Accidental Symmetry in Quantum Physics

Munir Al-Hashimi
Institute for Theoretical Physics
Bern University

May 13, 2008

2

Abstract
In this thesis we will study examples of accidental symmetry in quantum physics.
An indicator for the existence of conserved quantities is that all bound classical orbits
are closed in the corresponding classical system. In analogy of Kepler problem we
study accidental symmetry by constructing vectors that play the role of the Runge-
Lenz vector. First we consider a particle moving on a cone and bound to its tip
by 1/r or r 2 potentials. When the deficit angle of the cone divided by 2π is a
rational number, all bound classical orbits are closed. Correspondingly, the quantum
system has accidental degeneracies for the discrete energy spectrum. An accidental
SU(2) symmetry is generated by the rotations around the tip of the cone as well
as by a Runge-Lenz vector. Remarkably, some of the corresponding multiplets have
fractional “spin” and unusual degeneracies in both potentials.

A classical particle in a constant magnetic field moves in cyclotron motion on a

circular orbit. At the quantum level, againhere, all classical orbits are closed, and
this gives rise to degeneracies in the spectrum. It is well-known that the spectrum
of a charged particle in a constant magnetic field consists of infinitely degener-
ate Landau levels. Here, one also expects some hidden accidental symmetry with
infinite-dimensional representations. The position of the center of the cyclotron
circle plays the role of a Runge-Lenz vector. After identifying the corresponding
accidental symmetry algebra, we re-analyze the system in a finite periodic volume.
Interestingly, similar to the quantum mechanical breaking of CP invariance due to
the θ-vacuum angle in non-Abelian gauge theories, quantum effects due to two self-
adjoint extension parameters θx and θy explicitly break the continuous translation
invariance of the classical theory. This reduces the symmetry to a discrete magnetic
translation group and leads to finite degeneracy. Similar to a particle moving on
a cone, a particle in a constant magnetic field shows a very peculiar realization of
accidental symmetry in quantum mechanics.
Contents

1 Introduction 7

2 An Introduction to Symmetry in Physics 13

2.1 Symmetry in Classical Mechanics . . . . . . . . . . . . . . . . . . . . 13

2.1.1 The Casimir Operator of the Symmetry Group . . . . . . . . . 15

2.2 The Symmetries of Some Quantum Systems . . . . . . . . . . . . . . 15

2.2.1 The Kepler Problem . . . . . . . . . . . . . . . . . . . . . . . 16

2.2.2 Particle Moving on a Plane under the Influence of a Constant

Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3 Energy Levels and Degeneracy from just Symmetry . . . . . . . . . . 18

2.3.1 The Kepler Problem . . . . . . . . . . . . . . . . . . . . . . . 20

2.3.2 Particle under the Influence of a Magnetic Field . . . . . . . . 21

3 Accidental Symmetry for Motion on a Cone 25

3.1 Introduction to Motion on a Cone . . . . . . . . . . . . . . . . . . . 25

3.2 The Domains of Operators and Hermiticity . . . . . . . . . . . . . . . 27

3.2.1 The Hermiticity and Self-Adjointness of T . . . . . . . . . . . 29

3.3 The 1/r Potential on a Cone . . . . . . . . . . . . . . . . . . . . . . . 32

3.3.1 Classical Solutions . . . . . . . . . . . . . . . . . . . . . . . . 32

3
4 CONTENTS

3.3.2 Semi-classical Bohr-Sommerfeld Quantization . . . . . . . . . 34

3.3.3 Solution of the Schrödinger Equation . . . . . . . . . . . . . . 34

3.3.4 Runge-Lenz Vector and SU(2) Algebra . . . . . . . . . . . . . 35

3.4 Runge-Lenz Vector at the Quantum Level . . . . . . . . . . . . . . . 37

3.4.1 Casimir Operator . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.4.2 The Raising and Lowering Operator Acting on the Wave Func-
tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.4.3 Domains and Hermiticity of the Components of the Runge-

Lenz Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.4.4 Unusual Multiplets . . . . . . . . . . . . . . . . . . . . . . . . 46

3.4.5 Counting the Degeneracies in the Kepler Case . . . . . . . . . 49

3.5 The r 2 Potential on a Cone . . . . . . . . . . . . . . . . . . . . . . . 50

3.5.1 Classical Solutions . . . . . . . . . . . . . . . . . . . . . . . . 50

3.5.2 Semi-classical Bohr-Sommerfeld Quantization . . . . . . . . . 51

3.5.3 Solution of the Schrödinger Equation . . . . . . . . . . . . . . 52

3.5.4 Runge-Lenz Vector . . . . . . . . . . . . . . . . . . . . . . . . 53

3.5.5 The Runge-Lenz Vector as a Quantum Mechanical Operator . 54

3.5.6 Casimir Operator . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.5.7 The Raising and Lowering Operators Acting on the Wave

Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.5.8 Unusual Multiplets . . . . . . . . . . . . . . . . . . . . . . . . 61

3.5.9 Counting Degeneracies in the Case of the Isotropic Harmonic

Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4 Particle in Constant Magnetic Field 65

4.1 Introduction to Cyclotron Motion . . . . . . . . . . . . . . . . . . . . 65

4.2 Particle in the Infinite Volume . . . . . . . . . . . . . . . . . . . . . . 66

CONTENTS 5

4.2.1 Classical Treatment . . . . . . . . . . . . . . . . . . . . . . . . 66

4.2.2 Semi-classical Treatment . . . . . . . . . . . . . . . . . . . . . 68

4.2.3 Quantum Mechanical Treatment . . . . . . . . . . . . . . . . . 69

4.2.4 Creation and Annihilation Operators . . . . . . . . . . . . . . 72

4.2.5 Alternative Representation of the Hamiltonian . . . . . . . . . 73

4.2.6 Energy Spectrum and Energy Eigenstates . . . . . . . . . . . 73

4.2.7 Coherent States . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.3 Particle on a Torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.3.1 Constant Magnetic Field on a Torus . . . . . . . . . . . . . . 76

4.3.2 Discrete Magnetic Translation Group . . . . . . . . . . . . . . 79

4.3.3 Spectrum and Degeneracy on the Torus . . . . . . . . . . . . . 82

4.3.4 Coherent States on the Torus . . . . . . . . . . . . . . . . . . 84

4.3.5 Calculating the Expectation Value of Txl and Tyl . . . . . . . . 88

5 Conclusions 93
6 CONTENTS
Chapter 1

Introduction

Symmetry is a fundamental feature of the physical world. It plays a pivotal role

in studying classical systems since the beginning of the development of classical
mechanics because it leads to the identification of conserved quantities of classical
systems. Geometrical symmetries of space itself lead to conserved dynamical quan-
tities. For example, the assumption that space is homogeneous or, in other words,
has translation symmetry, leads to the conservation of linear momentum of a closed
isolated system, and isotropy of space leads to conservation of angular momentum
of an isolated system. On the other hand, if the system has time translation sym-
metry, then the total energy is conserved [1]. For a particle in an external potential
the status of the space symmetry is decided by the symmetry of the potential under
which the particle is moving. Therefore the symmetry of the potential dictates the
conserved quantities. For example, a particle moving under the influence of a spher-
ically symmetric potential has a conserved angular momentum. The importance of
the relation between symmetry and conserved quantities led to the development of
theorems like the Noether theorem that allow us to obtain these conserved quantities
from the knowledge of symmetry.

In quantum mechanics symmetry also leads to conserved quantities which man-

ifest themselves by good quantum numbers. States of different quantum numbers
with the same energy lead to degeneracy. We can understand degeneracy in classical
mechanical terms as different possible trajectories that belong to the same energy.
For example, rotating the plane of motion of a particle moving under the influence
of a spherically symmetric potential (a 1/r potential, for example) by an angle α
does not change the energy. Changing the plane of rotation means changing the
direction of the angular momentum vector L ~ by an angle α without changing the
~ There is an infinite number of trajectories that belong to the same
length of L.
energy. Analogously, in quantum mechanics the value of the z-component of the

7
8 CHAPTER 1. INTRODUCTION

angular momentum Lz = m~, with m being the integer magnetic quantum number,
characterizes different degenerate states. Different values of Lz do not affect the en-
ergy since Lz commutes with the Hamiltonian operator H for spherically symmetric
potentials. In contrast to the classical case, in quantum mechanics the degeneracy
is a finite number, equal to 2l + 1 where l is the integer-valued angular momentum
quantum number.

There is another kind of symmetry in addition to the geometrical symmetry of

space itself. This is a dynamical symmetry that results from special features of the
Lagrangian of the system, and not from homogeneity or isotropy of space. This type
of symmetry is also known as accidental symmetry. It is called accidental because
the system possesses such a symmetry only due to special values of some parameters
in the Lagrangian of the system, or due to a special functional form of the potential
of the system. For example, for spherically symmetric potentials of the form r α ,
only for the special values α = −1 and α = 2 there is accidental symmetry. Another
example is a particle moving on a cone. Only for special values of the deficit angle
δ of the cone there is accidental symmetry. Interestingly, there is a relation between
accidental symmetry and the fact that the bound classical orbits are closed. The
uniqueness of the 1/r and r 2 potentials was realized by Bertrand in 1873. He proved
that they are the only spherically symmetric scalar potentials in Euclidean space
for which all bound orbits are closed [2]. Conserved quantities due to geometrical
symmetry in discrete systems can usually be easily identified. It is sufficient to find
the cyclic canonical coordinate in the Lagrangian. Then the conserved quantity is
just the conjugate canonical momentum, i.e.
∂L
= ṗc = 0, (1.0.1)
∂qc
which implies that pc is constant. In the case of accidental symmetry, the conserved
quantities are not always so easy to identify. For the Kepler problem, after some
calculations involving Newton’s second law, it is relatively easy to find a conserved
quantity that is called the Runge-Lenz vector [3]. It is the three-vector pointing
from the center of the force towards the perihelion of the elliptic orbit of the parti-
cle. On the other hand, for other systems with accidental symmetries, finding the
conserved quantities may be a complicated process. For example, this is the case
for the problem of a particle moving under the influence of a 1/r potential (the
Kepler problem), or an r 2 potential (the isotropic harmonic oscillator) modified by
an angular momentum-dependent potential of the form

L2 (λ2 − 1)
V (r) = , (1.0.2)
2Mr 2
where M is the mass of the particle, and λ is a constant. It can be proved that for
λ = p/q a rational number, all classical orbits are closed. Finding the conserved
 9

quantities for a modified Kepler problem is highly nontrivial, and was studied in
[4], [5]. In fact, in [4] it was argued that the conserved quantities are tensors of
ranks depending on the value of the constant λ. Other later attempts investigated
the possibility of finding a general scheme to identify accidental symmetry from the
expression of the Hamiltonian [6]. Here I refer to unpublished work by Wiese in 1982
who studied the problem of the modified Kepler problem quantum mechanically [7].

