Lectures On Quantum Mechanics

a
r
X
i
v
:
m
a
t
h
-
p
h
/
0
5
0
5
0
5
9

v
4

2
8

M
a
y

2
0
0
5
1
Lectures on Quantum Mechanics
(nonlinear PDE point of view)
A.I.Komech
Wien 2002/2004
Abstract
We expose the Schrodinger quantum mechanics with traditional applications to Hydrogen atom:
the calculation of the Hydrogen atom spectrum via Schrodinger, Pauli and Dirac equations, the Heisen-
berg representation, the selection rules, the calculation of quantum and classical scattering of light
(Thomson cross section), photoeect (Sommerfeld cross section), quantum and classical scattering of
electrons (Rutherford cross section), normal and anomalous Zeemann eect (Lande factor), polariza-
tion and dispersion (Kramers-Kronig formula), diamagnetic susceptibility (Langevin formula).
We discuss carefully the experimental and theoretical background for the introduction of the
Schrodinger, Pauli and Dirac equations, as well as for the Maxwell equations. We explain in detail
all basic theoretical concepts: the introduction of the quantum stationary states, charge density and
electric current density, quantum magnetic moment, electron spin and spin-orbital coupling in vector
model and in the Russel-Saunders approximation, dierential cross section of scattering, the Lorentz
theory of polarization and magnetization, the Einstein special relativity and covariance of the Maxwell
Electrodynamics.
We explain all details of the calculations and mathematical tools: Lagrangian and Hamiltonian
formalism for the systems with nite degree of freedom and for elds, Geometric Optics, the Hamilton-
Jacobi equation and WKB approximation, Noether theory of invariants including the theorem on
currents, four conservation laws (energy, momentum, angular momentum and charge), Lie algebra
of angular momentum and spherical functions, scattering theory (limiting amplitude principle and
limiting absorption principle), the Lienard-Wiechert formulas, Lorentz group and Lorentz formulas,
Pauli theorem and relativistic covariance of the Dirac equation, etc.
We give a detailed oveview of the conceptual development of the quantum mechanics, and expose
main achievements of the old quantum mechanics in the form of exercises.
One of our basic aim in writing this book, is an open and concrete discussion of the problem of a
mathematical description of the following two fundamental quantum phenomena: i) Bohrs quantum
transitions and ii) de Broglies wave-particle duality. Both phenomena cannot be described by au-
tonomous linear dynamical equations, and we give them a new mathematical treatment related with
recent progress in the theory of global attractors of nonlinear hyperbolic PDEs. Namely, we suggest
that i) the quantum stationary states form a global attractor of the coupled Maxwell-Schrodinger or
Maxwell-Dirac equations, in the presence of an external conning potential, and ii) the wave-particle
duality corresponds to the soliton-like asymptotics for the solutions of the translation-invariant coupled
equations without an external potential.
We emphasize, in the whole of our exposition, that the coupled equations are nonlinear, and
just this nonlinearity lies behind all traditional perturbative calculations that is known as the Born
approximation. We suggest that both fundamental quantum phenomena could be described by this
nonlinear coupling. The suggestion is conrmed by recent results on the global attractors and soliton
asymptotics for model nonlinear hyperbolic PDEs.
2
Contents
0 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1 Introduction: Quantum Chronology 1859-1927 . . . . . . . . . . . . . . . . . . . . . . 15
1.1 Missing Matter Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2 Thermodynamics, Optics and Electrodynamics . . . . . . . . . . . . . . . . . . 15
1.3 Atomic Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
I Lagrangian Field Theory 23
2 Euler-Lagrange Field Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1 Klein-Gordon and Schrodinger Equations . . . . . . . . . . . . . . . . . . . . . 25
2.2 Lagrangian Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Free Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4 The Equations with Maxwell Field . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.5 Action Functional . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Hamilton Least Action Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3 Four Conservation Laws for Lagrangian Fields . . . . . . . . . . . . . . . . . . . . . . . 29
3.1 Time Invariance: Energy Conservation . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Translation Invariance: Momentum Conservation . . . . . . . . . . . . . . . . . 31
3.3 Rotation Invariance: Angular Momentum Conservation . . . . . . . . . . . . . 31
3.4 Phase Invariance: Charge Conservation . . . . . . . . . . . . . . . . . . . . . . 32
4 Lagrangian Theory for the Maxwell Field . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.1 Maxwell Equations and Potentials. Lagrangian Density . . . . . . . . . . . . . 34
4.2 Lagrangian for Charged Particle in Maxwell Field . . . . . . . . . . . . . . . . 36
4.3 Hamiltonian for Charged Particle in Maxwell Field . . . . . . . . . . . . . . . . 38
II Schrodinger Equation 39
5 Geometric Optics and Schrodinger Equation . . . . . . . . . . . . . . . . . . . . . . . . 41
5.1 Straight Line Propagation for the Free Equations . . . . . . . . . . . . . . . . . 41
5.2 WKB Asymptotics for Schrodinger Equation with a Maxwell Field . . . . . . . 44
6 Schrodinger Equation and Heisenberg Representation . . . . . . . . . . . . . . . . . . . 47
6.1 Electrons and Cathode Rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.2 Quantum Stationary States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.3 Four Conservation Laws for Schrodinger Equation . . . . . . . . . . . . . . . . 48
6.4 Quantum Observables and Heisenberg Representation . . . . . . . . . . . . . . 49
7 Coupling to the Maxwell Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.1 Charge and Current Densities and Gauge Invariance . . . . . . . . . . . . . . . 52
7.2 Electron Beams and Heisenbergs Uncertainty Principle . . . . . . . . . . . . . 53
7.3 Quantum Stationary States and Attractors . . . . . . . . . . . . . . . . . . . . 54
3
4 CONTENTS
7.4 Charge Continuity Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.5 Born Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
III Application of Schrodinger Theory 59
8 Spectrum of Hydrogen Atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
8.1 Spherical Symmetry and Separation of Variables . . . . . . . . . . . . . . . . . 61
8.2 Spherical Laplacian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
8.3 Radial Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
9 Spherical Spectral Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
9.1 Hilbert-Schmidt Argument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
9.2 Lie Algebra of Angular Momenta . . . . . . . . . . . . . . . . . . . . . . . . . . 66
9.3 Irreducible Representations of Commutation Relations . . . . . . . . . . . . . . 67
9.4 Spherical Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
9.5 Angular Momenta in Spherical Coordinates . . . . . . . . . . . . . . . . . . . . 69
10 Atom Dipole Radiation and Selection Rules . . . . . . . . . . . . . . . . . . . . . . . . 70
10.1 Rydberg-Ritz Combination Principle . . . . . . . . . . . . . . . . . . . . . . . 70
10.2 Selection Rules for Cylindrical Symmetry . . . . . . . . . . . . . . . . . . . . . 72
10.3 Selection Rules for Orbital Momentum . . . . . . . . . . . . . . . . . . . . . . . 74
11 Classical Scattering of Light: Thomson formula . . . . . . . . . . . . . . . . . . . . . . 77
11.1 Incident Plane Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
11.2 Scattering Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
11.3 Neglecting the Self-Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
11.4 Scattering in Dipole Approximation . . . . . . . . . . . . . . . . . . . . . . . . 79
12 Quantum Scattering of Light. Zero Order Approximation . . . . . . . . . . . . . . . . 81
12.1 Atom Form Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
12.2 Energy ux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
13 Light Scattering at Small Frequencies: Short-Range Scattering . . . . . . . . . . . . . 84
13.1 First Order Approximation to the Ground State . . . . . . . . . . . . . . . . . 84
13.2 Polarization and Dispersion: Kramers-Kronig Formula . . . . . . . . . . . . . . 86
13.3 Combinational Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
14 Light Scattering in Continuous Spectrum: Photoeect . . . . . . . . . . . . . . . . . . 88
14.1 Radiation in Continuous Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . 88
14.2 Limiting Amplitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
14.3 Long-Range Scattering: Wentzel Formula . . . . . . . . . . . . . . . . . . . . . 89
14.4 Coulomb Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
14.5 Shift in Angular Distribution: Sommerfeld Formula . . . . . . . . . . . . . . . 90
14.6 Photoeect for Excited States . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
15 Scattering of Particles: Rutherford Formula . . . . . . . . . . . . . . . . . . . . . . . . 91
15.1 Classical Scattering by a Nucleus . . . . . . . . . . . . . . . . . . . . . . . . . . 91
15.2 Quantum Scattering of Electrons by the Hydrogen Atom . . . . . . . . . . . . 93
16 Hydrogen Atom in a Magnetic Field. Normal Zeemann Eect . . . . . . . . . . . . . . 96
16.1 Uniform Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
16.2 Normal Zeemann Eect: Triplet Spectrum . . . . . . . . . . . . . . . . . . . . . 97
17 Diamagnetism and Paramagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
17.1 Electric Current at Stationary States . . . . . . . . . . . . . . . . . . . . . . . . 99
17.2 Langevin Formula for Diamagnetic Susceptibility . . . . . . . . . . . . . . . . . 100
17.3 Paramagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
CONTENTS 5
IV Electron Spin and Pauli Equation 101
18 Electron Spin: Experiments and Interpretation . . . . . . . . . . . . . . . . . . . . . . 103
18.1 Einstein-de Haas Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
18.2 Double Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
19 Pauli Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
19.1 Additional Magnetic Moment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
19.2 Additional Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
19.3 Pauli Equation. Uniform Magnetic Field . . . . . . . . . . . . . . . . . . . . . . 106
19.4 Pauli Equation. General Maxwell Field . . . . . . . . . . . . . . . . . . . . . . 107
19.5 Application to the Stern-Gerlach Experiment . . . . . . . . . . . . . . . . . . . 107
20 Einstein-de Haas and Anomalous Zeemann Eects . . . . . . . . . . . . . . . . . . . . 109
20.1 Spin-Orbital Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
20.2 Russell-Saunders Quantum Numbers . . . . . . . . . . . . . . . . . . . . . . . . 110
20.3 Lande Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
20.4 Application to the Einstein-de Haas and Anomalous Zeemann Eects . . . . . 112
V Special Relativity 115
21 Electromagnetic Nature of Light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
21.1 Maxwell Equations. Empirical Synthesis . . . . . . . . . . . . . . . . . . . . . . 117
21.2 Equations for Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
21.3 Problem of Luminiferous Ether: Michelson and Morley Experiment . . . . . . . 119
21.4 Time in a Moving Frame: Lorentz transformations . . . . . . . . . . . . . . . . 120
22 The Einstein Special Relativity and Lorentz Group . . . . . . . . . . . . . . . . . . . . 121
23 Covariant Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
VI Relativistic Dirac Theory 127
24 Relativistic Equation for Electron Field . . . . . . . . . . . . . . . . . . . . . . . . . . 129
25 Problem of Negative Energies for Dirac Equation . . . . . . . . . . . . . . . . . . . . . 131
26 Angular Momentum for Dirac Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 132
27 Pauli Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
28 Lorentz Covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
29 Lorentz Transformation of Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
29.1 Factorization of Lorentz Transformations . . . . . . . . . . . . . . . . . . . . . 136
29.2 Rotations of Dirac Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
30 Coupling to Maxwell Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
30.1 Dirac Equation in the Maxwell Field . . . . . . . . . . . . . . . . . . . . . . . . 140
30.2 Gauge Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
31 Pauli Equation as Nonrelativistic Approximation . . . . . . . . . . . . . . . . . . . . . 141
32 Charge Continuity Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
33 Charged Antiparticles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
34 Hydrogen Atom via Dirac Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
34.1 Spectral Problem and Spherical Symmetry . . . . . . . . . . . . . . . . . . . . 145
34.2 Spherical Spinors and Separation of Variables . . . . . . . . . . . . . . . . . . . 145
34.3 Radial Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
34.4 Hydrogen Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6 CONTENTS
VII Mathematical Appendices 153
35 Newton Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
35.1 One Particle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
35.2 Many Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
35.3 Symmetry Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
36 Lagrangian Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
36.1 One Particle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
36.2 Many Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
37 Noether Theory of Invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
37.1 Symmetry and Noether Theorem on Invariants . . . . . . . . . . . . . . . . . . 163
37.2 Application to Many-Particle Systems . . . . . . . . . . . . . . . . . . . . . . . 164
38 Hamilton Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
38.1 Legendre Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
38.2 Hamilton-Jacobi Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
39 Theory of Noether Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
39.1 Field Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
39.2 Noether Current and Continuity Equation . . . . . . . . . . . . . . . . . . . . . 171
40 Application to Four Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . . . 174
40.1 General Lagrangian Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
40.2 Klein-Gordon Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
40.3 Schrodinger Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
41 Cauchy Problem for Maxwell Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 179
42 Lorentz Molecular Theory of Polarization and Magnetization . . . . . . . . . . . . . . 182
42.1 Constitutive Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
42.2 Stationary Molecular Fields in Dipole Approximation . . . . . . . . . . . . . . 182
42.3 Non-Stationary Fields in Multipole Approximations . . . . . . . . . . . . . . . 183
42.4 Magnetic Moment of a Molecule . . . . . . . . . . . . . . . . . . . . . . . . . . 184
42.5 Macroscopic Limit: Maxwell Equations in Matter . . . . . . . . . . . . . . . . . 185
43 Long-Time Asymptotics and Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 187
43.1 Retarded Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
43.2 Limiting Amplitude Principle in Scattering Problems . . . . . . . . . . . . . . . 188
VIII Exercises 193
44 Exercises for Part I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
44.1 Exercise 1: Main Lemma of the Calculus of Variations . . . . . . . . . . . . . . 195
44.2 Exercise 2: Euler-Lagrange Equations . . . . . . . . . . . . . . . . . . . . . . . 195
44.3 Exercise 3: Light Propagation in a Stratied Medium . . . . . . . . . . . . . . 196
45 Exercises for Part III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
45.1 Exercise 4: The Kepler Problem in 5 Easy Steps . . . . . . . . . . . . . . . . . 197
45.2 Exercise 5: Bohr-Sommerfeld Quantization of the H Atom . . . . . . . . . . . 201
45.3 Exercise 6: Rutherford Scattering Formula . . . . . . . . . . . . . . . . . . . . 203
45.4 Exercise 7: Energy Flow in Maxwell Field . . . . . . . . . . . . . . . . . . . . . 205
45.5 Exercise 8: Electromagnetic Plane Waves . . . . . . . . . . . . . . . . . . . . . 206
45.6 Exercise 9: Hertzian Dipole Radiation . . . . . . . . . . . . . . . . . . . . . . . 208
45.7 Exercise 10: Polarizability in an External Electromagnetic Wave . . . . . . . . 210
45.8 Exercise 11: Fresnels Formulae for Reection and Refraction . . . . . . . . . . 211
45.9 Exercise 12: Classical Zeemann Eect . . . . . . . . . . . . . . . . . . . . . . . 216
CONTENTS 7
45.10 Exercise 13: Diamagnetism and Paramagnetism . . . . . . . . . . . . . . . . . 218
45.11 Exercise 14: Stark Eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
45.12 Exercise 15: Vector Model for Spin-Orbital Interaction . . . . . . . . . . . . . . 221
8 CONTENTS
0 Preface
These lectures correspond to a three-semester course given by the author at the Faculty of Mathematics
of the Vienna University in the academic years 2002/2003 and 2003/2004. We expose well known
Schrodinger quantum mechanics with traditional applications to Hydrogen atom, but the form of the
exposition is intended for a mathematically oriented reader of a graduate level.
The reader might ask why this new textbook could be useful as there are many other well estab-
lished introductions to quantum mechanics. Let us explain our motivations in writing this book.
Our principal aim is to give a reasonable introduction which provides a unied mathematical
strategy in applications to dierent problems. Many modern textbooks mainly focus on calculations
of matrix elements of a more formal nature, which are referred to as a rst approximation of a
perturbative procedure. We go one small step further in discussing the full nonperturbative problems
which form the background of such calculations. This makes the strategy of the applications of the
theory to specic problems more transparent.
Almost all existing texts avoid dealing with such questoins, merely mentioning them. This makes
the subject less accessible for understanding, and also impoverishes both its mathematical and physical
aspects. On the other hand, the recognition of the status of the original nonperturbative problems
leads to many questions which are open mathematical problems at the moment. Some of them are
suggested by the Heisenberg Program [45]. The open discussion of these problems is also one of the
principal aims of this book.
On Open Questions Quantum Mechanics exists as an axiomatic theory operating with Quantum
Stationary States, Elementary Particles, Probabilities, etc. It is well established as a set of rules
which give an excellent description of many experimental facts: atomic and molecular spectra, light
and particle scattering, periodic system of elements, chemical reactions, etc. However, a rigorous
foundation, i.e. a mathematical model of the axiomatics, is unknown at the moment because there
are many open mathematical questions. The cornerstone of the theory is Schrodingers dynamical
equation
(ih
t
e(x))(t, x) =
1
2
(ih
e
c
A(x))
2
(t, x) (S)
(or, analogously, the Klein-Gordon, Dirac Eqn, etc) for the wave function (t, x), where (x) and
A(x) are external electrostatic and magnetic vector potentials, respectively.
In a number of important particular cases, this equation could be solved exactly. Besides, the per-
turbation theory of the rst, second, or fourth orders sometimes allows to nd an approximate solution
with an amazing accuracy. This gives an impression that, in principle, if one did not restrict oneself
with the rst order approximation, then the computations could be carried out with an arbitrarily
high precision. Yet, this impression leads to a very deep confusion, since many of the fundamental
quantum eects could not be described by this linear equation. Let us mention some of them.
I. Bohrs transitions between Quantum Stationary States,
[E
) [E
+
), (0.1)
II. De Broglies Wave-Particle Duality: diraction of the electrons, etc.
III. Borns probabilistic interpretation of the wave function.
The transitions (0.1) are responsible for the spectra of atom radiation which coincide with the
eigenvalues of the stationary equation corresponding to (S). However the transitions are not an
inherent property of the solutions of the linear equation (S).
Similarly, the equation (S) explains the diraction pattern in the Davisson-Germer double-slit
experiment by the Bragg rules. However, the discrete registration of electrons (known as reduction
0. PREFACE 9
of wave packets), with the counting rate corresponding to this pattern, is not related to a property
of the solutions. Note that the stability of elementary particles is not explained yet as was pointed
out by Heisenberg [45] (see [6, 35, 36] for a recent progress in this direction for nonlinear equations).
Finally, Born identied [(t, x)[
2
with the density of probability just to explain the counting rate
in the Davisson-Germer experiment. The identication plays the key role in quantum mechanical
scattering problems. However, it has never been justied in the sense of Probability Theory.
Among other open questions: the explanation of statistic equilibrium in atom radiation (see Com-
ment 10.1), in the photoeect (see Comment 14.3), etc.
On Dynamical Treatment The common strategy in applications of Quantum Mechanics is a skilful
combination of the postulates I-III with the dynamical equation (S) in the description of various
quantum phenomena. The strategy is not mathematically selfconsistent since the postulates formally
are not compatible with the linear autonomous equation (S). Hence, the equation requires a suit-
able (nonlinear) modication which would imply the postulates as inherent properties of the modied
dynamics.
Fortunately, an obvious choice for the nonlinear version is well established: it is the system of
coupled Maxwell-Schrodinger equations. The coupled equations are known since the Schrodinger
paper [85, IV], where the charge and current densities have been expressed in terms of the wave
function. Namely, the static potentials (x) and A(x) in (S) should be replaced by the time-dependent
potentials (t, x) and A(t, x) which obey the Maxwell equations with the charge and current densities
corresponding to the Schrodinger equation. The coupled equations constitute a nonlinear system
for the electron wave function and the Maxwell eld, though the Schrodinger equation is linear with
respect to the wave function. The coupling is inevitable in a description of the atom spectra since the
spectral lines correspond to the wavelengths of the atom electromagnetic radiation which is a solution
to the Maxwell equations. Hence, the coupled equations give an authentic framework within which
Quantum Mechanics, or at least some of its aspects, may be mathematically selfconsistent formulated.
We suggest a new treatment of basic quantum phenolena based on the coupled nonlinear dynamical
equations. Namely, the phenomena I and II inspire the following dynamical treatment respectively:
A. The transitions (0.1) can be treated mathematically as the long-time asymptotics of the solutions
to the coupled equations,
((t, x), A(t, x)) (
(x)e
i
t
, A
(x)), t . (0.2)
Here A(t, x) = ((t, x), A(t, x)), and the limit functions (
(x)e
i
t
, A
(x)) correspond to the

stationary states [E
). The asymptotics would mean that the set of all Quantum Stationary States
is the global point attractor of the coupled dynamical equations (see [3, 43, 47, 95]).
B. The elementary particles seem to correspond to traveling waves (or solitons) which are the
solutions of type ((x vt)e
i(v,x,t)
, A(x vt)), to the coupled equations. Respectively, the de
Broglies wave-particle duality can be treated mathematically as the soliton-type asymptotics
((t, x), A(t, x))
k
(
k
(x v
k
t)e
i(v
k
,x,t)
, A
k
(x v
k
t)), t . (0.3)
Note that the asymptotics A correspond to the bound system, with an external conning potential,
like atom with a Coulomb nuclear potential, while A correspond to the translation invariant systems,
witout an external potential. More detailed asymptotics would include also a dispersive wave which
is a solution to the corresponding free linear system (see [64]). The dispersive wave in (0.2) should
describe the electromagnetic radiation (the photon emission) following the quantum transitions for
an atom: the radiation of the wave function is impossible by the charge conservation and the neutrality
of the atom.
10 CONTENTS
The long time asymptotics of type A and B are proved at present time for some model nonlin-
ear hyperbolic partial dierential equations (see below). For the coupled Maxwell-Schrodinger and
Maxwell-Dirac equations, the proving of the asymptotics A and B are open problems. Only the ex-
istence of the solutions is proved for the coupled equations, [13, 42], and the existence of solitons is
proved for the Maxwell-Dirac Equations, [31]. Note that the coupling of the elds is the main sub-
ject of Quantum Field Theory: Quantum Electrodynamics, etc. We consider only the semiclassical
coupled equations and do not touch the second quantized version which is the subject of Quantum
Electrodynamics. The coupled second quantized Maxwell-Dirac system of Quantum Electrodynamics
is also nonlinear (see [10, 52, 82, 83]). The corresponding analogs of the conjectures are also open
questions, and are very likely more dicult than in the semiclassical context.
A mathematical treatment of the Born statistical interpretation III is still unknown. Note that
the Born identication of the counting rate with a probability is equivalent to the ergodicity of the
corresponding random process. The statistical description is also necessary in the analysis of atom
radiation (see Comments 10.1 and 14.3).
On Perturbative Approach The coupled equations are commonly used by perturbation techniques.
The formal, perturbative approach is very successful as far as the classical quantum mechanical
results on the electron-nucleon interactions through the Maxwell eld are concerned. Our above
discussion mainly serves to complement this approach with a slightly more mathematical viewpoint.
Our suggestion is equivalent to the explicit recognition of the status of the coupled dynamics which
is responsible for the Quantum Transitions and Wave-Particle Duality. The recognition is inevitable
since the both phenomena seem to be genuine nonperturbative properties of the coupled nonlinear
equations.
Classical textbooks (for example, [12, 84, 89]) also use the coupled equations implicitly. That is,
they treat both equations separately which corresponds to a perturbation approach for the coupled
system. We follow the same strategy explicitly adding certain comments on possible relations to the
coupled system and the suggested long-time asymptotics A and B. For example,
The asymptotics A clarify Schrodingers identication of the Quantum Stationary States with eigen-
functions.
The asymptotics B claim an inherent mechanism of the reduction of wave packets. It claries the
de Broglies wave-particle duality in the Davisson-Germer diraction experiment and in the description
of the electron beam by plane waves.
The great success of the perturbative approach to the electron-nucleon interactions, however, cru-
cially depends on the following two main facts:
i) the linear part of the coupled Maxwell-Schrodinger equations is completely known from the Ruther-
ford experiment, which both detects the universal fact of the positive charge concentration and
uniquely identies the Coloumbic potential of the nuclei,
ii) the nonlinear terms are small due to the smallness of the Sommerfeld coupling constant 1/137,
thereby providing the numerical convergence of the perturbation series (corresponding to the Feynman
diagrams in the second-quantized version).
Both of these facts no longer hold in the case of the strong nuclear interaction. Therefore, a genuine
nonlinear approach might be necessary even from a more phenomenological point of view in this latter
case.
On a Mathematical Justication of the Asymptotics It is natural to think that the long-time
asymptotics of the type A and B are common features of a very general class of nonlinear Hamiltonian
equations. Otherwise, the dynamical equations of Quantum Mechanics would be very exceptional that
does not correspond to the universal character of the physical theory.
Note that the soliton-type asymptotics of type B have been discovered initially for integrable
0. PREFACE 11
equations: KdV, sine-Gordon, cubic Schrodinger, etc. (see [77] for a survey of these results). Let us
list the results on the asymptotics of type A and B obtained in the last decade for nonintegrable
nonlinear Hamiltonian equations.
The asymptotics have been proved for all nite energy solutions of i) 1D nonlinear wave equations
(see [59]-[61]) for a 1D singular relativistic Klein-Gordon equation, [5]), ii) nonlinear systems of 3D
wave, Klein-Gordon and Maxwell equations coupled to a classical particle (see [49]-[51] and [65]-[67]),
iii) the Maxwell-Landau-Lifschitz-Gilbert Equations [56] (see [60] for a survey of these results).
For U(1)-invariant 3D nonlinear Schrodinger equations, the rst result on the attraction of type A
has been established in [87, 88] (see also [79]). The rst results on the asymptotics of type B have been
established in [15, 16] for translation-invariant U(1)-invariant 1D nonlinear Schrodinger equations. The
results [15, 16] are extended in [18] to the dimension n = 3. An extension to the relativistic-invariant
nonlinear Klein-Gordon equation is still an open problem (see [19, 80] for a progress in this direction).
All the results [15, 16, 18, 79, 87, 88] concern initial states which are suciently close to the solitary
manifold.
The global attraction of type A of all nite energy states is established for the rst time in [63] for
the U(1)-invariant 1D nonlinear Klein-Gordon equation with a nonlinear interaction concentrated at
one point.
In [65] an adiabatic eective dynamics is established, for solitons of a 3D wave equation coupled to a
classical particle, in a slowly varying external potential. The eective dynamics explains the increment
of the mass of the particle caused by its interaction with the eld. The eective dynamics is extended in
[35, 36] to the solitons of the nonlinear Schrodinger and Hartree equations. An extension to relativistic-
invariant equations is still an open problem. On the other hand, the existence of solitons and the
Einstein mass-energy identity for them are proved, respectively, in [6] and [28], for general relativistic
nonlinear Klein-Gordon equations. Numerical experiments [69] demonstrate that the soliton-type
asymptotics B hold for all 1D relativistic-invariant equations, however the proof is still an open
problem.
There are also known results concerning stability of solitary waves for general nonlinear Hamilto-
nian equations, [40]. A large variety of the stability results can be found in the survey [93] concerning
the nonlinear Schrodinger equations.
Note that the mathematical theory of the linear Schrodinger and Dirac equations is well estab-
lished now: see for example, [7, 41, 48, 74, 81, 96]. However, the asymptotics A and B generally do
not hold for the linear autonomous Schrodinger and Dirac equations.
Main Goals of the Exposition We pursue the following two principal goals:
I. To explain why the theory has admitted its present form of a dynamical system described by the
Schrodinger, Pauli or Dirac equations coupled to the Maxwell equations.
We follow all details of the development of the coupled dynamical equations, from experimental facts
to the related mathematical context. We introduce the Schr odinger equation as a wave equation for
which rays coincide with trajectories of the Lorenz equation for the classical electron. This introduc-
tion is based on the geometric optics and WKB short wave asymptotics, and is close to the original
Schrodingers idea on the Hamilton optical-mechanical analogy, [85, II]. We explain in detail the ge-
ometric optics and the WKB asymptotics. For the introduction of the Pauli equation we analyze
the double splitting in the Stern-Gerlach experiment, the Einstein-de Haas eect and the anomalous
Zeemann eect. The Dirac equation corresponds to a relativistic energy-momentum relation similarly
to the Schrodinger equation which corresponds to a non-relativistic one.
II. To demonstrate that the coupled dynamical equations allow us to describe the basic quantum phe-
nomena of interaction of matter and electromagnetic radiation as inherent properties of the dynamics
if the asymptotics A and B would hold.
12 CONTENTS
Our analysis shows that the asymptotics would play the key role in a mathematical foundation of
Quantum Mechanics. Our exposition cannot be rigorous when we solve the nonlinear coupled Maxwell-
Schrodinger equations in the rst order approximation. We try to be careful in the solution of the
corresponding linear problems but we did not strain to be everywhere mathematically rigorous in
order not to overburden the exposition.
Among Other Novelties of the Exposition are the following:
An application of the Lagrangian formalism for the identication i) of the energy, momentum and
angular momentum for the Schrodinger equation (Lecture 6), ii) of the coupling of the Maxwell and
Schrodinger equations (Lecture 7, cf. [12, 84, 101]).
The straightforward derivation of the Rydberg-Ritz combination principle and intensities for the
dipole radiation from the coupled Maxwell-Schrodinger equations (Lecture 10). The derivation is a
formalized version of the approach [89].
An update version of the Lorentz theory of molecular polarization and magnetization (Lecture 42),
necessary for the quantum theory of dispersion and diamagnetism. We follow mainly [4] applying the
theory of distributions. The theory gives a framework for quantum versions of the Kramers-Kronig
dispersion theory (Lecture 14) and the Langevin theory of diamagnetism (Lecture 18).
The application of the nonstationary scattering theory to the explanation of the Einstein formula
for the photoeect, and the calculation of the limiting amplitude via the limiting absorption principle
(Lecture 15, cf. [89]). We demonstrate the fundamental role of retarded potentials in this calculation.
A concise presentation of the Russell-Saunders theory of the coupling between the orbital and spin
angular momentum (Lecture 21). We follow mainly [8, 17, 71, 84] and comment on probable relations
with the coupled Maxwell-Pauli equations.
An update form of the Noether Theorem on Currents with the complete proof (Lecture 40, cf.
[37, 103]).
We demonstrate a parallelism of quantum and classical description to clarify their relations and to
motivate an introduction of the corresponding quantum phenomenology: for example, the classical
and quantum description of diamagnetism, Zeemann eect, scattering of light and particles, the in-
troduction of dierential cross section, magnetic moment, etc.
We explain carefully all details of the calculations and all necessary methods of modern Mathemati-
cal Physics: the Lagrange and Hamilton theory for the elds, the Maxwell Electrodynamics and the
Einstein Special Relativity, scattering theory and representation theory of the rotation group. Let
us note that we consider only one-electron problems (Hydrogen atom, alkali atoms, etc). We do not
touch multi-particle problems, Hartree-Fock methods, etc (see [17, 21, 72, 94]).
Further Reading Our main goal is a concise explanation of basic theoretical concepts. More tech-
nical details and a systematic comparison with the experimental data can be found in [9, 17, 84, 89].
Plan of Exposition In the Introduction we describe the chronology of the conceptual development
of Quantum Mechanics. Then in Lectures 2-4 we provide the mathematical background for a concise
introduction of the Schrodinger equation in Lectures 5-7. Lectures 8-21 concern various applications
of the Schrodinger and Pauli equations. The relativistic Dirac theory is exposed in Lectures 22-35,
where we benet a lot from the book of Hannabus [44]. In Lectures 36-43 we collect mathematical
appendices. In Sections 44-45 we solve numerous exercises containing main achievements of the old
quantum mechanics. Let us explain the plan of the lectures in some detail.
Quantum Chronology The genesis of the Schrodinger equation has been inspired by the lack of a
matter equation in classical electrodynamics. The Schrodinger theory is the result of a synthesis of
a theoretical development with various experimental observations. The quantum chronology starts
from Kirchhos spectral law (1859) and the invention of the vacuum tube by Croockes (1870). The
0. PREFACE 13
next main steps are the identication of the cathode rays in the vacuum tube with the electrons by
Thomson (1897), and the fundamental Planck relation E = h
(1901) inspired by the comparizon

of the experimental Wien formula for the spectral density in the Kirchho law, with the Boltzmann-
Gibbs canonical distribution. The Planck relation has been applied by Einstein to the photoeect
(1905). The Einstein theory has been developed further by Bohr (1913) to explain the Rydberg-Ritz
combination principle and the stability of Rutherfords planetar model for the atom. Finally, Bohrs
theory of atom stationary states has lead to the quantum rules of Debye and Sommerfeld-Wilson (1913
and 1916) for the action function, which is a solution to the Hamilton-Jacobi equation. The quantum
rules provide the atom stationary states and allow to reproduce some basic experimentally observed
spectral lines.
In 1926 Schrodinger introduced an equation for the wave function [85]. The equation extends
de Broglies wave-particle theory (1922) from free particles to bound particles. De Broglies theory
is based on an algebraic argument relying on Einsteins special relativity and the Planck relation.
Schrodingers extension combines the algebraic argument with Hamiltons optical-mechanical analogy.
The analogy is justied by the WKB short wave asymptotics for the solutions to the Schrodinger
equation. That is, the corresponding eikonal equation coincides with the Hamilton-Jacobi equation
of the Maxwell-Lorentz theory for the classical electron in the Maxwell eld. This means that the
Schrodinger equation describes the cathode rays as short-wave solutions. An important role of the
Hamilton-Jacobi equation has been recognized previously related to the Debye-Sommerfeld-Wilson
quantization rules for the action.
Formally, a wave equation of the Schrodinger type can be introduced directly by an identication
of the trajectories of the classical electron in the Maxwell eld with the rays of short-wave solutions.
However, the magnitude h
of the corresponding small parameter, and the central role of Quantum

Stationary States, can be recognized only from the whole development, starting with the Kirchho
spectral law and leading to de Broglies wave-particle theory.
Mathematical Background In Lectures 2-4 we expose the Lagrangian eld theory, and the La-
grangian form of Maxwells electrodynamics. Lecture 5 concerns the geometric optics and short-wave
WKB asymptotics. In Lecture 6 we introduce the Schrodinger equation, the Heisenberg represen-
tation, and prove conservation laws. In Lecture 7 we derive the coupling of the Schrodinger wave
function to the Maxwell eld through the Lagrangian formalism.
Applications In Lectures 8-18 we apply the Schrodinger-Maxwell equations to the derivation of the
spectrum of the hydrogen atom, the dipole radiation and selection rules, the dierential cross sections
of the scattering of light and particles by an atom, the refraction coecient, the diamagnetism, and
the normal Zeemann eect. Lecture 9 contains a complete theory of quantum angular momentum,
including the representations of the corresponding Lie algebra. In Lecture 15 we explain the photoeect
by the limiting amplitude principle for the scattering in the continuous spectrum. In Lectures 19-21
we introduce the Pauli equation with electron spin and apply it to the calculation of the gyromagnetic
ratio. In Lectures 22-35 we expose the special relativity (Lorentz Transformations and Covariant Elec-
trodynamics), introduce the relativistic Dirac equation, prove its covariance and the Pauli theorem,
and calculate the spectrum of the hydrogen atom via the Dirac equation.
Mathematical Appendices In Lectures 36-43 we give an introduction to the Lagrange theory for
nite-dimensional systems and for elds: variational principle, conservation of energy, momentum,
angular momentum, Noether invariants, Hamilton equations, Hamilton-Jacobi theorem. In Lecture
40 we give a new simple proof of the Noether theorem on currents. In Lecture 42 we expose the
Lorentz theory of molecular polarization and magnetization. In Lecture 43 we explain the limiting
amplitude principle, limiting absorption principle and the role of retarded potentials for the calcula-
tion of limiting amplitudes.
14 CONTENTS
Exercises In Part VIII, written by C.Adam, we collect the solutions to the related classical prob-
lems: the Kepler problem, Bohr-Sommerfeld quantization, energy and momentum in the Maxwell
eld, electromagnetic plane waves and Fresnels formulae, Hertzian dipole radiation, vector model for
the spin-orbital coupling, etc.
Acknowledgments I thank H.Brezis, I.M.Gelfand, P.Joly, J.Lebowitz, E.Lieb, I.M.Sigal, A.Shnirel-
man, M.Shubin, H.Spohn, M.I.Vishik and E.Zeidler for fruitful discussions. I thank very much Dr.
C.Adam who wrote the Part VIII, for the collaboration in running the seminar following my lectures,
and also for his help in preparation of this book. Finally, I am indebted to the Faculty of Mathematics
of Vienna University, the Max-Planck Institute for Mathematics in the Sciences (Leipzig), the Project
ONDES (INRIA, Rocqencourt), and Wolfgang Pauli Institute for their hospitality.
A.Komech
Leipzig, 10.04.2005
1. INTRODUCTION: QUANTUM CHRONOLOGY 1859-1927 15
1 Introduction: Quantum Chronology 1859-1927
1.1 Missing Matter Equation
We mark important points of the development of Quantum Theory. Every point is signicant in either
obtaining new experimental results or inventing a new treatment, or developing new mathematical
methods.
The Maxwell theory (1865) perfectly describes the motion of charged particles in a given elec-
tromagnetic eld and also the propagation of electromagnetic waves generated by known charge and
current densities , j. However, generally it cannot describe the simultaneous evolution of the un-
known densities and elds since the microscopic evolution equation for the densities is missing. The
representation of the densities as an aggregation of moving charged particles does not help, since the
corresponding mass density is unknown. That is, the ratio / is not known and even does not have
a reasonable meaning, since the ratio e/m takes dierent values for dierent elementary particles.
The situation is better on the macroscopic level in simple media with known electric and magnetic
permeability and conductivity, since then the the macroscopic charge and current densities
mac
, j
mac
are functions of the elds (polarization, magnetization, Ohms law, etc). However, this is not the case
for the vacuum. Hence, classical electrodynamics is not sucient to explain the structure of matter
at the microscopic level.
Quantum mechanics just provides various matter equations: Schrodinger, Klein-Gordon, Dirac equa-
tions, etc. The equations arise inside classical electrodynamics, thermodynamics, optics and atomic
physics from experimental observations of various aspects of the eld-matter interaction and their
theoretical treatment. Let us briey sketch the chronology of the development.
1.2 Thermodynamics, Optics and Electrodynamics
Kirchho-Planck: The Black-Body Radiation
In 1859 Kirchho stated the radiation law which predicts a specic spectral density of light waves
radiated by a black body at a xed temperature. Light was identied with the electromagnetic eld
by Maxwell in 1865. Hence, the Kirchho law concerns the spectrum of the equilibrium distribution
of the electromagnetic eld at a xed temperature. Therefore, it provides a deep indirect information
on the interaction of matter with the Maxwell eld.
The experimental measurements have been performed by Tyndall in 1865, Crova in 1880, Langley
in 1886, Weber in 1887, and Paschen in 1895-1899. Very precise measurements were made in 1899 by
Lummer and Pringscheim, and Kurlbaum and Rubens. They conrmed the Wien formula (1896)
I()
3
exp(
T
) . (1.4)
Note that the traditional reference to the black body just means that its equilibrium radiation coincides
with the equilibrium Maxwell eld since the absorption of the black body is zero by denition. The
comparison of (1.4) with the general equilibrium Boltzmann-Gibbs distribution exp(
E
kT
) (where k
is the Boltzmann constant) suggests the famous Planck relation (1901)
E = h
, (P)
where E is the energy of the emitted photon and h
= k 1.05 10
27
erg sec is the Planck
16 CONTENTS
constant. Using this relation, Planck adjusted the formula (1.4) as
I()
2
exp(
a
T
)
1 exp(
a
T
)
. (KP)
Rydberg-Ritz: Atom Spectra
Atom spectra provide extremely important information on the structure of the atom. In 1885 Balmer
discovered the representation
2n
= R
_
1
2
2

1
n
2
_
(n 3) for a spectral series in the spectrum of
the hydrogen atom. Later, similar representations were found for other series by Paschen (1908)
3n
= R
_
1
3
2

1
n
2
_
(n 4), Lyman (1909)
1n
= R
_
1
1
n
2
_
(n 2), and Brackett (1914)
4n
=
R
_
1
4
2

1
n
2
_
(n 5). Similar structure
mn
=
m
n
, (R)
has been discovered experimentally by Rydberg (1900) for all the lines in several series of other
elements. The importance of these observations was also stressed by Ritz (1908), so it is now commonly
known as the Rydberg-Ritz combination principle, and the numbers
m
are called terms.
Crookes-Herz-Perrin-Thomson: The Cathode Rays and the Electron
The cathode rays were discovered rst in vacuum tube by Crookes in 1870 (he made a discharge tube
with a vacuum level higher than that of the Geissler tube used by Faraday in 1836-1838). The rays
demonstrated the continuous motion of charge in the vacuum in the presence of a Maxwell eld. This
is just one of the situations which is not covered by classical electrodynamics.
The deection of cathode rays in a magnetic eld has been observed in 1880-1890 by Hertz,
Lenard, Perrin and many others. Some physicists thought, like Goldstein, Hertz, and Lenard, that
this phenomenon is like light, due to vibrations of the ether or even that it is light of short wavelength.
It is easily understood that such rays may have a rectilinear path, excite phosphorescence, and eect
photographic plates. Others thought, like Crookes, J.J. Thomson, Perrin and others, that these
rays are formed by matter which is negatively charged and moving with great velocity, and on this
hypothesis their mechanical properties, as well as the manner in which they become curved in a
magnetic eld, are readily explicable. In 1895, Perrin collected the cathode rays, obtaining a negative
charge.
In 1897 J.J.Thomson showed that the rays are also deected by an electrostatic eld. He system-
atized all previous observations and demonstrated the particle-like behavior of the cathode rays which
is described by the Lorentz equation,
a =
e
(E +v B), (L)
where
e
< 0. Concretely, he identied the cathode rays with a beam of particles with negative charge
and introduced the name electron for these particles. This study led to the rst measurement of the
ratio
e
close to its present value. Kaumann [58] also observed the magnetic deection of cathode
rays and obtained a ratio
e
which is close to the value of J.J.Thomson.

J.J.Thomsons identication of the cathode rays led to many fundamental problems concerning
the size and the structure of the electron:
i) Abraham (1906) noted that the energy and the mass of the electron are innite if its radius is zero.
He introduced the model of the extended electron and calculated its radius.
ii) The extended electron cannot be stable because of the electrostatic repulsion (Poincare 1908).
So, classical electrodynamics had to be complemented with a matter equation which could describe
i) the cathode rays and their particle-like behavior, and ii) the stability of the electron.
Herz-Einstein: Photoeect
In 1887 Herz discovered the photoeect (the light electricity) via the generation of electric charge
by the sun radiation. Later, the photoeect has been observed with dierent types of electromagnetic
radiation by Stoletov, Elster, Geitel, Righi, Townsend, Rutherford, Compton and many others. The
experimental observations led to the relation
h
= E
el
A. (E)
Here, E
el
is the (maximal) energy of the photoelectrons detached from the metal by light of the
frequency . The constant A depends on the metal. The photoeect occurs only for large frequencies
of light >
red
= A/h
, where
red
is called the red bound of the photoeect.
In 1905 Einstein proposed the theory of the photoeect [30]: he identied the relation (E) with
energy conservation. That is, Einstein
I. Identied the quantity h
with the energy of the absorbed photon with frequency in accordance

with the Planck relation (P) (which concerns the emitted photon !), and
II. Identied A with the escape energy of the metal.
This explanation treats light as a collection of particle-like photons that cannot be explained by
using a wave picture of light and the classical representation of electrons as particles.
1.3 Atomic Physics
Rutherford: The Nucleus of the Atom and Atom Stability
In 1913 Rutherford discovered the nucleus of the atom in an experiment on the scattering of -particles.
This discovery suggested to him the classical model of the atom, where a nite number of electrons
moves around a point-like nucleus with positive charge. The electrons are governed by the classical
Lorentz eq. (L). However, the model is unstable due to the radiation of the rotating electrons in
accordance with Maxwell electrodynamics. Therefore, the Maxwell theory is insucient to explain
the stability of the atoms.
Bohr: Quantum Stationary States and Transitions
In 1913 Niels Bohr has proposed a new phenomenology for description of the atom stability. Namely,
he represented the Rydberg-Ritz combination principle (R) in the form
h
mn
= E
m
E
n
, (1.5)
which was suggested by the comparison of (R) with the Planck relation (P) and the Einstein treatment
of the relation for the photoeect, (E). Moreover, Bohr interpreted (B) generalizing the Planck and
Einstein ideas:
I. For an atom, there exist Stationary States [E
n
) with the energies E
n
. The atom is always
in a stationary state, and, sometimes, make transitions (or jumps) from one Stationary State to
another,
18 CONTENTS
[E
m
) [E
n
). (1.6)
II. The transition is accompanied by the radiation or absorption of light with frequency
mn
.
III. The identity (1.5) is energy balance in the transition, in accordance with the identication of
Planck and Einstein of the quantum h
mn
with the energy of an emitted or absorbed photon.
Both, the role of the Planck constant h
and the discreteness of the energies E

n
of the stationary states,
cannot be explained by the Maxwell theory. The discreteness of the energies is related to a restriction
to certain stable orbits of the electron in the atom.
Debye-Sommerfeld-Wilson: Old Quantum Theory
In 1913 Debye stated the quantum rule for the determination of stable periodic orbits of the electrons
in the atom,
S = 2nh
, n = 1, 2, 3, ... (D)
where S is the action integral corresponding to the time-periodic orbit of the electron. The rule
was motivated by the Ehrenfest idea of adiabatic invariance. The quantum rules allowed to nd the
hydrogen spectral terms
n
=
R
n
2
, n = 1, 2, ... which exactly agree with the series of Lyman, Balmer
etc. In 1916 Sommerfeld and Wilson extended the rule to more general quasiperiodic orbits.
In 1923 Bohr has developed the correspondence principle which allowed him to discover the selection
rules for the magnetic and azimuthal quantum numbers. The selection rules play the key role in the
explanation of atom spectra and agree with experimental observations, see [12] and [89, Vol.I].
Zeemann-Stern-Gerlach: Atoms in Magnetic Fields
In 1896 Zeemann discovered the splitting of the spectral lines of atoms in a magnetic eld. Lorentz
explained the splitting by the Maxwell theory in the simplest case of the normal Zeemann eect when
the line splits into three lines: and
= , where is proportional to the magnetic

eld. However, the explanation of the general anomalous Zeemann eect cannot be deduced from the
Maxwell theory. One example is the double splitting of the spectra of alkali atoms.
The Maxwell theory predicts a unique value r for the gyromagnetic ratio [m[/[J[ =
B
, where
B
=
[e[h
2mc
is the Bohr magneton and [m[ resp. [J[ are the magnetic and mechanical momenta of
the electron in an atom. In 1915 Einstein and de Haas rst measured the gyromagnetic ratio by an
observation of a magnetization of an iron in an external weak magnetic eld. However, the observed
ratio was 2r, i.e., two times larger than the theoretical value.
In 1921 Stern and Gerlach observed the double splitting of a beam of silver atoms in a strong
non-uniform magnetic eld. This implies that the stationary state of the atom is split into two states
with dierent gyromagnetic ratios [m[/[J[, which again contradicts the Maxwell theory.
Compton: Scattering of Light by Electrons
In 1923, Compton discovered that the scattered light has a wavelength
dierent from the wavelength

of the incident light:

2h
c
sin
2

2
,
where is the angle between the incident and scattered waves, and is the electron mass. Similarly
to the photoeect, the scattering also cannot be explained by using a wave picture of light, where the
wavelength does not change.
De Broglie: Wave-Particle Duality for Free Particles
In 1924 de Broglie, in his PhD thesis, introduced a wave function for a possible description of matter
by waves, [55], in analogy with the particle-wave duality of light which is demonstrated by the Maxwell
theory and the photoeect. Namely, he has applied Einsteins Special Relativity Theory to a beam of
free particles with the energy-momentum vector (E, p).
I. The beam is identied with a plane wave by the following wave-particle relation:
(t, x) = Ce
i(kxt)
beam of free particles. (1.7)
II. The Einstein relativity principle and the Planck relation (P) imply the identity
(E, p) = h
(, k). (1.8)
The identity plays a crucial role everywhere in quantum theory. In particular, it implies the famous
de Broglie relation for the particle wave length = 2/[k[,
=
2h
[p[
.
It also implies the relativistic dispersion relation
h
2
c
2
= h
2
k
2
+
2
c
2
, (1.9)
where is the particle mass and c is the speed of light. It follows from the expression for the
Hamiltonian of the relativistic particle (see (38.6))
E
2
c
2
= p
2
+
2
c
2
. (1.10)
iv) For small values of [p[ c the non-relativistic approximation holds,
E =
_
p
2
c
2
+
2
c
4
c
2
+
p
2
2
. (1.11)
Dropping here the unessential additive constant c
2
, we get the non-relativistic dispersion relation
h
=
h
2
k
2
2
. (1.12)
The dispersion relations (1.9) resp. (1.12) implies the free Klein-Gordon resp. Schrodinger equation
for the corresponding wave function (t, x) = e
i(kxt)
:
1
c
2
[ih
t
]
2
(t, x) = [(ih
x
)
2
+
2
c
2
](t, x), (KG
0
)
ih
t
(t, x) =
1
2
[ih
x
]
2
(t, x). (S
0
)
20 CONTENTS
Klein-Gordon-Schr odinger: Wave Equation for Bound Particles
In 1925-1926 Klein, Gordon and Schrodinger extended de Broglies wave equation to the bound electron
in an external Maxwell eld. The free equations (KG
0
) resp. (S
0
) formally follow from the energy-
momentum relations (1.10) resp. (1.11) by the substitutions
E ih
t
, p ih
x
. (1.13)
For an electron in the external scalar potential (t, x) and magnetic vector potential A(t, x), the
(conserved) energy E is given by [E e(t, x)]
2
/c
2
= [p
e
c
A(t, x)]
2
+
2
c
2
, where e is the charge of
the electron (see (4.36)). Then Klein, Gordon and Schrodinger generalized (KG
0
) to
1
c
2
[ih
t
e(t, x)]
2
(t, x) = [ih
e
c
A(t, x)]
2
(t, x) +
2
c
2
(t, x). (KG)
Schrodinger also generalized the nonrelativistic approximation (1.12) to
E e(t, x) = (p
e
c
A(t, x))
2
/(2),
which transforms into the wave equation
[ih
t
e(t, x)](t, x) =
1
2
[ih
e
c
A(t, x)]
2
(t, x). (S)
The next crucial step of Schrodingers theory is the identication of stationary states with solutions of
the type exp(it)(x) for the static external Maxwell elds (t, x) (x) and A(t, x) A(x). This
identication is suggested by the de Broglie plain wave exp(it) exp(ikx), where only the spatial
factor has to be modied since the external eld twists space but not time. The energy is again
E = h
. This identication leads to the corresponding stationary equations which are the eigenvalue
problems,
1
c
2
[ e(x)]
2
(x) = [ih
e
c
A(x)]
2
(x) +
2
c
2
(x),
[ e(x)](x) =
1
2
[ih
e
c
A(x)]
2
(x)
for the determination of the energies E = h
and the amplitudes (x) of the stationary states.

Schrodinger calculated all solutions to the last equation for the hydrogen atom: (x) = e/[x[
is the Coulomb potential of the nucleus, and A(x) = 0. The agreement with the experimentally
observed spectrum was perfect. The calculation uses the standard separation of variables in spherical
coordinates, which involves some integer numbers as in the Debye quantum condition (D). It was
just this analogy which suggested to Schrodinger an eigenvalue problem for the determination of the
stationary states of the atom.
Heisenberg: Matrix Mechanics
In 1925 Heisenberg have extended signically the Bohr correpondence principle. He assigned the in-
nite matrix i.e. the operator in the Hilbert space, to each classical observable: energy, mometum,
coordinate, angular momentum, etc. This assignment is called now the quantization of the corre-
sponding classical dynamical system. In particular, the substitutions (1.13) give an example of the
assignment at a xed time. In this matrix mechanics the operators depend on time and obeys the
dynamical equations which formally coincide with the equations for the classical observables. The
Heisenberg approach allowed him to discover the key Uncertainty Principle (see Lecture 7) which
plays a fundamental role in the theoretical analysis and applications of the quatum theory.
The Heisenberg ideas were developed by Born, Jordan and others. Later on it was proved that the
Heisenberg approach is completely equivalent to the Schrodinger theory. The equivalence has played
a crucial role in further development of the quantum theory leading to the second quantization and
quantum eld theory.
Uhlenbeck-Goudsmith-Pauli: Nonrelativistic Theory of Spin
In 1925 Uhlenbeck and Goudsmith introduced the hypothesis of the existence of the spin of the electron,
i.e., of its own angular momentum s (i.e. not related to its rotation) and a magnetic moment m (not
related to the corresponding convection current) with the gyromagnetic ratio g := [m[/[s[ = 2
B
. The
hypothesis had been inspired by the double splitting in the Stern-Gerlach experiment, the anomalous
Zeemann eect and the Einstein-de Haas experiment. In 1927 Pauli obtained the wave equation which
takes the spin of the electron into account,
[ih
t
e(t, x)](t, x) =
1
2
[ih
e
c
A(t, x)]
2
(t, x) +g
B
3
1
s
k
B
k
(t, x), (1.14)
where the wave function (t, x) = (
1
(t, x),
2
(t, x)) with two complex-valued functions
j
(t, x).
Further, B
k
are the components of the uniform magnetic eld and s
k
=
k
/2, where
k
are the
complex 2 2 Pauli matrices. Pauli obtained the equation by just postulating the double splitting,
the gyromagnetic ratio g = 2, and covariance with respect to space rotations.
This equation leads to the correct gyromagnetic ratio observed in the Einstein-de Haas experiment.
It also explains the anomalous Zeemann eect and the Stern-Gerlach double splitting. The agreement
with many experimental observations was a great triumph of the quantum theory.
Dirac: Relativistic Theory of Spin
In 1927 Dirac discovered the relativistic invariant equation which generalizes the Pauli and Klein-
Gordon equations,
3
=0
[ih
e
c
/
(x)] = c(x), x IR
4
,
where
0
=
t
/c, (
1
,
2
,
3
) =
x
, /
0
(x) = (x) and /
k
= A
k
(x), k = 1, 2, 3,
are the 4 4
Dirac matrices, and (x) C
4
for x IR
4
.
The Dirac equation automatically provides the correct gyromagnetic ratio g = 2 for the electron.
It gives a much more precise description of the hydrogen atom spectrum than the Schrodinger and
Pauli equations.
Davisson-Germer: Interference of Electrons
In 1927 Davisson and Germer observed the interference of electron beams. Later the experiments were
repeated and conrmed by many authors: Thomson, Rupp, Kikouchi and others. In 1949 Biberman,
Sushkin and Fabrikant observed the interference pattern with a weak beam with a very low rate of
registration of the electrons.
Born: The Probabilistic Interpretation of the Wave Function
In 1927 Born proposed the following interpretation of the wave function to explain the Davisson-
Germer experiment: [(t, x)[
2
is a density of probability.
22 CONTENTS
Part I
Lagrangian Field Theory
23
2. EULER-LAGRANGE FIELD EQUATIONS 25
2 Euler-Lagrange Field Equations
We introduce two equations important for Quantum Mechanics, namely, the Klein-Gordon and Schrodinger
equations. We prove that they are of Lagrangian form. Then we introduce an action functional and
prove the Hamilton least action principle.
The quantum mechanical evolution equations for charged matter in an electromagnetic eld are
provided by the Schrodinger equation and Klein-Gordon equation in the non-relativistic and relativistic
cases, respectively. Both equations may be derived from an action functional via the Hamiltonian least
action principle, and, therefore, a Lagrangian description exists for both cases.
2.1 Klein-Gordon and Schrodinger Equations
Let us dene x
0
= ct, x = (x
1
, x
2
, x
3
), x = (x
0
, x) and consider the Klein-Gordon equation
[ih
e
c
(x)]
2
(x)
= [ih
e
c
A(x)]
2
(x) +
2
c
2
(x), x IR
4
, (2.1)
where the function (x) takes complex values and > 0 is the electron mass. Let us also consider
the Schrodinger equation
[ih
0
e(x)](x)
=
1
2
[ih
e
c
A(x)]
2
(x), x IR
4
, (2.2)
where x
0
:= t.
2.2 Lagrangian Density
Denition 2.1 We will identify the complex vectors C
M
with the real vectors IR := (Re, Im)
IR
2M
and the multiplication by a complex number with an application of the corresponding matrix. We
will denote by the real scalar product in IR
2M
.
This denition implies the formulas
IRu IRv = Re(uv), (2.3)
IRu
(u iv) = iv,
IRv
(u iv) = iu,
IRu
(iu v) = iv,
IRv
(iu v) = iu, (2.4)
for u, v C since u iv = iu v and iu v = u iv.
Let us introduce the Lagrangian densities L for Eqs. (2.1) and (2.2) as the real functions dened
by (cf. (35.3), (35.10)),
L
KG
(x, , ) =
[[ih
e
c
(x)][
2
2

[[ih
e
c
A(x)][
2
2
2
c
2
[[
2
2
, (2.5)
L
S
(x, , ) = [ih
0
e(x)]
1
2
[[ih
e
c
A(x)][
2
. (2.6)
26
We will demonstrate below that the eld equations (2.1), (2.2) can be represented in the Euler-
Lagrange form,
L
(x, (x), (x))

3
=0
(x, (x), (x)) = 0, x IR

4
, (2.7)
where L is the corresponding Lagrangian density.
Remark 2.2 In (2.5) (2.7) the x and ,
are considered as independent variables with the

values in IR
4
and IR
2
, respectively.
Denition 2.3 The Lagrangian eld (x) with values in IR
N
is the dynamical system described by
the N real scalar equations (2.7) with a given Lagrangian density L(x, , ) C
2
(IR
4
IR
N
IR
4N
).
Denition 2.4 The canonically conjugate elds
(x) are dened by
(x) = L
(x, (x), (x)), = 0, ..., 3. (2.8)

With these notations the Euler-Lagrange equations (2.7) read
(x) = L
(x, (x), (x)), x IR

4
. (2.9)
Here and below we use the Einstein convention
(x) :=
(x) etc. Also = 0, 1, 2, 3

and k = 1, 2, 3.
2.3 Free Equations
First consider the free equations without Maxwell eld and without nonlinear self-interaction, and
with h
= 1:
2
0
(x) =
2
x
(x)
2
c
2
(x), x IR
4
, (2.10)
i
0
(x) =
1
2
2
x
(x), x IR
4
, (2.11)
Then the Lagrangian densities (2.5), (2.6) become
L
0
KG
(x, , ) =
[
0
[
2
2

[[
2
2

2
c
2
[[
2
2
, (2.12)
L
0
S
(x, , ) = i
0

1
2
[[
2
(2.13)
Exercise 2.5 Check the Euler-Lagrange form (2.9) for the Klein-Gordon and Schr odinger equations
(2.10), (2.11).
Solution
I For the Klein-Gordon equation (2.10): by Formulas (2.4) we get
0
(x) =
0
(x),
k
(x) =
k
(x), L
=
2
c
2
. (2.14)
Hence (2.9) is equivalent to (2.10).
II For the Schrodinger equation (2.11): by Formulas (2.4) we get
0
(x) = i(x),
k
(x) =
1
k
(x), L
= i
0
(x). (2.15)
2. EULER-LAGRANGE FIELD EQUATIONS 27
2.4 The Equations with Maxwell Field
Exercise 2.6 Check the Euler-Lagrange form (2.9) for the Klein-Gordon and Schr odinger equations
(2.1), (2.2).
Solution
I For the Klein-Gordon equation (2.1): by Formulas (2.4) we get
_
0
(x) = ih
[ih
e
c
(x)](x),
k
(x)=ih
[ih
e
c
A
k
(x)](x),
L
=
e
c
(x)[ih
e
c
(x)](x) +
e
c
A
k
(x)[ih
e
c
A
k
(x)](x)
2
c
2
.
(2.16)
II For the Schrodinger equation (2.2): by Formulas (2.4) we get
_
0
(x) = ih
(x),
k
(x) =
1
ih
[ih
k

e
c
A
k
(x)](x),
L
= ih
0
(x) 2e(x)(x) +
e
c
A
k
(x)[ih
k

e
c
A
k
(x)](x).
(2.17)
2.5 Action Functional
Denition 2.7 For k = 1, 2, ... and > 0 the space C
k
is the set of the functions (t, x) C

k
([0, )
IR
3
, IR
N
) with the space-decay
||k
[
(t, x)[ C
T
(1 +[x[)
, (t, x) C
k
([0, T] IR
3
). (2.18)
We will consider the real-valued functionals T on C
1
. By denition, T is a map C
1
IR.
Denition 2.8 For C
1
the Gateau derivative DT() is the linear functional

DT(), h) =
d
d
=0
T( +h) (2.19)
for h() C
1
if the derivative exists.

