Heidelberg University

Institute for Theoretical Physics

version April 26, 2025

Quantum Physics
Script to the undergraduate lecture in spring 2022
Script to the graduate lecture in spring 2025
Maurits W. Haverkort
©(2020 / 2025) Maurits W. Haverkort All rights reserved
2
Contents

Preface 11

1 The limits of classical physics 15

1.1 Black body radiation . . . . . . . . . . . . . . . . . . . . . . 18
1.1.1 Density of modes . . . . . . . . . . . . . . . . . . . . 22
1.1.2 Energy density according to classical physics . . . . 24
1.1.3 Planck’s law . . . . . . . . . . . . . . . . . . . . . . . 25
1.2 The photoelectric effect . . . . . . . . . . . . . . . . . . . . 27
1.3 The third law of thermodynamics and Einsteins model of the
specific heat of solids . . . . . . . . . . . . . . . . . . . . . . 29
1.4 The Compton effect . . . . . . . . . . . . . . . . . . . . . . 33
1.5 Dual wave and particle nature of matter . . . . . . . . . . . 36
1.6 The Zeeman effect and the Stern-Gerlach Experiment . . . 37
1.7 Atomic physics . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.7.1 Experimental findings . . . . . . . . . . . . . . . . . 44
1.7.2 Theoretical models . . . . . . . . . . . . . . . . . . . 50

2 Schrödinger’s conjecture and the postulates of quantum

mechanics 55
2.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.2 The wave function and the Hamiltonian defining the micro
state and its time evolution . . . . . . . . . . . . . . . . . . 57
2.3 The Schrödinger equation for many particle states . . . . . 58

3
Contents

2.4 The interpretation of the wave-function . . . . . . . . . . . 60

3 Mathematical properties of the Schrödinger equation 67

3.1 Continuity equation, density conservation and current density 68
3.2 Linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3 Hermitian and Self-adjoint . . . . . . . . . . . . . . . . . . . 73
3.4 Time evolution operator . . . . . . . . . . . . . . . . . . . . 75
3.5 Unitarity of time evolution . . . . . . . . . . . . . . . . . . 75
3.6 Uniqueness of the time evolution and determinacy of the
wave function . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.7 Stationary states and Eigenfunctions . . . . . . . . . . . . . 78
3.8 Continuity of the wave function and its derivative . . . . . . 80

4 Wave mechanics 83
4.1 Time evolution of a delta function . . . . . . . . . . . . . . 83
4.2 Wave packets . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.2.1 The wave function in position and momentum space 91
4.2.2 Gaussian wave packet with finite momentum . . . . 92
4.2.3 Group and phase velocity . . . . . . . . . . . . . . . 94
4.2.4 Moment expansion in position and momentum space 97
4.2.5 Heisenberg’s uncertainty relation . . . . . . . . . . . 103

5 Wave mechanics in exemplary potentials 107

5.1 Potential step . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.2 Potential barrier . . . . . . . . . . . . . . . . . . . . . . . . 116
5.3 Delta function potential . . . . . . . . . . . . . . . . . . . . 119
5.3.1 Continuity of the wave function and its derivative . 120
5.3.2 The wave function for a delta function potential . . 121
5.4 Particle in a box . . . . . . . . . . . . . . . . . . . . . . . . 122
5.5 Potential well . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.5.1 Bound states in a potential well . . . . . . . . . . . . 127
5.6 Harmonic oscillator . . . . . . . . . . . . . . . . . . . . . . . 132
5.6.1 Stationary states of the harmonic oscillator solution 1 133
5.6.2 Stationary states of the harmonic oscillator solution 2 136

4
 Contents

5.6.3 Wave packets of the harmonic oscillator . . . . . . . 141

5.6.4 Coherent states . . . . . . . . . . . . . . . . . . . . . 142
5.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

6 Linear algebra 147

6.1 States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.1.1 Linearity and algebra with kets . . . . . . . . . . . . 149
6.1.2 Complex Vector space . . . . . . . . . . . . . . . . . 149
6.1.3 Ray and Vector . . . . . . . . . . . . . . . . . . . . . 151
6.1.4 Linear dependence . . . . . . . . . . . . . . . . . . . 151
6.1.5 Dimension of Vector space . . . . . . . . . . . . . . . 152
6.1.6 Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
6.1.7 Projection on a basis . . . . . . . . . . . . . . . . . . 152
6.2 Hilbert space . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.2.1 Bra vectors as dual representations of ket vectors . . 154
6.2.2 Inner product of a bra and ket vector . . . . . . . . 154
6.2.3 The norm of a state vector . . . . . . . . . . . . . . 155
6.2.4 Orthonormal basis . . . . . . . . . . . . . . . . . . . 155
6.2.5 Infinite dimensional vector spaces . . . . . . . . . . . 156
6.2.6 Hilbert space . . . . . . . . . . . . . . . . . . . . . . 157
6.2.7 Vector subspace . . . . . . . . . . . . . . . . . . . . 158
6.2.8 Tensor product of vector spaces . . . . . . . . . . . . 158
6.3 Relation between bra ket notation and wave functions . . . 159
6.4 Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
6.4.1 Properties of operators . . . . . . . . . . . . . . . . . 161
6.4.2 Multiplication and commutation . . . . . . . . . . . 163
6.4.3 Outer product of a ket and bra . . . . . . . . . . . . 164
6.4.4 Completeness relation . . . . . . . . . . . . . . . . . 164
6.4.5 Matrix representation . . . . . . . . . . . . . . . . . 165
6.4.6 Basis transformation . . . . . . . . . . . . . . . . . . 167
6.4.7 Eigenbasis representation of an operator . . . . . . . 169
6.4.8 Projection operators . . . . . . . . . . . . . . . . . . 173
6.5 Distributions and expectation values . . . . . . . . . . . . . 175
6.5.1 (partial) Density Of States . . . . . . . . . . . . . . 178

5
Contents

7 Symmetry and conservation laws 181

7.1 Symmetry and commutation relations . . . . . . . . . . . . 182
7.2 Symmetry and conservation laws . . . . . . . . . . . . . . . 184
7.2.1 Translational symmetry and momentum conservation 185
7.2.2 Time invariance and energy conservation . . . . . . 189
7.2.3 Conservation laws and distribution functions . . . . 192
7.3 Symmetry groups . . . . . . . . . . . . . . . . . . . . . . . . 193
7.4 Eigenfunctions and irreducible representations . . . . . . . . 195
7.5 Rotational symmetry and angular momentum conservation 200
7.5.1 Commutation relations of the angular momentum op-
erators . . . . . . . . . . . . . . . . . . . . . . . . . . 203
7.5.2 Raising and lowering operators for angular momentum204
7.5.3 Angular momentum operators in spherical coordinates 208
7.5.4 The eigenfunctions and eigenvalues of Lz . . . . . . 209
7.5.5 The spherical harmonics . . . . . . . . . . . . . . . . 210
7.5.6 Tesseral harmonics . . . . . . . . . . . . . . . . . . . 212
7.5.7 Addition of angular momentum . . . . . . . . . . . . 214
7.5.8 Rotation of total angular momentum states, Wigner
D matrices . . . . . . . . . . . . . . . . . . . . . . . 229
7.5.9 Product of two spherical harmonic functions . . . . 233
7.5.10 Expectation values between angular momentum states 234
7.5.11 Vector and Tensor operators . . . . . . . . . . . . . 238
7.5.12 Wigner-Eckart theorem and reduced matrix elements 243
7.5.13 Rotation group, SO(3) and complex rotation group
SU(2) . . . . . . . . . . . . . . . . . . . . . . . . . . 245

8 The Hydrogen atom 249

8.1 The central potential . . . . . . . . . . . . . . . . . . . . . . 249
8.2 Angular solution . . . . . . . . . . . . . . . . . . . . . . . . 253
8.3 Radial solution . . . . . . . . . . . . . . . . . . . . . . . . . 254
8.3.1 Bound states . . . . . . . . . . . . . . . . . . . . . . 256
8.3.2 Scattering states . . . . . . . . . . . . . . . . . . . . 259
8.4 Cusp of the wave function at the origin . . . . . . . . . . . 261
8.5 The periodic table . . . . . . . . . . . . . . . . . . . . . . . 262

6
 Contents

8.6 The introduction of electron spin . . . . . . . . . . . . . . . 265

8.6.1 The spin of an electron as half integer valued angular
momentum . . . . . . . . . . . . . . . . . . . . . . . 266
8.6.2 Magnetic moments introduced by angular momen-
tum and spin . . . . . . . . . . . . . . . . . . . . . . 269
8.6.3 Spin-orbit interaction . . . . . . . . . . . . . . . . . 271

9 N particle systems 279

9.1 The Schrödinger equation alone is not enough . . . . . . . . 279
9.1.1 Pauli’s principle . . . . . . . . . . . . . . . . . . . . 280
9.1.2 Second quantization . . . . . . . . . . . . . . . . . . 285
9.1.3 Bosons in second quantization . . . . . . . . . . . . 292
9.1.4 Wave functions beyond Single Slater determinants . 292
9.2 Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
9.3 Entanglement and correlations . . . . . . . . . . . . . . . . 295
9.4 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . 300
9.4.1 Classical statistical physics for the ideal gas . . . . . 301
9.4.2 Boltzmann statistics in quantum mechanics . . . . . 303
9.4.3 Reduced density matrix . . . . . . . . . . . . . . . . 305
9.4.4 Brownian motion, thermal noise and thermal fluctu-
ations . . . . . . . . . . . . . . . . . . . . . . . . . . 307

10 Perturbation theory 309

10.1 Time independent perturbation theory without degenerate
states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
10.1.1 Perturbation example on two-level system . . . . . . 319
10.2 Time independent perturbation theory with degenerate states 324
10.3 Time dependent perturbation theory . . . . . . . . . . . . . 325
10.3.1 Dyson series . . . . . . . . . . . . . . . . . . . . . . . 328
10.3.2 Feynman diagrams . . . . . . . . . . . . . . . . . . . 331
10.3.3 Fermi’s golden rule . . . . . . . . . . . . . . . . . . . 334

11 Interaction with electromagnetic fields 339

11.1 Electro magnetic fields in Hamilton formalism . . . . . . . . 341

7
Contents

11.2 Classical fields in the Schrödinger equation . . . . . . . . . 343

11.3 Gauge invariance . . . . . . . . . . . . . . . . . . . . . . . . 344
11.3.1 Gauge invariance as a local phase transformation . . 345
11.3.2 Coulomb Gauge . . . . . . . . . . . . . . . . . . . . 349
11.4 Quantized fields in the Schrödinger equation . . . . . . . . . 350
11.5 Zeeman effect . . . . . . . . . . . . . . . . . . . . . . . . . . 353

12 Relativistic effects 355

12.1 The Dirac equation . . . . . . . . . . . . . . . . . . . . . . . 356
12.2 Dirac equation with electromagnetic field . . . . . . . . . . 362
12.3 Properties of the Dirac α matrices and alternative notations 363
12.4 Solutions for a free particle . . . . . . . . . . . . . . . . . . 364
12.5 Negative energy states, holes and anti-particles . . . . . . . 366
12.6 The non-relativistic limit . . . . . . . . . . . . . . . . . . . 367
12.6.1 The non-relativistic limit in a spherical potential . . 373
12.6.2 Spin orbit coupling constant in atoms . . . . . . . . 375
12.7 Symmetries and conservation laws for the Dirac equation . 377
12.7.1 Space and time translation symmetry or Momentum
and Energy conservation . . . . . . . . . . . . . . . . 377
12.7.2 The continuity equation . . . . . . . . . . . . . . . . 379
12.7.3 Rotational symmetry or Angular momentum conser-
vation . . . . . . . . . . . . . . . . . . . . . . . . . . 382
12.7.4 Helicity of a particle . . . . . . . . . . . . . . . . . . 384
12.7.5 Relativistic Quantum number κ . . . . . . . . . . . . 385
12.8 Solutions for the hydrogen atom . . . . . . . . . . . . . . . 389
12.8.1 Spherical coordinates . . . . . . . . . . . . . . . . . . 390
12.8.2 Angular solutions . . . . . . . . . . . . . . . . . . . . 393
12.8.3 Operator αr . . . . . . . . . . . . . . . . . . . . . . . 401
12.8.4 Radial solutions . . . . . . . . . . . . . . . . . . . . 404
12.9 Lamb shift . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404

13 Response theory 405

13.1 Classical example . . . . . . . . . . . . . . . . . . . . . . . . 409
13.1.1 Real variables without friction . . . . . . . . . . . . 410

8
 Contents

13.1.2 Real variables with damping . . . . . . . . . . . . . 413

13.1.3 Complex variables with damping . . . . . . . . . . . 417
13.1.4 Linear response functions . . . . . . . . . . . . . . . 417
13.2 Quantum example . . . . . . . . . . . . . . . . . . . . . . . 423
13.3 General response derived in the frequency domain . . . . . 434
13.4 General response derived in the time domain . . . . . . . . 438
13.5 Relation between frequency and time domain . . . . . . . . 444
13.6 Dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . 446
13.7 Kramers Kronig relations . . . . . . . . . . . . . . . . . . . 449

A The Dirac delta function (distribution) 451

A.1 Properties of the Dirac delta function . . . . . . . . . . . . 453
A.1.1 Convolution . . . . . . . . . . . . . . . . . . . . . . . 453
A.1.2 Scaling . . . . . . . . . . . . . . . . . . . . . . . . . 453
A.1.3 Derivative . . . . . . . . . . . . . . . . . . . . . . . . 453

B Gaussian Integral 455

C Fourier Transform 459

C.1 Proof that the inverse Fourier transform is the inverse oper-
ator of the Fourier transform . . . . . . . . . . . . . . . . . 460
C.2 Properties of the Fourier transform . . . . . . . . . . . . . . 461
C.2.1 Linearity . . . . . . . . . . . . . . . . . . . . . . . . 461
C.2.2 Derivative . . . . . . . . . . . . . . . . . . . . . . . . 462
C.2.3 Convolution . . . . . . . . . . . . . . . . . . . . . . . 462
C.3 The Fourier transform of some useful functions . . . . . . . 463
C.3.1 Fourier transform of a delta function . . . . . . . . . 463
C.3.2 Fourier transform of a Gaussian function . . . . . . 464
C.3.3 Fourier transform of a constant . . . . . . . . . . . . 464
C.3.4 Fourier transform of a polynomial . . . . . . . . . . 465

D Cauchy-Schwarz inequality 467

E Fundamental Theorem of Algebra 469

9
Contents

F The eigensystem of a Matrix 471

G Euler angles 473

10
Preface

This script is written for the lecture Quantum Physics (Theoretical Physics
4) taught in the Summer semester 2020 and 2022 at Heidelberg University.
There are many good textbooks available on the theory of Quantum
Physics [1, 2, 3, 4, 5] and this script is not meant to replace any of those.
Based on these textbooks, and several scripts of lectures taught in different
years, I have created a set of lectures to introduce the theory of Quantum
Physics. Reading these notes should be accompanied by visiting the lectures
and by solving problems. I can not stress enough how important it is to
practice by solving problems yourselves. Ten years from now it will not
matter much if you can remember an equation. It will matter a lot if you
can solve a practical problem encountered in your profession or research.
You only learn this by practicing a lot.
We start our lecture with a few experimental examples showing the
limits of classical physics. At the end of the 19th century there were multiple
experiments showing the limits of classical physics. Understanding these
experiments motivates the need to introduce the quantization of energy,
the correspondence between particles and waves, and the need of a linear
equation to describe the quantum state of matter. This script introduces
several of these experiments. In the lecture we will only go into depth on
one or two examples as I assume that you are familiar with most of them.
In the next two chapters I will introduce the Schrödinger equation. Just
as Newton did not derive his laws of mechanics, Schrödinger by no means
derived his equation. We will give a reasonable background (although not
historically correct) on how one might come to the idea to propose this

11
Contents

equation. In his introductory paper, Schrödinger writes that he proposes

this equation to explain the discrete line spectrum of radiation emitted
and absorbed by atoms. The reason we still use his original conjecture
(equation) is that it describes many physical phenomena really well. With
the appropriate relativistic (Dirac equation and Quantum electron dynam-
ics) corrections this is one of the most accurately and stringently tested
equations we know.
One of the open problems for the Schrödinger equation is the interpre-
tation of the wave-function. The wikipedia site on the interpretations of
quantum mechanics writes: “Although quantum mechanics has held up to
rigorous and extremely precise tests in an extraordinarily broad range of
experiments (not one prediction from quantum mechanics is found to be
contradicted by experiments), there exist a number of contending schools
of thought over their interpretation.” [6]. Wikipedia might not be the
leading authority on the field of philosophy of science, but it does give
an impression on how people in the field currently think about this topic.
Within this script we aim to be agnostic with respect to the different in-
terpretations of quantum mechanics. In section 2.4 we touch upon a few
concepts that are common to all interpretations. This section contains sev-
eral references to later chapters in the script. For a full discussion of the
different interpretations of quantum mechanics one should have a profound
understanding of the mathematical framework and experimental findings
that underly quantum mechanics. Section 2.4 provides a minimal under-
standing of the relation between the equations of quantum mechanics and
experimental observations, needed to discuss these equations effectively.
For a more thorough understanding, one can always return to this section
once the later chapters are studied.
After the introduction of the Schrödinger equation we will first discuss
solutions of the Schrödinger equation from a wave mechanics point of view.
This description solves the wave equations by solving differential equations.
The mathematics involved here are very similar to the solutions found for
electro magnetic waves in Maxwell’s equations or the propagation of sound
waves, water waves, or seismic waves, each governed by their own wave
equation and energy momentum dispersion relations. Even the propagation

12
 Contents

of heat through a medium leads, following the original work of Fourier,

to similar equations. Once we’ve seen a few solutions of the Schrödinger
equation using wave mechanics we will introduce the concepts of linear
algebra. This relates the Schrödinger equation [7, 8, 9, 10, 11, 12] to the
matrix mechanics as introduced by Heisenberg, Born and Jordan [13, 14,
15].
With the mathematical framework of linear algebra we will introduce
conservation laws based on symmetry. Only then we will look at the so-
lution for the hydrogen atom. The physical observation on the absorption
and emission spectrum of atoms was one of the leading motivations for
Schrödinger and Heisenberg to search for a new theory.
After we discuss the hydrogen atom in the lecture there are several
important and beautiful topics left. We will not be able to treat them all
and need to make a selection. There will be many occasions where you can
study each of these topics in more detail.
The appendices deal with mathematical concepts we need during this
lecture. These are really important and deserve your attention. They are
not in the appendix because they are not important. They are in the
appendix as I did not want to disturb the flow of physical reasoning too
much by the need to introduce mathematical methods.
Several of the references point towards original literature. For under-
standing the equations and becoming a good physicist there is no need to
read these. If you are interested in the history of physics these are nice to
read. Last but not least I believe one should cite original literature instead
of reviews of others as much as possible.

Maurits W. Haverkort (2022)

13
Contents

14
Chapter 1

The limits of classical

physics

In order to understand why we define the equations of quantum physics as

we do, it is useful to look at the experimental limits of classical physics.
We will do this partially using the historical framework of the physics of
the early 20th century, as available to the scientists who developed the
theories of quantum physics. This chapter gives by no means a complete
historical overview. I picked several topics not because they were the most
prevailing at that time, but because they allow me to show the need for
the introduction of a new theory. Several experiments that show the limits
of classical physics discus the relation between particles and waves. The
Zeeman effect or the Stern Gerlach experiment show the limits of classical
physics with a different quality. These experiments emphasize that we not
only need to quantize energy, but that the resulting quantum particles
also need to be described by a linear theory. A superposition of two valid
solutions of the quantum equations of motion needs to be a valid solution
on its own.
At the end of the 19th century and first quarter of the 20th century
a revolution in physics took place. Several experiments were found that
contradicted the classical view of the world. At the end of the 19th cen-

15
Chapter 1. The limits of classical physics

tury, science and physics had seen many years of steady progress. People
understood many physical phenomena and had equations with predictive
power. After many years of progress in the understanding of the natural
world many people had great confidence in the laws of physics. Based on
empirical findings the classical laws of mechanics, electro magnetism and
thermodynamics were formed. A few equations describe the motion of small
objects on earth, the movements of the planets, electro magnetic phenom-
ena, the transport of heat or the expansion of a gas when heated. These
findings were not only philosophical in nature. Being able to predict the po-
sition of the planets and their moons allowed one to better navigate at sea.
Newton’s laws had a profound influence on the development of structural
engineering. The understanding of thermodynamics allowed one to opti-
mize steam engines. The understanding of electro-magnetism contributed
to the development of electrical lighting. The improved understanding of
the natural sciences was a key ingredient to the industrial revolution and
enormously changed our way of life.
For many people in the late 19th century it was therefore hard to be-
lieve that anything could be wrong with classical physics, condensed in the
laws by Newton, Maxwell and thermodynamics. Some even went so-far as
to claim that physics was finished and one only needs to fill in the details.
It might have been shocking for many people to find more and more exper-
iments that seemed to indicate that something is wrong with the classical
picture. Many of these experiments are not related to our daily observation.
This might be a reason why these contradictions of classical physics have
gone unnoticed for so long. Classically, motion is continuous and particles
can have any velocity, mass or energy as desired. The experiments listed in
this chapter challenged that picture in such fundamental ways that people
started to look at alternative theories.
During the first quarter of the 20th century, several theories of Quantum
physics emerged. When Max Planck presented his Nobel lecture in 1920, he
did this with the title ”The genesis and present state of development of the
quantum theory” [16]. He was less modest in the translation of his original
speech, published in 1922 with the title ”The origin and development of the
quantum theory” [17]. Max Planck speaks about the quantum theory four

16
years before Schrödinger published his version of quantum theory we use to-
day. There were several theories of quantum physics produced between 1850
and 1926 [18]. Many contain some truth. A partial breakthrough came with
the matrix mechanics introduced by Heisenberg in 1925 [13]. Heisenberg
builds on previous experimental findings and theoretical considerations. He
specifically focused on the photon absorption spectra of atoms in magnetic
(Zeeman effect) and electric (Stark effect) fields. Based on these experi-
ments, although he never mentioned it in these words, Heisenberg basically
came to the conclusion that quantum states must be described by a com-
plex vector in a Hilbert space, following all rules of linear algebra. This
theory was rapidly developed further by Born, Heisenberg, and Jordan in
Göttingen [14, 15].
In his first publication, especially in the last paragraph, Heisenberg is
rather careful if his theory is sufficiently well developed to be the final the-
ory of quantum mechanics. As we now know it is not. Within a year,
Schrödinger proposed a differential equation to describe quantum mechan-
ics. Having a differential equation that determines the motion of particles
placed the understanding of quantum mechanics on the same footing as clas-
sical mechanics or electro dynamics. With Schrödingers equation, quantum
physics quickly became a well accepted theory. As Schrödingers differential
equation is linear, we can describe its solutions as complex vectors on a
Hilbert space that follow all rules of linear algebra. Schrödinger’s equa-
tion is thus in accordance with Heisenberg’s theory, but now also provides
explicit representations of the matrices and vectors present in Heisenbergs
equations. A dispute between Schrödinger and Heisenberg on how to inter-
pret the equations of quantum mechanics and which physical picture they
might represent remained till the death of Schrödinger.
In this chapter, I will provide a few examples of the experiments that
lead to the introduction of quantum physics. They are all beautiful by
themselves. Understanding these experiments in detail helps to develop
an intuitive feeling of quantum mechanics. We all have seen a ball fall as
described by the laws of classical mechanics. We have less experience with
observing single electrons scattering off the surface of a solid. Not having
that experience makes quantum mechanics less intuitive. In the end, the

17
Chapter 1. The limits of classical physics

equations and mathematical framework of quantum mechanics are not that

much harder than those of classical mechanics.

1.1 Black body radiation

All objects emit electromagnetic ra-
diation (light). The colour of an ob-
ject, i.e. the amount of light emit- Solid at
temperature T
ted as a function of wave length, Cavity with
depends on the temperature of the Radiation
object. These observations of the
nature of materials were very well Opening behaves as
a black body
know to the people of the 19th cen-
tury. If you burn coal, the colour
of the flame tells you how hot the Figure 1.1: A black body imple-
fire is. From cold to hot one sees mented by a small hole in a cavity
the colour of burning coal changing inside a solid.
from red to yellow to white to blue.
Glowing ashes are a dark red, a burning flame red yellow, if one adds oxygen
to the flame it turns white and ultimately blue. These relations between
colour and temperature are important if you run a steam engine, or if you
need to heat iron ore to produce steel.
In order to quantify this effect, Kirchhoff defined the concept of a Black
body and its radiation [19]. A black body is a solid that has no reflection
or transmission of light and absorbs all light that falls on it. Many black
insulating materials are, within a given frequency range, relatively good
absorbers of light. Normal materials do not represent a true black body
for all frequencies. A nearly perfect approximation of a black body is given
by a small opening of a cavity surrounded by a solid at a homogeneous
temperature as depicted in figure 1.1.
Kirchhoff noticed that such an object, in thermal equilibrium with its
surrounding, radiates the same amount of energy per second as it absorbs.
This is not only true for the total radiation, but this is true for each fre-

18
 1.1. Black body radiation

quency of the emitted and absorbed light. Kirchhoff derived this relation
by considering two of such bodies in thermal equilibrium in vacuum. If one
of them would absorb more than it radiates, its temperature would rise, si-
multaneously cooling the other object. This would violate the observations
of thermodynamics, from which we know that all objects strive towards
thermal equilibrium.
Using geometrical arguments, one can relate the amount of radiation
passing through a small opening of a cavity to the energy density in the
cavity.

4
u(λ, T ) = E(λ, T ) (1.1)
c
with E(λ, T ) the emissive power defined as the energy emitted at wave-
length λ per unit area, per unit time and u(λ, T ) the energy density inside
the cavity per unit of wavelength. Instead of working with the energy den-
sity as a function of wavelength, one can work with the energy density per
unit of frequency as a function of frequency. With λ = c/ν we have

dλ
u(ν, T ) = u(λ, T ) (1.2)
dν
c
= u(λ, T ).
ν2
At the end of the 19th century, several experiments were done to quan-
tify the radiation of a black body. Theoretical considerations were only
partially successful at explaining these observations.
Stefan observed that the total energy radiated by a black body and
thus the total energy density inside a cavity scales with T 4 [20]. Soon after
Stefan’s publication Boltzmann [21] explained this behaviour. He assumed
the second law of thermodynamics and the relation between the photon
pressure and energy inside the cavity as deduced by Maxwell. We thus find
that
Z ∞
4
u(ν, T )dν = σT 4 , (1.3)
0 c

19
Chapter 1. The limits of classical physics

2k4 π 5
with σ = 15cB2 h3 the Stefan-Boltzmann constant. The relation between the
Stefan-Boltzmann constant and the Boltzmann constant kB , the Planck
constant h, and the speed of light c was only derived later.
Wien considered what would
happen if the walls of a cavity con- 40 40
Planck
30 30 Wien

(λ2 /c) u(λ,T)

taining black body radiation would Rayleigh

u(ν,T)
20 20

move [22]. This would induce a 10 10

Doppler shift of the light absorbed 0

0 2 4 6 8 10
0
0 1 2 3 4

and emitted. Based on these argu- ν λ

ments, he showed that the energy

density inside the cavity should be Figure 1.2: Energy density inside a
of the form cavity according to Planck’s, Wien’s,
and Rayleigh-Jeans’ law. Left as a
3
u(ν, T ) = ν g(ν/T ), (1.4) function of frequency, right as a func-
tion of wavelength.
with g an unknown function that
only depends on a single parameter. Indeed, a plot of the experimental radi-
ation density inside a cavity divided by the temperature cubed u(ν, T )/T 3 ,
as a function of ν/T shows the same functional dependence independent
of the temperature or frequency at which the black body radiation energy
density is measured. Wien proposed the function

g(ν/T ) = αe−βν/T , (1.5)

with α and β fit parameters to the experiment. For large frequencies (ν) or
small wave lengths (λ) this approximates the experimental black body radi-
ation well. For small frequencies or large wave lengths there is a substantial
deviation between the theory of Wien and the experimental data.
Rayleigh and Jeans [23, 24, 25] derived the black body radiation as a
function of frequency using very general arguments of statistical physics,
classical mechanics and classical electro dynamics. They found, as we will
discuss in more detail below
8π kB T
u(ν, T ) = ν 3 . (1.6)
c3 ν

20
 1.1. Black body radiation

For low frequencies or large wavelengths, this equation agrees well with
experiment. For large frequencies or short wavelengths it does not agree
with experiment. Another major problem with Rayleigh-Jeans’ formula
is that the total energy density and thus the total emitted black body
radiation diverges
Z ∞
8π kB T
ν3 3 dν = ∞, (1.7)
0 c ν
not in agreement with the Stefan-Boltzmann equation. This is a major
problem as there is nothing wrong with the argumentation of Rayleigh and
Jeans. If one assumes the laws of classical mechanics, electrodynamics, and
statistical physics, the Rayleigh Jeans formula for black body radiation is
the result.
Max Planck published in 1900 a set of papers resolving the discrepancy
between the two theoretical models for black body radiation [26, 27, 28, 29,
30]. Planck’s work does not directly cite the theoretical work of Rayleigh
or Wien, but looks at the experimental data, which can be fitted to a
function of the form derived by Rayleigh at low frequencies of the light
and to function of the form derived by Wien at high frequency of the light.
In order to build a logical theory that can be derived from simple physical
principles, Planck had to assume that energy is quantized. For photons with
frequency ν the total energy of a ray can only occur in integer multiples of
hν. The physical consequences of this quantization of energy only became
clear during the next years and the work of Albert Einstein. Max Planck did
not start with the assumption that energy needs to be quantized. There is
no logical ground for such an assumption in classical physics. The first thing
Planck did is to interpolate the functions of Wien and Rayleigh. Wien’s
function for the energy density inside a cavity worked for high frequencies,
Rayleigh’s function for low frequencies. Planck found the following function

8πhν 3 1
u(ν, T ) = , (1.8)
c3 e khν
BT − 1

with h and kB the Planck and Boltzmann constant that needed to be fit

21
Chapter 1. The limits of classical physics

to the experiment. This equation described the experimental results very

accurately over the complete frequency range.
In the next part of this section, we will derive this result and show
how this leads us to assume that the energy of photons is quantized.
We will use the same argumentation to derive the Rayleigh-Jeans’ law
as we do for Planck’s law. The difference between the two theories is
minute. For the latter we will quantize the photon energy, for the for-
mer we do not. In order to derive the energy density inside a cavity,
we will first consider the number of electromagnetic modes that are al-
lowed inside the cavity. These can be counted as a function of their fre-
quency. Once we know the density of modes as a function of frequency,
we can consider the energy present in each mode as a function of tem-
perature T . For the latter, we will use the concept of equipartition of
energy from statistical physics and the fact that each mode has a kinetic
energy of 1/2kB T . The latter is the result of the observation that in ther-
mal equilibrium energy is shared equally among all of its various forms.

1.1.1 Density of modes

λ/2 = L
We will start the derivation of both
n=1
Planck’s and Rayleigh’s law for
black body radiation by calculating
the density of modes available in- 2 λ/2 = L
side a cavity. In order to do so,
n=2
we will first calculate the number of
modes N present with a wave vec-
3 λ/2 = L
tor k = 2π/λ = 2πν/c smaller than
kN and then take the derivative of n=3
N with respect to kN to find the
density of modes at k. For a square
cavity with volume V and edges L
such that V = L3 we do not allow Figure 1.3: Allowed wavelengths in a
all possible modes. If we take the cavity of length L.

22
 1.1. Black body radiation

boundary of the cavity to be per-

fectly absorbing, the electric field at
the boundary needs to be zero. The
allowed wave lengths perpendicular to one of the sides of the cavity is

nλ = 2L, (1.9)

with n ∈ N+ (See figure 1.3). The allowed wave vectors ~k are

~k = ( nx π , ny π , nz π ), (1.10)
L L L
with ni ∈ N+ .
In order to calculate the density
of modes, we consider the number
of modes with a wave vector smaller L ky
π
than kN . We can consider a sphere 8
in k space of radius kN and deter- 7
mine how many modes fit into this 6
sphere. The spacing between two 5
kN
modes is π/L in the x, y and z 4
direction. The volume per mode 3
is thus (π/L)3 . The volume of a 2
sphere with radius kN in k-space 1
is 4/3πkN 3 . The number of modes
1 2 3 4 5 6 7 8 L kx
with |~k| < kN is thus π

1 4 πk 3
N (kN ) = 2 3
N3 (1.11)
8 π Figure 1.4: Number of modes with
L
3
V kN wave vector smaller than kN
= .
3π 2
The factor of 1/8 comes from the
fact that we only consider modes
with positive wave-length. The factor of 2 comes from the fact that photons

23
Chapter 1. The limits of classical physics

can be polarised in two ways perpendicular to the k vector. We further-

more replaced L3 by the volume of the cavity. This equation seems like an
approximation, as modes with wave vectors |~k| ≈ kN can be just inside or
outside the sphere. In order to fix this, we can take the limit of L → ∞,
such that the spacing between modes in k space tends to zero. As we are
interested in the energy density, i.e. total energy per volume, it turns out
that there is no dependence on L and taking this limit is trivial. If we look
at the number of modes per volume n = N/V per wave vector we have

dn(k) 1
= 2 k2 . (1.12)
dk π
ck
With ν = 2π we can express the density of modes per frequency as

dn(ν) dn(k) dk
= (1.13)
dν dk dν
1 2πν 2 2π
 
= 2
π c c
8πν 2
= 3 .
c

1.1.2 Energy density according to classical physics

We can now derive the energy density according to the Rayleigh-Jeans’ law
by multiplying the density of modes with the average energy per mode.
In order to determine the average energy per electromagnetic mode, we
assume that the electromagnetic radiation is in thermal equilibrium with
the cavity. In thermal equilibrium, each degree of freedom has an average
kinetic energy of 12 kB T . Rayleigh, Jeans and Planck assumed the photon
to be the oscillation of something (the aether). The oscillation of this
something is harmonic. For an harmonic oscillation, the energy oscillates
between kinetic and potential energy. An oscillator with a kinetic energy
of 12 kB T has also a potential energy of 21 kB T such that the total average
energy we classically assign per photon mode is hEi = kB T . The average
denoted by brackets hi is over time. Each mode, i.e. each set of photons

24
 1.1. Black body radiation

with frequency ν, is in thermal equilibrium with all other modes as well as

the encasement of the cavity. Energy fluctuates between these modes. If
we focus on a single mode the energy is not constant over time, but changes
erratically. Only the time averaged energy of each mode is equal to kB T .
Knowing the average energy per mode as well as the number of modes
per frequency in the cavity, we can derive the classical energy density of
black body radiation inside a cavity by multiplying these

dn(ν)
u(ν, T ) = hEi (1.14)
dν
8πν 2
= 3 kB T.
c

The results we obtain with the arguments of classical physics lead to

formulas for the intensity of black body radiation that agree with experi-
ment for small frequencies ν, but not for large. Even worse, the total energy
flux of black body radiation according to classical physics diverges.

1.1.3 Planck’s law

We can derive Planck’s law by assuming that a mode of electromagnetic

radiation cannot contain any possible amount of energy, but that the energy
contained in each mode needs to be an integer multiple of hν. The density
of modes dn(ν)
dν remains the same as for the derivation of the Rayleigh-Jeans
equation. In Planck’s theory, we change the average energy assigned to each
mode. In order to derive Planck’s law, we need to calculate the average
energy of a mode under the assumption that the energy is quantized. The
mode can be in different states n with energy En = nhν. For this, we can
use Boltzmann statistics. With the use of statistical physics Boltzmann
derived the probability to find a system in state n with energy En as

1 − kEnT
pn = e B , (1.15)
Z

25
Chapter 1. The limits of classical physics

with Z the partition function needed to normalize the probability summed

over all states to one
∞
− kEnT
X
Z= e B . (1.16)
n=0

For the time averaged energy of an electromagnetic mode with frequency ν

we have
n=∞
X
hEi = En pn (1.17)
n=0
nhν
P∞ −
n=0 nhνe
kB T
= P∞ − k T .
nhν

n=0 e
B

We can simplify equation 1.17 by realizing that the denominator is given

P n − hν
x and the numerator as hν nxn with x = e kB T . The sum
P P n
as x
represents a geometric series and as 0 < x < 1 we have
∞
X 1
xn = . (1.18)
1−x
n=0

We find the numerator by differentiating with respect to x.

∞ ∞
X d X n
nxn = x x (1.19)
dx
n=0 n=0
d 1
=x
dx 1 − x
x
= .
(1 − x)2

We find for the time averaged energy of an electromagnetic mode

hνx
hEi = (1.20)
1−x

26
 1.2. The photoelectric effect

hν
= (1.21)
x−1 −1
hν
= hν . (1.22)
e kB T − 1
With this average energy per mode, the energy density inside the cavity
becomes
dn(ν)
u(ν, T ) = hEi (1.23)
dν
8πν 2 hν
= 3 hν ,
c e kB T − 1

such that we find for the emissive power (the energy emitted at frequency
ν per unit area per unit time) of a black body

2πν 2 hν
E(ν, T ) = 2 hν . (1.24)
c e kB T − 1

The Planck equation 1.24 interpolates between the Rayleigh-Jeans’ equa-

tion at small frequencies and the Wien equation at large frequencies. The
integral over frequency scales as T 4 confirming Steffan’s observation and
Boltzmann’s derivation of the total radiation energy per unit area per unit
time. The Planck equation fits experiment really well. The derivation of
Planck’s equation requires one to assume that the energy of electromagnetic
waves is quantized in portions of hν.

1.2 The photoelectric effect

Although Planck introduced the quantization of the electromagnetic field,
and with that the photon as a fundamental particle, he did not oversee
at that time the physical implications of his results. In order to see how
fundamental the quantization of a light wave in portions of hν is one can
compare the way one describes light and matter.

27
Chapter 1. The limits of classical physics

In 1905 and 1906 Einstein published two papers on the photo-electric

effect [31, 32]. In 1921 he was rewarded the Nobel prize ”for his services to
Theoretical Physics, and especially for his discovery of the law of the pho-
toelectric effect” which is laid out in these papers. His first paper on the
photo electric effect starts with a philosophical text on the way solids, flu-
ids and gasses are described compared to the description of electromagnetic
radiation, i.e. light. Macroscopically we describe gasses as objects with a
local particle, and current density. The particle and current density of a
gas are continuous functions of position and time. At the same time, for
matter the macroscopic quantities are time averages over the microscopic
behaviour which is governed by small particles, i.g. atoms and molecules.
In 1905, light, or electromagnetic fields, were macroscopically described by
a continuous field, similar to the macroscopic description of matter. Each
point in space and time can be assigned a density and flux, related to the
norm square of the electromagnetic field described by the Maxwell equa-
tions. Microscopically, the physics of 1905 assumed that one could zoom
in infinitely far and the electromagnetic field would always be a continuous
function. The classical description of light is thus fundamentally different
from the description of matter.
After pointing out this discrepancy in thinking between matter and
light, Einstein proposed to interpret the quantization of the light field in
quanta of hν as independent light particles. Einstein called these particles
light quanta. Only in 1926 the term photon can be found in the liter-
ature [33], which then gradually became accepted. Quantum mechanics
thus brings a unified picture for the description of both matter and light.
Macroscopically both are described by continues fields. Microscopically
both matter and light consist of particles. Einstein proposed a way to
test his hypotheses. It was known from earlier works [34] that light, when
focused on a solid, can lead to the emission of electrons.
Einstein proposed that light is made of photons (light quanta) and each
photon can excite a single electron. The energy contained in a photon is
hν. When this photon is absorbed, the energy is transferred to the electron
that is emitted. The energy of the photon is not converted completely to
the kinetic energy of the emitted electron. The electron is bound in a solid

28
1.3. The third law of thermodynamics and Einsteins model of
the specific heat of solids

by the positive charge of the ion. In a solid we call the minimal energy
difference between a bound electron and a free electron with zero kinetic
energy the work function W . The maximal kinetic energy of a photon-
electron is thus

Ek = hν − W (1.25)

From this equation we can draw a few conclusions. If the frequency of the
light is such that the energy of each photon hν is smaller than the work
function W , there will be no photon current. The maximal kinetic energy
of the photo electrons is linearly proportional to the frequency of the light.
The slope of Ek as a function of ν is equal to Planck’s constant h. The
intensity of the photon current (number of electrons produced) does not
depend on the frequency (ν), but on the intensity of the light.
These predictions were confirmed in 1914 by Millikan [35]. In 1921,
Einstein got the Nobel prize for laying one of the most fundamental corner
stones of quantum physics. All objects, both matter and light can macro-
scopically be described by a continuous density. Microscopically, one can
not divide them in infinitely small parts but needs to consider their particle
nature.

1.3 The third law of thermodynamics and Ein-

steins model of the specific heat of solids

Besides the discrepancy of black body radiation with the laws of classical
mechanics and statistical physics, also the third law of thermodynamics
conflicts with the notions of classical physics. In 1905 Nernst proposed,
based on experimental considerations, that the change in entropy of a sys-
tem (S) as a function of temperature (T ) is zero if cooled to zero Kelvin

29
Chapter 1. The limits of classical physics

[36] 1 .
dS
= 0. (1.26)
dT T =0

Instead of looking at the entropy of a system, we can consider the heat

capacity at constant volume CV . The heat capacity is defined as the amount
of heat (Q) needed to raise the temperature of the system
dQ
CV = . (1.27)
dT
The energy added to the system can either be distributed to internal degrees
of freedom and raise the internal energy U or provide work. At constant
volume, the system does not provide work such that the heat added to
the system is equal to the change in internal energy dQ = dU . From the
second law of thermodynamics, we know, that there is a relation between
the change in entropy of a closed system at constant volume and the heat
added to the system (dQ)

dQ = T dS. (1.28)

We thus find for the relation between the heat capacity and change of
entropy as a function of temperature
dS
CV = T . (1.29)
dT
From the third law of thermodynamics it follows that the heat capacity, at
constant volume of a system, must go to zero faster than T . Experimentally
the heat capacity indeed follows this rule.
In order to see why this conflicts with the theories of classical mechanics
and statistical physics we can use the results of the equipartition theorem
1
Nernst wrote a series of papers and talks in which he proposed this formula. In his
first papers he did not use the term entropy, but focused on possible useful work and
heat produced in different chemical reactions. Although historically not quite correct, we
here use the modern notation and definition of the third law of thermodynamics using
entropy.

30
 1.3. The third law of thermodynamics ...

to calculate the specific heat. We assume a system of N particles in ther-

mal equilibrium with a heat bath. For our system, the particle number
is conserved, but the total energy and distribution of the energy between
the particles changes. The micro-state of our classical system is uniquely
determined by the N position and momentum coordinates. In 3 dimen-
sional space this yields 6N degrees of freedom (independent coordinates)
that quantify the micro state.
Due to the exchange with a heat bath the actual micro state realized,
i.e. the position and momentum of the particles, changes in time. If we
assume that the probability to find a micro-state with energy E to be
present at time t is solely a function of E and the temperature T , then the
equipartition theorem states that
 
∂H
xm = δm,n kB T, (1.30)
∂xn
 
∂H
pm = δm,n kB T,
∂pn
with xn or pn one of the 6N degrees of freedom (either position x or mo-
mentum p) and H the classical Hamiltonian Rt of the system. The brackets
h. . .i denote the time average limt→∞ 1t 0 . . . dt.
We can use this result to calculate the heat capacity of N particles as
a function of temperature. We will first focus on the heat capacity due to
the kinetic energy. In a next step we generalize this result to the full heat
capacity.
(n)
The kinetic energy per particle and spatial direction n is given as Hkin =
p2n
2m such that
∂ p2n
   2 
p
pn = 2 n (1.31)
∂pn 2m 2m
D E
(n)
= 2 Ekin ,
or
D E 1
(n)
Ekin = kB T. (1.32)
2

31
Chapter 1. The limits of classical physics

With 3 spatial directions and N particles, we find that the total kinetic
energy is equal to
D E 3
(N )
Ekin = N kB T. (1.33)
2
The heat capacity due to the kinetic energy of N particles becomes

3
C V = N kB . (1.34)
2
For ideal gasses at large enough temperatures, the heat capacity is in-
deed constant. In this regime the heat capacity can be used to determine
the Boltzmann constant. At low temperatures the heat capacity of all ob-
jects experimentally measured goes to zero. The experiment is in agreement
with the third law of thermodynamics. One might argue that in solids or
at low temperatures things are more complicated as atoms interact with
each other. One could try to add interaction terms to the Hamiltonian
to compensate for the heat capacity due to the kinetic energy. There is,
however, no easy way to do this, nor is it logical why interactions between
particles should have this feature for all systems encountered in nature.
Einstein, charmed by the idea of discretizing the available energy, pro-
posed that also for the specific heat of solids one should not allow all values
of the kinetic energy, but assume that the energy comes in multiples of
some basic quantity, E0 [37]. All forms of energy are quantized.
Once we quantized the energy, we can use the same derivation we used
for the derivation of Planck’s law to calculate the total energy of the system
as a function of temperature.

3 E0
hEi = E0 . (1.35)
2
e kB T
−1

The heat capacity in Einsteins model becomes

d
CV = hEi (1.36)
dT

32
 1.4. The Compton effect

 2
3  E
 E 0

= kB  E  .

2  0 − 2k 0T
kB T e 2kB T
−e B

For T  E0 the heat capacity becomes equal to the classical limit of 3/2kB .
For temperatures smaller than E0 the heat capacity falls off to zero, which
is indeed observed experimentally.
Einstein assumed that each atom in a solid can oscillate independent of
its neighbours. This is not very realistic. A more appropriate consideration
of a solid and the treatment of its specific heat was given by De Bye in
1912 [38]. He starts his paper with a sentence stating that experimental
observations of the temperature dependence of the specific heat of solids
clearly show that one can not assume that energy is equally spread over
classical modes as one would obtain from classical physics. In order to
describe the specific heat, one needs to quantize the energy. The energy
unit this time not being hν as for photons, but a material specific constant
which we know as the De Bye temperature θD (or energy kB θD ).

1.4 The Compton effect

The previous sections showed that energy in all its forms is quantized. For
massive objects we are accustomed to the idea that materials are made from
atoms. There is a smallest quantum in which we can split the object. The
same holds for electromagnetic waves, which consists of photons. Photons
are particles with energy and momentum that can interact and scatter off
other objects just as atoms or other massive particles can do. Scattered
photons need to fulfill the same energy and momentum conservation rules
as we know for the scattering of two massive objects. One experiment,
where we need to describe the light field in terms of photons and particle
scattering instead of electromagnetic waves following Maxwells equations,
is the Compton effect.
In 1923, when Compton presented his work on the scattering of x-rays
from solids [39], it was well established that x-rays are electromagnetic

33
Chapter 1. The limits of classical physics

radiation with very short wavelengths. When high energy electrons (several
10’s of keV) hit a solid, high energy photons or x-rays are produced. These
rays were discovered in 1895 by Röntgen [40]. Directly afterwards, x-rays
were intensively studied, as they were very useful for imaging objects that
are otherwise hidden. Bragg [41] had established that one can use x-rays to
do diffraction experiments from solid crystals. If one knows the wavelength
of the light this allows one to determine atomic distances and the atomic
structure in crystals. For crystals with known atomic distances one can use
Bragg scattering to determine the wave length of the x-rays. In 1923, the
x-ray wavelengths, produced by different materials when bombarded with
high energy electrons in x-ray tubes, were well characterized.
In order to calculate the amount
of light scattered from a particle,
one can solve Maxwell’s equations.
One assumes the particle to be a
dielectric. Electric charges within
the particle will start to oscillate
in the electric field of the incoming
radiation. These oscillating elec-
tric charges will in turn produce
an electric field which radiates the
outgoing scattered light. In 1923,
this theory was well known and for Figure 1.5: Figure from the original
scattering of photons in the optical paper of Compton [39] showing the
wave length these equations work scattering of a photon on an electron
really well. For the inelastic scat- by the angle θ (left) and the momen-
tering of x-rays in solids this the- tum transfer relations that need to be
ory seemed to fail. Compton starts fulfilled (right).
his paper with the sentence: ”J. J.
Thomson’s classical theory of the
scattering of X-rays, ..., has been found incapable of explaining many of
the more recent experiments” [39].
Knowing the results of Planck and Einstein relating electromagnetic
waves to particles with energy E = hν and momentum |~ p| = hν
c , Comp-

34
 1.4. The Compton effect

ton used a different starting point to describe his inelastic scattering ex-
periments from Thomson. He assumed photons to be particles scattering
inelastic from electrons by the laws of classical physics of energy and mo-
mentum conservation. Given an incident photon with energy E0 = hν0
and momentum |~ p0 | = hνc 0 , one can scatter from an electron at rest by an
angle θ. The resulting energy and momentum of the photon areEθ = hνθ
hνθ 1
and |~
pθ | = c . For the electron we have the energy Ee = mc2 √
1−β 2
pe | = √mβc 2 , with β = ve /c. The momentum vectors for
and momentum |~
1−β
the particles involved are shown in figure 1.5. The equation of momentum
conservation is
!2 
hν0 2 hνθ 2
  
mβc hν0 hνθ
p = + +2 cos(θ). (1.37)
1 − β2 c c c c

The equation of energy conservation yields

1
hν0 + mc2 = hνθ + mc2 p . (1.38)
1 − β2

The dependence of the scattered frequency of the light as a function of

scattering angle becomes
1 hν0
νθ = ν0 where α= . (1.39)
1 + 2α sin2 ( 12 θ) mc2

By applying the same line of arguments for the scattering of photons

from moving electrons, Compton could explain his experimental observa-
tions. These findings have a deep philosophical implication for the nature of
electromagnetic waves. Compton showed that there are cases where electro-
magnetic radiation truly behaves more like particles than waves. In chapter
11, we will discuss the interaction between matter and electromagnetic ra-
diation within the Schrödinger equation. At that point, we will come back
to this dual nature of electromagnetic waves or photons and discuss the two
different scattering mechanisms.

35
Chapter 1. The limits of classical physics

1.5 Dual wave and particle nature of matter

In 1923, De Broglie presented his PhD thesis [42]. In this, he provides
several arguments to treat not only photons and electromagnetic waves, but
all matter with a dual particle and wave nature. He introduces concepts
like wave packets with a phase and group velocity to represent a localized
particle by a wave. These concepts we will discuss later within this lecture in
chapter 4. He also proposed to associate a wavelength to massive particles.
The De Broglie wavelength of a particle is
h
λ= , (1.40)
p
where h is the Planck constant and p the momentum of the particle. This
hypothesis could be tested by preforming a diffraction experiment involv-
ing for example a double slit with electrons. The wavelength of electrons
is rather short. An electron with a kinetic energy of 1 electron volt has
a wavelength of 1.7 nano meter. This makes a simple double slit experi-
ment difficult. Luckily, this wavelength fits with atomic distances in solids.
Electrons with a kinetic energy of a few eV to several keV can be used to
scatter of a crystal, producing similar diffraction patterns as one observes
with x-ray scattering.
De Broglies hypothesis was confirmed by electron diffraction experi-
ments [43]. With modern apertures one can do double slit experiments
[44, 45] with electrons. These experiments show the complete dual nature
of both matter and light in terms of particles and waves. By rule of thumb,
one finds that if the De Broglie wavelength of the particle is small compared
to the length-scale of the experiment one has a good change to observe par-
ticles. If the wavelength of the particle is comparable to the scale of the
experiment, one can describe the system by waves. A description that
includes both will follow from the Schrödinger equation.

36
 1.6. The Zeeman effect and the Stern-Gerlach Experiment

1.6 The Zeeman effect and the Stern-Gerlach Ex-

periment

The examples in the previous sections showed the dual nature of light and
matter. Some experiments are described well by treating light and matter
as particles, other by treating light and matter as waves. The Zeeman ef-
fect and the Stern-Gerlach experiment show an other property of quantum
physics. In order to explain these experiments, it follows that quantum
states can be considered as superpositions of other quantum states. This
accounts for things like entanglement and is a reason that quantum com-
puting might become much more efficient than classical computing.
Zeeman observed that atomic absorption and emission line spectra in a
magnetic field broaden [46]. With better resolution, one can observe that
they split into multiple discrete lines [47]. Single lines where observed to
split either into double lines (doublets), triple lines (triplets), or higher mul-
tiplicities, depending on the material. In order to explain this phenomenon,
it was concluded that the magnetic state of an electron or atom consists of
a discrete set of possibilities. There are two internal micro states, repre-
senting the magnetic moment for a system that shows two lines (doublet),
three internal micro states representing the magnetic moment for a triplet
etc. With the use of these finite number of internal micro states, one can
generate a magnetic moment in any spacial direction possible. We can ex-
emplify this for a doublet, i.e. for example an electron with spin 12 in a
magnetic field. For a doublet, the energy difference of the two observed
lines show that the magnetic moments of these two states are either paral-
lel or anti-parallel to the applied magnetic field. This is true independent
of the direction of the magnetic field. The two states related to the two
separate lines in the spectra for a field in one direction thus must be linear
combinations of the two states that emerge for a field in an other direction.
In three papers published in 1921 to 1922 [48, 49, 50], Stern and Gerlach
tested the assumption on the quantization of magnetization into a set of
discrete states with a simple experiment. They were looking for an experi-
ment that does not depend on spectroscopy. Obtaining results from spectra

37
Chapter 1. The limits of classical physics

requires often tractable, but involved theoretical considerations. Stern and

Gerlach set out to measure the possible orientations the magnetic moment
of a silver atom can have with respect to a magnetic field. Classically, one
would expect that any angle is possible. Based on the results of the Zeeman
effect, one should expect that a silver atom in a magnetic field can have
only a few, discrete orientations of its magnetic moment with respect to
the magnetic field.
The basic idea of Stern’s exper-
iment is that a magnet in an inho-
mogeneous magnetic field will en-
dure a force. Silver atoms have a
small magnetic moment. If a sil-
ver atom thus travels through an in-
homogeneous magnetic field, it will
bend according to the orientation
of the magnetic moment of the sil-
ver atom with respect to the mag- Figure 1.6: Results from the origi-
netic field. We will take the mag- nal Stern-Gerlach experiment. Left
netic moment of the silver atom the silver pattern without a magnetic
to be µ ~ . The magnetic energy in field. Right the silver pattern with an
a magnetic field for this atom is inhomogeneous magnetic field [49].
E(~r) = −~ ~ r). We assumed the
µ · B(~
magnetic field to be position dependent. The force on the silver atom is

F~ = ∇(~ ~
µ · B), (1.41)
~ r) = B
If we take the field to be B(~ ~ 0 + zB1 ẑ, we find a force to be in the z
direction and of the form
∂
Fz = (zµz B1 ) = µz B1 . (1.42)
∂z
In the following, we will assume that we always take the field gradient to
be parallel to the direction of the field.
In the Stern Gerlach experiment, silver atoms are created by heating a
piece of metallic silver in an oven. Hot silver metal evaporates and produces

38
 1.6. The Zeeman effect and the Stern-Gerlach Experiment

a silver vapour above the metal. Through a small opening in the oven the
silver vapour can move into the measurement chamber. As the measure-
ment chamber has been evacuated from air, the pressure difference allows
the silver atoms to come out in a beam-like fashion. This procedure of
creating a ray of silver atoms has no favoured orientation for the magnetic
moments of the silver atoms. The magnetic moment of the silver atoms
should be considered randomly orientated. If the atom traverses through
an inhomogeneous field as described above, one would expect that the sil-
ver atoms with the moment down (µz = −µ) would accelerate down-wards,
those with their moment up (µz = +µ) upwards and those with their mo-
ments perpendicular to the z direction (µz = 0) have their paths unchanged
by the magnetic field. In total, one would expect a semicircular distribution
centred around z = 0 representing the possible ways the magnetic moment
of the silver atoms could be orientated with respect to the field gradient.
The experimental result is different. In figure 1.6 we show the original
results of Stern and Gerlach. Without a magnetic field they focus the
silver atoms in a relatively sharp line. With an inhomogeneous field, the
line splits into two separate lines. The magnetic moment of the silver atom
is either parallel or anti-parallel to the field gradient, but seemingly never
perpendicular.
Although not performed by Stern and Gerlach, one can repeat such an
experiment multiple times. First of all we should notice that the choice of
the z direction is arbitrary. One can put the field gradient and magnetic
force in any direction acting on any component of the magnetic moment.
For a field and field gradient in the ~r direction, one should observe a split-
ting in the ~r direction of the ray of silver atoms. If the silver atoms pass
through an inhomogeneous field in the z direction with a field gradient in
the z direction, we observe a splitting into silver atoms with their magnetic
moment either parallel (+) or anti-parallel (−) to the magnetic field. We
can now block the silver atoms with their magnetic moment anti-parallel to
the magnetic field and make the silver atoms with their magnetic moment
parallel to the field pass through an other inhomogeneous magnetic field.
If this field is again in the z direction, we see no further splitting. This
situation is depicted in panel (a) of figure 1.7. The silver atoms are, after

39
Chapter 1. The limits of classical physics

Inhomogeneous Inhomogeneous Inhomogeneous

field parallel to field parallel to field parallel to
(a)
z+ z+
Source z z
z-
(b)
z+ x+
Source z x
z- x-
(c)
z+ x+ z+
Source z x z
z- x- z-

Figure 1.7: In a Stern-Gerlach experiment, we can choose the direction of

the field gradient and as such, choose the allowed magnetic orientations.
If we repeat the experiment with different directions for the magnetic field
gradient, the observation that quantum states magnetized in one direction
are superpositions of quantum states magnetized in a different direction
becomes visible.

the first pass through a field with gradient in the z direction, polarized in
the z direction. This polarization is conserved such that in a second pass
were we only enter with silver atoms that have their moment parallel to
the field, we will only observe silver atoms with their moment parallel to
the field.
If we polarize the silver atoms to have their moment in the positive z
direction, we can perform a Stern-Gerlach experiment with the field and
gradient in the x direction. This is depicted in panel (b) of figure 1.7. A
silver atom with its moment in the z direction has no moment in the x
direction. At the same time, the atom has to bend either in the x+ or
x− direction. Because of symmetry considerations, it is easy to see that
the silver atoms split equally into silver atoms with their moment either

40
 1.6. The Zeeman effect and the Stern-Gerlach Experiment

parallel or anti-parallel to the x direction. Naturally the same result would

have been obtained if one blocks the silver atoms with spin parallel to the
z direction and lets the silver atoms pass with their moment anti-parallel.
From this we must conclude that silver atoms with their moments polarized
in the z direction can be described as linear combinations of states that have
their moments polarized in the x direction. The same arguments hold if
the second field and field gradient were in the y direction.
If we add a third field and field gradient, again in the z direction, such
that we subsequently polarize the silver atoms in the z, x, and z direction,
we find silver atoms with their moments either parallel or anti-parallel to the
field gradient. Again by symmetry arguments we find both directions with
equal weight. Polarizing the silver atoms in the x direction and blocking
one component destroys the polarization in the z direction.
Mathematically, we can describe this by assuming that the silver atom
has two internal states. We will depict these by a vector. For state one we
write the vector
 
1
, (1.43)
0
and for state two we write the vector
 
0
. (1.44)
1
A specific silver atom will be in a superposition of these two states which
we can write as
     
1 0 α
α +β = , (1.45)
0 1 β

with the normalization condition that |α|2 + |β|2 = 1.  

α
The orientation of the moment for a silver atom in the state can
β
be calculated with the use of the Pauli matrices.
 
∗ ∗ α
~ = µ0 (α , β ) · (σx x̂ + σy ŷ + σz ẑ) ·
µ , (1.46)
β

41
Chapter 1. The limits of classical physics

with
 
0 1
σx = , (1.47)
1 0
 
0 −i
σy = ,
i 0
 
1 0
σz = .
0 −1

With the use of this mathematical framework, we can explain the Stern-
Gerlach experiment. After the silver atoms moved through a field and field
gradient in the z direction, we have two sets of silver atoms. The silver
atoms with their moments parallel to the field are in the state
 
1
. (1.48)
0

The silver atoms with their moments anti-parallel to the field are in the
state
 
0
. (1.49)
1

You should convince yourselves that with the use of equation 1.46 this
indeed yields the correct orientation of the magnetic moment. If the field
and gradient would have been in the x direction, we would have found the
polarized states to be
p   p 
1/2 1/2
p , and p . (1.50)
1/2 − 1/2

For the y direction the polarized states are

p   p 
1/2 p1/2 .
p , and (1.51)
i 1/2 −i 1/2

42
 1.6. The Zeeman effect and the Stern-Gerlach Experiment

We can now understand what happens if silver atoms polarized in the

z+ direction enter a field and field gradient in the x direction. The state
with moments in the positive z direction can be written as
  r p   p 
1 1 1/2 1/2
= p + p . (1.52)
0 2 1/2 − 1/2
On the right hand side of the equal sign we wrote the state polarized in the
z direction as a superposition of two states polarized in the x direction. In
a Stern-Gerlach experiment, where silver atoms polarized in the z direction
enter a field and field gradient in the x direction, one should expect two
lines corresponding to silver atoms with their moments either parallel or
anti-parallel to the x direction.
Some rays of atoms split, when moving through a magnetic field with
field gradient into 3 (or more) lines. For these we define a number of internal
states that is equivalent to the number of lines observed. For a ray that
splits into n lines, the different states are represented by the n unit vectors
of dimension n. In order to relate these states to the orientation of the
magnetic moment, we can define matrices similar to the Pauli matrices and
use these to calculate the magnetic moment.
The results of the Zeeman effect and Stern-Gerlach experiment put
severe restrictions on the allowed theories that are compatible with exper-
iment. If we assume that the quantum state that describes a magnetic
silver atom is given by a differential equation, then without a magnetic
field there are two solutions. One solution stands for the state with spin
up, one solution stands for the state with spin down. We are allowed to
‘quantize’ the magnetic state in any direction. The state with spin up in
the z direction can be represented by a linear superposition of the state
with the magnetic moment up and down in the x direction. Without a
magnetic field, any linear combination of the two solutions of the equations
describing the propagation of the silver atom needs to be valid solutions of
these equations as well. The set of equations that fulfill these properties are
linear differential equations. We thus know that in order to be in agreement
with experiment, any differential equation that describes the propagation
of a quantum system needs to be a linear differential equation.

43
Chapter 1. The limits of classical physics

1.7 Atomic physics

Many considerations that lead to the introduction of the theories of quan-
tum physics originate from the study of atomic physics. Schrödinger starts
his paper introducing the Schrödinger equation with the sentence: “In this
communication I show, in the simplest case of the (non-relativistic and
undisturbed) Hydrogen atom, that the usual quantization rule can be re-
placed by another requirement in which there is no longer a word of whole
numbers.” Understanding the atomic theory at that time and the prob-
lems that arise if one wants to understand atomic theory with classical
(electro-) mechanics of 1925 helps to understand the motivation behind the
Schrödinger equation.
Our understanding of atoms is the result of three different branches of
science: Chemistry, Spectroscopy and Diffraction. Each of these scientific
fields contributed to our understanding of the internal structure of atoms.
We start with a subsection containing a brief summary and a selection of
different experimental findings. In the next subsection, we describe the
different atomic models that were proposed based upon these findings.

1.7.1 Experimental findings

Historically, many experiments have been conducted that lead to our mod-
ern understanding of atoms. Especially before 1700, experiments dealing
with atoms were mostly related to alchemy. These experiments are gener-
ally depicted in popular literature as a mix of science, religious sects and
searching for gold. Besides the search for gold, Alchemist found out how
to change the properties (strength and stiffness) of metals. The proper-
ties of metals can be very different based on how the material is treated.
This is even more so when small amounts of other elements are allowed to
mix during the treatment. Copper alloyed with some small amount of tin
becomes bronze, which is much harder. Iron with some small amounts of
Carbon becomes Steel, which does not rust as easily. Trying to change lead
into gold might not have sounded so strange to people in the 18th or even
the 19th century as it does to us. Our understanding of atoms, molecules

44
 1.7. Atomic physics

and solids has come a long way since the 18th century and transformed in
an accurate science with quantitative predictive power.

Chemistry
Atomic physics studies the proper-
ties of the different elements. We
know about 118 different atomic el-
ements, with several different iso-
topes. In the periodic table they
are listed according to their atomic
number and grouped in rows and
columns according to their chemical
properties. Elements in the same
column have the same atomic va-
lence. Valence can be defined as the Figure 1.8: Elements in the peri-
number of hydrogen atoms that can odic table ordered according to their
combine with an element in a bi- atomic number and filling of different
nary hydride or twice the number atomic shells.
of oxygen atoms combining with an
element in its oxide or oxides. The atomic mass of each element is roughly
twice the atomic number.
In 1789, Lavoisier [51] wrote a text book discussing the chemical prop-
erties of the different elements. Naturally at that point in time not all
elements were known but basic chemical principles started to emerge. At
a given temperature and pressure, two different gasses have roughly the
same number of molecules per volume. A cubic cm of oxygen gas reacts
with two cubic cm of hydrogen gas, whereas a cubic cm of fluorine gas
only reacts with one cubic cm of hydrogen gas. This allows one to classify
the elements according to how they react with hydrogen. Based on these
and similar considerations Mendeleev arranged the elements in the peri-
odic table. His ordering allowed him to predict undiscovered elements. He
predicted the existence of the element Gallium and some of its chemical
properties. Gallium was subsequently found in the year 1875.

45
Chapter 1. The limits of classical physics

Why the periodic table has the structure it has was hard to understand.
With the publication of Schrödingers equation (combined with the relativis-
tic corrections of Dirac), one can immediately understand the structure of
the periodic table. With the Schrödinger equations, one can calculate the
magnetic, optical, and chemical properties of the elements. As a side note,
it turns out that the Schrödinger equation is highly non-trivial to solve
accurately for atoms with more than two electrons, and even harder for
molecules or solids. For complex systems including molecules and solids,
getting qualitative results is sometimes even a challenge, as we will discuss
in chapter 9 on many particle systems.

Optical absorption and emission spectroscopy

In his 1704 book Optics, Newton
describes an experiment where he
separates the white light of the sun
into its constituent colours (see Fig.
1.9). Newton used a dark room
with the windows blinded. In one
of the blinds there is a small pin-
hole. As the sun is far away the
light that comes trough the pinhole
travels in a near straight line. With
the help of a lens and an optical Figure 1.9: Sketch from Newtons
prism one can create a rainbow on notes showing the experimental setup
a screen. Newton laid the foun- separating the white light of the sun
dations of modern optics. In the in its different colours.
next hundred years that followed,
the quality of lenses gradually improved. Better telescopes were developed
to look at stars with sharper images (and ships at the horizon with more
accuracy). At the same time, optical devices were developed and improved
to analyze chemical compositions of liquids and gasses.
With these improved optical devices, discrete absorption lines in the
spectrum of the sun (see figure 1.10) and discrete photon emission lines

46
 1.7. Atomic physics

of heated elements (either in a flame or electric spark) were observed (see

figure 1.11). Around 1860, Kirchhof summarized these findings in his three
laws of spectroscopy

• A heated solid, liquid, or

dense gas will radiate light at
all wavelengths.

• A heated low density gas

emits light at discrete wave-
lengths specific to the ele-
ment.

• If light composing a continu-

ous spectrum passes through Figure 1.10: Solar spectrum in the
a dilute gas or dilute liquid visible range with discrete absorp-
solution, the result will be a tion lines as seen with the McMath-
discrete set of element specific Pierce Solar Telescope. Each of the 50
absorption lines. slices covers a wavelength range of 60
angstroms, for a complete spectrum
The discrete lines observed in from 4000 to 7000 angstroms (NASA
emission or absorption spectra were 2018) [52].
of practical use. These lines helped
to identify the chemical composi-
tion of materials and find new el-
ements. In the years to come, sev-
eral people tried to find regularities
in the wave-length of the emission and absorption lines. In 1879, Liveing
and Dewar [53] noticed that the Spectra of sodium and potassium in the
visible wave-length showed some linesFigure 1.11:sharp
that were Emission lines of
and some Nawere
that and
K with discrete lines as sketched
diffuse. Zeeman (1896) showed that these lines can split in a magnetic field by
[46]. This allowed for a classification Liveing
of lines and Dewar [53]
as singlets, (1879).triplets
doublets,
etc. according to the number of lines that emerged in a magnetic field.

47
Chapter 1. The limits of classical physics

Balmer, interested in numerology, searched for relations between numbers

and in 1885 noticed a simple regularity in the energy of the emission and
absorption lines of Hydrogen [54]. The wave-length of the spectral lines in
the Balmer series of hydrogen follows the simple rule:
n22
λ=h , (1.53)
n22 − 4
with n2 an integer equal or larger than 3 and the proportionality constant
h = 3.6456 10−7 meter. Rydberg generalized [55] this equation to not only
include a single spectral line series, but several of the series in the Hydrogen
spectrum.
 
1 1 1
= RH − , (1.54)
λ n21 n22
with RH the Rydberg constant for Hydrogen (RH = 4/h) and n1 and n2
positive integers with n2 > n1 . The Rydberg formula with n1 = 2 becomes
the Balmer series, with n1 = 1 becomes the Lyman series and with n1 = 3
becomes the Paschen series.
Observing these discrete emission and absorption lines must have been
very puzzling for somebody with the physical knowledge of the 19th century
(i.e. classical physics). The equations describing classical physics have
solutions for all values of energy. For atoms, the light emitted or absorbed
has only a discrete set of allowed energies. The internal energies of an
atom seemed to be discrete. These discrete set of energies are different for
each element. Really puzzling was that the discrete energies can be shifted
continuously with the use of a magnetic field.
At the beginning of the 20th century it was firmly experimentally es-
tablished that the internal energy of an atom can only increase in discrete
steps. For some cases there were strange rules involving integer rations be-
tween the different energy levels. At the same time this just did not fit into
any theoretical model of that time. We will see that Schrödingers equation
explains these energy levels. This is one of the reasons that despite the
very different way of thinking by Schrödinger, the equation was accepted
relatively fast and by the complete physics community.

48
 1.7. Atomic physics

Diffraction

In 1897, Thomson [56] published

an article that argued that β rays NaCl C14H10

are made of negatively charged par-

ticles. These particles are much
lighter than the lightest known
atom (hydrogen) and have a large
charge over mass ratio. We now
know these particles under the
name of electrons. In 1906, he
looked at the scattering and ab-
Figure 1.12: Electron density of NaCl
sorption of x-rays and beta rays in
(left) and C14H10 (right) retrieved
gasses. Thomson showed that the
from x-ray diffraction data.
absorption and scattering of x-rays
is proportional to the number of
atoms times the atomic weight of the atoms [57]. From this he concluded
that the number of particles in an atom is proportional to its atomic weight.
Shortly thereafter, Rutherford [58] reanalyzed the scattering of α (He2+ )
and β (e− ) particles from solids. In particular, he looked at the experiment
of Geiger and Marsden [59, 60] where α particles are scattered from a thin
foil of metal. Most α particles are only deflected by a very small angle (less
than 1 degree), but some of them were deflected by more than 90 degree.
Assuming that the deflection happens due to electro-static interactions, he
came to the conclusion that the atom must consist of a heavy nucleus with
positive charge surrounded by the light electrons with negative charge.
In 1912 Laue [61] and in 1913 Bragg [41] reported on the diffraction of
x-rays from crystals. The wavelength of the photons used in Bragg’s exper-
iment is about 100 pm (12 keV). The distances between nearest neighbour
atoms in a crystal are between 70 pm (Hydrogen bonds) to 400 pm, with
many bonds between atoms around 200 pm. A regular crystal of atoms
functions as a grating for x-rays. Bragg and Bragg observed a diffraction
pattern. They systematically measured the diffraction maxima of differ-
ent crystals for different orientations. From this they were able to resolve

49
Chapter 1. The limits of classical physics

the crystal structure. All of a sudden, atoms were no longer philosophical

constructs, but could be visualized. Or more precisely, their ordering and
shape given by the accompanying electron density in a crystal could be
reconstructed. In NaCl a distance of about 270 pm between the Na and Cl
atoms was found. By comparing different crystals, one can assign a typical
size to each atom or ion. If one looks at a periodic table, one can often find
numbers for the atomic or ionic radii of the different elements. These are
related to typical distances these elements will have to their neighbours in
crystal structures.
As x-rays scatter mainly of the electrons in a crystal, one actually can
reconstruct the electron density. Figure 1.12 shows the electron density in
the crystal NaCl and the molecule C14 H10 [62].

1.7.2 Theoretical models

Based on the experimental considerations discussed in the previous sec-
tion, several theories for the atom arose. The ancient Greek (Democritus
of Abdera, Leucippus of Miletus, and Epicurus amongst others) already
philosophized about the existence of small particles surrounded by empty
space to make up matter. With the chemical findings of elements reacting
in fixed ratio’s with each-other to form different solids, liquids or gasses,
these theories got a scientific motivation. In 1808, Dalton published a text-
book on chemistry, where he assumed all material to be made of element
specific atoms [63].
About a century later, Thomson discovered the electron as a particle
[56]. He also noticed that the amount of electrons or x-rays scattered from
a solid scales with the atomic number (or mass) times the number of atoms
in a solid. Based on this he invented a model of an atom which has, besides
a large distributed positive charge, several electrons randomly distributed
within the atom. The number of electrons in the atom is proportional to
the atomic number.
Shortly afterwards, in 1911, Rutherford analyzed scattering data of al-
pha particles of a gold foil [58]. Based on the fact that most particles pass
through the gold foil with small scattering angles, but a few particles are

50
 1.7. Atomic physics

scattered by a large angle, he came to the conclusion that the positive

charge needs to be centred in a volume with radius about 3 orders of mag-
nitude smaller than the typical size of the atom, determined by distances
between atoms in a solid. His atomic model consists of a small nucleus with
positive charge equal to the atomic number Z × e. Around this Z electrons
with negative charge of e orbit. The positively charged nucleus attracts the
negatively charged electrons with a force proportional to 1/r2 . The result
of this model would be electrons traveling in a circular or elliptical orbit
around the nucleus. There are two problems with this model. Firstly, an
electron that is accelerated (in an elliptical orbit) emits electromagnetic
radiation. This loss of energy results in a collapse of the atom (in about
10−10 seconds). Secondly, this model does not describe the chemical, nor
optical properties of atoms observed in experiment.
In order to explain the periodic table and the discrete line spectra,
Niels Bohr introduced his planetary like model in 1913. He assumed that
negatively charged electrons move in discrete orbitals around a positively
charged nucleus. Not all distances for the possible orbits are allowed. The
Bohr orbitals are grouped in shells that are numbered and sub-shells with
the names s, p, d and f . The discrete line spectrum of hydrogen and
the alkali metals can be explained by transitions of the electron between
different orbitals. In a flame, the single valence electron of an alkali metal is
excited into the principle shell (p) and can transition with a sharp line into
the s shell or with a diffuse line into the d shell. With higher resolution,
the defuse lines can be resolved into several sharp lines. In a magnetic
field, the splitting of the transition lines revealed that the s shell contained
one orbital, the p shell contained 3 orbitals and the d shell contained 5
orbitals. Combined with the Pauli principle that each orbital can contain
two electrons with opposite spin, this explained the structure of the periodic
table.
Bohr’s model can be captured in two postulates:

1. The electrons move in orbits restricted by the requirement that the

circumference (2πr) of the orbit is an integer multiple (n) of the wave-

51
Chapter 1. The limits of classical physics

length (λ) associated with the electron

nλ = 2πr. (1.55)

The wave length is related to the momentum (p) or velocity (v = p/m)

by the De Broglie wavelength such that

h
λ= =⇒ (1.56)
mv
nh
= 2πr =⇒ (1.57)
mv
nh
mvr = . (1.58)
2π
With l = mvr the angular momentum of the electron quantized as
integer multiples of ~ = h/2π.

2. Electrons can change their orbit by absorbing or emitting photons.

These are discrete transitions responsible for the line spectra observed
of atomic radiation. The photon energy released or absorbed in these
transitions is given by the difference in the kinetic and potential en-
ergy of the electron in the different allowed orbitals.

E2 − E1
ν= , (1.59)
h
under the assumption that E1 > E2 . For an hydrogen-like atom (H,
or He+ , Li2+ , but also the Alkali metals with Z ≈ 1), we can calculate
the energy by stating that in a Bohr orbital the Coulomb force needs
to provide a centrifugal force that generates a circular orbit. From
Newtonian mechanics and electrostatics we know
Ze2 mv 2
= . (1.60)
4π0 r2 r
If one would use Maxwell’s equations to solve the electro dynamical
problem of a charge traveling in a point like potential, this would

52
 1.7. Atomic physics

not be a solution. Compton already showed that there are cases

where electrodynamical problems are better solved by not assuming
Maxwells equations, but using arguments from macroscopic particles
and forces acted upon them. Assuming this balance between forces
on the electron we find
e2 Z
v= , (1.61)
20 hn
for the velocity and
1 0 n2 h2
r= (1.62)
π Ze2 m
for the electron radius. The energy associated with a Bohr orbital
becomes
1 Ze2
E = mv 2 − (1.63)
2 4π0 r
e4 Z 2 m
= − 2 2 2.
80 h n
From this equation, we can directly derive the experimentally ob-
served Rydberg series.
1 E2 E1
= − (1.64)
λ hc hc
e4 Z 2 m 1

1
= 2 3 − .
80 h c n21 n22
The Rydberg constant thus has the value
e4 Z 2 m
RH = , (1.65)
820 h3 c
which agrees well with experiment
What remained was the question why we have discrete energy levels in
an atom. The answer to this question is a result of the Schrödinger equation
and its solution for the atom. We will discuss this in chapter 8 after we
introduce the mathematical framework needed to do this.

53
Chapter 1. The limits of classical physics

54
Chapter 2

Schrödinger’s conjecture and

the postulates of quantum
mechanics

In 1926 Erwin Schrödinger published 4 articles [7, 8, 9, 10] laying the fun-
daments of Quantum physics as we know it today. In his first paper he
conjectured an equation that describes the discrete spectrum of the Hy-
drogen atom. These papers are very well written. If you are interested in
the original argumentation I can recommend you to read these. An his-
torical overview of the introduction of quantum mechanics by Schrödinger
is given in the Arxive paper by Norbert Straumann [64]. This paper also
refers to letters written by Schrödinger. If you read the original articles of
Schrödigner I would recommend to read the article by Straumann as well.
In this chapter we will not follow the historical line of thought but
motivate the Schödinger equation by the equivalence between momentum
and wave-length as postulated by De Broglie [65]. In the end it must
be clear that the only reason we use the Schrödinger equation is that its
predictions agree astoundingly well with experiment.

55
Chapter 2. Schrödinger’s conjecture and the postulates of
quantum mechanics

2.1 Motivation
According to De Broglie there is a relation between the momentum of a par-
ticle and the corresponding wave-length this particle displays in a diffraction
experiment.
h
|~
p| = , (2.1)
λ
with p~ = (px , py , pz ) the momentum of the particle, h the Planck constant
and λ the wavelength. If one performs a diffraction experiment (i.g. elec-
trons passing through two slits with distances comparable or smaller than
the De Broglie wavelength) one needs to represent the particle by a wave.
We can represent a plane wave by a function of space coordinates ~r and
time coordinate t as:
~
ψ(t, ~r) = ei(k·~r−ω t) , (2.2)

with the wave-vector ~k = p~/~ and the angular frequency ω = 2πν. Using
the relation between energy and frequency E = hν = ~ω we have:
i
ψ(t, ~r) = e ~ (~p·~r−E t) (2.3)

We can obtain the energy E and the momentum p~ from the function ψ(t, ~r)
∂ ∂ ∂
by differential operators 1 . The gradient (∇ = ∂x x̂ + ∂y ŷ + ∂z ẑ) of the
wave-function results in the momentum times the wave-function
~
p~ ψ(t, ~r) = ∇ψ(t, ~r). (2.4)
i
The time derivative of the wave-function results in the energy times the
wave-function
∂
E ψ(t, ~r) = i~ ψ(t, ~r). (2.5)
∂t
1
Operators in Quantum mechanics map a function (ψ(t, ~r) in our example) to an other
function.

56
 2.2. The wave function and the Hamiltonian defining the
micro state and its time evolution

In section 7.2.1 and 7.2.2 we will give an additional symmetry argument

why these definitions for the momentum and energy operator make sense.
The classical non-relativistic energy-momentum relation
p~2
E= (2.6)
2m
becomes the differential equation
 2
∂ 1 ~
i~ ψ(t, ~r) = ∇ ψ(t, ~r) (2.7)
∂t 2m i
~2
=− ∆ψ(t, ~r),
2m
∂ 2 ∂2 ∂ 2
with ∆ the gradient squared or Laplace operator ∆ = ∂x 2 + ∂y 2 + ∂z 2 .
Equation 2.7 represents the Schrödinger equation for a particle without
any external potentials.
We can extend this equation to include an external potential. The
energy for a classical particle with momentum p~ at position ~r in a potential
V (~r) is:

p~2
E= + V (~r). (2.8)
2m
The Schrödinger equation for a particle in an external potential becomes:
∂ ~2
i~ ψ(t, ~r) = − ∆ψ(t, ~r) + V (~r)ψ(t, ~r). (2.9)
∂t 2m

2.2 The wave function and the Hamiltonian defin-

ing the micro state and its time evolution
For a classical particle moving in a (conservative) potential V with coordi-
nates ~r and momentum p~ we can define the Hamiltonian function
p~2
H(~r, p~) = + V (~r). (2.10)
2m

57
Chapter 2. Schrödinger’s conjecture and the postulates of
quantum mechanics

The classical Hamiltonian function specifies the energy of the particles un-
der consideration. The quantum mechanical equivalent is obtained by re-
placing the momentum p~ by the momentum operator p ~ = ~i ∇. With this
substitution the Hamilton function, H becomes the Hamilton operator, H

~2
p
H= + V (~r) (2.11)
2m
~2
=− ∆ + V (~r).
2m
A quantum mechanical state of a system is defined by its wave-function
ψ(t, ~r). If one knows the function at time t = 0 as ψ(0, ~r) = ψ0 (~r) one
can calculate the wave-function at any specific time with the Schrödinger
equation:
∂
i~ ψ(t, ~r) = Hψ(t, ~r). (2.12)
∂t

2.3 The Schrödinger equation for many particle

states
Equation 2.12 has been introduced to describe the propagation of a single
particle in a potential. For a system with N particles we can extend our
formalism relatively straight forward. If the particles do not interact with
each other we can define a Hamilton operator for each particle

~2
Hi = − ∆i + V (~ri ), (2.13)
2mi
with index i labelling the different particles at position ~ri with mass mi .
The total Hamiltonian is given by the sum of the Hamiltonians of each
particle
N
X
H= H i. (2.14)
i=1

58
 2.3. The Schrödinger equation for many particle states

This is however not the most general form of the Hamiltonian we encounter
in nature. In general particles can interact with each other. In this case
the potential plus interaction term V is not given as a sum of potentials
acting on the single particles
N
X
V (~r1 , . . . , ~rN ) 6= V (~ri ). (2.15)
i=1

For most realistic systems the energy will depend on the position of the
individual particles as well as on the distance between the particles
N
X
V (~r1 , . . . , ~rN ) = V1 (~ri ) (2.16)
i=1
N,N
X
+ V2 (|~r1 − ~rj |)
i=1,j>i

That said, we at this point do not need to go into detail on the specific
interactions between possible particles. It P
is enough to note that the sum
over the single particle potential energies N i=1 V (~
ri ) for the total poten-
tial energy should be replaced by a more general function V (~r1 , . . . , ~rN )
that includes the potential energy of each particle due to an externally ap-
plied field, plus the potential energy due to interaction terms between the
particles. The result is the many-particle Hamiltonian operator
N
X ~2
H= − ∆i + V (~r1 , . . . , ~rN ). (2.17)
2mi
i=1

The solution of the Schrödinger equation with this Hamilton operator is

given by the wave-function ψ(t, ~r1 , . . . , ~rN ).
We will show in a later chapter, where we deal with many particle prob-
lems in more detail, that for identical particles the Schrödinger equation
leads to too many solutions. Only an N factorial small fraction of all solu-
tions of the Schrödinger equation is physically realized. We need to restrict

59
Chapter 2. Schrödinger’s conjecture and the postulates of
quantum mechanics

the solutions to those that fulfil the Pauli principle. The wave-function
must be totally symmetric (Bosons) or anti-symmetric (Fermions) under
the permutation of the coördinates of identical particles. For two particles
this yields for Bosons

ψ(t, ~r1 , ~r2 ) = ψ(t, ~r2 , ~r1 ) (2.18)

and for Fermions

ψ(t, ~r1 , ~r2 ) = −ψ(t, ~r2 , ~r1 ). (2.19)

This is an additional restriction on the solution of the Schrödigner equation

needed to make the predictions of this equation in agreement with reality.