Any problem of a particle moving under the influence of a spherically symmetric

potential modified by an angular momentum-dependent potential of the form in
Eq.(1.0.2) in 2-dimensions is equivalent to a particle moving on a 2-dimensional
cone with the center of force at the tip of the cone. The problem of a particle
moving on a cone under the influence of 1/r and r 2 potentials centered at the tip
of the cone was studied previously. A solution of the Schrödinger equation for
this problem was given for both cases of 1/r and r 2 potentials [8]. However, the
accidental symmetry of such systems has never been studied. In this thesis, from
which some results have already been published [9], the accidental symmetry of the
problem will be investigated classically and quantum mechanically. The first part
of the work is to construct the conserved quantities. They are identified as two
components of what is equivalent to a Runge-Lenz vector, as well as one component
of the angular momentum vector. From the commutation relations between these
quantities, one finds that they generate an SU(2) symmetry for both potentials.
What makes this problem particularly interesting is the value of the Casimir operator
which behaves like a spin operator. Unlike the usual value of an integer or a half-
integer of the Casimir spin that one finds in all previously studied quantum systems,
the Casimir spin for this problem can take any value. However, in order to have
extra degeneracies one must have s = p/q = 1 − δ/2π as a rational number. The
Casimir spin for the Kepler problem is S = nr + |m|/s, where nr = 0, 1, 2, ... is
the radial quantum number, and δ is the deficit angle of the cone. Accordingly,
in this case the Casimir spin can be a fractional number. This is unusual for an
SU(2) representation. What makes this SU(2) representation so special? We can
understand this by recalling the basic procedure by which the spectrum of other
systems was derived. Take, for example, a general spherically symmetric potential
problem in quantum mechanics. If we take the angular momentum components
as generators, then the Casimir operator for this case is nothing but L ~ 2 . Deriving
the spectrum of this operator involves obtaining the matrix elements of the raising
operator L+ = Lx + iLy , and the lowering operator L− = Lx − iLy , and using
eigenfunctions of Lz as a basis. The whole treatment is based on the assumption
that the functions are normalizable and constitute a complete set. (This will be
explained in more details in section (2.3)). For motion on a cone the situation is
different. The generators Rx and Ry have different features because the Runge-Lenz
~ which is in general different from the domain of
vector is Hermitian in a domain D[R]
the Hamiltonian D[H]. Therefore the raising and lowering operators R± = Rx ± iRy
10 CHAPTER 1. INTRODUCTION

can take a wave function outside D[H]. Thus the usual treatment for finding the
spectrum of the Casimir operator and the degeneracies does not apply. The issue
of the domains of the operators in a corresponding Hilbert space, and the issue
of Hermiticity versus self-adjointness plays a role in this thesis. In a usual SU(2)
representation the number of wave functions in the multiplet is finite and can be
found by a purely algebraic treatment (see section (2.3)). Here it is not possible to
do that. Counting the number of wave functions in a multiplet requires a different
technique. The counting is done by direct application of the explicit form of the
raising and lowering operators on the wave function. After lengthy calculations, it
can be shown that the multiplet terminates for both 1/r and r 2 potentials when
S is an integer or a half-integer, while it never terminates otherwise, and then the
multiplet contains an infinite number of wave functions. Not all the wave functions
in the multiplet are physical. To count the degeneracies we must count only the
physical wave functions. Therefore counting degeneracies is also a different process
than the one in the usual SU(2) representations. In our case, it requires finding a
rule by induction that relates the degeneracy g to the Casimir spin S, as well as to
p and q which are the parameters of s = p/q. The same argument applies to the
case of the isotropic harmonic oscillator.

Another important case of accidental symmetry is given by a particle moving

under the influence of a constant magnetic field undergoing cyclotron motion on a
circular orbit. Here the classical bound orbits are also closed, and again this leads to
conserved quantities. The conserved quantity is a vector that is pointing from the
origin to the center of the circular orbit. Since the coordinates of the center of the
circle do not change with time, they are conserved quantities and play the role of
the Runge-Lenz vector. In the case of infinite space we have an additional geomet-
rical symmetry, the rotational symmetry that leads to a conserved one-component
angular momentum. The Poisson brackets (as well as the commutator) for these
three generators do not give an SU(2) symmetry because {Rx , Ry } = c, where Rx ,
Ry are the components of the Runge-Lenz vector, and c is a constant (see sub-
section(2.2.2)). The generators constitute the Heisenberg group [10]. The radius
of the circle r is a constant of motion since it is proportional to the total energy
E = Mr 2 ω 2 /2. Quantum mechanically the fact that Rx , and Ry do not commute
implies that we cannot measure the components of the center of the circle simulta-
neously with absolute precession, although the radius of the circle can be measured
with absolute precession. In the cyclotron problem, translation invariance disguises
itself as an “accidental” symmetry. As a consequence, the symmetry multiplets —
i.e. the Landau levels — are infinitely degenerate. In order to further investigate
the nature of the accidental symmetry, in [11] the charged particle in the magnetic
field was coupled to the origin by an r 2 harmonic oscillator potential. This ex-
plicitly breaks translation invariance and thus reduces the degeneracy to a finite
amount, while rotation invariance remains intact. We put the system on a torus
 11

by imposing a boundary condition over a rectangular region Lx × Ly , which leads

to a quantization condition for the magnetic flux. This leads to interesting results
regarding the degeneracy of the system, which turns out to be given by nΦ , the
number of magnetic flux quanta. This explicitly breaks rotation invariance, while
leaving translation invariance (and hence the accidental symmetry)intact. Remark-
ably, the Polyakov loops, which are a consequence of the non-trivial holonomies of
the torus, give rise to non-trivial Aharonov-Bohm phases which are observable at
the quantum level but not at the classical level. In analogy to the quantum me-
chanical breaking of CP invariance due to the θ-vacuum angle in non-Abelian gauge
theories, here two self-adjoint extension parameters θx and θy explicitly break the
continuous translation invariance of the classical system down to a discrete magnetic
translation group. This reduces the degeneracy to a finite number, and allows us to
further investigate the nature of the accidental symmetry. In particular, just like for
motion on a cone, symmetry manifests itself in a rather unusual way in this quantum
system. In particular, due to its relevance to the quantum Hall effect, the Landau
level problem has been studied very extensively (for a recent review see [12]). For
example, the problem has already been investigated on a torus in [13, 14], however,
without elaborating on the accidental symmetry aspects. In the second part of this
thesis we concentrate exactly on those aspects, thus addressing an old and rather
well-studied problem from an unconventional point of view.
12 CHAPTER 1. INTRODUCTION
Chapter 2

An Introduction to Symmetry in
Classical and Quantum Physics

In this chapter we will highlight the importance of symmetry in studying classical

and quantum mechanics. Symmetry plays a pivotal role in understanding quantum
mechanics. In most cases the symmetry of a quantum system is the same as the
one of the corresponding classical system. That is aside from anomalies which break
some of the symmetries of the quantum system. However, such cases are beyond
the scope of this thesis.

In the beginning of this chapter we will discuss symmetry in classical physics, and
briefly discuss the symmetry of some classical systems. After that we discuss in more
details the symmetries of quantum systems. Two systems will be studied in detail,
the Kepler problem and the Landau level problem. Through the discussions, we will
distinguish between two kinds of symmetries, dynamical symmetry and geometrical
symmetry.

The last section concentrates on one of the important benefits of studying sym-
metry. We will show that the knowledge of symmetry leads to the energy spectrum
and degeneracies without even solving the Schrödinger equation.

2.1 Symmetry in Classical Mechanics

A continuous symmetry of a classical system is generated by a number of symmetry

generators. A generator is a mechanical variable that can be written in terms of
canonical momenta pk and coordinates qk , and has a vanishing Poisson bracket
with the classical Hamilton function H. Let us consider a classical system with n

13
14 CHAPTER 2. AN INTRODUCTION TO SYMMETRY IN PHYSICS

generators, and let us denote the ith generator by Gi (pk , qk ). Then

 
dGi X ∂H ∂G ∂H ∂G
{H, Gi } = = − = 0, (2.1.1)
dt k
∂qk ∂pk ∂pk ∂qk

which implies that Gi is a constant of motion.

The Poisson bracket of two generators of some classical system may constitute
a Lie group bracket relation of the form [3]
n
X
{Gi , Gj } = fijk Gk , (2.1.2)
k=1

where fijk are the so-called structure constants, and n is the number of generators of
the classical system. (From now on we will use the Einstein summation convention
unless stated otherwise). For the rotation group SO(3) we have fijk = ǫijk .

Consider a particle with mass M moving under the influence of spherically sym-
metric potential in three dimensions. In this case, we have three generators that
represent the three components of the angular momentum vector L. ~ The generators
constitute an SO(3) algebra. In the case of the 1/r potential, the so-called Kepler
problem, it is easy to prove that the three components of the vector

R ~ − Mκ~r ,
~ = p~ × L (2.1.3)
r
are also conserved. Here κ is the strength of the 1/r potential. This vector is known
as the Runge-Lenz vector and it will play a major role in this thesis.
~e
~ to R,
Let us re-scale R where

~e 1 ~
R =√ R. (2.1.4)
−2ME
Here E is the total energy of the particle. It is straightforward to obtain the following
Lie group brackets

{Li , Lj } = ǫijk Lk , ei , Lj } = ǫijk R

{R ek , ei , R
{R ej } = ǫijk R
ek . (2.1.5)

The structure constants in the above equations can be identified as the ones of
SO(4). Here it should be noted that this is only valid for the bound Kepler problem
in which the total energy is negative. For the unbound Kepler problem E is positive
~e
and R is a vector with purely imaginary components. This changes the structure
constants in the above equations.
2.2. THE SYMMETRIES OF SOME QUANTUM SYSTEMS 15

Another important example is the isotropic harmonic oscillator. In this case the
potential is given by V (r) = Mω 2 r 2 /2, where ω is the angular frequency of the os-
cillation. Here, in addition to the conserved components of the angular momentum,
there are other conserved quantities. For example, in two dimensions they are the
components of a second-rank tensor given by
1
Qij = (pi pj + M 2 ω 2 xi xj ). (2.1.6)
2M
From the components of this tensor we construct the following conserved quantities
Q12 + Q21 Q22 − Q11 L xpy − ypx
G1 = , G2 = , G3 = = . (2.1.7)
2ω 2ω 2 2
After using Eqs.(2.1.6) and Eqs.(2.1.7), it can be proved that the values of the
Poisson brackets are given by the following relation

{Gi , Gj } = ǫijk Gk . (2.1.8)

The structure constants of the above Lie bracket identify the symmetry group as
SO(3). It is worth noting that SO(3) is homomorphic to the group SU(2).