We will assume the following bounds for the Lagrangian density,
[L(x, , )[ C([[ +[[)
2
[L
(x, , )[ +[L
(x, , )[ C([[ +[[)
, [[+[[ const. (2.20)

Let us x a T > 0.
Denition 2.9 The action for the eld is the functional on C
1
, > 3/2 dened by

S
T
() =
_
T
0
[
_
IR
3
L(x, (x), (x))dx]dx
0
, () C
1
. (2.21)
28
Note that for > 3/2 the action is dened on the whole of C
1
by (2.18) and the rst inequality of

(2.20). Moreover, the functional is dierentiable:
Lemma 2.10 The Gateau derivative DS
T
(), h) exists for , h C
1
if > 3/2.
Proof From Denition 2.8 we get by the theorem of the dierentiation of integrals,
DS
T
(), h) :=
d
d
=0
_
T
0
[
_
IR
3
L(x, (x) +h(x), (x) +h(x))dx]dx
0
=
_
T
0
[
_
IR
3
_
L
(x, (x), (x))h(x) +

3
=0
L
(x, (x), (x))
h(x)
_
dx]dx
0
(2.22)
since the integrals converge uniformly by (2.18) and (2.20).
2.6 Hamilton Least Action Principle
Let us introduce the space of variations.
Denition 2.11 C
1
(T) is the space of functions h() C
1
([0, T] IR
3
, IR
N
) such that
h(0, x) = h(T, x) = 0, x IR
3
, (2.23)
h(t, x) = 0, [x[ const, t [0, T]. (2.24)
Denition 2.12 The function C
1
satises the Hamilton Least Action Principle (LAP) if for any

T > 0,
DS
T
(), h) = 0, h() C
1
(T). (2.25)
Theorem 2.13 For C
2
with > 3/2 the Hamilton LAP is equivalent to the Euler-Lagrange

equations (2.7).
Proof (2.22) implies by partial integration in x
, = 0, ..., 3,
DS
T
(), h)
=
_
T
0
[
_
IR
3
_
L
(x, (x), (x))h(x)

3
=0
(x, (x), (x))

_
h(x)dx]dx
0
. (2.26)
Therefore, (2.25) is equivalent to (2.7) by the main lemma of the calculus of variations.
3. FOUR CONSERVATION LAWS FOR LAGRANGIAN FIELDS 29
3 Four Conservation Laws for Lagrangian Fields
By the general Noether theorem 39.9, an invariance of the Lagrangian density of a eld with respect
to a symmetry group provides a conservation law. In particular, the theorem implies four classical
conservation laws: conservation of total energy, momentum, angular momentum and charge. Here, the
energy conservation follows from an invariance under time translations, the momentum conservation
follows from translation invariance in space, the angular momentum conservation follows from rotation
invariance, and the charge conservation follows from a phase invariance.
Here we state the four conservation laws. All these conservation laws are deduced from the Noether
theorem of Section 27.2. Let us x a > 3/2.
3.1 Time Invariance: Energy Conservation
Denition 3.1 The energy of the Lagrangian eld (x) C
1
is dened by (cf. (36.5), (36.17))

E(t) =
_
IR
3
_
0
(x)
0
(x) L(x, (x), (x))
_
dx, t IR. (3.1)
Remark 3.2 By denition, t = x
0
and
0
(x)
k
(x) =
j
0
(x)
k
j
(x).
Note that by (2.18) and (2.20) the energy and momentum are well dened for the solutions (t) C
1
since > 3/2.

Example 3.3 The linear wave equation
(t, x) = (t, x), x IR

3
(3.2)
is a particular case of the Klein-Gordon equation (2.1) with m = 0. The Lagrangian density L is given
by (2.5), i.e.
L(t, x, ,

,
) =
[

[
2
2

[[
2
2
. (3.3)
Then the canonically conjugate eld is
0
=

. Hence the energy is given by
E(t) =
_
IR
3
_
[

(t, x)[
2
2
+
[(t, x)[
2
2
_
dx, t IR. (3.4)
Theorem 3.4 Let the Lagrangian density L not depend on x
0
:= t,
L(x, , ) L
1
(x, , ). (3.5)
Then for any trajectory (x) C
2
of the equations (2.9) the energy is conserved, E(t) = const.

Proof for a particular case Let us prove the theorem for an equation of type (3.2) in dimension
one,
(t, x) =
(t, x), x IR. (3.6)

First, E(t) = lim
R
E
R
(t) where
E
R
(t) =
_
R
R
_
[

(t, x)[
2
2
+
[
(t, x)[
2
2
_
dx. (3.7)
30
Dierentiating, we get
E
R
(t) =
_
R
R
_

(t, x)
(t, x) +
(t, x)

(t, x)
_
dx. (3.8)
Substituting here

(t, x) =
(t, x), we get by partial integration
E
R
(t) =
_

(t, x)
(t, x)
_
R
R
=

(t, R)
(t, R)

(t, R)
(t, R). (3.9)

Now, Eq. (2.18) implies that for every xed t we have

E
R
(t) 0 as R , hence E(t) =const.
Indeed, for any T > 0,
E
R
(T) E
R
(0) =
_
T
0
E
(
t)dt, (3.10)
hence in the limit R ,
E(T) E(0) =
_
T
0
lim
R
E
R
(t)dt = 0. (3.11)
Proof for the general case By Denition (3.1), E(t) = lim
R
E
R
(t) where
E
R
(t)=
_
|x|<R
_
0
(x)
0
(x) L(x, (x), (x))
_
x
0
=ct
dx. (3.12)
E
R
(t) =
_
|x|<R
_
0
(x)
0
(x) +
0
(x)
2
0
(x)
L
(x, (x), (x))

0
(x)
3
=0
(x)
0
_
dx. (3.13)
By (2.9), we have
0
0
(x) = L
(x, (x), (x))
3
1
k
(x). Substituting into (3.13), we get by
the Stokes theorem,
E
R
(t) =
_
|x|<R
_
3
k
(x)
0
+
3
=1
(x)
0
_
dx
=
_
|x|<R
3
k
_
k
(x)
0
(x)
_
dx
=
_
|x|=R
3
1
n
k
(x)
_
k
(x)
0
(x)
_
dS, (3.14)
where n
k
(x) := x
k
/[x[, and dS is the Lebesgue measure on the sphere [x[ = R. Now, Eq. (2.18)
implies that for every xed t we have

E
R
(t) 0 as R , hence E(t) =const.
Remark 3.5 The identity (3.14) means that the vector-function
S
k
(x) :=
k
(x)
0
(x) :=
j
k
(x)
0
j
(x), k = 1, ..., 3, x IR
4
(3.15)
is the energy current density.
3.2 Translation Invariance: Momentum Conservation
Denition 3.6 The momentum of the Lagrangian eld at time t is the vector
p
n
(t) =
_
IR
3
_
0
(x)
k
(x)
_
dx, n = 1, 2, 3, t IR. (3.16)
Example 3.7 For the linear wave equation (3.2), the momentum is given by
p(t) =
_
IR
3
(t, x) (t, x) dx, t IR. (3.17)

Theorem 3.8 Assume that the Lagrangian density L(x, , ) does not depend on the coordinate
x
n
. Then for any trajectory (x) C
2
of Eq. (2.9) the corresponding component of momentum is

conserved, p
n
(t) = const .
Proof for a particular case Let us prove the theorem for the dAlembert equation (3.6). By
denition, P(t) = lim
R
P
R
(t) where
P
R
(t) =
_
R
R
(t, x)
(t, x)dx. (3.18)

P
R
(t) =
_
R
R
_
(t, x)
(t, x) +

(t, x)

(t, x)
_
dx. (3.19)
Substituting here

(t, x) =
(t, x), we get,
P
R
(t) =
_
[
(t, x)[
2
2
+
[

(t, x)[
2
2
_
R
R
. (3.20)
Now, Eq. (2.18) implies that, for every xed t, we have

P
R
(t) 0 as R , hence P(t) =const.
Exercise 3.9 Prove Theorem 3.8 for the general case. Hint: Dierentiate in time and apply partial
integration in x.
3.3 Rotation Invariance: Angular Momentum Conservation
Let us denote by O
n
(s) the rotations of the space IR
3
around a unit vector e
n
IR
3
(e
1
:= (1, 0, 0),
etc), with an angle of s radian. Let us consider the Lagrangian densities which are invariant with
respect to rotations:
L(x
0
, x, ,
0
,
x
) L(x
0
, O
n
(s)x, ,
0
, O
t
n
(s)
x
). (3.21)
Example 3.10 Let us consider the case of e
3
:= (0, 0, 1). Then the invariance (3.21) holds for
Lagrangian densities of the following structure:
L(x, (x), (x)) L
(x
0
, [(x
1
, x
2
)[, x
3
, (x),
0
(x), [(
1
(x),
2
(x))[,
3
(x)). (3.22)
Denition 3.11 The angular momentum of the Lagrangian eld at time t is vector
J
n
(t) =
_
IR
d
_
0
(x) (x
x
)
n
(x)
_
dx, n = 1, 2, 3, t IR, (3.23)
where
x
:= (
1
,
2
,
3
): for example, (x
x
)
3
:= x
1
2
x
2
1
etc.
32
Note that the integral converges for solutions (x) C
1
with > 2, by (2.18) and (2.20).

Remark 3.12 By denition,
0
(x) (x
x
)
n
(x) =
j
0
(x) (x
x
)
n
j
(x).
Theorem 3.13 Let the Lagrangian density satisfy (3.22) and > 5/2. Then, for any trajectory
(x) C
2
of Eq. (2.9), the corresponding component of angular momentum is conserved, J

n
(t) =
const.
Exercise 3.14 Prove Theorem 3.13.
Exercise 3.15 Calculate the angular momentum for the Klein-Gordon and Schr odinger equations
(2.1), (2.2).
3.4 Phase Invariance: Charge Conservation
Consider complex-valued elds with Lagrangian densities which are invariant with respect to rota-
tions in :
L(x, e
i
(x), e
i
(x)) L(x, (x), (x)), IR. (3.24)
Exercise 3.16 Check that (3.24) holds for Eqns (2.1) and (2.2).
Denition 3.17 The charge of the Lagrangian eld at time t is dened by
Q(t) =
_
IR
d
_
0
(x) i(x)
_
dx, t IR, (3.25)
where
0
(x) and i(x) are identied with real vectors in IR
2
, and is the real scalar product in IR
2
(cf. (2.4)).
Note that the integral converges for the solutions (t) C
1
by (2.18) and (2.20) since > 3/2.

Theorem 3.18 Let the Lagrangian density satisfy (3.24) and > 3/2. Then for any trajectory
(x) C
2
of Eq. (2.9) the charge is conserved, Q(t) = const .

Proof for a particular case Let us prove the theorem for the free linear 1D Klein-Gordon equation
(2.1),
(t, x) =
(t, x)
2
(t, x), x IR. (3.26)
Then =

and Q(t) = lim
R
Q
R
(t), where
Q
R
(t) =
_
R
R
(t, x) i(t, x)dx. (3.27)

Q
R
(t) =
_
R
R
(t, x) i(t, x)dx +

_
R
R
(t, x) i

(t, x)dx. (3.28)
The second integral on the RHS is zero since z
1
iz
2
is an antisymmetric bilinear form in C by (2.4).
Hence, substituting

(t, x) =
(t, x)
2
(t, x) from (3.26), we get by partial integration,
Q
R
(t) =
(t, x) i(t, x)
R
R
_
R
R
(t, x) i
(t, x)dx
2
_
R
R
(t, x) i(t, x)dx
=
(t, x) i(t, x)
R
R
= j(t, R) j(t, R), j(t, x) :=
(t, x) i(t, x). (3.29)

since both integrals are zero by antisymmetry. Therefore, Eq. (2.18) implies that for every xed t we
have

Q
R
(t) 0 as R , hence Q(t) =const.
Exercise 3.19 Prove Theorem 3.18 for the nonlinear 1D Schr odinger equation (2.2),
i

(t, x) =
1
2
(t, x) +F(), x IR, (3.30)

where F() = U([[), C IR
2
.
Exercise 3.20 Prove Theorem 3.18 for the general case and check the continuity equation,
(t, x) + div j(t, x) = 0, (3.31)
where (t, x) :=
0
(x) i(x) is the charge density and j
k
(t, x) :=
k
(t, x) i(t, x) is the current
density.
Exercise 3.21 Calculate the charge- and current densities for the Klein-Gordon and Schr odinger
equations (2.1), (2.2).
34
4 Lagrangian Theory for the Maxwell Field
The evolution of the electromagnetic eld is governed by the Maxwell equations, which describe,
among other features, the propagation of light. The Maxwell equations already obey the symmetries
of special relativity. The formulation of the Maxwell equations in terms of potentials rather than elds
is especially useful, because these potentials show up in the matter equations (e.g., in the Schrodinger
and Klein-Gordon equations).
The Maxwell equations, as well, can be derived from a least action principle, and, therefore, a La-
grangian formulation of the Maxwell theory may be given. A classical, point-like charged particle in an
electromagnetic eld is described by the non-relativistic or relativistic Lorentz equations, respectively.
Again, a Lagrangian and Hamiltonian formulation for the Lorentz equations can be found.
4.1 Maxwell Equations and Potentials. Lagrangian Density
In 1862, Maxwell completed the Coulomb, Faraday and Biot-Savart-Laplace equations by the dis-
placement current and wrote the complete system of classical electrodynamics. In Gaussian units it
reads
_
_
div E(t, x) = 4(t, x), rot E(t, x) =
1
c
B(t, x),
div B(t, x) = 0, rot B(t, x) =
4
c
j(t, x) +
1
c
E(t, x),
(t, x) IR
4
. (4.1)
where (t, x) and j(t, x) stand for the charge and current density, respectively.
Remark 4.1 The Maxwell equations imply the charge continuity equation,
(t, x) + div j(t, x) = 0, (t, x) IR
4
. (4.2)
Potentials and gauge invariance
Let us introduce scalar and vector potentials to rewrite (4.1) in a relativistic covariant four-dimensional
form. Namely, div B(t, x) = 0 implies that B(t, x) = rot A(t, x). Then rot E(t, x) =
1
c
B(
t, x)
implies rot [E(t, x) +
1
c
A(t, x)] = 0 hence E(t, x) +

1
c
A(t, x) =
x
(t, x). Finally,
B(t, x) = rot A(t, x), E(t, x) =
x
(t, x)
1
c
A(t, x), (t, x) IR

4
. (4.3)
The justication of all these relations follows from the Fourier transform. The choice of the potentials
is not unique since the gauge transformation
(t, x) (t, x) +
1
c
(t, x), A(t, x) A(t, x)
x
(t, x) (4.4)
does not change the elds E(t, x), B(t, x) for any function (t, x) C
1
(IR
4
). Therefore, it is possible
to satisfy an additional gauge condition. Let us choose for example the Lorentz gauge
1
c
(t, x) + div A(t, x) = 0, (t, x) IR

4
. (4.5)
Exercise 4.2 Prove the existence of the potentials satisfying (4.5). Hint: Derive an equation for the
function (t, x).
4. LAGRANGIAN THEORY FOR THE MAXWELL FIELD 35
Let us express the Maxwell equations (4.1) in terms of the potentials. Substituting Eq. (4.3) into
the rst Maxwell equation leads to 4(t, x) = div E(t, x) = (t, x)
1
c
div

A(t, x). Eliminating
div

A(t, x) by the dierentiation of (4.5) in time,
1
c
(t, x) + div

A(t, x) = 0, we get
2(t, x) := [
1
c
2
2
t
](t, x) = 4(t, x), (t, x) IR
4
. (4.6)
Similarly, substituting Eq. (4.3) into the last Maxwell equation, we get
rot rot A(t, x) =
4
c
j(t, x) +
1
c
E(t, x) =
4
c
j(t, x)
1
c

x

(t, x)
1
c
2
A(t, x). (4.7)

Exercise 4.3 Prove the identity
rot rot = +
x
div . (4.8)
Substituting Eq. (4.8) into (4.7) and eliminating
1
c

x

(t, x) by application of
x
to (4.5), we get
2A(t, x) =
4
c
j(t, x), (t, x) IR
4
. (4.9)
Remark 4.4 The arguments above show that the Maxwell equations (4.1) are equivalent to the system
of two wave equations (4.6), (4.9) for the potentials with the Lorentz gauge condition (4.5).
4D vector potential
Let us introduce the four-dimensional notations
_
x
0
= ct, x
= (x
0
, ..., x
3
),
= (
0
,
1
,
2
,
3
),
x
:= g
= (x
0
, x
1
, x
2
, x
3
),
:= g
= (
0
,
1
,
2
,
3
),
(4.10)
where g
= g
= diag(1, 1, 1, , 1). Let us also introduce the four-dimensional elds and currents
_
_
_
/
(x) = ((t, x), A(x)), /
(x) := g
(x) = ((x), A(x)),
(x) = ((x),
1
c
j(x)),
(x) := g
(x) = ((x),
1
c
j(x)).
x IR
4
(4.11)
Then the Maxwell equations (4.6), (4.9) become
2/
(x) = 4
(x), x IR
4
. (4.12)
Similarly, the charge continuity equation (4.2), gauge transformation (4.4) and Lorentz gauge (4.5)
become
(x) = 0, /
(x) /
(x) +
(x),
(x) = 0, x IR
4
. (4.13)
Tensor eld
Denition 4.5 The Maxwell tensor is dened by
T
(x) =
(x)
(x), x IR
4
. (4.14)
36
Exercise 4.6 Check the formula
(T
(x))
3
,=0
=
_
_
_
_
_
0 E
1
(x) E
2
(x) E
3
(x)
E
1
(x) 0 B
3
(x) B
2
(x)
E
2
(x) B
3
(x) 0 B
1
(x)
E
3
(x) B
2
(x) B
1
(x) 0
_
_
_
_
_
(4.15)
Hint: Use the formulas (4.3) and (4.10), (4.11).
Proposition 4.7 The Maxwell equations (4.12) are equivalent to
(x) = 4
(x), x IR
4
. (4.16)
Proof The Maxwell tensor does not depend on the choice of gauge since
(x)
(x) = 0.
Therefore, we can assume the Lorentz gauge (4.5) without loss of generality. Then (4.12) implies
(x)=
(x)
(x))=
(x)=2/
(x)=4
(x), x IR
4
. (4.17)
Lagrangian density
Denition 4.8 The Lagrangian density for the Maxwell equations (4.16) with given external charge-
current densities
(x) is dened by
L(x, /
, /
) =
1
16
T
(x)/
, (x, /
, /
) IR
4
IR
4
IR
16
. (4.18)
where T
:=
and T
:=
.
Proposition 4.9 The Maxwell equations (4.16) with given charge-current densities
(x) are equiv-

alent to the Euler-Lagrange equations (2.9) with the Lagrangian density (4.18) for the elds /
(x).
Proof T
is a quadratic form in /. Therefore, the canonically conjugate elds
with com-
ponents
are given by
:=
L =
1
8
T
=
1
8
(T
) =
1
4
T
. (4.19)
Therefore,
(x) =
1
4

(x), x IR
4
. (4.20)
On the other hand,
A
L =
(x), hence (4.16) is equivalent to (2.9).

4.2 Lagrangian for Charged Particle in Maxwell Field
In a continuous Maxwell eld the motion of a charged particle with small velocity [ x(t)[ c is
governed by the Lorentz equation (L) from the Introduction,
x(t) = e[E(t, x(t)) +
1
c
x(t) B(t, x(t))], t IR, (4.21)
where is the mass of the particle and e its charge. For large velocities [ x(t)[ c the equation must
be replaced by
p(t) = e[E(t, x(t)) +
1
c
x(t) B(t, x(t))], t IR, (4.22)
4. LAGRANGIAN THEORY FOR THE MAXWELL FIELD 37
where p := x/
_
1 ( x/c)
2
is the relativistic momentum of the particle (see (38.6)). Let us assume
that the elds E, B are C
1
vector functions in IR
4
. Then the dynamical equations (4.21), (4.22) dene
the corresponding dynamics uniquely. Let us dene the charge-current densities
(t, x) corresponding
to the trajectory x(): by (4.11),
0
(t, x) = e(x x(t)),
k
(t, x) =
1
c
x
k
(t)e(x x(t)), k = 1, 2, 3. (4.23)
Let us show that the dynamical equations (4.21), (4.22) automatically follow from the Hamilton
LAP applied to the Lagrangian density (4.18) with xed elds E(t, x), B(t, x). We consider (4.21)
for concreteness. Let /(t, x) be the 4-potential corresponding to the elds E(t, x), B(t, x). The
Lagrangian density (4.18) consists of two parts: the eld part L
f
and the eld-matter interaction part
L
fm
. Substituting (4.23) into L
fm
, we get the eld-matter action in the form
S
T
fm
: =
_
T
0
_
IR
3
(t, x)/
(t, x)dxdt
= e
_
T
0
_
(t, x(t))
1
c
x(t) A(t, x(t))
_
dt. (4.24)
The interaction term corresponds to the following Lagrangian function for the nonrelativistic particle,
L(x, v, t) =
v
2
2
e(t, x) +
e
c
v A(t, x). (4.25)
Theorem 4.10 Let the potential /(t, x) be xed. Then the Lorentz equation (4.21) for the trajectory
x() is equivalent to the Euler-Lagrange equations corresponding to the Lagrangian (4.25)
Proof First let us evaluate the momentum. By denition, p:= L
v
= v +
e
c
A(t, x), hence
p(t) := L
v
(x(t), x(t), t) = x(t) +
e
c
A(t, x(t)). (4.26)
Now (36.15) becomes,
p
k
(t) = L
x
k
(x(t), x(t), t) = e
k
(t, x) +
e
c
x
k
A(t, x)], k = 1, 2, 3. (4.27)
Let us calculate the derivative on the LHS:
p
k
(t) =
d
dt
( x
k
(t) +
e
c
A
k
(t, x(t))) = x
k
+
e
c
[

A
k
(t, x) +
j
A
k
(t, x) x
j
]. (4.28)
Substituting this expression into the LHS of (4.27), we get
x
k
+
e
c
[

A
k
(t, x) +
j
A
k
(t, x) x
j
] = e
k
(t, x) +
e
c
x
j
k
A
j
(t, x). (4.29)
Let us rewrite this as follows:
x
k
= e[
k
(t, x)
1
c
A
k
(t, x)] +
e
c
x
j
[
k
A
j

j
A
k
]. (4.30)
The rst square bracket on the RHS is E(t, x) by (4.3). Hence, it remains to check that x
j
[
k
A
j

j
A
k
] = xrot A(t, x). Let us note that
k
A
j
j
A
k
= (rot A)
l
kjl
where
kjl
is the antisymmetric
tensor. Therefore, x
j
[
k
A
j

j
A
k
] = x
j
(rot A)
l
kjl
= [ x rot A(t, x)]
k
by denition of the vector
product.
Remark 4.11 The derivation of the expression for the Lorentz force from the Maxwell equations is not
very surprising since the expression also follows from the Coulomb and Biot-Savart-Laplace equations.
38
4.3 Hamiltonian for Charged Particle in Maxwell Field
Nonrelativistic particle
We evaluate the Hamilton function as the Legendre transform of the nonrelativistic Lagrangian (4.25):
rst,
H := pv L = pv
v
2
2
+e(t, x)
e
c
v A(t, x) = e(t, x) +v(p
e
c
A(t, x))
v
2
2
. (4.31)
Next, we eliminate v by the relation p
e
c
A(t, x) = v. Then we get nally,
H = e(t, x) +
v
2
2
= e(t, x) +
1
2
(p
e
c
A(t, x))
2
. (4.32)
Relativistic particle
Let us consider the relativistic Lagrangian (cf. (38.6))
L(x, v, t) = c
2
_
1
2
e(t, x) +
e
c
v A(
t, x), (4.33)
where := [v[/c. Let us note that the rst term on the RHS is asymptotically c
2
+ v
2
/2 for
1. First we evaluate the momentum: by denition, p:= L
v
= v/
_
1
2
+
e
c
A(t, x), hence
H : = pv L = pv +c
2
_
1
2
+e(t, x)
e
c
v A(t, x)
= e(t, x) +v(p
e
c
A(t, x)) +c
2
_
1
2
. (4.34)
Next, we eliminate p by the relation p
e
c
A(t, x) = v/
_
1
2
. Then we get,
H = e(t, x) +
v
2
_
1
2
+c
2
_
1
2
= e(t, x) +
c
2
_
1
2
= e(t, x) +c
2
_
1 + (p
e
c
A(t, x))
2
/(c)
2
. (4.35)
We can rewrite this relation in the following standard form
(H/c
e
c
(t, x))
2
=
2
c
2
+ (p
e
c
A(t, x))
2
. (4.36)
Remark 4.12 This expression coincides with (1.10) for = 0, A = 0.
Part II
Schrodinger Equation
39
5. GEOMETRIC OPTICS AND SCHR
ODINGER EQUATION 41
5 Geometric Optics and Schrodinger Equation
In this lecture we show the wave propagation along straight lines for the simple cases of free Klein-
Gordon and Schrodinger equations. For the Schrodinger equation coupled to the Maxwell eld we
analyze the propagation along rays. The proof is based on constructing a formal Debye expansion.
Wave equations of type (4.6) describe the wave propagation in electrodynamics, acoustics and many
other elds. They describe well the diraction and the interference of waves. On the other hand, the
wave processes also demonstrate the straight-line propagation of waves, thereby justifying geometric
optics. The mathematical description of this feature by the wave equation has been discovered by
Hamilton around 1830 and developed by Liouville in 1837, Debye in 1911, Rayleigh in 1912, Jereys
in 1923, and Schrodinger, Wentzel, Kramers and Brillouin in 1926.
5.1 Straight Line Propagation for the Free Equations
Let us analyze the straight line propagation in the concrete example of the free Klein-Gordon equation
_

(t, x) = (t, x)
2
(t, x)
(0, x)=
0
(x),

(0, x)=
0
(x)
(t, x) IR IR
3
. (5.1)
Let us choose the initial data
0
(x),
0
(x) from the Schwartz space o(IR
3
) of test functions.
Denition 5.1 o(IR
3
) is the space of functions (x) C
(IR
3
) such that
sup
xIR
3
(1 +[x[)
N
[
x
(x)[ < (5.2)
for any N = 1, 2, ... and multiindices = (
1
,
2
,
3
).
Denition 5.2 For o(IR
3
) the Fourier transform is dened by
F(k) :=

(k) := (2)
3
_
IR
3
e
ikx
(x)dx, k IR
3
. (5.3)
Proposition 5.3 Let
0
,
0
o(IR
3
). Then the Cauchy problem (5.1) admits a unique solution
(t, x) satisfying the bounds
sup
xIR
3
(1 +[x[)
N
[
0
t

x
(t, x)[ < C(
0
, , N)(1 +[t[)
N
, t IR (5.4)
for any N = 1, 2, ...,
0
= 0, 1, 2, ... and multiindices = (
1
,
2
,
3
).
Proof Let us calculate the solution to (5.1) with
0
,
0
o(IR
3
) by using the Fourier transform. Let
us apply the transform to Equations (5.1) using the well-known formulas
(F
1
)(t, k)=ik
1

(t, k), (F)(t, k)=k
2

(t, k), (t, k) IR IR
3
. (5.5)
The bounds (5.4) imply also that (F

)(t, k) =

(t, k) and (F

)(t, k) =

(t, k) hence (5.1) becomes

_
_
_
(t, k) = k
2

(t, k)
2

(t, k)
(0, k)=

0
(k),

(0, k)=
0
(k)
(t, k) IR IR
3
. (5.6)
42
This is the Cauchy problem for an ordinary dierential equation which depends on the parameter
k IR
3
. The solution is well-known,
(t, k) =

0
(k) cos t +
0
(k)
sin t
, = (k) :=
_
k
2
+
2
. (5.7)
Therefore, the solution (t, x) is given by the inverse Fourier transform,
(t, x) =
_
IR
3
e
ixk
_
0
(k) cos t +
0
(k)
sin t
_
dx
=
1
2
_
IR
3
e
ixk
_
e
it
_
0
(k) +

0
(k)
i
_
+e
it
_
0
(k)

0
(k)
i
__
dk
=
+
(t, x) +
(t, x). (5.8)

The representation obviously implies the bounds (5.4) since

0
,
0
o(IR
3
).
Now let us choose the initial functions
0
,
0
with a localized spectrum, i.e.
supp

0
, supp
0
B
r
(k
), (5.9)
where B
r
(k
) is an open ball with center k
IR
3
and a small radius r [k
[. Then the same is true

for the spectra of the functions
(t, x) by (5.8):
supp

(t, ) B
r
(k
), t IR. (5.10)
The solutions of type
(t, x) are called wave packets.

Theorem 5.4 Let
0
,
0
o(IR
3
), and (5.9) holds. Then the corresponding wave packets
(t, x)
are localized solutions moving with the group velocities v
= (k
) in the following sense:

i) For a constant a > 0 and any N > 0,
[
(t, x)[ C
N
(1 +[t[ +[x[)
N
, [x v
t[ >
ar
[k
[
[t[. (5.11)
ii) For any constant A > 0,
[
(t, x)[ C(1 +[t[)

3/2
, [x v
t[
Ar
[k
[
[t[. (5.12)
Proof Let us prove the theorem for
+
since for
just the same arguments hold. Let us consider

the function
+
(t, x) along a ray x = vt with an arbitrary v IR
3
: by (5.8),
+
(t, vt) =
_
e
i
+
(k)t
+
(k)dk. (5.13)
Here the phase function is given by
+
(k) := vk (k) and the amplitude
+
(k) :=

0
(k)/2 +

0
(k)/(2i(k)). Let us apply the method of stationary phase [32] to the integral (5.13). Then we get
that the asymptotics for t depends on the existence of the critical points k supp
+
of the
phase function
+
(k),
+
(k) = v (k) = 0, k supp
+
. (5.14)
In other words, v = (k) with a k supp
+
. Now let us take into account that supp
+
B
r
(k
)
by (5.9). Then the system (5.14) admits a solution i v V
r
(k
) := (k) : k B
r
(k
). Now let
us analyze two distinct situations separately.
ODINGER EQUATION 43
i) First let us consider v , V
r
(k
). Then the asymptotics of the integral (5.13) is [t[

N
for any N > 0
which corresponds to (5.11). Let us deduce the asymptotics by a partial integration in (5.13) with the
help of an obvious identity
e
i
+
(k)t
=
D
it
e
i
+
(k)t
, k supp
+
, (5.15)
where D is the following dierential operator, D = ((k)/[(k)[
2
). It is important that (k) ,=
0 for k supp
+
since v , V
r
(k
). Applying the identity (5.15) N times in (5.13), we get by partial

integration,
+
(t, vt) = (it)
N
_
D
N
e
i
+
(k)t
+
(k)dk = (it)
N
_
e
i
+
(k)t
(D
)
N
+
(k)dk, (5.16)
where D
(k) = [(k)(k)/[(k)[
2
] is the adjoint operator to D. Therefore,
+
(t, vt) C
N
(v)(1 +[t[)
N
, t IR (5.17)
if v , V
r
(k
). Hence, for any B > 0,
+
(t, vt) C
N
(B)(1 +[t[ +[vt[)
N
, t IR (5.18)
if v , V
r
(k
) and [v[ B.
Exercise 5.5 Prove that the bounds (5.18) hold for all v , V
r
(k
) with a uniform constant C

N
instead of C
N
(B).
Hint: Consider the function
+
(t, x) along the ray t = w, x = v with [w[ < 1, [v[ = 1 and apply
the partial integration with the phase function kv w(k) and large parameter .
Finally, the bounds (5.18) with C
N
instead of C
N
(B) imply (5.11) since the diameter of the set V
r
(k
)
does not exceed ar/[k
[. The last fact follows from the bound [(k)[ a/[k[ which is obvious
since
(k) =
k
(k)
. (5.19)
ii) It remains to consider v V
r
(k
). This means that v = (k) with a point k B

r
(k
) which is a
solution to the system (5.14). Then the integral (5.13) is called the Fresnel integral and its asymptotics
is [t[
3/2
(see [32]).
Remark 5.6 The asymptotics (5.11), (5.12) means that the energy of the eld
(t, x) outside a ball

[xv
t[ ar[t[/[k
[ decays rapidly, while inside it is about constant since the energy is a quadratic
form. Therefore, the wave packet
(t, x) of the free Klein-Gordon equation (5.1) moves like a free

particle of the size O(r[t[/[k
[) and with the group velocity v
= (k
). Formula (5.19) means

that the velocity v
corresponds to the relativistic momentum k
of the particle.
Exercise 5.7 Analyze the wave packet propagation for the free Schr odinger equation
_
_
i

(t, x) =
1
2
(t, x)
(0, x) =
0
(x),
(t, x) IR IR
3
. (5.20)
Prove that the packets move like free non-relativistic particles of the size O(r[t[/[k
[) with mass m
and momentum k
.
Hint: = k
2
/2m, hence the group velocity v equals = k.
44
5.2 WKB Asymptotics for Schrodinger Equation with a Maxwell Field
Let us write the Lorentz equation (4.21) in Hamilton form with the Hamiltonian (4.32):
_
_
x = H
p
(x, p, t) =
1
(p
e
c
A(t, x)),
p =H
x
(x, p, t) = e
x
(t, x) +
e
c
A
x
(t, x),
(5.21)
where e is the charge of the particle and its mass. E.Schrodinger associated with the Hamilton
system the wave equation (2.2):
(ih
t
e(t, x))(t, x) =
1
2
(ih
e
c
A(t, x))
2
(t, x), (t, x) IR
4
. (5.22)
Let us demonstrate that the short-wave solutions to (5.22) are governed by the Hamilton equations
(5.21). More precisely, let us consider the Cauchy problem for (5.22) with the initial condition
[
t=0
= a
0
(x)e
iS
0
(x)/h
, x IR
3
, (5.23)
where S
0
(x) is a real function. Let us denote by (x(t, x
0
), p(t, x
0
)) the solution to the Hamilton
equations (5.21) with the initial conditions (38.8):
x[
t=0
= x
0
, p[
t=0
= S
0
(x
0
). (5.24)
The solution exists for [t[ < T(x
0
) where T(x
0
) > 0.
Denition 5.8 The curve x = x(t, x
0
) is the ray of the Cauchy problem (5.22), (5.23) starting at
the point x
0
.
Denition 5.9 The ray tube or ray beam emanating from the initial function (5.23) is the
set T = (t, x(t, x
0
)) IR
4
: [t[ < T(x
0
), x
0
supp a
0
.
The following theorem means roughly speaking that for h
1 the set T is the support of the

wave function (t, x), the solution to the Cauchy problem (5.22), (5.23). Let us assume that the
potentials (t, x), A(t, x) C
(IR
4
) and a
0
, S
0
C
(IR
3
). Then the map x
0
x(t, x
0
) is a local
C
-dieomorphism of IR
3
for small [t[. We will construct the formal Debye expansion
(t, x)
_
k=0
h
k
a
k
(t, x)
_
e
iS(t, x)/h
, h
0. (5.25)
Theorem 5.10 Let the map x
0
x(t, x
0
) be a dieomorphism of IR
3
for [t[ < T. Then the formal
expansion (5.25) exists for [t[ < T and is identically zero outside T , i.e.
a
k
(t, x) = 0, (t, x) , T , [t[ < T, k = 0, 1, 2, ... (5.26)
Proof First, let us dene the phase function S(t, x) as the solution to the Cauchy problem (38.7) with
N = 3:
_
S(t, x) = H(x, S(t, x), t), x IR

3
, [t[ < T,
S[
t=0
= S
0
(x), x IR
3
.
(5.27)
ODINGER EQUATION 45
The solution exists by Theorem 38.10. Further, let us substitute (t, x) = a(t, x)e
iS(t, x)/h
into the
Schrodinger equation (5.22). Then Equation (5.27) implies the following transport equation for the
amplitude a(t, x):
_
_
a(t, x) =
1
m
[S(t, x)
e
c
A(t, x)] a(t, x)
+
1
2
[S(t, x) + A(t, x)]a(t, x) +
ih
2
a(t, x)
=: La(t, x) +B(t, x)a(t, x) +
ih
2
d(t, x), [t[ < T,
a[
t=0
= a
0
(x), x IR
3
,
(5.28)
where L is the rst order dierential operator La(t, x) :=
1
m
[S(t, x)
e
c
A(t, x)] a(t, x), B(t, x) is
the function
1
2
[S(t, x) + A(t, x)] and d(t, x) := a(t, x). Let us note that (5.21) implies
a(t, x) +La(t, x) =
d
dt
a(t, x(t, x
0
)). (5.29)
Let us express all functions in ray coordinates (t, x
0
): a(t, x
0
) := a(t, x(t, x
0
)),

B(t, x
0
) := B(t, x(t, x
0
))
etc. Then (5.28) can be rewritten as
_
_
d
dt
a(t, x
0
) =

B(t, x
0
) a(t, x
0
) +
ih
d(t, x
0
), [t[ < T,
a(0, x
0
) = a
0
(x
0
), x
0
IR
3
.
(5.30)
Let us substitute in the rst equation the formal expansion a(t, x
0
)
k=0
h
k
a
k
(t, x
0
). Equating
formally the terms with identical powers of h
, we get the recursive transport equations

d
dt
a
0
(t, x
0
) =

B(t, x
0
) a
0
(t, x
0
),
d
dt
a
1
(t, x
0
) =

B(t, x
0
) a
1
(t, x
0
) +
i
2
d
0
(t, x
0
),
...
d
dt
a
k
(t, x
0
) =

B(t, x
0
) a
k
(t, x
0
) +
i
2
d
k1
(t, x
0
),
...
[t[ < T, (5.31)

where

d
0
(t, x
0
) is the function d
0
(t, x) := a
0
(t, x) expressed in ray coordinates, etc. It remains to
substitute the same expansion into the initial conditions (5.30) which gives
a
0
(0, x
0
) = a
0
(x
0
), a
1
(0, x
0
) = 0, ..., a
k
(0, x
0
) = 0, ... (5.32)
Now (5.31) and (5.32) imply that a
0
(t, x
0
) = 0, [t[ < T, if x
0
, supp a
0
. Hence, also

d
0
(t, x
0
) = 0,
[t[ < T, if x
0
, supp a
0
. Therefore, (5.31) and (5.32) imply that a
1
(t, x
0
) = 0, [t[ < T, if x
0
, supp a
0
,
etc.
46
Exercise 5.11 Prove the transport equation (5.28). Hint: Substitute = ae
iS/h
into the Schr odinger

equation (5.22) and divide by e
iS/h
. Then Equation (5.27) formally follows by setting h
= 0, and
afterwords, (5.28) also follows.
6. SCHR
ODINGER EQUATION AND HEISENBERG REPRESENTATION 47

6 Schrodinger Equation and Heisenberg Representation
According to Schrodinger, the quantum mechanical electron in an external electromagnetic eld is
described by the corresponding Schrodinger equation for the electron eld. The results of the last
section then imply that, in the classical limit, a ray of electrons propagates according to the Lorentz
dynamics. The symmetries of the Schrodinger equation lead to conserved physical observables of
the electron eld, like energy, momentum, angular momentum, and electric charge. All observables
are quadratic forms of linear operators in the Hilbert space of electron wave functions, therefore the
quantum observables are identied with linear operators in this Hilbert space.
6.1 Electrons and Cathode Rays
The electron was discovered by J.J.Thomson around 1897 in cathode rays. He systematized obser-
vations of the deection of rays in Maxwell elds. The observations demonstrated that the deec-
tions may be described by the Lorentz dynamics (4.21) or (5.21), with a xed ratio e/ close to
its present value. Later Kaumann [58] conrmed these observations with high accuracy, e/
1.76 10
8
C/g. The charge of the electron can be evaluated from Faradays electrolytic law to be
e 96500C/6.06 10
23
1.6 10
19
C 4.77 10
10
esu. In 1913 Millikan conrmed this result,
e 1.60 10
19
C 4.77 10
10
esu, leading to an electron mass of 9.1 10
28
g. Therefore, by
Theorem 5.10, the short-wave cathode rays can also be described by the Schrodinger equation (5.22)
with the electron mass and the negative electron charge e < 0:
[ih
t
e(t, x)](t, x) =
1
2
[ih
e
c
A(t, x)]
2
(t, x), (t, x) IR
4
. (6.1)
Remark 6.1 Let us note that the charge of the cathode rays is not xed, hence the equation (6.1)
describes the short-wave cathode rays of an arbitrary charge. It means that the Schrodinger equation
describes rather the dynamics of the electron eld, with a xed ratio e/, than the dynamics of a
particle with the charge e.
Remark 6.2 Theorem 5.10 means that the Schrodinger equation (6.1), with any small h
1, agrees
with the classical Lorentz equations (4.21) or (5.21). The actual value of the constant is xed by the
Planck relation h
= a/k 1.05 10
27
erg sec. Here a is the parameter in the Wien experimental
formula (1.4) and k is the Boltzmann constant. This identication of a small parameter h
in the
Schrodinger equation follows by the experimental and theoretical development from the Kirchho
spectral law to de Broglies wave-particle duality.
6.2 Quantum Stationary States
Let us consider the case of a static Maxwell eld with the potentials (t, x) (x) and A(t, x) A(x).
The corresponding Schrodinger equation (6.1) becomes,
ih
t
(t, x) = H(t, x) :=
1
2
[ih
e
c
A(x)]
2
(t, x) +e(x)(t, x), (t, x) IR
4
. (6.2)
Then the energy is conserved (see (3.1)):
_
IR
3
_
1
2
[[ih
e
c
A(x)](t, x)[
2
+e(x)[(t, x)[
2
_
dx=E, t IR. (6.3)
48
Denition 6.3 Quantum Stationary States for Equation (6.2) are nonzero nite energy solutions
of the type
(t, x) =
(x)e
it
(6.4)
with an IR.
Substituting (6.4) into (6.2), we get the stationary Schr odinger equation
h
(x) = H
(x), x IR
3
, (6.5)
which constitutes an eigenvalue problem. Substituting (6.4) into (6.3) and using (6.5), we get for the
energy
E = E
= h
_
IR
3
[
(x)[
2
dx < . (6.6)
Hence, the quantum stationary states correspond to the eigenfunctions
from the Hilbert space

c := L
2
(IR
3
). Let us assume the standard normalization condition (see Remark 7.1)
_
IR
3
[
(x)[
2
dx = 1. (6.7)
Then (6.6) becomes (cf. (P) from the Introduction)
E
= h
, (6.8)
and (6.5) takes the form
E
(x) = H
(x), x IR
3
. (6.9)
6.3 Four Conservation Laws for Schrodinger Equation
Let us state the four classical conservation laws for the Schrodinger equation (6.2). We adjust the
conserved quantities by some factors.
I. Energy Conservation: The total energy is conserved if the Maxwell potentials do not depend on
time t, i.e. (t, x) = (x), A(t, x) = A(x):
E(t) :=
_
_
1
2
[[ih
e
c
A(x)](t, x)[
2
+e(x)[(t, x)[
2
_
dx = const, t IR. (6.10)
This follows from Theorem 3.4 by (3.1) and (40.17) .
II. Momentum Conservation The component p
n
, n = 1, 2, 3, of the total momentum, is conserved
if the Maxwell potentials do not depend on x
n
:
p
n
(t) := ih
_
_
(t, x)
n
(t, x)
_
This follows from Theorem 3.8 by (3.16) and (40.19).
III. Angular Momentum Conservation Let us assume that the Maxwell elds are axially sym-
metric with respect to rotations around the axis x
n
, i.e.
(t, R
n
(s)x) (t, x) and A(t, R
n
(s)x) R
t
n
(s)A(t, x), s IR. (6.12)
Then corresponding component of the total angular momentum is conserved:
L
n
(t) := ih
_
_
(t, x) (x )
n
(t, x)
_
6. SCHR

This follows from Theorem 3.13 by (3.23) and (40.21) since the Lagrangian density (2.6) satises the
rotation-invariance condition (3.21) by (6.12).
IV. Charge Conservation The total charge is conserved:
Q(t) := e
_
[(t, x)[
2
This follows from Theorem 3.18 by (3.25) and (40.21).
Exercise 6.4 Check the rotation-invariance condition (3.21) for the Maxwell potentials satisfying
(6.12).
Exercise 6.5 Check that the axial symmetry condition (6.12) is equivalent to the corresponding axial
symmetry of the Maxwell eld: for every s IR,
E(t, R
n
(s)x) R
n
(s)E(t, x), B(t, R
n
(s)x) R
n
(s)B(t, x). (6.15)
6.4 Quantum Observables and Heisenberg Representation
The charge conservation (6.14) implies that the complex Hilbert space c := L
2
(IR
3
) is invariant with
respect to the dynamics dened by the Schrodinger equation. For
1,2
c, let us denote by
1
,
2
)
the Hermitian scalar product:
1
,
2
) :=
_

1
(x)
2
(x)dx. (6.16)
All conserved quantities E, p
n
, L
n
, Q are quadratic forms in the phase space c:
E = ,

E), p
n
= , p
n
), L
n
= ,

L
n
), Q = ,

Q). (6.17)
Here

E, p
n
,

L
n
,

Q stand for the corresponding operators of the quadratic forms:
E = H, p
n
= ih
n
,

L
n
= ih
(x )
n
,

Q = I. (6.18)
This motivates the following
Denition 6.6 i) A quantum observable is a linear operator

M in the Hilbert space c.
ii) The quadratic form M() := ,

M) is a mean value of the observable at the state c.
Example 6.7 The multiplication operator x
k
= x
k
(x) is the quantum observable of the k-th co-
ordinate.
Example 6.8 The (squared) absolute value of the total angular momentum is dened by

L
2
:=

L
2
1
+
L
2
2
+

L
2
3
.
Exercise 6.9 Prove that all mean values ,

M) of a quadratic form are real i the operator

M is
symmetric.
Exercise 6.10 Check that all the operators

E, p
n
,

L
n
are symmetric. Hint: All the mean values are
real since they coincide with the total energy, momentum, etc, dened by the real Lagrangian density
(2.6).
50
Exercise 6.11 Check the formula
L
n
= ih
n
, (6.19)
where
n
is the angle of rotation around the vector e
n
in a positive direction. Hint: Consider n = 3
and choose the polar coordinate in the plane x
1
, x
2
.
Now let us assume that (t, x) = (x), A(t, x) = A(x) and denote by U(t), t IR, the dynamical
group of the Schrodinger equation in the Hilbert space c: by denition, U(t)(0, ) = (t, ) for any
solution (t, x) to the equation. The Schrodinger equation implies that
U(t) =
i
h
HU(t) =
i
h
U(t)H. (6.20)
Exercise 6.12 Prove (6.20).
The charge conservation (6.14) means that U(t) is a unitary operator for all t IR.
Let us note that
E = H =
1
2
[ p
e
c
A( x)]
2
+e( x),

L = x p, [ x
k
, p
n
] = ih
kn
, (6.21)
where [, ] stands for the commutator [A, B] := AB BA.
Exercise 6.13 Check the identities (6.21). Hint: For polynomial potentials A(x), (x), the operator
of multiplication by A(x)) resp. (x) coincides with A( x)) resp. ( x).
Denition 6.14 The Heisenberg representation of a quantum observable

M is an operator func-
tion

M(t) := U(t)

MU(t), t IR.
The denition implies that
M((t)) = (0),

M(t)(0)), t IR (6.22)
for any solution (t) := (t, ) to the the Schrodinger equation.
Exercise 6.15 Check (6.22).
Exercise 6.16 Check the identities
E(t)=H=
1
2
[ p(t)
e
c
A( x(t))]
2
+e( x(t)),

L(t)= x(t) p(t), [ x
k
(t), p
n
(t)] =ih
kn
. (6.23)
Hint: First consider polynomial potentials A(x), (x) and use (6.21) together with the commutation
HU(t) = U(t)H from (6.20).
Exercise 6.17 Check the identities
[ x
k
(t), p
N
n
(t)]=ih
kn
N p
N1
n
, [A( x(t)), p
n
(t)]=ih
kn
A
x
n
( x(t)), [( x(t)), p
n
(t)]=ih
kn
x
n
( x(t)), (6.24)
for any N = 1, 2, .... Hint: First consider polynomial potentials A(x), (x) and apply the last formula
of (6.23) by induction.
6. SCHR

Let us derive a dynamical equation for

M(t): (6.20) implies the Heisenberg equations
M(t) =
i
h
[H,

M(t)] =
i
h
[

E(t),

M(t)], t IR. (6.25)
Lemma 6.19 The mean value, M((t)), is conserved if [H,

M] = 0.
Proof By (6.2),
d
dt
M((t)) =

(t),

M(t)) +(t),

M

(t)) =
i
h
H(t),

M(t)) (t),

M
i
h
H(t))
=
i
h
_
H(t),

M(t)) (t),

MH(t))
_
=
i
h
(t), [H,

M](t)), t IR. (6.26)
Exercise 6.20 Prove the angular momentum conservation (6.13), by using Lemma 6.19. Hint:
Check the commutation [H,

L
n
] = 0 under condition (6.12).
Let us write the Heisenberg equations for the observables x(t) and p(t). Then we obtain the system
_
x(t) =
1
( p(t)
e
c
A(t, x(t))),
p(t) = e
x
(t, x(t)) +
e
c
A
x
(t, x(t)).
(6.27)
Exercise 6.21 Check (6.27). Hint: Use (6.25) and (6.24) with N = 1, 2.
Let us note that Equations (6.27) formally coincide with the Hamilton system (5.21).
52
7 Coupling to the Maxwell Equations
The simultaneous evolution of the full system of electron wave function and Maxwell eld is determined
by the coupled Schrodinger and Maxwell equations. Stationary states of this coupled system would
provide natural candidates for physical states and transitions, but the existence of these stationary
states as well as the transition of general solutions to these stationary states for large times is an
open problem. Approximate, iterative solutions to the coupled system are provided by the Born
approximation.
7.1 Charge and Current Densities and Gauge Invariance
Let us determine the dynamics of the Maxwell eld in presence of the wave eld governed by the
Schrodinger equation. The Lagrangian density of the free Maxwell eld is known from (4.18), so we
have to modify only the interaction term in (4.18). The interaction term gives the dynamics of the
Maxwell eld in the presence of given charge-current densities (4.18), so it remains to express the
densities in terms of the wave eld .
On the other hand, we have shown that the interaction term in (4.18) uniquely determines the
Lorentz dynamics (4.21). The Schrodinger equation (6.1) substitutes the Lorentz dynamics (4.21).
This suggests to identify the Maxwell-Schrodinger interaction with the interaction term from the
Lagrangian density L
S
of the Schrodinger equation (6.1). That is, we add the Lagrangian density L
f
of the free Maxwell eld to the Lagrangian density L
S
of the Schrodinger equation (6.1) and get the
Lagrangian density L
MS
of the coupled Maxwell-Schrodinger equations,
L
MS
= [ih
t
e]
1
2
3
k=1
[[ih
e
c
A
k
][
2
1
16
T
, (7.1)
where T
:=
and T
:=
, /
= (, A
1
, A
2
, A
3
). The corresponding
Euler-Lagrange equations read
_
_
[ih
t
e(t, x)](t, x) =
1
2
[ih
e
c
A(t, x)]
2
(t, x),
1
4

(t, x) =
_
_
_
_
:= e[(t, x)[
2
, = 0
j
c
:=
e
c
[ih

e
c
A
(t, x)](t, x) (t, x), = 1, 2, 3

_
_
_
_
.
(7.2)
The system (7.2) describes the dynamics of the wave eld in its own, induced Maxwell eld
(t, x), A(t, x) generated by the charges and currents of the wave eld.
Remark 7.1 i) The charge-current densities from (7.2) coincide, up to a factor, with the Noether
currents (40.23) of the group of internal rotations (39.12).
ii) The factor is choosen in accordance to the normalization condition (6.7). Namely the condition
means that the total charge of the corresponding quantum stationary state is equal to e.
Now let us introduce external potentials
ext
(t, x), A
ext
(t, x) of the Maxwell eld generated by
some external sources. We formalize the introduction through the Lagrangian density
L
MS
=[ih
t
e( +
ext
)]
1
2
3
k=1
[[ih
e
c
(A
k
+A
ext
k
)][
2
1
16
T
. (7.3)
7. COUPLING TO THE MAXWELL EQUATIONS 53
The corresponding equations read
_
_
[ih
t
e((t, x)+
ext
(t, x))](t, x)=
1
2
[ih
e
c
(A(t, x)+A
ext
(t, x))]
2
(t, x),
1
4

(t, x)=
_
_
_
_
:= e[(t, x)[
2
j
c
:=
e
c
[ih
e
c
(A
(t, x)+A
ext
(t, x))](t, x)(t, x)

_
_
_
_
. (7.4)
Remark 7.2 The tensor T
is dened by the potentials (t, x), A(t, x), hence the system admits the
solution (t, x) = 0, (t, x) = 0, A(t, x) = 0 corresponding to the absence of matter.
The gauge transformation (4.4) does not change the Maxwell elds E(t, x), B(t, x) for any function
(t, x) C
1
(IR
4
). Therefore, it would be natural to expect that the solutions to the coupled equations
(7.2), (7.4) also do not change too much under this transformation. Indeed, we can complete the
transformation of the potentials (4.4) with a corresponding transformation of the wave function:
(t, x) (t, x) +
1
c
(t, x), A(t, x) A(t, x)
x
(t, x),
(t, x) e
i
e
ch
(t, x)
(t, x).
(7.5)
It is easy to check that the new functions also provide a solution to Equations (7.2) resp. (7.4).
Moreover, the transformations (7.5) do not change the electric charge and current densities (t, x) and
j(t, x) in (7.2) and (7.4).
7.2 Electron Beams and Heisenbergs Uncertainty Principle
Plane wave as an electron beam
According to the de Broglies wave-particle duality (1.7), a free electron beam is described by a plane
wave
(t, x) = Ce
i(kxt)
, k ,= 0 (7.6)
which is a solution to the free Schrodinger equation, i.e., without an external Maxwell eld. Hence,
h
=
h
2
2
k
2
> 0. (7.7)
We can dene corresponding density of the electrons from the charge density (7.2),
(t, x) := e[(t, x)[
2
= e[C[
2
. (7.8)
The velocities of the electrons may be dened from the electric current density (7.2),
j(t, x) :=
e
[ih
(t, x)]
in
(t, x) =
eh
[C[
2
. (7.9)
Respectively, the electron density, velocity and momentum are dened by
_
_
d(t, x):=(t, x)/e=[C[
2
,
v(t, x):=j(t, x)/(t, x)=
h
,
p(t, x):=v(t, x)=h
k.
(7.10)
54
Then the energy density (see (6.10)) admits the following representation:
e(t, x):=
1
2
[[ih
](t, x)[
2
=
h
2
2
[k[
2
[C[
2
=
v
2
2
d
in
(t, x). (7.11)
Heisenbergs Uncertainty Principle
The interpretation of the plane wave as a beam of free electrons cannot be completed by an assign-
ment of coordinates to the electrons. This assignment is possible only asymptotically, for small h
.
Corresponding bounds were discovered by Heisenberg [44, 75]:
xp h
, (7.12)
where x := , ( xx)
2
)
1/2
resp. p := , ( pp)
2
)
1/2
is an uncertainty of an electron coordi-
nate resp. momentum (see (6.17) and (6.18) for the notations). For the plane wave the momentum is
known exactly, hence p = 0. Now (7.12) implies that and x = , in other words, the coordinate
is not well-dened.
7.3 Quantum Stationary States and Attractors
Let us consider the system (7.4) corresponding to a static external Maxwell eld
ext
(x), A
ext
(x):
_
_
[ih
t
e((t, x)+
ext
(x))](t, x)=
1
2
[ih
e
c
(A(t, x)+A
ext
(x))]
2
(t, x),
1
4

(t, x)=
_
_
_
_
:= e[(t, x)[
2
j
c
:=
e
c
[ih
e
c
(A
(t, x)+A
ext
(x))](t, x)(t, x)
_
_
_
_
. (7.13)
Let us generalize the denition of quantum stationary states for Eq. (7.13).
Denition 7.3 ([40]) i) solitary waves for Eq. (7.13) are nonzero nite energy solutions of the
type
(t, x) =
(x)e
it
, (t, x) =
(x), A(t, x) = A
(x) (7.14)
with an IR.
ii) The set := IR : there exist a solitary wave with (
(x),
(x), A
(x)) ,= 0 is the set of

the solitary eigenfrequencies and / := (
(x),
(x), A
(x)) : is the set of the solitary

amplitudes.
Substituting (7.14) into (7.13), we get the corresponding stationary nonlinear eigenfunction problem:
in the Lorentz gauge we get, as in (4.17),
_
_
[h
e(
(x)+
ext
(x))]
(x)=
1
2
[ih
e
c
(A
(x)+A
ext
(x))]
2
(x),
1
4

(x) = := e[
(x)[
2
,
1
4
A
(x) =
j
c
:=
e
c
[ih
e
c
(A
(x)+A
ext
(x))]
(x)
(x).
(7.15)
The problem is nonlinear in contrast to (6.5). For the coupled nonlinear Dirac-Maxwell equations
with zero external potentials, the existence of solitary waves is proved in [31].
Bohrs quantum transitions (1.6) could be interpreted mathematically as the following convergence
(or attraction) for every nite energy solution (x, t) to the Schrodinger equation:
(t, x)
(x)e
i
t
, t , (7.16)
where h
= E
m
and h
+
= E
n
. This would mean that the set of all quantum stationary states
(x)e
it
is an attractor of the dynamical system dened by the Schrodinger equation (6.2). How-
ever, this is generally impossible for the linear autonomous equation by the principle of superposition.
Namely, a linear combination of two linearly independent stationary states,
(x)e
i
t
,
(t, x) = C
(x)e
i
t
+C
+
+
(x)e
i
+
t
, (7.17)
is a solution to the Schrodinger equation, but the asymptotics (7.16) does not hold if C
1
,= 0 and
C
2
,= 0.
On the other hand, the asymptotics could hold for solutions to the nonautonomous Schrodinger
equation (6.1) with the potentials (t, x), A(t, x) sin(
+
)t. Let us note that just this type of

Maxwell eld results from transition (7.16) according to the second Bohr postulate (see Introduction).
Comment 7.4 Summing up, we suggest that the transitions (7.16) hold for the autonomous coupled
nonlinear Maxwell-Schr odinger equations (7.13): for every nite energy solution ((t, x), (t, x), A(t, x)),
the following long-time asymptotics hold (cf. (0.2))
((t, x), (t, x), A(t, x)) (
(x)e
i
t
,
(x), A
(x)), t , (7.18)
where
IR and
(x), A
(x) are the solutions to the stationary system (7.15) with =
.
This means that the set of all solitary amplitudes / is the point attractor of the coupled equations.
Let us note that an analog of these transitions was proved recently in [59] [68] for some model
nonlinear autonomous wave equations (see the survey [60]): then the asymptotics of type (7.18) hold
in appropriate local energy seminorms. However, the proof of the transitions (7.18) for the coupled
equations (7.13) is still an open problem.
Comment 7.5 The nonlinear eigenvalue problem (7.15) could be handled by the iterative procedure.
Namely, rst let us neglect the elds
(x), A
(x)) in the Schrodinger equation of (7.15) and get the

rst Born approximation
h
(x) =
1
2
[ih
e
c
A
ext
(x)]
2
(x) +e
ext
(x)
(x) (7.19)
which is the linear eigenvalue problem of type (6.5). Further we could express
(x), A
(x) from
the Poisson equations in (7.15) (neglecting A
(x) in the RHS for the rst approximation), substitute

them into the rst equation, and continue by induction. The corrections
(x), A
(x)) are small in

a sense since the RHS of the Poisson equations contain the small factors e and e/c. Hence we could
expect that the iterations converge, and even the solution of the linear problem (7.19) give a good
(rst) approximation to a solution of the nonlinear problem (7.15). If the iterations converge, the
limit functions obviously give one of the solutions to (7.15).
56
Comment 7.6 i) The (orthodox) Copenhagen interpretation of Quantum Mechanics (see [74]) postu-
lates that each observation of a quantum observable

M gives one of its eigenvalue with the probability
one.
ii) This postulate agrees with our previous comment in the case when

Mcommutes with the Schrodinger
operator H (with the external potentials
ext
(x) and A
ext
(x)): in particular, this is true for the en-
ergy

c = H. Namely, between two successive macroscopic observations, the quantum system could
be considered as unperturbed even if the observation provide a disturbance. Therefore, the attraction
(7.18) implies that before the observation the wave function (t) := (, t) is almost sure close to
a solitary wave
e
it
. Furthermore,

k
, where
k
is an eigenfunction of the linear operator
H (see previous comment). On the other hand,
k
is also an eigenfunction of the observable

M.
Hence the (mean) value is equal to M(t) := (t),

M(t))
k
(t),

M
k
(t)) that coincides with the
corresponding eigenvalue.
iii) In the case when

M does not commute with the Schrodinger operator H, this interpretation fails.
One could expect then that the measuring process should modify the Schrodinger operator of the
combined system (observable system + measuring instrument) to insure the commutation.
7.4 Charge Continuity Equations
The Lagrangian density (7.3) is invariant with respect to the internal rotations (39.12). Furthermore,
the densities , j from (7.4) coincide with the corresponding Noether currents of the form (40.23) up
to a factor. Therefore, the Noether theorem II implies
Lemma 7.7 For any solution to the Schr odinger equation
[ih
t
e(t, x)](t, x) =
1
2
[ih
e
c
A(t, x)]
2
(t, x), (7.20)
the corresponding charge-current densities , j satisfy the continuity equation (4.2).
Let us prove a more general relation:
Lemma 7.8 For any two solutions
1
(t, x),
2
(t, x) to the Schr odinger equation in (7.20) the follow-
ing identity holds:

12
(t, x) + div j
12
(t, x) = 0, (t, x) IR
4
, (7.21)
where
_
12
(t, x) := e
1
(t, x)
2
(t, x),
j
12
(t, x) :=
e
2
_
[ih
e
c
A(t, x)]
1
(t, x)
_
2
(t, x)
+
e
2
_
[ ih
e
c
A(t, x)]
2
(t, x)
_
1
(t, x).
(7.22)
Remark 7.9 For
1
=
2
the expressions (7.22) become (7.2) by (2.3), and the identity (7.21)
becomes (4.2).
Proof of Lemma 7.8 Let us write the equation (7.20) for
1
and
2
:
_
_
[ih
t
e(x)]
1
(t, x) =
1
2
[ih
e
c
A(t, x)]
2
1
(t, x),
[ih
t
e(x)]
2
(t, x) =
1
2
[ih
e
c
A(t, x)]
2
2
(t, x)
(t, x) IR
4
. (7.23)
Let us multiply the rst equation by i
2
(t, x) and add the second equation multiplied by i
1
(t, x).
Then we get,
ih
1
(t, x)i
2
(t, x) +ih
2
(t, x)i
1
(t, x)
=
1
2
_
[ih
e
c
A(t, x)]
2
1
(t, x)
_
i
2
(t, x)
1
2
_
[ih
e
c
A(t, x)]
2
2
(t, x)
_
i
1
(t, x). (7.24)
This may be rewritten as
t
[
1
(t, x)
2
(t, x)]
=
1
2
__
[ih
e
c
A(t, x)]
1
(t, x)
_
2
(t, x)
+
_
[ih
e
c
A(t, x)]
2
(t, x)
_
1
(t, x)
_
. (7.25)
Exercise 7.10 Check the identity (7.25).
7.5 Born Approximation
Let us assume that the induced Maxwell eld (t, x), A(t, x) in (7.4) is small compared to the external
eld. Then we can neglect the induced eld in the rst equation and consider the approximate equation
(cf. (7.19))
[ih
t
e
ext
(t, x)](t, x) =
1
2
[ih
e
c
A
ext
(t, x)]
2
(t, x). (7.26)
Its solution is the Born approximation to the wave eld in (7.4). Let us solve the equation and
substitute the solution into the RHS of the second equation of (7.4) also neglecting the induced
Maxwell eld:
1
4

(t, x) =
_
_
_
_
:= e[(t, x)[
2
j
c
:=
e
c
[ih

e
c
A
ext
(t, x)](t, x) (t, x)

_
_
_
_
. (7.27)
Then the solution T
(t, x) is the Born approximation to the Maxwell eld radiated by the atom.
This process of approximations can be iterated.
58
Part III
Application of Schrodinger Theory
59
8. SPECTRUM OF HYDROGEN ATOM 61
8 Spectrum of Hydrogen Atom
For a hydrogen atom we study the corresponding Schrodinger equation and nd the spectrum as the
solution to the eigenvalue problem. The proof is based on splitting the space L
2
(S
2
) in a sum of
orthogonal eigenspaces of the spherical Laplacian.
Let us consider the hydrogen atom, which has precisely one electron of negative charge e. The
Rutherford experiment shows that the positive charge e is concentrated in a very small region called
nucleus, so its Maxwell eld is Coulombic (t, x) = e/[x[. We assume the magnetic eld of the
nucleus to be zero: A(t, x) = 0. Then the Schrodinger equation (6.2) becomes
ih
t
(t, x) = H(t, x) :=
h
2
2
(t, x)
e
2
[x[
(t, x), (t, x) IR
4
, (8.1)
and the stationary Schrodinger equation (6.9) becomes,
E
(x) = H(x)
. (8.2)
Theorem 8.1 The quantum stationary states
c := L
2
(IR
3
) of the hydrogen atom exist for
energies E
= E
n
:= 2h
R/n
2
, where R := e
4
/(4h
3
) is the Rydberg constant and n =
1, 2, 3, .... For other energies, stationary states do not exist.
We will prove the theorem in this lecture.
8.1 Spherical Symmetry and Separation of Variables
Rotation invariance
A basic issue in solving the eigenvalue problem (8.2) is its spherical symmetry which implies the
angular momentum conservation. Namely, the Schrodinger operator H is invariant with respect to all
rotations of the space IR
3
since the Laplacian and the Coulomb potential are invariant under the
rotations. The Schrodinger operator H is invariant with respect to all rotations of the space IR
3
. This
means that for any k = 1, 2, 3,
H
R
k
() =

R
k
()H, IR, (8.3)
where (

R
k
())(x) := (R
k
()x) and R
k
() is a space rotation around the unit vector e
k
with
an angle of radian in a positive direction, e
1
= (1, 0, 0), etc. The commutation holds by (8.2)
since the Laplacian and the Coulomb potential are invariant under all rotations of the space IR
3
.
Dierentiating (8.3) in at = 0, we get
[H,
k
] = 0, k = 1, 2, 3, (8.4)
where
k
:=
d
d
=0
R
k
().
Then H also commutes with H
k
:= i
k
and the angular momentum operators

L
k
since

L
k
= h
H
k
by (6.19):
[H, H
k
] = 0, [H,

L
k
] = 0, k = 1, 2, 3, . (8.5)
Hence the angular momenta L
k
(t) := (t),

L
k
(t)) are conserved.
Remark 8.2 Let us note that the angular momentum conservation played a crucial role in the de-
termination of the hydrogen spectrum in the Bohr-Sommerfeld old quantum theory (see Exercise
5).
62
Spherical harmonics
Now we can explain our general strategy in proving Theorem 8.1. Namely, the commutation (8.4)
provides a solution of the spectral problem (8.2) by a separation of variables. More precisely, the
strategy relies on the following three general arguments:
I. The commutation (8.5) obviously implies that the operator H
2
:= H
2
1
+ H
2
2
+ H
2
3
commutes with
H by (8.5):
[H, H
2
] = 0. (8.6)
Hence, each eigenspace of the Schrodinger operator H is invariant with respect to each operators H
k
and H
2
. Moreover, the operator H
2
commutes also with each operator H
k
, for example,
[H
k
, H
2
] = 0, k = 1, 2, 3. (8.7)
Exercise 8.3 Check (8.7). Hint: First, prove the commutation relations [H
k
, H
j
] = ih
kjl
H
l
where
kjl
is a totally antisymmetric tensor.
Since the operators H
k
, k = 1, 2, 3, do not commute, they cannot be diagonalized simultaneously. On
the other hand, the operators H, H
2
and, for example, H
3
, commute with each other. Hence, we
could expect that there is a basis of common eigenfunctions for them.
II. First, let us diagonalize simultaneously H
3
and H
2
. We consider the eigenvalue problem (8.2)
in the Hilbert space c := L
2
(IR
3
). On the other hand, both operators, H
3
and H
2
, act only on
the angular variables in spherical coordinates. Hence, the operators act also in the Hilbert space
c
1
:= L
2
(S, dS), where S stands for the two-dimensional sphere [x[ = 1.
Theorem 8.4 i) In c
1
there exist an orthonormal basis of Spherical Harmonics Y
m
l
(, ) which are
common eigenfunctions of H
3
and H
2
:
H
3
Y
m
l
= mY
m
l
, H
2
Y
m
l
= l(l + 1)Y
m
l
, m = l, l + 1, ..., l, (8.8)
where l = 0, 1, 2, ....
ii) Y
m
l
(, ) = F
m
l
()e
im
, where F
m
l
() are real functions.
We will prove this theorem in Section 10. It is important that each space of the common eigen-
functions is one-dimensional since the eigenvalues depend on l and m. This suggests that we could
construct the eigenfunctions of the Schrodinger operator H, by separation of variables, in the form
(x) = R(r)Y
m
l
(, ). (8.9)
The following theorem justify this choice of a particular form for the eigenfunctions.
Lemma 8.5 (On Separation of Variables) Each solution to the spectral problem (8.2) is a sum (or a
series) of the solutions of the particular form (8.9).
Formal Proof Let
m
l
denote the orthogonal projection in c
1
onto the linear span c
m
l
of Y
m
l
. Let us
dene its action also in c, by the formula
(
m
l
)(r, , ) :=
m
l
[(r, , )], r > 0, (8.10)
in the spherical coordinates r, , . Then
m
l
commutes with the Schrodinger operator H since the
latter commutes with H
3
and H
2
:
[H,
m
l
] = 0. (8.11)
Hence, applying
m
l
to (8.2), we get formally
E
m
l

= H
m
l

. (8.12)
It remains to note that
i) The function
m
l

has the form (8.9) since

m
l
[
(r, , )] c
m
l
, and the space c
m
l
is one-
dimensional;
ii)
l,m

m
l

.
Remarks 8.6 i) A complete solution to the spectral problem (8.8) relies on an investigation of all
commutation relations of the operators H
k
, k = 1, 2, 3, i.e. the Lie algebra generated by them.
ii) It remains still to determine the radial function in (8.9). We will substitute (8.9) into the equation
(8.8). This gives a radial eigenvalue problem which will be solved explicitly.
8.2 Spherical Laplacian
To determine the radial functions in (8.9), let us express the Laplacian operator in spherical coor-
dinates r, , : by denition,
x
3
= r cos , x
1
= r sin cos , x
2
= r sin sin . (8.13)
The operator is symmetric in the real Hilbert space L
2
(IR
3
). Hence, it is dened uniquely by the
quadratic form (, ), where D := L
2
(IR
3
) :
()
L
2
(IR
3
) C(IR
3
), [[ 2. In
spherical coordinates
(, ) = (, ) =
_

0
dr
_

0
d
_
2
0
d[(r, , )[
2
r
2
sin . (8.14)
Geometrically it is evident that
(r, , ) = e
r
r
+e
r
+e
r sin
, (8.15)
where e
r
, e
, e
are the orthogonal unit vectors proportional to

r
,
, respectively. Therefore,
(8.14) becomes
(, ) =
_

0
dr
_

0
d
_
2
0
d
_
2
+
2
+
r sin
2
_
r
2
sin . (8.16)
Integrating by parts, we get
(, ) =
_

0
dr
_

0
d
_
2
0
d(r
2
r
r
2
r
+

sin
r
2
sin
+
r
2
sin
2
)r
2
sin
= (r
2
r
r
2
r
+

sin
r
2
sin
+
r
2
sin
2
, ). (8.17)
Therefore, we get the Laplacian operator in spherical coordinates,
= r
2
r
r
2
r
+

sin
r
2
sin
+
r
2
sin
2
= r
2
r
r
2
r
+r
2
, (8.18)
where is the dierential operator on the sphere S with coordinates , :
=

sin
sin
+
sin
2
, (8.19)
64
Exercise 8.7 Check the integration by parts (8.17).
Denition 8.8 The operator is called Spherical Laplacian Operator.
Exercise 8.9 Check the identity
= H
2
. (8.20)
Hint: Both operators are second order spherically symmetric elliptic operators.
8.3 Radial Equation
Here we deduce Theorem 8.1 from Theorem 8.4. Let us consider a nonzero nite energy solution
(x) c := L
2
(IR
3
) to the problem (8.2) of the form (8.9). Substituting into (8.8), we get by (8.18)
and (8.20) that
2E
2
R(r) = r
2
r
r
2
r
R(r)
l(l + 1)
r
2
R(r) +
2e
2
h
2
r
R(r), r > 0. (8.21)
For r , the equation becomes
2E
2
R(r) R
l
(r). (8.22)
This suggests that E
< 0 and the asymptotics R(r) e

r
, r holds, where
=
_
2E/h
> 0. (8.23)
Next, let us write R(r) = e
r
F(r). Substituting into (8.21), we get
F
+
_
2
r
2
_
F
+
_
d
r

l(l + 1)
r
2
_
F = 0, r > 0, (8.24)
where d = b 2 with b = 2e
2
/h
2
. Finally, let us introduce the new variable = 2r, then (8.24)
becomes
f
+
_
2
1
_
f
+
_
1

l(l + 1)
2
_
f = 0, > 0, (8.25)
where f() = F(r) and = b/(2). Now let us seek for a solution f of the form
f() =
s
(a
0
+a
1
+a
2
2
+...) L()
s
(8.26)
with a
0
,= 0. We will nd two linearly independent solutions: one with s 0 and another with s 1.
Only the solution with s 0 is appropriate. For s 1 the corresponding eigenfunction
(x) is not
a function of nite energy since
(x) , L
2
(IR
3
) and
_
(x)[(x)[
2
dx = . Substituting (8.26)
into (8.25), we get
2
L
+
_
2s +
_
2
1
_
2
_
L
+
_
s(s 1) +
_
2
1
_
s +
_
1
r

l(l + 1)
2
_
2
_
L = 0 (8.27)
for > 0. After some evaluation, we have
2
L
+[2(s + 1) ]L
+ [( 1 s) +s(s + 1) l(l + 1)]L = 0, > 0. (8.28)

Setting = 0, we get formally that s(s + 1) l(l + 1) = 0. Hence, s = l since l 1 1. With
s = l, this equation becomes
L
+ [2(l + 1) ]L
+ [ 1 l]L = 0, > 0. (8.29)

Let us substitute here L() = a
0
+a
1
+a
2
2
+... and equate the coecients with identical powers of
:
_
0
: 2(l + 1) + ( 1 l)a
0
= 0,
1
: 2a
2
+ 2(l + 1)2a
2
a
1
+ ( 1 l)a
1
= 0,
2
: 3 2a
3
+ 2(l + 1)3a
3
2a
2
+ ( 1 l)a
2
= 0,
...
k
: (k + 1)ka
k+1
+ 2(l + 1)(k + 1)a
k+1
ka
k
+ ( 1 l)a
k
= 0,
...
(8.30)
Therefore, we get the recursive relation
a
k+1
=
k ( 1 l)
(k + 1)(k + 2l + 2)
a
k
. (8.31)
This implies a
k+1
/a
k
1/k if all a
k
,= 0. Then we have [L()[ Ce
= Ce
2r
with a C > 0. Hence,
R(r) = F(r)e
r
as r , which contradicts the denition of a quantum stationary state.
Therefore, a
k+1
= 0 and a
k
,= 0 for some k = 0, 1, 2, .... Then k ( 1 l) = 0, so
=
b
2
= k +l + 1 = n = 1, 2, ... (8.32)
Substituting here b = 2e
2
/h
2
and =
2E/h
, we get nally,
E = E
n
:=
2h
R
n
2
, n = 1, 2, ..., (8.33)
where R := e
4
/(4h
3
).
Corollary 8.10 The eigenfunctions
for = 2R/n
2
= e
4
/(2h
3
n
2
) have the following form
in spherical coordinates:
=
lmn
:= Ce
r/a
n
P
ln
(r)F
m
l
()e
im
. (8.34)
Here
a
n
:= 1/ = h
/
_
2E
n
= h
2
n/(e
2
), (8.35)
and P
ln
(r) is a polynomial function of degree n 1 l, F
m
l
()e
im
D(l) and m = l, ..., l.
Exercise 8.11 Calculate the multiplicity of the eigenvalue E
n
.
Hint: It is equal to

0ln1
(2l + 1).
Remark 8.12 For the stationary state (8.34)
i) The function L() =
l
(a
0
+... +a
k
k
) is a polynomial of degree k +l = n 1. Hence, n 1 equals
the number of zeros of the function L(), 0, and n is called the principal quantum number of the
stationary state.
ii) The component L
3
of angular momentum is equal to mh
. This follows from (6.13), (6.17) and

(6.19). The number m is called the magnetic quantum number of the stationary state since it is
responsible for the change of the energy in the magnetic eld (see (16.10)).
iii) The (squared) absolute value of angular momentum L
2
:= L
2
1
+L
2
2
+L
2
3
is equal to l(l +1)h
2
. This
follows from (8.8) since L
2
= h
2
H
2
. The number l is called the azimuthal quantum number of the
stationary state.
66
9 Spherical Spectral Problem
We split the space L
2
(S
2
) in a sum of orthogonal eigenspaces of the spherical Laplacian. The proof
uses rotation invariance of the Schrodinger operator and classication of all irreducible representations
of commutation relations for a Lie algebra of angular momenta.
9.1 Hilbert-Schmidt Argument
We start the proof of Theorem 8.4 with a diagonalization of the operator H
2
. This lemma and the
elliptic theory [86] imply
Lemma 9.1 The operator H
2
:= H
2
1
+H
2
2
+H
2
3
in the space c
1
:= L
2
(S, dS) is selfadjoint and admits
the spectral resolution
c
1
=
n=0
L(n), (9.1)
where L(n) C
(S) are nite-dimensional orthogonal subspaces of c

1
, and H
2
[
L(n)
=
n
with
n
as n .
Proof Step i) Each H
k
is a symmetric operator in c
1
. Indeed, the rotations

R
k
() form a unitary
group in c
1
= L
2
(S, dS), hence its generators
k
are skew-adjoint. Therefore, the operator H
2
is a
symmetric nonnegative operator in c
1
.
Step ii) The operator H
2
is a nonnegative elliptic second order operator on the sphere S. This follows
from the formula (8.19) by the identity (8.20). Hence, the operator H
2
+ 1 : H
2
(S) H
0
(S) is
invertible, where H
k
(S) stand for the Sobolev spaces on the sphere S. Respectively, the operator
(H
2
+ 1)
1
: H
0
(S) H
2
(S) is selfadjoint and compact in H
0
(S) = L
2
(S, dS) by the Sobolev em-
bedding theorem. Hence, the resolution (9.1) exists by the Hilbert-Schmidt theorem for the operator
(H
2
+ 1)
1
.
We will prove that we can choose the resolution (9.1) with dimL(n) = 2l + 1 and
n
= l(l + 1),
where l = l(n) = 0, 1, 2, ....
Lemma 9.2 All spaces L(n) are invariant with respect to rotations of the sphere.
Proof This follows from the rotation invariance of H
2
.
Corollary 9.3 All spaces L(n) are invariant with respect to the operators H
k
, k = 1, 2, 3.
9.2 Lie Algebra of Angular Momenta
The linear span (H
1
, H
2
, H
3
) is a Lie algebra since the following commutation relations hold:
[H
1
, H
2
] = iH
3
, [H
2
, H
3
] = iH
1
, [H
3
, H
1
] = iH
2
. (9.2)
Exercise 9.4 Check the commutation relations. Hint: H
k
= i(x
x
)
k
.
Lemma 9.5 For each k = 1, 2, 3,
i) All spaces L(n) are invariant with respect to H
k
.
ii) [H
k
, H
2
] = 0.
Proof ad i) The invariance holds by Corollary 9.3.
ad ii) The commutation holds by (8.7).
9. SPHERICAL SPECTRAL PROBLEM 67
Denition 9.6 H
= H
1
iH
2
.
Lemma 9.7 All spaces L(n) are invariant with respect to H
, and
[H
3
, H
] = H
, (9.3)
H
2
= H
+
H
+H
3
(H
3
1) = H
H
+
+H
3
(H
3
+ 1), (9.4)
H
2
=
1
2
(H
+
H
+H
H
+
) +H
2
3
. (9.5)
Proof The invariance follows from Lemma 9.5 i), (9.3) and (9.4) follow from (9.2), (9.5) follows from
(9.4).
9.3 Irreducible Representations of Commutation Relations
Here we give a complete classication of all possible triples of Hermitian operators satisfying the
commutation relations (9.2). It implies a part of Theorem 8.4 for integer or half odd-integer values of
l.
Proposition 9.8 [75] Let E be a nonzero nite-dimensional complex linear space with a Hermitian
scalar product, and
i) linear Hermitian operators H
k
, k = 1, 2, 3, in E, satisfying the commutation relations (9.2).
ii) H
2
is a scalar: H
2
= C,
iii) The space E is irreducible, i.e. it does not contain any nontrivial subspace which is invariant
under all H
k
.
Then there exists a spin number J = 0,
1
2
, 1,
3
2
, 2, ... and an orthonormal basis in E, e
m
: m =
J, J + 1, ..., J 1, J, such that = J(J + 1), and the action of the operators is given by
H
1
e
m
=
s
+
Jm
2
e
m+1
+
s
Jm
2
e
m1
, H
2
e
m
=
s
+
Jm
2i
e
m+1
Jm
2i
e
m1
, H
3
e
m
= me
m
, (9.6)
where
s
Jm
=
_
(J m)(J m+ 1). (9.7)
Proof Set H
:= H
1
iH
2
, then all relations (9.3)-(9.5) hold by (9.2), as in Lemma 9.5. Since H
3
is a Hermitian operator, there exists a basis in E, provided by its eigenvectors: for each vector e
m
of
this basis we have H
3
e
m
= me
m
with a real eigenvalue m IR.
Lemma 9.9 Either e
:= H
e
m
= 0, or e
is an eigenvector of H
3
with the eigenvalue m1.
Proof (9.3) implies that H
3
e
= H
3
H
e
m
= H
H
3
e
m
H
e
m
= me
.
Further, (9.5) implies that is real and m
2
since H
= H
. Therefore, = J(J +1) for a unique

J 0. Finally, the identities (9.4) imply that H
= H
3
(H
3
1) and H
+
H
+
= H
3
(H
3
+1).
Hence,
0 |H
e
m
|
2
= e
m
, H
e
m
) = e
m
, ( H
3
(H
3
1))e
m
)
= [J(J + 1) m(m1)]|e
m
|
2
= [(J +m)(J m+ 1)]|e
m
|
2
, (9.8)
0 |H
+
e
m
|
2
= e
m
, H
+
H
+
e
m
) = e
m
, ( H
3
(H
3
1))e
m
)
= [J(J + 1) m(m+ 1)]|e
m
|
2
= [(J m)(J +m+ 1)]|e
m
|
2
. (9.9)
68
Therefore, H
e
m
= 0 implies that either m = J or m = J + 1. But the last value is impossible
since J 0 and J(J + 1) = m
2
. Consequently, m = J 0. Similarly, H
+
e
m
= 0 implies that
m = J 0. Therefore, m has to run in integral steps from J to J. Hence, 2J +1 = 1, 2, .... Further,
e
m
: m = J, J + 1, ..., J 1, J is a basis since E is irreducible.
Finally, for the normalized vectors e
m
, (9.8) and (9.9) imply that
H
+
e
m
= s
+
Jm
e
m+1
, H
e
m
= s
Jm
e
m1
. (9.10)
Hence, the formulas (9.6) follow.
Corollary 9.10 Let us set Z
+
Jm
= s
+
J,J
...s
+
J,m1
. Then the vectors Y
m
J
= H
J+m
+
e
J
/Z
+
Jm
, m =
J, ...J, constitute an orthonormal basis in the space E.
Denition 9.11 For J = 0,
1
2
, 1,
3
2
, 2, ..., denote by D(J) the space E endowed with the operators H
k
,
k = 1, 2, 3, described by Proposition 9.8.
Example 9.12 For J =
1
2
, the operators H
k
in the orthonormal basis (Y
1/2
1/2
, Y
1/2
1/2
) are represented
by the matrices s
k
:=
1
2
k
, where
k
are the Pauli matrices:
1
=
_
0 1
1 0
_
,
2
=
_
0 i
i 0
_
,
3
=
_
1 0
0 1
_
. (9.11)
Exercise 9.13 Deduce (9.11) from (9.6).
9.4 Spherical Harmonics
Proposition 9.8 implies that each eigenspace L(n) is isomorphic to a direct sum of a number M(n, J)
of spaces D(J) with a unique value of J = J(n) = 0,
1
2
, 1,
3
2
, 2, .... We can assume that
n
are distinct
for distinct n, then distinct values of J for a xed n are impossible since the eigenvalue of H
2
in D(J),
equal to J(J + 1), is a strictly increasing function of J 0. Hence, Theorem 8.4 i) will follow from
the next lemma.
Lemma 9.14 i) For every half-integer J we have M(n, J) = 0 for each n.
ii) For every integer l = 1, 2, 3, ..., there exists a unique n = n(l) such that M(n, l) = 1.
Proof i) Each eigenspace L(n) is a direct sum of the irreducible subspaces D(J). Let us consider
one of the subspaces. For each eigenvector e
m
D(J), of the operator H
3
, we have the following
dierential equations,
H
3
e
m
(, ) = i
e
m
(, ) = me
m
(, ). (9.12)
Therefore, e
m
(, ) = F
m
l
()e
im
, so m is an integer number since the function e
m
(, ) has to be
single-valued. Hence all possible J also are only integer: J = l.
ii) Let us consider the lowest eigenvalue m = l in the space D(l). Then H
e
l
(, ) = 0 by Lemma
9.9. We will prove below that in spherical coordinates,
H
= e
i
[
i cot
], H
+
= e
i
[
+i cot
]. (9.13)
Taking into account that e
l
(, ) = F
l
l
()e
il
, we get from H
e
l
(, ) = 0 the dierential equa-
tion (
l cot )F
l
l
= 0. Hence, F
l
l
= C sin
l
which means that l is a simple eigenvalue. Hence,
M(n, l) 1 for all n.
9. SPHERICAL SPECTRAL PROBLEM 69
It remains to check that M(n, l) ,= 0 for some n. This is equivalent to the existence of an eigenvector
of the operator H
2
with eigenvalue l(l +1). However, this is obvious for the function e
l
(, ). Namely,
Eqn. (9.4) implies that
H
2
e
l
= H
+
H
e
l
+H
3
(H
3
1)e
l
= l(l + 1)e
l
since H
e
l
= 0 and H
3
e
l
= le
l
.
Now Theorem 8.4 i) is proved. Further, Corollary 9.10 implies that the functions H
l+m
+
e
l
/Z
+
lm
,
m = l, ..., l, provide an orthonormal basis in the space D(l). This implies Theorem 8.4 ii) since
H
l+m
+
e
l
=
_
e
i
[
+i cot
]
_
l+m
(sin
l
e
il
) = F
m
l
()e
im
, (9.14)
where F
m
l
() is a real function.
Denition 9.15 The normalized functions Y
m
l
(, ) = F
m
l
()e
im
/Z
+
lm
are called spherical har-
monics.
9.5 Angular Momenta in Spherical Coordinates
We prove (9.13). First, let us rewrite (8.15) as
(r, , ) = e
r
r
+e
r
+e
r sin
= e
1
1
+e
3
3
+e
3
3
, (9.15)
where e
1
:= (1, 0, 0), etc. Then it is evident geometrically that
_
_
e
r
= (e
1
cos +e
2
sin ) sin +e
3
cos ,
e
= (e
1
cos +e
2
sin ) cos e
3
sin ,
e
= e
2
cos e
1
sin .
(9.16)
Substituting this into (9.15), we get
_
1
= sin cos
r
+ cos cos
r
sin

r sin
,
2
= sin sin
r
+ cos sin
r
+ cos

r sin
,
3
= cos
r
sin
r
.
(9.17)
Substituting (9.17) and (8.13) into H
k
= i(x
x
)
k
, we get
_
_
H
1
= i(sin
+ cot cos
)
H
2
= i(cos
+ cot sin
)
H
3
= i
.
(9.18)
At last, the rst two formulas imply (9.13).
70
10 Atom Dipole Radiation and Selection Rules
We evaluate a radiation of an atom with an electrostatic potential
ext
(t, x) =
ext
(x) in a static
external magnetic eld with vector potential A
ext
(x).Then the Schrodinger equation (7.26) becomes
[ih
t
e
ext
(x)](t, x) =
1
2
[ih
e
c
A
ext
(x)]
2
(t, x). (10.1)
We apply the Bohr approximation and derive the Rydberg-Ritz combination principle and selection
rules in the case of cylindric symmetry.
10.1 Rydberg-Ritz Combination Principle
Consider the atom in a statistical equilibrium with an appropriate environment. Then the wave
function (t, x) admits an eigenfunction expansion
(t, x) =
k
c
k
k
(x)e
i
k
t
+
_

0
c
(x)e
it
d , (10.2)
where
k
and
are the eigenmodes of the time-independent Schrodinger equation belonging to the

discrete and continuous spectrum, respectively. Further, the c
k
and c
describe the degree of excitation

of the individual eigenmodes.
Comment 10.1 The amplitudes c
k
and c
are random variables corresponding to an equilibrium

distribution. The equilibrium is maintained by an external source, like the electric discharge or heat
bath, X-rays, etc. The equilibrium distribution depends on the accelerating voltage (or the heat bath
temperature, X-rays wave-lenght, etc): the number of the excited modes generally increases with the
voltage (or with the heat bath temperature, etc). A mathematical justication of the equilibrium
distribution for the Maxwell-Schrodinger equations is an open problem. The justication of the con-
vergence to equilibrium distribution in a thermal bath is done in [22]-[27] for a list of linear hyperbolic
equations, and in [54] for a nonlinear one. The role of the thermal bath plays the initial wave eld.
The convergence to the equilibrium distribution is caused by the uctuations of the initial eld. The
uctuations are provided by the mixing condition of Rosenblatt or Ibragimov-Linnik type.
Let us analyze the Maxwell eld produced by the corresponding charge and current densities from
(7.27). The potentials of the eld satisfy the Maxwell equations (4.6), (4.9) and their large time
asymptotics for bounded [x[ is given by the retarded potentials (43.4):
_
_
(t, x)
ret
(t, x) :=
_
(t [x y[/c, y)
[x y[
dy
A(t, x) A
ret
(t, x) :=
1
c
_
j(t [x y[/c, y)
[x y[
dy
t , [x[ R. (10.3)
Let us assume that the atom is located at the point y = 0. Then the densities (t, y), j(t, y) are
localized in a ball [y[ a 1, where a > 0 is an atom radius, a 10
8
cm:
(t, y) = 0, j(t, y) = 0, [y[ > a. (10.4)
10. ATOM DIPOLE RADIATION AND SELECTION RULES 71
Therefore, any macroscopic observation at a distance [x[ 1 coincides with high precision with the
dipole approximation (cf. (42.5))
_
_
(t, x)
1
[x[
_
|y|a
(t [x y[/c, y)dy
A(t, x)
1
c[x[
_
|y|a
j(t [x y[/c, y)dy
t , [x[ R. (10.5)
Remark 10.2 The approximation [x[
1
corresponds to the Hertzian dipole radiation (45.67).
Next, we calculate the charge-current densities corresponding to the wave function (10.2) in the Born
approximation (7.27). We can write it using (2.3), in the form
_
_
(t, y) = e(t, y)(t, y)
j(t, y) = Re
_
_
e
[ih
e
c
A
ext
(y)](t, y)
_
(t, y)
_
(10.6)
The integral over the continuous spectrum in the RHS of (10.2), is responcible only for the continuous
spectrum of the atom radiation. Hence, the discrete spectrum is completely determined by the rst
sum. Substituting the sum into (10.6), we get the discrete components of the charge and current
densities in the form
_
d
(t [x y[/c, y) eRe

kk
c
k
c
k
e
i(
k
k
)(t|xy|/c)
k
(y)
k
(y)
j
d
(t [x y[/c, y)
e
Re
kk
c
k
c
k
e
i(
k
k
)t
_
[ih
e
c
A
ext
(y)]
k
(y)e
i
k
|xy|/c
_
k
(y)e
i
k
|xy|/c
.
(10.7)
If we substitute this into (10.5), we get the following approximation for the discrete component of the
radiation:
_
_
(t, x)
1
[x[
Re
kk
kk
(x)e
i(
k
k
)t
A(t, x)
1
c[x[
Re
kk
A
kk
(x)e
i(
k
k
)t
t , [x[ R. (10.8)
Corollary 10.3 The Rydberg-Ritz combination principle holds: the discrete spectrum of the atom
radiation is contained in the set
kk
:=
k

k
.
Further, applying the formula Re
kk
S
kk
=
1
2
kk
(S
kk
+S
k
k
), we get the following expressions for
the limiting amplitudes in (10.8):
_
kk
(x) c
k
c
k
e
_
|y|a
e
i
kk
|xy|/c
k
(y)
k
(y)dy
A
kk
(x) c
k
c
k
e
2
_
|y|a
e
i
kk
|xy|/c
__
[ih
e
c
A
ext
(y)]
k
(y)
_
k
(y)
+
_
[ih
e
c
A
ext
(y)]
k
(y)
_
k
(y)
_
dy
(10.9)
72
with an error of order h
k
/c in the last formula. Further let us assume that
[
kk
[a/c 2 (10.10)
which means that the wave length
kk
:= 2c/[
kk
[ a.
Under condition (10.10) the exponent in the integrands of (10.9) is close to a constant exp(i
kk
[x[/c)
since [y[ a. In addition, the ball of integration [y[ a can be substituted by the whole space since
the eigenfunctions
k
,
k
are well localized. Therefore, (10.9) becomes
_
kk
(x) ec
k
c
k
e
i
kk
|x|/c
_

k
(y)
k
(y)dy
A
kk
(x)
e
2
c
k
c
k
e
i
kk
|x|/c
_
__
[ih
e
c
A
ext
(y)]
k
(y)
_
k
(y)
+
_
[ih
e
c
A
ext
(y)]
k
(y)
_
k
(y)
_
dy
(10.11)
First formula of (10.11) implies that
kk
(x) 0 for
kk
,= 0 by the orthogonality of dierent
eigenfunctions. Hence, the radiation is represented by the magentic potential
A(t, x)
1
c[x[
e
2
Re
kk
c
k
c
k
e
i
kk
(t|x|/c)
J
kk
, t , (10.12)
where J
kk
stands for the last integral in (10.11).
Note that the radiation eld (10.12) is identical with the sum of the Hertzian dipole radiation elds
of type (45.67) corresponding to the dipole moments J
kk
.
Comment 10.4 The radiation eld (10.12) brings the energy to innity like the Hertzian dipole
radiation eld. Hence, it cannot be stationary without an external source of the energy (cf. Comment
10.1).
The formulas (10.12) dene the intensity of the spectral line
kk
corresponding to simple
eigenvalues
k
and
k
:
A
kk
) c
k
c
k
)J
kk
, (10.13)
where ) stands for the mathematical expectation. For the multiple eigenvalues the intensity is also
proportional to the multiplicity of the spectral line which is dened by its splitting in a weak
magnetic eld (see [89]).
Traditional identication of the relative intensities (see [89]) means that
A
kk
)
A
nn
)
=
c
k
c
k
)J
kk
c
n
c
n
)J
nn
J
kk
J
nn
. (10.14)
It holds if c
k
c
k
) varies slowly in k, k
(N
1
, N
2
) with N
2
N
1
suciently small.
10.2 Selection Rules for Cylindrical Symmetry
We consider the case of a radial electrostatic potential
ext
(x) =
ext
([x[) and a static uniform
magnetic eld B. Then the vector potential A
ext
(x) = B x/2. Let us consider a xed spectral line
kk
corresponding to the frequencies
k
and
k
and eigenfunctions (see (16.12))
= R
nl
(r)F
m
l
()e
im
,
= R
n
l
(r)F
l
m
()e
im
. (10.15)
The next theorem demonstrates that the spectrum of the atom radiation is a very small subset of the
set of all dierences
kk
:=
k

k
from the Rydberg-Ritz combination principle.
Theorem 10.5 Let the condition (10.10) hold,
ext
(x) =
ext
([x[) and A
ext
(x) = B x/2. Then
for the dipole approximations of the limiting amplitudes we have J
kk
= 0 if either l
,= l 1 or
m
,= m, m1.
Proof Let us use the identity (7.21) for the solutions
k
(x)e
i
k
t
,
k
(x)e
i
k
t
to Equation (7.26).
Eqs. (7.22) give the densities
_
kk
(t, x) := e
i
kk
t
e
k
(x)
k
(x),
j
kk
(t, x) := e
i
kk
t
__
[ih
e
c
A
ext
(x)]
k
(x)
_
k
(x)
+
_
[ih
e
c
A
ext
(x)]
k
(x)
_
k
(x)
_
(10.16)
Then (7.21) becomes
i
kk

kk
(t, x) + div j
kk
(t, x) = 0, (t, x) IR
4
. (10.17)
We see that the integral in the second formula of (10.11) is proportional to the integral of j
kk
(t, y):
J
kk
e
i
kk
|x|/c
e
i
kk
t
_
j
kk
(t, y)dy. (10.18)
It remains to prove
Lemma 10.6
_
j
kk
(t, x)dx = 0, t IR, (10.19)
if either l
,= l 1 or m
,= m, m1.
Proof Let us check the case m
,= m, m1. The case l
,= l 1 will be considered in next section. Let

us multiply (10.17) by the coordinate x
p
, p = 1, 2, 3, and integrate over IR
3
. Then we get by partial
integration,
i
kk
_
x
p
kk
(t, x)dx
_
j
p
kk
(t, x)dx = 0, (10.20)

where j
p
kk
is the p-th component of j

kk
. Hence,
_
j
p
kk
(t, x)dx c(t)

_
x
p
k
(x)
k
(x)dx. (10.21)
Let us rewrite the last integral in spherical coordinates. Then we get by (10.15),
_
x
p
k
(x)
k
(x)dx
=
_

0
R
nl
(r)R
n
l
(r)r
3
dr
_
S
x
p
r
F
m
l
()e
im
F
l
m
()e
im
dS. (10.22)
Obviously, the last integral is equal to zero if m
,= m, m1.
Remark 10.7 The RHS of (10.21) equals to the matrix element of the dipole moment operator
x
p
. Hence, the intensities J
kk
are proportional to the dipole moment like classical Hertzian dipole
radiation (45.67). This is an example of the Bohr correspondence principle.
Comment 10.8 The second Bohr postulate states that the radiation of the spectral line
kk
is pro-
vided by the transition
k

k
between the stationary states. With this identication, the formulas
(10.18) and (10.21) mean that the RHS of (10.21) is proportional to the probability of the transition.
74
10.3 Selection Rules for Orbital Momentum
The proof of the selection rule l
= l 1 is more involved. We follow the method of [12, Appendix XXI].

First, let us re-express the 2l + 1 eigenvectors F
m
l
()e
im
of the spherical Laplacian with eigenvalue
l(l + 1) in terms of real eigenvectors, e.g., like
Z
l,m
= F
m
l
() cos m , m = 0, . . . , l
Z
l,m
= F
m
l
() sin(ml) , m = l + 1, . . . , 2l, (10.23)
then let us prove the following
Lemma 10.9 The Z
l,m
may be expressed as r
l
U
l,m
(x), where each U
l,m
is a homogeneous polynomial
of degree l in the variables x, y, z such that
U
l,m
= 0 (10.24)
holds.
Proof of Lemma 10.9
1) r
l
U
l,m
(x) only depends on the angular coordinates , .
2) From
=
1
r
2
2
r
+
2
r
r
+
1
r
2
it follows that
0 = U
l,m
= r
l2
(l(l + 1)Z
l,m
+ Z
l,m
) (10.25)
therefore Z
l,m
r
l
U
l,m
are eigenvectors of with eigenvalues l(l + 1).
3) There are exactly 2l + 1 linearly independent U
l,m
with the property U
l,m
= 0. This may be
shown as follows: There are
1
2
(l + 2)(l + 1) linearly independent homogeneous polynomials of
degree l, e.g., the monomials
x
l
, (x
l1
y, x
l1
z), (x
l2
y
2
, x
l2
yz, x
l2
z
2
), . . . , (xy
l1
, . . . , xz
l1
), (y
l
, . . . , z
l
)
(here, all monomials with the same power of x are grouped together). Further, the Laplacian
acting on a general homogeneous polynomial V
l
of degree l produces a general homogeneous
polynomial

V
l2
of degree l 2, V
l
=

V
l2
. The condition that

V
l2
vanishes introduces
therefore
1
2
l(l 1) conditions on the
1
2
(l + 2)(l + 1) coecients of V
l
, therefore there exist
1
2
(l + 2)(l + 1)
1
2
l(l 1) = 2l + 1 (10.26)
linearly independent V
l
such that V
l
= 0. Together with Theorem 8.4 this proves the above
lemma.
From Theorem 8.4 (orthogonality of the D(l) subspaces) it follows that
_
sin ddZ
l,m
Z
l
,m
= 0
unless l = l
. Therefore, all that remains to prove is that

x
p
r
Z
l,m
= Y
l+1
+Y
l1
(10.27)
where Y
l+1
and Y
l1
are some general eigenvectors of D(l +1) and D(l 1), respectively, and x
p
is one
component of the coordinate vector x = (x, y, z). Eq. (10.27) may be re-expressed, e.g., for x
p
= x,
like
xU
l
= U
l+1
+r
2
U
l1
(10.28)
where the U
k
are homogeneous polynomials of degree k which obey U
k
= 0. Eq. (10.28) may be
proved with the help of the following
Lemma 10.10 Every homogeneous polynomial F
n
of degree n may be uniquely expressed like
F
n
= U
n
+r
2
U
n2
+r
4
U
n4
+. . . (10.29)
where the U
k
are homogeneous polynomials of degree k which obey U
k
= 0.
The proof of this lemma will be given below. Certainly, xU
l
of Eq. (10.28) is a homogeneous polynomial
of degree l + 1, therefore it may be re-expressed like
xU
l
= U
l+1
+r
2
U
l1
+r
4
U
l3
+. . . (10.30)
Next, we act with the Laplace operator on both sides of Eq. (10.30). For the l.h.s. we get
xU
l
= 2
x
U
l
(10.31)
and for the r.h.s. we use
_
r
2h
U
l+12h
_
= r
2h
U
l+12h
+ 2
_
r
2h
_
(U
l+12h
) +U
l+12h
r
2h
= 0 + 4hr
2h2
x U
l+12h
+ 2h(2h + 1)r
2h2
U
l+12h
= 2h(2l + 3 2h)r
2h2
U
l+12h
(10.32)
where we used Eulers theorem for homogeneous polynomials,
x U
n
= nU
n
.
Therefore, we get altogether
2
x
U
l
= 2(2l + 1)U
l1
+ 4(2l 1)r
2
U
l3
+. . . . (10.33)
Now the important observation is that the l.h.s. vanishes upon a further application of the Laplace
operator,
x
U
l
=
x
U
l
= 0, therefore the r.h.s. must vanish, as well, when the Laplace operator
is applied. But, according to (10.32), this is possible only if all the U
l+12h
except for U
l1
vanish
(i.e., U
l3
= U
l5
= . . . = 0). This proves Eq. (10.28) and, therefore, the original statement that the
integral (10.22) vanishes unless l
= l 1.
We still have to prove Lemma (10.10).
Proof of Lemma 10.10 We prove the lemma by induction. It is obviously true for n = 0, 1. We
assume that it is true for all k < n. Then it is also true for F
n
, which is a homogeneous polynomial
of degree n 2, i.e.,
F
n
=

U
n2
+r
2

U
n4
+. . . +r
2h

U
n2h
+. . . (10.34)
The general solution to this equation, which is subject to the additional condition that it is a homo-
geneous polynomial of degree n, is
F
n
= G
n
+
1
2(2n 1)
r
2

U
n2
+. . . +
1
2h(2n 2h + 1
r
2h

U
n2h
(10.35)
76
as may be shown easily with the help of Eq. (10.32). Here G
n
is an arbitrary homogeneous polynomial
of degree n subject to the condition G
n
= 0. By choosing
U
n
= G
n
, U
n2h
= 2h(2n 2h + 1)
U
n2h
this can be brought into the form stated in Lemma (10.10).
11. CLASSICAL SCATTERING OF LIGHT: THOMSON FORMULA 77
11 Classical Scattering of Light: Thomson formula
Electromagnetic waves may be scattered by a charged matter. Here, we describe a light as a plane wave
satisfying the free Maxwell equation and a matter as composed of classical, point-like particles obeying
the Lorentz equation. The inconsistencies inherent to the concept of point-like charged particles
are circumvented by applying a certain approximation (neglecting self-interactions). We apply the
Hertzian dipole radiation formula and derive the Thomson dierential cross-section.
11.1 Incident Plane Wave
J.C.Maxwell identied light with electromagnetic waves which are solutions to the Maxwell equations
in free space with (t, x) = 0 and j(t, x) = 0. A general solution to the Maxwell equations is expressed
through the potentials by (4.3):
E(t, x) =
x
(t, x)
1
c
A(t, x), B(t, x) = rot A(t, x), (t, x) IR

4
. (11.1)
For the case (t, x) = 0 and j(t, x) = 0 the potentials A
0
(t, x) and
0
(t, x) are solutions to the
homogeneous wave equations (4.9) and (4.6):
2A
0
(t, x) = 0, (t, x) IR
4
. (11.2)
2
0
(t, x) = 0, (t, x) IR
4
. (11.3)
Let us choose (t, x) = 0 for concreteness, and
A
0
(t, x) = Asin k(x
3
ct)(1, 0, 0) (11.4)
with a wave number k > 0 and the corresponding frequency = kc > 0. Then (11.1) gives
E
0
(t, x) = kAcos k(x
3
ct)(1, 0, 0), B
0
(t, x) = kAcos k(x
3
ct)(0, 1, 0). (11.5)
The corresponding energy ux (i.e., Pointing vector) is
S
0
(t, x) =
c
4
E
0
(t, x) B
0
(t, x) =
cE
2
0
4
cos
2
k(x
3
ct)(0, 0, 1), (t, x) IR
4
, (11.6)
where E
0
:= kA. Let us note that the energy ux is directed along e
3
= (0, 0, 1) and its intensity is
I
0
:= lim
T
1
T
[
_
T
0
S
0
(t, x)dt[ =
cE
2
0
8
. (11.7)
11.2 Scattering Problem
We consider the scattering of the plane wave (11.5) by a classical electron. The scattering is described
by the coupled Maxwell-Lorentz equations
_
_
div E(t, x) = 4e(x x(t)), rot E(t, x) =
1
c
B(t, x),
1
c
E(t, x) +
4
c
e x(x x(t)),
(t, x) IR
4
, (11.8)
x(t) = e[E(t, x(t)) +
1
c
x(t) B(t, x(t))], t IR. (11.9)
78
The free electron is governed by the Lorentz equation (11.9) with the initial conditions
x(t) = 0, x(t) = 0, t < 0. (11.10)
The incident plane wave appears in the initial conditions for the elds,
E(t, x) = (x
3
+ct)E
0
(t, x) e
x
[x[
2
, B(t, x) = (x
3
+ct)B
0
(t, x), t < 0, (11.11)
where is the Heaviside function and ex/[x[ is the static Coulomb eld generated by the electron
at the position (11.10). Let us note that the incident wave is a solution to the homogeneous Maxwell
equations.
11.3 Neglecting the Self-Interaction
Let us split the solution to (11.8) like
_
E(t, x) = (x
3
+ct)E
0
(t, x) +E
r
(t, x),
B(t, x) = (x
3
+ct)B
0
(t, x) +B
r
(t, x),
t IR, (11.12)
where E
r
(t, x), B
r
(t, x) stand for the radiated elds. The elds are dened by the splitting. Then the
Maxwell equations (11.8) read
_
_
div E
r
(t, x) = 4e(x x(t)), rot E
r
(t, x) =
1
c
B
r
(t, x),
div B
r
(t, x) = 0, rot B
r
(t, x) =
1
c
E
r
(t, x) +
4
c
e x(x x(t)),
(t, x) IR
4
, (11.13)
since the incident wave in (11.11) is a solution to the homogeneous Maxwell equations. The initial
conditions (11.11) become
E
r
(t, x) = e
x
[x[
2
, B
r
(t, x) = 0, t < 0. (11.14)
The Lorentz equation (11.9) now reads
x(t) = e[(x
3
(t) +ct)E
0
(t, x(t)) +E
r
(t, x(t))
+
1
c
x(t) ((x
3
(t) +ct)B
0
(t, x(t)) +B
r
(t, x(t)))], t IR. (11.15)
Unfortunately, the problem (11.13), (11.15) is not well posed. Namely, the solutions E
r
(t, x), B
r
(t, x)
to (11.13) are innite at the points (t, x(t)), which is obvious from (11.14). Therefore, the RHS of
equation (11.15) does not make sense for the elds E
r
(t, x), B
r
(t, x), which are solutions to (11.13).
Remark 11.1 To make the problem well posed it is necessary to replace the point-like electron by the
extended electron suggested by M.Abraham [1]. For this model, the well-posedness is proved in [67].
Here we use another traditional approach to make the problem well posed, which is similar to the
Born approximation. Concretely, we omit the radiation elds at the RHS of (11.15):
x(t) = e[(x
3
+ct)E
0
(t, x) +
1
c
x(t) (x
3
+ct)B
0
(t, x)], t IR. (11.16)
Then we substitute the solution x(t) into the RHS of the Maxwell equations (11.13) to calculate the
radiated elds E
r
(t, x), B
r
(t, x).
11. CLASSICAL SCATTERING OF LIGHT: THOMSON FORMULA 79
The approximation (11.16) means that we neglect the self-interaction of the electron, which can be
justied for the non-relativistic electron. This means that the electron velocities are small compared
to the speed of light:
:= max
tIR
[ x(t)[/c 1. (11.17)
Then we can neglect the contribution of the magnetic eld to the RHS of (11.16). Finally, we consider
the equation
x(t) = eE
0
(t, x), t > 0. (11.18)
This equation and the initial conditions (11.10) dene the trajectory x(t) uniquely:
x(t) =
eA
kc
2
(1 cos kct)(1, 0, 0). (11.19)
The condition (11.17) for the solution is equivalent to
=
[e[A
c
2
1. (11.20)
Remark 11.2 This relation means that the amplitude of the oscillations
[e[A
kc
2
is small compared to
the wave length 2/k of the incident wave.
11.4 Scattering in Dipole Approximation
Now we have to solve the Maxwell equations (11.13) to dene the radiation elds E
r
(t, x), B
r
(t, x). Our
goal is an analysis of the energy ux at innity, i.e., of the Poynting vector S
r
(t, x) := (c/4)E
r
(t, x)
B
r
(t, x) as [x[ . We use the traditional dipole approximation which leads to the well-known
Thomson formula.
For this purpose, let us expand the charge density in the Maxwell equations (11.13) in a Taylor
series of the type (42.11):
e(x x(t)) = e(x) +ex(t) (x) +
1
2
e(x(t) )
2
(x) +..., t > 0. (11.21)
Here, the rst term is static, and the corresponding Maxwell eld is static with zero energy ux. The
second term corresponds to the Hertzian dipole with dipole moment p(t) := ex(t). The next terms of
(11.21) give small contributions to the total energy radiation to innity by (11.17).
Exercise 11.3 i) Prove the convergence of the Taylor series (11.21) in the sense of (42.14) with
H
a
(IR
3
) for a > max [x(t)[ = 2
[e[A
kc
2
.
ii) Prove that the energy ux S
r
(t, x) for large [x[ is determined by the second term of the expansion
(11.21) up to an error O()/[x[
2
. Hint: Use the methods of Exercise 45.6.
So we can use the Hertzian formula (45.72) for the dipole radiation (Section 45.6);
S
r
(t, x)
sin
2
4c
3
[x[
2
p
2
(t [x[/c)n, [x[ , (11.22)
where is the angle between p(t [x[/c) and n := x/[x[. By (11.18) and (11.5), we have
p(t) = e x(t) =
e
2
E
0
cos k(ct). (11.23)
80
Hence,
p(t)
2
=
_
e
2
_
2
E
2
0
cos
2
kct. (11.24)
Denition 11.4 Denote by the angle between n and e
3
, and by the (azimuthal) angle between e
1
and the plane (n, e
3
).
Exercise 11.5 Check that cos = cos sin . Solution: cos is the projection of n onto the vector
p(t [x[/c), which is parallel to e
1
. The projection of n onto the plane = (e
1
, e
2
) has length sin .
Finally, the angle between this projection and e
1
equals .
Corollary 11.6 sin
2
= 1 cos
2
sin
2
.
Therefore,
S
r
(t, x)
1 cos
2
sin
2
4c
3
[x[
2
_
e
2
_
2
E
2
0
cos
2
kc(t [x[/c)n, [x[ . (11.25)
Hence, for large [x[, the energy ux is directed along n. The corresponding intensity is obtained by
the standard replacement of cos
2
kc(t [x[/c) by 1/2:
I
r
(x) := lim
T
1
T
[
_
T
0
S
r
(t, x)dt[
1 cos
2
sin
2
8[x[
2
_
e
2
c
2
_
2
E
2
0
(11.26)
=
_
e
2
c
2
_
2
1 cos
2
sin
2
[x[
2
I
0
.
Therefore, the intensity per unit angle 1
r
:= I
r
(x)[x[
2
is
1
r
(, )
_
e
2
c
2
_
2
(1 cos
2
sin
2
)I
0
. (11.27)
Hence, the dierential cross-section is given by the Thomson formula
D(, ) :=
1
r
(, )
I
0
_
e
2
c
2
_
2
(1 cos
2
sin
2
). (11.28)
Finally, the total cross-section is given by
T :=
_
1
r
(, )d/I
0
=
_
D(, )d
_
e
2
c
2
_
2
_
(1 cos
2
sin
2
)d
=
_
e
2
c
2
_
2
_
4
_
2
0
cos
2
d
_

0
sin
3
d
_
=
_
e
2
c
2
_
2
8
3
. (11.29)
Remark 11.7 The dierential cross-section (11.28) depends on , hence it is not invariant with
respect to rotations around e
3
. This reects the fact that the incident wave is linearly polarized. If
we consider light with random polarization, then the dierential cross-section is given by (11.28) with
1/2 instead of cos
2
.
12. QUANTUM SCATTERING OF LIGHT. ZERO ORDER APPROXIMATION 81
12 Quantum Scattering of Light. Zero Order Approximation
We study scattering of light described as a plane wave by a quantum hydrogen atom in its ground
state, i.e. in the state of the lowest possible energy. We derive an energy ux and a dierential
cross-section.
The ground state energy is E
1
= 2h
R = e
4
/(2h
2
), and the corresponding wave function
(8.34) is
1
(x) = C
1
e
|x|/a
1
(we assume that the atom is situated at the origin). Then corresponding
solution to the Schrodinger equation is
1
(t, x) = C
1
e
|x|/a
1
e
i
1
t
,
1
= E
1
/h
= e
4
/(2h
3
). (12.1)
We want to describe the scattering of the plane wave (11.5) by the atom in the ground state. The
scattering is described by the coupled Maxwell-Schrodinger equations in the Born approximation
(7.26), (7.27). In the Lorentz gauge we have
[ih
t
e
ext
(x)](t, x) =
1
2
[ih
e
c
A
ext
(t, x)]
2
(t, x), (12.2)
_
_
1
4
2(t, x) = (t, x) = e[(t, x)[
2
,
1
4
2A(t, x) =
j(t, x)
c
=
e
c
[ih
e
c
A
ext
(t, x)](t, x) (t, x),
(12.3)
where
ext
= e/[x[ is the Coulomb eld of the nucleus and A
ext
stands for the incident wave (11.4)
(cf. (11.11)),
A
ext
(t, x) = Asin k(x
3
ct)(x
3
+ct)(1, 0, 0). (12.4)
12.1 Atom Form Factor
At zero order approximation (in A), the wave function is unperturbed, (t, x) =
1
(t, x). The
corresponding approximation to the radiation elds , A is given by the solutions to the Maxwell
equations (12.3) with (t, x) =
1
(t, x). Then the charge density (t, x) is static and the corresponding
potential is static with zero energy ux. Therefore, it suces to solve the equation for A with the
current
j(t, x)
c
=
e
c
[ih
1
(t, x)
1
(t, x)
e
c
A
ext
(t, x)
1
(t, x)
1
(t, x)]. (12.5)
Here, the rst term on the RHS is zero since the corresponding eigenfunction e
|x|/a
1
is real:
ih
1
(t, x)
1
(t, x) := Re
_
ih
1
(t, x)
1
(t, x)
_
= [C[
2
Re
_
ih
(e
|x|/a
1
)e
|x|/a
1
_
= 0. (12.6)
Therefore, the current is reduced to
j(t, x)
c
=
e
2
c
2
A
ext
(t, x)[
1
(x)[
2
. (12.7)
Let us split the solution
A(t, x) = A
ext
(t, x) +A
r
(t, x), (12.8)
where A
r
(t, x) is the radiated eld. Then (12.3) becomes
2A
r
(t, x) = 4
e
2
c
2
Asin k(x
3
ct)[
1
(x)[
2
(x
3
+ct)(1, 0, 0)
= 4
e
2
c
2
ImAe
ik(x
3
ct)
[
1
(x)[
2
(x
3
+ct)(1, 0, 0) =: f(t, x) (12.9)
82
since A
ext
is a solution to the homogeneous equation.
The radiation eld could be characterized uniquely by the initial condition of type (11.14) if the
atom radius would be zero as in the classical case of previous lecture. In our case the radiation eld
is of nite energy, i.e.
E
r
(t) :=
_
IR
3
_
[

A
r
(t, x)[
2
+ ([A
r
(t, x)[
2
_
dx < , t IR. (12.10)
An exact characterization of the radiation eld would be given by the limit E
r
(t) 0, t ,
however (12.10) is sucient for our purposes. Namely, let us demonstrate that Theorem 43.3 and
(12.10) with t = 0 imply the limiting amplitude principle holds, i.e.
A
r
(t, x) ImA
r
(x)e
ikct
, t . (12.11)
In particular, the zero order approximation to the radiation eld has the same frequency as the incident
wave.
Indeed, the asymptotics (12.11) follow similarly to (43.30). That is, Theorem 43.3 implies that the
long-time asymptotics of the nite-energy solution A
r
(t, x) is given by the retarded potential
A
r
(t, x)
_
f(t [x y[/c, y)dy
4[x y[
=
e
2
c
2
AIm
_
e
ik(y
3
c(t|xy|/c))
[
1
(y)[
2
dy
[x y[
(1, 0, 0)
=
e
2
c
2
AIme
ikct
_
e
ik(y
3
+|xy|)
[
1
(y)[
2
dy
[x y[
(1, 0, 0). (12.12)
Let us nd the asymptotics of the integral as [x[ . For any xed y IR
3
,
[x y[ = [x[ y n +o(1), [x[ , (12.13)
where n = n(x) = x/[x[. Let us also write y
3
= y e
3
. Then we get
A
r
(t, x)
e
2
c
2
AIme
ikct
e
ik|x|
[x[
_
e
iky(e
3
n)
[
1
(y)[
2
dy(1, 0, 0), [x[ . (12.14)
Next, we evaluate the last integral. Set K := k[e
3
n[ and denote by the angle between n and e
3
.
Then
K = K(k, ) = k
_
n
2
1
+n
2
2
+ (1 n
3
)
2
= k
_
2(1 n
3
) = k
_
2(1 cos ) = 2k sin

2
. (12.15)
Denote by the angle between y and e
3
n, and by the azimuthal angle around e
3
n. Finally,
let us take into account that the ground state
1
(y) =
r
1
([y[) is spherically symmetric. Then the
integral becomes
_

0
[y[
2
d[y[
_

0
sin d
_
2
0
de
iK cos |y|
[
1
(y)[
2
= 4
_

0
sin K[y[
K[y[
[
r
1
([y[)[
2
[y[
2
d[y[ =: F
a
(k, ) (12.16)
which is called the atom form factor.
Exercise 12.1 Calculate the last integral.
Since F
a
() is real, the asymptotics (12.14) becomes
A
r
(t, x)
e
2
c
2
A
sin k([x[ ct)
[x[
F
a
(k, )(1, 0, 0), [x[ . (12.17)
12. QUANTUM SCATTERING OF LIGHT. ZERO ORDER APPROXIMATION 83
12.2 Energy ux
We still have to calculate the Maxwell eld and the Poynting vector corresponding to this vector
potential. It suces to compare the expressions (12.17) with the vector potential of the Hertzian
dipole (formula (45.67) of Exercise 45.6):
A(t, x) =
1
c
p(t r/c)
r
. (12.18)
This is identical to (12.17) if F
a
() = 1 and
p(t) :=
e
2
c
2
k
Acos kct e
3
. (12.19)
Therefore, the energy ux S(t, x) corresponding to (12.17) is given, up to an error O([x[
3
), by the
Hertzian formula (11.22), with a factor [F
a
()[
2
. This follows from the fact that any dierentiation of
the form factor F
a
() in x
k
gives an additional factor with the decay O([x[
1
) since the form factor is
a homogeneous function of x. Finally, for the function (12.19), p(t) coincides with (11.23). Therefore,
(11.25) gives in our case
S(t, x) [F
a
()[
2
1 cos
2
sin
2
4c
3
[x[
2
_
e
2
_
2
E
2
0
cos
2
kc(t [x[/c)n, [x[ . (12.20)
Then the intensity and dierential cross-section in our case also contain the additional factor [F
a
()[
2
.
Hence, the dierential cross-section coincides with the Thomson formula (11.28) up to the atom form
factor:
D(k, , ) = [F
a
(k, )[
2
_
e
2
c
2
_
2
(1 cos
2
sin
2
). (12.21)
84
13 Light Scattering at Small Frequencies: Short-Range Scattering
Now we take into account the change of the ground state induced by the incident light (11.4), in the
rst order with respect to A, for small frequencies of light. This change describes the polarization of the
hydrogen atom (see previous lecture). It denes the corresponding dispersion and the combinational
scattering.
13.1 First Order Approximation to the Ground State
To calculate the rst approximation to the ground state, let us use the Schrodinger equation (12.2):
for small amplitudes [A[,
(t, x) =
1
(t, x) +Aw(t, x) +O(A
2
). (13.1)
Substituting into (12.2), we obtain in rst order in A,
A[ih
t
e
ext
(x)]w(t, x) = A
1
2
[ih
x
]
2
w(t, x) +
ih
e
c
A
ext
(t, x)
x
1
(t, x), (13.2)
since
1
(t, x) is a solution to Equation (12.2) with A
ext
= 0 and the Coulomb potential
ext
= e/[x[
of the nucleus. By (11.4) and (12.1), we have the following long-time asymptotics for the RHS of
(13.2),
ih
e
c
A
ext
(t, x)
x
1
(t, x)
ih
e
c
Asin k(x
3
ct)(1, 0, 0)
x
1
(x)e
i
1
t
=
Ah
e
2c
[e
ik(x
3
ct)
e
ik(x
3
ct)
]e
i
1
t
1
(x)
=
+
(x)e
i(
1
+)t
(x)e
i(
1
)t
, (13.3)
where := kc. Now let us apply the limiting amplitude principle (43.28):
w(t, x) = w
+
(x)e
i(
1
+)t
w
(x)e
i(
1
)t
+
l
C
l
l
(x)e
i
l
t
+r(t, x), (13.4)
where w
(x) are the limiting amplitudes,

l
(x) stand for the eigenfunctions of the discrete spectrum
of the homogeneous Schrodinger equation [ih
t
e
ext
(x)]w(t, x) =
1
2
[ih
x
]
2
w(t, x), which corre-
sponds to (13.2), and r(t, x) 0, t , in an appropriate norm. Below, we will omit the sum over
the discrete spectrum in the RHS of (13.4). According to Remark 43.11,
i) the limiting amplitudes w
(x) characterize the response of the atom to the incident wave and depend
on the frequency and Amplitude A of the incident wave, and do not depend on initial data.
ii) The sum over the discrete spectrum on the RHS of (13.4) depends only on initial data and does
not depend on the incident wave.
Hence, the contribution of the incident wave to the electric current and magnetization (see below),
in principle, can be singled out in an experimental observation of the considered scattering problem.
Finally, we identify
w(t, x) w
+
(x)e
i(
1
+)t
w
(x)e
i(
1
)t
, t +. (13.5)
For w
we get the equations (cf. (43.23))

[h
(
1
) e
ext
(x)]w
(x)
1
2
[ih
x
]
2
w
(x) =
(x) :=
h
e
2c
e
ikx
3
1
(x). (13.6)
Now we introduce the following condition:
spectral bound: [[ < [
1
[. (13.7)
13. LIGHT SCATTERING AT SMALL FREQUENCIES: SHORT-RANGE SCATTERING 85
Remark 13.1 Let us note that for the case of the hydrogen atom we have [
1
[ =
e
4
2h
3
20, 5
10
15
rad/sec. Hence, the bound (13.7) holds for the frequencies [[ < 3, 27 10
15
Hz or wave numbers
k < [
1
[/c 68 10
7
m
1
and wave lengths > 0.91176 10
5
cm = 911.76
A
.
The frequency bound implies that
1
< 0, hence the values h
(
1
) do not belong to the continuous
spectrum of the stationary Schrodinger equation (13.6). We will also assume that h
(
1
) do not
belong to the discrete spectrum. The last assumption is unessential since the coincidence h
(
1
)
with an eigenvalue h
n
is an event of codimension one, i.e. probability zero. Therefore, the solutions
decay at innity in the following sense
w
L
2
(IR
3
) (13.8)
(see Remark 43.10 ii)). For a moment let us neglect the external potential
ext
(x) in the equation
(13.6). Then it can be rewritten in the form
[z +
2
x
]w(x) = f(x), x IR
3
, (13.9)
where z < 0 and [f(x)[ Ce
|x|
with an > 0.
Exercise 13.2 Prove that w
(x) decay exponentially at innity, like
(x):
[w
(x)[ Ce
1
|x|
, x IR, (13.10)
where
1
> 0. Hint: Apply the Fourier transform to prove that i) the solution is unique in the class of
tempered distributions and ii) the solution is a convolution, w = f E, where E(x) is the fundamental
solution E(x) = e
|x|
/(4[x[) with :=
z > 0. Then
1
= min(, ) > 0.
Remark 13.3 We have supposed that the atom is in its groundstate,
1
, in the remote past, t .
The representation (13.1), the asymptotics (13.4) and (13.8) imply that, roughly speaking, the light
scattering modify the groundstate in the rst order in A, in the long-time limit t . The condition
(13.8) means that the scattering process is essentially concentrated in a bounded region of space.
Hence, the process corresponds to a short-range scattering.
Further, let us calculate the modied groundstate using the spectral resolution of the Schrodinger
operator in the equation (13.6). First, let us expand the RHS,
h
e
2c
e
ikx
3
1
(x) =
l
a
l

l
(x), (13.11)
where
l
stands for the sum over the discrete spectrum and the integral over the continuous spectrum.
Then the solutions w
have the form

w
(x) =
l
a
l

l
(x)
h
(
1

l
)
. (13.12)
Therefore, (13.5) becomes
w(t, x)
l
a
+
l

l
(x)
h
(
1
+
l
)
e
i(
1
+)t
l
a
l

l
(x)
h
(
1

l
)
e
i(
1
)t
. (13.13)
Let us calculate the coecients a
l
. Formally,
a
l
=
h
e
2c
_

l
(x)e
ikx
3
1
(x)dx. (13.14)
86
Let us assume that k 1 and set e
ikx
3
= 1. Then a
l
are approximately identical, and by partial
integration,
a
l

h
e
4c
_
[
l
(x)
1
1
(x)
1
l
(x)
1
(x)]dx
=
h
e
4c
e
i(
1
l
)t
_
[
l
(t, x)
1
1
(t, x)
1
l
(t, x)
1
(t, x)]dx
= i
1
2c
e
i(
1
l
)t
_
j
1
1l
(t, x)dx. (13.15)
Here
l
(t, x) := e
i
l
t
l
(x) and j
1
1l
(t, x) is the rst component of the current (7.22) corresponding
to
1
(t, x) and
l
(t, x), since the functions are solutions to the Schrodinger equation in (7.4) with
A(t, x) +A
ext
(t, x) = 0. Then the identity (10.20) implies,
a
l

1l
2c
_
x
1
e
1
(x)
l
(x)dx =
e
1l
2c
x
1
1l
=: a
l
, (13.16)
where
1l
:=
1
l
and x
1
1l
:=
_
x
1
1
(x)
l
(x)dx. Finally, (13.13) becomes,
w(t, x)
l
a
l
l
(x)
_
e
it
h
(
1l
+)

e
it
h
(
1l
)
_
e
i
1
t
, (13.17)
and the wave function (13.1) reads
(t, x) =
_
1
(x) +A
l
a
l
l
(x)
_
e
it
h
(
1l
+)

e
it
h
(
1l
)
__
e
i
1
t
+O(A
2
)
=
_
1
(x) +A(t, x)
_
e
i
1
t
+O(A
2
). (13.18)
13.2 Polarization and Dispersion: Kramers-Kronig Formula
Let us calculate the corresponding electric dipole moment. First, the charge density is given by
(t, x) = e(t, x)(t, x) = e(
1
(x) +A(t, x))(
1
(x) +A(t, x)) +O(A
2
)
= e[
1
(t, x)[
2
+eA
_
+
e
it
+
e
it
_
+O(A
2
), (13.19)
where
+
=
l
_

1
a
l
l
h
(
1l
+)

a
l
1
h
(
1l
)
_
,
=
+
. (13.20)
Therefore, the electric dipole moment equals, modO(A
2
),
p(t) :=
_
x(t, x)dx = p
11
+P(t), (13.21)
where p
11
:=
_
xe[
1
(t, x)[
2
dx = 0 by spherical symmetry, and
P(t) = eA
_
l
_
a
l
x
1l
h
(
1l
)

a
l
x
1l
h
(
1l
+)
_
e
it
+
l
_
a
l
x
1l
h
(
1l
)

a
l
x
1l
h
(
1l
+)
_
e
it
_
, (13.22)
where x
1l
:=
_
x
1
l
dx. By symmetry arguments, the vector P(t) is directed along e
1
. Indeed, the
invariance of p(t) with respect to the reection x
2
x
2
is obvious. The invariance with respect
13. LIGHT SCATTERING AT SMALL FREQUENCIES: SHORT-RANGE SCATTERING 87
to the reection x
3
x
3
follows from (13.11)(13.13) since we set k = 0. Therefore, substituting
a
l
= e
1l
x
1
1l
/(2c) and projecting x
1l
onto e
1
, we get
p(t) = Ae
1
4ke
2
h
1l
[x
1
1l
[
2
2
1l
2
cos t (13.23)
At last, let us average this expression with respect to all orientations of the atom. Then we obtain
p(t) := Ae
1
4ke
2
h
1l
[x
1
1l
[
2
2
1l
2
cos t, (13.24)
since the average of p
11
is obviously zero. Let us calculate the permittivity (42.30) of the atomic
hydrogen in the ground state
1
. We denote by E(t) the electric eld at the position x = 0 of the
atom: by (11.5), we have E(t) = kAcos t e
1
.
Corollary 13.4 The permittivity of the atomic hydrogen in the ground state is given by the Kramers-
Kronig formula
e
= n[p(t)[/[E(t)[ = n
4e
2
h
1l
[x
1
1l
[
2
2
1l
2
, (13.25)
where n is the number of atoms per unit volume.
This formula has the same analytic structure as its classical analog (45.80) of Exercise 45.7. This
allows us to express the hydrogen electric susceptibility = 1 + 4 (see (42.31)) and hence the
refraction coecient n
in the case of the magnetic susceptibility 1 (see (45.99)).