2.4 The interpretation of the wave-function

If one looks in the literature one finds that there is an ongoing discussion
on the interpretation of the wave function. This discussion started the mo-
ment Schödinger introduced his equation. Before the seminal 1926 papers
of Schödinger [7, 8, 9, 10, 11, 12] a lot of work was done on the theory of
quantum mechanics. Most notably by Heisenberg [13], Born and Jordan
[14, 15], who worked on a theory to describe, for example, the Stark and
Zeeman effect in atomic spectroscopy (See section 1.6 on the Stern Gerlach
experiment for a historically later, but simpler to understand experiment).
Born, Heisenberg and Jordan, worked until 1926, in the same institute in
Göttingen. In Göttingen they developed a matrix formalism of quantum
mechanics. These matrices represent operators on a linear vector space as
we will introduce in chapter 6. Born described the theories they developed
as: ”The matrix form of quantum mechanics, founded by Heisenberg and
developed by him together with Jordan and the author of this communi-
cation, is based on the idea that an exact representation of the processes
in space and time is impossible at all, and is therefore content with the
establishment of relations between observable quantities that can only be
interpreted as the characteristics of movements in the classic limit.” [66].
Schödinger introduced a wave function into quantum mechanics. In his

60
 2.4. The interpretation of the wave-function

original theory these functions represent the state of a system, independent

of observation.
In the current literature there are several different interpretations of the
wave-function available, with the set of interpretations known as the Copen-
hagen interpretations of Heisenberg and others the most prevailing ones.
We will, in this section, not give a complete overview of all interpretations
of quantum mechanics, nor in this lecture follow one specific interpretation.
Most text books choose to follow roughly the Copenhagen interpretation
of quantum mechanics. This, in my opinion, does not do justice to the
current state of the theory of quantum mechanics. We will try to keep a
relatively agnostic view in this lecture. At the same time, when we discuss
simple models, or the solution of the Hydrogen atom, we do want to talk
about physical systems and not just some beautiful mathematical equa-
tions. There are several concepts to which (almost) all interpretations of
quantum mechanics agree. These are sufficient to obtain predictive power
from the equations of quantum mechanics and allow the reader to, after the
study of the mathematics of quantum mechanics, form an informed opinion
on the different interpretations of quantum mechanics.
One point where Heisenberg and Schrödinger did agree upon, and which
holds for many interpretations of quantum mechanics today, is that the
wave function uniquely and completely defines the microscopic quantum
state. In Heisenbergs picture, or at least in its modern extension, the
quantum state is given by a complex vector on a Hilbert space. This is
mathematically equivalent to describing a quantum state by a wave function
whose time evolution is given by a linear differential equation (See chapter
6).
In order to understand the meaning of the microscopic state of a quan-
tum system we can start with the definition of the micro state of a classical
system. Classically, the state of a system of N particles is defined by the N
position coordinates and N momentum coordinates of the particles. Within
the Schrödinger equation the microscopic state is represented by a function
that assigns a complex number (complex vector for particles with a spin) to
each possible set of coordinates for the N particles. The question on how
to interpretate the meaning of the wave function relates to the question

61
Chapter 2. Schrödinger’s conjecture and the postulates of
quantum mechanics

on how to relate the wave-function defining the quantum state to the N

position and N momentum coordinates that define the classical state of a
system.
One limit where all interpretations of quantum mechanics agree is the
limit of macroscopically many particles. For macroscopically many particles
we can replace the positions of the individual particles by a particle density
ρ(t, ~r) and the velocities of the individual particles by a current density
~j(t, ~r). The density is defined as the number of particles per volume at the
point given by ~r. The current density describes the rate of flow through an
area at point ~r. Knowledge on the particle and current density of a classical
system defines its macro state. For a gas or fluid one can write down
the Navier-Stokes equations. These equations define the time evolution
of the system, based on the classical macro state, i.e. the particle and
current density. Following Einsteins reasoning [31], similar considerations
must hold for photons. For photons the particle density ρphotons (t, ~r) is
proportional to the energy density of the electromagnetic field. With the
energy per photon given by Planck’s unit of energy quanta hν. The energy
density of an electromagnetic field is proportional to the norm square of
the local electric and magnetic field.
1
ρphotons (t, ~r) = U (t, ~r) (2.20)
hν  
1 1 1 ~∗ ~ ~ ∗ ~
= B (t, ~r) · B(t, ~r) + 0 E (t, ~r) · E(t, ~r) ,
hν 2 µ0
~ ~r) the local magnetic field and E(t,
with U (t, ~r) the energy density, B(t, ~ ~r)
~
the local electric field. The photon current density jphotons (t, ~r) is propor-
tional to the Poynting vector

~jphotons (t, ~r) = 1 S(t,

~ ~r) (2.21)
hν
1 1 ~ 
~ ~r) .
= E(t, ~r) × B(t,
hν µ0
The time evolution for photons is not written on the level of the photon
densities, but given by the Maxwell equations which act on the fields.

62
 2.4. The interpretation of the wave-function

Similar as the electromagnetic field defines the photon density we can

use the wave function to define the particle density. For a system with
macroscopically (infinitely) many particles the particle density becomes a
smooth function of space. For N particles we define the particle density as
(see chapter 9)
Z Z
ρ(t, ~r) = N 3
d ~r2 . . . d3~rN ψ(t, ~r, ~r2 , . . . , ~rN )∗ ψ(t, ~r, ~r2 , . . . , ~rN ).
R3 R3
(2.22)

The particle density of a many particle system is defined as the integral

of the norm-squared of the wave-function over all coordinates of the wave-
function except for one. Due to the restrictions given by the Pauli-principle
on the wave-function it does not matter which coordinate one does not
include in the integration.
The particle density ρ(t, ~r) is a continuous function in space. For macro-
scopically many particles the experimental result can be approximated by
a continues density. It is in this limit where all interpretations of quantum
mechanics agree with each other and the experimental observations. ρ(t, ~r)
as given in equation 2.22 defines the particle density. The definition of the
current density will follow in section 3.1.
Instead of looking at the particle density one can look at any other phys-
ical observable such as the energy, velocity or magnetic moment. Physical
observables in quantum mechanics are defined by an Hermitian (section
3.3) operator, O. For a macroscopic system the observables measured in
an experiment are not given by a single number, but form a distribution
of values, ρ(o). For example, if one would measure the kinetic energy ε
of a particle in an ideal gas, the value measured would not be the same
for each measurement. The kinetic energy of a single gas particle fluctu-
ates in time due to the exchange of energy between particles. For an ideal
gas we know the temperature dependent distribution ρ(ε) of the kinetic
energy per particle which is given by the Maxwell Boltzmann distribution
3/2
p  − ε
ρ(ε) = 2 πε kB1T e kB T .

63
Chapter 2. Schrödinger’s conjecture and the postulates of
quantum mechanics

We now briefly run ahead and make a statement we will prove and
discus later, in section 6.5. Given the quantum state described by the wave
function ψ(t, ~r1 , . . . , ~rn ). It follows from the Schrödinger equation, that if
we for this state measure an observable, described by operator O and find as
a result the value o, then if we repeat this experiment many times, starting
from the same state ψ(t, ~r1 , . . . , ~rn ) we will find a distribution of values
ρ(o). The possible values of o are given by the eigenvalues (spectrum) of
the operator O. The frequency with which these values are found are given
by the norm square of the inner product of the eigenstates of operator O
and the many particle wave function of the system. More on this later,
after we introduced the proper math.
Finding a distribution of values for a repeated experiment on an en-
semble of macroscopically many particles is not surprising. Thermal fluc-
tuations are expected in classical systems and described by statistical me-
chanics. Despite the similarities between thermal and quantum fluctuations
there are however important differences. Firstly, thermal fluctuations av-
erage out over time, whereby quantum fluctuations all happen at the same
time (again more on this later). Secondly thermal fluctuations disappear
if one considers an isolated single particle. Quantum fluctuations remain,
also for a single particle.
The quantum mechanical description and the fluctuations of observ-
ables, described above is not restricted to the case of macroscopically large
values of N . The previous equations also work for the case where N = 1, i.e.
a single particle. Quantum mechanics, or at least the Schrödinger equation,
predicts that for a single particle, observables are given by a distribution
of values. Schrödinger in his original interpretation of quantum mechanics
took this quite literally and proposed to replace the concept of point par-
ticles by wave packets, with finite, but small widths [11]. Observations on
a single particle would then see the full distribution function of values.
If one looks at the physics of a single atom Schrödinger’s original inter-
pretation has several appealing features. For these cases Schrödinger’s pic-
ture provides a natural explanation for several experiments. Experiments
where one performs single atom diffraction observe the Fourier transform of
a spatial distribution of the atomically bound electron charge and spin den-

64
 2.4. The interpretation of the wave-function

sity, known as the atomic form factor. Similar arguments and observations
are true for experiments whereby the typical length scale in the experiment
is of the same order as the (de Broglie) wave length of the particles involved
and the temperature (kB T ) is small compared to the low energy excitations
in the system. Typical examples are scanning tunnelling microscopy and
spectroscopic variants of this method where single atoms and the orbital
structure of the electronic states is seen in the experimental results. In
order to determine the typical de Broglie wave length one should realize
that the kinetic energy of a particle at rest, i.e. a stationary state is not
necessarily zero.
If one on the other hand does experiments whereby the typical length
scale is much larger than the wave length of the particles involved something
seems to go wrong with Schrödinger’s original proposal for the interpreta-
tion of quantum mechanics. Cosmic rays originating from far away stars
do not appear as spread out wave packets. A cosmic particle appears in
a bubble chamber as a well localized point like particle on a clear trajec-
tory. It is possible to perform experiments on single particles whereby the
outcome is not a distribution function, but a single value. Before we will
touch the philosophical question on what this has to mean for the inter-
pretation of the wave function we will have a look at what possible single
values are allowed for such a measurement. For a microscopic ensemble of
N non-interacting particles we know that the solution of the Schrödinger
equation (section 6.5) dictates that we observe for the observable values of
an operator O a distribution function ρ(o) given by the eigenvalues of oper-
ator O with frequency given by the norm square of the inner product of the
eigenvectors of operator O with the N particle wave-function of the system.
If we have an apparatus that measures a single unique value for each par-
ticle, then in order to be in agreement with the solution for the ensemble
of N particles these values must be given by o, the eigenvalues of operator
O. The probability to obtain the eigenvalue o is given by the distribution
function ρ(o). An experiment that beautifully exemplifies this is the double
slit diffraction experiment. With a low flux electron beam one can register
each electron separately. As such one can see how the diffraction pattern
builds up as a distribution from individually events [44, 45].

65
Chapter 2. Schrödinger’s conjecture and the postulates of
quantum mechanics

The dispute on the interpretation of the wave-function relates to the

philosophical meaning of this probability. Is the probabilistic outcome
something inherent to the nature of quantum particles as assumed in the
Copenhagen interpretation, is this related to the entanglement of the ob-
server and the quantum particle as one assumes in the many world inter-
pretation of quantum mechanics, or are there additional hidden variables
involved that determine the outcome of the experiment for a single parti-
cle as assumed in the pilot wave interpretation of quantum mechanics. In
order to progress on this question it has become evident that one needs to
understand the process of measurement in more detail. This by all means
is a non-trivial task. Measurement apparatus are normally given by large
quantum systems including many interacting particles. Describing such
systems is, due to the exponential growth of the many-body quantum com-
putation problem nearly impossible. As we will discuss in chapter 9, such
many-body quantum systems can lead to emergent behaviour whereby the
outcome of the equations of quantum mechanics for complex systems of-
ten defy the intuition one obtains from solving simple small model systems
[67, 68].
We can largely evade the question on the ontology of the wave-function
and quantum state by treating the probabilistic outcome of an experiment
that measures a single value as a phenomenological theory. We do not know
how a measurement apparatus that measures a single value for an observ-
able of a single particle on a fundamental level really works. We do know
that we can predict its outcome with great accuracy. For this we follow
the rules that for the measurement given by the operator O the possible
outcome for the measured values are the eigenvalues of this operator. The
frequency distribution of these values is given by the distribution ρ(t, o) as
defined and derived from the Schrödinger equation in section 6.5.

66
Chapter 3

Mathematical properties of
the Schrödinger equation

In this chapter we will discuss the mathematical properties of the Schrödinger

equation. We will use these mathematical relations in the next chapters to
find solutions to the Schrödinger equation. In principle, we are interested in
the Schrödinger equation in d = 3 spatial dimensions, as this corresponds
to the space we live in. For several examples it will be easier to look at
solutions in d = 1 or d = 2 spatial dimensions. As such we will not specify
the dimension of space in this chapter, but use d for the number of spatial
dimensions. We furthermore look at a single particle in this chapter. The
extension to N particles will be made in chapter 9.
The state of the system is given by the wave function ψ(t, ~r), with
~r = (r1 , . . . , rd ) a d dimensional vector.
The Schrödinger equation, which specifies the time evolution of a quan-
tum state defined by the wave function ψ(t, ~r) is defined as
∂
i~ ψ(t, ~r) = Hψ(t, ~r), (3.1)
∂t
with H the Hamilton operator (or Hamiltonian)
~p2
H= + V (~r) (3.2)
2m

67
Chapter 3. Mathematical properties of the Schrödinger
equation

~2
=− ∆ + V (~r).
2m
Where we used that the momentum operator is given as

~p = −i~∇. (3.3)

The gradient ∇ of a function for an d dimensional system is defined as the

vector
i=d
X ∂
∇= rˆi . (3.4)
∂ri
i=1

The gradient squared or Laplacian ∆ = ∇ · ∇ is defined as the scalar

i=d
X ∂2
∆= . (3.5)
i=1
∂ri2

3.1 Continuity equation, density conservation and

current density
In the previous chapter we declared for a system with N particles
Z Z
ρ(t, ~r) = N d~r2 . . . d~rN ψ(t, ~r, ~r2 , . . . , ~rN )∗ ψ(t, ~r, ~r2 , . . . , ~rN ), (3.6)

as the particle density. This definition for the particle density only makes
physical sense if the total number of particles is conserved. Furthermore, for
ρ(t, ~r) to be a physically sensible density one needs the continuity equations
to be fulfilled. The continuity equation states that density cannot disappear
at one point in space ~r0 to instantaneously reappear at a distant point ~r,
which we can formalize using the -δ definition of continuity. For any small
spacial distance  = |~r − ~r0 |, there is a small time period δ = t − t0 during
which the density stays within a distance of  around its original position
~r0 . Density flows in a continuous motion from one position to another. In

68
 3.1. Continuity equation, density conservation and current
density

order to formulate this physical observation mathematically we can define

a closed surface S ⊂ Rd−1 that encloses a volume V ⊂ Rd in d dimensions,
as shown in figure 3.1.
The change in total number of particles in volume V as a function of
time is equal to the flux flowing through the surface S enclosing this volume.
Z Z
∂ d ~j(t, ~r) · d~n,
ρ(t, ~r)d ~r = − (3.7)
∂t V S

with vector ~n the outward point-

ing surface normal at each point of
the surface S and ~j(t, ~rR) the current S
density in point ~r, and S ~j(t, ~r)·d~n
the total particle current leaving V
volume V through surface S.
Subject to the condition that all
integrals are well behaved and con-
vergent (more on that later) we can
simplify these equations. On the Figure 3.1: A d = 2 dimensional com-
left we can interchange the deriva- pact volume V enclosed by a piece-
tive with respect to time and the in- wise smooth boundary surface S.
tegral over volume V . On the right
we can use the divergence theorem (one of Gauss’s theorems) to find
Z Z
∂ d
ρ(t, ~r)d ~r = − ∇ · ~j(t, ~r)dd~r. (3.8)
V ∂t V

This equation needs to be valid for any volume V such that the equality
needs to hold not only for the integral, but also for the integrants.
∂
ρ(t, ~r) = −∇ · ~j(t, ~r). (3.9)
∂t
The change in density at point ~r as a function of time is equal to minus the
density that flows out of this point at time t.
In equation 3.6 we defined the particle density in terms of the wave
function. In quantum mechanics the continuity equation not only holds for

69
Chapter 3. Mathematical properties of the Schrödinger
equation

an ensemble of N particles, but holds for each particle separately. This

will simplify the notation in the rest of the derivation. The single particle
density is given as

ρ(t, ~r) = ψ(t, ~r)∗ ψ(t, ~r). (3.10)

We now need to find a proper definition of the current density ~j(t, ~r) such
that equation 3.9 is fulfilled. We start with the left hand side of equation
3.9. For the time derivative of the density in point ~r we find
∂ ∂
ρ(t, ~r) = (ψ(t, ~r)∗ ψ(t, ~r)) (3.11)
∂t ∂t
   
∂ ∗ ∗ ∂
= ψ(t, ~r) ψ(t, ~r) + ψ(t, ~r) ψ(t, ~r) .
∂t ∂t
Next, we replace the time derivative of the wave function by a spatial
derivative using the Schrödinger equation. We need both the time derivative
of the wave function as well as the time derivative of the complex conjugate
of the wave function. The complex conjugate of a function that is a solution
of the Schrödinger equation is by itself not necessarily a solution of the
Schrödinger equation. We can find the relation between the time derivative
of the complex conjugate of a wave function and its spatial derivatives by
looking at the complex conjugate of the Schrödinger equation.

~2
 
∂
i~ ψ(t, ~r) = − ∆ + V (~r) ψ(t, ~r) =⇒ (3.12)
∂t 2m
∗  ∗
~2
 
∂
i~ ψ(t, ~r) = − ∆ + V (~r) ψ(t, ~r) =⇒ (3.13)
∂t 2m
~2
 
∂ ∗
−i~ ψ(t, ~r) = − ∆ + V (~r) ψ(t, ~r)∗ , (3.14)
∂t 2m
whereby in the last step we used that the potential V (~r) is real.
We can substitute this into equation 3.11 to find

~2
   
∂ 1 ∗
ρ(t, ~r) = − ∆ + V (~r) ψ(t, ~r) ψ(t, ~r) (3.15)
∂t −i~ 2m

70
 3.1. Continuity equation, density conservation and current
density

~2
   
∗ 1
+ ψ (t, ~r) − ∆ + V (~r) ψ(t, ~r)
i~ 2m
~
=− (ψ(t, ~r)∗ ∆ψ(t, ~r) − ψ(t, ~r)∆ψ(t, ~r)∗ )
2mi
~
= −∇ · (ψ(t, ~r)∗ ∇ψ(t, ~r) − ψ(t, ~r)∇ψ(t, ~r)∗ ) .
2mi
We now define the current density as

~j(t, ~r) = ~ (ψ(t, ~r)∗ ∇ψ(t, ~r) − ψ(t, ~r)∇ψ(t, ~r)∗ ) (3.16)
2mi
1
= (ψ(t, ~r)∗ p pψ(t, ~r)∗ ) ,
~ψ(t, ~r) − ψ(t, ~r)~
2m
such that the continuity equation for the density defined for a single par-
ticle in quantum mechanics is obeyed. This leaves the question if the cur-
rent density as defined in equation 3.16 is a sensible definition. A sensible
definition requires that the obtained theory is internally mathematically
consistent. This is given by construction. It furthermore requires that the
theory describes the experimental observations. The density and current
as defined in equation 3.10 and 3.16 describe experiments with a very high
accuracy.
We can now also prove that the total density of a system is conserved
if certain boundary conditions of ψ(t, ~r) are considered. From equation 3.7
we immediately seeR that if the current through a surface S that encloses
volume V is zero ( S ~j(t, ~r) · d~n = 0) the total number of particles within
this volume is constant in time.
For infinite volumes V = Rd we have conservation of particle number if
the density is finite and the surface integral for V → Rd converges to zero.
This sets restrictions on the possible wave functions we can allow.
Z Z
ψ(t, ~r)∗ ψ(t, ~r)dd~r = ψ(t = 0, ~r)∗ ψ(t = 0, ~r)dd~r (3.17)
Rd Rd

if
Z
ψ(t = 0, ~r)∗ ψ(t = 0, ~r)dd~r < ∞. (3.18)
Rd

71
Chapter 3. Mathematical properties of the Schrödinger
equation

We require that wave functions that represent physical systems can be

normalized, i.e. they are square-integrable and that the integral over the
norm-square is a positive number. We furthermore require that Hψ(t, ~r)
is square-integrable. The second order partial derivative of ψ(t, ~r) as well
as V (~r)ψ(t, ~r) need to be square-integrable. We will call the domain of all
allowed functions that are possible physical solutions of the Schrödinger
equation DH
This constraint is actually important and not just an exotic boundary
case. We will see for the solutions of the Hydrogen atom that the require-
ment for the wave function to be square-integrable is an essential part to
find discrete solutions. Furthermore, when we discuss model systems we
~
will encounter plane wave solutions. The plane wave eik·~r is not normaliz-
able in Rd . There are several ways to circumvent this, which we will discuss
later in this lecture.

3.2 Linearity
The Schrödinger equation is a linear partial differential equation. It is a
multi variable differential equation with partial derivatives. It is linear
while it only depends linearly on partial derivatives of ψ(t, ~r), i.g. there are
∂
no quadratic terms of the form ∂t ψ(t, ~r) ∂r∂ i ψ(t, ~r), ∂r∂ j ψ(t, ~r) ∂r∂ i ψ(t, ~r), or
other higher order polynomial functions of the partial derivatives. Linear
differential equations have the property that if ψ1 (t, ~r) and ψ2 (t, ~r) are both
solutions of the differential equation then ψ(t, ~r) = c1 ψ1 (t, ~r) + c2 ψ2 (t, ~r),
with c1 and c2 complex numbers is also a solution of the differential equa-
tion.
For the Schrödinger equation we can prove the latter statement. Let
ψ1 (t, ~r) and ψ2 (t, ~r) be solutions of equation 3.1
∂
i~ ψ1 (t, ~r) = Hψ1 (t, ~r), (3.19)
∂t
and
∂
i~ ψ2 (t, ~r) = Hψ2 (t, ~r). (3.20)
∂t

72
 3.3. Hermitian and Self-adjoint

We now find for the Schrödinger equation on ψ(t, ~r) = c1 ψ1 (t, ~r)+c2 ψ2 (t, ~r):

∂ ∂ ∂
i~ (c1 ψ1 (t, ~r) + c2 ψ2 (t, ~r)) = c1 i~ ψ1 (t, ~r) + c2 i~ ψ1 (t, ~r) (3.21)
∂t ∂t ∂t
= c1 Hψ1 (t, ~r) + c2 Hψ2 (t, ~r)
~2
 
= c1 − ∆ + V (~r) ψ1 (t, ~r)
2m
~2
 
+ c2 − ∆ + V (~r) ψ2 (t, ~r)
2m
~2
=− ∆ (c1 ψ1 (t, ~r) + c2 ψ2 (t, ~r))
2m
+ V (~r) (c1 ψ1 (t, ~r) + c2 ψ2 (t, ~r))
= H (c1 ψ1 (t, ~r) + c2 ψ2 (t, ~r)) ,

which concludes the proof.

From the previous proof we can also see that the Hamiltonian is a linear
operator. For all functions ψ(t, ~r), φ(t, ~r) ∈ DH we find that

H (ψ(t, ~r) + φ(t, ~r)) = Hψ(t, ~r) + Hφ(t, ~r). (3.22)

3.3 Hermitian and Self-adjoint

The Hamiltonian (3.2) is Hermitian. This means that for all functions
ψ(t, ~r), φ(t, ~r) ∈ DH we have
Z Z
∗ d
φ(t, ~r) Hψ(t, ~r)d ~r = (Hφ(t, ~r))∗ ψ(t, ~r)dd~r. (3.23)
Rd Rd

2
As the Hamiltonian is linear we can proof this for the two terms, − 2m
~
∆
and V (~r) in the Hamiltonian separately. For the potential energy term we
find
Z Z
∗ d
φ(t, ~r) V (~r)ψ(t, ~r)d ~r = V (~r)φ(t, ~r)∗ ψ(t, ~r)dd~r (3.24)
Rd Rd

73
Chapter 3. Mathematical properties of the Schrödinger
equation

Z
= V (~r)∗ φ(t, ~r)∗ ψ(t, ~r)dd~r (3.25)
Rd
Z
= (V (~r)φ(t, ~r))∗ ψ(t, ~r)dd~r, (3.26)
Rd

where we used that the potential V (~r) is real.

For the kinetic energy term we will use partial integration
Z b Z b 
d b d
u(x) v(x)dx = [u(x)v(x)]a − u(x) v(x)dx. (3.27)
a dx a dx

In our case, both ψ(t, ~r) and φ(t, ~r) are in DH . Specifically we need that
ψ(t, ~r) and φ(t, ~r) as well as their first and second derivative are square
integrable. We thus find for the boundary conditions
 
∂ ∗
lim φ(t, ~r) ψ(t, ~r) = 0, (3.28)
|ri |→∞ ∂ri

and
 
∂
lim φ(t, ~r)∗ ψ(t, ~r) = 0, (3.29)
|ri |→∞ ∂ri

for all but a finite set of values of ~r. The later set of points do not contribute
to the integral over this function and we can neglect the boundary term.
We can now proof the Hermiticity of the Hamiltonian for the kinetic
energy term
Z ∞ 2 Z ∞ 
∗ ∂ ∂ ∗ ∂
φ(t, ~r) 2 ψ(t, ~r)dri = − φ(t, ~r) ψ(t, ~r)dri (3.30)
−∞ ∂ri −∞ ∂ri ∂ri
Z ∞ 2 
∂ ∗
= 2 φ(t, ~
r) ψ(t, ~r)dri
−∞ ∂ri
Z ∞ 2 ∗
∂
= 2 φ(t, ~
r) ψ(t, ~r)dri
−∞ ∂ri

Which concludes the proof that the Hamiltonian is Hermitian.

74
 3.4. Time evolution operator

For a general linear operator A we define the adjoined operator as A†

such that
Z Z
φ(t, ~r)∗ A† ψ(t, ~r)dd~r = (Aφ(t, ~r))∗ ψ(t, ~r)dd~r, (3.31)
Rd Rd

for all φ(t, ~r) in DA and ψ(t, ~r) in DA† . If operator A is Hermitian and the
domains of A and A† are equivalent we call the operator A self-adjoint.
For self-adjoint operators we can write that A = A† .

3.4 Time evolution operator

If we know the wave function at time t then we can calculate the wave
function at time t0 using the time evolution operator.
ψ(t, ~r) = U (t, t0 )ψ(t0 , ~r). (3.32)
Operator U (t, t0 ) turns a wave function at time t0 into a function at time t.
For time independent potentials we can guess the form of U (t, t0 ) as
0
U (t, t0 ) = e−iH(t−t )/~ . (3.33)
With this definition the wave function ψ(t, ~r) = U (t, t0 )ψ(t0 , ~r) is a solution
of the Schrödinger equation for all functions ψ(t0 , ~r)
∂  −iH(t−t0 )/~ 0   0

i~ e ψ(t , ~r) = H e−iH(t−t )/~ ψ(t0 , ~r) . (3.34)
∂t

3.5 Unitarity of time evolution

The time evolution operator is unitary. This means that
U (t, t0 )† U (t, t0 ) = I, (3.35)
with I the identity operator. From this it follows that the conjugate trans-
pose of the time evolution operator is the inverse of the time evolution
operator
U (t, t0 )† = U (t, t0 )−1 . (3.36)

75
Chapter 3. Mathematical properties of the Schrödinger
equation

We can proof this by looking at the time derivative of the integral of

the product of two wave functions ψ(t, ~r) and φ(t, ~r) that each are solutions
of the Schrödinger equation. We first consider the time derivative of the
product of ψ(t, ~r) and φ(t, ~r).
 ∗
∂ ∂ ∂
i~ (φ(t, ~r)∗ ψ(t, ~r)) = φ(t, ~r)∗ i~ ψ(t, ~r) + −i~ φ(t, ~r) ψ(t, ~r) (3.37)
∂t ∂t ∂t
= φ(t, ~r) Hψ(t, ~r) − (Hφ(t, ~r))∗ ψ(t, ~r)
∗

Next we integrate ~r over Rd and change the order of the integral and time
derivative. For this to be allowed we need to assume that ψ(t, ~r) and φ(t, ~r)
as well as their first and second derivative are square integrable. We find
Z Z
∂ ∗
i~ d
φ(t, ~r) ψ(t, ~r)d ~r = φ(t, ~r)∗ Hψ(t, ~r)dd~r (3.38)
∂t Rd Rd
Z
− (Hφ(t, ~r))∗ ψ(t, ~r)dd~r
Rd
= 0,

as the Hamiltonian is Hermitian. If we now insert the time evolution oper-

ator we find
Z
∂
∗
U (t, t0 )φ(t0 , ~r) U (t, t0 )ψ(t0 , ~r) dd~r = 0


i~ (3.39)
∂t Rd
or
Z Z
0 0

∗ 0 0 d
φ(t0 , ~r)∗ ψ(t0 , ~r)dd~r. (3.40)


U (t, t )φ(t , ~r) U (t, t )ψ(t , ~r) d ~r =
Rd Rd

With
Z

∗
U (t, t0 )φ(t0 , ~r) U (t, t0 )ψ(t0 , ~r) dd~r =


(3.41)
Rd Z
φ(t0 , ~r)∗ U (t, t0 )† U (t, t0 )ψ(t0 , ~r)dd~r,
Rd

76
3.6. Uniqueness of the time evolution and determinacy of the
wave function

we conclude that

U (t, t0 )† U (t, t0 ) = I. (3.42)

The time evolution operator is unitary and

U (t, t0 )† = U (t, t0 )−1 (3.43)

0
= U (t , t)

3.6 Uniqueness of the time evolution and deter-

minacy of the wave function
For Hamilton operators that are self-adjoint we have a unitary time evolu-
tion operator and the definition of ψ(t0 , ~r) defines a unique wave function
for all other times

ψ(t, ~r) = U (t, t0 )ψ(t0 , ~r), (3.44)

and from ψ(t, ~r) we can uniquely retrieve ψ(t0 , ~r)

ψ(t0 , ~r) = U (t, t0 )† ψ(t, ~r), (3.45)

The Schrödinger equation is deterministic. Given the wave function at

time t0 the wave function is uniquely determined for all other times and all
previous times. Note that although this is formally true, practically we can
not determine these solutions. This is more than just a technical difficulty.
The Schrödigner equation for systems with macroscopically many particles
turns out to be extremely complex to solve. At the same time, these systems
show very rich and unexpected behaviour. The quantum dynamics of many
particles often defy expectations we achieve from focusing on simpler one
particle systems. In the words of P.W. Anderson: ”The ability to reduce
everything to simple fundamental laws does not imply the ability to start
from those laws and reconstruct the universe” [67].

77
Chapter 3. Mathematical properties of the Schrödinger
equation

3.7 Stationary states and Eigenfunctions

For potentials and thus Hamiltonians that are time independent we can
find solutions of the Schrödinger equation that are a product of a function
of the time coordinate t and position coordinates ~r. With the ansatz

ψ(t, ~r) = χ(t)φ(~r), (3.46)

we find for the Schrödinger equation

∂
φ(~r)i~ χ(t) = χ(t)Hφ(~r). (3.47)
∂t
From which we have
∂
i~ ∂t χ(t) Hφ(~r)
= (3.48)
χ(t) φ(~r)
= E,

with E a time and position independent constant. The latter follows from
the fact that the left hand side of the previous equation is independent of
position and the right hand side is independent of time and the equation
needs to be valid for all positions and time.
We thus find two independent eigenfunction equations
∂
i~ χ(t) = Eχ(t), (3.49)
∂t
and

Hφ(~r) = Eφ(~r). (3.50)

As H is an Hermitian operator we have

Z Z
∗ d
φ(~r) Hφ(~r)d ~r = (Hφ(~r))∗ φ(~r)dd~r = (3.51)
R d R d
Z Z
∗ d
φ(~r) Eφ(~r)d ~r = (Eφ(~r))∗ φ(~r)dd~r =
Rd Rd

78
 3.7. Stationary states and Eigenfunctions

Z Z
E φ(~r)∗ φ(~r)dd~r = E ∗ φ(~r)∗ φ(~r)dd~r =⇒
Rd Rd
E = E∗.
E is a real constant and associated with the energy of the state.
For the time dependent part we can solve the eigensystem equation to
find
i
χ(t) = c1 e− ~ Et , (3.52)
with c1 and arbitrary complex constant.
The wave function is normalized such that
Z
χ(t)∗ φ(~r)∗ χ(t)φ(~r)dd~r = 1. (3.53)
Rd
We can take χ(t) outside the spatial integral and find that
Z
∗
χ(t) χ(t) φ(~r)∗ φ(~r)dd~r = 1 (3.54)
Rd
holds for all times. From which it follows that
χ(t)∗ χ(t) = |c1 |2 , (3.55)
and
Z
φ(~r)∗ φ(~r)dd~r = |1/c1 |2 . (3.56)
Rd
For convenience we normally choose
c1 = 1. (3.57)
There are operators that do not have eigenfunctions that are square-
integrable. These contain important physical systems. In this case, we
can still use the eigenfunctions, but not directly as a representation for
the wave function of a physical system. One solution, which we will see
later, is to use these eigenfunctions as a basis to generate functions that are
square-integrable. Another option is to consider not an infinite-dimensional
system, but contain the entire world in a finite box. One can then look at
the evolution of the solutions as one takes the limit of the box size to
infinite.

79
Chapter 3. Mathematical properties of the Schrödinger
equation

3.8 Continuity of the wave function and its deriva-

tive
The Schrödinger equation for a particle in a Stationary state in real space
with external potential energy V (~r) is

~2
− ∆φ(~r) = (E − V (~r))φ(~r), (3.58)
2m
or
2m
∆φ(~r) = (V (~r) − E)φ(~r). (3.59)
~2
If V (~r) is finite ∀~r we know that the second partial derivative of φ(~r) is
finite. From this, it follows that the wave function φ(~r) and its first partial
derivatives ∇φ(~r) need to be continuous.
Not all potentials we encounter will be finite. The Coulomb potential
acting on a particle due to another charged particle at position ~r0 is of the
form
1
V (~r) ∝ . (3.60)
|~r − ~r0 |
This potential diverges for ~r = ~r0 . A divergent potential leads to a diver-
gence in the second partial derivative of φ(~r) and thus to a discontinuity
in the gradient of φ(~r). The wave function φ(~r) is continuous, but it has
a cusp at the point of the divergence of the potential. With the use of
Gauss divergence theorem, we can relate the cusp in the wave function to
the residue of the pole in the diverging potential.
Without loss of generality, we can shift the coordinate system such that
the divergence we want to investigate is at ~r = 0. In order to determine
the size of the cusp in the wave function we investigate the derivative of
the wave function with respect to the radial coordinate r = |~r|. In order
for the derivative of the wave function to be continuous, it must hold that
∂ ∂
ψ(~r) = − ψ(−~r), (3.61)
∂r ∂r

80
 3.8. Continuity of the wave function and its derivative

or equivalently in spherical coordinates

∂ ∂
ψ(r, θ, φ) = − ψ(r, π − θ, π + φ). (3.62)
∂r ∂r
For this to be valid for any θ and φ, it is must necessarily be true that
Z 2π Z π
1 ∂
ψ(~r) sin(θ)dθdφ = 0, (3.63)
4π 0 0 ∂r r=0

or equivalently
Z 2π Z π
1
lim ∇ψ(~r) · r̂ sin(θ)dθdφ = 0. (3.64)
R0 →0 4π 0 0 r=R0

The integral of the gradient of the wave function in the radial direction over
a sphere with radius R0 around the point ~r = 0 in the limit R0 → 0 has
to be zero for the wave function to be smooth. If this integral is non zero,
there must be a cusp in the wave function.
In order to relate this integral to the divergence of the potential, we use
Gauss’s theorem and relate the surface integral to a volume integral.
Z 2π Z π
1
lim ∇ψ(~r) · r̂ sin(θ)dθdφ = (3.65)
R0 →0 4π 0 0 r=R0
Z 2π Z π Z R0
1 1
lim 2 ∇ · ∇ψ(~r) sin(θ)r2 drdθdφ.
R0 →0 4π R0 0 0 0

As ψ(~r) is a solution of the Schrödinger equation, we know that

~2 ∇2
− ψ(~r) + V (~r)ψ(~r) = Eψ(~r), (3.66)
2m
and thus find that
Z 2π Z π
1 ∂
ψ(~r) sin(θ)dθdφ (3.67)
4π 0 0 ∂r r=0
Z 2π Z π Z R0
1 1 2m
= lim 2 (V (~r) − E) ψ(~r) sin(θ)r2 drdθdφ.
R0 →0 4π R0 0 0 0 ~2

81
Chapter 3. Mathematical properties of the Schrödinger
equation

We now write the potential minus the energy in a Laurent series,

Z
V (~r) − E = + c0 + ~c1 · ~r + . . . , (3.68)
|~r|

to find
Z 2π Z π
1 ∂
ψ(~r) sin(θ)dθdφ (3.69)
4π 0 0 ∂r r=0
Z 2π Z π Z R0  
1 1 2m Z
= lim 2 2
+ c0 + . . . ψ(~r) sin(θ)r2 drdθdφ
R0 →0 4π R0 0 ~ |~
r |
0 0 
1 1 2m 2 4π 3
= lim 2 ~2 2πZR0 + c0 R0 + . . . ψ(~r = 0)
R0 →0 4π R0 3
m
= Z 2 ψ(~r = 0).
~
For a divergent Coulomb potential, there is a cusp in the wave function at
the divergence, proportional to the residue of the pole times m/~2 times
the value of the wave function at the divergence.

82
Chapter 4

Wave mechanics

In this chapter, we explore how we can represent localized particles by wave

functions. We will study the time evolution of an exemplary wave function
representing a localized particle. The time evolution in our example will be
given by the Schödinger equation without an external potential (V (~r) = 0)
in one dimension. The reason to choose one dimension over three is to
simplify the math. This lets us focus on several important and general
concepts of quantum physics. Actually, the mathematical concepts in this
chapter are valid for any type of wave propagation. If one studies the prop-
agation of seismic waves through the earth’s crust or heat conductivity in a
complex medium one would need to study similar examples and equations.

4.1 Time evolution of a delta function

We want to look at the evolution of a particle created at time t = 0 at the
position x = 0 without an external potential (V (x) = 0). We will define
a wave function for such a particle and then look at the time evolution of
this particle. In order to create a wave function, we can start by creating a
density distribution function ρ(t = 0, x) that represents the particle. If we
integrate the distribution that represents this particle over the whole space
(x in this one-dimensional example) the total number of particles we have

83
Chapter 4. Wave mechanics

in our state should be one. We thus require that:

Z ∞
ρ(t, x)dx = 1. (4.1)
−∞

Furthermore we require that at t = 0 the particle can be found at the

position x = 0 and nowhere else. We implement this by the requirement
that

ρ(x) = 0 ∀x 6= 0. (4.2)

The distribution that fulfils these requirements is the Dirac delta function,
or more accurately the Dirac delta distribution. As argued in appendix A,
we can define this distribution as a limiting case of a sequence of functions.
A Gaussian function with the width going to zero would suffice
r
1 − x22
δ(σ, x) = e 2σ . (4.3)
2πσ 2
The limit of σ → 0 should be taken outside the integral (see appendix A)
such that
Z ∞ Z ∞
δ(x)dx = lim δ(σ, x)dx (4.4)
−∞ σ→0 −∞

= 1.

With this we can define the density distribution of our particle at t = 0.

ρ(t = 0, x) = δ(x). (4.5)

We can define several wave-functions that generate such a density distribu-

tion. A simple wave-function would be to take the square root of the Dirac
delta function
p
ψ(t = 0, x) = δ(x). (4.6)

84
 4.1. Time evolution of a delta function

In order to prevent several quirks of the Dirac delta function we will con-
tinue here with one of the functions that can be used to define the delta
function. In particularly we will use the Gaussian function as defined in
equation 4.4. The results we derive in this chapter are valid for any value
of σ, but for a localized particle we are interested in the case where σ is
very small and approaches zero. The wave function at t = 0 of our particle
localized at x = 0 is thus given as
 1/4
1 x2
ψ(t = 0, x) = e− 4σ2 , (4.7)
2πσ 2
for very small values of σ.
We can calculate the wave function at a later time t > 0, by solving the
partial differential equation defined by the Schrödinger equation

∂ ~2
i~ ψ(t, x) = − ∆ψ(t, x), (4.8)
∂t 2m
with the boundary condition for ψ(t, x) at t = 0 as given in equation 4.7.
In order to solve the time evolution we will use a method called spectral
decomposition. For a plane-wave we can relatively easy solve the time
evolution according to the Schrödinger equation. The function

ψk (t, x) = c ei(k x−E t/~) , (4.9)

with E = ~2 k 2 /(2m) and c a complex number (c ∈ C) is a valid solution for

all values of E ≥ 0 or k ∈ R. As the Schrödinger equation is linear we can
write our sought after solution as a sum of plane waves with k dependent
coefficients c.
r Z ∞  2t

1 i k x− ~k
ψ(t, x) = c(k) e 2m
dk. (4.10)
2π −∞

At time t = 0 we have
 1/4 2
r Z ∞
1 − x2 1
2
e 4σ = c(k) eik x dk, (4.11)
2πσ 2π −∞

85
Chapter 4. Wave mechanics

from which we can determine c(k). On the left hand side we have a Gaus-
sian function. On the right hand side we see the Fourier transform of the
function c(k). We now can use the equation that the Fourier transform (See
the appendix C.3.2 for a derivation) of a Gaussian function is a Gaussian
function
Z ∞ r
1 −σ 2 k2 ixk 1 − x22
√ e e dk = e 4σ , (4.12)
2π −∞ 2σ 2

to find
 1/4 √
1 2 2
c(k) = 2
2σ 2 e−σ k (4.13)
2πσ
 2 1/4
2σ 2 2
= e−σ k .
π

Next we substitute the equation for c(k) into the expansion of ψ(t, x) on
plane waves (eq. 4.10) to obtain
r ∞ 1/4
2σ 2 ~tk2
Z   
1 −σ 2 k2 i k x− 2m
ψ(t, x) = e e dk. (4.14)
2π −∞ π

This is a Gaussian integral for which we have (See appendix B)

Z ∞ r
−αk2 +βk π β2
e dk = e 4α . (4.15)
−∞ α

With β = ix and α = σ 2 + i 2m
~t
we find
1/4 s x2
σ2 −

π ( i~t
4 σ 2 + 2m )
ψ(t, x) = i~t
e (4.16)
2π 3 σ2 + 2m
!1/4 x2
1 −
4σ 2 1+ i~t 2
( )
= e 2mσ
i~t 2


2πσ 2 1+ 2mσ 2

86
 4.1. Time evolution of a delta function

With the knowledge of the wave function we can calculate the particle
density ρ(t, x). (i.e. the distribution function of the particle density one
would find if one repeats such an experiment an infinite number of times).

ρ(t, x) = ψ ∗ (t, x)ψ(t, x) (4.17)

x2
− 
v 
1 2 + t2 ~2
u 2 σ 2 σ2
=t e 4m
u 
t2 ~2
2π σ 2 + 4m 2 σ2

We find that for all times the particle distribution will be given by a Gaus-
sian. We defined σ to be the width (standard deviation) of the Gaussian
at time t = 0. We find that at time t the particle distribution spreads and
the width becomes
s
t2 ~2

σ(t) = 2
σ + . (4.18)
4m2 σ 2

The rate at which the width increases is given as

∂σ(t) t~2
= . (4.19)
∂t 4m2 σ 2 σ(t)

The rate depends on the inverse of σ 2 . The more localized we prepare the
initial wave function the faster the particle distribution will spread.
For our particle that we prepared to be at t = 0 exactly at position
x = 0, this result is somewhat problematic. We would like to take the limit
of σ to 0. This would yield for any finite time t > 0 an infinite width. The
particle instantly spreads from a delta distribution to a distribution that
is constant over the whole space and integrates to 1 over the whole space.
This also means that at times slightly larger than zero the particle can be
at a very large distance from the starting position. To do so its velocity
needs to be very large. A very large velocity means the particle must have
a very large energy.
We can actually calculate the energy of the particle. We have seen in eq.
2.5 that for a particle whose wave function is given by a single plane wave

87
Chapter 4. Wave mechanics

i
such that ψk (t, x) = e ~ (px−Et) the time derivative of the wave-function
is proportional to the energy times the wave-function i~∂/(∂t)ψk (t, x) =
Eψk (t, x). In section 3.7 we discussed stationary states and found that the
previous relation for plane waves holds for all stationary states ψstat (t, x). In
particular we found that ψstat (t, x) is an eigenstate of both the Hamiltonian
as well as the time derivative operator:
∂
ı~ ψstat (t, x) = Eψstat (t, x), and (4.20)
∂t
Hψstat (t, x) = Eψstat (t, x).

For both operators the eigenvalue is E, which we associate with the energy
of the state. Furthermore we know that for normalized states we have
Z ∞
ψstat (t, x)∗ ψstat (t, x)dx = 1, (4.21)
−∞

such that
Z ∞
∂
ψstat (t, x)∗ i~ ψstat (t, x)dx = E. (4.22)
−∞ ∂t

An argument based on symmetry will follow in chapter 7, but we now

extend the relation of the energy of a stationary state, as given in equation
4.22 to also apply to states that show a non-trivial time evolution. We take
Z ∞
∂
E= ψ(t, x)∗ i~ ψ(t, x)dx. (4.23)
−∞ ∂t

We can expand ψ(t, x) on plane waves and insert equation 4.10 for the
wave-function
Z ∞ r Z ∞ !∗
2t
 
1 i k x− ~k
E= c(k) e 2m
dk (4.24)
−∞ 2π −∞
r Z ∞ !
~k2 t
 
∂ 1 i k x− 2m
× i~ c(k) e dk dx
∂t 2π −∞

88
 4.1. Time evolution of a delta function

2 −k2 )
 
∞ ∞ ∞
c(k2 )∗ c(k1 ) ~2 k12 i
Z Z Z ~t(k1 2
(k1 −k2 )x− 2m
= √ √ e dk2 dk1 dx.
−∞ −∞ −∞ 2π 2π 2m
R∞ i(k1 −k2 )x dx
We can separate the integral over x and realize that √1
2π −∞ e
√
is the Fourier transform of 1, which (see Appendix C.3.3) is 2π times a
Dirac delta function δ(k1 − k2 ).
Z ∞Z ∞
~2 k12 −i ~t(k12 −k22 )
E= δ(k1 − k2 )c(k2 )∗ c(k1 ) e 2m dk2 dk1 (4.25)
−∞ −∞ 2m
Z ∞
~2 k 2
= c(k)∗ c(k) dk.
−∞ 2m
The first thing to notice is that the energy is conserved, i.e. it does not
depend on time. Actually the reason to define equation 4.23 as the energy
of the state is that it is conserved in time. A particle prepared in a given
Gaussian wave packet has a fixed amount of energy that does not change
2 k2
with time. Furthermore we notice that ~2m is the energy we associated
with a plane wave of momentum k. At the same time c(k) is the expansion
coefficient of our Gaussian wave packet on this plane wave. We find that
2 k2
each plane wave with momentum k contributes an energy ~2m multiplied
∗
by c(k) c(k), i.e. the norm-square of the weight in the expansion of the
Gaussian wave packet on this plane wave.
If we insert the values of c(k) for the expansion of the delta distribution
(equation 4.14) we find
Z ∞r 2
2σ −2σ2 k2 ~2 k 2
E= e dk (4.26)
−∞ π 2m
~2
=
8mσ 2
The sharper we make the initial distribution the larger the energy we have
in the system. If we want to localize p the particle exactly at x = 0 such
that ρ(t = 0, x) = δ(x) (ψ(t = 0, x) = δ(x)) we need an infinite amount
of energy. Our result that a particle prepared with a delta function dis-
tribution for its density instantly spreads to the whole space now becomes
understandable. The particle also has an infinite amount of kinetic energy.

89
Chapter 4. Wave mechanics

There is an other, maybe more physical, conclusion we can draw from

this example. With a finite amount of energy (E), we can not prepare
aqparticle with a Gaussian distribution that is better localized than σ =
~2
8mE .Naturally, we can try different distributions than the Gaussian
distribution. For those one will find different relations between the standard
deviation of the distribution and the energy. As we will show later in this
lecture there is a fundamental limit on how well-localized a particle can be
if you have a finite amount of energy.

4.2 Wave packets

In the previous section, we have seen how one can represent a particle by a
wave. Or better said, by a sum of waves. A particle perfectly localized in
point x = 0 is represented by a Dirac delta distribution. The Dirac delta
distribution can be written as a sum of plane waves with all possible k
vectors weighted equally (equation 4.13.). The energy of a particle repre-
sented by a plane wave is proportional to the wave-vector of the plane wave
2 k2
E = ~2m . For a particle that is defined by a wave function that is the sum
or superposition of several plane waves, the energy is given by the weighted
sum of these plane waves. Realistic particles are sums of plane waves such
that the resulting density distribution is localized. The resulting function
from such a sum we call a wave packet.
In the previous section, we already encountered an example of a wave
packet. In order to define the Dirac delta distribution, we used a Gaus-
sian function and took the limit of its width (σ) to zero. The Gaussian
function is a wave packet. The wave packet used in the previous section
represented a particle with zero velocity. As a function of time, the wave
packet broadened, but it was always centered around zero. In this section,
we will look at the movement of different wave packets and introduce the
concept of group and phase velocity. We will furthermore look at wave
packets of different shapes and use a moment expansion to characterize
these. There is a special relation between the shape of the wave packet in
real space and momentum space. In the last subsection on wave packets,

90
 4.2. Wave packets

we will discuss these relations in more detail using a moment expansion of

the wave function in real and momentum space.

4.2.1 The wave function in position and momentum space

We can write any wave function as a sum over plane waves
r Z ∞
1
ψ(t, x) = c(k)ei(kx−E(k)t/~) dk, (4.27)
2π −∞
2 2 k
with the energy - wave vector dispersion relation E(k) = ~2m . We define
c(k) as the expansion coefficient for the wave function ψ(t, x) on plane
waves. We can obtain c(k) by a Fourier transform of ψ(t, x).
r Z ∞ Z ∞Z ∞
1 −ikx 1 0 0
ψ(t, x)e dx = c(k 0 )ei(k x−E(k )t/~) e−ikx dk 0 dx
2π −∞ 2π
Z ∞ −∞ −∞ Z ∞
0 1 0
= c(k 0 )e−iE(k )t/~ ei(k −k)x dxdk 0
2π
Z−∞∞
−∞
0
= c(k 0 )e−iE(k )t/~ δ(k 0 − k)dk 0
−∞
= c(k)e−iE(k)t/~
≡ φ(t, k). (4.28)

We have introduced φ(t, k) as the wave function in momentum space. The

wave function in momentum space (φ(t, k)) is a dual representation of the
wave function in real space ψ(t, x). We will discuss this in more detail in a
later chapter on linear algebra 6.
In order to get a feeling for the wave function in real and momentum
space, we can have a look at two examples. In figure 4.1 we show the time
evolution of two different shaped wave packets. We start with two wave
packets that at t = 0 are given by real functions. Within one panel the top
row represents a Gaussian distribution in real space, and the bottom row
represents a rectangular distribution in real space. The left column shows
the wave function in real space (ψ(t, x)), and the right column shows the

91
Chapter 4. Wave mechanics

t=0 t=0.1 t=0.2 t=0.5

1.0 1.0 1.0 1.0
, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))
0.6 0.6 0.6 0.6

0.4 0.5 0.4 0.5 0.4 0.5 0.4 0.5

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2

2

2
2

2
-0.4 -0.5 -0.4 -0.5 -0.4 -0.5 -0.4 -0.5
|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.6 -0.6 -0.6 -0.6
-1.0 -1.0 -1.0 -1.0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
x k x k x k x k

1.0 1.0 1.0 1.0

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))
0.6 0.6 0.6 0.6

0.4 0.5 0.4 0.5 0.4 0.5 0.4 0.5

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2

2

2
2

2
-0.4 -0.5 -0.4 -0.5 -0.4 -0.5 -0.4 -0.5
|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.6 -0.6 -0.6 -0.6
-1.0 -1.0 -1.0 -1.0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
x k x k x k x k

t=1 t=2 t=5 t=10

1.0 1.0 1.0 1.0
, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))
0.6 0.6 0.6 0.6

0.4 0.5 0.4 0.5 0.4 0.5 0.4 0.5

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2

2

2
2

2
-0.4 -0.5 -0.4 -0.5 -0.4 -0.5 -0.4 -0.5
|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.6 -0.6 -0.6 -0.6
-1.0 -1.0 -1.0 -1.0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
x k x k x k x k

1.0 1.0 1.0 1.0

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))
0.6 0.6 0.6 0.6

0.4 0.5 0.4 0.5 0.4 0.5 0.4 0.5

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2

2

2
2

2
-0.4 -0.5 -0.4 -0.5 -0.4 -0.5 -0.4 -0.5
|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.6 -0.6 -0.6 -0.6
-1.0 -1.0 -1.0 -1.0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
x k x k x k x k

Figure 4.1: Time evolution of differently shaped wave packets with zero
momentum.

wave function in momentum space (φ(t, k)). The different panels show the
wave function at different times. The black lines show the norm of the wave
function, the blue and orange the real and imaginary part. For plotting we
took σ = 1, m = 1, ~ = 1 and t runs from 0 to 10. Both wave packets
spread over time in real space. The wave function in momentum space does
not broaden. φ(t, k) changes its phase for each k value, but not its norm
|φ(t, k)|, as a function of t.

4.2.2 Gaussian wave packet with finite momentum

In the previous section we discussed wave packets representing particles

with zero velocity. The wave packet represents a particle with zero velocity
in the sense that the center of the wave packet does not move. The wave
packet does broaden in time. In order to represent a particle with finite
velocity we need to give it a finite momentum, p0 = k0 ~. This we can do
by replacing c(k) by c(k − k0 ) in equation 4.13. After substituting k by
k − k0 we still have a valid solution of the Schrödinger equation. As we’ve
discussed before, any superposition of plane waves is, for the Schrödinger
equation with zero potential (V (x) = 0), a valid solution. A Gaussian wave

92
 4.2. Wave packets

t=0 t=0.1 t=0.2 t=0.5

1.0 1.0 1.0 1.0
, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))
0.6 0.6 0.6 0.6

0.4 0.5 0.4 0.5 0.4 0.5 0.4 0.5

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2

2

2
2

2
-0.4 -0.5 -0.4 -0.5 -0.4 -0.5 -0.4 -0.5
|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.6 -0.6 -0.6 -0.6
-1.0 -1.0 -1.0 -1.0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
x k x k x k x k

t=1 t=2 t=3 t=4

1.0 1.0 1.0 1.0
, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))

, Re(ϕ(t,k)), Im(ϕ(t,k))
0.6 0.6 0.6 0.6

0.4 0.5 0.4 0.5 0.4 0.5 0.4 0.5

0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2

2

2
2

2
-0.4 -0.5 -0.4 -0.5 -0.4 -0.5 -0.4 -0.5
|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)

|ϕ(t,k)
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.6 -0.6 -0.6 -0.6
-1.0 -1.0 -1.0 -1.0
-10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10 -10 -5 0 5 10
x k x k x k x k

Figure 4.2: Time evolution of a Gaussian wave packet in free space with
momentum k = 5, mass m = 1, ~ = 1, and width given by σ = 1. The time
evolution runs from 0 to 5.

packet with momentum p0 = k0 ~ is thus given by the wave functions

1/4
2σ 2

2 (k−k 2 ~k2 t
φ(t, k) = e−σ 0)
e−i 2m , (4.29)
π

in momentum space and

∞
r Z
1
ψ(t, x) = φ(t, k)eikx dk (4.30)
2π −∞
!1/4 x2 −4ik0 xσ 2 +2ik0
2 tσ 2 ~/m
1 − i~t
= e
2
4σ 1+
2mσ 2
( ) ,
i~t 2


2πσ 2 1+ 2mσ 2

in real space.
The wave function of a particle represented by a Gaussian wave packet
with finite momentum k0 looks surprisingly complicated. If we compare
the wave packet with a finite momentum to the wave packet with zero
momentum k0 = 0 we find that firstly, both are represented by a Gaussian
with modified width that spreads over time. The wave packet at finite
momentum has an additional phase proportional to k0 x for one term and
to k02 t for the second term. The equation simplifies if we look at the particle
distribution

ρ(t, x) = ψ(t, x)∗ ψ(t, x) (4.31)

93
Chapter 4. Wave mechanics

 2
k0 ~
x− m t
− 
2

1 2σ 2 1+ ( t~
2mσ 2
)
=√ q
2 e .
t~
2πσ 1 + 2mσ 2

Here we see that the particle is represented by a Gaussian distribution that

moves with a velocity

v = k0 ~/m (4.32)
= p/m.

This is the velocity momentum relation one would expect for a classical
particle. An example of such a wave packet can be seen in figure 4.2.

4.2.3 Group and phase velocity

In the previous section we have seen that a Gaussian wave packet for a
particle with mass m 6= 0 with φ(t, k) centered around k0 travels in real
space with the velocity v = ~k/m = p/m. This is the classical result one
should expect. This result is not obvious. A plane wave representing a
particle is defined as
i
ψ(t, x) = e ~ (px−Et) . (4.33)

The velocity we assign to this wave is v = E/p. For massless particles, we

have the energy-momentum relation E = pc with c the speed of light. This
yields that the velocity is equal to v = E/p = c. The wave travels with the
speed of light. For particles with a mass, we have E = p2 /(2m). This yields
that the velocity is equal to v = E/p = p/(2m). For a classical particle we
should expect v = p/m for its velocity and not v = p/(2m). The velocity
defined directly from a plane wave in this example is only half the classical
velocity one would expect. We have seen that once we look at the particle
distribution we do find the correct velocity. We can differentiate between
two velocities for a wave packet. The phase velocity and the group velocity.
The phase velocity is the velocity that corresponds to a single wave and is

94
 4.2. Wave packets

defined as

vp = E/p. (4.34)

This is not the velocity with which the envelope of the wave packet moves.
The latter is called the group velocity (vg ), which we will discuss in more
detail below.
The difference between the phase and group velocity can be seen in
figure 4.2. Figure 4.2 shows the propagation of a wave packet. In black, we
see the particle distribution. It moves with a constant velocity of vg = p/m
as one would expect for a classical particle. The real and the imaginary
part of the wave function seem to move slower than the norm of the wave
function. If one follows a zero crossing of either the real or imaginary part
one finds that it only moves with half the velocity vp = p/(2m).
We can derive the group velocity of a wave packet for an arbitrary
shaped wave and arbitrary dispersion relation E(p) or ω(k). A general
wave packet with a wave vector centered around k0 has the form
r Z ∞
1
ψ(t, x) = c(k − k0 )ei(kx−ω(k)t) dk. (4.35)
2π −∞

In order to derive the group velocity we make a Taylor expansion of ω(k)

in k around k0
∞
∂ω(k) X (k − k0 )n ∂ n ω(k)
ω(k) = ω(k0 ) + (k − k0 ) + . (4.36)
∂k k0 n! ∂k n k0
n=2

Substituting this into the equation of ψ(t, x) yields

 
∞
r Z ∂ω(k)
1 i(k−k0 ) x− t
ψ(t, x) = ei(k0 x−ω(k0 )t)
∂k
c(k − k0 )e k0
(4.37)
2π −∞
n
n 1 ∂ ω(k)
P∞
−i n=2 (k−k0 ) n! ∂kn t
×e k0 dk

We see that this equation has an x dependence in two different places.

95
Chapter 4. Wave mechanics

The first factor that depends on x gives a factor

ei(k0 x−ω(k0 )t) . (4.38)

This describes a monochrome plane wave with phase velocity vp = ω(k0 )/k0 .
The real and the imaginary part of ψ(t, x) form waves that travel through
the wave packet with velocity vp . For the particle distribution, this phase
modulation does not matter.
The second factor that dependents on x occurs inside the Fourier trans-
form of c(k − k0 ). This factor modulates the wave packet. The space
dependent factor is given as
 
∂ω(k)
i(k−k0 ) x− ∂k
t
e k0
. (4.39)

The velocity of this wave is defined as the group velocity and given by

∂ω(k)
vg = , (4.40)
∂k k0

or in energy and momentum instead of wave vector and angular frequency

variables
∂E(p)
vg = , (4.41)
∂p p0

Besides the terms defining the group and phase velocity of the wave
packet, there is a term that depends on the second and higher derivatives
of the energy-momentum dispersion. These terms are responsible for the
deformation of the shape of the wave packet in time. For systems that
have a linear energy-momentum dependence, i.e. E = c0 + c1 p one can
make wave packets that stay localized. For other energy-momentum rela-
tions, the wave packet will generally deform over time. Assuming different
energy-momentum relations than E = pc or E = p2 /(2m) is not just a
mathematical exercise. A wave traveling through a material will gener-
ally have a different energy-momentum dispersion than a wave traveling

96
 4.2. Wave packets

through vacuum. This can either be light traveling through a dielectric

material with an index of refraction different from n = 1, or an elec-
tron traveling through a metal or semiconductor. For electronic transport
through solids, one can engineer materials with different electron energy-
momentum dispersion relations to optimize specific electronic properties.
The electron energy-momentum dispersion relations in a solid are often
called their band-structure. A web search on ”band structure engineering”
yields many results from optimized light-emitting diodes and solar cells to
sensitive measurement devices.

4.2.4 Moment expansion in position and momentum space

A quantum system is uniquely defined by its wave function ψ(t, x) in posi-
tion space, or equally well by its wave function in momentum space φ(t, k).
These two functions are Fourier transforms of each other, so if one knowns
one the other is known and vice versa
r Z ∞
1
ψ(t, x) = φ(t, k)eikx dk ⇐⇒ (4.42)
2π −∞
r Z ∞
1
φ(t, k) = ψ(t, x)e−ikx dx.
2π −∞
The particle distribution in position and momentum space is given as

ρx (t, x) = ψ(t, x)∗ ψ(t, x) (4.43)

∗
ρk (t, k) = φ(t, k) φ(t, k).

We can represent these functions by their moments.

Z ∞
µ(n)
x (t) = xn ρx (t, x)dx (4.44)
Z−∞
∞
(n)
µk (t) = k n ρk (t, k)dk.
−∞

The zeroth moment relates to the total particle density in the system,
which is one in our case. The first moment gives the center of the particle

97
Chapter 4. Wave mechanics

distribution or the expected value for its position / momentum. The higher
moments relate to the spread of the distribution. If we define the central
moment with respect to the expected value as
Z ∞
(n)
µ̃x (t) = (x − µ(1) n
x ) ρx (t, x)dx (4.45)
Z−∞
∞
(n) (1)
µ̃k (t) = (k − µk )n ρk (t, k)dk,
−∞

then the second central moment is the moment of inertia of the distribution,
or equivalent, its variance. The third and fourth moments are referred to
as the skewness and kurtosis of the distribution.
In the following we will assume that both ρx (t, x) and ρk (t, k) tend to
zero fast enough as |x| → ∞ or |k| → ∞ such that the integrals are well
behaved and the moments of the distribution finite.

Zeroth moment and wave function normalization in position and

momentum space
(0)
The zeroth moment µx (t) in position space is taken to be 1 by choosing
the appropriate normalisation of the wave function. In this way ρ(t, x)
represents the particle distribution. Note that, for massive particles, we
equally well could have normalised our wave functions to m in which case
ρ(t, x) would represent the mass distribution. For charged particles one
can normalise ρ(t, x) to the charge (e for electrons) and have the charge
(0)
distribution. In momentum space we can calculate µk as
Z ∞
(0)
µk (t) = ρk (t, k)dk (4.46)
−∞
Z ∞
= φ(t, k)∗ φ(t, k)dk
−∞
Z ∞Z ∞Z ∞
1 0
= ψ(t, x0 )∗ ψ(t, x)eik(x−x ) dx0 dxdk
2π
Z−∞ −∞ −∞
∞ Z ∞
= ψ(t, x0 )∗ ψ(t, x)δ(x0 − x)dx0 dx
−∞ −∞

98
 4.2. Wave packets

Z ∞
= ψ(t, x)∗ ψ(t, x)dx
−∞
=1

We thus find the same normalization for the particle momentum distribu-
tion (ρk (t, k)) as well as for the particle position distribution (ρx (t, x)).

First moment and expectation value of position and momentum

The first moment gives the center of gravity or expectation (expected) value
of the distribution. The expectation value of the position of a particle is
Z ∞
µ(1)
x (t) = xρx (t, x)dx. (4.47)
−∞

When the particle distribution is given by a localized function, i.g. a Gaus-

sian, we find that the expectation value is at the center of this distribution
where ρ(t, x) is large. This intuitive relation between expectation value be-
ing at a position where one finds a large value of the particle distribution
function must not be true. If the particle is in a state that is given by
two Gaussians centered around +x0 and −x0 with x0 much larger than the
width of the Gaussians x  σ one finds that ρx (t, x = 0) ≈ 0 at the same
(1)
time the expectation value of the position is µx = 0.
The expectation value for the momentum is
Z ∞
(1)
µk (t) = kρk (t, k)dk. (4.48)
−∞

In order to get a better understanding of this equation we can transform it

to position space
Z ∞
(1)
µk (t) = kρk (t, k)dk (4.49)
−∞
Z ∞
= kφ(t, k)∗ φ(t, k)dk
−∞

99
Chapter 4. Wave mechanics

Z ∞ Z ∞ Z ∞
1 0
= kψ(t, x0 )∗ ψ(t, x)eik(x −x) dx0 dxdk
2π
Z−∞ −∞
∞ Z ∞
−∞
Z ∞
1 0 ∗ 0
= ψ(t, x ) ψ(t, x) keik(x −x) dkdx0 dx.
−∞ −∞ 2π −∞

Where in the last step we changed the order of integration.

We now use that the Fourier transform of a polynomial n
√ xn is given by the
nth derivative of the Dirac delta distribution times 2πi . See appendix
C.3.4 and A for a derivation and definition of the derivative of the Dirac
delta distribution. The equations we need are
1
Z ∞ √
xn e−ikx dk = 2πin δ (n) (k), (4.50)
2π −∞
1
Z ∞ √
k n eikx dk = 2πin δ (n) (−x),
2π −∞
and
Z ∞ ∞
∂n
Z
(n) n
δ (x)f (x)dx = (−1) δ(x) f (x)dx. (4.51)
−∞ −∞ ∂xn
With the further substitution of x → u + x0 we have,
Z ∞Z ∞
(1)
µk (t) = ψ(t, x0 )∗ ψ(t, u + x0 )iδ (1) (u)dudx0 (4.52)
Z−∞ −∞
∞ Z ∞
∂
= ψ(t, x0 )∗ δ(u)(−i) ψ(t, u + x0 )dudx0
−∞ −∞ ∂u
Z ∞Z ∞  
0 ∗ ∂
= ψ(t, x ) δ(u)(−i) ψ(t, v) dudx0
∂v u+x 0
Z−∞
∞
−∞
∂
= ψ(t, x0 )∗ (−i) 0 ψ(t, x0 )dx0 .
−∞ ∂x
We can calculate the expectation value of the momentum p = ~k for a
particle using the equation in momentum space as
Z ∞
(1)
~µk (t) = φ(t, k)∗ ~kφ(t, k)dk (4.53)
−∞

100
 4.2. Wave packets

or using the wave function in real space as

Z ∞
(1) ~ ∂
~µk (t) = ψ(t, x)∗ ψ(t, x)dx. (4.54)
−∞ i ∂x
This result should not surprise us. When we introduced the Schrödinger
equation we motivated the equation by the introduction of the momentum
operator
~ ∂
p= . (4.55)
i ∂x
We just reproduced that statement via the Fourier transform of the ex-
pected value of momentum calculated in momentum space.

Second moment of position and momentum

The second moment of the position and momentum are defined as
Z ∞
(2)
µx (t) = ψ(t, x)∗ x2 ψ(t, x)dx, (4.56)
−∞

and
Z ∞
(2)
µk (t) = φ(t, k)∗ k 2 φ(t, k)dk (4.57)
−∞
Z ∞
∂2
 
= ψ(t, x)∗ − 2 ψ(t, x) dx.
−∞ ∂x
The second central moments can be calculated as
µ̃(2) (2) (1)
x (t) = µx (t) − (µx (t))
2
(4.58)
(2) (2) (1)
µ̃k (t) = µk (t) − (µk (t))2 .
These are known as the variances of the distribution and contain infor-
mation on the squared width of the distribution. The square root of the
variance is the standard deviation of the distribution
q
(2)
σx (t) = µ̃x (t), (4.59)

101
Chapter 4. Wave mechanics

and
q
(2)
σk (t) = µ̃k (t). (4.60)
For a Gaussian wave packet with zero velocity (k0 = 0), normalisation
(0) (1)
µx = 1, first moment µx = 0, and standard deviation σx at t = 0 we
have
 2
1 − 21 x
ρx (t = 0, x) = √ e σx
, (4.61)
σx 2π
and
 2
1 − 21 k
ρk (t = 0, k) = √ e σk
, (4.62)
σk 2π
with
1
σk = . (4.63)
2σx
The relation between σx and σk for general times and wave packets of
different shapes are known as Heisenberg’s uncertainty relations. We will
discuss them after we briefly note the calculation of general moments of the
particle distribution function.

nth moment of position and momentum

For the nth moment of the particle position and momentum distribution
we can follow a similar derivation as we had for the first moment. The
result for the moments of the position distribution is
Z ∞
(n)
µx (t) = ψ(t, x)∗ xn ψ(t, x)dx. (4.64)
−∞

For the momentum distribution we find

Z ∞
(n)
µk (t) = φ(t, k)∗ k n φ(t, k)dk (4.65)
Z−∞
∞
∂n
= ψ(t, x)∗ (−i)n n ψ(t, x)dx
−∞ ∂x

102
 4.2. Wave packets

4.2.5 Heisenberg’s uncertainty relation

The wave function in position space and momentum space are related by a
Fourier transform. One thus should expect that there is a relation between
the different moments of these functions. There are several, but the best
known is the Heisenberg uncertainty relation relating the second moment
of the position and momentum distribution to each other. Heisenberg’s
uncertainty relation states that
(2) 1
µ̃(2)
x (t)µ̃k (t) ≥ . (4.66)
4
The proof of this inequality becomes much simpler with the use of tech-
niques from linear algebra. We will non the less present it here. The
statement that the product of the second central moment of a function
and its Fourier transform is larger than 1/4 is a mathematical property of
Fourier transforms. Any system described by a wave equation is bound by
this inequality. This becomes clear if one uses the mathematical tools used
to describe wave mechanics. So let us evaluate a few integrals and prove
Heisenberg’s uncertainty relation.
We will start the proof with the Cauchy-Schwarz inequality (see ap-
pendix D).
Z ∞ Z ∞ Z ∞ 2
∗ ∗
f (x)f (x)dx g (x)g(x)dx ≥ f ∗ (x)g(x)dx . (4.67)
−∞ −∞ −∞

For f (x) and g(x) we will take

 
f (x) = x − µ(1)
x ψ(t, x), (4.68)

and
 
∂ (1)
g(x) = −i − µk ψ(t, x). (4.69)
∂x
such that
Z ∞ Z ∞
∗ ∗
f (x)f (x)dx = (x − µ(1) (1)
x )ψ (t, x)(x − µx )ψ(t, x)dx (4.70)
−∞ −∞

103
Chapter 4. Wave mechanics

Z ∞
= ψ ∗ (t, x)(x2 − 2µ(1) (1) (1)
x x + µx µx )ψ(t, x)dx
−∞
= µ̃(2)
x ,

and
Z ∞ Z ∞   ∗
∗ ∂ (1)
g (x)g(x)dx = −i − µk ψ(t, x) (4.71)
−∞ −∞ ∂x
  
∂ (1)
−i − µk ψ(t, x) dx
∂x
Z ∞   
∗ ∂ (1) ∂ (1)
= ψ (t, x) −i − µk −i − µk ψ(t, x)dx
−∞ ∂x ∂x
!
∂ 2
Z ∞    
∗ (1) ∂ (1) (1)
= ψ (t, x) −i − 2µk −i + µk µk
−∞ ∂x ∂x
ψ(t, x)dx
(2)
= µ̃k ,

whereby for the second step we used that the momentum operator is Her-
mitian, which we can proof with the use of integration by parts.
To proof Heisenbergs uncertainty relation we note that the Cauchy-
Schwarz inequality has on the right hand side the modulus squared of a
complex number. This can be expressed as

|z|2 = (Re(z))2 + (Im(z))2 (4.72)

≥ (Im(z))2
z − z∗ 2
 
≥ ,
2i
or
R∞ R∞ !2
∞ 2 ∗ ∗
−∞ f (x)g(x)dx − −∞ g (x)f (x)dx
Z
f ∗ (x)g(x)dx ≥ . (4.73)
−∞ 2i

104
 4.2. Wave packets

The integrals on the right hand side are equal to

Z ∞
f ∗ (x)g(x) − g ∗ (x)f (x)dx (4.74)
−∞
Z ∞
= ψ ∗ (t, x)x2 ψ(t, x) − ψ ∗ (t, x)x2 ψ(t, x)dx
−∞
∞
∂ 2 ∂ 2
Z    
∗ ∗
+ ψ (t, x) −i ψ(t, x) − ψ (t, x) −i ψ(t, x)dx
−∞ ∂x ∂x
Z ∞    
∂ ∂
+ ψ ∗ (t, x)x −i ψ(t, x) − ψ ∗ (t, x) −i xψ(t, x)dx
−∞ ∂x ∂x
Z ∞    
∗ ∂ψ(t, x) ∂xψ(t, x)
=i ψ (t, x) −x + dx
−∞ ∂x ∂x
Z ∞    
∗ ∂ψ(t, x) ∂ψ(t, x)
=i ψ (t, x) −x + ψ(t, x) + x dx
∂x ∂x
Z−∞
∞
=i ψ ∗ (t, x)ψ(t, x)dx
−∞
= i.
Bringing it all together we have
Z ∞ 2
(2) ∗
µ̃(2)
x µ̃ k ≥ f (x)g(x)dx (4.75)
−∞
R∞ ∗ R∞ !2
− ∗ (x)f (x)dx
−∞ f (x)g(x)dx −∞ g
≥
2i
 2
i
≥
2i
1
≥ .
4
If we take the square root of the previous equation we find a relation
between the standard deviation of the position and momentum distribution
functions:
1
σx σk ≥ . (4.76)
2

105
Chapter 4. Wave mechanics

Using the momentum instead of the wave vector in our equations we have
~
σx σp ≥ . (4.77)
2
Instead of σ for the standard deviation of the particle or momentum dis-
tribution several authors use ∆x and ∆p as variables.
If we prepare a macroscopic large number of particles at time t = 0 in a
given wave function we can measure at a later time the particle momentum
and position distribution. It does not matter how we prepare our state,
we will always find that the standard deviation of the position distribution
times the standard deviation of the momentum distribution is larger than
~
2 . In particular, this is true at t = 0, i.e. right at the time where we
prepared our particle wave function. One can never prepare an ensemble of
particles with a product of the standard deviation of energy and momentum
distribution that is smaller than ~2 . For a single particle one can always
think of repeating the experiment nearly infinite many times. This again
will produce an ensemble of particles and for these, we can define a position
and momentum distribution. The product of the standard deviation of
these distributions built from single events will be larger than ~2 .

106
Chapter 5

Wave mechanics in
exemplary potentials

In this chapter, we evaluate the Schödinger equation for a single particle

propagating in an external potential. We will discuss several effects such
as reflection and transmission on potential steps, tunnelling through a po-
tential barrier, the continuity of the wave function and its derivative at
diverging potentials, or trapped particles in a box, well, or harmonic oscil-
lator potential. The aim is to become fully acquainted with the possible
quantum solutions of a propagating particle. The examples in this chap-
ter will be discussed in one spatial dimension. The reason to choose one
dimension over three is to simplify the math. One needs to be slightly
careful when making such simplifications. There are several effects unique
to one-dimensional systems that do not occur in higher dimensions. The
examples chosen here are safe in that respect and what we learn from these
one-dimensional examples can be transferred to other dimensions.

107
Chapter 5. Wave mechanics in exemplary potentials

5.1 Potential step

The first example we consider is the
transmission and reflection of a par-
ticle from a potential step. We take V(x)

V0
(
0 x0 ≤ 0
V (x) = (5.1)
V0 x0 > 0. x = x0

We introduced the variable dk to Figure 5.1: The potential used in the

simplify the notation later on and first example.
define

dk 2 ~2
V0 = . (5.2)
2m
Solutions to the Schrödinger equation that represent physical particles
localized in space are wave packets. If we locate the wave packet far away
from the step in the potential, i.e. at µ  x0 or µ  x0 , with µ the center,
or first moment, of the wave packet, then we should find similar solutions
for the wave packet and its time evolution as we did for the case without
a potential step 4. As a possible solution, we can assume a Gaussian wave
packet with a width σ  |x − x0 | and finite momentum k0 . If we choose
the centre of the wave packet to be left of the potential step (µ  x0 ) and
k0 > 0 the particle should travel towards the potential step and at time
t = µ2m
~k0 pass through or reflect from the step potential. Whereby we used
that the center of the wave will propagate with the group velocity vg = km0~
.
In order to understand what happens when the particle hits the potential
step, we can have a look at the two limiting cases where µ  x0 or µ  x0 .
The kinetic energy of the wave packet on the left-hand side of the potential
step is

k02 ~2
E= . (5.3)
2m

108
 5.1. Potential step

0.6 0.6 0.6 0.6 0.6 0.6 0.6

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
32 ℏ2 32 ℏ2 32 ℏ2 32 ℏ2 32 ℏ2 32 ℏ2 32 ℏ2
V= V= V= V= V= V= V=
2m 2m 2m 2m 2m 2m 2m
0.4 0.4 0.4 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2 -0.2 -0.2 -0.2

2

2
-0.4 -0.4 -0.4 -0.4 -0.4 -0.4 -0.4
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.6 -0.6 -0.6 -0.6 -0.6 -0.6 -0.6

-10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30

x x x x x x x

0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
V= V= V= V= V= V= V=
2m 2m 2m 2m 2m 2m 2m
0.4 0.4 0.4 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2 -0.2 -0.2 -0.2

2

2
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.4 -0.4 -0.4 -0.4 -0.4 -0.4 -0.4

-0.6 -0.6 -0.6 -0.6 -0.6 -0.6 -0.6

-10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30

x x x x x x x

0.6 72 ℏ2
0.6 72 ℏ2
0.6 72 ℏ2
0.6 72 ℏ2
0.6 72 ℏ2
0.6 72 ℏ2
0.6 72 ℏ2
, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
V= V= V= V= V= V= V=
2m 2m 2m 2m 2m 2m 2m
0.4 0.4 0.4 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2 -0.2 -0.2 -0.2

2

2
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.4 -0.4 -0.4 -0.4 -0.4 -0.4 -0.4

-0.6 -0.6 -0.6 -0.6 -0.6 -0.6 -0.6

-10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30 -10 0 10 20 30

x x x x x x x

Figure 5.2: Time evolution of a wave packet originally centred around x = 0

traveling with k0 = 5 across a potential step at x = 10 with a potential
give by dk = 3, 5, 7 from top to bottom respectively. m, ~, and σ(t = 0) for
the Gaussian wave packet are all set to 1.

Once the wave packet traveled past the potential step we should expect
that the kinetic energy is reduced by the potential energy of the step. The
kinetic energy of the wave packet past the potential step becomes
(k02 − dk 2 )~2
E − V0 = . (5.4)
2m
The central wave vector of the wave packet once the wave packet is centred
around a position µ > x0 therefor is
q
k1 = k02 − dk 2 . (5.5)

This requires that E > V0 or k0 > dk for the wave to propagate past the
potential step. For the case where V0 > E one would classically expect the
particle to reflect from the potential step. We will discuss the quantum
mechanical solution for both cases below.
Before we will investigate analytical solutions we will have a look at
some numerical results. In figure 5.2 we show numerical solutions to the
Schrödinger equation for an initial Gaussian wave packet traveling across a
potential step. The different columns represent different times. The size of
the potential step increases from top to bottom, with E > V0 for the top
panels, E = V0 for the middle panels, and E < V0 for the bottom panels. In

109
Chapter 5. Wave mechanics in exemplary potentials

all cases, we see that besides a transmitted wave packet there is a reflected
wave packet. This is similar to the transmission of light through an interface
between two materials with different dielectric constants. For the solutions
of the Maxwell equations, one finds for light propagating through such a
boundary a transmitted and reflected wave. For potential steps with a
potential energy smaller than the kinetic energy of the wave packet, we
find a large transmitted wave and a small reflected wave. The larger the
potential step the larger the reflected wave. For potential steps with a
potential energy larger than the kinetic energy of the original wave packet,
one finds complete reflection and an exponentially decaying wave above the
potential step.
In order to quantify the amount of transmission and reflection we can
have a look at the stationary states of the Schrödinger equation. In chapter
4 we wrote our wave packets as linear combinations of plane waves. Without
a potential a plane wave is a stationary solution of the Schrödinger equation.
That is, the plane wave is an eigenstate of the Hamiltonian. With a position
dependent potential, plane waves are no longer stationary solutions of the
Schrödinger equation. Following section 3.7 we can find the stationary
states of the Schrödinger equation and use these to determine the particle
currents related to these stationary states. This is a similar exercise as one
would encounter in a lecture on electro magnetism where one calculates the
transmittance and reflectivity of an electro magnetic wave from a surface
of a dielectric material. The Schrödinger equation for a potential step is
∂
i~ ψ(t, x) = Hψ(t, x), (5.6)
∂t
with
~2
H=− ∆ + V (x), (5.7)
2m
and V (x) defined in equation 5.1. We can separate the time and position
dependence and write

ψ(t, x) = χ(t)φ(x). (5.8)

110
 5.1. Potential step

This leads to two eigenvalue equations (see section 3.7). The solution for
χ(t) is
i
χ(t) = e− ~ Et (5.9)

and φ(x) is a solution of the eigenvalue equation

~2
− ∆φ(x) + V (x)φ(x) = Eφ(x). (5.10)
2m
For x < x0 the solution of 5.10 are plane waves of the form

φ− (x) = c1 eik0 x + c2 e−ik0 x (5.11)

with
2mE
k02 = , (5.12)
~2
and c1 and c2 complex constants. For x > x0 the solution of 5.10 are plane
waves of the form

φ+ (x) = c3 eik1 x + c4 e−ik1 x (5.13)

with
2m(E − V0 )
k12 = , (5.14)
~2
= k02 − dk 2 ,

and c3 and c4 complex constants. For the case where E > V0 , k12 is posi-
tive and the wave propagates on both sides of the potential step. For the
case where dk > k0 , k12 is negative and k1 becomes imaginary. In that case
φ+ (x) becomes a linear combination of an exponentially decaying and expo-
nentially growing function. The part that exponentially grows for x → ∞
can not be used to form square integrable wave packets and will get a
zero coefficient. The exponentially decaying wave is physical and shows the

111
Chapter 5. Wave mechanics in exemplary potentials

penetration depth of the particle when it reflects off a surface or boundary

between two media.
In order to determine c1 , c2 , c3 and c4 we can use that the first and
second derivative of φ(x) need to be continues (see section 3.8). The
Schrödinger equation states that the second derivative of φ(x) is finite.
This can only be realized if the first derivative and the function itself are
continues. This leads to the equations

φ− (x0 ) = φ+ (x0 ) =⇒ (5.15)

ik0 x0 −ik0 x0 ik1 x0 −ik1 x0
c1 e + c2 e = c3 e + c4 e ,

and
∂ ∂
φ− (x) = φ+ (x) =⇒ (5.16)
∂x x=x0 ∂x x=x0
−ik0 x0
k0 (c1 eik0 x0 − c2 e ) = k1 (c3 eik1 x0 − c4 e−ik1 x0 ).

Solving these sets of equations we find for each energy two possible
solutions
k0 − k1
c2 = c1 ei2k0 x0 , (5.17)
k0 + k1
2k0
c3 = c1 ei(k0 −k1 )x0 ,
k0 + k1
c4 = 0,

and
k0 + k1
c2 = c1 ei2k0 x0 , (5.18)
k0 − k1
c3 = 0,
2k0
c4 = c1 ei(k0 +k1 )x0 .
k0 − k1
A general solution is a linear combination of these two solutions.

112
 5.1. Potential step

In figure 5.3 we show the solu-

1.0
32 ℏ2
tions for k0 = 5, c4 = 0 and dk = 3,

, Re(ψ(t,x)), Im(ψ(t,x))
V=
2m

dk = 5, or dk = 7. For dk ≥ k0 we 0.5

find for x < x0 a standing wave and 0.0

for x > x0 an exponential decaying

2
function (infinite decay length for -0.5

|ψ(t,x)
dk = k0 ). For dk < k0 we find for -1.0

x > x0 a wave propagating to the 6 8 10 12 14

right and for x < x0 a mixture of x

a propagating wave and a standing 1.0

52 ℏ2
, Re(ψ(t,x)), Im(ψ(t,x))

V=
2m
wave. 0.5
These standing waves near po-
0.0
tential steps are observed in na-
ture (see figure 5.4). If one
2

-0.5
|ψ(t,x)

looks with the use of a scan-

-1.0
ning tunneling microscope (STM)
6 8 10 12 14
at a surface of crystalline Cop- x
per one finds boundaries or steps 1.0
72 ℏ2
, Re(ψ(t,x)), Im(ψ(t,x))

related to places where the crys- V=

2m

0.5
tal changes its thickness by one
atomic layer. At these steps 0.0

ripples in the STM image can

2

-0.5
be seen. These are related
|ψ(t,x)

to the standing waves of the -1.0

electron wave function one finds 6 8 10 12 14

x
in the Schrödigner equation of
a single particle scattering off
a potential step. Figure 5.3: Stationary solutions of a
These stationary solutions of potential step with m = 1, ~ = 1,
the Schrödinger equations, i.e. k0 = 5, and dk = 3, 5, 7 from top to
eigenstates of the Hamiltonian, can bottom respectively.
not be normalized and thus do not
directly related to a physical state the system can be in. Non the less we
can use these solutions to calculate the current density in the different re-

113
Chapter 5. Wave mechanics in exemplary potentials

gions. From the different currents we can calculate the transmitted and
reflected portion of a wave with wave vector k. In section 3.1 we defined
the current density as
1
j(t, x) = (ψ(t, x)∗ pψ(t, ~r) − ψ(t, x)pψ(t, x)∗ ) , (5.19)
2m
i
~ ∂
with p = i ∂x . For a plane wave, ψ(t, x) = e ~ (px−Et) the current density is
given as
 ∗ ~ ∂ i 
1 i
(px−Et) (px−Et) i
(px−Et) ~ ∂  i (px−Et) ∗
j(t, x) = e ~ e~ − e~ e~
2m i ∂x i ∂x
(5.20)
p
=
m
~k
= .
m
With these equations we can
calculated the reflected and trans-
mitted currents for the stationary
solutions of the Schrödinger equa-
tion with a potential step. We as-
sume a particle flux moving from
x = −∞ towards the potential step.
This particle current is partially re-
flected and partially transmitted.
In terms of the two solutions pre-
viously found we need the solution
with c4 = 0 as no particles enter the
described setup from x = +∞. Un- Figure 5.4: STM image of a copper
der the condition that k1 ∈ R, i.e. surface (Property of IBM)
V0 < E we find that the current for
ψ− (t, x) is
~k0
|c1 |2 − |c2 |)2


j− (t, x) = (5.21)
m

114
 5.1. Potential step

 2 !
~k0 k0 − k1
= |c1 |2 1−
m k0 + k1
~ 4k02 k1
= |c1 |2
m (k0 + k1 )2
and the current for ψ+ (t, x) is
~k1
|c3 |2 )


j+ (t, x) = (5.22)
m
 2
~k1 2 2k0
= |c1 |
m k0 + k1
~ 4k02 k1
= |c1 |2
m (k0 + k1 )2
We see that the current for x < x0 is equal to the current for x >
x0 . This needs to be the case as for a stationary state we do not want
particle density to collect at the potential step. The ratio of reflected flux
to incoming flux is

m
~k0
|c2 |2
R = ~k (5.23)
0
m |c1 |2
 2
k0 − k1
= ,
k0 + k1
and the ratio of transmitted flux to incoming flux is
~k1 2
m |c3 |
T = ~k (5.24)
2
m |c1 |
0

4k0 k1
= .
(k0 + k1 )2
For V0 > E we find that k1 becomes complex and
k0 − i|k1 | ∗
  
k0 − i|k1 |
R= (5.25)
k0 + i|k1 | k0 + i|k1 |

115
Chapter 5. Wave mechanics in exemplary potentials

= 1.

