Script

QUANTUM THEORY I
2nd Edition January 2024

by Jakob Weiß, Helmut Hörner, and Tobias Schäfer

based on the lecture by Prof. Andreas Grüneis, Prof. Stefan Rotter
for the Bachelor’s program Technical Physics
at the Technische Universität Wien
 Version 2.0
This work is licensed under a Creative Commons Attribution - NonCommercial -
NoDerivatives 4.0 International License. L M N Q
 0 Preface

0 Preface
Preface to the zeroth edition This script is based on the contents of the course Quantum
Theory I, which is offered to students of the Bachelor’s program in Technical Physics at the
Vienna University of Technology. The main topics of this script can be divided into the following
areas: (i) failure of classical physics and beginnings of quantum theory, (ii) foundations and
formal structure of non-relativistic quantum theory, (iii) applications of quantum theory to
fundamental problems and questions. The last two topics are particularly extensive. The script
often uses the Dirac notation, which enables an elegant and compact notation of mathematical
expressions. We particularly focus on the quantum-theoretical description of the harmonic
oscillator and the hydrogen atom, with a closer look at the theory of angular momentum and
spin in the latter. The topic of approximation methods is only briefly touched upon, as a more
detailed treatment will be given in the continuation course Quantum Theory II in the Master’s
program.

About the design of this script In order not to impede readability, different text boxes are used
in this script. We distinguish between motivation boxes , in-depth boxes , and example boxes
of the form:

Motivation
Will delve into the motivation behind various concepts of quantum theory.

In-Depth
Will provide context for various concepts of quantum theory or offer mathematical in-depth
explanations.

Example
Allows for a more detailed investigation of specific examples.

The content of the motivation and in-depth boxes is intended as learning support and does
not have to be considered as additional test material - the example boxes are mainly used to
examine different problem situations in more detail. In order to improve readability in complex
formulas and derivations, terms that cancel each other out are colored red (e.g. b + a − a = b,
or abc
b = ac). If we use +0 or ·1 within a formula, we color these terms blue (e.g. b = a − a + b
or ab = ab cc ).

In addition, important formulas, central statements of a section or fundamental results are

explicitly highlighted - the reader is encouraged to pay special attention to formulas of the
following form:

E = mc2 (0.1)

i
 0 Preface

The script is adapted to the lecture Quantum Theory I at the Vienna University of Technology
by Prof. A. Grüneis and Prof. S. Rotter. The purpose of this is to use the script in addition to
or as a supplement to the blackboard and slides.

Literature In addition, the following books are recommended for further study:

Quantum Mechanics, Vol. 1 by C. Cohen-Tannoudji

Modern Quantum Mechanics by J.J. Sakurai

Principles of Quantum Mechanics by R. Shankar

Theoretische Physik 3 by M. Bartelmann

About the cover page The cover page shows the spherical harmonics function Y20 (ϑ, φ). We
will get to know the spherical harmonics functions as the eigenfunctions in the spherically
symmetric position space of angular momentum. In the special case of an electron in the
hydrogen atom (which we will discuss in detail), we use Y20 (ϑ, φ) to localize the electron around
the proton with a certain probability.

Preface to the first edition In this updated edition, numerous corrections have been made
based on the feedback from the students. In addition, chapters 1, 7, and 8 („Failure of Classical
Physics“, „Spin“, and „Perturbation Theory“) have been revised and expanded. Furthermore,
a new chapter 9 („Concepts of Quantum Theory“) has been added, in which advanced concepts
and technical applications of quantum theory are explained. The authors are of course still
grateful for any feedback that helps to further improve the script!

Preface to the second edition In the second edition, not only were errors corrected throughout
the script (partly based on feedback from the students, for which the authors are very grateful),
but also chapters 2 and parts of chapter 3 were extensively revised and expanded, with the aim
of presenting the material and mathematical derivations even more clearly.

Preface to the English version of the second edition This English version is a translation of
the second edition of the original script in German. It was created by Helmut Hörner with the
assistance of AI in January 2025.

Collaboration on GitHub An online version of the script is freely available at https://github.

com/Quantentheorie-1/. Similarly, the forum at https://github.com/Quantentheorie-1/
Skriptum/issues on GitHub can be used to report errors or provide suggestions for improving
the script. The authors are happy to incorporate the comments and suggestions received into
the script and will provide corresponding updated versions on GitHub.

ii
Quantum Theory I

Contents
0 Preface i

1 Failure of Classical Physics 1

1.1 Radiation of Black Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Mode Density of the Black Body . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Spectral Energy Density of the Black Body . . . . . . . . . . . . . . . . . 3
1.2 Photoelectric Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Compton Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 De-Broglie Hypothesis for Matter Waves . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Discrete Spectral Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Schrödinger Equation 13
2.1 Motivation of the Schrödinger Equation . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.1 Correspondence Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.2 Probability Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.3 Probability Current Density . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.4 Time-independent Schrödinger Equation . . . . . . . . . . . . . . . . . . . 19
2.1.5 Properties of the Schrödinger Equation . . . . . . . . . . . . . . . . . . . 21
2.2 Wave Packets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.2.1 Calculation of the Wave Packet in Position Space . . . . . . . . . . . . . . 25
2.2.2 Heisenberg’s Uncertainty Principle . . . . . . . . . . . . . . . . . . . . . . 29
2.2.3 Expectations of the Wave Packet . . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Particle in a Potential Well . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.1 Convexity Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3.2 Symmetry Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3.3 Infinitely Deep Potential Well . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3.4 Finitely Deep Potential Well . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4 Particles at the Potential Barrier . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4.1 Particles at the Potential Step . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4.2 Transfer and Scattering Matrix . . . . . . . . . . . . . . . . . . . . . . . . 41
2.4.3 Transmission and Reflection Probability . . . . . . . . . . . . . . . . . . . 44
2.4.4 Probability Current Density in Scattering . . . . . . . . . . . . . . . . . . 45
2.4.5 Tunneling Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3 Formal Structure of Quantum Theory 51

3.1 Dirac Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1.1 Hilbert Space and Ket Vectors . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1.2 Dual Space and Bra Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.3 The Scalar Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.4 Complete Orthonormal System and Dimension of the Hilbert Space . . . 56
3.1.5 Operators in the Braket Notation . . . . . . . . . . . . . . . . . . . . . . . 57
3.2 Hermitian Operators, Eigenfunctions, and Eigenvalues . . . . . . . . . . . . . . . 58
3.2.1 Real Eigenvalues of Hermitian Operators . . . . . . . . . . . . . . . . . . 58
3.2.2 Orthogonality of Eigenstates of Hermitian Operators . . . . . . . . . . . . 59
3.3 Matrix Representation of an Operator . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3.1 Projections and Spectral Representation . . . . . . . . . . . . . . . . . . . 61
3.4 Distribution Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.4.1 Position and Momentum Eigenvalues and Eigenfunctions . . . . . . . . . 63
3.4.2 From the Abstract Ket Vector to the Concrete Wave Function . . . . . . 64
3.4.3 Position and Momentum Eigenfunctions in Position and Momentum Basis 65

iii
 Contents

3.4.4 Continuous Spectrum and Fourier Transformation . . . . . . . . . . . . . 68

3.4.5 General Spectral Representation . . . . . . . . . . . . . . . . . . . . . . . 70
3.4.6 Direct Sum and Tensor Product of Vector Spaces . . . . . . . . . . . . . . 70
3.5 Operator Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.5.1 Commutator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.5.2 Complementarity, Canonical Commutation Relation, and Uncertainty Prin-
ciple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.5.3 Compatibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.6 Measurements in Quantum Theory and Collapse of the Wave Function . . . . . . 80
3.6.1 Measurement on a Non-Degenerate System . . . . . . . . . . . . . . . . . 82
3.6.2 Measurement on a Degenerate System . . . . . . . . . . . . . . . . . . . . 84
3.6.3 Measurement on a Continuous System . . . . . . . . . . . . . . . . . . . . 84
3.6.4 Collapse of the Wave Function and Compatible Operators . . . . . . . . . 85

4 Harmonic Oscillator 87
4.1 Analytical Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.1.1 The classical harmonic oscillator . . . . . . . . . . . . . . . . . . . . . . . 87
4.1.2 Solution of the Schrödinger equation with the potential of the harmonic
oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.1.3 Limits of the Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . 94
4.2 Algebraic Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2.1 Ladder Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2.2 Commutators of the Ladder Operators and Number State Operator . . . 97
4.2.3 Action of Raising and Lowering Operators . . . . . . . . . . . . . . . . . . 99
4.2.4 Heisenberg Uncertainty Relation . . . . . . . . . . . . . . . . . . . . . . . 102
4.3 Time-Evolved Oscillator States . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.3.1 Time-Dependent Oscillator Wave Functions . . . . . . . . . . . . . . . . . 104
4.3.2 Coherent Glauber States . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5 Angular Momentum 111

5.1 Angular Momentum Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.1.1 Commutator Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.1.2 Polar and Axial Vector Operators . . . . . . . . . . . . . . . . . . . . . . 116
5.1.3 Commutator Relations with Scalar Operators . . . . . . . . . . . . . . . . 117
5.1.4 Ladder Operators of the Angular Momentum . . . . . . . . . . . . . . . . 118
5.2 Eigen System of Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.2.1 Construction of an Angular Momentum Multiplet . . . . . . . . . . . . . 120
5.2.2 Angular Momentum Uncertainty . . . . . . . . . . . . . . . . . . . . . . . 124
5.3 Angular Momentum in Position Space . . . . . . . . . . . . . . . . . . . . . . . . 127
5.3.1 Angular Momentum in Spherical Coordinates . . . . . . . . . . . . . . . . 127
5.3.2 Legendre Polynomials and Spherical Harmonics . . . . . . . . . . . . . . . 134
5.3.3 Symmetry Properties of the Eigenfunctions . . . . . . . . . . . . . . . . . 137
5.3.4 Representations of the Spherical Harmonics . . . . . . . . . . . . . . . . . 139

6 Hydrogen Atom 141

6.1 Schrödinger Equation as a Two-Body Problem . . . . . . . . . . . . . . . . . . . 141
6.1.1 Transformation to the Center of Mass System . . . . . . . . . . . . . . . . 141
6.1.2 Separation Ansatz in Spherical Coordinates . . . . . . . . . . . . . . . . . 145
6.1.3 Effective Potential and Form of the Wave Function . . . . . . . . . . . . . 146
6.2 Solution for Radial Wave Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.2.1 Solution Using Frobenius Method . . . . . . . . . . . . . . . . . . . . . . . 149
6.2.2 Solution via Laguerre Differential Equation . . . . . . . . . . . . . . . . . 154

iv
Quantum Theory I

6.3 Energy Levels and Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

7 Spin 161
7.1 Magnetic Moment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
7.1.1 Energy in the Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . 163
7.2 Stern-Gerlach Device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
7.3 Postulate of Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7.4 Spin Eigen System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7.4.1 Spin Operators and Commutators . . . . . . . . . . . . . . . . . . . . . . 166
7.4.2 Matrix Representation and Pauli Matrices . . . . . . . . . . . . . . . . . . 167
7.4.3 Bloch Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
7.5 Total Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
7.5.1 Product Basis and Coupled Basis . . . . . . . . . . . . . . . . . . . . . . . 174
7.5.2 States in the Product Basis and Coupled Basis . . . . . . . . . . . . . . . 177
7.5.3 Clebsch-Gordan Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . 179

8 Perturbation Theory 187

8.1 Zero-Order Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
8.2 First-Order Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
8.3 Second-Order Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

9 Concepts of Quantum Theory 193

9.1 Series of Stern-Gerlach Apparatuses . . . . . . . . . . . . . . . . . . . . . . . . . 193
9.2 Axioms of Quantum Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
9.3 The Measurement Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
9.3.1 Separable and Entangled States . . . . . . . . . . . . . . . . . . . . . . . . 196
9.3.2 Schrödinger’s Cat and the Limit of Quantum Mechanics . . . . . . . . . . 199
9.4 Bell’s Inequalities and Hidden Variables . . . . . . . . . . . . . . . . . . . . . . . 201
9.4.1 Hidden Variables and Spooky Action at a Distance . . . . . . . . . . . . . 202
9.4.2 Bell’s Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
9.5 Technological Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
9.5.1 Random Number Generators . . . . . . . . . . . . . . . . . . . . . . . . . 209
9.5.2 Quantum Computers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
9.5.3 “No-cloning” Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
9.5.4 Quantum Cryptography . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

10 Appendix 215
10.1 Transition from Classical to Quantum Mechanics . . . . . . . . . . . . . . . . . . 215
10.2 Hermite Differential Equation and Polynomials . . . . . . . . . . . . . . . . . . . 217
10.2.1 Derivation of the Rodrigues Formula for Hermite Polynomials . . . . . . . 217
10.2.2 Recursive Representation of Hermite’s Differential Equation . . . . . . . . 218
10.2.3 Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.2.4 Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.2.5 Normalizability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.3 Legendre Differential Equation and Polynomials . . . . . . . . . . . . . . . . . . . 219
10.3.1 Legendre Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
10.3.2 Associated Legendre Polynomials . . . . . . . . . . . . . . . . . . . . . . . 220
10.3.3 Orthogonality and Normalization . . . . . . . . . . . . . . . . . . . . . . . 222
10.4 Laguerre Differential Equation and Polynomials . . . . . . . . . . . . . . . . . . . 224
10.4.1 Rodrigues Formula of the Laguerre Differential Equation . . . . . . . . . 224
10.4.2 Orthogonality and Normalizability . . . . . . . . . . . . . . . . . . . . . . 226

v
 Contents

11 Useful Relations 229

vi
Quantum Theory I

1 Failure of Classical Physics

At the beginning of the 20th century, physics seemed almost complete. The physical laws known
at that time could explain nearly all observations, and there were only a few discrepancies
between experiment and theory in some isolated areas. This is why Max Planck was even
advised against studying theoretical physics by Philipp von Jolly in 1874, with the following
justification:
“[Theoretical physics is] a highly developed, almost fully matured science that has
now, after the discovery of energy, been crowned in a sense, and would soon probably
have assumed its final stable form. There might perhaps be a speck of dust or a
bubble to examine and to categorize in one corner or another, but the system as
a whole stood fairly secure, and theoretical physics was approaching that degree of
completion which geometry, for example, had possessed for centuries.”
It was assumed that these remaining contradictions would soon resolve with the existing theories
of what was then called classical physics. Instead, in attempting to clarify these apparent trifles,
a revolution in physics was set in motion: quantum theory was born.

1.1 Radiation of Black Bodies

Motivation: The Problem with Blackbody Radiation
Every real body emits electromagnetic (EM) radiation depending on its temperature T .
The radiation spectrum depends both on the temperature of the body and on the re-
flectance of the surface. A so-called Black Body is an idealized real body: it is assumed
that all incident EM radiation is completely absorbed. Therefore, in a Black Body, the
spectrum of the emitted electromagnetic radiation depends only on the temperature T .

Although no real Black Body exists, it can be experimentally mimicked by a cavity with
opaque, sufficiently absorbing inner walls. If this cavity has a small hole, radiation en-
tering through the hole undergoes so many reflections inside the cavity at the absorbing
inner walls that it can practically no longer escape from the hole. By heating the walls of
the cavity to a temperature T , the hole behaves like a Black Body with a surface equiva-
lent to the cavity. Thus, Blackbody radiation is also synonymously called Cavity Radiation.

The problem now is: when attempting to calculate the spectrum of cavity radiation using
the laws of classical physics, the Rayleigh-Jeans radiation law (1.12) is derived. While
it agrees well with measurements at low frequencies, it does not at high frequencies. It
suggests that the energy density increases quadratically with frequency. A Black Body
would, therefore, have to emit an infinite amount of energy! This problem is known as the
“ultraviolet catastrophe”. Another radiation law from classical physics is Wien’s Radiation
Law (1.20), which was justified using methods from statistical physics. While it can ade-
quately represent the spectrum at high frequencies, it completely fails at low frequencies.

Max Planck tried to unify these radiation laws. However, he only succeeded when, in
December 1900, in an „act of desperation“, he hypothesized that light is emitted only in
discrete energy packets, so-called “quanta” E = hν = ℏω. This was the birth of quantum
physics. The natural constant h = 6.62607015 · 10−34 Js is today known as the “Planck
constant”, while we refer to ℏ = h/2π as the “Reduced Planck Constant”.

The spectral energy density of a Black Body can be defined as follows:

ε(ω)dω = ⟨E(ω)⟩ n(ω)dω (1.1)

1
 1 Failure of Classical Physics

Here, ⟨E(ω)⟩ is the average energy of the vibrational modes of the radiation field in the Black
Body, and n(ω) is the mode density as a function of the angular frequency ω.

1.1.1 Mode Density of the Black Body

In a closed cavity, only certain vibrational modes can form. The mode density n(ω) describes
the absolute number of modes N (ω) per volume V in an infinitesimal frequency interval dω:
Number of modes 1 dN (ω)
n(ω) = = (1.2)
Volume · Frequency interval V dω
In the following, the mode density will be derived: We consider a cubic volume V = L3 with
edge length L, where a standing electromagnetic field E(r) forms. The surface of the cube is
assumed to be perfectly conducting; that is, the electromagnetic field has no component parallel
to the surface. Solving Maxwell’s equations inside the volume yields a possible field of the form:
Ex (r) Ex0 cos(kx x) sin(ky y) sin(kz z)
   

E(r) = Ey (r) = Ey0 sin(kx x) cos(ky y) sin(kz z) (1.3)

   
Ez (r) Ez0 sin(kx x) sin(ky y) cos(kz z)
Consider the case where x = L. According to our boundary conditions, the components of
the field E(x = L, y, z) parallel to the cube surface must vanish, which means Ey = Ez = 0
must hold. Consequently, the terms Ey0 sin (kx x) and Ez0 sin (kx x) must vanish, which only
holds if sin (kx x) vanishes with kx = nx πx/L with ni ∈ N+ . This leads to a quantization of
the corresponding wave number kx = nx π/L. The same applies analogously in the y and z
directions, so we can generally write:
π π π
kx = nx , ky = ny , kz = nz with nx , ny , nz ∈ N+ (1.4)
L L L
Thus, only those waves E(r) whose wave vector k is defined through the integer coefficients nx ,
ny , and nz can be realized in the cubic box:
   
kx nx
  (1.4) π  
k =  ky  = ny  with nx , ny , nz ∈ N
+
(1.5)
L
kz nz
The magnitude of this wave vector can therefore also be expressed through the coefficients nx ,
ny , and nz :
(1.5) π
q
|k| = n2x + n2y + n2z
L
Each possible k-vector („state“) can thus be assigned a point in a three-dimensional space with
the axes kx , ky , and kz (the so-called k-space)–two adjacent values of kx always have the constant
distance ∆kx = π/L due to (1.4). The same holds for ∆ky and ∆kz :
π
∆kx = ∆ky = ∆kz = (1.6)
L
We now want to know how many possible k-vectors there are whose magnitude |k| is smaller
than a given wave number k0 . To do this, we first assign each state k(nx , ny , nz ), which we
have so far represented only as a point in k-space, now a cubic volume Vk . Since all points have
constant distances, each point has the same associated volume:
 3
(1.6) π
Vk = ∆kx ∆ky ∆kz = (1.7)
L
To determine how many states N (k0 ) there are that satisfy the condition |k| < k0 , we check
how often the volume Vk fits into the volume of a sphere with radius k0 .
Note that:

2
Quantum Theory I

ky

kx

Abb. 1: (left) Radiation forms a standing wave field in the Black Body, in which only the
existence of certain wave vectors is possible. (right) Possible states in wave vector
space are represented by points. Therefore, only a finite number of wavelengths are
allowed in a sphere with a finite radius k.

kx , ky , and kz can only take positive values. Thus, not the whole sphere volume VSphere =
3 πk0 , but the positive octant with Voct = 8 3 πk0 is considered.
4 3 14 3

due to the two possible polarization directions of the electromagnetic field E(r), two states
can be assigned to all possible kx , ky , and kz . This results in an additional factor of 2.

We thus find for the number of realizable states with the boundary condition |k| < k0 :
1 4π 3  3
Voct k π L
N (k0 ) = 2 = 2 8 3 03 = k03
Vk (π/L) 3 π

Since we want to express the number of modes as a function of angular frequency ω = k0 c, not
wave number k0 , we substitute k0 = ω/c in the above expression, yielding:

π ω 3 L3 L3 ω 3 V ω3
N (ω) = = = (1.8)
3 c3 π 3 3π 2 c3 3π 2 c3
From (1.8), the mode density can now be derived using (1.2). Thus we get:

(1.2) 1 d (1.8) ω
2
n(ω) = N (ω) = 2 3 (1.9)
V dω π c

1.1.2 Spectral Energy Density of the Black Body

Deeper Understanding: Classical Energy Density of the Black Body

The average frequency-dependent energy ⟨E⟩ according to the laws of classical physics can
be calculated using occupancy probability. We use the Boltzmann distribution P (E) =
e−βE with the Boltzmann factor β −1 = kB T . If the energy can be varied continuously,
the averaging is done by integration. We calculate ⟨E⟩ as the first moment of E using the
following relation: R∞ R∞
0R dE E · P (E)
−βE
0R dE E · e
⟨E⟩ = ∞ = ∞ (1.10)
0 dE P (E)
−βE
0 dE e
The integral in the denominator of the fraction is the normalization and can be solved by

3
 1 Failure of Classical Physics

trivial integration:
1 h −βE i∞ 1 1
Z ∞
N= dE e−βE = − e = − (0 − 1) =
0 β E=0 β β
For evaluating the integral in the numerator, one can either integrate by parts, or use
Feynman’s trick, where clever differentiation can simplify integrations:
∂ 1 1
Z ∞ Z ∞ Z ∞
−βE ∂ −βE ∂
E= dE E · e =− dE e =− dE e−βE = − = 2
0 0 ∂β ∂β 0 ∂β β β

We obtain for the expected value of the energy ⟨E⟩ in the classical case:

E 1 1
⟨E⟩ = = β 2 = = kB T (1.11)
N β β

Note that according to the equipartition theorem, an average energy of kB T /2 is Rassigned to

each degree of freedom that appears quadratically in the Hamiltonian H0 = ε22 d3 r(E2 +
c2 B2 ). In using the Boltzmann distribution, it is already taken into account here that each
mode with energy E contains both potential energy (quadratic in position) and kinetic
energy (quadratic in momentum). The spectral energy density in the classical case is thus
given as:
2
(1.1) (1.11) (1.9) kB T ω
ε(ω) = ⟨E(ω)⟩ n(ω) = kB T n(ω) = (1.12)
π 2 c3
This corresponds to the Rayleigh-Jeans radiation law. It asserts that the energy density
ε increases quadratically with the frequency ω. A Black Body would, therefore, have to
emit an infinite amount of energy! This is referred to as the ultraviolet catastrophe.

Max Planck made the momentous assumption that we are dealing with discrete energy values
in electromagnetic radiation, not continuous ones:
En = ℏωn (1.13)
Therefore, in forming the expectation value ⟨E⟩ the integral must be replaced by a sum. ⟨E⟩ is
thus given by: P∞
En · P (En ) ℏω ∞ n=0 n · e
−βℏωn
P
⟨E⟩ = n=0 P∞ = P∞ (1.14)
n=0 P (En )
−βℏωn
n=0 e
The sum in the denominator of the fraction is again a normalization. It can be determined with
the geometric series formula since the exponential function is less than one in all cases. We
substitute x = e−βℏω in the following:
∞ ∞  ∞
n
subst 1
Ndis = e−βℏωn = e−βℏω = xn = (1.15)
X X X

n=0 n=0 n=0

1−x

For the numerator in (1.14), a similar trick can be used as in the evaluation of the average
energy E in the case of continuous energies:
∞ ∞ ∞ ∞ ∞
 n
subst d n
E dis = ne−βℏωn = n e−βℏω = nxn = xnxn−1 = x =
X X X X X
x
n=0 n=0 n=0 n=0 n=0
dx
∞
d X (1.15) d 1 x
=x xn = x = x (1 − x)−2 = (1.16)
dx n=0 dx 1 − x (1 − x)2

Thus, the averaging for discrete energy values results finally in:
E dis x/(1 − x)2 x(1 − x) ℏω subst ℏω
⟨E⟩ = ℏω = ℏω = ℏω = = βℏω (1.17)
Ndis 1/(1 − x) (1 − x)2 1/x − 1 e −1

4
Quantum Theory I

Therefore, the spectral energy density ε(ω) ultimately follows:

(1.1) (1.17) ℏω (1.9) ℏω ω2
ε(ω) = ⟨E(ω)⟩ n(ω) = n(ω) =
eβℏω − 1 eβℏω − 1 π 2 c3
This result is none other than the famous Planck’s Radiation Law:

ℏω 3 1
ε(ω) = (1.18)
2 3
π c e βℏω −1

Rayleigh-Jeans Radiation Law

1.25T

Planck’s Radiation Law

1.1̇T
ε(ω)

ε(ω)
T

Wien’s Radiation Law

0 0
ω ω

Abb. 2: (left) Planck’s radiation law as a function of angular frequency ω at different temper-
atures. (right) Planck’s radiation law with approximations at low and high energies.

Approximations for the limiting cases ω → 0 and ω → ∞ can now be investigated. Considering
only low frequencies ω → 0 leads to the classical Rayleigh-Jeans Radiation Law (1.12) using the
approximation eβℏω ≈ 1 + βℏω:
1 1
" #
ℏω 3
(1.18) ℏω 3 ω2 kB T
lim ε(ω) = lim = ≈ = 2 3 ω2 (1.19)
ω→0 ω→0 π 2 c3 eβℏω − 1 π 2 c3 1 + βℏω + O(ω 2 ) − 1 βπ 2 c3 π c

Forming the limit for high frequencies yields eβℏω − 1 ≈ eβℏω , resulting in the Wien’s Radiation
Law:
1 ℏω 3 1
" #
(1.18) ℏω 3 ℏω 3 −βℏω
lim ε(ω) = lim ≈ = e (1.20)
ω→∞ ω→∞ π 2 c3 eβℏω − 1 π 2 c3 eβℏω π 2 c3
The derivative with respect to ω from Wien’s Radiation Law (1.20) also follows Wien’s Dis-
placement Law. It states that the frequency ωmax , at which a Black Body with temperature T
emits its most intense radiation, is directly proportional to the temperature T :
d  3 −βℏω 
0= ω e = 3ω 2 e−βℏω − ℏω 3 βe−βℏω = ω 2 (3 − ℏωβ) e−βℏω
dω
The value ω = ωmax , at which the last part of the expression vanishes, corresponds to the
intensity maximum. Both ℏ and kB are constant quantities, whereby Wien’s Displacement Law
can be simplified to the following equation:
ωmax
= const. (1.21)
T
We have thus managed to show that Wien’s Displacement Law, named after Wilhelm Wien,
directly follows from Planck’s Radiation Law.

5
 1 Failure of Classical Physics

1.2 Photoelectric Effect

Motivation: The Photoelectric Effect
When illuminating a metal plate, which is negatively charged with respect to its surround-
ings, with light of sufficiently high frequency (e.g., UV light), electrons are released from
the metal plate. According to the laws of classical physics, the energy of a light wave is
given by its intensity. Therefore, it would be expected that the kinetic energy of the elec-
trons released from the plate (and therefore, also their speed in the non-relativistic limit)
would increase proportionally with the light intensity.

However, as physicist Philipp Lenard discovered in 1902, this is not the case. The speed
of the outgoing electrons does not depend on the intensity, but on the frequency of the
incoming light! When monochromatic light is shone on the metal plate, all outgoing elec-
trons have the same speed. While the number of electrons increases when the intensity is
increased, the speed of each electron remains unchanged.

In 1905, Albert Einstein succeeded in explaining this effect using the light quan-
tum hypothesis. The incoming light is imagined as being composed of individual „light
quanta“(photons). For monochromatic light, each photon has the same energy Eγ = ℏω
and can only transfer this energy to the knocked out electron. Thus, all electrons have
the same speed. An increased intensity of the monochromatic light beam results in more
photons hitting the metal plate per unit time, hence more electrons are emitted, but each
electron still has the same speed.

In 1921, Albert Einstein was awarded the Nobel Prize for this light quantum hypothesis,
although against the opposition of numerous colleagues. Ironically, the antisemitic physicist
Philipp Lenard, whose experiments initially prompted the light quantum hypothesis, was
among those who protested against awarding the Nobel Prize to Einstein.

The energy density ωEM of electromagnetic radiation is given in the classical electromagnetic
theory by the electric field strength E, the electric flux density D, as well as the magnetic field
strength H and the magnetic flux density B. With the electric and the magnetic constants
ε0 and µ0 , we can express the flux densities in terms of their respective field strengths. Thus,
we find a relationship for ωEM , which depends on the square of the field strengths, i.e., the
intensities:
1 1
ωEM = (E · D + H · B) = D = ε0 E, H = B
2 µ0
1 ε0 2
 
= ε0 E 2 + B =
2 ε0 µ 0
ε0 1 1
 
= E +
2
B =
2
= c2
2 ε0 µ 0 ε0 µ 0
ε0
= (E2 + c2 B2 ) (1.22)
2
The electromagnetic radiation is attributed a particle character according to Einstein’s quantum
mechanical interpretation, and the „elementary charges“of the field are referred to as photons.
According to the heuristic assumption of Planck in deriving his radiation law, these carry the
following energy Eγ :

Eγ = ℏω (1.23)

The energy of light radiation (or electromagnetic radiation in general) can take only integer
multiples of the energy of a photon. However, the energy of a photon is determined exclusively

6
Quantum Theory I

by the frequency of the photon and is completely independent of the light beam’s intensity. The
number of photons thus determines the amplitude of the light and is therefore responsible for
the intensity of the radiation.
But what momentum does a single photon have? According to the special theory of relativity,
for a moving particle with a rest mass m0 , it holds that:

p2 c2 = E 2 − m20 c4 (1.24)

A photon has a vanishing rest mass m0 = mγ = 0; therefore, the relativistic energy relationship
(1.24) simplifies to E 2 = p2γ c2 . Consequently, the momentum of a photon can be calculated by
taking the square root, resulting in:
Eγ (1.23) ℏω h
pγ = = = ℏ|k| = (1.25)
c c λ
For the vectorial momentum pγ of a single photon, it follows:

pγ = ℏk (1.26)

When a photon collides with a bound electron, it can release the electron from the potential if
the energy Eγ is sufficient.

red light green light blue light

Abb. 3: Photons hit a surface and depending on the frequency, electrons are ejected from it.
The Ekin of the electrons is directly dependent on the light frequency.

The photon energy Eγ = ℏω must exceed the work function W (which is the binding energy
of the electron), and the remaining energy is converted into kinetic energy Ekin of the now free
electron:
Ekin = Eγ − W = ℏω − W (1.27)
If the intensity of the light and thus the number of photons is increased, more electrons are
indeed emitted, but the kinetic energy (and thus the speed) of the electrons depends solely on
the frequency of the light! If Eγ = ℏω < W , no electrons are emitted from the metal plate, even
if exposure duration and intensity are increased.

1.3 Compton Effect

Motivation: The Compton Effect

The American physicist Arthur Holly Compton discovered the following phenomenon
in 1922: When a material is irradiated with X-rays, the scattered radiation experiences a
frequency shift to smaller frequencies (i.e., larger wavelengths). This shift is greater the
larger the scattering angle and is almost independent of the material.

Compton could explain this effect with the photon model as an elastic collision between a
photon with energy Eγ = ℏω and momentum k = ℏk and a weakly bound electron of the
scattering material. For his discovery, today called the Compton Effect, he was awarded

7
 1 Failure of Classical Physics

the Nobel Prize in 1927.

The Compton Effect is the elastic scattering of photons by (almost) free electrons, which leads
to the redshift (∆λ = λ′ − λ > 0) of the scattered light. The photon is characterized by initial
and final wavelengths (λ and λ′ ), as well as by an initial and final momentum (p and p′ ). The
electron with mass me is assumed to be at rest before the interaction (Ee = me c2 and pe = 0).
The scattering geometry is shown in Figure 4.

λ ′ , p′

ϑ
λ, p
φ

Ee′ , p′e

Abb. 4: The Compton Effect as the scattering of a photon by an electron.

The description follows the special theory of relativity. Therefore, the total energy of the electron
before the scattering process Ee and after the scattering process Ee′ is given by:
q
Ee = me c2 and Ee′ = e c + me c
p′2 2 2 4 (1.28)

During the scattering process, both energy and momentum are conserved. Energy conservation
can be expressed as:
Eγ + Ee = Eγ′ + Ee′ (1.29)
With (1.23), we can express the photon energies in terms of their respective frequencies ω and
ω ′ , while for the electrons, the quantities from (1.28) are used. Simple rearrangement of the
energy conservation (1.29) leads us to:
q
ℏ ω − ω ′ + me c2 = e c + me c
p′2 (1.30)


2 2 4

To eliminate the square root expression, both sides are squared. Using the relation ω = c|k| = ck,
(1.30) can now be specified as follows:

[cℏ k − k ′ + me c2 ]2 = p′2
e c + me c =⇒
2 2 4

ℏ2 (k − k ′ )2 + m2e c2 + 2ℏme c(k − k ′ ) = p′2

e + me c
2 2
(1.31)

Momentum is also conserved, which we can present in a vectorial relation:

pγ + pe = p′γ + p′e (1.32)

The electron is at rest before scattering, so pe = 0 holds. Substituting this into the momentum
conservation equation (1.32) simplifies to pγ = p′γ + p′e , where the photon’s momentum follows
from (1.26). Solving for the momentum of the scattered electron p′e , we get the relationship
p′e = pγ − p′γ = ℏ (k − k′ ). Thus, for the magnitude of momentum p′2 e = |pe | :
′ 2

 
p′2 ′ ′ 2 ′ ′ ′
e = pe · pe = ℏ (k · k + k · k − 2k · k ) = ℏ
2
k 2 + k ′2 − 2kk ′ cos(ϑ) (1.33)

8
Quantum Theory I

The angle ϑ is the scattering angle of the photon according to Figure 4. Expression (1.33) for
the scalar p′2
e can now be substituted into the energy conservation equation (1.31):

ℏ2 (k − k ′ )2 + m2e c2 + 2ℏme c(k − k ′ ) = ℏ2 (k 2 + k ′2 − 2kk ′ cos(ϑ)) + m2e c2 =⇒

ℏ(k 2 + k ′2 − 2kk ′ ) + 2me c(k − k ′ ) = ℏ(k 2 + k ′2 − 2kk ′ cos(ϑ)) =⇒
−2ℏkk ′ + 2me c(k − k ′ ) = −2 ℏkk ′ cos(ϑ) =⇒
me c(k − k ′ ) = ℏkk ′ − ℏkk ′ cos(ϑ)) =⇒
me c(k − k ′ ) = ℏkk ′ (1 − cos(ϑ))

A trigonometric relationship for the cosine term is 1−cos(ϑ) = 2 sin2 (ϑ/2); additionally applying
the relation k = ω/c = 2πν/c = 2π/λ, the wavelength change ∆λ = λ′ − λ can explicitly be
determined:
ϑ me c k − k ′ me c 1 1 me c ′ me c
   
1 − cos(ϑ) = 2 sin 2
= = − = (λ − λ) = ∆λ
2 ℏ kk ′ ℏ k ′ k ℏ2π h
Overall, when photons scatter off electrons, their wavelengths change according to the Compton
effect, depending on the scattering angle ϑ:
2h ϑ ϑ
   
∆λ = sin2 = 2λc sin 2
me c 2 2
We call λc = 2.4262 × 10−11 m the Compton wavelength of the electron; in other words, the
wavelength of a photon with an energy equivalent to the rest energy of an electron. Overall, we
find the wavelength shift of a scattered photon in the context of the Compton Effect as:

ϑ
 
∆λ = 2λc sin2 (1.34)
2

It must hold that ∆λ > 0: If ∆λ < 0 were the case, the wavelength would become shorter after
scattering, corresponding to an increase in energy and thus violating energy conservation. The
energy loss of the photon after scattering is transferred to the electron’s kinetic energy.

According to (1.34), the frequency shift and thus also the energy transfer to the electron is
greatest when the photon is reflected by the electron, resulting in backscattering (ϑ = π), and
smallest when the photon does not change angles, experiencing forward scattering (ϑ = 0).

1.4 De-Broglie Hypothesis for Matter Waves

Motivation: Particle and Wave Nature

After it was shown that light has both wave and particle characteristics, the French physi-
cist Louis-Victor de Broglie postulated a – somewhat complementary – hypothesis in
1924 in his dissertation: He postulated that every (moving) matter particle also exhibits
wave properties, and that the wave-particle duality is a universal phenomenon.

This was first confirmed in 1927 in experiments on the diffraction of electrons on thin
metal foils by Clinton Davisson and Lester Germer. Distinct interference patterns
were observable in these experiments, which can only be explained by the wave nature of
electrons. De Broglie received the Nobel Prize in 1929 for his theory of matter waves.

According topthe special theory of relativity, each moving particle with rest mass m is assigned
energy E = p2 c2 + m2 c4 . De Broglie postulates that this energy – just like for photons –
occurs only in „energy packets“:

9
 1 Failure of Classical Physics

E = ℏω (1.35)

This allows us to write the angular frequency ω as:

q
1q 2 2
ℏω = p2 c2 + m2 c4 =⇒ ω = p c + m2 c4 (1.36)
ℏ
De Broglie assumes that a moving, massive particle with velocity v corresponds to a wave
packet with group velocity vG , which is defined generally as:
∂ω
v ≡ vG = (1.37)
∂k
Substituting the expression for ω from (1.36) allows for the following calculation:
1 ∂ q 2 2 1 ∂p ∂ q 2 2 1 1 ∂p 1
v= p c + m2 c4 = p c + m2 c4 = 2pc2 (1.38)
ℏ 2 ∂k p c + m2 c4
p
ℏ ∂k ℏ ∂k ∂p 2 2

In the non-relativistic regime, where m2 c4 ≫ p2 c2 , it simplifies to p2 c2 + m2 c4 ≈ mc2 . Hence,

p

expression (1.38) simplifies as follows:

1 ∂p pc2 1 ∂p mv ∂p
v≈ 2
= =⇒ ℏv = v =⇒ ∂p = ℏ ∂k (1.39)
ℏ ∂k mc ℏ ∂k m ∂k
If both sides of (1.39) are integrated, a relation for the momentum p is obtained (in the non-
relativistic approximation):
Z Z
dp = ℏ dk + p0 =⇒ p = ℏk + p0

p0 is the integration constant in this context. Since k = 0 implies p = 0, it can be deduced that
p0 = 0 – we finally obtain the following relationship between the momentum p and the wave
number k of a moving, massive particle:

p = ℏk (1.40)

Thus, the de-Broglie wavelength λ can be assigned to a particle with momentum p = mv (with
p ≡ |p| and v ≡ |v|):
2π 2πℏ 2πℏ
λ= = = (1.41)
k p mv
Utilizing the classic Newtonian relationships E = mv 2 /2 and p = mv, it is also easy to establish
a relationship between angular frequency ω and wave number k using the de-Broglie equations:
mv 2 m2 v 2 p2 (1.40) ℏ2 k 2 !
E= = = = = ℏω
2 2m 2m 2m
Dividing both sides of the above equation by ℏ produces the dispersion relation ω(k) for free
matter waves:
ℏk 2
ω(k) = (1.42)
2m
From (1.42), the phase velocity vPh and the group velocity vG (for the non-relativistic limit
v ≪ c) can be derived:
ω(k) (1.42) ℏk ∂ (1.42) ℏk
vPh = = and vG = ω(k) = (1.43)
k 2m ∂k m
Thus, a free matter wave exhibits a quadratic dispersion relation (E ∝ k 2 ), as opposed to the
linear dispersion relation of photons (Eγ ∝ k).

10
Quantum Theory I

1.5 Discrete Spectral Lines

Motivation: Spectral Lines of the Hydrogen Atom

When a hydrogen atom is excited, it emits light not in a continuous spectrum, but in a
line spectrum, where only particular frequencies appear. Conversely, during absorption,
discrete absorption lines are also observed at the same positions in the spectrum. This
phenomenon cannot be explained by classical physics. Nonetheless, in 1888, the Swedish
physicist Johannes Rydberg could empirically establish the following formula, today
known as the Rydberg Formula, which describes the hydrogen line spectrum very accu-
rately:
1 1 1
 
= RH −
λ n21 n22
The spectroscopic quantity RH = 1.096 775 83 × 107 m−1 is referred to as the Rydberg
Constant for hydrogen, while the parameters n1 and n2 are whole numbers greater than
zero, with n2 > n1 . Certain series of spectral lines, each distinguished by its n1 , could be
found experimentally: n1 = 1 (Lyman series), n1 = 2 (Balmer series), n1 = 3 (Paschen
series), etc.

Only with the help of quantum physics could the Rydberg Formula be justified and derived.
Today, it is understood that the parameters n1 and n2 represent possible values of the principal
quantum number n, describing the quantized energy levels of the electron in the hydrogen atom’s
shell. The spectral lines arise from the discrete “allowed” transitions between different energy
levels. In quantum physics, the Rydberg Formula can thus be expressed as:

1 ∆E Ry 1 1
 
= = 2 − 2 (1.44)
λ hc hc n1 n2

Ry = 13.605 693 eV is the Rydberg constant in energy units. The precise derivation of the energy
levels of the hydrogen atom (and thus the Rydberg Formula) is provided in a later chapter of
this text.

Figure 5 shows both the emission spectrum of helium He and the absorption spectrum of H. The
emission spectrum occurs when energy is supplied to a helium gas, causing excitation–successive
de-excitation leads to the emission of photons at certain frequencies according to (1.44), which
manifests as discrete spectral lines. This spectral “fingerprint” enables the identification of
elements within a gas. If white light is shone through a gas, excitation of the gas atoms occurs
at particular wavelengths; photons of that energy are absorbed, hence do not appear in the
initial beam after passing through the gas.

394 416 427 441 469 502 589 668 710

Helium
Hydrogen

410 434 486 656

Abb. 5: (top) The emission spectrum of helium; the spectral lines are indicated as wavelengths
in nm. (bottom) The absorption spectrum of hydrogen.

11
Quantum Theory I

2 Schrödinger Equation

Significance of the Schrödinger Equation

If particles and matter also have wave properties according to the hypothesis of de
Broglie, the question arises, which wave equation describes these matter waves. The
answer was given by Erwin Schrödinger when he formulated in 1926 the today named
after him Schrödinger equation (2.8), which represents the fundamental equation of non-
relativistic quantum mechanics. The Schrödinger equation proved to be so successful,
among other things in the description of the hydrogen atom, that Erwin Schrödinger
was awarded the Nobel Prize for his work as early as 1933.

In stark contrast to classical mechanics, where the state of a system at a given time
can be described by the positions ri (t) and momenta pi (t) of all particles, the state of
a quantum mechanical system is represented by a so-called wave function ψ(r, t). Here,
the Schrödinger equation is the differential equation with which one can determine the
temporal evolution of this wave function (and thus of the quantum mechanical system).

2.1 Motivation of the Schrödinger Equation

The Schrödinger equation cannot be directly derived from classical physics. However, starting
from the idea of the matter wave and the quantization of energy in the form of “energy packets”
E = ℏω, it can at least be “motivated” as we will demonstrate in the following.

We consider a moving mass particle (for simplicity initially only in one dimension), based on de
Broglie’s idea of the matter wave in free space. In other words, we postulate that the moving
mass particle can be described by a wave equation. The general solution of a one-dimensional
wave equation is a superposition of plane waves of the form:

ψ(x, t) = C ei(kx−ωt) (2.1)

If we take the second spatial derivative of that wave function (2.1), we obtain:

∂2
ψ(x, t) = −k 2 C ei(kx−ωt) = −k 2 ψ(x, t) (2.2)
∂x2
According to de Broglie, for the momentum of matter waves p = ℏk applies following (1.40)
and thus p2 = ℏ2 k 2 (with |p|2 ≡ p2 and |k|2 ≡ k 2 ). From this follows k 2 = p2 /ℏ2 , which we can
substitute into (2.2):
∂2 p2
ψ(x, t) = − ψ(x, t)
∂x2 ℏ2
The kinetic energy of a moving particle can be represented (in the non-relativistic case) as
Ekin = mv 2 /2 = m2 v 2 /(2m) = p2 /(2m). This can be rearranged to p2 = 2mEkin . Substituting
into the above equation, we get the following differential equation:

∂2 2m
2
ψ(x, t) = − 2 Ekin (x, t)ψ(x, t) =⇒
∂x ℏ
ℏ2 ∂ 2
Ekin (x, t) ψ(x, t) = − ψ(x, t) (2.3)
2m ∂x2
For the total energy, we have E = Ekin + Epot . We represent the potential energy Epot initially
via a time-independent energy potential V (x), from which follows: E = Ekin (x, t) + V (x), or

13
 2 Schrödinger Equation

Ekin (x, t) = E − V (x). Substituting again into (2.3), we obtain a differential equation for the
total energy E:

ℏ2 ∂ 2
[E − V (x)] ψ(x, t) = − ψ(x, t) =⇒
2m ∂x2
ℏ2 ∂ 2
Eψ(x, t) = − ψ(x, t) + V (x)ψ(x, t) =
2m ∂x2 !
ℏ2 ∂ 2
= − + V (x) ψ(x, t) (2.4)
2m ∂x2

With this, we have nearly the (one-dimensional) Schrödinger equation! But only nearly, because
there is a (second) derivative with respect to position, but not yet a time derivative. Also, we
need an expression for the total energy E. Here, de Broglie helps us again. Let us consider
the time derivative of the wave function (2.1):
∂
ψ(x, t) = −iω Cei(kx−ωt) = −iω ψ(x, t) (2.5)
∂t
According to (1.35), de Broglie postulates for matter waves E = ℏω, or ω = E/ℏ. Thus, we
can rewrite (2.5) as:
∂ i ∂
ψ(x, t) = − Eψ(x, t) =⇒ Eψ(x, t) = iℏ ψ(x, t) (2.6)
∂t ℏ ∂t
Substituting this insight into (2.4), we finally obtain the complete (one-dimensional) Schrödinger
equation: !
∂ ℏ2 ∂ 2
iℏ ψ(x, t) = − + V (x) ψ(x, t) (2.7)
∂t 2m ∂x2
If you go from one dimension to R3 , you need to replace the second spatial derivative with the
Laplace operator ∆. In Cartesian coordinates, ∆ is simply defined as:

∂2 ∂2 ∂2
∆ = ∇2 = + +
∂x2 ∂y 2 ∂z 2
Combining all the above points, you get a general expression for the Schrödinger equation of a
particle:
!
ℏ2 ∂
− ∆ + V (r) ψ(r, t) = iℏ ψ(r, t) (2.8)
2m ∂t

The kinetic and potential energy terms on the left in (2.8) lead to an expression corresponding to
the classical Hamiltonian H(r, p). This similarity will be discussed in more detail in Appendix
10.1.

For a free particle, that is, for vanishing potential (V (r) = 0), a simplified expression applies in
three-dimensional space:
ℏ2 ∂
− ∆ψ(r, t) = iℏ ψ(r, t) (2.9)
2m ∂t
For a system consisting of N interacting particles in a potential V1 (ri ), the interaction force
V2 (rrel ) between the individual particles with respect to each other (rrel = ri − rj ) must also be
included:
 
N N
1X
!
−ℏ2 ∂
∆i + V1 (ri ) + V2 (ri − rj ) ψ(r1 , . . . , rN , t) = iℏ ψ(r1 , . . . , rN , t) (2.10)
X
2mi 2 i̸=j

i=1
∂t

14
Quantum Theory I

If we consider only a system of N particles that are not interacting with each other, the
Schrödinger equation (2.10) simplifies again to:
N
!
−ℏ2 ∂
∆i + V1 (ri ) ψ(r1 , . . . , rN , t) = iℏ ψ(r1 , . . . , rN , t) (2.11)
X

i=1
2mi ∂t

2.1.1 Correspondence Principle

Motivation: Representation of Classical Measurable Quantities in Quantum

Mechanics
In classical mechanics, measurable system properties A, called observables, are described
as a function of the phase space, i.e., as a function of all (generalized) coordinates qi and
momenta pi : A(t) = A(q1 , . . . , qN , p1 , . . . , pN , t).

However, as we have seen, a quantum mechanical system is not described by the coordi-
nates and momenta of all particles, but by a wave function. Thus, instead of a function
with qi and pi as function arguments for each classical observable A(t), we need a mathe-
matical formalism that allows us to calculate the desired physical observables for a given
wave function ψ(r, t). This is achieved with the help of operators on Hilbert spaces, where
certain classical observables are replaced by their corresponding quantum mechanical op-
erators.

In fact, it turns out that for each classical observable A(t), there is an equivalent operator
 that can act on the wave function ψ(r, t): A(q1 , . . . , qN , p1 , . . . , pN , t) ↔ Âψ(r, t). This
is called the correspondence principle. The properties of such operators will be discussed
in more detail in later chapters.

Although we shall rigorously define operators later, we want to briefly introduce some important
operators at this point, which will simplify the representation of future equations.

For the energy E, the already derived correspondence identity in (2.6) applies:
∂
E ↔ Ê = iℏ (2.12)
∂t
If we let Ê act on a wave function, we obtain the corresponding energy. Let’s next see what the
operator −iℏ ∂x∂
(which appears in second order in the Schrödinger equation) yields when acting
directly on ψ(x, t):
∂ ∂  i(kx−ωt)   
−iℏ ψ(x, t) = −iℏ Ce = ℏk C ei(kx−ωt) = ℏk ψ(x, t) (2.13)
∂x ∂x
According to de Broglie, p = ℏk, thus −iℏ ∂x
∂
is associated with the momentum operator:

∂
p ↔ p̂ = −iℏ (2.14)
∂x
For the momentum operator in three-dimensional space R3 , we analogously find:

p ↔ p̂ = −iℏ∇ (2.15)

In view of the three-dimensional Schrödinger equation (2.8), we find for the quadratic momentum
operator p̂2 the following relation:

p2 ↔ p̂2 = −ℏ2 ∆ (2.16)

15
 2 Schrödinger Equation

Let’s look once more at the Schrödinger equation (2.8). As noted in (2.12), the right, time-
dependent side gives the total energy E. But then, of course, the left side of (2.8) must also
provide the total energy! Thus, the following expression is also an energy operator, which we
henceforth refer to as the Hamiltonian operator Ĥ:

ℏ2
Ĥ = − ∆ + V (r) (2.17)
2m

For a particle with mass m in an arbitrary potential V (r), Ĥ can be written with the help of
(2.15) as:
ℏ2 (2.16) p̂
2
E ↔ Ĥ = − ∆ + V (r) = + V (r̂) (2.18)
2m 2m
In this way, the Schrödinger equation from (2.8) can be compactly written:

∂
Ĥψ(r, t) = iℏ ψ(r, t) (2.19)
∂t

2.1.2 Probability Density

The probability density ρ(r, t), which is described by the expression ρ(r, t)dr with which prob-
ability the particle is located at time t in an infinitesimally small volume element dr, can be
calculated from the wave function ψ(r, t) as follows:

ρ(r, t) = ψ ∗ (r, t)ψ(r, t) = |ψ(r, t)|2 (2.20)

One can see that from experimentally determined values of ρ(r, t) ≡ ρ, the underlying wave
function ψ(r, t) can never be completely reconstructed, because the magnitude square eliminates
all phase information: A wave function ψ̃ = ψeiφ with real phase φ yields exactly the same
probability density as ψ. By calculating the complex conjugate of (2.20), one obtains the relation:

ρ∗ = (ψ(r, t)∗ ψ(r, t))∗ = ψ(r, t)ψ(r, t)∗ = ψ(r, t)∗ ψ(r, t) = ρ

This must obviously be the case, because the probability density ρ(r, t) is a real number.
In the one-dimensional case, the probability w(t) that a particle is located at time t within a
specific range between x = a and x = b is given by:
Z b Z b
w(t) = dx ρ(x, t) = dx |ψ(x, t)|2 (2.21)
a a

Since the particle considered (in the one-dimensional case) must with absolute certainty at any
time be located somewhere between x = −∞ and x = +∞, the integration of ρ(x, t) from
x = −∞ to x = +∞ must yield exactly 1 in the sense of a probability interpretation:
Z ∞ Z ∞
dx ρ(x, t) = dx |ψ(x, t)|2 = 1 (2.22)
−∞ −∞

In the three-dimensional case, similarly the integration over the entire space V must again yield
exactly 1: Z Z
dV ρ(r, t) = dV |ψ(r, t)|2 = 1 (2.23)
V V
It should be noted that this normalization does not automatically result from the Schrödinger
equation. Rather, the wave function ψ(r, t) must be provided with an appropriate constant in
a separate calculation step to satisfy the normalization conditions (2.22) or (2.23).

16
Quantum Theory I

For a normalization of the wave function with a suitable constant to be possible, all valid wave
functions must belong to the so-called square-integrable functions, i.e., satisfy the following
condition: Z
dV |ψ(r, t)|2 < ∞ (2.24)
V
The fact that w(t) can be regarded as a probability is motivated by two assumptions: w(t) is
real and positive definite (w(t) ≥ 0). The first condition follows from ρ∗ = ρ:
Z b Z b
w∗ = dx ρ∗ = dx ρ = w
a a

Since w∗ = w, w must be real. Together with the postulate of the positive definite metric, w(t)
is consistent with the interpretation as a probability.

Example: Calculation of a Probability Density

We assume a wave function ψ, which is composed of the superposition of two separate
wave functions ψ1 and ψ2 . It holds:

ψ = a ψ1 + b ψ2

The probability density can now be calculated via the square of the modulus:

|ψ|2 = ψ ∗ ψ = (a∗ ψ1∗ + b∗ ψ2∗ )(a ψ1 + b ψ2 ) =

= |a|2 |ψ1 |2 + |b|2 |ψ2 |2 + a∗ ψ1∗ bψ2 + b∗ ψ2∗ aψ1 = | z = a∗ ψ1∗ bψ2 , z ∗ = b∗ ψ2∗ aψ1
= |a|2 |ψ1 |2 + |b|2 |ψ2 |2 + z + z ∗ = | z + z ∗ = 2 Re(z)
= |a|2 |ψ1 |2 + |b|2 |ψ2 |2 + 2 Re(a∗ ψ1∗ bψ2 )

Besides the expected terms describing the probability of the presence of the respective wave
functions ψ1 and ψ2 , a further expression called the interference term is obtained. This
term is responsible for special phenomena in quantum mechanics, for example, explaining
the interference pattern in the double-slit experiment. The form of the interference term
is to be understood using two complex numbers α = aψ1 and β = bψ2 :

α∗ β + αβ ∗ = (αβ ∗ )∗ + αβ ∗ = Re(αβ ∗ ) − i Im(αβ ∗ ) + Re(αβ ∗ ) + i Im(αβ ∗ ) = 2 Re(αβ ∗ )

Another computation possibility for the complex numbers is:

α∗ β + αβ ∗ = [Re(α) − i Im(α)][Re(β) + i Im(β)] + [Re(α) + i Im(α)][Re(β) − i Im(β)] =

= 2 Re(α) Re(β) + 2 Im(α) Im(β) = 2 [Re(αβ) + Im(αβ)]

2.1.3 Probability Current Density

Motivation: Probability as a Conservation Quantity

As we have already determined, the probability of finding an observed particle somewhere
in the entire space must always be one. Strictly speaking, this is a constraint of Schrödin-
gerian quantum mechanics. In high-energy physics, particles can indeed be created or
annihilated – an aspect considered in quantum field theory (QFT). The probability of
presence can be higher at time t1 in one space region than in another, and at a later time
t2 , it may be the reverse. But overall, the total probability across the entire space always
remains one, thus the total probability is a conservation quantity!

17
 2 Schrödinger Equation

The entire space thus resembles a container in some way, in which the probability density
can “flow” from one place to another, but additional probability is never “created or
destroyed”. Therefore, there must be a continuity equation, and a probability current
density must be definable.

For the probability density of presence, we have previously determined the following expression:
ρ(r, t) = |ψ(r, t)|2 , where ρ is a positive definite quantity. Integrating over the entire space, ρ
always yields 1, meaning that the total probability remains conserved. Simultaneously, ρ(r, t) can
change over time in each space region. Therefore, we expect there must be a probability current
density j that, together with ρ, satisfies a continuity equation (similar to, e.g., the charge carrier
density and charge current density in electrodynamics fulfilling a continuity equation):

∂ρ
= −∇ · j (2.25)
∂t

It becomes evident: The faster the probability density ρ changes over time, the greater must be
the spatial change of the probability current density j. Now we attempt to find an expression
for the probability current density j such that the above equation (2.25) is satisfied. To do this,
we need to form the time derivative of ρ = ψ ∗ ψ:
∂ρ ∂ ∂ψ ∗ ∂ψ
= (ψ ∗ ψ) = ψ + ψ∗ (2.26)
∂t ∂t ∂t ∂t
To convert the time derivative of ψ into a space derivative, we use the Schrödinger equation
(2.9) for a free particle:

ℏ2 ∂ψ ∂ψ ℏ
− ∆ψ = iℏ =⇒ =− ∆ψ (2.27)
2m ∂t ∂t 2im
We obtain an expression for the time derivative of ψ ∗ by taking the complex conjugate of both
sides of the Schrödinger equation (2.9):
!∗ ∗
ℏ2 ∂ψ ℏ2 ∂ψ ∗ ∂ψ ∗

ℏ
− ∆ψ = iℏ =⇒ − ∆ψ ∗ = −iℏ =⇒ = ∆ψ ∗ (2.28)
2m ∂t 2m ∂t ∂t 2im

Substituting (2.28) and (2.27) into (2.26), we get:

∂ρ ℏ
= [(∆ψ ∗ )ψ − ψ ∗ ∆ψ] (2.29)
∂t 2im
We now have an equation that on the left side includes the time derivative of ρ and on the right
side includes space derivatives of ψ and ψ ∗ . However, we cannot yet bring the right side of (2.29)
into direct agreement with the continuity equation (2.25) because the latter is determined by
the divergence of a (probability) current. Therefore, a further transformation is required, for
which we must first derive the following identities using the product rule:

∇ · [(∇ψ ∗ )ψ] = (∆ψ ∗ )ψ + (∇ψ ∗ )(∇ψ) =⇒ (∆ψ ∗ )ψ = ∇ · [(∇ψ ∗ )ψ] − (∇ψ ∗ )(∇ψ)
(2.30)
∇ · (ψ ∗ ∇ψ) = (∇ψ ∗ )(∇ψ) + ψ ∗ ∆ψ =⇒ ψ ∗ ∆ψ = ∇ · (ψ ∗ ∇ψ) − (∇ψ ∗ )(∇ψ)

18
Quantum Theory I

By substituting both relationships from (2.30) into (2.29), we finally get:

∂ρ ℏ
= [∇ · [(∇ψ ∗ )ψ] − (∇ψ ∗ )(∇ψ) − ∇ · (ψ ∗ ∇ψ) + (∇ψ ∗ )(∇ψ)] =
∂t 2im
ℏ
= ∇ · [(∇ψ ∗ )ψ − ψ ∗ ∇ψ] =
2im
h ℏ  ℏ  i ℏ
= −∇ · ψ ∗ ∇ψ + ψ − ∇ψ ∗ = | α = ψ ∗ ∇ψ
2im 2im 2im
= −∇ · (α + α∗ ) = −∇ · [2 Re (α)] =
 
∗ ℏ (2.25)
= −∇ · Re ψ ∇ψ = −∇ · j (2.31)
im
| {z }
j

Comparing this result with the continuity equation (2.25), we can immediately read off the
sought expression for the probability current density j(r, t):
 
ℏ ∗ ℏ
j(r, t) = Re ψ ∇ψ = [ψ ∗ ∇ψ − (∇ψ ∗ )ψ] (2.32)
im 2im

In this derivation, we used the Schrödinger equation for a free particle. Expression (2.32) and the
continuity equation (2.25) are valid (without proof) for a particle in a (real-valued) potential
as well. Moreover, it follows from (2.25) that the probability density remains conserved. To
demonstrate this, we apply Gauss’ theorem (with surface normal n):
Z I
− dV ∇ · j = − dS n · j = 0
V ∂V

If ψ(r, t) is a square-integrable function, it can be normalized, so that the probability of pres-

ence at the edge of the system vanishes. From this follows the conservation of norm with real
potentials.

2.1.4 Time-independent Schrödinger Equation

In this chapter, we show that the Schrödinger equation can be solved by separating the wave
function into a space-dependent and a time-dependent part, and how the solution that depends
only on space relates to the so-called time-independent Schrödinger equation.

The Schrödinger equation (2.8) and the resulting wave function ψ(r, t) are, evidently, time and
space-dependent: !
ℏ2 ∂
− ∆ + V (r) ψ(r, t) = iℏ ψ(r, t)
2m ∂t
As long as the potential V (r) depends only on the location, we can choose an ansatz where the
wave function ψ(r, t) is factorizable into a space-dependent part and a time-dependent part:
ψ(r, t) = ψ̃(r)χ(t) (2.33)
Substituting the ansatz (2.33) for these special solutions into the Schrödinger equation (2.8), we
succeed in separating space and time variables:
ℏ2   ∂  
− ∆ ψ̃(r)χ(t) + V (r)ψ̃(r)χ(t) = iℏ ψ̃(r)χ(t)
2m ∂t
ℏ2 dχ(t) h i
− χ(t)∆ψ̃(r) + V (r)ψ̃(r)χ(t) = iℏψ̃(r) | ÷ ψ̃(r)χ(t)
2m dt
ℏ2 1 1 dχ(t)
∆ψ̃(r) + V (r) = iℏ (2.34)
2m ψ̃(r) χ(t) dt

19
 2 Schrödinger Equation

The left side of (2.34) depends only on r, the right side depends only on t. To satisfy the
equation for all r and t, both sides must correspond to a constant. We wisely call this constant
E, and first look at the right side of equation (2.34), which we also multiply by χ(t):

d
iℏ χ(t) = Eχ(t) (2.35)
dt

A comparison with (2.12) shows: Since Ê = iℏ dt d

is the energy operator, the chosen constant E
is obviously the total energy. A solution for χ(t) can be easily found:

dχ 1 1 dχ E 1 E 1 E
Z Z
iℏ = Eχ | · =⇒ = | · dt =⇒ dχ = dt =⇒ dχ = dt =⇒
dt iℏχ χ dt iℏ χ iℏ χ iℏ
E Et
ln(χ) = t + c = −i + c =⇒ χ(t) = e−iEt/ℏ+c = e−iEt/ℏ ec = N e−iEt/ℏ
iℏ ℏ
We simply choose the constant N = 1, because we can, due to the ansatz ψ(r, t) = ψ̃(r)χ(t),
shift all normalization into ψ̃(r). It holds E = ℏω or ω = E/ℏ, enabling us to write the solution
for χ(t) as:
χn (t) = e−iEn t/ℏ = e−iωn t (2.36)
Due to the mathematical form of χn (t), the time-dependent part of the solution corresponds to
a phase factor. The index n suggests, as we will see, that there may be several (often infinitely
many) solutions of the Schrödinger equation with different energies En . Let’s now look at the
left – space-dependent – side of equation (2.34) and equate it again to the constant energy E.
After multiplying with ψ̃(r):

ℏ2
− ∆ψ̃(r) + V (r)ψ̃(r) = E ψ̃(r) (2.37)
2m

Equation (2.37) is the time-independent Schrödinger equation. We can write it compactly using
the Hamiltonian operator (2.18):

Ĥ ψ̃n (r) = En ψ̃n (r) (2.38)

Again, we use the index n, indicating that there may be several solutions with different energies.
The time-independent Schrödinger equation is generally a linear, elliptic, partial differential
equation, finding its solution will be the task of the following chapters.

Stationary Solutions We have shown that, given a time-independent potential V (r), special
solutions of the Schrödinger equation can be factorized into a space-dependent part ψ̃n (r) and
a time-dependent part χn (t) so that:
(2.36)
ψn (r, t) = ψ̃n (r)χn (t) = ψ̃n (r)e−iωn t (2.39)

The index n again characterizes different solutions with different energies En . Observing these
special solutions closely, the associated probability of presence is time-independent (stationary):

ρ = ψ ∗ (r, t)ψ(r, t) = ψ̃ ∗ (r)eiωt ψ̃(r)e−iωt = |ψ̃(r)|2 (2.40)

One speaks of stationary solutions of the Schrödinger equation, which are determined by solving
the time-independent Schrödinger equation (2.38) and are each associated with a well-defined
energy En .

20
Quantum Theory I

Superposition of States When is it not sufficient to solve only the time-independent Schrödinger
equation (2.38)? Since the Schrödinger equation is a linear differential equation (meaning, ψ
appears only as a first power), the superposition principle applies: arbitrary linear combina-
tions of solutions are again solutions. Thus, we may be interested in solutions representing a
superposition of several stationary solutions, such as:
 
Ψ(r, t) = N (ψ1 (r, t) + ψ2 (r, t)) = N ψ̃1 (r)e−iω1 t + ψ̃2 (r)e−iω2 t (2.41)

The constant N is a normalization constant to be chosen such that the normalization condition
(2.23) is satisfied. Calculating the probability density ρ for Ψ(r, t), it is seen that it is not
time-independent:
ρ = Ψ∗ (r, t)Ψ(r, t) =
h ih i
= |N |2 ψ̃1∗ (r)eiω1 t + ψ̃2∗ (r)eiω2 t ψ̃1 (r)e−iω1 t + ψ̃2 (r)e−iω2 t =
h i
= |N |2 |ψ̃1 (r)|2 + |ψ̃2 (r)|2 + ψ̃1∗ (r)ψ̃2 (r)ei(ω1 −ω2 )t + ψ̃1 (r)ψ̃2∗ (r)ei(ω2 −ω1 )t = | α = ψ̃1∗ ψ̃2 ei∆ωt
h i
= |N |2 |ψ̃1 (r)|2 + |ψ̃2 (r)|2 + α + α∗ = | α + α∗ = 2 Re(α)
h  i
= |N |2 |ψ̃1 (r)|2 + |ψ̃2 (r)|2 + 2 Re ψ̃1∗ (r)ψ̃2 (r)ei(ω1 −ω2 )t (2.42)

The wave function Ψ(r, t), which cannot be assigned to a definite energy En , is not a solution
to the time-independent Schrödinger equation, and its time development results from the super-
position of several stationary solutions (with time-dependent phase factors). Alternatively, one
can attempt a direct solution of the time-dependent Schrödinger equation.

2.1.5 Properties of the Schrödinger Equation

An exact solution of the Schrödinger equation is only possible for very few systems. Through a
thorough mathematical classification of the Schrödinger equation, helpful insights into possible
solutions can nevertheless be derived. Properties of the Schrödinger equation are:
Sturm-Liouville Problem: The one-dimensional, time-independent Schrödinger equa-
tion can be understood as a Sturm-Liouville eigenvalue problem. This means that we can
find real, ordered eigenvalues λn (λ1 < λ2 < . . . ), whose corresponding eigenfunctions will
have n − 1 zeros. These “nodes” lead us to the node theorem, allowing an ordering of the
eigenvalues. The eigenfunctions are orthogonal to each other and form a complete basis.
Partial Differential Equation: The time-independent Schrödinger equation is a partial
differential equation because it includes both partial derivatives with respect to position
r and time t. However, the time derivative is a simple derivative, while there is a second
derivative concerning the position: thus, it is a first-order differential equation in time
and a second-order one in space. The lack of symmetry in space and time derivatives is
problematic for a relativistic application of the Schrödinger equation because, in relativ-
ity, space and time coordinates are treated equitably. This issue is solved by the Dirac
equation, which, however, is not covered in this lecture.
Parabolic Differential Equation: The time-dependent Schrödinger equation belongs
to the class of parabolic differential equations because the time component is influenced
by a first time derivative. Due to the dependence on position and time, solving the
Schrödinger equation is both a boundary value and an initial value problem. Another
parabolic differential equation is the heat equation.
Linear Differential Equation: The Schrödinger equation is a linear differential equa-
tion, meaning ψ appears only as a first power. This, in turn, means that even a su-
perposition of solutions must again be a solution of the Schrödinger equation. To show

21
 2 Schrödinger Equation

this, assume that ψ and ϕ represent valid wave functions. Then the same must hold for
Ψ = αψ + βϕ:
∂ ∂ ∂ ∂
       
Ĥ − iℏ Ψ = Ĥ − iℏ (αψ + βϕ) = α Ĥ − iℏ ψ + β Ĥ − iℏ ϕ=0 □
∂t ∂t ∂t ∂t

Homogeneous Differential Equation: The homogeneity of the Schrödinger equation

is a prerequisite for the wave function to be normalizable at all times. If this assumption is
abandoned, the conservation of the norm of the Schrödinger equation is violated (likewise,
this is the case when the potential contains imaginary terms).

Born’s Probability Interpretation: The wave functions ψ(r, t) as solutions of the

Schrödinger equation escape a direct physical interpretation. Yet ρ(r, t) = |ψ(r, t)|2 from
(2.20) can be understood as a probability density.

In-Depth: No Norm Conservation with Complex Potential

Typically (and throughout this script), the potential term in the Schrödinger equation is
real-valued. However, in certain contexts, it can be practical to assume a complex-valued
potential V (r):
!
ℏ2 ∂ ℏ2
− ∆ + V (r) ψ(r, t) = iℏ ψ(r, t) with Ĥ = − ∆ + V (r) (2.43)
2m ∂t 2m

Since V (r) now has imaginary parts, V (r) ̸= V ∗ (r) and the Hamiltonian is thus no longer
Hermitian overall (Ĥ ̸= Ĥ ∗ ). We now want to clarify the effect on the continuity equation.
We begin with the same approach (2.26) as with the free particle:

∂ρ ∂ ∂ψ ∗ ∂ψ
= (ψ ∗ ψ) = ψ + ψ∗
∂t ∂t ∂t ∂t

To establish a relation between ∂ψ

∂t and ∆ψ, we divide both sides of the (complex-conjugated)
Schrödinger equation (2.43) by iℏ:

∂ψ ℏ V ∂ψ ∗ ℏ V∗ ∗
=− ∆ψ + ψ and = ∆ψ ∗ − ψ (2.44)
∂t 2im iℏ ∂t 2im iℏ
We now substitute both relations from (2.44) into the approach (2.26):

∂ρ V∗ ∗ V
   
ℏ ℏ
= ∆ψ ∗ − ψ ψ + ψ∗ − ∆ψ + ψ =
∂t 2im iℏ 2im iℏ
ℏ 1
= [(∆ψ ∗ )ψ − ψ ∗ ∆ψ] + ψ ∗ ψ (V − V ∗ ) (2.45)
2im
| {z } iℏ
−∇·j

The first term on the right side of equation (2.45) is identical to expression (2.29), which
in further progression was identified in (2.31) as −∇ · j. For the expression (V − V ∗ ), the
following calculation rule for complex numbers applies: V − V ∗ = 2i Im(V ). Thus, we
obtain the following modified continuity equation:
∂ρ 2
= −∇ · j + Im(V ) (2.46)
∂t ℏ
The continuity equation now includes an additional term due to the complex-valued po-
tential, which, depending on the sign of Im(V ), corresponds to a particle source or sink.

22
Quantum Theory I

Consequently, the total probability ρ (i.e., the integrated probability density over the en-
tire space) is generally not conserved. This allows relatively simple simulation of physical
situations where additional particles can emerge or disappear.

In-Depth: No Norm Conservation with Inhomogeneous Schrödinger Equation

The physically correct Schrödinger equation is a homogeneous differential equation. To
simplify the simulation of specific physical configurations, it might be helpful to add an
inhomogeneity term. Therefore, we assume a Schrödinger equation for a free particle with
a (complex) inhomogeneity s(r, t):

ℏ2 ∂ ℏ2
− ∆ψ(r, t) + s(r, t) = iℏ ψ(r, t) with Ĥ = − ∆ (2.47)
2m ∂t 2m

At least here Ĥ = Ĥ ∗ (the Hamiltonian is Hermitian). This is, however, not the case for
a complex-valued inhomogeneity: s(r, t) ̸= s∗ (r, t). Again, we aim to derive a continuity
equation. Proceeding similarly as in (2.26):

∂ρ ∂ ∂ψ ∗ ∂ψ
= (ψ ∗ ψ) = ψ + ψ∗
∂t ∂t ∂t ∂t
As in the preceding example, we divide the Schrödinger equation (2.47) by iℏ to create a
simple connection between the temporal and spatial derivatives. Following, we take the
complex conjugate of both sides of the equation:
∂ψ ℏ 1 ∂ψ ∗ ℏ 1
=− ∆ψ + s and = ∆ψ ∗ − s∗ (2.48)
∂t 2im iℏ ∂t 2im iℏ
With this, we can proceed with the tried-and-true approach by substituting (2.48) into
(2.26):

∂ρ 1 1
   
ℏ ℏ
= ∆ψ ∗ − s∗ ψ + ψ ∗ − ∆ψ + s =
∂t 2im iℏ 2im iℏ
ℏ 1
= [(∆ψ ∗ )ψ − ψ ∗ ∆ψ] − [s∗ ψ − ψ ∗ s] = | α = s∗ ψ
2im iℏ
ℏ 1
= [(∆ψ )ψ − ψ ∆ψ] − (α − α∗ )
∗ ∗
(2.49)
2im
| {z } iℏ
−∇·j

The first term on the right of equation (2.49), as we have learned from (2.29) and (2.31), is
identical to −∇ · j. For the expression (α − α∗ ), the following calculation rule for complex
numbers applies: α − α∗ = 2i Im(α) ≡ 2i Im(s∗ ψ). Consequently, we obtain the following
modified continuity equation:
∂ρ 2
= −∇ · j − Im(s∗ ψ) (2.50)
∂t ℏ
Once again, the continuity equation includes an additional term which, depending on the
sign of Im(s∗ ψ), acts absorbingly or enhancingly, breaking the conservation property of
the continuity equation. In this manner, one can fairly easily calculate damping effects.

23
 2 Schrödinger Equation

2.2 Wave Packets

Motivation: Describing Classical Processes with Quantum Physics
We have already shown that the one-dimensional Schrödinger equation (2.7) is solved
by plane waves of the form ψ(x, t) = Aei(kx−ωt) . However, when we want to represent
something as elementary as the motion of a single particle in space with it, we encounter
multiple problems:

Classical particles are precisely localizable. We expect the quantum theory to repre-
sent the position of a particle, at least with the corresponding quantum mechanical
probability.

In contrast to this expectation, plane waves are entirely delocalized: The probability
density of a plane wave ρ = |ψ(x, t)|2 = |Aei(kx−ωt) |2 is constant at every point in
space ρ = |A|2 .

Thus, ψ(x, t) cannot be normalized according to the normalization condition (2.22):

The integral of a constant over an infinite space region goes against infinity!

We will show in this chapter that the solution to these problems lies in combining an
infinite number of plane waves in superposition, so that the overall solution ultimately
yields a spatially localized wave packet, identifying it with a particle that (even if with
corresponding “quantum fuzziness”) follows the laws of classical physics in specific limit
cases.

We seek a suitable superposition of solutions of the Schrödinger equation ψi . Thus, one might
try the following ansatz (A here being the overall normalization constant):

ϕ(x, t) = A ai ψi (x, t) = A (2.51)

X X
ai ei(ki x−ωi t)
i i

The sum in (2.51) represents a discrete superposition of plane waves. In this discrete superpo-
sition, the various coefficients ai are complex numbers defining the weightings of the individual
solutions in the superposition. However, such a discrete superposition never leads to a solution
localized in space. We can achieve this with a continuous superposition of the following form:
Z ∞ Z ∞
ϕ(x, t) = A dk a(k)ψk (x, t) = A dk a(k)ei(kx−ωk t) (2.52)
−∞ −∞

Instead of the discrete weighting coefficients ai in (2.51), a weighting function a(k) must be used
in the continuous superposition (2.52). It turns out that the following weighting function is well
suited for the formation of localized wave packets:
1 2 2
a(k) = √ e−(k−k0 ) d (2.53)
2π
This corresponds to a Gaussian function shifted by k0 ; a(k) clearly has the property of tending
towards zero in infinity, possibly ensuring normalizability over the entire space. By substituting
(2.53) into (2.52), we obtain:
Z ∞
A 2 d2
ϕ(x, t) = √ dk e−(k−k0 ) ei(kx−ωk t) (2.54)
2π −∞

For the time t = 0, the superposition (2.54) could also be written in the following form:
1
Z ∞
2 d2
ϕ(x, 0) = √ dk ϕk (k)eikx with ϕk (k) = Ae−(k−k0 ) (2.55)
2π −∞

24
Quantum Theory I

But this now corresponds exactly to the (inverse) Fourier transform of √ the function ϕk (k) from
the k-space into the real space (with the convention of the pre-factor 1/ 2π in both the forward
and the inverse transformation)! This means we can also understand our approach this way:
First, we “build” a wave packet in real space using the Gaussian function ϕk (k) and inverse
Fourier transform at the time t = 0. In this context, the factor k0 in the Gaussian function
corresponds to the initial momentum p0 = ℏk0 of the wave packet. We then ask how this
wave packet will evolve over time. Finally, we realize that the term eikx in the Fourier integral
corresponds to a plane wave, and we already know the (trivial) time evolution e−iωk t for plane
waves. Therefore, we can simply add e−iωk t to the integrand to describe the time evolution,
leading us back to exactly the same solution as before in (2.54):
1
Z ∞
2 d2
ϕ(x, t) = √ dk ϕk (k)eikx e−iωk t with ϕk (k) = Ae−(k−k0 ) (2.56)
2π −∞

We must still express the dependence between the angular frequency ωk and the wave number
k: It holds Ek = ℏωk and therefore ωk = Ek /ℏ; based on E = p2 /2m and p = ℏk, we can thus
express the angular frequency ωk as a function of the wave number k as follows:

Ek p2 ℏ2 k 2 ℏk 2
ωk = = = = (2.57)
ℏ 2mℏ 2mℏ 2m
After substituting this relationship into (2.54), we finally derive the following expression for the
wave packet ϕ(x, t): Z ∞
A ℏk2


2 2
ϕ(x, t) = √ dk e−(k−k0 ) d ei kx− 2m t (2.58)
2π −∞
By substituting (2.57) into (2.56), we obtain the equivalent representation in the “Fourier view”:
1
Z ∞
ℏk2 2 d2
ϕ(x, t) = √ dk ϕk (k)eikx e−i 2m t with ϕk (k) = Ae−(k−k0 ) (2.59)
2π −∞

If we substitute (2.59) into the Schrödinger equation (2.9) for a free particle, we can confirm
that the wave packet ϕ(x, t) is a valid solution of the Schrödinger equation for all times t:

ℏ2 d2 d (2.59)
− ϕ(x, t) = iℏ ϕ(x, t) =⇒
2m dx 2 dt
ℏ2 d2 1 1
Z ∞ Z ∞
d
   
ℏk2 2
ikx −i 2m t ikx −i ℏk
− √ dk ϕk (k)e e = iℏ √ dk ϕk (k)e e 2m
t
=⇒
2m dx2 2π −∞ dt 2π −∞
ℏ2 1 1
!
d2 ikx −i ℏk2 t
Z ∞ Z ∞
d −i ℏk2 t
 
− √ dk ϕk (k) e e 2m = iℏ √ dk ϕk (k)e ikx
e 2m =⇒
2m 2π −∞ dx2 2π −∞ dt
ℏ2 k 2 1 ℏ2 k 2 1
Z ∞ 2 ∞ ℏk2
Z
ikx −i ℏk (2.59)
√ dk ϕk (k)e e = 2m
t
√ dk ϕk (k)eikx e−i 2m t =⇒
2m 2π −∞ 2m 2π −∞
ℏ2 k 2 ℏ2 k 2
ϕ(x, t) = ϕ(x, t) □
2m 2m
We have thus found a solution for the time evolution of the wave packet without explicitly
solving the Schrödinger equation as a differential equation.