13.3 Combinational Scattering
Formula (13.18) gives the rst order correction to the unperturbed ground state
1
(x)e
i
1
t
in the
presence of the incident wave (11.5). Similarly, for any unperturbed stationary state
j
(x)e
i
j
t
, the
corresponding rst order correction is obtained by changing the index from 1 to j. Therefore, the rst
order correction to a general unperturbed solution of the type (10.2) reads
(t, x) =
j
c
j
(T)
_
j
(x) +A
l
a
l
l
(x)
_
e
it
h
(
jl
+)

e
it
h
(
jl
)
__
e
i
j
t
. (13.26)
Hence, we can apply the methods of Section 10 to calculate the spectrum of the dipole radiation of
the hydrogen atom in the presence of light. Now we get a new set of frequencies: in the rst order
approximation in A, the spectrum is contained in the set
jj
,
jj
. The corresponding selection
rules also can be obtained by the methods of Section 10.
88
14 Light Scattering in Continuous Spectrum: Photoeect
For light with a frequency which is larger than the binding energy of the stationary electron state,
scattering of the electron into the continuous spectrum may occur. Within the coupled Maxwell-
Schrodinger system this scattering is described by a weakly decaying contribution to the electron
wave function, which corresponds to a non-zero electron current radiating to innity.
14.1 Radiation in Continuous Spectrum
Let us consider the scattering of light with frequency
> [
1
[. (14.1)
Then
1
< 0 and
1
+ = [
1
[ > 0. This means that h
(
1
+ ) belongs to the continu-
ous spectrum of the stationary Schrodinger equation (13.6) with a slow decay of the wave function
[w
+
(x)[ 1/[x[ at [x[ . We will calculate the long-range asymptotics of w
+
(x) and obtain the
main term of the radiation in the form
Aw
+
(x)e
i(
1
+)t
A
a(, )
[x[
e
i[k
r
|x|(
1
+)t]
, [x[ . (14.2)
On the other hand, h
(
1
) does not belong to the continuous spectrum. Let us also assume that
h
(
1
) does not belong to the discrete spectrum, hence
w
L
2
(14.3)
similarly to (13.8). We will deduce from the asymptotics (14.2) the following stationary electric current
at innity,
j(t, x) A
2
eh
k
r
a
2
(, )
[x[
2
n(x), [x[ , (14.4)
where n(x) := x/[x[. This describes the radiation of a non-zero electric current from the atom to
innity.
Remark 14.1 i) The Schr odinger theory explains the red bound of the photoeect, >
red
, and
provides the value
red
= [
1
[.
ii) Similarly to (7.10), the radiated wave function (14.2) can be identied with a beam of electrons
with the density
d(x) = (t, x)/e = [Aw
+
(x)[
2
= A
2
a
2
(, )
[x[
2
(14.5)
and the velocities
v(x) = j(t, x)/(ed(x)) =
h
k
r
n(x). (14.6)
14.2 Limiting Amplitude
Let us calculate the limiting amplitude w
+
(x). First, we rewrite Equation (13.6) in the form
[
2
x
+k
2
r
()]w
+
(x) =
e
h
c
e
ikx
3
1
(x)
2e
2
2
[x[
w
+
(x), (14.7)
14. LIGHT SCATTERING IN CONTINUOUS SPECTRUM: PHOTOEFFECT 89
where k
r
() =
_
2(
1
+)/h
> 0. For a moment, let us neglect the last term on the RHS. So, we
get the Helmholtz equation
[
2
x
+k
2
r
()]w
+
(x) = f
+
(x) :=
e
h
c
e
ikx
3
1
(x). (14.8)
The limiting amplitude is given by the convolution (cf. (43.27))
w
+
(x) =
_
e
ik
r
()|xy|
4[x y[
f
+
(y)dy. (14.9)
This follows from the limiting absorption principle (43.32) since the fundamental solution
e
ik
r
(+i)|xy|
4[x y[
is a tempered distribution for small > 0. Indeed, Imk
r
( +i) > 0 for the xed branch k
r
() > 0.
Now we can calculate the asymptotics (14.2). Let us substitute the expression (14.8) for f
+
into
(14.9). Then, after partial integration, we get
w
+
(x) =
e
h
c
_

1
e
ik
r
|xy|
4[x y[
e
iky
3
1
(y)dy
=
ik
r
e
h
c
_
e
ik
r
|xy|
(x
1
y
1
)
4[x y[
2
e
iky
3
1
(y)dy +O([x y[
2
). (14.10)
Substituting here x
1
y
1
= sin cos [x y[, we get (14.2) with the amplitude
a(, ) = C sin cos , (14.11)
since the ground state
1
(y) decays rapidly at innity.
14.3 Long-Range Scattering: Wentzel Formula
Let us deduce (14.4) from (14.2). In the Born approximation the current is given by (7.27),
j :=
e
[ih
(t, x)] (t, x)

e
2
c
A
ext
(t, x)[(t, x)[
2
. (14.12)
The function w
decays at innity by (14.3). Therefore, (13.1) becomes

(t, x)
1
(t, x) +Aw
+
(x)e
i(
1
+)t
. (14.13)
The contribution of the function
1
to the current at innity also is negligible due to the exponential
decay at innity. Therefore, (14.4) follows by substituting the asymptotics (14.2) into the rst term
on the RHS.
Finally, (14.11) implies the following stationary electric current at innity,
j(t, x)
sin
2
cos
2
[x[
2
n(x), [x[ (14.14)
which has been discovered rst by G.Wentzel in 1927 (see [100]).
Remark 14.2 The slow decay (14.2), (14.14) implies that the total electric current to innity does
not vanish, i.e.
lim
R
_
|x|=R
j(t, x)dS(x) = J ,= 0. (14.15)
This corresponds to the long-range scattering under the condition (14.1). If the condition fails, the fast
decay (13.10) implies that in this case the total electric current to innity vanishies that corresponds
to the short-range scattering under the condition (13.7) (see Remark 13.3).
90
Comment 14.3 The stationary nonvanishing current (14.15) formally contardicts the charge con-
servation for the atom. The contradiction is provided by the perturbation strategy which leaves the
current (12.5) unchanged. To maintain the stationary photocurrent, one needs an external source
(galvanic element, etc) to reimburse the charge decay.
14.4 Coulomb Potential
Now let us discuss the general equation (14.7) instead of (14.8). Then (14.9) changes to
w
+
(x) =
_
G
k
r
(x, y)f
+
(y)dy, (14.16)
where G
k
r
is the corresponding Green function. For each xed y IR
3
the following asymptotics
holds (cf. [89, Vol. II, formula (II.7.33)]),
G
k
r
(x, y) C
1
e
i log |xy|
e
ik
r
|xy|
4[x y[
, (14.17)
where = (k
r
) IR. Therefore, the asymptotics (14.2), (14.4) follow by the same arguments as
above, with the amplitude of the form (14.11).
Exercise 14.4 Deduce (14.2), (14.4), (14.11) from (14.16) and (14.17).
14.5 Shift in Angular Distribution: Sommerfeld Formula
Next correction to Wentzel formula has been obtained by Sommerfeld and Shur [90]. Namely, they have
taken into account the pressure of the incident light upon the ougoing photocurrent. The corresponding
corrected formula reads (see [89, Vol. II])
j(t, x)
sin
2
cos
2
[x[
2
n(x), [x[ . (14.18)
Comment 14.5 The Sommerfeld-Shur calculation continues the perturbation procedure for the cou-
pled Maxwell-Schr odinger equations. It would be interesting to prove that the photoeect is an inher-
ent feature of the coupled Maxwell-Schr odinger equations and holds under the same spectral condition
(14.1).
14.6 Photoeect for Excited States
The angular dependence (14.11) of the limiting amplitude is characteristic for the light scattering by
spherically-symmetric ground state (12.1). The photoeect is observed also in the light scattering by
excited stationary states of atoms which are not spherically-symmetric. In this case, the formula of
type (14.16) for the limiting amplitude also holds and admits a long-range asymptotics of type (14.2).
However, its angular distribution is dierent from (14.11): see [89], where the distribution is obtained
for the excited states from K-, L-, and M-shells of many-electron atoms.
15. SCATTERING OF PARTICLES: RUTHERFORD FORMULA 91
15 Scattering of Particles: Rutherford Formula
Here, the scattering of charged particles at a heavy, charged object (nucleus) is considered. First,
the classical case is considered, when the scattering particles are described by the trajectories of
classical mechanics. Then, the quantum mechanical scattering of an electron with large momentum
by a hydrogen atom is studied.
15.1 Classical Scattering by a Nucleus
Let us consider scattering of charged particles with charge Q and mass M by a nucleus with charge
[e[Z > 0. We will derive the Rutherford formula in both the repulsive case, when Q > 0, and the
attractive case, when Q < 0. The repulsive case corresponds, for example, to the scattering of -
particles with Q = 2[e[. The attractive case corresponds, for example, to the scattering of electrons
with Q = e < 0. Let the heavy nucleus be situated at the origin. The incident particle moves along
the trajectory x = x(t), < t < . We assume that the particles come from innity, i.e.
lim
t
[x(t)[ = . (15.1)
Then the trajectory is a hyperbola as described, e.g., in Exercise 45.1. Let us choose the coordinates
x
k
in space in such a way that x
3
(t) 0 and
lim
t
x(t) = (v, 0, 0), lim
t
x
1
(t) = , lim
t
x
2
(t) = b, (15.2)
where b is the impact parameter. Let us further choose standard polar coordinates in the plane x
3
= 0,
x
1
= r cos , x
2
= r sin (15.3)
and denote by r(t), (t) the trajectory of the particle in these coordinates. Then the initial scattering
conditions (15.2) imply
lim
t
(t) = , lim
t
r(t) sin (t) = b. (15.4)
Angle of scattering
Let us calculate the nal scattering angle
:= lim
t
(t). (15.5)
Lemma 15.1 The nal scattering angle satises the equation
cot

2
=
Mbv
2
Q[e[Z
. (15.6)
Proof First, let us write the angular momentum and energy conservation:
r
2
(t)
(t) = bv,
M
2
( r
2
(t) +r
2
(t)
2
(t)) +
Q[e[Z
r(t)
=
M
2
v
2
. (15.7)
Let us substitute here
r(t) :=
dr
dt
=
dr
d
d
dt
= r
. (15.8)
92
Then the energy conservation becomes,
M
2
2
(t)([r
(t)[
2
+r
2
(t)) +
Q[e[Z
r(t)
=
M
2
v
2
. (15.9)
Now let us make the Clerot substitution r = 1/. Then r
/
2
and the momentum conserva-
tion gives

(t) = vb
2
. Therefore, (15.9) reads,
M
2
b
2
v
2
([
[
2
+
2
) +Q[e[Z =
M
2
v
2
. (15.10)
Let us dierentiate this expression in . Then after division by
, we get the Clerot equation
+ = C :=
Q[e[Z
Mbv
2
. (15.11)
The general solution to this equation is
() = Acos +Bsin +C. (15.12)
The initial scattering conditions (15.2) give,
lim
() = 0, lim
()
sin()
=
1
b
. (15.13)
Substituting here (15.12), we get A+C = 0 and B = 1/b, hence
() = C(1 + cos ) +
1
b
sin . (15.14)
At last, for the nal scattering angle we get from (15.5) that () = 0. Hence,
C(1 + cos ) +
1
b
sin = 0. (15.15)
This implies (15.6).
Remark 15.2 The solution (0, ) (, 2) to Equation (15.6) exists and is unique. The repulsive
and attractive cases correspond to Q > 0, (0, ) and Q < 0, (, 2), respectively.
Dierential cross section
Now let us assume that the incident particles constitute a beam with a ux density of n particles per
cm
2
sec. Let us denote by N = N(b, b + db) the number of incident particles per sec with an impact
parameter within the interval [b, b +db]. By axial symmetry, we get for innitesimal db,
N(b, b +db) = n2bdb (15.16)
The particles are scattered into the spatial angle d = 2 sin d.
Denition 15.3 The dierential cross section of the scattering is dened by
D() :=
N/d
n
=
bdb
sin d
. (15.17)
Let us calculate the cross section. Rewriting (15.6) in the form
b
2
=
_
Q[e[Z
Mv
2
_
2
cot
2

2
(15.18)
and dierentiating, we get
2bdb =
_
Q[e[Z
Mv
2
_
2
2 cot

2
1
sin
2
2
d
2
. (15.19)
Substituting this into (15.17), we get the Rutherford formula
D() =
_
Q[e[Z
Mv
2
_
2
4 sin
4
2
. (15.20)
15.2 Quantum Scattering of Electrons by the Hydrogen Atom
We consider the scattering of the electron beam by a hydrogen atom in its ground state
1
(t, x),
(12.1). The incident electron beam is described by the plane wave (7.6):
in
(t, x) = Ce
i(kxt)
, k ,= 0, (15.21)
which is a solution to the free Schrodinger equation, i.e., without an external Maxwell eld. Hence,
(see (7.7))
h
=
h
2
2
k
2
> 0. (15.22)
Corresponding electric current density is given by (7.9):
j
in
(t, x) :=
e
[ih
in
(t, x)]
in
(t, x) =
eh
[C[
2
. (15.23)
We will assume that [C[ C
1
(see (12.1)) and consider the incident plane wave as a small perturbation.
Hence, the total wave eld, approximately, is a solution to the Schrodinger equation of the type (7.2),
[ih
t
e(x)](t, x) =
1
2
[ih
e
c
A(x)]
2
(t, x), (15.24)
where (x), A(x) are the potentials of the total static Maxwell eld corresponding to the ground state
1
. In particular (cf. (12.3)),
1
4
(x) = (x) = e[
1
(t, x)[
2
+[e[(x), (15.25)
where [e[(x) is the charge density of the nucleus. The potential decays at innity like [x[
2
, since
_
(x)dx = 0. For simplicity of calculations, we assume rst that
(x) = e
e
|x
[x[
, (15.26)
where > 0 is small. At the end we will perform the limit 0.
94
Radiated wave
Let us split the total wave eld into three terms,
(t, x) =
in
(t, x) +
1
(t, x) +
r
(t, x), (15.27)
where
r
(t, x) is a small radiated wave. Substituting (15.27) into the Schrodinger equation (7.2), we
get
[ih
t
e(x)][
in
(t, x) +
r
(t, x)] =
1
2
[ih
e
c
A(x)]
2
[
in
(t, x) +
r
(t, x)], (15.28)
since
1
(t, x) is an exact solution. Neglecting the relativistic corrections due to the small term
e
c
A(x), we rewrite the equation as
_
ih
t
e(x)
1
2
[ih
x
]
2
_
r
(t, x) = e(x)
in
(t, x) = e(x)Ce
i(kxt)
. (15.29)
Since > 0 by (15.22), the frequency belongs to the continuous spectrum of the Schrodinger
operator. Now let us apply the limiting amplitude principle (43.28) to derive the long-time asymptotics
of the solution
r
(t, x). Up to a contribution of the discrete spectrum (see Remark 43.11), the solution
admits the asymptotics
r
(t, x)
r
(x)e
it
, t . (15.30)
Substituting this asymptotics into Equation (15.29), we get a stationary equation for the limiting
amplitude
_
h
e(x)
1
2
[ih
x
]
2
_
r
(x) = eC(x)e
ikx
. (15.31)
Neglecting the term with the spatial decay at the LHS, we nally get the Helmholtz equation
_
k
2
+
2
x
_
r
(x) =
2eC
h
2
(x)e
ikx
, (15.32)
since k
2
=
2
h
> 0 by (15.22). This is an equation of the type (14.8). Therefore, the solution is
given by a convolution similar to (14.9),
r
(x) =
2eC
h
2
_
e
ik|xy|
4[x y[
(y)e
iky
dy, k := [k[. (15.33)
The convolution (15.33) is almost identical to the last integral (12.12) if we identify k = (0, 0, k).
Evaluating this by the method (12.13)-(12.16), we get
r
(x) C
e
ik|x|
[x[
f(k, ), [x[ , (15.34)
where
f(k, ) =
2e
h
2
_

0
sin K[y[
K[y[
(y)[y[
2
d[y[, K := 2k sin

2
. (15.35)
Now (15.30) becomes for large [x[
r
(t, x) C
e
ik|x|
[x[
f(k, )e
it
, t . (15.36)
The limiting amplitude
r
has a slow decay at innity and innite energy. This corresponds to the
fact that the frequency > 0 belongs to the continuous spectrum (see Remark 43.10 ii)). Physically
this describes the radiation of electrons to innity as we will see below.
Dierential cross section
Let us calculate the electric current corresponding to the radiated wave
r
(t, x). As in (15.23), we
have for large [x[,
j
r
(t, x) :=
e
[ih
r
(t, x)]
r
(t, x)
eh
kn(x)
[f(k, )[
2
[x[
2
[C[
2
, n(x) :=
x
[x[
. (15.37)
Let us note that the current at innity is radial.
Denition 15.4 The dierential cross section of the scattering is dened by
D(n) := lim
x/|x|=n,|x|
[j
r
(t, x)[
[j
in
(t, x)[
[x[
2
, n IR
3
. (15.38)
Remark 15.5 The denition obviously agrees with Denition 15.3.
Now (15.37) and (15.23) imply that
D(n) = [f(k, )[
2
. (15.39)
Finally, substituting (15.26) into (15.35), we get
f(k, ) =
2e
2
Kh
2
_

0
sin K[y[e
|y|
d[y[ =
2e
2
h
2
(K
2
+
2
)
. (15.40)
Here K = 2k sin

2

2
for ,= 0 and large values of k > k(). Then we can drop
2
in the limit
0 and obtain
f(k, ) =
2e
2
h
2
K
2
=
e
2
2h
2
k
2
sin
2

2
. (15.41)
We rewrite this expression using (7.10):
f(k, ) =
e
2
2v
2
sin
2

2
. (15.42)
Now (15.39) implies the formula
D(n) =
_
e
2
v
2
_
2
4 sin
4

2
, (15.43)
which coincides with the classical Rutherford formula (15.20) with Q = e and Z = 1.
Remark 15.6 Born considered the agreement of (15.43) with the classical formula (15.20) as the
crucial conrmation of the probabilistic interpretation of the wave function: [(t, x)[ is the density
of the probability of the particle registration, and the expression (15.37) is e times the ux of
the probability, [12].
96
16 Hydrogen Atom in a Magnetic Field. Normal Zeemann Eect
We derive quantum stationary states and the corresponding energies for a hydrogen atom in a uniform
magnetic eld. We also analyze the normal Zeemann eect.
Let us consider a hydrogen atom in an external static magnetic eld B(x) with vector potential
A(x). Then the Lorentz gauge condition (4.5) is equivalent to the Coulomb gauge condition
x
A(x) = 0. (16.1)
The Schrodinger equation (6.2) becomes
ih
t
(t, x) =
1
2
(ih
e
c
A(x))
2
(t, x)
e
2
[x[
(t, x), (t, x) IR
4
. (16.2)
Evaluating, we get by (16.1)
ih
t
(t, x) =
1
2
h
2
(t, x) +i
h
e
c
A(x)
x
(t, x)
1
2
h
2
A
2
(x)(t, x)
e
2
[x[
(t, x), x IR
3
. (16.3)
Let us assume that the potential A(x) is small:
[A(x)[ 1. (16.4)
Then we can neglect the term with A
2
(x) in (16.9) and get the equation
ih
t
(t, x) =
1
2
h
2
(t, x) +i
h
e
c
A(x)
x
(t, x)
e
2
[x[
(t, x), x IR
3
. (16.5)
16.1 Uniform Magnetic Field
Now let us consider a hydrogen atom in an external uniform static magnetic eld B with vector
potential A(x) = Bx/2. Let us choose the coordinates in such a way that B = (0, 0, B) with B 0.
Then the vector potential is given by A(x) =
1
2
B(x
2
, x
1
, 0) =
1
2
B[x[ sine
. Therefore, we have the

static Maxwell eld with the potentials
(x) :=
e
[x[
,
A(x) :=
1
2
B(x
2
, x
1
, 0)
x IR
3
. (16.6)
Remark 16.1 The angular momentum conservation (6.13) holds for n = 3:
L
3
(t) := (t, x),

L
3
(t, x)) = const, (16.7)
where

L
3
is the operator in (6.18) with n = 3. The conservation reects the axial symmetry of Equation
(16.5) with the potentials (16.6).
16. HYDROGEN ATOM IN A MAGNETIC FIELD. NORMAL ZEEMANN EFFECT 97
Let us note that ih
A(x)
x
= Bih
/2 = B
L
3
/2 by (6.18). Here, is the angular coordinate of
the rotation around the vector e
3
:= (0, 0, 1), i.e. around B. Then (16.5) becomes
ih
t
(t, x) =
1
2
h
2
(t, x)
e
2c
B
L
3
(t, x)
e
2
[x[
(t, x), x IR
3
. (16.8)
Hence, the corresponding stationary equation reads
E
(x) =
1
2
h
(x)
L
L
3
(x)
e
2
[x[
(x), x IR
3
, (16.9)
where
L
:= eB/(2c) < 0 is the Larmor frequency. Therefore, Theorem 8.1, Theorem 8.4 ii) and
(8.32) imply the following theorem.
Theorem 16.2 i) The quantum stationary states
of the hydrogen atom in a uniform magnetic

eld coincide with the functions
lmn
from (8.34).
ii) The corresponding energies are equal to
E
mn
:= 2h
R/n
2
+mh
L
, n = 1, 2, 3..., m = l, ..., l, l n 1. (16.10)
Radial scalar potential Let us consider more general equations of the type (16.6) with a static
radial external potential ([x[) and the static uniform magnetic eld B. Then the corresponding
eigenvalue problem of the type (16.8) becomes
E
(x) =
1
2
h
(x) ih
(x) +e([x[)
(x), x IR
3
. (16.11)
The following theorem can be proved by the same arguments as in Theorems 8.1, 16.2.
Theorem 16.3 i) The quantum stationary states of the electron in a static central electric and a
uniform magnetic eld have the following form in spherical coordinates (cf. (8.34)):
lmn
= R
ln
(r)F
m
l
()e
im
. (16.12)
ii) The corresponding energies are equal to
E
lmn
= E
ln
+mh
L
, n = 1, 2, 3..., m = l, ..., l, l n 1, (16.13)
where E
ln
are the energies corresponding to B = 0.
16.2 Normal Zeemann Eect: Triplet Spectrum
In 1895 Zeemann observed the inuence of a magnetic eld on the spectral lines of atoms and molecules.
Consider the spectrum of radiation of the hydrogen atom in a uniform magnetic eld. Let us calculate
the level splitting and the polarization of the radiation emitted by the atom: we will see that both are
as in the classical description (see Exercise 45.9). Namely, the unperturble spectral lines, corresponding
to the zero magnetic eld B = 0, split into the normal triplet in the case B ,= 0.
First, by Theorem 16.2, the bound state eigenfunctions are (8.34):
k
= R
nl
(r)F
m
l
()e
im
, k = (nlm), (16.14)
and the eigenvalues are
k
=
2R
n
2
+m
L
,
L
=
eB
2c
. (16.15)
98
In the case B = 0, we have also
L
= 0. Hence, the unperturbed spectral lines,
0
kk
= 2R[
1
n
2

1
n
2
],
where k = (nlm) and k
= (n
).
Next consider the case B ,= 0. The calculations (10.18)-(10.22) imply the formula for the intensity
of the spectral line
kk
in the dipole approximation
J
kk

_

0
R
nl
(r)R
n
l
(r)r
3
dr
_
S
xF
m
l
()F
m
()e
i(mm
)
dS. (16.16)
We know this integral is nonzero only for m
= m, m 1 and l
= l 1. Hence, for B ,= 0, the

unperturbed spectral line
0
kk
generates the triplet of spectral lines with frequencies
_
_
m
= m :
kk
=
0
kk
= m1 :
kk
=
0
kk
+
L
m
= m+ 1 :
kk
=
0
kk

L
,
(16.17)
precisely like in the classical case. To nd the polarizations we re-write the vector x like
x = sin (e
i
e
+
+e
i
e
) + cos e
z
, e
=
1
2
( e
x
i e
y
). (16.18)
Inserting back into (16.16) we nd that for m
= m only the e
z
component contributes, and for
m
= m1 the e
component. Therefore, we nd from (10.12) the radiation in the form

_
_
m
= m : A
kk
(t, x)
1
[x[
Rec
k
c
k
e
z
e
i
0
(t|x|/c)
m
= m1 : A
kk
(t, x)
1
[x[
Rec
k
c
k
e
+
e
i(
0
+
L
)(t|x|/c)
m
= m+ 1 : A
kk
(t, x)
1
[x[
Rec
k
c
k
e
e
i(
0
L
)(t|x|/c)
.
(16.19)
Note that the resulting polarizations of the radiation elds are exactly like in the classical case (see
Exercise 45.9).
Remark 16.4 The spectral lines (16.17) do not depend on the azimuthal quantum number l as well as
the eigenvalues (16.15). Therefore, the radiations, induced by the pairs k = (mln), k
= (m
) with
m
= m, m 1 and all possible azimuthal quantum number l, l
, contribute to the same frequencies

(16.17). On the other hand, in many cases the atom spectra in a magnetic eld demonstrate the
multiplet structure drustically dierent from the triplet structure (16.17). This anomalous Zeemann
eect cannot be explained by the Schrodinger equation (16.2). The explanation is provided by the
Pauli equation which takes into account the electron spin (see Section 22.3).
17. DIAMAGNETISM AND PARAMAGNETISM 99
17 Diamagnetism and Paramagnetism
We compute a magnetic moment of a hydrogen atom and split it in two parts, one of which describes
a diamagnetism of the atom and the other describes a paramagnetism. This allows us to calculate the
magnetic susceptibility of the hydrogen atom (see Lecture 14).
Namely, we will consider the stationary states of the atom in the uniform external magnetic eld
B and calculate their magnetic moment (42.22)
m =
1
2c
_
y j(y)dy. (17.1)
Let us choose the coordinates in such a way that B = (0, 0, B) with B 0. Then the corresponding
vector potential is given by A(x) =
1
2
B(x
2
, x
1
, 0) =
1
2
B[x[ sine
.
17.1 Electric Current at Stationary States
First we calculate the electric current j dened by the Born approximation (7.27),
j(t, x) =
e
[ih
e
c
A(x)](t, x) (t, x). (17.2)
The stationary states are given by (8.34) (see Theorem 16.2). Let us write (8.34) in the form
(t, x) = a
lmn
(r, )e
im
e
it
, (17.3)
where a
lmn
is a real function according to Theorem 8.4 ii). Let us express the gradient operator in
spherical coordinates (see (9.15)):
= e
r
r
+e
r
+e
r sin
. (17.4)
Then (17.2) and (17.3) give,
j(t, x) =
e
[ih
e
r
r
a
lmn
(r, )] a
lmn
(r, ) +
e
[ih
r
a
lmn
(r, )] a
lmn
(r, )
+
e
[mh
1
r sin
](t, x) (t, x)
e
2
c
A(x)(t, x) (t, x). (17.5)
The rst and second term on the RHS are zero since the function a
lmn
(r, ) is real. Hence nally,
j(t, x) =
e
[mh
1
r sin
][(t, x)[
2
e
2
c
B[x[ sin e
[(t, x)[
2
= j
(x) +j
(x). (17.6)
The magnetic moment (17.1) becomes
m =
1
2c
_
y j
(y)dy +
1
2c
_
y j
(y)dy = m
+m
. (17.7)
The currents j
(x), j
(x) are axially symmetric with respect to the axis Ox

3
, and
(y e
)
3
= [y[e
3
sin . (17.8)
100
17.2 Langevin Formula for Diamagnetic Susceptibility
By (17.8), the direction of m
is opposite to B, hence, m
describes the diamagnetism of the atom.

The corresponding diamagnetic susceptibility (per atom) is dened by m
=
m
B, hence
m
:=
[m
[
B
=
e
2
2c
2
_
[y[
2
sin
2
[[
2
dy =
e
2
2c
2
3
, (17.9)
where
3
denotes the moment of inertia of the distribution [[
2
with respect to the axis Ox
3
. For
spherically symmetric stationary states we have
3
=
1
3
, :=
_
[y[
2
[[
2
dy. (17.10)
Then (17.9) becomes the Langevin formula (cf. Exercise 45.10)
m
=
e
2
6c
2
. (17.11)
For spherically non-symmetric states the formula also holds in the mean due to the random orientation
of the atoms with respect to the direction of the magnetic eld.
17.3 Paramagnetism
The moment m
is dierent from zero even in the absence of the magnetic eld,

m
= mh
e
2c
_
(y e
)
3
e
3
[y[ sin
[(t, y)[
2
dy = mh
e
3
e
2c
_
[(t, y)[
2
dy =
e
2c
mh
e
3
(17.12)
by (17.8). Hence, m
describes the paramagnetism of the atom. However, its mathematical expectation

is zero due to the random orientation of the atom since the probabilities of the values m and m are
identical. On the other hand, the probabilities are not identical if the magnetic eld does not vanish.
The corresponding statistical theory has been developed by Langevin.
Part IV
Electron Spin and Pauli Equation
101
18. ELECTRON SPIN: EXPERIMENTS AND INTERPRETATION 103
18 Electron Spin: Experiments and Interpretation
We describe some experiments which have inspired the ideas of a spin and a magnetic moment of an
electron, namely, the Einstein-de-Haas experiment, the Stern-Gerlach experiment, and the anomalous
Zeemann eect.
The concept of spin and magnetic moment of the electron has been introduced by Goudsmith and
Uhlenbeck in 1925: Every electron has an intrinsic angular momentum (spin) s with the magnitude
[s[ =
h
2
, and a magnetic moment m
s
with the magnitude [m
s
[ =
B
:=
[e[h
2c
, which is the Bohr
magneton. Here the term intrinsic means that the spin angular momentum is not related to a
rotation of the particle, and spin magnetic moment is not related to the corresponding convection
current.
Let us consider the experimental data which inspired this conjecture.
18.1 Einstein-de Haas Experiment
Let us consider an iron bar positioned vertically in the earths gravitational eld, which is attached
to a vertical string in such a way that it can rotate about its axis. Let us magnetize the bar by a
vertical external weak magnetic eld. The eld orients the elementary Ampere molecular currents,
so that their magnetic moments increase the external magnetic eld. Therefore, the corresponding
elementary angular momenta are also oriented in the same direction, and the sum of the microscopic
angular momenta increases. By axial symmetry, the total angular momentum of the bar, macro-
scopic+microscopic, is conserved. Hence, the bar as a whole must change its macroscopic angular
momentum, too. In 1915 Einstein and de Haas measured the changes of magnetic moment and
macroscopic angular momentum of the bar to check their ratio. The classical and quantum theo-
ries predict the ratio
e
2c
. However, the experimental results contradict this value. This suggests the
existence of an additional magnetic moment of the atoms which is responsible for the anomalous ratio.
Classical theory
Let us assume that in each molecule, the currents are caused by the rotation of electrons with the
same angular velocity . Then the angular momentum and magnetic moment of the molecule are
given by
L :=
x
k
( x
k
), m :=
1
2c
x
k
e( x
k
), (18.1)
where x
k
are the electron positions. Therefore, the following key relation holds for each molecule,
m =
e
2c
L. (18.2)
Hence, an increment of the magnetization, m
3
, of the bar, is followed by the corresponding increment,
L
3
=
e
2c
m
3
, of its macroscopic angular momentum. Both quantities, m
3
and L
3
, can
be measured experimentally: L
3
by the torsion vibration of the string, and m
3
by the residual
magnetism. The experiment was performed by Einstein and de Haas in 1915. However, the result was
m
3
= g
e
2c
L
3
(18.3)
with the Lande factor g 2, which contradicts (18.2). This contradiction inspired the ratio [s[/[m
s
[ =
e
c
in the Stern-Gerlach conjecture. This ratio corresponds to the Lande factor g = 2 and allows to
explain the result of the Einstein-de Haas experiment.
104
Quantum theory
Let us check that the relation (18.2) also holds for the Schr odinger equation (16.8) for small B. Indeed,
(17.1) and (17.2) imply for small B,
m
1
2c
_
y
e
[ih
y
](t, y) (t, y)dy
=
e
2c
_
[ih
y
y
](t, y) (t, y)dy =
e
2c
L. (18.4)
In particular, for the stationary state (17.3) we have the magnetization (cf. (17.12))
m
3

e
2c
mh
. (18.5)
On the other hand, the energy for the stationary state is given by (16.13). The external magnetic eld
causes the transitions to stationary states with lower values of the energy (16.13), which corresponds
to greater values of m since
L
< 0. Therefore, the magnetization (17.12) increases in the transitions.
18.2 Double Splitting
An additional suggestion for the spin is provided by the observation of the splitting of stationary states
of atoms and molecules in a magnetic eld B.
Stern-Gerlach experiment
In 1922 Stern and Gerlach sent a beam of silver atoms through an inhomogeneous magnetic eld.
Later similar experiments have been performed with hydrogen atoms. The atoms are in the ground
state (12.1), which is non-degenerate according to the Schr odinger equation (16.8). Let us write (16.8)
in the form
ih
t
(t, x)=
1
2
h
2
(t, x) +e([x[)(t, x)
e
2c
LB(t, x), x IR
3
. (18.6)
The interaction term

LB(t, x) vanishes for the ground state. Therefore, the ground state also satises
Equation (18.6) with B ,= 0. However, Stern and Gerlach observed a splitting of the beam into two
components. This obviously contradicts the identity of all atoms in the ground state and suggests
that
i) For B = 0 the eigenspace corresponding to the ground state has a dimension at least two.
ii) For B ,= 0 the eigenspace splits in two distinct eigenspaces.
This suggests the existence of an additional magnetic moment of the atoms which is responsible for
the splitting in the magnetic eld.
Anomalous Zeemann eect
The Schrodinger equation gives a satisfactory explanation of the normal Zeemann eect with splitting
into three lines. However, Zeemann demonstrated in 1895 that, for B ,= 0, most of the spectral lines
are split into a dierent number of lines: two, ve, etc. This anomalous Zeemann eect contradicts the
Schrodinger theory, which only predicts the normal Zeemann eect. This also suggests the existence
of an additional magnetic moment responsible for the anomalous splitting.
19. PAULI EQUATION 105
19 Pauli Equation
The Einstein-de Haas, Stern-Gerlach and anomalous Zeemann eects demonstrate that the Schrodinger
equation requires a modication. To the Schrodinger equation, we introduce an additional spin mag-
netic moment corresponding to spin angular momentum. Then the Schrodinger equation becomes the
Pauli equation which explains the Stern-Gerlach eect. Let us analyze the details.
19.1 Additional Magnetic Moment
Relation (18.4) implies that, for small [B[, the last term in (18.6) can be rewritten as
e
2c
LB(t, x) mB(t, x), (19.1)

where m is the quantum observable corresponding to the magnetic moment. This corresponds to the
fact that mB is the energy of the magnetic moment m in the magnetic eld B. The Stern-Gerlach
experiment demonstrates that the magnetic moment m does not vanish even for the ground state of
the hydrogen atom. Therefore, we have to postulate the existence of an additional magnetic moment
m
s
,= 0. The Einstein-de Haas experiment suggests the Goudsmith-Uhlenbeck conjecture that the
additional magnetic moment corresponds to an additional spin angular momentum of the electron,
s =spin, with the Lande factor 2, i.e.
m
s
= 2
e
2c
s. (19.2)
Next, we have to specify a mathematical construction for this spin angular momentum s.
19.2 Additional Angular Momentum
Orbital angular momentum
First, let us analyze the orbital angular momentum L = (L
1
, L
2
, L
3
) for the case A
k
= 0, k = 1, 2, 3
and spherically symmetric scalar potential ([x[):
ih
t
(t, x)= H(t, x) :=
1
2
h
2
(t, x) +e([x[)(t, x), x IR
3
. (19.3)
The angular momentum is a conserved quantity which corresponds to an invariance of the Lagrangian
(2.6) with respect to the regular representation of the rotation group SO(3). The regular rep-
resentation R
g
acts on the phase space c := L
2
(IR
3
) by the formula R
g
(x) := (gx), x IR
3
, for
g SO(3), c. The Schrodinger operator H commutes with the representation and with the
generators H
k
of the rotations around the axis Ox
k
(see (8.5)). The generators automatically satisfy
the commutation relations (9.2) of the Lie algebra of the rotation group SO(3). The corresponding
conserved quantities are given by the Noether theorem and have the form (6.13):
L
k
= ,

L
k
), k = 1, 2, 3, (19.4)
where

L
k
= h
H
k
. An alternative proof of the conservation follows from the commutation [H, H
k
] = 0
(see (9.2)).
Spin angular momentum
The analysis suggests that the spin angular momentum corresponds in a similar way to another action
of the rotation group, which is dierent from the regular representation. To construct this spinor
representation, we denote its generators by h
k
. Then similarly to (9.2), we necessarily have
[h
1
, h
2
] = ih
3
, [h
2
, h
3
] = ih
1
, [h
3
, h
1
] = ih
2
. (19.5)
106
Further, dene s
k
= h
h
k
and the spin angular momenta
s
k
= , s
k
), k = 1, 2, 3. (19.6)
The momenta are conserved quantities if the modied Schrodinger operator commutes with h
k
.
The double splitting of the ground state in the Stern-Gerlach experiment with the hydrogen atom
suggests that the dimension of the unperturbed ground eigenspace is two. Therefore, in analogy with
the orbital angular momentum, the ground state formally corresponds to 2l + 1 = 2, i.e. l = 1/2,
and, for B ,= 0, it generates two stationary states which are eigenfunctions of the operator h
3
with the
eigenvalues 1/2. Therefore, the action of the group SO(3) in the unperturbed ground eigenspace is
an irreducible two-dimensional representation S
g
.
Such a two-dimensional representation is given by Proposition 9.8: it corresponds to the generators
h
k
:=
1
2
k
, where
k
are the Pauli matrices (see (9.11))
1
=
_
0 1
1 0
_
,
2
=
_
0 i
i 0
_
,
3
=
_
1 0
0 1
_
. (19.7)
Exercise 19.1 Check the commutation relations (19.5) for the generators h
k
=
1
2
k
.
The simplest way to make the ground eigenspace two-dimensional is to dene the modied phase
space as the tensor product c
:= c C
2
and consider the tensor product of the regular representation
T
g
in c and the spinor representation S
g
in C
2
.
Remark 19.2 By the denition of the tensor product of representations, all generators H
k
commute
with all generators h
j
.
19.3 Pauli Equation. Uniform Magnetic Field
Let us summarize our discussion and change the Schrodinger equation (18.6) to the Pauli equation
(1.14),
ih
t
(t, x) = T(t, x)
:=
1
2
h
2
(t, x) +e([x[)(t, x)
e
2c
LB(t, x)
e
c
sB(t, x), (19.8)
where (t, x) = (
1
(t, x),
2
(t, x)) c
= c C
2
and the additional factor 2 in the last term
corresponds to (19.2).
Denition 19.3 i) For the Pauli Equation, the total angular momentum is the following mean value:
J = L +s. (19.9)
ii) The total magnetic moment is
m =
e
2c
L +
e
c
s. (19.10)
The unperturbed Pauli operator T with B = 0 commutes with all T
g
and S
g
. Hence, the orbital
and spin momenta, L and s, are conserved quantities if B = 0. Therefore, J is also conserved if B = 0.
Remark 19.4 For the Pauli equation with B = 0 we have three conserved angular momenta L, s and
J. Hence, the identication of the total angular momentum with the vector J, is not well justied.
19. PAULI EQUATION 107
For B = (0, 0, B) ,= 0, the equation becomes
ih
t
(t, x) = T(t, x)
=
1
2
h
2
(t, x) +e([x[)(t, x)
e
2c
L
3
B(t, x)
e
c
s
3
B(t, x). (19.11)
Remark 19.5 The operator T commutes with

L
2
:=

L
2
1
+

L
2
2
+

L
2
3
,

L
3
, s
2
:= s
2
1
+ s
2
2
+ s
2
3
, s
3
and
J
2
:=

J
2
1
+

J
2
2
+

J
2
3
,

J
3
. Therefore, the corresponding mean values L
2
, L
3
, s
2
= 3/4, s
3
and J
2
,
J
3
= L
3
+s
3
, are conserved.
19.4 Pauli Equation. General Maxwell Field
The natural extension of the Pauli equation for a general Maxwell eld reads
[ih
t
e(t, x)](t, x) =
1
2
[ih
e
c
A(t, x)]
2
(t, x)
e
c
sB(t, x)(t, x). (19.12)
It corresponds to the Lagrangian density (cf. (2.6))
L
P
(x, , ) = [ih
0
e(x)]
1
2
[[ih
e
c
A(x)][
2
e
c
[sB(x)] , (19.13)
where B(x) = rot A(x) and stands for the scalar product in C
2
IR
4
. This suggests the following
extension to the Lagrangian density of the coupled Maxwell-Pauli equations (cf. (7.1))
L
MP
= [ih
0
e(x)]
1
2
[[ih
e
c
A(x)][
2
e
c
[sB(x)]
1
16
T
(19.14)
where T
:=
and T
:=
. The corresponding Euler-Lagrange equations

are the coupled Maxwell-Pauli equations (cf. (7.2))
_
_
[ih
t
e(t, x)](t, x) =
1
2
[ih
e
c
A(t, x)]
2
(t, x)
e
c
sB(t, x)(t, x).
1
4

(t, x)=
_
_
_
_
:= e[(t, x)[
2
j
c
:=
e
c
[ih
e
c
A(t, x)](t, x)(t, x)+
e
c
rot
_
[s(t, x)](t, x)
_
_
_
_
_
(19.15)
19.5 Application to the Stern-Gerlach Experiment
Theorem 16.2 allows us to construct quantum stationary states corresponding to the Pauli equation
(19.11) describing the hydrogen atom. Let us dene the vector-functions
+
lmn
=
_

lmn
0
_
,
lmn
=
_
0
lmn
_
, n = 1, 2, 3..., l n 1, m = l, ..., l, (19.16)
where the functions
lmn
are given by (8.34). They are eigenfunctions of the operator T corresponding
to the energies
E
mn
:= 2h
R/n
2
+mh
L
h
L
. (19.17)
In particular, the Schrodinger ground state
001
generates two stationary states
001
with distinct
energies E
01
:= 2h
R h
L
and distinct spin magnetic moments
eh
2c
. This explains the double
splitting of the beam in the Stern-Gerlach experiment.
108
Remark 19.6 The double splitting of spectral lines is observed experimentally even in the absence
of the external magnetic eld, when formally
L
:= eB/(2c) = 0. This means that empirically
the Larmor frequency
L
,= 0 even for B = 0. It is natural to think that it is produced by the intrinsic
magnetic eld generated by the electric current in the coupled Maxwell-Pauli equations (19.15). Let
us recall that the formulas (19.16), (19.17) are obtained in the Born approximation neglecting the
intrinsic eld. So the double splitting in the absence of the external magnetic eld might be explained
by the nonlinear self-interaction in the coupled Maxwell-Pauli equations. An alternative mechanism
of generation of the magnetic eld has been proposed by Thomas [97] and Frenkel [34] (see below).
Comment 19.7 The classical interpretation of the splitting term h
L
in (19.17) means that the
projection of the spin magnetic moment to the axis Ox
3
is equal to
e
2c
. The magnitude of
the projection is attained instantly with the magnetic eld, which contradicts the classical picture
(nonzero moment of inertia, etc). In the quantum context the instant reaction is not surprising in
view of the conjecture A (see Preface). Namely, it should be interpreted as the change of the attractor
of the system, and the formulas (19.17) determine the instant bifurcation of the (point) attractor.
20. EINSTEIN-DE HAAS AND ANOMALOUS ZEEMANN EFFECTS 109
20 Einstein-de Haas and Anomalous Zeemann Eects
We consider a modication of the Pauli equation taking into account the interaction between the
orbital and spin angular momentum. The Russell-Saunders method gives a satisfactory explanation
of the Einstein-de Haas and anomalous Zeemann eects.
The Pauli equation in the form (19.8) is not sucient for the explanation of the anomalous Zeemann
eect. This follows from the formulas (19.17), (19.17). Namely, the selection rules m m, m 1
give the splitting
0

0
,
0
L
of the unperturbed spectral line
0
. Hence, the splitting gives the
triplet corresponding to the normal Zeemann eect (cf. Remark 16.4).
This situation is related to the fact that the Pauli equation (19.8) does not take into account the
interaction between the orbital and spin angular momenta (see Remark 19.4), while the interaction is
suggested by the great success of the phenomenological vector model (see Exercise 15). That model
suggests the interaction through the Maxwell eld. Namely, the orbital angular momentum is related
to the rotational motion of the electron, i.e., the circular current which generates the magnetic eld
acting on the spin magntic moment. Similarly, the spin angular momentum is related to the intrinsic
rotation of the electron which generates the magnetic eld acting on the orbital motion of the electron.
20.1 Spin-Orbital Coupling
To explain the anomalous Zeemann eect, we have to modify the equation (19.8) taking into account
the interaction between the orbital and spin angular momenta. In spirit, the interaction might be
described by the coupled Maxwell-Pauli equations (19.15) (see Remark 19.6). An alternative rela-
tivistic modication has been found by Thomas [97] and Frenkel [34]. Namely, the modication takes
into account the magnetic eld arising in the moving frame of the electron. The magnetic eld is
expressed in the electrostatic radial potential ([x[) by the Lorentz formulas (23.17): in the rst order
approximation w.r.t. = v/c, we get
B =
1
c
Ev =
1
c
([x[) v =
1
c
([x[)
x
[x[
p =
1
c[x[
([x[)
L (20.1)
where we set formally v = p/. This magnetic eld produces the corresponding correction term
e
c
s
B(t, x) in the RHS of (19.8). Furthermore, the Thomas and Frenkel phenomenological argu-
ments imply an additional factor 1/2 which is justied by the relativistic Dirac theory [82]. Hence,
nally, the modied equation reads
ih
t
(t, x)= T
m
(t, x)
(20.2)
:=
1
2
h
2
(t, x) +e([x[)(t, x) +
e
2
2
c
2
([x[)
[x[
Ls(t, x)
e
2c
LB(t, x)
e
c
sB(t, x).
In the case B = 0 the equation becomes
ih
t
(t, x)= T
m
(t, x)
(20.3)
:=
1
2
h
2
(t, x) +e([x[)(t, x) +
e
2
2
c
2
([x[)
[x[
Ls(t, x).
Remarks 20.1 i) The correction term

Ls is called the Russell-Saunders spin-orbital cou-
pling.
110
ii) A rigorous justication for the correction term follows from the Dirac relativistic equation.
Namely, just this term appears in an asymptotic expansion of the Dirac equation in the series of
c
1
(see [89, Vol. II] and [9, 84]).
iii) The term is a genuine relativistic correction since it is provided by the Lorentz transformation
(23.17) of the Maxwell eld.
We will see that the stationary states and the energies for the equation (20.2) depend on the azimuthal
quantum number (in contrast to (19.16), (19.17)) and describe perfectly the anomalous Zeemann eect
as well as the Einstein-de Haas experiment. First let us assume that B = (0, 0, B) and rewrite (20.2)
in the form
ih
t
(t, x) = T
m
(t, x) m
3
B(t, x), (20.4)
where m =
e
2c
L +
e
c
s in accordance with the denition (19.10). We will apply the perturbation
theory to calculate the spectrum of the equation (20.4) for small [B[. More precisely, we will calculate
the corrections to the eigenvalues of the unperturbed equation (20.4) with B = 0. This is sucient
since just these corrections describe the splitting of the spectral lines in the magnetic eld.
We will also calculate the Lande factor for quantum stationary states of the equation (20.4). The
factor explains the Einstein-de Haas experiment and anomalous Zeemann eect. By Denition (19.10),
the magnetic moment for a state is dened as
m = , m). (20.5)
The Lande factor g for a state is dened by
m
3
J
3
= g
e
2c
, (20.6)
and this factor (or rather its maximal value) is measured in the Einstein-de Haas experiment.
20.2 Russell-Saunders Quantum Numbers
The main point of the Russell-Saunders method is the analysis of the symmetry properties of the
problem which correspond to the correction term

Ls. Namely, the operators T
m
,

L
2
,
J
2
,
J
3
:=

L
3
+s
3
commute with each other. On the other hand, the momenta

L
3
and s
3
do not commute with

Ls
and hence with the generator T
m
. Hence, we cannot use the classication of type (19.16) for the
stationary states by the quantum numbers n, l, m, 1 which correspond to the eigenvalues of the
operators T,

L
2
,

L
3
, s
3
. Let us choose dierent quantum numbers.
Exercise 20.2 Check that the operators T
m
,

L
2
,
J
2
,
J
3
commute with each other. Hints:
i) (20.2) implies that
T
m
=
1
2
h
2
+e([x[) +
e
2
2
c
2
([x[)
[x[
Ls
e
2c
[
L
3
+ 2s
3
]B. (20.7)
ii)

Ls commutes with

L
2
since each

L
n
and s
n
commutes with

L
2
.
iii)

J
2
commutes with

L
2
since

J
2
=

L
2
+ 2
Ls +s
2
.
iv)

Ls commutes with

J
2
and

J
3
since 2
Ls =

J
2

L
2
s
2
.
Therefore, we could expect that the operators in the space c
= c C
2
can be simultaneously
diagonalized. Let us recall that possible eigenvalues of the operators are, respectively, E, h
2
L(L +
1), h
2
J(J + 1) and h
M, where J = 0,
1
2
, 1,
3
2
, ... and M = J, ..., J (see Sections 10.2 and 10.3). Let
us introduce Diracs notation.
Denition 20.3 [E, L, J, M) denotes an eigenvector which corresponds to the eigenvalues
E, h
2
L(L + 1), h
2
J(J + 1), h
M, of the operators T,

L
2
,
J
2
,
J
3
, respectively.
20.3 Lande Formula
Theorem 20.4 The Lande factor g = g() corresponding to the stationary state = [E, L, J, M) is
given by
g =
3
2
+
3/4 L(L + 1)
2J(J + 1)
. (20.8)
Proof The vector belongs to one of the irreducible subspaces invariant with respect to the operators
J
1
,
J
2
,
J
3
. Indeed,

J
2
= h
2
J(J+1) and

J
3
= h
M. Therefore is an element e
M
of the canonical
basis e
J
, ..., e
J
by Proposition 9.8, and all basis vectors e
M
are obtained from e
M
by an application
of the operators

J
:=

J
1
i
J
2
.
For a linear operator A in c
, let us denote the matrix element A

M,M
= Ae
M
, e
M
). Then (19.9)
and (19.10) imply that
m
3
= m
M,M
3
=
e
2c
(
J
M,M
3
+s
M,M
3
) =
e
2c
(h
M +s
M,M
3
). (20.9)
It remains to nd the matrix element s
M,M
3
and calculate g from the denition (20.6). Let us collect
the commutators of the operators:
_
_
[
J
1
, s
1
] = 0, [
J
1
, s
2
] = ih
s
3
, [
J
1
, s
3
] = ih
s
2
,
[
J
2
, s
2
] = 0, [
J
2
, s
3
] = ih
s
1
, [
J
2
, s
1
] = ih
s
3
,
[
J
3
, s
3
] = 0, [
J
3
, s
1
] = ih
s
2
, [
J
3
, s
2
] = ih
s
1
,
(20.10)
where the second and the third line follow from the rst one by cyclic permutations. Let us use the
commutators (20.10) to calculate the commutators of the operators

J
with s
:= s
1
is
2
. We nd
[
, s
+
] = 2h
s
3
, [
J
+
, s
+
] = 0. (20.11)
The rst formula implies the identity
J
M,M+1
s
M+1,M
+
s
M,M1
+

J
M1,M
= 2h
s
M,M
3
, (20.12)
since all other matrix elements

J
M
,M
= 0 for M
,= M
+1 by (9.10). The non-zero matrix elements

of

J
are known from (9.10) and (9.7):
J
M+1,M
+
= h
_
(J M)(J +M + 1),

J
M1,M
= h
_
(J +M)(J M + 1). (20.13)
On the other hand, s
+
can be derived from the second identity (20.11). Taking the matrix element
()
M+1,M1
, we derive
J
M+1,M
+
s
M,M1
+
s
M+1,M
+

J
M,M1
+
= 0. (20.14)
Then (20.13) implies that
s
M+1,M
+
_
(J M)(J +M + 1)
=
s
M,M1
+
_
(J M + 1)(J +M)
=: A. (20.15)
Substituting this into (20.12) and using the matrix elements of

J
from (20.13), we get

s
M,M
3
= AM. (20.16)
It remains to calculate the constant A. We start with the identity
J
2
= (
L +s)
2
=

L
2
+ 2
Ls +s
2
=

L
2
+ 2
Js s
2
. (20.17)
112
Hence
(
Js)
M,M
= h
2
J(J + 1) L(L + 1) + 3/4
2
(20.18)
since s
2
= 3/4. On the other hand, the same matrix element can be expressed with the help of another
identity
2
Js = s
+
+s
J
+
+

