r0
O(r) .
Using (r) we can say that solving our problem is like solving the one-dimensional
problem of a particle which feels the potential
U
e
(r) = U(r) +
l(l + 1)
2
2mr
2
,
for r > 0 while for r < 0 the potential is that of an innite potential well, because we
are requiring (r) to be zero at r = 0.
We now choose the radial equation to be the one of the electron in a hydrogen
atoms which feels the Coulomb potential given by
U(r) =
e
2
r
, (9.6)
or in general to be the one of an electron spinning around a nucleus of charge Ze:
U(r) =
Ze
2
r
. (9.7)
The radial equation in R(r) becomes
2
2m
1
r
2
d
dr
_
r
2
dR
dr
_
Ze
2
R
r
+
l(l + 1)
2
2mr
2
R = ER . (9.8)
184 Chapter 9. Hydrogen atom
m is not the mass of the electron but its reduced mass:
m =
m
e
m
N
m
e
+ m
N
m
e
.
Now we shall introduce the following dimensionless expressions :
r , where
2
8m[E[
2
,
and let us denote with
the following quantity:
2mZe
2
2
Ze
2
_
m
2 [E[
_
1/2
.
The radial equation becomes
1
2
d
d
_
2
dR
d
_
+
_
1
4
l(l + 1)
2
_
R = 0 . (9.9)
It is easy to prove that for suciently large , an R() of the form R()
n
e
1
2
satises the above equation. In fact, the kinetic term splits into three parts:
_
A
n1
+ B
n
+ C
n2
1
2
(9.10)
and the same happens for the potential term, thus it is always possible to determine
and l in such a way to make the kinetic and potential part equal and opposite to
each other. Since we are assuming R() 0 for r in order to have properly
normalized wave functions,
_
0
drr
2
R
2
(r) < ,
the solution with the plus sign in (9.10) cannot be taken. Therefore, we will seek for
an exact solution of the form
R() = F() e
1
2
,
where F() is a polynomial of nite order in .
185
The equation for R() becomes one for F():
F
() +
_
2
1
_
F
() +
_
l(l + 1)
2
_
F() = 0 . (9.11)
Let us choose F() of the form
F() =
s
_
a
0
+ a
1
+ a
2
2
+ . . .
. .
L()
,
with a
0
,= 0 and s 0. The condition s 0 is needed for R() to be nite for = 0.
The equation for the expression L() becomes
2
L
+ [2(s + 1) ] L
+
_
(
s 1) + s(s + 1) l(l + 1)
_
L = 0 . (9.12)
If = 0, Eqn. (9.12) turns out to be simply:
[s(s + 1) l(l + 1)] L = 0 . (9.13)
L was given by
L() = a
0
+ a
1
+ a
2
2
+ . . . .
With a
0
,= 0 we get L(0) = a
0
,= 0, and as a consequence Eqn. (9.13) implies
s(s + 1) l(l + 1) = 0 which has the following two solutions:
s = l , s = (l + 1) .
However, the latter, namely s = (l +1), is not acceptable since R() would diverge
for 0:
R() =
(l+1)
_
a
0
+ a
1
+ a
2
2
+ . . .
1
2
0
.
So, the only acceptable solution is s = l and the equation for L(), [Eqn. (9.12)],
becomes
2
L
+ [2(l + 1) ] L
+ (
l 1)L = 0 .
186 Chapter 9. Hydrogen atom
By inserting the expression for L() = a
0
+ a
1
+ a
2
2
+ . . . one gets the following
recursive relation among the coecients of L():
a
+1
=
+ l + 1
( + 1)( + 2l + 2)
a
. (9.14)
If the series did not stop, its asymptotic behavior would be
a
+1
a
1
+ 1
1
,
which is exactly the kind of behavior which we would have got from a solution of the
form
n
e
n
e
=
n
=0
1
!
;
in fact in this case the coecients would be
a
+1
=
1
( + 1)!
, and a
=
1
!
.
So their ratio is
a
+1
a
1
+ 1
1
,
i.e., of the type found above. But we have already said that a behavior
n
e
would
result in a radial solution R() which diverges for . Thus, the only way out is
that the series expansion of L() terminates. This must happen in order for the wave
function not to diverge.
By using the recursive equation (9.14) one has that a
+1
= 0 (and the same for all
successive coecients) when
= + l + 1 ,
From now on we shall indicate with n
+ 1)
th
power. Thus the above relation reads
= n
+ l + 1 .
187
n
is referred to as radial quantum number for it is the highest power of the series
representation of L() and thus of the radial wave function. The equation above tells
us that also
is an integer number, and we shall denote it with n:
n = n
+ l + 1 .
n is called total quantum number.
Let us now recall the denition of
, namely
=
Ze
2
_
m
2 [E[
_
1/2
,
since
= n we get
n =
Ze
2
_
m
2 [E[
_
1/2
,
and by inverting we get the following expression for the energy levels
E
n
= [E
n
[ =
mZ
2
e
4
2
2
1
n
2
, n 1 . (9.15)
The rule here is that free electrons are at zero energy and bound states have negative
energies. (We shall explain later the reason for n 1.)
Notice that we have obtained the behavior E
n
1/n
2
which is the same behavior
experimentalists had discovered. The constant
R
H
mZ
2
e
4
2
2
is called Rydbergs constant, after the name of the experimentalist who rst observed
the 1/n
2
behavior. Quantum mechanics has been capable of deriving this constant in
terms of more basic quantities.
Let us now look at the wave functions. The radial solution R(r) with energy
eigenvalue E
n
carries both a label n and a label l since l enters Eqn. (9.9). So we
shall denote the R with R
nl
(r). However, energy levels do not depend on l and
this means there is degeneracy. The total wave functions contain also the spherical
harmonics Y
lm
(, ) and so there is a dependence also on the quantum number m:
nlm
(r, , ) = R
nl
(r)Y
lm
(, ) ,
188 Chapter 9. Hydrogen atom
Besides l, the energy levels do not depend on m, i.e., we get a further degeneracy.
The polynomials L(), solutions of Eqn. (9.12), are called Laguerre polynomials.
Their general expression is:
L
2l+1
n+l
() =
nl1
k=0
(1)
k+2l+1
[(n + l)!]
2
k
(n l 1 k)!(2l + 1 + k)!k!
.
In terms of these polynomials the radial wave function R
ln
() can be written as
R
ln
() = e
1
2
l
L
2l+1
n+l
()c
n,l
,
where c
n,l
are proper normalization coecients.
Let us now turn back to the expression for the energy levels and let us dene the
quantity
a
0
2
me
2
.
The energy levels can then be written as:
E
n
=
Z
2
e
2
2a
0
n
2
.
Let us remember that n = n
+ l + 1 and that n
n=1,l=0,m=0
[
n=1,l=0,m=0
.
It is in r = a
0
. So a
0
has a specic physical meaning: it is the position at which it is
highest the probability of nding the electron and it is called Bohr radius.
9.1. Degeneracy of the energy levels 189
9.1 Degeneracy of the energy levels
As we said, the energy levels depend only on n, while the eigenfunctions
depend also
on l and m, thus there is degeneracy. Let us determine this degeneracy. Remember
that n = n
+ l + 1, where n
and
l in a way that does not aect the values of n. In particular, since n
0, l can vary
from 0 to n 1, i.e., these are all the values l can take once n is xed. Moreover, for
every value of l there are correspondingly 2l + 1 allowed values of m. Therefore, the
total number of states sharing the same value of n is
n1
l=0
(2l + 1) =
2n(n 1)
2
+ n = n
2
.
This is the total degeneracy.
The degeneracy in m is due to rotational invariance along z (or about any generic
axis) and it is thus a common feature of all central potentials U(r). On the contrary,
degeneracy on l is typical of the Coulomb potential 1/r. If you consider a valence
electron in an atom with many electrons, the potential energy is only approximately
of the Coulomb form due to the mutual repulsion forces among the electrons, and the
degeneracy in l is no longer present.
This degeneracy in l is due to an extra symmetry present in the 1/r potential. It is
a symmetry whose associated conserved charge is the Runge-Lenz vector. This vector
is the one joining the focus of an ellipse with its perihelion. Using this extra conserved
quantity Pauli gave a solution of the Hydrogen atom entirely in terms of operators in
the same period during which Schr odinger was working out its solution.
9.2 Form of the eigenfunctions
Let us now discuss the structure of the energy levels as usually given in chemistry
and how to translate it in the language of physicists.
The K shell is dened as the one with n = 1, and therefore l = 0 and m = 0. So,
190 Chapter 9. Hydrogen atom
there is only one eigenstate (it is referred to as 1s state, where 1 stays for n = 1 and
s for l = 0). Its eigenfunction is given by
u
n=1,l=0,m=0
=
1
_
Z
a
0
_
3/2
exp
_
Z
r
a
0
_
.
There can be two electrons with this radial wave function, one with spin up and the
other with spin down.
The L shell is dened as the one having n = 2, so l can assume two dierent values,
namely l = 0 and l = 1. Therefore we have two kind of states: the state 2s (n = 2
and l = 0) whose eigenfunction is
u
n=2,l=0,m=0
=
1
4
2
_
Z
a
0
_
3/2
_
2
Z
a
0
r
_
exp
_
Zr
2a
0
_
,
with two electrons with opposite spin, and the states 2p (n = 2, l = 1), with m =
1, 0, 1 and correspondingly we have three dierent eigenfunctions
u
n=2,l=1,m=1
=
1
8
_
Z
a
0
_
3/2
Z
a
0
r exp
_
Zr
2a
0
_
sin exp (i) ,
u
n=2,l=1,m=0
=
1
4
2
_
Z
a
0
_
3/2
Z
a
0
r exp
_
Zr
2a
0
_
cos ,
u
n=2,l=1,m=1
=
1
8
_
Z
a
0
_
3/2
Z
a
0
r exp
_
Zr
2a
0
_
sin exp (i) ,
Thus the L shell can carry eight electrons.
9.3 Zeeman eect
Let us now see how to remove the degeneracies:
the degeneracy in l is removed by perturbing the original Coulomb potential in
such a way that we move from U(r) = 1/r to
U(r) = 1/r +F(r) where F(r) is
any polynomial in r or 1/r.
the degeneracy in m is removed by switching on a magnetic eld. For example,
the p levels (with l = 1) split themselves into three levels with dierent m and
dierent energies. This is called Zeeman eect.
9.3. Zeeman eect 191
Let us try to explain the Zeeman eect. We know that if there is a magnetic eld
described through a vector potential A(x, y, z) the Hamiltonian becomes
H =
1
2m
0
_
p
e
c
A
_
2
.
This is the so-called minimal coupling scheme. At the quantum level we get
H =
1
2m
0
_
i
q
e
c
A( q)
_
2
=
1
2m
0
_
p
e
c
A( q)
_
2
.
Let us recall that p, being an operator, does not commute with A and thus the above
expression becomes
_
p
e
c
A( q)
_
2
= p
2
e
c
p
A
e
c
A p +
e
2
c
2
A
2
.
Moreover, since p = i
q
, we have
p AA p = [p, A] = i A .
Thus,
H =
p
2
2m
0
e
m
0
c
A p + i
e
2m
0
c
A+
e
2
2m
0
c
2
A
2
.
Let us suppose we have a constant magnetic eld along z:
B = A = [B[ z .
A possible choice of the vector potential A that produces such magnetic eld is the
following one:
A
x
=
1
2
By , A
y
=
1
2
Bx , A
z
= 0 . (9.16)
Notice that, with this choice, A = 0 and the Hamiltonian in particular becomes
H =
p
2
2m
0
+
e
m
0
c
A p +
e
2
2m
0
c
2
A
2
.
Now, by inserting Eqs. (9.16) into the right-hand side above we get
H =
p
2
2m
0
+
eB
2m
0
c
( x p
y
y p
x
) +
e
2
2m
0
c
2
A
2
.
192 Chapter 9. Hydrogen atom
The latter term is order of
1
c
2
and we shall neglect it. The Hamiltonian becomes
H =
p
2
2m
0
+
e
2m
0
c
B
L
z
.
where
L
z
is the z component of the angular momentum.
We can also add a central potential (like in the case of the Hydrogen atom):
H =
p
2
2m
0
+ U(r)
. .
Hr
+
e
2m
0
c
B
L
z
. .
H
B
. (9.17)
This Hamiltonian does commute with both
L
z
and
L
2
. Moreover, it commutes also
with
H
r
alone,
H
r
=
p
2
2m
0
+ U(r) ,
which represents the Hamiltonian of the system when the magnetic eld is turned o.
So, we are allowed to diagonalize simultaneously the four operators
H =
H
r
+
H
B
,
H
r
,
L
z
and
L
2
. We get
H
r
n,l,m
= E
n,l
n,l,m
, (9.18)
(in the particular case of the Coulomb potential E
n,l
does not depend on l). We
diagonalize in this base
H
B
:
H
B
n,l,m
=
E
n,l,m
n,l,m
eB
2m
0
c
L
z
n,l,m
=
eB
2m
0
c
m
n,l,m
, (9.19)
that is,
E
n,l,m
=
eB
2m
0
c
m = m
0
B ,
0
=
e
2m
0
c
, (9.20)
where
0
is the Bohrs magneton. By combining Eqs. (9.18)(9.20) we get
H
n,l,m
= (E
n,l
+ m
0
B)
n,l,m
,
thus the new energy levels are
E
n,l,m
= E
n,l
+ m
0
B .
9.4. Landau Levels Problem 193
E
n
m
E
n,m
+
0
gB
z
0
gB
z
E
n,m
Fig. 9.1.
and they change with m. This eect is known as Zeeman eect.
Indeed, since the electron carries spin (and hence an intrinsic magnetic moment) if
we switch on a magnetic eld the Hamiltonian will be no longer given by Eqn. (9.17),
but by
H =
H
r
+ m
0
B
z
+ 2
s
B
z
, (9.21)
where
s
=
0
s
z
and s
z
=
1
2
. The factor two in front of the last term in Eqn. (9.21)
is the gyromagnetic ratio of the electron. By diagonalizing Eqn. (9.21) one nds that
the energy levels split according to the scheme in Fig. 9.1. This eect is called
anomalous Zeeman eect.
Homework 9.1. Diagonalize Eqn. (9.21).
9.4 Landau Levels Problem
Look for the eigenfunctions and the energy levels of a charged particle in a constant
magnetic eld along z. Choose the following vector potential:
A
x
= 0 , A
y
= Bx , A
z
= 0 . (9.22)
194 Chapter 9. Hydrogen atom
Lets start by checking that Eqn. (9.22) gives the correct (i.e., constant) magnetic
eld:
B
x
=
A
z
y
A
y
z
= 0 ,
B
y
=
A
x
z
A
z
x
= 0 ,
B
z
=
A
y
x
A
x
y
= B .
The Hamiltonian is built according to the minimal coupling rule:
H =
1
2m
_
p
2
x
+
_
p
y
e
c
B x
_
2
+ p
2
z
_
,
The three operators
H, p
y
and p
z
do commute among themselves since
H does not
contain y and z. On the contrary,
H does not commute with p
x
since x appears in
the expression of
H.
So, we are allowed to diagonalize simultaneously
H, p
y
and p
z
. Let us begin with
p
y
and p
z
: we have
p
y
(x, y, z) = p
y
(x, y, z) ,
i.e.,
i
y
(x, y, z) = p
y
(x, y, z) ,
whose solution is
py
(x, y, z) = e
ipyy/
(x, z) , (9.23)
with eventually a normalization factor 1/
py
(x, y, z) = p
z
py
(x, y, z) ,
9.4. Landau Levels Problem 195
i.e.,
i
z
_
e
ipyy/
(x, z)
_
= p
z
e
ipyy/
(x, z) ,
which yields
i
z
(x, z) = p
z
(x, z) ,
whose solution is
pz
(x, z) = e
ipzz/
(x) ,
with (x) a function to be determined plus eventually a normalization constant
1/
2. Thus,
py,pz
(x, y, z) = e
ipyy/
e
ipzz/
(x) .
Now, we use this expression to diagonalize
H:
H
E,py,pz
(x, y, z) = E
E,py,pz
(x, y, z) ,
which is nothing but
1
2m
_
p
2
x
+
_
p
y
e
c
B x
_
2
+ p
2
z
_
e
ipyy/
e
ipzz/
(x) = E e
ipyy/
e
ipzz/
(x) .
This equation becomes the following one:
1
2m
_
p
2
x
+
_
p
y
e
c
B x
_
2
+ p
2
z
_
E
(x) = E
E
(x) . (9.24)
Note that in the equation above p
y
and p
z
are no longer operators, but numbers,
while we are still using p
x
as operator. We have labelled (x) with the corresponding
energy eigenvalue E since the solutions of Eqn. (9.24) depend on E. Let us rewrite
Eqn. (9.24) in the following way:
2
2m
E
(x) +
1
2m
_
p
y
e
c
Bx
_
2
E
(x) =
_
E
p
2
z
2m
_
E
(x) .
Let us introduce the following quantity
E
T
E
p
2
z
2m
,
196 Chapter 9. Hydrogen atom
where T means transverse. E
T
is the total energy in the xy-plane. Now, we change
variables and we replace x with
x
p
y
eB
c
x .
The derivatives also change according to the chain rule:
d
2
dx
2
=
eB
c
2
d
2
d(x
)
2
.
Eqn. (9.24) thus becomes
2
2m
eB
c
E
(x
) +
1
2m
(x
)
2
E
(x
) = E
T
E
(x
) .
By dividing it by [ eB/c[
2
we obtain
2
2m
E
(x
) +
1
2m
c
eB
2
(x
)
2
E
(x
) =
c
eB
2
E
T
E
(x
) . (9.25)
Let us now recall the form of the Hamiltonian of an harmonic oscillator:
H =
2
2m
2
x
2
+
1
2
m
2
x
2
.
If we compare this with the left-hand side of Eqn. (9.25) we recognize that Eqn. (9.25)
is formally the equation of an harmonic oscillator if we identify
m
2
=
1
m
c
eB
2
,
hence
=
c
eBm
, (9.26)
where we have chosen the positive solution for obvious reasons. So, Eqn. (9.25) is the
equation of a quantum harmonic oscillator with frequency given by Eqn. (9.26). The
right-hand side of Eqn. (9.25) tells us that the energy of this harmonic oscillator is
c
eB
2
E
T
, and knowing the expression for the energy levels of the harmonic oscillator
we get
c
eB
2
E
T
=
_
n +
1
2
_
,
9.5. Review Problems and Solutions 197
where is given by Eqn. (9.26). By substituting the expression for E
T
, we get
c
2
e
2
B
2
_
E
p
2
z
2m
_
=
_
n +
1
2
_
,
and therefore
E
n
=
eB
mc
_
n +
1
2
_
+
p
2
z
2m
, (9.27)
where we labelled the energy eigenvalues with the integer index n = 0, 1, . . . since n
appears on the right-hand side, however they are not discretized energy levels since
p
z
is continuous. The corresponding energy eigenfunctions are
E,py,pz
(x
, y, z) =
1
2
exp
_
i
(p
y
y + p
z
z)
_
n
(x
) , (9.28)
where
n
are the energy eigenfunctions of the one-dimensional quantum harmonic os-
cillator. Note that while the energy eigenfunctions do depend on p
y
, the energy eigen-
values do not. This means there is degeneracy, and in particular an -degeneracy
since p
y
is a continuous variable. These states are called Landau levels. They are
localized in x (in fact, in x are Gaussian functions multiplied by Hermite polynomials)
but delocalized in y and z where they are simply plane waves.
9.5 Review Problems and Solutions
Problem 9.1. Lets consider a quantum-mechanical system whose Hamiltonian
has two eigenvalues E
+
= and E
2
([+ +[) , [b =
1
2
([+ [) , (9.30)
respectively. Consider the two following situations.