2.1.1 The Casimir Operator of the Symmetry Group

In general, for a semi-simple Lie algebra with n generators one can define a quadratic
Casimir operator as
Xn
C= ai G2i , (2.1.9)
i=1

where ai are constants that can be chosen such that C commutes with all generators
Gi . The number of independent Casimir operators is given by the rank of the group,
i.e. by the number of commuting generators. For example, in the case of spherically
symmetric potentials in three dimensions with a geometric SO(3) symmetry only,
the only Casimir operator is the square of the length of the angular momentum
vector L~ 2 = L2 + L2 + L2 .
x y z

2.2 The Symmetries of Some Quantum Systems

In this section we will discuss the symmetries of some important quantum systems
by finding the conserved quantities for each of these systems. This leads to the
definition of the generators of the symmetry group. For a quantum system the
16 CHAPTER 2. AN INTRODUCTION TO SYMMETRY IN PHYSICS

classical dynamical variables will be replaced by the corresponding operators. The

symmetry will be studied here by replacing the Poisson brackets by commutators
and the classical generators will be replaced by their corresponding operators. For
example, the Poisson brackets of the symmetry group of a particle moving in a
spherically symmetric potential in the first equation of Eqs.(2.1.5) will be written
for the corresponding quantum mechanical system as
[Li , Lj ] = iǫijk Lk . (2.2.1)
In this thesis we will use natural units in which ~ = 1, c = 1.

2.2.1 The Kepler Problem

For the Kepler problem the Hamiltonian takes the form

 
∆ κ −1 2 2 1 2 cot2 θ csc2 θ 2 κ
H=− − = ∂r + ∂r + 2 ∂θ + 2
∂θ + 2
∂ϕ − . (2.2.2)
2M r 2M r r r r r
The expression of the Runge-Lenz vector as an operator can be obtained from
Eq.(2.1.3). There is an ordering ambiguity in the first term. We can overcome
this difficulty by writing
 
R~ = 1 p~ × L
~ −L~ × p~ − Mκ~r . (2.2.3)
2 r
In order to verify that the components of the above operator are constants of motion
we must prove that they commute with the Hamiltonian in Eq.(2.2.2). This can be
done, for example, by writing the components of the Runge-Lenz vector in terms of
~ and ~p in spherical coordinates, and then
explicit expression for the components of L
proving that the commutator is zero for all three components of R. ~ The calculation
either can be done by hand, or by a Mathematica program that we constructed
which calculates the commutator for any two operators. The components of L ~ in
spherical coordinates are
Lx = i (cot θ cos ϕ∂ϕ + sin ϕ∂θ ) ,
Ly = i (cot θ sin ϕ∂ϕ − cos ϕ∂θ ) ,
Lz = −i∂ϕ , (2.2.4)
and of ~p in spherical coordinates are
 
cos ϕ cos θ sin ϕ
px = −i ∂ϕ + sin θ sin ϕ∂r + ∂θ ,
r sin θ r
 
sin ϕ cos θ cos ϕ
py = −i − ∂ϕ + sin θ cos ϕ∂r + ∂θ ,
r sin θ r
 
sin θ
pz = −i cos θ∂r − ∂θ . (2.2.5)
r
2.2. THE SYMMETRIES OF SOME QUANTUM SYSTEMS 17

Substituting the values of the components of L ~ and p~ in Eq.(2.2.3) one then calcu-
lates the commutator with the Hamiltonian. The result is zero for all components.
This proves that the components of R ~ are indeed constants of motion. In addition
to this, one can prove that the components of L ~ commute with the Hamiltonian of
Eq.(2.2.2). Accordingly, in total we have six generators. This gives the following
commutation relations
[Li , Lj ] = iǫijk Lk , (2.2.6)

ei , Lj ] = iǫijk R
[R ek , (2.2.7)

ei , R
[R ej ] = iǫijk R
ek , (2.2.8)
where we again have
~e 1 ~
R =√ R. (2.2.9)
−2ME
It is also clear here that the symmetry group is SO(4). This can be realized by
labelling R ex = L14 , R
ey = L24 , R
ez = L34 , and L
~ = (L23 , L31 , L12 ). The extension of
the definition of the components of the angular momentum Lij = xi pj −xj pi to i, j =
1, 2, 3, 4 gives six generators of the group of proper rotations in four dimensions.
However, we must keep in mind that unlike the purely geometrical SO(3) symmetry,
the SO(4) symmetry is an accidental dynamical symmetry because r4 and p4 are not
a real coordinate and momentum of the system in a fourth dimension. Therefore
this symmetry is considered to be a dynamical symmetry instead of a geometrical
one.

2.2.2 Particle Moving on a Plane under the Influence of a

Constant Magnetic Field

Consider a particle moving on an infinite 2-dimensional plane under the influence of

a constant magnetic field. When the direction of the magnetic field is perpendicular
to the plane of motion, the Hamiltonian of the system takes the form
1

H=− ∂x2 + ∂y2 + 2ieBx∂y − (eBx)2 . (2.2.10)
2M
Here we have used the asymmetric gauge with A ~ = (0, Bx, 0). (This problem will
be discussed in more details in chapter 4). It can be proved that there are three
conserved quantities that commute with the above Hamiltonian. They are
eB eB
L = x(py + x ) − y(px + y ), (2.2.11)
2 2
18 CHAPTER 2. AN INTRODUCTION TO SYMMETRY IN PHYSICS

py px
Rx = − , Ry = y + , (2.2.12)
Be eB
where B is the magnetic field. As it is clear from Eq.(2.2.11), L is the component
of a conserved quantity that can be identified with the angular momentum. The
quantities Rx and Ry represent two components of a vector that is pointing towards
the center of what is classically a circular orbit of the particle on the plane (as we
will explain later in chapter 4). The commutation relations for these three operators
are given by

ex ] = −iR
[L, R ey , ey ] = iR
[L, R ex , ex , R
[R ey ] = i, (2.2.13)
p p
where Rex = − 1/BeRx and R ey = 1/BeRy . From the structure constants in the
above equations we can identify the group as the Heisenberg group.

2.3 Energy Levels and Degeneracy from just Sym-

metry

One of the useful aspects of studying the symmetry of a quantum system is that
it may enable us to find the energy levels and degeneracies without solving the
Schrödinger equation [1]. To understand this, consider a Lie algebra with three
generators, each of which being a Hermitian dynamical variable, say the components
of the angular momentum Lx , Ly , and Lz . As we saw in section (2.2) in such a case
these three generators constitute an SU(2) algebra. The Casimir operator for this
case is just
L2 = L2x + L2y + L2z . (2.3.1)
Let us define two operators as

L+ = Lx + iLy ,
L− = Lx − iLy . (2.3.2)

Using the above equations as well as Eq.(2.2.6), we obtain the following commutation
relations

[Lz , L+ ] = L+ , [Lz , L− ] = −L− , [L+ , L− ] = 2Lz . (2.3.3)

An eigenfunction of Lz with eigenvalue m can at the same time be chosen as an

eigenfunction of L2 since these operators commute. Then we have

hl, m|L2 |l′ , m′ i = f (l)δmm′ δll′ , hl, m|Lz |l′ , m′ i = mδmm′ δll′ , (2.3.4)
2.3. ENERGY LEVELS AND DEGENERACY FROM JUST SYMMETRY 19

where l is the quantum number associated with the eigenvalues of L2 . If we assume

that the states |l, mi are normalized and form a complete set, then the first of the
Eqs.(2.3.3) leads to
hl, m|Lz |l,′ m′ ihl′ , m′ |L+ |l′′ , m′′ i − hl, m|L+ |l′ , m′ ihl′ , m′ |Lz |l′′ , m′′ i
= hl, m|L+ |l′′ , m′′ i. (2.3.5)
The above equation together with Eqs.(2.3.3) gives
(m − m′′ − 1)hl, m|L+ |l, m′′ i = 0. (2.3.6)
It is clear from the above equation that hl, m|L+ |l, m′′ i vanishes unless m = m′′ + 1 .
Repeating the same process with the second commutator of Eqs.(2.3.3) we obtain a
vanishing hl, m|L− |l, m′′ i unless m = m′′ − 1. This means that L+ is indeed a raising
operator, while L− is a lowering operator. The only non-vanishing matrix elements
of these two operators are
hl, m + 1|L+ |l, mi = λm , hl, m|L− |l, m + 1i = λ∗m . (2.3.7)
The above Eq.(2.3.7) and the third commutation relation in Eqs.(2.3.3) give
|λm−1 |2 − |λm |2 = 2m. (2.3.8)
The above equation is a first order difference equation. The general solution is
|λm |2 = C − m(m + 1), (2.3.9)
where C is a constant. Since |λm |2 must be positive or zero and m is an integer, we
have two values of m = m1 and m = m2 at which |λm |2 = 0, while it is positive in
between these two values. The solution of the equation C − m(m + 1) = 0 gives
1 1 1 1 1 1
m1 = − + (1 + 4C) 2 , m2 = − − (1 + 4C) 2 . (2.3.10)
2 2 2 2
From the above equations we obtain m2 + 1 = −m1 , which implies that m1 changes
to −m1 by an integer number of steps. This leads to m1 either being an integer or
a half-integer (m1 = 0, 21 , 1, 32 , 2...).

In order to evaluate f (l) we use the relation

1
L2 = (L+ L− + L− L+ ) + L2z . (2.3.11)
2
Using the above equation as well as Eqs.(2.3.4) gives
1
f (l) = (|λm−1 |2 + |λm |2 ) + m2 = C = m1 (m1 + 1) = l(l + 1). (2.3.12)
2
Since m1 is the largest value of m, we can identify it with l. Accordingly, the
eigenvalue of L2 is l(l + 1), and for each value of l we have 2l + 1 values of m.

As we saw here, just knowing the symmetry, we found the spectrum of the
operators Lz , L2 and the degeneracy of states with quantum number l without
solving any eigenvalue equation.
20 CHAPTER 2. AN INTRODUCTION TO SYMMETRY IN PHYSICS

2.3.1 The Kepler Problem

The Kepler problem with κ = Ze2 is the case of the hydrogen atom. We can define
~ and R
two quantities using L ~ [1].