J
3
s
3
(20.19)
as
(
Js)
M,M
=
1
2
s
M,M1
+

J
M1,M
+
1
2
s
M,M+1

J
M+1,M
+
+h
Ms
M,M
3
. (20.20)
Note that the matrix elements s
M+1,M
+
and s
M,M+1
are complex conjugate. Therefore, we have by

(20.15) that
s
M+1,M
+
= A
_
(J M)(J +M + 1) = s
M,M+1
, (20.21)
since the constant A is real by (20.16). Finally, let us substitute (20.21), (20.18), (20.16) and (20.13)
into (20.15). Then we get an equation for A which implies that
A = h
J(J + 1) L(L + 1) + 3/4

2J(J + 1)
. (20.22)
At last, (20.16) and (20.9) give that
m
3
=
e
2c
(h
M +AM) =
e
2c
(1 +A/h
)h
M =
e
2c
(1 +A/h
)J
3
. (20.23)
Substituting (20.22) for A, we get (20.6), where g coincides with (20.8).
Remarks 20.5 i) Our proof follows the calculations in [71].
ii) The formula (20.8) has been obtained by Lande, [70], using the phenomenological vector model and
the Bohr correspondence principle (see also [12] and [89, Vol.I]).
20.4 Application to the Einstein-de Haas and Anomalous Zeemann Eects
The Lande formula (20.8) explains satisfactorily the Einstein-de Haas experiments (see [12] and [89,
Vol.I]). Furthermore, by perturbation theory (see [102]), the Lande factor (20.8) determines the splitting
of the eigenvalues for (20.4) in the magnetic eld (0, 0, B) with small [B[. The resulting correction to
the frequency = E/h
of the eigenstate [E, L, J, M) is equal to
L,J,M
= m
M,M
3
B/h
= m
3
B/h
= g
e
2c
MB = g
L
M (20.24)
up to corrections O(B
2
). This formula, together with the selection rules J J
= J 1, M
M
= M, M 1 (suggested by the Bohr correspondence principle [12] and [89, Vol.I]), explains the
multiplet structure in the anomalous Zeemann eect. Namely, the magnitude of the splitting (20.24),
g
e
2c
h
B, now depends on the quantum numbers L, J in contrast to (19.17) and (16.15). This explains
the multiplet structure of the spectrum since (16.17) now becomes
kk
=
0
kk

L
[g(k)M g(k
)M
] =
0
kk

L
g(k)[M M
]
L
[g(k) g(k
)]M
, (20.25)
where k = (E, L, J, M), k
= (E
, L
, J
, M
) and
0
kk
is the unperturbed spectral line corresponding
to B = 0. Hence, the radiation in the allowed transition does not reduce to the normal triplet (16.17).
The calculations are in excellent agreement with experimental observations (see [9, 12, 17] and [89,
Vol.I]). This agreement was one of the greatest successes of quantum theory.
Remark 20.6 The calculation of the splitting (20.25) does not depend on the magnitude of the un-
perturbed spectral line
0
kk
. The magnitude can be calculated also with the help of the perturbation
theory (see [9, 17] and [89, Vol.II]).
114
Part V
Special Relativity
115
21. ELECTROMAGNETIC NATURE OF LIGHT 117
21 Electromagnetic Nature of Light
We deduce the Maxwell equations from the Coulomb and Biot-Savart-Laplace laws for the interaction
of charges and currents. We also discuss connections of the Maxwell equations to the Einstein ideas
on relativity, in particular, to the Lorentz transformation.
21.1 Maxwell Equations. Empirical Synthesis
Maxwell stated the equations written in the MKS rationalized (or SI) system of units:
_
_
div E(t, x) =
1
0
(t, x), rot E(t, x) =
B(t, x),
div B(t, x) = 0,
1
0
rot B(t, x) = j(t, x) +
0

E(t, x).
(21.1)
The equations contain the dielectric permittivity and magnetic permeability of the vacuum,
0
and
0
,
and do not contain explicitly the speed of light c. Maxwell deduced the equations from the Coulomb
and Biot-Savart-Laplace laws for the interaction of charges and currents. The Coulomb law states the
force of an electrostatic interaction of two charges q
1,2
at a distance r:
F
2
=
1
4
0
q
1
q
2
e
1,2
r
2
, (21.2)
where e
1,2
is the unit vector directed from the charge at q
1
to the charge at q
2
, and F
2
is the force
acting onto the second elementary charge. Similarly, the Biot-Savart-Laplace law states the force of a
magnetic interaction of two (stationary) elementary currents I
k
dl
k
, k = 1, 2, at a distance r:
F
2
=

0
4
I
2
dl
2
(I
1
dl
1
e
1,2
)
r
2
, (21.3)
where F
2
is the force acting onto the second elementary current I
2
dl
2
. The values of
0
and
0
were
measured in a laboratory for electromagnetic experiments with a high accuracy. In the MKS system
0

1
4 9 10
9
As
V m
,
0
410
7
V s
Am
. (21.4)
The rst equation
First, (21.2) implies that the electric eld of the rst elementary charge is
E(x
2
) :=
F
2
q
2
=
1
4
0
q
1
e
1,2
[x
1,2
[
2
, (21.5)
where x
1,2
:= x
2
x
1
and x
1
, x
2
are the vector-positions of the charges. Then for a distribution of
the charges, (x
1
)dx
1
, we obtain by the principle of superposition,
E(x
2
) =
1
4
0
_
(x
1
)e
1,2
[x
1,2
[
2
dx
1
, (21.6)
Dierentiation gives
div E(x
2
) =
1
0
(x
2
), (21.7)
which coincides with the rst equation of (21.1).
Exercise 21.1 Check (21.7) for (x) C
0
(IR
3
). Hint: use that div
x
2
e
1,2
|x
1,2
|
2
=
x
2
1
[x
1,2
[
= 4(x
1,2
).
118
Last equation for stationary currents
Second, (21.3) means, by Amperes law, that the magnetic eld of the rst elementary current is
B(x
2
) =

0
4
I
1
dl
1
e
1,2
[x
1,2
[
2
. (21.8)
Integrating, we obtain by the principle of superposition,
B(x
2
) =

0
4
_
j(x
1
) e
1,2
[x
1,2
[
2
dx
1
, (21.9)
where j(x
1
) is the current density at the point x
1
. Dierentiating, we get
rot B(x
2
) =
0
j(x
2
), (21.10)
which coincides with the last equation of (21.1 in the case of the stationary currents.
Exercise 21.2 Check (21.9) for j(x) C
0
(IR
3
). Hint: use the charge continuity equation div j(x)
0 for the stationary currents.
Last equation for nonstationary currents: Maxwells Displacement Current
The divergence of the LHS of the last Maxwell equation vanishes since div rot = 0. On the other
hand, the divergence of the RHS, div j(t, x) generally does not vanish for nonstationary currents.
To solve this contradiction, Maxwell has completed the last equation by the displacement current
0

E(t, x). The current provides a vanishing divergence of the RHS:
div j(t, x) + div
0

E(t, x) = 0 (21.11)
by the charge continuity equation div j(t, x) + (t, x) = 0 and the rst Maxwell equation.
Faradays equations
The second and third Maxwell equations have been divined and checked experimentally by Faraday.
21.2 Equations for Potentials
Let us repeat the introduction of the Maxwell potentials (see Lecture 4) in the MKS system. Namely,
div B(t, x) = 0 implies that B(t, x) = rot A(t, x). Then rot E(t, x) =
B(
t, x) implies rot [E(t, x) +
A(t, x)] = 0 hence E(t, x) +

A(t, x) =
x
(t, x). Finally,
B(t, x) = rot A(t, x), E(t, x) =
x
(t, x)

A(t, x), (t, x) IR
4
. (21.12)
Exercise 21.3 Prove the existence of the potentials (21.12). Hint: Use the Fourier transform.
The choice of the potentials is not unique since the gauge transformation
(t, x) (t, x) + (t, x), A(t, x) A(t, x)
x
(t, x) (21.13)
does not change the elds E(t, x), B(t, x) for any function (t, x) C
1
(IR
4
). Therefore, it is possible
to satisfy an additional gauge condition. Let us choose for example the Lorentz gauge
0

(t, x) +
1
0
div A(t, x) = 0, (t, x) IR
4
. (21.14)
21. ELECTROMAGNETIC NATURE OF LIGHT 119
Let us express the Maxwell equations (21.1) in terms of the potentials. Substitution of (21.12) into
the rst Maxwell equation leads to
4
0
(t, x) = div E(t, x) = (t, x) div

A(t, x). Eliminating
div

A(t, x) by the dierentiation of (21.14) in time,
0
(t, x) +
1
0
div

A(t, x) = 0, we get
[
0
2
t
](t, x) =
1
0
(t, x), (t, x) IR
4
. (21.15)
Similarly, substituting (21.12) into the last Maxwell equation, we get
1
0
rot rot A(t, x) = j(t, x) +
0

E(t, x) = j(t, x)
0
x

(t, x)
0

A(t, x). (21.16)
Exercise 21.4 Prove the identity
rot rot = +
x
div . (21.17)
Substituting (21.17) to (21.16) and eliminating
x

(t, x) by application of
x
to (21.14), we get
[
0
2
t
]A(t, x) =
0
j(t, x), (t, x) IR
4
. (21.18)
Remark 21.5 The arguments above show that the Maxwell equations (21.1) are equivalent to the
system of two wave equations (21.15), (21.18) for the potentials with the Lorentz gauge condition
(21.14).
Maxwell deduced the wave equations (21.15) and (21.18) with the coecient
0
0
and at that time it
was known that this coecient is equal to 1/v
2
, where v is the propagation velocity of the solutions.
He calculated v = 1/
0
using the values (21.4). The values (21.4) give v 3 10
8
m/s which almost
coincides with the speed of light c. Hence the equations (21.15) and (21.18) become
2(t, x) =
1
0
(t, x), 2A(t, x) =
0
j(t, x), (t, x) IR
4
, (21.19)
where 2 :=
1
c
2
2
t
. Furthermore, Maxwell found that the electromagnetic waves are transversal
like the light waves. This is why Maxwell suggested to identify the electromagnetic waves with light.
Below we will use Gaussian units and write (21.19) as
2(t, x) = 4(t, x), 2A(t, x) =
4
c
j(t, x), (t, x) IR
4
. (21.20)
21.3 Problem of Luminiferous Ether: Michelson and Morley Experiment
The discovery of Maxwell led to a new very dicult question. The propagation velocity can be equal
to c only in a unique preferred reference frame in which the Maxwell equations have the form
(21.1) or (21.20). The preferred reference frame is called the frame of the luminiferous ether. In all
other frames, the propagation velocity depends on a direction and vector of the relative velocity of
the frame of reference, if space-time is transformed by the classical Galilean transformations
_
_
_
_
_
t
x
1
x
2
x
3
_
_
_
_
_
_
_
_
_
_
t
1
x
2
x
3
_
_
_
_
_
=
_
_
_
_
_
t
x
1
vt
x
2
x
3
_
_
_
_
_
. (21.21)
120
Therefore, the Maxwell equations are not invariant with respect to the Galilean transformations. This
is why Michelson and Morley started around 1880 the famous experiment to identify the preferred
luminiferous ether frame with the frame in which the Sun is at rest. They have tried to check that
the Earth moves with respect to the luminiferous ether, i.e. the velocity of light along and against
the velocity of the Earth diers by twice the velocity of the Earth. Concretely, they compared the
wavelengths of light along and against the velocity of the Earth motion around the Sun. However,
the result (1887) was negative and very discouraging: the wave lengths were identical with a high
accuracy, hence the propagation velocity does not depend on the frame of reference! Astronomical
observations of double stars by de Sitter (1908) conrmed the negative result of Michelson and Morley.
Also the experiment of Trouton and Noble conrmed the negative result.
21.4 Time in a Moving Frame: Lorentz transformations
Various partial explanations of the negative results were proposed by Ritz, Fitzgerald, Lorentz and
others. The complete explanation was provided in 1905 by Einstein, who was able to cumulate the
Maxwell and Lorentz ideas into a new complete theory. The main novelty was the following postulate
of the Einstein theory:
The time in a moving frame is distinct from the time in the rest frame!
We will prove that the transformation of space-time coordinates from the rest frame to the moving
frame of reference is given by the Lorentz formulas
_
_
_
_
_
_
_
_
ct
x
1
x
2
x
3
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
ct
1
x
2
x
3
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
ctx
1
1
2
x
1
ct
1
2
x
2
x
3
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
1
1
2
1
2
0 0
1
2
1
1
2
0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
ct
x
1
x
2
x
3
_
_
_
_
_
_
_
_
(21.22)
where (t, x
1
, x
2
, x
3
) stands for the time-space coordinates in the rest frame and (t
, x
1
, x
2
, x
3
) corre-
sponds to the moving frame if the relative velocity is (v, 0, 0), [v[ < c and := v/c .
Exercise 21.6 Check that the wave equation
[
1
c
2

2
t
](t, x) = f(t, x), (t, x) IR
4
(21.23)
is invariant with respect to the transformations (21.22).
Hints: i) Set c = 1 and use that
2
=
2
,
3
=
3
. ii) Check that the 1D equation (
2
t

2
1
)(t, x) =
f(t, x), (t, x) IR
2
is equivalent to [(
t
)
2
(
1
)
2
)]
(t
, x
) = f
(t
, x
) if t
= at bx and x
= ax bt
with a
2
b
2
= 1. By denition,
(t
, x
) = (t, x) and f
(t
, x
) = f(t, x).
Exercise 21.7 Check that the wave equation (21.23) is not invariant with respect to the Galilean
transformation (21.21). Hint: In the new variables, the equation (21.23) becomes
[
1
c
2
(
t
v
x
)
2
x
]
(t
, x
) = f
(t
, x
), (t
, x
) IR
4
. (21.24)
Remark 21.8 For small velocities, [v[ c, the Galilean transformation (21.21) is close to the Lorentz
one, (21.22), and the coecients of the equation (21.24) are close to the ones of (21.23).
22. THE EINSTEIN SPECIAL RELATIVITY AND LORENTZ GROUP 121
22 The Einstein Special Relativity and Lorentz Group
The formulas (21.22) dene a particular transformation of the Lorentz group. We dene the Lorentz
group as transformations of the space-time preserving the form of the wave equation. This property
follows from the Einsteins postulate of Special Relativity Theory extending the Galileis principle of
relativity from mechanics to electrodynamics. Namely, the main point of Einsteins special relativity
is the following postulate:
E: Maxwell Equations have an identical form in all Inertial Frames
The postulate implies also an invariance of the wave equation (21.23) in the case f(x) = 0. Let us
calculate all possible transformations which leave the equation with f(x) ,= 0 invariant.
First, introduce a new time variable x
0
:= ct and rewrite the wave equation (21.23) in the form
2(x) := g
(x) = f(x), x IR
4
, (22.1)
where g
,
= diag(1, 1, 1, 1),
:=

x
and the summation in the repeated indexes is under-

stood, , = 0, 1, 2, 3.
We look for maps : IR
4
IR
4
such that the equation keeps its form in new variables x
= x,
i.e., (22.1) is equivalent to
2
(x
) = g
(x
) = f
(x
), x
IR
4
, (22.2)
where, by denition,
(x
) := (x) and f
(x
) := f(x). The map , naturally, i) is continuous and

moreover, ii) (x +x) = x + x, x IR
4
, which expresses the fact that space-time is homogeneous.
Let us assume, additionally, the normalization condition iii) (0) = 0. The three conditions imply
necessarily that is a linear transformation of the space IR
4
dened by a matrix:
x
= (x)
= 0, ..., 3. (22.3)
Exercise 22.1 Check that the conditions i), ii), iii), imply (22.3). Hint: First, check the correspond-
ing one-dimensional statement with x IR.
Finally, rewrite the equation (22.1) in new variables (22.3). By the chain rule
=
x
. (22.4)
Hence, the equation in the new variables becomes
g
,
(x
) = f
(x
), x
IR
4
. (22.5)
Comparing with (22.2), we get the system of algebraic equations
g
= g
= 0, 1, 2, 3. (22.6)
In matrix form, the system is equivalent to the equation
g
t
= g, (22.7)
where
t
stands for the transposed matrix of .
Exercise 22.2 Check that [det [ = 1. Hint: take the determinant of both sides of (22.7).
122
Hence is invertible and (22.7) is also equivalent to
(
t
)
1
g
1
= g, (22.8)
The matrix equation is equivalent to the invariance, with respect to the map
1
, of the quadratic
form
g(x, x) := (x, gx) = g
, x IR
4
(22.9)
which is called the Lorentz interval. Then this form is invariant also with respect to the map ,
hence
t
g = g, (22.10)
or equivalently,
g
= g
= 0, 1, 2, 3. (22.11)
Denition 22.3 L is the set of all linear maps : IR
4
IR
4
satisfying (22.10).
Exercise 22.4 Check that the set L is a group.
Denition 22.5 i) Minkowski space is the space IR
4
endowed with the quadratic form (22.9).
ii) L is called the Lorentz group.
Example 22.6 The simplest example of a Lorentz transform is given by the matrices
=

R :=
_
1 0
0 R
_
, (22.12)
where R SO(3) is a rotation of the 3D space.
Remark 22.7 The wave equation (21.23) is invariant with respect to the transformation x
=

Rx
since the Laplacian is invariant with respect to the rotations.
Exercise 22.8 Check (22.7) for the matrix (22.12).
Exercise 22.9 Check that the map R

R is an isomorphism of the rotation group SO(3) onto a
subgroup of the Lorentz group L.
Exercise 22.10 Construct all Lorentz transformations of the form
=
_
_
_
_
_
a b 0 0
c d 0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
, (22.13)
Solution: (22.7) is equivalent to the matrix equation
_
a c
b d
__
1 0
0 1
__
a b
c d
_
=
_
1 0
0 1
_
(22.14)
which is equivalent to a system a
2
c
2
= 1, b
2
d
2
= 1, abcd = 0. Then a = cosh , c = sinh ,
d = cosh , b = sinh, and cosh sinh sinh cosh = 0. Therefore, tanh = tanh or
= . Finally, we get four one-parametric families of the matrices
:=
_
_
_
_
_
cosh sinh 0 0
sinh cosh 0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
,
:=
_
_
_
_
_
cosh sinh 0 0
sinh cosh 0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
(22.15)
which are hyperbolic rotations (or boosts) in the angle in the plane x
0
, x
1
.
22. THE EINSTEIN SPECIAL RELATIVITY AND LORENTZ GROUP 123
Remark 22.11 The Lorentz formulas (21.22) correspond to the matrix
+
+
from (22.15) with cosh =
1/
_
1
2
and sinh = /
_
1
2
, so tanh = .
124
23 Covariant Electrodynamics
Here we complete a justication of the Einstein postulate E. Namely, we still have to determine the
transformation of the Maxwell Field by the Lorentz group.
Let us introduce the 4D elds and currents
/(x) = ((t, x), A(x)), (x) = ((x),
1
c
j(x)), x IR
4
. (23.1)
In this notation, the Maxwell equations (21.20) become
2/(x) = 4(x), x IR
4
. (23.2)
Hence, the transformations for the potentials /
(x) and the currents
(x) must be identical since

the wave operator 2 is invariant with respect to the Lorentz group. We will prove below that the
convective currents are transformed, like the 4-vector x, by the matrix . Hence, the same holds for
the potentials and all currents, so
/
(x
) = /(x),
(x
) = (x), x
= x. (23.3)
The Lagrangian approach gives another conrmation for (23.3). Namely, in Lecture 5 we have shown
that the Maxwell equations are the canonical Euler-Lagrange equations corresponding to the La-
grangian density
L(x, /, /) =
1
16
T
g(, /), (x, /, /) IR

4
IR
4
IR
16
. (23.4)
where T
:=
, T
:= g
and
g(, /) := g
. (23.5)
The Einstein postulate will hold if the density is invariant under the corresponding transformations
for the elds. It is easy to check that the density is invariant if the potentials /
(x) and the currents
(x) are transformed by (23.3).

Exercise 23.1 Check that the density (23.4) is invariant under the transformations (23.3). Hint:
The bilinear form (23.5) is symmetric and invariant since the corresponding quadratic form is invari-
ant.
At last, let us check that the transformation law (23.3) holds for the convective currents. It suces
to consider only point charges since a general convective current is a combination of moving point
charges. Let x = x(x
0
) be a C
1
trajectory of a point charge, so its 4-current is given by
0
(x) = e(x x(x
0
)),
k
(x) = ev
k
(x
0
)(x x(x
0
)), (23.6)
where v
k
(x
0
) :=
dx
k
(x
0
)
dx
0
.
Remark 23.2 The expressions (23.6) for the convective current are equivalent to the postulating of the
charge-invariance under the motion which is referred traditionally as to an experimental fact (charge
invariance, [46, 53]).
Proposition 23.3 The function (23.6) is Lorentz-covariant, i.e. for every Lorentz transformation
L,
(x
) = (x), x
= x. (23.7)
23. COVARIANT ELECTRODYNAMICS 125
Proof Since (x) is a distribution, (23.7) means that
(x
),
(x
)) = (x), (x)) (23.8)

for any test function (x) C
0
(IR
4
), where
(x
) := (x), x
= x. Here we have used that

[det [ = 1 by (22.7). Since (x) is a measure, it suces to check (23.8) for characteristic functions
(x) of any open bounded set IR
4
. Then (23.8) becomes,
_
(x
)d
4
x
=
_
(x)d
4
x. (23.9)
Let us assume that the set x : x = x(x
0
) is an interval of the trajectory (x
0
, x(x
0
)) : a < x
0
<
b. Then integrating rst in dx, we get
_
(x)d
4
x = e
_
b
a
(1, v(x
0
))dx
0
= e(x
0
, x(x
0
))
b
a
. (23.10)
It implies (23.9) since (x
0
, x(x
0
)) is an invariant 4-vector, i.e. (x
0
, x
(x
0
)) := (x
0
, x(x
0
)).
Remarks 23.4 i) The charge invariance (23.6) implies the transformation (23.7) for the convective
current. For all other currents, the transformation is postulated.
ii) The density (1, v(x
0
)) is not an invariant 4-vector. On the other hand, the dierential form
(1, v(x
0
))dx
0
is Lorentz-invariant which follows from the integration (23.10).
Let us note that the form (1, v(x
0
))dx
0
admits the factorization
(1, v(x
0
))dx
0
=
(1, v(x
0
))
_
1 v
2
(x
0
)
dx
0
_
1 v
2
(x
0
), (23.11)
where the fraction in RHS is an invariant 4-vector while the remaining factor is an invariant dierential
form. Indeed,
w(x
0
) :=
(1, v(x
0
))
_
1 v
2
(x
0
)
=
d(x
0
, x(x
0
))
dx
0
_
1 v
2
(x
0
)
=
d(x
0
, x(x
0
))
_
(dx
0
)
2
(dx(x
0
))
2
(23.12)
is an invariant 4-vector since the Lorentz interval (dx
0
)
2
(dx(x
0
))
2
is Lorentz-invariant. Respectively,
dx
0
_
1 v
2
(x
0
) =
_
(dx
0
)
2
(dx(x
0
))
2
is a Lorentz-invariant dierential form.
Denition 23.5 i) The vector w(x
0
) from (23.12) is the 4-dimensional velocity of a particle.
ii) The function (b) :=
_
b
a
_
(dx
0
)
2
(dx(x
0
))
2
=
_
b
a
_
1 v
2
(x
0
)dx
0
is the proper time of the
particle along the trajectory.
Now (23.11) can be written as
(1, v(x
0
))dx
0
= w(x
0
)d(x
0
). (23.13)
Corollary 23.6 The formulas (23.3) imply the corresponding transformation for the Maxwell tensor
(4.14):
T
(x) =
(x), x
. (23.14)
In the matrix form
T
(x) = T(x)
t
. (23.15)
126
Example 23.7 Applying this formula to the Lorentz bust (21.22), we get the Lorentz transfromation
for the Maxwell elds :
_
_
E
1
(x
) = E
1
(x), B
1
(x
) = B
1
(x),
E
2
(x
) =
E
2
(x) B
3
(x)
_
1
2
, B
2
(x
) =
B
2
(x) +E
3
(x)
_
1
2
,
E
3
(x
) =
E
3
(x) +B
2
(x)
_
1
2
, B
3
(x
) =
B
3
(x) E
2
(x)
_
1
2
.
(23.16)
Equivalently,
E
(x
) =
E(x) +
v
c
B(x)
_
1
2
, B
(x
) =
B(x)
v
c
E(x)
_
1
2
. (23.17)
Exercise 23.8 Check the formulas (23.16). Hints: The formulas follow from (4.15) by (23.15) with
the Lorentz matrix (21.22).
Part VI
Relativistic Dirac Theory
127
24. RELATIVISTIC EQUATION FOR ELECTRON FIELD 129
24 Relativistic Equation for Electron Field
The Special Relativity of Einstein has to be extended to the quantum mechanical theory, as well.
We introduce the Dirac equation which is the relativistic covariant generalization of the Schrodinger
equation.
The Schrodinger equation obviously is not invariant with respect to the Lorentz group. One
possible choice is the Klein-Gordon equation
h
2
c
2
2
t
(x) = h
2
(x)
2
c
2
(x), x IR
4
, (24.1)
which is Lorentz-invariant as well as the wave equation (21.23). However, it leads to negative energies
which is not satisfactory from a physical point of view. Namely, the Klein-Gordon equation is obtained
from the relativistic energy-momentum relation
E
2
c
2
= p
2
+
2
c
2
(24.2)
by the Schrodinger identication E ih
t
, p ih
x
. In the Fourier transform,

(x
0
, p) :=
_
e
i
px
h
(x
0
, x)dx, the Klein-Gordon equation becomes
h
2

2
(x
0
)
2
(x
0
, p) = p
2

(x
0
, p)
2
c
2

(x
0
, p), x
0
IR, p IR
3
. (24.3)
Then the solutions are linear combinations of e
i
E
h
t
where E/c =
_
p
2
+
2
c
2
coincides with (24.2).
The solutions with E/c =
_
p
2
+
2
c
2
seem to correspond to negative energies unbounded from
below, hence the physical interpretation of the Klein-Gordon equation requires an additional analysis.
This is why Dirac tried to nd a relativistic invariant equations of the rst order in time, like
the Schrodinger equation, to avoid the negative roots. Let us follow Diracs arguments. First, the
relativistic invariance requires then the rst order in space. Second, we will look for a linear equation
of the form
ih
(x) = c(x), x IR
4
. (24.4)
Then the corresponding energy-momentum relation can be written as
(p) = c, p IR
4
, (24.5)
where p
:= (E/c, p) and (p) is a linear form:

(p) =
, p IR
4
. (24.6)
The third condition is a correspondence principle: the equation (24.4) must imply the Klein-Gordon
equation (24.1). Namely, applying the operator
ih
to both sides, we get

[
ih
]
2
(x) =
2
c
2
(x), x IR
4
. (24.7)
Then the correspondence principle is equivalent to the algebraic identity
2
(p) = p
2
0
p
2
, p = (p
0
, p) IR
4
(24.8)
since (24.2) can be written as p
2
0
p
2
=
2
c
2
. Diracs extra idea was the choice of the coecients
in the matrix algebra since the scalar coecients do not exist. The existence of the scalar coecients
would mean that the polynomial p
2
0
p
2
is reducible which is not true.
130
Exercise 24.1 Check that (24.8) is impossible with any scalar coecients
.
Theorem 24.2 In 2 2 block form, the matrices
(p) =
_
p
0
p
p p
0
_
(24.9)
satisfy the identity (24.8), where := (
1
,
2
,
3
) are the Pauli spin matrices.
Proof By direct multiplication of 2 2 block matrices, we get
2
(p) =
_
p
2
0
( p)
2
0
0 p
2
0
( p)
2
_
. (24.10)
It remains to use that ( p)
2
= [p[
2
.
Now let us consider the matrices
= (e
), where e
0
= (1, 0, 0, 0), etc. From (24.9), we get
0
=
_
1 0
0 1
_
,
j
=
_
0
j
j
0
_
, j = 1, 2, 3. (24.11)
Proposition 24.3 The matrices
satisfy
_
_
(
0
)
2
= 1, (
j
)
2
= 1, j = 1, 2, 3,
= 0, ,= .
(24.12)
Proof Let us rewrite (24.8) in the form
2
(p) = g(p), p IR
4
, (24.13)
where g(p) := p
2
0
p
2
. It implies
(p)(q) +(q)(p) = 2g(p, q), (24.14)
where g(p, q) = p
0
q
0
pq is the corresponding symmetric bilinear form. In particular, for p = e
and
q = e
= 2g(e
, e
) (24.15)
which implies (24.12).
Remark 24.4 Obviously the matrices (24.11) are not unique solutions to the relations (24.12): for
example, we can replace
by
for certain indexes . We will prove below the Pauli Theorem:

the matrices
are unique up to a change of an orthonormal basis.

Let us rewrite the Dirac equation (24.4) in a Schrodinger form. First, it is equivalent to
ih
0

(t, x) = c(c ih
j

x
j
)(t, x). (24.16)
Second, multiplying by
0
and using that (
0
)
2
= 1, we get
ih

(t, x) = H
D
(t, x) := c
0
(c ih
j

x
j
)(t, x), (24.17)
where the operator H
D
is called the free Dirac Hamiltonian.
25. PROBLEM OF NEGATIVE ENERGIES FOR DIRAC EQUATION 131
25 Problem of Negative Energies for Dirac Equation
We prove the energy conservation for the Dirac equation and check that the energy is not bounded
from below. The Dirac equation (24.17) is a Hamiltonian system with the Hamilton functional
H() := (x), H
D
(x)), (25.1)
in analogy with the Schrodinger equation. It is conserved, i.e.
H((t, )) = const, t IR (25.2)
for the solutions to (24.17).
Exercise 25.1 Check (25.2). Hint: Dierentiate (25.1) and use the Dirac equation (24.17) and the
symmetry of the Dirac operator H
D
.
The quadratic form (25.1) is not positive denite, hence energy conservation (25.2) does not provide
an a priori estimate for the solutions. To see this, it is useful to split each Dirac spinor into a pair of
two-component vectors
(x) =
_

+
(x)
(x)
_
. (25.3)
Dene the Fourier transform
(p) =
_
e
ipx/h
(x)dx, p IR
3
. (25.4)
Then the quadratic form (25.1) becomes by the Parseval identity
H() = (2)
3
(p), c
0
_
c p
p c
_

(p)) = c(2)
3
(p),
_
c p
p c
_

(p))
= c(2)
3
_
c
+
(p),

+
(p)) 2
+
(p), p

(p)) c
(p),

(p))
_
.
In particular,
H
_
+
(x)
0
_
=(2)
3
c
2
+
(p),

+
(p)), H
_
0
(x)
_
=(2)
3
c
2
(p),

(p)) (25.5)
The negative energy might lead to an instability of the Dirac dynamics due to a possible transition of
the solution to the states
_
0
(x)
_
. On the other hand, this instability has never been proved.
Dirac suggested that the transition of all particles is forbidden by the Pauli exclusion principle
since almost all states with negative energy have been occupied long ago. (A more satisfactory solution
of the problem of negative energies, which is applicable also for bosons, is provided by Quantum Field
Theory.) On the other hand, by the Dirac theory, the transitions for certain particles are possible,
and the negative states can be interpreted as the states with positive energy for antiparticles which
are positrons, i.e., electrons with positive charge e. The positrons have been discovered in cosmic
rays by Anderson in 1932.
132
26 Angular Momentum for Dirac Equation
We prove the angular momentum conservation for the Dirac equation. The conservation justies
the Goudsmith-Uhlenbeck conjecture on the electron spin as an intrinsic property of the dynamical
equations.
The conserved orbital momentum for the Schrodinger equation is dened by (6.17) and (6.18):
L = L() := ,

L), where

L = ih
x . For the solutions to the Dirac equation the orbital

momentum generally is not conserved, since the operator

L does not commute with the Dirac operator
H
D
. Hence, for the Dirac equation the denition of the angular momentum requires a modication.
Denition 26.1 For the Dirac equation the angular momentum is J = J() = ,
J) where
J :=

L +
1
2
h
, :=
_
0
0
_
. (26.1)
Theorem 26.2 The angular momentum J is conserved for the solutions to the Dirac equation,
J((t, )) = const, t IR. (26.2)
Proof Dierentiating, we obtain similarly to (6.26),
d
dt
J((t)) =

(t),
J(t)) +(t),
J

(t)) =
i
h
H
D
(t),
J(t)) (t),
J
i
h
H
D
(t))
=
i
h
_
H
D
(t),
J(t)) (t),
JH
D
(t))
_
=
i
h
(t), [H
D
,
J](t)). (26.3)
It suces to check the commutation [H
D
,
J] = 0. First, we know the commutators [
L
k
, p
j
] = ih
kjl
p
l
,
where p
j
:= ih
j
and
kjl
is the totally antisymmetric tensor. Therefore,
[
L
k
, H
D
] = ch
j
[L
k
, p
j
] = ich
kjl
p
l
. (26.4)
This shows that H
D
does not commute with the orbital angular momentum operators

L
k
, hence the
orbital momentum L generally is not conserved. It remains to calculate the commutators [ , H
D
].
First we note that
l
=
_
1 0
0 1
__
0
l
l
0
_
=
_
0
l
l
0
_
.
Hence,
[
k
, H
D
] = [
k
, c
2
0
c
0
l
p
l
] = c
_
0 [
k
,
l
]
[
k
,
l
] 0
_
p
l
.
The commutation relations for the Pauli spin matrices allows us to reduce this to
2ic
klj
_
0
j
j
0
_
p
l
= 2ic
klj
j
p
l
.
Multiplying this by h
/2 and adding (26.4), we get the commutation [H

D
,
J] = 0 by the antisymmetry
of
klj
.
Remarks 26.3 i) Theorem 26.2 means that
1
2
h
represents an intrinsic spinor angular momentum

of the relativistic electron.
ii) We will demonstrate below that the coupling of the Dirac equation to the magnetic eld provides
automatically the correct Lande factor g = 2 for the spinor angular momentum.
27. PAULI THEOREM 133
27 Pauli Theorem
We prove the Pauli theorem on the uniqueness of the Dirac matrices. The theorem states that the
Dirac matrices (24.11) are the unique representations of the relations (24.12), up to equivalence.
Theorem 27.1 Let
, = 0, ..., 3 be operators on a nite-dimensional complex vector space V ,

dimV 4, which satisfy the relations (24.12). Then dimV = 4 and the operators
, for a suitable
choice of a basis, take the matrix form (24.11).
Proof Step i) A key idea of the proof is the following simple characterization of the basis vectors.
Namely, for the standard Dirac matrices (24.11), the matrix
1
2
is diagonal and hence commutes
with
0
which is also diagonal:
0
and
1
2
have the diagonal block matrix form
0
=
_
1 0
0 1
_
,
1
2
=
_
i
3
0
0 i
3
_
. (27.1)
Therefore, the basis vectors e
0
, ..., e
3
are common eigenvectors of the matrices
0
and
1
2
with the
eigenvalues 1 and i, 1 and i, 1 and i, 1 and i respectively.
Step ii) Now apply this observation to the general matrices
from the theorem. The anticommutation

relations from (24.12) imply that the matrices
0
and
1
2
commute with each other:
2
=
1
0
. (27.2)
Exercise 27.2 Check this commutation. Solution:
0
2
=
1
2
=
1
0
.
Hence there exists at least one common eigenvector
1
for both (since V is a complex vector space!):
1
=
1
and
1
1
=
1
(27.3)
where and are suitable complex numbers.
Step iii)
2
= 1 since (
0
)
2
= 1, and similarly
2
= 1 since (
1
2
)
2
= (
1
)
2
(
2
)
2
= 1. Hence,
= 1 and = i. Let us check that all four combinations of the signs are possible for suitable
eigenvectors
1
. Namely,
1
=
3
1
=
3
1
,
and
1
=
3
1
=
3
1
,
hence the vector
3
:=
3
1
is also a common eigenvector with the eigenvalues and . Similarly,
2
:=
3
1
resp.
4
:=
1
1
are common eigenvectors of the operators
0
and
1
2
with the
eigenvalues and resp. and . Since all four possible signs occur, we may permute the four
vectors to ensure that = 1 and = i.
Step iv) On the subspace with basis
1
,
2
,
3
and
4
the operators
0
and
1
2
have the diagonal
block matrix form (27.1). Moreover, in this basis the operators
1
and
3
have the form
1
=
_
0
1
1
0
_
,
3
=
_
0
3
3
0
_
(27.4)
which coincide with (24.11). From these we may check that
2
=
1
(
1
2
) also has the desired form.
Step v) The vectors
j
, j = 1, ..., 4, span the space V since dimV 4.
134
Exercise 27.3 Check the formulae (27.1) for
0
and
1
2
.
Exercise 27.4 Check the formulae (27.4) for
1
and
3
.
Corollary 27.5 For any Lorentz transformation , there exists a nondegenerate matrix ()
GL(4, C) such that
(p) = ()(p)
1
(), p IR
4
. (27.5)
Proof (24.13) implies
2
(p) = g(p) = g(p), p IR
4
, (27.6)
since is a Lorentz transformation. Hence, the matrices (e
) satisfy the relations (24.12) as well

as
:= (e
). Therefore, by the Pauli Theorem, we have

(e
) = ()(e
)
1
(), = 0, ..., 3, (27.7)
where () is an invertible operator in IR
4
(which transforms the vector e
into
, = 0, ..., 3).
Then (27.5) follows by linearity.
28. LORENTZ COVARIANCE 135
28 Lorentz Covariance
We justify the Einstein postulate of Special Relativity Theory for the Dirac equation: the equation
takes an identical form in all Inertial Frames.
Let us consider two frames of reference of two observers related by a Lorentz transformation:
x
= x. By a natural extension of the Einstein postulate E, the Dirac equation (24.4) has the same
form in both frames of reference. This extended postulate allows us to determine the corresponding
transformation of the wave function. The next theorem gives a transformation of the wave function
which leaves the Dirac equation invariant.
Theorem 28.1 Let (x) be a solution to the Dirac equation (24.4) and
(x
) := (
#
)(x), x IR
4
, (28.1)
where x
= x and
#
:= (
t
)
1
where
t
is the transposed matrix to . Then the function
(x
) is
also a solution to the Dirac equation.
Proof Let us translate the Dirac equation (24.4) into the Fourier transform
(p) :=
_
e
ipx
h
(x)dx, p IR
4
, (28.2)
where px := p
. Then (24.4) becomes

(p)
(p) = c
(p), p IR
4
. (28.3)
The Fourier transform translates (28.1) into
(p
) = (
#
)
(p), p IR
4
, (28.4)
where p =
t
p
.
Exercise 28.2 Check (28.4). Hint: formally, we have
(p
) :=
_
e
ip
(x
)dx
=
_
e
ip
x
h
(
#
)(x)[ det [dx = (
#
)
_
e
i
t
p
x
h
(x)dx = (
#
)
(p) (28.5)
since [ det [ = 1 for the Lorentz transformation .
Now express (28.3) in terms of the wave function

(p
):
(p)
1
(
#
)
(p
) = c
1
(
#
)
(p
), p
IR
4
. (28.6)
This is equivalent to the Dirac equation (28.3) i
(
#
)(p)
1
(
#
) = (p
), p
IR
4
. (28.7)
This is equivalent to (27.5) for
#
instead of since p
=
#
p. Finally, it is true since
#
also belongs
to the Lorentz group.
Exercise 28.3 Check that
#
is a Lorentz transformation for any L.
Remarks 28.4 i) The formal calculations (28.5) can be justied by the properties of the Fourier
transformation of tempered distributions. This is necessary since the integrals in (28.5) generally
never converge for the solutions to the Dirac equation by energy and charge conservations (see below).
ii) The theorem implies that the variance rule (28.1) leaves the Dirac equation unchanged. This
means, by denition, that the Dirac equation is covariant with respect to the Lorentz group.
136
29 Lorentz Transformation of Spinors
We give a construction of the operator () acting on the spinor wave functions. For the rotations
the operator will be calculated explicitly.
29.1 Factorization of Lorentz Transformations
First let us derive a useful formula for the transformations of the Dirac matrices.
Lemma 29.1 i) For any two vectors p, q IR
4
with q non-null w.r.t. g (i.e. g(q, q) ,= 0), we have
(q)(p)
1
(q) = (R
q
p), (29.1)
where R
q
is a reection R
q
p = 2
g(q, p)
g(q, q)
q p.
ii) The transformation R
q
is in the Lorenz group but not proper, i.e. det R
q
= 1.
Proof i) Since
2
(q) = g(q, q), we have
1
(q) = (q)/g(q, q). Therefore, multiplying the relation
(q)(p) +(p)(q) = 2g(q, p)
on the right by
1
(q), we obtain
(q)(p)
1
(q) +(p) = 2
g(q, p)
g(q, q)
(q).
This implies (29.1) by the linearity of .
ii) It is easy to check that R
q
is in the Lorentz group, i.e.
g(R
q
p, R
q
p) = g(p, p), p IR
4
. (29.2)
It remains to check that
det R
q
= 1. (29.3)
This is obvious since
i) R
q
preserves the components along q, i.e. R
q
q = q.
ii) R
q
reverses the components g-orthogonal to q, i.e. R
q
p = p if g(q, p) = 0.
Example 29.3 q = (1, 0): then R
q
p = (2p
0
, 0, 0, 0) p = (p
0
, p) for p = (p
0
, p).
Example 29.4 q = (0, q) with q a unit vector: then g(q, q) = [q[
2
= 1 and R
q
p = 2(q p)q p =
(p
0
, 2(qp)qp). Hence the spatial component along q is unchanged while the orthogonal components
are reversed, which is precisely the eect of a rotation through about the axis q.
Now let us factorize the Lorentz transformations in two reections.
Lemma 29.5 Let be a proper Lorentz transformation that xes two linearly independent vectors v
1
and v
2
. Then it admits the factorization
v = R
w
R
u
v, v IR
4
. (29.4)
Here u is a non-null vector (ie. g(u, u) ,= 0), g-orthogonal to v
1
and v
2
, and we set w = u + u if
u ,= u. Otherwise, we take w which is g-orthogonal to all three vectors v
1
, v
2
and u.
29. LORENTZ TRANSFORMATION OF SPINORS 137
Proof First let us note that u can be chosen non-null. Indeed. the plane , g-orthogonal to v
1
and
v
2
, is two-dimensional. Hence, its intersection with the null cone is at most one-dimensional cone
since the null cone does not contain two-dimensional planes. Therefore, all vectors u are non-null
except for at most two lines.
Second, note that
g(u, v
j
) = g(u, v
j
) = g(u, v
j
) = 0. (29.5)
since xes v
j
and preserves g. Hence, u and w = u +u are g-orthogonal to v
1
and v
2
.
Further consider the case u ,= u. Then + 1 is an invertible operator. Namely, has two
eigenvectors v
1
, v
2
with the eigenvalue 1. If it has an eigenvector v
3
with the eigenvalue 1, then also
a vector v
4
g-orthogonal to v
1
, v
2
and v
3
is an eigenvector with the eigenvalue 1. Then also u = u
which contradicts our assumption.
Therefore, ( + 1) is also a two-dimensional plane. Hence, by the same arguments, w = u +u
is a non-null vector for almost all u .
Now let us check the identity (29.4). First, the composition R
w
R
u
xes v
1
and v
2
since each factor
reverses both vectors. Hence, (29.4) holds for v = v
1
and v = v
2
. Further, let us prove (29.4) for
v = u. Namely,
g(w, w) = g(u, u) + 2g(u, u) +g(u, u)
= g(u, u) + 2g(u, u) +g(u, u)
= 2[g(u, u) +g(u, u)] = 2g(u, w).
(29.6)
Therefore,
R
w
u =
2g(u, w)
g(w, w)
w u = w u = u. (29.7)
Then also u = R
w
R
u
u since R
u
u = u.
So (29.4) holds for three linearly independent vectors v
1
, v
2
and u. Hence their action on the whole
space IR
4
is identical since both R
w
R
u
and are proper Lorentz transformations. The case u = u
can be checked similarly.
Exercise 29.6 Check the lemma for the case u = u.
The next lemma demonstrates that each Lorentz transformation admits a factorization into three ones,
each preserving two linearly independent vectors. Let us recall that the rotation group SO(3) is a nat-
ural subgroup of the Lorentz group L: for each R SO(3) the corresponding Lorentz transformation
is given by (22.12).
Lemma 29.7 Let be a proper Lorentz transformation. Then it admits the factorization
=

R
1
B

R
2
, (29.8)
where R
1
, R
2
SO(3) and B is a proper boost of type (22.15).
Proof Let us consider the vector e
0
= (v
0
, v), where e
0
:= (1, 0, 0, 0). There exists a rotation
Q
1
SO(3) such that Q
1
v = (v
1
, 0, 0). Then the vector e
0
:=

Q
1
e
0
has the form (v
0
, v
1
, 0, 0) with
g(e
0
) = v
2
0
v
2
1
= 1 since g(e
0
) = 1 and

Q
1
L.
Exercise 29.8 Check that there exists a proper boost D of type (22.15) such that De
0
= e
0
. Hint:
Choose D in the form
+
+
if v
0
> 0 and
if v
0
< 0.
Now D

Q
1
e
0
= e
0
, hence the Lorentz transformation B

R
2
has a matrix of type
D

Q
1
=
_
1 w
0 Q
2
_
, (29.9)
138
Exercise 29.9 Check that w = 0. Hint: use (22.6) and (22.11) for the matrix D

Q
1
.
Therefore, Q
2
SO(3) and (29.9) implies that D

Q
1
=

Q
2
, hence
=

Q
2
D
1

Q
1
1
. (29.10)
Lemmas 29.5 and 29.7 imply that each proper Lorentz transformation admits a factorization
= R
q
1
...R
q
6
.
Corollary 29.10 The matrix () := (q
1
)...(q
6
) satises
(p) = ()(p)
1
(), p IR
4
. (29.11)
Proof Lemma 29.1 implies by induction that
()(p)
1
() = (q
1
)...(q
6
)(p)
1
(q
6
)...(q
1
)
1
= (R
q
1
...R
q
6
p) = (p). (29.12)
29.2 Rotations of Dirac Spinors
This corollary allows us to construct explicitly the matrix (). Consider, for example, a rotation
=

R(n) through (, ) about an axis n. Then (28.1) becomes
(x
) := (

R(n))(

R(n)x
), x
IR
4
(29.13)
since

R
#
(n) =

R(n) for the orthogonal matrix R(n). The rotation

R(n) xes two vectors n
and e
0
. Hence we can apply Lemma 29.5 to factorize

R(n) and then Corollary 29.10 to construct
(

R(n)).
According to Lemma 29.5, we choose any non-null vector u which is perpendicular to n and note
that
R(n)u = cos u sin n u (29.14)

for the rotation about n in a positive direction. According to Lemma 29.5, we dene
w = u +

R(n)u = (1 + cos )u sin n u
= 2 cos

2
[cos

2
u sin

2
n u]. (29.15)
It is a nonzero vector since (, ), hence

R(n) =

R
w

R
u
by Lemma 29.5. Therefore, by Corollary
29.10, we can choose
(

R(n)) = (w)(u) =
_
0 w
w 0
__
0 u
u 0
_
=
_
(w)(u) 0
0 (w)(u)
_
. (29.16)
Since the reection

R
w
is unaected by any normalization of w, let us redene w as the unit vector
cos

2
u + sin

2
n u. Then
(w)(u) = uw+i(wu) = cos

2
i sin

2
n = exp(
i
2
n) (29.17)
29. LORENTZ TRANSFORMATION OF SPINORS 139
by the Euler trick. Finally,
(

R(n)) =
_
exp(
i
2
n) 0
0 exp(
i
2
n)
_
. (29.18)
Finally, (29.13) becomes
(x
) :=
_
exp(
i
2
n) 0
0 exp(
i
2
n)
_
(

R(n)x
), x
IR
4
. (29.19)
Remarks 29.11 i) The matrices (29.18) are unitary, because they represent pure rotations.
ii) A map

R(n) (

R(n)) is a homomorphism of the one-parametric subgroup of the Lorentz group
L, into the unitary group U(4).
iii) Our calculation prove that (29.18) satises the identity (29.11) for (, ). However it holds
for any IR by analytic continuation and gives, for = 2, that
(

R(2n)) = 1. (29.20)
Hence, (29.19) implies that the Dirac spinor changes sign by a rotation through 2.
Corollary 29.12 The innitesimal generator G of rotations about the axis n satises the identity
ih
G =
h
2
_
n 0
0 n
_
+

Ln, (29.21)
where

L stands for the standard orbital angular momentum operator.
Proof The innitesimal generator is dened as the derivative in of the rotation around n. Formula
(28.1) implies that the generator is given by
G(x) :=
d
d
=0
_
(

R(n))(

R(n)x)
_
, x IR
4
. (29.22)
Therefore, ih
G consists of two terms:

ih
G(x) := ih
_
d
d
=0
(

R(n))
_
(x) ih
d
d
=0
(

R(n)x). (29.23)
Dierentiating (29.18), we obtain that the rst term coincides with the rst summand in (29.21).
Exercise 29.13 Check that the second term in (29.23) coincides with the second summand in (29.21).
Hints: i) Choose the coordinates with e
3
= n. Then

Ln =

L
3
and is an angle of rotation around e
3
in a positive direction. ii) Use the formula
L
3
(x) = ih
(

R(n)x). (29.24)
Now (29.21) is proved.
Exercise 29.14 Check the formula (29.24).
140
30 Coupling to Maxwell Field
We dene the Dirac equation with an external Maxwell eld and prove gauge invariance.
30.1 Dirac Equation in the Maxwell Field
Let us denote
0
=
0
,
j
=
j
and
P
0
= ih

x
0
, P
j
= ih

x
j
. (30.1)
Write the Dirac equation (24.4) in the form
(P)(x) = c(x), x IR
4
. (30.2)
where the dierential operator (P) :=
is called the Dirac operator.

Let us recall the Schrodinger equation with the Maxwell eld:
[ih
t
e(t, x)](t, x) =
1
2
[ih
e
c
A(t, x)]
2
(t, x), (30.3)
In the notations (30.1)
c[P
0
e
c
(t, x)](t, x) =
1
2
[P
e
c
A(t, x)]
2
(t, x). (30.4)
In other words, in the presence of the Maxwell eld, we change P
0
to P
0
e
c
(t, x) and P to P
e
c
A(t, x). This suggests the following generalization of the Dirac equation (30.2) for the electron eld
in an external Maxwell eld:
(P
e
c
/(x))(x) = c(x), x IR
4
, (30.5)
where /(x) = ((t, x), A(t, x)).
Denition 30.1 We accept (30.5) as the denition of the Dirac equation for the electron eld (x)
in the presence of an external Maxwell eld /(x).
30.2 Gauge Invariance
Let us recall that the gauge transformation
(t, x) (t, x) +
1
c
(t, x), A(t, x) A(t, x)
x
(t, x) (30.6)
does not change the Maxwell elds corresponding to the potentials (t, x) and A(t, x). Let us rewrite
it as
(x)
(x) := (x) +
1
ih
P
0
(x), A(x) A
(x) := A(x) +
1
ih
P(x) (30.7)
Theorem 30.2 Let (x) be a solution of the Dirac equation (30.5) with the potentials (x), A(x).
Then
(x) := exp(
e(x)
ih
c
)(x) satises (30.5) with the potentials
(x), A
(x).
Proof This follows from a direct calculation, since P exp(
e(x)
ih
c
) = exp(
e(x)
ih
c
)
e
ih
c
P(x).
31. PAULI EQUATION AS NONRELATIVISTIC APPROXIMATION 141
31 Pauli Equation as Nonrelativistic Approximation
We demonstrate that the Pauli Equation corresponds to the nonrelativistic approximation of the Dirac
equation. This justies the correspondence between relativistic and nonrelativistic quantum theories
and the Lande factor g = 2, predicted by Goudsmith and Uhlenbeck, for the spin.
First, rewrite the Dirac equation (30.5), similarly to (30.4), as
(ih
t
e) =
0
[c(P
e
c
A) +c
2
], (31.1)
where g := (
1
,
2
,
3
) := (
1
,
2
,
3
). Let us substitute here the block form of the matrices
(24.11) and the splitting of the wave function (25.3). Then the Dirac equation reduces to the coupled
equations
_
_
(ih
t
e)
+
= c(P
e
c
A)
+c
2
+
(ih
t
e)
= c(P
e
c
A)
+
c
2
(31.2)
Now let us replace by a gauge transformation

:= exp(ic
2
t/h
). This cancels the term c

2
in the
rst equation and doubles it in the second one:
_
_
(ih
t
e)
+
= c(P
e
c
A)
(ih
t
e)
= c(P
e
c
A)
+
2c
2

(31.3)
Let us assume that the LHS of the last equation is small compared to c
2
. This is true in the limit
c which corresponds to the nonrelativistic approximation. Namely, the Lorentz transformations
become then the Galilean ones, the retarded potentials becomes the Coulomb ones, etc.
Remark 31.1 The LHS represents a kinetic energy and an electrostatic potential. Usually both are
small compared to the rest energy c
2
.
So the last equation can be approximated by
2c
2

= c(P
e
c
A)
+
. (31.4)
Substituting it into the rst equation (31.3), we obtain
(ih
t
e)
+
=
1
2
((P
e
c
A))
2

+
. (31.5)
Now we have to evaluate the operator on the RHS.
Lemma 31.2 The identity
((P
e
c
A))
2
= (P
e
c
A)
2
+
e
c
h
B, (31.6)
holds, where B := rot A is the magnetic eld.
Proof First, we have the standard identity similar to (29.17):
((P
e
c
A))
2
= (P
e
c
A)(P
e
c
A) +i((P
e
c
A) (P
e
c
A)). (31.7)
142
Let us note that its proof does not depend on the commutation of the components of the vector
P
e
c
A. On the other hand, the vector product of the vector P
e
c
A with itself does not vanish since
its components do not commute. For example, let us calculate the rst component:
(P
2
e
c
A
2
)(P
3
e
c
A
3
) (P
3
e
c
A
3
)(P
2
e
c
A
2
) = [P
2
e
c
A
2
, P
3
e
c
A
3
]. (31.8)
The commutator obviously reduces to
e
c
([P
2
, A
3
] + [A
2
, P
3
]) = i
e
ch
(
2
A
3
3
A
2
) = i
e
ch
(rot A)
1
= i
e
ch
B
1
. (31.9)
Now (31.5) becomes
ih
+
=
1
2
_
(P
e
c
A)
2
e
c
h
B
+
+e
_
+
(31.10)
which coincides with the Pauli equation (19.12).
Remark 31.3 This agreement with the Pauli spin theory was one of the great triumphs of the Dirac
theory. It means that the Dirac equation automatically explains the Lande factor g = 2 for the spinor
electron magnetic moment, suggested by Goudsmith and Uhlenbeck to explain the Einstein-de Haas
and Stern-Gerlach experiments.
Remark 31.4 For the Dirac equation only the sum of orbital and spinor angular momentum is
conserved, while for the Pauli equation both are conserved separately. This means that the Dirac
equation requires both moments for relativistic invariance, while the Pauli equation is a degenerate
version of the Dirac one.
32. CHARGE CONTINUITY EQUATION 143
32 Charge Continuity Equation
We dene the charge-current densities of the Dirac equation and prove the continuity equation.
Denition 32.1 For a spinor wave function (x)
i) The corresponding charge-current densities are dened by
_
_
(x) = e
(x)(
0
)
2
(x) := e
(x)
(x)
j
k
(x) = e
(x)
0
k
(x) := e
(x)(
0
k
(x))
x IR
4
, (32.1)
where
stands for the conjugate transpose to .

ii) The vector s(x) := ((x), j(x)) with the components s
(x) = e
(x)
0
(x) is called four-

current density.
Remark 32.2 The charge density is non-positive and equal to
(x) = e
(x)(x) := e[(x)[
2
. (32.2)
Theorem 32.3 For any solution (x) to the Dirac equation (30.5), the four-current density satises
the continuity equation
s
= 0, x IR
4
. (32.3)
Proof Substituting the denitions (32.1), we obtain
ih
= ih
e
_

x
0
_
_
+

x
k
_
_
_
. (32.4)
Each term produces two summands, with derivatives of or
. The contribution coming from is

ih
0
_
0

x
0
+
k

x
k
_
= e
0
(P) = e
0
_
c +
e
c
(/)
_
(32.5)
by the Dirac equation (32.1). The contribution from
can be evaluated similarly, with the help of

the identities (
0
k
)
=
0
k
: it is equal to
e
_
ih
0
_
0

x
0
+
k

x
k
_
= e(
0
(P))
= e
_
0
_
c +
e
c
(/)
_
. (32.6)
Finally, the sum of the contributions from and
vanishes by the same identities.