2.2.1 Calculation of the Wave Packet in Position Space

In what follows, the integral (2.58) shall be executed. To do this, we first rewrite (2.58) as
follows:
!
ℏk 2
Z ∞
A
(2.58)
ϕ(x, t) = √ dk e ζ(k,x,t)
with ζ(k, x, t) = −(k − k0 ) d + i kx −
2 2
t (2.60)
2π −∞ 2m

25
 2 Schrödinger Equation

Next, we convert the exponent ζ(k, x, t) to complete a square in k:

ℏk 2
   
ζ(k, x, t) = i kx − t − k 2 − 2kk0 + k02 d2 =
2m
iℏt 2
= ixk − k − d2 k 2 + 2k0 d2 k − k02 d2 =
2m
iℏt
   
=− + d2 k 2 + ix + 2k0 d2 k − k02 d2 = (2.61)
2m | {z }
c
| {z | {z
} }
a 2b

= −ak 2 + 2bk − c | ÷ (−a) =⇒

 2  2 2
ζ(k, x, t) b b b c b b2 c

= k 2 − 2k + − + = k− − 2
+ | · (−a) =⇒
−a a a a a a a a
| {z }
(k−b/a)2
2
b b2

ζ(k, x, t) = −a k − + −c (2.62)
a a
In placeholders a, b, and c, all terms constant from the perspective of the integral have been
combined. The integration variable k now appears only in the expression −a(k − b/a)2 . This
is advantageous as all terms not dependent on k can be extracted from the integral, and the
remaining
p term of integration over a (shifted) Gaussian curve, for which we already know the
result π/a:

A +∞ (2.62) A +∞ b 2 b2
dk e−a(k− a ) + a −c =
Z Z
(2.60)
ϕ(x, t) = √ dk eζ(k,x,t) = √
2π −∞ 2π −∞
A b2 −c +∞ A b2
dk e−a(k− a ) = √ e a −c
Z
b 2
= √ ea (2.63)
2π −∞ 2a
| √{z }
π/a

The offset b/a does not change the result π/a of the integral in (2.63), which becomes clear
p

when b/a is real-valued (it doesn’t matter where the Gaussian function has its maximum, as
long as we integrate from −∞ to +∞). Verification shows this is also the case when b/a is
complex-valued, as in our case. Substituting the original values for a, b, and c from (2.61)
provides the analytical representation of the wave packet in real space:

( ix
2 + k0 d )
" #
2 2
A
ϕ(x, t) = r  exp − k02 d2 ≡ N eφ(x,t) (2.64)
iℏt
+ d2

2 2m
iℏt
+ d2 2m

The wave function ϕ(x, t) itself is not directly measurable. However, the probability of finding
a particle (based on ρ = |ϕ(x, t)|2 ) can be measured. Thus, we assess the square of the modulus
|ϕ(x, t)|2 .
(2.64)
|ϕ(x, t)|2 = |N |2 |eφ |2 = |N |2 eRe(φ) +i Im(φ) eRe(φ) −i Im(φ) = |N |2 e2 Re(φ) (2.65)

First, we handle the factor |N |2 in (2.65):

(2.64) A∗ A |A|2 |A|2

|N |2 = N ∗ N = r     = r h i = r h i =⇒
ℏ2 t 2 ℏ2 t2
2 d2 − iℏt
2m 2 d2 + iℏt
2m 4 d4 + 4m2
4d4 1 + 4m2 d4

|A|2 |A|2 def ℏt

|N |2 = 2 ≡ 2d2 p1 + ∆2 (t) with ∆(t) = (2.66)
2md2
r 
2d2 1 + ℏt
2md2

26
Quantum Theory I

Next, we need to compute the exponent 2 Re(φ) in (2.65):

( ix + k0 d2 )2
" #
(2.64)
2 Re(φ) = 2 Re 2 2 iℏt − k02 d2 =
d + 2m
2
(− x4 + k02 d4 + ixk0 d2 ) (d2 − 2m )
iℏt
" #
= 2 Re − k02 d2 =
(d2 + 2m iℏt
) (d2 − 2m )
iℏt

k02 d4 ℏt
 
2 2 2 xk0 d2 ℏt
− x 4d + k02 d6 + ixk0 d4 + i x8mℏt − i +
= 2 Re  2m 2m
− k02 d2  =
d4 + ( 2m )
ℏt 2
 
2 2 xk0 d2 ℏt 1
− x 4d + k02 d6 + d2
= 2 h 2m
i 1 − k02 d2  =
d4 1+ ( 2md 2)
ℏt 2
d2
 
x2
− + k02 d4 + xk0 d2 2md
ℏt
2 def ℏt
= 2 4
− k02 d2  = ∆(t) =
2md2
h i
d2 1 + ( 2md 2)
ℏt 2

2
− x + k02 d4 + xk0 d2 ∆(t) − k02 d4 1 + ∆2 (t)


= 2 4 =
d2 (1 + ∆2 (t))
1 x2 − 4xk0 d2 ∆(t) + 4k02 d4 ∆2 (t)
= −2 =
4 d2 (1 + ∆2 (t))
x2 − 4xd2 k0 ∆(t) + 4d4 [k0 ∆(t)]2
= − (2.67)
2d2 (1 + ∆2 (t))

For the quantum particle, the momentum relationship p0 = ℏk0 holds (where p0 represents the
initial momentum). With p0 = mv0 , this turns into mv0 = ℏk0 , which simplifies to k0 = mv0 /ℏ
(using v0 as the initial velocity). The auxiliary quantity ∆(t) can consequently be rewritten as:

mv0 ℏt v0 t
k0 ∆(t) = = 2
ℏ 2md2 2d
Substituting this back in the numerator of the recently derived expression (2.67) gives:
v 2 t2
x2 − 4xd2 2d 2 + 4d 4d4
v0 t 4 0
x2 − 2xv0 t + v02 t2 (x − v0 t)2
2 Re(φ) = − = − = − (2.68)
2d2 (1 + ∆2 (t)) 2d2 (1 + ∆2 (t)) 2d2 (1 + ∆2 (t))

Now we can substitute (2.66) and (2.68) into (2.65), yielding an (as yet unnormalized) expression
for |ϕ(x, t)|2 :
2
|A|2 − 2
(x−v0 t)
|ϕ(x, t)|2 = e 2d (1+∆2 (t)) (2.69)
2d2 1 + ∆2 (t)
p

Now let’s compare this result with the normalized Gaussian distribution:

1 −(x−µ0 )2
N (x; µ0 , σ 2 ) = √ e 2σ2 (2.70)
2πσ
∞
For the normalized Gaussian distribution N (x), it holds that −∞ dx N (x) = 1. We expect the
R

same for our result |ϕ(x, t)| , which also takes the form of a Gaussian function. By comparing
2

the exponents in (2.70) and (2.69), we can immediately read off the variance σ 2 and the standard
deviation σ for |ϕ(x, t)|2 :
h i q
σ 2 = d2 1 + ∆2 (t) ⇐⇒ σ = d 1 + ∆2 (t) (2.71)

27
 2 Schrödinger Equation

wave packet
t=0
Re(ϕ)

t=2

|ϕ|2
t=4

x x

Abb. 6: (left) Representation of the real part of a wave packet as a superposition of plane waves.
(right) Schematic representation of the temporal evolution of a wave packet: As time
progresses, there is a shift in the maximum, but also a spreading of the probability of
presence in space. The norm of the probability density remains conserved.

Substituting this into (2.69), we obtain:

|A|2 − (x−v02t)2
|ϕ(x, t)|2 = e 2σ (2.72)
2dσ
By comparing the prefactors of (2.72) and (2.70), the missing normalization |A|2 can finally be
determined:
|A|2 1 2dσ 2
r
=√ =⇒ |A|2 = √ = d (2.73)
2dσ 2πσ 2πσ π

Substituting this into (2.69), we finally obtain the following normalized expression for the prob-
ability density of the wave packet:
2
1 − 2
(x−v0 t)
ℏt
|ϕ(x, t)|2 = p e 2d (1+∆2 (t)) with ∆(t) = (2.74)
2πd2 (1 + ∆2 (t)) 2md2

Or, even more compactly:

1 (x−v0 t)2 q ℏt
|ϕ(x, t)|2 = √ e− 2σ2 with σ = d 1 + ∆2 (t) and ∆(t) = (2.75)
2πσ 2md2

From (2.74) and (2.75), we can immediately read two facts: Firstly: The maximum of the wave
packet moves with x0 (t) = v0 t to larger x values, which corresponds exactly to the classical
equation of motion x(t) = v0 t for a moving free particle. Secondly: As the wave packet (for
positive v0 ) moves towards larger x values, the variance σ 2 = d2 (1 + ∆2 (t)) increases as time t
progresses, which means that the wave packet “spreads out” in the position representation.

28
Quantum Theory I

2.2.2 Heisenberg’s Uncertainty Principle

Motivation: Thought Experiments on the Uncertainty Principle

In this section, we will demonstrate another fundamental property of quantum physics:
The position and momentum of a particle are so-called “complementary observables”,
which means it is not possible to determine both observables with arbitrary precision.
This is not due to technical inadequacies of the measuring instruments used. Rather, it is
a fundamental property of quantum systems.

One way to understand this phenomenon visually is to illustrate the process of measuring
a location: The most obvious way to determine the location of a particle is to observe
the particle. However, this necessarily means interacting with the particle. Because, even
if you just “look at” a particle, you have to illuminate it – the light interacts with the
particle and changes its momentum. In classical physics, you can make the interaction
arbitrarily small by reducing the light intensity as much as you want. In quantum physics,
however, you are forced to shoot at least one photon with momentum |p| = ℏk = h/λ at
the presumed location of the particle and then see if the photon was deflected. Therefore,
the interaction during the location measurement cannot be reduced arbitrarily! An ex-
perimental physicist might come up with the clever idea of using a photon with a larger
wavelength to reduce the momentum transfer to the particle. This idea has only one flaw:
It also reduces the spatial resolution of the measurement because a photon with a larger
wavelength is less well localized.

But what if we shoot a stream of particles in the z-direction at a plate with a very small
hole? Suppose all particles have a very precisely determined momentum before they hit
the plate. Behind the plate, you can only find those particles that happened to hit the
hole exactly. That would mean that the location (at least in the x and y directions,
i.e., transverse to the direction of flight) is precisely determined and the momentum was
already known before the position measurement! But it turns out: even this attempt to
circumvent the laws of quantum physics fails. Because the location of all particles that
made it through the hole is so precisely known in the transverse direction, the momentum
of these particles in the transverse direction is now particularly uncertain. Each particle
flies away in a different transverse direction after the hole. And this is precisely what we
would expect from the diffraction of a wave at a small opening! This thought experiment
demonstrates that at a fundamental level, a quantum particle does not have a well-defined
momentum once its location is known very precisely.

In the previous section, we determined the standard deviation σ of |ϕ(x, t)|2 in (2.71). It is the
standard deviation of the probability of presence in position space. To emphasize this, we will
call it σx in this section:
(2.71)
q
σx = d 1 + ∆2 (t) (2.76)

We now want to correlate σx with the standard deviation σk of |ϕk (k)|2 in k-space. We already
know the function ϕk (k) from the “Fourier perspective” (2.59). It is important to note that
ϕk (k) does not depend on time because the time evolution in the Fourier integral affects only
the position space. First, we determine |ϕk (k)|2 :
(2.59) 2 d2 2 2 d2
|ϕk (k)|2 = Ae−(k−k0 ) = |A|2 e−2(k−k0 ) (2.77)

By comparing the exponent in (2.77) with the exponent of the normalized Gaussian distribution

29
 2 Schrödinger Equation

(2.70), we find:
1 1 1
2d2 = =⇒ σk2 = 2 =⇒ σk = (2.78)
2σk2 4d 2d
σk does not depend on time, just as ϕ(k) doesn’t. By forming the product of both standard
deviations, we can estimate its minimum size:
1 q 1
σx σk = d 1 + ∆2 (t) ≥ (2.79)
2d | {z } 2
≥1

When transitioning from the standard deviation σk in k-space to the standard deviation σp in
momentum space, an additional factor of ℏ appears in the weight function a(k) from (2.53)
due to p = ℏk. This results in the Heisenberg Uncertainty Principle named after Werner
Heisenberg:

ℏ
σx σp ≥ (2.80)
2

This means: The smaller σx is (i.e., the more precisely the position is determined), the larger σp
must become to satisfy the uncertainty principle (2.80). Therefore, we cannot determine position
and momentum with arbitrary precision. A more accurate measurement of one observable results
in the complementary quantity being determined less precisely.

2.2.3 Expectations of the Wave Packet

It has already been shown that the maximum of the Gaussian wave packet in position represen-
tation changes over time with x0 + v0 t. From this, one can conclude (because of the symmetry
of the Gaussian function) that this will also apply to the expected value ⟨x̂⟩. This will now
be explicitly shown in the formalism of quantum mechanics, by integrating x over the entire
one-dimensional space with the probability density ρ(x, t):
Z +∞ Z +∞
⟨x̂⟩ = dx ρ(x, t)x = dx |ϕ(x, t)|2 x (2.81)
−∞ −∞

The following two remarks apply:

1. The notation ⟨x̂⟩ (with “operator hat”) suggests that even the position measurement cor-
responds to an operator x̂. In position representation, however, the operator with x̂ → x
is quite trivial.

2. If there were only a countable, discrete set of positions xi , at which the particle could
be found with probability Pi (t), one would intuitively calculate the expectation ⟨x̂⟩ as a
weighted probability as follows:

⟨x̂⟩ = Pi (t)xi (2.82)

X

The integral (2.81) is simply the generalization of the sum (2.82) for an uncountable,
continuous set of possible positions.

30
Quantum Theory I

Continuing the calculation started in (2.81):

Z +∞
⟨x̂⟩ = dx |ϕ(x, t)|2 (x − v0 t + v0 t) =
−∞
Z +∞ Z +∞
(2.75)
= dx |ϕ(x, t)|2 (x − v0 t) + v0 t dx |ϕ(x, t)|2 =
−∞ −∞
| {z }
=1 (normalized)
1 +∞ (x−v0 t)2
Z
=√ dx e− 2σ (x − v0 t) + v0 t = | ξ = x − v0 t
2πσ −∞
1
Z +∞
ξ2
=√ dξ e| −{z2σ} · ξ +v0 t = v0 t (2.83)
2πσ 2 −∞ even
|{z}
odd
| {z }
=0

In the last step, we used the fact that an integral with symmetric limits over the product of
an even and an odd function vanishes. As already suspected, the spatial expectation ⟨x̂⟩ = v0 t,
making it clear that the center of the wave packet follows the classical equation of motion and
moves over time in the direction determined by the sign of v0 .

To calculate the momentum expectation value, we use the previously motivated (one-dimensional)
momentum operator p̂ from (2.14):
d ℏ d
p̂ → −iℏ ≡
dx i dx
We showed in (2.13) that this momentum operator acts on a plane wave as follows:
ℏ d i(kx−ωt)
e = p ei(kx−ωt)
i dx
If we multiply the complex conjugate plane wave on the left side, we obtain as a result the
momentum of the individual plane wave:
ℏ d i(kx−ωt)
e−i(kx−ωt) e = e−i(kx−ωt) p ei(kx−ωt) = p
i dx
Now our wave packet ϕ(x, t) according to (2.52) is simply at any time a continuous superposition
of plane waves. Therefore, we can express the expectation value ⟨p̂⟩ again as a weighted average
as follows:
Z ∞
ℏ d
⟨p̂⟩ = dx ϕ∗ (x, t) ϕ(x, t) (2.84)
−∞ i dx
Comparing this with our approach (2.81) for calculating ⟨x̂⟩, we see that we could have also
written (2.81) as follows:
Z ∞
⟨x̂⟩ = dx ϕ∗ (x, t) x ϕ(x, t) (2.85)
−∞

The integrals (2.84) and (2.85) have the same structure, and it turns out that one can calculate
the expectation value ⟨Â⟩ of any observable  for a wave function ϕ(x, t) as follows:
Z ∞
⟨Â⟩ = dx ϕ∗ (x, t) Â{x} ϕ(x, t) (2.86)
−∞

The notation Â{x} is meant to suggest that in the integral, the operator  must be written in
the position basis. This means that, for example, ℏi dx
d
can only be used as a momentum oper-
ator if the wave function is a function of x, and we integrate over x. If the wave function, for

31
 2 Schrödinger Equation

instance, were a function of p, and we were integrating over p, the momentum operator would
look different. We will learn more about this in the next chapter.

Therefore, we use the approach (2.84) and insert the Fourier representation (2.59) of ϕ(x, t):
Z ∞
d
 
∗ (2.59)
⟨p̂⟩ = dx ϕ (x, t) −iℏ ϕ(x, t) =
−∞ dx
1 1
Z ∞ Z ∞ ∗ Z ∞
′2 ℏ d
  2

′ ′ ik′ x −i ℏk ikx −i ℏk
= dx √ dk ϕk (k )e e 2m
t
√ dk ϕk (k)e e 2m
t
=
−∞ 2π −∞ i dx 2π −∞
1 ∞ ∞ ℏk′2 ∞ ℏ d ikx −i ℏk2 t
Z Z  Z 
′
= dx dk ′ ϕ∗k (k ′ )e−ik x ei 2m t dk ϕk (k) e e 2m =
2π −∞ −∞ −∞ i dx
1 ∞
Z ∞  Z ∞
ℏk′2

ℏk2
Z
′ ℏ
= dx dk ′ ϕ∗k (k ′ )e−ik x ei 2m t dk ϕk (k) ikeikx e−i 2m t =
2π −∞ −∞ −∞ i
ℏ ∞
Z ∞ Z ∞
ℏk′2 ℏk2
Z
′
= dx dk ′ dk ϕ∗k (k ′ )ϕk (k) e−ik x eikx ei 2m t ei 2m t k =
2π −∞ −∞ −∞
Z ∞ Z ∞ Z ∞
ℏ ′ ℏ ′2 2
= dx dk ′ dk ϕ∗k (k ′ )ϕk (k) ei(k−k )x ei 2m (k −k )t k =
2π −∞ −∞ −∞
ℏ ∞ ′ ∞
Z Z Z ∞ 
∗ ′ i(k−k′ )x ℏ ′2 2
= dk dk ϕk (k )ϕk (k) dx e ei 2m (k −k )t k =
2π −∞ −∞ −∞
| {z }
2πδ(k−k′ )
Z ∞ Z ∞
ℏ ℏ ′2 −k 2 )t
= dk ′ dk ϕ∗k (k ′ )ϕk (k) 2πδ(k − k ′ )ei 2m (k k=
2π −∞ −∞
Z ∞ ℏ 2 −k 2 )t
Z ∞
=ℏ dk ϕ∗k (k)ϕk (k) ei 2m (k k=ℏ dk |ϕk (k)|2 k (2.87)
−∞ −∞

A comparison of (2.87) with (2.81) shows that the integrals look entirely equivalent. We therefore
attempt a solution strategy analogous to (2.83):
Z +∞
⟨p̂⟩ = ℏ dk |ϕk (k)|2 (k − k0 + k0 ) =
−∞
Z +∞ Z +∞
(2.59)
=ℏ dk |ϕk (k)| (k − k0 ) + ℏk0
2
dk |ϕk (k)|2 =
−∞ −∞
| {z }
=1 (normalized)
Z +∞
2 d2 2
=ℏ dk A e−(k−k0 ) (k − k0 ) + ℏk0
−∞
Z +∞
2 d2
= ℏ|A| 2
dk e−2(k−k0 ) (k − k0 ) + ℏk0 = | ξ = k − k0
−∞
Z +∞
2 2
= ℏ|A|2 dξ e| −2ξ d
· ξ + ℏk0 = ℏk0 = p0 = mv0 (2.88)
−∞
{z } |{z}
even odd
| {z }
=0

In the last line, the same argument as in (2.83) is applied again: The integration with symmetric
limits over the product of an even and an odd function vanishes. The normalization in the second
line is a consequence of the Parseval’s equation:
Z +∞ Z +∞
dx |ϕ(x, t)|2 = dk |ϕk (k, t)|2 (2.89)
−∞ −∞

The normalization of a state ϕ(x, t) remains even after a Fourier transformation to ϕk (k, t) (and
vice versa). Since in (2.74) the position wave function was constructed so that the total inte-
grated probability is equal to one, the probability in momentum space must also lead to the

32
Quantum Theory I

same result.

In summary, one obtains the following results for the expected values of position and momentum:

d ⟨x̂⟩
⟨x̂⟩ = v0 t and ⟨p̂⟩ = mv0 = m
dt
These relations satisfy the equations of motion p = mv = mẋ of classical mechanics. In the
lecture Quantentheorie II (Quantum Theory II), we will see that this property is related to
the Ehrenfest theorem, named after Paul Ehrenfest, which states that for special systems,
the expected values of quantum mechanical observables correspond to the classical quantities of
the equations of motion. This is the case when we have potentials V (x) that include at most
polynomial terms up to the second order in x (i.e., V (x) ∝ x0 , x1 , x2 ). For a potential of the
form V (x) ∝ x3 , the Ehrenfest theorem would no longer be satisfied. In such a case, there would
be discrepancies between quantum mechanical and classical measured quantities. In the case of
our wave packet, we solved the Schrödinger equation for a free particle (V (x) = 0), making the
Ehrenfest theorem valid.

2.3 Particle in a Potential Well

Motivation: A Simple Model for Bound Particles Reveals Astonishing Quan-
tum Properties
The term potential well refers to a spatial region where a potential V (r) or, in the one-
dimensional case, V (x) has a local minimum. Earth’s gravitational field is an example of
such a potential well in classical physics. The potential of the strong nuclear force, which
causes protons to remain together in the atomic nucleus, is an example of a potential well
in quantum physics.

In this section, we consider two simple one-dimensional, model potential wells: the in-
finitely deep potential well and the finitely deep potential well. Both are suitable for
demonstrating essential differences between classical physics and quantum physics.

The “infinitely deep potential well” is formed by a potential that is only zero within a
limited spatial region and infinite otherwise. One can imagine a quantum particle being
perfectly enclosed in this potential well. However, if a quantum particle is confined to
a certain spatial region, its position is to some extent determined: It is certainly inside
the potential well and not outside. As we previously explained, this means its momen-
tum is no longer precisely determinable! That is, in every measurement, we would find
that the particle would have a certain momentum in the negative or positive x-direction.
For a quantum particle, it is utterly impossible to simply “lie completely motionless” on
the bottom of a potential well! It always has a minimum kinetic energy, and thus (un-
like a classical particle) a so-called ground state energy that is non-zero. Additionally, we
will see that (unlike in classical physics) only discrete (quantized) energy states are allowed.

Similar effects can be observed (although with different numerical values) in the finitely
deep potential well, where the potential is again zero within a limited spatial region but
otherwise takes on a finite positive value. In such a finitely deep potential well, another
effect can be observed that does not exist in classical physics: A particle trapped in the
well can penetrate into the walls of the potential well!

33
 2 Schrödinger Equation

2.3.1 Convexity Relations

Based on the stationary Schrödinger equation and the form of the potential, certain statements
about the characteristic of the wave function ψ(x) can be made. In the following, the time-
independent, one-dimensional Schrödinger equation for stationary states is considered further:

d2 2m
ψ(x) = 2 (V (x) − E) ψ(x)
dx2 ℏ
The left side of the differential equation is related to the kinetic energy (since the position
representation of the quadratic momentum operator is used here and p̂2 ∝ Êkin applies) and
can be used within the framework of an analysis of the wave function to describe the curvature
behavior of ψ(x).

ψ>0
ψ

ψ<0
x

Abb. 7: Curvature behavior of wave functions at different energies. While the red function
corresponds to E > V (x), the blue functions concern the case E < V (x).

It is distinguished whether the wave function curves towards or away from the x-axis. This is
determined at fixed E by the energy of the potential and the sign of the wave function (assumed
here to be a real function); in other words, the cases in which the potential is greater or less than
the energy of the particle are distinguished. Specifically, the following cases can be distinguished:

a.) V (x) < E: If the energy of the particle exceeds the respective potential, this is referred to
as a classically allowed region, and the wave function curves towards the axis. Whether
the wave function is concave or convex is determined by the sign of ψ(x):
1.) ψ(x) > 0: One obtains a concave wave function, the curvature occurs in the direction
of the axis.
2.) ψ(x) < 0: One obtains a convex wave function. If one examines the transition
between negative and positive wave function, a point of inflection is found at this
point.

b.) V (x) > E: The energy of the potential exceeds that of the particle, and this is referred
to as a classically forbidden region. However, as will be shown in the following chapters,
the probability of stay is not zero even in such regions. The convexity behavior is again
determined by the sign of the wave function:
3.) ψ(x) > 0: One obtains a convex wave function.
4.) ψ(x) < 0: One obtains a concave wave function.

34
Quantum Theory I

2.3.2 Symmetry Relations

In addition to the energy of the system and the sign of the wave function, the actual form
of the potential V (r) can also be examined. Particularly interesting is the case in which the
potential is completely symmetric, as this property directly carries over to the wave functions
as solutions of the Schrödinger equation. That is, if we solve the Schrödinger equation with a
symmetric potential V (r), we will obtain wave functions ψ(x) that are either symmetric (even)
or antisymmetric (odd):

ψ(x) = ψ(−x): The wave function is symmetric or even.

ψ(x) = −ψ(−x): The wave function is antisymmetric or odd.

The reason why the symmetry properties of the potential V (r) influence the wave functions in
this manner will be explained more rigorously in later chapters and in the lecture Quantentheorie
II (Quantum Theory II).

2.3.3 Infinitely Deep Potential Well

Instead of considering a free particle, an interaction with a one-dimensional potential V (x) is
introduced for the first time. This has the simple form:

0, 0≤x≤L
(
V (x) = (2.90)
∞, otherwise

The potential well has a width L and is bounded by infinitely high potential walls. Inside the
well, V (x) = 0, and the Schrödinger equation simplifies to (2.9). However, the limitation due to
the potential form now provides boundary conditions. The wave function ψ(x) cannot penetrate
into the infinitely high walls. This physical condition, therefore, suggests Dirichlet boundary
conditions:
ψ(0) = ψ(L) = 0 (2.91)
To determine the time-independent eigenfunctions, we can choose as a solution approach to the
Schrödinger equation a superposition of a left-traveling and a right-traveling plane wave:

ψ(x) = Ae+ikx + Be−ikx (2.92)

The wavenumber k can be associated with energy using the de Broglie formalism: k 2 = 2mE/ℏ2 .
Substituting into (2.9), one obtains:

−ℏ2 ∂ 2 ψ(x) −ℏ2 ∂ 2  +ikx  −ℏ2 

ikx (2.92)

= Ae + Be ikx
= (−k 2
) Ae+ikx
+ Be = Eψ(x)
2m ∂x2 2m ∂x2 2m
The wave function ψ(x) vanishes at the boundaries of the potential, allowing the constants A,
B, and k to be determined. It holds:

ψ(0) = A + B = 0 =⇒ A = −B

At x = L, the wave function also vanishes; we immediately use the result A = −B:
  nπ
ψ(L) = A e+ikL − e−ikL = 2iA sin(kL) = C sin(kL) = 0 =⇒ kn =
L
The sine function vanishes only when the argument kL takes whole multiples of π, where n ∈ N+
(even n = 0 is not a valid solution, as in this case ψ(x) would not be normalizable). We denote n
as a quantum number - this means that n determines the microscopic state of a quantum system
but can only take discrete values. The constant C must now be determined by normalizing ψ(x):

35
 2 Schrödinger Equation

The particle is with absolute certainty inside the potential well, as it cannot penetrate into its
walls. The probability of stay is therefore:

1 ! 1
Z L Z L
L
= dx sin (kn x) =
2
dx [1 − cos(2kn x)] =
C2 0 2 0 2
The wave function that describes the particle in the potential well is thus given by the following
real function:
2
r
nπ
 
ψn (x) = sin x (2.93)
L L
From (2.93) and the graphically represented probability of stay in Figure 8, the convexity be-
havior of the wave function can also be recognized: For the infinitely high potential well, only a
classically allowed region is possible, so ψ(x) always curves towards the axis.
The quantized values of kn are a consequence of the boundary conditions and accordingly trans-
late to the energies of the stationary states n:

ℏ2 k 2 ℏ2 π 2 2
En = = n (2.94)
2m 2mL2
The energy En thus increases quadratically with n, and energy levels can only exist at permitted
n values.

The set of these eigenstates ψn (x) forms a complete orthonormal system. The orthogonality of
the wave functions will be demonstrated in the following:

2 L πnx πmx
Z    
ψn (x)ψm (x) = dx sin sin =
L 0 L L
1 L π(n − m)x π(n + m)x
Z     
= dx cos − cos =
L 0 L L
L
1 L sin(π(n − m)x/L) L sin(π(n + m)x/L)

= − =
L π(n − m) π(n + m) 0
sin(π(n − m)) sin(π(n + m))
= −
π(n − m) π(n + m)

This relation vanishes for all n ̸= m, as each sine function involves a natural multiple of π - only
if n = m the first term must be further considered. Since in this case, zero is divided by zero,
L’Hôpital’s rule can be applied (let it be x = n − m):
d
sin(πx) π cos(πx)
lim dx
d
= lim =1
x→0
dx πx
x→0 π

The product only does not vanish if the indices are equal and thus represent the same wave
function, allowing us to speak of orthogonality. Due to the fact that ψn (x) is also normalized,
it even holds orthonormality:
ψn (x)ψm (x) = δnm (2.95)
The number of zero crossings ñ0 of ψn (x) can be used to establish an energetic hierarchy of the
system (nodel theorem). The larger ñ0 becomes, the higher the frequency of the eigenstate and
thus its respective energy. The principal or energy quantum number n can be determined over:

n = ñ0 + 1 (2.96)

Counting the zero crossings (or nodes) of the wave function, we can thus determine the quantum
number n of a state in an infinitely deep potential well.

36
Quantum Theory I

2.3.4 Finitely Deep Potential Well

Again, a particle is to be trapped in a one-dimensional potential well, but this time it has a
finite depth:
0,
(
−L ≤ x ≤ L
V (x) = (2.97)
V0 > 0, otherwise
Two cases can generally be distinguished here: The energy of the particle E exceeds or falls below
the energy of the potential V0 . The case where E > V0 will not be addressed in the following, as
it leads to no bound state but to a non-normalizable scattering state. When E < V0 , the particle
cannot leave the potential well, and there exist normalizable eigenstates at discrete energies.
Unlike in (2.91), the wave function of the particle can now penetrate the finite potential barrier
- to find a solution, the entire space can be divided into three regions: ψI (x) in (−∞, −L),
ψII (x) in [−L, L], and ψIII (x) in (L, ∞). The wave function ψII (x) corresponds to that of a free
particle, and the wave number k yields the known de Broglie relation:
√
2mE
k= (2.98)
ℏ
For the wave functions ψI (x) and ψIII (x), the finite potential needs to be considered in both
cases in the same way. The Schrödinger equation corresponds in this case to:
d2 2m
2
ψI,III (x) = 2 (V0 − E) ψI,III (x)
dx ℏ
The energy terms on the right side of the equation can be replaced by a constant, which is cer-
tainly a positive quantity due to E < V0 - for solving the differential equation, a real exponential
function is chosen. For the constant κ, it holds:

2m(V0 − E)
p
κ= (2.99)
ℏ
A general approach for a solution can then be given:

−κx + A e+κx , x < −L
 A1 e 2


ψ(x) = B1 + B2
e−ikx e+ikx , −L ≤ x ≤ L (2.100)

C e−κx + C e+κx ,

L<x
1 2

For the wave functions ψI (x) and ψIII (x) in the potential, it must hold that A1 = C2 = 0, as the
wave function must not grow exponentially (otherwise it would no longer be normalizable). At
the junctions x = −L and x = L, however, it must hold:
ψI (−L) = ψII (−L) and ψII (L) = ψIII (L)
(2.101)
ψI′ (−L) = ψII
′ ′
(−L) and ψII ′
(L) = ψIII (L)
The transition should occur without discontinuities and be continuously differentiable.
A further simplification of the problem can be made by incorporating the symmetry of the
potential. The potential in (2.97) is symmetric around x = 0, which results in the entire
Hamiltonian operator Ĥ exhibiting these symmetry properties. In later chapters, it will be
shown that the symmetry of the Hamiltonian operator impacts the eigenstates: only symmetric
or antisymmetric eigenstates can exist, making it clear that A2 = C1 = A must hold. For the
symmetric or even solution, it generally holds f (x) = f (−x), allowing the wave function ψ+ (x)
to be written as: 
+κx , x < −L
Ae


ψ+ (x) = B cos(kx), −L ≤ x ≤ L

Ae−κx ,

L<x

37
 2 Schrödinger Equation

L = 5 nm V0 = 1 eV, L = 5 nm
n=4
|ϕ|2

|ϕ|2
n=3
n=2

n=2

n=1

n=1

x x

Abb. 8: (left) Infinitely deep potential well with the probabilities of stay of stationary states
with the four lowest eigenenergies. (right) The finite potential well has, for a given
potential depth V0 and width L, only a finite number of bound states, whose proba-
bilities of stay are depicted here.

For the antisymmetric or odd solution, it generally holds f (x) = −f (−x), resulting in the
waveform ψ− (x): 
+κx ,
−Ae x < −L


ψ− (x) = B sin(kx), −L ≤ x ≤ L

Ae−κx ,

L<x
The boundary conditions allow us to express A through B; B can later be determined through
normalization conditions:

ψII,+ (L) = ψIII,+ (L) ⇐⇒ B cos(kL) = Ae−κL thus A = B cos(kL)e+κL

Additionally, the differential boundary condition allows us to establish a connection between k

and κ: ′
ψII,+ (L) ′
ψIII,+ (L) −k sin(kL) −κAe−κL
= ⇐⇒ = so k tan(kL) = κ
ψII,+ (L) ψIII,+ (L) cos(kL) Ae−κL
Through the given parameters E and V0 , constraints for k and κ can be defined - only certain
(but arbitrarily many) pairs of k and κ are possible through V0 :
2m
k 2 + κ2 = V0 = η 2
ℏ2
Thus, one has two conditions available for k and κ to calculate concrete values. For even
eigenfunctions, it thus holds:

k tan(kL) = κ and k 2 + κ2 = η 2 (2.102)

Depending on the depth V0 and width L of the potential, only finitely many pairs of k and κ
can be found. For odd eigenfunctions, one obtains a similar relationship:

−k cot(kL) = κ and k 2 + κ2 = η 2 (2.103)

Both (2.102) and (2.103) provide solutions for k and κ only after numerical (or graphical)
evaluation. Unlike with the square dependence on n in the infinite potential well from (2.94), the

38
Quantum Theory I

eigenenergies follow no clear progression of a quantum number n. The wider the potential well
becomes, the smaller the spacing between energy levels, and the number of possible eigenstates
increases. This is also the case when V0 increases. A calculated example of the probability
density of a particle in two possible states is given in Figure 8. The graphical ordering of the
wave function is justified by the node count (or in the case of the probability density by the zero
points) according to (2.96).

2.4 Particles at the Potential Barrier

Motivation: Scattering Phenomena and Tunnel Effect
While we dealt in the previous sections with potential wells where a particle is bound in a
specific spatial region, we now examine the effects that occur when particles coming from
infinity encounter simple potential barriers or potential steps.

In the following sections, not only will important mathematical tools for dealing with
such cases (such as the transfer and scattering matrix) be presented, but the astonishing
phenomenon of the tunnel effect will also be described. In this phenomenon, a particle
– despite not having sufficient energy according to the laws of classical physics – can
penetrate a potential barrier (in other words, “tunnel through”).

2.4.1 Particles at the Potential Step

Let us consider, as the simplest case, a one-dimensional potential V (x) with the energy height
V0 , which has the following step form (see Fig. 9, left):

0, x < 0 (Region I)
(
V (x) = (2.104)
V0 , x ≥ 0 (Region II)

It is assumed that the kinetic energy satisfies E > V0 (this is referred to as superbarrier scat-
tering). The entire available space can be divided into two subregions: In Region I (−∞, 0) the
wave function ψI (x) holds, and in Region II [0, +∞) the wave function ψII (x). Assuming an
incoming wave from the left, we expect it to be partially reflected, meaning ψI (x) is described by
both an incoming and an outgoing wave. On the other hand, Region II with the wave function
ψII (x) will only be characterized by an outgoing wave. We therefore choose the Ansatz:

ψI (x) x < 0 ψI (x) = Ae+ikx + Be−ikx

(
ψ(x) = with ′ (2.105)
ψII (x) x ≥ 0 ψII (x) = Ce+ik x

Anticipating the next step, we calculate the first spatial derivative of ψI (x) and ψII (x):

ψI′ (x) = ikAe+ikx − ikBe−ikx

′ ′ (2.106)
ψII (x) = ik ′ Ce+ik x

The wave number k is equivalent to the expression already found in (2.98), while for k ′ , from
the Schrödinger equation in analogy to (2.99), the following relation is obtained:
s
2m(E − V0 )
k′ = (2.107)
ℏ2

The boundary conditions (continuity of the wave function and its derivative) are as follows:

ψI (0) = ψII (0) and ψI′ (0) = ψII

′
(0) (2.108)

39
 2 Schrödinger Equation

From the first of these conditions in (2.108), a direct connection between the amplitudes of the
wave functions is derived:
(2.105)
ψI (0) = ψII (0) =⇒ A + B = C (2.109)
The differential boundary condition from (2.108) provides another relation:

(2.106) k′
ψI′ (0) = ψII
′
(0) =⇒ ikA − ikB = ik ′ C ⇒ A−B = C (2.110)
k
The amplitudes B and C are to be expressed in terms of the amplitude A of the incoming wave.
Inserting (2.109) into (2.110) yields:

k′ k − k′
A−B = (A + B) ⇒ kA − kB = k ′ A + k ′ B ⇒ B= A (2.111)
k k + k′
By substituting (2.111) into the first boundary condition (2.109), an expression for the amplitude
C of the outgoing wave is obtained:

k − k′ k + k′ + k − k′ 2k
C =A+ A = A ⇒ C= A (2.112)
k + k′ k + k′ k + k′
Since a plane wave over the entire space cannot be normalized, A remains undetermined. The
solution to the scattering problem can thus be represented by inserting (2.111) and (2.112) into
the Ansatz (2.105) as follows:

k − k ′ −ikx
(
ψI (x) x<0 ψI (x) = Ae+ikx + A e
ψ(x) = with k + k′ (2.113)
ψII (x) x≥0 2k +ik′ x
ψII (x) = A e
k + k′
How can we understand the wave function ψ(x) from (2.113)? In the region x < 0, we deal
with the superposition of a wave moving to the right (incoming) and a wave moving to the left
(reflected).
Re(ψ(x))

Re(ψ(x))

3 3
E= V
2 0
E= V
2 0
V0 V0

1
E= V
2 0

x x

Abb. 9: (left) A particle wave incoming from the left interacts with a positive potential step,
and depending on the energy E, either undergoes a frequency reduction or exponential
decay. (right) Viewing a negative potential step, the frequency increases.

Note, however, that the direction of motion of the plane waves becomes clear only in the context
of the time-dependent Schrödinger equation, which according to (2.36) provides a phase factor

40
Quantum Theory I

e−iωt (thus ei(kx−ωt) is a wave moving to the right). The amplitude of the reflected, left-moving
part depends on the ratio of the energy E to the step height V0 . In the region of the potential
(x ≥ 0), there is a change in the frequency of the non-reflected component of the plane wave.
Depending on the sign of the potential V0 , there is either an increase or decrease. With the wave
vector k ′ from (2.107), one obtains:
c c ′ c q
ν′ = = k = 2m(E − V0 ) (2.114)
λ′ 2π 2πℏ

2.4.2 Transfer and Scattering Matrix

Let us consider a one-dimensional potential barrier V (x) with finite width L and height V0 (see
Fig. 10, left): 
0, x<0 (Region I)


V (x) = V0 , 0 ≤ x ≤ L (Region II) (2.115)

0,

x>0 (Region III)
In this example, we assume that E < V0 , which causes an exponential decay of the wave function
in the barrier (this is also known as sub-barrier scattering). The space is again divided into three
distinct regions: Region I (−∞, 0) with ψI (x), Region II [0, L] with ψII (x), and Region III (L, ∞)
with ψIII (x). As in the previous chapter, the constants k and κ can be determined from the
Schrödinger equation, which also match in this case with (2.98) and (2.99):
√
2mE 2m(V0 − E)
p
k= and κ =
ℏ ℏ
The following solution approach to the Schrödinger equation for the potential barrier is chosen,
assuming this time that there can be incoming, and therefore also reflected (outgoing) waves
both from the left (in Region I) and from the right (in Region III):

ψI (x) = Ae+ikx + Be−ikx


ψI (x),

 x<0
ψ(x) = ψII (x), 0 ≤ x ≤ L with ψII (x) = Ce−κx + De+κx (2.116)

ψ (x), x > L −ikx
ψIII (x) = F e+ikx
+ Ge

III

Again, we calculate the first spatial derivative of ψI , ψII , and ψIII for later use:

ψI′ (x) = ikAe+ikx − ikBe−ikx

′
ψII (x) = −κCe−κx + κDe+κx (2.117)
′ −ikx
ψIII (x) = ikF e +ikx
− ikGe

The following boundary conditions apply:

ψI (0) = ψII (0) and ψII (L) = ψIII (L)

(2.118)
ψI′ (0) = ψII
′ ′
(0) and ψII ′
(L) = ψIII (L)

The boundary conditions for x = 0 lead to:

(2.116)
(I)| ψI (0) = ψII (0) =⇒ A + B = C + D
(2.117) κ
(II)| ψI′ (0) = ψII
′
(0) =⇒ ik(A − B) = κ(D − C) =⇒ A − B = (D − C)
ik

41
 2 Schrödinger Equation

The addition of both equations, or the subtraction of the second equation from the first, respec-
tively gives:
κ ikC + ikD + κD − κC ik − κ ik + κ
(I)+(II)| 2A = C + D + (D − C) = = C+ D =⇒
ik ik ik ik
1
A= [(ik − κ)C + (ik + κ)D] (2.119)
2ik
κ ikC + ikD − κD + κC ik + κ ik − κ
(I)–(II)| 2B = C + D − (D − C) = = C+ D =⇒
ik ik ik ik
1
B= [(ik + κ)C + (ik − κ)D] (2.120)
2ik
This can be written more compactly as a matrix equation:

1 1
! ! ! ! ! !
A ik − κ ik + κ C k + iκ k − iκ C C
= = =P (2.121)
B 2ik ik + κ ik − κ D 2k k − iκ k + iκ D D
| {z }
P

Here, the matrix P is a transfer matrix, which immediately relates the amplitudes in Region I
to the amplitudes in Region II. For the boundary condition at x = L, the relations are a bit
more complicated due to the non-vanishing exponential function. To simplify the next step, the
first equation is first multiplied by κ:
(2.116)
(III)| κψII (L) = κψIII (L) =⇒ κCe−κL + κDe+κL = κF e+ikL + κGe−ikL
′ ′ (2.117)
(IV)| κψII (L) = κψIII (L) =⇒ − κCe−κL + κDe+κL = ikF e+ikL − ikGe−ikL

The addition of both equations, or the subtraction of the second equation from the first, gives:

(III)–(IV)| 2Cκe−κL = (κ − ik)F e+ikL + (κ + ik)Ge−ikL

(III)+(IV)| 2Dκe+κL = (κ + ik)F e+ikL + (κ − ik)Ge−ikL

By simple rearrangement, one obtains the following expressions for C and D:

κ − ik +ikL+κL κ + ik −ikL+κL
C= Fe + Ge
2κ 2κ
κ + ik +ikL−κL κ − ik −ikL−κL
D= Fe + Ge
2κ 2κ
Here too, the equation system can be more clearly presented in matrix form:

1 (κ − ik)e+L(ik+κ) (κ + ik)e−L(ik−κ)
! ! ! !
C F F
= =Q (2.122)
D 2κ (κ + ik)e+L(ik−κ) (κ − ik)e−L(ik+κ) G G
| {z }
Q

The matrix Q is also a transfer matrix. It relates the amplitudes in Region II to the amplitudes
in Region III. Transfer matrices have the practical property that they can be easily multiplied
to obtain a superior transfer matrix. Thus, if you multiply the transfer matrices P and Q, you
obtain a new transfer matrix M , which describes the overall relationship between the amplitudes
in Region I (left of the barrier) and the amplitudes in Region III (right of the barrier):
! ! ! ! !
A F F M11 M12 F
= PQ =M = (2.123)
B |{z} G G M21 M22 G
M

A transfer matrix therefore always relates the amplitudes on the right to the amplitudes on the
left of a scatterer. If you already know the amplitudes F and G of the outgoing and incoming

42
Quantum Theory I

wave on the right side of the barrier, then the transfer matrix M can be used to calculate the
amplitudes A and B of the incoming and outgoing wave on the left side of the barrier. With the
matrix M −1 , the calculation is possible in reverse direction. However, this is obviously not very
helpful for most problem settings. If you want to use the transfer matrix M , for example, you
must already know the amplitude of the outgoing wave on the right side. However, this will be
somehow derived from the reflection of the incoming wave from the right and the transmission of
the incoming wave from the left. It would therefore be better if there were a matrix that could
calculate the unknown amplitudes of the outgoing waves on the left and right (B and F ) from
the known amplitudes of the incoming waves on the left and right (A and G). Such a matrix is
called a scattering matrix. We want to denote it with S:
! ! ! !
B A S11 S12 A
=S = (2.124)
F G S21 S22 G

We will now attempt to express the components Sij of the scattering matrix S in terms of the
components Mij of the transfer matrix M . For this, we first write the matrix equation (2.123)
of the transfer matrix M as a system of equations:

A = M11 F + M12 G (2.125)

B = M21 F + M22 G (2.126)

In contrast, the matrix equation (2.124) of the scattering matrix S corresponds to the following
system of equations:

B = S11 A + S12 G (2.127)

F = S21 A + S22 G (2.128)

Our goal should be to express the amplitudes B and F as a function of the amplitudes A and G,
but only using the matrix entries Mij . For F , this is simple: We just need to rearrange equation
(2.125) accordingly:

1 M12
F = A− G (2.129)
M11 M11
By comparing coefficients between (2.129) and (2.128), we can immediately deduce:

1 M12
S21 = and S22 = − (2.130)
M11 M11
Finally, inserting the relation (2.129) into equation (2.126), we obtain:

1 M12 M21 M21 M12

 
B = M21 A− G + M22 G = A− G + M22 G =⇒
M11 M11 M11 M11
M21 M22 M11 − M21 M12
B= A+ G (2.131)
M11 M11
A further coefficient comparison between (2.131) and (2.127) finally gives the following expres-
sions for the matrix components S11 and S12 :
M21 M22 M11 − M21 M12
S11 = and S12 = (2.132)
M11 M11
For the scattering matrix S, expressed in the components of the transfer matrix M , the following
holds according to (2.130) and (2.132):

43
 2 Schrödinger Equation

1
!
M21 M22 M11 − M21 M12
S= (2.133)
M11 1 −M12

2.4.3 Transmission and Reflection Probability

We will now consider how large the reflection and transmission probability is for an incoming
particle at a potential barrier according to (2.115), depicted in Fig. 10 (left). To do this,
let’s assume that a particle approaches the potential barrier from the left. There is a certain
probability of transmission or reflection. Because we assume that the particle is only incoming
from the left (but no particle from the right), we can set the amplitude G of the wave incoming
from the right in the Ansatz (2.116) to G = 0. Doing this, the equations (2.127) and (2.128)
simplify as follows:
B = S11 A (2.134)
F = S21 A (2.135)
The simplified scattering equation (2.135) allows us to define and calculate the complex-valued
transmission coefficient for transmission from left to right tL→R as the ratio between the ampli-
tude F of the right outgoing wave and the amplitude A of the left incoming wave. The square
of the magnitude of this quantity leads to the transmission probability TL→R :
F (2.135) (2.130) 1 2
tL→R = = S21 and TL→R = |tL→R |2 = |S21 |2 = (2.136)
A M11
The complex-valued reflection coefficient rL for particles incoming from the left is instead defined
as the ratio between the amplitude B of the left outgoing wave and the amplitude A of the left
incoming wave. Using equation (2.134), an expression can also be found for this. In a manner
equivalent to before, the reflection probability RL can also be calculated:
2
B (2.134) (2.132) M21
rL = = S11 and RL = |rL |2 = |S11 |2 = (2.137)
A M11
For the transmission coefficient tL→R ∈ C, it holds that ψIII (x) = tL→R A e+ikx (in the case of
an exclusively left incoming wave). The transmission probability TL→R ∈ R must lie within the
interval TL→R ∈ [0, 1].

Now let’s assume that a particle exclusively hits the potential barrier from the right. This
particle will either be transmitted from right to left or reflected back to the right. Since we
assume that this time no particle is incoming from the left, we can now set the amplitude A = 0
in the Ansatz (2.116). This simplifies the equations (2.127) and (2.128) as follows:
B = S12 G (2.138)
F = S22 G (2.139)
The transmission coefficient from right to left tR→L is the ratio between the amplitude B of the
left outgoing wave and the amplitude G of the right incoming wave. The associated transmission
probability TR→L is the square of the magnitude of tR→L . Both can be calculated using the
relation (2.138):
B (2.138)
tR→L = = S12 and TR→L = |tR→L |2 = |S12 |2 (2.140)
G
The reflection coefficient from the right rR is finally the ratio between the amplitude F of the
right outgoing wave and the amplitude G of the right incoming wave. We can calculate it (and
the associated reflection probability RR ) using equation (2.139).
F (2.139)
rR = = S22 and RR = |rR |2 = |S22 |2 (2.141)
G
Thus, we can also write the scattering matrix S as follows:

44
Quantum Theory I

!
rL tR→L
S= (2.142)
tL→R rR

2.4.4 Probability Current Density in Scattering

Since V (x) is purely real, there is no absorption or amplification within the potential. Therefore,
the probability current density j(x) must also be preserved during transmission through the
potential barrier. For j(x), it holds according to (2.32) in the one-dimensional, time-independent
case:
ℏ d
 
∗
j(x) = Re ψ ψ (2.143)
im dx
Let’s assume a matter wave incoming from the left, which scatters at the barrier. It will be
partially reflected back to the left and will partially run out to the right. There should be
no wave incoming from the right. We can therefore set the amplitude G = 0 in our Ansatz
(2.116). Additionally, we can express the amplitudes B and F in terms of the amplitude A of
the incoming wave using the relations (2.134) and (2.135). We then obtain for ψI (x) and ψIII (x):
(2.134)
ψI (x) = Ae+ikx + Be−ikx =⇒ ψI (x) = Ae+ikx + S11 Ae−ikx
(2.144)
ψIII (x) = F e+ikx =⇒ ψIII (x) = S21 Ae
(2.135) +ikx

Without loss of generality, we now additionally assume that the incoming wave has an amplitude
A = 1. This results in:

ψI (x) = e+ikx + S11 e−ikx and ψIII (x) = S21 e+ikx (2.145)

V0
Re(ψ(x))

x x

Abb. 10: (left) Transmission through a rectangular potential barrier. While the amplitude
significantly decreases after transmission, there is no change in the particle’s energy
in this case. (right) Transmission of an α-particle, bound by the strong nuclear force,
through the Coulomb barrier. Besides the decrease in amplitude, there is a reduction
in the particle’s kinetic energy.

45
 2 Schrödinger Equation

The probability current density on the left side of the barrier yields:
ℏ d
 
(2.143) (2.145)
jI (x) = Re ψI∗ ψI =
im dx
 ∗ ℏ d  
= Re e+ikx + S11 e−ikx e+ikx + S11 e−ikx =
im dx
 
 ℏ
= Re e−ikx + S11
∗ +ikx
e ike+ikx − ikS11 e−ikx =
im
ℏk h i
= Re 1 − S11 e−2ikx + S11
∗ +2ikx
e ∗
− S11 S11 =
m | {z } | {z } | {z }
α α∗ |S11 |2
ℏk h i
= Re 1 − (α − α∗ ) − |S11 |2 α − α∗ = 2i Im(α)
m
ℏk h i
= Re 1 − 2i Im (α) − |S11 |2 =
m
ℏk   (2.137) ℏk
= 1 − |S11 |2 = (1 − RL ) (2.146)
m m
On the right side of the barrier, with analogous calculations, we find:
ℏ d
 
(2.143) ∗ (2.145)
jIII (x) = Re ψIII ψIII =
im dx
 ∗ ℏ d  
= Re S21 e+ikx S21 e+ikx =
im dx
 
∗ −ikx ℏ ℏk h ∗ i
= Re S21 e ikS21 e+ikx = Re S21 S21 e−ikx e+ikx =
im m
ℏk (2.136) ℏk
= |S21 |2 = TL→R (2.147)
m m
Since the probability current density must also be conserved in a scattering process due to
particle conservation, it holds that jI (x) = jIII (x), and thus:
(2.146)
(2.147) ℏk ℏk
jI (x) = jIII (x) =⇒ (1 − RL ) = TL→R =⇒ RL + TL→R = 1 (2.148)
m m
The calculation carried out above can of course also be done for a particle incident exclusively
from the right, leading to an analogous result for RR and TR→L :

1 = RR + TR→L (2.149)

Due to the functional dependence of RL and TL→R on the components M11 and M21 of the
transfer matrix according to (2.136) and (2.137), it is possible in the case of resonance that
TL→R = 1 and RL = 0, meaning the particle is transmitted with a hundred percent probability
and not reflected. This is the case if M21 = 0. Contrary to classical intuition, however, for a
finite potential barrier, it must always be the case that T ̸= 0, because M11 must not vanish,
since in this case T would diverge according to (2.136). Therefore, there is always a certain
transmission probability with a finite potential barrier. In classical physics, however, the case
R = 1 is indeed possible at E < V0 . The argument presented can be equally applied to RR and
TR→L .

From the conservation of probability current density, it follows also that the scattering matrix
is unitary:
S −1 = S † ⇐⇒ SS † = 1 (2.150)

46
Quantum Theory I

The proof of (2.150) follows through the conservation of probability current density on the left
and right side of the barrier, starting from the Ansatz (2.116). By substituting (2.116) into
(2.143), it follows:
ℏ d
 
(2.143) (2.116)
jI (x) = Re ψI∗ ψI =
im dx
 ∗ ℏ d  
−ikx −ikx
= Re Ae +ikx
+ Be Ae + Be =
+ikx
im dx
  ℏ  
= Re A∗ e−ikx + B ∗ e+ikx ikAe+ikx − ikBe−ikx =
im
ℏk h ∗ −ikx  i
= Re A e + B ∗ e+ikx Ae+ikx − Be−ikx =
m
ℏk h ∗ i
= Re A A − B ∗ B + AB ∗ +2ikx
e − A ∗
Be −2ikx
} =
m | {z
α
} | {z
∗ α
ℏk h 2 i
= Re |A| − |B|2 + (α − α∗ ) = α − α∗ = 2i Im(α)
m
ℏk h 2 i
= Re |A| − |B|2 + 2i Im (α) =
m
ℏk  2 
= |A| − |B|2 (2.151)
m

ℏ d
 
(2.143) ∗ (2.116)
jIII (x) = Re ψIII ψIII =
im dx
 ∗ ℏ d  
= Re F e+ikx + Ge−ikx F e+ikx + Ge−ikx
=
im dx
  ℏ  
= Re F ∗ e−ikx + G∗ e+ikx ikF e+ikx − ikGe−ikx =
im
ℏk h  i
= Re F ∗ e−ikx + G∗ e+ikx F e+ikx − Ge−ikx =
m
ℏk h ∗ i
= Re F F − G∗ G + F G ∗ +2ikx
e − F ∗
Ge −2ikx
} =
m | {z
α
} | {z
∗ α
ℏk h 2 i
= Re |F | − |G|2 + (α − α∗ ) = α − α∗ = 2i Im(α)
m
ℏk h 2 i
= Re |F | − |G|2 + 2i Im (α) =
m
ℏk  2 
= |F | − |G|2 (2.152)
m
It must hold that jI (x) = jIII (x). Therefore, we find:
(2.151)(2.152)
jI (x) = jIII (x) =⇒
ℏk  2  ℏk  
|A| − |B|2 = |F |2 − |G|2 =⇒
m m
|B| + |F | = |A|2 + |G|2
2 2
(2.153)

The amplitudes A and G correspond to the waves incoming on the left and right sides of the
barrier, and the amplitudes B and F correspond to the waves outgoing on the left and right
sides. We can therefore define the amplitude vectors ain and aout :
! !
A B
ain = and aout = (2.154)
G F

47
 2 Schrödinger Equation

† †
The left side of Equation (2.153) can thus be expressed as aout aout , and the right side as ain ain .
Additionally, we can write the scattering matrix equation (2.124) compactly as follows:

aout = S ain (2.155)

By expressing Equation (2.153) with the help of the amplitude vectors (2.155), and then using
(2.155):

† † (2.155)
aout aout = ain ain =⇒
†
(Sain )† Sain = ain ain =⇒
† †
ain S † S ain
|{z} = ain ain (2.156)
1

In order for Equation (2.156) to be satisfied, it must hold that:

S†S = 1 □

Thus, we have proven the unitarity of the scattering matrix as postulated in (2.150).

2.4.5 Tunneling Effect

We have already shown in (2.123) that we can represent the transmission matrix M of a potential
barrier as a product of the matrices P and Q, which we have already determined in (2.121) and
(2.122). We now want to calculate the transmission probability through a barrier of length
∆x = L. As already discussed, the penetration of the barrier (classically impossible) in the case
E < V0 , as shown in Figure 12, is not forbidden. We call this phenomenon the tunneling effect.
In the case of E > V0 , transmission is also classically permitted, resulting in a classical T = 1.
In quantum physics, however, reflection R > 0 also occurs (except in resonance cases), resulting
in transmission below one (Fabry-Pérot interferences).

Example: Transmission Probability through Barrier

The transmission probability of a particle incident from the left through a potential barrier
is given by (2.136):
1
TL→R =
|M11 |2
The potential barrier has height V0 and the particle has energy E < V0 . Since we are only
interested in the entry M11 of the transfer matrix M = P Q to be calculated according to

48
Quantum Theory I

(2.123), we need not calculate the entire matrix M , but only the entry M11 :
(2.121)(2.122)
M11 = P11 Q11 + P12 Q21 =
k + iκ κ − ik L(ik+κ) k − iκ κ + ik L(ik−κ)
= e + e =
2k 2κ  2k 2κ
1 κ k L(ik+κ) κ k L(ik−κ)
     
= 1+i 1−i e + 1−i 1+i e =
4 k κ k κ
1 κ k κ k
    
= 1 + i − i + 1 eκL + 1 − i + i + 1 e−κL eikL =
4 k κ k κ
1 κ k κL κ k −κL ikL
       
= 2 eκL + e−κL + i − e −i − e e =
4 k κ k κ
1  κL κ k  κL κ k
    
= 2 e + e−κL + i − e − e−κL eikL = ε = −
4 k κ k κ
1 κL iε +κL
     
= e + e−κL + e − e−κL eikL =
2 4
iε 2 cosh(κL) + iε sinh(κL) ikL
 
= cosh(κL) + sinh(κL) eikL = e (2.157)
2 2

The quantity ε = κk − κk has been defined. By the magnitude square of |M11 |2 , one obtains
the transmission probability TL→R from (2.136):
1 (2.157)
TL→R = =
|M11 |2
4
= a cosh2 (x) + b sinh2 (x) = a + (b + a) sinh2 (x)
4 cosh (κL) + ε2 sinh2 (κL)
2

4
= (2.158)
4 + (ε2 + 4) sinh2 (κL)

k and κ are equivalent to the expressions found in (2.98) and (2.99), but now κ should be
represented in terms of the ratio of the energy of the incoming particle to the potential
x = E/V0 :
s s s
2mE 2m(V0 − E) 2mV0 (1 − x) √
k= and κ = = = α 1−x
ℏ2 ℏ2 ℏ2

In α, all constants are summarized. ε can thus be written as a function of x:

s s
V0 (1 − x) 1−x 1 − 2x
r
κ k E x
r
ε= − = − = − =p
k κ E V0 (1 − x) x 1−x x(1 − x)

For E < V0 , the already-derived formula for T still applies; for E > V0 , however, it must
be noted that the root in the hyperbolic function becomes imaginary. Utilizing the relation
sinh(ix) = −i sin(x), one obtains a piecewise transmission probability as a function of the
ratio x = E/V0 in the following form:

4
, x<1
 4+ (1−2x)2 +4 sinh2 (α√1−x)

  

x(1−x)
T (x) = 4
, x>1
 4− (1−2x)2 +4 sin2 (α√x−1)

 

x(1−x)

49
 2 Schrödinger Equation

E/V0 = 1 T =1 κL = π κL = 2π ... T =1
ε=3
ε=4 ε=5
T

T
α=3

α=1

α=5

E/V0 κL

Abb. 12: (left) Representation of the transmission probability for different energy ratios
E/V0 . (right) At constant potential characteristics V0 and L, the nπ-periodic
structure of the resonances is apparent.

50
Quantum Theory I

3 Formal Structure of Quantum Theory

Motivation: Dirac Notation and Hilbert Space

In the preceding chapters, we have shown that the state of a quantum system at a given
time can be described with a normalized wave function ψ(r). If we consider the space of
all possible states a quantum system can take, then this is a space of wave functions ψi (r),
for which (due to normalization) the following holds: d3 r|ψN (r)|2 = 1 (the N stands for
R

“normalized”). To achieve this, the (non-normalized) wave functions are provided with a
normalization factor N , so that ψN = N ψ.

For a wave function to be normalizable at all, it must be square-integrable. Therefore,

wave functions belong to the Rspace L2 of square-integrable functions in R3 , where for all
functions the following holds: d3 r|f (r)|2 < ∞. It can be shown that the L2 – with certain
additional conditions – forms a Hilbert space H.

In the following, we will describe the properties of this Hilbert space, and show how one can
use, among other things, the so-called Dirac notation to represent and compute quantum
states very elegantly and compactly using the properties of the Hilbert space.

3.1 Dirac Notation

The origin of the Dirac notation introduced by Paul Dirac becomes clear when one imagines
how the scalar product of two wave functions ϕ(r) and ψ(r) can be written:
Z
⟨ϕ|ψ⟩ = d3 r ϕ∗ (r)ψ(r) (3.1)

In the Dirac notation, one also uses the two halves of ⟨ϕ|ψ⟩ outside of the scalar product: The
quantum state corresponding to the wave function ψ(r) is written as |ψ⟩ and is called a Ket
vector. The left part of ⟨ϕ|ψ⟩ can also be written as ⟨ϕ|. This is then referred to as a Bra
vector. Bra and Ket together form a complete bracket, English “Bra-c-ket”. Therefore, the
Dirac notation is also called “Braket notation”. The reason one speaks of vectors – as we will
see – is because all elements of a Hilbert space can be understood as vectors.

In this notation, the following must be considered: The wave function ψ(r) explicitly depends on
the location r. However, one could also represent exactly the same quantum state, for instance,
using a wave function ψ̃(p), which depends on the momentum p. The wave function then
appears completely different. The necessary abstraction is provided by Dirac notation: The
quantum state is written as |ψ⟩, and this notation is independent of how the wave function is
concretely represented in a further course.

3.1.1 Hilbert Space and Ket Vectors

The physical quantum state of a system is thus described by the state vector |ψ⟩, which is
interpreted as an element of the Hilbert space H:
|ψ⟩ ∈ H (3.2)
Since a Hilbert space is a vector space, all properties of a vector space apply to a Ket state,
such as associativity, commutativity, distributivity, and the existence of a neutral as well as an
inverse element. The definition of a vector space essentially means that there must be vectors
and scalars, that we can add the vectors “in a meaningful way”, and that we can multiply scalars
and vectors “in a meaningful way”. We can now ascertain that the above-mentioned properties
also apply to Ket state vectors and complex numbers in the Hilbert space H.

51
 3 Formal Structure of Quantum Theory

In-Depth: Definition of a Vector Space

A vector space over a field K is the set V, for which, together with addition V × V → V
and scalar multiplication V × K → V, the following properties must apply (u, v, w ∈ V and
α, β ∈ K):

u + (v + w) = (u + v) + w (Associative law for vector addition)

u + v = v + u (Commutative law for vector addition)

u + 0 = 0 + u = u (∃ neutral vector element 0)

u + (−u) = (−u) + u = 0 (∀u : ∃ inverse vector element −u)

α · (u + v) = α · u + α · v (1st Distributive law for scalar product)

(α + β) · u = α · u + β · u (2nd Distributive law for scalar product)

(α · β) · u = α · (β · u) (Associative law for scalar product)

1 · u = u (the neutral scalar 1 is “sensibly” processed by the scalar product)

For the Ket vectors |ψ1 ⟩ , |ψ2 ⟩ , |ψ3 ⟩ ∈ H and scalars α, β ∈ C, the following holds:
|ψ1 ⟩ + (|ψ2 ⟩ + |ψ3 ⟩) = (|ψ1 ⟩ + (|ψ2 ⟩) + |ψ3 ⟩ (Associative law for vector addition)
|ψ1 ⟩ + |ψ2 ⟩ = |ψ2 ⟩ + |ψ1 ⟩ (Commutative law for vector addition)
|ψ1 ⟩ + |0⟩ = |0⟩ + |ψ1 ⟩ = |ψ1 ⟩ (∃ neutral state vector |0⟩)
|ψ1 ⟩ + (− |ψ1 ⟩) = (− |ψ1 ⟩) + |ψ1 ⟩ = |0⟩ (∀ |ψ⟩ : ∃ inverse state vector − |ψ⟩)
α (|ψ1 ⟩ + |ψ2 ⟩) = α |ψ1 ⟩ + α |ψ2 ⟩ (1st Distributive law for scalar product)
(α + β) · |ψ1 ⟩ = α |ψ1 ⟩ + β |ψ1 ⟩ (2nd Distributive law for scalar product)
(αβ) |ψ1 ⟩ = α (β |ψ1 ⟩) (Associative law for scalar product)
1 · |ψ1 ⟩ = |ψ1 ⟩ (The multiplication of 1 with a vector leaves it unchanged)
That all these points are fulfilled becomes intuitively clear when you mentally replace the (ab-
stract) Ket vectors with concrete wave functions: Of course, you can add multiple functions,
where the order does not matter. Similarly, a function can be multiplied by a (complex) number,
etc. And the “neutral state vector” |0⟩ simply corresponds to the zero function, which returns
zero everywhere in space, where we must deviate from the requirement of normalizability as
an exception; but we are still in the L2 -space. Physically, |0⟩ corresponds to a state “with-
out particles”. Therefore, it is occasionally acceptable that the probability of finding a particle
somewhere in space is equal to zero.

In-Depth: Wave Functions as Vectors in Hilbert Space

We have thus just shown that Ket vectors and all wave functions that represent such Ket
vectors, together with the complex numbers, satisfy the definition of a vector space. Ket
vectors and wave functions therefore represent vectors in Hilbert space. How can one con-
ceptualize this? What does it “really” (intuitively) mean for a Ket vector to be orthogonal
to another Ket vector, especially when these Ket vectors represent wave functions?

To understand this, it helps to visualize the analogy between Ket vectors and simple vec-
tors, for instance, in R3 . You can think of a simple vector v ∈ R3 as an arrow with

52
Quantum Theory I

direction and length that exists independent of the chosen coordinate system: v thus rep-
resents the “vector in itself”, without reference to any coordinate system. If the vector
v is instead written as a triplet of numbers v = (v1 , v2 , v3 )⊺ , then one has decided on a
particular coordinate system. You can represent the same vector in another coordinate
system with a different triplet of numbers v = (ṽ1 , ṽ2 , ṽ3 )⊺ . The Ket vector |ψ⟩ in Hilbert
space now corresponds to the vector v ∈ R3 . It represents the vector in Hilbert space “in
itself”, without committing to a coordinate system. A wave function ψ(x) belonging to
the Ket vector |ψ⟩ corresponds to the representation of the Ket vector |ψ⟩ in a particular
coordinate system. We can also represent the same vector |ψ⟩ in a different coordinate
system, for instance with the wave function ψ̃(p).

But how can a function ψ(x) in the Hilbert space H be analogous to a triplet of numbers
(v1 , v2 , v3 )⊺ in R3 ? To understand this, we first consider that we could replace the triplet
of numbers with the three numbers v1 , v2 , and v3 with a function v(i), which would allow
us to write the vector (in this coordinate system) as (v(1), v(2), v(3))⊺ . For v(i), i can
only take on the values 1, 2, or 3. However, since we want to go beyond R3 , we next
assume an infinitely-dimensional space R∞ . By this point, it becomes clear that we can
no longer represent a vector by simply writing triplets of numbers (v1 , v2 , v3 , v4 , v5 , . . . )⊺
with infinitely many entries. There is no sensible alternative but to represent the infinitely
many values vi by the function v(i). In the final step of generalization, we assume that
our vector space becomes so vast that not even the countably infinite number of values
v(i) with i = 1, 2, 3, 4, 5, . . . is sufficient to represent the vector. There exists
√ an infinite
continuum of values between v(1) and v(2), for instance v(1.5), v(1.3333), v( 2), etc. Our
vector is finally represented definitively by a function v(i) with i ∈ R!

The vector space was enlarged so that we can no longer manage with the countable values
v1 , v2 , v3 , etc. Instead of v, we now write |ψ⟩ (to illustrate that we are located in a space of
significantly higher dimensions – the Hilbert space H), and we replace vi ≡ v(i) or ṽi ≡ ṽ(i)
with (for example) ψ(x) or ψ̃(p). With this visualization, it becomes easier to understand
how two vectors in Hilbert space can be orthogonal, and how this can be calculated. We
first visualize two vectors a and b in R3 , and then the inner product follows:
3 3
a·b= ai bi = (3.3)
X X
ãi b̃i
i=1 i=1

If this value is equal to zero, then a and b are orthogonal. The value of the inner product is
independent of the coordinate system used. If we denote the distance between the possible
index values i as ∆i, then ∆i = 1, and for a and b in R3 or in R∞ , the following holds:
3 3 ∞ ∞
a·b= ∆i ai bi = ∆i ãi b̃i and a · b = ∆i a(i)b(i) = ∆i ã(i)b̃(i) (3.4)
X X X X

i=1 i=1 i=1 i=1

In the case of R∞ , we no longer write the values ai and bi , rather express these solely with
the functions a(i) and b(i). If we shift from R∞ to C∞ , it must additionally be considered
that one of the values must be conjugated complexly:
∞ ∞
(a, b) = ∆i a∗ (i)b(i) = ∆i ã∗ (i)b̃(i) (3.5)
X X

i=1 i=1

In the Hilbert space of square-integrable functions, where there are no countable indices

53
 3 Formal Structure of Quantum Theory

anymore, the inner product between the states |a⟩ and |b⟩ is finally calculated as follows:
Z ∞ Z ∞
∗
⟨a|b⟩ = dx a (x)b(x) = dp ã∗ (p)b̃(p) (3.6)
−∞ −∞

The “index” is now, for instance, x or p instead of i, the “index boundaries” now run
continuously over the entire space; ultimately the sum becomes an integral.

3.1.2 Dual Space and Bra Vectors

We have described in the introduction that in the Dirac notation, the scalar product ⟨ϕ|ψ⟩ is
somewhat dissected and that we not only give the right part |ψ⟩ as a “Ket vector” an independent
identity but also the left part ⟨ϕ| as a “Bra vector”. For fundamental understanding, it is usually
enough to simply imagine the Bra vector as the “left part of the scalar product”. However, as
we are indeed considering Bra vectors as independent objects, it makes sense to reflect on the
space to which Bra vectors belong and what properties they have.
So, what is a Bra vector? The answer lies in the scalar product from which we originally
“extracted” the Bra vector: Whenever we connect a Bra vector ⟨ϕ| with a Ket vector |ψ⟩, we
obtain a scalar product ⟨ϕ|ψ⟩, which gives us a (complex) number. You could say that the Bra
vector is a mathematical object that takes a Ket vector as input and outputs a number. Such
an object is called a functional. Specifically, a Bra vector is a linear functional because the
following holds:
Z
⟨ϕ| (α |ψ1 ⟩ + β |ψ2 ⟩) = dr ϕ∗ (r) (αψ1 (r) + βψ2 (r)) =
Z Z
∗
=α dr ϕ (r)ψ1 (r) + β dr ϕ∗ (r)ψ2 (r) =
= α ⟨ϕ|ψ1 ⟩ + β ⟨ϕ|ψ2 ⟩ (3.7)

Therefore, the space of all Bra vectors is the space of all linear functionals of the Hilbert space!
This space is called the dual space. It can be shown that all properties of the vector space
from the previous section are also satisfied for dual vectors, which is why the dual space also
represents a vector space and thus justifies the label as Bra vectors.

Through transposition and complex conjugation (briefly: “daggering”), a Ket vector can be
rewritten into a Bra vector, and vice versa:

|ψ⟩† = ⟨ψ| and ⟨ψ|† = |ψ⟩ (3.8)

The usage of complex conjugation is easily understood due to the way in which the scalar product
is calculated. The significance of the added transposition will only become clear later on.

3.1.3 The Scalar Product

In-Depth: Definition of Hilbert Space

A Hilbert space H is a (linear)pvector space with a scalar product ⟨ψ|ψ⟩, which induces
a corresponding norm ||ψ|| = ⟨ψ|ψ⟩. Additionally, the space must be complete, which
means that for any convergent Cauchy sequence ψi ∈ H, it holds that the limit limi→∞ ψi
is again in H.

Let |ψ⟩ now be an element of the Hilbert space H, then for the scalar product ⟨ψ|ψ⟩ ≥ 0.
Thus, the scalar product is positive semidefinite; equality ⟨ψ|ψ⟩ = 0 is only fulfilled if

54
Quantum Theory I

|ψ⟩ = 0.

According to definition, we can only speak of a Hilbert space if it is a vector space equipped
with a scalar product for the vectors, and we can define a norm via the scalar product. This
should now be the case: A scalar product of the form ⟨ψ1 |ψ2 ⟩ is already available as a starting
point for our considerations.

|a2 ⟩

|a1 ⟩
⟨a1 |a2 ⟩
||a1 ||2 |a1 ⟩

Abb. 13: Schematic representation of the scalar product as the projection of a vector |a2 ⟩ onto
|a1 ⟩ in the two-dimensional space R2 .

This scalar product ⟨ψ1 |ψ2 ⟩ can be interpreted, quite analogous to the tangible position vectors
in R3 , as the projection of the state |ψ2 ⟩ onto |ψ1 ⟩. Just as with position vectors in R3 , we can
define with the abstract vectors in Hilbert space: If |ψ1 ⟩ and |ψ2 ⟩ are orthogonal with respect
to each other, then the following holds:

⟨ψ1 |ψ2 ⟩ = 0 (3.9)

In this case, |ψ1 ⟩ contains no component parallel to |ψ2 ⟩. With the help of ⟨ψ|ψ⟩, one can now
define the norm ||ψ|| of a state vector |ψ⟩:
q
||ψ|| = ⟨ψ|ψ⟩ ≥ 0 (3.10)

That the expression (3.10) always yields a real number greater than or equal to zero is a necessary
condition for it to be a norm. Another requirement is also fulfilled: Expression (3.10) produces
a value of zero if and only if |ψ⟩ equals the zero vector |0⟩:

⟨ψ|ψ⟩ = 0 ⇐⇒ |ψ⟩ = |0⟩ (3.11)

Another property of the scalar product is that it is linear in the second argument (this is
according to our definition, in some fields of mathematics and physics, linearity in the first
argument is also required):

⟨ψ1 |αψ2 + βψ3 ⟩ = ⟨ψ1 |αψ2 ⟩ + ⟨ψ1 |βψ3 ⟩ = α ⟨ψ1 |ψ2 ⟩ + β ⟨ψ1 |ψ3 ⟩ (3.12)

For the first argument, the scalar product is semi-linear, meaning the following:

⟨αψ1 + βψ2 |ψ3 ⟩ = ⟨αψ1 |ψ3 ⟩ + ⟨βψ2 |ψ3 ⟩ = α∗ ⟨ψ1 |ψ3 ⟩ + β ∗ ⟨ψ2 |ψ3 ⟩ (3.13)

(3.12) and (3.13) together indicate that the scalar product is a so-called sesquilinear form with
the following property:
⟨ψ1 |ψ2 ⟩∗ = ⟨ψ2 |ψ1 ⟩ (3.14)
Scalar quantities can therefore always be pulled out of the Ket-bracket:

|αψ⟩ = α |ψ⟩ (3.15)

For Bra vectors, we must note that scalars standing within the Bra-bracket are complex conju-
gated and transposed (briefly: “daggered”). For a scalar α, α† = α∗ ; for Bra vectors, therefore:

⟨ψα| = ⟨ψ| α∗ (3.16)

55
 3 Formal Structure of Quantum Theory

Furthermore, for a Ket vector due to the general rule for (complex-conjugated) transposition
(XY )† = Y † X † :
(α |ψ⟩)† = |ψ⟩† α† = ⟨ψ| α∗ (3.17)
For the scalar product, we can state the Cauchy-Schwarz inequality:
| ⟨ψ1 |ψ2 ⟩ | ≤ ||ψ1 || · ||ψ2 || (3.18)
With (3.18), the triangle inequality for arbitrary state vectors |ψ1 ⟩ and |ψ2 ⟩ can also be moti-
vated:
||ψ1 + ψ2 || ≤ ||ψ1 || + ||ψ2 || (3.19)

Example: Triangle Inequality

From the Cauchy-Schwarz inequality, the validity of the triangle inequality can be easily
demonstrated. The norm of the sum of two state vectors |ψ⟩ and |ϕ⟩ is squared to eliminate
roots, resulting in:
(3.18)
||ϕ + ψ||2 = ||ϕ||2 + ||ψ||2 + ⟨ϕ|ψ⟩ + ⟨ψ|ϕ⟩ ≤
≤ ||ϕ||2 + ||ψ||2 + ||ϕ|| · ||ψ|| + ||ψ|| · ||ϕ|| =
= ||ϕ||2 + ||ψ||2 + 2||ϕ|| · ||ψ|| =
= (||ϕ|| + ||ψ||)2

Indeed, this inequality follows from (3.18), allowing for an estimation of the magnitude of
⟨ϕ|ψ⟩. Subtracting the square root from both sides exactly yields the triangle inequality
(3.19):
||ϕ + ψ|| ≤ ||ϕ|| + ||ψ||

3.1.4 Complete Orthonormal System and Dimension of the Hilbert Space

Just like with the tangible position vectors in R3 , one can also represent every state vector
|ψ⟩ in an N -dimensional Hilbert space H as a linear combination of orthonormal basis states
{|ϕ1 ⟩ , |ϕ2 ⟩ , . . . |ϕN ⟩}:
|ψ⟩ = (3.20)
X
ci |ϕi ⟩
i
Unlike in R3 , however, the expansion coefficients ci are complex-valued: ci ∈ C. When all basis
vectors |ϕi ⟩ are orthogonal to each other and normalized to a length of 1, so that ⟨ϕi |ϕj ⟩ = δij
holds, one speaks of an orthonormal system (ONS). The dimension of a Hilbert space is given
by the maximum possible number of linearly independent basis vectors. If an orthonormal basis
in an N -dimensional Hilbert space HN contains exactly N orthonormal vectors, then this basis
is termed a complete orthonormal system (CONS). With a CONS as basis, any arbitrary vector
in HN can be represented with N expansion coefficients c1 , . . . , cN .
Depending on the type of quantum system being considered, the dimension of the underlying
Hilbert space can be finite or even (countably) infinite. For many phenomena, it is even necessary
to extend the Hilbert space to an uncountably infinite-dimensional space, the so-called extended
Hilbert space. However, for now, we will focus on finite-dimensional Hilbert spaces.
Using the decomposition in basis vectors shown in (3.20), an abstract Ket vector |ψ⟩ can be
represented as a column vector in CN that contains the expansion coefficients ci :
 
c1
{|ϕi ⟩}  c2 
 
|ψ⟩ = c1 |ϕ1 ⟩ + c2 |ϕ2 ⟩ + · · · + cN |ϕN ⟩ −−−→   .. 
 (3.21)
 . 
cN

56
Quantum Theory I

The corresponding Bra vector ⟨ψ| = |ψ⟩† is then represented by the complex conjugated, trans-
posed column vector of (3.21) – thus a row vector:
{⟨ϕi |}
 
⟨ψ| = c∗1 ⟨ϕ1 | + c∗2 ⟨ϕ2 | + · · · + c∗N ⟨ϕN | −−−→ c∗1 c∗2 . . . c∗N (3.22)

It should be noted that the numbers in the vectors (3.21) and (3.22) depend on the chosen basis.
Therefore, one should use an arrow instead of an equals sign, which also specifies the underlying
basis (in our case {ϕi }). The scalar product ⟨ψ|ψ⟩ can be expressed as the multiplication of row
and column vectors, and is independent of the chosen basis, which is why an equals sign can be
used again:  
c1
  c2  ∗ ∗ ∗
⟨ψ|ψ⟩ = c1 c2 . . . cN  . 
∗ ∗ ∗
 = c1 c1 + c2 c2 + · · · + cN cN (3.23)
 
.

