QM PDF

Theoretical Physics III
Quantum Theory
Peter E. Blöchl
Caution! This version is not completed and may contain

errors
Institute of Theoretical Physics; Clausthal University of Technology;

D-38678 Clausthal Zellerfeld; Germany;
http://www.pt.tu-clausthal.de/atp/
2
© Peter Blöchl, 2000-July 6, 2011

Source: http://orion.pt.tu-clausthal.de/atp/phisx.html
Permission to make digital or hard copies of this work or portions thereof for personal or classroom
use is granted provided that copies are not made or distributed for profit or commercial advantage and
that copies bear this notice and the full citation. To copy otherwise requires prior specific permission
by the author.
1
1 Tothe title page: What is the meaning of ΦSX? Firstly, it reads like “Physics”. Secondly the symbols stand for
the three main pillars of theoretical physics: “X” is the symbol for the coordinate of a particle and represents Classical
Mechanics. “Φ” is the symbol for the wave function and represents Quantum Mechanics and “S” is the symbol for the
Entropy and represents Statistical Physics.
Foreword and Outlook
At the beginning of the 20th century physics was shaken by two big revolutions. One was the theory
of relativity, the other was quantum theory. The theory of relativity introduces a large but finite
parameter, the speed of light. The consequence is a unified description of space and time. Quantum
theory introduces a small but finite parameter, the Planck constant. The consequence of quantum
theory is a unified concept of particles and waves.
Quantum theory describes the behavior of small things. Small things behave radically different
than what we know from big things. Quantum theory becomes unavoidable to describe nature at
small dimensions. Not only that, but the implications of quantum mechanics determine also how our
macroscopic world looks like. For example, in a purely classical world all matter would essentially
collapse into a point, leaving a lot of light around.
Quantum mechanics is often considered difficult. It seems counterintuitive, because the picture
of the world provided in classical physics, which we have taken for granted, is inaccurate. Our mind
naturally goes into opposition, if observations conflict with the well proven views we make of our
world and which work so successfully for our everyday experiences. Teaching, and learning, quantum
mechanics is therefore a considerable challenge. It is also one of the most exciting tasks, because
it enters some fundamentally new aspects into our thinking. In contrast to many quotes, it is my
strong belief that one can “understand” quantum mechanics just as one can “understand” classical
mechanics2 . However, on the way, we have to let go of a number of prejudices that root deep in our
mind.
Because quantum mechanics appears counterintuitive and even incomplete due to its probabilistic
elements, a number of different interpretations of quantum mechanics have been developed. They
correspond to different mathematical representations of the same theory. I have chosen to start
with a description using fields which is due to Erwin Schrödinger, because I can borrow a number of
important concepts from the classical theory of electromagnetic radiation. The remaining mystery
of quantum mechanics is to introduce the particle concept into a field theory. However, other
formulations will be introduced later in the course.
I have structured the lecture in the following way:
In the first part, I will start from the experimental observations that contradict the so-called
classical description of matter.
I will then try to sketch how a theory can be constructed that captures those observations. I
will not start from postulates, as it is often done, but demonstrate the process of constructing the
theory. I find this process most exciting, because it is what a scientist has to do whenever he is faced
with a new experimental facts that do not fit into an existing theory. Quantum mechanics comes in
two stages, called first and second quantization. In order to understand what quantum mechanics
is about, both stages are required, even though the second part is considered difficult and is often
taught in a separate course. I will right from the beginning describe what both stages are about, in
order to clear up the concepts and in order to put the material into a proper context. However, in
order to keep the project tractable, I will demonstrate the concepts on a very simple, but specific
2 This philosophical question requires however some deep thinking about what understanding means. Often the
problems understanding quantum mechanics are rooted in misunderstandings about what it means to understand
classical mechanics.
3
4
example and sacrifice mathematical rigor and generality. This is, however, not a course on second
quantization, and after the introduction, I will restrict the material to the first quantization until I
come to the last chapter.
Once the basic elements of the theory have been developed, I will demonstrate the consequences
on some one-dimensional examples. Then it is time to prepare a rigorous mathematical formu-
lation. I will discuss symmetry in some detail. In order to solve real world problems it is important
to become familiar with the most common approximation techniques. Now we are prepared to ap-
proach physical problems such as atoms and molecules. Finally we will close the circle by discussing
relativistic particles and many particle problems.
There are a number of textbooks available. The following citations3 are not necessarily complete
or refer to the most recent edition. Nevertheless, the information should allow to locate the most
recent version of the book.
• Cohen-Tannoudji, Quantenmechanik[6]. In my opinion a very good and modern textbook with

lots of interesting extensions.
• Schiff, Quantum Mechanics[1]. Classical textbook.
• Gasiorewicz, Quantenphysik[2].
• Merzbacher, Quantum mechanics[3]
• Alonso and Finn, Quantenphysik[4],
• Atkins, Molecular Quantum Mechanics[7]. This book is a introductory text on quantum me-
chanics with focuses on applications to molecules and descriptions of spectroscopic techniques.
• Messiah, Quantum Mechanics[5]. A very good older text with extended explanations. Very
useful are the mathematical appendices.
• W. Nolting, Grundkurs Theoretische Physik 5: Quantenmechanik[8]. Compact and detailed
text with a lot of problems and solutions.
• C. Kiefer Quantentheorie [9] Not a text book but easy reading for relaxation. [11]J.-L. Bas-
devant and J. Dalibard, Quantum mechanics. A very good course book from the Ecole Poly-
technique, which links theory well upon modern themes. It is very recent (2002).
3 Detailed citations are compiled at the end of this booklet.

Contents
1 Waves? Particles? Particle waves! 13
2 Experiment: the double slit 15

2.1 Macroscopic particles: playing golf . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Macroscopic waves: water waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Microscopic waves: light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Microscopic particles: electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3 Towards a theory 25
3.1 Particles: classical mechanics revisited . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1.1 Hamilton formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Waves: the classical linear chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.1 Equations of motion ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.2 ... and their solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Continuum limit: transition to a field theory . . . . . . . . . . . . . . . . . . . . . . 32
3.4 Differential operators, wave packets, and dispersion relations . . . . . . . . . . . . . 33
3.5 Breaking translational symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.6 Introducing mass into the linear chain (Home study) . . . . . . . . . . . . . . . . . 39
3.7 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.8 Postulates of Quantum mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.9 Is Quantum theory a complete theory? . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.10 Planck constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.11 Wavy waves are chunky waves: second quantization . . . . . . . . . . . . . . . . . . 45
3.12 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4 Quantum effects 47
4.1 Quasi-one-dimensional problems in the real world . . . . . . . . . . . . . . . . . . . 48
4.2 Time-independent Schrödinger equation . . . . . . . . . . . . . . . . . . . . . . . . 50
4.3 Method of separation of variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.4 Free particles and spreading wave packets . . . . . . . . . . . . . . . . . . . . . . . 54
4.4.1 Wave packets delocalize with time . . . . . . . . . . . . . . . . . . . . . . . 55
4.4.2 Wick rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.4.3 Galilei invariance (home study) . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.5 Particle in a box, quantized energies and zero-point motion . . . . . . . . . . . . . . 58
4.5.1 Time-dependent solutions for the particle in the box . . . . . . . . . . . . . 62
5
6 CONTENTS
4.6 Potential step and splitting of wave packets . . . . . . . . . . . . . . . . . . . . . . 65

4.6.1 Kinetic energy larger than the potential step . . . . . . . . . . . . . . . . . 65
4.6.2 Construct the other partial solution using symmetry transformations . . . . . 67
4.6.3 Kinetic energy smaller than the potential step . . . . . . . . . . . . . . . . . 72
4.6.4 Logarithmic derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.7 Barrier and tunnel effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.8 Particle in a well and resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5 Language of quantum mechanics 81

5.1 Kets and the linear space of states . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.1.1 Axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.1.2 Bra’s and brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1.3 Some vocabulary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.2 Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.2.1 Axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.2.2 Some vocabulary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.3 Analogy between states and vectors . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.4 The power of the Dirac notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.5 Extended Hilbert space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6 Application: harmonic oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6.1 Algebraic treatment of the one-dimensional harmonic oscillator . . . . . . . 93
5.6.2 Wave functions of the harmonic oscillator . . . . . . . . . . . . . . . . . . . 96
5.6.3 Multidimensional harmonic oscillator . . . . . . . . . . . . . . . . . . . . . . 97
6 Representations 101
6.1 Unity operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.2 Representation of a state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.3 Representation of an operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.4 Change of representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.5 From bra’s and ket’s to wave functions . . . . . . . . . . . . . . . . . . . . . . . . 104
6.5.1 Real-space representation . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.5.2 Momentum representation . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.5.3 Orthonormality condition of momentum eigenstates (Home study) . . . . . 107
6.6 Application: Two-state system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.6.1 Pauli Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.6.2 Excursion: The Fermionic harmonic oscillator (Home study) . . . . . . . . . 111
7 Measurements 113
7.1 Expectation values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
7.2 Certain measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
7.2.1 Schwarz’ inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
7.3 Eigenstates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.4 Heisenberg’s uncertainty relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.5 Measurement process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.6 Kopenhagen interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
CONTENTS 7
7.7 Decoherence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

7.8 Difference between a superposition and a statistical mixture of states . . . . . . . . 123
7.8.1 Mixture of states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7.8.2 Superposition of states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7.8.3 Implications for the measurement process . . . . . . . . . . . . . . . . . . . 124
8 Dynamics 125
8.1 Dynamics of an expectation value . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
8.2 Quantum numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
8.3 Ehrenfest Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.4 Particle conservation and probability current . . . . . . . . . . . . . . . . . . . . . . 131
8.5 Schrödinger, Heisenberg and Interaction pictures . . . . . . . . . . . . . . . . . . . 132
8.5.1 Schrödinger picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
8.5.2 Heisenberg picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
8.5.3 Interaction picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
9 Symmetry 137
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
9.1.1 Some properties of unitary operators . . . . . . . . . . . . . . . . . . . . . . 140
9.1.2 Symmetry groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.2 Finite symmetry groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
9.3 Continuous symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.3.1 Shift operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
9.4.1 Translation operator from canonical momentum . . . . . . . . . . . . . . . 144
10 Specific Symmetries 147

10.1 Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
10.2 n-fold rotation about an axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
10.3 Exchange of two particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
10.3.1 Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
10.3.2 Bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
10.4 Continuous translations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
10.5 Discrete translations, crystals and Bloch theorem . . . . . . . . . . . . . . . . . . . 154
10.5.1 Real and reciprocal lattice: One-dimensional example . . . . . . . . . . . . . 154
10.5.2 Real and reciprocal lattice in three dimensions . . . . . . . . . . . . . . . . 155
10.5.3 Bloch Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
10.5.4 Schrödinger equation in momentum representation . . . . . . . . . . . . . . 161
10.6 Some common Lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
10.6.1 Cubic lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
10.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
10.7.1 Symmetrization of states . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
10.7.2 Dirac comb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
11 Rotations, Angular Momentum and Spin 171

11.1 Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
11.1.1 Derivation of the Angular momentum . . . . . . . . . . . . . . . . . . . . . 172
8 CONTENTS
11.1.2 Another derivation of the angular momentum (Home study) . . . . . . . . . 172

11.2 Commutator Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
11.3 Algebraic derivation of the eigenvalue spectrum . . . . . . . . . . . . . . . . . . . . 174
11.4 Eigenstates of angular momentum: spherical harmonics . . . . . . . . . . . . . . . . 180
11.5 Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.6 Addition of angular momenta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
11.7 Products of spherical harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
11.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
12 Atoms 191
12.1 Radial Schrödinger equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
12.2 The Hydrogen Atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
13 Approximation techniques 203

13.1 Perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
13.2 General principle of perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . 204
13.3 Time-independent perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . . 204
13.3.1 Degenerate case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
13.3.2 First order perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . . 208
13.3.3 Second order perturbation theory . . . . . . . . . . . . . . . . . . . . . . . 209
13.4 Time-dependent perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . . . 210
13.4.1 Transition probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
13.5 Variational or Rayleigh-Ritz principle . . . . . . . . . . . . . . . . . . . . . . . . . . 215
13.6 WKB-Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
13.6.1 Classical turning points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
13.7 Numerical integration of one-dimensional problems . . . . . . . . . . . . . . . . . . 222
13.7.1 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
13.7.2 Natural boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 224
13.7.3 Radial wave functions for atoms . . . . . . . . . . . . . . . . . . . . . . . . 224
13.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
13.8.1 A model for optical absorption . . . . . . . . . . . . . . . . . . . . . . . . . 225
13.8.2 Gamow’s theory of alpha decay . . . . . . . . . . . . . . . . . . . . . . . . 227
13.8.3 Transmission coefficient of a tunneling barrier . . . . . . . . . . . . . . . . . 227
13.8.4 Fowler-Nordheim tunneling . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
14 Relativistic particles 229

14.1 A brief review of theory of relativity . . . . . . . . . . . . . . . . . . . . . . . . . . 229
14.2 Relativistic Electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
14.3 Electron in the electromagnetic field . . . . . . . . . . . . . . . . . . . . . . . . . . 232
14.4 Down-folding the positron component . . . . . . . . . . . . . . . . . . . . . . . . . 233
14.5 The non-relativistic limit: Pauli equation . . . . . . . . . . . . . . . . . . . . . . . . 234
14.6 Relativistic corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
14.6.1 Spin-Orbit coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
14.6.2 Darwin term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
14.6.3 Mass-velocity term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
15 Many Particles 239

CONTENTS 9
15.1 Bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

15.1.1 Quantum mechanics as classical wave theory . . . . . . . . . . . . . . . . . 239
15.1.2 Second quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
15.1.3 Coordinate representation . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
15.2 Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
15.3 Final remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
I Appendix 245
A Galilei invariance 247
B Spreading wave packet of a free particle 249

B.1 Probability distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
C The one-dimensional rectangular barrier 253

C.1 E > V , the barrier can be classically surmounted . . . . . . . . . . . . . . . . . . . 254
C.2 Tunneling effect E < V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
D Particle scattering at a one-dimensional square well 259

D.1 Square well E > 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
D.2 Square well E < 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
E Alternative proof of Heisenberg’s uncertainty principle 263
F Spherical harmonics 267

F.1 Spherical harmonics addition theorem . . . . . . . . . . . . . . . . . . . . . . . . . 267
F.2 Unsöld’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
F.3 Condon-Shortley phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
G Random Phase approximation 269

G.1 Repeated random phase approximation . . . . . . . . . . . . . . . . . . . . . . . . . 271
G.2 Transition matrix element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
G.3 Rate equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
G.4 Off-diagonal elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
H Propagator for time dependent Hamiltonian 275
I Matching of WKB solution at the classical turning point 277
J Positrons 281
J.1 The action of quantum electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . 283
K Method of separation of variables 287
L Trigonometric functions 289
M Gauss’ theorem 291

10 CONTENTS
N Fourier transform 293

N.1 General transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
N.2 Fourier transform in an finite interval . . . . . . . . . . . . . . . . . . . . . . . . . . 293
N.3 Fourier transform on an infinite interval . . . . . . . . . . . . . . . . . . . . . . . . 294
N.4 Table of Fourier transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
N.5 Dirac’s δ-function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
O Pauli Matrices 297
P Linear Algebra 299

P.1 Half Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
P.2 Vector space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
P.3 Banach space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
P.4 Hilbert space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
Q Matrix identities 301

Q.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Q.2 Identities related to the trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Q.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Q.2.2 Invariance under commutation of a product . . . . . . . . . . . . . . . . . . 301
Q.2.3 Invariance under cyclic permutation . . . . . . . . . . . . . . . . . . . . . . 302
Q.2.4 Invariance under unitary transformation . . . . . . . . . . . . . . . . . . . . 302
Q.3 Identities related to the determinant . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Q.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Q.3.2 Product rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Q.3.3 Permutation of a product . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Q.3.4 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
R Special Functions 305

R.1 Bessel and Hankel functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
R.2 Hermite Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
R.3 Legendre Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
R.4 Laguerre Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
S Principle of least action for fields 309
T ℓ-degeneracy of the hydrogen atom 311

T.1 Laplace-Runge-Lenz Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
T.1.1 Commutator relations of Runge-Lenz vector and angular momentum . . . . 312
T.1.2 Rescale to obtain a closed algebra . . . . . . . . . . . . . . . . . . . . . . . 312
T.2 SO(4) symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
T.3 ℓ-Degeneracy of the hydrogen atom . . . . . . . . . . . . . . . . . . . . . . . . . . 313
T.4 Derivation of commutator relations used in this chapter . . . . . . . . . . . . . . . . 314
~ˆ 2 . . . . . . . . . . . . . . . . . . . . .
T.4.1 Calculation of M . . . . . . . . . . . 316
T.4.2 Commutator of M ~ˆ and H~ˆ . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
T.4.3 Commutators of M ~ˆ and L
~ˆ . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
T.4.4 Commutators of J~ and J~ˆ(2) . . . . . . . . . . . . . .
ˆ (1)
. . . . . . . . . . . 323
CONTENTS 11
U A small Dictionary 325
V Greek Alphabet 329
W Philosophy of the ΦSX Series 331
X About the Author 333

12 CONTENTS
Chapter 1
Waves? Particles? Particle waves!
Quantum mechanics is about unifying two concepts that we use to

describe nature, namely that of particles and that of waves.
• Waves are used to describe continuous distributions such as wa-

ter waves, sound waves, electromagnetic radiation such as light
or radio waves. Even highway traffic has properties of waves with
regions of dense traffic and regions with low traffic. The char-
acteristic properties of waves are that they are delocalized and
continuous.
• On the other hand we use the concept of particles to describe the

motion of balls, bullets, atoms, atomic nuclei, electrons, or in the
example of highway traffic the behavior of individual cars. The
characteristic properties of particles are that they are localized
and discrete. They are discrete because we cannot imagine half
a particle. (We may divide a bigger junk into smaller pieces, but
Fig. 1.1: Louis Victor
then we would say that the larger junk consisted out of several
de Broglie, 1892-1987.
particles.)
French physicist. Nobel
price in Physics 1929 for
However, if we look a little closer, we find similarities between these
the de Broglie relations
concepts: E = ~ω and p = ~k,
linking energy to frequency
• A wave can be localized in a small region of space. For example,
and momentum to inverse
a single water droplet may be considered a particle, because it
wavelength, which he pos-
is localized, and it changes its size only slowly. Waves can also
tulated 1924 in his PhD
be discrete, in the case of bound states of a wave equation: A
thesis.[1]
violin string (g: Violinensaite) can vibrate with its fixed natural
vibrational frequency, the pitch, or with its first, second, etc. harmonic. The first overtone
has two times the frequency of the pitch (g:Grundton), the second overtone has three times
the frequency of the pitch and so on. Unless the artist changes the length of the vibrating
part of the chord, the frequencies of these waves are fixed and discrete. We might consider
the number of the overtone as analogous to number of particles, which changes in discrete,
equi-spaced steps. This view of particles as overtones of some vibrating object is very close to
particle concept in quantum field theory.
• If we consider many particles, such as water molecules, we prefer

to look at them as a single object, such as a water wave. Even if
we consider a single particle, and we lack some information about
it, we can use a probability distribution to describe its whereabouts
13
14 1 WAVES? PARTICLES? PARTICLE WAVES!
at least in an approximate manner. This is the realm of statistical

mechanics.
Quantum mechanics shows that particles and waves are actually two aspects of more general
objects, which I will call particle waves. In our everyday life, the two aspects of this generalized
object are well separated. The common features become evident only at small length and time
scales. There is a small parameter, namely the Planck constant h, that determines when the two
aspects of particle waves are well separated and when they begin to blur1 . In practice, the reduced
def h
Planck constant “hbar”, ~ = 2π , is used2 . Compared to our length and time scales, ~ is so tiny
that our understanding of the world has sofar been based on the assumption that ~ = 0. However,
as one approached very small length and time scales, however, it became evident that ~ > 0, and a
few well established concepts, such as those of particles and waves, went over board.
A similar case, where a new finite parameter had to be introduced causing a lot of philosophical
turmoil is the theory of relativity. The speed of light, c, is so much larger than the speed of the
fastest object we can conceive, that we can safely regard this quantity as infinite in our everyday
life. More accurate experiments, however, have found that c < ∞. As a result, two very different
quantities, space and time, had to be unified.
Interestingly, the concept of a maximum velocity such as the speed of light, which underlies the
theory of relativity, is a natural (even though not necessary) consequence of a wave theory, such as
quantum mechanics. Thus we may argue that the theory of relativity is actually a consequence of
quantum mechanics.
1 “to
blur” means in german “verschwimmen”
2 The
Planck constant is defined as h. This has been a somewhat unfortunate choice, because h appears nearly
always in combination with 1/(2π). Therefore, the new symbol ~ = h/(2π), denoted reduced Planck constant has
been introduced. The value of ~ is ≈ 10−34 Js to within 6%.
Chapter 2
Experiment: the double slit
The essence of quantum mechanics becomes evident from a single

experiment[2, 3, 4, 5]: the double-slit experiment. The experiment
is simple: We need a source of particles or waves, such as golf balls,
electrons, water waves or light. We need an absorbing wall with two
holes in it and behind the wall, at a distance, an array of detectors or
a single detector that can be moved parallel to the wall with the holes.
We will now investigate what the detector sees.
2.1 Macroscopic particles: playing golf

Let us first investigate the double-slit experiment for a simple case,
namely macroscopic particles such as golf balls.
Fig. 2.1: Thomas Young,

1773-1829. English physi-
cian and physicist. Es-
tablished a wave theory of
light with the help of the
double-slit experiment.
Fig. 2.2: Double-slit experiment for particles. Particles are shot randomly against a wall, which
contains two slits. Some particles pass through one or the other slit in the wall and proceed to the
next wall. The number of particles arriving at the second wall is counted as function of their position,
yielding a probability distribution as function of their position. This probability distribution, shown as
full line at the very right, is the superposition of the probability distributions, shown as dashed lines,
which one obtains if one or the other slit is closed.
We take a golfer as a source of golf-balls. We build a wall with two slits. As detector we simply
use little baskets that collect the golf balls that pass through the holes. By counting the number
of balls in each basket that arrive per unit time, we obtain a distribution P12 (y ) as function of the
basket position. We observe a distribution with two maxima, one behind each hole. If the two slits
15
16 2 EXPERIMENT: THE DOUBLE SLIT
are close together the two maxima may also merge into one.
Let us repeat the experiment with the second slit closed. The golfer shots the balls at the same
rate at the wall with the single remaining slit. This experiment yields a distribution of golf balls P1 (y ).
Then we repeat the experiment with the first hole closed and obtain the distribution P2 (y ).
For classical particles the three distributions are not independent but they are related by
P12 (y ) = P1 (y ) + P2 (y ) ,
that is the two distributions P1 (y ) and P2 (y ) simply add up to P12 (y ).

The explanation for this result is that the ball passes either through one or the other hole. If
it passes through one hole, it does not feel the other hole, and vice versa. Hence, the probabilities
simply add up.
2.2 Macroscopic waves: water waves

Now, let us turn to waves we understand well, namely water waves.
Fig. 2.3: Double-slit experiment for waves. A source on the left emits waves, which pass through
two slits in a first wall. The intensity, which is proportional to the maximum squared amplitude at
the second wall is monitored and shown on the very right. One observes an interference pattern
for the intensity as the waves emerging from the two slits interfere. They interfere constructively in
certain directions from the double slit, indicated by the dashed lines, leading to a maximum of the
intensity, and they interfere destructively in other directions, leading to vanishing intensity. While the
amplitudes of the waves emerging from the two slits add up to the total amplitude, the intensities
obtained, when one or the other slit is closed, do not add up to the total amplitude.
The water wave is created in this experiment by a finger that periodically moves up above and
down below the water surface. The frequency is characterized by the angular frequency ω.1 The
water wave moves radially outward. The height of the water surface has the form
h ω i
φ(~r, t) = A(|~r|) sin (k|~r| − ωt) = A(|~r|) sin k |~r| − t (2.1)
k
It is assumed that the finger is located at the origin of the coordinate system. We see2 that the crests
and troughs of the wave move outward with constant velocity ω/k. The wave crests are separated
by the wave length λ = 2π/k. The amplitude of the oscillations decreases with the distance from
the source, because the energy of the wave is distributed over a larger region as the wave moves
1 The
angular frequency is 2π times the frequency or ω = 2π/T where T is the period of the oscillation.
2 Consider
a wave crest, which forms a line defined by those points for which the argument of the sinus has the
r (t)| = π+2πn
value π + 2πn with an integer n. Hence the line is defined by |~ k
+ ωk t. Thus the velocity of the wave crest
d|~
r (t)| ω
is v = dt
= k
.
2 EXPERIMENT: THE DOUBLE SLIT 17
further away from the source3 . This effect is described by the function A(|~r|). At a given point in
space, the water surface oscillates up and down with a period of T = 2π/ω. The frequency of the
oscillation is f = ω/(2π).
Fig. 2.4: Spherical wave
When the water wave hits the wall, it is reflected except for the two slits, where it can pass, as
shown in Fig. 2.3. We have chosen the slits smaller than the wave length of the water wave. As
a result, the water level inside the slits moves up and down. The slits can therefore be considered
as point sources for water waves themselves. Important is that these sources are not independent,
but they are synchronized. We call such waves coherent. Behind the wall, two waves extend in
half-circles away from the two slits. Where the water waves penetrate each other, their amplitudes
are superimposed. The detector observes, at each point in space, a periodic oscillation of the water
surface.
Depending on the position of the detector, the waves originating from the two slits had to travel
different distances to the detector. If the traveling distances from the slits to the detector differ by
one-half of a wave length, one of the waves moves upward at the site of the detector, while the other
wave moves downward. Since the two waves add up to zero, the detector will show no signal. If the
traveling distances of the two waves are equal or if they differ by a multiple of a full wave length λ,
the two waves oscillate synchronously, and, as a result, they add up to a large signal at the detector.
This effect is called interference. As the detector is displaced parallel to the wall, it passes through
regions with large and small amplitudes of oscillation of the water surface, that is through regions of
constructive interference and destructive interference. Thus the distribution oscillates with the
distance from the center of the two slits, as it is sketched on the right side of Fig. 2.3.
Let us make this result a little more quantitative and work out the distribution: A wave φ(~r, t)
3 The wave transports energy away from the center. Because of energy conservation, the energy flux through the
surface of any surface enclosing the source must be identical. Note that, in two dimensions, this surface is a line. Since
the surface area increases with distance and because the energy density must depend in some way on the amplitude, the
energy-flux density must decrease with distance. The energy flux density is proportional to the square of the amplitude,
and the surface, a half-circle has the length ℓ = πr is, in two dimensions, proportional to the distance r , we obtain
A(~r ) ∼ √1πr .
originating from a single point source, such as one of the slits, has the form
Eq. 2.1
φ(~r, t) = A(ℓ) sin(kℓ − ωt), (2.2)
where ℓ = |~r − ~r0 | is the distance of the detector at position ~r from the point source at ~r0 . In our
experiment the two slits play the role of the point sources.
1
y
111
000
000
111
000
111
d 2
000
111
Fig. 2.5: Demonstration of the variables used for the calculation of the distribution of the double slit
experiment for classical waves.
The two waves originate fromqslits separated by a distance q d. We denote the distances of the
detector from each slit as ℓ1 = x + (y − 2 ) and ℓ2 = x 2 + (y + d2 )2 respectively. x is the
2 d 2
distance of the screen from the slits and y is the position on the screen relative to the projection of
the mid plane of the slits. Similarly we denote the wave amplitudes as φ1 and φ2 .
In order to design the experiment such that we can compare it with the corresponding particle
experiment, we consider here the intensity of the wave instead of the amplitude.
The intensity is the time averaged energy flux of the wave. We have rescaled the wave amplitude
such that the proportionality constant between energy flux and half of the squared wave amplitude
is absorbed in the latter. Thus the intensity can be written in the form
Z
1 2 1 T 1
I(~r) = φ (~r, t) = lim dt φ2 (~r, t) (2.3)
2 t T →∞ T 0 2
The angular brackets denote the time average.


1 1
Let us now calculate the intensity I12 = 2 2 (φ1 + φ2 ) 2
of the superposed waves from the two
t
slits, where φ1 = A(ℓ1 ) sin(kℓ1 − ωt) and φ2 = − ωt).
A(ℓ2 ) sin(kℓ2
def 1 def
We introduce the mean distance ℓ0 = 2 (ℓ1 +ℓ2 ) and the difference ∆ = ℓ2 −ℓ1 , so that ℓ1 = ℓ0 −∆/2
def def
and ℓ2 = ℓ0 + ∆/2. Similarly we use the short hand A1 = A(ℓ1 ) and A2 = A(ℓ2 )
The sinus can be decomposed according to
Eq. L.4
sin(a + b) = sin(a) cos(b) + cos(a) sin(b) , (2.4)
where a = kℓ0 − ωt and b = ±k∆/2, to obtain
∆ Eq. 2.4 ∆ ∆
sin(kℓ0 ± k − ωt) = sin(kℓ0 − ωt) cos(k ) ± cos(kℓ0 − ωt) sin(k ) (2.5)
2 2 2
Similarly we can write
1 1
A1 = (A1 + A2 ) + (A1 − A2 )
2 2
1 1
A2 = (A1 + A2 ) − (A1 − A2 ) (2.6)
2 2
so that we obtain for the superposition of the two waves
φ1 + φ2 = A(ℓ1 ) sin(kℓ1 − ωt) + A(ℓ2 ) sin(kℓ2 − ωt)

Eq. 2.5,Eq. 2.6
= (A1 + A2 ) sin(kℓ0 − ωt) cos(k∆/2)
+ (A1 − A2 ) cos(kℓ0 − ωt) sin(k∆/2) (2.7)
Now we introduce the following shorthand notations R, S, α

def
R = (A1 + A2 ) cos(k∆/2) (2.8)
def
S = (A1 − A2 ) sin(k∆/2) (2.9)
def
p
cos(α) = R/ R2 + S 2 (2.10)
p
sin(α) = S/ R2 + S 2 (2.11)
The last relation for the sinus follows directly from the identity cos2 (α) + sin2 (α) = 1.
Eq. 2.7
φ1 + φ2 = R sin(kℓ0 − ωt) + S cos(kℓ0 − ωt)
p h i
= R2 + S 2 sin(kℓ0 − ωt) cos(α) + cos(kℓ0 − ωt) sin(α)
Eq. 2.4
p
= R2 + S 2 sin(kℓ0 − ωt + α) (2.12)
In the last step we used again Eq. 2.4. cont’d

From the last equation we can read that the intensity of the oscillation averaged over time is given
by the squared pre-factor R2 + S 2 multiplied by a factor 1/2 that results from the time average of
the oscillatory part.
Z
Eq. 2.3 1 T 1 Eq. 2.12 1
I12 = lim dt (φ1 (t) + φ2 (t))2 = R2 + S 2
T →∞ T 0 2 4
Eqs. 2.8,2.9 1 1
= (A1 + A2 )2 cos2 (k∆/2) + (A1 − A2 )2 sin2 (k∆/2)
4 4
Eq. L.2 1 2 1 2 1
= A + A + A1 A2 (cos (k∆/2) − sin2 (k∆/2))
2
4 1 4 2 2
Eq. L.2 1 2 1 2 1
= A + A + A1 A2 (1 − 2 sin2 (k∆/2))
4 1 4 2 2
Eq. L.7 1 2 1 2 1
= A + A2 + A1 A2 cos(k∆))
4 1 |{z}
|{z} 4 2
I1 I2
One factor 12 in the proportionality between intensity and A2 stems from the definition of the intensity
and the other from the time average of the squared sinus function.
INTERFERENCE PATTERN IN THE INTENSITY

D E
Eq. 2.3 1
Thus we obtain the intensity I12 (y ) = 2 (φ 1 (y , t) + φ 2 (y , t)) 2
when both slits are open as
t
p
I12 (y ) = I1 (y ) + I2 (y ) + 2 I1 (y )I2 (y ) cos(k∆(y )) (2.13)
| {z }
Interference term

1 2 1 2
from the intensities I1 (y ) = 2 φ1 (y ) and I2 (y ) = 2 φ2 (y ) that are obtained, when one or the
t t
other slit is closed
∆(y ) in Eq. 2.13 is a function 4 that varies monotonously from −d for y = −∞ to +d at y = +∞

as shown in Fig. 2.6. Close to the center, that is for y = 0 and if the screen is much farther from
the slits than the slits are separated, ∆ is approximately ∆(y ) = y d/x + O(y 2 ). Hence the spatial
oscillation near the center of the screen has a period 5 of 2πx/(kd).
An important observation is that the intensities do not simply add up as the probabilities do for
particles. The first two terms are just the superposed intensities of the contribution of the two slits.
If we interpret the intensities as probabilities for a particle to arrive, they correspond to the probability
that a particle moves through the left slit and the probability that it moved to the right slit. The
4
r r s s
1 2 1 2 d2 d2
∆(y ) = ℓ2 − ℓ1 = x2 2
+ (y + d) − x + (y − d) = x + y + 2 2 + dy − x 2 + y 2 + + dy
2 2 4 4
s "s s #
d2 yd yd
= x2 + y 2 + 1+ 2 − 1 − 2
4 x 2 + y 2 + d4 x 2 + y 2 + d4
"s s #
d<<x,y p yd yd Taylor p 2 yd y
≈ x2 + y 2 1+ 2 2
− 1− 2 2
= x + y2 2 =dp
x +y x +y x + y2 x2 + y 2
def
sin γ = √ y
x 2 +y 2
= sin(γ)d
5 The period Λ of cos(k∆(y ) is given by k∆(y + Λ) = k∆(y ) + 2π. In the center of the screen, i.e. y = 0 and for
−1
small wave length, we can Taylor-expand ∆ and obtain k d∆
dy
Λ = 2π. Thus the period is Λ = 2π
k
d∆
dy
= 2πx
kd
.
1
∆
d
0
−1
−40 −20 0 20 y
Fig. 2.6: ∆(y ) = ℓ2 − ℓ1 for a a double slit with d = 1 and a screen separated from the slit by
x = 10. The function ∆ is approximately linear ∆(y ) ≈ d/x in the center of the interference pattern,
i.e. for y ≈ 0, and it is nearly constant for x >> d. For large deflections, it saturates at a value
∆(|y | >> x) = ±d.
last, oscillatory term is a property of waves called interference term.

The interference term oscillates more rapidly if the wave length λ = 2π/k becomes small com-
pared to the distance of the slits, i.e. λ << d. This is shown in Fig. 2.7. In the extreme case of a
very short wave length, the interference term oscillates so rapidly that the oscillations are averaged
out by any experiment that has a finite resolution. In this limit, the oscillations become invisible. In
such an ’inaccurate’ experiment, the wave intensities behave additive just as probabilities for particles.
Hence, waves with very short wave lengths behave in a certain way like particles. Waves with longer
wave lengths do not.
DISAPPEARANCE OF THE INTERFERENCE TERM FOR SMALL WAVE LENGTH
In the limit of very small wave length, the interference term is averaged out in any experiment having
a finite resolution. The non-oscillatory term of the intensities, which can still be observed, behaves
additive just like the probability of particles. Thus the wave character of particles disappears in the
limit of short wavelength.
This is a very important result, which lies at the heart of measurement theory and the classical
limit of quantum mechanics that will be described later.
2.3 Microscopic waves: light

Sofar, everything sounds familiar. Now let us switch to the microscopic world and investigate light,
which has a much shorter wave length:
The wave length of visible light varies from 750 nm for red light to 400 nm for blue light. If
the wave length is longer than for visible light, we pass through infrared, namely heat radiation, to
microwaves as they are used for radio and mobile telephony. For wave lengths shorter than that of
visible light, we pass through ultraviolet light up to X-ray radiation. The wave length of visible light
is still large compared to atomic dimensions. Atoms in a crystal have a typical distance of 1-2 Å or
0.1-0.2 nm, which is why X-rays need to be used to determine crystal structures.
If we perform the two-slit experiment with light, we need to use photo-detectors. A photo-
detector contains electrons loosely bound to defect atoms in a crystal. When such a crystal is placed
in an electric field, such as that of a light beam, the electrons can follow the applied electric field,
resulting in a measurable current.
−10 −5 0 5 10
Fig. 2.7: Interference pattern of a two dimensional wave through a double slit with separation
d = 5, on a screen located at a distance z = 1 behind the slits. The wave length has values
λ = 0.2, 2, 20, 200 with increasing values from the bottom to the top. Each figure contains in
addition the sum of intensities from the individual slits (yellow dashed line), which would correspond
to a pure particle picture.
Let us now return to the double-slit experiment. For a regular light source, we obtain the same
interference pattern as for water waves, only a lot smaller in its dimensions.
What happens as we slowly turn down the light? At first, the intensity measured by the photo
detector becomes smaller, but the signal maintains its shape. If we turn the light further down, after
the intensity is far too small to be observed with the naked eye, the signals begins to flicker and
for even lower intensity random independent flashes of very low intensity are observed. If we use an
array of photo detectors at different positions, only one detects a flash at a time, while the others
are completely silent. (The experimentalists assured me that this is not a problem with their photo
detectors.) If we add up the signals over time, the original interference pattern is observed.
Thus if the light is turned down, it behaves lumpy, like particles! These particles are called
photons.
Apparently, if the light is strong, many photons are detected in short time intervals, so that we
obtain a signal that appears continuous. In reality we see many particles. Thus our experience has
misguided us, and we have introduced a concept of a continuous wave, even though we actually
observe particles.
2.4 Microscopic particles: electrons

Let us now go to the microscopic world of particles. We choose electrons as example. We mount a
small metallic tip so that it points towards a metallic plate that has two slits in it. Between tip and
metal plate we apply a large voltage. Near the tip the electric field becomes so large, that it rips
electrons out of the tip, which in turn travel towards the metallic plate. The metallic plate has two
slits, which allows some electrons to pass through.
Electrons that pass through the slits are collected by a detector. We count the electrons that
arrive at the detector, when the detector is at a given position. The electrons pass one by one, and
enter the detector as individual particles, but when we count a large number of them as a function of
position of the detector, we obtain a clear interference pattern as in the case of light! This is shown
in Fig. 2.8.
We remember that the interference pattern for water
waves resulted from the wave passing through both holes
simultaneously. We expect that a particle, which passes
through either one or the other hole cannot have an inter-
ference pattern.
Let us change the experiment. We close one hole,
collect the signals, then close the other, and add up the
signals. The result are two maxima, but no interference
pattern. Apparently there is not only the possibility that
the particle passes through one or the other slit, but also
the probability that it passes through both slits simultane-
ously. The latter possibility makes the difference between
the two humps, which are expected for particles, and the
interference pattern, expected for waves.
A movie and demonstration of the double-slit experi-
ment can be found at the Hitachi Web site http://www.
hqrd.hitachi.co.jp/em/doubleslit.cfm
2.5 Summary
We have learned that waves such as light are composed of
fixed quanta, which we may identify with particles. Par-
ticles such as electrons on the other hand are wavy in
their collective behavior. In the macroscopic world, either Fig. 2.8: Results of a double-slit-
the particle properties or the wave properties dominate, so experiment performed by Dr. Tonomura
that we have developed two different concepts for them. showing the build-up of an interference
In reality they are two aspects of the same concept, which pattern of single electrons. Numbers of
is the realm of quantum theory. electrons are 10 (a), 200 (b), 6000 (c),
The observations of the double-slit experiment cannot 40000 (d), 140000 (e).[6, 5, 7]
be derived from any theory of classical waves or classical
particles independently. What we observe is a new phe-
nomenon. A new theory is required, and that theory is
quantum theory. Together with Einstein’s theory of rel-
ativity, the advent quantum theory changed radically our
view of the world in the early 20th century.
Chapter 3
Towards a theory
In the previous section, we learned that particle beams exhibit interfer-

ence patterns as they are known from waves. To be more precise, the
probability of finding a particle at a point in space exhibits an interfer-
ence pattern. Therefore, we need to construct a theory that on the one
hand is able to describe interference patterns on a small length scale and
on the other hand it must match classical mechanics on a macroscopic
length scale. One possibility is to consider particles as wave packets. If
the size of the wave packet and the wave length is small on macroscopic
length scales, we may be fooled in thinking that the particle is in fact a
point. On the other hand, the wavy nature may reveal itself on a length
scale comparable to the wave length.
In the following, we will try to find a theory of waves in the form of
differential equations, in which wave packets behave, on a macroscopic
length scale, like classical particles.
Fig. 3.1: Erwin
Schrödinger, 1887-1961.
3.1 Particles: classical mechanics revisited Photograph from 1933.
Austrian Physicist. Nobel
Before we investigate the behavior of wave packets, let us revisit the price 1933 in physics for
main features of classical mechanics. One requirement of our theory is the Schrödinger equation.
that the wave packets behave like classical particles, if the wave length
is small. Therefore, we first need to have a clear understanding of the dynamics of classical particles.
In classical mechanics, the motion of particles is predicted by Newton’s equations of motion
mẍ = F = −∂x V (x)
where each dot denotes a time derivative, and ∂x V is the derivative of the potential V (x) with respect
to the position x. Newton’s equations relates the acceleration ẍ of a particle to the force F acting
on the particle with the mass m as proportionality constant.
The equations of motion can be derived from another formalism, the principle of least action1 .
The principle of least action is particularly useful to obtain a consistent set of forces from, for example,
symmetry arguments.
We start from a so-called Lagrangian L(v , x), which depends on positions x and velocities v = ẋ.
For a particle in a potential, the Lagrangian has the form kinetic energy minus potential energy
1
L(v , x) = mv 2 − V (x)
2
1 “principle of least action” translates as “Wirkungsprinzip; Hamilton principle” into German
25
26 3 TOWARDS A THEORY
The principle of least action says that from all conceivable paths x(t) starting at time t1 from x1
and arriving at time t2 at their destination x2 , the particle will choose that path which minimizes the
action S.2 The action is a functional 3 of the path defined as follows:
ACTION
The action S is a functional of a path x(t). It is the integral of the Lagrangian L over a path
Z x2 ,t2
S[x(t)] = dt L(ẋ, x) (3.1)
x1 ,t1
The integration bounds indicate that the path begins at a specified space-time point (x1 , t1 ) and
ends at another one, namely (x2 , t2 ). When the path is varied within the framework of the principle
of least action, these boundary conditions remain unchanged.
x
x(t)+ δx(t)
δx(t)
(x1,t1)
x(t)
(x2,t2)
t
Fig. 3.2: Two paths, x(t) and x(t) + δx(t), connecting two space-time points, (x1 , t1 ) and (x2 , t2 ).
Each path has a different value for the action S[x(t)]. The action depends on the path as a whole.
If the action does not change to first order with respect to all possible small variations δx(t), the
action is stationary for this path, and the path is a physical trajectory from the initial to the final
points.
The minimum, or better a stationary point, of the action is obtained as follows: We consider a
path x(t) and a small variation δx(t). The variation δx(t) vanishes at the initial and the final time.
Now we test whether an arbitrary variation δx(t) can change the action to first order. If that is not
2 One should better say that the physical path makes the action stationary rather than minimize it. Historically this
more general formulation has been noticed only long after the principle of least action has been formulated, so that we
often speak about minimizing the action, even though this statement is too limited.
3 A functional maps a function onto a number. To show the analogy, a function maps a number onto a number. A
functional can be considered a function of a vector, where the vector index has become continuous.
3 TOWARDS A THEORY 27
the case, the action is stationary for the path x(t).
Z
S[x(t) + δx(t)] = dt L(ẋ + δ ẋ, x + δx)
Zh i
Taylor ∂L ∂L
= dt L(ẋ , x) + δ ẋ + δx + O(δx 2 )
∂v ∂x
Z h d ∂L d ∂L i
part. int. ∂L
= S[x(t)] + dt δx − δx + δx + O(δx 2 )
dt ∂v dt ∂v ∂x
h ∂L ix2 ,t2 Z h d dL dL i
= S[x(t)] + δx + dt − + δx(t) + O(δx 2 )
∂v x1 ,t1 dt dv dx
| {z }
=0
Z h d ∂L ∂L i
δS = S[x(t) + δx(t)] − S[x(t)] = dt − + δx + O(δx 2 )
dt ∂v ∂x
Thus, the functional derivative of the action is
δS d ∂L ∂L
=− + (3.2)
δx(t) dt ∂v ∂x
The notation O(δx 2 ) is a short hand notation for all terms that contains the second and higher
powers of δx.
The derivation contains the typical steps, which are worth to remember:
(1) Taylor expansion of the Lagrangian to first order in coordinates and velocities.
(2) Conversion of derivatives δ ẋ into δx using partial integration.
(3) Removing the complete derivatives using boundary conditions δx(t1 ) = δx(t2 ) = 0.
The requirement that dS = S[x + δx] − S[x] vanishes for any variation δx in first order of δx,
translates, with the help of Eq. 3.2, therefore into the
EULER-LAGRANGE EQUATIONS
d ∂L(v , x) ∂L(v , x)
= with v = ẋ (3.3)
dt ∂v ∂x
The Euler-Lagrange equations are identical to Newton’s equation of motion which is easily verified
for the special form L = 21 mv 2 − V (x) of the Lagrangian.
3.1.1 Hamilton formalism
The Hamilton formalism is identical to the principle of least action, but it is formulated in more
convenient terms. Noether’s Theorem4 says that every continuous symmetry results in a conserved
quantity. For a system that is translationally symmetric in space and time, the two conserved quan-
tities are momentum and energy, respectively. The Hamilton formalism is built around those two
quantities.
The Hamilton formalism is obtained by a Legendre transformation of the Lagrangian. We define
a new function, the Hamilton function
4 see ΦSX: Klassische Mechanik

HAMILTON FUNCTION
def
H(p, x) =pv − L(v , x, t), (3.4)
The Hamilton function depends on positions and momenta p, but not on on the velocities! As
described in the following, the velocities are considered a function v (p, x) of momenta and coordinates
themselves, i.e.
H(p, x) = p · v (p, x, t) − L(v (p, x, t), x, t),
The physical meaning of the Hamilton function is the total energy.
In order to get rid of the dependence on the velocity v in Eq. 3.4, we define the momentum p so
that the Hamilton function H does not explicitly depend on ẋ . This leads to
∂H ∂L
=p− =0
∂v ∂v
which defines the canonical momentum as follows:

CANONICAL MOMENTUM
def ∂L
p = (3.5)
∂v
The difference between the canonical momentum described above and the kinetic momentum de-
fined as p ′ = mv , that is as mass times velocity. The kinetic momentum is the canonical momentum
of a free particle.
From Eq. 3.5 we obtain the momentum p(x, v , t) as function of positions and velocities. This
expression can be converted into an expression for the velocity v (p, x, t) as function of momenta and
positions. The latter is used in Eq. 3.4 to obtain the Hamilton function.
In the following, we derive a new set of equations of motion by forming the derivatives of the
Hamilton function with respect to positions and momenta and by using the Euler Lagrange equations
Eq. 3.3. The resulting equations of motion are called Hamilton’s equations. Hamilton equations,
contain exactly the same information as Newton’s equation of motion and the Euler-Lagrange equa-
tions.

∂H Eq. 3.4 ∂L ∂L ∂v (p, x) Eq. 3.3 d ∂L Eq. 3.5
= − + p− = − = −ṗ
∂x ∂x ∂v ∂x dt ∂v
| {z }
= ∂H
∂v
=0

∂H Eq. 3.4 ∂L ∂v (p, x)
= v + p− = v = ẋ
∂p ∂v ∂p
| {z }
= ∂H
∂ ẋ
=0
We generalize the result to higher dimensions

HAMILTON’S EQUATIONS OF MOTION
∂H
ṗi = −
∂xi
∂H
ẋi = (3.6)
∂pi
We will come back to the Hamilton equations in quantum mechanics, when we investigate the dy-
namics of wave packets. Euler-Lagrange equations, Hamilton equations and Newton’s equations of
motion have the same content and are mathematically identical.
3.2 Waves: the classical linear chain

Let us now remind ourselves of some properties of classical waves. As a working example, I am
choosing the most simple wave I can think of, namely the vibrations of a linear chain. The linear
chain consists of particles, or beads, distributed along a line. The beads are free to move along the
direction of the line, but each bead is connected to the two neighboring beads by springs.
The linear chain is a simple model for lattice vibrations, where atoms oscillate about some equilib-
rium positions. If an atom in a crystal is displaced, the bonds to the neighboring atoms are stretched
on one side and compressed on the other. The resulting forces try to restore the original atomic
distances. If we let the beads loose, they will start to oscillate. These oscillations usually propagate
along the entire chain. If we consider the material on a larger length scale, and if we ignore that the
material is made of individual atoms, we can consider those lattice deformations as continuous waves
similar to water waves or light. These lattice-distortion waves are called phonons. Phonons are the
quanta of lattice vibrations just as photons are the quanta of light.
• Phonons are used to describe the transport of sound waves in solids.
• Phonons are responsible for heat transport in insulators. (In metals the electronic contribution
to the heat transport is usually a lot larger than the lattice contribution.)
3.2.1 Equations of motion ...

Let us now describe the linear chain mathematically:
The positions of the particles are at xj (t) = x̄j + φj (t). The equi- α
librium positions x̄j = j∆ are equally spaced with a distance ∆ between m
neighboring particles. The displacements from the equilibrium positions
are denoted by φj (t).
Let us write down the Lagrangian, from which we obtain the equa- ∆
tions of motion. The Lagrangian is kinetic energy minus potential en-
ergy. φj xj
Xh 1 1 i
L({φ̇j }, {φj }) = mφ̇2j − α(φj − φj −1 )2 (3.7)
2 2
j
α is the spring constant. Thus the energy of a spring is 21 α(d − d0 )2 , where d = xj − xj −1 is the
actual length of the spring and d0 = x̄j − x̄j −1 is the length of the spring without tension.
The action is defined as the integral of the Lagrange function.

Z {φ2,j },t2
S [{φj (t)}] = dt L({φ̇j (t)}, {φj (t)})
{φ1,j },t1
We use curly brackets to denote all elements in a vector, and brackets to denote functions. The
action is a functional of the time-dependent displacements [{φj (t)}]. According to the variational
principle, the physical displacements φj (t) are the extremum of the action under the condition that
the displacements at the initial time t1 and the final time t2 have fixed values.
We determine the equations of motion for the displacements φj (t) from the extremum condition,
that is the Euler-Lagrange equations Eq. 3.3.
d ∂L Eq. 3.3 ∂L
=
dt ∂ φ̇n ∂φn

Eq. 3.7 d ∂ 1 2 1 2
⇒ mφ̇n = − α(φn − φn−1 ) − α(φn+1 − φn )
dt ∂φn 2 2
⇒ mφ̈n = −α(φn − φn−1 ) + α(φn+1 − φn )
⇒ mφ̈n = α (φn+1 − 2φn + φn−1 ) (3.8)
The right hand side reminds of the differential quotient of the second derivative. 5
3.2.2 ... and their solutions

The equations of motion Eq. 3.8 are those of a multidimensional harmonic oscillator, since the forces
are linear in the displacements {φn }. The multidimensional harmonic oscillator can be solved with
the techniques described in ΦSX: Klassische Mechanik.
The problem Eq. 3.8 is even simpler than the multidimensional harmonic oscillator, because it
is translationally invariant. Translationally invariant problems are best solved with an exponential
Ansatz.
Z
dk
φj (t) = a(k)ei (k x̄j −ω(k)t) (3.9)
2π
We insert this Ansatz into the equations of motion Eq. 3.8 for the linear chain. This will provide
us with the dispersion relation of the linear chain. A dispersion relation relates the frequency ω(k)
to the wave vector k. Another form of the dispersion relation connects the energy of a particle to its
momentum. Later, we will see that there is an intimate connection between energy and frequency
on the one hand and the momentum and the wave vector on the other hand.
Eq. 3.8
mφ̈n = α (φn+1 − 2φn + φn−1 )
Z Z
Eq. 3.9 dk dk
⇒ (−mω 2 (k))a(k)ei (k x̄n −ωt) = α (ei k∆ − 2 + e−i k∆ ) a(k)ei (k x̄n −ωt)
2π 2π | {z }
h ∆ ∆ i2
eik 2 −eik 2 =(2i )2 sin2 (k ∆2 )
k∆
mω 2 (k) = 4α sin2 ( )
r 2
α k∆
ω(k) = ±2 sin( ) (3.10)
m 2
Eq. 3.10 is the dispersion relation of the linear chain. Its graph is shown in Fig. 3.3. Like the dispersion
relation of light, the one of the linear chain behaves almost linear for small wave vectors. In contrast
5 Differential quotient of the second derivative
d2f f (x + ∆) − 2f (x) + f (x − ∆)
= lim
dx 2 ∆→0 ∆2
p
to light, however,the linear chain has a maximum angular frequency6 of ωmax = 2 α/m. This is a
consequence of the discrete nature of the linear chain.
3
2
1
pα
ω/ m 0
−1
−2
−3
−1.5 −1 −0.5 0 0.5 1 1.5
∆
2k
Fig. 3.3: Dispersion relation ω(k) of the discrete linear chain. The straight (dotted) lines are the
dispersion relation of a continuous linear chain. The dispersion relation of the discrete chain is periodic
with period 2π 4π
∆ . (Each branch by itself has a period of ∆ .)
With the dispersion relation Eq. 3.10, we can write down the solution Eq. 3.9 for the linear chain
as
Z
dk
φj (t) = a+ (k)ei (k x̄j −|ω(k)|t) + a− (k)ei (k x̄j +|ω(k)|t)
2π
with
a+ (k) = a− (−k)∗
The last requirement ensures that the displacements are real. 7

Usually, we can work with each component independently of the other. The coefficients a± (k)
are determined from the initial conditions.
Z
dk
φj (t = 0) = (a+ (k) + a− (k))ei k x̄j
2π
Z
dk
∂t φj (t = 0) = − i |ω(k)|(a+ (k) − a− (k))ei k x̄j
2π
6 The angular frequency ω is defined as
2π
ω= = 2πf (3.11)
T
where T is the duration of a period and f = 1/T is the frequency.
7 Here we show that the requirement that a field has only real values implies the relation a(k) = a∗ (−k) of its
Fourier components: Let φj = φ(xj ), where xj = ∆j. The requirement that the displacements φ(x) are real has the
form
Z Z
φj = d2πk eikx a(k)
R
dk i kx dk −i kx ∗
φ(x) = φ∗ (x) ⇒ e a(k) = e a (k)
2π 2π
| {z }
d k ik∆j ∗
R
= 2π e a (−k)
Z
dk i kx
⇒ e a(k) − a∗ (−k) = 0 ⇒ a(k) = a∗ (−k)
2π
q.e.d.
3.3 Continuum limit: transition to a field theory

Sofar, the linear chain has been formulated by discrete beads. For our purposes, the discrete nature
of the chain is of no interest.8 Therefore, we take the continuum limit ∆ → 0 of the linear chain. In
this way make the transition to a field theory.
We start out with the Lagrangian Eq. 3.7 for the discrete linear chain, and modify it introducing
the equilibrium distance ∆ at the proper positions without changing the result.

Eq. 3.7 X h1 m 1 φj (t) − φj −1 (t) 2 i
L({∂t φj }, {φj }) = lim ∆ (∂t φj )2 (t) − |{z}
α∆
∆→0 2 |{z}
∆ 2 ∆
j →E | {z }
| {z } →̺
R →(∂x φ)2
→ dx
As ∆ approaches zero, we can convert the sum into an integral and the differential quotient will
approach the derivative. We also see that a meaningful limit is obtained only if the mass is scaled
linearly with ∆, that is m(∆) = ̺∆ and if the spring constant is scaled inversely with ∆, that is
α(∆) = ∆E . ̺ is the mass density of the linear chain and E is its linear elastic constant9. Thus
the potential-energy term in the Lagrangian can be identified with the strain energy. Finally, the
displacements φj (t) of the individual beads must be replaced by a displacement field φ(x, t) so that
φ(xj , t) = φj (t) for xj = j∆.
The limit ∆ → 0 then yields
Z
1 2 1 2
L[φ(x), ∂t φ(x)] = dx ̺ (∂t φ(x)) − E(∂x φ(x))
2 2
def
q
c= ̺ 1E Z
1
= E dx 2 (∂t φ(x))2 − (∂x φ(x))2 (3.12)
2 c
q
where we have introduced the speed of sound c = E̺ . The identification of c with the speed of
sound will follow later from the Euler-Lagrange equation and the corresponding dispersion relation.
The Euler-Lagrange equations for fields are only slightly more complicated compared to those for
particles. First we extract the Lagrangian density ℓ. The spatial integral of the Lagrangian density
yields the Lagrange function and the integral over space and time yields the action.
For the linear chain, the Lagrangian density is
Eh 1 i
ℓ(φ, ∂t φ, ∂x φ, x, t) = 2
(∂t φ)2 − (∂x φ)2 (3.13)
2 c
EULER-LAGRANGE EQUATIONS FOR FIELDS
The Euler-Lagrange equations for fields have the form (see ΦSX: Klassische Mechanik)
∂ℓ ∂ℓ ∂ℓ
∂t + ∂x − =0 (3.14)
∂(∂t φ) ∂(∂x φ) ∂φ
For the linear chain, we obtain the following equations of motion from Eq. 3.14
1 2
∂ φ(x, t) = ∂x2 φ(x, t)
c2 t
8 Our ultimate goal is to develop a field theory for particles in free space. Since space and time do not exhibit any
structure, we can as well consider an infinitely fine-grained linear chain.

9 The elastic constant (or Young modulus) E = σ/ǫ is the ratio between stress σ and strain ǫ. The strain is the
relative displacement per unit length. The stress is the force per area that is to be applied to a solid bar. Whereas the
elastic constant of a tree-dimensional object has the unit of pressure or “energy-divided-by-volume”, what we call linear
elastic constant has the unit “energy-divided-by-length”.
and the dispersion relation

ω = ±ck
If we compare this dispersion relation with that of the discrete chain in Fig. 3.3, we see that the
continuum limit is identical to the k → 0 limit of the discrete chain, which is the long wave length
limit.
3.4 Differential operators, wave packets, and dispersion rela-

tions
Let us now investigate the dynamics of a wave packet. In order to show the full generality of the
argument, we start from a general linear differential equation of a translationally invariant system in
space and time.
For a translationally invariant system, the corresponding time and space coordinates do not ex-
plicitly enter the differential operator, but only derivatives.10 Hence the differential equation of such
a system has the form11 :
Q(∂t , ∂x )φ(x, t) = 0
In the special case of the continuous linear chain discussed before, the function Q has the form
Q(a, b) = c12 a2 − b2 so that the differential operator has the form Q(∂t , ∂x ) = c12 ∂t2 − ∂x2 . 12,13
The Ansatz φ(x, t) = ei (kx−ωt) converts the differential equation of a translationally invariant
system into an algebraic equation, because
∂t ei (kx−ωt) = −i ωei (kx−ωt)
∂x ei (kx−ωt) = +i kei (kx−ωt) (3.15)
so that
0 = Q(∂t , ∂x )ei (kx−ωt) = Q(−i ω, i k)ei (kx−ωt)
Resolving the equation Q(−i ω, i k) = 0 for ω yields the dispersion relation ωσ (~k), which provides
the angular frequency ω as function of the wave vector k. Depending on the order14 of the differential
equation with respect to time, we may obtain one or several solutions ω for a given k. Thus the
dispersion relation may have several branches, which we denote by an index σ.
2
In the special case of the continuous linear chain Q(−i ω, i k) = − ωc 2 + k 2 , so that the dispersion
relation has two branches, namely ω1 = ck and ω2 = −ck.
The most general solution of the differential equation is then
X Z dk
φ(x, t) = ei (kx−ωσ (k)t) aσ (k) (3.16)
σ
2π
with some arbitrary coefficients aσ (k). Once the initial conditions are specified, which determine
aσ (k), we can determine the wave function for all future times.
10 Any dependence on x or t in the differential operator would imply that it is not invariant with respect to the
corresponding translations. The differential operator is called translationally invariant, if a shifted solution of the
differential equation is again a solution of the same differential operator.
11 Any differential equation of two variables x and t can be written in the form Q(x, t, ∂ , ∂ )φ(x, t) = 0, where
x t
Q(x, t, u, v ) is an arbitrary function of the four arguments. The differential operator is defined by the power series
expansion, where u is replaced by ∂x and v is replaced by ∂t
12 Here we use a strange concept, namely a function of derivatives. The rule is to write out the expression as if the
derivative were a number, but without interchanging any two symbols in a product, because ∂x x 6= x∂x . Note that
∂x x = 1 + x∂x , because in operator equations we should always consider that this operator is applied to a function, i.e.
∂x xf (x) = f (x) + x∂x f (x) = (1 + x∂x )f (x). Any differentiation occurring in a product acts on all symbol multiplied
to it on its right-hand side.
13 The operator def ~ 2 is called the d’Alambert operator. The Laplacian is defined as ∆ def
= c12 ∂t2 − ∇ = ∇~ 2.
14 The order of a differential equation is the highest appearing power of a derivative in a differential equation.
Wave packets
Let us now investigate the behavior of a wave packet. We select an envelope function χ(x, t = 0),
a wave vector k0 and one particular branch, i.e. σ, of the dispersion relation.
WAVE PACKET
A wave packet has the form
φ(x, t) = ei (k0 x−ωσ (k0 )t) χ(x, t) (3.17)
An example is shown in Fig. 3.4. We call χ(x, t) the envelope function. The envelope function
modulates a plane wave. k0 is the dominating wave vector. We usually assume that the envelope
function is smooth and extended over several wave lengths λ = 2π/|k0 | of the plane wave part.
We want to investigate the behavior of the dynamical evolution of an envelope function χ(x, t).
By equating
Z
Eq. 3.17 Eq. 3.16 dk i (kx−ωσ (k)t)
ei (k0 x−ωσ (k0 )t) χ(x, t) = φ(x, t) = e aσ (k)
2π
we obtain a expression of χ(x, t)

Z
dk i [k−k0 ]x−[ωσ (k)−ωσ (k0 )]t
χ(x, t) = e aσ (k) (3.18)
2π
This is an exciting result: For a system, that is translationally invariant in space and time, the
dispersion relation is sufficient to predict the complete future of a wave packet.
ψ ψ 2πδ( k−k0)
vph
ψ( x,t )
χ(k, t )
ψ( k,t )
λ
x
k0 k
ei(k0x−ω t) χ(x,t) vgr
Fig. 3.4: Wave packet in real and reciprocal space. In real space, a wave packet Ψ(x, t) is the
product between a plane wave ei (kx−ωt) and a smooth envelope function χ(x, t). The plane wave
moves with the phase velocity vph , while the envelope function moves with the group velocity vgr . In
reciprocal space, the plane wave part is a δ-function times 2π centered at the dominant wave vektor
k0 . Because the envelope functionR is smooth, it contributes only near k = 0. The wave packet
′
Ψ(k, t) is a convolution Ψ(k, t) = d2πk 2πδ(k − k0 − k ′ )χ(k ′ , t) = χ(k − k0 , t).
Sofar, everything has been exact. Now, we need to introduce an approximation that is only valid
for very smooth envelope functions χ(x, t) . We assume that the envelope function is smooth on the
length scale of the wave length λ = 2π/k. This implies that its Fourier transform aσ (k) has large
components only for k close to |k0 |.15 In this region, we can replace ω(k) by its Taylor expansion16
15 If
R dk R dk 1 k
f (x) = 2π f (k)ei kx then we find by variable substitution x → αx and k → k/α that f (αx) = 2π α
f ( α x)ei kx .
If α becomes smaller f (αx) becomes smoother in real space and more localized near k = 0 in k-space.
16 The notation O(x n ) is a shorthand notation for all terms containing powers of x of order n and higher.
dω
ω(k) ≈ ω(k0 ) + − k0 ) + O(k − k0 )2 to first order in k − k0 .
d k |k0 (k
Z
Eq. 3.18 dk i [k−k0 ](x− ∂ωσ t) Eq. 3.18 ∂ω
χ(x, t) ≈ e ∂k aσ (k) = χ(x − t, 0) (3.19)
2π ∂k
|{z}
vg
∂ω
Thus, the wave packet moves with the so-called group velocity vg = ∂k , without much of a change
of its shape.
Let us generalize the result to three dimensions.
GROUP VELOCITY
The wave packet as a whole, i.e. its envelope function, moves with velocity

Eq. 3.19
~ ~ ω(~k) ∂ω
vg = ∇
~ k ~ or vg,i = (3.20)
k0 ∂ki ~k0
It is not quite true that the wave packet does not change its shape. If we violate the assumption
that the wave packet is very extended and consider envelope functions that are less smooth, then
the Fourier transform is no more localized around k0 , and higher order terms of the Taylor expansion
of ωσ (~k) become important. These terms contribute to a delocalization, or spreading, of the wave
packet. This effect can be thought of in the following way: The wave packet consists of components
with different velocities, so that faster components run ahead and the slower components lag behind
the rest of the wave packet. As a result, the wave packet usually broadens with time. This broadening
of the wave packet is called dispersion.
In general, the group velocity is distinct from the phase velocity. The phase velocity is derived
from a single plane wave
vp
z}|{
ω
i k(x− t)
ei (kx−ωt) =e k (3.21)
The phase velocity is the velocity of a single wave crest or wave trough.
Again, we list here the three-dimensional expression for the phase velocity
PHASE VELOCITY
Eq. 3.21 ω(~k0 ) ω(~k0 ) ~k0

~
vp = ek0 =
~ (3.22)
|~k0 | |~k0 | |~k0 |
where ~ek = ~k/|~k| is the unit vector pointing along the wave vector ~k. The phase velocity is the
velocity of a wave crest. It can be read from the dispersion relation ω(k) as the slope of the line
connecting the origin with a particular point at the dispersion relation.
The dispersion relation of the continuous linear chain is purely linear, which implies that the
waves have a constant uniform velocity. This velocity is the velocity of sound, and it is analogous to
the speed of light. A medium with a linear dispersion relation is convenient for signal transmission
because, given a constant value for ω/k, a wave packet does maintain its shape during time even
if it is localized in space. Hence, the information encoded in the shape of the wave packet remains
intact during transmission.
3.5 Breaking translational symmetry

The wave packets discussed so-far move with a constant speed. This is a consequence of the
translational symmetry we have imposed. In order to explore, how we can teach wave packets to
ω
slope=vg
slope=vph
Fig. 3.5: Group and phase velocity in a dispersion relation ω(~k) for a classical particle. The group
velocity is the slope of the tangent of the dispersion relation, while the phase velocity is the slope of
the line connecting the origin of the graph with a given point on the dispersion relation.
behave like particles, we need to understand their behavior when they respond to external forces.
What happens, if the translational symmetry is slightly broken, such as in the presence of an
external potential? In that case, the Lagrangian and the equation of motion, i.e. the differential
operator Q(∂t , ∂x , x, t), contain terms that depend explicitly on the space coordinate.
To simplify the discussion, we assume that these terms are, on a length scale of the wave length,
only slowly varying. The dispersion relation ω(k) can be written down rigorously only in a translation-
ally symmetric situation. Without translational symmetry in space and time there is no one-to-one
correspondence between frequency and wave vector. However, if we allow only slow spatial variations,
we can still evaluate a spatially dependent ω(k, x) by fixing the spatially dependent terms at their
local values: at each point in space, we compare to a translationally invariant system with the same
local dispersion relation.
Let us define the position of the wave packet as17 .
R 2
def dx |φ(x, t)| x
x0 (t) = R (3.23)
dx |φ(x, t)|2
The mean k-vector is denoted by k0 . The velocity of the wave packet is the group velocity
Eq. 3.20 ∂
∂t x0 (t) = ω (x0 (t), k0 (t)) . (3.24)
∂k
Since we still have translational symmetry in the time variables, the value of the conjugate variable
ω(x0 (t), k0 (t)) remains constant18 , so that
d ∂ω ∂ω Eq. 3.20 ∂ω ∂ω
0= ω(x0 (t), k0 (t)) = ∂t x0 + ∂t k0 = + ∂t k0
dt ∂x |{z} ∂k ∂k ∂x
∂ω
∂k
∂
⇒ ∂t k0 = − ω(x0 (t), k0 (t)) (3.25)
∂x
17 Consider |φ(x)|2
P (x) = R dx|φ(x)2 |
as a density. Then the definition of the local position is the average position.
18 Ihave not really explained why ω is conserved with time. Let us assume a differential equation Q(∂t , ∂x , x)Ψ(x, t) =
0. We can then make an Ansatz Ψ(x, t) = e−i ωt φ(x, ω). Inserting the Ansatz into the differential equation yields
Q(−iω, ∂x , x) = 0. For
P each
R value of ω, we obtain one or more solutions φσ (x, ω). The general solution can therefore be
written as Ψ(x, t) = σ dω φ (x, ω)aσ (ω)ei ωt with some arbitrary coefficients aσ . Thus if we choose the coefficients
2π σ
so that only a small ω-region contributes, this region will be the same for all times. Hence, ω does not change in time.
600 1100
Fig. 3.6: The wave function of a linear chain. The wave packets obtained for subsequent times
have been displaced upwards with increasing time. The vertical displacement for each wave packet
corresponds to the longitudinal displacement of the chain-beads. The tic-marks on the x-axis denote
units of the chain. The functions have been obtained by numerical solution of the differential equation
for each chain bead.
Let us take a closer look at what we have got: the wave packet travels according to
Eq. 3.24 dω Eq. 3.25 dω

∂t x0 = ; ∂t k0 = − (3.26)
dk dx
This sounds familiar! Let us compare this result with Hamilton’s equation of motion, Eq. 3.6, for
classical particles
dH dH
∂t x = ; ∂t p = − (3.27)
dp dx
The two sets of equations, Eq. 3.26 and Eq. 3.27, are identical, if we perform the replacement
H(p, x) → ~ω(x, ~k)

p → ~k
where ~ is some arbitrary constant, which we will later identify with the reduced Planck constant.
(We have dropped here the subscripts of x and k.)
What we have done sofar, is to derive the equations of motion for wave packets from a specified
differential operator. We have seen that the dynamics of wave packets is identical to that of classical
particles, if the dispersion relations are related.
We are almost there!
CORRESPONDENCE PRINCIPLE
In order to construct wave packets that behave like classical particles, we start from a Lagrangian
L for the classical particle. From the Lagrangian we derive the Hamilton function H. The equation
E − H(p, x) = 0 is converted with the help of Eq. 3.15 into a differential equation by replacing
E → ~ω → i ~∂t
~
p → ~k → ∂x (3.28)
i
def
Ĥ = H( ~i ∂x , x) is called the Hamilton operator. The correspondence principle was invented by de
Broglie in his PhD thesis. It is also called de Broglie relation[1].
The correspondence principle is used to translate the classical equations of motion into the quantum
mechanical wave equations and vice versa.
The correspondence principle, when applied to the classical equation E = H(p, x), directly leads
to the
SCHRÖDINGER EQUATION [8, 9, 10, 11]
~
Q(∂t , ∂x , x, t)ψ(x, t) = [i ~∂t − H( ∂x , x, t)]ψ(x, t) = 0 (3.29)
i
I introduced a new symbol, namely ψ(x, t) instead of φ(x, t), in order to distinguish the displacement
field φ of a linear chain from the Schrödinger wave function ψ .19
We found a theory that is capable of describing the interference patterns, while reproducing
classical mechanics for wave packets on a macroscopic length scale. What we have done sofar is to
quantize the motion of a classical particle with a given Lagrangian. This is called first quantization.
The Schrödinger equation is for quantum mechanics what Newton’s law is for classical mechanics.
Most of this course will be devoted to the Schrödinger equation, how to understand it, and how to
solve it.
Without proof, I add here the form of the corresponding Lagrangian, that produces the Schrödinger
equation as the corresponding Euler-Lagrange equation.
Z h i
~
L[ψ, ∂t ψ, ∂x ψ, x, t] = dx ψ ∗ (x, t) i ~∂t − H( ∂x , x, t) ψ(x, t) (3.30)
i
It is a simple exercise to show that the action for a Hamilton function H = p 2 /(2m) + V (x) is real,
and, to show, via Eq. 3.14, that the Euler-Lagrange equation for Eq. 3.30 is indeed the Schrödinger
equation.
In most practical cases, the translation from a classical system in to the quantum description is
unique. However, there are exceptions. In classical physics we need not pay attention to the order in
which momenta and positions appear. When we replace the momentum and energy by its differential
operators, however, the product px results in a different differential equation than the product xp
because ∂x x 6= x∂x . There is no unique recipe that shows us how to arrive at the correct wave-
description from a classical theory alone. Experiments provided guidance to set up the Hamilton
function for a classical system. When the resulting Hamiltonian is not sufficient to determine the
quantum mechanical Hamilton operator, we need to resort to experiments again.
19 The field φ of the hanging linear chain will be analogous to that of a relativistic particle. In the non-relativistic
limit, the wave function φ does not directly turn into the Schrödinger field ψ(x, t). An additional transformation is
required to get rid of the energy related to the rest-mass.
3.6 Introducing mass into the linear chain (Home study)

We have ventured far from the linear chain, with which we began this chapter. The wave packets of
the linear chain do not seem to have too much to do with particles. The reason is that the dispersion
relation of the linear chain is similar to that of photons, the mass-less light-particles. However, it
does not seem to be able to describe particles with mass such as electrons.
Here, I will show that a small variation of the linear chain actually leads to to the relativistic
dispersion relation of particles that have a mass. While being a bit of a curiosity, the comparison
of a purely classical system that is analogous to a quantum mechanical wave description may be
very useful to imagine quantum mechanical principles. Furthermore, it may even guide the interested
reader into the fundamental principles underlying the Theory of Relativity.
1111111111111111111111111111111111111111111111111111111111111111111111111111
0000000000000000000000000000000000000000000000000000000000000000000000000000
0000000000000000000000000000000000000000000000000000000000000000000000000000
1111111111111111111111111111111111111111111111111111111111111111111111111111
0000000000000000000000000000000000000000000000000000000000000000000000000000
1111111111111111111111111111111111111111111111111111111111111111111111111111
0000000000000000000000000000000000000000000000000000000000000000000000000000
1111111111111111111111111111111111111111111111111111111111111111111111111111
0000000000000000000000000000000000000000000000000000000000000000000000000000
1111111111111111111111111111111111111111111111111111111111111111111111111111
0000000000000000000000000000000000000000000000000000000000000000000000000000
1111111111111111111111111111111111111111111111111111111111111111111111111111
00
11 1010 00
11 00
11 1010 1010 00
11 000
111 000
111
0000000000000000000000000000000000000000000000000000000000000000000000000000
1111111111111111111111111111111111111111111111111111111111111111111111111111
00
11 00
11 00
11 00
11 000
111 000
111
00
11 1010 00
11 00
11 1010 1010 00
11 000
111 000
111
00
11 00
11 00
11 00
11 000
111 000
111
00
11 0
1 00
11 00
11 0
1 0
1 00
11 000
111 000
111
00
11 0
1 00
11 00
11 0
1 0
1 00
11 000
111 000
111
00
11 0
1 00
11 00
11 0
1 0
1 00
11 000
111 000
111
00
11 0
1 00
11 00
11 0
1 0
1 00
11 000
111 000
111
00
11 0
1 00
11 00
11 0
1 0
1 00
11 000
111 000
111
00
11 0
1 00
11 00
11 0
1 0
1 00
11 000
111 000
111
00
11 10 00
11 00
11 10 10 00
11 000
111 000
111
Let us again have a look at the continuous linear chain. In Eq. 3.12 (p. 32) we obtained the
Lagrangian, which had the form
Z h1 i
E
L[φ(x), φ̇(x)] = dx 2 |∂t φ|2 − |∂x φ|2
2 c
where E is the linear elastic constant. There is a simple generalization of that Lagrangian. The
properties we want to preserve are
• spatial and temporal translational symmetry and the so-called
• gauge symmetry20 21 . Gauge symmetry means that the physics depends only on the absolute
value of φ and not its phase. Furthermore, we require that the Hamiltonian has
• only terms up to quadratic order in φ in order to maintain its simple form.
There is exactly one additional term, which we can include, namely a constant multiplied with the
m2 c 2
absolute square of φ. We choose this constant equal to ~02 , where m0 is a free parameter. We
could have chosen any other prefactor. The explicit form chosen will be motivated later, when we
compare the dispersion relation for the hanging linear chain with that of a relativistic particle.
Z h1 m c 2 i
E 0
L[φ(x), φ̇(x), ∂x φ(x)] = dx 2 |∂t φ|2 − |∂x φ|2 − |φ|2 (3.31)
2 c ~
The Euler-Lagrange equation Eq. 3.14 for this Lagrange function is the Klein-Gordon equation
KLEIN-GORDON EQUATION
h1 m0 c 2 i
2 2
∂ t − ∂ x + ( ) φ(x, t) = 0 (3.32)
c2 ~
20 “gauge symmetry” translates as “Eichsymmetrie” into German

21 Gauge is pronounced “geidsch”.
If we apply the correspondence principle Eq. 3.28 backwards to the Klein-Gordon equation, we
obtain the dispersion relation of a relativistic particle with rest mass m0 .
DISPERSION RELATION OF RELATIVISTIC PARTICLES
E 2 = p 2 c 2 + m02 c 4 ,
Now we can identify the constants in the Klein-Gordon equation.
• c is the speed of light. This is the maximum speed c = vmax = max ∂E

∂p = c
• m0 is the rest mass of the particle. For non-relativistic velocities the dispersion relation is
p2
E = m0 c 2 + 2m 0
. Thus, in this limit, m0 is the known inertia mass. m0 c 2 is the energy related
to the mere existence of the particle. If a particle collides with its antiparticle, this energy will
be turned into radiation energy.
What is the relation between a wave-packet of the Klein-Gordon equation and a relativistic par-
ticle? A wave packet of the Klein-Gordon equation cannot move faster than the speed of light c 22 ,
which is the basis for the theory of relativity. Historically, the theory of relativity indeed has its roots
in a wave theory, namely electrodynamics: Lorentz observed that the Maxwell’s equations had certain
symmetries, that were incompatible with the concept of an infinite speed of light. It took Einstein to
realize the radical implications of this symmetry and to propose that the same symmetries are valid
for the dynamics of particles.
E
nonrelativistic
dispersion relation
E=m 0c 2+p2/(2m0)
rest energy
particles m 0c 2
p
cp
lig
E=
ht
:E
:
ht
=−
lig
pc
anti particles
Fig. 3.7: Relativistic dispersion relation. The relativistic dispersion relation approaches E = cp
for large momenta, indicating that a particle cannot become faster than the speed of light. The
nonrelativistic dispersion relation is obtained as Taylor expansion of E(p) at p = 0. The relativistic
dispersion relation predicts the existence of anti particles. Even at rest, a particle has an energy the
rest energy. If for example, an electron and a positron, its anti particle, collide, they annihilate each
other. The rest energy is emitted in the form of two photons leaving in opposite directions.
We obtained the linear chain as a purely classical model of a chain of beads connected by springs.
The mechanical analogue of the Klein-Gordon equation is the hanging linear chain: A chain of
pendulums connected by springs. The first term in the Klein-Gordon equation Eq. 3.32 is the inertia
q 1
22 The velocity is v = ∂E/∂p = pc 2 / p2 c 2 + m02 c 4 = c[1 + ( mp0 c )2 ]− 2 → c for p → ∞
force of the individual beads. The second term is the force due to the springs between neighboring
beads. The last term can be described by a force between the displaced bead and its equilibrium
position.
Thus we have found a purely classical model, that describes a quantized relativistic particle. There
is only one caveat: the displacements of the linear chain must be real, whereas the wave functions
belonging to one branch of the dispersion relation are always complex.
3.7 Measurements
The purpose of any theory is to make predictions for measurable quantities. A measurable quantity
is called an Observable. Examples for observables are energy, momentum, position etc. However
~ r , t) of electrodynamics is
there are also quantities that are not observable: The vector potential A(~
such an example: A gauge transformation yields different vector potentials, which describe exactly
the same physics. Thus the vector potential cannot be measured. Nevertheless, the magnetic field,
which is directly related to the vector potential, can be measured.
In classical mechanics, the state of a system is described by the positions and momenta of all
particles. Thus all observables can be expressed as functions A(p, x) of positions and momenta.
An example is the energy, which is expressed by the Hamilton function H(p, x) , or the angular
momentum defined as L(~ ~ r , p~) = ~r × p~.
Quantum mechanical wave packets, however, are spread over some volume. Therefore, they do
not have a precise position nor a precise momentum. Thus, we need to perform an average in order
to attribute a value to an observable for a given wave packet.
We relate the square of the wave function to an intensity, just as in optics. On the other hand,
the intensity has been related to the density of particles or the probability density to find a particle
at a given point in time and space.
PHYSICAL MEANING OF THE WAVE FUNCTION
The absolute square φ∗ (x)φ(x) of the wave function is interpreted as probability density for the
particle to be at a given position.
As a consequence of this probabilistic interpretation of the wave function, we need to require

that a intensity of a wave function integrates to unity. We call this condition the normalization
condition.
NORMALIZATION CONDITION
Z
dx φ∗ (x, t)φ(x, t) = 1 (3.33)
The normalization condition reflects that the probability for the system being in any state is equal to
one.
Most of the previous discussion has been based on wave packets of the form φ(x, t) = χ(x, t)ei (kx−ωt) ,
which are very localized on a macroscopic length scale, but delocalized on a length scale set by the
wave length. Intuitively we identify the position hxi of a wave packet with the center of the intensity.
Z Z
∗
hxi = dx φ (x)xφ(x) = dx |χ(x)|2 x (3.34)
Thus we have defined what we mean with the position of a wave packet.
Now let us turn to the momentum. The momentum of an extended wave packet should be related
to the wave vector k0 of the plane wave part.
The correspondence principle identified ~k with the momentum of a classical particle. We can
obtain the momentum of a wave packet as
Z Z
Eq. 3.28 ~ Eq. 3.17
hpi = dx φ (x) ∂x φ(x) = ~k − i ~ dx χ∗ (x)∂x χ(x)
∗
i
Z Z
i~
= ~k − dx χ∗ (x)∂x χ(x) + dx χ(x)∂x χ∗ (x)
2
| {z }
(A)
Z Z
i~
− dx χ∗ (x)∂x χ(x) − dx χ(x)∂x χ∗ (x)
2
| {z }
(A)
Z Z
i~ i~ ∗
= ~k − dx ∂x χ∗ (x)χ(x) − m dx χ (x)∂x χ(x) − χ(x)∂x χ∗ (x)
2 |2m {z }
Jχ (x)
∞ Z
i~
= ~k − χ∗ (x)χ(x) +~ dx Im[χ∗ (x)∂x χ(x)]
2 −∞
| {z }
=0
Im(χ)=0
= ~k
The assumptions are that the envelope function is purely real, i.e. Im(χ) = 0, and that it vanishes
at infinity.23
Thus we have defined what we mean with the momentum of a wave packet. We have used here
the Symbol hAi to denote the quantum mechanical mean value of an observable A consistent with
the expression Eq. 3.34 for the positions. Furthermore we required that the wave packet be localized,
so that its amplitude vanishes at infinity.
Now, we need to generalize these results to more general observables A. In classical mechanics,
an observable can always be expressed by a function A(x, p) of the positions and momenta. This
is because positions and momenta characterize a classical state uniquely, two generalizations are
possible:
• we could attribute the observable to the function
hAi = A(hpi, hxi) (3.35)
of the mean values for positions and momenta obtained above.
• we could attribute the observable to the differential operator A( ~i ∂x , x) and evaluate the ex-
pectation value as
Z
~
hAi = dx φ∗ (x)A( ∂x , x)φ(x) (3.36)
i
similar to the expressions for the positions and momenta.
Only the second choice, Eq. 3.36, is correct. Using the former choice, Eq. 3.35, a variable transform
would result in different expressions for the expectation value.
Another indication in favor of the second choice Eq. 3.36 is the following: The correspondence
principle, Eq. 3.28 on p. 38, provides us, together with the Schrödinger equation Eq. 3.29, with a
prescription for the energy in the form:
Z Z
Eqs. 3.28,3.36 ∗ Eq. 3.29 ~
hEi = dx φ (x)i ~∂t φ(x) = dx φ∗ (x)H( ∂x , x)φ(x)
i
23 The expression J resembles the expression for the flux density. that is derived later in Section 8.4 on p. 131. As
χ
it is applied only to the envelope function it has a different meaning in this case, and with the assumption of a real
envelope function the expression for Jχ vanishes altogether.
Since the energy H(p, q) is a perfectly allowed measurable quantity, our generalization must be
consistent with the expression given above. The alternative rule using hAi = A(hpi, hxi) would give
an inconsistent result.
Therefore, we set up the general rule for the evaluation of observables: start from a classical
observable A(p, x, t) and replace momenta by its differential operator according to the correspondence
principle.
EXPECTATION VALUE OF A GENERAL OBSERVABLE
The expectation value of an observable, which is represented by a function of momenta and position
A(p, x), is obtained for a quantum mechanical system from its wave function as
R
dx φ∗ (x)A( ~i ∂x , x, t)φ(x)
hAi = R (3.37)
dx φ∗ (x)φ(x)
We have used here a form that does not require explicit normalization.
3.8 Postulates of Quantum mechanics

In this section, we have gone through the process of constructing a theory, that describes new
experimental facts from the double-slit experiment. We can now summarize our choices in a set of
postulates.[7]
1. State: The state of a quantum mechanical system with one particle is fully described by
the wave function Ψ(~r, t), that depends on the position of the particle and the time. For
systems with several particles, the wave function depends on the position s of all particles, i.e.
Ψ(~r1 , . . . , ~rN , t).
2. Dynamics: The time evolution of a wave function is determined by the Schrödinger equation
Eq. 3.29.
3. Observable: Observables are represented by linear differential operators acting on the wave
function. We will see later that these differential operators need to have the property of being
hermitian, a term that will be described later.
4. Measurement: For a given state of the system only the probabilities for the outcome of an
experiment can be predicted. The mean value for the of the predicted outcomes is the ex-
pectation value defined by Eq. 3.37. In particular, the expectation value of the observable A
that is defined by A(~r, p~) = δ(~r − ~r0 ) provides the probability density for finding the particle at
position ~r0 . This probability density is the square of the wave function at position ~r0 , that is
ψ ∗ (~r, t)ψ(~r, t).
This set of postulates is not formulated in its most general form. This set is also not complete and
a few postulates for the measurement process will be added later.
The connection with classical mechanics is given by the following two requirements.
1. if the classical dynamics is governed by Hamilton equations with a Hamiltonian H(p, q), so that
E = H(~
p, ~
x)
the wave function evolves according the differential equation
~~
i ~∂t Ψ(~r, t) = H( ∇, ~r)Ψ(~r, t)
i
The connection is provided by the correspondence principle Eq. 3.28.
2. If a classical observable is represented by a function A(~ x ), then the quantum mechanical

p, ~
operator representing the observable is obtained by the correspondence principle.
3. The transformation of a function into an operator via the correspondence principle is not always
unique. In those cases one has to choose that operator, which is hermitian.
3.9 Is Quantum theory a complete theory?