198 Chapter 9. Hydrogen atom
1. Let the system evolve from t = 0 to t = 2 and calculate the probability
of nding the value a if a measurement of
O is performed on the system at
t = 2.
2. Alternatively, evolve the system until the time t = is reached, then perform
a measurement of
O without reading the outcome (such kind of measure-
ments are called non-selective ones), then let the system evolve again until
time 2 and calculate the probability to obtain a if a measurement of
O is
performed at the time t = 2.
Compare the two cases.
Solution. We work in the basis in which the Hamiltonian
H is diagonal, namely
H =
_
_
0
0
_
_
,
or in an equivalent manner
H = [+ +[ [ [ .
The time-evolution of the state [
0
from t = 0 to t = 2 is given by:
[(2) = e
H2
[
0
= e
H2
_
1
2
[+ +
_
3
4
[
_
=
1
2
[+ e
2i
+
_
3
4
e
2i
[ .
Now we look for the probability of nding a if a measurement of
O is performed on the
state [(2). According to the basic rules of quantum mechanics, such probability is given
9.5. Review Problems and Solutions 199
by
P(
O = a[2) = [a[(2)[
2
=
2
(+[ +[)
_
1
2
[+ e
2i
+
_
3
4
[ e
2i
_
2
=
1
2
_
1 +
_
3
4
cos (4)
_
. (9.31)
The notation P(
O = a[2) denotes the probability of nding the value a for the observable
O at the time t = 2. We will use a similar notation also for the other probabilities which
we shall calculate in the following.
Now, let us consider the second case. In this case, the system evolves until time t = ,
then at t = a measurement of
O is performed on the system but we do not read the
resulting outcome. Since we do not know what the result of this measurement is, we need
to calculate the probabilities of nding both results. The time-evolution of [
0
from t = 0
to t = is
[() =
1
2
[+ e
i
+
_
3
4
e
i
[ .
Now, the probability of nding a in a measurement of
O at t = is
P(
O = a[) = [a[()[
2
=
1
2
_
1 +
_
3
4
cos 2
_
.
But we can also have b as possible outcome of the measurement, which implies that after
the measurement the state [() would have collapsed into the eigenstate [b, that is
[b =
1
2
[[+ [] .
The probability of nding b is
P(
O = b[) = [b[()[
2
=
2
([+ [)
_
1
2
[+ e
i
+
_
3
4
[ e
i
_
2
=
1
2
_
1
_
3
4
cos 2
_
.
200 Chapter 9. Hydrogen atom
Now, suppose we have performed a measurement of
O and we have obtained a, then the
state would become
[a =
1
2
([+ +[) ,
and this will then evolve from t = to t = 2. At t = 2 the state will be:
[a, 2 = e
i
H/
[a =
1
2
_
e
i
[+ + e
i
[
_
.
If instead b had been obtained at t = , the time evolution from t = to t = 2 will lead
at t = 2 to the state
[b, 2 = e
i
H/
[b =
1
2
_
e
i
[+ e
i
[
_
.
Now, we have all the informations needed to calculate the probability of nding a at the
time t = 2. We shall denote this probability with
P(
O = a[2). Considering the whole
process, this composite probability is given by the sum of products of simpler probabilities
and it has the following expression:
P(
O = a[2) = P(
O = a[) [a[a, 2 [
2
+P(
O = b[) [a[b, 2 [
2
.
Since
[a[b, 2[
2
=
2
(+[ +[)
1
2
_
e
i
[+ e
i
[
_
2
= sin
2
,
and
[a[a, 2 [
2
=
2
(+[ +[)
1
2
_
e
i
[+ + e
i
[
_
2
= cos
2
,
we get
P(
O = a[2) =
1
2
+
1
2
_
3
4
cos
2
2 .
Note that this result is dierent from that of the previous case given by Eqn. (9.31). This
means that a measurement done at t = can aect the future results even if we do not read
its outcomes.
9.5. Review Problems and Solutions 201
Problem 9.2. A particle which is rotating with uniform velocity in the xy-plane
around a xed point is prepared in the state
() =
2
2
cos
2
.
1. Find the possible outcomes of a measurement of
M
z
and the corresponding
probabilities.
2. Calculate the mean value of
M
z
.
3. Find the possible outcomes of a measurement of the square of
M
z
, i.e., of
M
2
z
, and the corresponding probabilities.
4. Calculate the mean value of
M
2
z
.
5. Calculate the mean square deviation.
Solution. First of all, we diagonalize
M
z
in order to get its spectrum and so the possible
outcomes of a measurement of
M
z
. We will use = 1. We have
M
z
= i
d
d
,
and thus the eigenvalue problem for
M
z
reads
i
d
d
m
() = m
m
() ,
whose solutions are
m
() = Ae
im
,
and in order
m
to be single-valued we must have
m
(0) =
m
(2) and thus m must be
integer:
m = 0, 1, 2 . . .
A is found by imposing that
m
is properly normalized:
|
m
|
2
= A
2
_
2
0
e
im
e
im
d = A
2
_
2
0
d = 2A
2
= 1 ,
202 Chapter 9. Hydrogen atom
therefore A = 1/
0
=
1
2
,
1
=
1
2
e
i
,
1
=
1
2
e
i
, (9.32)
2
=
1
2
e
i2
,
2
=
1
2
e
i2
. . .
The initial state is
() =
2
2
cos
2
.
Since we have to compute probabilities we have to check that this state is actually normal-
ized. Indeed, it is not normalized:
||
2
=
_
2
0
()() d =
3
2
,
hence the normalized initial states is given by
() =
_
2
3
2
2
cos
2
=
_
4
3
cos
2
.
In order to obtain the possible outcomes of a measurement of
M
z
and their correspond-
ing probabilities on the state () we need to expand the state () on the basis of the
eigenfunctions of
M
z
given in Eqn. (9.32), i.e.:
() =
m=0
c
m
m
() ,
where c
m
=
m
()[(). In this case, there is an easier way to get the result without
having to calculate all these scalar products. In fact, it is sucient to rewrite the expression
for the initial state in the following way:
() =
_
2
3
2
2
cos
2
=
_
2
3
2
2
_
1 + cos 2
2
_
=
_
2
3
1
2
. .
0
+
1
6
1
2
e
2i
. .
2
+
1
6
1
2
e
2i
. .
2
.
9.5. Review Problems and Solutions 203
Since our state contains only
0
,
2
and
2
the only possible outcomes of a measurement
of
M
z
are m = 0, m = 2. The corresponding probabilities, i.e., [c
m
[
2
, can be read directly
from the above expression of () and in particular they are given by
P
0
=
_
2
3
2
=
2
3
, P
2
=
2
=
1
6
, P
2
=
2
=
1
6
.
The sum of these probabilities is
P
0
+P
2
+P
2
= 1 ,
as it should be.
The mean value of
M
z
is
M
z
=
m
mP
m
= 0
2
3
+ 2
1
6
+ (2)
1
6
= 0 .
The possible outcomes of a measurement of
M
2
z
are its eigenvalues and since for
M
z
they
were 0, 2, 2, for
M
2
z
we will have 0, 4, 4:
M
2
z
0
= 0 ,
M
2
z
2
= 4
2
,
M
2
z
2
=
M
z
M
z
2
= 2
M
z
2
= +4
2
.
Thus, the possible outcomes are only 0 and 4. The corresponding probabilities will be given
by
P
0
=
_
2
3
2
=
2
3
, P
4
= P
2
+P
2
=
2
+
2
=
1
3
.
The mean value of
M
2
z
is
M
2
z
=
m
m
2
P
m
= 0 + 4
1
3
=
4
3
.
The mean square deviation is
M
z
=
_
M
2
z
M
z
2
,
and since M
z
= 0 we get
M
z
=
_
4
3
=
2
3
.
204 Chapter 9. Hydrogen atom
Problem 9.3. The state of a quantum particle is given by
(x) = C exp
_
i
p
0
x
_
(x) ,
with (x) real. Calculate the probability that a measurement of the momentum
gives 3p
0
as outcome.
Solution. First of all, (x) needs to be normalized:
||
2
=
_
+
(x)(x) dx = 1 =
_
+
[C[
2
[[
2
dx ,
hence
[C[
2
=
__
+
[(x)[
2
dx
_
1
.
Next we expand (x) on the basis of the eigenstates of the momentum operator p, which
are given by plane waves
1
2
e
ipx/
:
(x) =
_
+
c(p)
e
ipx/
2
dp ,
and [ c(p)[
2
= P(p) is the probability density of nding p as outcome of a measurement of
p. Multiplying by e
ip
x/
/
(x)
e
ip
x/
2
dx =
1
2
_
+
c(p) e
i(pp
)x/
dp dx
=
_
+
c(p)
_
p p
_
dp
= c(p
) .
In the former manipulations, we have made use of the relation
1
2
_
+
e
i(kk
)x
dx = (k k
) .
9.5. Review Problems and Solutions 205
We have found therefore that
c(p
) =
_
+
(x)
e
i
p
2
dx
=
_
+
C(x)
e
i
p
0
x
2
e
i
p
dx
=
1
2
_
C(x) e
i
(p
0
p
)x
dx .
Now, if we want the probability of nding 3p
0
in measuring p we have simply to put p
= 3p
0
in the above relation:
c(3p
0
) =
1
2
_
+
C(x) e
i2
p
0
x
dx ,
thus (since (x) is real)
P(3p
0
) = [ c(3p
0
)[
2
=
1
2
[C[
2
_
+
(x) e
2i
p
0
x
dx
_
+
(x
) e
+2i
p
0
x
dx
=
1
2
[C[
2
_
(x)
(x
) e
i2
p
0
(x
x)
dxdx
,
where
[C[
2
=
1
_
+
[(x)[
2
dx
.
Problem 9.4 (spin-orbit). In atoms with many electrons there might be inter-
action between the intrinsic magnetic moment of the electrons and the magnetic
eld created by their motion. This eect is called spin-orbit interaction, and
in particular it might become important for the electrons in the most external
shells. The interaction Hamiltonian is
H = g
L s ,
where L is the angular momentum operator, s is the spin of the electron and g
is a coupling which can have dimensions and it is not the gyromagnetic ratio.
206 Chapter 9. Hydrogen atom
Suppose that an electron is in the p state (i.e., l = 1) nd the eigenvalues of
J =
L+ s ,
from which it follows
J
2
=
L
2
+ s
2
+
L s + s
L =
L
2
+ s
2
+ 2
L s .
We get this expression because
L and s commute with each other because they refer to
dierent quantities. From the expression above we obtain
L s =
J
2
L
2
s
2
2
,
and thus
H =
g
2
_
J
2
L
2
s
2
_
.
Let us now show that
H commutes with
J
2
,
J
z
,
L
2
, s
2
. Let us begin with
J
2
. We need to
prove that
J
2
commutes with
L
2
and s
2
. Remember that
J
2
=
L
2
+ s
2
+ 2
L s .
Clearly,
L
2
commutes with
L
2
itself, moreover it surely commutes with s
2
since they are
dierent degrees of freedom. Finally,
L
2
commutes with 2
J
2
[j, m, l, s =
2
j(j + 1) [j, m, l, s
J
z
[j, m, l, s = m[j, m, l, s
L
2
[j, m, l, s =
2
l(l + 1) [j, m, l, s
s
2
[j, m, l, s =
2
s(s + 1) [j, m, l, s
. (9.33)
We know that l = 1 and s = 1/2, so
[l s[ =
1
2
j
3
2
= l +s ,
that is, we can have j = 1/2 or j = 3/2. Now, we diagonalize
H in the basis of Eqn. (9.33)
because all these operators commute with
H:
H[j, m, l, s =
g
2
_
J
2
L
2
s
2
_
[j, m, l, s
=
2
g
2
[j(j + 1) l(l + 1) s(s + 1)] [j, m, l, s .
Inserting l = 1 and s = 1/2 we get
H[j, m, 1, 1/2 =
2
g
2
_
j(j + 1) 2
3
4
_
[j, m, 1, 1/2
=
2
g
2
_
j(j + 1)
11
4
_
[j, m, 1, 1/2 ,
therefore the energy is
E
j,1,1/2
=
g
2
2
_
j(j + 1)
11
4
_
,
that is, it depends on j, l and s in general but not on m and thus m signals a degeneracy.
If we choose j = 3/2, we get:
E
3/2,1,1/2
=
g
2
2
_
3
2
5
2
11
4
_
=
g
2
2
.
There are 2j +1 = 2
3
2
+1 = 4 eigenstates associated with this energy and in particular they
have the following form:
[3/2, 3/2, 1, 1/2
[3/2, 1/2, 1, 1/2
[3/2, 1/2, 1, 1/2
[3/2, 3/2, 1, 1/2
_
_
E
3/2,1,1/2
.
208 Chapter 9. Hydrogen atom
The other value of the energy is obtained by taking j = 1/2 and it is given by
E
1/2,1,1/2
=
g
2
2
_
1
2
3
2
11
4
_
= g
2
,
the degeneracy is 2j + 1 = 2
1
2
+ 1 = 2 and the associated eigenstates are
[1/2, 1/2, 1, 1/2
[1/2, 1/2, 1, 1/2
_
_
_
E
1/2,1,1/2
.
Problem 9.5. At the time t = 0 an harmonic oscillator is prepared in the state
[
0
=
_
2 a
+ 1
_
[0 ,
where a and a
2m
p ,
a
=
_
m
2
x i
1
2m
p .
Calculate the mean values of position and momentum operators as functions of
time.
Solution. Let us recall that the Hamiltonian has the form
H =
_
a
a +
1
2
_
,
and our initial state is
[
0
= 2 [1 +[0 ,
where [1 = a
0
[
0
= (0[ +1[ 2) (2 [1 +[0) = 5 ,
9.5. Review Problems and Solutions 209
thus the normalized state is
[
0
=
1
5
[2 [1 +[0] .
The evolution in time gives
[
t
= e
i
Ht/
1
5
[2 [1 +[0]
=
1
5
2 e
iE
1
t/
[1 +
1
5
e
iE
0
t/
[0 .
Since
E
n
=
_
n +
1
2
_
,
the state at the time t can be written as
[
t
=
2
5
e
i
3
2
t
[1 +
1
5
e
it/2
[0 .
The mean values of x and p can be computed by inverting the expressions of a and a
in
terms of x and p, i.e.,
x =
_
2m
_
a
+ a
_
,
p = i
_
m
2
_
a
a
_
,
thus
X(t) = (t)[ x[(t) =
_
2
5
e
i
3
2
t
1[ +
1
5
e
i
1
2
t
0[
_
_
_
2m
_
a + a
_
_
_
2
5
e
i
3
2
t
[1 +
1
5
e
i
1
2
t
[0
_
,
and since
a [n =
n[n 1 , a
[n =
n + 1 [n + 1 ,
the previous expression becomes
X(t) =
_
2m
_
2
5
e
i
3
2
t
1[ +
1
5
e
i
1
2
t
0[
_
_
2
5
e
i
3
2
t
[0 +
2
5
e
i
3
2
t
[2 +
1
5
e
i
1
2
t
[1
_
=
_
2m
_
2
5
e
i
2
2
t
+
2
5
e
i
2
2
t
_
=
4
5
_
2m
cos t .
210 Chapter 9. Hydrogen atom
In the same way one can nd
P(t) = (t)[ p[(t) =
2
5
2m sin t .
Now, we consider the mean value of the energy:
(t)[
H[(t) =
_
2
5
e
i
3
2
t
1[ +
1
5
e
i
1
2
t
0[
_ _
_
N +
1
2
__
_
2
5
e
i
3
2
t
[1 +
1
5
e
i
1
2
t
[0
_
=
_
2
5
e
i
3
2
t
1[ +
1
5
e
i
1
2
t
0[
_
__
1 +
1
2
_
2
5
e
i
3
2
t
[1 +
1
2
1
5
e
i
1
2
t
[0
_
=
_
2
5
3
5
+
1
5
1
5
1
2
_
=
5
_
6 +
1
2
_
=
13
10
.
Problem 9.6. Consider two particles of spin one-half. Show that the following
operator associated with the two particles
S
12
=
_
3
_
(1)
r
_ _
(2)
r
_
r
2
(1)
(2)
_
, (9.34)
can be written as
S
12
= 2
_
3
( s r)
2
r
2
( s)
2
_
,
where
s =
1
2
_
(1)
+
2
_
,
and the labels 1 and 2 refer to particles 1 or 2.
9.5. Review Problems and Solutions 211
Solution. Let us begin with the relation
( s r)
2
=
1
4
_
(1)
r +
(2)
r
2
=
1
4
_
_
(1)
r
_
2
+
_
(2)
r
_
2
+ 2
_
(1)
r
_ _
(2)
r
_
_
,
where
( r)
2
=
2
x
x
2
+
2
y
y
2
+
2
z
z
2
.
Since
2
x
=
2
y
=
2
z
= 11, we have that
( r)
2
= r
2
11 .
Using this relation we get
( s r)
2
=
1
4
_
2r
2
11 + 2
_
(1)
r
_ _
(2)
r
_
=
1
2
_
r
2
11 +
_
(1)
r
_ _
(2)
r
_
,
or equivalently
2 ( s r)
2
r
2
11 =
_
(1)
r
_ _
(2)
r
_
.
Using the L.H.S of the above expression in (9.34), we get
S
12
=
_
3
2 ( s r)
2
r
2
11
r
2
(1)
(2)
_
= 6
( s r)
2
r
2
311
(1)
(2)
.
Let us now consider the last term on the right-hand side. Remember that
s =
1
2
_
(1)
+
(2)
_
,
thus
( s)
2
=
1
4
_
2
(1)
+
2
(2)
+ 2
(1)
(2)
_
=
1
4
_
(3 + 3)11 + 2
(1)
(2)
=
3
2
11 +
1
2
(1)
(2)
,
212 Chapter 9. Hydrogen atom
where we have used the fact that
2
(1)
=
2
(1),x
+
2
(1),y
+
2
(1),z
= 11 + 11 + 11 = 311 .
Therefore,
(1)
(2)
= 2 ( s)
2
311 ,
and so for
S
12
we get
S
12
= 6
( s r)
2
r
2
311 + 311 2 ( s)
2
= 6
( s r)
2
r
2
2 ( s)
2
,
which is the expression we wanted to derive.
Problem 9.7. Three Hermitian operators
A,
B and
C satisfy the following com-
mutation rules:
[
A,
C] = 0 = [
B,
C] , [
A,
B] ,= 0 .
Prove that one of the eigenvalues of
C must necessarily be degenerate.
Solution. Let
c
be an eigenstate of
C with eigenvalue c. We apply [
A,
C] = 0 to
c
:
[
A,
C]
c
= 0 ,
thus
A
C
c
C
A
c
= c
_
A
c
_
C
_
A
c
_
= 0 ,
which means that
A
c
is another eigenstate of
C with the same eigenvalue c of
c
. If we
assume that c is not degenerate then
A
c
must be proportional to
c
:
A
c
= a
c
.
Notice that in this way
c
turns out to be eigenstate of
A with eigenvalue a. The same
argument can be applied to [
B,
C]
c
= 0, yielding
B
c
= b
c
,
9.5. Review Problems and Solutions 213
where b is the proportionality constant but also the eigenvalue.
Now, we use the fact that [
A,
B] ,= 0. We apply this commutator to the state
c
and we
get
[
A,
B]
c
=
A
B
c
B
A
c
= ab
c
ab
c
= 0 .
Since
c
form a complete set, this result holds for arbitrary states . This contradicts
the fact that [
A,
B] ,= 0. So we have to give up the hypothesis that
c
is not degenerate.
It has to be given up at least for one of the eigenstates of
C.