 . 
cN

3.1.5 Operators in the Braket Notation

There is also a compact notation for operators in the Dirac formalism. We have already learned
operators as mathematical objects that “receive as input a wave function” (or, as typically
phrased: “act on a wave function”) and produce another function as a result. When we transfer
this to the more abstract Dirac notation with vectors in Hilbert space, an operator  is a (in
our case linear) mapping from the Hilbert space HN into the same Hilbert space HN (Â : HN →
HN ). An operator  always acts only from the left on a Ket state |ψ⟩; the action of the operator
 from the right on the Ket vector |ψ⟩ is not defined:

|Âψ1 ⟩ = Â |ψ1 ⟩ = |ψ2 ⟩ (3.24)

However, the complex conjugated (adjoint) operator † can indeed act from the right on a Bra
state:
⟨Âψ1 | = ⟨ψ1 | † = ⟨ψ2 | (3.25)
That we can state ⟨Âψ1 | = ⟨ψ1 | † in (3.25) is due to the fact that the Bra-bracket ⟨·| acts like
a “dagger” operator on  and ψ, and the rule (XY )† = Y † X † comes into effect.

Important Terms Regarding Operators In quantum mechanics, hermitian operators have par-
ticular significance. These are defined by the relation  = † . Such operators act on a Ket
vector (to the right) the same way they act on a Bra vector (to the left). Unitary operators,
on the other hand, have the characteristic that the inverse and the adjoint of an operator Û
coincide:
Û −1 = Û † and Û Û −1 = Û Û † = 1 (3.26)
In general, an adjoint operator † should be defined as follows for all |ϕ⟩ and |ψ⟩:
∗
∀ |ψ⟩ , |ϕ⟩ : ⟨ψ|† |ϕ⟩ = [(⟨ψ|† |ϕ⟩)† ]∗ = ⟨ϕ|Â|ψ⟩ (3.27)

An operator  is homogeneous in its action on a state |ψ⟩, meaning that a scalar α and an
operator  can always be interchanged:

 |αψ⟩ = Âα |ψ⟩ = α |ψ⟩ (3.28)

We only consider linear operators. This means that when  acts on a sum of states |ψ1 ⟩ and
|ψ2 ⟩, it is the same as if the operator acts on each state individually, and then the sum is formed:

 |ψ1 + ψ2 ⟩ =  (|ψ1 ⟩ + |ψ2 ⟩) =  |ψ1 ⟩ +  |ψ2 ⟩ (3.29)

57
 3 Formal Structure of Quantum Theory

3.2 Hermitian Operators, Eigenfunctions, and Eigenvalues

When an operator  acts on a state |ψ1 ⟩, a new state is generally created, which we can call,
for example, |ψ2 ⟩:
 |ψ1 ⟩ = |ψ2 ⟩
However, there are also states |ai ⟩ for an operator  such that the following relation holds:

 |ai ⟩ = ai |ai ⟩ (3.30)

The operator  maps the state |ai ⟩ back onto itself, but with a scaling factor ai (where ai must
be a scalar). Since there can be several such special states |ai ⟩ for each operator, we number
them with the subindex i. Generally, (3.30) reminds us of the eigenvalue problem from linear
algebra, where for a matrix A, the corresponding eigenvectors vi and scalars λi are determined,
satisfying the eigenvalue equation Avi = λi vi . The relationship (3.30) is thus an eigenvalue
equation for the operator Â, with the eigenstates |ai ⟩ and the eigenvalues ai ! The set of all
eigenstates |ai ⟩ of the operator  spans a subspace in the underlying Hilbert space, the so-called
eigenspace of Â. As already mentioned, hermitian operators play a special role in quantum
mechanics. To reiterate, for these operators the following holds:

 = † (3.31)

3.2.1 Real Eigenvalues of Hermitian Operators

And just as a hermitian matrix A = A† has only real eigenvalues λi , a hermitian operator
 = † has only real eigenvalues ai . This can be confirmed as follows: If an operator  and a
normalized eigenvector |ai ⟩ are given, the associated eigenvalue ai can be expressed as follows:

⟨ai | Â |ai ⟩ = ⟨ai |ai |ai ⟩ = ai ⟨ai |ai ⟩ = ai (3.32)

| {z } | {z } | {z }
 acts to Scalar ai normalized
the right interchanged with ⟨ai |

However, if the operator  is hermitian, you can also let  act to the left, and (3.32) is written
as follows:
⟨ai |  |ai ⟩ = ⟨ai | † |ai ⟩ = ⟨ai |a∗i |ai ⟩ = a∗i ⟨ai |ai ⟩ = a∗i (3.33)
| {z } | {z } | {z } | {z }
Â=† † acts to Scalar a∗i normalized
the left interchanged with ⟨ai |

For a hermitian operator, it should make no difference whether it acts to the right on a Ket
vector or to the left on a Bra vector. Thus, equations (3.32) and (3.33) must be simultaneously
fulfilled. From this, it follows:
ai = a∗i ⇐⇒ ai ∈ R (3.34)
Thus, hermitian operators indeed possess real eigenvalues. But what does this mean physically?
Recall that Ket vectors like |ai ⟩ correspond to states of quantum systems. The application of
a hermitian operator  to a state is linked in quantum theory to the measurement of the cor-
responding observable. The fact that there are corresponding eigenstates |ai ⟩ with eigenvalues
ai for each operator  means the following initially: If the system is in the eigenstate |ai ⟩, we
always measure the measurement value ai .
This may seem trivial, but recall: Measurements on quantum systems are usually accompanied
by some degree of uncertainty. Even if an experiment is prepared the same way every time,
the measurement outcome can deviate from the previous one every time. Therefore, a quantum
state that yields the same measurement result each time is rather an exception than the rule in
the quantum world and is thus something special!

58
Quantum Theory I

Alongside, we now understand why we need hermitian operators (with real eigenvalues) to
represent observables in the first place: The eigenvalues of the operators correspond to the
measurement values of eigenstates, and measurement values are always represented by real
numbers.

3.2.2 Orthogonality of Eigenstates of Hermitian Operators

We have shown in (3.34) that hermitian operators have real eigenvalues ai ∈ R. Thankfully,
this means that we don’t need to worry about whether a hermitian operator  = † acts to the
right on a Ket vector or to the left on a Bra vector. Due to the hermiticity of Â, the following
holds equally for the eigenvectors |ai ⟩ and ⟨ai |:

 |ai ⟩ = ai |ai ⟩ and ⟨ai |  = ⟨ai | ai

With this, we can show that for a hermitian operator Â, any two arbitrary eigenstates |ai ⟩ and
|aj ⟩ (i ̸= j) are always orthogonal to each other.

Non-degenerate System First, we let  act on a state |ai ⟩ and project the result onto ⟨aj | –
due to the hermiticity of the operator, we could also act on ⟨aj | with  and project that result
onto |ai ⟩. We obtain two equations:

(I)| ⟨aj |Â|ai ⟩ = ai ⟨aj |ai ⟩ | Acts to the right

(3.35)
(II)| ⟨aj |Â|ai ⟩ = aj ⟨aj |ai ⟩ | Acts to the left

Subtracting the lower equation from the upper in (3.35) gives us:

(I) – (II)| 0 = (ai − aj ) ⟨aj |ai ⟩ (3.36)

(3.36) is fulfilled when either the eigenvalues ai and aj are equal, or ⟨aj |ai ⟩ vanishes; the latter
occurs when an eigenstate equals the zero vector or both eigenstates are orthogonal. If the
eigenvalues all satisfy ai ̸= aj , then they are termed non-degenerate, and the only non-trivial
way to satisfy (3.36) is as follows:
⟨ai |aj ⟩ = 0 (3.37)
This means that the eigenstates of hermitian operators with non-degenerate eigenvalues are
orthogonal to each other! If the eigenvalues of two or more (linearly independent) eigenvectors
are the same (ai = aj ), then one speaks of degeneracy, and the eigenfunctions are not necessarily
orthogonal to each other: ⟨ψi |ψj ⟩ ̸= 0. If n linearly independent states belong to the same
eigenvalue, one speaks of n-fold degeneracy.

System with Degeneracy The case of degeneracy shall now be considered in more detail:
Suppose we have a1 = a2 for the first two eigenvalues. The associated eigenfunctions |a1 ⟩ and
|a2 ⟩ are normalized but not orthogonal due to degeneracy: ⟨a1 |a2 ⟩ = ̸ 0. By forming linear
combinations of |a1 ⟩ and |a2 ⟩, orthogonal states can always be constructed – this can be done
through a simple algorithm within the Gram-Schmidt orthogonalization process.
The goal of the Gram-Schmidt process is to find two new (normalized) states |ã1 ⟩ and |ã2 ⟩ for
which now applies: ⟨ã1 |ã2 ⟩ = 0. For simplicity, the first new state is chosen as:

|ã1 ⟩ = |a1 ⟩ (3.38)

For |ã2 ⟩, we initially project |a2 ⟩ onto |a1 ⟩, thus obtaining the parallel component of the two
eigenstates – this is depicted by the red vector ⟨a1 |a2 ⟩ |a1 ⟩ in Figure 14. Subtracting the par-
allel component from the original state |a2 ⟩ leaves only the component |ã2 ⟩ orthogonal to |a1 ⟩.
Therefore, it holds:
|ã2 ⟩ = |a2 ⟩ − ⟨a1 |a2 ⟩ |a1 ⟩ (3.39)

59
 3 Formal Structure of Quantum Theory

|a2 ⟩

|ã2 ⟩

|a1 ⟩ = |ã1 ⟩
⟨a1 |a2 ⟩ |a1 ⟩

Abb. 14: Construction of a state |ã2 ⟩ that satisfies the condition ⟨ã1 |ã2 ⟩ = ⟨a1 |ã2 ⟩ = 0. Note
that |ã2 ⟩ must be normalized first.

Orthogonality was required between the newly constructed states, which can be easily checked:
(3.39)
⟨ã1 |ã2 ⟩ = ⟨a1 |ã2 ⟩ =
= ⟨a1 | (|a2 ⟩ − ⟨a1 |a2 ⟩ |a1 ⟩) =
= ⟨a1 |a2 ⟩ − ⟨a1 | ⟨a1 |a2 ⟩ |a1 ⟩ = | ⟨a1 |a2 ⟩ is a scalar ⇒ factor out
= ⟨a1 |a2 ⟩ − ⟨a1 |a2 ⟩ ⟨a1 |a1 ⟩ = | ⟨a1 |a1 ⟩ = 1
= ⟨a1 |a2 ⟩ − ⟨a1 |a2 ⟩ = 0 □ (3.40)

For general number n of eigenfunctions to be orthogonalized, the Gram-Schmidt process states

the following:
i−1
⟨aj |ai ⟩
|ãi ⟩ = |ai ⟩ − (3.41)
X
|aj ⟩
j=1
⟨aj |aj ⟩

Since the parallel component is always subtracted between the affected vectors, in the end only
the desired orthogonal state vector remains.

3.3 Matrix Representation of an Operator

We now consider a hermitian operator  whose eigenfunctions |aj ⟩ corresponding to the discrete
eigenvalues aj are known, fulfilling the eigenvalue equation  |aj ⟩ = aj |aj ⟩. If this eigenvalue
equation is projected onto another eigenstate ai , we obtain:

Aij = ⟨ai |Â|aj ⟩ = ⟨ai |aj |aj ⟩ = aj ⟨ai |aj ⟩ (3.42)

Since aj is a scalar quantity, we could factor it out of the Braket-bracket in the last step. Aij
is now the matrix representation of the operator  in the space of its eigenfunctions, where the
index i corresponds to the row index and index j to the column index. Since i and j share the
same range of values, Aij is a square matrix.

Non-degenerate System If no degeneracies occur, further statements can be made about Aij :
Since each eigenvalue ai is associated with a single eigenfunction |ai ⟩, these must be orthogonal
to each other, and (3.42) simplifies to:

Aij = ⟨ai |Â|aj ⟩ = ai ⟨ai |aj ⟩ = ai δij (3.43)

Thus, the matrix is diagonal without degeneracy and can be written in the space of eigenfunctions
(in the so-called eigenbasis) as:
 .. 
a1 0 . 
 0 a2
Aij =   = diag{a1 , a2 , . . . } (3.44)
.. ..

. .

60
Quantum Theory I

System with Degeneracy If the assumption is dropped that no degeneracies occur, strict
diagonal format is lost, though block-diagonal format remains. Assuming for example two
eigenfunctions |ak ⟩ and |al ⟩ have the same eigenvalue ak , then ⟨ak |al ⟩ ̸= δkl holds. The matrix
is ultimately block-diagonal:

a1 0 . . .
 
0 a 
2
.
 
. .. 
Aij =  . . 
 (3.45)
..
  
. . . .
 

.. ..
  
. .

The form of the block matrix can be quickly written in the special case (twofold degeneracy
of the eigenstates (|a3 ⟩ and |a4 ⟩ with a common eigenvalue a3 from a total of four possible
eigenstates): For the off-diagonal elements, it holds ⟨a3 |Â|a4 ⟩ = a3 ⟨a3 |a4 ⟩ (we let  act to the
left here), and ⟨a4 |Â|a3 ⟩ = a3 ⟨a4 |a3 ⟩ (here it acts to the right), which allows us to write the
block in matrix representation as:
! !
1 ⟨a3 |a4 ⟩ 1 ⟨a3 |a4 ⟩
ABlock = a3 = a3 (3.46)
34
⟨a4 |a3 ⟩ 1 ⟨a3 |a4 ⟩∗ 1

3.3.1 Projections and Spectral Representation

According to (3.20), any arbitrary state |ψ⟩ in an N -dimensional Hilbert space HN can always be
represented by a set of basis states {|ϕi ⟩}. It is especially practical to use normalized, orthogonal
basis vectors (an orthonormal system, briefly ONS) for this. Then it holds:

|ψ⟩ = ci |ϕi ⟩ with ⟨ϕi |ϕj ⟩ = δij (3.47)

X

Each expansion coefficient ci in (3.47) can be easily calculated for an ONS – just as with position
vectors in R3 – by projecting onto the respective basis vector |ϕi ⟩:

ci = ⟨ϕi |ψ⟩ (3.48)

From the completeness of the Hilbert space HN , it follows that a complete orthonormal basis
system (briefly: CONS) can be found. By forming the tensor product |ϕi ⟩ ⟨ϕi | from the basis
vectors and summing over all possible states i, a “completeness relation” (i.e., a unity operator
1 that does not change the state) can be constructed:

1= (3.49)
X
|ϕi ⟩ ⟨ϕi |
i

If we apply a completeness relation to a wave function |ψ⟩, we naturally obtain |ψ⟩ again, hence
1 |ψ⟩ = |ψ⟩. Let us take a closer look at what happens when we apply the completeness relation,
as defined in (3.49), to |ψ⟩:
(3.49) (3.48) (3.47)
|ψ⟩ = 1 |ψ⟩ = |ϕi ⟩ ⟨ϕi |ψ⟩ = ⟨ϕi |ψ⟩ |ϕi ⟩ = ci |ϕi ⟩ = |ψ⟩ □
X X X

i i i

Hence, we can regard |ϕi ⟩ ⟨ϕi | as an operator P̂i that projects |ψ⟩ onto the ith eigenstate |ϕi ⟩,
scaling with ci . The effect of P̂i is defined as follows:

P̂i |ψ⟩ = ci |ϕi ⟩ with P̂i = |ϕi ⟩ ⟨ϕi | (3.50)

61
 3 Formal Structure of Quantum Theory

P̂i is a projection operator, which is recognizable by its idempotence: P̂i2 = P̂i (and accordingly
holds over complete induction: P̂in = P̂i ). That this is so can be easily demonstrated:
(3.50) (3.50)
P̂i2 = |ϕi ⟩ ⟨ϕi |ϕi ⟩ ⟨ϕi | = |ϕi ⟩ ⟨ϕi | = P̂i □

The trick that one can insert the unity operator 1 at any time is also used to represent an
operator  in its eigenbasis {|ai ⟩}:
(3.49)
 = 1Â1 =
X  X 
= |ai ⟩ ⟨ai | Â |aj ⟩ ⟨aj | =
i j

= |ai ⟩ ⟨ai |Â|aj ⟩ ⟨aj | = | Â |aj ⟩ = aj |aj ⟩

XX

i j

= |ai ⟩ ⟨ai |aj |aj ⟩ ⟨aj | = | aj is a scalar ⇒ pull out

XX

i j

= aj |ai ⟩ ⟨ai |aj ⟩ ⟨aj | = | ⟨ai |aj ⟩ = δij

XX

i j

= aj |ai ⟩ δij ⟨aj | =

XX

i j

= (3.51)
X
ai |ai ⟩ ⟨ai |
i

This representation of an operator is called the spectral representation in the eigenbasis. In this
eigenbasis, Â is then representable as a diagonal matrix with the corresponding eigenvalues in
the diagonal:
a1 0 . . .
 

{|ai ⟩}  0 a2
 
(3.52)

 −−−→   .. .. 
. . 

aN
If the operator  is hermitian and thus satisfies  = † , all eigenvalues ai = a∗i are real
quantities. This can be utilized in the following relation:

† = (ai |ϕi ⟩ ⟨ϕi |)† = a∗i ⟨ϕi |† |ϕi ⟩† = ai |ϕi ⟩ ⟨ϕi | =  □ (3.53)
X X X

i i i

3.4 Distribution Space

Motivation: Continuous Observables

In the previous chapters, we have (more or less tacitly) assumed that for an operator Â
there is always a countable number of eigenstates. The eigenvalue equation  |ai ⟩ = ai |ai ⟩
expresses this through the discrete index i. The physical meaning was that with every
single measurement, we can only measure one value from the list of possible operator
eigenvalues {ai }. The corresponding eigenstates {|ai ⟩} form an ONB, which is why we can
represent general states |ψ⟩ as a superposition of the eigenstates |ai ⟩: |ψ⟩ = i ci |ai ⟩ with
P

ci = ⟨ai |ψ⟩. Even more simply, we can represent |ψ⟩ in the basis {|ai ⟩} as a column vector
and operators as square N × N matrices.

This makes physical sense in many problems: For instance, if one measures the energy of a
particle in an infinitely deep potential well, the measuring device will display a value from
the list of eigenenergies belonging to the Hamiltonian operator with every energy measure-

62
Quantum Theory I

ment. There is a lowest energy E1 (ground state energy), then a next higher energy E2
(energy of the first excited state), etc. Even if we assume infinitely many energy values:
The energy basis remains (in this example) always countable.

Apparently, there are observables that behave differently: If we measure, for example, the
position of a particle in a one-dimensional system along the x-axis, this is expressed by
the corresponding position operator x̂. However, in general, we will not expect that there
exists only a countable set {xi } of places where we can find the particle during a measure-
ment. The notation x̂ |xi ⟩ = xi |xi ⟩, which implies discrete eigenvalues {xi } and position
eigenvectors {|xi ⟩}, therefore no longer makes sense.

In the following chapters, we will deal with how to solve this problem. Obviously, one must
move from a countable set of basis vectors to an uncountable set; accordingly, a general
state |ψ⟩ can no longer be represented as a discrete superposition of basis vectors, but only
through a corresponding integral. Related to this, we will also reveal the secret of how
to transition from the abstract ket representation |ψ⟩ to the more concrete wave function
ψ(x), or, for example, to ψ̃(p).

3.4.1 Position and Momentum Eigenvalues and Eigenfunctions

In a one-dimensional task, we will generally not expect that there exists only a countable set {xi }
of places where a particle can be located during a measurement. The notation x̂ |xi ⟩ = xi |xi ⟩
is therefore meaningless if one assumes that i may only take integer values. To fix the problem,
one could, of course, agree that the index i can take not only integer values but any real values.
For example, x1.234 would be the eigenvalue corresponding to a measurement where we find the
particle at position x = 1.234. However, from this example, one can already see that we no longer
need the index i in this case. If x can already take any real value, it makes sense in our notation
to use x instead of xi . And since for every real eigenvalue x, there is a corresponding eigenvector,
we can also write this with |x⟩ instead of |xi ⟩. This allows us to write the (uncountably infinite)
set of eigenvalue equations for the position operator x̂ in one dimension as follows:
x̂ |x⟩ = x |x⟩ (3.54)
Correspondingly, we write for the three-dimensional position operator r̂ a similar eigenvalue
equation:

r̂ |r⟩ = r |r⟩ (3.55)

The (uncountably infinite) set of position eigenvectors {|r⟩} is called the position eigenbasis.
It should be noted that in the three-dimensional case, the corresponding eigenvalues r are, of
course, position vectors that must be fully expressed with three numerical values.
Analogously, one can define the momentum eigenbasis {|p⟩} for the three-dimensional momen-
tum operator p̂. For every momentum eigenvector |p⟩, there exists an eigenvalue p, so the
following eigenvalue equation holds:

p̂ |p⟩ = p |p⟩ (3.56)

Of course, an eigenvalue equation for each component of the momentum operator p̂i can also be
written. Because p = ℏk, one can equivalently also write for the wave number operator k̂:
k̂ |k⟩ = k |k⟩ (3.57)
At this point, we must note: The action of p̂ or k̂ from (3.56) and (3.57) is so simple only
because we write the eigenvalue problem in the respective eigenbasis {|p⟩} or {|k⟩}. What the

63
 3 Formal Structure of Quantum Theory

action of the momentum operator in the position basis {|r⟩} looks like, we will discuss in a later
section.

3.4.2 From the Abstract Ket Vector to the Concrete Wave Function
Let us recap how a wave function can be decomposed with a countable basis {|ai ⟩} and how the
normalization appears:

|ψ⟩ = 1 |ψ⟩ = |ai ⟩ ⟨ai |ψ⟩ = and ⟨ψ|ψ⟩ = c∗i ci = 1

X X X
ci |ai ⟩
i i i
| {z }
ci

If we are dealing with a continuous, uncountable basis such as the one-dimensional position basis
{|x⟩}, then we must replace the sum with an integral:
Z Z
|ψ⟩ = 1 |ψ⟩ = dx |x⟩ ⟨x|ψ⟩ = dx c(x) |x⟩ (3.58)
| {z }
c(x)

The normalization can also be represented in a form similar to the discrete case, which we will
later justify more precisely: Z
⟨ψ|ψ⟩ = dx c∗ (x)c(x) = 1 (3.59)

We also know that the scalar product ⟨ψ|ψ⟩ is independent of the representation of the vectors;
therefore, the following relation must also be valid:
Z
⟨ψ|ψ⟩ = dx ψ ∗ (x)ψ(x) = 1 (3.60)

A comparison of equations (3.59) and (3.60) shows that ψ(x) and c(x) are apparently identical,
i.e., ψ(x) ≡ c(x). From (3.58) we additionally know that c(x) = ⟨x|ψ⟩. Thus, we have now
discovered how to formally transition from the abstract vector |ψ⟩ to the wave function ψ(x):

ψ(x) = ⟨x|ψ⟩ (3.61)

The abstract wave function |ψ⟩ is thus projected into the position basis {|x⟩} to enable a mean-
ingful representation of a state. In three-dimensional space, for a wave function, the same
reasoning as in one dimension applies analogously:

ψ(r) = ⟨r|ψ⟩ (3.62)

If one wants a wave function ψ̃(p), which does not have the position as a function argument
but a momentum, one can project the abstract ket vector |ψ⟩ onto the momentum eigenbasis
analogously:

ψ̃(p) = ⟨p|ψ⟩ (3.63)

Because of the relation p = ℏk we can write equivalently for a wave function ψ̄(k) with a wave
vector as argument:
ψ̄(k) = ⟨k|ψ⟩ (3.64)
So one recognizes the following: Just as, for example, the same position vector r in Euclidean
space can be represented by completely different numeric values (expansion coefficients ci ), the
same state vector |ψ⟩ in the Hilbert space H can be represented by entirely different mathe-
matical functions ψ(r) or ψ̃(p). These ultimately represent nothing more than an uncountably
large set of “expansion coefficients” for different bases.

64
Quantum Theory I

3.4.3 Position and Momentum Eigenfunctions in Position and Momentum Basis

For orthogonal, normalized base systems with countably many basis vectors {|ai ⟩}, we have
already stated that (due to orthogonality and normalization) it must trivially hold:

⟨ai |aj ⟩ = δij

In the transition to orthogonal, normalized base systems with uncountably many basis vectors,
such as the position basis {|x⟩}, it holds analogously:

⟨x|x′ ⟩ = δ(x − x′ ) (3.65)

With δ(x − x′ ) is meant the Dirac delta function (or more precisely: delta distribution), which
is defined by: Z ∞
dx δ(x − x′ ) f (x) = f (x′ ) (3.66)
−∞

In-Depth: Delta Distribution

The delta distribution δ(x) assigns the evaluation of the function at x = 0 to every arbi-
trarily often differentiable function f (x). One can write the functional:

⟨δ, f ⟩ = f (0) (3.67)

This pairing takes place formally in an integral of the following form:

Z ∞
⟨δ, f ⟩ = dx δ(x)f (x) = f (0) (3.68)
−∞

This shows that the delta distribution δ(x) must have the property of taking the value 0
everywhere for x ̸= 0. Around the value x = 0, δ(x) must – graphically speaking – form
an “infinitely narrow” peak that has the exact area of 1, so that it holds:
Z ∞
dx δ(x) = 1 (3.69)
−∞

Although in physics one often speaks of the “delta function,” strictly speaking, there is no
function that meets this definition. δ(x) is only defined by (3.67) and (3.68) and is thus
more accurately called delta distribution. However, δ(x) can be represented as the limit of
a sequence of functions, whereby there are many possibilities for representation here. One
of these uses an ever narrower Gaussian function or sinc function:
1 2 2
δ(x) = lim √ e−x /(2σ )
σ 2π
σ→0
(3.70)
1
δ(x) = lim sinc(x/n)
n→0 n

The most common calculation rules with the δ function are given below (noting that we
always have to integrate over δ(x)):

f (x)δ(x − x0 ) = f (x0 )δ(x − x0 ) (3.71)

δ(x) = δ(−x) (3.72)
xδ(x) = 0 (3.73)
δ(ax) = δ(x)/|a| (3.74)
|x| δ(x2 ) = δ(x) (3.75)

65
 3 Formal Structure of Quantum Theory

In the case of the following relation (3.76), xi are the simple zeros of the function f (x).
X δ(x − xi )
δ(f (x)) = (3.76)
i
|f ′ (xi )|

In (3.61), we have learned that ⟨x|ψ⟩ = ψ(x) holds. Thus, the expression ⟨x|x′ ⟩ in equation
(3.65) is nothing other than the eigenfunction |x′ ⟩ of the position operator (to the eigenvalue
x′ ) represented in the position basis {|x⟩} – in other words, the projection of one position state
|x′ ⟩ onto another position state |x⟩. Because the eigenstates of the position operator form an
eigenbasis, different position states are orthogonal to each other, and it follows:

⟨x|x′ ⟩ = ϕx′ (x) = δ(x − x′ ) (3.77)

The scalar product only does not vanish if both position states coincide (x′ = x). Completely
analogously for k-space or momentum space, one can write:
⟨p|p′ ⟩ = ϕp′ (p) = δ(p − p′ )
(3.78)
⟨k|k ′ ⟩ = ϕk′ (k) = δ(k − k ′ )
In three-dimensional space, similar relations hold, but at this point, we have to use the three-
dimensional delta distribution:
⟨r|r′ ⟩ = ϕr′ (r) = δ (3) (r − r′ )
⟨p|p′ ⟩ = ϕp′ (p) = δ (3) (p − p′ ) (3.79)
′ ′
⟨k|k ⟩ = ϕk′ (k) = δ (3)
(k − k )

Representation in Eigen Space We now want to determine the effect of the position operator
x̂ in the position basis (its eigenbasis). The eigenvalue equation to be fulfilled is:

x̂ |x′ ⟩ = x′ |x′ ⟩

To exploit relationship (3.77), we project this relation from the left onto ⟨x|:

⟨x|x̂x′ ⟩ = ⟨x|x′ |x′ ⟩ = x′ ⟨x|x′ ⟩ (3.77)
= x′ δ(x − x′ )
⟨x|x̂|x′ ⟩ = (3.77)
(3.80)

⟨xx̂|x′ ⟩ = ⟨x|x|x′ ⟩ = x ⟨x|x′ ⟩ = x δ(x − x′ )

Because the position operator is Hermitian x̂ = x̂† , we can let it act both to the right and
to the left – from this, it can be concluded that xδ(x − x′ ) = x′ δ(x − x′ ) holds. In the one-
and three-dimensional position space, we recognize through this the (very simple) form of the
position operator x̂ or r̂:
{|x⟩}
x̂ −−−→ x (3.81)
{|r⟩}
r̂ −−−→ r (3.82)

The same applies correspondingly for the momentum operator p̂i ≡ p̂ or p̂ in momentum space,
where we can argue analogously to position space:
{|p⟩}
p̂ −−−→ p (3.83)
{|p⟩}
p̂ −−−→ p (3.84)

Next, we determine the momentum eigenfunctions |px ⟩ ≡ |p⟩ in the one-dimensional position
basis ϕp (x) = ⟨x|p⟩. The eigenvalue equation to be fulfilled for the operator p̂x ≡ p̂ reads:

p̂ |p⟩ = p |p⟩

66
Quantum Theory I

Since we are interested in the representation in the position basis, we project the equation onto
a position eigenstate ⟨x|. In this case, we can use the already motivated position representation
of the momentum operator from (2.15):
(2.14) ∂
⟨x|p̂|p⟩ = ⟨x|p̂p⟩ = ⟨x|p|p⟩ = p ⟨x|p⟩ = p ϕp (x) −−−→ −iℏ ϕp (x) = p ϕp (x) (3.85)
∂x
We obtain a differential equation for the function ϕp (x). For the momentum eigenfunctions in
the position basis, we assume plane waves, which we know solve the Schrödinger equation:
1 1
ϕp (x) = √ eikx = √ eipx/ℏ (3.86)
2πℏ 2πℏ
It should be noted that ϕp (x) is not an element of the Hilbert space because plane waves are
not spatially limited and therefore not square-integrable (the integral over the entire space
approaches infinity). However, by superimposing plane waves, as shown in Chapter ??, wave
packets with limited extension can be formed, which then do form an element of H. To show
that the Ansatz (3.86) solves the eigenvalue problem, we must substitute into (3.85):
∂ 1 1 1 1
      
−iℏ √ eipx/ℏ = −iℏ − p √ eipx/ℏ = p √ eipx/ℏ □
∂x 2πℏ iℏ 2πℏ 2πℏ
Thus, we have shown that the expression (3.86) represents the momentum eigenfunctions in
the position basis. We can therefore write the eigenfunction |p⟩ of the momentum operator,
expressed in the position basis, as follows:
1
⟨x|p⟩ = ϕp (x) = √ eipx/ℏ (3.87)
2πℏ
Because p = ℏk, we can choose, up to constants, the same ansatz for ϕk (x) (which solves the
eigenvalue equation p̂ ϕk (x) = ℏkϕk (x)):
1
⟨x|k⟩ = ϕk (x) = √ eikx (3.88)
2π
In three-dimensional space, we can choose a similar Ansatz for each:
1
⟨r|p⟩ = ϕp (r) = √ eip·r/ℏ
2πℏ
(3.89)
1
⟨r|k⟩ = ϕk (r) = √ eik·r
2π
As the last point in this section, the eigenfunction |x⟩ of the position operator, expressed in the
momentum or k-basis, is still missing. Because ⟨p|x⟩ = ⟨x|p⟩∗ , we can immediately write this
with a glance at (3.87) and (3.88):
1
⟨p|x⟩ = ϕx (p) = √ e−ipx/ℏ
2πℏ
(3.90)
1
⟨k|x⟩ = ϕx (k) = √ e−ikx
2π
In three-dimensional space, it holds once again analogously:
1
⟨p|r⟩ = ϕr (p) = √ e−ip·r/ℏ
2πℏ
(3.91)
1
⟨k|r⟩ = ϕr (k) = √ e−ik·r
2π
We can already recognize that the momentum eigenfunctions in position space have a great
similarity to the Fourier transformation between momentum and position space. In the next
section, we want to take advantage of this to enable a continuous transformation between the
two eigenbases.

67
 3 Formal Structure of Quantum Theory

3.4.4 Continuous Spectrum and Fourier Transformation

In-Depth: Fourier Transformation

For Fourier transformations (and the corresponding inverse transformations), there are dif-
ferent conventions regarding the prefactors and the sign in the exponent of the exponential
function. In this script, we follow the following conventions:

Transformation from Position Space to Momentum Space (1D):

1 1
Z ∞ Z ∞
ψ̃(p) = √ dx ψ(x) e−ipx/ℏ and ψ̄(k) = √ dx ψ(x) e−ikx (3.92)
2πℏ −∞ 2π −∞

Transformation from Position Space to Momentum Space (3D):

1 1
Z Z
−ip·r/ℏ
ψ̃(p) = 3
d r ψ(r) e and ψ̄(k) = d3 r ψ(r) e−ik·r (3.93)
(2πℏ)3/2 V (2π)3/2 V

Inverse Transformation from Momentum Space to Position Space (1D):

1 1
Z ∞ Z ∞
ψ(x) = √ dp ψ̃(p) e +ipx/ℏ
and ψ(x) = √ dk ψ̄(k) e+ikx (3.94)
2πℏ −∞ 2π −∞

Inverse Transformation from Momentum Space to Position Space (3D):

1 1
Z Z
ψ(r) = 3 +ip·r/ℏ
d p ψ̃(p) e and ψ(r) = d3 k ψ̄(k) e+ik·r (3.95)
(2πℏ)3/2 p3 (2π)3/2 k3

Transformation of the Delta Distribution from Position Space to Momentum

Space:
To determine which result the transformation of the Delta distribution into momentum
space yields, we simply set ψ(x) = δ(x − x0 ), and perform the corresponding integrals from
(3.92):
1 ∞ 1
Z
δ̃(p) = √ dx δ(x − x0 ) e−ip(x−x0 )/ℏ = √ (3.96)
2πℏ −∞ 2πℏ
1 1
Z ∞
δ̄(k) = √ dx δ(x − x0 ) e−ik(x−x0 ) = √ (3.97)
2π −∞ 2π
The Delta distribution transformed into momentum space δ(x − x0 ) therefore yields a con-
stant!

Integral Representation of the Delta Distribution: √ √

By substituting the transformed delta distributions δ̃(p) = 1/ 2πℏ and δ̄(k) = 1/ 2π
into the corresponding inverse transformations (3.94), we obtain integral expressions for
the delta distribution that are often used in practice:
1 ∞ 1 1 ∞
Z Z
δ(x − x0 ) = √ dp √ eip(x−x0 )/ℏ = dp eip(x−x0 )/ℏ (3.98)
2πℏ −∞ 2πℏ 2πℏ −∞
1 1 ik(x−x0 ) 1 ∞
Z ∞ Z
δ(x − x0 ) = √ dk √ e = dk eik(x−x0 ) (3.99)
2π −∞ 2π 2π −∞

Just as in section 3.3.1, a “completeness relation” 1 was formed by summing over all basis vectors
of a discrete CONS – see equation (3.49) – one can also construct a “completeness relation” with
position or momentum eigenfunctions. Since the set of eigenvectors is uncountable in this case,

68
Quantum Theory I

the sum must be replaced by an integral:

Z Z Z
1= dx |x⟩ ⟨x| = dp |p⟩ ⟨p| = dk |k⟩ ⟨k| (3.100)

Similar to the case of discrete eigenfunctions, one can again express an arbitrary state |ψ⟩ in a
certain basis by “inserting a one”:
R
R dx |x⟩ ⟨x|ψ⟩ ,

 (Position space)
|ψ⟩ = 1 |ψ⟩ = dp |p⟩ ⟨p|ψ⟩ , (Momentum space) (3.101)

R dk |k⟩ ⟨k|ψ⟩ , (k-space)


The expressions P̂x = |x⟩ ⟨x|, P̂p = |p⟩ ⟨p|, and P̂k = |k⟩ ⟨k| can again be regarded as projection
operators that project the state |ψ⟩ onto the respective basis vector.

In the following, we briefly demonstrate how one can, for example, prove the identity ⟨x|x′ ⟩ =
δ(x − x′ ) using the Fourier integral representation of the delta function (3.98):
Z +∞ Z +∞
(3.100)
⟨x|x′ ⟩ = ⟨x|1|x′ ⟩ = dk ⟨x|k⟩ ⟨k|x′ ⟩ = dk ⟨k|x⟩∗ ⟨k|x′ ⟩ =
−∞ −∞
1 1
Z +∞ Z +∞
(3.87) ′
= dk ϕ∗x (k)ϕx′ (k) = dk √ eikx √ e−ikx =
−∞ −∞ 2π 2π
1
Z +∞
ik(x−x′ ) (3.98)
= dk e = δ(x − x′ ) (3.102)
2π −∞

At this point, it can also be easily shown that a general state ψ̄(k) is nothing other than the
Fourier transform of ψ(x):
Z +∞ Z +∞
(3.100)
ψ̄(k) = ⟨k|ψ⟩ = ⟨k|1|ψ⟩ = dx ⟨k|x⟩ ⟨x|ψ⟩ = dx ϕx (k)ψ(x) =
−∞ −∞
1
Z +∞ Z +∞
(3.87)
= dx ψ(x)ϕx (k) = √ dx ψ(x)e−ikx (3.103)
−∞ 2π −∞

Conversely, ψ(x) is of course nothing other than the inverse Fourier transform of ψ̄(k):
Z +∞ Z +∞
(3.100)
ψ(x) = ⟨x|ψ⟩ = ⟨x|1|ψ⟩ = dk ⟨x|k⟩ ⟨k|ψ⟩ = dk ϕk (x)ψ̄(k) =
−∞ −∞
1
Z +∞ Z +∞
(3.87)
= dk ψ̄(k)ϕk (x) = √ dk ψ̄(k)eikx (3.104)
−∞ 2π −∞

In-Depth: Action of the Momentum Operator in Position Space

Already in (2.15), the 1D momentum operator p̂, which acts on wave functions in position
space ψ(x), was derived. Here we show another derivation, based on the methods of the
last chapters:
Z ∞ Z ∞
(3.100)
p̂ψ(x) = ⟨x|p̂|ψ⟩ = ⟨x|p̂1|ψ⟩ = dp ⟨x|p̂|p⟩ ⟨p|ψ⟩ = dp ⟨x|p|p⟩ ⟨p|ψ⟩ =
−∞ −∞
1
Z ∞ Z ∞
3.86
= dp p ⟨x|p⟩ ⟨p|ψ⟩ = dp √ p eipx/ℏ ⟨p|ψ⟩ (3.105)
−∞ −∞ 2πℏ
We conduct a side calculation here to simplify the expression from (3.105):

∂ ipx/ℏ ip ipx/ℏ ℏ ∂ ipx/ℏ ∂

e = e =⇒ p eipx/ℏ = e ≡ −iℏ eipx/ℏ (3.106)
∂x ℏ i ∂x ∂x

69
 3 Formal Structure of Quantum Theory

The momentum p can thus simply be replaced by a derivative with respect to position!
We substitute (3.106) into (3.105):

1 1 ∂
Z ∞ ∞
∂ ipx/ℏ
Z
p̂ψ(x) = −iℏ √ dp e ⟨p|ψ⟩ = −iℏ √ dp eipx/ℏ ⟨p|ψ⟩ =
2πℏ−∞ ∂x 2πℏ ∂x −∞
1
Z ∞ Z ∞
∂ 3.86 ∂ (3.100)
= −iℏ dp √ eipx/ℏ ⟨p|ψ⟩ = −iℏ dp ⟨x|p⟩ ⟨p|ψ⟩ =
∂x −∞ 2πℏ ∂x −∞
∂ ∂ ∂
= −iℏ ⟨x|1|ψ⟩ = −iℏ ⟨x|ψ⟩ = −iℏ ψ(x) (3.107)
∂x ∂x ∂x
In three dimensions, this expression can be generalized to the already known expression:

p̂ = −iℏ∇ (3.108)

3.4.5 General Spectral Representation

There are physical systems for which both discrete spectral decomposition and continuous spec-
tral decomposition are needed. For example, consider the possible eigenenergies of an electron
in the Coulomb potential of the nucleus of a hydrogen atom: the bound states correspond to
the discrete eigenvectors of the Hamiltonian operator Ĥ. The unbound (scattering) states cor-
respond to the continuous eigenstates of the Hamiltonian operator. In such a case, to consider
the whole, we must break down the Hamiltonian operator into two parts:
Z
Ĥ = En |n⟩ ⟨n| + (3.109)
X
dE E |ϕE ⟩ ⟨ϕE |
n

|n⟩ represents the (countable) bound energy eigenstates, while |ϕE ⟩ represents the eigenstates
in the continuous spectrum region, thus describing the unbound states.

3.4.6 Direct Sum and Tensor Product of Vector Spaces

In the present case, an operator was decomposed into its discrete spectrum (with an N -dimensional
Hilbert space HN ) and its continuous spectrum (with the distribution space D of continuous
states). To understand how the entire Hilbert space can be described, we look at the generalized
case of two vector spaces V1 and V2 , which are supposed to span a new vector space via the
direct sum:
V = V1 ⊕ V2 (3.110)
The dimension of V is simply obtained via the sum of the dimensions of the individual vector
spaces:
dim V = dim V1 + dim V2 (3.111)
Now, consider two elements from V, namely
! !
u1 ∈ V1 w1 ∈ V1
u= and w = .
u2 ∈ V2 w2 ∈ V2

When forming the scalar product between u and w, we must consider the affiliation to the
respective vector space:
⟨u|w⟩ = ⟨u1 |w1 ⟩ + ⟨u2 |w2 ⟩ (3.112)
| {z } | {z }
∈V1 ∈V2

Alternatively, a vector space V can also be formed using the so-called tensor product between
two vector spaces:
V = V1 ⊗ V2 (3.113)

70
Quantum Theory I

As a concrete example for the use of the tensor product, separable wave functions ψ(x1 , x2 ) can
be used, for which: ψ(x1 , x2 ) = ψ1 (x1 )ψ2 (x2 ). If one wants to calculate the dimension of the
vector space V in this case, it follows:

dim V = dim V1 · dim V2 (3.114)

When calculating the scalar product between u and w, we must form the product between the
elements from the respective vector spaces:

⟨u|w⟩ = ⟨u1 |w1 ⟩ · ⟨u2 |w2 ⟩ (3.115)

| {z } | {z }
∈V1 ∈V2

This can be easily illustrated with the example of a separable function ψ(x1 , x2 ) = ψ1 (x1 )ψ2 (x2 ).
For the projection onto itself, we get:
ZZ
⟨ψ|ψ⟩ = dx1 dx2 ψ ∗ (x1 , x2 )ψ(x1 , x2 ) =
ZZ
= dx1 dx2 ψ1∗ (x1 )ψ2∗ (x2 )ψ1 (x1 )ψ2 (x2 ) =
Z Z
= dx1 ψ1∗ (x1 )ψ1 (x1 ) dx2 ψ2∗ (x2 )ψ2 (x2 ) =
= ⟨ψ1 |ψ1 ⟩ ⟨ψ2 |ψ2 ⟩ (3.116)

We recognize that in the case of the separable function, the elements of the respective sub-vector
space can be naturally separated.

3.5 Operator Algebra

We have already learned about the meaning and functioning of operators, as well as discussed
their properties. Fundamentally, operators can be understood as the mapping of one state to
another state, with both lying in the same vector space. This chapter will explain how to handle
multiple operators and delve deeper into complementary or compatible operators.

Let a general state |ψ⟩ be given, and two arbitrary operators  and B̂ whose effect on |ψ⟩ we
do not know in detail. If we let both operators act on the state, the following notation should
apply:
ÂB̂ |ψ⟩ = Â(B̂ |ψ⟩) and B̂ Â |ψ⟩ = B̂(Â |ψ⟩
The first equation means that  acts on a state that has already been changed by the action of
B̂. The second equation similarly means that B̂ acts on a state that has already been changed
by the action of Â. In general, it holds that ÂB̂ ̸= B̂ Â, and thus the two equations above will
yield different results!

Example: Commutativity of Position and Momentum Operator

Let a state ψ(x) be given, which is represented in the one-dimensional position space.
Likewise, we know the effect of the position and momentum operators x̂ = x and p̂ ≡ p̂x =
∂
−iℏ ∂x in the position basis. If x̂ acts on ψ(x), we obtain the eigenvalue x, while p̂ is first
characterized by the derivative. If we let p̂ act first and then x̂, it results in:
∂
x̂p̂ψ(x) = −iℏx ψ(x)
∂x
The effect of x̂ on the spatial derivative of ψ(x) corresponds again to x. However, in the

71
 3 Formal Structure of Quantum Theory

reverse order, we obtain:

∂ ∂ ∂
   
p̂x̂ψ(x) = −iℏ (xψ(x)) = −iℏ ψ(x) + x ψ(x) = −iℏ 1 + x ψ(x)
∂x ∂x ∂x
The results are obviously different. Calculating the difference x̂p̂ − p̂x̂ and their effect on
ψ(x) yields the important relation:

∂ ∂
 
(x̂p̂ − p̂x̂)ψ(x) = −iℏ x − 1−x ψ(x) = iℏψ(x)
∂x ∂x
We will later use this to identify position and momentum operator as complementary
operators. If we omit the state ψ(x) on which the operators act in the notation, we obtain
a so-called commutator relation:
x̂p̂ − p̂x̂ = iℏ (3.117)

The fact that two operators  and B̂ do not commute in general (ÂB̂ − B̂  ̸= 0) leads us in
the following section to the commutator relationships of operators.

3.5.1 Commutator
That operators generally do not commute leads to the introduction of the commutator. This is
defined over two operators  and B̂:

[Â, B̂] = ÂB̂ − B̂ Â (3.118)

If the commutator between two operators vanishes ([Â, B̂] = 0), Â and B̂ fulfill the commuta-
tivity relation and they can henceforth be referred to as “commutative”. This is equivalent to
the statement that the two operators  and B̂ commute. The commutator is an antisymmetric
function, which can be easily shown by:
[B̂, Â] = B̂ Â − ÂB̂ = −(ÂB̂ − B̂ Â) = −[Â, B̂] (3.119)
The commutator between two identical operators necessarily leads to the zero operator:
[Â, Â] = ÂÂ − ÂÂ = 0 (3.120)
With a third operator Ĉ and the two scalars β and γ, it can also be shown that the commutator
satisfies the distributive law:
[Â, β B̂ + γ Ĉ] = β ÂB̂ + γ ÂĈ − β B̂ Â − γ Ĉ Â =
= β(ÂB̂ − B̂ Â) + γ(ÂĈ − Ĉ Â) =
= β[Â, B̂] + γ[Â, Ĉ] (3.121)
Another useful property of distributivity arises in the commutator of  and the product of two
operators B̂ Ĉ:
[Â, B̂ Ĉ] = ÂB̂ Ĉ − B̂ Ĉ Â =
= (ÂB̂ − B̂ Â + B̂ Â)Ĉ − B̂(Ĉ Â − ÂĈ + ÂĈ) =
= ([Â, B̂] + B̂ Â)Ĉ − B̂([Ĉ, Â] + ÂĈ) =
= [Â, B̂]Ĉ + B̂ ÂĈ − B̂[Ĉ, Â] − B̂ ÂĈ =
= B̂[Â, Ĉ] + [Â, B̂]Ĉ (3.122)
Commutators can also be nested and in the simplest case can even be cyclically permuted in
this way – the Jacobi identity plays an important role here:
[Â, [B̂, Ĉ]] + [Ĉ, [Â, B̂]] + [B̂, [Ĉ, Â]] = 0 (3.123)

72
Quantum Theory I

Example: Proof of the Jacobi Identity

It should be shown that the Jacobi identity from (3.123) is fulfilled. All commutators are
evaluated explicitly with (3.122), otherwise no special assumptions are made:
(3.122)
(I)| [B̂, [Ĉ, Â]] = [B̂, Ĉ Â − ÂĈ] = [B̂, Ĉ Â] − [B̂, ÂĈ] =
= Ĉ[B̂, Â] + [B̂, Ĉ]Â − Â[B̂, Ĉ] − [B̂, Â]Ĉ =
= −([Ĉ, B̂]Â + Â[B̂, Ĉ] + [B̂, Â]Ĉ + Ĉ[Â, B̂])
(3.122)
(II)| [Ĉ, [Â, B̂]] = [Ĉ, ÂB̂ − B̂ Â] = [Ĉ, ÂB̂] − [Ĉ, B̂ Â] =
= Â[Ĉ, B̂] + [Ĉ, Â]B̂ − B̂[Ĉ, Â] − [Ĉ, B̂]Â =
= ÂĈ B̂ − ÂB̂ Ĉ + Ĉ ÂB̂ − ÂĈ B̂ − B̂ Ĉ Â + B̂ ÂĈ − Ĉ B̂ Â + B̂ Ĉ Â =
= −ÂB̂ Ĉ + Ĉ ÂB̂ + B̂ ÂĈ − Ĉ B̂ Â =
= [B̂, Â]Ĉ + Ĉ[Â, B̂]
(3.122)
(III)| [Â, [B̂, Ĉ]] = [Â, B̂ Ĉ − Ĉ B̂] = [Â, B̂ Ĉ] − [Â, Ĉ B̂] =
= B̂[Â, Ĉ] + [Â, B̂]Ĉ − Ĉ[Â, B̂] − [Â, Ĉ]B̂ =
= B̂ ÂĈ − B̂ Ĉ Â + ÂB̂ Ĉ − B̂ ÂĈ − Ĉ ÂB̂ + Ĉ B̂ Â − ÂĈ B̂ + Ĉ ÂB̂ =
= −B̂ Ĉ Â + ÂB̂ Ĉ + Ĉ B̂ Â − ÂĈ B̂ =
= [Ĉ, B̂]Â + Â[B̂, Ĉ]

From the results of the individual terms, it is quickly apparent that the Jacobi identity is
fulfilled, as (I) is compensated by (II) and (III) and it must therefore hold that:

[Â, [B̂, Ĉ]] + [Ĉ, [Â, B̂]] + [B̂, [Ĉ, Â]] = 0

It is also possible to define a function called the anticommutator. It differs from the commutator
by the internal sign:
{Â, B̂} = ÂB̂ + B̂ Â (3.124)
The notation varies in the literature; note that {Â, B̂} and [Â, B̂]+ both symbolize the anticom-
mutator. Unlike the commutator, the anticommutator is a symmetric function:

{Â, B̂} = ÂB̂ + B̂ Â = B̂ Â + ÂB̂ = {B̂, Â} (3.125)

Together with the commutator relation, for an arbitrary product of two operators  and B̂, the
following decomposition can be found:
1  1  1 1
ÂB̂ = 2ÂB̂ + B̂ Â − B̂ Â = ÂB̂ − B̂ Â + ÂB̂ + B̂ Â = [Â, B̂] + {Â, B̂} (3.126)
2 2 2 2

Restricting ourselves again to Hermitian operators  = † and B̂ = B̂ † , we can find new
relationships:

[Â, B̂]† = (ÂB̂)† − (B̂ Â)† = B̂ † † − † B̂ † = B̂  − ÂB̂ = [B̂, Â] = −[Â, B̂] (3.127)

By taking the adjoint of the commutator, we obtain the same commutator except for a sign.
We refer to this property as anti-Hermiticity – unlike a Hermitian operator, an anti-Hermitian
operator only has imaginary eigenvalues.

73
 3 Formal Structure of Quantum Theory

Proof: Imaginary Eigenvalues of an Anti-Hermitian Operator

Let an anti-Hermitian operator Ĉ = −Ĉ † be given. As we have just seen, a commutator of

two Hermitian operators  and B̂ is anti-Hermitian, so we could define such an operator
Ĉ, for example, by the relationship Ĉ = [Â, B̂]. The operator Ĉ is related to the state |ϕ⟩
in the eigenvalue relationship Ĉ |ϕ⟩ = c |ϕ⟩. The eigenvalue c can be calculated directly by:

c = ⟨ϕ|Ĉ|ϕ⟩ . (3.128)

Conjugating and transposing equation (3.128) gives:

(3.128) †
c∗ = ⟨ψ|Ĉ|ψ⟩ = ⟨ψ|Ĉ † |ψ⟩ = |Ĉ † = −Ĉ (anti-Hermitian)
(3.128)
= − ⟨ψ|Ĉ|ψ⟩ = −c (3.129)

For the eigenvalue c it follows that: c∗ = −c. This is only fulfilled if c is a purely imaginary
quantity. This can be verified by assuming that c is a general complex number c = α + iβ
with real part α and imaginary part β:
!
c∗ = −c
(α + iβ)∗ = −(α + iβ)
α − iβ = −α − iβ
only possible if α = 0 and c = iβ is purely imaginary. □ (3.130)

It can also be easily shown that the anticommutator is Hermitian:

{Â, B̂}† = (ÂB̂)† + (B̂ Â)† = B̂ † † + † B̂ † = B̂  + ÂB̂ = {B̂, Â} = {Â, B̂} (3.131)

This implies that we only obtain real eigenvalues for the anticommutator. We have already
provided the corresponding proof in (3.32–3.34).

3.5.2 Complementarity, Canonical Commutation Relation, and Uncertainty Principle

The commutator between the position and momentum operator r̂ and p̂ yields an important
relationship, referred to as the canonical commutation relation. Already in (3.117) the canonical
commutation relation [x̂, p̂] = iℏ in one-dimensional position space was shown. This can be
generalized to three dimensions:
[r̂, p̂] = iℏ (3.132)
If the commutator contains operators that act along the direction of the basis vectors of an
orthonormal system (for example, r̂1 = x̂, r̂2 = ŷ, r̂3 = ẑ or p̂1 = p̂x , p̂2 = p̂y , p̂3 = p̂z ), then the
expression in (3.132) can be further generalized to:

[r̂i , p̂j ] = iℏδij (3.133)

Operators  and B̂ that fulfill the relation (3.133) are referred to as complementary operators:

[Â, B̂] = iℏ (3.134)

Here the statement in (3.127), that the commutator carries purely imaginary eigenvalues, can
be confirmed, as the effect is obviously characterized by the imaginary quantity iℏ. A similar
relationship can also be found in classical mechanics, where instead of the commutator, the
Poisson bracket is used.

74
Quantum Theory I

Motivation: Poisson Brackets in Classical Mechanics

Let an arbitrary observable of the form f (qk , pk , t) be given with k = 1, . . . , N degrees of
freedom, where qk describe the generalized position and pk the generalized momentum.
The time derivatives of these quantities lead to Hamilton’s equations of motion, which are
introduced in (10.2) and have the following form:

∂H ∂H
q̇k = and ṗk = − with k = 1, . . . , N
∂pk ∂qk

These relationships can be used to calculate the total time differential of f (qk , pk , t). It
holds:
N
df ∂f ∂qk ∂f ∂pk ∂f
 
= + + =
X
dt k=1 ∂qk ∂t ∂pk ∂t ∂t
N 
∂f ∂H ∂f ∂H ∂f

= + =
X
−
k=1
∂qk ∂pk ∂pk ∂qk ∂t
∂f
= {f, H} + (3.135)
∂t
We denote {f, H} as the Poisson bracket between the observable and the Hamiltonian. If
we additionally assume that f (qk , pk , t) ≡ f (qk , pk ) does not explicitly depend on time, the
last term ∂f∂t vanishes. If it further holds that the observable is a conserved quantity, the
total differential must also vanish, leading to: {f, H} = 0. If this relationship is fulfilled
for any f (qk , pk ), we can, by reverse, conclude that it is a conserved quantity.

The Poisson bracket can generally be formed between any quantities. We now specifically
consider the generalized position and momentum and find the following relationship:
N  N
∂qi ∂pl ∂qi ∂pl

{qi , pl } = = δik δlk = δil (3.136)
X X
−
k=1
∂qk ∂pk ∂pk ∂qk k=1

We thus obtain a similar form to the canonical commutation relation from (3.133). Fur-
thermore, the analogy also holds for Poisson brackets between identical quantities, which
can be easily compared in the case of commutators with (3.120):

{qi , ql } = {pi , pl } = 0 (3.137)

Let us examine the commutator relationship from (3.134) more closely and restrict ourselves to
two Hermitian operators  = † and B̂ = B̂ † , then the Heisenberg uncertainty relation shown in
(2.80) can now be derived for any operators  and B̂. First, we want to introduce a fluctuation
operator ∆Â, which describes the deviation of the effect from the expected value (mean of many
measurements):
∆ =  − ⟨Â⟩ (3.138)
⟨Â⟩ = ⟨ψ|Â|ψ⟩ is the expected value of  for a given, normalized wave function |ψ⟩. The spread
(or variance) can be calculated as the expected value of the squared fluctuation operator (∆Â)2 ;
the standard deviation is described by the square root of the variance:
q
σA = ⟨(∆Â)2 ⟩ (3.139)

If we now want to obtain the expected value of the variance (∆Â)2 , we can use that the following
general calculation rules apply to expectation values: ⟨X̂ + Ŷ ⟩ = ⟨X̂⟩ + ⟨Ŷ ⟩ and ⟨X̂ Ŷ ⟩ =

75
 3 Formal Structure of Quantum Theory

⟨X̂⟩ ⟨Ŷ ⟩. Furthermore, for the expectation value of an expectation value, it holds: ⟨⟨X̂⟩⟩ = ⟨X̂⟩.
(3.138) 2
⟨(∆Â)2 ⟩ = ⟨( − ⟨Â⟩)2 ⟩ = ⟨( − ⟨Â⟩)( − ⟨Â⟩)⟩ = ⟨Â2 − ⟨Â⟩  −  ⟨Â⟩ + ⟨Â⟩ ⟩ =
2
= ⟨Â2 − 2 ⟨Â⟩ Â + ⟨A⟩2 ⟩ = ⟨Â2 ⟩ − ⟨2 ⟨Â⟩ Â⟩ + ⟨⟨Â⟩ ⟩ =
2 2
= ⟨Â2 ⟩ − ⟨2⟩ ⟨⟨Â⟩⟩ ⟨Â⟩ + ⟨Â⟩ = ⟨Â2 ⟩ − 2 ⟨Â⟩ ⟨Â⟩ + ⟨Â⟩ =
2 2 2
= ⟨Â2 ⟩ − 2 ⟨Â⟩ + ⟨Â⟩ = ⟨Â2 ⟩ − ⟨Â⟩ (3.140)

⟨Â2 ⟩ is called the second moment of Â. Generally, ⟨Âx ⟩ is called the x-th moment of Â, so
⟨Â⟩ = ⟨Â1 ⟩ is the first moment. Without loss of generality, let us now assume an expected value
of 0 for simplification; namely ⟨Â⟩ = ⟨B̂⟩ = 0. Then the variance ⟨(∆Â)2 ⟩ is:

⟨(∆Â)2 ⟩ = ⟨ψ|Â2 |ψ⟩ = ⟨Âψ|Âψ⟩ = || |ψ⟩ ||2 (3.141)

| {z }
Square of
length of Â|ψ⟩

Similar calculation steps can also be carried out for an operator B̂, so that we obtain an expected
variance of ⟨(∆B̂)2 ⟩ = ||B̂ |ψ⟩ ||2 . Let us now try to estimate the product of the variances ⟨(∆Â)2 ⟩
and ⟨(∆B̂)2 ⟩:
(3.18)
⟨(∆Â)2 ⟩ ⟨(∆B̂)2 ⟩ = || |ψ⟩ ||2 ||B̂ |ψ⟩ ||2 ≥
(3.126)
≥ | ⟨Âψ|B̂ψ⟩ |2 = | ⟨ψ|ÂB̂|ψ⟩ |2 =
D 1 1 E 2
= ψ [Â, B̂] + {Â, B̂} ψ =
2 2
1 2
= ⟨ψ| [Â, B̂] |ψ⟩ + ⟨ψ| {Â, B̂} |ψ⟩ = | α = ⟨[Â, B̂]⟩ , β = ⟨{Â, B̂}⟩
4 | {z } | {z }
anti-Herm. Hermitian
1 2 1  
= α + iβ = α2 + β 2 ≥
4 4
1 2 1 2
≥ β = ⟨ψ| [Â, B̂] |ψ⟩ (3.142)
4 4
In the first line of the above derivation, we used the squared Cauchy-Schwarz inequality from
(3.18). In the fourth line, we use that the commutator of Hermitian operators is anti-Hermitian,
and thus according to (3.130) only provides imaginary eigenvalues, while the anticommutator of
Hermitian operators is itself a Hermitian quantity and thus according to (3.131) only provides
real eigenvalues. Taking the square root of the expression, we obtain the uncertainty relation
for initially any Hermitian operators  and B̂:

1
σA σB ≥ ⟨ψ| [Â, B̂] |ψ⟩ (3.143)
2

In the case of operators that represent complementary observables, the relation [Â, B̂] = iℏ
holds. We can substitute this into (3.143), and we get:

1 1 ℏ2
⟨(∆Â)2 ⟩ ⟨(∆B̂)2 ⟩ ≥ |⟨ψ|iℏ|ψ⟩|2 = |iℏ⟨ψ|ψ⟩|2 = (3.144)
4 4 4
By taking the square root on both sides, we finally obtain the Heisenberg uncertainty relation
for complementary observables:

76
Quantum Theory I

ℏ
[Â, B̂] = iℏ =⇒ σA σB ≥ (3.145)
2

3.5.3 Compatibility
Another special relationship that two operators  and B̂ can satisfy in the commutator is:

[Â, B̂] = 0 ⇐⇒ ÂB̂ = B̂ Â (3.146)

In this case, Â and B̂ are interchangeable and are therefore called compatible operators. Op-
erators that satisfy these relations possess a common, complete orthonormal system, and this
statement is also valid in the reverse direction (a common orthonormal system leads to inter-
changeable operators). Compatible operators play an outstanding role in quantum physics and
should therefore be discussed in more detail.

Before we turn to the proof that compatible operators and a common orthonormal system
mutually imply each other, a brief excursion into the notation using an example: Suppose we
have the compatible operators  and B̂ and a common orthonormal system with the basis
vectors {|ϕ1 ⟩ , |ϕ2 ⟩ , |ϕ3 ⟩ , |ϕ4 ⟩ , |ϕ5 ⟩ , |ϕ6 ⟩}. If we let the operator  or B̂ act separately on each
of the basis vectors, we get the following results in our example:

 |ϕ1 ⟩ = a1 |ϕ1 ⟩ B̂ |ϕ1 ⟩ = b1 |ϕ1 ⟩

 |ϕ2 ⟩ = a2 |ϕ2 ⟩ B̂ |ϕ2 ⟩ = b1 |ϕ2 ⟩

 |ϕ3 ⟩ = a2 |ϕ3 ⟩ B̂ |ϕ3 ⟩ = b2 |ϕ3 ⟩

 |ϕ4 ⟩ = a3 |ϕ4 ⟩ B̂ |ϕ4 ⟩ = b1 |ϕ4 ⟩

 |ϕ5 ⟩ = a3 |ϕ5 ⟩ B̂ |ϕ5 ⟩ = b2 |ϕ5 ⟩
 |ϕ6 ⟩ = a3 |ϕ6 ⟩ B̂ |ϕ6 ⟩ = b3 |ϕ6 ⟩

Most of the eigenvalues of both operators are thus degenerate. Neither an eigenvalue ai , nor an
eigenvalue bj uniquely identifies an eigenvector, but the combination of ai and bj does. Because
the operators  and B̂ are compatible (commute), we can let them act consecutively in any
order:

ÂB̂ |ϕ1 ⟩ = B̂ Â |ϕ1 ⟩ = a1 b1 |ϕ1 ⟩

ÂB̂ |ϕ2 ⟩ = B̂ Â |ϕ2 ⟩ = a2 b1 |ϕ2 ⟩

ÂB̂ |ϕ3 ⟩ = B̂ Â |ϕ3 ⟩ = a2 b2 |ϕ3 ⟩

ÂB̂ |ϕ4 ⟩ = B̂ Â |ϕ4 ⟩ = a3 b1 |ϕ4 ⟩

ÂB̂ |ϕ5 ⟩ = B̂ Â |ϕ5 ⟩ = a3 b2 |ϕ5 ⟩
ÂB̂ |ϕ6 ⟩ = B̂ Â |ϕ6 ⟩ = a3 b3 |ϕ6 ⟩

Since certain eigenvalues ai and bj are characteristic for each basis vector, we now want to
introduce the following notation:

|ϕ1 ⟩ ≡ |a1 b1 ⟩ , |ϕ2 ⟩ ≡ |a2 b1 ⟩ , |ϕ3 ⟩ ≡ |a2 b2 ⟩ , |ϕ4 ⟩ ≡ |a3 b1 ⟩ , |ϕ5 ⟩ ≡ |a3 b2 ⟩ , |ϕ6 ⟩ ≡ |a3 b3 ⟩

That means, we now generally refer to a basis vector no longer as |ϕi ⟩, but as |ai bj ⟩. By this,
we mean that the following eigenvalue equations should be satisfied for such a basis vector:

 |ai bj ⟩ = ai |ai bj ⟩ and B̂ |ai bj ⟩ = bj |ai bj ⟩ (3.147)

77
 3 Formal Structure of Quantum Theory

Just as in our example, the following generally applies to compatible operators: If we let both
operators act consecutively on |ai bj ⟩, the eigenstate does not change (because of the compatibility
of the operators):

ÂB̂ |ai bj ⟩ = Âbj |ai bj ⟩ = bj  |ai bj ⟩ = bj ai |ai bj ⟩ = ai bj |ai bj ⟩

(3.148)
B̂ Â |ai bj ⟩ = B̂ai |ai bj ⟩ = ai B̂ |ai bj ⟩ = ai bj |ai bj ⟩

We must only be careful with the summation indices in this notation when we want to represent
a sum over basis vectors. For example, we would simply write a superposition over all basis
vectors {|ϕi ⟩} like this:
6
|Ψ⟩ = c1 |ϕ1 ⟩ + c2 |ϕ2 ⟩ + c3 |ϕ3 ⟩ + c4 |ϕ4 ⟩ + c5 |ϕ5 ⟩ + c6 |ϕ6 ⟩ =
X
ci |ϕi ⟩
i=1

In the newly introduced notation for a common orthonormal system {|ai bj ⟩}, we must instead
write this sum as follows:
3 gX
a (i)

|Ψ⟩ = c11 |a1 b1 ⟩ + c21 |a2 b1 ⟩ + c22 |a2 b2 ⟩ + c31 |a3 b1 ⟩ + c32 |a3 b2 ⟩ + c33 |a3 b3 ⟩ =
X
cij |ai bj ⟩
i=1 j=1

Here, ga (i) is the degeneracy of the eigenvalue ai . Having clarified this, we want to prove the
statement

[Â, B̂] = 0 ⇐⇒ ∃ common orthonormal system

first “from right to left.” We want to show: “If the operators  and B̂ have a common, complete
orthonormal system (CONS), then  and B̂ are compatible.“ To do this, we write  and B̂ in
their spectral representation according to (3.147):

X gX
a (i)
X gX
a (i)

 = |ai bj ⟩ ai ⟨ai bj | and B̂ = |ai bj ⟩ bj ⟨ai bj | (3.149)

i j i j

We now want to specify the products ÂB̂ and B̂ Â in this common eigenbasis in the spectral
representation according to (3.149):
  
ga (i) ga (l)
(3.149)
ÂB̂ =  |ak bl ⟩ bl ⟨ak bl | =
X X X X
|ai bj ⟩ ai ⟨ai bj | 
i j k j

X gX
a (i)
X gX
a (l)
X gX
a (i)

= |ai bj ⟩ ai ⟨ai bj |ak bl ⟩ bl ⟨ak bl | = |ai bj ⟩ ai bj ⟨ai bj | (3.150)

i j k j i j
| {z }
δik δjl
  
ga (i)
(3.149) X X X gX
a (l)

B̂ Â =  |ai bj ⟩ bj ⟨ai bj |  |ak bl ⟩ ak ⟨ak bl | =

i j k j
ga (i) ga (l)
X gX
a (i)

= |ai bj ⟩ bl ⟨ai bj |ak bl ⟩ ak ⟨ak bl | = (3.151)

XXXX
|ai bj ⟩ ai bj ⟨ai bj |
i j k j i j
| {z }
δik δjl

Comparing (3.150) and (3.151), we see: If we assume that there is a common orthonormal system
for the operators  and B̂, then there is no difference between ÂB̂ and B̂ Â, and therefore:
[Â, B̂] = 0.
The proof should now also be carried out in “forward direction”. We want to prove: “If  and

78
Quantum Theory I

B̂ are compatible (commute), then a common, complete orthonormal basis can be found”. So,
we initially assume only that there is an ONBS {|ai ⟩} for the operator  and independently an
ONBS {|bj ⟩} for the operator B̂, such that the following eigenvalue relations hold:

 |ai ⟩ = ai |ai ⟩ and B̂ |bj ⟩ = bj |bj ⟩ (3.152)

For the time being, we do not make any claim about whether these two basis systems are
identical or not. Let’s now consider, starting from the eigenvalue equation for Â, the following
simple calculation:

 |ai ⟩ = ai |ai ⟩ | let B̂ act from the left

B̂ Â |ai ⟩ = B̂ai |ai ⟩ | B̂ Â = ÂB̂, B̂ai = ai B̂
ÂB̂ |ai ⟩ = ai B̂ |ai ⟩ =⇒
Â(B̂ |ai ⟩) = ai (B̂ |ai ⟩) (3.153)

In the last line, we have added (actually unnecessary) parentheses to clarify the following:
Clearly, not only is |ai ⟩ an eigenstate of  for the eigenvalue ai , but also B̂ |ai ⟩ is an eigenstate
of Â, and indeed for the same eigenvalue ai ! We must now discuss two cases:

ai is non-degenerate: If ai is non-degenerate, there is a unique association between the

eigenvalue ai and the eigenvector |ai ⟩. Therefore, if B̂ |ai ⟩ is also an eigenvector of  with
the eigenvalue ai , this is only possible if B̂ |ai ⟩ represents a (complex-valued) stretched or
contracted version of |ai ⟩, so that:

B̂ |ai ⟩ = c |ai ⟩ (3.154)

However, this is a new eigenvalue equation showing that |ai ⟩ is thus also an eigenstate of
the operator B̂ for the eigenvalue c! The state |ai ⟩ is therefore a common eigenvector of
both  and B̂ (which was to be proven).

ai is degenerate: If ai is degenerate, there are two or more eigenvectors for the same
eigenvalue ai . The association between the eigenvalue ai and the corresponding eigenvector
is no longer unique. In other words: There are multiple linearly independent eigenstates,
all lying in the degenerate subspace for the eigenvalue ai . Therefore, the earlier argumen-
tation no longer works. The vector B̂ |ai ⟩ indeed lies in this degenerate subspace, but it is
no longer necessarily just a stretched or contracted version of |ai ⟩.

Thus, we need a different line of proof. First, let’s represent the eigenvector |ai ⟩, which is
part of the degenerate subspace, in the eigenstates of B̂:

|ai ⟩ = (3.155)
X
βj |bj ⟩
j

Here, βj are the expansion coefficients. Let’s consider the eigenvalue equation (3.153)
again:
Â(B̂ |ai ⟩) = ai (B̂ |ai ⟩)

79
 3 Formal Structure of Quantum Theory

Simple transformations now allow us to achieve a clearer picture:

0 = Â(B̂ |ai ⟩) − ai (B̂ |ai ⟩)
(3.155)
= (Â − ai )(B̂ |ai ⟩) =
= (Â − ai )B̂ βj |bj ⟩ =
X

= (Â − ai ) βj B̂ |bj ⟩ = | B̂ |bj ⟩ = bj |bj ⟩

X

= (Â − ai )
X
βj bj |bj ⟩
j

= βj bj (Â − ai ) |bj ⟩
X

j
| {z }
|b′j ⟩

The superposition of the vectors |b′j ⟩ results in a null vector, even though all |bj ⟩ are
linearly independent. To fulfill this equation, the following must hold true: For every
value of the summation index j, either the expansion coefficient βj or the vector |b′j ⟩ must
vanish. However, we can be certain that not all βj vanish since in such a case, due to
(3.155), |ai ⟩ = 0 would result, which is certainly not true. Thus, there are terms in the
sum in which (for certain values of j) βj does not vanish. In these terms, the vector |b′j ⟩
must vanish. This leads us directly to the desired result:
|bj ⟩ = 0 =⇒ (Â − ai ) |bj ⟩ = 0 =⇒ Â |bj ⟩ = ai |bj ⟩ =⇒ Â |bj ⟩ = ai |bj ⟩
Thus, we have found an eigenvalue equation showing that even in the degenerate subspace,
the basis vector |bj ⟩ of the operator B̂ is simultaneously an eigenvector of the operator Â.
The assertion that for compatible operators a common, complete orthonormal system can
be found is thus confirmed in full generality.
These considerations, which at first glance are very theoretical, are of great practical relevance.
When one solves the Schrödinger equation for a specific problem (e.g., the hydrogen atom), one
generally tries to find a complete orthonormal system that can describe a maximum number of
mutually compatible observables:
[Ĥ, Âi ] = 0 and [Âi , Âj ] = 0, ∀i, j
The maximum number of compatible observables determines the maximum amount of infor-
mation that can be determined in a quantum measurement in accordance with the Heisenberg
uncertainty principle. All observables that commute with each other can also be measured
precisely simultaneously!

3.6 Measurements in Quantum Theory and Collapse of the Wave Function

Motivation: The Mystery of Measurement in Quantum Mechanics

We have already shown what happens when a measurement is performed on a system that
is in an eigenstate |ai ⟩ of the operator  associated with the measurement. It is rather
unspectacular: every time, we measure the eigenvalue associated with the eigenstate with
absolute certainty. But what if the system is not in an eigenstate of the operator? Some-
thing strange happens.

Let’s look at a concrete example: Suppose we measure the energy of a state |ψ⟩. The
operator associated with energy measurement is the Hamiltonian operator Ĥ. The eigen-
values (eigenenergies) of our operator are {E1 , E2 , . . . , EN } and the corresponding energy

80
Quantum Theory I

eigenstates are denoted as {|e1 ⟩, |e2 ⟩ , . . . , |eN ⟩}. The state to be measured |ψ⟩ is now a
superposition of the first two energy eigenstates: |ψ⟩ = c1 |e1 ⟩ + c2 |e2 ⟩. Classically, one
would probably expect to measure a total energy E in each measurement, which somehow
lies between the energies E1 and E2 . After all, the state |ψ⟩ is a “mixture” of the corre-
sponding states |e1 ⟩ and |e2 ⟩.
But this is not the case! In each individual measurement on |ψ⟩, the measuring device only
shows either the energy E1 or the energy E2 ‘!T hef actthatthestate|ψ⟩ is a superposition
of the energy eigenstates E1 and E2 is only revealed when numerous measurements are
performed: only after averaging over many measurement results does an average measure-
ment result emerge, which approaches the quantum mechanically calculable expectation
value. One can imagine it as if the system has not yet decided, as long as we have not
measured, which state it is “really” in. Only our measurement creates this reality.

But that’s not all: with the measurement, we have irreversibly destroyed the original state
|ψ⟩! If we (randomly) obtained the measurement value E1 in a measurement process, the
system is no longer in the state |ψ⟩ after the measurement, but in the state |e1 ⟩. If we
happened to obtain the measurement value E2 , the system is in the state |e2 ⟩ after the
measurement. Our quantum system, which was in a superposition of two eigenstates be-
fore the measurement, is forced by the measurement to (randomly) decide on one of the
eigenstates. And our calculations cannot predict which of the two variants will occur. All
that quantum mechanics can do in this case is predict the probability that our measuring
device will display E1 or E2 during a measurement. This reduction of the wave function
is called the collapse of the wave function during the measurement. It is not something
that automatically follows from the calculation, but must be inserted into the calculation
separately. The collapsed wave function must also be renormalized.

Finally, this example also shows the following: In quantum mechanics, measurement and
preparation of an experiment cannot be distinguished. What we have presented above as
a measurement can also be used to prepare a state. If we want to prepare the state |e2 ⟩
in our example, we simply measure the state |ψ⟩ until our measuring device displays the
measurement value E2 , and then use only this state as the prepared state!

It is important to understand the following: If we start from an operator  that corresponds

to a certain physical observable, and also from a quantum system in a general state |ψ⟩, then
quantum mechanics can generally not predict the result of a single measurement. However, it
is possible to calculate the expectation value of the measurement value to which the average
of many measurements on the state |ψ⟩ will converge. The expectation value of the observable
associated with the operator  is usually denoted by ⟨Â⟩. Note: Indicating ⟨Â⟩ by itself only
makes sense if it is clear to which state |ψ⟩ this value relates! The expectation value ⟨Â⟩ can be
calculated as follows:
⟨Â⟩ = ⟨ψ|Â|ψ⟩ (3.156)
To prove that ⟨Â⟩ is a real quantity, one can use the same proof as in (3.32–3.34), replacing |ai ⟩
with |ψ⟩, and ⟨ai | with ⟨ψ|.

In a measurement on a quantum system, only eigenvalues ai of the operator  associated with

the observable are ever measured, even if the system is in a state that can be represented as a
superposition of these eigenstates. The eigenvalues thus describe possible measurement values
of the respective observables, and must therefore be real. The operator  associated with the
observable is therefore Hermitian:  = † or ai ∈ R. We assume that there is a bound state
and thus a discrete spectrum of Â:
 |ϕi ⟩ = ai |ϕi ⟩

81
 3 Formal Structure of Quantum Theory

Now let the state of the system be measured at time t0 . Again, we must distinguish between
the two cases of a degenerate and non-degenerate system. For simplicity, we will first examine
the non-degenerate case.

3.6.1 Measurement on a Non-Degenerate System

We assume that before the measurement, there is a general wave function |ψ(t)⟩, which we can
represent as a coherent superposition of the eigenstates of any operator. For practical purposes,
we choose the eigensystem of the measurement operator Â:

|ψ(t)⟩ = |ϕi ⟩ ⟨ϕi |ψ(t)⟩ = ci (t) |ϕi ⟩ (3.157)

X X

i i

|ψ(t0 )⟩
|ψ(t1 )⟩

|ψ(0)⟩
|ψ(t0 + ε)⟩ = |ϕi ⟩

t
0 t0 t1

Abb. 15: A measurement of the observable  is performed on a system in the state |ψ(t)⟩ at
time t0 . Owing to the “intervention” of the measurement process, |ψ(t)⟩ collapses into
an eigenstate |ϕi ⟩ of the measurement operator. The Schrödinger equation describes
the propagation of the state up to time t = t0 . After the measurement, the “collapsed”
state can continue to propagate.

At the time of measurement t0 , this state of the superposition “collapses.” The measurement
results in the collapse of the wave function to an eigenstate |ϕi ⟩ of the measurement operator.
Therefore, it holds:
|ψ(t0 )⟩ = |ϕi ⟩
In general, it is not possible to predict to which eigenstate |ϕi ⟩ the wave function will collapse
in a single experiment! One can always only calculate the probability for a particular eigenstate
|ϕi ⟩ for a given experiment.