1 ~ ~e ~ = 1 (L ~e
I~ = (L + R), K ~ − R). (2.3.13)
2 2

It is easy to see that the components of I~ and K

~ constitute separately an SU(2)
algebra. In addition, these vectors commute with the Hamiltonian of Eq.(2.2.2) and
they have a vanishing commutator with each other. Therefore we have

[Ii , Ij ] = iǫijk Ik , [Ki , Kj ] = iǫijl Kl . (2.3.14)

From the previous section we know that when generators constitute an SU(2) alge-
bra, then the spectrum of the Casimir operator is l(l+1) (see Eq.(2.3.12)). Therefore
the spectrum of I~2 is i(i + 1) and the spectrum of K~ 2 is k(k + 1). As we said, the
symmetry group for this system is SO(4) = SU(2) ⊗ SU(2) accordingly, there are
two Casimir operators, they are

~ 2 = 1 (L
C = I~2 + K e~2 ),
~2 + R C ′ = I~2 − K
~2 =L ~e
~ · R. (2.3.15)
2

~ in Eq.(2.2.3) and L
~ that L ~e
~ ·R
It can be proved from the expression of R = 0.
′ ~2 ~ 2
Accordingly, C = 0, and this gives I = K , which implies i = k. This result
together with the first of Eqs.(2.3.15) lead to
1
C = I~2 + K
~ 2 = 2k(k + 1), k = 0, , 1, ... (2.3.16)
2
~e ~ from
Substituting for the value of R from Eq.(2.1.4) and using the definition of R
Eq.(2.2.3) we get
1 ~ 2 M e~2 Mκ2 1
C = (L − R )=− − . (2.3.17)
2 2E 4E 2
From the above and Eq.(2.3.15) we obtain the relation between the total energy E
and k as
Mκ2 Mκ2
E=− = − , (2.3.18)
2(2k + 1)2 2n2
where n = 1, 2, 3, ... From the symmetry approach we know from section (2.3)
that the eigenvalues of L~ 2 are l(l + 1) with l = 0, 1 , 1, ... By adding an additional
2
physical restriction that makes l just an integer, and knowing from Eqs.(2.3.13) that
~ = I~ + K,
L ~ this yield l with any value between i + k = 2k − 1 = n − 1 down to
|i − k| = 0. As we found in section (2.3), it was proved using symmetry only that for
2.3. ENERGY LEVELS AND DEGENERACY FROM JUST SYMMETRY 21

each value of l there are 2l + 1 values for Lz = m. Accordingly, the total degeneracy
of an energy level with quantum number n is
n−1
X
g= (2l + 1) = n2 . (2.3.19)
l=0

As we saw, it is possible to find the energy levels, the spectrum of the operator
~ 2 , the spectrum of the operator Lz , and the total degeneracy of an energy level g
L
without solving the Schrödinger equation.

2.3.2 Particle under the Influence of a Magnetic Field

Another important example of using symmetry to calculate energy levels and de-
generacies is the case of a particle moving on an infinite plane under the action of a
magnetic field perpendicular to the plane.

As we mentioned before, in this case the Hamiltonian of the system is given by

Eq.(2.2.10) and the conserved quantities are the angular momentum in Eq.(2.2.11)
and the components of the Runge-Lenz vector given by Eqs.(2.2.12). The Hamilto-
nian in Eq.(2.2.10) can be written as
ω

H= 2L + ω(Rx2 + Ry2 ) . (2.3.20)
2
The energy levels and their degeneracies can be calculated using a similar approach
as the one that has been used in section (2.3). To do this, we first define the raising
and lowering operators

R+ = Rx + iRy , R− = Rx − iRy . (2.3.21)

Using the commutation relations in Eqs.(2.2.13) and the above equations give

2M
[L, R+ ] = R+ , [L, R− ] = −R− , [R+ , R− ] = .
ω
(2.3.22)
Using basis states that are eigenfunctions of L gives the matrix element

hm|L|m′ i = mδmm′ . (2.3.23)

Assuming that the basis states are normalized, the first commutator in Eqs.(2.3.22)
gives

hm|L|m′ ihm′ |R+ |m′′ i − hm|R+ |m′ ihm′ |L|m′′ i = hm|R+ |m′′ i. (2.3.24)
22 CHAPTER 2. AN INTRODUCTION TO SYMMETRY IN PHYSICS

From the above equation and Eq.(2.3.23) we get

(m − m′′ − 1)hm|R+ |m′′ i = 0. (2.3.25)

This means that the only non-vanishing matrix elements are the ones with m′′ =
m − 1, and R+ is indeed a raising operator. Using the same method for R− , we
can prove that it is indeed a lowering operator. Accordingly, the only non-vanishing
matrix elements are the ones with m′′ = m + 1. We write the non-vanishing matrix
elements as
hm + 1|R+ |mi = λm , hm|R− |m + 1i = λ∗m . (2.3.26)
The third commutator of Eqs.(2.3.22) gives
2M
hm|R+ |m′ ihm′ |R− |mi − hm|R− |m′ ihm′ |R+ |mi = . (2.3.27)
ω
Substituting the matrix elements from Eqs.(2.3.26) into the above equation gives
2M
|λm−1 |2 − |λm |2 = . (2.3.28)
ω
The solution of the above difference equation is
2M
|λm |2 = −m + C. (2.3.29)
ω
Since |λm |2 ≥ 0 we obtain C = 2nM/ω, where n is the upper value for m ≥ 0. Unlike
the case of the SU(2) symmetry, the solution of the difference equation Eq.(2.3.28)
gives no condition on the value of m whether it is an integer, half-integer, or not.
On the other hand, if we assume that there is a state in which m = 0, then definitely
all the values of m ∈ Z occur, since the state with m = 0 can be raised or lowered
by unit steps using R+ or R− . It is obvious that we have an infinite degeneracy with
m ∈ [−∞, n], where n is a positive integer.

In order to calculate the energy levels we write the Hamiltonian as

 
ω 1
H= 2L + ω (R+ R− + R− R+ ) . (2.3.30)
2 2
Since H and L commute they can have the same eigenfunctions. Accordingly the
matrix representing the above operator is diagonal in the m representation and from
the above Eq.(2.3.30) and Eq.(2.3.26) we get
 
1 2 1 2 2
Em = hm|H|mi = 2Mωm + ω M (|λm−1 | + |λm | )
2M 2
 
1
= ω n+ . (2.3.31)
2
2.3. ENERGY LEVELS AND DEGENERACY FROM JUST SYMMETRY 23

It is worth noting that there is another solution of Eq.(2.3.28) with negative n

and negative m ≤ n. However this solution must be discarded because it leads to
negative energy according to the above equation. Since m ∈ [−∞, n] we can write

m + |m|
n= + k, k = 0, 1, 2, ... (2.3.32)
2
Here the degeneracy is infinite but countable. This is because we are working in the L
representation. Working in a different representation like the px or py representation
also leads to infinite degeneracy, however, in this case it is continuous rather than
countable. This will be discussed later in chapter 3.

To show that the previous results are correct, we will now obtain the energy
levels and degeneracies by solving the Schrödinger equation for this problem. For
technical reasons the calculations in the following will be carried out using the
symmetric gauge. We must stress here that the choice of gauge does not change the
physics. Nevertheless it will simplify the solution of the Schrödinger equation. The
Hamiltonian in symmetric gauge A ~ = (−By/2, Bx/2, 0) takes the following form.
 2  2 !
1 eBx eBy
H=− ∂x2 + ∂y2 + ieBx∂y − ieBy∂x − − . (2.3.33)
2m 2 2

In this gauge the angular momentum takes the form

L = xpy − ypx . (2.3.34)

After writing the Hamiltonian in polar coordinates, it can be proved that a solu-
tion can be chosen which is simultaneously an eigenfunction of the Hamiltonian in
Eq.(2.3.33) as well as of the angular momentum operator of Eq.(2.3.34). It has the
following general form
ψ(r, ϕ) = exp(imϕ)ψ(r). (2.3.35)
Accordingly, the Schrödinger equation for this problem can be written as

(2m + Mωr 2 )2 ψ(r) − 16r 2 ∂r2 ψ(r) − 16r 4 ∂r22 ψ(r) = 8MEr 2 ψ(r). (2.3.36)

By making the substitutions

ME 2 2ω ρµ
ρ= r = α2 r 2 , µ= , ψ(r) = exp(− )F (ρ), (2.3.37)
2 E 4
Eq.(2.3.36) takes the form
 2 
m µ(m + 1) µ
+( − 1)ρ F (ρ) + (ρ − 1)ρ∂ρ F (ρ) − ρ2 ∂ρ2 F (ρ) = 0. (2.3.38)
4 4 2
24 CHAPTER 2. AN INTRODUCTION TO SYMMETRY IN PHYSICS

The solution can be found using Frobenius’ method, by assuming that it is a series
of the form ∞
X
F (ρ) = an ρs+n . (2.3.39)
n=0

Substituting the above solution in Eq.(2.3.38), and after that expressing the value of
the coefficients an and s, we find that s = ±|m|/2. Here the an ’s are the coefficients
of the confluent hypergeometric function. Therefore we can write
 2   
2 2 |m|/2 r µα m |m| 1 2 2 2
ψ(r) = A(r α ) exp − 1 F1 + + − , |m| + 1, α r ,
4 2 2 2 µ
(2.3.40)
where the solution with s = −|m|/2 has been discarded because it diverges at the
origin. The confluent hypergeometric function diverges as r → ∞ unless the series
terminates (see chapter 3). This can happen only if the first argument of 1 F1 (a, b, x)
is a negative integer. Accordingly we get

m |m| 1 2
−k = + + − . (2.3.41)
2 2 2 µ

However, in Eqs.(2.3.37) we defined µ = 2ω/E. Therefore the above equation leads

to  
m + |m| 1
E=ω + +k , k = 0, 1, 2, 3, ... (2.3.42)
2 2
This is exactly the same result that was reached in the beginning of this section by
investigating just the symmetry.
Chapter 3

Runge-Lenz Vector and Accidental

SU (2) Symmetry for Motion on a
Cone

3.1 Introduction to Motion on a Cone

Among countless classical systems the 1/r and harmonic oscillator potentials are
exceptional because in addition to rotation invariance they have accidental dynam-
ical symmetries as was explained in chapters 1 and 2. There are other conserved
quantities in addition to the angular momentum vector that arises from spherical
symmetry. In the case of the Kepler potential the additional quantity is the three-
component Runge-Lenz vector, and in the case of the isotropic harmonic oscillator
it is a rank two tensor.

At the classical level the accidental symmetries imply that all bound orbits are
closed, while at the quantum level they lead to additional degeneracies in the discrete
energy spectrum. In general we can say that the SO(d) rotational symmetry of
the d-dimensional 1/r potential (the Coulomb potential for d = 3) is enlarged to
the accidental symmetry SO(d + 1). The additional conserved quantities form the
d-components of the Runge-Lenz vector. Similarly, the d-dimensional harmonic
oscillator has an SO(d) rotational symmetry which is contained as a subgroup in an
accidental SU(d) symmetry.