Problem 9.8. Consider a particle moving on a circle with radius 1 and angle
. Its angular momentum is
P
= i
d
d
. Since the commutator is [ ,
P
] = i,
do we have
P
2
, (9.35)
or not? Actually the range of is 2, so the square variation must be smaller
than 2, i.e.: 2, and if Eqn. (9.35) were true we expect that it would
be impossible to build a packet having P
_
2
= [(
A
A)
2
[ ,
where the label on
A
A
a
= a
a
,
214 Chapter 9. Hydrogen atom
it is easy to show that
A
a
= 0, in fact we have (assuming
a
is normalized) :
a
[(
A
A)
2
[
a
=
a
[(
A
a
[
A[
a
)
2
[
a
=
a
[[
A
2
+ (
a
[
A
a
)
2
2
A
a
[
A[
a
][
a
=
a
[
A
2
+a
2
2
Aa[
a
=
a
[a
2
+a
2
2a
2
[
a
= 0 .
So for example the states =
exp i
m
2
have
P
H =
2
2m
d
2
d
2
, 0 2 ,
and let us study its self-adjointness:
1
[
H
2
=
H
1
[
2
2
2m
_
2
(
1
)
2
0
. (9.36)
H is Hermitian if the surface term is zero. There is a compact way to rewrite the surface
term by introducing the following notation:
(1)
_
_
1
_
_
,
(2)
_
_
2
_
_
.
So,
2
(
1
)
2
=
(1)
_
_
0 1
1 0
_
_
(2)
,
and as a consequence the surface term in Eqn. (9.36) can be written as
2
2m
_
_
(1)
(2)
_
_
0 1
1 0
_
_
(2)
(2)
(1)
(0)
_
_
0 1
1 0
_
_
(2)
(0)
_
_
.
In order for this term to vanish, let us suppose that
1,2
(2) =
A
1,2
(0) , (9.37)
where
A is a 2 2 matrix, because
1,2
are 2- vectors. By inserting (9.37) in the surface
term we get that it is zero if:
_
_
0 1
1 0
_
_
A =
_
_
0 1
1 0
_
_
. (9.38)
9.5. Review Problems and Solutions 215
The matrices which satisfy this relation are
A = e
[i+
2
]
,
2
=
_
_
0 i
i 0
_
_
, (9.39)
where and are real numbers.
So,
H is Hermitian on a space of wave functions which are labelled by two parameters
and . They are called self-adjoint extensions of
H because it is a space larger than the
space of periodic functions.
If we take = 0, the eigenfunctions of
H are
()
n
() =
1
2
exp
_
i
_
2
+n
_
_
, (9.40)
and their corresponding eigenvalues are
E
n
=
2
2m
_
2
+n
_
2
.
()
n
() satisfy Eqn. (9.37) with
A given by Eqn. (9.39). In fact, Eqn. (9.37) and Eqn. (9.39)
with = 0 are
1
(2) = e
i
1
(0) ,
2
(2) = e
i
2
(0) .
Of course, these functions are not periodic, but only the eigenfunctions of
L
z
must be
periodic.
So, we can say that the domain of self-adjointness of
H is C
2
(0, 2) with
(2) = e
i
(0) . (9.41)
Let us now apply the operator to ():
() =
() .
This
does no longer fulll Eqn. (9.41). In fact, we have
(2) = 2(2) ,
(0) = 0 (0) = 0 ,
216 Chapter 9. Hydrogen atom
and so it is not true that
(2) = e
i
(0). This means that the operator pushes outside
the domain of self-adjointness of
H, which is our Hilbert space. A crucial ingredient in the
proof of the Heisenberg uncertainty principle is that
() must belong to the same Hilbert
space we started from. As a consequence of this, what we can say is that the uncertainty
principle does not hold anymore for the operators and
P
, () = e
ip/
/
2, even if < 2.
Note that all this goes through also in the case of periodic eigenfunctions.
Problem 9.9. Prove that the expectation value of the force on the stationary
states of the Schr odinger equation is always zero.
Solution. What we have to prove is that
E
[
F[
E
= 0 , (9.42)
with
H
E
= E
E
.
Using
F =
p and the correspondence rule
p = p, H
P.B.
p =
1
i
[ p,
H] ,
Eqn. (9.42) becomes
_
1
i
[ p,
H]
E
_
=
1
i
_
_
p
H
H p
_
E
_
=
E
i
E
[ p[
E
E
[ p[
E
= 0 .
Chapter 10
TIME-INDEPENDENT PERTURBATION THEORY
There is a small number of potentials which are exactly solvable in quantum me-
chanics. Because of this fact, people have developed techniques to solve other poten-
tials at least in an approximate way.
These techniques basically belong to two separate classes: perturbative techniques
and non-perturbative ones. Here, we will deal with the rst ones.
Let us suppose that the Hamiltonian
H of our system can be splitted into two
parts, say
H
0
and g
V , as follows:
H =
H
0
+ g
V , (10.1)
where we are assuming to know exactly the eigenvalues and the eigenstates of
H
0
,
which is not necessarily the free Hamiltonian. For the time being, let us assume that
the spectrum of
H
0
is a purely discrete one:
H
0
[E
(0)
n
= E
(0)
n
[E
(0)
n
. (10.2)
Furthermore, let us assume that
g
H
0
1 , (10.3)
where denotes the mean value with respect to arbitrary states.
Condition Eqn. (10.3) is sometimes expressed in a qualitative fashion by saying
that g (the coupling) is somehow small and thus the perturbation g
V with respect
to
H
0
is small.
218 Chapter 10. Time-independent perturbation theory
A more precise way of saying that the perturbation is small is the following one:
E
(0)
n
[g
V [E
(0)
n
E
n+1
E
n
1 ,
that is to say, the shift on the energy levels E
(0)
n
due to the perturbation g
V is smaller
than the dierence between the energy of the n-th level and the next level.
10.1 Non-degenerate case
In this section we deal with the case in which the energy levels of the unperturbed
Hamiltonian
H
0
are non degenerate. The cases in which there is degeneracy will be
covered in a later section.
Our task is to nd an approximate expression of the energy levels E
n
and their
corresponding eigenstates [E
n
of the complete Hamiltonian
H:
H[E
n
= E
n
[E
n
. (10.4)
Obviously, we want to require that, if we switch o the perturbation, then the energy
levels get into those of the unperturbed Hamiltonian
H
0
, i.e.,
lim
g0
E
n
= E
(0)
n
, lim
g0
[E
n
= [E
(0)
n
.
We assume that the exact values of E
n
and [E
n
can be written as a series in g,
since g is small. That is to say,
E
n
= E
(0)
n
+ g
1
+ g
2
2
+ . . . + g
m
m
+ . . . , (10.5)
[E
n
= [E
(0)
n
+ g [1 + g
2
[2 + . . . + g
m
[m + . . . , (10.6)
where the quantities
1
,
2
, . . . and the states [1, [2, . . . are to be determined.
We impose
E
n
[E
(0)
n
= 1 . (10.7)
10.1. Non-degenerate case 219
We will see in a while that it is always possible to meet such requirement even if the
eigenstates [E
(0)
n
are normalized, i.e.,
E
(0)
n
[E
(0)
n
= 1 . (10.8)
In fact, taking into account Eqn. (10.6), the latter requirement is fullled if
E
(0)
n
[1 = 0 , E
(0)
n
[2 = 0 , . . . , E
(0)
n
[m = 0 , . . . (10.9)
Using Eqs. (10.5)(10.6) into Eqn. (10.4) we get
_
H
0
+ g
V
_
_
[E
(0)
n
+ g [1 + g
2
[2 + . . .
_
=
_
E
(0)
n
+ g
1
+ g
2
2
+ . . .
_
[E
(0)
n
+ g [1 + g
2
[2 + . . .
_
,
By equating terms with the same power of g on the left and right hand side in the
previous expression we get the system of equations:
_
H
0
E
(0)
n
_
[E
(0)
n
= 0 ,
_
H
0
E
(0)
n
_
[1 +
_
V
1
_
[E
(0)
n
= 0 ,
_
H
0
E
(0)
n
_
[2 +
_
V
1
_
[1
2
[E
(0)
n
= 0 , . . .
(10.10)
The reason for equating terms sharing the same power of g is that g is a small but
otherwise arbitrary parameter.
Now we will see how it is possible to choose the states [m in such a way to satisfy
Eqn. (10.9) and thus Eqn. (10.7) to be fullled. The state [E
n
solution of
H[E
n
= E
n
[E
n
, (10.11)
is dened up to a constant k:
[E
n
k [E
n
,
in fact [E
n
[1 + k [E
(0)
n
, (10.12)
and insert [1
H
0
E
(0)
n
_
[1
+
_
V
1
_
[E
(0)
n
= 0 .
Inserting the expression of [1
yields
_
H
0
E
(0)
n
_
[1 + k
_
H
0
E
(0)
n
_
[E
(0)
n
. .
0
+
_
V
1
_
[E
(0)
n
= 0 ,
hence
_
H
0
E
(0)
n
_
[1 +
_
V
1
_
[E
(0)
n
= 0 .
Therefore, we can say that solutions of Eqn. (10.10) are dened up to a constant
proportional to [E
(0)
n
. The same reasoning applies to [2 and to the other states.
We can use this freedom to impose
E
(0)
n
[m
= 0 E
(0)
n
[m + k E
(0)
n
[E
(0)
n
. .
1
= 0 ,
10.1. Non-degenerate case 221
i.e., it is sucient to choose
k = E
(0)
n
[m . (10.13)
This choice of k produces a state [m
orthogonal to E
(0)
n
[. Of course, k changes with
[m so it would be better to put a label on k:
k
m
= E
(0)
n
[m .
Now, let us come back to Eqn. (10.10) and let us try to determine
1
,
2
, . . . and
the states [1 , [2 , . . .
10.1.1 First order corrections
Let us suppose we know only the state [E
(0)
n
and its energy E
(0)
n
.
Consider the second equation in Eqs. (10.10) and braket with E
(0)
n
[:
E
(0)
n
[
H
0
E
(0)
n
[1 +E
(0)
n
[
V
1
[E
(0)
n
= 0 ,
so
E
(0)
n
[E
(0)
n
E
(0)
n
[1
. .
0
+E
(0)
n
[
V [E
(0)
n
1
E
(0)
n
[E
(0)
n
. .
1
= 0 ,
and thus
1
= E
(0)
n
[
V [E
(0)
n
. (10.14)
This is the rst-order perturbative correction to the energy and it is a function only
of known things, namely,
V and [E
(0)
n
.
10.1.2 Second order corrections
Consider the third equation in Eqs. (10.10) and bracket it with respect to E
(0)
n
[:
E
(0)
n
[
H
0
E
(0)
n
[2 +E
(0)
n
[
V
1
[1
2
E
(0)
n
[E
(0)
n
. .
1
= 0 ,
so
E
(0)
n
[ E
(0)
n
E
(0)
n
. .
0
[2 +E
(0)
n
[
V [1
1
E
(0)
n
[1
. .
0
2
= 0 ,
222 Chapter 10. Time-independent perturbation theory
from which it follows that
2
= E
(0)
n
[
V [1 . (10.15)
In the same way (do this as an homework) it is possible to prove that
i
= E
(0)
n
[
V [i 1 . (10.16)
A key point in the derivation above is the fact that the states [m can be chosen
orthogonal to E
(0)
n
[.
From Eqs. (10.15)(10.16) we notice that
2
can be determined once we know [1
and, more generally, in order to calculate the correction
m
we should know the state
[m1 and so forth.
10.1.3 Correction to the states
To build the state [1 we should determine its components along a given orthogo-
nal basis. As a basis we consider the one made by the eigenstates of
H
0
, namely,
[E
(0)
1
, [E
(0)
2
, . . . , [E
(0)
i
, . . .. Thus, we should build
E
(0)
i
[1 ,
with i ,= n since, from Eqn. (10.9), we know that E
(0)
n
[1 = 0.
Leu us start from Eqn. (10.10) and project along E
(0)
i
[. From the rst equation
we get
E
(0)
i
[
H
0
E
(0)
n
[E
(0)
n
= 0 ,
thus
E
(0)
i
[E
(0)
i
E
(0)
n
[E
(0)
n
=
_
E
(0)
i
E
(0)
n
_
E
(0)
i
[E
(0)
n
. .
0
= 0 ,
thus the rst equation is trivially fullled; It happens that E
(0)
i
[E
(0)
n
= 0 because
the basis of the eigenstates of
H
0
is an orthonormal one.
Now, we consider the second equation in Eqn. (10.10) and bracket it with E
0
i
[
E
(0)
i
[
H
0
E
(0)
n
[1 +E
(0)
i
[
V
1
[E
(0)
n
= 0 ,
10.1. Non-degenerate case 223
thus
E
(0)
i
[1
_
E
(0)
i
E
(0)
n
_
= E
(0)
i
[
V
1
[E
(0)
n
,
therefore
E
(0)
i
[1 =
E
(0)
i
[(
V
1
)[E
(0)
n
E
(0)
n
E
(0)
i
.
This equation gives the components of the state [1 on the basis of [E
(0)
i
and it
requires only things which are known , namely, E
(0)
i
[, E
(0)
n
[,
V and
1
, which has
been already determined through Eqn. (10.14) without having to know anything about
[1.
Now we start from the third equation in Eqs. (10.10) and we bracket with E
(0)
i
[:
E
(0)
i
[
H
0
E
(0)
n
[2 +E
(0)
i
[
V
1
[1
2
E
(0)
i
[E
(0)
n
. .
0
= 0 ,
thus
_
E
(0)
i
E
(0)
n
_
E
(0)
i
[2 = E
(0)
i
[
V
1
[1 ,
and therefore
E
(0)
i
[2 =
E
(0)
i
[(
V
1
)[1
E
(0)
n
E
(0)
i
. (10.17)
Eqn. (10.17) gives the components of [2 in the basis of the eigenstates [E
(0)
i
in terms
of given quantities on the right-hand side, namely, once again
1
,
V , [E
(0)
i
and [1
which has been already determined in the previous section without needing to know
[2.
The procedure seems to be a recursive one. In fact, the general formula for the
state [l is
E
(0)
i
[l =
1
E
(0)
n
E
(0)
i
_
E
(0)
i
[
V
1
[l 1
2
E
(0)
i
[l 2
. . .
l1
E
(0)
i
[1
_
,
(10.18)
with i ,= n.
Homework 10.1. Prove Eqn. (10.18).
224 Chapter 10. Time-independent perturbation theory
From Eqn. (10.18) it follows that [l is completely determined from quantities which
are already known, namely,
1
, . . . ,
l1
and [1 , . . . , [l 1. Thus, it is a recursive
procedure.
At this point, it should be emphasized that the string of states [1, [2, . . . , [l,
. . . depends on which state [E
(0)
n
we are considering. Therefore, it would have been
better to write Eqn. (10.6) in the form
[E
n
= [E
(0)
n
+ g [1
n
+ g
2
[2
n
+ . . . g
m
[m
n
+ . . . ,
where the label [
n
on the states [1, [2, . . . denotes that these states are dierent if
we use dierent [E
(0)
n
.
From Eqn. (10.18) it follows that the states [l depend on which energy level E
(0)
n
and which state [E
(0)
n
we are dealing with. In fact, E
(0)
n
appear on the right-hand
side of Eqn. (10.18).
Eqn. (10.18) can be written in a more elegant way by introducing suitable projec-
tors.
10.1.4 Projection operators in perturbation theory
Let us introduce the operator
Q
0
11 [E
(0)
n
E
(0)
n
[ =
i=n
[E
(0)
i
E
(0)
i
[ . (10.19)
The last result follows from the completeness of the eigenstates of
H
0
:
j
[E
(0)
j
E
(0)
j
[ = 11 .
We can also introduce the following operator:
Q
0
Q
0
1
E
(0)
n
H
0
Q
0
. (10.20)
It is not dicult to prove, using Eqn. (10.19), that
Q
0
=
i=n
[E
(0)
i
E
(0)
i
[
E
(0)
n
E
(0)
i
. (10.21)
10.1. Non-degenerate case 225
Using
Q
0
we can write Eqn. (10.18) as
[l =
Q
0
__
V
1
_
[l 1
2
[l 2 . . .
l1
[1
_
. (10.22)
Eqn. (10.22) tells us that from [E
(0)
n
and
1
we get [1, from [1 we get
2
, then from
[1 and
2
we get [2, from [2 we get
3
, and so.
Furthermore, notice that
Q
0
[E
(0)
n
= 0 , (10.23)
with n ,= i. This follows from Eqn. (10.21) e from the fact that E
(0)
n
[E
(0)
i
=
i,n
.
Notice that also
Q
0
should carry a label n since the right-hand side of Eqn. (10.21)
contains E
(0)
n
, thus there are as many
Q
0
and strings of states [1, [2, . . . , as E
(0)
n
.
Homework 10.2. Prove Eqn. (10.21).
Homework 10.3. Prove Eqn. (10.22).
10.1.5 Summary of formulas for rst-order perturbation theory
Let us recall Eqn.(10.14):
1
= E
(0)
n
[
V [E
(0)
n
,
and also Eqn. (10.5)
E
n
= E
(0)
n
+
1
g +O(g
2
)
= E
(0)
n
[
H
0
[E
(0)
n
+ g E
(0)
n
[
V [E
(0)
n
+O(g
2
)
= E
(0)
n
[(
H
0
+ g
V )[E
(0)
n
+O(g
2
)
E
n
= E
(0)
n
[
H[E
(0)
n
+O(g
2
) .
Therefore, the energy at the rst order in g is nothing but
H (the complete Hamilto-
nian) evaluated on the unperturbed state [E
(0)
n
.
From the formula in Eqs. (10.6) we get
[E
n
= [E
(0)
n
+ g [1 +O(g
2
) ,
226 Chapter 10. Time-independent perturbation theory
where
[1 =
Q
0
_
V
1
_
[E
(0)
n
,
and since
Q
0
[E
(0)
n
= 0 we get
[1 =
Q
0
V [E
(0)
n
,
thus
[E
n
= [E
(0)
n
+ g [1 +O(g
2
)
=
_
1 + g
V
_
[E
(0)
n
+O(g
2
) .
It is also possible to check that
E
n
[E
n
= 1 +O(g
2
) , E
n
[E
(0)
n
= 1 .
Another remark: Eqn. (10.18) can be simplied as follows:
E
(0)
i
[1 =
E
(0)
i
[
V [E
(0)
n
1
E
(0)
i
[E
(0)
n
E
(0)
n
E
(0)
i
,
and from E
(0)
i
[E
(0)
n
= 0 for n ,= i we obtain
E
(0)
i
[1 =
E
(0)
i
[
V [E
(0)
n
E
(0)
n
E
(0)
i
,
i.e., the components of [1 on E
(0)
i
[ are given (modulo the denominator E
(0)
n
E
(0)
i
)
from the o-diagonal matrix elements of
V .
10.2 Perturbation theory: degenerate case
We will now treat the case in which the states are degenerate. For completeness, let
us rewrite some of the previous equations. We are interested in solving
H =
H
(0)
+ g
V , (10.24)
10.2. Perturbation theory: degenerate case 227
and E
(0)
1
, E
(0)
2
, E
(0)
3
, . . . , E
(0)
m
, . . . are the energy levels of the unperturbed Hamiltonian
H
(0)
which are exactly known.
Lets now suppose, dierently from the previous section, that these energy levels
are degenerate:
H
(0)
[E
(0)
n
, = E
(0)
n
[E
(0)
n
, , (10.25)
where is the set of quantum numbers which are necessary to label the degeneracy, or
if you wish, the quantum numbers associated with a set of operators which together
with
H
(0)
form a complete set.
Like for the non-degenerate case, we require that
lim
g0
E
n
= E
(0)
n
, (10.26)
where E
n
is the eigenvalue of
H.