If the measurement operator is non-degenerate, each possible eigenstate of  is associated with

a unique eigenvalue (i.e., measurement value), namely ai = ⟨ϕi |Â|ϕi ⟩ = ⟨ψ(t0 )|Â|ψ(t0 )⟩. There
is no uncertainty immediately after the measurement; the state after the collapse of the wave
function can be uniquely determined by the eigenvalue ai . If the measurement process lasts a
short duration ∆t = ε, we can continue to assume after this short duration:

|ψ(t0 + ε)⟩ = |ϕi ⟩

However, the state is changed again by the influence of the system after the measurement.
Mathematically speaking, the Schrödinger equation, according to its Hamiltonian operator Ĥ,
dictates the time evolution of the state.
For conducting a quantum-mechanically significant measurement experiment, it is necessary
that we can perform a large number of measurements on the same initial state |ψ(t)⟩. This
is possible by repeatedly performing the same experiment (with the same preparation of the

82
Quantum Theory I

initial state). This allows an interpretation of the measured observables in a statistical sense.
If we were to measure only once, we could claim that our superposed state from (3.157) was
already in the measured state before the measurement! Only repeated measurements allow for
the experimental determination of the probability pi , that the general state |ψ(t)⟩ collapses into
the eigenstate |ϕi ⟩ at the time of measurement t0 . The probability is given by:

pi = | ⟨ϕi |ψ⟩ |2 = |ci (t)|2 (3.158)

Measuring once is therefore generally not meaningful. Only repeated measurements enable the
application of statistical methods so that we can, for example, assign a higher probability to
measurement values that occur more frequently than those observed less frequently.

N =1 N =3

N = 16 N = 38

Abb. 16: An ensemble of similar states is measured: Each measurement of an observable leads
to the collapse of the wave function and the identification of the state. Only repeated
execution of the same measurement (by multiple repetitions of the experiment) allows
for a statistical interpretation of the measurement.

If we know the probabilities with which a state |ϕi ⟩ will occur during a measurement, we can
calculate the expectation value of the observable Â:

⟨ψ|Â|ψ⟩ = ⟨ψ|1Â1|ψ⟩ =
= ⟨ψ|ϕi ⟩ ⟨ϕi |Â|ϕj ⟩ ⟨ϕj |ψ⟩ =
X

ij

= c∗i cj aj ⟨ϕi |ϕj ⟩ =

X

ij

= c∗i cj aj δij =
X

i
(3.158)
= =
X
|ci |2 ai
i
= (3.159)
X
pi ai
i

The expectation value is therefore a weighted average of the eigenvalues ai of the observable
Â. It is clearly distinct from the measured value in each individual measurement: While the
measurement result in each individual measurement can only be one of the eigenvalues {ai } of
the measurement operator Â, the expectation value arises from a multitude of measurements
and can numerically lie between the discrete eigenvalues.

(3.159) can be written more concisely if we directly include beforehand that  is Hermitian and
it necessarily follows that ⟨ϕi |ϕj ⟩ = 0. We can thus directly neglect the sum over j and simply

83
 3 Formal Structure of Quantum Theory

state the operator in its spectral representation:

⟨ψ|Â|ψ⟩ = ⟨ψ|ai |ϕi ⟩ ⟨ϕi |ψ⟩ = ai ⟨ψ|P̂i |ψ⟩ =

X X X
ai p i
i i i

Using the projection operator ⟨ψ|P̂i |ψ⟩ = ⟨ψ|ϕi ⟩ ⟨ϕi |ψ⟩ = |ci |2 , the expectation value of all
possible eigenvalues (measurement values) at time t0 can be briefly written as:

⟨ψ|Â|ψ⟩ = ⟨ψ|ai P̂i |ψ⟩ = (3.160)

X X
pi ai
i i

The expectation value of many measurements therefore corresponds to the averaging of the
possible eigenvalues ai , weighted with the probabilities pi for each eigenvalue.

3.6.2 Measurement on a Degenerate System

Now consider the more complex case with additional degeneracy, this degree of degeneracy g
must be taken into account (i.e., how often a state is degenerate). The eigenvalue equation for
the states |ϕi ⟩ must therefore be extended by the degenerate wave functions:

 |ϕni ⟩ = ai |ϕni ⟩ (3.161)

For an eigenvalue ai , there exist several eigenstates |ϕni ⟩, whose degeneracy is indicated by
the additional index n = 1, . . . , g. Even in the projector, the degeneracy must be considered,
resulting in a new definition:
g
P̂in = |ϕni ⟩ ⟨ϕni | where P̂in = P̂i (3.162)
X

n=1

Applying this to a state |ψ⟩ yields: P̂in |ψ⟩ = cni |ϕni ⟩. If we expand the wave function |ψ(t0 )⟩ ≡
|ψ⟩ into the eigenstates of the observable Â, we obtain an expression analogous to (3.159):
g g
|ψ(t0 )⟩ ≡ |ψ⟩ = 1 |ψ⟩ = P̂in |ψ⟩ = (3.163)
XX XX
cni |ϕni ⟩
i n=1 i n=1

The expectation value of P̂i at time t = t0 corresponds to the probability that the system is in
a state with eigenvalue ai . Accounting for degeneracy, we obtain:
g
⟨ψ|P̂i |ψ⟩ =
XX
|cni |2
i n=1

If we again measure the expectation value of the observable Â, we can proceed as in (3.160),
but must take into account the additional summation over all degeneracies:
g g g
⟨ψ|Â|ψ⟩ = = = (3.164)
XX XX XX
⟨ψ|ai P̂in |ψ⟩ ai |cni |2 ai pni
i n=1 i n=1 i n=1

pni is intended to express the probability that the system is in a state ai with the degeneracy n.

3.6.3 Measurement on a Continuous System

Let’s briefly step away from discrete states and turn to a continuous system. For example, for
the one-dimensional position operator x̂, the eigenvalue equation x̂ |x⟩ = x |x⟩ from (3.80) holds.
Again, there is initially a general superposition state |ψ⟩ that is to be measured at a certain

84
Quantum Theory I

time. The considered observable is the position x̂, whereby the expectation value after many
measurements is determined by:
(3.100)
⟨x̂⟩ = ⟨ψ|x̂|ψ⟩ = ⟨ψ|1x̂1|ψ⟩ =
Z Z
= dx dx′ ⟨ψ|x⟩ ⟨x|x̂|x′ ⟩ ⟨x′ |ψ⟩ = | x̂ |x′ ⟩ = x′
Z Z
(3.77)
= dx dx′ x′ ⟨ψ|x⟩ ⟨x|x′ ⟩ ⟨x′ |ψ⟩ =
Z Z
= dx dx′ x′ δ(x − x′ ) ⟨ψ|x⟩ ⟨x′ |ψ⟩ =
Z Z
= dx x ⟨ψ|x⟩ ⟨x|ψ⟩ = dx x | ⟨x|ψ⟩ |2 =
Z
= dx x |ψ(x)|2 (3.165)

At the end of the first line, (3.100) was used. The expectation value of x̂ thus corresponds to
the weighted mean of x with the probability density |ψ(x)|2 .

3.6.4 Collapse of the Wave Function and Compatible Operators

In the case of a non-degenerate, discrete system, the state |ψ⟩ changes to the collapsed state
|ϕi ⟩ through measurement:

measurement P̂i |ψ⟩ P̂i |ψ⟩

|ψ⟩ −−−−−−−−→ |ϕi ⟩ = q =q with P̂i = |ϕi ⟩ ⟨ϕi |
⟨ψ|P̂i2 |ψ⟩ ⟨ψ|P̂i |ψ⟩

To obtain a normalized wave function, one divides by the norm, exploiting the fact that P̂i is
2
idempotent (i.e., it holds that: P̂i = P̂i ).
For the degenerate case, a measurement leads to a projection onto the multidimensional (degen-
erate) subspace associated with the measurement value ai :
g
measurement P̂i |ψ⟩
with P̂i =
X
|ψ⟩ −−−−−−−−→ q P̂in
⟨ψ|P̂i |ψ⟩ n=1

Let’s consider the case where two compatible operators  and B̂ exist, which satisfy the relation
[Â, B̂] = 0. It follows that for both observables, a common, complete orthonormal system
{|ai bj ⟩} must exist. A new projection operator, the product projector P̂ij , can thus be defined
again:
P̂ij = |ai bj ⟩ ⟨ai bj | (3.166)
The eigenvalues of P̂ij are determined by pij = ⟨ψ|P̂ij |ψ⟩ = ⟨ψ|ai bj ⟩ ⟨ai bj |ψ⟩, where pij can be
interpreted as the product probability. Furthermore, it must hold that due to completeness, the
following relation is fulfilled:
1= P̂ij = (3.167)
X X
|ai bj ⟩ ⟨ai bj |
ij ij

The summation indices represent the respective affiliation to an eigenvalue of the corresponding
observable. Due to the compatibility of  and B̂, both observables can be measured simul-
taneously, and the independent eigenvalues ai and bj are obtained. According to (3.160), the
expectation value of the observable  can be quickly written as:
gj
X 
⟨Â⟩ = ai pij = = (3.168)
X X X
ai pij ai pi
ij i j i

85
 3 Formal Structure of Quantum Theory

ai is independent of the sum over j and can therefore be factored out. If the product probability
pij is summed over all states of B̂, a “new” variable pi can be defined as a reduced probability
which implicitly already contains that sum. The same applies to the expectation value of B̂:
gi
X 
⟨B̂⟩ = bi pij = = (3.169)
X X X
bj pij bj pj
ij j i j

Here the corresponding summation index changes, so that in the end only the summation over
the reduced probability pj remains.

86
Quantum Theory I

4 Harmonic Oscillator
Motivation: The fundamental role of the harmonic oscillator in physics
The concept of the harmonic oscillator, where the restoring force increases linearly with the
displacement for arbitrarily large displacements, is of fundamental importance for physics.
Many potentials that have a local minimum can be well approximated by a harmonic
potential and thus described analytically. This allows us to approximate many systems
very well if only small displacements from the equilibrium position are considered. The
concept of the harmonic oscillator also has central importance in the quantization of the
electromagnetic field in quantum electrodynamics (QED).

4.1 Analytical Solution

4.1.1 The classical harmonic oscillator

t0 t1 t2
t0 t2
t
V t1

k=m=1

Abb. 17: (left) Model for the spring pendulum: As time progresses, the mass m describes a
trajectory x(t) from (4.3). (right) The motion in the quadratic oscillator potential
(4.10) leads to the same trajectory (in the illustration, this sine curve corresponds to
the height in the potential).

As an introduction, we recap the classical harmonic oscillator using the example of the one-
dimensional spring pendulum, which is displaced in the x-direction. Its motion is determined
by the second Newton’s law F = ma:
d2 x(t)
F (t) = m (4.1)
dt2
The assumed restoring force F (t) = −kx(t) increases linearly with the displacement x(t). The
linearity factor k is the spring constant and always acts against the current displacement (hence
the negative sign). Thus the equation of motion is:

d2 x(t)
−kx(t) = m (4.2)
dt2
We can define the angular frequency ω as ω = k/m. The differential equation (4.2) can be
p

solved with the following ansatz:

x(t) = A sin(ωt) + B cos(ωt) (4.3)

√
Defining the amplitude C = A2 + B 2 and the phase angle φ = arctan(A/B), the relation (4.3)
is equivalent to the ansatz:
x(t) = C cos(ωt − φ) (4.4)

87
 4 Harmonic Oscillator

But for now, we stick with the ansatz (4.3). Based on this, the velocity v(t) follows for the first
time derivative:
v(t) = ẋ(t) = ω [A cos(ωt) − B sin(ωt)] (4.5)
Choosing as initial conditions the lower turning point −xu of the oscillation (xu > 0), so that
x(0) = −xu and v(0) = 0. Then (4.3) and (4.5) simplify to:

x(t) = −xu cos(ωt) (4.6)

v(t) = ωxu sin(ωt) (4.7)

Equation (4.7) can be rewritten to generate a dependence on the positional coordinate:

q
v(t) = ωxu sin2 (ωt) =
q
= ωxu 1 − cos2 (ωt) =
q
= ω x2u − x2u cos2 (ωt) =
(4.6)
q
= ω x2u − (−xu cos(ωt))2 =
q
= ω x2u − x2 (t)

Considering a large number of such harmonic oscillators, the statistical probability P (x) of
finding an oscillator at a certain position x can be obtained. The trick here is to convert
the corresponding probability P (t) of finding the oscillator at a certain time t within a period
between t = 0 and t = T into P (x). In particular, since P (t) = 1/T , we obtain:

dt dx dx
P (t)dt = = = p 2 = P (x)dx (4.8)
T v(x)T 2π xu − x2

Thus, we find the result that the probability of finding the particle at a specific location x is
inversely proportional to the speed of the particle – the probability of measuring the particle is
greatest at the two turning points. Later we will see that the probabilities of finding calculated
wave functions in the classical limit will approximate the form of P (x)!

4.1.2 Solution of the Schrödinger equation with the potential of the harmonic oscillator
To solve the Schrödinger equation, we need, instead of the restoring force F = −kx, an equiva-
lent potential. This leads us to the one-dimensional harmonic oscillator, whose potential V (x)
is:

1
V (x) = mω 2 x2 (4.9)
2

The force corresponds to the negative gradient of the potential – we recognize that from V (x)
the spring force F = −kx follows:

dV d 1 k
 
F =− =− mω 2 x2 = −mω 2 x = −m x = −kx □ (4.10)
dx dx 2 m

If we insert this potential into the time-independent, one-dimensional Schrödinger equation (2.8)
in the position representation, we get:

1
!
ℏ2 d2
− + mω 2 x2 ψ(x) = Eψ(x) (4.11)
2m dx2 2

88
Quantum Theory I

Transformation to the dimensionless Schrödinger equation To simplify the Schrödinger equa-

tion, we introduce the following dimensionless quantities („reduced units“) ε and y for the energy
E and the position coordinate x:

E = ℏωε (4.12)
x = x0 y (4.13)

The purpose of this transformation is to bring the Schrödinger equation to a form that is
independent of all units. For position and momentum, we need the conversion factors x0 and
p0 :
s
ℏ
x0 = (4.14)
mω
ℏ √
p0 = = mℏω (4.15)
x0
Since we substitute x with y, we must adjust both the differentials dx and the wave function
ψ(x). From (4.13), it follows:
dy 1
= (4.16)
dx x0
The substitution of the wave function ψ(x) is performed by changing the dependency; we swap
x to y(x), so that ψ(x) → ψ(y(x)) ≡ ψ(y). It follows:

d2 d dψ(y) df(y) df(y) df(y) dy

2
ψ(y) = = = =
dx dx | dx
{z } dx dx dy dx
f(y)

df(y) dy (4.16) 1 df(y) 1 d dψ(y) dψ(y) dψ(y) dy

= = = = =
dy |{z}
dx x0 dy x0 dy dx dx dy dx
1/x0

1 d  dψ(y) dy  (4.16) 1 d  dψ(y) 1  1 d2

= = = ψ(y) (4.17)
x0 dy dy dx
|{z} x0 dy dy x0 x20 dy 2
1/x0

We can now insert (4.12), (4.13), and (4.17) into the Schrödinger equation from (4.11):

ℏ2 1 d2 1
!
(4.14)
ℏωεψ(y) = − + mω 2 x20 y 2 ψ(y) =
2m x0 dy
2 2 2
1
!
ℏ2 mω d2 ℏ 2
= − + mω 2 y ψ(y) =
2m ℏ dy 2 2 mω
!
ℏω d2 ℏω 2
= − + y ψ(y)
2 dy 2 2

On both sides ℏω can be cancelled – thus obtaining a compact, dimensionless representation of

the Schrödinger equation:
1
!
d2
− 2 + y ψ(y) = εψ(y)
2
(4.18)
2 dy
Before we actively solve this equation, we attempt to learn something about the structure of
the solution in advance: The potential of the harmonic oscillator (4.9) is symmetric around
x = 0. Thus, we can expect the wave functions of the stationary (bound) states of the system
to exhibit the same symmetry behavior. Formally, this follows from the fact that the parity
operator Π̂ commutes with the Hamilton operator Ĥ, causing the solution functions of Ĥ to

89
 4 Harmonic Oscillator

also be eigenstates of Π̂. According to the node rule, we can again arrange the eigenstates ψ n (y)
and eigenenergies εn by the number of “nodes” n.
In the limit x → ±∞, the V (x) is infinitely high – as with the finite deep potential well, the wave
functions will also be able to penetrate into the “classically forbidden region” of the potential
for arbitrary, smaller values of x, but they will decrease exponentially there.

Solution of the Ground State With these considerations, we can motivate the following ap-
proach for the dimensionless Schrödinger equation (4.18):

1 2
ψ(y) = N e− 2 y h(y) (4.19)

We assume a Gaussian function that meets the required symmetry and decay behavior for
y → ±∞. For h(y) we initially assume an arbitrary power series:

∞
h(y) = (4.20)
X
ai y i
i=0

Perhaps the simplest case is the special case h(y) = 1 (with a0 = 1 and all other ai = 0), which
results in a pure Gaussian function:
1 2
ψ 0 (y) = N e− 2 y (4.21)

Substituting this wave function approach into the Schrödinger equation, we can determine the
associated eigenvalue ε0 . Since the wave function (4.21) has no nodes (the Gaussian function is
positive for all x, hence n = 0), we can associate ε0 with the ground state energy. It follows:
!
− 12 y 2 N d2 1 2
ε0 N e = − 2 + y 2 e− 2 y =
2 dy
N d 
 
1 2 1 2

= − −y e− 2 y + y 2 e− 2 y =
2 dy
N  1 2 1 2 1 2
 1 1 2
= e− 2 y − y 2 e− 2 y + y 2 e− 2 y = N e− 2 y
2 2

The wave function ψ 0 (y) cancels on both sides. We thus obtain the reduced ground state energy
ε0 = 1/2. With (4.12) we can back-substitute to the actual energy value and find:

1
E0 = ℏω (4.22)
2

Unlike in classical physics, the harmonic oscillator in quantum physics has a ground state energy
that lies above the minimum of the potential V (0) = 0.

In-Depth: Ground State with Zero Energy E0 = 0

If the harmonic quantum oscillator could have a ground state with zero energy, then
the position of the particle “at the bottom of the potential” would be precisely defined.
However, without energy, the momentum must also vanish, meaning that both position
and momentum would be sharply defined. This, however, is fundamentally impossible
according to Heisenberg’s uncertainty principle. Therefore, in quantum physics, there can
be no harmonic oscillator with a ground state energy of zero!

90
Quantum Theory I

General Solution Approach To obtain all solutions, we must substitute the general solution ap-
proach (4.19) into the dimensionless Schrödinger equation (4.18). To do this, we first determine
the second derivative of the general approach for ψ(y):

1 d2 d2 h − 1 y2 i d h ′
− 1 y2 i
ψ(y) = e 2 h(y) = h (y) − yh(y) e 2 =
N dy 2 dy 2 dy
1 2 1 2
h i h i
= h′′ (y) − yh′ (y) − h(y) e− 2 y − h′ (y) − yh(y) ye− 2 y =
1 2
h i
= h′′ (y) − 2yh′ (y) − (1 − y 2 )h(y) e− 2 y (4.23)

Now we substitute the approach (4.19) and the result (4.23) into the Schrödinger equation (4.18)
to obtain a new differential equation of the function h(y):

1 2 N h   1 2 2 /2
i
εN e− 2 y h(y) = − h′′ (y) − 2yh′ (y) − (1 − y 2 )h(y) e− 2 y + y 2 h(y)e−y =
2
N h i 2 /2
= −h′′ (y) + 2yh′ (y) + (1 − y 2 )h(y) + y 2 h(y) e−y =
2
N 2 /2
= −h′′ (y) + 2yh′ (y) + h(y) e−y
 
2
Reducing the Gaussian function terms on both sides of the equation gives us the so-called
Hermite differential equation:

−h′′ (y) + 2yh′ (y) + (1 − 2ε)h(y) = 0 (4.24)

It is now up to us to solve this Hermite differential equation (4.24). We already know the
development coefficient for n = 0 based on our assumption that the ground state is described
by a Gaussian function: a0 = 1. As mentioned in (4.20), we use a power series approach for the
solution of the Hermite differential equation for the function h(y). To substitute this approach
into equation (4.24), we must first compute both derivatives h′ (y) and h′′ (y):
∞
d X d  
h′ (y) = ai y i = a0 + a1 y + a2 y 2 + a3 y 3 + . . . =
dy i=0 dy
∞
= 1a1 + 2a2 y + 3a3 y 2 · · · = (4.25)
X
nai y i−1
i=1
′′ d ′ (4.25) d
 
h (y) = h (y) = 1a1 + 2a2 y + 3a3 y 2 + . . . =
dy dy
∞
= 2 · 1a2 + 3 · 2a3 y + · · · = i(i − 1)ai y i−2 (4.26)
X

n=i

Special attention must be paid to the initial summand and its index due to changes depending
on the degree of differentiation, avoiding negative exponents (here to i = 1 or i = 2). Now we
substitute our Ansatz (4.20) and the derived power series (4.25) and (4.26) into the Hermite
differential equation (4.24):
∞ ∞ ∞
i(i − 1)ai y i−2 + 2y nai y i−1 +(1 − 2ε) ai y i = 0 (4.27)
X X X
−
i=2 i=1 i=0
| {z } | {z }
(I) in (4.28) (II) in (4.29)

We want to modify all y terms in the sums, especially in the (I) and (II) terms, to have the
same power y i to combine everything under a single sum starting at i = 0. In the second sum
(II), this can be easily achieved by multiplying the term y inside the sum. The sum can now

91
 4 Harmonic Oscillator

begin at i = 0 instead of i = 1, as the factor of i results in each summand being naturally zero,
avoiding negative exponents:
∞ ∞  
(II)| 2y iai y i−1 = 2 nai y i = 2 1 · a1 y + 2 · a2 y 2 + . . . =
X X

i=1 i=1
  ∞
= 2 0 · a0 + 1 · a1 y + 2 · a2 y 2 + . . . = 2 (4.28)
X
iai y i
i=0

Note that the inclusion of i ensures the term a0 vanishing at i = 0. We can similarly modify the
first sum (4.29):
∞ ∞
(I)| i(i − 1)ai y i−2 = 2 · 1 a2 y 0 + 3 · 2 a3 + · · · = (j + 2)(j + 1) aj+2 y j (4.29)
X X

i=2 j=0

For clarification, the index i starting at zero can be reintroduced on the right-hand side of
equation (4.29), allowing us to re-express it in equation (4.27):
∞ ∞ ∞
0=− (i + 2)(i + 1)ai+2 y i + 2 iai y i + (1 − 2ε) ai y i =
X X X

i=0 i=0 i=0

∞ ∞
= [−(i + 2)(i + 1)ai+2 + (2i + 1 − 2ε)ai ] y i = (4.30)
X X
Pi y i
i=0 i=0
| {z }
Pi

Equation (4.30) holds only if each individual polynomial Pi is zero, as all separate powers y i are
linearly independent. By setting Pi = 0 in (4.30) and solving for ai+2 , we obtain the following
recursion relation:
2i + (1 − 2ε)
ai+2 = ai (4.31)
(i + 2)(i + 1)
With this recursion, we can determine the polynomials h(y). For example, assuming (without
normalization): a0 = 1. Then we can determine all even expansion coefficients (a2 , a4 , a6 , . . . )
using (4.31). If a1 is known, then the odd expansion coefficients (a3 , a5 , a7 , . . . ) can be deter-
mined.
Due to the symmetry properties of the oscillator potential V (x), we expect ψ(y) to have both
even and odd solutions (but no solutions with mixed symmetry). Due to the definition of ψ(y)
2
in (4.21), and the fact that e−y /2 has even parity, the polynomial h(y) determines whether ψ(y)
is even or odd: If a0 = 0, then all ai with even i are zero, making h(y) and ψ(y) have odd parity.
If a1 = 0, then all ai with odd i are zero, making h(y) and ψ(y) have even parity.

We now consider whether the series in h(y) can truly have infinitely many terms without causing
wave function divergence for y. We examine the asymptotic behavior of ai as i → ∞. The ratio
of coefficients ai+2 /ai initially yields:

ai+2 (4.31) 2i 1 − 2ε 2i 1 − 2ε
= + = 2 +
ai (i + 2)(i + 1) (i + 2)(i + 1) i + 3i + 2 (i + 2)(i + 1)

While the second term will clearly approach zero as n → ∞, this is not true for the first term.
Here, the quadratic term dominates in the denominator:
ai+2 n→∞ 2i 2
−−−→ 2 = (4.32)
ai i i
Finding a function having the same behavior in the limit i → ∞ allows us to associate this with
h(y) and analyze whether the entire, dimensionless wave function ψ(y) converges or diverges for

92
Quantum Theory I

very large y. As y increases indefinitely, the series term y i with larger i dominates. To verify
this, consider a Gaussian function expressed as a series:
∞ ∞ ∞ ∞
2 1  2 k X 1 2k 1
ey = = y = yi = (4.33)
X X X
y bi y i
k=0
k! k=0
k! i=0,2,4,...
(i/2)! i=0,2,4,...
| {z }
bi

The counting variable i in the last sum only takes even number values. Examining the conver-
gence behavior:
bi+2 (i/2)! (i/2)! 1 2 i→∞ 2
= = = = −−−→
bi (i/2 + 1)! (i/2 + 1)(i/2)! (i/2 + 1) i+2 i
It becomes clear that the Gaussian function demonstrates the same convergence behavior for
large i as the recursion found in (4.32). If the series does not cease, it indicates that h(y) behaves
like a Gaussian function from (4.33) for large i. This would imply the following problem:
2 /2 i→∞ 2 /2 2 2 /2
ψ(y) = e−y h(y) −−−→ e−y ey = ey (4.34)
Yet, this function is not normalizable, implying divergence for y → ±∞ (contrary to our initial
analysis of the potential). This necessitates finding solutions where the infinite series in the
Ansatz (4.20) terminates!

n=4

n=3
|ϕ|2

n=2
ϕ

V (x) V (x)
n=1

n=0

x x

Abb. 18: Representation of the wave functions (left) and the probabilities (right) of the har-
monic quantum oscillator. The first five energy states are shown; the symmetry of
the eigenstates and the correlation between the number of nodes and energy level are
clearly evident.

This circumvents the divergence issue because a finite series can no longer be equated to an
analytic function like (4.33). If the series (4.20) terminates, there must be a largest value n for
the index i. Setting this value n into the recursion relation (4.31) shows an+2 = 0 to ensure the
series stops. Consequently, all subsequent ai vanish when i > n. Setting this n in the recursion
relation (4.31), where an+2 = 0, defines the (dimensionless) eigenenergy εn such that the series
terminates at index i = n:
! (4.31) 2n + (1 − 2εn ) 1
ai+2 = 0 =⇒ = 0 =⇒ 2n + 1 − 2εn = 0 =⇒ εn = n + (4.35)
(n + 2)(n + 1) 2
Using the substitution E = ℏω from (4.12), the eigenenergies En arise:

93
 4 Harmonic Oscillator

1
 
En = ℏω n + (4.36)
2
Thus, the harmonic quantum oscillator has equidistant eigenenergies! Using the energy quantum
number n, we can hierarchically organize individual eigenstates and energies. Recall this applies
to the nodes of each wave function as much as energy levels; the greater the energy, the more
nodes a wave function manifests (thus stronger oscillation).

Eigenfunctions of the Harmonic Oscillator The obtained expression for the possible (reduced)
eigenenergies (4.35) can now be substituted into the differential equation (4.24):
(4.35)
0 = −h′′ (y) + 2yh′ (y) + (1 − 2εn )h(y) =
1
  
= −h′′ (y) + 2yh′ (y) + 1 − 2 n + h(y) =
2
= −h′′ (y) + 2yh′ (y) + (1 − 2n − 1) h(y) = | · (−1)
= h′′ (y) − 2yh′ (y) + 2nh(y) (4.37)
This differential equation (4.37) is solved by our power series Ansatz (4.20), assuming a specific
(reduced) eigenenergy εn of the n-th excited state based on (4.35), and then determining the
coefficients ai of the power series using the recursion relation (4.31) (normalization not included).
This is equivalent to the solution h(y) = Hn (y) for the n-th eigenenergy εn , where Hn (y) denotes
the n-th Hermite polynomial, which can be defined via the Rodrigues formula:
n
2 d 2
Hn (y) = (−1)n ey e−y (4.38)
dy n
A detailed derivation of the Hermite polynomials can be found in Appendix 10.2. By inserting
h(y) = Hn (y) in Ansatz (4.19), we finally obtain the (dimensionless) solution:
1 2
ψ n (y) = N e− 2 y Hn (y) (4.39)
With the back substitution from (4.13), we get the final result for the position-dependent wave
function ψn (x). The normalization N is determined by the Hermite polynomials:
s
1 1 x
 
1 2 ℏ
ψn (x) = 1/4 √ e− 2 (x/x0 ) Hn with x0 = (4.40)
π x0 2 n!
n x0 mω

4.1.3 Limits of the Harmonic Oscillator

In the framework of the classical calculation of the harmonic oscillator, we encountered in (4.8)
a correlation of the probability proportional to P (x) ∝ (xu − x)−1/2 . A particle in a classical
oscillator potential will therefore be least likely found in the potential minimum, but mainly
near the two turning points of its motion. The question arises as to how probability behaves in
quantum theory.

The state with the most distinct “quantum-character” is the ground state with n = 0, where the
probability is maximal at the potential minimum: This is in complete contradiction to classical
intuition. Likewise, when energy increases and the principal quantum number n becomes larger,
the likelihood of finding the particle in the potential minimum decreases, and we realize we
approach the classical limit. A schematic representation of these states is shown in Figure
19. Here, although we still see a significantly higher probability around x ≈ 0, the behavior
of P (x) increasingly approaches the classical expression from (4.8). This behavior is found in
other quantum systems: With increasingly large quantum numbers, the system exhibits more
“classical-like” behavior (another example is the Rydberg state for very high principal quantum
numbers n of the hydrogen atom, which can be approximately described classically).

94
Quantum Theory I

P (x) n=0 P (x)

|ϕ|2

|ϕ|2
n = 15

x x

Abb. 19: Limits of the Harmonic Quantum Oscillator: In addition to the classical probability
P (x) (red), the wave functions for n = 0 and n = 15 are shown. The discrepancy
between classical and quantum physics becomes particularly evident in the case n = 0.

4.2 Algebraic Solution

Motivation: Algebraic Solution Path of the Harmonic Oscillator

In this section, we describe a very elegant solution method for the quantum mechanical
harmonic oscillator, developed by Paul Dirac. For this solution path, two so-called lad-
der operators â and ↠are defined that respectively extract or add a quantum of energy
ℏω to an oscillator.

Although we initially only use the algebraic method to describe the harmonic oscillator
within this text, its significance extends far beyond. We will subsequently also use ladder
operators to solve problems with quantum mechanical angular momentum more easily.
Moreover, ladder operators are essential in the quantization (that is, the quantum me-
chanical representation) of fields like, for example, the electromagnetic field, which is the
foundation of the „second quantization“.

4.2.1 Ladder Operators

Let’s start from the one-dimensional, reduced Schrödinger equation of a particle in (4.18) and
extract the (reduced) Hamiltonian operator H̄:

1 d2
 
H̄ = y 2 − 2 (4.41)
2 dy

95
 4 Harmonic Oscillator

Foresightedly, we can rewrite this Hamiltonian operator H̄:

1 1 2 d2 1 2 d2
    
H̄ = y − + y − =
2 2 dy 2 2 dy 2
1 1 2 d2 d d 1 2 d2 d d
    
= y − −y +y + y − 2 +y −y =
2 2 dy 2 dy dy 2 dy dy dy
1 1 2 d2 d d 1 2 d2 d d
    
= y − − y + 1 + y + y − + y − 1 − y =
2 2 dy 2 dy dy 2 dy 2 dy dy
1 1 2 d2 d d 1 2 d2 d d
       
= y − − y + 1 + y + y − + y − 1 + y (4.42)
2 2 dy 2 dy dy 2 dy 2 dy dy

To simplify the curly braces, it helps to investigate the effect of a general operator  = d
dy y on
a function ψ(y):
d  d d
  
 ψ(y) = yψ(y) = ψ(y) + y ψ(y) = 1 + y ψ(y) =⇒
dy dy dy
d d d
   
 = 1 + y =⇒ y = 1+y (4.43)
dy dy dy
Thus, we can substitute the two curly braces in (4.42) according to (4.43):
1 1 2 d2 d d 1 2 d2 d d
    
H̄ = y − 2 −y + y + y − 2 +y − y =
2 2 dy dy dy 2 dy dy dy
1 1 d 1 d 1 d 1 d
        
= √ y+ √ y− + √ y− √ y+ (4.44)
2 2 dy 2 dy 2 dy 2 dy
| {z }| {z } | {z }| {z }
â ↠↠â

We replace the operations in parentheses by the new operators â and ↠. This allows us a new,
more compact notation for the Hamiltonian H̄:
1 †  1
H̄ = ââ + ↠â = {â, ↠} (4.45)
2 2
The dimensionally based Hamiltonian Ĥ arises from H̄ via Ĥ = ℏω H̄. We have now reduced the
Hamiltonian to the ladder operators â and ↠. The operator ↠is called the creation operator
or raising operator:
s
1 1 1
r
d x d mω ℏ d
    
†
â = √ y − =√ − x0 =√ x− (4.46)
2 dy 2 x0 dx 2 ℏ mω dx
The operator â is the annihilation operator or lowering operator:
s
1 1 1
r
d x d mω ℏ d
    
â = √ y + =√ + x0 =√ x+ (4.47)
2 dy 2 x0 dx 2 ℏ mω dx
Both operators are generally termed the ladder operators; the reason for this naming will soon
become clear. With the definitions (4.46) and (4.47), we express the one-dimensional posi-
tion operator x̂ and momentum operator p̂ as a linear combination of the raising and lowering
operators:
x0  
x̂ = √ â + ↠(4.48)
2
ip0  
p̂ = − √ â − ↠(4.49)
2

96
Quantum Theory I

Example: Position and Momentum Operator as Ladder Operators

To show that the representations of x̂ and p̂ from (4.48) and (4.49) actually correspond
to the position and momentum operator, we substitute the definitions of creation and
annihilation from (4.46) and (4.47):

x0  x0 x d x d
  
x̂ = √ â + ↠= + x0 + − x0 =x □
2 2 x 0 dx x 0 dx
iℏ  iℏ x d x d d
  
†
p̂ = − √ â − â = − + x0 − + x0 = −iℏ □
2x0 2x0 x0 dx x0 dx dx

The momentum operator is traditionally represented in the position space – therefore the deriva-
tive with respect to x can be reversed by the action of hatp:
d 1 i
= − p̂ = p̂
dx iℏ ℏ
If we substitute this into the definitions (4.46) and (4.47), due to the complex term, it becomes
immediately apparent that â and ↠are not Hermitian:
1 x̂ x0
 
†
â = √ − i p̂ (4.50)
2 x0 ℏ
1 x̂ x0
 
â = √ + i p̂ (4.51)
2 x0 ℏ

It holds â ̸= ↠because they differ in the sign before the complex number in (4.50) and (4.51).
Using the provided definitions (4.13), and (4.15), we express the position operator (4.48) and
momentum operator (4.49) in terms of the dimensionless position operator ŷ and momentum
operator p̂y :

1
ŷ = √ (â + ↠) (4.52)
2
i
p̂y = − √ (â − ↠) (4.53)
2

4.2.2 Commutators of the Ladder Operators and Number State Operator

The two trivial commutators between the faller â and the raiser ↠follow:

[â, â] = [↠, ↠] = 0 (4.54)

However, the two operators â and ↠do not commute with each other. The following important
commutator relationship hence follows:

[â, ↠] = â↠− ↠â = 1 (4.55)

Example: Commutator between â and â†

To prove the relation from (4.55), we substitute the definitions (4.46) and (4.47) for â†
and â. As these are operators with specific effects, carefully consider their operation on a

97
 4 Harmonic Oscillator

preceded wave function ψ(y):

 
[â, ↠] ψ(y) = â↠− ↠â ψ =
1 d d d d
      
= y+ y− ψ− y− y+ ψ =
2 dy dy dy dy
1
" ! !#
d dψ d dψ
  
= y+ yψ − − y− yψ + =
2 dy dy dy dy
1
" ! !#
dψ d   d2 ψ dψ d   d2 ψ
= 2
y ψ−y + yψ − 2 − y ψ+y
2
− yψ − 2 =
2 dy dy dy dy dy dy
1 2
" #
dψ d   d2 ψ dψ d   d2 ψ
= y ψ−y + yψ − 2 − y 2 ψ − y + yψ + 2 =
2 dy dy dy dy dy dy
d   dψ dψ dψ
= yψ − y = ψ+y −y =1·ψ □
dy dy dy dy
Equivalently, a more concise approach avoids using position space representation but uses
the canonical commutation relation [x̂, p̂] = iℏ from (3.133). We omit the wave function
for this proof and imply the operation of the operators:
1 x̂ x0 x̂ x0
   
†
[â, â ] = + i p̂ , − i p̂ =
2 x0 ℏ x0 ℏ
1 1
!
i i x20
= [x̂, x̂] − [x̂, p̂] + [p̂, x̂] + 2 [p̂, p̂] =
2 x20 ℏ ℏ ℏ
1 i i i
 
(3.133)
=− [x̂, p̂] + [x̂, p̂] = − [x̂, p̂] =
2 ℏ ℏ ℏ
i
= − iℏ = 1 □
ℏ

With a simple transformation of (4.55), the following useful relation can be derived:

â↠= 1 + ↠â (4.56)

Using this relation, we can simplify the dimensionless Hamiltonian (4.45):

1 †  (4.56) 1   1
H̄ = ââ + ↠â = 1 + ↠â + ↠â = ↠â +
2 2 2

Typically, instead of ↠â, a new operator called the number state operator N̂ is defined:

N̂ = ↠â (4.57)

Hence, we find a new, compact notation for the dimensioned Hamiltonian Ĥ using the number
state operator N̂ :

1
 
Ĥ = ℏω N̂ + with N̂ = ↠â (4.58)
2

Properties of theNumber State Operator How does the number state operator act on an
eigenstate of the Hamiltonian operator Ĥ? From the analytical solution of the Schrödinger
equation in the oscillator potential, we know the solutions (4.36) of the eigenenergies; based on

98
Quantum Theory I

the eigenvalue equation Ĥ |n⟩ = En |n⟩, where |n⟩ are the wave functions of the n-th eigenenergy
state of the harmonic oscillator:
1 1
   
(4.36)
   
0 = Ĥ − En |n⟩ = ℏω N̂ + − n+ |n⟩ = ℏω N̂ − n |n⟩
2 2

Reorganizing this equation leads to another crucial expression, namely the effect of the number
state operator N̂ on an eigenenergy state |n⟩ of the harmonic oscillator. Specifically, by its
eigenvalue n, N̂ conveys the information on which excitation state |n⟩ resides.

N̂ |n⟩ = n |n⟩ (4.59)

An eigenvalue n of the number state operator N̂ can only take values greater than or equal to
zero:
n = ⟨n|N̂ |n⟩ = ⟨n|↠â|n⟩ = ⟨ân|ân⟩ = || |ân⟩ ||2 ≥ 0 (4.60)
For n = 0, this denotes the ground state of the system, and for n > 0, an excited state. Also,
for arbitrary functions, the expectation value formation ensures the relationship ⟨ψ|N̂ |ψ⟩ ≥ 0
(positivity). The number state operator is Hermitian and positive; thus, we get positive, real
eigenvalues in the eigen-system. Hermiticity follows directly from (N̂ )† = (↠â)† = ↠â = N̂ .

Commutators of the Number State Operator Additionally, commutator relations exist be-
tween N̂ and the ladder operators, utilisng the already known commutator relation (4.55):

(4.55)
[N̂ , â] = [↠â, â] = ↠ââ − â↠â = (↠â − â↠)â = [↠, â]â = −[â, ↠]â =
= −â (4.61)

From a straightforward transformation of (4.61), the following useful relation can be derived:

N̂ â − âN̂ = −â =⇒ N̂ â = âN̂ − â (4.62)

Exploiting the identity [AB, C] = A[B, C] + [A, C]B, the commutator [N̂ , ↠] is simply deter-
mined:
(4.55)
[N̂ , ↠] = [↠â, ↠] = ↠[â, ↠] − [↠, ↠]â = ↠[â, ↠] =
= ↠(4.63)

This yields another beneficial relation via rearrangement of (4.63):

N̂ ↠− ↠N̂ = ↠=⇒ N̂ ↠= ↠+ ↠N̂ (4.64)

4.2.3 Action of Raising and Lowering Operators

Let N̂ act on a state |↠n⟩, we can explicitly determine the action of ↠:

(4.64)
N̂ |↠n⟩ = N̂ ↠|n⟩ = (↠+ ↠N̂ ) |n⟩ = ↠(1 + N̂ ) |n⟩ = ↠(1 + n) |n⟩ =
= (n + 1)↠|n⟩ = (n + 1) |↠n⟩ (4.65)

The state |↠n⟩ is thus an eigenfunction of N̂ with the eigenvalue n + 1. Or, in other words:
the number state or the excitation level of |n⟩ is increased by one by the application of ↠; this
is why ↠is also called the raising operator! We get the same eigenvalue n + 1 when we let the
number state operator N̂ act on a state |n + 1⟩ – up to a scaling factor c, this will correspond to

99
 4 Harmonic Oscillator

|↠n⟩ ≡ ↠|n⟩ = c |n + 1⟩. The specific form of c can be determined from the modulus squared
construction of |α† n⟩:

|| |↠n⟩ ||2 = ⟨n + 1|c∗ c|n + 1⟩ = |c|2 ⟨n + 1|n + 1⟩ = |c|2

(4.56)
|| |↠n⟩ ||2 = ⟨n|â ↠|n⟩ = ⟨n|↠â + 1|n⟩ = ⟨n|N̂ + 1|n⟩ = ⟨n|n + 1|n⟩ =
= (n + 1)⟨n|n⟩ = n + 1 (4.66)
√
We find two solutions for c = ± n + 1 (where we disregard a global phase eiφ ). The negative
solution is neglected to guarantee a positive eigenvalue of N̂ according to (4.60). From (4.65)
and (4.66) it follows thus for the action of the raising operator ↠:
√
↠|n⟩ = n + 1 |n + 1⟩ (4.67)

In a similar way, we can also determine the action of the lowering operator. First, we consider
the action of N̂ on a state |ân⟩:
(4.62)
N̂ |ân⟩ = N̂ â |n⟩ = (âN̂ − â) |n⟩ = â(N̂ − 1) |n⟩ = â(n − 1) |n⟩ =
= (n − 1)â |n⟩ = (n − 1) |ân⟩ (4.68)

The state |ân⟩ is thus an eigenfunction of N̂ with the eigenvalue n − 1, just like the state |n − 1⟩.
Or, in other words: the number state or the excitation level of |n⟩ is decreased by one by the
application of â; this is why â is also called the lowering operator! The eigenstate |ân⟩ again
corresponds, up to a scaling factor c, to the state |n − 1⟩ – it follows: |ân⟩ = c |n − 1⟩. From
the construction of the modulus squared, we can determine the form of c:

|| |ân⟩ ||2 = ⟨n − 1|c∗ c|n − 1⟩ = |c|2 ⟨n − 1|n − 1⟩ = |c|2

(4.57)
|| |ân⟩ ||2 = ⟨n|↠â|n⟩ = ⟨n|N̂ |n⟩ = ⟨n|n|n⟩ = n⟨n|n⟩ =
=n (4.69)
√
Again, we find two solutions c = n, whereby due to the positivity of the number state operator
N̂ , the negative scaling factor can be excluded (and a global phase eiφ ignored). Thus, from
(4.68) and (4.69) the complete action of the lowering operator results:

√
â |n⟩ = n |n − 1⟩ (4.70)

Applying the lowering operator to the ground state |0⟩, one obtains zero:

â |0⟩ = 0 (4.71)

Intuitively, this is also logical: if a state is already in the ground state, it cannot be lowered
further. Starting from the ground state |0⟩, we can generate a state |n⟩ by applying the raising
operator ↠n times: √
(↠)n |0⟩ = n! |n⟩ (4.72)
Rearranging (4.72) for |n⟩ gives us an expression algebraically equivalent to Hermite polynomials
from (4.40):

1
|n⟩ = √ (↠)n |0⟩ (4.73)
n!

100
Quantum Theory I

n+1

â N̂
n
â†

n−1

Abb. 20: Schematic action of the raising operator ↠, the lowering operator â, and the number
state operator N̂ .

Example: Action of N̂ , ↠and â on the Ground State in Position Space

We have now learned two equivalent formalisms for describing the harmonic quantum os-
cillator: the analytical path, which led us to the Hermite polynomials (4.40) as functions of
position x, and the algebraic path, via which new operators N̂ , ↠and â were introduced.
How do number state, raising, and lowering operators act now when the oscillator eigen-
states |n⟩ are projected into the “dimensionless position space” |y⟩? We will demonstrate
this using the simplest case, the ground state |0⟩:
(4.21) 1 2
ψ 0 (y) = ⟨y|0⟩ = N e− 2 y

From (4.46) and (4.47), the dimensionless number state operator can be derived:

1 1 1
! !
d d d2 d d d2
  
N̂ = y− y+ = y2 − 2 − y +y −1 = y2 − 2 − 1
2 dy dy 2 dy dy dy 2 dy

For the action of the number state operator N̂ in position space, it holds:
!
N d2 1 2
⟨y|N̂ |0⟩ = y 2 − 2 − 1 e− 2 y =
2 dy
N d h − 1 y2 i
 
2 − 12 y 2 − 1 2
= y e + ye 2 −e 2 y
=
2 dy
N  2 − 1 y2 1 2 1 2 1 2

= y e 2 − y 2 e− 2 y + y 2 e− 2 y − y 2 e− 2 y = 0
2
We obtain the expected result: the ground state corresponds to an occupation of n = 0
(since ⟨y|N̂ |0⟩ = 0 · ⟨y|n⟩ = 0). For the action of the raising operator, it follows:

N d N  Ñ
 
1 2 1 2 1 2 1 2

⟨y|↠|0⟩ = √ y − e− 2 y = √ ye− 2 y + ye− 2 y = √ 2y e− 2 y
2 dy 2 2
The new term 2y corresponds to the first Hermite polynomial H1 (y) = 2y (without recal-
culating the normalization Ñ explicitly). For the lowering operator, it holds:

N d N 
 
1 2 1 2 1 2

⟨y|â|0⟩ = √ y + e− 2 y = √ ye− 2 y − ye− 2 y = 0
2 dy 2
The ground state can thus no longer be further reduced by the annihilation operator â.

101
 4 Harmonic Oscillator

Example: Ground State in Position Space from â |0⟩ = 0

According to (4.71), it holds: â |0⟩ = 0; actually, from this, we can calculate the position
representation of the ground state ψ 0 (y) ≡ ψ 0 . Projecting the whole equation into the
reduced position space |y⟩, the lowering operator changes according to (4.47), and we can
formulate the following differential equation:

1 d dψ 0 1
 
⟨y|â|0⟩ = √ y − ψ 0 = 0 =⇒ = −ydy =⇒ ln(ψ 0 ) = − y 2 + C
2 dy ψ0 2

In the last step, both sides were integrated independently; exponentiating both sides yields
the already known result (4.21) for the ground state:
1 2
ψ 0 = N e− 2 y

In-Depth: Algebraic and Analytical Solutions

How can the algebraic and the analytical solution be transformed into one another? Project
a general eigenstate |n⟩ into the reduced position space ⟨y| and use (4.40), (10.17), and
(4.38), we obtain:
1 1 1 2 (4.38)
⟨y|n⟩ = ψn (y) = √ e− 2 y Hn (y) =
π 1/4 2n n!
1 1 n y2 d
n
 
− 12 y 2 −y 2 (10.17)
= √ e · (−1) e e =
π 1/4 20 n! dy n
1 1
n
d
  
1 2 1 2 1 2
= √ e− 2 y · e 2 y y − e− 2 y =
(2π) 1/4
n! dy
1 1
n 
d
 
1 2 (4.46)
= √ y− e− 2 y =
n! dy (2π) 1/4

1
= √ ⟨y|(↠)n |0⟩ (4.74)
n!
We have managed to derive an expression from the wave function in position space and
the Rodrigues formula of Hermite polynomials, which matches the representation of the
raising operator in position space (4.46). However, we must note that the normalization
changes by a factor of 2−1/4 in the transition from the Rodrigues formula (4.38) to (10.17),
since the Gaussian function now decreases at half the rate.

4.2.4 Heisenberg Uncertainty Relation

To construct the Heisenberg uncertainty relation concerning the eigenstates of the harmonic
oscillator, we need to calculate the position variance ⟨(∆x̂)2 )⟩ and momentum variance ⟨(∆p̂)2 )⟩,
as described in (3.140). For reminder: for an arbitrary operator Â, it is generally true with the
fluctuation operator ∆Â:
2
⟨(∆Â)2 ⟩ = ⟨Â2 ⟩ − ⟨Â⟩

To calculate the expectation value of the position ⟨x̂⟩ for a particular excitation state |n⟩, we
use the representation of the position operator with ladder operators from (4.48), and the action

102
Quantum Theory I

of the ladder operators on an eigenstate |n⟩ according to (4.67) and (4.70):

(4.48)
⟨x̂⟩ = ⟨n|x̂|n⟩ =
x0
= √ ⟨n|â + ↠|n⟩ =
2
x0   (4.67,4.70)
= √ ⟨n|â|n⟩ + ⟨n|↠|n⟩ =
2
x0 √ √ 
=√ n⟨n|n − 1⟩ + n + 1⟨n|n + 1⟩ = 0 (4.75)
2

The eigenstates |n + 1⟩ and |n − 1⟩ are both orthogonal to |n⟩, as the energy eigenfunctions
form a complete orthonormal system {|n⟩}. We thus expect an average localization at x = 0 in
the middle of the potential. At this point, we also recognize the power of the ladder operators:
instead of working with the complicated representation in position space, we find here only ab-
stract eigenstates, whose behavior we can exactly describe under certain operations (↠,â and
N̂ )! Complicated computational operations can thus be elegantly and compactly expressed in
algebraic notation.

For the expectation value of the momentum ⟨p̂⟩, we can analogously use the representation of
the momentum operator with raising operators (4.49):

(4.49)
⟨p̂⟩ = ⟨n|p̂|n⟩ =
i ℏ
= −√ ⟨n|â − ↠|n⟩ =
2 x0
i ℏ   (4.67,4.70)
= −√ ⟨n|â|n⟩ − ⟨n|↠|n⟩ =
2 x0
i ℏ √ √ 
= −√ n⟨n|n − 1⟩ − n + 1⟨n|n + 1⟩ = 0 (4.76)
2 x0

The expectation value of the momentum thus also vanishes! For the second moment of x̂, we
find an expression different from zero:

(4.48)
⟨x̂2 ⟩ = ⟨n|x̂2 |n⟩ =
x20
= ⟨n|(â + ↠)(â + ↠)|n⟩ =
2
x2 (4.56)
= 0 ⟨n|ââ + ↠↠+ â↠+ ↠â|n⟩ =
2
x2
= 0 ⟨n|ââ + ↠↠+ 1 + ↠â + ↠â|n⟩ =
2
x2   (4.57)
= 0 ⟨n|ââ|n⟩ + ⟨n|↠↠|n⟩ + 2 ⟨n|↠â|n⟩ + ⟨n|n⟩ =
2
x2   (4.67,4.70)
= 0 ⟨n|ââ|n⟩ + ⟨n|↠↠|n⟩ + 2 ⟨n|N̂ |n⟩ + 1 =
2
x 2 √ √ √ √ 
= 0 n n − 1⟨n|n − 2⟩ + n + 1 n + 2⟨n|n + 2⟩ + 2n⟨n|n⟩ + 1 =
2
1 (4.13) ℏ 1
  
= x0 n +
2
= n+ (4.77)
2 mω 2

While the expectation value seems to be centered around the midpoint of the oscillator potential
⟨x̂⟩ = 0, the variance of the position does not vanish; there are therefore fluctuations around
the equilibrium position. The second moment of the momentum operator can be calculated

103
 4 Harmonic Oscillator

analogously:
(4.49)
⟨p̂2 ⟩ = ⟨n|p̂2 |n⟩ =
p20
=− ⟨n|(â − ↠)(â − ↠)|n⟩ =
2
p2
= − 0 ⟨n|ââ + ↠↠− â↠− ↠â|n⟩ =
2
p20 (4.56)
= ⟨n| − ââ − ↠↠+ â↠+ ↠â|n⟩ =
2
p20
= ⟨n| − ââ − ↠↠+ 1 + ↠â + ↠â|n⟩ =
2
p2   (4.57)
= 0 − ⟨n|ââ|n⟩ − ⟨n|↠↠|n⟩ + 2 ⟨n|↠â|n⟩ + ⟨n|n⟩ =
2
p20   (4.67,4.70)
= − ⟨n|ââ|n⟩ − ⟨n|↠↠|n⟩ + 2 ⟨n|N̂ |n⟩ + 1 =
2
p2  √ √ √ √ 
= 0 − n n − 1⟨n|n − 2⟩ − n + 1 n + 2⟨n|n + 2⟩ + 2n⟨n|n⟩ + 1
2
1 (4.15) 1
  
= p20 n + = mℏω n + (4.78)
2 2
Thus, for the variance of position and momentum, we obtain:
1
 
2 (4.77,4.75)
ℏ
⟨(∆x̂) ⟩ = ⟨x̂ ⟩ − ⟨x̂⟩
2 2
= n+
mω 2
1
 
(4.78,4.76)
⟨(∆p̂)2 ⟩ = ⟨p̂2 ⟩ − ⟨p̂⟩2 = mℏω n +
2
The square root of each variance represents the standard deviation; forming the product ∆x
and ∆p we finally obtain the following expression for the Heisenberg uncertainty relation:
1
q q  
∆x∆p = ⟨(∆x̂)2 ⟩ ⟨(∆p̂)2 ⟩ =ℏ n+ (4.79)
2
If the harmonic quantum oscillator is in the ground state |0⟩ (with n = 0), it possesses the
minimum uncertainty according to (4.79):

ℏ
∆x∆p = (4.80)
2

4.3 Time-Evolved Oscillator States

4.3.1 Time-Dependent Oscillator Wave Functions
Let’s now investigate the time-dependent, one-dimensional Schrödinger equation for a particle
in the harmonic oscillator potential V (x):

1
!
ℏ2 ∂ 2 ∂
Ĥ(x)ψn (x, t) = − + mω 2 x2 ψn (x, t) = iℏ ψn (x, t) (4.81)
2m ∂x 2 2 ∂t

The Hamiltonian operator Ĥ is not explicitly time-dependent. Therefore, we can apply the
separation Ansatz ψn (x, t) = ϕn (x)φn (t) as described in Chapter ??. For the time-dependent
part, we have the following differential equation according to (2.35) and (4.36):
d 1
 
(4.36)
iℏ φn (t) = En φn (t) = ℏω n + φn (t) (4.82)
dt 2

104
Quantum Theory I

This differential equation can, as shown in (2.36), be easily solved, and we obtain the time
evolution of the eigenstates of the harmonic oscillator:

ψn (x, t) = ϕn (x) e−iEn t/ℏ = ϕn (x) e−iω(n+ 2 )t = ϕn (x) e−iωnt e− 2 iωt

(2.36) (4.36) 1 1
(4.83)

The time-dependent part leads, with increasing excitation of the system, to an ever faster os-
cillation of the phase of the stationary eigenstate. A general state in space (at a fixed time t0 )
ψ(x, t0 ) can be constructed as a linear combination of the time-independent eigenstates of the
quantum oscillator ϕn (x). The further time evolution of ψ(x, t) for times t > t0 then simply
follows from each of these base states ϕn (x) evolving with its associated time development:
∞ ∞ ∞
1 1
ψ(x, t) = an ψn (x, t) = an ϕn (x)e−iωnt e− 2 iωt = e− 2 iωt an ϕn (x)e−iωnt (4.84)
X X X

n=0 n=0 n=0

The expansion coefficients an can, as usual, be calculated according to the following relation:
Z +∞
an = dx′ ϕ∗n (x′ )ψ(x′ , t) (4.85)
−∞

We will use the above-found relationships to construct states that can behave like classical
particles.

4.3.2 Coherent Glauber States

Motivation: Classical Oscillator Motion with Quantum Physics

In this chapter, we ask the question, which time-dependent wave functions in the harmonic
oscillator potential come closest to the harmonic motion of a particle known from classical
physics. Erwin Schrödinger already succeeded in constructing such states in 1926,
which are today called “coherent states” or “Glauber states” (after Roy Glauber). As
we will see, these states correspond to Gaussian wave packets that run back and forth in
the harmonic potential and maintain the minimal position and momentum uncertainty at
all times.

The desired coherent or Glauber state is referred to as |α⟩ and corresponds to a superposition
of time-dependent eigenstates of the harmonic oscillator. The wave packets resulting from this
superposition should have the following properties:

they follow the classical equations of motion, so that ⟨x̂⟩ = C cos (ωt − δ);

they have minimal uncertainty for all times t > 0;

the uncertainty is equally distributed over y and py .

As a starting point, we consider how x(t) and p(t) of a particle behave in a classical harmonic
oscillator (assuming, without loss of generality, a phase φ = 0 for x(t) from (4.4)):

x(t) = A cos (ωt) = A Re(eiωt )

π
p(t) = −B sin (ωt) = B cos ( + ωt) = B Re(e| iπ/2 (4.86)
{z } e ) = B Re(ie )
iωt iωt
2
i

In the second line, we use the identity − sin (x) = cos (x + π/2). After this transformation, we
recognize that the momentum p(t) is shifted by the phase eiπ/2 = i compared to the position
x(t). Let us now formulate the condition that the sought quantum states |α⟩ must satisfy for the

105
 4 Harmonic Oscillator

quantum mechanical uncertainty to be equally distributed over the (dimensionless) quantities y

and py :

(ŷ − ⟨ŷ⟩) |α⟩ = −i (p̂y − ⟨p̂y ⟩) |α⟩ (4.87)

The factor −i = e−iπ/2 on the right-hand side of equation (4.87) serves to compensate for the
phase shift of eiπ/2 = i between py and y. As py and y are dimensionless, no further pre-factors
are necessary. By transforming (4.87), we arrive at an eigenvalue equation that the desired state
|α⟩ must satisfy:

(ŷ + ip̂y ) |α⟩ = (⟨ŷ⟩ + i ⟨p̂y ⟩) |α⟩ (4.88)

The right-hand side of (4.88) corresponds to a complex eigenvalue. Substituting the definitions
(4.52) and (4.53) into (4.88), we obtain:
1  1 
   
(⟨ŷ⟩ + i ⟨p̂y ⟩) |α⟩ = √ â + ↠+ √ â − ↠|α⟩ =
2 2
1  
= √ â + ↠+ â − ↠|α⟩ =
2
√
= 2â |α⟩

The lowering operator can thus be expressed in terms of the corresponding expectation values
of the position and momentum operators in the case of the Glauber state, here abbreviated with
α:
1
â |α⟩ = √ (⟨ŷ⟩ + i ⟨p̂y ⟩) |α⟩ = α |α⟩
2
We are looking for states |α⟩ that are eigenstates of the lowering operator â and thus satisfy the
following eigenvalue equation:

â |α⟩ = α |α⟩ (4.89)

The complex-conjugate version of the eigenvalue equation also holds analogously:

⟨α| ↠= ⟨α| α∗ (4.90)

The solution for the above eigenvalue problem (4.88) at a fixed time t = 0 is:
∞ ∞
αn (4.73) (α↠)n
|α⟩t=0 = N |n⟩ = N (4.91)
X X
√ |0⟩
n=0 n! n=0
n!

For the latter expression, we use that |n⟩ can always be expressed by the ground state through
the raising operator. For the normalization factor N , we calculate:

1 ! (4.91)
2
= ⟨α|α⟩t=0 =
N
∞ X ∞
αn (α∗ )m
= ⟨m|n⟩ =
X
√
n=0 m=0 m!n!
∞ X
∞
αn (α∗ )m
= δnm =
X
√
n=0 m=0 m!n!
∞
(α α)n
∗
= =
X

n=0
n!
∞
|α|2n 2
= = e|α|
X

n=0
n!

106
Quantum Theory I

Rearranging and taking the square root yields the finished normalization, depending on the
magnitude of the eigenvalue:
1 2
N = e− 2 |α| (4.92)

Thus, the normalized Glauber state |α⟩t=0 at time t = 0 can finally be represented as fol-
lows:

∞
1 2 (α↠)n
|α⟩t=0 = e− 2 |α| (4.93)
X
|0⟩
n=0
n!

Letting â act on |α⟩t=0 , we see that it is indeed an eigenstate of the lowering operator:

∞
αn
 
1 2
â |α⟩t=0 = â e− 2 |α| =
X
√ |n⟩
n=0 n!
∞
1 2 αn (4.70)
= e− 2 |α| √ â |n⟩ =
X

n=0 n!
∞
1 2 αn √
= e− 2 |α| n |n − 1⟩ =
X
√
n=0 n!
∞
1 2 ααn−1
= e− 2 |α| |n − 1⟩ =
X
(n − 1)!
p
n=1
∞
− 12 |α|2 αm−1 m=n+1
= αe |m − 1⟩ =
X
(m − 1)!
p
m=1
∞ n
1 2 X α
= αe− 2 |α| √ |n⟩ =
n=0 n!
= α |α⟩t=0 (4.94)

All eigenstates of the lowering operator are normalized by construction, so it always holds:

⟨α|α⟩ = 1 (4.95)

However, two arbitrary, different eigenstates |α⟩ and |β⟩ of the lowering operator are not or-
thogonal to each other:
⟨α|β⟩ =
̸ δαβ (4.96)

This was expected because â is not a Hermitian operator. Let’s verify now whether the require-
ment of minimal uncertainty in position and momentum is met. For this, we need to calculate
the first and second moments of x and p. Let’s start with ⟨x̂⟩:

(4.48) x0
⟨x̂⟩ = ⟨α|x̂|α⟩ = √ ⟨α|â + ↠|α⟩ =
2
x0  
= √ ⟨α|â|α⟩ + ⟨α|↠|α⟩ =
2
x0   x0
= √ α⟨α|α⟩ + α∗ ⟨α|α⟩ = √ (α + α∗ ) =
2 2
x0 √
= √ 2 Re(α) = 2x0 Re(α) (4.97)
2

107
 4 Harmonic Oscillator

For the expectation value ⟨p̂⟩, a completely equivalent calculation can be performed:

(4.49) i ℏ
⟨p̂⟩ = ⟨α|p̂|α⟩ = − √ ⟨α|â − ↠|α⟩ =
2 x0
i ℏ 
= −√ ⟨α|â|α⟩ − ⟨α|↠|α⟩ =
2 x0
i ℏ  i ℏ
= −√ α⟨α|α⟩ − α∗ ⟨α|α⟩ = − √ (α − α∗ ) =
2 x0 2 x0
1 ℏ √
=√ 2 Im(α) = 2p0 Im(α) (4.98)
2 x0

To calculate the variance of position and momentum, we also need the second moments ⟨x̂2 ⟩
and ⟨p̂2 ⟩. Let’s consider the second moment of position:

(4.48)
⟨x̂2 ⟩ = ⟨α|x̂2 |α⟩ =
x20
= ⟨α|(â + ↠)(â + ↠)|α⟩ =
2
x20 (4.56)
= ⟨α|ââ + ↠↠+ â↠+ ↠â|α⟩ =
2
x20
= ⟨α|ââ + ↠↠+ 1 + ↠â + ↠â|α⟩ =
2
x20  
= ⟨α|ââ|α⟩ + ⟨α|↠↠|α⟩ + 2 ⟨α|↠â|α⟩ + ⟨α|α⟩ =
2
x20  2 
= α + (α∗ )2 + 2αα∗ + 1 =
2
x20  
= (α + α∗ )2 + 1 (4.99)
2

For the second moment of momentum, the same calculation applies:

(4.49)
⟨p̂2 ⟩ = ⟨α|p̂2 |α⟩ =
1 ℏ2
=− ⟨α|(â − ↠)(â − ↠)|α⟩ =
2 x20
1 ℏ2 (4.56)
= − 2 ⟨α|ââ + ↠↠− â↠− ↠â|α⟩ =
2 x0
1 ℏ2
= − 2 ⟨α|ââ + ↠↠− 1 − ↠â − ↠â|α⟩ =
2 x0
1 ℏ2  
= − 2 ⟨α|ââ|α⟩ + ⟨α|↠↠|α⟩ − 2 ⟨α|↠â|α⟩ − ⟨α|α⟩ =
2 x0
1 ℏ2  
= − 2 α2 + (α∗ )2 − 2αα∗ − 1 =
2 x0
1 ℏ2  
= − 2 (α − α∗ )2 − 1 (4.100)
2 x0

108
Quantum Theory I

Thus, we can finally calculate the variance for position and momentum:

(4.99,4.97)
⟨(∆x̂)2 ⟩ = ⟨x̂2 ⟩ − ⟨x̂⟩2 =
x20  
= (α + α∗ )2 + 1 − (α + α∗ )2 =
2
x20 (4.14) 1 ℏ
= = (4.101)
2 2 mω
(4.100,4.98)
⟨(∆p̂)2 ⟩ = ⟨p̂2 ⟩ − ⟨p̂⟩2 =
1 ℏ2 
∗ 2 ∗ 2

= 1 − (α − α ) + (α − α ) =
2 x20
1 ℏ2 (4.14) 1
= = ℏmω (4.102)
2 x20 2

Finally, we obtain the following expression for the Heisenberg uncertainty relation:
s s
q q ℏ ℏmω ℏ
∆x∆p = ⟨(∆x̂)2 ⟩ ⟨(∆p̂)2 ⟩ = = (4.103)
2mω 2 2

We recognize that this corresponds to the same uncertainty as in the ground state of the harmonic
oscillator (4.80). It corresponds – as required – to the minimal uncertainty.

Time Evolution of Glauber States Until now, we have only considered coherent states at a
fixed time. If we drop this assumption, we need to take into account the time evolution of the
eigenstates |n⟩ derived in (4.84). For this, we write again the expression for the time-independent
state from (4.93) and add the time evolution from (4.84), renaming the eigenvalue α to α0 to
express that this is the eigenvalue at time t = 0:

∞
1 2 α0n 1
|α(t)⟩ = e− 2 |α0 | e−iωnt e− 2 iωt |n⟩ =
X
√
n=0 n!
∞
− 21 iωt − 21 |α0 |2 αn  n
=e √ 0 e−iωt |n⟩ =
X
e
n=0 n!
∞
1 1 2 1  n
= e− 2 iωt e− 2 |α0 | α0 e−iωt |n⟩ = | α(t) = α0 e−iωt
X
√
n=0 n!
∞
− 21 iωt − 21 |α0 |2 α(t)n
=e |n⟩ =
X
e √
n=0 n!
∞
1 1 2 α(t)n
= e− 2 iωt e− 2 |α(t)| √ |n⟩ =
X

n=0 n!
1
= e− 2 iωt |α′ ⟩α′ =α(t) (4.104)

Since the phase term does not play a role in the modulus, it follows immediately |α0 | = |α(t)|.
We have just shown that the time-dependent Glauber state |α(t)⟩ can be expressed for every
1
time t over the expression for the time-independent state |α′ ⟩ if it is multiplied by e− 2 iωt and
the factor α′ is set to α0 e−iωt . This means that all previous proofs remain valid since we have
1
always calculated with a general (arbitrary) value for α, and the phase pre-factor e− 2 iωt does
not play a role. Therefore, for instance, the minimal uncertainty (4.103) remains preserved for
all times t. The wave packet does not “disperse”.

109
 4 Harmonic Oscillator

Analogy to the Classical Oscillator We can now calculate how the expectation value of the
position ⟨x̂(t)⟩ changes over time. For this, we use expression (4.97), replacing α with α(t):
x0 (4.104) x0
⟨x̂⟩ = √ 2 Re(α(t)) = √ 2 Re(α0 e−iωt ) (4.105)
2 2
The complex number α0 can also be represented by its magnitude |α0 | and its phase φ, so
α0 = |α0 |eiφ . Substituting this into (4.105), we get:
x0   √
⟨x̂⟩ = √ 2 Re |α0 | eiφ e−iωt = 2 x0 |α0 | cos (ωt − φ) (4.106)
2
The coherent wave packet thus swings in the harmonic oscillator potential like a classical particle;
just as we derived it for the classical harmonic oscillator in (4.4).

110
Quantum Theory I

5 Angular Momentum

Motivation: The Orbital Angular Momentum (and Other Angular Momenta)

In this chapter, we will discuss the important quantum mechanical observable of angular
momentum L. Using the correspondence principle, we will first derive a quantum me-
chanical angular momentum operator L̂ from the classical definition of L. The observable
thus defined is called the orbital angular momentum because it is the quantum mechanical
equivalent of the classical angular momentum exhibited, for example, by electrons orbiting
the atomic nucleus.

Since angular momentum is a vector quantity, the angular momentum operator L̂ is conse-
quently a vector operator. Subsequently, we will determine a set of commutation relations
that the angular momentum operator L̂ satisfies. It turns out that these relations are so
general that we can call any operator that satisfies them an angular momentum operator.

Indeed, in quantum mechanics, there are other angular momentum operators besides the
orbital angular momentum operator L̂, such as the spin operator Ŝ, which we will examine
in more detail later.

5.1 Angular Momentum Operator

The classical angular momentum is defined as L = r × p; from the correspondence principle, it
follows for the angular momentum operator that L̂ = r̂ × p̂. In position representation, for the
position and momentum operators it holds that r̂ → r and p̂ → −iℏ∇. Consequently, we write
in position space {|r⟩}:
   
x ∂x
{|r⟩}
L̂ = r̂ × p̂ −−−→ −iℏ r × ∇ = −iℏ y  × ∂y  (5.1)
   
z ∂z

By explicitly carrying out the cross product, we can find a position representation for each
individual component of the angular momentum:
     
L̂x ŷ p̂z − ẑ p̂y y∂z − z∂y
 {|r⟩}
L̂ = L̂y  = ẑ p̂x − x̂p̂z  −−−→ −iℏ z∂x − x∂z  (5.2)
    

L̂z x̂p̂y − ŷ p̂x x∂y − y∂x

In the course of this chapter, no strict distinction is made between the abstract operator notation
and the expression in the position basis; instead of “→”, “=” is written directly.

z
L

x
r
p

Abb. 21: Representation of the classical angular momentum.

111
 5 Angular Momentum

L̂ is also referred to as a vector operator. In index notation, the cross product can be expressed
using the Levi-Civita tensor εijk :

L̂i = εijk r̂j p̂k = −iℏεijk rj ∂k (5.3)

In addition to the individual components, a magnitude operator of the angular momentum L̂2
can also be defined:
L̂2 = L̂2x + L̂2y + L̂2z (5.4)

In-Depth: Index Notation

General: In index notation, instead of writing a vector r = (rx , ry , rz )T , one simply writes
ri , where the index i stands for the i-th element of a vector r, so: r1 = rx , r2 = ry , r3 = rz .
Of course, any other letter can be used as a placeholder instead of i.

(r)i =r
ˆ i

(Einstein’s) Summation Convention: If the same index appears exactly twice in a

product term (“is saturated”), it is summed over, for example:
3
ai ri = ai ri = a1 r1 + a2 r2 + a3 r3
X

i=1

The above example corresponds to the inner product of vectors a and r: ai ri = a · r. Two
variables with the same index in a product term represent the inner product, and thus a
scalar.

Free and Saturated Indices: If a product term consists only of “double” (“saturated”)
indices, then this term represents a scalar expression. For example: ai bi cj dj . However, in
all product terms that are additively linked in an equation, only saturated indices must
appear (adding a vector to a scalar is not possible). Conversely: If a product term contains
a “free” index (i.e., an index that appears only once), then this product term represents a
vector. The left and right sides of the equation, along with all the terms involved, must
have the same free index, for example: aj = bi ci dj + bi ci aj .

Kronecker Delta: The Kronecker delta δij is defined as follows:

1, i = j
(
δij = (5.5)
0, i ̸= j

Levi-Civita Symbol: The Levi-Civita symbol is defined as follows:


 1, if ijk is an even permutation of 123


εijk = −1, if ijk is an odd permutation of 123 (5.6)

 0, otherwise (if at least two indices are the same)


Derivative: The derivative operator in index notation is defined as follows:

∂ ∂ ∂ ∂
∂i = (∇)i = ⇐⇒ ∂1 = , ∂2 = , ∂3 = (5.7)
∂ri ∂x ∂y ∂z

112
Quantum Theory I

Calculation Rules: Here’s a list of calculation rules, where the occasionally used letter
n stands for the number of dimensions (in our case n = 3). We mainly consider important
rules for δij , εijk , and ∂i :

δij = δji
δii = n
δij δjk = δik
δij xj = xi
εijk εklm = δil δjm − δim δjl
εijk εijl = 2δkl
εijk εijk = 3! = 6
εijk δij = εijk δik = εijk δjk = 0 (5.8)
εijk = εjki = εkij (5.9)
εijk = −εikj = −εjik
εiik = εiki = εikk = εiii = 0
∂i xj = δij
∂i xi = δii = n

Connection between Vector and Index Notation:

Operation Vector Notation Index Notation
Dot Product x·y xi yi
Magnitude Squared ||x||2 = x · x xi xi
Gradient (scalar) ∇f (r) ∂i f (r)
Divergence ∇·x ∂i xi
Cross Product x×y εijk xj yk
Curl ∇×x εijk ∂j xk
Laplacian (scalar) ∇2 f (r) = ∇ · (∇f (r)) ∂i ∂i f (r)
Vector Gradient ∇ ⊗x  ∂j xi
∇2 x1
Laplacian (vectorial) ∇ x =  ∇ x2 
2  2 
∂j ∂j xi
∇2 x3

We now turn to the question of whether the angular momentum operator L̂ satisfies the Hermitic-
ity condition. To do this, we need to prove that each component L̂x , L̂y , and L̂z is individually
Hermitian. As a starting point for the proof, we take equation (5.3) and consider in the last
step that p̂k and r̂j are Hermitian.

L̂†i = (εijk r̂j p̂k )† = εijk (r̂j p̂k )† = εijk p̂†k r̂j† = εijk p̂k r̂j (5.10)

To complete the proof of the Hermiticity of L̂i , we would need to reverse the order of p̂k and r̂j .
However, position and momentum operators do not commute per se, but satisfy the canonical
commutation relation [r̂i , p̂j ] = iℏδij according to (3.133). This implies: r̂j p̂k − p̂k r̂j = iℏδjk ,
and thus p̂k r̂j = r̂j p̂k − iℏδjk . Substituting this into (5.10), we obtain:
(5.3)
L̂†i = εijk (r̂j p̂k − iℏδjk ) = εijk r̂j p̂k − iℏεijk δjk = L̂i − iℏεijk δjk = Li □ (5.11)

Considering the calculation rule (5.8), the last term vanishes, and we have successfully proven
that all components of the operator, and thus L̂ as a whole, are Hermitian:

L̂† = L̂ and L̂†i = L̂i (5.12)

113
 5 Angular Momentum

For a Hermitian operator Â, it holds that any powers of it are also Hermitian: (Â2 )† = † † =
ÂÂ = Â2 ; it follows immediately that the magnitude operator of the angular momentum is also
Hermitian:
(L̂2 )† = L̂2 and (L̂2i )† = L̂2i (5.13)

5.1.1 Commutator Relations

The angular momentum operator satisfies a special commutator relation with both the position
and the momentum operators. Let’s first consider the position operator r̂:
[L̂i , r̂j ] = iℏεijk r̂k (5.14)

Example: Commutator between L̂x and r̂

We first use the relation (5.14). This allows the commutator relationships between L̂x and
the components of the position operator x̂, ŷ, and ẑ to be determined very easily (the same
calculation steps naturally also apply to L̂y and L̂z ):

[L̂x , x̂] ≡ [L̂1 , r̂1 ] = iℏε11k r̂k = iℏε111 r̂1 + iℏε112 r̂2 + iℏε113 r̂3 = 0 (5.15)
[L̂x , ŷ] ≡ [L̂1 , r̂2 ] = iℏε12k r̂k = iℏε121 r̂1 + iℏε122 r̂2 + iℏε123 r̂3 = iℏr̂3 ≡ iℏẑ (5.16)
[L̂x , ẑ] ≡ [L̂1 , r̂3 ] = iℏε13k r̂k = iℏε131 r̂1 + iℏε132 r̂2 + iℏε133 r̂3 = −iℏr̂2 ≡ −iℏŷ (5.17)

To demonstrate that these relationships are indeed correct, we can also use the explicit
representation (5.3) for L̂i . Note that the operators are only defined by their effect on a
following state ψ(x) ≡ ψ, which must be implicitly included in the calculation. We explic-
itly carry out the calculation for [L̂x , x̂] and [L̂x , ŷ] (where [L̂x , ẑ] again follows through
equivalent calculation steps):

[L̂x , x̂]ψ ≡ [L̂1 , r̂1 ]ψ = L̂1 r̂1 ψ − r̂1 L̂1 ψ = −iℏ [ε1jk rj ∂k (r1 ψ) − r1 ε1jk rj ∂k ψ] =
= −iℏ [ε1jk rj ψ∂k r1 + ε1jk rj r1 ∂k ψ − r1 ε1jk rj ∂k ψ] = | ε1jk = ε123 + ε132
= −iℏ [ε123 r2 ψ∂3 r1 + ε132 r3 ψ∂2 r1 ] = −iℏ [r2 ψδ31 − r3 ψδ21 ] = 0

As expected, we obtain the same result (5.15) in the position representation! For [L̂x , ŷ],
we proceed in an entirely analogous manner:

[L̂x , ŷ]ψ ≡ [L̂1 , r̂2 ]ψ = L̂1 r̂2 ψ − r̂2 L̂1 ψ = −iℏ [ε1jk rj ∂k (r2 ψ) − r2 ε1jk rj ∂k ψ] =
= −iℏ [ε1jk rj ψ∂k r2 + ε1jk rj r2 ∂k ψ − r2 ε1jk rj ∂k ψ] = | ε1jk = ε123 + ε132
= −iℏ [ε123 r2 ψ∂3 r2 + ε132 r3 ψ∂2 r2 ] = −iℏ [r2 ψδ32 − r3 ψδ22 ] =
= iℏr3 ψ = iℏr̂3 ψ ≡ iℏẑψ

The commutator between the angular momentum operator and the momentum operator looks
quite identical:
[L̂i , p̂j ] = iℏεijk p̂k (5.18)

Example: Commutator between L̂z and p̂

We first use the relation (5.18). This allows the commutator relationships between L̂z
and the components of the momentum operator p̂x , p̂y , and p̂z to again be determined
very easily. Since we continue to remain in the position representation, at the end we

114
Quantum Theory I

substitute the respective momentum operator components p̂i with their explicit position
representation p̂i = −iℏ∂i . Analogous to the position operator components, it follows:

[L̂z , p̂x ] ≡ [L̂3 , p̂1 ] = iℏε31k p̂k = iℏε312 p̂2 = iℏp̂2 = iℏ (−iℏ∂2 ) = ℏ2 ∂2 (5.19)
[L̂z , p̂y ] ≡ [L̂3 , p̂2 ] = iℏε32k p̂k = iℏε321 p̂1 = −iℏp̂1 = −iℏ (−iℏ∂1 ) = −ℏ2 ∂1 (5.20)
[L̂z , p̂z ] ≡ [L̂3 , p̂3 ] = iℏε33k p̂k = 0 (5.21)

To demonstrate that these relationships hold, we can again use the explicit representation
in the position basis (5.3) for L̂i and −iℏ∂i for p̂i . The state ψ, on which the operators act,
must be included in the calculation. We demonstrate the calculation explicitly for [L̂z , p̂x ]
and [L̂z , p̂y ] (again the proof for [L̂z , p̂y ] works analogously). It holds:

[L̂z , p̂x ]ψ ≡ [L̂3 , p̂1 ]ψ = L̂3 p̂1 ψ − p̂1 L̂3 ψ =

= (−iℏε3jk rj ∂k ) (−iℏ∂1 ) ψ − (−iℏ∂1 ) (−iℏε3jk rj ∂k ) ψ =
= −ℏ2 ε3jk rj ∂k ∂1 ψ + ℏ2 ε3jk ∂1 (rj ∂k ψ) =
= − ℏ2 ε3jk rj ∂k ∂1 ψ + ℏ2 ε3jk (∂1 rj ) (∂k ψ) + ℏ2 ε3jk rj ∂k ∂1 ψ = | ε3jk = ε312 + ε321
= ℏ2 ε312 (∂1 r1 ) (∂2 ψ) + ℏ2 ε321 (∂1 r2 ) (∂1 ψ) =
∂ψ
= ℏ2 ε312 δ11 (∂2 ψ) + ℏ2 ε321 δ12 (∂1 ψ) = ℏ2 ∂2 ψ ≡ ℏ2
∂y

All calculation steps can be analogously transferred to the commutator [L̂z , p̂y ]:

[L̂z , p̂y ]ψ ≡ [L̂3 , p̂2 ]ψ = L̂3 p̂2 ψ − p̂2 L̂3 ψ =

= (−iℏε3jk rj ∂k ) (−iℏ∂2 ) ψ − (−iℏ∂2 ) (−iℏε3jk rj ∂k ) ψ =
= −ℏ2 ε3jk rj ∂k ∂2 ψ + ℏ2 ε3jk ∂2 (rj ∂k ψ) =
= − ℏ2 ε3jk rj ∂k ∂2 ψ + ℏ2 ε3jk (∂2 rj ) (∂k ψ) + ℏ2 ε3jk rj ∂k ∂2 ψ = | ε3jk = ε312 + ε321
= ℏ2 ε312 (∂2 r1 ) (∂2 ψ) + ℏ2 ε321 (∂2 r2 ) (∂1 ψ) =
∂ψ
= ℏ2 ε312 δ21 (∂2 ψ) + ℏ2 ε321 δ22 (∂1 ψ) = −ℏ2 ∂1 ψ ≡ −ℏ2
∂x

In general, the following holds: If the components of an operator V̂ satisfy the following com-
mutator relationship, then V̂ is a vector operator:

[L̂i , V̂j ] = iℏεijk V̂k (5.22)

This relation expresses that the angular momentum operator L̂ is the generator of rotations. A
vector operator is characterized by transforming like a vector under rotations. Vector operators
are special cases of tensor operators whose matrix elements can be described by the Wigner-
Eckart theorem. The commutator of L̂ with a scalar operator (5.30) vanishes (since scalar
quantities are invariant under rotations). This topic and all the above claims will be discussed
in more detail in the lecture Quantum Theory II.