The uniqueness of the 1/r and r 2 potentials was realized by Bertrand in 1873. He
proved that they are the only spherically symmetric scalar potentials in Euclidean
space for which all bound orbits are closed [2]. On the other hand, there exist

25
26 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

a number of other systems with accidental symmetries involving vector potentials,

direction-dependent potentials or non-Euclidean spaces. For example, a free particle
confined to the surface of the d-dimensional hyper-sphere S d moves along a great
circle (which obviously is closed). Indeed the rotational SO(d + 1) symmetry of
this system corresponds to the accidental symmetry of the 1/r potential. At the
quantum mechanical level it was first realized by Fock in 1935 that the hydrogen
atom possesses “hyper-spherical” symmetry [15]. Based on this work, Bargmann
[16] has shown that the generators of the accidental symmetry are the components
of the Runge-Lenz vector [17]
 
R~ = 1 ~p × L ~ −L~ × ~p − κ~er . (3.1.1)
2M
Accidental symmetry may also involve vector potentials, for example, the symmetry
involving the vector potential in cyclotron motion [18, 19]. In all these cases, there
is a deep connection between the fact that all bound classical orbits are closed and
additional degeneracies in the discrete energy spectrum of the corresponding quan-
tum system. The subject of accidental symmetry has been reviewed, for example,
by McIntosh [20].

In order to further investigate the phenomenon of accidental symmetries in this

chapter, we study a particle confined to the surface of a cone. A cone is obtained
from the plane by removing a wedge of deficit angle δ and gluing the open ends
back together. As a consequence, the polar angle χ no longer extends from 0 to
2π, but only to 2π − δ. The geometry of the cone is illustrated in figure (3.1). The
surface of a cone can be mapped on a 2-dimensional plane with points defined by
polar coordinates r and the angle ϕ such that it again covers the full interval, i.e.
χ
ϕ = ∈ [0, 2π], (3.1.2)
s
with the scale factor
δ
s=1− . (3.1.3)
2π
The kinetic energy of a particle of mass M then takes the form
M 2 M 2
T = (ṙ + r 2 χ̇2 ) = (ṙ + r 2 s2 ϕ̇2 ). (3.1.4)
2 2
The radial component of the momentum pr is canonically conjugate to r, i.e.
∂T
pr = = M ṙ. (3.1.5)
∂ ṙ
Similarly, the canonically conjugate momentum corresponding to the rescaled angle
ϕ is given by the angular momentum
∂T
L= = Mr 2 s2 ϕ̇, (3.1.6)
∂ ϕ̇
3.2. THE DOMAINS OF OPERATORS AND HERMITICITY 27

r
Χ

Figure 3.1: We can think of a cone as a result of cutting a wedge of deficit angle δ
out of the 2-dimensional plane, and by gluing the open ends back together. A point
on the cone is defined by the distance r from the tip and an angle χ which varies
between 0 and 2π − δ. Unlike the cone in the figure (3.1), the actual cone considered
in this work extends over the whole range r ∈ [0, ∞].

such that  
1 2 L2
T = pr + 2 2 . (3.1.7)
2M r s
From Eq.(3.1.7) we can see that a positive deficit angle δ (i.e. s < 1) leads to an
increase of the centrifugal barrier, while a negative deficit angle (s > 1) leads to its
reduction. As usual, upon canonical quantization (and again using natural units in
which ~ = 1) the angular momentum conjugate to the rescaled angle ϕ is represented
by the operator
L = −i∂ϕ . (3.1.8)

3.2 The Domains of Operators and Hermiticity

The domains of the operators in a corresponding Hilbert space, Hermiticity and self-
adjointness play an important role in this work. For some mathematical background
we refer to [21, 22].The Hilbert space can be denoted as H = L2 ((0, ∞) × (0, 2π); r).
For a particle moving on a cone it consists of the square-integrable functions Ψ(r, ϕ)
with r ∈ (0, ∞), ϕ ∈ [0, 2π], and with the norm hΨ|Ψi < ∞. The norm here can be
evaluated using the scalar product which is defined by
Z ∞ Z 2π
hΦ|Ψi = dr r dϕ Φ(r, ϕ)∗ Ψ(r, ϕ). (3.2.1)
0 0
28 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

As it is obvious from the definition of the Hilbert space, the functions in that space
need not be continuous nor differentiable, they even need not be periodic. In the
following argument we will use the angular momentum operator as an example. In
order to completely define a quantum mechanical operator O, the domain of wave
functions on which the operator acts must be defined, say D[O] ⊂ H. This is because
L acts on a wave function that must be differentiable at least once with respect to
ϕ.

An operator O is defined to be Hermitean or called symmetric in other mathe-

matical terms if
hOΦ|Ψi = hΦ|OΨi, (3.2.2)
for all wave functions Φ, Ψ ∈ D[O]. To examine the Hermiticity of L, let us perform
a partial integration that gives
Z ∞ Z 2π
hOΦ|Ψi = dr r dϕ [−i∂ϕ Φ(r, ϕ)]∗ Ψ(r, ϕ)
0 0
Z ∞ Z 2π
= dr r dϕ Φ(r, ϕ)∗ [−i∂ϕ Ψ(r, ϕ)]
Z0 ∞
0

+ i dr rΦ(r, ϕ)∗ Ψ(r, ϕ)|ϕ=2π

ϕ=0
0
Z ∞
= hΦ|OΨi + i dr rΦ(r, ϕ)∗ Ψ(r, ϕ)|ϕ=2π
ϕ=0 . (3.2.3)
0

Thus, the operator L is Hermitean if

Φ(r, ϕ)∗ Ψ(r, ϕ)|ϕ=2π

ϕ=0 = 0. (3.2.4)

Understanding self-adjointness and Hermiticity is very important here because

Hermiticity alone does not guarantee a real-valued spectrum. On the other hand,
self-adjointness means that the operator indeed has a real-valued spectrum. An
operator O is self-adjoint (i.e. O = O † ) if it is Hermitean and the domain of its
adjoint O † coincides with the domain of O, i.e. D[O †] = D[O], for all Ψ ∈ D[O].
For example, let us consider the operator L in the domain of differentiable functions
Ψ ∈ H (with LΨ ∈ H). There are two possibilities for the operator L to be
Hermitean. It is either that Ψ(r, 0) = Ψ(r, 2π) = 0 then D[L] is smaller than D[L† ]
because Φ is unrestricted on the boundary, or Φ(r, 0) = Φ(r, 2π) = 0 then D[L] is
bigger than D[L† ]. For neither case the operator L is self-adjoint restricted to such
domains. Self-adjointness of L can be reached by choosing a different domain for L.
Let us impose the boundary condition

Ψ(r, 2π) = z ∗ Ψ(r, 0), z ∈ C. (3.2.5)

3.2. THE DOMAINS OF OPERATORS AND HERMITICITY 29

Then the condition in Eq.(3.2.4) implies that

Φ(r, 2π)∗ Ψ(r, 2π) − Φ(r, 0)∗ Ψ(r, 0) = [Φ(r, 2π)∗ z − Φ(r, 0)∗ ]Ψ(r, 0) = 0, (3.2.6)

such that
1
Φ(r, 2π) = Φ(r, 0). (3.2.7)
z∗
For L to be self-adjoint (i.e. to have D[L† ] = D[L]) the functions Φ ∈ D[L† ] must
obey the same condition as Ψ ∈ D[L]. This implies z = 1/z ∗ = exp(iθ). The angle
θ characterizes a one-parameter family of self-adjoint extensions of the operator L
to the domain of differentiable functions obeying the boundary condition

Ψ(r, 2π) = exp(iθ)Ψ(r, 0). (3.2.8)

Since the coordinates ϕ = 0 and ϕ = 2π describe the same physical point on the
cone, this requires a single-valued wave function. Therefore, for wave functions on
the cone, the domain D[L] ∈ H consists of the periodic differentiable functions Ψ
(with LΨ ∈ H) which obey
Ψ(r, 2π) = Ψ(r, 0). (3.2.9)

3.2.1 The Hermiticity and Self-Adjointness of T

Let us consider the operator of the kinetic energy,

 
1 1 1 2
T =− ∂r2 + ∂r + 2 2 ∂ϕ . (3.2.10)
2M r r s

Since ∂ϕ = s∂χ , this operator seems to be identically the same as the standard one
operating on wave functions on the plane. For a complete definition of T , its domain
D[T ] must be identified. The wave functions of D[T ] should again obey Eq.(3.2.9),
which means that they should be periodic in the rescaled angle ϕ (not in the original
polar angle χ of the full plane). Writing

Ψ(r, ϕ) = ψ(r) exp(imϕ), (3.2.11)

Eq.(3.2.9) leads to m ∈ Z , and we have

 
1 2 1 m2
T =− ∂r + ∂r + . (3.2.12)
2M r 2Mr 2 s2
30 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

The radial wave functions belong to the radial Hilbert space Hr = L2 ((0, ∞)); r). It
is well-known that the operator −i∂r is not Hermitean in Hr . This follows from
Z ∞
hφ|∂r ψi = dr rφ(r)∗∂r ψ(r)
0
Z ∞
= − dr ∂r [rφ(r)∗] ψ(r) + rφ(r)∗ψ(r)|∞
0
Z0 ∞
= − dr [r∂r φ(r)∗ + φ(r)∗] ψ(r) + rφ(r)∗ψ(r)|∞
0
0
Z ∞  
1
= dr r −∂r φ(r)∗ − φ(r)∗ ψ(r) + rφ(r)∗ ψ(r)|∞ 0
0 r
= h∂r† φ|ψi + rφ(r)∗ψ(r)|∞ 0 . (3.2.13)

Therefore the Hermitean conjugate of ∂r is

1
∂r† = −∂r − . (3.2.14)
r

The above equation is correct if the boundary term vanishes (i.e rφ(r)∗ψ(r)|∞
0 = 0).
On the other hand, it can be proved that the operator Dr given below is Hermitean
in the domain D[Dr ] of differentiable functions ψ(r) (with Dr ψ ∈ Hr ) obeying
ψ(0) = 0, with
 
1 1 √
Dr = −i ∂r + = −i √ ∂r r. (3.2.15)
2r r

However, that does not mean that it represents a physical observable because it is
not self-adjoint. On the other hand,

 2
1 1 1
Dr2 = − ∂r + = −∂r2 − ∂r + 2 , (3.2.16)
2r r 4r

which is closely related to the kinetic energy operator T , possesses a family of self-
adjoint extensions. Eq.(3.2.14) also seems to readily imply Hermiticity of the kinetic
energy operator T because, at least formally,

 †  2  
2 1 †2 †1 1 1 1
∂r + ∂r = ∂r + ∂r = ∂r + − ∂r +
r r r r r
2 1 1
= ∂r2 + ∂r − ∂r = ∂r2 + ∂r . (3.2.17)
r r r
3.2. THE DOMAINS OF OPERATORS AND HERMITICITY 31