There is a fundamental question related to measurements, that is heavily discussed until today. Let us
go back to the diffraction pattern of photons. A theory based on waves, namely Maxwell’s equations
can describe the intensity of the electromagnetic field. However the experiment also shows that when
the intensity is turned down, we observe a discrete signal here or there, but no more a continuous
signal.
There are two conclusions following out of this observation.
• The events come in chunks.
• The intensity corresponds to the probability of an event.
The second observation is the reason for calling the wave function the probability amplitude:
φ∗ (x)φ(x) is the probability to find the particle at position x.
The probabilistic nature of quantum mechanical predictions appears unsatisfactory, and seems to
indicate that the theory is incomplete. The sentence “Gott würfelt nicht” (“God does not play dice”)
was Einstein’s comment to quantum theory, but all attempts to find an inconsistency in the theory,
or any contradiction between the predictions of the theory and experimental observation failed.
It is very difficult to imagine that any deterministic theory is capable of describing the outcome
of the double-slit experiment with an individual particle. The experimental setup we described for the
double-slit experiment has been completely symmetric with respect to a mirror reflection about the
y = 0 line connecting the source with the midpoint of the two slits. The outcome of the individual
signals, however, is not symmetric. A single particle can appear anywhere, as long as the probability
is finite, which in general is off-center and therefore not symmetric, unlike the experimental setup.
Hence we cannot use a symmetric theory to predict the outcome for a single particle. We need a
mechanism to break the symmetry. But where do we take the information how the symmetry is
broken? Until we find some uncontrolled environmental effects not considered in the experimental
setup as we described it, we need to be content with the probabilistic theory.
Do we have to accept a theory that is less complete and predictive than classical mechanics? Not
really: think of chaos. If we would know the exact values of positions and velocities of a particle and
the equations of motion, classical mechanics can, in principle, predict the outcome for all future. But
is the exact specification of the initial conditions meaningful? Try to devise an experiment so that
the trajectory is prepared in a unique way far into the future. It does not suffice to determine the
initial conditions very, very accurate. Accurate is not what we mean with exact. After some time,
trajectories in a chaotic system with even slightly different initial conditions deviate completely from
each other. Hence, also in classical mechanics there are experiments, whose outcome we cannot
predict with arbitrary certainty.
On the other hand, there are experiments for which quantum mechanics provides exact predictions.
Namely those for which the probability distribution shrinks to a point.
When we compare quantum mechanics and classical mechanics closely, the difference in the
completeness appears more gradual than qualitative.
3.10 Planck constant

We still need to determine the value of one undetermined constant, namely the Planck constant
def h
h. In practice, the reduced Planck constant ~ = 2π is used24 and often both are named in the
same way, namely as Planck constant. Sofar the Planck constant played the role of an arbitrary
proportionality constant in the correspondence principle E = ~ω, p = ~k.
Only experiment can tell us the value of ~. The best value of ~ is currently[12, 13]
m2
~ = 1.05457168 × 10−34 kg (3.38)
s
What is the physical meaning of ~?
~ = ~r × p~ or of a phase-space-volume element
• The unit of ~ is that of an angular momentum L
dp dx.
• Planck constant determines the length scale on which the wavy nature becomes evident, namely
the wave length λ. Given a classical particle with velocity v , the wave length is λ = 2π 2π~
k = p =
2π~
mv . We recognize that the wavelength vanishes if Planck constant would be zero. In this case,
we would return to classical physics, where the oscillations are so rapid that they cannot be
observed.
• The limit of fast particles is similar to the classical limit ~ → 0. Indeed, quantum mechanical
effects seem to become unimportant at high temperatures.
• Let us consider a few real world examples:
– What is the wave length of a walking person? Let us make an estimate: the person should
weigh 72 kg and walk with a speed of 3.6 km/h=1m/sec. Its wave length is therefore
λ ≈ 10−35 m. This is the 1025 th fraction of an interatomic distance. Thus we can safely
describe a person by classical physics.
– The situation is different for an electron. An electron has a mass of about 10−30 kg. The
wave length of an electron at the same speed is already three millimeter or about 107
atomic distances.
3.11 Wavy waves are chunky waves: second quantization

We have mentioned that we have an intensity which we identify with a particle number. However,
sofar we have no way to explain why the intensity comes in chunks, as we expect it from particles.
Here I will give a brief outlook on how the chunks are explained. I cannot give here the theory in its
full glory, because that fills another lecture series on its own. Rather I want to sketch the main lines
of thought.
Our starting point was the linear chain. We built it up from particles connected by springs. A
modification of the linear chain revealed, that the collective motion of the beads produces wave
packets, that behave like classical particles. Hence we identified particles with wave packets. To
be consistent, we have to describe also every bead of the chain by a wave packet. What are the
consequences?
Let us consider a violin chord as a realization of the continuous linear chain.25 The violin chord
24 The Planck constant is defined as h. This has been a somewhat unfortunate choice, because h appears nearly
always in combination with 1/(2π). Therefore, the new symbol ~ = h/(2π), denoted reduced Planck constant has
been introduced. The value of ~ is ≈ 10−34 Js to within 6%.
25 The linear chain is not exactly the correct model for a violin chord. The transversal oscillation of the violin chord
differs from the longitudinal oscillation of the linear chain. However, for small displacements the transversal oscillation
behaves mathematically similar to a longitudinal oscillation. A better realization of the linear chain would be a rubber
chord stretched along its direction.
produces waves that are enclosed by its end points. A violin chord oscillate with frequencies that are
multiples of the pitch. It does not oscillate with any other frequencies.
What the violin chord is for a massless particle, the harmonic oscillator is for a massive particle.
Remember, that we identified the linear chain with a massless particle. The wave function of a
particle with mass in a harmonic potential oscillates with frequencies that are multiples of the classical
eigenfrequency (plus a constant). The frequency is related to the energy. Thus the energy of such
a system can change in discrete steps.
Let us consider a chain where we freeze all beads except one. This bead can be described by a
harmonic oscillator. Its energy is discretized, like the frequency of a violin chord. Thus we have a
special wave, which comes in chunks. Each frequency corresponds to a given number of particles.
Since the energy of a wave is always quadratic in the wave function amplitudes, it can always be related
to harmonic oscillators and all harmonic oscillators have discrete and equidistant energy levels. Hence,
when the wave function amplitude (or the position of the beads) is described by wave packets, all
field theories of the type discussed so far result in chunky waves.
Thus we have an additional dimension in addition to space and time, namely the field of a given
particle. This system is described by a wave function φ(x, t). If we quantize this world, we obtain
chunky fields and particle waves.
3.12 Summary
In this section we introduced the Schrödinger equation. In three dimensions the Schrödinger equation
has the form:
~~
i ~∂t ψ(~r, t) = H( ∇, ~r)ψ(~r, t) (3.39)
i
It has been obtained from the classical energy expression E = H(~ p, ~r) which is transformed into a
differential equation by using the correspondence principle, E → i ~∂t and p~ → ~i ∇, ~ and multiplication
by the wave function.
• ψ(~r, t) is the wave function. ψ is a complex function and the absolute square value P (~r, t) =
ψ ∗ (~r, t)ψ(~r, t) is the probability density to find the particle at point ~r in space at time t.
~ ~r) is obtained from the Hamilton H(~
• the Hamilton operator H( ~i ∇, p, ~r) function by replacing
the momentum p~ by the momentum operator
~~
p~ˆ = ∇ (3.40)
i
• ~ = 1.054 × 10−34 Js is Planck’s constant (germ: Planck’sches Wirkungsquantum). Note

that h = 2π~ is called by the same name. If Planck’s constant were zero, we would obtain our
traditional classical point mechanics.
Instead of describing a particle as a point with defined position and a momentum, in quantum
mechanics we describe particles by a wave packet with a intensity distribution φ∗ (x)φ(x) and a
momentum distribution.
The wavy nature of the wave packets becomes evident only at small length scales, so that it
evaded observation for a long time. One exception is again electromagnetism. There the wave
length can be very large. Here a field theory has been used a long time before quantum mechanics
has been invented.
We have also sketched how the quantization of the wave function itself naturally leads to the
concepts of chunks of the wave, which are identified as individual particles. Thus we changed our
perspective radically, from point-like particles to “waves of waves”. The macroscopic predictions of
quantum theory are, however, still in accord with the particle picture classical mechanics.
Chapter 4
Quantum effects
In this section, we will explore the physical effects that result from
the Schrödinger equation:
QUANTUM EFFECTS
• Spreading of wave packets,

• Splitting of wave packets,
• Tunnel effect,
• Quantized energy levels,
• Zero-point energy, and
• Resonances.
One of the main tasks of this section is to become familiar with

the similarities and differences of quantum-mechanical particle waves Fig. 4.1: Max Planck, 1858-
on one side and either classical particles or classical waves on the 1947. German Physicist. No-
other side. The comparison will lead us to conceptual difficulties that bel price 1918 for the discov-
we have to become familiar with. The reason we try to compare to ery of energy quanta. Photo
classical problems is that this allows us to use the experience that aus dem Jahr 1901
we have already gained. However, it is important to also gain a
thorough understanding of the conceptual problems.
In order to illustrate the quantum effects, I am choosing a class of very simple problems that
can be solved analytically, namely systems where the potential varies only in one dimension and
is piecewise constant. Thus the particles behave like free particles except where the value of the
potential changes abruptly to another value. The problems are
• the free particle,
• the potential step,
• the particle in a box,
• the potential barrier, and
• the potential well.
47
48 4 QUANTUM EFFECTS
4.1 Quasi-one-dimensional problems in the real world

Even though the examples are chosen to introduce quantum effects, the one-dimensional problems
discussed in the following play an important role in practical problems. All these problems can be
realized and have been used in electronic devices.
Electronic devices are build by bringing different materials into contact with each other, in order to
control the flow of electrons through these devices. Three basic types of materials play a role: metals,
which conduct electrons freely, insulators which do not conduct electrons, and semiconductors which
can be made conducting or insulating depending on the electrostatic fields, temperature and doping.
Each material can be considered to be homogeneous and each ma-
terial has its own dispersion relation for the electrons. This dispersion E
relation is called band structure1 . Every state can be occupied by a
EF
maximum of two electrons. The bands, the branches in the dispersion
relation, are filled with electrons up to an energy, the so-called Fermi
energy EF . In Fig. 4.2 we divide materials into two classes: If a band
crosses the Fermi energy the material is a metal. Otherwise it is an p
insulator or a semiconductor. In this case, a forbidden energy region,
E
where electrons cannot travel, separate the filled states, called the va-
lence bands2, from the empty states, called the conduction bands.
Ec
The forbidden energy region is called the band gap. EF
Eg
If there is no electric field, there are as many electrons moving right
Ev
as there are electrons moving left. As a consequence there is no net
electric current. An electrical current flows only when the balance of
electrons moving into opposite directions is broken, which can only hap- p
pen when electrons are excited above the Fermi energy. In a metal there
is no minimum energy to excite electrons. Therefore, metals conduct Fig. 4.2: Schematic band
electrons at all temperatures. In an insulator, the band gap is so large structure of a metal (top)
that thermal energies are insufficient to lift electrons from the valence and an insulator (bottom).
band into the conduction band. Therefore, hardly any current flows in EF is the Fermi level. Eg is
an insulator. In a semiconductor, the band gap is sufficiently small, so the band gap of the insula-
that the conductance can be easily modulated by temperature, dopant tor. Ev is the top of the
atoms and external fields. valence band and Ec is the
We can learn a lot about electrons in materials from the free particle bottom of the conduction
model, which underlies much of the discussion in this chapter. In a band.
semiconductor, only the lowest part of the conduction band and the
uppermost part of the valence band matter for electron transport, because thermal energies are small
and cannot lift many electrons much higher. Therefore, we can perform a Taylor expansion of the
conduction band about the minimum up to second order. This gives a parabola shaped dispersion
relation, which can be written as
~2~k 2
E(k) = Ec +
2mc∗
Ec is the conduction band minimum and mc∗ is called the effective mass of the conduction electron.
Similarly we can describe the hole by a dispersion relation
~2~k 2
E(k) = Ev −
2mv∗
where Ev is the valence band top and mv∗ is the effective mass of the holes. Holes behave as particles
with a positive charge, similar to positrons.
1 The concept of band structures is more general. There are band structures for phonons, that are the quanta
of lattice vibrations. There are even band structures for photons in so-called photonic crystals, a recent and highly
interesting research area.
2 Usually one distinguishes the valence states from so-called core states, which are also occupied. Core states are
located so close to the nucleus, that they do not interact with neighboring atoms.
4 QUANTUM EFFECTS 49
We have seen that particles which interact with a material behave like free particles but with
changed properties such as the mass or the charge. Such particles are called quasiparticles. The
situation is similar to that of light entering a material. Inside the material the speed of light is lower
than in vacuum, which explains the refractive index of materials. Also here we consider the photon
and the response of the material to the electric fields of the photon as one quasi particle.
We have done some simplifications: In reality (1) the effective mass can be a 3 × 3 tensor. (2)
The minimum need not to be located at k = 0. (3) There may be several minima and several nearby
branches of the bands at the conduction and valence band edges.
In a metal, the band structure can be described by a parabola with a minimum below the Fermi
energy. Only electrons near the Fermi energy contribute to the current. Close to the Fermi energy,
the dispersion relation is approximately linear with the wave vector ~k, so that the electrons have a
constant velocity similar to light and sound waves.
Here I will describe the realization of the model systems discussed in this section in electronic
devices:
• The free particle describes for example the transport in a bulk material
• The potential step is a model for an interface between two materials with different band edges.
• In the particle in the box a particle is confined in all three spatial directions as in a quantum
dot. A quantum dot could be realized by surrounding a semiconductor by an insulator, so
that any conduction electrons cannot escape. Quantum dots are used for example for so-called
single electron transistors. If the particle is confined only in one direction, the particle in the box
describes a two-dimensional electron gas, which was important to obtain the so-called quantum
Hall effect, for which v. Klitzing was awarded the Nobel prize.
• The finite barrier is realized by an insulating layer between two metals or semiconductors.
The finite barrier is a model of a gate oxide in a MOSFET3 transistor to describe its tunnel
current. Also the scanning tunneling microscope (Nobel prize H. Rohrer and G. Binnig) uses
the tunneling current from a surface to a tip to monitor surfaces with atomic resolution.
• A particle in a well describes electrons in a structure that can be made by depositing two
different semiconductors A and B in the sequence ABA. Since the conduction band edge of
the semiconductor B is higher or lower than that of A, the conduction band edge forms either
a finite barrier or a finite well. The transport through a finite well is relevant for the transport
through a “resonant tunnel diode”
3 MOSFET=Metal-oxide-semiconductor field-effect transistor
• Dirac atom (not yet included) The Dirac atom is the most simple one-dimensional model for
an atom. It is a model for a two-dimensional electron gas or a one-dimensional conductor.
Almost any of these quasi-one-dimensional problems can be realized by growing thin layers of
semiconductors onto other semiconductors. Such structures can be very complex and are called
multilayer structures.
4.2 Time-independent Schrödinger equation

Before we enter the discussion of quantum effects, we need to become familiar with the time-
independent Schrödinger equation. If the Hamiltonian is time independent, the Schrödinger equation
can be substantially simplified. Time-independent problems are so abundant, so that most of our
discussion will address those problems.
For time-independent problems, the wave function can be broken up into a product of a time-
dependent part, and a time independent part, which depends only on the atomic position. The
time-dependent part is trivial. The time-independent part of the wave function obeys the time-
independent Schrödinger equation.
Let us Fourier transform the Schrödinger equation with respect to the time.4 The time-dependent
wave function ψ(r, t) is related to its Fourier-transformed counterpart ψ(r, ~ω) by5
Z
dω −i ωt
ψ(~r, t) = e ψ(~r, ~ω) (4.1)
2π
Inserting the Fourier transform into the Schrödinger equation Eq. 3.29 yields6

~~
i ~∂t − H( ∇, ~r) Ψ(r, t) = 0
i
Z
Eq. 4.1 ~ ~ ~r) dω
⇒ i ~∂t − H( ∇, e −i ωt ψ(~r, ~ω) = 0
i 2π
Z
dω ~~
⇒ i ~∂t − H( ∇, ~r) e −i ωt ψ(~r, ~ω) = 0
2π i
Z
dω −i ωt ~~
⇒ e ~ω − H( ∇, ~r) ψ(~r, ~ω) = 0
2π i
| {z }
equation holds only if this vanishes
4 Note here that we denote the Fourier transform with the same symbol as the original functions. Mathematically
this is incorrect! However, this notation is very convenient when used cautiously. When we evaluate the function value
we need not only the numeric value of the argument, but also the type. So if we evaluate the function for a given
time, we obtain a different result than if we evaluate the function for a frequency with the same numeric value. We
will later see how such a notation can be formulated rigorously in the bracket formalism of Dirac.
5 The reader may be wonder why we use e−i ωt instead of ei ωt , which would be our regular notation of a Fourier
transform. The origin of te minus sign is that we derived the Schrödinger equation from plane waves ei (kx−ωt) . The
sign convention for ω is a matter of choice, but if we use this form of the Schrödinger equation, we need to adhere to
this choice. P
6 Plane waves ei ωt are linear independent. Linear independency implies that no superposition i ωi t of different
i ci e
(ωi 6= ωj ) plane waves can result in a vanishing function. Thus the expression
Z
dω
c(ω)ei ωt = 0
2π
can hold for all t only, when c(ω) = 0. Because plane waves are linear independent, the Fourier transform is unique.
We obtain
~~
H( ∇, ~r)ψ(~r, ~ω) = ψ(~r, ~ω)~ω
i
after realizing that the equation must be valid for every Fourier component.
By replacing ~ω with E, we obtain the time-independent Schrödinger equation.
TIME-INDEPENDENT SCHRÖDINGER EQUATION
~~
H(~r, ∇) ψ(~r, E) = ψ(~r, E) E (4.2)
i
We call such an equation an eigenvalue equation. The equation has solutions only for specific
values of E. These values are the eigenvalues of the Hamiltonian. The set of all eigenvalues is
called the spectrum. The solution ψ(~r, E) is called an eigenstate of the Hamiltonian.
An eigenvalue equation is similar to the problem of diagonalizing a matrix, A c~n = ~ cn λn . The
only difference is that the differential operator is replaced by a matrix and the vector is replaced
by a function in space. We will see later that this difference is smaller than one might think. The
difference is related only to the fact that in one case we describe finite matrices and vectors and in
the other case infinite vectors and matrices. We will return to this aspect in detail in chapter 5.
There may be several solutions to a given energy. We call such solutions degenerate. The
time-independent Schrödinger equation may have solutions for discrete values for the energy E, or
for continuous sections of the energy axis. The set of allowed energy values is called the eigenvalue
spectrum of the Hamilton operator. The energy spectrum of the Hamilton operator is bound from
below, which is exploited in the variational principle described in Section 13.5.
Once we know all eigenstates ψσ (~r, E), of the time-independent Schrödinger equation, we can
determine the wave function for all times as
Z
Eq. 4.1 dE −i Et/~ X
ψ(~r, t) = e ψσ (~r, E)aσ (E)
2π~ σ
We use the index σ to distinguish different eigenfunctions for the same eigenvalue.
The initial conditions, represented by the values ψ(r, t = 0), determine the coefficients aσ (E) of
this expansion in stationary states.
Z
dE X
ψ(r, t = 0) = ψσ (r, E)aσ (E)
2π~ σ
4.3 Method of separation of variables

In this chapter we will work almost exclusively with one-dimensional problems. One may ask how
much this has to do with the real three-dimensional world? The answer is given by the method of
separation of variables.
As in classical mechanics, it is often possible to exploit symmetries to break down a problem into
problems of lower dimensions, and, if we are lucky, into one-dimensional problems. An example is a
problem with a potential that varies only in one dimension, so that it is translationally invariant in
the other directions.
Quasi-onedimensional problem in three dimensions
Let us try to exploit this symmetry on a quasi-onedimensional problem: consider the Schrödinger
equation with a potential that varies only in x-direction:
h −~2 i
i ~∂t Ψ(~r, t) = ∇2 + V (x) Ψ(~r, t) (4.3)
2m
We can write the Hamiltonian Eq. 4.3 as sum H = H1 + H2 + H3 of three terms, where each acts
on one variable
−~2 2 −~2 2 −~2 2

H1 = ∂ + V (x); H2 = ∂ ; H3 = ∂
2m x 2m y 2m z
We make a product ansatz for the wave function
Ψ(~r, t) = φ1 (x, t)φ2 (y , t)φ3 (z, t) (4.4)
where each term φi solves the corresponding one-dimensional Schrödinger equation.

−~2 2
i ~∂t φ1 (x, t) = ∂ + V (x) φ1 (x, t)
2m x
−~2 2
i ~∂t φ2 (y , t) = ∂ φ2 (y , t)
2m y
2
−~ 2
i ~∂t φ3 (z, t) = ∂ φ3 (z, t) (4.5)
2m z
In order to show that this ansatz yields the correct result, we insert it into the Schrödinger equation
Eq. 4.3. We divide the equation by the wave function, which allows us to write the equations in a
more economical fashion.
1 product rule i ~∂t φ1 (x, t) i ~∂t φ2 (y , t) i ~∂t φ3 (z, t)

i ~∂t Ψ(~r, t) = + +
Ψ(~r, t) φ1 (x, t) φ2 (y , t) φ3 (z, t)
2
Eq. 4.5 1 −~ 2
= ∂ + V (x) φ1 (x, t)
φ1 (x, t) 2m x
2 2
1 −~ 2 1 −~ 2
+ ∂y φ2 (y , t) + ∂ φ3 (z, t)
φ2 (y , t) 2m φ3 (z, t) 2m z
Now there is the only tricky part: We can multiply the first term on the right-hand side with φ2 φ3
from the right side and divide by the same on the left hand side. This does not change the result,
because the Hamiltonian H1 does not act on φ2 φ3 . Similarly, we multiply the second term with φ1 φ3
on the right and divide by the same factor on the left. Finally, we multiply the second term with
φ1 φ2 on the right and divide by the same factor on the left.
Thus we obtain
2
1 1 −~ 2
i ~∂t Ψ(~r, t) = ∂x + V (x) Ψ(~r, t)
Ψ(~r, t) Ψ(~r, t) 2m
2 2
1 −~ 2 1 −~ 2
+ ∂y Ψ(~r, t) + ∂ Ψ(~r, t)
Ψ(~r, t) 2m Ψ(~r, t) 2m z
2
−~ 2 −~2 2 −~2 2
⇒ i ~∂t Ψ(~r, t) = ∂x + V (x) + ∂y + ∂z Ψ(~r, t) = HΨ(~r, t)
2m 2m 2m
This proves that our ansatz, Eq. 4.4 with Eq. 4.5, yields the correct solution.
General case
The method of separation of variables is not limited to quasi-one-dimensional problems. It can be

applied to a whole class of Hamiltonians. Let us state it in some greater generality.
METHOD OF SEPARATION OF VARIABLES
If the Hamiltonian is a sum of several parts, where each part only acts on one variable such as
 
X ~
i ~∂t Ψ(x1 , x2 , . . . , t) =  Hj ( ∂j , xj , t) Ψ(x1 , x2 , . . . , t) (4.6)
i
j
then the wave function can be expressed as

Y
Ψ(x1 , x2 , . . . , t) = φj (xi , t) (4.7)
j
with
~
i ~∂t φj (x, t) = Hj ( ∂j , x, t)φj (x, t) (4.8)
i
Thus we have broken down the multidimensional problem into many one-dimensional problems, that
are much easier to solve. Of course, due to the superposition principle, any superposition of such
solutions is again a solution.
In the proof, we exploit that the Hamilton operator Hi ( ~i ∂i , xi , t) does not act on wave functions
φj (xj , t) with j 6= i . They do not depend on the parameter xi on which the Hamilton operator
Hi ( ~i ∂i , xi , t) acts. Therefore, Hi ( ~i ∂i , xi , t)φj (xj , t) = φj (xj , t)Hi ( ~i ∂i , xi , t) for j 6= i .
The proof of Eq. 4.6 with Eqs. 4.7,4.8 goes as follows
Y product rule
i ~∂t φi (xi , t) = [i ~∂t φ1 (xi , t)] φ2 (x2 , t) · · · + φ1 (x1 , t) [i ~∂t φ2 (x2 , t)] · · · + . . .
i
| {z }
r ,t)
Ψ(~
" #
Y X 1
= φi (xi , t) i ~∂t φi (xi , t)
φi (xi , t)
i i
" #
Eq. 4.8 Y X 1 ~
= φi (xi , t) Hi ( ∂i , xi , t)φi (xi , t)
φi (xi , t) i
i i
" #
X ~ Y
= Hi ( ∂i , xi , t) φi (xi , t) q.e.d.
i
i i
| {z }
r ,t)
Ψ(~
The method of separation of variables can be used, if we are able to express the Hamilton operator
by a sum of differential operators that only depend on a single variable. Sometimes the choice is not
immediately apparent and a coordinate transform is required to bring it into the desired form. The
methods of finding such variable transforms are similar to the techniques used in classical mechanics,
such as the transformation on center-of-mass and relative coordinates in a many particle system and
the division into angular and radial coordinates in a spherical system.
Most Hamiltonians, however, cannot be written in this simple form. Even if the problem cannot
be separated exactly, the method of separation of variables is invaluable in finding approximate so-
lutions. As most quantum mechanical problems cannot be solved analytically or even numerically, it
is important to find models that represent the true system closely – even though not exactly –, but
contain symmetries that allow to break down the problem. This approximate model is the starting
point for approximation techniques discussed in section 13 to include whatever has been neglected in
the model system.
4.4 Free particles and spreading wave packets

A free particle does not experience forces. It experiences a constant potential. Thus, it obeys the
Schrödinger equation
h −~2 i
∂x2 + V0 ψ(x, E) = Eψ(x, E) (4.9)
2m
with a spatially constant potential V (x) = V0 .
The solutions can be found using the Ansatz ψ(x, E) = ei kx . By insertion we find the dispersion
relation
~2 k 2
E(k) = + V0
2m
The energy eigenstates of a free particle are plane waves
Ψ(x, E) = ei k(E)x (4.10)
1
p
with a wave vector k that has length k(E) = ± ~ 2m(E − V0 ).
We have discarded the solutions with E < V0 . They are not plane waves, but exponential
functions. Neither plane waves nor exponential functions are normalizable. Therefore, they cannot
describe the wave function of a single particle. While for the plane waves, a way out of this dilemma
has been found, for the exponential function the problem cannot be cured. Nevertheless, exponential
solutions will play a role as partial solution in problems that have been pieced together.
A few properties are noteworthy:
• The intensity φ∗ (x)φ(x) = e−i kx ei kx = 1 of the plane wave is uniformly distributed over all
space.
d xψ∗ (x) ~ ∂x ψ(x)
R
• The momentum of a plane wave is hpi = R d xψ∗ (ri)ψ(r ) = ~k. Furthermore, the spread of
momenta h(p − hpi)2 i = hp 2 i − hpi2 vanishes for a plane wave.
Thus a plane wave describes a beam of particles flying with the same momentum p, or in other words,
constant velocity v = ∂p E(p) = mp .
Free particle in three dimensions
We can use the result for the free particle in one dimension and the method of separation of variables
to obtain a solution for a free particle in three dimensions: The wave function in three dimensions is
the product of the three one-dimensional free particles.
~
Ψ(~r) = ei kx x ei ky y ei kz z = ei k~r
Here it becomes apparent why we denote ~k as wave vector.
The energy is the sum of the three one-dimensional problems
(~~k)2
E(~k) = + V0
2m
and the momentum is
R
Eqs. 3.28,3.37 ~ r)
d 3 r Ψ∗ (~r) ~i ∇Ψ(~
h~
pi = R
3 ∗
= ~~k
d r Ψ (~r)Ψ(~r)
Eq. 3.6
The velocity7 of the particles in the beam is then ~
v ~ p) =
∇E(~ 1
~.
= mp
7 The group velocity of a wave packet has been linked to the dispersion relation ω(k). The Schrödinger equation has
been constructed such that its dispersion relation is analogous to that of classical particles E = H(p, x, t). Therefore,
we may use Hamilton’s equations, that in principle are equations of motion for classical particles.
Wave functions that are not normalizable
There is a problem related to solution Eq. 4.10: It cannot be normalized, because the integral Eq. 3.33
is infinite. This would contradict the interpretation of the wave function as probability amplitude.
There are two possible ways around the problem.
1. Instead of an infinite integration volume we consider a large but finite volume, such as the
universe, which in fact is finite. In the finite volume the wave function is normalizable. Further-
more, the spectrum in a finite volume is discrete, and does not contain continuous sections.
The disadvantage is that the amplitude of the wave function is extremely tiny, if the volume is
very large. This point of view, which we will adopt in most of this book, avoids mathematical
subtleties. Before the expressions are evaluated in practice, one has to take the limit of infinite
volume.
2. We consider the non-normalizable wave function only as one part of the true wave function.
The true wave function would be a product of the non-normalizable wave function with an
envelope function that itself is normalizable, such as the wave packet
~
ψ(~r, t) = χ(~r, t)ei (k~r−ωt)
where the plane wave is not normalizable, but the envelope function χ(x, t) is normalized. By
multiplication with a normalizable envelope function, the complete wave function is normal-
izable. The corresponding energy eigenvalue spectrum may contain discrete eigenvalues and,
simultaneously, energy regions with a continuous set of eigenvalues.
The second choice is the mathematically more sophisticated approach. In a way, it limits the in-
tegration volume like the first suggestion. It does it in a way so that the wave function remains
differentiable everywhere in space.
4.4.1 Wave packets delocalize with time

Let us explore, what happens with a free particle that is localized in space. We start with a wave
packet of Gaussian shape8 . We use the ansatz
1 2
ψ(x, t) = A(t)e− 2 B(t)x ,
which leads to differential equations for A(t) and B(t). Because the derivation is a bit lengthy, we
solve the problem in the App. B. Here, we only report the solution:
s
Eq. B.16 4 C0 1 2 Eq. B.11 i ~t
ψ(x, t) = e− 2C(t) x with C(t) = C0 + (4.11)
π|C(t)|2 m
The symbol C0 is the initial value of C(t). The wavefunction acquires a complex phase, because
C(t) becomes complex. As a result the wave function oscillates in space.
What we are mostly interested in is the probability to find the particle at a certain point in space.
Let us therefore evaluate the probability distribution P (x) = ψ ∗ (x, t)ψ(x, t).The derivation is found
again in appendix B. Here we only report the result.
s 2
2
1 ~t

Eq. B.17 x
− 12 d (t) Eq. B.18
P (x) = √ e with d(t) = d(0) 1 + (4.12)
2π · d(t) 2md 2 (0)
√
The initial width is related to the coefficient C in Eq. B.11 by d(0) = 21 C0 . The density maintains
its Gaussian shape, but the width, which is proportional to d(t) grows, after a certain while, almost
linearly with time, that is d(t) → 2md ~t
(0) for t → ∞.
2
8A Gaussian has the form e−ax .
d(t)
0 t
Fig. 4.3: Decay of a Gaussian wave packet with time. The horizontal axis is a spatial coordinate.
The full line is the real part of the wave function and the dashed line is the imaginary part. The
three figure represent the initial state and equi-distant later times. Note that the absolute square
of the wave packet maintains its Gaussian shape. Real and imaginary part however oscillate. The
oscillation is more rapid in the wings: The wings are made of the contributions that run away most
quickly, which also have the smallest wavelength.(vg = m1 ~k).
Something strange happens here: the smaller the wave packet is initially, the faster it spreads
out. In other words, a small width d(0) leads to a rapid growth of the width as seen from Eq. 4.12.
This observation can be rationalized as follows: When we build a narrow Gaussian from plane waves,
Z
1 1 x 2 dk i kx − 1 (d ·k)2
√ e− 2 ( d ) = e e 2 , (4.13)
2π · d 2π
we have to include plane waves ei kx with large kinetic energy ~ω(k). Those contributions have high
velocity and run away quickly.
What we see here, is another manifestation of Heisenberg’s uncertainty relation , which will
be derived later in section 7.4. Heisenbergs uncertainty relation says that if the position of a particle
is well known, there is a large uncertainty about its momentum. In other words, if the position is
precisely determined, that is if the width of the wave packet is small, the spread in velocities is large.
Delocalization in practice
Let us consider a person who has a mass of 50 kg and let us consider that its position is initially
certain to within the thickness of a hair that is d(0) = 1 µm. How long does it take until the
Eq. 4.12
position becomes uncertain to about d(t) = 1 mm. With ~ ≈ 10−34 kg m2 /s, we obtain t =
1
~ 2md(0)d(t) = 10
27
s. Because a year has about 3 × 107 s, we have to wait 1020 years!
On the other hand, let us consider a microscopic particle such as an electron: The electron has
a mass of about 10−32 kg. To spread from 1 µm to 1 mm, takes only one tenth of a microsecond.
Now let us consider a truly microscopic situation: Let us localize an electron on a single atom,
which has a radius of about 0.1 nm=10−10 m. What is the localization region after a typical atomic
oscillation period of about 10−13 s? Its diameter has grown to a few µm or 10,000 times the size
of an atom. This indicates that the atoms do not experience the electrons as “particles” but as
delocalized waves.
This discussion explains why delocalization as quantum effect is not observable in our daily life,
while it is essential in the nano-world.
4.4.2 Wick rotation

The decaying wave packet reminds us of a diffusion process. While the process of a diffusing particle
is entirely different from that of a spreading wave packet, there is an intrinsic connection on a
mathematical level. The diffusion equation is closely related to the Schrödinger equation. Consider
the diffusion equation
~ 2 ρ(~r, t)
∂t ρ(~r, t) = D∇
where ρ is the density and D is the isotropic diffusion constant. Now let us perform the so-called
Wick-rotation
~2 it
D→ and t→ and ρ→ψ
2m ~
then we obtain from the diffusion equation the Schrödinger equation of a free particle.
~ ~2
i ~∂t ψ(~r, t) = − ∇ ψ(~r, t)
2m
The close connection between quantum mechanics and statistical mechanics, which has become
evident here, has been extremely fruitful in the past and has lead to the Quantum Monte Carlo
method and the Renormalization theory for the description of phase transitions (Kenneth G. Wilson,
Nobel prize for Physics 1982).
4.4.3 Galilei invariance (home study)

We have seen that the wave function of a resting wave packet spreads equally in all directions. The
analogy with the diffusion equations may lead to a misunderstanding that this is also true if the
paerticle have initiually a non-zero mean velocity. Such a behavior would violate Galilei invariance,
saying that the laws of nature are independent of the velocity of an observer. This is the reason for
investigating a wave function with a finite velocity and its relation to the resting wave packet.
Galilei invariance implies that the laws of nature are independent of a (constant) velocity of the
observer. In other words, there is no absolute velocity.9 Here we illustrate, rather than prove, Galileo
invariance of the Schrödinger equation for a free particle by investigating how a wave packet changes
its shape with time.
We will see that envelope function of a free particle does not remain constant with time: Firstly,
it moves with a velocity v = ddEp and secondly it changes its shape slowly so that eventually it covers
all space. We will show that a moving wave packet behaves exactly the same way as a resting one.
In order to show this equivalence between a moving wave packet and a resting one, we propose
a transformation from a resting wave packet to one moving with velocity v . Then we show that the
transformed wave packet obeys the Schrödinger equation if the resting does as well.
If Ψ0 (x, t), the “resting wave packet”, obeys the Schrödinger equation for a free particle, then
also the wave function
def
Ψv (x, t) = Ψ0 (x − v t, t)ei (kx−ω(k)t) , (4.14)
9 Note that the problem with an external, spatially dependent, potential V (x) is not Galileo invariant, because that
potential is fixed in space. However, if we step back and include whatever creates that potential into the description.
This enlarged system is Galileo invariant.
the “moving wave packet,does”, if v is the group velocity
v = ∂k ω
Interesting is that the transformation does not only correspond to a time dependent shift of the wave
function, but that it obtains in addition a phase factor, the plane wave. A more detailed derivation
is given in appendix A on p.247.
Now we show by insertion that the moving wave packet Ψv (x, t) as defined in Eq. 4.14 obeys the
Schrödinger equation.
?
h d −~2 2 i
0 = i~ − ∂ Ψ0 (x| − t )ei [kx−ωt]
{zv }t , |{z}
dt 2m x
def def
=y (x,t) =τ(x,t)
h ~2 2 i
= ei [kx−ωt] −i ~v ∂y + i ~∂τ + ~ω + (∂y + 2i k∂y − k 2 ) Ψ0 (y (x, t), τ (x, t))
2m
2
h ~ ~k ~2 2 i
= ei [kx−ωt] i ~∂τ + ∂y2 − (v − ) i ~∂y + (~ω − k ) Ψ0 (y , τ )
2m | {z m } | {z2m }
=0 =0

−~2 2
= ei [kx−ωt] i ~∂τ − ∂y Ψ0 (y , τ ) = 0 (4.15)
2m
| {z }
=0
Thus, we have shown that Ψv (x, t), as defined in Eq. 4.14, is a solution to the Schrödinger equation
for a free particle, if
• k = ~1 mv ,
1 ~2 k 2 11
• ω= ~ 2m = ~ 2 mv
2
and, if
• Ψ0 obeys the Schrödinger equation for a free particle as well.
Note that the total derivative d/dt acts also on the argument y (x, t) = x − v t, whereas the
partial derivative only acts on the second argument of χ. In the last step we simplified using the
dispersion relation and v = ∂ω
∂k = m
~k
Eq. 4.15 is nothing but the Schrödinger equation for a particle at rest. Thus the shape of the
envelope function is independent of the velocity of the wave packet. This is Galileo invariance: The
equations of motion do not depend on the relative velocity between the object and the observer.
4.5 Particle in a box, quantized energies and zero-point motion

The particle in a box is our simple model for an atom. In a hydrogen atom, the electron is bound to the
nucleus by electrostatic attraction. Under normal circumstances, the electron cannot overcome this
attraction. We simplify the problem by replacing the Coulomb potential of the nucleus by a constant
inside the attraction region and infinitely high potential walls, which ensures that the electron does
not get lost.
0
0 L x
The potential V (x) for particle in a box enclosed by impenetrable walls is
(
0 for x ∈ [0, L]
V (x) =
∞ for x ∈ / [0, L]
where L is the width of the box.

Apart from the walls, the problem is similar to that of a free particle. The Schrödinger equation
in the well is
−~2 2
∂ ψ(x, E) = Eψ(x, E) for x ∈ [0, L]
2m x
In the well, we use the same Ansatz as for the free particle
(
Aei kx + Be−i kx for x ∈ [0, L]
ψ(x) =
0 otherwise
√
with k(E) = ~1 2mE. Outside the well the wave function vanishes.
The potential imposes an additional boundary condition on the wave function:
ψ(0) = 0 and ψ(L) = 0 (4.16)
The wave function vanishes continuously beyond the boundary of the well. At this point the motivation
for these boundary conditions is based only on physical grounds, namely that the probability density
in the region with infinite potential, that is outside the well must vanish. At the boundary of the box
the wave function is not differentiable. This reflects that the kinetic energy in the forbidden region,
and hence the curvature of the wave function, is infinite. A more rigorous justification is obtained
from the solution for the particle in a box with finite walls, the particle well, which is discussed in
section 4.8. We can use that solution and form the limit when the walls grow to infinity.
We satisfy the first boundary condition Eq. 4.16
ψ(0) = A + B = 0 ⇒ B = −A
i kx

−i kx
⇒ ψ(x) = = A e −e 2i A sin(kx) = C sin(kx)
= |{z} (4.17)
=:C
def
where we introduced a new variable C =2i A.
The second boundary condition Eq. 4.16 translates into
ψ(L) = C sin(kL) = 0 ⇒ kL = πn
where n is an arbitrary integer.10

Thus the wave vector can assume one of the discrete values kn , where
π
kn = n (4.18)
L
The dispersion relation E = (~k)2 /2m provides the corresponding energy eigenvalues.
(~kn )2 Eq. 4.18

~2 π 2
En = = n2 (4.19)
2m 2mL2
where n is an integer11 . Thus, in a box, only discrete wave vectors and eigenvalues are permitted.
They are quantized.
25
Fig. 4.4: The lowest five wave func-

tions of the particle in the box. The
20
vertical axis is the energy in units of
~2 π 2 /(2mL2 ). The energy value has
been taken as the zero axis for the cor-
responding wave function. The latter is
given in arbitrary units. Note that the
15
lowest state is not at energy zero, which
is a consequence of the zero point vi-
bration or Heisenberg’s uncertainty re-
lation. Note that the wave functions
can be classified according to their be-
10
havior under mirror operation about the
center of the box: Some are symmetric
and some are antisymmetric. This is a
consequence of the mirror symmetry of
the potential. Note that the number of
5
nodes (the zeros) increases with energy.
Also this is a general feature of Problems
with a local potential.
0
0 0.2 0.4 0.6 0.8 1
Finally we need to normalize the wave function, which determines C.
Z ∞
dx ψ ∗ (x)ψ(x) = 1
−∞
10 The zeros of the sinus function lie at multiples of π.

11 integer number=ganze Zahl
E3
E2
E1
k
trivial solution ϕ=0
zero−point energy
Fig. 4.5: The energy levels for a particle in a box are obtained from the dispersion relation E(k) of a
free particle by selecting equi-distant values kn = πL n for the k-points. The spacing of the k-points
changes inversely with the size of the box. The number of nodes increase with energy. Note that
the wave function for k = 0 does not exist, because it is identical to zero. Every wave function is
a standing wave composed of two plane waves with positive and negative k. They are shown in the
insets.
which implies
Z ∞ Z L
Eq. 4.17
1= dx ψ ∗ (x)ψ(x) = dx C ∗ C sin2 (kn x)
−∞ 0
Z L
1 1 L
= C ∗C dx − cos(2kn x) = C ∗C
0 2 2 2
r
2
⇒ |C| =
L
We may use C as real number.
Thus we obtain the wave functions and their energies as
r
2 Eq. 4.18 π
ψn (x) = sin(kn x) with kn = n
L L
2 2
Eq. 4.19 ~ π
En = n2
2mL2
We make some interesting observations:
• The lowest energy is not zero! Since the particle in the box has only kinetic energy, this implies
that the particle can not be at rest.12 This effect, which is a general feature of bound systems,
is called the zero-point motion and the corresponding energy is called zero-point energy.
We observe here an effect of Heisenberg’s uncertainty principle, which says that the position
and the momentum of a particle cannot be determined simultaneously with absolute certainty.
12 Note that we use here the language from classical mechanics for a quantum particle.
If the particle were at rest, its momentum would be zero and therefore certain. Consequently,
the position should be completely unknown, which contradicts with the fact that the particle is
confined in a box. The smaller the box the larger is the zero point energy. We will later learn
about a more rigorous formulation of Heisenberg’s uncertainty principle.
Loosely termed, Heisenberg’s uncertainty principle says that one cannot form a localized wave
packet with a single plane wave. Very localized wave packets require large sets of plane waves
and therefore have wide distribution of momenta. The average momentum, on the other hand
is still a single quantity. In a language of particles, the uncertainty principle is most puzzling.
However, in a wave description it is almost a triviality.
In a classical description an electron would fall right in the nucleus, since as accelerated charged
particles it emits electromagnetic waves, and therefore looses energy. The zero point energy
prevents this process.
• The allowed energies are quantized, that is they are discrete. This implies that we cannot
add energy to the system or remove energy from the system in arbitrary amounts. The energy
spacing is inversely proportional with the mass of the particle. Hence, for a very heavy particle
it would not be observable and the behavior would appear classical.
A word of caution is required here. The quantization only concerns stationary states. It is
possible to prepare a state with arbitrary energies. This state however is not a stationary state,
but a superposition of stationary states, which has a probability distribution that changes with
time.
• The number of nodes in the wave function increases with the energy. This is a general property
of second order differential equations in one dimension having a local potential. This statement
is a consequence of the Knotensatz by Richard Courant[14].
4.5.1 Time-dependent solutions for the particle in the box

Note that the solutions for the stationary Schrödinger equation, provide us with the energy levels,
but not with the dynamics of a general wave function. A general wave function has the form
∞
X i
ψ(x, t) = φn (x)e− ~ ǫn t an
n=1
where the coefficients an are determined by the initial conditions of the problem.
Let us consider several situations:
• If only one eigenstate contributes, so that

i
ψ(x, t) = φn (x)e− ~ ǫn t
we observe that the probability density
P (x, t) = ψn∗ (x, t)ψn (x, t) = φ∗n (x)φn (x)
does not change with time. Note that this is also true for excited states. This is the reason
that one calls these solutions stationary. The wave function itself, however, oscillates. The
higher the energy the faster the oscillation.
• Consider now the superposition of the first and the second eigenstate
i i
ψ(x, t) = φ1 (x)e− ~ ǫ1 t a1 + φ2 (x)e− ~ ǫ2 t a2
Fig. 4.6: Probability density P(x,t) for a superposition of ground and first excited state of the particle
in the box as function of time. The time axis points from the front to the back.
The probability density is no more constant in time

P (x, t) = ψ ∗ (x, t)ψ(x, t) = φ∗1 (x)ψ1 (x)a1∗ a1 + φ∗2 (x)ψ2 (x)a2∗ a2
i

+2Re φ∗1 (x)φ2 (x)a1∗ a2 e− ~ (ǫ1 −ǫ2 t
The term that does not vary in time is mirror-symmetric, while the oscillating term is antisym-
metric. This indicates that the particle oscillates back an forth with circular frequency given by
~ω = ǫ2 − ǫ1 . If the particle is charged, the particle acts like an antenna and it sends out elec-
tromagnetic radiation, i.e. light. The circular frequency of the radiation is ω, which indicates
that light is emitted only with certain spectral lines corresponding to the energy differences
if the discrete energy eigenlevels. Thus the system looses energy until it ends up in a pure
eigenstate, which in this case is the ground state.13 This is a first indication that atoms and
molecules end up after a certain while in pure eigenstates of the Hamilton operator.
• How does the quantum mechanical result relate to the classical result? Let us create a wave
packet and follow its motion.
We start from the general Ansatz for the wave function as superposition of the solutions of
the time-independent Schrödinger equation. The time dependence is included by attaching the
13 This is not obtained from the Schrödinger equation in the present form, because the electromagnetic fields do not
enter the Hamiltonian.

i
phase e−i ωn t = e− ~ ǫn t to each eigenstate φn (x).
∞ r
X 2
ψ(x, t) = sin(kn x) e−i ω(kn )t an
n=1 |
L
{z }
φn (x)
∞ r
X 2 1 i (kn x−ω(kn )t)
= e − e−i (kn x+ω(kn )t) an
n=1
L 2i
∞ ∞
X 1 X 1
= ei (kn x−ω(kn )t) √ an − e−i (kn x+ω(kn )t) √ an
n=1
i 2L n=1
i 2L
| {z } | {z }
ψ1 (x,t) ψ2 (x,t)
The solution seems to consist of two wave packets, ψ1 (x, t) and ψ2 (x, t), running in opposite
directions. There is even more to it: we can show that each of the two solution actually is
an array of partner solutions separated by a distance of 2L. These presence of those periodic
images is due to the equispaced discrete grid of wave vectors.
−2πn
∞ z }| { ∞
X
i (kn x−2kn L −ω(kn )t) 1 X 1
ψ1 (x, t) = e √ an = ei (kn (x−2L)−ω(kn )t) √ an
n=1
i 2L n=1
i 2L
= ψ1 (x − 2L, t) (4.20)
In our problem, the periodic images are not real because at most one of them is located in
the box. The other partners are only mathematical artefacts. Nevertheless they allow us to
imagine the processes that occur.
The two solutions ψ1 (x, t) and ψ2 (x, t) are also related by a geometic transformtion: they are
mirror images at the mirror plane at the right wall of the box, x = 0.
∞ ∞
X
−i (kn x+ω(kn )t) 1 X 1
ψ2 (x, t) = e √ an = ei (kn (−x)−ω(kn )t) √ an = ψ1 (−x, t) (4.21)
n=1
i 2L n=1
i 2L
Furthermore, there is a mirror plane at the right wall of the box, x = L.

Eq. 4.21 Eq. 4.20
ψ2 (x − L) = ψ1 (−x + L) = ψ1 (−x − L)
We choose the initial conditions such that the solution ψ1 (x, t) is a wavepacket somewhere
inside the box and with a velocity vg . Our analysis tells us that this wave packet will move to
one the walls, where it disappears “in the wall”. While this wave packet disappears, one of the
periodic images of ψ2 (x, t) appears out “of the wall”. At the walls these two solutions exactly
cancel consistent with the requirement that the wave function goes to zero at the boundaries
of the box. The wave packet ψ2 (x, t) moves to the opposite wall, where it disappears. At the
same time the next periodic image of ψ1 (x, t) is ready to appear out of the wall. The result is
that the wave packet bounces back and forth exactly as a classical particle would do. As the
wave packet moves back an forth, it delocalizes exactly as we described it for the free particle.
2L 2L
4.6 Potential step and splitting of wave packets

The potential step is the basic building block for a large class of problems that can be solved fairly
easily. The potential step will also make us familiar with a new quantum effect, namely the splitting
of wave packets.
Imagine an interface between a metal and a semiconductor. Electrons can propagate in the semi-
conductor, if they have sufficient energy to enter the conduction band. Thus the electrons experience
a potential step. Metal-semiconductor interface are present at electric contacts to semiconductor
devices, so-called Schottky contacts. They are the key element of Schottky-diodes used as rectifiers.
V VII
VI
∆V I
I II
x
The problem is defined as follows: Space is divided into two regions, region I for x < 0 and region
II for x > 0. In region I we have a constant potential VI and a solution ψI and in region II we have a
potential VII and a solution ψII so that
V (x) = θ(−x)VI + θ(x)VII

ψ(x) = θ(−x)ψI (x) + θ(x)ψII (x)
where θ(x) is the Heaviside function or step function defined by


0 for x < 0

θ(x) = 12 for x = 0 .

1 for x > 0

4.6.1 Kinetic energy larger than the potential step

Let us first consider the case when the energy is sufficient to mount the step. A classical particle in
such a situation will always mount the step and proceed with a smaller velocity as some of its kinetic
energy has been used up to mount the step. For a quantum particle everything is different.
For a given energy, the energy eigenstates are
ψI (x) = Aei kI x + Be−i kI x

ψII (x) = Cei kII x + De−i kII x (4.22)
with
p p
~kI = 2m(E − VI ) and ~kII = 2m(E − VII ) (4.23)
For a given energy, two of the four coefficients

(A, B, C, D) are determined by the condition that the wave 2 2
E II (k)= h k +V II
E 2m
function is continuous and differentiable at the potential
step. The remaining coefficients correspond to two inde- 2 2
E I(k)= h k +V I
pendent solutions of the problem, each having a variable 2m
pre-factor. To pick one solution we select D = 0 and E
A = 1. This corresponds to a particle beam that arrives
from the left and scatters into the forward and the back-
ward direction. No particles arrive at the step from the
right side, i.e. D = 0.
Now we determine B and C from the requirement that
the wave function is differentiable at the step. −kI −kII 0 kII kI k
Eq. 4.22
ψI (0) = ψII (0) ⇒ 1 +B = C + |{z}
|{z} 0
A D Fig. 4.7: Dispersion relations from both
Eq. 4.22
∂x ψI (0) = ∂x ψII (0) ⇒ i kI − Bi kI = i kII C sides of the step. Observe that the
phase velocity vph = ωk in this example
which results in if higher ontop of the step than at the
kI − kII
B= (4.24) lower side. Nevertheless the group ve-
kI + kII locity vgr = ∂ω
∂k is lower ontop of the
2kI step, because some of the kinetic energy
C= (4.25)
kI + kII is used up by mounting the step.
Thus, all parameters are determined by the energy E = ~ω, which in turn determines the values
for kI (E) and kII (E) via Eq. 4.23. We obtain the time-independent solution as

i kI x kI − kII −i kI x 2kI i kII x
φ(x, E) = θ(−x) e + e +θ(x) e (4.26)
kI + kII kI + kII
| {z } | {z }
ψI (x,ω) ψII (x,ω)
The time-dependent solution is obtained by multiplication of the stationary solutions with e−i ωt .
We obtain the time-dependent partial solutions
kI − kII i (−kI x−ωt) 2kI
ψkI >0 (x, t) = θ(−x)ei (kI x−ωt) + θ(−x) e + θ(x) ei (kII x−ωt) (4.27)
kI + kII kI + kII
| {z } | {z }
ΨI ΨII
The solution can be characterized by the wave vector kI of the incoming wave instead of its
energy. The requirement that, in the beginning, the wave packet approaches the step from the left,
translates into kI ≥ 0.
Already from Fig. 4.7 a lot of the relevant physics is obtained: The momentum ~k ontop of the
step is smaller than on the lower ledge. Also the velocity ontop of the step is smaller as seen from
the slope of the dispersion relation at the respective wave vectors.
The striking result is however, that the particle beam is split into a transmitted beam and a
reflected beam. In contrast, a classical particle either mounts the step, or, if the energy is not
sufficient, it is completely reflected. A quantum particle has a certain probability to be reflected,
even if the energy is sufficient to mount the step. Splitting of wave packets is one of the quantum
effects.
The probability density for a beam of particles, i.e. for a plane wave, is constant in space. Thus
the particle density on the right-hand side of the step is constant. On the left-hand side, the incident
beam interferes with the reflected beam, leading to ripples in the density distribution. These ripples
are due to standing waves. 14
14 A standing wave is for which the amplitude oscillates with time, but the wave crests remain in place. A standing
wave is constructed from two running waves with opposite wave vectors, such as cos(~ k~r − ωt) + cos(−~ r − ωt) =
k~
2 cos(~k~
r ) cos(ωt)
The situation described here is rather similar to the refraction of light at the surface of an optically
denser medium. Partial reflection and partial transmission at an interface is a well known observation.
However some properties differ: For a particle with a mass, the momentum ~k is smaller ontop of
the barrer than on the lower ledge. For light entering a denser medium the wave vector grows larger.
As a result the angle of refraction is smaller than the incident angle for light, while for particles it is
larger than the incident angle.
4.6.2 Construct the other partial solution using symmetry transformations

A second-order differential equation in one dimension has always two linear independend solutions.
In order to obtain the general solution, we need to construct one other independent solution of the
problem.
The second partial solution can be obtained using symmetry arguments from the one just derived
and presented in Eq. 4.27. Using symmetry arguments to construct solutions is an extremely powerful
technique that allows one to save a lot of computations. This is why I describe this example in some
detail:
I II II I II I
The second partial solution shall describe a wave packet, which approaches the step from the right
(higher) side and which hops down the step. We construct this solution by two successive transfor-
mations of the problem.
1. First we apply a mirror operation at the step. The Schrödinger equation remains independent
except for the potential. The new potential V ′ (x) = V (−x) describes a step with the upper
side on the left. The partial solution given in Eq. 4.27 is transformed by interchanging x by
−x, i.e. x ′ = −x, so that ψ ′ (x, t) = ψ(−x, t).
ψ ′ (x ′ , t) = ψkI >0 (x, t)

′
⇒ ψ (x, t) = ψkI >0 (−x, t)
Eq. 4.27 kI − kII i (+kI x−ωt) 2kI
⇒ ψ ′ (x, t) = θ(x)ei (−kI x−ωt) + θ(x) e + θ(−x) ei (−kII x−ωt)
kI + kII kI + kII
| {z } | {z }
Ψ′II Ψ′I
We have defined Ψ′I = ΨII (−x, t) so that it describes the transformed solution on the left side
of the step and Ψ′II = ΨI (−x, t) described the transformed solution on the right of the step.
The solution obtained after this operation describes a particle approaching the step from the
right-hand side, which is partially reflected and which partially hops up the step.
2. In a second operation, we restore the original potential: We interchange VI and VII , so that
the elevated part of the step is again on the right hand side. We need to control that the
derivation of Eq. 4.27 does not make use that VII > VI . The second transformation transforms
VI′′ = VII′ = VII and VII′′ = VI′ = VI . This implies replacing kI and kII , i.e. kII′′ = kI and kI′′ = kII .
In this way, we obtain our final result for the second partial solution.
kI − kII i (+kII x−ωt) 2kII i (−kI x−ωt)
ψ ′′ (x, t) = θ(x)ei (−kII x−ωt) − θ(x) e + θ(−x) e (4.28)
k + kII kI + kII
| {z I } | {z }
Ψ′′II Ψ′′I
In summary, the second solution ψ ′′ (x, t) given in Eq. 4.28 is obtained from the first partial
solution in Eq. 4.27 by the transformation
x ′ = −x and kII′ = kI and kI′ = kII .
This solution ψ ′′ describes a beam of particles approaching from the right, that is partially reflected
and partially transmitted down the potential step. Surprisingly, we observe that the particle are
partially reflected even at a downhill step.
Scattering of wave packets
In order to explore the physical content of these solutions, let us consider the scattering of a wave
packet. We may choose one that approaches the step from the left side with coefficients a(k).
Z
dk
ψ1 (x, t) = χ(x, t)ei (k̄I x−ω(k̄I )t) = a(k)ei (kx−ω(k)t) (4.29)
2π
ψ1 (x, t) describes the initial wave packet before it arrives at the potential step.
If the wave packet is localized initially in region I, it can be expressed entirely by partial solutions
given in Eq. 4.27. The wave packet can then be expressed by three wave packets, which we denote
as Ψ1 , Ψ2 and Ψ3 . The first one is, per construction, our initial wave packet Eq. 4.29.
Z h i
Eq. 4.27 dkI
ψ(x, t) = a(kI ) θ(−x)ei (kI x−ωt) + θ(−x)B(kI )ei (−kI x−ωt) + θ(x)C(kI )ei (kII x−ωt)
2π
Z Z
dkI dkI
= θ(−x) a(kI )ei (kI x−ωt) +θ(−x) a(kI )B(kI )ei (−kI x−ωt)
2π 2π
| {z } | {z }
ψ1 ψ2
Z
dkI
+ θ(x) a(kI )C(kI )ei (kII x−ωt)
2π
| {z }
ψ3
= θ(−x)ψ1 (x, t) + θ(−x)ψ2 (x, t) + θ(x)ψ3 (x, t) (4.30)
The coefficients B(kI ) and C(kI ) are those of Eq. 4.24 and Eq. 4.25. The wave packet ψ1 (x, t)
describes the incoming particles. The wave packed ψ2 (x, t) describes the reflected wave particles
and the wave packet ψ3 (x, t) describes the particles transmitted through the barrier.
In order to simplify the discussion, let us choose a very smooth envelope function for the wave
packet so that the wave function coefficients a(k) are localized near a given wave vector k̄. Then I
can approximate B(k) and C(k) by the value at k̄.
Z
Eq. 4.30 dkI
ψ2 (x, t) = a(kI )B(kI )ei (−kI x−ωt)
2π
Z
dkI
≈ B(k̄I ) a(kI )ei (−kI x−ωt)
2π
Eq. 4.24 k̄I − k̄II
= B(k̄I )ψ1 (−x, t) = ψ1 (−x, t) (4.31)
k̄I + k̄II
Thus, the reflected wave packet is identical to the incoming wave packet, but mirrored at the step
and reduced in size by the prefactor B(k̄I ). While the incoming wave disappears at the step due to
the step function, the reflected wave appears “from under its own step function”.
Let us now consider the third wave packet ψ3 (x, t): Here the wave vector kII occurs in the plane
wave. Since both, kI and kII , are linked to the energy, one can also express kII as function of kI . We
use Eq. 4.23
~2 kII2 = 2m(E − VII ) ∧ ~2 kI2 = 2m(E − VI )
⇒ ~2 kII2 − ~2 kI2
= −2m(VII − VI ) = −2m ∆V
q
⇒ kII (kI ) = kI2 − 2m ∆V /~2 (4.32)
Because only values of kI close to k̄I contribute appreciably to the integral, we replace this function
by its Taylor expansion about k̄I :

∂kII
kII (kI ) = kII (k̄I ) + (kI − k̄I ) + O(kI − k̄I )2
| {z } ∂kI k̄I
k̄II | {z }
=k̄I /k̄II
Eq. 4.34 k̄I

= k̄II + (kI − k̄I ) + O(kI − k̄I )2 (4.33)
k̄II
We inserted here the derivative

∂kII Eq. 4.32 1 2k̄I Eq. 4.32 k̄I def
k̄II =kII (k̄I k̄I
= q = = (4.34)
∂kI k̄I 2 k̄ 2 − 2m ∆V /~2 kII (k̄I ) k̄II
I
We use this expression in the following when we determine the relation between ψ3 and ψ1 .
Z
Eq. 4.30 dkI
ψ3 (x, t) = a(kI )C(kI )ei (kII (kI )x−ωt)
2π
≈k (k )
II I
z }| {
k̄I
Z
dkI i ( k̄II + (kI − k̄I ) x−ωt)
≈ C(k̄I ) a(kI )e k̄II
2π
Z
k̄
i (k̄ − I k̄ )x dkI k̄
i ( I k x−ωt)
= C(k̄I )e II k̄II I a(kI )e k̄II I (4.35)
2π
| {z }
∂kII
=ψ1 ( ∂kI
x,t)
Now we represent the incoming wave packet by a plane wave part and an envelope function
Eq. 4.29
ψ1 (x, t) = χ(x, t)ei (k̄I x−ωt) (4.36)
Eq. 4.35
k̄
i (k̄II − k̄ I
k̄I )x k̄I
ψ3 (x, t) = C(k̄I )e II ψ1 ( x, t)
k̄II
Eq. 4.36
k̄
i (k̄II − k̄ I k̄I )x k̄I k̄
i (k̄ I x−ωt)
= C(k̄I )e II χ( x, t)e I k̄II
k̄II
k̄I
= C(k̄I )χ( x, t)ei (k̄II x−ωt)
k̄II
Eq. 4.25 2k̄I k̄I
= χ( x, t)ei k̄II x−ωt (4.37)
k̄I + k̄II k̄II
Thus the transmitted wave packet ψ3 can again be related to the incoming wave packet ψ1 by a few
transformations:
• The leading wave k̄I vector is replaced by a smaller one, namely k̄II .
• The envelope function χ(x, t) is compressed by the factor k̄II /k̄I
• The velocity of the wave packet is reduced by the same factor k̄II /k̄I .
Eq. 4.25 2k̄I
• The amplitude is reduced by the factor C = k̄I +k̄II
Thus we can obtain the wave packets Ψ2 and Ψ3 from Ψ1 by simple geometric operations and a
rescaling of the amplitude and main wave vector. Ψ2 is a simple mirror image of Ψ1 and approaches
the step from the right side. Ψ3 is located on the same side as Ψ1 but it is stretched. While Ψ3 has
about the same envelope function as Ψ1 , the envelope function modulates a different plane wave.
Fig. 4.8: schematic representation of a 2-dimensional wave packet scattering at a potential step.
The lines represent the wave crests. The white areas represent the shape of the wave packet. The
arrows indicate the velocity, respectively the wave vector.
Initially the wave function Ψ is equal to Ψ1 and it is localized in the left half plane. Ψ2 and Ψ3 do
not appear yet because they are localized in that half plane, where the corresponding step function
vanishes. The wave packet approaches the step from the left-hand side. When it hits the step, Ψ1
starts to disappear under its step function. At the same time the wave packets Ψ2 and Ψ3 appear
from under their step functions, so that Ψ1 is gradually replaced by Ψ2 and Ψ3 . The wave packet
Ψ2 describes the particles reflected to the left, while Ψ3 describes the particles transmitted to the
right.
Reflection and transmission coefficient
The weight of the reflected wave packet is called the reflection coefficient
Z 2
def Eq. 4.31 |k̄I − k̄II |
R = dx ψ2∗ (x)ψ2 (x) =
|k̄I + k̄II |2
It is the probability that a particle is reflected at the potential step.
The probability that the wave packet transmitted is called transmission coefficient.
Z Z Z
def ∗ ∗ k̄I k̄I 2 k̄II
T = dx ψ3 (x)ψ3 (x) = |C(k̄I )| 2
dx ψ1 ( x)ψ1 ( x) = |C(k̄I )| dx ψ1∗ (x)ψ1 (x)
k̄II k̄II k̄I
|k̄II | |2k̄I |2 4|k̄I ||k̄II |
= =
|k̄I | |k̄I + k̄II |2 |k̄I + k̄II |2
| {z }
|C(k̄I )|2
It is easily seen that R + T = 1, that is the particle is either transmitted or reflected. R/T is
small when the kinetic energy of the incoming wave packet is much larger than the size of the step.
Thus we recover the classical result in this limit. When the energy is barely sufficient to mount the
step, that is when kII approaches zero, the reflection coefficient becomes unity.
Thus the wave packet splits at the step in two pieces. Such a behavior is impossible for a classical
particle, which always maintains its integrity. If its energy is sufficient it will always climb up the
step, and does not reflect. For a wave it is not surprising it can be split. A beam of light, that is
electromagnetic waves, can be partially reflected and partially transmitted through a window of glass.
The reflection and transmission coefficients as function of the energy is shown in Fig. 4.9.
1
R T
0 V E
Fig. 4.9: Transmission (green) and reflection coefficient (golden) for a step of height V as function
of energy. The transmission coefficient vanishes when the energy is lower than the step height and
it approaches unity for high energies.
What does it mean when the wave packet splits? The particle is not really divided into two distinct
pieces. The wavefunction provides us only with the probability to find the particle somewhere in space.
Hence a splitting wave packet tells us that the particle has a finite probability to pass on or to be
reflected.
There is one open question: Why do we not observe this beam splitting for macroscopic particles?
The answer lies in the abruptness of the potential step. A macroscopic potential step is not abrupt.
If the wave packet is small compared to the region over which the potential changes from one value
to another, the transmission coefficient approaches one and the wave packet behaves like a classical
particle. 15
Interestingly also a wave packet hopping down the step is reflected with finite probability.
It is possible to evaluate the transmission and reflection coefficients without using wave packets.
What we need are the amplitudes and wave vectors of the transmitted and reflected waves, that is
Eqs. 4.23,4.24 and 4.25. We can evaluate the particle current as density times velocity, that is
∂ω ~k
j(x) = Ψ∗ (x)Ψ(x) = Ψ∗ (x)Ψ(x)
∂k
|{z} m
vg
Thus we can define the reflection coefficient as the ratio of reflected versus incoming current and
the transmission coefficient as the ratio of transmitted versus incoming current.
j3 |C|2 kII 4kI kII

T = = =
j1 kI (kI + kII )2
j2 (kI − kII )2
R = = |B|2 =
j1 (kI + kII )2
Note that both, kI and kII are functions of the energy, and therefore transmission and reflection
coefficient are functions of the energy as well.
15 This can be seen as follows: we divide the potential step into many smaller steps, so that the total height is
preserved. Each plateau of the potential must be much larger than the size of the wave packet. The total transmission
coefficient is then the product of the transmission coefficients of the smaller steps. If the step is divided into many
small steps, the total transmission coefficient tends to one. (Hint: Use steps ki = kI eαi /n with eα = kII /ki )
4.6.3 Kinetic energy smaller than the potential step

If the energy is smaller than the potential step, a classical particle is always reflected at the step.
We will see that a quantum particle behaves similarly. However, it can enter the classically forbidden
region to a certain extent.
When the step is larger than the kinetic energy, kII = i λ is imaginary so that the plane wave is
replaced by a exponential function. Thus, the solution in region II can be written as
ψII (x) = Ce−λx + Deλx
We can immediately exclude a solution with D 6= 0, because the wave function is not normalizable,
since it grows to infinity for large x. Thus we conclude that the wave function decays exponentially
in region II.
On the left-hand side we can match a cosine function16 , that is ψI (x) = A cos(kx − δ). From
the matching condition for value and derivative, we obtain two equations
A cos(−δ) = C (4.38)
−kA sin(−δ) = −Cλ (4.39)
for the coefficients A, B and δ. We can choose one parameter freely, so we set A = 1.
λ
δ = −atan( )
k
It is not completely straightforward to work out C from this starting point. We first determine cos(δ)
from the expression Eq. 4.40 for δ. Then we obtain C from Eq. 4.38.
λ λ
δ =−atan( ) ⇒ tan(−δ) =
k k
2
λ2 sin (−δ) 1 − cos 2
(−δ)
⇒ = tan2 (−δ) = =
k2 cos2 (−δ) cos2 (−δ)
2
λ
⇒ 1 − cos2 (−δ) = 2
cos2 (−δ)
k

λ2
⇒ 1 = 1 + 2 cos2 (−δ)
k
s
Eq. 4.38 1
⇒ C = A cos(−δ) = 2
1 + λk 2
Thus we obtain
r
k2
C=
k2 + λ2
Therefore, we find
λ
ψI (x) = cos[kx + atan( )]
r k
k 2
ψII (x) = e−λx
k 2 + λ2
Important is that the wave function has an exponential tail that enters the classically forbidden
region, even though only for a small distance. This implies that the particle has a finite probability
16 We could also have chosen a superposition of an incoming and a reflected plane wave with complex coefficients.
However, we can choose on the right hand side a real dependence, i.e. Im[C] = 0. Then also ψI (x) is real. The real
part of the two partial solutions has the general form of a cosinus function with a phase shift.
to enter the classically forbidden region. For a classical particle this is impossible. The same effect
will lead us later to the effect of tunneling: A quantum particle can even penetrate a hard wall if it
is sufficiently thin.
The exponential damping of the wave function in region II corresponds to total reflection of light.
If we try to look through the water surface into the pond, there is a viewing certain angle where
the water surface reflects completely and all that is underneath the surface is hidden from us. If we
stand up and peek into the water at a steeper angle, we suddenly observe that the water becomes
more transparent.
4.6.4 Logarithmic derivative

Before we continue, we will learn a useful way to solve matching problems. The general problem is
one of two regions in a one-dimensional space. We assume that we have only one solution ψ3 in
region II, and the superposition of two solutions ψ1 and ψ2 in region I. Without limitation of generality
we choose region I to be the region left of x = 0 and region II to be the region right of x = 0.
In the case of the potential step we have selected one outgoing wave ψ3 = ei kx in region II, but
no incoming wave. In region I we have a superposition of ψ1 and ψ2 . The Ansatz for the wave
function is
h i
Ψ(x) = θ(−x) AΨ1 (x) + BΨ2 (x) + θ(x)Ψ3 (x) (4.40)
Without loss of generality, I have chosen x = 0 as the boundary between the regions I (x < 0) and II
(x > 0). θ(x) is the Heaviside function with θ(x < 0) = 0 and θ(x > 0) = 1. Thus θ(x) equals one
for region II and vanishes in region I, while θ(−x) equals one for region I and vanishes for region II.
def def
The matching problem has the following form, where we abbreviate ψi =ψi (0) and ∂x ψi =∂x |0 ψi .
! ! !
ψ1 ψ2 A ψ3
=
∂x ψ1 ∂x ψ2 B ∂x ψ3
where values and derivatives are taken at the boundary x = 0 between the two regions I and II. After
multiplication with the inverse of the matrix on the left hand side, we obtain the coefficients A and
B
! ! !
A 1 ∂x ψ2 −ψ2 ψ3
=
B ψ1 ∂x ψ2 − ψ2 ∂x ψ1 −∂x ψ1 ψ1 ∂x ψ3
!
1 ψ2 ψ3 (D2 − D3 )
= (4.41)
ψ1 ψ2 (D2 − D1 ) ψ1 ψ3 (D3 − D1 )
where we used the shorthand notation Di = Dψi (0) for the logarithmic derivative defined below.
Definition 4.1 LOGARITHMIC DERIVATIVE
∂x |x0 ψ(x)
Dψ (x0 ) = = ∂x |x0 ln[ψ(x)] (4.42)
ψ(x0 )
The logarithmic derivative is independent of a scale factor, that is of the normalization.
Thus we find the solution17 for the solution that is valid across the boundary.
h ψ (x) D − D ψ2 (x) D1 − D3 i
Eqs. 4.40,4.41 1 2 3
ψ = ψ3 (0)θ(−x) + + θ(x)ψ3 (x) (4.43)
ψ1 (0) D2 − D1 ψ2 (0) D1 − D2
17 It is not required to memorize this equation itself, but only its derivation.
Using this formula we can easily do the matching for a system with many potential steps. One
simply chooses one partial solution on one end of the steps, which plays the role of ψ3 (x). Then one
matches the solution of the next segment to ψ3 (x). The new solution in this segment now plays the
role of ψ3 (x) to do the matching across the next boundary. In this ways one can proceed step by
step from one end to the other and thus obtains one partial solution for the entire problem.
4.7 Barrier and tunnel effect

Let us consider a particle that approaches a barrier of finite width. If a classical particle approaches
such a situation, it will overcome the barrier, if its kinetic energy is sufficiently large, and if its energy
is lower than the barrier, it will be reflected. There is no grey zone. However, we have seen for the
potential step, that the particle can penetrate a little into the step, even if its energy is insufficient
to mount the step. This implies that if the barrier is sufficiently thin, there is a certain probability
that the particle appears on the other side of the step, even if that is classically forbidden. This is
the essence of the tunnel effect, which we will investigate here in some detail.
A famous application of the tunnel effect is the scanning
tunneling microscopy18 , which is one of the very few experi-
mental techniques that can make individual atoms visible. In the
scanning tunneling microscope a very fine needle is brought into
contact with a surface. If a voltage is applied between the needle
and the surface a current can flow. The distance between the
tip of the needle and the surface is of the order of a nanome-
ter. Due to the gap between surface and tip electrons cannot
flow classically. However they can tunnel from surface to the
tip. The resulting current is extremely sensitive to the distance,
which is exploited to measure the topography of the surface and
to control the motion of the tip.
V
I II I
III
V0
x
L
In the following, we will determine the quantum mechanical behavior of particles impinging on a
barrier. The potential has the following form
( ) ( )
V0 0<x <L
V (x) = for
0 x < 0; x > L
We call the region to the left of the barrier region I, the barrier itself region II and what is to the
right of the barrier is region III.
V =0 √
The wavenumber of the particle in region I and III is kI I= ~1 2mE. Two different cases need
to be distinguished in the middle region II:
18 Heinrich Rohrer and Gerd Binnig received, together with Ernst Ruska, the Nobel prize in Physics 1986 for the
invention of the Scanning tunneling microscope.

• Passing the barrier is classically allowed. That is the case of the kinetic energy in region I is
larger than thepbarrier V0 . The wave
p function is oscillatory in the barrier region with a wave
vector kII = ~1 2m(E − VII ) = ~1 2m(E − V0 )
• Passing the barrier is classically forbidden. In this case, the energy is below VII and the ki-
netic energy in the barrier region is negative! A classical particle would be reflected by the
barrier. Instead of an oscillating solution we
pobtain a wave function that decays (or increases)
exponentially with a decay constant λ = 1~ 2m(V0 − E).
We begin with the case where the kinetic energy of the particle is sufficient to mount the barrier.
We only sketch the outline of the derivation and leave the detailed derivation to appendix D.1 on
p. 259. We form an Ansatz of the wave function
ψ(x) = θ(−x)ψI (x) + θ(x)θ(L − x)ψII (x) + θ(x − L)ψIII (x)
with the partial solutions for the three regions:
ψI (x) = AI ei kI x + BI e−i kI x
ψII (x) = AII ei kII x + BII e−i kII x
ψIII (x) = AIII ei kI x + BIII e−i kI x
The Ansatz has six unknowns.
• Four of the six unknowns are determined by the conditions that the wave function must be, at
each boundary, continuous with value and derivative. We call a wave function that is continuous
with value and derivative, differentiable.
• The remaining two variables can be determined by fixing the initial condition of a physical
problem: If we are interested in a particle beam impinging from the left onto the barrier, we
would fix AI = 1 and BIII = 0. It is sufficient to determine two independent sets of initial
conditions. Every other solution can be formed by superposition of these two partial solutions.
Of interest is how many of the impinging electrons are transmitted. Thus we evaluate the
transmission coefficient
2
def AIII
T (E) =
AI
which we evaluated in the appendix in Eq. C.19 on p. 256. as

 −1
E>V  V02 2 L
p 
T (E) = 1 + 4E(E − V0 ) sin ( ~ 2m(E − V0 ))
 (4.44)
| {z }
kII L
The result is shown in Fig. 4.11.

The transmission coefficient has certain values where it becomes equal unity. Those states,
respectively their energies, are called resonances. The resonance can be understood as follows: The
wave packet on top of the barrier is partially reflected at the border of the barrier. If it is reflected
twice it can positively interfere with itself. The interpretation would be that the particle is caught
into a state with finite lifetime, where it remains for a while before it is transmitted again. Because it
pauses on top of the barrier the probability density there is enhanced. In the resonance the particles
are not reflected at all by the barrier, but every particle is transmitted.
In between the resonance energies the transmission coefficient is strongly suppressed, indicating
that a large fraction of the particle beam is reflected.
$4$ $4$
$3$ $3$
$2$ $2$
$T;R$
$T;R$
$1$ $1$
$0$ $0$
$-1$ $-1$
$-2$ $-2$
$-2$ $-1$ $0$ $1$ $2$ $3$ $-2$ $-1$ $0$ $1$ $2$ $3$
$E$ $E$
$4$ $4$
$3$ $3$
$2$ $2$
$T;R$
$T;R$
$1$ $1$
$0$ $0$
$-1$ $-1$
$-2$ $-2$
$-2$ $-1$ $0$ $1$ $2$ $3$ $-2$ $-1$ $0$ $1$ $2$ $3$
$E$ $E$
$4$ $4$
$3$ $3$
$2$ $2$
$T;R$
$T;R$
$1$ $1$
$0$ $0$
$-1$ $-1$
$-2$ $-2$
$-2$ $-1$ $0$ $1$ $2$ $3$ $-2$ $-1$ $0$ $1$ $2$ $3$
$E$ $E$
$4$ $4$
$3$ $3$
$2$ $2$
$T;R$
$T;R$
$1$ $1$
$0$ $0$
$-1$ $-1$
$-2$ $-2$
$-2$ $-1$ $0$ $1$ $2$ $3$ $-2$ $-1$ $0$ $1$ $2$ $3$
$E$ $E$
$4$
$3$
$2$
$T;R$
$1$
$0$
$-1$
$-2$
$-2$ $-1$ $0$ $1$ $2$ $3$
$E$
Fig. 4.10: Wave function and density of a wave scattering at a barrier of width one and height
10. The dashed line is the real part. The dotted line is the imaginary part and the full line is the
probability density. The energies of the impinging particles chosen are 1 below the barrier height,
and above the barrier height: 0.25, 0.5, 0.75, 1, 1.25, 1.5, 2, 3. in units of (~π)2 /(2mL2 ). Editor:
more explanation!
T(E)
R(E)
0
0 5 6 9 14
E
√
Fig. 4.11: Transmission coefficient of a barrier of height 5 and width π/ 2. The transmission
coefficient becomes nonzero, even if the energy is not sufficient to surmount the barrier classically.
If the kinetic energy is larger than the barrier height, the transmission coefficient shows resonances
2
at those energies above the barrier, where the infinite well has its eigenvalues, i.e. E = V + (~π)
2mL2 n .
2
The origin for these resonances lies in the fact that the particle is reflected partly even if it jumps
down a potential step.
Let us now consider the second case, where the kinetic energy of the incoming particle is smaller
than the barrier height. We can obtain the result for the transmission coefficient, by extending it
into the complex plane.
In this case, we have to express
ei (i x) − e−i (i x) e−x − ex ex − e−x

sin(i x) = = =i = i sinh(x)
2i 2i 2
so that out of Eq. 4.44 we obtain the result for E < V
−1
V02 1p
T (E) = 1 + sinh2 ( 2m(V0 − E)L) (4.45)
4E(V0 − E) ~
which is identical to Eq. C.20 on p. 256.