Furthermore, we assume an expansion to be of the form
_
_
_
E
n
= E
(0)
n
+ g
1
+ g
2
2
+ . . . + g
m
m
+ . . .
[E
n
= [E
(0)
n
+ g [1 + g
2
[2 + . . . + g
m
[m + . . .
, (10.27)
Inserting Eqs. (10.27) into the stationary Schr odinger equation
H[E
n
= E
n
[E
n
,
we get, as before, the string of equations
_
_
_
H
(0)
E
(0)
n
_
[E
(0)
n
= 0 ,
_
H
(0)
E
(0)
n
_
[1 +
_
V
1
_
[E
(0)
n
= 0 ,
_
H
(0)
E
(0)
n
_
[2 +
_
V
1
_
[1
2
[E
(0)
n
= 0
.
.
.
(10.28)
Again, we impose
E
n
[E
(0)
n
= 1 ,
and we get from the second equation of (10.27) that
1[E
(0)
n
= 2[E
(0)
n
= . . . = m[E
(0)
n
= . . . = 0 .
228 Chapter 10. Time-independent perturbation theory
H
|E
n
E
n
|
E
n
E
n
|
E
n
E
n
H
(0)
g 0
E
(0)
n
|E
(0)
n
, 1
|E
(0)
n
, 2
.
.
.
|E
(0)
n
, l
E
n
of
H which go to E
(0)
n
for g 0 and the corresponding states
[E
n
, [
E
n
, [
E
n
go into dierent vectors of the l-dimensional space of eigenstates
associated with E
(0)
n
.
It may also happen that these states
[E
n
, [
E
n
, [
E
n
l
n
,
l
n
respectively. We require (and this is
true if the perturbation does not make
H too dierent from
H
(0)
) that the sum of the
degeracy of [E
n
, [
E
n
, [
E
n
, namely l
n
+
l
n
+
l
n
, is equal to the degeneracy of E
(0)
n
,
10.2. Perturbation theory: degenerate case 229
namely l, i.e:
l
n
+
l
n
+
l
n
= l .
Let us denote with
1
,
2
,
3
the subspaces of E
n
with degeneracies l
n
,
l
n
,
l
n
respec-
tively, so :
lim
g0
(
1
+
2
+
3
) =
(0)
n
,
where
(0)
n
is the subspace spanned by the eigenstates [E
(0)
n
, . The projector
P onto
1
+
2
+
3
for g 0 will go into the projector
P
(0)
(which was dened previously):
lim
g0
1
+
P
2
+
P
3
=
P
(0)
.
Let us now move back to the equation of the unperturbed Hamiltonian:
_
H
(0)
E
(0)
_
[E
(0)
, = 0 ,
where we have removed, for the moment, the label n of the energy level. [E
(0)
,
is a linear combination of the degenerate states [E
(0)
n
, of Eqn. (10.29). The only
requirement is that this state belongs to the subspace
(0)
, i.e.,
P
(0)
[E
(0)
, = [E
(0)
, .
Now, consider Eqs. (10.28). The second equation is
_
H
(0)
E
(0)
_
[1 +
_
V
1
_
[E
(0)
n
= 0 ,
where [E
(0)
n
is one generic state which belongs to
(0)
. By projecting this equation
onto
(0)
we get
P
(0)
_
H
(0)
E
(0)
_
[1 +
P
(0)
_
V
1
_
[E
(0)
n
= 0 ,
i.e.,
P
(0)
H
(0)
[1 E
(0)
P
(0)
[1
. .
0
+
P
(0)
V [E
(0)
n
1
P
(0)
[E
(0)
n
. .
|E
(0)
n
= 0 ,
230 Chapter 10. Time-independent perturbation theory
and hence
P
(0)
H
(0)
[1 +
P
(0)
V [E
(0)
n
1
[E
(0)
n
= 0 . (10.30)
P
(0)
is the projector onto the subspace labelled by the states [E
(0)
n
,
i
, i.e.:
P
(0)
=
i
[E
(0)
n
,
i
E
(0)
n
,
i
[ .
If now we put this expession into Eqn. (10.30) we get
i
[E
(0)
n
,
i
E
(0)
n
,
i
[
H
(0)
[1 +
P
(0)
V
P
(0)
[E
(0)
n
=
1
[E
(0)
n
,
so
E
(0)
n
i
[E
(0)
n
,
i
E
(0)
n
,
i
[1 +
P
(0)
V
P
(0)
[E
(0)
n
=
1
[E
(0)
n
,
therefore we end up in
_
P
(0)
V
P
(0)
_
[E
(0)
n
=
1
[E
(0)
n
. (10.31)
The latter is an eigenvalue equation. It is basically the diagonalization of the operator
V in the space
(0)
, since it is there that we have projected
V via the projector
P
(0)
. Moreover, the eigenstates are of the type [E
(0)
n
, i.e., linear combinations (to be
determined) of the states [E
(0)
n
,
i
.
Now, remember that
P
(0)
=
i
[E
(0)
,
i
E
(0)
,
i
[ ,
and that we can always choose [E
(0)
,
i
to be mutually orthogonal. Eqn. (10.31)
then becomes
j
[E
(0)
,
i
E
(0)
,
i
[
. .
P
(0)
V [E
(0)
,
j
E
(0)
,
j
[E
(0)
n
=
1
[E
(0)
n
.
If we bracket from the left with E
(0)
,
k
[ we get
j
E
(0)
,
k
[
V [E
(0)
,
j
E
(0)
,
j
[E
(0)
n
=
1
E
(0)
,
k
[E
(0)
n
,
10.2. Perturbation theory: degenerate case 231
which can be written as
j
V
k
,
j
(0)
j
=
1
(0)
k
, (10.32)
where
V
k
,
j
= E
(0)
,
k
[
V [E
(0)
,
j
,
and
(0)
j
= E
(0)
,
j
[E
(0)
n
.
The
1
are determined by Eqn. (10.31) or equivalently by Eqn. (10.32).
From Eqn. (10.32), we see that
1
are eigenvalues of the matrix V
k
,
j
. In the
non-degenerate case, we had only one eigenvalue:
1
= E
(0)
n
[
V [E
(0)
n
.
Now instead we have many of them, since we have to diagonalize a matrix V
k
,
j
of
dimension l l. If the various
1
are all dierent one from the other, we will have
exactly l of them and the corresponding energy levels will be
_
_
E
1
= E
0
+ g
1
(1)
E
2
= E
0
+ g
2
(1)
.
.
.
E
l
= E
0
+ g
l
(1)
,
(10.33)
where the index (1) denotes that we are dealing with the rst-order perturbative
correction, and we have l dierent corrections since Eqn. (10.32) has l dierent eigen-
values (as we have assumed). From Eqn. (10.33), if
j
(1)
are all dierent, one says that
the perturbation has completely removed the degeneracy. In fact, one goes from one
value for the energy E
(0)
to l dierent values E
1
, E
2
, . . . , E
l
.
232 Chapter 10. Time-independent perturbation theory
V (x)
x
a
a
2
0
V
0
Fig. 10.2. Innite well with a perturbation
10.3 Problems and Solutions
Problem 10.1. For a particle in an innite-well potential of width a, i.e., 0 <
x < a, nd the rst-order perturbative shift of the energy levels under the action
of a piecewise linear perturbation potential as plotted in the gure (10.2).
Solution. It is not dicult to show that the analytic form of this potential is
V (x) =
V
0
a
(a [2x a[) , (10.34)
where V
0
/a will be the small parameter. The exactly solvable system, i.e.,
H
0
in Eqn. (10.1),
is the innite potential well between 0 and a. Its energy levels and eigenfunctions are
E
(0)
n
=
2
2
(n + 1)
2
2ma
2
, (10.35a)
(0)
n
(x) =
_
2
a
sin
(n + 1) x
a
, (10.35b)
where n =0,1,2,. . . By using Eqn. (10.14), which gives the energy correction at rst order in
perturbation theory
(1)
n
=
(0)
n
[
V [
(0)
n
,
10.3. Problems and Solutions 233
and using Eqs. (10.34)(10.35) we get
(1)
n
=
_
a
0
dx
_
_
2
a
sin
(n + 1)x
a
_
V
0
a
(a [2x a[)
_
_
2
a
sin
(n + 1) x
a
_
=
2
a
V
0
a
_
a
0
sin
2
_
(n + 1) x
a
_
a [2x a[ dx
=
2V
0
a
2
_
a
0
a sin
2
_
(n + 1) x
a
_
dx
2V
0
a
2
_
a
0
[2x a[ sin
2
_
(n + 1) x
a
_
dx
=
2V
0
a
_
a
0
sin
2
(n + 1) x
a
dx
2V
0
a
2
_
a/2
0
(a 2x) sin
2
_
(n + 1) x
a
_
dx
2V
0
a
2
_
a
a/2
(2x a) sin
2
_
(n + 1) x
a
_
dx .
The rst integral above is evaluated by means of the duplication formula
2V
0
a
_
a
0
1 cos
2(n+1)x
a
2
dx ,
for the second one we use the identity
_
xsin
2
x =
_
x
_
1 cos 2x
2
_
=
_
x
2
_
x
2
cos 2x ,
where the latter is evaluated by parts. The result is
(1)
n
= V
0
_
1
2
+
1 + (1)
n
2
(n + 1)
2
_
.
FURTHER PROBLEMS WILL BE PROVIDED IN HANDWRITTEN
FORM BY THE TEACHER
Chapter 11
SCHR
O
O
=
U
O
, (11.1)
where states change as
[ [
=
U [
[
[ = [
U
.
(11.2)
In this way, we have
[
O[ = [
U
O
U[ =
. (11.3)
The usual picture of time evolution which we have used so far is an exception to
this rule. It is the one in which the states evolve in time as:
[(t) = e
i
Ht/
[(0) ,
according to the Schr odinger equation and the observables instead do not change in
time. This should be considered as a further postulate of quantum mechanics. If for
the observable x we want to compute the mean value at time t we do the following:
(t)[ x[(t) = x(t) . (11.4)
That means we take x as given and use the states at time t. This is called Schrodinger
picture.
236 Chapter 11. Schr odinger, Heisenberg and interaction pictures
As we said in the previous chapters, we will not apply the rules (11.1) to the unitary
transformation
U(t) representing time evolution. If we had applied rules (11.1)(11.2)
we would have got the analogous of Eqn. (11.3) and thus
x(t) = x(0) .
This does not make any sense, since it would tell us that the mean position does not
change in time, as if the system had not evolved in time.
We shall denote quantities in the Schr odinger picture with the label S, for example,
[
S
(t) means that this is a state in the Schr odinger picture where the states, but not
the observables, evolve in time.
It may seem unnatural to let the states evolve in time while letting the observables
unchanged. Let us ask ourselves if it is possible to use a dierent picture in which the
observables evolve in time but in such a way that we get the same mean value of the
Schr odinger picture, namely Eqn. (11.4), which have been conrmed by experiments.
Such dierent picture does in fact exist, and it is called Heisenberg picture. The
states in the Heisenberg picture will be denoted by [
H
(t). Such states are obtained
from those in the Schr odinger picture via a unitary transformation of this type:
[
H
(t) =
S
[
S
(t) , (11.5)
where
= e
i
Ht/
.
Recall that
[
S
(t) = e
i
Ht
[
S
(0) ,
by inserting this expression into Eqn. (11.5) we obtain
[
H
(t) =
S
[
S
(t) = e
i
Ht/
[
S
(t) = e
i
Ht/
e
i
Ht/
[
S
(0) = [
S
(0) . (11.6)
Therefore, the states in the Heisenberg picture at any time t are equal to the states
in the Schr odinger picture at the initial time t = 0.
11.1. Schr odinger and Heisenberg pictures 237
Notice that on the left-hand side of Eqn. (11.6) t appears but it does not on
the right-hand side, which means that in fact also on the left-hand side there is no
eective dependence on time. Thus, also if we change t on the left nothing changes:
[
H
(t
) = [
H
(t) = [
H
(0) = [
S
(0) .
As
S
O
H
=
S
O
S
S = e
i
Ht/
O
S
e
i
Ht/
. (11.7)
In this way, it is clear that, even if
O
S
does not depend explicitly on time,
O
H
is
time-dependent, and for this reason we shall denote it with
O
H
(t).
11.1.1 Time evolution of
O
H
(t)
Lets try to understand how
O
H
(t) evolves in time. By taking the time-derivative, we
obtain:
d
O
H
(t)
dt
=
d
dt
_
e
i
Ht/
O
S
e
i
Ht/
_
=
d
dt
_
e
i
Ht/
_
O
S
e
i
Ht/
+e
i
Ht/
O
S
d
dt
_
e
i
Ht/
_
=
i
H e
i
Ht/
O
S
e
i
Ht/
. .
O
H
(t)
+e
i
Ht/
O
S
_
H
_
e
i
Ht/
=
i
H
O
H
(t) + e
i
Ht/
O
S
e
i
Ht/
. .
O
H
(t)
_
H
_
=
i
H
O
H
(t)
i
O
H
(t)
H =
i
[
H,
O
H
(t)] ,
hence
d
O
H
(t)
dt
=
i
[
H,
O
H
(t)] . (11.8)
238 Chapter 11. Schr odinger, Heisenberg and interaction pictures
Notice that this is exactly the same equation we would get by using the correspondence
principle to the classical equations of motion
dO
dt
= O, H
P.B.
= H, O
P.B.
, (11.9)
The corrispondence rules say that
,
P.B.
1
i
[, ] ,
so by applying these to Eqn. (11.9) we exactly get Eqn. (11.8).
Among all operators, only one is the same in both Schr odinger and Heisenberg
pictures, and this is the Hamiltonian operator
H. In fact,
H
H
(t) = e
i
Ht/
H
S
e
i
Ht/
= e
i
Ht/
e
i
Ht/
H
S
=
H
S
.
These two dierent pictures reect the dierent approaches Schr odinger and Heisen-
berg used to develop Quantum Mechanics. Heisenberg was using all the time the
analogies with classical mechanics and so he made heavy use of the correspondence
principle, while Schr odinger had postulated his equation and was calculating every-
thing using that. Only afterwards they realized their approaches were giving the same
physical results.
Now, we shall prove that the expectation value of the observables are the same in
the Schr odinger and Heisenberg pictures, like Eqn. (11.4). The expectation value of
an observable
O
H
in the Heisenberg picture is
H
(t)[
O
H
(t)[
H
(t) =
S
(t)[ e
i
Ht/
. .
H
(t)|
e
i
Ht/
O
S
e
i
Ht/
. .
O
H
(t)
e
i
Ht/
[
S
(t)
. .
|
H
(t)
=
S
(t)[
O
S
[
S
(t) .
Expectation values are the quantities that experimentalists measure in their experi-
ments, so it is very important that they do not depend on the particular picture we
use to describe time-evolution.
11.1. Schr odinger and Heisenberg pictures 239
Since the expectation values are those measured by experimentalists, it is also
important to understand how these expectation values evolve in time. This is inves-
tigated in the following section.
11.1.2 Time evolution of the average values.
Let us consider an observable
O and let us calculate its expectation value working in
the Schr odinger picture:
O(t) = (t)[
O[(t) .
Taking the time derivative we get
d
dt
O(t) =
d
dt
_
(t)[
O[(t)
_
=
d (t)[
dt
O[(t) +(t)[
O
d [(t)
dt
. (11.10)
Remember that, using the time-dependent Schr odinger equation,
i
d [(t)
dt
=
H[(t) ,
which can also be written as:
d [(t)
dt
=
i
H[(t) , (11.11)
and whose dual is
d
dt
(t)[ = (t)[
H
i
. (11.12)
So we get from Eqn. (11.10) that
d
dt
O(t) =
i
(t)[
H
O[(t)
i
(t)[
O
H[(t)
=
i
(t)[[
H,
O][(t)
=
i
[
H,
O] (11.13)
240 Chapter 11. Schr odinger, Heisenberg and interaction pictures
This is the equation of motion of the expectation values. Notice that, in general, on
the right-hand side of Eqn. (11.13) it is not possible to factorize the mean value on
each term of the commutator, i.e.,
[
H,
O] ,= [
H,
O] . (11.14)
Homework 11.1. Explain why in general Eqn. (11.14) does hold.
11.1.3 Constants of motion
Let us suppose that
O commutes with
H:
[
O,
H] = 0 .
Then, the right-hand side of Eqn. (11.13) is zero and so
d
O(t)
dt
= 0 , (11.15)
which means that the mean value does not change in time. In this case,
O is said to
be a constant of motion.
If
O is a constant of motion, then it is possible to prove that its eigenvalues do not
depend on time. Let us start from Eqn. (11.15) and in particular we consider the
expectation value with respect to an eigenstate [
n
of
O:
O[
n
= o
n
[
n
.
The time derivative of the mean value of
O with respect to the eigenstate [
n
is
easily evaluated to be
d
O
dt
n
=
d
dt
_
n
(t)[
O[
n
(t)
_
=
d
dt
_
n
(0)[ e
i
Ht/
Oe
i
Ht/
[
n
(0)
_
,
11.1. Schr odinger and Heisenberg pictures 241
which must be zero according to Eqn. (11.15).
O commutes with
H, thus it is possible
to interchange
O with e
i
Ht/
in the previous expression to obtain
d
O(t)
dt
n
=
d
dt
_
n
(0)[ e
i
Ht/
e
i
Ht/
O[
n
(0)
_
=
d
dt
_
n
(0)[
O[
n
(0)
_
=
d
dt
[
n
(0)[o
n
[
n
(0)] =
d
dt
[o
n
n
(0)[
n
(0)]
=
do
n
dt
= 0 ,
and this proves also the result that the eigenvalues of observable which are constants
of motion do not depend on time.
11.1.4 Evolution equation of q and p and Ehrenfest equation
Let us consider the position and momentum operators in the Heisenberg picture: q
H
and p
H
. Their time evolution is obtained by use of Eqn. (11.8), i.e.,
d q
H
dt
=
i
[
H, q
H
] ,
d p
H
dt
=
i
[
H, p
H
] , (11.16)
These are the Heisenberg equations, and are the quantum analogous of the Hamilton
equations.
Let us now consider the expectation values of Eqn. (11.16):
d
dt
q
H
=
i
[
H, q
H
] ,
d
dt
p
H
=
i
[
H, p
H
] . (11.17)
These are called Ehrenfest equations.
Some remarks are needed here. Let us suppose that
H =
p
2
2m
+ V ( q) ,
then Eqs. (11.17) becomes
d
dt
q
H
=
_
p
H
m
_
,
d
dt
p
H
=
_
V
q
( q)
_
. (11.18)
242 Chapter 11. Schr odinger, Heisenberg and interaction pictures
It is important to stress that these equations do not tell us that the average values obey
classical mechanics. In fact in the second of equations. (11.18) we are not allowed to
replace
_
V
q
( q)
_
with
V ( q)
q
. If it were possible, then Eqs. (11.18) would be identical,
in form, to the classical equations of motion:
d
dt
q
H
=
p
H
m
,
d
dt
p
H
=
V ( q)
q
. (11.19)
But it is not so, since
_
V
q
_
,=
V ( q)
q
. (11.20)
Why? It is sucient to consider, for example, a potential such as
V ( q) = q
3
,
to show that
_
V
q
_
=
3 q
2
_
,
V ( q)
q
= 3 q
2
. (11.21)
These two expressions are not the same, in fact their dierence is given by
3 q
2
3 q
2
= 3
_
q
2
q
2
_
= 3 (q)
2
,
i.e., it is equal to the mean square displacement.
Homework 11.2. For an arbitrary potential V ( q), calculate the dierence
_
V
q
_
V ( q)
q
,
the result should be expressed in terms of 1) higher-order derivatives of V , 2) the
mean square displacement and 3) higher-order moments (generalization of the mean
square displacement).