In more details
  Z ∞  
2 1 ∗ 2 1
hφ| ∂r + ∂r ψi = dr rφ(r) ∂r + ∂r ψ(r)
r 0 r
Z ∞
= − dr ∂r [rφ(r)∗ ] ∂r ψ(r) + rφ(r)∗ ∂r ψ(r)|∞
0
Z ∞0
− dr ∂r φ(r)∗ ψ(r) + φ(r)∗ ψ(r)|∞
0
Z0 ∞
= dr ∂r2 [rφ(r)∗] ψ(r) − ∂r [rφ(r)∗] ψ(r)|∞0
Z0 ∞
− dr ∂r φ(r)∗ ψ(r) + [rφ(r)∗∂r ψ(r) + φ(r)∗ ψ(r)]∞
0
Z0 ∞   
2 1
= dr r ∂r2 + ∂r − ∂r φ(r)∗ ψ(r)
0 r r
+ [rφ(r) ∂r ψ(r) − r∂r φ(r) ψ(r)]∞
∗ ∗
0
 
1
= h ∂r2 + ∂r φ|ψi + [rφ(r)∗ ∂r ψ(r) − r∂r φ(r)∗ψ(r)]∞ 0 .
r
(3.2.18)

However, this imposes a delicate condition so that T is Hermitean. The condition is

[rφ(r)∗∂r ψ(r) − r∂r φ(r)∗ ψ(r)]∞

0 = 0. (3.2.19)

The condition implies that the component of the probability current density jr along
~r must vanish at ∞. This become clear when we replace φ(r) by ψ(r). The condition
admits a one-parameter family of self-adjoint extensions. The self-adjoint extensions
of T have been studied in [23]. It turns out that the tip of the cone is a singular
point that may be endowed with non-trivial physical properties. These properties
are described by a real-valued parameter that defines a family of self-adjoint ex-
tensions. Physically speaking, the different self-adjoint extensions correspond to
properly renormalized δ-function potentials of different strengths located at the tip
of the cone. In this paper, we limit ourselves to the case without δ-function poten-
tials, which corresponds to the so-called Friedrichs extension [21] characterized by
the boundary condition
lim r∂r ψ(r) = 0. (3.2.20)
r→0

If we impose this condition on ψ ∈ D[T ], and also want to satisfy Eq.(3.2.19), the
function φ ∈ D[T † ] must also obey Eq.(3.2.20). As a result, D[T † ] = D[T ], such
that T = T † is indeed self-adjoint.

While the cone is as flat as the plane, its singular tip and its deficit angle δ
have crucial effects on the dynamics. In the following, we will consider a particle
32 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

moving on a cone and bound to its tip by a 1/r or r 2 potential. Interestingly, when
the deficit angle divided by 2π (or equivalently s) is a rational number, all bound
classical orbits are again closed and once more there are additional degeneracies in
the discrete spectrum of the Hamilton operator H. Just like in the plane, the 1/r
and r 2 potentials on a cone have accidental SU(2) symmetries. However, unlike in
the plane, the corresponding multiplets may now have fractional “spin” and unusual
degeneracies. This unusual behavior arises because, in this case, the Runge-Lenz
vector R~ — although Hermitean in its appropriate domain D[R] ~ — does not act as
a Hermitean operator in the domain D[H] of the Hamiltonian and thus does not
represent a proper physical observable.

The problem of space-times with a conical singularity was investigated by ’t Hooft

[24] and by Deser and Jackiw [25] concerning (2 + 1)-dimensional Einstein gravity.
In this context, the Klein-Gordon and the Dirac equation have also been studied
[26]. Conical space-times also arise in the study of cosmic strings and are related
to the Aharonov-Bohm effect [27]. Indeed, 1/r and r 2 potentials have already been
considered in this context [8, 28], however, without discussing accidental symmetries.
Furthermore, graphene — a single sheet of graphite, i.e. a honeycomb of carbon
hexagons — can be bent to form cones by adding or removing a wedge of carbon
atoms and by replacing one hexagon by a carbon hepta- or pentagon [29, 30]. While
the low-energy degrees of freedom in graphene are massless Dirac fermions, in this
thesis we limit ourselves to studying the Schrödinger equation for motion on a cone.

3.3 The 1/r Potential on a Cone

In this section we will study a particle moving on the surface of a cone bound to its
tip by a 1/r potential
κ
V (r) = − . (3.3.1)
r
The corresponding total energy is thus given by
 
1 2 L2 κ
H =T +V = pr + 2 2 − . (3.3.2)
2M r s r

3.3.1 Classical Solutions

The classical orbit can be found by solving Newton’s equation of motion for this
problem. The Newton orbit equation is
d2 u f (u−1)
+ u = − , (3.3.3)
d2 θ h2 u2
3.3. THE 1/R POTENTIAL ON A CONE 33

where u = 1/r, h = r 2 θ̇, f (u−1 ) is component of the central force in the er -direction.
The above equation is valid only for central potentials, i.e, f (~r) = f (r). The solution
of this equation for the 1/r potential is

1 Mκs2
= [1 + e cos(s(ϕ − ϕ0 ))] , (3.3.4)
r L2

with the eccentricity given by

r
2EL2
e= 1+ , (3.3.5)
Mκ2 s2

where E < 0 is the energy and L is the angular momentum. The radial component
of the momentum takes the form

Mκs
pr = e sin(s(ϕ − ϕ0 )). (3.3.6)
L

Here ϕ0 is the angle between the positive x- direction and a vector pointing towards
the perihelion. From Eq.(3.3.4) we see that the classical orbit is closed only when
s = p/q is a rational number (with p, q ∈ N not sharing a common divisor). For
s = p/q, after q revolutions around the tip of the cone, both r and pr return to their
initial values. Some examples of classical orbits are shown in figure (3. 2). Without
this condition the orbit never closes which means that both r and pr never return
to their initial value after any length of time.

Figure 3.2: Examples of bound classical orbits for the 1/r potential with s = 3 (left),
s = 12 (middle), and s = 1 (right). The latter case represents a standard Kepler
ellipse. The orbits are shown in the x-y-plane with (x, y) = r(cos ϕ, sin ϕ) where
ϕ = χ/s ∈ [0, 2π] is the rescaled polar angle.
34 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

3.3.2 Semi-classical Bohr-Sommerfeld Quantization

Let us consider Bohr-Sommerfeld quantization. The quantization condition for the

angular momentum takes the form
I
dϕ L = 2πL = 2πm, (3.3.7)

such that L = m ∈ Z. Similarly, the quantization condition for the radial motion is
given by I  
1
dr pr = 2π nr + , nr ∈ {0, 1, 2, ...}. (3.3.8)
2
The factor 1/2, which is sometimes not taken into account in Bohr-Sommerfeld
quantization, arises for librations but is absent for rotations. Using Eqs.(3.3.4),
(3.3.5), and (3.3.6) and integrating over the period 2π/s it is straightforward to
obtain
I Z 2π/s r !
|L|e2 sin2 (s(ϕ − ϕ0 )) Mκ2 |L|
dr pr = dϕ = 2π − − , (3.3.9)
0 (1 + e cos(s(ϕ − ϕ0 ))2 2E s

which leads to
Mκ2
E=−  2 . (3.3.10)
2 nr + |m|
s
+ 1
2

The above result is exact and not just limited to the semi-classical regime.

3.3.3 Solution of the Schrödinger Equation

The radial Schrödinger equation takes the form

   
1 2 1 m2 κ
− ∂r + ∂r + − ψ(r) = Eψ(r), (3.3.11)
2M r 2Mr 2 s2 r

which can be solved using Frobenius’ method [8].

αr 2|m|
ψnr ,m (r) = A exp(− )(αr)|m|/s 1 F1 (−nr , + 1, αr), (3.3.12)
2 s
with p
α= 8M|E|, (3.3.13)
where 1 F1 is a confluent hyper-geometric function. The corresponding quantized
energy values are given by Eq.(3.3.10). Now nr is the number of nodes of the radial
3.3. THE 1/R POTENTIAL ON A CONE 35

wave function and m ∈ Z is the angular momentum quantum number. Parity

symmetry together with the SO(2) rotational symmetry ensures the degeneracy of
states with quantum numbers m and −m. The relation between a closed bound
orbit and additional accidental degeneracies in the spectrum can be realized here
from Eq.(3.3.4). When the scale factor s is a rational number the orbit is closed.
Equivalently, Eq.(3.3.10) gives additional degeneracies.

It is important to note that Eq.(3.3.11) has other solutions that diverge at the
origin. These solutions will play a pivotal role in counting the degeneracies of differ-
ent states. For example, let us consider the case s = 21 corresponding to the deficit
angle δ = π. In that case, a wave function without nodes (i.e. with nr = 0) and
with angular momentum m = ±1 is degenerate in energy with a wave function with
two nodes (nr = 2) and with m = 0. As another example, let us consider s = 2
which corresponds to the negative deficit angle δ = −2π. In this case, one builds a
“cone” by cutting two planes open and gluing them together in the same way as the
double-layered Riemann surface of the complex square root. This effectively lowers
the centrifugal barrier by a factor of s2 = 4. In this case, a wave function without
nodes (nr = 0) and with m = ±2 is degenerate with a wave function with one node
(nr = 1) and with m = 0. Similarly, for s = n ∈ N, one glues n cut planes to a
“cone” in the same way as the multi-layered Riemann surface of the complex n-th
root. Now, a wave function without nodes (nr = 0) and with m = ±n is degenerate
with a wave function with one node (nr = 1) and with m = 0. Some features of the
energy spectrum are illustrated in figure (3.3).

3.3.4 Runge-Lenz Vector and SU (2) Algebra

When we realize that all classical orbits are closed, we expect that there must be a
hidden conserved quantity. Quantum mechanically this leads to a Hamilton operator
with a discrete spectrum and accidental degeneracies, and this suggests that there
must be a corresponding accidental symmetry. From the Kepler problem in 2-d
(with s = 1) we know that the corresponding conserved quantity is the Runge-Lenz
vector and the accidental symmetry is an SU(2) symmetry. The Runge-Lenz vector
can be constructed for any s = p/q. However, it should be pointed out that, due to
the conical geometry, the resulting object no longer transforms as a proper vector.
We still continue to refer to it as the “Runge-Lenz vector”.