Since the angular momentum operator L̂ is itself a vector operator, the following transformation
behavior must be valid according to (5.22):

[L̂i , L̂j ] = iℏεijk L̂k (5.23)

The relationship (5.23) leads to an important conclusion: Two different components of the
angular momentum operator L̂ never commute and therefore exclude simultaneous measurement.
Later, we will recognize that this is the cause of angular momentum uncertainty!

115
 5 Angular Momentum

Example: Commutator between L̂x and L̂y

We first use the relationship (5.23), to calculate the commutator [L̂x , L̂y ]:

[L̂x , L̂y ] ≡ [L̂1 , L̂2 ] = iℏε12k L̂k = iℏε123 L̂3 = iℏL̂3 ≡ iℏL̂z

To show that this relation is valid, we now perform the calculation explicitly with the
representation (5.3):

[L̂x , L̂y ]ψ ≡ [L̂1 , L̂2 ]ψ = L̂1 L̂2 ψ − L̂2 L̂1 ψ =

= (−iℏε1jk rj ∂k ) (−iℏε2lm rl ∂m ) ψ − (−iℏε2lm rl ∂m ) (−iℏε1jk rj ∂k ) ψ =
= ℏ2 ε1jk ε2lm [−rj ∂k (rl ∂m ψ) + rl ∂m (rj ∂k ψ)] =
= ℏ2 ε1jk ε2lm [−rj (∂k rl ) (∂m ψ) − rj rl ∂m ∂k ψ + rl (∂m rj ) (∂k ψ) + rl rj ∂m ∂k ψ] =
= −ℏ2 ε1jk ε2lm δkl rj ∂m ψ + ℏ2 ε1jk ε2lm δmj rl ∂k ψ =
= −ℏ2 ε1jk εkm2 rj ∂m ψ + ℏ2 εk1j εj2l rl ∂k ψ =
= −ℏ2 (δ1m δj2 − δ12 δjm )rj ∂m ψ + ℏ2 (δk2 δ1l − δkl δ12 )rl ∂k ψ =
= −ℏ2 δj2 rj δ1m ∂m ψ − ℏ2 δ1l rl δk2 ∂k ψ =
= ℏ2 (−r2 ∂1 + r1 ∂2 ) ψ = | ε321 = −1; ε312 = +1
= ℏ2 (ε321 r2 ∂1 + ε312 r1 ∂2 ) ψ =
= ℏ2 (ε3jk rj ∂k ) ψ =
= iℏ (−iℏε3jk rj ∂k ) ψ = iℏL̂z ψ

5.1.2 Polar and Axial Vector Operators

In this section, we consider how the angular momentum, by construction, behaves differently
under symmetry transformations than position vectors r or momentum vectors p. To do so, we
first introduce the parity operator Π̂, which will be discussed more thoroughly in the Quantum
Theory II lecture. The parity operator Π̂ performs a point reflection about the origin of the
coordinate system. It acts on a general wavefunction in position representation as follows:

Π̂ψ(r) = ψ(−r) (5.24)

There are wavefunctions φπ (r) with positive parity, so that Π̂φπ (r) = φπ (−r) = φπ (r), and
those with negative parity, so that Π̂φπ (r) = φπ (−r) = −φπ (r). Wavefunctions that behave
this way are evidently eigenstates of the parity operator with an eigenvalue of plus or minus
one. Therefore, we can write:

Π̂ |φπ ⟩ = pπ |φπ ⟩ with pπ = ±1 (5.25)

Any arbitrary wavefunction can be decomposed into a part with positive parity and a part with
negative parity (i.e., into the eigenstates |φπ ⟩ of the parity operator Π̂). The parity operator is
both Hermitian and unitary; therefore, the following relationship holds:

Π̂ = Π̂† = Π̂−1 (5.26)

Depending on the symmetry of a state |φπ ⟩, the eigenvalue is either pπ = 1 (then we speak of
even states) or pπ = −1 (then we speak of odd states). This principle can also be applied to
operators. We will briefly discuss the symmetry behavior of the operators r̂, p̂, and L̂ here.

116
Quantum Theory I

A polar vector (or, due to the correspondence principle, a vector operator) indicates a direction
in space. Accordingly, when a space reflection occurs, the direction in space should also logically
change. We are already familiar with two polar operators:

Π̂r̂Π̂−1 = −r̂
(5.27)
Π̂p̂Π̂−1 = −p̂

What applies to the operators holds also for the associated polar wavefunctions:

Π̂ |r⟩ = − |r⟩
(5.28)
Π̂ |p⟩ = − |p⟩

On the other hand, an axial vector (or an axial vector operator) indicates the rotation sense
in space and does not change even when a space reflection occurs! For example, the angular
momentum operator is subject to the following:

Π̂L̂Π̂−1 = +L̂ (5.29)

This can be easily verified by inserting 1 = Π̂−1 Π̂:

Π̂L̂Π̂−1 = Π̂ (r̂ × p̂) Π̂−1 = Π̂r̂Π̂−1 × Π̂p̂Π̂−1 = (−r̂) × (−p̂) = L̂ □

It should be noted here: The parity operator leads to a point reflection and is thus not to be
confused with a plane reflection, which would lead to an entirely different symmetry behavior.

L̂

L̂
Π̂
y p̂
r̂

x
r̂
p̂

Abb. 22: Behavior of polar and axial vectors under a point reflection.

5.1.3 Commutator Relations with Scalar Operators

Similar to the commutator relationship between the angular momentum and vector operator L̂
and V̂ from (5.22), a relation can also be defined between the angular momentum operator and
a scalar operator Ŝ:

[L̂i , Ŝ] = 0 (5.30)

Such scalar operators are, for example, the squared forms of vector operators, such as the
operator L̂2 defined in (5.4), or the squared position operator r̂2 = x̂2 + ŷ 2 + ẑ 2 . In general,
scalars behave the same under rotations as (5.30).

117
 5 Angular Momentum

Example: Commutator between L̂x and r̂2

Here, we examine the commutator relationship between the angular momentum component
in the x-direction and the magnitude of the position operator r̂2 . For the individual terms
of r̂2 , the following holds:
(5.15)
[L̂x , x̂2 ] = x̂[L̂x , x̂] + [L̂x , x̂]x̂ = 0 + 0 = 0
(5.16)
[L̂x , ŷ 2 ] = ŷ[L̂x , ŷ] + [L̂x , ŷ]ŷ = ŷ(iℏẑ) + (iℏẑ)ŷ = 2iℏŷẑ
(5.17)
[L̂x , ẑ 2 ] = ẑ[L̂x , ẑ] + [L̂x , ẑ]ẑ = ẑ(−iℏŷ) + (−iℏŷ)ẑ = −2iℏŷẑ

Thus, for the sum of the three terms, we derive the desired expression; the scalar operator
r̂2 commutes, as expected, with the angular momentum operator:

[L̂x , r̂2 ] = [L̂x , x̂2 + ŷ 2 + ẑ 2 ] = [L̂x , x̂2 ] + [L̂x , ŷ 2 ] + [L̂x , ẑ 2 ] = 2iℏŷẑ − 2iℏŷẑ = 0 □

An important special case of a commutator between a vector and a scalar operator is the com-
mutator between a component of the angular momentum operator and the magnitude operator
L̂2 :
[L̂i , L̂2 ] = 0 (5.31)
This relation can be verified through a simple calculation using index exchange and renaming:
(5.23)
[L̂i , L̂2 ] = [L̂i , L̂j L̂j ] = L̂j [L̂i , L̂j ] + [L̂i , L̂j ]L̂j =
   
= L̂j iℏεijk L̂k + iℏεijk L̂k L̂j =
 
= iℏ εijk L̂j L̂k + εijk L̂k L̂j = | εijk = −εikj
 
= iℏ εijk L̂j L̂k − εikj L̂k L̂j = | second term: k ↔ j
 
= iℏ εijk L̂j L̂k − εijk L̂j L̂k = 0 □

L̂i and L̂2 are thus compatible operators and thereby lead to a common, complete orthonormal
system between the angular momentum operator and the magnitude of the angular momentum
operator. Therefore, a simultaneous, accurate measurement of the eigenvalues of L̂i and L̂2 is
possible. We will engage in constructing such an eigen system in the following chapters.

5.1.4 Ladder Operators of the Angular Momentum

Similar to the raising and lowering operators of the harmonic oscillator, raising and lowering
operators L̂± for angular momentum L̂ can also be constructed. Their specific effect will be
discussed in the next chapter, as only there the eigenstates of the angular momentum opera-
tor are introduced. Here, we primarily focus on defining L̂± and familiarizing ourselves with
useful commutator relations. The raising and lowering operator for angular momentum can be
expressed as follows:

L̂± = L̂x ± iL̂y (5.32)

In this definition, the z-component of the angular momentum is chosen as the preferred direction
according to convention. How exactly the sign is interpreted will be discussed in the next
chapter. In anticipation, we denote L̂+ as the raising operator and L̂− as the lowering operator.
Transposition and complex conjugation lead to the switch between the raising and lowering
operators:
 † (5.12)
L̂†± = L̂x ± iL̂y = L̂†x ∓ iL̂†y = L̂x ∓ iL̂y = L̂∓ (5.33)

118
Quantum Theory I

Relevant Relationships A useful relation is obtained by forming the product between the
raising and lowering operators (and vice versa). In the last step, the relationship (5.23) is used:
    
L̂+ L̂− = L̂x + iL̂y L̂x − iL̂y = L̂2x + L̂2y + i L̂y L̂x − L̂x L̂y =
(5.23)
= L̂2x + L̂2y − i[L̂x , L̂y ] = L̂2x + L̂2y + ℏL̂z (5.34)
    
L̂− L̂+ = L̂x − iL̂y L̂x + iL̂y = L̂2x + L̂2y − i L̂y L̂x − L̂x L̂y =
(5.23)
= L̂2x + L̂2y + i[L̂x , L̂y ] = L̂2x + L̂2y − ℏL̂z (5.35)

From (5.4), it follows that L̂2x + L̂2y = L̂2 − L̂2z . Substituting this into (5.34) and (5.35), we get:

L̂± L̂∓ = L̂2x + L̂2y ± ℏL̂z = L̂2 − L̂2z ± ℏL̂z (5.36)

The sum of the expressions from (5.34) and (5.35) (i.e., the anticommutator between the raising
and lowering operators) leads us to:
   
L̂+ L̂− + L̂− L̂+ = 2 L̂2x + L̂2y = 2 L̂2 − L̂2z (5.37)

We can rewrite this expression by again using the relationship L̂2x + L̂2y = L̂2 − L̂2z derived from
(5.4) and substituting this into (5.37):
1 
L̂2 = L̂+ L̂− + L̂− L̂+ + L̂2z (5.38)
2
In the next chapter, we want to derive the effect of L̂2 , L̂z , and L̂± . Both the relationship
(5.38) and the following commutator relations are very helpful. We start with the commutator
between the raising and lowering operators L̂+ and L̂− , which is easily calculated (in the last
step, the relationship (5.23) using [L̂x , L̂y ] = iℏL̂z is used):

[L̂± , L̂∓ ] = [L̂x ± iL̂y , L̂x ∓ iL̂y ] =

= [L̂x , L̂x ] + [L̂y , L̂y ] ± i([L̂y , L̂x ] − [L̂x , L̂y ]) =
(5.23)
= ∓2i[L̂x , L̂y ] = ±2ℏL̂z (5.39)

Another important relationship is the commutator between L̂z and L̂± (we use (5.23); more
precisely, the variants [L̂z , L̂x ] = iℏL̂y and [L̂z , L̂y ] = −iℏL̂x ):
(5.23)
[L̂z , L̂± ] = [L̂z , L̂x ± iL̂y ] = [L̂z , L̂x ] ± i[L̂z , L̂y ] =
  (5.32)
= iℏL̂y ± ℏL̂x = ±ℏ L̂x ± iL̂y =
= ±ℏL̂± (5.40)

Writing out the commutator in (5.40) explicitly, it becomes L̂z L̂± − L̂± L̂z = ±ℏL̂± , which can
be rewritten as:

L̂z L̂± = L̂± L̂z ± ℏL̂± (5.41)

Due to (5.31), the relationship between the commutator of L̂2 and L̂± is trivially zero:
(5.32) (5.31)
[L̂2 , L̂± ] = [L̂2 , L̂x ± iL̂y ] = [L̂2 , L̂x ] ± i[L̂2 , L̂y ] = 0 (5.42)

Now that we are equipped with a “toolbox” of useful relationships, we want to investigate the
eigenstates of L̂2 and L̂z in the next chapter.

119
 5 Angular Momentum

5.2 Eigen System of Angular Momentum

Due to (5.23), the components L̂x , L̂y , and L̂z do not commute among themselves and thus
cannot be measured simultaneously. This is a typical quantum phenomenon. While in classical
physics, angular momentum can be represented by a vector whose components obviously exist
and can be measured independently, this is not the case in quantum physics. As soon as
we measure one vector component of angular momentum, we automatically affect the other
two components of angular momentum! Because compatible operators can be found to have
a common eigen system, at least the simultaneous measurement of the magnitude of angular
momentum via L̂2 and a single angular momentum component is possible as already shown in
(5.31). We conventionally settle on the L̂z operator, i.e., the z-component of angular momentum.
Without initially imposing restrictions, we define {|a, b⟩} as the common, complete eigen system
of L̂z and L̂2 . The eigenvalue equations are:

L̂2 |a, b⟩ = a |a, b⟩ (5.43)

L̂z |a, b⟩ = b |a, b⟩ (5.44)

While the eigenvalue a of L̂2 represents the length of the angular momentum vector, the eigen-
value b of L̂z represents the z-component of the angular momentum. However, the eigenvalues
a and b must satisfy the following relationship:

0 ≤ |b|2 ≤ a with a ≥ 0 (5.45)

This makes sense: A single component of angular momentum cannot be longer than the entire
length of the angular momentum (in the classical case: L2 = x2 + y 2 + z 2 with L2 = a and
z 2 = |b|2 ). We focus on the inequality |b|2 ≤ a, which will be discussed in more detail later in
the context of the uncertainty relation.

5.2.1 Construction of an Angular Momentum Multiplet

We want to construct our angular momentum eigenstate |a, b⟩ as a linear multiplet: This means
that starting from a fixed quantum number a, we can reach every state of the multiplet {|a, b⟩}
through ladder operators. L̂± should therefore only lead to a change in the quantum number
b while leaving a unchanged. But how do the ladder
 operators
 act on an eigenstate of angular
momentum? To find this out, we let L̂z act on L̂± |a, b⟩ :

(5.41) (5.44)
 
L̂z L̂± |a, b⟩ = L̂z L̂± |a, b⟩ = (L̂± L̂z ± ℏL̂± ) |a, b⟩ = L̂± (L̂z ± ℏ) |a, b⟩ =
= L̂± (b ± ℏ) |a, b⟩ = (b ± ℏ)L̂± |a, b⟩ =
 
= (b ± ℏ) L̂± |a, b⟩ (5.46)

As required, the eigenvalue (b ± ℏ) of the L̂z operator shows that the ladder operators only act
on b. We now perform the same calculation for L̂2 , using the fact that according to (5.42), the
operators L̂2 and L̂± commute:
(5.42) (5.43)
 
L̂2 L̂± |a, b⟩ = L̂2 L̂± |a, b⟩ = L̂± L̂2 |a, b⟩ = L̂± a |a, b⟩ = aL̂± |a, b⟩ =
 
= a L̂± |a, b⟩ (5.47)

In summary, we have shown: The raising operator increases the eigenvalue of L̂z by ℏ, while
the lowering operator decreases it by ℏ. However, the raising and lowering operators do not
influence the eigenvalue of L̂2 and thus do not change the magnitude of the angular momentum.
We have thus constructed a multiplet state! By repeatedly applying L̂± , we can switch between

120
Quantum Theory I

all states corresponding to b for a fixed a. By associating the eigenvalue b ± ℏ in (5.47) with the
corresponding part of the eigenvector, we can express the action of L̂± as follows (with N as a
normalization factor):
L̂± |a, b⟩ = N |a, b ± ℏ⟩
Since the state |a, b⟩ was changed by the action of L̂± , the relation above is no longer an
eigenvalue equation! We can now determine the norm N , and thus the complete action of the
ladder operator. To do this, we rewrite the ladder operators in terms of L̂2 and L̂z , using (5.36),
as we already know their action from (5.43) and (5.44):

|N |2 ⟨a, b ± ℏ|a, b ± ℏ⟩ = ⟨L̂± (a, b)|L̂± (a, b)⟩ =

(5.33)
= ⟨a, b|L̂†± L̂± |a, b⟩ =
(5.36)
= ⟨a, b|L̂∓ L̂± |a, b⟩ =
= ⟨a, b|L̂2 − L̂2z ∓ ℏL̂z |a, b⟩ =
(5.43,5.44)
= ⟨a, b|L̂2 |a, b⟩ − ⟨a, b|L̂2z |a, b⟩ ∓ ℏ ⟨a, b|L̂z |a, b⟩ =
= ⟨a, b|a|a, b⟩ − ⟨a, b|b |a, b⟩ ∓ ℏ ⟨a, b|b|a, b⟩ =
2
 
= a − b2 ∓ ℏb ⟨a, b|a, b⟩ =
= a − b(b ± ℏ) (5.48)

We thus obtain the following relation for the action of the ladder operator on a multiplet state
|a, b⟩: q
L̂± |a, b⟩ = a − b(b ± ℏ) |a, b ± ℏ⟩ (5.49)

Let us now consider the states with the maximum and minimum eigenvalues of L̂z , calling
them bmax and bmin . Assuming such states exist is justified because we assume a fixed a, and
according to (5.45), b is bounded. For example, if we let the raising operator act on the state
with maximum angular momentum in the z-direction, we have:

L̂+ |a, bmax ⟩ = 0 (5.50)

This must be so because the maximum value of L̂z is already maximal and cannot be increased
further. The same applies for the action of the lowering operator on the minimum value of L̂z :

L̂− |a, bmin ⟩ = 0 (5.51)

We also know that the ladder operator L̂+ (except for the state |a, bmax ⟩) increases the eigenvalue
b by ℏ, and correspondingly the ladder operator L̂− (except for the state |a, bmin ⟩) decreases the
eigenvalue b by ℏ. Consider a state with an arbitrary eigenvalue b. Since a maximum value
bmax exists, there must be a corresponding number nmax times the ladder operator L̂+ needs
to be applied to reach bmax starting from b. Correspondingly, nmin is the necessary number of
applications of L̂− to reach the minimum eigenvalue bmin . We can express this as follows:

bmax = b + nmax ℏ
(5.52)
bmin = b − nmin ℏ

Therefore, the difference between the boundary eigenvalues is as follows:

bmax − bmin = ℏ(nmax + nmin ) = ℏn (5.53)

Here, n is the maximum number of consecutive identical ladder operations that can be performed.
Thus, starting from a boundary eigenvalue of the multiplet state, we can reach any other state

121
 5 Angular Momentum

of the multiplet with consistent application of a ladder operator. To show the size of n, we apply
L̂− L̂+ to |a, bmax ⟩ and consider that the result must be zero due to relation (5.50):
(5.36)
0 = L̂− L̂+ |a, bmax ⟩ =
(5.46,5.47)
 
= L̂2 − L̂2z − ℏL̂z |a, bmax ⟩ =
 
= a − b2max − ℏbmax |a, bmax ⟩ =
= b2max + ℏbmax − a (5.54)

We can solve a quadratic equation for bmax and obtain two solutions. Here we choose only the
positive solution, as we are searching for the maximum value bmax :
s s
ℏ ℏ2 ℏ ℏ2
bmax =− ± +a=− + +a
2 4 2 4

We perform an equivalent calculation for the action of the lowering operator on the minimum
state, using (5.51):
(5.36)
0 = L̂+ L̂− |a, bmin ⟩ =
(5.46),(5.47)
 
= L̂2 − L̂2z + ℏL̂z |a, bmin ⟩ =
 
= a − b2min + ℏbmin |a, bmin ⟩ =
= b2min − ℏbmin − a (5.55)

Again, we solve a quadratic equation, but this time we choose the negative solution to obtain
the minimum state: s s
ℏ ℏ2 ℏ ℏ2
bmin = ± +a= − +a
2 4 2 4
We see that bmax = −bmin must hold. This allows us, using the definition n = 2j, to rewrite the
relation (5.53) as follows:

bmax − bmin = 2bmax = nℏ = 2jℏ =⇒ bmax = jℏ (5.56)

While n must always be an integer, j can, due to our definition, indeed be a half-integer (and
thus the eigenvalue b). At the moment, however, we want to restrict j also to integer values. In
this case, by convention, j is renamed to j = l and this quantum number is called the (orbital)
angular momentum quantum number. We can now express the eigenvalue a as a function of j
(or rather l). First, we rewrite relation (5.54) for a, then insert (5.56). Finally, we rename j to
l as explained earlier:
(5.56)
a = b2max + ℏbmax = ℏ2 (j 2 + j) = ℏ2 j(j + 1) = ℏ2 l(l + 1) (5.57)

Since we now express the eigenvalue a as a function of l, it is more logical to describe the
quantum state not as |a, b⟩ but as |l, b⟩. We also perform another substitution: b → m:

b = mℏ (5.58)

Here, m is referred to as the magnetic quantum number. Consequently, the quantum state
|a, b⟩ is now represented as |l, m⟩. Thus, for the angular momentum magnitude operator L̂2 , we
obtain, using (5.43) and (5.57), the following eigenvalue equation:

122
Quantum Theory I

L̂2 |l, m⟩ = ℏ2 l(l + 1) |l, m⟩ (5.59)

For the z-component of the angular momentum operator L̂z , we obtain, taking into account
(5.44) and (5.58), an equally simple expression:

L̂z |l, m⟩ = ℏm |l, m⟩ (5.60)

The action of the raising and lowering operators can also be expressed using the quantum
numbers l and m. To do this, we substitute (5.57) and (5.58) into (5.49). This yields the
following relation:
q
L̂± |l, m⟩ = ℏ l(l + 1) − m(m ± 1) |l, m ± 1⟩ (5.61)

Here, |a, b ± ℏ⟩ was transformed into |l, m ± 1⟩ because, according to (5.58), m = b/ℏ. The
range of values for the magnetic quantum number m can be determined using inequality (5.45).
By inserting the substitutions (5.57) and (5.58), we obtain:

m2 ≤ l(l + 1) (5.62)

Assuming the extremal values of the magnetic quantum number are m = ±l, then the inequality
(5.62) is evidently satisfied, because l2 < l(l+1). On the other hand, if we try the extremal value
m = ±(l + 1), the inequality (5.62) would not hold, because (l + 1)2 > l(l + 1). Therefore, we
can write the following relationship between the orbital angular momentum quantum number l
and the magnetic quantum number m:

|m| ≤ l (5.63)

All m fulfilling (5.63) belong to a multiplet with quantum number l and can be reached via
ladder operators. Since ladder operators only affect the eigenvalue m, we cannot leave the mul-
tiplet of l using L̂± .

We will also focus on integer angular momentum quantum numbers in the following chapter and
explore the significance of half-integer angular momentum quantum numbers (still denoted by j)
later. How many possible settings of m are there for a fixed quantum number l? From (5.56), we
already know that the maximum and minimum values of the magnetic quantum number m are
fixed at mmax = l and mmin = −l. We sum over all possible m, using the Gaussian summation
formula and the substitution m = n − l:
l 2l
p(l) = 1= 1 = 2l + 1
X X

m=−l n=0

Thus, according to the orbital angular momentum quantum number l, we find a number p(l) of
independent magnetic states:

p(l) = 2l + 1 (5.64)

Figure 23 illustrates a two-dimensional angular momentum system with l = 3: While the length
of the angular momentum clearly remains constant for each realization of the multiplet, in this
case, p(3) = 2 · 3 + 1 = 7 settings for m are allowed. Additionally, the action of the raising and
lowering operators is schematically depicted.

123
 5 Angular Momentum

z z z
m=3 m=3 m=3

L̂+ L̂−
m=2 m=2 m=2

L̂+ L̂−
m=1 m=1 m=1

m=0 m=0 m=0

x x x

m = −1 m = −1 m = −1

m = −2 m = −2 m = −2

m = −3 m = −3 m = −3

Abb. 23: (left) Schematic, two-dimensional representation of an angular momentum system

with l = 3. (center) Action of the raising operator L̂+ and (right) the lowering
operator L̂− .

5.2.2 Angular Momentum Uncertainty

The commutator relation (5.23) shows that the commutator between two different angular mo-
mentum components does not vanish, and the L̂x -, L̂y -, and L̂z -components cannot be measured
sharply at the same time. Typically, the measurable quantity by convention (arbitrarily) is the
z-component of the angular momentum. This also means that the poles of the angular momen-
tum sphere (see Figure 24) can never be reached by L̂. This must naturally be so because if the
angular momentum vector pointed exactly in the (positive or negative) z-direction, both the x-
and y-components would be precisely set to zero. However, this is excluded by the commuta-
tor relation (5.23)! According to convention, only L̂z is supposed to be a sharply measurable
quantity, so it holds:
2
⟨(∆L̂z )2 ⟩ = ⟨L̂2z ⟩ − ⟨L̂z ⟩ = 0 (5.65)
For ⟨(∆L̂x )2 ⟩ and ⟨(∆L̂y )2 ⟩, this relationship no longer holds. They are smeared in the xy-plane.
That means: If one measures the x- or y-component in a system with a fixed z-component,
you will (in many repetitions of the experiment) get a different value each time! To calculate
⟨(∆L̂x )2 ⟩ and ⟨(∆L̂y )2 ⟩ concretely, it helps to express L̂x and L̂y through ladder operators. To
this end, we consider the two relations from (5.32):

L̂+ = L̂x + iL̂y and L̂− = L̂x − iL̂y (5.66)

By adding L̂+ and L̂− from (5.32), one obtains the relation L̂+ + L̂− = 2L̂x , which can be
rewritten as:

1 
L̂x = L̂+ + L̂− (5.67)
2

If L̂− is subtracted from L̂+ from (5.32), one obtains L̂+ − L̂− = 2iL̂y , which can again be
rewritten as:

1  
L̂y = L̂+ − L̂− (5.68)
2i

124
Quantum Theory I

Let us first consider the uncertainty of the x-component of the angular momentum. For the first
moment ⟨L̂x ⟩, we obtain:
(5.67)
⟨L̂x ⟩ = ⟨l, m|L̂x |l, m⟩ =
1
= ⟨l, m|L̂+ + L̂− |l, m⟩ =
2
1  (5.61)
= ⟨l, m|L̂+ |l, m⟩ + ⟨l, m|L̂− |l, m⟩ =
2
1
= (N+ ⟨l, m|l, m + 1⟩ + N− ⟨l, m|l, m − 1⟩) = 0 (5.69)
2
On average, L̂x fluctuates around zero, so the first moment ⟨L̂x ⟩ = 0 vanishes. However, because
each individual measurement yields a non-zero value, we expect the second moment ⟨L̂2x ⟩ to be
non-zero. This will be checked in the following:
(5.67)
⟨L̂2x ⟩ = ⟨l, m|L̂2x |l, m⟩ =
1
= ⟨l, m|(L̂+ + L̂− )2 |l, m⟩ =
4
1  (5.37)
= ⟨l, m|L̂2+ |l, m⟩ + ⟨l, m|L̂2− |l, m⟩ + ⟨l, m|L̂+ L̂− + L̂− L̂+ |l, m⟩ =
4
1 2 
= N+ ⟨l, m|l, m + 2⟩ + N−2 ⟨l, m|l, m − 2⟩ + 2 ⟨l, m|L̂2 − L̂2z |l, m⟩ =
4
1  (5.59,5.60)
= ⟨l, m|L̂2 |l, m⟩ − ⟨l, m|L̂2z |l, m⟩ =
2
1h 2 i
= ℏ l(l + 1)⟨l, m|l, m⟩ − m2 ℏ2 ⟨l, m|l, m⟩ =
2
ℏ2 h i
= l(l + 1) − m2 (5.70)
2
The variance ⟨(∆L̂x )2 ⟩ as a measure of the uncertainty of the x angular momentum component
can be expressed concretely as:
2 (5.69)(5.70) ℏ2 h i
⟨(∆L̂x )2 ⟩ = ⟨L̂2x ⟩ − ⟨L̂x ⟩ = l(l + 1) − m2 (5.71)
2
If we want to know the minimal uncertainty, we must insert the maximum value for m in (5.71).
According to (5.63), this is m = l, so we obtain for the minimum uncertainty:
ℏ2     ℏ2
⟨(∆L̂x )2 ⟩ ≥ l(l + 1) − l2 = l2 + 1 − l2 = l (5.72)
2 2
Finally, to obtain the value for the smallest possible uncertainty, we only need to replace l with
the smallest possible value of the quantum number l in (5.72). Since l is the quantum number
for angular momentum (and can only take integer values), the minimum angular momentum
setting is l = lmin = 1. Inserting this into (5.72), we find:
ℏ2
⟨(∆L̂x )2 ⟩ ≥
for l = lmin = 1 (5.73)
2
While l = 0 would also be a valid orbital angular momentum quantum number, the angular
momentum would vanish and thus would not be meaningful for a discussion of uncertainty. If
half-integer angular momentum values are possible, we replace l with j. The minimum value is
j = jmin = 21 , giving:
ℏ2 1
⟨(∆L̂x )2 ⟩ ≥ for j = jmin = (5.74)
4 2
The same results are obtained following an analogous calculation for the y angular momentum
component.

125
 5 Angular Momentum

In-Depth: Robertson-Schrödinger Uncertainty Relations

The starting point is the Robertson Uncertainty Relation, which can be expressed in the
form of the following inequality for the product of two standard deviations of the operators
 and B̂:
1
∆A · ∆B ≡ σA σB ≥ ⟨[Â, B̂]⟩
2
It is also supposed to hold that [Â, B̂] ̸= 0. To abbreviate the notation of the following
scalar products, we introduce the definitions for the standard deviations σA and σB :
2
2
σA = ⟨Â − ⟨Â⟩ |Â − ⟨Â⟩⟩ = ⟨Â2 ⟩ − ⟨Â⟩ =⇒ |α⟩ = |Â − ⟨A⟩⟩
= ⟨B̂ − ⟨B̂⟩ |B̂ − ⟨B̂⟩⟩ = ⟨Bˆ2 ⟩ − ⟨B̂⟩
2
2
σB =⇒ |β⟩ = |B̂ − ⟨B⟩⟩

Note that, actually, |Â − ⟨A⟩⟩ = |(Â − ⟨A⟩)ψ⟩ holds, since the operators always act on an
underlying state. Over the Cauchy-Schwarz inequality from (3.18), we can estimate the
product σA σB (it holds ||α|| = ⟨α|α⟩):
p

σB = ⟨α|α⟩ ⟨β|β⟩ ≥ | ⟨α|β⟩ |2

2 2
σA (5.75)

The scalar product of α and β can be expressed using the appropriate commutator and
anticommutator:
1 1 1
⟨α|β⟩ = ⟨αβ⟩ = ⟨αβ + αβ⟩ = ⟨αβ − βα + αβ + βα⟩ = ⟨[α, β] + {α, β}⟩ (5.76)
2 2 2
In general, the scalar product corresponds to a complex number. We could already show
that the commutator is always anti-hermitian, and therefore, its eigenvalue is purely com-
plex, while the anticommutator is hermitian and has a purely real eigenvalue. Taking the
magnitude of ⟨α|β⟩ follows the same rules as taking the magnitude of a complex number
|z|2 = |x + iy|2 = |x|2 + |y|2 :
(5.76) 1 1 1
| ⟨α|β⟩ |2 = | ⟨[α, β]⟩ + ⟨{α, β}⟩ |2 = | ⟨[α, β]⟩ |2 + | ⟨{α, β}⟩ |2 (5.77)
4 4 4
We insert (5.77) into (5.75) and take the square root of the entire expression; the result is
termed the Robertson-Schrödinger Uncertainty Relation:
1q 1
σA σB ≥ | ⟨[α, β]⟩ |2 + | ⟨{α, β}⟩ |2 ≥ | ⟨[α, β]⟩ | (5.78)
2 2
Only the last estimation leads us to the desired expression of the Robertson Uncertainty,
which is trivial by neglecting the second term. We can verify the relation from (5.78) on
the already well-known Heisenberg uncertainty relation σx σp ≥ ℏ/2:

(5.78) 1 (3.132) 1 ℏ
σx σp ≥ | ⟨[x̂, p̂]⟩ | = | ⟨iℏ⟩ | =
2 2 2

The relationship from (5.74) can also be easily confirmed with the Robertson Uncertainty Rela-
tion from (5.78); the following connection holds:

1 ℏ (5.60) ℏ
2
σLx σLy ≥ | ⟨[L̂x , L̂y ]⟩ | = | ⟨L̂z ⟩ | = m (5.79)
2 2 2
In the case of an angular momentum setting at maximum m, it also holds that m = l. Since the
uncertainty product σLx σLy is non-zero, there will always be angular momentum components in
the x and y directions for every quantum number m (if l ̸= 0, this is also true for m = 0, as seen

126
Quantum Theory I

from (5.73)). Only in the case l = 0 do the x and y components of angular momentum vanish;
this does not contradict the uncertainty relation since no angular momentum exists when l = 0.
z
m=3

m=2

m=1

m=0
y

m = −1
x

m = −2

m = −3

Abb. 24: Schematic, three-dimensional representation of an angular momentum system with

l = 3: While the z-component is sharply defined, the x and y components are smeared
and lie in this sketch on the intersections of the sphere and the corresponding cones.

Let’s recall the start of the chapter, specifically (5.45). Here, the inequality 0 ≤ |b|2 ≤ a
or 0 ≤ |m|2 ≤ l(l + 1) becomes clear: If equality |m|2 = l(l + 1) were to hold, the product
σLx σLy would vanish according to (5.74) and thus allow all angular momentum components
to be determined exactly. We have, however, already recognized that this is non-physical and
cannot correspond to reality. Only without angular momentum l = 0 may the equality relation
be fulfilled!

5.3 Angular Momentum in Position Space

Until now, we have come to know the eigenstates of angular momentum as angular momentum
multiplets as abstract states in the angular momentum eigenspace. However, to verify our
predictions in experiments, we must continue our considerations in position space. This is
achieved by projection onto {|r⟩}, where we will transition from Cartesian coordinates to a
spherically symmetric coordinate system {|r⟩} ≡ {|ϑ, φ⟩}. It holds:

Ylm (r) = ⟨r|l, m⟩ =⇒ Ylm (ϑ, φ) = ⟨ϑ, φ|l, m⟩ (5.80)

Later, we will associate Ylm (r) with spherical harmonics, but for now, Ylm (r) should just rep-
resent the position representation of an angular momentum eigenstate. The relevant eigenvalue
equations from (5.59) and (5.60) thus result in:

L̂2 Ylm (r) = ℏ2 l(l + 1)Ylm (r) (5.81)

L̂z Ylm (r) = mℏYlm (r) (5.82)

But what do the eigenfunctions Ylm (r) look like concretely? In the following chapter, we will
perform the transformation of L̂ into spherical coordinates and will find an explicit representation
of |l, m⟩ in position space!

5.3.1 Angular Momentum in Spherical Coordinates

As a repetition: The angular momentum in operator notation is given by L̂ = r̂ × p̂. We know
from position and momentum operators the cartesian position representation, and have found

127
 5 Angular Momentum

the following expression in (5.2):

 
y∂z − z∂y
L̂ = −iℏ z∂x − x∂z 
 
x∂y − y∂x
However, it is more natural to express the angular momentum in spherically-symmetric coor-
dinates. We must therefore represent both {x, y, z} and the partial derivatives {∂x , ∂y , ∂z } in
spherical coordinates. For a position vector r, the transformation occurs via:
r sin(ϑ) cos(φ)
   
x
r = y  =  r sin(ϑ) sin(φ)  (5.83)
   
z r cos(ϑ)
For the present problem, however, the inversion of (5.83) is relevant, since we want to transform
from the Cartesian coordinate system to spherical coordinates. Respectively, it holds:
q
r= x2 + y 2 + z 2 (5.84)
!
z
ϑ = arccos p 2 (5.85)
x + y2 + z2
y
 
φ = arctan (5.86)
x
From Figure 25, the relations in (5.84) become easily understandable. While there are no
restrictions for the Cartesian coordinate system ({x, y, z} ∈ R), for {r, ϑ, φ}, we must note that
the radius r ∈ [0, +∞) must be positive and the polar angle ϑ ∈ [0, π] is not defined over the
entire angle range, while this is the case for the azimuthal angle φ ∈ [0, 2π].
z

y
φ

Abb. 25: Coordinates of a spherically symmetric system; here, only the first octant is shown for
simplicity. We recognize that every point of the coordinate system can be uniquely
identified by {r, ϑ, φ}.

Let us turn to the partial derivatives. From (5.83), it becomes apparent that in the Cartesian
coordinate system x(r, ϑ, φ), y(r, ϑ, φ), and z(r, ϑ, φ) all depend on the radius r, as well as
the polar angle ϑ and the azimuthal angle φ. This allows for an extension of the respective
differential operators ∂x , ∂y , and ∂z using the chain rule – thus, it follows:
∂ ∂r ∂ ∂ϑ ∂ ∂φ ∂
∂x ≡ = + +
∂x ∂x ∂r ∂x ∂ϑ ∂x ∂φ
∂ ∂r ∂ ∂ϑ ∂ ∂φ ∂
∂y ≡ = + + (5.87)
∂y ∂y ∂r ∂y ∂ϑ ∂y ∂φ
∂ ∂r ∂ ∂ϑ ∂ ∂φ ∂
∂z ≡ = + +
∂z ∂z ∂r ∂z ∂ϑ ∂z ∂φ

128
Quantum Theory I

For the total differential of a function that depends on Cartesian coordinates (for example,
f = r(x, y, z), f = ϑ(x, y, z), or f = φ(x, y, z)), it can be written:

∂f ∂f ∂f
df = dx + dy + dz (5.88)
∂x ∂y ∂z

The actual forms of dr, dϑ, and dφ can be found in (5.92), (5.96), and (5.100).

Example: Derivation of dr, dϑ, and dφ

Radius r: This part of the derivation is quite simple: We differentiate (5.84) with respect
to x, y, and z, then substitute x + y 2 + z 2 in the denominator with r and convert the
p
2

numerator using (5.83) into spherical coordinates. Thus, the following results:

∂r 1 2x (5.84) x (5.83) r sin(ϑ) cos(φ)

= p 2 = = = sin(ϑ) cos(φ) (5.89)
∂x 2 x + y2 + z2 r r
∂r 1 2y (5.84) y (5.83) r sin(ϑ) sin(φ)
= p 2 = = = sin(ϑ) sin(φ) (5.90)
∂y 2 x + y2 + z2 r r
∂r 1 2z (5.84) z (5.83) r cos(ϑ)
= p 2 = = = cos(ϑ) (5.91)
∂z 2 x + y2 + z2 r r

For the total differential dr, we finally insert (5.89–5.91) into (5.88):

∂r ∂r ∂r
dr = dx + dy + dz = sin(ϑ) cos(φ)dx + sin(ϑ) sin(φ)dy + cos(ϑ)dz (5.92)
∂x ∂y ∂z
Polar Angle dϑ: This derivation is similar, but involves a little more calculation effort.
We start by differentiating the expression (5.85) with respect to x, y, and z:

∂ϑ 1 1 z 1 xz
 
(5.84)
= −q − 2x = q =
∂x 1− 2 z2 2 (x2 + y 2 + z 2 )
3
2 1− z2 r3
x +y 2 +z 2 r2
1 xz xz (5.84) xz (5.83)
= 1√ = √ = =
3
x2 + y 2
p
r r2 − z2 r r r2 − z 2
2 r2
r2 sin(ϑ) cos(φ) cos(ϑ)
= q =
r2 r2 sin2 (ϑ) cos2 (φ) + r2 sin2 (ϑ) sin2 (φ)
sin(ϑ) cos(φ) cos(ϑ) cos(φ) cos(ϑ)
=q = (5.93)
r2 sin2 (ϑ) cos2 (φ) + sin2 (φ)
  r
∂ϑ 1 1 z 1 yz
 
(5.84)
= −q − 2y = q =
∂y 1− 2 z2 2 (x2 + y 2 + z 2 )
3
2 1− z2 r3
x +y 2 +z 2 r2
1 yz yz (5.84) yz (5.83)
= 1√ = √ = =
3
x2 + y 2
p
r r2 − z2 r r r2 − z 2
2 r2
r2 sin(ϑ) sin(φ) cos(ϑ)
= q =
r2 r2 sin2 (ϑ) cos2 (φ) + r2 sin2 (ϑ) sin2 (φ)
sin(ϑ) sin(φ) cos(ϑ) sin(φ) cos(ϑ)
=q  = (5.94)
r sin (ϑ) cos (φ) + sin (φ)
2 2

2 2 r

1 1 1
!
∂ϑ z (5.84)
= −q − 2z =
x +y +z 2 (x2 + y 2 + z 2 ) 32
p
∂z 1− 2 z2 2 2 2
x +y 2 +z 2

129
 5 Angular Momentum

2
1 1 z2 1 1 z2 1 − zr2
! !
(5.84)
= −q − = − 1√ − = −√ =
1− z2 r r3 r r2 − z2 r r3 r2 − z 2
r2
z2
1− r2 (5.83) 1 − cos2 (ϑ)
= −p = −q =
x2 + y 2 r2 sin2 (ϑ) cos2 (φ) + r2 sin2 (ϑ) sin2 (φ)
sin2 (ϑ) sin(ϑ)
= −q  =− (5.95)
r2 sin2 (ϑ)

cos2 (φ) + sin2 (φ) r

Substituting (5.93–5.95) into the total derivative (5.88), we obtain the expression:

∂ϑ ∂ϑ ∂ϑ cos(φ) cos(ϑ) sin(φ) cos(ϑ) sin(ϑ)

dϑ = dx + dy + dz = dx + dy − dz (5.96)
∂x ∂y ∂z r r r
Azimuthal Angle dφ: This is finally the analogous computation of the total derivative
dφ:

∂φ 1 y x2 y
 
(5.83)
= − 2 =− =
∂x 2
1 + xy 2 x x2 + y 2 x2
r sin(ϑ) sin(φ)
=− =
r2 sin2 (ϑ) cos2 (φ)
+ r2 sin2 (ϑ) sin2 (φ)
r sin(ϑ) sin(φ) sin(φ)
=− 2 2  2  =− (5.97)
r sin (ϑ) cos (φ) + sin (φ)
2 r sin(ϑ)
∂φ 1 1 1 x x (5.83)
= = = 2 =
∂y y 2
1+ 2 x y
1+ 2 x
2 2 x +y 2
x x
r sin(ϑ) cos(φ)
= 2 2 =
r sin (ϑ) cos2 (φ) + r2 sin2 (ϑ) sin2 (φ)
r sin(ϑ) cos(φ) cos(φ)
= 2 2  2  = (5.98)
r sin (ϑ) cos (φ) + sin2 (φ) r sin(ϑ)
∂φ
=0 (5.99)
∂z
Substituting results (5.97–5.99) again into (5.88), we obtain for dφ:

∂φ ∂φ ∂φ sin(φ) cos(φ)
dφ = dx + dy + dz = − dx + dy (5.100)
∂x ∂y ∂z r sin(ϑ) r sin(ϑ)

Transformation Matrix: We can summarize all partial derivatives in a transformation

matrix T:
cos(ϑ) cos(φ)
− rsin(φ)
 
∂φ  sin(ϑ) cos(φ)
 ∂r ∂ϑ
∂x ∂x ∂x r sin(ϑ)
∂φ 
T =  ∂y
∂r ∂ϑ
=
 sin(ϑ) sin(φ)
cos(ϑ) sin(φ) cos(φ)
(5.101)
  

∂y ∂y  r r sin(ϑ) 
∂r ∂ϑ ∂φ sin(ϑ)
∂z ∂z ∂z cos(ϑ) − r 0

Through matrix-vector multiplication, we can transform the vector (∂x , ∂y , ∂z )T into spher-
ically symmetric coordinates in one step – the partial derivatives are accordingly written in
the required form in (5.87). Using the inverse of the transformation matrix T , we can also
calculate the partial differentials in spherical coordinates based on the Cartesian deriva-
tives. For the matrix-vector multiplication, the following holds in our case in a compact

130
Quantum Theory I

form:    
∂x ∂r
∂y  = T ∂ϑ  (5.102)
   
∂z ∂φ

To write the Cartesian, partial derivatives explicitly over the spherical coordinates {r, ϑ, φ}, we
substitute the respective components from the transformation matrix (5.101) into (5.102). The
following are three expressions for the Cartesian, partial derivatives, expressed in the desired
coordinates:
∂ ∂ cos(ϑ) cos(φ) ∂ sin(φ) ∂
∂x ≡ = sin(ϑ) cos(φ) + − (5.103)
∂x ∂r r ∂ϑ r sin(ϑ) ∂φ
∂ ∂ cos(ϑ) sin(φ) ∂ cos(φ) ∂
∂y ≡ = sin(ϑ) sin(φ) + + (5.104)
∂y ∂r r ∂ϑ r sin(ϑ) ∂φ
∂ ∂ sin(ϑ) ∂
∂z ≡ = cos(ϑ) − (5.105)
∂z ∂r r ∂ϑ

Example: Derivation of L̂x , L̂y , and L̂z

To represent the individual components of the angular momentum operator L̂ in spherically-

symmetric coordinates, we use the previously found relationships of the Cartesian partial
derivatives from (5.103), (5.104), and (5.105). Substituting them in, considering the re-
spective transformations in (5.2), we obtain the following relations.

x-Component: According to (5.2), it holds for L̂x = −iℏ(y∂z − z∂y ). Substituting the
transformed relationships for y, z, ∂y , and ∂z , we obtain:

∂ ∂
 
(5.83)
L̂x = − iℏ y −z =
∂z ∂y
∂ ∂ (5.104)(5.105)
 
= − iℏ r sin(ϑ) sin(φ) − r cos(ϑ) =
∂z ∂y
∂ sin(ϑ) ∂
  
= − iℏ r sin(ϑ) sin(φ) cos(ϑ) − −
∂r r ∂ϑ
∂ cos(ϑ) sin(φ) ∂ cos(φ) ∂
 
− r cos(ϑ) sin(ϑ) sin(φ) + + =
∂r r ∂ϑ r sin(ϑ) ∂φ
 ∂ cos(ϑ) ∂
  
= − iℏ − sin(φ) cos2 (ϑ) + sin2 (ϑ) − cos(φ) =
∂ϑ sin(ϑ) ∂φ
∂ ∂
 
= − iℏ − sin(φ) − cot(ϑ) cos(φ)
∂ϑ ∂φ

y-Component: According to (5.2), it holds for L̂y = −iℏ(z∂x − x∂z ). Substituting the
transformed relationships for x, z, ∂x , and ∂z , we can determine the partial differential

131
 5 Angular Momentum

operator in spherically-symmetric coordinates and thus obtain the expression:

∂ ∂ (5.83)
 
L̂y = − iℏ z −x =
∂x ∂z
∂ ∂ (5.103)(5.105)
 
= − iℏ r cos(ϑ) − r sin(ϑ) cos(φ) =
∂x ∂z
∂ cos(ϑ) cos(φ) ∂ sin(φ) ∂
  
= − iℏ r cos(ϑ) sin(ϑ) cos(φ) + − −
∂r r ∂ϑ r sin(ϑ) ∂φ
∂ sin(ϑ) ∂
 
− r sin(ϑ) cos(φ) cos(ϑ) − =
∂r r ∂ϑ
 ∂ cos(ϑ) ∂
  
= − iℏ cos(φ) cos2 (ϑ) + sin2 (ϑ) − sin(φ) =
∂ϑ sin(ϑ) ∂φ
∂ ∂
 
= − iℏ cos(φ) − cot(ϑ) sin(φ)
∂ϑ ∂φ

z-Component: According to (5.2), it holds again for L̂z = −iℏ(x∂y − y∂x ). Thus, sub-
stituting the transformed relationships for x, y, ∂x , and ∂y , it follows:

∂ ∂
 
(5.83)
L̂z = − iℏ x −y =
∂y ∂x
∂ ∂ (5.103)(5.104)
 
= − iℏ r sin(ϑ) cos(φ) − r sin(ϑ) sin(φ) =
∂y ∂x
∂ cos(ϑ) sin(φ) ∂ cos(φ) ∂
  
= − iℏ r sin(ϑ) cos(φ) sin(ϑ) sin(φ) + + −
∂r r ∂ϑ r sin(ϑ) ∂φ
∂ cos(ϑ) cos(φ) ∂ sin(φ) ∂
 
− r sin(ϑ) sin(φ) sin(ϑ) cos(φ) + − =
∂r r ∂ϑ r sin(ϑ) ∂φ
cos(φ) ∂ sin(φ) ∂
    
= − iℏ r sin(ϑ) cos(φ) + r sin(ϑ) sin(φ)
r sin(ϑ) ∂φ r sin(ϑ) ∂φ
  ∂ ∂
= − iℏ cos2 (φ) + sin2 (φ) = −iℏ
∂φ ∂φ

With more or less elaborate calculations, we could express the Cartesian angular momentum
components in spherical coordinates (via the angle coordinates ϑ and φ). The radius r actually
plays no role! Here are the results found for L̂x , L̂y , and L̂z summarized again. For the x-
component of the angular momentum, it holds:

∂ ∂
 
L̂x = iℏ sin(φ) + cot(ϑ) cos(φ) (5.106)
∂ϑ ∂φ

In its form, the y-component of the angular momentum is very similar to L̂x . It follows:

∂ ∂
 
L̂y = iℏ − cos(φ) + cot(ϑ) sin(φ) (5.107)
∂ϑ ∂φ

For L̂z , we obtain the simplest differential expression – the z-component of the angular momen-
tum only acts on the azimuthal angle. We can write L̂z as:

∂
L̂z = −iℏ (5.108)
∂φ

132
Quantum Theory I

Example: Derivation of L̂2

In the derivation of L̂2 , we use the previously found relationships from (5.106), (5.107), and
(5.108). Compared to the previous calculations, the derivation of the magnitude operator
of the angular momentum L̂2 = L̂2x + L̂2y + L̂2z in spherical coordinates is somewhat more
laborious, since we must consider that through differential operators, the chain rule must
be applied. However, we will recognize that a large portion of the terms obtained cancel
out:
(5.106)(5.107)(5.108)
L̂2 =L2x + L2y + L2z =
∂ ∂ ∂ ∂
  
=−ℏ 2
sin(φ) + cot(ϑ) cos(φ) sin(φ) + cot(ϑ) cos(φ) +
∂ϑ ∂φ ∂ϑ ∂φ
∂ ∂ ∂ ∂ ∂2
   
+ − cos(φ) + cot(ϑ) sin(φ) − cos(φ) + cot(ϑ) sin(φ) + =
∂ϑ ∂φ ∂ϑ ∂φ ∂φ2
∂2 ∂2 ∂ ∂
  
= − ℏ2 sin2 (φ) 2 + cot2 (ϑ)cos2 (φ) 2 + sin(φ) cot(ϑ) cos(φ) +
∂ϑ ∂φ ∂ϑ ∂φ
∂ ∂ ∂2 ∂2
 
+ cot(ϑ) cos(φ) sin(φ) + cos2 (φ) 2 + cot2 (ϑ)sin2 (φ) 2 −
∂φ ∂ϑ ∂ϑ ∂φ
∂ ∂ ∂ ∂ ∂2
    
− cos(φ) cot(ϑ) sin(φ) − cot(ϑ) sin(φ) cos(φ) + =
∂ϑ ∂φ ∂φ ∂ϑ ∂φ2
 2
∂   ∂ 2 sin(φ) cos(φ) ∂ ∂2
= − ℏ2 + 1 + cot 2
(ϑ) − + cot(ϑ) sin(φ) cos(φ) +
∂ϑ2 ∂φ2 sin2 (ϑ) ∂φ ∂ϑ∂φ
∂ ∂2 sin(φ) cos(φ) ∂
 
+ cot(ϑ)cos(φ) cos(φ) + sin(φ) + −
∂ϑ ∂ϑ∂φ sin2 (ϑ) ∂φ
∂2 ∂ ∂2
 
− cot(ϑ) sin(φ) cos(φ) − cot(ϑ)sin(φ) − sin(φ) + cos(φ) =
∂ϑ∂φ ∂ϑ ∂ϑ∂φ
sin2 (ϑ) + cos2 (ϑ) ∂ 2
 2
∂ ∂

= − ℏ2 + + cot(ϑ) =
∂ϑ2 sin2 (ϑ) ∂φ2 ∂ϑ
1
 2
∂ ∂ ∂2

= − ℏ2 + cot(ϑ) + =
∂ϑ2 ∂ϑ sin2 (ϑ) ∂φ2
1 ∂ ∂ 1 ∂2
   
=−ℏ 2
sin(ϑ) + =
sin(ϑ) ∂ϑ ∂ϑ sin2 (ϑ) ∂φ2
ℏ2 ∂ ∂ ∂2
   
=− sin(ϑ) sin(ϑ) +
sin2 (ϑ) ∂ϑ ∂ϑ ∂φ2

The magnitude of the angular momentum operator L̂2 can also be expressed as a function of ϑ
and φ. From the above derivation, we obtain the following result:
" #
ℏ2 ∂ ∂ ∂2
 
L̂ = − 2
2
sin(ϑ) sin(ϑ) + (5.109)
sin (ϑ) ∂ϑ ∂ϑ ∂φ2

Since the individual components L̂x , L̂y , and L̂z were independent of the radius, L̂2 is naturally
independent of r as well. As we will show in more detail later, the last term from (5.109)
essentially corresponds to the square of L̂z ; this is practical as it simplifies the calculation of the
following differential equation. Moreover, the functional form of L̂2 is very similar to the Laplace
operator ∆ in spherical coordinates, which we will utilize for deriving the wave functions of the
hydrogen atom.

133
 5 Angular Momentum

Example: Derivation of L̂±

The derivation of L̂± = L̂x ± iL̂y in spherical coordinates is again significantly simpler. We
obtain just by inserting (5.106) and (5.107):

∂ ∂ ∂ ∂
   
L̂± = iℏ sin(φ)
+ cot(ϑ) cos(φ) ± i − cos(φ) + cot(ϑ) sin(φ) =
∂ϑ ∂φ ∂ϑ ∂φ
∂ ∂
 
= iℏ (sin(φ) ∓ i cos(φ)) + cot(ϑ) (cos(φ) ± i sin(φ)) =
∂ϑ ∂φ
∂ ∂
 
= iℏ ∓i (cos(φ) ± sin(φ)) + cot(ϑ) (cos(φ) ± i sin(φ)) =
∂ϑ ∂φ
∂ ∂
 
= ℏe±iφ ± + i cot(ϑ)
∂ϑ ∂φ

For the raising and lowering operators L̂+ and L̂− , we therefore find the following expression
collectively:

∂ ∂
 
L̂± = ℏe±iφ ± + i cot(ϑ) (5.110)
∂ϑ ∂φ

Canonical Commutation Relation For the position and momentum operators x̂ and p̂, we
learned the canonical commutation relation [x̂, p̂] = iℏ in Cartesian coordinates in (3.133). In
spherically symmetric spherical coordinates, we can now observe an equivalent relation between
the azimuthal angle φ̂ and the z-component of the angular momentum operator L̂z :

[φ̂, L̂z ] = iℏ (5.111)

The relation (5.111) can be easily verified by transitioning into the position representation
(keeping in mind that the operators act on a wave function, and therefore the chain rule must
be considered):
∂ ∂ ∂ ∂ ∂
     
[φ̂, L̂z ]ψ = −iℏ φ, ψ = −iℏ φ − φ ψ = −iℏ φ ψ − (φψ) (5.112)
∂φ ∂φ ∂φ ∂φ ∂φ
∂ ∂φ ∂
 
= −iℏ φ − −φ ψ = iℏψ □ (5.113)
∂φ ∂φ ∂φ
The azimuthal angle φ is thus the complementary quantity to the z-component of the angular
momentum L̂z .

5.3.2 Legendre Polynomials and Spherical Harmonics

In the previous chapter, we got to know the position representation of the operators L̂2 and L̂z ,
where we initially left out a specific representation of the eigenfunctions. First, let’s recap our
two fundamental eigenvalue equations (5.59) and (5.60) in the required representation space:
" #
ℏ2 ∂ ∂ ∂2
 
(5.109) (5.59)
L̂ 2
Ylm (ϑ, φ) = − 2 sin(ϑ) sin(ϑ) + 2 Y m (ϑ, φ) =
sin (ϑ) ∂ϑ ∂ϑ ∂ φ l
= ℏ2 l(l + 1)Ylm (ϑ, φ) (5.114)
(5.108) ∂ (5.60)
L̂z Ylm (ϑ, φ) = −iℏ Ylm (ϑ, φ) = ℏmYlm (ϑ, φ) (5.115)
∂φ
We recognize that the eigenvalue equation (5.115) of the z-component of the angular momentum
operator depends solely on φ and there is also a clear separation between the ϑ- and φ-dependent

134
Quantum Theory I

terms in (5.114). It is therefore likely that the eigenvalue equation of L̂2 will be solvable by a
separation approach. We define Ylm = F (ϑ)Φ(φ) and directly substitute this approach into
(5.115). F (ϑ) cancels out, and we obtain a simple differential equation for Φ(φ):
∂ i
−iℏF (ϑ) Φ(φ) = ℏmF (ϑ)Φ(φ) | ·
∂φ ℏ
dΦ(φ))
= imΦ(φ) (5.116)
dφ
This differential equation can be solved quite simply with the following approach:
Φ(φ) = C e+imφ (5.117)
The quantum number m in this simple form thus determines the phase of the solution functions.
Let’s now consider the eigenvalue equation for L̂2 in (5.114) more closely. The derivative term
∂2
∂2φ
appears in this equation. But we know from (5.108) that L̂z = −iℏ∂φ contains precisely
this derivative. If we square L̂z in this representation, we find a way to introduce the operator
L̂z into the eigenvalue equation of L̂2 (5.114):
∂2 ∂2 L̂2z
L̂2z = −ℏ2 =⇒ = − (5.118)
∂φ2 ∂φ2 ℏ2
We can insert the expression obtained in this way into (5.114), and transform the ϑ-dependent
differential equation accordingly:
( " # )
ℏ2 ∂ ∂ ∂2
 
(5.118)
0= − 2 sin(ϑ) sin(ϑ) + 2 − ℏ2 l(l + 1) Ylm (ϑ, φ) =
sin (ϑ) ∂ϑ ∂ϑ ∂ φ
1 ∂ ∂ 1
     
(5.115)
= sin(ϑ) sin(ϑ) − 2 L̂2z + l(l + 1) F (ϑ)Φ(φ) =
sin2 (ϑ) ∂ϑ ∂ϑ ℏ
1
( " # )
∂ ∂ ℏ2 m2
 
= sin(ϑ) sin(ϑ) − + l(l + 1) F (ϑ)Φ(φ) =
sin2 (ϑ) ∂ϑ ∂ϑ ℏ2
1
( " #)
∂ ∂ m2
  
= sin(ϑ) sin(ϑ) + l(l + 1) − F (ϑ)Φ(φ)
sin (ϑ)
2 ∂ϑ ∂ϑ sin2 (ϑ)
Admittedly, this expression is somewhat unwieldy. Therefore, we shall introduce a substitution
that will significantly simplify writing the differential equation and its solution in the further
course: p
u = cos(ϑ) =⇒ sin(ϑ) = 1 − u2
Whatever applies for u must also apply for the differentials du. Considering this, we obtain:
du
= − sin(ϑ) =⇒ du = − sin(ϑ)dϑ
dϑ
The derivative with respect to ϑ can also be adapted to the new variable u:
d du d d
= = − sin(ϑ)
dϑ dϑ du du
If we now apply the substitution directly, the representation of the differential equation from
(5.114) changes noticeably. It becomes a Legendre’s Differential Equation:
1
( " #)
∂ ∂ m2
  
0= (1 − u2 ) (1 − u2 ) + l(l + 1) − f (u) =
1−u2 ∂u ∂u 1 − u2
( " #)
∂ ∂ m2
 
= (1 − u2 ) + l(l + 1) − f (u) =
∂u ∂u 1 − u2
( " #)
∂2 ∂ m2
= (1 − u ) 2 − 2u
2
+ l(l + 1) − f (u) (5.119)
∂u ∂u 1 − u2

135
 5 Angular Momentum

How (5.119) can indeed be solved will be discussed in more detail in Appendix 10.3. Here we
anticipate the solution in the form of the Rodrigues Formula of Legendre’s Differential Equa-
tion:
s
(−1)l+m 2l + 1 (l − m)! +imφ m dm+l
Ylm (ϑ, φ) = e sin (ϑ) sin2l (ϑ) (5.120)
2l l! 4π (l + m)! d cos(ϑ)m+l

Apart from the more or less complicated prefactor, we have a φ-dependent part that we call
a phase factor and is already known from (5.117). The ϑ-dependent term corresponds to the
associated Legendre polynomials and depends on both the orbital angular momentum and the
magnetic quantum number l and m. If we are interested in spherical harmonics with specific
quantum numbers l and m, we either evaluate (5.120) or look it up in tabulated works.

In-Depth: Construction of Spherical Harmonics through Ladder Operators

Similar to the eigenfunctions of the harmonic oscillator, we can also construct the eigen-
functions of L̂2 and L̂z by the application of ladder operators. If we carry this out directly
in the position space, the resulting functions of the multiplet agree with (5.120).

Let’s begin with the position representation of the ladder operators L̂± of angular momen-
tum:
∂ ∂
 
L̂± = ℏe±iφ ± + i cot(ϑ)
∂ϑ ∂φ
If L̂± acts on a state at the edge of the multiplet, the expression vanishes since, for example,
a minimal state cannot be further decreased. If L̂− acts on a spherical harmonic function
with m = −l, we obtain the trivial expression L̂− Yl−l = 0. From (5.110) and (5.120) it
follows:
∂ ∂ ∂F (ϑ)
   
−iφ
0 = ℏe − + i cot(ϑ) F (ϑ)e−ilφ = ℏe−i(l+1)φ − + l cot(ϑ)F (ϑ)
∂ϑ ∂φ ∂ϑ
The expression in the bracket must vanish for the equation to be satisfied. To solve the
resulting differential equation, we substitute with u = sin(ϑ):

1 cos(ϑ)
Z Z
ln(F (ϑ)) + C = l dϑ cot(ϑ) = l du = ln(ul )
cos(ϑ) u

This way, we have found – without considering normalization – the minimal state of the
angular momentum multiplet l! It corresponds to:

F (ϑ) = N sinl (ϑ) (5.121)

To construct arbitrary spherical harmonic functions Ylm from (5.121), we need to know the
position representation of a p-fold application of a ladder operator L̂± . Simple application
gives:
∂ ∂
 
L̂± Yln = ℏe±iφ ± + i cot(ϑ) F (ϑ)e+inφ =
∂ϑ ∂φ
d
 
= ℏei(n±1)φ ± − n cot(ϑ) F (ϑ) =
dϑ
1 d
 
= ℏei(n±1)φ ± sin(ϑ) − n cos(ϑ) F (ϑ) =
sin(ϑ) dϑ

136
Quantum Theory I

sin1±n (ϑ) d  ∓n
= ±ℏei(n±1)φ sin (ϑ)F (ϑ) =

sin(ϑ) dϑ
d  ∓n
= ∓ℏei(n±1)φ sin1±n (ϑ) sin (ϑ)F (ϑ)

d cos(ϑ)

With the right extensions, one can exploit the fact that we can transform the additive
expression in the brackets into a product term using the chain rule. Applying L̂± a second
time allows us to ignore the additive representation of L̂± and start directly with the
simplified expression in product form. It is easily noticeable that primarily the degree of
differentiation within the function changes, as well as the power of the sinus and exponential
functions not affected by differentiation. We obtain the following expression:
d
 
 ∓n
L̂2± Yln = L̂± ∓ℏei(n±1)φ sin1±n (ϑ) sin (ϑ)F (ϑ) =

d cos(ϑ)
d d
 
∓n+1  ∓n
= (∓ℏ) e
2 i(n±2)φ
sin 2±n
(ϑ) sin (ϑ) sin 1±n
(ϑ) sin (ϑ)F (ϑ) =

d cos(ϑ) d cos(ϑ)
d2  ∓n
= (∓ℏ)2 ei(n±2)φ sin2±n (ϑ) sin (ϑ)F (ϑ)

d cos(ϑ) 2

Iteratively, we can generalize this result for a p-fold application of L̂p± :

dp
L̂p± Yln = (∓ℏ)p ei(n±p)φ sinp±n (ϑ)
 ∓n
sin (ϑ)F (ϑ) (5.122)

d cos(ϑ) p

Let n = −l and p = l + m, then we obtain – again without normalization – the spherical

harmonic Ylm :
!l+m
L̂+ dl+m h i (5.121)
Ylm ∝ Yl−l = (−1)l+m eimφ sinm (ϑ) sin l
(ϑ)F (ϑ) =
ℏ d cos(ϑ)l+m
dl+m
= (−1)l+m eimφ sinm (ϑ) sin2l (ϑ)
d cos(ϑ)l+m

5.3.3 Symmetry Properties of the Eigenfunctions

We now let the parity operator Π̂ act on the eigenstate of the relevant angular momentum
operators L̂2 and L̂z to investigate the corresponding symmetry properties. In general, for a
state in position space, it holds:
Π̂ |r⟩ = − |r⟩
In spherical coordinates, we must consider that the radius remains unchanged under a reflection
(r → r), as it cannot become negative. However, the angles ϑ and φ will change. The following
relationships hold:
ϑ → π − ϑ and φ → π + φ (5.123)
Why (5.123) is valid can be easily seen in Figures 26 and 27. But what effect does Π̂ have on the
explicit position representation of the eigenfunction? Let’s consider the Rodrigues Formula of
the Legendre Differential Equation – our chosen representation of the spherical harmonics Ylm
in (5.120) – and substitute the relations from (5.123).
It follows for the polar angle ϑ:

cos(ϑ) → − cos(ϑ)
(
ϑ→π−ϑ:
sin(ϑ) → sin(ϑ)

137
 5 Angular Momentum

z y
r r
ϑ
φ+π φ
π−ϑ
x x

−r −r

Abb. 26: Parity transformation in spherical coordinates: Only the two-dimensional projection
of the angles ϑ and φ is considered.

ϑ
φ y
ϑ φ

x
−r

Abb. 27: A three-dimensional representation of the parity transformation of a position vector:

While r remains unchanged, only the angles ϑ and φ change.

The azimuthal angle only appears in the phase factor of the spherical harmonics and behaves
under the symmetry transformation in the following manner:

φ → π + φ =⇒ e+imφ → e+imφ e+imπ = (−1)m e+imφ

Using these three relations in (5.120), we obtain (noting that only the relevant terms are explic-
itly written out):

dm+l
Ylm (ϑ, φ) = N e+imφ sinm (ϑ) sin2l (ϑ) →
d cos(ϑ)m+l
dm+l
→ N (−1)m e+imφ sinm (ϑ)(−1)m+l sin2l (ϑ) =
d cos(ϑ)m+l
dm+l
= (−1)l N e+imφ sinm (ϑ) sin2l (ϑ)
d cos(ϑ)m+l

In short form, we can write the effect of the parity operator on the spherical harmonics Ylm (ϑ, φ)
as:
Π̂Ylm (ϑ, φ) = (−1)l Ylm (ϑ, φ) (5.124)
A parity transformation, i.e., a reflection about the origin of the used spherically symmetric
coordinate system, thus depends only on the eigenvalue l and is not influenced by the magnetic
quantum number m.

138
Quantum Theory I

5.3.4 Representations of the Spherical Harmonics

In the following, some spherical harmonics are depicted in two-dimensional polar representation
and in three dimensions. In both cases, Y00 (ϑ, φ) was omitted, since it can simply be represented
by a circle or a sphere.
90 90 90
120 60 Y10 120 60 Y20 120 60 Y30

150 30 150 30 150 30

180 0 180 0 180 0

210 330 210 330 210 330

240 300 240 300 240 300

270 270 270

Abb. 28: Polar representations of Yl0 (ϑ, φ) for l = 1, 2, 3.

90 90 90
120 60 Y11 120 60 Y21 120 60 Y31

150 30 150 30 150 30

180 0 180 0 180 0

210 330 210 330 210 330

240 300 240 300 240 300

270 270 270

Abb. 29: Polar representations of Yl1 (ϑ, φ) for l = 1, 2, 3.

90 90 90
120 60 Y22 120 60 Y32 120 60 Y42

150 30 150 30 150 30

180 0 180 0 180 0

210 330 210 330 210 330

240 300 240 300 240 300

270 270 270

Abb. 30: Polar representations of Yl2 (ϑ, φ) for l = 2, 3, 4.

For example, we can see from this representation that with greater difference ∆lm = l − m,
the number of nodes increases. In three dimensions, the same spherical harmonics are shown in
Figure 31-33.

139
 5 Angular Momentum

z z
z

x y x y
x y

Abb. 31: The spherical harmonics Yl0 (ϑ, φ) for l = 1, 2, 3.

z z

x y
x y x y

Abb. 32: The spherical harmonics Yl1 (ϑ, φ) for l = 1, 2, 3.

z
z
z

x y
x y
x y

Abb. 33: The spherical harmonics Yl2 (ϑ, φ) for l = 2, 3, 4.

140
Quantum Theory I

6 Hydrogen Atom

Motivation: The Hydrogen Atom

The good agreement between experimentally measured and quantum mechanically calcu-
lated spectral lines in the absorption and emission spectrum of the hydrogen atom was one
of the greatest successes of quantum theory since 1925. Besides explaining the series in dif-
ferent wavelength ranges, the solution of the Schrödinger equation for the hydrogen atom
also provides the geometric shape of the electron orbitals. This enabled the development
of the shell model of electrons, and the understanding of the structure of the periodic table
of elements began without further postulates. Additionally, it became possible to explain
chemical bonds, which play a central role in the formation of molecules and solids.

The hydrogen atom also played a central role in the extension of quantum theory to
relativistic quantum mechanics by Paul Dirac and quantum electrodynamics by Hans
Bethe and Edwin Salpeter.

6.1 Schrödinger Equation as a Two-Body Problem

The hydrogen atom consists of a positively charged proton and a negatively charged electron.
The proton forms the nucleus of the atom (the “nucleus”). To describe the entire system
“hydrogen atom” sensibly, we must consider both the electron and the nucleus in our Hamiltonian
operator. Thus, we are dealing with a 6-dimensional two-body problem and a complication of
the Schrödinger equation (2.10):

p̂2e p̂2
Ĥ = + N + V (|re − rN |) (6.1)
2me 2mN

While the kinetic terms in Ĥ can be easily separated, the potential term, which explicitly depends
on re and rN , poses a problem. We will subsequently transform our coordinate system so that
we achieve a decoupling of the 6-dimensional problem into two separable three-dimensional
problems.

6.1.1 Transformation to the Center of Mass System

We are initially in the laboratory system. Electron and nucleus can be described (at least in the
thinking of classical physics) with the position and momentum coordinates re and rN as well as
pe and pN . The goal is to transition into the center of mass system with the relative coordinates
r and p and the center of mass coordinates R and P. This will have the advantage that we
can express the relative distance re − rN of the electron and proton in our new coordinates very
simply.

With the mass of the electron me = 9.109 383 × 10−31 kg and the mass of the nucleus (in other
words, the proton mass) mN = 1.672 621 × 10−27 kg we can trivially express the total mass as:

M = mN + me ≈ mN (6.2)

Since the nucleus is approximately 1000 times heavier than the electron, M is practically domi-
nated only by the proton. For the center of mass system, the reduced mass µ is relevant, which
is a product of the force balance between the electron and the proton:

1 1
−1
me mN

µ= = + ≈ me (6.3)
mN + me me mN

141
 6 Hydrogen Atom

For the center of mass coordinate R, the following expression holds, whereby R lies very close
to the nucleus due to the much greater mass mN :
m e re mN rN mN
R= + ≈ rN (6.4)
mN + me mN + me M
We already know the expression for the relative coordinate r, it corresponds to the difference of
both components:
r = re − rN (6.5)
Thus, after the transformation, the potential term reduces to V (|re − rN |) → V (|r|) and we
achieve a separation of variables. By deriving the position coordinate with respect to time, we
also obtain the velocity and thus the momentum. For the total momentum, the following holds:
dR me ve mN vN
 
P=M =M + = me ve + mN vN = pe + pN (6.6)
dt mN + me mN + me
The relative motion is determined by the relative mass µ and thus results in:
dr mN pe m e pN
p=µ = µ (ve − vN ) = − (6.7)
dt mN + me mN + me
In the center of mass system, the relation for the magnitude of the individual momentum
components holds: |pe | = |pN |. Through the correspondence principle, we can also associate the
relative momentum p and the center of mass momentum P with the corresponding operators p̂
and P̂!

Example: Transformation to the Center of Mass System

With (6.4) and (6.5) as well as (6.6) and (6.7), we have found all relevant expressions for
the transformation. While we can already write the potential term V (|re − rN |) → V (|r|)
in the new coordinates, we need to rewrite the momentum operators for the two kinetic
terms. We will first do this explicitly in the position representation, where p̂i = −iℏ∇i ,
and then in an abstract operator notation.

Let’s start by expressing the Laplace operators ∆e and ∆N in center of mass coordinates.
We use an expression similar to (5.87), applying the chain rule:

∂ ∂ri ∂ ∂Ri ∂ (6.5,6.4) me ∂ ∂

= + = +
∂re,i ∂re,i ∂ri ∂re,i ∂Ri M ∂Ri ∂ri
∂ ∂ri ∂ ∂Ri ∂ (6.5,6.4) mN ∂ ∂
= + = −
∂rN,i ∂rN,i ∂ri ∂rN,i ∂Ri M ∂Ri ∂ri

By squaring the upper two expressions, we obtain the individual components of the Laplace
operators ∆e = ∂x2e + ∂y2e + ∂z2e and ∆N = ∂x2N + ∂y2N + ∂z2N :

∂2 ∂2 m2e ∂ 2 2me ∂ 2
2 = + +
∂re,i ∂ri2 M 2 ∂Ri2 M ∂ri ∂Ri
∂2 ∂2 m2e ∂ 2 2mN ∂ 2
2 = + −
∂rN,i ∂ri2 M 2 ∂Ri2 M ∂ri ∂Ri

By Schwarz’s theorem, we know that the order of derivatives can be exchanged, and thus
the upper two expressions simplify.
If we replace the indices i with {x, y, z} and move to three dimensions according to
Einstein’s summation convention, we obtain expressions dependent on the corresponding

142
Quantum Theory I

Laplace operators ∆r and ∆R :

m2e 2me
∆e = ∆r + ∆R + ∇r ∇R
M2 M (6.8)
m2 2mN
∆N = ∆r + N2 ∆R − ∇r ∇R
M M

If we substitute these into our Hamiltonian operator Ĥ, which is now independent of the
relative potential, we obtain the following simple expression in the position representation:

ℏ2 ℏ2 (6.8)
Ĥ(V = 0) = − ∆e − ∆N =
2me 2mN
ℏ2 1 me 2 1 mN 2
 
=− ∆r + 2 ∆R + ∇ r ∇ R + ∆r + 2 ∆R − ∇r ∇R =
2 me M M mN M M
2 1 1 me + mN
  
ℏ (6.2,6.3)
=− + ∆r + ∆R =
2 me mN M2
ℏ2 ℏ2
= − ∆r − ∆R (6.9)
2µ 2M

This corresponds in operator notation to the following relation with p̂ and P̂:

p̂2 P̂2
Ĥ(V = 0) = + (6.10)
2µ 2M

With (6.7) and (6.6), we can derive (6.10) directly through the respective operators. We
first rewrite p̂e and p̂N so that they depend on p̂ and P̂:
me
(me + mN ) p̂e = me P̂ + M p̂ =⇒ p̂e = P̂ + p̂
M
mN
(me + mN ) p̂N = mN P̂ − M p̂ =⇒ p̂N = P̂ − p̂
M
Squaring the above expressions, we obtain the kinetic terms of the Hamiltonian (6.1).
The center of mass momentum operator P̂ and the relative momentum operator p̂ also
commute since they act in distinct spaces ({|R⟩} and {|r⟩}).

m2e 2 2me
p̂2e = p̂2 + P̂ + P̂p̂
M2 M (6.11)
m2 2mN
p̂2N = p̂2 + N2 P̂2 − P̂p̂
M M

(6.11) corresponds to (6.8), and the calculation of Ĥ is analogous to (6.9).

We obtain a new Hamiltonian operator, with individual terms dependent on a single (center of
mass) coordinate:
P̂2 p̂2
Ĥ = + + V (|r|) (6.12)
2M 2µ
Without initially going into the form of the potential V (|r|), we can provide the stationary
Schrödinger equation as:
!
P̂2 p̂2
+ + V (|r|) ψ(r, R) = Eψ(r, R) (6.13)
2M 2µ
E represents the eigenenergies of the entire system. (6.13) is still a differential equation in 6
dimensions. At first glance, we seemingly have not improved the problem in its complexity

143
 6 Hydrogen Atom

compared to the laboratory system. However, upon closer inspection: It is now possible to
separate the variables!