In that case, we see that the transmission coefficient Fig. 4.11 does not immediately drop to
zero. Even though a classical particle with that energy would be reflected a wave packet is partly
transmitted. This is the tunnel effect.
The tunnel effect is analogous to the observation that we can shine light through a very thin
metal foil. Even though the metal should reflect light completely, some of it can pass through, if the
foil is sufficiently thin.
4.8 Particle in a well and resonances
Let us consider a particle in a box with walls that allow it in principle to escape. This is a simple
model for an electron bound by a nucleus. Given sufficient energy the electron may escape from the
nucleus. This is the case we are considering here.
V
I II I
III
0
|V0|
V0 x
( ) ( )
V0 0<x <L
V (x) = for
0 x < 0; x > L
If the energy of a state is smaller than zero the solutions are similar to those of the infinite box.
The main difference is that the wave function does not vanish abruptly at the box boundary, but leaks
out with an exponential tail, as in the finite step. As the wave function is now less constrained the
energy is lowered as compared to the box with infinite walls.
Interesting are the scattering states with energy larger than zero. This is a model for an electron
scattering at an atom or an impurity in a solid. We introduce three regions: Region I to the left of
the well, region II is the well itself and region III lies to the right of the well.
−5 −4 −3 −2 −1 0
E
Within each region the potential is constant so that we choose an Ansatz built up from piecewise
solutions of the free particle with appropriate energy. For a given energy the wave vector in the well
is k = ±k0 and the wave vector in regions I and III is k = ±k1 with k0 and k1 defined as
1p
k0 = 2m(E − V0 )
~
1√
k1 = 2mE
~
Note that the wave vector k can be imaginary. This happens if the energy lies below the potential
energy. In that case, we expect an exponential instead of an oscillatory solution.
The detailed calculation can be found in appendix D.1. Here we only discuss the results.
1
T
√
Fig. 4.12: Transmission coefficient of a potential well with depth V0 = −5 and width π/ 2. The
transmission coefficient becomes unity at the resonances. The resonances, indicated by the double
arrows, lie at the energies, where the box with infinite walls has its eigenvalues, i.e. Ei = n2 − 5. If
the kinetic energy of the particle is large compared to the well depth, the particle does not experience
the well any longer and the transmission coefficient becomes unity.
2
We obtain the transmission coefficient T = AAIIII as

Eq. C.19
h V2 Lp i−1
T = 1+ sin2 2m(E − V0 )
4E(E − V ) ~
The transmission coefficient has maxima where k0 L0 = nπ, n being an integer, where it is unity.
This is where the infinite potential well of same width has its eigenvalues. In between it drops to
a value T = 4E(E−V )
(V −2E)2 , which is zero if the kinetic energy E outside the well vanishes, and which
approaches one, if the kinetic energy is much larger then the well potential.
The energies, for which the transmission coefficient has its maxima, are called resonances. At
those energies the electron is not reflected at all. The energies of the resonances can be estimated
from those of the particle in a box. They are approximately at
~2 π 2 2
E ≈ V0 + n
2mL2
4.9 Summary
We have learned that wave packet of a free particle becomes broader with time. This is due to the
fact that probability distribution for the momenta is broad. Heisenberg’s uncertainty relation indicates
that the momentum distribution is broader if the position distribution is narrow. If the dispersion
relation E(p) is non-linear the broad momentum distribution results in a broad velocity distribution.
This velocity distribution is responsible for the broadening.
√
An energy eigenstate for the free particle is a simple plane wave with wave vector k = ~1 2mEkin
where Ekin = E − V is the kinetic energy. The higher the momentum of a particle, the smaller is its
wave length.
For negative kinetic energy we would obtain exponentially decaying and increasing functions as
eigenfunctions. They do not lead to eigenstates, because they violate the boundary conditions.
From the particle in the box we have seen that eigenstates are discrete, i.e. quantized, when the
particle is enclosed in a finite region.
The lowest energy for the particle in the box is not zero, but is has a minimum energy, the
zero-point energy. This observation can again be rationalized by Heisenberg’s uncertainty relation,
saying that a particle enclosed into a small spatial region has a large momentum uncertainty. This
implies that the momentum cannot be zero, which implies that it has a finite kinetic energy. This
zero-point energy becomes smaller the larger the box is. For an infinite box, that is for a free particle,
the zero-point energy vanishes.
The energy eigenvalues for the particle in a box can be estimated from the quantization condition
for the momenta, i.e. kn L = πn, and the dispersion relation E(p) with the de-Broglie relation p = ~k.
A wave packet arriving at a potential step with a kinetic energy, that is sufficient to classically
mount the step, is split into a reflected and a transmitted wave package. If the energy is smaller than
the barrier, the probability density does not disappear abruptly at the step, but the wave function
exhibits an exponential tail that enters the classically forbidden region. Nevertheless, all particles are
reflected. For light, the exponential tails in the classically forbidden region correspond to the so-called
evanescent waves. If the kinetic energy is very different from the potential step, either much lower
or much higher, the particle behaves classically, that is either the reflected or the transmitted wave
packet vanishes.
The potential well exhibits quantized energy levels for negative energies and a continuum of
scattering states for positive energies. The bound states are closely related to the states of the hard
box. It differs by the fact that the wave function does not go to zero abruptly at the boundaries of
the box, but that it has exponential tails reaching out into the classical forbidden region. The energy
levels are slightly below those of the hard box, because they have a little more room, because they
can leak out of the box a little.
For positive energies, the wave functions of the particle well are plane waves, that are modulated
in the well: The wave length is shorter, due to the larger kinetic energy and the probability density
may be larger. The particle is partly reflected at each boundary of the well. Hence it can, after being
reflected twice, interfere constructively or destructively with the original wave. For constructive
interference the probability density in the well can be much larger than outside. This is an indication
of a resonance. The resonance is the relict of the bound states of the hard box. This state can be
described classically by a particle that is captured in the well for a while, after which it is emitted again.
The life-time of the particle in the well is related to the width of the resonance, i.e. the probability
of finding the particle in the well as function of energy. The life-time is inverse proportional to the
width of the resonance. This behavior is reminiscent of the behavior of a driven harmonic oscillator
with friction. The width of the resonance peak in the susceptibility of a harmonic oscillator is inverse
proportional to the decay time (see PhiSX:Klassische Mechanik).
For the barrier, the particle will penetrate the barrier even if its energy is insufficient to classically
mount the barrier. Inside the barrier, the amplitude decays exponentially. Hence there is a finite
amplitude at the opposite side of the well, which leads to a transmitted wave. Hence the particle can
penetrate a barrier that, classically is too high. This is the so-called tunneling effect.
If the kinetic energy is larger than the barrier, the particle is reflected on both sides of the barrier,
which again leads to resonances just as in the potential well. The resonances are roughly at the
location where the quantization condition for the reflected wave is met, that is at the location where
the kinetic energy ontop of the barrier is related to the energy levels of the hard box.
Editorial remark: describe wave functions at the resonances, amplitude in the well, phase
shifts, delay and lifetime
4.10 Exercises
Chapter 5
Language of quantum mechanics
In this section we will become familiar with the mathematical basis of

quantum mechanics.
Any predictive theory must have three main elements:
• States that describe momentary physical situations
• Equations of motion that describe how a physical state evolves

in time. They are needed to extrapolate into the future.
• Measurements that allow to extract observable quantities from a

state which can be compared with experiment. Measurements are
needed to specify the initial state by comparison with an experi-
mental situation, and to verify or falsify the result by comparison
with experiment. Note that here measurement has the meaning
of extracting a prediction on outcome of a real experiment from
a state. Therefore, it is a theoretical concept.
Fig. 5.1: Paul Dirac,
In this and the following section we will define these ingredients for 1902-1984. British Physi-
quantum theory. cist. Nobel price 1933 in
Physics on the Dirac equa-
tion describing relativistic
5.1 Kets and the linear space of states electrons and their anti-
particles, the positrons.
A quantum mechanical state is represented by a ket written as |ψi. A
state can be expressed by a wave function in coordinate space as we have done before. There is
nothing wrong to have a wave function in mind when thinking about a state. The same state can
be represented either as a wave function in real space Ψ(x) or, alternatively, by its Fourier transform
Ψ(k). Both carry the same information. We call these two ways of describing the state different
representations of the same state. The concept of representations will be described more rigorously
later. We will see that there are a number of different and equally important representations of a
state, and finding a suitable representation is the key to solve many problems. The concept of a state
or ket abstracts from the particular representation. A state simply stands for the physical object.
All possible states, or kets, span the Hilbert space1 .
The space of all kets has the same mathematical structure as the space of all square-integrable2 ,
complex functions in real space, the wave functions. It also has the same structure as the space of
1 Strictly speaking, the set of quantum mechanical states with the operations defined below is a Hilbert space, but
not the only one. Physicists often refer to the set of all quantum mechanical states as the Hilbert space. Hilbert spaces
have been defined by von Neumann[15] R
2 A square integrable complex function in one dimension has a defined, finite value of the integral dx f ∗ (x)f (x).
81
82 5 LANGUAGE OF QUANTUM MECHANICS
vectors in infinite dimensions. Thus, in the following, rather formal, description of the basic postulates
of the Hilbert space, it is useful to compare and verify the relations also for square-integrable functions
ψ(x) and vectors ~r.
It may be useful to compare the abstraction from wave functions to kets, by comparing it to the
abstraction from n-tuples of numbers to vectors. When we begin to work with vectors, we usually
use n-tuples of numbers such as (rx , ry , rz ) to describe a vector. Later we introduce an abstract
symbol ~r for it. We realize that the n-tuple can represent a vector, if a coordinate system, that
is three basis vectors ~ ex , ~
ey and ~ ez , is chosen. The basis vectors are considered normalized and
orthogonal. That is we consider a cartesian coordinate system. We obtain the components of a
vector by rx = ~ ex ~r, ry = ~ ey ~r and rz = ~ ez ~r. We can also go the other way round and compose a
vector from its components as ~r = ~ ex rx + ~ ez rz .
ey ry + ~
Now we do the analogous abstractions from wave functions to kets: We start out from a wave
function ψ(x). Note that a function maps the argument x to a value namely ψ(x), similar as the
n-tuple maps the index i ∈ {x, y , z} to a number ri . A function is a very similar object as a vector: if
we make the index i continuous, we obtain a function ψ(x). A more abstract description of a wave
function is a ket |Ψi, just as the abstract notation of a vector is ~r. In order to obtain the wave
functions we need to introduce a basis. One possible basisset are the states |xi which correspond to
δ-functions in real space. That is, the wave function related to the ket |x0 i is δ(x − x0 ).
In order to do the next step we need to define a bra hψ| corresponding to a ket |ψi. When a ket
|ψi is related to a wave function ψ(x) the corresponding bra hψ| is related to the complex conjugate
wave function ψ ∗ (x).
Bra’s and ket’s are related to each other like standing and lying vectors. A lying vector can be
seen as an 1 × n matrix and a standing vector can be seen as a n × 1 matrix. A lying vector on
the left multiplied with a standing vector on the right gives the scalar product of the two vectors.
The scalar product is thus obtained using the usual rules of matrix multiplication. Whereas in the
common vector notation we use symbols such as a dot for the scalar-product and ⊗ for the dyadic
product, the matrix-like notation of vectors as standing and lying vectors does not require different
symbols for the different products. Instead, the type of product immediately follows from the order
with which the vectors are multiplied.
We proceed similarly for bra’s and ket’s. The scalar product between two states is defined as the
product of a bra on the left and a ket on the right-hand side.
Z
def
hψ|φi = dx ψ ∗ (x)φ(x)
P
This expression is analogous to the definition of the scalar product of vectors as a~ · ~b = j aj bj .
The main difference is that the index j is converted into a continuous argument x and the sum is
correspondingly converted into an integral.
The form of the expression is also the reason for the names “Bra” and “Ket”: A bra and a ket
together make up a “Bracket”. Often an operator is sandwiched “bracketed” between a bra and a
ket such as hψ|Â|φi.
Now we can obtain the “real space components” of a ket |ψi, namely the wave function, as
Z
hx|ψi = dx ′ δ(x − x ′ )ψ(x ′ ) = ψ(x)
This expression is analogous to the way we obtain vector components as ~

ei ~r = ri .
We can also reconstruct the ket from its wave function as
Z
|ψi = dx |xiψ(x)
P
which is analogous to the vector expression ~r = i ei ri .
~
5 LANGUAGE OF QUANTUM MECHANICS 83
5.1.1 Axioms
In the following I will present the definition of the so called Hilbert space, which defines the basic rules
for doing computations with kets. The term “Hilbert space” has been coined by von Neumann[15],
who also laid down the axiomatic foundation of quantum mechanics. His classic book[16] is an
excellent reference on this topic.
It is helpful to compare and verify these rules for vectors and wave functions. The analogy with
the familiar expressions will help to memorize the abstract rules.
Addition of states
The addition is defined as follows:
• for each pair of states in the Hilbert space H the sum |ψ1 i + |ψ2 i is a state of H. We may
def
also use a short hand notation |ψ1 + ψ2 i = |ψ1 i + |ψ2 i for the sum.
• Commutative Law with respect to addition:
|ψ1 i + |ψ2 i = |ψ2 i + |ψ1 i (5.1)
We say that the addition of states is commutative
• Associative Law with respect to addition of states:

|ψ1 i + |ψ2 i + |ψ3 i = |ψ1 i + ψ2 i + |ψ3 i (5.2)
We say that the addition of states is associative.
• There is a zero element |∅i:
|ψi + |∅i = |ψi (5.3)
For teaching purposes, I have introduced the symbol |∅i for the zero state. However, commonly
one denotes the zero state simply as 0, just like the number zero. Also this notation cannot
lead to misunderstandings.
Note that the symbol |0i is often used denote the ground state, which is not the zero state.
Because of this ambiguity I recommend to avoid using the symbol |0i for the zero state.
• For each state |ψi there is a negative state | − ψi:
|ψi + | − ψi = |∅i (5.4)
A set of states with the addition defined as just shown is called an abelian group under addition.
A abelian group is a set with some operation that obeys the properties shown above for the addition.
It is called abelian, because the commutative law holds.
Multiplication by a scalar
We define the multiplication by a scalar by
• For each state |ψi in H and each complex number c, there is a state |ψic in H.
• Associative Law with respect to multiplication by a scalar:
|ψi(ab) = (|ψia)b (5.5)

• Distributive Law with respect to multiplication by a scalar: The multiplication with a scalar is
distributive
(|ψ1 i + |ψ2 i)c = |ψ1 ic + |ψ2 ic (5.6)
We may also use a short hand notation |ψ1 + ψ2 i = |ψ1 i + |ψ2 i for the sum and |ψci = |ψic for
the product with a scalar.
Up to now, that is after defining addition and the multiplication with a scalar, we have defined a
linear vector space3 over the complex numbers.
Scalar product
A linear vector space with the scalar product as defined below is a Hilbert Space[15, 16].
We define a scalar product hψ1 |ψ2 i between two states |ψ1 i and |ψ2 i by
• hψ1 |ψ2 i is a complex number.
• Hermitean Symmetry: Exchanging the left and right state in the scalar product yields the
complex conjugate of the original scalar product.
hψ2 |ψ1 i = hψ1 |ψ2 i∗ (5.7)
• Linearity: Linearity requires distributivity Eq. 5.8, and associativity for a factor of the ket
Eq. 5.9.
hψ1 |ψ2 + ψ3 i = hψ1 |ψ2 i + hψ1 |ψ3 i (5.8)

hψ1 |ψ2 ci = hψ1 |ψ2 ic (5.9)
Note however, that hcψ1 |ψ2 i = c ∗ hψ1 |ψ2 i! This is a consequence of the Hermitean symmetry
Eq. 5.7. The scalar product is therefore a sesquilinear form, which implies
• Positive definiteness: The scalar product of a state with itself is positive definite. It is zero
exactly if the state is the zero state.
hψ|ψi ≥ 0
Note that from the preceding axioms we can show that the scalar product of a state with itself
is a real number, so that we can use here the greater or equal relation.
• Completeness of the Hilbert space: The limit of every Cauchy sequence in the Hilbert space
is again a member of the Hilbert space. A Cauchy sequence is a sequence |Ψ1 i, |Ψ2 i, . . . of
elements (states) with the property that
hψi − ψj |ψi − ψj i
for all elements |ψi i and |ψj i in the sequence can be made smaller than any positive number
ǫ, by dropping a finite number of elements from the beginning of the sequence.
• Separability of the Hilbert space: We will not discuss this axiom here and refer instead to the
Book of von Neumann[16].
3 A nonlinear vector space would be the space of distances in a curved space such as a sphere surface. Here for
example the distributive law for the multiplication with a scalar is not valid. Take one vector from the north pole to the
equator and a second one a quarter turn along the equator. If we add the two vectors we obtain a vector that points
from the north pole to the equator. If we double the two vectors, the first leads to the south pole and the second
moves back to the north pole. This result is clearly different than twice the sum of the two original vectors, which
points from north pole to south pole.Editor: Confirm that this is a nonlinear vector space
5.1.2 Bra’s and brackets

A scalar product is called a bracket. If speak about the left state in the BRAcket, we call it a bra. A
bra is written as hψ|. We can use the following rules for bras, which derive from the axioms mentioned
above.
hψc| = c ∗ hψ| (5.10)

hψ1 + ψ2 | = hψ1 | + hψ2 | (5.11)
5.1.3 Some vocabulary

• The norm of a state is
p
hψ|ψi (5.12)
the square root of the scalar product of the state with itself. Note that the scalar product of
a state with itself is often called the norm, which can lead to misunderstandings.
• A state is called normalized, if the norm equals unity, i.e. hψ|ψi = 1.
• Two states are called orthogonal, if their scalar product vanishes. That is hψ1 |ψ2 i = 0.
• A set of states {|ψi i} is called orthonormal, if hψi |ψj i = δi ,j . The states in an orthonormal
set are both orthogonal to each other and they are normalized.
• A set of states {|ψi i} is called linear independent, if we can create the zero state only by the
trivial superposition of those states, i.e.
X
|ψi ici = 0 ⇒ ci = 0
i
• A set of states is called linear dependent, if they are not linear independent.
• A set of states is {|ψi i} called complete if any state |φi can be expressed as superposition of
the states in the set, that is
X
|φi = |ψi ici .
i
5.2 Operators
Operators are transformations of states like functions are transformations of numbers. An operator
can be defined by defining the result of the operation for every possible state. We denote an operator
by Â and the result of the operation on a state |ψi as A|ψi.
Here some analogies
• A vector maps (i.e. transforms) a natural number from a finite set onto a number, its corre-
sponding vector component: i → y = yi
• a matrix maps a vector onto a vector: ~

x → y~ = A~
x
• a function maps a number onto a number: x → y = f (x)
• a functional maps a function onto a number. f (x) → y = F [f (x)] An example of a functional

is the definite integral.
• an operator maps a state onto a state: |ψi → |φi = Â|ψi. We will see that a state can be
represented by a vector. In that case the operator will be represented as a matrix mapping
a vector onto a vector. If the state is represented by a wave function, the operator will map
functions onto functions. An example for an operator is the derivative, which maps a function
ψ(x) onto a different function ddψx , namely that function which is the corresponding derivative
in each point in space.
5.2.1 Axioms
Not all operators are linear, but in quantum mechanics one uses almost exclusively linear operators.
Linear operators have the following two properties:
• Â(|ψ1 i + |ψ2 i) = Â|ψ1 i + Â|ψ2 i
• Â(|ψic) = (Â|ψi)c
From the above definition and the axioms for states, we can derive the basic rules for the algebra
of linear operators.
Addition
We begin to define the addition of operators as
(Â + B̂)|ψi = Â|ψi + B̂|ψi
which connects the addition of operators with that of states. The addition of operators has the
properties
• for any two linear operators Â and B̂ there is a linear operator Â + B̂
• Commutative Law with respect to addition of linear operators
Â + B̂ = B̂ + Â (5.13)
• Associative Law with respect to addition of linear operators:
(Â + B̂) + Ĉ = Â + (B̂ + Ĉ) (5.14)
• There is a zero operator ∅

ˆ
ˆ = Â
Â + ∅ (5.15)
• For each operator Â there is a negative operator.
ˆ
Â + (−Â) = ∅ (5.16)
Multiplication of operators with a scalar
The rules for the multiplication with a scalar follow directly from

Âc |ψi = Â|ψi c (5.17)
and the preceding axioms.

Multiplication of operators
Now we consider the multiplication of operators with themselves:
• for any two linear operators Â and B̂ there is a linear operator ÂB̂
• Associative Law
(ÂB̂)Ĉ = Â(B̂ Ĉ) (5.18)
• There is a unity operator 1̂:
Â1̂ = 1̂Â = Â (5.19)
The unit operator is the identity

• Distributive Law with respect to multiplication of two operators
(Â + B̂)Ĉ = ÂĈ + B̂ Ĉ (5.20)
The rules for operators and numbers are fairly similar. However, two axioms are missing!
• Two operators do not necessarily commutate, i.e. we cannot assume ÂB̂ = B̂ Â.
• Not every operator has an inverse.
5.2.2 Some vocabulary

• An operator Â† is the Hermitian conjugate or adjunct to the operator Â, if , for each ket
|ψi = Â|φi, the corresponding bra is hψ| = hφ|Â† .
ADJUNCT OPERATOR
|ψi = Â|φi ⇔ hψ| = hφ|Â† (5.21)
so that
hφ|Â† |ψi = hψ|Â|φi∗ = hÂφ|ψi (5.22)
Hermitian conjugate operators are required to interchange the order of bra’s and kets of matrix
elements.
Let us make the connection with matrices: for a Matrix A with matrix elements Ai ,j , the
hermitian conjugate matrix is defined as (A† )i ,j = A∗j,i or A † = (A
A⊤ )∗ . The hermitian conjugate
is obtained by transposing a matrix and taking the complex conjugate of the result.
We can easily show4 a few important rules that can be derived from the above axioms:
†
Â† = Â
(Â + B̂)† = Â† + B̂ †
(ÂB̂)† = B̂ † Â†
4 We show that (ÂB̂)† = B̂ † Â†
Eq. 5.22 Eq. 5.22 Eq. 5.22 Eq. 5.22
hψ|(ÂB̂)† |φi = hφ|ÂB̂|ψi∗ = hÂ† φ|B̂|ψi∗ = hB̂ † Â† φ|ψi∗ = hψ|B̂ † Â† |φi
• An operator that fulfills

Â† = Â (5.23)
is called hermitian.
Only hermitian operators can be related to observables (i.e. measurable quantities). Measurable
quantities must correspond to real expectation values. Only hermitian operators have the
property that all their expectation values are real.
hψ|Â|ψi = hψ|Â† |ψi∗ = hψ|Â|ψi∗ (5.24)
Example: Show thatR the differential operator p̂ =

~
i ∂x is hermitian, if the scalar product is
defined as hψ|φi = dx ψ ∗ (x)φ(x).
Z ∞ ∗
† ∗ ~
∗
hψ|p̂ |φi = hφ|p̂|ψi = dx φ (x) ∂x ψ(x)
−∞ i
Z ∞ ∗
~
=− dx [∂x (φ∗ (x)ψ(x)) − ψ(x)∂x φ∗ (x)]
i −∞
 ∗
Z ∞
~
= − [φ∗ (x)ψ(x)]∞ dx ψ(x)∂x φ∗ (x)

−
i | {z −∞} −∞
=0
Z Z
~ ~
= dx ψ ∗ (x)∂x φ(x) = dx ψ ∗ (x) ∂x φ(x) = hψ|p̂|φi
i i
†
⇒ p̂ = p̂
• The commutator [Â, B̂] of two operators is defined as

[Â, B̂] = ÂB̂ − B̂ Â (5.25)
The commutator will be important because observables that are represented by operators that
do not commutate, cannot be measured at the same time with arbitrary precision. This will be
explained in connection with Heisenberg’s uncertainty relation. Sets of commutating operators
can be used as quantum numbers.
Some rules for commutators can be derived from the definition Eq. 5.25.
[c Â, B̂] = c[Â, B̂]
[Â, B̂] = −[B̂, Â]
[Â + B̂, Ĉ] = [Â, Ĉ] + [B̂, Ĉ]
ÂB̂ = B̂ Â + [Â, B̂]
[ÂB̂, Ĉ] = Â[B̂, Ĉ] + [Â, Ĉ]B̂
[Â, B̂]† = [B̂ † , Â† ]
Note that the commutator is sometimes written with a minus-sign as subscript.
• Two operators with vanishing commutator are called commutating operators.
• A canonical momentum P̂ for a generalized coordinate Q̂ is an operator, which obeys the
commutator relation
~
[P̂ , Q̂] = (5.26)
i
The generalized coordinate Q must be a hermitian operator. For example momentum and
position are canonical conjugate of each other.
The canonical momentum p for the position x has been defined in classical mechanics by Eq. 3.5
as derivative of the Lagrange function with respect to the velocity. The physical meaning of
the canonical momentum p results from the Noether theorem as conservation number related
to a symmetry of the action with respect to translation in the coordinate x.
The quantum mechanical version of the canonical momentum follows from the correspondence
principle, Eq. 3.28, which relates the momentum to the wave vector of a wave packet. This
yields the expression p → ~k → ~i ∂x . This relation in turn determines the commutator [p, x] =
~
i between the coordinate and the canonical momentum.
5
• An inverse Â−1 of an operator fulfills
Â−1 Â = ÂÂ−1 = 1̂
Not every linear operator has an inverse: For example if an operator maps two different states
onto the same resulting state, we cannot find an operator, that knows from which initial state
the result was obtained.
• An operator Û is called unitary if
Û † Û = 1̂ or Û −1 = Û †
If a unitary operator Û has an inverse, then its inverse is its hermitian conjugate Û † .
While hermitian operators are important as they represent observables, unitary operators are
important as they describe transformations between physical states: Such a transformation
must map a normalized state onto a normalized state, which follows directly from the following
theorem:
Unitary operators and only unitary operators preserve the scalar product of two states if both
states are transformed by the same operation |φ′ i = Û|φi and |ψ ′ i = Û|ψi.
!
hφ′ |ψ ′ i = hÛφ |Ûψi = hφ| Û †
|{z} Û |ψi = hφ|ψi (5.27)
|{z}
hφ′ | 1̂
Furthermore, any transformation between states, that preserves the norm of all states is unitary.
The analogy of unitary operators are unitary matrices, which describe rotations and mirror
operations of vectors. This has been discussed in ΦSX:Klassische Mechanik. Rotations and
mirror images, and translations are those transformations that preserve the distance, the norm,
between any two points in space.
• We can also form functions of operators . The function of an operator is defined by its power
series expansion (Taylor expansion): In order to work out the function of an operator, one starts
from a regular function f (x), which depends on numbers x. One build the Taylor expansion of
the function, and then inserts the operator A for the argument.
∞
X 1 d n f
f (x) = n
(x − x0 )n
n=0
n! dx x0
∞ n
X 1 d f
⇒ f (Â) = (Â − x0 1̂)n (5.28)
n=0
n! dx n x0
5 This argument does not prove the reverse, namely that every pair forming the commutator relation Eq. 5.26 can
be mapped onto a momentum as defined in classical mechanics. This proof can be found in the article[17] by John v.
Neumann.
Since additions and multiplication of operators have been defined, the function of the operator
can be performed. One should remember, though, that a Taylor expansion has a finite radius
of convergence. This implies that the result may not be defined for all operators.
The function f of an operator A can also be defined by its eigenstates
Â|ψi = |ψia ⇒ f (Â)|ψi = |ψif (a) (5.29)
Example for the function of a matrix

We will see in the future that operators can be represented as matrices. Let us therefore calculate
a function of a matrix. We use the matrix
!
0 1
A=ˆ
−1 0
Let us form the exponential function of A . The power series expansion of the exponential function is
∞
X 1 n
ex = x
n=0
n!
We insert the argument, which is now a matrix, as argument into the function and the power-series
expansion.
∞
X 1
eA t = At)n
(A
n=0
n!
Now we need to determine the powers of A :
A0 = 1 !
1 0 1
A =
ˆ Â
=A
−1 0
! ! !
2 0 1 0 1 −1 0
A =
ˆ = ˆ −1
=
−1 0 −1 0 0 −1
A 3 = A 2 A = −1
1A = −AA
4 3
A = A A = −A AA = 1
5
A =A
We can combine these results in the following form:

( n−1
n (−1) 2 A for odd n
A = n
(−1) 2 1 for even n
Now we insert the powers of A into the power-series expansion:

∞ ∞
At
X 1 k
X 1
e = 1
(−1) t + 2k
(−1)k At 2k+1
2k! (2k + 1)!
k=0 k=0
∞
! ∞
!
X 1 k 2k
X 1 k 2k+1
= (−1) t 1+ (−1) t A
2k! (2k + 1)!
k=0 k=0
| {z } | {z }
cos(t) sin(t)
We recognize the power-series expansions of the sinus and cosinus functions

∞
X 1
cos(x) = (−1)k x 2k
(2k)!
k=0
∞
X 1
sin(x) = (−1)k x 2k+1
(2k + 1)!
k=0
Thus we obtain our final result

!
cos(t) sin(t)
eAt = cos(t)1 A=
1 + sin(t)A ˆ
− sin(t) cos(t)
Thus eAt is just the rotation matrix in two dimensions.

I would like to mention here some short-hand notations that are commonly used.
Â|ψi = |Âψi
hÂψ| = hψ|Â†
We do not use hψ Â| because that is confusing, if one does not write the hat on the operator, so that
one cannot distinguish between operator and state.
5.3 Analogy between states and vectors

Most of what we have learned here is known from vectors. The main difference between states and
vectors results from a wave function being complex, while we are used to vectors that are real. The
second complication is that states span a space of infinite dimensions, while vectors normally have a
finite number of dimensions.
Because wave functions are complex, we needed to introduce bra’s and ket’s, whereas we use the
same symbol for a vector, irrespective of on which side of the scalar product it stands. The vector
notation is a specialty of Euclidean spaces. For example when we deal in a curved space, we have to
introduce a metric and also then we need to distinguish what are called covariant and contravariant
vectors. They are distinguished in the notation by one having a subscript index and the other a
superscript index.
A result of the infinite dimension is that the index can be continuous. In that case we have a
wave function. The argument of Ψ(x) is nothing more than an index that has become continuous.
When one thinks of a vector one usually has in mind an n-tuple of real numbers. In reality a vector
is an abstract object, which has exactly two properties, namely a length and a direction. Describing a
vector by an n-tuple of numbers only makes sense after defining a basis. The same is true for states.
We usually have in mind wave function in real space, when we think of a state or ket. The state is
the abstract object and to imagine it as a function we have introduced a special basis, namely {|xi}.
As we can change the basis and transform into a different coordinate system, we can transform the
wave function into a different basis such as momentum eigenstates. The object remained the same,
but the basis and hence the numerical representation of the object has changed.
Below, I will show some analogies between vectors and states. Some of the objects described
here have not been discussed before, but this will be the subject of the following sections.
Note that the dyadic or outer product ~s ⊗ ~r of two vectors ~s and ~r is a matrix with elements
s ⊗ ~r)i ,j = si rj . The dyadic product is defined by
(~
a ⊗ ~b)~
(~ c = a~(~b~
c)
The dyadic product must not be mistaken by the vector product or cross product ~r × ~s, which is
defined in the three dimensional vector space as a vector.
Abstract object ~r |ψi

Inner (scalar) product ~s∗~r hχ|ψi
Outer (dyadic) product ~r ⊗ ~s∗ |χihψ|
Ortho-normality of the basis ~
ei ~
ej = δi ,j hφi |φj i = δi ,j
Basis {~ e} {|φi i}
Pi P
Representation ~r = i ~ ei ri |ψi = i |φi ici
Coefficients ri = ~ ei ~r ci = hφi |ψi
Operator/Matrix ~s = A~r |χi = Â|ψi
P P
Operator/Matrix si = j Ai ,j rj hφi |χi = j hφi |Â|φj ihφj |ψi
5.4 The power of the Dirac notation

The bracket notation of Dirac makes rather complex quantum mechanical expressions very simple.
It seems to me that the beauty is related with the fact that it makes the difference between “lying”
and “standing vectors” explicit. Thus we do not need different symbols between scalar and dyadic
product as demonstrated in the following equation.

|Ψ1 ihΨ2 | |ψ3 i = |Ψ1 i hΨ2 |ψ3 i

a~1 ⊗ ~b c~ = a~1 ~b~
c
In the Dirac notation there is only one product and no parentheses are required.
5.5 Extended Hilbert space

There is an important extension of Hilbert space. This space contains also those elements that can
be reached as the limit of a converging series in Hilbert space. It contains states with infinite norm
hψ|ψi = ∞. An example for a state that is not an element of Hilbert space, is the δ-function: We
can obtain the δ-function as a series of increasingly localized Gaussians. If we consider the norm in
this series, it tends to infinity. Another example would be a wave packet with an envelope function
that becomes broader until the wave packet becomes a plane wave. If we keep the wave packet
normalized, while performing the limit we obtain zero. But if we keep the amplitude at some point
in space constant we obtain a plane wave.
The scalar product of a state in this extended Hilbert space with a state in Hilbert space is always
finite.
Editorial Remark: Include something about discrete and continuous spectra
5.6 Application: harmonic oscillator

The harmonic oscillator is probably the single most important system in physics. The harmonic
oscillator describes a system, for which the Hamilton function is quadratic in the coordinates. In
other words, the forces are proportional to the deviation from some equilibrium position. The reason
for the importance of the harmonic oscillator is that, at low temperatures, any system will occupy the
regions where the potential is the lowest. These are the minima of the potential energy surface. If
the potential is smooth, we can approximate it by a Taylor series. If we truncate the Taylor expansion
after the first nontrivial term, the result is quadratic. This approximation is a harmonic oscillator.
The harmonic oscillator is, as an example, a model for vibrations of a molecule. Even though a
molecule has many degrees of freedom, the model system–a multidimensional harmonic oscillator–
can be broken up into decoupled one-dimensional harmonic oscillators.
We discuss the harmonic oscillator at this point, because it illustrates how the abstract notation
can be very efficient as compared to wave functions. It will be shown later that one has to rely
completely on the abstract notation for certain effects such as the spin of a particle.
5.6.1 Algebraic treatment of the one-dimensional harmonic oscillator

The Hamilton operator for the one dimensional harmonic oscillator has the following form.
p̂ 2 1 1 h 2 i
Ĥ = + c x̂ 2 = p̂ + m2 ω02 x̂ 2 (5.30)
2m 2 2m
pc 2π
where ω0 = m is the eigenfrequency of the oscillator defined as ω0 = T by the oscillation period
T.
Creation and Annihilation operators
In the following we will show that the the Hamiltonian can be expressed essentially as a product of an
operator with its hermitian conjugate. The special property of this operator will allow us to construct
the energy spectrum without determining the wave functions.
Let me first sketch the idea of the derivation given below using classical variables: The structure
of the Hamiltonian Eq. 5.30 is of the form a2 + b2 . We do know that a2 − b2 = (a + b)(a − b). The
minus sign can be converted by replacing b → i b so that a2 + b2 = (a + i b)(a − i b). If a and b are
real we find a2 + b2 = (a − i b)∗ (a − i b). What follows is analogous to these steps, only that now we
do them with operators instead of numbers.
The Hamilton operator can be written, up to a constant, as a product of two operators.
Eq. 5.30 1 2
Ĥ = p̂ + m2 ω02 x̂ 2
2m
1
= (p̂ + i mω0 x̂)(p̂ − i mω0 x̂) + [p̂, i mω0 x̂]−
2m | {z } | {z }
p̂ 2 −p̂i mω0 x̂+i mω0 x̂ p̂+m2 ω02 x̂ 2 +p̂i mω0 x̂−i mω0 x̂ p̂
1 h i
= (p̂ + i mω0 x̂)(p̂ − i mω0 x̂) + m~ω0
2m
h (p̂ + i mω x̂)(p̂ − i mω x̂) 1 i
0 0
= ~ω0 +
2m~ω0 2
Thus, the Hamiltonian can be written in the form

HAMILTON OPERATOR OF THE HARMONIC OSCILLATOR
1
Ĥ = ~ω0 â† â + (5.31)
2
where the operators â and â† are the creation and annihilation operators6 for the harmonic oscillator
as follows:
6 The definition of these creation and annihilation operators is not unique. Other definitions often encountered in
√
the literature differ by a prefactor i and/or the factor 2m~ω for the creation operator. The annihilation operator
obtains the corresponding prefactors consistent with hermitean conjugation.
CREATION AND ANNIHILATION OPERATORS FOR THE HARMONIC OSCILLATOR
r
1
â = (p̂ − i mω0 x̂) (5.32)
2m~ω0
r
1
â† = (p̂ + i mω0 x̂) (5.33)
2m~ω0
We will call the operator a annihilation operator and a† creation operator.
We will see below that the operators create and annihilate excitations of the harmonic oscillator. In
quantum field theory, the excitations of the harmonic oscillator will be interpreted as particles, so
that a† and a create and annihilate particles. Note that a and a† are indeed hermitian conjugates of
each other.
Spectrum of the harmonic oscillator
In the following we obtain the eigenvalue spectrum of the harmonic oscillator solely from abstract
operator equations. We will also find a first order differential equation for the ground state and a
recursive equation for the higher states.
Let us evaluate the commutator algebra and let us in addition consider the anti-commutators as
well.
ANTI-COMMUTATOR
An anti-commutator [Â, B̂]+ is defined as
[Â, B̂]+ = ÂB̂ + B̂ Â (5.34)
It looks like the commutator only with the sign reversed. In order to distinguish the commutator
clearly from the anticommutator, a “minus” sign is usually attached to the commutator [A, B] =
[A, B]− = AB − BA. Another common notation is to use curly brackets for the anticommutator.
1 i mω
0 2 1 1
ââ† = p̂ 2 + p̂x̂ − x̂ p̂ + x̂ = Ĥ + (5.35)
2~ω0 m 2~ | {z } 2~ ~ω0 2
=[p̂,x̂]=~/i
1 i mω
0 2 1 1
â† â = p̂ 2 − p̂x̂ − x̂ p̂ + x̂ = Ĥ − (5.36)
2~ω0 m 2~ 2~ ~ω0 2
Eq. 5.35,5.36 2Ĥ

[â, â† ]+ = (5.37)
~ω0
Eq. 5.35,5.36
[â, â† ]− = 1̂ (5.38)
Let us further consider the commutation relations between Ĥ and â, â†
1
[â, Ĥ]− = [â, ~ω0 (â† â + )]−
2
= ~ω0 (ââ† â − â† ââ) = ~ω0 [â, â† ]− â
= ~ω0 â
1
[â† , Ĥ]− = [â† , ~ω0 (â† â + )]−
2
= ~ω0 (â† â† â − â† ââ† ) = ~ω0 â† [â† , â]−
= −~ω0 â†
Thus we found two important relations,
[â, Ĥ]− = ~ω0 â

[â† , Ĥ]− = −~ω0 â† ,
which characterize â† and â as creation and annihilation operators. They are a kind of shift operators
that will play an important role when we discuss symmetries.
The creation and annihilation operators allow us to create new eigenstates from one given eigen-
state. If we apply the above equation to an eigenstate of the Hamiltonian |ψn i with eigenvalue En ,
we can show that a|ψn i is an eigenstate of the Hamiltonian with energy En − ~ω0 and that a† |ψn i
is and eigenstate with energy En + ~ω0 .
Ĥ â = (âĤ − [â, Ĥ]− ) = â(Ĥ − ~ω0 )

Ĥ|ψ i=|ψ iE
n n n
⇒ Ĥ â|ψn i = â|ψn i (En − ~ω0 ) (5.39)
† † † †
Ĥ â = (â Ĥ − [â , Ĥ]− ) = â (Ĥ + ~ω0 )
Ĥ|ψ i=|ψ iE
⇒ Ĥ â† |ψn i â† |ψn i (En + ~ω0 )
n n n
= (5.40)
Thus we can create a seemingly infinite number of eigenstates |Ψn−k i = ak |ψn i with equi-spaced
eigenvalues En−k = En −k~ω0 . Is there an end to it? Yes, the spectrum is bound below and therefore
the series must end. The spectrum is bound, because the expectation value of the Hamiltonian can
be written as the norm of a state, which is always nonnegative and a constant, namely
1
hψ|Ĥ|ψi = hψ|~ω0 (â† â + )|ψi
2

hψ|Ĥ|ψi hψ|â† â|ψi 1 hâψ|âψi 1
⇒ = ~ω0 + = ~ω0 + ≥ 0.
hψ|ψi hψ|ψi 2 hψ|ψi 2
We denote the ground state by |ψ0 i. We have assumed that the states are normalized.
The ground state fulfills
â|ψ0 i = 0 (5.41)
because otherwise there would be another eigenstate a|ψ0 i with lower energy.
What is the energy of this ground state?
Eq. 5.31 1 Eq. 5.41 1

Ĥ|ψ0 i = ~ω0 (â† â + )|ψ0 i = ~ω0 |ψ0 i
2 2
The energy of the ground state is E0 = 12 ~ω0 .

ENERGY SPECTRUM OF THE HARMONIC OSCILLATOR
The energy eigenvalue spectrum of the harmonic oscillator is equi-spaced

1
En = ~ω0 (n + ) (5.42)
2
where n is a non-negative integer.
Again we observe that the ground state energy does not vanish, just as
for the particle in a box. The ground state energy is due to the so-called E
zero-point vibrations or zero-point motion. It is not possible to remove
that remaining energy out of the system. The origin of the zero-point hω
vibrations is Heisenberg’s uncertainty relation , which says that it is not
possible to find quantum mechanical system with a determined position and
momentum. That would be required for the classical system to be at rest. hω
Can we observe the zero point vibration? We cannot observe it by removing
that energy, however if we compare the energy of an hydrogen molecule and
a deuterium molecule, which have identical forces acting between the two hω
atoms. (Deuterium is an isotope of hydrogen with a mass about two times
larger.) We find that the deuterium molecule is more stable by a few tenths
of an eV which is entirely due to the different zero-point energies. hω
1 hω
2
0
5.6.2 Wave functions of the harmonic oscillator
From the ground state all excited states can be constructed by repeatedly applying the creation
operator. However, we need to add normalization factor so that all states are properly normalized.
Eq. 5.35 Ĥ 1
hâ† ψn |â† ψn i = hψn |ââ† |ψn i = hψn | + |ψn i
~ω 2
Eq. 5.42
= hψn |ψn i(n + 1)
1
⇒ |ψn+1 i = √ â† |ψn i (5.43)
n+1
Eq. 5.36 Ĥ 1
hâψn |âψn i = hψn |â† â|ψn i = hψn | − |ψn i
~ω 2
Eq. 5.42
= hψn |ψn in
1
⇒ |ψn−1 i = √ â|ψn i (5.44)
n
EIGENSTATES OF THE HARMONIC OSCILLATOR
The eigenstates of the harmonic oscillator can be constructed by successively applying the creation
operator to the ground state |ψ0 i.
Eq. 5.43 1
|ψn i = √ (â† )n |ψ0 i (5.45)
n!
The ground state is defined by Eq. 5.41.
â|ψ0 i = 0 (5.46)
What is the wave function of the ground state? We can obtain it from
â|ψ0 i = 0 (5.47)
which is a first order differential equation

~
( ∂x − i mω0 x)ψ0 (x) = 0
i
A first order differential equation has a tremendous advantage over a second order differential equation
because it can be mapped onto a simple integration using the Ansatz
ψ0 (x) = eS(x)
With this Ansatz we obtain

~
∂x S − i mω0 x = 0
i Z x mω
0 ′ mω0 2
S(x) = dx ′ − x =C− x
−∞ ~ 2~
r
mω0 2 mω0 − mω0 x 2
ψ0 (x) = eS(x) = Ae− 2~ x = 4 e 2~
π~
where A = eC is the normalization constant. The ground state wave function of the harmonic
oscillator is a simple Gaussian.
The higher eigenstates are obtained by applying a† several times to the ground state. This is
even simpler that finding the ground state, because we only need to form the derivatives of the
already known states. The result are a polynomials times the Gaussian. The polynomials are the
so-called Hermite polynomials. Some useful relations about Hermite polynomials can be found in
the appendix of “Quantum Mechanics” Vol. I by A. Messiah.
5.6.3 Multidimensional harmonic oscillator

The multidimensional harmonic operator describes for example the vibrations of a molecule. The
vibrations can be tested using infrared (IR) spectroscopy, where the absorption of light is determined
as function of frequency. Light is absorbed when a vibration is excited from the ground state to an
excited state. The excitation energy taken from a photon that is absorbed and a phonon (or lattice
vibration) is “emitted”. As the vibrational energies are characteristic of the atomic structure, IR
spectroscopy is an important tool to understand the nature of the material under study. The atomic
vibrations can be described as a multidimensional harmonic oscillator. To lowest order the atoms
oscillate about their equilibrium positions. We will see that higher order terms of the Taylor-expansion
of the potential about the equilibrium structure results in an interaction between vibrations, so that,
in a crystal, we talk about phonon-phonon interactions. For further information on IR spectroscopy

see Atkins[7]
The multidimensional oscillator has the Hamiltonian
X p̂ 2 1X
i
H= + x̂i ci ,j x̂j
2mi 2
i i ,j
It is always possible to transform this Hamiltonian onto a system of independent one-dimensional

harmonic oscillators shown as follows.
P
First we introduce new coordinates X̂i = j Ai ,j x̂j with a real matrix A .
7
Per definition, the
canonical conjugate momentum P̂ of X̂ must fulfill the commutator relation 8 9
[P̂ , X̂] = ~i , which
P −1,⊤
determines the momentum as P̂i = j Ai ,j p̂j . Because we required the matrix A to be real, we
can replace A −1,⊤ by A −1,† .
TRANSFORMATION OF MOMENTA UNDER A COORDINATE TRANSFORM
X X
x̂i′ = Ai ,j xj ⇔ p̂i′ = Ai−1,†
,j pj (5.48)
j j
The transformation matrix A is real. The transformation of momenta is chose such that the trans-
formed coordinates and momenta obey the commutator relation [p̂ ′ i , xˆ′ j ] = ~i δi ,j
We will use two such transformations in the following derivation. The first transformation is a
rescaling of the positions and momenta to get rid of the masses.
def√
ξ̂i = mi x̂i
def 1
π̂i = √ p̂i
mi
so that the Hamiltonian has the form
X1 1X 1 1
H= π̂i2 + ξ̂i √ ci ,j √ ξ̂j
2 2 mi mj
i i ,j | {z }
Di,j
7 The
generalization to complex transformations is at least problematic.
8 The
transformation rules for the momenta can be obtained from classical physics as follows: Let us consider a
transformation of the coordinates.
X ∂fi ∂fi
x ′ , t)
xi = fi (~ ⇒ ẋi = ẋ ′ +
′ j
j
∂xj ∂t
We already calculated the transformation for the velocities, which will be required in the following
Now we express the action by the old and the new coordinates
Z Z X ∂ f~ Z
∂ f~
S[x(t)] = dt L(~ x˙ , t) =
x, ~ x ′ , t),
dt L(f~(~ ẋj
′
+ , t) = dt L′ (~ x˙ ′ , t) = S[x ′ (t)]
x′, ~
j
∂xj′ ∂t
The momentum is defined related to the old and the new (primed) coordinates is
∂L ∂L′ X ∂L ∂fj X ∂fj
pi = ; and pi′ = = = pj ′
∂ ẋi ∂ ẋi′ j
∂ ẋj ∂xi
′
j
∂xi
This provides the transformation rule for momenta for a given transformation of positions.
9
X X ~X ~
[P̂i , X̂j ] = [A−1
m,j p̂m , Ai ,n x̂n ] = A−1
m,j Ai ,n [p̂m , x̂n ] = Ai ,m A−1
m,j =
m,n m,n
| {z } i m i
~ | {z }
δm,n
i
δi,j
def
The matrix Di ,j = √1mi ci ,j √1mj is called dynamical matrix.
We diagonalize the dynamical matrix so that
X
Di ,k Uk,j = Ui ,j dj
k
the vector {Uk,j } for a given j is an eigenvector of D and dj is the corresponding eigenvalue. We
assume that the eigenvectors are normalized. Since the eigenvectors of a Hermitian matrix such as
D are orthonormal, U is a unitary matrix.
X
∗
Uk,i Uk,j = δi ,j
k
Especially for real matrices, the eigenvectors can be chosen to be real.10 This corresponds to a
transformation, where the transformed coordinate axes point along the principal axes.11
Now we are ready to do the second transformation12
def
X † X †√
X̂i = Ui ,j ξ̂j = Ui ,j mj x̂j
j j
def
X X 1
P̂i = Ui†,j π̂j = Ui†,j √ p̂j
mj
j j
which yields13
X 1 1
H= P̂i2 + di X̂i2
2 2
i
Qn Finally we use the method of separation of variables: We use the product Ansatz Ψ(X1 , . . . , Xn ) =
i =1 φi (Xi ), where each wave function φi (X) fulfills the Schrödinger equation for the one-dimensional
harmonic oscillator
Ĥi |φi i = |φi iei
1 1
Ĥi = P̂i2 + di X̂i2
2 2
There are of course many eigenvalues and eigenstates for each one-dimensional harmonic oscillator.
For reasons of simplicity we suppress the corresponding index.
The eigenvalues for the multidimensional harmonic oscillator are
X
E= ei
i
and the wave functions are obtained as

n
Y X √
Ψ(x1 , . . . , xn ) = φi Ui k mk xk
i =1 k
10 The eigenvectors are not necessarily real,which is important to know when using canned tools for the diago-
nalization.
P For real, hermitian matrices, however, the eigenvectors can be chosen real, because otherwise the form
∗ would not be real.
Di ,j = k Ui k dk Uj,k
11 germ(principal axis) = Hauptachse
U † )−1† = U † since U is unitary.

12 Note that (U
13
!
X X U∗
U =U
X X X †
X
π̂i2 = Ui ,j Ui ,k P̂j P̂k = Ui∗,j Ui ,k P̂j P̂k = P̂j Uj,i Ui ,k P̂k = P̂j2
i i ,j,k i ,j,k j,k i j
| {z }
δj,k
Chapter 6
Representations
In the last section we introduced abstract states and operators that

transform them into each other. The only way to produce numbers is
to evaluate scalar products and matrix elements. If we wish to represent
a state numerically, for example as wave function, we need to choose a
representation. Choosing a representation in quantum mechanics is like
choosing a coordinate system for vectors. In quantum mechanical cal-
culations, the choice of representations and the transformation between
representations is very important.
In this section we will learn about the matrix formulation of quantum
mechanics, where a state is represented by a vector and an operator is
represented by a matrix. The only difficulty is that we need to deal with
infinitely many dimensions.
We will also see the connection to wave functions as used in
Schrödinger’s wave-mechanical formulation of quantum mechanics,
Fig. 6.1: Werner Heisen-
where states are functions in space. The wave function can be thought
berg, 1901-1976. Ger-
of as a vector with a continuous vector index.
man Physicist. Nobel price
1933 in physics for the in-
vention of Quantum Me-
chanics, which he published
only 23 years old.
6.1 Unity operator
Let us consider a complete set of states {|ui i}. Then we can express the unity operator as
X
1̂ = |ui iKi ,j huj | (6.1)
i ,j
where K is the inverse of the so-called overlap matrix O defined as Oi ,j = hui |uj i.
X
Ki ,k huk |uj i = δi ,j (6.2)
k
101
102 6 REPRESENTATIONS
PROOF: we show that every state is mapped P onto itself, if the unity operator in this form is applied to it.
Every state |ψi can be expressed as |ψi = i |ui ici if {|ui i} is complete.
X X X
|ui iKi ,j huj | |ψi = |ui iKi ,j huj | |uk ick
i ,j i ,j k
X X X
= |ui i Ki ,j huj |uk i ck = |ui ici = |ψi
i ,k j i
| {z }
=δi,k Eq. 6.2
q.e.d.
If the set {|ui i} is not only complete but also orthonormal, the overlap matrix O and its inverse
K are unity matrices. Hence,
X
1̂ = |ui ihui |
i
for an orthonormal basis {ui }. In the following we assume {|ui i} to be orthonormal.
6.2 Representation of a state

We can now express a state |ψi in a certain representation. A representation is defined by a complete
basis of orthonormal states
P {|ui i}. A state can be represented in a certain basis by multiplication
with the unity operator i |ui ihui |.
X
|ψi = |ui ihui |ψi
i
Thus we can represent the state |ψi by a vector ψ, ~ where each vector component is a number,
namely ψi = hui |ψi. However, we need to specify the basis, namely the set {|ui i}. Compared to
ordinary vector algebra, the only uncommon points is that the vector has infinitely many components
and that the vector components are complex numbers.
What we have done is similar to representing a vector ~r in three dimensions in a basis spanned by
three orthonormal basis vectors ~ex , ~ ez . We can write
ey , ~
~r = ~
ex rx + ~
ey ry + ~
ez rz
Usually the vector ~r is then represented by its three components (rx , ry , rz ). In order to know the
vector ~r we need to know the choice of basis vectors ~ ex , ~ ez and the vector components rx , ry , rz
ey , ~
in this basis.
A special representation is the real space representation. Here the basis {|r i} is continuous. The
scalar products between states of the basis are defined as hr |r ′ i = δ(r − r ′ ). Instead of a sum we
need to choose the integral. The unity operator has the form
Z
1̂ = d 3 r |r ihr |
Thus we obtain
Z Z
|ψi = d 3 r |r ihr |ψi = d 3 r |r iψ(r )
so that we can represent the state by a function, the wave function ψ(r ) = hr |ψi.
6 REPRESENTATIONS 103
6.3 Representation of an operator

Let us now represent an operator A in the basis
X X
Â = |ui ihui | Â |uj ihuj |
i j
X
= |ui ihui |Â|uj ihuj |
i ,j
The operator A in the representation is a matrix A with matrix elements Ai ,j = hui |A|uj i.
If we consider a general matrix element of A we can write it in Bracket or vector representation
X X
hφ|Â|ψi = hφ|ui ihui |Â|uj ihuj |ψi = φ∗i Ai ,j ψj
i ,j i ,j
as vector-matrix-vector product. Thus a general matrix element can be written in a given represen-
tation as vector-matrix-vector product.
6.4 Change of representations

It is convenient to choose the representation according to the problem at hand. Therefore, it is
described here how to get from one representation to the other.
Let us start from a representation {|ui i}, and pass over to a representation defined by the base
{|vi i}. In order to discriminate between the representations we place the symbol of the representation
in parenthesis as superscript, when necessary.
†
Ui,j ψj(u)
X X X X z }| { z }| {
|ψi = |vi ihvi | |uj ihuj | |ψi = |vi i hvi |uj i huj |ψi
i j i j
| {z }| {z } | {z }
=1̂ =1̂ ψi(v )
~ is transformed by multiplication with a matrix U † ,

Thus the state vector ψ
X † (u)
ψi(v ) = Ui ,j ψj
j
Ui ,j = hui |vj i
The meaning of the matrix U is that it transforms the orthonormal basis {|ui i} into the other
orthonormal basis {|vi i}, that is
X X
|vi i = |uj i huj |vi i = |uj iUj,i
| {z }
j j
Uj,i
The matrix U , which describes the transformation between two orthonormal basissets is a unitary matrix.
PROOF: The matrix U is defined as Ui ,j = hui |vj i, where {|ui i} and {|vi i} are orthonormal and complete.
X † X X
Ui ,k Uk,j = hvi |uk ihuk |vj i = hvi | |uk ihuk | |vj i = hvi |vj i = δi ,j
k k k
q.e.d.
The transformation of matrix elements proceeds similar and we obtain

X X
hvi |Â|vj i = hvi | |uk ihuk | Â |ul ihul | |vj i
k l
X
A(v
i ,j
)
= Ui†,k A(u)
k,l Ul ,j
k,l
Let us summarize the transformation rules:

CHANGE OF REPRESENTATION
Under a change from one orthonormal basis to another, the vector components of a state and the
matrix elements transform as
(v )
X † (u)
ψi = Ui ,j ψj (6.3)
j
(v )
X (u)
Ai ,j = Ui†,k Ak,l Ul ,j (6.4)
k,l
where
Ui ,j = hui |vj i (6.5)
is a unitary matrix.
6.5 From bra’s and ket’s to wave functions

6.5.1 Real-space representation
Now we have learned to deal with abstract states and with wave functions. What is missing is how
to translate the expressions from one formulation to the other. While the connection is better shown
in the context of representation theory discussed later in chapter 6, I will show here already the main
arguments.
The connection between the two formulations lies in
def
Ψ(x) = hx|Ψi (6.6)
where the states |xi form a complete set of states. Note, however, that the states |xi are not
elements of the Hilbert space, but lie in the extended Hilbert space.
Similarly, the meaning of of the somewhat sloppy expression Âψ(x), where Â is an arbitrary
operator, is as follows
def
Âψ(x) =hx|Â|ψi
The states |xi obey the conditions:
x̂|xi = |xix eigenvalue equation (6.7)

′ ′
hx|x i = δ(x − x ) orthonormality (6.8)
Z
1̂ = dx |xihx| completeness (6.9)
The second equation, Eq. 6.8, shows that the wave function related to the state |x0 i, defined
according to Eq. 6.6, is a δ-function located at position x0 .
Position and momentum operators have the form1

Z
x̂ = dx|xixhx| (6.10)
Z
~
p̂ = dx|xi ∂x hx| (6.11)
i
The Schrödinger equation has the following form in the abstract notation and the wave function
notation.
2
p̂
+ v (x̂) |ψi = |ΨiE (6.12)
2m
2
−~ 2
∂ + v (x) ψ(x) = ψ(x)E (6.13)
2m x
The two above versions of the Schrödinger equations are mathematically correct. However, an
equation like the following is, strictly speaking, meaningless

−~2 2
∂x + v (x) |ψi = |ψiE
2m
because a function in space is multiplied with a abstract state. Nevertheless, such a notation is often
used in practice.
The difficult steps in the derivation of the above expression Eq. 6.13 from Eq. 6.12 are
1. to represent a function of an operator
Eq. 5.29
v (x̂)|xi = |xiv (x)
Z Z
⇒ v (x̂) = v (x̂) dx|xihx| = dx |xiv (x)hx|
| {z }
1̂
1 We first show that the first equation Eq. 6.10 holds:

Z Z Z
Eq. 6.9 Eq. 6.7
x̂ = x̂ dx |xihx| = dx x̂ |xihx| = dx |xixhx|
|{z}
| {z } |xix
1̂
Now we show that the second equation Eq. 6.11 holds: The momentum operator is defined as that operator that
has a commutator [p̂, x̂] = ~i . Thus we need to show that the commutator has the correct value.
Eq. 5.25
[p̂, x̂ ] = p̂x̂ − x̂ p̂
Z Z Z Z
Eq. 6.11 ~ ~
= dx|xi ∂x hx| dx ′ |x ′ ix ′ hx ′ | − dx ′ |x ′ ix ′ hx ′ | dx|xi ∂x hx|
i i
 
Z Z
 ~ ~ 
= dx dx ′ 
|xi i ∂x |hx|x
′
i x ′ hx ′ | − |x ′ ix ′ hx ′ |xi ∂x hx|
{z } | {z } i 
δ(x−x ′ ) δ(x−x ′ )
Z
~ ~
= dx |xi ∂x xhx| − |xix ∂x hx|
i i
Z ~ ~ ~
= dx |xi ∂x x −x ∂x hx| =
i i i
| {z }
~ ~
i +x i ∂x
2. to square the momentum operator

Z Z
2 Eq. 6.11 ~ ′ ′ ~ ′
p̂ = dx |xi ∂x hx| dx |x i ∂x hx |
′
i i
 
Z Z
~  ~
dx |xi ∂x  dx ′ hx|x ′ i ∂x ′ hx ′ |

=
i | {z } i
δ(x−x ′ )
Z
~ ~
= dx |xi ∂x ∂x hx|
i i
Z
= dx |xi(−~2 ∂x2 )hx|
6.5.2 Momentum representation

What are the momentum eigenstates? We start from the eigenvalue equation for the momentum
operator, which defines the momentum eigenstates |pi up to a constant.
p̂|pi = |pip eigenvalue equation (6.14)
We would like to know the momentum eigenstates as wave functions: With the help of Eq. 6.11
we obtain an eigenvalue equation for hx|pi.
Z Z
~ Eq. 6.14
dx |xi ∂x hx|pi = dx |xihx|pip
i
Now, we exploit that the kets {|xi} are linear independent. This implies that the coefficients for the
kets |xi must be pairwise identical.
~
⇒ ∂x hx|pi = hx|pip
i
def
This is a differential equation with constant coefficients for the function f (x) =hx|pi, which can be
solved with the exponential ansatz. As solution, one obtains a simple plane wave
i
hx|pi = e ~ px (6.15)
One should be careful not to write hx|pi = p(x) even though this is formally correct: One could
easily mistake it for a function that expresses the momentum as function of position.
Note that the prefactor of the wave function can be chosen in different ways. In our notation the
orthonormality condition and the representation of the unity operator are.
ORTHONORMALITY AND COMPLETENESS OF MOMENTUM EIGENSTATES
hp|p ′ i = 2π~ δ(p − p ′ ) orthonormality (6.16)

Z
dp
1̂ = |pihp| completeness (6.17)
2π~
i
1
In many textbooks, the momentum eigenstates are defined as |pi = √2π~ e ~ px . That is they
1
are normalized differently. Our notation has the advantage that the factors √2π~ do not appear in
the real space wave function corresponding to the momentum eigenstate. On the other hand, our
notation requires the factors 2π~ in the normalization condition Eq. 6.16 and the representation of
the unit operator, Eq. 6.17, which do not occur in the other notation.
6.5.3 Orthonormality condition of momentum eigenstates (Home study)

Here, we derive the normalization condition of the momentum eigenstates, namely Eq. 6.16. We use
the expression Eq. 6.15 for the real space wave function of a momentum eigenstate.
1̂
z
Z }| { Z Z
∞ ∞ ∞
i ′
′
hp|p i = hp| dx |xihx| |p ′ i = dx hp|xihx|p ′ i = dx e ~ (p−p )x (6.18)
−∞ −∞ −∞
This expression is awkward, because it contains an integral that does not even converge, if the
bounds of the integral are at infinity. Therefore, we have to play a common trick to make the integral
converge anyway2 : We multiply the integrand with a function, namely e−λ|x| , that attenuates the
oscillations for large arguments, i.e. at ±∞. For each value of λ, the integral is well defined. As λ
approaches zero, the function e−λ|x| approaches 1. Hence, we recover the original integral Eq. 6.18
in the limit λ → 0.
Z ∞
i ′
′
hp|p i = lim+ dx e( ~ (p−p )x e−λ|x|
λ→0 −∞
Z 0 Z ∞
( ~i (p−p ′ )+λ)x ( ~i (p−p ′ )−λ)x
= lim dx e + dx e
λ→0 −∞ 0
−1 −1 !
i i
= lim (p − p ′ ) + λ − (p − p ′ ) − λ
λ→0 ~ ~
−2λ
= lim
λ→0 −1 − p ′ )2 − λ 2
~2 (p
~λ
= 2~ lim (6.19)
λ→0 (p − p ′ )2 + ~ 2 λ2
The function in the limes has the form of a Lorentzian. A Lorentzian is defined as L(x) =
1 Γ/2
π (x−x0 )2 +(Γ/2)2 .
The Lorentzian has a peak at x0 and a width related to Γ . The Lorentzian has the
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
−4 −2 0 2 4
Fig. 6.2: Lorentzian with width Γ = 1
form of a spectral line in an absorption spectrum.

We recognize already that the function in Eq. 6.19 vanishes in the limit for p 6= p ′ and that it
diverges for p = p ′ , which suggests that the limes could be a delta function. What needs to be shown
is that the integral over p is equal to a well-defined constant.
2 The background is the theory of Distributions, which are generalizations of functions such as the Deltafunction of
Dirac. See [18]. Editor: In the first line we interchange the limit and the integral. Is this legitimate?
We use the integral formula taken from Bronstein[13]

Z
1 1 x
dx 2 = atan( )
x + a2 a a
Thus we obtain
 
Z ∞ ∞
Eq. 6.19 p − p′
dp hp|p ′ i
 
= 2~ lim atan( ) = 2~ atan(∞) − atan(−∞)
−∞ λ→0 ~λ −∞ | {z } | {z }
π/2 −π/2
= 2π~
Since a function that integrates to one, but vanishes everywhere except one point is a delta function3
, we can directly express the scalar product as
hp|p ′ i = 2π~ δ(p − p ′ )
which proves Eq. 6.16.
Now it is a simple matter to determine the unity operator in momentum representation as given
in Eq. 6.17. The form of the unity operator can be verified by showing that it does not change the
state to which it is applied. It is sufficient to show this for an arbitrary state from a complete set of
states, which we choose to be the momentum eigenstates |pi.
Due to the superposition principle, it holds for any state if it holds for a complete set of states
such as |pi. The identity is shown as follows:
Z Z
Eq. 6.17 dp ′ ′ ′ Eq. 6.16 dp ′ ′
1̂|pi = |p ihp |pi = |p i2π~δ(p − p ′ ) = |pi
2π~ 2π~
6.6 Application: Two-state system

After the harmonic oscillator, we learn here about the second most important model in quantum
mechanics, the two-state system. The two-state system is useful as a model for any system where
only the ground state and the first excited state is accessible. At low temperatures for example, it
may happen that we only have to consider the lowest two states. There are also systems that only
have two states. We will learn that the spin of an electron, which is the rotation of an electron about
itself, can only have two states namely a clock-wise and a counter-clock wise rotation about an axis
with a well defined absolute value of the angular momentum.
We denote the two states with the symbols |ψ0 i and |ψ1 i. |ψ0 i is the ground state of the system
and |ψ1 i is the first excited state.
Let us assume that they are eigenstates of the Hamilton operator with eigenvalues ǫ0 for the
ground state and ǫ1 for the excited state. Therefore,
Ĥ|ψ0 i = |ψ0 iǫ0
Ĥ|ψ1 i = |ψ1 iǫ1 (6.20)
We will see later that we can always choose the eigenstates to be orthonormal, that is
hψi |ψj i = δi ,j
We can now express the Hamilton operator in the basis of its eigenstates, so that it is represented
by a matrix
! !
hψ0 |Ĥ|ψ0 i hψ0 |Ĥ|ψ1 i Eq. 6.20 ǫ0 0
= =: H
hψ1 |Ĥ|ψ0 i hψ1 |Ĥ|ψ1 i 0 ǫ1
3 This conclusion is not entirely strict because there are also distributions (functions) that are zero everywhere and
integrate to zero. An example is the derivative of the delta function. An admixture of these functions cannot be
excluded.
The Hamilton operator has the form

1
X
Ĥ = |ψi iHi ,j hψj | = |ψ0 iǫ0 hψ0 | + |ψ1 iǫ1 hψ1 | (6.21)
i ,j =0
Hence we can write the Hamilton matrix H .

! ! !
ǫ0 0 ǫ0 + ǫ1 10 ǫ1 − ǫ0 −1 0
H= = +
0 ǫ1 2 01 2 0 1
Without loss of generality, we can choose the energy zero to be equal with ǫ0 +ǫ
2 , the meanvalue of
1
the two energies, which deletes the first term in the above equation. Furthermore we introduce the
frequency ω0 via the excitation energy
def
~ω0 = ǫ1 − ǫ0
so that
!
1 −1 0
H = ~ω0
2 0 1
Now we would like to create ladder operators just as for the harmonic oscillator they are defined
by
â|ψ0 i = |∅i
â|ψ1 i = |ψ0 i (6.22)
Note here that |∅i 6= |ψ0 i! |∅i is the zero-state and is associated with a wave function that is zero
everywhere in space. |ψ0 i on the other hand is the ground state.
Let us evaluate a in the representation (|ψ0 i, |ψ1 i)
! !
hψ0 |â|ψ0 i hψ0 |â|ψ1 i Eq. 6.22 0 1
=
hψ1 |â|ψ0 i hψ1 |â|ψ1 i 00
The states are orthonormal. We will later see that eigenstates of any hermitian operator (such as
the Hamiltonian) to different eigenvalues are always orthogonal.
We can now evaluate the matrix elements of a† as
! ! !
hψ0 |â† |ψ0 i hψ0 |â† |ψ1 i hψ0 |â|ψ0 i∗ hψ1 |â|ψ0 i∗ 00
= =
hψ1 |â† |ψ0 i hψ1 |â† |ψ1 i hψ0 |â|ψ1 i∗ hψ1 |â|ψ1 i∗ 10
Let us now investigate the commutator relations: We start working out the products of the ladder
operators. We can use the matrix notation, if we stay in the representation of eigenstates.
! !
† 00 † 10
â â = ˆ ; ââ = ˆ
01 00
! !
0 0 0 0
â2 =
ˆ =
ˆ∅ˆ ; â†2 =
ˆ =∅ˆ
00 00
Hence we can write the Hamiltonian as

!
1 1 −1 0
Ĥ = ~ω0 [â† , â]− =
ˆ ~ω0
2 2 0 1
while the anti-commutator4 is a constant
[â, â† ]+ = 1̂
[â, â]+ = ∅ˆ
† † ˆ
[â , â ]+ = ∅
Note here, that the commutator relations are similar to those of the harmonic oscillator as seen in
Eq. 5.38 only that commutators and anti-commutators are interchanged. This will be important if
we pass to a many particle description of Fermions such as the electron.
6.6.1 Pauli Matrices

It can be shown that all matrices in two dimensional space can be expressed in terms of 4 Hermitian
matrices, the three so-called Pauli matrices and the unity operator. Pauli matrices play an important
role for the description of spin and in general for certain particles called Fermions. Therefore, we
shall introduce them here. Some useful expressions that facilitate working with Pauli matrices are
provided in App. O.
We can now look for the most general form of an observable in the space spanned by the two
states. The most general Hermitian two-by-two matrix has the form
!
a b + ic
A=
b − ic d
which can be expressed in the form

! ! ! !
a+d 10 01 0 −i a−d 1 0
A= +b −c +
2 01 10 i 0 2 0 −1
a+d a−d
= + bσx − cσy + σz
2 2
with real coefficients and the so-called Pauli matrices σx , σy and σz .
PAULI MATRICES
! ! !
01 0 −i 1 0
σx = ; σy = ; σz = . (6.23)
10 i 0 0 −1
The multiplication table, commutators and anti-commutators of Pauli matrices are derived in
App. O. Since these relations will be useful for what follows I recommend to go through that
appendix at this point.
While we started with observables, that is with hermitian matrices, it can be shown that any 2 × 2
matrix, and not only Hermitian matrices, can be expressed as superposition of the identity and the
Pauli matrices with complex coefficients.
As an example, the ladder operators have the form
1 1
â = (σ̂x − i σ̂y ); â† = (σ̂x + i σ̂y )
2 2
and the Hamilton operator can be written as
1
Ĥ = ~ωσ̂z (6.24)
2
4 See Eq. 5.34 on p. 94 for a definition of the anti-commutator
6.6.2 Excursion: The Fermionic harmonic oscillator (Home study)

This section is a fun chapter. There is no harm if you skip to the next section, at least for the
moment. If you are still with me, lean back and enjoy! We will get a glimpse into the field of
super-symmetry, which plays a role in elementary particle physics.
Super-symmetry is important because of the zero point motion. We will see later that when
we quantize fields the zero point motion of the harmonic oscillator results in an infinite energy
even for free space. At first this does not seem like a big problem, because we can always add a
compensating energy. However, the theory of general relativity says that energy, which is a type of
mass in this theory, distorts space and therefore results in gravitational forces. So, when we consider
also gravitational forces, the zero-point energy is observable and we cannot simply get rid of it.
One attempt to get rid of the zero-point energy is to introduce for every harmonic oscillator also a
corresponding two-state system, so that their positive and negative zero point motions cancel out
exactly. This is the basic idea of super-symmetry. We will later talk about particles, Bosons and
Fermions, where the Fermions are particles that are represented by a two-state system in each point
in space and Bosons are particles that are represented by harmonic oscillators. But now we have
already gone way beyond what we learned sofar.
What we try to do here is to develop the two-state system analogously to the harmonic oscillator.
In the two-state system and in the harmonic oscillator we used creation and annihilation operators
to construct the eigenvalue spectrum. We have seen that a lot is similar but that commutators and
anti-commutators interchange their role.
Here we follow up on that analogy between two-state systems and the harmonic oscillator and
develop what is called the Fermionic harmonic oscillator. The main question we will ask is: What are
the coordinate and momentum operators that describe the two-state system.
We build the momenta and position operators from odd and even combinations of the creation
and annihilation operators roughly analogous to the ordinary harmonic oscillator.
r
~ †
χ̂ = i (â − â)
2
r
~ †
π̂ = (â + â)
2
Both are Hermitian operators

Let us work out the product table
~ ~ ~ ~
χ̂2 = ; π̂ 2 = ; π̂ χ̂ = −i σ̂z ; χ̂π̂ = i σ̂z ;
2 2 2 2
in order to derive the commutation and anti-commutation relations
[χ̂, χ̂]+ = ~ [χ̂, χ̂]− = 0

[π̂, π̂]+ = ~ [π̂, π̂]− = 0
[π̂, χ̂]+ = 0 [π̂, χ̂]− = −i ~σz
Thus we can express the Hamilton operator in the form
Eq. 6.24 ~ω i [π̂,χ̂]+ =0

Ĥ = σ̂z = ω[π̂, χ̂]− = i ωπ̂ χ̂
2 2
Now we see the analogy in full glory. It seems that all we need to do is to interchange plus and
minus signs.
harmonic Oscillator two-state system

Ladder operators [â, â† ]− = 1 [â, â† ]+ = 1
[â, â]− = [â† , â† ]− = 0 [â, â]+ = [â† , â† ]+ = 0
Hamilton operator Ĥ = 21 ~ω0 [â† , â]+ Ĥ = 12 ~ω0 [â† , â]−
q q
Momentum p̂ = mω20 ~ (â† + â) π̂ = ~2 (â† + â)
q q
Position x̂ = i 2mω ~
0
(â† − â) χ̂ = i ~2 (â† − â)
q p 0 q q
â = 2~mω 1
0
p − i mω 2~ x â = 2~ 1
π̂ − i 2~ 1
χ̂
(Anti)commutators [p̂, x̂]− = ~i [π̂, χ̂]+ = 0
[p̂, p̂]− = [x̂ , x̂]− = 0 [π̂, π̂]+ = [χ̂, χ̂]+ = ~
p̂ 2
Hamilton operator 2m + mω x̂
2 2
i ωπ̂ χ̂
The classical limit is really funny. If we let ~ go to zero, we find that interchanging π̂’s and χ’s
in a product still changes its sign. This is the so-called Grassman algebra. In the Grassman algebra
we work with numbers, but any interchange of two numbers changes the sign of the product.
Why do we call this two-state system a harmonic oscillator? The regular harmonic oscillator has
the property that the classical system has exactly one frequency. Let us therefore derive the classical
Hamilton equations.
∂H
χ̇ = = i ωχ ⇒ χ(t) = ei ωt A
∂π
∂H
π̇ = − = i ωπ ⇒ π(t) = ei ωt B
∂χ
(Note, that there seems to be a sign error in the last equation. This is not so, because we have
to interchange χ and π in the Hamiltonian before we can form the derivative. Due to the anti-
commutation relation for χ and π this introduces a sign change.) Thus the “positions” χ and
momenta perform oscillations with a single frequency just as the regular harmonic oscillator.
Chapter 7
Measurements
The main new feature of quantum mechanics as discussed so far was

to replace particles by wave packets. Thus make the transition from a
classical particle theory to a, still classical, field theory. The quantum
mechanical measurement process is what is really puzzling about quan-
tum mechanics. The rules are fairly clear, but there is still an ongoing
discussion about its meaning.
In classical mechanics, measurable quantities are associated with
functions A(q, p) of the coordinates and momenta– or alternatively the
velocities–in configuration space. A measurable quantity for a state is
obtained by evaluating the value of the function A for this state. In
quantum mechanics a measurable quantity is associated with an Her-
mitian operator.
7.1 Expectation values

Fig. 7.1: Niels Bohr, 1892-
Quantum theory does not predict the outcome of an experiment with 1987. Danish Physicist.
certainty. However, it predicts the mean value of a measurable quantity Nobel price 1922 for his in-
when the experiment is repeated sufficiently often. This is less of a vestigations of the struc-
limitation than it seems, because we may very well measure the spread of ture of atoms and the
the measured values. Hence, we can predict under which circumstances radiation emanating from
our prediction will be certain. them.
The mean value of a measured quantity is called expectation value.
The expectation value of an operator A and a quantum mechanical state |ψi is obtained as
hψ|Â|ψi
hAi = (7.1)
hψ|ψi
In the following we assume that the state |ψi is normalized, so that we can express the expectation
value simply as
hψ|ψi=1
hAi = hψ|Â|ψi
7.2 Certain measurements
If a state predicts a sharp value of an observable Â, the state is an eigenstate of Â.
113
114 7 MEASUREMENTS
Let us consider the square variation1 σ 2 of an observable from its expectation value in some
state. Given a state |ψi we define the deviation ∆Â from the expectation value of that state.
def
∆Â = Â − hψ|Â|ψi1̂ (7.2)
Note that ∆Â describes one operator. The operator depends on the state |ψi. The expectation value
in Eq. 7.2 is considered a fixed number. The square deviation is then the expectation value of (∆Â)2 .
Let us work out the expectation value of the square deviation
D 2 E
hψ|(∆Â)2 |ψi = ψ| Â − hÂi |ψ
* !2 +
hψ|Â|ψi
= ψ| Â − |ψ
hψ|ψi
!2
2 hψ|Â|ψi hψ|Â|ψi
= hψ|Â |ψi − 2hψ|Â|ψi + hψ|ψi
hψ|ψi hψ|ψi
!2
hψ|(∆Â)2 |ψi hψ|Â2 |ψi hψ|Â|ψi
= − (7.3)
hψ|ψi hψ|ψi hψ|ψi
of a measurable quantity in an experiment for a given wave function.
Using the expression Eq. 7.1 for the expectation value we find
Eq. 7.3
h(∆Â)2 i = hÂ2 i − hÂi2
This result reminds strongly of the classical expression for the square deviation σA2 of a quantity A.
We consider here an experiment, which contains some randomness. The measurement can produce
any of the possible values ai of A. The frequency of the outcome ai is Pi , which is the number of
times ai was the measured result relative to the total number of measurements. The frequency Pi
can be interpreted as probability of the outcome ai . The mean value of the measured quantity A is
def
X
hAi = Pi ai
i
and the square deviation is

X X
σA2 = h(A − hAi)2 i = Pi (ai − hAi)2 = Pi (ai2 − 2ai hAi + hAi2 )
i i
X X
= Pi ai2 −2hAi Pi ai +hAi2 )
i i
| {z } | {z }
hA2 i hAi
2 2
= hA i − hAi
After this little detour we return back to quantum mechanics.

Since the operator is Hermitian – it corresponds to a measurable quantity –, the condition for a
certain prediction without spread is
h(∆Â)2 i = 0
which implies
Eq. 7.3
h(∆Â)2 i = hψ|Â2 ψihψ|ψi − hÂψ|ψi2
⇒ hÂψ|Âψihψ|ψi = hÂψ|ψi hψ|Âψi
| {z } | {z } | {z }
hφ|φi hφ|ψi hψ|φi
1 germ: Quadratische Abweichung
7 MEASUREMENTS 115
def
Let us introduce |φi = Â|ψi. Schwarz’ inequality, which is derived below in 7.4, says that for
any two states |φi and |ψi the inequality
Eq. 7.4
hφ|φihψ|ψi − hφ|ψihψ|φi ≥ 0
holds and that the equal sign holds, if the two states |ψi and |φi are linear dependent, i.e. |φi = |ψia,
with some proportionality factor a.
For our problem with h(∆Â)2 i = 0, Schwarz’ inequality implies
Â|ψi = |ψia
We call such an equation an eigenvalue equation. The state |ψi is called eigenstate, and the value
a is called an eigenvalue. Thus we have shown that a sharp measurement is only possible for an
eigenstate of the corresponding operator.
7.2.1 Schwarz’ inequality
Let us now derive Schwarz’ inequality we used before.

SCHWARZ’ INEQUALITY
Schwarz’ inequality says that for two arbitrary states |φi and |ψi
hφ|φihψ|ψi − hφ|ψihψ|φi ≥ 0 (7.4)
and that the equality holds only when |φi and |ψi are linearly dependent, i.e. if |φi = |ψia with some
(complex) proportionality constant a.
The essence of Schwarz inequality is that a vector becomes shorter if we project something out
of it. The above relation can be written as

1 Eq. 7.4
hψ| 1̂ − |φi hφ| |ψi ≥ 0
hφ|φi
In order to grasp the meaning of Schwarz’ inequality consider regular vectors instead of quantum
mechanical states. For regular vectors Schwarz inequality says
a~2~b2 − (~
a~b)2 ≥ 0
Dividing this expression by the squared length of the two vectors yields
!2
a~ ~b
1− = 1 − cos2 (α) = sin2 (α) ≥ 0
a||~b|
|~
where α is the angle between the two vectors. Clearly, the squared sinus is always positive and it
vanishes for an angle of 0◦ and 180◦ , that is for parallel or antiparallel vectors. The analogy with
regular vectors may be instructive in the proof of Schwarz’ inequality for quantum mechanical states.
The proof of Schwarz inequality goes as follows:
116 7 MEASUREMENTS
PROOF of Schwarz inequality Eq. 7.4: Let us say we have two states |ψi and |φi. Out of the two states,
we construct one, |δi = |φihψ|ψi − |ψihψ|φi, that is orthogonal to |ψi. We express |φi as
h i 1
|φi = |δi + |ψihψ|φi
hψ|ψi
and evaluate its norm exploiting the orthogonality hψ|δi = 0

h i 1
hφ|φi = hδ|δi + hφ|ψihψ|ψihψ|φi
hψ|ψi2
hδ|δi
⇒ hφ|φihψ|ψi − hφ|ψihψ|φi =
hψ|ψi
|ψ><ψ|
|φ>
<ψ|ψ>
|ϕ>
|δ> |ψ>
<ψ|ψ>
1. Since the norm of every state is positive or zero, the right hand side is positive or zero.
2. The right hand side vanishes, when |δi = 0. In that case, however, we find from the equation defining
|δi that
hψ|φi
|φi = |ψi
hψ|ψi
Hence |φi and |ψi are linearly dependent if the equal sign holds.
q.e.d.
7.3 Eigenstates
We have learned that the uncertainty of a measurement vanishes, if the state is an eigenstate
of the operator corresponding to the measured quantity. Eigenstates play an important role in
measurements.
There are a few important points to know about eigenstates:
The eigenvalues of a Hermitian operator are real.

1.
PROOF: Let ψ be an eigenstate of the Hermitian operator Â with eigenvalue a.

Â=Â†
hψ|ψia = hψ|Â|ψi = hψ|Â† |ψi∗ = hψ|Â|ψi∗ = hψ|ψi∗ a∗
Because the norm of a state is real, we find that the eigenvalue a = a∗ is real, too.
q.e.d
The eigenstates of a Hermitian operator with different eigenvalues are orthogonal.