At the classical level, we can use eqs.(3.3.4) and (3.3.6) to write

L2
κe cos(s(ϕ − ϕ0 )) = κe [cos(sϕ) cos(sϕ0 ) + sin(sϕ) sin(sϕ0 )] = − κ,
Mrs2
pr L
κe sin(s(ϕ − ϕ0 )) = κe [sin(sϕ) cos(sϕ0 ) − cos(sϕ) sin(sϕ0 )] = , (3.3.14)
Ms
36 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

Αr
3
2
2

Figure 3.3: The 1/r potential (solid curve) together with an effective potential in-
cluding the centrifugal barrier with m = ±1 (dashed curve) for s = 3. The energies
of the ground state and the first three excited states are indicated by horizontal lines.
The numbers besides the lines specify the degree of degeneracy. The ground state
(with nr = 0, m = 0) is non-degenerate, while the first and second excited states
(with nr = 0, m = ±1 and nr = 0, m = ±2, respectively) are two-fold degenerate due
to parity symmetry. The third excited level has an accidental three-fold degeneracy
and consists of the states with nr = 0, m = ±3 and nr = 1, m = 0.

such that
 
L2 pr L
Rx = κe cos(sϕ0 ) = − κ cos(sϕ) + sin(sϕ),
Mrs2 Ms
 2 
L pr L
Ry = κe sin(sϕ0 ) = 2
− κ sin(sϕ) − cos(sϕ), (3.3.15)
Mrs Ms
are indeed independent of time or constants of motion. It is easy to see that for
s = 1, Rx and Ry are just the components of the familiar Runge-Lenz vector. There
are two ways to verify that these two components are constants of motion, either by
proving that they have a vanishing Poisson bracket with the Hamilton function or
by direct substitution of Eq.(3.3.4) and Eq.(3.3.6) in the above expression of Rx and
Ry . If the result is independent of ϕ then they are constants of motion. It should be
noted that, for non-integer values of s, the quantities Rx and Ry are not conserved
quantities in the usual sense. In particular, they are not single-valued functions of
the coordinates x = r cos ϕ and y = r sin ϕ, but depend on the angle ϕ itself. As a
consequence, the values of Rx and Ry depend on the history of the motion, i.e. on
the number of revolutions around the tip of the cone. However, quantities that are
3.4. RUNGE-LENZ VECTOR AT THE QUANTUM LEVEL 37

“conserved” only because they refer back to the initial conditions, do not qualify
as proper physical constants of motion. To further clarify this issue, it is useful to
construct the complex variables
 2 
L pr L
R± = Rx ± iRy = −κ∓i exp(±isϕ). (3.3.16)
Mrs2 Ms

For rational values s = p/q (with p, q ∈ N) the quantities

 2 q
q L pr L
R± = −κ∓i exp(±ipϕ) (3.3.17)
Mrs2 Ms

are single-valued functions of x = r cos ϕ and y = r sin ϕ, and hence qualify as

proper conserved quantities.

The length of the Runge-Lenz vector is given by

 2 2  2
2 2 2 L pr L
R = Rx + Ry = −κ +
Mrs2 Ms
2
 2 2

L pr L κ 2 2HL2
= 2 + − + κ = + κ2 . (3.3.18)
Ms2 2M 2Mr 2 s2 r Ms2

In the quantum mechanical treatment it will turn out to be useful to introduce the
rescaled variables
r r
ex = − M ey = − M Ry , L e = L,
R Rx , R (3.3.19)
2H 2H s
which makes sense for bound orbits with negative energy. We then obtain
2
e2 + R e2 = − Mκ
e2 + L Mκ2
C=Rx y ⇒ H=− . (3.3.20)
2H 2C
It will turn out that the quantum analogue of C is the Casimir operator of an
accidental SU(2) symmetry.

3.4 Runge-Lenz Vector at the Quantum Level

To treat the problem quantum mechanically, the components of the Runge-Lenz

vector must be turned into operators. The procedure that is presented here is
useful not just for this particular problem, but for any problem involving finding a
conserved quantity generating a symmetry group of a quantum mechanical system.
The steps of the procedure are as follows:
38 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

1) An ansatz must be written down based on the classical form of the operator with
unknown coefficient functions of coordinates in front of the derivatives that compose
the operator. For example, in our case, there are functions of ϕ such that
1 1
R = A1 (ϕ)∂ϕ2 + A2 (ϕ)∂ϕ + A3 (ϕ) + A4 (ϕ)∂ϕ ∂r + A5 (ϕ)∂r . (3.4.1)
r r
2) Commuting the proposed operator with the Hamiltonian, we get coefficients of
ψ(r, ϕ), ∂r ψ(r, ϕ),∂ϕ ψ(r, ϕ),∂ϕ ψ(r, ϕ), ∂ϕ2 ψ(r, ϕ) etc. If the proposed operator shall
commute with the Hamiltonian, these coefficients of ψ(r, ϕ) and its derivatives must
equal zero. This gives a set of differential equations. In our case, these equations
are,
2kms2 A5 (ϕ) + ∂ϕ2 A3 (ϕ) = 0,
s2 A2 (ϕ) + ∂ϕ2 A2 (ϕ) = 0,
kms2 A4 (ϕ) + ∂ϕ A3 (ϕ) = 0,
s2 A5 (ϕ) + ∂ϕ2 A5 (ϕ) = 0, (3.4.2)
2 2
−2s A2 (ϕ) + s A4 (ϕ) + 2∂ϕ A5 (ϕ) + ∂ϕ2 A4 (ϕ) = 0,
2
−s A1 (ϕ) + ∂ϕ A4 (ϕ) = 0,
s2 A1 (ϕ) + 2A5 (ϕ) + 2∂ϕ A2 (ϕ) + ∂ϕ2 A1 (ϕ) = 0.

3) The differential equations are solved using the appropriate boundary conditions.
In our case the solution must be sϕ -periodic. Solving Eqs.(3.4.2) gives two solutions
for which all A’s are non-zero functions of ϕ. These two solutions represent the
two components of the Runge-Lenz operator. In addition to that, we have a third
solution with all the A’s equal to zero, except A2 equal to a constant. This solution
represents the one-component angular momentum operator. We then obtain
1 1
Rx = − 2
cos(sϕ)∂ϕ2 + sin(sϕ)∂ϕ − κ cos(sϕ)
Mrs 2Mrs
1 1
− sin(sϕ)∂r ∂ϕ − cos(sϕ)∂r ,
Ms 2M
1 1
Ry = − 2
sin(sϕ)∂ϕ2 − cos(sϕ)∂ϕ − κ sin(sϕ)
Mrs 2Mrs
1 1
+ cos(sϕ)∂r ∂ϕ − sin(sϕ)∂r . (3.4.3)
Ms 2M
The result can be checked by commuting the components of the Runge-Lenz vector
with the Hamiltonian, and indeed the result is zero, i.e.
[Rx , H] = [Ry , H] = 0. (3.4.4)
It can be proved that Rx , Ry , and L obey the following algebra
2HL
[Rx , Ry ] = −i , [Rx , L] = −isRy , [Ry , L] = isRx . (3.4.5)
Ms
3.4. RUNGE-LENZ VECTOR AT THE QUANTUM LEVEL 39

Applying the rescaling of Eq.(3.3.19), this leads to

ex , R
[R ey ] = iL,
e [R
ey , L]
e = iR
ex , [L,
e Rex ] = iR
ey . (3.4.6)

Hence, Rex , R
ey , and L
e generate an SU(2) algebra. The raising and lowering operators
are defined by
e± = R
R ex ± iR
ey . (3.4.7)

3.4.1 Casimir Operator

The Casimir operator for the SU(2) algebra is

e2 + R
e2 + L
e2 = − Mκ2 1
C=Rx y − , (3.4.8)
2H 4
such that
Mκ2 Mκ2
H=−
= −
2 . (3.4.9)
2 C + 14 2 S + 21
As it was explained in chapter 2, we have one Casimir operator for SU(2) with
spectrum S(S + 1) where S is either an integer or a half-integer as we explained in
chapter 2. Accordingly we have
 2
1 1 1
C + = S(S + 1) + = S+ . (3.4.10)
4 4 2

By comparison with Eq.(3.3.10) for the energy spectrum, we thus identify

|m|
S = nr + . (3.4.11)
s
This result is unexpected because for 2|m|/s ∈ / N the abstract spin S is not an
integer or a half-integer. For a general scale factor s corresponding to a general
deficit angle δ, the abstract spin is continuous. Even for general rational s, for
which all bound classical orbits are closed and there are accidental degeneracies in
the discrete spectrum of the Hamiltonian, the spin S is not always an integer or
a half-integer. The treatment that was considered in chapter 2 assumed that the
components of the Runge-Lenz vector form raising and lowering operators that will
never take the wave function outside the Hilbert space. This leads to integer or half-
integer S. Here we are facing a different situation because the raising and lowering
operators can take one outside the Hilbert space and the treatment in section (2.3)is
no longer valid.
40 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

3.4.2 The Raising and Lowering Operator Acting on the

Wave Function

In order to understand the degeneracy of the energy levels and other aspects like
domains of operators and Hermiticity, it is important to understand the effect of
the raising and lowering operator on the wave function. For our problem this is
possible by direct application of the raising and lowering operators on the wave
function. The explicit form of the lowering operators can be found by substituting
the expressions for Rx and Ry from Eq.(3.4.3) into Eq.(3.4.7). One then applies
the lowering operator on the wave function of Eq.(3.3.12). The result depends on
whether |m|/s is an integer or half-integer or not. We will study both cases and
prove that, upon repeated application of R− the multiplet terminates when |m|/s is
an integer or a half-integer, while it does not terminate otherwise. One application
of R− on the wave function with m > 0 gives
r
−α2 |m| αr
R− ψnr ,m (r, ϕ) = −A (αr) s −1 exp(− ) exp(iϕ(m − s))
8HM 2
2|m|
( s − 1) 2|m|
× αrnr 2|m| 1 F1 (1 − nr , + 2, αr)
( s + 1) s
  !
2|m| 2|m| 2|m| 2|m|
+ αr( + nr ) − ( − 1) 1 F1 (−nr , + 1, αr) .
s s s s
(3.4.12)
The confluent hypergeometric function is defined by
k=∞
X (a)k xk
a a(a + 1) x2 a...(a + k) xk+1
1 F1 (a, b, x) = 1+ x+ +...+ = . (3.4.13)
b b(b + 1) 2! b...(b + k) (k + 1)! k=0
(b)k k!
The above series terminates only if a is zero or a negative integer. For such cases the
wave function in Eq.(3.3.12) goes to zero as r goes to infinity. Therefore the only
physically acceptable cases are a equal zero or a negative integer. Two identities,
proved using the definition in Eq.(3.4.13), will play an important role in our work
later,
lim b1 F1 (a, b, x) = ax1 F1 (a + 1, 2, x), (3.4.14)
b−→0
x2
lim (b + 1)1 F1 (a, b, x) = −a(a + 1)1 F1 (a + 2, 3, x). (3.4.15)
b−→−1 2
Using well-known recurrence relations [31], after lengthy calculations the following
identity can be proved
(2 − b)a
(2 − 3b + b2 + x(1 + a − b))1 F1 (a, b, x) − x 1 F1 (a + 1, b + 1, x)
b
= (1 − b)(2 − b)1 F1 (a − 1, b − 2, x). (3.4.16)
3.4. RUNGE-LENZ VECTOR AT THE QUANTUM LEVEL 41

Using the above identity, it is easy to see that Eq.(3.4.12) reduces to

r
−α2 |m| αr
R− ψnr ,m (r, ϕ) = −A (αr) s −1 exp(− ) exp(iϕ(m − s))
8HM 2
2|m| 2|m| 2|m|
× ( − 1) 1 F1 (−1 − nr , − 1, αr)
s s s
= Bψnr +1,m−s (r, ϕ), (3.4.17)

where B is a constant that results from one application of R− or R+ on wave function.