11.2 Interaction picture
This is an intermediate picture, in the sense that the time-dependence of the states
is only partially moved on to the observables.
11.2. Interaction picture 243
Let us split the Hamiltonian in the Schr odinger picture into two parts: the free
part
H
(0)
S
(more generally,
H
(0)
S
is the exactly-solvable part of the Hamiltonian) and
the interacting part
V
S
(i.e., the perturbative part):
H
S
=
H
(0)
S
+
V
S
. (11.22)
Next, we introduce the following unitary operator:
0
= e
i
H
(0)
S
t/
.
By acting with this operator, we dene a new picture (the interaction picture, thus
the label I) according to the following rules:
[
I
(t) =
S
0
[
S
(t) = e
i
H
(0)
S
t/
[
S
(t)
O
I
(t) =
S
O
S
S
0
= e
i
H
(0)
S
t/
O
S
e
i
H
(0)
S
t/
.
First of all, notice that the mean values do not change in the new picture, as it should
be:
I
[
O
I
[
I
=
S
[
S
0
S
O
S
S
0
S
0
[
S
=
S
[
O
S
[
S
.
Notice that
O
I
(t) depends on time, while
O
S
does not, and also [
I
depends on time,
but in a way dierent from that of [
S
. Let us see how.
Let us derive the evolution equation for [
I
(t). Our starting point is
[
I
(t) = e
i
H
(0)
S
t/
[
S
(t) .
Taking the time-derivative of both sides yields
d [
I
(t)
dt
=
i
H
(0)
S
e
i
H
(0)
S
t/
[
S
(t)
. .
|
I
(t)
+e
i
H
(0)
S
t/
d [
S
(t)
dt
=
i
H
(0)
S
[
I
(t)
i
e
i
H
(0)
S
t/
_
H
(0)
S
+
V
S
_
[
S
(t) =
i
e
i
H
(0)
S
t/
V
S
[
S
(t)
=
i
e
i
H
(0)
S
t/
V
S
e
i
H
(0)
S
t/
. .
V
I
(t)
e
i
H
(0)
S
t/
[
S
(t)
. .
|
I
(t)
,
244 Chapter 11. Schr odinger, Heisenberg and interaction pictures
thus
i
d [
I
(t)
dt
=
V
I
[
I
(t) . (11.23)
This is the analogous of the Schr odinger equation but in the interaction picture.
From this equation we discover why this picture is called interaction picture, namely
because the role of the Hamiltonian is played here by the interaction
V
I
. This picture
is particularly suitable for doing time-dependent perturbation theory, as we will see.
11.2.1 Evolution of the operators in the interaction picture
Our starting point is
O
I
(t) = e
i
H
(0)
S
t/
O
S
e
i
H
(0)
S
t/
.
Taking the time-derivative of both sides, we get
d
O
I
(t)
dt
=
i
H
(0)
S
e
i
H
(0)
S
t/
O
S
e
i
H
(0)
S
t/
. .
O
I
(t)
e
i
H
(0)
S
t/
O
S
e
i
H
(0)
S
t/
. .
O
I
(t)
H
(0)
S
=
i
H
(0)
S
O
I
i
O
I
H
(0)
S
=
i
[
H
(0)
S
,
O
I
] (11.24)
In the interaction picture,
H
(0)
S
becomes
H
(0)
I
= e
i
H
(0)
S
t/
H
(0)
S
e
i
H
(0)
S
t/
=
H
(0)
S
e
i
H
(0)
S
t/
e
i
H
(0)
S
t/
=
H
(0)
S
.
Therefore,
H
(0)
I
is the same as
H
(0)
S
. Substituting this result into Eqn. (11.24) yields
d
O
I
dt
=
i
[
H
(0)
I
,
O
I
] . (11.25)
Therefore, in the interaction picture the evolution of the states is ruled by
V while
the evolution of the observables is ruled by
H
(0)
.
Chapter 12
TIME-DEPENDENT PERTURBATION THEORY
12.1 General derivation
This is a perturbative technique to obtain solutions of the ordinary Schr odinger equa-
tion
i
d [
S
(t)
dt
=
H[
S
(t) . (12.1)
In the previous sections, we dealt with stationary perturbation theory, i.e., our goal
was to nd approximated solutions of the stationary Schr odinger equation
H[
n
= E
n
[
n
. (12.2)
Solving Eqn. (12.1) in an approximated way means we have to nd an approximated
form for the evolution operator
U(t, t
0
) = e
i
H(tt
0
)/
. (12.3)
By taking the time-derivative of both sides with respect to t we get
i
d
U(t, t
0
)
dt
=
H
U(t, t
0
) . (12.4)
The standard initial condition which has to be fullled by the solution of this equation
is
U(t
0
, t
0
) = 11 .
Now, consider the interaction picture. Working in the Schr odinger picture, we write
down the Hamiltonian in the following way:
H =
H
(0)
S
+
V
S
,
246 Chapter 12. Time-dependent perturbation theory
where
H
(0)
S
is an Hamiltonian which we are able to solve exactly.
V
S
is an extra
interaction in Schr odinger picture which can eventually depend explicitly on time.
We have already seen that the states in the interaction picture are found from those
in the Schr odinger picture in the following way:
[
I
(t) = e
i
H
(0)
S
t/
[
S
(t) =
S
0
(t) [
S
(t) , (12.5)
while operators are found according to
O
I
= e
i
H
(0)
S
t/
O
S
e
i
H
(0)
S
t/
=
S
0
(t)
O
S
S
0
(t) , (12.6)
where
S
0
(t) = e
i
H
(0)
S
t/
.
We have already seen that the states in the interaction picture evolve in time
according to
i
d [
I
(t)
dt
=
V
I
[
I
(t) , (12.7)
where
V
I
= e
i
H
(0)
S
t/
V
S
e
i
H
(0)
S
t/
.
Operators evolve as
d
O
I
dt
=
1
i
[
O
I
,
H
(0)
I
] . (12.8)
Notice that
H
(0)
S
and
H
(0)
I
are the same. In fact,
H
(0)
I
= e
i
H
(0)
S
t/
H
(0)
S
e
i
H
(0)
S
t/
=
H
(0)
S
e
i
H
(0)
S
t/
e
i
H
(0)
S
t/
=
H
(0)
S
.
Let us now write down the evolution operator
U in the interaction picture. We have
U
I
(t) =
S
0
(t)
U(t)
S
0
(t) . (12.9)
By taking the time-derivative of both sides we can check that
U
I
(t) does not satisfy
the usual equation of motion of the operators in the interaction picture, namely
i
d
dt
U
I
(t) =
V
I
(t)
U
I
(t) . (12.10)
12.1. General derivation 247
In fact, we get
d
U
I
(t)
dt
=
d
S
0
(t)
dt
U(t)
S
0
(t) +
S
0
(t)
d
U(t)
dt
S
0
(t) +
S
0
(t)
U(t)
d
S
0
(t)
dt
=
1
i
H
(0)
U
S
0
+
S
0
1
i
H(t)
U(t)
S
0
(t) +
S
U(t)
1
i
H
(0)
S
0
=
1
i
H
(0)
U
S
0
+
1
i
0
_
H
(0)
+
V
_
U(t)
S
0
(t) +
S
U(t)
1
i
H
(0)
S
0
=
1
i
H
(0)
U
S
0
+
1
i
H
(0)
U
S
0
+
1
i
V
U
S
0
+
1
i
U(t)
H
(0)
S
0
=
1
i
V
S
0
. .
V
I
U
S
0
. .
U
I
+
1
i
U
S
0
. .
U
I
H
(0)
S
0
. .
H
(0)
I
=
1
i
V
I
U
I
+
1
i
U
I
H
(0)
I
.
We can see that in the expression above there is an extra term
U
I
H
(0)
I
/i which does
not appear in Eqn. (12.10).
Consider instead the following operator
I
(t)
S
U(t) . (12.11)
Let us derive the evolution equation for this operator in the interaction picture. By
taking the time-derivative of both sides of Eqn. (12.11) we get
d
U
I
(t)
dt
=
d
S
0
dt
U(t) +
S
0
d
U(t)
dt
=
1
i
H
0
U(t) +
S
0
1
i
H
U
=
1
i
H
0
U(t) +
1
i
0
_
H
0
+
V
_
U =
1
i
H
0
U +
1
i
H
0
U +
1
i
V
U
=
1
i
V
U =
1
i
V
S
0
S
U =
1
i
V
I
U
I
(t) .
Thus, the evolution is
d
U
I
(t)
dt
=
1
i
V
I
U
I
(t) . (12.12)
This equation tells us that it is
U
I
the operator which run the evolution of the states
[
I
(t). In fact,
d
U
I
(t)
dt
[
I
(0) =
1
i
V
I
U
I
(t) [
I
(0)
d
dt
_
I
(t) [
I
(0)
_
=
1
i
V
I
_
I
(t) [
I
(0)
_
(12.13)
248 Chapter 12. Time-dependent perturbation theory
By comparing Eqn. (12.13) to Eqn. (12.7), it is clear that we can identify
[
I
(t) =
U
I
(t) [
I
(0) . (12.14)
Eqn. (12.12) can be integrated to nd
I
(t, t
0
) = 11
i
_
t
t
0
V
I
()
I
(, t
0
) d . (12.15)
Taking the derivative of Eqn. (12.15) yields Eqn. (12.12). Eqn. (12.15) is called
integral equation, and it can be solved iteratively. We get
I
(t, t
0
) = 11 + (i)
1
_
t
t
0
V
I
()
. .
(1)
d
+ (i)
2
_
t
t
0
d
_
t
0
d
V
I
()
V
I
(
)
. .
(2)
I
(
, t
0
) . (12.16)
From this new expression of
U
I
we see that we have a term with one power of
V
I
[the
term (1) in Eqn. (12.16)] and another term with two powers of
V
I
[the term (2) in
Eqn. (12.16)]. If we think at
V
I
as a quantity proportional to some small parameter
g, i.e.,
V
I
= g
V ,
then Eqn. (12.16) contains successive powers of g until the second one. By iterat-
ing again the procedure, i.e., by inserting again
U
I
(
, t
0
) on the right-hand side of
Eqn. (12.16) and using the expression of
U
I
(t, t
0
) given by that equation (see the
left-hand side), we get
I
(t, t
0
) = 11 +
n=1
U
n
I
(t, t
0
) , (12.17)
where
U
n
I
(t, t
0
) = (i)
n
_
d
n
d
n1
d
1
_
V
I
(
n
)
V
I
(
1
)
_
,
with t >
n
>
n1
> . . . >
1
> t
0
. All the integrals are computed between t
0
and the corresponding
i
. Notice that in these integrals it appears only
V
I
, i.e., the
12.1. General derivation 249
interaction part. Furthermore, if
V
I
is proportional to g, Eqn. (12.17) is practically
an expansion on g, and thus it is a perturbative series.
Let us remember that we are interested in
U(t) and not
U
I
. To obtain the former,
we can simply use Eqn. (12.11) which can also be written as
U(t) =
S
0
U
I
.
By using Eqn. (12.17) and the relation among
U(t) and
U
I
(t) mentioned above and
the denition of
V
I
, we get
U(t, t
0
) =
U
(0)
(t, t
0
) +
n=1
U
n
(t, t
0
) , (12.18)
where
U
n
(t, t
0
) = (i)
n
_
d
n
d
n1
d
1
S
(0)
(t,
n
)
V (
n
)
S
(0)
(
n
,
n1
)
S
(0)
(
2
,
1
)
V (
1
)
S
(0)
(
1
, t
0
) . (12.19)
Here we have indicated with
S
(0)
what previously we wrote as
S
0
and also notice
that
V
I
has been turned into
V via the associated transformations. Note that the
whole expression in Eqn.(12.19) above depends only on V and
S
(0)
but calculated at
dierent instants and in dierent intervals of time.
Now, let us come back to the original unperturbed Hamiltonian
H
(0)
which can be
solved exactly. Suppose that
H
(0)
has a discrete set of eigenvalues
_
H
(0)
[a = E
(0)
a
[a
H
(0)
[b = E
(0)
b
[b
.
.
.
H
(0)
[k = E
(0)
k
[k
.
.
.
(12.20)
Now, let us dene the following two quantities:
k,l
1
_
E
(0)
k
E
(0)
l
_
,
V
k,l
k[
V (t)[l .
250 Chapter 12. Time-dependent perturbation theory
k,l
is referred to as Bohr frequency. Assume that at the time t
0
the system is in
the state [a. We want to compute the probability that at time t the system can be
found in the state [b, i.e.,
P
ab
=
b[
U(t, t
0
)[a
2
.
If the perturbation were zero, i.e.,
V = 0, then we would have
U(t, t
0
) = e
i
H
(0)
(tt
0
)/
and the probability would be
P
ab
=
_
b
e
i
H
(0)
(tt
0
)/
a
_
2
=
e
iE
(0)
a
(tt
0
)/
b[a
2
= 0 ,
since b[a = 0. If
V ,= 0, we can use Eqn. (12.18) and we obtain
b[
U(t, t
0
)[a =
n=1
b[
U
n
[a .
The rst terms in this expression are:
b[
U
(1)
[a = i
1
_
t
t
0
d
_
e
iE
(0)
b
(t)/
. .
S
(0)
(t,)
V
b,a
() e
iE
(0)
a
(t
0
)/
_
b[
U
(2)
[a = (i)
2
k
_
t
t
0
d
_
t
0
d
_
e
iE
(0)
b
(t)/
V
b,k
() e
iE
(0)
k
(
)/
V
k,a
(
) e
iE
(0)
a
(
t
0
)
_
b[
U
(3)
[a = (i)
3
l
_
t
t
0
d
_
t
0
d
t
0
d
_
e
iE
(0)
b
(t)/
V
b,k
()
e
iE
(0)
k
(
)/
V
k,l
(
) e
iE
(0)
l
(
)/
V
l,a
(
) e
iE
(0)
a
(
t
0
)/
_
.
and so on. In order to obtain the various exponential terms in the expressions above
we have made use of the states b[ and [a and the completeness relations
k
in
between.
12.1. General derivation 251
a
b
V ()
t
0
t
(a) b[
U
(1)
[a
a
k
V (
)
b
V ()
t
0
t
(b) b[
U
(2)
[a
a
k
V (
)
l
V (
)
b
V ()
t
0
t
(c) b[
U
(3)
[a
Fig. 12.1. Feynman diagrams
The matrix elements b[
U
(1)
[a, b[
U
2
[a, etc., have a graphical representation which
is given in Fig. 12.1.
The intermediate states [k, [l are called virtual (and we sum over these states)
while [a and [b are named physical states.
Notice that the perturbation seems to occur at the time but we integrate all over
the in between t
0
and t, and analogously for
.
If in the expansion above we stop at the n-th order, we get
P
ab
b[
U
(1)
[a +b[
U
(2)
[a + . . . +b[
U
(n)
[a
2
.
Thus at the rst order we have
P
(1)
ab
=
b[
U
(1)
[a
2
=
2
_
t
t
0
e
i
b,a
V
b,a
() d
2
. (12.21)
In the integral above there should be also an extra term e
iE
(0)
b
t
e
iE
(0)
a
t
. This is a
phase which does not depend on and it can be pushed outside the integral since the
integral is done with respect to . Next we have to take the modulus of this phase
but and its modulus is one.That is the reason why we neglected it.
Homework 12.1. Prove that P
(1)
ab
= P
(1)
ba
. This relation does not hold for higher
perturbative orders.
252 Chapter 12. Time-dependent perturbation theory
-
2
t
2
t
-
4
t
4
t
-
6
t
6
t
0
t
2
2
t
2(1cos t)
2
Fig. 12.2. Plot of f(t, ).
If
V does not depend on , it can be extracted from the integral and we get
P
(1)
ab
[V
b,a
[
2
f(t,
b,a
)/
2
, (12.22)
where
f(t, )
_
t
0
e
i
d
2
=
2 (1 cos t)
2
. (12.23)
The plot of f(t, ) as a function of is given in Fig. 12.2. The function has a narrow
maximum at = 0 with width 2/t, furthermore
_
+
f(t, ) d = 2t . (12.24)
Homework 12.2. Prove Eqn. (12.24).
As t , the width 2/t goes to zero and the value of the maximum at t = 0 goes
to . Thus the plot of f(t, ) becomes like a Diracs delta, and in order to reproduce
Eqn. (12.24) we must have
f(t, )
t
2t() .
By considering Eqn. (12.22), it can be seen that in order to obtain the transition
probability we have to compute the square modulus of b[
V [a weighted by f(t,
b,a
),
the latter being peaked around
a,b
= 0 and with 2/t as width. As a consequence,
12.1. General derivation 253
the transition is more likely to occur between states whose energy is near that of [a
(because the maximum of f(t,
b,a
) is at
a,b
= 0) and in an energy band given by
E
0
2/t .
So we can conclude that the pertubation leaves the energy unchanged within a bound
of 2/t.
12.1.1 Structure of f(t,
b,a
)
If
b,a
= 0, i.e., if the state [b has the same energy of [a, the perturbation preserves
the energy and f(t, = 0) t
2
, which means the longer is the time t the larger will
be the probability of having a transition.
If
b,a
,= 0, then f(t, ) as a function of t is an oscillating function between 0 and
4/
2
b,a
with period 2/
b,a
, see Eqn. (12.23). So, P
ab
will oscillate with the same
period [see, Eqn. (12.22)] around 2 [V
b,a
[
2
/(E
0,b
E
0,a
)
2
. The factor of two comes from
2
of the denominator of [1 cos t]. For small values of t with respect to 2/
b,a
,
the behavior of (1 cos t) is like t
2
.
12.1.2 Transition to nearby states
In this section we consider transitions to a set of nearby states.
Of course, so far we have considered only the case of a discrete spectrum, but noth-
ing changes if we suppose that some states are surrounded by a continuous spectrum.
Furthermore, let us suppose such states to be normalized according to
b[b
=
1
n(b)
(b b
) , (12.25)
where n(b) is a positive real-valued function of b, and b is some parameter used to
label the states. The case n(b) = 1 gives the usual normalization condition
b[b
= (b b
) .
254 Chapter 12. Time-dependent perturbation theory
The projector onto a domain B is
P
B
=
_
[b b[ n(b) db , (12.26)
Homework 12.3. Show that the term n(b) must be present in Eqn. (12.26) in order
for
P
B
to full the relation
P
B
P
B
=
P
B
,
which must hold because
P
B
is a projector.
Let E(b) be the energy of the state [b. By changing the integration variable in
Eqn. (12.26) from b to E(b) we get
P
B
=
_
B(E)
[b b[
b
(E) dE , (12.27)
where
b
(E) = n(b)
db
dE
.
b
(E) is the density of states [b at energy E, i.e., the number of states [b per unit
of energy.
The transition probability from [a to one of the states belonging to B is
T
aB
=
P
B
U(t, 0) [a
2
,
where
U(t, 0) is the time-evolution operator from time 0 to time t, and
P
B
is the
projector onto B. Thus,
T
aB
=
_
a
(t, 0)
P
P
B
U(t, 0)
a
_
= a[
(t, 0)
P
B
U(t, 0)[a , (12.28)
the latter equality follows from
B
=
P
B
,
P
B
P
B
=
P
B
.
Let us remember that
P
B
=
_
B(E)
[b b[
b
(E) dE, hence
T
aB
=
_
B(E)
P
ab
b
(E) dE ,
12.1. General derivation 255
where
P
ab
=
b[
U(t, 0)[a
2
.
Now, for P
ab
we use Eqn. (12.22) calculated at rst order in perturbation theory:
T
aB
1
2
_
B(E)
[V
b,a
[
2
b
(E)f(t,
b,a
) dE , (12.29)
where V
b,a
depends on E via the parameter b.