2.
7 MEASUREMENTS 117
PROOF: Let us consider two eigenstates |ψ1 i and ψ2 i of a Hermitian operator Â with different eigen-
values a1 and a2 .
Because Â is Hermitian we obtain
0 = hφ|Â − Â† |ψi
We continue by using the eigenvalue equations Â|ψ1 i = |ψ1 ia1 and Â|ψ2 i = |ψ2 ia2 .
0 = hψ1 |Â − Â† |ψ2 i

= hψ1 |Â|ψ2 i − hψ2 |Â|ψ1 i∗
= hψ1 |ψ2 ia2 − hψ2 |ψ1 i∗ a1∗
= hψ1 |ψ2 i[a2 − a1 ]
(Note that a1∗ = a1 , since Â is Hermitian.) If the eigenvalues are different this equation can only be
fulfilled if the scalar product vanishes, thus the two eigenstates are orthogonal.
q.e.d.
Any linear combination of eigenstates of an operator with the same eigenvalue is itself an
eigenstate with this eigenvalue.
3.
PROOF: We start with two eigenstates to Â with the same eigenvalue a
Â[|ψ1 ic1 + |ψ2 ic2 ] = [Â|ψ1 ic1 + Â|ψ2 ic2 ] = [|ψ1 iac1 + |ψ2 iac2 ] =
= [|ψ1 ic1 + |ψ2 ic2 ]a
q.e.d.
The eigenstates to the same eigenvalue can be orthogonalized. The number of orthogonal
eigenstates for a given eigenvalue is called the degeneracy of that eigenvalue.
4.
PROOF: This is a trivial statement, because any set of states can be orthogonalized. Here we show
a procedure to do this in practice. The method is called Gram-Schmidt orthogonalization.
We start out from a set of states {|ψi i} and orthogonalize them recursively. The resulting states {|φi i}
will not only be orthogonalized but also normalized, so that hφi |φj i = δi ,j .
n−1
X
|ξn i = |ψn i − |φj ihφj |ψn i
j=1
1
|φn i = |ξn i p
ξn |ξn i
In a first step, the orthogonalization step, we produce the states |ξn i. Only those states |φn i are required
that have already been constructed. We can assume that the states |φj i are normalized, because, in
the second step, the normalization step, the orthonormalized state |φi is obtained by scaling |ξn i. The
recursive equation can produce a zero state |ξn i = 0, which produces a divide-by-zero in the next step.
When this happens |ψn i is linearly dependent to the other states and must be excluded from the set.
Thus, the orthogonal set {|φj i} may have less elements than the initial set {|ψi i}. q.e.d.
The set of all eigenvalues of an operator Â is called the spectrum of Â. The spectrum can be partly
discrete and partly continuous. We denote the eigenstates with discrete eigenvalues with in integer
index |φi i and those from the continuous spectrum |φα i with a continuous index α.
118 7 MEASUREMENTS
We normalize the states that
hφi |φj i = δi ,j
hφα |φα′ i = δ(α − α′ )
hφi |φα i = 0
If the orthogonal system of eigenstates is complete, that is every state |ψi can be represented as
superposition of eigenstates φi ,
X Z
|ψi = |φi ici + dα|φα icα
i
then the Hermitian operator is called an observable. The coefficients are ci = hφi |ψi and cα =
hφα |ψi.
Important Note! In the following we will often use the sum alone even if the eigenvalue spectrum
may be completely or partly continuous. In this case, we always imply summation over all states of
the discrete part of the spectrum and integration over all states from the continuous spectrum.
We have already made contact with operators with discrete and with continuous eigenvalue
spectra. The Hamiltonian of the harmonic oscillator has a purely discrete eigenvalue spectrum. The
Hamiltonian for the free particle on the other hand has a purely continuous eigenvalue spectrum.
7.4 Heisenberg’s uncertainty relation
HEISENBERG’S UNCERTAINTY RELATION
Heisenberg’s uncertainty principle [19] says that two canonical conjugate quantities such as po-
sition and momentum cannot be determined simultaneously with absolute accuracy. Furthermore it
gives a lower limit
q q
~
hψ|(∆Q̂) |ψi hψ|(∆P̂ )2 |ψi ≥
2 (7.5)
2
for the certainty with which both quantities can be determined.
Here we used the operator ∆Â that describes the deviation of an operator Â from its mean value
hψ|Â|ψi is
def
∆Â = Â − hψ|Â|ψi (7.6)
The uncertainty principle can be generalized[20] to a general pair of observables Â and B̂ saying
q q
1

hψ|(∆Â)2 |ψi hψ|(∆B̂)2 |ψi ≥ hψ| [A, B] |ψi (7.7)
2
Two quantities are canonical conjugate of each other if they are both Hermitian and if their
commutator is [P̂ , Q̂] = ~i .
As shown earlier, we define the mean square deviation of an observable from the mean value of
an operator Q̂ in a state |ψi as
q q q
2 Eq. 7.6
hψ|(∆Q̂) |ψi = hψ|(Q̂ − hψ|Q̂|ψi) |ψi = hψ|Q̂2 |ψi − hψ|Q̂|ψi2
2
Heisenberg’s uncertainty relation says that one cannot prepare a system in a state such that two
observables have sharp values, unless their operator vanishes.
7 MEASUREMENTS 119
The relevance of Heisenberg’s uncertainty principle reaches further than only quantum mechanics.
It is important for optics and for signal processing in general. Translated into optics it says that we
can only make a light-pulse shorter at the cost of its frequency resolution. A femtosecond light pulse
will have a wide frequency distribution. In signal processing it says that, if our frequency band for
transmission is limited, this also limits the sharpness of the transmitted signal in the time domain.
The underlying principle is that the Fourier transform of a wave packet, becomes wider as we
decrease the width of the wave packet. This is an intrinsic property of a Fourier transform. Two
conjugate variables can always be considered as the variables in a Fourier transform. For example
Z Z Z
dp dp dp i px
hx|ψi = hx| |pihp| ψi = hx|pihp|ψi = e ~ hp|ψi
| {z } 2π~ 2π~ | {z } 2π~ | {z }
ψ(x)
| {z } i px ψ(p)
1̂ e~
Since time and frequency ω are also related by a Fourier transform, just like position and momenta,
it is clear that the results of Heisenberg’s uncertainty principle can be translated into the above
problems in the time-frequency domain.
PROOF: The proof of Heisenberg’s uncertainty principle goes as follows: For two observable Â and B̂ let us
define the corresponding deviation as
def def
∆Â = Â − hÂi and ∆B̂ = B̂ − hB̂i
Next, we introduce two states, namely

def def
|φA i = ∆Â|ψi and |φB i = ∆B̂|ψi
We express the product of the deviations by the states |φA i and |φB i. Then we exploit Schwarz’ inequality
Eq. 7.4 to obtain a lower bound.
∆Â=∆Â†
hψ|(∆Â)2 |ψihψ|(∆B̂)2 |ψi = hψ|(∆Â)† (∆Â)|ψi hψ|(∆B̂)† (∆B̂)|ψi
| {z } | {z } | {z } | {z }
hφA | |φA i hφB | |φB i
Eq. 7.4
= hφA |φA ihφB |φB i ≥ |hφA |φB i|2
2

= hψ|(∆Â)† (∆B̂)|ψi
| {z } | {z }
hφA | |φB i
2
hψ| 1 [∆Â, ∆B̂]− + 1 [∆Â, ∆B̂]+ |ψi

= (7.8)
2 2
We wrote the product in terms of commutator and anti-commutator, because one can show that the expec-
tation value of the commutator is purely imaginary, and that of the anticommutator is purely real. This fact
will be useful in the following and therefore we derive it here: We begin with
†
hψ|∆Â∆B̂|ψi = hψ| ∆B̂ † ∆Â† |ψi = hψ|∆B̂ † ∆Â† |ψi∗ = hψ|∆B̂∆Â|ψi∗
where we exploited only that the two operators are hermitean. Now, we can show that the expectation values
of commutator and anticommutator are either purely real or purely imaginary.
hψ|[∆Â, ∆B̂]− |ψi = hψ|∆Â∆B̂|ψi − hψ|∆B̂∆Â|ψi = hψ|∆Â∆B̂|ψi − hψ|∆Â∆B̂|ψi∗

= 2iIm hψ|∆Â∆B̂|ψi (7.9)
hψ|[∆Â, ∆B̂]+ |ψi = hψ|∆Â∆B̂|ψi + hψ|∆B̂∆Â|ψi = hψ|∆Â∆B̂|ψi + hψ|∆Â∆B̂|ψi∗

= 2Re hψ|∆Â∆B̂|ψi (7.10)
...cont’d
120 7 MEASUREMENTS
cont’d...
With Eqs. 7.9,7.10 we can express the absolute square by the squares of real and imaginary part.
2
2 2 Eq. 7.8 1 1
hψ|(∆Â) |ψihψ|(∆B̂) |ψi = hψ| 2 [∆Â, ∆B̂]− |ψi + hψ| 2 [∆Â, ∆B̂]+ |ψi

2 2
hψ| 1 [∆Â, ∆B̂]− |ψi + hψ| 1 [∆Â, ∆B̂]+ |ψi
Eqs. 7.9,7.10
= 2 2
2 2
hψ| 1 [Â, B̂]− |ψi + hψ| 1 [∆Â, ∆B̂]+ |ψi

= (7.11)
2 2
In the last expression we exploited that the commutator of the deviations is the same as that of the original
operators. A similar simplification is not possible for the anticommutator.
The extended form of the uncertainty principle from Eq. 7.11 is due to Schrödinger[21]. It is probably the
mosts stringent form of the uncertainty principle.
By dropping the term with the anticommutator, we obtain the even lower bound specified in Eq. 7.7.
q q
1
hψ|(∆Â)2 |ψi hψ|(∆B̂)2 |ψi ≥ hψ|[Â, B̂]− |ψi
2
~
and by insertion of the commutator of canonical conjugate variables [P, Q]− = i one obtains the special form
of Heisenberg Eq. 7.5.
q.e.d
7.5 Measurement process

Now we are prepared to investigate the measurement process. The measurement process in quantum
mechanics exhibits some surprising elements. I will approach this topic as follows. Firstly I will describe
the so-called Kopenhagen interpretation of the measurement process, which goes back to Niels Bohr.
The resulting rules can easily be memorized and they chapter everything needed for everyday use of
quantum mechanics.
The Kopenhagen interpretation had some ad-hoc definitions that were not quite satisfactory. One
of these problems leads to puzzles such as the one of Schrödinger’s cat. The theory of decoherence
digs deeper into the measurement process and arrives at a conceptionally satisfactory description of
the measurement process.
7.6 Kopenhagen interpretation

The quantum mechanical measurement process can be described by the following postulates, which
are part of the conceptual basis of quantum mechanics.
If we perform a measurement of an observable Â, that is denoted by a Hermitian operator, the
outcome will be one value a. After the measurement, we know the value of a with certainty. Hence
the state must be in an eigenstate of the operator Â with eigenvalue a. This tells us a few important
lessons:
• Spectrum of measurable values: The only values, one can possibly obtain by measuring a
quantity related to an operator Â, are its eigenvalues. If the spectrum of Â is discrete, one can
only obtain certain discrete values.
• Probabilistic interpretation of quantum mechanics: Quantum mechanics does not predict

which one of the eigenvalues of Â is measured in the actualPexperiment, but it predicts the
probability with which one obtains a certain value by Pa = i hΨ|χi ihχi |Ψi, where the sum
runs again over all eigenstates |χi i with eigenvalue a.
7 MEASUREMENTS 121
• Collapse of the wave function: During the measurement, the wave function “collapses”. The
measurement process changes the wave function into an eigenstatePof Â. If the wave function
i |χi ihχi |Ψi
before the measurement was |Ψi, after the measurement it is √P 2
, where |χi i are
i |hχi |Ψi|
normalized eigenstates of Â to the eigenvalue of a.
Let us describe the process of performing a laboratory experiment including a measurement in

more detail:
1. Set up the experimental conditions, that is define the Hamilton operator Ĥ(t)
2. Define the initial conditions. A certain state |ψ(t0 )i is prepared
3. The system is allowed to evolve. The dynamics is governed by Schrödinger’s equation. At
some point in time, t1 the actual measurement takes place. Just before the measurement, the
state is |ψ(t1− )i. The minus sign indicates that we consider a time arbitrary close to that of
the measurement, but still before the actual measurement takes place.
4. The measurement is performed. We obtain one of the possible outcomes ai . Even though we
could conceive a larger variety of outcomes, quantum mechanics says that we will obtain only
those numbers, that correspond to eigenvalues of the measured quantity Â. If we repeat the
experiment, we do not obtain the same value for A each time the experiment is performed, but
we obtain the eigenvalues of Â in a seemingly random manner. Thus, the experiment is not
deterministic. While we cannot describe the experiment by a single outcome, we can determine
the frequencies Pi of outcomes ai . The frequency is the ratio of experiments with a particular
outcome ai . Quantum theory predicts the frequencies as
Pi = |hχi |ψi|2
where |χi i is a normalized eigenstate of Â with eigenvalue ai , that is Â|χi i = |χi iai . If
the eigenvalue is degenerate, the frequency is obtained as the sum over all states with that
eigenvalue.
5. After each measurement, we can perform additional measurements of the same or different
observable. The system behaves as if the state would be changed during the measurement
abruptly from |ψ(t1− )i to
hχi |ψ(t1− )i
|ψ(t1+ )i = |χi i q
|hχi |ψ(t1− )i|2
The term in the denominator ensures that the new state is again normalized. Thus the state
|ψ(t1− )i is projected onto the eigenstate of Â that was the outcome of the experiment.
If the eigenvalue ai is degenerate we need to include all projections onto eigenstates with the
same eigenvalue, i.e.
P −
+ j :Â|χj i=|χj iai |χj ihχj |ψ(t1 )i
|ψ(t1 )i = P
q
− 2
j :Â|χj i=|χj iai |hχj |ψ(t1 )i|
Repeated measurements
If the measurement of Â is repeated twice on the same system, we obtain the same result, namely
a, with certainty . Something unexpected happens, if we measure two different observables Â and B̂
one after the other. Unless the two operators commutate, implying that there is a set of common
eigenstates, every measurement of the variable B̂ destroys the information of Â and vice versa. If we
measure Â twice on the same system, the result depends if we measured a quantity B̂ between the
two measurements of Â or not.
122 7 MEASUREMENTS
7.7 Decoherence
The Kopenhagen interpretation is a completely satisfactory working solution to the measurement
process. The predictions are undisputed. Nevertheless, it has one unsatisfactory aspect: While the
dynamics is described by the Schrödinger equation, a measurement causes an irreversible collapse of
the wave function that is in direct conflict with the Schrödinger equation.
Decoherence theory investigates the measurement process in more detail and describes how seem-
ingly classical behavior results from the dynamics of the measurement process in full accord with the
Schrödinger equation. The “trick” is to consider our system as an open system, that is in contact
with an environment. This interaction has to fulfill certain criteria in order to introduce irreversibility,
such that a description with classical alternatives for the outcome of the measurement results. An
excellent introduction to decoherence theory is given by Zurek[22].
Decoherence divides the measurement into several parts, each of which can be attributed to
a well-defined physical process. In the following, I will not present a formal derivation, but only
summarize my view on it.
P
• Let us start out with an initial wave function |ψ1 i = σ |σiασ for our system, which may be in
a superposition of eigenstates of the observable Â that we wish to measure, i.e. Â|σi = |σiσ.
Messprozess
• Transfer of the information: We bring the system described by the wave function |ψi in
contact with a detector d, which can be in any of the states |dσ i. The interaction must be
exactly such that, after the measurement, the combined system of measured system and the
detector is
X
|ψ2 i = |σ, dσ iασ
σ
Now the detector contains all the information of the measured system, but it is itself in a super-
position of states. One says that the detector is entangled with the system to be measured.
An important concept, often missed, is that a superposition of states is different from a set of
states with some probabilities, i.e. an ensemble of states. A superposition of states is still a
single wave function. A set of states describes different alternatives for the wave function.
• Decoherence: An integral condition for a detector is that it is coupled to an environment.
There must be a small but non-trivial interaction of the detector with the environment. If we
then investigate our local system, it behaves just like an ensemble of states with probabilities
equal to Pσ = |ασ |2 . This is an important step, because, from now on, our description is one
that gives each eigenvalue of the observable a classical probabilitie.
• Collapse: From now on, the situation is analogous to what happens with a classical system
for which we do not have the full information. We describe the system with a probability
distribution. By performing the measurement we gain additional information, so that we choose
a better description by replacing the possible outcomes of the measurement by only one of the
alternatives, namely the one that has been measured. This is the collapse of the probability
distribution.
7 MEASUREMENTS 123
7.8 Difference between a superposition and a statistical mixture

of states
The essence of the measurement process is that a superposition of states is converted into a weighted
mixture of states. In the Kopenhagen interpretation this is done by a postulate, while in decoherence
theory this is due to a physical process that results from a specific coupling of the measurement
aparatus to an environment.
The distinction between a superposition of states and an mixture of states is however a bit subtle,
which is the reason for discussing it in greater detail.
7.8.1 Mixture of states

Let us first explain a statistical mixture: Imagine that we do not know the state of the system. In
order to describe the situation, we may provide several guesses for the state and attribute a probability
to each of them. We may also select a complete set of eigenstates |ai i of an observable Â and provide
each with a probability.
In order to obtain the expectation value of the observable Â we can weight the expectation value
of each state with the probability of the state, i.e.
X X
hAi = Pi hai |Â|ai i = Pi ai
i i
We obtain the weighted average of the eigenvalues of Â

However, we have accidentally considered the expectation value of a special observable, namely
the one which provided the states to describe our mixture. For another observable B̂, we obtain
X
hBi = Pi hai |B̂|ai i
i
The result looks similar. Instead of the eigenvalues of B we have to use the expectation values of
B̂ with the set of states from our mixture. The probabilities are the same as those that we used to
determine the expectation value of the observable Â.
7.8.2 Superposition of states

Let us now compare these equations for the expectation value with those from a superposition of
states.
Westart with a quantum-mechanical state |Ψi. If we want to measure a certain observable Â it
is convenient to expand this state into (orthonormal) eigenstates |ai i of the operator Â, i.e.
X X
|ψi = |ai ihai | ψi = |ai i ci
|{z}
i i
| {z } hai |ψi
1̂
This is a superposition of states |ai i. The components of the wave function are coherent.
The expectation value of the observable Â for such a superposition is
X X X X (A)
hψ|Â|ψi = ci∗ cj hai |Â|aj i = ci∗ cj hai |aj iaj = ci∗ ci ai = Pi ai
i ,j i ,j i i
Thus the expectation value can be expressed as a sum over the eigenvalues of A with probabilities
Pi = ci† ci = hψ|ai ihai |ψi for each eigenvalue.
124 7 MEASUREMENTS
This equation is basically identical to the one obtained for a mixture of states. However, let us
now look at the expectation value of some other operator B̂.
X X X
hψ|B̂|ψi = ci∗ cj hai |B̂|aj i = ci∗ ci hai |B̂|ai i + ci∗ cj hai |B̂|aj i
i ,j i i 6=j
This expression differs from the expression for the mixture by the presence of the off-diagonal terms of
the double sum. These off-diagonal terms are a sign of a coherent superposition of states. Coherent
(A)
means the following: For a given set of probabilities Pi , we can determine the coefficients
√
c i = P i e i ϕi
only if we also specify the relative phases ϕi of the states.

While a mixture of states is uniquely defined by a set of states, say {|ai i} and their probabilities
Pi , a superposition carries additional information, namely the phase information.
If we average the expectation value for B̂ over all states that only differ by the relative phases,
the off-diagonal terms cancel out, and we are left with an expression for expectation values that is
identical to that for a mixture. This is the essence of the so-called random-phase approximation.
The random phase approximation converts a superposition of states into a mixture of states.
It may be instructive to consider the expectation value for a superposition of states in terms of
eigenvalues of the observable B. Let us simply introduce the eigenstates |bi i and the corresponding
eigenvalues bi .
X X X (B)
hψ|B̂|ψi = hψ|bi ihbi |B̂|bj ihbj |ψi = |hψ|bi i|2 bi = Pi bi
i ,j i i
While we can express the expectation value of B as a weighted sum of eigenvalues, the probabilities
Pi(B) differ from the probabilities Pi(A) , we have used before to obtain the expectation value of A.
7.8.3 Implications for the measurement process

We have seen that there is a fundamental difference between a mixture and a superposition of states.
An instrument can only be considered a detector for an observable A, if it introduces an interactions
with the environment in such a way that the coherent superposition of states is transformed into a
mixture of states of the subsystem to be measured.
Despite the fact that we arrive at a statistical mixture of states for our subsystem, there is
still a single quantum state for the universe, that is for the measured system, the detector and its
environment taken together. However, if we consider this state only in a subsystem, it has the form
of a statistical mixture of states from the
Chapter 8
Dynamics
8.1 Dynamics of an expectation value

Here we investigate the time evolution of an expectation value
d d Â
hψ|Â|ψi = h∂t ψ|Â|ψi + hψ|( )|ψi + hψ|Â|∂ψi
dt dt
We use the time-dependent Schrödinger equation1 i ~∂t |ψ(t)i =
Ĥ|ψ(t)i to obtain the derivatives of bras and kets
i ~∂t |ψi = Ĥ|ψi

−i ~∂t hψ| = hψ|Ĥ
Thus we obtain
d d Â
i~ hψ|Â|ψi = hψ|(−Ĥ Â + ÂĤ)|ψi + hψ|i ~( )|ψi
dt ! dt

d Â
= hψ| Â, Ĥ + i ~ |ψi (8.1)
dt Fig. 8.1: Paul Ehrenfest,
1880-1933.
This provides us with an important rule:
COMMUTATOR AND CONSERVATION LAWS
If an observable is not explicitly time dependent and if it commutates with the Hamilton operator,
the expectation value of that observable does not change with time.
As a special case we find the energy conservation: the expectation value of the Hamilton operator,
the energy, is conserved if the Hamilton operator is not explicitly time dependent.
8.2 Quantum numbers

From the above relation Eq. 8.1 follows the following:
The expectation value of a observable, that commutates with the Hamilton operator is invariant.
Such observables are called constants of motion.
1 The rule for the bra is obtained from the rule for the ket by forming a matrix element with an arbitrary bra hφ|,
and using the rules for interchanging bras and kets in a matrix element, i.e. hψ|Â|φi = hφ|Â† |ψi
125
126 8 DYNAMICS
Definition 8.1 COMPATIBLE OPERATORS

Two operators are called compatible if their commutator vanishes.
We see immediately, that for a time-independent Hamilton operator, the energy is a constant of
motion, because the Hamilton operator commutates with itself.
If an operator Â commutates with the Hamilton operator, we can find a basis of states which are
eigenstates both of Â and of Ĥ. The eigenvalues of such an operator Â are called good quantum
numbers.
A basis of common eigenstates of Ĥ and Â, which have a vanishing commutator, is constructed
as follows: We start with the eigenstates of the Hamilton operator.
• If an eigenvalue of the Hamiltonian is non-degenerate, the corresponding eigenstate |Ψi is also

an eigenstate of the operator Â. This is shown as follows:
Ĥ|ψi = |ψiE
⇒ ÂĤ|ψi = Â|ψiE
[H,A]=0
⇒ Ĥ Â|ψi = Â|ψi E (8.2)
Thus, the state Â|ψi is also an eigenstate of the Hamilton operator with the same eigenvalue.
Because the eigenvalue is non-degenerate, the state Â|ψi and the state |ψi must be linear
dependent. Hence
Â|ψi = |ψia
with some scale factor a. Because the above equation, is nothing but an eigenvalue equation,
the scale factor a is the corresponding eigenvalue of Â. Thus, the state |ψi is an eigenstate to
the operator Â.
• If the eigenvalue E of the Hamiltonian is m-fold degenerate, we can use one of the eigenstates
from the corresponding set of degenerate states to form a sequence of states
|ψj i = Âj |ψi
which are all eigenstates of the Hamiltonian with the same eigenvalue E. The proof is analogous
to the non-degenerate case as shown in Eq. 8.2.
We determine a linear independent sub-set of the states |ψ̄j i for example by Gram-Schmidt
orthogonalization. While there seems to be an infinite number of states |ψj i in the sequence,
there are only m linear independent eigenstates with this eigenvalue of the Hamilton operator.
Thus the linear independent set of states |ψ̄j i resulting from the Gram-Schmidt orthogonaliza-
tion has at most m members.

A state Â|ψ̄j i can itself be expressed as a superposition of the states in the set |ψ̄k i; k = 1, . . . , m
of degenerate states.
X
Â|ψ̄j i = |ψ̄k ick,j
k
X
⇒ hψ̄i |Â|ψ̄j i = hψ̄i |ψ̄k i ck,j = ci ,j
| {z } | {z }
k
=Ai,j δi,k
The matrix elements ci ,j are equal to the matrix elements Ai ,j of the operator Â in the basis
{|ψ̄j i}. Therefore, we denote it in the following as Ai ,j = ci ,j .
Since the matrix A is hermitian, it can be diagonalized, which yields the eigenvectors c~n and
eigenvalues an .
A c~n = c~n an
8 DYNAMICS 127
The eigenvectors can then be used to create the eigenstates |φn i of Â

def
X
|φn i = |ψ̄j icj,n
j
which fulfill the eigenvalue equation

def
Â|φn i = |φn ian
and which are simultaneously eigenstates of the Hamiltonian.
Thus it has been shown that it is possible to construct a complete orthonormal basisset of states,
which are simultaneously eigenstates of Ĥ and Â, if the two operators commutate.
A complete set of compatible operators is a set of compatible operators, with the property that the
set of their eigenvalues completely characterizes each state of a complete basis in Hilbert space.
An example are the states of the Hydrogen atom. They are characterized completely by the three
quantum numbers n, ℓ, m through the eigenvalue equations
Z
Ĥ|ψℓ,m,n i = |ψℓ,m,n i H
2n2
L̂z |ψℓ,m,n i = |ψℓ,m,n i~m
~ˆ2 |ψ
L ℓ,m,n i = |ψ i~2 ℓ(ℓ + 1)
ℓ,m,n
where Z is the atomic number, Ĥ is the Hamilton operator of the hydrogen atom, H is the Hartree
energy unit, Z is the atomic number and L ~ˆ is the angular-momentum operator. n is the principal
quantum number, ℓ is the main angular-momentum quantum number and m is the magnetic quantum
number.
Another example is the wave vector of a plane wave. Plane waves form a complete set of states.
They are eigenstates of the Hamiltonian for a free particle and they are characterized by the eigenvalue
equation.
p~ˆ|ψ~k i = |ψ~k i~~k
8.3 Ehrenfest Theorem

If we apply the above rule to the position and momentum operators, we find that the expectation
values for these quantities evolve with the classical equations of motion. The Ehrenfest theorem
plays an important role to justify the argument that we used in the second chapter to derive the
Schrödinger equation. It says, that if a wave packet remains localized, it behaves like a classical
particle. We need to read this sentence carefully, though. The wave packet can still split or spread,
and then the correspondence between classical mechanics and quantum theory is not as clear cut.
EHRENFEST THEOREM
The Ehrenfest Theorem[23] says that the time evolution of a quantum mechanical expectation
values is identical to that of a classical variable under the expectation value of the forces.
d
hψ|x̂|ψi = hψ|∂p H(p̂, x̂)|ψi
dt
d
hψ|p̂|ψi = hψ| − ∂x H(p̂, x̂)|ψi (8.3)
dt
These equations compare to Hamilton’s equations Eq. 3.6 on p. 29
128 8 DYNAMICS
Let us first consider a special Hamiltonian

p̂ 2
Ĥ = + V (x̂)
2m
The general proof will be shown later in this section.
We obtain for the position operator
d Eq. 8.1 i p̂ 2 i
hψ|x̂ |ψi = hψ| [ + V (x̂), x̂]|ψi = hψ| [p̂ 2 , x̂]|ψi
dt ~ 2m 2m~
i
= hψ| (p̂p̂x̂ −p̂x̂ p̂ + p̂x̂ p̂ −x̂ p̂ p̂)|ψi
2m~ | {z }
=0
i
= hψ| (p̂[p̂, x̂] + [p̂, x̂]p̂)|ψi
2m~
p̂
= hψ| |ψi
m
For the momentum operator we find similarly the equation corresponding to the Hamilton equation
for the momentum2 .

d Eq. 8.1 i p̂ 2
hψ|p̂|ψi = hψ| + V (x̂), p̂ |ψi
dt ~ 2m
i
= hψ| [V (x̂), p̂]|ψi
~ 
 Z 
Eq. 6.11  
= 
 |ψi
hψ| V (x̂), dx |xi∂x hx|
 | {z }
i
~
p̂
Z
V (x̂)|xi=|xiV (x)
= dx hψ|xi [V (x), ∂x ] hx|ψi
Z
= − dx hψ|xi [∂x , V (x)]hx|ψi
| {z }
(∂x V )
= −hψ|(∂x V (x̂))|ψi
The meaning of ∂x V (x̂) is the following: First form the derivative (∂x V (x)) of the function V (x).
The parenthesis shall indicate that the derivative does not act on any following functions. Then
replace the argument x by the operator x̂.
We have shown the Ehrenfest theorem for a special form of the Hamilton operator. In order to
derive the Ehrenfest theorem in its most general form, we need the equation of motion for operators
derived earlier in Eq. 8.1
d Eq. 8.1 i
hψ|x̂|ψi = hψ| [Ĥ, x̂]|ψi
dt ~
d Eq. 8.1 i
hψ|p̂|ψi = hψ| [Ĥ, p̂]|ψi
dt ~
and the commutator expressions
~
[H(p̂, x̂), x̂] = ∂p H(p̂, x̂)
i
~
[H(p̂, x̂), p̂] = − ∂x H(p̂, x̂)
i
2 In the derivation, note that ∂ V (x) = (∂ V (x)) + V (x)∂ . This is an operator expression that must be read as if
x x x
it were applied to a state following the expression on the right side.
8 DYNAMICS 129
which are shown as follows:

130 8 DYNAMICS
Proof: Here we show that the operator equation

~
[f (p̂, x̂), x̂] = ∂p f (p̂, x̂ ) (8.4)
i
~
[f (p̂, x̂ ), p̂] = − ∂x f (p̂, x̂ ) (8.5)
i
is valid for any function of the canonical conjugate operators x̂ and p̂, that is those having the commutator
relation [p̂, x̂ ] = ~i
A function of an operator is defined by its power-series expansion. Therefore, any operator function f (p̂, x̂ ) =
P i j k l
α cα gα (p̂, x̂) can be expressed by a sum of weighted powers gα (p̂, x̂) = x̂ p̂ x̂ p̂ . . ., where i, j, k, l . . . are
non-negative integers. The order of the power is nα = i + j + k + l + . . .. It is important to remember,
that the order of the operators p and x in the function f matters, and this order must be observed when
performing the derivatives. In the following, we show first that the above relation holds for arbitrary powers
gα (p̂, x̂) and than derive from that the equations above for general operator functions f (x̂ , p̂).
The proof for powers gα (p̂, x̂) proceeds by induction.
1) We start with gα (p̂, x̂ ) = p̂ 1 and gα′ (p̂, x̂ ) = x̂ 1 , which are the only powers of order one.
~ ~
[p̂, x̂ ] = = [∂p p̂] (8.6)
i i
~ ~
[x̂, p̂] = − = − [∂x x̂] (8.7)
i i
Thus, the result is obviously true for powers of order one.
2) Now we assume that the relations Eq. 8.4 and Eq. 8.5 are true for powers of order n − 1 in x̂ and p̂ and
derive from that the relation for powers of order n. All powers of order n can be written as one of the four
possibilities x̂ g(p̂, x̂), g(p̂, x̂ )x̂ , p̂g(p̂, x̂ ), g(p̂, x̂ )p̂, where g is a power of x̂ and p̂ of order n − 1.

Eq. 8.4 ~
[x̂g(p̂, x̂ ), x̂ ] = x̂g(p̂, x̂ )x̂ − x̂ 2 g(p̂, x̂ ) = x̂[g(p̂, x̂), x̂] = x̂ ∂p g(x̂, p̂)
i
~
= ∂p x̂g(x̂, p̂)
i
Eq. 8.4 ~
[g(p̂, x̂)x̂ , x̂ ] = g(p̂, x̂)x̂ 2 − x̂g(p̂, x̂)x̂ = [g(p̂, x̂ ), x̂ ]x̂ = ∂p g(x̂ , p̂) x̂
i
~
= ∂p g(x̂, p̂)x̂
i
=0
z }| {
[g(p̂, x̂ )p̂, x̂ ] = g(p̂, x̂)p̂x̂ −g(p̂, x̂ )x̂ p̂ + g(p̂, x̂)x̂ p̂ −x̂g(p̂, x̂)p̂

Eq. 8.4 ~ ~
= g(p̂, x̂)[p̂, x̂ ] + [g(p̂, x̂), x̂]p̂ = g(p̂, x̂) + ∂p g(p̂, x̂) p̂
i i
~
= ∂p g(p̂, x̂)p̂
i
=0
z }| {
[p̂g(p̂, x̂ ), x̂ ] = p̂g(p̂, x̂)x̂ −p̂x̂g(p̂, x̂ ) + p̂x̂g(p̂, x̂ ) −x̂ p̂g(p̂, x̂ )

Eq. 8.4 ~ ~
= p̂[g(p̂, x̂ ), x̂ ] + [p̂, x̂ ]g(p̂, x̂ ) = p̂ ∂p g(p̂, x̂) + g(p̂, x̂ )
i i
~
= ∂p p̂g(p̂, x̂)
i
Thus we have shown that the relation Eq. 8.4 carries over from a power of order n to the next higher order
n + 1.
The proof for Eq. 8.5, the commutator with p̂, proceeds analogously. We may even exploit Eq. 8.4 is
converted into Eq. 8.5 by an exchange of p̂ and x̂ and sign-change of ~.
From the commutator relationsa follows that
X X
[ cα gα (x̂, p̂), x̂ ] = cα [gα (x̂ , p̂), x̂]
α α
which completes the proof for a general function of the operators x̂ and p̂. q.e.d.
a If Â, B̂, Ĉ are operators and c is a number the commutator relations [c Â, B̂] = c[Â, B̂] and [Â + B̂, Ĉ] = [Â, Ĉ] +
[B̂, Ĉ] hold.
8 DYNAMICS 131
8.4 Particle conservation and probability current

As the system evolves, the probability to find the system anywhere in space does not change. This
fundamental conservation law, the particle conservation law, led to the requirement that the Hamil-
tonian must be Hermitian.3 Here we investigate the probability for the particle to be at a given point
or in a given finite volume respectively. The probability to find the system in a given volume can
change only if the particle appears somewhere else. This can be described by the probability that
particle flow through the boundary of the volume. In this section we will develop an expression for
the probability flux.
The condition that the number of particles in a given volume Ω changes only when particles flow
through the surface ∂Ω of the volume, can be written as follows:
Z I
∂t 3
d r P (~r, t) = − dA~ J(~
~ r , t)
Ω ∂Ω
Here P (~r, t) is the probability density and J(~ ~ r , t) is the probability current. Note that per convention
~
the surface element vector d A always points outward. We use Gauss Theorem to formulate this
condition in differential form of the conservation law.
Z Z
d 3 r ∂t P (~r, t) = − d 3 r ∇ ~ r , t)
~ J(~
Ω
Z
d 3 r ∂t P (~r, t) + ∇ ~ r , t) = 0
~ J(~
Ω
∂t P (~r, t) + ∇ ~ r , t) = 0
~ J(~ (8.8)
We have used here that if the integral vanishes for any arbitrary volume Ω, the integrand must vanish
anywhere in space.
Let us evaluate the probability of finding the particle in the volume Ω.
Z
∂t d 3 r Ψ∗ (~r, t)Ψ(~r, t)
Z Ω h i
= d 3 r ∂t Ψ∗ (~r, t) Ψ(~r, t) + Ψ∗ (~r, t) ∂t Ψ(~r, t)
Ω
Z h ~2 2 i
i ~ 2 + V (~r) Ψ∗ (~r, t)) − Ψ∗ (~r, t) − ~ ∇ ~ 2 + V (~r) Ψ(~r, t)
= d 3 r Ψ(~r, t) − ∇
~ Ω 2m 2m
Z h i
−i ~ ~ 2 Ψ∗ (~r, t) − Ψ∗ (~r, t)∇
~ 2 Ψ(~r, t)
= d 3 r Ψ(~r, t)∇
2m Ω
Z h
−i ~ ~ Ψ(~r, t)∇Ψ ~ ∗ (~r, t) − ∇Ψ(~ ~ r , t) ∇Ψ ~ ∗ (~r, t)
= d 3r ∇
2m Ω
i
−∇ ~ Ψ∗ (~r, t)∇Ψ(~
~ r , t) + ∇Ψ ~ r , t)
~ ∗ (~r, t) ∇Ψ(~
Z h i
−i ~ ~ Ψ(~r, t)∇Ψ~ ∗ (~r, t) − Ψ∗ (~r, t)∇Ψ(~
~ r , t)
= d 3r ∇
2m Ω
3 Particle conservation requires that ∂t hψ|ψi = 0.
0 = ∂t hψ|ψi = h∂t ψ|ψi + hψ|∂t ψi
Now we use the Schrödinger equation
i i
i~∂t |ψi = Ĥ|ψi ⇒ |∂t ψi = − Ĥ|ψi ⇒ h∂t ψ| = h∂t ψ|Ĥ†
~ ~
and insert into the above equation
i i
0 = h∂t ψ|ψi + hψ|∂t ψi = hψ|Ĥ† |ψi − hψ|Ĥ|ψi = hψ|Ĥ† − Ĥ|ψi
~ ~
Since this equation holds for any wave function |ψi it follows that Ĥ† − Ĥ has only zero eigenvalues and therefore is
the zero operator. hence Ĥ† = Ĥ, which is the condition that the Hamiltonian is hermitian.
132 8 DYNAMICS
Let us define4 now the current density

PROBABILITY-CURRENT DENSITY
h i h ˆ i
~ r , t) = i ~ Ψ(~r, t)∇Ψ
J(~ ~ r , t) = Re Ψ∗ (r,~t) p~ ψ(~r, t)
~ ∗ (~r, t) − Ψ∗ (~r, t)∇Ψ(~ (8.9)
2m m0
Note that the current density has a different functional form for other Hamiltonians.5
The current density could also have been derived from the Noether theorem as a consequence
of gauge symmetry. Gauge symmetry implies that a multiplication of the wave functions with a
complex phase factor ei φ with a real angle φ does not change the physics. More precisely stated,
the action is invariant with respect to a multiplication with a phase factor. The Noether theorem is
applied for the Klein-Gordon field in ΦSX:Elektrodynamik.
8.5 Schrödinger, Heisenberg and Interaction pictures

8.5.1 Schrödinger picture
We will learn several representations of wave functions and operators. These are the Heisenberg
picture and the Interaction picture. In contrast, the conventional picture is the Schrödinger picture.
In the Schrödinger picture the operators are time independent, except for an explicit time dependence
such as a time dependent potential, and the dynamics of the system is described by a differential
equation for the wave function, the Schrödinger equation.
i ~∂t |ψS (t)i = ĤS |ψS (t)i
The subscript S indicates that the wave functions and operators are given in the conventional picture,
the schrödinger equation. The expectation value is given by
hAi = hψS |ÂS |ψS i
8.5.2 Heisenberg picture

In order to introduce the different pictures, we will require the concept of a so-called propagator 6 .
In this section we limit ourselfes to time-independent Hamiltonians. The more general case is
discussed in section H on p. 275.
Formally, we can directly solve the Schrödinger equation for a time-independent Hamiltonian in
the following way
|ψS (t)i = Û(t)|ψS (0)i. (8.10)
where the propagator fulfills the equation
i ~∂t Û(t) = ĤS Û(t) (8.11)
which is proven by inserting Eq. 8.10 into the time-dependent Schrödinger equation.
4 This is actually the definition of the particle current.The usual approach is to convert the time derivative of the
Probability density into the divergence of a vector field. If this is successful, we can claim that the particle number is
conserved and the vector field is the corresponding current density of those particles.
5 For a particle in an electric and magnetic field, so that H = (~ ~ ))2 /2m + eΦ(r ), the current density has the
p − q A(r
h i
~ i~ ~ ∗ ∗ ~ e ~ ∗
form J = 2m
Ψ(r, t)∇Ψ (r, t) − Ψ (r, t)∇Ψ(r, t) − AΨ Ψ. For massless particles with H = e~cp such as photons
m
the current is J = ~
e cP , where ~
e is the direction of the light beam.
6 “propagator” translates as “Zeitentwicklungsoperator” into German
8 DYNAMICS 133
We obtain the propagator as

i
Û(t) = e− ~ ĤS t
which is proven by insertion into the defining equation Eq. 8.11.

The propagator is unitary7 , that is
Û(−t) = Û −1 (t) = Û † (t)
The expectation value of an operator at a particular time can be written in two different ways,
once using the state at that time, and once the state at a reference time 0.
hAi = hψS (t)|ÂS |ψS (t)i = hψS (0)|Û † (t)ÂS Û(t)|ψS (0)i
The operator
def
ÂH (t) = Û † (t)ÂS (t)Û(t)
is the corresponding Heisenberg operator and the Heisenberg state is
def
|ΨH i = |ΨS (0)i = Û † (t)|ΨS (t)i
As the Heisenberg state is time-independent, the dynamics is entirely contained now in the oper-
ators. We can write down the equations of motion for this operator as 8
∂A
i ~∂t ÂH (t) = [ÂH (t), ĤS ] + i ~( )H
∂t
† d ÂS def
Note here that the last term means ( ∂A
∂t )H = Û d t Û.
If Ĥ is time independent, the transformation can be written explicitly in terms of eigenvalues ǫn
and eigenstates |φn iof the reference Hamiltonian.
i
X i
Û(t) = e− ~ ĤS t = |φn ie− ~ ǫn t hφn |
n
so that
X i
ÂH = |φn ihφn |ÂS |φm ie ~ (ǫn −ǫm )t hφm |
n,m
7 without proof
8 In the derivation we use that Ĥ commutates with the propagator Û, and that the time derivative of the propagator
is ~i Ĥ Û.
d d h † i
ÂH = Û (t)ÂS Û(t)
dt dt

d † d d
= Û (t) ÂS Û(t) + Û † (t) ÂS Û(t) + Û † (t)ÂS Û
dt dt dt

i † † d † i
= ĤS Û (t) ÂS Û(t) + Û (t) ÂS Û(t) + Û (t)ÂS − Û Ĥ
~ dt ~

i h † †
i
† d
= Ĥ(t)Û (t)ÂS Û(t) − Û (t)ÂS Û(t)Ĥ + Û (t) ÂS Û(t)
~ dt
i ∂ Â
= Ĥ(t), ÂH + ( )H
~ ∂t
134 8 DYNAMICS
8.5.3 Interaction picture

Often we can divide the Hamiltonian into a Hamiltonian H0 of a reference system, for which the
Schrödinger equation can be solved exactly, and a remainder W . W is often called the interaction.
ĤS = Ĥ0,S + ŴS

i ~∂t |ψS (t)i = Ĥ0,S + ŴS |ψS (t)i
Let us now introduce the propagator Û0 (t) of the unperturbed system
i ~∂t Û0 (t) = ĤS Û0 (t) (8.12)
For a time-independent Hamiltonian we can integrate this equation directly and obtain
i
Û0 (t) = e− ~ Ĥ0,S t
Now we introduce a state in the interaction picture
def
|ΨW i = Û0† (t)|ΨS (t)i (8.13)
and an operator in the interaction picture
def
ÂW (t) = Û0† (t)ÂS Û0 (t) (8.14)
We observe that the state in the interaction picture is just the initial state in the absence of a
perturbation Ŵ . If the perturbation is small also the time dependence is slow. This is the main
reason to introduce the interaction picture, namely to work with states for which the dynamics is
slow and governed only by the “difficult” part of the Hamiltonian.
We can show that expectation values have the usual form also in the interaction picture, namely
Û0† Û0 =1̂ Eqs. 8.13,8.14

hAi = hψS |ÂS |ψS i = hψS | Û0 Û0† ÂS Û0 Û0† |ψS i = hψW |ÂW |ψW i
| {z } | {z }
1̂ 1̂
Let us now investigate the equation of motion for a state in the Interaction picture

i ~∂t |ψS (t)i = Ĥ0,S + ŴS |ψS (t)i

Eq. 8.13
⇒ i ~∂t Û0 (t)|ψW (t)i = Ĥ0,S + Ŵ Û0 (t)|ψW (t)i

Eq. 8.12
⇒ Û0 (t) H0 + i ~∂t |ψW (t)i = Ĥ0,S + ŴS Û0 (t)|ψW (t)i
[Ĥ0,S Û0 ]=0 i i
⇒ i ~∂t |ψW (t)i = e ~ Ĥ0 t ŴS e− ~ Ĥ0 t |ψW (t)i = ŴW |ψW (t)i
Hence
i ~∂t |ψW (t)i = ŴW |ψW (t)i (8.15)

8 DYNAMICS 135
The dynamics for the operator is obtained as follows

i ~∂t ÂW = i ~∂t Û0† ÂS Û0

= i ~∂t Û0† ÂS Û0 + Û0† i ~∂t ÂS Û0 + Û0† ÂS i ~∂t Û0
| {z }
†
− i ~∂t Û0
†
= − Ĥ0,S Û0 Û0 Û0† ÂS Û0 + i ~Û0† ∂t ÂS Û0 + Û0† ÂS Û0 Û0† Ĥ0,S Û0
| {z } | {z } | {z }
=1̂ =1̂
Ĥ0,S Û0†

= [ÂW , Ĥ0,W ] + i ~ ∂t Â
W
[Ĥ0 ,Û0 ]=0

= [ÂW , Ĥ0,S ] + i ~ ∂t Â (8.16)
W
Thus we see the the operators are propagated via the simple dynamics that is the dynamics is governed
by the unperturbed system.
136 8 DYNAMICS
Chapter 9
Symmetry
9.1 Introduction
Symmetry is one of the most fundamental and useful concepts in

physics. This is so, because an underlying symmetry can often be rec-
ognized by inspection. We are biologically prepared by our mental and
sensory system to recognize symmetries and their violation.
Our sensitivity towards symmetries has probably developed evolu-
tionary from the necessity to detect, for example, a predator in the
high-growing grass of the Savanna.
Human art is largely about symmetries, i.e. order, and the violation
of expected symmetries. Symmetry or the absence of symmetry are
concepts that have a special access to our mind and emotions. Music,
for example, is nearly unconceivable without some tonal and rhythmic
systems. Free jazz is an extreme example of destroying such order to
a degree that it is nearly unrecognizable. The inexperienced listener is
appalled, but the trained ear is fascinated and stimulated by the mental
act of recognizing the barely noticeable ordering principles, that are
more implicit than explicit.
Fig. 9.1: Wolfgang Pauli,
We say that a system is symmetric under an operation Ŝ, if it be- 1900-1958. Austrian
haves in exactly the same way, after all relevant parts are transformed Physicist. Nobel price
by the same transformation. When we speak about the behavior, we in physics 1945 for the
mean the equations of motion. In quantum mechanics the dynamics is Pauli-exclusion principle.
governed by the Hamilton operator.
137
138 9 SYMMETRY
Thus we say that the system is symmetric under an operation Ŝ, if every solution of Hamilton operator
is transformed by the operation Ŝ into a solution of the Hamilton operator.
To be more concise, the system described by the Hamilton operator Ĥ is symmetric under the
operation Ŝ, if for any solution of
i ~∂t |ψ(t)i = Ĥ|ψ(t)i
the transformed solution

def
|ψ ′ (t)i = Ŝ|ψ(t)i
fulfills the equation
i ~∂t |ψ ′ (t)i = Ĥ|ψ ′ (t)i
If the Hamilton operator Ĥ is symmetric under Ŝ, we call Ŝ a symmetry operation of Ĥ.
Symmetry plays a very important role in quantum theory, because it allows us to break up the problem
of solving the Schrödinger equation into smaller problems of reduced dimensions. Since the difficulty
to solve a quantum mechanical system is substantial, the use of symmetry is very important. One
may object that symmetry is often broken and that the number of systems obeying a symmetry is
very limited. This may be so, but symmetry is useful even if the symmetry is not exactly obeyed. We
will learn in the section about approximations, that we can start solving a system accurately, even
though approximately, if we can find only a similar system that can be solved exactly. Here we can
use a symmetric model, even though the symmetry only holds approximately in the real system.
R
The most simple symmetry operation is the inversion in one dimension Ŝ = dx | − xihx|. If we
def
apply the inversion to a state |ψi we obtain the new state |ψ ′ i = Ŝ|ψi. The wave function is given
by
Z
ψ (x) = hx|ψ i = hx|Ŝ|ψi = hx| dx ′ | − x ′ ihx ′ |ψi
′ ′
Z
= dx ′ hx| − x ′ i hx ′ |ψi = h−x|ψi = ψ(−x)
| {z } | {z }
δ(x+x ′ ) ψ(x ′ )
The most important symmetry operations are translations and rotations. In order to grasp the
general idea of a symmetry operation, let us explore symmetry operations by collecting some general
features that all symmetry operations must obey.
• .
All symmetry operations have in common that they transform a physical state again into a phys-
ical state. This implies that the total probability remains unchanged by a symmetry operation
Ŝ.
Thus, if the state |φi is obtained from an arbitrary state |ψi by a symmetry operation Ŝ, i.e.
|φi = Ŝ|ψi
then the norm of |φi must be identical to that of the original state |φi, that is
hφ|φi = hψ|ψi
†
⇒ hφ|φi = hŜψ|Ŝψi = hψ|Ŝ Ŝ|ψi = hψ|1̂|ψi
true for all |ψi
⇒ Ŝ † Ŝ = 1̂
9 SYMMETRY 139
In the last step we exploited that two operators are identical, if the expectation values of both
operators are identical for all states.
This implies that a symmetry transformation is represented by a unitary operator S, i.e.
Ŝ † Ŝ = 1̂ or Ŝ −1 = Ŝ † . (9.1)
• An operator Â is symmetrical under the transformation Ŝ, if its expectation values are un-
changed for each state when applying the transformation1
|φi = Ŝ|ψi ⇒ hφ|Â|φi = hψ|Â|ψi
From this identity we find that Ŝ † ÂŜ = Â. Applying Ŝ = (Ŝ † )−1 to the left on both sides of
the equation shows the following
An operator Â is symmetric under a transformation Ŝ, if it commutates with the transformation,

i.e.
[Â, Ŝ] = 0 (9.2)
We have already seen that if two operators commutate, we can find a complete system of
states, that are eigenstates to both operators.
• .
A system is called symmetric with respect to a transformation, when the Hamilton operator is
symmetric.
Now we can come to the main point of symmetry operations: the knowledge of symmetries can
tremendously simplify the solution of the Schrödinger equation. We will see in the following, that
it is fairly easy to determine the eigenstates of a symmetry operator. We have seen that we can
determine the eigenstates of the Hamilton operator from the eigenstates of the symmetry operator.
In a basisset of eigenstates of a symmetry operator Ŝ, the Hamilton matrix is block diagonal. That
is, for each pair of eigenstates of Ŝ with different eigenvalues s1 , s2 , i.e.
Ŝ|ψ1 i = |ψ1 is1 ∧ Ŝ|ψ2 i = |ψ2 is2
the Hamilton matrix element vanishes
hψ1 |Ĥ|ψ2 i = 0 if s1 6= s2 (9.3)
The Hamilton matrix is block diagonal in the basis of eigenstates of the symmetry operator.
1 As an example imagine an operator that describes a spherical potential and the rotation as symmetry operation.
Clearly if we rotate a state about the center of the potential, the matrix elements are not affected by the potential.
140 9 SYMMETRY
s1 s2 s3
s1 0 0
s2 0 0
s3 0 0
Fig. 9.2: If the Hamilton operator commutates with a symmetry operator Ŝ, the matrix of the
Hamilton operator in a basisset of eigenstates of the Symmetry operator Ŝ, is block-diagonal. This
means that the matrix elements between eigenstates with different eigenvalues of Ŝ vanish. As
a consequence one only needs to consider, for each eigenvalue of Ŝ a smaller problem with less
basis-functions. Each block typically still contains an infinite number of eigenstates. However, the
spectrum of the Hamilton operator for each block is less dense, and this is what matters.
PROOF: If we consider two eigenstates of a symmetry operator with different eigenvalues, say
Ŝ|ψ1 i = |ψ1 is1

Ŝ|ψ2 i = |ψ2 is2
we can immediately show that the Hamilton matrix elements between the two states vanishes.
In the derivation, following below, we will use
Ŝ † |ψ1 i = |ψ1 is1∗
which is obtained as follows: We exploit Eq. 9.1 saying that the symmetry transformation is unitary, and
Eq. 9.4, specified below, saying that the absolute value of all eigenvalues of a unitary operator is equal to
one.
Ŝ Ŝ † =1̂ Ŝ|ψ1 i=|ψ1 is1 s1∗ Ŝ Ŝ † =1̂
Ŝ † |ψ1 i = Ŝ −1 |ψ1 i = |ψ1 is1−1 = |ψ1 i = |ψ1 is1∗
s1 s1∗
Now we are ready for the actual proof:

Ŝ|ψ2 i=ψ2 is2 [Ĥ,Ŝ]=0
hψ1 |Ĥ|ψ2 is2 = hψ1 |Ĥ Ŝ|ψ2 i = hψ1 |Ŝ Ĥ|ψ2 i
(ŜĤ)†
z }| { ∗
= hψ2 | Ĥ † Ŝ † |ψ1 i
Ŝ † |ψ1 i=|ψ1 is1∗
∗
= hψ2 |Ĥ † |ψ1 is1∗
= hψ1 |Ĥ|ψ2 is1
Thus
⇒ hψ1 |Ĥ|ψ2 i(s1 − s2 ) = 0
Thus if s1 6= s2 , the matrix element hψ1 |Ĥ|ψ2 i vanishes. q.e.d
9.1.1 Some properties of unitary operators

Unitary operators have a special property that will be useful in the following:
9 SYMMETRY 141
The absolute value of all eigenvalues of a unitary operator is one.
Ŝ Ŝ † = 1̂ ∧ Ŝ|ψi = |ψis ⇒ |s| = 1 (9.4)
Hence we can write its eigenvalues in the form ei γ with a real phase-angle γ.
PROOF: We use hψ|Ŝ † Ŝ|ψi = hψ|ψi with an eigenstate |ψi of Ŝ to an eigenvalue α, i.e. Ŝ|ψi = |ψiα.
We find α∗ hψ|ψiα = hψ|ψi and therefore α∗ α = 1. The most general form for α fulfilling this condition is
α = ei γ with a real γ. q.e.d.
9.1.2 Symmetry groups

A group G is a set of elements Ŝn with a binary operation (we will refer to it as multiplication, even
though it may be a generalization of a multiplication) such that
• the product of two elements of the group is again an element of the group, that is
∀Ŝn ,Ŝm ∈G Ŝn Ŝm ∈ G
• there is a unity element 1 in the group so that
∀Ŝn ∈G Ŝn 1 = 1Ŝn = Ŝn
• for each element Ŝn in the group there is an inverse Ŝn−1 so that
∀Ŝn ∈G ∃Ŝn−1 ∈G Ŝn Ŝn−1 = Ŝn−1 Ŝn = 1
For example, the set of integers with the addition is a group. In this case the addition plays the
role of the operation in the group definition. The addition by zero plays the role of the multiplication
by the unit element.
It is important to remember that the multiplication does not need to commutate. If the multipli-
cation commutates, ∀Ŝn Ŝm ∈G Ŝn Ŝm = Ŝm Ŝn , then the group is called Abelian, otherwise it is called
non-Abelian.
A group is called finite, if it has a finite number of elements.
9.2 Finite symmetry groups

First, we consider finite symmetry groups, because they are simple to handle. An example for a finite
group is the mirror reflection.
The eigenvalues of a symmetry operator Ŝ of a finite symmetry group, that is Ŝ n = 1̂ with some
positive integer n, are e i 2πj /n where j is any integer number from 0 to n − 1.
142 9 SYMMETRY
PROOF: Let us consider a series of operators in the group (. . . , Ŝ −1 , 1̂, Ŝ, Ŝ 2 , . . .) If the group is finite, this
series must be cyclic, i.e. we can find a number n so that Ŝ j+n = Ŝ j for all j . n must be smaller than or equal
to the number of elements in the group. Thus we find, by applying Ŝ −j , that Ŝ n = 1.
Let us consider an eigenstate |ψi of Ŝ, so that
Ŝ|ψi = |ψiei φ
We have shown that the eigenvalues of a unitary operator can be written as ei φ . Then we can show that |ψi
is an eigenstate of S j with eigenvalue ei φj .
Ŝ j |ψi = |ψiei φj
Since on the other hand S n = 1 we find that ei φn = 1 and therefore φ = 2π/n. Thus the eigenvalues are
ei 2πj/n . q.e.d.
One can construct eigenstates of the symmetry operator Ŝ from any state |φi by
n−1
X
|ψk i = Ŝ j |φie−i 2πj k/n (9.5)
j =0
Note however, that the result of the sum may vanish.
Here we showa that the state defined in Eq. 9.5 is an eigenstate of Ŝ with eigenvalue ei 2πk/n , where the series
Ŝ i is cyclic with period n.
Proof: We multiply |ψk i from Eq. 9.5 with Ŝ
n−1
X n−1
X n−1
X
Ŝ Ŝ j |φie−i 2πjk/n = Ŝ j+1 |φie−i 2π(j+1)k/n ei 2πk/n = Ŝ j |φie−i 2πjk/n ei 2πk/n
j=0 j=0 j=0
| {z } | {z }
|ψk i |ψk i
In the last step, we shifted the indices j + 1 → j and exploited that the series (S i ) is cyclic with period n.
Thus we have shown that |ψk i as defined in Eq. 9.5 obeys the eigenvalue equation.
Ŝ|ψk i = |ψk iei 2πk/n
q.e.d.
a One
Pn−1 j
can derive the result by inserting an Ansatz |ψk i = j=0 Ŝ |φicj,k into the eigenvalue equation Ŝ|ψk i =
|ψk ie−i 2πjk/n which results in a recursive relation for the coefficients cj,k .
9.3 Continuous symmetries
9.3.1 Shift operator
Let us now consider an operator Q and its canonical momentum P . The canonical momentum has
been defined in Eq. 5.26 by
Eq. 5.26 ~
[P̂ , Q̂] = .
i
We define a shift operator

9 SYMMETRY 143
SHIFT OPERATOR
i
Ŝ(∆) = e− ~ P̂ ∆ . (9.6)
As shown below, the shift operator has the property
Ŝ † (∆)Q̂Ŝ(∆) = Q̂ + ∆1̂. (9.7)
This implies that if we transform a state |ψi to a new state |φi = S(∆)|ψi, the expectation value
of the new state is that of the old state shifted by ∆ if the state is normalized, i.e. hφ|Q̂|φi =
hψ|Q̂|ψi + ∆.
Here we prove Eq. 9.7 for a shift operator defined by Eq. 9.6.
PROOF: First, we show that the commutator of a function of a momentum with the position can be expressed
by the derivative of that function with respect to the momentum, i.e.
~ d
[f (P̂ ), Q̂] = f (P̂ ) (9.8)
i dP
Because the function of an operator is defined by its power series expansion, we show first that the relation
holds for all powers of the momentum,
[P̂ n , Q̂] = P̂ n Q̂ − Q̂P̂ n = P̂ n−1 [P̂ , Q̂] + P̂ n−1 Q̂P̂ − Q̂P̂ n

~
= P̂ n−1 + [P̂ n−1 , Q̂]P̂
i
We apply this equation recursively n times until we obtain on the right side the commutator [P̂ 0 , Q̂] = [1̂, Q̂] =
ˆ Thus we obtain
∅.
~ n−1
[P̂ n , Q̂] = n P̂ ,
i
which we can rewrite in the form
~ d n
[P̂ n , Q̂] = P̂ . (9.9)
i dP
which is exactly Eq. 9.8 specialized to powers, i.e. f (x) = x n . Addition and multiplication with scalar does
not affect the result so that we can generalize Eq. 9.9 to a power-series expansion an such to Eq. 9.8.
This is the first time that we see a derivative with respect to an operator. It is the elementary step defining
the derivative with respect to an operator. The derivative is formed here just as if P̂ would be a simple
number. If the function depends on operators that do not commutate, the derivative is not as simple and
must be traced back to the individual terms of the power series expansion.
Finally, we only need to specialize the result to the shift operator from Eq. 9.6.
Eq. 9.8 ~ d − ~i P̂ ∆
⇒ [Ŝ(∆), Q̂] = e = −∆ Ŝ(∆)
i dP | {z }
i
− P̂ ∆
e ~
⇒ Q̂Ŝ(∆) = Ŝ(∆)Q̂ + ∆Ŝ(∆)

S −1 =S † †
⇒ Ŝ (∆)Q̂Ŝ(∆) = Q̂ + ∆1̂
q.e.d. which proves Eq. 9.7.
This shows that if we have an eigenstate of the operator Q̂, the shift operator, Eq. 9.6, creates
144 9 SYMMETRY
a new eigenstate of Q̂ with an eigenvalue increased by ∆.
Q̂|ψi = |ψiq

⇒ Q̂ Ŝ(∆)|ψi = Ŝ(∆) Q̂ + ∆ |ψi = Ŝ(∆)|ψi (q + ∆)
So far we have started from the canonical momentum P̂ and constructed the shift operator
via Eq. 9.6. Now we go the opposite direction and start out from the shift operator and derive
the canonical momentum from it. This will show the intricate relationship between momenta and
symmetry.
If the shift operator S for an operator Q̂ obeying Eq. 9.7 is known, then we can construct the
canonical momentum using
i i
Ŝ(∆) = e− ~ P̂ ∆ = 1̂ − P̂ ∆ + O(∆2 )
~
Ŝ(∆) − 1̂
i~ = P̂ + O(∆)
∆
Thus we obtain an expression that allows us to obtain the canonical momentum to any variable,
for which the corresponding shift operator is known.
CANONICAL MOMENTUM
Ŝ(∆) − 1̂
P̂ = i ~ lim (9.10)
∆→0 ∆
9.4 Exercises
9.4.1 Translation operator from canonical momentum
i
Use the relation Ŝ = e− ~ p̂x with the momentum
Z
~
p̂ = dx |xi ∂x hx| (9.11)
i
to show that ŜΨ(x) = Ψ(x − ∆)

Solution:
The shift operator given above is a function of the momentum operator, which is defined by its
power series expansion. Whenever the function depends only on one operator one best changes the
representation into eigenstates of that operator. Thus we need to transform the momentum operator
into the basis of its eigenstates. Most of the derivation consists of introducing again the momentum
representation, which has been worked out previously.
The eigenstates of the momentum operator are denoted by |pi. They fulfill the eigenvalue
equation
p̂|pi = |pip (9.12)
and are normalized as follows:
hp|p ′ i = 2π~δ(p − p ′ ) (9.13)
The inclusion of the factor 2π~ is the convention used in this text. Other books often drop the
factor.
9 SYMMETRY 145
The real-space wave functions corresponding to momentum eigenstates have the form
i
hx|pi = e ~ px (9.14)
Eq. 9.14 is shown as follows: The eigenvalue equation Eq. 9.12 can easily be verified
Z Z
Eq. 9.11 ′ ~ i ′ i
hx|p̂|pi = dx hx|x i ∂x ′ hx |pi = dx ′ δ(x − x ′ ) pe ~ px = e ~ px p = hx|pip
′ ′
| {z } i | {z }
δ(x−x ′ ) eipx ′ /~
Now we need to check the normalization condition Eq. 9.13

1̂
z
Z }| { Z Z
i ′
′ ′ ′
hp|p i = hp| dx |xihx| |p i = dx hp|xihx|p i = dx e ~ (p−p )x
Clearly the integral does not converge. Therefore, we introduce a limiting factor to the integrand
Z 0 Z ∞
i ′ i ′
hp|p ′ i = lim dx e( ~ (p−p )+λ)x + dx e( ~ (p−p )−λ)x
λ→0 −∞ 0
−1 −1 !
i ′ i ′
= lim (p − p ) + λ − (p − p ) − λ
λ→0 ~ ~
−2λ
= lim
λ→0 −1 − p ′ )2 − λ 2
~2 (p
~λ
= 2~ lim
λ→0 (p − p ′ )2 + ~ 2 λ2
We recognize already that the function vanishes for p 6= p ′ and that it diverges for p = p ′ which
suggests that the limes could be a delta function. What needs to be shown is that the integral over
p is equal to a well-defined constant.
We use the integral formula taken from Bronstein[13]
Z
1 1 x
dx 2 = atan( )
x + a2 a a
Thus we obtain
 
Z ∞ ′
∞
p−p
dphp|p ′ i = 2~ lim atan(
 
) = 2~ atan(∞) − atan(−∞)
−∞ λ→0 ~λ −∞ | {z } | {z }
π/2 −π/2
= 2π~
Thus, we can identify the scalar product as

hp|p ′ i = 2π~δ(p − p ′ )
which verifies the normalization condition Eq. 9.13.
The unity operator in momentum representation has the form
Z
dp
1̂ = |pihp|
2π~
The form of the unity operator can be verified by applying it to an arbitrary state |pi. Due to the
superposition principle, it holds for any state if it holds for a complete set of states such as |pi. The
identity is shown as follows:
Z Z
dp ′ ′ ′ dp ′ ′
1̂|pi = |p ihp |pi = |p i2π~δ(p − p ′ ) = |pi
2π~ 2π~
146 9 SYMMETRY
Now we can evaluate the shift operator in momentum representation

Z Z Z
i dp dp − i p̂∆ dp i
Ŝ(∆) = |e−{z~
p̂∆
} 2π~ |pihp| = e ~ |pihp| = |pie− ~ p∆ hp|
2π~ 2π~
Ŝ(∆) | {z }
1̂
Note that the exponential does not carry the momentum operator, but its eigenvalue, which is a
number. Now we can determine the result of the shift operator acting on a wave function ψ(x).
Z
dp i
Ŝ(∆)ψ(x) = hx|Ŝ(∆)|ψi = hx| |pie− ~ p∆ hp||ψi
2π~
Z Z
dp i dp i p(x−∆)
= hx|pi e− ~ p∆ hp|ψi = e~ ψ(p)
2π~ | {z } | {z } 2π~
i ψ(p)
e ~ px
We compare the result with the original wave function expressed by its momentum representation
Z
dp
1̂ψ(x) = hx|1̂|ψi = hx| |pihp||ψi
2π~
Z Z
dp dp i px
= hx|pi hp|ψi = e ~ ψ(p)
2π~ | {z } | {z } 2π~
i ψ(p)
e ~ px
We observe directly that
Ŝψ(x) = ψ(x − ∆) (9.15)

Chapter 10
Specific Symmetries
10.1 Parity
An example of a finite group is the inversion:
Z
Ŝ = dx| − xihx|
Z Z
Ŝψ(x) = hx|Ŝ|ψi = hx| dx ′ | − x ′ ihx ′ | |ψi = dx ′ hx| − x ′ i hx ′ |ψi
| {z } | {z }
| {z } δ(x+x ′ ) ψ(x ′ )
Ŝ
Z
= dx ′ δ(x + x ′ )ψ(x ′ ) = ψ(−x)
The group has the elements (1̂, Ŝ)

Since Ŝ 2 = 1̂ we find two eigenvalues s1 = 1 and s2 = −1. If the eigenvalue is +1 we say the
state has parity one or the state is even. If it is −1 we say the state has parity minus one or that
the state is odd.
Example: We expect that a system is inversion symmetric, if the potential is inversion symmetric,
such as the particle in the box. If we place the zero in the center of the box, then the inversion does
not change the potential. As a consequence of the symmetry, the eigenstates shown in Fig. 4.4 on
p. 60, are either symmetric or antisymmetric with respect to the inversion: The symmetric states
have parity s = 1, while the antisymmetric states have parity s = −1.
In the following we show that the inversion symmetry can be traced back to the inversion symmetry
of the potential. It seems obvious that an inversion symmetric potential leads to a inversion symmetric
problem. Nevertheless this nearly trivial exercise may be helpful to demonstrate the principle of
showing the symmetry of a quantum mechanical problem.
Z
−~2 2
Ĥ = dx|xi ∂x + v (x) hx| = T̂ + V̂
2m
Z
−~2 2
T̂ = dx|xi ∂ hx|
2m x
Z
V̂ = dx|xiv (x)hx|
In order to verify that the system is symmetric with respect to inversion we need to show that
147
148 10 SPECIFIC SYMMETRIES
the Hamilton operator commutes with the inversion operator.

Z Z Z Z
−~2 2 −~2 2
[T̂ , Ŝ] = dx|xi ∂ hx| dx | − x ihx | − dx ′ |x ′ ih−x ′ |
′ ′ ′
dx |xi ∂ hx|
2m x 2m x
| {z } | {z } | {z } | {z }
T̂ Ŝ Ŝ T̂
Z 2 Z Z Z 2
−~ 2 −~ 2
= dx|xi ∂ dx ′ hx| − x ′ ihx ′ | − dx ′ dx |x ′ i h−x ′ |xi ∂ hx|
2m x | {z } | {z } 2m x
δx+x ′ δ(x+x ′ )
Z 2 Z 2
−~ 2 −~ 2
= dx|xi ∂ h−x| − dx | − xi ∂ hx| = 0
2m x 2m x
In the last step we performed a variable transform x → −x on the right-hand part.1 2 Thus we have
shown that the kinetic energy commutates with the inversion operator.
Now we need to investigate the potential operator
Z Z Z Z
[V̂ , Ŝ] = dx|xiv (x)hx| dx ′ | − x ′ ihx ′ | − dx ′ |x ′ ih−x ′ | dx |xiv (x)hx|
| {z } | {z } | {z } | {z }
Z Z V̂ Ŝ
Z Z Ŝ V̂
= dx dx ′ |xiv (x) hx| − x ′ ihx ′ | − dx ′ dx |x ′ i h−x ′ |xi v (x)hx|

| {z } | {z }
δ(x+x ′ ) δ(x+x ′ )
Z Z
= dx|xiv (x)h−x| − dx ′ |x ′ iv (−x ′ )h−x ′ |
Z
= dx|xi v (x) − v (−x) h−x|
Thus the commutator [Ĥ, Ŝ] between the hamiltonian and the parity operator vanishes, if the potential
is itself inversion symmetric, i.e. v (x) = v (−x). Thus the Hamiltonian is symmetric under the parity
operation, if the parity operation transforms the potential on itself. Remember that this is not true
for the wave function: The wave functions of a parity symmetric system are either symmetric or
antisymmetric under the parity operation.
10.2 n-fold rotation about an axis

Consider an n-fold rotation in space, which is described by a unitary matrix U
r~′ = U ~r
Since U describes a n-fold rotation, U n = 1 .

1 Variable transform:
dy y =−x
y (x) = −x ⇒ dy = dx = −dx
dx
Z b Z y (b) Z −b Z −a
dy y =−x
dxf (x) = dy f (x(y )) = − dy f (−y ) = + dy f (−y )
a y (a) dx −a −b

dy d2y dy 2 2 y =−x 2
∂x2 = ∂x ∂y = ∂y + ∂y = ∂y
dx dx 2 dx
2 Here we show the identity for an arbitrary matrix element. which proves the operator identity.
Z Z
−~2 2 −~2 2
hφ|[T̂ , Ŝ]|ψi = dxhψ|xi ∂x h−x|φi − dx hψ| − xi ∂ hx|φi
2m 2m x
Z 2 Z 2
−~ 2 −~ 2
= dx ψ ∗ (x) ∂ φ(−x) − dx ψ ∗ (−x) ∂ φ(x) = 0
2m x 2m x
10 SPECIFIC SYMMETRIES 149
The rotation in real space corresponds to a transformation Ŝ in Hilbert space.

Z
Ŝ = d 3 r |U
U ~rih~r|
Z
Ŝψ(~r) = U r~′ i hr~′ |ψi
h~r|Ŝ|ψi = d 3 r ′ h~r|U
| {z } | {z }
U r~′ ) ψ(r~′ )
r −U
δ(~
−1 †
U ]=1;U
det[U U U
=U
= U †~r)
ψ(U
Note that the resolution of the delta function was non-trivial.3

Just as the rotation of real space point form a group with elements {1 1, U , U 2 , . . . U n−1 } the group
of the corresponding symmetry operations in real space forms a group. The group has n elements,
{1̂, Ŝ, . . . , Ŝ n−1 }. The eigenvalues of Ŝ are sj = ei 2πj /n for j = 0, . . . , n − 1.
Editor: introduce here the example of ammonia, to show how eigenstates for a symmetry
operator are constructed
10.3 Exchange of two particles

If the exchange of two particles does not change the system, we call those particles identical. Thus
the presence of identical particles implies a symmetry. Unlike classical mechanics, this symmetry has
radical consequences in quantum mechanics.
The symmetry operation is the permutation P̂ of two particles.
Z Z
P̂i ,j = d 3 r1 · · · d 3 rN |~r1 , . . . , ~rj , . . . , ~ri , . . . , ~rN ih~r1 , . . . , ~ri , . . . , ~rj , . . . , ~rN |
The wave function that corresponds to the state |~r1 , . . . , |~rN i is a product of δ-functions in the
3N-dimensional configuration space centered at the positions ~r1 , . . . , ~rN , i.e.
|ψi = |~r1 , . . . , ~rN i ⇒ ψ(r~′ 1 , . . . , r~′ N ) = hr~′ 1 , . . . , r~′ N |~r1 , . . . , ~rN i = δ(r~′ 1 − ~r1 ) · · · δ(r~′ N − ~rN )
The permutation can be considered as a geometric operation in the 3N-dimensional configuration

space of all N particles. This is illustrated best by considering two one-dimensional particles as the
simplest possible case. The positions of the two particles are described by two coordinates (x1 , x2 ).
The permutation of the two particles corresponds to a mirror operation about the main diagonal of
the x1 − x2 -plane. This is illustrated in Fig. 10.3.1 on p. 152.
Let us now consider a wave function ψ(~r1 , ~r2 , . . . , ~rN ) for many particles and the permutation of
two particles as symmetry operator.
P̂i ,j ψ(. . . , ~ri , . . . , ~rj . . .) = ψ(. . . , ~rj , . . . , ~ri . . .)
For the interchange of every two particles we have a symmetry group with two elements and eigen-
values 1 and −1.
For identical particles there exists no Hamilton operator that is not invariant with respect to
particle exchange. If there would be one, then it would be possible to devise and experiment that
distinguishes between the particles. If we can distinguish between particles, they are obviously no
more be identical. Since this is not the case, each Hamilton operator commutates with the particle
exchange of identical particles. Thus all particles are either symmetric or antisymmetric with respect
to particle exchange. Accordingly, particles can be classified as being either Fermions or Bosons.
3 In one dimension the identity δ(ax) = 1 δ(x) holds. In three dimension this translates into δ(A
A~x) = 1
δ(~
x ). In
a det A
the operation used in the text, the prefactor vanishes, because U is unitary and therefore det U = 1
FERMIONS AND BOSONS
Fermions Bosons
wave function antisymmetric under permutation wave function symmetric under permutation
half-integer spin integer spin
Pauli principle macroscopic quantum states possible
(e.g. Bose Einstein condensation, Laser)
“particles” “interactions”
Examples: Examples
electron photon (electrodynamics)
proton graviton (gravity)
neutron Z- and W-boson (weak force)
quark gluon (strong force)
Higgs boson
phonon
exciton
10.3.1 Fermions
Particles that are antisymmetric under particle exchange are called Fermions. All Fermions have a
half-integer spin. (The concept of spin will be introduced later.) Examples for Fermions are electrons,
protons, quarks. An important consequence of the antisymmetry is the Pauli principle, which says
that no two Fermions can occupy the same site or the same one-particle state.
Let us show that two identical Fermions cannot occupy the same site. The wave function ampli-
tude for two particles at the same position ~r0 and all others on arbitrary positions is
~r(0) , . . . , |{z}
ψ(. . . , |{z} ~r(0) , . . .)
pos.i pos.j
Exchanging the two particle coordinates, one obtains , on the one hand, the same result, on the
other hand we obtain a sign change due to particle permutation. Thus:
ψ(. . . , |{z} ~r(0) , . . .) = −ψ(. . . , |{z}
~r(0) , . . . , |{z} ~r(0) , . . .)
~r(0) , . . . , |{z}
pos.i pos.j pos.i pos.j
This is only possible if the wave function amplitude vanishes, whenever two particles have the same
position.4 The probability that the two particles are at ~r0 is obtained from the absolute square of the
wave function integrated over all particle positions except particles i and j. This probability vanishes
as well.
PAULI PRINCIPLE
• The probability of finding two identical Fermions in the same position vanishes
• Two identical Fermions cannot occupy the same one-particle state.
Let us now consider a product state Ψ(~r1 , ~r2 ) = φ1 (~r1 )φ2 (~r2 ). Product states are the most
simple many particle wave functions. They play an important role for non-interacting particles.
4 If the two particles are at the same position the value of the wave function does not change under permutation.
On the other hand the wave function has to change its sign. Thus the wave function vanishes for identical positions.
P (1)
For non-interacting particles the N-particle Hamiltonian Ĥ (N) = N i =1 Ĥi is a sum of one-particle
(1)
Hamiltonians, where the one-particle Hamiltonian Ĥi only acts on the coordinates of the i -th
particle. For non-interacting particles, we can use the method of separation of variables to express
the eigenstates as a product of wave functions of the one-particle wave functions, that is a product
state. The energy is then the sum of the corresponding one-particle energies.
However, a product state does not have the symmetry or antisymmetry required for identical
particles. Thus we need to antisymmetrize the wave function. We can anti-symmetrize this state an
according to our recipe from Eq. 9.5. In the two-particle case we obtain
1 h i
ΨF (~r1 , ~r2 ) = √ φ1 (~r1 )φ2 (~r2 ) − φ2 (~r1 )φ1 (~r2 )
2!
Here we can easily verify that the Pauli principle holds: the wave function vanishes if φ1 (r ) = φ2 (r ).
Thus we have shown that one cannot place two identical Fermions into a single one-particle state.
An antisymmetric wave function can be written in a more general way as a Slater determinant
 
φ1 (~r1 ) φ1 (~r2 ) · · · φ1 (~rN )
 
1  φ2 (~r1 ) φ2 (~r2 ) · · · φ2 (~rN ) 
ΨF (~r1 , . . . , ~rN ) = √ det 
 .. .. .. .. 

N!  . . . . 
φN (~r1 ) φN (~r2 ) · · · φN (~rN )
From the property of that the determinant it changes its sign under exchange of two columns,
we obtain antisymmetry. Not every antisymmetric wave function can be represented as a Slater
determinant. The most general form of a Fermionic wave function of N particles is a superposition
of all N-particle Slater determinants that can be formed from a complete set of one-particle states.
Not every N-particle state can be expressed as a single Slater determinant. However, if we
form all possible the Slater determinants from the eigenstates of a one-particle Hamiltonian ĥ, we
obtain a complete basisset of the Hilbert space of Fermionic, i.e. antisymmetric, N-particle states.
These
PN Slater-determinants are the eigenstates of the non-interacting N-particle Hamiltonian Ĥ (N) =
(1) (1)
i =1 Ĥi . For identical particles, each one-particle Hamiltonian Ĥi is identical to all others except
that it acts on the coordinates of a specified, namely the i -th, particle. The eigenstates of an
observable such as the N-particle Hamiltonian forms a complete basisset. A general N-particle wave
function can thus be written as a Multiconfiguration wave function, namely as a sum of all Slater
determinants that can be built from a complete one-particle basisset.
The energy eigenstates of non-interacting Fermions can be expressed as single Slater determinants
made from the eigenstates of the corresponding one-particle Hamiltonian. The wave functions of
interacting particles must be written as multiconfiguration wave functions. Multiconfiguration wave
functions are notoriously difficult to describe. However there are techniques such as density functional
theory, that allow to map a system of interacting electrons onto a system of non-interacting particles.
The latter allow us to effectively deal with wave functions that are single Slater determinants.
10.3.2 Bosons
Particles that are symmetric under particle exchange are called Bosons. All Bosons have an integer
spin. (The concept of spin will be introduced later.) Examples for bosons are photons, mesons.
Typically they are related to interactions such as the Coulomb interaction.
I can form a Bosonic state as a symmetrized product of a single one-particle states. For a two
particle state built from two one-particle orbitals φ1 (~r) and φ2 (~r)one obtains
1
ΨB (~r1 , ~r2 ) = √ (φ1 (~r1 )φ2 (~r2 ) + φ2 (~r1 )φ1 (~r2 ))
2
This can be generalized to many particles as follows:
Y 1
ΨB (~r1 , ~r2 , . . . , ~rN ) = √ (1 + Pi ,j )φ(~r1 )φ(~r2 ) . . . φ(~rN )
i <j
2
x2
x1
Fig. 10.1: Wave function of two fermions in a one-dimensional box. One axis represents the position
of the first particle and the other axis that of the other. The problem of two particles in a one-
dimensional box is analogous to that of one-particle in a two dimensional box. The symmetry with
respect to particle exchange corresponds to a mirror plane at the diagonal x1 = x2 . The antisymmetry
forces a node plane to lie at the diagonal x1 = x2 . The lower figure shows a contour plot of the wave
function. The different signs are indicated by the different colors. The functions on the right indicate
the wave function (dashed) and the density (filled line) of the first particle for a fixed position of the
second. The position of the second is marked by the black ball.
where the operator Pi ,j permutes the coordinates ~ri and ~rj Thus we can place an macroscopic number
of Bosons into a single one-particle state. Such a state is called a macroscopic quantum state,
and it allows quantum properties to be observed in the macroscopic . An example for such a state is
laser light. Another example is the superfluid state of liquid Helium, which flows without dissipation
and has infinite viscosity. Superconductivity is the result of conduction electrons forming pairs, which
are Bosons, and the condensation of these pairs into a single one-particle orbital. Even Atoms can
be prepared in a Bosonic state, which undergoes Bose-Einstein condensation and then all atoms are
placed in a macroscopic quantum state.
10.4 Continuous translations

The translation operator, which translates a state in x-direction, is defined by
Z
Ŝ = dx |x + ∆ihx|
Z
Ŝ(∆)Ψ(x) = hx|Ŝ|ψi = dx ′ hx|x ′ + ∆ihx ′ |ψi = hx − ∆|ψi = Ψ(x − ∆)
| {z }
δ(x ′ +∆−x)
Note here the minus sign in the argument: If Ŝ(∆) shifts the wave function to the right, the values
for the new wave function are obtained from the unshifted wave function at a position to the left.
We obtain the conjugate momentum by
Z
Eq. 9.10 1
p̂x = i ~ lim dx |x + ∆ihx| − |xihx|
∆→0 ∆
Z
1
= i ~ lim dx |xihx − ∆| − |xihx|
∆→0 ∆
Z
hx − ∆| − hx|
= i ~ dx|xi lim
∆→0 ∆
| {z }
−∂x hx|
Z
~
= dx |xi ∂x hx|
i
Z
~
p̂x Ψ(x) = hx|p̂x |ψi = dx ′ hx|x ′ i ∂x ′ hx ′ |ψi
| {z } i | {z }
δ(x−x ′ ) ψ(x ′ )
~
= ∂x ψ(x)
i
Thus we have derived the explicit form for the momentum operator from the shift operator.
The generalization to a translation in three dimensions is straightforward:
Z
~ ~ r|
Ŝ(∆) = d 3 r |~r + ∆ih~
~ ~
r ) = Ψ(~r − ∆)
Ŝ(∆)Ψ(~
Z
~~
p~ˆ = d 3 r |~ri ∇h~
r|
i
The eigenstates |ψ~k i of the shift operator can be obtained from the eigenvalue equation of the
momentum operator
Eq. 9.6 ~~
p̂ψ~k (~r) = ψ~k (~r)~~k ⇒ ~ ~ (~r) = Ψ~ (~r − ∆)
Ŝ(∆)Ψ k k
~ = Ψ~k (~r)e−i k ∆ (10.1)
where ~~k are here the eigenvalues of the momentum operator. By setting ~r = ~0 and replacing ∆
~ by
−~r in Eq. 10.1, we find that the eigenstates of the translation operator, which are also momentum
eigenstates have the from of plane waves.
~ Eq.=10.1 Ψ~ (~r)e−i ~k ∆
Ψ~k (~r − ∆)
~
k
r →~0;
~ ~ r
∆→−~ ~
⇒ Ψ~k (~0 − (−~r)) = Ψ~k (~0)e−i k(−~r)
~
⇒ Ψ~k (~r) = Ψ~k (~0)ei k~r (10.2)
EIGENSTATES OF THE TRANSLATION OPERATOR
The eigenstates of the translation operator,
~ −i k ∆ ~~
Ŝ(∆)|ψk i = |ψ~
~ k ie (10.3)
are plane waves

Eq. 10.2 ~
Ψ~k (~r) = A ei k~r (10.4)
The eigenstates of the shift operator are also the eigenstates of the corresponding momentum oper-
ator, i.e.
p̂|ψ~k i = |ψ~k i~~k
This is the reason for using plane waves as Ansatz, when we search a solution for a differential
equation with translational symmetry. Because the result rests on symmetry, this Ansatz is very
general and works for any kind of differential equation.
10.5 Discrete translations, crystals and Bloch theorem
Most solids are crystals. This implies that the atoms are located on regular positions. The clearest
evidence of this regular arrangement are the crystal faces of gem stones. This arrangement results in
a symmetry, the lattice periodicity. Lattice translation symmetry is a discrete translational symmetry.
10.5.1 Real and reciprocal lattice: One-dimensional example
Let us consider a linear chain of hydrogen atoms as the most simple model of a one-dimensionally
periodic crystal of hydrogen atoms shown in Fig. 10.2. It is a model system, which is not stable in
reality. However it allows to develop the concepts.
T a
Fig. 10.2: Linear chain of hydrogen atoms. The lattice spacing is a. The balls represent the atoms.
The graph sketched below represents the potential as function of position. The primitive lattice
~ which has the length a. The primitive reciprocal lattice vector has the length 2π/a.
vector is T
The periodicity of the chain implies that for every atom at position R ~ 0 there are also equivalent
~ ~ ~ ~
atoms at the positions Rn = R0 + nT , where n is any integer. T is the primitive lattice vector.
General lattice vectors are ~ tn = nT~ . The primitive lattice vector is the smallest possible lattice
vector, that fully describes the lattice periodicity. Its length is called the lattice constant a.
Not only the atomic positions are regular, but also the potential is periodic, that is v (~r + ~tn ) =
v (~r). A periodic potential has a discrete Fourier transform
X ~
v (~r) = ei Gn~r Vn
n
where the vectors G ~n = n~g are the general reciprocal lattice vectors. The primitive reciprocal
lattice vector is denoted by ~ g . The length of the primitive reciprocal lattice vector is inversely
g | = 2π
proportional to the real-space lattice constant a, that is |~ a
10.5.2 Real and reciprocal lattice in three dimensions

Let us now generalize the expressions to three a three-dimensional lattice. Fig. 10.5.2 demonstrates
the concepts developed in the following. Corresponding to the three spatial directions there are now
three primitive lattice vectors, which we denote by T ~1 , T
~2 , T
~3 . Note that in two or three dimensions
there is no unique choice of primitive lattice vectors. The three primitive lattice vectors span the
primitive unit cell.
A general lattice vector ~
ti ,j,k can be expressed by the primitive lattice vectors T ~1 , T
~2 , T
~3 as
  
Tx,1 Tx,2 Tx,3 i
~
ti ,j,k ~ ~ ~
= i T1 + j T2 + k T3 =

ˆ  Ty ,1
 
Ty ,2 Ty ,3   j 
Tz ,1 Tz ,2 Tz ,3 k
| {z }
T
where i , j, k are arbitrary integers.