Separation of Variables The potential term V (|r|) and one of the two kinetic terms now depend
only on the relative coordinates, while the second kinetic term is determined by the center of
mass component R̂. Through a separation Ansatz, we can separate relative and center of mass
coordinates: ψ(r, R) = ϕ(r)Ψ(R). If we substitute this into (6.13), we can write:
!
P̂2 p̂2
Eϕ(r)Ψ(R) = + + V (|r|) ϕ(r)Ψ(R) =
2M 2µ
1 2 1 2
= P̂ [ϕ(r)Ψ(R)] + p̂ [ϕ(r)Ψ(R)] + V (|r|)ϕ(r)Ψ(R)
2M 2µ

However, since P̂2 only acts on Ψ(R), and p̂2 only acts on Ψ(r), in the first term ϕ(r) and in
the second term Ψ(R) can be pulled before the operator:

ϕ(r) 2 Ψ(R) 2
P̂ Ψ(R) + p̂ ϕ(r) + V (|r|)ϕ(r)Ψ(R) = Eϕ(r)Ψ(R) =⇒
2M 2µ
Ψ(R) 2 ϕ(r) 2
p̂ ϕ(r) + V (|r|)ϕ(r)Ψ(R) = Eϕ(r)Ψ(R) − P̂ Ψ(R) | ÷ Ψ(R)ϕ(r)
2µ 2M
1 1
! !
p̂2 P̂2
ϕ(r) + V (|r|)ϕ(r) = EΨ(R) − Ψ(R) =⇒
ϕ(r) 2µ Ψ(R) 2M
1 1
! !
p̂2 P̂2
+ V (|r|) ϕ(r) = E− Ψ(R) (6.14)
ϕ(r) 2µ Ψ(R) 2M

Both sides of the differential equation (6.14) now depend on only one variable each and thus
must each result in a constant ε, so that the equation can be satisfied. We first consider the
right side of differential equation (6.14), and replace (due to P̂ = −iℏ∇R ) the operator P̂2 by
−ℏ2 ∆R :

1
!
ℏ2
E+ ∆R Ψ(R) = ε | · Ψ(R)
Ψ(R) 2M
!
ℏ2
E+ ∆R Ψ(R) = εΨ(R) (6.15)
2M

The relation (6.15) describes the motion of the center of mass of the atom. Since the equation
does not depend on the potential, it can be easily solved by a plane wave:

Ψ(R) = eikR (6.16)

We substitute this solution into (6.15) and consider that ∆R Ψ(R) = ∆R eikR = k2 eikR =
k2 Ψ(R). We then obtain the following expression for the total energy E:
!
ℏ2 k2 ℏ2 k2
E− Ψ(R) = εΨ(R) =⇒ E = ε +
2M 2M

The total energy E thus consists (as expected) of the kinetic energy of the center of mass
ℏ2 k2 /2M and ε. The latter corresponds to the relative kinetic energy between the electron and
the nucleus – i.e., the interaction energy. We have thus reduced the 6-dimensional problem to a
three-dimensional one by separating the center of mass coordinates.

144
Quantum Theory I

6.1.2 Separation Ansatz in Spherical Coordinates

We now turn to the actually interesting wave function ϕ(r) and the associated eigenenergy ε.
For this, we now consider the left part of equation (6.14), which must also correspond to the
constant energy ε due to the separation Ansatz:

1
!
p̂2
+ V (|r|) ϕ(r) = ε | · ϕ(r)
ϕ(r) 2µ
!
p̂2
+ V (|r|) ϕ(r) = εϕ(r) (6.17)
2µ

We now solve the Schrödinger equation from (6.17) in the position space of the relative coordi-
nates, and thus replace (due to p̂ = −iℏ∇r ) the operator p̂2 by −ℏ2 ∆r :
!
−ℏ2
∆r + V (|r|) ϕ(r) = εϕ(r) (6.18)
2µ

The potential depends only on the magnitude of the coordinate |r| ≡ r, which is why a trans-
formation into spherical coordinates is suitable. In this case, the angular dependent parts are
independent of the potential and can be solved separately. For the transformation, however, the
Laplace operator ∆r must be transformed into spherical coordinates.

The expression for ∆r in spherical coordinates results in a more or less complex expression, which
we can simplify using our knowledge about the spatial representation of angular momentum:

1 ∂ 1
" #
∂ ∂ ∂ ∂2
   
(5.109)
∆r = 2 r2 − 2 2 sin(ϑ) sin(ϑ) + =
r ∂r ∂r r sin (ϑ) ∂ϑ ∂ϑ ∂φ2
1 ∂ ∂ 1
 
= 2 r2 − 2 2 L̂2 (6.19)
r ∂r ∂r ℏ r
The first term of the operator acts only on the radial component r, while the second term, as
shown in (5.109), involves only the angular components ϑ and φ through L̂2 . Incidentally, this
separation can also be motivated through the square of the classical angular momentum L:

L2 = (r × p)2 = (rp sin(α))2 = r2 p2 (1 − cos2 (α)) = r2 p2 − (rp cos(α))2 = r2 p2 − (r · p)2

This can be transformed into r2 p2 = (r · p)2 + L2 and finally divided by the radius r:

(r · p)2 1
p2 = + 2 L2 (6.20)
r2 r
Thus, we have obtained an expression corresponding to the kinetic term p̂2 /2µ in (6.18). While
the first term of the momentum square projects the momentum vector p in the direction of
r – thus considering only radial components – the second term corresponds to the angular
components due to L2 = L2 (ϑ, φ). Without specifying the potential in more detail, we can
rewrite the Schrödinger equation in spherical coordinates by inserting (6.19) into (6.18) as
follows: " #
−ℏ2 ∂ ∂ L̂2
 
r2 + + V (r) ϕ(r) = εϕ(r) (6.21)
2µr2 ∂r ∂r 2µr2
Again, we can perform a separation ansatz to separate the radial wave functions from the angle-
dependent part:
ϕ(r) = R(r)Ylm (ϑ, φ) (6.22)
The angular component is completely described by the magnitude operator of angular momen-
tum L̂2 ; we already know its eigenfunction in the spatial representation, namely the spherical

145
 6 Hydrogen Atom

harmonics Ylm (ϑ, φ). The effect of L̂2 on the Ylm (ϑ, φ) is also known from (5.114) and we can
thus write in a simplifying manner:

ℏ2 l(l + 1)
" #
−ℏ2 ∂ ∂
 
r2 + + V (r) R(r)Ylm (ϑ, φ) = εR(r)Ylm (ϑ, φ) (6.23)
2µr ∂r
2 ∂r 2µr2

The spherical harmonics can be canceled out on both sides, and we thereby obtain the differential
equation for the radial part of the spherical Schrödinger equation. Only here does it become
necessary to consider the specific form of the potential. Due to the electrostatic attractive forces
between the electron and proton, V (r) is the Coulomb potential (note the prefactors, depending
on the units used a 1/4πε0 might also appear here):

Ze2
V (r) = − (6.24)
r
Z is the charge number of the nucleus in units of the elementary charge e = 1.602 176 × 10−19 C.
In the case of the hydrogen atom, Z = 1. We maintain a general Z to be able to describe
hydrogen-like atoms (that is, atoms and ions with only one electron).

6.1.3 Effective Potential and Form of the Wave Function

Inserting the Coulomb potential (6.24) into (6.23) gives us the following equation for the radial
part R(r) of the wave function:

ℏ2 l(l + 1) Ze2
" #
−ℏ2 ∂ ∂
 
r2 + − R(r) = εR(r) (6.25)
2µr ∂r
2 ∂r 2µr2 r

In (6.25) we call the sum of the angular momentum term and the Coulomb potential the effective
potential; it corresponds to the following form:

ℏ2 l(l + 1) Ze2
Veff (r) = − (6.26)
2µr 2 r}
| {z
| {z }
Centrifugal- Coulomb-
potential potential

Both terms lie approximately in the same order of magnitude. At small distances, the repulsive
centrifugal potential associated with the angular momentum dominates, at large distances the
attractive Coulomb potential dominates. This leads in cases where l > 0 to electrons hardly
being found near the nucleus, since the energy induced by angular momentum increases more
strongly at very small distances than the electrostatic attraction. This fact is clearly visible in
Figure 34: The larger the orbital angular momentum quantum number l becomes, the greater
the influence of angular momentum and the farther the electron will be found from the nucleus.

Before we solve the Schrödinger equation for the central potential, we can make assertions about
the properties of the wave function based on the form of the potential. In the case of l = 0, the
potential term V (r) is consistently negative and thus allows for bound states; if ε < 0, a bound
state occurs while an electron with ε > 0 is scattered by the purely electrostatic potential.
Assuming l > 0, it is classically forbidden for an electron wave function to penetrate into the
positively diverging region of the effective potential – however, due to the tunneling properties
of wave functions, we will also find electrons with l > 0 at small distances r. Generally, in this
chapter, we will only deal with electrons with negative eigenenergy ε < 0, which exhibit a bound
state for all l.
Additionally, V (r) is rotationally symmetric. This means that the central potential always has
the same value regardless of the angles ϑ and φ. Thus, [Π̂, V (r)] = 0 and consequently [Π̂, Ĥ] = 0.

146
Quantum Theory I

Parity and Hamiltonian operators are therefore compatible and share the same eigenfunctions.
The wave function ϕ(r) will thus remain invariant (except for a sign) under point reflections.
For the spherical harmonics Ylm (ϑ, φ), the behavior under that symmetry transformation has
already been shown in (5.123): Applying Π̂ results in an additional (−1)l term.

VL2 (r; l = 3)
VL2 (r; l = 1)
Veff (r)

Veff (r)
Veff (r)
Veff (r)
VC

VC

r r

Abb. 34: Effective potential for the cases l = 1 and l = 3: Dashed lines indicate the pure
contributions of the Coulomb potential VC and the angular momentum term VL2 .

For bound states, the node rule applies again: We will find a quantum number corresponding to
the number of “nodes” of the wave function, thus allowing the placement of ϕ(r) in an energetic
hierarchy.

6.2 Solution for Radial Wave Functions

Let us now discuss how the differential equation for the radial wave function R(r) can be solved.
To do this, we first bring the right-hand side of equation (6.25) to the left. Considering that we
assume ε < 0 to allow for bound states (we use ε = −|ε|), we can write this as:
ℏ2 l(l + 1) Ze2
" #
ℏ2 ∂ ∂
 
− r2 + − + |ε| R(r) = 0 (6.27)
2µr2 ∂r ∂r 2µr2 r
The first derivative term with respect to the radius r can, following the product rule, be rewritten
as:
1 ∂ 1 2
! !

∂

∂ ∂ 2 ∂ ∂ 2
r2 = 2 2r + r2 2 = +
r2 ∂r ∂r r ∂r ∂r r ∂r ∂r2
Inserted into (6.27), we can further simplify the radial wave equation using transformations and
substitutions from (6.29) and (6.30):
2 ∂ ℏ2 l(l + 1) Ze2
" ! #
ℏ2 ∂2 −2µ
0= − + 2 + − + |ε| R(r) = ·
2µ r ∂r ∂ r 2µr2 r ℏ2
2 ∂ l(l + 1) 2Z µe2 2µ|ε|
!
∂2 (6.29,6.30)
= + 2 − 2
+ 2
− 2 R(r) =
r ∂r ∂ r r r ℏ ℏ
2 ∂ l(l + 1) 2Z
!
∂2
= + 2 − + − k 2 R(r) (6.28)
r ∂r ∂ r r2 a0 r
The new constant a0 is referred to (in Planck units) as the Bohr radius. It represents the radius
of the hydrogen atom in the lowest energy state. However, we should remind ourselves that

147
 6 Hydrogen Atom

we are using the reduced mass µ and a0 is actually defined using the electron mass me . The
difference is only 0.1‰. The Bohr radius a0 is defined as:

me e2
a0 = ≈ 0.529 Å (6.29)
ℏ2
The energy ε and the wave number k are related as follows:

ℏ2 k 2 2µ|ε|
|ε| = ⇐⇒ k 2 = (6.30)
2µ ℏ2

First Modification of the Radial Wave Function Without knowing the explicit form of the
radial wave functions, we know that it must be normalized over the range r ∈ [0, ∞). When
calculating the normalization integral, we must remember that we are working in spherical
coordinates and must therefore include the Jacobian determinant when integrating. It follows
that:
Z ∞ Z ∞ Z ∞
1= dr r |R(r)| = dr |rR(r)| = dr |u(r)|2 with u(r) = rR(r) (6.31)
2 2 2 def

0 0 0

By including the Jacobian determinant and defining a new function u(r) = rR(r), we can
interpret the radial wave function R(r) as a sort of spherical wave:

u(r)
R(r) = (6.32)
r
As a side calculation, we now need to explicitly evaluate the first and second derivatives of
R(r) ≡ u/r:

∂ ∂ u u′ r − u u′ u
R(r) = = 2
= − 2 (6.33)
∂r ∂r r r r r
∂2 ∂2 u ∂ u′ r − u (u′ r − u)′ r2 − (u′ r − u)2r
 
R(r) = = =
∂r2 ∂r2 r ∂r r2 r4
(u r + u − u )r − 2r u − 2ru
′′ ′ ′ 2 2 ′ u ′′ u′ u
= 4
= − 2 2
+2 3 (6.34)
r r r r
By substituting (6.32), (6.33), and (6.34) into the radial wave equation (6.28), we obtain:

2 u l(l + 1) 2Z
  ′
u u′′ u′ u u
   
0= − 2 + −2 2 − 2 3 + − 2
+ − k2 =
r r r r r r r a0 r r
1 ′′ 1 l(l + 1) 2Z 1 ∂2 l(l + 1) 2Z
  !
= u + − + −k u=
2
− + −k u= |·r
2
r r r2 a0 r r ∂r2 r2 a0 r
l(l + 1) 2Z
!
∂2
= 2
− 2
+ − k 2 u(r) (6.35)
∂r r a0 r

This equation has the form of a single-particle Schrödinger equation with the effective potential
as the potential term. However, in this equation, the derivative is not taken with respect to
Cartesian coordinates but with respect to the radius. To obtain physical solutions, we introduce
the following two Dirichlet boundary conditions for u(r):

u(r → ∞) −→ 0 and u(r → 0) −→ 0 (6.36)

The first boundary condition is necessary for the radial wave function to be normalizable. The
second boundary condition must be fulfilled because otherwise the radial wave function R(r) =
u(r)/r would diverge at r = 0.

148
Quantum Theory I

Second Modification of the Radial Wave Function We have now transformed the initial
differential equation into the following form:

l(l + 1) 2Z
!
∂2
2
− 2
+ − k 2 u(r) = 0
∂r r a0 r

We can find another ansatz for u(r) that modifies the differential equation and brings us closer
to a solution:
u(r) = e−kr y(r) (6.37)
The form of y(r) ≡ y will be discussed in the following sections. For now, we substitute the
ansatz (6.37) into the above differential equation:

l(l + 1) 2Z
!
∂2
0= − + − k 2 ye−kr =
∂r2 r2 a0 r
∂ 2  −kr  l(l + 1) −kr 2Z −kr
= ye − ye + ye − k 2 ye−kr =
∂r2 r2 a0 r
∂  ′ −kr  l(l + 1) 2Z −kr
= ye − kye−kr − 2
ye−kr + ye − k 2 ye−kr =
∂r r a0 r
l(l + 1) −kr 2Z −kr
= y ′′ e−kr − ky ′ e−kr − ky ′ e−kr + k 2 ye−kr − ye + ye − k 2 ye−kr =
r2 a0 r
l(l + 1) 2Z
   
= y ′′ − 2ky ′ − − y e−kr
r2 a0 r

We can now cancel out the exponential function e−kr ; similarly, the k 2 term has vanished and
the wave number k now appears as an exponential damping term in (6.37). This leads to the
following differential equation:

1 l(l + 1) 2Z
" #
∂2 ∂

2
− 2k − − y(r) = 0 (6.38)
∂r ∂r r r a0

To determine y(r), two different methods can be chosen. We will first solve the above differential
equation using the so-called Frobenius method by inserting a power series ansatz. Later, we will
also introduce the Laguerre differential equation, for which ready-made solution formulas exist.

6.2.1 Solution Using Frobenius Method

As an ansatz solution, starting with the Frobenius method, we will choose a power series at this
point. We can represent it as:
∞ ∞
y(r) = rα+1 ai r i = (6.39)
X X
ai rα+1+i
i=0 i=0

In (6.38), we need the first and second derivatives of our ansatz:

∞ ∞ ∞
!
∂y ∂ ∂ α+1+i X
= = = ai (α + i + 1)rα+i (6.40)
X X
α+1+i
ai r ai r
∂r ∂r i=0 i=0
∂r i=0
∞ ∞
!
∂2y ∂ ∂y ∂
 
(6.40)
= = ai (α + i + 1)r = ai (α + i + 1)(α + i)rα−1+i (6.41)
X X
α+i
∂r 2 ∂r ∂r ∂r i=0 i=0

149
 6 Hydrogen Atom

If we substitute the ansatz (6.39) and the derived derivative expressions (6.41) and (6.40) into
(6.38), we obtain:
∞
1 l(l + 1) 2Z
" #
∂2 ∂

(6.40)(6.41)
0= − 2k =
X
− − rα+1 ai r i
∂r2 ∂r r r a0 i=0
∞
1 l(l + 1) 2Z
   
= ai (α + i + 1)(α + i)r − 2k(α + i + 1)r =
X
α−1+i α+i α+1+i
− − r
i=0
r r a0
∞
2Z
   
−1
= ai (α + i + 1)(α + i)r − 2k(α + i + 1)r − l(l + 1)r =
X
α−1+i α+i α+i
− r
i=0
a0
∞
2Z α+i
 
= ai (α + i + 1)(α + i)rα−1+i − 2k(α + i + 1)rα+i − l(l + 1)rα−1+i + =
X
r
i=0
a0
∞
2Z
h i   
= (α + i + 1)(α + i) − l(l + 1) r + − 2k(α + i + 1) rα+i (6.42)
X
α−1+i
ai
i=0
a0

We find that r appears with the powers rα−1+i and rα+i . If we explicitly write out the sum,
each individual r power term must vanish to satisfy the differential equation. This allows us to
determine α. Let’s first consider the term with the smallest power of r: This occurs at i = 0.
In this case, there is a term with rα−1 and one with rα . Obviously, rα−1 is the smallest power
term that can occur. It must also be zero, which means:

0 = a0 [(α + 1)α − l(l + 1)] rα−1 = | ÷ a0 rα−1

= α(α + 1) − l(l + 1) = α2 + α − l(l + 1)

We can now solve a simple quadratic equation for α, obtaining two solutions:

1 1 1 1q −1 ± (2l + 1)
r
α1,2 =− ± + l(l + 1) = − ± (2l + 1)2 =
2 4 2 2 2
Only one of these solutions will have actual physical relevance. We find:

α1 = l
(
α =⇒ (6.43)
α2 = −(l + 1) (unphysical)

How do we know that only the solution α1 is physical? For this, we must first link the original
radial wave function R(r) over u(r) and y(r) with our ansatz (6.39):
∞ ∞
(6.32) u(r) (6.37) 1 −kr (6.39) 1 −kr α+1 X
R(r) = = y(r) = ai ri = e−kr (6.44)
X
e e r ai rα+i
r r r i=0 i=0

Let’s try inserting the two possible solutions for α from (6.43) into (6.44) and check if we can
find solutions over the entire existence range of the radius r. We distinguish:

α = l: Replacing α with the orbital angular momentum quantum number α1 = l, we

obtain:
∞ ∞ ∞
R(r) = e−kr ai rl+i =⇒ u(r) = re−kr ai rl+i = e−kr
X X X
ai rl+i+1
i=0 i=0 i=0

For all sum terms i ≥ 0, it holds that R(r) and u(r) remain normalizable since there can
be no singularity in the allowed interval of r. Especially r → 0 poses no problems here,
not even for i = 0 and l = 0.

150
Quantum Theory I

α = −(l + 1): When we now insert our found result α2 = −(l + 1) for α, it follows:
∞ ∞ ∞
R(r) = e−kr ai ri−l−1 =⇒ u(r) = re−kr ai ri−l−1 = e−kr
X X X
ai ri−l
i=0 i=0 i=0

We recognize that, in the case of r → 0 and i = 0, l ≥ 1, a singularity arises in u(r), and

u(r) would not be normalizable in this form. Considering the special case l = 0, although
there is no problem with normalizing u(r), the actual radial wave function R(r) would
have a singularity at the sum term i = 0:
∞
1
R(r → 0) = e−kr
X
ai ri−1 −→
i=0
r

At the transition to the limit, the r−1 term dominates! R(r) is then inserted into the
Schrödinger equation (6.23), resulting in a ∆r−1 term that needs further discussion. For
this, it holds:
1
∆ = −4πδ(r)
r
Thus, we obtain a δ function in the Schrödinger equation, preventing a comprehensive
solution of the differential equation. Therefore, the special case l = 0 is not a possible
solution for the radial wave functions, and α2 = −(l + 1) must be excluded overall.

Example: ∆r−1
In the previous argument, we used the following relation (6.45), but we now want to derive
in more detail how this expression arises. Initially, it holds:
1
∆ = −4πδ(r) (6.45)
r
r represents the radius in a spherical coordinate system; we must therefore also adjust our
Laplace operator ∆ according to (6.19) to the chosen coordinates. When we insert ∆ in
spherical coordinates into (6.45) for the case of r ̸= 0, we obtain:

1 1 ∂ 1 1 ∂
!
∂ r2
 
∆ = 2 r2 =− 2 =0
r r ∂r ∂r r r ∂r r2

If we now perform the integration over an arbitrary sphere with radius r0 , using Gauss’s
theorem, or the Nabla operator ∇ in polar coordinates, we can derive the following ex-
pression:

1 1 1 ∂ 1
Z Z Z Z Z
dV ∆ = dV ∇∇ = dS r∇ = r02 dΩ rr =− dΩ = −4π
V r V r ∂V r ∂V ∂r r r=r0 ∂V

As we previously recognized that in the case of r ̸= 0, the relation ∆r−1 = 0 must hold, we
only obtain this result of the integration if r = 0. With the delta function, we can enforce
the corresponding behavior, and it follows:
1
∆ = −4πδ(r) □
r

Only α = l is therefore physically meaningful as a solution. We further require that the Dirichlet
boundary condition, i.e., u(r → 0) −→ 0, must be fulfilled. We now use the found relation α = l

151
 6 Hydrogen Atom

in our sum equation from (6.42) and further simplify it to obtain a recursive relationship between
the development coefficients ai :
∞
2Z
h i   
0= (l + i + 1)(l + i) − l(l + 1) r + − 2k(l + i + 1) rl+i =
X
l−1+i
ai
i=0
a0
∞
2Z
h i   
= l + li + l + li + i + i + i − i − l(l + 1) r + − 2k(l + i + 1) rl+i =
X
2 2 l−1+i
ai
i=0
a0
∞
2Z
h i   
= l2 + l + 2li + 2i + i2 − i − l(l + 1) rl−1+i + − 2k(l + i + 1) rl+i =
X
ai
i=0
a0
∞ 
2Z
h i   
= ai l(l + 1) + 2i(l + 1) + i(i − 1) −l(l + 1) r + ai − 2k(l + i + 1) rl+i =
X
l−1+i

i=0
| {z } a0
vanishes at i=0
∞ ∞
Z l+i
h i  
= ai 2i(l + 1) + i(i − 1) rl−1+i − 2 ai k(l + i + 1) − r = |i = n + 1
X X

i=1 i=0
a0
∞ ∞
Z l+i
h i  
= 2(n + 1)(l + 1) + (n + 1)n r 2ai k(l + i + 1) − r = |n → i
X X
l+n
an+1 −
n=0 i=0
a0
∞ 
Z
h i  
= 2(l + 1)(i + 1) + i(i + 1) − 2ai k(l + i + 1) − (6.46)
X
ai+1 rl+i
i=0
a0

We have managed to manipulate both sum terms in such a way that instead of two different
powers of r, we now have two different development coefficients ai+1 and ai . A coefficient
comparison is now very easy: To fulfill the equation, the entire expression in the curly brackets
must vanish for each power rl+i :
Z
 
ai+1 [2(l + 1)(i + 1) + i(i + 1)] − 2ai k(l + i + 1) − = 0 =⇒
a0
Z
 
ai+1 [2(l + 1)(i + 1) + i(i + 1)] = 2ai k(l + i + 1) −
a0
From this, we can derive the following recursion relation between ai+1 and ai :

2[k(l + i + 1) − Z/a0 ]
ai+1 = ai (6.47)
2(l + 1)(i + 1) + i(i + 1)

This corresponds to a two-term recursion formula for the Laguerre functions. If we know the
starting value a0 , we can write down all radial wave functions using (6.47) and a subsequent
normalization.

Recursion Formula We must check again whether it will diverge at any point. For this, we
calculate the ratio between two successive development coefficients by dividing the recursion
equation (6.47) by ai :

ai+1 2[k(l + i + 1) − Z/a0 ] 2[ki + k(l + 1) − Z/a0 ]

= = =
ai 2(l + 1)(i + 1) + i(i + 1) 2(l + 1)(i + 1) + i(i + 1)
2ki 2[k(l + 1) − Z/a0 ]
= 2 +
i + i + 2(l + 1)(i + 1) i2 + i + 2(l + 1)(i + 1)

Letting i go to infinity, we can immediately neglect the second term and also recognize that in
the denominator, the quadratic term will dominate. Therefore:
ai+1 i→∞ 2ki 2k
−−−→ 2 = (6.48)
ai i i

152
Quantum Theory I

Similar to the harmonic oscillator, we now try to determine which function corresponds to this
asymptotic behavior (6.48). We find:
∞ ∞
(2kr)i
e2kr = (6.49)
X X
≡ bi ri
i=0
i! i=0

Again, we form the ratio between two successive development coefficients based on this Taylor
expansion:
bi+1 (2k)i+1 i! 2k i→∞ 2k
= = −−−→
bi (i + 1)! (2k) i i+1 i
We have found a function that exhibits the same behavior as the development coefficients ai of
the radial wave function (6.44) in the limit i → ∞. The i-th term of the series (6.44) reads:
R(i) (r) = e−kr ai rl+i
Setting i to infinity, we can, as just shown, replace ai with e2kr :
∞ ∞ ∞ ∞
i→∞ 2k X (6.49) X −kr 2kr l+i
R(r) = e−kr e−kr ai rl+i = e e r =
X X
ai rl+i −−−→ ekr rl+i
i=0
i i=0 i=0 i=0

We now have the problem that with our exponential ansatz for the development coefficients, the
Dirichlet boundary condition u(r → ∞) = R(r → ∞) −→ 0 cannot be satisfied, as the exponen-
tial function ekr diverges. To avoid divergence in the limit case, the recursion relationship must
be truncated at some point. We therefore define a cut-off condition i = nr for the recursion
relation from (6.47), where we refer to nr as the radial quantum number. For all i > nr , all ai
must vanish. For this condition to be met:
! (6.47) 2[k(l + nr + 1) − Z/a0 ] Z
0 = anr +1 = anr =⇒ k (l + nr + 1) − = 0 (6.50)
2(l + 1)(nr + 1) + nr (nr + 1) | {z } a0
n

Since nr indicates the maximum polynomial degree of the power series (6.44), the radial quantum
number naturally also determines the number of zeros or the number of nodes of the radial wave
function. We can now, as shown in (6.50), define the energy or principal quantum number n:
n = l + nr + 1 with n ≥ 1 (6.51)
Unlike the energy quantum number in the harmonic oscillator (which is also represented by n),
in the case of the hydrogen atom, n cannot be zero. The orbital angular momentum quantum
number l can be written according to (6.51) as:
l = n − nr − 1 with 0 ≤ l ≤ n − 1 (6.52)
We already know that l ≥ 0 must hold from the theory of angular momentum. l = n − 1 applies
in the special case that nr = 0. By simple transformation of (6.50), we can also express the
wave number k ≡ kn as a discrete function of the principal quantum number n:
Z
kn = (6.53)
a0 n
Substituting this into (6.30), we can assign an energy εn to each principal quantum number n:
ℏ2 kn2 ℏ2 Z 2
εn = − =− (6.54)
2µ 2µ a20 n2
The principal quantum number n thus represents different (discrete) energy levels of the hy-
drogen atom! We can summarize the constants in (6.54) in the so-called Rydberg constant Ry
and obtain the following compact expression for the eigenenergies, which increases quadratically
with the principal quantum number n:

153
 6 Hydrogen Atom

Z2 ℏ2
εn = −Ry with Ry = (6.55)
n2 2µa20

The energy εn is now written without absolute value, reminding us that as a bound state,
negative eigenenergies must have. Through the inverse proportionality to 1/n2 , we approach
with increasing energy quantum number εn −→ 0; the energy necessary to lift an electron from
a level n to εn = 0 is called ionization energy. Reaching this energy, the electron can „escape“
from the atom, leaving a proton behind in the case of hydrogen: The hydrogen atom H is ionized
to H+.
Applying less energy results in a spectral transition, formally described earlier in (1.44). By
adding energy, we lift the electron from a lower level to a higher one; the electron can also fall
from a higher-energy level to a lower one. For n > m, we convert the following energy in the
form of a photon: !
Z2 Z2
ℏωnm = εn − εm = Ry − 2
m2 n
The radial eigenfunctions can be expressed by using the expression (6.44), remembering that
α = l, and letting the series run only up to i = nr . We then obtain:

nr
Rnl r (r) = e−kr (6.56)
X
ai rl+i
i=0

The development coefficients must be calculated successively using the recursion relation from
(6.47). In the following section, we will deal with a solution strategy that leads directly to a
wave function for nr and l.

6.2.2 Solution via Laguerre Differential Equation

We have so far treated the differential equation of the radial wave function using the Frobenius
method with a power series ansatz and successfully determined the eigenenergies. At this point,
we want to abandon the power series ansatz and instead switch to the following ansatz with a
general function q(r) ≡ q:
y(r) = rα+1 q(r) (6.57)
Before we substitute this ansatz into the initial equation (6.38), we calculate the first and second
derivatives of our ansatz from (6.57) as a side calculation:

d dq (α + 1) dq
   
rα+1 q = (α + 1)rα q + rα+1 = rα+1 q+
dr dr r dr
d2 d dq
   
rα+1 q = (α + 1)rα q + rα+1 =
dr2 dr dr
(6.58)
dq dq d2 q
= α(α + 1)rα−1 q + (α + 1)rα + (α + 1)rα + rα+1 2 =
dr dr dr
α(α + 1) 2(α + 1) dq d q
" #
2
= rα+1 q+ +
r2 r dr dr2

154
Quantum Theory I

Inserting now (6.57) and the two expressions from (6.58) into the initial differential equation
(6.38), we obtain:

1 l(l + 1) 2Z
" #
d2 d

0= − 2k − − rα+1 q(r) =
dr2 dr r r a0
d2  α+1  d  α+1  l(l + 1) 2Z (6.58)
= r q − 2k r q − rα+1 q + rα+1 q =
dr2 dr r2 a0 r
α+1 α(α + 1) 2(α + 1) dq d2 q (α + 1) dq l(l + 1) 2Z
" #
=r q+ + − 2k q − 2k − q+ q =
r2 r dr dr2 r dr r2 a0 r
α(α + 1) − l(l + 1) 2(α + 1) dq
dr − 2k(α + 1)q 2Z
" #
d2 q dq
=r α+1
2
q + + 2
− 2k + q =
r r dr dr a0 r
dq
α(α + 1) − l(l + 1) 2Z
" #
dr − kq d2 q dq
=r α+1
q + 2(α + 1) + − 2k + q (6.59)
r2 r dr2 dr a0 r

At this point, we perform the following substitution:

y dr 1 dy
r= =⇒ = =⇒ =k (6.60)
k dy k dr

With this substitution, the function q ≡ q(r) becomes a function q ≡ q(y), and we must
appropriately consider the following transformation:
dq dq dy (6.60) dq
= = k (6.61)
dr dy dr dy

We again substitute (6.60) and (6.61) into (6.59):

 
dq
α(α + 1) − l(l + 1) − kq dq 2Zk 
 α+1 k dy 2
y 2d q
0= q + 2(α + 1) + k − 2k 2 + q =
(y/k)

k 2 y/k dy 2 dy a0 y

α(α + 1) − l(l + 1) 2 2Z
 α+1 " #
y k2 dq d2 q dq
 
= k q + 2(α + 1) − q + k 2 2 − 2k 2 + q =
k y2 y dy dy dy a0 k
α(α + 1) − l(l + 1) 2Z
 α+1 " #
k2 y dq d2 q dq
 
= q + 2(α + 1) − q + y 2 − 2y + q =
y k y dy dy dy a0 k
α(α + 1) − l(l + 1)
(  )
yα d2 q dq Z

= y 2 + [2(α + 1) − 2y] + − 2(α + 1) + 2 q (6.62)
k α−1 dy dy y a0 k

Correspondingly, we can, through a further simple transformation – corresponding to the last

term in (6.62) – from (6.53) write:
Z
n= (6.63)
a0 k
We need to substitute again to achieve a simplification. We define:
x dy 1 dx
y= =⇒ = =⇒ =2 (6.64)
2 dx 2 dy

With this substitution, the function q ≡ q(y) becomes a function q ≡ q(x), and we must
appropriately consider the following transformation:
dq dq dx (6.64) dq
= = 2 (6.65)
dy dx dy dx

155
 6 Hydrogen Atom

We now insert (6.63), (6.64), and (6.65) into (6.62):

(x/2)α α(α + 1) − l(l + 1)

(  )
x d2 q dq

0 = α−1 4 2 + [2(α + 1) − x] 2 + − 2(α + 1) + 2n q =
k 2 dx dx x/2
α(α + 1) − l(l + 1)
(  )
xα d2 q dq

= α α−1 2x 2 + 2 [2(α + 1) − x] +2 − (α + 1) + n q =
2 k dx dx x
α(α + 1) − l(l + 1)
(  )
xα d2 q dq

= x 2 + [2(α + 1) − x] + − (α + 1) + n q
(2k)α−1 dx dx x

Since the entire expression should vanish, we can cancel the term before the curly brackets:

α(α + 1) − l(l + 1)
( )
d2 d

x 2 + [2(α + 1) − x] + − (α + 1) + n q(x) = 0
dx dx x

Additionally, we know from (6.43) that α = l. Therefore, another term cancels out:

l(l + 1) − l(l + 1)
( )
d2 d

0= x 2 + [2(l + 1) − x] + − (l + 1) + n q(x) =
dx dx x
( )
d2 d
= x 2 + [2l + 1} +1 − x] + (n − l − 1) q(x) =
dx | {z dx | {z }
β w
′′ ′
= xq + (β + 1 − x)q + wq with β = 2l + 1 and w = n − l − 1 (6.66)

With the substitutions β = 2l + 1 and w = n − l − 1, we have brought this equation into

a form corresponding to the Laguerre differential equation from (10.28). In Chapter 10.4 of
the appendix, it is derived that the Laguerre polynomials Lβw (x) represent the solutions of
this differential equation, which can be expressed, for example, through the Rodrigues formula
(10.31):
1 −β x dw (xw+β e−x )
Lβw (x) = x e (6.67)
w! dxw
The solutions of the differential equation (6.66) depending on the quantum numbers n and l can
be written as:

q(x) = Lβw (x) with β = 2l + 1 and w = n − l − 1 (6.68)

To determine the radial wave function R(r) ≡ Rnl (r), the substitutions (6.32), (6.37), (6.57),
and (6.43) must first be reversed before we can use the solution (6.68):
1
(6.32) (6.37) 1 −kr (6.57) 1 −kr α+1 (6.43) (6.68)
Rnl (r) = u(r) = e y(r) = e r q(r) = e−kr rl q(r) =⇒
r r r
Rnl (r) = e−kr rl Lβw (x(r)) (6.69)

By reversing the substitutions (6.64), (6.60), and (6.63), we obtain x = 2y, y = kr, and k =
Z/(a0 n). Thus, we achieve that the right side of equation (6.69) directly depends on r:

(6.64) (6.60) (6.63) 2Zr

x = 2y = 2kr = (6.70)
a0 n

2Zr
 
(6.69) −kr l (6.63) − aZrn l (6.70) − aZrn l
Rnl (r) = e r Lβw (x) = e 0 r Lβw (x) = e 0 r Lβw (6.71)
a0 n

156
Quantum Theory I

We substitute for x in Lβw (x) the expression (2Zr)/(a0 n). Thus, we must also replace the w-
th derivative with respect to x in the Rodrigues formula (6.67) with the w-th derivative with
respect to r. For this, we calculate:
dx (6.70) 2Z dr a0 n
= =⇒ = (6.72)
dr a0 n dx 2Z
and
dw (xw+β e−x ) dw (xw+β e−x ) dw r (6.72) dw (xw+β e−x )
w
a0 n

(6.70)
= = =
dx w drw dxw drw 2Z
2Zr
w w " w+β #
a0 n d
 2Zr
−a
= e 0 =n
2Z drw a0 n
a0 n w 2Z w 2Z β dw
     
2Zr
 
w+β − a0 n
= r e =
2Z a0 n a0 n drw
2Z β dw
   2Zr

w+β − a0 n
= r e (6.73)
a0 n drw

Finally, substituting
R∞ 2 l
β = 2l + 1 and w = n − l − 1 into (6.71), and further considering the
normalization 0 r |Rn (r)|2 dr = 1, we can express the following expression for the radial wave
function:

s 3 
2 (n − l − 1)! 2Zr 2Zr
l
Z
  
2 − aZrn
Rnl (r) = e 0 L2l+1 (6.74)
n2 (n + l)! a0 a0 n n−l−1
a0 n
w
(10.30) (w + β)!
with Lβw (x) = (−1)m
X
xm
m=0
(w − m)!(β + m)!m!

Using the Rodrigues formula (6.67) and the substitution (6.73), the radial wave function (6.74)
can be rewritten as follows:
s
2 (n − l − 1)!Z 3 2Zr 1 2Zr
l −2l−1
dn−l−1
   
− aZrn 2Zr 2Zr
n+l − a0 n
Rnl (r) = 2 e 0 e a0 n r e
n (n + l)! a30 a0 n (n − l − 1)! a0 n drn−l−1
| {z }|   {z }
corresponding to (6.74) L2l+1 2Zr
according to (6.67) and (6.73)
n−l−1 a0 n

After canceling, the following alternative expression for the radial wave function remains:

s 3 
2 1 2Z
l
Z dn−l−1
 Zr
 
2
−l−1 − 2Zr
Rnl (r) = 2 r e a0 n rn+l e a0 n (6.75)
n (n − l − 1)!(n + l)! a0 a0 n dr n−l−1

We recognize that only for l = 0 at r = 0 is the radial wave function Rn0 different from zero; if
the electron has an angular momentum component l > 0, the wave function vanishes at r = 0.

As a reminder: To fully represent an eigenfunction ϕnlm (r, ϑ, φ) of the hydrogen atom in the
state n, l, m, we must only use the result (6.74) or (6.75), and then reverse the separation
(6.22):

(6.22)
ϕnlm (r, ϑ, φ) = Rnl (r) Ylm (ϑ, φ) (6.76)

157
 6 Hydrogen Atom

R10 (r)
Rnl (r)

Rnl (r)
R20 (r)

R31 (r) R30 (r)

R21 (r) R32 (r)

r r

r2 |R20 (r)|2
r2 |Rnl (r)|2

r2 |Rnl (r)|2

r2 |R32 (r)|2
r2 |R21 (r)|2

r2 |R30 (r)|2
r 2
|R10 (r)|2 r 2
|R31 (r)|2

r r

Abb. 35: Selected examples of radial wave functions Rnl (x) (top) and their probability densities
|Rnl (x)| (bottom). (left) No nodes. (right) One (R20 , R31 ) or two (R30 ) radial nodes.

6.3 Energy Levels and Notation

The energy levels in the hydrogen atom follow a relatively simple law of the form:
1
εn = −Ry
n2
The Rydberg constant Ry ≈ 13.605 eV sets the lowest energy value, i.e., the ground state, at
n = 1; due to the quadratic dependence on the principal quantum number n, the energy increases
with increasing n. We notice that the energy εn does not depend on the other quantum numbers
used to describe the angular components of the hydrogen atom: the orbital angular momentum
and the magnetic quantum number l and m. Our energy levels are thus degenerate in l and
m! Although this seems quite convenient at first glance, it is actually due to a “flaw” in our
derivation: We calculated the eigenenergies of the hydrogen atom non-relativistically!
The flaw can be corrected by solving the Dirac equation instead of the Schrödinger equation, as
it can merge quantum theory and relativity theory. In this case, however, we will find that our
“solution” is an approximation of the actual solution, which sometimes lifts the degeneracy of l.
However, the degeneracy of the magnetic quantum number m remains without further assump-
tions – only when we place our hydrogen atom in an external magnetic field B does it lead to

158
Quantum Theory I

l=0 l=1 l=2 l=3 l=4

n≥5
n=4
n=3

n=2

m = +2
m = +1
l=2 m=0
m = −1
m = −2

n=1

Abb. 36: Schematic representation of the energy levels of the hydrogen atom: Additionally
highlighted is the state |n, l⟩ = |3, 2⟩ and its magnetic splitting according to m.

l=0 l=1 l=2 l=3 l=4

n≥5
n=4
Brackett
n=3
Paschen

n=2
Balmer

n=1
Lyman

Abb. 37: Emission sequences of spectral transitions: According to (1.44), the energy of the
emitted photon depends on the spacing of the spectral lines.

159
 6 Hydrogen Atom

an energetic splitting according to m.

If an electron now finds itself in an orbital n > 1, it wants to fall back to the ground state n = 1
(or another more favorable state) for energetic reasons. This releases energy, which is emitted
as a photon with frequency νnm = ∆ε/h and corresponds to the difference in energy levels:

1 1
 
∆ε = Ry 2
− 2
m n

We find various emission series characterized by the final state n of the decay (for example, a
decay in the Balmer series always leads to n = 2).

So far, we have identified three quantum numbers to characterize the state of an electron bound
in a hydrogen-like system: the principal quantum number n, the orbital angular momentum
quantum number l, and the magnetic quantum number m. While n determines the energy of
the electron and its maximum achievable angular momentum (n−1 ≥ l), l determines the parity
of the state through a (−1)l term from (5.123). m actually plays a role energetically only if an
external magnetic field is present – otherwise, the magnetic quantum number with l influences
the actual geometry of our state.
We can introduce a new notation for the individual angular momentum quantum numbers l:

l = 0, 1, 2, 3, 4, . . . =⇒ l = s, p, d, f, g, . . . (6.77)

Let’s consider an example of a state: Our electron is characterized by the quantum numbers
n = 3 and l = 1. We designate this state in our (spectroscopic) shorthand notation as a 3p-state.

160
Quantum Theory I

7 Spin

Motivation: Spin
According to the Bohr’s atomic model, electrons move in orbits around the nucleus, to
which a quantized angular momentum can be assigned. Now the question arises whether
the electron – similar to a planet – also has an intrinsic angular momentum (spin), or
whether this is even possible for a point-like particle.

The Stern-Gerlach experiment conducted by Otto Stern and Walther Gerlach in

1922 represents the central experiment in this context, in which (unknowingly) the spin of
the electron was first demonstrated (see Figure 39). A beam of silver atoms passes through
an inhomogeneous magnetic field. The electron configuration of a silver atom consists of a
closed shell plus a 5s-electron. Due to this configuration, the total angular momentum of a
silver atom should actually be zero. However, it is observed that the incoming atom beam
is split into two outgoing beams in the inhomogeneous magnetic field. If the electron were
in a p-orbital (l = 1), there would have to be three outgoing beams due to the degeneracy
(2l + 1). The observation of two beams can only be explained by a half-integer angular
momentum of the electron. This internal angular momentum or spin of the electron was
theoretically explained by Paul Dirac in 1928 using relativistic quantum mechanics.

7.1 Magnetic Moment

In the previous chapters, we have already discussed the electron in the hydrogen atom and cal-
culated the geometric shape of the wave functions analytically in the position representation.
Another quantity, which is related to the angular momentum L, however, we have not yet dis-
cussed: the magnetic (dipole) moment µ.

Before we specifically describe the magnetic moment, let’s recall the classical angular momentum
L. It can be written as follows:

L = r × p = m (r × v) = m (r × (ω × r)) (7.1)

Using the “bac-cab” rule a × (b × c) = b (a · c) − c (a · b) and |r| ≡ r, this can be converted to:

L = m[ω(r · r) − r(r · ω)] = mr2 ω (7.2)

If we are interested in the magnitude of the angular momentum |L|, we get with |ω| ≡ ω and
the rotation period T = 2π/ω or ω = 2π/T :
2π
|L| = mr2 (7.3)
T
The classical magnetic (dipole) moment is given in classical physics as the product of the current
I and the area A:
µ = IA (7.4)
Let’s specifically consider a charged point particle (e.g., an electron) with charge q on a circular
orbit with rotation period T – the current I then results as I = q/T . For the circular area,
|A| = πr2 holds, where A points in the direction of the surface normal en and has the same
orientation as L or ω. By inserting into (7.4), we get:
q 2
µ= πr en (7.5)
T

161
 7 Spin

Rearranging the magnitude of the angular momentum |L| from (7.3) for T and inserting into
(7.5) provides a relationship between µ and L.

q|L| q e
µ= πr2 en = L=− L (7.6)
2mπr 2 2m 2me
In the last step, since we mainly deal with the electron, we have set the charge q as the negative
elementary charge −e, and m as the electron mass me . Due to the negative charge, the magnetic
moment µ points in the opposite direction of the angular momentum L. The relative length of
µ with respect to L is called the gyromagnetic ratio γL and can be calculated as follows:

|µ|
γL = (7.7)
|L|

A consequence of the semi-classical Bohr model of the atom for the electron is the Bohr magneton
µB = 9.274 × 10−24 J T−1 . It is defined as the magnitude of the magnetic moment produced by
an electron with the lowest orbital angular momentum quantum number (l = 1 with |L| = 1ℏ)
in the Bohr model on its circular orbit (with Bohr radius a0 ) around the nucleus:

eℏ
µB = (7.8)
2me
For the gyromagnetic ratio γL of the electron, we find using the Bohr magneton µB the relation:
e ℏ (7.8) µB
γL = = (7.9)
2me ℏ ℏ
So far the classical consideration. The transition from classical to quantum mechanics is achieved
via the correspondence principle; the angular momentum operator L̂ has already been defined
in (5.1) and is given by:
L̂ = r̂ × p̂
Also, through the correspondence principle, we can now form a new operator for the observable
of the magnetic moment µ̂. Here we initially introduce an additional factor gL without further
assumptions, which we will generally refer to as the Landé factor of the orbital angular momen-
tum. Later we will also find different Landé factors for other angular momentum sizes; in the
case of the orbital angular momentum L̂, however, it trivially holds that gL = 1. Considering
gL and with the help of (7.6), we find:

µB
µ̂ = sgn(q) gL L̂ (7.10)
ℏ

Here sgn(q) is the sign of the charge q of the considered particle; in the case of the electron,
sgn(q) = sgn(−e) = −1. An equivalent, shortened formulation would be µ̂ = sgn(q)γL L̂. It can
be seen from (7.10): If L̂ is quantized in units of ℏ, then µ̂ is also quantized in units of µB . Also,
if we now let µ̂ act on an eigenstate of the angular momentum operator, we can derive simple
results using the already known relations (5.59), (5.60), and (5.61).

In the previous chapters, we always chose the z-direction as the preferred axis of the angular
momentum; in the x- and y-directions, we cannot obtain “sharp” measurement values due to
the uncertainty relation. However, we can always rotate our coordinate system such that we
can use the simple eigenvalue relationships (5.60) and (5.115). Also within this chapter, in the
case of spin and general angular momenta, we will adhere to this convention and arbitrarily set
the z-direction as the preferred axis.

162
Quantum Theory I

z
ml L

µlz µl

Abb. 38: Schematic representation of angular momentum L̂ and magnetic moment µ̂l , where
for simplicity the gyromagnetic ratio γL = 1 is set.

7.1.1 Energy in the Magnetic Field

The existence of a magnetic moment is interesting in the sense that in the presence of a magnetic
field B, there is a change in energy. Accordingly, the eigenenergies and eigenfunctions of the
Hamiltonian operator Ĥ will change. Let’s first consider the classical case of a magnetic dipole
in an external magnetic field B. The potential, magnetic energy is given by:
(7.6) µB
Eµ = −µB = LB (7.11)
ℏ
If there is an inhomogeneous magnetic field, there is also a force effect Fµ on the dipole moment:

(7.11) µB
Fµ = −∇Eµ = − ∇(LB) (7.12)
ℏ
For simplicity, we choose our coordinate system so that our magnetic field only points in the
z-direction: B = Bez . The potential energy Eµ from (7.11) simplifies so that we can neglect the
x- and y-directions of the angular momentum (and hence in a following quantum mechanical
description the angular momentum operators L̂x and L̂y ):

µB B
Eµ,z = Lz (7.13)
ℏ

Through the correspondence principle, we can define from (7.13) an energy operator ĤB,z for
an electron in the directed magnetic field (where we also take the Landé factor gL into account).
At the same time, we will write ĤB,z in spherical coordinates – we already know L̂z in these
coordinates from (5.108):

µB gL B (5.108) ∂
ĤB,z = L̂z = −iµB gL B (7.14)
ℏ ∂φ

The eigenfunctions of ĤB,z are in the position representation the spherical harmonics. Starting
from an abstract eigenstate of the angular momentum operator |l, m⟩, according to (5.60), the
eigenvalues can be easily calculated to be:
µB B (5.60)
ĤB,z |l, m⟩ = gL L̂z |l, m⟩ = µB BgL m |l, m⟩ (7.15)
ℏ
If an external magnetic field is present, the states, which are otherwise degenerate, characterized
by the magnetic quantum number m, can split! This is referred to as the normal Zeeman effect.

163
 7 Spin

7.2 Stern-Gerlach Device

Let’s consider the setup of a Stern-Gerlach device as shown in Figure 39. A beam of silver atoms
with a well-defined speed is passed through a strongly inhomogeneous magnetic field B with an
orientation n given by the pole shoes of the magnet.

+ℏ/2

−ℏ/2

Abb. 39: Sketch of the Stern-Gerlach experiment with the double splitting of the particle beam.

Analogous to (7.12), the following force F acts on each particle:

F = −∇ (−µ · B) = ∇ (µ · B) (7.16)

µ represents the magnetic moment. It is important to understand that this deflecting force is
not the Lorentz force, but arises from the magnetic moment. Therefore, it is also important to
use neutral particles for the experiment, as the Lorentz force would otherwise completely overlay
the result. Assuming the orientation of the magnetic field as the z-direction, expression (7.16)
simplifies to:
∂B
 
F = µz ez (7.17)
∂z
For a finite force on the particles, it is essential for the experiment that the magnetic field indeed
has a finite gradient ∂B∂z ̸= 0. However, if the magnetic field is aligned along any direction n,
expression (7.16) can also be written as:

∂B
 
F = |µ| cos(α) n (7.18)
∂n

Here, ∂B
∂n should be the derivative of B ≡ |B| in the direction of n, and α represents the angle
between µ and B.

In the Stern-Gerlach experiment, a splitting of the beam of silver atoms into two sub-beams
was observed. When attempting to interpret the splitting classically, one might conclude that
the particles’ angular momentum is not statistically distributed, but that only two groups of
particles are generated in the particle source: those with angular momentum parallel to the
magnetic field and those with angular momentum antiparallel to the magnetic field. However,
the problem with this interpretation is that the measurement result is completely independent
of the orientation of the magnetic field. No matter how the magnetic field is oriented, one
always observes the division into two sub-beams! From a classical perspective, this makes no
sense, and Stern and Gerlach were confused by the result of their experiment. They hoped
to demonstrate the (observed) quantization of the magnetic moment, but an orbital angular
momentum with an integer angular momentum quantum number l = 1 should have resulted in
three sub-beams according to the triple magnetic degeneracy p(1) = (2 · 1 + 1) = 3 according to
(5.64).
A possible solution to this problem offers itself when we assume half-integer angular momenta.

164
Quantum Theory I

7.3 Postulate of Spin

In the chapter on angular momentum, we decided to rename the angular momentum quantum
number j from (5.56) to l for the orbital angular momentum, and to allow only positive integer
values. However, for the general angular momentum quantum number j, it is also permissible to
choose positive, half-integer values. Such half-integer angular momentum quantum numbers we
will now assign to “spin”, where we want to focus at this point on the spin of a single electron.
Instead of an integer angular momentum j, for spin j = s = 21 is chosen, in accordance with the
double splitting of the atomic beam in the Stern-Gerlach experiment. The associated operators
are labeled Ŝ2 and Ŝz . They function analogous to the orbital angular momentum operators L̂2
and L̂z , and follow the same eigenvalue relationships with the eigenfunctions |s, ms ⟩.

To actually explain the two lines in the Stern-Gerlach experiment, a magnetic (spin) moment µ̂s
must also exist, analogous to the orbital angular momentum. Depending on the actual value of
the spin (here s = 12 ), we obtain the following number of realization possibilities for the magnetic
component ms :
2s + 1 = 2
In general, for the magnetic component ms of spin s, it must lie in the value range ms =
−s, −s + 1, . . . , s − 1, s. In the case of a single spin with s = 12 , ms can therefore only take two
values, which we can write as:
1
ms = ± (7.19)
2
A clear interpretation of the spin is difficult; it is suggested to understand the electron spin as a
kind of “intrinsic angular momentum”, but upon closer inspection, this classical interpretation
fails. This is because the speed v with which the “equator” of an electron with classical electron
radius would have to rotate would exceed the speed of light (v ≫ c).

Spin has another astonishing peculiarity: Only after a rotation by φ = 4π does a wave function
return to itself again. For orbital angular momentum, we find φ = 2π, which corresponds to
classical expectations. It is as if an electron “has to rotate twice around its own axis” before it
“looks the same again”. This can be explained by the magnetic quantum number ms of spin,
which is half-integer and thus lets the phase of a wave function rotate only half as fast:
φ φ (φ + 4π)
      
!
exp(ims φ) = exp i = exp i + 2π = exp i
2 2 2
A coherent description of spin becomes possible only through the solution of the Dirac equation,
which is a relativistic extension of the Schrödinger equation. Instead of postulating the spin
in the non-relativistic case and introducing it artificially, the solution of the Dirac equation
naturally provides both the spin and antiparticles.

7.4 Spin Eigen System

In the following chapter, we will deal with a pure spin system s = 21 (i.e., there is no orbital
angular momentum L̂) to learn about the relevant eigenvalue equations, operators, and repre-
sentations. Let’s begin with a consideration of the magnetic moment of spin µ̂s :
µB
µ̂s = sgn(q) gs Ŝ (7.20)
ℏ
Here, sgn(q) is the sign of the charge q of the considered particle; in the case of the electron,
sgn(q) = sgn(−e) = −1. For the electron, the Landé factor gs no longer takes the trivial value
1; it specifically holds for electron spin:

gs = 2.002319... ≈ 2 (7.21)

165
 7 Spin

The approximately correct factor gs = 2 can be derived from the solution of the Dirac equation
for fermions. A consideration within the framework of quantum field theory allows for an even
clearer picture and the determination of the decimal places.

7.4.1 Spin Operators and Commutators

As an angular momentum operator, the spin operator Ŝ satisfies the same commutator relations
as the orbital angular momentum L̂ and is also a generator of rotations. When calculating the
commutator between two spin operators, the following relation is always fulfilled (instead of Ŝj ,
any vector operator V̂j could be used here):

[Ŝi , Ŝj ] = iℏεijk Sk (7.22)

We can simultaneously measure two spin-related quantum numbers exactly: the observable Ŝ2
and a single component of spin Ŝi , where we here, as usual, choose the z-direction. Ŝ2 and Ŝz
are compatible and can therefore be measured simultaneously:

[Ŝ2 , Ŝz ] = 0 (7.23)

Relation (7.23) again allows us to find a common eigen system between the magnitude of the
spin operator Ŝ2 and an arbitrary spin component Ŝz . (7.22) also forbids the possibility of
simultaneously measuring different components of spin, so that our choice of the z-component
leads to the fact that we cannot simultaneously measure the x- and y-components exactly.

We now know which observables we can determine simultaneously. But what is the action of the
associated operators? Since we only renamed the general angular momentum quantum number
j as s in the case of spin (instead of, as with the orbital angular momentum, to l), we already
know the answer: The already known relations (5.59), (5.60), and (5.61) apply, only we replace
the orbital angular momentum quantum number l with the spin quantum number s, and the
magnetic angular momentum quantum number m with the magnetic spin quantum number
ms . The magnitude operator of spin Ŝ2 thus acts on a spin eigenstate |s, ms ⟩ in the following
way:

Ŝ2 |s, ms ⟩ = ℏ2 s(s + 1) |s, ms ⟩ (7.24)

The z-component of spin leads us to an eigenvalue, which depends on the magnetic spin quantum
number. It holds:

Ŝz |s, ms ⟩ = ℏms |s, ms ⟩ (7.25)

We also find variations equivalent to the angular momentum for raising and lowering operators
Ŝ± :
q
Ŝ± |s, ms ⟩ = ℏ s(s + 1) − ms (ms ± 1) |s, ms ± 1⟩ (7.26)

For the spin of a single electron, we already know s = 21 or ms = ± 12 . We can thus rewrite the
eigenstate anew, using the following notation variants optionally:

| 12 , + 12 ⟩ = |↑⟩ = |+⟩ = |0⟩

(
|s, ms ⟩ = (7.27)
| 12 , − 12 ⟩ = |↓⟩ = |−⟩ = |1⟩

166
Quantum Theory I

We thus obtain a clearer picture for (7.24) and (7.25). For a single electron, the eigenvalue of
the magnitude operator Ŝ2 is always the same, independent of the spin direction:
(7.27) (7.24) (7.27) 3 2
 
Ŝ2 |↑⟩ = Ŝ2 | 12 , + 12 ⟩ = ℏ 12 1
2 + 1 | 12 , + 12 ⟩ = 4 ℏ |↑⟩
(7.27) (7.24)
  (7.27) 3 2
(7.28)
Ŝ2 |↓⟩ = Ŝ2 | 12 , − 12 ⟩ = ℏ 12 1
2 + 1 | 12 , − 12 ⟩ = 4 ℏ |↓⟩

The eigenvalue of Ŝz allows for a bit more variation; depending on the spin orientation, it results
in a positive or negative sign:
(7.27) (7.26) (7.27)
Ŝz |↑⟩ = Ŝz | 12 , + 12 ⟩ = + 12 ℏ | 21 , + 12 ⟩ = + 12 ℏ |↑⟩
(7.27) (7.26) (7.27)
(7.29)
Ŝz |↓⟩ = Ŝz | 21 , − 12 ⟩ = − 12 ℏ | 21 , − 12 ⟩ = − 12 ℏ |↓⟩

The ladder operators also behave as expected, with Ŝ+ |↑⟩ = Ŝ− |↓⟩ = 0 trivially disappearing:
r 
(7.27) (7.26)
  
Ŝ+ |↑⟩ = Ŝ+ | 12 , + 21 ⟩ = ℏ 1
2
1
2 +1 − 1
2
1
2 + 1 | 12 , + 32 ⟩ = 0
(7.27) (7.26)
Ŝ+ |↓⟩ = Ŝ+ | 12 , − 21 ⟩ = ℏr
|↑⟩
(7.27) (7.26)
    (7.27) (7.30)
Ŝ− |↑⟩ = Ŝ− | 21 , + 12 ⟩ = ℏ 1
2
1
2 +1 − 1
2
1
2 − 1 | 12 , − 12 ⟩ = ℏ |↓⟩
r 
(7.27) (7.26)
  
Ŝ− |↓⟩ = Ŝ− | 12 , − 12 ⟩ = ℏ 1
2
1
2 +1 + 1
2 − 12 − 1 | 21 , − 32 ⟩ = 0

7.4.2 Matrix Representation and Pauli Matrices

So far, all our wave functions could be represented in the position space Hr . The representation
of the newly added spin eigenstates occurs exclusively in a spin Hilbert space Hs . To describe
all the characteristics of an electron, we introduce a total Hilbert space H, which consists of a
position component and a spin component:

H = Hr ⊗ Hs with dim H = 2 · dim Hr (7.31)

The fact that we consider spin at all when solving the Schrödinger equation is an ad hoc as-
sumption, as only the Dirac equation naturally provides spin eigenstates (so-called spinors).
With a single electron, only two spin states (|+⟩ and |−⟩), or their superposition, are realized.
According to (7.31), the corresponding spin Hilbert space is therefore two-dimensional.

Let |ψ⟩s ∈ Hs be an eigenstate of the operators Ŝ2 and Ŝz , then orthogonality holds for the
individual eigenstates:

⟨↑ | ↑⟩ = ⟨↓ | ↓⟩ = 1 and ⟨↓ | ↑⟩ = ⟨↑ | ↓⟩ = 0 (7.32)

The eigen system must also be complete; for our two-dimensional Hilbert space, the simple
relationship holds:
1 = |↑⟩ ⟨↑| + |↓⟩ ⟨↓| (7.33)
We can represent any spin state as a superposition of |↑⟩ and |↓⟩, where the coefficients α, β ∈ C
satisfy:
|ψ⟩s = α |↑⟩ + β |↓⟩ (7.34)
α and β must satisfy the relationship: |α|2 + |β|2 = 1. This can be quickly shown by calculating
the inner product:

⟨ψ|ψ⟩s = (α∗ ⟨↑| + β ∗ ⟨↓|) (α |↑⟩ + β |↓⟩) =

= α∗ α ⟨↑ | ↑⟩ + β ∗ β ⟨↓ | ↓⟩ = |α|2 + |β|2 = 1 □

167
 7 Spin

In analogy to R2 as a two-dimensional vector space, we can also represent the spin eigenstates
|+⟩ and |−⟩ as vectors (where |ϕi ⟩ ∈ {|+⟩, |−⟩}):
! !
{|ϕi ⟩} 1 {|ϕi ⟩} 0
|↑⟩ −−−→ and |↓⟩ −−−→ (7.35)
0 1

We could also define the vectorial representation of |↑⟩ and |↓⟩ the other way around, but the
notation must remain consistent. The general spin state from (7.34) is given in this vector
notation as: ! !
⟨↑ |ψ⟩s α
ψs = = (7.36)
⟨↓ |ψ⟩s β

In the eigen space of the operators Ŝ2 and Ŝz , the eigenstates can be represented as vectors; a
spectral decomposition allows representing the operators as matrices as well:

3ℏ2
(7.28) {|ϕi ⟩}
Ŝ2 = Ŝ2 1 = Ŝ2 (|↑⟩ ⟨↑| + |↓⟩ ⟨↓|) = (|↑⟩ ⟨↑| + |↓⟩ ⟨↓|) −−−→
4 #
{|ϕi ⟩} 3ℏ 3ℏ2 3ℏ2
" ! ! " ! !# !
2
1   0   1 0 0 0 1 0
−−−→ 1 0 + 0 1 = + =
4 0 1 4 0 0 0 1 4 0 1

In the spin eigenbasis, we can thus represent Ŝ2 as a matrix of the following form:

3ℏ2
!
1 0
S2 = (7.37)
4 0 1

Analogously, using (7.29) for the z-component of the spin operator Ŝz :
ℏ (7.28) {|ϕi ⟩}
(|↑⟩ ⟨↑| − |↓⟩ ⟨↓|) −−−→
Ŝz = Ŝz 1 = Ŝz (|↑⟩ ⟨↑| + |↓⟩ ⟨↓|) =
" ! ! 2 # " ! !# !
{|ϕi ⟩} ℏ 1   0   ℏ 1 0 0 0 ℏ 1 0
−−−→ 1 0 − 0 1 = − =
2 0 1 2 0 0 0 1 2 0 −1

We can give the matrix representation of Ŝz in the spin eigenbasis and find the corresponding
representation as a matrix as:
!
ℏ 1 0
Sz = (7.38)
2 0 −1

As expected, Ŝ2 and Ŝz are diagonal in matrix representation – since the description takes place
in the common eigen system of the two operators, we can identify the diagonal elements as
eigenvalues.

Example: Matrix Representation of Ŝx and Ŝy

The representation of Ŝx and Ŝy as matrices is somewhat trickier compared to the z-
component, as we are not in the spin eigen system of the two operators. Expressing Ŝx
and Ŝy in analogy to (5.67) and (5.68) via Ŝ+ and Ŝ− , we have:

1  1  
Ŝx = Ŝ+ + Ŝ− and Ŝy = Ŝ+ − Ŝ− (7.39)
2 2i

Let’s first consider what the matrix representations of the ladder operators Ŝ+ and Ŝ−

168
Quantum Theory I

look like. As in (7.37) and (7.38), we introduce a complete identity 1 and write
!
(7.30) 0 1 {|ϕi ⟩}
Ŝ+ = Ŝ+ 1 = Ŝ+ |↑⟩ ⟨↑| + Ŝ+ |↓⟩ ⟨↓| = ℏ |↑⟩ ⟨↓| −−−→ ℏ
0 0
!
(7.30) 0 0 {|ϕi ⟩}
Ŝ− = Ŝ− 1 = Ŝ− |↑⟩ ⟨↑| + Ŝ− |↓⟩ ⟨↓| = ℏ |↓⟩ ⟨↑| −−−→ ℏ
1 0

We have utilized the action of the ladder operators on a multiplet state: The highest state
|↑⟩ cannot be increased further, just as the lowest state |↓⟩ cannot be decreased further.
Ŝx and Ŝy result from the appropriate summation of the ladder operators according to the
relations (7.39):

1
" ! !# !
(7.39)
 {|ϕ ⟩} ℏ 0 1 0 0 ℏ 0 1
= Ŝ+ + Ŝ− −−−→ + = (7.40)
i
Ŝx
2 2 0 0 1 0 2 1 0
1 
" ! !# !
(7.39)
 {|ϕ ⟩} ℏ 0 1 0 0 ℏ 0 −i
= = (7.41)
i
Ŝy Ŝ+ − Ŝ− −−−→ −
2i 2i 0 0 1 0 2 i 0

These matrices are not diagonal. Letting Ŝx and Ŝy act on an eigenstate |↑⟩ or |↓⟩, the
eigenstate changes. For example, for Ŝx |↓⟩ = (ℏ/2) |↑⟩, we obtain:
! ! !
ℏ {|ϕi ⟩} ℏ 0 1 0 ℏ 1
Ŝx |↓⟩ = |↑⟩ −−−→ =
2 2 1 0 1 2 0

Pauli Matrices Ŝx , Ŝy , and Ŝz are usually expressed with the Pauli matrices σx , σy , and σz
(often called σ1 , σ2 , and σ3 in literature), which have the following form:
! ! !
0 1 0 −i 1 0
σx ≡ σ1 ≡ and σy ≡ σ2 ≡ and σz ≡ σ3 ≡ (7.42)
1 0 i 0 0 −1

It is clear from (7.38), (7.40), and (7.41) that the components of the spin operator Ŝ and the
Pauli matrices differ only by a constant factor ℏ/2, which corresponds to the eigenvalue of Ŝz :
ℏ
Ŝi = σi (7.43)
2
Since the difference is only in a multiplicative factor, all eigenvalue relations and commutators
for the Pauli matrices function the same as for Ŝx , Ŝy , and Ŝz , differing only by the factor ℏ/2.
For the commutator from (7.22), we find the equivalent relation:

[σi , σj ] = 2iεijk σk (7.44)

The correctness of the commutator (7.44) can be easily verified by substituting (7.43) into
(7.22). Furthermore, for the anticommutator (in relativistic quantum theory, this relationship
is a determining part of the Clifford algebra), the following holds:

{σi , σj } = 2δij (7.45)

(7.45) can again be verified by substituting the actual Pauli matrices. We can use (7.44) and
(7.45) to find a generally valid relationship for arbitrary vector operators  and B̂:

(σ Â)(σ B̂) = ÂB̂ + iσ(Â × B̂) (7.46)

169
 7 Spin

Example: Product of Pauli Matrices and Vector Operators

Let’s start with the anticommutator relation {σi , σj } = 2δij and manipulate it. We find:

{σi , σj } = σi σj + σj σi = σi σj − σj σi + σj σi + σj σi = [σi , σj ] + 2σj σi

Rearranging this equation allows us to represent the product of two Pauli matrices through
the commutator and the anticommutator:
1
σj σi = ({σi , σj } + [σi , σj ]) (7.47)
2

We can use (7.47) to derive (σ Â)(σ B̂) = ÂB̂ + iσ(Â × B̂) from (7.46). Â and B̂ should
be arbitrary vector operators. To facilitate this derivation, we switch to index notation:
(7.47)
(σ Â)(σ B̂) = σi σj Âi B̂j =
1 (7.44,7.45)
= ({σi , σj } + [σi , σj ]) Âi B̂j =
2
1
= (2δij + 2iεijk σk ) Âi B̂j =
2
= δij Âi B̂j + iεijk σk Âi B̂j =
= Âi B̂i + iσk εijk Âi B̂j =
= (σ Â)(σ B̂) + iσ(Â × B̂) □

In the last step, it was used that εijk Âi B̂j corresponds to the cross product of the quantities
 and B̂.

Properties of the Pauli Matrices All Pauli matrices must satisfy the following properties; that
these statements are true can be easily confirmed by explicitly inserting the Pauli matrices from
(7.42):

Hermiticity: Each of the Pauli matrices is hermitian, thus:

σi† = σi (7.48)

Unitarity: Likewise, all Pauli matrices are unitary, hence:

σi−1 = σi† (7.49)

Determinant: The determinant of any Pauli matrix is always:

det{σi } = −1 (7.50)

Trace: The trace of any Pauli matrix vanishes:

Tr{σi } = 0 (7.51)

Magnitude squared: The magnitude squared of each component of the vector of Pauli
matrices σi results in unity (which follows from the hermiticity and unitarity of σi ):

σx2 = σy2 = σz2 = 1 (7.52)

170
Quantum Theory I

7.4.3 Bloch Sphere

So far, to describe spin, we have always assumed Ŝ2 and Ŝz to be the sharply measurable
observables, while Ŝx and Ŝy are “smeared”. Let us now abandon this simplification and assume
spin states oriented parallel or antiparallel in the spin space to an arbitrarily oriented unit
vector n = (nx , ny , nz )⊺ . We seek a spin operator Ŝn , for which such states are eigenstates. A
description in spherically symmetric coordinates (for a unit sphere r = 1) is suitable, where for
the spin operator in the direction n the following applies: Ŝn ≡ Ŝ · n:

sin(ϑ) cos(φ)
       
σ n σ
ℏ ℏ  x  x ℏ  x 
Ŝn = Ŝ · n = σ · n = σy  · ny  = σy  ·  sin(ϑ) sin(φ)  =

2 2 2
σz nz σz cos(ϑ)
" ! ! ! #
ℏ 0 1 0 −i 1 0
= sin(ϑ) cos(φ) + sin(ϑ) sin(φ) + cos(ϑ) (7.53)
2 1 0 i 0 0 −1

The expression from (7.53) can be simplified using Euler’s formula e±iφ = cos(φ) ± i sin(φ).
Therefore, we find the following expression for Ŝn :
!
ℏ cos(ϑ) sin(ϑ)e−iφ
Ŝn ≡ Ŝn (ϑ, φ) = (7.54)
2 sin(ϑ)e+iφ − cos(ϑ)

Thus, we have found a compact expression for the spin operator Ŝn for an arbitrary orientation
n of the spin. However, an eigenstate of Ŝn can always be expressed in the eigenbasis of Ŝz !
If we solve the eigenvalue problem with respect to Ŝn , we find the eigenvalues and eigenstates
(7.55) with the help of the following calculation.

The eigenvalues of Ŝn are calculated using det(Ŝn − 1λ) = 0, which yields the following charac-
teristic polynomial:

ℏ2 ℏ2
  
ℏ ℏ
0=− cos(ϑ) + λ cos(ϑ) − λ − sin2 (ϑ) = λ2 −
2 2 4 4

We find two eigenvalues λ1 = +ℏ/2 and λ2 = −ℏ/2, just like with the already known operator Ŝz .
To calculate the corresponding eigenstates, we use the two trigonometric relations cos2 (ϑ/2) =
(1 + cos(ϑ))/2 and sin2 (ϑ/2) = (1 − cos(ϑ))/2 as well as sin(ϑ) = 2 sin(ϑ/2) cos(ϑ/2). It follows:
! !
cos(ϑ) − 1 sin(ϑ)e−iφ 0 − sin2 (ϑ/2) sin(ϑ/2) cos(ϑ/2)e−iφ 0
λ1 : =
sin(ϑ)e+iφ − cos(ϑ) − 1 0 sin(ϑ/2) cos(ϑ/2)e +iφ − cos2 (ϑ/2) 0
! !
cos(ϑ) + 1 sin(ϑ)e−iφ 0 cos2 (ϑ/2) sin(ϑ/2) cos(ϑ/2)e−iφ 0
λ2 : =
sin(ϑ)e+iφ − cos(ϑ) + 1 0 sin(ϑ/2) cos(ϑ/2)e+iφ sin2 (ϑ/2) 0

By solving these two systems of equations, we obtain for the new eigenstates |↑n ⟩ and |↓n ⟩,
expressed in the eigenbasis {|↑⟩ , |↓⟩} of Ŝz , the two general states:
!
cos(ϑ/2)e−iφ ℏ
|↑n ⟩ = = cos(ϑ/2)e−iφ |↑⟩ + sin(ϑ/2) |↓⟩ for λ1 = +
sin(ϑ/2) 2
! (7.55)
sin(ϑ/2)e−iφ −iφ ℏ
|↓n ⟩ = = sin(ϑ/2)e |↑⟩ − cos(ϑ/2) |↓⟩ for λ1 = −
− cos(ϑ/2) 2

A general spin state |↑n ⟩ or |↓n ⟩ is thus developed into the eigenstates |↑⟩ or |↓⟩ of Ŝz , where
the development coefficients describe the position on the surface of a unit sphere. This can be
illustrated graphically using the Bloch sphere (also called Poincaré sphere), on which each spin

171
 7 Spin

state |ψ⟩ = |↑n ⟩ can be uniquely represented on the sphere surface by the angles ϑ and φ. At
the same time, this representation indicates the spatial angles in which the corresponding spin
state can be measured sharply.
The fact that each state in a two-dimensional complex vector space can be uniquely associated
with a spatial angle ultimately follows from the isomorphism between the Lie algebras for the
SU(2) and the SO(3). For spin quantum numbers higher than s = 21 , this unique correspondence
no longer holds.

|↑⟩

|ψ⟩

φ y

|↓⟩

Abb. 40: Schematic representation of the Bloch sphere with an arbitrary state |ψ⟩ = |↑n ⟩ with
the spatial angles ϑ and φ.

Example: Rotation of Spin Eigenstates

To illustrate spin states in arbitrary directions, we choose the simple case for a fixed
azimuth angle φ = 0 and a variable polar angle ϑ. Substituting into the finished formula
for the spin eigenfunctions (7.55), we find for a general ϑ the state |↑ϑ,0 ⟩, respectively for
ϑ = π/2 the state |↑x ⟩:

ϑ=π/2 1
|↑ϑ,0 ⟩ = cos(ϑ/2) |↑⟩ + sin(ϑ/2) |↓⟩ −−−−→ |↑x ⟩ = √ (|↑⟩ + |↓⟩)
2

z z
|↑⟩ |↑⟩
|ψ⟩

ϑ ϑ = π/2
x |ψ⟩ x

Abb. 41: Rotation of an eigenstate of the Ŝz operator by (left) the arbitrary angle ϑ and
(right) by ϑ = π/2. In both cases, φ = 0 applies.

172
Quantum Theory I

With ϑ = π/2, we have rotated the general spin state |↑ϑ,0 ⟩ directly to the x-axis. We
can now calculate the probability with which we measure +ℏ/2 for the observable Ŝz for
a system in the state |↑ϑ,0 ⟩ or |↑x ⟩:

ϑ=π/2 1
⟨P̂↑ϑ,0 ⟩ = | ⟨↑ | ↑ϑ,0 ⟩ |2 = cos2 (ϑ/2) −−−−→ ⟨P̂↑x ⟩ =
2

In-Depth: Modern Interpretation of the Stern-Gerlach Experiment

When there is an interaction between the magnetic field in the Stern-Gerlach apparatus
and the spin of the particle, its energy changes. The energy during such a process can be
expressed using the Hamiltonian operator Ĥ, which has the following form:
µB
Ĥ = −µ̂ · B ≈ 2 B Ŝ · n = µB B σ̂ · n = α σ̂ · n (7.56)
ℏ
For an arbitrary direction n of the magnetic field, we can write (7.56) as a matrix in the
eigenbasis of σ̂z analogous to (7.54), expressing n in spherically symmetric coordinates
according to (5.83) (assuming a unit sphere with r = 1):
!
cos(ϑ) sin(ϑ)e−iφ
Ĥ = α (7.57)
sin(ϑ)e+iφ − cos(ϑ)

Depending on the polar and azimuth angles ϑ and φ, we can manipulate the structure of
the Hamiltonian operator. Three configurations of the two angles {ϑ, φ} should be high-
lighted: {π/2, 0}, {π/2, π/2}, and {0, 0}, which correspond to the x-, y-, and z-directions
of the magnetic field B – it is evident that these configurations each generate one of the
Pauli matrices.

All particles that are located in the “upper” sub-beam after the measurement are particles
in the state |↑n ⟩, and have an eigenvalue of +ℏ/2. So if a particle takes the upper path,
and the Stern-Gerlach apparatus is interpreted as a measuring device, then one can say
that the state |↑n ⟩ has been measured on the particle. If one interprets it as a preparation
device, one can also say that the particle has been prepared in the state |↑n ⟩. Conversely,
all particles in the “lower” sub-beam are in the state |↓n ⟩, and have an eigenvalue of −ℏ/2.
The Stern experiment is depicted in Figure 39.

Only with quantum physics and the concept of spin does the result of the Stern-Gerlach
experiment make sense: The Stern-Gerlach apparatus is a measuring device that measures
the spin of the particles in a certain direction (e.g., in the z-direction). If a particle has
not been prepared before the measurement, then it does not have a well-defined spin ori-
entation – in a very fundamental sense. It is not just that we do not know the orientation;
it does not actually exist yet. Only through the measurement does the observed reality
“arise” in a way, where the spin- 21 particle is either aligned parallel or antiparallel to the
measurement direction – and this regardless of which measurement direction was chosen.

However, after the measurement with the Stern-Gerlach apparatus, there are two sub-
beams, and in each sub-beam, the particles have a well-defined spin orientation along the
chosen axis. If one then measures along this axis again, the measurement result can be pre-
dicted with certainty. This is an example that in quantum mechanics every measurement
process is also a preparation process.

173
 7 Spin

7.5 Total Angular Momentum

Motivation: Total Angular Momentum

We now want to move on to the total angular momentum of a system that has more than
one angular momentum as a degree of freedom. This can involve single particles with, for
example, spin and orbital angular momentum, as well as the total angular momentum of
multiple particles, which may have different intrinsic angular momenta/spins. To under-
stand the behavior of such systems, it is necessary to perform the addition of the individual
angular momentum operators and find a quantum mechanical representation of the new
eigenstates.

A relevant example of total angular momentum is once again the hydrogen atom. The
electron in the hydrogen atom has both an orbital angular momentum and a spin. In an
external magnetic field, one observes a splitting of the emission and absorption lines – this
splitting and the corresponding number of allowed transitions can only be described by a
total angular momentum that takes into account the coupling of the individual angular
momentum operators.

Another example is the helium atom, which consists of two electrons and the nucleus. In
the case where the orbital angular momentum of both electrons is neglected or l = 0 is
assumed, describing the total angular momentum of both electrons requires taking into
account the coupling of the spins. In the coupled basis, one obtains a so-called singlet
state with total spin S = 0 (parahelium) and a triplet state with S = 1 (orthohelium).
These states can also be used for understanding the so-called quantum entanglement.