The transmitted wave in this case is non-zero, but exponentially decays

such that no current is transported
2k0
φ+ (x) = c1 eik0 x0 e−|k1 |(x−x0 ) , (5.26)
k0 + i|k1 |
and

j+ (t, x) = 0. (5.27)

5.2 Potential barrier

Instead of a potential step, we can
make two successive potential steps
such that for x > x0 or x < −x0 the V(x)
potential is zero, but for |x| < x0
the potential is V0 . Naturally, the V0
potential barrier does not need to
be centered around x = 0, but a x = -x0 x = x0
shift of coordinates is trivially exe-
cuted afterward. The potential bar- Figure 5.5: A potential barrier.
rier is an idealized model system for
several relevant physical problems, including nuclear decay, field-effect tran-
sistors, and superconducting tunnel junctions used for highly sensitive de-
tectors and electronic circuits. In this section, we will follow the same line
of argumentation as in the previous section. We will first look at the prop-
agation of a Gaussian wave packet across a potential barrier. This we will
solve numerically and gives an idea of the possible effects. We will then
look at the solutions for a stationary state to analytically quantify the re-
sults. Physical particles are described by wave packets, which we can write
as superpositions of stationary states.
In figure 5.6 we show the numerical results for a wave packet traveling
across a potential barrier. Based on the results for a potential step we can

116
 5.2. Potential barrier

0.6 0.6 0.6 0.6 0.6 0.6 0.6

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
42 ℏ2 42 ℏ2 42 ℏ2 42 ℏ2 42 ℏ2 42 ℏ2 42 ℏ2
V= V= V= V= V= V= V=
2m 2m 2m 2m 2m 2m 2m
0.4 0.4 0.4 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2 -0.2 -0.2 -0.2

2

2
-0.4 -0.4 -0.4 -0.4 -0.4 -0.4 -0.4
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.6 -0.6 -0.6 -0.6 -0.6 -0.6 -0.6

-5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25
x x x x x x x

0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
0.6 52 ℏ2
, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
V= V= V= V= V= V= V=
2m 2m 2m 2m 2m 2m 2m
0.4 0.4 0.4 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2 -0.2 -0.2 -0.2

2

2
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.4 -0.4 -0.4 -0.4 -0.4 -0.4 -0.4

-0.6 -0.6 -0.6 -0.6 -0.6 -0.6 -0.6

-5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25
x x x x x x x

0.6 62 ℏ2
0.6 62 ℏ2
0.6 62 ℏ2
0.6 62 ℏ2
0.6 62 ℏ2
0.6 62 ℏ2
0.6 62 ℏ2
, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))

, Re(ψ(t,x)), Im(ψ(t,x))
V= V= V= V= V= V= V=
2m 2m 2m 2m 2m 2m 2m
0.4 0.4 0.4 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0 0.0 0.0

-0.2 -0.2 -0.2 -0.2 -0.2 -0.2 -0.2

2

2
|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)

|ψ(t,x)
-0.4 -0.4 -0.4 -0.4 -0.4 -0.4 -0.4

-0.6 -0.6 -0.6 -0.6 -0.6 -0.6 -0.6

-5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25 -5 0 5 10 15 20 25
x x x x x x x

Figure 5.6: Time evolution of a wave packet originally centred around x = 0

traveling with k0 = 5 across a potential barrier from x = 10 to x = 11 with
a potential give by dk = 4, 5, 6 from top to bottom respectively. m, ~, and
σ(t = 0) for the Gaussian wave packet are all set to 1.

already guess what should happen. When the particle hits the potential
barrier part of the wave packet is reflected and a part is transmitted. When
the barrier is at a higher energy than the kinetic energy of the particle one
would classically expect the particle to be fully reflected. This is almost,
but not completely the case in quantum mechanics. Inside the barrier, the
wave function exponentially decays. What is left over of the wave function
after this exponential decay at the other end of the potential barrier will
generate a new wave packet that corresponds to part of the particle density
that is transmitted through the potential barrier.
In order to quantify the amount of particles that is transmitted and
reflected form a potential barrier we can look at the stationary states. We
follow the same line of argument as in the previous section where we solved
the stationary states of a potential step. We define ψ(t, x) = χ(t)φ(x),
χ(t) = e−i/~Et and

ik0 x + c e−ik0 x x ≤ −x
φ− (x) = c1 e
 2 0
φ(x) = φ0 (x) = c3 eik1 x + c4 e−ik1 x −x0 ≤ x ≤ x0 (5.28)

 ik
φ+ (x) = c5 e 0 x x0 ≤ x,
q q
with k0 = 2mE
~2
and k1 = 2m(E−V ~2
0)
.

117
Chapter 5. Wave mechanics in exemplary potentials

We can solve these equations

1.0
42 ℏ2
by the requirement that ψ(x) and

, Re(ψ(t,x)), Im(ψ(t,x))
V=
2m

its first derivative are continuous at 0.5

−x0 and +x0 . For V0 < E (k1 ∈ R). 0.0

This gives values for c2 to c5 that all

2
scale with c1 . From these equations -0.5

|ψ(t,x)
we can calculate the current in the -1.0

different regions, or the transmit- -4 -2 0 2 4

tance and reflectively. The math is x

very similar to the algebra needed , Re(ψ(t,x)), Im(ψ(t,x))

1.0
V=
5.012 ℏ2
2m
in the previous section about a po- 0.5
tential step. We find for the trans-
0.0
mittance
2

-0.5
|c5 |2
|ψ(t,x)

T = (5.29)
|c1 |2 -1.0

-4 -2 0 2 4
1
=  2 , x

1 k0 k1
1+ 4 k1 − k0 sin2 (2k1 x0 ) 1.0
62 ℏ2
, Re(ψ(t,x)), Im(ψ(t,x))

V=
2m

0.5

and for the reflectance

0.0

|c2 |2
2

-0.5
R= (5.30)
|ψ(t,x)

|c1 |2
-1.0
= 1 − T. -4 -2 0 2 4
x

For the case where 2k1 x0 = nπ we

find that T = 1, i.e. the barrier be- Figure 5.7: Stationary solutions for a
comes transparent. There is a re- potential barrier.
flection on both sides of the barrier
in such a fashion that the reflections of each of these negatively interfere
and cancel each other.
For the case where E < V0 , k1 is complex. The derivation is similar
as before with the addition that one needs to be careful with all complex

118
 5.3. Delta function potential

conjugates. In this case, we find for the transmittance

1
T =  2 . (5.31)
1 k0 |k1 | 2
1+ 4 |k1 | + k0 sinh (2|k1 |x0 )

5.3 Delta function potential

We can now ask the question what happens to the scattering of a potential
barrier when the width of the barrier is much smaller than the De Broglie
wavelength of the particles involved. In order to study this, we can take
the limit of x0 in the previous section to zero. If we do this for a fixed
potential V0 that is independent of the width we quickly convince ourselves
that the transmittance becomes 1, independent of the size of V0 as both
sin(0) and sinh(0) are zero. We can take a nontrivial limit of x0 → 0 by
requiring that V0 × x0 is constant and taking the limit of x0 to zero. The
resulting potential is exactly the Dirac delta function (distribution). For
cases where the internal structure of the barrier is much smaller than the
De Broglie wavelength of the particles our system is good enough described
by the potential

V (x) = 2x0 V0 δ(x), (5.32)

corresponding to a barrier of width 2x0 and potential V0 . In this section,

we take 2x0 V0 > 0. Bound states with V0 < 0 will be discussed in the next
section.
In order to find solutions we can have a look at stationary states. We
define ψ(t, x) = χ(t)φ(x), χ(t) = e−i/~Et and
(
φ− (x) = c1 eikx + c2 e−ikx x < 0
φ(x) = (5.33)
φ+ (x) = c3 eikx + c4 e−ikx x > 0,
q
2mE
with k = ~2
. In the previous two sections we used the requirement
∂φ(x)
that both φ(x) and ∂x are continues. This argument was based on the

119
Chapter 5. Wave mechanics in exemplary potentials

Schrödinger equation
∂ 2 φ(x) 2m
2
= 2 (V (x) − E)φ(x). (5.34)
∂x ~
As long as 2m
~2
(V (x) − E)φ(x) is finite the second derivative of φ(x) is finite
and φ(x) as well as its first derivative are continues. For V (x) = 2x0 V0 δ(x)
we need to realize that at x = 0 the potential is not finite. We thus need
to reanalyze the continuity of the wave function and its derivative.

5.3.1 Continuity of the wave function and its derivative

In order to prove the continuity of φ(x) and its derivative at the point x
for finite potentials we can integrate equation 5.34 from x −  to x + .
To shorten the notation we will assume in the following that the potential
diverges at x = 0. For general values of x, one can shift all coordinates. The
integral over an infinitesimal range around x = 0 of equation 5.34 becomes:
Z + 2 Z 
d φ(x) 2m
lim dx = 2 lim (V (x) − E)φ(x)dx (5.35)
→0+ − dx2 ~ →0+ −
  Z 
dφ(x) dφ(x) 2m
lim − = 2 lim (V (x) − E)φ(x)dx
→0+ dx  dx − ~ →0+ −

whereby the 0+ in the limit means that  is a positive number. When V (x)
and φ(x) are continues functions around x = 0 the limit of the integral
on the right converges to zero and we find that the derivative of φ(x) is a
continues function at x = 0. For the case where V (x) is given by a delta
function at x = 0 we find
  Z 
dφ(x) dφ(x) 2m
lim − = 2 lim (2x0 V0 δ(x) − E)φ(x)dx (5.36)
→0+ dx  dx − ~ →0+ −
2m
= 2 2x0 V0 φ(0).
~
There is a discontinuity in the derivative of the wave function proportional
to the wave function at the point of the delta function and the strength of
the delta function potential (2x0 V0 ).

120
 5.3. Delta function potential

5.3.2 The wave function for a delta function potential

We now return to the problem of calculating the stationary solutions for a
delta function potential. Requiring φ(x) to be constant and the derivative
to have a jump of 2m
~2
2x0 V0 φ(0) yields the conditions

c1 + c2 = c3 + c4 (5.37)
2m
−ik(c1 − c2 ) + ik(c3 − c4 ) = 2 2x0 V0 (c1 + c2 ).
~
There are two sets of solutions. For odd functions ψ(0) = 0 and the delta
function potential has no effect

φodd (x) = eikx − e−ikx (5.38)

For even functions we find

m 2x0 V0
φeven (x) = cos(kx) + sin(k|x|). (5.39)
k~2
Which is a valid approximate solution p
for potential barriers where the thick-
ness 2x0 is small compared to 1/k = ~2 /(2mE).

121
Chapter 5. Wave mechanics in exemplary potentials

5.4 Particle in a box

Up to this point we have focused
in this chapter on ”free” particles.
That is, the stationary states we
discussed extended to ±∞. Wave V→∞ V→∞
packets we build from these states
V(x)=0
continuously broaden until their
x = -x0 x = x0
density at each point in space goes
to zero. The next several sections
Figure 5.8: The potential for a parti-
focus on bound particles. Bound
cle confined to the region −x0 to x0 .
particles are for all times localized
in space. The first potential we will
use is an infinitely deep potential well, or a particle in a box. We define

∞ x < −x0

V (x) = 0 −x0 < x < x0 (5.40)

∞ x0 < x


In previous sections, we first looked at the numerical time evolution

of a Gaussian wave packet and then focused on the stationary states to
understand the behavior of these wave packets analytically. A Gaussian
wave packet would have density (albeit small) outside the region −x0 to
x0 . Any density outside this region is multiplied by an infinite potential
and leads to infinite total energy. If we want to generate a localized wave
packet and discuss its time evolution the wave packet has to be within the
box. It will be simpler to first look at the stationary states and then use
these to build our wave packets.
For a stationary state we have ψ(t, x) = χ(t)φ(x), χ(t) = e−i/~Et and
φ(x) is an eigenfunction of the Hamiltonian
d2 φ(x) 2mE
+ 2 φ(x) = 0, (5.41)
dx2 ~
with the side condition that
φ(x) = 0 ∀|x| > x0 . (5.42)

122
 5.4. Particle in a box

The solution to equation 5.41 are plane waves

φ(x) = c1 eikx + c2 e−ikx (5.43)

= C1 sin(kx) + C2 cos(kx),

with C1 = i(c1 − c2 ) and C2 = c1 + c2 . As we have seen in the previous

section for delta function potentials we do not require the derivative of
the wave function to be continues at x = ±x0 . The potential diverges
at x = ±x0 such that, following the Schrödinger equation also the second
derivative of φ(x) is divergent at these points. We do require that φ(x) itself
is continues for all x. This generates the side condition that φ(±x0 ) = 0
and restricts the solution to certain k values. We find the odd solutions
(q
1 nπx
x0 sin( x0 ) |x| ≤ x0
φs (x) = , (5.44)
0 |x| ≥ x0

and the even solutions

(q (n−1/2)πx
1
x0 cos( x0 ) |x| ≤ x0
φc (x) = , (5.45)
0 |x| ≥ x0

with n ∈ N+ and the normalisation C1 and C2 chosen such that

Z x0
|φ(x)|2 dx = 1. (5.46)
−x0

We can combine the even and odd solutions into a single set of functions
(q nπ(x+x0 )
1
x0 sin( 2x0 ) |x| ≤ x0
φn (x) = , (5.47)
0 |x| ≥ x0

with n ∈ N+ .
The energy of these states is given as
R x0
φn (x)∗ Hφn (x)dx
E = −xR x00
∗
(5.48)
−x0 φn (x) φn (x)dx

123
Chapter 5. Wave mechanics in exemplary potentials

 
∗ − ~2 d2 φ (x) dx
R x0
−x0 φ n (x) 2m dx 2 n
= R x0
∗
−x0 φn (x) φn (x)dx
∗ ~2 π 2 n2
R x0
−x0 φn (x) 8mx20 φn (x)dx
= R x0
∗
−x0 φn (x) φn (x)dx
~2 π 2 n2
=
8mx20

We find that the more nodes the function has the higher its (kinetic)
energy is. This is a general property of wave functions. We can relate the
kinetic energy density at point x to the the norm square of the slope of the
wave function at point x.
Z ∞
~2 d2
Ekin =− φ(x)∗ 2 φ(x)dx (5.49)
2m −∞ dx
2
∞ Z ∞
~2

~ ∗ d d d
=− φ(x) φ(x) + φ(x)∗ φ(x)dx
2m dx −∞ 2m −∞ dx dx
2 Z ∞ 2
~ d
= φ(x) dx
2m −∞ dx

Whereby we used that the wave function is square integrable such that
d 2
limx→±∞ φ(x) = 0. As the integrant dx φ(x) is a strict positive function
we see that the kinetic energy increases if the norm of the slope of wave
function is large.
The set of stationary states that we find form an orthonormal set. (In
the next chapter we will generalise this statement ). Orthonormality of a
set of functions is defined as
Z ∞
φn (x)∗ φm (x)dx = δnm . (5.50)
−∞

We can prove the orthonormality of the set of stationary states found for a

124
 5.4. Particle in a box

particle in a box by evaluating the integral

Z x0 Z x0 r  r  
∗ 1 nπ(x + x0 ) 1 mπ(x + x0 )
φn (x) φm (x)dx = sin sin dx
−x0 −x0 x0 2x0 x0 2x0
Z 2x0    
1 (n − m)πx (n + m)πx
= cos − cos dx
2x0 0 2x0 2x0
sin((n − m)π) sin((n + m)π)
= −
(n − m)π (n + m)π
= δnm . (5.51)

We now use this property to evaluate the time evolution of a general

wave packet. We will write our wave packet as a sum over the stationary
states with a state dependent complex prefactor.
X
Ψ(t, x) = cn φn (x)χn (t) (5.52)
n
r   2 n2
X 1 nπ(x + x0 ) −i ~π
8mx2
t
= cn sin e 0

n
x0 2x0
We will not discuss the convergence of this expansion here but come back
to this in the next chapter. For now, we assume that there is a set of cn ’s
such that the left and right-hand sides become equal. We can calculate cn
by multiplying equation 5.52 with ψm (t, x)∗ = φm (x)∗ χm (t)∗ and integrate
over x.
Z ∞ Z ∞ X
ψn (t, x)∗ Ψ(t, x)dx = φn (x)∗ χn (t)∗ cm φm (x)χm (t)dx (5.53)
−∞ −∞
X Zm ∞
= cm χn (t)∗ χm (t) φn (x)∗ φm (x)dx
m −∞
X
= cm χn (t)∗ χm (t)δnm
m
= cn χn (t)∗ χn (t)
= cn .

125
Chapter 5. Wave mechanics in exemplary potentials

In the last step we used that

i
χn (t) = e− ~ En t and |χn (t)|2 = 1.
As the right hand side of equation
5.53 is time independent we can re-
place t by 0 and find
Z ∞
cn = φn (x)∗ Ψ(t = 0, x)dx.
−∞
(5.54)

In order to calculate the time

evolution for a particle defined by
a wave packet given by the func-
tion Ψ(t = 0, x) we can calculate
the expansion coefficients cn on the
stationary states of this function
Figure 5.9: Time propagation of a
and use equation 5.52 to calculate
wave packet in a box.
Ψ(t, x) for arbitrary times. In fig-
ure 5.9 we did exactly this. One
finds that a Gaussian wave packet
in a box propagates as a free wave
packet and bounces off the walls of the box. Once the wave packet broadens
such that its width is similar to the size of the box interference patterns
start to emerge. These results in beautiful structures, which one can ana-
lyze and understand in great detail.

5.5 Potential well

Instead of looking at a particle in a
box, i.e. an infinitely deep well we
can look at a finite potential well. V(x)
x = -x0 x = x0
The potential for a well is given by
V0

126

Figure 5.10: A potential well.

 5.5. Potential well


0
 x < −x0
V (x) = V0 −x0 < x < x0

0 x0 < x


(5.55)

This is the same potential as we de-

fined for the potential barrier. For
the potential barrier, we took V0 > 0. If we take V0 < 0 the potential
describes a well. The solutions for the potential barrier can directly be
used for the potential well. With one major difference. For a potential
barrier, all stationary states have positive energy. For a potential well,
the stationary states with positive energy have the same structure as the
states for a potential barrier. In a well, we can find additional bound states
with negative energy. It is these bound solutions we want to discuss in this
section.

5.5.1 Bound states in a potential well

Just as in the previous section, we are looking for stationary solutions of the
Schödinger equation. We will use these stationary solutions as a basis to
build other solutions from. For E < 0 and E > V0 the piecewise continuous
solution for the Hamilton eigenfunction equation Hφ(x) = Eφ(x) are

k0 x x < −x0
c1 e

φ(x) = c2 cos(k1 x) + c3 sin(k1 x) −x0 < x < x0 , (5.56)

 −k0 x
c4 e x0 < x
q q
with c1 to c4 ∈ C, k0 = − 2mE ~2
and k 1 = 2m
~2
(E − V0 ). We can solve for
c1 to c4 by requiring that both φ(x) and its first derivative are continuous
at x = ±x0 . We find

c1 e−k0 x0 = c2 cos k1 x0 − c3 sin k1 x0 (5.57)

127
Chapter 5. Wave mechanics in exemplary potentials

-1
Tan[ x0 k1 ]

2 m x20 V0
ℏ2 k21 x20
-V0 large
2

-
1

-V0 small
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
x0 k1 / π

sin k1 x0
Figure 5.11: Graphical solution for the equation k0 = k1 cos k1 x0 , yielding
the discrete set of k vectors for the even bound states.

3
-1
-V0 large
-CoTan[ x0 k1 ]

2 m x20 V0
ℏ2 k21 x20

2
-

1
-V0 small
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
x0 k1 / π

Figure 5.12: Graphical solution for the equation k0 = −k1 cos k1 x0

sin k1 x0 , yielding
the discrete set of k vectors for the odd bound states.

128
 5.5. Potential well

c4 e−k0 x0 = c2 cos k1 x0 + c3 sin k1 x0

c1 k0 e−k0 x0 = k1 (c2 sin k1 x0 + c3 cos k1 x0 )
−c4 k0 e−k0 x0 = k1 (−c2 sin k1 x0 + c3 cos k1 x0 )

This yields the relation between k0 and k1 as

c2 sin k1 x0 + c3 cos k1 x0
k0 = k1 (5.58)
c2 cos k1 x0 − c3 sin k1 x0
and
c2 sin k1 x0 − c3 cos k1 x0
k0 = k1 . (5.59)
c2 cos k1 x0 + c3 sin k1 x0
These two equations can only be fulfilled simultaneously when either c2 = 0
or c3 = 0. For these two different cases we find even solutions with
sin k1 x0
k0 = k1 , (5.60)
cos k1 x0
e−k0 x0
c2 = c1 ,
cos(k1 x0 )
c3 = 0,
c4 = c1 .

or odd solutions with

cos k1 x0
k0 = −k1 , (5.61)
sin k1 x0
c2 = 0,
e−k0 x0
c3 = −c1 ,
sin(k1 x0 )
c4 = −c1 .

We do not have a closed ex-

pression for the relation between k0

129
Chapter 5. Wave mechanics in exemplary potentials

and k1 . The easiest way to get

insight into the allowed quantized
wave functions is to solve for k0 and
k1 graphically. With the additional
condition that
2mV0
k02 = −k12 − (5.62)
~2
we can plot
s
2mx2 V0
− 2 20 2 − 1, (5.63)
~ k 1 x0

as a function of x0 k1 in red and ad-

ditionally plot

tan(x0 k1 ), (5.64)

for the even or

1
− , (5.65)
tan(x0 k1 )

for the odd solutions in blue. The allowed vectors can be found at the
intersections of these graphs. This is done in figure 5.11 and figure 5.12 for
the odd and even solutions respectively. r
2mx2 V
In figure 5.11 we see in blue tan(x0 k1 ) and in red − ~2 k20x20 − 1. In
1 0
figure 5.13 we plotted a set of eigenfunctions for a potential well with
− 2mV
~2 kx2
0
= 100/π 2 , a value chosen to generate an aesthetic pleasing plot.
For states with V0 − E  −V0 , i.e. those deeply bound in the potential
well we find solutions that are similar to the stationary states in a box.
For states where the binding energy is nearly zero E . 0, we find that the
eigenstates have exponential tails that substantially enter into the region
outside the potential well.

130
 5.5. Potential well

Just as in the case of a particle in a box we can generate wave packets

as sums over the bound state solutions. If we label the different stationary
bound states by φn (x) with energy En we can write the wave function of a
wave packet as
N
X
Ψ(t, x) = cn φn (x)χn (t). (5.66)
n=1

The sum runs from 1 to N for a potential well with N bound states. As
N is finite it becomes immediately clear that in this case not all possible
wave packets can be represented as bound states inside the potential well.
Besides the bound states, there are also infinitely many stationary states
that are unbound. These functions can be found in the section on the
(b)
potential barrier. We can now use the set of bound (φn (x)) and unbound
(u)
(φk (x)) states to build our general wave packet.

N Z ∞
(u) (u)
X
Ψ(t, x) = cn φ(b) (b)
n (x)χn (t) + c(k)φk (x)χk (t)dk (5.67)
n=1 −∞

Naturally, such a wave packet will partially stay inside the well (the
part that is made of the bound stationary states) and partially propagate
towards infinity (the part that is made of the unbound stationary states
of a potential well). An important finding is that even if one finds a set
(b)
of stationary states (φn (x)) these might not be sufficient to express all
physically relevant wave packets as a sum over these states. We will discuss
the completeness of a set of stationary states, i.e. the question if we can
express all possible physical wave packets as a convergent sum over the set
of stationary states in the next chapter.

131
Chapter 5. Wave mechanics in exemplary potentials

5.6 Harmonic oscillator

The last example discussed in this chapter is the Harmonic oscillator. In one
dimension this is a model of a particle bound to point x = 0. The force that
acts on the particle is proportional to the distance between the particle and
the point x = 0. In stead of the force as a function of distance, F = −kx
we can look at the potential energy of the particle. The potential energy
has the form V (x) = 12 kx2 such that the Schrödinger equation becomes

∂ ~2 ∂ 2 1
i~ ψ(t, x) = − ψ(t, x) + kx2 ψ(t, x). (5.68)
∂t 2m ∂x2 2
In order to describe the dynamics of a quantum particle in a harmonic
potential, we will find solutions for the stationary states and use these to
build wave packets. For stationary states, we can write the wave function
as a product of a time and position-dependent part. The time-dependent
part is trivial to solve and yields χ(t) = e−i/~Et . The eigenvalue equation
for the position-dependent part is

~2 d2 φ(x) 1
− + mω 2 x2 φ(x) = Eφ(x), (5.69)
2m dx2 2
whereby we replaced k by the angular
q eigenfrequency of an undamped
k
classical harmonic oscillator ω = m . In order to reduce the writing
further we introduce the dimensionless variables
2E
= (5.70)
~ω
and
r
mω
y= x. (5.71)
~
With these substitutions we obtain the differential equation

d2 φ(y)
+ ( − y 2 )φ(y) = 0. (5.72)
dy 2

132
 5.6. Harmonic oscillator

Within this section, we will use two different methods to solve this dif-
ferential equation. The first method uses general concepts on how to solve
partial differential equations, the second method uses operator methods
that we will formalize in the next chapter. After we found the stationary
states we discuss the resulting wave packets we can create.

5.6.1 Stationary states of the harmonic oscillator solution 1

Equation 5.72 defines an eigenvalue equation. In order to solve this equa-
tion, we can first focus on the limits where y → ±∞. We will show that
in this limit the function φ(y) exponentially decays to zero. All states are
bound states within a harmonic potential. We subsequently use that our
solutions are analytical and write them as a polynomial of finite order in y
multiplied with the solution for the limit of y → ±∞.
In the limit of y → ±∞ we can neglect  in the term involving  − y 2 .
All stationary states need to go asymptotically to solutions of
d2 φ∞ (y)
− y 2 φ∞ (y) = 0. (5.73)
dy 2
Instead of neglecting  (i.e. setting it to zero) we can take any value that is
much smaller than y 2 in the limit of y 2 → ∞. We can guess the solutions
in the region y → ±∞ to be of the form
2 /2
φ∞ (y) = e−y . (5.74)

Which we can test to find that this is an exact solution for the differential
equation
d2 φ∞ (y)
+ (1 − y 2 )φ∞ (y) = 0. (5.75)
dy 2
As  − 1  y 2 for y → ±∞ we find that φ∞ (y) is a valid solutions in the
limit of y 2 → ∞ for all values of .
For our full solution we write
2 /2
φ(y) = h(y)e−y . (5.76)

133
Chapter 5. Wave mechanics in exemplary potentials

As φ(y) needs to be analytical we can hope to find stationary solutions of

the Schrödinger equation with h(y) a polynomial of finite order, such that
lim φ(y) = φ∞ (y). (5.77)
y→±∞

This line of argument, i.e. finding the solution of a differential equation

in the limits where the argument of the function goes to infinity and then
multiplying these with a finite polynomial works often, but not always.
2
Substituting φ(y) = h(y)e−y /2 into equation 5.72 yields
d2 h(y) dh(y)
2
− 2y + ( − 1)h(y) = 0. (5.78)
dy dy
We now assume that h(y) is a polynomial of finite order and justify our
assumption once we have found valid solutions. (or try something else if
this leads to no possible solutions). For a polynomial of order n we write
n
X
hn (y) = αm,n y m . (5.79)
m=0

Substituting the expansion for hn (y) into equation 5.78 we find the relation

(m + 1)(m + 2)am+2,n = (2m − (n) + 1)am,n . (5.80)

One should realize that the different polynomials of different order corre-
spond to different eigenfunctions with possible different eigenenergies . We
indicate this relation between n and  by making  an explicit function of
the order of the polynomial (n). As am,n is only related to am+2,n we see
that even and odd solutions are strictly separate. This is related to the fact
that our system is invariant under space inversion x 7→ −x which will be
discussed in more detail in chapter 7 on symmetries.
If (n) is the eigenenergy of φ(y) in the case where hn (y) is given as a
polynomial of order n then αn,n is non zero, but αn+2,n must be zero in
this expansion. We thus know that
0 = (2n − (n) + 1)αn,n , (5.81)

134
 5.6. Harmonic oscillator

with αn,n 6= 0 this yields that

 = 2n + 1, (5.82)

with n ∈ N0 . The allowed eigenenergies are

1
E = ~ω(n + ), (5.83)
2
2
We can now generate all solutions φn (y) = hn (y)e−y /2 by taking αm=0,n =
1 for even n and αm=1,n = 1 for odd n and use the recurrent relations to
generate αm,n for all m ≤ n. We have made an educated guess on what
the solution might be and thus far found consistent results. The only thing
that we still need to check is that h0 (y) = 1 and h1 (y) = y yield valid
solutions. For h0 (y) = 1 the eigenvalue  should be 1. This yields equation
2
5.75 which we constructed such that 1e−y /2 is the solution. Similarly we
can check that h1 (y) = y generates a valid eigenfunction with eigenvalue
 = 3.
The functions hn (y) are besides a normalization constant known as the
Hermite polynomials. The Hermite polynomials are normalized such that
Z ∞
2 √
Hn (y)2 e−y dy = 2n n! π. (5.84)
−∞

Combining all together we find that our normalized stationary states are
given as
s
1 2
φn (y) = n
√ Hn (y)e−y /2 , or (5.85)
2 n! π
r
1  mω  41
r 
mω mω 2
φn (x) = n
Hn x e− 2~ x ,
2 n! ~π ~

with energy
1
E = ~ω(n + ). (5.86)
2

135
Chapter 5. Wave mechanics in exemplary potentials

5.6.2 Stationary states of the harmonic oscillator solution 2

We will now rederive equation 5.85 with the use of operator methods that
relate state φn±1 (x) to state φn (x). In equation 5.69 we defined the Hamil-
tonian as

~2 d2 mω 2 2
H=− + x , (5.87)
2m dx2 2

such that Hφ(x) = Eφ(x). We will now rewrite the Hamiltonian as a

product of two operators that are each others Hermitian conjugate. If we
define
r r r !
1 mω ~ d
a† = a+ = x− , and (5.88)
2 ~ mω dx
r r r !
1 mω ~ d
a = a− = x+ ,
2 ~ mω dx

we find for the Hamiltonian H

 
1
H = ~ω a+ a− + (5.89)
2
 
1
= ~ω a† a +
2

In order to prove that a+ is the hermitian conjugate or adjoint operator

of a− we can use the two identities,
Z ∞ Z ∞
φ(x)∗ xϕ(x)dx = (xφ(x))∗ ϕ(x)dx, (5.90)
−∞ −∞

and
Z ∞ Z ∞
d ∗ d
φ(x) ϕ(x)dx = (− φ(x))∗ ϕ(x)dx (5.91)
−∞ dx −∞ dx

136
 5.6. Harmonic oscillator

which hold for all φ(x) and ϕ(x), ∈ DH (see section 3.3). From this we see
that
Z ∞ Z ∞
∗
φ(x) a− ϕ(x)dx = (a+ φ(x))∗ ϕ(x)dx, (5.92)
−∞ −∞
or
a†+ = a− . (5.93)
Justifying the notation in terms of a† and a in equation 5.88.
We now solve for the eigenfunctions of H by first finding the ground-
state, i.e. the eigenstate with the lowest energy and then using our op-
erators
R∞ a† and a to build all other eigenstates. For all φ(x) ∈ DH with
2
−∞ |φ(x)| dx = 1, we have
Z ∞ Z ∞  
∗ ∗ † 1
φ(x) Hφ(x)dx = ~ω φ(x) a a + φ(x)dx (5.94)
−∞ 2
−∞
Z ∞ 
1
= ~ω (aφ(x))∗ (aφ(x))dx +
−∞ 2
1
≥ ~ω.
2
In the last step we used that the integral over a real function that is equal
or larger than zero for all values of its argument is equal or larger than zero.
More precisely the integral can only be zero if aφ(x) = 0. As equation 5.94
defines the energy of the system we find that E ≥ 12 ~ω for all states. For
the state with the lowest eigenenergy we find that E = 21 ~ω. This state is
a solution to the differential equation aφ(x) = 0:
aφ0 (x) = 0 (5.95)
r r r !
1 mω ~ d
x+ φ0 (x) = 0.
2 ~ mω dx

The solution of this differential equation is

 mω 1/4 mω 2
φ0 (x) = e− 2~ x , (5.96)
π~

137
Chapter 5. Wave mechanics in exemplary potentials

which is an eigenfunction of H or a stationary state of the Schrödinger

equation with the lowest possible energy of
1
E0 = ~ω. (5.97)
2
The next task is to find the energy spectrum of H, i.e. all eigenvectors
and eigenvalues. For this we will use the operators a and a† and their
commutator. The commutator of two operators plays an important role
in physics as we will see in more detail in the next chapter(s). For two
operators A and B we define the commutator [A, B] as
[A, B] = AB − BA. (5.98)
For the operators a and a† we find
[a, a† ] = aa† − a† a (5.99)
= 1,
which can be derived using the definitions of the operators a and a† (equa-
d d


tion 5.88) and by realizing that dx (xφ(x)) = x dx + 1 φ(x) such that
 
d
, x = 1. (5.100)
dx
In order to find the spectrum of H we look at the commutator of a or a†
with H. We find
[H, a] = Ha − aH (5.101)
 
† 1 † 1
= ~ω (a a + )a − a(a a + )
2 2
 
= ~ω a† aa − aa† a
= −~ωa,
whereby in the last step we used that aa† = 1 + a† a. Similarly we find for
the commutator of the Hamiltonian with a†
[H, a† ] = ~ωa† . (5.102)

138
 5.6. Harmonic oscillator

If φn (x) is an eigenstate of H with energy En we can find a new eigenstate

by applying the operator a† to φn .

Ha† φn (x) = ~ωa† φn (x) + a† Hφn (x) (5.103)

† †
= ~ωa φn (x) + a En φn (x)
= (~ω + En )a† φn (x).

We thus find that a† φn (x) is also an eigenstate of H with an eigenenergy

of ~ω + En . We know that the ground state of H or the eigenstate with
lowest energy is φ0 (x) with energy E0 = ~ω 12 . By induction we find that
for n ∈ N0

φn (x) = αn (a† )n φ0 (x) (5.104)

are eigenstates of H with energy

 
1
En = ~ω n + , (5.105)
2
and αn an arbitrary complex constant.
A similar set of functions can be found from the state aφn (x) and we
find that

Haφn (x) = (En − ~ω)aφn (x). (5.106)

Whereby from equation 5.95 we know that by construction

aφ0 (x) = 0, (5.107)

such that the energy is bound from below and the ground state φ0 (x) is
the state with lowest energy.
Given that φn (x) are eigenstates of the operator H = ~ω(a† a + 1/2)
with eigenenergies ~ω(En = n + 1/2) one finds that the states φn (x) are
also eigenstates of the operator N = a† a with eigenvalues n.
The operators a† and a relate the state φn to the state φn±1 . These
states differ in energy by a constant ±~ω. The operator a† that brings

139
Chapter 5. Wave mechanics in exemplary potentials

you from a state n to n + 1 is called the raising operator. The operator a

that brings you from state n + 1 to state n is called the lowering operator.
Instead of thinking of a particle with a kinetic energy that is quantized
in portions of ~ω, we can also think of a field to which we add particles.
The quantized modes of electromagnetic radiation in a cavity, which Planck
considered, behave in this way. Each time we raise the energy we can think
of it as adding a particle to the field of energy ~ω. As such the operators are
also called creation and annihilation operators. They create or annihilate
a quantum (particle) with ~ω of energy in the ”mode”R of frequency ω.
∞
We often normalize our wave functions such that −∞ |φn (x)|2 dx = 1,
which simplifies several equations. We can find the constant αn in equation
5.104 that makes the norm of the wave function one by looking at the state
a† φn (x):

a† φn (x) = αn (a† )n+1 φ0 (x) (5.108)

αn
= φn+1 (x).
αn+1

If we look at the expectation value of the operator aa† we find

Z ∞ Z ∞
φn (x)∗ aa† φn (x)dx = (a† φn (x))∗ a† φn (x)dx (5.109)
−∞ −∞
 2 Z ∞
αn
= φn+1 (x)∗ φn+1 (x)dx
αn+1 −∞
Z ∞ Z ∞
φn (x)∗ aa† φn (x)dx = φn (x)∗ (1 + a† a)φn (x)dx
−∞ −∞
Z ∞
= (n + 1) φn (x)∗ φn (x)dx
−∞
R∞ 2
For wave functions that are normalized such that −∞ |φn (x)| dx = 1 we
find

αn 2
 
= n + 1, (5.110)
αn+1

140
 5.6. Harmonic oscillator

or
r
1
αn = . (5.111)
n!
To conclude we find
r
1 † n
φn (x) = (a ) φ0 (x), (5.112)
n!

√
aφn (x) = nφn−1 (x), (5.113)

and
√
a† φn (x) = n + 1φn+1 (x). (5.114)

5.6.3 Wave packets of the harmonic oscillator

With the use of the stationary states of the harmonic oscillator we can build
wave packets. With φn (x) the eigenstates of the Hamiltonian with energy
i
En and χ(t) = e− ~ En t we find that

ψn (t, x) = φn (x)χn (t) (5.115)

are stationary solutions of the Schrödinger equation. General wave packets

can be written as
∞
X
Ψ(t, x) = cn ψn (t, x), (5.116)
n=0

with cn a set of complex numbers defining the state Ψ(t, x). We now follow
the same line of argumentation we used for a particle in a box (around
equation 5.52) and can calculate the expansion coefficients cn as
Z ∞
cn = ψn (t, x)∗ Ψ(t, x)dx. (5.117)
−∞

141
Chapter 5. Wave mechanics in exemplary potentials

This derivation requires that the eigen states of the Hamiltonian are or-
thogonal, i.e. that
Z ∞
φn (x)∗ φm (x)dx = δm,n , (5.118)
−∞

which one can check using the explicit representation of our eigen functions.
In the next chapter we will prove that non-degenerate eigenstates (En 6=
Em ) of any hermitian operator are orthogonal to each other.

5.6.4 Coherent states

We have seen that wave packets generally deform in time. This is re-
lated to the energy-momentum relation E(p) or ω(k). The group velocity
d
vg = dk ω(k) relates to the propagation of the wave packet. The higher
dn
derivatives of the energy-momentum dispersion relation dk n ω(k) for n > 1

relate to the deformation in time of the wave packet (see equation 4.37). For
a harmonic potential, the equidistant spacing of the eigenenergies allows us
to create a coherent state. We can define a wave packet that oscillates back
and forward in the harmonic potential trap like a classical particle, without
the density of the wave packet deforming.
We take

Ψ(t = 0, x) = φ0 (x − x0 ). (5.119)

For x0 = 0 this is a stationary state with energy E0 = 12 ~ω, for a general

x0 the state is given as a sum over all other stationary states
∞
X
Ψ(t, x) = cn φn (x)χn (t), (5.120)
n=0

with
Z ∞
cn = φn (x)∗ φ0 (x − x0 )dx. (5.121)
−∞

142
 5.6. Harmonic oscillator

In order to determine cn we first show that φ0 (x − x0 ) is an eigenstate

of the lowering operator a. We have
r r r !
1 mω ~ d
aφ0 (x − x0 ) = x+ φ0 (x − x0 ) (5.122)
2 ~ mω dx
with u = x − x0
r r r !
1 mω ~ d
= (u + x0 ) + φ0 (u)
2 ~ mω du
with aφ0 (u) = 0
r
mω
= x0 φ0 (x − x0 )
2~

p φ0 (x − x0 ) is an eigenstate of the lowering operator a

Which proves that
with eigenvalue mω 2~ x0 .
We can now determine the expansion coefficients cn starting with the
eigenvalue equation for a on the wave function at t = 0.
∞
X
aΨ(t = 0, x) = a cn φn (x) (5.123)
n=0
∞
X √
= cn nφn−1 (x)
n=1
∞
X √
= cn+1 n + 1φn (x).
n=0

At the same time we have

aΨ(t = 0, x) = aφ0 (x − x0 ) (5.124)

r
mω
= x0 φ0 (x − x0 )
2~
∞ r
X mω
= x0 cn φn (x).
2~
n=0

143
Chapter 5. Wave mechanics in exemplary potentials

It follows that
√
r
mω
cn+1 n + 1 = x0 cn , (5.125)
2~
or
r r n
1 mω
cn = x0 c0 . (5.126)
n! 2~
The value of c0 can be determined by directly integrating
Z ∞
c0 = φ0 (x)∗ φ0 (x − x0 )dx (5.127)
−∞
mωx20
= e− 4~ .

The wave function at time t is

∞ r n
mωx2
− 4~ 0 1 mω i
X
Ψ(t, x) = e √ x0 φn (x)e− ~ En t (5.128)
n! 2~
n=0
∞ r n
mωx2
− 4~ 0 − 12 iωt 1 mω
X
= e √ x0 e−iωt φn (x).
n! 2~
n=0

This result is periodic in t with

2π
Ψ(t = n, x) = (−1)n φ0 (x − x0 ), (5.129)
ω
and
2π 1
Ψ(t = (n + ), x) = −i(−1)n φ0 (x + x0 ), (5.130)
ω 2
with n ∈ Z.
We can investigate the time evolution of the wave packet further by
writing
mωx20 − 1 iωt mω −iωt )2
Ψ(t, x) = e− 4~ 2 e 4~ (x0 e (5.131)

144
 5.6. Harmonic oscillator

∞ r n
X
− mω (x0 e−iωt )2 1 mω
× e 4~ √ x0 e−iωt φn (x)
n! 2~
n=0
1 −2iωt −1 mω x2 −2iωt
= e 4 ((e )
~ 0 ) φ (x − x e−iωt )
0 0

If we now look at the particle density ρ(t, x) = Ψ(t, x)∗ Ψ(t, x) we find
r
mω − mω (x−x0 cos(ωt))2
ρ(t, x) = e ~ (5.132)
~π

A particle represented by a Gaussian density distribution oscillates with

frequency ω in a harmonic potential.
Schrödinger wrote in 1926 an article where he investigated this exam-
ple. The article is about the relation between micro and macro-mechanics
[11], i.e. quantum and classical mechanics. His intention was to show,
by an example, how one could generate wave packets that do not diffuse
in space but stay localized. The generalization made in his paper is not
correct. For a harmonic oscillator, one can find such coherent states that
do not diffuse. For general systems, quantum wave packets do diffuse over
time. For Schrödinger this was quite a disappointment. He wanted to inter-
pret Ψ(t, x)∗ Ψ(t, x) for a wave packet as the particle density representing a
classical particle. For that, the wave packet can be distributed over length
scales comparable to the De Broglie wavelength of the particle, but not over
much larger length scales as that would seemingly defy our experimental
observations. Making localized wave packets that stay localized works for
a harmonic oscillator, but can not easily be generalized to other potentials.
Even for the harmonic oscillator, we only showed that a localized coherent
state stays local, we did not provide a reason why macroscopic systems
should favor such states, which is what we observe.

145
Chapter 5. Wave mechanics in exemplary potentials

5.7 Summary
We have seen in this chapter several examples of solutions of the Schrödinger
equation with simple potentials. In order to understand the dynamics of a
particle, we define a wave packet (Ψ(t, x)) that represents this particle. We
can calculate the dynamics of a wave packet once we know the stationary
states:

ψn (t, x) = φn (x)χn (t). (5.133)

The time dependent part of a stationary state is given as

i
χn (t) = e− ~ En t , (5.134)

with En the energy of the state. For the position dependent part φ(x) we
need to solve the eigenvalue equation of the Hamiltonian

Hφn (x) = En φn (x). (5.135)

Once we know the eigenstates of the Hamiltonian we can use the sta-
tionary states to build general wave packets
∞
X
Ψ(t, x) = cn ψ(t, x), (5.136)
n=0

with
Z ∞
cn = φn (x)∗ Ψ(t = 0, x)dx. (5.137)
−∞

In several places, we referred to the next chapter as we have not always

proven all steps. There are in principle many more beautiful examples we
can discuss, but let us first define the necessary framework.

146
Chapter 6

Linear algebra

In this chapter we will formalize the methods used in the previous two
chapters to solve problems in quantum mechanics. For a particle in free
space we have seen that we can Fourier transform the wave function Ψ(t, ~r)
to get
 d Z
1 ~
φ(t, ~k) = √ Ψ(t, ~r)e−ik·~r dd~r. (6.1)
2π R d

The function φ(t, ~k) defines the Fourier components with wave vector ~k
of the wave packet Ψ(t, ~r). The time evolution of the Fourier components
can straightforwardly be calculated. For a particle in free space V (~r) is
constant and plane waves are stationary states such that
φ(t, ~k) = φ(~k)χ(t), (6.2)
~
with φ(~k) = c(~k)eik·~r , c(~k) a complex-valued function independent of ~r and
~
t, and χ(t) = e−iE(k)t . We can use these time-dependent Fourier compo-
nents φ(t, ~k) to write an arbitrary wave packet. For arbitrary potentials
plane waves are no longer the stationary solutions of the Schrödinger equa-
tion. For those systems a Fourier transform is not as efficient to solve the
Schrödinger equation. The stationary solutions of the Schrödigner equa-
tion are always given by the eigenstates of the Hamiltonian, but not always

147
Chapter 6. Linear algebra

by plane waves. In the last section of the previous chapter, we looked at

solutions of the Harmonic oscillator. General solutions for the Schrödinger
equation were not written as linear combinations of plane waves, (one could,
but it does not simplify the problem), but as linear combinations of the sta-
tionary solutions. This simplified the problem and allowed us to study for
example coherent wave packets. In this chapter, we will generalize the
method of expanding wave packets on a basis of the stationary states of the
Schrödinger equation.

6.1 States
In classical physics a microstate is described by all positions and momenta
of all particles involved. In quantum physics the microstate is defined by
the wave function. We will often drop the term micro in the reference to a
quantum (micro) state. A state in quantum mechanics is always considered
to be the description on a microscopic level.
We characterize the state of a quantum system using the notation in-
troduced by Dirac and write it as:

|ψi . (6.3)

We will call this the state vector ”ket” belonging to ψ. The ket for a wave
function ψ(t, ~r) is given as |ψi. We do not need to know the wave function
to write down a ket. For a Harmonic oscillator we have a set of normalized
stationary states that we can label by the quantum number n. The kets for
these states are |ni. We can work with these kets and know their associated
energy due to the lowering and raising operators, even if we do not know an
exact representation of the wave function. The fact that we can write down
the ket representing a state for which we do not know the wave function
becomes even more apparent if we look at the state of a Silver atom. We
have seen in chapter 1 where we discussed the Stern-Gerlach experiment
that the magnetic moment of a Silver atom is described by two internal
states. These two states are related to the angle the magnetic moment has
with respect to an external magnetic field. For the states of a Silver atom,

148
 6.1. States

we can write the kets representing these states as |αi and |βi, without
detailed knowledge on the wave function they represent.

6.1.1 Linearity and algebra with kets

Just as we can add wave functions that represent states we can add kets

|αi + |βi = |γi , (6.4)

or multiply them by complex numbers

c |αi = |αi c, (6.5)

with c ∈ C. For c 6= 0, |αi and c |αi describe the same state. The multipli-
cation with c = 0 is special as this results in a null ket.

0 |αi = 0 |βi = 0. (6.6)

We will write the null ket as 0.

6.1.2 Complex Vector space

The mathematical structure that can represent elements with the previously
described properties is that of complex vector spaces. A complex vector
space (also called a linear space) V , is a collection of objects (|αi) often
called vectors, which may be added together and multiplied by a complex
number. The latter is called scaling the vector. The complex number with
which we multiply a vector is called respectively a scalar.
For vector spaces V and vectors that are elements of this vector space
|αi ∈ V , we have the following properties
• Closed under addition

|αi + |βi ∈ V ∀ |αi , |βi ∈ V (6.7)

• Abelian

|αi + |βi = |βi + |αi ∀ |αi , |βi ∈ V (6.8)

149
Chapter 6. Linear algebra

• Associative

(|αi + |βi) + |γi = |αi + (|βi + |γi) ∀ |αi , |βi , |γi ∈ V (6.9)

• Existence of a 0-vector

∃0 ∈ V : |αi + 0 = |αi ∀ |αi ∈ V (6.10)

• Existence of an inverse-vector

∀ |αi ∈ V ∃ |βi ∈ V : |αi + |βi = 0 (6.11)

For scaling (multiplying with a complex constant) a vector (|αi) that is

element of a vector space (|αi ∈ V ) we have the following properties and
rules

• Closed under scaling

c |αi ∈ V ∀ |αi ∈ V, c ∈ C (6.12)

• Scaling with unity

1 |αi = |αi ∀ |αi ∈ V (6.13)

• Associative

(ab) |αi = a(b |αi) ∀ |αi ∈ V, a, b ∈ C (6.14)

• Distributive

(a + b) |αi = a |αi + b |αi ∀ |αi ∈ V, a, b ∈ C (6.15)

c(|αi + |βi) = c |αi + c |βi ∀ |αi , |βi ∈ V, c ∈ C (6.16)

The property that 0 |αi = 0 |βi = 0 follows from the above definitions.

150
 6.1. States

6.1.3 Ray and Vector

A physical state is defined by an equivalence class of vectors. For |αi ∈ V
and c ∈ C, |αi and c |αi with c 6= 1 represent different vectors in the complex
vector space V . For c 6= 0 they do represent the same physical state. A
state is therefore defined as a Ray given by all vectors c |αi with c 6= 0. The
complex phase and normalization of a state has a physical meaning, but
only in relation to the normalization and phase of other states. The phase
of a quantum state is, for example, important if one wants to understand
interference between particles. In the next chapter, we will show that the
phase and norm of a state are also important if one discusses symmetry in
quantum mechanical systems. Symmetry operations that do not change the
state of a system might well change the phase of the state, i.e. the state ray
does not change under symmetry operations, the state vector does change.

6.1.4 Linear dependence

In section 3.2 we have seen that if |αi and |βi are solutions of the Schrödinger
equation then also |γi = |αi + |βi is a solution of the Schrödigner equation.
In general, we can always write a state as a superposition of a set of other
states. We can ask the question if there is a unique solution for such an ex-
pansion or if one can find several different expansions that define the same
state. In order to answer this question, we will look at the state vectors |αi
instead of states (rays).
A set of state vectors {|α1 i , . . . , |αn i} is called linearly independent if
the equation
n
X
ci |αi i = 0, (6.17)
i=1

has only one solution for ci ∈ C. This unique solution necessarily is ci =

0 ∀i. If the previous equation has more than one solution the set of state
vectors is called linearly dependent. Later in this chapter we will define
methods to test the linear dependence of a set of state vectors with the use
of the inner product.

151
Chapter 6. Linear algebra

6.1.5 Dimension of Vector space

The dimension of a vector space V is given by the largest number of linearly
independent vectors one can find. In our equations we silently assume that
the dimension of the vector space is finite. We will discuss the extension to
infinite vector spaces later in this chapter. For a full discussion of infinite-
dimensional vector spaces we refer to the literature.

6.1.6 Basis
The basis B of a vector space V is defined as a set of linearly independent
vectors with the dimension (n) of B equal to the dimension (n) of V .

6.1.7 Projection on a basis

For a given set of basis vectors B for complex vector space V , any state
vector |αi in V , can be written as a unique linear superposition of these
basis vectors. In order to proof this statement we write the basis B of an
n dimensional vector space as the set

{|β1 i , . . . , |βn i} (6.18)

We first proof that there is a set of complex numbers {bi , . . . , bn } ∈ Cn such

that
n
X
|αi = bi |βi i (6.19)
i=1

The equation
n
X
a |αi − bi |βi i = 0, (6.20)
i=1

has the trivial solution a = 0 and bi = 0 ∀i. If this was the only solution
then vector |αi would be linearly independent of the vectors |βi i and the
set B would not be the largest possible set of linearly independent vectors

152
 6.1. States

in V . Therefore the set B would not be the basis of V , in contradiction

with our original assumptions that the set B is the basis of V . It follows
that equation 6.20 must have a second solution. For a = 0 the only solution
allowed is the solution with all bi s equal to zero as B is the basis of V . We
must have a solution of equation 6.20 with a 6= 0. We can now scale all
bi ’s such that a = 1. The solution to equation 6.20 with a = 1 is the state
vector for which equation 6.19 holds.
The uniqueness of this solution can be proven by writing two different
solutions for the expansion of the state vector |αi on the basis.
n
X
|αi = bi |βi i (6.21)
i=1
Xn
= ci |βi i .
i=1

We now need to show that these two expansions necessarily need to be

equivalent, i.e. that bi = ci ∀i. If we subtract the two expansions we find
n
X
(bi − ci ) |βi i = 0. (6.22)
i=1

As the set {|β1 i , . . . , |βn i} is the basis of V we know that the only solution
for equation 6.22 is the one where bi − ci = 0 ∀i. We thus find that
bi = ci ∀i and there is only one solution possible for the projection of a
state onto the basis.
This is an important mathematical result that has physical implications.
For a Silver atom we can define two state vectors and describe the magnetic
orientation for any Silver atom one can encounter in nature as a linear
superposition of these two state vectors. This expansion is unique and
complete.

153
Chapter 6. Linear algebra

6.2 Hilbert space

6.2.1 Bra vectors as dual representations of ket vectors
For each vector |αi i we can define a vector hαi | such that |αi i† = hαi |.
With † the dagger or complex conjugate transposed of the vector. We will
gennerally associate kets with column and bras with row vectors. If V is
the vector space of all vectors |αi i than V † is the vector space of all vectors
hαi |. We call V † a dual vector space of V . For each element in V we find
a dual element in V † . It follows that

(c |αi)† = hα| c∗ . (6.23)

All properties derived for the ket vectors in the previous section also hold
for the bra vectors. In particular if the vectors {|β1 i , . . . , |βn i} form a basis
for an n dimensional vector space V , then the vectors {hβ1 | , . . . , hβn |} form
a basis for the n dimensional vector space V † .

6.2.2 Inner product of a bra and ket vector

We define the inner product of a bra vector hα| ∈ V † and a ket vector
|βi ∈ V as

c = hα|βi , (6.24)

with c ∈ C. The inner product assigns a complex number to two vectors

from two vector spaces that are dual to each other. For the inner product
we have the following relations:

• Complex conjugation under commutation

hα|βi = hβ|αi∗ ∀ |αi , |βi ∈ V (6.25)

• Linearity

hα| (a |βi + b |γi) = a hα|βi + b hα|γi (6.26)

154
 6.2. Hilbert space

∀ |αi , |βi , |γi ∈ V, a, b ∈ C

(a |βi + b |γi)† |αi = a∗ hβ|αi + b∗ hγ|αi
∀ |αi , |βi , |γi ∈ V, a, b ∈ C

• Positive inner product of the vector with its own dual representation

hα|αi ≥ 0 ∀ |αi ∈ V (6.27)

• Zero inner product only for the zero vector

hα|αi = 0 ⇐⇒ |αi = 0 ∀ |αi ∈ V (6.28)

Two vectors are called orthogonal if their inner product is zero.

6.2.3 The norm of a state vector

The norm of a state vector can be defined as the square root of the inner
product of a state vector with its own dual representation. The norm for
state vector |αi is
p
k |αi k = hα|αi. (6.29)

6.2.4 Orthonormal basis

A basis B = {|β1 i , . . . , |βn i} for vector space V is called orthogonal if
hβi |βj i = 0 ∀i 6= j, i ≤ n, j ≤ n. (6.30)
The basis is called orthonormal if
hβi |βj i = δi,j ∀i ≤ n, j ≤ n. (6.31)
If B = {|β1 i , . . . , |βn i} is an orthonormal basis of vector space V then
we can write any vector |αi ∈ V as
n
X
|αi = hβi |αi |βi i . (6.32)
i=1

155
Chapter 6. Linear algebra

We can prove this by realizing that any state vector can be written as a
linear combination of the basis state vectors (equation 6.19). If we mul-
tiply equation 6.19 with hβj | from the left the relation for the expansion
coefficients bi follow.

6.2.5 Infinite dimensional vector spaces

In the previous sections, we assumed that the dimension of the vector space
is at most countable, i.e. we have a set of discrete states that we can label
with an integer. As we have seen in the previous chapter this is the case
for bound states, but not for unbound states. For unbound states we have
a continuum of states. As an example, we can think of plane waves in
free space labeled by the wave vector ~k in d dimensions. For states labeled
with a continuous index ~k, we can still define
E a vector space with inner
~
product and norm. If the state vectors k define an infinite-dimensional
(innumerable) basis for vector space V Ethen we E define the orthogonality
conditions between two state vectors ~k1 and ~k2 as
D E
~k1 ~k2 = δ(~k1 − ~k2 ), (6.33)

Whereby we replaced the Kronecker delta in case of an at most countable

set by the Dirac delta distribution. The projection of a state on a basis is
in the case of an infinite dimensional (innumerable) vector space given by
an integral instead of a sum. For a general state vector |αi we have
Z E
|αi = c(~k) ~k d~k, (6.34)
Rd

with c(~k) the complex expansion coefficients. We can determine c(~k) as

D E
c(~k) = ~k α , (6.35)
D
which we can check by multiplying equation 6.34 with ~k from the left
and using expression 6.33.

156
 6.2. Hilbert space

6.2.6 Hilbert space

Vector spaces that represent physical states in quantum mechanics are com-
plete. Every physical relevant state can be written as a sum over the ba-
sis states. For finite-dimensional vector spaces, this is trivially given, for
infinite-dimensional vector spaces this is not always true. We must restrict
the set of vector spaces used to describe physical systems to a subset of
vector spaces where completeness is given.
We call a vector space with norm complete if every Cauchy sequenceE1
of vectors in V converges to a vector in V . The sequence of vectors ~k1 ,
E E E
~k2 , ~k3 , . . . is a Cauchy sequence that converges to the state vector ~k
iff:
E E
∀ > 0 ∃n0 : ~kn − ~k <  ∀n ≥ n0 . (6.36)

Cauchy sequencesE by their definition converge. For a Cauchy sequence

E
whose vectors ~kn ∈ V we do not necessarily have that limn→∞ ~kn ∈ V
An example of a sequence of elements in a set that converges to a result
that is not in the set can be constructed with the rational numbers (Q).
The sequence
xn 1
xn+1 = + , (6.37)
2 xn
with x0 = 1 is given by the fraction of two integer numbers, i.e. a rational
number for each finite integer value of n. The limit of n → ∞ gives
√
lim xn = 2, (6.38)
n→∞

which is not a rational number.

E
1
We define a Cauche sequence as a sequence of states ~kn for which the distance
between the states for large n becomes arbitrarily small:
E E
∀ > 0 ∃N ∈ N ∀m, n ≥ N : ~kn − ~km < .

157
Chapter 6. Linear algebra

We call a vector space a Hilbert space if it is complete and if its norm

is defined by the inner product as in equation 6.29. Physical systems are
described by Hilbert spaces.

6.2.7 Vector subspace

For a Hilbert space V with dimension n we define the Hilbert space W with
dimension m as a subspace of V if m ≤ n and all basis vectors of W can
be written as a linear combination of the basis vectors of V . In particular,
we can generate a subspace of V by choosing m ≤ n different basis vectors
of V and use these as the basis vectors of W . We can generalize this to
infinite-dimensional Hilbert spaces with the definition that a subset W ⊂ V
is called a vector subspace of V iff W is a vector space under the operations
of V . In order for W to be a Hilbert space, it must additionally be complete.

6.2.8 Tensor product of vector spaces

Let Va and Vb be two Hilbert spaces with dimensions na and nb and bases
Ba = {|α1 i , . . . , |αnb i} and Bb = {|β1 i , . . . , |βnb i}. We define the tensor
product space

V = Va ⊗ Vb , (6.39)

as the Hilbert space with basis

B = {|α1 i |β1 i , |α1 i . . . , |α1 i |βnb i , . . . |αna i |β1 i , |αna i . . . , |αna i |βnb i}.
(6.40)

The product states can equally well be written as

|αi i |βj i = |αi , βj i (6.41)

= |αi i ⊗ |βj i .

The choice on how to write down the product states is a matter of prefer-
ence.

158
 6.3. Relation between bra ket notation and wave functions

The inner product between two states in V is defined as

hαi , βj |αk , βl i = hα1 |αk i hβj |βl i . (6.42)

The dimension of the tensor product space V is the product of the

dimension of Va and Vb

n = na nb . (6.43)

6.3 Relation between bra ket notation and wave

functions
The description of quantum mechanics with the use of state vectors is
equivalent to the description with the use of wave functions. A quantum
state of N particles can be represented by the wave function

ψ(t, ~r1 , . . . , ~rN ), (6.44)

or by the state vector

|ψi . (6.45)

For a single particle at position ~r we can define the state vectors |~ri.
The vectors |~ri define the basis of an infinite dimensional vector space. A
general state |ψi can be expressed on the basis |~ri as
Z
|ψi = ψ(t, ~r) |~ri d~r, (6.46)
Rd

with ψ(t, ~r) the wave function representing the state |ψi. In order to express
the wave function in terms of an inner product of two state vectors we can
multiply equation 6.46 from the left with h~r1 |
Z
h~r1 |ψi = ψ(t, ~r) h~r1 |~ri d~r (6.47)
Rd

159
Chapter 6. Linear algebra

Z
= ψ(t, ~r)δ(~r1 − ~r)d~r
Rd
= ψ(t, ~r1 ).
The wave function ψ(t, ~r) at point ~r describes the complex coefficient of the
vector for the position basis state |~ri. Similarly the state |~r1 i corresponds
to the wave function δ(~r − ~r1 ).
With the relation between states and wave functions we can define the
inner product of two wave functions. Starting from the inner product of
two states hψ1 |ψ2 i we insert equation 6.46 twice to get
Z † Z 
hψ1 |ψ2 i = ψ1 (t, ~r1 ) |~r1 i d~r1 ψ2 (t, ~r2 ) |~r2 i d~r2 (6.48)
Rd Rd
Z Z
= ψ1∗ (t, ~r1 )ψ2 (t, ~r2 ) h~r1 |~r2 i d~r1 d~r2
d d
ZR ZR
= ψ1∗ (t, ~r1 )ψ2 (t, ~r2 )δ(~r1 − ~r2 )d~r1 d~r2
d d
ZR R
= ψ1∗ (t, ~r)ψ2 (t, ~r)d~r.
Rd

We can relate the inner product of two state vectors to the integral over
the wave functions. We can generalize the previous equations to N particles.
Two N particle quantum states can be represented by the wave functions
ψ1 (t, ~r1 , . . . , ~rN ) and ψ2 (t, ~r1 , . . . , ~rN ). These states can equivalently be
represented by the state vectors |ψ1 i and |ψ2 i. The inner product between
these two states is given as
Z Z
hψ1 |ψ2 i = ... ψ1 (t, ~r1 , . . . , ~rN )∗ ψ2 (t, ~r1 , . . . , ~rN )d~r1 . . . d~rN . (6.49)
Rd Rd

The notation of a quantum state in terms of state vectors or in terms of

wave functions are mathematically equivalent and can be used to describe
the same physics. In particular, all properties derived for our state vectors
also hold for the wave functions. We already used these mathematical
tools in the previous chapter where we used the coherent states as a basis
to expand the wave packets on.

160
 6.4. Operators

6.4 Operators
Besides state vectors, we have operators in quantum mechanics. An oper-
ator takes one state vector and transforms it into another state vector. In
the previous chapters we have already encountered several operators acting
on wave functions. Due to the possible mapping between wave functions
and state vectors ψ(t, ~r) ⇐⇒ |ψi we can identify operators acting on
wave functions and rewrite them as operators acting on state vectors. An
operator acting on a wave function takes one wave function as its argument
and produces a wave function as its output. The Schrödinger equation
∂
(i~ ∂t ψ(t, ~r) = Hψ(t, ~r)) is given by two operators. The Hamilton operator
∂
H on the right-hand side as well as the time derivative operator i~ ∂t on
the left-hand side of the equal sign. We have also seen the position oper-
ator ~ r = ~r as well as the momentum operator p ~ = −i~∇ in the previous
chapters.
Operators are not only used for the description of the Schrödinger equa-
tion, they are also related to the values obtained in an experiment, called
observables. We have discussed the measurement of the position and ve-
locity (momentum) of a particle and have seen that these are given by
distribution functions derived from the wave function in coordinate or mo-
mentum space. In the last section of this chapter, we will generalize this
to the observation made for any type of experiment. Here we first define
several properties of operators.

6.4.1 Properties of operators

Operators acting on a ket vector change one state vector into an other. For
an operator O acting on the state vector |αi we can define
O |αi = |βi . (6.50)
If state |αi is an element of Hilbert space V1 then state O |αi must not
necessarily be an element of this Hilbert space. An operator O acts between
two Hilbert spaces V1 and V2 if
∀ |αi ∈ V1 O |αi ∈ V2 . (6.51)

161
Chapter 6. Linear algebra

An operator O acts within a Hilbert space V if

∀ |αi ∈ V O |αi ∈ V. (6.52)

We say that two operators A and B are equivalent (A = B) within a

Hilbert space V if

∀ |αi ∈ V A |αi = B |αi . (6.53)

Operator O is the null operator within a Hilbert space V if

∀ |αi ∈ V O |αi = 0. (6.54)

For three operators A, B and C acting within Hilbert space V addition of

operators is commutative

A + B = B + A, (6.55)

and associative

A + (B + C) = (A + B) + C (6.56)

Most operators encountered in quantum physics are linear. For a linear

operator O acting within a Hilbert space V and two state vectors |αi and
|βi in V we have

O(a |αi + b |βi) = aO |αi + bO |βi , (6.57)

with a, b ∈ C.
Operators act on a ket vector from the left and on a bra vector from
the right. Given the Hilbert space V with element |αi we can define the
dual Hilbert space V † with element hα|. For an operator O acting between
Hilbert spaces V and V † we define

O |αi = |β1 i , and,

hα| O = hβ2 | . (6.58)

162
 6.4. Operators

An operator acting on the right on a ket vector or on the left of a bra

vector is undefined. The ket vector O |αi and the bra vector hα| O are in
general not dual to each other, i.e. |β1 i† and hβ2 | in the previous equation
can but must not be equivalent. We define the Adjoined, or equivalently
the Hermitian conjugate, of an operator O acting on Hilbert space V as
operator O † acting on Hilbert space V † such that

∀ |αi ∈ V (O |αi)† = hα| O † . (6.59)

For two state vectors |αi and |βi in Hilbert space V as well as an operator
acting between V and V † we have
D E D E∗
β O α = α O† β . (6.60)

Operator O acting between Hilbert spaces V and V † is Hermitian if ∀ |αi ∈

V and ∀ |βi ∈ V
D E D E∗
βO α = αO β . (6.61)

If an Hermitian operator acts within Hilbert-space V and the domain of

operator O and O † are equivalent we call the operator self-adjoined and
can write

O † = O. (6.62)

For finite sized Hilbert spaces all Hermitian operators are self-adjoint.

6.4.2 Multiplication and commutation

We can multiply two operators, but generally multiplication of operators is
not commutative. For two operators A and B we can define

AB = BA + [A, B] (6.63)

Whereby the commutator [A, B] is defined as

[A, B] = AB − BA, (6.64)

163
Chapter 6. Linear algebra

such that the previous equation holds trivially.

The Hermitian conjugate of the product of two operators is given as

(AB)† = B † A† (6.65)

6.4.3 Outer product of a ket and bra

For state vectors |αi and |βi in Hilbert space V we can define the product
between a ket and bra state vector as the outer product. The outer product
defines an operator

O |αi→|βi = |βihα| . (6.66)

We can see that this is an operator by applying it to a third state vector

|γi ∈ V :

O |αi→|βi |γi = |βi hα|γi . (6.67)

This operator projects any state vector to the state vector |βi with a com-
plex coefficient given by the inner product hα|γi.

6.4.4 Completeness relation

For an n dimensional Hilbert space V with an orthonormal basis B =
{|β1 i , . . . , |βn i} we can define the operator
n
X
I= |βi ihβi | . (6.68)
i=1

This operator is equivalent to the identity operator and leaves all state
vectors unaltered. For a general state |αi ∈ V we can write
n
X
|αi = aj |βj i . (6.69)
j=1

164
 6.4. Operators

If we calculate the effect of operator I on this state we find

n
X
I |αi = aj |βi i hβi |βj i (6.70)
i,j
Xn
= ai |βi i
i=1
= |αi .

6.4.5 Matrix representation

Any operator O acting within Hilbert space V with orthonormal basis
B = {|β1 i , . . . , |βn i} can be represented as a matrix. In order to derive
the matrix representation of an operator we can focus on how the operator
transforms the basis state vectors and expand an arbitrary state vector
|αi ∈ V on the basis vectors. For the transformation of a single basis state
vector |βi i to a single basis state vector |βj i we define the operator

O |βi i→|βj i = |βj ihβi | , (6.71)

P
such that for a state |αi = k ak |βk i we have
X
O |βi i→|βj i |αi = ak |βj i hβi |βk i (6.72)
k
= ai |βj i .

For a general linear operator in V we can write

X
O= oj,i |βj ihβi | , (6.73)
i,j

with the coefficients oj,i given as

oj,i = hβj |O|βi i . (6.74)

165
Chapter 6. Linear algebra

The elements oj,i define a matrix O. This is the matrix representation of

the operator O
 
o1,1 o1,2 ... o1,n
 o2,1 o2,2 ... o2,n 
O= . ..  . (6.75)
 
.. ..
 .. . . . 
on,1 on,2 . . . on,n

Within this notation a ket state vector in V defines a column vector

 
a1
n
 a2 
X  
|αi = ai |βi i ⇔  .  , (6.76)
 .. 
i=1
an

and a bra state vector in V defines a row vector

n
X
c∗i hβi | ⇔ c∗1 c∗2 . . . c∗n .
 
hγ| = (6.77)
i=1

We can write the expectation value of an operator with the use of matrix
multiplication as
   
o1,1 o1,2 ... o1,n a1
  o2,1 o2,2
 ... o2,n   a2 
 
hγ|O|αi = c∗1 c∗2 . . . c∗n ·  .

..  ·  ..  . (6.78)

.. ..
 .. . . .  .
on,1 on,2 ... on,n an

Whereby one must take care to use the same dual basis for both the operator
representation as well as for the representation of the bra and ket state
vectors.

166
 6.4. Operators

Matrix representation of the Hermitian conjugate of an operator

The matrix representation of the Hermitian conjugate of operator O is the
complex conjugate of the transposed matrix representing operator O.
Let V be a Hilbert space with basis B = {|β1 i , . . . , |βn i}. From equa-
tion 6.60 we know that
D E D E∗
βi O † βj = βj O βi . (6.79)
†
For the matrix elements Oi,j of operator O and Oi,j of operator O † this
yields
(O† )i,j = (O∗ )j,i . (6.80)
From which we see that for the matrix representation O of operator O we
have
(O† ) = (OT )∗ . (6.81)

6.4.6 Basis transformation

Let V be an n dimensional Hilbert space with basis B = {|β1 i , . . . , |βn i}.
We can define a second set of orthonormal states C = {|γ1 i , . . . , |γn i} that
also defines a basis for the Hilbert space V , i.e. the basis of an Hilbert
space is not unique. In order to transform from one basis to an other we
can express the basis vectors of C in terms of the basis vectors B.
X
|γi i = ui,j |βj i , (6.82)
j

with
ui,j = hβj |γi i . (6.83)
For the back transformation we can also express the basis vectors of B in
the basis vectors of C
X
|βj i = ũj,i |γi i , (6.84)
i

167
Chapter 6. Linear algebra

with

ũj,i = hγi |βj i (6.85)

= u∗i,j .

We can define the matrix U with matrix elements ui,j such that a basis
transformation from B to C is given as

C = UB (6.86)

and the back transformation as

B = U †C (6.87)

whereby we used the dagger as symbol for the Hermitian conjugate which we
take within a Hilbert space equivalent to the complex conjugate transpose
of a matrix.
In particularly we find that

B = U † U B, (6.88)

from which it follows that

U †U = 1 (6.89)

or

U † = U −1 . (6.90)

the basis transformation matrix is unitary. As we did not specify any

information on the two basis sets B and C we can state that any basis
transformation is given by a unitary matrix and a unitary matrix can be
used to define a basis transformation.
Instead of transforming the basis vectors |βi i we can transform a state
vector defined on this basis. For |αi ∈ V we can expand the state vector
on basis B and C and require that these result in the same vector.
X
|αi = bj |βj i (6.91)
j

168
 6.4. Operators

X
= ci |γi i .
i

We now can replace the basis vectors of B by those of C and those of C by

those of B to find
X
|αi = bj u∗i,j |γi i (6.92)
i,j
X
= ci ui,j |βj i .
i,j

From which it follows that

X
bj = ui,j ci (6.93)
i
X
ci = u∗i,j bj .
j

If we define the vectors b and c with the elements bj and ci we have

b = UT c (6.94)
∗
c = U b.

Vectors on a basis transform with the complex conjugate of the matrix with
which the basis states themselves transform.

6.4.7 Eigenbasis representation of an operator

Eigenfunctions take a special role in quantum mechanics. We have seen in
previous chapters that the time dependence of an eigenstate of the Hamil-
tonian is particularly simple and given as χ(t) = e−iEt/~ . If we know for
a wave packet the expansion on eigenstates of the Hamiltonian at time
t = 0 we can calculate the wave packet at arbitrary time without too much
effort. In the next section we will see that also for the determination of
the moments of distribution functions obtained from averaging over macro-
scopically many measurements of some observable given by a Hermitian
operator O simplifies, if we know the eigenstates of the operator O.

169
Chapter 6. Linear algebra

Let V be a Hilbert space with dimension n and O a linear Hermitian

operator acting within this Hilbert space. The eigenvectors |ψi i with eigen-
values λ are solutions of the equation

O |ψi i = λ |ψi i . (6.95)

We can rewrite this equation as

(O − λI) |ψi i = 0. (6.96)

For an arbitrary choice of basis, we can write this as a matrix equation with
O an operator represented as an n by n matrix and 1 an n by n dimensional
unity matrix acting on a coefficient vector ψi of |ψi i

(O − λ1)ψi = 0. (6.97)

This equation has the solution ψi = 0 which is not the one we are after, and
solutions for all values of λ where the determinant of the matrix (O − λ1)
is zero. For an n-dimensional matrix the determinant

|O − λ1| = 0, (6.98)

defines an n-th order characteristic polynomial of the complex variable λ.

Using the fundamental theorem of algebra (see appendix E), we can write

|O − λ1| = Πni=1 (oi − λ), (6.99)

with oi ∈ C. We thus have n eigenvalues given by the zero’s of the charac-

teristic polynomial oi .
If |ψi i is an eigenfunction of the Hermitian operator O with eigenvalue
oi and normalized such that hψi |ψi i = 1 we have

hψi |O|ψi i = hψi |oi ψi i (6.100)

= oi .

We also have
D E
ψi O † ψi = hoi ψi |ψi i (6.101)

170
 6.4. Operators

= o∗i .

For a Hermitian operator these are equivalent. We find that the eigenvalues
for a Hermitian operator are real. This is a restatement of the results from
equation 3.51 in section 3.7.

Orthonormality of an eigenbasis
For a Hilbert space of dimension n, we can define operators that act within
that Hilbert space as n by n matrices. The eigenvalues of such a matrix are
given by the zeros of an n-th order characteristic polynomial, which has n
solutions. We find n eigenvectors |ψi i. In this section we will show that
these eigenvectors define an orthonormal basis for V .
If |ψi i is an eigenvector of O then c |ψi i with c ∈ C is also an eigenvector
of O. We can choose the normalization of our eigenvectors such that

hψi |ψi i = 1. (6.102)

We can always choose our eigenvectors to be normalized. We now need to

proof that they are orthogonal. As our eigenvalues are real we know that
for two eigenstates |ψi i and |ψj i we have
D E D E
0 = ψi O † ψj − ψi O ψj (6.103)
= hOψi |ψj i − hψi |Oψj i
= hoi ψi |ψj i − hψi |oj ψj i
= o∗i hψi |ψj i − oj hψi |ψj i
= (oi − oj ) hψi |ψj i

Either the two eigenstates are orthogonal (hψi |ψj i = 0) or they have the
same eigenvalue (oi = oj ). States with the same eigenvalue are called
degenerate states.
For an operator without degenerate eigenstates, the normalized eigen-
vectors are all orthogonal to each other. For an n dimensional Hilbert space
there are n eigenstates and they define an orthonormal basis for V .

171
Chapter 6. Linear algebra

For an operator with degenerate eigenvalues, we can always add a ran-

dom Hermitian operator multiplied with  to this operator such that the
sum of the two operators does not have degenerate eigenvalues. We can
take the limit of  to zero and find that also in this case the normalized
eigenvectors define an orthonormal basis for the Hilbert space.

The eigenbasis as a unitary transformation to diagonalize an op-

erator
Let V be a Hilbert space with dimension n and basis B = {|β1 i , . . . , |βn i}
and let O be a linear operator acting within this Hilbert space with the ma-
(B)
trix representation OB on basis B with matrix elements oi,j = hβi |O|βj i.
The eigenstates of operator O are given as

O |ψi i = oi |ψi i , (6.104)

and can be written as vectors on basis B as

n
X
|ψi i = uj,i |βj i . (6.105)
j=1

The expansion coefficients uj,i define an n by n dimensional unitary matrix

U , whereby the ith row corresponds to the eigenvector of state |ψi i. As the
eigenvectors are orthonormal we find that the matrix U is unitary

U ∗ U T = 1. (6.106)

We can use matrix U to transform the matrix representation OB of operator

O on basis B and find

OA = U ∗ OB U T , (6.107)

with matrix elements

(A)
oi,j = δi,j oi . (6.108)

172
 6.4. Operators

The matrix OA is diagonal with the eigenvalues of operator O on its diag-

onal. Matrix OA is a representation of operator O on the eigenbasis A of
operator O given as

A = {|ψ1 i , . . . , |ψn i}. (6.109)

The relation between the basis A and B is given as

A = U B, (6.110)

whereby we treat the multiplication of a matrix times a set as matrix-vector

multiplication.

6.4.8 Projection operators

The projection operator P B projects states from a Hilbert space A to a
subspace B. In general, we can define the basis of Hilbert space A and
B by the eigenvectors of an operator O. A useful projection operator P
projects a general state |ψk i to a subspace defined by some, but not all,
eigenvectors of operator O.
Given an operator O and its complete orthonormal eigenbasis |ωi,j i with
eigenvalues oi such that

O |ωi,j i = oi |ωi,j i , (6.111)

and oi 6= oi0 ∀ i 6= i0 . Note that we have chosen the indices such that i labels
states with different eigenvalues and j labels different states with the same
eigenvalue. For each oi we will take Ni as the number of linear independent
eigenstates of O that have eigenvalue oi . Ni is called the degeneracy or
multiplicity of the eigenstates with eigenvalue oi of operator O.
A general state |ψk i can be expanded on the eigenbasis as

(k)
X
|ψk i = ci,j |ωi,j i . (6.112)
i,j

173
Chapter 6. Linear algebra

(k)
We can obtain the expansion coefficients ci,j with the help of the projection
operator that projects a state onto the state |ωi,j i

P i,j = |ωi,j ihωi,j | . (6.113)

Acting with the projection operator on a general state |ψk i yields

X (k)
P i,j |ψk i = ci0 ,j 0 P i,j ωi0 ,j 0 (6.114)
i0 ,j 0
(k)
X
= ci0 ,j 0 |ωi,j i ωi,j ωi0 ,j 0
i0 ,j 0
(k)
= ci,j |ωi,j i .

Instead of projecting on a specific state we can project to the Hilbert

space Hoi given by all eigenstates of operator O with eigenvalue oi . The
basis of Hoi is given as

Boi = {|ωi,j=1 i , . . . , |ωi,j=Ni i}. (6.115)

The projection operator to Hoi is given as

Ni
X
Pi = P i,j . (6.116)
j=1

We can express P i in terms of operator O and its eigenvalues as

Y 1
Pi = (O − oi0 ) . (6.117)
o i − o i0
i0 6=i

Acting with P i on a general state |ψk i yields

X (k)
P i |ψk i = ci0 ,j P i ωi0 ,j (6.118)
i0 ,j
X (k)
Y 1
= ci0 ,j (O − oi00 ) ωi0 ,j
oi − oi00
i0 ,j i00 6=i

174
 6.5. Distributions and expectation values

X (k)
Y 1
= ci0 ,j (oi0 − oi00 ) ωi0 ,j .
oi − oi00
i0 ,j i00 6=i

We now notice that oi −o 1

never diverges as i00 = i is excluded from the
i00
00
product and oi 6= oi00 ∀i 6= i . We furthermore notice that
Y
(oi0 − oi00 ) = 0, (6.119)
i00 6=i

except when i0 = i, as for all other values of i0 we have for i00 = i0 a

multiplication by 0. We thus find
X (k)
Y oi0 − oi00
P i |ψk i = ci0 ,j δi0 ,i ωi0 ,j (6.120)
oi − oi00
i0 ,j 00
i 6=i
(k)
X
= ci,j |ωi,j i .
j

The operator P i projects a state onto the Hilbert space Hoi . Note that we
can express P i in terms of O, but can not express P i,j in terms of operator
O.

6.5 Distributions and expectation values

In chapter 2 where we introduced the Schrödigner equation we made the
following statement: ”Given the quantum state described by the wave func-
tion ψ(t, ~r1 , . . . , ~rn ). It follows from the Schrödinger equation, that If we
for this state measure an observable, described by operator O and find as
a result the value o, then if we repeat this experiment many times, start-
ing from the same state ψ(t, ~r1 , . . . , ~rn ) we will find a distribution of values
ρ(t, o). The possible values of o are given by the eigenvalues (spectrum) of
the operator O. The frequency with which these values are found is given
by the norm square of the inner product of the eigenstates of operator O
and the many-particle wavefunction of the system.” We now can argue why
this is the case. Whereby we should note that we can not explain why a

175
Chapter 6. Linear algebra

complex measurement apparatus can yield a single value for the measure-
ment of an observable. But if it does, the values and their frequency follow
the rules outlined above.
We start with the experimental observation that the density of a particle
is given by ρ(t, ~r) = ψ ∗ (t, ~r)ψ(t, ~r). With this, we mean that if we have
an ensemble of N particles, each given by the same state |ψi then ρ(t, ~r)
defines the ensemble-averaged particle density. Whereby it is irrelevant if
we measure all particles at once or one after the other. We now need to
understand what this implies for the measurement of the position of a single
particle and what this implies for the measurements of other operators than
the position operator for a single particle.
If we know that the average over the measurement of the position of
many particles starting from the same state ψ(t, ~r) yields the distribution
function ρ(t, ~r) then the probability to find the position ~r to be in the
volume V for a single measurement of the position is given by the value
r)dd~r. The more involved problem is to understand what this implies
R
V ρ(t, ~
for the measurement of other physical properties.
Instead of the position of a particle, we can look at the velocity, or
momentum, of a particle. For this we can Fourier transform the wave func-
tion from position space to momentum space (see section 4.2.4 or equation
4.42). As a result, we obtain the wave function as a function of the momen-
tum of the particle φ(t, p~). Consistent with the definition of the particle
density distribution we take the momentum distribution of a particle to be
ρ(t, p~) = φ∗ (t, p~)φ(t, p~). For a single particle given by the state φ(t, p~) you
will
R measure the momentum p~ to be in the volume V with the probability
ρ(t, p
~)ddp
~ .
V

We can quantify the density distribution ρ(t, ~r) or the momentum dis-
tribution ρ(t, p~) by the moments of these distributions. In section 4.2.4 we
defined the nth moment of the momentum distribution as the expectation
value of the operator (~p)n (see equation 4.44). In equation 4.53 and 4.54 we
defined two equivalent ways to determine the moments of the momentum
distribution. This concept is so important that we repeat these equations

176
 6.5. Distributions and expectation values

here (We change to the one-dimensional case to simplify notation).

Z ∞
(n)
µp (t) = φ(t, p)∗ pn φ(t, p)dp (6.121)
−∞
∂ n
Z ∞  
= ψ(t, x)∗ −i~ ψ(t, x)dx.
−∞ ∂x
The momentum operator in momentum space is represented as p = p. In
∂
real space the momentum operator is represented as p = ~i ∂x . Without
specifying the basis of the Hilbert space we have
µ(n) n
p (t) = hψ(t)|p |ψ(t)i . (6.122)
In the previous sentence we presented the momentum operator on two
different sets of basis vectors. The relation between them is given by a basis
transformation from position space to momentum space. For a general
operator O which measures the variable o one would obtain, if one repeats
an experiment very often, a frequency distribution ρ(t, o). We can quantify
this function by its moments.
Z ∞
(n)
µo (t) = on ρ(t, o)do (6.123)
−∞
= hψ(t)|On |ψ(t)i .
(n)
In order to derive the distribution ρ(t, o) from its moments µo (t) we
can write the operator O on its eigenbasis. We take the eigenbasis of O as
Ω = {|ω1 i , . . . , |ωn i} with
O |ωi i = oi |ωi i . (6.124)
The operator O on its eigenbasis is written as
X
O= oi |ωi ihωi | (6.125)
i

and is diagonal. Such that the nth power of operator O yields

!n
X
On = oi |ωi ihωi | (6.126)
i

177
Chapter 6. Linear algebra

X
= oni |ωi ihωi | .
i

The n-th moment of the distribution ρ(t, o) then becomes

µ(n) n
o (t) = hψ(t)|O |ψ(t)i (6.127)
X
= hψ(t)|ωi i oni hωi |ψ(t)i
i
X
= |hωi |ψ(t)i|2 oni .
i

The distribution that has these moments is given as

X
ρ(t, o) = |hωi |ψ(t)i|2 δo,oi . (6.128)
i

6.5.1 (partial) Density Of States

One special distribution function is the energy distribution. If we take an
ensemble of particles in the state |ψi (t)i then these generally will have a
distribution of different energies. In order to calculate this distribution we
need to project the state |ψi (t)i onto the eigenstates of the Hamiltonian
|φj i with H |φj i = Ej |φj i. The states |φj i are the stationary states of the
t
system that show a trivial time evolution χj (t) = e−i ~ Ej . We find for the
distribution function of the energy for state |ψi (t)i
X
ρi (t, E) = |hφj |ψi (t)i|2 δE,Ej . (6.129)
j

The function ρi (t, E) is known as the partial density of states. Note that
Z ∞ Z ∞X
ρi (t, E)dE = |hφj |ψi (t)i|2 δE,Ej dE (6.130)
−∞ −∞ j
X
= |hφj |ψi (t)i|2
j

178
 6.5. Distributions and expectation values

X
= hψi (t)|φj i hφj |ψi (t)i
j

= hψi (t)|ψi (t)i

= 1.

Instead of looking at the partial density of states for a single state we

can look at the density of states for all states in the system. We define the
Density Of States as
X
DOS(E) = ρi (t, E) (6.131)
i
XX
= |hφj |ψi (t)i|2 δE,Ej
i j
X
= δE,Ej .
j

R E+dE
The integral E DOS(E)dE gives the information on the number of
states one finds in the energy interval E to E + dE. The density of states is
a very useful concept to describe many properties in physics. In section 1.1
we calculated the black body radiation. In order to do so, we determined
the density of electromagnetic modes, which is the density of states, inside
a cavity. As you will see in later lectures, the density of states plays a
central role in understanding many physical properties, from the stability
of Neutron stars to the functioning of semiconductor solar cells.