It is often convenient to combine the lattice vectors into a 3 × 3 matrix T as shown above. Often
the atomic positions R ~ n are provided in relative positions X
~ which are defined as
~n = T X
R ~n ⇔ ~ n = T −1 R
X ~n
A potential is called periodic with respect to these lattice translations, if V (~r + ~

ti ,j,k ) = V (~r).
RECIPROCAL LATTICE
The reciprocal lattice is given by those values of the wave vector G, ~ for which the corresponding
~
plane waves ei G~r have the same periodicity as the real-space lattice.
The primitive reciprocal-lattice vectors ~
gn for n = 1, 2, 3 are defined in three dimensions as
~ ~m = 2πδn,m
gn T (10.5)
Thus, in three dimensions the reciprocal lattice vectors can be obtained as
~ ×T
T ~3 ~ ×T
T ~1 ~ ×T
T ~2
~
g1 = 2π 2 , ~
g2 = 2π 3 , ~
g3 = 2π 1
~1 T
T ~2 × T
~3 ~1 T
T ~2 × T
~3 ~1 T
T ~2 × T
~3
It is easily shown that these expressions fulfill the defining equation Eq. 10.5 for the reciprocal lattice
vectors. Note that the expressions for the second and third lattice vector are obtained from the first
by cyclic commutation of the indices.
Because the vector product is perpendicular to the two vectors forming it, every reciprocal lattice
vector is perpendicular to two real-space lattice vectors. This brings us to the physical meaning of
the reciprocal lattice vectors: Two real space lattice vectors define a lattice plane. A plane can be
defined either by two linearly independent vectors in the plane or alternatively by the plane normal.
The reciprocal lattice vectors apparently define lattice planes, because they are plane normals.
Let us now consider the distance ∆n of two neighboring lattice planes. It is obtained by projecting
one vector pointing T ~n from one plane to the other onto a normalized plane normal ~ gn |.
gn /|~
∆n = T ~n ~
gn Eq. 10.5
=
2π
|~
gn | |~
gn |
2π
⇒ |~
gn | =
∆n
PHYSICAL MEANING OF RECIPROCAL LATTICE VECTORS
• The reciprocal lattice vectors are perpendicular to the lattice planes of the real-space lattice.
• The length of the primitive reciprocal lattice vectors is 2π divided by the distance of the lattice
planes.
We can form a 3 × 3 matrix g from the three primitive reciprocal lattice vectors.
 
gx,1 gx,2 gx,3
ˆ
def 
g =  gy ,1 gy ,2 gy ,3 
gz ,1 gz ,2 gz ,3
The definition Eq. 10.5 of the reciprocal lattice vectors can be expressed in matrix form as
Eq. 10.5
g ⊤T = 1
2π1 (10.6)
Thus the matrix g is obtained as

Eq. 10.6 ⊤
g = 2π T −1 (10.7)
The general reciprocal lattice vectors are
~i ,j,k = i~
G g1 + j~
g2 + k~
g3
The reciprocal lattice vectors play an important role in the Fourier transform of periodic functions.
~n are characterized by a reciprocal lattice
All plane waves that are periodic with the lattice vectors T
vector.
~ ~ ~
ei G(~r+Tn ) = ei G~r
~ r +T
⇒ G(~ ~n ) = G~
~ r + 2πmn
~ ~
⇒ G Tn = 2πmn
~ T = 2π m
⇒ GT ~
⊤
⇒G ~ T −1 = 2π T −1 m
~ = 2π mT ~
~ are exactly those of the reciprocal lattice. This was

where mn are integers. The allowed values for G
the motivation for the definition Eq. 10.6.
t3,−2,0 g2
∆2
C3 C6
T2
∆1
g1
T1
Fig. 10.3: Translational and rotational symmetry of a single graphite sheet and demonstration of
reciprocal lattice vectors. Graphite is a layered material, which consists of sheets as the one shown.
The yellow balls indicate the carbon atoms and the blue sticks represent the bonds. In graphite these
two-dimensional sheets are stacked on top of each other, where every second sheet is shifted such
that an atom falls vertically below the point indicated by C6 . Here we only consider the symmetry
of a single sheet. The elementary unit cell is indicated by the red parallelogram, which is spanned by
the elementary lattice vectors T ~1 and T~2 . The third lattice vector points perpendicular to the sheet
towards you. An example for a general lattice vector is ~ ~1 − 2T
t3,−2,0 = 3T ~2 + 0T
~3 . The lattice planes
indicated by the dashed lines. The lattice planes are perpendicular to the sheet. The distance of the
lattice planes are indicated by ∆1 and ∆2 . The elementary reciprocal lattice vectors are ~ g1 and ~g2 .
The third reciprocal lattice vector points perpendicular out of the plane towards you. Note that the
reciprocal lattice vectors have the unit “inverse length”. Thus their length is considered irrelevant in
this real space figure. The axis through C3 standing perpendicular to the plane is a three-fold rotation
axis of the graphite crystal. The axis through C6 perpendicular to the plane is a 6-fold rotation axis of
the graphite sheet, but only a three-fold rotation axis of the graphite crystal. (Note that a rotation
axis for the graphite crystal must be one for both sheets of the crystal). In addition there is a mirror
plane lying in the plane. Furthermore there are several mirror planes perpendicular to the plane: One
passing through every atom with one of the three bonds lying in the plane and one perpendicular
each bond passing through the bond center.
10.5.3 Bloch Theorem
BLOCH THEOREM:
The eigenstates of a system with lattice symmetry can be written as product of a periodic function
~
uk (~r) and a modulating plane wave ei k~r , that is
def ~
Ψk (~r) = uk (~r)ei k~r (10.8)
where the Bloch vector ~k is a good quantum number of the System.

More precisely: Lattice translation symmetry implies that there are lattice translation vectors
T~1 , T~2 , T
~3 , so that the potential obeys V (~r + T
~i ) = V (~r). In that case the periodic function obeys
~ ~
u~k (~r + Ti ) = u~k (~r), and the Bloch vectors k lie in one reciprocal unit cell.
PROOF: Because the system has lattice symmetry, the Hamilton operator commutates with the lattice
translation operator(Eq. 9.2). As a consequence we can form common eigenstates of the lattice translation
operator and the Hamilton operator. Thus the Hamilton operator is block diagonal in the basis of eigenstates
of the translation operator (Eq. 9.3). The eigenstates of the Hamilton operator can then be obtained by
diagonalizing the Hamilton operator in the subspace of all eigenvalues of the translation operator with a
given eigenvalue of the lattice translation vector. This is to be repeated for every eigenvalue of the lattice
translation operator. All this follows directly from what we have derived already.
Now we need to find the eigenstates and eigenvalues for the lattice translation vector. The lattice translation
operator is defined by
Z
ŜT~n = d 3 r |~r + T
~n ih~r|
where T~n for n = 1, 2, 3 are the three primitive lattice translation vectors in real space. The result on a wave
function is
def
~n )
ŜT~n Ψ(~r) = h~r|ŜT~n |Ψi = Ψ(~r − T
We need not consider general translations, because the lattice translation operator for a general translation
vector ~ ~1 + j T
ti ,j,k = i T ~2 + k T
~3 is the product of the those for the primitive lattice translation vectors.
Ŝ~ti,j,k = ŜTi~1 ŜTj~ ŜTk~3

2
The lattice translation operator used for the lattice symmetry is the same as the continuous lattice translation
vector, specialized to the lattice translations. The eigenvalues for the general lattice translation vector are
already known. The eigenvalue equation is
~n ) = Ŝ ~ ψ(~r) Eq.=10.3 ψ(~r)e−i ~k T~n

ψ(~r − T Tn
~ ~
·e−i k(~r−Tn )
⇒ ~n )e−i k(~r−T~n ) = ψ(~r)e−i ~k~r
ψ(~r − T
Thus the function

def ~
uk (~r) = ψ(~r)e−i k~r (10.9)
is periodic, that is
~n ) = uk (~r)
uk (~r − T
This is already what had to be proven: The eigenstates of the lattice translation operator for a given eigenvalue
~~
e−i k Tn , and thus the eigenstates of the Hamilton operator, have the form of a product of a periodic function
~
uk (~r) and modulated by a plane wave ei k~r . The wave function has the form
Eq. 10.9 ~
Ψk (~r) = uk (~r)ei k~r
q.e.d
We can define a k-dependent Hamiltonian for the periodic part u~k (~r) of the Bloch waves.
~ ~
Ĥei k~r u~k (~r) = ei k~r u~k (~r)En,~k
~ ~
[e−i k~rĤei k~r]u~k (~r) = u~k (~r)En,~k
p~2 p + ~~k)2
(~
H̄ˆ(~k) = e−i k~r Ĥei k~r = e−i k~r [
~ ~ ~ ~
+ V (~r)]ei k~r = + V (~r)
2m 2m
2
−~ ~
= (∇ + i ~k)2 + V (~r)
2m
The energies En (~k) are called band structure. They are the dispersion relations of the particles in a
periodic potential. Band structures exist for electrons, phonons and photons.
The band structure is periodic in k-space. We show this as follows: We begin with some Bloch
wave Ψ~k,n (~r), which shall be also an eigenstate of the Hamiltonian with energy En (~k). We multiply
Fig. 10.4: Schematics of a calculated band structure of crystalline silicon [24].
~
the periodic function u~k,n (~r) with e−i G~r , which is itself periodic, and compensate by multiplication of
~r
i G~
the phase factor with e .
~ ~ ~ ~
Ψ~k,n (~r) = u~k,n (~r)ei k~r = [u~k ,n (~r)e−i G~r]ei (k+G)~r
Thus we have written the very same state as a Bloch wave
~
i (k+G)~
r ~
ψ~k +G,n
~ (~r ) = u~k+G,n
~ e
for the k-point ~k + G,

~ with a periodic part
def ~
u~k+G,n
~ (~r ) = u~k (r )e−i G~r
~ ~
and a phase ei (k+G)~r .
Because both function are identical, i.e. Ψ~k,n (~r) = Ψ~k+G,n
~ (~r ), also the energies are the same.
~
This shows that the band structure En (k) is periodic in reciprocal space, i.e
En (~k) = En (~k + G)
~
10.5.4 Schrödinger equation in momentum representation

Editor: Construction! We can use the periodicity of the band structure to simplify its graphical
representation. Let us consider a free particle. Clearly a completely translationally symmetric system
also has lattice symmetry. The dispersion relation of a free particle is a parabola.
The wave functions of the free particle have the form
~ ~
φG~ (~r) = ei G~r = h~r|Gi
and the eigenvalues are
~ 2
(~G)
EG~ =
2m
Now we can consider the Hamiltonian in plane waves. A general state can be expressed in its Fourier
~ as
representation |Gi
Z
d 3G ~ ~
|Ψi = |Gic(G)
(2π)3
Z
d 3G′ ~ ~ ′ ic(G)
~ =0
hG|Ĥ − E|G
(2π)3
Z Z
d 3G′ ~ 3 ~2 ~ 2 ~ ′ ic(G)
~ =0
hG| d r |~
r i − ∇ + V (~r) − E h~r| |G
(2π)3 2m
| {z }
Ĥ
Z Z
d 3G′ 2
~ ri − ~ ∇ ~ 2 + V (~r) − E h~r|G
~ ′ i c(G)
~ =0
d 3 r hG|~
(2π)3 | {z } 2m | {z }
~r ~
e−i G~ ei G ′ ~r
Z 3 ′ Z
d G ~r
−i G~ ~2 ~ 2 ~′ ~ =0
3
3
d re − ∇ + V (~r) − E ei G ~r c(G)
(2π) 2m
!
Z Z 2 ~′2
d 3G ′ ~ G ~′ ~ ~ =0
d 3r + V (~r) − E ei (G −G)~rc(G)
(2π)3 2m
!
Z
d 3G ′ ~′2
~2 G 3 ~ −G
(2π) δ(G ~ ′ ) + V (G ~ −G ~ ′ ) − E(2π) δ(G
3 ~ −G~ ′ ) c(G)
~ =0
(2π)3 2m
Z
~ ~
V (G) = d 3 r V (~r)ei G~r
Thus we obtain the form
2
! Z
~′
~2 G d 3G ′
−E ~ +
c(G) ~ −G
V (G ~ ′ )c(G
~ ′) = 0
2m (2π)3
~ −G
We can describe this in the form that the potential transfers a momentum ~(G ~ ′ ) to an electron.
If the potential is periodic its Fourier transform isEditor: Check
X
~ =
V (G) ~i ,j,k )δ(G
(2π)3 V (G ~ −G
~i ,j,k )
i ,j,k
Thus we obtain the form
2
!
~′
~2 G X
−E ~ +
c(G) ~i ,j,k )c(G
V (G ~ +G
~i ,j,k ) = 0
2m
i ,j,k
P
If we make the Ansatz c(G) ~ = c~ δ G ~ i ,j,k
~ − ~k + i ,j,k G we immediately observe that the
k,i ,j,k
Schrödinger equation can be solved independently for each set of coefficients with different ~k.
Thus it makes sense to draw the energies of the basis states for each ~k on top of each other.
Thus we obtain the reduced zone scheme. If we again fold out the states, we obtain the extended
zone scheme.
We will see later in the chapter on first order perturbation theory, that the energies that lie close
to each other “repel” each other. This repulsion is strongest if the energy spacing is small. This leads
to an opening of the band gap wherever two bands in the reduced zone scheme meet.
2π
a
Fig. 10.5: If a free particle experiences a potential with periodicity a, states with wave vectors that
differ by a reciprocal lattice vector Gn = 2πa n couple. The coupling shifts the upper state up and
the lower state down, which results in an opening of the gap. As a result of the coupling, the wave
vector of a free particle state is no more a good quantum number. However, the wave vector in the
reduced zone scheme remains to be a good quantum number.
10.6 Some common Lattices

10.6.1 Cubic lattices
a a a
a a a
a a a
Fig. 10.6: The simple-cubic, the face-centered cubic and the body-centered cubic lattice in the simple
cubic unit cell.
The most simple lattice is the simple cubic (sic) lattice which has lattice vectors
     
1 0 0
~1 = 
T 0
 

a , ~
T 2 =
 
 1 a , ~
T 3 =
 
0a
0 0 1
where a is the lattice constant. I am not aware of any crystal with one atom per unit cell that is
stable in the simple cubic lattice.
More common is the face-centered cubic (fcc) lattice with lattice vectors
     
1 1
0 2 2
T~1 =  1  a ,
  ~2 =  0  a ,
T
  ~3 =  1 
T

2 2 a
1 1
2 2 0
The face centered cubic lattice corresponds to one of the possible lattices of closed packed spheres.
It is very common for atoms which do not form directed bonds. One example are noble gases, when
they freeze. Another are metals such as Sodium, Aluminium, Gold, Copper, Silver. Austenite, the
major component of Steel is another example for a fcc-crystal.
The third cubic lattice is the body-centered cubic (bcc) lattice with lattice vectors
     
− 21 1
2
1
2
~1 = 
T  12  a ,
 ~2 = 
T  − 12  a ,

T~3 = 
 12  a

1 1
2 2 − 21
The body centered cubic lattice is found for metals, that form some extend of directed bonds, such
as the transition metals in the middle of the transition metal row. One example is iron in its magnetic
form
10.7 Exercises
10.7.1 Symmetrization of states
Consider three hydrogen atoms placed at the corners of an equilateral triangle. The wave function of
one atom at the origin ψ(~r) = e−λ|~r| . Construct eigenstates of the rotation operator about the three-
fold rotation axis, which are superpositions of the states of the three atoms. Only one wave function
per atom is considered. Provide the coefficients of the three eigenstates and their eigenvalues.
10.7.2 Dirac comb

Note: This exercise is very instructive, but cannot completely solved analytically. Furthermore it
must be known how to solve the Schrödinger equation for a Deltafunction potential.
The Dirac comb is a one-dimensional model system for a crystal of atoms. The potential of a
single atom is simply modeled by a δ-function vn (x) = −V0 δ (x − na), where n is the lattice constant.
The wave function in a δ-function like potential has a step in the derivative, which is obtained by
inserting it in the one-dimensional Schrödinger equation

~2 2
− ∂ + V (x) ψk, n(x) = ψ(x)Ek,n
2m x
Use the Bloch theorem to determine the eigenstates and the band structure of the Dirac comb.
Solution
The potential is
∞
X ∞
X
v (x) = vn (x) = −V0 δ(x − na)
n=−∞ n=−∞
Let us consider a single dirac potential at the origin.

Z ǫ 2
−~ 2
dx ∂ − V0 δ(x) − E ψ(x) = 0
−ǫ 2m x
Because the second derivative of the wave function contains a δ function, which follows from the
Schrödinger equation, the first derivative contains a step function and the wave function itself is
continuous with a step in the first derivative. Therefore, we first match the two partial solutions such
that the resulting wave function is continuos.
lim ψ(x) = lim+ ψ(x)

x→0− x→0
where 0− denotes that the origin is approached from the left and 0+ denotes that the origin is
approached from the right.
The trick to deal with the delta function potential is to integrate the Schrödinger equation from
a small negative value −ǫ to a small positive value +ǫ.
Z ǫ 2
−~ 2
0= dx ∂x − V0 δ(x) − E ψ(x)
−ǫ 2m
2 Z ǫ Z ǫ Z ǫ
−~ 2
= dx ∂x ψ(x) − V0 dx δ(x)ψ(x) − E dx ψ(x)
2m −ǫ −ǫ −ǫ
Z ǫ
−~2
= ∂x |ǫ ψ − ∂x |−ǫ ψ − V0 ψ(0) − E dx ψ(x)
2m −ǫ
Now we let ǫ approach zero. In this limiting case ǫ → 0, the last term vanishes. (Note that the wave
function is still continuous as shown above.) Thus we obtain
2mV0
∂x |0+ ψ − ∂x |−0+ ψ = − ψ(0)
~2
Now we can approach the crystal: We use Bloch theorem to express the wave function by a
periodic part and a phase factor
ψk (x) = uk (x)ei kx
The Schrödinger equation for the periodic part is
" #
~2 2 X
0= − ∂ − V0 δ(x − na) − E uk (x)ei kx
2m x n
" #
2
~ X
= ei kx − (∂x + i k)2 − V0 δ(x − na) − E uk (x) = 0
2m n
" #
~2 2
X
⇒ − (∂x + i k) − V0 δ(x − na) − E uk (x) = 0
2m n
Now we play the same trick as for the single delta-potential. Like ψ(x) also uk (x) is continuous.
We integrate from ǫ to +ǫ:
Z ǫ " #
~2 2
X
0= dx − (∂x + i k) − V0 δ(x − na) − E uk (x)
−ǫ 2m n
Z ǫ " #
~2 2mV 0
X 2mE
=− dx ∂x2 + 2i k∂x − k 2 + δ(x − na) + 2 uk (x)
2m −ǫ ~2 n
~
~2 h
=− ∂x |−ǫ uk − ∂x |+ǫ uk + 2i k (uk (ǫ) − uk (−ǫ))
2m Z ǫ Z ǫ
2mV0 X 2mE i
2
+ dx δ(x − na)u k (x) + + k dxu k (x)
~2 n −ǫ ~2 −ǫ
2 h
~
=− ∂x |−ǫ uk − ∂x |+ǫ uk + 2i k (uk (ǫ) − uk (−ǫ))
2m | {z }
→0
Z ǫ i
2mV0 2mE 2
+ 2
uk (0) + 2
+k dx uk (x)
~ ~
| −ǫ {z }
→0
Now we let ǫ approach zero:

2mV0
∂x |−ǫ uk − ∂x |+ǫ uk = − uk (0)
~2
Now we concentrate on the wave function in between the delta potentials: In between the δ
potentials, the wave function has the form of a plane wave.
By insertion into the Schrödinger equation in the region between the atoms we obtain G as
function of the energy.

−~2
(∂x + i k) − E ei Ḡx = 0
2
2m
2
−~
(i G + i k) − E ei Ḡx = 0
2
2m
1√
Ḡ1/2 (E) = −k ± 2mE
|~ {z }
=:G
Thus the periodic part uk (x) has the general form

uk (x) = Aei Ḡ1 x + Bei Ḡ2 x = Aei (G−k)x + Be−i (G+k)x = Aei Gx + Be−i Gx e−i kx
def√
We have defined G = 2mE
Now we consider the matching of the two solutions at the origin, where the delta potential resides.
uk (+ǫ) = uk (−ǫ) = uk (a − ǫ)
2mV0
∂x |ǫ uk − ∂x |a−ǫ uk = − 2 uk (0)
~
If we insert our ansatz we obtain two conditions for our coefficients:
The condition of continuity yields
A + B = Aei (G−k)a + Be−i (G+k)a

1 − ei (G−k)a
⇒B=− A
1 − e−i (G+k)a
Thus the wave function has the form

h i
uk (x) = C 1 − e−i (G+k)a ei (G−k)x − 1 − ei (G−k)a e−i (G+k)x
where C is a normalization constant.

Next we consider the condition from the step in the first derivative.
2mV0
i (G − k)A − i (G + k)B = i (G − k)Aei (G−k)a − i (G + k)Be−i (G+k)a − (A + B)
~2
2mV
0
⇒ 0 = i (G − k) −1 + ei (G−k)a − A
~2
2mV
0
+ i (G + k) 1 − e−i (G+k)a − B
~2
Now we insert the result for B into the last equation and obtain an equation for G.
2mV
i (G−k)a 0
0 = i (G − k) −1 + e − A
~2
2mV 1 − ei (G−k)a
−i (G+k)a 0
− i (G + k) 1 − e − A
~2 −i (G+k)a
|1 − e {z }
−B
2mV
i (G−k)a 0 −i (G+k)a
0 = i (G − k) −1 + e − 1 − e
~2
2mV
0
− i (G + k) 1 − e−i (G+k)a − 1 − e i (G−k)a
~2
2mV
i (G−k)a −i (G+k)a 0 −i (G+k)a
= −i (G − k) 1 − e 1−e − 1−e
~2
2mV
−i (G+k)a i (G−k)a 0 i (G−k)a
− i (G + k) 1 − e 1−e − 1−e
~2
2mV
0
= −2i G 1 − ei (G−k)a 1 − e−i (G+k)a − 2
ei (G−k)a − e−i (G+k)a
~
i Ga −i Ga
−i ka −2i ka
2mV0 i Ga −i Ga
−i ka
= −2i G 1 − e + e e +e − e − e e
~2
ika
·e /(4i ) 1 mV0 1 i Ga
⇒ 0 = G ei ka − ei Ga + e−i Ga + e−i ka + 2 e − e−i Ga
2 ~ 2i
mV0
= G (cos(ka) − cos(Ga)) + 2 sin(Ga)
~
mV0 a
= G cos(ka) − cos(Ga) + 2 sin(Ga)
~ Ga
mV0 a
⇒ cos(ka) = cos(Ga) − 2 sin(Ga)
| {z Ga
~ }
Z(Ga)
We defined the function
def mV0 a
Z(x) = cos(x) − sin(x)
~2 x
It seems to be difficult to resolve the equation for G(k). However it is more straight forward to
determine inverse function k(G).
1 a√
k(E) = arccos Z( 2mE)
a ~
Thus we need to determine the zeros of the function Zk (G).

The figure shows the function Zπ/2 (Ga) for mV 0a
~2 = 5. We have chosen k = π/2 because then
the cos(ka) drops out. States can only exist where |Zk | < 1. Thus we can directly determine the
band edges.
5
4
3
2
Zπ/2 (G) 1
0
−1
−2
−3
0 5 10 15 20
Ga
Note that we also need to consider possible states with negative energies! In that case we need
to replace G = i λ by an imaginary quantity. We use
e−x − ex 1
sin(i x) = = − sinh(x) = i sinh(x)
2i i
The function Z(i λ) has the form
mV0
Z(i λ) = cosh(λa) − sinh(λa)
~2 λ
10
8
6
Zπ/2 (i λ) 4
2
0
−2
0 1 2 3 4 5 6
λa
We see immediately that there is a very narrow band at energy E = −5, which corresponds to the
bound states for the isolated atom.
PROGRAM MAIN
IMPLICIT NONE
INTEGER(4),PARAMETER :: NE=1000 ! #(ENERGY GRID POINTS)
REAL(8) ,PARAMETER :: EMIN=-13.D0 ! LOWER BOUND OF ENERGY WINDOW
REAL(8) ,PARAMETER :: EMAX=100.D0 ! UPPER BOUND OF ENERGY WINDOW
REAL(8) ,PARAMETER :: A=1.D0 ! LATTICE CONSTANT
REAL(8) ,PARAMETER :: HBAR=1.D0 ! HBAR
REAL(8) ,PARAMETER :: M=1.D0 ! MASS
REAL(8) ,PARAMETER :: V=5.D0 !
INTEGER(4) :: I
REAL(8) :: E,G,Y,Z,K
REAL(8) :: PI,G0
! *******************************************************************
PI=4.D0*DATAN(1.D0) ! CALCULATE PI
G0=2.D0*PI/A ! RECIPROCAL LATTICE VECTOR
DO I=1,NE
E=EMIN+(EMAX-EMIN)/REAL(NE-1)*REAL(I-1)
IF(E.GT.0) THEN ! SELECT SCATTERING SOLUTION
G=SQRT(2.D0*M*E) ! PSI(X)=A*E^(I*G*X)+B*E^(-I*G*X)+B
Y=G*A ! SCALED VARIABLE
Z=COS(Y)-M*V*A/(HBAR**2)*SIN(Y)/Y ! Z(G*A)
ELSE ! SELECT BOUND SOLUTION
G=SQRT(-2.D0*M*E) ! PSI(X)=A*E^(I*G*X)+B*E^(-I*G*X)+B
Y=G*A ! SCALED VARIABLE
Z=COSH(Y)-M*V*A/(HBAR**2)*SINH(Y)/Y !Z(I*LAMBDA*A)
END IF
! == IN THE FORBIDDEN REGION WE SET Z TO THE MAXIMUM OR MINIMUM
! == ALLOWED VALUE. THE RESULT WILL OCCUR AT THE BOUNDARY OF THE
! == FIGURE OR ON THE E-AXIS AND WILL BE HIDDEN BEHIND THE FRAME

Z=MIN(Z,1.D0) ! SET Z TO MAXIMUM ALLOWED VALUE
Z=MAX(Z,-1.D0) ! SET Z TO MINIMUM ALLOWED VALUE
K=ACOS(Z)/A ! CALCULATE K(Z)
IF(K.GT.0.5D0*G0)K=K-G0 ! MAP K INTO FIRST BRILLOUIN ZONE
PRINT*,E,K/G0,-K/G0 ! PRINT RESULT WITH K IN UNITS OF G0
ENDDO
STOP
END
We can now determine the band structure
100
90
80
70
60
50
40
30
20
10
-10
-20
Fig. 10.7: Band structure En (k) of the Dirac comb with potential V0 = −5. The horizontal axis
ranges from −g/2 to +g/2, where g = 2π/a is the primitive reciprocal lattice vector.The state at
negative energies corresponds to the bound states of the atoms, while those at positive energies
are scattering states. Compare this dispersion relation with the folded dispersion relation of a free
particle (parabola).
Chapter 11
Rotations, Angular Momentum and

Spin
Systems with rotational symmetry play such an important role, that we devote a separate chapter to
them, even though what is done here is only an extension of what we learned in the section about
symmetry.
11.1 Rotations
Let us consider first a continuous rotation about the z-axis. The symmetry operator transforms ~r
into r~′ = R (ϕ)~r, where R (ϕ) is the corresponding rotation matrix.
! ! !
x′ cos(ϕ) − sin(ϕ) x
=
y′ sin(ϕ) cos(ϕ) y
| {z }
R
The angle is positive for counterclockwise rotation.

A rotation of a wave function by an angle ϕ is then defined as
Z
def
Ŝ(ϕ) = d 3 r |R
R (ϕ)~rih~r| (11.1)
Z Z
h~r|Ŝ(ϕ)|ψi = d 3 r h~r|R
R~rih~r|ψi = R −1~r|~rih~r|ψi = hR
d 3 r hR R −1~r|ψi
| {z } | {z }
Ŝ(ϕ)ψ(~
r) R −1 ~
ψ(R r)
Note, that we need to use the inverse rotation, because we obtain the value of the rotated wave
function at a point r0 by looking up the value of the original wave function at the position which
fell into r0 by the rotation. The inverse of the rotation is simply its transpose, because the rotation
matrix is unitary.
171
172 11 ROTATIONS, ANGULAR MOMENTUM AND SPIN
11.1.1 Derivation of the Angular momentum

Let us construct the canonical conjugate momentum L to the angle ϕ. We will see that L is the
angular momentum.
Ŝ(ϕ) − 1̂
L̂z ψ(x, y ) = i ~ lim ψ(x, y )
ϕ→0 ϕ
1+O(ϕ2 ) ϕ+O(ϕ2 ) ϕ+O(ϕ2 ) 1+O(ϕ2 )
z }| { z }| { z }| { z }| {
ψ[ cos(ϕ) x + sin(ϕ) y , − sin(ϕ) x + cos(ϕ) y ] − ψ[x, y ]
= i ~ lim
ϕ→0 ϕ
ψ(x,y )+(x+ϕy +O(ϕ)2 )∂x ψ+(y −ϕx+O(ϕ)2 )∂y ψ+...
z }| {
ψ[x + ϕy + O(ϕ2 ), y − ϕx + O(ϕ2 )] −ψ[x, y ]
= i ~ lim
ϕ→0 ϕ
ϕy ∂ ψ − ϕx∂ ψ + O(ϕ2 )
x y
= i ~ lim
ϕ→0 ϕ
~
= −y ∂x + x∂y ψ
i
L̂z = x̂ p̂y − ŷ p̂x
Thus we obtained the form of the expression of the angular-momentum operator.

Editor: Use the above example as exercise for a Taylor expansion
We can generalize this result into three dimensions for a rotation about an axis, which is described
by a vector ϕ.
~ The direction of the vector describes the rotation axis and the length of the vector
denotes the angle of rotation.
Definition 11.1 ANGULAR-MOMENTUM OPERATOR
~ˆ has the form.
The angular-momentum operator L
~ = ~rˆ × p̂~
L̂ (11.2)
With the angular-momentum operator we can express the rotation operator Ŝ(ϕ) as
ROTATION OPERATOR
Z
Eq. 11.1 Eq. 9.6 i
Ŝ(ϕ) = d 3 r |R
R~rih~r| = e− ~ L̂x ϕ (11.3)
11.1.2 Another derivation of the angular momentum (Home study)

This derivation is more complex but may also show some aspects more clearly. It does not refer
to wave functions and rests entirely within Dirac’s bracket notation. In any case it is a very good
exercise for the latter.
Z Z
1 1 3 3
L̂z = i ~ lim Ŝ(ϕ) − 1̂ = i ~ lim R(ϕ)~rih~r| − d r |~rih~r|
d r |R
ϕ→0 ϕ ϕ→0 ϕ
Z
1
= i ~ d 3 r lim [|R R (ϕ)~ri − |~ri] h~r
ϕ→0 ϕ
11 ROTATIONS, ANGULAR MOMENTUM AND SPIN 173
Next we Taylor-expand the rotated vector R (ϕ)~r up to first order in ϕ.

   
cos(ϕ) − sin(ϕ) 0 1 + O(ϕ2 ) −ϕ + O(ϕ2 ) 0
   
R (ϕ) =  sin(ϕ) cos(ϕ) 0  =  ϕ + O(ϕ2 ) 1 + O(ϕ2 ) 0 
0 0 1 0 0 1
 
0 −1 0
= 1 +  1 0 0  ϕ + O(ϕ2 )
 
0 0 0
 
−y
⇒ R~r = ~r +  x  ϕ + O(ϕ2 ) = ~r + ϕ~ ez × ~r + O(ϕ)2
 
0
R
~ def
Let us now use the translation operator T̂ (∆) ~ r | to express the rotated state
= d 3 r |~r + ∆ih~
R (ϕ)~ri
|R = e × ~r + O(ϕ2 )i = T̂ (ϕ~
|~r + ϕ~ e × ~r + O(ϕ2 ))|~ri
| z {z } | z {z }
∆ ∆
∆
z }| {
Eq. 9.6
=
i
− ~ p~ˆ ez × ~r + O(ϕ2 ) |~ri
ϕ~
|e {z }
T̂ (∆)

T ay l or i ˆ 2
= 1̂ − p~ϕ(~ ez × ~r) + O(ϕ ) |~ri
~

~(~
a c )=~
b×~ c ×~
b(~ a) i
= 1̂ − ~ez (~r × p~ˆ)ϕ + O(ϕ2 ) |~ri
~
Thus we obtain
Z Z
i
L̂z = i ~ d 3 r lim − ~ ez ( d 3 r |~ri~rh~r| ×p~ˆ)
ez (~r × p~ˆ)ϕ + O(ϕ2 ) |~rih~r| = ~
ϕ→0 ~
| {z } | {z }
1
~
i~ z
~ˆ)
r ×p
e (~ rˆ
~
ez (~rˆ × p~ˆ)
=~
Since we can choose our coordinate system arbitrarily we can immediately generalize the result directly
to arbitrary directions of the rotation axis, which identifies the angular-momentum operator as
~ˆ = ~rˆ × p~ˆ
L
11.2 Commutator Relations

It turns out that the components of the angular momentum operator do not commutate with each
other. Therefore, it is not possible to determine the direction of the angular momentum with certainty.
Let us now explore the commutator relations among the angular momentum operators.
[L̂x , L̂y ] = [ŷ p̂z − ẑ p̂y , ẑ p̂x − x̂ p̂z ]

= y p̂z ẑ p̂x − ŷ p̂z x̂ p̂z − ẑ p̂y ẑ p̂x + ẑ p̂y x̂ p̂z
− ẑ p̂x ŷ p̂z + ẑ p̂x ẑ p̂y + x̂ p̂z ŷ p̂z − x̂ p̂z ẑ p̂y
= ŷ p̂x [p̂z , ẑ ] + p̂y x̂[ẑ , p̂z ] = i ~(x̂ p̂y − ŷ p̂x )
= i ~L̂z
The other commutators are obtained by cyclic permutation of indices.
[L̂x , L̂y ] = i ~L̂z ; [L̂y , L̂z ] = i ~L̂x ; [L̂z , L̂x ] = i ~L̂y
~ 2 commutates with all components

Furthermore we can show that L̂
~ˆ2 , L̂x ] = [L̂2x , L̂x ] +[L̂2y , L̂x ] + [L̂2z , L̂x ]

[L
| {z }
=0
=0 =0
z }| { z }| {
= L̂2y L̂x −L̂y L̂x L̂y + L̂y L̂x L̂y −L̂x L̂2y + L̂2z L̂x −L̂z L̂x L̂z + L̂z L̂x L̂z −L̂x L̂2z
= L̂y [L̂y , L̂x ] + [L̂y , L̂x ]L̂y + L̂z [L̂z , L̂x ] + [L̂z , L̂x ]L̂z

= i ~ −L̂y L̂z − L̂z L̂y + L̂z L̂y + L̂y L̂z = 0
Cyclic permutation of indices yields the commutator relations with the remaining angular momentum
components. Thus we can measure the amplitude of the angular momentum and one component of
the angular momentum simultaneously.
[L̂i , L̂x ] [L̂i , L̂y ] [L̂i , L̂z ] ~ˆ2 ]

[L̂i , L
[L̂x , L̂i ] 0 i ~L̂z −i ~L̂y 0
[L̂y , L̂i ] −i ~L̂z 0 i ~L̂x 0
[L̂z , L̂i ] i ~L̂y −i ~L̂x 0 0
In short, we can write the result using the Levi-Civita Symbol1 as

COMMUTATOR RELATION OF ANGULAR MOMENTA
X
[L̂i , L̂j ] = i ~ ǫi ,j,k L̂k (11.4)
k
11.3 Algebraic derivation of the eigenvalue spectrum

In the following we will show the algebraic description of angular momentum, which will provide us
with the eigenvalues of angular momenta. The algebraic description is important, because we will
find eigenstates, that cannot be represented by a single-valued wave function. Any Ansatz of a real
space wave function will therefore fail to produce those eigenstates.
Shift operators revisited

We already made contact with shift operators in connection with continuous translations. There,
it was possible use a shift operator to create new eigenstates from an existing eigenstate with a
shifted eigenvalue. Here we use shift operators, that are not unitary operators and therefore are no
symmetry transformations. However, we can still use them to create new eigenstates of an operator.
The concept of shift operators to create new eigenstates is analogous to that of the creation and
1 The Levi-Civita symbol is the fully antisymmetric tensor defined as being equal to one if the indices are ascending
ǫ1,2,3 = ǫ2,3,1 = ǫ3,1,2 = 1 am being −1 if the indices are descending, that is ǫ3,2,1 = ǫ1,3,2 = ǫ2,1,3 = −11. If
two indicesP are equal the Levi-Civita symbol vanishes. The Levi-Civita Symbol is used to describe the cross-product
a × ~b)i = j,k ǫi ,j,k aj bk .
(~
annihilation operators, that have been used in the algebraic treatment of the harmonic oscillator and
the two state system.
A shift operator Â has a commutator with the shifted operator B̂, which is proportional to the
shift operator itself. As a result, the shift operator produces a new eigenstate Â|ψb i of B̂ from an
existing one, |ψb i with a shifted eigenvalue, that is

[Â, B̂] = c Â and B̂|ψb i = |ψb ib ⇒ B̂ Â|ψb i = Â|ψb i (b − c)
We can confirm this easily:
B̂ Â|ψb i = Â B̂|ψb i −c Â|ψi

| {z }
|ψb ib

B̂ Â|ψb i = Â|ψb i (b − c)
Furthermore we can show that [Â† , B̂] = −c Â† if B̂ is hermitian and c is a real number.
†
B=B † c=c ∗
[Â† , B̂] = Â† B̂ − B̂ Â† = B̂ † Â − ÂB̂ † = −[Â, B̂]† = −c Â†
Hence we can use Â† to create an eigenstate of B̂ with eigenvalue b + c

B̂ Â† |Ψi = Â† |Ψi (b + c)
Thus we can produce a whole series of eigenstates of B̂ with equi-spaced eigenvalues, if we can
only find an operator with the property [Â, B̂] = c Â, where c is real. The series of eigenvalues is
bound from below and above if there is a solution to Â|ψ0 i = 0 and to Â†(n+1) |ψ0 i = 0.
Shift operators for L̂z

Such a shift operator Â is a handy thing to have. Let us look for one that may be useful for angular
momentum operators. We consider
[L̂y , L̂z ] = i ~L̂x

[L̂z , L̂x ] = i ~L̂y
which suggests that an operator L̂− = aL̂x + bL̂y may do the job. If we use this as Ansatz for
[L̂− , L̂z ] = c L̂− , we find one solution2 with a = 1, b = −i and C = ~.
Thus we define the shift operators
L̂− = L̂x − i L̂y (11.5)

L̂+ = L̂x + i L̂y = L̂†− (11.6)
They obey the commutation relations
[L̂− , L̂z ] = ~L̂− ; [L̂+ , L̂z ] = −~L̂+ ; ~ˆ2 ] = 0;

[L̂− , L ~ˆ2 ] = 0
[L̂+ , L
2 We use the ansatz L̂− = aL̂x + bL̂y and require
[L̂− , L̂z ] = C L̂−
!
⇒ [aL̂x + bL̂y , L̂z ]− = a[L̂x , L̂z ]− + b[L̂y , L̂z ]− = −i~aL̂y + i~bL̂x = ib~L̂x − ia~L̂y = C aL̂x + bL̂y
we obtain two conditions, namely aC = ib~ and bC = −ia~. Thus C = i~ ba = −i~ ba , which requires b2 = −a2 and
therefore b = ±ia. If we choose a = 1 and b = −ia, then C = ~. The other solution leads to L+ with an arbitrary
scale factor.
The first two equations identify the operators L̂+ and L̂− as shift operators for eigenstates of
L̂z . Thus for an eigenstate |µi of L̂z with eigenvalue µ, the state L̂+ |µi is an eigenstate of L̂z with
eigenvalue µ + ~. Similarly the state L̂− |µi is an eigenstate of L̂z with eigenvalue µ − ~, i.e.


 Lˆz L̂+ |µi = L̂+ |µi(µ + ~)
 | {z } | {z }

|µ+~i |µ+~i
Lˆz |µi = |µiµ ⇒ (11.7)

 ˆ
Lz L̂− |µi = L̂− |µi(µ − ~)

 | {z } | {z }
|µ−~i |µ−~i
~ˆ2 by L̂+ , L̂− and L̂z , which we will need later.

We can now express L
L̂+ L̂− = (L̂x + i L̂y )(L̂x − i L̂y ) = L̂2x + L̂2y −i [L̂x , L̂y ]
| {z } | {z }
~ 2 −L2z
L i ~Lz
=L~ˆ2 − L̂2z + ~L̂z (11.8)

L̂− L̂+ = L~ˆ2 − L̂2z − ~L̂z (11.9)
~ˆ2 = L̂2z + 1 (L̂− L̂+ + L̂+ L̂− )
L (11.10)
2
Angular-momentum eigenvalue spectrum

Now we are ready to explore the spectrum of eigenvalues. Let us now assume we have eigenstates3
~ˆ2 and L̂z , so that
|λ, µi to L
~ˆ2 |λ, µi = |λ, µiλ

L (11.11)
L̂z |λ, µi = |λ, µiµ. (11.12)
hλ, µ|λ′ , µ′ i = δλ,λ′ δµ,µ′ (11.13)
The requirement that the norm of L̂± |λ, µi is positive provides us with bounds for the spectrum
of eigenvalues of L̂z . This, somewhat involved, derivation will be shown in the following:
~ˆ2 and L̂ . We form two new states, namely
Let us consider an eigenstate |λ, µi of L z
def
|φ+
λ,µ i = L̂+ |λ, µi
def
|φ−
λ,µ i = L̂− |λ, µi
by operating once with L̂+ and once with L̂− on it. Because the scalar product of any state must be
positive, we obtain two inequalities.
0 ≤ hφ+ +
λ,µ |φλ,µ i
= hλ, µ|L̂†+ L̂+ |λ, µi

L̂†+ =L̂−
= hλ, µ|L̂− L̂+ |λ, µi
Eq. 11.9
= ~ˆ2 − L̂2z − ~L̂z |λ, µi
hλ, µ|L
Eq. 11.11,11.12,11.13
= λ − µ2 − ~µ
⇒ µ2 + ~µ ≤ λ (11.14)
The equal sign holds exactly when |φ+

λ,µ i is the zero state.
3 Note that |λ, µi should be read as |Ψ
λ,µ i. The symbol Ψ seems superfluous– it is the same symbol for all members
in the specified set of states– it is normally dropped. The new notation just includes the indices in the ket symbol. For
example we would write Ψλµ (~ r ) = h~
r |λ, µi
Completely analogously we obtain the second inequality from the condition that the scalar product
of |φ−
λ,µ i with itself is positive.
0 ≤ hφ− −
λ,µ |φλ,µ i
= hλ, µ|L̂†− L̂− |λ, µi

L̂†− =L̂+
= hλ, µ|L̂+ L̂− |λ, µi
Eq. 11.8
= ~ˆ2 − L̂2z + ~L̂z |λ, µi
hλ, µ|L
Eqs. 11.11,11.12,11.13
= λ − µ2 + ~µ
⇒ µ2 − ~µ ≤ λ (11.15)
Again, the equal sign holds exactly when |φ−

λ,µ i is the zero state.
The two inequalities are graphically shown in figure 11.1. Let us assume that there is a state
µ(µ+h) µ(µ−h)
λ
h h h
h h
−h h µ
Fig. 11.1: Demonstration of the allowed values for λ and µ. The inequalities Eq. 11.14 and Eq. 11.15
require that the allowed values of (λ, µ) lie in the white region above both parabolas. The ladder of
µ values terminates towards the right only on the left parabola, for which L̂+ |λ, µi = |∅i. Similarly
it terminates towards the left only on the right parabola. For a bound set of µ values each ladder
must start and end at the boundary of the white region. The discrete spacing, Eq. 11.7, of µ values
by ~ specifies the allowed values of µ.
|λ, µi in the allowed region, that is both inequalities are fulfilled. This state is represented by a point
in the white region of the diagram above. We have shown that each application of L̂− creates a
new state with µ reduced by ~. This creates a new point shifted to the left by ~. If we repeatedly
apply L̂− we create a point in the forbidden region. This however would contradict the inequalities.
However, if we exactly hit the boundary, we are fine, because the next state is a zero state. That is,
there is no next state, and therefore there is no contradiction. This consideration tells us that the
only allowed states |λ, µi must be located a multiple of ~ to the right of the left boundary.
Now let us perform the same steps but we operate with L̂+ on our state |λ, µi. L̂+ shifts the
point in the graph by ~ to the right. The contradiction given by the inequalities can only be avoided
if the horizontal distance between left and right boundary is exactly a multiple of ~. This condition
immediately provides us with the allowed values for λ.
We will now express the same by formulas in order to determine the allowed values for λ and µ:
The inequalities Eq. 11.14 and Eq. 11.15 require µmi n ≤ µ ≤ µmax with
Eqs. 11.14,11.15
µ2mi n/max ± ~ µmi n/max − λ = 0
2
~ 2 ~
⇒ µmi n/max ± = +λ
2 2
 s 
2
~ ~
⇒ µmi n/max (λ) = ± ∓ + λ
2 2
The choice of the second sign ∓ determines, whether we choose the inner or the outer branches of
the two parabola in Fig. 11.1. Because λ is positive, as seen in Fig. 11.1, the negative sign yields
the smaller absolute value for µmi n/max .
Thus, the the lowest allowed value µmi n of µ and the highest allowed value µmax of µ for a given
λ is
r !
~ ~ 2
µmi n (λ) = + − ( ) +λ
2 2
r !
~ ~ 2
µmax (λ) = − − ( ) +λ (11.16)
2 2
Note, that the value of the expression in parenthesis is negative.

The requirement that these values differ by a multiple of ~, obtained from Eq. 11.7, yields
µmi n − µmax = ~n
where n is an arbitrary non-negative integer. Thus we obtain

r r
~ ~ 2 ~ ~ p
− + ( ) + λ − + ( )2 + λ = −~ + ~2 + 4λ = ~n
2 2 2 2
| {z }| {z }
µmax (λ) −µmin (λ)
~2 (n + 1)2 = ~2 + 4λ
(n + 1)2 − 1 n2 + 2n n n
λ = ~2 = ~2 = ~2 [( )2 + ]
4 4 2 2
λ = ~2 ℓ(ℓ + 1)
where we replaced n/2 by ℓ. ℓ can assume half-integer values ℓ = 0, 12 , 1, 32 , 2, . . .. Thus we have

~ˆ2 .
obtained the eigenvalue spectrum of L
The eigenvalues for L̂z are obtained for each ℓ from
r r
~ ~ 2 ~ ~
µmi n = − ( ) + λ = − ( )2 + ~2 ℓ(ℓ + 1)
2 2 2 2
r r
~ 1 ~ 1 ~ 1
= − ~ (ℓ + ℓ + ) = − ~2 (ℓ + )2 = − ~(ℓ + )
2 2
2 4 2 2 2 2
= −~ℓ
Now we change our notation. Instead of characterizing a state by its eigenvalues λ = ~2 ℓ(ℓ + 1)
and µ = ~m as |λ, µi, we use the angular quantum numbers ℓ and m as |ℓ, mi. Thus we denote the
eigenstates of the angular momentum as |λ, µi = |ℓ, mi.
EIGENVALUE SPECTRUM OF THE ANGULAR MOMENTUM
~ˆ2 |ℓ, mi = |ℓ, mi~2 ℓ(ℓ + 1)

L
L̂z |ℓ, mi = |ℓ, mi~m
where the quantum numbers ℓ = 0, 12 , 1, 23 , . . . can have any non-negative, half-integer value and
m = −ℓ, −ℓ + 1, . . . , ℓ − 1, ℓ.
L h
h
h l(l+
1)
h
p
Fig. 11.2: The absolute value of the angular momentum can have the values ~ ℓ(ℓ + 1). The
eigenvalues of L̂z can assume the values ~m, where m = −ℓ, −ℓ + 1, . . . , ℓ − 1, ℓ. In an eigenstate the
values of L̂z and L̂y cannot be determined. It is as if the angular momentum would be precessing.
However, the angular momentum does not move, but it simply is delocalized on the circumference
~ˆ2 − L̂2z does have sharp values.
of the circles shown. Note that L̂2x + L̂2y = L
Recursion relation for eigenstates
So far we have obtained the eigenvalues and a procedure to create new eigenstates for different
eigenvalues of L̂z if one eigenstate is known. However these states are not automatically normalized.
Here we want to renormalize the shift operators so that they create normalized states from normalized
states
~ˆ2 − L̂2z − ~L̂z |ℓ, mi

hℓ, m|L̂†+ L̂+ |ℓ, mi = hℓ, m|L
= hℓ, m|ℓ, mi(~2 ℓ(ℓ + 1) − ~2 m2 − ~2 m)
= hℓ, m|ℓ, mi~2 (ℓ(ℓ + 1) − m(m + 1))
1
|ℓ, m + 1i = L̂+ |ℓ, mi p
~ ℓ(ℓ + 1) − m(m + 1)
1
= L̂+ |ℓ, mi p
~ (ℓ − m)(ℓ + m + 1)
Similarly we obtain the renormalization for L̂−
~ˆ2 − L̂2z + ~L̂z |ℓ, mi

hℓ, m|L̂†− L̂− |ℓ, mi = hℓ, m|L
= hℓ, m|ℓ, mi(~2 ℓ(ℓ + 1) − ~2 m2 + ~2 m)
= hℓ, m|ℓ, mi~2 (ℓ(ℓ + 1) − m(m − 1))
1
|ℓ, m − 1i = L̂− |ℓ, mi p
~ ℓ(ℓ + 1) − m(m − 1)
1
= L̂− |ℓ, mi p
~ (ℓ + m)(ℓ − m + 1)
Thus if we know just one eigenstate of the angular momentum for each ℓ we can construct all
others by the recursion relation above.
Let us summarize here the effect of the angular momentum operators on the basis of eigenstates
~ˆ2 and L̂z :
of L
~ˆ2 |ℓ, mi = |ℓ, mi~2 ℓ(ℓ + 1)

L (11.17)
L̂z |ℓ, mi = |ℓ, mi~m (11.18)
p
L̂+ |ℓ, mi = |ℓ, m + 1i~ (ℓ − m)(ℓ + m + 1) (11.19)
p
L̂− |ℓ, mi = |ℓ, m − 1i~ (ℓ + m)(ℓ − m + 1) (11.20)
Because we can express all angular momentum components by L̂z , L̂+ , L̂− , we can derive the matrix
elements of any operator composed from angular momenta just from these four relations.4
11.4 Eigenstates of angular momentum: spherical harmonics
~ˆ2 , which are called

In the following we will derive the functional form of the eigenstates of L̂z and L
spherical harmonics.
Let us introduce polar coordinates:
p
r= x2 + y 2 + z2
x = r cos(φ) sin(θ) z
θ = acos( )
y = r sin(φ) sin(θ) r
x
z = r cos(θ) φ = acos( p )sgn(y )
x2 + y2
with θ ∈ [0, 2π] and φ ∈ [−π, π]
4 The first of these relations is actually redundant, and follows from the remaining three.
θ y
ϕ
x
Using Eq. 9.6, we can express the rotation operator Ŝ(α)

hr, θ, φ|Ŝ(α)|ψi = hr, θ, φ − α|ψi (11.21)
| {z }
ψm (r,θ,φ−α)
which shifts the angle by φ about the z-axis in the form

Eq. 11.3 i
Ŝ(φ) = e− ~ L̂z φ (11.22)
If the shift operator is applied to an eigenstate |ψm i of the z-component of the angular momentum
L̂z with eigenvalue ~m, we obtain
Eq. 11.22 i L̂z |ψm i=|ψm i~m
Ŝ(α)|ψm i = e− ~ L̂z α |ψm i = |ψm ie−i mα
Eq. 11.21
hr, θ, φ|Ŝ(α)|ψm i = hr, θ, φ − α|ψm i = hr, θ, φ|ψm ie−i mα
setting first φ = 0 and then α = −φ we obtain
ψm (r, θ, φ) = ψm (r, θ, 0)ei mφ
Thus we obtained the φ-dependence of the angular momentum. This works fine for integer angular
momentum. For a half-integer m, the wave function changes its sign for a full turn, because ei π = −1.
Hence they cannot be associated with a single valued wave function if m is not an integer. Half-integer
angular momenta are associated to spin, which we will discuss later.
The θ-dependence
p can be obtained from L̂− Yℓ,−ℓ = 0 and
L̂+ Yℓ,m = Yℓ,m+1 ~ ℓ(ℓ + 1) − m(m − 1). Here we only summarize the main results.
The full eigenstates are[12]
s
2ℓ + 1 (ℓ − m)! m
Yℓ,m (θ, φ) = P (cos θ)ei mφ (11.23)
4π (ℓ + m)! ℓ
The Pℓm are called associated Legendre polynomials defined as

p |m| d |m|
Pℓ|m| (u) = (−1)m 1 − u2 Pℓ (u). (11.24)
du |m|
(ℓ − |m|)! |m|
Pℓ−|m| (u) = (−1)m P (u) (11.25)
(ℓ + |m|)! ℓ
with the help of the Legendre polynomials
1 ℓ 2
Pℓ (u) = ∂ (u − 1)ℓ (11.26)
2ℓ ℓ! u
Numerically, associated Legendre polynomials are be evaluated by efficient recursion relations. Most
recursion relations amplify numerical errors, rendering them useless. A recursion relation for asso-
ciated Legendre polynomials that works is found in the “Numerical Recipes”[12]. It is in general
advisable to consult a book on numerical mathematics such as the “Numerical Recipes”, before eval-
uating special functions on the computer, as they often pose substantial difficulties that are easily
overlooked.
Yℓ,m (~r) Yℓ,m (θ, φ)

q q
1 1
Y0,0 4π 4π
q q
3 x−i y 3
Y1,−1 8π r sin θe−i φ
q q8π
3 z 3
Y1,0 4π r cos(θ)
q q4π
3 x+i y 3
Y1,1 − 8π r − 8π sin θei φ
q q
15 (x 2 −y 2 )−i xy
Y2,−2 32π r2
15
32π sin2 θe−2i φ
q q
15 xz −i y z 15
Y2,−1 r2 sin θ cos θe−i φ
q 8π q 8π
5 3z 2 −r 2 5
Y2,0 16π r2 16π (3 cos2 θ − 1)
q q
15 xz +i y z 15
Y2,1 − 8π r2 − sin θ cos θei φ
q q8π
15 (x 2 −y 2 )+i xy
Y2,2 32π r2
15
32π sin2 θe2i φ
Table 11.1: Spherical harmonics Yℓ,m in Cartesian and polar coordinates. Polar coordinates are defined
as x = r sin θ cos φ, y = r sin θ sin φ and z = r cos θ. Source: Jackson: Electrodynamics and Atkins
Molecular Quantum Mechanics.
Often the real spherical harmonics are used, which are fully real, but are not eigenstates of L̂z .
We denote them with the symbol Ȳ . For m = 0 real and complex spherical harmonics are identical.
For m 6= 0 we use the following definition.
1 √
Ȳℓ,|m| = √ Yℓ,|m| + (−1)m Yℓ,−|m| = 2Re[Yℓ,m ]
2
i √
Ȳℓ,−|m| = − √ Yℓ,|m| − (−1)m Yℓ,−|m| = 2Im[Yℓ,m ]
2
~ˆ2 , but they are no eigenstates of L̂z

Note that the real spherical harmonics are still eigenstates to L
The following is a list of real spherical harmonics from which the complex spherical harmonics are
readily obtained. Here the real spherical harmonics are identified by a running index n = ℓ2 + ℓ + m
q
1
s Ȳ1 Y0 4π
q
√1 (Y1,1 3 x
px Ȳ2 − Y1,−1 ) r
2 q 4π
3 z
pz Ȳ3 Y1,0
q 4π r
−i 3 y
py Ȳ4 √ (Y
2 1,1
+ Y1,−1 ) 4π r
q
√1 (Y2,2 15 x 2 −y 2
dx 2 −y 2 Ȳ5 2
+ Y2,−2 ) 16π r2
q
√1 (Y2,1 60 xz
dxz Ȳ6 2
− Y2,−1 ) 16π r 2
q
5 3z 2 −r 2
d3z 2 −r 2 Ȳ7 Y2,0 16π r2
q
−i 60 y z
dy z Ȳ8 √ (Y
2 2,1
+ Y2,−1 ) r2
q 16π
−i 60 xy
dxy Ȳ9 √ (Y2,2
2
− Y2,−2 ) 16π r 2
(ϕ )|
|Y l,m
ϕ
x
Fig. 11.3: Representation of real spherical harmonics (cos(ϕ)) in polar coordinates. For a given
direction defined by the angles θ, φ, a vector with a length equal to the absolute value of the amplitude
of the real spherical harmonics points from the origin to the graph of the curve. Because this graph
does not distinguish the different signs of the function, they are indicated by coloring the lobes with
positive and negative values differently. Note that spherical harmonics defined as eigenstates of the
angular momentum L ~ˆ2 and L̂z are complex, so that real and imaginary part of a spherical harmonics
must be drawn separately. Real spherical harmonics are not eigenstates of L̂z , but they are already
real.
px py pz
dxy dxz dyz
dx2 y2 dz2
Fig. 11.4: Real spherical harmonics: The graphical representation refers to a surface | |r1| Ȳℓ,m (~r)| = 1.
That is for every angle the distance of the surface from the origin is equal to the value of the spherical
harmonic. Typically the lobes are colored differently to distinguish regions with positive and negative
sign. The motivation for these plots is the representation of atomic orbitals, which are eigenstates
of angular momentum and have the form R(|r |)Yℓ,m (~r). If we plot a surface where the orbital has
one value c in one color, and another surface for a value −c with another color, we obtain similarly
looking lobes.
11.5 Spin
An important result of the algebraic derivation of the spectrum of the angular momentum is the
existence of states with half-integer m quantum number. There is no single valued wave function in
real space that is an eigenstate to a non-integer m so that L̂z |ψi = |ψi~m. Therefore, we could not
have found these eigenvalues by trying to solve the differential equation in real space. The way out
is to treat spin algebraically.
In the following we concentrate on eigenstates with ℓ = 21 , which we also call spin 12 eigenstates.
States with other spin values ℓ can be represented analogously.
def 1 1 def
Let us introduce a basis of spin eigenstates, namely | ↑i =| , i and | ↓i =| 21 , − 21 i defined
2 2
|{z} |{z}
ℓ m
by
~ ~
L̂z | ↑i = | ↑i L̂z | ↓i = | ↓i(− )
2 2
so that
! !
h↑ |L̂z | ↑i h↑ |L̂z | ↓i ~ 1 0
=
h↓ |L̂z | ↑i h↓ |L̂z | ↓i 2 0 −1
Now we use
p
L̂+ |ℓ, mi = |ℓ, m + 1i~ ℓ(ℓ + 1) − m(m + 1)
p
L̂− |ℓ, mi = |ℓ, m − 1i~ ℓ(ℓ + 1) − m(m − 1)
with ℓ = 12 and m = ± 21 to express the action of the ladder operators L̂+ and L̂− in terms of these
basis functions
L̂− | ↑i = | ↓i~; L̂− | ↓i = 0

L̂+ | ↑i = 0; L̂+ | ↓i = | ↑i~
! !
h↑ |L̂− | ↑i h↑ |L̂− | ↓i 00
=~
h↓ |L̂− | ↑i h↓ |L̂− | ↓i 10
! !
h↑ |L̂+ | ↑i h↑ |L̂+ | ↓i 01
=~
h↓ |L̂+ | ↑i h↓ |L̂+ | ↓i 00
Using L̂x = 12 (L̂+ + L̂− ) and L̂y = 1

2i (L̂+ − L̂− ) we obtain the remaining two components of the
angular momentum
! !
h↑ |L̂x | ↑i h↑ |L̂x | ↓i ~ 01
=
h↓ |L̂x | ↑i h↓ |L̂x | ↓i 2 10
! !
h↑ |L̂y | ↑i h↑ |L̂y | ↓i ~ 0 −i
=
h↓ |L̂y | ↑i h↓ |L̂y | ↓i 2 i 0
Thus we can write

! !
ψ↑ ~ ψ↑
L̂i = σi
ψ↓ 2 ψ↓
where Ψ↑ = h↑ |Ψi and Ψ↓ = h↓ |Ψi are vector components of a state |Ψi in the basis of eigenstates
to L̂z , and where the σi are the so-called Pauli Matrices
! ! !
01 0 −i 1 0
σx = ; σy = ; σz =
10 i 0 0 −1
that we have already learned about in the section on the two state system.
Most particles have an angular momentum that is not related to its “trajectory”. One interprets
it as the rotation of a particle about its axis. A spin 12 particle of this type is the electron. It is
described by a double valued spatial wave function (Ψ↑ (r ), Ψ↓ (r )), that is able to describe both its
position and angular momentum coordinates.
The spin at a certain point is obtained from a two-component wave function by
! !
~ Ψ∗↑ (r ) Ψ↑ (r )
Si (r ) = σi
2 Ψ∗↓ (r ) Ψ↓ (r )
~
Sx (r ) = Ψ∗↑ (r )Ψ↓ (r ) + Ψ∗↓ (r )Ψ↑ (r ) = ~Re[Ψ∗↑ (r )Ψ↓ (r )]
2
~ ∗
Sy (r ) = Ψ↑ (r )Ψ↓(r ) − Ψ∗↓ (r )Ψ↑ (r ) = ~Im[Ψ∗↑ (r )Ψ↓ (r )]
2i
~ ∗
Sz (r ) = Ψ↑ (r )Ψ↑ (r ) − Ψ∗↓ (r )Ψ↓ (r )
2
The particle density is obtained as the sum of the squared components.

n(r ) = Ψ∗↑ (r )Ψ↑ + Ψ∗↓ (r )Ψ↓ (r )
Often the wave functions are constrained to have exclusively either only a spin-up component or
exclusively a spin-down component. Then the states are eigenstates to Sz = ~2 σz .
11.6 Addition of angular momenta

Let us consider two electrons orbiting in an atom. Obviously this is a rotationally invariant problem,
and therefore the angular momentum must be a quantum number. However, if the particles interact,
they can exchange angular momentum between themselves, as only the total angular momentum is
conserved.
We can describe the two electrons by their individual angular momenta L ~ 1 and L
~ 2 , where L
~ 1 acts
only on the coordinates and momenta of the first electron and L ~ 2 acts only on the coordinates and
momenta of the second atom. Examples:
• For example if the two electrons are described by a two-particle wave functions Ψ(~r1 , ~r2 ), the
angular momentum acting on the first particle would be
h~r1 , ~r2 |L ~ˆ1 Ψ(~r1 , ~r2 ) = ~ ~r1 × ∇

~ˆ1 |Ψi =: L ~ ~r1 Ψ(~r1 , ~r2 )
i
~ ~ ~r2 Ψ(~r1 , ~r2 )
h~r1 , ~r2 |L̂2 |Ψi =: L̂2 Ψ(~r1 , ~r2 ) = ~r2 × ∇
i
The total angular momentum is denoted as
J~ˆ= L
def ˆ
~ˆ2
~1 + L
Its eigenvalues will be denoted by j = ℓJ and jz = mJ .

• Similarly we can consider a single particle with spin orbiting an atom. In this case, the total
angular momentum has the form
J~ˆ = L
~ˆ + S
~ˆ
where L~ˆ is the orbital angular momentum and S

~ˆ is the spin of the particle. In two-component
spinor representation the total angular momentum has the form
~
J~ˆ = ~rˆ × p~ˆ 1 + 1̂ σ~
2!
~~ 10 ~
ˆ ~r × ∇
= + ~σ
i 01 2
This is a difficult but very instructive example: We have operators, denoted by a hat, “ ˆ ”, that
acts on states that correspond square integrable complex functions in the three-dimensional
coordinate space and those that act on the two-dimensional spin states. The complication
comes from the complex nature of the Hilbert space for that system. This Hilbert space is a
product of a two-dimensional Hilbert space, describing the spin, and the Hilbert space of wave
functions in three-dimensional coordinate space. A basis state in that space is
|~r, σi
where σ can have the values ↑ or ↓. The two dimensional matrices act on the second variable,
whereas the differential operators act on the first variable.
When the electrons interact, each electron does not have a conserved angular momentum. The
total angular momentum J~ = L ~1 + L~ 2 is still conserved. Hence we are looking for eigenstates of J~ˆz
and J~ˆ2 . Note that the eigenvalues for J~ˆ2 and Jˆz are denoted by ~j(j + 1) and ~jz respectively.
The question we address in the following is, how we transform between these two sets of eigen-
states.
We start out with a basis of angular momentum eigenstates for each electron individually
~ˆ21 |ℓ1 , m1 , ℓ2 , m2 i = |ℓ1 , m1 , ℓ2 , m2 i~2 ℓ1 (ℓ1 + 1)

L
~ˆ2 |ℓ , m , ℓ , m i = |ℓ , m , ℓ , m i~2 ℓ (ℓ + 1)
L 2 1 1 2 2 1 1 2 2 2 2
L̂1,z |ℓ1 , m1 , ℓ2 , m2 i = |ℓ1 , m1 , ℓ2 , m2 i~m1
L̂2,z |ℓ1 , m1 , ℓ2 , m2 i = |ℓ1 , m1 , ℓ2 , m2 i~m2
p
L̂1,− |ℓ1 , m1 , ℓ2 , m2 i = |ℓ1 , m1 − 1, ℓ2 , m2 i~ (ℓ1 + m1 )(ℓ1 − m1 + 1)
p
L̂2,− |ℓ1 , m1 , ℓ2 , m2 i = |ℓ1 , m1 , ℓ2 , m2 − 1i~ (ℓ2 + m2 )(ℓ2 − m2 + 1)
p
L̂1,+ |ℓ1 , m1 , ℓ2 , m2 i = |ℓ1 , m1 + 1, ℓ2 , m2 i~ (ℓ1 − m1 )(ℓ1 − m1 + 1)
p
L̂2,+ |ℓ1 , m1 , ℓ2 , m2 i = |ℓ1 , m1 , ℓ2 , m2 + 1i~ (ℓ2 − m2 )(ℓ2 − m2 + 1)
Let us express J~ˆ2 by the angular momenta of individual particles L

~ˆ1 and L
~ˆ2
J~ˆ2 = L
~ˆ21 + 2L~ˆ1 L ~ˆ2
~ˆ2 + L
h 2 i
=L~ˆ2 + L
1
~ˆ2 + 2 L̂ L̂
2 1,x 2,x+ L̂1,y L̂2,y + L̂1,z L̂2,z
h i
= L21 + L22 + 2L1,z L2,z + L1,+ L2,− + L1,− L2,+
Apparently the states are not eigenstates of the total angular momentum. To show this let us
apply Jz and J 2 to the eigenstates of L1 and L2 .
Jz |ℓ1 , m1 , ℓ2 , m2 i = ~(m1 + m2 )|ℓ1 , m1 , ℓ2 , m2 i

J 2 |ℓ1 , m1 , ℓ2 , m2 i = ~2 [ℓ1 (ℓ1 + 1) + ℓ2 (ℓ2 + 1) + 2m1 m2 ]|ℓ1 , m1 , ℓ2 , m2 i
+ |ℓ1 , m1 + 1, ℓ2 , m2 − 1i
p
× ~2 [ℓ1 (ℓ1 + 1) − m1 (m1 + 1)][ℓ2 (ℓ2 + 1) − m2 (m2 − 1)]
+ |ℓ1 , m1 − 1, ℓ2 , m2 + 1i
p
× ~2 [ℓ1 (ℓ1 + 1) − m1 (m1 − 1)][ℓ2 (ℓ2 + 1) − m2 (m2 + 1)]
In order to find an eigenstate of J 2 we need to superimpose the states with a given ℓ1 , ℓ2 and
m = m1 + m2 .
The new basis will have the eigenvalues
J 2 |ℓ1 , ℓ2 , ℓ, mi = ~2 ℓ(ℓ + 1)|ℓ1 , ℓ2 , ℓ, mi

Jz |ℓ1 , ℓ2 , ℓ, mi = ~m|ℓ1 , ℓ2 , ℓ, mi
L21 |ℓ1 , ℓ2 , ℓ, mi = ~2 ℓ1 (ℓ1 + 1)|ℓ1 , ℓ2 , ℓ, mi
L21 |ℓ1 , ℓ2 , ℓ, mi = ~2 ℓ1 (ℓ1 + 1)|ℓ1 , ℓ2 , ℓ, mi
which is a definition of the basis functions |ℓ1 , ℓ2 , ℓ, mi.

The general form of the basis functions is
min(ℓ1 ,ℓ2 )−m/2
X m m
|ℓ1 , ℓ2 , ℓ, mi = |ℓ1 , + i , ℓ2 , − i ici
2 2
i =− min(ℓ1 ,ℓ2 )−m/2
where the coefficients ci also depend on ℓ1 , ℓ2 , ℓ, m. The coefficients ci = hℓ1 , m1 , ℓ2 , m2 |ℓ1 , ℓ2 , ℓ, mi

with i = m1 −m
2
2
, which also depend on ℓ1 , ℓ2 , m, ℓ are called Clebsch-Gordan coefficients, Wigner
or vector coupling coefficients.
They are fairly complex so that we will not evaluate them here.
Editorial remark: Work on this section. Also include pictorial description of vector addition
components as in Atkins
11.7 Products of spherical harmonics

Extremely useful is the product rule of spherical harmonics.
GAUNT COEFFICIENTS
The Gaunt coefficients[26] are defined as

Z
ℓ′′ ,m′′ def
Gℓ,m,ℓ′ ,m′ = dΩ Yℓ∗′′ ,m′′ (~r)Yℓ,m (~r)Yℓ′ ,m′ (~r) (11.27)
The Gaunt coefficients can be used to express the product of two spherical harmonics by a sum of
single spherical harmonics, called the Gaunt series.
X ′′ ′′
ℓ ,m
Yℓ,m (~r)Yℓ′ ,m′ (~r) = Gℓ,m,ℓ r)
′ ,m′ Yℓ′′ ,m′′ (~
ℓ′′ ,m′′
Note that the definition

√ of Gaunt coefficients is not uniform, and I have seen definitions with an
additional factor of 4π (Cohen Tannoudji). In order to confirm numerically of the definition agrees
0,0
with the one given above, simply test the coefficients Gℓ,m,ℓ,m = √14π . This rule can be derived, from
1
knowing that Yℓ,m = √4π .
The product rule is for example needed to express the particle density of an atom in spherical
harmonics, if the wave function are given as radial function times spherical harmonics. If Ψn (~r) =
Rn (|~r|)Yℓ,m (~r), we obtain
X ′′ ′′
ρ(~r) = qΨ∗ (~r)Ψ(~r) = ℓ ,m
Gℓ,−m,ℓ,m |R(|~r|)|2 Yℓ′′ ,m′′ (~r)
ℓ′′ ,m′′
Here, we used the property that Yℓ,m ∗

(~r) = Yℓ,−m (~r), which is obtained from the definition Eq. 11.23
with the help of Eq. 11.25.
The Gaunt coefficients are closely related to the Clebsch-Gordan coefficients:
s
ℓ,m m2 (2ℓ1 + 1)(2ℓ2 + 1)
Gℓ1 ,m1 ,ℓ2 ,m2 = (−1) hℓ1 , 0, ℓ2 , 0|ℓ1 , ℓ2 , ℓ, 0ihℓ1 , m1 , ℓ2 , −m2 |ℓ1 , ℓ2 , ℓ, mi
4π(2ℓ + 1)
In the following we sketch how these coefficients are derived: We start out with a two particle
wave function, which depends on the positions ~r1 and ~r2 of the two particles.
hr1 , r2 |ℓ1 , m1 , ℓ2 , m2 i = Yℓ1 ,m1 (~r1 )Yℓ2 ,m2 (~r2 )
hr1 , r2 |ℓ1 , ℓ2 , ℓ, mi = Φ(~r1 , ~r2 )
We can relate Φ to the products of two wave functions using the Clebsch Gordan coefficients.
X
h~r1 , ~r2 |ℓ1 , m1 , ℓ2 , m2 i = h~r1 , ~r2 |ℓ1 , ℓ2 , ℓ, mihℓ1 , ℓ2 , ℓ, m|ℓ1 , m1 , ℓ2 , m2 i
ℓ,m
X
Yℓ1 ,m1 (~r1 )Yℓ2 ,m2 (~r2 ) = Φℓ1 ,ℓ2 ,ℓ,m (~r1 , ~r2 )hℓ1 , ℓ2 , ℓ, m|ℓ1 , m1 , ℓ2 , m2 i
ℓ,m
and therefore also

X
Yℓ1 ,m1 (~r)Yℓ2 ,m2 (~r) = Φℓ1 ,ℓ2 ,ℓ,m (~r, ~r)hℓ1 , ℓ2 , ℓ, m|ℓ1 , m1 , ℓ2 , m2 i
ℓ,m
It can be shown that Φ(~r, ~r) is proportional to a spherical harmonic itself. Once that is shown,
we only need to work out the normalization constants.
Let us show that Fℓ1 ,ℓ2 ,ℓ,m (~r) = Φℓ1 ,ℓ2 ,ℓ,m (~r, ~r) is a spherical harmonic. Therefore, we apply an
angular momentum operator to F : Using product and chain rule we relate the derivatives of F . We
can consider ~r1 and ~r2 to be functions of r namely ~r1 (~r) = ~r and ~r2 (~r) = ~r.
~r1 (~r) = ~r
~r2 (~r) = ~r
~r × ∇F (~r) = ~r × ∇Φ(~r1 (~r), ~r2 (~r))
= [~r × ∇1 + ~r × ∇2 ]Φ(~r1 (~r), ~r2 (~r))
= [~r1 × ∇1 + ~r2 × ∇2 ]Φ(~r1 (~r), ~r2 (~r))
LF (~r) = [L~1 + L
~ 2 ]Φ(~r, ~r)
where L~ 1 acts only on the first argument and L ~ 2 on the second.

Because Φ(~ r1 , ~r2 ) is an eigenstate of angular momentum (L ~1 + L~ 2 )Φ(~ r1 , ~r2 )~2 ℓ(ℓ + 1)
r1 , ~r2 ) = Φ(~
~ 1,z + L
and (L ~ 2,z )Φ(~ r1 , ~r2 )~m we find that also F (r ) is an eigenstate of angular momentum
r1 , ~r2 ) = Φ(~
Lz and L2 ,
Lz |F i = ~m|F i
L |F i = ~2 ℓ(ℓ + 1)|F i
2
which implies that they are related to spherical harmonics.

The derivation of the remaining coefficients can be found in Cohen-Tannoudji[6].
11.8 Exercises
Chapter 12
Atoms
12.1 Radial Schrödinger equation
Let us consider a particle in a spherically symmetric potential
h p~ˆ2 i
+ V (|~rˆ|) |ψi = |ψiE
2m
Editor: Include hats to distinguish numbers from operators

We introduce the radial momentum p̂r = |r1| (~rˆ~rˆ) and obtain an expression that contains radial
and angular momenta separately.
h 1 1 1 i
p̂r2 rˆ + 2 L̂2 + V (ˆ
r ) |ψi = |ψiE
2m rˆ rˆ
where r = |~r|. Note, that[16]
1 2 1
pr r = 2 pr r 2 pr
r r
191
192 12 ATOMS
~2 can be written as
Here we show that p
1 2 ~2
L
p~2 = pr r + 2
r r
PROOF: We start with squaring one component of the angular momentum
L2x = (y pz − z py )2 = y pz y pz − y pz z py − z py y pz + z py z py
= y 2 pz2 − y [pz , z ]py − y z pz py − z [py , y ]pz − z y py pz + z 2 py2
~
= (y 2 pz2 + z 2 py2 ) − 2(y py )(z pz ) − (y py + z pz )
i
with cyclic permutation we find
~ 2 = (y 2 pz2 + z 2 py2 ) − 2(y py )(z pz ) − ~ (y py + z pz )

L
i
2 2 2 2 ~
+ (z px + x pz ) − 2(z pz )(xpx ) − (z pz + xpx )
i
2 2 2 2 ~
+ (x py + y px ) − 2(xpx )(y py ) − (xpx + y py )
i
= x 2 (py2 + pz2 ) + y 2 (px2 + pz2 ) + z 2 (px2 + py2 )

− xpx (y py + z pz ) + y py (xpx + z pz ) + z pz (xpx + y py )
2~
− xpx + y py + z pz )
i
Now we add the identity x 2 px2 − (xpx )2 + ~i xpx = 0, and those obtained by exchanging x by y and z to the
expression above
~ 2 = ~r2 p ~
L ~2 − (~rp~)2 − ~rp ~
i
~
= ~r2 p
~2 − (r pr ) r pr −
i
= r 2p~2 − (r pr )(pr r )
1 1
p~2 = pr2 r + 2 L2
r r
where we used ~rp~ = r pr and [pr , r ] = ~i . q.e.d.
Now we use as Ansatz the eigenstates of the angular momentum
ψ(~r) = Rℓ (|~r|)Yℓ,m (~r)
By insertion we find1
h 1 1 1 i
0= pr2 r + 2 L2 + V (r ) − E Rℓ (r )Yℓ,m (~r)
2m r r
h 1 1 ~2 ℓ(ℓ + 1) i
= Yℓ,m (~r) pr2 r + + V (r ) − E Rℓ (r )
2m r r2
We can divide by the spherical harmonics and obtain an one-dimensional differential equation for the
radial component of the wave function.
We can rewrite2 the radial Schrödinger equation in the form
h −~2 ~2 ℓ(ℓ + 1) i
∂r2 + + V (r ) − E r R(r ) = 0
2m 2mr 2
1 We used here that the radial momentum commutates with the spherical harmonics, which is obvious, as the
spherical harmonics do not depend on the radial coordinate.
2 Multiply from the left with r .
12 ATOMS 193
Thus the radial Schrödinger equation has the form of a regular one-dimensional Schrödinger equation
with an angular dependent potential, that also includes a potential which represents the centrifugal
2
force. It is as if a one-dimensional particle would move in a potential Vℓ (r ) = ~ ℓ(ℓ+1)
2m
1
r 2 + V (r ).
Ze 2 1
For an electron in that experiences the electrostatic attraction to the nucleus V (r ) = − 4πǫ 0 r
. The
ℓ-dependent potentials for the hydrogen atom are shown in Fig. 12.1.
0.1
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
0 2 4 6 8 10
Fig. 12.1: Radial potential of the hydrogen atom (Z=1) including the centrifugal part for ℓ =
2
2 2
0 ~ ℓ(ℓ+1) e m Z2
0, 1, 2, 3, 4. The minima are at rmi n = 4πǫ
me 2 Z and the well depth is V mi n = − 1
2 4πǫ0 ~2 2ℓ(ℓ+1) .
Formal relation of the radial Schrödinger equation to Newton’s equation of motion
The radial Schrödinger equation for r R(r ) is an ordinary differential equation, and can also be solved
with the methods developed for the latter. It may be instructive to show the relation to Newton’s
equations of motion. Because we may have a good imagination on the behavior of classical particles
compared to that of quantum mechanical wave functions.
Consider the Newtons equation of motion for a harmonic oscillator with a time-dependent force
constant.
m∂t2 x(t) = −c(t)x(t)
We do now the following transformations in order to convert Newtons equation of motion into the
radial Schrödinger equation.
• The time is converted into the radial distance, i.e. t → r
• the position x(t) is converted into r R(r )
~2
• the mass of the harmonic oscillator is replaced by 2m

~2 ℓ(ℓ+1)
• the force constant c(t) is converted into the radial kinetic energy E − V (r ) + 2mr 2

−m∂t2 − c(t) x(t) = 0

~2 2
→ − ∂ + V (r ) − E r R(r ) = 0
2m r
194 12 ATOMS
Boundary conditions at the origin

The main difference of the radial Schrödinger equation to the one-dimensional Schrödinger equation
are the boundary conditions at the origin r = 0. We proceed here with an expansion of R in powers
of r , that is
X
R(r ) = aj r j
j

We maintain only the leading order in r of the potential, that is V (r ) = Cr α 1 + O(r ) .
h −~2 ~2 ℓ(ℓ + 1) iX
α
0= ∂r2 + + Cr − E aj r j +1
2m 2mr 2
j
Xh −~2 2
~ ℓ(ℓ + 1) j −1 i
= j(j + 1)aj r j −1 + aj r + Caj r α+j +1 − Eaj r j +1
2m 2m
j
X h −~2 i
= r j −1 j(j + 1) − ℓ(ℓ + 1) aj + Caj −α−2 − Eaj −2
2m
j
which results in a recursion relation for the coefficients

−~2
0= j(j + 1) − ℓ(ℓ + 1) aj + Caj −α−2 − Eaj −2
2m
2mC 2mE
= (j − ℓ)(j + ℓ + 1)aj − 2 aj −α−2 + 2 aj −2
~ ~
1 2mC 2mE
aj = a j −α−2 − a j −2 (12.1)
(j − ℓ)(j + ℓ + 1) ~2 ~2
Let us consider the case where
• α = −1. This case is important for atoms, because the electrostatic potential of an electron
Ze 2
that experiences the Coulomb attraction of the nucleus is of this form V (r ) = − 4πǫ 0r
. We
need to consider a solution for which aj <j0 = 0. In Eq. 12.1, aj0 can only be nonzero, if the
denominator vanishes, which determines the possible values of j0 .
(j0 − ℓ)(j0 + ℓ + 1)aj0 = 0 ⇒ j0 = ℓ or j0 = −ℓ − 1
The solution j0 = −ℓ − 1 can be discarded because it will be irregular at the origin and not
normalizable. The coefficient aj0 is arbitrary and will be determined by the normalization con-
dition.
In order to obtain the next term in the expansion, we insert this result into Eq. 12.1 for
j = j0 + 1 = ℓ + 1
mC 1
aℓ+1 = aℓ
~2 ℓ + 1
Thus the radial wave function behaves near the origin as
h mC i
Rr (r ) = A 1 + 2 r + O(r 2 ) r ℓ
~ (ℓ + 1)
and is not differentiable for ℓ = 0, which is a consequence of the Coulomb singularity. This
kink is not energy dependent, while the higher order terms would depend explicitly on energy.
For the “hydrogen atom” with atomic number Z we obtain C = −Ze 2 /(4πǫ0 ) and a recursion
relation with k = j − ℓ
1 2mZe 2 2mE
aj = − 2
aj −1 + 2 aj −2 = 0
(j − ℓ)(j + ℓ + 1) 4πǫ0 ~ ~
ak<0 = 0
12 ATOMS 195
so that
h mZe 2 i
Rr (r ) = a0 1 − 2
r + O(r 2 ) r ℓ
4πǫ0 ~ (ℓ + 1)
Note, that the kink is independent of the energy.
• α = 0 which describes a continuous potential. The term proportional r ℓ+1 vanishes, and the
second radial function behaves like
h i
Rr (r ) = A 1 + O(r 2 ) r ℓ
The recursion relation Eq. 12.1 for the coefficients could be used to determine a Taylor expansion
of R(r ) to arbitrary high order. More accurate results are obtained, when one uses the Taylor
expansion to obtain value and derivative at a point close to the nucleus and to use these values as
initial conditions for a numerical solution of the Schrödinger equation.
12.2 The Hydrogen Atom

The potential for the hydrogen atom is
Ze 2
V (~r) = −
4πǫ0 |r |
where Z is the atomic number, e is the elementary charge, and ~r is the position relative to the nuclear
site. The radial potentials including the centrifugal term are shown in Fig. 12.1.
h −~2 ~2 ℓ(ℓ + 1) Ze 2 i
∂r2 + − − E r R(r ) = 0
2me 2me r 2 4πǫ0 r
Hartree atomic units

Let us simplify the equation by introducing dimensionless quantities. We multiply with me /~2 and
introduce a length unit and an energy unit so that the equation is dimensionally consistent.
h ℓ(ℓ + 1) 2me Ze 2 1 2me E i
−∂r2 + + − r R(r ) = 0
r2 4πǫ0 ~2 r ~2
h 1 ℓ(ℓ + 1) Z E i
− ∂r2 + + − r R(r ) = 0
2 2r 2 a0 r E0 a02
h 1 ℓ(ℓ + 1) Z Ei
− ∂y2 + + − y R(y a0 ) = 0 (12.2)
2 2y 2 y E0
Thus we have defined a length scale, the Bohr radius a0 , and an energy scale, the Hartree E0 = H.
Definition 12.1 BOHR RADIUS
4πǫ0 ~2
a0 = (12.3)
me e 2
The Bohr radius is the length unit of the Hartree atomic unit system. It corresponds to the radius of
a classical electron circling a proton at the ground state energy of the quantum mechanical hydrogen
atom.
196 12 ATOMS
Definition 12.2 HARTREE
~2 me e 4
H= = (12.4)
me a02 (4πǫ0 ~)2
The Hartree is the energy unit of the Hartree atomic unit system. One Hartree is twice the binding
energy of an electron in the hydrogen atom.
In this way we can define a new system of units, the so-called Hartree atomic units.
Definition 12.3 HARTREE ATOMIC UNITS
The Hartree atomic units are defined by
~ = me = e = 4πǫ0 = 1
where me is the electron mass, e is the elementary charge, and ǫ0 is the dielectric constant of the
vacuum.
The Hartree atomic unit system is commonly used for atomic and molecular physics. Unlike the official
SI unit system, Hartree atomic units are defined entirely by measurable fundamental quantities.
• the length unit of the Hartree atomic unit system is the Bohr radius, which corresponds to the
radius of the classical orbit of an electron in the hydrogen atom.
• the energy unit is the Hartree, which is twice the binding energy of a hydrogen atom.
Instead of using the symbols a0 , H, etc. one often uses a.u. which stands for “atomic unit” in general.
Note that there are two atomic unit systems. The so-called Rydberg atomic unit system is different.
This leads to confusion, because often it is not spelled out, which of the two atomic unit systems is
meant.
In Hartree atomic units, the Schrödinger equation of an electron in the Coulomb potential of the
nucleus has the form
1 Z
− ∇2 − − E |Ψi = 0
2 r
The values of the fundamental constants and the conversion factors can be found in the book
“Quantities, Units and Symbols in Physical Chemistry” [17] or on the web[18].
Bound wave functions

Before we solve Eq. 12.2 let us investigate its qualitative behavior, which will guide us to useful ansatz
for the solution. We can simplify the differential equation, if we consider the two limits r → 0 and
r →∞
• For r → 0 we can use the Power series expansion, which tells us that the wave function at the
origin is proportional to r ℓ .
• For r → ∞ we can ignore the potential and the centrifugal term. Thus the differential equation
becomes independent of Z and ℓ. We obtain with y = r /a0
h 1 Ei
− ∂y2 − y R(y a0 ) ≈ 0
2 H
which can easily be solved. The solution has the form
1 −
q
2E y
R(y a0 ) ∼ e H
y
12 ATOMS 197
The wave function decays exponentially with the decay governed by the energy. For bound
states the wave functions decay exponentially, while they oscillate for positive energies. Here
we only consider the bound states.
These considerations suggest an Ansatz of the form
R(y a0 ) = F (y )y ℓ e−λx
where
r
def E
λ= 2 (12.5)
H
The goal is now to determine the function F (y )
We insert this Ansatz into the radial Schrödinger equation. Eq. 12.2:
h 1 ℓ(ℓ + 1) Z Ei
− ∂y2 + − − F (y )y ℓ+1 e−λy = 0 .
2 2y 2 y H
We use
ℓ+1
∂y y ℓ+1 e−λy = y ℓ+1 e−λy [ − λ + ∂y ]
y
ℓ+1 ℓ+1
⇒ ∂y2 y ℓ+1 e−λy = y ℓ+1 e−λy [ − λ + ∂y ][ − λ + ∂y ]
y y
" 2 2 #
ℓ+1 −λy ℓ+1 ℓ+1 ℓ+1 2
=y e − λ + ∂y −λ + − λ ∂y + ∂y
y y y

ℓ+1 −λy ℓ+1 2 ℓ+1 ℓ+1 2
=y e ( − λ) − 2 + 2[ − λ]∂y + ∂y
y y y

ℓ+1 −λy 2 ℓ+1 ℓ(ℓ + 1) ℓ+1 2
=y e ∂y + 2[ − λ]∂y + − 2λ +λ
y y2 y
and insert into the radial Schrödinger equation Eq. 12.2
h ℓ+1 2 2E i
∂y2 + 2[ − λ]∂y − λ(ℓ + 1) − Z + + λ2 F (y ) = 0
y y | H {z }
=0
h i
⇒ y ∂y2 + 2[ℓ + 1 − λy ]∂y − 2 λ(ℓ + 1) − Z F (y ) = 0 (12.6)
This differential equation is a special case of the general associated Laguerre differential equation.

x∂x2 + (k + 1 − x)∂x + n L(k)n (x) = 0
where k, n are integers. The solutions are the so-called associated Laguerre Polynomials
ex x −k d n −x n+k
L(k)
n (x) = e x
n! dx n
This can be used if one find the corresponding tables. A useful source is the Book of Abramowitz
and Stegun[19].
However, here we want to investigate
P how such a problem can be solved directly: We make a
power-series Ansatz for F (y ) = j aj y j in Eq. 12.6.
Xh i
j(j − 1)aj y j −1 + 2(ℓ + 1)jaj y j −1 − 2λjaj y j − 2 λ(ℓ + 1) − Z aj y j = 0
j
X Xh i
[j(j − 1) + 2(ℓ + 1)j] aj y j −1 − 2λj + λ(ℓ + 1) − Z aj y j = 0
j j
 
X Xh i 
[(j + 1)j + 2(ℓ + 1)(j + 1)] aj +1 − 2λj + λ(ℓ + 1) − Z aj y j = 0
 
j j
198 12 ATOMS
Because the identity holds for all values of y , it must also hold in every power of y .3 If we collect
all terms that are proportional to a certain power of y , namely y j −1 , we obtain

j(j + 1)aj +1 + 2(ℓ + 1)(j + 1)aj +1 − 2λjaj − 2 λ(ℓ + 1) − Z aj = 0

⇒ j(j + 1) + 2(ℓ + 1)(j + 1) aj +1 − 2 λj + λ(ℓ + 1) − Z aj = 0,
which provides us with a recursion relation for the coefficients aj
λ(j + ℓ + 1) − Z
aj +1 = 2 aj (12.7)
(j + 1)(j + 2ℓ + 2)
The power series can start following that index, where the denominator vanishes that is with j = 0
or j = −(2ℓ + 1). The second choice leads to the irregular solution. The irregular solution has a
singularity at the origin. The resulting wave function would obey the Schrödinger equation everywhere,
except at the origin.Therefore, the irregular solution must be excluded. Thus the powerseries starts
with jmi n = 0, which leads to the regular solution.
The power series also has an upper end, where the numerator of Eq. 12.7 vanishes: The power
series expansion finishes, if there is an integer jmax so that
Z
0 = λ(jmax + ℓ + 1) − Z ⇒ jmax = −ℓ − 1 +
λ
jmax must itself be a positive integer, because otherwise the power series does not terminate.
(Some further calculations are required to show that such solutions lead to exponentially increasing
functions that cannot be normalized.4 )
The requirement that jmax is an integer, determines the allowed values for λ. For jmax we obtain
Z
λjmax = (12.8)
jmax + ℓ + 1
We introduce a new variable the so-called principal quantum number
n = jmax + ℓ + 1 (12.9)
Now we take the definition of λ, Eq. 12.5, resolve it for the energy,
1 2
E= λ H
2
and insert Eq. 12.8 with jmax expressed by the main quantum number as defined in Eq. 12.9. Re-
member that H is the Hartree energy unit defined in Eq. 12.4
Thus, we obtain an expression for the allowed energy eigenvalues and thus the eigenvalue spectrum
of the bound states of the hydrogen atom.
3 These becomes obvious by forming higher derivatives of the expression, which must all vanish, because the expres-
sion is zero for all values of y .