144
33 Charged Antiparticles
We establish a one-to-one correspondence between solutions to the Dirac equations with charge e and
e, respectively. Namely, consider the Dirac equation (30.5) with e instead of e:
(P +
e
c
/(x))(x) = c(x), x IR
4
, (33.1)
It describes the positron wave eld corresponding to particles with positive charge e = [e[. We will
establish an isomorphism between the solutions to (30.5) and (33.1).
Denition 33.1 The charge conjugation operator K maps a Dirac wave function to
c
= K :=
2
, and
c
is called the charge-conjugated wave function.
Let us note that the operator K interchanges the rst two components with the last two components of
the Dirac spinor. Hence it changes the energy sign: the factor exp(iEt) interchanges with exp(iEt)
Theorem 33.2 Let satisfy the Dirac equation (P
e
c
/(x))(x) = c(x) for mass and charge
e. Then
c
satises the Dirac equation (33.1) for mass and charge e.
Proof Step i) Let us check that for any vector p C
4
,
2
(p)
2
= (p). (33.2)
That is, all matrices
are real except

2
which is purely imaginary, hence
=
_

, ,= 2
, = 2.
(33.3)
On the other hand, the anticommutation relations for the Dirac matrices imply
2
=
_

, ,= 2
, = 2.
(33.4)
Therefore, (33.2) follows.
Step ii) Conjugating the Dirac equation for , we obtain
2
(P
e
c
/(x))
2
= c. (33.5)
Here
e
c
/(x) is real, while P includes the imaginary factor i. Hence, multiplying (33.5) by
2
, we get
(P +
e
c
/(x))
c
= c
c
. (33.6)
34. HYDROGEN ATOM VIA DIRAC EQUATION 145
34 Hydrogen Atom via Dirac Equation
We determine the quantum stationary states of the hydrogen atom from the Dirac equation. The
main issue is the spherical symmetry of the problem. We apply the method of separation of variables
and take into account the information on the irreducible representations of the rotation group from
Lecture 9.
34.1 Spectral Problem and Spherical Symmetry
Let us consider the Dirac equation for the electron eld in the hydrogen atom. The corresponding four-
potential of the nucleus is / = (, 0, 0, 0) with = e/[x[. Then the corresponding Dirac equation
becomes
ih
t
= H
D
:= c
0
(c (P)) +e(x). (34.1)
We are going to determine all quantum stationary states which are the solutions of the form
E
(x)e
iEt/h
with nite charge

Q(
E
) =
_
[
E
(x)[
2
dx < (34.2)
(see (32.2)). Substituting into the Dirac equation (34.1), we get the corresponding stationary eigen-
value problem
E
E
= H
D
E
. (34.3)
It reduces to the coupled equations for the components of the spinor
E
=
_
_
:
_
(E c
2
e)
+
= cP
(E +c
2
e)
= cP
+
.
(34.4)
The main issue for the solution of the problem is its spherical symmetry. Namely, the nucleus potentials
are spherically symmetric and it is possible to prove that the angular momentum J = J() = ,
J)
is conserved for the solutions to (34.1). Here, by Denition 26.1,
J :=

L +

S (34.5)
is the total angular momentum operator, where

L := ih
x and

S :=
1
2
h
are the orbital and

spin angular momentum operator, respectively. The conservation follows from the commutation
[
J, H
D
] = 0. (34.6)
Exercise 34.1 Check (34.6). Hint: For the free Dirac equation with = 0, the commutation is
proved in Theorem 26.2. It remains to check the commutation of

J with (x) which follows obviously
from the spherical symmetry of the potential.
34.2 Spherical Spinors and Separation of Variables
Let us recall that we have solved the spectral problem for the nonrelativistic Schrodinger equation by
a general strategy of separation of variables (Section 9.1). Now we are going to develop it analogously
for the relativistic problem (34.3). In this case, the role of the orbital angular momentum L is played
by the total angular momentum J since it is conserved. Hence, the strategy now has to be modied
correspondingly:
146
I. First, (34.6) implies that the operator

J
2
:=

J
2
1
+

J
2
2
+

J
2
3
commutes with H
D
:
[
J
2
, H
D
] = 0. (34.7)
Second, H
D
also commutes with each

J
n
. Hence, each eigenspace of the Dirac operator H
D
is invariant
with respect to each operator

J
n
and

J
2
. Moreover, the operator

J
2
commutes also with each operator
J
n
, for example,
[
J
2
,
J
3
] = 0. (34.8)
Exercise 34.2 Check (34.8). Hint: First, prove the commutation relations [
J
k
,
J
j
] = ih
kjl
J
l
where
kjl
is a totally antisymmetric tensor. The relations follow from similar ones for the orbital and
spinor angular momenta, and from the commutation of the momenta.
Hence, we could expect that there is a basis of common eigenfunctions for the operators H
D
,

J
3
and
J
2
. Therefore, it would be helpful to rst diagonalize simultaneously

J
2
and

J
3
.
II. The condition (34.2) means that we consider the eigenvalue problem (34.3) in the Hilbert space
c := L
2
(IR
3
) C
4
. On the other hand, both operators,

J
3
and

J
2
, act only on the spinor variables
and angular variables in spherical coordinates. Hence, the operators act also in the Hilbert space
c
1
:= L
2
(S, dS) C
4
, where S stands for the two-dimensional sphere [x[ = 1. Moreover, both
operators

J
3
and

J
2
have the block form
J
3
=
_

L
3
+s
3
0
0

L
3
+s
3
_
,

J
2
=
_
(
L +s)
2
0
0 (
L +s)
2
_
, (34.9)
where s :=
1
2
h
. Therefore, we can split the space c

1
into the sum c
1
= E
+
1
E
1
, where E
1
= E
1
:=
L
2
(S, dS) C
2
, and the action of the operators is identical in each summand. In the following section
we will prove
Lemma 34.3 i) In the space E
1
there exists an orthonormal basis of Spinor Spherical Harmonics
o
jk
(, ) which are common eigenfunctions of the operators

J
3
and

J
2
:
J
3
o
jk
(, )=h
ko
jk
(, ),

J
2
o
jk
(, )=h
2
j(j + 1)o
jk
(, ), k=j,j+1, ..., j, (34.10)
where j =
1
2
,
3
2
, ....
ii) The space of the solutions to (34.10) is two-dimensional for each xed j, k except for a one-
dimensional space for j = 1/2, k = 1/2.
III. The lemma suggests that we could construct the eigenfunctions of the Dirac operator H
D
, by
separation of variables, in the form
E
=
_

+
_
=
_
_
_
R
+
+
(r)o
+
jk
(, ) +R
+
(r)o
jk
(, )
R
+
(r)o
+
jk
(, ) +R
(r)o
jk
(, )
_
_
_ (34.11)
where o
+
jk
, o
jk
is the basis of the solutions to the equations (34.10) with the xed j, k. The following
theorem justies this particular choice for the eigenfunctions.
Lemma 34.4 (On Separation of Variables) Each solution to the spectral problem (34.3) is a sum (or
a series) of solutions of the particular form (34.11).
Formal Proof Let
jk
denote the orthogonal projection in E
1
onto the linear span E
jk
of o
jk
. Let
us dene its action also in c = E
1
L
2
(IR
+
) E
1
L
2
(IR
+
) as
jk
1
jk
1, or equivalently,
[
jk
_

+
_
](r, , ) :=
_

jk
[
+
(r, , )]
jk
[
(r, , )]
_
, r > 0. (34.12)
in the spherical coordinates r, , . Then
jk
commutes with the Schrodinger operator H since the
latter commutes with H
3
and H
2
:
[H,
jk
] = 0. (34.13)
Hence, applying
jk
to (34.3), we get formally
E
jk
= H
jk
. (34.14)
It remains to note that
i) The function
jk
has the form (34.11) since

jk
[
(r, , )] E
jk
, and dimc
jk
2;
ii)
j,k

jk
.
Remarks 34.5 i) A complete solution to the spectral problem (34.11) relies on an investigation of all
commutation relations of the operators

J
k
,

L
k
and

S
k
, k = 1, 2, 3, i.e. the Lie algebra generated by
them.
ii) It still remains to determine the radial functions in (34.11). We will substitute (34.11) into the
equation (34.4). This gives a radial eigenvalue problem which will be solved explicitly.
Tensor product and Clebsch-Gordan Theorem
We prove Lemma 34.3. In (34.10), we consider the action of the operator

J =

L + s in the space
E
1
:= L
2
(S, dS) C
2
. Let us note that the operator

L acts on L
2
(S, dS) while s acts on the second
factor C
2
, and

L obviously commutes with s. Therefore, the operator

J is a generator of the tensor
product of the regular and spinor representations of the rotation group SO(3) since

L, s are the
generators of the representations. Then the eigenfunctions and eigenvalues of

J
2
can be found by the
Clebsch-Gordan theorem [38].
Namely, we know the spectral decomposition of the operator

L
2
in the space L
2
(S, dS):
L
2
(S, dS) =
l=0
L(l). (34.15)
Here the L(l) are nite-dimensional orthogonal eigenspaces of the operator

L
2
corresponding to the
eigenvalues h
2
l(l + 1), where l = 0, 1, .... In L(l) there is an orthonormal basis e
l
, ..., e
l
where
e
m
=

H
m+l
+
e
l
(here

H
+
:=

H
1
+i
H
2
, where H
k
:= h
1
L
k
) and (cf. (34.10))
L
3
e
lm
= h
me
lm
,

L
2
e
lm
= h
2
l(l + 1)me
lm
, m = l, ..., l. (34.16)
Namely, e
lm
= h
Y
m
l
, where Y
m
l
are Spherical Harmonics (8.8). Similarly, in C
2
there is an orthonormal
basis f
1/2
, f
1/2
where f
1/2
= s
+
f
1/2
and
s
3
f
s
= h
sf
s
, s
2
f
s
= h
2
s(s + 1)f
s
, s = 1/2, 1/2. (34.17)
Namely, f
1/2
=
_
0
1
_
and f
1/2
=
_
1
0
_
. Therefore, we have
L
2
(S, dS) C
2
=
l=0
L(l) C
2
, (34.18)
148
and the tensor products e
lm
f
s
with m = l, ..., l and s = 1/2, 1/2 form an orthonormal basis in
the space E(l) := L(l) C
2
. The relations (34.16), (34.17) imply that
J
3
e
lm
f
s
= h
(l +s)e
lm
f
s
. (34.19)
Now let us state the Clebsch-Gordan theorem for our particular case. It is known as the addition
of angular momenta.
Lemma 34.6 For each l = 0, 1, ...
i) The space E(l) is an orthogonal sum of two eigenspaces E
(l) of the operator

J
2
:
E(l) = E
+
(l) E
(l),

J
2
[
E
(l)
= h
2
l
(l
+ 1), (34.20)
where l
= l 1/2 and dimE
(l) = 2l
+ 1.
ii) For l 1, in the space E
(l) there exists a basis o
k
, k = l
, l
+ 1, ..., l
, satisfying the
eigenvalue problem (34.10) with j = l
. For l = 0, the space E
(0) = 0 and in E
+
(0) there exists a
basis o
+
k
, k = 1/2, 1/2, satisfying (34.10) with j = 1/2.
Exercise 34.7 Prove the lemma. Hints: i) Let us denote

J
+
:=

J
1
+ i
J
2
. Then o
k
:=

J
k+l
+
+
o
l
+
,
k=l
, l
+ 1, ..., l
.
ii) Hence, the space E
(l) is uniquely determined by o
E(l) which is an eigenvector of the

operator

J
3
with the eigenvalue h
.
iii) Obviously, o
+
l
+
= e
l
f
1/2
and it remains to construct o
. It is orthogonal to E
+
(l) and
belongs to the subspace F E(l) which consists of all eigenvectors of the operator

J
3
, with the
eigenvalue h
, in the space E(l). We have to choose a nonzero vector o
F which is orthogonal
to E
+
(l). Further, consider two cases, l 1 and l = 1, separately:
l 1. In this case the eigenspace F is the two-dimensional linear span of the vectors e
l
f
1/2
and e
l+1
f
1/2
if l 1. The intersection E
+
(l) F is the one-dimensional linear span of the vector
J
+
o
+
l
+
. Hence, o
F is determined uniquely (up to a factor) as a vector orthogonal to

J
+
o
+
l
+
.
l = 0. Now the space F is the one-dimensional span of e
0
f
1/2
since the vector e
1
does not exist
in this case. Hence, E
(0) = 0.
Proof of Lemma 34.3 i) The functions o
,k
:= J
k+l
3
o
, k = l
, ..., l
, are solutions to the

problem (34.10) with j = l
, and form an orthogonal basis in the space E(l).

ii) The solutions are linear combinations of the orthogonal functions o
+
j,k
and o
j,k
if (j, k) ,= (1/2, 1/2).
Otherwise, all solutions are proportional to o
+
1/2,1/2
since the function o
1/2,1/2
does not exist.
It turns out that the spaces E
(l) are eigenspaces also for the operator L:

Lemma 34.8
L takes the value h
l resp. h
(l + 1) on the space E
+
(l) resp. E
(l).
Proof This follows immediately from the identity
L = [
L +
1
2
h
]
2

L
2
1
4
h
2
(34.21)
since l
(l
+ 1) l(l + 1) equals either h
l or h
(l + 1).
34.3 Radial Equations
We are going to substitute the expansion (34.11) into the coupled equations (34.4) to derive ordinary
dierential equations for the radial functions R
(r). For this purpose we need an expression of the

operator P in terms of the orbital angular momentum and related operators. The following lemma
gives the necessary relations.
Lemma 34.9 The following relations hold:
P = [x[
2
x(xP+iL), (34.22)
x(L +h
) + (L +h
)x = 0. (34.23)
Proof The formula for products of spin matrices gives
(x)(P) = xP+i(x P) = xP+iL. (34.24)
Now the equation (34.22) follows on multiplying this equation on the left by x. The equation (34.23)
follows on multiplying the commutation relations [L
j
, x
k
] = ih
jkl
x
l
by
j
k
= 2
jk

k
j
and
simplifying.
Substituting the expression (34.22) into the equations (34.4) and using (34.23), we get
_
_
(E c
2
e)
+
= c[x[
2
x(xP+iL)
= c[x[
2
(xPi(L +h
))x
(E +c
2
e)
= c[x[
2
x(xP+iL)
+
.
(34.25)
The last equation can be rewritten as
(E +c
2
e)x
= c(xP+iL)
+
. (34.26)
Together with the second equation in (34.25), this suggests the substitution
:=
x
[x[
and
+
:=
+
. Then we get, rewriting the equations (34.25) in spherical coordinates,
_
_
(E c
2
e)
+
= c(ih
d
dr
ir
1
(L + 2h
))
(E +c
2
e)
= c(ih
d
dr
+ir
1
L)
+
.
(34.27)
By Lemma 34.4, it suces to construct all nonzero solutions to (34.4) in the form (34.11). For example,
we can assume that
R
+
+
(r) , 0. (34.28)
Let us denote by
+
jk
the orthogonal projection in the space E
1
onto the linear span of the function
o
+
jk
. Denote by
+
jk
:=
+
jk
1 the corresponding projector in the space E
1
L
2
(IR) = L
2
(IR
3
) C
2
.
Then
+
jk
commutes with the operator
L by Lemma 34.8. Hence, applying

+
jk
to the equations
(34.27), we get
_
_
(E c
2
e)
+
jk
+
= c(ih
d
dr
ir
1
(h
l + 2h
))
+
jk
(E +c
2
e)
+
jk
= c(ih
d
dr
+ir
1
h
l)
+
jk
+
,
(34.29)
150
where l = j
1
2
by the lemma. Let us note that
+
jk
+
= R
1
(r)o
+
jk
(, ),
+
jk
= R
2
(r)o
+
jk
(, ), (34.30)
where
R
1
(r) R
+
+
(r) , 0 (34.31)
by (34.28). Let us denote by R(r) :=
_
R
1
(r)
R
2
(r)
_
and substitute the representations (34.30) into the
equations (34.29). Then (34.29) is equivalent to the following radial equation for the vector-function
R(r):
(E e c
2
3
)R(r) = ich
_
_
d
dr
+
1
r
_
+
(l + 1)
r

3
_
1
R(r), r > 0, (34.32)
where
k
are the Pauli matrices.
Corollary 34.10 The eigenvalue problem for the Dirac equation with a spherically symmetric elec-
trostatic potential can be written in the form
(E e c
2
3
)R(r) = ich
_
_
d
dr
+
1
r
_
1
+i
(l + 1)
r

2
_
R(r), r > 0 (34.33)
which follows from (34.32) since
3
1
= i
2
.
34.4 Hydrogen Spectrum
Here we calculate the eigenvalues E of the problem (34.3). As in the nonrelativistic case, substitute
R(r) = e
r
P(r). Then the equation (34.32) reduces to
(E e c
2
3
)P(r) = ich
_
_
d
dr
+
1
r

_
1
+i
(l + 1)
r

2
_
P(r), r > 0, (34.34)
or equivalently, to
(E c
2
3
ich
1
)P(r) = ich
_
_
d
dr
+
1
r
_
1
+i
(l + 1)
r

2
+
ie
ch
_
P(r), r > 0. (34.35)
For the above matrix we introduce the notation
M
i
ch
(E c
2
3
ich
1
), (34.36)
because it frequently appears in the calculations. Let us also rewrite the Coulombic potential as
e = ch
/r, where
:=
e
2
ch

1
137
is the dimensionless Sommerfeld ne structure constant.
We nd the parameter from the asymptotic condition at innity. Namely, we suggest that R(r)
is a polynomial
R(r) = r
0
R
k
r
k
(34.37)
with an R
n
,= 0. Then the equation (34.35) implies that MR
n
= 0, hence det M = 0:
det(E c
2
3
ich
1
) = 0. (34.38)
This is equivalent to
c
2
h
2
=
2
c
4
E
2
, (34.39)
so, in particular, E < c
2
. We have to choose the positive root for to have a solution satisfying
(34.2). To justify (34.37), we seek a solution in the general form
R(r) = r
0
R
k
r
k
(34.40)
where we can assume that R
0
,= 0 for a nontrivial solution. Substituting into (34.35), we get the
equation
0
r
k+
MR
k
=
0
_
(k + + 1)
1
+i(l + 1)
2
+i
_
r
k+1
R
k
. (34.41)
This gives the recurrence equation
MR
k1
=
_
(k + + 1)
1
+i(l + 1)
2
+i
_
r
k+1
R
k
, k = 0, 1, ... (34.42)
This equation with k = 0 implies that
_
( + 1)
1
+i(l + 1)
2
+i
_
R
0
= 0. (34.43)
This implies the indicial equation
det
_
( + 1)
1
+i(l + 1)
2
+i
_
= 0 (34.44)
since R
0
,= 0. It is equivalent to
( + 1)
2
= (l + 1)
2
2
. (34.45)
Then [ +1[ l +1 since is small. Therefore, we have to choose the positive root for +1 since for
the negative root we get l + 2 2 while > 3/2 by the condition (34.2).
Finally, an investigation of the recurrence equation (34.42) shows that the series (34.40) should
terminate by the condition (34.2) as in the case of the Schrodinger equation. Hence, we arrive at (34.37)
with an R
n
,= 0. This implies again (34.38) and (34.39), however it is not sucient to determine the
eigenvalues E since we have the additional unknown parameter . Therefore, we need an additional
equation which we will derive from the recurrence equation (34.42) with k = n:
MR
n1
=
_
(n + + 1)
1
+i(l + 1)
2
+i
_
r
n+1
R
n
. (34.46)
Namely, the characteristic equation for the matrix M reads,
(M 2iE/ch
)M = 0 (34.47)
since its determinant is zero and the trace is 2iE/ch
. Therefore, multiplying both sides of (34.42) by

M 2iE/ch
, we get
0 = (M 2iE/ch
)
_
(n + + 1)
1
+i(l + 1)
2
+i
_
R
n
. (34.48)
152
Multiplying here the Pauli matrices, we arrive at
0 =
_
2(n + + 1) 2E/ch
_
R
n
. (34.49)
This gives us the new quantization condition
E = ch
(n + + 1), (34.50)
which together with the equation (34.39) determines the eigenvalues E: solving the system of equa-
tions, we get
E = E
ln
=
c
2
_
1 +
_
2
/(n + + 1)
2
_
, (34.51)
where = (l) is given by (34.45).
Since is small, we can approximate the eigenvalues by the binomial expansion:
E
ln
c
2
c
2
2
2(n + + 1)
2
. (34.52)
Remarks 34.11 i) The approximation (34.52) with = 0 coincides with the nonrelativistic spectrum
of the hydrogen atom up to the unessential additive constant c
2
.
ii) The relativistic formula depends on the angular momentum j through = (l), while the nonrela-
tivistic formula does not depend on the angular momentum. This was another triumph of the Dirac
theory since it corresponds to the experimental observation of the ne structure.
Remark 34.12 The above analysis gives also the corresponding eigenfunctions
E
=
_
_
_
_
_
_ =
x
[x[
_
_
_
_
_
_ =
x
[x[
_
_
_
R
1
(r)o
+
jk
(, )
R
2
(r)o
+
jk
(, )
_
_
_ (34.53)
Part VII
Mathematical Appendices
153
35. NEWTON MECHANICS 155
35 Newton Mechanics
We recall the Newton mechanics of one and many particles, for potential force elds. In the case of a
certain symmetry of the potential we derive the corresponding conservation laws.
35.1 One Particle
Newton equation
The motion of one particle of mass m > 0 is governed by the Newton dierential equation
m x(t) = F(x(t), t), t IR. (35.1)
Here x(t) IR
3
is the particle position at time t and F() is the force eld. Let us assume that F
C
1
(IR
3
IR, IR
3
). Then the solution x(t) is dened uniquely by the initial conditions x(0) = x
0
IR
3
,
x(0) = v
0
IR
3
by the main theorem of ordinary dierential equations. The solution exists for [t[ ,
where > 0 depends on the initial data x
0
, v
0
.
Energy conservation
Let us assume that the force eld F has a potential function (or simply potential) V () C
2
(IR
3
IR),
F(x, t) = V (x, t), x IR
3
, t IR. (35.2)
Denition 35.13 i) c := IR
3
IR
3
is the phase space of the Newton equation, c
+
:= IR
3
IR
3
IR
is the extended phase space of the Newton equation.
ii) The energy E(x, v, t) is the function on the extended phase space dened by
E(x, v, t) =
mv
2
2
+V (x, t), (x, v, t) c
+
. (35.3)
Theorem 35.14 Let the condition (35.2) hold. Further, we assume that the potential does not depend
on t,
V (x, t) V (x), (x, t) IR
3
IR. (35.4)
Then for every solution x(t) C
2
([t
0
, t
1
], IR
3
) to the Newton equation, the energy is conserved,
E(t) := E(x(t), x(t)) = const, t [t
0
, t
1
]. (35.5)
Proof By the chain rule of dierentiation, the Newton equation (35.1), and (35.2),
E(t) = m x(t) x(t) +V (x(t)) x(t) = [m x(t) +V (x(t))] x(t) = 0, t [t

0
, t
1
]. (35.6)
Well-posedness condition
Theorem 35.15 Let the condition (35.4) hold, and let the potential be bounded from below by some
constant C IR,
V (x) C, x IR
3
. (35.7)
Then every solution x(t) to the Newton equation (35.1) exists globally in time, i.e., for all t IR.
Proof The energy conservation (35.5) implies that the velocity is bounded, [ x(t)[ v. Then also
[x(t)[ is bounded by vt + const. This provides the existence of the global solution for all t IR.
156
35.2 Many Particles
Newton equations
The motion of n particles with masses m
i
> 0 is governed by the Newton dierential equation
m
i
x
i
(t) = F
i
(x(t), t), t IR, i = 1, ..., n. (35.8)
Here
i) x
i
(t) IR
3
is the position of the i-th particle at time t, x(t) = (x
1
(t), ..., x
n
(t)) IR
3n
and
ii) F
i
(x(t), t) IR
3
is the force acting on the i-th particle.
Let us assume that the force eld F(x, t) := (F
1
(x, t), ..., F
n
(x, t)) C
1
(IR
3n
IR, IR
3n
). Then
the solution x(t) = (x
1
(t), ..., x
n
(t)) to the system (35.8) is dened uniquely by the initial conditions
x(0) = x
0
IR
3n
, x(0) = v
0
IR
3n
by the main theorem of ordinary dierential equations. The
solution exists for [t[ , where > 0 depends on the initial data x
0
, v
0
.
Energy conservation
Let us assume that the force eld F(x, t) has a potential function (or simply potential) V () C
2
(IR
3n
),
F
i
(x, t) =
x
i
V (x, t), x IR
3n
, i = 1, ..., n. (35.9)
Denition 35.16 i) c = IR
3n
IR
3n
is the phase space of the Newton system (35.8), c
+
:=
IR
3n
IR
3n
IR is the extended phase space of the Newton system (35.8).
ii) The energy E(x, v, t) is the function on the extended phase space dened by
E(x, v, t) =
i
m
i
v
2
i
2
+V (x, t), (x, v) c, (35.10)
where v = (v
1
, ..., v
n
).
Let us call a trajectory any solution x(t) to the Newton system (35.8).
Theorem 35.17 Let the condition (35.9) hold. Further, we assume that the potential does not depend
on t,
V (x, t) V (x), (x, t) IR
3n
IR. (35.11)
Then for any trajectory x(t) C
2
([t
0
, t
1
], IR
3n
), the energy is conserved,
E(t) := E(x(t), x(t)) = const, t [t
0
, t
1
]. (35.12)
Proof By the chain rule of dierentiation, the Newton system (35.8), and (35.9),
E(t) =
i
m
i
x
i
(t) x
i
(t) +
x
i
V (x(t)) x
i
(t)
=
i
[m
i
x
i
(t) +
x
i
V (x(t))] x
i
(t) = 0, t [t
0
, t
1
]. (35.13)
Well-posedness condition
Theorem 35.18 Let the condition (35.11) hold, and let the potential be bounded from below by some
constant C IR:
V (x) C, x IR
3n
. (35.14)
Then every solution x(t) to the Newton equation (35.8) exists globally in time, i.e. for all t IR.
Proof The energy conservation (35.12) implies that the velocity is bounded, [ x(t)[ const. Then also
[x(t)[ is bounded by vt + const. This provides the existence of the global solution for all t IR.
35. NEWTON MECHANICS 157
35.3 Symmetry Theory
The invariance of the potential V with respect to translations in time, (35.11), provides the energy
conservation (35.12). Let us show that the invariance of the potential V (x) with respect to some
transformations of the conguration space Q := IR
3n
leads to new conservation laws.
Translation group
Let us x a vector h ,= 0 in IR
3
and consider the translations x x+hs of IR
3
and the corresponding
action in IR
3n
:
T
s
(x
1
, , , ., x
n
) = (x
1
+hs, , , ., x
n
+hs), (x
1
, , , ., x
n
) IR
3n
. (35.15)
Denition 35.19 The system (35.8) is invariant with respect to the translations (35.15) if
V (T
s
(x), t) = V (x, t), (x, t) IR
3n+1
, s IR. (35.16)
Example 35.20 The Newton system (35.8) is invariant with respect to the translations (35.15) with
every h IR
3
, if the potential energy has the structure
V (x
1
, ..., x
n
, t) = W(x
1
x
n
, ..., x
n1
x
n
, t), (x
1
, ..., x
n
, t) IR
3n+1
(35.17)
with a function W of 3n 2 variables.
Denition 35.21 i) The momentum p
i
of the i-th particle is the vector function on the phase space
c dened by
p
i
:= m
i
v
i
IR
3
, (x, v) c. (35.18)
ii) the (total) momentum p of the system (35.8) is the vector function on the phase space c dened by
p :=
i
p
i
=
i
m
i
v
i
IR
3
, (x, v) c. (35.19)
iii) The center of mass of the system of n particles is
X :=
1
M
i
m
i
x
i
, x = (x
1
, ..., x
n
) IR
3n
, (35.20)
where M :=
i
m
i
is the total mass of the system.
Theorem 35.22 Let (35.9) hold and the system (35.8) be invariant with respect to the translations
(35.15) along a xed vector h IR
3
. Then for any trajectory x(t) C
2
([t
0
, t
1
], IR
3n
), the projection of
the momentum p(t) onto h is conserved,
p
h
(t) := p(t) h = const, t [t
0
, t
1
]. (35.21)
Proof By (35.9), (35.8), and the chain rule of dierentiation,
p
h
(t) =
i
m
i
x
i
(t) h =
x
i
V (x(t), t) h =
d
ds
s=0
V (T
s
x(t), t) = 0, t [t
0
, t
1
] (35.22)
by (35.16).
Corollary 35.23 Let the Newton system (35.8) be invariant with respect to the translations (35.15)
along all vectors h IR
3
. Then for any trajectory, the momentum p(t) is conserved, p(t) = const,
and the center of mass X(t) =
i
m
i
x
i
(t)/M moves uniformly: X(t) = vt +X(0).
Proof Since (35.22) holds for every h, we have p(t) = const and
X(t) =
1
M
i
m
i
x
i
= 0.
158
Rotation group
Let us x a unit vector e IR
3
and consider the rotation around e in IR
3
with an angle of s radian.
Let us denote by R
e
(s) SO(3) the corresponding orthogonal matrix and dene the corresponding
transformation in IR
3n
by
R
s
(x
1
, ..., x
n
) = (R
e
(s)x
1
, ..., R
e
(s)x
n
), (x
1
, ..., x
n
) IR
3n
. (35.23)
Denition 35.24 The system (35.8) is invariant with respect to the rotations (35.23) if
V (R
s
(x), t) = V (x, t), (x, t) IR
3n+1
, s IR. (35.24)
Example 35.25 The system (35.8) is invariant with respect to the rotations (35.15) with every e
IR
3
, if the potential energy has the structure
V (x
1
, ..., x
n
, t) = W([x
i
x
j
[ : 1 i < j n, t), (x
1
, ..., x
n
, t) IR
3n+1
. (35.25)
Denition 35.26 i) The angular momentum L
i
of the i-th particle is the vector function on the phase
space c dened by
L
i
(x, v) := x
i
p
i
IR
3
, (x, v) c. (35.26)
ii) the angular momentum L of the Newton system (35.8) is the vector function on the phase space c
dened by
L(x, v) :=
i
L
i
=
i
x
i
p
i
IR
3
, (x, v) c. (35.27)
Theorem 35.27 Let the Newton system (35.8) be invariant with respect to the rotations (35.23)
around a xed vector e IR
3
2
([t
0
, t
1
], IR
3n
), the projection of the
angular momentum L(x(t), x(t)) onto e is conserved,
L
r
(t) := L(x(t), x(t)) e = const, t [t
0
, t
1
]. (35.28)
Proof The dierentiation gives,
L
e
(t) = [
i
x
i
(t) p
i
(t) +
i
x
i
(t) p
i
(t)] e
= [
i
x
i
(t) m
i
x
i
(t) +
i
x
i
(t) m
i
x
i
(t)] e
=
i
[x
i
(t) m
i
x
i
(t)] e =
i
m
i
x
i
(t) [e x
i
(t)]. (35.29)
Therefore, the Newton system (35.8) and (35.27) imply by the chain rule of dierentiation,
L
e
(t) =
x
i
V (x(t), t) [e x
i
(t)] =
d
ds
s=0
V (R
s
x(t), t) = 0, t [t
0
, t
1
]. (35.30)
by (35.24), since
e x
i
(t) =
d
ds
s=0
R
e
(s)x
i
(t). (35.31)
This identity follows from the fact that the vectors on both sides are orthogonal to the plane containing
e and x
i
(t), and have the same length. Indeed, the length of the LHS is [x
i
(t)[ sin , where is the
angle between e and x
i
(t), and the length of the RHS is the radius of the circle R
e
(s)x
i
(t) : s [0, 2]
which is equal to [x
i
(t)[ sin .
36. LAGRANGIAN MECHANICS 159
36 Lagrangian Mechanics
We introduce Lagrangian systems corresponding to one particle and to many particles, formulate
the Hamilton least action principle, derive the Euler-Lagrange equations and check that the Newton
equations are of the Euler-Lagrange form. In the case of a certain symmetry of the Lagrangian function
we derive the corresponding conservation laws.
36.1 One Particle
Lagrangian function
We expose the Lagrangian form of the Newton equation (35.1) with the potential (35.2),
m x(t) = V (x(t), t), t IR. (36.1)
Denition 36.1 The Lagrangian L(x, v, t) of the system is the following function on the extended
phase space c
+
= IR
3
IR
3
IR (cf. (35.3)),
L(x, v, t) =
mv
2
2
V (x, t), (x, v, t) c
+
. (36.2)
Exercise 36.2 Check that the Newton equation (36.1) can be represented in the Euler-Lagrange
form,
d
dt
L
v
(x(t), x(t), t) = L
x
(x(t), x(t), t), t IR. (36.3)
Let us consider more general Lagrangian systems with an arbitrary function L(x, v, t).
Denition 36.3 i) The Lagrangian system for one particle is the dynamical system described by the
Lagrangian equation (36.3) with a function L(x, v, t) C
2
(c
+
).
ii) The momentum of the Lagrangian system is the vector-function on the extended phase space c
+
dened by
p = L
v
(x, v, t), (x, v, t) c
+
. (36.4)
iii) The energy of the Lagrangian system is the function on the extended phase space c
+
dened by
E(x, v, t) = pv L(x, v, t), (x, v, t) c
+
. (36.5)
Example 36.4 The Newton equation (36.1) results from the Lagrangian system with the Lagrangian
functional (36.2), momentum p = mv, and energy E =
mv
2
2
+V (x, t).
Theorem 36.5 Let the Lagrangian not depend on time,
L(x, v, t) = L(x, v), (x, v, t) c
+
. (36.6)
2
([t
0
, t
1
], IR
3
), the energy is conserved, (35.5).
Proof The dierentiation of (36.5) with x = x(t) and v = x(t) gives,
E(t) = pv +p v L
x
x L
v
v = 0 (36.7)
by Equations (36.3) and Denition (36.4).
Exercise 36.6 Calculate the momentum and the energy for the Lagrangian L(x, v) = m
1 v
2
.
160
Action functional
Denition 36.7 C
1
= C
1
([0, ), IR
3
) is the space of all paths in three-dimensional space.
1
1
IR.
Example 36.8 T(x) =
_
T
0
[ x(t)[dt is the length of the path x() C
1
, t [0, T].
Denition 36.9 The Gateau dierential DT(x) is the linear functional C
1
IR dened by
DT(x), h) =
d
d
=0
T(x +h), h() C
1
(36.8)
if the derivative on the RHS exists.
Let us x a T > 0.
Denition 36.10 The action is the functional on C
1
dened by
S
T
(x()) =
_
T
0
L(x(t), x(t), t)dt, x() C
1
. (36.9)
Note that the functional is dened on the whole of C
1
if L(x, v, t) C(c
+
). Moreover, the functional
is dierentiable if L(x, v, t) C
1
(c
+
):
Lemma 36.11 The Gateau dierential DS
T
(x) exists for x C
1
.
DS
T
(x), h) : =
d
d
=0
_
T
0
L(x(t) +h(t), x(t) +
h(t), t)dt
=
_
T
0
[L
x
(x(t), x(t))h(t) +L
v
(x(t), x(t), t)
h(t)]dt, (36.10)
since L(x, v, t) C
2
(c
+
) by our basic assumptions.
Hamilton least action principle
Denition 36.12 C
1
(T) = h C
1
: h(0) = h(T) = 0.
Denition 36.13 The function x C
1
satises the Hamilton least action principle (LAP) if, for any
T > 0,
DS
T
(x), h) = 0, h() C
1
(T) (36.11)
Theorem 36.14 For x C
2
([0, ), IR
3
) the Hamilton LAP is equivalent to the Euler-Lagrange equa-
tions (36.3) with t [0, T].
Proof The partial integration in (36.10) gives
DS
T
(x), h) =
_
T
0
[L
x
(x(t), x(t), t)
d
dt
L
v
(x(t), x(t), t)]h(t)dt, h C
1
0
(T). (36.12)
Therefore, (36.11) is equivalent to (36.3) by the following lemma:
Lemma 36.15 Main lemma of the calculus of variations (du Bois-Reymond).
Let a function f(t) C[0, T] and
_
T
0
f(t)h(t)dt = 0 for any function h(t) C[0, T] with the boundary
values h(0) = h(T) = 0. Then f(t) = 0, t [0, T].
Exercise 36.16 Prove the lemma.
36. LAGRANGIAN MECHANICS 161
36.2 Many Particles
Lagrangian function
We extend the Lagrangian formalism to the Newton equations (35.8) with the potential (35.9),
m
i
x
i
(t) =
x
i
V (x(t), t), t IR. (36.13)
We introduce the Lagrangian L(x, v, t) of the system (36.13) as the following function on the extended
phase space c
+
= IR
3n
IR
3n
IR (cf. (36.2)),
L(x, v, t) =
i
m
i
v
2
i
2
V (x, t), (x, v, t) c
+
, (36.14)
where v = (v
1
, ...v
n
).
Exercise 36.17 Check that the Newton equation (36.13) can be represented in the Euler-Lagrange
form,
d
dt
L
v
(x(t), x(t), t) = L
x
(x(t), x(t), t), t IR. (36.15)
Let us consider more general Lagrangian systems with the extended phase space c
+
:= IR
N
IR
N
IR, where N = 1, 2..., and an arbitrary function L(x, v, t).
Denition 36.18 i) The Lagrangian system in the extended phase space c
+
:= IR
N
IR
N
IR is the
dynamical system described by the equations (36.15) with a function L(x, v, t) C
2
(c).
ii) The momentum of the Lagrangian system is the vector function on the extended phase space c
+
dened by
p = L
v
(x, v, t), (x, v, t) c
+
. (36.16)
iii) The energy of the Lagrangian system is the function on the phase space c
+
dened by
E(x, v, t) = pv L(x, v, t), (x, v, t) c
+
. (36.17)
Exercise 36.19 Check that the Newton equations (36.13) result from the Lagrangian system with the
Lagrangian functional (36.14), momentum p = (p
1
, ..., p
n
) where p
i
= m
i
v
i
, and the energy
E =
i
m
i
v
2
i
2
+V (x, t). (36.18)
Theorem 36.20 Let the Lagrangian not depend on time,
L(x, v, t) = L(x, v), (x, v, t) c
+
. (36.19)
2
([t
0
, t
1
], IR
N
), the energy is conserved, (35.12).
Proof The dierentiation of (36.17) with x = x(t) and v = x(t) gives,
E(t) = pv +p v L
x
x L
v
v = 0 (36.20)
by Equations (36.15) and Denition (36.16).
162
Action functional
Denition 36.21 C
1
= C
1
([0, ), IR
N
) is the space of all paths in N-dimensional space.
1
1
IR.
Example 36.22 T(x) =
_
T
0
[ x(t)[dt is the length of the path x() C
1
, t [0, T].
Denition 36.23 The Gateau dierential DT(x) is the linear functional C
1
IR dened by
DT(x), h) =
d
d
=0
T(x +h) (36.21)
for h() C
1
if the derivative on the RHS exists.
Let us x a T > 0.
Denition 36.24 The action is the functional on C
1
(T) dened by
S
T
(x) =
_
T
0
L(x(t), x(t))dt, x() C
1
. (36.22)
Note that the functional is dened on the whole of C
1
(T) if L(x, v) C(c). Moreover, the functional
is dierentiable if L(x, v) C
1
(c):
Lemma 36.25 The Gateau dierential DS
T
(x) exists for x() C
1
.
DS
T
(x), h) : =
d
d
=0
_
T
0
L(x(t) +h(t), x(t) +
h(t))dt
=
_
T
0
[L
x
(x(t), x(t))h(t) +L
v
(x(t), x(t))
h(t)]dt, (36.23)
since L(x, v) C
2
(c) by our basic assumptions.
Hamilton least action principle
Denition 36.26 C
1
(T) = h() C
1
: h(0) = h(T) = 0.
Denition 36.27 The function x C
1
satises the Hamilton least action principle (LAP) if for any
T > 0
DS
T
(x), h) = 0, h() C
1
(T). (36.24)
Theorem 36.28 For x C
2
([0, ), IR
N
) the Hamilton LAP is equivalent to the Euler-Lagrange
equations (36.15).
Proof The partial integration in (36.23) gives
DS(x), h) =
_
T
0
[L
x
(x(t), x(t))
d
dt
L
v
(x(t), x(t))]h(t)dt. (36.25)
Therefore, (36.24) is equivalent to (36.15) by the main lemma of the calculus of variations.
Remark 36.29 The expositions in Sections 36.1 and 36.2 formally are almost identical.
37. NOETHER THEORY OF INVARIANTS 163
37 Noether Theory of Invariants
We construct a Noether invariant for a system whose Lagrangian is invariant with respect to a one-
parametric transformation group. This construction is applied to one particle and many particles
mechanical systems.
The invariance of the Lagrangian L with respect to translation in time, (36.19), provides the
conservation of energy (35.5). Let us show that the invariance of the Lagrangian with respect to some
transformations of the conguration space X := IR
N
leads to new conservation laws.
37.1 Symmetry and Noether Theorem on Invariants
Consider a group G = g of dierentiable transformations g C
2
(X, X) of the conguration space
X of a Lagrangian system.
Denition 37.1 G is a symmetry group of the Lagrangian system if the identity
L(g(x), dg(x)v, t) = L(x, v, t), (x, v, t) c
+
, g G (37.1)
holds, where dg : IR
N
IR
N
is the dierential of g.
Let us recall the denition of the dierential:
dg(x)v :=
dg(X())
d
=0
(37.2)
if X(0) = x and

X(0) = v.
Consider a one-parametric subgroup g
s
G : s IR of the symmetry group G,
L(g
s
(x), dg
s
(v), t) = L(x, v, t), (x, v, t) c
+
, s IR. (37.3)
Remark 37.2 Since g
s
is a one-parametric subgroup, we have g
0
= Id, hence
(g
0
(x), dg
0
(v)) = (x, v), (x, v) c. (37.4)
Denition 37.3 The Noether invariant is a function on the extended phase space c
+
, which is dened
by
I(x, v, t) = L
v
(x, v, t)
dg
s
(x)
ds
s=0
, (x, v, t) c
+
. (37.5)
Theorem 37.4 (E.Noether [76]) Let x(t) C
2
(IR, IR
N
) be a solution to the Euler-Lagrange equations
(36.15) and g
s
: s IR a one-parametric symmetry group of the Lagrangian system. Then I(t) :=
I(x(t), x(t), t) =const, t IR.
Proof Dierentiation gives
I(t) =
d
dt
L
v
(x(t), x(t), t)
d
ds
s=0
g
s
(x(t)) +L
v
(x(t), x(t), t)
d
dt
d
ds
s=0
g
s
(x(t)). (37.6)
For the rst summand on the RHS, Eqns. (36.15) give
d
dt
L
v
(x(t), x(t), t)
d
ds
s=0
g
s
(x(t)) = L
x
(x(t), x(t), t)
d
ds
s=0
g
s
(x(t)). (37.7)
164
For the second summand we have
L
v
(x(t), x(t), t)
d
dt
d
ds
s=0
g
s
(x(t)) = L
v
(x(t), x(t), t)
d
ds
s=0
d
dt
g
s
(x(t)), (37.8)
since

2
ts
=

2
st
. At last,
d
dt
g
s
(x(t)) = dg
s
(x(t)) x(t) (37.9)
by Denition (37.2) of the dierential of the map g
s
. Now (37.6)-(37.9) gives, by the chain rule and
(37.4),
I(t) = L
x
(x(t), x(t), t)
d
ds
s=0
g
s
(x(t)) +L
v
(x(t), x(t), t)
d
ds
s=0
dg
s
(x(t)) x(t)
=
d
ds
s=0
L(g
s
(x(t)), dg
s
(x(t)) x(t), t) = 0 (37.10)
according to (37.1).
37.2 Application to Many-Particle Systems
Let us apply the Noether theorem to a Lagrangian system of n particles, i.e., with N = 3n and
x = (x
1
, ..., x
n
), where x
i
IR
3
.
Translation group
Let us x a vector h ,= 0 in IR
3
and consider the transformation T
s
from (35.15). By Denition 37.1
the Lagrangian system is invariant with respect to the translations (35.15) if
L(T
s
(x), dT
s
(x)v, t) = L(x, v, t), (x, v, t) c
+
, s IR. (37.11)
Exercise 37.5 Check that for the translations T
s
, s IR, the dierential is given by
dT
s
(x)v = v, x, v IR
3n
. (37.12)
Exercise 37.6 Check that the Lagrangian system is invariant with respect to the translations (35.15)
for every h IR
3
, if the Lagrangian has the structure (cf. (35.17)),
L(x
1
, ..., x
n
, v, t) = (x
1
x
n
, ..., x
n1
x
n
, v, t), (x
1
, ..., x
n
, v, t) c
+
. (37.13)
Denition 37.7 i) The momentum p
i
of the i-th particle of the Lagrangian system is the vector
function on the space c
+
dened by
p
i
(x, v, t) := L
v
i
(x, v, t) IR
3
, (x, v, t) c
+
. (37.14)
ii) the momentum p of the Lagrangian system is the vector function on the space c
+
dened by
p(x, v, t) :=
i
p
i
=
i
L
v
i
IR
3
, (x, v, t) c
+
. (37.15)
37. NOETHER THEORY OF INVARIANTS 165
Theorem 37.8 Let the Lagrangian system (35.8) be invariant with respect to the translations (35.15)
along a xed vector h IR
3
2
([t
0
, t
1
], IR
3n
), the projection of the
momentum p(x(t), x(t)) onto h is conserved,
p
h
(t) := p(x(t), x(t), t)h = const, t [t
0
, t
1
]. (37.16)
Proof 1 The conservation follows from the Noether theorem for the one-parametric symmetry group
g
s
= T
s
since ph coincides with the corresponding Noether invariant. Indeed, the invariant reads
I := L
v
d
ds
s=0
T
s
(x) =
i
L
v
i
d
ds
s=0
(x
i
+hs) =
i
p
i
h = ph. (37.17)
Proof 2 By Denition 37.7, the Euler-Lagrange equations (36.13), and the chain rule of dierentiation,
p
h
(t) =
i
p
i
(t)h =
x
i
L(x(t), x(t), t)h
=
d
ds
s=0
L(T
s
(x(t)), dT
s
(x(t), t) x(t)) = 0, t [t
0
, t
1
] (37.18)
by (37.12) and (37.11).
Rotation group
Let us x a unit vector e IR
3
and consider the transformation R
s
from (35.23). By Denition 37.1
the Lagrangian system is invariant with respect to the rotations (35.23) if
L(R
s
(x), dR
s
(x)v, t) = L(x, v, t), (x, v, t) c
+
, s IR. (37.19)
Exercise 37.9 Check that for the rotations R
s
, s IR, the dierential is given by
dR
s
(x)v = R
s
v, v IR
3n
. (37.20)
Exercise 37.10 Check that the Lagrangian system (35.8) is invariant with respect to the rotations
(35.23) with every r IR
3
, if the Lagrangian has the structure (cf. (35.25))
L(x
1
, ..., x
n
, v
1
, ..., v
n
, t) =
1
([x
i
x
j
[ : 1 i < j n; [v
i
[ : 1 i n, t),
(x
1
, ..., x
n
, v
1
, ..., v
n
, t) c
+
. (37.21)
Denition 37.11 i) The angular momentum L
i
of the i-th particle is the vector function on the space
c
+
dened by
L
i
(x, v, t) := x
i
p
i
, (x, v, t) c
+
. (37.22)
ii) the angular momentum L of the Newton system (35.8) is the vector function on the space c
+
dened by
L(x, v, t) :=
i
L
i
=
i
x
i
p
i
, (x, v, t) c
+
. (37.23)
Theorem 37.12 Let the Lagrangian system be invariant with respect to the rotations (35.23) around
a xed vector e IR
3
. Then, for any trajectory x(t) C
2
([t
0
, t
1
], IR
3n
), the projection of the angular
momentum L(x(t), x(t)) onto e is conserved,
L
e
(t) := L(x(t), x(t))e = const, t [t
0
, t
1
]. (37.24)
166
Proof 1 The conservation follows from the Noether theorem for the one-parametric symmetry group
g
s
= R
s
, since L
e
coincides with the corresponding Noether invariant. Indeed, the invariant reads
I : = L
v
d
ds
s=0
R
s
(x) =
i
L
v
i
d
ds
s=0
(R
e
(s)x
i
)
=
i
p
i
(e x
i
) =
i
e(x
i
p
i
) = eL (37.25)
according to (35.31).
Proof 2 By Denition 37.11, the Euler-Lagrange equations (36.13), and the chain rule of dierentia-
tion,
L
e
(t) = [
i
x
i
(t) p
i
(t) +
i
x
i
(t) p
i
(t)]e
= [
i
x
i
(t) L
v
i
(x(t), x(t)) +
i
x
i
(t) L
x
i
(x(t), x(t))]e
=
i
(e x
i
(t))L
v
i
(x(t), x(t)) +
i
(e x
i
(t))L
x
i
(x(t), x(t)). (37.26)
However, (35.31) and (37.20) imply that
e x
i
(t) =
d
ds
s=0
R
e
(s)x
i
(t), e x
i
(t) =
d
ds
s=0
dR
e
(s) x
i
(t). (37.27)
Therefore, (37.26) implies by the chain rule and (37.19),
L
e
(t) =
d
dt
L(R
s
(x(t)), dR
s
(x(t)) x(t)) = 0. (37.28)
38. HAMILTON MECHANICS 167
38 Hamilton Mechanics
38.1 Legendre Transform
We introduce the Legendre transform which translates the Euler-Lagrange equations to the Hamilto-
nian form and study the Hamilton-Jacobi equation. Let us consider a Lagrangian system with the
extended phase space c
+
:= IR
N
IR
N
IR and the Lagrangian functional L C
2
(c
+
). We will
identify IR
N
with its dual space.
Denition 38.1 i) The Legendre transform corresponding to the Lagrangian L is the map of the ex-
tended phase space c
+
into itself which is dened by : (x, v, t) (x, p, t) with p := L
v
(x, v, t).
ii) The Legendre transform of the function L(x, v, t) on c
+
is the function on c
+
dened by (L)(x, p, t)
pv L(x, v, t) with (x, v, t) =
1
(x, p, t), if the Legendre map is a C
1
-dieomorphism c
+
c
+
.
Exercise 38.2 Prove that : c
+
c
+
is a C
1
-dieomorphism i the following inequality holds:
[L
vv
(x, v, t)[ , = 0, (x, v, t) c
+
.
Example 38.3 The inequality holds for the Lagrangian (36.14) because then the Jacobian matrix
J := L
vv
(x, v, t) is diagonal and [J[ = m
1
...m
n
,= 0 since all m
i
> 0.
Example 38.4 v
2
= p
2
/4, v
4
= 3v
4
= 3(p/4)
4/3
,...
Exercise 38.5 Prove that (L) = L if : c
+
c
+
is a C
1
-dieomorphism.
Theorem 38.6 Let the Legendre transform be a C
1
-dieomorphism c
+
c
+
. Then transforms
the Euler-Lagrange equations (36.15) into the Hamiltonian form,
x(t) = H
p
(x(t), p(t), t), p(t) = H
x
(x(t), p(t), t), (38.1)
where H(x, p, t) is the Legendre transform of the Lagrangian,
H(x, p, t) = pv L(x, v, t), p = L
v
(x, v, t). (38.2)
Proof The rst equation of (38.1) follows by dierentiation of the identity H(x, p, t) pv L(x, v, t):
H
p
= v +pv
p
L
x
x
p
L
v
v
p
= v = x (38.3)
since p = L
v
by denition, and x
p
= 0. The second equation of (38.1) follows from the Euler-Lagrange
equation (36.15):
H
x
= p
x
v +pv
x
L
x
L
v
v
x
= L
x
= p. (38.4)
since p = L
v
by denition, and p
x
= 0.
Remark 38.7 In (38.3) and (38.4) the derivatives L
x
mean the derivatives with xed v and t, but in
all other terms the derivatives in x mean the derivatives with xed p and t.
Example 38.8 For the Lagrangian (36.14) the energy has the form (36.18), hence H(x, p, t) =
i
p
2
i
2m
i
+V (x, t).
Exercise 38.9 Calculate the momentum, energy and the Hamilton function for the Lagrangian of the
relativistic particle
L(x, v) = c
2
1
_
v
c
_
2
, (x, v) IR
3
IR
3
. (38.5)
168
Solution:
p :=
v
1
_
v
c
_
2
, v =
p
2
+
_
p
c
_
2
, E =
c
2
1
_
v
c
_
2
, H = c
2
2
+
_
p
c
_
2
. (38.6)
38.2 Hamilton-Jacobi Equation
Let us consider the Lagrangian function L(x, v, t) C
2
(c
+
) with the corresponding Hamilton function
H(x, p, t) C
2
(c
+
) on the extended phase space c
+
:= IR
N
IR
N
IR, and the Cauchy problem of
the type
S(t, x) = H(x, S(t, x), t), (t, x) IR

N+1
S[
t=0
= S
0
(x), x IR
N
,
(38.7)
where S
0
(x) C
1
(IR
N
) is a given function. Let us describe the Hamilton-Jacobi method of the
construction of the solution to the problem (38.7).
First, consider the corresponding Cauchy problem for the Hamilton system,
x(t) = H
p
(x(t), p(t), t), p(t) = H
x
(x(t), p(t), t)
x[
t=0
= x
0
, p[
t=0
= S
0
(x
0
)
(38.8)
with x
0
IR
N
. Let us denote the solution by (x(t, x
0
), p(t, x
0
)). The solution exists and is C
1
-smooth
for small [t[ depending on x
0
. Let us dene the function o by the action integral
o(t, x
0
) = S
0
(x
0
) +
_
t
0
L(x(s, x
0
), x(s, x
0
), s)ds (38.9)
for x
0
IR
N
and small [t[. At last, let us express x
0
in x(t, x
0
) for small [t[: this is possible since the
Jacobian x
x
0
(t, x
0
) = E for t = 0. Thus, x
0
= x
0
(t, x) where x
0
(t, x) C
1
(IR IR
N
), hence we can
dene
S(t, x) = o(t, x
0
(t, x)), x IR
N
(38.10)
for small [t[.
Theorem 38.10 Let T > 0 and the map x
0
x(t, x
0
) be a C
1
-dieomorphism of IR
N
for t [0, T].
Then the function S(t, x) from (38.10) is the unique solution to the Cauchy problem (38.7) for t
[0, T].
Proof The theorem follows from the properties of the dierential 1-form
1
= pdx Hdt in IR
N+1
called the Poincare-Cartan integral invariant [2].
Step i) Let us consider x
0
, x
0
+ x
0
IR
N
and , + [0, T]. Let /
denote the following

two-dimensional submanifold in the extended phase space c
+
,
/
= x(t, x
0
+sx
0
), x(t, x
0
+sx
0
), t) : s [0, 1], t [0, +s]. (38.11)
The boundary /
is the union
1

0
, where
:= (x
0
+sx
0
, S
0
(x
0
+sx
0
), 0) : s [0, 1],
:= (x(t +s, (x
0
+sx
0
)), p(t +s, (x
0
+sx
0
)), t +s) : s [0, 1],
0
:= (x(t, x
0
)), p(t, x
0
), t) : t [0, ],
1
:= (x(t, x
0
+ x
0
)), p(t, x
0
+ x
0
), t) : t [0, + ]
(38.12)
38. HAMILTON MECHANICS 169
are oriented according to the increment of the parameters s, t. Therefore, by the Stokes theorem,
_
M
d
1
=
_
1
+
_
1
. (38.13)
Step ii) The central point of the proof is the observation that the restriction of the form d
1
onto
the submanifold /
vanishes,
d
1
[
M
= 0. (38.14)
This follows from two facts: i) the Hamilton vector eld H := (H
p
, H
x
, 1) in the extended phase
space c
+
is tangent to /
at every point, and ii) d

1
(H, 1) 0 for every vector eld 1 in c
+
. The
last identity follows from the following calculations: rst, d
1
= dpdxdHdt = dpdx(H
p
dp+
H
x
dx) dt is the antisymmetric bilinear form with the matrix
A =
_
_
_
0 E H
x
E 0 H
p
H
x
H
p
0
_
_
_, (38.15)
where E is the N N identity matrix. Second, AH 0, hence
d
1
(H, 1) = AH, 1) 0. (38.16)
Step iii) Now (38.13) reads
_
(pdx Hdt) =
_
S
0
(x)dx +
_
1
Ldt
_
0
Ldt, (38.17)
since dt[
= 0, p[
= S
0
(x) and
1
[
i
= Ldt. The rst term on the RHS of (38.17) is equal to
S
0
(x
0
+ x
0
) S
0
(x
0
). Therefore, (38.17) becomes by (38.9), (38.10),
_
(pdx Hdt) = S
0
(x
0
+ x
0
) +
_
1
Ldt (S
0
(x
0
) +
_
0
Ldt)
= S(x + x, + ) S(x, ), (38.18)
where x + x = x( + , x
0
+ x
0
) and x = x(, x
0
). Finally, (38.18) implies

S(t, x) = H(x, p, t)
and S(t, x) = p, hence Equation (38.9) follows.
170
39 Theory of Noether Currents
We consider a generalization of the symmetry theory to Lagrangian elds. Let the corresponding
Lagrangian density be invariant with respect to a symmetry group. Then the Noether free divergent
currents and the corresponding invariants can be constructed. We prove the Noether theorem providing
the construction.
39.1 Field Symmetry
Consider the one-parametric group of transformations g
s
: IR
4
IR
N
IR
4
IR
N
of the form
g
s
:
_
x
_
y
s
_
=
_
a
s
(x)
b
s
()
_
s IR (39.1)
where a
s
and b
s
are some dierentiable transformations a
s
: IR
4
IR
4
and b
s
: IR
N
IR
N
, respec-
tively. Let us dene the corresponding transformations of the elds
(x)
s
(y) := b
s
((x)). (39.2)
This denition implies a corresponding transformation for the derivatives: by the chain rule,
(x)
y
s
(y) :=
b
s
((x))(x)
x(y)
y
. (39.3)
Remark 39.1 At s = 0 all transformations are identities since g
0
= Id for the group g
s
(cf. (37.4)):
(a
0
x, b
0
) = (x, ), (x, ) IR
4
IR
N
. (39.4)
Denition 39.2 The transformation g
s
, s IR, is a symmetry of a Lagrangian eld with the La-
grangian density L(x, , ) if the following identity holds,
L(x, , ) = L(y,
s
,
y
s
)
y(x)
x
, (x, , ) IR
d+1
IR
N
IR
4N
, (39.5)
where y := a
s
(x),
s
:= b
s
() and
y
s
:=
b
s
()
x(y)
y
.
Example 39.3 Time-translations Consider the time translation along e
0
= (1, 0, 0, 0),
g
s
:
_
x
_
y
s
_
=
_
x se
0
s IR. (39.6)
Then
s
(y) = (y +se
0
),
y
s
(y) = ()(y +se
0
),
y(x)
x
1. (39.7)
Hence, (39.5) for this transformation is equivalent to (3.5).
Example 39.4 Space-translations Consider the space-translation along e
1
= (0, 1, 0, 0),
g
s
:
_
x
_
y
s
_
=
_
x +se
1
s IR. (39.8)
Then
s
(y) = (y se
1
),
y
s
(y) = ()(y se
1
),
y(x)
x
1. (39.9)
Hence, (39.5) for this transformation means that the Lagrangian density L does not depend on x
1
.
39. THEORY OF NOETHER CURRENTS 171
Example 39.5 Space-rotations Consider the group
g
s
:
_
(x
0
, x)
_
y
s
_
=
_
(x
0
, R
n
(s)x)
s IR, (39.10)
where R
n
(s) is the rotation of IR
3
around e
n
with an angle of s radian. Then
s
(y)=(y
0
, R
n
(s)y),
y
s
(y)=()
_
1 0
0 R
n
(s)
_
,
y(x)
x
1. (39.11)
Example 39.6 Phase rotations For C
M
dene
g
s
:
_
x
_
y
s
_
=
_
x
e
is
s IR. (39.12)
Then
s
(y) = e
is
(y),
y
s
(y) = e
is
(y),
y(x)
x
1. (39.13)
39.2 Noether Current and Continuity Equation
Denition 39.7 For a given one-parametric group g
s
of transformations (39.1) and a given trajectory
(x), let us dene the vector elds
v(x) =
a
s
x
s
s=0
, w(x) =

s
(x)
s
s=0
, x IR
d+1
. (39.14)
Denition 39.8 The Noether current corresponding to a given one-parametric group g
s
of transfor-
mations (39.1) is the following vector eld
S
(x)=
(x)w(x)+L(x, (x), (x))v
(x), x IR
d+1
, = 0, ..., 3. (39.15)
Theorem 39.9 (E.Noether [76]) Let g
s
be a one-parametric symmetry group, i.e. (39.5) holds for
s IR. Let (x) C
2
(IR
4
, IR
N
) be a solution to the equations (2.9), and w(x) C
1
(IR
4
, IR
N
),
v(x) C
1
(IR
4
, IR
4
) are dened by (39.14). Then the continuity equation holds,
(x) = 0, x IR
4
. (39.16)
Corollary 39.10 Let i) all conditions of Theorem 39.9 hold, ii) the bounds (2.20) hold, iii) (x) C
2
,
w(x) C
1
with a > 3/2, and v(x) C

1
0
. Then the conservation law holds,
o
0
(t) :=
_
IR
3
S
0
(t, x)dx = const, t := x
0
IR. (39.17)
Proof We have o
0
(t) = lim
R
o
R
0
(t) where
o
R
0
(t) :=
_
|x|R
S
0
(t, x)dx, t IR. (39.18)
Dierentiating, we get by (39.16) and the Stokes theorem,
o
R
0
(t) :=
_
|x|R
(t, x)dx =
_
|x|=R
n
(x)S
(t, x)d, t IR, (39.19)

where n
(x) := x
/[x[ and d is the Lebesgue measure on the sphere [x[ = R. Therefore,

o
R
0
(t) 0
as R since S
(x) C
1
with > 3. Hence (39.17) follows.