179
Chapter 6. Linear algebra

180
Chapter 7

Symmetry and conservation

laws

Symmetry plays an immensely important role in physics. We say a sys-

tem has a certain symmetry if a given observable is unchanged under some
transformation. For example, most physical systems are invariant under
time translation. If I do an experiment today or tomorrow we should get
the same answer. Time translation is a symmetry of our Hamiltonian in
free space. The same holds true for position translation. It does not matter
where we do an experiment, we should get the same answer. We will show
that for each symmetry operation of a system we can define a conserved
quantity in that system. The relation between symmetry and conserved
quantities is known as Noether’s theorem [69, 70]. The Noether theorem is
valid in both classical and quantum physics and holds for continues sym-
metry operators (as in the original paper of Noether) as well as for discrete
symmetry operations. Well-known examples of continues symmetries re-
lated to conserved quantities are time translation with the conserved quan-
tity the energy of the system and position translation with the conserved
quantity the momentum of the system. The statement that a system is
invariant under time (position) translation is equivalent to the statement
that we have energy (momentum) conservation. An example for discrete

181
Chapter 7. Symmetry and conservation laws

symmetry would be inversion symmetry with the conserved quantity the

parity of a state. Besides conservation laws that follow from symmetry
operations, we will also find that symmetry helps us to separate a large
problem into several smaller problems. The use of symmetry arguments
makes involved problems easier to solve.

7.1 Symmetry and commutation relations

We start our discussion of sym-
metry with the example of trans- 𝞧(x) Tx0(𝞧(x))
lation symmetry for a one dimen-
sional system. We define the oper- x0
ator T x0 that shifts all functions to
the right by a value x0
Figure 7.1: The translation operator
T x0 ψ(x) = ψ(x − x0 ). (7.1) T : x 7→ x − x applied to the func-
x0 0
The operator T x0 can be imple- tion ψ(x).
mented via a coordinate transformation
T x0 : x 7→ x − x0 . (7.2)
We will use the same symbol for the operator, independent if it acts on
coordinates or on functions. In figure 7.1 we plot both ψ(t, x) as well
as T x0 ψ(t, x). In order for T x0 to be a symmetry operator of our sys-
tem the Hamiltonian must be invariant under this transformation. For
the Hamiltonian to be invariant under transformation we require that
T x0 V (x) = V (x), the potential is translational invariant. This is the case
when V (x + x0 ) = V (x). The potential must be a constant, V (x) = V0 .
The Hamiltonian that is invariant under translation becomes
~2 ∂ 2
H=− + V0 . (7.3)
2m ∂x2
The stationary solutions of the Schrödinger equation with constant poten-
tial are
ψk (t, x) = φk (x)χk (t), (7.4)

182
 7.1. Symmetry and commutation relations

φk (x) = eikx ,
χk (t) = e−iE(k)t/~ ,
~2 k 2
E(k) = + V0 .
2m
It is tempting to think that for a system that is translational invariant
the equation T x0 H = H should hold. This is not true. We can easily see
that T x0 H 6= H by evaluating both T x0 H and H appleid to a stationary
state.

T x0 Hψk (t, x) = T x0 (Hψk (t, x)) (7.5)

= T x0 E(k)ψk (t, x)
= E(k)ψk (t, x − x0 )
= E(k)e−ikx0 ψk (t, x),

but

Hψk (t, x) = E(k)ψk (t, x). (7.6)

The energy of the shifted state is the same as the energy of the unshifted
state (E(k)). The shifted state is (for non degenerate states) given by the
same state ray, but not by the same state vector. What is true is that

T −1
x0 HT x0 = H, (7.7)

with T −1x0 T x0 = I. We can shift the state, apply the Hamiltonian and
shift it back and find the same state as we would have obtained if we just
applied the Hamiltonian to the original state. Multiplying equation 7.7 by
T x0 yields that

HT x0 = T x0 H. (7.8)

Expressed with the use of the commutator we find that

[T x0 , H] = 0. (7.9)

183
Chapter 7. Symmetry and conservation laws

In general we define that operator O is a symmetry operation of a

system with Hamiltonian H if

[O, H] = 0. (7.10)

We furthermore restrict ourselves, for the moment, to symmetry operators

that are linear and unitary when acting within an Hilbert space, such that

O † = O −1 . (7.11)

7.2 Symmetry and conservation laws

Each symmetry operation of a system can be related to a conserved quantity
of the system. The conserved quantity is the expectation value of the
symmetry operator. In order to proof this we can have a look at the time
derivative of the expectation value of a symmetry operator. We need to
proof that
∂
hψ(t)|O|ψ(t)i = 0, (7.12)
∂t
for all states |ψ(t)i if [H, O] = 0. If the time derivative is zero the expec-
tation value is time independent and we call the value conserved. As a first
i
step to proof this we insert the time evolution operator (U (t, 0) = e− ~ Ht
equation 3.33) into equation 7.12
∂ ∂ D − i Ht i
E
hψ(t)|O|ψ(t)i = e ~ ψ(t = 0) O e− ~ Ht ψ(t = 0) . (7.13)
∂t ∂t
i
The
P∞ exponent of an operator can be written as a Taylor series e− ~ Ht =
1 i n
n=0 n! (− ~ Ht) . With the use that H is Hermitian we have

∂ ∂ D i i
E
hψ(t)|O|ψ(t)i = ψ(t = 0) e+ ~ Ht Oe− ~ Ht ψ(t = 0) . (7.14)
∂t ∂t
We changed the time dependence from the wave function to the opera-
tor, which sometimes is called the Heisenberg representation (contrasting

184
 7.2. Symmetry and conservation laws

the Schrödinger representation where the time dependence is in the wave

functions).
As a next step to proof that the expectation value of O is time in-
dependent we use that O is a symmetry operation of the system with
Hamiltonian H. From the definition of a symmetry operator we know that
[O, H] = 0. If O and H commute then also their powers commute, in
particular [O, H n ] = 0 ∀n ∈ N. We find:

∂ ∂ D E
hψ(t)|O|ψ(t)i = ψ(t = 0) e+iHt/~ Oe−iHt/~ ψ(t = 0) . (7.15)
∂t ∂t
∂ D E
= ψ(t = 0) e+iHt/~ e−iHt/~ O ψ(t = 0) .
∂t
∂
= hψ(t = 0)|O|ψ(t = 0)i .
∂t
= 0.

If O is a symmetry operation of a system with Hamiltonian H such

that [O, H] = 0 then the expectation value of O is time independent and
conserved.
Note that equation 7.15 is valid for all states |φ(t)i and thus not only
the expectation value is conserved, but also the operator e+iHt/~ Oe−iHt/~
is time independent such that

∂  +iHt/~ −iHt/~ 
e Oe = 0. (7.16)
∂t

7.2.1 Translational symmetry and momentum conservation

In the beginning of section 7.1 we discussed the example of translational
symmetry. In this section we ask the question which conserved quantity is
related to translational symmetry. We take the operator T ~r0 that translates
a wave function by ~r0 . We can represent this operator in different ways,
depending on the object it acts upon. We can act on a function

T ~r0 Ψ(t, ~r) = Ψ(t, ~r − ~r0 ), (7.17)

185
Chapter 7. Symmetry and conservation laws

make a coordinate transformation

T ~r0 : ~r 7→ ~r − ~r0 , (7.18)

or act on a state in a complex Hilbert space

Z
T ~r0 |Ψ(t)i = |~r + ~r0 i h~r|Ψ(t)i dd~r. (7.19)
Rd

In oder to know the conserved quantity that belongs to this operator we

need to calculate

hΨ(t)|T ~r0 |Ψ(t)i , (7.20)

for some arbitrary wave packet Ψ(t, ~r).

It is useful to represent the operator T ~r0 by a differential operator. This
is a similar procedure as we did in chapter 2 for the energy and momen-
tum operator. For the translation operator T ~r0 acting on the plane wave
~
ψ(t, ~r) = ei(k·~r−ωt) we find
~ ~
T ~r0 ei(k·~r−ωt) = ei(k·(~r−~r0 )−ωt) (7.21)
r0 ·~k i(~k·~
−i~ r−ωt)
=e e
∞
X 1  n
~
= −i~r0 · ~k ei(k·~r−ωt)
n!
n=0
∞
X 1 ~
= (−~r0 · ∇)n ei(k·~r−ωt)
n!
n=0
~
= e−~r0 ·∇ ei(k·~r−ωt) .

We thus guess that a differential operator representation of a translation

by ~r0 is given as

T ~r0 = e−~r0 ·∇ (7.22)

− ~i ~
r0 ·~
p
=e .

186
 7.2. Symmetry and conservation laws

In order to proof that the translation operator is indeed given by equation

7.22 we can look at the effect of this operator on a general state ψ(t, ~r).
Any state can be written as a sum over plane waves (see for example section
4.2.1)
 d/2 Z
1 ~ ~
ψ(t, ~r) = c(~k)ei(k·~r−E(k)t/~) dd~k. (7.23)
2π Rd

For the translation of this state by a vector ~r0 we find

 d/2 Z
1 ~ ~
T ~r0 ψ(t, ~r) = T ~r0 c(~k)ei(k·~r−E(k)t/~) dd~k (7.24)
2π Rd
 d/2 Z
1 ~ ~
= e−~r0 ·∇ c(~k)ei(k·~r−E(k)t/~) dd~k
2π R d
 d/2 Z
1 ~ ~
= c(~k)e−~r0 ·∇ ei(k·~r−E(k)t/~) dd~k
2π R d
 d/2 Z
1 ~ ~ ~
= c(~k)e−ik·~r0 ei(k·~r−E(k)t/~) dd~k
2π Rd
 d/2 Z
1 ~ ~
= c(~k)ei(k·(~r−~r0 )−E(k)t/~) dd~k
2π Rd
= ψ(t, ~r − ~r0 ).
Which proofs that equation 7.22 indeed defines a translation operator. Note
that we can derive equation 7.22 also by writing the translation operator
as an infinite product of infinitesimal small translations.
N
Y
T ~r0 ψ(t, ~r) = lim T 1
~
r (7.25)
N →∞ N 0
i=1
N  
Y 1
= lim 1− ~r0 · ∇
N →∞ N
i=1
N
Y 1
= lim e− N ~r0 ·∇
N →∞
i=1

187
Chapter 7. Symmetry and conservation laws

= e−~r0 ·∇ .

We now can use the differential operator representation of the trans-

lation operator to determine the conserved quantity of a system that has
∂
translation symmetry. When [H, T ~r0 ] = 0, then ∂t hψ(t)|T ~r0 |ψ(t)i = 0 for
all states |ψ(t)i, such that hψ(t)|T ~r0 |ψ(t)i is a conserved quantity. We find
that
D i
E
hψ(t)|T ~r0 |ψ(t)i = ψ(t) e− ~ ~r0 ·~p ψ(t) (7.26)
Z
i
= φ∗ (t, p~)e− ~ ~r0 ·~p φ(t, p~)dd p~
d
ZR
i
= φ∗ (t, p~)e− ~ ~r0 ·~p φ(t, p~)dd p~
d
ZR
i
= e− ~ ~r0 ·~p ρp~ (t, p~)dd p~,
Rd

is conserved for all values of ~r0 . With ρp~ (t, p~) = φ∗ (t, p~)φ(t, p~) the momen-
tum distribution of the particle. If the Fourier transform of the momentum
distribution of the particle is a conserved quantity, then the momentum dis-
(n)
tribution itself is a conserved quantity. In particular we find that µr̂0 ·~p , the
nth moment of the momentum distribution in the ~r0 direction (see section
4.2.4) is conserved.
∂
If [H, T ~r0 ] = 0 for all ~r0 in Rd then ∂t ρp~ (t, p~) = 0. It follows that
ρp~ (t, p~) = ρp~ (t = 0, p~). For systems with translational invariance the mo-
mentum distribution of a particle is conserved. In particular the first mo-
ment of this distribution is conserved. Translation symmetry leads to mo-
mentum conservation. This is a result we, without commenting, already
obtained in chapter 4 on wave mechanics. The propagating wave packets
discussed in that chapter had a momentum distribution, or equivalent a
wave vector distribution ρk (t, k) that is independent of time.
In chapter 2 we defined the momentum operator as ~p = −i~∇. We here
find that if a system has translational symmetry, then the expectation value
of the operator −i~∇ is conserved. This is an other reason to define the
operator −i~∇ as the operator representing the momentum of the particle.

188
 7.2. Symmetry and conservation laws

Note that this is not a proof that the momentum operator needs to be
~p = −i~∇, but shows the internal consistency of our definitions.

7.2.2 Time invariance and energy conservation

Instead of a translation in space, as discussed in the previous section, we
can translate our system in time. The operator

T δt : t 7→ t − δt, (7.27)

translates the state by a time δt. If the potential in the Hamiltonian is time
independent we find that

[H, T δt ] = 0, (7.28)

and
∂
hΨ(t)|T δt |Ψ(t)i = 0. (7.29)
∂t
The value hΨ(t)|T δt |Ψ(t)i is a conserved quantity for all states |Ψ(t)i. We
here want to ask which physical observable this conserved quantity corre-
sponds to.
Following the same line of reasoning as we did in the previous section
(7.2.1) we want to express T δt as a differential operator. For a system
that is translational invariant we know that the potential is a constant
and the stationary states of the Hamiltonian are plane wave functions. A
system that is invariant under time translation can have an arbitrary, time
independent potential. The stationary states are not necessarily given by
plane waves. The stationary states do define a complete set that we can use
as a basis. We will write our states |Ψ(t)i as a sum over stationary states.
Any state |Ψ(t)i ∈ DH can be expanded on the set of stationary states as
X Z
Ψ(t, ~r) = cn φn (~r)χn (t) + c(κ)φ(κ, ~r)χ(κ, t)dκ. (7.30)
n

189
Chapter 7. Symmetry and conservation laws

Whereby φn (~r)χn (t) form the discrete set of stationary states and φ(κ, ~r)χ(κ, t)
form a continues set of stationary states of the Schrödinger equation. For
the stationary states we know that
i i
χn (t) = e− ~ En t , χ(κ, t) = e− ~ E(κ)t , (7.31)
Hφn (~r) = En φn (~r), Hφ(κ, ~r) = E(κ)φn (~r).

For the time translation operator acting on a general state we find

Z !
X
T δt Ψ(t, ~r) = T δt cn φn (~r)χn (t) + c(κ)φ(κ, ~r)χ(κ, t)dκ (7.32)
n
X Z
= cn φn (~r)χn (t − δt) + c(κ)φ(κ, ~r)χ(κ, t − δt)dκ
n
X i
= cn φn (~r)e− ~ En δt χn (t)
n
Z
i
+ c(κ)φ(κ, ~r)e− ~ E(κ)δt χ(κ, t)dκ
X ∂
= cn φn (~r)e−δt ∂t χn (t)
n
Z
∂
+ c(κ)φ(κ, ~r)e−δt ∂t χ(κ, t)dκ
Z !
∂ X
−δt ∂t
=e cn φn (~r)χn (t) + c(κ)φ(κ, ~r)χ(κ, t)dκ .
n

The time translation operator in differential form is thus given as

∂
T δt = e−δt ∂t (7.33)
−δt ~i H
=e ,
∂
whereby we used that H = i~ ∂t as defined in the Schrödinger equation.
Note that we just derived the unitary time translation operator U (t+δt, t) =
e−iHδt/~ as encountered in equation 3.33 but now starting from symmetry
arguments.

190
 7.2. Symmetry and conservation laws

D ∂
E D i
E
We find that both Ψ(t) e−δt ∂t Ψ(t) and Ψ(t) e−δt ~ H Ψ(t) are con-
served quantities of motion. AsDthey are conservedE quantities of motion for
∂ n
any value of δt we know that Ψ(t) ∂t Ψ(t) and hΨ(t)|H n |Ψ(t)i are

conserved quantities of motion for all n ∈ N. The previous statements

follow from the general argumentation of the relation between operators
that commute with the Hamiltonian and conserved quantities as given in
the beginning of this section. For the specific case

of time translation we
∂ n
can also show that the expectation values of ∂t or H n are conserved by
evaluating the inner product
of a general state expanded on the stationary
∂ n
states. For the operator ∂t we find
  n   n
∂ ∂ ∂ X ∗ ∗ ∂
Ψ(t) Ψ(t) = cn1 cn2 hφn1 |φn2 i χn1 (t) χn2 (t)
∂t ∂t ∂t n n ∂t
1 2
Z Z  n
∂ ∗ ∗ ∂
+ c(κ1 ) c(κ2 ) hφκ1 |φκ2 i χ (κ1 , t) χ(κ2 , t)dκ1 dκ2
∂t ∂t
 n
∂ X ∗ i
= cn1 cn2 δn1 ,n2 χ∗n1 (t) − En2 χn2 (t)
∂t n n ~
1 2
Z Z  n
∂ ∗ ∗ i
+ c(κ1 ) c(κ2 )δ(κ1 − κ2 )χ (κ1 , t) − E(κ2 ) χ(κ2 , t)dκ1 dκ2
∂t ~
 n Z  n
∂ X
2 i ∂ 2 i
= |cn1 | − En1 + |c(κ1 )| − E(κ1 ) dκ1
∂t n ~ ∂t ~
1

= 0. (7.34)

Similarly we find for the operator H n

∂ ∂ X ∗
hΨ(t)|H n |Ψ(t)i = c cn hφn1 |H n |φn2 i χ∗n1 (t)χn2 (t)
∂t ∂t n n n1 2
1 2
Z Z
∂
+ c(κ1 )∗ c(κ2 ) hφκ1 |H n |φκ2 i χ∗ (κ1 , t)χ(κ2 , t)dκ1 dκ2
∂t
∂ X ∗
= c cn δn ,n (En2 )n χ∗n1 (t)χn2 (t)
∂t n n n1 2 1 2
1 2

191
Chapter 7. Symmetry and conservation laws

Z Z
∂
+ c(κ1 )∗ c(κ2 )δ(κ1 − κ2 ) (E(κ2 ))n χ∗ (κ1 , t)χ(κ2 , t)dκ1 dκ2
∂t
Z
∂ X ∂
= |cn1 |2 (En1 )n + |c(κ1 )|2 (E(κ1 ))n dκ1
∂t n ∂t
1

= 0. (7.35)
The energy of a single stationary state φn (~r)χn (t) (φ(κ, ~r)χ(κ, t)) is En
(E(κ)). A general state Ψ(t, ~r) can be written as a linear superposition
of all stationary states as given in equation 7.30. The energy of Ψ(t, ~r) is
determined by the expansion coefficients of Ψ(t, ~r) on the stationary states
X Z
EΨ = |cn |2 En + |c(κ)|2 E(κ)dκ. (7.36)
n

For systems that are time translation invariant EΨ is a conserved quantity

∂
( ∂t EΨ = 0). To be more specific, the coefficients cn and c(κ) define an
energy distribution
X Z
ρE (t, E) = |cn | δ(E − En ) + |c(κ)|2 δ(E − E(κ))dκ.
2
(7.37)
n

This distribution of energy is conserved in time.

Given a system with N strictly non-interacting particles in the state
|Ψ(t)i. If we have an experimental setup that measures the energy of these
particles, then we should find the distribution function ρE (t, E) for the
energies measured. Whereby it does not matter at which time we do the
measurement.

7.2.3 Conservation laws and distribution functions

In section 6.5 we argued that if we for a macroscopic number of strictly
independent particles in state |ψ(t)i measure an observable described by
operator O we find a distribution function ρo (t, o) for the measured values
(equation 6.128)
X
ρo (t, o) = |hωi |ψ(t)i|2 δo,oi . (7.38)
i

192
 7.3. Symmetry groups

Whereby |ωi i are the eigenstates of operator O with eigenvalues oi . If O is

a symmetry operator of the system then the distribution function is time
independent and ρo (t, o) = ρo (t = 0, o) ∀t.
If we have an experimental setup that gives a particular value of o for
each measurement on a single particle we will find that each measurement
at time t will have a different value. Only if we build a frequency distribu-
tion ρo (t, o) stating how often we measure the value o as a function of the
value measured, do we find that this distribution is time-independent. In
order to understand how one goes from a wave function Ψ(t, ~r) with a given
distribution function ρo (t, o) for given operator O to the measurement of
a specific value of o in the instance of a measurement on a single parti-
cle is topic of discussion between the different interpretations of quantum
mechanics.

7.3 Symmetry groups

In the previous section we have seen that if a system has some symmetry
there is a conserved quantity related to this symmetry. Most systems have
more than one symmetry, i.g. a system can be both invariant under time
and space translation. The set of all symmetry operations of a physical
system forms a group. Groups are sets with special properties. Several of
these properties are useful if one wants to determine the eigenstates of the
Hamiltonian. This section is by no means complete but forms a minimal
introduction to symmetry groups.
We first define some notation. If we successively apply two symmetry
operations, O 1 and O 2 to our system then we will write the combined
transformation by the composition of the two operators, i.e. O 2 O 1 . For
operators acting on a finite Hilbert space, we can express the symmetry
operators as matrices. Operator composition in this case is equivalent to
matrix multiplication.
A symmetry group G is a set of symmetry operators O i for which under
composition, i.e. the successive application of the symmetry operator, four
conditions are satisfied.

193
Chapter 7. Symmetry and conservation laws

• The symmetry group is closed under composition

∀O 1 , O 2 ∈ G O 1 O 2 ∈ G. (7.39)

• Composition is associative

∀O 1 , O 2 , O 3 ∈ G (O 1 O 2 )O 3 = O 1 (O 2 O 3 ). (7.40)

• The symmetry group contains the identity operator I defined such

that

∃O 1 = I ∈ G : O 1 O 2 = O 2 ∀O 2 ∈ G. (7.41)

• The symmetry group contains the inverse of each element

∀O 2 ∈ G ∃O 1 ∈ G : O 1 O 2 = O 2 O 1 = I. (7.42)

Symmetry operations can either be discrete or continues. For example,

translational symmetry is a continues symmetry. The set of all translation
operators T x0 : x 7→ x − x0 ∀x0 ∈ R define a continues symmetry group.
We can test that the set of all translation operators is a group by going
through the four properties of a group discussed above. Under composition
we have T x0 T x1 = T x2 =x0 +x1 . In particular T x2 is a translation and an
element of the symmetry group. Translation operators are associative. A
translation by x0 = 0 yields the identity operator T 0 = I. We furthermore
have T −x0 T x0 = T 0 = I ∀x0 ∈ R, such that each translation has an inverse
translation. For the set of all translation operators the four conditions of a
group are fulfilled. Discrete groups are related to non-continuous changes
of a system. The inversion operator in the point x0 , for example, defines
a discrete group. The inversion group consists of two symmetry elements,
I : x 7→ x and ix0 : x 7→ x0 − x. We can test that the inversion group is a
group by going through the four conditions that need to be fulfilled for a
group.
A system can be invariant under the symmetry operations of more than
one group. The Hamiltonian in one-dimensional space with V (x) = 0

194
 7.4. Eigenfunctions and irreducible representations

is invariant under translation, but also under inversion. The combined

symmetry operations that leave the system invariant need to form a group.
If a system is translation invariant and invariant under inversion in the point
x0 = 0 then it must be inversion invariant for each point x0 . All inversion
operators in all points x0 become part of the same group if a system has
translational and inversion symmetry. We can classify the different groups
by the type of symmetry operations they have. The number of relevant
symmetry groups is modest such that one can make lists and tables of
these symmetry groups.

7.4 Eigenfunctions and irreducible representations

With the use of the symmetry operators of the group we can split the
Hilbert space of the full problem into several invariant subspaces. Given a
set of symmetry operators G, with elements O k that act on a Hilbert space
V . We call W an invariant subspace, if W is a subspace of V and

O k |ψi ∈ W ∀O k ∈ G and ∀ |ψi ∈ W. (7.43)

Note that V is an invariant subspace of itself. We can express the symmetry

operators O k as matrices Ok on the invariant subspace W . The set of all
matrices Ok for all symmetry operators in G defines a representation of the
group. We call a subspace W irreducible if W 6= ∅ and the only invariant
subspaces of W are W and ∅. The corresponding representation, given by
the matrices Ok , is called an irreducible representation. The trace over the
matrices Ok defines a set of complex numbers χ = {o1 , o2 , . . .}, which we
will call the character of the irreducible representation1 . We can classify
the different irreducible representations according to their character. We
say a state |ψi belongs to a given irreducible representation if it can be
used as a basis state for the corresponding irreducible subspace W .
1
We do not need to consider all symmetry operations to define the character. We
can group several symmetry operations together, that have the same trace for all matrix
representations of the symmetry operators, into so-called conjugacy classes.

195
Chapter 7. Symmetry and conservation laws

The previous paragraph made several definitions about groups and rep-
resentations. We can use these definitions to make several important state-
ments about the eigenstates of the Hamiltonian. The eigenfunctions of the
Hamiltonian are related to the eigenfunctions of the symmetry operators
of the system. The relations between the eigenstates of the Hamiltonian
and the eigenstates of the symmetry operators, or the irreducible repre-
sentations, become much easier if we assume that the eigenvalues of the
symmetry operators and the eigenvalues of the Hamiltonian are all non-
degenerate. We call the eigenvalues on of operator O degenerate if several
linearly independent eigenstates have the same eigenvalue. We call two
states degenerate (with respect to operator O) if they are linearly indepen-
dent and have degenerate eigenvalues:

O |ψm i = om |ψm i , and (7.44)

O |ψn i = on |ψn i , and
hψn |ψm i = 0, and
on = om .

For a state that is degenerate (with other states), we define the multiplicity,
or degeneracy N as the number of linear independent states |ψn i with
eigenvalue on , such that all other eigenstates of O have an eigenvalue that
is different from on .
We will restrict our proofs to non-degenerate states, the statements
made can be extended to the case with degeneracies. If [H, O] = 0 and
|ψi is a non-degenerate eigenstate of O then |ψi is also an eigenstate of H.
Given

O |ψi = o |ψi , (7.45)

we find

OH |ψi = HO |ψi (7.46)

= Ho |ψi
= oH |ψi .

196
 7.4. Eigenfunctions and irreducible representations

The state |ψi and H |ψi are both eigenstates of the operator O with the
same eigenvalue o. As the state |ψi is a non-degenerate eigenstate of opera-
tor O we know that |ψi and H |ψi describe the same state. Not necessarily
the same state vector, but the same state ray. We find

H |ψi = E |ψi , (7.47)

with E in principle a complex number. As H is hermitian we know that

E is real. Similarly we find that if |ψi is a non-degenerate eigenstate of H
then

O |ψi = o |ψi , (7.48)

with o a complex eigenvalue, as symmetry operators do not need to be

Hermitian. Furthermore we know that |o| = 1, as symmetry operators
should not change the norm of a state. We can classify the different non-
degenerate eigenstates of the Hamiltonian H |ψi = E |ψi by the complex
eigenvalues ok of all the symmetry operations O k of the symmetry group
of the system. For a symmetry group G with symmetry operations O k we
find

O k |ψi = ok |ψi . (7.49)

The eigenvalues ok define a set of complex numbers χ = {o1 , o2 , . . .} which

we call the character of state |ψi 2 . Two non-degenerate eigenstates of
the Hamiltonian with a different character belong to a different irreducible
representation. The idea of classifying objects by how they behave under
different symmetry operations is something one has encountered before.
When we discussed if a function is even or odd under inversion symmetry we
look at the eigenvalue (+1 or −1) of this function under inversion symmetry.
The classification into irreducible representations is an extension of this
notion to the case where we have more than one symmetry operation.
2
We do not need to consider the eigenvalues of all symmetry operations for the char-
acter of a state. Those symmetry operators that belong to the same conjugate class have
the same eigenvalue.

197
Chapter 7. Symmetry and conservation laws

A linear combination of two states that belong to the same irreducible

representation also belongs to that irreducible representation. As an ex-
ample, the linear combination of two odd states is still an odd state. It is
often much easier to find the eigenstates of the symmetry operators than
to find the eigenstates of the Hamiltonian. As such we can often determine
sets of states which belong to different irreducible representations, but are
not necessarily eigenstates of the Hamiltonian. We can use these states
that belong to an irreducible representations as convenient basis for the
Hamiltonian. Two states that belong to a different irreducible representa-
tion do not interact with each other. Given two states |ψ1 i and |ψ2 i that
are eigenstates of symmetry operator O such that

O |ψ1 i = eiα1 |ψ1 i (7.50)

O |ψ2 i = eiα2 |ψ2 i

we find

hψ1 |H|ψ2 i = ψ1 O −1 HO ψ2 (7.51)

D E
= ψ1 O † HO ψ2
= hOψ1 |H|Oψ2 i
= eiα1 ψ1 H eiα2 ψ2
= ei(α2 −α1 ) hψ1 |H|ψ2 i .

It follows that either α1 = α2 or hψ1 |H|ψ2 i = 0. If α1 = α2 for all symmetry

operators O k that belong to the symmetry group of the system the states
|ψ1 i and |ψ2 i belong to the same irreducible representation. If α1 6= α2
for one of the symmetry operators O k we find that the states belong to a
different irreducible representation and hψ1 |H|ψ2 i = 0.
As an example we can look at the eigenstates of the Hamiltonian of a
particle in one dimension without an external potential

p2
H= . (7.52)
2m

198
 7.4. Eigenfunctions and irreducible representations

The eigenstates of the Hamiltonian Hψk (x) = Ek ψk (x) are given by plane
waves

ψk (x) = eikx , (7.53)

with eigenvalues

~2 k 2
E(k) = . (7.54)
2m
The Hamiltonian commutes with the translation operator T x0 : x 7→
x − x0 and the inversion operator ix0 : x 7→ x0 − x. These two symmetry
operators define a group. The eigenstates and eigenvalues of the translation
operator become evident if we act with the translation operator on a plane
wave

T x0 eikx = e−ikx0 eikx . (7.55)

Such that the states eikx are eigenvectors of the translation operator with
eigenvalue e−ikx0 . For infinitesimal small x0 each eigenstate of T x0 has its
own eigenvalue such that the states are non-degenerate. We thus find that
the eigenstates of the translation operator are also the eigenstates of the
Hamiltonian. The eigenstates and eigenvalues of the inversion operator ix0
are given as
Z ∞  
1 1
ix0 c(k) eik(x− 2 x0 ) ± e−ik(x− 2 x0 ) dk = (7.56)
−∞
Z ∞  
1 1
± c(k) eik(x− 2 x0 ) ± e−ik(x− 2 x0 ) dk,
−∞

with c(k) a complex function. The inversion symmetry operator only dis-
tinguishes between even and odd states. If we look at the eigenstates of the
Hamiltonian we find that the state with momentum k is degenerate with
the state with momentum −k. The state eikx is not an eigenstate of the
inversion symmetry operator. The state eikx is neither even nor odd. We
can make a linear combination between the degenerate states at k and −k

199
Chapter 7. Symmetry and conservation laws

2 2 k
to generate an even and an odd state with energy ~2m . This is possible
because [T x0 , H] = 0 and [ix0 , H] = 0, but [T x0 , ix00 ] 6= 0.
All non-degenerate eigenstates of the symmetry operator T x0 are si-
multaneously eigenstates of the Hamiltonian. Linear combinations of the
degenerate eigenstates of the Hamiltonian with momentum k and −k are
eigenstates of the inversion operator ix0 .

7.5 Rotational symmetry and angular momentum

conservation
In two or more dimensions objects can rotate. Rotational invariance leads
to the conservation of angular momentum. Rotations are special symmetry
operations and immensely important for the discussion of many physics
problems. As such the results we find in this section will be crucial for
the explanation of many of the following chapters. Any Hamiltonian of the
form
~2
H=− ∆ + V (|~r|) (7.57)
2m
is invariant under rotation around the point ~r = 0. Rotations in R3 can
be represented as 3 by 3 matrices that transform a coordinate vector ~r via
matrix multiplication

R : ~r 7→ R~r (7.58)

The rotation matrices for a rotation around the x, y or z axis are given as
 
1 0 0
Rx (α) = 0 cos(α) − sin(α) , (7.59)
0 sin(α) cos(α)

 
cos(β) 0 sin(β)
Ry (β) =  0 1 0 , (7.60)
− sin(β) 0 cos(β)

200
 7.5. Rotational symmetry and ...

and
 
cos(γ) − sin(γ) 0
Rz (γ) =  sin(γ) cos(γ) 0 . (7.61)
0 0 1

A general rotation of angle θ around the vector R~ can be written by the

product of three rotations around x, y and z with angles α, β, and γ

Rα,β,γ = Rz (γ)Ry (β)Rx (α). (7.62)

Defining a rotation in this way is known as an extrinsic Euler rotation (see

appendix G for more information). As rotations are non-commutative one
can make a different, but equally valid definition by changing the order
of the rotation matrices. As such many different definitions for the Euler
angles exist and care has to be taken which order is used for equations
found in the literature.
In order to study the conserved quantities related to rotations, we want
to express the rotation operator as a differential operator acting on the
state. For a small rotation with an angle dγ around the z axis, we can
expand the rotation operator in a Taylor series. Up to linear order in the
angle dγ, we find for the rotation of a point [x, y, z] in R3 around the z axis
the rotated coordinates:

Rz (dγ) [x, y, z] = [x cos(dγ) − y sin(dγ), x sin(dγ) + y cos(dγ), z] (7.63)

= [x − ydγ, y + xdγ, z]
= T x (−ydγ)T y (xdγ) [x, y, z] .

We expressed an infinitesimal small rotation around the z axis in terms of

translation operators. The differential form of the translation operator is
i
given in equation 7.22 T ~r0 = e− ~ ~p·~r0 . We can express the rotation operator
for an infinitesimal small rotation around the z axis thus as the differential
operator
i
Rz (dγ) = e− ~ (xpy −ypx )dγ (7.64)

201
Chapter 7. Symmetry and conservation laws

For finite angels γ = limN →∞ N dγ, with dγ = γ/N we have Rz (γ) =

Rz (dγ)N or
i
Rz (γ) = e− ~ (xpy −ypx )γ . (7.65)

By similar argumentation we find for the rotation around the x and y

direction
i
Rx (α) = e− ~ (ypz −zpy )α , (7.66)

and
i
Ry (β) = e− ~ (zpx −xpz )β . (7.67)

If we define the angular momentum operator as

~ = ~r × p
L ~ (7.68)
= [Lx , Ly , Lz ]
 
= ypz − zpy , zpx − xpz , xpy − ypx ,

we find for a rotation around the i direction by an angle θ the operator

i
Ri (θ) = e− ~ Li θ . (7.69)
i
By analogy to the translation operator T ~r0 = e− ~ ~p·~r0 we find that the
conserved quantity for a system with rotational symmetry is the angular
momentum. For a state |ψ(t)i that evolves according to an Hamiltonian
with spherical symmetry such that [Ri (θ), H] = 0 we have

∂ D i
E
ψ(t) e− ~ Li θ ψ(t) = 0. (7.70)
∂t
For such a system we can define a distribution function for the measurement
of the angular momentum ρ~l(t, ~l), which is time-independent.

202
 7.5. Rotational symmetry and ...

7.5.1 Commutation relations of the angular momentum op-

erators
In order to simplify the diagonalization of the Hamiltonian (For now we
have not assumed an explicit form, and only assumed it has spherical sym-
metry) we can look at the eigenstates of the rotation operators. As the rota-
tion operator commutes with the Hamiltonian we know that non-degenerate
eigenstates of the rotation operator are also eigenstates of the Hamiltonian.
The rotation operators are exponential functions of the angular momentum
operators. As the rotation operator commutes with H for any value of the
angle we find that non-degenerate eigenstates of the angular momentum
operator will also be eigenstates of the Hamiltonian. Eigenstates of the
operator Lx are in general no eigenstates of the operator Ly or Lz . There
must be degenerate eigenstates of the angular momentum operators. We
can check the previous claim by looking at the commutator of the angular
momentum operators. We find
[Lx , Ly ] = [ypz − zpy , zpx − xpz ] (7.71)
= [ypz , zpx ] − [zpy , zpx ] − [ypz , xpz ] + [zpy , xpz ]
= y[pz , z]px + x[z, pz ]py
~
= (ypx − xpy )
i
= i~Lz .
Similarly we find
[Ly , Lz ] = i~Lx (7.72)
[Lz , Lx ] = i~Ly ,
but
[H, Li ] = 0, ∀i ∈ {x, y, z}. (7.73)
Non-degenerate eigenstates of the operator Li are eigenstates of H, but
eigenstates of Lx are no eigenstates of Ly . There must be several degen-
erate eigenstates of the operator Li . We can search for a set of symmetry

203
Chapter 7. Symmetry and conservation laws

P
operators {O 1 , O 2 , . . .}, such that the operator O = i ci O i for some val-
ues ci is non-degenerate. For the first operator of this set we can take one
of the components of the operator L. ~ For a further operator we can search
for an operator that commutes both with L ~ and H. In order to find an
operator that commutes with the operator L ~ as well as with the Hamilto-
nian we can have a look at polynomial functions of L. ~ These polynomial
~
functions in L will, by construction, also commute with H. The simplest
symmetric combination we can make is L2 = L ~ · L.
~ For the commutator of
2
Lz with L we find

[Lz , L2 ] = [Lz , L2x + L2y + L2z ] (7.74)

= [Lz , L2x + L2y ]
= Lx [Lz , Lx ] + [Lz , Lx ]Lx + Ly [Lz , Ly ] + [Lz , Ly ]Ly
= i~(Lx Ly + Ly Lx − Ly Lx − Lx Ly )
= 0.

The same is true for the commutator between Lx and L2 or Ly and L2 .

With the use of the L2 operator, we can search for simultaneous eigen-
~ and L2 . If these eigenstates have a unique pair
states of one component of L
of eigenvalues (which they have) then these eigenstates will also be eigen-
states of the Hamiltonian. We thus can diagonalize the angular part of the
Hamiltonian, if we have spherical symmetry, without complete knowledge
of the exact form of the potential.

7.5.2 Raising and lowering operators for angular momen-

tum
In order to diagonalize Lz and L2 we will use similar techniques as we used
for the eigenstates of the Harmonic oscillator. We will define

L+ = Lx + iLy , (7.75)

and

L− = Lx − iLy , (7.76)

204
 7.5. Rotational symmetry and ...

~
as the raising and lowering operators. The angular momentum operator L
†
is Hermitian L~ = L.~ We thus find that

L†+ = L− . (7.77)

The reason why we call these the raising and lowering operators will become
clear later in this chapter. For the commutation relations we find

[L+ , L− ] = 2~Lz , (7.78)

[L+ , Lz ] = −~L+ , (7.79)

[L− , Lz ] = +~L− , (7.80)

and

[L± , L2 ] = 0. (7.81)

Since L†+ = L− we have

hψ|L+ L− |ψi = hL− ψ|L− ψi ≥ 0, (7.82)

and

hψ|L− L+ |ψi = hL+ ψ|L+ ψi ≥ 0. (7.83)

With the use of

L− L+ = (Lx − iLy )(Lx + iLy ) (7.84)

= L2x + i[Lx , Ly ] + L2y
= L2 − L2z − ~Lz ,

205
Chapter 7. Symmetry and conservation laws

and

L+ L− = L2 − L2z + ~Lz , (7.85)

we have

ψ L2 − L2z − ~Lz ψ ≥ 0 (7.86)

2
ψL − L2z + ~Lz ψ ≥ 0,

or

|hψ|~Lz |ψi| ≤ ψ L2 − L2z ψ . (7.87)

We will call the eigenstates of Lz the states |mi with eigenvalues

Lz |mi = m~ |mi . (7.88)

We can now have a look at the state L+ |mi and find

Lz L+ |mi = (L+ Lz + [Lz , L+ ]) |mi (7.89)

= (L+ m~ + ~L+ ) |mi
= (m + 1)~L+ |mi .

When |mi is an eigenfunction of Lz with eigenvalue m then L+ |mi is

an eigenfunction of Lz with eigenvalue m + 1, or L+ |mi = 0. The later
constraint is important as we know from equation 7.86 that there is a
maximum for the value of m. We can do the same with the operator L−
and find that L− |mi is an eigenfunction of Lz with eigenvalue m − 1, or
L− |mi = 0. We take l as the maximum value for |m|. We now can no
longer label our states by m alone. There can be different states with
different values of l and for each l we have a set of states with different m
always such that |m| ≤ l. From now on we thus label our states by the
quantum numbers l and m as |l, mi. For the operator L+ acting on a state
with m = l we have L+ |l, m = li = 0 and

L− L+ |l, m = li = 0 = (L2 − L2z − ~Lz ) |l, m = li (7.90)

206
 7.5. Rotational symmetry and ...

= (L2 − l2 ~2 − l~2 ) |l, m = li ,

or
L2 |l, m = li = l(l + 1)~2 |l, m = li . (7.91)
We find the same relation if we would act with the operator L+ L− on the
state |l, m = −li. The state |l, m = ±li is an eigenstate of the operator L2
with eigenvalue l(l + 1). We now can calculate
L2 (L+ )n |l, m = −li = (L+ )n L2 |l, m = −li (7.92)
= l(l + 1)~2 (L+ )n |l, m = −li ,
such that we find that |l, mi is an eigenstate of L2 for all values of |m| ≤ l
with eigenvalue l(l + 1) independent of m.
We now discuss which values of l we can obtain. With hl, m|l, mi = 1
such that our states are normalized we have
hlm|L− L+ |lmi = l, m L2 − L2z − ~Lz l, m (7.93)
hL+ lm|L+ lmi = (l(l + 1) − m2 − m)~2
= (l − m)(l + m + 1)~2
and
hlm|L+ L− |lmi = l, m L2 − L2z + ~Lz l, m (7.94)
2 2 2
hL− lm|L− lmi = (l(l + 1) − m + m)~ ~
= (l + m)(l − m + 1)~2
such that
√ √
L+ |l, mi = ~ l − m l + m + 1 |l, m + 1i (7.95)
√ √
L− |l, mi = ~ l + m l − m + 1 |l, m − 1i .
In oder for L+ |l, m = li = 0 and L− |l, m = −li = 0 we require that 2l ∈
N0 . Angular momentum is quantized by half integer values. We find a
multiplicity of the eigenstates of the total angular momentum operator L2
of 2l + 1.

207
Chapter 7. Symmetry and conservation laws

7.5.3 Angular momentum operators in spherical coordinates

In order to find explicit represen-
tations of the eigenfunctions of the z

angular momentum operators it is ᵠ

P(r,θ, )

useful to express these operators in r

spherical coordinates. We define, as θ

r cos θ
shown in figure 7.2 rs y

θ
in

ᵠ sin
ᵠ θ
x = r cos ϕ sin θ (7.96)

os
rc
r sin ᵠ sin θ
y = r sin ϕ sin θ x

z = r cos θ. Figure 7.2: Definition of the spherical

The Cartesian partial derivatives in coordinates used.
spherical coordinates are
∂ ∂r ∂ ∂θ ∂ ∂ϕ ∂
= + +
∂x ∂x ∂r ∂x ∂θ ∂x ∂ϕ
(7.97)
∂ 1 ∂ 1 sin(ϕ) ∂
= sin(θ) cos(ϕ) + cos(θ) cos(ϕ) −
∂r r ∂θ r sin(θ) ∂ϕ
∂ ∂r ∂ ∂θ ∂ ∂ϕ ∂
= + +
∂y ∂y ∂r ∂y ∂θ ∂y ∂ϕ
∂ 1 ∂ 1 cos(ϕ) ∂
= sin(θ) sin(ϕ) + cos(θ) sin(ϕ) +
∂r r ∂θ r sin(θ) ∂ϕ
∂ ∂r ∂ ∂θ ∂ ∂ϕ ∂
= + +
∂z ∂z ∂r ∂z ∂θ ∂z ∂ϕ
∂ 1 ∂
= cos(θ) − sin(θ)
∂r r ∂θ

such that
 
∂ ∂
Lz = −i~ x −y (7.98)
∂y ∂x

208
 7.5. Rotational symmetry and ...

∂
= −i~ ,
∂ϕ
and
    
∂ ∂ ∂ ∂
L± = ~ −i y −z ± z −x (7.99)
∂z ∂y ∂x ∂z
 
∂ cos θ ∂
= ~e±iϕ ± +i .
∂θ sin θ ∂ϕ

7.5.4 The eigenfunctions and eigenvalues of Lz

We will denote our eigenfunctions as Yl,m (θ, ϕ). Such that |Yl,m i = |l, mi
The eigenfunctions of Lz can be calculated relatively straight forward. We
have
∂
−i~ Yl,m (θ, ϕ) = m~Yl,m (θ, ϕ). (7.100)
∂ϕ

We write Yl,m (θ, ϕ) = P̃l,m (θ)Φm (ϕ) and find

r
1 imϕ
Φm (ϕ) = e . (7.101)
2π
The wave-function assigns to each point in space a complex number the
points ϕ + 2πj with j ∈ Z are equivalent. We thus require that

Φm (ϕ) = Φm (ϕ + 2πj) ⇐⇒ (7.102)

imϕ im(ϕ+2πj)
e =e ⇐⇒
ei2πjm = 1 ⇐⇒
m ∈ Z.

In equation 7.95 we derived that we needed 2m ∈ Z to find a natural

restriction of |m| ≤ l. We here find that only for m ∈ Z do we find
solutions for the wave function that assign a single complex value to each
point in space. There might be other solutions for the angular momentum

209
Chapter 7. Symmetry and conservation laws

state, with half-integer values of the angular momentum. These states are
however not representable by a wave function that assigns a single complex
number to each point in space. In section 8.6.1 we will encounter the half-
integer angular momentum to represent spin and the accompanying wave
function to be given by a function that assigns a complex vector to each
point in space.

7.5.5 The spherical harmonics

We now want to find explicit expressions for the θ dependence. We know
that L+ |l, m = li = 0 which yields the equation
  r
iϕ ∂ cos θ ∂ 1 ilϕ
~e +i P̃l,m=l (θ) e = (7.103)
∂θ sin θ ∂ϕ 2π
  r
i(l+1)ϕ ∂ cos θ 1
~e −l P̃l,m=l (θ) = 0.
∂θ sin θ 2π
The solutions to this differential equation are
s
(2l + 1)!
P̃l,m=l (θ) = (−1)l sinl (θ). (7.104)
(l!)2 2l+1
Similarly we find for m = −l the state
s
(2l + 1)!
P̃l,m=−l (θ) = sinl (θ). (7.105)
(l!)2 2l+1

We used the conventional (Condon Shortley) phases for the spherical har-
monics. The functions are normalized such that the integral over the norm
square of the wave function is one. We can find the states for general m by
applying the raising and lowering operators to the two states with extremal
m = ±l. The general solutions are known as the spherical harmonics. For
m ≥ 0 we have
r s
2l + 1 (l − m)!
Yl,m (θ, ϕ) = Pl,m (cos(θ))eimϕ , (7.106)
4π (l + m)!

210
 7.5. Rotational symmetry and ...

Figure 7.3: Pictures of the spherical harmonics for l = 0 (top row) to l = 3

bottom row, with m = −l to m = l from left to right. Shown is the surface
given by the equation r = Yl,m∗ (θ, ϕ)Y
l,m (θ, ϕ). The surface colour is given
by the argument of Yl,m (θ, ϕ) according to the colour scheme shown in
figure 7.4.

211
Chapter 7. Symmetry and conservation laws

with Pl,m (cos(θ)) the associated Legendre function

l   l+k−1 
m l m
X k! l
Pl,m (cos(θ)) = (−1) 2 sin(θ) 2 cos(θ)k−m
(k − m)! k l
k=m
(7.107)

For m < 0 we can use the relation

∗
Yl,−m (θ, ϕ) = (−1)m Yl,m (θ, ϕ). (7.108)

Figure 7.3 shows images of the spherical har-

monics for l = 0 to l = 3. In order to represent
these complex valued functions of two variables we
plot the surface defined by the equation
∗
r = Yl,m (θ, ϕ)Yl,m (θ, ϕ). (7.109)

The phase of the function is represented by the Figure 7.4: Colour

colour. We use the argument of the spherical har- coding used to plot
monic to define the hue of the surface as shown in complex functions.
figure 7.4. For m = 0 the spherical harmonics are The argument of
real valued functions with l zonal (at constant φ) the complex number
∗ (θ, ϕ)Y defines the hue of
nodes (Yl,m l,m (θ, ϕ) = 0). For m = ±l we
find a donut shaped surface with l repetitions of the surface. Real
the rotation of complex phase from 0 to 2π as a valued functions are
function of φ. For general values of m one finds red (positive) or cyan
|m| repetitions of the complex phase as a function (negative).
of φ and l − |m| zonal nodes.

7.5.6 Tesseral harmonics

Instead of simultaneously diagonalizing the operator Lz and L2 we could
have diagonalized the operators Lx and L2 . Actually any linear combina-
tion of Lx , Ly and Lz in combination with L2 works. There is nothing
special about the z direction or the Lz operator. For all possible choices

212
 7.5. Rotational symmetry and ...

Figure 7.5: Pictures of the tesseral harmonics for l = 0 (top row) to l = 3

bottom row, with m = −l to m = l from left to right. Shown is the surface
∗ (θ, ϕ)Z
given by the equation r = Zl,m l,m (θ, ϕ). The surface colour is given
by the argument of Zl,m (θ, ϕ) according to the colour scheme shown in
figure 7.4.

213
Chapter 7. Symmetry and conservation laws

of the quantization axis, one finds a set of functions that are linear com-
binations of the states we have found for the specific choice of quantizing
in the z direction, i.e. taking the Lz operator. The quantum number l
labels the total angular momentum. The total angular momentum l distin-
guishes states that belong to different irreducible representations. States
with a different total angular momentum l will generally have a different
energy. The quantum number m labels the degenerate states within this
representation. We can thus make linear combinations of states with the
same l but different m without loss of generality. The spherical harmonics
are complex functions We can make linear combinations of the complex
spherical harmonics that are real. These real functions are known as the
Tesseral harmonics
We define the tesseral Harmonics Zl,m (θ, ϕ) as:

Yl,m (θ, ϕ) m=0


1 m
Zl,m (θ, ϕ) = √2 (Yl,−m (θ, ϕ) + (−1) Yl,m (θ, ϕ)) m > 0 (7.110)

 √i (Yl,m (θ, ϕ) − (−1)m Yl,−m (θ, ϕ)) m < 0

2

such that all functions are real. We plot the tesseral harmonics for l ≤ 3 in
figure 7.5.

7.5.7 Addition of angular momentum

In the previous sections we showed that for a single particle rotational
symmetry leads to the conservation of angular momentum. In this section
we consider the case where we have two particles. We want to calculate
the total angular momentum of the two particles combined. In classical
physics the angular momentum of each particle is given by a vector ~li .
For two particles the total angular is just the vector sum of the angular
momentum of each particle ~l = ~l1 + ~l2 . In quantum mechanics the angular
momentum is given by the expectation value of the angular momentum
~ For two particles with coordinates ~r1 and ~r2 we can define the
operator L.
angular momentum operators
~ 1 = ~r1 × ~p1
L (7.111)

214
 7.5. Rotational symmetry and ...

~ 2 = ~r2 × ~p2 .
L

~ of particle 1 and 2
We will define the total angular momentum operator L
as the sum of the angular momentum operators of the individual particles.

~ =L
L ~1 +L
~ 2. (7.112)

In this section we will show that the operator L ~ represents the physical
sensible total angular momentum. We will furthermore show how to diag-
~ relate to the eigenstates
onalize this operator and how the eigenstates of L
~ 1 and L
of L ~ 2.

~ =L
Commutation relations of L ~1 +L
~2

~ =L
For operator L ~ 1 +L ~ 2 to be a useful definition of the total angular mo-
mentum it must fulfill the commutation relations of an angular momentum
~ 1 and L
operator. As L ~ 2 act on different particles with different coordinates
we have

[L1i , L2j ] = 0 ∀i, j ∈ [x, y, z]. (7.113)

The angular momentum operators acting on different particles commute.

~ =L
With the use of L ~1+L ~ 2 and the known commutation relations for L
~1
~
and L2 we can show that

[Lx , Ly ] = i~Lz (7.114)

[Ly , Lz ] = i~Lx
[Lz , Lx ] = i~Ly .

We furthermore find that all components of L ~ commute with L2 and we

can define the raising and lowering operators as

L+ = L + 1 + L+ 2 (7.115)
L− = L − 1 + L− 2 ,

215
Chapter 7. Symmetry and conservation laws

which have the same commutation relations as those for the angular mo-
mentum operators acting on a single particle. With these relations we can
derive all equations we found for the properties of the angular momentum
operators in section 7.5.2. In particularly we find that the operator L2 has
eigenstates with eigenvalue l(l + 1)~2 that are 2l + 1 fold degenerate. The
operator Lz has simultaneous eigenstates with the operator L2 . One can
use the eigenstates of Lz to label the degenerate states of the operator L2
uniquely from m = −l to m = l.

Angular momentum coupling and tensor product states

Lets consider a state |l1 m1 i that is an eigenstate of L21 and L1z . We can
furthermore consider a state |l2 m2 i that is an eigenstate of L22 and L2z .
These two states represent different particles. Even when the values of the
angular momenta are the same such that l1 = l2 and m1 = m2 these two
states are not the same. The first relates to particle 1 with coordinates
~r1 , the second to particle 2 with coordinates ~r2 . A quantum state which
represents both particles is given as a tensor product of the two states
|l1 m1 i ⊗ |l2 m2 i. For a given total angular momentum l1 of particle 1 the
basis of the Hilbert space for particle 1 is given as

B1 = {|l1 −l1 i , |l1 −l1 +1i , . . . , |l1 l1 i}. (7.116)

Similar we find for particle 2 with total angular momentum l2 the basis of
the Hilbert space to be

B2 = {|l2 −l2 i , |l2 −l2 +1i , . . . , |l2 l2 i}. (7.117)

The basis of the Hilbert space for all product states is given by the tensor
product of B1 and B2

B = B1 ⊗ B2 (7.118)
= {|l1 −l1 l2 −l2 i , . . . |l1 l1 l2 −l2 i , |l1 −l1 l2 −l2 +1i , . . . ,
|l1 l1 l2 −l2 +1i , . . . , |l1 l1 l2 l2 i}.

216
 7.5. Rotational symmetry and ...

Note that we will use all three of the notations |l1 m1 i⊗|l2 m2 i = |l1 m1 l2 ml2 i =
|l1 m1 i |l2 m2 i for product states, depending on the situation at hand. For a
given l1 there are 2l1 +1 states |l1 m1 i with different m1 possible. Similar for
a given l2 there are 2l2 + 1 states |l2 m2 i with different m2 possible. The re-
sult is that for a given l1 and l2 there are (2l1 + 1)(2l2 + 1) states |l1 m1 l2 m2 i
with different m1 and m2 possible. The dimension of the Hilbert spaces are
2l1 + 1 and 2l2 + 1 for particle 1 and 2 respectively and (2l1 + 1)(2l2 + 1)
for the tensor product space of the two particles simultaneously.
Instead of labeling the states as product states |l1 m1 i ⊗ |l2 m2 i we can
arrange the states in the product basis according to their total angular
momentum l and m into the states |lmi. The states |lmi are eigenvectors
of L2 and Lz . We now want to find the allowed values of l and m for the
states |lmi such that we generate a basis for the product Hilbert space.
We immediately know that we need to find (2l1 + 1)(2l2 + 1) of such states
|lmi as that is the dimension of the product Hilbert space. Classically we
would expect that l relates to the length of the angular momentum vector.
If classically ~l1 and ~l2 are parallel we find that |~l| = |~l1 | + |~l2 |, whereas when
~l1 and ~l2 are anti-parallel we find that |~l| = |~l1 | − |~l2 |. We thus could expect
that quantum mechanically the total angular momentum l is given by
l ∈ [|l1 − l2 |, l1 + l2 ] . (7.119)
In order to prove that this is indeed the case for the quantum mechanical
description of a system with spherical symmetry, we use the fact if a state
|lmi with total angular momentum l is in the basis of the Hilbert space,
then all 2l + 1 states with m ∈ [−l, l] must be in the Hilbert space. We
will furthermore count the number of states with total angular momentum
m = m1 + m2 in the basis B. We know that
Lz |l1 m1 l2 m2 i = (L1z + L2z ) |l1 m1 l2 m2 i (7.120)
= (m1 + m2 ) |l1 m1 l2 m2 i
= m |l1 m1 l2 m2 i ,
with m = m1 + m2 . As such we can count the number of states |l1 m1 l2 ml2 i
with a given eigenvalue m = m1 + m2 of operator Lz in the basis B and
know the number of states |lmi with a given value of m.

217
Chapter 7. Symmetry and conservation laws

The maximum value of m for a state in basis B is equal to l1 + l2 .

There is one state with m = l1 + l2 , which is the state |l1 l1 l2 l2 i. There
are two states with m = m1 + m2 − 1, namely the states |l1 l1 −1l2 l2 i and
|l1 l1 l2 l2 −1i. In general, for n ≤ 2 min(l1 , l2 ) there are n + 1 states with
angular momentum m = l1 + l2 − n. For 2 min(l1 , l2 ) < n ≤ l1 + l2 the
number of states with m = l1 + l2 − n is 2 min(l1 , l2 ) + 1, independent of
m. In the second column of table 7.1 we list the possible states for different
values of m. In the table we assumed l1 > l2 and only list the states for
m ≥ 0.

Table 7.1: Product basis of two states |l1 m1 i and |l2 m2 i for fixed value of
l1 and l2 (l2 < l1 ) sorted according to m = m1 + m2 . Either given as states
on the product basis |l1 m1 l2 m2 i = |m1 m2 i or on the basis listing total
angular momentum |lmi = |li. The value of n = 2 min(l1 , l2 ) labels the
border between different behaviour of the number of states as a function of
m. For m < 0 the table is similar as the table for m > 0.

m Basis |l1 m1 l2 m2 i = |m1 m2 i Basis |lmi = |li

l1 + l2 {|l1 l2 i} {|l1 +l2 i}
l1 + l2 − 1 {|l1 −1l2 i , |l1 l2 −1i} {|l1 +l2 i , |l1 +l2 −1i}
l1 + l2 − 2 {|l1 −2l2 i , |l1 −1l2 −1i , {|l1 +l2 i , |l1 +l2 −1i ,
|l1 l2 −2i} |l1 +l2 −2i}
.. .. ..
. . .
l1 + l2 − n {|l1 −nl2 i , . . . , |l1 l2 −ni} {|l1 +l2 i , . . . , |l1 +l2 −ni}
l1 − l2 − 1 {|l1 −n−1l2 i , . . . , |l1 − 1l2 −ni} {|l1 +l2 i , . . . , |l1 +l2 −ni}
.. .. ..
. . .
0 {|−l2 l2 i , . . . , |l2 −l2 i} {|l1 +l2 i , . . . , |l1 +l2 −ni}

In the third column of table 7.1 we show the possible states |lmi for
different values of l at a fixed value of m. The maximum value for m is
l1 + l2 . The maximum value of l is thus l1 + l2 . There are two states with
m = l1 + l2 − 1 in basis B. There are thus two eigenstates |lmi of L2 and

218
 7.5. Rotational symmetry and ...

Lz with expectation value m = l1 + l2 − 1. These must be the state with

l = l1 + l2 and the state with l = l1 + l2 − 1. There are n + 1 states with
m = l1 + l2 − n. These n + 1 states |l1 l1 −n1 l2 l2 −n2 i with n1 + n2 = n
are linear combinations of the states |lmi with l ∈ [l1 + l2 − n, l1 + l2 ] and
m = l1 + l2 − n. Whereby we require that n ≤ 2 min(l1 , l2 ). We thus find
that the values of l are given by l ∈ [|l1 − l2 |, l1 + l2 ], in agreement with the
classical result.
As a sanity check we can consider the dimension of the Hilbert space.
The basis is given by all states |lmi with l ∈ [|l1 −l2 |, l1 +l2 ] and m ∈ [−l, l].
The dimension of the Hilbert space with basis B becomes

lX
1 +l2

Dim(B) = 2l + 1 (7.121)
l=|l1 −l2 |

= (2l1 + 1)(2l2 + 1),

in agreement with the dimension found for the tensor product basis of B1
and B2 .

Unitary basis transformation between |lmi and |l1 m1 l2 m2 i

For a fixed l1 and l2 we can generate (2l1 + 1)(2l2 + 1) product states

|l1 m1 l2 m2 i, with −l1 ≤ m1 ≤ l1 and −l2 ≤ m2 ≤ l2 that form the basis
of the Hilbert space of these two particles. We can equally well write the
basis as the eigenstates of L2 and Lz as |lmi with |l1 − l2 | ≤ l ≤ l1 + l2
and for each l the restriction on m given by −l ≤ m ≤ l. The states |lmi
and |l1 m1 l2 m2 i are a different basis for the same Hilbert space. As such
there exists a unitary transformation that transforms the states |lmi into
the states |l1 m1 l2 m2 i and vice versa.

l1
X l2
X
|lmi = Cllm
1 m1 l2 m2
|l1 m1 l2 m2 i , (7.122)
m1 =−l1 m2 =−l2

219
Chapter 7. Symmetry and conservation laws

and
lX
1 +l2 l
X
|l1 m1 l2 m2 i = Cllm
1 m1 l2 m2
|lmi , (7.123)
l=|l1 −l2 | m=−l

with Cllm1 m1 l2 m2
real numbers known as the Clebsch-Gordan coefficients.
Note that in principle the basis transformation could be complex. As we
will show below it turns out that the Clebsch-Gordan coefficients are real
for the coupling of angular momentum. If we multiply equation 7.122 from
the left with hl1 m1 l2 m2 | we find

Cllm
1 m1 l2 m2
= hl1 m1 l2 m2 |lmi . (7.124)

The Clebsch-Gordan coefficients tell you how the states |l1 m1 i and |l2 m2 i
couple to form the state |lmi, an eigenstate of the total angular momentum
operators L2 and Lz .

Properties of the Clebsch-Gordan coefficients

Before we will derive a closed expression for the Clebsch-Gordan coefficients
we can already find a few useful properties of these numbers.
The state |l1 m1 l2 m2 i is an eigenfunction of L1z with eigenvalue m1 , of
L2z with eigenvalue m2 and of Lz = L1z +L2z with eigenvalue m1 +m2 . As
such the states |lmi projected on the product states |l1 m1 l2 m2 i has only a
non-zero overlap if m = m1 + m2 . For the Clebsch-Gordan coefficients we
find

Cllm
1 m1 l2 m2
=0 if m 6= m1 + m2 . (7.125)

As we derived above based on the necessity to have 2l + 1 states with m

ranging from −l to l for each l and the multiplicity of the eigenstates of Lz
we find that

Cllm
1 m1 l2 m2
=0 if l ∈
/ [|l1 − l2 |, l1 + l2 ]. (7.126)

220
 7.5. Rotational symmetry and ...

These conditions on m and l for which the Clebsch-Gordan coefficients are

zero are known as the triangular relations.
The basis states |lmi are orthonormal such that
δm,m0 = lm lm0 (7.127)
l1
X l2
X
= hlm|l1 m1 l2 m2 i l1 m1 l2 m2 lm0
m1 =−l1 m2 =−l2
l1 l2
0
X X
= Cllm C lm
1 m1 l2 m2 l1 m1 l2 m2
.
m1 =−l1 m2 =−l2

Similarly we find using the orthonormality of the basis states |l1 m1 l2 m2 i

δm1 ,m01 δm2 ,m02 = l1 m1 l2 m2 l1 m01 l2 m02 (7.128)
lX
1 +l2 l
X
= hl1 m1 l2 m2 |lmi lm l1 m01 l2 m02
l=|l1 −l2 | m=−l
lX
1 +l2 l
X
= Cllm C lm 0 0 .
1 m1 l2 m2 l1 m l2 m
1 2
l=|l1 −l2 | m=−l

The phase of the Clebsch-Gordan coefficients is fundamentally unde-

fined. The state rays |ai = |lmi and |bi = eiθ |lmi with θ ∈ R correspond
to the same physical normalized state. The Clebsch-Gordan coefficients for
state |ai and |bi will differ by a complex phase eiθ . We can use the phase
ambiguity of the states to make all Clebsch-Gordan coefficients real. That
leaves the freedom of choosing the sign of the states. We can fix some of
the signs with the use of the ladder operators L+ and L− . If we take the
standard definition of the ladder operators
√ √
L− |lmi = ~ l + m l − m + 1 |lm − 1i , (7.129)
we fix the phase between states with the same l, but different m. In or-
der to fix the states with different l we choose the Condon-Shortley phase
convention
Clll1 l1 l2 l−l1 = hl1 l1 l2 l−l1 |lli > 0. (7.130)

221
Chapter 7. Symmetry and conservation laws

Derivation of the Clebsch-Gordan coefficients using ladder oper-

ators and orthogonality
There are several ways to derive the Clebsch-Gordan coefficients. We will
present two methods within these lecture notes. The first, based on ladder
operators and orthogonality is physically insightful. This is the derivation
we make in this section. At the end of the derivation, we will encounter
a closed form that needs severe algebraic simplification to get to the final
result. We can simplify the derivation with the use of projector methods
introduced by Löwdin [71, 72]. This is the path we will follow in the next
section.
Because of the ladder operators L± and orthonormality conditions of
the angular momentum states |lmi, |l1 m1 i, and |l2 m2 i we can derive all
Clebsch-Gordan coefficients once a single one is known. For the case where
l = l1 + l2 and m = m1 + m2 the Clebsch-Gordan coefficients are trivial as
there is only a single state with m = m1 + m2

|l1 +l2 l1 +l2 i = |l1 l1 i ⊗ |l2 l2 i , (7.131)

and

Cll11+l 2 l1+l2
l1 l2 l2 = 1. (7.132)

We now need to find an expression that relates the general Clebsch-Gordan

coefficients Cllm
1 m1 l1 m2
to Cll11+l 2 l1+l2
l1 l2 l2 .
We start by expressing |lmi in terms of |lli. With the use of the ladder
operator L− we find

|lli
 m=l
1 √ 1 −
|lmi = √
l+m+1 l−m ~
L |lm+1i −l ≤ m < l (7.133)

0 |m| > l

 
l−m  l−m
Y 1 1
= √ √  L− |lli .
p=1
l+m+p l−m−p+1 ~

222
 7.5. Rotational symmetry and ...

Next we express the state |lli in terms of the product states |l1 m1 l2 m2 i
and Clebsch-Gordan coefficients
min(l1 ,l+l2 )
X
|lli = Clll1 ql2 l−q |l1 ql2 l−qi (7.134)
q=max(−l1 ,l−l2 )

Combining equation 7.133 and 7.134 yields for the state |lmi the ex-
pression
 l−m
1
|lmi ∝ L− |lli (7.135)
~
 l−m
1
= (L1− + L2− ) |lli
~
l−m
X l − m  1 r  l−m−r
1
= L1− L2− |lli
r ~ ~
r=0
l−m
X X l − m  r  l−m−r
ll 1 1
= Cl1 ql2 l−q L1− L2− |l1 qi |l2 l−qi
q
r ~ ~
r=0
l−m min(l1 ,l+l2 )  
X X l−m
= Clll1 ql2 l−q
r
r=0 q=max(−l1 ,l−l2 )
 
r−1
Yp p
× l1 + q − p l1 − q + p + 1 |l1 q − ri
p=0
 
l−m−r−1
Y p p
× l2 + l − q − p l2 − l + q + p + 1 |l2 m + r − qi .
p=0

We now multiply equation 7.135 from the left with hl1 m1 l2 m2 | and find
with q − r = m1 the expression
 
l1 l−m
X Y 1
Cllm
1 m1 l2 m2
=δm1 +m2 ,m  √ √  (7.136)
q=m p=1
l + m + p l − m − p + 1
1

223
Chapter 7. Symmetry and conservation laws

 
Y1 −1
q−m
p p
× l1 + q − p l1 − q + p + 1
p=0
 
Y 1 −1
l−m−q+m
p p
× l2 + l − q − p l2 − l + q + p + 1 
p=0
  
l−m ll
× C .
q − m1 l1 ql2 l−q

If we know the values for Clll1 ql2 l−q we found a closed expression for all
Clebsch-Gordan coefficients.
As a next step we express Clll1 ql2 l−q in terms of Clll1 l1 l2 l−l1 with the use of
the recurrence relations between Clebsch-Gordan coefficients. Acting with
the ladder operator L± on the state |lmi yields
√ √
L± |lmi = ~ l ∓ m l ± m + 1 |lm ± 1i , (7.137)

or on the coupled basis

X
0 0
L± |lmi = (L1± + L2± ) Cllm 0
1 m l2 m
0 l1 m1 l2 m2 (7.138)
1 2
m01 m02
X q q
= ~ l1 ∓ m1 l1 ± m01 + 1 l1 m01 ±1l2 m02 +
0

m01 m02
q q 
0 0
l2 ∓ m2 l2 ± m2 + 1 l1 m1 l2 m2 ±1 Cllm
0 0
0
1 m l2 m
0 .
1 2

We now multiply both sides with hl1 m1 l2 m2 | from the left and find
√ √
l ∓ m l ± m + 1 hl1 m1 l2 m2 |lm ± 1i = (7.139)
p p p p
l1 ∓ m1 + 1 l1 ± m1 Cllm
1 m1∓1l2 m2
+ l2 ∓ m2 + 1 l2 ± m2 Cllm1 m1 l2 m2∓1
.

With hl1 m1 l2 m2 |lm±1i = Cllm±1

1 m1 l2 m2
this yields a recurrence relation be-
tween Clebsch-Gordan coefficients with different values of m but the same
l.

224
 7.5. Rotational symmetry and ...

If we take m = l in equation 7.139 we find

√ √
ll l2 − m2 + 1 l2 + m2 ll
Cl1 m1−1l2 m2 = − √ √ C . (7.140)
l1 − m1 + 1 l1 + m1 l1 m1 l2 m2−1
We now take m2 = l − m1 + 1 and replace m1 by l1 − r to find
√ √
ll l2 − l + l1 − r l + l2 − l1 + r + 1 ll
Cl1 l1−r−1l2 l−l1+r+1 = − √ √ Cl1 l1−rl2 l−l1+r .
r + 1 2l1 − r
(7.141)

Repeating equation 7.141 l1 − q times yields

l1 −q−1 √ √
Y  l2 − l + l1 − r l + l2 − l1 + r + 1

l1 −q
Clll1 ql2 l−q = (−1) √ √ Clll1 l1 l2 l−l1 .
r=0
r + 1 2l1 − r
(7.142)

Lastly we need to calculate Clll1 l1 l2 l−l1 . In order to do this we use that

the state
l1
X
|lli = Clll1 m1 l2 l−m1 |l1 m1 l2 l−m1 i (7.143)
m1 =−l1

is normalized. We find for the norm

X
hll|lli = Clll1 m0 l2 l−m0 Clll1 m1 l2 l−m1 l1 m01 l2 l−m01 l1 m1 l2 l−m1 (7.144)
1 1
m1 ,m01
l1
X  2
= Clll1 m1 l2 l−m1 .
m1 =−l1

If we now insert equation 7.142 we find

l1 l1 −q−1 √ √ 2
X Y  l2 − l + l1 − r l + l2 − l1 + r + 1  ll 2
√ √ Cl1 l1 l2 l−l1 = 1.
q=−l1 r=0
r + 1 2l1 − r
(7.145)

225
Chapter 7. Symmetry and conservation laws

With the use of the Condon-Shortley phase definition we have

u l1 l1 −q−1  √ √ 2 −1
v 
uX Y l2 − l + l1 − r l + l2 − l1 + r + 1 
Clll1 l1 l2 l−l1 = t √ √ .
q=−l r=0
r + 1 2l1 − r
1

(7.146)
We now combine equation 7.136, 7.142 and 7.146 to come to the final
result
 
l1 l−m
X Y 1
Cllm
1 m1 l2 m2
= δm1 +m2 ,m  √ √  (7.147)
q=−m1 p=1
l + m + p l − m − p + 1
 
Y1 −1 p
q−m
p
× l1 + q − p l1 − q + p + 1
p=0
 
Y 1 −1
l−m−q+m
p p
× l2 + l − q − p l2 − l + q + p + 1
p=0
 
l−m
×
q − m1
l1 −q−1 √ √
Y  l2 − l + l1 − r l + l2 − l1 + r + 1

l1 −q
× (−1) √ √
r=0
r + 1 2l1 − r

√ √ 2 −1/2
 0

l1 l1 −q
Y−1 
X l2 − l + l1 − r l + l2 − l1 + r + 1 
× √ √ .

q 0 =−l1 r=0
r + 1 2l1 − r

This result can, with the aid of some algebra, be rewritten. The stan-
dard form in the literature one commonly finds is
s
(l1 + l2 − l)! (l1 − l2 + l)! (−l1 + l2 + l)!
Cllm
1 m1 l2 m2
= δm1 +m2 ,m (7.148)
(l1 + l2 + l + 1)!
X √
× (−1)t t! (l1 +l(2l+1) (l1 +m1 )! (l1 −m1 )! (l2 +m2 )! (l2 −m2 )! (l+m)! (l−m)!
2 −l−t)! (l1 −m1 −t)! (l2 +m2 −t)! (l−l2 +m1 +t)! (l−l1 −m2 +t)!
.
t

226
 7.5. Rotational symmetry and ...

Whereby the sum over t includes all integers for which the factorials are
not taken of negative numbers.

Derivation of the Clebsch-Gordan coefficients using projection

operators
In the previous section we derived the Clebsch-Gordan coefficients with
the use of properties of the angular momentum operators. Although the
derivation gives great physical insight into to coupling of angular momen-
tum, the derivation needs several steps that are not obvious at first sight.
One derivation, that goes back to Löwdin [71, 72], allows one to get a closed
expression for the Clebsch-Gordan coefficients within a few steps. It then
still takes some work to bring these coefficients into an algebraic beautiful
form.
The Clebsch-Gordan coefficients allow one to express a state with two
particles with angular momentum l1 and l2 and projection on the z direction
m1 and m2 into states with total angular momentum l and projection on
the z axis m.
lX
1 +l2 l
X
|l1 m1 l2 m2 i = Cllm
1 m1 l2 m2
|lmi . (7.149)
l=|l1 −l2 | m=−l

If one defines a projector operator 6.4.8 that projects states from the full
Hilbert space for two electrons with angular momentum l1 and l2 onto the
subspace of eigenvectors of the operator L2 /~2 and Lz /~ with eigenvalues
l(l + 1) and m we obtain an expression containing only a single Clebsch-
Gordan coefficient.
lX 0
1 +l2 l
0 0
X
0 0
P lm |l1 m1 l2 m2 i = Cll1m
m1 l2 m2 P lm l m (7.150)
l0 =|l1 −l2 | m0 =−l0

= Cllm
1 m1 l2 m2
|lmi .

As Cllm
1 m1 l2 m2
= 0 when m 6= m1 + m2 we do not really need a projection
onto the eigenstates of Lz /~ but can use that a projection on the eigenstates

227
Chapter 7. Symmetry and conservation laws

of L2 /~2 with eigenvalues l(l + 1) yields

lX 0
1 +l2 l
0 0
X
0 0
P l |l1 m1 l2 m2 i = Cll1m
m1 l2 m2 Pl l m (7.151)
l0 =|l 1 −l2 | m0 =−l0

= δm,m1 +m2 Cllm

1 m1 l2 m2
|lmi .

This expression still contains the unknown state |lmi. In order to re-
move this we take the inner product of two coupled state projected to the
same total angular momentum l
D E
l1 m01 l2 m02 P †l P l l1 m1 l2 m2 = δm0 ,m01 +m02 δm,m1 +m2 (7.152)
0
× Cllm 0
1 m l2 m
lm
0 Cl m l m
1 1 2 2
lm0 lm
1 2

= δm,m01 +m02 δm,m1 +m2 Cllm 0

1 m l2 m
lm
0 Cl m l m .
1 1 2 2
1 2

For m01 = m1 and m02 = m2 we obtain the square of the Clebsch-Gordan

coefficients. Actually, as we did not define the phase of the state |lmi
but only required it to be an eigenstate of the operators L2 and Lz this
is the only well-defined evaluation of the Clebsch-Gordan coefficients we
can make. If we take m01 6= m1 and m02 6= m2 we can obtain the relative
phase between different Clebsch-Gordan coefficients for the elements with
the same l. For elements with different l we have the freedom to choose
the phase or sign of the Clebsch-Gordan coefficients. Equation 7.152 thus
contains a full closed expression of all properties of the Clebsch-Gordan
coefficients. D E
What is left is to evaluate the expectation value of l1 m01 l2 m02 P †l P l l1 m1 l2 m2 .
Operator P l is ( see equation 6.117 )

+l2  ~12 L2 −l0 (l0 +1)
l1Y
l(l+1)−l0 (l0 +1) l 6= l0
Pl = (7.153)
0
1
l =|l1 −l2 | l = l0

We note that the operator P l is Hermitian such that P †l = P l and that

228
 7.5. Rotational symmetry and ...

P 2l = P l by construction such that we need to evaluate

0 0
δm,m01 +m02 δm,m1 +m2 Cllm 0
1 m l2 m
lm
0 Cl m l m = l1 m1 l2 m2 P l l1 m1 l2 m2 .
1 1 2 2
1 2
(7.154)

The further algebraic simplification of this expression is lengthy. We do

not include it here but refer to the literature [71] and [72]. The final result
for the Clebsch-Gordan coefficients under the assumption of the Condon-
Shortley phase is given in equation 7.148.

7.5.8 Rotation of total angular momentum states, Wigner

D matrices
The total angular momentum states |lmi are eigenstates of the operators
L2 and Lz . As such we know that if we rotate the state |lmi around the z
axis the phase of the state can change, but the state itself should stay the
i
same. With Rz (θ) = e− ~ Lz θ (see equation 7.69) we indeed find
i
Rz (θ) |lmi = e− ~ Lz θ |lmi (7.155)
= e−imθ |lmi .

For the rotation around the x or y direction the relation is not so simple.
The commutator between Lx or Ly and Lz is non-zero. Therefore also the
commutators

[Lz , Rx (θ)] 6= 0, (7.156)

[Lz , Ry (θ)] 6= 0,

are nonzero. A rotation around the x or y axis thus leads to a mixing of

states with different m.
i
Ry (θ) |lmi = e− ~ Ly θ |lmi (7.157)
1
(L− −L+ )θ
=e 2~ |lmi
X
= dlm0 m (θ) lm0 .
m0

229
Chapter 7. Symmetry and conservation laws

We first of all note that a rotation of an angular momentum state does not
change the total angular momentum l. This is a direct result from the fact
that all components of L ~ commute with L2 . The numbers dl 0 define, for
mm
each l a square matrix d of dimension 2l + 1. The matrix d is known as a
Wigner d matrix. The elements of the Wigner d matrix are given as
D 1
E
dlm0 m (θ) = lm0 e 2~ (L− −L+ )θ lm (7.158)

In the previous example we rotated around the y axis. We are interested

in general rotations. For this we will use Euler angles and rotations (see
appendix G for more info). We have seen before that rotations around the
z axis are particularly simple and rotations around the y axis yield a real
rotation operator. We thus choose intrinsic Euler angles around z, y, and
z to represent the rotation matrix
i i i
R(α, β, γ) = e− ~ Lz α e− ~ Ly β e− ~ Lz γ . (7.159)

We can relate the operator R(α, β, γ) to the Wigner D matrix, Dl (α, β, γ),
with matrix elements
l 0
Dm 0 m (α, β, γ) = lm R(α, β, γ) lm . (7.160)

We note that
l l
Dm 0 m (0, β, 0) = dm0 m (β), (7.161)

for a rotation around the y axis. Furthermore for a rotation around the z
axis we have
l −im α0
Dm 0 m (α, 0, 0) = e δm,m0 , (7.162)

and
l −imγ
Dm 0 m (0, 0, γ) = e δm,m0 , (7.163)

such that
0
−im α l
l
Dm 0 m (α, β, γ) = e dm0 m (β)e−imγ . (7.164)

230
 7.5. Rotational symmetry and ...

Unitarity
The Wigner D matrix defines a unitary transformation between the states
|lmi and the rotated states |lm0 i. As such that matrix Dl (α, β, γ) is unitary

Dl (α, β, γ)Dl (α, β, γ)† = 1. (7.165)

For the matrix elements of Dl (α, β, γ) this yields

l
X
l
Dm 1m
l
(α, β, γ)Dm 2m
(α, β, γ)∗ = δm1 ,m2 . (7.166)
m=−l

For a rotation and inverse rotation acting on the state |lmi we find
X
l 0
R(α, β, γ) |lmi = Dm 0 m (α, β, γ) lm , (7.167)
m0
X
R(α, β, γ)−1 |lmi = l
Dmm 0 (α, β, γ)
∗
lm0 .
m0

Relation to spherical harmonics

The matrix elements of the Wigner D matrices can be related to the spher-
ical harmonics. We know that states with angular momentum l and m can
be represented by spherical harmonics. For a single particle we have

h~r|lmi = f (~r)Yl,m (θ, φ). (7.168)

We thus find that

X
l
R(α, β, γ)Yl,m (θ, φ) = Dm 0 m (α, β, γ)Yl,m0 (θ, φ). (7.169)
m0

At the same time we know that R(α, β, γ) represents a rotation such that

R(α, β, γ)Yl,m (θ, φ) = Yl,m (θ0 , φ0 ), (7.170)

231
Chapter 7. Symmetry and conservation laws

with θ0 and φ0 the rotated coordinates. In particular we have

R−1 (φ, θ, 0)Yl,m (0, 0) = Yl,m (θ, φ). (7.171)

We thus find

Yl,m (θ, φ) = R−1 (φ, θ, 0)Yl,m (0, 0) (7.172)

X
l ∗
= Dmm 0 (φ, θ, 0) Yl,m0 (0, 0).

m0
q
2l+1
We now realize that Yl,m (0, 0) = δm,0 4π and find
r
2l + 1
Yl,m (θ, φ) = l
Dm0 (φ, θ, γ)∗ , (7.173)
4π
or
r
4π
l
Dm0 (α, β, γ)∗ = Yl,m (β, α). (7.174)
2l + 1

Clebsch-Gordan series
One property of the Wigner D matrix that we will need in the next section
l1 l2
is the Kronecker product of two matrices Dm 0 m (α, β, γ)Dm0 m (α, β, γ). In
1 1 2 2
order to derive this product we start from the expansion of a product state
|l1 m1 i |l1 m1 i into states of angular momentum |lmi
X
|l1 m1 i |l2 m2 i = Cllm
1 m1 l2 m2
|lmi . (7.175)
lm

Next we rotate the states, with

X
l 0
R(α, β, γ) |lmi = Dm 0 m (α, β, γ) lm , (7.176)
m0
X
l1 0
R(α, β, γ) |l1 m1 i = Dm 0 m (α, β, γ) l1 m1 ,
1
1
m01

232
 7.5. Rotational symmetry and ...

X
l2 0
R(α, β, γ) |l2 m2 i = Dm 0 m (α, β, γ) l2 m2 .
2
2
m02

We find
X
0
Dml1 l2
0 m (α, β, γ)Dm0 m (α, β, γ) l1 m1
1 2
l2 m02 = (7.177)
1 2
m01 m02
XX
0
Cllm
1 m1 l2 m2
l
Dm 0 m (α, β, γ) lm .
lm m0

We now multiply from the left with hl2 m02 | hl1 m01 | to get
l1 l2
Dm 0 m (α, β, γ)Dm0 m (α, β, γ) = (7.178)
1 1 2 2
0
XX
Cllm C lm 0 0 Dm
1 m1 l2 m2 l1 m l2 m
l
0 m (α, β, γ).
1 2
lm m0

7.5.9 Product of two spherical harmonic functions

One quantity that one needs to calculate often for problems with spherical
symmetry is the product of two spherical harmonics. With the use of
equation 7.178 we can write the product of two spherical harmonic functions
as a sum over single spherical harmonic functions with different values of
l. If we insert equation 7.174 into equation 7.178 we find, with m01 = 0,
m02 = 0, and m0 = 0
r r
4π 4π
Yl ,m (θ, φ) Yl ,m (θ, φ) = (7.179)
2l1 + 1 1 1 2l2 + 1 2 2
r
X 4π
Cllm C l0
1 m1 l2 m2 l1 0l2 0
Yl,m (θ, φ).
2l + 1
lm
Reordering terms yields
Yl1 ,m1 (θ, φ)Yl2 ,m2 (θ, φ) = (7.180)
lX
1 +l2
r r
1 (2l1 + 1)(2l2 + 1)
Cllm C l0
1 m1 l2 m2 l1 0l2 0
Yl,m (θ, φ),
4π 2l + 1
l=|l1 −l2 |

with m = m1 + m2 and |l1 − l2 | ≤ l ≤ l1 + l2 .

233
Chapter 7. Symmetry and conservation laws

7.5.10 Expectation values between angular momentum states

One of the things we will need to calculate are integrals over products of
spherical harmonics. With the previous considerations we are now in a
situation where we can find analytical expressions for these. The spherical
harmonics are eigenstates of the rotation operators. The only function that
can have a non-zero integral is the function that has the eigenvalue 1 for
all symmetry operations, i.e. rotations. We thus immediately know that
Z 2π Z π
√
Yl,m (θ, φ) sin(θ)dθdφ = δm,0 δl,0 2 π. (7.181)
0 0

The integrals over two spherical harmonics yield the orthogonality con-
ditions between the spherical harmonics
Z 2π Z π
Yl∗0 ,m0 (θ, φ)Yl,m (θ, φ) sin(θ)dθdφ = δm,m0 δl,l0 . (7.182)
0 0

The integrals over three spherical harmonics appear when we calculate

expectation values of potentials, interactions or other operators. For the
integral over three spherical harmonics we can use equation 7.180 to reduce
the problem to an integral over two spherical harmonics. For this we then
use the orthogonality relations. We find
Z 2π Z π
∗
Yl,m (θ, φ)Yl1 ,m1 (θ, φ)Yl2 ,m2 (θ, φ) sin(θ)dθdφ (7.183)
0 0
s
X (2l1 + 1)(2l2 + 1) 0 0
l0 0
= 0
Cll1m
m1 l2 m2 Cl1 0l2 0
4π(2l + 1)
l 0 m0
Z 2π Z π
∗
× Yl,m (θ, φ)Yl0 ,m0 (θ, φ) sin(θ)dθdφ
0 0
s
(2l1 + 1)(2l2 + 1) lm
= Cl1 m1 l2 m2 Cll01 0l2 0 .
4π(2l + 1)

The latter expression is useful when we need to calculate expectation

values of states expressed on a basis of spherical harmonics. Given a basis

234
 7.5. Rotational symmetry and ...

state |nlmi represented by the wave function ψn,l,m (~r) = Rnl (r)Yl,m (θ, φ),
with Rnl (r) some radial function depending on the quantum numbers nl
and Yl,m (θ, φ) describing the angular dependence. We will often need to
evaluate integrals of the form

o = nlm O n0 l0 m0 . (7.184)

If we can write the operator O as

X
O(~r) = O k,q (r)Yk,q (θ, φ), (7.185)
k,q

we find
XZ ∞
0 0 0 ∗
nlm O n l m = Rnl (r)O k,q (r)Rn0 l0 (r)dr (7.186)
k,q 0
Z 2π Z π
∗
× Yl,m (θ, φ)Yk,q (θ, φ)Yl0 ,m0 (θ, φ) sin(θ)dθdφ
0 0
s
XZ ∞ (2k + 1)(2l0 + 1) lm
∗
= Rnl (r)O k,q (r)Rn∗ 0 l0 (r)dr l0
Ckql0 m0 Ck0l 00.
0 4π(2l + 1)
k,q

We can reduce the 3 dimensional integral hnlm|O|n0 l0 m0 i to a one dimen-

sional integral over r. The angular dependence is captured by the analytical
expression in terms of the Clebsch-Gordan coefficients. In order for this ex-
pression to be useful, we need to expand the operator O(~r) on spherical
harmonics. Luckily, as we will show below for a few common examples,
there are well-known equations for such an expansion.