4 Power series expansions that do not truncate will fulfill an approximate recursion relation
λ−Z
aj+1 ≈ 2 aj
(j + 1)
1
aj ≈ 2j (λ − Z)j
j!
F (x → ∞) → e2(λ−Z)x
Thus F (x) will diverge exponentially, which violates the boundary conditions at infinity.Editor: For λ < Z the function
F (x) does not converge! Check if those values are allowed and what is their meaning!
12 ATOMS 199
EIGENVALUE SPECTRUM OF THE HYDROGEN ATOM
Z2 Z 2 me e 4
En,ℓ,m = − H = −
2n2 2n2 (4πǫ0 ~)2
for n = jmax + ℓ + 1 = 1, 2, 3, . . ..
States for different ℓ and m quantum numbers are degenerate. n is the principal quantum number,
ℓ the angular momentum quantum number, and m is called the magnetic quantum number.
Only states with ℓ ≥ n − 1 contribute to one multiplet of degenerate states.
n=4; 0.03.. H
n=3; 0.06.. H
n=2; 0.125 H
n=1; 0.5 H
Fig. 12.2: Electron levels of the hydrogen atom.
We can now write down the recursion relation for the coefficients depending on n as
r
−2En Z
λn = =
H n
2Z j +ℓ+1−n
aj +1 = aj
n (j + 1)(j + 2ℓ + 2)
X
Zr j +ℓ −Zr /(na0 )
Rn,ℓ (r ) = bj e
na0
j
j +ℓ+1−n
bj +1 = 2 bj
(j + 1)(j + 2ℓ + 2)
Note that these wave functions are not yet normalized. The polynomials appearing in this equation
are, up to a normalization constant, the associated Laguerre polynomials.
Zr ℓ − naZr
R(r ) = L2ℓ+1
n+ℓ (2 )r e 0
na0
200 12 ATOMS
n l E[H] R(r)
2 3
1 0 − Z2 2(a0 )− 2 e−r

/a0

2 3
2 0 − Z8 2(2a0 )− 2 1 − r
2a0 e
−r /(2a0 )
2 3
2 1 − Z8 2(2a0 )− 2 ar0 e −r /(2a0 )
1.5
−3 1
R[a0 2 ]
0.5
0
0 2 4 6 8 10
r [a0 ]
0.6
0.5
0.4
R 2 −3
dA |Ψ(r )| [a0.3
0 ]
0.2
0.1
0
0 2 4 6 8 10
r [a0 ]
Fig. 12.3: Radial parts of the 1s,2s and 2p wave function of the hydrogen atom. Formulas taken
from [7]. These functions are normalized. The results need to be checked.Editor: check formulas
and refine figures
Two observations
• The number of node planes of the wave function increases with n. Angular node planes must
be counted as well.
• wave functions and energies for general atomic number can be obtained from the hydrogen
atom by a simple scaling law.
Periodic Table
There are states with angular momentum greater than zero are degenerate with those with ℓ = 0.
This explains already a large part of the structure of the periodic table.
12 ATOMS 201
H He
Li Be B C N O F Ne
Na Mg Al Si P S Cl Ar
K Ca Sc Ti V Cr MnFe Co Ni Cu Zn Ga Ge As Se Br Kr
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
Fr Ra Ac
• the lowest state in singly degenerate (or doubly degenerate if we include spin degeneracy).
Thus we can fill this state with one or two electrons resulting in H and He.
• The next shell contains one state with ℓ = 0 and three states with ℓ = 1. The ones with only
s-electrons (ℓ = 0) are Li and Be. Alkali and Earth alkali metals. Then come the ones with
partially filled p-shell (ℓ = 1), B,C,N,O,F. The last atom in this row is Ne a noble gas that
does not form any chemical bonds, because it fulfills the octet rule. The octet rule is fulfilled
if the (s and p like valence wave functions) are occupied with just eight electrons.
• Electronegativity: The electrons in the first groups (columns in the periodic table) are loosely
bound, while those in the last group have the most strongly bound electrons, because the
electron energies decrease with increasing atomic number. Atoms with low lying electron levels
are called electronegative. Noble gas atoms, in the rightmost column of the periodic table, are
inert, because on the one hand they hold on to their electrons most strongly. On the other
hand they cannot accept any more electrons, because those would need to occupy the next
shell having much higher energies. If we combine atoms from an early group and one of a late
group, we form ionic materials and electrons are transferred so that the atom from an early
group is positive (cation) and the one from a late group is negative (anion) as in NaCl. Anions
can accept electrons until their shell is filled, while cations donate electrons until their shell is
empty. Thus we can estimate the composition of most oxides: Na2 O, MgO, Al2 O3 , SiO2 etc.
Because electrons in the early groups are loosely bound, they often form metals such as Na, Mg,
where the electrons are free to wander around. In contrast, towards the right of the periodic
table they form covalently bound materials such as F2 , O2 , N2 .
• Due to Coulomb repulsion between electrons the ℓ degeneracy is lifted in real atoms, so that
the orbitals with lower angular momentum are lower in energy than those with higher angular
momentum. The reason is that the valence electrons experience a lower effective atomic
number than the core electrons. The electrostatic potential of the nucleus is shielded by the
core electrons. This makes the core states more compact. The valence states with low angular
momentum can therefore come closer to the nucleus as well, because they are kept at larger
radius mostly by the Pauli repulsion from the core electrons. As a consequence the s-electrons
are occupied before p-electrons are occupied.
This effect also explains an anomaly in the periodic table, namely that the third period contains
only s and p electrons, while the 3d (n = 3, ℓ = 2) electrons are occupied only after the 4s
(n = 4, ℓ = 0) electrons have been filled.
202 12 ATOMS
Chapter 13
Approximation techniques
In most cases quantum mechanical problems cannot be solved exactly. Therefore, for real-world
problems, we usually need to resort to approximations or numerical methods, or a combination of
both. There is no limit to the number of approximation techniques one can use for certain problems.
However, there is a set of techniques, that already get us a long way, and the ideas behind these
approaches can be used to use new approximations for a particular problem.
13.1 Perturbation theory

Perturbation theory can be used to determine the properties of a complex system if a similar system
that is much simpler can explicitly be solved. For example the complex system may have an approxi-
mate symmetry. We can solve the system that has the symmetry exactly fulfilled, which is simpler,
and then work out the differences of the real system to our model system.
Perturbation theory is not only an approximation method but in many cases it provides information
more directly than a brute force approach. This is the case when we are interested in the linear
response of a quantity to a perturbation. For example we would like the splitting of energy levels of
an atom in an external field, because the splitting may experimentally be accessible. There is not
point to evaluate all the energy values, when it is possible to calculate the splitting directly.
The first order perturbation is also what is needed in linear response theory. For example we
would like to evaluate the dielectric constant of a gas. The dielectric constant is related to the linear
response1 of the polarization P~ to an electric field. This response is called the electric susceptibility.
Let us consider a system that can be described by an unperturbed Hamiltonian Ĥ0 and a small
perturbation Ŵ .2 We assume that we can solve the Schrödinger equation for the unperturbed
system, and we want to use that knowledge to obtain an estimate for the perturbed system. The
full Hamiltonian Ĥ is given as
Ĥ = Ĥ0 + Ŵ
If the perturbation is small, we need not solve the entire problem, but we can instead use pertur-
bation theory. The basic idea of the procedure is as follows: We introduce a scale parameter λ for
the perturbation so that
Ĥ(λ) = Ĥ0 + λŴ

dP
1 The electric susceptibility is defined by χi ,j = ǫ1 dEi , where P~ is the Polarization or electric dipole density and E
~
0 j
is the electric field. The relative dielectric constant is related to the electric susceptibility by ǫr = 1 + χ. See “ΦSX:
Elektrodynamik”.
2 We use the symbol Ŵ standing for the German word “Wechselwirkung” for Interaction, because the interaction
between particles is often treated by perturbation theory.
203
204 13 APPROXIMATION TECHNIQUES
Thus for λ = 0 we obtain the unperturbed system and for λ = 1 the fully perturbed system. For this
scaled system, the energy eigenvalues and the wave functions depend on λ.
13.2 General principle of perturbation theory

Let us now consider that we would like to evaluate a quantity F , which can be the energy, the wave
function or the density. Since the Hamiltonian depends on the parameter λ, also the quantity F will
depend on λ. Thus one can write down a Taylor expansion for F (λ) as
∞
X 1 d n F
F (λ) = n
λn
n=0
n! dλ λ=0
By inserting λ = 1 we obtain the result for the perturbed system.

∞
X 1 d n F
F (1) =
n=0
n! dλn λ=0
The first few derivatives are easy to evaluate, but the evaluation of higher derivatives is in general
cumbersome. If the perturbation is sufficiently small, higher derivatives also do not contribute much.
Therefore, the expression for the quantity F in the perturbed system is approximated by
nmax
X 1 d n F
F (1) ≈
n=0
n! dλn λ=0
This is the result for F (1) in perturbation theory of order nmax . Typically, nmax is chosen equal to
one or two, leading to first-order (F 1st ) and second-order (F 2nd ) perturbation theory.

dF
F 1st (1) = F (0) +
dλ λ=0

2nd dF 1 d 2 F
F (1) = F (0) + +
dλ λ=0 2 dλ2 λ=0
From the λ-derivative of the Schrödinger equation, we obtain expressions for the λ derivative of
the energy-eigenvalues and the wave functions. Higher derivatives can be obtained recursively. Then
we use these derivatives to set up a Taylor-series expansion of energies and wave functions in λ. This
Taylor-series expansion is truncated, usually already after the linear term, and the truncated series is
used as approximation for the values at λ = 1.
Our derivation, shown in the following, deviates from those in most text books, in that we
first determine the exact λ dependent first derivatives and then form higher derivatives from those
expressions instead of forming the first and higher derivatives for λ = 0. The reason for this choice is
two-fold. First it is a little simpler to construct the higher derivatives from the first derivative, which
we need to evaluate anyway, as compared to determining the second derivative from scratch. More
importantly, the λ dependent derivatives can be used to obtain the correct result by integration. This
procedure is at times used in numerical evaluations of the perturbed energies and eigenstates, and it
is also underlies a number of analytical derivations. The latter, however, are not worked out in this
text.
13.3 Time-independent perturbation theory

Here we show the procedure in detail: We consider now the λ-dependent, orthonormal eigenstates
|ψ(λ)i of the perturbed Hamiltonian Ĥ0 + λŴ . Since the Schrödinger equation
h i
Ĥ0 + λŴ −En (λ) |ψn (λ)i = 0
| {z }
Ĥ(λ)
13 APPROXIMATION TECHNIQUES 205
must be fulfilled for any value of λ, we can also form its derivative with respect to λ.
h i h i
Ĥ(λ) − En (λ) |ψn′ (λ)i + Ŵ − En′ (λ) |ψn (λ)i = 0
Here we use En′ = ddEλn and |ψn′ i = d d|ψλn i .

Next we multiply from the left with hψm (λ)|, which corresponds to an eigenstate of the perturbed
Hamiltonian.
h i
0 = hψm (λ)|Ĥ(λ) − En (λ)|ψn′ (λ)i + hψm (λ)| Ŵ − En′ (λ) |ψn (λ)i
= hψm (λ)|Em (λ) −En (λ)|ψn′ (λ)i + hψm (λ)|Ŵ − En′ (λ)|ψn (λ)i (13.1)
| {z }
hψm (λ)|Ĥ(λ)
For m = n the first term vanishes, so that Eq. 13.1 yields an expression for En′ :
DERIVATIVE OF THE ENERGY EIGENVALUE WITH RESPECT TO THE STRENGTH OF THE
PERTURBATION
def dEn
En′ (λ) = = hψn (λ)|Ŵ |ψn (λ)i (13.2)
dλ
Thus the changes of the eigenvalues are obtained from the expectation value of the perturbation op-
erator alone.
Let us now consider Eq. 13.1 with m 6= n.
hψm (λ)|Em (λ) − En (λ)|ψn′ (λ)i + hψm (λ)|Ŵ − En′ (λ)|ψn (λ)i = 0
hψm |Ŵ |ψn (λ)i
⇒ hψm (λ)|ψn′ (λ)i = (13.3)
En (λ) − Em (λ)
so that we can express the perturbation of the wave functions as
X
|ψn′ (λ)i = |ψm (λ)ihψm (λ)| |ψn′ (λ)i
m
| {z }
1̂
Eq. 13.3
X hψm (λ)|Ŵ |ψn (λ)i
= |ψn (λ)ihψn (λ)|ψn′ (λ)i + |ψm (λ)i
En (λ) − Em (λ)
m(m6=n)
This solution is, however, not yet unique, because it still depends on the unknown matrix element
hψn |ψn′ i. We will show in the following that (1) the real part of hψn |ψn′ i vanishes, and (2) that its
imaginary part can be chosen equal to zero.
1. We can use the condition that the wave function is normalized for every value of λ:
d
0= hψn (λ)|ψn (λ)i = hψn′ (λ)|ψn (λ)i + hψn (λ)|ψn′ (λ)i = 2Re{hψn (λ)|ψn′ (λ)i}
dλ | {z }
=1
Thus the real part of hψn′ |ψn i vanishes.

2. What about the imaginary part? It is arbitrary! This is shown as follows: Consider a λ-
dependent wave function |ψn (λ)i. For each value of λ the wave function can be multiplied
with a phase factor ei φ(λ) . The resulting wave function
|ψ̃n (λ)i = |ψn (λ)iei φ(λ)

is an eigenstate for each strength of the perturbation if the same is true for |ψn (λ)i. Now we
form the derivative
d
|ψ̃n′ (λ)i = |ψn (λ)iei φ(λ) = |ψn′ iei φ(λ) + |ψn ii φ′ (λ)ei φ(λ)
dλ
⇒ hψ̃n (λ)|ψ̃n′ (λ)i = e−i φ(λ) hψn (λ)| |ψn′ (λ)i + |ψn (λ)ii φ′ (λ) ei φ(λ)
= hψn |ψn′ i + hψn |ψn i i φ′ (λ) = hψn |ψn′ i + i φ′ (λ)
| {z }
=1
Thus, simply by changing the λ-dependent phase factor φ(λ), the imaginary part of hψ̃n |ψ̃n′ i
can be changed into an any value. All these solutions are equally valid. Therefore, we choose
the most convenient result by setting the imaginary part of hψ̃n |ψ̃n′ i to zero. In the following
we drop the tilde on the wave function.
With the argument just provided, the matrix element hψn |ψn′ i vanishes. Thus, with the appropriate
choice of the phase factor, we obtain the final result:
DERIVATIVE OF THE WAVE FUNCTION WITH RESPECT TO THE STRENGTH OF THE
PERTURBATION
def d|ψn (λ)i X hψm (λ)|Ŵ |ψn (λ)i

|ψ ′n (λ)i = = |ψm (λ)i (13.4)
dλ En (λ) − Em (λ)
m(m6=n))
Note that Eqs. 13.2 and 13.4 are exact.
13.3.1 Degenerate case

Eq. 13.4 becomes problematic if En is degenerate, because, in this case, terms with a divide-by-zero
appear in the expression for the wave function.
Without limitation of generality, we assume that the degeneracy is present at λ = 0. This is also
the most relevant case.
The remedy is the following recipe:
1. Determine all the states |Ψn (0)i within one multiplet, that is n ∈ M where M is the set of
indices of the states in the same multiplet. A multiplet is the set of degenerate states, that is
En (0) = Em (0) for all n, m ∈ M.
2. determine the matrix elements of the perturbation with the states in the multiplet
Wm,n = hψm (0)|Ŵ |ψn (0)i with n, m ∈ M
and diagonalize it so that

X
W c~j = c~j wj or Wn,m cm,j = cn,j wj
m
The eigenvalues are wj and the eigenvectors are c~j .
3. Form a new basisset so that

X
|ψ̃j (0)i = |ψm (0)icm,j
m
4. Now we set up the equation Eq. 13.1 that we obtained before,with the difference that we use
the new states with a tilde.
hψ̃m |ψ̃ ′ n i(ǫm − ǫn ) + hψ̃m |Ŵ |ψ̃n i − hψ̃m |ψ̃n iǫ′n = 0
• Case 1: ǫm = ǫn
hψ̃m |Ŵ |ψ̃n i − hψ̃m |ψ̃n iǫ′n = 0

⇒ wn δm,n − ǫ′n δm,n = 0
⇒ ǫ′n = wn
• Case2:
X X hψ̃m |Ŵ |ψ̃n i
|ψ̃ ′ n i = |ψ̃m ihψ̃m |ψ̃ ′ n i + |ψ̃m i
ǫn − ǫm
m∈M m∈M
/
Next we need to show that the matrix elements hψ̃m |ψ̃ ′ n i can be chosen equal to zero.
This is shown as follows: we begin with the λ-derivative of the normalization condition
hψ̃m |ψ̃n i = δm,n

∂λ
⇒ hψ̃ ′ m |ψ̃n i + hψ̃m |ψ̃ ′ n i = 0
⇒ hψ̃ ′ m |ψ̃n i = −hψ̃ ′ n |ψ̃m i
This shows that the matrix hψ̃m |ψ̃ ′ n i is anti-hermitian. In the following we proceed anal-
ogously to what has been done before by adding a lambda-dependent phase factor. Here,
however we need to generalize the argument to several states.
We consider a unitary transformation U (λ) = eA(λ) of the states in the multiplet,
X X
|ψ̃¯n (λ)i = |ψ̃m (λ)iUm,n (λ) = |ψ̃m (λ)i eA(λ)
m,n
m m
where A (λ) is a antihermitian matrix, that is A † = −AA.

It can be shown as follows, that U := eA is unitary, if A is antihermitian:
† †
U † U = eA eA = eA eA = e−AA eA = 1
We require that the transformation is the identity for λ = 0, that is U (λ = 0) = 1 or

A (λ = 0) = 0 .
Now we evaluate our expression

¯
|ψ̃n′ (λ)i = |ψ̃ ′ m (λ)i eA(λ) + |ψ̃m (λ)i A eA(λ)
m,n m,n
1
U (λ=0)=1 ¯
⇒ |ψ̃n′ (0)i = |ψ̃ ′ m (λ)i + |ψ̃m (λ)i A ′ m,n
¯ X
hψ̃¯m (0)|ψ̃n′ (0)i = hψ̃m (0)|ψ̃ ′ n (0)i + hψ̃ (λ)|ψ̃ (λ)i A′
| m {z k } k,n
k
δm,k
λ=0
= hψ̃m (λ)|ψ̃ ′ n (λ)i + A′m,n
Thus one can always find a λ-dependent, unitary transformation so that the states within
the multiplet do not contribute to the derivative.
Thus, for the degenerate result we obtain

DERIVATIVES OF ENERGIES AND WAVE FUNCTIONS IN THE DEGENERATE CASE
ǫ′n (λ) = hψn (λ)|Ŵ |ψn (λ)i

X hψm (λ)|Ŵ |ψn (λ)i
|ψn′ (λ)i = |ψm (λ)i
ǫn (λ) − ǫm (λ)
m(ǫm (λ)6=ǫn (λ))
where
hψn (λ)|Ŵ |ψm (λ)i = wn δm,n
for all states in the same multiplet.
13.3.2 First order perturbation theory

Now we use the derivatives Eq. 13.2 and Eq. 13.4 for the unperturbed system, and express the
perturbed values as Taylor series.
X 1 d j En
En (λ) = λj
j! dλj λ=0
j
This Taylor expansion is evaluated for the full perturbation that is for λ = 1.
X 1 d j En
⇒ En (λ = 1) =
j! dλj λ=0
j
Then we do the actual perturbations by truncating the expansion after the n-th-order term in λ for
the n-th order perturbation theory. In practice only the first and second order perturbation theory is
used frequently. For the first order perturbation theory we calculate

1st order dEn
En ≈ En (λ = 0) +
dλ λ=0
and for the second order perturbation theory we include also the next term

2nd order dEn 1 d 2 En
En ≈ En (λ = 0) + +
dλ λ=0 2 dλ2 λ=0
For the wave function we proceed completely analogously.

FIRST ORDER PERTURBATION THEORY
The energy En (1) and the energy eigenstates |ψn (1)i for a perturbed Hamiltonian H(1) = Ĥ0 + Ŵ
have the form
En (1) = En (0) + hψn (0)|Ŵ |ψn (0)i + O(Ŵ 2 )

X hψm (0)|Ŵ |ψn (0)i
|ψn (1)i = |ψn (0)i + |ψm (0)i + O(Ŵ 2 )
En (0) − Em (0)
m(6=n)
where En (0) are the energy eigenvalues and |ψn (0)i are the eigenstates of the unperturbed Hamilto-
nian Ĥ0
Note, that special precaution is necessary for degenerate states!
Linear response
Once we have an approximation for the wave function in a certain order, we can evaluate the pertur-
bation not only of the energy but of any other observable up to this order. Note, that some terms,
which turn up during squaring the wave functions, need to be dropped. We illustrate the principle on
the first-order result

hψn (λ)|Â|ψn (λ)i = hψn (0)| + λhψn′ (0)| Â |ψn (0)i + |ψn′ (0)iλ + O(λ2 )

= hψn (0)|Â|ψn (0)i + λ hψn (0)|Â|ψn′ (0)i + hψn′ (0)|Â|ψn (0)i

+λ2 λhψn′ (0)|Â|ψn′ (0)i + O(λ2 )
= hψn (0)|Â|ψn (0)i

+λ hψn (0)|Â|ψn′ (0)i + hψn′ (0)|Â|ψn (0)i + O(λ2 )
Thus we obtain
1st order
hψn (1)|Â|ψn (1)i ≈ hψn (0)|Â|ψn (0)i + hψn (0)|Â|ψn′ (0)i + hψn′ (0)|Â|ψn (0)i
LINEAR RESPONSE
Thus the linear response δAn of an expectation value An of an observable Â in an eigenstate |ψn i of
the Hamiltonian due to a perturbation Ŵ is obtained by the following expression.
 
def
X hψn |Â|ψm ihψm |Ŵ |ψn i 
δ Ân = hψn (1)|Â|ψn (1)i − hψn (0)|Â|ψn (0)i = 2Re 
En − Em
m(m6=n)
To give an example, this expression may be useful to determine the dipole of a molecular molecule
that is induced by an external electrical field. The proportionality constant is the polarization of the
molecule.
13.3.3 Second order perturbation theory

The second-order result for the energy eigenvalues is sufficiently simple so that we show it here. For
higher order terms even perturbation theory becomes increasingly difficult so that it is rarely done.
The second derivatives of En (λ) and |ψn (λ)i are obtained from Eq. 13.2 and Eq. 13.4, which
are valid for any value of λ, by deriving them once more with respect to λ. The resulting expression
contains first derivatives that are known already. Finally, we insert the result into the Taylor expansion
up to second order about λ = 0 and set λ = 1
En′ = hψn |Ŵ |ψn i

Eq. 13.4 X |hψm |Ŵ |ψn i|2
En′′ = hψn′ |Ŵ |ψn i + hψn |Ŵ |ψn′ i = 2
En − Em
m(6=n)
1
En = Ēn + En′ + En′′
2
X |hψ̄m |Ŵ |ψ̄n i|2
= Ēn + hψ̄n |Ŵ |ψ̄n i +
Ēn − Ēm
m(6=n)
Note that the sum extends over all states |ψm i except |ψn i. (The index n is fixed.)
E"n
Em En
Fig. 13.1: Second order correction En′′ by a second level Em . The result of the second order correction
is a “level repulsion”. It is evident that the correction diverges for En = Em indicating the breakdown
of the perturbation series for degenerate energy levels.
It may be interesting to consider the effect of the second order correction of the energy levels.
The contribution of nearby energy levels is largest, given that the matrix elements are roughly of
similar size.
13.4 Time-dependent perturbation theory

The prototypical application of time-dependent perturbation theory is the determination of optical
absorption spectra. The light-wave creates an oscillating electric field at the site of the transition,
and will induce transitions of electrons from an occupied orbital into an un-occupied orbital. Typically,
the electric field is sufficiently small to be treated as a small perturbation. The wave length is usually
sufficiently large that we only need to consider a fluctuating, but spatially constant electric field at
the point of the transition. For monochromatic light, the perturbation of the Hamiltonian due to the
light has the following form
Ŵ (t) = −~rˆq E
~ 0 sin(ωt)
~ 0 the amplitude of the electric field in the light wave,

where q is the charge of the excited electron, E
and ~ω the photon energy.
E2 E2 E2
hω hω hω
hω hω
E1 E1
E1
Fig. 13.2: Sketch of time dependent processes. Left optical absorption: A photon is absorbed and
an electron is lifted to a higher energy level. Right: Stimulated emission: If the higher energy level
is occupied and the lower level is filled, a photon can trigger the desorption of another one, while
the electron drops into the lower energy level. Stimulated emission underlies the workings of a laser.
Middle: Spontaneous emission: An electron can drop into the lower level while emitting a photon.
This process is similar to stimulated emission, but the photon that triggers the emission is a virtual
photon, i.e. a vacuum fluctuation of the electromagnetic field.
In the following we do not specialize on monochromatic light, but treat a general time-dependent
perturbation Ŵ (t) of the unperturbed and time-independent Hamiltonian Ĥ0 .

i ~∂t |ψi = Ĥ0 + Ŵ (t) |ψi (13.5)
We assume that we know the eigenstates of Ĥ0
Ĥ0 |φn i = |φn iEn (13.6)
Then we make the Ansatz

X i
|ψ(t)i = |φn ie − ~ En t an (t) (13.7)
n
and insert it into the time-dependent Schrödinger equation Eq. 13.5. Furthermore, we multiply the
equation from the left with hφm |,
Eq. 13.5
hφm |i ~∂t |ψ(t)i = hφm |Ĥ0 + Ŵ (t)|ψ(t)i
X X
Eq. 13.7 − ~i En t i
⇒ hφm |φn ie En an (t) + i ~∂t an (t) = hφm |Ĥ0 + Ŵ (t)|φn ie− ~ En t an (t)
n n
X X
Eq. 13.6 − ~i En t i
⇒ hφ |φ i e i ~∂t an (t) = hφm |Ŵ (t)|φn ie− ~ En t an (t)
n
| m{z n} n
δm,n
X
− ~i Em t i
⇒e i ~∂t am (t) = hφm |Ŵ (t)|φn ie− ~ En t an (t)
n
Division by the common phase factor ei Em t/~ yields

X i
i ~∂t am (t) = hφm |Ŵ (t)|φn ie− ~ (En −Em )t an (t) (13.8)
n
This equation is equivalent to the Schrödinger equation.

Now, we convert the differential equation Eq. 13.8 into an integral equation.
Z
i X t ′ i ′
am (t) = am (t0 ) − dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t an (t ′ ) (13.9)
~ n t0
Sofar we have not done any approximation. We simply changed to the interaction picture discussed
earlier in Section 8.5.3.
Just as in the time-independent perturbation theory, we introduce a scale factor for the pertur-
bation Ŵ .
Ĥ(λ) = Ĥ0 + λŴ
So that Ĥ(λ = 1) = Ĥ0 + Ŵ is the full Hamiltonian and Ĥ(λ = 0) = Ĥ0 is the unperturbed
Hamiltonian. Secondly we introduce the truncated Taylor expansions bn,k (t) for the coefficients
an (λ, t), which now not only have a time dependence but also a λ dependence.
k
X 1 d j an (t)
bn,k (t) = λj
j! dλj λ=0
j =0
The truncated Taylor expansions will converge to the correct result3 , if k approaches infinity, that is
lim bn,k (t) = an (t)

k→∞
3 We assume that λ = 1 lies within the convergence radius of the Taylor expansion.
The initial condition is chosen independent of λ.
bn,k (0) = an (0)
Now we divide4 the above equation Eq. 13.8 into orders of λ and drop all terms with a power of
λ higher than λp
Z
i X t ′ i ′
am (λ, t) = am (λ = 0, 0) − dt hφm |λŴ (t ′ )|φn ie− ~ (En −Em )t an (λ, t)
~ n 0
Z
i X t ′ i ′
⇒ bm,p (t) = am (0) − dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t λbn,p−1 (t)
~ n 0
Thus we obtain a recursive relation for increasingly improved estimates bn,p (t) for the coefficients,
where the first term is simply bn,0 = an (0). The first two terms have the form
Z
i X t ′ i ′
bm,1 (t) = am (0) − dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t an (0) (13.10)
~ n 0
Z
i X t ′ i ′
bm,2 (t) = am (0) − dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t an (0)
~ n 0
Z Z ′
1 X t ′ t
− 2 dt dt ′′ hφm |Ŵ (t ′ )|φk ihφk |Ŵ (t ′′ )|φn i
~ 0 0
n,k
i ′ i ′′
· e− ~ (Ek −Em )t e− ~ (En −Ek )t an (0) (13.11)
The first approximation bm,1 is the result from first order perturbation theory, and bm,2 is the result
from second-order perturbation theory. Notice that the highest power of Ŵ in the result is equal to
the order of the interaction.
13.4.1 Transition probabilities

Here we will use the expression from first-order perturbation theory to determine the excitation
probability of a system, such as an optical excitation due to a light pulse. The description here differs
from that in many text books. Instead of investigating the increase of the excited state probability
with time for a constant beam of light, we use here the excitation probability for a light pulse. This
derivation avoids some of the conceptual problems of the other derivation. This description seems
to be closer to a real experiment.
A more fundamental and explicit derivation for the transition probabilities can be found in ap-
pendix G on p. 269.
4 If there is an equation of the form
F (t, λ) = G(t, λ)
which is fulfilled for all values of λ, we can form the derivatives.
∂ n F (t, λ) ∂ n G(t, λ)
n
=
∂λ ∂λn
Since all the derivatives of left and right-hand side are equal, also every term of the Taylor expansion
1 ∂ n F (t, λ) 1 ∂ n G(t, λ)
n
=
n! ∂λ n! ∂λn
and the truncated Taylor expansions must be equal.
k k
X 1 ∂ n F (t, λ) k X 1 ∂ n G(t, λ) k
n
λ = λ
n=0
n! ∂λ n=0
n! ∂λn
This is a fundamental step done in perturbation theory.

We begin with the first order expression Eq. 13.10 for the time-dependent perturbation theory.
Z
Eq. 13.10 i X t ′ i ′
am (t) = am (0) − dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t an (0) + O(Ŵ 2 )
~ n 0
Let us insert a pulse with a given temporal shape and a finite duration.
Ŵ (t) = Ŵ0 ei ωt χ(t) (13.12)
where χ(t) is the shape of the pulse. We choose the pulse shape function wlog5 so that it is square
normalized.
Z ∞
dt χ∗ (t)χ(t) = 1 (13.13)
−∞
Note that a short pulse duration automatically leads to a poor frequency resolution. A monochromatic
pulse is necessarily of infinite duration. The Fourier transform of the pulse shape is obtained as
Z
χ(t) = dω χ̃(ω)ei ωt (13.14)
Z ∞
1
χ̃(ω) = dt χ(t)e−i ωt (13.15)
2π −∞
The final probability amplitude am

f
in the final state is
Z ∞
(f ) Eq. 13.10 i X ′ ′ − ~i (En −Em −~ω)t ′
am = am (0) − hφm |Ŵ0 |φn i dt χ(t )e an (0)
~ n
| 0 {z }
2π χ̃( 1~ (En −Em −~ω))
The expression in parenthesis is simply the Fourier transform of the pulse shape. With Eq. 13.15 we
obtain
(f ) i X 1
am = am (0) − hφm |Ŵ0 |φn i 2π χ̃( (En − Em − ~ω))an (0)
~ n ~
After the pulse, the probability of finding the system in the excited state |φm i, given that it started
out in state |φn i, is
TRANSITION PROBABILITY
4π 2 1
Am,n = |hφn |Ŵ0 |φm i|2 |χ̃( (En − Em − ~ω))|2 (13.16)
~2 ~
Now we would like to extract the essentials of the term contributed by the pulse shape.
Z ∞ Z Z Z ∞
∗ ′ ∗ ′ ′
1= dt χ (t)χ(t) = dω dω χ̃ (ω )χ̃(ω) dt ei (ω−ω )t
−∞ −∞
Z Z ′ ′
ei (ω−ω )T − e−i (ω−ω )T
= lim dω dω ′ χ̃∗ (ω ′ )χ̃(ω)
T →∞ i (ω − ω ′ )
Z Z
sin((ω − ω ′ )T )
= dω dω ′ χ̃∗ (ω ′ )χ̃(ω)2 lim
T →∞ (ω − ω ′ )
5 wlog is a abbreviation of “without loss of generality”
Here we show that
sin(xT )
lim = πδ(x)
T →∞ x
R
We use that the integral dx sin(ax)
x = π.
Z Z Z
sin(xT ) sin(y ) y sin(y )
dx f (x) lim = dy lim f ( ) = dy f (0)
T →∞ x y T →∞ T y
| {z }
π
So that
Z ∞ Z
∗
1= dt χ (t)χ(t) = 2π dω |χ̃∗ (ω)|2
−∞
This immediately shows us if the pulse is very monochromatic, the square of the pulse shape can be
related to a delta function.
h(ω−hωi)2 i−→0 1
|χ̃(ω)|2 → δ(ω)
2π
When we insert this result into Eq. 13.16 we obtain the special case of monochromatic light.

h(ω−hωi)2 i−→0 2π 2 1
Am,n → |hφn |Ŵ0 |φm i| δ ω − (ǫm − ǫn )
~2 ~
2π
= |hφn |Ŵ0 |φm i|2 δ (ǫm − ǫn − ~ω))
~
This is Fermi’s golden rule
FERMI’S GOLDEN RULE
The transition probability due to a monochromatic pulse with shape Ŵ (t) = Ŵ0 ei ωt χ(t) with a
smooth, normalized pulse shape function χ(t) is
2π
Am,n = |hφn |Ŵ0 |φm i|2 δ (ǫm − ǫn − ~ω)) (13.17)
~
Fermi’s golden rule says that an excitation with monochromatic light is only possible if the energy of
the incident light beam have exactly that energy which is necessary to lift the system into a higher-,
or lower-energetic state.
We recognize that Fermi’s golden rule is a special case of a more general expression Eq. 13.16
for the transition probability with arbitrary pulse shapes.
From the more general expression Eq. 13.16 we can also see why femtosecond spectroscopy
cannot have a good frequency resolution. This fact is closely related to Heisenberg’s uncertainty
principle which links a good spatial resolution to a poor momentum resolution.
With these transition rates we can investigate the stimulated absorption and stimulated emis-
sion of light. The stimulated emission of light underlies the function of a laser. Electrons are prepared
in their excited state. They remain there until light stimulates emission, which pumps further en-
ergy into the laser beam. Important for the function of a laser is that the spontaneous emission of
light is sufficiently small so that the system can be prepared in its excited state. The latter effect
can be considered as the spontaneous emission of light induced by the zero point vibrations of the
electromagnetic field[7].
13.5 Variational or Rayleigh-Ritz principle

The variational principle is a useful scheme to find the ground state of a system. The ground state
is the energy eigenstate with lowest energy. It is commonly known under the name Rayleigh-Ritz
principle or Ritz-method.[27]6
Nearly all methods to investigate the electrons in real solids or molecules are based on the varia-
tional principle.
The basis for the variational principle is the following observation:
hψ|Ĥ|ψi
• The energy E[ψ] = hψ|ψi evaluated for an arbitrary state |ψi is higher than the energy of the
ground state, and
• that it this energy becomes equal to the ground-state energy, if we insert the ground state for
|ψi
This is very handy if we cannot solve the Schrödinger equation exactly. We just choose an arbitrary
function, and we can evaluate, if not the ground state energy itself, at least a strict upper bound for
it. The better the guess for the wave function the closer we approach the ground state energy.
This is shown as follows:
P
PROOF: let us expand |ψi = n |φn icn with cn = hφn |ψi into the orthonormal system of eigenstates |φn i
of the Hamiltonian with eigenvalues En , that is Ĥ|φn i = |φn iEn and hφn |φm i = δn,m .
1̂
z
X }| {
hψ|Ĥ |φn ihφn | |ψi
def hψ|Ĥ|ψi n
E[ψ] = = X
hψ|ψi hψ| |φn ihφn | |ψi
n
| {z }
1̂
Ĥ|φn i cn
P z }| { z }| { P 2
n hψ |φn iEn hφn |ψi n |cn | En
= P = P 2
n hψ|φn ihφn ψi n |cn |
P 2
n |cPn | (En − E0 )
E[ψ] − E0 = 2
≥0
n |cn |
Since each term is non-negative the result must be non-negative. The equal sign holds, if all coefficients
vanish except that for the ground state, i.e. for En = E0 . Hence the equal sign holds only for the ground
state. q.e.d
The variational principle is the basis for almost all electronic structure methods, that try to
determine the electronic wave functions for real materials. Here we work with N-particle systems,
which are often approximated in one or the other way by non-interacting N-particle systems.
Since these systems are often enormously complex, we choose a set of basis functions |χi i. These
functions could be atomic orbitals. In this case, the method would be called “linear superposition of
atomic orbitals” (LCAO).7
Then we make an ansatz of these basis functions.
X
|ψn i = |χi ici ,n
i
6 Walter Ritz, born 1878 in Sion, Switzerland, died 1909,Goettingen, Germany. swiss theoretical physicist, known
for the variational principle named after him as well as for the Rydberg-Ritz formula, describing the spectral lines of
the hydrogen atom. The contribution of Ritz was that the spectral lines can be constructed as differences of energy
levels. Walter Ritz died in the young age of only 31 years on tuberculosis.
7 Note that LCAO is not a very good technique for practical electronic structure methods.
and minimize the total energy for this wave function.

If we multiply the Schrödinger equation from the left with a basis function, we obtain a eigenvalue
equation in matrix form
Ĥ|ψn i = |ψn iEn

X X
⇒ hχi |Ĥ|χj icj,n = hχi |χj icj,n En
j j
⇒ H~
cn = O ~
cn En
where Hi ,j = hχi |Ĥ|χj i are the Hamilton matrix elements and Oi ,j = hχi |χj i are the Overlap matrix
elements. If the set of basis functions |χn i is complete, this was nothing but a change of the
representation. No approximation has been done so far.
However, if the basis set is complete, it is also infinite. Such a problem cannot be handled on the
computer. In practice we thus have to limit the dimension of the vectors and matrices. Only a finite
set of wave functions is used. This however is a fairly uncontrolled approximation. The variational
principle provides us with two important conclusions about the approximate solution
• The energy is an upper bound of the ground state energy
• The deviations of the approximate wave functions only induces changes of second order in the
total energy.
The groundstate of a non-interacting N-Fermion system is a Slater determinant formed by the N

eigenstates of the corresponding one-particle Hamiltonian. The energy of this Slater determinant is
given by the sum of the corresponding eigenvalues of the one-particle Hamiltonian. The variational
principle shows that every Slater determinant of orthonormal orbitals has an energy larger or equal
to the ground state.
E0 ≤ min F (~
c0 , λ)
c0
~
F (~ c0 − λ (~
c0 , λ) = c~0 H ~ c0 O c~0 − 1)
∂F X
= Hi ,j cj,0 − λO
∂ci ,0
j
where λ is a Lagrange multiplier for the constraint ~ c0 = 1, that the wave function is normalized.
c0 O ~
P
PROOF: We describe the trial wave function |ψi = n |φn icn as a linear superposition of basis functions |φn i
with variable coefficients cn . Note, that the basis functions need not be complete, orthogonal or normalized.
P ∗
hψ|Ĥ|ψi n,m cn hφn |Ĥ|φm icm
E[ψ] = = P ∗
hψ|ψi n,m cn hφn |φm icm
The minimum condition is dE/dcn = 0 and dE/dcn∗ = 0. We have to consider here the coefficient and its
complex conjugate as independent quantities. There are two derivatives, because each coefficient has a real
and an imaginary part. The coordinate transformation from real and imaginary part to a complex coefficient
and its complex conjugate is a common trick for handling complex numbers.
P ∗
dE[ψ] d n,m cn hφn |Ĥ|φm icm
= P
dck∗ dck∗ ∗
n,m cn hφn |φm icm
P
hφk |Ĥ|φm icm
= Pm ∗
P ∗ P
n,m cn hφn |Ĥ|φm icm m hφk |φm icm
− P ∗ × P ∗
c
n,m n hφ |φ
n m mic n,m cn hφn |φm icm
From the condition that the derivatives vanish, we obtain an equation for the coefficients
X X
⇒0= hφk |Ĥ|φm icm − hφk |φm icm Ē
m m
P ∗
n,m cn hφn |Ĥ|φm icm
Ē = P ∗
What we obtained is a set of nonlinear equations for the coefficients, because Ē depends itself on the
coefficients. However, the problem can be solved as eigenvalue problem, after identifying Ē with the lowest
eigenvalue. q.e.d
The eigenvalue equation is nothing more than solving the Schrödinger equation in a limited Hilbert
space defined by the basis functions. The more basis functions are included, the lower the energy
and the more accurate is the result. By including more and more basis functions we can therefore
approach the correct solution arbitrarily close.
The variational principle is so powerful, because any deviation in the wave function from the
ground-state wave function enters only to second order into the energy expression.
PROOF: Here we show that the deviation of the trial wave function from the ground state wave function
enters only to second order in the energy.
We expand the trial wave function |ψi into Peigenstates |φn i of the Hamiltonian with energies En . The
coefficients of the trial wave function |ψi = n |φn icn are written as cn = δn,0 + δcn , where δcn describes the
deviation from the ground state |φ0 i.
hψ|Ĥ|ψi
E=
hψ|ψi
P 2
P 2
n |δn,0 + δcn | En n (δn,0 + 2Re[δcn ]δn,0 + |δcn | )En
= P 2
= P 2
n |δn,0 + δcn | n (δn,0 + 2Re[δcn ]δn,0 + |δcn | )
2
P
P n |δcn | En
E0 (1 + 2Re[δc0 ]) + n |δcn |2 En E0 + 1+2Re[δc 0]
= P = P 2
(1 + 2Re[δc0 ]) + n |δcn |2 n |δcn |
1 + 1+2Re[δc 0]
X En − E0 2 4
= E0 + |δcn | + O(δcn )
n
1 + 2Re[δc0 ]
X
= E0 + (En − E0 )|δcn |2 + O(δcn )3
n
1
In the last steps we used 1+x = 1 − x + O(x 2 ). Thus we find that E = E0 + O(δcn2 ) q.e.d.
A small error in the trial wave function produces an even smaller error in the energy. Hence, a
very good estimate of the ground state energy can be obtained with fairly poor trial wave functions.
It must be noted however, that the error in other observables is still of first order. One must not be
misguided from a converged total energy, that the wave function itself is already of good quality. (A
common mistake!)
Proof of the variational principle for non-interacting many-particle systems (Home study)
Let us start with N one particle states |ψn i, which are orthonormal, that is hψn |ψm i = δn,m . We
need to show that
N
X N
X
hψn |Ĥ|ψn i = En
n=1 n=1
where En are the eigenvalues of the one-particle Hamiltonian. Let |φi i be the exact eigenstates of
the one-particle Hamiltonian.
Ĥ|φj i
N
X N
X X N X
X z }| {
hψn |Ĥ|ψn i = hψn |Ĥ |φj ihφj | |ψn i = hψn |φj iEj hφj |ψn i
n=1 n=1 j n=1 j
| {z }
1̂
 
X  XN  X
 
= Ej hφj |
 |ψn ihψn | |φj i
= Ej hφj |P̂n |φj i
j  n=1  j
| {z }
=:P̂
We have introduced P̂ , the projection operator on the approximate wave functions chosen. We can
now show that
• 0 ≤ hφj |P̂ |φj i ≤ 1.
P
• j hφj |P̂ |φj i = N. This is shown as follows:
X X N
X N
X X
hφj |P̂ |φj i = hφj | |ψn ihψn |φj i = hψn | |φj ihφj | |ψn i
j j n=1 n=1 j
| {z }
=1̂
N
X
= hψn |ψn i = N
| {z }
n=1
=1
Now we can calculate the difference from the groundstate energy as

N
X N
X N
X
E[ψ1 , . . . , ψN ] − Ej = hψn |Ĥ|ψn i − En
j =1 n=1 n=1
N
X ∞
X
= Ej hφj |P̂ − 1̂|φj i + Ej hφj |P̂ |φj i
| {z } | {z }
j =1 j =N+1
≤0 ≥0
N
0≤hφj |P̂ |φj i≤1 X ∞
X
≥ EN hφj |P̂ − 1̂|φj i + EN+1 hφj |P̂ |φj i
| {z } | {z }
j =1 j =N+1
≤0 ≥0
∞
X
= (EN+1 − EN ) hφj |P̂ |φj i ≥ 0
j =N+1
13.6 WKB-Approximation
The WKB approximation is useful for quasi-one-dimensional potentials, which have little structure on
a length scale of the wave length of the particle considered. It is used for example in semiconductor
physics to investigate the transport of charge carriers.
The WKB approximation8 named after Wentzel, Kramers and Brillouin[28, 29, 24] is also called
the quasi-classical approximation.
A very good description of he WKB approximation can be found in the book by Griffiths[10].
Apparently, the essential features of the WKB approximation have been known as early as
1912[30]. The WKB method is applicable if the potential is rather smooth, so that is it
varies only little over a wave length of the particle. It is also called quasi-classical approximation,
because it is a type of perturbation theory, where the small parameter is Planck’s constant ~. (Recall
that ~ = 0 is the classical limit.)
For a given potential, the wavelength of a particle with a given energy scales with ~. Thus the
smaller ~, the better the condition, that the potential varies slowly over the wavelength of a particle,
is fulfilled. Even though the WKB approximation can also be applied in higher dimensions, it most
commonly applied only in one dimension, where it is particularly simple. (The generalization of the
WKB method to higher dimensions is nontrivial.)
The WKB method starts with the following Ansatz for the wave function
i
ψ(x) = e ~ S(x)
~ S(x) S(x)

⇒ p̂ψ(x) = ∂x ei ~ = ei ~ ∂x S
i
~ S(x) S(x)
~
⇒ p̂ ψ(x) = [ ∂x ]2 ei ~ = ei ~ (∂x S)2 + ∂x2 S
2
i i
When we insert this result into the one-dimensional Schrödinger equation

h −~2 i
Eψ = ∂x2 + V (x) ψ (13.18)
2m
we obtain
1 ~
E= (∂x S)2 + V (x) + (∂ 2 S) (13.19)
2m 2mi x
8 In the Netherlands it is called KWB, in France BWK and in Britain JWKB (for Jeffreys). Kramers was dutch,
Brillouin french and Jeffreys was british.

Note that we keep the energy fixed in this case. The typical application is to find an approximation
for a wave function at a given energy irrespective of the boundary conditions.9
Now, we expand S in powers of ~, that is
S = S0 (x) + ~S1 (x) + . . .
We insert this series into the equation (13.19) and order the terms according to powers of ~.10
1 ~
0= (∂x S0 + ~∂x S1 + . . .)2 + V (x) − E + (∂ 2 S0 + ~∂2 S1 + . . .)
2m 2mi
h 1 i
= (∂x S0 )2 + V (x) − E
2m
h1 1 2 i
+ (∂x S0 )(∂x S1 ) + ∂ S0 ~ + O(~2 ) (13.20)
m 2mi x
For the lowest order, we obtain an expression for S0
(∂x S0 )2
+ V (x) − E = 0
2m
This is nothing but the classical energy momentum relation, if we identify S0 with the classical action
and ∂x S0 with the momentum.11
p
p(x) = ∂x S0 = ± 2m(E − V (x)) (13.21)
Z x Z x p
′ ′
S0 (x) = ± dx p(x ) + const. = ± dx ′ 2m(E − V (x ′ )) + const.
We obtained a closed expression for the lowest term. The corresponding wave function is oscillatory,
but with a local wave vector ~k = dd Sx that depends on space.
i i
ψ(x) = Ae ~ S0 (x) + Be− ~ S0 (x)
9 This differs from the applications of first order perturbation theory, where we wanted to find a better bound state.
Thus the boundary conditions have been fixed, but the energy depended on the perturbation.
10 Note that this step is analogous to that in first order perturbation theory.
11 The momentum is indeed the gradient of the action which can be shown as follows. We start with the variation
of the action. However, in contrast to the derivation of the Euler Lagrange equations, we allow also variations of the
end points of the trajectory.
Z
S[x(t)] = dt L(x(t), ẋ (t), t)
Z
∂L ∂L
δS = S[x(t) + δx(t)] − S[x(t)] = dt δx(t) + δ ẋ (t)
∂x ∂ ẋ
Z
∂L d ∂L d ∂L
= dt δx(t) + δx(t) − δx(t)
∂x dt ∂ ẋ dt ∂ ẋ
Z Z
∂L d ∂L d ∂L
= dt δx(t) δx(t) − + dt δx(t)
∂x dt ∂ ẋ dt ∂ ẋ
| {z }
ELG
If we consider now physical trajectories, which fulfill the Euler-Lagrange equation, the term denoted by “ELG” vanishes
and only the last term remains. Noting that the canonical momentum is defined as p = ∂L/∂ ẋ , we obtain
 
Z tf  ∂L 
d  
δS = dt  δx(t) = p(tf )δx(tf ) − p(ti )δx(ti )
ti dt  ∂ ẋ 
|{z}
p
If we now consider the variation of the action under a variation of the end point along the trajectory, we obtain
dS
= p(tf )
dx(tf )
In the classically allowed region, i.e. where V (x) < E, the maxima and minima of the wave function
remain constant in space.
p In the classically forbidden region, the momentum is imaginary. p(x) = ±i λ(x) with λ(x) =
2m(V − E). is imaginary. Therefore, also the action is imaginary, so that the wave function
consist out of an exponentially decreasing and an exponentially increasing contribution.
Let us now consider the next higher order in Eq. 13.20.
Eq. 13.20 1 1 2
0 = (∂x S0 )(∂x S1 ) + ∂ S0
m 2mi x
i
0 = p(x)(∂x S1 ) − ∂x p(x)
2
i ∂x p(x) i p
∂x S1 = = ∂x ln[|p(x)|] = ∂x i ln[ |p(x)|]
2 p(x) 2
p
S1 = i ln[ |p(x)|] + const.
Thus we obtain for as result for the wave function which now includes also an amplitude modulation.
WAVE FUNCTION IN THE WKB APPROXIMATION
1 h iR i
R i
ψ(x) = p Ae ~ d xp(x) + Be− ~ d xp(x)
|p(x)|
where the momentum p(x) is according to Eq. 13.21

p
p(x) = 2m(E − V (x))
The WKB approximation always requires the first two terms, S0 and S1 , to provide a meaningful
result. This is because there is a prefactor 1/~ in front of the expansion. Thus, the second term,
S1 , is required to obtain the wave function to zeroth order in ~.
The zeroth order, however, plays an important historical role, because it captures Bohr’s quan-
tization condition, which was one of the first steps towards quantum theory. In that theory, the
energy levels have been quantized by the requirement that only those closed classical trajectories
are allowed, for which the change ofR the phase S0 (x) along a closed classical path is a multiple of
Planck’s quantum h = 2π~, that is dxp(x) = ~n. This empirical assumption lead to Bohr’s atom
model, which captured the observed excitation energies between the states of the hydrogen atom.
Note that along a closed trajectory S0 must change by 2πn along a closed trajectory in order to
avoid a discontinuity in the wave function.
13.6.1 Classical turning points

We can already see that the wave function diverges at classical turning points, that is the points
where E = V (x). At those points the series expansion Eq. 13.20 does not converge.
Let us investigate Eq. 13.19 and resolve it for ∂x S
r
~
(∂x S) = ± 2m(E − V (x)) − (∂x2 S) (13.22)
i
s
p ~(∂x2 S)
= ± 2m(E − V (x)) 1 + i (13.23)
2m(V − E)
The expansion of S in orders of ~ is identical to a Taylor expansion of the last factor. Since the
√
convergence radius of the Taylor series of 1 + x is equal to one12 , the expansion converges only if
2m|E − V |
|∂x2 S| <
~
This term is only fulfilled at the classical turning point, if ∂x2 S vanishes to all orders in ~.
Near classical turning points we need to evaluate the Schrödinger equation directly. The problem
can be simplified if the potential is linearized near the turning point. The comparison of asymptotic
solutions of the exact solution and that of the WKB approximation provides connection formulas.
We will not go into that detail, but only state the final result. The derivation is given in the appendix
App. I. Here we only list the final result.
MATCHING CONDITION FOR THE WKB SOLUTION AT THE CLASSICAL TURNING POINT
The WKB wave function for a potential with positive slope at the classical turning point can be
written as
Z
1 1 xc l ′ π
Ψ(x < xcl ) = p cos( [ dx p(x ′ )] − )
|p(x)| ~ x 4
1 1
R ′ ′
Ψ(x > xcl ) = p e− ~ d x p(x )
|p(x)|
and for negative slope

Z
1 1 x ′ π
Ψ(x > xcl ) = p cos( [ dx p(x ′ )] − )
|p(x)| ~ xc l 4
1 1
R ′ ′
Ψ(x < xcl ) = p e ~ d x p(x )
|p(x)|
13.7 Numerical integration of one-dimensional problems

The one-dimensional Schrödinger equation can readily be integrated numerically.
−~2 2
[ ∂ + V ]ψ = Eψ
2m x
We use here a method, that is completely analogous to the Verlet, which has been introduced in
ΦSX: Klassische Mechanik to numerically solve Newton’s equations of motion. The basic idea is to
place the wave function onto a grid xi = ∆i , where ∆ is the grid spacing and i is the integer index of
the grid point. Now we replace the derivative by the corresponding differential quotients
ψi +1 − ψi −1
∂x ψ(xi ) ≈
2∆
2 ψi +1 − 2ψi + ψi −1
∂x ψ(xi ) ≈
∆2
12 The convergence radius of a Taylor expansion is at most the distance of the nearest singularity. The squareroot
√
has a diverging derivative at the origin, that is for xsi ng = −1 in 1 + x. Thus the convergence radius is at most
|xsi ng | = 1
and obtain the discretized Schrödinger equation13 . We resolve it for ψi +1
−~2
(ψi +1 − 2ψi + ψi −1 ) + V ψi = Eψi
2m∆2
2m(E − V (xi ))∆2
ψi +1 = 2ψi − ψi −1 − ψi
~2
Thus knowing the value of ψ at two successive grid points we can evaluate the wave function
recursively on all other grid points. With the formula given above we obtain the wave function for
increasing x. If we resolve the discretized Schrödinger equation for ψi −1 , we obtain a recursive
equation that leads towards smaller x.
13.7.1 Stability
Numerical solutions are not exact. Two aspects need to be understood. The first is accuracy
and the second is stability. Accuracy can be improved by decreasing the grid spacing ∆. The
computation becomes more involved so that we have to find a good compromise between accuracy
and computational effort. When we try to minimize the computational effort by increasing the grid
spacing we run into the problem that the scheme may become unstable, and that it will produce
qualitatively incorrect results.
Consider for example the classically forbidden region, where we have exponentially increasing and
exponentially decreasing solutions. If we integrate downward an exponentially decreasing solution, a
small error will create a small component of the increasing solution, and therefore magnify. Such
a procedure is therefore numerically unstable. If, on the other hand, we integrate in the opposite
direction, and integrate the solution in choosing the direction in which it increases, any error will
exponentially die out. Therefore, one is not entirely free in the direction: In the classical forbidden
region one should always propagate the solution in the direction where the wave function increases
in absolute value.
If one follows an exponentially increasing wave function we need to remember that a computer
can only represent numbers up to a maximum value, which is 1.79 × 10308 for an 8 byte real number.
While this is a large number, it can still be reached quickly with an exponential increasing function.
It is therefore necessary to check the size of the wave function and rescale the entire solution (on
all grid points). A computer is not only limited in representing large numbers, but there is also a
smallest number the computer can describe. Where the absolute value of the wave function drops
below 2 × 10−308 , the smallest number on an 8-Byte computer, the wave function will be set exactly
to zero, but there the result can be considered insignificant for most purposes.
For an oscillatory solution we need to choose the step size sufficiently small. One should choose
the step size smaller than one percent of the period of the oscillation, that is
1 2π~
∆< p
10 2m(E − min[V ])
If the step size is larger than one third of the oscillation period, the algorithm becomes unstable.
13 The accuracy of these expressions can be tested by inserting the Taylor expansion of ψ
1 2 2 1 1 4 4
ψi +1 = ψi + ∆∂x ψ + ∆ ∂x ψ + + ∆3 ∂x3 ψ + ∆ ∂x ψ + . . .
2 6 24
1 1 1 4 4
ψi −1 = ψi − ∆∂x ψ + ∆2 ∂x2 ψ + − ∆3 ∂x3 ψ + ∆ ∂x ψ + . . .
2 6 24
ψi +1 − ψi −1 1
= ∂x ψ + ∆2 ∂x3 ψ + O(∆)4
2∆ 6
ψi +1 − 2ψi + ψi −1 1 2 4
= ∂x2 ψ + ∆ ∂x ψ + O(∆)4
∆2 12
Thus the error introduced is of order ∆2
13.7.2 Natural boundary conditions

For bound states an important question is how to start up the solution. We cannot use two zero
values of the wave function because then the result will be the zero state. Here we can use the WKB
approximation. We use the function
√
− ~1 d x 2m(V (x)−E)
R
1
ψ(x) = √ 4
e
2m(V (x)−E)
to evaluate the wave function on a grid starting from a point where the WKB approximation is
safe. Then we use two such values to follow the solution into the classically allowed region.
We can do this from the left and from the right, but since we do not yet know the correct energy,
the solution will not match, where they meet. Here we can use another trick.
We expand the solution in a Taylor series ψ(E) = ψ(E0 ) + (E − E0 ) dd ψ E about one energy. To
first order
−~2 2 dψ
[ ∂x + V − E] = ψ(E)
2m dE
We discretize this differential equation to obtain dd ψ

E . To start the solution up, we must of course
also form the energy derivative of the WKB solution. Then we decide on a matching point in the
classically allowed region, and evaluate values and derivatives using the differential quotient.
A prediction of the energy can be obtained from the condition that the logarithmic derivatives
match
∂x ψ+ + dE∂x ψ̇+ ∂x ψ− + dE∂x ψ̇−
=
ψ+ + dE ψ̇+ ψ− + dE ψ̇−
which results in a quadratic equation for dE = E − E0
A + dEB + CdE 2 = 0
A = ψ− ∂x ψ+ − ψ+ ∂x ψ−
B = ψ̇− ∂x ψ+ − ψ̇+ ∂x ψ− + ψ− ∂x ψ̇+ − ψ+ ∂x ψ̇−
C = ψ̇− ∂x ψ̇+ − ψ̇+ ∂x ψ̇−
We first test if there is a solution by checking if B 2 − 4AC > 0. If there is no such solution, we
linearize the equation and use E = E0 − B/A as the next better estimate for the energy.
If there is such a solution for dE from the quadratic equation, we solve this equation and take the
solution for dE with the smaller absolute value to obtain a new estimate for the energy E = E0 + dE.
Using this value, we repeat the procedure until the energy is converged. Now we rescale the solutions
so that their values are identical at the matching points.
13.7.3 Radial wave functions for atoms

The radial Schrödinger equation has the form ψ(~r) = φℓ,m (|r |)Yℓ,m (~r)
h −~2 1 ~2 ℓ(ℓ + 1) i
0= ∂r2 r + + V (r ) − E φℓ,m (r )
2m r 2mr 2
h 2 2 i
1 −~ 2 ~ ℓ(ℓ + 1)
= ∂ + + V (r ) − E r φℓ,m (r )
r 2m r 2mr 2
For an atom the potential has singularities at the origin. Therefore, a uniform step-size will not
be appropriate. We need to choose a step size that becomes finer near the origin. Therefore, we
introduce a logarithmic grid. First we introduce a variable transform
r (x) = eαx
which results in
df df dr df 1
= = αr ⇒ ∂r = ∂x
dx dr dx dr αr
1 1
∂r2 = − 2 ∂x + 2 2 ∂x2
αr α r
and obtain, after multiplication from the left with r , the transformed radial Schrödinger equation
h −~2 −~2 ~2 ℓ(ℓ + 1) i
2
∂ x − ∂ x + + V (r ) − E r φℓ,m (r ) = 0
2mα2 r 2 2mαr 2 2mr 2
h −~2 2 2
α ~ ℓ(ℓ + 1) i
⇒ (∂x2 − α∂x ) + + α2 r 2 (V (r ) − E) r φℓ,m (x) = 0
2m 2m
This form has no more divergent terms at the origin. We discretize along x and use standard
procedures. The grid becomes infinitely fine at the origin, so that the divergent terms can be easily
integrated. At the origin of r that is for x → −∞ we need to obey the correct initial conditions. We
choose the innermost grid point and use the asymptotic solution at the origin to evaluate the wave
function at the first two grid points.
13.8 Exercises
13.8.1 A model for optical absorption
Consider a on-dimensional particle in a box with light shining on it. The light creates an electric field
h i
E(x, t) = E0 x θ(−t)ei (ω−i λ)t + θ(t)ei (ω+i λ)t
λ describes, how the light is turned on and off again.

Calculate the occupation of the first and second excited state as function of time, if the system
starts in the ground state.
Solution:
The box has the length L and extends from − L2 to L2 . Inside the box, the first three states have
the form
r
2 πx
φ0 (x) = cos( )
L L
r
2 2πx
φ1 (x) = sin( )
L L
r
2 3πx
φ2 (x) = cos( )
L L
Outside of the box the wave functions vanish. Their energies are
~2 π 2 ~2 π 2
E0 = =
2m L 2mL2
2
2
~ 2π ~2 π 2
E1 = =4
2m L 2mL2
2
2
~ 3π ~2 π 2
E2 = =9
2m L 2mL2
The time dependent states without perturbation have the form
i
Ψj (x, t) = φj (x)e ~ Ej t
Now we use Eq.?? to evaluate the coefficients.

Z
i X ∞ ′ i ′
am (t) = am (−∞) − dt hφm |W (t ′ )|φn ie− ~ (Ej −En )t an (t ′ )
~ n −∞
First we need the coefficients at t = −∞. If the system is in the ground state,
aj (−∞) = δj,0
Next we need to evaluate the matrix elements

h i
hφm |W (t ′ )|φn i = hφm |E0 x|φn i θ(−t)ei (ω−i λ)t + θ(−t)ei (ω+i λ)t
We use
   
Z Z 
 

 i kx
 i kx
 − 1 1 ei kx = xe + e
1 i kx 

dx |{z} ei kx =
x |{z} dx ∂x  x e


|{z} i k  |{z} i k 
 ik k2
f g′  f | {z } f ′ | {z }
g g
to evaluate the formulas needed for the evaluation of the matrix elements
Z
dx x cos(k1 x) sin(k2 x)
Z
ei k1 x + e−i k1 x ei k2 x − e−i k2 x
= dx x
2 2i
Z
1
= dx x ei (k1 +k2 )x − ei (k1 −k2 )x + e−i (k1 −k2 )x − e−i (k1 +k2 )x
4i

−x ei (k1 +k2 )x ei (k1 −k2 )x e−i (k1 −k2 )x e−i (k1 +k2 )x
= − + −
4 k1 + k2 k1 − k2 k1 − k2 k1 + k2
i (k1 +k2 )x i (k1 −k2 )x −i (k1 −k2 )x

1 e e e e−i (k1 +k2 )x
+ − + −
4i (k1 + k2 )2 (k1 − k2 )2 (k1 − k2 )2 (k1 + k2 )2

−i x sin((k1 + k2 )x) sin((k1 − k2 )x)
= −
2 k1 + k2 k1 − k2

1 sin((k1 + k2 )x) sin((k1 − k2 )x)
+ −
2 (k1 + k2 )2 (k1 − k2 )2
Z a
sin((k1 + k2 )a) sin((k1 − k2 )a)
⇒ dx x cos(k1 x) sin(k2 x) = −
−a (k1 + k2 )2 (k1 − k2 )2
Now we can evaluate the matrix elements The matrix elements between two identical states
vanishes, because the probability density of a state is inversion symmetric, while x is antisymmetric
with respect to inversion. Thus the integrand is antisymmetric. Hence the integral vanishes.
hφi |E0 x̂|φi i = 0
Similarly the product of φ1 and φ3 is symmetric with respect to inversion, so that the matrix
elements between these two states vanishes as well.
hφ1 |E0 x̂|φ3 i = 0

We still need to evaluate the matrix elements with the second state:
Z L
2 2 π 2π
hφ1 |E0 x̂|φ2 i = E0 dx cos( x) sin( x)
L −2L L L
" #
3π L πL
2 sin( L 2 ) sin(− L 2 )
= E0 −
3π 2
2
L − Lπ
L
2 " #
2 L sin( 3π
2 ) π
= E0 − sin(− )
L π 9 2
16
= LE0
9π 2
and between the second and the third
Z L
2 2 3π 2π
hφ3 |E0 x̂|φ2 i = E0 x) sin( x)
dx cos(
L − L2 L L
" #
L
2 sin( 5π
L 2) sin( πL L2 )
= E0 −
L 5π 2 π 2
L L
" #
2 L 2 sin( 5π 2 ) π
= E0 − sin( )
L π 25 2
48
=− LE0
25π 2
Now we insert these matrix elements into the first-order expression

Z ∞ h i i
i 16 ′ ′ i (ω−i λ)t ′ ′ i (ω+i λ)t ′ ′
a2 (∞) = − 2
LE 0 dt θ(−t )e + θ(t )e e− ~ (E2 −E1 )t
~ 9π −∞
Z 0 Z ∞
i 16 ′ ~i (~ω−(E2 −E1 )−i ~λ)t ′ ~i (~ω−(E2 −E1 )+i ~λ)t ′
=− LE 0 dt e + dt e
~ 9π 2 −∞ 0
16 2i ~λ
= − 2 LE0
9π (~ω − (E2 − E1 ))2 + ~2 λ2
13.8.2 Gamow’s theory of alpha decay

See Introduction to quantum mechanics by David J. Griffiths
13.8.3 Transmission coefficient of a tunneling barrier

See Introduction to quantum mechanics by David J. Griffiths
13.8.4 Fowler-Nordheim tunneling

Editor: not ready!!!
Fowler-Nordheim tunneling is used to determine the band offsets between materials, e.g. between
a metal and an oxide.
p 3
!
2m ∗
2 4 ox (qφ B ) 2
JF N = CF N Eox exp −
3 q~ Eox
Here Eox is the electric field in the oxide, φB is the barrier height, that is the conduction band offset.
The effective mass of electrons in SiO2 is mox

∗
= 0.42me

JF N 1
One plots ln E 2 versus Eox
ox
Fowler nordheim tunneling is applicable when Eox d > φB , that is the oxide must be thick enough
that the electrons can tunnel into the oxide conduction band.
This formula is derived using the WKB method.
======================================
Other transport processes are the Pole Frenkel emission. This mechanism dominates if the oxide
has a high density of traps. (e.g.Si3 N4 ). The current is a simple drift mechanism
J~ = qnµE
~
where the carrier density N depends exponentially on the depth of the trap, which is connected to
the electric field.
" r !#
q qE
n = n0 exp − φB −
kB T πǫN
so that the total current equals

" r !#
~ q qE ~
J = qn0 exp − φB − µE
kB T πǫN
============================== Space charge limited current see www B. Van Zegh-

broeck.
Chapter 14
Relativistic particles
In the first chapter we have learned that the hanging linear chain describes relativistic particles, and
that the Schrödinger equation is the non-relativistic approximation. While the Schrödinger equation is
appropriate to describe most observations, there are instances where relativistic effects are important.
For example, close to the nucleus, electrons gain sufficient kinetic energy that their velocity becomes
comparable to the speed of light. Therefore, atoms contain relativistic effects from that region.
14.1 A brief review of theory of relativity

We are already familiar with the Klein-Gordon equation which we derived from the linear chain after
introducing a mass term. After applying the correspondence principle, the Klein-Gordon equation
resulted in the relativistic dispersion relation E(p) with E 2 = p 2 c 2 + m02 c 4 . All relativistic particles
obey the same dispersion relation. This dispersion relation lead Einstein to propose the famous
equivalence of energy and mass E = m0 c 2 , which is at the root of nuclear power, peaceful or deadly.
This dispersion relation is symmetric with respect to transformations that leave E 2 − p 2 c 2 in-
variant. Energy and momentum are related through the correspondence principle to derivatives of
space and time. This implies that c12 ∂t2 − ∇2 is an invariant of the symmetry groups of free space. It
might be interesting to find out what these symmetries are and if we can learn something from the
symmetry of the dispersion relation about the geometric structure of space and time.
The relativistic dispersion relation can be written with the help of the metric tensor1
 
−1 0 0 0
 0 1 0 0
 
gi ,j =  
 0 0 1 0
0 001
and the four-dimensional momentum p = (E/c, px , py , pz ) = ~i (− c1 ∂t , ∇) as
3
X
(pi c)gi−1 2 4
,j (pj c) + m0 c = 0
i ,j =0
We use here the notation that the momentum has a zeroth component, which is related to the
energy by p0 = E/c. The momentum is a four dimensional vector just as time and space are written
in the formulation of the theory of relativity as four dimensional vector (ct, x, y , z). The components
1 There are a number of different notations that are equivalent. Often the element −1 of the metric tensor is written
as − c12 and often the metric tensor has opposite sign. These are all equivalent notations. I have chosen one where the
metric tensor is dimensionless, and where the four-dimensional momentum vector has dimensions mass*length/time
229
230 14 RELATIVISTIC PARTICLES
of the four dimensional momentum vector are the canonical conjugate momenta of the corresponding
four-dimensional coordinate vector. For example the energy is conserved if the equations of motion
are translationally invariant in time. The momentum is conserved if they are translational invariant
in space.
Let us now consider coordinate transformations that leave c12 ∂t2 − ∇2 invariant. Let us write the
coordinate transformation as
4 4
X ∂x ′ X
xi′ = i
xj = Ri ,j xj
∂xj
j =0 j =0
What is the corresponding transformation of derivatives?

4 4
X ∂xj′ X
∂i ψ = (∂j′ ψ) = (Rj,i ∂j′ )ψ
∂xi
j =0 j =0
2 ∂
We have used here the vector ∂i = ∂xi .
In a vector notation we can write
x ′ = Rx
∂ = RT ∂ ′
The symmetry operations of free space are coordinate transformations R that leave ∂g −1 ∂ =
− c12 ∂t2
+ ∇2 invariant. These operations are called Lorentz transformations What are those trans-
formations?
∂g −1 ∂ = ∂ ′ Rg −1 RT ∂ ′ = ∂ ′ g −1 ∂ ′
Rg −1 RT = g −1
This is the defining equation for the symmetry operations.