Example
The frequency of oscillation in of the function f is
2
t
as can be easily seen by
plotting f versus . Let us consider the transition to the levels b contained in the
energy interval (E
1
2
, E
1
+
2
), and suppose to be small enough that both V
b,a
and
b
are constant in the interval. As a consequence we can pull them outside the
integration in Eqn. (12.29).
Let us assume t to be suciently large to satisfy: t
2
or
2
t
. (12.30)
So the relation above indicates that is much bigger than the frequency of oscilla-
tion of f as a function of (if we put = 1). Let us turn to the integral :
T
aB
1
2
[V
b,a
[
2
b
_
B(E)
f(t,
b,a
) dE . (12.31)
There are two possible cases:
1. that the maximum of f at
b,a
= 0 lies outside the integration domain . In
this case, f inside the integral can be replaced by its mean value on several
oscillations in . Remember the expression of f:
f(t, ) =
2(1 cos t)
2
.
256 Chapter 12. Time-dependent perturbation theory
E
1
2
E
1
+
2
E
1
E
a
B
Fig. 12.3.
By taking the average of cos t over many complete oscillations we get zero,
thus the average of f is 2/
2
. Hence,
T
aB
2
b
(E
1
)
[V
a,b
[
2
(E
1
E
a
)
2
, (12.32)
where E
1
is the center of the interval
E
B
=
_
E
1
1
2
, E
1
+
1
2
_
.
2. that the maximum of f lies within the domain of integration, see Fig. 12.3. In
this case, the leading contribution to the integral comes from the peak of f and
only a small error results from taking the integral between and +. We
know that
_
+
f(t, ) d = 2t ,
so
T
aB
2
1
[V
b,a
(E
a
)[
2
b
(E
a
)t . (12.33)
In this case, E
a
E
b
since E
a
E
b
=
a,b
and 0 is the peak of f.
12.1.3 Transition probability per unit of time and Fermis golden rule
The transition probability per unit of time is dened by
T
aB
=
dT
aB
dt
.
12.1. General derivation 257
Consider the two previous cases. In the rst case, this quantity is zero because the
transition probability is time-independent, Eqn. (12.32). These are the transitions
which do not preserve the energy.
In the second case by taking the derivative with respect to t of Eqn. (12.33) we get
T
aB
=
2
[V
b,a
[
2
b
, (12.34)
these are the transitions which preserve the energy of the states because
ba
= 0 is
inside the integral we performed before. Eqn. (12.34) is called Fermis golden rule.
b
is the density of states whose energy is equal to that of the initial state, since
everything is concentrated at the peak of f.
If we consider (12.30), i.e.,
t
2
, (12.35)
t must be very long in order for Eqn. (12.34) to be valid. But, on the other end, we
have used the perturbation theory at the rst order and so it must be T
aB
1,
i.e.,
T
aB
=
T
(1)
aB
t 1 . (12.36)
So we conclude that indeed t cannot be too large.
Chapter 13
PATH INTEGRAL APPROACH TO QUANTUM
MECHANICS
13.1 General derivation
This approach to quantum mechanics was presented by R. F. Feynman in his Ph.D.
thesis of 1942. It was later published (1948) in Rev. Mod. Phys. with the title
Space-time approach to non-relativistic Quantum Mechanics.
Feynman wanted somehow a formulation in which space-time played a role and
not just the Hilbert space, like in the traditional approach to quantum mechanics.
His approach is very intuitive and helped in visualizing many quantum mechanical
phenomena and in developing various techniques, like the Feynman diagrams, non-
perturbative methods ( 0, N ), etc. Somehow, Dirac had got close to the
Feynman formulation of quantum mechanics in a paper in which he asked himself
which is the role of the Lagrangian in quantum mechanics.
Let us rst review the concept of action which everybody has learned in classical
mechanics. Its denition is
S[x(t)] =
_
(x
1
,t
1
)
(x
0
,t
0
)
/(x(t), x(t)) dt , (13.1)
where x(t) is any trajectory between (x
0
, t
0
) and (x
1
, t
1
), not necessarily the classical
one, and / is the Lagrangian of the system. The action S[x(t)] is what in mathemat-
ical terms is known as a functional. Remember that a functional is a map between a
space of functions x(t) and a set of numbers (the real or complex numbers, or similar
structures). From Eqn. (13.1) one sees that S[x(t)] is a functional because, once we
260 Chapter 13. Path integral approach to quantum mechanics
x
cl
(t)
x(t)
x(t)
x(t)
x(t)
(x
0
, t
0
)
(x
1
, t
1
)
Fig. 13.1.
insert the function x(t) on the right-hand side of Eqn. (13.1) (and perform the inte-
gration) we get a real number which is the value of the action on that trajectory. If
we change the trajectory, we get a dierent number. A functional is indicated with
square brackets, S[x(t)], dierently than a function whose argument is indicated with
round brackets: f(x). A function is a map between the set of numbers (real, complex,
or similar) to another set of numbers (real, complex, etc.). So, if we restrict to the
real numbers
Function: R R ,
Functional: [functions] R .
Given these denitions, let us now see what is the path integral formulation of
quantum-mechanics given by Feynman.
We know that in quantum mechanics a central element is the transition kernel to
go from (x
0
, t
0
) to (x
1
, t
1
) which is dened as
K(x
1
, t
1
[x
0
, t
0
) x
1
[ e
i
H(t
1
t
0
)
[x
0
. (13.2)
What Feynman proved is the following formula:
K(x
1
, t
1
[x
0
, t
0
) =
_
(x
1
,t
1
)
(x
0
,t
0
)
T [x(t)] e
i
S[x(t)]
, (13.3)
13.1. General derivation 261
where, on the right-hand side of Eqn. (13.3), the symbol
_
(x
1
,t
1
)
(x
0
,t
0
)
T [x(t)] stays for a
functional integration which roughly means the sum over all trajectories between
(x
0
, t
0
) and (x
1
, t
1
).
So, in Eqn. (13.3) we insert a trajectory in e
i
S[x(t)]
, calculate this quantity and
sum it to the same expression with a dierent trajectory inserted and so on for all
trajectories between (x
0
, t
0
) and (x
1
, t
1
). This is the reason why this method is called
path-integral. Note that all trajectories enter Eqn. (13.3) and not just the classical
one.
We shall give a rigorous derivation of Eqn. (13.3) but for the moment let us try
to understand a more physical reason why trajectories enter the expression of the
quantum transition kernel.
Let us recall the double slit experiment, see Fig. 13.2. In Fig. 13.2(a) both slits A
and B are open while in the other two gures only one is open. We know that the
probabilities P
AB
, P
A
, P
B
satisfy the inequality
P
AB
,= P
A
+ P
B
,
while for the probability amplitudes
AB
,
A
,
B
we have
AB
=
A
+
B
. (13.4)
Let us now put more screens with dierent openings, like in Fig. 13.3. Let us suppose
we close the slits 1B, 2A, 2C, 2D,3A, 3B and let us call the associated wave function
as
_
1A
2B
3C
_
where we have indicated on the wave-functions which slits are open. For example for
the wave-function above only the slits 1A, 2B and 3C are open as shown in Fig. 13.3.
We can associate this amplitude with the path that join the slits 1A, 2B , 3C.
Let us also remember that Eqn. (13.4) can be generalized to any set of screens with
262 Chapter 13. Path integral approach to quantum mechanics
A
B
Screen
Source
(a)
A
B
Screen
Source
(b)
A
B
Screen
Source
(c)
A
B
Screen
Source
(d)
Fig. 13.2. (a) The probability P
AB
with both slits open. (b) The probability P
A
obtained
keeping open the slit A (c) The probability P
B
obtained keeping open the slit B (d) The
probability P
A
+P
B
. We see that P
AB
,= P
A
+P
B
.
Screen
Source
1A
1B
2A
2B
2C
2D
3A
3B
3C
Fig. 13.3. More screens with dierent openings.
13.1. General derivation 263
any set of open and closed slits, so:
i
(some slits closed)
. (13.5)
In turn the amplitudes on the right-hand side of Eqn. (13.5) can be written as the
sum of all the amplitudes which have some of the remaining slits closed. The basic
blocks of all these amplitudes will be those which have only one slit open per screen
and to these we can associate a path running through the open slits. So Eqn. (13.5)
can be formally written as
=
(paths)
(path)
, (13.6)
where we have substituted the labels on the wave functions with the paths which join
the open-slits on the various screens. Using this trick Feynman had the following
idea: the open space between a source and a screen can be thought as if it were lled
by an innite set of screens each with an innite set of slits. So now if we want the
transition amplitudes from x
0
to x
1
, i.e.,
x
0
,x
1
we could write it as
x
0
,x
1
=
(paths)
(path)
, (13.7)
where the (paths) are the labels of the amplitude associated to a conguration of
screens with only one slit open through which the path passes. Of course, the paths
will be all possible paths between x
0
and x
1
because the screens have innite slits. Let
us stress that the paths in Eqn. (13.6) and (13.7) are nothing else than a symbol
to indicate a set of slits open.
This somehow gives a physical intuition of why pathseven if they are only sym-
bols or labelsenter the transition amplitudes. Of course, one cannot say that the
particles follow one path or the other because, to check that, one should do a set
of measurements along the whole path while in the transition x
0
, t
0
[x
1
, t
1
the only
measurements are made at x
0
and x
1
. What we can say from Eqn. (13.3) is that,
if we do measurements only at x
0
and x
1
then the transition amplitude is the sum
264 Chapter 13. Path integral approach to quantum mechanics
of transition amplitudes each one associated to a dierent path between x
0
and x
1
.If
instead we do a set of innte measures on all points of a path to nd out if the
particle follows that particular path, then the interference eect, between the various
amplitudes entering the path-integral, will be destroyed.
Let us now give a rigorous derivation of Eqn. (13.3). For small time intervals it was
rst done by Dirac in 1932, for arbitrary time interval it was derived by Feynman in
1942 in his Ph.D. thesis.
Before proceeding in the derivation let us review some formula regarding the expo-
nential of operators. If
A and
B are two operators, then e
A
e
B
is not equal to e
A+
B
in general, like it would happen if
A and
B were numbers. The general formula was
derived by Baker and Hausdor and is
e
A
e
B
= e
H(
A,
B)
, (13.8)
where
H(
A,
B) =
A +
B +
1
2
[
A,
B] +
1
12
_
A, [
A,
B]
_
+
1
12
_
B, [
B,
A]
_
+ . . . (13.9)
if
A and
B commutes then H(
A,
B) =
A +
B like in the case of c-numbers.
Let us now go back to physics and calculate the transition kernel which is dened
as
K(x, t[x
0
, 0) x[ e
t
H
[x
0
, (13.10)
where
H is the Hamiltonian of the system
t
H
_
=
_
exp
_
it
N
H
__
N
. (13.11)
This is so because the operators t
H/N commute among themselves in the Baker-
Hausdor formula.
Note that this is not the transition amplitude which would be xt[x
0
0 = x[ e
i
t
H
[x
0
.
13.1. General derivation 265
Let us now remember that the Hamiltonian
H is the sum of two parts
H =
A+
B,
which do not commute because
A = p
2
/2m and
B =
V ( x). So, using again the
Baker-Hausdor formula we can write
exp
_
t
N
H
_
= exp
_
t
N
_
A+
B
_
_
= exp
_
t
N
B
_
exp
_
t
N
A
_
+O
_
_
t
N
_
2
_
, (13.12)
where the terms O((t/N)
2
) are those that come from the commutators of t
A/N and
t
B/N present in Eqn. (13.9). Of course, if we take the limit N those terms are
negligible with respect to the rst. So combining Eqn. (13.12) with Eqn. (13.11) we
can write
_
x
t
H
x
0
_
= lim
N
_
x
_
e
it
N
B
e
it
N
A
_
N
x
0
_
(13.13)
Let us now write all terms of the operator
_
exp
_
it
B
N
_
exp
_
it
N
A
_
_
N
,
and so Eqn. (13.13) reads
_
x
_
e
it
N
B
e
it
N
A
_
_
e
it
N
B
e
it
N
A
_
_
e
it
N
B
e
it
N
A
_
. .
N times
x
0
_
. (13.14)
Next let us divide the interval from x
0
to x in N intervals labelled by the points
x
1
, x
2
x
N1
and corresponding to the N intervals of time. Let us now insert in
Eqn. (13.14) after the rst square bracket a completeness of the form
_
dx
N1
[x
N1
x
N1
[,
where x
N1
is the point before x in Fig. 13.4, and we continue by inserting the com-
pleteness
_
dx
N2
[x
N2
x
N2
[ after the second square bracket in Eqn. (13.14) and
so on. What we get is that Eqn. (13.13) can be written as
x[ e
t
H
[x
0
= lim
N
_
dx
N1
dx
1
_
_
x
it
N
B
e
it
N
x
N1
_
_
x
N1
it
N
B
e
it
N
x
N2
_
_
x
1
it
N
B
e
it
N
x
0
_
_
. (13.15)
266 Chapter 13. Path integral approach to quantum mechanics
x
0
t
0
x
1
t
1
x
2
t
2
.
.
.
x
N1
t
N1
x
t
Fig. 13.4. The discretized approximation of the path integral
All this can be put in the compact form
x[ e
t
H
[x
0
= lim
N
_
N
j=1
Q
j,j1
N1
j=1
dx
j
,
where Q
j,j1
is
Q
j,j1
_
x
j
it
N
B
e
it
N
x
j1
_
.
In the expression above we can apply
B =
V ( x) to the state x
j
[ and get
Q
j,j1
= e
it
N
V (x
j
)
_
x
j
it
N
x
j1
_
. (13.16)
The operator
A is instead p
2
/2m so we cannot apply it directly to the state [x
j1
.
What we will do is to insert a completeness
_
[p p[ dp = 11 before the state [x
j1
in
Eqn. (13.16). We get in this way
_
x
j
it
N
x
j1
_
=
_ _
x
j
it
N
p
2
2m
p
_
p[x
j1
dp (13.17)
Remembering that
p[x =
1
2
e
px
,
and applying p
2
on [p we get from Eqn. (13.17) the following expression
_
x
j
it
N
x
j1
_
=
_
1
2
_
e
_
t
N
p
2
2m
p(x
j
x
j1
)
_
dp
13.1. General derivation 267
which, by performing the integration in p, becomes
_
x
j
it
N
x
j1
_
=
_
m
2i
t
N
_1
2
e
im
2
N
t
(x
j
x
j1
)
2
. (13.18)
Inserting this in Eqn. (13.16) and next in Eqn. (13.15) we get
K(x, t[x
0
, 0) = lim
N
_
m
2i
_N
2
_
e
i
N
j=1
m
2
(x
j
x
j1
)
2
V (x
j
)
N1
j=1
dx
j
, (13.19)
where = t/N. Let us note two things:
1. In the exponential of Eqn. (13.19) we have the discretized form of the action.
In fact,
_
/dt
j
/(x
j
, x
j
)
=
j
_
m
2
x
2
j
V (x
j
)
_
j
_
m
(x
j
x
j1
)
2
2
2
V (x
j
)
_
=
j
_
m
(x
j
x
j1
)
2
2
V (x
j
)
_
.
2. If we look at Fig. 13.4 we see that a trajectory in the discretized form is the
broken line we have drawn between x
0
and x. If we keep x
0
and x xed and
move the intermediate points (x
1
, t
1
), (x
2
, t
2
), . . . , (x
N1
, t
N1
), we get all pos-
sible trajectories between x
0
and x. This is exactly what is achieved by the
integrations over x
j
, j = 1, 2, . . . , N 1 in Eqn. (13.19). So the sum over all
trajectories indicated by the functional integration
_
T [x(t)] in Eqn. (13.3) is
actually realized by the measure of integration contained in Eqn. (13.19), i.e.,
_
T [x(t)] = lim
N
_
m
2i
_N
2
_
N1
j=1
dx
j
.
268 Chapter 13. Path integral approach to quantum mechanics
So we can conclude that Eqn. (13.19) is nothing else than the discretized form of the
expression (13.3) and that the functional integral can be reduced to an innite set
of standard Lebesque integrals.
The path-integral not only brings to light the role of the non-classical trajectories
in quantum mechanics but also the role of the action in quantum mechanics. In fact
the action had played a role in classical mechanics but never in quantum mechanics.
This idea was the one which triggered in 1932 the work of Dirac.
The last thing we want to bring to the attention of the reader is which are the
paths which contribute most in the path-integral. We shall show that are the paths
which are continuous but nowhere-dierentiable.
The proof goes as follows. If we look at the kinetic piece in Eqn. (13.19) we see
that in the limit of 0 we must have that (x
j
x
j1
)
2
/ remains nite otherwise
the exponent would not make sense. This means
(x
j
x
j1
)
2
nite ,
i.e.,
(x
j
x
j1
)
2
, (x)
2
t . (13.20)
From Eqn. (13.20) we notice two things:
1. When t 0 we have x 0, which means that the paths are continuous;
2. the velocities x/t goes as 1/x because, using Eqn. (13.20) , we have:
x
t
x
(x)
2
=
1
x
,
so when x 0 the velocity has a singularity. As this happens at every point
it means the path is non-dierentiable.
The paths which are dierentiable, i.e.:
x
t
nite
13.2. Perturbation theory via path integrals 269
i.e., x t have a kinetic piece in the action which goes as follows:
(x)
2
t
(t)
2
t
t 0 .
So in the continuum t 0 these kinetic pieces go as e
0
. These are constants that can
be factorized out of the path-integral and we can get rid of them in the normalization.
So these paths do not give any phase which may interphere with the other paths and
create typical quantum mechanical eects.
The non-dierentiable paths are a typical indicator of quantum mechanical eects.
In classical mechanics the paths are smooth instead. More details and examples on
the path-integrals can be found in the many books on the subject.
13.2 Time dependent perturbation theory via path integrals and proto-
Feynmans diagrams
In this section we will derive the time-dependent perturbative expansion for the time-
evolution kernel via path-integrals, without using the interaction picture.
We have seen that the transition Kernel between x = x
a
and x = x
b
from time t
a
to time t
b
has the following path integral expression
:
K
V
(b, a) =
_
b
a
T [x(t)] exp
_
i
__
t
b
ta
_
m x
2
2
V (x, t)
_
dt
__
. (13.21)
If we suppose that the potential is small or better that
_
t
b
ta
V (x, t) dt is small with
respect to , then we can expand the integrand as follows:
exp
_
_
t
b
ta
V (x, t) dt
_
= 1
i
_
t
b
ta
V (x, t) dt
+
1
2!
_
i
_
2
__
t
b
ta
V (x, t) dt
_
2
+ . . . (13.22)
If we want to get the transition amplitude and not the transition kernel, we should just put a
minus sign in front of the Lagrangian.
270 Chapter 13. Path integral approach to quantum mechanics
Of course,
_
t
b
ta
V (x, t) dt is a functional, so the statement that
_
t
b
ta
V (x, t) dt should be
small with respect to needs to be better claried but we will leave this clarication
for more advanced courses.
Using the above expansion, we can rewrite K
V
(b, a) in the following way:
K
V
(b, a) = K
(0)
(b, a) + K
(1)
(b, a) + K
(2)
(b, a) . . . , (13.23)
where
K
(0)
(b, a) =
_
b
a
T [x(t)]
_
exp
_
i
_
t
b
ta
m x
2
2
dt
__
,
K
(1)
(b, a) =
i
_
T [x(t)]
___
t
b
ta
V (x(s), s) ds
_
exp
_
i
_
t
b
ta
m x
2
2
dt
__
,
K
(2)
(b, a) =
1
2
2
_
T [x(t)]
___
t
b
ta
V (x(s), s) ds
_ __
t
b
ta
V (x(s
), s
) ds
_
exp
_
i
_
t
b
ta
m x
2
2
dt
_ _
Let us now proceed to evaluate the various terms. K
(0)
(b, a) is the free particle
transition kernel. K
(1)
(b, a) can instead be written as
K
(1)
(b, a) =
i
_
t
b
ta
F(s) ds , (13.24)
where
F(s) =
_
b
a
T [x(t)] V (x(s), s) exp
_
i
_
t
b
ta
m x
2
2
dt
_
. (13.25)
Basically, F(s) is the path-integral of the free particle but with inserted at t = s
the potential. So the time-evolution before t = s is the one of the free particle, at
t = s it gets perturbed by V (x(s), s), afterwards it is again the evolution of the free
particle. The picture that we can associate to Eqn. (13.25) is the one in Fig. 13.5.