Let us now transition from a single, free electron to the hydrogen atom from the previous
chapter, and try to find a more complete picture for the electron bound in this system using
the knowledge about spin we have gained here. The electron bound in the central potential
(without spin) can be characterized by the quantum numbers {n, l, ml }; describing the spin
state requires the additional quantum numbers {s, ms }. From (7.31), we know that the orbital
angular momentum and spin exist in different spaces Hr and Hs – a product state |ψ⟩ ∈ H is
given accordingly by:
|ψ⟩ = |n, l, ml ⟩ ⊗ |s, ms ⟩ ≡ |n, l, ml ; s, ms ⟩ (7.58)
For the inner product of two normalized states |ψ⟩ and |ψ ′ ⟩, orthogonality follows. It holds:

⟨ψ|ψ ′ ⟩ = ⟨n, l, ml ; s, ms |n′ , l′ , m′l ; s′ , m′s ⟩ = δnn′ δll′ δml m′l δss′ δms m′s (7.59)

For simplicity, we will always assume n = n′ in the following, as we are initially only interested in
the relations between orbital angular momentum and spin. Both L̂ and Ŝ are angular momentum
operators and can be described analogously. The difference lies in the space in which these
operators exist. The goal now is to transform |ψ⟩ in a way that allows the description of a total
angular momentum.

7.5.1 Product Basis and Coupled Basis

So far, in position space, we have found the following compatible operators: {Ĥ, L̂2 , L̂z }. We
are thus able to measure the energy, as well as the magnitude of the angular momentum and
its z-component simultaneously with certainty. We identify this set of observables with the
following “good” quantum numbers:

{Ĥ, L̂2 , L̂z } =⇒ {n, l, ml }

174
Quantum Theory I

By adding the spin system, we now have two more operators available, which can also be
measured simultaneously with certainty: {Ŝ2 , Ŝz }. Again, we find a set of “good” quantum
numbers that can be determined simultaneously:

{Ŝ2 , Ŝz } =⇒ {s, ms }

However, the question arises as to whether {Ĥ, L̂2 , L̂z } and {Ŝ2 , Ŝz } are compatible operators,
making a simultaneous measurement of the respective observables possible. Since the operators
of the position space L̂2 and L̂z cannot act on the spin part of a state and vice versa, we
immediately recognize that a simultaneous measurement of these quantities will be possible.
For Ĥ, such a clear statement is not a priori possible, as the Hamiltonian operator can indeed
contain terms that act on the spin (for example, in the hydrogen atom in a magnetic field). If
this is the case, and there are terms that do not commute with Ŝ2 and Ŝz , a common eigenstate
system cannot generally be found.

Product Basis Let’s assume, for simplicity, that the Hamiltonian is entirely independent of
spin, whereby {Ĥ, L̂2 , L̂z , Ŝ2 , Ŝz } is a set of compatible observables, and a characterization of a
wave function in the form (7.58) is possible. In this case, we speak of the product basis with the
quantum numbers:
{L̂2 , L̂z ; Ŝ2 , Ŝz } =⇒ {l, ml ; s, ms } (7.60)
In this representation, L̂ and Ŝ are independent of each other, and, as shown in Figure 42, a
separate measurement of ml and ms is possible!

L
ml

ms
S

Abb. 42: Schematic representation of the product basis: orbital angular momentum L̂ and spin
Ŝ are independent of each other and a separate measurement of the projections in
the z-direction is possible (ml and ms ).

However, this set of compatible observables is not unique, as we can build another set of com-
patible observables using the total angular momentum Ĵ:

Ĵ = L̂ + Ŝ = L̂ ⊗ 1s + 1r ⊗ Ŝ (7.61)

Ĵ thus acts both on eigenstates of the orbital angular momentum and on eigenstates of the
spin, and generally still fulfills the function of an angular momentum operator. Again, in all
generality, the already known commutation relation holds:

[Jˆi , Jˆj ] = iℏεijk Jˆk (7.62)

175
 7 Spin

That this must be satisfied can be easily verified by using the already known commutators (5.23)
and (7.22). We show:

[Jˆi , Jˆj ] = [L̂i + Ŝi , L̂j + Ŝj ] =

(5.23,7.22)
= [L̂i , L̂j ]+[L̂i , Ŝj ] + [Ŝi , L̂j ] + [Ŝi , Ŝj ] =
= iℏεijk L̂k + iℏεijk Ŝk =
(7.61)
= iℏεijk (L̂k + Ŝk ) =
= iℏεijk Jˆk □ (7.63)

Additionally, it holds that the operator for the magnitude of the total angular momentum
operator Ĵ2 and the operator for the z-component Jˆz are compatible, and thus can form a
common eigenstate system:
[Ĵ2 , Jˆz ] = 0 (7.64)
However, it is no longer clear in advance whether the set of compatible observables {Ĵ2 , Jˆz } can
be added to the observables of the product basis (7.60). To address this, we ask whether Ĵ2 and
Jˆz commute with the observables {L̂2 , L̂z , Ŝ2 , Ŝz } – Ĥ has been omitted here as the Hamiltonian
will have terms that are at most linear combinations of L̂ and Ŝ.

Coupled Basis For Jˆz = L̂z + Ŝz , it can be quickly ensured that this observable is compatible
with all other observables {L̂2 , L̂z , Ŝ2 , Ŝz }. But what about Ĵ2 ? Ĵ2 can be expressed as:
 
Ĵ2 = (L̂ + Ŝ)2 = L̂2 + Ŝ2 + 2L̂Ŝ = L̂2 + Ŝ2 + 2 L̂x Ŝx + L̂y Ŝy + L̂z Ŝz (7.65)

The term L̂2 +Ŝ2 poses no problem, as it is again compatible with all other observables. However,
the cross components L̂x Ŝx + L̂y Ŝy do not commute with the observables {L̂z , Ŝz }! We thus find
no common eigenstate system of the form {L̂2 , L̂z , Ŝ2 , Ŝz , Ĵ2 , Jˆz }!

mj J

Abb. 43: Schematic representation of the total angular momentum Ĵ and its components L̂
and Ŝ. It is evident that we cannot measure orbital angular momentum and spin
sharply, as they precess around the total angular momentum.

Without proof, the following theorem holds: In a system with more than one set of simul-
taneously diagonalizable operators, the number of compatible and independent operators is a
conserved quantity. For our case, this means that we can only find four mutually compatible
observables. We choose the set:

{L̂2 , Ŝ2 , Ĵ2 , Jˆz } =⇒ {l, s, j, mj } (7.66)

176
Quantum Theory I

While Jˆz was already identified as compatible with the other operators, the problematic oper-
ators {L̂z , Ŝz } have now been removed. Ĵ2 is compatible with L̂2 and Ŝ2 , and of course, with
Jˆz . In the case of (7.66), we speak of the coupled basis.

In the coupled basis {L̂2 , Ŝ2 , Ĵ2 , Jˆz }, only the magnetic component of the total angular momen-
tum mj can actually be measured “sharply” simultaneously with l, s, and j – ml and ms , on the
other hand, cannot. On the other side, it is not possible in the product basis {L̂2 , L̂z , Ŝ2 , Ŝz }
to measure the magnitude of the total angular momentum and its z-component simultaneously
with the other observables of the product basis. The choice of basis depends on the respective
problem. In a strong external magnetic field, the interactions between the individual angular
momenta and the magnetic field are so great that the total spin quantum number j is no longer
a good quantum number to describe the states – the angular momenta L̂ and Ŝ each couple
individually to the external magnetic field, which allows the quantum numbers l and s to be
determined (Paschen-Back effect). The product basis is therefore favorable for measurements
with a strong external magnetic field. In a weak external magnetic field, on the other hand, the
interaction between spin and orbital angular momentum is the dominant influence on the system
behavior, and therefore the coupled basis is preferable. Here, the total angular momentum Ĵ
couples to the magnetic field.

7.5.2 States in the Product Basis and Coupled Basis

Given two general angular momenta Ĵ1 and Ĵ2 . For example, Ĵ1 could be the orbital angular
momentum L̂, and Ĵ2 the spin Ŝ of a hydrogen atom. However, we deliberately choose a general
formulation here to clarify that the following relations hold for any angular momentum operators.
The total angular momentum of the system thus becomes:

Ĵ = Ĵ1 + Ĵ2 (7.67)

In what interval do the quantum numbers j and mj of the operators Ĵ2 and Jˆz lie? While
the magnitudes of Ĵ1 and Ĵ2 are fixed, their direction relative to each other can change – it is
therefore possible for the individual angular momenta to point in the same direction, or in the
opposite direction. It follows:

|j1 − j2 | = jmin ≤ j ≤ jmax = j1 + j2 (7.68)

The calculation of mj is simpler, as each combination of mj1 + mj2 must exist. The magnetic
quantum number of the total angular momentum is constrained by the maximum (anti-)parallel
alignment of Ĵ1 and Ĵ2 . It holds:

mj,max = mj1 ,max + mj2 ,max (7.69)

For each ji , (2ji + 1) states can be found, which differ only in their magnetic quantum number.
Considering a system composed of two angular momenta Ĵ1 and Ĵ2 with the quantum numbers
j1 and j2 , in the context of the product basis, we obtain the following total number of states:

p(j1 , j2 ) = (2j1 + 1)(2j2 + 1) (7.70)

The angular momenta Ĵ1 and Ĵ2 are independent of each other when viewed in the product
basis and each provides for itself a certain number of states (2j1 + 1) and (2j2 + 1). At the same
time, exactly the same number of states must exist in the coupled basis; for every possible value
of j between jmin ≤ j ≤ jmax , there should again be exactly (2j + 1) realizable states. We thus
need to verify that this is consistent with the total number of allowed values for j according to

177
 7 Spin

(7.70). For j1 ≥ j2 , it holds:

j=j1 +j2

p(j1 , j2 ) = (2j + 1) =
X

j=j1 −j2
2j2
=2 [2 (k + j1 − j2 ) + 1] =
X

k=0
2j2 2j2 n
n
=2 k+ (2j1 − 2j2 + 1) = k= (n + 1)
X X X

k=0 k=0 k=0

2
2j2
=2 (2j2 + 1) + (2j2 + 1) (2j1 − 2j2 + 1) =
2
= 2j2 (2j2 + 1) − (2j2 + 1) 2j2 + (2j2 + 1) (2j1 + 1) =
= (2j1 + 1)(2j2 + 1) □ (7.71)

As expected, we obtain the same number of realizable states in both the product basis and the
coupled basis. The total number of states remains conserved during the transformation into the
new basis!

Example: Product Basis and Coupled Basis for a System with j1 = 5

2 and j2 = 1

To illustrate the concept of the coupled basis, let’s consider a concrete example with j1 = 25
and j2 = 1. In the sense of the product basis (7.70), we obtain a total of (2· 52 +1)·(2·1+1) =
6 · 3 = 18 basis vectors, each with different quantum numbers {j, mj }. For the magnitude
quantum number j of the operator Ĵ2 , it holds according to (7.68):

5 3 7 5
−1 = ≤j ≤ = +1
2 2 2 2
The schematic representation below shows how the angular momenta can align in the
extreme cases jmax = j1 + j2 and jmin = |j1 − j2 |.

j1 j2
j2 j j1
j

j1 + j2 j1 − j2

Abb. 44: Maximum and minimum total angular momentum quantum number j.

We find all values for j by starting from jmin = 32 and repeatedly adding +1, until we reach
jmax = 72 . In total, the possible values for j are

3 5 7
 
j∈ , ,
2 2 2
All values for mj must now fall within the interval −j ≤ mj ≤ j for each j. Starting
at mj,min , successive increasing by +1 yields all possible values of the magnetic quantum
number. Since in this example there are three values for j, we need to repeat this procedure
three times for each individual j. Thus, we find all possible states in the coupled basis to

178
Quantum Theory I

be:
3 3 1 1 3
 
j= : mj ∈ − , − , ,
2 2 2 2 2
5 5 3 1 1 3 5
 
j = : mj ∈ − , − , − , , ,
2 2 2 2 2 2 2
7 7 5 3 1 1 3 5 7
 
j = : mj ∈ − , − , − , − , , , ,
2 2 2 2 2 2 2 2 2
We can also graphically identify all possible mj : Using mj1 and mj2 as coordinates, we
mark every pair {mj1 , mj2 } as shown in the left graphic of Figure 45. We find p(j1 , j2 ) = 18
distinct points, and by summing mj = mj1 + mj2 at each point, it immediately becomes
clear that all points on the diagonals have the same mj value.

mj2 j
7/2
1
5/2
0 mj1
3/2
-1
1/2
-5/2 -3/2 -1/2 1/2 3/2 5/2 mj
-7/2 -5/2 -3/2 -1/2 1/2 3/2 5/2 7/2

Abb. 45: (Left) Representation of all states in the product basis. (Right) The same system
is shown in the coupled basis.

We draw another coordinate system, as shown in the right graphic of Figure 45, this time
with the axes mj and j. Here, we can once again plot all 18 points: At the height of j = 23 ,
we plot horizontally all points from mj = − 32 to mj = + 32 ; at the height j = 52 , there are
horizontal points from mj = − 52 to mj = + 52 , and finally, at j = 27 , mj runs from − 72 to + 27 .
Both graphs can be directly transformed into each other: In the left graph (product basis),
the diagonal lines become vertical coordinate lines in the right graph (coupled basis) – a
strong indication that we can switch from the product basis to the coupled basis through
linear transformations. Except for mj = ± 72 , there are multiple values for j for every mj .

7.5.3 Clebsch-Gordan Coefficients

Let us consider a system whose eigenstates |l, ml ; s, ms ⟩ are known in the product basis for the
respective observables {L̂2 , L̂z ; Ŝ2 , Ŝz }. We want to represent the eigenstate |l, s, j, mj ⟩ in the
coupled basis {L̂2 , Ŝ2 , Ĵ2 , Jˆz } as a linear combination of the eigenstates of the product basis.
For this, we perform the following linear transformation:

|l, s, j, mj ⟩ = 1 |l, s, j, mj ⟩ = ⟨l, ml ; s, ms |l, s, j, mj ⟩ |l, ml ; s, ms ⟩ (7.72)

X

ml ,ms

Thus |l, s, j, mj ⟩ is expanded in the eigenfunctions |l, ml ; s, ms ⟩: The expansion coefficient

⟨l, ml ; s, ms |l, s, j, mj ⟩ is called the Clebsch-Gordan coefficient. For the eigenstates of the coupled
basis, the same applies as for the product basis in (7.59):

⟨l, s, j, mj |l, s, j ′ , m′j ⟩ = δll′ δss′ δjj ′ δmj m′j (7.73)

If we carry out the transformation explicitly, we must note that we keep the quantum numbers
l and s the same and only transform ml and ms into the corresponding j and mj . For (7.73) it

179
 7 Spin

can be further elaborated as follows:

δjj ′ δmj m′j = ⟨l, s, j, mj |1|l, s, j ′ , m′j ⟩ =

= ⟨l, s, j, mj |l, ml ; s, ms ⟩ ⟨l, ml ; s, ms |l, s, j ′ , m′j ⟩ =
X

ml ,ms

= ⟨l, s, j, mj |l, ml ; s, ms ⟩ ⟨l, s, j ′ , m′j |l, ml ; s, ms ⟩ (7.74)

X

ml ,ms

We use ⟨l, ml ; s, ms |l, s, j ′ , m′j ⟩ = ⟨l, s, j ′ , m′j |l, ml ; s, ms ⟩∗ = ⟨l, s, j ′ , m′j |l, ml ; s, ms ⟩ for real Clebsch-
Gordan coefficients. Associating ⟨l, s, j, mj |l, ml ; s, ms ⟩ with the real transformation matrix Aij
and forming the matrix product Aij Akj – with the same assumptions as in (7.74) we obtain:
!
Aij Akj = Aij ATjk = δik = 1 = AA−1
X X

j j

From the comparison with the transformation matrices A, we immediately recognize that the
transformation matrix from the product basis to the coupled basis is represented by an orthog-
onal matrix A−1 = AT .
An analogous derivation can be made for the transformation from the coupled basis to the prod-
uct basis. This gives the inverse transformation matrix (in this case (7.59) would apply). If we
already know A, we can obtain the inverse by simply transposing the matrix due to orthogonality!

So far, the calculation of the Clebsch-Gordan coefficients or the explicit transformation between
the two representations has been treated very abstractly: In the following examples, we will
take a closer look at the transition between the product and the coupled basis. We will deal
with very simple cases here, but the same procedures can easily be transferred to more complex
systems.

Example: Clebsch-Gordan coefficients for a system with s1 = 1

2 and s2 = 1
2

In this example, the angular momenta Ĵ1 and Ĵ2 , previously referred to generally, are the
spins Ŝ1 and Ŝ2 of two spin-1/2 particles. The product basis is thus {Ŝ21 , Ŝz,1 ; Ŝ22 , Ŝz,2 }, and
a corresponding eigenstate is generally represented as |s1 , ms,1 ; s2 , ms,2 ⟩ with four quantum
numbers. Since the spin quantum numbers s1 = 21 and s2 = 12 are fixed (and remain the
same in the coupled basis), we do not have to write them explicitly all the time, and we
can therefore indicate all states in the following, shortened notation:

|s1 , ms,1 ; s2 , ms,2 ⟩ ≡ |ms,1 ; ms,2 ⟩

We can indicate the product basis concisely as {Ŝz,1 , Ŝz,2 }. According to (7.70), the total
number of states is (2 · 21 + 1) · (2 · 21 + 1) = 2 · 2 = 4. The realizable states are thus given
by:
|ms,1 ; ms,2 ⟩ ∈ {|↑↑⟩ , |↑↓⟩ , |↓↑⟩ , |↓↓⟩} = {|ϕi ⟩}
(1)
Moreover, it should be noted that an operator from the Hilbert space Hs can only act on
(1)
the first state. For example, Ŝz acts as follows:
(7.25)
Ŝz(1) |ms,1 ; ms,2 ⟩ ≡ (Ŝz(1) ⊗ 1(2) )(|ms,1 ⟩ ⊗ |ms,2 ⟩) = ℏms,1 |ms,1 ; ms,2 ⟩
(2) (2)
Analogously, we can determine the effect for Ŝz , an operator from Hs . The other state
is thus not affected by the operator. But what if we act on a state from |ms,1 ; ms,2 ⟩ with

180
Quantum Theory I

(1) (2)
Ŝ2 and Ŝz ? Let’s start with Ŝz = Ŝz + Ŝz :

(1) (2) (1) (2) (7.29) ℏ

Ŝz |↑↑⟩ = (Ŝz + Ŝz ) |↑↑⟩ = Ŝz |↑↑⟩ + Ŝz |↑↑⟩ = 2 |↑↑⟩ + 2
ℏ
|↑↑⟩ = ℏ |↑↑⟩
(1) (2) (1) (2) (7.29) ℏ
Ŝz |↑↓⟩ = (Ŝz + Ŝz ) |↑↓⟩ = Ŝz |↑↓⟩ + Ŝz |↑↓⟩ = 2 |↑↓⟩ − ℏ2 |↑↓⟩ = 0
(1) (2) (1) (2) (7.29)
Ŝz |↓↑⟩ = (Ŝz + Ŝz ) |↓↑⟩ = Ŝz |↓↑⟩ + Ŝz |↓↑⟩ = − ℏ2 |↓↑⟩ + ℏ
2 |↓↑⟩ = 0
(1) (2) (1) (2) (7.29)
Ŝz |↓↓⟩ = (Ŝz + Ŝz ) |↓↓⟩ = Ŝz |↓↓⟩ + Ŝz |↓↓⟩ = − ℏ2 |↓↓⟩ − ℏ
2 |↓↓⟩ = −ℏ |↓↓⟩

We can represent the effect of Ŝz in the product basis as a matrix:

1 0 0 0
    
|↑↑⟩ |↑↑⟩
|↑↓⟩ {|ϕ ⟩} 0 0 0 0  |↑↓⟩
 
(7.75)
i
Ŝz   −−−→ ℏ 
   
|↓↑⟩ 0 0 0 0  |↓↑⟩
 

|↓↓⟩ 0 0 0 −1 |↓↓⟩

The blue-colored area of the matrix corresponds to a degenerate subspace. For the eigen-
states of the product basis |↑↓⟩ and |↓↑⟩, we find the eigenvalue 0 in both cases. A similar
degeneracy will also be found for the modulus operator Ŝ2 = (Ŝ(1) + Ŝ(2) )2 , where we must
first bring the operator Ŝ2 to a usable form:

Ŝ2 = (Ŝ(1) + Ŝ(2) )2 = Ŝ(1)2 + Ŝ(2)2 + 2Ŝ(1) Ŝ(2) =

 
= Ŝ(1)2 + Ŝ(2)2 + 2 Ŝx(1) Ŝx(2) + Ŝy(1) Ŝy(2) + Ŝz(1) Ŝz(2)
= Ŝ(1)2 + Ŝ(2)2 + 2Ŝz(1) Ŝz(2) + 2Σ

(1) (2) (1) (2)

The last term is problematic insofar as Ŝx Ŝx and Ŝy Ŝy will definitely lead to off-
diagonal elements. With the help of (7.39), we can rewrite Σ:
(7.39)
Σ = Ŝx(1) Ŝx(2) + Ŝy(1) Ŝy(2) =
1  (1) (1) 1
 
(2) (2)
 1  (1) (1) 1
 
(2) (2)

= Ŝ+ + Ŝ− Ŝ+ + Ŝ− + Ŝ+ − Ŝ− Ŝ+ − Ŝ− =
2 2 2i 2i
1  (1) (2) (1) (2) (1) (2) (1) (2)

= Ŝ Ŝ + Ŝ+ Ŝ− + Ŝ− Ŝ+ + Ŝ− Ŝ− +
2 + +
1 (1) (2) (1) (2) (1) (2) (1) (2)

+ − Ŝ+ Ŝ+ + Ŝ+ Ŝ− − Ŝ− Ŝ+ + Ŝ− Ŝ− =
2
1  (1) (2) (1) (2)

= Ŝ+ Ŝ− + Ŝ− Ŝ+
2

Plugging Σ back into our expression for Ŝ2 finally takes us to

(1) (2) (1) (2)
Ŝ2 = Ŝ(1)2 + Ŝ(2)2 + 2Ŝz(1) Ŝz(2) + Ŝ+ Ŝ− + Ŝ− Ŝ+ (7.76)

With (7.76) as well as (7.28) and (7.30), the effect of Ŝ2 on all product states results in:
(7.28,7.30) 3 2
Ŝ2 |↑↑⟩ = 4 ℏ |↑↑⟩ + 34 ℏ2 |↑↑⟩ + 2 ℏ2 ℏ2 |↑↑⟩ + 0 + 0 = 2ℏ2 |↑↑⟩
(7.28,7.30) 3 2
 
Ŝ2 |↑↓⟩ = 4 ℏ |↑↓⟩ + 34 ℏ2 |↑↓⟩ + 2 ℏ2 ℏ
2 |↑↓⟩ + 0 + ℏℏ |↓↑⟩ = ℏ2 |↑↓⟩ + ℏ2 |↓↑⟩
(7.28,7.30) 3 2
 
Ŝ2 |↓↑⟩ = 4 ℏ |↓↑⟩ + 34 ℏ2 |↓↑⟩ + 2 − ℏ2 ℏ
2 |↓↑⟩ + ℏℏ |↑↓⟩ + 0 = ℏ2 |↓↑⟩ + ℏ2 |↑↓⟩
(7.28,7.30) 3 2
  
Ŝ2 |↓↓⟩ = 4 ℏ |↓↓⟩ + 34 ℏ2 |↓↓⟩ + 2 − ℏ2 − ℏ2 |↓↓⟩ + 0 + 0 = 2ℏ2 |↓↓⟩

181
 7 Spin

If we represent Ŝ2 as a matrix in the product basis, we immediately recognize that we now
have to deal with off-diagonal elements – thus {|ms,1 ; ms,2 ⟩} are not eigenfunctions of Ŝ2 :

2 0 0 0
    
|↑↑⟩ |↑↑⟩
|↑↓⟩ {|ϕ ⟩} 0 1 1 0 |↑↓⟩
 
(7.77)
i
Ŝ2   −−−→ ℏ2 
   
|↓↑⟩ 0 1 1 0 |↓↑⟩
 

|↓↓⟩ 0 0 0 2 |↓↓⟩

(i)
This is a consequence of the non-vanishing commutator [Ŝ2 , Sz ] ̸= 0. The goal is now
to diagonalize Ŝ2 and Ŝz . We have already found two eigenstates of the coupled basis,
namely |↑↑⟩ and |↓↓⟩, as they are already diagonal for Ŝ2 and Ŝz ! However, the degenerate
subspace (in blue) from (7.77) still needs to be diagonalized.

Let’s first make some general considerations about the coupled basis {Ŝ2 , Ŝz }, whose eigen-
states we will denote as |S, Ms ⟩. The possible values of S are determined by the relation
(7.68):
1 1 1 1
|s1 − s2 | = − = 0 ≤ S ≤ 1 = + = s1 + s2
2 2 2 2
For each value of S, MS,max = S and MS,min = −S holds. In the case of this example, we
obtain four different realizations of an eigenstate in the coupled basis. This must of course
be so, as there are also four states in the product basis. The four states of the coupled
basis are:
|S, Ms ⟩ ∈ {|0, 0⟩ , |1, −1⟩ , |1, 0⟩ , |1, +1⟩}
The state with S = 0 is referred to as the singlet state, while we refer to the three states
with S = 1 as triplet states. A graphical representation allows a simple confirmation of
the form of the upper states in the coupled basis!

(2)
ms S
1/2 0
(1)
ms
−1/2 Ms
−1 1
−1/2 1/2

Abb. 46: (Left) Representation of the states of our system in the product basis. (Right)
These same states are represented in the coupled basis.

Each state of such a multiplet can be transferred into another state using a ladder operator.
Suppose, for example, that we repeatedly apply the lowering operator on the state with
the highest quantum number Ms,max , we will traverse all realizable states of the multiplet
until we reach the state with the lowest quantum number Ms,min .

Another possibility to generate |S, Ms ⟩ is a rotation of |ms,1 ; ms,2 ⟩ into the desired state
through the diagonalization of the degenerate subspace. Both options will be discussed in
the following.

Diagonalization of the Subspace Consider only the degenerate subspace from (7.77) and
diagonalize this. The subspace has the basis {|bi ⟩} = {|↑↓⟩ , |↓↑⟩}. The eigenvalues (divided
by ℏ) will correspond to the possible values of the quantum number S of the coupled basis

182
Quantum Theory I

– the eigenvalues Ms we already know from (7.75), they are in both cases Ms = 0. It holds:
! ! !
|↑↓⟩ {|bi ⟩} 1 1 |↑↓⟩
Ŝ2 −−−→ ℏ2 (7.78)
|↓↑⟩ 1 1 |↓↑⟩

The eigenvalues of the matrix from (7.78) are λ1 = 2ℏ and λ2 = 0, with which we can
calculate the following eigenvectors:

1 1 1
!
{|bi ⟩}
λ1 = 2ℏ : |1, 0⟩ = √ (|↑↓⟩ + |↓↑⟩) −−−→ √
2 2 1
1 1
!
{|bi ⟩} 1
λ1 = 0 : |0, 0⟩ = √ (|↑↓⟩ − |↓↑⟩) −−−→ √
2 2 −1

Following the eigenvectors, we can give the transformation matrix T from the product
basis to the coupled basis. The elements of this matrix correspond to the Clebsch-Gordan
coefficients! It holds:

1 1 1
! !
2 0
T =√ with Λ = (7.79)
2 1 −1 0 0

Creation Using Ladder Operators According to (7.77), we know that the states |↑↑⟩ and
|↓↓⟩ of the product basis {Ŝz,1 , Ŝz,2 } are equivalent to the eigenstates |1, +1⟩ and |1, −1⟩
of the coupled basis {Ŝ2 , Ŝz }. These two states belong to the multiplet with S = 1, which
contains 2S + 1 = 3 distinct states, and the third state must evidently be |1, 0⟩. However,
the representation of |1, 0⟩ and also |0, 0⟩ in states of the product basis is not as clear as
in the cases of |1, +1⟩ and |1, −1⟩, which represent the edges of the triplet.

In the triplet (S = 1), we can proceed as follows: We look for the state with the highest
quantum number (here MS = MS,max = S = 1), i.e., |1, 1⟩ = |↑↑⟩, and apply the lowering
operator from (7.26) to this coupled as well as product state:
√ 
Ŝ− |1, 1⟩ = ℏ 2 |1, 0⟩  1
=⇒ |1, 0⟩ = √ (|↑↓⟩ + |↓↑⟩)
Ŝ− |↑↑⟩ =
(1)
(Ŝ− +
(2)
Ŝ− ) |↑↑⟩ = ℏ (|↑↓⟩ + |↓↑⟩) 2

Thus, |1, 0⟩ consists of a linear combination of the mixed states, which is symmetric with
respect to the exchange of both particles. Reapplying brings us into the state with the
lowest quantum number (MS = −1) in the triplet:
√ 
Ŝ− |1, 0⟩ = ℏ 2 |1, 0⟩ 
(1) (2) √ =⇒ |1, −1⟩ = |↓↓⟩
Ŝ− √12 (|↑↓⟩ + |↓↑⟩) = √12 (Ŝ− + Ŝ− )(|↑↓⟩ + |↓↑⟩) = ℏ 2 |↓↓⟩

This gives us the „lowest“ state |1, −1⟩ of the triplet. The single state |0, 0⟩ of the singlet
(S = 0) can be generated by exploiting the orthogonality of the coupled basis states.
By assuming a general linear combination |0, 0⟩ = α |↑↓⟩ + β |↓↑⟩, we can determine the
coefficients α and β by projection on |1, 1⟩, |1, −1⟩, and |1, 0⟩. In fact, only |1, 0⟩ is relevant
(which lies with |0, 0⟩ in the degenerate subspace), from which it follows:

! 1 1
0 = ⟨1, 0|0, 0⟩ = √ (⟨↑↓| + ⟨↓↑|) (α |↑↓⟩ + β |↓↑⟩) = √ (α + β) =⇒ β = −α
2 2

183
 7 Spin

√ √
Since the state |0, 0⟩ must be normalized, we choose α = 1/ 2 and thus also β = −1/ 2.
By applying the ladder operators and the orthonormality condition, we have found a
complete set of states in the coupled basis and their representation in the product basis:



|0, 0⟩ = √12 (|↑↓⟩ − |↓↑⟩)

|1, 0⟩ = √1 (|↑↓⟩ + |↓↑⟩)

|S, Ms ⟩ = 2
|1, −1⟩ = |↓↓⟩



|1, +1⟩ = |↑↑⟩



√
The factors ±1/ 2 again correspond to the sought Clebsch-Gordan coefficients. Thus with
both calculation methods we equally find the following transformation:

1
! ! !
|1, 0⟩ 1 1 |↑↓⟩
=√
|0, 0⟩ 2 −1 1 |↓↑⟩

For the inverse transformation, the transposed transformation matrix can be used, as we
are dealing with a unitary transformation matrix:

1 1 −1
! ! !
|↑↓⟩ |1, 0⟩
=√
|↓↑⟩ 2 1 1 |0, 0⟩

Overall, for the action of Ŝ2 on the coupled eigenstates in the matrix representation, we
get:
|1, +1⟩ 2 0 0 0 |1, +1⟩
    
 |1, 0⟩  {|S,M ⟩} 0 2 0 0  |1, 0⟩ 
(7.80)
s
Ŝ2   −−−−−→ ℏ2 
    
 |0, 0⟩  0 0 0 0  |0, 0⟩ 
 

|1, −1⟩ 0 0 0 2 |1, −1⟩

As expected, there are no off-diagonal elements left in the coupled basis! At the same
time, Ŝz remains diagonal in the matrix representation of the coupled basis, because the
transformation was carried out in a degenerate subspace that diagonalizes Ŝz :

|1, +1⟩ 1 0 0 0 |1, +1⟩

    
 |1, 0⟩  {|S,M ⟩} 0 0 0 0   |1, 0⟩ 
 
(7.81)
s
Ŝz   −−−−−→ ℏ 
  
 |0, 0⟩  0 0 0 0   |0, 0⟩ 
 

|1, −1⟩ 0 0 0 −1 |1, −1⟩

In practice, Clebsch-Gordan coefficients do not have to be calculated individually for each system
of angular momenta. Usually, it is sufficient to look up the correct values in tables. A small
excerpt of such tables can be found in the lower three tables: Here, all coefficients for j1 × j2 =
2 × 2 , 1 × 2 , and 1 × 1 are listed. The tables are read as follows: The gray shaded values
1 1 1

(bottom right) correspond to the Clebsch-Gordan coefficients, where a shortened representation

is common: A value 1/2 thus means 1/2! To the left of the gray shaded areas are the values
p

of m1 and m2 in the product basis, and above the gray shaded area are the quantum numbers
J and mj of the coupled basis. In all three systems, the „highest“ and „lowest“ states (mjmax
and mjmin ) are the same in the product basis and the coupled basis.

1
1/2 × 1/2
+1 1 0
J
j1 × j2 +1/2 +1/2 1 0 0
mj
+1/2 −1/2 1/2 1/2 1
m1 m2 CG-coefficient
−1/2 +1/2 1/2 −1/2 −1
−1/2 −1/2 1

184
Quantum Theory I

3/2
1 × 1/2
+3/2 3/2 1/2
+1 +1/2 1 +1/2 +1/2
+1 −1/2 1/3 2/3 3/2 1/2
0 +1/2 2/3 −1/3 −1/2 −1/2
0 −1/2 2/3 1/3 3/2
−1 +1/2 1/3 −2/3 −3/2
−1 −1/2 1
2
1×1
+2 2 1
+1 +1 1 +1 +1
+1 0 1/2 1/2 2 1 0
0 +1 1/2 −1/2 0 0 0
+1 −1 1/6 1/2 1/3
0 0 2/3 0 −1/3 2 1
−1 +1 1/6 −1/2 1/3 −1 −1
0 −1 1/2 1/2 2
−1 0 1/2 −1/2 −2
−1 −1 1

Example: Time Evolution of Particles with l = 1 in the Anisotropic Crystal

Field
We consider a system of a free particle with angular momentum l = 1, which is in an
anisotropic crystal field. The Hamilton operator Ĥ thus includes not only the rotational
energy but also an energy component influenced by the external field. For Ĥ in this
example (we abbreviate here |l, m⟩ ≡ |m⟩) the following applies:

L̂2
ĤKF = + γ (|−1⟩ ⟨+1| + |+1⟩ ⟨−1| + |0⟩ ⟨0|) (7.82)
2I
I is in this case the moment of inertia of the particle, γ is the coupling to the crystal field.
We already know the eigenstates of the modulus operator of the angular momentum and
its action on these:
|1, m⟩ ≡ |m⟩ ∈ {|−1⟩ , |0⟩ , |+1⟩}
Applying Ĥ to the eigenstates of L̂2 , we obtain in the matrix representation the following
result:
2 0 0 0 0 1
        
|−1⟩ 2 |−1⟩
 ℏ 
ĤKF  |0⟩  =  0 2 0 + γ 0 1 0  |0⟩  (7.83)
     
2I
|+1⟩ 0 0 2 1 0 0 |+1⟩
The expression in the bracket is obviously not diagonal but leads to mixtures of |m⟩; the
eigenfunctions of L̂2 are therefore not eigenfunctions of ĤKF . We find these by diagonalizing
the matrix HKF – not only do we obtain the eigenfunctions of the Hamilton operator but
also its eigenvalues! It holds:

0
 
ℏ2 /I γ
HKF = 0 ℏ /I + γ
2 0 
 
γ 0 ℏ2 /I

The eigenvalues follow from det |HKF − λ1| = 0. In this case we obtain λ1 = ℏ2 /I − γ
and the doubly degenerate eigenvalue λ2,3 = ℏ2 /I + γ. For each eigenvalue, we find the

185
 7 Spin

respective eigenfunction |λi ⟩:

1 0 1
     
1   |−1⟩ − |+1⟩ 1   |−1⟩ + |+1⟩
|λ1 ⟩ = √  0  = √ , |λ2 ⟩ = 1 = |0⟩ , |λ3 ⟩ = √ 0 = √
 
2 −1 2 0 2 1 2

Compiling our eigenvectors into the transformation matrix T and the eigenvalues into a
diagonal matrix Λ, we obtain:

1 √0 1 0 0
   
ℏ2 /I − γ
1 
T = √  0 2 0 and Λ =  0 ℏ /I + γ
2 0 (7.84)
  
2 −1 0 1

0 0 ℏ2 /I + γ

We are now interested in the time evolution of any state of this system. Arbitrarily,
we choose |+1⟩ here and proceed according to (2.39) – the only time-dependent term
comes from the exponential function (or rather the time evolution operator, which we will
learn more rigorously in Quantentheorie II). We already know |ψ(t)⟩ at time t = 0:
|ψ(0)⟩ = |+1⟩. For arbitrary times t it holds:

|ψ(t)⟩ = eiĤt/ℏ |+1⟩

Without delving further into operator-valued functions: If the description is in the eigen-
basis of the Hamilton operator, Ĥ in the exponential function simply corresponds to the
eigenvalue from (7.84). However, we then also have to give |+1⟩ √ in the eigenbasis of Ĥ –
by subtracting |λ1 ⟩ from |λ3 ⟩, we obtain |+1⟩ = (|λ3 ⟩ − |λ1 ⟩) / 2. It follows:
1  
|ψ(t)⟩ = √ eiλ3 t/ℏ |λ3 ⟩ − eiλ1 t/ℏ |λ1 ⟩ =
2
1 h iλ3 t/ℏ i
= e (|−1⟩ + |+1⟩) − eiλ1 t/ℏ (|−1⟩ − |+1⟩) =
2
1 h i( ℏ + γ )t
e I ℏ − ei( I − ℏ )t |−1⟩ + ei( I + ℏ )t + ei( I − ℏ )t |+1⟩ =
ℏ γ   ℏ γ ℏ γ  i
=
2 
γ γ
    
i ℏI t
=e i sin t |−1⟩ + cos t |+1⟩ (7.85)
ℏ ℏ
The rotation of the phase is determined solely by the size of the moment of inertia I; the
coupling to the external field γ, on the other hand, controls in which „direction“ the state
|t⟩ points at any time t.

186
Quantum Theory I

8 Perturbation Theory

Motivation: Approximation Methods and Perturbation Theory in Quantum

Theory
The Schrödinger equation can only be solved exactly for very few, highly simplified sys-
tems. The potentials that are dealt with within the framework of this script (Coulomb
potential, harmonic oscillator) thus represent the exception and not the rule in terms of
their exact solvability. As will be shown in this chapter, the Schrödinger equation can
be approximately solved for potentials that slightly deviate from these exactly solvable
potentials.

The theory that describes such “slightly perturbed” potentials is perturbation theory, which
has established itself as a widespread method to compute approximate solutions for ana-
lytically not closed solvable quantum mechanical problems.

This chapter deals with the Rayleigh-Schrödinger perturbation theory. A central idea of this
approximation method is that the Hamiltonian operator Ĥ can be divided into an unperturbed
part Ĥ0 and a perturbation V̂ :
Ĥ = Ĥ0 + λV̂ (8.1)
The method is helpful when we can analytically solve the eigenvalue problem of the unperturbed
Hamiltonian Ĥ0 , but not the actual problem with the perturbed Hamiltonian Ĥ. We thus start
from the following initial situation:
Ĥ0 |ϕn ⟩ = εn |ϕn ⟩ (8.2)
If the order parameter λ is small enough, the perturbation V̂ has only a weak effect, and we
can approach the unknown solution for Ĥ in a series development from the known solution for
Ĥ0 . Although the choice of Ĥ0 and V̂ is not always clear, the following conditions must be
met to obtain the most accurate results with the least computational effort for perturbative
calculations:
The entire spectrum of eigenfunctions |ϕn ⟩ and eigenenergies εn of Ĥ0 can be calculated
exactly in the sense of (8.2).
λV̂ is a weak perturbation compared to Ĥ0 , with the order parameter λ scaling the strength
of the perturbation.
The last condition will be illustrated in an example. We assume that the eigenenergies are non-
degenerate – it follows that for all eigenenergies εn ̸= εn′ (∀n ̸= n′ ). It will later be clear why this
condition is an important prerequisite for the functioning of Rayleigh-Schrödinger perturbation
theory (in the case of degeneracy, one must use degenerate perturbation theory). Additionally,
only time-independent Hamiltonians will be examined in this chapter.

Our goal is now to determine the general eigenenergies En and wavefunctions |ψn ⟩ of the Hamil-
tonian Ĥ, which fulfill the perturbed Schrödinger equation:
Ĥ |ψn ⟩ = En |ψn ⟩ (8.3)
The index n represents the energetic hierarchy of the state (quantum number). Note: All
following derivations apply to a fixed n, such as for the ground state n = 0 with |ψ0 ⟩ and E0 ,
or any other (fixed) energy state |ψn ⟩ and En . Let’s start with the expansion of the desired
eigenfunction |ψn ⟩ over the (small) order parameter λ:
∞
|ψn ⟩ = |0⟩ + λ |1⟩ + λ2 |2⟩ + · · · = (8.4)
X
λi |i⟩
i=0

187
 8 Perturbation Theory

We assume that the functions {|i⟩} form an orthonormal system: ⟨i|j⟩ = δij . An analogous
series expansion is also established for the desired eigenenergy En of the perturbed Hamiltonian
Ĥ: ∞
En = E0 + λE1 + λ2 E2 + · · · = (8.5)
X
λi Ei
i=0

We currently know neither the expansion functions |i⟩ nor the expansion eigenenergies Ei . To
determine these, we proceed from the Schrödinger equation (8.3), and insert the above series
expansions:
∞ ∞ ∞
! ! !
  X
Ĥ0 + λV̂ = (8.6)
X X
λi |i⟩ λi Ei λi |i⟩
i=0 i=0 i=0
Since we want to expand in powers of λ, it is helpful for clarity to write out the first terms of
this equation explicitly:
     
Ĥ0 + λV̂ |0⟩ + λ |1⟩ + λ2 |2⟩ + . . . = E0 + λE1 + λ2 E2 + . . . |0⟩ + λ |1⟩ + λ2 |2⟩ + . . .

We simplify the expression above by multiplying out and reorganizing; for clarity, we group all
terms with the same powers of the order parameter λ, including at most the second order λ2 :
 
0 = λ0 Ĥ0 |0⟩ − E0 |0⟩ +
 
+ λ1 Ĥ0 |1⟩ + V̂ |0⟩ − E0 |1⟩ − E1 |0⟩ +
 
+ λ2 Ĥ0 |2⟩ + V̂ |1⟩ − E0 |2⟩ − E1 |1⟩ − E2 |0⟩ + O(λ3 ) (8.7)

Each term corresponding to each order n must individually vanish, otherwise the equation (8.7)
cannot be satisfied. Solving for each coefficient, we can compute correction terms of higher order
step by step.

8.1 Zero-Order Solution

To determine the zero-order solution, the crudest approximate solution, we consistently consider
only the zero order λ0 in (8.7), obtaining:

Ĥ0 |0⟩ = E0 |0⟩ (8.8)

(8.8) is determined only by the unperturbed Hamiltonian and thus corresponds to the system
without the presence of perturbation. A direct comparison with (8.2) immediately shows:

E0 = εn and |0⟩ = |ϕn ⟩ (8.9)

If only these results are considered in the series expansions (8.4) and (8.5), one recognizes:
Unsurprisingly, in zero order, the solutions |ψn ⟩ and En for the perturbed eigenvalue problem
are identical to the solutions |ϕn ⟩ and εn of the unperturbed Hamiltonian Ĥ0 ! For λ → 0, one
can write the following solution for Ĥ:

En ≈ εn and |ψn ⟩ ≈ |ϕn ⟩ (8.10)

8.2 First-Order Solution

Analogously, we now also proceed to determine the first correction term in the series expansions
(8.4) and (8.5). Considering the term related to λ1 from (8.7), we can reformulate it with respect
to |0⟩ and |1⟩, resulting in:    
Ĥ0 − E0 |1⟩ = E1 − V̂ |0⟩ (8.11)
| {z } | {z }
(I) (II)

188
Quantum Theory I

Projecting the two relations (I) and (II) from (8.11) onto the bra vector ⟨0|, we obtain a closed
expression for E1 , through which the first-order energy correction can be computed. Note
according to (8.9) that |0⟩ = |ϕn ⟩, leading to the relation Ĥ0 |0⟩ = E0 |ϕn ⟩. For the terms (I)
and (II), we find:

(I) | ⟨0|Ĥ0 − E0 |1⟩ = ⟨0|Ĥ0 |1⟩ − ⟨0|E0 |1⟩ = E0 ⟨0|1⟩ − E0 ⟨0|1⟩ = 0

(II) | ⟨0|E1 − V̂ |0⟩ = E1 ⟨0|0⟩ − ⟨0|V̂ |0⟩ = E1 − ⟨ϕn |V̂ |ϕn ⟩

Due to the hermiticity of Ĥ0 , the Hamiltonian operator can act to the left – the advantage of
this is that we know the action of Ĥ0 on |0⟩, but not on |1⟩. We thus find by equating (I) and
(II) a simple relation for E1 , for which holds (again with |0⟩ = |ϕn ⟩):

E1 = ⟨ϕn |V̂ |ϕn ⟩ (8.12)

By considering this energy correction along with the zero-order solution in the series development
(8.5), one obtains the solution for En in first order:

En ≈ εn + λ ⟨ϕn |V̂ |ϕn ⟩ (8.13)

To determine the first correction term in the series development (8.4) of |ψn ⟩, we now project an
arbitrary eigenstate of the unperturbed Hamiltonian ⟨ϕk | (thus any eigenstate except ⟨0| = ⟨ϕn |)
from the left onto (8.11):

⟨ϕk |Ĥ0 − E0 |1⟩ = ⟨ϕk |Ĥ0 |1⟩ − ⟨ϕk |εn |1⟩ = εk ⟨ϕk |1⟩ − εn ⟨ϕk |1⟩ = (εk − εn ) ⟨ϕk |1⟩ =
= ⟨ϕk |E1 − V̂ |0⟩ = E1 ⟨ϕk |ϕn ⟩ − ⟨ϕk |V̂ |ϕn ⟩ = − ⟨ϕk |V̂ |ϕn ⟩

Due to the orthogonality of eigenfunctions, the term ⟨ϕk |ϕn ⟩ disappears if k ̸= n. We have thus
found a relation for an (almost) arbitrary expansion coefficient ⟨ϕk |1⟩ of |1⟩ in the eigenbasis of
the unperturbed Hamiltonian:

⟨ϕk |V̂ |ϕn ⟩

⟨ϕk |1⟩ = for k ̸= n (8.14)
εn − εk

Obviously, we encounter a problem in the solution (8.14) due to the denominator if k = n –

our initial assumption that |ϕk ⟩ is not |ϕn ⟩ = |0⟩ is therefore justified afterward. Let’s now use
(8.14) to explicitly calculate |1⟩:
∞
(8.14) ⟨ϕk |V̂ |ϕn ⟩
|1⟩ = 1 |1⟩ = |ϕk ⟩ ⟨ϕk |1⟩ = |ϕn ⟩ ⟨ϕn |1⟩ + |ϕk ⟩ ⟨ϕk |1⟩ =
X X X
|ϕk ⟩
k=0 k̸=n k̸=n
εn − εk

Again, from |ϕn ⟩ = |0⟩, it follows that ⟨ϕn |1⟩ = ⟨0|1⟩ = 0 vanishes. The first expansion coefficient
for the eigenfunction |ψn ⟩ of the Hamiltonian Ĥ thus only encompasses the off-diagonal elements
k ̸= n; the smaller the distance between the neighboring energy levels εn and εk , the greater the
deviation from the unperturbed state |ϕk ⟩. It follows:

X ⟨ϕk |V̂ |ϕn ⟩

|1⟩ = |ϕk ⟩ (8.15)
k̸=n
εn − εk

By considering this correction along with the zero-order solution in the series development (8.4),
one obtains the following solution for |ψn ⟩ in first order:

189
 8 Perturbation Theory

X ⟨ϕk |V̂ |ϕn ⟩

|ψn ⟩ ≈ |ϕn ⟩ + λ |ϕk ⟩ (8.16)
k̸=n
εn − εk

8.3 Second-Order Solution

Lastly, to also determine the second correction terms in the series developments (8.4) and (8.5),
we again use the expression related to λ2 from (8.7). We obtain:
   
Ĥ0 − E0 |2⟩ = E1 − V̂ |1⟩ + E2 |0⟩ (8.17)
| {z } | {z }
(I) (II)

Analogous to the first order correction, we again project (8.17) onto the bra vector ⟨0|:

(I) | ⟨0|Ĥ0 − E0 |2⟩ = ⟨0|Ĥ0 |2⟩ − ⟨0|E0 |2⟩ = E0 ⟨0|2⟩ − E0 ⟨0|2⟩ = 0

(II) | ⟨0|E1 − V̂ |1⟩ + ⟨0|E2 |0⟩ = E1 ⟨0|1⟩ − ⟨0|V̂ |1⟩ + E2 ⟨0|0⟩ = E2 − ⟨0|V̂ |1⟩

We thus find that the second order energy correction depends on |1⟩. By substituting the
aforementioned expression, we obtain:

(8.9) ⟨ϕk |V̂ |ϕn ⟩

E2 = ⟨0|V̂ |1⟩ =
X
⟨ϕn |V̂ |ϕk ⟩
k̸=n
ε n − εk

Thereby, we find a valid expression for the second energy correction – we notice that unlike the
first energy correction (8.12), only the off-diagonal elements k ̸= n are considered here:
X | ⟨ϕk |V̂ |ϕn ⟩ |2
E2 = (8.18)
k̸=n
εn − εk

Including this second-order energy correction along with the first-order correction and the zero-
order solution in the series expansion (8.5), one obtains the complete solution for En in second
order:
X | ⟨ϕk |V̂ |ϕn ⟩ |2
En ≈ εn + λ ⟨ϕn |V̂ |ϕn ⟩ + λ2 (8.19)
k̸=n
εn − εk

To determine the second correction term in the series expansion of |ψn ⟩, we again project an
arbitrary eigenstate of the unperturbed Hamiltonian ⟨ϕk | (assuming again |ϕk ⟩ =
̸ |ϕn ⟩) from the
left onto (8.17):

⟨ϕk |Ĥ0 − E0 |2⟩ = ⟨ϕk |Ĥ0 |2⟩ − ⟨ϕk |εn |2⟩ = εk ⟨ϕk |2⟩ − εn ⟨ϕk |2⟩ = (εk − εn ) ⟨ϕk |2⟩ =
= ⟨ϕk |E1 − V̂ |1⟩ + ⟨ϕk |E2 |ϕn ⟩ = E1 ⟨ϕk |1⟩ − ⟨ϕk |V̂ |1⟩ + E2 ⟨ϕk |ϕn ⟩

Once more, we find an expression for an expansion coefficient ⟨ϕk |2⟩ in the eigenbasis of the
unperturbed system; however, this time, the first order correction terms E1 from (8.12) and |1⟩
from (8.15) appear. We write:

E1 ⟨ϕk |1⟩ − ⟨ϕk |V̂ |1⟩ (8.12) ⟨ϕk |V̂ |1⟩ − ⟨ϕn |V̂ |ϕn ⟩ ⟨ϕk |1⟩
⟨ϕk |2⟩ = = (8.20)
εn − εk εn − εk
As before, the initial assumption k ̸= n prevents the energy denominator from diverging. Ex-
plicitly, we find, analogously to the procedure for |1⟩, now for |2⟩:

(8.20) ⟨ϕk |V̂ |1⟩ − ⟨ϕn |V̂ |ϕn ⟩ ⟨ϕk |1⟩

|2⟩ = 1 |2⟩ = |ϕn ⟩ ⟨ϕn |2⟩ + |ϕk ⟩ ⟨ϕk |2⟩ =
X X
|ϕk ⟩
k̸=n k̸=n
εn − εk

190
Quantum Theory I

Specifically, the expression for |2⟩ results in the following:

X ⟨ϕk |V̂ |1⟩ − ⟨ϕn |V̂ |ϕn ⟩ ⟨ϕk |1⟩

|2⟩ = |ϕk ⟩ (8.21)
k̸=n
εn − εk

For |1⟩, in applying the formula, expression (8.15) must still be inserted. This results in the
following approximation solution for |ψn ⟩ in second order:

X ⟨ϕk |V̂ |1⟩ − ⟨ϕn |V̂ |ϕn ⟩ ⟨ϕk |1⟩

|ψn ⟩ ≈ |ϕn ⟩ + λ |1⟩ + λ2 |ϕk ⟩ (8.22)
k̸=n
εn − εk

Example: Perturbed 2 × 2 Matrix

We consider the Hamiltonian operator Ĥ of an unspecified two-dimensional system in a

complete basis {|ϕi ⟩}:
! ! !
{|ϕi ⟩} A1 W A1 0 0 W
Ĥ −−−→ = + (8.23)
W ∗ A2 0 A2 W∗ 0
| {z } | {z }
H0 V

It is supposed to hold that A1 < A2 and W ∈ C! By decomposing the matrix into diagonal
and off-diagonal elements, it is evident that the two eigenvalues of the unperturbed system
Ĥ0 are given by ε1 = A1 and ε2 = A2 .
f (x)

2nd order
exact solution

Abb. 47: For small |W | or large distances between the energy levels A1 ≫ A2 , the second-
order correction and the exact solution agree very well.

According to (8.8), this is simultaneously the estimate for the eigenenergies of the perturbed
(0) (0)
system in zero order, which we want to write here as E1 and E2 (corresponding to the
energy E0 in the previous sections where we always considered only one energy level).
According to (8.8), the corresponding eigenvectors of the unperturbed system, |ϕ1 ⟩ and
|ϕ2 ⟩, are simultaneously the estimation of the eigenvectors of the perturbed system in zero
(0) (0)
order, which we want to write here as |E1 ⟩ and |E2 ⟩ (corresponding to the vector |0⟩ in

191
 8 Perturbation Theory

the previous sections where we always considered only one energy level). It follows:
!
(0) {|ϕi ⟩} 1
|E1 ⟩ = |ϕ1 ⟩ −−−→ at ε1 = A1
0
! (8.24)
(0) {|ϕi ⟩} 0
|E2 ⟩ = |ϕ2 ⟩ −−−→ at ε2 = A2
1

Here, we want to apply perturbation theory up to the second order, and thus can accord-
(0) (1) (2)
ingly infer from (8.5): Ei ≈ Ei + Ei + Ei (the order parameter λ is evidently equal
to 1 in our example according to (8.23)). But what do the eigenvalues E1 and E2 of the
entire system Ĥ look like when only a small perturbation is present? For the first-order
(1)
correction Ei , we obtain:
! !
(1) (8.12) (8.24)
  0 W 1
E1 = ⟨ϕ1 |V̂ |ϕ1 ⟩ = 1 0 =0
W∗ 0 0
! ! (8.25)
(1) (8.12) (8.24)
  0 W 0
E2 = ⟨ϕ2 |V̂ |ϕ2 ⟩ = 0 1 =0
W∗ 0 1

The first-order correction vanishes! We will directly see that this is not the case for the
second-order correction – the following result is found:
!2
1
!
(2) (8.18) | ⟨ϕ2 |V̂ |ϕ1 ⟩ |2 (8.24)
  0 W 1 |W |2
E1 = = 0 1 =
ε1 − ε2 A1 − A2 W ∗ 0 0 A1 − A2
!2 (8.26)
1
!
(2) (8.18) | ⟨ϕ1 |V̂ |ϕ2 ⟩ |2 (8.24)
  0 W 0 |W |2
E2 = = 1 0 =
ε2 − ε1 A2 − A1 W ∗ 0 1 A2 − A1

By summing the results (8.24), (8.25), and (8.26), we obtain the energy eigenvalues of Ĥ
according to second-order Rayleigh-Schrödinger perturbation theory:

|W |2 |W |2
E1 ≈ A1 − and E2 ≈ A2 + (8.27)
A2 − A1 A2 − A1

Exact Solution The simple form of the perturbed Hamiltonian Ĥ would indeed also allow
for an exact solution by directly solving the eigenvalue problem det |Ĥ − 1E| = 0:

0 = det |H − 1E| = (A1 − E)(A2 − E) − |W |2 = E 2 − E(A1 + A2 ) + A1 A2 − |W |2

Solving this quadratic equation leads√us to the following expression, which we can expand
up to the second order using f (x) = 1 + x2 ≈ 1 + x2 /2:
s
2
A1 + A2 A1 − A2 2|W |

E1,2 = ± 1+ ≈
2 2 A1 − A2
A1 + A2 A1 − A2 2|W |2
!
≈ ± 1+ (8.28)
2 2 (A1 − A2 )2

The exact solution developed up to the second order ((8.28) (depending on whether one
evaluates it for E1 or E2 ) is completely identical with the result (8.27) of the perturbation
calculation! In the figure below, the approximation for f (x) with x = 2|W |/(A1 − A2 ) is
graphically illustrated.

192
Quantum Theory I

9 Concepts of Quantum Theory

In this chapter, we deal with advanced concepts, or rather, we try to understand very fundamen-
tal principles of quantum theory. We also delve into specific technical realizations of quantum
mechanical phenomena, such as the quantum computer or quantum cryptography.

9.1 Series of Stern-Gerlach Apparatuses

In Chapter 7, we got to know the Stern-Gerlach apparatus. In the following, we want to
illustrate some of the fundamental properties of quantum mechanics using multiple Stern-Gerlach
apparatuses arranged in sequence.

Simple Stern-Gerlach Apparatus In Figure 48, the following experimental setup is sketched:
A stream of uncharged spin- 21 particles moving in the x direction passes through a Stern-Gerlach
apparatus with an arbitrary magnetic field angle ϑ (measured in the yz-plane, relative to the
positive z direction). We assume that the particles emitted from the source (on the left) are
polarized in the z direction and are in an initial state |ψ⟩ = |↑⟩. The particles leaving the Stern-
Gerlach apparatus on the lower path and thus in the state |↓ϑ ⟩ (with the energy eigenvalue
λ↓ = −ℏ/2) are blocked. The particles that leave the Stern-Gerlach apparatus on the upper
path and thus are in the state |↑ϑ ⟩ (with λ↑ = +ℏ/2) can pass and are detected. What is the
probability that a particle hits our detector?
B = Bn

Abb. 48: The number of particles originally prepared in the state |↑⟩ that leave the Stern-
Gerlach apparatus, set at an arbitrary angle ϑ, in the state |↑ϑ ⟩ (upper path) should
be determined. The lower path (state |↓ϑ ⟩) is blocked.

The particles leaving the Stern-Gerlach apparatus are in the following state according to (7.55)
(we assume that φ = 0):
(7.55)
|↑ϑ ⟩ = cos(ϑ/2) |↑⟩ + sin(ϑ/2) |↓⟩ for λ↑ = +ℏ/2 (9.1)

The probability that the measurement of the particle with initial state |↑⟩ subsequently results
in the state |↑ϑ ⟩ is:
(9.1)
P (↑, ↑ϑ ) = | ⟨↑ | ↑ϑ ⟩ |2 = | ⟨↑| (cos(ϑ/2) |↑⟩ + sin(ϑ/2) |↓⟩) |2 =
= | cos(ϑ/2)⟨↑ | ↑⟩ + sin(ϑ/2)⟨↑ | ↓⟩|2 = cos2 (ϑ/2)

Naturally, due to particle conservation, with P (↑, ↓ϑ ) = sin2 (ϑ/2) (as the probability of obtaining
the measurement value −ℏ/2), the relation P (↑, ↑ϑ ) + P (↑, ↓ϑ ) = cos2 (ϑ/2) + sin2 (ϑ/2) = 1 must
be fulfilled. As a reminder: The concept of probability here should be understood such that we
can specify the following expected value ⟨N ′ ⟩ in a multitude of equivalent experiments conducted
with individual particles (i.e., the source generates a beam of a total of N particles):

⟨N ′ ⟩ = N cos2 (ϑ/2) (9.2)

193
 9 Concepts of Quantum Theory

Double Stern-Gerlach Apparatus Let’s now consider the more complicated case where two
Stern-Gerlach apparatuses with measurement angles ϑ and ϑ′ , respectively, are placed in se-
quence. Again, it should be the case that the original particle beam is polarized in the z
direction (|ψ⟩ = |↑⟩). From both apparatuses, only the states |↑ϑ ⟩ and |↑ϑ′ ⟩, corresponding to
the +ℏ/2 eigenvalues, are allowed through (see Figure 49). We have:
|↑ϑ ⟩ = cos(ϑ/2) |↑⟩ + sin(ϑ/2) |↓⟩ for λ↑ = +ℏ/2
(9.3)
|↑ϑ′ ⟩ = cos(ϑ′ /2) |↑⟩ + sin(ϑ′ /2) |↓⟩ for λ′↑ = +ℏ/2

B = Bn B = Bn′

Abb. 49: Two Stern-Gerlach apparatuses connected in series with different magnetic field orien-
tations ϑ and ϑ′ . In both apparatuses, only the states corresponding to the eigenvalues
+ℏ/2 are allowed through, and the states corresponding to the eigenvalues −ℏ/2 are
each blocked.

The probability that the state |ψ⟩ = |↑⟩ transitions to the state |↑ϑ ⟩ after the first Stern-Gerlach
apparatus is (as already calculated) given by:
P (↑, ↑ϑ ) = ⟨P̂↑ϑ ⟩ = | ⟨↑ | ↑ϑ ⟩ |2 = cos2 (ϑ/2)
The projector P̂↑ϑ ensures that every particle is in the state |↑ϑ ⟩ after passing the first Stern-
Gerlach apparatus (the other particle beam is blocked). We now want to calculate the probability
that the particle is in the state |↑ϑ′ ⟩ after the second apparatus:
(9.3)
P ′ (↑ϑ , ↑ϑ′ ) = | ⟨↑ϑ | ↑ϑ′ ⟩ |2 =

2
= (cos (ϑ/2) ⟨↑| + sin (ϑ/2) ⟨↓|) cos ϑ′ /2 |↑⟩ + sin ϑ′ /2 |↓⟩ =

′ ′
= | cos(ϑ/2) cos(ϑ /2)⟨↑ | ↑⟩ + sin(ϑ/2) sin(ϑ /2)⟨↓ | ↓⟩+
+ cos(ϑ/2) sin(ϑ′ /2)⟨↑ | ↓⟩ + sin(ϑ/2) cos(ϑ′ /2)⟨↓ | ↑⟩ |2 =
2
= cos(ϑ/2) cos(ϑ′ /2) + sin(ϑ/2) sin(ϑ′ /2) =


′
= cos ((ϑ − ϑ)/2)
2
(9.4)
For general angles ϑ and ϑ′ , we thus expect that with initially N particles, ultimately ⟨N ′′ ⟩
particles in the state |↑ϑ′ ⟩ will hit the detector:
ϑ ϑ′ − ϑ
   
′′ ′ (9.2,9.5)
⟨N ⟩ = N · P (↑, ↑ϑ )P (↑ϑ , ↑ϑ′ ) = N cos 2
cos2 (9.5)
2 2
Analogous to a series of optical polarization filters, we can set the Stern-Gerlach apparatuses
so that, for example, at the end, we measure no particles anymore. In the case that the first
Stern-Gerlach apparatus allows the incoming beam completely through (ϑ = 0) and the second
apparatus is aligned exactly orthogonal to it (ϑ′ = π), such an extinction would occur. However,
if the first Stern-Gerlach apparatus is rotated to ϑ = π/2 (thus letting only half of the incoming
beam through), while the second apparatus remains unchanged (ϑ′ = π), the total probability
P = 0.25 for transmission results. Interestingly, an additional filter can thus lead to greater
transmission!

We can understand the result as the successive realization of a final state; the “path” to the
final state is then divided into “partial paths”: The probability of measuring the final state
corresponds to the product of the probabilities of the individual part steps.

194
Quantum Theory I

9.2 Axioms of Quantum Theory

Let’s now attempt to formulate some of the rules of quantum mechanics for calculating measur-
able predictions in terms of axioms:

1.) Information Content: The state of a physical system is fully characterized by a state
vector |ψ⟩. All information about a system is thus contained in the state vector. However,
this applies only to pure states; for mixed states, the entire information about the system
is contained in the density operator ρ̂ = |ψ⟩ ⟨ψ| (see Quantum Theory II).

2.) Observable: A physical observable A, i.e., a real measurable quantity, always corresponds
to a hermitian operator  = † . Such operators act on state vectors  |ψ⟩ to access the
information contained therein.

3.) Measurement Value: The eigenvalues an of an operator  are the only possible mea-
surement values of the observable A. The eigenvalue equation  |ϕn ⟩ = an |ϕn ⟩ follows,
where we call |ϕn ⟩ an eigenstate of  with eigenvalue an .

4.) Measurement Probability: Given an arbitrary state |ψ⟩, the probability of obtaining
the measurement value ân of the observable A can be given via the projection onto the
corresponding eigenstate pn = | ⟨ϕn |ψ⟩ |2 = ⟨ψ|P̂n |ψ⟩. The state |ψ⟩ is mapped to a state
|ϕn ⟩ using the projection operator P̂n = |ϕn ⟩ ⟨ϕn |. A measurement leads to the “collapse
of the wave function” onto |ϕn ⟩.

5.) Time Evolution: The time evolution of a state |ψ(t)⟩ is determined by the Schrödinger
equation iℏ |ψ̇(t)⟩ = Ĥ |ψ(t)⟩.

6.) Classical Analogue: Operators with classical analogues are determined by canonical
quantization by replacing the canonical coordinates ri and pi with complementary op-
erators that satisfy the commutation relation [r̂i , p̂j ] = iℏδij . Operators whose classical
analogues are functions of ri and pi are symmetrized in the order of the non-commuting
operators.

Calculation Rules for Probabilities If we assume a completely prepared system, the following
calculation rules for probabilities can be noted (according to Dietrich Grau, Übungsaufgaben zur
Quantentheorie, V5.52, 2020):

(1) Final States: The probability for a “final state” is given by the absolute square of a
complex number; namely, the probability amplitude of the final state. A final state refers
to the distinguishable, alternative outcomes of the experiment under consideration.

(2) Alternative Realizations: If there are several, within the experiment, fundamentally
indistinguishable alternative ways (“virtual paths”) to achieve a final state, the probability
amplitude for reaching that final state is the sum of the probability amplitudes for the
relevant virtual paths (interference).

(3) Successive Alternative Realizations: If one of the virtual paths consists of several
“sub-paths”, the probability amplitude for the entire virtual path is the product of the
probability amplitudes for the sub-paths.

(4) Events: If multiple final states are grouped into a “event”, the probability for this event
is given by the sum of the probabilities for the respective final states.

195
 9 Concepts of Quantum Theory

1 2 3 4
i i i i

f5 f4 f3 f2 f1 f f f2 f1

Abb. 50: Schematic representation for calculating probabilities of experiment outcomes: (1)
different final states, (2) alternative realizations, (3) successive realization, and (4)
events from multiple final states.

9.3 The Measurement Process

The measurement process plays a central role in quantum mechanics, since we not only intervene
explicitly in a quantum system but also understand the measurement process as a means of
preparing the system. In the following, we will discuss different terms for quantum states |ψ⟩
(which we will revisit in the Quantum Theory II in the context of the density operator) and
also touch on the most frequently “killed” animal in the history of physics – Schrödinger’s cat.

9.3.1 Separable and Entangled States

Consider two systems HA and HB , which interact with each other and can be described together
in HA+B . For the individual systems, we find the corresponding observables  and B̂, as well
as a complete orthonormal basis {|ai ⟩} and {|bi ⟩}, respectively. These bases are however not
the eigenbases of the operators  and B̂; it should only be that  acts in HA , while B̂ acts
exclusively in HB . If a state |ψ⟩ from HA+B can be written as follows, we call this state separable:
X  X 
|ψ⟩ = αi |ai ⟩ ⊗ βj |bj ⟩ = |ψa ⟩ ⊗ |ψb ⟩ with αi , βj ∈ C (9.6)
i j

For the coefficients of such a product state of the linear superpositions of the states |ψa ⟩ and
|ψb ⟩, it holds that i |αi |2 = j |βj |2 = 1. If such a representation as a sum of product states
P P

(|ai ⟩ ⊗ |bj ⟩ ≡ |ai bj ⟩) is not possible, we call the state entangled. Such a state is given in full
generality as follows:
|ϕ⟩ = γij |ai ⟩ ⊗ |bj ⟩ with γj ∈ C (9.7)
X

ij

The state |ϕ⟩ is not separable. The separability criterion is only met if all γij can be represented
as γij = αi βj , making an explicit separation between |ψa ⟩ and |ψb ⟩ possible. A clear separation
of γij into αi and βj (where i, j = 1, . . . , N ) is in principle only possible by solving a N 2 -large
system of equations. Whether this solution exists can be verified by writing all coefficients γij
in the form of a matrix and then checking if rank(γij ) = 1 holds. If instead rank(γij ) > 1, then
the state is entangled.
It follows that in the case of separability, the matrix γij = αi βj corresponds to the outer product
of the vectors, α = (α1 , α2 , · · · , αN )⊺ and β = (β1 , β2 , · · · , βN )⊺ , and a matrix formed by α ⊗ β
must always have a rank of 1.
This method is not limited to two-state systems and can be generalized to arbitrarily large
systems (NH > 2).

Example: Separability of a Two-Particle System

We consider a system of two interacting electrons i = 1, 2 with spins {|↑⟩i , |↓⟩i }. To
decide whether two different states are separable or entangled, we examine the rank of the

196
Quantum Theory I

matrix formed from the coefficients γij . First, we generally compare a separable state in a
representation like (9.6) with a representation like (9.7):

|ψ⟩ = (α↑ |↑⟩1 + α↓ |↓⟩1 ) ⊗ (β↑ |↑⟩2 + β↓ |↓⟩2 ) =

= α↑ β↑ |↑↑⟩ + α↓ β↓ |↓↓⟩ + α↑ β↓ |↑↓⟩ + α↓ β↑ |↓↑⟩
|ϕ⟩ = γ↑↑ |↑↑⟩ + γ↓↓ |↓↓⟩ + γ↑↓ |↑↓⟩ + γ↓↑ |↓↑⟩

A comparison of |ψ⟩ and |ϕ⟩ should allow a clear assignment of the coefficients γij with
αi and βj in the form γij = αi βj for a separable state. Summarized as a matrix, we can
better understand the decomposition into the outer product:
! ! !
γ↑↑ γ↑↓ α↑ β↑ α↑ β↓ α↑  
= = β↑ β↓ (9.8)
γ↓↑ γ↓↓ α↓ β↑ α↓ β↓ α↓

At this point, it also becomes clear why a matrix formed by α ⊗ β must have a rank of
1: The first and second columns are linearly dependent on each other. The first column
corresponds to the second column multiplied by a factor β↑ /β↓ .

Separable State As a simple example, consider the following state:

|ψ1 ⟩ = |↑⟩1 ⊗ |↑⟩2 ≡ |↑↑⟩

In the context of (9.6), |ψ1 ⟩ = 1 |↑↑⟩+0 |↓↓⟩+0 |↑↓⟩+0 |↓↑⟩ applies. The coefficients matrix
according to (9.7) has rank one and can therefore be written as an outer product:
! ! !
1 0 1 0 1  
Rank = 1 =⇒ = 1 0
0 0 0 0 0

The state is thus separable, only the coefficients α↑ = β↑ = 1 and as a direct consequence
γ↑↑ = 1 are non-zero.

Entangled State Now consider the following state:

1
|ψ2 ⟩ = √ (|↑↓⟩ − |↓↑⟩)
2
The coefficients matrix according to (9.7) has rank two and therefore cannot be written as
the outer product of two coefficient vectors α and β:

0 √1
!
Rank 2 = 2 =⇒ entangled state
− √12 0

We infer that the state |ψ2 ⟩ is not separable but entangled.

Expectation Values for Separable States We now want to determine the expectation value of
 for a general separable state according to (9.6) in the system HA+B , where both αi ̸= 0 and
βj ̸= 0 should hold. We need to note that  acts only on states from HA ! The construction of

197
 9 Concepts of Quantum Theory

⟨Â⟩ leads us to:

X  X 
⟨ψ|Â|ψ⟩ = αi∗ βj∗ ⟨ai | ⊗ ⟨bj | Â αk βl |ak ⟩ ⊗ |bl ⟩ =
ij kl

= βj∗ βl ⟨bj |bl ⟩ αi∗ αk ⟨ai |Â|ak ⟩ =

X X

jl ik
X 
= βj∗ βl δjl αi∗ αi ⟨ai |Â|ai ⟩ + αi∗ αk = |βj |2 = 1
X XX X
⟨ai |Â|ak ⟩
jl i k i̸=k j

= |αi | ⟨ai |Â|ai ⟩ + αi∗ αk (9.9)

X XX
2
⟨ai |Â|ak ⟩
i k i̸=k

We do not know the explicit action of  on any arbitrary state |ai ⟩ in this general example, so
⟨ai |Â|ak ⟩ must remain the final result here. However, we know that  does not act on |bj ⟩, which
allowed us to pull the scalar product ⟨bj |bl ⟩ to the front. The calculation of the expectation value
in (9.9) leads us to diagonal terms (i = k), which can be interpreted as classical probabilities, as
well as off-diagonal terms (i ̸= k), which we call interference terms. Due to these terms, we speak
of a coherent superposition – the coherence of the states allows the formation of interferences in
analogy to optics.

Expectation Values for Entangled States However, we have already recognized that not every
state is separable. Therefore, let’s calculate the expectation value ⟨Â⟩ again for a general
entangled state as described in (9.7):
X  X 
∗
⟨ϕ|Â|ϕ⟩ = γij ⟨ai bj | Â γkl |ak bl ⟩ =
ij kl
∗
= γkl ⟨bj |bl ⟩ ⟨ai |Â|ak ⟩ =
XX
γij
ij kl
∗
= γkl δjl ⟨ai |Â|ak ⟩ =
XX
γij
ij kl
∗ ∗
= ⟨ai |Â|ak ⟩ = γkj = ⟨ak bj |ϕ⟩ , γij = ⟨ϕ|ai bj ⟩
XX
γkj γij
ij k

= ⟨ak bj |ϕ⟩ ⟨ϕ|ai bj ⟩ ⟨ai |Â|ak ⟩ =

XX

ij k

= ⟨ak | ⟨bj |ϕ⟩ ⟨ϕ|bj ⟩ |ai ⟩ ⟨ai |Â|ak ⟩ = |ai ⟩ ⟨ai | = 1

XX X

ij k i

= ⟨ak | ⟨bj |ϕ⟩ ⟨ϕ|bj ⟩ Â|ak ⟩ =

XX

j k
n o
= TrA TrB {|ϕ⟩ ⟨ϕ|} Â (9.10)

No interference term appears here, i.e., it is an incoherent superposition or a statistical mixture.

We will learn about the formalism of the density operator ρ̂ in Quantum Theory II, which can
sensibly describe statistical mixtures.

Example: Coherence of a Two-Particle System

Let’s discuss the coherence behavior of a two-particle system using the states |ψ1 ⟩ and |ψ2 ⟩
more precisely. Let |ψ1 ⟩ be the most general two-particle product state with the realizable
states {|↑a ⟩ , |↓a ⟩} for the first particle and {|↑b ⟩ , |↓b ⟩} for the second particle. We find the

198
Quantum Theory I

following wavefunction according to (9.6):

|ψ1 ⟩ = (α1 |↑a ⟩ + α2 |↓a ⟩) ⊗ (β1 |↑b ⟩ + β2 |↓b ⟩)

We calculate the expectation value of the observable  of the system HA . However, it

must be noted that  can only act on the first particle. It follows for the quantity ⟨Â⟩:

⟨Â⟩ = ⟨ψ1 |Â|ψ1 ⟩ =

= [(α1∗ ⟨↑a | + α2∗ ⟨↓a |) ⊗ (β1∗ ⟨↑b | + β2∗ ⟨↓b |)] Â [(α1 |↑a ⟩ + α2 |↓a ⟩) ⊗ (β1 |↑b ⟩ + β2 |↓b ⟩)] =
= (α1∗ ⟨↑a | + α2∗ ⟨↓a |) Â (α1 |↑a ⟩ + α2 |↓a ⟩) · (β1∗ ⟨↑b | + β2∗ ⟨↓b |) (β1 |↑b ⟩ + β2 |↓b ⟩) =
 
= α1∗ α1 ⟨↑a |Â| ↑a ⟩ + α2∗ α2 ⟨↓a |Â| ↓a ⟩ + α1∗ α2 ⟨↑a |Â| ↓a ⟩ + α2∗ α1 ⟨↓a |Â| ↑a ⟩ ·
(β1∗ β1 ⟨↑b | ↑b ⟩ + β2∗ β2 ⟨↓b | ↓b ⟩ + β1∗ β2 ⟨↑b | ↓b ⟩ + β2∗ β1 ⟨↓b | ↑b ⟩) =
 
= |α1 |2 ⟨↑a |Â| ↑a ⟩ + |α2 |2 ⟨↓a |Â| ↓a ⟩ + α1∗ α2 ⟨↑a |Â| ↓a ⟩ + α2∗ α1 ⟨↓a |Â| ↑a ⟩ ·
 
|β1 |2 + |β2 |2 =
=|α1 |2 ⟨↑a |Â| ↑a ⟩ + |α2 |2 ⟨↓a |Â| ↓a ⟩ + α1∗ α2 ⟨↑a |Â| ↓a ⟩ + α2∗ α1 ⟨↓a |Â| ↑a ⟩
| {z }
Interference terms

The result corresponds exactly to the previously derived equation (9.9), and we recognize
that the product state generates interference terms. But how does this look in the case of
an entangled state |ψ2 ⟩? We define the following wavefunction:

|ψ2 ⟩ = α |↑a ⟩ ⊗ |↓b ⟩ + β |↓a ⟩ ⊗ |↑b ⟩ ≡ α |↑↓⟩ + β |↓↑⟩

The rank of the coefficients matrix is 2; the state is therefore not separable. We calculate
the expectation value of the same observable  as before, which acts only on the first
particle.

⟨Â⟩ = ⟨ψ2 |Â|ψ2 ⟩ =

= (α∗ ⟨↑↓| + β ∗ ⟨↓↑|) Â (α |↑↓⟩ + β |↓↑⟩) =
= α∗ α ⟨↑a |Â| ↑a ⟩ ⟨↓b | ↓b ⟩ + β ∗ β ⟨↓a |Â| ↓a ⟩ ⟨↑b | ↑b ⟩ +
+ α∗ β ⟨↑a |Â| ↓a ⟩ ⟨↓b | ↑b ⟩ + β ∗ α ⟨↓a |Â| ↑a ⟩ ⟨↑b | ↓b ⟩ =
=|α|2 ⟨↑a |Â| ↑a ⟩ + |β|2 ⟨↓b |Â| ↓b ⟩

Obviously, there are no off-diagonal elements; a state completely entangled with the system
HB cannot interfere in the system HA .

9.3.2 Schrödinger’s Cat and the Limit of Quantum Mechanics

Motivation: An Explanation for the Behavior of Macroscopic Objects.

The question arises why in everyday life with macroscopic objects, we cannot observe
the same phenomena as with microscopic quantum particles. For example, electrons sent
through two closely spaced slits form the known interference pattern of the double-slit ex-
periment (which can be explained through the interference terms described above), while
such an effect has never been observed with tennis balls sent through two slits.

If, however, quantum physics is supposed to be a fundamental, universal theory, then it

must also be able to describe the behavior of macroscopic objects such as tennis balls in

199
 9 Concepts of Quantum Theory

the limiting case – after all, these too are ultimately made of atoms that obey the laws
of quantum physics. But how come that the conglomeration of atoms forming two tennis
balls apparently never manifests in a product state? An answer to this is provided by the
effect of decoherence.