172
Remark 39.11 The integral identity (39.19) means that the vector eld S
k
(t, x) is the current density
of the eld S
0
(t, x).
Proof of Theorem 39.9 (cf. [37, 103]) Consider an arbitrary open region IR
3
with a smooth
boundary. Integrating the symmetry condition (39.5) over , we get
_
L(x, (x), (x))dx =

_
s
L(y,
s
(y),
y
s
(y))dy, s IR, (39.20)
where
s
:= a
s
(). Let us make the change of variables y = a
s
(x) on the RHS. Then we get the
identity
_
L(a
s
(x), b
s
((x)), D
s
(x))I
s
(x)dx = const, s IR, (39.21)
where I
s
(x) :=
a
s
(x)
x
and
[D
s
(x)]
:=

s
(y)
y
y=a
s
(x)
, = 0, ..., d. (39.22)
Dierentiating (39.21) in s, we get by (39.4),
_
_
L
x
(x, (x), (x))
d
ds
s=0
a
s
(x)+L
(x, (x), (x))

d
ds
s=0
b
s
((x))
+L
(x, (x), (x))

d
ds
s=0
[D
s
(x)]
+L(x, (x), (x))

d
ds
s=0
I
s
((x))
_
dx=0. (39.23)
Let us calculate the four derivatives in s.
i) By Denition (39.14),
d
ds
s=0
a
s
(x) = v(x). (39.24)
ii) By Denition (39.2), the chain rule and (39.14), (39.4),
d
ds
s=0
b
s
((x)) =
d
ds
s=0
s
(a
s
x) =
d
ds
s=0
[
s
(a
0
x) +
0
(a
s
x)] = w(x) +(x) v(x). (39.25)
iii) By denition, [D
s
(x)]
:=
b
s
((x))
[a
s
(x)]
=

s
(a
s
(x))
[a
s
(x)]
. Hence the same arguments imply

d
ds
s=0
[D
s
(x)]
=
d
ds
s=0
s
(a
s
(x))
[a
s
(x)]
=
d
ds
s=0
_
s
(a
0
(x))
[a
0
(x)]
+

0
(a
s
(x))
[a
0
(x)]
+

0
(a
0
(x))
[a
s
(x)]
_
=
w(x) +
((x)v(x)) +
d
ds
s=0
(x)
[a
s
(x)]
. (39.26)
To calculate the last derivative, let us use the matrix identity
(x)
[a
s
(x)]
[a
s
(x)]
=
(x)
x
. (39.27)
39. THEORY OF NOETHER CURRENTS 173
Dierentiating in s, we get by (39.4),
d
ds
s=0
(x)
[a
s
(x)]
+
(x)
x
= 0. (39.28)
Therefore, the last derivative in (39.26) equals
v. Hence (39.26) becomes,

d
ds
s=0
[D
s
(x)]
w(x) +(
(x)) v(x). (39.29)

iv) Finally, the derivative of the determinant I
s
of the Jacobian matrix is the trace of the derivative,
d
ds
s=0
I
s
((x)) =
d
ds
s=0
a
s
(x)
x
= tr
d
ds
s=0
a
s
(x)
x
= tr
v(x)
x
= v(x), (39.30)
since the Jacobian matrix is diagonal:
a
0
(x)
x
= E.
Collecting all calculations i) iv) in (39.23), we get
_
_
L
x
(x, (x), (x)) v(x) + L
(x, (x), (x)) (w(x) +(x) v(x))

+
(x) (
w(x) +(
(x)) v(x))
+L(x, (x), (x)) v(x)
_
dx = 0. (39.31)
Since the region is arbitrary, the integrand is zero by the main lemma of the calculus of variations.
We rewrite it as follows,
L
w(x) +
(x)
w(x) +
_
L(x, (x), (x))v(x)
_
= 0. (39.32)
Finally, let us substitute L
(x) from the Euler-Lagrange equations (2.9). Then (39.32)

becomes
(x) w(x)
_
+
_
L(x, (x), (x))v(x)
_
= 0, (39.33)
which coincides with (39.16) by (39.15).
Remark 39.12 The justication of the formal proof (39.26) (39.28) of (39.29) follows from the
identity of type (39.27),
[D
s
(x)]
[a
s
(x)]
=
(a
s
x)
x
, (39.34)
by dierentiation similar to (39.28) and (39.26).
174
40 Application to Four Conservation Laws
We obtain the formulas for Noether currents corresponding to concrete symmetry groups: time- and
space translations, space- and phase rotations. The formulas imply, by the Noether theorem, the
corresponding conservation laws for energy, momentum, angular momentum and charge. We consider
rst general Lagrangian elds and then specify the formulas for the Klein-Gordon and Schrodinger
equations.
40.1 General Lagrangian Fields
Let us apply the Noether theorem 39.9 to the four groups of the examples 39.3 39.6.
I. Proof of Theorem 3.4 For the group (39.6), Denition (39.14) implies by (39.7),
v(x) = e
0
, w(x) = (x)e
0
=
0
(x), x IR
4
. (40.1)
Hence, the Noether current (39.15) becomes,
_
S
0
(x) =
0
(x)
0
(x)L(x, (x), (x)),
S
k
(x) =
k
(x)
0
(x), k = 1, 2, 3.
(40.2)
The identity (3.5) implies that the Lagrangian density satises the invariance condition (39.5) with the
group (39.6). Therefore, Theorem 39.9 implies the continuity equation (39.16) for the current (40.2),
and Corollary 39.10 implies (39.17), which means the conservation of energy by Denition (3.1).
II. Proof of Theorem 3.8 Let us consider the case n = 1 for concreteness. For the group (39.8),
Denition (39.14) implies by (39.9),
v(x) = e
1
, w(x) = (x)e
1
=
1
(x), x IR
4
. (40.3)
_
_
S
0
(x) =
0
(x)
1
(x),
S
1
(x) =
1
(x)
1
(x)+L(x, (x), (x)),
S
k
(x) =
k
(x)
1
(x), k = 2, 3.
(40.4)
The Lagrangian density satises the invariance condition (39.5) with the group (39.8). Therefore,
Theorem 39.9 implies the continuity equation (39.16) for the current (40.4), and Corollary 39.10
implies (39.17), which means the conservation of the rst component of momentum by Denition
(3.16).
III. Proof of Theorem 3.13 For the group (39.10), Denition (39.14) implies by (39.11) and (35.31),
v(x) = (0, e
n
x), w(x) = (x)(0, e
n
x) = (x
x
)
n
(x), x IR
4
. (40.5)
_
S
0
(x) =
0
(x)(x
x
)
n
(x),
S
k
(x) =
k
(x)(x
x
)
n
(x) +Le
n
x, k = 1, 2, 3.
(40.6)
The identity (3.21) implies that the Lagrangian density satises the invariance condition (39.5) with
the group (39.10). Therefore, Theorem 39.9 implies the continuity equation (39.16) for the current
(40.6), and Corollary 39.10 implies (39.17), which means the conservation of the n-th component of
40. APPLICATION TO FOUR CONSERVATION LAWS 175
angular momentum by Denition (3.23).
IV. Proof of Theorem 3.18 For the group (39.12), Denition (39.14) implies by (39.13),
v(x) = 0, w(x) = i(x), x IR
4
. (40.7)
Hence, the components of the Noether current (39.15) equal the charge-current densities,
S
(x)=
(x) i(x), = 0, ..., 3. (40.8)

The identity (3.24) implies that the Lagrangian density satises the invariance condition (39.5) with
the group (39.12). Therefore, Theorem 39.9 implies the continuity equation (39.16) for the current
(40.8), and Corollary 39.10 implies (39.17), which means the conservation of charge by Denition
(3.25).
40.2 Klein-Gordon Equation
Let us substitute Expressions (2.16) and (2.5) into (40.2), (40.4), (40.6) and (40.8). Then Theorem
39.9 implies for solutions to Equation (2.1):
I. Energy ux The continuity equation (39.16) holds for the energy- and energy current densities
_
_
S
0
(x) = ih
(ih
e
c
(x))(x)
0
(x)L(x, (x), (x))
=
(ih
e
c
(x))(x) (ih
0
+
e
c
(x))(x)
2
,
+
3
k=1
[(ih
e
c
A
k
(x))(x)[
2
2
,
S
k
(x) = ih
(ih
e
c
A
k
(x))(x)
0
(x), k = 1, 2, 3,
(40.9)
if the potentials (x), A(x) do not depend on time x
0
= ct.
For the free equation (2.10):
_
_
S
0
(x) =
[
0
(x)[
2
2
+
3
k=1
[
k
(x)[
2
2
+
2
[(x)[
2
2
,
S
k
(x) =
k
(x)
0
(x), k = 1, 2, 3.
(40.10)
176
II. Momentum ux The continuity equation (39.16) holds for the rst components of the momentum-
and momentum current densities
_
_
S
0
(x) = ih
(ih
e
c
(x))(x)
1
(x),
S
1
(x) = ih
(ih
e
c
A
1
(x))(x)
1
(x) +L(x, (x), (x))
=
[(ih
e
c
(x))(x)[
2
2
+
(ih
e
c
A
1
(x))(x) (ih
1
+
e
c
A
1
(x))(x)
2
+
3
k=2
[(ih
e
c
A
k
(x))(x)[
2
2
,
S
k
(x) = ih
(ih
e
c
A
k
(x))(x)
1
(x), k = 2, 3,
(40.11)
if the potentials (x), A(x) do not depend on x
1
.
_
_
S
0
(x) =
0
(x)
1
(x),
S
1
(x) =
[
0
(x)[
2
2
+
[
1
(x)[
2
2

k=2,3
[
k
(x)[
2
2

2
[(x)[
2
2
,
S
k
(x) =
k
(x)
1
(x), k = 2, 3.
(40.12)
III. Space rotations The continuity equation (39.16) holds for the n-th component of the angular
momentum- and angular momentum current densities
_
_
S
0
(x) = ih
(ih
e
c
(x))(x) (x
x
)
n
(x),
S
k
(x) = ih
(ih
e
c
A
k
(x))(x) (x
x
)
n
(x) +L(e
n
x)
k
, k = 1, 2, 3,
(40.13)
if Eq. (3.22) holds for the density (2.5).
_
_
S
0
(x) =
0
(x)(x
x
)
n
(x),
S
k
(x) =
k
(x) (x
x
)
n
(x) +L(e
n
x)
k
, k = 1, 2, 3.
(40.14)
IV. Phase rotations The continuity equation (39.16) holds for the charge- and charge current
densities
_
_
S
0
(x) = ih
(ih
e
c
(x))(x) i(x),
S
k
(x) = ih
(ih
k

e
c
A
k
(x))(x) i(x), k = 1, 2, 3.
(40.15)
S
0
(x)=
0
(x) i(x), S
k
(x)=
k
(x) i(x), k = 1, 2, 3. (40.16)
40. APPLICATION TO FOUR CONSERVATION LAWS 177
40.3 Schrodinger Equation
Let us substitute Expressions (2.17) and (2.6) into (40.2), (40.4), (40.6) and (40.8). Then Theorem
39.9 implies for solutions to Equation (2.2):
I. Energy ux The continuity equation (39.16) holds for the energy- and energy current densities
_
_
S
0
(x) = ih
(x)
0
(x)L(x, (x), (x))
= e(x)(x) (x) +
1
2
3
k=1
[(ih
k

e
c
A
k
(x))(x)[
2
S
k
(x) =
1
m
ih
(ih
k

e
c
A
k
(x))(x)
0
(x), k = 1, 2, 3,
(40.17)
if the potentials (x), A(x) do not depend on time x
0
= t.
_
_
S
0
(x) =
1
2
3
k=1
[
k
(x)[
2
,
S
k
(x) =
1
m
k
(x)
0
(x), k = 1, 2, 3.
(40.18)
II. Momentum ux The continuity equation (39.16) holds for the rst component of the momentum-
and momentum current densities
_
_
S
0
(x) = ih
(x)
1
(x),
S
1
(x) =
1
m
ih
(ih
e
c
A
1
(x))(x)
1
(x) +L(x, (x), (x))
= e(x)(x) (x)+
1
2
(ih
e
c
A
1
(x))(x) (ih
1
+
e
c
A
1
(x))(x)
1
2
3
k=2
[(ih
k

e
c
A
k
(x))(x)[
2
,
S
k
(x) =
1
m
ih
(ih
k

e
c
A
k
(x))(x)
1
(x), k = 2, 3,
(40.19)
if the potentials (x), A(x) do not depend on x
1
.
_
_
S
0
(x) = i(x)
1
(x),
S
1
(x) = i
0
(x) (x) +
1
2
[
1
(x)[
2
1
2
k=2,3
[
k
(x)[
2
,
S
k
(x) =
1
m
k
(x)
1
(x), k = 2, 3.
(40.20)
178
III. Space rotations The continuity equation (39.16) holds for the n-th component of the angular
momentum- and angular momentum current densities
_
_
S
0
(x) = ih
(x) (x
x
)
n
(x),
S
k
(x) =
1
m
ih
(ih
e
c
A
k
(x))(x) (x
x
)
n
(x)+L(e
n
x)
k
, k = 1, 2, 3
(40.21)
if Eq. (3.22) holds for the density (2.6).
_
_
S
0
(x) = i(x) (x
x
)
n
(x),
S
k
(x) =
1
m
k
(x) (x
x
)
n
(x) +L(e
n
x)
k
, k = 1, 2, 3.
(40.22)
IV. Phase rotations The continuity equation (39.16) holds for the charge- and charge current
densities
S
0
(x)= h
(x) (x), S
k
(x)=
1
m
h
(ih
k
+
e
c
A
k
(x))(x) (x), k = 1, 2, 3. (40.23)
S
0
(x)= (x) (x), S
k
(x)=
1
m
i
k
(x) (x), k = 1, 2, 3. (40.24)
41. CAUCHY PROBLEM FOR MAXWELL EQUATIONS 179
41 Cauchy Problem for Maxwell Equations
We prove the existence of dynamics for inhomogeneous Maxwell equations and construct an integral
representation for the solutions.
Let us consider the Cauchy problem for the Maxwell equations (4.1) with the initial conditions
E[
t=0
= E
0
(x), B[
t=0
= B
0
(x), x IR
3
. (41.1)
We assume (E
0
(x), B
0
(x)) L
2
L
2
, where L
2
= L
2
(IR
3
) IR
3
, (t, x) C(IR, L
2
(IR
3
)), j(t, x)
C(IR, L
2
) and also (E(t, x), B(t, x)) C(IR, L
2
L
2
). Then the system (4.1) leads to the identities
div E
0
(x) = 4(0, x) div B
0
(x) = 0, x IR
3
, (41.2)
which are necessary constraints for the existence of solutions to the overdetermined system (4.1).
Theorem 41.1 Let E
0
(x), B
0
(x) and (t, x), j(t, x) satisfy all conditions mentioned above and the
constraints (4.2) and (41.2). Then
i) The Cauchy problem (4.1), (41.1) has a unique solution (E(t, x), B(t, x)) C(IR, L
2
L
2
).
ii) Let j(t, x) 0. Then the energy is conserved:
_
IR
3
[E
2
(t, x) +B
2
(t, x)]dx = const, t IR
3
. (41.3)
iii) The convolution representation holds
_
E(t)
B(t)
_
= M(t)
_
E
0
B
0
_
+ 4
_
t
0
G(t s)
_
_
(s)
1
c
j(s)
_
_
ds, t IR, (41.4)
where E(t) := E(t, ) etc, and M(t) resp. G(t) is 6 6 resp. 6 4 matrix-valued distribution concen-
trated on the sphere [x[ = [t[, for every xed t IR:
M(t)(x) = 0, G(t, x) = 0, for [x[ , = [t[. (41.5)
Proof ad i) We introduce the complex eld C(t, x) = E(t, x) +iB(t, x) and rewrite (4.1) as
1
c
C(t, x) = i rot C(t, x)

4
c
j(t, x), C[
t=0
= C
0
(x), (41.6)
div C(t, x) = 4(t, x), (41.7)
where C
0
(x) = E
0
(x) +iB
0
(x). Fourier transformation

C(k, t) =
_
exp(ik x)C(t, x)dx leads to the
equations
C(t, k) = c m(k)
C(t, k) 4
j(t, k),

C[
t=0
=

C
0
(k), (41.8)
ik

C(k, t) = 4 (k, t), (41.9)
where m(k) denotes the 3 3 skew-adjoint matrix of the operator k in C
3
. The solution

C(t, k)
is dened uniquely from the rst equation (41.8) of the overdetermined system (41.8), (41.9),
C(t, k) = exp(c m(k)t)
C
0
(k) 4
_
t
0
exp(c m(k)(t s))
j(s, k)ds, k IR
3
. (41.10)
180
We still have to show that (41.10) satises the constraint (41.9). Indeed, the Fourier transformed
equations (41.2), (4.2) are
ik

C
0
(k) = 4 (0, k), k IR
3
, (41.11)
(t, k) ik
j(t, k) = 0, k IR
3
, t IR. (41.12)
With S(t, k) = 4 (t, k) +ik

C(t, k) they imply by (41.8)
S(k, 0)=4 (k, 0)+ik

C
0
(k)= 0,

S(t, k)=4
(t, k)4ik
j(t, k)=0, k IR
3
, (41.13)
since k m(k)
C(t, k) = 0. Therefore, S(t, k) = 0 which means (41.9). Since m(k) is a skew-adjoint

matrix, its exponent exp(c m(k)t) is unitary. Now i) follows from (41.10).
ad ii) The Parseval identity implies
_
IR
3
[E
2
(t, x) +B
2
(t, x)]dx =
_
IR
3
[C(t, x)[
2
dx = (2)
3
_
IR
3
[
C(t, k)[
2
dk. (41.14)
Therefore, (41.3) follows from (41.10) since j(t, k) 0 and exp(c m(k)t) is a unitary matrix.
ad iii) We have to transform (41.10) back to position space in order to check (41.5). We have
m = m(k) = k, m
2
= k
2
+[k >< k[, m
3
= [k[
2
m, . . .. Hence,
m
2j+1
= (1)
j
[k[
2j
m = (1)
j
m
[k[
[k[
2j+1
for j 0,
m
2j
= m
2j1
m = (1)
j1
[k[
2j2
m
2
= (1)
j
_
m
[k[
_
2
[k[
2j
, j 1,
which yields by Eulers trick for the exponential
exp( m(k)t) =
0
( mt)
n
/n! =
0
( mt)
2j
/(2j)! +
0
( mt)
2j+1
/(2j + 1)! (41.15)
= 1 +
_
m
[k[
_
2
(1 cos [k[t) +
m
[k[
sin [k[t = cos [k[t + m
sin [k[t
[k[
+ (1 cos [k[t)
[k >< k[
[k[
2
.
Let us denote by

K(t, k) = sin [k[t/[k[, m(t, k) =
t

K(t, k) + m

K(t, k), and

D(t, k) = 1 cos [k[t.
Then we nally obtain,
exp(c m(k)t) = m(ct, k) +[k >
D(ct, k)
[k[
2
< k[. (41.16)
Inserting this into (41.10) and using the constraints (41.11) and (41.12), we get
C(t, k) = m(ct, k)
C
0
(k) + 4i[k >
D(ct, k)
[k[
2
(k, 0) (41.17)
4
_
t
0
[ m(c(t s), k)
j(s, k) i[k >
D(c(t s), k)
[k[
2
(s, k)]ds, (41.18)

which through integration by parts becomes
C(t, k) = m(ct, k)
C
0
(k) 4c
_
t
0
[ m(c(t s), k)
j(s, k) i[k >

t

D(c(t s), k)
[k[
2
(s, k)]ds. (41.19)
41. CAUCHY PROBLEM FOR MAXWELL EQUATIONS 181
Using
t

D(t, k) = [k[ sin[k[t = [k[
2

K(t, k), we get the inverse Fourier transforms,
m(t, x) := F
1
kx
m(t, k) =
t
K(t, x) i rot K(t, x), (41.20)
g(t, x) := F
1
kx
_
i[k >

K(t, k), m(t, k)
_
= (K(t, x), m(t, x)), (41.21)
where K(t, x) denotes the Kirchho kernel
K(t, x) := F
1

K(t, k) =
1
4t
([x[ [t[). (41.22)
With these notations, (41.19) implies (41.4) in the complex form
C(t) = m(ct) C
0
+ 4
_
t
0
g(c(t s))
_
c(s)
j(s)
_
ds t IR. (41.23)
Separating into real and imaginary parts, we obtain
E(t, x) = E
(r)
(t, x) +E
(0)
(t, x), B(
t, x) = B
(r)
(t, x) +B
(0)
(t, x), (41.24)
where we denote
_
E
(0)
(t)
B
(0)
(t)
_
=
_

t
K(ct) rot K(ct)
rot K(ct)
t
K(ct)
_
_
E
0
B
0
_
, (41.25)
and
_
E
(r)
(t)
B
(r)
(t)
_
= 4
_
t
0
_
K(c(t s))
t
K(c(t s))
0 rot K(c(t s))
_
_
c(s)
j(s)
_
ds. (41.26)
Here K(t, x) coincides with the Kirchho kernel
K(t, x) = F
1
kx
K(t, k) =
1
4t
([x[ [t[). (41.27)
Now (41.4) and (41.5) follow immediately. 2
Remark The formula (41.26) coincides with standard Lienard-Wiechert representation of the re-
tarded elds E
(r)
(t, x) and B
(r)
(t, x) through the Kirchho retarded potentials (t, x), A(t, x), [83]:
E
(r)
(t, x) = (t, x)

A(t, x), B
(r)
(t, x) = rot A(t, x),
(t, x) =
_
d
3
y
()
4[x y[
(y, ), A(t, x) =
_
d
3
y
()
4[x y[
j(y, ),
(41.28)
where = t [xy[ is the retarted time, and () is the Heaviside step function. We emphasize, that
(E
(r)
(t, x), B
(r)
(t, x)) is not a solution to Maxwell equations (41.1) with prescribed (t, x) and j(t, x),
since E
(r)
[
t=0
= 0, and hence div E(t, x) = (t, x) is not satised at t = 0. For the same reason,
(E
(0)
(t, x), B
(0)
(t, x)) is not a solution to the Maxwell equations (41.1) with = 0, j = 0.
182
42 Lorentz Molecular Theory of Polarization and Magnetization
We analyze the macroscopic Maxwell eld generated by charge and current distributions of a molecule
at rest.
42.1 Constitutive Equations
The Maxwell equations (4.1) dene the electromagnetic eld generated by a given charge and current
distribution. On the other hand, the Lorentz equations (4.21), (4.22) dene the motion of charged
particles in a given Maxwell eld. However, the classical theory cannot explain the structure of
matter, i.e., the stability of particles, the constitution of atoms and molecules, solid states etc. This
is related to the fact that the coupled system (4.1), (4.21) (or (4.1), (4.21)) is not well dened for
point particles. Hence, we miss the correct dynamical equation for matter. In particular, we need an
additional hypothesis to get a satisfactory theory of matter in a Maxwell eld.
Such a theory has been constructed by Lorentz [73] to justify the transition from a microscopic
electron theory to the macroscopic Maxwell equations. First, the theory postulates that matter is
a collection of identical small cells called molecules. Second, it is necessary to introduce an additional
hypothesis concerning the molecular response to an external Maxwell eld. The state of a neutral
molecule is characterized by its dipole moment and magnetic moment. The parameters allow to de-
scribe the eld generated by the molecule, at large distances from the molecule, with high precision.
Hence, the parameters give a complete description of the molecular eld for any macroscopic obser-
vation. Indeed, observations can be made only at distances which are much larger than the size of a
molecule. Hence, it is sucient to specify the inuence of the external elds onto the parameters by
the corresponding constitutive equations.
We start with an analysis of the distant Maxwell eld generated by the charge and current distri-
butions of a molecule at rest.
42.2 Stationary Molecular Fields in Dipole Approximation
Let us denote by a > 0 the size of the molecule and choose the origin in its center, i.e. assume that
(t, y) = 0, j(t, y) = 0, [x[ > a, t IR. (42.1)
Let us assume that
(t, ) L
1
(IR
3
), j(t, ) L
1
(IR
3
) IR
3
, t IR. (42.2)
Static elds First, let us consider the static case when the densities do not depend on time. Then
Equations (4.6), (4.9) become the stationary Poisson equations and their solutions are the Coulomb
potentials
(x) =
_
(y)dy
[x y[
, A(x) =
1
c
_
j(y)dy
[x y[
, x IR
3
. (42.3)
Let us expand 1/[x y[ in a Taylor series for small [y[ a:
1
[x y[
=
1
_
x
2
+y
2
2yx
=
1
[x[
+
yx
[x[
3
+O
_
1
[x[
3
_
, [x[ . (42.4)
Then (42.3) becomes,
_
_
(x) =
Q
[x[
+
px
[x[
3
+O
_
1
[x[
3
_
A(x) =
J
c[x[
+
/x
[x[
3
+O
_
1
[x[
3
_
[x[ a, (42.5)
42. LORENTZ MOLECULAR THEORY OF POLARIZATION AND MAGNETIZATION 183
where we denote
Q =
_
(y)dy, p =
_
y(y)dy,
J =
_
j(y)dy, /
kl
=
1
c
_
j
k
(y)y
l
dy.
(42.6)
We will identify the molecular elds with the rst two terms in the expansions (42.5), since [x[/a 1
in all macroscopic observations.
Let us note that the remainder in (42.4) is O(y
2
). Therefore, the rst two terms in the expansions
(42.5) correspond to the following dipole approximations for (y) and j(y):
d
(y) = Q(y) p
y
(y), j
d
(y) = J(y) c/
y
(y). (42.7)
42.3 Non-Stationary Fields in Multipole Approximations
It is easy to see that an asymptotic behavior of the type (42.5) holds for the retarded Kirchho
potentials (43.4) generated by non-stationary localized densities satisfying (42.1):
(t, x) =
_
(t [x y[/c, y)
[x y[
dy, A(t, x) =
1
c
_
j(t [x y[/c, y)
[x y[
dy, (t, x) IR
4
. (42.8)
For this purpose, let us continue the Taylor expansion (42.4) and obtain a complete expansion of the
type (42.5), including all negative powers of [x[:
1
[x y[
=
1
[x[
+
yx
[x[
3
+
||2
y
(n)
[x[
||+1
, (42.9)
where we denote n := x/[x[ and P
is a polynomial. The expansion is a convergent series for [x[ > a.

Hence, the retarded potentials may be expressed by the converging series
_
_
(t, x) =
Q(t)
[x[
+
p(t)x
[x[
3
+
||2
(t)P
(n)
[x[
||+1
A(t, x) =
J(t)
c[x[
+
/(t)x
[x[
3
+
||2
A
(t)P
(n)
[x[
||+1
[x[ > a. (42.10)

These expansions correspond to the following multipole approximations for (t, y) and j(t, y):
_
m
(t, y) = Q(t)(y) p(t)
y
(y) +
||2
(t)
y
(y),
j
m
(t, y) = J(t)(y) c/(t)
y
(y) +
||2
j
(t)
y
(y).
(42.11)
The coecients are dened by
_
_
Q(t) =
_
(t [x y[/c, y)dy, p(t) =
_
y(t [x y[/c, y)dy,
J(t) =
_
j(t [x y[/c, y)dy, /
kl
(t) =
1
c
_
j
k
(t [x y[/c, y)y
l
dy,
(42.12)
(t) =
_
IR
3
(y)
!
(t [x y[/c, y)dy, j
(t) =
_
IR
3
(y)
!
j(t [x y[/c, y)dy. (42.13)
Let us justify the convergence of the series (42.11) in the sense of distributions.
184
Denition 42.1 The space H
a
(IR
3
) consists of test functions (y) which are real analytic in the ball
B
a
= y IR
3
: [y[ < a, and the Taylor series (y) =
!

()
(0) converges uniformly in B
a
.
Exercise 42.2 Check that the function
x
(y) := 1/[x y[ belongs to the space H
a
(IR
3
) for [x[ > a.
Proposition 42.3 Let (42.1) hold and (t, ), j(t, ) L
1
(IR
3
). Then the series (42.11) converges and
coincides with (t, y) or j(t, y) in the following sense:
_
_
(t, y), (y)) = Q(t)(y) p(t)
y
(y) +
||2
(t)
y
(y), (y))
j(t, y), (y)) = J(t)(y) c/(t)
y
(y) +
||2
j
(t)
y
(y), (y))
(42.14)
for every test function from the space H
a
(IR
3
) or H
a
(IR
3
) IR
3
, respectively.
Exercise 42.4 Prove Proposition 42.3. Hint: Substitute the Taylor expansion for and into the
LHS of (42.14) and use (42.1).
Corollary 42.5 The multipole approximations (42.11) satisfy the charge continuity equation

m
(t, y) +
y
j
m
(t, y) = 0, (t, y) IR
4
. (42.15)
Exercise 42.6 Prove the corollary. Hint: Use the identity
m
(t, y) +
y
j
m
(t, y), (t, y)) =
(t, y) +
y
j(t, y), (t, y)) for C
0
(IR
4
) such that (t, ),

(t, ) H
a
(IR
3
), t IR.
Substituting the series (42.11) into (42.15), we get
Q(t)(y) p(t)
y
(y) +J(t)
y
(y) +
||2
C
(t)
y
(y) = 0. (42.16)
Therefore, we have
Q(t) 0, J(t) p(t). (42.17)

42.4 Magnetic Moment of a Molecule
Let us consider a molecule in a stationary state, i.e. (t, y) (y) and j(t, y) j(y). Then the
multipole expansions (42.11) do not depend on time, i.e.,
_
m
(t, y)
m
(y) = Q(y) p
y
(y) +
||2
y
(y)
j
m
(t, y) j
m
(y) = J(y) c/
y
(y) +
||2
j
y
(y).
(42.18)
Proposition 42.7 Let the multipole charge-current densities (42.18) correspond to a stationary state
of the molecule. Then the matrix / is skewsymmetric, and
/
y
= m
y
, (42.19)
where m IR
3
is the magnetic moment of the molecule, i.e., of the current density j(y) in the
stationary state.
42. LORENTZ MOLECULAR THEORY OF POLARIZATION AND MAGNETIZATION 185
Proof Substituting (42.18) into (42.15), we get
y
j
m
(y) = 0, y IR
3
. (42.20)
Therefore, in particular J = 0 and
y
[/
y
(y)] = 0. Then /
kl
+/
lk
= 0 and the vector m IR
3
is dened by the following matrix identity:
/=
_
_
_
_
_
_
_
0 m
3
m
2
m
3
0 m
1
m
2
m
1
0
_
_
_
_
_
_
_
. (42.21)
Formulas (42.21) and (42.6) imply that
m =
1
2c
_
y j(y)dy. (42.22)
Remark 42.8 The integral does not depend on the choice of the origin since J:=
_
j(y)dy = 0 by
(42.20).
Adiabatic Condition Let us assume that the molecular dynamics can be described as an adiabatic
evolution of stationary states with corresponding dipole electric moment p(t) and magnetic moment
m(t).
Finally, let us assume that the molecule is neutral, i.e., Q = 0. We will identify (t, x) and j(t, x)
with the rst two terms on the RHS of (42.18). Then, by (42.17) and (42.19), we get the dipole
approximation
(t, x)
d
(t, x):=p(t)
x
(x), j(t, x)j
d
(t, x):= p(t) (x) +c
x
m(t)(x). (42.23)
Let us stress that the approximations are sucient for any macroscopic observation of the molecular
elds at distances much larger than the size of the molecules.
42.5 Macroscopic Limit: Maxwell Equations in Matter
Macroscopic Limit The total charge-current densities (t, x), (t, x) in matter are sums of contribu-
tions of all molecules concentrated at the points x
n
IR
3
. In the dipole approximation, the densities
are
_
mol
(t, x) =
n
p
n
(t)
x
(x x
n
)
j
mol
(t, x) =
n
_
p
n
(t) (x x
n
) +c
x
m
n
(t)(x x
n
)
_
.
(42.24)
Now let us consider the macroscopic limit, when the diameter a of a molecule converges to zero, and
each singular density converges to a limit distribution. More precisely, let us assume that for every
xed t we have the following asymptotics in the sense of distributions of x:
_
n
p
n
(t)(x x
n
) P(t, x)
n
p
n
(t)(x x
n
)

P(t, x),

n
m
n
(t)(x x
n
) M(t, x)
a 0. (42.25)
186
Then by the continuity of the dierentiation of distributions, we have in the limit a 0,
_
mol
(t, x)
x
P(t, x)
j
mol
(t, x)

P(t, x) +c
x
M(t, x)
. (42.26)
Denition 42.9 i) The vector functions P(t, x) and M(t, x) are called the electric polarization
and magnetization of the molecules at point x and time t, respectively.
ii) D(t, x) = E(t, x) + 4P(t, x) is called the dielectric displacement and H(t, x) = B(t, x)
4M(t, x) is called the magnetic eld intensity.
Let us separate the macroscopic and molecular charge and current densities:
(t, x) =
mac
(t, x) +
mol
(t, x), j(t, x) = j
mac
(t, x) +j
mol
(t, x), (42.27)
where the molecular densities are identied with the macroscopic limits (42.26). Let us substitute the
expressions (42.27), (42.26) into the Maxwell equations (4.1) and express the elds E, B in terms of
D, P, H, M. Then we obtain the Maxwell equations in matter:
_
_
div D(t, x)=4
mac
(t, x), rot E(t, x)=
1
c
B(t, x),
div B(t, x)=0, rot H(t, x)=
1
c
D(t, x)+
4
c
j
mac
(t, x)
(t, x) IR
4
. (42.28)
Constitutive Equations Equations (42.28) contain two additional unknown vector elds D, H.
Therefore, we need two additional vector equations. For isotropic materials they are the constitutive
equations
D(t, x) = E(t, x), B(t, x) = H(t, x), (42.29)
where is called the permittivity and is called the permeability of matter. The constitutive
equations are equivalent to
P(t, x) =
e
E(t, x), M(t, x) =
m
H(t, x), (42.30)
where
e
is called the electric susceptibility and
m
is called the magnetic susceptibility of
matter. The following relations follow from Denition 42.9 ii):
= 1 + 4
e
, = 1 + 4
m
. (42.31)
43. LONG-TIME ASYMPTOTICS AND SCATTERING 187
43 Long-Time Asymptotics and Scattering
The retarded potentials are particular solutions of the wave and Maxwell equations. Here we want to
explain the outstanding role of these potentials.
43.1 Retarded Potentials
Formula (41.26) may be rewritten as
E
(r)
(t, x) =
(r)
(t, x)
1
c
A
(r)
(t, x), B
(r)
(t, x) = rot A
(r)
(t, x), (43.1)
where the potentials are given by
(r)
(t, x) =
_
(t
ret
)
[x y[
(t
ret
, y)dy, A
(r)
(t, x) =
1
c
_
(t
ret
)
[x y[
j(t
ret
, y)dy, (43.2)
and t
ret
= t [x y[/c. Let us assume that the charge and current densities are continuous and
localized in space,
(t, x) = 0, j(t, x) = 0, [x[ > R, t IR. (43.3)
Then (43.2) for large t > 0 become the standard retarded potentials [53],
_
(r)
(t, x) =
ret
(t, x) :=
_
(t [x y[/c, y)
[x y[
dy
A
(r)
(t, x) = A
ret
(t, x) :=
1
c
_
j(t [x y[/c, y)
[x y[
dy
t > R +[x[. (43.4)

Further, the elds (43.1) become the retarded elds
_
E
(r)
(t, x) =
ret
(t, x)

A
ret
(t, x),
B
(r)
(t, x) = rot A
ret
(t, x),
t > R +[x[. (43.5)

The distinguished role of the particular retarded solutions (43.4) to the wave equations (4.6), (4.9) is
justied in scattering theory. In general, the solutions to the wave equations (4.6), (4.9) are dened
uniquely by the initial conditions at time zero:
[
t=0
=
0
(x),

[
t=0
=
0
(x), x IR
3
. (43.6)
A[
t=0
= A
0
(x),

A[
t=0
=
0
(x), x IR
3
. (43.7)
However, the asymptotic behavior of the solutions for t + and any xed point x does not
depend on the initial data
0
,
0
, /
0
,
0
and coincides with the retarded potentials (43.4).
Space-localized initial data For example let us consider the initial functions with compact supports.
Proposition 43.1 Let (43.3) hold, and let the initial functions
0
(x),
0
(x), A
0
(x),
0
(x) be contin-
uous and localized in space,
0
(x) =
0
(x) = 0, A
0
(x) =
0
(x) = 0, [x[ > R. (43.8)
Then for large time the solutions to the Cauchy problems (4.6), (43.6) and (4.9), (43.7) coincide with
the retarded potentials (43.4):
(t, x) =
ret
(t, x), A(t, x) = A
ret
(t, x), t > R +[x[. (43.9)
188
Proof Let us prove the proposition for the scalar potential (t, x). The Kirchho formula for the
solution reads
(t, x) =
1
4t
_
S
t
(x)
0
(y)dS(y) +
t
_
1
4t
_
S
t
(x)
0
(y)dS(y)
_
+
ret
(t, x), t > 0, x IR
3
, (43.10)
where S
t
(x) denotes the sphere y IR
3
: [x y[ = t and dS(y) is the Lebesgue measure on the
sphere. Now (43.9) follows from (43.8).
A similar theorem holds for the Maxwell equations.
Theorem 43.2 Let (43.3) and the conditions of Theorem 41.1 hold, and
E
0
(x) = B
0
(x) = 0, [x[ > R. (43.11)
Then
E(t, x) = E
ret
(t, x), B(t, x) = B
ret
(t, x) t > R +[x[. (43.12)
Proof (43.12) follows from (41.24) and (43.5) since E
(0)
, B
(0)
vanish for t > R+[x[ by (43.11), (41.25)
and (41.5).
Finite energy initial data For the Maxwell equations with j = 0 the energy is conserved, (41.3).
For the wave equations (4.6) and (4.9) with = 0 and j = 0 the energy conservations read (see (40.9)),
_
IR
3
[
1
c
2
[
(t, x)[
2
+[(t, x)[
2
]dx=const,
_
IR
3
[
1
c
2
[

A(t, x)[
2
+[A(t, x)[
2
]dx=const. (43.13)
For solutions with nite initial energy, (43.9) and (43.12) hold in the local energy semi-norms in the
limit t +.
Theorem 43.3 Let (43.3) hold and, further, (t, x), A(t, x), and E(t, x), B(t, x) are nite energy
solutions to the wave equations (4.6) and (4.9), or to the Maxwell Eqns (4.1), respectively. Then for
any R > 0,
_
_
_
|x|<R
[[

(t, x)

ret
(t, x)[
2
+[ (t, x)
ret
(t, x)[
2
]dx 0
_
|x|<R
[[

A(t, x)

A
ret
(t, x)[
2
+[A(t, x) A
ret
(t, x)[
2
]dx 0
_
|x|<R
[[E(t, x) E
ret
(t, x)[
2
+[B(t, x) B
ret
(t, x)[
2
]dx 0
t +. (43.14)
Proof Let us split the initial functions into two components: a rst, space-localized component
similar to (43.8), (43.11), and the rest. For the rst component we construct a solution to the non-
homogeneous equations, and for the rest we construct a solution to the homogeneous equations. Then,
for the rst component the convergence (43.14) follow from (43.9) and (43.12). The contribution of
the second component is uniformly small in time by energy conservation.
43.2 Limiting Amplitude Principle in Scattering Problems
Time-periodic source and spectral problem Let us consider a wave problem with an external
periodic source
(
2
t
)(t, x) = b(x)e
it
, (43.15)
where IR and the amplitude b(x) decays rapidly as [x[ (see for example (12.9)). We assume
c = 1 for simplicity. Similar problems arise in scattering problems for the Schrodinger equation
ih

(t, x) = H(t, x) +b(x)e
it
(43.16)
(see for example (13.2), (13.3)). Here H is the Schrodinger operator corresponding to a static Maxwell
eld with the potentials (t, x) (x) and A(t, x) A(x) (see (2.2)),
H :=
1
2m
(ih
e
c
A(x))
2
+e(x). (43.17)
Let us write Equations (43.15), (43.16) in a unied form,
i

(t, x) = A(t, x) +B(x)e
it
. (43.18)
For the wave equation (43.15) we set (t, x) := ((t, x),

(t, x)) and
A = i
_
0 1
0
_
, B(x) = i
_
0
b(x)
_
. (43.19)
The corresponding free equation reads
i

(t, x) = A(t, x). (43.20)
Let us denote by U(t) the corresponding dynamical group if it exists:
(t) = U(t)(0), (43.21)
where (t) := (t, ).
We consider nite energy solutions to Equations (43.18), (43.20). This means that (t, )
C(IR, E), where E is the corresponding Hilbert phase space. That is, E := L
2
(IR
3
) for the Schrodinger
equation (43.16), and E := ((x), (x)) : (x), (x) L
2
(IR
3
) is the Hilbert space with the norm
|((x), (x))|
2
E
:= |(x)|
2
+|(x)|
2
, (43.22)
where | | stands for the norm in L
2
(IR
3
).
Remark 43.4 U(t), t IR, is a unitary operator in the corresponding Hilbert space by energy con-
servation.
Let us look for a solution of the type
(x)e
it
to Eq. (43.18). Substituting, we get the Helmholtz
stationary equation,
(A)
(x) = B(x). (43.23)

Denition 43.5 i) The spectrum of Equation (43.18), specA, is the set of all C such that the
operator A+i is not invertible in the corresponding Hilbert space E.
ii) The resolvent of the operator A is dened by
R() := (A+i)
1
, C specA. (43.24)
Equation (43.23) admits a unique solution
(x) E for every B E if , specA:
= R()B, C specA. (43.25)

190
Exercise 43.6 Calculate specA for the free Schr odinger equation corresponding to (43.16) with h
= 1
and m = 1:
i

(t, x) = (t, x). (43.26)
Hint: Use the Fourier transform

(k) :=
_
e
ikx
(x)dx.
Solution: The operator A = becomes the multiplication operator

A(k) = k
2
, and the phase space
E = L
2
(IR
3
) E by the Parseval theorem. Therefore,

R(, k) is multiplication by (k
2
+)
1
. It is
bounded in

E i the function (k
2
+)
1
is bounded in IR
3
. Hence, specA = IR
= IR : 0.
Exercise 43.7 Calculate the resolvent R() corresponding to (43.26).
Hint: R() is a convolution with a fundamental solution c
(x) of the operator . Choose the

fundamental solution from the space of tempered distributions.
Solution: The tempered fundamental solution is unique,
c
(x) = c
(x) :=
e
|x|
4[x[
, C IR
, (43.27)
where we choose Re
> 0 for all C IR
.
Remark 43.8 i) The distribution c
(x) is tempered since its Fourier transform is a bounded function.

ii) The distribution c
+
(x) :=
e
|x|
4[x[
also is a fundamental solution, however, it is not tempered for
all C IR
.
iii) For < 0, both fundamental solutions, c
(x), are tempered.

Limiting amplitude principle The principle states (see [29]) the following long-time asymptotics
for a general class of solutions to (43.18) with (0, x) E:
(t, x) =
(x)e
it
+
l
C
l
l
(x)e
i
l
t
+r(t, x), (43.28)
where
(x) is a limiting amplitude,

l
(x) stand for the eigenfunctions of the discrete spectrum of
the operator A, and r(t, x) 0, t , in an appropriate norm.
For the wave equation (43.15), the asymptotics (43.28) follows immediately from the Kirchho formula
(43.10), if the initial functions decay rapidly. Let us assume, for example, that
[(0, x)[ +[(0, x)[ +[(0, x)[ = O(e
|x|
), [x[ , (43.29)
with an > 0. Then the rst and second terms on the RHS of (43.10) decay exponentially, like e
t
,
while the last is just proportional to e
it
. Indeed, substituting (t, x) =
1
4
b(x)e
it
and c = 1 into
(43.4), we get
ret
(t, x) :=
_
b(y)e
i(t|xy|)
4[x y[
dy = e
i(t)
(x), (x) =
_
b(y)e
i|xy|
4[x y[
dy. (43.30)
This gives (43.28) without the sum. This corresponds to the absence of the discrete spectrum for the
wave equation with constant coecients.
For the Schrodinger equation (43.16) the asymptotics (43.28) is proved now for a restricted class
of potentials. For example, it can be deduced from the Kato and Jensen results [57] for
(x) = O([x[
4
), A(x) 0, (43.31)
where > 0. The asymptotics holds for the initial function (0, x) from the Agmon spaces, H
=
(x) : (1 + [x[)
(1 ) L
2
(IR
3
) with > 1/2. Then the remainder r(t, x) 0 in a dual space
H
to H
.
Proposition 43.9 i) For the Schr odinger equations (43.16), (43.17), with potentials satisfying (43.31),
(0, x) H
, > 1/2, and ,=

l
, ,= 0, the asymptotics (43.28) holds and |r(t, )|
H
0, t .
ii) The limiting amplitude is given by (cf. (43.25))
= lim
0+
R( +i)B, IR, B H
, > 1/2, (43.32)

where the limit holds in the space H
.
Remarks 43.10 i) The limiting amplitude
E := L
2
(IR
3
) if , specA. The results [57] imply
that
for 0, though spec

c
A then, where spec
c
A stands for the continuous spectrum
of the operator A. However generally
, E: this is related to a slow decay of
(x) at innity,
[x[ , which physically means a radiation of electrons to innity.
ii) The Coulomb potential (x) = C[x[
1
does not t the bound (43.31).
iii) Formula (43.32) is called the limiting absorption principle since R( + i) = (A + i)
1
is
the resolvent of the damped equation
i

(t, x) = A(t, x) i(t, x) +B(x)e
it
, (43.33)
where the term i(t, x) describes an absorption of energy.
Let us sketch a formal proof of the proposition using the results [57]. Namely, the Duhamel represen-
tation for the solutions to the inhomogeneous equation (43.18) reads,
(t) = U(t)(0) i
_
t
0
U(t s)Be
is
ds, (43.34)
where U(t) := e
iAt
is the dynamical group of the corresponding free equation.
The results [57] imply that the rst term on the RHS admits an asymptotics (43.28) with
= 0.
This is a central achievement of the Kato-Jensen theory [57]: the contribution of the continuous
spectrum to the solution U(t)(0) of the homogeneous equation vanishes as t .
It remains to analyze the integral term. It can be rewritten as
I(t) = ie
it
_
t
0
U(s)Be
is
ds. (43.35)
Therefore, we have asymptotically,
I(t) = ie
it
_
t
0
e
iAs
Be
is
ds e
it
, t , (43.36)
since
i
_

0
e
iAs
Be
is
ds = i lim
0+
_

0
e
iAs
Be
i(+i)s
ds = lim
0+
(A i)
1
B =:
, (43.37)
where the last limit exists in the space H
by the results of [57] since ,=

l
and ,= 0.
Remark 43.11 Our arguments above demonstrate that
i) The term (x)e
it
on the RHS of (43.28) characterizes the response of a physical system to the
external periodic source B(x)e
it
, and it does not depend on the initial data.
ii) On the other hand, the sum over the discrete spectrum on the RHS depends only on the initial data
and does not depend on the external source.
Hence, in an experimental observation of a scattering problem, the response, in principle, can be singled
out if the experiment is concentrated on the eld which is built by the external source.
192
Part VIII
Exercises
193
44. EXERCISES FOR PART I 195
44 Exercises for Part I
44.1 Exercise 1: Main Lemma of the Calculus of Variations
Denition 44.1 C[a, b] is a space of continuous functions on the interval [a, b].
C
0
[a, b] := h(x) C[a, b] : h(a) = h(b) = 0.
Lemma 44.2 Let f(x) C[a, b] and
T
_
0
f(x)h(x) dx = 0 for every h(x) C[a, b]. Then f(x)
0.
a) Prove this lemma.
b) Prove the analogous lemma for h(x) C
0
[a, b].
Proof of a) Let there exist a point x
0
such that f(x
0
) ,= 0. Without loss of generality, we
assume f(x
0
) = a > 0. Since f(x) C[a, b], then for any > 0 there exists a > 0 such that
[f(x)a[ < for all x O
(x
0
) (x
0
, x
0
+). We choose = a/2. Then f(x) > a = a/2,
x O
(x
0
). Now we choose the function h(x) C[a, b] such that supp h(x) O
(x
0
) and
h(x) 0. For example, h(x) = 0 for [x x
0
[ and h(x) = [x x
0
[ for [x x
0
[ . Then
T
_
0
f(x)h(x) dx
x
0
+
_
x
0
f(x)h(x) dx >
a
2
x
0
+
_
x
0
h(x) dx > 0. 2
The Proof of a) covers the case b), because the function h(x) used in the proof h(x) C
0
[a, b].
44.2 Exercise 2: Euler-Lagrange Equations
a) Given a functional F : C
1
[a, b] IR,
F[y] :=
b
_
a
f(x, y(x), y
(x)) dx,
introduce a space
L[a, b] := y C
1
[a, b], y(a) = y
a
, y(b) = y
b
.
Show that for a y which minimizes F, the Euler-Lagrange equation
f
y
(x, y, y
)
d
dx
f
y
(x, y, y
) = 0
holds. Here f
y
:= f/y.
Proof: Let F(y
) := min
yL[a,b]
F(y). Let h(x) C
1
0
[a, b] := h(x) C
1
[a, b] : h(a) = h(b) = 0.
Fix such a function h(x). Introduce the function
S() := F[y
+h] =
b
_
a
f(x, y
(x) +h(x), y
(x) +h
(x)) dx, IR.

196
Since y
(x) + h(x) L[a, b] for any , S() S(0) = F(y
). Hence, min S() = S(0) and

S
(0) = 0.
S
()
=0
=
b
_
a
[f
y
(x, y
(x), y
(x))h(x) +f
y
(x, y
(x), y
(x))h
(x)] dx = 0.
Integrating by parts we obtain
b
_
a
[f
y
(x, y
(x), y
(x))
d
dx
f
y
(x, y
(x), y
(x))]h(x) dx = 0.
Exercise 1 implies f
y
(x, y
(x), y
(x))
d
dx
f
y
(x, y
(x), y
(x)) = 0.
b) Geodesic line: Use the Euler-Lagrange equations to calculate the minimum of the length
functional in the plane
L[y] =
b
_
a
_
1 +[y
(x)[
2
dx
with xed boundary conditions y(a) = y
a
, y(b) = y
b
.
Solution: f(x, y, y
) :=
_
1 +[y
(x)[
2
. Then f
y
= 0. Substituting into the Euler-Lagrange
equations we get
d
dx
f
y
(x, y(x), y
(x)) = 0.
Hence, f
y
(x, y(x), y
(x)) =
y
1+|y
(x)|
2
=const. Hence, y
=const, and y(x) = a +bx.

44.3 Exercise 3: Light Propagation in a Stratied Medium
Consider a medium where the index of refraction only depends on one coordinate y, n(y) =
1/v(y) (here v(y) is the speed of light.) The resulting system is eectively two-dimensional, i.e.,
the z coordinate may be suppressed.
a) Use the Fermat principle (minimization of the ight time)
T[y] := T
y(a),y(b)
:=
b
_
a
dl
v(y)
with xed starting point (x = a, y = y
a
) and end point (x = b, y = y
b
)) to prove the Snellius
law of refraction
n(y) sin (y) = const .
where (y) is the angle between the light ray and the y-axis.
b) Use the Snellius law to sketch the light rays of a wave guide with n(y) = 2 + cos y.
45. EXERCISES FOR PART III 197
c) Use the Snellius law to sketch the light rays for ordinary refraction n(y) = n
1
for y < 0, and
n(y) = n
2
for y > 0.
Proof of a) Since dl =
_
1 + (y
(x))
2
and n(y) = 1/v(y), we rewrite T[y] in the form
T[y] =
b
_
a
n(y)
_
1 + (y
(x))
2
dy.
Applying the Euler-Lagrange equations to f(x, y, y
) := n(y)
_
1 +[y
(x)[
2
, we get
n
(y)
_
1 +[y
(x)[
2
d
dx
n(y)y
_
1 +[y
(x)[
2
= 0.
Dierentiating in x we get
n
(y)
_
1 +[y
(x)[
2
n(y)
y
(1 +[y
(x)[
2
)
3/2
= 0.
This is equivalent to
d
dx
_
n(y)
1
_
1 +[y
(x)[
2
_
= 0.
Hence, n(y)
1
_
1 +[y
(x)[
2
= const. Since sin (y) =
1
_
1 + ctg
2
=
1
_
1 +[y
(x)[
2
, we get
n(y) sin (y) = const. 2
45 Exercises for Part III
45.1 Exercise 4: The Kepler Problem in 5 Easy Steps
The Kepler problem is dened by the two-particle Lagrangian
L =
m
1
2
x
2
1
+
m
2
2
x
2
2
+
m
1
m
2
[x
1
x
2
[
.
Solve it by going through the following steps:
a) Reduction to a one-particle problem (coordinate x) via separation of the center-of-mass coordi-
nate X.
b) Reduction to a two-dimensional problem by the observation that the plane spanned by (x(0), x(0))
is invariant.
c) Reduction to a one-dimensional problem (coordinate r) by the introduction of polar coordinates
(r, ). Show that Keplers third law holds, nd the eective potential of the one-dimensional
problem, and check for energy conservation.
d) Investigate the orbits of the one-dimensional problem.
e) Integrate the orbit equation to nd the possible orbits. Show that, depending on the values of
the integration constants (angular momentum, energy), the orbits are indeed ellipses, parabolae
or hyperbolae.
198
Solution:
a) Denote by X(t) the center of mass, X(t) :=
m
1
x
1
(t) +m
1
x
1
(t)
M
, where the total mass is M :=
m
1
+m
2
, and x := x
1
x
2
. Then
2
i=1
m
i
x
2
i
=
m
1
m
2
x
2
(t)
M
+M

X
2
.
Finally, we can rewrite the Lagrangian in the form
L =
m
1
m
2
x
2
2M
+
M

X
2
2
+
m
1
m
2
[x
1
x
2
[
= L
1
(x, x,

X).
Applying the Euler-Lagrange equation to the coordinate X
d
dt
L
L
X
= 0,
we get
d
dt
M

X = 0. Hence,
X = const, (45.1)
then X(t) = at +b, where a, b are some constants.
Remark 45.3 We can obtain Eqn (45.1) from the following arguments. Note that the potential
V (x
1
, x
2
) :=
m
1
m
2
[x
1
x
2
[
depends only on the dierence x
1
x
2
. Hence, V is invariant with respect to translations
T
s
: (x
1
, x
2
) (x
1
+hs, x
2
+hs), s IR, h IR
3
.
Hence, the total momentum (see Section 3.2) is conserved, p :=
2
i=1
m
i
x
i
= const, and the center
of mass moves uniformly, X(t) = At +B.
Finally, we get that it suces to consider the Lagrangian
L
1
L
1
(x, x) :=
m
1
m
2
x
2
2M
+
m
1
m
2
[x[
.
b) We can omit the constants m
1
m
2
and consider the equivalent Lagrangian
L L(x, x) :=
x
2
2M
+

[x[
,
which gives the same trajectories. Denote by M(t) the angular momentum M(t) := x(t) p(t),
where p =
L
x
=
x
M
. Then,
M(t) = x(t)
x(t)
M
= const, t IR. (45.2)
Indeed,
M(t) = x(t) p(t) +x(t) p(t) = x(t)

x(t)
M
+x(t)
x(t)
M
= x(t)
x(t)
M
.
On the other hand, the Euler-Lagrange equation implies
x(t)
M
+
x
[x[
3
= 0. Hence, x|x and
M(t) = x(t)
x(t)
M
= 0. 2
Remark 45.4 We can obtain Eqn (45.2) from the following arguments. Note that the potential
V (x) :=

[x[
depends only on [x[. Hence, V (x) is invariant with respect to rotations (see section
2.3.2), and the angular momentum M is conserved (see Theorem 2.15).
Finally, Eq. (45.2) implies
x(t) x(t) = x(0) x(0) =: n.
Hence, x(t) belongs to the plane perpendicular to the vector n. We can choose the coordinates
such that x
3
| n. Then, x(t) = (u(t), v(t), 0) and the Lagrangian for the variables u(t), v(t) is
L(u, v, u, v) :=
u
2
+ v
2
2M
+

u
2
+v
2
. (45.3)
c) We choose polar coordinates (r, ):
_
u = r cos ,
v = r sin .
(45.4)
Then u
2
+v
2
= r
2
, u
2
+ v
2
= r
2
+r
2

2
. We substitute into the Lagrangian and get
L(r, , r, ) :=
r
2
+r
2

2
2M
+

r
. (45.5)
Applying the Euler-Lagrange equation to the coordinate ,
d
dt
L

= 0, we get
d
dt
_
r
2

M
_
= 0.
Hence, we proved Keplers third law
r
2
= const =: I. (45.6)
Remark 45.5 (45.6) is equivalent to (45.2). Namely, if we calculate the angular momentum
M(t) in polar coordinates, we get
M(t) =
u v uv
M
=
r
2

M
= M(0).
Now applying the Euler-Lagrange equation to the coordinate r:
d
dt
L
r

L
r
= 0, we get the
reduced radial equation
r = r
2
M
r
2
.
By (45.6) we rewrite it as follows:
r =
I
2
r
3

M
r
2
= V
(r), V (r) :=
I
2
2r
2

2M
r
.
Multiplying this equation by r and integrating, we obtain the reduced energy conservation
r
2
+V (r) = E
0
. (45.7)
200
Remark 45.6 Since the Lagrangian (45.3) does not depend on t, the energy is conserved,
E(t) :=
u
2
+ v
2
2M

u
2
+v
2
= const.
In polar coordinates, taking into account (45.6), we get
E(t) =
r
2
+r
2

2
2M

r
=
r
2
2M
+
I
2
2Mr
2

r
= const,
which coincides with (45.7).
d) Now we nd r(). Since
dr
d
=
r

, we get by (45.7) and (45.2)
dr
d
=
I
2
r
2
+
2M
r
+ 2E
0
I
r
2
.
Hence
_
Idr
r
2
I
2
r
2
+
2M
r
+ 2E
0
=
_
d.
We introduce the Clerot substitution := 1/r. Then dr =
2
d. Hence,
_
Idr
r
2
I
2
r
2
+
2M
r
+ 2E
0
=
_
Id
_
I
2
2
+ 2M + 2E
0
=
_
Id
_
D I
2
( B)
2
.
Here D := 2E
0
+ (M)
2
I
2
, B := MI
2
. Note that the constant D must be positive for any
non-empty trajectory. Further,
_
Id
_
D I
2
( B)
2
= arcsin
I( B)
D
= +
0
.
Finally, we get
I( B)
D
= sin( +
0
). Since := 1/r, we get
A
1
r
= A
2
+ sin( +
0
), (45.8)
where
A
1
:=
I
D
, (45.9)
A
2
:=
IB
D
=
M
_
2E
0
I
2
+ (M)
2
. (45.10)
Note that r is bounded below by (45.8).
e) Let, for simplicity,
0
= 0. Then (45.8) becomes
A
2
r = A
1
r cos . (45.11)
We remember that r sin = v and r =
u
2
+v
2
, and now we re-write (45.11) in the form
(u
2
+v
2
)A
2
2
= (A
1
v)
2
. Hence,
u
2
+v
2
_
1
1
A
2
2
_
+ 2
A
1
A
2
2
v =
A
2
1
A
2
2
.
Obviously, A
1
0. Let us assume that A
1
> 0. Then, if A
2
> 1 we have an ellipse, if A
2
< 1
we have a hyperbola, and if A
2
= 1 we have a parabola (A
2
is a positive constant). By (45.10)
we get: if E
0
< 0 we have an ellipse, if E
0
> 0 we have a hyperbola, and if E
0
= 0 we have a
parabola.
When A
1
= 0 (i.e. I = 0) then we have either r = 0, arbitrary (the point mass is in the origin)
or sin = A
2
= const (the point mass moves along a line through the origin).
45.2 Exercise 5: Bohr-Sommerfeld Quantization of the H Atom
The Hamiltonian for an electron with charge e and mass m in the Coulomb eld of an innitely
heavy nucleus with charge e is
H =
1
2
_
p
2
r
+
p
2
r
2
_
e
2
r
, (45.12)
where the reduction to the eective two-dimensional problem (in polar coordinates (r, ), see
(45.4)) has already been performed (cf (45.5)). Find the rst two integrals of motion (conser-
vation of angular momentum and energy). Assume that the energy E is less that zero to have
periodic orbits (ellipses). Then use the BohrSommerfeld quantization rules
_
p
r
dr = kh,
_
p
d = lh (45.13)
(here both integrals are over one period, k and l are integers, and h is Plancks constant) to
derive the quantization of angular momentum and the Balmer formula for the energy levels of
the electron. Use the principal quantum number n = k +l in the latter case.
Solution: By (45.12) it follows that
= H
p
=
p
r
2
, (45.14)
p
= H
= 0, (45.15)
r = H
p
r
=
p
r
, (45.16)
p
r
= H
r
=
p
2
r
3

e
2
r
2
, (45.17)
By (45.15) we have
p
= M = const . (45.18)
Then by (45.16) we get
p
r
=
M
2
r
3

e
2
r
2
. (45.19)
202
Since
dp
r
dr
=
p
r
r
, (45.16), (45.19) imply
dp
r
dr
=
M
2
r
3

e
2
r
2
p
r
, or
p
r
dp
r
=
_
M
2
r
3

e
2
r
2
_
dr. (45.20)
Hence,
p
2
r
2
+
M
2
2r
2

e
2
r
= c = const . (45.21)
Let c < 0. Then (see Exercise 4) the orbits are ellipses. Hence,
p
r
=
2(
e
2
r
E)
M
2
r
2
. (45.22)
where E [c[ > 0 was introduced for convenience.
Now we nd the extrema of r. (45.16) implies that r = 0 i p
r
= 0. By (45.22) we have
2(
e
2
r
E)
M
2
r
2
= 0.
r
2
e
2
r
E
+
M
2
2E
= 0.
r =
e
2
2E

e
4
4E
2

M
2
2E
=
e
2
2E

1
2E
e
4
2M
2
E
=: r
.
Hence, r
max
= r
+
and r
min
= r
. Now we use (45.13). At rst, (45.13) and (45.18) imply

_
p
d =
_
M d = 2M = hl. (45.23)
Hence, M = hl, l ZZ. Then (45.22) implies
_
p
r
dr = 2
r
max
_
r
min
p
r
dr = 2
r
max
_
r
min
2(
e
2
r
E)
M
2
r
2
dr
= 2
r
max
_
r
min
_
2Er
2
+ 2e
2
r M
2
dr
r
. (45.24)
We use the formula
_
_
ax
2
+bx +c
dx
x
=
_
ax
2
+bx +c
b
2
a
arcsin
2ax +b
b
2
4ac
c arcsin
bx + 2c
x
b
2
4ac
, a < 0, c < 0.
We apply this formula with a = 2E < 0, b = 2e
2
, c = M
2
< 0 to
_
p
r
dr = 2
_
2(
e
2
r
E)
M
2
r
2
r
max
r
min
2e
2
2
2E
arcsin
4Er + 2e
2
_
4
2
e
4
8M
2
E
r
max
r
min
[M[ arcsin
2e
2
r 2M
2
r
_
4
2
e
4
8M
2
E
r
max
r
min
_
= 2
_
0 +
e
2
2E
[M[
_
= 2
_
e
4
2E
[M[
_
. (45.25)
Now (45.13) implies
2
_
e
4
2E
[M[
_
= hk, k ZZ.
Applying (45.23), we get
e
4
2E
= h
(k +[l[).
Denoting k +[l[ = n, we get
e
4
2E
= h
2
n
2
.
Hence, (cf. Theorem 8.1)
E =
e
4
2h
2
n
2
=
R
n
2
, R :=
e
4
2h
2
.
45.3 Exercise 6: Rutherford Scattering Formula
The orbit equation for a particle in a central potential U(r) is
=
_
r
r
0
M/r
2
dr
_
2(E U(r
)) M
2
/r
2
, (45.26)
where E is the energy and M is the angular momentum. For the Coulomb potential U(r) = /r
(with IR), and for E > 0 this is a scattering orbit (a hyperbola).
a) Calculate the angle
0
, which is obtained from Eq. (1) by integrating from r
min
to . Express
the total angle of deection of the orbit for a particle starting at r = and going back to r =
(i.e., the scattering angle ) with the help of
0
.
b) Re-express the constants of motion (E, M) by (E, b), where b is the scattering parameter, i.e.,
the normal distance between the asymptote of the incident particle and the scattering center at
r = 0.
Then, assume that a constant ux of particles (i.e., a constant number n of particles per area
and time) with xed energy and direction is approaching the scattering center. The number of
particles per area and per time in a ring between b and b+db is then dN = 2nb db. If there is a
one-to-one functional relation between b and the scattering angle , then dN is at the same time
the number of particles per time that is scattered in an angle between and + d. Use this
to calculate the dierential cross section for Coulomb scattering (i.e., the Rutherford scattering
formula)
d
dN
n
= 2b()
db()
d
d.
Solution:
a) For the Lagrangian L =
x
2
2
+

|x|
in polar coordinates, we get (45.26) (see Exercise 4). Let
E > 0, then the orbits are hyperbolae. The scattering angle is = [ 2
0
[, where
0
:=
_

r
min
Mdr
_
2Er
2
2r
M
2
. (45.27)
204
We use the formula
_
dx
x
ax
2
+bx +c
=
1
c
arcsin
bx + 2c
x
b
2
4ac
, c := M
2
< 0.
Hence,
0
:= M
1
M
arcsin
2r
2M
2
r
_
4
2
2
+ 8M
2
E
[
r
min
. (45.28)
Now we nd r = r
min
:
2Er
2
2r M
2
= 0
Since r > 0, we nd
r =