Expansion of V (~r) on spherical Harmonics

Given a basis of our Hilbert state defined by the functions ψn,l,m (~r) =
Rnl (r)Yl,m (θ, φ) or |nlmi we can represent the Hamiltonian as a matrix
with matrix elements

Hi,j = hni li mi |H|nj lj mj i . (7.187)

235
Chapter 7. Symmetry and conservation laws

For the potential energy we have the operator H (pot.) = V (~r) and matrix
elements
(pot.)
Hi,j = hni li mi |V (~r)|nj lj mj i . (7.188)

We expand the angular part of V (~r) on spherical Harmonics and make

a Taylor series expansion in r to find
∞ X
X l
V (~r) = Al,m rl Yl,m (θ, φ). (7.189)
l=0 m=−l

The coefficients Al,m are

4π
Al,m = p p (7.190)
(2l + 1) (l − m)! (l + m)!
× ∂zl−|m| (sign(m)∂x + i∂y )|m| V (r, θ, φ) .
r=0

With this expansion we find for the potential energy matrix elements
(pot.)
Hi,j = hni li mi |V (~r)|nj lj mj i (7.191)
∞ X
X l D E
= Al,m ni li rl nj lj hli mi |Yl,m (θ, φ)|lj mj i
l=0 m=−l
s
∞ X
l
X D
l
E (2l + 1)(2lj + 1) li mi li 0
= Al,m ni li r nj lj Clmlj mj Cl0l j0
.
4π(2li + 1)
l=0 m=−l

~
Expansion of eik·~r on spherical Harmonics
For the interaction ofE particles with light we encounter the matrix elements
D
~ ~
ni li mi eik·~r nj lj mj . In order to evaluate these we can expand eik·~r on
spherical harmonics. In principle we can use equation 7.189 and 7.190 and
expand the plane wave in powers of r. For large r this leads to a divergent
interaction. This is true for general potential expansions as discussed in

236
 7.5. Rotational symmetry and ...

the previous section as well. If as a function of increasing r the radial wave

function R(r) becomes negligible small before the Taylor expansion of V (~r)
in r diverges there is no problem. For relatively smooth potentials on the
length scale of the extension of R(r) we can use a Taylor expansion. For
~
V (~r) = eik·~r we can choose k = |~k| always so large that the potential is not
smooth in the range where R(r) is non-zero. For small wavelengths, i.e.
large wave vectors k a plane wave is not smooth over the spatial extend of
the basis functions R(r). As such we should not expand plane waves in a
Taylor series expansion. We can circumvent this problem by using spherical
Bessel functions for the radial part
∞ X
l
~ X
∗
eik·~r = 4π il jl (kr)Ylm (θk , φk )Ylm (θr , φr ). (7.192)
l=0 m=−l

With jl (kr) the spherical Bessel function

 2
l 1 ∂ sin(ρ)
jl (ρ) = (−ρ) . (7.193)
ρ ∂ρ ρ

The matrix elements become

∞ X
X l
∗
µi,j = 4π il hni li |jl (kr)|nj lj i Yl,m (θk , φk ) hli mi |Yl,m (θ, φ)|lj mj i .
l=0 m=−l
(7.194)

The matrix element

(l)
µi,j = hni li |jl (kr)|nj lj i , (7.195)

is known as the lth multipole interaction strength between state j and i at

the angular wavenumber k. The expansion on spherical Bessel functions
and spherical harmonics is known as a multipole expansion.

237
Chapter 7. Symmetry and conservation laws

1
Expansion of |~
r1 −~r2 | on spherical harmonics

Many interactions between two particles fall off as 1/r with r the distance
between the two particles. In order to evaluate the matrix elements of
the interaction between two particles we thus want to expand the function
1
|~
r1 −~r2 | on spherical harmonics. One finds

∞ X
l
1 X 1 min(r1 , r2 )l ∗
= Yl,m (θ1 φ1 )Yl,m (θ2 φ2 ). (7.196)
|~r1 − ~r2 | 2l + 1 max(r1 , r2 )l+1
l=0 m=−l

For the calculation of matrix elements, one needs to take into account
that the wave function of two fermions (bosons) needs to be antisymmetric
(symmetric) under the exchange of coordinates. We will discuss this in
more detail in chapter 9.

7.5.11 Vector and Tensor operators

We define a vector operator as an operator with 3 components that trans-
form under rotation as a vector in R3 . In order to see what the implications
are for such an operator we can have a look at the expectation value of a
vector operator for a general state |φi and a rotated state

φ0 = R(θ, n̂) |φi . (7.197)

The operator R(θ, n̂) rotates the state |φi by an angle θ around the unit
vector n̂ and is given as
i ~
R(θ, n̂) = e− ~ θL·n̂ . (7.198)

For a vector in R3 we can also define a rotation by a 3 by 3 matrix R(α, β, γ)

as shown in equation 7.59, using Euler angles. For a vector operator O ~ we
require that
D E D E
ψ0 O~ ψ 0 = ψ R† (θ, n̂)OR(θ,
~ n̂) ψ (7.199)
D E
= R(α, β, γ) ψ O ~ ψ .

238
 7.5. Rotational symmetry and ...

Whereby the first line follows from the definition of the rotated state |φ0 i
~ being a vector operator. Note
and the second line from the definition of a O
that in the second line the expectation value is a vector and R(α, β, γ)
a matrix such that the multiplication is a matrix vector multiplication.
Equation 7.199 is valid for all states |ψi such that we find for the vector
operator

R† (θ, n̂)OR(θ,
~ ~
n̂) = R(α, β, γ)O. (7.200)

For an infinitesimal small rotation δθ around n̂ we find for the right

hand side of equation 7.200
~ =O
R(δθ, n̂)O ~ + δθn̂ × O.
~ (7.201)

For the left hand side of equation 7.200 we find for an infinitesimal small
rotation
i~ ~ − δθ i L
R† (δθ, n̂)OR(δθ,
~ n̂) = (1 + δθ L · n̂)O(1 ~ · n̂) (7.202)
~ ~
~ + δθ i L
 
=O ~ · n̂O
~ −O
~L~ · n̂ .
~
Combining the left and right hand side of equation 7.200 for infinitesimal
small rotations we find that for vector operators the following commutation
relations must hold
~ · n̂, O]
[L ~ = −i~ n̂ × O.
~ (7.203)

For Cartesian coordinates x, y and z this yields the commutation relations

[Lx , O y ] = i~O z , (7.204)

[Ly , O z ] = i~O x ,
[Lz , O x ] = i~O y ,
[Lx , O z ] = −i~O y ,
[Ly , O x ] = −i~O z ,
[Lz , O y ] = −i~O x ,

239
Chapter 7. Symmetry and conservation laws

[Li , O i ] = 0.
We can use the commutation relations for the definition of a vector operator.
Any operator with 3 components that fulfills the commutation relations of
equation 7.203 is a vector operator. Some well known vector operators
are the angular momentum L, ~ the position operator ~r and the momentum
operator ~p.
As we will see in the next section, the rotation properties of products
of vector operators becomes much nicer if we do not work with Cartesian
vector operators, but with spherical vector operators. We define
r
1
O −1 = (O x − iO y ) (7.205)
2
O0 = Oz
r
1
O +1 = (−O x − iO y ) .
2
The spherical vector operators rotate like a spherical harmonic with l = 1.
This becomes clear if one writes down the explicit equations for the spherical
harmonics with l = 1.
r r 
4π 1 x y
Y1,−1 (θ, φ) = −i (7.206)
3 2 r r
r
4π z
Y1,0 (θ, φ) =
3 r
r r 
4π 1 x y
Y1,1 (θ, φ) = − −i .
3 2 r r
We can express a rotation of a spherical vector operator with the use of
a Wigner D matrix.
X
R† (α, β, γ)O m R(α, β, γ) = Dml=1
0 m (α, β, γ)O m0 . (7.207)
m0
For the commutator of the angular momentum operator with a spherical
vector operator we find
[Lz , O m ] = ~mO m (7.208)

240
 7.5. Rotational symmetry and ...

√ √
[L± , O m ] = ~ 1 ± m 1 ± m + 1O m±1 .

Tensor operators
We can generate a rank k Cartesian tensor operator by taking the tensor
product of k Cartesian vector operators. The elements of tensor operator
T are given as
(k)
T i,j,... = O i O 0j . . . . (7.209)

A Cartesian tensor operator of rank k has 3k elements. The tensor operator

(k)
T i,j,... rotates by rotating each vector operator separately

(k)
X
R† T i,j,... R = Rii0 O i0 Rjj 0 O 0j 0 . . . , (7.210)
i0 ,j 0 ,...

with R the rotation operator and Rii0 element i, i0 of a 3 by 3 Cartesian

rotation matrix.
The rotation of a Cartesian tensor operator mixes all elements of this
operator. At the same time we know that we can make linear combinations
of the elements of such an operator that behave simple under rotations. For
a rank 2 tensor combining operator O ~ and O ~ 0 we know that the operator

(k=2)
T 0,0 =O ~ 0,
~ ·O (7.211)

does not change under rotations. Furthermore the operator

~ (k=2) = O
T ~ 0,
~ ×O (7.212)
1

rotates like a vector operator. In total a rank 2 tensor operator has 9 ele-
ments. We see that we can create one linear combination of these elements
that rotates like a scaler, 3 linear combinations of the tensor elements rotate
like a vector and the other 5 are not determined yet.
The rotation properties of tensor operators are much nicer if we look at
spherical tensors instead of Cartesian tensors. If we couple two spherical

241
Chapter 7. Symmetry and conservation laws

vectors operators, then we couple two objects that rotate like spherical
harmonics with l = 1. The coupling of these two operators will yield 3 sets
of operators that rotate like a state with angular momentum l = 0, l = 1,
or l = 2. The l = 0 operator rotates like a scalar, the l = 1 operators like
a spherical vector, and the l = 2 operators are the 5 elements of a rank 2
Cartesian tensor we did not identify before. We define the spherical tensor
operators of rank 1 to be equal to the spherical vector operators

T (k=1)
q = O m=q . (7.213)

The coupling of two spherical tensors can be done with the use of the
Clebsch-Gordan coefficients
X kq
T (k)
q = Ck1 q1 k2 q2 T (k1) (k2 )
q1 T q2 . (7.214)
q1 q2

We will call k the rank of the spherical tensor operator, which has a slightly
different meaning than the rank of a Cartesian tensor operator. The indices
k and q are similar to the indices l and m used for states with angular
momentum l and z projection m.
(k)
The spherical tensor operator T q has 2k + 1 elements with q ∈ [−k, k].
The tensor rotates like a spherical harmonic with angular momentum l = k.
As such the rotation can be expressed with the use of a Wigner D matrix
(k)
X
R† (α, β, γ)T (k)
q R(α, β, γ) = Dql=k
0 q (α, β, γ)T 0 .
q (7.215)
q0

As spherical tensor operators transform like angular momentum states

with k = l under rotation we find that the commutators of spherical tensor
operators and the angular momentum operator are

[Lz , T (k) (k)

q ] = ~qT q (7.216)
(k)
p p
[L± , T (k)
q ] = ~ k ∓ q k ± q + 1T q±1 .

242
 7.5. Rotational symmetry and ...

7.5.12 Wigner-Eckart theorem and reduced matrix elements

We will encounter spherical tensor operators in several parts of physics. The
~
multipole expansion of eik·~r as shown in equation 7.192, yields a spherical
tensor operator. As such for the interaction of a free moving particle in
space with a locally bound state we will need to evaluate expectation values
of spherical tensor operators. We will be interested in integrals of the form
D E
µkq 0
lml m0 = lm T (k) 0 0
q l m . (7.217)

µkq 0 0
lml0 m0 describes the transition from state |l m i to state |lmi due to the
(k)
operator T q . For a given l, l0 and k there are (2l + 1) × (2l0 + 1) × (2k + 1)
different matrix elements µkq lml0 m0 . These matrix elements with different m,
0
m , or q are all related to each other. The Wigner-Eckart theorem states
that
D E D E
0 0 (k) 0
lm T (k)
q l m = C lm
0 0
l m kq l T l , (7.218)

with l T (k) l0 the reduced matrix element that is independent of m, m0 or

q and Cllm 0 m0 kq the Clebsch-Gordan coefficient coupling angular momentum

state |l0 m0 i with state |kqi to state |lmi. Whereby we now used the Clebsch-
Gordan coefficients not to couple two states with each other, but to describe
the rotational dependence of the product of a spherical tensor operator with
a state.
Equation 7.218 is valid for all values of m, m0 and q, such that we can
pick a convenient value for them to calculate the reduced matrix elements.
For example
D E 1 D (k)
E
l T (k) l0 = l0 l0 T 0 l0 0 , (7.219)
Cl0 0k0

and
D
0 0
E Cllm
0 m0 kq
D
(k)
E
lm T (k)
q lm = l0 l0 T 0 l0 0 . (7.220)
Cl0 0k0

243
Chapter 7. Symmetry and conservation laws

We can proof the Wigner-Eckart theorem (equation 7.218) by look-

(k) (k)
ing at the commutator of L± and T q . We know that [L± , T q ] =
√ √ (k)
~ k ∓ q k ± q + 1 T q±1 , such that
D h i E D E
0 0 (k) 0 0
p p
lm L± , T (k)
q l m = ~ k ∓ q k ± q + 1 lm T q±1 l m . (7.221)

At the same time we have

D h i E D E D E
0 0 (k) 0 0 0 0
lm L± , T (k)
q l m = lm L T
± q l m − lm T (k)
q L ± l m (7.222)
√ √ D
0 0
E
=~ l ± m l ∓ m + 1 lm ∓ 1 T (k) q lm
√ √ D
0 0
E
− ~ l0 ∓ m0 l0 ± m0 + 1 lm T (k)
q l m ± 1

If we combine
D equationE7.221 and 7.222 we find a recurrence relation be-
(k)
tween lm L± T q l0 m0 for different values of m and m0

√ √ D
(k) 0 0
E
l ± m l ∓ m + 1 lm ∓ 1 T q l m = (7.223)
√ √ D
0 0
E
l0 ∓ m0 l0 ± m0 + 1 lm T (k)
q l m ± 1
D E
(k)
+ k ∓ q k ± q + 1 lm T q±1 l0 m0 .
p p

The solution to equation 7.223 can be found in equation 7.139 where we

found a recurrence
D relation
E between the Clebsch-Gordan coefficients. We
(k) 0 0
find that lm T q l m = Cllm 0 m0 kq is a solution to equation 7.223. It is

however not the only solution. If we multiply the Clebsch-Gordan coeffi-

cient with an m, m0 , and q independent constant we still have a solution.
This constant is exactly the reduced matrix element l T (k) l0 we intro-
duced in equation 7.218. With this we proved the Wigner-Eckart theorem.
In later chapters we will encounter a few cases where we will use the theo-
rem. Either to simplify calculations or to assert the existence of a solution
of a given form.

244
 7.5. Rotational symmetry and ...

7.5.13 Rotation group, SO(3) and complex rotation group

SU(2)
The set of all rotations around a point (the origin) in R3 form a group. We
know that successive application of two rotations yield a new rotation, suc-
cessive rotations are associative, a rotation by 0 degrees yields the identity
operator and each rotation can be undone by a rotation. These properties
of all rotations are, as stated in section 7.3, the necessary and sufficient
conditions to form a group. The operator L2 commutes with all rotation
operators. Based on the discussion in the previous sections we furthermore
know that the eigenstates of the angular momentum operator L2 can be
used to represent an irreducible representation of the rotation group SO(3).
The eigenstates of L2 can be written as |lmi with eigenvalue ~2 l(l +1). The
eigenstates are 2l + 1 fold degenerate with m running from −l to l in steps
of 1 and 2l ∈ N. Acting with a rotation operator R(θ, n̂) that rotates an
eigenstate |lmi of L2 around the n̂ axis through the origin by the angle θ
yields the state
l
X
l 0
R(θ, n̂) |lmi = Dm 0 m (θ, n̂) lm , (7.224)
m0 =−l

with Dm l
0 m (θ, n̂) the Wigner D matrix. A rotation can mix the degenerate

states of L2 , but does not mix states with different l. We can assign to each
irreducible representation, i.g. the set of all 2l + 1 states |lmi with different
m for a fixed l, a character χ. The character is a set of (complex) numbers
that indicates how the representation, i.g. the states |lmi, behave under
rotation. The element of the character of an irreducible representation for
a specific rotation is given by the trace of the Wigner D matrix for that
rotation
l
X
l
χ(l, θ, n̂) = Dmm (θ, n̂). (7.225)
m=−l

In order to understand the physical importance of the previous para-

graph we can have a look at an example. As an example we take the set

245
Chapter 7. Symmetry and conservation laws

of one dimensional real functions f (x) and their behaviour under inversion
symmetry ix7→−x . The eigenfunctions of Ix7→−x are even or odd functions
with eigenvalue ±1
ix7→−x fg (x) = +fg (x) (7.226)
ix7→−x fu (x) = −fu (x).
Each function f (x) can be written as a sum of an even (gerade) and odd
(ungerade) contribution
f (x) = fg (x) + fu (x). (7.227)
Looking at the behavior of the even and odd functions separately often
greatly simplifies the problem at hand. We can recast the previous state-
ments in the language of groups. The inversion operator combined with
the identity operator forms a group. This group has two irreducible rep-
resentations. The even and odd representation. The character of the even
representation is χg = {1, 1} the character for the odd representation is
χu = {1, −1}. Whereby the two elements of the character refer to the sign
one obtains if one acts with the identity operator or with the inversion
operator on the irreducible representation.
It is useful to calculate the character for the irreducible representations
of the rotation group. We first of all notice, that, while we have rotational
symmetry, it should not matter around which axis we rotate. We thus
find that χ(l, θ, n̂) = χ(l, θ, ẑ) = χ(l, θ). We can choose to calculate the
character for a rotation around the ẑ direction. For a rotation around the
ẑ direction we have
i
R(θ, ẑ) |lmi = e− ~ Lz θ |lmi (7.228)
−imθ
=e |lmi .
l
The Wigner D matrix for a rotation around z is diagonal, with Dmm (θ, ẑ) =
e−imθ . The trace of the Wigner D matrix yields
l
X
χ(l, θ) = e−imθ (7.229)
m=−l

246
 7.5. Rotational symmetry and ...

sin((l + 12 )θ)
= .
sin( 12 θ)

For θ = 0 the character is equal to the multiplicity of the representation,

i.e. χ(l, 0) = 2l + 1. For θ = 2π we find however that the character is
equal to χ(l, 2π) = (−1)2l (2l + 1). Irreducible representations with an even
multiplicity, i.e. those with half-integer angular momentum, obtain a minus
sign under a rotation of 2π. This will become important when we discuss
the spin of an electron in the next chapter.

SO(3) v.s. SU(2)

Rotations in R3 can be represented by real orthonormal 3 by 3 matrices.
The inverse of a rotation matrix is equal to its transpose. The determinant
of a matrix describing a rotation is 1 as rotations do not change the length
of the rotated vector. With the use of these matrices, a rotation becomes
equivalent to matrix multiplication. The group of all rotations in R3 can
be represented by the group of all real 3 by 3 orthonormal matrices with
determinant 1 under matrix multiplication. These matrices are called the
special orthogonal matrices of dimension 3, explaining the name SO(3).
The rotation matrices for spin 1/2 particles are however given by 2 by 2
complex matrices. The inverse of these matrices is not given by their trans-
pose, but by their complex conjugate transpose. The matrices are unitary.
These matrices are called the special unitary matrices of dimension 2, hence
the name SU(2). The spin 1/2 particles are not part of representations of
the rotation group SO(3) but of the complex rotation group SU(2).

247
Chapter 7. Symmetry and conservation laws

248
Chapter 8

The Hydrogen atom

Before Schrödinger introduced his equation, atomic physics was poorly un-
derstood. The Bohr model based on results from atomic spectroscopy is
able to describe some of the phenomena in atomic physics. It was, from the
moment the Bohr model was introduced, well understood that the Bohr
model has major physical internal problems (see section 1.7). The Bohr
model can be used to describe the energy levels, but the model has little
predictive power. Schrödinger in his first paper [7] derived the excitation
spectrum of the Hydrogen atom. These results made his theory an immedi-
ate success. With the knowledge of the previous chapters, we are now able
to reproduce these results and explain some of the properties of a Hydrogen
atom.

8.1 The central potential

An Hydrogen atom is made from an electron and proton that interact via
the Coulomb interaction. We consider an Hydrogen atom isolated from
other atoms without excitations. In this case we expect to find a system
that is not changing over time. A wave function that does not change over
time is described by a stationary state. As such, we will not generate a
wave packet to represent the electron rotating around the proton, but look

249
Chapter 8. The Hydrogen atom

for the stationary solutions of the Schrödinger equation. These are the
eigenstates of the Hamiltonian. The Hamiltonian is given by

1 2 1 2 e2 1
H= ~pp + ~pe − , (8.1)
2mp 2me 4π0 |~re − ~rp |

with mp (me ) the mass of the proton (electron), ~rp (~re ) the position of the
proton (electron) and ~pp = −i~∇p (~pe = −i~∇e ) the momentum operator
for the proton (electron). The last term represents the Coulomb interaction.
As a first step we replace the two particle differential equation by two
decoupled differential equations related to the centre of mass and the rel-
ative distance between the proton and electron. We define the centre of
mass and relative coordinates as

~ = mp~rp + me~re
R (8.2)
mp + me
~r = ~re − ~rp .

Next we define the corresponding momentum operators. For this we can

look at the partial derivatives and find for the i-th component of the coor-
dinates

∂ ∂~ri ∂ ∂R~i ∂
= + (8.3)
∂~rp,i ~i
∂~rp,i ∂~ri ∂~rp,i ∂ R
∂ mp ∂
=− + ,
∂~ri mp + me ∂ R ~i
∂ ∂~ri ∂ ∂R~i ∂
= +
∂~re,i ~i
∂~re,i ∂~ri ∂~re,i ∂ R
∂ me ∂
=+ + ,
∂~ri mp + me ∂ R ~i

such that
∂ ∂ ∂
= + , (8.4)
~
∂ Ri ∂~rp,i ∂~re,i

250
 8.1. The central potential

 
∂ 1 ∂ ∂
= mp − me .
∂~ri mp + me ∂~re,i ∂~rp,i
The centre of mass and relative momentum operators become
~ = ~pp + ~pe = −i~∇R
P (8.5)
mp~pe − me~pp
~p = = −i~∇r .
mp + me
With

M = me + mp (8.6)
me mp
µ= ,
me + mp
we find for the Hamiltonian
1 ~2 1 2 e2 1
H= P + ~p − . (8.7)
2M 2µ 4π0 |~r|
There is no interaction between the centre of mass and relative coordi-
nates and we can separate the solution for the Schrödinger equation in a
product wave function for each coordinate separately. With
~ = φ(~r)ϕ(R),
Φ(~r, R) ~ (8.8)

and,
1 ~2
H R~ = P (8.9)
2M
1 2 e2 1
H ~r = ~p − ,
2µ 4π0 |~r|
we find
~ = Etot φ(~r)ϕ(R)
(H R~ + H ~r )φ(~r)ϕ(R) ~ =⇒ (8.10)
H ~r φ(~r) H R~ ϕ(R)~
+ = Etot =⇒
φ(~r) ~
ϕ(R)

251
Chapter 8. The Hydrogen atom

~ = E ~ ϕ(R)
H R~ ϕ(R) ~
R
H ~r φ(~r) = E~r φ(~r)
E~r + ER~ = Etot .

For the wave function of the centre of mass we find

~ = 1 ~ ~
ϕ(R) eiK·R , (8.11)
(2π)3/2

with energy
~ 2
~2 |K|
ER~ = . (8.12)
2M
This is a plane wave. One can make a wave packet from these plane waves to
describe the position and velocity of the Hydrogen atom as a free quantum
particle.
The degrees of freedom related to the center of mass are generally not
the ones we are interested in. If we look at excitations of an atom we look at
the internal degrees of freedom represented by the relative coordinates. In
order to find the eigenstates of the Hamiltonian for the relative coordinates,
we need to do a bit more work.
The operator H ~r is invariant under rotation (It has the same form
as the operator given in equation 7.57). We thus know that angular mo-
mentum commutes with the Hamiltonian. We furthermore know that the
angular momentum is a conserved quantity, i.e. the expectation value of
the angular momentum operator (and all powers of the angular momen-
tum operator) for an arbitrary wave packet Epropagating in this potential
~ n Ψ(t, ~r) = 0). Last but not least we
D
∂
are time independent ( ∂t Ψ(t, ~r) L
know that the eigenstates of the Hamiltonian are also eigenstates of the
angular momentum operator L2 . The eigenstates of L2 have eigenvalues
l(l + 1) and a degeneracy 2l + 1. We label the different degenerate states
with equal l by the projection of the angular momentum on the z axis going
from m = −l to m = +l. The easiest way to continue with our quest to
find the eigenstates of the Hamiltonian is to write the Hamilton operator in

252
 8.2. Angular solution

spherical coordinates. Using the coordinate definitions from equation 7.96

we find

~2 1 ∂ ∂2
     
2 ∂ 1 ∂ ∂ 1
H ~r = − r + 2 sin θ + 2 2
2µ r2 ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2
(8.13)
e2 1
− ,
4π0 r

whereby the first line of this equation represents the kinetic energy in spher-
ical coordinates and the second line the potential or Coulomb energy be-
tween the electron and proton. With the use of the angular momentum
operator (L~ = ~r × ~p)

1 ∂2
   
~ 2 = −~2 1 ∂ ∂
L sin θ + (8.14)
sin θ ∂θ ∂θ sin2 θ ∂ϕ2

we have

~2 L2 e2 1
   
1 ∂ 2 ∂
H ~r = − r − 2 2 − . (8.15)
2µ r2 ∂r ∂r ~ r 4π0 r

8.2 Angular solution

In equation 8.15 we see that the only angular dependence on θ and ϕ enters
via the operator L2 . As a next step we assume that the eigenfunction of
H ~r can be written as a product of a radial function R(r) and an angular
dependent function Y (θ, ϕ)

φ(~r) = R(r)Y (θ, ϕ). (8.16)

The angular dependent function needs to be an eigenfunction of L2 , which

are the spherical harmonics ( see section 7.5.5 )

L2 Yl,m (θ, ϕ) = l(l + 1)~2 Yl,m (θ, ϕ). (8.17)

253
Chapter 8. The Hydrogen atom

The angular wave functions shown in figure 7.3 or 7.5 are besides eigen-
states of the angular momentum operator also the solutions for the atomic
orbitals. One uses the term orbital for a wave function that is relatively
well localized, similar to the term wave packet. The term orbital is mostly
used for stationary states, the term wave packet for states that show some
dynamics and as such are superpositions of stationary states. The atomic
orbitals play a very important role in understanding the chemical and phys-
ical properties of molecules and solids. Why, if, and how atoms bind and
form molecules or solids and what the properties of the resulting materials
are is intimately related to these angular wave functions.

8.3 Radial solution

With L2 Yl,m (θ, ϕ) = l(l + 1)~2 Yl,m (θ, ϕ) we find for the Hamiltonian acting
on the radial part of the wave function:

~2 1 ∂ e2 1
   
2 ∂ l(l + 1)
Hr = − r − − , (8.18)
2µ r2 ∂r ∂r r2 4π0 r

such that we seek solutions of the eigen value equation

H r R(r) = ER(r). (8.19)

We can simplify the equation by introducing the function

P (r) = rR(r). (8.20)

With
 
∂ 1 ∂
R(r) = 2 r P (r) − P (r) , (8.21)
∂r r ∂r

and thus
∂2
 
∂ 2 ∂
r R(r) = r 2 P (r), (8.22)
∂r ∂r ∂r

254
 8.3. Radial solution

and given that rH r R(r) = EP (r), we find the Hamiltonian acting on P (r)
to be:
~2 ∂ 2 l(l + 1)~2 e2 1
H=− + − , (8.23)
2µ ∂r2 2µr2 4π0 r
such that
HP (r) = EP (r). (8.24)
Form equation 8.23 we see that we are effectively looking for the stationary
solutions of a particle in a one-dimensional potential. The mass of the
electron is replaced by the reduced mass µ. The Coulomb potential is
2
modified by the additional term l(l+1)~
2µr2
.
In order to solve this eigenvalue equation, we use the same method as
we did for the solution of the Harmonic oscillator (see section 5.6.1). We
first look at the solution in the limit where r → ∞ and r → 0 and then use
this solution multiplied with a polynomial of order n as a trial solution for
the full eigenvalue equation. One might wonder why we need to test also
the limit for r → 0, as we did not test this limit for the harmonic oscillator.
One should realize that the harmonic oscillator defines a potential that
is analytic over the full domain of the function. This is not true for the
Coulomb potential. The Coulomb potential diverges for ~r = 0. If the
potential diverges we know that the derivative of φ(~r) at ~r = 0 might show
a discontinuity. Following the solution for the delta function potential (see
section 5.3.1) we should expect a cusp in φ(~r) at r = 0 that is proportional
to the value of φ(~r = 0).
For r → 0 we know that E and 1/r are small compared to 1/r2 . We can
thus neglect the terms proportional to E and 1/r and only keep the terms
proportional to 1/r2 in equation 8.23. In the limit r → 0 we thus find the
eigenvalue equation
∂2 l(l + 1)
− 2
P0 (r) + P0 (r) = 0. (8.25)
∂r r2
This equation is satisfied if
P0 (r) = rl+1 , (8.26)

255
Chapter 8. The Hydrogen atom

or

P0 (r) = r−l . (8.27)

The solution that behaves like rl+1 is called the regular solution and can
lead to a wave function that is square-integrable. The solution that behaves
like r−l is called the irregular solution. For l > 0 the irregular solution is
ill-defined for ~r → 0. The radial function P (r) diverges, but the angular
function can have any phase such that the wave function is discontinuous.
As such the irregular solutions can not be used to build quantum states.
The irregular solution for l = 0 will be discussed later in this chapter 8.4.
We find that for small r, P (r) should behave like rl+1 (R(r) behaves like
rl ).
Next, we focus on the limit where r → ∞. In this limit we can neglect
the potential terms and need to find a solution for

~2 ∂ 2
− P∞ (r) = EP∞ (r). (8.28)
2µ ∂r2

with k 2 = 2µE
~2
we find two separate types of solutions depending if E > 0
for the free, scattering states and E < 0 for the bound states:
(
c1 eikr + c2 e−ikr E > 0
P∞ (r) = , (8.29)
e−|k|r E<0

with c1 and c2 complex constants that can be chosen later.

8.3.1 Bound states

We will first discuss the bound states with E < 0. The bound eigenstates
are the normalizable stationary solutions of the Schrödinger equation for
the hydrogen atom. One should realize that these are not all eigenstates of
the Hamiltonian. The resulting set of all bound eigenstates does not define
a complete orthonormal basis. Only if one includes also the continuous
spectrum of the scattering states one obtains a complete basis set. The

256
 8.3. Radial solution

Hydrogen excitation spectrum is determined by the normalizable stationary

states and as such, we focus on the bound states first.
Based on the solution in the limit r → 0 and r → ∞ we make the ansatz
that our radial equation is given by
n0
X
P (r) = r l+1
αj rj e−|k|r . (8.30)
j=0

This initial guess function has the correct behaviour for r → 0 and r → ∞.
With αn,l,q = αj=q−l−1 we can rewrite this trial function as
n
X
Pn,l (r) = αn,l,q rq e−|k|r , (8.31)
q=l+1

whereby we now included the explicit dependence on l and n in our notation

of both P and α. We require that 0 ≤ l < n for Pn,l (r) to be different
from 0. With this trial function we can look for eigenfunction solutions of
the Hamiltonian 8.23. We insert the trial function into the eigenfunction
equation and after evaluating the derivatives we find
n
X ~2 l(l + 1) − ~2 m(m − 1)
αn,l,q rq−2 e−|k|r (8.32)
2µ
q=l+1
 2  !
~ |k|m e2
+ − r = 0.
µ 4π0

We now extend the sum to start at q = 1 and set all αn,l,q that were not in
the original sum to zero. With αn,l,q = 0 ∀q ≤ l, q > n we have
∞
X l(l + 1) − q(q + 1)
αn,l,q+1 (8.33)
2
q=1
!
µe2
 
+ |k|q − αn,l,q rq−1 = 0.
4π0 ~2

257
Chapter 8. The Hydrogen atom

This equation is valid for all values of r. As such each term needs to be
zero. We find
µe2
 
l(l + 1) − q(q + 1)
αn,l,q+1 = − |k|q αn,l,q . (8.34)
2 4π0 ~2
As αn,l,n+1 = 0 and αn,l,n 6= 0 we have
µe2
− |k|n = 0, (8.35)
4π0 ~2
or
µe2
|k| = . (8.36)
n4π0 ~2
Which yields for the energy
 2 2
e µ 1
En = − . (8.37)
4π0 2~2 n2
This is the same result as Bohr obtained (see equation 1.65) within the Bohr
model and describes the experimental findings as described by Rydberg (see
equation 1.54). It is this result that Schrödinger obtained in his first paper
and shows the power of his equation.
With the energy of each eigenstate given by the quantum numbers n, l
and m we can calculate the wave function. Equation 8.34 defines the rela-
tion between the coefficients of the polynomial part of the radial function
Pn,l (r) given by equation 8.31. If we do not worry about the normalization
we can take αn,l,q=l+1 = 1 and use equation 8.34 to calculate the full radial
wave function. The radial wave function is given by an exponential function
times a polynomial. The polynomial contains the terms rl+1 to rn . These
polynomials are so important that they have gotten a name. We call them
(up to some scaling and a factor rl+1 ) the associated Laguerre polynomials
L2l+1
n−l−1 . In terms of these polynomials we find for the normalized radial
wave function
s
2 3 (n − l − 1)! 2r l − aµr n 2l+1
    
2r
Rn,l (r) = e Ln−l−1 , (8.38)
naµ 2n(n + l)! aµ n aµ n

258
 8.3. Radial solution

2
with aµ = 4π 0~
µe2
the (reduced) Bohr radius.
The Bohr radius aµ (5.29 10−11 meter) defines a typical distance in
atomic physics and accordingly a length scale for the distances between
atoms in molecules and solids. We furthermore find that the radial function
scales with n and for larger n we find a more extended wave function. For
all l, except for l = 0 we find that R(r = 0) = 0. The divergence in the
potential at ~r = 0 leads to a cusp in the wave function for l = 0, but not
for states with finite angular momentum l > 0. Last but not least, the
Laguerre polynomials have n − l − 1 zeros.

8.3.2 Scattering states

Next, we will consider the free, unbound, or scattering states. These states
extend up to infinity. As such the stationary states or eigenstates of the
Hamiltonian are not normalizable. We can however use the stationary
states as a basis to generate wave packets that describe physical particles.
From equation 8.27 and 8.29 we know that the radial wavefunction at r → 0
must behave as P0 (r) = r−l and at r → ∞ as P∞ (r) = c1 eikr + c2 e−ikr .
In equation 3.16 we defined the current operator, the current density as-
p
sociated with a plane wave eikr is equal to ~km = m . We can not have a
stationary state with a radial current. For a state with a radial wavefunc-
tion given by eikr the particle would disappear to infinity and for a radial
wavefunction given by e−ikr the particle would accumulate at ~r = 0. The
radial wavefunction in the limit for r → ∞ thus needs to be a standing
wave, composed of two propagating waves with equal amplitude inward
and outward.
As a trial function we will take the polynomial expansion
∞
X
Pk,l = αk,l,q (kr)q . (8.39)
q=l+1

For large q the relation

αk,l,q 1
lim =− , (8.40)
q→∞ αk,l,q−2 q(q − 1)

259
Chapter 8. The Hydrogen atom

must hold in order to fulfill the boundary condition at r → ∞.

If we insert the trial wave function 8.39 into the Schrödinger equation
8.23 we find
∞  
q+1 2
X
q q+2
αk,l,q (q(q − 1) − l(l + 1)) (rk) + (rk) + (rk) = 0.
aµ k
q=l+1
(8.41)
Just like the case for the radial wave equation of bound states, we extend
the summation limits and set αk,l,q = 0 if q ≤ l. We furthermore group the
sum according to powers of rk and find
∞  
X
q 2
(rk) (q(q − 1) − l(l + 1)) αk,l,q + αk,l,q−1 + αk,l,q−2 = 0. (8.42)
aµ k
q=1

This equation holds for all values of r such that

2
(q(q − 1) − l(l + 1)) αk,l,q + αk,l,q−1 + αk,l,q−2 = 0, (8.43)
aµ k
or

0

 q≤l
αk,l,q = c   q =l+1, (8.44)
1 2

l(l+1)−q(q−1) aµ k αk,l,q−1 + αk,l,q−2 q >l+1



with c a constant used to scale the function.

We indeed find the correct behaviour of αk,l,q in the limit of q → ∞ such
that our trial ansatz was successful. We canPwrite the resulting polynomials
for the radial function rRk,l (r) = Pk,l (r) = ∞ q
q=l+1 αk,l,q (kr) as generalized
Laguerre polynomials
Rk,l (r) = ie−ikr (kr)l L2l+1
−l−1− i
(2ikr). (8.45)
k

Whereby we used the scaling

i 2l+1
c= L i (0). (8.46)
k −1−l− k

260
 8.4. Cusp of the wave function at the origin

Note that Rk,l (r) is real. This can be seen from the series expansion in
equation 8.39 and 8.44 with c real. It is not so obvious in equation 8.45.

8.4 Cusp of the wave function at the origin

One of the properties of the Schrödinger equation is that for finite potentials
the gradient of its solutions are continues (see section 3.1). The solutions
we found for the hydrogen atom are, by construction, analytical for |~r| > 0.
We thus find that for |~r| > 0 all partial derivatives of ψn,l,m (~r) are continues.
At r = 0 we find that the eigenfunctions are continues but for l = 0 the
first derivative (gradient) is discontinues. The wave function with l = 0 at
~r = 0 has a cusp. As shown in section 3.1 this happens when the potential
diverges. For a divergent potential the size of the cusp must be such that
∂ m
ψn,l,m (~r) =Z ψn,l,m (~r = 0), (8.47)
∂r r→0+ ~2
with Z the residue of the pole of the divergent potential. For the electron
in the hydrogen atom we find
e2
Z=− . (8.48)
4πε0
With the knowledge on the eigenstates of the Hamiltonian we can test
if the cusp in the wave function indeed has the correct magnitude. We find
that ψn,l,m (~r = 0) = 0 for l 6= 0. Only when l = 0 there will be a cusp in
the wave function. For l = 0 we can express the size of the cusp in the 3
dimensional wave function by looking at the kink in R(|r|) around r = 0.
From equation 8.31 with Rn,l (r) = 1r Pn,l (r) we know
n
X − a rn
Rn,l=0 (r) = αn,l=0,q rq−1 e µ . (8.49)
q=1

At r → 0 only the term with q = 1 and r0 remains and we find

Rn,l=0 (r)|r→0 = αn,l=0,q=1 . (8.50)

261
Chapter 8. The Hydrogen atom

For the derivative we find

n  
∂ X r
q−2 − aµ n 1 q−1 − aµr n
Rn,l=0 (r) = αn,l=0,q (q − 1)r e − r e . (8.51)
∂r aµ n
q=1

For the limit r → 0 this yields

∂ 1
Rn,l=0 (r) = αn,l=0,q=2 − αn,l=0,q=1 . (8.52)
∂r r→0 aµ n

We now use the relation between αn,l=0,q=2 and αn,l=0,q=1 as found in the
recurrence relation 8.34
 
1 1
αn,l=0,q=2 = − − αn,l=0,q=1 , (8.53)
aµ aµ n

and get

∂ 1
Rn,l=0 (r) =− αn,l=0,q=1 (8.54)
∂r r→0 aµ
1
=− Rn,l=0 (r)|r→0
aµ
µe2
=− Rn,l=0 (r)|r→0 .
4π0 ~2
The cusp in the wave function at ~r = 0 is proportional to the value of the
function at ~r = 0. It is furthermore multiplied by µ/~2 the effective mass
divided by ~2 and the residue of the pole in the potential e2 /(4π0 ). This
is the same result as we obtained in section 3.1 based on Gauss’s theorem.

8.5 The periodic table

Besides the excitation spectrum of the hydrogen atom, the Schrödinger
equation can be used to explain the structure of the periodic table. A
relation between atomic spectroscopy and the periodic table was already

262
 8.5. The periodic table

inferred before the Schrödinger equation was known. Pauli in 1925 pub-
lished several papers where he compared atomic spectroscopy in magnetic
fields for different atoms. For a state with angular momentum l, the Zee-
man effect lifts the degeneracy of these 2l + 1 states. For a spectroscopic
transition one observes that a single absorption line without magnetic field
transforms into 2l + 1 lines in a magnetic field. Pauli related the spec-
troscopic line splitting due to the Zeeman effect to the positions of atoms
in the periodic table [73, 74]. Pauli concluded that in order to explain
the periodic table one can use the Bohr model and fill each orbital in this
model with a maximum of two electrons. This assumption remains valid
if one wants to explain the periodic table with the use of the Schrödinger
equation. Each atomic orbital with the quantum numbers n, l, and m
can at maximum host two electrons. This assumption is not part of the
Schrödinger equation and needs to be added as a separate side condition.
The number two for the filling per orbitals seems at first sight arbitrary.
In statistical physics one normally counts each degree of freedom or state
once. An atomic orbital with the quantum numbers n, l, and m seems to
have two possible internal quantum states. The degeneracy of these states
can be lifted by magnetic fields, as observed by doublet lines in absorption
spectroscopy, or in the Stern Gerlach experiment. An additional quantum
number σ was introduced; the spin of the electron [75]. We will label this
additional degree of freedom with the quantum number σ. The values of
σ can be ± 21 . The spin degree of freedom couples to magnetic fields. It
thus seems natural to relate this degree of freedom to an internal angular
momentum. The size of this angular momentum needs to be such that the
degeneracy is 2. We find
2 = 2lspin + 1, (8.55)
1
lspin = ,
2
for the internal spin-angular momentum of an electron.
In order to explain the periodic table, we find that for multi-particle
systems not all solutions of the Schrödinger equation represent physical
states. Only solutions of the Schrödinger equation where electrons occupy

263
Chapter 8. The Hydrogen atom

one-electron quantum states with different quantum numbers are realized

in nature. As we will show in chapter 9 this statement is imprecise as it is
not invariant under basis transformations between degenerate states. How
to formulate this exclusion principle, known as the Pauli principle, mathe-
matical proper will be discussed in chapter 9 on multi-particle systems.
With the additional quantum number for spin one can reproduce the
periodic table by filling up the one particle atomic states with one electron
each. These one-particle states are labeled with the quantum numbers n, l,
m, and σ. The quantum number n is called the principal quantum number
and all states with the same n are said to belong to the same atomic shell.
States with the same n but different l belong to a different subshell.
In the periodic table the atoms are ordered according to the atomic
number. The atomic number relates to the number of protons in the nucleus
and the number of electrons in the different atomic shells and sub-shells.
Furthermore, the atoms are arranged into different rows, called periods,
and columns called groups. Historically the elements are placed in groups
as they have similar chemical and physical properties. With the use of the
shell and sub-shell structure of the solutions of the Schrödinger equation,
we can understand the ordering of the elements in the periodic table on
a fundamental level. Atoms in the same group have similar fillings of the
different subshells.
The atomic orbitals are filled according to the energy of the different
shells and sub-shells. Each time one. starts to fill an electron in a shell with
a new principle quantum number n one starts a new row. The elements
in the first column in the periodic table (H, Li, Na, K, ...) are known as
the Alkali metals and contain one electron in the shell with n equal to the
period or row of the periodic table. Historically the atomic orbitals are
named s, p, d, or f instead of listing them by the angular momentum of
the electrons in these sub-shells. This somewhat arbitrarily naming of the
orbitals is related to spectroscopic transitions found for the Alkali metals.
Transitions between the np to n0 s states yield a series of sharp lines with
visible wavelengths. Transitions between the np to n0 d states yield a series
of diffuse lines in the visible wavelength range.
A specific filling of the atomic orbitals is known as an electron configu-

264
 8.6. The introduction of electron spin

ration. We write an electron configuration with a given number of electrons

(N ) in a sub-shell nl by the symbol nlN . For a state with electrons in more
than one subshell, we write the shells successively. For hydrogen the elec-
tronic configuration is 1s1 , for helium 1s2 and for lithium 1s2 2s1 . Based
on the energy of the orbitals in the hydrogen atom it is not clear if one
fills the 2s or 2p shell first. States with a different l but the same n are
degenerate. For a multi-electron atom, the degeneracy between the elec-
trons within one shell but different sub-shells is lifted. The interaction that
lifts this degeneracy is the Coulomb interaction between the electrons. The
Coulomb repulsion raises the energy of the state with the largest angular
momentum the most. Such that one first fills the 1s sub-shell, then the
2s, the 2p, the 3s, the 3p, and then not the 3d, but the 4s sub-shell. At a
filling of 19 electrons, the splitting of the n = 3 shell into sub-shells due to
the Coulomb interaction becomes larger than the energy difference between
states with different principal quantum numbers (n = 3 and n = 4). As a
result, the 4s sub-shell is filled before the 3d subshell.

8.6 The introduction of electron spin

In the previous section we have seen that in order to explain the periodic
table with the Schrödinger equation we need to assume that each state
given by the quantum numbers n, l, and m is two-fold degenerate. This
degeneracy is also found in the Stern-Gerlach experiment. Silver has the
electronic configuration 1s2 2s2 2p6 3s2 3p6 3d10 4s2 4p6 4d10 5s1 . This configu-
ration has no angular momentum. The magnetic moment of a Silver atom
thus must be related to the additional quantum number σ we introduced.
The same finding is true for atomic spectroscopy. The splitting of several
spectroscopic lines into two lines in a magnetic field, i.e. the Zeeman effect,
leads to the conclusion that electrons must have, besides a charge also a
magnetic moment. Goudsmit and Uhlenbeck [75, 76] related this additional
degree of freedom to the internal rotation or better said, internal angular
momentum of the electron.

265
Chapter 8. The Hydrogen atom

8.6.1 The spin of an electron as half integer valued angular

momentum
At the end of section 7.5.2, we discussed the allowed values of angular mo-
mentum. We found that −l ≤ m ≤ l and derived the raising and lowering
operators for angular momentum. In order for these to truncate when m > l
or m < −l we realized below equation 7.95 that the angular momentum is
quantized by half-integer values. At the same time, for the eigenfunctions
of the angular momentum operators, we only found eigenfunctions with
integer values of angular momentum. This is a direct consequence of the
requirement that the solutions to the eigenvalue equation are wave func-
tions that assign a single value (complex number) to each point in space.
If we rotate a function by 2π (around the z axis) then its value can not
change.

Rz (2π)ψ(r, θ, ϕ) = ψ(r, θ, ϕ + 2π) (8.56)

= ψ(r, θ, ϕ).

This limits the angular momentum of states that can be represented by

single-valued functions R3 → C to integer values.
For a state in a Hilbert space, we know that acting with a symmetry
operator on that state can change the phase of that state. The state ray
does not change, the state vector can change. At the same time, we know
that by definition a scalar transforms under rotations as a state with l =
0. This is the only irreducible representation of the angular momentum
operator that is single-fold degenerate. The character for states with l = 0
is χ(l = 0, θ) = 1. States with l = 0 do not acquire additional phases due to
rotations. The canonical extension of a scalar field would be to assign each
point in space two complex values. We can identify this local object with a
state with l = 12 . The state with l = 21 is the only irreducible representation
of the angular momentum operator that is two-fold degenerate. We thus
can explore the possibility to represent the spin by wave functions that
assign two complex numbers to each point in space. These local complex
vectors transform under rotations like l = 12 states. As a first step, we write

266
 8.6. The introduction of electron spin

the basis of the corresponding Hilbert space for each point in space as

B = {|1/2, 1/2i , |1/2, −1/2i}, (8.57)

representing the states by vectors labeled with |l, mi. It is worth noting
that several equivalent notations used are

B = {|↑i , |↓i}, (8.58)

or
   
1 0
B={ , }. (8.59)
0 1

The angular momentum operators on this basis are (see section 7.5.2)
 
1 1 0
Sz = ~ , (8.60)
2 0 −1

 
0 1
S+ = ~ , (8.61)
0 0
and
 
0 0
S− = ~ . (8.62)
1 0
We used the letter S for spin angular momentum instead of L, which
we will reserve for orbital angular momentum from now on. Both are
angular momentum operators. With the use of S x = 21 (S + + S − ) and
S y = −i 21 (S + − S − ) we furthermore have
 
1 0 1
Sx = ~ , (8.63)
2 1 0
and
 
1 0 −i
Sy = ~ . (8.64)
2 i 0

267
Chapter 8. The Hydrogen atom

These matrices are, besides a factor of ~/2 called the Pauli matrices.
Next we have a look on how these spin states behave under rotations.
i
With Ri (θ) = e− ~ S i θ (equation 7.69) for a rotation around the i direction
we have
cos 2θ
−i sin
2θ



Rx (θ) = , (8.65)
−i sin 2θ cos 2θ

θ
− sin 2θ




cos 2

Ry (θ) = θ , (8.66)
sin 2 cos 2θ

iθ
!
e− 2 0
Rz (θ) = iθ . (8.67)
0 e2

We find that for a rotation of θ = 2π each of these rotation matrices becomes

equal to
 
−1 0
Ri (2π) = , (8.68)
0 −1
such that

Ri (2π) |σi = − |σi . (8.69)

This is valid behaviour for a state represented by a ray in a complex vector

space, but can not be represented by a single valued function in R3 . We
represent the spin state by a complex vector field
 
ψ↑ (t, ~r)
ψ(t, ~r) = , (8.70)
ψ↓ (t, ~r)

which for each point ~r defines a complex vector within the Hilbert space
with basis B = {|↑i , |↓i}. For the normalization of this state we find
Z
ψ(t, ~r)† ψ(t, ~r)dd~r (8.71)
Rd

268
 8.6. The introduction of electron spin

Z
= ψ↑ (t, ~r)∗ ψ↑ (t, ~r) + ψ↓ (t, ~r)∗ ψ↓ (t, ~r)dd~r.
Rd

Instead of looking at a vector in spin space (on the basis B) we can

look at the local magnetization which defines a vector in real space (on the
basis of the unit vectors x̂, ŷ, and ẑ.
 
~ ~r) = ψ(t, ~r)† S x ψ(t, ~r), ψ(t, ~r)† S y ψ(t, ~r), ψ(t, ~r)† S z ψ(t, ~r) .
S(t, (8.72)

The vector field S(~~ r) defines a vector to each point in space. This vector
is real and defines the spin direction of the electron at this point in space.
The length of this vector is related to the local charge density. For states
that can be written in the form of equation 8.70 we find
q
~ ~r) · S(t,
S(t, ~ ~r) 1
†
= ~. (8.73)
ψ(t, ~r) ψ(t, ~r) 2
Both the spin density and the charge density are functions of position. Their
quotient (without integration) is a constant. This relation generally holds
for states describing single particles. For compound particles (for example
an atom with several electrons) with total spin 1/2, the wave function de-
pends on several coordinates and the above equation not necessarily holds.

8.6.2 Magnetic moments introduced by angular momentum

and spin
In order to calculate the magnetic moment generated by orbital and spin
angular momentum one needs to consider the relativistic interaction of a
charged particle with an electromagnetic field. Based on classical theory
one can obtain an idea what the form of this interaction should be. An
electron rotating with angular momentum ~l around the nucleus generates a
magnetic dipole moment. If it rotates in a circle with radius r and velocity
v the current is
ev
I=− . (8.74)
2πr

269
Chapter 8. The Hydrogen atom

The area included by the rotating electron is

A = πr2 , (8.75)

such that the magnetic dipole moment is

e
µ=− mrv (8.76)
2m
e ~
=− |L|.
2m
Generalizing to vectors and operators we have
e ~
~ =−
µ L. (8.77)
2m
For magnetic moments related to orbital angular momentum this relation
inferred from classical arguments also holds in quantum mechanics. For
magnetic moments related to the spin of an electron an additional factor
g = 2.0023 . . . enters, such that the general relation between the magnetic
moment and angular momentum becomes
e ~ e ~
~ = −g
µ S− L. (8.78)
2m 2m
The additional factor for spin is called the Landé g-factor and arises due to
relativistic effects. Within the Dirac equation the Landé g-factor has the
value g = 2. Further corrections arrises within quantum electrodynamics.
There is no classical explanation for the deviation of the Landé g-factor
from 1.
The relation between magnetic moment and angular momentum of a
body can actually be probed without the understanding of quantum me-
chanics. The idea that magnetism of solids arises from electrons rotating
around the nucleus was initiated by Ampère. The currents that gener-
ate these magnetic moments are known in the literature around 1900 as
Ampère’s currents. There are countable problems with a model of persistent
rotating currents generating the internal magnetic fields in solids. Most no-
tably the existence of persistent rotating currents violates Maxwell’s equa-
tions, (see section 1.7). As we now know the idea was very worth pursuing

270
 8.6. The introduction of electron spin

as it lead to the foundations of quantum mechanics. If we assume that

magnetism is generated by rotating electrons, then by changing the mag-
netization direction one changes the angular momentum of the electrons.
A change in the angular momentum of the electrons changes the total an-
gular momentum of the solid. As such the solid will start to rotate if one
changes the magnetization direction. At the beginning of the 20th century,
there were several attempts to measure this effect. The most noted paper
is written by Einstein and De Haas in 1915 [77]. This paper has an excel-
lent introduction summarizing all problems of classical physics of that time.
The solutions to these problems are the equations of quantum mechanics.
Einstein and De Haas found a relation between magnetization and angular
momentum such that g ≈ 1. This is the value one would expect from the
Maxwell equations if one assumes that the rotating electrons are static.
The internal conflict in these assumptions was very well known to Einstein.
Later experiments corrected the Landé g-factor to 2.00231930436256(35).
The electron g-factor is one of the most precisely determined constants in
physics. The experimental value is in complete agreement with the theory
of quantum mechanics including corrections from quantum electrodynam-
ics. The theory of quantum mechanics is one of the most accurately tested
theories.

8.6.3 Spin-orbit interaction

A rotating electron around the nucleus generates a magnetic moment µ ~L =
e ~
− 2m L. At the same time an electron has an intrinsic magnetic moment
e ~
~ S = −g 2m
due to its spin µ S. These two magnetic moments interact with
each other. If one solves the Maxwell equations for an electron rotating in a
spherical potential V (r) to first order in (v/c)2 one finds for this interaction:

1 1 ∂ ~ · S.
~
HSOC = V (r)L (8.79)
2m2 c2 r ∂r
This is known as the spin-orbit interaction. As discussed before, the Maxwell
equations for this situation are slightly troublesome. According to the

271
Chapter 8. The Hydrogen atom

Maxwell equations one can not have a persistent rotating current. The
real motivation for the spin-orbit interaction comes from the Dirac equa-
tions, which to first order in (v/c)2 yields for a single electron in a spherical
potential the same result. The Hamiltonian for the hydrogen atom with
spin-orbit interaction becomes
~p2 e2 1 e2 ~ ~
H= − + L · S. (8.80)
2µ 4π0 r 2m2 c2 4π0 r3
There are other relativistic effects that are of the same order of mag-
nitude as the spin-orbit interaction. These are the mass velocity term and
the Darwin term, which modify the kinetic energy and the potential energy.
The mass velocity and Darwin term make quantitative changes, whereas the
spin-orbit coupling makes a qualitative change. The spin-orbit interaction
reduces the symmetry of the Hamiltonian. Without spin-orbit interaction
the Hamiltonian commutes with both L and S. With spin-orbit interaction
it commutes only with L + S as we will show below. Without spin-orbit
coupling an nl shell has a (2s + 1)(2l + 1) fold degeneracy with s = 12 . With
spin-orbit coupling this degeneracy is lifted.

The spin-orbitals
In order to diagonalize the Hamiltonian of the hydrogen atom with spin-
orbit interaction, we can have a look at the commutator of the Hamiltonian
with the angular momentum operators. We know that without spin-orbit
interaction the Hamiltonian commutes with both L ~ and S.
~ We find however
that
~ · S,
[L ~ S z ] = Lx [S x , S z ] + Ly [S y , S z ], (8.81)
and
~ · S,
[L ~ Lz ] = S x [Lx , Lz ] + S y [Ly , Lz ]. (8.82)
The orbital and spin angular momentum operators do not commute with
the Hamiltonian 8.80. Therefore we know that the orbital angular momen-
tum nor the spin angular momentum are conserved quantities. If we write

272
 8.6. The introduction of electron spin

the interaction between spin and orbital angular momentum as

L ~ = Lz S z + 1 (L+ S − + L− S + ) ,
~ ·S (8.83)
2
we see that the spin-orbit coupling operator can transfer angular momen-
tum from the spin sector to the orbital sector and vice versa. For a con-
served quantity we should look at the total angular momentum J ~ of an
electron rotating around the nucleus with orbital angular momentum L ~
~
and spin S.

~ =L
J ~ + S.
~ (8.84)

Indeed we find that

~ · S,
[L ~ J~ ] = 0. (8.85)

We thus find that

2
~ ] = 0,
[H, J (8.86)

and

~ ] = 0.
[H, J (8.87)

such that the eigenstates of the Hamiltonian are simultaneously eigenstates

~ 2 and J z . For the case of a single electron, which we discuss here, we
of J
find that the Hamiltonian also commutes with L ~ 2 and S
~ 2 . Note that this
is not true for many electron atoms.
We label our states as |lsjjz i such that

~ 2 |lsjjz i = j(j + 1)~2 |lsjjz i

J (8.88)

and

J z |lsjjz i = jz ~ |lsjjz i . (8.89)

273
Chapter 8. The Hydrogen atom

jjz
With the use of the Glebsch-Gorden coefficients Css z llz
we can write these
states as linear combinations of the eigenstates of the orbital angular mo-
mentum and spin angular momentum.
X jj
|lsjjz i = Csszz llz |ssz llz i , (8.90)
sz ,lz

whereby |ssz llz i are the tensor product states of the eigenstates of the spin
|ssz i and orbital |llz i angular momentum. The eigenstates of the orbital
angular momentum are given by the spherical harmonics. The eigenstates
of the spin are given by a vector in a complex Hilbert space, which we denote
as |↑i and |↓i. Using explicit values of the Glebsch-Gordon coefficients for
s = 1/2 we find for the case where j = l + 1/2
r r
l + 1/2 + jz l + 1/2 − jz
|lsjjz i = Yl,jz −1/2 (θ, ϕ) |↑i + Yl,jz +1/2 (θ, ϕ) |↓i
2l + 1 2l + 1
(8.91)

and for the case that j = l − 1/2

r r
l + 1/2 − jz l + 1/2 + jz
|lsjjz i = − Yl,jz −1/2 (θ, ϕ) |↑i+ Yl,jz +1/2 (θ, ϕ) |↓i
2l + 1 2l + 1
(8.92)

These states are called spin-orbitals and are given as angle-dependent

functions on the spin Hilbert space. This Hilbert space of dimension two has
a basis of states representing a spin up and a spin down electron. Following
equation 8.72 we can define a vector field that assigns to each point in
coordinate space a spin direction. The spin direction does not change as
a function of r such that we only need to look at the spin direction as a
function of θ and φ. The magnetization we find around an atom due to the
spin is given by a vector field. Each point in space can have a magnetization
in a different direction. In figure 8.1,8.2,8.3, and 8.4 we show this vector
field. We plot the vectors on the surface given by r = |ψ(θ, ϕ)|2 . In order to
make the vector field easier to understand we show, besides the 3D image

274
 8.6. The introduction of electron spin

Figure 8.1: The Hydrogen atomic spin-orbitals with l = 0, j = 1/2, and

jz = (−1/2, 1/2)

a cut in the yz, zx, and xz plane. Spins parallel to the x-axis are red,
parallel to the y-axis are green, and parallel to the z-axis blue. For the s
spin-orbital with j = 1/2, the spin is coaligned for each point in space. For
the p spin-orbital with j = 1/2 the magnetization direction rotates once
around its axis as a function of θ going from 0 to π.
Spin-orbit coupling relates the direction of the spin orientation to the
spatial direction around the atom. For atoms embedded in a solid, where
the local environment is no longer spherical this leads to magnetic anisotropy.
Magnetic anisotropy creates magnetic moments that are larger in one spa-
tial direction than in another. Magnetic anisotropy is very important for
the fabrication of permanent magnets. Without anisotropy one can always
rotate the magnetization direction at very little energy cost. The intriguing
entanglement of spin direction and spatial direction induced by spin-orbit
coupling lifts in a non-spherical environment this degeneracy and makes
a Ferromagnet stable. These magnetization distributions around an atom
can be observed in experiments. If one diffracts a particle (photon, elec-

275
Chapter 8. The Hydrogen atom

Figure 8.2: The Hydrogen atomic spin-orbitals with l = 1, j = 1/2, and

jz = (−1/2, 1/2)

tron, neutron) from an atom then one becomes an image of this object.
Either measured in momentum (Fourier) space as an interference pattern,
or with the help of lenses directly in real space. The momentum-dependent
scattering of an atom is known as the (magnetic) atomic form factor.

276
 8.6. The introduction of electron spin

Figure 8.3: The Hydrogen atomic spin-orbitals with l = 1, j = 3/2, and

jz = (−3/2, −1/2, 1/2, 3/2)

277
Chapter 8. The Hydrogen atom

Figure 8.4: The Hydrogen atomic spin-orbitals with l = 2, j = 3/2, and

jz = (−3/2, −1/2, 1/2, 3/2)

278
Chapter 9

N particle systems

9.1 The Schrödinger equation alone is not enough

The Schrödinger equation, with the appropriate relativistic corrections is
one of the best tested theories we have. No experiment is known that shows
results that are not in agreement with this equation. There is one problem
with this statement. The He atom (and all other many electron systems) is
not correctly described by the exact solution of the Schrödinger equation,
including its relativistic corrections. For simplicity we can approximate
the He atom to be described by a wave function that has two electrons in
the first shell, with n = 1 and therefor l = 0. According to Schrödingers
equation there are 4 possible solutions that fulfil this orbital filling. The
two electron wave function can be a function with both electrons spin up,
both electrons spin down, the first electron spin up and the second electron
spin down or the first electron spin down and the second electron spin up.
The possibility to have four different wave functions for the ground-state
of He, i.e. a four fold degeneracy, does not agree with experiment. There
is only one realization of the ground state of the He atom. One can test
this by looking at the heat capacity of He gas. If there would be several
different states for the He atom these different degrees of freedom would
increase the heat capacity.

279
Chapter 9. N particle systems

Not all possible solutions of the Schrödinger equation are realized in

nature. Actually only a tiny fraction of all possible allowed solutions of
the Schrödinger equation represent physical states. In order to explain the
periodic table in chapter 8.5 we assumed that each spin orbital can only be
occupied by a single electron. This is an extra restriction on the Schrödinger
equation and known as the Pauli principle. In the following section we will
investigate this statement in a bit more detail.

9.1.1 Pauli’s principle

Based on the structure of the periodic table, atomic spectra, and the Bohr
model for atoms Pauli in 1925 formulated the Pauli exclusion principle: Two
or more identical fermions cannot occupy the same quantum state simul-
taneously [73, 74]. The problem with the definition of equivalent quantum
states is that the question if two quantum states are equivalent or not de-
pends on the basis one uses for the Hilbert space. Physical properties are
not allowed to depend on the basis one chooses. We thus need to formulate
the Pauli principle with a little bit more care. As we will show below, the
requirement for two fermions to be in a different state, independent of the
basis choice of the Hilbert space, requires the wave function to be fully
antisymmetric under the exchange of coordinates.
In order to understand the Pauli principle in more detail we will have
a look at the example of an He atom. We approximate the ground state of
the He atom to be a state with two electrons in the 1s shell. Without the
Pauli principle there are four different possible states.
?
|ψHe (~r1 , ~r2 )i = |ψ1s↑ (~r1 )ψ1s↑ (~r2 )i (9.1)
?
|ψHe (~r1 , ~r2 )i = |ψ1s↑ (~r1 )ψ1s↓ (~r2 )i
?
|ψHe (~r1 , ~r2 )i = |ψ1s↓ (~r1 )ψ1s↑ (~r2 )i
?
|ψHe (~r1 , ~r2 )i = |ψ1s↓ (~r1 )ψ1s↓ (~r2 )i .

In our notation we included the variables ~r1 and ~r2 in the bracket notation.
We normally do not need to do this. As we will show later in this section,

280
 9.1. The Schrödinger equation alone is not ...

for physically allowed states the coordinate dependence for N particles is

uniquely given if we know N wave functions. At this point we want to dis-
cuss how this comes about and need to discuss states that violate the Pauli
principle. In order to be able to discuss this we need the explicit depen-
dence of our wave functions on the coordinates of the different particles. It
is important to note that the 4 states given in equation 9.1 are linear inde-
pendent states, i.e. the inner product between two different states is zero.
Furthermore, if we neglect the Coulomb interaction between the electrons,
each of the states in equation 9.1 is a solution to the Schrödinger equation.
In order to implement the Pauli principle for the four states given in
equation 9.1 we can remove the states with two electrons with spin up or
two electrons with spin down. For these two states we see that the electrons
obviously occupied the same state. After doing so we are still left with two
possible solutions for the wave function of the He atom.

?
|ψHe (~r1 , ~r2 )i = |ψ1s↑ (~r1 )ψ1s↓ (~r2 )i (9.2)
?
|ψHe (~r1 , ~r2 )i = |ψ1s↓ (~r1 )ψ1s↑ (~r2 )i .

That is still one more than nature provides.

We now need to think how to define the same quantum state. For two
states that belong to the same Hilbert space we can take the inner product
of the two states. For normalized states two states are the same if their
inner product is one. A state with two electrons can be created by building
a complete Hilbert space for electron one and a separate complete Hilbert
space for electron two. The Hilbert space of two electron states is generated
by the tensor product, as a combination of each possible state of electron
one with each possible state of electron two. An inner product between the
states for electron one and the states for electron two is not defined.
We did not define our quantization axis for the previous wave functions
of He. Up and down can refer to the z direction, but could equally well
relate to the x direction. We need the electrons to be in different spin
orbitals, independent of the choice of basis we take. We can write the states
quantized in the x direction as linear combinations of the states quantized

281
Chapter 9. N particle systems

in the z direction
r
1
|ψ1s↑x i = (|ψ1s↑z i + |ψ1s↓z i) (9.3)
2
r
1
|ψ1s↓x i = (|ψ1s↑z i − |ψ1s↓z i)
2
To exemplify the relation between these two basis sets we can have a look
at the spin operator S x as a matrix on these two basis sets:
   
1
0 1 hψ1s↑z |
S x = ~ |ψ1s↑z i |ψ1s↓z i · · (9.4)
2 1 0 hψ1s↓z |
   
1
1 0 hψ1s↑x |
= ~ |ψ1s↑x i |ψ1s↓x i · ·
2 0 −1 hψ1s↓x |
Returning to the problem of the wave function of the He atom, we can write
one of the two possible solutions as
?
|ψHe (~r1 , ~r2 )i = |ψ1s↑z (~r1 )ψ1s↓z (~r2 )i (9.5)
r r +
1 1
= (ψ1s↑x (~r1 ) + ψ1s↓x (~r1 )) (ψ1s↑x (~r2 ) − ψ1s↓x (~r2 ))
2 2

1
= |ψ1s↑x (~r1 )ψ1s↑x (~r2 )i − |ψ1s↑x (~r1 )ψ1s↓x (~r2 )i
2

+ |ψ1s↓x (~r1 )ψ1s↑x (~r2 )i − |ψ1s↓x (~r1 )ψ1s↓x (~r2 )i .

After rotating the basis one finds that two electrons that appear to be in
different spin orbitals for a quantization in the z direction are at least partly
in the same spin orbital for a quantization in the x direction.
We can guarantee that two electrons are not in the same spin orbital,
independent of the basis used, when we make the wave function anti sym-
metric under exchange of the particle coordinates. We find
r  
1
|ψHe (~r1 , ~r2 )i = |ψ1s↑x (~r1 )ψ1s↓x (~r2 )i − |ψ1s↓x (~r1 )ψ1s↑x (~r2 )i (9.6)
2

282
 9.1. The Schrödinger equation alone is not ...

r  +
1
= ψ1s↑x (~r1 )ψ1s↓x (~r2 ) − ψ1s↓x (~r1 )ψ1s↑x (~r2 )
2
= |ψ1s↑x ψ1s↓x i

For fermions, only fully anti symmetric solutions of the Schrödinger

equation represent physical states. This is a tiny fraction of all possi-
ble solutions of the Schrödinger equation. For N particles that occupy N
spin-orbitals we can distribute these spin-orbitals in N ! ways over the N
coordinates of these particles. Only one of these solutions, namely the anti
symmetric one, represents a physically realized state.
The same argument holds for a state with three electrons. For example
a wave function for Lithium with one electron per spin-orbital would be:
r 
1
|ψLi (~r1 , ~r2 , ~r3 )i = |φ1s↑ (~r1 ) φ1s↓ (~r2 ) φ2s↑ (~r3 )i (9.7)
6
− |φ1s↓ (~r1 ) φ1s↑ (~r2 ) φ2s↑ (~r3 )i
+ |φ1s↓ (~r1 ) φ2s↑ (~r2 ) φ1s↑ (~r3 )i
− |φ2s↑ (~r1 ) φ1s↓ (~r2 ) φ1s↑ (~r3 )i
+ |φ2s↑ (~r1 ) φ1s↑ (~r2 ) φ1s↓ (~r3 )i

− |φ1s↑ (~r1 ) φ2s↑ (~r2 ) φ1s↓ (~r3 )i .

In 1929 Slater [78] published a paper on the theory of complex spectra.

Here he introduced a compact notation for such fully anti symmetric wave
functions in terms of the determinant of a matrix. For N particles we
define a matrix where each column has a different coordinate and each row
a different wave function. The state is then given by the determinant of
this matrix.
 
r φ1s↑ (~r1 ) φ1s↑ (~r2 ) φ1s↑ (~r3 ) +
1
|ψLi (~r1 , ~r2 , ~r3 )i = det φ1s↓ (~r1 ) φ1s↓ (~r2 ) φ1s↓ (~r3 ) (9.8)
6
φ2s↑ (~r1 ) φ2s↑ (~r2 ) φ2s↑ (~r3 )

283
Chapter 9. N particle systems

We can extend this notation to four electrons which one encounters in

Beryllium. A four electron wave function represented by a single Slater
determinant is
 
r φ1s↑ (~r1 ) φ1s↑ (~r2 ) φ1s↑ (~r3 ) φ1s↑ (~r4 ) +
1 φ1s↓ (~r1 ) φ1s↓ (~r2 ) φ1s↓ (~r3 ) φ1s↓ (~r4 )
|ψBe (~r1 , ~r2 , ~r3 , ~r4 )i = det 
φ2s↑ (~r1 ) φ2s↑ (~r2 ) φ2s↑ (~r3 ) φ2s↑ (~r4 )

24
φ2s↓ (~r1 ) φ2s↓ (~r2 ) φ2s↓ (~r3 ) φ2s↓ (~r4 )
(9.9)

In Gold (Au) we have 79 electrons. A state represented by a single Slater

determinant state for Au is thus given by a determinant of a 79 by 79
matrix
r  +
1 79 × 79
|ψAu (~r1 , . . . , ~r79 )i = det (9.10)
8.9 · 10116 matrix

For atomic Au a Single Slater determinant state has more terms than we
have atoms in the universe. If we want to write down a wave function
for a macroscopic piece of solid we need about 1023 atoms. If each atom
has one electron we need a wave function with 1023 electrons. In this case
the wave function is given by the determinant of a 1023 by 1023 matrix.
36
Such a matrix has 1023 ! ≈ 1010 terms. Calculating expectation values
36 2
 
of such an object would require to evaluate 1010 integrals. Quantum
mechanics yields a description of the world so complex that even writing
down the state of a moderately sized system becomes impossible. The use
of Slater determinants for states is however not the most compact notation,
such that there is partly a solution to the immense complexity encountered.
This we will discuss in the next section.

284
 9.1. The Schrödinger equation alone is not ...

9.1.2 Second quantization

With the use of second quantization we can at least find a way to be able
to write down some states and calculate expectation values of these states.
As we will discuss in the next section there is also a non trivial scaling
problem for many body quantum mechanics. The name ”Second quan-
tization” is historically a bad choice. We do not quantize anything new.
Second quantization is just a powerful notation to circumvent the problem
of the ridiculously many terms in a Slater determinant that render any
computation impractical.
The idea is to generate an N particle wave function by creating electrons
from the vacuum. For this we define creation operators that create electrons
in certain basis functions. As it turns out we will be able to implement
Pauli’s principle and the anti symmetrization of the wave function with the
use of anti commutation relations of these operators.
We define |0i to be the vacuum state. This is the state with zero
particles. We have h0|0i = 1 and clearly |0i 6= 0. The vacuum is not equal
to zero, but describes a space without particles.
We define a†τ as the operator creating a single electron in the one particle
wave function φτ (~r). Given a Hilbert space with states labeled by τ we find
a†τ |0i = |φτ i. For N electrons we define
 
φ1 (~r1 ) φ2 (~r1 ) ...φN (~r1 ) +
r
1  φ1 (~r2 ) φ2 (~r2 ) ...φN (~r2 ) 

a†1 a†2 . . . a†N |0i = det  . (9.11)

N!  ..


φ1 (~rN ) φ2 (~rN ) . . . φN (~rN )

For two electrons this reduces to

r +
1
a†1 a†2 |0i = (φ1 (~r1 ) φ2 (~r2 ) − φ2 (~r1 ) φ1 (~r2 )) . (9.12)
2

Instead of first creating particle 2 and then particle 1 we could also have
changed the order. This should give the same state, i.e. the same ray, but

285
Chapter 9. N particle systems

not necessarily the same vector.

r +
† † 1
a2 a1 |0i = (φ2 (~r1 ) φ1 (~r2 ) − φ1 (~r1 ) φ2 (~r2 )) (9.13)
2
= −a†1 a†2 |0i .

We find that the order in which we create the particles changes the sign
of the wave function. We can generate the desired behaviour if we define
the following anti-commutation behaviour for our creation operators

{a†α , a†β } = a†α a†β + a†β a†α (9.14)

= 0,

The anti-commutator between the creation of two Fermions is zero. This

does not only work for two Fermions, but works for any particle number.
Interchanging two adjacent rows or columns of a matrix changes the sign
of the determinant of that matrix, such that

a†1 a†2 . . . a†i a†j . . . a†N |0i = −a†1 a†2 . . . a†j a†i . . . a†N |0i . (9.15)

With the known definition that

|φτ (~r)i† = hφτ (~r)| , (9.16)

we find that
 †
a†τ |0i = h0| aτ . (9.17)

We call the operator aτ the annihilation operator. The annihilation opera-

tor aτ annihiles the particle one creates with the creation operator a†τ such
that
D E
0 aτ a†τ 0 = 1. (9.18)

We can derive the anti-commutation relations for the annihilation operator

by taking the hermitian conjugate of the anti-commutation relations of

286
 9.1. The Schrödinger equation alone is not ...

the creation operator. We find for the anti-commutation relations of the

annihilation operator

{aα , aβ } = aα aβ + aβ aα (9.19)
= 0.

In the end we want to replace an N particle wave function, which is given

by a Slater determinant with N ! terms, by a single product of N operators
acting on the vacuum. The anti symmetrization of the wave function is
achieved by enforcing the correct anti-commutation relations between the
creation operators for ket vectors and annihilation operators for bra vectors.
The only thing left to define are the anti-commutator relations between the
creation and annihilation operators.
If we define that the states |ψτ i for different values of τ define an or-
thonormal basis set of a Hilbert space and a†τ |0i = |ψτ i then we know that

D E
0 aα a†β 0 = δα,β . (9.20)

We see that annihilating a particle with quantum number α from a state

that does not occupy this state yields zero. It thus makes sence to define

aτ |0i = 0, (9.21)

and

h0| a†τ = 0. (9.22)

For the expectation value of an anti-commutator between a creation and

an annihilation operator we know that
D E
0 aβ a†α + a†α aβ 0 = δα,β . (9.23)

This relation is fulfilled if we define for the anti-commutator

{a†α , aβ } = a†α aβ + aβ a†α (9.24)

287
Chapter 9. N particle systems

= δα,β .

With this definition we find

0 = aτ aτ a†τ (9.25)
= −aτ a†τ aτ + aτ
= +a†τ aτ aτ
= 0.

The anti-commutators between the creation and annihilation operators of

particles, as described in equation 9.14, 9.19 and 9.24, define a valid al-
gebra that can be used to define states. The anti-symmetrization of the
wave-function is replaced by the anti-commutation relations between the
operators.

Operators in Second quantization

Up to this point we have defined a notation where we can write wave
functions consisting of N particles as a string of N creation operators. We
reduced the complexity of writing down single Slater determinant states
from N ! to N . In order for the method to be useful we also need to be
able to evaluate expectation values of operators. In general an operator O
depends on the coordinates used in the wave function to describe the N
particles in the problem. For N particles we can write the operator as
N
X N
X
O(~r1 , . . . , ~rN ) = O (0) + O (1) (~ri ) + O (2) (~ri , ~rj ) + . . . . (9.26)
i=1 i,j=1

We call O (0) , O (1) , O (2) , . . . the zero, one, two, etc. particle operator. The
p2
potential energy V (~ri ), and the kinetic energy 2mi are given by a one particle
1
operator. The Coulomb interaction between two electrons ∝ |~ri −~ rj | , as well
as the angular momentum operator L2 = ( i (~ri × ~pi ))2 are given by two
P
particle operators.

288
 9.1. The Schrödinger equation alone is not ...

We can rewrite an N particle operator O (N ) in second quantization by

multiplying it from the left and the right by a complete set of ket times bra
N particle states. The zero particle operator is just a constant and needs
no further consideration.
In order to write the one particle operator in second quantization we
definePthe complete set of one particle states a†α |0i = |φα i for α ∈ [1, ∞].
With α |φα ihφα | = 1 we have
X
O (1) = |φβ ihφβ | O (1) |φα ihφα | (9.27)
α,β
XD E
= φβ O (1) φα |φβ ihφα |
α,β
XD E
= φβ O (1) φα a†β aα .
α,β

In the last step we used that |φβ ihφα | = a†β aα . If these operators act on a
many particle state where orbital α is unoccupied both operators yield 0. If
they act on a many particle state where orbital α is occupied the occupied
orbital α is replaced by the orbital β.
In order to write the two particle operator in second quantization we
define the complete set of two particle states a†α a†β |0i = |φα φβ i for α ∈
[1, ∞] and β ∈ [1, α − 1]. Note that we need to take β < α as a†β a†α =
−a†α a†β . P
The identity operator written on a basis of two particle states
becomes β<α |φα φβ ihφβ φα | = 1, such that
X
O (2) = |φα φβ ihφβ φα | O (2) |φδ φγ ihφγ φδ | (9.28)
β<α,δ<γ
X D E
= φβ φα O (2) φδ φγ |φα φβ ihφγ φδ |
β<α,δ<γ
D E
φβ φα O (2) φδ φγ a†α a†β aγ aδ .
X
=
β<α,δ<γ

289
Chapter 9. N particle systems

We can rewrite the sum to include all values of β and δ by writing out the
two particle expectation value of operator O (2) .

ZZ
1 X
O (2) = φ∗α (~r1 )φ∗β (~r2 ) − φ∗β (~r1 )φ∗α (~r2 ) O (2) (~r1 , ~r2 )


2 R3
β<α,δ<γ

× (φδ (~r1 )φγ (~r2 ) − φγ (~r1 )φδ (~r2 )) d3~r1 d3~r2 a†α a†β aγ aδ (9.29)
X Z Z
= φ∗α (~r1 )φ∗β (~r2 )O (2) (~r1 , ~r2 )φδ (~r1 )φγ (~r2 )d3~r1 d3~r2
β<α,δ<γ R3

X ZZ
+ φ∗β (~r1 )φ∗α (~r2 )O (2) (~r1 , ~r2 )φγ (~r1 )φδ (~r2 )d3~r1 d3~r2
β<α,δ<γ R3

X ZZ
− φ∗β (~r1 )φ∗α (~r2 )O (2) (~r1 , ~r2 )φδ (~r1 )φγ (~r2 )d3~r1 d3~r2
β<α,δ<γ R3
ZZ !
X
− φ∗α (~r1 )φ∗β (~r2 )O (2) (~r1 , ~r2 )φγ (~r1 )φδ (~r2 )d3~r1 d3~r2
β<α,δ<γ R3

1
× a†α a†β aγ aδ
2
X ZZ
= φ∗α (~r1 )φ∗β (~r2 )O (2) (~r1 , ~r2 )φδ (~r1 )φγ (~r2 )d3~r1 d3~r2
β<α,δ<γ R3

X ZZ
+ φ∗α (~r1 )φ∗β (~r2 )O (2) (~r1 , ~r2 )φδ (~r1 )φγ (~r2 )d3~r1 d3~r2
β>α,δ>γ R3

X ZZ
+ φ∗α (~r1 )φ∗β (~r2 )O (2) (~r1 , ~r2 )φδ (~r1 )φγ (~r2 )d3~r1 d3~r2
β>α,δ<γ R3
ZZ !
X
+ φ∗α (~r1 )φ∗β (~r2 )O (2) (~r1 , ~r2 )φδ (~r1 )φγ (~r2 )d3~r1 d3~r2
β<α,δ>γ R3

1
× a†α a†β aγ aδ
2

290
 9.1. The Schrödinger equation alone is not ...

ZZ
1 X
= φ∗α (~r1 )φ∗β (~r2 )O (2) (~r1 , ~r2 )φδ (~r1 )φγ (~r2 )d3~r1 d3~r2
2 R3
αβγδ

× a†α a†β aγ aδ
1 XD ∗ E
= φα (~r1 )φ∗β (~r2 ) O (2) (~r1 , ~r2 ) φδ (~r1 )φγ (~r2 ) a†α a†β aγ aδ .
2
αβγδ

We can write any operator in second quantization as a sum over prod-

ucts of creation and annihilation operators. For a one particle operator we
found
N
X
(1)
O = O (1) (~ri ) (9.30)
i=1
X
= εαβ a†α aβ ,
α,β

with

εαβ = hφα (~r)| O (1) (~r) |φβ (~r)i . (9.31)

For a two particle operator we found

N X
X N
O (2) = O (2) (~ri , ~rj ) (9.32)
i=1 j=1

Uα,β,γ,δ a†α a†β aγ aδ ,

X
=
α,β,γ,δ

with
1
Uα,β,γ,δ = hφα (~r1 ) φβ (~r2 )| O (2) (~r1 , ~r2 ) |φδ (~r1 )φγ (~r2 )i . (9.33)
2
With this notation the calculation of expectation values reduces to the
repeated use of commutation relations. Most importantly we never need
to explicitly implement the Pauli principle as a Slater determinant with N !
terms.

291
Chapter 9. N particle systems

9.1.3 Bosons in second quantization

In the previous two sections we extensively discussed the Pauli principle, the
resulting anti-symmetrization of the wave function and how to implement
this with the use of second quantization. Those arguments are related
to Fermions. We can make a similar argumentation for Bosons. In the
derivation of the black body radiation in chapter 1.1 we implicitly assumed
that there is only a single state with two photons with energy hν1 and
hν2 . There is one major difference between Fermions and Bosons. Two
Bosons can occupy the same quantum state, or Bosonic mode. A laser
with frequency ν can have any field strength that is an integer multiple of
hν. The wave-function for N bosons in the same mode is fully symmetric.
The resulting commutation relations for the creation (b† ) and annihilation
(b ) operators for bosons are

[b†α , b†β ] = b†α b†β − b†β b†α (9.34)

= 0,

[bα , bβ ] = bα bβ − bβ bα (9.35)
= 0,

and

[bα , b†β ] = bα b†β − b†β bα (9.36)

= δα,β .

9.1.4 Wave functions beyond Single Slater determinants

Single Slater determinant states, i.e. wave functions written as a single
string of creation operators acting on the vacuum are not the most general
states one encounters. In most cases these are not the eigen states of a
many particle Hamiltonian.

292
 9.2. Complexity

Starting from an n dimensional one particle Hilbert space with basis

functions labeled by τ ranging from 1 to n the one particle states are defined
as

a†τ |0i = |φτ (~r)i . (9.37)

For a state with N electrons we can define single Slater determinant wave
functions as
Y
|ψi (~r1 , ~r2 , . . . , ~rN )i = a†τ |0i . (9.38)
τ ∈Di

Di is a subset of N one particle
wave functions from a total of n basis

n n!
functions. In total there are N = N !(n−N )! different N particle single
Slater determinants possible. A general function is written as a linear
combination of these single Slater determinants
n!
N !(n−N )!
X
|ϕj (~r1 , ~r2 , . . . , ~rN )i = αi,j |ψi (~r1 , ~r2 , . . . , ~rN )i . (9.39)
i=1

9.2 Complexity
In the previous section we have seen that one is generally not able to write
down single Slater determinant states. For a moderate number of electrons
there are just too many terms in such an expression. We also showed that
by changing notation we can write single Slater determinant states as a
string of creation operators acting on the vacuum. The work we need to
do to write such a state scales linearly with the number of electrons. The
Pauli principle is implemented via the anti-commutation relations between
the creation and annihilation operators.
Sadly the single Slater determinant representable states are generally
not the eigenstates of the problem. The eigenstates are linear superpositions
of single Slater determinant states. The number of determinants needed to
represent a many particle system scales badly with system size. As an

293
Chapter 9. N particle systems

example, if we assume that we need a modest number of 5 basis functions

per electron to accurate describe the physics the N particle Hilbert space
has dimension
 
5N (5N )!
= . (9.40)
N N !(4N )!
For an Oxygen atom with 8 electrons this would already yield a Hilbert
space of dimension 7.6 107 . Vectors representing states are 76 million long
and matrices representing operators are 76 million by 76 million large. If
we would want to do calculations for gold with 79 electrons in this way we
would have a basis of dimension 3.4 1084 . Exact solutions for a moderate
set of orbitals per electron scale binomially and become intractable for
relatively modest systems. This becomes even more so if one realizes that
we only discussed a single atom and have not started to talk about solids
with 1023 atoms.
The scaling of the complexity of the equations to solve for a quantum
system is very different from the scaling for classical systems. For classical
systems with N particles we have 3N variables to label the coordinates and
3N coordinates to label the momentum of the particles. We can label the
state with a vector that has 6N real variables. Classical systems show a
linear scaling for the representation of the micro-state with system size. The
quantum mechanical representation of a state shows binomial scaling. This
difference in scaling has severe consequences on the physics one describes.
The complexity of a quantum mechanical system makes finding exact or
even numerically accurate solutions often impossible. This becomes even
worse as the intuition one develops from small systems does not always
transfer to larger systems. For example the statistical laws that hold for a
classical system do not necessarily hold for a quantum system, as we will
discuss later when we discuss phase transitions and spontaneous symmetry
breaking.
In order to deal with the complex solutions of quantum mechanics sev-
eral approximations and effective models exist. These approximations gen-
erate solutions that are better tractable than the general approach of sum-
ming Slater determinants. One of the more powerful approximations in-

294
 9.3. Entanglement and correlations

troduces a set of quasi particles that emerge from the interaction of many
particles. These quasi particles are described by an effective Hamiltonian
which is valid in a low energy regime. As these quasi particles are often non-
interacting the resulting low energy Hamiltonian is easier to solve. These
approximations sometimes work and allow us for example to predict the
properties of semi conductors and with that the electronics we use in our
devices.
What all of these models and approximations have in common is that
they strive to find ways to deal with the complexity that emerges from
the simple Schrödinger equation. Although the basic laws that govern the
interactions between particles are extremely well understood the resulting
states that emerge for a large number of these particles are extremely rich.
This richness in these equations allows for the complex structures we en-
counter in every day life. The complex behaviour of many particle systems
is often not easy to predict by extrapolating the behaviour of a few parti-
cles. In the words of P.W. Anderson: ”The ability to reduce every thing
to simple fundamental laws does not imply the ability to start from those
laws and reconstruct the universe.” [67, 68].