√
We can show that all those transformations leave a certain distance d = c 2 t 2 − r 2 in the
four-dimensional space invariant.
x ′ gx ′ = xRT gRx = x(gR−1 g −1 )gRx = xgx

c t − r ′2 = c 2 t 2 − r 2
2 ′2
The invariance of this length is taken to define the basic symmetry of space. What is the distance
between two space-time points. Two points that are spatially distant can still have zero distance in
space time, namely when it takes the speed of light to get from (r1 , t1 ) to (r2 , t2 ).
Now we can investigate the symmetry operations in more detail: Let us consider a few examples.
Translations in time and space conserve d. Also Rotations conserve the length. The rotation in the
x,y-plane can be written as
! ! !
x′ cos(φz ) sin(φz ) x
=
y′ − sin(φz ) cos(φz ) y
When we replace x in the four-dimensional distance by x ′ , we obtain the same result. Similarly if
a physical path of a particle is x(t) then also x ′ (t ′ ) is a physical path, that obeys the equations of
motion.
2 Here a more detailed derivation:
∂Ψ(x1′ , x2′ , x3′ , x4′ ) X ∂Ψ(x ′ , x ′ , x ′ , x ′ ) ∂xj′ (x1 , x2 , x3 , x4 ) X ∂Ψ X ∂Ψ
1 2 3 4
= ′
= R =
′ j,i
RiT,j ′
∂xi j
∂x j ∂x i j
∂xj j
∂xj
14 RELATIVISTIC PARTICLES 231
There must be also a similar transformation in the ct, x plane that conserves the four-dimensional
distance
! ! !
ct ′ cosh( vc ) sinh( vc ) ct
=
x′ sinh( vc ) cosh( vc ) x
It can be easily shown using cosh2 (x) − sinh2 (x) = 1 that the distance c 2 t 2 − x 2 is conserved. What
is the physical meaning of this transformation? Let us investigate the nonrelativistic limit, that is
c → ∞. We obtain the Galileo transformation
x′ = x − vt
t′ = t
Invariance with respect to this transformation simply means that if a path physical x(t) obeys the
equations of motion, also the path x ′ (t ′ ) = x(t) + v t obeys the same equations of motion. This is of
course true for the free particle. The transformed path is the original x(t) as it is seen by an observer
moving with velocity −v . If the speed of light c is finite, as it is in reality, the transformation of the
equations of motion is more complex.
In order to build a relativistic theory, we need only to start combining so-called invariants of the
symmetry group to build up the Lagrangian, from which the equations of motion are derived. Once
we have such a theory, we might look for particles in the real world that are described by such a
theory. This is one way to come up with theories for new particles.
We have already seen that E 2 − p 2 c 2 is such an invariant and another is m02 c 4 . Adding the two
invariants, we arrive at the relativistic dispersion relation, from which we started. The correspondence
principle brings us back to the Klein-Gordon equation, which describes so-called scalar particles.
However there are other invariants, for example we might express the wave function as an array of
wave functions in each point. The wave function of such a particle is called a spinor wave function.
Now we can build up new invariants from operators acting on the spinor components, which still are
Lorentz invariant in space-time. One very important vector particle are photons, which are spin-one
particles. Another are electrons and other particles with spin 12 .
14.2 Relativistic Electrons

The Dirac equation can be written as
X
E− αi cpi − βm0 c 2 |Ψi = 0 (14.1)
i
where the energy is here an operator, namely E = i ~∂t and the momentum is the momentum operator
p = ~i ∇. The factors αi and βi are themselves operators acting not on the spatial variable but in a
four-dimensional spinor space. We will come to that later. For now we need to remember that αi
and β may not commutate among each other.
One condition for a relativistic wave equation is that it reproduces the relativistic dispersion
relation
P E = p c +2 m0 c . We get there by multiplying the Dirac equation from the left with
2 2 2 2 4
E + i αi cpi + βm0 c .
3
X 3
X
E2 − αi pi c 2 αj pj − m0 c 3 [αi β + βαi ]pi − m02 c 4 β 2 |Ψi = 0
i ,j =1 i =1
The relativistic dispersion operator is obtained when the conditions

1 1 1
[αi , αj ]+ = δi ,j [αi , β]+ = 0 [β, β]+ = 1
2 2 2
These equations cannot be fulfilled when α and β are simple numbers. A solution can be found by
assuming that the αi and βi are matrices in a four dimensional space. This statement does not refer
to the space-time, but to four components of the wave function |Ψi.
There is one solution, namely
! !
0 σi 1 0
αi = β=
σi 0 0 −1
where the σi ’s are the Pauli matrices and the 1’s are two-by-two unit matrices.
! ! ! !
01 0 −i 1 0 10
σx = σy = σz = 1=
10 i 0 0 −1 01
The four spinor components of the wave function |Ψi describe
• an electron with spin up
• an electron with spin down
• a positron with spin up
• a positron with spin down
Positrons are antiparticles of the electron. They have identical mass but opposite charge as the
electron. A positron and an electron can annihilate each other by creating two photons. The photons
carry the energy of the rest mass away.
14.3 Electron in the electromagnetic field

The electron(positron) is coupled to the electromagnetic field by replacing
E → E − qΦ
~
p~ → p~ − q A
~ is the vector potential. They are defined as E

where Φ(r ) is the electric potential and A ~ = −∇φ−∂
~ ~
tA
~ =∇
and B ~
~ × A.The charge of the electron is negative, that is q = −e where e is the elementary
charge, which is positive.
Thus the Dirac equation has the form
X
E − qΦ − αi c(pi − qAi ) − βm0 c 2 |Ψi = 0 (14.2)
i
If we split the four-component spinor into two two-component spinors (|φi, |χi), the Dirac equa-
tion has the form
P ! !
m0 c 2 + qΦ − E i σi (pi − qAi )c |φi
P 2 =0 (14.3)
i σi (pi − qAi )c −m0 c + qΦ − E |χi
where the first two component spinor |φi is attributed to electronic degrees of freedom for spin-up
and spin-down electrons and |χi describes spin-up and spin-down positron.
14.4 Down-folding the positron component

The four components of the Dirac spinor can be attributed to electrons with spin up and spin down
and positrons with spin up and spin down. The positron is the anti-particle of the electron. While a
Dirac particle is usually a superposition of electronic and positronic components, in the non-relativistic
limit, the two different components can be decoupled exactly.
In this case, we are interested in describing electrons. Therefore, we consider solution with
an energy close to E = +m0 c 2 . In this case, we can show that the positronic components are
fairly small. Therefore, one often names the electronic components |φi the “large components”,
while the positronic components |χi are named “small component”. We do not want to ignore the
positronic components, but we can simplify the equations by eliminating the positronic components
in an effective way.
In order to simplify the following equations let us introduce a quantity
ǫ − qΦ
M(ǫ, r ) = m0 +
2c 2
and the energy ǫ with the rest-energy removed
ǫ = E − m0 c 2
q
p2
Note that M is not the relativistic mass defined as mr = m0 1 + ( mp0 c )2 = m0 + 2m0 c 2 + O( c12 ) =
m0 + ǫ−V
c2 .
Thus we obtain
P ! !
qΦ − ǫ i σi (pi − qAi )c |φi
P =0
i σi (pi − qAi )c −2Mc 2 |χi
We can now use the second line of the equation to obtain |χi as function of the electron com-
ponents |φi,
1 X
|χi = σi (pi − qAi )|φi,
2Mc
i
which we then insert in to the first equation to eliminate the explicit dependence on the positrons.
X 1 X
σi (pi − qAi ) σj (pj − qAj ) + qΦ − ǫ |φi = 0
2M
i j
We can use the equation

~ σ B)
σ A)(~
(~ ~ =A
~B~ + i~ ~ × B)
σ (A ~ (14.4)
which connects the vectors A ~ and B
~ and the vector formed by the Pauli matrices. When we apply
this equation to operators one has to take care of the commutation relations.
hX 1 X i
(pi − qAi ) (pi − qAi ) + qΦ + i ~ × 1 (~
σi p~ − q A) ~
p − q A) − ǫ |φi = 0 (14.5)
2M 2M i
i i
Note that this equation is still nonlinear in energy and electromagnetic field, via the dependence on
the relativistic mass.
Furthermore the wave function φ is not normalized. The correct normalization is
~ 1 ~
1 = hφ|φi + hχ|χi = hφ|1 + (~
σ p~ − q A) σ p~ − q A)|φi
(~
(2Mc)2
In the last term the magic Eq. 14.4 has been used.
14.5 The non-relativistic limit: Pauli equation

Let us now form the non-relativistic limit by setting c = ∞ in Eq. 14.5. In that case the relativistic
"mass" M is equal to the rest mass m0 , which is constant in space. Thus we obtain the so-called
Pauli equation.
PAULI EQUATION
h (~ ~ 2
p − q A) q ~~ i
+ qΦ − S B − ǫ |φi = 0 (14.6)
2m0 m0
We have used here that
~ × (~
p − q A)
(~ ~ = −q(~
p − q A) ~+A
p×A ~ × p~) = i ~q(∇ ~+A
~ ×A ~ × ∇) ~
~ = i ~q B
~ =∇
B ~ ×A ~+A ~ × ∇ is the magnetic field. Here we have expressed the magnetic field as operator
equation, where ∇ not only acts on A ~ but also any function tow which this operator is applied to.
The magnetic field obtained this way is a simple vector function of real space.
The spin is written as
~ = ~~
S σ
2
where ~
σ = (σx , σy , σz ) are the Pauli matrices.
Equation 14.6 differs from the Schrödinger equation only by the Stern-Gerlach term in the
Hamiltonian3
q ~ˆ ~ ˆ
ĤSG = − S B(~r)
m
q ~ ~ Because the spin

which describes the energy of a magnetic moment m ~ = m S in a magnetic field B.
~q
~
of an electron has the eigenvalues ± 2 , the magnetic moment of the electron is µb = 2m 0
. This
fundamental constant is called the Bohr magneton. The magnetic moment in the Dirac equation is
the origin of interpreting the spin as the rotation of a particle about its own axis, which is a strange
concept for a point particle.
For a classical particle rotating about a center, magnetic moment and angular momentum are
q ~
related by the relation m~ = 2m L. The ratio between magnetic moment and angular momentum is
called the gyromagnetic moment. For a classical particle the gyromagnetic moment is therefore
q
γ = 2m . For other charge distributions the ratio between angular momentum and magnetic moment
differs, which is taken into account by the g-factor g so that in general m ~ Comparison
~ = g mq L.
with the Pauli equation indicates that the g-factor of the electron is ge = 2 and the gyromagnetic
q q
ratio of the electron is γe = ge 2m = m . The magnetic moment of the electron is ge = 2.0023. The
deviation from the value 2 is explained by quantum electrodynamics.
~ The vector
Consider a classical particle circling about the origin in a constant magnetic field B.
3 It is disturbing that the Pauli equation is a two-component equation, while the Schrödinger equation only has a
single component. In most cases, one can decouple the two components of the Pauli equation by choosing the magnetic
field in the z -direction. In that case there are no additional terms that couple spin-up and spin-down components. Thus
one arrives at a description that distinguishes between spin-up and spin-down electrons. Each fulfills a Schrödinger
equation, albeit with different potentials.
~ = 1B
potential for a constant magnetic field can be written as A ~ × ~r. The energy can be written as
2
1 ~ 2 = 1 (~ q~
E= p − q A)
(~ p− B × ~r)2
2m 2m 2
p2 q 2
~ × ~r) + q (B~ × ~r)2
= − p~(B
2m 2m 8m
p2 q ~ q2 ~
= − B(~r × p~) + (B × ~r)2
2m 2m 8m
~ = E0 − q L
E(B) ~B ~ + O(B 2 )
2m
Now consider the limit where the orbit of the particle goes to zero, so that we obtain a point particle,
and the angular momentum approaches the spin of the electron, that is |L| → ~/2. We obtain to
q~ ~
first order E(B) = E0 − 4m B, assuming that the magnetic moment is parallel to the magnetic field.
Thus we obtain half the value of that obtained from the Dirac equation. The factor missing is called
the g-factor, and is considered an anomaly. We can attribute it to the fact that the classical particle
in our model becomes itself relativistic as the radius is shrunk to zero.
If the magnetic field is parallel to the z-axis, the spin-up and spin-down components are indepen-
dent. Each component is a solution of a Schrödinger equation, with the only difference that spin-up
and spin-down components experience different potentials V↑ = qΦ − µB Bz and V↑ = qΦ + µB Bz .
If the electron is free to occupy the state with lowest energy, the spin will align parallel and not
antiparallel to the magnetic field.
Just as we have determined the Pauli equation for the electrons from the Dirac equation for
energies near E = +m0 c 2 after forming the non-relativistic limit, we can similarly determine a Pauli
equation for the positrons, that is the solutions with energies near E = −m0 c 2 . This is done explicitly
in appendix J on p. 281. There we also show that the motion of a positron corresponds indeed to a
particle with the opposite charge of the electron.
14.6 Relativistic corrections

Relativistic effects are quite large near the nucleus, where the kinetic energy of an electron approaches
infinity, and the velocity of the electrons approaches that of light.
Relativistic corrections can be observed macroscopically in every compass. A ferromagnet could
not produce permanent magnetization direction without spin-orbit coupling. Spin-orbit coupling
links the magnetization to the crystal geometry. In a compass without spin orbit coupling, the
magnetization direction inside the material would orient towards the north-pole, but the needle would
stay put.
In order to separate out the relativistic effects, let us introduce a new dimension-less variable
m 1 1 D
D= −1⇒ = +
M 2M 2m0 2m0
D vanishes in the nonrelativistic limit c → ∞, where M = m.
1 1 D
By replacing 2M by 2m 0
+ 2m 0
in the two-component Dirac equation Eq. 14.5, and after collecting
all terms proportional to D, we obtain the relativistic corrections W to the Pauli equation
1 n ~ ~ + i~

~ × (~ ~
o
W = p − q A)D(~
(~ p − q A) σ (~p − q A)D p − q A) (14.7)
2m0
The two component Dirac equation can then be written as (Ĥ0 + W − ǫ)|φi = 0, where Ĥ0 is the
Pauli Hamiltonian and W are the relativistic corrections. Note, that D contains not only a spatial
dependence, but also an energy dependence. If we only consider the lowest order of the energy
dependence, we can remove the explicit energy dependence.
Before we continue, let us investigate some properties of the function D
• As mentioned before, D vanishes in the non-relativistic limit.

• close to the relativistic limit that is for sufficiently slow particles, D has the form
ǫ−V 1 p2 1
D=− 2
+ O( 4 ) = − 2 2 + O( 4 ). (14.8)
2m0 c c 4m0 c c
where V = qΦ. The function D is negative and measures approximately the ratio of kinetic
energy and the rest-energy of the particle.
• Let us consider the spatial dependence of D for an electron near a nucleus: We use V (~r) =
qΦ = −Ze 2 /(4πǫ0 |r |) and ǫ = 0 as model for the potential in the center of an atom.
m0 1 2m0 c 2 −1
D= −1= ǫ−V
− 1 = − 1 +
M 1 + 2m 2 ǫ−V
0c
2
8πǫ0 m0 c |r | −1 |~r|
≈− 1− ≈ −1 + 2 + O(r 2 )
Ze 2 rT
where
Ze 2
rT =
4πǫ0 m0 c 2
is the so-called Thomson radius. The Thomson radius gives an indication about the size of the
region near the nucleus, where relativistic effects are important. The Thomson radius lies in
the range of 10−4 − 10−2 a0 , where a0 ≈ 0.529 × 10−10 m is the Bohr radius. As a comparison,
a typical atomic radius is in the range of 3 a0 .
Let us now return to the relativistic corrections of Eq. 14.7. To simplify the discussion we will
introduce a number of approximations. As the first simplification we ignore the dependence of the
relativistic corrections on the magnetic field. The idea is that usually both, relativistic corrections and
the magnetic field, can be considered as small perturbations, so that they can be treated independently
to first order.
We obtain
1 n o
W = p~D~ σ p~D × p~
p + i~ (14.9)
2m0
The last term in Eq. 14.9 will be identified with the spin-orbit coupling. The first term can be divided
into two terms that will be identified with the Darwin term and the mass-velocity correction. We
rewrite p~D~
p in two different ways and average the resulting expressions.
p = p~[D, p~] + p~2 D = −~
p~D~ p , D] + p~2 D = −[~
p[~ p , D], p~ + p~2 D
p , D]] − [~
p , [~
p~D~
p = [~
p , D]~
p + D~ p2
1 1 2
p = − ([~
p~D~ p , [~
p , D]] + (~p D + D~p2)
2 2
~2 2 1 2
= (∇ D) + (~ p D + D~p2)
2 2
where (∇2 D(~r)) is considered as a function and not an operator. 4 p 2 on the other hand is considered
in both cases as an operator. We can rewrite the relativistic corrections as
~2 ~ 2 1 i
W = (∇ D) + p 2 D + D~
(~ p2) + σ p~
~ ~D × p~
~ (14.10)
4m0 4m0 2m0
= WD + Wmv + WLS (14.11)
4 Unfortunately the notation of operator equations is not completely consistent. Often, an additional pair of paren-
theses is used to indicate that the operator, in this case ∇2 acts only on the function(s) inside the same pair of
parentheses and not further onto any function in the operator or standing to the right of the operator. The true
meaning must be obtained in the end from the context of the equation. It is always possible, even though in some
cases utterly inconvenient to rewrite an operator equation such that every operator acts on everything standing to the
right of it.
where the first term WD is the Darwin term, Wmv is the mass-velocity correction and WLS is the
spin-orbit coupling. We will investigate the three contributions separately.
14.6.1 Spin-Orbit coupling

The last term of the relativistic corrections in Eq. 14.11 is the so-called spin-orbit coupling
1
WLS = σ p~D × p~
i~ (14.12)
2m0
We will now introduce a some assumptions to allow a physical interpretation.
Let us assume that the electrostatic potential is spherically symmetric. As a result, also D is
spherically symmetric, so that we can write
~ ~r
p~D = [~
p , D]− + D~
p= (∂r D) + D~p
i |r |
where we consider (∂r D) as a single function. Note again the inconsistency in the notation, because
one might expect that the radial derivative would need to be applied to any function standing to the
right of D as well. Here and in the following we consider (∂r D) as a symbol for a function.
Thus we find
1 ~ (∂ D) 1 ~ (∂r D)
r
WLS = i~
σ ~r × p~ + D~
p × p~ = i~
σ ~r × p~
2m0 i |r | 2m0 i |r |
(∂r D) ~ ~
= LS
m0 |r |
where we used that ~~ ~ is the spin operator and ~r × p~ = L

σ /2 = S ~ is the angular momentum operator.
The energy splitting resulting from spin-orbit coupling in first order perturbation theory is
1 (∂r D) ~ ~ ~2 ℓ (∂r D)
∆ǫLS = hψ| LS|ψn i = ± hψ| |ψn i
m0 |r | 2m0 |r |
The spin-orbit coupling splits an atomic energy level with an angular momentum quantum number ℓ
into a upper level, where the spin is parallel to the orbital angular momentum, and a lower level with
antiparallel spin and orbital momentum. Because the total angular momentum for the parallel spin
level is j = ℓ + 12 the degeneracy of this upper orbital is 2j + 1 = 2ℓ + 2. The degeneracy of the lower
level is 2ℓ.
For historical reasons one often finds the expression for the spin-orbit coupling in a different
form. Here one employs first the assumption that the relativistic effects are small, that is D =
ǫ−V 1
− 2m 2 + O( c 4 ), then one introduces the potential of the hydrogen atom and the assumption that
0c
the energy of the electron is small, that is ǫ = 0. This yields
Ze 2 ~2 ℓ 1
∆ǫLS ≈ 2 2
hψ| 3 |ψi
16πǫ0 m0 c r
This is a good (or bad) example how one can make incompatible approximations and still get away:
Without corrections, D is not singular. However, we now used the approximation of small relativistic
effects, that is relatively slow particles in combination with infinite potential energy, that is fairly fast
particles, which lead us to a divergent expression for D near the nucleus. This is clearly nonsense!
What saves the construction, is that (1) the ℓ = 0 term vanishes, and (2) for other angular momenta
the wave functions vanish sufficiently rapid near the nucleus, that they compensate the divergent
term in the integral for the energy splitting.
The spin-orbit splitting is important for the conducting properties of semiconductors with heavier
elements such as GaAs. Here the top of the valence band, which has p character is split into 4 bands
with spin and orbital angular momentum parallel and two bands, where they are antiparallel. The
spin-orbit splitting of these levels in silicon is 0.044 eV and in the heavier Ge it is already 0.3 eV.
The spin-orbit term is also visible in core level spectroscopy, where the core levels are split due to
spin-orbit coupling into two multiplets that are separated by several eV.
14.6.2 Darwin term

In order to make the Darwin term
~2 ~ 2
WD = (∇ D) (14.13)
4m0
more transparent, let us now introduce again our usual approximations. First we keep only the leading
V −ǫ
order in c1 , so that D = 2m 0c
2 . Note that this approximation is not valid near the nucleus. I list it
here, because that is what has been done historically. We obtain

~ 2 Φ)
~2 e(∇ 1
WD = + O( 4 )
8m02 c 2 c
~ 2Φ = ρ
We can simplify this expression further using the Poisson equation ∇ ǫ0 .
Ze 2
If we specialize our result further to the Coulomb potential of the nucleus V = qΦ = − 4πǫ0 |r |
and
ǫ = 0, we obtain
~2 Z 2 e 2
WD ≈ δ(r )
8ǫ0 m02 c 2
Thus the Darwin term acts only on s-wave function and shifts those up in energy.
14.6.3 Mass-velocity term

In order to give a physical interpretation to the mass-velocity
1
Wmv = p 2 D + D~
(~ p2) (14.14)
4m0
from Eq. 14.11, we use
p~2
ǫ − V |ψi ≈ |ψi
2m0
p~2
⇒ hψ|D|ψi ≈ hψ| − |ψi
4m02 c 2
p~4
⇒ hψ|WMV |ψi ≈ hψ| − |ψi
8m03 c 2
where we use the wave functions of the Pauli equation, implying that we only consider the leading
order of the correction in c1 . The classical version of the mass velocity term is the leading term of
the relativistic energy momentum relation.
p2 p4
q
1
ǫ(p) = m02 c 4 + p 2 c 2 − m0 c 2 = − + O( 4 )
2m0 8m03 c 2 c
√
where we used 1 + x = 1 + 12 x − 81 x 2 + O(x 3 )
Chapter 15
Many Particles
Sofar, quantum theory was actually a theory of classical fields. Just as we describe light by electro-
magnetic fields, we introduced fields to describe other particles such as electrons. The new ingredients
are the interpretation of the field amplitude as probability amplitude for particles to occur, and that
we consequently normalized this probability to one.
15.1 Bosons
Let us return to the continuous linear chain that we investigated in one of the first sections. We
showed there, that we could form a quantum theory by constructing wave packets that behave
like classical particles. However, since the linear chain itself was build up from particles, we should
treat every bead of the chain itself as a quantum particle. This idea is the central idea of second
quantization and the treatment of many particles.
15.1.1 Quantum mechanics as classical wave theory

Let us consider the Lagrangian of the hanging linear chain, which can be also used to describe a
relativistic particle. There are several reasons for choosing this model: we can look at it both as a
classical model and a quantum mechanical problem. (1) If we remove the “mass” term that results
from hanging the linear chain, we have a model for a phonon in a crystal. (2) If we ignore the
fact that the displacements are real, the displacement field follows the same equations as the Klein-
Gordon field, which describes relativistic spin-0 particles, such as mesons. Furthermore we avoid the
intricacies of complex field amplitudes in the derivations.
We start out from the Lagrangian Eq. 3.31, which has been derived early in this booklet.
Z L
1 m2 c 2
L=K dx 2 (∂t φ)2 − (∂x φ)2 − 02 φ(x, t)2
0 c ~
The first step is to decompose φ(x, t) into eigenmodes.1 First we derive the Euler-Lagrange
equations, which are the classical field equations such as the Schrödinger equation. Here we obtain:
1 m2 c 2
− 2 ∂t2 + ∂x2 − 02 φ(x, t) = 0
c ~
We obtain the time independent equation using the ansatz φ(x, t) = uj (x)e−i ωj t
ω2 m02 c 2
j
+ ∂x2 − uj (x) = 0
c2 ~2
1 We could also have staid in the real space formulation, but that would have made the derivation a bit more
complex. Here I rather choose to transform first into Eigenmodes and later transform back into real space.
239
240 15 MANY PARTICLES
The solutions are

1 2π
uj (x) = √ sin(kj x) with kj = j
L L
q
~ωj = m02 c 4 + (~kj c)2
where we use periodic boundary conditions φ(x) = φ(x + LL) and the ortho-normalization condition
RL
huj |uj ′ i = 0 dxuj (x)uj ′ = δj,j ′ .
There is one tricky point: We only keep the solution ω > 0 even though the equation has a
positive and a negative branch. This is analogous to the harmonic oscillator where the classical
motion is built from solutions x(t) = Ae+i ω0 t + Be−i ω0 t . Nevertheless, we only consider positive
frequency for the harmonic oscillator.
P
Now we need to introduce the mode decomposition φ(x, t) = j uj (x)φj (t) into the Lagrangian,
which we rewrite with the help of Gauss theorem and the Euler-Lagrange equations obtained above.
Z L 1
2 2 m02 c 2
L=K dx (∂ t φ) + φ(∂ x − )φ
0 c2 ~2
X 1 Z L
′
=K 2
(∂ φ
t j )(∂ φ
t j ) dxuj (x)uj′ (x)
c 0
n,n′
Z L
m2 c 2
+φj φ′j dxuj (x)(∂x2 − 02 )uj′ (x)
0 ~
K X
= 2 (∂t φj )2 − ωj2 φ2j
c
j
In the next step we need to find the Hamilton function for the classical displacement field. We
obtain the the canonical momentum of the displacement field amplitude φj (t),
∂L 2K c2
Πj = = 2 ∂t φj ⇒ ∂t φj = Πj
∂ φ̇j c 2K
and set up the Hamilton function in the usual way

X
H(Πj , φj ) = Πj ∂t φj − L(φj ∂t φj )
j
c 2 X 2 4K 2 ωj2 2
= Πj + φj = E
4K c4
j
15.1.2 Second quantization

Now we quantize the displacement field φ(x). The procedure is completely analogous to the first
quantization, where we made the transition from a position coordinate x to a wave function φ(x) on
the coordinate space. Here we make the transition from a displacement field or probability amplitude
to a wave function Φ defined on the space spanned by the probability amplitudes. This implies that
the wave function amplitude φ(x, t) does not have a well defined value but is itself described by a
wave function Φ[φ(x)]. Note that Φ is a functional of the displacement field. For the linear chain,
this is a natural step, because it corresponds to the first quantization of the beads. If however we
interpret φ(x, t) as the probability amplitude of a quantized point particle, things become different.
We use the correspondence principle
~
E → i ~∂t Πj → ∂φ (15.1)
i j
15 MANY PARTICLES 241
to convert the Hamilton function into a Hamilton operator.

We obtain a Schrödinger equation
c 2 X 2 4K 2 ωj2 2
i ~∂t |Φi = H|Φi = −~ (∂φj )2 + φj |Φi
4K c4
j
where Φ is a wave function that depends on the displacements of all the eigenmodes Φ(φ1 , φ2 , . . .).
Note here, that the difference between the Hamilton operator used here and the Hamilton operator
~2 2
H(p, x) = − 2m ∂x +V (x), that defines the “classical field equations” such as the Schrödinger equation.
Let us try to give some meaning to this Schrödinger equation. What we find is a harmonic
oscillator for each vibrational mode. We know how to deal with this from Sec. 5.6. By comparing
with Eq. 5.30, we identify the mass of the harmonic oscillator with m = 2K c 2 and the frequency with
ω0 = ωj . We introduce the ladder operators a and â† analogous to Eq. 5.33.
s r
c2 Kωj
âj = Π̂j − i φ̂j
4~Kωj ~c 2
s r
† c2 Kωj
âj = Π̂j + i φ̂j
4~Kωj ~c 2
Note that Π̂j = ~i ∂φj is an operator and no more a classical field with a well defined value at each
position.
With the help of ǫj = ~ωj , the Hamilton operator can be written as
X 1
Ĥ = |uj iǫj âj† âj + huj |
2
j
What we get are discrete energy levels with values

X 1
E(nj ) = ǫj ( + nj ).
2
j
The energy is discretized and comes in bunches, where each quantum carries the energy correspond-
ing to the dispersion relation ǫ(kj ) = ~kj . Thus we can interpret each excitations of a vibrational
mode as particle with a wave function with the corresponding eigenmode of the hanging linear chain.
We obtained the particle properties, namely discretized energy quanta from a pure field theory!
All states can be built up from a vacuum state |Oi, which corresponds to no particles at all.
Note that the vacuum state is not the zero state! A state with n1 particles in the first vibrational
mode, n2 particles in the second and so on, can be constructed as
Q † n
j (aj )
j
|n1 , n2 , . . .i = Q |Oi
j nj !
We have actually used here two equivalent notations for a given state, the number representation
on the left side and the one using creation operators. We have extended Hilbert space used for
one-particle states, and formed the so-called Fock space, which contains the Hilbert space for one
particle, the one for two particles and so on.
Now we can understand the double-slit experiment for photons. For the quantized Maxwell’s
equation the double-slit experiment has the following interpretation: In order to find out if electro-
magnetic radiation has been absorbed, we need to measure the intensity of the electromagnetic
P field
before and after absorption. The intensity is related to the particle number operator j aj† aj . The
resulting eigenvalues, however only change by integer amounts. If the intensity absorbed is very large,
the changes of the intensity relative to the total intensity are so small that they become irrelevant,
so that we obtain the classical interpretation of the intensity. Editorial remark: Where are the
anti-particles?
15.1.3 Coordinate representation

Editorial remark: This is pretty strong!
We can also transform this into the coordinate representation using
X
φ(x, t) = uj (x)φj (t)
j
X
Π(x, t) = uj (x)Πj (t)
j
Let us consider the commutator relations among the two operator fields. We use
~
[Πj , φj ′ ] = δj,j ′ [φj , φj ′ ] = 0 [Πj , Πj ′ ] = 0
i
which is a direct consequence of the correspondence principle.
We obtain
X
[Π(x), φ(x ′ )] = uj (x)uj ′ (x ′ )[Πj , φj ′ ]
j,j ′
X
= uj (x)uj (x ′ ) = δ(x − x ′ )
j
and therefore
~
[Π(x), φ(x ′ )] = δ(x − x ′ ) [φ(x), φ(x ′ )] = 0 [Π(x), Π(x ′ )] = 0
i
The creation and annihilation operators can be transformed between the mode representation and
the real space transformation.
X
a† (x) = uj∗ (x)aj†
j
X
a(x) = uj (x)aj
j
The inverse transformation is

Z
aj† = dxuj (x)a† (x)
Z
aj = dxuj∗ (x)a(x)
Hence the commutator relations of the creation and annihilation operators are
[a(x), a† (x ′ )] = δ(x − x ′ ) [a† (x), a† (x ′ )] = 0 [a(x), a(x ′ )] = 0
Let us now consider the Hamilton operator and transform it to real space
X 1
H= ǫj (aj† aj + )
2
j
X ǫj Z Z Z Z
= ( dxuj (x)a (x))( dx uj (x )a(x )) + ( dxuj (x)a(x))( dx ′ uj (x ′ )a† (x ′ ))
† ′ ∗ ′ ′ ∗
2
j
Z Z X Z Z X
= dx dx ′ a† (x) uj (x)ǫj uj∗ (x ′ ) a(x) + dx dx ′ a(x) uj∗ (x)ǫj uj (x ′ ) a† (x)
j j
Z h q q
1
= dx a† (x) m02 c 4 + c 2 ∂x2 a(x) + a(x) m02 c 4 + c 2 ∂x2 a† (x)
2
15 MANY PARTICLES 243
Here we used
Z Z X
dx dx ′ f (x)uj∗ (x)ǫj uj (x ′ )g(x ′ ) = hf |uj ihuj |H|uj ihuj |gi
j
X
= hf | |uj ihuj | H |uj ihuj | |gi
j
= hf |H|gi
Here a space derivative appears under a square root, which is dealt with in the usual way, namely
as power series expansion of the square root.
Note, that we deal now with wave functions on an infinite dimensional space. Thus it is practically
impossible to write down such a wave function. If we resort to the abstract operator algebra the
expressions are quite intuitive.
Now some serious business: let us consider free space, then we are left with the zero point motion
of the particle coordinate, which carries an infinite energy
Z
dk 1
E= ǫ(k)
2π 2
Since ǫ(k) grows with increasing k the vacuum has infinite energy. Should we worry? Yes! We know
from the theory of relativity that energy is related to mass, and mass results in gravitational forces.
They should be observable.
This turned out to be one of the major puzzles in physics, namely that the two main theories,
general theory of relativity, which describes gravitational forces, and quantum theory seem to be
incompatible.
15.2 Fermions
Fermions are constructed similarly. The main difference is that instead of introducing the ladder
operators of the harmonic oscillator for each vibrational mode we introduce those of the two state
system. Again the Fermionic waves come in bunches, but only one Fermion can be created in one
vibrational mode of the linear chain.
Unfortunately, I do not know of a good classical analogue for Fermions and people say there is
none. Therefore, you need to rely on your imagination.
Fermions could solve the problem of infinite zero point energy of the bosons. As Fermions have
negative zero-point energy, the two energies would cancel if they would exist in equal numbers. This
is the basic idea of super-symmetry, which assumes that for every Boson, there is also a Fermion, but
now we have ventured very far from the material of an elementary course of quantum mechanics.
15.3 Final remarks

We only scratched the surface. We have seen that if we use the correspondence principle for the
wavefunction amplitudes we obtain discretized excitations, which we interpret as particles. The same
principle, which made waves out of point particle, made particles out of continuous wave functions.
Thus we can now describe all the puzzling observations in the double-slit experiment both for electrons
and for photons. The price is, however, a theory of some complexity: In the end we deal with wave
functions on infinite dimensional spaces!
For further reading on quantum field theory, I recommend the book by Haken [15]. It should be
noted that quantum field theory, the description of many particle states, comes in many different
appearances, such as Green’s functions, Feynman diagrams, creation and annihilation operators etc.,
which all have their own virtues and disadvantages.
Part I
Appendix
245
Appendix A
Galilei invariance
Here we show, that the dynamics of the envelope function χ(x, t) of a wave packet is Galilei invariant,
respectively that it is invariant under the transformation from a resting to a moving reference frame.
For the derivation it is important to distinguish between partial and total derivatives. In order to
make this explicit we consider
def
χv (x, t) = χ0 (x − v t, t) = χ0 (y (x, t), τ (x, t)) (A.1)
with the transformed variables
def def
y (x, t) = x ′ = x − v t and τ (x, t) = t ′ = t
where χ0 (x, t) is the resting wave packet, and χv (x, t) is the moving wave packet.
Now, we derive an equation for the envelope function χv (x, t) from the Schrödinger equation for
the moving wave packet
φv (x, t) = χv (x, t)ei (kx−ωt) (A.2)
The following expressions will be used:
i ~∂t ei (kx−ωt) χv (x, t) = ei (kx−ωt) [~ω + i ~∂t ]χv (x, t)

~ ~
∂x ei (kx−ωt) χv (x, t) = ei (kx−ωt) [~k + ∂x ]χv (x, t)
i i
2 2
~ i (kx−ωt) i (kx−ωt) ~
∂x e χv (x, t) = e ~k + ∂x χv (x, t)
i i
From the Schrödinger equation for the moving wave packet we derive a differential equation for
its envelope function χv (x, t) alone.

~2
0= i ~∂t + ∂x φv (x, t) (A.3)
2m
" 2 #
Eq. A.2 ~2 i (kx−ωt) i (kx−ωt) 1 ~
⇒ 0 = i ~∂t + ∂x e χv (x, t) = e ~ω + i ~∂t − ~k + ∂x χv (A.4)
(x, t)
2m 2m i
" 2 #
1 ~
⇒ 0 = ~ω + i ~∂t − ~k + ∂x χv (x, t) (A.5)
2m i
247
248 A GALILEI INVARIANCE
Now we can rewrite

Eq. A.1
i ~∂t χv (x, t) = i ~∂t χ0 (y (x, t), τ (x, t))
∂y ∂τ
= i ~∂y χ0 (y (x, t), τ (x, t)) +i ~∂τ χ0 (y (x, t), τ (x, t))
∂t ∂t
|{z} |{z}
−v 1
=−v i ~∂y χ0 (y , τ ) + i ~∂τ χ0 (y , τ )
~ Eq. A.1 ~
∂x χv (x, t) = ∂x χ0 (y (x, t), τ (x, t))
i i
~ ∂y ~ ∂τ
= ∂y χ0 (y (x, t), τ (x, t)) + ∂τ χ0 (y (x, t), τ (x, t))
i ∂x i ∂x
|{z} |{z}
1 0
~
= ∂y χ0 (y , τ )
i
With this we can turn the equation for the envelope function χv into an equation for the envelope
function χ0 (x, t) for the resting system.
" 2 #
Eq. A.5 1 ~
0 = ~ω + i ~∂t − ~k + ∂x χ (y (x, t), τ (x, t))
2m i |0 {z }
χv (x,t)
" 2 #
1 ~
= ~ω − v i ~∂y + i ~∂τ − ~k + ∂y χ0 (y , τ )
2m i

(~k)2 ~k ~ −~2 2
= ~ω − v i ~∂y + i ~∂τ − − ∂y − ∂y χ0 (y , τ )
2m m i 2m
2
2

~ 2 (~k) ~k ~
= i ~∂τ + ∂ + ~ω − + v− ∂y χ0 (y , τ )
2m y 2m m i
~2 k 2 ∂E
If ~ω = E(k) = 2m and v = ∂(~k) , we obtain the Schrödinger equation itself

~2 2
i ~∂τ + ∂ χ0 (y , τ )
2m y
Thus the transformation into a moving reference frame involves a simple transformation of the
envelope function
x′ = x − vt
|{z} and t′ = t
|{z} and χ′ (x ′ , t ′ ) = χ(x, t)
| {z } | {z }
y τ χ0 (y ,τ) χ0 (x,t)
Thus a wave packet at rest χ0 (x, t) is transformed into a moving one by firstly replacing the
spatial argument by x − v t and secondly by multiplication with a plane wave ei (kx−ω(k)t) , where
v = ∂ω/∂k.
Eq. A.2
ψ ′ (x, t) = χv (x, t)ei (kx−ω(k)t) = χ0 (x − v t, t)ei (kx−ω(k)t)
Eq. A.2
= = ψ(x − v t, t)ei (kx−ω(k)t)
In order to transform the wave function itself, we need to add a phase factor. The reason is the
following: The envelope function only describes the spatial probability distribution. The information
about the velocities is not contained in the envelope function. This information must be included in
the transformation of the wave function. The phase factor tells that the velocity distribution of the
wave packet has to be shifted by the relative velocity v .
Appendix B
Spreading wave packet of a free

particle
Let us explore a one-dimensional free particle with a wave packet that initially has Gaussian shape.
The Schrödinger equation has the form
−~2 2
i ~∂t ψ(x, t) = ∂ ψ(x, t) (B.1)
2m x
We use an Ansatz for the wave function that has the same form as the initial wave packet, but
with time dependent coefficients
1 2
ψ(x, t) = A(t)e− 2 B(t)x (B.2)
The parameters A, B and their time dependence are obtained by insertion into the time-dependent
Schrödinger equation on the one side and the normalization condition on the other side. 1
Let us first evaluate the derivatives that appear in the Schrödinger equation Eq. B.1.
h i h∂ A 1 ih i
Eq. B.2 1 2 t 1 2
i ~∂t ψ(x, t) i ~∂t A(t)e− 2 B(t)x = i ~
= − (∂t B)x 2 A(t)e− 2 B(t)x (B.3)
A 2
−~2 2 Eq. B.2 −~
2 h 1 2
i −~2 h 1 2
i
∂x ψ(x, t) = ∂x2 A(t)e− 2 B(t)x = A(t)∂x −B(t)xe− 2 B(t)x
2m 2m 2m
−~2 h i 1 2
= −B(t) + B (t)x A(t)e− 2 B(t)x
2 2
(B.4)
2m
From the requirement that the left-hand side, Eq. B.3, and the right-hand side, Eq. B.4, of the
Schrödinger equation Eq. B.1, are identical everywhere in space, we obtain the two conditions for
A(t) and B(t)
∂ A
t −~2
i~ = −B(t) (B.5)
A 2m
1 −~2 2
i ~ − ∂t B = B (t) (B.6)
2 2m
This is a system of ordinary differential equations that can be solved as follows: We begin with
1 The factor 1
2
has been included only because the formulas for Gaussians with this choice turn out to be more
pleasant.
249
250 B SPREADING WAVE PACKET OF A FREE PARTICLE
the second equation for B(t)

Eq. B.6 −i ~ 2
∂t B(t) = B (t) (B.7)
m
Z t Z t
−i ~
dt B −2 (t)∂t B(t) = dt (B.8)
0 0 m
1 1 −i ~
−
+ = t (B.9)
B(t) B(0) m
1 ∂t B(t) Eq. B.7 i~
⇒ ∂t =− 2 =
B(t) B (t) m
Z t
1 i~ i~
⇒ = C0 + dt ′ = C0 + t
B(t) 0 m | {z m }
def
=C(t)
Eq. B.11 1
⇒ B(t) = (B.10)
C(t)
where we introduced the integration constant C0 and the variable
def i ~t
C(t) = C0 + (B.11)
m
to simplify the following equations
Next we insert this result, Eq. B.10, into the first equation, Eq. B.5, to obtain A(t)
Eq. B.5 −i ~
A−1 ∂t A(t) = ∂t ln[A(t)] = B(t)
2m
Z Z
−i ~ t ′ Eq. B.10 −i ~ t ′ 1
⇒ ln[A(t)] = ln[A(0)] + dt B(t ′ ) = ln[A(0)] + dt
2m 0 2m C(t ′ )
Z t Z t
Eq. B.11 1 dC(t ′ ) 1 1 d
= ln[A(0)] − dt ′ ′ ′
= ln[A(0)] − dt ′ ln[C(t ′ )]
2 0 dt C(t ) 2 0 dt ′
1 1 Eq. B.13 1
= ln[A(0)] − ln[C(t)] + ln[C0 ] = D − ln[C(t)]
2 2 2
1
⇒ A(t) = p eD (B.12)
C(t)
where
def 1
D = ln[A(0)] + ln[C0 ] (B.13)
2
Thus the wave function has the form
Eqs. B.2,B.10,B.12 1 1 x2
ψ(x, t) = eD p e− 2 C(t) (B.14)
C(t)
R
Finally we fix the factor eD using2 the normalization condition dx ψ ∗ ψ.
Z ∞ ∗ Z ∞
eD+D

−1 1
+ 1 x2
1= dx ψ ∗ (x, t)ψ(x, t) = dx e 2 C(t)∗ C(t)
−∞ |C(t)| −∞
Z s
∞
e2Re[D] − 12 2Re[C(t)] x 2
2Re[D] 1 |C(t)|2 √
= dx e |C(t)| 2
=e 2π
|C(t)| −∞ |C(t)| 2Re[C(t)]
r r
2Re[D] π Eq. B.11 2Re[D] π
=e = e
Re[C(t)] C0
r
Re[D] 4 C0
⇒ e = (B.15)
π
R∞ 1 2 √
2 We also use the Gauss integral −∞ dx e− 2 x = 2π
B SPREADING WAVE PACKET OF A FREE PARTICLE 251
The imaginary part of D is irrelevant because that only changes the overall phase of the wave function.
We obtain the wave function
s
Eqs. B.14,B.15 4 C0 1 x2
ψ(x, t) = 2
e− 2 C(t) (B.16)
π|C(t)|
with C(t) given by Eq. B.11.
B.1 Probability distribution

Let us now evaluate the probability distribution P (r ) = ψ ∗ (x, t)ψ(x, t).
s
C0

∗ Eq. B.16 −1 1
+ 1 x2
P (r, t) = ψ (x, t)ψ(x, t) = 2
e 2 C(t) C∗ (t)
π|C(t)|
s s
C0 − Re[C(t)] x 2 1 2C0 −1
2C0
x2
= 2
e |C(t)| 2
=√ 2
e 2 |C(t)|2
π|C(t)| 2π |C(t)|
2
1

−1 x
= √ e 2 d (t)
(B.17)
2π · d(t)
with
s r s 2 s 2
def |C(t)|2 Eq. B.11 C0 ~t ~t
d(t) = = 1+ = d(0) 1 + (B.18)
2C0 2 mC0 2md(0)2
We see that the density maintains its Gaussian shape, but the width d(t) of the Gaussian increases
after a certain while almost linearly with time.
The spreading of a Gaussian probability distribution reminds of the behavior of a diffusing particle.
Such a particle undergoes Brownian motion. The related probability distribution behaves similarly
to that of a quantum particle, but the width of the distribution grows only like the square root of
time.
def p 2
We define the width of the wave packet as ∆ = hx i − hxi2 . Now we investigate how the
width changes with time.
Let us consider the first few moments3 of the probability distribution.
Z ∞
Eq. B.17
h1i = dx P (x, t) = 1
−∞
Z ∞
Eq. B.17
hxi = dx P (x, t)x = 0
−∞
Z ∞ Z ∞ 2
1

Eq. B.17 −1 x
hx 2 i = dx P (x, t)x 2 = √ dx e 2 d (t) x 2
−∞ 2π · d(t) −∞
√ 2 Z ∞
√
u=x/( 2d ) 2d(t) 2 2d(t)2 π
= √ du e−u u 2 = √ = d(t)2 (B.19)
π −∞ π 2
This reveals the physical meaning of the variable d(t).

3 The nth moment of a real function f (x) is defined as
Z
def
hx n i = dx f (x)x n
252 B SPREADING WAVE PACKET OF A FREE PARTICLE
We combine Eq. B.19 to obtain the width

rD E p
def Eq. B.19
∆(t) = (x − hxi)2 = hx 2 i − hxi2 = d(t)
s 2
Eq. B.18 ~t ~t
= d(0) 1 + 2
→ for t → ∞
2md (0) 2md 2 (0)
Thus the width of the probability distribution grows linearly with time.
Appendix C
The one-dimensional rectangular

barrier
V
I II I
III
V0
x
L
The potential of a one-dimensional rectangular barrier has the following form
( ) ( )
V0 0<x <L
V (x) = for
0 x < 0; x > L
We call the region to the left of the barrier region I, the barrier itself region II and what is to the
right of the barrier is region III.
The wavenumber of the particle in region I and III is
1√
kI = 2mE (C.1)
~
and the wavenumber in region II is
1p
kII = 2m(E − V ) (C.2)
~
At first we consider the case where the energy of the particle is larger than the barrier. Later we will
use this result to obtain the results for particles with an energy that is lower than the barrier.
253
254 C THE ONE-DIMENSIONAL RECTANGULAR BARRIER
C.1 E > V , the barrier can be classically surmounted

We start with the Ansatz
ψI (x) = AI ei kI x + BI e−i kI x
ψIII (x) = AIII ei kI x + BIII e−i kI x
We select one particular partial solution, namely the one corresponding to a particle beam approaching
the barrier from the left, meaning that no particles arrive from the right-hand side or
BIII = 0 (C.3)
A second coefficient can be fixed by the choosing a common prefactor. We choose the prefactor so
that
AI = 1 (C.4)
The remaining four coefficients are determined by the requirement that the wave function is
continuous and differentiable, i.e.
ψI (0) = ψII (0) (C.5)

∂x ψI (0) = ∂x ψII (0) (C.6)
ψII (L) = ψIII (L) (C.7)
∂x ψII (L) = ∂x ψIII (L) (C.8)
These requirements translate into
Eqs. C.5,C.4
1 + BI = AII + BII
Eqs. C.6,C.4
i kI − i kI BI = i kII AII − i kII BII
i kII L −i kII L Eqs. C.7,C.3
AII e + BII e = AIII ei kI L
Eqs. C.8,C.3
i kII AII ei kII L − i kII BII e−i kII L = i kI AIII ei kI L
We simplify the equations
1 + BI = AII + BII (C.9)

kI
(1 − BI ) = AII − BII (C.10)
kII
AII e−i (kI −kII )L + BII e−i (kI +kII )L = AIII (C.11)
kI
AII e−i (kI −kII )L − BII e−i (kI +kII )L = AIII (C.12)
kII
By adding and subtracting the last two equations Eq. C.11 and Eq. C.12 we can express AII and BII
by AIII :

Eqs. C.11,C.12 1 i (kI −kII )L kI
AII = e 1+ AIII (C.13)
2 kII

Eqs. C.11,C.12 1 i (kI +kII )L kI
BII = e 1− AIII (C.14)
2 kII
C THE ONE-DIMENSIONAL RECTANGULAR BARRIER 255
Inserting this result, Eq. C.13 and Eq. C.14, into the first equation Eq. C.9 we obtain
Eq. C.9
1 + BI = (AII + BII )

Eqs. C.13,C.14 1 i (kI −kII )L kI 1 kI
= e 1+ AIII + ei (kI +kII )L 1 − AIII
2 kII 2 kII
i (kI −kII )L
e + ei (kI +kII )L kI ei (kI −kII )L − ei (kI +kII )L
= + AIII
2 kII 2

kI
= ei kI L cos(kII L) − i sin(kII L) AIII (C.15)
kII
The second equation, Eq. C.10, together with the expressions for AII and BII , Eq. C.13 and Eq. C.14,
yields
kII
1 − BI = (AII − BII )
kI

kII 1 i (kI −kII )L kI 1 kI
= e 1+ AIII − ei (kI +kII )L 1 − AIII
kI 2 kII 2 kII

1 k II 1 k II
= ei kI L e−i kII L 1 + + ei kII L 1 − AIII
2 kI 2 kI

kII
= ei kI L cos(kII L) − i sin(kII L) AIII (C.16)
kI
Finally we add Eqs. C.15,C.16 to obtain AIII :

i kI L kI i kI L kII
2=e cos(kII L) − i sin(kII L) AIII + e cos(kII L) − i sin(kII L) AIII
kII kI

kI kII
= ei kI L cos(kII L) − i sin(kII L) + cos(kII L) − i sin(kII L) AIII
kII kI

kI kII
= ei kI L 2 cos(kII L) − i + sin(kII L) AIII
kII kI
−1
−i kI L i kI2 + kII2
⇒ AIII = e cos(kII L) − sin(kII L) (C.17)
2 kI kII
Knowing AIII , we can determine the other parameters from Eqs. C.15,C.13, and Eq. C.14.
The transmission coefficient T of the barrier can be evaluated as

AIII 2
T =
AI

−1
Eq. C.17 i kI2 + kII2 i kI2 + kII2
= cos(kII L) − sin(kII L) cos(kII L) + sin(kII L)
2 kI kII 2 kI kII
 −1
2 2 2

k + kII
= cos2 (kII L) + I sin2 (kII L)
 
| {z } 2kI kII
1−sin2 (kII L)
" 2 ! #−1
kI2 + kII2 2
= 1+ − 1 sin (kII L) (C.18)
2kI kII
Now we express the wave vectors kI and kII by the energy 1
" 2 ! #−1
Eq. C.18 kI2 + kII2
T = 1+ − 1 sin2 (kII L)
2kI kII
−1
V2 Lp
= 1+ sin2 ( 2m(E − V )) (C.19)
4E(E − V ) ~
In this derivation we assumed that the energy of the particle is greater than the barrier, so that
the motion over the barrier is classically allowed and the wave function is oscillatory also within the
barrier.
C.2 Tunneling effect E < V

Now we consider the case when the energy is lower than the barrier.
In that case the wave vector within the barrier becomes imaginary, which says, that the wave
function is exponentially decaying or exponentially increasing. There is nothing special about a
imaginary wavevector, but for convenience we introduce a new quantity, which is real: The decay
factor λ.
1p
kII = 2m(E − V ) =: i λ
~
We just proceed with the derivation given for E > V and make sure that no operation depends
on the fact that the wave vector kII is real.
The only complication is when we introduce the cosinus and the sinus.
ei kII L + ei kII L e−λL + e+λL

cos(kII L) = = = cosh(λL)
2 2
ei kII L − ei kII L e−λL − e+λL
sin(kII L) = = = i sinh(λL)
2i 2i
Thus we have to do this replacement in the final expression. We obtain for the transmission coeffi-
cient2
" 2 2 ! #−1
kI − λ2
T = 1− − 1 sinh2 (λL)
2i kI λ
−1
V2 2 1
p
= 1+ sinh ( 2m(V − E)L) (C.20)
4E(V − E) ~
1 We 1
√ p 1
use the following, where kI = ~
2mE and kII = 2m(E − V ):
~
!2 !2
kI2 + kII2 2E − V 4E 2 − 4EV + V 2
−1 = p −1 = −1
2kI kII 2 E(E − V ) 4E(E − V )
4E 2 − 4EV + V 2 − 4E 2 + 4EV V2
= =
4E(E − V ) 4E(E − V )
2
!2 !2
kI2 − λ2 E − (V − E) 4E 2 − 4EV + V 2
−1 = p −1= −1
2ikI λ 2i E(V − E) −4E(V − E)
4E 2 − 4EV + V 2 + 4EV − 4E 2 V2
= =−
−4E(V − E) 4E(V − E)
C THE ONE-DIMENSIONAL RECTANGULAR BARRIER 257
If the barrier is very wide or the barrier is much higher than the energy of the particle, we can
simplify the expression to3
4E(V − E) − 1 √2m(V −E)L

T ≈ e ~
V2
3 Inthis limit, the term with the sinus hyperbolicus will always be much larger than one. so that the latter can
be neglected. Secondly for large arguments the sinus hyperbolicus will be dominated by one of the two exponential
functions, so that we can ignore the other. These arguments sound hand waiving, but root firmly on the ground of a
Taylor expansion.
Appendix D
Particle scattering at a
one-dimensional square well
V
I II I
III
0
|V0|
V0 x
D.1 Square well E > 0

The barrier can be converted into a well discussed in Appendix by changing the sign of the barrier
height.
Thus, for particles that can travel classically outside the well, we can directly use the transmission
coefficients of the barrier Eq. C.19.
D.2 Square well E < 0

For the energies for which a classical particle is confined to the well, we use the technique of the
logarithmic derivatives from Section 4.6.4 on p. 73.
The wave function consists of the three parts: Inside the well we have travelling waves, but outside
we have functions that exponentially decay away from the well.
ψI (x) = AI eλx
ψIII (x) = BIII e−λ(x−L)
We have to solve two subsequent matching problems. We start from the left and match the
259
260 D PARTICLE SCATTERING AT A ONE-DIMENSIONAL SQUARE WELL
solutions ψII to ψI . In this case, we assign the following values to the expressions φi (x) in Eq. 4.43.
ψ1 (x) = ei kII x D1 = i kII

ψ2 (x) = e−i kII x D2 = −i kII
ψ3 (x) = eλx D3 = λ
Insertion into Eq. 4.43 yields

−i kII − λ
i kII x −i kII x i kII − λ
ψII (x) = AI e +e
−i kII − i kII i kII + i kII

i kII x 1 λ −i kII x 1 λ
= AI e −i +e +i
2 2kII 2 2kII

λ
= AI cos(kII x) + sin(kII x)
kII
Now we come to the second matching problem at x = L. The solution, we just obtained for
ψII , plays now the role of ψ3 , to which we match an exponentially decaying and an exponentially
increasing function in region III.
ψ1 (x) = eλx D1 = λ
ψ2 (x) = e−λx D2 = −λ

λ −kII sin(kII L) + λ cos(kII L)
ψ3 (x) = AI cos(kII x) + sin(kII x) D3 =
kII cos(kII L) + kλII sin(kII L)
kII
1− λ tan(kII L)
= λ λ
1+ kII tan(kII L)
We insert the result again into Eq. 4.43 in order to obtain the solution in region III.

λ −λ − D3 λ − D3
ψIII (x) = AI cos(kII L) + sin(kII L) eλ(x−L) + e−λ(x−L)
kII −λ − λ λ+λ

λ 1 D3 1 D3
= AI cos(kII L) + sin(kII L) eλ(x−L) 1+ + e−λ(x−L) 1−
kII 2 λ 2 λ
The boundary condition require that the prefactor of the exponentially increasing term vanishes,
which provides us with a condition for D3
D3 = −λ
kII
1− λ tan(kII L)
λ λ
= −λ
1+ kII tan(kII L)
kII λ
1− tan(kII L) = −1 − tan(kII L)
λ kII

kII λ
2= − tan(kII L)
λ kII
2kII λ
= tan(kII L)
kII2 − λ2
Now we introduce the expressions for kII and λ
p
2 −E(E − V ) 1p
= tan( 2m(E − V )) (D.1)
2E − V ~
This system of equations cannot be solved analytically, but the structure of the solutions can be
visualized graphically as shown in Fig. D.2.
D PARTICLE SCATTERING AT A ONE-DIMENSIONAL SQUARE WELL 261
−5 −4 −3 −2 −1 0
E
Fig. D.1: Graphical representation of Eq. D.1 and the resulting eigenstates. The left-hand side of
Eq. D.1 is the full line, wile the tan-function on the right-hand side is the dashed lines. The crossing
points, denoted by the red dots represent the energies of the bound states. The Insets show the
corresponding wave functions. The color coding indicates the extent of the well.
262 D PARTICLE SCATTERING AT A ONE-DIMENSIONAL SQUARE WELL
Appendix E
Alternative proof of Heisenberg’s

uncertainty principle
Now let us work out the proof of Heisenberg’s uncertainty principle, which is not difficult to follow,
but requires a quite clever step. I will show here the proof for the more general uncertainty relation
Eq. 7.7, which is not more complicated than the one for the more restricted form Eq. 7.7.
def def
1. Firstly, we show that the pair ∆Â = Â − hÂi and ∆B̂ = B̂ − hB̂i obeys the same commutation
relations as the operators Â and B̂ themselves
PROOF:
[∆Â, ∆B̂] = [Â − hÂi, B̂ − hB̂i]

= (ÂB̂ − ÂhB̂i − hÂiB̂ + hÂihB̂i) − (B̂ Â − B̂hÂi − hB̂iÂ + hÂihB̂i)
= [Â, B̂]
q.e.d
2. Now we show that, for anyq pair of hermitian operators

Â and B̂,
the product of expectation
1
values obeys the relation hψ|Â |ψihψ|Q̂ |ψi ≥ 2 hψ|[Â, B̂]|ψi.
2 2
263
264 E ALTERNATIVE PROOF OF HEISENBERG’S UNCERTAINTY PRINCIPLE
PROOF: We start by defining the operator Ŝ(λ) = Â + iλB̂

def
F (λ) = hψ|Ŝ † Ŝ|ψi = hŜψ|Ŝψi ≥ 0 (E.1)
Since Ŝ † = Â − iλB̂
F (λ) = hψ|(Â − iλB̂)(Â + iλB̂)|ψi

= hψ|(Â2 + iλ[Â, B̂]− + λ2 B̂ 2 )|ψi
Eq. E.1
= hψ|Â2 |ψi + λhψ|i[Â, B̂]|ψi + λ2 hψ|B̂ 2 |ψi ≥ 0
F (λ) is a quadratic function in λ. F (λ) is real because i times the commutator of two hermitian
operators is hermitian. This is shown as follows:
† † †
Â =Â;B̂ † =B̂
i[Â, B̂] = −i ÂB̂ − B̂ Â = −i B̂ † Â† − Â† B̂ † = −i B̂ Â − ÂB̂ = i[Â, B̂]
Because the expectation value of a hermitian operator is real, also the middle coefficient in F (λ) is
real.
The parabola is always upwards open, which can be seen by determining the sign of the prefactor of
λ2 . We obtain
B̂=B̂ † d 2F
hψ|B̂ 2 |ψi = hψ|B̂ 2 |ψi = hB̂ψ|B̂ψi ≥ 0 ⇒ ≥0
dλ2
The requirement that the function is always non-negative tells us that the function has at most one
real zero. (A parabola has always two complex zeros.)
F( λ)= A2 + λ i[A,B] + λ2 B 2
λ λ λ
The zeros of F (λ) are
s 2
hψ|i[Â, B̂]|ψi hψ|i[Â, B̂]|ψi hB̂ 2 i
λ± = − ± −
2hÂ2 i 2hÂ2 i hÂ2 i
The expression under the square-root on the right-hand side must be zero or negative, because otherwise
there would be two real zeros of the parabola, which would violate the requirement that the parabola
is non-negative for all values of λ.
2
hψ|i[Â, B̂]|ψi hψ|B̂ 2 |ψi
− ≤0
2hψ|Â2 |ψi hψ|Â2 |ψi
q q
1

⇒ hψ|B̂ 2 |ψi hψ|Â2 |ψi ≥ hψ|i[Â, B̂]|ψi
2
3. Similarly we can insert ∆Â for Â and ∆B̂ for B̂ and obtain Heisenberg’s uncertainty principle
Eq. 7.7.
q q
1
(E.2)

hψ|(∆Â)2 |ψi hψ|(∆B̂)2 |ψi ≥ hψ|i [Â, B̂]|ψi
2
4. The original formulation of Heisenberg’s uncertainty principle is obtained as a special case of

E ALTERNATIVE PROOF OF HEISENBERG’S UNCERTAINTY PRINCIPLE 265
Eq. E.2. For a generalized coordinate Q̂ and its canonical conjugate momentum P̂ we obtain
q q
1 ~

hψ|(∆Q̂)2 |ψi hψ|(∆P̂ )2 |ψi ≥ hψ|i [P̂ , Q̂]|ψi = (E.3)
2 2
because of their specific commutator relation. q.e.d.
266 E ALTERNATIVE PROOF OF HEISENBERG’S UNCERTAINTY PRINCIPLE
Appendix F
Spherical harmonics
F.1 Spherical harmonics addition theorem
ℓ
~r1 ~r2 4π X ∗
Pℓ ( )= Yℓ,m (~r1 )Yℓ,m (~r2 )
|~r1 | |~r2 | 2ℓ + 1
m=−ℓ
where Pℓ (x) are the Legendre polynomial as defined in Eq. 11.26 on p. 182.
Proof: Rotate the coordinate system so that ~r1 points in z-direction.
ℓ
r2 ~ez 4π X ∗
Pℓ ( ) = Yℓ,m (~
ez )Yℓ,m (~r2 )
|~r2 | 2ℓ + 1
m=−ℓ
4π
= ]Y ∗ (~
ez )Yℓ,0 (~r2 )
2ℓ + 1 ℓ,0
Eq. 11.23 4π 2ℓ + 1 0
= P (1)Pℓ0 (cos(θ))
2ℓ + 1 4π ℓ
Eq. 11.25
= Pℓ (1)Pℓ (cos(θ))
Eq. 11.26 1 ℓ
= ℓ
∂u (u 2 − 1)ℓ u=1 Pℓ (cos(θ))
2 ℓ!
To complete the proof, we need to show that Pℓ (1) = 1, which is not done yet....
F.2 Unsöld’s Theorem

[31]
ℓ
4π X ∗
Yℓ,m (~r)Yℓ,m (~r) = 1
2ℓ + 1
m=−ℓ
F.3 Condon-Shortley phase

The Condon-Shortley phase (−1)m is often introduced in quantum mechanical calculations. It can
be introduced either the in the definition of the associated Legendre polynomials or in the expression
of the spherical harmonics expressed by associated Legendre polynomials.
267
268 F SPHERICAL HARMONICS
Appendix G
Random Phase approximation
Caution: The notation is not yet consistent with that in the section 13.4.1.
In section 13.4.1 on p.212, we extracted the transition probabilities from one initial state to a
specified final state. The argument can however be driven much further, leading to kinetic equations
for the probabilities of the kind
X
∂t Pm (t) = Γm,n Pn (t) − Γn,m Pm (t) (G.1)
n
with transition probabilities Γm,n . In section 13.4.1 we made a number of simplifying assumptions.
Here we we will either circumvent them or make them explicit.
The wave function |ψ(t)i is represented as a superposition of the time dependent solutions of
the unperturbed Schrödinger equation
Eq. 13.7
X i
|ψ(t)i = |φn ie− ~ En t an (t)
n
where |φn i are eigenstates of the unperturbed Hamiltonian Ĥ0 with eigenvalues En .
In Eq. 13.11, the coefficients am (t) are given up to second order in the perturbation hatW (t) as
Z
Eq. 13.11 i X t ′ i ′
am (t) = am (0) − dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t an (0)
~ n 0
Z Z ′
1 X t ′ t i ′ i ′′
− dt dt ′′ hφm |Ŵ (t ′ )|φk ihφk |Ŵ (t ′′ )|φn ie− ~ (Ek −Em )t e− ~ (En −Ek )t an (0)
~2 0 0
n,k
+ O(Ŵ 3 ) (G.2)
P
The matrix elements of the density matrix ρ̂(t) = |ψ(t)ihψ(t)| = m,m′
∗
|φm iam (t)am ′ (t)hφm |
269
270 G RANDOM PHASE APPROXIMATION
are
Z
∗ ∗ i X t ′ i ′
am (t)am ′ (t) = am (0)am ′ (0) − dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t an∗ (0)am′ (0)
~ n 0
Z
i X t ′ i ′ ∗
+ dt hφm′ |Ŵ (t ′ )|φn i∗ e+ ~ (En −Em′ )t am (0)an (0)
~ n 0
Z Z t
1 X t ′ i ′ i ′′
+ 2 dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t dt ′′ hφm′ |Ŵ (t ′′ )|φn′ i∗ e+ ~ (En′ −Em′ )t an (0)an∗′ (0)
~
n,k 0 0
Z t Z t′
1 X i ′ i ′′
− 2 dt ′ dt ′′ hφm |Ŵ (t ′ )|φk ihφk |Ŵ (t ′′ )|φn i · e− ~ (Ek −Em )t e− ~ (En −Ek )t an (0)am ∗
′ (0)
~ 0 0
n,k
Z Z ′
1 X t ′ t i ′ i ′′
− 2 dt dt ′′ hφm′ |Ŵ (t ′ )|φk i∗ hφk |Ŵ (t ′′ )|φn′ i∗ · e+ ~ (Ek −Em′ )t e+ ~ (En′ −Ek )t am (0)an∗′ (0)
~ ′ 0 0
n ,k
+ O(Ŵ 3 ) (G.3)
Now let us average over the initial phases, i.e. we multiply each initial amplitude an (0) by a factor
ei φn and then average over the phases independently.
Z Z
1 1
dφn dφm an∗ (0)e−i φn am (0)ei φm = 0 for n 6= m
2π 2π
Z
1
dφn an∗ (0)e−i φn an (0)ei φn = |an (0)|2 = Pn (0)
2π
so that

∗
am (0)am′ (0) = δm,m′ Pm (0)
initial phases
G RANDOM PHASE APPROXIMATION 271
We obtain

∗
am (t)am′ (t)
initial phases
Z
i t ′ i ′
= Pm (0)δm,m′ − dt hφm |Ŵ (t ′ )|φm′ ie− ~ (Em′ −Em )t Pm′ (0)
~ 0
Z
i t ′ i ′
+ dt hφm′ |Ŵ (t ′ )|φm i∗ e+ ~ (Em −Em′ )t Pm (0)
~ 0
Z Z t
1 X t ′ i ′ i ′′
+ 2 dt hφm |Ŵ (t ′ )|φn ie− ~ (En −Em )t dt ′′ hφm′ |Ŵ (t ′′ )|φn i∗ e+ ~ (En −Em′ )t Pn (0)
~ n 0 0
Z t Z t′
1 X i ′ i ′′
− 2 dt ′ dt ′′ hφm |Ŵ (t ′ )|φk ihφk |Ŵ (t ′′ )|φm′ i · e− ~ (Ek −Em )t e− ~ (Em′ −Ek )t Pm′ (0)δm,m′
~ 0 0
k
Z t Z t′
1 X i ′ i ′′
− 2 dt ′ dt ′′ hφm′ |Ŵ (t ′ )|φk i∗ hφk |Ŵ (t ′′ )|φm i∗ e+ ~ (Ek −Em′ )t e+ ~ (Em′ −Ek )t Pm (0)δm,m′
~ 0 0
k
+ O(Ŵ 3 )
Z
i t ′ ′ − ~i (Em′ −Em )t ′
= Pm (0)δm,m′ − dt hφm |Ŵ (t )|φm′ ie Pm′ (0) − Pm (0)
~ 0
Z Z
1 X t ′ t ′′ i ′ i ′′
+ 2 dt dt hφm |Ŵ (t ′ )|φn ihφn |Ŵ (t ′′ )|φm′ ie− ~ (En −Em )t e+ ~ (En −Em′ )t Pn (0)
~ n 0 0
Z t Z t
1 X
− 2 dt ′ dt ′′ hφm |Ŵ (t ′ )|φk ihφk |Ŵ (t ′′ )|φm i
~ 0 0
k
i ′ i ′′
·θ(t ′ − t ′′ )e− ~ (Ek −Em )t e− ~ (Em −Ek )t Pm (0)δm,m′
Z Z
1 X t ′ t ′′
− 2 dt dt hφm |Ŵ (t ′′ )|φk ihφk |Ŵ (t ′ )|φm i
~ 0 0
k
i ′ i ′′
·θ(t − t ′′ )e− ~ (Em −Ek )t e− ~ (Ek −Em )t Pm (0)δm,m′
′
+ O(Ŵ 3 ) (G.4)
In the last step we introduced the Heaviside step function θ(t). The last two terms can be combined,
after the two time variables have been interchanged in the last term. Then the two step functions
simply add up to one.

∗
am (t)am ′ (t)
initial phases
Z
i t ′ ′ − ~i (Em′ −Em )t ′
= Pm (0)δ − m,m′ dt hφm |Ŵ (t )|φm ie ′ Pm (0) − Pm (0)
′
~ 0
Z Z
1 X t ′ t ′′ i ′ i ′′
+ 2 dt dt hφm |Ŵ (t ′ )|φn ihφn |Ŵ (t ′′ )|φm′ ie− ~ (En −Em )t e+ ~ (En −Em′ )t Pn (0)
~ n 0 0
Z t Z t
1 X i ′ i ′′
− 2 dt ′ dt ′′ hφm |Ŵ (t ′ )|φk ihφk |Ŵ (t ′′ )|φm ie− ~ (Ek −Em )t e− ~ (Em −Ek )t Pm (0)δm,m′
~ 0 0
k
+ O(Ŵ 3 ) (G.5)
G.1 Repeated random phase approximation

The density matrix is not diagonal. Therefore we make an additional approximation, namely that we
simply drop the off-diagonal terms. We can also say that we perform an additional phase average
at time t. This may be justified, if coupling to an environement destroys the phase-relation be-
tween different eigenstates of the unperturbed Hamiltonian. Such a phase relationship is also called
entanglement or coherence.
The probability at time t is obtained as

∗
Pm (t) = am (t)am′ (t)
initial and final phases
RANDOM PHASE APPROXIMATION
Eq. G.5
X
Pm (t) = Pm (0) + Λm,n (t)Pn (t) − Λn,m (t)Pm (0) (G.6)
n
Z t Z t
1 ′ i ′ ′′
Λm,n (t) = dt dt ′′ hφm |Ŵ (t ′ )|φn ihφn |Ŵ (t ′′ )|φm ie− ~ (En −Em )(t −t ) + O(Ŵ 3 )
~2 0 0
Z t 2
1 ′ ′

− ~i (En −Em )t ′ 3
=
~ dt hφm |Ŵ (t )|φn ie + O(Ŵ ) (G.7)
0
The rate equation Eq. G.6 can be understood as follows: The population of state m increases by
particles that are excited from the n-th state into the state |φm i. On the other hand the population
of this state is decreased by transitions from the state |φm i into any other state.
G.2 Transition matrix element

The matrix element can be brought into a more convenient form. This is basically a repetition of
section 13.4.1.
Analogously to section 13.4.1, we consider a perturbation Ŵ (t) that has the form of a pulse with
a given frequency ω.
Eq. 13.12
Ŵ (t) = Ŵ0 ei ωt χ(t) (G.8)
where χ(t) is a smooth envelope function for the light pulse, which obeys the normalization
Z ∞
dt |χ(t)|2 = 1
−∞
Without loss of generality, we furthermore require that the pulse, i.e. χ(t), is not switched on before
t = 0 and that it completely died out after time t.
Using the Fourier transform
Z ∞
Eq. 13.15 1
χ̃(ω) = dt χ(t)e−i ωt
2π −∞
we obtain with t → ∞
2
2π 1
Λm,n = hφm |Ŵ0 |φn i χ̃ (En − Em − ~ω)
~ ~
which is identical to the result Eq. 13.16 obtained earlier and which turns, for a monochomatic pulse,
into
Eq. 13.17 2π 2
Λm,n = hφm |Ŵ0 |φn i δ(En − Em − ~ω)
~
G RANDOM PHASE APPROXIMATION 273
G.3 Rate equation

A rate equation can be obtained by comparing the transition matrix element with the result of the
kinetic equation Eq. G.1. Integrating the latter, leads us to an expression for the probabilities at time
t
X
Pm (t) = eΓ̄Γ t Pn (0)
n m,n
P
where Γ̄m,n = Γm,n − k Γk,m δm,n .
By comparing this expression with Eq. G.6, we obtain the identity
def
X
eΓ̄Γ t = Λ̄m,n = Λm,n (t) − Λk,m (t)δm,n
k

1
Γ = lim ln Λ̄(t)
Γ̄
t→∞ t
G.4 Off-diagonal elements

The off diagonal elements, which we have dropped in the random phase approximation above, are
not zero. They describe the coherence, or entanglement of the occupations of the states.
Z
∗ m6=m′ i t ′ ′ − ~i (Em′ −Em )t ′
am (t)am′ (t) = − dt hφm |Ŵ (t )|φm ie ′ Pm (0) − Pm (0)
′
~ 0
Z
X i t ′ ′ − ~i (En −Em )t ′
+ − dt hφm |Ŵ (t )|φn ie
n
~ 0
Z t
i ′′ ′′ + ~i (En −Em′ )t ′′
· dt hφn |Ŵ (t )|φm ie ′ Pn (0) 1 − δm,m + O(Ŵ 3 )
′
~ 0
X
m6=m′
= Ωm,m′ Pm′ (0) − Pm (0) + Ωm,n Pn Ω∗n,m + O(Ŵ 3 )
n
Z t
i i ′
Ωm,m′ = − dt ′ hφm |Ŵ (t ′ )|φm′ ie− ~ (Em′ −Em )t (G.9)
~ 0
The off-diagonal elements reflect that the system remains, after the pulse in a superposition of several
excited states. One says that the individual states are entangled.
Appendix H
Propagator for time dependent

Hamiltonian
We define the propagator such that it transforms the initial wave function into the time dependent
wave function
|Ψ(t)i = Û(t, t0 )|Ψ(t0 )i (H.1)
We look for a compact functional form for the propagator. We start from the Schrödinger
equation, where we express the state by the propagator and the initial state.
i ~∂t |ψ(t)i = Ĥ(t)|ψ(t)i
⇒ i ~∂t Û(t, t0 )|ψ(t0 )i = Ĥ(t)Û(t, t0 )|ψ(t0 )i
Since this equation holds for arbitrary initial states |ψ(t0 )i, we obtain a differential equation for the
propagator
i ~∂t Û(t, t0 ) = Ĥ(t)Û(t, t0 )
with the initial condition Û(t0 , t0 ) = 1̂.
Integration converts the differential equation into an integral equation
Z
i t ′
Û(t, t0 ) = = 1 − dt Ĥ(t ′ )Û(t ′ , t0 )
~ t0
This integral equation can be solved by successive iteration, su that each approximation Û (n) (t, t0 )
contains terms up to the n-the power of the Hamiltonian.
Z
i t ′
Û (n+1) (t, t0 ) = 1̂ − dt Ĥ(t ′ )Û (n) (t ′ , t0 )
~ t0
Û (0) (t, t0 ) = 1̂
Z t
(1) i
Û (t, t0 ) = 1̂ − dt1 Ĥ(t1 )
~ t0
Z t Z
(2) i i t1 ′ ′
Û (t, t0 ) = 1̂ −
dt1 Ĥ(t1 ) 1 − dt1 Ĥ(t1 )
t0 ~ ~ t0
Z Z Z t2
i t i 2 t
= 1̂ − dt1 Ĥ(t1 ) + − dt2 Ĥ(t2 ) dt1 Ĥ(t1 )
~ t0 ~ t0 t0
n Z
i j t
X Z tj Z t3 Z t2
(n)
Û (t, t0 ) = − dtj dtj −1 . . . dt2 dt1 Ĥ(tj )Ĥ(tj −1 ) · · · Ĥ(t2 )Ĥ(t1 )
~ t0 t0 t0 t0
j =0
275
276 H PROPAGATOR FOR TIME DEPENDENT HAMILTONIAN
If the series of Û (n) converges, that is Û (n+1) = Û (n) in the limit n → ∞, we obtain
∞ Z
i j t
X Z tj Z t2
Û(t, t0 ) = − dtj dtj −1 . . . dt1 Ĥ(tj )Ĥ(tj −1 ) . . . Ĥ(t2 )Ĥ(t1 )
~ t0 t0 t0
j =0
We introduce the time-ordering operator T

(
Ĥ(t1 )Ĥ(t2 ) für t1 > t2
T Ĥ(t1 )Ĥ(t2 ) =
Ĥ(t2 )Ĥ(t1 ) für t2 > t1
The time-ordering operator does not belong to the kind of operators we are used to: “Normal”
operators act on states. This operator rearranges a product of time-dependent operators. This is
why I use a caligraphic letter instead of putting a hat ontop of the symbol.
The Hamiltonians in the expression for the propagator are arranged such that the time ascends
from the right to the left, i.e. tj ≥ tj −1 ≥ . . . ≥ t2 ≥ t1 . Now let us allow the integration to disregard
this requirement, that is each integration runs from t0 to t. Instead of the correct order of Hamilton
operator all j! (j-factorial) permutation occur equally. Using the time ordering operator we bring the
Hamiltonians in the right order and the factor j1! corrects for the multiple inclusion of the products.
∞ Z
i j t
Z t Z t
X 1
Û(t, t0 ) = − dtj dtj −1 . . . dt1 T Ĥ(tj )Ĥ(tj −1 ) . . . Ĥ(t2 )Ĥ(t1 )
j! ~ t0 t0 t0
j =0
 
∞ Z j
X 1 i t ′
=T  − dt Ĥ(t ′ ) 
j! ~ t0
j =0
h i Rt i
− Ĥ(t ′ )
= T e ~ t0
Thus we obtain
PROPAGATOR
h i Rt i
Ĥ(t ′ )
Û(t, t0 ) = T e− ~ t0 (H.2)
Appendix I
Matching of WKB solution at the

classical turning point
The linearized equation

 
 ~2 
 2 dV 
− ∂ x + x  ψ(x) = 0
 2m dx x=0 

| {z }
V (x)−E
has the form of the so-called Airy’s equation
∂z2 f (z) = zf (z) (I.1)
Apart from the factors Airy’s equation describes the Schrödinger equation for a particle in a constant
field. Thus it corresponds to the free fall.
In order to arrive at Airy’s equation, we perform a variable transform1
s !
3
2m dV
z= − 2 x
~ dx x=0
The solutions f (z) of Airy’s equation, Eq. I.1, are the Airy’s functions Ai (z) and Bi (z).
Z
1 ∞ 1
Ai (z) = ds cos( s 3 + sz
π 0 3
Z
1 ∞ 1 3 1
Bi (z) = ds e− 3 s +sz + sin( s 3 + sz
π 0 3
1 We define the new variable z by z = cx with a constant c. We insert x = z /c and
d dz d
∂x = = = c∂z
dx dx dz
into the Schrödinger equation at the classical turning point Eq. I.1

~2 2 2 dV 1 Eq. I.1
− c ∂z + z ψ(x) = 0
2m dx x=0 c

Thus the constants drop out if

s
~2 2 dV 1 3 2m dV
− c =+ ⇒ c= −
2m dx x=0 c ~2 dx x=0
277
278 I MATCHING OF WKB SOLUTION AT THE CLASSICAL TURNING POINT
Their asymptotic forms far from the origin, the classical turning point, are
1 2 3
Ai (z >> 0) = 1 e− 3 z 2
2πz 4
1 2 3
Bi (z >> 0) = 1 e3z 2
πz 4
1
2 3 π
Ai (z << 0) = sin( (−z) 2 + )
1
π(−z) 3 44
1 2 3 π
Bi (z << 0) = 1 cos( (−z) 2 + )
π(−z) 4 3 4
The wave function in the neighborhood of the classical turning point can be described by the
Airy’s functions
ψ(x) = aAi (x/c) + bBi (x/c)
For ~ → 0, that is for z → ∞ this wave function must become identical with the WKB approximation
of the wave function for a linear potential.
1 h i Rx i
Rx i
ψ(x) = p Ae ~ 0 d x p(x) + Be− ~ 0 d x p(x)
|p(x)|
s
p V (x)−E= ddVx x dV p
p(x) = 2m(E − V (x)) = 2m x = ~c −z(x)
dx x=0

where c = z/x.
Thus we obtain for z > 0

1 i 1
Rz ′ ′
1 i 1
Rz ′ ′
1
ψ(x) = p √ Ae ~ c 0 d z ~i cz 2 + Be− ~ c 0 d z i ~cz 2
~c −z
1
(−1)− 4 = 1+i
√
(1 + i )π 1 3 2 3
2 − 23 z 2 z2
= √ 1 Ae + Be 3
2~c πz 4
(1 + i )π
→ √ [2A · Ai (z) + B · Bi (z)]
2~c
and for z < 0

1 i −1
R (−z ) 1
d (−z ′ ) ~c(−z ′ ) 2 − ~i −1
R (−z ) 1
d (−z ′ ) ~c(−z ′ ) 2
ψ(x) = p √ Ae ~ c 0 + Be c 0
~c −z

π 1 −i 32 (−z ) 23
3
+i 23 (−z ) 2
= √ 1 Ae + Be
~c π(−z) 4
3 3
π 1

2 π 2 π
i π4 −i 3 (−z ) 2 + 4 i π4 +i 3 (−z ) 2 + 4
= √ 1 Ae e + Be e
~c π(−z) 4
π 2
π 1 i π4 i π4 2 3 π i4 i π4 3 π
= √ 1 Ae + Be cos (−z) + 2 + i Ae + Be sin (−z) +
2
~c π(−z) 4 3 4 3 4
π h π π
π π
i
→ √ Aei 4 + Bei 4 · Ai (z) + i Aei 4 + Bei 4 · Bi (z)
~c
The coefficients of the Airy functions must be identical left and right to the turning point. This
provides us with a relation for the coefficients A, B of the solution left of the barrier with those to
the right of the barrier.
I MATCHING OF WKB SOLUTION AT THE CLASSICAL TURNING POINT 279
Matching condition for the WKB solution at the classical turning point. The WKB wave function
for a potential with positive slope at the classical turning point can be written as
Z
1 1 xc l ′ π
Ψ(x < xcl ) = p cos( [ dx p(x ′ )] − )
|p(x)| ~ x 4
1 1
R ′ ′
Ψ(x > xcl ) = p e− ~ d x p(x )
|p(x)|
and for negative slope

Z
1 1 x ′ π
Ψ(x > xcl ) = p cos( [ dx p(x ′ )] − )
|p(x)| ~ xc l 4
1 1
R ′ ′
Ψ(x < xcl ) = p e ~ d x p(x )
|p(x)|
280 I MATCHING OF WKB SOLUTION AT THE CLASSICAL TURNING POINT
Appendix J
Positrons
Here we want to show how that the last two components of the Dirac spinor describe spin-up and
spin-down positrons. We do this by deriving the nonrelativistic limit for E ≈ −m0 c 2 . We begin with
Eq. 14.3 from p. 232.
P ! !
m0 c 2 + qΦ − E i σi (p i − qAi )c |φi
P 2 =0
i σi (pi − qAi )c −m0 c + qΦ − E |χi
Let us we introduce a small parameter ǫ = E + m0 c 2 and let us introduce the relativistic mass
−ǫ + qΦ
M(ǫ, ~r) = m0 +
2c 2
P ! !
2m c 2 + qΦ − ǫ σi (pi − qAi )c |φi
P 0 i
=0
i σi (pi − qAi )c −ǫ + qΦ |χi
so that we obtain
P ! !
2Mc 2 i σi (pi − qAi )c |φi
P =0
i σi (pi − qAi )c −ǫ + qΦ |χi
Now we use the first equation to express the electronic components by the positronic ones.
1 X
|φi = − σi (pi − qAi )c|χi
2Mc 2
i
and insert it into the second equation
X
0= σi (pi − qAi )c|φi − ǫ + qΦ|χi
i
" #
X 1 X
= − σi (pi − qAi ) σi (pi − qAi ) + qΦ − ǫ |χi
2M
i i
" #
X 1 X
=− σi (pi − qAi ) σi (pi − qAi ) − qΦ + ǫ |χi
2M
i i
" #
X 1 X 1

=− (pi − qAi ) (pi − qAi ) + i σi (~ ~ ×
p − q A) ~
p − q A)
(~ − qΦ + ǫ |χi
2M 2M i
i i
281
282 J POSITRONS
Now we are ready to form the non-relativistic limit. As c approaches infinity, the electronic
component |φi goes to zero. The positronic components fulfil
" #
(~ ~ 2
p − q A) q ~~
0= − B S − qΦ + ǫ |χi
2m0 m
Let us compare this result with the Pauli equation for the electrons
" #
(~ ~ 2
p − q A) q ~~
0= − B S + qΦ − ǫ |φi
2m0 m
The equation for the positrons differs from that for the electrons by the sign reversal of the energy
and the potential energy. In order to extract the implications for the dynamics, we use this equation
to extract a classical Hamilton function. From that we can derive the classical equations of motion
using Hamilton’s equations.
One open problem is that we still have a two component spinor wave function. We want to avoid
to deal with a classical spinor particle, even though such a generalization could be made. In order
to obtain two one-component wave functions, we restrict the equation to magnetic fields oriented
along the z-axis. In this case, the positronic Pauli equation separates exactly into one equation for
spin-up and one for spin-down positrons. nevertheless we keep the vector notation for S ~ B.
~
" #
(~ ~ 2
p − q A) q ~~
p , ~r, t) = −
Hp (~ − B S − qΦ
2m0 m
" #
∂H (~ ~
p − q A) q
ṗi = − = ~ − (∂i B)
(−q∂i A) ~ S~ − q∂i Φ
∂ri m0 m
∂H pi − qAi
r˙i = =−
∂pi m0
m0 rï = −ṗi + q~r˙ ∇A
~ i + q∂t Ai
" #
p − q A)
(~ ~ q
=− (−q∂i A) − (∂i B)S − q∂i Φ + q~r˙ ∇A
~ ~ ~ ~ i + q∂t Ai
m0 m
~ + q (∂i B)
= −q~r˙ ∂i A ~ S~ + q∂i Φ + q~r˙ ∇A ~ i + q∂t Ai
m
X q ~ S~
= −q [−∂i Φ − ∂t Ai ] −q [r˙j ∂i Aj − r˙j ∂j Ai ] − (∂i B)
| {z } m
j
Ei | {z }
~ i
v ×B)
(~
h i
m0~r¨ = −q E~ +~ ~ − q (∂i B)
v ×B ~ S~
m
The first contribution is the Lorentz force acting on a particle with charge −q. The second part is
q ~ ~
the force related to the energy of a particle with magnetic moment m ~p = m S in a magnetic field B.
The energy of a magnetic moment m ~ is E = −m
~ in a magnetic field B ~ The magnetic moment of a
~ B.
q ~ ~ is the angular momentum.
classical charge distribution is m
~ = g 2m L, where g is the g-factor, and L
Remember that the charge q of the electron is negative q = −e so that the charge of the positron
is −q = +e, where e is the elementary charge.
Thus we can identify the Dirac spinor
|Ψ1 i = |φ1 i → e↑−
|Ψ2 i = |φ2 i → e↓−
|Ψ3 i = |χ1 i → p↓
|Ψ4 i = |χ2 i → p↑
The positron pσ behaves exactly like an electron eσ with the opposite charge.
J POSITRONS 283
J.1 The action of quantum electrodynamics

Quantum electrodynamics is the most simple quantum field theory involving particles and fields,
namely electrons, positrons and the electromagnetic field. The “classical action” is the starting point
for setting up the quantum field theory.
ACTION OF QUANTUM ELECTRODYNAMICS
 
Z
1 1
d 4 x Ψ † γ 0 γ µ (i ~∂µ −qAµ )c − m0 c 2  Ψ − Fµν F µν
 
S=
c |{z} 4µ0
p̂µ
Here the wave function Ψ of the electron is a four-component spinor function, which we write as
4 × 1 matrix. F µν is the electric field tensor defined as
def
F µν = ∂ µ Aν − ∂ ν Aµ
The Dirac matrices obey the anticommutation relation

γ µ , γ ν ]+ = 2g µ,ν
[γ
where g µ,ν is the metric tensor. The Dirac matrices can have the form
 
! 10 0 0
1 0 0 1 0 0 
γ0 =
 
= 
0 −1 1  0 0 −1 0 
0 0 0 −1
 
! 0 0 01
0 σx  0 0 1 0
γ1 =
 
= 
−σ σx 0  0 −1 0 0 
−1 0 0 0
 
! 0 0 0 −i
0 σy  0 0i 0 
γ2 =
 
= 
−σ σy 0  0 i 0 0 
−i 0 0 0
 
! 0 01 0
0 σz  0 0 0 −1 
γ3 =
 
= 
−σ σz 0  −1 0 0 0 
0 10 0
In order to avoid transforming the Dirac matrices under Lorentz transformations, we can alter-
natively transform the spinors so that
′
a µ = Λµ ν aν
Ψ ′† γ µ Ψ ′ = Λµ ν Ψ † γ ν Ψ
Thus
Ψ′ = S Ψ
S † γ µ S = Λµ ν γ ν
284 J POSITRONS
Mathematical appendices
285
Appendix K
Method of separation of variables
The method of separation of variables is a method to solve partial differential equations. While it is
not applicable only in a subset of systems, it leads to so dramatic simplification that it is usually the
first method that is tried.
A simple example
Let us consider a simple example first, namely the one-dimensional wave equation
1 2
∂x2 Φ(x, t) = ∂ Φ(x, t)
c2 t
We start with an product-Ansatz
Φ(x, t) = f (x)g(t)
During the calculation we have to verify if the Ansatz leads to a solution. If inconsistencies appear,
we need to discard the Ansatz.
By dividing through the wave function we obtain
1 2 1 1 2
∂x f (x) = 2 ∂ g(t)
f (x) c g(t) t
Since the right side does not depend on x, the left-hand side must be a constant. The analogous
argument holds for the right-hand side. This allows us to break the problem up into two one-
dimensional problems, namely
1 2
∂ f (x) = C
f (x) x
1 1 2
∂ g(t) = C
c 2 g(t) t
which are two eigenvalue equations.
∂x2 f (x) = Cf (x)

1 2
∂ g(x) = Cg(x)
c2 t
The breakup of the partial differential equation into several differential equations of lower dimension-
ality is the essence of the method of separation of variables. The appearance of Eigenvalue equations
is also a fairly common feature of the method of separation of variables.
287
288 K METHOD OF SEPARATION OF VARIABLES
The method
The general idea of the method of separation of variables is to break up a partial differential equation
into problems of lower dimensionality. It starts with a ansatz of a product wave function.
While I am not sure if the following describes the complete class of problems to which the method
of separation of variables can be applied, it is by far the most common one.
The method of separation of variables works always, if the differential operator can be written as
a sum of terms, where each only acts on a subset of the coordinates. In that case we can break the
partial differential equation into those subsets.
~ x, ∇
D(∇ ~ y,~
x , y~ )Φ(~
x , y~) = 0
with
~ x, ∇
D(∇ ~ y,~ ~ x,~
x , y~) = D1 (∇ ~ y , y~)
x ) + D 2 (∇
We choose the Ansatz
x , y~ ) = f (~
Φ(~ x )g(~
y)
and obtain

~ x,~
D 1 (∇ ~ y , y~) f (~
x ) + D 2 (∇ x )g(~
y) = 0
1 ~ x,~ 1 ~ y , y~)g(~
D 1 (∇ x )f (~
x) + D 2 (∇ y) = 0
f (~
x) g(~y)
The important step is that D1 f g = gD1 f and D2 f g = f D2 g, which is valid, because the differential
operators do not act on the coordinates of the function, that is shifted in front of the expression.
This leads to
1 ~ x,~
D 1 (∇ x )f (~
x) = C
f (~
x)
1 ~ y , y~)f (~
D 2 (∇ y) = C
g(~y)
which is equivalent to two eigenvalue equations for the individual subsets

~ x,~
D 1 (∇ x ) = −Cf (~
x )f ~ x)
D 2 (∇~ y , y~ )g(~
y ) = Cg(~
y)
Appendix L
Trigonometric functions
π
The zero’s of the cosine function lie at odd integer multiples of 2 and those of the sin function lie
at even multiples of π2 .
π
cos(x) = 0 ⇒ x= (1 + 2k)
2
π
sin(x) = 0 ⇒ x = (2k)
2
with integer k.
The cosinus is an even function with respect to sign-change of the argument and the sinus is an
odd function
cos(x) = cos(−x)
sin(x) = − sin(x)
The formulas involving trigonometric functions appear quite complex and hard to memorize.
However, they are extremely simple to derive, if one only considers them as real and imaginary part
of the exponential function with imaginary argument.
One needs for this the identity
ei x = cos(x) + i sin(x) (L.1)
the elementary rules for the exponential function
ex+y = ex ey
e0 = 1
∂x e x = e x .
We start with the first formula that we will need frequently:

ei x e−i x = ei 0 = 1
1 = cos(x) + i sin(x) cos(x) − i sin(x) = cos2 (x) + sin2 (x)
Thus we obtain
cos2 (x) + sin2 (x) = 1 (L.2)
The addition theorem is then obtained in the following way

ei (x+y ) = ei x ei y

Eq. L.1
cos(x + y ) + i sin(x + y ) = cos(x) + i sin(x) cos(y ) + i sin(y )

= cos(x) cos(y ) − sin(x) sin(y ) + i cos(x) sin(y ) + sin(x) cos(y )
289
290 L TRIGONOMETRIC FUNCTIONS
By comparing real and imaginary part individually we obtain
cos(x + y ) = cos(x) cos(y ) − sin(x) sin(y ) (L.3)

sin(x + y ) = cos(x) sin(y ) + sin(x) cos(y ) (L.4)
Similarly we obtain the square of sinus and cosinus as

2
ei x = e2i x

Eq. L.1
cos(x) + i sin(x) = cos(2x) + i sin(2x)
cos2 (x) − sin2 (x) + 2i sin(x) cos(x) = cos(2x) + i sin(2x) (L.5)
With the help of Eq. L.2 we can extract the square of cosinus and sinus functions from the real
part of the above identity.
1 2 1
cos2 (x) = cos (x) + sin2 (x) + cos2 (x) − sin2 (x)
2| {z } 2| {z }
=1 cos(2x)
1
= 1 + cos(2x) (L.6)
2
1 2 1
sin2 (x) = cos (x) + sin2 (x) − cos2 (x) − sin2 (x)
2| {z } 2| {z }
=1 cos(2x)
1
= 1 − cos(2x) (L.7)
2
The derivatives are obtained as
dei x
= i ei x
dx
d cos(x) d sin(x)
+i = − sin(x) + i cos(x)
dx dx
Thus we obtain
d cos(x)
= − sin(x)
dx
d sin(x)
= + cos(x)
dx
Appendix M
Gauss’ theorem
Gauss theorem says that the divergence of a vector field integrated over a volume is equal to the
component of the vector field normal to the surface integrated over the surface of the volume
Z X XZ
dr ∇i fi (r ) = dAi fi (r ) (M.1)
Ω i i ∂Ω
It is the multidimensional analogue of

Z b
df
dx = f (b) − f (a)
a dx
A variant of Gauss’ theorem

Z Z
dr ∇i g(r ) = dAi g(r ) (M.2)
Ω ∂Ω
is obtained by using fi = ei g(r ), where ei = δi ,j is a unity vector pointing into the Cartesian direction
j.
Z X Z X Z XZ
dr ∇i f (r ) = dr ∇i δi ,j g(r ) = dr ∇j g(r ) = dAi fi (r )
Ω i Ω i Ω i ∂Ω
XZ Z
= dAi δi ,j g(r ) = dAj g(r )
i ∂Ω ∂Ω
291
292 M GAUSS’ THEOREM
Appendix N
Fourier transform
N.1 General transformations

We would like to expand the function in a set of functions {φi (x)}. This set of functions should be
complete so that any function f (x) is can be expressed as
X
f (x) = φi (x)fi
i
We obtain the expansion coefficients fi in the following way. We pick one function out of the set,
multiply the complex conjugate from the left and integrate
Z XhZ i
∗
dxφj (x)f (x) = dxφ∗j (x)φi (x) fi
i
We construct an inverse U of the overlap matrix, so that

X Z
Ui ,k dxφ∗k (x)φi (x) = δi ,j
k
Note that if functions in the set {φi } is over complete1 , the matrix U is not uniquely defined and
cannot be obtained using the conventional matrix inversion techniques. One way to obtain an inverse
is singular value decomposition.
Then we multiply the matrix U from the left and obtain
X Z
fi = Ui ,j dxφ∗j (x)f (x)
j
N.2 Fourier transform in an finite interval

Let us now consider a function f (x)that is defined on the interval [− L2 , L2 ]. Now we choose a special
set of functions φn (x) = Cei Gn x , where Gn L = 2πn. where C is an arbitrary constant.
The overlap matrix is
Z L Z Z
2
L/2 L/2
dx φ∗n (x)φm (x) = C 2 dx ei (Gn −Gm )x = C 2 dxe2πi (m−n)x/L = C 2 Lδn,m
− L2 −L/2 −L/2
1 A set of functions is over complete, if one function out of the set can be represented as a linear superposition of
the others.
293
294 N FOURIER TRANSFORM
which is easily inverted so that

Z Z
1 −i Gn x 1
fi = 2 dxCe f (x) = dxe−i Gn x f (x)
C L CL
Forward an backward transform are therefore
X
f (x) = C ei Gn x fn
n
Z L/2
1
fn = dxe−i Gn x f (x)
CL −L/1
with Gn L = 2πn
The choice of the parameter C is arbitrary. The following choices are common.
• A good choice is C = 1, so that
X
f (x) = ei Gn x fn
n
Z L/2
1
fn = dx e−i Gn x f (x)
L −L/1
• Often the symmetric form is preferred.

r
1 X i Gn x
f (x) = e fn
L n
r Z L/2
1
fn = dx e−i Gn x f (x)
L −L/1
N.3 Fourier transform on an infinite interval

Let us now consider the limit L → ∞. We introduce the spacing ∆ = 2π L between reciprocal lattice
points. Since
P theRspacing vanishes in the limit of a large interval, we can convert the sum into an
integral ∆ n → dG of the continuous variable G.
Z Z
C CL
f (x) = dGei Gx f (G) = dGei Gx f (G)
∆ 2π
Z L/2
1
f (G) = dx ei Gx f (x)
CL −L/2
The limit can be performed if we assure that CL = β remains constant.
Z
β
f (x) = dGei Gx f (G)
2π
Z
1
f (G) = dxei Gx f (x)
β
Different conventions are in use:
• A symmetric form can be obtained
√ by equating the pre factors for the forward and the backward
transformation, that is β = 2π.
Z
1
f (x) = √ dGei Gx f (G)
2π
Z
1
f (G) = √ dxe−i Gx f (x)
2π
N FOURIER TRANSFORM 295
• Personally, I prefer the non-symmetric version with β = 2π, which I am using in the present
lecture notes.
Z
dG i Gx
f (x) = e f (G) (N.1)
2π
Z
f (G) = dxe−i Gx f (x) (N.2)
N.4 Table of Fourier transforms

R d G i Gx
R
f (x) = 2π e f (G) f (G) = dxe−i Gx f (x)
δ(x) 1
ei G0 x 2πδ(G − G0 )
y (x − x0 ) y (G)e−i Gx0
R
dx g(x − x ′ )y (x)
′
g(G)y (G)
N.5 Dirac’s δ-function

The defining equation
Z
f (x0 ) = dxδ(x − x0 )f (x)
Variable transform of the δ-function

Z
f (y (x0 )) = dy δ(y − y (x0 ))f (y )
Z
dy
= dx δ(y (x) − y (x0 ))f (y (x))
dx
On the other hand we can write
Z
f (y (x0 )) = dxδ(x − x0 )f (y (x))
which follows directly from the definition of the delta function. By comparing the two expressions,
we obtain
dy
δ(x − x0 ) = δ(y (x) − y (x0 )
dx
The Fourier transform of the δ-function is a a plane wave in reciprocal space. If the delta function
is located at the origin, it is a constant.
Z
dxδ(x − x0 )e−i kx = e−i kx0
We can now use the transformation equations for the Fourier transform Eqs. N.1,N.2 to identify
f (k) = e−i kx0 with the Fourier transform of the delta function. Inserting this into the backward
transform Eqs. N.1, we obtain the representation of the delta function as integral over k-space
Z
dk i k(x−x0 )
δ(x − x0 ) = e
2π
296 N FOURIER TRANSFORM
Appendix O
Pauli Matrices
All two-by-two matrices can be expressed as a superposition of the unity matrix and three Hermitian
two-by-two matrices. These matrices are called Pauli matrices.
! ! !
01 0 −i 1 0
σx = ; σy = ; σz = .
10 i 0 0 −1
They play an important role for the description of the electron spin, for any two-state system and
for the quantum field theory of Fermions.
In order to become familiar with Pauli matrices, let us now investigate the commutator and
anti-commutator relations. We use the multiplication table, which is obtained using the two-by-two
matrices
σ j σ j +1 = i σ j +2
σ j σ j −1 = −i σ j −2
σj σj = 1
where the j refers to the indices (x, y , z) with cyclic permutation. That is if j = y then j + 1 = z
and j + 2 = x. etc.
The multiplication table has the form
σ1σ 2 σx σy σz
σx 1 iσ
σz -iσ
σy
σy -iσ
σz 1 iσ
σx
σz iσ
σy -iσ
σx 1
where σ 1 refers to the rows, and σ 2 to the columns.