E
+

2
4E
2
+
M
2
2E
=

E
+
1
2E
_
2
+ 2M
2
E/
0
:= arcsin
M
2
/(r)
_
2
+ 2M
2
E/
[
r
min
= arcsin

_
2
+ 2M
2
E/
arcsin(1)
= arcsin

_
2
+ 2M
2
E/
+/2. (45.29)
Hence,
= [ 2
0
[ = 2 arcsin
[[
_
2
+ 2M
2
E/
. (45.30)
b) Since E = v
2
/2, v
=
_
2E/. Hence, M = [ r v[ = v
b =
2Eb. (45.30) implies

= 2 arcsin
[[
2
+ 4E
2
b
2
. (45.31)
Hence,
2
+ 4E
2
b
2
= sin
2

2
.
Hence,
b
2
=

2
4E
2
_
1
sin
2
(/2)
1
_
=

2
4E
2
ctg
2

2
,
b =

2E
ctg

2
. (45.32)
At rst, note that dN = 2nb db = 2nb()
db()
d
d and d
dN
n
= 2b()
db()
d
d. On
the other side, then (45.32) implies
db
d
=

2E
1
2
_
1
sin
2
(/2)
_
.
Hence,
d = 2

2E
ctg

2
2E
1
2
1
sin
2
(/2)
d =

2
4E
2
cos /2
sin
3
/2
d.
Hence, since d = 2 sin d = 4 sin /2 cos /2 d, we get for the dierential scattering cross
section
d =

2
16E
2
1
sin
4
/2
d.
45.4 Exercise 7: Energy Flow in Maxwell Field
Use the Maxwell equations in matter and the fact that
_
d
3
xE j is the change of kinetic energy
per time of the charged matter contained in the volume to show that u =
1
8
(E
2
+B
2
) is the
energy density of the electromagnetic eld, and S
c
4
EB is the density of energy ow.
Solution: Let us recall the Maxwell equations
_
_
div E(t, x) = 4(t, x), rot E(t, x) =
1
c
B(t, x),
1
c
E(t, x) +
4
c
j(t, x),
(t, x) IR
4
,
where (t, x) and j(t, x) stand for the charge and current density, respectively. Subtracting
B E from E B we obtain
E rot BB rot E =
1
c
_
E

E+B

B
_
+
4
c
E j. (45.33)
The l.h.s. may be rewritten like
E rot BB rot E = ( B) E( E) B = (EB), (45.34)
where the brackets indicate the action of the derivative. We arrive at
t
1
8
(E
2
+B
2
) = E j S, (45.35)
where
S
c
4
EB (45.36)
is called Poynting vector. Integration over a region of space leads to
t
1
8
_
d
3
x(E
2
+B
2
) =
_
d
3
xE j
_
da n S. (45.37)
Next, we need the fact that
_
d
3
xE j is the change of kinetic energy per time unit of the matter
described by j in the electric eld E,
_
d
3
xE j =
t
c
kin
. (45.38)
This we will prove below. Now we integrate over the whole space in (45.37) and use that the
elds go to zero for large distances to nd
t
__
d
3
x
1
8
(E
2
+B
2
) +c
kin
_
= 0, (45.39)
which is the conservation of the total energy. Therefore,
u =
1
8
(E
2
+B
2
) (45.40)
is the energy density of the electromagnetic eld, and the integral c
em
=
_
d
3
xu is its energy
in the volume . For nite volume we nally nd
t
(c
em
+c
kin
) =
_
da n S, (45.41)
206
therefore
_
da n S is the ow of electromagnetic energy through the surface , and S is the

density of the energy ow.
We still have to prove that
_
d
3
xE j is the power (change of kinetic energy per time unit) of
the charged matter in the volume in the electric eld E. Here we assume that the matter is
composed of point particles and prove the assumption for a single particle, where j = ev(x
x(t)). We nd
_
d
3
xE j = ev E(x), (45.42)
if the particle is contained in the volume . Further, we have in the non-relativistic case
c
kin
=
m
2
v
2

t
c
kin
= mv v (45.43)
and the expression for the Lorentz force
m v = eE+
e
c
v B. (45.44)
Inserted into the expression for
t
c
kin
this gives
t
c
kin
= ev E, (45.45)
because the second (magnetic) term in the Lorentz force does not contribute. This is what we
wanted to prove. The proof for relativistic matter particles is similar.
45.5 Exercise 8: Electromagnetic Plane Waves
Find the electric and magnetic elds of a plane electromagnetic wave by inserting the plane wave
ansatz E = E
0
exp(ik x it) into the free Maxwell equations. Specialize to k = k e
3
, nd
the two linearly independent solutions and construct a circularly polarized wave as an example.
Calculate the corresponding energy ow per time unit.
Solution: The Maxwell equations in vacuum are
B =
1
c
E , E =
1
c
B
B = 0 , E = 0 . (45.46)
From here the wave equation may be derived easily, e.g.,
(E) =
1
c
B,
( E) E =
1
c
2
E,
E =
1
c
2
E, (45.47)
and, in a similar fashion
B =
1
c
2
B. (45.48)
Inserting the plane-wave ansatz
E = E
0
e
ikxit
(45.49)
into the wave equation leads to the dispersion relation
k
2
=

2
c
2
. (45.50)
Further, the plane wave ansatz may be veried for the magnetic eld with the help of the Maxwell
equations. From
E = ik E
0
e
ikxit
=
1
c
B (45.51)
we nd upon integration
B =
c
k E+B
0
(x) (45.52)
where B
0
is a constant of integration which may depend on x only. Further, we have
B =
1
c
E, (45.53)
i
c
k
2
E+B
0
= i
c
E, (45.54)
and, therefore, B
0
= 0, which, together with B
0
= 0, implies B
0
= const. If we assume
that such a constant magnetic eld in all space is absent, then
B =
k
[k[
E =
k
[k[
E
0
e
ikxit
(45.55)
follows. Obviously, [E[ = [B[ and E B hold.
Inserting the plane wave ansatz into the Maxwell equations, we get
k B =
c
E , k E =

c
B,
k B = 0 , k E = 0 . (45.56)
Here we allow for a complex notation for the plane wave eld for convenience, but one should
keep in mind that only the real part of the eld is physical. For a plane wave eld (45.49) with
E
0
= nE
0
= n[E
0
[e
i
the physical eld is
E
ph
= Re
_
E
0
e
ikxit
_
= n[E
0
[ cos(k x t +) (45.57)
where n is a real constant unit vector.
Now let us assume without loss of generality that k = k e
z
. Then Equation (45.56) has two
linearly independent solutions pointing into the x and y directions, and, consequently, the general
solution is
E = E
0,x
e
x
e
ikzit
+E
0,y
e
y
e
ikzit
(45.58)
E
ph
= [E
0,x
[ e
x
cos(kz t +
1
) +[E
0,y
[ e
y
cos(kz t +
2
) . (45.59)
If
1
=
2
, then the electromagnetic wave is linearly polarized. If we assume that
1
=
2
= 0
and focus on the x, y plane z = 0, then the electric eld points into the [E
0,x
[ e
x
+ [E
0,y
[ e
y
direction at a time t = 0, is zero at a time t = (/2), points into the minus [E
0,x
[ e
x
+[E
0,y
[ e
y
direction at t = (/), etc. Therefore, the electric eld vector oscillates along a line parallel to
the [E
0,x
[ e
x
+[E
0,y
[ e
y
direction.
Circular polarization we have for [E
0,x
[ = [E
0,y
[ E
0
and
2

1
= (/2). For
1
= 0, e.g.,
we have in the z = 0 plane an electric eld vector with length E
0
pointing into the e
x
direction
208
at time t = 0, a vector with the same length pointing into the e
y
direction at time t = (/2),
etc. Therefore, the eld vector rotates in a counter-clockwise direction with angular velocity
without changgng its length. We nd for the physical elds in this case
E
ph
= E
0
( e
x
cos(kz t) e
y
sin(kz t)) (45.60)
B =
c
k E = e
z
E = E
0
( e
y
cos(kz t) + e
x
sin(kz t)) (45.61)
and, therefore, for the energy ow density
S =
c
4
EB =
c
4
E
2
0
e
z
=
c
4
E
2
0
k
[k[
. (45.62)
The energy ow density points into the direction of the wave vector k.
Remark 45.7 For general polarizations the energy ow is still time dependent. E.g., for linear
polarization E = E
0
ncos(kz t) (E
0
real, n in the x, y plane), the energy ow density is
S =
c
4
E
2
0
k
[k[
cos
2
(kz t). (45.63)
Here, one is more interested in the average energy ow, where a time average for times t >>
1
is performed,
lim
T
1
T
_
T
0
cos
2
(kz t)dt =
1
2
, (45.64)
which leads to
S =
c
8
E
2
0
k
[k[
(45.65)
(the factor of (1/2) compared to (45.62) is due to the fact that in (45.65) we took into account
only one of the two degrees of freedom).
45.6 Exercise 9: Hertzian Dipole Radiation
For a time-dependent dipole, (t, x) = p(t)
3
(x), calculate the radiation eld, i.e., the
electric and magnetic elds in the radiation (far) zone. Calculate the corresponding energy ux
and discuss its direction dependence. Hint: use the retarded potentials and the large distance
expansion.
Solution: With the help of the current conservation equation + j = 0, the corresponding
dipole current density may be calculated,
j = p(t)
3
(x). (45.66)
The retarded vector potential generated by this dipole is
A(x, t) =
1
c
_
d
3
x
_
dt
j(x
, t
)
[x x
[
(t
+
[x x
[
c
t) =
1
c
_
dt
p(t
)
1
[x[
(t
+
[x[
c
t) =
1
c
p(t (r/c))
r
, (45.67)
where r = [x[. The scalar potential may be calculated in an easy way by employing the Lorentz
gauge condition A+ (1/c)

= 0,
= c A =
t
p(t (r/c))
r
,
=
t
p(t (r/c))
r
=
1
r
p(t (r/c)) ep
1
r
=
x p(t (r/c))
cr
2
+
x p
r
3

x p(t (r/c))
cr
2
, (45.68)
where the large distance approximation was used in the last step.
Now the magnetic eld may be calculated,
B = A =
1
c
p(t (r/c))
r
=
1
c
p(t (r/c))
x
r
3
+
1
cr
(t
r
c
) p(t
r
c
)
1
c
2
p(t
r
c
) x
r
2
, (45.69)
and, in a similar fashion, the electric eld
E =
1
c
A
1
c
2
p
r

x p(t
r
c
)
cr
2
=
1
c
2
p
r

1
cr
2
( p +x
j
p
j
(t
r
c
))
1
c
x pr
2
1
c
2
p
r
+
x(x p)
c
2
r
3
=
1
c
2
r
3
( p(t
r
c
) x) x, (45.70)
where for E only the rst and third terms in the second line contribute in the large distance
limit. Observe that x E B, i.e., this is a radiation eld.
For the energy ow (Poynting vector) we nd
S =
c
4
EB =
1
4c
3
r
5
_
( p x)x pr
2
_
( p x) =
1
4c
3
r
5
_
r
2
p
2
( p x)
2
_
x (45.71)
If we assume that the dipole is oriented toward the z direction, p = (0, 0, p) then we nd
S =
sin
2
4c
3
r
2
p
2
n, (45.72)
where
n =
x
r
, cos =
n p
[p[
. (45.73)
Therefore, S behaves like r
2
for large distances (which it must due to energy conservation),
and there is no radiation (energy ow) in the direction of the dipole (for = 0, ).
210
45.7 Exercise 10: Polarizability in an External Electromagnetic Wave
In many instances, a classical, point-like electron in matter may be described approximately
as a damped harmonic oscillator which oscillates around a xed, positively charged center. In
an external electromagnetic eld the corresponding equation of motion for the position of the
electron is
m( x(t) + x(t) +
2
0
x(t)) = e(E+
1
c
x B) .
Here m is the mass and e the charge of the electron. is a damping constant and x is the
distance vector from the center. For [v[ << c the second, magnetic part of the Lorentz force
may be neglected. Under this assumption, c alculate the electric dipole moment p = ex of the
electron in the direction of the external eld for an oscillating eld E = E
0
exp(it). Assume
that the dipole oscillates with the same frequency, p = p
0
exp(it). Discuss the frequency
dependence of the resulting polarizability.
Solution: Inserting the dipole moment into the above equation and assuming that the velocity
is suciently small (neglecting the magnetic part of the Lorentz force), we get
p + p +
2
0
p =
e
2
m
E
0
e
it
. (45.74)
Further assuming that the distance vector (and the dipole moment) oscillates with the same
frequency like the external electric eld (driven oscillator) leads to
(
2
i +
2
0
)p
0
=
e
2
m
E
0
(45.75)
or
p
0
=
e
2
m
1
2
0

2
i
E
0
()E
0
(45.76)
with the atomic polarizability . The complex nature of the polarizability means that there
exists a phase dierence between E and p. This phase dierence is 0 for low frequency and in
the limit of very high frequency. The real and imaginary parts of are
Re =
e
2
m
2
0

2
(
2
0

2
)
2
+
2
2
(45.77)
Im =
e
2
m
(
2
0

2
)
2
+
2
2
. (45.78)
This shows that there is anomalous dispersion (weaker refraction for light with higher frequency)
for values of near the eigenfrequency
0
.
If there are n atoms per volume and f
k
denotes the fraction of electrons per atom that oscillate
with eigenfrequency
k
, then the total polarization per volume is
p =
ne
2
m
k
f
k
2
k

2
i
E, (45.79)
and, with p =
e
E, = 1 + 4
e
, we nd for the electric permittivity
= 1 +
4ne
2
m
k
f
k
2
k

2
i
, (45.80)
which is the formula of Drude for the permittivity. For many materials 1 and n

. For
frequencies near the eigenfrequencies anomalous dispersion occurs.
45.8 Exercise 11: Fresnels Formulae for Reection and Refraction
Two homogeneous, sotropic transparent media (with constant, real permittivities
1
,
2
and
permeabilities
1
,
2
) are separated by a plane. Use the Maxwell equations in macroscopic
matter, the resulting boundary conditions, and the plane wave ansatz to derive the laws of
reection and refraction.
Solution: The Maxwell equations in macroscopic matter, but without further microscopic
charge and current distributions, are
D = 0 (45.81)
B = 0 (45.82)
E =
1
c
B (45.83)
H =
1
c
D (45.84)
where
B = H , D = E. (45.85)
Now we assume that we have a medium 1 with permittivity
1
and permeability
1
in the region
z < 0, and a medium 2 with permittivity
2
and permeability
2
in the region z > 0, so the
separating plane is the (x, y) plane z = 0. Then the following continuity conditions for the elds
innitesimally below and above the separating plane follow from the Maxwell equations:
D
(2)
= D
(1)
(45.86)
B
(2)
= B
(1)
(45.87)
E
(2)
||
= E
(1)
||
(45.88)
H
(2)
||
= H
(1)
||
. (45.89)
Here, e.g., D
(2)
is the component perpendicular to the separating plane of the D eld innitesi-

mally above the plane (i.e., in region 2), etc.
To prove Eq. (45.86), we integrate the Maxwell equation (45.81) over the volume of a cylinder,
such that the bottom and top of the cylinder are parallel discs of the same radius innitesimally
below and above the plane z = 0, and the cylinder mantle is perpendicular to the plane and has
innitesimal height. We get
0 =
_
d
3
x D =
__
bottom
+
_
mantle
+
_
top
_
df D
=
_
disc
df(n
2
+n
1
) D, (45.90)
where we used the facts that the mantle area is innitesimal and that the two discs have the
same area. The two discs have opposite orientation, therefore n
2
= n
1
,
0 =
_
disc
df (D
2
D
1
), (45.91)
and Eq. (45.86) follows (n
i
are perpendicular to the plane z = 0). The proof of Eq. (45.87) is
analogous.
212
To prove Eq. (45.88), we integrate Eq. (45.83) over a rectangle with corners at (x
1
, y
1
, ),
(x
1
, y
1
, ), (x
2
, y
2
, ), and (x
2
, y
2
, ). Here is innitesimal, whereas the distance l = [(x
2

x
1
)
2
+ (y
2
y
1
)
2
]
1/2
is nite. We nd
_
rect
dfn E =
1
c
_
rect
dfn

B (45.92)
where n is the unit vector perpendicular to the rectangle. Here the r.h.s. may be neglected,
because the area of the rectangle is innitesimal,
lim
0
_
rect
dfn

B = 0. (45.93)
The l.h.s. is then
0 =
_
rect
dfn E =
_
dl (E
(2)
E
(1)
), (45.94)
where dl is the line element along the upper nite side of the rectangle, and we neglected the
line integrals along the innitesimal sides. As the orientation of the nite sides of the rectangle
in the (x, y) plane is arbitrary, Eq. (45.88) follows. The proof of eq. (45.87) is analogous.
Within a given medium (i.e., with xed and ), the Maxwell equations may be re-written like
E = 0 (45.95)
B = 0 (45.96)
E =
1
c
B (45.97)
B =
n
c
E (45.98)
where n stands for the index of refraction:
n
. (45.99)
As for the Maxwell equations in vacuum it follows that E (as well as B, D and H) obeys the
wave equation
E
n
2
c
2
E = 0 (45.100)
with propagation velocity v = c/n c. Therefore, we may use the plane wave ansatz for the
electromagnetic eld in a region with xed index of refraction,
E = E
0
e
ikxit
(45.101)
where the dispersion relation is [k[ = (n/c).
Now we want to study reection and refraction. We assume that an incident plane wave in
region 1 (z < 0) propagates toward the separating plane z = 0 (i.e., k
z
> 0), and we want to
determine a further plane wave in region 1 with k
z
< 0 (reected wave) and a plane wave in
region 2 (refracted wave) from the boundary conditions (45.86)(45.87). Therefore we make
the ansatz
E = E
0
e
ikxit
, k E
0
= 0 , z < 0 (45.102)
E
= E
0
e
ik
xi
t
, k
0
= 0 , z > 0 (45.103)
E
= E
0
e
ik
xi
t
, k
0
= 0 , z < 0 (45.104)
where [k[ = (n
1
/c), [k
[ = (n
2
/c)
and [k
[ = (n
1
/c)
. The other elds may be computed

easily from the Maxwell equation (45.82)
B =
c
k E, (45.105)
B
=
c
, (45.106)
B
=
c
, (45.107)
and from relations (45.85).
The total solution in region 1 is now E
(1)
= E +E
, and the solution in region 2 is E

(2))
= E
,
and analogously for the other elds. Now we may use the continuity conditions (45.86)(45.87).
As they must hold everywhere along the z = 0 plane and for all times, the phase functions
multiplying the constant vectors must be the same, i.e.,
=
, (45.108)
k
x
= k
x
= k
x
, k
y
= k
y
= k
y
. (45.109)
Therefore, all three wave vectors must lie in the same plane (plane of incidence), and we
may choose without loss of generality k
y
= k
y
= k
y
= 0 (incidence plane = (x, z) plane). The
dispersion relations are now
[k[ = [k
[ = (n
1
/c) , [k
[ = (n
2
/c) (45.110)
and imply
k
z
= k
z
. (45.111)
Dening the angles between the wave vectors and the z axis,
cos =
k
z
[k[
, cos
=
k
z
[k
[
, cos
=
k
z
[k[
, (45.112)
we have therefore =
, i.e., the incidence angle equals the reection angle. We further have
sin
sin
=
k
x
/[k[
k
x
/[k
[
=
n
2
n
1
, (45.113)
which is the law of Snellius.
With the help of the unit vector e
z
(which is perpendicular to the plane z = 0) the continuity
conditions (45.86)(45.87) may be expressed like follows: for the tangential component of E
(E
0
+E
0
E
0
) e
z
= 0, (45.114)
normal component of B
(k E
0
+k
0
k
0
) e
z
= 0, (45.115)
tangential component of H = (1/)B
_
1
1
(k E
0
+k
0
)
1
2
k
0
_
e
z
= 0, (45.116)
214
normal component of D = E
_
1
(E
0
+E
0
)
2
E
0
_
e
z
= 0. (45.117)
For a further investigation we specialize to two linearly independent, linearly polarized incident
waves. This is sucient, because a general incident electromagnetic plane wave may always be
expressed as a linear combination of these two. The rst case is a vector E
0
perpendicular to
the incidence plane, i.e., E
0
= E
0
e
y
. First, we nd that E
0
and E
0
also have only components
perpendicular to the incidence plane, E
0
= E
0
e
y
, E
0
= E
0
e
y
. For assume that E
i
= E
e
y
(and analogous for E
) are nonzero. Then we nd from (45.114)

_
E
i

k
e
y
E
e
y
_
e
z
= 0, (45.118)
E
i
cos E
i
cos
= 0 E
i
=
cos
cos
E
i
. (45.119)
cos and cos
are greater than zero by denition, therefore the two amplitudes have dierent
sign. On the other hand, we nd from (45.115)
_
E
1
k
e
y
)
E
2
k
e
y
)
_
e
z
= 0, (45.120)
_
1
[k
[ +
E
2
[k
[
_
e
y
e
z
= 0, (45.121)
i
n
1
1
+
E
i
n
2
2
= 0 E
i
=
n
2
1
n
1
2
E
i
, (45.122)
which implies that the two amplitudes must have the same sign. This requires that both ampli-
tudes are zero.
For the perpendicular components we nd from (45.114)
(E
0
+E
0
E
0
) e
y
e
z
= 0 E
0
+E
0
E
0
= 0, (45.123)
and from (45.116)
_
1
1
(E
0
k e
y
+E
0
k
e
y
)
1
2
E
0
k
e
y
_
e
z
= 0,
_
1
1
(E
0
k e
z
+E
0
k
e
z
)
1
2
E
0
k
e
z
_
e
y
= 0,
n
1
1
cos (E
0
E
0
)
n
2
2
cos
0
= 0. (45.124)
Eq. (45.115) gives the same result like Eq. (45.114), and Eq. (45.117) is trivially zero. Inserting
(45.123) into (45.124) leads to
_
1
cos (E
0
E
0
) =
_
2
cos
(E
0
+E
0
),
E
0
E
0
=
_
1
/
1
cos
_
2
/
2
cos
1
/
1
cos +
_
2
/
2
cos
=
1
_
2
cos
cos
1 +
_
2
cos
cos
=
1

1
2
tan
tan
1 +

1
2
tan
tan
. (45.125)
Further,
E
0
E
0
= 1 +
E
0
E
0
=
2
1 +
_
2
cos
cos
=
2
1 +

1
2
tan
tan
. (45.126)
For
1
=
2
(which is true for many substances) we nd
E
0
E
0
=
1
sin cos
cos sin
1 +
sin cos
cos sin
=
sin(
)
sin( +
)
, (45.127)
E
0
E
0
=
2
1 +
sin cos
cos sin
=
2 cos sin
sin( +
)
, (45.128)
which are the Fresnel formulae.
The second case to study is for E
0
parallel to the incidence plane, E
0
= E
0
k e
y
. Again, it may
be proven easily that both E
0
and E
0
must be parallel to the incidence plane, as well, in this
case, E
0
= E
e
y
and E
0
= E
e
y
. We nd from (45.114)
_
E
0
k e
y
+E
e
y
E
e
y
_
e
z
= 0,
_
E
0
k e
z
+E
e
z
E
e
z
_
e
y
= 0,
(E
0
E
0
) cos E
0
cos
= 0, (45.129)
and from (45.116)
_
1
1
[E
0
k (
k e
y
) +E
0
k
e
y
)]
1
2
E
0
k
e
y
)
_
e
z
= 0,
_
1
1
[E
0
[k[ +E
0
[k
[]
1
2
E
0
[k
[
_
e
y
e
z
= 0,
n
1
1
(E
0
+E
0
)
n
2
2
E
0
= 0 E
0
=
_
1
(E
0
+E
0
) (45.130)
(again (45.115) and (45.117) do not contain additional information). Inserting this into the
above equation leads to
(E
0
E
0
) cos =
_
1
(E
0
+E
0
) cos
,
E
0
E
0
=
cos
_
2
/
2
1
cos
cos +
_
2
/
2
1
cos
=
1
_
1
cos
cos
1 +
_
1
cos
cos
=
1

1
2
tan
tan
1 +

1
2
tan
tan
. (45.131)
Further,
E
0
E
0
=
_
1
(1 +
E
0
E
0
) =
2
_
1
1 +

1
2
tan
tan
. (45.132)
216
For
1
=
2
these formulae simplify to
E
0
E
0
=
1
n
2
1
n
2
2
tan
tan
1 +
n
2
1
n
2
2
tan
tan
=
1
sin
cos
sin cos
1
sin
cos
sin cos
=
sin cos sin
cos
sin cos + sin
cos
, (45.133)
E
0
E
0
=
2
n
1
n
2
1 +
n
2
1
n
2
2
tan
tan
=
2
sin
sin
1 +
sin
cos
sin cos
=
2 sin
cos
sin cos + sin
cos
, (45.134)
which are the Fresnel formulae for the case of electric eld parallel to the incidence plane.
From these results some physical applications immediately follow. For n
2
< n
1
total reection
for >
max
follows from the Snellius law, where sin
max
= n
2
/n
1
.
Another interesting consequence is the possibility to generate linearly polarized light from un-
polarized light by choosing the so-called Brewster angle for the incident light beam. Indeed, it
follows from (45.133) that the component parallel to the incidence plane is not reected at all if
the condition
sin cos = sin
cos
(45.135)
holds. The solution =
is of course forbidden if n
2
,= n
1
. But there exists the possible
solution
= /2 which leads to sin
= cos , cos
= sin and
n
2
n
1
=
sin
sin
=
sin
cos
= tan . (45.136)
Therefore, if an unpolarized light beam is incident with an angle
B
such that tan
B
= n
2
/n
1
,
then the reected beam will contain only light linearly polarized into the direction perpendicular
to the incidence plane, E
0
e
y
.
45.9 Exercise 12: Classical Zeemann Eect
Again electrons in matter are described by spherical harmonic oscillators which oscillate around
a xed center (nucleus), as in Exercise 45.7. Here we assume that the electrons are excited
with their eigenfrequency by some unspecied mechanism (e.g. heat or radiation), therefore we
ignore a possible damping. These harmonic oscillators are now exposed to a constant, external
magnetic eld, therefore the equation of motion for the position of one electron is
m( x(t) +
0
x(t)) =
e
c
x B. (45.137)
Find the three eigenfrequencies and eigenmodes of this equation of motion. (Hint: use the
symmetry of the problem and the known behavior of a charged particle in a constant magnetic
eld to guess the right ansatz.) Use the resulting dipole moments p = ex to calculate the dipole
radiation of all three modes and their polarization patterns both parallel and perpendicular to
the constant, external magnetic eld.
Solution: We assume without loss of generality that the constant, external magnetic eld points
into the positive z direction, B = B e
3
, B > 0. Then we have the following equations of motion,
x
1
+
2
0
x
1
=
eB
mc
x
2
, (45.138)
x
2
+
2
0
x
2
=
eB
mc
x
1
, (45.139)
x
3
+
2
0
x
3
= 0. (45.140)
Obviously, the third, decoupled equation is solved by x
3
= x
0
sin(
0
t
0
) where
0
is an
irrelevant integration constant and is set to zero in the sequel. Therefore the rst eigenmode is
x
||
= x
0
sin(
0
t) e
3
(45.141)
with eigenfrequency
0
(the eigenfrequency is not altered by the magnetic eld). Now we use the
following facts about a charge moving in a constant magnetic eld: a charge moving parallel to
the magnetic eld feels no force at all, whereas a charge moving perpendicular to the magnetic
eld is deected such that it moves along a circle with constant angular velocity. Further, the
free two-dimensional harmonic oscillator composed of Eqs. (45.138) and (45.139) has circles with
constant angular velocity as possible eigenmodes, therefore we try a circle with constant angular
velocity as an ansatz:
x
1
= x
0
cos t , x
1
= x
0
sin t , x
1
= x
0
2
cos t, (45.142)
x
2
= x
0
sin t , x
2
= x
0
cos t , x
2
= x
0
2
sin t. (45.143)
We nd from (45.138)
x
0
2
cos t +
2
0
x
0
cos t =
eB
mc
x
0
cos t (45.144)
and an analogous equation with cos t sin t from (45.139). Therefore the ansatz is a solution
provided that
2
+
2
0
=
eB
mc
(45.145)

=
eB
2mc

2
0
+
e
2
B
2
4m
2
c
2

eB
2mc

0
, (45.146)
where in the last step we assumed that the eigenfrequency
0
is much larger in absolute value
than the Larmor frequency
L
=
eB
2mc
. The minus sign for
means that this eigenmode

rotates in a clockwise direction, and the minus sign of
L
is the law of induction. Therefore, we
nd the following two perpendicular eigenmodes
x
1
= x
0
cos(
1
t) e
1
+x
0
sin(
1
t) e
2
(45.147)
x
2
= x
0
cos(
2
t) e
1
x
0
sin(
2
t) e
2
(45.148)
where
1
=
+
and
2
=
.
Next, we want to calculate the dipole radiation emitted by these three eigenmodes. Here we use
the formulae for Hertzian dipole radiation (see Exercise 45.6), and p = ex,
B(t, x) =
e
c
2
r
2
p(t
r
c
) x, (45.149)
E(t, x) =
e
c
2
r
3
[r
2
p(t
r
c
) + (x p(t
r
c
))x], (45.150)
S =
e
2
4c
3
r
5
[r
2
p
2
(x p)
2
]x. (45.151)
For the parallel eigenmode, p
||
= ex
0
sin(
0
t) e
3
, we nd linearly polarized radiation with polar-
ization direction parallel to the external B eld (the radiation B eld is always perpendicular
to the e
3
direction). The linear polarization is extracted most easily from the Poynting vector
S =
e
4
4
0
x
2
0
4c
3
r
5
(r
2
z
2
) sin
2
(
0
t
)x =
e
4
4
0
x
2
0
4c
3
r
5
sin
2
sin
2
(
0
t
)n, (45.152)
218
where we used spherical polar coordinates, n (x/r), and t
t (r/c). For linear polarization

at each point in space in the radiation zone the electric (and magnetic) eld vector oscillates
and vanishes for certain times (instead of rotating along an ellipse or a circle), therefore also the
energy ow vanishes at certain times. Further , there is no radiation in the e
3
direction.
For the eigenmodes perpendicular to the constant, external magnetic eld the Poynting vector
is (where =
)
S =
e
4
4
x
2
0
4c
3
r
5
[r
2
(xcos(t
) +y sin(t
))
2
]x =
e
4
4
x
2
0
4c
3
r
2
[1 sin
2
cos
2
(t
)]n. (45.153)
In the (x, y) plane perpendicular to e
3
we get (with n
= cos e
1
+ sin e
2
)
S =
e
4
4
x
2
0
4c
3
r
2
sin
2
(t
)n
(45.154)
and, therefore, linearly polarized light. Further, the polarization direction is perpendicular to
the e
3
direction (because the magnetic radiation eld is parallel to e
3
). For the radiation in e
3
direction we get
S =
e
4
4
x
2
0
4c
3
r
2
e
3
(45.155)
and the energy ow does not depend on time. Therefore, the radiation is circularly polarized
(the radiation eld vectors rotate without changing their length). Further, the + mode has left
circular polarization (the E eld rotates counter-clockwise) and the mode has right circular
polarization.
45.10 Exercise 13: Diamagnetism and Paramagnetism
The relation between the magnetic induction B, the magnetic eld intensity H and the magne-
tization M of matter is (for isotropic matter)
B = H+ 4M, B = H, M =
m
H, = 1 + 4
m
.
Here a substance is called diamagnetic if < 1. Use again the harmonic oscillator model in
a constant external magnetic eld (like in Exercise 45.9) to derive the diamagnetic behavior.
Here the electrons moving on circular orbits have to be interpreted as currents which induce a
magnetic eld. Assume that the net induced magnetic eld without external magnetic eld is
zero.
Solution: We again assume that the external magnetic eld is along the e
3
direction, B = B e
3
.
Further, we are interested only in the d.o.f. perpendicular to the external magnetic eld. We
know from Exercise 45.9 that there are two degrees of freedom, +, , with frequencies
=
eB
2mc

0
. (45.156)
Without external eld these are just two modes rotating with the same angular velocity
0
,
one (the + mode) in the counter-clockwise direction, the other in the clockwise direction. The
two modes induce magnetic moments of equal strengths but opposite orientations, therefore,
macroscopically, their net contribution to the magnetization is zero. With external eld, however,
the additional contribution is in the clockwise (negative) direction in both cases, and a net
contribution remains. Further, the electron current and induced magnetic eld depend linearly
on , therefore we may just omit the
0
piece and use just
L
=
eB
2mc
. The magnetic moment
of a current distribution is
m =
1
2c
_
d
3
xx j, (45.157)
where j is the current density. In our case the electron is in a circular orbit in the (x, y) plane,
therefore
j = e(x x(t)) x (45.158)
(with x
3
= 0). With
x x = (x
1
x
2
+x
2
x
1
) e
3
= x
2
0
L
e
3
(45.159)
we get
m =
1
2c
e
L
x
2
0
e
3
=
e
2
Bx
2
0
4mc
2
e
3
, (45.160)
and, therefore, the magnetic moment is opposite to the external magnetic eld. For general
orbits of the electrons, x
2
0
must be replaced by the orbit average x
2
1
+x
2
2
; if the force law for the
electron is spherically symmetric, as in our case, then x
2
1
+x
2
2
= (2/3)x
2
. If there are several
electrons per atom with dierent average orbit radii and if there are n atoms per volume, then
the magnetization is
M = n
e
2
B
6mc
2
j
x
2
j
e
3
= n
e
2
6mc
2
j
x
2
j
B (45.161)
and the susceptibility is approximately (if it is small)
m
= n
e
2
6mc
2
j
x
2
j
. (45.162)
The assumptions here were rather general (electrons orbiting around nuclei) and, in fact, all
materials are diamagnetic, but the diamagnetism may be over-compensated by other eects
with positive magnetic susceptibility (like paramagnetism).
Paramagnetism is present when the following conditions hold: 1) the atoms or molecules of
the material already have a xed magnetic moment m
0
(even without external magnetic eld),
and 2) the orientations of these magnetic moments are randomly distributed in the absence of
an external magnetic eld. Without external magnetic eld there is, therefore, no macroscopic
magnetization of the paramagnetic material. However, in an external magnetic eld the magnetic
moments are partially aligned along the direction of the external eld, because this aligned
position is energetically favorable. This alignment is thwarted by thermal uctuations, and the
average magnetic moment per molecule is
m =
[m
0
[
2
kT
B
ext
, (45.163)
where is some constant depending on the molecule type, T is the temperature and k the
Boltzmann constant. Both quantum mechanical and thermodynamic considerations are needed
to derive equation (45.163), which is beyond the scope of the present discussion.
220
45.11 Exercise 14: Stark Eect
A hydrogen atom is exposed to a constant electric eld. Use perturbation theory to calculate
the energy shifts for the ground state
100
of the hydrogen atom (with no energy degeneracy of
the unperturbed system) and for the states
2lm
(with a four-fold energy degeneracy).
Solution: The Hamiltonian for the electron wave function in a Coulomb eld and in a constant
electric eld is
H =
h
2
2
e
2
r
eE x, (45.164)
and we assume E = f e
3
. The eigenfunctions
nlm
= R
nl
(r)F
m
l
()e
im
of the hydrogen atom are
no longer eigenfunctions of the full Hamiltonian, therefore perturbation theory in the perturbing
term V = efz = efr cos is needed.
For the ground state, no degeneracy of the unperturbed system occurs, because there is only one
ground state
100
with energy E
100
= R. In this case, the perturbation series for the energy is
E = E
0
k
+V
kk

i=k
[V
ki
[
2
E
0
i
E
0
k
+ (45.165)
where E
0
i
are the eigenenergies of the unperturbed Hamiltonian H
0
= H V , H
0
i
= E
0
i
i
,
and the perturbation about the level k is calculated. Further,
V
ki
=
_
d
3
x
k
(x)V
i
(x) (45.166)
is the matrix element of the perturbation V w.r.t. the eigenstates of the unperturbed system.
In our case
k
=
100
,
V
ki
= ef
_

0
drr
3
R
10
(r)R
nl
(r)
_
dcos F
0
0
()F
m
l
()e
im
, (45.167)
which is non-zero only for m = 0, l = 1. Therefore, V
kk
is zero, and there is no rst order (linear
in the external eld) contribution. Up to second order in the perturbation, we get
E = R e
2
f
2
const, (45.168)
which is quadratic in the external eld (quadratic Stark eect). For the electric dipole moment
we nd
p = ex) =

E
V ) =

E
E = const e
2
E. (45.169)
The interpretation of this result is as follows: the spherically symmetric ground state
100
has
no permanent electric dipole moment. But the external electric eld induces a dipole moment
proportional to its strength by polarizing the atom (displacing the charge center of the electron
wave function relative to the nucleus).
For higher states (e.g., n = 2) degeneracy occurs because all states with quantum number n
have energy (R/n
2
). E.g. for n = 2 there are four states
1

200
,
2

210
,
3

211
,
4

211
, (45.170)
with unperturbed energy E
0
2
= (R/4). When degeneracy occurs, the energies within the
degenerate subspace have to be determined exactly from the degeneracy condition
det
_
(E
0
2
E)
+V
_
= 0, (45.171)
where labels the degeneracy subspace. Here
V
= ef
_

0
drr
3
R
2l
(r)R
2l
(r)
_
dcos F
m
()F
m
()e
i(m
)
. (45.172)
It follows that
m
= m
= 0 , l
= l
1 (45.173)
and, therefore, only V
12
and V
21
are non-zero. They may be calculated and one nds
V
12
= V
21
= . . . = 3efa
0
, (45.174)
where a
0
is the Bohr radius. The determinant condition is
det
_
_
_
_
_
E
0
2
E V
12
0 0
V
12
E
0
2
E 0 0
0 0 E
0
2
E 0
0 0 0 E
0
2
E
_
_
_
_
_
= 0, (45.175)
(E
0
2
E)
2
[(E
0
2
E)
2
V
2
12
] = 0, (45.176)
with the solutions
E = E
0
2
(45.177)
(two-fold degenerate) and
E E
= E
0
2
V
12
. (45.178)
This expression is linear in the external eld (linear Stark eect). The electric dipole moment
for the two non-degenerate modes E
is
p =

E
E
= 3ea
0
e
3
. (45.179)
The interpretation is as follows: The spherically non-symmetric wave functions give rise to a
permanent electric dipole moment.
45.12 Exercise 15: Vector Model for Spin-Orbital Interaction
A phenomenological vector model has been used in the old quantum theory for an addition of the
magnetic moments in many-electron atoms and molecules. It is not rigorous and contradicts the Pauli
equation, however, sometimes it gives exact results.
Precession of angular momentum
Let us consider the orbital angular momentum L.
Theorem 45.8 In the uniform magnetic eld B = (0, 0, B) the components L
3
(t) and s
3
(t) are
conserved, while the vectors (L
1
(t), L
2
(t)) IR
2
and (s
1
(t), s
2
(t)) IR
2
rotate with angular velocity
L
and 2
L
, respectively, where
L
=
eB
2c
is the Larmor frequency.
Proof For concreteness we consider the orbital momentum. The conservation of L
3
is already proved.
Since

L
k
= h
H
k
, we have by (9.2),
[
L
1
,

L
2
] = ih
L
3
, [
L
2
,

L
3
] = ih
L
1
, [
L
3
,

L
1
] = ih
L
2
. (45.180)
222
Therefore,
[T,

L
1
] = i
L
h
L
2
, [T,

L
2
] = i
L
h
L
1
. (45.181)
since all

L
k
commute with all s
j
by Remark 19.2. Hence, analogously to the Heisenberg equation
(6.25), we get
L
1
(t) = (t), ih
1
[T
L
1

L
1
T](t))
= (t),
L
L
2
(t)) =
L
L
2
(t). (45.182)
Similarly, we have

L
2
(t) =
L
L
1
(t), hence
d
dt
(L
1
(t) +iL
2
(t)) = i
L
(L
1
(t) +iL
2
(t)). (45.183)
The theorem implies the precession of the vectors L(t) and s(t) with the angular velocities
L
and
2
L
, respectively, around the magnetic eld B.
Remark 45.9 A similar precession can be proved for a classical system of electrons rotating as rigid
bodies in the uniform magnetic eld [53].
Vector model
Let us apply the idea of the precession to the addition of the orbital and spin magnetic moments.
This addition explains the Einstein-de Haas and anomalous Zeemann experiments.
The model explains the magnetization in the Einstein-de Haas experiment by the classical mech-
anism of the reorientation in a magnetic eld of a total magnetic moment m which exists even in the
absence of the magnetic eld. m is the sum of the orbital and spin magnetic moments m
o
:=
e
2c
L
and m
s
:=
e
c
s:
m = m
o
+m
s
=
e
2c
L +
e
c
s. (45.184)
The moments are dened by the orbital and spin angular momenta L and s, respectively, with the
corresponding distinct gyromagnetic ratios
e
2c
and
e
c
.
In the absence of an external magnetic eld, the total angular momentum J = L+s is conserved,
while L and s are generally not conserved. For example, the spin angular momentum s precesses in the
magnetic eld generated by the orbital angular momentum. Similarly, the orbital angular momentum
L precesses in the magnetic eld generated by the spin angular momentum. Therefore, the lengths
of the vectors L and s are conserved. Hence, the conservation of the sum J = L +s implies that the
vectors L and s rotate around J. Then the total magnetic moment m =
e
2c
L +
e
c
s also rotates
around J. Since the angular velocity of the rotation is very high, we have to take into account only the
eective value of the magnetic moment, m
e
, which is the projection of the total magnetic moment
onto J. Let us calculate this projection.
The angle between the vectors J and L is conserved as well as the angle between the vectors
J and s, and
cos =
[J[
2
+[L[
2
[s[
2
2[J[[L[
, cos =
[J[
2
+[s[
2
[L[
2
2[J[[s[
. (45.185)
Then the projection equals
m
e
=
e
2c
[L[ cos +
e
c
[s[ cos . (45.186)
A nal, highly nontrivial approximation is as follows: we redene the lengths of the vectors J, L and
s as
[J[
2
:=, (
J
2
1
+
J
2
2
+
J
2
3
)), [L[
2
:=, (
L
2
1
+
L
2
2
+
L
2
3
)), [s[
2
:=, (s
2
1
+s
2
2
+s
2
3
)). (45.187)
The operators

J
2
1
+
J
2
2
+
J
2
3
,

L
2
1
+
L
2
2
+
L
2
3
, and s
2
1
+s
2
2
+s
2
3
= 3/4 commute. Hence, the quantum stationary
states can be classied by the eigenvalues of the operators which are equal to J(J +1), L(L+1), and
3/4, where J, L = 0, 1, 2, ... . For the states we have [J[
2
= J(J + 1), [L[
2
= L(L + 1), and [s[
2
= 3/4.
Substituting this into (45.186), we obtain the Lande formula (20.8) for the eective gyromagnetic ratio
g
e
:=
m
e
[J[
e
2c
=
[J[
2
+[L[
2
[s[
2
2[J[
2
+ 2
[J[
2
+[s[
2
[L[
2
2[J[
2
=
3
2
+
[s[
2
[L[
2
2[J[
2
=
3
2
+
3/4 L(L + 1)
2J(J + 1)
. (45.188)
The formula is conrmed experimentally by the Einstein-de Haas and anomalous Zeemann eects.
224
Bibliography
[1] M.Abraham, Theorie der Elektrizitat, Bd.2: Elektromagnetische Theorie der Strahlung, Teubner,
Leipzig, 1905.
[2] V.Arnold, Mathematical Methods of Classical Mechanics, Springer, Berlin, 1978.
[3] A.V.Babin, M.I.Vishik, Attractors of Evolutionary Equations, North-Holland, Amsterdam, 1992.
[4] R.Becker, Elektronentheorie, Teubner, Leipzig, 1933.
[5] A.Bensoussan, C.Iliine, A.Komech, Breathers for a relativistic nonlinear wave equation, Arch.
Rat. Mech. Anal. 165 (2002), 317-345.
[6] H.Berestycki, P.L.Lions, Arch. Rat. Mech. and Anal. 82 (1983), no.4, 313-375.
[7] F.A.Berezin, M.A.Shubin, The Schrodinger equation, Kluwer Academic Publishers, Dordrecht,
1991.
[8] H.Bethe, Intermediate Quantum Mechanics, Benjamin, NY, 1964.
[9] H.Bethe, E.Salpeter, Quantum Mechanics of One- and Two-Electron Atoms, Berlin, Springer,
1957.
[10] J.D.Bjorken, S.D.Drell, Relativistic Quantum Mechanics, McGraw-Hill, NY, 1964; Relativistic
Quantum Fields, McGraw-Hill, NY, 1965.
[11] N.Bohr, Phil. Mag. 26 (1913) 1; 476; 857.
[12] M.Born, Atomic Physics, Blackie & Son, London-Glasgow, 1951.
[13] N.Bournaveas, Local existence for the Maxwell-Dirac equations in three space dimensions, Com-
mun. Partial Dier. Equations 21 (1996), no.5-6, 693-720.
[14] L. de Broglie, The Current Interpretation of Wave Mechanics, Elsevier, NY, 1964.
[15] V.S.Buslaev, G.S.Perelman, On the stability of solitary waves for nonlinear Schrodinger equations,
Amer. Math. Soc. Trans. (2) 164 (1995), 75-98.
[16] V.S.Buslaev, C.Sulem, On asymptotic stability of solitary waves for nonlinear Schrodinger equa-
tions, Ann. Inst. Henri Poincar, Anal. Non Linaire 20 (2003), no.3, 419-475.
[17] E.U.Condon, E.U., G.H.Shortley, The Theory of Atomic Spectra, Cambridge University Press,
Cambridge, 1963.
[18] S.Cuccagna, Stabilization of solutions to nonlinear Schrodinger equations, Comm. Pure Appl.
Math. 54 (2001), no.9, 1110-1145.
225
226 BIBLIOGRAPHY
[19] S.Cuccagna, On asymptotic stability in 3D of kinks for the
4
model, preprint 2004.
[20] C.Davisson, L.Germer, Nature 119 (1927), 558.
[21] M.Defranceschi, C.Le Bris, Mathematical Models and Methods for ab Initio Quantum Chemistry,
Lecture Notes in Chemistry. 74, Springer, Berlin, 2000.
[22] T.Dudnikova, A.Komech, E.A.Kopylova, Yu.M.Suhov, On convergence to equilibrium distribu-
tion, I. Klein-Gordon equation with mixing, Comm. Math. Phys. 225 (2002), no.1, 1-32.
[23] T.Dudnikova, A.Komech, N.Mauser, On the convergence to a statistical equilibrium for the Dirac
equation, Russian Journal of Math. Phys. 10 (2003), no.4, 399-410.
[24] T.Dudnikova, A.Komech, N.Mauser, On two-temperature problem for harmonic crystals, Journal
of Statistical Physics 114 (2004), no.3/4, 1035-1083.
[25] T.Dudnikova, A.Komech, N.E.Ratanov, Yu.M.Suhov, On convergence to equilibrium distribution,
II. Wave equation with mixing, Journal of Statistical Physics 108 (2002), no.4, 1219-1253.
[26] T.Dudnikova, A.Komech, H.Spohn, On a two-temperature problem for wave equation with mix-
ing, Markov Processes and Related Fields 8 (2002), no.1, 43-80.
[27] T.Dudnikova, A.Komech, H.Spohn, On convergence to statistical equilibrium for harmonic crys-
tals, Journal of Mathematical Physics 44 (2003), no.6, 2596-2620.
[28] T.Dudnikova, A.Komech, H.Spohn, Energy-momentum relation for solitary waves of relativistic
wave equation, Russian Journal Math. Phys. 9 (2002), no.2, 153-160.
[29] D.M.Eidus, The principle of limit amplitude, Russ. Math. Surv. 24 No.3, 97-167 (1969).
[30] A.Einstein,

Uber einen die Erzeugung und Verwandlung des Lichtes betreenden heuristischen
Gesichtspunkt (Concerning the generation and transformation of light as seen from a heuristic
point of view), Annalen der Physik, March 18, 1905.
[31] M.Esteban, V.Georgiev, E.Sere, Stationary solutions of the Maxwell-Dirac and the Klein-Gordon-
Dirac equations, Calc. Var. Partial Dier. Equ. 4 (1996), no.3, 265-281.
[32] M. Fedoriuk, Asymptotic Methods for Partial Dierential Equations, Encyclopaedia of Mathe-
matical Sciences Vol. 34, Springer, Berlin, 1999.
[33] R.Feynman, Quantum Electrodynamics, Addison-Wesley, Reading, Massachusetts, 1998.
[34] Ya.Frenkel, Zs. Phys. 37 (1926), 43.
[35] J.Frohlich, S.Gustafson, B.L.G.Jonsson, I.M.Sigal, Solitary wave dynamics in an external poten-
tial, arXiv:math-ph/0309053v1.
[36] J.Frohlich, T.P.Tsai, H.T.Yau, On the point-particle (Newtonian) limit of the nonlinear Hartree
equation, Comm. Math. Physics 225 (2002), no.2, 223-274.
[37] I.Gelfand, S.Fomin, Calculus of Variations, Dover Publications, Mineola, NY, 2000.
[38] I.M.Gelfand, R.A.Minlos, Z.Ya.Shapiro, Representations of the Rotation and Lorentz Groups and
their Applications, Pergamon Press, Oxford, 1963.
BIBLIOGRAPHY 227
[39] M.Gell-Mann, Phys. Rev. 125 (1962), 1067.
[40] M.Grillakis, J.Shatah, W.A.Strauss, Stability theory of solitary waves in the presence of symme-
try, I; II. J. Func. Anal. 74 (1987), no.1, 160-197; 94 (1990), no.2, 308-348.
[41] S.J.Gustafson, I.M.Sigal, Mathematical Concepts of Quantum Mechanics, Springer, Berlin, 2003.
[42] Y.Guo, K.Nakamitsu, W.Strauss, Global nite-energy solutions of the Maxwell-Schrodinger sys-
tem, Comm. Math. Phys. 170 (1995), no. 1, 181196.
[43] J.Hale, Asymptotic Behavior of Dissipative Systems, AMS, Providence, 1988.
[44] K.Hannabus, An Introduction to Quantum Theory, Clarendon Press, Oxford, 1997.
[45] W.Heisenberg, Der derzeitige Stand der nichtlinearen Spinortheorie der Elementarteilchen, Acta
Phys. Austriaca 14 (1961), 328-339.
[46] W.Heitler, The Quantum Theory of Radiation, Oxford University Press, Oxford, 1954.
[47] D.Henry, Geometric Theory of Semilinear Parabolic Equations, Lecture Notes in Mathematics,
840, Springer, Berlin, 1981.
[48] P.D.Hislop, I.M.Sigal, Introduction to Spectral Theory. With applications to Schrodinger opera-
tors, Springer, NY, 1996.
[49] V.Imaikin, A.Komech, P.Markowich, Scattering of solitons of the Klein-Gordon equation coupled
to a classical particle, Journal of Mathematical Physics 44 (2003), no.3, 1202-1217.
[50] V.Imaikin, A.Komech, H.Spohn, Soliton-like asymptotics and scattering for a particle coupled to
Maxwell eld, Russian Journal of Mathematical Physics 9 (2002), no.4, 428-436.
[51] V.Imaikin, A.Komech, H.Spohn, Scattering theory for a particle coupled to a scalar eld, Journal
of Discrete and Continuous Dynamical Systems 10 (2003), no.1&2, 387-396.
[52] C.Itzykson, J.B.Zuber, Quantum Field Theory, McGraw-Hill, NY, 1980.
[53] Jackson, Classical Electrodynamics, 3rd ed., John Wiley & Sons, New York, 1999.
[54] V.Jaksic, C.-A.Pillet, Ergodic properties of classical dissipative systems. I. Acta Math. 181 (1998),
no.2, 245-282.
[55] M.Jammer, The Conceptual Development of Quantum Mechanics, McGraw-Hill, NY, 1966.
[56] P.Joly, A.Komech, O.Vacus, On transitions to stationary states in a Maxwell-Landau-Lifschitz-
Gilbert system, SIAM J. Math. Anal. 31 (1999), no.2, 346-374.
[57] T.Kato, A.Jensen, Spectral properties of Schrodinger operators and time-decay of the wave func-
tions, Duke Math. J. 46 (1979), 583-611.
[58] W.Kaumann, Ann. Physik 61 (1897), 544.
[59] A.Komech, On stabilization of string-nonlinear oscillator interaction, J. Math. Anal. Appl. 196
(1995), 384-409.
[60] A.Komech, On transitions to stationary states in Hamiltonian nonlinear wave equations, Phys.
Letters A 241 (1998), 311-322.
228 BIBLIOGRAPHY
[61] A.Komech, On transitions to stationary states in one-dimensional nonlinear wave equations, Arch.
Rat. Mech. Anal. 149 (1999), no.3, 213-228.
[62] A.Komech, Attractors of nonlinear Hamiltonian one-dimensional wave equations, Russ. Math.
Surv. 55 (2000), no.1, 43-92.
[63] A.Komech, On attractor of a singular nonlinear U(1)-invariant Klein-Gordon equation, p. 599-611
in: Proceedings of the 3
r
d ISAAC Congress, Freie Universitat Berlin, Berlin, 2003.
[64] A.Komech, On Global Attractors of Hamilton Nonlinear Wave Equations, preprint 2005.
[65] A.Komech, M.Kunze, H.Spohn, Eective Dynamics for a mechanical particle coupled to a wave
eld, Comm. Math. Phys. 203 (1999), 1-19.
[66] A.Komech, H.Spohn, Soliton-like asymptotics for a classical particle interacting with a scalar
wave eld, Nonlinear Analysis 33 (1998), no.1, 13-24.
[67] A.Komech, H.Spohn, Long-time asymptotics for the coupled Maxwell-Lorentz equations, Comm.
Partial Di. Equs. 25 (2000), no.3/4, 558-585.
[68] A.Komech, H.Spohn, M.Kunze, Long-time asymptotics for a classical particle interacting with a
scalar wave eld, Comm. Partial Di. Equs. 22 (1997), no.1/2, 307-335.
[69] A.Komech, N.Mauser, A.Vinnichenko, On attraction to solitons in relativistic nonlinear wave
equations, to appear in Russian J. Math. Phys.
[70] A.Lande, Zs. f. Phys. 5 (1921), 5.
[71] V.G.Levich, Yu.A.Vdovin, V.A.Mamlin, Course of Theoretical Physics, v. II, Nauka, Moscow,
1971. [Russian]
[72] E.H.Lieb, The Stability of Matter: from Atoms to Stars, Springer, Berlin, 2001.
[73] H.A.Lorentz, The Theory of Electrons and its Applications to the Phenomena of Light and
Radiant Heat. Lectures from a course held at Columbia University, New York, NY, USA, March
and April 1906, Repr. of the 2nd ed., Sceaux: ditions Jacques Gabay, 1992.
[74] J. von Neumann, Mathematical Foundations of Quantum Mechanics, Princeton University Press
Princeton, 1955.
[75] R.Newton, Quantum Physics, Springer, NY, 2002.
[76] E.Noether, Gesammelte Abhandlungen. Collected papers, Springer, Berlin, 1983.
[77] S.P.Novikov, S.V.Manakov, L.P.Pitaevskii, V.E.Zakharov, Theory of Solitons: The Inverse Scat-
tering Method, Consultants Bureau [Plenum], New York, 1984.
[78] J.Perrin, New experiments on the cathode rays, Nature, 53 (1896), 298-299.
[79] C.-A.Pillet, C.E.Wayne, Invariant manifolds for a class of dispersive, Hamiltonian, partial dier-
ential equations, J. Dier. Equations 141 (1997), No.2, 310-326.
[80] R.Pyke, A.Soer, M.I.Weinstein, Asymptotic stability of the kink soliton of the
4
nonlinear
equation, preprint, 2003.
BIBLIOGRAPHY 229
[81] M.Reed, B.Simon, Methods of Modern Mathematical Physics, Academic Press, NY, I (1980), II
(1975), III (1979), IV (1978).
[82] J.J.Sakurai, Advanced Quantum Mechanics, Addison-Wesley, Reading, Massachusets, 1967.
[83] G.Scharf, Finite Quantum Electrodynamics. The Causal Approach, Springer, Berlin, 1995.
[84] L.Schi, Quantum Mechanics, McGraw-Hill, NY, 1955.
[85] E.Schrodinger, Quantisierung als Eigenwertproblem, Ann. d. Phys. I, II 79 (1926) 361, 489; III
80 (1926) 437; IV. 81 (1926) 109.
[86] M.A.Shubin, Pseudodierential Operators and Spectral Theory, Springer-Verlag, Berlin, 2001.
[87] A.Soer, M.I.Weinstein, Multichannel nonlinear scattering for nonintegrable equations, Comm.
Math. Phys. 133 (1990), 119-146.
[88] A.Soer, M.I.Weinstein, Multichannel nonlinear scattering for nonintegrable equations. II. The
case of anisotropic potentials and data. J. Dierential Equations 98 (1992) no. 2 376.
[89] A.Sommerfeld, Atombau und Spektrallinien, Vol. I and II, Friedr. Vieweg & Sohn, Brounschweig,
1951.
[90] A.Sommerfeld, G.Schur, Ann. d. Phys. 4 (1930), 409.
[91] H.Spohn, Dynamics of Charged Particles and Their Radiation Field, Cambridge University Press,
Cambridge, 2004.
[92] W.A.Strauss, Nonlinear Invariant Wave Equations, Lecture Notes in Phys.73 (1978), Springer,
Berlin, 197-249.
[93] C.Sulem, P.L.Sulem, The Nonlinear Schrodinger Equation. Self-Focusing and Wave Collapse,
Springer, NY, 1999.
[94] A.Szabo, N.S.Ostlund, Modern Quantum Chemistry: Introduction to Advanced Electronic Struc-
ture Theory, Dover, 1996.
[95] R.Temam, Innite-Dimensional Dynamical Systems in Mechanics and Physics, Applied Mathe-
matical Sciences, Springer, Berlin, 1988.
[96] B.Thaller, The Dirac Equation, Springer, Berlin, 1991.
[97] L.H.Thomas, Nature 117 (1926), 514.
[98] J.J.Thomson, Phil. Mag. 44 (1897), 298.
[99] B.R.Vainberg, Asymptotic Methods in Equations of Mathematical Physics, Gordon and Breach,
New York, 1989.
[100] G.Wentzel, ZS. f. Phys. 43 (1927), 1, 779.
[101] H.Weyl, The Theory of Groups and Quantum Mechanics, Dover, NY, 1949.
[102] E.P.Wigner, Theory of Group, Academic Press, NY, 1959.
[103] E.Zeidler, Applied Functional Analysis. Main Principles and Their Applications, Applied Math-
ematical Sciences Vol.109, Springer, Berlin, 1995.

Lectures On Quantum Mechanics

Uploaded by

Copyright:

Lectures On Quantum Mechanics

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Lectures On Quantum Mechanics

Uploaded by

Copyright:

a

(x)) correspond to the

(1901) inspired by the comparizon

of the corresponding small parameter, and the central role of Quantum

which is close to the value of J.J.Thomson.

with the energy of the absorbed photon with frequency in accordance

and the discreteness of the energies E

= , where is proportional to the magnetic

dierent from the wavelength

and the amplitudes (x) of the stationary states.

(x, (x), (x))

(x, (x), (x)) = 0, x IR

are considered as independent variables with the

(x) are dened by

(x, (x), (x)), = 0, ..., 3. (2.8)

(x, (x), (x)), x IR

(x) etc. Also = 0, 1, 2, 3

is the set of the functions (t, x) C

the Gateau derivative DT() is the linear functional

if the derivative exists.

(x, , )[ C([[ +[[)

, [[+[[ const. (2.20)

, > 3/2 dened by

by (2.18) and the rst inequality of

(x, (x), (x))h(x) +

(x, (x), (x))

satises the Hamilton Least Action Principle (LAP) if for any

with > 3/2 the Hamilton LAP is equivalent to the Euler-Lagrange

(x, (x), (x))h(x)

(x, (x), (x))

is dened by (cf. (36.5), (36.17))

since > 3/2.

(t, x) = (t, x), x IR

of the equations (2.9) the energy is conserved, E(t) = const.

(t, x), x IR. (3.6)

(t, x), we get by partial integration

(t, R). (3.9)

(x, (x), (x))

(x, (x), (x))

(t, x) (t, x) dx, t IR. (3.17)

of Eq. (2.9) the corresponding component of momentum is

(t, x)dx. (3.18)

(t, x), we get,

with > 2, by (2.18) and (2.20).

of Eq. (2.9), the corresponding component of angular momentum is conserved, J

by (2.18) and (2.20) since > 3/2.

of Eq. (2.9) the charge is conserved, Q(t) = const .

(t, x) i(t, x)dx. (3.27)

(t, x) i(t, x)dx +

(t, x) i(t, x). (3.29)

(t, x) +F(), x IR, (3.30)

A(t, x)] = 0 hence E(t, x) +

A(t, x), (t, x) IR

(t, x) + div A(t, x) = 0, (t, x) IR

A(t, x). (4.7)

(x) = ((t, x), A(x)), /

(x) = ((x), A(x)),

(x) are equiv-

is a quadratic form in /. Therefore, the canonically conjugate elds

(x), hence (4.16) is equivalent to (2.9).

(t, k) hence (5.1) becomes

(t, x). (5.8)