9.3 Entanglement and correlations

A very important ingredient of the equations governing quantum mechanics
is that they are linear. If ψ1 is a solution of the Schrödinger equation and
ψ2 is a solution of the Schrödinger equation then also ψ = ψ1 + ψ2 is a
solution of the Schrödinger equation. This can lead to the formation of
entanglement between particles. For two particles, the state of one particle
can depend on the state of the other particle, even if they do not (no longer)
interact with each other. We will here show two examples of this effect.
The first example relates to an abstract state on an abstract Hilbert space.
After the abstract example we will follow with an example that is slightly
closer to reality.
For the abstract example we choose an Hilbert space with 2 electrons
occupying 4 different orthonormal one particle states labeled by A, B, C,

295
Chapter 9. N particle systems

and D. The two particle basis B of the Hilbert space is given as

n o
B = a†A a†B |0i , a†A a†C |0i , a†A a†D |0i , a†B a†C |0i , a†B a†D |0i , a†C a†D |0i ,
(9.41)

and has dimension 6. For a given two particle state we can calculate the
occupation of the four different one particle states with the operator a†τ aτ ,
with τ ∈ {A, B, C, D}. In matrix form on basis B these operators are
   
1 0 0 0 0 0 1 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0 0 0
   
† 0 0 1 0 0 0 † 0 0 0 0 0 0
aA aA = 
0
 aB aB =   (9.42)
 0 0 0 0 0
0
 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 1 0
0 0 0 0 0 0 0 0 0 0 0 0
   
0 0 0 0 0 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0 0 0
   
† 0 0 0 0 0 0 † 0 0 1 0 0 0
aC aC = 
0
 aD aD =  .
 0 0 1 0 0
0
 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 1 0
0 0 0 0 0 1 0 0 0 0 0 1

From the matrix representation we can directly see that the eigen values of
the operators a†τ aτ are 0 and 1. Following section 6.5 we thus know that
in an experiment where one measures the occupation of state τ such that
a single number is the outcome of such an experiment this value can either
be 0 or 1. The frequency with wich one finds these values, if one repeats
the experiment very often, is determined by the wave function.
We now use the fact that we can represent a solution of the Schrödinger
equation as a superposition of other solutions of the Schrödinger equation.
We consider the following two states which we assume to be solutions of
the Schrödinger equation

|ψ1 i = a†A a†B |0i (9.43)

296
 9.3. Entanglement and correlations

|ψ2 i = a†C a†D |0i

or as wave functions
p
ψ1 (~r1 , ~r2 ) = 1/2(ψA (~r1 )ψB (~r2 ) − ψB (~r1 )ψA (~r2 )) (9.44)
p
ψ2 (~r1 , ~r2 ) = 1/2(ψC (~r1 )ψD (~r2 ) − ψD (~r1 )ψC (~r2 )).

We can build a linear combination of these states such that

r 
1 
|ψi = a†A a†B + a†C a†D |0i , (9.45)
2
or function
1
ψ(~r1 , ~r2 ) = ψA (~r1 )ψB (~r2 ) − ψB (~r1 )ψA (~r2 ) (9.46)
2

+ ψC (~r1 )ψD (~r2 ) − ψD (~r1 )ψC (~r2 ) ,

is also a solution of the Schrödinger equation. For state |ψi we find that
the expectation value for the occupation of each of the one particle states
is 1/2, i.e.
D E 1
ψ a†τ aτ ψ = , (9.47)
2
for τ ∈ {A, B, C, D}. With the combined knowledge that for a single mea-
surement we either find the value 0 or 1 we know that in 50% of all cases we
will find the value 0 and in 50% of the cases the value 1 for the occupation
of the state A, B, C and D.
There is a strong correlation between the occupation we find for the
different one particle states. We can have a look at the operator a†A aA a†B aB
and the operator a†A aA a†C aC . If we calculate the expectation value of these
operators for the state |ψi we find
D E 1
ψ a†A aA a†B aB ψ = , (9.48)
D E 2
ψ a†A aA a†C aC ψ = 0.

297
Chapter 9. N particle systems

From these expectation values we can deduce that in an experiment with

a single particle we will either find one particle states A and B occupied
or one particle states C and D occupied. If we do a measurement and find
that A is occupied we can immediately deduce that also B will be occupied.
These correlations between the outcome of different experiments is known
as entanglement.
Now let’s have a look at a more realistic example. For this example we
will still only sketch the physics we want to discribe instead of looking at
real solutions. The later we can do numerically, but that does not provide
the insight we want to convey. We will take the example of two particles,
each with spin 1/2 bound together in a singlet state. We then assume that
at some point in time these two particles disconnect and the two particles
propagate in different directions. We want to look at the spin of the different
particles, after they propagated far away from each other.
Just as in the abstract example we will define a Hilbert space consisting
of four one particle states that we fill in different ways with two electrons.
The one particle states are Gaussian wave packets with momentum k(t)
and −k(t) as given in equation 4.30. At t = 0 we assume the wave packets
to be centred around x = 0. For t ≤ 0 we assume the momentum k(t)
to be zero and for t  0 we assume a constant positive value. The exact
acceleration of the particles is of no importance for our model. Each of
these wave functions can either represent a state with spin up or spin down
such that the one particle wave packets are given as ψk↑ (x),ψ−k↑ (x),ψk↓ (x),
and ψ−k↓ (x) or equivalent as |k ↑i, |−k ↑i, |k ↓i, and |−k ↓i. For t ≤ 0 we
find that k = −k = 0. At k = 0 our one particle states are not orthogonal
to each other. This is inconvenient, but not a real problem.
We want the particles to be in the singlet state. Ignoring the fact that
our basis is not always orthonormal we can define the spin operators as

1  
S z = ~ a†k↑ ak↑ − a†k↓ ak↓ + a†−k↑ a−k↑ − a†−k↓ a−k↓ (9.49)
2
S + = a†k↑ ak↓ + a†−k↑ a−k↓
S − = a†k↓ ak↑ + a†−k↓ a−k↑

298
 9.3. Entanglement and correlations

and the standard relations S x = 1/2(S + + S − ), S y = −i/2(S + − S − ), and

S2 = S~ · S.
~ For t  0 where the basis orbitals are orthonormal the singlet
state with zero total momentum is given as
r 
1 
|ψi = a†k↑ a†−k↓ − a†k↓ a†−k↑ |0i , (9.50)
2
or equivalently
1
ψ(t, x1 , x2 ) = ψk↑ (t, x1 )ψ−k↓ (t, x2 ) − ψk↑ (t, x2 )ψ−k↓ (t, x1 ) (9.51)
2

− ψk↓ (t, x1 )ψ−k↑ (t, x2 ) + ψk↓ (t, x2 )ψ−k↑ (t, x1 ) .

For t ≤ 0 where k = −k = 0 we find the state to be,

r 
1 
|ψi = a†0↑ a†0↓ − a†0↓ a†0↑ |0i (9.52)
2
r 
1 
= a†0↑ a†0↓ + a†0↑ a†0↓ |0i
2
√ † †
= 2a0↑ a0↓ |0i ,
√
which is besides a normalization error of a factor of 2 the correct singlet
state for two particles occupying the same local orbital.
The physics these wave functions describe is, for example, a nuclear
decay of a singlet nucleus emitting an electron with spin 1/2 and a daughter
nucleus with spin 1/2. Before the decay these two particles are in a singlet
state t ≤ 0. As the spin angular momentum is conserved we find that after
decay these two particles are still in a singlet state. This yields, just as in the
case of our abstract example with single particle states an entangled state.
For these states the measurement of the spin of one particle is correlated
with the value one obtains when the other particle is measured. One can
map these two examples on each other if we identify the state A as k ↑, B
as −k ↓, C as −k ↑ and D as k ↓.
Entanglement and the resulting correlations between electrons is that
what makes quantum physics so different from classical physics. It is a

299
Chapter 9. N particle systems

direct consequence of the linearity of the Schrödinger equation, which allows

the sum of different solutions to be a solution by it-selves. Particles entangle
when they interact with each other. In the previous example we started
with two particles locally bound in a singlet state. Even if there is no
direct interaction term in the Hamiltonian, for two particles bound in the
same spacial quantum state, the Pauli principle requires that their spin is
different.
Entanglement does not only make quantum computations involved and
scale badly in complexity with particle number, it also opens up possibilities
for quantum computation.

9.4 Temperature
We have seen in the early chapters of this script that a particle in motion
can be described by a wave packet. For the Hydrogen atom we used the
eigenstates of the Hamiltonian with negative energy as the bound states.
The eigenstates of the Hamiltonian are stationary such that the states are
stable over time. Being able to describe stable states with angular momen-
tum for the electron for the hydrogen atom, solved a big internal problem
of the Bohr model.
What we have not discussed is temperature. There are currently two
competing ways to model temperature. On a macroscopic scale both de-
scriptions yield the same result. We will discuss both theoretical models.
Based on some general argumentation we can derive what the macroscopic
limit for averaged values in a system with finite temperature needs to be.
(See also the section on Black body radiation 1.1 and Einsteins model for
the heat capacity 1.3.) Once we know the macroscopic averaged limit we
can discuss the two different microscopic mechanisms to realize the statis-
tical macroscopic state.
We find a correct description of systems at finite temperature if we use
Boltzmann statistics. Boltzmann statistics tells one what the time averaged
probability is to find a system in a state with energy E at temperature T .
For Boltzmann statistics to be applicable we need to assume that the system

300
 9.4. Temperature

is in thermal equilibrium with a bath. The coupling of the system to a bath

allows one to exchange energy (heat) between the system and the bath, i.e.
a large reservoir. As such, the energy of the system at finite temperature
is a function of time E(t) due to the heat exchange between system and
bath. Boltzmann statistics tells one what the time averaged energy is
Rt
−t E(t)dt
hEit = lim Rt . (9.53)
t→∞
−t dt

In this equation we used brackets to indicate the average over time. This
yields the thermodynamical properties of a system. We realize that in a
book on quantum mechanics this notation can be confusing. We use the
same bracket notation for inner products of vectors in a complex Hilbert
space within the Dirac notation. As will become clear later in the lecture or
future lectures there are clear parallels between the Dirac bracket and the
temperature averaging. There is also a major difference. The Dirac bracket
integrates over states in a complex vector space. The temperature bracket
integrates over time. It is important to distinguish the difference between
temperature averages and quantum averages over distribution functions.
We choose to add the index t for temperature averaging, indicating that
one averages over fluctuations in time (contrary to building an average
over quantum fluctuations in configurational space. More on that later in
this lecture and other lectures on Green’s / correlations functions or path
integrals).

9.4.1 Classical statistical physics for the ideal gas

We can justify the use of the Boltzmann equation by looking at the be-
haviour of a classical ideal gas. The energy of a single atom in a classical
ideal gas will fluctuate over time. If we focus on a single particle in such
a gas then sometimes the particle will move faster, sometimes slower. The
energy distribution for a single particle in an ideal gas is in this case de-
termined by the velocity distribution. Or better said, by the norm square
of the velocity distribution ~v 2 . In one of your earlier statistical physics

301
Chapter 9. N particle systems

lectures we defined

p(vx , vy , vz )dvx dvy dvz (9.54)

the probability distribution to find an atom with a velocity in the range

between (vx , vy , vz ) and (vx + dvx , vy + dvy , vz + dvz ). For an ideal gas the
energy in the different Cartesian directions is conserved. Furthermore the
different directions need to have the same energy distribution. We thus
know that

p(~v 2 ) = p(vx2 + vy2 + vz2 ) (9.55)

= p(vx2 )p(vy2 )p(vz2 ).

The function that fulfils this requirement is the exponential function. We

find
1 −a~v2
p(~v 2 ) = e . (9.56)
Z
The function p(~v 2 ) represents a Gaussian distribution for the energies of
the atoms in an ideal gas. The average velocity of each particle is zero.
This should be the case as we assume the total volume of our gas to be at
rest. The factor Z can be used to normalize the probability distribution.
The function p represents a probability such that the integral over p should
be 1, therefor 1/Z = (a/π)3/2 , such that
Z ∞Z ∞Z ∞
1 −a~v2
e dvx dvy dvz = 1. (9.57)
−∞ −∞ −∞ Z

For this integral we assumed that a > 0, otherwise the function would not
be normalizable.
We can relate a to the temperature of the system. We define temper-
ature in such a way that the average energy of an atom in an ideal gas is

3
hEit = kB T, (9.58)
2

302
 9.4. Temperature

with kB the Boltzmann constant. We can also relate the average energy to
the average velocity square. For a particle with mass m we find
m 2 
hEit = vx t + vy2 t + vz2 t (9.59)
2
3
= m vx2 t .
2

With the probability to find the particle with the velocity ~v 2 given by p we
can calculate vx2 t as

∞
Z r
a −av~x 2
vx2 t = vx2 e (9.60)
−∞ π
1
= .
2a
Combining the two expressions for hEit yields
m
a= . (9.61)
2kB T

The velocity distribution becomes

1 − 2km~v2T
p(~v 2 ) = e B . (9.62)
Z
Instead of making the probability a function of the particles velocity we can
look at the probability to find the particle with energy E = 21 m~v 2 . This
yields

1 − k ET
p(E) = e B . (9.63)
Z

9.4.2 Boltzmann statistics in quantum mechanics

We now generalize the classical Boltzmann statistics to quantum systems. If
|ni is an eigenstate of the Hamiltonian with energy En and our sample is in

303
Chapter 9. N particle systems

thermal equilibrium with a heat bath at temperature T then the probability

to find the system in state |ni is given as
1 − kEnT
pn = e B . (9.64)
Z
The normalization Z is given by a sum over all eigenstates of the Hamilto-
nian
X − En
Z= e kB T . (9.65)
n

If we want to calculate the temperature dependent expectation value of

an operator we find
X
hoit = pn hψn |O|ψn i , (9.66)
n

whereby |ψn i are the eigenstates of the Hamiltonian with eigenenergy En .

The temperature dependent energy is
X
hEit = pn hψn |H|ψn i . (9.67)
n

The heat capacity of a system is given by the temperature derivative of the

temperature dependent energy.
There are two possible competing microscopic models that can lead
to this behaviour. The first is related to the concept of reduced density
matrices, the second assumes the total micro state of the system is given
by a single wave function. The second model describes the time evolution in
all its detail and as such describes thermal fluctuations, Brownian motion
or thermal noise. It is however not completely clear how the second model
can lead to thermalization.

304
 9.4. Temperature

9.4.3 Reduced density matrix

In order to equilibrate the temperature of a system we can place the system
into contact with a heat bath. We take |αn i to be the eigenstates of our
system without a heat bath. The states |αn i form a basis for a Hilbert
space on which all possible states of the system can be described. We can
do the same for the heat bath. The states |βm i are the eigenstates of the
heat bath and form a basis for the Hilbert space on which all possible states
in the heat bath can be described. In order to thermalize the system we
couple the system to the heat bath. We can assume that the coupling is
weak. In the limit of infinitesimal weak coupling the eigenstates of the
combined system with heat bath are given as tensor product states of the
eigenstates of the system and the eigenstates of the heat bath

|ψn,m i = |αn i ⊗ |βm i (9.68)

We are interested in properties of our system. For this we will look at

operators O that act on the system, but not on the heat bath. We find
for the expectation value of operator O for an eigenstate of the combined
system plus heat bath

ψn0 ,m0 O ψn,m = δm0 ,m hαn0 |O|αn i . (9.69)

A system in thermal equilibrium is not stationary. Atoms in a classical

ideal gas move, energy fluctuates as a function of time between the sys-
tem and the heat bath. We can however write the thermal state |ψi as a
superposition of the eigenstates of the system plus heat bath
X
|ψi = cn,m |ψn,m i , (9.70)
n,m

with cn,m some set of complex numbers whose exact values determine the
system, but are irrelevant for the current discussion. The coefficients cn,m
contain the information on how much the product of system state |αn i and
bath state |βm i contributes to the thermal state |ψi. If we calculate an

305
Chapter 9. N particle systems

expectation value of the thermal state we find

X
hψ|O|ψi = c∗n0 ,m0 cn,m ψn0 ,m0 O ψn,m (9.71)
n,m,n0 ,m0
X
= c∗n0 ,m0 cn,m δm0 ,m hαn0 |O|αn i
n,m,n0 ,m0
X
= c∗n0 ,m cn,m hαn0 |O|αn i .
n,m,n0

The sum where n 6= n0 relates to the case where two different system
eigenstates |αn i and |αn0 i are multiplied with the same bath state |βm i.
If we assume that the bath
P to∗ be large and randomly entangled with the
system of interest then 0
m cn0 ,m cn,m becomes zero unless n = n. We
retrieve the Boltzmann equation if we choose
X
c∗n,m cn,m = pn (9.72)
m
1 − kEnT
= e B .
Z
A quantum system obtains Boltzmann statistics for its local expectation
values when the system entangles with a heat bath. The local system is in
a quantum super position of different eigenstates. Each term in this super
position is represented with a weight given by Boltzmann statistics. Each
term in this superposition is entangled with different states of the heat
bath, such that there is no interference between the different terms if one
looks at expectation values of operators acting on the system. The thermal
average one makes in this description of temperature is not over time, but
is an average over different configurations of local states that entangle with
states of the heat bath. This microscopic mechanism for temperature is not
able to describe Brownian motion, thermal noise and thermal fluctuations.
This microscopic model to describe temperature does allow for non thermal
states to thermalize.

306
 9.4. Temperature

9.4.4 Brownian motion, thermal noise and thermal fluctua-

tions
Instead of requesting a heat bath to thermalize a system we can define a
wave function that has a time evolution, such that the time average results
in a Boltzmann average. For this to work we need to require that the system
has no degenerate states. We assume an Hamiltonian with eigenstates |ψn i
and eigenenergies En . These states form an orthonormal basis for the
Hilbert space of our problem. The claim is that the wave function
X√
|ψi = pn |ψn i (9.73)
n
− 2kEnT
X
= e B |ψn i ,
n

describes a state whose time averaged expectation values follow Boltzmann

statistics. The time dependent state has the form
X√
|ψ(t)i = pn e−iHt/~ |ψn i (9.74)
n
X√
= pn e−iEn t/~ |ψn i .
n

If we now calculate an expectation value we find

X √ √
hψ(t)|O|ψ(t)i = e−i(En −En0 )t/~ pn0 pn hψn0 |O|ψn i . (9.75)
n,n0

The time average expectation value is

X 1 Z t √ √
hψ(t)|O|ψ(t)it = lim e−i(En −En0 )t/~ dt pn0 pn ψn0 O ψn
t→∞
0
2t −t
n,n
(9.76)
X √ √
= δn,n0 pn0 pn hψn0 |O|ψn i
n,n0

307
Chapter 9. N particle systems

X
= pn hψn |O|ψn i .
n

A thermal state defined as a single state, obeys Boltzmann statistics

once time averaged. At a specific time the expectation values are not
perfectly thermal, but show fluctuations around the thermal average as
a function of time. As such this description of temperature naturally in-
cludes thermal noise and would be able to describe Brownian motion. It is
however not clear how one would reach such a state. There is no heat bath
with which energy is exchanged to thermalize the state.

308
Chapter 10

Perturbation theory

For many interesting cases in quantum mechanics, we are not able to find
analytical solutions to the Schrödinger equation. This is not only true for
the complex many-particle systems but even for a single particle traveling
through a potential. The number of potentials for which we have analytical
solutions of the Schrödinger equation is rather limited. Sadly enough we
have already discussed a large part of the analytically solvable problems.
With the use of numerical methods, one is able to solve with good accuracy
basically all single-particle problems of interest, if the Hamiltonian is time-
independent. Time-dependent potentials can lead to numerical instabilities
in the differential equation that need to be solved. With some attention one
normally can find numerical solutions for a single particle traveling through
a time-dependent potential. For many-particle systems we often do not have
analytical, nor numerically accurate solutions. Instead of trying to find
exact analytical solutions or numerical accurate solutions one can look at
approximate solutions. Perturbation theory is an approximate analytical
method or series expansion that can help to find solutions to otherwise
intractable problems. Even for systems that can be solved numerically,
perturbation theory can help to give insight into the physical reason why
and how things happen.
The general idea behind perturbation theory is that we split the Hamil-

309
Chapter 10. Time independent perturbation ...

tonian (H) of a system that does not have a simple solution into a part
that we can solve (H 0 ) and a part that we treat as a correction to the
solutions of the solvable part (H 1 ). We then make a series expansion of
the true solution in the perturbing Hamiltonian. We can use perturbation
theory for example if we want to calculate the eigenstates of the hydrogen
atom in an external electromagnetic field. We can first solve the solutions
of the Schrödinger equation of the hydrogen atom without an external elec-
tromagnetic field and then treat the field as a perturbation to the solutions
without a field. The same is true when we look at relativistic corrections.
For most problems we can first solve the Schrödinger equation and then
treat the relativistic effects as captured in the Dirac equation or by quan-
tum electrodynamics as a perturbation. Even for simpler systems such
as the Harmonic oscillator one often finds that these only approximately
resemble experimental situations and one needs to take anharmonic per-
turbations into account.

10.1 Time independent perturbation theory with-

out degenerate states
∂
Given a general Hamiltonian that is time independent, such that ∂t H = 0.
We are interested in the eigenstates (|φn i) and eigenenergies (En ) of H
with

H |φn i = En |φn i , (10.1)

but have no method to analytically solve this equation. In order to use

perturbation theory we define

H = H 0 + H 1, (10.2)

with
E E
H 0 φ(0)
n = En(0) φ(0)
n , (10.3)

310
an eigenvalue equation that we know how to solve. Within perturbation
theory we will express the eigenstates and eigenenergies of H using a power
series, or Taylor series, expansion in H 1 around the point H 0 .
We will for now assume that H 0 has no degenerate eigenstates, i.e.

En(0) 6= Em
(0)
∀n, m 6= n. (10.4)

Whereby we could be slightly less strict as it suffice to require

E that we
(0)
are only interested in the eigenstates |φn i obtained from φn for which
(0) (0)
En 6= Em ∀m 6= n. This requirement is needed as we will divide by
(0) (0)
En − Em at some point in the derivation. The case with degeneracies will
be discussed in the next section. In order to make a Taylor series expansion
of |φn i and En we need to know how they change if we change H 1 . We
∂
are thus interested in something of the form ∂H 1
|φn i. We can prevent the
need to define the derivative of a state vector with respect to an operator
by writing

H λ = H 0 + λH 1 , (10.5)

with

H λ |φλn i = Eλn |φλn i , (10.6)

defining the λ dependent eigenstates |φλn i and eigenenergies Eλn . Instead

of making an expansion in H 1 we can now make the expansion in terms of
powers of λ. As the λ dependence of H λ only enters as λH 1 we will find
that a power series in λ also yields a power series in H 1 . We will make the
expansion around the point λ = 0 and need to evaluate the series in the
point λ = 1 to get the approximation of the original problem. Given that
both Eλn and |φλn i are functions of λ we find their Taylor series in powers
of λ to be:
∞
X
Eλn = λi En(i) (10.7)
i=0

311
Chapter 10. Time independent perturbation ...

∞
X E
|φλn i = λi ϕ(i)
n ,
i=0
E
(i) (i)
with En and ϕn the expansion coefficients we need to determine. For
λ = 1 we find
∞
X
En = En(i) (10.8)
i=0
X∞ E
|φn i = ϕ(i)
n .
i=0
E
(i) (i)
As En and ϕn appear as the ith order in λ they also appear as the ith
E
(i) (i)
order in H 1 in the expansion. The energy En and state ϕn are thus
the ith order corrections to the energy and state in powers of H 1 . Given
that equation 10.7 can be calculated as a Taylor series expansion we find
that the ith order corrections to the energy and to the state vector of the
nth eigenstate are

1 ∂i
En(i) = Eλn (10.9)
i! ∂λi λ=0
E 1 ∂i
ϕ(i)
n = |φλn i
i! ∂λi λ=0

A Taylor series has a range of convergence and this is important to re-

member. We assume that Eλ,n and |φλ,n i are analytical functions of λ for
|λ| ≤ 1. This condition is often phrased in a statement that H 1 needs to
be weak in comparison to H 0 for a perturbation series to converge. There
are no simple generally valid hard criteria that determine if a perturbation
series is convergent. For realistic cases convergence is by far not always
given. An example will be discussed at the end of this section.
The 0th order term in the Taylor series expansion is given by the
eigenenergies and eigenstates evaluated for the case where λ = 0. We

312
find that

En(0) = En(0) (10.10)

E E
ϕ(0)
n = φ (0)
n .
D E
(0) (0)
The 0th order states form an orthonormal set ϕn ϕm = δn,m . For a
E
(i)
general i we find that the states ϕn are not normalized nor orthonormal
to each other
D E
ϕ(i)
n ϕ (j)
n = Sn(i,j) . (10.11)
D E
(i) (0)
In particular there always is an i for which ϕn ϕn is non-zero un-
E E
(0) P∞ (i)
less ϕn = |φn i. This follows from the fact that |φn i = i=0 ϕn ,
D E P D (i) (j) E D E
(0) (0) (i) (i)
ϕn ϕn = 1, hφn |φn i = i,j ϕn ϕn = 1, and ϕn ϕn ≥ 0.
Before we focus on the calculation of the expansion terms we want
to slightly rewrite the Taylor series. It is useful to modify the expansion
slightly such that at least the 0th order expansion states are orthogonal to
all higher order expansion states. We will do this at the expanse of loosing
the normalization of the final solution. This is not a big problem as we can
always normalize the states afterwards. We define for i > 0
E E D E E
φ̃(i)
n = ϕ(i)n − φ(0)
n ϕn
(i)
φ(0)
n , (10.12)

such that for i > 0

D E
φ(0)
n φ̃ (i)
n = 0. (10.13)

For the eigenstates of H λ we find

∞
X E
|φλn i = λi ϕ(i)
n (10.14)
i=0

313
Chapter 10. Time independent perturbation ...

E X∞ E
= φ(0)
n + λ i
ϕ (i)
n
i=1
∞ ∞
!
X D E E X E
i
= λ φ(0)
n ϕ(i)
n φ(0)
n + λ i
φ̃ (i)
n .
i=0 i=1
D E
P∞ (0) (i)
We now divide |φλn i by i=0 λi φn ϕn . We will call the new, unnor-
malized state |ψλn i. We find
1
|ψλn i = P D E |φλn i (10.15)
∞ (0) (i)
i=0 λi φn ϕn
∞
E X λi E
= φ(0)
n + D E φ̃(i)
n .
j φ(0) ϕ(j)
P∞
i=1 j=0 λ n n

The second term in equation 10.15 is some function of λ. We can make a

Taylor series expansion of this function and find
E X ∞ E
|ψλn i = φ(0)
n + λ i
φ (i)
n . (10.16)
i=1
E E
(i) (i)
The relation between φn and φ̃n is non-trivial. Luckily we do not need
to know this relation to make the perturbation expansion work. The norm
of |ψλn i has a special meaning and defines the overlap of the n-th eigenstate
of H with the nth eigenstate of H 0 . If we define |ψn i = |ψλn i |λ=1 we have

D E
φ(0)
n ψn = 1, (10.17)

1
|φn i = p |ψn i , (10.18)
hψn |ψn i
and thus
D E 1
φ(0)
n φ n = p . (10.19)
hψn |ψn i

314
 With these definitions we are able to derive the perturbation expansion.
We insert the expansion for the wave function |ψλn i in powers of λ as given
in equation 10.16 and the energy expanded in powers of λ as given in
equation 10.7, into the eigenvalue equation for H λ (equation 10.6) and find

∞
X E X∞ X
∞ E
(H 0 + λH 1 ) λi φ(i)
n = λ i j (j)
λ En φ (i)
n . (10.20)
i=0 i=0 j=0

Rewriting this to collect equal powers in λ yields

∞
X  E E X∞ X
i E
λi H 0 φ(i)
n + H 1 φ (i−1)
n = λ i (i−j)
En φ (j)
n . (10.21)
i=0 i=0 j=0

As this equation needs to hold for all values of λ we find

E E i
X E
H 0 φ(i)
n + H1 φ(i−1)
n = En(i−j) φ(j)
n . (10.22)
j=0

Reordering terms gives the recurrence relation

  E Xi−1 E E
H 0 − En(0) φ(i)
n = En
(i−j)
φ (j)
n − H 1 φ (i−1)
n . (10.23)
j=0

We can express the ith order correction term to the wave function as a sum
over the 0th to i − 1th terms.
(i)
In order to calculate the ithDeigenenergy correction
D En weD multiply
(0) (0) (0) (0)
equation 10.23 from the left with φn . Using that φn H 0 = φn En
we find
i−1
X D E D E
0= En(i−j) φ(0)
n φn
(j)
− φ(0) (i−1)
n H1 φn . (10.24)
j=0

315
Chapter 10. Time independent perturbation ...

D E E
(0) (j) (j)
We now use that φn φn = δj,0 as we orthogonalized all states φn
E
(0)
to φn to find
D E
 φ(0) (0)
n H 0 φn i=0
En(i) = D (0) E . (10.25)
 φn H 1 φ(i−1)
n i>0

The ith order correction to the energy is given by the matrix element of H 1
between the i − 1th correction to the state and the 0th order state around
which we make the perturbative E series expansion.
(i)
In order to determine φn we return to equation 10.23 but now mul-
D
(0)
tiply it from the left with φm such that we obtain

D E Xi−1 D E D E
φ(0)
m H 0 − En
(0) (i)
φ n = En
(i−j)
φ (0) (j)
m φ n − φ (0)
m H 1 φ (i−1)
n . (10.26)
j=0
D D
(0) (0) (0)
We use that φm H 0 = φn Em to get

D E i−1
X D E D E
(0)
(Em − En(0) ) φ(0)
m φ(i)
n = En(i−j) φ(0)
m φn
(j)
− φ(0) (i−1)
m H 1 φn .
j=0
(10.27)
E
(0)
As φn forms a complete basis set (Eigenstates of H 0 ) we can write
E X ED E
φ(i6
n
=0)
= φ (0)
m φ (0) (i)
m φ n . (10.28)
m6=n
D E
(0) (i)
We can remove the case where m = n from the sum as φn φn = 0 for
(0)
i 6= 0. Next we use the fact that our state of interest φn is non degenerate
(0) (0) (0) (0)
such that En 6= Em for all m 6= n and divide equation 10.27 by En −Em .

316
 E E
(i) (0)
With the combination of the expansion of φn on φm (equation 10.28)
we find
 
i−1
E X E 1 D E X E
φ(i)
n = φ(0)
m (0) (0)
φ(0)
m
H 1 φ(i−1)
n − En(i−j) φ(j)
n
.
m6=n En − Em j=0

(10.29)
This equation expresses the ith order corrections to the perturbative state
(0)
around φn in terms of the eigenenergy E and eigenstate correctionsEof lower
(i) (i)
order. Once we obtain the states φn we can sum the states φn to find
the unnormalized eigenstates of H
∞
X E
|ψn i = φ(i)
n . (10.30)
i=0

We can normalize the states |ψn i

1
|φn i = p |ψn i , (10.31)
hψn |ψn i
to getE the normalized eigenstates of H, |φn i in terms of the eigenstates,
(0) (0)
φn , and energies, En , of H 0 .
To summarize: The 0th order eigenstate and energy of H are the eigen-
state and energy of H 0 given by the equation
E E
H 0 φ(0)
n = E (0)
n φ (0)
n . (10.32)

The first order perturbative corrections to are:

D E
En(1) = φ(0)
n H 1 φ (0)
n (10.33)

for the energy and

D E
E X E φ(0) (0)
m H 1 φn
φ(1)
n = φ(0)
m (0) (0)
(10.34)
m6=n En − Em

317
Chapter 10. Time independent perturbation ...

for the state. The second order perturbative corrections are:

(0) 2
D E
(0)
X φ m H φ
1 n
En(2) = (0) (0)
, (10.35)
m6=n E n − Em

for the energy and

D ED E
E X X E φ(0)m2 H 1 φm1
(0) (0)
φm 1 H 1 φn
(0)

φ(2)
n = φ(0)
m2    (10.36)
(0) (0) (0) (0)
m1 6=n m2 6=n En − E m2 E n − Em1
D ED E
X E φ(0)
m H 1 φn
(0) (0)
φn H 1 φn
(0)
(0)
− φm   
(0) (0) (0) (0)
m6=n En − Em En − Em

for the state. The third order energy correction is

D ED ED E
(0) (0) (0) (0) (0) (0)
X X φn H 1 φm2 φm2 H 1 φm1 φm1 H 1 φn
En(3) =    (10.37)
(0) (0) (0) (0)
m2 6=n m1 6=n En − Em2 En − Em1
(0) 2
D E
D E X φ(0) m H 1 φn
− φ(0)n H 1 φn
(0)
 .
(0) 2

(0)
m6=n En − Em

There are two very important conclusions to draw from these equations.
The first conclusion is that ”levels repel each other” when additional inter-
actions are added. If we look at equation 10.35 for the second-order energy
correction then we find that for a state |φn i the states with an energy larger
(0)
than En will lead to a lowering of the energy and states with an energy
(0)
smaller than En will lead to an increase of the state energy. The second
conclusion immediately follows from the first. For the groundstate there
are only states with larger energies in the Hilbert space. We thus find that
the ground-state energy including the 1-st order perturbative correction is
higher than the true ground-state energy.

318
10.1.1 Perturbation example on two-level system
The first example for perturbation theory we discuss uses a model E system E
(0) (0)
with two internal states. As a basis we take the states φ1 and φ2 .
D E D E
(0) (0) (0) (0)
These states have an energy φ1 H φ1 = E1 and φ2 H φ2 =
D E
(0) (0)
E2 > E1 . Furthermore there is a coupling φ1 H φ2 = t between
them. The model described here could be a magnetic atom with a doublet
state (Silver atom) in a strong magnetic field in the z direction responsible
for the energies E1 and E2 and a small field in the x direction responsible
for the interaction t. The full Hamiltonian is
D 
(0)

(0)
E
(0)
E φ 1
H = φ1 φ2 · H · D (0)  , (10.38)
φ2

with H the matrix

!
(0)
E1 t
H= (0) . (10.39)
t E2

In order to make a perturbation series we take

(0)
H = E1 1 + H0 + H1 , (10.40)

with
 
0 0
H0 = , (10.41)
0 ∆
(0) (0)
∆ = E2 − E1 , and
 
0 1
H1 = t . (10.42)
1 0

We introduced a shift of all energies by E1 and introduced ∆ > 0 as the

energy difference between the unperturbed first excited state and ground

319
Chapter 10. Time independent perturbation ...

state. This will simplify the notation later on. In principle we do not need
to do perturbation theory to find the eigenstates and eigenenergies of H.
We can solve this problem exactly and find
(0) 1 p 
E1 = E1 + ∆ − ∆2 + 4t2 (10.43)
2 r
 
4t2
t √
2 +4t2 2
+1
 1 (∆+ ∆ ) 
|φ1 i =  r , − √ 
 4t2 ∆ 2 + 4t2 
√
2 +4t2 2
+1
(∆+ ∆ )
(0) 1  p
2 2

E2 = E1 + ∆ + ∆ + 4t
2
 r 
4t2
t √ 2 + 1
 (∆+ ∆2 +4t2 ) 1 
|φ2 i =  √ ,r .
 ∆2 + 4t2 4t2

√ 2 + 1
(∆+ ∆ +4t )
2 2

We can use these exact results to compare to the results we obtain in

perturbation theory. In order to facilitate the comparison, we first make
a Taylor series expansion of Ei and |φi i. We note that H1 scales linearly
in t. Instead of introducing Hλ = H0 + λH1 and making a Taylor series
expansion in λ we can equivalently make a Taylor series expansion in t. For
the energy and wave functions, we find

(0) t2 t4 2t6
+ 3 − 5 + O t8


E1 = E1 − (10.44)
∆ ∆ ∆
       
1 2 1 3 3
|φ1 i = 1, 0 + t 0, − +t − , 0 + t 0, (10.45)
∆ 2∆2 2∆3
   
11 31
+ t4 , 0 + t5 0, − 5 + O t6


8∆ 4 8∆

(0) t2 t4 2t6
− 3 + 5 + O t8


E2 = E2 + (10.46)
∆ ∆ ∆

320
        
1 2 1 3 3
|φ2 i = 0, 1 + t , 0 + t 0, − 2 + t − ,0 (10.47)
∆ 2∆ 2∆3
   
11 31
+ t4 0, + t5 , 0 + O t6 .


8∆ 4 8∆ 5

The different termsE for the states in equation 10.45 and 10.47 are equal to
(i)
the states ϕn as defined in equation 10.7. The order i of the expansion
E
(i)
is given by the power of t, i.e. ϕn is the term proportional to ti . The
E
(i)
states ϕn are not orthonormal to each other. There are infinitely many
E
(i)
states ϕn but the dimension of the Hilbert space is only 2.
With the use of perturbation theory, i.e. following equation 10.29 and
10.25 we find

(0) t2 t4 2t6
+ 3 − 5 + O t8


E1 = E1 − (10.48)
∆ ∆ ∆

       
1 2 3 1
|ψ1 i = 1, 0 + t 0, − + t 0, 0 + t 0, 3 (10.49)
∆ ∆
   
2
+ t4 0, 0 + t5 0, − 5 + O(t6 ).
∆

(0) t2 t4 2t6
− 3 + 5 + O t8


E2 = E2 + (10.50)
∆ ∆ ∆

       
1 2 3 1
|ψ2 i = 0, 1 + t , 0 + t 0, 0 + t − 3 , 0 (10.51)
∆ ∆
   
2
+ t4 0, 0 + t5 , 0 + O(t6 ).
∆5

The different terms

E for the states in equation 10.49 and 10.51 are equal to
(i)
the states φn as defined in equation 10.16. The order i of the expansion

321
Chapter 10. Time independent perturbation ...

E
(i)
is given by the power of t, i.e. φn is the term proportional to ti . The
E E
(i) (0)
states φn for i > 0 are orthogonal to the state φn . For i > 0 and
E E
(i) (j)
j > 0 the states φn and φn are not orthogonal to each other. Actually
they are all linear dependent on each other as the E dimension of the Hilbert
(0)
space is 2 and they need to be orthogonal to φn .
|φn i and |ψn i as determined from a Taylor series or perturbation series
represent the same state. They differ in normalization such that hφn |φn i =
1 and
4t2
hψn |ψn i =  √ 2 + 1. (10.52)
∆+ ∆2 + 4t2

Equivalently we can write for the overlap with the unperturbed state
D E
ψn φ(0)
n = 1, (10.53)

and
D E 1
φn φ(0)
n =r . (10.54)
4t2
√ 2 +1
(∆+ ∆2 +4t2 )

Range of convergence
We now discuss a very important point of perturbation theory. The range
of convergence. For which values of t and ∆ is the above series expansion
valid. For that we can have a look at the Taylor series expansion of E1 =
(0) √
E1 + 21 ∆ − ∆2 + 4t2 . A seemingly harmless function that is smooth
and infinitely differentiable on the real axis. For the convergence of a Taylor
series expansion we need to look for poles or branch cuts in the complex
√
plain. The function z z ∈ C has a branch cut for z = x with x ∈ R and
x < 0. For x < 0 we find
√ p p
i −x = lim x + iy = − lim x − iy. (10.55)
y→0+ y→0+

322
Figure 10.1: Real (left) and imaginary (right) part of the ground-state
energy of a two level system as given in equation 10.43 for ∆ = 1 and
t = x + iy. In blue for the exact solution and in yellow for a 20th order
perturbation theory. The range of convergence is t = 21 .

For t = iτ we hit the branch cut of the square root when τ > ∆/2. The
Taylor series and thus also the perturbation expansion has a range of con-
vergence for t of ∆/2. The above equations are valid when

∆
t≤ . (10.56)
2
In figure 10.1 we show the energy of the two level ground-state 21 (∆ −
√
∆2 + 4t2 ) as well as the 20th order perturbation theory approximation to
the energy as a function of t in the complex plane. Although the function
is well behaved on the real axis the branch cut for |Im(t)| > 12 yields a finite
convergence radius, also on the real axis.
In general one can state that for a perturbation series to be valid the
coupling strength between two states needs to be small compared to the
energy between these states. There are many interesting cases where this
is not realized.

323
Chapter 10. Time independent perturbation ...

10.2 Time independent perturbation theory with

degenerate states
In order
E to derive theE equation for the ith order perturbative correction to
(0) (i) (0) (0)
φn given by φn we, in equation 10.29 divided by En − Em . This
(0)
only works if En0 6= Em for all m. We thus restricted ourselves to the case
where the eigenstates of H 0 are non-degenerate. We now want to extend
our perturbation theory to cases where we do have degeneracies. We group
the eigenstates of H 0 according to their energy as
E E
(0) (0)
H 0 φn,k = En(0) φn,k . (10.57)

We furthermore take
D E
(0) (0)
φn0 ,k0 φn,k = δk,k0 δn,n0 , (10.58)
E E
(0) (0)
to guarantee orthonormality. For two eigenstates φn0 ,k0 and φn,k with
n0 6= n we take
(0)
En0 6= En(0) . (10.59)

We group all degenerate states together with the same quantum number n.
(0)
The quantum number k labels different states with the same energy En .
For each quantum number n there are Nn orthonormal states with dif-
ferent k. These states form a basis for a Hilbert space. On this finite
dimensional Hilbert space we can express the Hamiltonian H = H 0 + H 1
as a finite dimensional matrix. In order to implement perturbation the-
ory with degenerate states we need to diagonalize
E this finite dimensional
(0)
matrix. We will call the eigenstates φn,κ . For the Hamiltonian matrix
elements we find
 D E
D E δκ0 , κ φ(0)
n,κ H 0 + H 1 φ
(0)
n,κ n0 = n
(0)
φn0 ,κ0 H 0 + H 1 φ(0)
n,κ = D E . (10.60)
 φ(0)0 0 H 0 + H 1 φ
(0)
n,κ n 0 6= n
n ,κ

324
 10.3. Time dependent perturbation theory

We can now rewrite equation 10.29 as

Nm
E XX
(0)
E 1 D
(0)
φ(i)
n,κ = φ m,k (0) (0)
φm,k (10.61)
m6=n k=1 En − Em
 
E i−1
X E
H 1 φ(i−1)
n,κ − En(i−j) φ(j)
n,κ
.
j=0

We do not have to sum over states with the same n, but different κ. The
states we perturb around already diagonalized the full Hamiltonian
E within
(0)
the Hilbert space with a basis defined by the states φn,k for different k
and the same n.

10.3 Time dependent perturbation theory

Most physical processes we want to describe are given by a time-dependent
Hamiltonian. If you turn on a light switch you change the state of a system
where no current flows through a light bulb into a state where a current
flows through the light bulb. In order to describe photosynthesis, we need
to describe the plant under illumination. The light is described by a time-
dependent, oscillating electromagnetic field. What these examples have in
common is that we have a large time-independent Hamiltonian, describing
the electric wires or the chemical composition of the plant and on top of
that a smaller time-dependent Hamiltonian describing the electromagnetic
potentials of the current outlet or the light field. We thus can write our
time-dependent Hamiltonian as

H(t) = H 0 + H 1 (t). (10.62)

For time-independent Hamiltonians we were interested in the calcula-

tion of the eigenstates of the Hamiltonian. These eigenstates form the
stationary states of the Hamiltonian and with those, we can describe wave
packets and the time evolution of these wave packets. For time-dependent

325
Chapter 10. Time independent perturbation ...

Hamiltonians the notion of a stationary state is no longer useful. If we

assume that H 0 is ”large” compared to H 1 (t) the stationary states of H 0
can still be useful. We can, for example, use the stationary states of H 0
as a basis. On this basis we can then describe the dynamics of the system.
A wave packet expanded on the eigenstates of H 0 will show dynamics due
to the combined time evolution of the wave packet by H 0 and the time-
dependent Hamiltonian H 1 (t). The time evolution of a stationary state of
H 0 due to H 0 is just a phase change, whereas H 1 (t) will allow the mixing
or scattering between different eigenstates of H 0 .
Let us assume that we know the state of our system at t = 0 and are
interested in the state at time t. We thus need to solve the differential
equation
∂
i~ |ψ(t)i = (H 0 + H 1 (t)) |ψ(t)i , (10.63)
∂t
with the boundary condition that |ψ(t = 0)i is known. We assume that
H 0 is simple enough such that we can determine its eigenstates |φ(t)i. We
know that
H 0 |φ(t)i = En(0) |φ(t)i , (10.64)
furthermore the time evolution of |φ(t)i due to H 0 is given by a change of
phase such that we define
t (0)
|φ(t)i = e−i ~ En |φi . (10.65)
We can use |φn i as a basis for |ψ(t)i and write
X
|ψ(t)i = cn (t) |φn (t)i (10.66)
n
X t (0)
= cn (t)e−i ~ En |φn i .
n

We now insert this expansion of our time dependent state into the Schrödinger
equation 10.63 to find
X  ∂ 
(0) i (0) (0)

−i ~t En −i ~t En
i~ cn (t) e − En cn (t)e |φn i = (10.67)
n
∂t ~

326
 10.3. Time dependent perturbation theory

(0)
X t
 
cn (t)e−i ~ En En(0) + H 1 (t) |φn i ,
n

subject to the boundary condition that the complex coefficients cn (t = 0)

are known. In order to determine cn (t) we multiply equation 10.67 from
t
the left with hφk (t)| = hφk | ei ~ Ek and find
 
∂ (0)
i~ ck (t) + Ek ck (t) = (10.68)
∂t
t (0) (0)
(0)
X
Ek ck (t) + cn (t)e−i ~ (En −Ek ) hφk |H 1 (t)|φn i .
n

Simplifying yields
∂ i X t (0) (0)
ck (t) = − cn (t)e−i ~ (En −Ek ) hφk |H 1 (t)|φn i . (10.69)
∂t ~ n

We can solve for ck (t) by integration and find

i X t
Z
t0 (0) (0)
ck (t) = ck (0) − cn (t0 )e−i ~ (En −Ek ) φk H 1 (t0 ) φn dt0 . (10.70)
~ n 0

(0)
If H 1 (t) is ”small” we can assume that ck (t) ≈ ck (0). We define ck (t) =
ck (0) and the recursive definition

i X t (i) 0 −i t0 (En(0) −E (0) )

Z
(i+1)
ck (t) = ck (0) − c (t )e ~ k φk H 1 (t0 ) φn dt0
~ n 0 n
(10.71)
(i)
for ck (t). For well behaved systems with a ”small” perturbation H 1 (t)
(i)
compared to H 0 and or short times t the sequence ck (t) in the limit for
i → ∞ might converge to ck (t). Calculating convergence conditions is by
far not a trivial task. However for physically relevant Hamiltonians H 1 (t),
without divergent interactions, one can always find some finite value of
t > 0 for which the series converges. We can see this from equation 10.69

327
Chapter 10. Time independent perturbation ...

where we can express the first order time derivative of ck (t) at time t = 0
as
∂ i X
ck (t) =− cn (0) hφk |H 1 (0)|φn i , (10.72)
∂t t=0 ~ n

combined with the notion that ck (t) must be an analytical function of t in

order not to violate the continuity equations.

10.3.1 Dyson series

We are interested in the question on how a state at time t = 0 evolves to a
state at time t. The operator that describes this transition is given by the
time evolution operator U (t, t0 ) with

|ψ(t)i = U (t, t0 ) |ψt0 i . (10.73)

If the Hamiltonian is time independent we showed in section 3.4 that the

time evolution operator is given as
i
U (t, t0 ) = e− ~ (t−t0 )H . (10.74)

If the Hamiltonian is time-dependent, i.e. H = H(t) this equation does

not hold. In this section we want to derive an equation for the time evolu-
tion operator for time-dependent Hamiltonians and generate a useful series
expansion for this operator.
We separate the time interval from t0 to t into N sections of length
t − t0
δt = , (10.75)
N
going from tn−1 to tn , with n ∈ [1, N ] and

tn = t0 + nδt. (10.76)

We find

U (t, t0 ) = U (tN , tN −1 )U (tN −1 , tN −2 ) . . . U (t1 , t0 ) (10.77)

328
 10.3. Time dependent perturbation theory

N
!
Y
=T U (tn , tn−1 ) .
n=1

Whereby in the last line we introduced the time ordering operator T . Given
the product of two time dependent operators, the time ordering operator
places the operator acting at a later time left of the operator acting at an
earlier time. For example
(
A(t1 )A(t2 ) t1 > t2
T (A(t1 )A(t2 )) = . (10.78)
A(t2 )A(t1 ) t2 > t1

For the limit of N → ∞ we find that tn → tn−1 and δt → 0. As such the

Hamiltonian H(t) does not vary in the interval tn−1 → tn for N → ∞. For
a Hamiltonian that is constant over the time interval of the time evolution
we know the time evolution operator U (tn , tn−1 ) and find
i
lim U (tn , tn−1 ) = lim e− ~ δtH(tn ) . (10.79)
N →∞ N →∞

For the time evolution from t0 to t we find

N
!
− ~i δtH(tn )
Y
U (t, t0 ) = lim T e . (10.80)
N →∞
n=1

The operators H(tn ) and H(tm ) do not necessarily commute such that
i i i
e− ~ δtH(tn ) e− ~ δtH(tm ) 6= e− ~ δt(H(tn )+H(tm )) . (10.81)

However the time ordering operator T will reorder the operators such that
operators at a later time will be left of operators acting on an earlier time.
As such we find
 i i
  i 
T e− ~ δtH(tn ) e− ~ δtH(tm ) = T e− ~ δt(H(tn )+H(tm )) . (10.82)

We thus can write

 i PN 
U (t, t0 ) = lim T e− ~ n=1 H(tn )δt (10.83)
N →∞

329
Chapter 10. Time independent perturbation ...

 i Rt
H(t0 )dt0

−
= T e ~ t0 .

Although this is a nice and compact analytical expression for the time evolu-
tion operator we will continue our argumentation with equation 10.80. We
can replace the time evolution operator with a time-dependent Hamiltonian
for a long interval by infinitely many evolutions of an infinitesimally small
interval. Whereby for each infinitesimal small interval the Hamiltonian is
constant in time.
As discussed in the previous section we often deal with a situation
whereby the Hamiltonian is given by a ”large” time independent part H 0
and a ”small” time dependent part H 1 (t) with

H(t) = H 0 + H 1 (t). (10.84)

Using this dependence of the Hamiltonian we can rewrite the time evolution
operator as

N
!
− ~i δt(H 0 +H 1 (tn ))
Y
U (t, t0 ) = lim T e (10.85)
N →∞
n=1
N  !
Y i
= lim T 1 − δt (H 0 + H 1 (tn )) + O(δt2 )
N →∞ ~
n=1
N    !
Y i i
= lim T 1 − δtH 1 (tn ) 1 − δtH 0 + O(δt2 )
N →∞ ~ ~
n=1
N   !
Y i − ~i δtH 0 2
= lim T 1 − δtH 1 (tn ) e + O(δt ) .
N →∞ ~
n=1

Whereby we note that for N → ∞ we find that δt → 0 such that O(δt2 ) → 0

compared to the other terms in the equation. We now reorder the terms in
equation 10.85 and collect different powers in H 1 (t).
i
U (t, t0 ) = e− ~ (t−t0 )H 0 (10.86)

330
 10.3. Time dependent perturbation theory

Z t
i i i
− e− ~ (t−t1 )H 0 H 1 (t1 )e− ~ (t1 −t0 )H 0 dt1
~ t0
Z tZ t1
1 i i
− e− ~ (t−t2 )H 0 H 1 (t2 )e− ~ (t2 −t1 )H 0 H 1 (t1 )
~2 t1 t0
i
e− ~ (t1 −t0 )H 0 dt1 dt2
+ ...
∞ 
−i 2
 Z Z
X i
= ··· e− ~ (t−tn )H 0 H 1 (tn ) . . .
~ t≥tn ≥...≥t1 ≥t0
n=0
i i
H 1 (t2 )e− ~ (t2 −t1 )H 0 H 1 (t1 )e− ~ (t1 −t0 )H 0 dt1 dt2 . . . dtn .
The time evolution from the state |ψ(t0 )i to |ψ(t)i for a system with Hamil-
tonian H 0 +H 1 (t) can be written as a sum of operators containing different
orders of H 1 (t). To 0th order in H 1 (t) the time evolution operator is given
by the time evolution operator given by the Hamiltonian H 0 . To 1st order
in H 1 (t) the time evolution operator from t0 to t is given by the time evo-
lution operator given by H 0 from t0 to t1 , multiplied by H 1 (t1 ), multiplied
by the time evolution operator given by H 0 from t1 to t. The nth order is
given by n + 1 time intervals where the time evolution operator is given by
H 0 and n interactions with H 1 (ti ) in between these intervals.

10.3.2 Feynman diagrams

Equation 10.86 describes the propagation of a state for a system whose
dynamics is given by the Hamiltonian H(t) = H 0 + H 1 (t). The equation
contains a very important expansion for the time propagation operator in
terms of the propagation of a state for a system whose dynamics is given by
the Hamiltonian H 0 and the interaction of the state with operator H 1 (t)
at a discrete number of times. You will encounter this equation in several
further lectures on quantum mechanics. One instructive way to represent
equation 10.86 is with the use of Feynman diagrams. In equation 10.86 we
encountered two different operators. The time evolution operator
i
U 0 (tj+1 , tj ) = e− ~ H 0 (tj+1 −tj ) , (10.87)

331
Chapter 10. Time independent perturbation ...

and the interaction at time tj with operator

H 1 (tj ) = H 1 (tj ). (10.88)

We can replace these operators by pictures, where acting the operator on

a state corresponds to the state progressing along the line of the diagram

U (t, t0 ) = (10.89)
t0 t

U 0 (tj+1 , tj ) = (10.90)
tj tj+1

i
− H 1 (tj ) = H1 . (10.91)
~ tj tj
The product of two equations is represented by writing the pictures next
to each other, connecting the lines at the left and right end of the pictures.
If a variable stands in the middle of a picture we assume that we integrate
over this variable. The integration boundaries are such that the arrows
always point towards a later time. The second term in equation 10.86 can
be written as
i t − i (t−t1 )H 0
Z
i
− e ~ H 1 (t1 )e− ~ (t1 −t0 )H 0 dt1 = H1
~ t0 t0 t1 t1 t
(10.92)

We now can write for the time evolution operator

= (10.93)
t0 t t0 t

+ H1
t0 t1 t1 t

+ H1 H1
t0 t1 t1 t2 t2 t

+ ...

332
 10.3. Time dependent perturbation theory

With tn+1 = t this becomes

∞ n
!
X Y
= H1 , (10.94)
t0 t t
n=0 0
t1 i=1
ti ti+1
whereby we assume a time ordered product with ti left of ti+1 and ti < ti+1 .
Equation 10.94 describes an operator. In general functions are much
easier to work with than operators. Especially for many particle systems
where operators work on a Hilbert space that grows exponentially with
system size. We are often interested in the complex (probability) amplitude
describing the propagation from a given state |ψ(t0 )i at time t0 to a state
|φ(t)i at time t. We will call the function U (ψ, t0 ; φ, t) the propagator. We
can express the propagator as the expectation value of operator U (t, t0 )
U (φ, t, ψ, t0 ) = hφ|U (t, t0 )|ψi . (10.95)
In order to calculate the propagator we still need to evaluate the Feynman
diagrams which we expressed in equation 10.89 as operators. We Pcan turn
these diagrams into functions by inserting the identity operator τi |τi ihτi |
for some complete orthonormal basis set |τ i at each vertex in equation 10.93
or 10.94. We introduce the following pictorial notation for the functions
U0 (τi , ti ; τj , tj ) = hτi |U 0 (ti , tj )|τj i (10.96)
= hτi |U 0 (ti , tj )|τj i
= τi τj
tj ti

=
τj tj τi ti
We remark that we generally read equations from right to left, i.e. the
equation hτi |U (ti , tj )|τj i defines to complex amplitude to travel from state
|τj i at time tj to state |τi i at time ti . Feynman diagrams are read from
left to right, in the direction of the arrow. We will generally reserve arrows
pointing to the left for anti-particles or holes traveling in the positive time
direction. Similar we define for the matrix elements of operator H 1 (t)
i
− hτi |H 1 (tk )|τj i = τi H1 τj (10.97)
~ tk tk
333
Chapter 10. Time independent perturbation ...

H1
=
τj tk τi tk
We find for the complex amplitude to travel from state |ψi at time t0 to
state |φi at time t

= (10.98)
ψt0 φt ψt0 φt

+ H1
ψt0 τ1 t1 τ2 t1 φt

+ H1 H1
ψt0 τ1 t1 τ2 t1 τ3 t2 τ4 t2 φt

+ ...,

with the basic diagrams

= U (φ, t; ψ, t0 ), (10.99)
ψt0 φt
D i
E
= τi e− ~ (ti −tj )H 0 τj , (10.100)
τj tj τi ti
and
i
H1 = − hτi |H 1 (tk )|τj i . (10.101)
τj tk τi tk ~
We implicitly integrate over all time variables that correspond to internal
vertices, i.e. the places where two basic diagrams connect. We furthermore
implicitly sum over all quantum states |τi ihτi |. For continuous basis sets
the sum over |τi ihτi | is replaced by an integral.

10.3.3 Fermi’s golden rule

One case where perturbation theory is essential is in the description of
measurements. If we want to learn something from a system we have to

334
 10.3. Time dependent perturbation theory

measure it. The way measurements work is that we have to make a small
change to the system to be able to observe something. The formal way to
describe measurements is with the use of response theory. We will discuss
response theory in chapter 13. With the use of the perturbation theory we
can derive a very important result from response theory, the transition rate
between states due to a small perturbation H 1 (t).
We assume the following situation. Before t = 0, i.e. for t < 0 the
dynamics of the system is described by the time independent Hamiltonian
H 0 . At time t = 0 we start the measurement and the Hamiltonian changes
to H 0 + H 1 (t). The Hamiltonian H 1 (t) can describe an electric field over
a conducting wire if we want to measure the resulting current, or an oscil-
lating electromagnetic field if we want to measure the optical conductivity.
The Hamiltonian valid for all times is given as
H = H 0 + θtH 1 (t), (10.102)
with θ(t) the Heaviside step function
(
0 t<0
θ(t) = . (10.103)
1 t>0
Note that for t = 0 we can define the function to be 0, 1, 1/2, or another
value one likes. We normally are interested in the integral over this function
such that a single point does not matter.
We want to describe how a wave packet defined at t < 0 changes its
propagation due to H 1 (t). Instead of considering a general wave packet we
can look at the propagation of the eigenstates of H 0 . Due to the linearity
of the Schrödinger equation we can always write the wave packet and its
time evolution as a sum over the eigenstates |φn i of operator H 0 . If we
start at t0 < 0 in the state |φn i we find at time t the state
|ψ(t)i = U (t0 , t) |φn i . (10.104)
If we, at time t measure the energy of this state we find the distribution
function (see equation 6.129 and the discussion in section 6.5)
X
ρ(t, E) = | hφm |ψ(t)i |2 δE,Em (10.105)
m

335
Chapter 10. Time independent perturbation ...

X
= |hφm |U (t0 , t)|φn i|2 δE,Em
m
X
= |U (m, t; n, 0)|2 δE,Em .
m

In the last step we used that for t0 < 0 the Hamiltonian that governs
the dynamics is given by H 0 such that for t0 < 0 we find U (n0 , 0; n, t0 ) =
i
−δn,n0 ~i e− ~ t0 En and we can replace t0 by 0. We can now define the transition
rate from state |φn i to state |φm i as

1
Γm,n (t) = |U (m, t; n, 0)|2 (10.106)
t
1 2
= + H1 + ... .
t φ n 0 φ m t φ n 0 φ τ1 t 1 φτ2 t1 φm t
With
i
= δn,m e− ~ tEn , (10.107)
φn 0 φm t
we find
1 i
Γm,n (t) = δn,m e− ~ tEn (10.108)
t
2
i t − i (t−t1 )Em
Z
i
− e ~ hφm |H 1 (t1 )|φn i e− ~ (t1 )En + . . . dt1 .
~ 0

For n 6= m we find in lowest order in H 1 (t)

Z t 2
1 1 − ~i t1 (En −Em )
Γm,n (t) = e hφm |H 1 (t1 )|φn i dt1 . (10.109)
t ~2 0

In order to evaluate the integral in equation 10.109 we will assume an

explicit time dependence of H 1 (t). In chapter 13 we will discuss a more
general derivation. One form one often encounters is

H 1 (t) = H 1 e−iωt + H †1 eiωt . (10.110)

336
 10.3. Time dependent perturbation theory

This Hamiltonian for example describes the interaction of a state with

light, an oscillating electromagnetic field, with angular frequency ω. For
the transition rate we find

1 1 D EZ t i
Γm,n (t) = 2 φm H 1 φn e− ~ t1 (En −Em +~ω) dt1 (10.111)
t~ 0

D EZ t i 2
† − ~ t1 (En −Em −~ω)
+ φm H 1 φn e dt1
0

1 D E 1 − e ~i t(En +~ω−Em )
= φm H 1 φn
t Em − En − ~ω
D E 1 − e ~i t(En −~ω−Em ) 2
†
+ φm H 1 φn
Em − En + ~ω
D E 2 t  sin( t (E + ω~ − E )) 2
2~ n m
= φm H 1 φn t
~2 2~ (E n + ω~ − E m)
D E 2 t sin( t (E − ω~ − E )) 2

† 2~ n m
+ m H1 n 2 t
2~ (En − ω~ − Em )
~
+ cross terms.

The time dependent transition rate Γm,n (t) has several terms that os-
cillate in time t. If we look at the limit where t is large, these oscillating
parts will cancel. For En ± ~ω − Em = 0 a contribution remains that does
not oscilate. We find

Γm,n = lim Γm,n (t) (10.112)

t→∞
D E2
= φm H 1 φn δ(En + ~ω − Em )
D E2
+ φm H †1 φn δ(En − ~ω − Em ).

We find first of all that the transition rate between the state |φn i and |φm i
is proportional to the matrix element |hm|H 1 |ni|2 . We furthermore see that

337
Chapter 10. Time independent perturbation ...

one can only have a transition if Em − En = ±~ω. The system can undergo
a transition that costs an energy Em − En only if the frequency of the
driving force is equal to ~ω. If we consider excitations with light, one can
only absorb a photon if the angular frequency of the light, times Planck’s
constant, matches the energy of the excitations. We furthermore see that
the system can also emit a photon if the energy of the state En > Em . The
energy of the emitted photon is En − Em . These are exactly the original
conditions proposed by Einstein for the quantization of light, that lead to
the development of quantum mechanics.

338
Chapter 11

Interaction with
electromagnetic fields

In this chapter we will investigate the interaction of a charged particle

with electro magnetic fields. We will start from the classical interaction
given by Maxwell’s equations and show how to incorporate these into the
Schrödinger equation.
In chapter 2 we introduced the Schrödinger equation using the argu-
ments of De Broglie. Based on experimental considerations De Broglie
related matter to wave packets. From this we found an operator form of
the Hamiltonian that acts on a particle represented by a wave function.
The Hamiltonian in the Schrödinger equation is

~p2
H= + V (~r), (11.1)
2m

with ~p = −i~∇ an operator. The Hamiltonian in the Schrödinger equation

is similar to the classical Hamiltonian. The main difference is that the
classical momentum is given by a variable, whereas in quantum mechanics
it is replaced by an operator.
Classically we can derive the equations of motions from the Hamilto-

339
Chapter 11. Interaction with electromagnetic fields

nian. We do this with the use of Hamiltons equations which are

d~ri ∂H
= (11.2)
dt ∂~
pi
d~
pi ∂H
=− .
dt ∂~ri
We start from the classical Hamiltonian for a particle in a potential H =
~2
p
2m + V (~r). We now can, with the use of Hamiltons equations, derive
the equations of motion for the particle. We find from the first Hamilton
equation the relation between momentum and velocity of the particle

d~r p~
= , (11.3)
dt m
d~
r
or p~ = m~v , with ~v = dt the velocity of the particle. The second Hamilton
equation yields

d~
p
= −∇H (11.4)
dt
= −∇V (~r)
= F~ .

Combining the two equations yields F~ = m~a, with F~ the force acting on
2
the particle due to the potential V (~r) and ~a = ddt2~r the acceleration of the
particle.
In the next section we will derive the equations of motion from the
Hamilton formalism for a classical charged particle in an electromagnetic
field. We will find a change in the form of the total momentum derived
from the first Hamilton equation. This changed definition of momentum
combined with the second Hamilton equation yields the Lorentz equations.
The Lorentz equations give the relation between the force on the particle
and the electromagnetic fields. In the second section of this chapter, we
will extrapolate the classical Hamilton formalism for electromagnetic fields
to the quantum world. As a result we will add a classical electromagnetic

340
 11.1. Electro magnetic fields in Hamilton formalism

field into the Schrödinger equation. From the consideration of black body
radiation, we know that we need to quantize electromagnetic fields. We
will do this in the third section of this chapter. The third section should
be seen as an outlook for future lectures on quantum field theory.

11.1 Electro magnetic fields in Hamilton formal-

ism
In this section we assume to have a classical electric field E(t, ~ ~r) and
~
magnetic field B(t, ~r). These fields are allowed to be time and position-
dependent. We want to describe how a classical particle with mass m and
charge q moves through these fields. We will not directly use the Lorentz
equations, but will construct an Hamiltonian and use the Hamilton equa-
tions to derive the equations of motion. We will in the next section use a
similar formalism to add the interaction of a charged particle with electro-
magnetic fields to the Schrödinger equation.
The classical force on a classical charged particle is the Lorentz force

F~ = q E
~ + q~v × B.
~ (11.5)

The equations of motion for a charged particle subject to the Lorentz force
are

F~ = m~a (11.6)
~ + q~v × B.
= qE ~

In this section we want to proof that the Hamiltonian that generates this
force is given by
1  2
~ + qΦ,
H= p~ − q A (11.7)
2m
~ the magnetic vector potential, such that
with Φ the potential and A
~
~ ~r) = −∇Φ(t, ~r) − ∂ A
E(t, (11.8)
∂t

341
Chapter 11. Interaction with electromagnetic fields

~ ~r) = ∇ × A.
B(t, ~

In order to proof that equation 11.7 indeed is the correct Hamiltonian

to describe the motion of a charged particle in an electromagnetic field we
will derive the equation of motion from this Hamiltonian. Using Hamilton’s
first equation for the time derivative of the position we find
d~ri ∂H
= (11.9)
dt ∂~
pi
1  ~i

= p~i − q A
m
If we multiply the previous equation with m we find
~ i.
m~vi = p~i − q A (11.10)
For the momentum of the particle we thus find the expression
~ i.
p~i = m~vi + q A (11.11)
The total momentum is not just the mass times the velocity of the particle
but is modified by the interaction with the electromagnetic field.
We can compare this with the expression for the time derivative of the
momentum as obtained from Hamiltons second equation
d~
pi ∂H
=− (11.12)
dt ∂~ri
q  ~
~ · ∂ A − q ∂Φ .

= p~ − q A
m ∂~ri ∂~ri
Combining the time derivative of equation 11.11 with equation 11.12 yields

dA~i q  ~
~ · ∂ A − q ∂Φ .

m~ai + q = p~ − q A (11.13)
dt m ∂~ri ∂~ri
dA~i ∂A~i ~i · ∂~
r
With dt = ∂t + ∇A ∂t we find

∂Φ ∂A~i q  ~
~ · ∂ A − q∇A
~ i · ∂~r

m~ai = −q −q + p~ − q A (11.14)
∂~ri ∂t m ∂~ri ∂t

342
 11.2. Classical fields in the Schrödinger equation

~
= qE~ i + q~v · ∂ A − q∇A~ i · ~v
∂~ri
 
~
= q Ei + q ~v × B ~
i

whereby we used in the second step that E~ i = − ∂Φ − ∂ A~ i and m~v = p~ − q A.

~
∂~
ri ∂t
~ ~r) = ∇ × A
For the last step we use that B(t, ~ and wrote out the ith
component of this vector. With this we showed that the Hamiltonian that
generates the Lorentz force is given by equation 11.7.

11.2 Classical fields in the Schrödinger equation

In the previous section we learned how write down the classical Hamiltonian
including electro magnetic fields. We will modify the Schrödinger equation
equation in a similar fashion. We will start from the standard Schrödinger
equation

∂
i~ |ψ(t)i = H |ψ(t)i . (11.15)
∂t
The Hamiltonian for several interacting charged particles with spin subject
to an external electromagnetic field will be written as a sum of several
terms related to the different interactions between charged particles and
electromagnetic fields.
Following the classical Hamiltonian for the kinetic energy and Lorentz
forces between a charged particle and electric field we define the quantum
Hamiltonian as
1  2
~ ~r) + qΦ(t, ~r).
H Lorentz = ~p − q A(t, (11.16)
2m
For N charged particles in an electro magnetic field, one should, besides
the Lorentz force, also consider the Coulomb forces between the particles.
We already encountered these in the previous chapter where we discussed
the hydrogen atom. The Coulomb Hamiltonian between two particles at

343
Chapter 11. Interaction with electromagnetic fields

coordinates ~r1 and ~r2 with charge q1 and q2 is

q1 q2 1
H Coulomb = . (11.17)
4π0 |~r1 − ~r2 |

Besides an electric charge, particles can have a spin. This intrinsic

angular momentum for a charged particle yields a magnetic moment that
interacts with the magnetic field. The resulting term in the Hamiltonian is
the spin contribution to the Zeeman term
q ~ ~
H Spin−Zeeman = −g S · B(t, ~r), (11.18)
2m
where we introduced the g factor. The g factor has no classical explanation,
but follows from the Dirac equations. For electrons we find ge ≈ 2, for
neutrons gn ≈ −3.8 and for protons gp ≈ 5.6.
For N charged particles interacting with a classical electromagnetic field
the total Hamiltonian is
N
X 1  2
~ ~ri ) + qi Φ(t, ~ri )
H= ~pi − qi A(t, (11.19)
2mi
i=1
qi ~ ~
− gi S i · B(t, ~ri )
2mi
X qi qj 1
+ .
4π0 |~ri − ~rj |
j>i

11.3 Gauge invariance

We defined the Hamiltonian including electro magnetic fields with the use
~ This introduces a Gauge invariance in the equations.
of the vector field A.
The magnetic field is defined as the curl of A. ~ The curl of the divergence
of a continues function (Λ(t, ~r)) is zero, i.e.

∇ × ∇Λ(t, ~r) = 0. (11.20)

344
 11.3. Gauge invariance

We can define the Gauge transformation

~0 = A
A ~ − ∇Λ(t, ~r) (11.21)
∂Λ(t, ~r)
Φ0 = Φ + ,
∂t
~ and the magnetic field B
such that the electric field E ~ derived from A
~ and
~ are equivalent. As a result the physics of the system should not change
A 0

under a Gauge transformation.

The Hamiltonian will change under a Gauge transformation. For a
single particle without spin we find for the original Hamiltonian
1  ~
2
H= ~p − q A(t, ~r) + qΦ(t, ~r) (11.22)
2m
and for the Gauge transformed Hamiltonian
 
0 1  
~
2 ∂Λ(t, ~r)
H = ~p − q A(t, ~r) + ∇Λ(t, ~r) + q Φ(t, ~r) + . (11.23)
2m ∂t

Eigenstates of H are not necessarily eigenstates of H 0 . As we will show in

the next section, if ψ(t, ~r) is an eigenstate of H, such that

Hψ(t, ~r) = Eψ(t, ~r) (11.24)

then
q
ψ 0 (t, ~r) = ei ~ Λ(t,~r) ψ(t, ~r) (11.25)

is an eigenstate of H 0 such that

H 0 ψ 0 (t, ~r) = Eψ 0 (t, ~r). (11.26)

11.3.1 Gauge invariance as a local phase transformation

A Gauge transformation is related to a local phase change of the wave
function. Instead of starting from the experimentally justified Maxwell

345
Chapter 11. Interaction with electromagnetic fields

equations and modifying the Hamiltonian such that the Maxwell equations
are retrieved one can investigate how a theory can be developed that is
invariant under a local phase change of the wave function. This property
is known as a Gauge symmetry.
Before we look at local phase changes of the wave function, i.e. a Gauge
symmetry, let us discus an almost obvious symmetry in quantum mechanics,
namely a global phase change of the wave function. The symmetry operator
that changes the global phase of a wave function is

T θ ψ(t, ~r) = eiθ ψ(t, ~r). (11.27)

The operator T θ multiplies the wave function with a phase. This operator
obviously commutes with the Hamiltonian of the Schrödinger equation

[T θ , H] = 0. (11.28)

We can now relate this symmetry operator to a conserved quantity. Look-

ing back at chapter section 7.2 we see that it must follow that the time
dependent expectation value of T θ is conserved

∂
hψ(t)|T θ |ψ(t)i = 0. (11.29)
∂t
The operator T θ acting on a state just multiplies the state by a complex
number. We thus find that
∂
hψ(t)|ψ(t)i = 0. (11.30)
∂t
In other words the fact that a phase change of the wave-function does not
change the system leads to a conservation of the norm of the wave-function.
For the Schrödinger equation the norm of the wave-function is related to
the total particle number. A system that is invariant under a global phase
change of the wave-function thus obeys particle number conservation. For
massive charged particles the mass and charge of the system is proportional
to the total number of particles. We thus also find that for a system where

346
 11.3. Gauge invariance

the Hamiltonian commutes with a global phase change of the wave function
the total mass and the total charge is conserved.
For a Gauge transformation we do not change the global phase, but
change locally the phase of the wave function. For a continues, differentiable
function Λ(t, ~r) we define the phase change as equation 11.21, ψ(t, ~r)0 =
q
ei ~ Λ(t,~r) ψ(t, ~r). The Gauge symmetry operation that generates this local
phase change is
q
T Λ(t,~r) ψ(t, ~r) = ei ~ Λ(t,~r) ψ(t, ~r). (11.31)
The operator T Λ(t,~r) does not commute with the Hamiltonian defined in
equation 11.22. We can however modify the Hamiltonian, and all other
operators such that all observables remain unchanged for a local phase
transformation. If we change all operators O by
q q
O 0 = ei ~ Λ(t,~r) Oe−i ~ Λ(t,~r) , (11.32)
then
q q q q
ψ10 (t, ~r)∗ O 0 ψ20 (t, ~r) = ψ1 (t, ~r)∗ e−i ~ Λ(t,~r) ei ~ Λ(t,~r) Oe−i ~ Λ(t,~r) ei ~ Λ(t,~r) ψ2 (t, ~r)
(11.33)
= ψ1 (t, ~r)∗ Oψ2 (t, ~r),
i.e. all observables are invariant under a simultaneous local phase transfor-
mation of the wave-function and all operators.
For operators that are functions of the position operator we know that
[~r , T Λ(t,~r) ] = 0 such that
q q
~r 0 = ei ~ Λ(t,~r)~r e−i ~ Λ(t,~r) (11.34)
= ~r .
Operators that only depend on the position of a particle, or on functions
of the position operator do not change under a Gauge transformation.
Next we investigate how the momentum operator transforms under a
Gauge transformation
q q
~p0 = ei ~ Λ(t,~r)~pe−i ~ Λ(t,~r) (11.35)

347
Chapter 11. Interaction with electromagnetic fields

= −q∇Λ(t, ~r) + ~p.

The momentum operator does not transform back to it selves under a Gauge
transformation. Without additional modifications to the Schrödinger equa-
tion quantum mechanics would not be invariant under a local phase change
of the wave function. We can make the equation invariant if one looks at
the Gauge transformation of ~p − q A(t, ~ ~r), with A(t,~ ~r) the electromagnetic
vector field. The electromagnetic vector field is a function of ~r and as such
does not change under a local phase transformation. The operator ~p does
change under a Gauge transformation. The additional term, −q∇Λ(t, ~r)
is the gradient of a continues function. The electromagnetic vector field
can be changed by the gradient of a continues function without changing
physical observables, such that the combined operator
 
~ 0 (t, ~r) = ei ~q Λ(t,~r) ~p − q A(t,
~p0 − q A ~ ~r) e−i ~q Λ(t,~r) (11.36)
 
= ~p − q A(t, ~ ~r) + ∇Λ(t, ~r) ,

does not change physical observables under a Gauge transformation.

In order to understand how the Schrödinger equation transforms under
a change of the local phase of a wave function, we need to know how the
time derivative transforms
∂0 q ∂ q
i~ = i~ei ~ Λ(t,~r) e−i ~ Λ(t,~r) (11.37)
∂t ∂t
∂ ∂
= i~ + q Λ(t, ~r).
∂t ∂t
Also the time derivative operator changes under a Gauge transformation.
We can absorb the additional term into the Gauge transformation of the
electric field potential.
If we combine the transformation of the time derivative, the momentum
operator and the transformation of functions of ~r we find for the transfor-
mation of the Schrödinger equation
 
∂ 1  ~
2
i~ ψ(t, ~r) = ~p − q A(t, ~r) + qΦ(t, ~r) ψ(t, ~r) ⇐⇒ (11.38)
∂t 2m

348
 11.3. Gauge invariance


∂ 1   2
i~ ψ 0 (t, ~r) = ~ ~r) + ∇Λ(t, ~r)
~p − q A(t,
∂t 2m
 
∂Λ(t, ~r)
+q Φ(t, ~r) + ψ(t, ~r)0 .
∂t
q
The local change in the phase between ψ(t, ~r) and ψ(t, ~r)0 = ei ~ Λ(t,~r) ψ(t, ~r)
changes the effect of the position and time derivatives. These changes can
be absorbed into the definitions of the electro magnetic vector field A(t, ~ ~r)
and Φ(t, ~r). As a result if ψ(t, ~r) is a solution of the Schrödinger equation
then ψ 0 (t, ~r) is a solution of the transformed Schrödinger equation, and
both describe the same physics. We call a theory for which the physics and
dynamics of the system does not change, i.e. is invariant under a continues
local transformation a Gauge theory. We have seen that in order to turn
the Schrödinger equation that describes electrons into a Gauge theory we
needed to add the electro magnetic field in order to absorb the changes in
the derivatives under a local phase change. This is a general property we
will find in later lectures when describing other elementary particles. All
particles and forces in the standard model are described by Gauge theories
of the form we introduced in this section.

11.3.2 Coulomb Gauge

~ do not commute.
In general one should realize that the operators ~p and A
 2
~ ~r) is equal to
The term ~p − q A(t,
 2  
~p − q A(t, ~ ~r) − A(t,
~ ~r) = ~p2 − q ~p · A(t, ~ ~r) · ~p + q 2 A(t,
~ ~r)2 (11.39)

~ ~r) = 0, which is known

We can choose the Gauge Λ(t, ~r) such that ∇ · A(t,
as the Coulomb Gauge. In the Coulomb Gauge ~p and A(t,~ ~r) commute such
that the Hamiltonian for a single particle without spin becomes
1  2 ~ ~r) · ~p + q 2 A(t,

~ ~r)2 + qΦ(t, ~r).
H= ~p − 2q A(t, (11.40)
2m

349
Chapter 11. Interaction with electromagnetic fields

11.4 Quantized fields in the Schrödinger equation

In section 1.1 we showed that in order to explain black body radiation
one needs to quantize the electromagnetic field. Electromagnetic waves,
defined by the vector field A ~ can not have an arbitrary amplitude. For
electromagnetic waves propagating through the vacuum, these amplitudes
are such that the energy contained in the electromagnetic field is an integer
multiple of ~ω. The field contains n ∈ N photons of energy ~ω. In order to
~ we will first choose a basis on which we can expand
quantize the field A,
the field. Given this basis we will define an operator A ~ that creates or
annihilates particles (photons) into these basis states. We want our basis
states to be normalized. Plane waves, the eigenstates for a non-interacting
particle in free space, are non-normalizable. A standard solution is to
choose a square box of length L and volume V = L3 and assume that the
field is zero outside this box. We can let the size of this box go to infinite
after we’ve calculated the properties of interest. If we choose the Coulomb
gauge such that ∇ · A ~ = 0 we can write the vector field inside the volume
V as
r
1 X X  (σ) (σ) ~ (σ)∗ (σ) ~

~
A(t, ~r) = ~~ c~ (t)eik·~r + ~~ c~ (t)∗ e−ik·~r , (11.41)
V σ
k k k k
~k

(σ)
with ~k the wave vector of the field, c~ (t) the time-dependent expansion
k
(σ)
coefficients of the field on the plane wave basis, and ~~ the polarization
k
vector of the electromagnetic field. We can choose two possible polarization
vectors that are perpendicular to each other and perpendicular to the wave
(σ)
vector ~~ · ~k = 0. The field is zero outside the volume V and in order to
k
be continuous also zero at the boundary of the box. As a result the allowed
wave vectors are discrete
~k = (nx , ny , nz ) 2π (11.42)
L
with n ∈ Z. The quantization of the wave vector is not the one leading to
the definition of photons but needed to make the wave functions normaliz-
able. The quantization of ~k will vanish once we take the limit of L → ∞.

350
 11.4. Quantized fields in the Schrödinger equation

We can quantize the amplitude of the field, by replacing the expansion

(σ)
coefficients c~ (t) with operators that create or annihilate a photon. We
k
define the quantization rules
r
(σ) ~
c~ (t) → b (11.43)
k 2ω0 σ,~k
r
(σ) ~ †
c~ (t)∗ → b .
k 2ω0 σ,~k

The operator b† ~ creates a photon with polarization σ and wave vector ~k,
σ,k
the operator b ~ annihilates a photon from the same state. The commu-
σ,k
tation relations for the photon creation and annihilation operators are

[b ,b ] =0 (11.44)
σ,~k σ 0 ,~k0

[b† ~ , b† 0 ~ 0 ] =0
σ,k σ ,k

[b ~ , b† 0 ~ 0 ] = δ~k,~k0 δσ,σ0 .
σ,k σ ,k

Photons are bosons and follow the corresponding commutation relations.

The electro magnetic vector field operator becomes
r
~  ~ ~

~σ b ~ eik·~r + ~σ∗ b† ~ e−ik·~r .
XX
~ r) =
A(~ (11.45)
σ
2ωV 0 σ,k σ,k
~k

The operators for the electro and magnetic field can be derived from this
and are respectively
r
~ω  ~ ~

~σ b ~ eik·~r − ~σ∗ b† ~ e−ik·~r
XX
~ r) = i
E(~ (11.46)
σ
2V 0 σ,k σ,k
~k
r
~ω ~ ~ ~

k × ~σ b ~ eik·~r − ~k × ~σ∗ b† ~ e−ik·~r
XX
~ r) = i
B(~
σ
2V 0 σ,k σ,k
~k

351
Chapter 11. Interaction with electromagnetic fields

We can now infer an Hamiltonian for the photons by calculating the

energy contained in the field
Z
1 ~ ~r)|2 + c2 |B(t,
~ ~r)|2 d3~r
H = 0 |E(t, (11.47)
2 V
XX † 1

= ~ω b ~b ~ + .
σ
σ,k σ,k 2
~k

The Hamiltonian for a single massive particle interacting with photons is

1  ~ r)
2
H= ~p − q A(~ (11.48)
2m
q ~ ~
−g S · B(~r)
2m
XX † 1

+ ~ω b ~b + .
σ,k σ,~k 2
~k σ

The first term in this operator describes the interaction between the
particle and the electro magnetic field. The operator A(~~ r) allows for the
creation (emission) and annihilation (absorption) of photons. Eigenstates
of this Hamiltonian are states with a single particle and a variable number
of photons. The electromagnetic field generated from a moving particle
follows naturally from this Hamiltonian. The second term describes the
interaction of the particle spin with the magnetic field. The last term de-
scribes the energy of the photons. The Hamiltonian for multiple charged
particles follows by summing the first two terms of equation 11.48 for each
particle. The last term describes the electromagnetic photon field, which is
the same for all particles. The Coulomb interaction between two particles
can be seen as the interaction of one particle with the electro magnetic
field generated by the other particle. As such we do not need to add the
Coulomb interaction to this Hamiltonian as we did for the case of the in-
teraction between charged particles and a classical electro magnetic field.
Each particle will generate an electro magnetic field by creating and an-
nihilating photons. The other particles will move through this field and

352
 11.5. Zeeman effect

interact with this field by creating and annihilating photons. The resulting
interaction is the Coulomb force.
The future study of this Hamiltonian is part of quantum field theory,
which is one of the lectures you can follow in your future studies. We
should note that equation 11.48 is not fully Lorentz invariant. This can
lead to problems and incorrect answers. The correct way would be to do
the quantization of the electromagnetic field on top of the Dirac equation.
Done properly this leads to the equations of quantum electro dynamics.

11.5 Zeeman effect

In this section we investigate the effect of a constant magnetic field on a
charged particle. We will assume the magnetic field to be in the z direction.
For the vector field we take

~ = 1 (xBz ŷ − yBz x̂) .

A (11.49)
2
The corresponding magnetic field is

~ =∇×A
B ~ (11.50)
= Bz ẑ.

The Hamiltonian for a charged particle in a constant magnetic field

~ = Bz ẑ is
B

1  ~ −g q S
2
~ ·B ~
H= ~p − q A (11.51)
2m 2m
 2
1 1 q
= ~p − q (xBz ŷ − yBz x̂) − g S z Bz
2m 2 2m
1 2 q2 2 2 q q
Bz (x + y 2 ) −


= ~p + xpy − ypx Bz − g S z Bz
2m 8m 2m 2m
1 2 q2 2 2 q q
= ~p + Bz (x + y 2 ) − L z Bz − g S z Bz .
2m 8m 2m 2m

353
Chapter 11. Interaction with electromagnetic fields

The first term is the kinetic energy, the second term is related to the in-
ternal energy of the magnetic field, the third and fourth terms define the
interaction of the orbital and spin magnetic moments in the system with
the external magnetic field. In chapter 8.6.2 we inferred the interaction
between the magnetic moments of the hydrogen atom with a magnetic field
from classical arguments. In this section we have given it a theoretical
foundation.