There, we have supposed that the point reached at time s is x
c
. In Eqn. (13.25) we
are eectively integrating over x
c
and as a consequence F(s) can be written in the
following manner:
F(t
c
) =
_
+
K
(0)
(b, c)V (x
c
, t
c
)K
(0)
(c, a) dx
c
.
13.2. Perturbation theory via path integrals 271
t
x
x
a
x
b
x
c
t
a
t
b
s
Fig. 13.5. Schematic representation of F(s).
Using this, K
(1)
(b, a) can be represented as
K
(1)
(b, a) =
i
_
t
b
ta
_
+
K
(0)
(b, c)V (x
c
, t
c
)K
(0)
(c, a) dx
c
dt
c
. (13.26)
The pictures that can be drawn associated to the perturbation series of Eqn. (13.23)
are, in analogy to that in Fig. 13.5, the ones in Fig. 13.6. These are proto-Feynman
diagrams in a potential theory.
We shall now derive an integral equation for K
V
(b, a). Using Eqn. (13.26) and
the analog for higher orders we get easily the following expression (where
_
d
c
=
_
dx
c
dt
c
):
K
V
(b, a) = K
(0)
(b, a)
i
_
K
(0)
(b, c)V (x
c
, t
c
)K
(0)
(c, a) d
c
+
_
_
2
_ _
K
(0)
(b, c)V (x
c
, t
c
)K
(0)
(c, d)V (x
d
, t
d
)K
(0)
(d, a) d
c
d
d
+ , (13.27)
which can also be written as
K
V
(b, a) = K
(0)
(b, a)
i
_
K
(0)
(b, c)V (x
c
, t
c
)
_
K
(0)
(c, a)
i
_
K
(0)
(c, d)V (x
d
, t
d
)K
(0)
(d, a) d
d
+ . . .
_
d
c
+ . . . (13.28)
272 Chapter 13. Path integral approach to quantum mechanics
V
a
b
(a)
V
a
b
c
(b)
V
a
b
c
d
(c)
V
a
b
c
d
(d)
Fig. 13.6. Schematic representation of series expansion Eqn. (13.23).
If we look at Eqn. (13.28) and in particular at the quantity inside the square brackets,
we notice that Eqn. (13.28) can be re-written as
K
V
(b, a) = K
(0)
(b, a)
i
_
K
(0)
(b, c)V (x
c
, t
c
)K
V
(c, a) d
c
. (13.29)
This is an integral equation for K
V
. It is completely equivalent to the dierential
Schr odinger equation: is basically the integral version of it
Chapter 14
SEMICLASSICAL (WKB) QUANTUM MECHANICS
This chapter is mostly based on the Bachelor Thesis of Taracchini Andrea, Univer-
sity of Trieste.
14.1 Introduction
The wave function (x, t) of a physical system is a complex-valued quantity and so
it can be written in general as
(x, t) = A(x, t) e
i
S(x,t)
, (14.1)
where A(x, t) and S(x, t) are real-valued functions. The presence of the factor at
exponent is just a convention and in this manner S acquires the dimension of an
action.
Let us now calculate various expressions entering the Schroedinger equation:
2
(x, t) =
2
_
A(x, t) e
i
S(x,t)
_
=
_
2
A(x, t) + 2
i
A(x, t) S(x, t) +
i
A(x, t)
2
S(x, t)
2
A(x, t) [S(x, t)]
2
_
e
i
S(x,t)
,
(x, t)
t
=
t
_
A(x, t) e
i
S(x,t)
_
=
_
A(x, t)
t
+
i
A(x, t)
S(x, t)
t
_
e
i
S(x,t)
.
Inserting Eqn. (14.1) into the time-dependent Schr odinger equation
i
(x, t)
t
=
_
2
2m
2
+ V (x, t)
_
(x, t) , (14.2)
274 Chapter 14. Semiclassical (WKB) quantum mechanics
we get
i
A(x, t)
t
A(x, t)
S(x, t)
t
=
2
2m
2
A(x, t)
i
m
A(x, t) S(x, t)
i
2m
A(x, t)
2
S(x, t) +
1
2m
A(x, t) [S(x, t)]
2
+ V (x, t)A(x, t). (14.3)
Equating the real and imaginary parts we get the two equations
S(x, t)
t
+
1
2m
[S(x, t)]
2
+ V (x, t) =
2
2m
2
A(x, t)
A(x, t)
, (14.4)
m
A(x, t)
t
+ A(x, t) S(x, t) +
1
2
A(x, t)
2
S(x, t) = 0 . (14.5)
Eqn. (14.5) is nothing but the continuity equation for the probability associated to
the Schr odinger equation:
t
+ j = 0 . (14.6)
In fact, dening the probability and the corresponding current j as
= [(x, t)[
2
= [A(x, t)]
2
, (14.7)
j =
_
(x, t)
im
(x, t)
_
= [A(x, t)]
2
S(x, t)
m
, (14.8)
Eqn. (14.5) becomes
t
+ j = 0 .
Let us now turn to Eqn. (14.4). This equation is similar to an equation of classical
mechanics known as Hamilton-Jacobi equation, which is
S
0
(x, t)
t
+
[S
0
(x, t)]
2
2m
+ V (x, t) = 0 . (14.9)
Notice that Eqn. (14.4) for S(x, t) reduces to the Hamilton-Jacobi equation in the
limit 0.
People sometimes say that the Schr odinger equation is the quantum correction
to the Hamilton-Jacobi, but this statement is not correct because the Schr odinger
equation involves both A(x, t) and S(x, t) while the Hamilton-Jacobi involves only S.
We have chosen to put the sub index 0 in the solution S
0
of the Hamilton-Jacobi
equation, since it is the solution at order zero in
2
of Eqn. (14.4).
14.2. Hamilton-Jacobi formalism 275
14.2 Review of the Hamilton-Jacobi formalism of classical mechanics
Let us briey review the Hamilton-Jacobi formalism of classical mechanics. The
Hamiltonian description of a system with f degrees of freedom involves f congura-
tional variables q
1
, . . . , q
f
and their conjugated momenta p
1
, . . . , p
f
. The dynamics is
governed by the Hamilton equations of motions, which can be written as
q
k
= q
k
, H
P.B.
, (14.10)
p
k
= p
k
, H
P.B.
, (14.11)
k = 1, . . . , f, where H is the Hamiltonian.
An important feature of the Hamiltonian formalism is that it allows transformations
on the phase space of the system
Q
k
= Q
k
(q
1
, . . . , q
f
, p
1
, . . . , p
f
, t) , (14.12)
P
k
= P
k
(q
1
, . . . , q
f
, p
1
, . . . , p
f
, t) , (14.13)
which mix the qs and ps yet preserve the Hamiltonian form of the equations of
motion, i.e., the equations of motion have the same form also with respect to the
new set of transformed variables. Such transformations are referred to as canonical
transformations.
For a transformation to be a canonical the fundamental condition (called symplectic
condition) that must be satised is
MJM
T
= J , (14.14)
k=1
_
F
q
k
G
p
k
F
p
k
G
q
k
_
,
for arbitrary dierentiable functions F and G dened on the phase space of the system.
276 Chapter 14. Semiclassical (WKB) quantum mechanics
where M is the Jacobian of the transformation, i.e.:
M =
(Q
1
, . . . , Q
f
, P
1
, . . . , P
f
)
(q
1
, . . . , q
f
, p
1
, . . . , p
f
)
=
_
_
_
_
_
_
_
_
_
_
_
_
_
_
Q
1
q
1
Q
f
q
1
P
1
q
1
P
f
q
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Q
1
q
f
Q
f
q
f
P
1
q
f
P
f
q
f
Q
1
p
1
Q
f
p
1
P
1
p
1
P
f
p
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Q
1
p
f
Q
f
p
f
P
1
p
f
P
f
p
f
_
_
_
_
_
_
_
_
_
_
_
_
_
_
,
and J is called Poisson matrix:
J =
_
_
0 11
11 0
_
_
,
which is written in an f f blocks form.
The well-known variational principle tells that the solutions of Eqs. (14.10)(14.11)
satisfy the variational equation
_
t
2
t
1
/(q
1
, . . . , q
f
, q
1
, . . . , q
f
, t) dt
=
_
t
2
t
1
_
f
k=1
q
k
p
k
H(q
1
, . . . , q
f
, p
1
, . . . , p
f
, t)
_
dt = 0 .
(14.15)
An analogous expression is valid for the new phase-space variables, i.e.,
_
t
2
t
1
_
f
k=1
Q
k
P
k
K(Q
1
, . . . , Q
f
, P
1
, . . . , P
f
, t)
_
dt = 0 , (14.16)
where we have denoted with K the Hamiltonian with respect to the new set of vari-
ables. In order for Eqs. (14.15) and (14.16) to describe the same system, the two
integrands must dier at most by the total derivative of a function F(q, P, t). F is
called the generating function of the canonical transformation. Because of this total
14.2. Hamilton-Jacobi formalism 277
derivative we get:
F(q, P, t)
q
k
= p
k
, (14.17)
F(q, P, t)
P
k
= Q
k
, (14.18)
F(q, P, t)
t
+ H(q, p, t) = K(Q, P, t) . (14.19)
By properly choosing the canonical transformation, it is possible to obtain a trans-
formation such that K(Q
1
, . . . , Q
f
, P
1
, . . . , P
f
) = 0, i.e., the Hamiltonian in the new
coordinates is simply zero. In this case, the new coordinates evolve according to
Q
k
= Q
k
, K
P.B.
= 0 ,
P
k
= P
k
, K
P.B.
= 0 ,
i.e., the new variables are constants of motion.
According to Eqn. (14.19), a canonical transformation of this type is generated by
an F(q, P, t) that satises
F(q, P, t)
t
+ H(q, p, t) = 0 . (14.20)
If in this equation we rewrite everything as a function of only (q, P, t) , by using
Eqn. (14.17), we get
F(q, P, t)
t
+ H
_
q,
F(q, P, t)
q
, t
_
= 0 . (14.21)
This is exactly the Hamilton-Jacobi equation: it is a non-linear partial dierential
equation for the generating function F(q, P, t). Using Hamilton variational methods
(with variations also of the time at the extremes) it is possible to prove that the
Hamilton-Jacobi equation is satised by the classical action function
F(q, P, t) = S
cl
(q, P, t) =
_
t
t
1
/(q
cl
(t
), q
cl
(t
), t
) dt
, (14.22)
Here and in the following with the symbol /q (i.e., partial derivative with respect to a vector)
we mean the gradient with respect to that vector, i.e., /q =
q
.
278 Chapter 14. Semiclassical (WKB) quantum mechanics
where q
cl
(t
2
[E V (x)] (x) = 0 . (14.26)
Remember that if V (x) = const. = V and E > V , Eqn. (14.26) has solutions of the
form
(x) = Ae
ikx
, k =
_
2m
2
(E V ) .
They are oscillating wave functions with constant amplitude A and constant wave-
length = 2/k.
14.3. WKB method 279
Now, let us assume that V (x) varies slowly as a function of the position x with
respect to . Then, it seems resonable to assume that the wave function is still an
oscillating one but with slowly-varying position-dependent amplitude and wavelength.
Analogously, if V (x) = const. = V but with E < V , then the solutions of
Eqn. (14.26) take the form
(x) = Ae
x
, =
_
2m
2
(V E) ,
i.e., a pure exponential. Note that the above is dierent than the k of the previous
formula. Again, if the potential V (x) varies slowly as a function of x with respect
to 1/, then we may assume that the wave function solution is still of exponential
type but with A and k slowly dependent on x.
It should already be clear that our considerations will give a reasonable approxi-
mation only within certain regions of the domain of (x). (We will be more precise
later on.) In fact, at the points where E = V both and 1/ diverge to + and it
does no longer make sense to assume V (x) to be a slowly varying function of x on
the scale of in the neighbor of such points. Such points are called turning points,
since classically those are the points at which the particle changes the direction of its
motion. The mathematical treatment of the solution in a neighbor of these points
is rather cumbersome and we will invite the interested reader to study it in more
advanced books.
14.3.2 WKB solutions and their validity
Let us come back to the stationary Schr odinger equation and write (x) as
(x) = A(x) e
i
S(x)
, (14.27)
where A and S are time-independent real-valued functions. Plugging Eqn. (14.27)
into the stationary Schr odinger equation and equating real and imaginary parts we
280 Chapter 14. Semiclassical (WKB) quantum mechanics
get two coupled equations for A and S which are
(S
(x))
2
2m[E V (x)] =
2
A
(x)
A(x)
, (14.28)
2A
(x)S
(x) + A(x)S
(x) = 0 , (14.29)
where the prime denotes the derivative with respect to x, i.e., S
)
1
2
. (14.30)
Substituting back into Eqn. (14.28) we obtain
(S
)
2
= 2m(E V ) +
2
_
3
4
_
S
_
2
1
2
S
_
. (14.31)
This equation is rigorously equivalent to the Schr odinger equation we started from,
Eqn. (14.26).
So far, no approximation has been made. The WKB approximation is basically an
expansion of S in powers of
2
, i.e.,
S = S
0
+
2
S
1
+ . . . . (14.32)
Of course, we shall keep only the lowest order terms in
2
. The reason why the
expansion Eqn. (14.32) has been done in
2
and not simply in is that in Eqn. (14.31)
2
(and not itself) makes its appearance.
At this point we can rewrite Eqn. (14.27) as
(x) = Ae
i
S
= Ae
i
(S
0
+
2
S
1
+...)
. (14.33)
Inserting the Eqn. (14.32) into (14.31) and looking only at the zero-th order in
2
we
get
(S
0
)
2
= 2m(E V ) . (14.34)
14.3. WKB method 281
Using the following denitions
(x) =
1
k(x)
=
_
2m[E V (x)]
, for E > V (x) , (14.35)
(x) =
1
(x)
=
_
2m[V (x) E]
, for E < V (x) , (14.36)
Eqn. (14.34) becomes
S
0
(x) =
(x)
= k(x) , for E > V (x) , (14.37)
S
0
(x) = i
(x)
= i(x) , for E < V (x) . (14.38)
So the WKB solutions have the form
(x) Ae
i
S
0
, (14.39)
which using Eqn. (14.30) becomes
(x)
WKB
(x) =
C
_
k(x)
e
i
_
k(x) dx
, for E > V (x) , (14.40)
(x)
WKB
(x) =
C
_
(x)
e
_
(x) dx
, for E < V (x) . (14.41)
The general approximate solutions will be a linear combinations of the two above
(with the plus and the minus sign). It is interesting to observe that, in the classically
allowed region (i.e., for E > V ), we have that
= [(x)[
2
[C[
2
k(x)
1
p(x)
, (14.42)
This equation shows that the probability to nd the particle at the point x is inversely
proportional to its linear momentum p(x), i.e., to its velocity, exactly as in Classical
Mechanics.
When one uses approximation methods, it is always necessary to specify mathe-
matically the limits of applicability of the method. To state that the WKB works well
for slowly varying potentials is not a mathematical precise statement. If we want to
282 Chapter 14. Semiclassical (WKB) quantum mechanics
be more precise, let us note from Eqn. (14.33) that the corrections of order
2
brings
into the WKB solutions only a further phase factor e
iS
1
. This factor can be neglected
only if S
1
1. At this point it is enough to insert the expansion Eqn. (14.32) into
Eqn. (14.31) and equate the terms of order
2
. For example, for E > V we nd after
some simple calculations
S
1
(x) =
_
1
4
(x)
1
8
[
(x)]
2
(x)
_
, (14.43)
and integrating both sides of this expression we get
S
1
=
_
1
4
(x)
1
8
_
[
(x)]
2
(x)
dx
_
. (14.44)
The criterion for the WKB approximation, i.e., S
1
1, becomes then:
WKB
(x) =
A
_
(x)
e
_
a
x
(x
) dx
+
B
_
(x)
e
+
_
a
x
(x
) dx
WKB
(x) =
C
_
k(x)
e
i
_
a
x
k(x
) dx
+
D
_
k(x)
e
+i
_
a
a
k(x
) dx
_
a
x
(x
) dx
+
B
_
(x)
e
+
_
a
x
(x
) dx
2A
_
k(x)
cos
__
x
a
k(x
) dx
4
_
B
_
k(x)
sin
__
x
a
k(x
) dx
4
_
. (14.48)
Clearly, the oscillating solutions are valid for x a while the exponential solutions
are valid for x a. Analogous formulae hold when V (x) E for x a.
14.4 decay
The WKB belongs to the class of so called non-perturbative approximation methods.
As we have already seen, the basic idea is to consider small, while the perturbative
coupling of the potential can be any value, even not small. Other non-perturbative
methods are the variational method and the adiabatic method, which we shall not
discuss here. Since non-perturbative methods do not require the coupling constant
to be small, they are the obvious choice when the interactions are strong, as it
happens for instance in nuclear physics.
In this section, we shall see an application of the WKB method in that sector of
physics. In particular we shall study the decay of the nuclei. A nucleus of radius
284 Chapter 14. Semiclassical (WKB) quantum mechanics
V (x)
x
a b
V
0
E
Region I Region II Region III
Fig. 14.1. Model of potential barrier in the decay of the nuclei.
a and charge Z can be roughly represented by a potential well of depth V
0
. (See
Fig. 14.1.) For x > a, the potential is of Coulomb type and goes to zero for x .
Let us suppose that inside the nucleus there is a small bunch of matter, i.e., particles
with charge z, and energy E > 0. Let us indicate with x = b the point at which the
Coulomb potential is equal to the energy, i.e.,
zZe
2
4
0
b
= E .
Our goal is to evaluate the probability amplitude that the particle escapes the nu-
cleus. We have to build a solution of the Schr odinger equation which in the region III
behaves as a wave (transmitted wave). In that region, using the WKB approximation
the solution has the form
III
(x)
C
_
k(x)
exp
_
i
_
k(x) dx
_
. (14.49)
To simplify the calculations, let us choose a wave which propagates in the direction
of positive x. Let us insert a phase of /4 (s are dened up to an overall phase)
14.4. decay 285
and C is given by choosing the integration range, so we obtain:
III
(x) =
1
_
k(x)
exp
_
i
_
x
b
k(x
) dx
+ i
4
_
=
1
_
k(x)
_
cos
__
x
b
k(x
) dx
+
4
_
+ i sin
__
x
b
k(x
) dx
+
4
__
. (14.50)
Now, we make use of the connection formulae in order to determine in the region II.
We extend only the sin function, since the cos function would give a exponentially
increasing function which would be not normalizable. We have
II
(x) =
i
_
(x)
exp
__
b
x
(x
) dx
_
=
i
_
(x)
exp
__
b
a
(x
) dx
_
exp
_
_
x
a
(x
) dx
_
(14.51)
Dening
=
_
b
a
(x) dx =
_
b
a
_
2m
2
[V (x) E] dx , (14.52)
Eqn. (14.51) becomes
II
(x) =
i
_
(x)
e
exp
_
_
x
a
(x
) dx
_
. (14.53)
In region I, the solution of the Schr odinger equation is always an oscillating function.