The transition from a product state, which exhibits interference terms during measurement,
to an entangled state, which no longer has interference terms during a measurement, is called
decoherence. Typically, decoherence occurs when a quantum mechanical system interacts with
its environment. It leads to an entanglement with the particles of the environment (which are
not measured), resulting in the loss of interference properties in the particles of the object to be
measured. This means that in everyday macroscopic objects – which inevitably interact with the
environment – a measurement can be described using classical probabilities. This limiting factor
of quantum states also plays a role in the paradox of Schrödinger’s Cat. Erwin Schrödinger
wrote in 1935 in his essay “The Present Situation in Quantum Mechanics”:
“One can even set up quite ridiculous cases. A cat is penned up in a steel cham-
ber, along with the following diabolical device (which must be secured against direct
interference by the cat): In a Geiger counter there is a tiny bit of radioactive sub-
stance, so small that perhaps in the course of an hour one of the atoms decays, but
also, with equal probability, perhaps none; if it happens, the counter tube discharges
and through a relay releases a hammer that shatters a small flask of hydrocyanic
acid. If one has left this entire system to itself for an hour, one would say that
the cat still lives if meanwhile no atom has decayed. The first atomic decay would
have poisoned her. The ψ-function of the entire system would express this by having
in it the living and the dead cat (pardon the expression) mixed or smeared out in
equal parts. Typical instances of this kind are actually found quite frequently in
the valid interpretation of quantum mechanics. These rather naive examples have
the disadvantage of allowing an incorrect interpretation which leaps unnoticed across
the precarious edge of direct observation. This prevents us from naively accepting
a ’blurred model’ as a reflection of reality. In itself, it would not contain anything
unclear or contradictory. There is a difference between a blurred or out-of-focus
photograph and a photograph of clouds and wafts of mist.”

R.I.P.

Abb. 51: Schrödinger’s hell machine: Only when opening the box can the observer verify
whether the cat inside is alive (left) or dead (right).

The answer to the problem posed by Schrödinger is provided by the phenomenon of decoher-
ence: In macroscopic objects (such as a cat), which consist of a very large number of quantum
particles, it becomes increasingly difficult to impossible to shield these from their environment.
Interactions inevitably occur between the particles of the cat and the environment. This in-
evitably leads to a state of decoherence where the observable quantum properties practically

200
Quantum Theory I

disappear. A superposition of a live and dead cat is therefore indeed meaningless because the
coherence condition is no longer fulfilled from the outset due to decoherence.

9.4 Bell’s Inequalities and Hidden Variables

Motivation: Completeness of Quantum Physics

Let us imagine the following experiment: A particle source between the physicists Alice and
Bob generates two entangled spin- 12 particles, which fly apart in opposite directions, one
towards Alice and the other towards Bob. These entangled spin- 12 particles are produced
through the spontaneous decay of spin-0 particles – due to the conservation of angular
momentum, the spin of the spin- 12 particles must always be opposite. That means, if Alice
and Bob measure along the same axis, they will always find an opposite spin, regardless
of the measurement direction they have agreed upon. It holds that:
1
|ψ⟩ = √ (|↑⟩A |↓⟩B − |↓⟩A |↑⟩B ) (9.11)
2

This representation is done in the eigenbasis of Ŝz ; it is immediately apparent that the spin
of the particles is anti-correlated in the z-direction. However, it turns out that this is also
the case when performing a basis transformation along any other measurement axis – see
for example (9.38), where this state was converted into the x-basis. Note: The minus sign
in (9.11) is important because the same state with a plus sign is no longer anti-correlated
in every direction!

Quantum mechanics now tells us the following: Before the measurement, the spin of both
particles is completely indeterminate, no matter in which direction one measures it. In a
much more fundamental sense, the spin direction before measurement is not existent at
all. If Alice measures the spin of the first particle on the left side of the laboratory in any
direction n, she will randomly obtain either the result |↑n ⟩A or |↓n ⟩A . Only through this
measurement does she define reality and fix the spin of “her” particle to the measurement
result. Simultaneously, her measurement causes the second particle, which has by then
arrived on the right side of the laboratory, to instantly adopt the opposite state. If, for
example, Alice measured |↑n ⟩A , Bob can be sure to measure the state |↓n ⟩B – even if his
measurement occurs so quickly after Alice’s measurement that the information could not
have traveled from Alice to Bob at the speed of light. It appears as if Alice’s measurement
influenced the spin of Bob’s particle faster than the speed of light! This seemed to con-
tradict relativity theory to Einstein, and he therefore rejected this notion. He called the
described effect “spooky action at a distance”.

In 1935, Albert Einstein published, together with Boris Podolsky and Nathan
Rosen („EPR“), a much-noted article. In it, the hypothesis is proposed that quantum
physics is incomplete. Essentially, it concerns the measurement of complementary ob-
servables which, according to the Heisenberg uncertainty principle, cannot be measured
simultaneously, or not even simultaneously exist as real. EPR took position and momen-
tum as an example; here we want to follow the argument based on the spin measurement
on two orthogonal axes, according to the revised version of the EPR experiment by David
Bohm.

We again assume the entangled state (9.11). Suppose Alice measures the spin of her parti-
cle as |↑⟩A in the z-direction, and Bob measures the spin of his (entangled) particle as |↓x ⟩B
in the x-direction. But what would have happened if Alice had also measured her parti-

201
 9 Concepts of Quantum Theory

cle in the x-direction, like Bob? EPR argue: Had Alice also decided on the x-direction,
then she would have measured |↑x ⟩A , because the particles are anti-correlated in every
measurement direction. Thus, according to EPR, Alice knows the spin in the z-direction
(a measurement she performed), but she also apparently knows the spin in the x-direction
because she can infer from Bob’s measurement how her measurement would have turned
out in the x-direction.
This would mean that Alice and Bob circumvented the uncertainty relation, because the
operators Sz and Sx do not commute and a simultaneous spin measurement in these or-
thogonal directions should therefore not be possible according to quantum physics. With
this thought experiment, as EPR argue, it would be proven that the spin is very much
part of physical reality in both the z-direction as well as in the x-direction.

EPR assume in their argumentation that a valid physical theory must fulfill the following
assumptions:

Reality assumption: If, without disturbing a system, a quantity can be predicted

with certainty, then there exists an element of physical reality that can be associated
with that quantity.

Locality assumption: Based on two systems; if both systems do not interact with
each other, a measurement on one system does not change the state of the other
system.

Since Alice must make the decision about the measurement axis only shortly before measur-
ing the first particle, this measurement can, assuming locality, have no disturbing influence
on the reality of the second particle. Nevertheless, with her measurement, she can always
deduce the spin of the second particle along the chosen measurement axis. From this,
EPR conclude that both the spin in the z-direction and the spin in the x-direction must be
part of physical reality, and therefore, according to EPR, quantum mechanics is incomplete.

For a long time, these problems were considered unanswerable by physics and seemed more
an issue for metaphysics. But then, in 1964, the physicist John Stewart Bell published
an article that made experimental verification of this problem possible.

9.4.1 Hidden Variables and Spooky Action at a Distance

Einstein, Podolsky, and Rosen argue in their work that the apparent paradoxes of quantum
mechanics could be solved with so-called hidden variables. This is to mean the following: Alice
and Bob conduct their spin measurements along different measurement axes, namely along the
z-axis (ϑ = 0, basis {|↑⟩ , |↓⟩}), at a 45° angle (ϑ = π/4, basis {|↗⟩ , |↙⟩}), and furthermore
along the x-axis (ϑ = π/2, basis {|→⟩ , |←⟩}).

Each particle already carries with it at its creation in so-called hidden variables λ the infor-
mation within itself of how the measurement outcome will be along each measurement axis. A
specific particle could for example have the hidden variables λ = (↓, ↗, →); thereby it would be
simultaneously determined that one would get the result |↑⟩ when measuring in the z-direction,
the result |↗⟩ when measuring at a 45°-direction, and the result |→⟩ when measuring in the
x-direction. Also in the entanglement of two particles, there would then be nothing mysterious
anymore. The two entangled spin- 12 particles are indeed created at a common point and simply
carry since their creation the opposite hidden variables within themselves. If Alice’s particle
contains the information λA = (↓, ↗, →), then Bob’s entangled twin particle has initially the
opposite information λB = (↑, ↙, ←). For the correlation of measurement results between Alice
and Bob, “spooky action at a distance” is no longer necessary.

202
Quantum Theory I

9.4.2 Bell’s Inequality

We can now consider one of the particles in our experiment (for example Alice’s particle) and
enter all possible state combinations of hidden variables in Table 1. Alice and Bob each measure
with the bases {|↑⟩ , |↓⟩} (ϑ = 0), {|↗⟩ , |↙⟩} (ϑ = π/4), and {|→⟩ , |←⟩} (ϑ = π/2). Further-
more, we can arrange the different state combinations in a set-theoretical Venn Diagram – see
Figure 52. In this diagram, three circles (A, B, and C) each symbolize the three sets.

Set A includes all states of the hidden variables where Alice, when measuring in the z-
direction, gets the state |↑⟩.

Set B includes all states of the hidden variables where Alice, when measuring at a 45°-
direction, gets the state |↗⟩.

Set C includes all states of the hidden variables where Alice, when measuring in the x-
direction, gets the state |→⟩.

B C
3 6 7
5
2 4

Abb. 52: Venn Diagram for a better understanding of Bell’s Inequality. It is assumed that
the measurement results are predetermined by hidden variables. In set A are all
particles whose measurement in the z-direction yields |↑⟩, in set B are all particles
whose measurement in a 45°-direction yields |↗⟩, and in set C are all particles whose
measurement in the x-direction yields |→⟩.

A particle in the state with the hidden variables λA = (↑, ↙, ←) belongs, for example, only to
set A, but not to set B and C. It thus lies in area 1 of the Venn Diagram. A particle in the state
λA = (↑, ↗, ←), on the other hand, belongs to both set A and set B (but not to set C), and
thus lies in the overlap area 2 of the Venn Diagram.
If Alice had the possibility of measuring the spin of every particle in all three directions, then
with the help of Table 1 she could uniquely assign every measurement result to an area within
the Venn Diagram in Figure 52. But exactly this simultaneous measurement is forbidden by
quantum physics. Alice can merely do the following: She can measure the spin of her particle
along one axis, and Bob can measure the spin of the entangled particle along another axis.
Alice measures, for example, along the z-axis and gets the result |↑⟩. So, she assumes that the
first entry of her “hidden variable” looks like this: λA = (↑, ∗, ∗). The second and third entries
she does not know after her measurement, which we symbolize with a ∗. But if Bob now also
conducts a measurement on his entangled particle, for instance in the direction ϑ = π/4, and
gets the result |↗⟩, then Alice deduces that she would have measured oppositely |↙⟩ if she
had chosen the same measurement direction as Bob. She thus knows at least two of the three
“hidden variables” of her particle: λA = (↑, ↙, ∗).

This can be useful! For instance, all states λA = (↑, ↙, ∗) are contained in the subset “A
without B” (in other words, the intersection of A and the inverse set of B; formally: A ∧ B̄) in

203
 9 Concepts of Quantum Theory

hidden set area

variables A B C in the Diagram
↑ ↗ → × × × 5
↑ ↗ ← × × 2
↑ ↙ → × × 4
↑ ↙ ← × 1
↓ ↗ → × × 6
↓ ↗ ← × 3
↓ ↙ → × 7
↓ ↙ ← –

Table 1: Possible states of the assumed hidden variables in Alice’s particle, and their assignment
in the Venn Diagram in Figure 52.

the Venn Diagram, which covers areas 1 and 4. If we restrict ourselves to the case where Alice
only measures along ϑ = 0 and ϑ = π/4 and Bob measures along ϑ = π/4 and ϑ = π/2, we are
left with the non-trivial areas B ∧ C̄ and A ∧ C̄, for which we can make similar considerations.

subset Alice Bob Alice’s hid- area

0◦ 45◦ 45◦ 90◦ den variables in the Diagram
A ∧ B̄ |↑⟩A |↗⟩B (↑, ↙, ∗) 1∪4
B ∧ C̄ |↗⟩A |→⟩B (∗, ↗, ←) 2∪3
A ∧ C̄ |↑⟩A |→⟩B (↑, ∗, ←) 1∪2

Table 2: Joint measurements by Alice and Bob allow the determination of 2 out of 3 (hypothet-
ical) hidden variables of Alice’s particle and thus the assignment to certain subareas
in the Venn Diagram.

Alice and Bob can conduct a certain number of measurements (for example 1000 measurements)
along the axes 0° (Alice) and 45° (Bob), then the same number of measurements along the axes
45° (Alice) and 90° (Bob), and finally the same number of measurements along the axes 0°
(Alice) and 90° (Bob). Whenever the measurements agree with the result from Table 2 (i.e.,
“spin-up” for both measurements in the respective measurement basis), it is noted as a hit for
the respective subset. In the end, Alice is able to specify the number of hits N for the respective
subsets.

If the hidden variables are real, then each particle, for which Alice has noted a hit, is in principle
(in objective reality) either to be assigned to areas 1, 2, 3, or 4 in the Venn Diagram, and there
is an (objective) hit count N (1), N (2), N (3), and N (4) for each area. Since these numbers can
only be positive, it must trivially hold:

N (1) + N (2) + N (3) + N (4) ≥ N (1) + N (2) (9.12)

Alice doesn’t know the individual values, but based on Table 2, she can easily conclude:

N (A ∧ B̄) = N (1) + N (4)

N (B ∧ C̄) = N (2) + N (3)
N (A ∧ C̄) = N (1) + N (2)

Plugging this into (9.12), you obtain Bell’s inequality:

204
Quantum Theory I

N (A ∧ B̄) + N (B ∧ C̄) ≥ N (A ∧ C̄) (9.13)

We can now actually compare this inequality with the predictions of quantum theory, and also
with experiments using Table 2! It turns out that with an entangled state like in (9.11), the
probability that Alice and Bob measure “up-spin” in their respective measurement basis depends
on the difference angle ∆ϑ of the two measurement bases as follows:

1 ϑ2 − ϑ1 1 ∆ϑ
   
P (↑ϑ1 ; ↑ϑ2 ) = sin2 = sin2 (9.14)
2 2 2 2

Example: Measurement Probability of Alice and Bob

The probability that Alice and Bob both measure “spins-up” along arbitrary measurement
axes ϑ1 and ϑ2 is to be derived. We define the first measurement direction (of Alice) as
the z-axis, i.e., ϑ = 0. We are allowed to do this because the entangled state (9.11) has
no preferred direction: No matter in which direction you measure, the particles are always
anti-correlated. If you express the Bell state (9.11) in any basis {|↑ϑ ⟩ , |↓ϑ ⟩}, the result
is again |ψ⟩ = √12 (|↑ϑ ⟩A |↓ϑ ⟩B − |↓ϑ ⟩A |↑ϑ ⟩B ). Bob’s rotated measurement direction then
corresponds to the angle ∆ϑ of the original problem.
Let’s first calculate the probability P1 that Alice measures the state |↑⟩A :

(9.11)
P1 = ⟨ψ|P̂↑A 1B |ψ⟩ =
1
= (⟨↑|A ⟨↓|B − ⟨↓|A ⟨↑|B ) (|↑⟩A ⟨↑|A ) (|↑⟩A |↓⟩B − |↓⟩A |↑⟩B ) =
2
1
= (⟨↑|A ⟨↓|B − ⟨↓|A ⟨↑|B ) (|↑⟩A ⟨↑ | ↑⟩A |↓⟩B + |↑⟩A ⟨↑ | ↓⟩A |↑⟩B ) =
2
1
= (⟨↑|A ⟨↓|B − ⟨↓|A ⟨↑|B ) |↑⟩A |↓⟩B =
2
1 1
= (⟨↑ | ↑⟩A ⟨↓ | ↓⟩B − ⟨↓ | ↑⟩A ⟨↑ | ↓⟩B ) = (9.15)
2 2
It’s not a surprise as the spin before the measurement is indefinite in every direction –
Alice therefore measures the state |↑⟩ with a probability of 50%. After the measurement
of Alice, the state |ψ⟩ collapses and becomes the state |ψ̃⟩.
1
|ψ̃⟩ = P̂↑A 1B |ψ⟩ = √ |↑⟩A ⟨↑|A (|↑A ⟩ |↓⟩B − |↓⟩A |↑⟩B ) =
2
1 1
= √ |↑⟩A ⟨↑ | ↑⟩A |↓⟩B − |↑⟩A ⟨↑ | ↓⟩A |↑⟩B = √ |↑⟩A |↓⟩B
2 2

The new state must still be normalized, however; we thus get for |ψ̃⟩:

|ψ̃⟩ = |↑⟩A |↓⟩B (9.16)

After Alice has measured the state |↑⟩A at her particle, the state of Bob’s particle is thus
fixed at |↓⟩B (along the same measurement direction). That is precisely the effect that
Einstein called “spooky action at a distance”! Now we only need to calculate with what
probability P2 Bob measures the state |↑ϑ ⟩ with the |ψ̃⟩ state after Alice’s measurement,

205
 9 Concepts of Quantum Theory

according to (9.16):

ϑ ϑ ϑ ϑ
          
(7.55)
P̂↑ϑ = |↑ϑ ⟩ ⟨↑ϑ | = cos
|↑⟩ + sin |↓⟩ cos ⟨↑| + sin ⟨↓| =
2 2 2 2
ϑ ϑ ϑ ϑ
       
= cos2 |↑⟩ ⟨↑| + sin2 |↓⟩ ⟨↓| + sin cos (|↑⟩ ⟨↓| + |↓⟩ ⟨↑|) (9.17)
2 2 2 2

This allows us to calculate the state 1A P̂↑ϑB |ψ̃⟩ after Bob’s measurement (under the
condition that Alice has measured |↑⟩A previously):

(9.16) (9.17)
|Ψ⟩ = 1A P̂↑ϑB |ψ̃⟩ = P̂↑ϑB |↑⟩A |↓⟩B =
2 ϑ 2 ϑ
     
= cos |↑⟩B ⟨↑|B + sin |↓⟩B ⟨↓|B |↑⟩A |↓⟩B +
2 2
ϑ ϑ
     
+ sin cos (|↑⟩B ⟨↓|B + |↓⟩B ⟨↑|B ) |↑⟩A |↓⟩B =
2 2
2 ϑ 2 ϑ
   
= cos |↑⟩A |↑⟩B ⟨↑ | ↓⟩B + sin |↑⟩A |↓⟩B ⟨↓ | ↓⟩B
2 2
ϑ ϑ
    
+ sin cos |↑⟩A |↑⟩B ⟨↓ | ↓⟩B + |↑⟩A |↓⟩B ⟨↑ | ↓⟩B
2 2
ϑ ϑ ϑ
     
= sin 2
|↑⟩A |↓⟩B + sin cos |↑⟩A |↑⟩B (9.18)
2 2 2

One sees that also Bob’s measurement changes the state to |Ψ⟩. Now we can calculate the
probability P2 that Bob measures his particle in the state |↑θ ⟩B , provided that Alice had
previously measured at her particle |↑⟩A :

(9.16,9.18)
P2 = ⟨ψ̃|Ψ⟩ =
ϑ ϑ ϑ
       
= ⟨↑|A ⟨↓|B sin 2
|↑⟩A |↓⟩B + sin cos |↑⟩A |↑⟩B
2 2 2
2 ϑ ϑ ϑ 2 ϑ
       
= sin ⟨↑ | ↑⟩A ⟨↓ | ↓⟩B + sin cos ⟨↑ | ↑⟩A ⟨↓ | ↓⟩B = sin (9.19)
2 2 2 2

The joint probability P (↑; ↑ϑ ) is now the product of P1 and P2 :

1 ϑ
 
(9.15,9.19)
P (↑; ↑ϑ ) = P1 P2 = sin2
2 2

As already argued in the beginning, this is the same probability P (↑ϑ1 , ↑ϑ2 ), that Alice
measures |↑ϑ1 ⟩ and Bob |↑ϑ2 ⟩:

1 ∆ϑ
 
P (↑ϑ1 ; ↑ϑ2 ) = sin2 (9.20)
2 2

If we still follow the orientation according to Table 2, it becomes apparent that:

for determining N (A ∧ B̄), Alice had to measure in the 0° direction, and Bob in the 45°
direction. Therefore, in this case, ∆ϑ = 45◦ = π/4.

for determining N (B ∧ C̄), Alice had to measure in the 45° direction, and Bob in the 90°
direction. Here, ∆ϑ = 45◦ = π/4.

for determining N (A ∧ C̄), Alice had to measure in the 0° direction, and Bob in the 90°
direction. Here, ∆ϑ = 90◦ = π/2.

206
Quantum Theory I

With this, we can compare the Bell inequality (9.13) with the quantum mechanically predicted
measurement result:
1 π/4 1 π/4 1 π/2
     
sin2 + sin2 ≥ sin2 =⇒ 0.1464 ≥ 0.25
2 2 2 2 2 2
This is obviously wrong! Thus, quantum theory clearly cannot be reconciled with the idea of
hidden variables (at least not if they act locally only). In fact, this prediction of quantum
theory was first verified in 1972 by Stuart Freedman and John Clauser, and thereafter
by Alain Aspect and many other experimenters in increasingly sophisticated experiments. It
indeed seems that Einstein was wrong in this case, and that in quantum physics, some sort of
measurement (the “observation”) creates reality: Before the measurement, Alice’s particle has
no definite spin in any direction, and after the measurement, the particle indeed has a definite
spin along the measurement direction but not in the orthogonal direction.

Although a measurement on one particle in an entangled state also collapses the other particle,
no information can be transmitted between the particles in this way. Thus, no information is
transmitted faster than the speed of light in keeping with relativity theory.

Deepening: Formal Derivation of Bell’s Inequalities

We define a continuous hidden variable λ (previously λ could only take three positions),
which should satisfy the following relation with an appropriate distribution function ρ(λ):
Z
dλ ρ(λ) = 1 (9.21)

We now define – with reference to the hidden variable λ – the expected correlation for
different measurement directions m and n (as well as p). Here, the correlation of the
measurements in any directions E(m, n) is to be understood as the product of the Pauli
matrices σ̂1 (n) and σ̂2 (m) (or spin operators):
Z
E(m, n) = dλ ρ(λ) ⟨σ̂1 (m)σ̂2 (n)⟩λ

Due to the locality assumed by EPR, it should also be valid that the spin measurements
are independent of each other and (after production) there is no interaction between the
particles:
⟨σ̂1 (m)σ̂2 (n)⟩λ = ⟨σ̂1 (m)⟩λ ⟨σ̂2 (n)⟩λ
Following the assumption of reality, it follows that the measurement of the spin gives the
observable quantity ⟨σ̂i ⟩ (n) ± 1 – and that for each measurement direction n (by using
the Pauli matrices, the usual factor ℏ/2 is omitted here). The spins of both particles are
always oriented oppositely due to our problem setting; it follows through E(m, m) = −1:

⟨σ̂1 (m)⟩λ = − ⟨σ̂2 (m)⟩λ

We can now summarize this for the expected correlation of the two measurements:
Z
E(m, n) = − dλ ρ(λ) ⟨σ̂1 (m)⟩λ ⟨σ̂1 (n)⟩λ (9.22)

Calculating the difference between two expected correlations, with n now shifted to p, we

207
 9 Concepts of Quantum Theory

can furthermore write with ⟨σ̂1 (m)⟩2λ = (±1)2 = 1:

Z
E(m, n) − E(m, p) = dλ ρ(λ) (⟨σ̂1 (m)⟩λ ⟨σ̂1 (p)⟩λ − ⟨σ̂1 (m)⟩λ ⟨σ̂1 (n)⟩λ ) =
Z  
= dλ ρ(λ) ⟨σ̂1 (m)⟩λ ⟨σ̂1 (n)⟩2λ ⟨σ̂1 (p)⟩λ − ⟨σ̂1 (m)⟩λ ⟨σ̂1 (n)⟩λ =
Z  
= dλ ρ(λ) ⟨σ̂1 (m)⟩λ ⟨σ̂1 (n)⟩λ ⟨σ̂1 (n)⟩λ ⟨σ̂1 (p)⟩λ − 1

However, we are only interested in the absolute difference of the correlations; because the
product of two expectation values | ⟨σ̂1 (m)⟩λ ⟨σ̂1 (n)⟩λ | ≤ 1 must be (at most we get ±1 as
an expectation value), we can perform the following estimate:
Z
|E(m, n) − E(m, p)| ≤ dλ ρ(λ) (1 − ⟨σ̂1 (n)⟩λ ⟨σ̂1 (p)⟩λ ) =
Z Z
(9.22)
= dλ ρ(λ) − dλ ρ(λ) ⟨σ̂1 (n)⟩λ ⟨σ̂1 (p)⟩λ = 1 + E(n, p)

Simple rearrangement thus leads us to the Bell inequality, which must be satisfied for a
locally-realistic theory with hidden variables:

1 + E(n, p) − |E(m, n) − E(m, p)| ≥ 0 (9.23)

Correlation between two operators For the expectation value of the correlation between
the spin operators σ̂1 (m) and σ̂2 (n), we find the following expression:

E(m, n) = ⟨ψ|σ̂1 (m)σ̂2 (n)|ψ⟩ = −mn = − cos(∆φ) (9.24)

We show this explicitly using the state |ψ⟩ = √12 (|↑↓⟩ − ↓↑) from (9.11) and the Pauli
matrices for arbitrary measurement directions according to (7.54). The evaluation takes
place in the matrix representation of σ̂i , hence for example ⟨↑ |σ̂| ↑⟩1 = cos(ϑ1 ):

(9.11)
E(m, n) = ⟨ψ|σ̂1 (m)σ̂2 (n)|ψ⟩ =
1
= (⟨↑↓| − ⟨↓↑|) σ̂1 σ̂2 (|↑↓⟩ − |↓↑⟩) = | e.g. ⟨↑ |σ̂| ↑⟩1 = ⟨↑ |σ̂| ↑⟩2
2
= ⟨↑ |σ̂| ↑⟩1 ⟨↓ |σ̂| ↓⟩2 − ⟨↑ |σ̂| ↓⟩1 ⟨↓ |σ̂| ↑⟩2 =
= cos(φ1 )(− cos(φ2 )) − sin(φ1 ) sin(φ2 ) = − cos(φ1 − φ2 ) = − cos(∆ϑ)

Now, if we explicitly substitute E(m, n) from (9.24) into Bell’s inequality (9.23), the fol-
lowing results:
1 − np − |mp − mn| ≥ 0
We understand m, n and p as direction vectors, where ϕ1 is the angle between m and n
and ϕ2 is the angle between m and p. Thus, for Bell’s inequality we find:

1 − cos(ϕ2 − ϕ1 ) − | cos(ϕ2 ) − cos(ϕ1 )| ≥ 0 (9.25)

This inequality is not satisfied for all angles ϕ1 and ϕ2 and thus violates (9.23)!

9.5 Technological Applications

In the following section, some technological applications such as random generators, quantum
computers, and quantum encryption will be discussed.

208
Quantum Theory I

9.5.1 Random Number Generators

For computers, generating random numbers is a great challenge. Most often, algorithms only
generate so-called pseudorandom numbers. The sequences of numbers generated by such al-
gorithms seem random but are not truly random since they are generated by a deterministic
algorithm. If a specific random number algorithm is initialized with the same starting value
(“seed”) every time, the same sequence of pseudorandom numbers will result each time.

However, through quantum theory, it is now possible for us to actually generate sequences of
true random numbers based on the superposition of a two-state system. Let’s look at Figure 48.
A source emits particles, which have been polarized in the z-direction and are in the state |↑⟩.
These electrons, prepared in such a way, subsequently move through a Stern-Gerlach apparatus,
whose magnetic field is rotated by ∆ϑ = π/2 relative to the prepared polarization direction.
The Stern-Gerlach apparatus is thus de facto a measuring device that measures the spin in the
x-direction. This causes the electrons prepared in the z-direction to be in a superposition of
the states |↑x ⟩ and |↓x ⟩ from the perspective of the measurement basis of the Stern-Gerlach
apparatus:
1
|↑⟩ = √ (|↑x ⟩ + |↓x ⟩)
2
The apparatus allows only the transmission of the |↑x ⟩ state; and that with a probability of
P (↑, ↑x ) = 0.5. Thus, with the same probability, the electron does not pass the Stern-Gerlach
apparatus! If we now assign the transmission of the electron the value |↑x ⟩ = |0⟩ and the
opposite state |↓x ⟩ = |1⟩, we generate |0⟩ and |1⟩ with an evenly distributed probability. The
randomness is inherent here – we cannot predict whether the initial state |↑⟩ will collapse into
|0⟩ or |1⟩. Thus, an absolutely random sequence of zeros and ones can be generated, which can
be interpreted by a computer program as a perfect sequence of random numbers.

9.5.2 Quantum Computers

While a classical computer works with the classical bits 0 and 1, a quantum computer uses
qubits (or Q-bits); that is, a system with two distinct quantum states |0⟩ and |1⟩. Apart from
the notation, this does not seem to be a big difference at first glance. However, the quantum
mechanical states |0⟩ and |1⟩ are capable of forming a superposition:

|ψ⟩ = α |0⟩ + β |1⟩

Furthermore, we can, depending on our computer system, entangle states. Analogue to the
classical computer, arbitrary numbers are to be represented in binary. In Table 3, the coding of
numbers 0 to 7 using three classical bits and three entangled qubits is shown.

class. entangled states

0 000 |0⟩ ⊗ |0⟩ ⊗ |0⟩ = |000⟩ ≡ |0⟩
1 001 |0⟩ ⊗ |0⟩ ⊗ |1⟩ = |001⟩ ≡ |1⟩
2 010 |0⟩ ⊗ |1⟩ ⊗ |0⟩ = |010⟩ ≡ |2⟩
3 011 |0⟩ ⊗ |1⟩ ⊗ |1⟩ = |011⟩ ≡ |3⟩
4 100 |1⟩ ⊗ |0⟩ ⊗ |0⟩ = |100⟩ ≡ |4⟩
5 101 |1⟩ ⊗ |0⟩ ⊗ |1⟩ = |101⟩ ≡ |5⟩
6 110 |1⟩ ⊗ |1⟩ ⊗ |0⟩ = |110⟩ ≡ |6⟩
7 111 |1⟩ ⊗ |1⟩ ⊗ |1⟩ = |111⟩ ≡ |7⟩

Table 3: Example for the binary representation of numbers with 3 classical bits or three entan-
gled qubits.

209
 9 Concepts of Quantum Theory

Of course, we can also bring the entangled states into superposition (although, unfortunately, in
practice this becomes increasingly difficult with the degree of entanglement). For n-entangled
states we find:

|ψ⟩ = a0 |00 . . . 0⟩ + a1 |00 . . . 1⟩ + · · · + a2n −1 |11 . . . 1⟩ =

= a0 |0⟩ + a1 |1⟩ + · · · + a2n −1 |2n − 1⟩ (9.26)

A single state thus contains the information of 2n − 1 complex numbers in the form of the
expansion coefficients. In the case of n = 3 qubits, as in Table 3, there are only 7 coefficients,
but for n = 100 qubits, there would already be 1030 coefficients! If we now perform a calculation,
that is, let an operator act on |ψ⟩, the linearity of quantum mechanics follows:

Û |ψ⟩ = a0 Û |0⟩ + a1 Û |1⟩ + · · · + a2n −1 Û |2n − 1⟩ (9.27)

The effect of Û on the superposition state thus corresponds to the parallel effect on the basis
states; we thus perform 2n − 1 calculations simultaneously! An operation does not destroy the
superposition and is also reversible – only when we measure the state does the superposition
collapse into a single basis state |i⟩. To better imagine the effect of any Û in the simple case
of the superposition of two states, one can use the Bloch sphere. We have shown in Figure 40
that a superposition of |0⟩ and |1⟩ (or |↑⟩ and |↓⟩) can be represented as a vector in the context
of this space. When Û acts on |ψ⟩, it results in a rotation of |ψ⟩ and a change in the expansion
coefficients. In the case of a superposition of more than two basis states, the illustrative concept
of the Bloch sphere is unfortunately no longer applicable.

9.5.3 “No-cloning” Theorem

Motivation: Cloning of Quantum States

As we already know, the measurement of a general quantum state destroys that state. For
example, if we measure the spin in the z direction of the state |ψ⟩ = α |↑⟩ − β |↓⟩ using Ŝz ,
the result of the individual measurement will either be |↑⟩ or |↓⟩, with probabilities |α|2
and |β|2 , respectively. However, after the measurement, the original state is irrevocably
destroyed and depending on the (random) measurement result, transitions to the state
|ψ⟩′ = |↑⟩ or |ψ⟩′ = |↓⟩. Therefore, multiple measurements are not possible, and there is
no way to determine the values of α and β from this single measurement.

But perhaps there is a unitary clone operator ÛK , which is somehow able to transfer the
(Source) state |ψS ⟩ onto another (Target) state |ψT ⟩, without changing the (Source) state?
In this section we will explore why this is not possible, why such a clone operator violates
fundamental laws of quantum physics and would also allow for superluminal communica-
tion.

Is it thus possible to copy a general, arbitrary (Source) state |ψS ⟩ ∈ HS onto a (Target) state
|0T ⟩ ∈ HT through a unitary operation ÛK ; or in other words: Is it possible to clone a quantum
state? To answer this question, we first define such a (hypothetical) clone operator through the
following relation:
ÛK (|ψS ⟩ ⊗ |0T ⟩) = |ψS ⟩ ⊗ |ψT ⟩ (9.28)
ÛK causes the state of the (Source) system HS to now also be established in the (Target) system
HT . The clone operator should be able to overwrite |0T ⟩ with any arbitrary state. We therefore
consider two arbitrary, different states |ϕ⟩ and |ψ⟩:

ÛK (|ϕS ⟩ |0T ⟩) = |ϕS ⟩ |ϕT ⟩

(9.29)
ÛK (|ψS ⟩ |0T ⟩) = |ψS ⟩ |ψT ⟩

210
Quantum Theory I

To determine if the assumed clone operator ÛK can truly clone any arbitrary state, we consider
the expression ⟨ϕ|ψ⟩. We also assume that |0⟩ is normalized, hence ⟨0|0⟩ = 1. Utilizing the
unitarity of ÛK , it holds:
(9.29)
⟨ϕ|ψ⟩ = ⟨ϕS |ψS ⟩ ⟨0T |0T ⟩ = ⟨ϕS | ⟨0T |ψS ⟩ |0T ⟩ = ⟨ϕS | ⟨0T |ÛK† ÛK |ψS ⟩ |0T ⟩ =
= ⟨ϕS | ⟨ϕT |ψS ⟩ |ψT ⟩ = ⟨ϕS |ψT ⟩ ⟨ϕT |ψS ⟩ = ⟨ϕ|ψ⟩2

Thus, we have found that the clone operator only works for the state |ϕ⟩, and also for the state
|ψ⟩ under the following condition:
⟨ϕ|ψ⟩ = ⟨ϕ|ψ⟩2
However, this means that |ϕ⟩ and |ψ⟩ are not arbitrary, independent states: First consider the
case where ⟨ϕ|ψ⟩ = 1. Then, it must be |ϕ⟩ = eiφ |ψ⟩. Since the global phase does not matter,
this means in other words, that the states |ϕ⟩ and |ψ⟩ are identical. Conversely, if ⟨ϕ|ψ⟩ = 0,
then the state |ϕ⟩ is orthonormal to |ψ⟩.

Therefore, we have shown: If there is a clone operator ÛK that works for any arbitrary state |ϕ⟩,
it only works (trivially) for identical, and otherwise only for orthogonal states. It does not work
for any other states. Thus, it is clear that a clone operator for arbitrary states cannot exist.

Example: Cloning of a General Qubit

Let’s consider a general qubit, represented by the superposition of the states |0⟩ and |1⟩
with the complex coefficients α and β as:

|ψS ⟩ = α |0⟩ + β |1⟩ (9.30)

We let the clone operator ÛK act on |ψS ⟩ |0T ⟩, to transfer the state |ψS ⟩ onto |0T ⟩. Following
(9.29):
(9.30)
ÛK (|ψS ⟩ |0T ⟩) = |ψS ⟩ |ψT ⟩ =
= (α |0S ⟩ + β |1S ⟩) (α |0T ⟩ + β |1T ⟩) =
= α2 |0S ⟩ |0T ⟩ + αβ |0S ⟩ |1T ⟩ + βα |1S ⟩ |0T ⟩ + β 2 |1S ⟩ |1T ⟩ ≡
≡ α2 |00⟩ + αβ |01⟩ + βα |10⟩ + β 2 |11⟩ (9.31)

Due to the linearity of the superposition we can also write:

ÛK (|ψS ⟩ |0T ⟩) = ÛK [(α |0S ⟩ + β |1S ⟩) |0T ⟩] =

(9.29)
= αÛK |0S ⟩ |0T ⟩ + β ÛK |1S ⟩ |0T ⟩ =
= α |0S ⟩ |0T ⟩ + β |1S ⟩ |1T ⟩ ≡ α |00⟩ + β |11⟩ (9.32)

Comparing (9.31) with (9.32), we recognize that α2 = α, β 2 = β and αβ = βα = 0 must

be satisfied. This is, however, only fulfilled if α = 1 and β = 0 (or vice versa) holds. It is
therefore evident that the clone operator ÛK can only copy the base states |0⟩ and |1⟩, but
not general superpositions as in (9.30).

The fact that a clone operator cannot exist may seem at first glance very unfortunate for certain
applications. For instance, it is an inherent problem of the quantum computer that it is not
possible to simply copy an arbitrary quantum state, composed of a superposition of many qubits.
However, if one considers the consequences of a potential clone operator more closely, one realizes
that they would set aside both the foundations of quantum physics itself and the special theory
of relativity.

211
 9 Concepts of Quantum Theory

Superluminal Communication with the Clone Operator If a clone operator existed, the follow-
ing superluminal communication protocol could be applied: Alice and Bob wish to communicate
with each other superluminally in binary code. Exactly in the middle between them is a particle
source that emits a steady stream of entangled spin- 12 particles in opposite directions, so that
Alice always receives a particle when the corresponding, entangled particle arrives at Bob. The
entangled particles have the following state:
1
|ψ⟩ = √ (|↑⟩A |↓⟩B − |↓⟩A |↑⟩B ) (9.33)
2
This means, if Alice measures in the z basis, there is a 50% probability that she will find her
particle in state |↑⟩A , but also a 50% probability that she will determine the state |↓⟩A . Due to
the entanglement, we can be sure: Every time Alice measures |↑⟩A , Bob will measure |↓⟩B , and
vice versa.
A crucial point in our argument is the use of two orthogonal measurement bases. One measure-
ment basis is simply the spin in the z direction {|↑⟩ , |↓⟩}, while the other measurement basis
is the spin in the x direction {|↑x ⟩ , |↓x ⟩} ≡ {|→⟩ , |←⟩}. We already know from (7.55) how to
express |→⟩ and |←⟩ in the basis {|↑⟩ , |↓⟩}:
(7.55) 1
|→⟩ ≡ |↑x ⟩ = cos (π/4) |↑⟩ + sin (π/4) |↓⟩ = √ (|↑⟩ + |↓⟩) (9.34)
2
(7.55) 1
|←⟩ ≡ |↓x ⟩ = sin (π/4) |↑⟩ − cos (π/4) |↓⟩ = √ (|↑⟩ − |↓⟩) (9.35)
2
If we add (9.34) and (9.35), or subtract (9.35) from (9.34), we get |↑⟩ and |↓⟩ in the basis
{|→⟩ , |←⟩}:
√ 1
|→⟩ + |←⟩ = 2 |↑⟩ =⇒ |↑⟩ = √ (|→⟩ + |←⟩) (9.36)
2
√ 1
|→⟩ − |←⟩ = 2 |↓⟩ =⇒ |↓⟩ = √ (|→⟩ − |←⟩) (9.37)
2
We can thus use (9.36) and (9.37) to express the entangled state |ψ⟩ from (9.33) in the basis
{|→⟩ , |←⟩} and show that it retains an equivalent form in the new basis too. It follows:
1 (9.36,9.37)
|ψ⟩ = √ (|↑⟩A |↓⟩B − |↓⟩A |↑⟩B ) =
2
1 1 1
 
=√ (|→⟩A + |←⟩A ) (|→⟩B − |←⟩B ) − (|→⟩A − |←⟩A ) (|→⟩B + |←⟩B ) =
2 2 2
1 1

=√ (|→⟩A |→⟩B − |←⟩A |←⟩B + |←⟩A |→⟩B − |→⟩A |←⟩B ) −
2 2
1

− (|→⟩A |→⟩B − |←⟩A |←⟩B − |←⟩A |→⟩B + |→⟩A |←⟩B ) =
2
1 1

=√ (|→⟩A |→⟩B − |←⟩A |←⟩B + |←⟩A |→⟩B − |→⟩A |←⟩B ) +
2 2
1

+ (− |→⟩A |→⟩B + |←⟩A |←⟩B + |←⟩A |→⟩B − |→⟩A |←⟩B ) =
2
1
= (|←⟩A |→⟩B − |→⟩A |←⟩B ) (9.38)
sqrt2

We thus recognize: If Alice measures her particle in the basis {|→⟩ , |←⟩}, she will also find
with a 50% probability the state |→⟩A and can be sure that the (distant) particle at Bob is
also instantaneously in the anti-correlated state |←⟩B . With a 50% probability, however, Alice

212
Quantum Theory I

measures the state |→⟩A , and then knows that the particle at Bob is likewise immediately in
the state |←⟩B .

Everything is therefore prepared for the superluminal communication protocol: Whenever Alice
wants to send a 0, she measures her particle in the basis {|↑⟩ , |↓⟩}. Her particle will then be
either in the state |↑⟩A or |↓⟩A . The entangled particle at Bob will be in the opposite state
|↓⟩B or |↑⟩B , as can be seen from (9.33). If Bob can only conduct one measurement on his
particle in the basis {|↑⟩ , |↓⟩}, it does not benefit him, since he will detect one or the other
state with a 50% probability. But if Bob had a clone operator, he could replicate the state of
his particle and conduct very many measurements. He would then observe (when measured in
the basis {|↑⟩ , |↓⟩} that the measurement on all cloned particles always yields uniformly |↓⟩B or
uniformly |↑⟩B . From this, he could infer that Alice coded a 0; instantaneously, even if she were
many million kilometers away.

If Alice, however, wants to code a 1, then she performs the measurement on her particle in the
basis {|→⟩ , |←⟩}. Her particle will then be either in the state |→⟩A or |←⟩A . The entangled
particle at Bob will be in the opposite state |←⟩B or |→⟩B , as can be seen from (9.38). Let us
assume that Bob continues to take measurements of his particle in the basis {|↑⟩ , |↓⟩}. If his
particle is in the state |→⟩B , he has according to (9.34) a 50-50 probability of measuring |↑⟩ or |↓⟩.
The same applies if his particle is in the state |←⟩B . If he can only conduct one measurement, he
is unable to determine this probability. But with the help of the clone operator, he could again
repeat this measurement on the cloned particles as often as he likes! He would then observe that
the measurement in the basis {|↑⟩ , |↓⟩} on the cloned states yields completely random results,
and could infer from this that Alice coded a 1.

9.5.4 Quantum Cryptography

Motivation: Key exchange in cryptography.

When two communication partners (as usual Alice and Bob) want to communicate secretly
using an encryption algorithm, they must first agree on a key and exchange it between
them. No matter how good the actual encryption algorithm is: there is always a risk that
an unauthorized third person, Eve (after “Eavesdropper”), intercepts this key or learns of
it through another way. Thus, any encryption strategy, no matter how good otherwise, is
at risk.

This is a seemingly unsolvable problem in cryptography. How are Alice and Bob supposed
to know that their key has indeed remained secret? It turns out that quantum physics
offers an elegant solution to this problem.

We want to discuss a particular form of “quantum key distribution”, namely the Bennett-
Brassard protocol (BB84). Let’s assume that Alice wants to send a secret message to Bob. An
encryption should be used, the key of which must be known to both Alice and Bob. To establish
a key, Alice could, for example, generate a stream of two entangled spin- 21 particles each, which
are (as already described in the previous chapter) in the following state:
1
|ψ⟩ = √ (|↑⟩A |↓⟩B − |↓⟩A |↑⟩B ) (9.39)
2
Alice could then perform spin measurements on “her” particle in the basis {|↑⟩ , |↓⟩} and send
the other, entangled particles to Bob, who measures them in the same basis. There is a 50
However, the key thus exchanged is by no means secure against potential eavesdropping, since

213
 9 Concepts of Quantum Theory

Eve could observe the communication channel and also perform a spin measurement in the z-
basis on the particle that Alice sends to Bob. This would allow her to obtain the key as well.
Eve’s measurement does not change the state of the particle, since she measures in the same
basis as Alice and Bob. Bob cannot, therefore, notice that the key has been “intercepted”.

This problem can, however, be elegantly solved: Alice randomly decides for each measure-
ment whether she measures the spin state in the basis {|↑⟩ , |↓⟩} or in an orthonormal direction
{|←⟩ , |→⟩}. For |→⟩ ≡ |↑x ⟩ and |←⟩ ≡ |↓x ⟩ we have already derived in the previous section
(9.38) that the entangled state |ψ⟩ can also be written as follows:
1
|ψ⟩ = √ (|←⟩A |→⟩B − |→⟩A |←⟩B ) (9.40)
2
This means, if Alice randomly determines the state |→⟩A , Bob will find the state |←⟩B (provided
that both measure in the basis {|←⟩ , |→⟩}), and both can note down a 1. In the opposite case,
both write down a 0.
The crucial point is that Alice does not tell Bob whether she measured in the basis {|↑⟩ , |↓⟩}
or {|←⟩ , |→⟩}. Bob must, therefore, decide (randomly) in which of these two bases he performs
measurements himself for each measurement. In cases where Alice and Bob happen to measure
in the same basis, both will have noted the same random bit. In all cases where they have
measured in different bases, the measurement results (and thus the noted bits) are completely
uncorrelated.

At the end of the particle transmission, Alice tells Bob in which bases she measured her spin
states. Bob can thus discard all measurements in which he measured in a different basis than
Alice, resulting in a shared random key at the end!

Basis ↕ ↕ ↔ ↕ ↔ ↕ ↔ ↔ ↔ ↕ ↔ ↕
Alice
Measurement 1 0 0 1 1 0 1 1 1 0 0 1
Basis ↕ ↔ ↔ ↕ ↔ ↔ ↕ ↕ ↔ ↕ ↕ ↕
Bob
Measurement 1 1 0 1 1 1 0 1 0 0 0 1
Key 1 - 0 1 - - - - 0 0 - 1

Table 4: Alice and Bob measure the spin of a particle in two (orthogonal) bases; the key corre-
sponds only to the measured states that were measured in the same bases (highlighted
in green). However, if Eve eavesdrops, she may measure in a different basis and thus
change the state (highlighted in red). This is manifested when comparing parts of the
bit sequence of Alice and Bob: Despite the same measurement basis, they have noted
a different digit!

But what happens if Eve wants to eavesdrop? She now has the same problem as Bob because
she also doesn’t know in which basis Alice measures her particle spins. Therefore, Eve also has
to randomly choose a basis for each individual measurement. If she chooses the wrong basis,
while Alice and Bob both measure in the same basis, her measurement alters the state of the
particle. Alice and Bob should ideally measure the opposite spin of their particle, but with a 50
To check whether someone has observed the key exchange (i.e., changed the state of the particles
through measurement), Alice and Bob compare a small subsequence of their bit sequence. If
there is a discrepancy (we recall that Bob keeps only the measurements that took place in the
same basis as Alice’s), both know they were eavesdropped on and will not use the key anymore.

214
Quantum Theory I

10 Appendix
In the following sections, aspects that were not discussed in detail in the script will be examined
more closely. Besides the transition from classical to quantum mechanics, the mathematical
background of the differential equations encountered in the previous chapters will be investi-
gated here, specifically the Hermite, Legendre, and Laguerre differential equations and the used
Rodrigues formulas for solutions.

10.1 Transition from Classical to Quantum Mechanics

One assumes a set of generalized coordinates with N independent degrees of freedom:
q = (q1 , . . . , qN ) and p = (p1 , . . . , pN ) (10.1)
H(q, p, t) as the Hamiltonian function shall depend functionally on these generalized coordi-
nates. The Hamiltonian equations of motion consist of 2N differential equations for 2N unknown
functions q(t) and p(t) and are determined by:
∂H ∂H
q̇k = and ṗk = − with k = 1, . . . , N (10.2)
∂pk ∂qk
If a transformation of the phase space onto itself is now performed, then (q, p) → (q′ , p′ ) or
H(q, p, t) → H̃(q′ , p′ , t). It is called a canonical transformation if the Hamiltonian equations of
motion remain invariant under such a transformation.
∂ H̃ ∂ H̃
q̇k′ = and ṗ′k = −
∂p′k ∂qk′

The transformed Hamiltonian function H̃ can differ from the original Hamiltonian function H
by the time derivative of a generating function S = S(q, p′ , t), as this vanishes during variation
of action:
∂S
H̃(q′ , p′ , t) = H(q, p, t) + (10.3)
∂t
S as a function of the original position coordinate and the transformed momentum coordinate
is chosen such that the following relations hold:
∂ ∂
pk = S(qk , p′k , t) and qk′ = ′ S(qk , p′k , t) (10.4)
∂qk ∂pk
An attempt is now made to use canonical transformations to produce a Hamiltonian function
H̃(q′ , p′ , t) that vanishes for all times:
∂S
H̃(q′ , p′ , t) = H(q, p, t) + =0 (10.5)
∂t
If that holds, the Hamiltonian equations of motion can be used, and the following simplified
results for transformed position and momentum are obtained:
∂ H̃
q̇k′ = = 0 =⇒ qk′ = const.
∂p′k
∂ H̃
ṗ′k = − = 0 −→ p′k = const.
∂qk′
qk′ and p′k = α are thus both constants of motion and are determined by the initial condition of
the system. With the obtained momentum, S(q, p′ , t) = S(q, α, t). The total derivative of this
generating function yields:
d ∂S(q, α, t) ∂S(q, α, t)
S(q, α, t) = dq + = pdq − H(q, p, t) = L(q, q̇, t)
dt ∂q ∂t

215
 10 Appendix

Integrating both sides over time, it is recognized that the generating function S corresponds to
the action: Z
S(q, α, t) = dt L(q, q̇, t) (10.6)

If a particle is considered in three-dimensional space, the momentum from (10.4) can be written
as:
p = ∇S
According to (10.3), in classical mechanics, for the Hamiltonian function H(r, p, t) = H(r, ∇S, t)
and the action S(r, α, t). The classical Hamilton-Jacobi equation (10.5) can then be written as

∂S 1 ∂S
H(r, p, t) + = (∇S)2 + V (r) + =0 (10.7)
∂t 2m ∂t
Now we move on to quantum theory: Bringing all terms to one side, the three-dimensional
Schrödinger equation (2.8) in quantum mechanics can be written as follows:

ℏ2 2 ∂
− ∇ ψ(r, t) + V (r)ψ(r, t) − iℏ ψ(r, t) = 0
2m ∂t
The wave function of a particle can be described as a superposition of plane matter waves, as
shown here:
iS
 
ψ(r, t) = Aei(kr−ωt)
= A exp (10.8)
ℏ
For the action S, known in wave optics as “eikonal”, holds:

S = p · r − Et (10.9)

The wave function from (10.8) can now be inserted into the Schrödinger equation:

−ℏ2 2 ∂
 
iS iS
− ∇ + V (r) Ae ℏ = iℏ Ae ℏ
2m ∂t

The constant factors A cancel out. The time derivative on the right-hand side of the equation
can be carried out as:
∂ iS i ∂S iS
 
iS
iℏ e = iℏ
ℏ e ℏ = −Ṡe ℏ
∂t ℏ ∂t
On the left side, only the derivative term is considered, as ψ(r, t) does not change due to the
potential:
ℏ2 2 iS ℏ2 i iℏ i
   
iS iS
− ∇ Ae = −
ℏ ∇ ∇Se ℏ =− ∇ S + (∇S) e ℏ
2 2
2m 2m ℏ 2m ℏ
The exponential functions cancel out, and we obtain:
1 iℏ
(∇S)2 − ∆S + V (r) = −Ṡ
2m 2m
The result can be brought to the form of the classical Hamilton-Jacobi equation (10.7):

1 ∂S iℏ
(∇S)2 + V (r) + = ∆S (10.10)
2m ∂t 2m
As can be seen, we obtain an additional term 2m iℏ
∆S on the right side, which is referred to as
the quantum correction. Letting ℏ approach zero (ℏ → 0) transitions from quantum physics to
classical physics, and we obtain the classical Hamilton-Jacobi equation (10.7) again.

216
Quantum Theory I

10.2 Hermite Differential Equation and Polynomials

We were able in the chapter on the harmonic oscillator to reduce the Schrödinger equation in the
oscillator potential to the Hermite differential equation in (4.37); however, we did not solve it
explicitly but instead used the Rodrigues formula for Hermite polynomials in (4.38) to anticipate
the solution directly. We will discuss one possible solution approach here in more detail and
examine the properties of Hermite polynomials more closely.

10.2.1 Derivation of the Rodrigues Formula for Hermite Polynomials

2
To derive the Rodrigues formula, we first start with a simple approach: u(y) = e−y . Differen-
tiating this expression with respect to y gives us a differential equation for the known solution:
u′ = −2yu =⇒ u′ + 2yu = 0
Differentiating this differential equation (n + 1) times leads us to an increasingly complex ex-
pression:
u(n+2) + 2(yu)(n+1) = 0
To explicitly perform the differentiation of the product term (yu)(n+1) , we apply the Leibniz
formula – we obtain a sum, which terminates already at the second term, because y only appears
linearly in (yu)(n+1) :
n
!
n (k) (n−k)
(y · u) = = yu(n) + (n + 1)u(n−1)
X
(n)
y u
k=0
k

Substituting (n + 1) instead of n in the Leibniz formula, we can directly insert into the above
expression. Furthermore, we substitute ξ(y) = (−1)n u(n) (the term (−1)n is only inserted ad
hoc here to obtain the desired solution structure):
u(n+2) + 2yu(n+1) + (n + 1)u(n) = 0 =⇒ ξ ′′ + 2yξ ′′ + 2(n + 1)ξ = 0
2
We carry out another substitution ξ = ζe−y , where we directly insert all derivatives and cancel
the Gaussian function:
h i
ζ ′′ − 4yζ ′ + 4y 2 ζ − 2ζ + 2y ζ ′ − 2yζ + 2(n + 1)ζ = 0 =⇒ ζ ′′ − 2yζ ′ + 2nζ = 0
 

Now we compare the obtained differential equation with the required Hermitian differential
equation:
h′′ (y) − 2yh′ (y) + 2nh(y) = 0 (10.11)
The two expressions match! Thus, ζ(y) = h(y) = Hn (y), and by substituting all substitutions
back into ζ(y), we obtain the Rodrigues formula:
2 dn −y2
Hn (y) = (−1)n ey e (10.12)
dy n
With this, all Hermite polynomials can be calculated very simply; the first five Hn (y) are formally
given below:
H0 (y) = 1
H1 (y) = 2y
H2 (y) = 4y 2 − 2 (10.13)
H3 (y) = 8y − 12y3

H4 (y) = 16y 4 − 48y 2 + 12

In Figure 53, the polynomials from (10.13) are plotted: considering the nodal rule of wave
functions, this behavior can already be recognized here.

217
 10 Appendix

H2 (x)

Hn (x)
H0 (x)
H4 (x)
H1 (x)

H3 (x)

Abb. 53: The first five Hermite polynomials.

10.2.2 Recursive Representation of Hermite’s Differential Equation

To represent Hermite’s differential equation as in (10.11), we can also find a recursive form.
We start from the Rodrigues formula (10.12): the index of Hn (y) represents both the energy
quantum number of the harmonic oscillator and the degree of differentiation in the formula. If
we differentiate (10.12) again, we get:
d n
(10.12) d 2 d
 
2
Hn (y) = (−1)n ey n
e−y =
dy dy dy
" #
dn −y2
y2 y2 d
n+1
2
= (−1) n
2ye n
e + e n+1
e−y =
dy dy
= 2yHn (y) − Hn+1 (y)

The minus sign results from the need to adjust the prefactor (−1)n to the higher degree of the
polynomial Hn+1 (y). We can thus write:
d
 
Hn+1 (y) = 2y − Hn (y) (10.14)
dy
Starting from H0 (y) = 1, this recursive representation derives all Hermite polynomials from
(10.13). By substituting the preceding Hn (y) into Hn+1 (y), we obtain:
n
d

Hn (y) = 2y − ·1 (10.15)
dy
The term ·1 indicates the first Hermite polynomial. Regarding the algebraic solution for the
Schrödinger equation of the harmonic oscillator, we can make another modification to (10.14):
n
d

1 2 1 2
Hn+1 (y) = e 2
y
y− e− 2 y Hn (y) (10.16)
dy
By explicitly setting Hn (y) as the Hermite polynomial (again H0 (y) = 1), we see that another
explicit expression arises from the recursive form:
n
d

1 2 1 2
Hn (y) = e 2 y y− e− 2 y (10.17)
dy

218
Quantum Theory I

10.2.3 Symmetry
It can be easily shown that the Hermite polynomials, depending on their degree, are either
symmetric or antisymmetric (this was already required at the beginning of the chapter due to
the symmetry of the oscillator potential). By substituting y → −y, we get:

(10.12) 2 dn −(−y)2 n y2 n d
n
2
Hn (−y) = (−1)n e(−y) n
e = (−1) e (−1) n
e−y = (−1)n Hn (y)
d(−y) dy

10.2.4 Orthogonality
The Hermite polynomials are orthogonal in a weighted Hilbert space L2 (R, exp{−y 2 }dy); this
means that scalar products can only be formed with the corresponding Gaussian weight function.
We require that n < m holds and thus obtain:
Z
2
⟨Hn (y)Hm (y)⟩ = dy e−x Hn (y)Hm (y) =
R
dm  −x2 
Z
2 2
= (−1) m
dy e−x e+x Hn (y) e =
R dy m
dHn (y) dm−1  −x2 
" #
dm−1   Z
2
= (−1) m
Hn (y) m−1 e−x − dy e =
dy R R dy dy m−1
dn+1 Hn (y) dm−n−1  −x2 
Z
= (−1) (−1) m n+1
dy e =0
R dy n+1 dy m−n−1
The first term of the partial integration vanishes because the Gaussian function retains its form
even after m − 1 derivations and falls off at the boundaries faster than any polynomial. We
perform the partial integration (n + 1) times, but in the last step, we derive a polynomial of
degree n. The term vanishes, and the orthogonality is proven.

10.2.5 Normalizability
It holds that n = m, and we directly perform all calculation steps of the upper derivative:

dn Hn (y) −x2
Z Z
−x2
⟨Hn (y)Hn (y)⟩ = dy e Hn (y)Hn (y) = (−1) (−1) n n
dy e =
R R dy n
dn Hn (y) √
Z
−x2
= dy e = 2n n! π (10.18)
dy n R

Considering (10.15), we recognize that Hn (y) ∝ (2y)n . The n-th derivative is therefore indepen-
(n)
dent of y and can be taken out of the integral; it holds Hn (y) = 2n n!. The integration over
√
the Gaussian function is a standard integral and yields the term π.

10.3 Legendre Differential Equation and Polynomials

At this point, we concern ourselves with deriving the Rodrigues formula for the Legendre dif-
ferential equation, considering not only the ordinary case but also the associated Legendre
differential equation. To maintain clarity and generality, we keep the substitution u = cos(ϑ).
The Legendre differential equation has the following form:
" #
d h i m2
0= (1 − u2 )f ′ + l(l + 1) − f=
du 1 − u2
" #
m2
= (1 − u2 )f ′′ − 2uf ′ + l(l + 1) − f
1 − u2

219
 10 Appendix

10.3.1 Legendre Polynomials

First, we solve the Legendre differential equation for the case m = 0. We start with an initial
guess and will cleverly manipulate this guess to arrive at the desired equation. We define:

f ≡ f (u) = (1 − u2 )n

The guess (1 − u2 ) can be justified by the prefactor in the second derivative from (5.119); the
exponent is used to allow multiple differentiations. We differentiate f (u) once and obtain a first
differential equation:

f ′ = −2nu(1 − u2 )n−1 =⇒ (1 − u2 )f ′ + 2nuf = 0

Differentiating again brings us to a form that already shows similarities with the required equa-
tion (5.119). We finally differentiate the second-order differential equation n more times:

0 = (1 − u2 )f ′′ + 2(n − 1)uf ′ + 2nf =

= [(1 − u2 )f ′′ ](n) + [2(n − 1)uf ′ ](n) + 2nf (n)

Using the Leibniz formula, we can further simplify the non-trivial terms; the polynomial term
also limits the number of possible differentiations:
n
!
′′ (n) n
[(1 − u )f ] = (1 − u2 )(k) f (n+2−k) =
X
2

k=0
k
= (1 − u2 )f (n+2) − 2nuf (n+1) − n(n − 1)f (n)
n
!
n (k) (n+1−k)
[2(n − 1)uf ′ ](n) = 2(n − 1) =
X
u f
k=0
k
= 2(n − 1)uf (n+1)
+ 2n(n − 1)f (n)

By inserting our found expressions into the processed guess, we recognize that it now almost
corresponds to the Legendre differential equation (5.119):

(1 − u2 )f (n+2) − 2uf (n+1) + n(n + 1)f (n) = 0

Only the n-th derivatives still differ from the required form. We substitute one last time:
dn
Pn ≡ Pn (u) = f (u)
dun
Only now do our guess and (5.119) fully agree – using the solution, we have reconstructed the
searched-for differential equation:

(1 − u2 )Pn′′ − 2uPn′ + n(n + 1)Pn′ = 0 (10.19)

We call our solution functions Legendre polynomials. These are generated by the following
Rodrigues formula:
dn
Pn (u) = n (1 − u2 )n (10.20)
du

10.3.2 Associated Legendre Polynomials

So far, we have only considered the special case m = 0; when we drop this restriction and allow
m ∈ Z, we can derive the complete Legendre differential equation from (5.119). In reference to
the solution of the ordinary Legendre differential equation, we replace n → l in (10.20).

220
Quantum Theory I

We start with the Ansatz from our previous result in (10.20). By inserting the solution into the
full associated Legendre differential equation and differentiating m times, we obtain:

0 = (1 − u2 )Pl′′ − 2uPl′ + l(l + 1)Pl = | ∂xm

(m)
= [(1 − u2 )Pl′′ ](m) − [2uPl′ ](m) + l(l + 1)Pl

We again apply the Leibniz formula to differentiate the multiplicative terms sensibly:
m
!
m (m+2−k)
[(1 − u )Pl′′ ](m) = (1 − u2 )(k) Pl =
X
2

k=0
k
(m+2) (m+1) (m)
= (1 − u2 )Pl − 2muPl − m(m − 1)Pl
m
!
′ (m) m (k) (m+1−k) (m+1) (m)
[2uf ] =2 = 2uPl + 2mPl
X
u Pl
k=0
k

The results thus obtained are inserted into the Ansatz of the differential equation, resulting in:
(m+2) (m+1) (m)
(1 − u2 )Pl − 2(m + 1)uPl + [l(l + 1) − m(m + 1)]Pl =0

The prefactor of the first derivative contains a term (m + 1), which does not match with (5.119)!
To remove this, we multiply the entire equation by (1 − u2 )m . This allows us to pull together
(m+2) (m+1)
the previously separate terms Pl and Pl , enabling a more compact notation:
d h (m+1)
i
(m)
(1 − u2 )m+1 Pl + [l(l + 1) − m(m + 1)](1 − u2 )m Pl
du
We substitute again. The Ansatz is by no means trivial and seems “reasonable” at first glance,
but it will lead us to the desired result. We assert:

ξ ≡ ξ(u) = (1 − u2 )m/2 Pl
(m+1)
However, for us, the first derivative of Pl is also relevant, leading to a more or less complex
expression:
(m+1) d
Pl = (1 − u2 )−m/2 ξ = mu(1 − u2 )−m/2−1 ξ + (1 − u2 )−m/2 ξ ′
du
To evaluate the entire term in the square bracket, we substitute the expression found above:
d h (m+1)
i d h m m+2
i
(1 − u2 )m+1 Pl = mu(1 − u2 ) 2 ξ + (1 − u2 ) 2 ξ ′ =
du du
m m−1 m
= m(1 − u2 ) 2 ξ − m2 u2 (1 − u2 ) 2 ξ + mu(1 − u2 ) 2 ξ ′ −
m m+2
− (m + 2)u(1 − u2 ) 2 ξ ′ + (1 − u2 ) 2 ξ ′′ =
m
nh i
= (1 − u2 ) 2 m − m2 u2 (1 − u2 )−1 ξ+
o
+ [mu − (m + 2)u] ξ ′ + (1 − u2 )ξ ′′
m
nh i
= (1 − u2 ) 2 m − mu2 − m2 u2 (1 − u2 )−1 ξ−
o
− 2uξ ′ + (1 − u2 )ξ ′′

We can insert the results obtained into our original differential equation and finally obtain an
expression that almost matches the result.
( " # )
m
′′ ′ m − mu2 − m2 u2
(1 − u )
2 2 (1 − u )ξ − 2uξ + l(l + 1)ξ +
2
− m(m + 1) ξ =0
1 − u2

221
 10 Appendix

m
Simplifying the fractional term and canceling out (1 − u2 ) 2 finally results in the associated
Legendre differential equation (5.119):
" #
′′ ′ m2
(1 − u )ξ − 2uξ + l(l + 1) −
2
ξ=0 (10.21)
1 − u2

By performing all our substitutions backwards, the associated Legendre polynomials can now
be explicitly given in the Rodrigues representation:

m dl+m
Plm (u) = (1 − u2 ) 2 (1 − u2 )n (10.22)
dul+m
It should be noted that in literature, an additional factor 1/(2l l!) appears – this serves to ensure
that the first Legendre polynomial P00 = 1. We will find this factor in the next section in the
form of normalization.

l m Plm (x)
0 0 √ 1
-1 1 − x2 /2
1 0 √x
+1 − 1 − x2
-2 3(1
√− x )/24
2

-1 x 1 − x2 /2
2 0 (3x2√− 1)/2
+1 −3x 1 − x2
+2 3(1 − x2 )

P00 (x)
P2+1 (x) P3+1 (x)
P30 (x) P30 (x)
P4+1 (x)
Pl+1 (x)
Pl0 (x)

P20 (x)
P1+1 (x)
P10 (x)

x x

Abb. 54: (left) Associated Legendre polynomials for m = 0 for l = 0, . . . , 4. (right) Associated
Legendre polynomials for m = +1 for l = 1, . . . , 4.

10.3.3 Orthogonality and Normalization

To form an orthonormal system, the spherical harmonics (and thus the associated Legendre
polynomials) must be orthogonal and normalized. We will investigate this behavior using Pkm

222
Quantum Theory I

and Plm . However, first, we perform an auxiliary calculation and evaluate the following integral
via a recurrence relation:
Z +1 Z +1
In = dx (1 − x2 )n = dx 1 · (1 − x2 )n =
−1 −1
Z +1
d h d
 i 
= dx x(1 − x2 )n − x (1 − x2 )n =
−1 dx dx
h i+1 Z +1
= x(1 − x2 )n + 2n dx x2 (1 − x2 )n−1 =
−1 −1
Z +1
= 2n dx [1−(1 − x2 )](1 − x2 )n−1 =
−1
Z +1 Z +1 
= 2n dx (1 − x ) 2 n−1
− dx (1 − x ) 2 n
= 2n [In−1 − In ]
−1 −1

We reinstate into the result found an additional recurrence relation for In−1 , obtaining iteratively:

2n (2n)(2n − 2) 2n n!
In = In−1 = In−2 = I0 =
2n + 1 (2n + 1)(2n − 1) (2n + 1)!!
2n n! n!n!
= 2n n! · 2 = 22n+1 (10.23)
(2n + 1)! (2n + 1)!

We will use this expression later. For the sake of avoiding excessive notation, we substitute
v = (1 − u2 ) and calculate the product between the polynomials Pkm and Plm :

1
Z 1 Z 1   
= du Pkm Plm = du v m ∂uk+m v k ∂ul+m v l =
N2 −1 −1
   1 Z 1   h  i
= v m ∂uk+m v k ∂ul+m−1 v l − du ∂ul+m−1 v l ∂u v m ∂uk+m v k =
−1 −1
Z 1   h  i
= (−1)1 du ∂ul+m−1 v l ∂u v m ∂uk+m v k = | × (l + m)
−1
Z 1 h  i Z 1
= (−1)l+m du v l ∂ul+m v m ∂uk+m v k = (−1)l+m du v l D(k; l)
−1 −1

We perform the partial integration (l + m) times to create a polynomial completely free from
derivation. The expression in front of the integral disappears on each iteration because v =
(1 − u2 ) vanishes at u = +1 and u = −1! We calculate the final derivation term ∂ul+m [. . . ]
separately:
h  i
D(k; l) = ∂ul+m (1 − u2 )m ∂uk+m (1 − u2 )k ∝ |k = l
h  i
∝ ∂ul+m (−1)m u2m ∂ul+m (−1)l u2l =
(2k)! l−m
 
= (−1)l+m ∂ul+m u2m u =
(l − m)!
(2k)! l+m h l+m i
= (−1)l+m ∂ u =
(l − m)! u
(l + m)!
= (−1)l+m (2l)! (10.24)
(l − m)!

We require that l ≥ k must hold (in the reverse case, we would have needed to integrate
partially differently)! The entire expression in the square bracket has a polynomial degree
deg[. . . ] = k + m, thus it is less than the degree of derivation. At this point, we can demand
that k = l holds, and Pkm and Plm are orthogonal to each other; if the polynomial and derivation

223
 10 Appendix

degrees are equal, we only need to consider the term with the largest exponent, as all other
terms vanish due to derivation.
Inserting (10.23) and (10.24) into the original integral we obtain a rounded expression for the
normalization of the associated Legendre polynomials:

1
Z 1 Z 1 h  i (10.24)
= du Plm Plm = (−1) l+m
du v l ∂ul+m v m ∂ul+m v l =
N2 −1 −1
(l + m)! 1
Z
(10.23)
= (−1)l+m (−1)l+m (2l)! du (1 − u2 )l =
(l − m)! −1
(l + m)! 2l+1 l!l!
= (−1)2(l+m) (2l)! 2 =
(l − m)! (2l + 1)!
h i2 2 (l + m)!
= (−1)(l+m) 2l l! (10.25)
2l + 1 (l − m)!

Returning to the spherical harmonics from (5.114); in addition to the normalization of the
Legendre polynomials (10.25), we also have to account for the normalization of the φ-dependent
part in (5.117). It follows:

1 1
Z 2π
= dφ e−imφ e+imφ = 2π =⇒ N = √ (10.26)
N2 0 2π
Together, (10.25) and (10.26) now provide the complete normalization of the spherical harmon-
ics: s
(−1)(l+m) 2l + 1 (l − m)!
N= (10.27)
2l l! 4π (l + m)!

10.4 Laguerre Differential Equation and Polynomials

Analogous to the Hermite and Legendre differential equations, we aim to investigate the Laguerre
differential equation and its solutions here. With the quantities β ≥ 0 and w ∈ R, the equation
is:
xy ′′ + (β + 1 − x)y ′ + wy = 0 (10.28)
To encounter the Rodrigues formula of the Laguerre differential equation, we will again begin
with an Ansatz; we conduct our derivation here directly for the general case β ≥ 0, while β > 0
is referred to as the associated Laguerre differential equation.

10.4.1 Rodrigues Formula of the Laguerre Differential Equation

Similar to the other cases, we begin from the “solution” and derive it through successive differ-
entiation back to our desired differential equation. We set:

y(x) ≡ y = e−x xw+β

Let’s start our derivation by differentiating the upper equation once and transforming it:
w+β
 
′ −x −x
y = −e x w+β
+ (w + β)e x w+β−1
= − 1 y =⇒ xy ′ + (x − w − β)y = 0
x
We obtain a first differential equation that, however, does not yet resemble the required form
from (10.28). We differentiate again:

xy ′′ + (x − w − β + 1)y ′ + y = 0

224
Quantum Theory I

Each of these terms will be differentiated another w times; to evaluate the product terms, we
apply the Leibniz formula:
w
!
dw ′′ X w (k) (w+2−k)
w
xy = x y = xy (w+2) + wy (w+1)
x k=0
k
dw
(x − w − β + 1)y ′ = (x − w − β + 1)y (w+1) + ny (w)
xw
Inserting the results of our auxiliary calculation into the differential equation, we can again
substitute: kw = y (w) . This yields the following new equation:
′′ ′
xy (w) + (x − β + 1)y (w+1) + (w)y (w) = kw + (x − β + 1)kw + (w + 1)kw = 0

While the second derivative term in this form already aligns, the prefactors of the first and
zeroth derivatives still present problems. Therefore, we substitute again:

kw (x) ≡ kw = xβ e−x Lβw

The first and second derivatives of kw are somewhat laborious to calculate, hence, without
delving into technical details, we ahead the result:
β β
 
′ β′
kw = L − Lβw + Lw xβ e−x
x w
α α(α − 1) 2α
    
′′
kw = Lβw ′′ + 2Lβw ′ − 1 + Lβw − + 1 xβ e−x
x x2 x
Inserting these results into the obtained differential equation and canceling, we obtain:

xLβw ′′ + (β + 1 − x)Lβw ′ + wLβw = 0

This corresponds exactly to the Laguerre differential equation! Thus, we have found a solution
formula with which all solutions Lβw , (so-called “associated Laguerre polynomials”), can be
constructed. This formula is known as the Rodrigues formula. By reversing all substitutions,
we obtain (without any particular normalization):

dw  w+β −x 
Lβw (x) ≡ Lβw = x−β ex x e (10.29)
dxw
A commonly used method for the simple generation of associated Laguerre polynomials Lβw is
the following summation formula:

w
(w + β)!
Lβw (x) = (−1)m (10.30)
X
xm
m=0
(w − m)!(β + m)!m!

This summation formula corresponds to the Rodrigues formula (10.29), provided it is supple-
mented with an additional prefactor 1/w!:

1 −β x dw  w+β −x 
Lβw (x) = x e x e (10.31)
w! dxw

225
 10 Appendix

10.4.2 Orthogonality and Normalizability

The Laguerre polynomials are only normalizable when a weighting function g(x) = xα e−x is
used. Substituting our Rodrigues formula; we will replace ξτ (x) ≡ ξτ = xτ +β e−x to shorten the
following expressions. This yields:
1
Z ∞ Z ∞
= dx xα e−x Lβv Lβw = dx xα e−x x−β ex ξv(v) x−β ex ξw
(w)
=
N2 0 0
Z ∞
= dx x−β ex ξv(v) ξw
(w)
=
0
∞
Z ∞  (1)
= x−β ex ξv(v) ξw
(w−1)
− (w−1)
dx ξw x−β ex ξv(v) = | ×w
0 0
Z ∞  (w) Z ∞
= (−1)w dx ξw x−β ex ξv(v) = (−1)w dx xw+β e−x D(w; v)
0 0

L02 (x)

L03 (x)
Lβw (x)

L00 (x)

L01 (x)

Abb. 55: A selection of Laguerre polynomials each at β = 0.

D(w; v) corresponds to a constant, as shown in the lower auxiliary calculation; the remaining
integral can be equated with the Γ function: Γ(w + β + 1). For the nested derivation, we finally
obtain, assuming w > v:
h  i
D(w; v) = ∂xw x−β ex ∂xv xv+β e−x =
v
" ! #
v  v+β (k) −x
(v−k)
= ∂xw x−β ex =
X
x e
k=0
k
v
" ! #
−β x −x v  v+β (k)
= (−1)v ∂xw (−1)−k =
X
x e e x
k=0
k
v
(v + β)!
!
v
 
= (−1)v (−1)k ∂xw x−β xv+β−k =
X

k=0
k (v + β − k)!
v
(v + β)!
!
v
= (−1) (−1)k ∂ w xv−k = | v = w
X
v

k=0
k (v + β − k)! x
(v + β)!
!
v
= (−1)v (−1)0 ∂ w xw = (−1)v w!
0 (v + β − 0)! x

226
Quantum Theory I

Evaluating the above integral gives us the following normalization: N = (w! Γ(w + β + 1))−1/2 .
Altogether, we can now represent the normalized Rodrigues formula of the Laguerre polynomials
as:
1 dw −β x  w+β −x 
L̃βw (x) = p x e x e (10.32)
w! Γ(w + β + 1) dxw

227
Quantum Theory I

11 Useful Relations
In the following table, a selection of useful eigenvalue relations of operators from this script is
given; found are, in addition to the notation used for the operator, the eigenstate as well as the
actual eigenvalue equation:

 Operator |a⟩ Eigenstate Effect

Common Operators
x̂ Position |x⟩ Position x̂ |x⟩ = x |x⟩
p̂ Momentum |p⟩ Momentum p̂ |p⟩ = p |p⟩
Ĥ Hamiltonian |ψn ⟩ Energy Ĥ |ψn ⟩ = En |ψn ⟩
Π̂ Parity |φπ ⟩ Parity Π̂ |φπ ⟩ = pπ |φπ ⟩
H. Oscillator
N̂ Number State |n⟩ H. Oscillator N̂ |n⟩ = n |n⟩
√
↠Raising |n⟩ H. Oscillator ↠|n⟩ = n + 1 |n + 1⟩
√
Lowering |n⟩ H. Oscillator â |n⟩ = n |n − 1⟩
â
State |α⟩ Glauber â |α⟩ = α |α⟩
Angular Momentum (Ĵ, L̂, Ŝ with Ĵ = L̂ + Ŝ)
Ĵ Magnitude |j, mj ⟩ Ang. Momentum Ĵ |j, mj ⟩ = ℏ2 j(j + 1) |j, mj ⟩
Jˆz z-Component |j, mj ⟩ Ang. Momentum Jˆz |j, mj ⟩ = ℏmj |j, mj ⟩
Jˆ+ Raising |j, mj ⟩ Ang. Momentum Jˆ+ |j, mj ⟩ = ℏ j(j + 1) − m(m + 1) |j, mj + 1⟩
p

Jˆ− Lowering |j, mj ⟩ Ang. Momentum Jˆ− |j, mj ⟩ = ℏ j(j + 1) − m(m − 1) |j, mj − 1⟩
p

Spin- 12 (σ̂ = ℏ2 Ŝ)

σ̂ Magnitude |s, ms ⟩ Spin- 12 σ̂ |s, ms ⟩ = 3 |s, ms ⟩
σ̂z z-Component |s, ms ⟩ Spin- 12 σ̂z |s, ms ⟩ = ± |s, ms ⟩
σ̂+ Raising |s, ms ⟩ Spin- 21
σ̂+ |s, −⟩ = |s, +⟩
σ̂− Lowering |s, ms ⟩ Spin- 12 σ̂− |s, +⟩ = |s, −⟩

The following table provides useful commutator relations: It should be noted that the commu-
tator must always act on an underlying wave function.

 Operator B̂ Operator Commutator

Common Operators
x̂i Position p̂j Momentum [x̂i , p̂j ] = iℏδij
Harmonic Oscillator
â Lowering ↠Raising [â, ↠] = 1
N̂ Number State â Lowering [N̂ , â] = −â
N̂ Number State â † Raising [N̂ , ↠] = â†
Angular Momentum
L̂i Angular Momentum V̂j Vector [L̂i , V̂j ] = iℏεijk Vk
L̂i Angular Momentum Ŝ Scalar [L̂i , Ŝ] = 0
L̂± Raising/Lowering L̂∓ Lowering/Raising [L̂± , L̂∓ ] = ±2ℏL̂z
L̂z z-Angular Momentum L̂± Raising/Lowering [L̂z , L̂± ] = ±ℏL̂±
L̂2 Angular Momentum Magnitude L̂± Raising/Lowering [L̂2 , L̂± ] = 0

229