354
Chapter 12

Relativistic effects

The Schrödinger equation is not Lorentz invariant. When Schrödinger in-

troduced his equation it was clear to him that this equation, at best, is
the non-relativistic approximation of a more general relativistic equation.
Schrödinger himself tried to find a differential equation to describe relativis-
tic quantum mechanics. Before Schrödinger published his non-relativistic
equation in 1926 [7] he, in his notes, derived a Lorentz invariant relativis-
tic differential equation that describes quantum mechanics. This equation
is now known as the Klein Gordon equation. Schrödinger showed, in his
private notes, that the Klein Gordon equation does not correctly described
the relativistic corrections for the energy spectrum of the hydrogen atom
and as such dismissed this equation. In the end Schrödinger failed to find
a relativistic wave equation for electrons and went ahead and published his
non-relativistic version.
After Schrödinger published his equation many people searched for a
relativistic version of his equation. Within a year both Klein [79] and Gor-
don [80] proposed independently a relativistic version of the Schrödinger
equation. We will discuss this equation below. Klein and Gordon used
their equation to describe Compton diffraction of x-rays. The Klein Gor-
don equation can be used to describe spin-less particles such as Pions. As
Schrödinger privately derived in his notes and knew at the moment the

355
Chapter 12. Relativistic effects

Klein Gordon equation was published, this equation does not yield predic-
tions for the relativistic corrections to the hydrogen atomic energy levels in
agreement with experiment. In the end it was Dirac who proposed the cor-
rect form for the relativistic generalization of the Schrödinger equation for
electrons [81]. With his equation Dirac was able to include the spin of elec-
trons in a natural way and find relativistic corrections to the eigenenergies
of the hydrogen atom.
In this chapter we will first describe the general idea behind the Dirac
equation. After we formulated the Dirac equation we will have a look at the
exact solutions for a particle in free space. We then test if the Dirac equation
is compatible with the Schrödinger equation and look at the non-relativistic
limit of the Dirac equation. Once we convinced ourselves that the Dirac
equation gives a mere correction to the Schrödinger equation we will derive
the solution for the hydrogen atom. Once we know the eigenstates of the
hydrogen atom we will discuss the Lamb shift [82]. This is a small shift
of the electronic levels of the hydrogen atom due to the electro-magnetic
vector potential. The challenge of calculating the Lamb shift will bring us
to the formulation and basic ideas behind quantum electro dynamics. We
will not go into the details of the calculation, but will lay out the general
framework.

12.1 The Dirac equation

In order to understand how Dirac came to the idea to propose his equation
we can have a look on how we can introduce the Schrödinger equation from
the classical energy momentum dispersion of a massive particle. We dis-
cussed in chapter 2 that one can obtain the Schrödinger equation by substi-
tuting classical variables in the energy momentum dispersion by operators.
We then use the same substitutions for the relativistic energy momentum
dispersion. This will yield the Klein Gordon equation. After we obtained
the Klein Gordon equation we will have a look on how to modify this to
obtain the Dirac equation.
In classical physics the micro state of a system is described by the

356
 12.1. The Dirac equation

position and momentum coordinates of all particles. In quantum mechanics

we replace the representation of a state by the wave function φ(~r, t). The
differential equation this wave function obeys, i.e. the Schrödinger equation,
can be obtained from the classical energy momentum dispersion relation
1 2
E= p~ . (12.1)
2m
In order to go from classical physics to quantum physics, we substitute
∂
E with the operator i~ ∂t and the momentum p~ with the operator −i~∇.
Finally we let both sides of equation 12.1 act on the wave function. With
these substitutions we obtain the Schrödinger equation

∂ ~2 2
i~ φ(~r, t) = − ∇ φ(~r, t). (12.2)
∂t 2m
The simplest assumption for a relativistic extension of the Schrödinger
equation would be to start from the relativistic energy momentum relation
for a massive particle

E 2 = c2 p~2 + m2 c4 . (12.3)
∂
We again can substitute E by the operator i~ ∂t and p~ by the operator
−i~∇. We let both sides act on a wave function φ(~r, t) and obtain the
differential equation

∂2
−~2 φ(~r, t) = (−c2 ~2 ∇2 + m2 c4 )φ(~r, t). (12.4)
∂t2
The resulting equation is the Klein Gordon equation. As Schrödinger al-
ready noted, this equation does not describe the relativistic corrections for
the energy levels of the hydrogen atom correctly.
In order to come to his equation, Dirac focused on two problems within
the Schrödinger equation. Firstly, the Schrödinger equation is not Lorentz
invariant, secondly the equation does not describe the spin of the electrons.
In the words of Dirac[81]: ”The discrepancies consist of duplexity phenom-
ena, the observed number of stationary states for an electron in an atom

357
Chapter 12. Relativistic effects

being twice the number given by the theory. To meet the difficulty, Goudsmit
and Uhlenbeck have introduced the idea of an electron with a spin angular
momentum of half a quantum and a magnetic moment of one Bohr magne-
ton [75]. This model for the electron has been fitted into the new mechanics
by Pauli [83] and Darwin [84].” Pauli and Darwin incorporated the spin of
an electron into the Schrödinger equation by writing the wave-function for
an electron as a two component vector. Each component of this vector is
a solution of its own Schrödinger equation. Magnetic fields and the spin-
orbit coupling can mix the two different components of the wave-function.
The doubled Schrödinger equation that incorporates the spin in the form
of Pauli and Darwin becomes:
 ∂   
i~ ∂t 0 φ↑ (~r, t)
∂ · = (12.5)
0 i~ ∂t φ↓ (~r, t)
p2
!
~ e ~ ~
2m + V (~
r ) − 2m B · L 0
p2
~ e ~ ~
0 r) − 2m
2m + V (~ B·L
 
1 e Bz Bx + iBy
−g ~
2 2m Bx − iBy −Bz
 !  
1 Lz Lx + iLy φ↑ (~r, t)
+ζ ~ · ,
2 Lx − iLy −Lz φ↓ (~r, t)

with B ~ =
~ an external magnetic field, V (~r) an external electric potential, L
~r × ~p the angular momentum of the particle, and g and ζ experimentally
determined constants.
With these considerations in mind Dirac looked for a relativistic exten-
sion of the Schrödinger equation that correctly describes the energy levels
of an hydrogen atom. Dirac noted that the Schrödinger equation is linear
in the derivative of time. He searched for a similar form for the relativistic
equation. Dirac thus set himself the task to find an equation linear in the
energy that would reproduce the equation that E 2 = m2 c4 + c2 p~2 . For this
he wrote
X3
E=c αi p~i + α4 mc2 , (12.6)
i=1

358
 12.1. The Dirac equation

with αi chosen such that

3
!2
X
E2 = c αi p~i + α4 mc2 (12.7)
i=1
= m c + c2 p~2 .
2 4

This yields the following requirements on α

αi αj + αj αi = 2δi,j . (12.8)

If the αi ’s need to be complex numbers there is no solution for this set

of equations. However the variables αi might be matrices and the wave
function on which the Hamilton operator acts a vector. If φ(~r, t) is a vector
of length two, then the two components could represent the spin of the
electron. From the Zeeman effect, the Stern-Gerlach experiment and the
structure of the periodic table it was well known that any electron state is
represented by a linear combination of two basis states. It would thus be
nice to represent the wave function of an electron as
 ↑ 
φ (~r, t)
φ(~r, t) = , (12.9)
φ↓ (~r, t)

and the matrices αi as

 
αi,↑,↑ αi,↑,↓
αi = . (12.10)
αi,↓,↑ αi,↓,↓

In order for the Hamiltonian to be Hermitian, and the energies real, we

must restrict the matrices αi to be Hermitian. The final form would yield an
equation of the same form as equation 12.5, phenomenologically introduced
by Darwin [84] and Pauli [83].
For a two by two matrix representation of the Dirac equation the re-
quirements on the Dirac Matrices αi are
 
1 0
αi αj + αj αi = 2δi,j . (12.11)
0 1

359
Chapter 12. Relativistic effects

The three Hermitian matrices that fulfill equation 12.11 are the Pauli ma-
trices:
     
0 1 0 −i 1 0
σx = , σy = , and σz = . (12.12)
1 0 i 0 0 −1

A two by two Hermitian matrix has 4 independent components. The two

real elements on the diagonal and the real and imaginary part of the off
diagonal element. These 4 components can be recombined to form 4 linearly
independent matrices. Three of them are the Pauli matrices. The Pauli
matrices fulfill equation 12.11. The forth linearly independent matrix is the
identity matrix. The identity matrix does not fulfill equation 12.11. The
relativistic energy momentum dispersion relation, which depends on the
square of the energy and momentum can not be factorized into the square
of a linear dependence, if we assume equation 12.8 to be a matrix equation
of dimension two.
Dirac must have felt he was on the right track and instead of giving up
when he saw that one can not factorize the relativistic energy momentum
relation when the Dirac matrices have dimension two he realized that one
can factorize the energy momentum dispersion in to a linear form if we as-
sume matrices of dimension four. The solution Dirac found to factorize the
relativistic energy momentum relation is given by the following αi matrices:
 
0 0 0 1
0 0 1 0
α1 = αx = 
0
, (12.13)
1 0 0
1 0 0 0

 
0 0 0 −i
0 0 i 0 
0 −i 0 0  ,
α2 = αy =   (12.14)
i 0 0 0

360
 12.1. The Dirac equation

 
0 0 1 0
0 0 0 −1
α3 = αz = 
1 0
, (12.15)
0 0
0 −1 0 0

 
1 0 0 0
0 1 0 0
α4 =  . (12.16)
0 0 −1 0 
0 0 0 −1
We now can replace the energy and momentum in equation 12.6 by opera-
tors as done within the Schrödinger equation. The Dirac equation becomes
3
!
∂ X ∂ 2
i~ φ(~r, t) = −i~c αi + α4 mc φ(~r, t). (12.17)
∂t ∂~ri
i=1
or
3
!
∂ X
i~ φ(~r, t) = c αi~pi + α4 mc2 φ(~r, t) (12.18)
∂t
i=1
α · ~p + α4 mc2 φ(~r, t),


= c~

whereby we introduced the vector α~ = (α1 , α2 , α3 ). The Dirac equation has

the same linear form for its time dependence as the Schrödinger equation,
with the Dirac Hamiltonian defined as

α · ~p + α4 mc2 .


H D = c~ (12.19)

Within the Schrödinger equation the wave function is a complex scaler.

The wave function in the Dirac equation is a complex vector of dimension
4.
 
φ↑,l (~r, t)
 φ↓,l (~r, t) 
φ(~r, t) = 
φ↑,s (~r, t) .
 (12.20)
φ↓,s (~r, t)

361
Chapter 12. Relativistic effects

The components of the wave function describe the 4 basis states that can
be used to create an electron or positron with spin up or spin down. The
first two components are called the large component, the second two the
small component of the wave function. The first and third index refer to
spin up, the second and forth to spin down. As we will show later in this
chapter only for a particle at rest the four basis states of the vector refer
to eigenstates of the Dirac equation and only in this case the basis states
corresponds to an electron or positron with either spin up or down. For
other situations physical particles are linear combinations of the 4 basis
states.

12.2 Dirac equation with electromagnetic field

The Dirac Hamiltonian given in equation 12.19 does not include the inter-
action between particles and the electro-magnetic field. Adding an elec-
tromagnetic field to the Dirac equation can be done relatively straight for-
ward. We start from the magnetic vector potential A(~ ~ r, t) and the elec-
tric potential Φ(~r, t) with the potential energy for an electron equal to
V (~r, t) = −eΦ(~r, t), and e the absolute value of the electronic charge. The
electric and magnetic fields follow from the electromagnetic potentials as
~ r, t) = −∇Φ(~r, t) − ∂ A(~
E(~ ~ r, t), (12.21)
∂t
and
~ r, t) = ∇ × A(~
B(~ ~ r, t). (12.22)

For the interaction we replace the momentum operator ~p by ~p + eA ~ and

add the potential energy V = −eΦ due to an electron moving in an external
potential Φ. The resulting Dirac Hamiltonian becomes
   
H D = c~ ~ i (~r, t) + V (~r, t) + α4 mc2 ,
α · ~p + eA (12.23)

and the wave functions are solutions of the equation

∂
i~ φ(~r, t) = H D φ(~r, t). (12.24)
∂t

362
 12.3. Properties of the Dirac α matrices and alternative
notations

12.3 Properties of the Dirac α matrices and alter-

native notations
The choice of the Dirac α matrices used to factorize the quadratic energy
momentum relation is not unique. Any unitary transformation with u† u =
1 can be used to redefine the Dirac matrices as

αi0 = u† αi u, (12.25)

such that the new Dirac matrices αi0 can be used to factorize the quadratic
energy momentum dispersion relation. The unitary transformation u repre-
sents a basis transformation of the Dirac four-vectors. In the choice made
by Dirac the components relate to the large and small part of the wave
function with either spin up or down in the z direction. After a transfor-
mation of α to α0 by the unitary matrix u the new basis is given by uT
times the old basis.
One property that all choices of the Dirac matrices have in common is
their anti-commutation relations. By doing 32 matrix multiplications it is
straightforward to show that for all unitary matrices u

{αi0 , αj0 } = αi0 αj0 + αj0 αi0 = 2δi,j 1, (12.26)

and thus

αi0 αi0 = 1. (12.27)

This is a property we will use later when calculating commutators of the

Dirac equation with spin operators.
Instead of using the Dirac α matrices a common choice for notation are
the Dirac γ matrices. The α and γ matrices are related by

γ4 = α4 , (12.28)
γ1 = iα1 α4 , α1 = iγ4 γ1 ,
γ2 = iα2 α4 , α2 = iγ4 γ2 ,
γ3 = iα3 α4 , α3 = iγ4 γ3 .

363
Chapter 12. Relativistic effects

The Dirac Hamiltonian using the γ matrices becomes

3
X
H D = icγ4 γi~pi + mc2 γ4 . (12.29)
i=1

The Dirac matrices can be written as two by two block matrices of two
by two matrices. Each block then refers to the large or small component of
the Dirac four-vector. Within a block the components refer to spin up or
down. By doing so one can rewrite the Dirac matrices with the use of the
Pauli matrices. One finds for i ∈ [1, 2, 3]
 
0 σi
αi = , (12.30)
σi 0
and
 
1 0
α4 = . (12.31)
0 −1

12.4 Solutions for a free particle

The wave-function for a free particle moving in space without external
potentials in the Schrödinger equation is given as
r
1 i (~p·~r−Et)
φ(~r, t) = e~ , (12.32)
V
whereby we confined the particle to a volume V to make the wave function
normalizable. We can find a similar solution for the Dirac equation. As the
Dirac wave function is given by a vector of dimension 4 we should expect
to find 4 independent states. The wave functions representing these states
are
 
r 1
E + mc2  0 r 1 i
φ1 (~r, t) = e ~ (~p·~r−Et) , (12.33)
 
 p~z c 
2E  E+mc2  V
(~
px +i~
py )c
E+mc2

364
 12.4. Solutions for a free particle

 
r 0
E + mc2  1 r 1 i
φ2 (~r, t) = e ~ (~p·~r−Et) ,
 
 (~px −i~py )c 
2E  E+mc2  V
−~
pz c
E+mc2

 p
~z c

r E−mc2 r
E− mc2  x py2)c 
 (~
p +i~ 1 i (~p·~r−Et)
φ3 (~r, t) =  E−mc  e~ ,

2E  1  V
0

px −i~
 
(~ py )c
r E−mc2 r
E− mc2  −~pz c  1 i (~p·~r−Et)
φ4 (~r, t) = 2
e~ .
 E−mc 
2E V
 
 0 
1

These four states are normalized such that

Z
hφi |φi i = φi (~r, t)† φi (~r, t)d3~r (12.34)
V
= 1.

The variable E is the eigenvalue of the Dirac Hamiltonian. For state φ1 (~r, t)
and state φ2 (~r, t) we find
p
E1,2 = p2 c2 + m2 c4 . (12.35)

For the state φ3 (~r, t) and state φ4 (~r, t) we however find

p
E3,4 = − p2 c2 + m2 c4 . (12.36)

365
Chapter 12. Relativistic effects

12.5 Negative energy states, holes and anti-particles

This would suggest that the energy of state φ3 (~r, t) and φ4 (~r, t) is negative.
The relation between energy and momentum is given as E 2 = m2 c4 + c2 p~2 ,
which in principle would allow for negative energy solutions. Negative en-
ergy solutions and a spectrum of the Hamiltonian that is not bounded
from below is problematic. If both state φ1 and state φ3 would represent
electron states then electrons in state φ1 (~r, t) could, once we add the inter-
action with the electromagnetic field, radiatively decay into state φ3 (~r, t).
During such ap transition the electron would produce a photon with energy
~ν = 2E = 2 m2 c4 + c2 p~2 . Such transitions are experimentally not ob-
served. Electrons are experimentally stable. In order to circumvent this
problem one could assume that the one particle states with negative energy
all are filled. This would forbid electrons in states with positive energy to
decay into the negative energy states. We thus assume that the vacuum
ground state is a state with all negative energy solutions occupied. This
assumption allows for a different kind of excitations. If we assume that
all states with negative energy are occupied we can make hole excitations.
Removing an electron from the filled set of states will create a hole. These
holes in the vacuum ground state with all negative energy states occupied,
would be an effective new particle with positive charge and an energy op-
posite to the binding energy of the original electron state.
Instead of thinking of holes as missing electrons in the vacuum ground
state we can take this idea one step further and declare the hole as a new
particle that one can create from the vacuum. This is just a change of
language and does not require us to change the mathematical formalism
nor the physical predictions the theory makes. Once we define the hole
to be its own particle it becomes an equally fundamental particle as the
electron. We will call these particles the electron and positron. These two
particles are each others antiparticles. Electrons can no longer decay, but
one can make a particle hole excitation, or in our new language make a pair
excitation creating
p an electron and positron pair. A photon with energy
~ν = 2E = 2 m c + c2 p~2 can create an electron and positron, both with
2 4

energy E. The states given by wave functions φ3 (~r, t) and φ4 (~r, t) describe

366
 12.6. The non-relativistic limit

electron holes in the vacuum, i.e the positrons. The charge of a hole is
positive and it costs energy to create them.
To summarize we find that for p~ = 0 the state given by the wave function
φ1 (~r, t) describes an electron with spin up in the z direction, the wave
function φ2 (~r, t) describes an electron with spin down in the z direction, the
wave function φ3 (~r, t) describes a positron with spin up in the z direction,
and the wave function φ4 (~r, t) describes a positron with spin down in the
z direction. For finite momentum we still have 4 particles and 4 basis
states. The eigenstates of the Dirac equation at finite momentum do not
correspond to the basis states, but are given as superpositions of these
states. An electron with spin up is thus given as a linear superposition of
the 4 basis states used to write down the Dirac equation.

12.6 The non-relativistic limit

In the non-relativistic limit the Dirac equation should reproduce the Schrö-
dinger equation. One obvious difference between the Dirac and Schrödinger
equation is the fact that the Schrödinger equation acts on a wave function
that is a scalar, whereas the Dirac equation acts on a vector of dimension 4.
At rest, i.e. the true non-relativistic limit, the large component describes
the electron states. The creation of an electron hole pair costs 2mc2 with
m the rest mass of the electron. The non-relativistic limit is defined as the
limit where the energies involved are small compared to 2mc2 . We are thus
only interested in the behaviour of the electron, which for the case of zero
velocity is given by the large part of the wave function. Starting from the
Dirac equation we can derive an equation that only involves the large part
of the wave function. We start by writing
 
φ↑,l (~r, t)
φl (~r, t) = , (12.37)
φ↓,l (~r, t)

and
 
φ↑,s (~r, t)
φs (~r, t) = , (12.38)
φ↓,s (~r, t)

367
Chapter 12. Relativistic effects

for the large and small part of the wave function. With the use of the Pauli
matrices
 
0 1
σ1 = , (12.39)
1 0
 
0 −i
σ2 = ,
i 0
 
1 0
σ3 = ,
0 −1

we can write the Dirac equation as given in 12.23 as

! 
∂
+ V (~r, t) + mc2 ~

−i~ ∂t c~σ · (~p + eA) φl (~r, t)
~ ∂ · = 0. (12.40)
−c~σ · (~p + eA) i~ ∂t − V (~r, t) + mc2 φs (~r, t)

whereby we introduced the vector notation for the Pauli matrices such that

3
X
~ r, t)) =
~σ · (~p + eA(~ ~ i (~r, t)).
σi (~pi + eA (12.41)
i=1

In order to go from this equation to the Schrödinger equation we will restrict

ourselves to stationary states. We set the time dependence in the electric
and magnetic vector potential to zero such that V (~r, t) = V (~r) and take
~ r, t) = 0. Next we use that i~ ∂ φ(~r, t) = Eφ(~r, t) for stationary states of
A(~ ∂t
the Dirac equation, involving particles, such that we find

− E − V (~r) − mc2

  
c~σ · ~p φl (~r, t)
· = 0. (12.42)
E − V (~r) + mc2


−c~σ · ~p φs (~r, t)

This yields two coupled differential equations

− E − V (~r) − mc2 φl (~r, t) + c~σ · ~p φs (~r, t) = 0

(12.43)
−c~σ · ~p φl (~r, t) + E − V (~r) + mc2 φs (~r, t) = 0


(12.44)

368
 12.6. The non-relativistic limit

If we divide the second equation by E − V (~r) + mc2 we find

E − V (~r) − mc2 φl (~r, t) = c~σ · ~p φs (~r, t)

(12.45)
c
φs (~r, t) = ~σ · ~p φl (~r, t) (12.46)
E − V (~r) + mc2

The second equation shows that once, for stationary states, the large part
of the wave function is known also the small part is uniquely determined. If
we substitute the second equation into the first we find a single differential
equation for the large part of the wave function

c2
E − V (~r) − mc2 φl (~r, t) = ~σ · ~p


~σ · ~p φl (~r, t). (12.47)
E − V (~r) + mc2

This is an exact representation for stationary states of the Dirac equation

involving only the large part of the wave function.
In order to obtain the Schrödinger equation from equation 12.47 we
need to replace the relativistic energy by the non relativistic energy

E (0) = E − mc2 . (12.48)

The result is
  c2
E (0) − V (~r) φl (~r, t) = ~σ · ~p ~σ · ~p φl (~r, t). (12.49)
E (0) − V (~r) + 2mc2

We now assume that the kinetic energy of the electron E (0) − V (~r) is small
compared to the energy needed to create a particle hole excitation (electron
positron pair production) at zero momentum i.e. 2mc2 . In this limit we
can make a Taylor series in (E (0) − V (~r))/(2mc2 ) for the right hand side of
the equation. For the energy dependent term this yields

∞
!n
c2 1 X E (0) − V (~r)
= (−1)n . (12.50)
E (0) − V (~r) + 2mc2 2m 2mc2
n=0

369
Chapter 12. Relativistic effects

If we keep the terms linear in E (0) we find for the Dirac equation on the
large part of the wave function
!
  1 E (0) − V (~
r , t)
E (0) − V (~r) φl (~r, t) = ~σ · ~p 1 − ~σ · ~p φl (~r, t).
2m 2mc2
(12.51)

If we rearrange terms, realizing that ~p and V (~r) do not commute, and use
that (~σ · ~p)2 = ~p2 we find
 
1 2 1 1 2 (0)
~p + V (~r) + ~σ · ~p V (~r)~σ · ~p − ~p E φl (~r, t)
2m 4m2 c2 4m2 c2
= E (0) φl (~r, t). (12.52)

If we neglect terms proportional to 4m12 c2 we find the Schrödinger equation

for stationary states acting on the large part of the wave function.
 
1 2
~p + V (~r) φl (~r, t) = E (0) φl (~r, t). (12.53)
2m
If we neglect all relativistic corrections we find that the wave function of the
electron φe (~r, t) is equal to the large part of the wave function in the Dirac
equation φl (~r, t). In this case the Dirac equation reduces to the Schrödinger
equation.
We now want to discuss the lowest order deviations from the Schrödinger
equation. Once relativistic effects play a role we can no longer take the
wave function of the electron to be equivalent to the large part of the Dirac
spinor. We have derived above an equation on the large part of the wave
function, but we want an equation for the electron, which is given by a
linear combination of the large and small part of the wave function. We
can write

|φe i = α |φl i + β |φs i , (12.54)

with α∗ α + β ∗ β = 1. In order to find which linear combination of the large

and small part of the wave function make an electron wave function we can

370
 12.6. The non-relativistic limit

use the normalization of the wave function. We know that

1 = hφe |φe i (12.55)

= hφl |φl i + hφs |φs i .

We now use that

c
|φs i = ~σ · ~p |φl i (12.56)
E0 − V (~r, t) + 2mc2
1
≈ ~σ · ~p |φl i ,
2mc
to find
 
1 2
hφe |φe i = φl 1 + ~p φl . (12.57)
4m2 c2

For the wave function we can take

r
1
φe (~r, t) = 1 + ~p2 φl (~r, t) (12.58)
4m2 c2
!
~p2
≈ 1+ φl (~r, t)
8m2 c2

Whereby ~p2 is an operator, such that φe (~r, t) is not just a scaling of φl (~r, t)
but a different linear independent function. For φl (~r, t) we can write
 −1/2
1 2
φl (~r, t) = 1+ ~p φe (~r, t) (12.59)
4m2 c2
!
~p2
≈ 1− φe (~r, t).
8m2 c2

We can now substitute the relation between the large part of the wave
function and the electron wave function into the Dirac equation 12.52 and

371
Chapter 12. Relativistic effects

find the Dirac equation acting on the electron wave function to first order
in 4m12 c2 to be
 
1 2 1 1 2 (0)
~p + V (~r) + ~σ · ~p V (~r)~σ · ~p − ~p E
2m 4m2 c2 4mc2
! !
~p2 (0) ~p2
× 1− φe (~r, t) = E 1− φe (~r, t). (12.60)
8m2 c2 8m2 c2

In order
 to simplify
 this equation we multiply the equation from the left
p2
~
with 1 − 8m2 c2
!
~p2

1 2 1 1 2 (0)
1− ~p + V (~r) + ~σ · ~p V (~r)~σ · ~p − ~p E
8m2 c2 2m 4m2 c2 4mc2
! ! !
~p2 ~p2 ~
p 2
× 1− φe (~r, t) = 1 − E (0) 1 − φe (~r, t).
8m2 c2 8m2 c2 8m2 c2
(12.61)
1
Keeping terms to first order in mc 2 we find

1 2 1 1
~p − ~p4 + V (~r) + ~σ · ~p V (~r)~σ · ~p (12.62)
2m 8m3 c2 4m2 c2
~p2 ~p2

− V (~r) − V (~r) 2 2 φe (~r, t) = E0 φe (~r, t)
8m2 c2 8m c
For the forth term of equation 12.62 we can write
1 1
~σ · ~p V (~r)~σ · ~p = (~σ · (~p V (~r))~σ · ~p + V (~r)~σ · ~p ~σ · ~p)
4m2 c2 4m2 c2
1
−i~~σ · (∇V (~r))~σ · ~p + V (~r)~p2


=
4m2 c2
1 
~ σ · ~p + V (~r)~p2

= −ie~~σ · E~
4m2 c2
1 
~ · ~p + ~e~σ · E

~ × ~p + V (~r)~p2 .
= −ie~E
4m2 c2
(12.63)

372
 12.6. The non-relativistic limit

In the last step we used the identity for the Pauli matrices that

~ σ · ~p = E
~σ · E~ ~ · ~p + i~σ · E
~ × ~p, (12.64)

which one can derive using the commutation relations of the Pauli matrices,
or check to be valid by writing out all terms. For the fifth term of equation
12.62 we write
1 1
− 2 2
~p2 V (~r) = ~2 ∇2 V (~r) (12.65)
8m c 8m2 c2
1
= ~2 ∇ ((∇V (~r)) + V (~r)∇)
8m2 c2
1
~2 (∇2 V (~r)) + 2(∇V (~r))∇ + V (~r)∇2


= 2 2
8m c
1  2 2 
~ · ~p − V (~r)~p2 .
= ~ (∇ V (~
r )) + 2ie~E
8m2 c2
The resulting Hamiltonian is

1 2 1 4 ~2 e~ ~ × ~p.
H= ~p − 3 2
~
p + V (~
r ) + 2 2
(∇2 V (~r)) + ~σ · E
2m 8m c 8m c 4m2 c2
(12.66)

The second term in equation 12.66 is a correction to the kinetic energy

of the electron. It is known as the mass velocity term. The fourth term in
equation 12.66 is a correction to the potential energy of the electron. It is
known as the Darwin term. The fifth term couples the outer product of the
electron moment with an external electric field to the spin of the electron.
It is known as the spin-orbit coupling term. The name becomes clear if we
look at the relevant case for an electron moving in a spherical potential,
such as we encounter in atoms.

12.6.1 The non-relativistic limit in a spherical potential

In this section we want to look at the spin-orbit coupling term of equation
12.66 for a spherical potential of the form V (~r) = V (r). For a spherical

373
Chapter 12. Relativistic effects

~ we have the relation

potential V (r), due to an external electric field E
between field and potential

~ = ∂
eE V (r)r̂. (12.67)
∂r
~ due to a spherical potential into equation
If we substitute the electric field E
12.66 we find for the spin-orbit coupling term
 
1 1 ∂ ~ · L,
~
H SOC = V (r) S (12.68)
2m2 c2 r ∂r

whereby we used that the angular momentum operator is defined as L ~ =

~ 1
~r × ~p and that the spin operator is S = 2 ~~σ on the basis of spin up and
spin down used for the wave function.
In a spherical potential we can write the wave-function as a product
of a radial wave function and an angular wave function. This allows us to
look at the radial and angular part of the wave function separately. If we
define

φτ (~r) = Rτ (r)Yτ (θ, ϕ), (12.69)

for a state with quantum numbers τ we can discuss the effect of the spin
orbit coupling Hamiltonian on the angular part of the wave function. The
integrals over the radial wave function determine the coupling strength
   
1 1 ∂
ζτ = Rτ (r) V (r) R τ (r) (12.70)
2m2 c2 r ∂r

of the spin-orbit coupling. The operator acting on the angular part of the
wave function then becomes
~ · L.
H SOC = ζτ S ~ (12.71)

374
 12.6. The non-relativistic limit

0.500

0.100 5p
5d
ℏ2 ζnl [eV]

0.050 4f
4p
4d

0.010
3d
0.005 3p

2p
0.001
0 20 40 60 80
Element number Z

Figure 12.1: The atomic values for the spin-orbit coupling constant ~2 ζn,l
for the valence electrons of the different elements.

12.6.2 Spin orbit coupling constant in atoms

With the aid of numerical methods and computer implementations we can

find approximate solutions for the wave-functions of the atoms in the pe-
riodic table. If we approximate the atomic N electron wave function to
be single Slater determinant representable we can write the many electron
state as a product state of electrons created into a single electron spin-
orbital. Each of these spin-orbitals is a solution of an electron propagating
in a spherical effective potential. The potential is determined by the nu-
clear charge and the charge density of all other electrons. The spin-orbital
wave function for each electron is the product of an angular times radial
part. The angular part of the atomic spin-orbitals is given by the states dis-
cussed in section 8.6.3. The radial part needs to be solved self-consistently

375
Chapter 12. Relativistic effects

and numerically. With the radial wave function and the effective potential
one can calculate the spin-orbit coupling constant, ζnlj as given in equation
12.70.
In figure 12.1 we show the size of the spin-orbit coupling constant for the
valence electrons of the elements in the periodic table. We approximated
the N = Z electron ground state as the single Slater determinant repre-
sentable state with lowest energy. For Boron, with one electron in the 2p
shell the coupling constant is the smallest and slightly larger than 1 meV.
One generally finds that the heavier the element the larger the spin-orbit
coupling constant is. We furthermore find that each time a new shell starts
the coupling constant is considerably reduced. We can understand these
trends from basic considerations. Spin-orbit coupling is a relativistic effect.
It should become larger when the electrons have large velocities. The larger
the nuclear charge, the faster a classical electron will rotate around the nu-
cleus and thus the more important relativistic corrections become. If one
starts with a new shell, then the electrons of the previous shell will be closer
to the nucleus and thus shield the nuclear potential. The effective potential
seen by the outer electrons is thus smaller, thereby reducing the angular
velocity of the electrons. Although a picture based on classical behaviour
of the electrons, it does give a basic feeling of what happens.
The numerical values presented here agree well with experimental val-
ues. These values are however notably different from theoretical values one
obtains when neglecting the interactions between the electrons. One can
compare the numerical results including the interaction between electrons
to the values one finds for a single electron in a potential of the form

Ze
V (r) = . (12.72)
4πr
This is the potential one finds for an effective H atom with nuclear charge
Z. The radial wave function in this case is given as
 l+ 3 s
Z 2 (n − l − 1)! − arZn sl+1 2rZ
Rn,l (r) = 2l+1 n−2−l rl e 0 Ln−l−1 ( ). (12.73)
a0 (n + l)! a0 n

376
 12.7. Symmetries and conservation laws for the Dirac
equation

For the spin-orbit coupling constant we find

   
1 1 ∂
ζn,l = Rn,l (r) V (r) Rn,l (r) (12.74)
2m2 c2 r ∂r
e2 Z4
= ,
4π0 a30 c2 m2 n3 l(l + 1)(2l + 1)

This value is about 10 times larger for the last element of each sub shell than
the numerical values we obtained. Furthermore is about 100 times larger
than the numerical values for the first element of each sub shell. This
discrepancy is due to the screened effective potential as described above.
The analytical result does explain the jumps in the trend each time a new
sub-shell is started. The coupling constant scales with Z 4 , but also with
1/n3 and 1/l3 .

12.7 Symmetries and conservation laws for the

Dirac equation
Constants of motion are given by operators that commute with the Dirac
~ ~r) = 0 and Φ(t, ~r) = 0, equation 12.19) or
Hamiltonian either without (A(t,
with (equation 12.23) electro magnetic field
   
H D = c~ ~ r, t) + V (~r, t) + α4 mc2 ,
α · ~p + eA(~ (12.75)

with ~p = −i~∇, and V (~r, t) = −eΦ(~r, t). By the Noether theorem sym-
metry operators that commute with the Hamiltonian relate to a conserved
quantity (see chapter 7 for more detail).

12.7.1 Space and time translation symmetry or Momentum

and Energy conservation
~ r, t) and Φ(~r, t) are constant in ~r and t, then both the space trans-
When A(~
lation operator T ~r0 : ~r 7→ ~r − ~r0 (equation 7.18), and the time translation

377
Chapter 12. Relativistic effects

operator T δt : t → t − δt (equation 7.27) commute with the Dirac Hamilto-

nian. We can readily show this if we represent the space and time transla-
i
tion operators in their differential form T ~r0 = e− ~ ~r0 ·~p (equation 7.22) and
i
T δt = e−δt ~ H D (equation 7.33). The term α4 mc2 in the Dirac Hamiltonian
is a constant in ~r and as such commutes with all components of the vector
operator ~p. The other term in the Dirac equation is given by c~ α · ~p and
as such actually represents 3 terms, each of them is proportional to one of
the components of ~p. All components of the momentum operator commute
with each other. We find that the momentum operator commutes with the
Dirac Hamiltonian. Therefor also all potentials of the momentum operator
commute with H D and in particular

[H D , T ~r0 ] = 0, and (12.76)

[H D , T δt ] = 0.

As a result of these symmetries, by Noether’s theorem, energy and momen-

tum are conserved quantities of the Dirac equation. To be more precise:
Given an arbitrary state |ψ(t)i, represented by the four-vector ψ(~r, t), that
follows the dynamics given by the Dirac equation, we know that for all
n ∈ N,
∂
hψ(t)|~pn |ψ(t)i = 0, and (12.77)
∂t
∂
hψ(t)|H nD |ψ(t)i = 0,
∂t
i.e. all moments of the energy and momentum distribution functions µE
and µp~ and thus the distributions themselves are conserved.
For an arbitrary state |ψ(t)i the energy and momentum are given by
distribution functions µE and µp~ (see section 6.5). Within the Copenhagen
interpretation of quantum mechanics the chance to measure a specific value
of the energy E or momentum p~ for a particle in state |ψ(t)i is given by
these distribution functions. Within the Dirac equation the distribution
functions of energy and momentum are conserved quantities. This is a
property that the Dirac and Schrödinger equation have in common.

378
 12.7. Symmetries and conservation laws for the Dirac
equation

12.7.2 The continuity equation

In section 3.1 we discussed the continuity equation that holds for particles
in quantum mechanics. Particles do not jump from one point in space to an
other point, but follow a continues flow. Mathematically we derived this by
looking at the change of density dρ(~r, t) in a volume V . We realized that
for all volumes the change in density in volume V is equal to the current
density ~j(~r, t) that flows out through the surface S of the volume V . The
resulting continuity equation is

∂
ρ(~r, t) = −∇ · ~j(~r, t), (12.78)
∂t
with for the Schrödigner equation the density defined as

ρ(~r, t) = φ(~r, t)∗ φ(~r, t) (12.79)

and the current density is, following section 3.1 defined as

~j(~r, t) = 1 φ(~r, t)∗~pφ(~r, t) − φ(~r, t)~pφ(~r, t)∗ . (12.80)

2m
For the Dirac equation the wave function is a vector that assigns 4
complex numbers to each point in space. We can still define a conserved
density, similar to the Schrödigner equation as

ρ(~r, t) = φ(~r, t)† φ(~r, t). (12.81)

In order to define the current density we can look for a definition that fulfills
the continuity equation. We have

∂
−∇ · ~j(~r, t) = ρ(~r, t) (12.82)
∂t
∂
= φ(~r, t)† φ(~r, t)
∂t
   
∂ † † ∂
= φ(~r, t) φ(~r, t) + φ(~r, t) φ(~r, t)
∂t ∂t

379
Chapter 12. Relativistic effects

Using the Dirac equation we have

∂ 1
α · ~p + α4 mc2 φ(~r, t)


φ(~r, t) = c~ (12.83)
∂t i~ α4 2 
= −c~ α·∇+ mc φ(~r, t).
i~
For the time derivative of the Hermitian conjugate of the wave function we
find
∂  α4 2  †
φ(~r, t)† = −c~ α·∇+ mc φ(~r, t) (12.84)
∂t  i~
 α4
= − ∇φ(~r, t)† · α~ c − φ(~r, t)† mc2 .
i~
Substituting these results into equation 12.82 for the divergence of the
current density we find
   
−∇ · ~j(~r, t) = −c φ(~r, t)† α
~ · (∇φ(~r, t)) + ∇φ(~r, t)† · α
~ φ(~r, t) (12.85)
 
= −c∇ φ(~r, t)† α~ φ(~r, t) .

From which it follows that

~j(~r, t) = cφ(~r, t)† α
~ φ(~r, t). (12.86)

The current density of a plane wave solutions

In order to get a feeling of the definition of the current in the Dirac equation
we can have a look at the current transported by the stationary plane wave
solutions of the Dirac equation without potential. pFor a particle with spin
up traveling with momentum p~ and energy E = p2 c2 + m2 c4 we have
 
r 1
E + mc2  0  i
 ~ (~p·~r−Et)
φ(~r, t) = e . (12.87)

p z c
2E

 E+mc2 
(px +ipy )c
E+mc2

380
 12.7. Symmetries and conservation laws for the Dirac
equation

Following equation 12.86 we find

~j(~r, t) = cφ(~r, t)† α
~ φ(~r, t) (12.88)
c2 p~
= .
E
We now use that we can relate the relativistic momentum p~ to the velocity
of a particle with the use of the Lorentz factor γ = √ 1 2 2 as
1−v /c

p~ = γm~v . (12.89)

With this relation we can write

E2 p2
= + m2 (12.90)
c4 c2
m2 γ 2 v 2
= + m2
c2
2

2 2v
=m 1+γ 2
c
v 2 /c2
 
2
=m 1+
1 − v 2 /c2
2 2 v 2 /c2
 
2 1 − v /c
=m +
1 − v 2 /c2 1 − v 2 /c2
= m2 γ 2 .

For the current density we thus find

~j(~r, t) = p~
(12.91)
E/c2
p~
=
mγ
= ~v .

The current density of a plane wave solution represent the velocity of the
particle.

381
Chapter 12. Relativistic effects

12.7.3 Rotational symmetry or Angular momentum conser-

vation
Within the Schrödinger equation we encountered two types of angular mo-
mentum. The orbital angular momentum L ~ related to the rotation of mas-
sive particles around a position ~r0 and the spin angular momentum S, ~ which
~ ~
is intrinsic to the particles. Both L and S commute with the Schrödinger
equation and are thus related to conserved quantities. Neither of the two
commute with the Dirac equation.
For the orbital angular momentum we recall that L ~ = ~r × ~p, ~p = −i~∇,
and [~pi , ~rj ] = −i~δi,j . For the commutator with the Dirac Hamiltonian we
find
h i
HD, L~ = c [~ α · ~p, ~r × ~p] (12.92)
  
X 3 y~pz − z~py
= c αi~pi , z~px − x~pz 
i=1 x~py − y~px
 
[αy ~py , y]~pz − [αz ~pz , z]~py
= c  [αz ~pz , z]~px − [αx~px , x]~pz 
[αx~px , x]~py − [αy ~py , y]~px
 
αy ~pz − αz ~py
= −i~c αz ~px − αx~pz 
αx~py − αy ~px
= −i~c~
α × ~p,

which is non-zero.
For the spin angular momentum we define the Pauli matrices acting on
a Dirac four-vector as
 
0 1 0 0
1 0 0 0
Σx = iα3 α2 = 
0 0 0 1 ,
 (12.93)
0 0 1 0

382
 12.7. Symmetries and conservation laws for the Dirac
equation

 
0 −i 0 0
i 0 0 0 
0 0 0 −i ,
Σy = iα1 α3 =  

0 0 i 0
 
1 0 0 0
0 −1 0 0 
Σz = iα2 α1 = 
0 0 1 0  .


0 0 0 −1
The spin operator on the same basis as the one defined for the Dirac α
matrices then has the matrix representation

~ = 1 ~Σ.
S ~ (12.94)
2
With these matrices and the definition of the spin operator on the basis
of a Dirac four-vector we can calculate the commutation relations between
the spin operator and the Dirac Hamiltonian
h i  1

~ = c~
HD, S 2
α · ~p + mc α4 , ~Σ ~ (12.95)
2
  
3 iα3 α2
X 1
= c αi~pi + mc2 α4 , ~ iα1 α3 
2
i=1 iα2 α1
  
3 α3 α2
1 X
= i~c ~pi αi , α1 α3 
 
2
i=1 α2 α1
 
αy ~pz − αz ~py
= i~c αz ~px − αx~pz 
αx~py − αy ~px
α × ~p.
= i~c~

Neither the spin nor the orbital angular momentum is conserved within
the Dirac equation. Comparing equation 12.92 and 12.95 it is straight-
forward to see that the sum of the spin and orbital angular momentum

383
Chapter 12. Relativistic effects

~ =L
J ~ +S ~ does commute with H D
h i
~ = 0.
HD, J (12.96)

The spin of a particle truly represents an angular momentum and for a

system that has rotational symmetry the total angular momentum, i.e.
spin plus orbital, is conserved.

12.7.4 Helicity of a particle

The orbital angular momentum of a particle, in the direction it travels, is
always zero. In classical physics this is easily seen by the fact that (~r×~
p) ⊥ p~
or equivalent (~r × p~) · p~ = 0. In quantum mechanics this still holds and
~ · ~p = (~r × ~p) · ~p
L (12.97)
= 0.

The operator L ~ · ~p acting on any state gives zero. If the orbital angular
momentum along a certain direction is zero, then the total angular momen-
tum along this direction is given by the spin angular momentum. Whilst
the total angular momentum is conserved we should expect that the spin
angular momentum along the direction the particle travels is a conserved
quantity. This is known as the helicity of the particle. For systems where
~ r, t) and Φ(~r, t) are constants such that [H D , ~p] = 0 we find that
A(~
h i h i
HD, S ~ · ~p = H D , S~ · ~p (12.98)
 
αy ~pz − αz ~py
= i~c αz ~px − αx~pz  · ~p
αx~py − αy ~px
= 0.

The helicity of a particle, given by the operator S ~ · ~p thus represents a

conserved quantity. Note that the helicity is a conserved quantity within
one reference frame, but the helicity of a massive particle is not an invariant
under a Lorentz transformation.

384
 12.7. Symmetries and conservation laws for the Dirac
equation

12.7.5 Relativistic Quantum number κ

In order to describe the electron eigenstates of the hydrogen atom, for the
Schrödinger equation we used 5 quantum numbers: n, l, lz , s, and sz . For
the angular part the operators L2 , Lz , S 2 , and S z all commute with each
other and with the Hamiltonian. Furthermore the combined eigenstates of
these operators had, for a given principle quantum number n no degen-
erate states, i.e. no two linear independent states have the same set of
quantum numbers. We thus could use these operators to uniquely identify
the eigenstates of the Schrödinger Hamiltonian. For the Dirac equation the
operators L~ and S ~ do not commute with the Dirac Hamiltonian. We thus
can not use the eigenstates of the operators L2 , Lz , S 2 , and S z to find the
eigenstates of the Dirac Hamiltonian. Instead we will have to look at the
eigenstates of J 2 and J z . We can however readily see that these two oper-
ators do not uniquely label all eigenstates of the Dirac Hamiltonian for the
hydrogen atom within a given atomic shell with principle quantum number
n. For a state with total angular momentum j we can in the non-relativistic
picture, either have a state with l = j − 1/2 or l = j + 1/2 coupled with the
spin of an electron s = 1/2 to form a total angular momentum j. In order
to label the different states with the same value of j, we need an additional
quantum number and operator. In the previous section we saw that the
spin along the direction of motion for a particle is a conserved quantity.
We can look for a similar definition for the spin projection along the total
angular momentum. For this we define the operator
~2
 
2 ~ ~
K = α4 S·J − (12.99)
~ 4
 
~ ~ 1
= α4 J · Σ − ~ ,
2
~ instead of J
or equivalently defined with operator L ~
 
2 ~ ~ 1
= α4 L · S + ~2 (12.100)
~ 2
 
= α4 L ~ ·Σ
~ +~ ,

385
Chapter 12. Relativistic effects

whereby we used that the operator S ~ 2 = 3 ~2 on the basis of a Dirac four-

4
vector for a spin 1/2 particle.
We now want to show that K commutes with both J ~ and H D such
that we can use the eigenstates of K and J z to find the eigenstates of the
Dirac equation for systems with spherical symmetry.
For the commutator of K with H D we find

~2
  
2 ~ ~
[H D , K] = H D , α4 S·J − (12.101)
~ 4
~2 ~2
     
2 ~ ~ 2 ~ ~
= [H D , α4 ] S·J − − α4 S·J − , HD
~ 4 ~ 4
~2
 
2 ~ ~ 2 h~ i
~,
= [H D , α4 ] S·J − − α4 S, HD · J
~ 4 ~

whereby we used that for 3 operators A, B and C it holds that [A, BC] =
[A, B]C − B[C, A] and that [H D , J~ ] = 0. The commutator between H D
~ is given in equation 12.95. The commutator between H D and α4 is
and S

α · ~p, α4 ]
[H D , α4 ] = [c~ (12.102)
" 3 #
X
= c αi~pi , α4
i=1
3
X
=c ~pi [αi , α4 ]
i=1
3
X
= −2cα4 αi~pi
i=1
= −2cα4 α
~ · ~p,

such that we can simplify the commutator between H D and K to

~2
 
4 ~ ~ ~
[H D , K] = − cα4 α
~ · ~p S · J − α × ~p) · J
+ α4 2ic (~ (12.103)
~ 4

386
 12.7. Symmetries and conservation laws for the Dirac
equation

 
3 X
3 3
X
~ j + i (~ ~ −
X ~
= −2cα4  αi~pi Σj J α × ~p) · J αi~pi
2
i=1 j=1 i=1
  
α1 α3 α2 α2 α3 α2 + α3 α3 α3 α2 − α2
= −2cα4 i~p · α1 α1 α3 − α3 α2 α1 α3 ~
α3 α1 α3 + α1  · J
α1 α2 α1 + α2 α2 α2 α1 − α1 α3 α2 α1

~
−~ α · ~p
2
 
~ +α ~
= 2cα4 iα1 α2 α3~p · J ~ · ~p
2
 
~ ~ ~
= 2cα4 iα1 α2 α3~p · (L + S) + α ~ · ~p
2
 
~ +α ~
= 2cα4 iα1 α2 α3~p · S ~ · ~p
2

~ ~ +α

= 2cα4 iα1 α2 α3~p · Σ ~ · ~p
2   
α3 α2
~
= 2cα4 −α1 α2 α3 α3 α2  · ~p + α ~ · ~p
2
α3 α2
   
α1
~
= 2cα4 − α2  · ~p + α ~ · ~p
2
α3
= 0.

For the commutator between K and J ~ we find

i  2
  
~ = α4 2 S
h
K, J ~ − ~ ,J
~ ·J ~ (12.104)
~ 4
~ 2S ~ + α4 2 S
h i h i h i
= α4 , J ~ ·J ~ ·J
~,J ~ − ~ α4 , J
~ .
~ ~ 2
h i h i
If we can show that α4 , J~ = 0 and S ~ ·J
~,J
~ = 0 then we showed that

387
Chapter 12. Relativistic effects

h i h i
K, J ~ = 0, which is what we wanted to show in the first place. For α4 , J
~
we find
h i h i
~ = α4 , L
α4 , J ~ +S ~ (12.105)
h i
~
= α4 , S
1 h ~
i
= ~ α4 , Σ
2
= 0.

For the commutator between S ~ ·J

~ and J
~ we find
h i h i h i
S~ ·J
~,J
~ = S ~ · L,
~ J~ + S~ · S,
~ J~ (12.106)
   h
1 ~ 2 ~ 2 ~ 2 ~ ~ · S,
~ J
i
~ .
= J − L − S ,J + S
2
In order to show that this is zero we will have a look at the commutator
between J~ and L~ 2 or S~ 2 . On the basis of a Dirac four-vector the spin
operator squared is given as
2
~2 = ~ Σ
S ~2 (12.107)
4
~2
= − (α3 α2 α3 α2 + α1 α3 α1 α3 + α2 α1 α2 α1 )
4
3~2
= .
4
h i
This is just a number such that S ~ · S,
~ J~ = 0. For the commutator be-

tween L~ 2 and J ~ we can write J~ =L ~ + S.

~ The commutator between L ~2
2
and L~ is trivially zero. The operator L
~ is given by a scalar operator times
the identity matrix on the basis of the Dirac four-vector space such that it
commutes with all α matrices and thus with the operator S. ~ We conclude
that
h i
K, J~ = 0. (12.108)

388
 12.8. Solutions for the hydrogen atom

 
In summary we have shown that the operator K = α4 ~2 S ~ − ~2
~ ·J
4
commutes with both the Dirac Hamiltonian H D and the total angular
~.
momentum of the system J

12.8 Solutions for the hydrogen atom

An hydrogen atom consists of two particles, a proton and an electron,
with mass and position coordinates mp , me , ~rp and ~re respectively. In
the non-relativistic case we transformed the Schrödinger equation for these
two particles to a new differential equation representing the centre of mass
coordinates R ~ = (mp~rp + me~re )/(mp + me ) and the difference coordinates
~r = ~re − ~rp . In these new coordinates we find two decoupled differential
equations. The differential equation in the centre of mass coordinate system
describes a free particle of mass M = me + mp , i.e. the hydrogen atom.
The differential equation in the relative mass coordinate system describes
the bound state or a particle with charge e in a central potential V (r) =
−e2 /(4π0 r) and reduced mass µ = me mp /(me + mp ). See section 8.1 for
more detail.
Within Dirac’s formalism there is no simple equivalence. We can not
simply make a coordinate transformation of the Dirac equation to centre
of mass and relative coordinates. One problem is that the Dirac equation
describes spin 1/2 particles. Both the proton and the electron are spin 1/2
particles. However if we would look at the centre of mass coordinates then
the particle that propagates with these coordinates is the hydrogen atom.
This is a coupled state of a proton and electron. The spin of the combined
particles is the vector coupling of the spin of the electron and proton. The
coupled spin of the electron and proton is either 0 or 1 and never 1/2.
We will in this chapter, not deal with the problem of the finite nuclear
mass, but instead assume that the nuclear mass is infinite and discuss the
problem of a relativistic electron propagating in a potential of the form
V (r) = −Ze2 /(4π0 r). This also means we should expect our energies to
be off by a factor of roughly 1/(1 + me /M ) with M the mass of the nucleus,
and M = mp for a hydrogen atom. The Dirac Hamiltonian we want to

389
Chapter 12. Relativistic effects

solve, i.e. find the eigenstates of is given as

α · ~p + α4 mc2 + V (r),
H D = c~ (12.109)
with
Ze2
V (r) = − . (12.110)
4π0 r
For the derivation of the eigenstates of the Dirac Hamiltonian we will ex-
tensively use the results we know for the non-relativistic case as derived in
chapter 8.

12.8.1 Spherical coordinates

We start by transforming the Dirac equation to spherical coordinates such
that we can look at the angular and radial solutions separately. The term
α4 mc2 and V (~r) already are in spherical coordinates the only part we need
to transform is c~α · ~p for which we find
 
i
~ · ~p = αr pr + α4 K ,
α (12.111)
r
with
1
αr = α~ · ~r (12.112)
r
and
1 ∂
pr = −i~ r. (12.113)
r ∂r
One way to prove the representation of α ~ ·~p is to explicitly write the operator
on a basis of a Dirac four-vector as a 4 by 4 matrix in Cartesian coordinates.
For the left hand side of equation 12.111 we find
 
0 0 pz px − ipy
 0 0 px + ipy −pz 
~ · ~p = 
α . (12.114)
 pz px − ipy 0 0 
px + ipy −pz 0 0

390
 12.8. Solutions for the hydrogen atom

For the right hand side of equation 12.111 we find

   
i 1 ∂ i ~ ~
αr pr + α4 K = αr −i~ r+ Σ·L+~ (12.115)
r r ∂r r
 
1 ∂ i ~ ~
= αr −i~ − i~ + Σ·L+~
r ∂r r
 
∂ i~
= αr −i~ + Σ · (~r × ~p) .
∂r r
∂
For the operator −i~ ∂r we can write

∂ ∂x ∂ ∂y ∂ ∂z ∂
−i~ = −i~ − i~ − i~ (12.116)
∂r ∂r ∂x ∂r ∂y ∂r ∂z
∂x ∂y ∂z
= p + p + p
∂r x ∂r y ∂r z
1
= ~r · ~p
r
We now write out the αr as a matrix
 
0 0 z x − iy
1 0 0 x + iy −z 
αr =  , (12.117)
r z x − iy 0 0 
x + iy −z 0 0

and find for the operator αr 1r ~r · ~p the matrix expression

xzpx + yzpy + z 2 pz
 
0 0 (x − iy)~r · ~p
1 1 0 0 (x + iy)~r · ~p −xzpx − yzpy − z 2 pz 
αr ~r · ~p = 2  2 (x − iy)~r · ~p
 .
r r xzpx(x++yzp y + z pz
iy)~r · ~p −xzpx − yzpy − z 2 pz
0
0
0
0


(12.118)

For the second term in equation 12.115 we can use that

 
yp − zpy
i~ i~  z
Σ · (~r × ~p) = Σ · zpx − xpz  (12.119)
r r
xpy − ypx

391
Chapter 12. Relativistic effects

 
Z X − iY 0 0
i X + iY −Z 0 0 
=  ,
r  0 0 Z X − iY 
0 0 X + iY −Z

with (X, Y, Z) = ~r × ~p. We now multiply this equation from the left with
αr to find
 
(ixy + y 2 + z 2 )px +

 0 0 −xzpx − yzpy + (x2 + y 2 )pz (−xy − ix2 − iz 2 )py −  

 (x − iy)zpz 

 (−ixy + y 2 + z 2 )px + 

i~ 1
 

 0 0 (−xy + ix2 + iz 2 )py − xzpx + yzpy − (x2 + y 2 )pz  
αr Σ · (~r × ~p) = 2 (x + iy)zpz
.
 
 
(ixy + y 2 + z 2 )px +
r r
 
 
−xzpx − yzpy + (x2 + y 2 )pz (−xy − ix2 − iz 2 )py − 0 0
 

(x − iy)zpz
 
 
 (−ixy + y 2 + z 2 )px +
 

(−xy + ix2 + iz 2 )py − xzpx + yzpy − (x2 + y 2 )pz
 
 0 0 
(x + iy)zpz

(12.120)

We now add equation 12.118 to equation 12.120 together. We find for the
right hand side of equation 12.111 the expression
 
i 1 0 0 (x + y + z )p (x + y + z )(p − ip )
 2 2 2
z
2 2 2
x y

αr pr + α4 K = 2  (x + y 0+ z )p (x + y + z0 )(p − ip ) (x + y + z0 )(p − ip ) (x + y 0+ z )p 

2 2 2 2 2 2
2 2 2
x y
2 2 2
z

r r (x + y + z )(p − ip ) (x + y + z )p
2 2 2
x
z
y 0 2 20
x
2
z
y

 
0 0 pz (px − ipy )
= 


0
pz
0
(px − ipy )
(px − ipy )
0
pz
0



(px − ipy ) pz 0 0

~ · ~p,
=α (12.121)

which proofs that equation 12.111 indeed represents the kinetic energy term
of the Dirac equation in spherical coordinates. Putting it all together we
find for the Dirac equation represented on spherical coordinates
 
i
H D = cαr pr + α4 K + α4 mc2 + V (r). (12.122)
r

392
 12.8. Solutions for the hydrogen atom

12.8.2 Angular solutions

We know that the eigenfunctions of equation 12.122 are simultaneous eigen-
functions of K, J 2 and J z . Whereby we picked an arbitrary component of
~ to define a set mutually commuting operators. We know from the eigen-
J
states of the Schrödinger equation that the non-relativistic eigenstates can
uniquely be labeled by the quantum numbers n, l, lz , s, and sz . For the
Dirac equation L ~ and S ~ do not commute with the Hamiltonian, but only
the total angular momentum J ~ does. We can label the states by n, j, and
jz , but this does not uniquely label all states. If we couple the orbital an-
gular momentum l with the spin angular momentum we find, within the
non-relativistic limit, two different sets of states with the same j. The s
electrons (l = 0) couple with the spin of the electron to a j = 1/2. The p
electrons l = 1 couple with the spin of the electron to a j = 1/2 or to a
j = 3/2. The d electrons couple to j = 3/2 or j = 5/2. In general a state
with orbital angular momentum l can couple with the spin of s = 1/2 to
a total angular momentum of l ± 1/2. We thus find for j = 1/2 either the
s1/2 or p1/2 states, for j = 3/2 either the p3/2 or the d3/2 states, etc.
We now can use the eigenvalues of operator K to assign each state a
unique set of quantum numbers. We define the eigenvalues of K to be −κ~
such that if ψ(~r) is an eigenstate of the Dirac Hamiltonian then
Kψ(~r) = −κ~ψ(~r). (12.123)
In order to solve the angular solutions we will have a closer look at the
eigenvalues of K. For this we will express the operator K 2 in terms of J 2
and a constant. We find
  2
K 2 = α4 Σ ~ ·L~ +~ (12.124)
 2
= Σ ~ +~
~ ·L
 2
= Σ ~ + 2~Σ
~ ·L ~ + ~2
~ ·L
3 X
X 3
= Σi Σj Li Lj + 2~Σ ~ + ~2 .
~ ·L
i=1 j=1

393
Chapter 12. Relativistic effects

We now use that

Σ1 Σ1 = −α3 α2 α3 α2 (12.125)
= 1,

and similar Σ2 Σ2 = 1 and Σ3 Σ3 = 1. For the terms where i 6= j we find

Σ1 Σ2 = −α3 α2 α1 α3 (12.126)
= −α2 α1
= iΣ3 ,

and similar Σ2 Σ3 = iΣ1 and Σ3 Σ1 = iΣ2 . With these equations on the Σ

matrices we find
 
~ · L
K 2 = L2 + i Σ ~ ×L
~ + 2~Σ~ ·L~ + ~2 . (12.127)

~ ×L
Note that L ~ 6= 0, but
 
Ly Lz − L z L y
~ ×L
L ~ =  L z Lx − L x Lz  (12.128)
L x Ly − L y L x
 
[Ly , Lz ]
= [Lz , Lx ]
[Lx , Ly ]
~
= i~L,

such that

K 2 = L2 − ~Σ ~ + 2~Σ
~ ·L ~ + ~2
~ ·L (12.129)
~ + ~2 .
~ ·L
= L2 + ~Σ

We now use that

J 2 = (L + S)2 (12.130)

394
 12.8. Solutions for the hydrogen atom

~ ·L
= L2 + 2 S ~ + S2

= L2 + ~Σ ~ + 3 ~2 ,
~ ·L
4
such that we, by comparison of K 2 to J 2 find the expression,
1
K 2 = J 2 + ~2 (12.131)
4
With this relation we can rewrite κ in terms of j. If ψ(~r) is an eigenstate
of H D , then ψ(~r) is also an eigenstate of K and J 2 . With eigenvalues −κ~
and j(j + 1) respectively. We now can use the expression of K 2 in terms
of J 2 to find

K 2 ψ(~r) = κ2 ~2 (12.132)
1
= j(j + 1)~2 + ~2 .
4
It follows that
1
κ2 = j(j + 1) + , (12.133)
4
or
 
1
κ=± j+ . (12.134)
2

The eigenvalues of K are thus related to the eigenvalues of J 2 . Where j is

always positive, as it relates to the length of the angular momentum vector
κ can be both positive or negative. If κ is known j can be determined by
1
j = |κ| − . (12.135)
2
We discussed before that the quantum numbers j and jz do not uniquely
determine the angular wave function, as there are two sets of states with the
same value for j. These are in the non-relativistic limit related to j = l+1/2
and j = l − 1/2. We will now show that the sign of κ is negative when, in

395
Chapter 12. Relativistic effects

the non-relativistic limit j = l + 1/2 and positive when j = l − 1/2. In the

non-relativistic limit only the large part of the wave function is non-zero.
In this limit we can replace the matrix α4 by the identity matrix. We then
find that the operator K acting on the large part only becomes

2~ ~
K→ S · L + ~2 (12.136)
~
1 2
J − L2 − S 2 + ~2


=
~ 
1 2 2 1 2
= J −L + ~ ,
~ 4

whereby we used in the last step that the spin of an electron is 1/2 ,
such that S 2 = 3/4. If we now use this operator to calculate, in the non-
relativistic limit, the expectation value of K for an eigenstate with quantum
numbers l, s, and j = l ± 1/2 we find
 
1 1 1
hψlsj |K|ψlsj i = ~ (l ± )(l + 1 ± ) − l(l + 1) + (12.137)
2 2 4
 
1 1
= ~ ±(l + ) + .
2 2

We thus find that

 
1 1
κ=∓ l+ + (12.138)
2 2
 
1
=± j+ ,
2

whereby κ < 0 when j > l and κ > 0 when j < l, i.e. κ = −1 corresponds
to the s1/2 state, κ = +1 corresponds to the p1/2 state, κ = −2 corresponds
to the p3/2 state, κ = +2 corresponds to the d3/2 state, etc.
Note that for the relativistic case L does not commute with the Hamil-
tonian. The operator K does commute with the Hamiltonian. We can
connect the non-relativistic limit to the relativistic case by changing the

396
 12.8. Solutions for the hydrogen atom

speed of light c form infinite to its real value. If we do so the wave func-
tions will change in a continues fashion. The eigenvalues of operator K for
eigenstates of the Dirac equation, will not change when changing c form
infinite to its real value. Such that our discussion that the sign of κ is
different for the two solutions with the same expectation value of J 2 still
holds also for the relativistic solution.
We now have a unique set of quantum numbers to label the different
angular states. Next we want to determine the angular wave functions. For
this we will show that although the eigenstates of the Dirac Hamiltonian are
not eigenstates of the operator L2 , but the small and large part separately
are eigenstates of L2 , albeit with different eigenvalues.
For the operator L2 we can write

L2 = (J − S)2 (12.139)
2 2
= J + S − 2S · J ,

with K = α4 ~2 S · J − 14 ~2 we find

1
L2 = J 2 + S 2 − α4 ~K − ~2 (12.140)
2
2 1 2
= J − α4 ~K + ~ .
4
We take ψκ,jz (θ, φ) to be an eigenfunction of the Dirac Hamiltonian. We
find
 
1
L2 ψκ,jz (θ, φ) = j(j + 1)~2 + α4 ~2 κ + ~2 ψκ,jz (θ, φ). (12.141)
4

We now use that κ = ±(j + 1/2) to find

   
2 2 1 1
L ψκ,jz (θ, φ) = ~ j(j + 1) ± α4 j + + ψκ,jz (θ, φ). (12.142)
2 4

With α4 a 4 by 4 matrix and ψκ,jz (θ, φ) a vector of length 4. We now take

(A) (B)
ψκ,jz (θ, φ) and ψκ,jz (θ, φ) to be vectors of length 2, related to the large and

397
Chapter 12. Relativistic effects

small part of the wave function, such that

(A)
!
ψκ,jz (θ, φ)
ψκ,jz (θ, φ) = (B) , (12.143)
ψκ,jz (θ, φ)

and
 
12 0
α4 = . (12.144)
0 −12

For the operator L2 acting on an eigenstate of the Dirac Hamiltonian we

find
(A)
!
1
(A) !
1
ψ (θ, φ) j(j + 1) ± j + + ψ (θ, φ)
L2 κ,jz
(B) = ~2 2
4
κ,jz
(B) (12.145)
ψκ,jz (θ, φ) j(j + 1) ∓ j + 21 + 14 ψκ,jz (θ, φ)

(A) !
1 1


j ± j ± + 1 ψ (θ, φ)
= ~2 2
2
κ,j
(B)
z

j ∓ 21 j ∓ 21 + 1 ψκ,jz (θ, φ)

(A) (B)
Both ψκ,jz (θ, φ) and ψκ,jz (θ, φ) are eigenstates of L2 albeit with different
eigenvalues. We find
(
j − 12 κ < 0
lA = , (12.146)
j + 12 κ > 0

and
(
1
j+ 2 κ<0
lB = 1
. (12.147)
j− 2 κ>0

For κ = −1 we find that the large part of the wave function has the angular
dependence of a s1/2 state, the small part of the wave function has the
angular dependence of a p1/2 state. For κ = +1 this is reversed and the
large part of the wave function has the angular dependence of a p1/2 state,
the small part of the wave function has the angular dependence of a s1/2

398
 12.8. Solutions for the hydrogen atom

state. For κ = −2 we find that the large part of the wave function has the
angular dependence of a p3/2 state, the small part of the wave function has
the angular dependence of a d3/2 state, etc.
In order to find explicit functions of the angular dependence we realize
that the angular dependence is given by the coupling of a state with l =
j ± 1/2 and s = 1/2 to a state with total angular momentum j. In general
we can express this coupling with the use of Clebsch-Gordan coefficients,
see section 7.5.7 and equation 7.148 for explicit values. In general we can
write
1/2
j,jz
X
|lsjjz i = Cl,lz =jz −sz ,s,sz
|l, jz − sz i |ssz i . (12.148)
sz =−1/2

If we write out the sum explicitly this becomes

j,jz
|lsjjz i = Cl,jz −1/2,s,+1/2
|l, jz − 1/2i |↑i (12.149)
j,jz
+ Cl,jz +1/2,s,−1/2
|l, jz + 1/2i |↓i .

We now define the vector function Yl,j,jz (θ, ϕ) as an explicit representation

of the state |lsjjz i
j,jz
!
Cl,j z −1/2,s,+1/2
Yl,jz −1/2 (θ, ϕ)
Yl,j,jz (θ, ϕ) = j,jz (12.150)
Cl,j Y
z +1/2,s,−1/2 l,jz +1/2
(θ, ϕ)

For the case where l+ = j + 21 this simplifies to

 q 
− j+1−j z
Y 1 (θ, ϕ)
2(j+1) l+ ,jz − 2
Yl+ ,j,jz (θ, ϕ) =  q j+1+j . (12.151)
z
Y
2(j+1) l+ ,jz + 1 (θ, ϕ)
2

For the case where l− = j − 12 we find

q 
j+jz
2j Y 1 (θ, φ)
l− ,jz − 2
Yl− ,j,jz (θ, ϕ) = q . (12.152)
j−jz
2j Yl− ,jz + 1 (θ, ϕ)
2

399
Chapter 12. Relativistic effects

We can combine these equations using the shorthand notation

Y± = Yl± ,j,jz (θ, ϕ), (12.153)

with l± = j ± 1/2 and s = 1/2. With the use of the notation jz± = jz ± 21
we find
 r 
l± + 12 ∓jz
∓ 2(l± + 12 ) Yl± ,jz− (θ, ϕ)
Y± =  r . (12.154)
 l± + 12 ±jz 
Y + (θ, ϕ)
2(l + 1 ) l± ,jz
± 2

As the last step we return to the original Dirac equation. We have found
solutions for the angular dependence of the small and large part of the wave
function. Following the argumentation in the non-relativistic limit, the full
solution is a product of angular and radial solutions. We realize that the
radial dependence for spin up and down is the same. The radial dependence
of the large and small part can be very different. We will call the radial
dependence of the large part g(r) and of the small part if (r). The reason
for the phases of the radial functions will become clear once we find explicit
solutions. With these definitions the Eigenstates of the Dirac Hamiltonian
for a problem with spherical symmetry become
 
Yl− ,j,jz (θ, φ)g(r)
ψκ<0,jz (~r) = (12.155)
Yl+ ,j,jz (θ, φ)if (r),
 
Yl+ ,j,jz (θ, φ)g(r)
ψκ>0,jz (~r) = .
Yl− ,j,jz (θ, φ)if (r)

In shorthand notation this becomes

 
Y± (θ, φ)g(r)
ψ± (~r) = . (12.156)
Y∓ (θ, φ)if (r)

400
 12.8. Solutions for the hydrogen atom

12.8.3 Operator αr
In order to solve the radial equations for the large and small part we return
to the equation 12.122 for the Dirac Hamiltonian in spherical coordinates.
We know that the eigenstates of this Hamiltonian are eigenstates of K with
eigenvalues −~κ and we replace the operator by its eigenvalues.
 
i
H D = cαr pr − ~ α4 κ + α4 mc2 + V (r). (12.157)
r

Our wave function factorizes into a radial and angular part. However the
Dirac Hamiltonian, although we replaced the operator K by its expectation
value still contains a part that depends on the angular coordinates. The
operator αr is defined as
1
αr = α~ · ~r (12.158)
r
1
 
0 r~σ· ~r
= 1 ,
σ · ~r
r~ 0

with
e−iϕ sin(θ)
 
1 cos(θ)
~σ · ~r = . (12.159)
r eiϕ sin(θ) − cos(θ)

We know that ψκ,jz (~r) is an eigenstate of H D , it therefor must be an

eigenstate of αr . In order to find the eigenvalues of αr we can look at the
square of this operator. If we square the operator αr we find

2 !
1
2 r ~
σ · ~
r 0
αr = 1

2 (12.160)
0 r~σ · ~r
 
12 0
= .
0 12

The eigenvalues of αr2 are clearly 1. The eigenvalues of αr are thus either
+1 or −1. The value one finds depends on the phase choice of the spherical

401
Chapter 12. Relativistic effects

harmonics. For the choice we make we find the eigenvalue c = −1. The
proof of this involves some recouping of angular momenta and can be found
below.
Using the notation Y± = Yl=j± 1 ,j,jz (θ, φ) we find
2

1
     
Y± 0 r~σ · ~r Y±
αr = 1 · (12.161)
Y∓ σ · ~r
r~ 0 Y∓
 
Y±
=c ,
Y∓
with c = +1 or c = −1. It follows that
1
cY± = ~σ · ~rY∓ (12.162)
r
e−iϕ sin(θ)
 
cos(θ)
= iϕ Y∓
e sin(θ) − cos(θ)
r  √ 
4π Y 1,0 (θ, ϕ) 2Y 1,−1 (θ, ϕ)
= √ Y∓
3 − 2Y1,1 (θ, ϕ) −Y1,0 (θ, ϕ)
We now insert the explicit angular dependence in terms of the spherical
harmonics for the spin up and down part of the wave function, whereby
we suppressed the notation of the arguments of the spherical harmonic
functions. We get
 r 
√ l∓ + 12 ±jz
± 2(l∓ + 12 ) Yl∓ ,jz − 12 
r  
4π Y1,0 2Y1,−1
cY± = √ · r  (12.163)
3 − 2Y1,1 −Y1,0  l∓ + 1 ∓jz 
2
Y
2(l∓ + 12 ) l∓ ,jz + 2
1

 r r 
l∓ + 12 ±jz l∓ + 12 ∓jz
r
4π  ± 1 Y1,0 Yl∓ ,jz − 1 + Y Y
(l + 1 ) 1,−1 l∓ ,jz + 2 
1

=  r 2(l∓ + 2 ) 2
r ∓ 2 
3  l∓ + 12 ±jz l∓ + 12 ∓jz 
∓ (l + 1 ) Y1,1 Yl∓ ,jz − 1 − 2(l + ) 1 Y1,0 Yl∓ ,jz + 1 .
∓ 2 2 ∓ 2 2

The product of two spherical harmonics with quantum numbers l1 , m1

and l2 , m2 can be combined into a sum over single spherical harmonic func-
tions with quantum numbers l, m, whereby due to conservation of angular

402
 12.8. Solutions for the hydrogen atom

momentum we know that |l1 − l2 | ≤ l ≤ l1 + l2 and m = m1 + m2 . The ex-

pansion coefficients can be found in equation 7.180. The relevant quantum
numbers we need yield
l∓ +1 r r
X
lm l0 2l∓ + 1 3
Y1,m1 Yl∓ ,m2 = C1m1 l∓ m2 C10l∓ 0 Yl,m . (12.164)
2l + 1 4π
l=|l∓ −1|

We now use that

r s
2l∓ + 1 l∓ −1 0 l∓
C10l∓ 0 = − (12.165)
2l + 1 (2l∓ − 1)
r
2l∓ + 1 l∓ 0
C =0
2l + 1 10l∓ 0
r s
2l∓ + 1 l∓ +1 0 l∓ + 1
C10l∓ 0 = ,
2l + 1 (2l∓ + 3)

and find
s r
l∓ −1 m l∓ 3
Y1,m1 Yl∓ ,m2 = − C1m 1 l∓ m2
Yl −1,m (12.166)
2l∓ − 1 4π ∓
s r
l∓ +1 m l∓ + 1 3
+ C1m 1 l∓ m2
Yl +1,m .
2l∓ + 3 4π ∓

Inserting this result into equation 12.163 yields

 r r q 
l∓ + 12 ±jz l∓ −1 jz− l∓ + 12 ∓jz l∓ −1 jz− l∓
 ∓ 2(l∓ + 12 ) C10l∓ jz− + (l + 1 )
C1−1l j + 2l∓ −1 Yl∓ −1,jz− 
cY± =   r r ∓ 2 ∓ z
q 
 l∓ + 12 ±jz l∓ −1 jz+ l∓ + 12 ∓jz l∓ −1 jz+ l∓

± 2(l + 1 ) C11l j − − 1
(l∓ + 2 )
C +
10l∓ jz
Y
2l∓ −1 l∓ −1,jz +
∓ 2 ∓ z
 r r q 
l∓ + 12 ±jz l∓ +1 jz− l∓ + 12 ∓jz l∓ +1 jz− l∓ +1
 ± 2(l∓ + 12 ) C10l∓ jz− + (l + 1 )
C1−1l j + 2l∓ +3 Yl∓ +1,jz− 
+  r r ∓ 2 ∓ z
q 
 l∓ + 12 ±jz l∓ +1 jz+ l∓ + 12 ∓jz l∓ +1 jz+ l∓ +1

∓ 2(l + 1 ) C11l j − − (l + ) 1 C 10l j + 2l∓ +3 Y l ∓ +1,j +
z
∓ 2 ∓ z ∓ 2 ∓ z

(12.167)

403
Chapter 12. Relativistic effects

with jz± = jz ± 12 . If one inserts explicit expressions of the Clebsch-Gordan

coefficients one finds that the coefficients in front of Yl− −1,m and Yl+ +1,m
are zero. we furthermore realize that

Yl∓ ±1,m = Yl=j± 1 ∓1,m (12.168)

2

= Yl± ,m .

 r 
l± + 12 ∓jz
∓ 2(l± + 12 ) Yl± ,jz− (θ, ϕ)
cY± = −  r . (12.169)
 l± + 12 ±jz 
Y + (θ, ϕ)
2(l + 1 ) l± ,jz
± 2

Comparing this to equation 12.154 we find that c = −1, i.e. the eigenvalue
of αr for all the eigenstates of K is −1.

12.8.4 Radial solutions

12.9 Lamb shift

To be done

404
Chapter 13

Response theory

If you want to obtain information about a system you make a measurement.

The way we do measurements is to take the system and see how it responds
to an external stimulus. If you want to measure the electric conductivity
you apply an electric field and see how much current will flow. If you, for
example, want to measure the strength of a beam of material you apply a
force to the beam and see when it brakes. The theory that deals with the
question on how an object reacts due to an external stimulus is known as
response theory. We are often interested in the regime where the external
stimulus is small. A small perturbation allows one to induce reversible
changes to the system. If you apply an electrical field of the order of volts
per meter to a system it will induce a current. If you turn off the field the
current will stop and the system relaxes back to the ground-state. One can
repeat this a near infinite times. As an example, if you apply a field of
the order of volts per angstrom to a molecule or crystal it will, irreversible,
change the chemical bonds and your crystal might fall apart. The use of
small fields allows one to use perturbation theory. If we can use first order
perturbation theory we talk about linear response theory. In second order
perturbation theory we talk about quadratic response theory etc. As we
will see below, the border when response theory is valid and when not is
not always trivial. The process has definitely to be such that you come

405
Chapter 13. Response theory

back to the original ground state when the external perturbation is turned
off. At the same time, extremely large fields applied for very short times
can still be described by response theory.
We will start this chapter with a short general introduction of response
functions and their Volterra expansion. We will then focus on the calcu-
lation of the response functions from the microscopic description of the
system. We will first calculate the response of an electron bound to an
atom due to an oscillating electrical field (i.e. light), before we generalize
the response theory to any system and arbitrary operator representing an
observable. The specific example gives a very clear and intuitive feeling of
what response theory is about. We will do the example once on a classical
system, representing the bound electron by a charged particle on a spring
and once for the hydrogen atom described by the Schrödinger equation. We
can already guess what the possible responses are for such a system. We
assume a positive charge at position z = 0 and the electron at position z.
At rest the average position of the electron will be at z = 0. If we move
the electron away from z = 0 we have generated a dipole moment p = −ez.
In order for the electron to get to the position z, starting at the position
z = 0, it needs to move. The movement of the electron is related to a cur-
rent. In an oscillating electric field the induced current will also oscillate.
In order for the velocity to change we need to accelerate the electron. An
accelerating charge produces an electro magnet field, such that we get scat-
tering of the light. The response functions of relevance are related to the
position, velocity and acceleration of the electron, i.e. the induced dipole
moment, the induced current and the resulting scattering of light, due to
the external oscillating electrical field. The reason we will do this both for
a classical and quantum model is that the response of the system in both
descriptions is very similar, making a direct comparison useful.
Within response theory we want to calculate the response of a system
x(t) in time as a function of some external action on the system F (t). We
can think of these as the position x of a particle influenced by an external
force F . In general the position we find at time t depends on the position
at t → −∞ and the force applied at all times t0 < t. We can express x(t)

406
in terms of a Volterra series expansion
x(t) = χ(0) (t) (13.1)
Z t
+ χ(1) (t; t1 )F (t1 )dt1
−∞
Z t Z t
+ χ(2) (t; t1 , t2 )F (t1 )F (t2 )dt1 dt2
−∞ −∞
+ ...
∞ Z
X t Z t n
Y
(0) (n)
=χ (t) + ... χ (t; t1 , t2 , . . . , tn ) F (tj )dtj ,
n=1 −∞ −∞ j=1

whereby χ(n) is called the n-th order response function, or n-th order sus-
ceptibility. For most of our systems it does not matter if we do an experi-
ment today or tomorrow. The susceptibility therefore should only depend
on the difference between tj and t
χ(n) (t; t1 , t2 , . . . , tn ) = χ(n) (t − t1 , t − t2 , . . . , t − tn ). (13.2)
With τn = t − tn we find
∞ Z ∞
X Z ∞ n
Y
(0)
x(t) = χ (t) + ... χ(n) (τ1 , τ2 , . . . , τn ) F (t − τj )dτj . (13.3)
n=1 0 0 j=1

For a system that is time translational invariant we should find that en-
ergy is a conserved quantity. For most force fields we encounter in physics
the energy is related to a frequency and it thus makes sense to Fourier
transform equation 13.3 to the frequency domain. Even without the phys-
ical background of the Noether theorem that states that for each symme-
try operation there should be a conserved quantity one should see that
Fourier transforming equation 13.3 will simplify the math. Equation 13.3
describes a convolution of the susceptibility and the time dependent force.
The Fourier transform of a convolution yields a simple product. We define
the Fourier transforms:
Z ∞
x(ω) = x(t)eiωt dt (13.4)
−∞

407
Chapter 13. Response theory

Z ∞
1
x(t) = x(ω)e−iωt dω
2π −∞
Z ∞
F (ω) = F (t)eiωt dt
−∞
Z ∞
1
F (t) = F (ω)e−iωt dω,
2π −∞

and find
Z ∞
x(ω) = χ(0) (t)eiωt dt (13.5)
−∞
Z ∞X∞ Z ∞ Z ∞ n
Y
(n)
+ ... χ (τ1 , τ2 , . . . , τn ) F (t − τj )dτj eiωt dt
−∞ n=1 0 0 j=1
(0)
= χ (ω)
Z ∞Z ∞
+ θ(τ1 )χ(1) (τ1 )F (t − τ1 )eiω(t−τ1 ) eiωτ1 dτ1 d(t − τ1 )
−∞ −∞
+ ...,

whereby we introduced the the Heaviside step function θ(τj ) such that we
can take the integration boundaries from −∞ to ∞ for the integral over τ1 ,
used that eiω(t−τ1 ) eiωτ1 = eiωt , and shifted the integration boundaries for t
by τ1 . With
Z ∞
(1)
χ (ω) = θ(τ1 )χ(1) (τ1 )eiωτ1 dτ1 , (13.6)
−∞

we find

x(ω) = χ(0) (ω) + χ(1) (ω)F (ω) + . . . . (13.7)

This chapter will deal with the question on how to calculate χ(n) for
various actuators and responses of different materials. We will start with
two model examples and then continue with a general treatment.

408
 13.1. Classical example

13.1 Classical example

Figure 13.1: Classical model used to calculate the linear response.

We define our model by a particle (electron) with mass m and charge

−e at position x. The particle is bound to the position x = 0 by the force
FS of a spring with spring constant k, as shown in figure 13.1

FS = −kx. (13.8)

The particle is subject to an external electric field Eext (ω) = E0 cos(ωt)

resulting in a force on the particle

Fext = F0 cos(ωt), (13.9)

with F0 = −eE0 . Once the particle moves it is subject to a friction force

dx
Ffriction = −mγ . (13.10)
dt
The friction is proportional to the velocity of the particle. The reason for
the proportionality constant of mγ becomes clear once we find the solution
of the equation.
The resulting equation of motion is
d2 x dx
m 2
= −kx − mγ + F0 cos(ωt). (13.11)
dt dt

409
Chapter 13. Response theory

13.1.1 Real variables without friction

We can solve equation 13.33 by assuming that x is proportional to cos(ω 0 t)
or sin(ω 0 t). The solution we find is

F0
x= cos(ωt) + A0 cos(ω0 t − φ0 ), (13.12)
mω 2 − mω02

with
r
k
ω0 = , (13.13)
m
the eigenfrequency of the harmonic oscillator. The term oscillating at fre-
quency ω0 are oscillations possible without a driving field. This term is
known as the transient solution and determined by the initial conditions.
The motion with frequency ω0 is not related to the response of the system
due to the external driving field. Without loss of generality for the deriva-
tion of the response function we can choose the initial settings such that
A0 = 0. With A0 = 0 the only movement of the particle is related to the
response of the system to the external driving force. The term that oscil-
lates with the frequency of the driving force is known as the steady-state
solution.

Amplitude and phase

If we set the amplitude of the transient solution to zero we find that the
particle oscillates at the same frequency as the driving force. We can write
the motion of the particle as

x = A cos(ωt − φ), (13.14)

with A ≥ 0 the amplitude of the oscillation

F0
A= (13.15)
mω02 − mω 2

410
 13.1. Classical example

Amplitude of oscilation depending of driving frequency ω Phase of oscilation depending of driving frequency ω

1.0
6
]
ω20 m
F0

5 0.8

Phase ϕ [units of π]
Amplitude A [units of [

4 0.6
3
0.4
2
0.2
1

0 0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
ω [units of ω0 ] ω [units of ω0 ]

Figure 13.2: The amplitude and phase of the response of the particle to an
oscillating driving force.

F0 1
= ,
m (ω0 − ω)(ω0 + ω)

and φ the phase of the oscillation with respect to the driving force

φ = π θ(ω − ω0 ), (13.16)

and θ(ω − ω0 ) the Heaviside step function that is zero for ω < ω0 and 1 for
ω > ω0 .
The amplitude and phase of the oscillation are shown in figure 13.2.
The amplitude is largest if the frequency of the driving force is equal to
the eigenfrequency of the harmonic oscillator (it diverges at this frequency)
1
and falls of as (ω0 −ω)(ω0 +ω)
away from the resonance. When |ω0 − ω|  1
we can take (ω0 + ω) to be roughly constant and the amplitude falls of like
1 1
ω . For the case where ω  ω0 the amplitude falls of as ω 2 .

411
Chapter 13. Response theory

Total energy in the oscillator

Besides the position, velocity and acceleration of the particle as a response

to the external driving force it is useful to discuss the energy of the particle,
such that we can calculate the absorption of energy by the system. The
kinetic energy within the harmonic oscillator is
 2
1 dx
Ekin = m (13.17)
2 dt
F 2 ω2
= 0 sin2 (ωt).
2m (ω 2 − ω02 )2

The potential energy within the harmonic oscillator is given as

1
Epot = kx2 (13.18)
2
F2 ω02
= 0 cos2 (ωt).
2m (ω 2 − ω02 )2

The total energy within the harmonic oscillator is given as

Etot = Ekin + Epot (13.19)

F02ω02 cos2 (ωt) + ω 2 sin2 (ωt)
=
2m (ω 2 − ω02 )2
F02 (ω 2 + ω02 )
 
1
= 2 2 2
+ 2 cos(2ωt) .
4m (ω − ω0 ) (ω0 − ω 2 )

The energy in the Harmonic oscillator can be written as a constant term

and a term that oscillates in time.