To make the calculations easier, and indicating as usual k
0
=
_
2m(E V
0
)/, we
can write
I
(x) = Asin [k
0
(x a) + ] =
A
2i
_
e
i[k
0
(xa)+]
e
i[k
0
(xa)+]
_
, (14.54)
where the constants A and are determined by imposing the continuity of the wave
function and of its rst-order derivative in x = a. In particular, we nd that
_
_
_
k
0
cot = (a)
Asin =
i
(a)
e
.
As we want to study the tunnel eect through the potential barrier, we need to
calculate the transmission coecient T, given by the ratio between the transmitted
286 Chapter 14. Semiclassical (WKB) quantum mechanics
probability current J
tr
III
which reaches region III and the incident current J
inc
I
coming
from region I, i.e.,
T =
J
tr
III
J
inc
I
.
In order to calculate J
inc
I
let is note that
I
is the sum of an incident and a reected
wave. Let us remember that in one dimension the probability current is
J =
_
(x)
p
m
(x)
_
=
_
(x)
im
d
dx
(x)
_
.
It is easy to prove that
J
inc
I
=
1
4
[A[
2
k
0
m
and J
tr
III
=
m
.
Thus
T = 4
_
(V
a
E)(E V
0
)
V
a
V
0
e
2
,
which means that
T exp
_
_
b
a
_
2m[V (x) E] dx
_
.
This one is the famous Gamow formula.
The reader may ask why we have used the WKB approximation. The reason is that
we could not solve exactly the Schr odinger equation associated with the potential
drawn in the previous gure. At the same time there was no small perturbative
parameter associated with the potential which would allow us to use perturbative
methods. Moreover, we had potentials which (especially the Coulomb potential felt
outside the nucleus) were changing slowly on the scale of the wavelength involved, so
it was the ideal ground to implement the WKB method.
Chapter 15
INTRODUCTION TO SCATTERING
15.1 introduction
The material for this chapter will be provided by the teacher via handwritten notes.
Chapter 16
THE SCHR
H r [ (t) (r, t) : R
n
R C
m
.
In what follows we will use for simplicity the position representation and so, besides
the states, also the operators ( p, r,
V , ecc.) will be given in that basis.
We know that the state of a quantum system in a potential V (r) evolves in time
according to the Schr odinger equation:
_
2
2m
2
+ V (r)
_
(r, t) = i
t
(r, t) (16.1)
290 Chapter 16. The Schr odinger Field and Second Quantization
Notice that (r, t), being a function, can be interpreted as a eld , and we can ask
ourselves if the evolution equation (16.1) can be derived via a variational principle
from some appropriate eld-Lagrangian. As (t, x) is a complex eld, it is equivalent
to two real and independent elds: e and m. Equivalently we can describe the
system using the eld and its complex coniugate:
_
_
= e + im
= e im.
We shall assume that and
is:
/ =
_
i
t
+
2
2m
2
_
. (16.2)
To get the Schr odinger equation for a particle in a a potential V , one just need to
add to the previous Lagrangian the term: V
V .
Let us now check that the EulerLagrange eld equation derived from the La-
grangian above is the (16.1). From the literature
we know that the EulerLagrange
equations for the elds
= 0 . (16.3)
The Schr odinger elds are just two ( and
= 0
_
i
t
+
2
2m
2
_
= 0 (16.4)
The students who do not know this should study the very rst chapters on elds equations
contained in many books like, for example, Goldstein or Landaus book.
291
In this manner we have eproduced (16.1).
To build the Hamiltonian density 1 associated to /, we have to build the momenta
associated to and
the momentum
coniugated to and with the momentum coniugated to
:
=
/
,
= 0
=
/
,
= i
.
Using all this the Hamiltonian density turns out to be :
1 (
, )
t
_
_
/ = i
t
i
2
2m
2
=
2
2m
2
_
.
(16.5)
If we integrate the expression above, the free Hamiltonian is :
H =
_
Q
1d
3
r =
_
Q
2
2m
2
_
d
3
r . (16.6)
Note that this expression coincides with the denition, in standard QM, of the ex-
pectation value of the kinetic operator
T on the state , which in the bra and ket
notation of Dirac can be written as:
T = [
T [ . (16.7)
Inserting the decomposition of the identity 11 in position space on the left and on the
right of the operator
T, we get the position picture representation of this expectation
value:
T =
_
Q
_
Q
[ x
[
T [x x[ d
3
xd
3
x
=
_
Q
(x)
_
2
2m
2
_
(x) d
3
x.
(16.8)
This is exactly the expression (16.6).
In the most general case in which the potential is not zero, the lagrangian (16.2)
becomes:
/ =
_
i
t
H
_
. (16.9)
292 Chapter 16. The Schr odinger Field and Second Quantization
Now let us concentrate on the Schr odinger equation for a free system and on the
interpretation of the wave function as a eld. Fields are now, like in the electromag-
netic case, our degrees of freedom but up to now we have not quantized these elds
(x, t), but only the operators x and p. Because of this, we can say that (16.2) is the
equation of motion of a classical eld. To work in analogy with the electromagnetic
eld, we can say that the Schr odinger equation corresponds to the Maxwell equations
and the (x, t) corresponds to the electromagnetic eld.
Let us take the Schr odinger equation for a free system and let us indicate with
+
n
the eigenfunctions associated to the positive eigenvalues E
0
n
:
2
2m
+
n
= E
0
n
+
n
. (16.10)
In this way we get the numerable set
+
n
which is a basis of the Hilbert space
where the eq. (16.10) is dened. A generic state H can be written as a linear
combinations of the states above because they make up a basis. The states
n
can
be represented on the position eigenstates. Moreover imposing periodic boundary
conditions and supposing we work in a nite space of volume V, they become:
+
n
(r, t) =
_
1
V
_1
2
e
i(pnrE
0
n
t)
(16.11)
where p
n
= 2(V )
1/3
(n
x
, n
y
, n
z
) and E
0
n
= p
2
n
/(2m). Remember that the ortogonal-
ity relation
_
V
j
dr =
i,j
holds. So a generic state can be written as:
(r, t) =
n
a
n
+
n
(r, t) (16.12)
where the expansion coecients a
n
are complex numbers. Using the expansion above
for the and inserting it into the Hamiltonian (16.6), we get:
H =
_
QR
3
_
n
a
n
__
2
2m
2
__
k
a
k
k
_
d
3
x =
_
Q
_
n
a
n
__
k
a
k
_
2
2m
2
_
k
_
d
3
x =
_
Q
_
n
a
n
__
k
a
k
E
0
k
k
_
d
3
x =
=
k
E
0
k
a
n
a
k
_
Q
k
d
3
x =
k
E
0
k
a
n
a
k
n,k
=
n
E
0
n
a
n
a
n
(16.13)
293
We can interpret this nal relation by saying that the energy of the generic eld , is
a combination of the energies of the single eigenstates of
H, and each of them is taken
with a weight related to the coecients a
n
of the projections of on the eigenstates
n
.
Next thing we want to do is quantize the eld that describes the system, that
means we must promote the eld (x) to be an operator. This procedure is called
second quantization. We can prove that what we obtain is equivalent to the rst
quantization of a system described by our Hamiltionian but containing an arbitrary
number of particles, this is a so called many body system.
The rest of the chapter is organized as follows: In section 1.1 we attack the many
body problem by building the wave function of the system (r
1
, ..., r
N
, t) and the
associated evolution equation. Next we perform a change of basis in order to write
the over-all Hamiltonian as a function of some creation and annihilation operators:
a
e a.
In 1.2 we show that the Hamiltonian derived in section 1.1 is formally equivalent
to the one built using the second quantization procedure. As we said earlier this
procedure turn the eld into an operator (
), and turn the Hamiltonian into
the following operator:
H
_
(x)H(x)
(x) dx.
The second quantization is widely used as the only correct technique to treat quan-
tum relativistic elds. This is a sector of physics where we always have to handle from
the beginning a many body system. In fact for example a photon, with its natural
creation of the many particle-antiparticle pairs from the vacuum along its motion , is
automatically a many body system.
294 Chapter 16. The Schr odinger Field and Second Quantization
16.1 The many body problem.
Let us consider a system of N identical particles interacting via a two body potential
V (r
i
, r
j
). In rst quantization the associated Hamiltonian has the following form:
H
(1
st
)
=
N
i=1
T
i
+
1
2
N
i=j=1
V (r
i
, r
j
) . (16.14)
The wave function, in position representation, which describes the system and its
associated Schr odinger equation are:
(r
1
, ..., r
N
, t) (r
N
, t) (16.15)
i
t
(r
N
, t) = H
(1
st
)
(r
N
, t) . (16.16)
can be build from the isingle particle, time independent states, which we shall
indicate with
(E)
(r
i
) for the iparticle. We choose these states which obey the
stationary Shr odinger eq.: H
(E)
(r
i
) = i
t
(E)
(r
i
) because they make up a basis
and have a clear physical meaning. As the particles are identical ones they all have
the same spectrum. Taking all of this into account we can then expand the many-body
wave function as:
(r
N
, t) =
E
i
C(E
1
, ...., E
N
, t)
E
1
(r
1
) ....
E
N
(r
N
) , (16.17)
where the sum over E
i
indicates the sum over all the energies of all single particles.
We have conned the time-dependence into the coecients C.
Let us insert the expansion (16.17) in (16.16), and multiply the equation for the
complex coniugate of a xed state:
E
1
(r
1
) ....
E
N
(r
N
). Next let us integrate what
we get over r
N
. In this way we get an equation for the coecients C associated to
a particular choice of the energies E
1
, ...., E
N
. We will call this equation as equation
for the coecients . Let us note that for a particular choice of the set of energies,
many dierent products of states
E
1
(r
1
) ....
E
N
(r
N
) may corresponds to that set of
energies, i.e. there may be degeneracies on that set of energies. To solve this problem
16.1. The many body problem. 295
let us dene the so called occupation number n
i
as the number of single-particle
states which happens to have energy E
i
.
At this point let us notice that it is possible to order, inside the coecients, the
energies which are equal to each other without having to change signs (this is possible
at least for bosoni ) and pass from a sum over the energies to a sum over the occupation
numbers:
E
i
(n
1
,....,n
x
)
(16.18)
where we have indicated with n
x
(x ,= N) , the last occupation number which is
dierent from zero. We can also imagine to do a sum over an innitly long strings of
occupation numbers (n
1
, .....; n
C(n
1
, ...., n
x
, t) C(E
1
, ...., E
N
, t) = C( E
1
..
n
1
, ..., E
x
..
n
x
, ..., E
N
, t). (16.19)
We can re-write the
C as :
C(n
1
, ...., n
x
, t)
_
N!
n
1
! .... n
x
!
_1
2
f(n
1
, ...., n
x
, t) . (16.20)
In the last step we have introduced some abstract functions f related to the C(E
1
, ...., E
N
, t).
In this manner we can re-write the total wave function in a new basis as:
(r
N
, t) =
(n
1
,....,n
x
)
f(n
1
, ...., n
x
, t)
(n
1
,....,n
x
)
(r
N
) (16.21)
where
(n
1
,....,n
x
)
(r
N
)
E
i
(n
1
,....,n
x
)
_
n
1
! .... n
x
!
N!
_1
2
E
1
(r
1
) ....
E
N
(r
N
) . (16.22)
The sum is done over all energies which are compatible with a given string of occu-
pation numbers (n
1
, ...., n
x
). It is for this reason that we put on a label given by a
296 Chapter 16. The Schr odinger Field and Second Quantization
string of occupation numbers. The states are in fact in a one to one correspondence
with the states [n
1
, ...., n
x
which belong to what is called a Fock space which will be
dened later on.
Let us remember that the C used in (16.17) have to obey what we called the
equation for the coecients, and as a consequence also f(n
1
, ...., n
x
, t) of equation
(16.21) must obey a similar equation modulo some factors in front. In order to be
able to use the dependence of f on the occupation numbers, we need to have only sum
over the energies. The sum over the numbers of particles in the Hamiltonian (16.14)
can be turned into a dierent sum as:
N
i=1
E
i
n
i
i
n
i
(16.23)
This replacement is very easy for the kinetic term
T while in the case of
V we have to
impose the constraint i ,= j which is present when we do the sum over the particles
number and we have to nd an analog one when we sum over the energies. This new
constraint is derived via the following reasoning : if E
i
,= E
j
, then the condition is
automatically satised and the occupation are simply n
i
ed n
j
. If instead the energies
are the same for two dierent particles, i.e. E
i
= E
j
, in that case we have to count
only once the particles labelled by dierent indices but that occupy that same energy
level. This means that we have to group together in couples, without repetition, the
number of particles at energy E
i
i.e, :
_
n
i
2
_
= n
i
(n
i
1) . (16.24)
To unify the two cases E
i
= E
j
e E
i
,= E
j
, we can write:
N
i=j
E
i
E
j
n
i
(n
j
i
j
)
j
n
i
(n
j
i
j
) , (16.25)
where the sums are over all admissible energies. The nal equation for the coecients
16.1. The many body problem. 297
f is:
i
t
f(n
1
, ...., n
x
, t) =
i
i[
T[i n
i
f(n
i
)+
i=j
i[
T[j (n
i
)
1
2
(n
j
+1)
1
2
f(n
i
1, n
j
+1)
+
1
2
i=j
l
ii[
V [kl (n
i
)
1
2
(n
i
1)
1
2
(n
k
+ 1)
1
2
(n
l
+ 1)
1
2
f(n
i
2, n
k
+ 1, n
l
+ 1)
+
1
2
i=j
l
ij[
V [kl (n
i
)
1
2
(n
j
)
1
2
(n
k
+1)
1
2
(n
l
+1)
1
2
f(n
i
1, n
j
1, n
k
+1, n
l
+1)
+ remaining terms for which k ,= l ,= (i o j). (16.26)
The compact notation used for f(n
i
+ 1), indicates that all n
j
, with j from 1 to
x, are the same as in the f(n
1
, ...., n
x
, t) which appears on the L.H.S of eq. (16.26)
except the literm which is turned into the (n
i
+ 1).
The crucial point embodied in (16.22) is that we have passed from the basis of the
energy eigenstates, (i.e. a sum over the energies), to an occupation number basis (i.e.
a sum over (n
1
, ...., n
x
)). It is possible to show that actually these are the position
picture representation of the abstract states which make up what is called the Fock
space. So we can write the total wave function as a linear combinations of states
of the Fock space as :
[(t) =
(n
1
,....,nx)
f(n
1
, ...., n
x
, t) [n
1
, ...., n
x
(16.27)
In order to derive (16.27) from (16.21) we just have to write
E
i
(r
i
) = r
i
[E
i
and
check that the coecients of the change of basis [E
1
, ...., E
N
[n
1
, ...., n
x
are the
same as in (16.22).
Let us now see how we can write the Hamiltonian in this new basis. Let us apply
i
t
to both sides of (16.27), and note that on the R.H.S the only dependence on time
is in the factors f, and so we can use the equation (16.26). Doing this and working
out the calculations in details, we can notice that it is possible to reconstruct on the
R.H.S the abstract state [(t) with the new Hamiltonian applied on it. This one is
basically all the R.H.S of (16.26) where each coecient f is multiplied by the same
298 Chapter 16. The Schr odinger Field and Second Quantization
ket [n
1
, ...., n
x
. Next we operate a change of variables n
i
1 n
i
etc. and we get
coecients f(n
1
, ...., n
x
) which are all equal and we get kets whose occupation numbers
corresponds exactly to the factors of the type (n
i
)
1
2
by which they are multiplied
example: f(n
1
, ...., n
x
) (n
i
+ 1)
1
2
[n
1
, ..., n
i1
, (n
i
+ 1), n
i+1
.... (16.28)
Let us remember that in the Fock space we can dene creation and annihilation
operators a
i
[...n
i
... = (n
i
+ 1)
1
2
[...n
i
+ 1...
a
i
[...n
i
... = (n
i
)
1
2
[...n
i
1...
and they derive from the well-known commuation relations.
As the new Hamiltonian contains terms of the form (16.28), we can rewrite it as a
function of the operators a
H
(many body)
i,j
a
i
i[
T[j a
j
+
i,j,k,l
a
i
a
j
ij[
V [kl a
k
a
l
. (16.29)
Note that the matrix elements i[
and a.
The Schr odinger equation for the many body problem, re-written in the Fock space,
is equivalent to the eq. eq:Schrodmany where we started from, i.e:
i
t
[(t)
E
=
H
(1
st
)
[(t)
E
i
t
[(t)
n
=
H
(many body)
[(t)
n
(16.30)
where with the symbols [(t)
E
and [(t)
n
we had indicated the ket [(t) in the two
dierent basis, respectively in the energy eigenstates and in the Fock space one. In
this last space the only information we need regarding the state [(t)
n
is its ground
state necause all the other states are obtained by applying the a
and a operators.
16.2. Second Quantization 299
16.2 Second Quantization
Second Quantization is basically a quick recipe to build an Hamiltonian equivalent
to the
H
(many body)
presented in the previous section. For semplicity we will use the
position representation. The prescriptions of that recipe are:
- Build the eld operator which is dened at every point in space as:
(r
N
)
i
(r
i
) a
i
(16.31)
where the
i
(r
i
) are the single particle eigenstate of energy E
i
(see the previous
sections).
- Build the Hamiltonian
i,j
i[
H
(1
st
)
[j a
i
a
j
=
_
(r
N
)H
(1
st
)
(r
N
)
(r
N
) dr
N
(16.32)
Remember that H
(1
st
)
(r
N
) is a complex number.
In general, the second quantization rule modies any operator not only the Hamilto-
nian and it does that using the analog of the (16.32).
The physical meaning of (16.32) is the following: let us suppose we take as [i, [j
two eigenstate of
H
(1
st
)
. Doing i[
H
(1
st
)
[j we get something proportional to
j
i
and
so only a sum over one index is left over:
i
i[
H
1
st [j a
i
a
i
=
i
E
i
(r) n
i
. (16.33)
So the operator
(n
1
,...,nx)
f(n
1
, ..., n
x
, t) [n
1
, ..., n
x
is given again by the Schroedinger equa-
tion:
i
t
[(t)
n
=
H[(t)
n
=
_
(r
N
)H
1
st (r
N
)
(r
N
) dr
N
[(t)
n
.
(16.34)
Let us now prove the equivalence between the two formulations: the many body and
the second quantization one. This proof will throw light also on the denition (16.31).
In eq. (16.29) let us insert some completness relation in position space between the
rst quantization operators
T and
V in order to represent everything in position
space and let us use the denition r
i
[i =
i
(r
i
). Moreover let us use the fact that
r[
T[r
= T(r)(r r
H
(many body)
=
_
dr
_
i
(r
i
) a
i
_
T(r)
_
j
(r
j
) a
j
_
+
_ _
i
(r
i
) a
i
__
j
(r
j
) a
j
_
V (r
i
, r
j
)
_
k
(r
k
) a
k
__
l
(r
l
) a
l
_
dr dr
(r)T(r)
(r) dr +
_
(r)
(r
)V (r, r
(r
(r) dr dr
H . (16.35)
This is what we wanted to show. So beside the relation (16.30), we also have:
i
t
[(t)
n
=
H
(many body)
[(t)
n
i
t
[(t)
n
=
H[(t)
n
(16.36)
16.3 Summary
We can say that we have obtained the Hamiltonian of second quantization following
two dierent roads:
- From the many body problem in rst quantization, via the change of basis
[E
1
, ...., E
N
[n
1
, ...., n
x
, we have built the Hamiltonian (16.29) as a function
of the creation and annihilation operators. As we showed in (16.35), it is then
sucient one more manipulation to get
H.
16.3. Summary 301
- Following the recipe of second quantization, we build the eld operator
and
with this the Hamiltonmian
H.
From the proof of the equivalence betyween
H
(many body)
and