The commutator table is obtained as the antisymmetric part of the multiplication table
σ 1 σ 2 ]−
[σ σx σy σz
σx 0 2iσ
σz -2iσσy
σy -2iσ
σz 0 2iσ
σx
σz 2iσ
σy -2iσ
σx 0
and the anti-commutator table is the symmetric part
σ 1 σ 2 ]+
[σ σx σy σz
σx 2 0 0
.
σy 0 2 0
σz 0 0 2
297
298 O PAULI MATRICES
We can derive a very useful formula for Pauli matrices, namely

MAGIC FORMULA
~σ )(B~
(A~ ~σ) = A
~B~ + i~ ~×B
σ (A ~ (O.1)
which is proven in the following.

Let us consider two superpositions of Pauli matrices
X
A= ~σ
Ai σ i = A~
i
X
B= ~σ
Bi σ i = B~
i
The identity Eq. O.1 can be verified by using the multiplication table.
X
~σ )(B~
(A~ ~σ) = Aj Bk σ j σ k
j,k
X X
= Aj Bj σ 2j + Aj Bk σ j σ k
j j 6=k
X X
= Aj Bj + ǫj,k,l Aj Bk i σ l
j j 6=k
~B
=A ~ + i~ ~ × B)
σ (A ~
Hereby ǫi ,j,k is the completely antisymmetric tensor with ǫi ,j,k = 0 if any pair of the indices are
equal. ǫi ,j,k = 1 if (i , j, k) is ascending, that is can be mapped onto the sequence (1, 2, 3, 1, 2), and
ǫi ,j,k = −1 if (i , j, k) is descending, that is can be mapped onto the sequence (2, 1, 3, 2, 1). The
P
vector product can be written as (A ~ × B)
~ i=
j,k ǫi ,j,k Aj Bk
Appendix P
Linear Algebra
P.1 Half Group

A half group G contains elements {a, b, . . .} and an operation ◦. The requirements for half-group
are
• completeness: The result of the operation between any two elements is within the group:
∀a,b∈G : a ◦ b ∈ G
• neutral element: There is an element of the group, called neutral element, which does not
change the group element it operates on.
∃1∈G : ∀a∈G : a ◦ 1 = 1 ◦ a = a
• associative law1
∀a,b,c∈G : a ◦ (b ◦ c) = (a ◦ b) ◦ c
A group is a half group with a definition of the inverse
• inverse element
∀a∈G : ∃a−1 ∈G : a ◦ a−1 = a−1 ◦ a = 1
For example the non-negative integers together with the addition constitute a half-group. The
integers with the addition form a group.
P.2 Vector space

A vector space ios a set that is closed under addition and multiplication with sa scalar. (http:
//mathworld.wolfram.com/VectorSpace.html)
The following conditions must be met
• Commutativity: a~ + ~b = ~b + a~
a + ~b) + ~
• Associativity of the vector addition: (~ c = a~ + (~b + ~
c)
1 “associative law” translates as “Assoziativgesetz” into German
299
300 P LINEAR ALGEBRA
• additive identity: ∀a~~0 + a~ = a~ + ~0 = a~

a) = ~0
• existence of an additive inverse: ∀a~∃−~a a~ + (−~
• associativity of scalar multiplication: x(y a~) = (xy )~
a
• distributivity of scalar sums: (x + y )~
a) = x a~ + y a~
a + ~b) = x a~ + x ~b
• distributivity of vector sums: x(~
• scalar multiplication identity: 1~
a = a~
P.3 Banach space

A Banach space is a complete Hilbert space with a norm.
P.4 Hilbert space

A Hilbert space is a vector space with a scalar product, so that the norm of a vector is the square
root of the scalar product of the vector with itself
Appendix Q
Matrix identities
Editor: This whole chapter is in preparation
Q.1 Notation
We introduce the complete orthonormal set of unit vectors
ei
~
which have the components
e )i = ~
(~ ei ~
ej = δi ,j
A matrix can
Q.2 Identities related to the trace

Q.2.1 Definition
The trace of a matrix is defined as the sum of its diagonal elements.
def
X X
A] =
Tr[A Ai ,i = δi ,j Ai ,j (Q.1)
i i ,j
A matrix can be written in the form

A = [~
a1 , a~2 , . . .]
where the vector components (~
ai )j = Ai ,j or as
X X X
A=[ ~
ei Ai ,1 , ei Ai ,2 ,
~ ei Ai ,3 , . . .]
~ (Q.2)
i i i
Q.2.2 Invariance under commutation of a product
AB ] = Tr[B
Tr[A B A]
Proof:
X X
AB ] =
Tr[A Ai ,j Bj,i = B A]
Bj,i Ai ,j Tr[B
i ,j i ,j
301
302 Q MATRIX IDENTITIES
Q.2.3 Invariance under cyclic permutation

The trace is invariant under cyclic permutation
AB C ] = Tr[C
Tr[A C AB ] (Q.3)
The proof follows directly from the invariance of the trace under commutation of a product of
two matrices: One simply considers the product A B as one matrix and C as the other.
Q.2.4 Invariance under unitary transformation

The trace is invariant under transformation by a unitary matrix, i.e.
U † A U ] = T r [A
Tr[U A] (Q.4)
X X U U † =1
1
U † AU ] =
Tr[U Ui†,j Aj,k Uk,i = Uk,i Ui†,j Aj,k = T r [U
U U † A] = A]
T r [A
i ,j,k i ,j,k
Q.3 Identities related to the determinant

Q.3.1 Definition
The determinant is defined by the fully antisymmetric tensor ǫ as
def
X
A] =
det[A ǫi ,j,k,... Ai ,1 Aj,2 Ak,3 · · · (Q.5)
i ,j,k,...
The fully antisymmetric tensor has the following elements
ǫ1,2,3,... = 1
• The matrix element vanishes if two indices are equal.
• The matrix element changes its sign for every permutation of two indices
In the reverse the epsilon tensor can be expressed by the determinant by

h i X
det ~
ei , ~
ej , ~
ek . . . , ] = ǫi ′ ,j ′ ,k ′ ,... δi ′ ,i δj ′ ,j δk ′ ,k · · · = ǫi ,j,k,...
i ′ ,j ′ ,k ′ ,...
The determinant can also be written in the following form

X X
A] =
det[A ǫi ,j,k,... Ai ,1 Aj,2 Ak,3 · · · = ei , ~
det[~ ek , . . .]Ai ,1 Aj,2 Ak,3 · · ·
ej , ~
i ,j,k,... i ,j,k,...
Q.3.2 Product rule
AB ] = det[A
det[A A] det[B
B] (Q.6)
Q MATRIX IDENTITIES 303
Proof: The proof is taken from http://algebra.math.ust.hk/determinant/05_proof/lecture3.

shtml#product. Fix the matrix A and consider
def
B ) = det[A
f (B AB ]
as a function of the matrix B
Let us use the notation A = [a~1 , a~2 , . . . , a~n ] where a matrix is represented by its column vectors.
A matrix product can then be written as
A~b1 , A~b2 . . .]
A B = [A
Now we derive two properties of the function f , that we will need later:
1.
f ([. . . , u~ + ~ A[. . . , u~ + ~
v , . . .]) = det[A v , . . .]) = det[[. . . , A (~
u+~
v ), . . .]) = det[[. . . , A u~ + A~
v , . . .])
??
= det[[. . . , A u~, . . .]] + det[[. . . , A~
v , . . .]] = f ([. . . , u~, . . .]) + f ([. . . , ~ (Q.7)
v , . . .])
2.
f [(. . . , u~, . . . , ~ A · [. . . , u~, . . . , ~
v , . . .]) = det[A v , . . .]] (Q.8)
A
= det[[. . . , u~, . . . , ~A v , . . .]] (Q.9)
= − det[[. . . , A~
v , . . . , A u~, . . .]] (Q.10)
A[. . . , ~
= − det[A v , . . . , u~, . . .]] (Q.11)
= −f [(. . . , ~
v , . . . , u~, . . .]) (Q.12)
Thus we can write

Eq. Q.2
X X
f ([[B]) = f[ ~
ei Bi ,1 , ei Bi ,2 , . . .]
~
i i
Eq. Q.7 X
= f [~ ej , . . .]Bi ,1 Bj,2 · · ·
ei , ~ (Q.13)
i ,j,k,...
Now we interchange the unit vectors in f until they are in ascending order. According to Eq. Q.12
every interchange causes a sign change. The sign changes can be encoded by the fully antisymmetric
tensor ǫ
Eq. Q.13 X
f ([[B]) = f [~ ej , . . .]Bi ,1 Bj,2 · · ·
ei , ~
i ,j,k,...
Eq. Q.12 X
= f [~ e2 , . . .]ǫi ,j,k,... Bi ,1 Bj,2 · · ·
e1 , ~
i ,j,k,...
X
= f [~
e ,~
e , . . .] ǫi ,j,k,... Bi ,1 Bj,2 · · ·
| 1 {z2 }
i ,j,k,...
A ·1
det[A 1] | {z }
B]
det[B
= A] det[B
det[A B]
Because f (B AB ] this proves Eq. Q.6.
B ) = det[A
Q.3.3 Permutation of a product

Permutation of a product of two square matrices does not change the determinant
AB ] = det[B
det[A B A] (Q.14)
The proof relies on the product rule Eq. Q.6
Eq. Q.6 Eq. Q.6
AB ]
det[A = A] det[B
det[A B ] = det[B
B ] det[A
A] = B A]
det[B
304 Q MATRIX IDENTITIES
Q.3.4 1
det[ecAA ] = ecTr[AA] (Q.15)
Proof: Assuming that A can be diagonalized, let
A = U †a U (Q.16)
where a is a diagonal matrix containing the eigenvalues of a, and U is a unitary matrix containing
the eigenvectors.
∞
" # "∞ #
X 1 X 1 n
A
cA n Eq. Q.16 †
det[e ] = det (cAA) = det U caa U
n=0
n! n=0
n!
" ∞
! #
U † U =1
1
X 1 n
= det U † (caa ) U
n=0
n!
h i
= det U † ecaa U
h i
Eq. ??
= det U U † ecaa
U † U =1
1
= det [ecaa ]
= eca11 · eca22 · · ·
= eca11 +ca22 +···
= ecTr[aa ]
†
= ecTr[UU a U ]
= ecTr[AA]
Appendix R
Special Functions
Quantum mechanical problems often characterize special functions. A good source to find details
about them is Abramovitz Mathematical Tables and in a more accessible form the appendices of
Messiah Quantum Mechanics.
R.1 Bessel and Hankel functions

Bessel and Hankel functions play a role as radial part of the Schrödinger equation with a constant
potential.
R.2 Hermite Polynomials

Hermite polynomials are used to construct the solutions of the Schrödinger equation for the one-
dimensional harmonic oscillator.
R.3 Legendre Polynomials

The Legendre polynomials appear in the θ dependence of spherical harmonics.
R.4 Laguerre Polynomials

(Source: Merzbacher)
Laguerre Polynomials are are used for the solutions of the hydrogen atom.
[x∂x2 + (1 − x)∂x + n]Ln (x) = 0

Ln (0) = n!
The solutions are the Laguerre polynomials
d n h n −x i
Ln (x) = ex x e = (−1 + ∂x )n x n
dx n
305
306 R SPECIAL FUNCTIONS
n Ln (x)
0 1
1 1−x
2 x 2 − 2x + 2
3 −x 3 + 9x 2 − 18x + 6
-2
-4
0 1 2 3 4 5 6 7 8
Fig. R.1: Laguerre polynomials rescaled so that the value at the origin is unity
Associated Laguerre Polynomials
[x∂x2 + (m + 1 − x)∂x + n − m]Lq (x) = 0
The associated Laguerre polynomials have the form
Lm m
n (x) = ∂x Ln (x)
We start out with the radial Schrödinger equation for the Coulomb potential in Hartree atomic
units.
h 1 ℓ(ℓ + 1) Z i
− ∂r2 + 2
− − E r ℓ+1 e−λr F (r )
2 2r r
Now we use
ℓ + 1
∂r r ℓ+1 e−λr = r ℓ+1 e−λr − λ + ∂r
r
ℓ + 12
∂r2 r ℓ+1 e−λr =r ℓ+1 −λr
e − λ + ∂r
r
ℓ + 1 2 ℓ+1 ℓ + 1 ℓ + 1 2
2
− λ + ∂r = ∂r − 2 + 2 − λ ∂r + −λ
r r r r
ℓ + 1 ℓ + 1 (ℓ + 1)2 ℓ+1
= ∂r2 + 2 − λ ∂r − 2 + − 2λ + λ2
r r r2 r
ℓ + 1 ℓ(ℓ + 1) ℓ+1
= ∂r2 + 2 − λ ∂r + − 2λ + λ2
r r2 r
R SPECIAL FUNCTIONS 307
Now we insert this to obtain an equation for F (r )

h
ℓ + 1 ℓ(ℓ + 1) ℓ+1
∂r2 + 2− λ ∂r + − 2λ + λ 2
r r2 r
ℓ(ℓ + 1) 2Z i
− + + 2E F (r ) = 0
h r 2 r i
r ∂r2 + 2(ℓ + 1) − 2λr ∂r + 2Z − 2λ(ℓ + 1) + r 2E + λ2 F (r ) = 0
h Z r i
x∂x2 + 2(ℓ + 1) − x ∂x + − (ℓ + 1) + 2E + λ2 F (x/(2λ) = 0
λ 2λ
√
Now we select λ = −2E and obtain
h Z i
x∂x2 + 2(ℓ + 1) − x ∂x + − (ℓ + 1) F (x/(2λ) = 0
λ
√
Now√we set p + 1 = 2(ℓ + 1) and q − p = Z 2 −2E − ℓ − 1 and obtain ... with p = 2ℓ + 1 and
q = Z 2 −2E + ℓ. In order to obtain an integer solution for q we need
Z2
E=−
2n2
with n = q − ℓ or q = n + ℓ.
Thus the solution is
ℓ −λr
R(r ) = L2ℓ+1
n+ℓ (2λr )r e
p Z
λ= −2En =
n
q
Z2
√
For n = −2E the solutions are the associated Laguerre polynomials F (x/ −2E) = Lℓn (x)
308 R SPECIAL FUNCTIONS
Appendix S
Principle of least action for fields
Now let us generalize the Formalism of least action to fields: in general we have now a Lagrangian
density that depends on the field amplitudes φ(x, t), their time and spatial derivatives1 ∂t φ and ∂x φ
respectively and of course on time and spatial coordinates.
L(φ, ∂t φ, ∂x φ, x, t)
The action is a functional of the field

Z Z
S([φ(x, t], x, t) = dt dxL(φ, ∂t φ, ∂x φ, x, t)
The Euler-Lagrange equations are obtained from the variation of φ(x, t) by δφ(x, t), which yields
Z Z h dL dL
δS([φ(x, t], x, t) = dt dx δφ(x, t) + δ∂t φ(x, t)
dφ(x, t) d∂t φ(x, t)
dL i
+ δ∂x φ(x, t)
d∂ φ(x, t)
Z x Z h dL i
dL dL
= dt dxδφ(x, t) − ∂t − ∂x
dφ(x, t) d∂t φ(x, t) d∂x φ(x, t)
We have used Gauss theorem, and the requirement that the variation of the field vanishes at the
integration boundary, that is at ti and tf and at x = ±∞.
The resulting Euler-Lagrange equations are
dL dL dL
∂t = − ∂x
d∂t φ(x, t) dφ(x, t) d∂x φ(x, t)
We can also derive the Hamilton equations by a Legendre transform of the Lagrangian. We first
introduce a field momentum density
δL
Π(x, t) =
δ∂t φ
The field momentum has a different physical meaning than the momentum of mass particles, even
though formally they are equivalent.
Next define a Hamilton function
Z h i
H([Π(x, t)], [φ(x, t)], x, t) = dx Π(x, t)∂t φ(x, t) − L
1 There may be more than one spatial derivatives but we remain here consistent with the one-dimensional description.
The generalization is obtained by indexing the spatial coordinates and introducing more of them.
309
310 S PRINCIPLE OF LEAST ACTION FOR FIELDS
We consider the derivatives of the field amplitudes as independent variables.The derivative of the
right hand side with respect to ∂t φ, vanishes if the Momenta are defined as above.
Let us now consider the derivatives with respect to field amplitudes and momenta.
δH
= ∂t φ(x, t)
δΠ(x, t)
δH ∂L
=−
δφ(x, t) ∂φ
Appendix T
ℓ-degeneracy of the hydrogen atom
For the hydrogen atom, the eigenvalues depend only on the main quantum number, but the eigen-
values with the same main angular momenta are degenerate. This degeneracy is not a consequence
of rotational symmetry. Rotational symmetry only leads to degenerate eigenvalues with different
magnetic quantum numbers m. The ℓ-degeneracy is peculiar1 for the non-relativistic Schrödinger
equation with a Coulomb potential. It is not caused by a geometric symmetry but a so-called dy-
namical symmetry. A symmetry is called dynamical if it originates from the equations of motion
and not directly from a space-time symmetry.
I am following a description[32] by M. Jacobi and H. Spahr taken from http://www.desy.de/
~jlouis/Vorlesungen/QMII06/vortrag_09.pdf.
T.1 Laplace-Runge-Lenz Vector

The non-relativistic electron in a hydrogen atom experiences a potential that is analogous to the
Kepler problem. The Hamilton operator has the form
p~ˆ2 κ
Ĥ = − (T.1)
2m |~rˆ|
Ze 2
where κ = 4πǫ 0
for the hydrogen atom.
The classical Kepler problem has a conserved quantity, called the Laplace-Runge-Lenz vector
~ˆ def 1 ˆ ~ˆ ~ˆ ˆ κ
M = p~ × L − L × p~ − ~rˆ (T.2)
2m |~rˆ|
This is already the hermitian translation of the classical Runge-Lenz vector.
The classical Laplace-Runge-Lenz vector lies in the plane of the trajectory and points along the
principal axis of the ellipse. Conservation of the Runge-Lenz vector implies that the trajectory is
a ellipse that is stationary in space. For general spherical potentials there will be a precession of
apsides2 of any orbit. The observation of the precession of the apsides of Mercury has been one
supporting experimental evidence for the general theory of relativity, as it was a prove for a deviation
of the gravitational force from the pure 1/r form.
The Runge-Lenz vector is a constant of motion:
~ˆ Ĥ]
[M,
Eq. T.45
= 0 (T.3)
1 Also the spherical harmonic oscillator in three dimensions exhibits ℓ-degeneracy.
2 The apsides are the point of smallest and farthest distance from the system’s center of gravity. Singular: apsis,
plural apsides. The point of closest approach is called periapsis or pericenter. The point of farthest distance is called
apoapsis or apocenter. The perihelion is the periapsis of an orbit about the sun.
311
312 T ℓ-DEGENERACY OF THE HYDROGEN ATOM
T.1.1 Commutator relations of Runge-Lenz vector and angular momentum

X
[L̂i , L̂j ] = i~ ǫi ,j,k L̂k (T.4)
k
Eq. T.53
X
[M̂i , L̂j ] = i~ ǫi ,j,k M̂k (T.5)
k
Eq. T.54 −2Ĥ X

[M̂i , M̂j ] = i~ ǫi ,j,k L̂k (T.6)
m
k
T.1.2 Rescale to obtain a closed algebra

The operators M ~ˆ and L
~ˆ would form a closed algebra, if the Hamilton operator would not appear in
the last expression Eq. T.6. Because the Hamiltonian commutates with the Runge-Lenz vector, we
can define a new operator
r
ˆ ′ def
~ m − 1 ~ˆ
M =i Ĥ 2 M (T.7)
2
This expression is defined because Hamilton operator and Runge-Lenz vector commutate. Thus, for
each cartesian direction, I can determine a common system of eigenstates |ni with
Ĥ|ni = |niǫn
M̂i |ni = |nian
Then I can write the operator

r
mX an
M̂ ′ i = |ni √ hn|
2 n −ǫn
Note, that the eigensystem n is different for the three components of M ~ˆ ′ , because the different
components of M ~ˆ do not commutate. The new operator is not hermitian, because it has imaginary
eigenvalues for states with positive energy.
Because the Hamiltonian commutates with the Runge-Lenz vector and with the angular momen-
tum, we can easily adapt the commutator relations to the ones for the scaled vector M ~ ′.
X
[L̂i , L̂j ] = i~ ǫi ,j,k L̂k
k
Eq. T.5 X
[M̂ ′ i , L̂j ] = i~ ǫi ,j,k M̂ ′ k (T.8)
k
Eq. T.6
X
[M̂ ′ i , M̂ ′ j ] = i~ ǫi ,j,k L̂k (T.9)
k
T.2 SO(4) symmetry

One introduces the generalized angular momenta

def 1 ~ˆ ′
J~ˆ(1) = L~ˆ + M (T.10)
2

def 1 ~ ~ˆ ′
J~ˆ(2) = ˆ−M
L (T.11)
2
T ℓ-DEGENERACY OF THE HYDROGEN ATOM 313
These operators obey the following commutator relations that have been derived in Eq. T.56
3
X
[Jî(1) , Jˆj(1) ] = i ~ ǫi ,j,k Jˆk(1) (T.12)
k=1
3
X
(2) (2) (2)
[Jî , Jˆj ] = i ~ ǫi ,j,k Jˆk (T.13)
k=1
(1) (2)
[Jî , Jˆj ] = 0 (T.14)
This algebra is can be characterized[32] as SO(3)×SO(3), as SU(2)×SU(2), or as SO(4).

SO(d) stands for special orthogonal group in d dimensions. It is the group of real d × d matrices
with determinant +1. These matrices are the rotation matrices in d dimensions. In d dimensions
there are 1 + 2 + . . . + (d − 1) = d (d2−1) independent rotation axes corresponding to the independent
matrix elements in an antisymmetric matrix in d dimensions. The antisymmetric matrices are the
generators of this group.
SU(d) stands for special unitary group. It is the group of complex d ×d matrices with determinant
+1.
The SO(4) symmetry corresponds to rotations in a four-dimensional cartesian space. In four
dimensions, there are six rotation angles. These rotations do not operate in some physical space.
T.3 ℓ-Degeneracy of the hydrogen atom

In order to make the link to the ℓ-degeneracy, we show that the Hamiltonian of the hydrogen atom
can be expressed entirely by the generalized angular momenta J~ˆ(1) and J~ˆ(2) .
~ˆ 2 Eq. T.39 2 ~ˆ2

M = Ĥ L + ~2 + κ2
m
m −1 ~ˆ 2 ~ˆ2 m
⇒ − Ĥ M + L + ~2 = − κ2 Ĥ −1
2 2
Eq. T.7
~ˆ ′ 2 + L~ˆ2 + ~2 m 2 −1
⇒ M = − κ Ĥ
2
2 2 m 2 −1
Eqs. T.10,T.11
⇒ 2 J ˆ
~(1)
+ J~ˆ (2)
+ ~2 = − κ Ĥ
2
Thus we can express the Hamiltonian by the squares of the generalized angular momenta J~ˆ(1) and
J~ˆ(2) .
−1
mκ2 2 2
⇒ Ĥ = − 2 J~ˆ(1) + 2 J~ˆ(2) + ~2 (T.15)
2
Each of the operators J~ˆ(1) and J~ˆ(2) obeys an algebra that is identical to the angular momentum
algebra, which has been obtained for the normal angular momentum L. ~ˆ Therefore, we obtain the
same eigenvalue spectrum, namely
2
J~ˆ(1) |j1 , j2 i = |j1 , j2 i~2 j1 (j1 + 1) (T.16)
2
J~ˆ(2) |j1 , j2 i = |j1 , j2 i~2 j2 (j2 + 1) (T.17)
where j1 and j2 may be non-negative, half-integer numbers

def
With Eq. T.55 we can show that the j1 = j2 = j.

0
Eq. T.55
= ~ˆ ′ L|j
M ~ˆ 1 , j2 i Eq.=T.11 J~ˆ(1) − J~ˆ(2) J~ˆ(1) + J~ˆ(2) |j1 , j2 i

J~ˆ(1) − J~ˆ(2) |j1 , j2 i~2 [j1 (j1 + 1) + j2 (j2 + 1)]
Eq. T.17
=
Eq. T.17
= |j1 , j2 i~2 [j1 (j1 + 1) − j2 (j2 + 1)] ~2 [j1 (j1 + 1) + j2 (j2 + 1)]
⇒ j1 = j2 ∧ j1 = −(j2 + 1)
Since j1 and j2 are non-negative, we can conclude that the two parameters are identical.
Now return to Eq. T.15 and determine the eigenvalues
−1
mκ2 2 2
2 J~ˆ(1) + 2 J~ˆ(2) + ~2
Eq. T.15
Ĥ|ji = − |ji
2

mκ2 2
2 −1
= |ji − 4~ j(j + 1) + ~
2

mκ2 −2
= |ji − 2 [2j + 1]
2~
Thus we obtain the eigenvalues

def
n =2j +1 mκ2 1
En = −
2~2 n2
This is the eigenvalue spectrum of the hydrogen atom. We can form a basisset
2
J~ˆ(1) |j, m1 , m2 i = |j, m1 , m2 i~2 j(j + 1)
J~ˆz(1) |j, m1 , m2 i = |j, m1 , m2 i~m1

J~ˆ(2) |j, m , m i = |j, m , m i~m
z 1 2 1 2 2
2
because J~ˆ(1) , J~ˆz(1) and J~ˆz(1) form with the Hamiltonian a set of mutually commutating operators.
Because the energy depends only on j, there are (2j + 1)2 pairs of quantum numbers (m1 , m2 ).
Hence each eigenvalue has a degeneracy n2 , which is exactly the degeneracy predicted for the hydro-
gen atom including ℓ-degeneracy.
T.4 Derivation of commutator relations used in this chapter

Here I shall in future include the derivation of the expressions used above: These calculations are not
finished. The derivations are from http://www.theochem.ru.nl/~pwormer/rungelenz.pdf[33].
This derivation is not completed!!!!
X X
[p̂i , L̂j ] = ǫj,m,n [p̂i , x̂m p̂n ] = ǫj,m,n (p̂i x̂m p̂n − x̂m p̂n p̂i )
m,n m,n
 
X ~ ~X
= ǫj,m,n  δi ,m p̂n + x̂m p̂i p̂n − x̂m p̂n p̂i  = ǫj,i ,n p̂n
m,n
i | {z } i n
=0
X
= i~ ǫi ,j,k p̂k (T.18)
k
X X
[x̂i , L̂j ] = ǫj,m,n [x̂i , x̂m p̂n ] = ǫj,m,n (x̂i x̂m p̂n − x̂m p̂n x̂i )
m,n m,n

X ~
= ǫj,m,n x̂i x̂m p̂n − x̂m δn,i − x̂m x̂i p̂n
m,n
i
X
= i~ ǫi ,j,k x̂k (T.19)
k
X
~p =
L~ ǫi ,j,k xj pk pi = 0 (T.20)
|{z} |{z}
i ,j,k pi pk
=−ǫk,j,i
X
~ L
L( ~ × p~) = ǫi ,j,k Li Lj pk = 0 (T.21)
|{z} |{z}
i ,j,k
=−ǫj,i,k =Lj Li
X
~ p × L)
L(~ ~ = ǫi ,j,k Li pj Lk
i ,j,k
!
Eq. T.18
X X X
= ǫi ,j,k Li i~ ǫj,k,m pm + ǫi ,j,k Li Lk pj
i ,j,k m i ,j,k
| {z }
=0
 
X X Eq. T.20
= i~  ~p
ǫi ,j,k ǫm,j,k  Li pm = i ~L~ = 0 (T.22)
i ,m j,k
| {z }
2δi,m
~ ~r = 1 ~ 1 1
X X X
L ǫi ,j,k xj pk xi = ǫi ,j,k xj δi ,k + ǫi ,j,k xj xi pk
|~r| |~r| i |~r| |~r|
i ,j,k i ,j,k i ,j,k
~X 1 X 1
= ǫi ,j,i xj + ǫi ,j,k =−ǫj,i,k xj xi pk =0 (T.23)
i
j,j
|{z} |~r| i ,j,k |{z} |~r|
=0 =xi xj

~M~ Eq. T.2
~ 1 ˆ ~ˆ ~ˆ ˆ κˆ
L = L p~ × L − L × p~ − ~r
2m |~rˆ|
 
ˆ
=
1 ~ ˆ ~ˆ
L(p~ × L) − ~
L(L ~ ~r Eqs.
~ˆ × p~ˆ) − κ L T.21,T.22,T.23
= 0 (T.24)
2m | {z } | {z } |~rˆ|
=0 =0 |{z}
=0
X
~ˆ × p~ˆ
L = ǫi ,j,k L̂j p̂k
i
j,k
X X
= ǫi ,j,k p̂k L̂j − ǫi ,j,k [p̂k , L̂j ]
j,k j,k
Eq. T.18 X X X
= − ǫi ,j,k p̂j L̂k − i ~ ǫi ,j,k ǫk,j,m p̂m
j,k j,k m
 
X X X
= − ǫi ,j,k p̂j L̂k − i ~ − ǫi ,j,k ǫm,j,k  p̂m
j,k m j,k
| {z }
2δi,m

= ~ˆ + 2i ~p̂i
− p~ˆ × L (T.25)
i
X
~ =
p × L)
p~(~ ǫi ,j,k pi pj Lk = 0 (T.26)
|{z} |{z}
i ,j,k
=−ǫj,i,k =pj pi
h i
~ × p~) Eq.=T.25 p~ −(~
p~(L ~ + 2i ~~
p × L) p
Eq. T.26
p2
= 2i ~~ (T.27)
X X
(~ ~ p
p × L)~ = ǫi ,j,k pj Lk pi = ~ × p~)
ǫj,k,i pj Lk pi = p~(L
i ,j,k i ,j,k
Eq. T.27 2
= p
2i ~~ (T.28)
T.4.1 ~ˆ 2
Calculation of M
~ˆ and L
Products of p~ˆ × L ~ˆ × p~ˆ
!
X X X
(~ ~ 2
p × L) = ǫi ,j,k ǫi ,m,n pj Lk pm Ln = ǫi ,j,k ǫi ,m,n pj L k pm L n
i ,j,k,m,n j,k,m,n i
| {z }
δj,m δk,n −δj,n δk,m
X
= (pj Lk pj Lk − pj Lk pk Lj )
j,k
X
= (pj [Lk , pj ]Lk + pj pj Lk Lk − pj Lk pk Lj )
j,k
Eq. T.18
X X
= i~ ~2 −
ǫk,j,i pj pi Lk + p~2 L ~ p) L
pj (L~
|{z} |{z} | {z} j
i ,j,k j
=−ǫk,i,j =pi pj =0
Eq. T.20
= ~2
p~2 L (T.29)
This is the result of Eq. 15 of Wormer[33].

h i
(~ ~ L
p × L)( ~ × p~) Eq.=T.25 (~
p × L) ~ˆ + 2i ~p~ˆ
~ −p~ˆ × L
Eq. T.29
= −~ ~ 2 + 2i ~(~
p2L ~ p~ˆ
p × L)
X
= −~ ~ 2 + 2i ~
p2L ǫi ,j,k pj Lk pi
i ,j,k
X
2~ 2 ~ 2 + 2i ~ p~(L
~ × p~)
= −~
p L + 2i ~ p2L
ǫj,k,i pj Lk pi = −~ (T.30)
| {z }
i ,j,k
p2
2i ~~
Eq. T.27
= −~ ~ 2 − 4~2 p~2
p2L (T.31)
h i
~ × p~)(~
(L p × L) ~ˆ + 2i ~p~ˆ (~
~ Eq.=T.25 −p~ˆ × L ~
p × L)
Eq. T.29
= −~ ~ 2 + 2i ~ p~(~
p2L ~
p × L)
| {z }
=0
Eq. T.26
= −~ ~2
p2L (T.32)
h ih i
~ × p~)(L
(L ~ × p~) Eq. T.25
= ~ˆ + 2i ~p~ˆ −p~ˆ × L
−p~ˆ × L ~ˆ + 2i ~p~ˆ
 
Eq. T.29
= ~ 2 − 2i ~ 
p~2 L ~ p + p~(~
p × L)~
(~ ~ 
p × L)  − 4~2 p~ˆ2
| {z } | {z }
p2
2i ~~ =0
Eq. T.28,T.26 2~ 2
= p~ L (T.33)

We can combine these four results to
2
~ −L
p~ × L ~ 2 − (−~
~ × p~ = p~2 L ~ 2 − 4~2 p~2 ) − (−~
p2L ~ 2 ) + p~2 L
p2L ~2

p2 L
= 4~ ~ 2 + ~2 (T.34)
~ × p~) ~r X 1
(L = ǫi ,j,k Lj pk xi
|~r| |~r|
i ,j,k
X ~ 1 X 1
= ǫi ,j,k Lj δk,i + ǫi ,j,k Lj xi pk
i |~r| |~r|
i ,j,k i ,j,k
X ~ 1 X X 1
= ǫi ,j,i Lj + Lj ǫj,k,i xi pk
|{z} i |~r| |~r|
i ,j,k j i ,k
=0 | {z }
=−Lj
~2
L
=− (T.35)
|~r|
~ˆ2 commutates with any spherical function of

This is the result of Eq. 19 of Wormer[33]. Note that L
~r.
~ ~r Eq. T.25
~ × p~) ~r ~r
p × L)
(~ = −(L p
+ 2i ~~
|~r| |~r| |~r|
Eq. T.35 ~2
L ~r
= p
+ 2i ~~ (T.36)
|~r| |~r|
~r ~ = 1
X 1 ~2
p × L)
(~ ǫi ,j,k xi pj Lk = L (T.37)
|~r| |~r| |~r|
i ,j,k
~r ~ Eq. T.25 ~ r ~ + 2i ~~
~r
~ + 2i ~ ~r p~
(L × p~) = −~
p×L p =− p~ × L
|~r| |~r| |~r| |~r|
Eq. T.37 ~ˆ2
L 1
= − + 2i ~ ~rp~ (T.38)
|~rˆ| |~r|
2
~ˆ 2 Eq. T.2 1 ˆ ~ˆ ~ˆ ˆ κˆ
M = p~ × L − L × p~ − ~r
2m |~rˆ|
2
1 ˆ ~ˆ ~ˆ ˆ2 κ ˆ ~ˆ ~ˆ ˆ ~r κ ~r ˆ ˆ
~ ˆ
~ ˆ

2 ~r
= p~ × L − L × p~ − p~ × L − L × p~ − p~ × L − L × p~ +κ
4m2 | {z } 2m ˆ
|~r| 2m ˆ
|~r| |~rˆ|
| {z } | {z }
Eq. T.34
Eq. T.36 and Eq. T.35 Eq. T.37 and Eq. T.38
! !
1 2 h~ 2 2
i κ ~2
L ~r κ ~2
L ~r
= p L +~
4~ − 2 + 2i ~~p − 2 − 2i ~ p~ + κ2
4m2 2m |~r| |~r| 2m |~r| |~r|
1 2 h~ 2 2
i κ L ~2 κ ~r
= p
4~ L + ~ − 4 − 2i ~[~ p , ] + κ2
4m2 2m |~r| 2m |~r|

1 2 h~ 2 i κ L ~2 κ 3 1 2
2 2
= p L +~
4~ − 4 − 2~ − ~r + κ2
4m2 2m |~r| 2m |~r| |~r|3
i
2 p~2 h~ 2 κ L ~2 2~2 κ
= L + ~2 − 4 − + κ2
m 2m 2m |~r| m |~r|
2 p~2 h~ 2 i 2 κ
~ 2 − 2 κ ~2 + κ2
= L + ~2 − L
m 2m m |~r| m |~r|
2 h i
2 p~ κ ~ 2 + ~2 + κ2
= − L
m 2m |~r|
Eq. T.1 2 h~ 2 i
= Ĥ L + ~2 + κ2 (T.39)
m
This is the third statement or Theorem 2 of Wormer[33].

T.4.2 ~ˆ and H
Commutator of M ~ˆ
X
[~ ~ p~2 ]
p × L, = 2
ǫi ,j,k pj Lk pm 2
− pm pj L k
i
i ,j,k
X
= ǫi ,j,k (pj (Lk pm )pm − pj pm (pm Lk ))
i ,j,k
X
= ǫi ,j,k (pj [Lk , pm ]pm + pj pm Lk pm − pj pm [pm , Lk ] − pj pm Lk pm )
i ,j,k
X
= ǫi ,j,k (−pj [pm , Lk ]pm − pj pm [pm , Lk ])
i ,j,k
!
Eq. T.18
X X
= − ǫi ,j,k pj pm i~ ǫm,k,n pn
i ,j,k n
!
X X
= −i ~ ǫk,i ,j ǫk,n,m pj pm pn
i ,j,n k
| {z }
δi,n δj,m −δi,m δj,n
 
X X
= −i ~  pj pj pi − δj,n pj pi pj  = 0 (T.40)
j j
!
X X X X X
~ =
~r × L ǫi ,j,k xj Lk = ǫi ,j,k xj ǫk,m,n xm pn = ǫk,i ,j ǫk,m,n xj xm pn
i
j,k j,k m,n j,n,m k
| {z }
δi,m δj,n −δi,n δj,m
X X X
= δi ,m δj,n xj xm pn − δi ,n δj,m xj xm pn = (xj xi pj − xj xj pi )
j,n,m j,n,m j

= ~r(~rp~) − ~r2 p~ i (T.41)
 
X X X X X
~ × ~r =
L ǫi ,j,k Lj xk = ǫi ,j,k ǫj,m,n xm pn xk =  ǫj,k,i ǫj,m,n  xm pn xn
i
j,k j,k m,n k,n,m j
| {z }
δk,m δi,n −δk,n δi,m
X X X
= δk,m δi ,n xm pn xk − δk,n δi ,m xm pn xk = (xk pi xk − xi pk xk )
k,n,m k,n,m k
X ~ ~ X
= (− δk,i xk + pi xk xk + δi ,k xk − pk xk xi ) = (pi xk xk − pk xk xi )
i i
k k

= p~~r2 − (~
p~r)~r i (T.42)


xj ~ δi ,j xi xj
[pi , ] = − 3
|~r| i |~r| |~r|
 
xj X
pi pi xj −pi xj pi + pi xj pi − xj pi pi 

→ [p 2 , ] =
|~r|  |~r |~r |~r |~r 
i | {z }
=0
X xj xj

= pi [pi , ] + [pi , ]pi
|~r |~r
i

~X δi ,j xi xj δi ,j xi xj
= pi − 3 + − 3 pi
i |~r| |~r| |~r| |~r|
i

~ 2 1 xj 1 2 xj
= pj ~r − (~ p~x ) 3 + 3 ~r pj − 3 (~ x p~)
i |~r|3 |~r| |~r| |~r|
 
~  1 1 
=  pj ~r2 − (~p~
x )xj + 3 ~r2 pj − xj (~
x p~) 
3
i | {z } |~r| |~r| | {z }
~ r )j
=(L×~ −(~ ~ j
r ×L)

Eqs. T.42,T.41 ~ ~ 1 1 ~ j
= (L × ~r)j 3 − 3 (~r × L)
i |~r| |~r|
r ,|~
[~ ~ r |−3 ]=0
r |−3 ]=0;[L,|~ i~ ~ −L ~ × ~r

= ~
r × L (T.43)
|~r|3 j

1 ~ ~
X 1 1
[ ,L × p~ − p~ × L] = ǫi ,j,k (Lj pk − pj Lk ) − (Lj pk − pj Lk )
r i |~r| |~r|
j,k

~ r |−1 ]=0
[L,|~ X 1 1 1 1
= ǫi ,j,k Lj pk − pj Lk − Lj pk + pj L k
|~r| |~r| |~r| |~r|
j,k

X 1 1
= ǫi ,j,k −Lj [pk , ] + [pj , ]Lk
|~r| |~r|
j,k

~X xk xk
= ǫi ,j,k +Lj 3 − 3 pj
i |~r| |~r|
j,k

~ ~ ~r ~r ~
= +L × 3 − 3 × L
i |~r| |~r| j
~ r |−3 ]=0 ~ 1
[L,|~

= ~ × ~r − ~r × L
L ~ (T.44)
i |~r|3 j
This is the result of Lemma 10 of Wormer[33].

" #
~ˆ Ĥ] Eq. T.1,T.2 1 ˆ ~ˆ ~ˆ ˆ κ ˆ p~ˆ2 κ
[M, = p~ × L − L × p~ − ~r, −
2m |~rˆ| 2m |~rˆ| −

1 h ˆ ~ˆ ~ˆ ˆ 2 i κ ˆ ~ˆ ~ˆ ˆ 1
= p
~ × L − L × p
~ , p
~ − p
~ × L − L × p
~ ,
4m2 | {z } 2m |
− |~r| −
{z }
=0 (Eq. T.40,Eq. T.25) Eq. T.44
" # " #
κ ~rˆ 2 ~rˆ 1
− , p~ +κ2 ,
2m |~rˆ| −
|~rˆ| |~r| −
| {z } | {z }
Eq. T.43 =0
κ ~ 1 ~
~ − κ −i ~ ~r × L
~ −L
~ × ~r

= − L × ~
r − ~
r × L
2m i |~r|3 2m |~r|3
= 0 (T.45)
T.4.3 ~ˆ and L
Commutators of M ~ˆ
XX
~ × p~)i , (L
[(L ~ × p~)j ] = ǫi ,k,l ǫj,m,n (Lk pl Lm pn − Lm pn Lk pl ) (T.46)
k,l m,n
XX
= ǫi ,k,l ǫj,m,n (Lk [pl , Lm ]pn − Lm [pn , Lk ]pl ) (T.47)
k,l m,n
! ! !
XX X X
= ǫi ,k,l ǫj,m,n Lk i~ ǫl ,m,o po pn − L m i~ ǫn,k,o po pl
k,l m,n o o
X
= i~ (ǫi ,k,l ǫj,m,n ǫl ,m,o Lk po pn − ǫi ,k,l ǫj,m,n ǫn,k,o Lm po pl ) (T.48)
k,l ,m,n,o
! ! !
X X X X X
= i~ ǫi ,k,l ǫm,n,j ǫm,o,l L k po pn − ǫi ,k,l ǫj,m,n ǫn,k,o L m po pl
k,l ,o n m m n
!
X X X
= i~ ǫi ,k,l (δn,o δj,l − δn,l δj,o ) Lk po pn − ǫi ,k,l (δj,k δm,o − δj,o δm,k ) Lm po pl
k,l ,o n m
X X X
= i~ ǫi ,k,l δn,o δj,l Lk po pn − ǫi ,k,l δn,l δj,o Lk po pn
k,l ,o n n
X X
− ǫi ,k,l δj,k δm,o Lm po pl + ǫi ,k,l δj,o δm,k Lm po pl
m m
 
 
X X X X 
 
= i~  ǫi ,k,j Lk pn pn − ǫi ,k,n Lk pj pn − ǫi ,j,l Lm pm pl + ǫi ,m,l Lm pj pl 
 
 k,n k,n l ,m l ,m 
| {z } | {z }
(A) (A)
!
X X
= −i ~ ~ p)
ǫi ,j,k Lk p~2 − (L~ ǫi ,j,k pk
k k
X
= −i ~ ~ p)pk
ǫi ,j,k Lk p~2 − (L~ (T.49)
k
This differs from the corresponding result of Lemma 12 of Wormer[33].

X
[(~ ~ i , (~
p × L) ~ j]
p × L) = ǫi ,k,l ǫj,m,n (pk Ll pm Ln − pm Ln pk Ll )
k,l ,m,n
X
= ǫi ,k,l ǫj,m,n (pk [Ll , pm ]Ln + pk pm Ll Ln − pm [Ln , pk ]Ll − pm pk Ln Ll )
k,l ,m,n
! ! !
Eq. T.18 X X X
= ǫi ,k,l ǫj,m,n pk −i ~ ǫm,l ,o p̂o L n − pm −i ~ ǫk,n,p p̂p Ll
k,l ,m,n o p
!
X X X
= −i ~ ǫi ,k,l ǫj,m,n ǫm,l ,o pk p̂o Ln − ǫk,n,p pm p̂p Ll
k,l ,m,n o p
! ! !
X X X X X X
= −i ~ ǫi ,k,l ǫm,n,j ǫm,l ,o pk p̂o Ln − ǫi ,k,l ǫn,j,m ǫn,o,k pm p̂o Ll
k,n,o l m l ,m,o k n
!
X X
= −i ~ ǫi ,k,l (δn,l δj,o − δn,o δj,l ) pk p̂o Ln
k,n,o l
!
X X
− ǫi ,k,l (δj,o δm,k − δj,k δm,o ) pm p̂o Ll
l ,m,o k
!
X X X X
= −i ~ ǫi ,k,n pk p̂j Ln − ǫi ,k,j pk p̂n Ln − ǫi ,m,l pm p̂j Ll + ǫi ,j,l pm p̂m Ll
k,n k,n l ,m l ,m
 
 
X X X X 
 
= −i ~  ǫi ,k,n pk p̂j Ln − ǫi ,k,j pk p̂n Ln − ǫi ,k,n pk p̂j Ln + ǫi ,j,n pk p̂k Ln 
 
 k,n k,n k,n k,n 
| {z } | {z }
(A) (A)
!
X X
= −i ~ + ~ˆ + p~2
ǫi ,j,k pk (p~ˆL) ǫi ,j,n Ln (T.50)
k,n n
X X
[(~ ~ i , pj ]
p × L) = ǫi ,m,n (pm Ln pj − pj pm Ln ) = ǫi ,m,n (pm [Ln , pj ] + pm pj Ln − pj pm Ln )
m,n m,n
| {z }
=0
!
Eq. T.18 X X
= ǫi ,m,n pm −i ~ ǫj,n,k p̂k
m,n k
!
X X
= −i ~ ǫn,i ,m ǫn,k,j p̂m p̂k
k,m n
X
= −i ~ (δi ,k δm,j − δi ,j δm,k ) p̂m p̂k
k,m

= −i ~ p̂i p̂j − δi ,j p~2 (T.51)
Eq. T.25
~ i , (L
p × L)
[(~ ~ × p~)j ] = −[(~ ~ i , (L
p × L) ~ × p~)j ] + 2i ~[(~
p × L)~ i , pj ]
!
X X
Eqs. T.50,T.51
= i~ + ˆ~ˆ
ǫi ,j,k pk (p~L) + p~2
ǫi ,j,n Ln + 2i ~ −i ~ p̂i p̂j − δi ,j p~2

k,n n
X X
~ˆ + i ~~

= i~ ǫi ,j,k pk (p~ˆL) p2 ǫi ,j,n Ln + 2~2 p̂i p̂j − δi ,j p~2 (T.52)
k,n n
These equations are still given without derivation.

X
[M̂i , L̂j ] = i ~ ǫi ,j,k M̂k (T.53)
k
−2Ĥ X
[M̂i , M̂j ] = i ~ ǫi ,j,k L̂k (T.54)
m
k
~ˆ ′ L
M ~ˆ = 0 (T.55)
T.4.4 Commutators of J~ˆ(1) and J~ˆ(2)

We determine the commutator relations of J~ˆ(1) and J~ˆ(2) . The signs in the definitions of the two
operators are encrypted by the variables σ. σ = +1 for the operator J~ˆ(1) and σ = −1 for the operator
J~ˆ(2) .

1 ′
1
′
1
L̂i + σ1 M̂ i , L̂j + σ2 M̂ j = L̂i + σ1 M̂ ′ i , L̂j + σ2 M̂ ′ j
2 2 4
 
1 
 L̂i , L̂j + σ2 L̂i , M̂ ′ j +σ1 M̂ ′ i , L̂j + σ1 σ2 M̂ ′ i , M̂ ′ j 
=
4  | {z } 
−[M̂ j ,L̂i ]
′
1 X
= i~ ǫi ,j,k L̂k + σ2 M̂ ′ k + σ1 M̂ ′ k + σ1 σ2 K̂k
4
k

X 1 1 + σ1 σ2 σ1 + σ2 ′
= i~ ǫi ,j,k L̂k + M̂ k
2 2 2
k
h i
ˆ(1) , Jˆ(1) σ1 =σ =1 P


 J i i =2 i ~ k ǫi ,j,k Jˆk(1)
h i
(1) (2) σ1 =−σ2 =1
⇒ Jî , Jî = 0 (T.56)

 h i
 ˆ(2) ˆ(2) σ1 =σ2 =−1 P
 Ji , Ji = i ~ k ǫi ,j,k Jˆk(2)
325
326 U A SMALL DICTIONARY
Appendix U
A small Dictionary
abelian abelsch
abbreviation Abkürzung (of a word)
action Wirkung
adjunct adjungiert
angle of refraction Brechungswinkel
angular frequency Kreisfrequenz
associative law Assoziativgesetz
arbitrary willkürlich, beliebig
bead glasperle
to blur verschwimmen
bracket Klammer
branch Zweig
canned aus der Dose
chord Saite
to commutate vertauschen, vertauschbar sein.
commutative Law Kommutativgesetz
continuous stetig
convolution Faltung
degeneracy Entartung
degenerate entartet
differentiable differenzierbar
distributive law Distributivgesetz
dyadic product dyadisches Produkt,äußeres Produkt
elevated erhöht, hoch
expectation value Erwartungswert
gauge symmetry Eichsymmetrie
gem stone Edelstein
Hamiltonian Hamilton operator
hermitian hermitesch
incident angle Einfallswinkel
integer ganze Zahl
intricacy Kniffligkeit
Lagrangian Lagrange Funktion
Lattice Gitter
Lettuce Salat
ledge Absatz, sims
mean square deviation mittlere quadratische Abweichung
momentum Impuls
U A SMALL DICTIONARY 327
normalizable normalisierbar
periodicity Periodizität
pitch Grundton
potential well Potentialtopf
power Potenz
power series expansion Potenzreihenentwicklung
principle of least action Wirkungsprinzip
precaution Vorsicht
principal axis Hauptachse
probability Wahrscheinlichkeit
q.e.d quod erat demonstrandum= Was zu zeigen war
radius of convergence Konvergenzradius
realm Fachgebiet / Gefilde / Reich
rectifier Gleichrichter
reference frame Bezugssystem
refraction Brechung
sequence Folge
series Reihe
set Menge
slope Steigung
sophisticated verfeinert, anspruchsvoll, ausgeklügelt
space Raum
spatial räumlich
spring constant Federkonstante
squared quadratiert
square root Wurzel
square variation quadratische Abweichung
state Zustand
strain Dehnung
string Saite
tension Spannung (elastisch)
transition probability Übergangswahrscheinlichkeit
trough Tal
wave crest Wellenberg
well Graben
328 U A SMALL DICTIONARY
Appendix V
Greek Alphabet
A α alpha N ν nu
B β beta Ξ ξ ksi
Γ γ gamma O o, omicron
∆ δ delta Π π, ̟ pi
E ǫ, ε epsilon P ρ, ̺ rho
Z ζ zeta Σ σ, ς sigma
H η eta T τ tau
Θ θ, ϑ theta Υ υ upsilon
I ι iota Φ φ, ϕ phi
K κ kappa X χ chi
Λ λ lambda Ψ φ phi
M µ mu Ω ω omega
329
330 V GREEK ALPHABET
Appendix W
Philosophy of the ΦSX Series
The ΦSX series tries to implement a few rules that I learned mostly from the feedback given to me
by the students attending the course and that relied on this books as background material.
The course should be self-contained. There should not be any statements “as shown easily...” if,
this is not true. The reader should not need to rely on the author, but he should be able to convince
himself, if what is said is true. I am trying to be as complete as possible in covering all material
that is required. The basis is the mathematical knowledge. With few exceptions the material is also
developed in a sequence so that the material covered can be understood entirely from the knowledge
covered earlier.
The derivations shall be explicit. The novice should be able to step through every single step of
the derivation with reasonable effort. An advanced reader should be able to follow every step of the
derivations even without paper and pencil.
All units are explicit. That is, formulas contain all fundamental variables, which can be inserted in
any desirable unit system. Expressions are consistent with the SI system, even though I am quoting
some final results in units, that are common in the field.
The equations that enter a specific step of a derivation are noted ontop of the equation sign.
The experience is that the novice does not immediately memorize all the material covered and that
he is struggling with the math, so that he spends a lot of time finding the rationale behind a certain
step. This time is saved by being explicit about it. The danger that the student, gets dependent on
these indications, is probably minor, as it is still some effort for the advanced reader to look up the
assumptions, rather than memorizing the relevant material.
Important results and equations are included in boxes. This should facilitate the preparations for
examinations.
Portraits of the key researchers and short biographical notes provide independent associations
to the material. A student may not memorize a certain formula directly, but a portrait. From
the portrait, he may associate the correct formula. The historical context provides furthermore an
independent structure to organize the material.
The two first books are in german (That is the intended native language) in order to not add
complications to the novice. After these first books, all material is in English. It is mandatory that the
student masters this language. Most of the scientific literature is available only in English. English
is currently the language of science, and science is absolutely dependent on international contacts.
I tried to include many graphs and figures. The student shall become used to use all his senses
in particular the visual sense.
I have slightly modified the selection of the material. Some topics, which I consider of mostly
historical relevance I have removed. Others such as the Noether theorem, I have added. Some, like
Chaos, Stochastic processes, etc. I have not added yet.
331
332 W PHILOSOPHY OF THE ΦSX SERIES
Appendix X
About the Author
Prof. Dr. rer. nat Peter E. Blöchl studied physics at Karlsruhe University of Technology in Germany.
Subsequently he joined the Max Planck Institutes for Materials Research and for Solid state Research
in Stuttgart, where he worked on development of electronic-structure methods related to the LMTO
method and on first-principles investigations of interfaces. He received his doctoral degree in 1989
from the University of Stuttgart.
Following his graduation, he joined the renowned T.J. Watson Research Center in Yorktown
Heights, NY in the US on a World Trade Fellowship. In 1990 he accepted an offer from the IBM
Zurich Research Laboratory in Ruschlikon, Switzerland, which had just received two Nobel prices in
Physics (For the Scanning Tunneling Microscope in 1986 and for the High-Temperature Supercon-
ductivity in 1987). He spent the summer term 1995 as visiting professor at the Vienna University
of Technology in Austria, from where was awarded the habilitation in 1997. In 2000 he left the
IBM Research Laboratory after a 10-year period and accepted an offer to be professor for theoretical
physics at Clausthal University of Technology in Germany. Since 2003, Prof. Blöchl is member of
the Braunschweigische Wissenschaftliche Gesellschaft (Academy of Sciences).
The main thrust of Prof. Blöchl’s research is related to ab-initio simulations, that is the parameter-
free simulation of materials processes and molecular reactions based on quantum mechanics. He
developed the Projector Augmented Wave (PAW) method, one of the most widely used electronic
structure methods to date. Next to the research related to simulation methodology, his research
covers a wide area from biochemistry, solid state chemistry to solid state physics and materials
science. Prof. Bloechl contributed to 8 Patents and published about 100 research publications,
among others in well-known Journals such as “Nature”. The work of Prof. Blöchl has been cited over
10,000 times, and he has an H-index of 37.
333
334 X ABOUT THE AUTHOR
Bibliography
[1] L. de Broglie. Recherches sur la théorie des quanta. Ann. Phys.(Paris), 3:22, 1925.
[2] C. Davisson and L.H. Germer. Diffraction of electrons by a crystal of nickel. Phys. Rev., 30:705,
1927.
[3] G.P. Thomson and A. Reid. Difraction of cathode rays by a thin film. Nature, 119:890, 1927.
[4] C. Jönsson. Elektroneninterferenzen an mehreren künstlich hergestellten feinspalten. Zeitschr.
Phys., 161:454, 1961.
[5] C. Jönsson. Electron diffraction at multiple slits. Am. J. Phys., 42:4, 1974.
[6] C. Jönsson. Elektroneninterferenzen an mehreren künstlich hergestellten feinspalten. Zeitschr.
Phys. A, 161:454, 1961.
[7] A. Tonomura, J. Endo, T. Matsuda, T. Kawasaki, and H. Ezawa. Demonstration of single-
electron build-up of an interference pattern. Am. J. Phys., 57:117, 1989.
[8] E. Schrödinger. Quantisierung als eigenwertproblem (erste mitteilung). Ann. Phys.(Leipzig),
1926.
[9] E. Schrödinger. Quantisierung als eigenwertproblem (zweite mitteilung). Ann. Phys.(Leipzig),
1926.
[10] E. Schrödinger. über das verhältnis der heisenberg-born-jordanschen quantenmechanik zu der
meinen. Ann. Phys.(Leipzig), 80:437, 1926.
[11] E. Schrödinger. An undulatory theory of the mechanics of atoms and molecules. Phys. Rev.,
28:1049, 1926.
[12] P.J. Mohr and B.N. Taylor. The fundamental physical constants. Physics Today, 56:guide,
2003.
[13] P.J. Mohr and B.N. Taylor. Codata recommended values of the fundamental physical constants:
1989. Rev. Mod. Phys., 72:351, 2000.
[14] R. Courant. Ein allgemeiner satz zur theorie der eigenfunktionen selbstadjungierter differen-
tialausdrücke,. Nachr. v. d. Ges. d. Wiss.zu Göttingen, page 81, 1923.
[15] J. von Neumann. Allgemeine eigenwerttheorie hermitescher funktionaloperatoren. Math. Annal.,
102:49, 1929.
[16] J. von Neumann. Mathematische Grundlagen der Quantenmechanik, volume 38 of Grundlehren
der mathematischen Wissenschaften. Springer, 1932.
[17] John v. Neumann. Die eindeutigkeit der schrödingerschen operatoren. Math. Annal., 104:1432,
1931.
[18] Wolfgang Walter. Einführung in die Theorie der Distributionen. BI-Wiss.Verl., 1994.
335
336 BIBLIOGRAPHY
[19] W. Heisenberg. Über den anschaulichen inhalt der quantentheoretischen kinematik und
mechanik. Zeitschr. Phys., 43:172, 1927.
[20] P.A. Robertson. The uncertainty principle. Phys. Rev., 34:163, 1929.
[21] E. Schrödinger. Zum heisenbergschen unschärfeprinzip. Sitzungsberichte Preuss. Akad. d. Wiss.
Phys.-Math. Klasse., 19:296, 1930.
[22] W.H. Zurek. Decoherence and the transition from the quantum to classical. Physics Today,
44:36, 1991.
[23] P. Ehrenfest. Bemerkung b̈er die angenäherte gültigkeit der klassisehen mechanik innerhalb der
quantenmechanik. Zeitschr. Phys., 45:455, 1927.
[24] J.R. Chelikowsky and M.L. Cohen. Electronic structure of silicon. Phys. Rev. B, 10:5095, 1974.
[25] W.H. Press, S.A. Teukolsky, W.T. Vetterling, and Brian P. Flannery. Numerical Recipes: The
Art of Scientific Computing. Cambridge University Press, 1992.
[26] J.A. Gaunt. The triplets of helium. Phil. Trans. R. Soc. (London) A, 228:151, 1929.
[27] Walter Ritz. über eine neue methode zur lösung gewisser variationsprobleme der mathematischen
physik. J. Reine Angew. Math., 135:1, 2009.
[28] G. Wentzel. Eine verallgemeinerung der quantenbedingungen für die zwecke der wellenmechanik.
Zeitschr. Phys., 38:518, 1926.
[29] H.A. Kramers. Wellenmechanik und halbzahlige quantisierung. Zeitschr. Phys., 39:828, 1926.
[30] Lord Rayleigh. On the propagation of waves through a stratified medium, with special reference
to the question of reflection. Proc. Roy. Soc. A, 86:207, 1912.
[31] A. Unsöld. Ann. der Phys., 82:355, 1927.
[32] Matthias Jacobi and Hendrik Spahr. Das noethertheorem in der quantenmechanik und die
so(4)-symmertrie des wasserstoffatoms. Technical report, http://www.desy.de/~jlouis/
Vorlesungen/QMII06/vortrag_09.pdf, 2006.
[33] P. Wormer. Aanvulling syllabus "lie theorie voor natuurkundigen" van g.j. heckman (mathema-
tisch instituut ku nijmegen) door p.e.s. wormer (instituut voor theoretische chemie ku nijmegen).
Technical report, http://www.theochem.ru.nl/ pwormer/rungelenz.pdf, 2003.
Bibliography
[1] L.I. Schiff, Quantum Mechanics, (McGraw-Hill, New York, 1968).

[2] S. Gasiorewicz, Quantenphysik, (R. Oldenburg Verlag, München, 1977); S. Gasiorewicz, Quan-
tum Physics, (John Wiley, 1974)
[3] E. Merzbacher, Quantum mechanics, (John Wiley, New York)
[4] M.Alonso and E.J. Finn, Quantenphysik, (Addison Wesley, Bonn, 1988); M.Alonso and E.J.
Finn, Fundamental University Physics, Vol III, Quantum and Statistical Physics (Addison Wesley,
1968)
[5] A. Messiah and G.M. Temmer, Quantum Mechanics Vol. I+II,(North Holland, Amsterdam,
1964).
[6] C.Cohen-Tannoudji and F. Laloe, Quantenmechanik, (Walter de Gruyter, Berlin, 1999).
[7] P.W. Atkins,Molecular Quantum mechanics, (Oxford University Press, Oxford 1983).
[8] W. Nolting, Grundkurs Theoretische Physik 5,Quantenmechanik Teil 1+2, (Verlag
Zimmermann-Neufang, Ulmen 1992).
[9] C. Kiefer, Quantentheorie, (Fischer Taschenbuch Verlag, 2002)
[10] D.J. Griffiths,Introduction to Quantum mechanics, (Prentice Hall, New jersey, USA, 1994)
[11] J.-L. Basdevant and J. dalibard Quantum Mechanics, (Springer Verlag, Berlin 2002)
[12] Numerical Recipes: The Art of Scientific Computing, W.H. Press, B.P. Flanney, S.A. Teukolsky
and W.T. Vetterling, (Cambridge University press, Cambridge,1986). An online version can be
found at http://www.nr.com/.
[13] Bronstein, Semendjajew, Musiol, Mühlig,Taschenbuch der Mathematik, (Verlag Harri Deutsch,
Thun, 2001)
[14] ”Relativistic Quantum Mechanics”, J..D. Bjorken and S.D. Drell, Int’ Ser. in Pure and Applied
Physics, (McGraw-Hill, 1964) ”Relativistische Quantenmechanik” J..D. Bjorken and S.D. Drell,
Bibliographisches Institut, Mannheim.
[15] H. Haken, Quantenfeldtheorie des Festkörpers, (Teubner, Stuttgart 1973)
[16] J.D. Jackson, Classical Electrodynamics, (John Wiley, New York, 1998)
[17] Quantities, Units and Symbols in Physical Chemistry I. Mills et al. (Blackwell Scientific, Oxford
1993).
[18] http://www.codata.org/databases/databases.html
[19] M. Abramowitz and I.A. Stegun,Pocketbook of Mathematical Functions, (Verlag Harri Deutsch,
Thun, 1984)
337
338 BIBLIOGRAPHY
[20] K. Baumann und R.U. Sexl, Die Deutungen der Quantentheorie, (Vieweg) [Überblick Über die
Interpretationsprobleme der Quantenmechanik sowie 12 wichtige Originalarbeiten in deutscher
Sprache]
[21] J. Audretsch, verschränkte Welt (Wiley-VCH) [Allgemeinverständliche, auf Vorträgen beruhende
Beiträge verschiedener Autoren.]
[22] R. Omnes,Understanding Quantum Mechanics (Princeton University
Press)[Populärwissenschaftlich gehaltrenes Buch zur Iterpretation der Quantenmechanik,
das auch das Dekohärenzkonzept behandelt, aber im wesentlichen auf den Consistent Theories
Zugang abhebt.]
[23] D. Giulini et al. Decoherence and the Appearance of the Classical World in Quantum the-
ory.(Springer)[Umfassendes, durchaus technisches Werk zu allen Aspekten der Dekoherenz]
[24] L. Brillouin, Comptes Rendus 183, 24 (1926)
Index
abelian group, 83 constant of motion, 125

absorption correspondence principle, 38
stimulated, 214 creation operator, 94
action, 26 current density, 132
adjunct operator, 87
Airy’s equation, 277 d’Alambert operator, 33
Airy’s function, 277 de Broglie relations, 38
angular frequency, 31 decoherence, 122
angular momentum, 172 degeneracy, 117
eigenvalues, 179 degenerate, 51
annihilation operator, 94 differentiable, 75
anti-commutator, 94 differential operator, 33
associative law Dirac matrix, 283
addition of states, 83 dispersion, 35
multiplication of a state by a scalar, 83 dispersion relation, 30
distributive law
band structure, 48 multiplication of a state by a scalar, 84
band gap, 48 dyadic product, 91
bands, 48 dynamical matrix, 99
basis, 102 dynamical symmetry, 311
Bloch Theorem, 159
Bloch vector, 159 Ehrenfest Theorem, 127
Bohr magneton, 234 eigenstate, 51
Bohr radius, 195 eigenvalue, 51
Bohr’s atom model, 221 eigenvalue equation, 51
Bohr’s quantization condition, 221 eigenvalue spectrum, 51
Boson, 151 elastic constant, 32
bra, 82, 85 emission
bracket, 82, 85 stimulated, 214
ensemble of states, 122
canonical momentum, 28 Entanglement, 122
canonical conjugate, 89 entanglement, 272
canonical momentum, 88 envelope function, 34
Cauchy sequence, 84 Euler-Lagrange equations, 27
classical limit, 21 evanescent waves, 80
coherence, 272 expectation value, 113
coherent, 17, 123, 124 Expectationvalue, 43
Collapse of the wave function, 121 extended zone scheme, 162
commutative law
addition of states, 83 Fermi energy, 48
commutator, 88 Fermi’s golden rule, 214
compatible operators, 126 Fermion, 150
Completeness, 84 frequency
completeness, 85 and probability, 121
conduction band, 48 angular, 31
339
340 INDEX
function Lagrangian, 25
of an operator, 89 Lagrangian density, 32
functional, 26 Laguerre DGl
associated, 197
g-factor, 234, 282 Laguerre polynomials
gauge symmetry, 132 associated, 197
Gaunt coefficients, 188 Laplace-Runge-Lenz vector, 311
gaunt series, 188 Laplacian, 33
Gauss integral, 250 laser, 214
Gram-Schmidt orthogonalization, 126 lattice constant, 155
Grassmann algebra, 112 lattice vector
ground state, 215 general, 155
group velocity, 35 primitive, 155
gyromagnetic moment, 234 reciprocal, 155
least action
Hamilton formalism, 27 principle of, 25
Hamilton’s equations, 28 Legendre Transformation, 27
Hamilton’s equations of motion, 37 linear independency, 85
harmonic oscillator, 92 linear chain, 29
Hartree, 196 linear dependency, 85
Hartree atomic units, 196 linear independent
Heaviside function, 65 states, 85
Heisenberg operator, 133 Linearity
Heisenberg picture, 132 scalar product, 84
Heisenberg state, 133 logarithmic derivative, 73
Heisenberg’s uncertainty principle Lorentzian, 107
derivation, 118
harmonic oscillator, 96 mass
particle in a box, 61 effective, 48
Heisenberg’s uncertainty relation, 56 measurement theory, 21
hermitean symmetry, 84 metal, 48
hermitian, 43 moment, 251
hermitian conjugate, 87 momentum
Hermitian operator, 88 canonical, 28
Hilbert space, 84 kinetic, 28
Hilbert space, 81, 83 momentum
hydrogen atom canonical, 142, 144
eigenvalues, 199 momentum
canonical, 88
identity, 87 multiconfiguration wave function, 151
insulator, 48
Noether theorem, 27
intensity, 18
Norm, 85
interaction picture, 134
normalization, 85
interference, 17
normalization condition, 41
constructive, 17
destructive, 17 Observable, 41
interference term, 21 operator
inverse operator, 89 differential, 33
irregular solution, 198 optical absorption, 210
orthogonality, 85
ket, 81, 82 orthonormality, 85
kinetic equations, 269 outer product, 91
Klein-Gordon equation, 39
Knotensatz, 62 Parity, 147
INDEX 341
PAULI EQUATION, 234 Slater determinant, 151

Pauli matrices, 110 spectrum, 51
periodic, 155 speed of sound, 32
periodic image, 64 spherical harmonics, 180
perturbation theory spin, 184
stationary, 208 square integrable, 81
perturbation theory square variation, 114
time-dependent, 210 Stern-Gerlach term, 234
phonons, 29 strain, 32
photons, 39 stress, 32
Planck constant super symmetry, 111
reduced, 14, 37, 45 Symmetry, 138
Planck constant, 14, 45 symmetry
polarization, 209 dynamical, 311
positions symmetry operation, 138
relative, 155
positive definiteness time-independent Schrödinger equation, 51
scalar product, 84 time-ordering operator, 276
positron, 232 transition probability, 214
principal quantum number, 198 transmission coefficient, 70
propagator, 132 tunnel effect, 74, 77
Quantization uncertainty principle

first, 38 derivation, 118
quantum electrodynamics, 283 harmonic oscillator, 96
quantum field theory, 13 particle in a box, 61
quantum number, 126 unit cell
principal, 198 primitive, 155
quasi particles, 49 unitary operator, 89
quasiclassical approximation, 219 unity operator, 101
valence band, 48
radius of convergence, 90
variational principle, 215
Reciprocal lattice vectors, 156
Vector space
reduced zone scheme, 162
linear, 84
reflection coefficient, 70
regular solution, 198 wave function
representation, 102 interpretation, 41
Resonance, 79 Wick-rotation, 57
resonances, 75 WKB approximation, 219
Runge-Lenz vector, 311
Rydberg atomic units, 196 zone scheme
extended, 162
scanning tunneling microscopy, 74 reduced, 162
Schrödinger equation, 38
Schrödinger equation
time independent, 50
Schrödinger picture, 132
Schwarz’ inequality, 115
semiconductor, 48
Separation of variables, 53
separation of variables, method of, 51
sesquilinear form, 84
shift operator, 95, 142
shift operators, 175

QM PDF

Uploaded by

Copyright:

Available Formats

QM PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

QM PDF

Uploaded by

Copyright:

Available Formats

What are the two big revolutions in physics in the early 20th century according to the text?

What are the two big revolutions in physics in the early 20th century according to the text?

What does quantum theory introduce according to the text?

What does quantum theory introduce according to the text?

Theoretical Physics III

Caution! This version is not completed and may contain

Institute of Theoretical Physics; Clausthal University of Technology;

© Peter Blöchl, 2000-July 6, 2011

• Cohen-Tannoudji, Quantenmechanik[6]. In my opinion a very good and modern textbook with

3 Detailed citations are compiled at the end of this booklet.

1 Waves? Particles? Particle waves! 13

2 Experiment: the double slit 15

4.6 Potential step and splitting of wave packets . . . . . . . . . . . . . . . . . . . . . . 65

5 Language of quantum mechanics 81

7.7 Decoherence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

10 Specific Symmetries 147

11 Rotations, Angular Momentum and Spin 171

11.1.2 Another derivation of the angular momentum (Home study) . . . . . . . . . 172

13 Approximation techniques 203

14 Relativistic particles 229

15 Many Particles 239

15.1 Bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

A Galilei invariance 247

B Spreading wave packet of a free particle 249

C The one-dimensional rectangular barrier 253

D Particle scattering at a one-dimensional square well 259

E Alternative proof of Heisenberg’s uncertainty principle 263

F Spherical harmonics 267

G Random Phase approximation 269

H Propagator for time dependent Hamiltonian 275

I Matching of WKB solution at the classical turning point 277

K Method of separation of variables 287

L Trigonometric functions 289

M Gauss’ theorem 291

N Fourier transform 293

O Pauli Matrices 297

P Linear Algebra 299

Q Matrix identities 301

R Special Functions 305

S Principle of least action for fields 309

T ℓ-degeneracy of the hydrogen atom 311

U A small Dictionary 325

V Greek Alphabet 329

W Philosophy of the ΦSX Series 331

X About the Author 333

Waves? Particles? Particle waves!

Quantum mechanics is about unifying two concepts that we use to

• Waves are used to describe continuous distributions such as wa-

• On the other hand we use the concept of particles to describe the

• If we consider many particles, such as water molecules, we prefer

at least in an approximate manner. This is the realm of statistical

Experiment: the double slit

The essence of quantum mechanics becomes evident from a single

2.1 Macroscopic particles: playing golf

Fig. 2.1: Thomas Young,

that is the two distributions P1 (y ) and P2 (y ) simply add up to P12 (y ).

2.2 Macroscopic waves: water waves

Fig. 2.4: Spherical wave

The angular brackets denote the time average.