Absorption of energy

We can calculate the absorption of energy by the harmonic oscillator by

looking at the time average power exerted by the field. The power, or the

412
 13.1. Classical example

energy transferred, is given by the movement of the particle against the

external force
dx
P =F (13.20)
dt
F02 ω
= sin(2ωt).
2m ω − ω02
2

Equivalently the power can be calculated by looking at the change of the

total potential plus kinetic energy of the particle
d
P = Etot (13.21)
dt
F2 ω
= 0 2 sin(2ωt).
2m ω − ω02
The absorption of energy by the Harmonic oscillator is given as the time
average of the power
R 2π/ω
P dt
hP i = R02π/ω (13.22)
0 1dt
= 0.

Half of the time the field adds energy to the oscillator, the other half of the
time the oscillator adds energy to the driving field. On average there is no
net absorption of energy by the particle in the driven harmonic oscillator.

13.1.2 Real variables with damping

The solution to the equation of motion 13.33 for the case where γ 6= 0 is
off the same form as for the case where γ = 0:
(
0 γ
− γ2 t cos(ω0 t + φ0 ) ω0 > 2
x = A0 e 0 (13.23)
e−ω0 t ω0 ≤ γ2
F0 ω02 − ω 2 cos(ωt) + γω sin(ωt)


+
2 ,
m γ 2ω2 + ω2 − ω2
0

413
Chapter 13. Response theory

q q
2
with ω0 = m k
and ω00 = |ω02 − γ2 |.
We find that the transient solution is exponentially damped by γ. We
furthermore find that when ω0 < γ/2 the oscillator is over-damped and
the transient solution behaves in total as an exponentially decay instead
of an oscillating function. The frequency with which the transient solution
oscillates is shifted from ω0 to ω00 . The movement of the particle due to
the driving field is at the same frequency of the driving force (ω), however
with a non-trivial phase shift.

Amplitude and phase

Figure 13.3: The amplitude and phase of the response of the particle to an
oscillating driving force for the damped harmonic oscillator.

In order to determine the amplitude and phase of the induced oscil-

lation with respect to the driving force we can use that A cos(ωt − φ) =
A cos(φ) cos(ωt) + A sin(φ) sin(ωt) such that the amplitude and phase are
given by the equations:

ω02 − ω 2


F0
A cos(φ) = (13.24)
m γ 2ω2 + ω2 − ω2 2


0

414
 13.1. Classical example

F0 γω
A sin(φ) =
. (13.25)
m γ 2ω2 + ω2 − ω2 2
0

We find for the amplitude

F0 1
A= q
2 , (13.26)
m
γ 2 ω 2 + ω02 − ω 2

and for the phase

 
−1 γω
φ = tan (13.27)
ω02 − ω 2

Total energy in the oscillator

Just as in the case without damping we can calculate the total energy within
the harmonic oscillator. The kinetic energy is given as
 2
1 dx
Ekin = m (13.28)
2 dt
F2 ω2 2 2

2
= 0
2 γω cos(ωt) + (ω − ω0 ) sin(ωt) ,
2m (ω 2 − ω )2 + γ 2 ω 2
2
0

and the potential energy as

1
Epot = kx2 (13.29)
2
F2 ω02
2
= 0 2 (ω 2 − ω02 ) cos(ωt) + γω sin(ωt) ,
2m (ω 2 − ω 2 )2 + γ 2 ω 2


0

such that the total energy is

Etot = Ekin + Epot (13.30)

F02 ω 2 + ω02
=
4m (ω 2 − ω02 )2 + γ 2 ω 2

415
Chapter 13. Response theory

(ω 2 − ω02 )((ω 2 − ω02 )2 − γ 2 ω 2 )

−
2 cos(2ωt)
(ω 2 − ω02 )2 + γ 2 ω 2
!
2γω(ω 2 − ω02 )2
+
2 sin(2ωt) .
(ω 2 − ω02 )2 + γ 2 ω 2
The change in total energy over time has two reasons. There is an
exchange of energy between the oscillator and the driving field and there
is a los of energy due to the friction term. The total energy averaged over
a complete oscillation is constant such that the average power supplied by
the field to the oscillator is equivalent to the averaged energy loss of the
oscillator due to the friction term.

Absorption of energy
We can calculate the absorption of energy by the oscillator from the time
averaged power exerted by the field. The power is given as the movement
against a force exerted by the field
dx
P =F (13.31)
dt
F02 1
= 2
2m (ω − ω0 )2 + γ 2 ω 2
2

× γω 2 + γω 2 cos(2ωt) + ω(ω 2 − ω02 ) sin(2ωt) .

The absorption of energy by the Harmonic oscillator is given as the time

average of the power
R 2π/ω
P dt
hP i = R02π/ω (13.32)
0 1dt
F02 γω 2
= .
2m (ω 2 − ω02 )2 + γ 2 ω 2
Although the energy in the system is constant once averaged over a com-
plete oscillation cycle, the field does provide energy, which is dissipated by
the friction term.

416
 13.1. Classical example

13.1.3 Complex variables with damping

In the previous sub section we skipped the algebra of several derivations and
just gave the solutions to the posed equations. One can check that these
equations are indeed solutions to the given differential equations. The math
becomes much more elegant with the use of complex variables. We will add
a complex part to the force and use that F0 cos(ωt) − F0 i sin(ωt) = F0 e−iωt .
The differential equation to solve becomes

d2 z dz
m 2
= −kz − mγ + F0 e−iωt , (13.33)
dt dt
and x = Re(z). The gain in using complex variables is that the phase
shift is now much easier derived and allows us to write the response as a
multiplication of a response function times the driving force. Setting the
amplitude of the transient solution to zero we find
1
z=
F0 e−iωt . (13.34)
m ω02 − ω 2 − iγω

With x = Re(z)qwe find the same solutions as in the previous section.

2
With ω00 = ω02 − γ2 the oscillation frequency of the transient term
we can rewrite the solution as
 
1 1 1
z= − + F0 e−iωt . (13.35)
2mω00 ω − ω00 + i γ2 ω + ω00 + i γ2

13.1.4 Linear response functions

The use of complex variables allows one to define elegant relations between
the force acting on the system and the response of the system. One can
define 3 useful response functions. The susceptibility describes how much
the position of the particle is changed due to the external force. For an
electric field and a charged particle this is proportional to the induced
dipole moment as a function of field strength. The conductivity describes
how much the particle moves due to the external force. For a charged

417
Chapter 13. Response theory

particle in an external field this is the induced current. The scattering

length describes how much the particles are accelerated due to the external
force. For a charged particle in an external electric field the scattering
length tells you how much the system re-emits photons to the field.
The electric susceptibility, conductivity and scattering length are de-
fined for a continues medium, relating the induced polarization density,
current density and scattering of a continues medium to the externally ap-
plied electrical field. Below we will use these response functions for a single
harmonic oscillator. In order to calculate a response density one can assume
a homogeneous distribution of these harmonic oscillators in space. For a
medium one should furthermore consider that the local field can be different
from the applied field. The local field seen by a single oscillator is the ap-
plied field plus the field induced by all other oscillators in the medium. For
a single oscillator one does not need to include these subtleties between the
applied and local field as they are equivalent. One should notice that there
is no common convention used for the phase and units used to express these
response functions between different fields of research. As the susceptibility
can be expressed as a dimensionless unit this can lead to great confusion.
This not only relates to factors of 4π and 10 (Tesla per Kilo Gauss) one can
encounter, but one can add a factor of i, −i, or −1 to the response function
without loss of generality. In principle both the force and the resulting
action are real such that also the response functions must be real. In order
to include the phase shift between an oscillating force and the response we
used complex variables, realizing we in the end need to look at the real or
complex part of these functions to get an object with physical meaning. If
we change the phase of the response function and simultaneously change
the interpretation of the real and complex part of our variables there is no
change in physical observables. There is no physical reason to choose one
phase definition over an other and it’s left to convention.
The reason for the flurry of possible units and phase conventions is
historical. The response functions relate to different observables: The ab-
sorption and refraction of light relate to the index of refractivity, which
directly follows from the susceptibility, the former are defined by Newton
in his book on optics. The propagation of electromagnetic fields and the

418
 13.1. Classical example

defined dielectric constants in Maxwell’s theory of electromagnetism in a

medium are defined by the susceptibility. Diffraction as studied by Laue
and Bragg is given by the scattering cross section. Each of these different
fields developed their own convention for the response functions, without
realizing the common framework. A unifying framework is presented by
the paper of Kubo on linear response theory. The fact that one can find
different definitions of the response function with different phases is not
such a big problem. Observables must be real, such that one can always
backwards engineer which phase an author used. The fact that one can
add seemingly arbitrary dimensionless constant of 4π or 10 (Tesla per Kilo
Gauss) is confusing, but related to roughly 200 years of conventions that
we are not likely to change soon.
In the next section we very deliberately only use proportionality con-
stants for the definition of our response functions and leave out the specific
constants. This will later allow us to generalize the response functions to
the observation of any response of a system to any type of action.

Susceptibility
We define the electric susceptibility for a single harmonic oscillator χs to
relate the induced electric dipole moment p to the applied external electric
field:

p ∝ χs (ω)E. (13.36)

The index s is there to remember that we calculate the susceptibility not for
a medium, but for a single oscillator. Instead of a relation between induced
dipole moment as a function of the electric field we can equally well define
the susceptibility as a relation between the particle position and the applied
force. The induced dipole moment is proportional to the position

p = −ez, (13.37)

and the electric field is proportional to the force

F = −eE. (13.38)

419
Chapter 13. Response theory

We thus find that the susceptibility for a single harmonic oscillator also
relates the position of the particle to the applied force:

z ∝ χs (ω)F. (13.39)

From equation 13.34 we see that

1
χs (ω) =∝ (13.40)
ω02
− ω 2 − iγω
 
1 1 1
= − + .
2ω00 ω − ω00 + i γ2 ω + ω00 + i γ2

The electric susceptibility is given by an analytical function with two

poles at energy ±ω00 − i γ2 . The poles are in the lower half of the complex
plane. The real part of the energy of the pole ±ω00 is given by the resonant
energy of the harmonic oscillator.
For large frequencies ω → ∞ the particle in the harmonic oscillator
should start to oscillate very fast. As the particle has a finite mass it
should take time for it to accelerate and move. The displacement of the
particle thus should go to zero for ω → ∞ or in other words χs (ω) → 0 for
ω → ∞. We indeed find that this is true, the real part of χs (ω) falls off as
1/ω 2 , the imaginary part as 1/ω 3 .

Conductivity
The electric conductivity (σs ) for our single harmonic oscillator relates the
induced current to the external electrical field

j = σs (ω)E (13.41)
d
with j = −ev the current of the electron, v = dt x the velocity of the
electron and E the applied electric field. We can relate the current to the
time derivative of the induced dipole moment such that
d
j= p. (13.42)
dt

420
 13.1. Classical example

Substituting σs (ω)E for j on the left hand side we find

d
σs (ω)E = p (13.43)
dt
d
∝ χs (ω) E
dt
∝ −iωχs (ω)E.

We can now divide both sides by E to find

σs (ω) ∝ −iωχs (ω) (13.44)

ω
∝ −i 2
ω0 − ω 2 − iγω
 
ω 1 1
=i 0 − .
2ω0 ω − ω00 + i γ2 ω + ω00 + i γ2

The conductivity is directly related to the absorption coefficient (α).

The conductivity relates the induced current to the applied electromag-
netic field. With j = −ev and F = −eE we can equally state that the
conductivity relates the induced velocity of the particles to the applied
force. The energy dissipation per unit time is given by the movement of
the particle agains the applied force. We can now use that

F v = Ej (13.45)
= Eσs (ω)E,

to find the total absorbed energy per unit time. The absorption coefficient is
defined as the absorbed energy per unit time divided by the energy density
in the electric field such that we find
!
~ ∗ · σs · E
E ~
α = Re . (13.46)
~ 2
|E|

In the previous equation we allowed the electric field to be a vector and

σs a tensor. We furthermore included that we used complex variables for

421
Chapter 13. Response theory

the definition of our response functions, such that we need to take the
real part to obtain the observables. In some fields it is common to define
σs (ω) ∝ ωχs (ω), which changes the phase of the conductivity. In that case
we find that the absorption coefficient is given by the equation
!
~ ∗ · σs · E
E ~
α = −Im . (13.47)
~ 2
|E|

For the harmonic oscillator we find that

ω
σs ∝ −i (13.48)
ω02 − ω 2 − iγω
!
γω 2 ω(ω 2 − ω02 )
=
2 + i
2 ,
γ 2 ω 2 + ω02 − ω 2 γ 2 ω 2 + ω02 − ω 2

such that Re(σ) = α in agreement with equation 13.32.

Scattering length
Photons can scatter of an electron. We define the scattering length to relate
the outgoing or scattered field to the applied field:

Eout ∝ fs (ω)E(t, ω). (13.49)

On a microscopic level the outgoing field is produced by the acceleration of

the electric charges. In our model we find

d2
Eout ∝ p(t). (13.50)
dt2
Using that p ∝ χs (ω)E(t, ω) we find that

d2
Eout ∝ χs (ω) E(t, ω) (13.51)
dt2
= −ω 2 χs (ω)E(t, ω).

422
 13.2. Quantum example

From which we find that

f (ω) ∝ −ω 2 χs (ω) (13.52)

ω2
∝
ω 2 − ω02 + iγω

ω2
 
1 1
= − .
2ω00 ω − ω00 + i γ2 ω + ω00 + i γ2

If the force on a charged particle is given by an oscillating electrical field

at frequency ω then χ relates the position of the particle to the applied field,
σ relates the velocity of the particle to the applied field and f relates the
acceleration of the particle to the applied field. For large frequencies a
massive particle should not have time to respond. We indeed see that both
limω→∞ χ(ω) = 0 and limω→∞ σ(ω) = 0. For the scattering length we find
that

lim f (ω) ∝ 1 (13.53)

ω→∞

Even at infinite frequencies an electromagnetic field scatters of an electron.

The scattering length at those frequencies does not depend on the internal
structure off the scattering centre, but becomes a constant as a function of
frequency.

13.2 Quantum example

In the previous section we described the linear response of a classical har-
monic oscillator to an external periodic driving force. Here we will look
what happens if we describe the particle on a quantum level. We will take
the H atom in an external oscillating electromagnetic field as an example.
We will restrict ourselves to a basis of the 1s and 2pz orbital and take the
field to be in the z direction. The Hamiltonian without an electromagnetic
field is given as

H 0 = ε1s a†1s a1s + ε2p a†2pz a2pz , (13.54)

423
Chapter 13. Response theory

or in matrix form
 
ε1s 0
H0 = , (13.55)
0 ε2p

and
!
  a1s
H0 = a†1s a†2p · H0 · , (13.56)
a2p

with ε1s = −1 Rydberg and ε2p = −1/4 Rydberg, as the eigenenergy for
the hydrogen atomic wave functions in the Schrödinger equation is equal
to −1/n2 Rydberg .
~ = E0 ẑ sin(ωt). Such that the Hamiltonian
The electric field is given as E
for the electromagnetic field is given as

H 1 = −eE0 sin(ωt)z, (13.57)

with z the position operator. In matrix form we find for the position
operator
 
hφ1s |z|φ1s i hφ1s |z|φ2pz i
z= , (13.58)
hφ2pz |z|φ1s i hφ2pz |z|φ2pz i

and
!
  a1s
z = a†1s a†2p · z · , (13.59)
a2p

We can evaluate the matrix elements with the use of the explicit form
of the hydrogen wave functions. With

φnlm (~r) = Ylm (θ, φ)Rnl (r), (13.60)

r
1 1
Y1s (θ, φ) = ,
2 π
r
1 3
Y2pz (θ, φ) = cos(θ),
2 π

424
 13.2. Quantum example

 3/2
−r/a0 1
R1s (r) = 2e ,
a0
 5/2
1 1
R2p (r) = √ re−r/(2a0 ) ,
2 6 a0
we find for the matrix representing the position operator
 
0 1
z = z0 , (13.61)
1 0

with
128 √
z0 = 2a0 (13.62)
243
≈ 0.744a0 . (13.63)

Instead of z0 one often encounters the dipole moment between the 1s and
2p hydrogen wave function defined as µ = −ez0 .
Both the s and pz orbitals are centred around z = 0, which is reflected
by the zero values on the diagonal of the position operator. The eigenstates
of the position operator are
r
1
φ±z (~r) = (φ1s (~r) ± φ2pz (~r)) , (13.64)
2
with eigenvalues, i.e. the centre of the charge of the resulting orbitals, ±z0 .
On a basis of the 1s and 2pz orbital we can still describe the electric
field as a driving force that drives the electrons away from z = 0. As we
truncated the one particle basis we can no longer bring the electron to
±∞, but find a maximum distance of z0 that the electron can be moved
away from the origin. If this is a sufficient approximation depends on the
experiment under consideration.
In order to describe the full response of the system we should not look
at the eigenstates of the position operator only. For the state

φα (~r) = cos(α)φ1s (~r) + sin(α)φ2pz (~r), (13.65)

425
Chapter 13. Response theory

0.5

0.4

0.3
ϕ 1 s , ϕ 2 pz

0.2

0.1

0.0

-0.1
-10 -5 0 5 10
z [a0 ]

φ(r) as a function of α

φ(r) as a function of β

Figure 13.4: The possible wave-functions on a basis of the 1s and 2pz hy-
drogen orbital. Top row from left to right, the 1s and 2pz wave-function
along the line x = y = 0, isosurface of the 1s and 2pz orbitals and isosur-
face of the 1s ± 2pz orbital. Middle rows isosurface of the wave function
cos(α)1s + sin(α)2pz shown in two different perspectives and the surface
defined by the equation r = | cos(α)Ys (θ, φ) + sin(α)Ypz (θ, φ)|. Bottom row,
surface defined by r = | cos(α)Ys (θ, φ) + eiβ sin(α)Ypz (θ, φ)| as a function of
β for α = π/2.

426
 13.2. Quantum example

we find that the expectation value for the position operator z is

hφα |z|φα i = z0 2 sin(α) cos(α) (13.66)

= z0 sin(2α),

such that any value of z between 0 and z0 can be realized. Naturally a

quantum state for a single particle contains more information than the
expectation value of the position (or momentum) of that particle. In figure
13.4 we show the wave function for the state φα (~r). On the top row we start
from left to right with the radial wave function of the 1s and 2p orbital and
then continue with a two dimensional contour plot of the wave function
in the xz plane as well as a 3 dimensional isosurface plot for the 1s and
2pz orbital. The right panel of the top row shows a 2 dimensional contour
plot and 3 dimensional isosurface plot of the eigenstates of the z operator.
The middle 3 rows of figure 13.4 show plots of φα (~r) as a function of α.
The top two rows show isosurface plots of φα (~r). As the radial function
of the 1s and 2pz are different these plots slightly depend on the value of
the isosurface used. We can simplify the plots by looking at the angular
dependent part of the function only. If we plot the surface defined by the
contour r = | cos(α)Ys (θ, φ) + sin(α)Ypz (θ, φ)| we obtain the bottom of the
3 plots that show the wave function as a function of α. The later plot
neglects the difference in radial wave functions. As one can see from all of
these three plot sequences, the centre of the φα (~r) orbital oscillates around
z = 0 as a function of α with a period of π. Both the 1s and 2pz orbitals
are centred around z = 0, the linear combinations of these orbitals move
away from the centre.
Naturally a wave function must not be real. The most general solution
for our state on the basis of the 1s and 2pz orbital is given by the function

φα,β (~r) = cos(α)φ1s (~r) + eiβ sin(α)φ2pz (~r). (13.67)

If we calculate the expectation value for the position operator for this gen-
eral state we find
 
hφα,β |z|φα,β i = z0 eiβ + e−iβ sin(α) cos(α) (13.68)

427
Chapter 13. Response theory

= z0 cos(β) sin(2α).

The phase of the pz orbital with respect to the s orbital allows one to
move the centre of the electron wave function in the z direction. In the
bottom row of figure 13.4 we show the wave function φα,β (~r) for α = π/2
as a function of β. The contour plotted is given by the equation r =
| cos(α)Ys (θ, φ) + eiβ sin(α)Ypz (θ, φ)| and the surface colour corresponds to
the phase of the wave function.
If we start at t = 0 in the state φ(t = 0, ~r) = φα,β (~r) without an external
electromagnetic field, then the state at finite time is given as

φ(t, ~r) = e−iε1s t/~ cos(α)φ1s (~r) + e−iε2p t/~ eiβ sin(α)φ2pz (~r). (13.69)

As εs 6= ε2p we find that the relative phase between the 1s and 2pz contri-
bution to the wave function changes as a function of time. Starting in a
state φα,β (~r) the electron thus oscillates around z = 0 with an amplitude
z0 sin(2α) and angular frequency ω0 = (ε2p − ε1s )/~.
In order for the electron to oscillate around z = 0 there must be a
restoring force that pulls the electron back to z = 0 if it is displaced.
We can calculate this force from the Hamiltonian H 0 . We can express
hφα,β |H 0 |φα,β i as a function of hφα,β |z|φα,β i. When doing so we should
realize that the amplitude of the oscillation is defined by α, the phase
within an oscillation by β. For a classical harmonic oscillator the kinetic
energy is zero when the particle is furthest away from the centre. In our
model this corresponds to β = 0. The expectation value of the energy and
position for β = 0 are given as

hφα |z|φα i = z0 2 sin(α) cos(α), (13.70)

and

hφα |H 0 |φα i = ~ω0 sin2 (α) (13.71)

2
= ~ω0 (1 − cos (α)),

428
 13.2. Quantum example

with ~ω0 = εpz − εs = 43 Rydberg. Combining these two equations we find

for the relation between energy and extremal position
s s
hφα |H 0 |φα i hφα |H 0 |φα i
hφα |z|φα i = z0 2 1 − . (13.72)
~ω0 ~ω0

If we solve for the energy we find

 s 
1 hφα |z|φα i 
hφα |H 0 |φα i = ~ω0 1− 1− . (13.73)
2 z0

In an Harmonic oscillator the energy depends quadratically on the position

of the particle. For the case where z is small we can make a Taylor series
expansion and find
 2
z4

z
hφα |H 0 |φα i = ~ω0 + + ... . (13.74)
4z02 16z04

The first term describes the potential of an Harmonic oscillator, the higher
order terms are anharmonic corrections to the potential. We should expect
these anharmonic corrections as the electron in an hydrogen atom is trapped
in a bare potential that is of the form −1/r and not equal to r2 .
Once we have established the similarity (not the equivalence) between
the quantum and classical model we can have a look at the time dependent
solution of the quantum problem. For this we need to solve the time depen-
dent Schrödinger equation. The time dependent Hamiltonian on a basis of
the 1s and 2pz orbital is
!
  a
H(t) = a†1s a†2p · H(t) · 1s , (13.75)
a2p

with the matrix representation H(t) of the Hamiltonian given as

 
0 V sin(ωt)
H(t) = , (13.76)
V sin(ωt) ~ω0

429
Chapter 13. Response theory

whereby we shifted H 0 by the energy of the 1s orbital and V = E0 µ, with

E0 the electric field strength and µ = −ez0 the dipole moment between the
1s and 2pz orbital.
We are interested in solutions of the Schrödinger equation
d
i~ φ(t) = H(t)φ(t). (13.77)
dt
For the case where V = 0, i.e. without a driving electromagnetic field the
solution would be

φ(t) = (cos(α), e−iω0 t sin(α)), (13.78)

with α depending on the starting conditions. For the case with an external
field we will take the trail soltution

φ(t) = (a(t), e−iω0 t b(t)), (13.79)

with |a(t)|2 + |b(t)|2 = 1. If we insert the trail solution into the Schrödinger
equation we find a set of two coupled equations
d
i~ a(t) = V sin(ωt)e−iω0 t b(t) (13.80)
 dt 
d
i~e−iω0 t b(t) − iω0 b(t) = V sin(ωt)a(t) + ~ω0 e−iω0 t b(t).
dt
These equations can either be solved numerically,
or approximately analyt-
ical. With the use that sin(ωt) = 21 i e−iωt − eiωt we have

d 1  
i~ a(t) = iV e−i(ω0 +ω)t − e−i(ω0 −ω)t b(t), (13.81)
dt 2
or
Z 
1 
a(t) = V e−i(ω0 +ω)t − e−i(ω0 −ω)t b(t)dt. (13.82)
2~

For ω ≈ ω0 the term with e−i(ω0 +ω)t oscillates much faster than the term
with e−i(ω0 −ω)t . As a result the integral over the first term will be much

430
 13.2. Quantum example

smaller and can be neglected. This approximation is known as the rotating

wave approximation. The resulting equations are
d 1
i~ a(t) = − iV eiωt e−iω0 t b(t) (13.83)
 dt  2
d 1
i~ b(t) − iω0 b(t) = iV e−iωt eiω0 t a(t) + ~ω0 b(t).
dt 2

The solutions for a(t = 0) = 1 and b(t = 0) = 0 are

 
− 21 i(−ω+ω0 )t i(ω − ω0 ) sin(Ωt)
a(t) = e cos(Ωt) − (13.84)
2Ω
1 sin(Ωt)
b(t) = eiω0 t e− 2 i(ω+ω0 )t V ,
2~Ω
with the Rabi frequency Ω given as
s 
1 V 2
Ω= + (ω − ω0 )2 . (13.85)
2 ~

We find that the phase between the 1s and 2pz contribution to ψ(t, ~r)
oscillates with frequency ω, i.e. the frequency of the driving field. This
will induce an oscillation of the electron and induced dipole moment with
the frequency of the driving force. We furthermore see that the coefficient
of the 2pz contribution is proportional to sin(Ωt). This will induce a time
dependence in the amplitude of the oscillations. We can quantify this by
looking at the time dependent expectation value of the position operator
z.

hφ(t)|z|φ(t)i = z0 a∗ (t)eiω0 t b(t) + a(t)e−iω0 t b∗ (t)

(13.86)
V (ω − ω0 ) sin(ωt)(1 − cos(2Ωt))
= z0
4~Ω2
!
V cos(ωt) sin(2Ωt)
+ .
2~Ω

431
Chapter 13. Response theory

The electron oscillates with frequency ω and the amplitude is modified with
the frequency 2Ω.
We now want to focus on the linear response of the system. For this we
assume that V is small and make a series expansion of hφ(t)|z|φ(t)i. We
find
sin(ωt) − sin(ω0 t)
hφ(t)|z|φ(t)i = z0 V + O(V 2 ). (13.87)
~(ω − ω0 )

The term proportional to sin(ω0 t) comes from the term proportional to

i(ω−ω0 ) sin(Ωt)
2Ω in a(t) and is related to an oscillation of the electron with the
eigenfrequency of the oscillator. This is always possible and not related
to the response of the system. The change in the position of the electron
induced by the oscillating electromagnetic field to linear order in V is given
by
sin(ωt)
z(t) = z0 V , (13.88)
~(ω − ω0 )
such that the susceptibility is given as
1
χ∝− . (13.89)
ω − ω0
The classical response function without damping is given as
 
1 1
χ∝ − + . (13.90)
ω − ω0 ω + ω0
We first of all find that our quantum model compares to a classical har-
monic oscillator without damping. We furthermore see that the quantum
response function in the rotating wave approximation only has a pole at
ω = ω0 , whereas the classical treatment for the response function also gives
a pole at ω = −ω0 . The fact that the second pole is missing is a real problem
and makes the quantum response function calculated on this level of theory
unphysical. The real part of χ decays as 1/ω for large ω. It follows that
the conductivity becomes constant for large ω and the scattering length

432
 13.2. Quantum example

diverges for large ω. We know that for massive particles there should not
be a response at large frequencies. There is thus clearly something not fully
correct with the approximations we made to solve the quantum mechanical
model.
The problem is in the rotating wave approximation. We neglected the
term proportional to e−i(ω+ω0 )t and only kept the term e−i(ω−ω0 )t in the
differential equation. This approximation is not completely valid. In order
to correct this we can, based on the solution found in the rotating wave
approximation, guess a solution for the time dependent wave function. For
the linear response we only need a correct solution to linear order in V ,
which makes such a guess feasible.
If we look at φ(t) obtained from the rotating wave approximation to
linear order in V we find
ω+ω0 !
e−i 2 t sin( ω−ω
2
0
t)
φ(t) = 1, V . (13.91)
~(ω − ω0 )

This equation does not fulfill the original Schrödinger equation to linear
order in V . If we test this function as a solution of the Schrödinger equation
we find,
 
d 1 iωt
i~ φ(t) − H(t)φ(t) = 0, ie V , (13.92)
dt 2

whereby we neglected all quadratic terms in V . For a solution to the

Schrödinger equation the right hand side should have been zero. The value
on the right hand side is not small, nor oscillates quickly inside an integral
such that we could assume it averages out.
Based on the classical result for the response function we can guess a
valid solution for φ(t) to linear order in V . We then only need to test this
solution in the Schrödinger equation to convince ourselves that our guessed
solution is indeed correct. In the susceptibility based on the equation de-
rived in the rotating wave approximation we found one term related to the
resonance at ω − ω0 = 0, but the term where the resonance is at ω + ω0 = 0

433
Chapter 13. Response theory

is missing. The time dependent wave function in the rotating wave approx-
imation has a projection of 1 onto the 1s basis orbital and a projection that
is linear in V and depends on ω on the 2pz basis orbital. If we subtract
a second term to the projection on the 2pz orbital that is also linear in V
whereby we change ω to −ω we find the function
ω+ω0 −ω+ω0 !
e−i 2 sin( ω−ω
t
2 t)
0
e−i 2 t sin( −ω−ω
2
0
t)
φ(t) = 1, V −V (13.93)
~(ω − ω0 ) ~(−ω − ω0 )
 
V −iω0 t


= 1, −iωe + iω cos(ωt) + ω0 sin(ωt) ,
~(ω 2 − ω02 )

which indeed is a solution to linear order in V for the Schrödinger equation

and has the susceptibility

2ω0
χ∝− (13.94)
− ω02
ω2
1 1
∝− + .
ω − ω0 ω + ω0

13.3 General response derived in the frequency

domain
In the previous section we derived the response of an Hydrogen atom in
an oscillating electromagnetic field with frequency ω. In this section we
want to generalize these equations for the response of a system given by
an arbitrary Hamiltonian to an action given by an arbitrary operator. We
will start our derivation from the results of the previous section. For a
system defined on a Hilbert space with two states we can define the matrix
representation of the Hamiltonian on the eigenbasis, |φ0 i and |φ1 i as
 
0 0
H0 − E0 = , (13.95)
0 ω1

434
 13.3. General response derived in the frequency domain

with E0 = hφ0 |H 0 |φ0 i. We can now look at the response of this system to
a perturbation proportional to an operator that couples state 0 and 1
 
0 µ1
O= . (13.96)
µ∗1 0

We define the total Hamiltonian as

H(t) = H 0 + H 1 (t), (13.97)

with

H 1 (t) = λ O sin(ωt). (13.98)

At t = 0 we assume the state of the system to be known as

|φ(t = 0)i = |φ0 i . (13.99)

Starting from this state we find, to linear order in λ, the time dependent
state to be
−iωe−iω1 t + iω cos(ωt) + ω1 sin(ωt)
|φ(t)i = |φ0 i + λµ∗1 2 2 |φ1 i + O(λ2 ).
~(ω − ω1 )
(13.100)

We note that the state is only normalized to linear order in λ. We find that
the occupation of the ground state |φ0 i is not altered in linear order in λ,
i.e. hφ0 |φ(t)i = 1. The linear response defines how one starts to deviate
from the state |φ0 i but the deviation always stays small. With the use of
|φ(t)i we can calculate the time dependent expectation value of operator
O. To linear order in λ we find:
 
2 1 1
hφ(t)|O|φ(t)i = λ sin(ωt)|µ1 | − , (13.101)
~(ω − ω1 ) ~(ω + ω1 )
such that the susceptibility becomes
hφ(t)|O|φ(t)i
χ(1) (ω) = (13.102)
λ sin(ωt)

435
Chapter 13. Response theory

 
2 1 1
= |µ1 | − .
~(ω − ω1 ) ~(ω + ω1 )

In order to generalize this result we can enlarge the Hilbert space from
2 to n + 1 states. We assume that |φ0 i does not couple to a single state,
but to n different states. On the eigenbasis of H 0 the Hamiltonian H 0 and
operator O can be written as
n
~ωi a†i ai ,
X
H 0 − E0 = (13.103)
i=1

and
n
µ∗i a†i a0 + µi a†0 ai .
X
O= (13.104)
1=1

The total Hamiltonian is given as H(t) = H 0 + λO sin(ωt), similar as

defined in equation 13.97. If we take the side condition that at t = 0 the
state is given by |φ0 i, then we find in equation 13.101 the time dependent
state to linear order in λ if one would couple to only a single other state. As
the projection of |φ(t)i onto |φ0 i is one for all times the coupling between
state |φi i and |φ(t)i is independent of the coupling of state |φ(t)i to other
states. We find for the time dependent wave function
n
X
|φ(t)i = |φ0 i + λ µ∗i f (t; ω, ωi ) |φi i + O(λ2 ), (13.105)
i=1

with
−iωe−iωi t + iω cos(ωt) + ωi sin(ωt)
f (t; ω, ωi ) = . (13.106)
~(ω 2 − ωi2 )

The expectation value of O to linear order in λ becomes

n  
X
2 1 1
hφ(t)|O|φ(t)i = λ sin(ωt) |µi | − , (13.107)
~(ω − ωi ) ~(ω + ωi )
i=1

436
 13.3. General response derived in the frequency domain

and the susceptibility

hφ(t)|O|φ(t)i
χ(1) (ω) = (13.108)
E0 sin(ωt)
n  
X
2 1 1
= |µi | − .
~(ω − ωi ) ~(ω + ωi )
i=1

Each state contributes separately to the response and thus the suscepti-
bility. On the eigenbasis of H 0 the response of the total system is given by
the sum of the response of the state |φ0 i coupling to each state separately.
This is a general feature of linear response theory in a quantum mechanical
system and extremely simplifies the treatment of response theory.
The previous derivation assumed the Hamiltonian to be on its eigen-
basis. In this case the matrix representation of H 0 is diagonal and the
response due to each state decouples from the response to all other states.
In a next step we can generalize the previous form of the susceptibility to
general Hamiltonians. We first note that
n
X
O |φ0 i = µ∗i |φi i . (13.109)
i=1

Furthermore the matrix

1 1
− , (13.110)
~ω − H0 + E0 ~ω + H0 − E0

is diagonal with the entries

1 1
− , (13.111)
~(ω − ωi ) ~(ω + ωi )

on the diagonal. If we put this together we find that

   
(1) 1 1
χ (ω) = φ0 O − O φ0 . (13.112)
~ω − H 0 + E0 ~ω + H 0 − E0

437
Chapter 13. Response theory

We can express the susceptibility as an expectation value of the state |φ0 i.

We derived equation 13.112 on an eigenbasis of the Hamiltonian. The
expectation value of an operator is however basis set independent, such that
equation 13.112 is general valid, also on a basis where the Hamiltonian is
non-diagonal.
We can make one more generalization. In equation 13.98 we assumed
that H 1 (t) is proportional to O. For many experiments this is the case,
but this must not necessarily hold. If one measures the electric (magnetic)
dipole moment induced by an electric (magnetic) field the perturbing oper-
ator is proportional to the operator whose expectation value we calculate.
However if one measures the electric dipole moment induced by a magnetic
field H 1 (t) is not proportional to O. For the latter case we can take

H 1 (t) = λH 1 sin(ωt), (13.113)

and
 
(1) 1 1
χ (ω) = φ0 O H1 − H1 O φ0 . (13.114)
~ω − H 0 + E0 ~ω + H 0 − E0

13.4 General response derived in the time domain

Instead of looking at the response of the system to a driving force that
oscillates in time at a given frequency ω we can have a look at the response
in the time domain. We assume that the Hamiltonian is given as

H(t) = H 0 + H 1 (t), (13.115)

with H 0 time independent and H 1 (t) a time dependent operator. Naturally

we can take the Fourier transform of H 1 (t) such that
Z ∞ Z ∞
1
H 1 (ω) = iωt
H 1 (t)e dt ⇐⇒ H 1 (t) = H 1 (ω)e−iωt dω,
−∞ 2π −∞
(13.116)

438
 13.4. General response derived in the time domain

defining a frequency dependent Hamiltonian. We then can calculate the

response to H 1 (ω) in the frequency domain, however it is educational to
also look at the response directly in the time domain.
The question we want to answer to derive the response functions is:
Given the Hamiltonian H(t) and a known state at t = 0
|φ(t = 0)i = |φ0 i , (13.117)
what is the state |φ(t)i at time t.
For a time independent Hamiltonian we can easily calculate the time
evolution of a state
E t
φ̃(t) = e−iH 0 ~ |φ0 i . (13.118)

For a time dependent Hamiltonian we can not simply replace H 0 by H(t).

However at each time t we can calculate the time evolution for an infinites-
imal time step δt as
δt
|φ(t + δt)i = e−iH(t) ~ |φ(t)i . (13.119)
If we take infinite of such infinitesimal time steps we arrive at the time
evolution from t = 0 to t = t
N
0 δt0
Y
|φ(t)i = lim e−iH(t ) ~ |φ0 i , (13.120)
N →∞
j=1

with δt0 = N1 t and t0 = jδt0 .

We can introduce a different notation for equation 13.120 using the time
ordering operator. In general we know that for commuting operators we can
combine the product of two exponential functions in to a single exponential
δt δt δt
e−iH(t+δt) ~ e−iH(t) ~ = e−i(H(t+δt)+H(t)) ~ , when [H(t), H(t + δt)] = 0.
(13.121)
When [H(t), H(t + δt)] 6= 0 the former equation is not valid. We do have
∞  n !
δt δt X 1 δt
e−iH(t+δt) ~ e−iH(t) ~ = −iH(t + δt) (13.122)
n! ~
n=0

439
Chapter 13. Response theory

∞  !
δt n

X 1
× −iH(t)
n! ~
n=0
∞  !
δt n

X 1
=T −i(H(t + δt)H(t))
n! ~
n=0
 δt

= T e−i(H(t+δt)+H(t)) ~ ,

with T the time ordering operator that places operators acting on a later
time left of operators acting on an earlier time.
With the use of the time ordering operator we find for equation 13.120
 P 
−i N H(t0 ) δt0
|φ(t)i = lim T e j=1 ~ |φ0 i (13.123)
N →∞
 i Rt 0 0

= T e− ~ 0 H(t )dt |φ0 i .

Even for time dependent Hamiltonians we can write the time evolution
operator as a single exponential, however we need to order different powers
in the Hamiltonian according to the time at which they act.
In order to solve for |φ(t)i we go back to equation 13.120 and expand
the exponential. We can stop at linear order in δt0 as we take the limit of
N → ∞ such that δt0 goes to zero. We find

N  
Y i 0 0 02
|φ(t)i = lim 1 − H(t )δt + O(δt ) |φ0 i (13.124)
N →∞ ~
j=1
N  
Y i 0 i 0 0
= lim 1 − H 0 δt − H1 (t )δt |φ0 i .
N →∞ ~ ~
j=1

We now sort this expression in orders of H 1 (t) to find

N  
Y i 0
|φ(t)i = lim 1 − H 0 δt |φ0 i (13.125)
N →∞ ~
j=1

440
 13.4. General response derived in the time domain

N N   
X Y i 0 i 0 0
+ lim 1 − H 0 δt − H 1 (nδt )δt
N →∞ ~ ~
n=1 j=n+1
n−1
Y !
i 0
× 1 − H 0 δt |φ0 i
~
j=1

+ ...
i
=e− ~ H 0 t |φ0 i
i t − i H0 (t−t1 )
Z
i
− e ~ H 1 (t1 )e− ~ H 0 t1 dt1 |φ0 i
~ 0
Z t Z t2
1 i i i
− 2 e− ~ H 0 (t−t2 ) H 1 (t2 )e− ~ H 0 (t2 −t1 ) H 1 (t1 )e− ~ H 0 t1 dt1 dt2 |φ0 i
~ 0 0
+ ....

If we evaluate this equation for a two level system in linear order in

H 1 we reproduce the result found in equation 13.101. For a general sys-
tem calculating the time dependent wave function is difficult. Whereas
the ground-state of a many electron system often can be expressed on a
mean-field level as a single Slater determinant, this is normally not a good
approximation for excited states and thus not a good approximation for the
time evolution of a system. Instead of looking at the state, we can ask the
question what is the expectation value of some given operator O at time
t. The latter contains less information than the full wave function, but is
easier to calculate.
To linear order in H 1 (t) we can calculate the time dependent expecta-
tion value of O as
D i i
E
hφ(t)|O|φ(t)i = φ0 e ~ H 0 t Oe− ~ H 0 t φ0 (13.126)
 Z t 
i i
H0t − ~i H 0 (t−t1 ) − ~i H 0 t1
− φ0 e ~ O e H 1 (t1 )e dt1 φ0
~ 0
 Z t 
i i
H 0 t1 i
H 0 (t−t1 ) − ~i H 0 t
+ φ0 e ~ H 1 (t1 )e ~ dt1 Oe φ0 .
~ 0

441
Chapter 13. Response theory

The first term originates from the zero-th order expansion of the bra and
ket state. The second and third term consider either the bra or ket state
in first order in H 1 (t) and the other state in 0-th order. The expectation
value simplifies to
D i i
E
hφ(t)|O|φ(t)i = φ0 e ~ H 0 t Oe− ~ H 0 t φ0 (13.127)
Z tD
i i i i i
E
− φ0 e ~ H 0 t Oe− ~ H 0 t e ~ H 0 t1 H 1 (t1 )e− ~ H 0 t1 φ0 dt1
~ 0
i tD
Z E
i i i i
+ φ0 e ~ H 0 t1 H 1 (t1 )e− ~ H 0 t1 e ~ H 0 t Oe− ~ H 0 t φ0 dt1
~ 0
D i i
E
= φ0 e ~ H 0 t Oe− ~ H 0 t φ0
i t D h i H 0 t1
Z i E
i i i
+ φ0 e ~ H 1 (t1 )e− ~ H 0 t1 , e ~ H 0 t Oe− ~ H 0 t φ0 dt1 .
~ 0
This equation simplifies if we use the definition of the operator in the
interaction picture:
i i
O H (t) = e ~ H 0 t Oe− ~ H 0 t , (13.128)

for time independent operators and

i i
H 1 (t)H = e ~ H 0 t H 1 (t)e− ~ H 0 t , (13.129)

for time dependent operators. With this notation we can write a compact
equation for the time dependent expectation value

i t
Z
hφ(t)|O|φ(t)i = hφ0 |O H (t)|φ0 i + hφ0 |[H 1 (t1 )H , O H (t), ]|φ0 i dt1
~ 0
(13.130)

We now can compare the result for the expectation value of the opera-
tor O with the definition of the response function as given by the Volterra
series in equation 13.1. The 0-th order term in the Volterra series of the sus-
ceptibility does not dependent on the time dependent perturbation H 1 (t)

442
 13.4. General response derived in the time domain

and is given by

χ(0) (t) = hφ0 |O H (t)|φ0 i . (13.131)

In order to write the linear term as a susceptibility we need to assume

that we can write the time dependence of H 1 (t) as

H 1 (t) = H 1 f (t), (13.132)

with f (t) some arbitrary function in time and H 1 time independent. We

then find for the term linear in H 1 of the expectation value of O:

i t
Z
hφ(t)|O|φ(t)i = hφ0 |[H 1H (t1 ), O H (t)]|φ0 i f (t1 )dt1 . (13.133)
~ 0

The linear susceptibility is thus defined as

i
χ(1) (t; t1 ) = θ(t1 )θ(t − t1 ) hφ0 |[H 1H (t1 ), O H (t)]|φ0 i , (13.134)
~
whereby the Heaviside step function θ is added to allow a freedom of inte-
gration boundaries.
For a general state the susceptibility explicitly depends on the time t
and t1 . If you are not in a stationary state of the Hamiltonian H 0 , there will
be a non-trivial time evolution of the state due to H 0 . This time evolution
leads to a change of expectation values, such that χ(1) (t; t1 ) depends both
on t and t1 explicitly. If |φ0 i is an eigenstate of H 0 the time evolution with
H 0 only changes the phase of the wave function. For the expectation value
of eigenstates of H 0 , the time dependent phase obtained by the bra and
get state cancels. As a result the expectation values are independent of the
time when the total experiment is done. When |φ0 i is and eigenstate of H 0
we find that the susceptibility only depends on the time difference between
t and t1 .

443
Chapter 13. Response theory

13.5 Relation between frequency and time domain

We can relate the response function in the time and frequency domain by
making a Fourier transforms of the susceptibility in the frequency domain
as given in equation 13.112. In order to simplify notation we introduce the
retarded Green’s operator
1
GR± (ω) = lim . (13.135)
η→0+ ω ∓ H 0 ± E0 + iη
The Green’s operator can be used to calculate Green’s functions, of some
systems, but also to calculate the response functions of the system. With
the use of the Green’s operator we can express the susceptibility as
χ(1) (ω) = φ0 OGR+ O φ0 − φ0 OGR− O φ0 . (13.136)
In order to obtain the response function in the time domain we can Fourier
transform the response function in the frequency domain. This results in
a Fourier transform of the Green’s operator. The Fourier transform of the
Green’s operator is
Z ∞
1
GR± (t) = e−iωt GR± (ω)dω. (13.137)
2π −∞
In oder to evaluate this integral we can close the contour in the upper half
complex plane. As there are no poles enclosed in the contour over the upper
half plane we have
I
1
e−iωt GR± (ω)dω = 0, (13.138)
2π upper half
or
Z
1
G R±
(t) = − e−iωt GR± (ω)dω. (13.139)
2π arc

With ω = R0 (cos(Θ) + i sin(Θ) the integral over the arc transforms in an

integral over Θ for R0 → ∞.
Z π
1
R±
G (t) = − lim e−iR0 cos(Θ)t eR0 sin(Θ)t GR± (ω(Θ))dΘ. (13.140)
2π R0 →∞ 0

444
 13.5. Relation between frequency and time domain

This integral is zero for t < 0 as limR0 →∞ eR0 sin(Θ)t = 0 for t < 0. We thus
find that

GR± (t < 0) = 0. (13.141)

For t > 0 we can look at the contour integral around the lower half
complex plane. Using the same argumentation as before we can conclude
that in the lower half complex plane the integral over the arc is zero for
t > 0 such that the contour integral is equal to the integral over the real
axis.
I
1
R±
G (t > 0) = e−iωt GR± (ω)dω (13.142)
2π lower half
 
−iωt 1
= −i res e .
ω ∓ H 0 ± E0 + i0+

The Green’s operator has poles in the lower half complex plane at ω =
±H 0 ∓ E0 with residue 1. We thus find

GR± (t > 0) = −ie−i(±H 0 ∓E0 )t . (13.143)

If we combine the solution for t < 0 and t > 0 we find

GR± (t) = −iθ(t)e−i(±H 0 ∓E0 )t . (13.144)

For the susceptibility that relates the expectation value of operator O

at time t for a system given by the Hamiltonian H 0 + H 1 f (t0 ) starting in
an eigenstate of H 0 we find
Z ∞
1
(1)
χ (t) = χ(1) (ω)e−iωt dω (13.145)
2π −∞
Z ∞
1
= φ0 OGR+ (ω)H 1 − H 1 GR− (ω)O φ0 e−iωt dω
2π −∞
= φ0 OGR+ (t)H 1 − H 1 GR− (t)O φ0
D E
= −iθ(t) φ0 Oe−i(H 0 −E0 )t H 1 − H 1 e−i(−H 0 +E0 )t O φ0

445
Chapter 13. Response theory

= −iθ(t) φ0 eiE0 t Oe−iH 0 t H 1 − H 1 eiH 0 t Oe−iE0 t φ0

= −iθ(t) φ0 eiH 0 t Oe−iH 0 t H 1 − H 1 eiH 0 t Oe−iH 0 t φ0
= −iθ(t) hφ0 |O H (t)H 1H (t = 0) − H 1H (t = 0)O H (t)|φ0 i
= iθ(t) hφ0 |[H 1H (t = 0), O H (t)]|φ0 i ,

which concludes the relation between χ(ω) and χ(t).

13.6 Dissipation
In equation 13.112 we derived the linear susceptibility for a general operator
O. We were able to express this as a ground-state expectation value. The
problem with this equation is that the operator from which we take the
expectation value is Hermitian. The resulting susceptibility is thus real. If
we then calculate the conductivity and from that the dissipation we find
that there is no absorption of energy in the system. This is indeed the
correct answer. When driven by an oscillatory force, a finite system might
absorb some energy during some time, but will restore the energy into the
driving field during some other time, such that there is no total dissipation.
Based on the classical result we can add dissipation to the system by
shifting the poles in the susceptibility from the real axis slightly into the
complex plain
   
(1) 1 1
χ (ω) = φ0 O − O φ0 .
~ω − H 0 + E0 + i γ2 ~ω + H 0 − E0 + i γ2
(13.146)
Although this works, this is not a completely satisfactory way to add dis-
sipation to our theory.
A finite size system described by quantum mechanics does not have
dissipation. We need an infinite system to get absorption. There are two
conceptual ways we can introduce an infinite basis. We either can assume
that the operator O does not couple state |φ0 i to a single state, but to
infinite many within a fixed energy interval, or we can assume that the
state |φ0 i couples to a single state O |φ0 i. This state O |φ0 i then couples

446
 13.6. Dissipation

to infinite many states in a finite energy interval around the energy of the
state O |φ0 i. Both descriptions are mathematically equivalent and describe
the same process.
An example for the first method to include dissipation, would be a
model where we have an energy band of states that can be excited from
the ground-state. We can make a simple model assuming that we have N
states in the interval ωmin
p to ωmax , equidistantly spaced in energy with a
coupling strength µi = µ 1/N . We then take the limit of N to infinity to
generate an excitation into the continuum. The Hamiltonian for this model
is given as
N  
i − 1/2
a†i ai .
X
H0 = ωmin + (ωmax − ωmin ) (13.147)
N
i=1

The coupling of the ground-state to the excited states by the time depen-
dent perturbation is given as H 1 (t) = H 1 (t) sin(ωt) with
N  
µi a†i a0 + a†0 ai .
X
H1 = (13.148)
i=1

The susceptibility due to each state a†i |0i is given by equation13.112,

and for N of such states we can sum the susceptibility induced by each
state to obtain the susceptibility of the total system. For N → ∞ the sum
becomes an integral and we find for the susceptibility of our model
Z ωmax  
(1) 2 1 1
χ (ω) = lim µ − dω0 (13.149)
γ→0+ ωmin ω − ω0 + i γ2 ω + ω0 + i γ2
2
+ (γ − 2iω)2 − log 4ωmax 2
+ (γ − 2iω)2 ,




= lim log 4ωmin
γ→0+

whereby we added a small imaginary part to have a unique definition of the

branch cuts to take for the integration and the logarithm. The imaginary
part of this function is zero except for ω in the interval from ωmin to ωmax .
In this interval the imaginary part of χ(1) (ω) is constant and has the value
πµ2 . We thus find a constant imaginary part of the susceptibility within

447
Chapter 13. Response theory

this model when the excitation energy is such that one can excite state |φ0 i
into the band of excited states. As a result this model shows dissipation
when the system is excited with light for which ω < ωmax and ω > ωmin .
Instead of looking at a single state that couples to a band of states we
can make a model where a single ground-state can be excited into a single
excited state. The excited state then couples to a continuum, which will
generate the dissipation. As an exemplary model for the second method to
induce dissipation we can look at the absorption of light in the hydrogen
atom. We will slightly simplify the coupling of light to our discrete states
as we do not need to calculate the exact form of the line broadening and
lifetime, but want to elude the general mechanism that induces dissipation
in quantum systems. As a model we look at the hydrogen atom where
we can excite the 1s state into the 2p state. If we include photons as
particles in our basis we reproduce the two level example discussed in the
previous section on a basis of the 1s orbital plus a photon and the 2p orbital
without a photon. These two states are coupled by the dipole moment of
the electron between the 1s and 2p orbital and the photon field at the
position of the atom. The resulting susceptibility is purely real with a pole
at ~ω ± ~ω0 with ~ω0 the energy difference between the 1s and 2p state
and ~ω the energy of the photon present in combination with the electron
in the 1s orbital. We can now extend the basis set with the infinite number
of states given by the electron in the 1s orbital and a photon with energy
~ωi . To be precise one should quantize the photons by their wave vector as
there are many states with the same energy possible, for the current model
quantizing by the energy is good enough. We will assume that the coupling
strength between the√2p state and the 1s state plus a photon with energy
~ωi is equal to µ(ωi ) δωi . The infinite set of states with the electron in
the 1s orbital and one additional photon now enter as an extension of the
Hilbert space only coupling the the state with the electron in the 2p orbital
and thus as a self energy in the susceptibility. We find

 
(1) 2 1 1
χ (ω) = µ − , (13.150)
ω − ω0 − Σ+ (ω) ω + ω0 + Σ− (ω)

448
 13.7. Kramers Kronig relations

with
Z ∞
1
Σ± (ω) = ± lim µ2 (ωi ) dωi . (13.151)
γ→0+ 0 ω ∓ ωi + i γ2

For atomic transitions the wavelength of the light at resonant is normally

large compared to the the size of the atom. As a result µ(ωi ) is nearly
constant and thus the imaginary part of Σ(ω) is roughly constant around
ω0 . The natural line-shape of atomic resonances is thus Lorentzian.
Both methods to make χ(ω) complex and induce dissipation in the
system are mathematically equivalent and thus describe the same physics.
If one would diagonalize the Hamiltonian of an Hydrogen atom on a basis of
the electron in the 2p orbital and the infinite set of states with the electron
in the 1s orbital plus a photon one would find a set of eigenstates with a
partial 2p character that is spread in energy and peaks around ω0 above
the energy of the 1s state. The susceptibility then can be calculated via
the first method whereby one can make a transition in the continues band
of states. In all cases, once we go to an infinite basis sets the susceptibility
becomes complex such that the system has dissipation.

13.7 Kramers Kronig relations

Response functions are given by a complex function with real and imagi-
nary part. The real and imaginary part of the response function are not
independent of each other. One can determine one from the other. In order
to guarantee causality, i.e. the system does not respond to actions that are
later than the time we are looking at, we know that χ(ω), σ(ω) and f (ω)
are analytical in the upper half complex plane. We furthermore know that
χ(ω) falls off to zero for |ω| → ∞ faster then 1/|ω|, or more precisely: For
all  > 0 there is an R ∈ R such that |ωχ(ω)| <  for all |ω| ≥ R. These
are sufficient conditions to uniquely relate the real and imaginary part of
the response functions. These relations are known as the Kramers Kronig
relations. We can start the derivation of the Kramers Kronig relations by
noting that as there are no poles in the upper half complex plane we know

449
Chapter 13. Response theory

that a contour integral around the upper half complex plain of χ(ω), σ(ω)
and f (ω) yields zero.
I Z +∞ Z π
χ(ω)dω = χ(ω)dω + lim χ(reiφ )rdφ (13.152)
γ −∞ r→∞ 0

= 0.

Furthermore we know that

Z π
lim χ(reiφ )rdφ = 0, (13.153)
r→∞ 0

as χ(ω) falls of to zero for |ω| → ∞ faster than 1/|ω|.

Instead of looking at the contour integral over χ(ω) we can look at the
function ωχ(ω) χ(ω)
0 −ω . The function ω 0 −ω has poles in the lower complex plane due

to χ(ω) and one pole at the real axis at ω = ω 0 . If we now use the Cauchy’s
residue theorem we find
Z ∞
χ(ω 0 ) χ(ω 0 )
I 
0 0
0
dω = P 0
dω (13.154)
γ ω −ω −∞ ω − ω
= iπ res(χ(ω 0 ), ω)
= iπ χ(ω),

with P the principle value integral. This yields the Kramers Kronig rela-
tions
Z ∞
χ(ω 0 )

0
P 0
dω = iπχ(ω), (13.155)
−∞ ω − ω

or equivalent
Z ∞
Re(χ(ω 0 )) 0

P dω = −πIm(χ(ω)), (13.156)
ω0 − ω
Z−∞∞
Im(χ(ω 0 )) 0

P 0
dω = πRe(χ(ω)).
−∞ ω − ω

450
Appendix A

The Dirac delta function

(distribution)

The Dirac delta function, δ(~r) is an object that is extremely useful for many
parts of physics. The function is defined on Rd with d ∈ N, and is zero for
all values, except for |~r| = 0. We furthermore define the function to be
normalized such that
Z
δ(~r)dd~r = 1. (A.1)
Rd

There is one problem. There is no function that can fulfill these prop-
erties. The integral over a function that is zero on its domain, except for a
singular point, is zero. As a solution we will define the Dirac delta function
as a special limiting case of a series of functions.
We can define a function δ(, ~r) on ~r ∈ Rd ,  ∈ R,  > 0, which is con-
tinues on some finite compact domain around ~r = 0 and has the following
properties
δ(, ~r) ≥ 0, and (A.2)
n
lim |~r| δ(, ~r) = 0, ∀n ∈ N, and
|~
r|→∞

lim δ(, ~r) = 0, ∀~r 6= 0, and

→0+

451
Appendix A. The Dirac delta function (distribution)

Z
lim δ(, ~r)dd~r = 1.
→0+ Rd

There are many functions δ(, ~r) that fulfil these properties. We can
take for example a rectangle distribution
(
d
1
 |~ri | ≤ /2 ∀i ∈ [1, d]
δ(, ~r) = , (A.3)
0 0

or use a Gaussian function

 d
1 r2
~
δ(, ~r) = √ e− 2ε2 . (A.4)
 2π
As long as the generating function fulfils the properties A.2 it can be used
to generate the delta function. Note that
Z
lim δ(, ~r)dd~r = 1, (A.5)
→0+ Rd

but that
Z
lim δ(, ~r)dd~r = 0. (A.6)
+
Rd →0

We thus find that, strictly speaking,

δ(~r) 6= lim δ(, ~r), (A.7)

→0+

although one can find the later statement often in the literature. The
limit should be understood to be taken after one integrates over the delta
function. The integral Rd δ(~r)dd~r is defined as
R

Z Z
δ(~r)dd~r = lim δ(, ~r)dd~r, (A.8)
Rd →0+ Rd

for any of the generating functions one chooses. The final result does not
depend on the choice of the generating function used, which is meant with
the statement δ(~r) = lim→0+ δ(, ~r).

452
 A.1. Properties of the Dirac delta function

A.1 Properties of the Dirac delta function

Some useful properties of the delta function include

A.1.1 Convolution
As δ(~r − ~r0 ) = 0 for all ~r 6= ~r0 it follows that for all continues functions f (~r)
for which the integral over δ(, ~r)f (~r) is convergent we have
Z
δ(~r0 − ~r)f (~r0 )dd~r0 = f (~r). (A.9)
Rd

A.1.2 Scaling
If we scale the delta function we find, for c ∈ R
 d
1
δ(c~r) = δ(~r), (A.10)
|c|

in particular

δ(−~r) = δ(~r). (A.11)

A.1.3 Derivative
d n
The n-th derivative of the delta function dx (n) (x) is well defined
n δ(x) = δ

and can be calculated using integration by parts.

Z ∞ ∞ Z ∞
(n)
f (x)δ (x)dx = f (x)δ (n−1)
(x) − f 0 (x)δ (n−1) (x)dx. (A.12)
−∞ −∞ −∞

As δ(x) = 0 for all x 6= 0 we know that δ(x) is constant except for x = 0.

The derivative of the delta function evaluated at ±∞ is thus 0. We find
Z ∞ Z ∞
f (x)δ (n) (x)dx = − f 0 (x)δ (n−1) (x)dx. (A.13)
−∞ −∞

453
Appendix A. The Dirac delta function (distribution)

This is true for all functions f (x) and thus not only the integrals, but also
the integrants must be equivalent

f (x)δ (n) (x) = −f 0 (x)δ (n−1) (x). (A.14)

By iterating this relation we find

f (x)δ (n) (x) = (−1)n f (n) (x)δ(x). (A.15)

454
Appendix B

Gaussian Integral

In several of the worked out problems and examples we encounter the fol-
lowing integral
Z ∞ r
ax2 +bx+c b2
c− 4a π
e dx = e , (B.1)
−∞ −a

for a, b, and c some complex numbers with Re(a) < 0. In this appendix we
will proof this relation.
We start by rewriting the exponent
Z ∞ Z ∞
b2 b 2
ax2 +bx+c c− 4a
e dx = e ea(x+ 2a ) dx (B.2)
−∞ −∞
b
2
b
Z ∞+ 2a
c− 4a 2
=e eau du,
b
−∞+ 2a

b b
with u = x + 2a . When 2a is real the change in integration limits does
b
not matter, while limx→∞ x = limx→∞ x + a for a ∈ R. However if 2a
is complex the path of integration in the complex plain actually becomes
different. As such we can not neglect the change in integration boundaries
2
without further commenting. Luckily we know that the integrant, eau has
no poles in the complex plain. The integral over a closed contour thus

455
Appendix B. Gaussian Integral

b b
vanishes. If we take the contour S : limR→∞ (−R, R, R + 2a , −R + 2a , −R)
then we find
I
2
eau du = 0 (B.3)
S

For Re(a) < 0 we find

b b
Z −R+ 2a Z R+ 2a
au2 2
lim e du = lim eau du (B.4)
R→∞ −R R→∞ R

= 0,

while the integrand becomes zero along this path. We thus find that
b
Z ∞+ 2a Z ∞
2 2
eau du = eau du. (B.5)
b
−∞+ 2a −∞

Note that for analytical functions without poles integrated from −∞ to

+∞ we can generally shift the boundaries of integration in the complex
plain. This is not true if the function has poles in the complex plain. In
that case, the value of the integral changes depending if the pole is above
or below the integration path. R
∞ 2
We can evaluate the integral −∞ eau du with the use of a little trick.
If we square the integral we have
Z ∞ 2 Z ∞ Z ∞
2 2 2
eau du = ea(x +y ) dxdy. (B.6)
−∞ −∞ −∞

We now shift to polar coordinates to find

Z ∞ 2 Z 2π Z ∞
au2 2
e du = ear rdrdθ (B.7)
−∞ 0
Z ∞0
2
= 2π ear rdr
0
Z ∞
2
=π ear dr2
0

456
 ∞
1 2
= π ear .
a 0

With the use of Re(a) < 0 we find

Z ∞ 2
au2 1
e du = π (0 − 1) (B.8)
−∞ a
π
=−
a
Combining all results we have
Z ∞ r
ax2 +bx+c b2
c− 4a π
e dx = e , (B.9)
−∞ −a

for a, b, and c some complex numbers if Re(a) < 0, which is the equation
we set out to derive.

457
Appendix B. Gaussian Integral

458
Appendix C

Fourier Transform

Given an integrable function f : R 7→ C we define the Fourier transform be-

tween position (x) and momentum (p) or wavevector (k = p/~) coordinates
as
r Z ∞
1
g(k) = f (x)e−ikx dx. (C.1)
2π −∞
r Z ∞
1
f (x) = g(k)eikx dk.
2π −∞

The Fourier transform between time (t) and energy (E) or angular fre-
quency (ω = E/~) coordinates is defined as
r Z ∞
1
g(ω) = f(t)eiωt dt. (C.2)
2π −∞
r Z ∞
1
f(t) = g(ω)e−iωt dω.
2π −∞

The sign definitions are chosen such that for the description of a plane
wave with wave vector k0 and angular frequency ω0 we can take g(ω) =
δ(ω − ω0 ), and g(k) = δ(k − k0 ), with δ the Dirac delta function (Appendix
A), such that the functions in frequency and wave vector domain peak at

459
Appendix C. Fourier Transform

the values of the plane wave they describe. The functions in the position
and time domain become
r Z ∞ ! Z
∞ 
1 ikx 1 −ikx
f (x)f(t) = g(k)e dk g(ω)e dω (C.3)
2π −∞ 2π −∞
r Z ∞ ! r Z !
∞
1 ikx 1 −iωt
= δ(k − k0 )e dk δ(ω − ω0 )e dω
2π −∞ 2π −∞
1 ikx −iωt
= e e
2π
1 i(kx−ωt)
= e .
2π
Which has the correct behaviour for a plane wave propagating in the posi-
tive x direction in time for positive wave vectors.

C.1 Proof that the inverse Fourier transform is

the inverse operator of the Fourier transform
In equations C.1 and C.2, we defined both the Fourier transform (f (x) 7→
g(k)) as well as the inverse Fourier transform (g(k) 7→ f (x)). In principle
we only needed to define one of the transformations as the other follows as
its inverse. We can proof the consistency of the definition of the Fourier
and inverse Fourier transform by transforming the Fourier transform with
the inverse Fourier transform and show we retrieve the original function:

∞
r Z
1
f (x) = g(k)eikx dk (C.4)
2π −∞
!
∞ ∞
r Z r Z
1 1 0 −ikx0 0
= f (x )e dx eikx dk
2π −∞ 2π −∞
Z ∞ Z ∞
1 0
= f (x0 )eik(x−x ) dx0 dk.
2π −∞ −∞

460
 C.2. Properties of the Fourier transform

With y = x − x0 we find
Z ∞ !
∞
r Z r
1 1 iky
f (x) = f (x − y) e dk dy. (C.5)
−∞ 2π −∞ 2π

Whereby the inner integral is the (Inverse) Fourier transform of a constant.

We now q use that the Fourier transform as well as the inverse Fourier trans-
1
form of 2π , a constant, is the Dirac delta function as we will show later,
(C.3.3). We thus find
Z ∞
f (x) = f (x − y)δ(y)dy (C.6)
−∞
= f (x),

which concludes the proof.

C.2 Properties of the Fourier transform

C.2.1 Linearity
The Fourier transform is a linear operation. For all integrable functions
f1 (x) and f2 (x) for which
r Z ∞
1
g1 (k) = f1 (x)e−ikx dx, (C.7)
2π −∞

and
∞
r Z
1
g2 (k) = f2 (x)e−ikx dx, (C.8)
2π −∞

we find
∞
r Z
1
g1 (k) + g2 (k) = (f1 (x) + f2 (x)) e−ikx dx. (C.9)
2π −∞

461
Appendix C. Fourier Transform

C.2.2 Derivative
For all n-fold differentiable functions f (x) that are integrable and fulfill
dm
that limx→±∞ dx m f (x) = 0 for all m ∈ N and m ≤ n we find that
Z ∞ n  Z ∞
1 d −ikx n 1
√ f (x) e dx = (ik) √ f (x)e−ikx dx. (C.10)
2π −∞ dxn 2π −∞
We can proof this using integration by parts
Z ∞ n   n−1  ∞
1 d −ikx 1 d −ikx
√ f (x) e dx = √ f (x) e (C.11)
2π −∞ dxn 2π dxn−1 −∞
Z ∞  n−1  
1 d d −ikx
−√ n−1
f (x) e dx
2π −∞ dx dx
Z ∞
dn−1
 
1
= (ik) √ n−1
f (x) e−ikx dx.
2π −∞ dx
d m
Whereby we used that limx→±∞ dx m f (x) = 0 for all m ∈ N and m ≤ n.

By applying integration by parts n times we arrive at equation C.10 and

conclude the proof.

C.2.3 Convolution
The Fourier transform of the convolution of two functions is the product
of the Fourier transform of each of these functions. Given two integrable
functions f1 (x) and f2 (x) their convolution f (x) is given as
Z ∞
f (x) = f1 (x − x0 )f2 (x0 )dx0 . (C.12)
−∞

Furthermore their Fourier transforms are given as

Z ∞
1
g1 (k) = √ f1 (x)e−ikx dx, (C.13)
2π −∞
and
Z ∞
1
g2 (k) = √ f2 (x)e−ikx dx. (C.14)
2π −∞

462
 C.3. The Fourier transform of some useful functions

For the convolution theorem we find that

1
Z ∞ √
√ f (x)e−ikx dx = 2πg1 (k)g2 (k). (C.15)
2π −∞
We can proof equation C.15 by inserting the inverse Fourier transform
of g1 (k) for f1 (x − x0 ) into the definition of the convolution and change the
order of integration over k and x0
Z ∞
f (x) = f1 (x − x0 )f2 (x0 )dx0 (C.16)
−∞
Z ∞ Z ∞ 
1 0
= √ g1 (k)eik(x−x ) dk f2 (x0 )dx0
−∞ 2π −∞
Z ∞ Z ∞ 
1 0 −ikx0 0
=√ g1 (k) f2 (x )e dx eikx dk
2π −∞ −∞
Z ∞
= g1 (k)g2 (k)eikx dk.
−∞

Next we Fourier transform both the left and√right hand side. Whereby we
realize that the right hand side is currently 2π times the inverse Fourier
transform of g1 (k)g2 (k). We find
1
Z ∞ √
√ f (x)e−ikx dx = 2πg1 (k)g2 (k), (C.17)
2π −∞
which is the relation we wanted to proof.

C.3 The Fourier transform of some useful func-

tions
C.3.1 Fourier transform of a delta function
The Fourier transform of the Dirac delta function yields
r Z ∞
1
g(k) = δ(x)e−ikx dx (C.18)
2π −∞

463
Appendix C. Fourier Transform

r
1
= .
2π

C.3.2 Fourier transform of a Gaussian function

Given a Gaussian function
1 (x−µ)2
f (x) = √ e− 2σ2 , (C.19)
σ 2π
with σ > 0. We find for its Fourier transform
r Z ∞
1 1 (x−µ)2
g(k) = √ e− 2σ2 e−ikx dx. (C.20)
2π −∞ σ 2π
1 k2 σ 2 +2ikµ
= √ e− 2 .
2π
Whereby we used in the last step the solution for a Gaussian integral as
discussed in appendix B.

C.3.3 Fourier transform of a constant

In order to calculate the Fourier transform of a constant f (x) = c with c ∈ C
we can use, as shown in C.3.1, that the Fourier transform of a delta function
yields a constant. We thus should expect that the Fourier transform of a
constant should yield the delta function. In order to proof this we can look
at the Fourier transform of a function that can be used to define the delta
function in an appropriate limit. For this we will use a Gaussian function.
We write
x2 σ 2
f (σ, x) = ce− 2 , (C.21)

and

f (x) = lim f (σ, x) (C.22)

σ→0
=c

464
 C.3. The Fourier transform of some useful functions

The Fourier transform of f (σ, x) is the Fourier transform of a Gaussian

function and yields
r Z ∞
1 x2 σ 2
g(σ, k) = ce− 2 e−ikx dx (C.23)
2π −∞
k2
e− 2σ
=c
√σ
= c 2πδ(σ, k),
k2
with δ(σ, k) = σ√12π e− 2σ a function that can be used to generate the delta
function.
If we are only interested in properties of g(σ, k) that can be calculated
from integrals over g(σ, k) we find for the Fourier transform of a constant
f (x) = c:
r Z ∞
1
g(k) = ce−ikx dx (C.24)
2π −∞
√
= c 2πδ(k).

In particularly we find that the Fourier transform of √12π is δ(x), the

Dirac delta function, and that the inverse Fourier transform of the Dirac
delta function is √12π (C.3.1).

C.3.4 Fourier transform of a polynomial

For the Fourier transform of xn we find
√ dn
r Z ∞
1
xn e−ikx dx = in 2π n δ(k). (C.25)
2π −∞ dk
We can proof equation C.26 by first Inverse Fourier transforming both the
left and right hand side of this equation.
√
r Z ∞  n 
n 1 n d
x = i 2π δ(k) eikx dk. (C.26)
2π −∞ dk n

465
Appendix C. Fourier Transform

We now use the properties of the delta function and its derivative, in par-
ticular equation A.15 and A.9 to find.
Z ∞ n 
n n n d ikx
x = i (−1) n
e δ(k)dk (C.27)
−∞ dk
Z ∞
n n n
= i (−1) (ix) δ(k)eikx dk
−∞
n
=x .

466
Appendix D

Cauchy-Schwarz inequality

For two square integrable functions f (x) and g(x) we find the following
Cauchy-Schwarz inequality
Z ∞ Z ∞ Z ∞ 2
f ∗ (x)f (x)dx g ∗ (x)g(x)dx ≥ f ∗ (x)g(x)dx . (D.1)
−∞ −∞ −∞

We can proof the Cauchy-Schwarz inequality by taking

R∞ ∗
g (x)f (x)dx
h(x) = f (x) − R−∞
∞ ∗
g(x), (D.2)
−∞ g (x)g(x)dx
R∞
g ∗ (x)f (x)dx R∞ ∗ (x)f (x)dx
Next we substitute f (x) = h(x) + R−∞
∞
g ∗ (x)g(x)dx
g(x) into −∞ f
−∞
and used that
Z ∞ Z ∞
∗
g (x)h(x)dx = h∗ (x)g(x)dx = 0, (D.3)
−∞ −∞

to get the relation

Z ∞ Z ∞
∗
f (x)f (x)dx = h∗ (x)h(x)dx (D.4)
−∞ −∞
R∞ ∗ 2Z
∞
−∞ g (x)f (x)dx
+ R∞
∗ (x)g(x)dx
g ∗ (x)g(x)dx
−∞ g −∞

467
Appendix D. Cauchy-Schwarz inequality

R∞ 2
∗
−∞ g (x)f (x)dx
Z ∞
= h∗ (x)h(x)dx + R∞
∗
−∞ −∞ g (x)g(x)dx
R∞ 2
∗
−∞ g (x)f (x)dx
≥ R∞
∗
,
−∞ g (x)g(x)dx

which proofs the Cauchy-Schwarz inequality.

468
Appendix E

Fundamental Theorem of
Algebra

469
Appendix E. Fundamental Theorem of Algebra

470
Appendix F

The eigensystem of a Matrix

471
Appendix F. The eigensystem of a Matrix

472
Appendix G

Euler angles

Intrinsic extrinsic rotations and transformation

relation between different order or rotation
relation to rotation around the axis ~r

473
Appendix G. Euler angles

474
Bibliography

[1] S. Gasiorowicz, Quantum Physics. Wiley, 1975, 1995, 2003.

[2] J. J. Sakurai, Modern Quantum Mechanics. Cambridge University
Press, 1967 - 2017.
[3] D. Griffiths, Introduction to Quantum Mechanics. Cambridge Univer-
sity Press, 2004, 2018.
[4] A. Messiah, Quantum Mechanics I+II. North-Holland Publishing Co,
1961.
[5] L. D. Landau and E. M. Lifshitz, Quantum Mechanics Non-Relativistic
Theory. Elsevier, 1958, 1965, 1977.
[6] Wikipedia, “Interpretations of quantum mechanics.” [Link]
[Link]/wiki/Interpretations_of_quantum_mechanics.
[7] E. Schrödinger, “Quantisierung als Eigenwertproblem,” Annalen der
Physik, vol. 384, p. 361, 1926.
[8] E. Schrödinger, “Quantisierung als Eigenwertproblem,” Annalen der
Physik, vol. 384, p. 489, 1926.
[9] E. Schrödinger, “Quantisierung als Eigenwertproblem,” Annalen der
Physik, vol. 385, p. 437, 1926.
[10] E. Schrödinger, “Quantisierung als Eigenwertproblem,” Annalen der
Physik, vol. 386, p. 109, 1926.

475
Bibliography

[11] E. Schrödinger, “Der stetige Übergang von der Mikro- zur

Makromechanik,” Naturwissenschaften, vol. 14, p. 664, 1926.
[12] E. Schrödinger, “Über das Verhältnis der Heisenberg-Born-
Jordanschen Quantenmechanik zu der meinem,” Annalen der
Physik, vol. 384, p. 734, 1926.
[13] W. Heisenberg, “Über quantentheoretische Umdeutung kinematischer
und mechanischer Beziehungen.,” Zeitschrift für Physik, vol. 33, p. 879,
1925.
[14] M. Born and P. Jordan, “Zur Quantenmechanik,” Zeitschrift für
Physik, vol. 34, p. 858, 1925.
[15] M. Born, “Zur Quantenmechanik II,” Zeitschrift für Physik, vol. 35,
p. 557, 1926.
[16] M. Planck, Nobel Lectures, Physics 1901-1921. Elsevier, 1967.
[17] M. Planck, The origin and development of the quantum theory. Oxford,
1922.
[18] W. Wilson, “The quantum-theory of radiation and line spectra,” Philo-
sophical Magazine, vol. 29, p. 759, 1915.
[19] G. Kirchhoff, “Ueber das Verhältniss zwischen dem Emis-
sionsvermögen und dem Absorptionsvermögen der Körper für Wärme
und Licht,” annalen der Physik, vol. 185, p. 275, 1860.
[20] J. Stefan, “Über die Beziehung zwischen der Wärmestrahlung und der
Temperatur,” Sitzungsberichte der Kaiserlichen Akademie der Wis-
senschaften: Mathematisch-Naturwissenschaftliche Classe Abteilung
II, vol. 79, p. 391, 1879.
[21] L. Boltzmann, “Ableitung des Stefan’schen Gesetzes, betreffend die
Abhängigkeit der Wärmestrahlung von der Temperatur aus der elec-
tromagnetischen Lichttheorie,” Annalen der Physik, vol. 258, p. 291,
1884.

476
 Bibliography

[22] W. Wien, “On the division of energy in the emission- spectrum of a

black body,” Philosophical Magazine. 5-th series, vol. 42, p. 214, 1897.

[23] F. Rayleigh, “LIII. Remarks upon the law of complete radiation,”

Philosophical Magazine and Journal of Science 5-th series, vol. 49,
p. 539, 1900.

[24] J. Jeans, “XI. On the partition of energy between matter and aether,”
The London, Edinburgh, and Dublin Philosophical Magazine and Jour-
nal of Science, vol. 10, p. 91, 1905.

[25] J. Jeans, “A Comparison between Two Theories of Radiation,” Nature,

vol. 72, p. 293, 1905.

[26] M. Planck, “Über eine Verbesserung der Wien’schen Spectralgle-

ichung,” Verhandlungen der Deutschen Physikalischen Gesellschaft,
vol. 2, p. 202, 1900.

[27] M. Planck, “Zur Theorie des Gesetzes der Energieverteilung im

Normalspectrum,” Verhandlungen der Deutschen Physikalischen
Gesellschaft, vol. 2, p. 237, 1900.

[28] M. Planck, “Entropie und Temperatur strahlender Wärme,” Annalen

der Physik, vol. 306, p. 719, 1900.

[29] M. Planck, “Ueber irreversible Strahlungsvorgänge,” Annalen der

Physik, vol. 306, p. 69, 1900.

[30] M. Planck, “Ueber das Gesetz der Energieverteilung im Normalspek-

trum,” Annalen der Physik, vol. 309, p. 553, 1901.

[31] A. Einstein, “Über einen die Erzeugung und Verwandlung des Lichtes
betreffenden heuristischen Gesichtspunkt,” Annalen der Physik,
vol. 322, p. 132, 1905.

[32] A. Einstein, “Zur Theorie der Lichterzeugung und Lichtabsorption,”

Annalen der Physik, vol. 325, p. 199, 1906.

477
Bibliography

[33] L. G. Newton, “The conservation of photons,” Nature, vol. 118, p. 874,

1926.

[34] H. Herz, “Ueber einen Einfluss des ultravioletten Lichtes auf die elec-
trische Entladung,” Annalen der Physik, vol. 267, p. 983, 1887.

[35] R. A. Millikan, “A Direct Determination of ”h”,” Physical Review,

vol. 4, p. 73, 1914.

[36] W. Nernst, “Ueber die Berechnung chemischer Gleichgewichte aus

thermischen Messungen,” Nachrichten von der Gesellschaft der Wis-
senschaften zu Göttingen, Mathematisch-Physikalische Klasse, vol. 1,
p. 1, 1906.

[37] A. Einstein, “Die Plancksche Theorie der Strahlung und die Theorie
der spezifischen Wärme,” Annalen der Physik, vol. 327, p. 180, 1907.

[38] P. de Bye, “Zur Theorie der spezifischen Wärmen,” Annalen der

Physik, vol. 344, p. 789, 1912.

[39] A. H. Compton, “A Quantum Theory of the Scattering of X-rays by

Light Elements,” Physical Review, vol. 21, p. 483, 1923.

[40] W. C. Röntgen, “On a new kind of Rays,” Philosophical Magazine.

5-th series, vol. 59, p. 227, 1896.

[41] W. H. Bragg and W. L. Bragg, “The reflection of X-rays by crystals,”

Proceedings of the Royal Society A, vol. 88, p. 428, 1913.

[42] D. Broglie, Recherches sur la théorie des Quanta. université en cours

d’affectation, 1924.

[43] G. P. Thomson and A. Reid, “Diffraction of Cathode Rays by a Thin

Film,” Nature, vol. 119, p. 890, 1927.

[44] A. Tonomura, J. Endo, T. Matsuda, T. Kawasaki, and H. Ezawa,

“Demonstration of single electron buildup of an interference pattern,”
American Journal of Physics, vol. 57, p. 117, 1989.

478
 Bibliography

[45] R. Bach, D. Pope, S.-H. Liou, and H. Batelaan, “Controlled double-slit

electron diffraction,” New Journal of Physics, vol. 15, p. 033018, 2013.

[46] P. Zeeman, “On the influence of magnetism on the nature of the light
emitted by a substance,” Philosophical Magazine. 5-th series, vol. 43,
p. 262, 1897.

[47] T. Preston, “Radiation Phenomena in the Magnetic Field,” Nature,

vol. 59, p. 224, 1899.

[48] O. Stern, “Ein Weg zur experimentellen Prüfung der Richtungsquan-

telung im Magnetfeld,” Zeitschrift für Physik, vol. 7, p. 249, 1921.

[49] W. Gerlach and O. Stern, “Der experimentelle Nachweis der Rich-

tungsquantelung im Magnetfeld,” Zeitschrift für Physik, vol. 9, p. 349,
1922.

[50] W. Gerlach and O. Stern, “Das magnetische Moment des Silberatoms,”

Zeitschrift für Physik, vol. 9, p. 353, 1922.

[51] Lavoisier, Elementary Treatise of Chemistry. 1789.

[52] N. P. F. [Link], “The solar spectrum.” [Link]

gov/resource/the-solar-spectrum/.

[53] G. D. Liveing and J. Dewar, “On the Spectra of Sodium and Potas-
sium,” Proceedings of the Royal Society of London, vol. 29, p. 398,
1879.

[54] J. J. Balmer, “Notiz über die Spectrallinien des Wasserstoffs,” Annalen

der Physik, vol. 261, p. 80, 1885.

[55] J. R. Rydberg, “On the structure of the line-spectra of the chemical

elements,” Philosophical Magazine. 5-th series, vol. 29, p. 331, 1890.

[56] J. J. Thomson, “Cathode Rays,” Philosophical Magazine and Journal

of Science 5-th series, vol. 44, p. 293, 1897.

479
Bibliography

[57] J. J. Thomson, “On the number of corpuscles in an atom,” Philo-

sophical Magazine and Journal of Science 6-th series, vol. 11, p. 769,
1906.

[58] E. Rutherford, “The scattering of α and β particles by matter and the

structure of the atom,” Philosophical Magazine, vol. 21, p. 669, 1911.

[59] H. Geiger and E. Marsden, “On a diffuse reflection of the α-particles,”

Proceedings of the Royal society A, vol. 82, p. 495, 1906.

[60] H. Gegier, “The scattering of α-particles by matter,” Proceedings of

the Royal society A, vol. 83, p. 42, 1910.

[61] Wikipedia, “Max von laue.” [Link]

von_Laue.

[62] V. C. Sinclair, J. M. Robertson, and A. M. Mathieson, “The crystal

and molecular structure of anthracene. II. Structure investigation by
the triple Fourier series method,” Acta Crystallographica, vol. 3, p. 257,
1950.

[63] J. Dalton, New system of chemical philosophy. S. Russel, 125, Deans-

gate, London, 1800.

[64] N. Straumannr, “Schrödingers Entdeckung der Wellenmechanik,”

2001. arXiv:quant-ph/0110097.

[65] L. de Broglie, “A tentative theory of light quanta,” Philosophical Mag-

azine and Journal of Science, vol. 47, p. 446, 1924.

[66] M. Born, W. Heisenberg, and P. Jordan, “Quantenmechanik der

Stoßvorgänge,” Zeitschrift für Physik, vol. 38, p. 803, 1926.

[67] P. W. Anderson, “More is different,” Science, vol. 177, p. 393, 1972.

[68] P. W. Anderson, “Physics: the opening to complexity,” PNAS, vol. 92,

p. 6653, 1995.

480
 Bibliography

[69] E. Noether, “Invarianten beliebiger Differentialausdrücke”,”

Nachrichten von der Gesellschaft der Wissenschaften zu Göttin-
gen, Mathematisch-Physikalische Klasse, vol. 1918, p. 235, 1918.
[70] E. Noether, “Invariante Variationsprobleme,” Nachrichten von
der Gesellschaft der Wissenschaften zu Göttingen, Mathematisch-
Physikalische Klasse, vol. 1918, p. 37, 1918.
[71] P.-O. Löwdin, “Angular Momentum Wavefunctions Constructed by
Projector Operators,” Rev. Mod. Phys., vol. 36, p. 966, 1964.
[72] J.-L. Calais, “Derivation of the Clebsch Gordan coefficients by means
of projection,” International Journal of Quantum Chemistry, vol. 2,
p. 715, 1968.
[73] W. Pauli, “Über den Einfluß der Geschwindigkeitsabhängigkeit der
Elektronenmasse auf den Zeemaneffekt,” Zeitschrift für Physik, vol. 31,
p. 373, 1925.
[74] W. Pauli, “Über den Zusammenhang des Abschlusses der Elektronen-
gruppen im Atom mit der Komplexstruktur der Spektren,” Zeitschrift
für Physik, vol. 31, p. 765, 1925.
[75] G. Uhlenbeck and S. Goudsmit, “Spinning Electrons and the Structure
of Spectra,” Nature, vol. 117, p. 264, 1926.
[76] L. H. Thomas, “The Motion of the Spinning Electron,” Nature,
vol. 117, p. 514, 1926.
[77] A. Einstein and W. de Haas, “Experimental proof of the existence of
Ampeére’s molecular currents,” KNAW Proceedings, vol. 18, p. 696,
1915.
[78] J. [Link], “The Theory of Complex Spectra,” Phys. Rev., vol. 34,
p. 1293, 1929.
[79] O. Klein, “Elektrodynamik und Wellenmechanik vom Standpunkt des
Korrespondenzprinzips,” Zeitschrift für Physik, vol. 41, p. 407, 1927.

481
Bibliography

[80] W. Gordon, “Der Comptoneffekt nach der Schrödingerschen Theorie.,”

Zeitschrift für Physik, vol. 40, p. 117, 1926.

[81] P. A. M. Dirac, “The quantum theory of the electron,” Proceedings of

the Royal Society London A, vol. 117, p. 610, 1928.

[82] W. E. Lamb and R. C. Retherford, “Fine Structure of the Hydrogen

Atom by a Microwave Method,” Phys. Rev., vol. 72, p. 241, 1947.

[83] W. Pauli, “Zur Quantenmechanik des magnetischen Elektrons,”

Zeitschrift für Physik, vol. 43, p. 601, 1927.

[84] C. G. Darwin, “The electron as a vector wave,” Proceedings of the

Royal Society London A, vol. 116, p. 227, 1927.

482