The explicit form or value of this constant is not important in our argument. We
must keep in mind that when B appears in different equations, it may not have the
same value, but this also does not affect our arguments. A general rule was found
for k applications of R− on the wave function, that is

Yk   
k j 2|m| 2|m|
R− ψnr ,m (r, ϕ) = A (−1) + 2 − 2j + 1 − 2j
j=1
s s
−α2 k |m| αr
× ( ) 2 (αr) s −k exp(− ) exp(iϕ(m − ks))
8HM 2
2|m|
× 1 F1 (−k − nr , + 1 − 2k, αr)
s
= Bψnr +k,m−sk (r, ϕ). (3.4.18)

Here we must stress that during successive applications of the operator R− we may
get non-integer m − sk depending on the values of s and k. Therefore this operator
may take the physical wave function to an unphysical one. For the case of m < 0,
the same procedure can be applied by using a well-known identity for the confluent
hypergeometric function [31]. The result of k applications of R− on the wave function
is

k
! !
Y 2|m|
k j nr + s
+j nr − j + 1
R− ψnr ,−|m| (r, ϕ) = A (−1) 2|m| 2|m|
j=1 s
+ 2j − 1 s
+ 2j
−α2 k |m| αr
× ( ) 2 (αr) s +k exp(− ) exp(iϕ(m − ks))
8HM 2
2|m|
× 1 F1 (k − nr , + 1 + 2k, αr)
s
= Bψnr −k,m−sk (r, ϕ). (3.4.19)
42 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

The case of k applications of R+ on the wave function with m ≥ 0 can be studied

similarly and the result is
k
! !
Y 2|m|
k nr + + j nr − j + 1
R+ ψnr ,m (r, ϕ) = A (−1)j 2|m| s 2|m|
j=1 s
+ 2j − 1 s
+ 2j
−α2 k |m| αr
× ( ) 2 (αr) s +k exp(− ) exp(iϕ(m + ks))
8HM 2
2|m|
× 1 F1 (k − nr , + 1 + 2k, αr)
s
= Bψnr −k,m+sk (r, ϕ). (3.4.20)

For the case of k application of R+ on the wave function with m < 0 we get
k
Y   
k j 2|m| 2|m|
R+ ψnr ,−|m| (r, ϕ) = A (−1) + 2 − 2j + 1 − 2j
j=1
s s
α2 k |m| αr
× (− ) 2 (αr) s −k exp(− ) exp(iϕ(m + ks))
8HM 2
2|m|
× 1 F1 (−k − nr , + 1 − 2k, αr)
s
= Bψnr +k,m−sk (r, ϕ). (3.4.21)

To count the number of wave functions in the multiplet we begin with nr = 0. Then,
Eq.(3.4.11) gives S = mmax /s with m = mmax . The value of m = mmax can not
be raised further because applying R+ gives zero according to Eq.(3.4.20). Now
applying the lowering operator R− a number of k0 times, with k0 = mmax /s + 1/2,
using Eq.(3.4.18) gives
k0
Y   
k0 j 2|mmax | 2|mmax |
R− ψ0,mmax (r, ϕ) = A (−1) + 2 − 2j + 1 − 2j
j=1
s s
−α2 k0 |mmax | αr
× ( ) 2 (αr)( s −k0 ) exp(− ) exp(iϕ(mmax − k0 s))
8HM 2
2|mmax |
× 1 F1 (−k0 , + 1 − 2k0 , αr). (3.4.22)
s
Here we have two cases in which the multiplet terminates.

1) Half-integer mmax /s

Applying the raising operator R+ once gives zero according to Eq.(3.4.20). On the
other hand, applying the lowering operator R− a number of k0 = mmax /s+1/2 times,
3.4. RUNGE-LENZ VECTOR AT THE QUANTUM LEVEL 43

and after that using the identity in Eq.(3.4.14), then Eq.(3.4.22) can be written as

k0 1 αr s |mmax | 1
R− ψ0,mmax (r, ϕ) = C(αr) 2 exp(− ) exp(−iϕ )1 F1 (− + , 2, αr)
2 2 s 2
= Bψ |mmax | − 1 ,−s/2 (r, ϕ), (3.4.23)
s 2

where C is just another constant that does not influence the argument. As we can
see, the resulting wave function has m = −s/2 < 0. The operator R− can be applied
′
further k0 times. However, we must use the appropriate relation Eq.(3.4.19) and we
obtain

k0
′ ! !
Y |mmax | |mmax |
k
′
k0 + 12 + j + 12 − j
R−0 (R− ψ0,mmax (r, ϕ)) = B (−1)j s s

j=1
2j 2j + 1
′
−α2 k0 1 ′ αr 1 ′
× ( ) 2 (αr) 2 +k0 exp(− ) exp(−iϕ( + k0 )s))
8HM 2 2
|mmax | 1 ′
× 1 F1 (k0 − + , 2k0 + 2, αr). (3.4.24)
s 2

It is obvious from the above equation that applying R− a number of |mmax |/s + 1/2
times gives zero. Therefore the number of wave functions in the multiplet is

′ |mmax |
N = k0 + k0 = 2 + 1. (3.4.25)
s

It is worth noting that for s = 1 one obtains N = 2|mmax | + 1 which is the number
of states in a multiplet of the Kepler problem in two dimensions.

2) Integer mmax /s

Applying the raising operator R+ gives zero according to Eq.(3.4.20). On the other
hand, applying the lowering operator R− a number of k0 = mmax /s + 1 times, and
after using the identity in Eq.(3.4.15), then Eq.(3.4.22) can be written as

k0 αr |mmax |
R− ψ0,mmax (r, ϕ) = C(αr)1 exp(− ) exp(−iϕs)1 F1 (− + 1, 3, αr)
2 s
= Bψ |mmax | −1,−s (r, ϕ). (3.4.26)
s

As we can see, the resulting wave function has m = −s < 0. The operator R− can
′
be applied further k0 times. However, we must use the appropriate relation for this
44 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

case, that is Eq.(3.4.19), and we obtain

′
k0
! !
Y |mmax | |mmax |
′
k0 k0 +1+j −j
R− (R− ψ0,mmax (r, ϕ)) = A (−1)j s s

j=1
2j + 1 2j + 2
′
−α2 k0 ′ αr ′
× ( ) 2 (αr)1+k0 exp(− ) exp(−iϕ(1 + k0 )s))
8HM 2
′ |mmax | ′
× 1 F1 (k0 − + 1, 2k0 + 3, αr). (3.4.27)
s
It is obvious from the above equation that applying a number of R− |mmax |/s times
gives zero. Therefore the number of wave functions in the multiplet is again

′ |mmax |
N = k0 + k0 = 2 + 1. (3.4.28)
s

Not integer or half-integer mmax /s

Applying the lowering operator R− any number of times will not make identity
(3.4.14) or identity (3.4.15) applicable, since neither (2|mmax |/s+1−2j) nor (2|mmax |/s+
2 − 2j) in Eq.(3.4.22) is equal to 0 or −1 for any value of j ∈ Z. After k0 > |mmax |/s
applications of R− , the power of (αr)(|mmax |/s)−k0 is negative, and the resulting wave
function diverges at the origin, although it is still a solution of the Schrödinger equa-
tion. The multiplet does not terminate for any number of applications of R− , and
the number of wave functions in such a multiplet is infinite. The multiple applica-
tion of R− transforms a well-behaved wave function into one that diverges at the
origin after k0 > |mmax |/s applications.

The same argument can be repeated for the case m = −|mmax |, this time by
using the raising operator R+ , as well as Eq.(3.4.20) and Eq.(3.4.21). The same
result is reached regarding the relation between |mmax |/s and the number of wave
functions in the multiplet.

3.4.3 Domains and Hermiticity of the Components of the

Runge-Lenz Vector

Let us now address the questions of Hermiticity and of the domains of the various
operators. Once it is endowed with an appropriate extension, the Hermitean kinetic
energy operator T becomes self-adjoint and thus qualifies as a physical observable.
The same is true for the full Hamiltonian including the potential. In this case,
3.4. RUNGE-LENZ VECTOR AT THE QUANTUM LEVEL 45

we assume the standard Friedrichs extension [21], which implies that there is no
δ-function potential located at the tip of the cone.

Using ∂r† = −∂r − 1/r as well as ∂ϕ† = −∂ϕ , it is straightforward to show that, at
least formally, Rex† = R
ex and R
ey† = Rey , which implies Re±†
=R e∓ . However, Hermitic-
ity also requires appropriate boundary conditions, which restrict the domains of the
corresponding operators. It is interesting to note that, using sϕ = χ, the operators
Rx and Ry of Eq.(3.4.3) formally agree with the components of the standard Runge-
Lenz vector for the plane from Eq.(3.1.1). The Runge-Lenz vector for the plane is
a Hermitean and even self-adjoint operator acting in a domain D[R] ~ that contains
the domain of the Hamiltonian. This domain contains smooth functions which are
2π-periodic in the polar angle χ of the plane. The operators R ex and R ey , on the other
hand, act on the Hilbert space of square-integrable wave functions on the cone. In
this case, the domain of the Hamiltonian D[H] contains smooth functions which are
2π-periodic in the rescaled angle ϕ and obey the boundary condition of Eq.(3.2.20).
While R ex and R ey on the cone are still Hermitean in their appropriate domain, in
contrast to the case of the plane, they are not Hermitean in the domain D[H] of
the Hamiltonian. In particular, for s 6= 1 the operators R ex and R ey map 2π-periodic
physical wave functions onto functions outside D[H], because they contain multi-
plications with the 2π/s-periodic functions cos(sϕ) and sin(sϕ). Proper symmetry
generators should map wave functions from the domain of the Hamiltonian back
into D[H]. Hence, for s ∈ / N, the operators R ex and R ey do not represent proper
symmetry generators.

It is interesting to consider the case of rational s = p/q with p, q ∈ N. In this

case, a single application of
e± = R
R ex ± iRey (3.4.29)

may take us out of the domain of the Hamiltonian, but a q-fold application of these
operators brings us back into D[H]. Indeed, just as for rational s the classical object
R±q
represents a proper conserved physical quantity, Re±q
(but not Re± itself) qualifies
as a proper symmetry generator. The case of integer s = n is also interesting,
because in that case cos(sϕ) and sin(sϕ) are indeed 2π-periodic. Hence, by acting
with Re± we might expect to stay within D[H], although for n ≥ 3 the abstract
spin S = nr + |m|/s = nr + |m|/n is still quantized in unusual fractional units.
However, we already saw this above, another subtlety arises because R e± may turn
a physical wave function that is regular at the origin (and thus obeys the boundary
condition of Eq.(3.2.20) into a singular one. This further limits the domain of the
operators Re± . The unusual (not properly quantized) value of the Casimir spin can
be traced back to the mathematical fact that the Runge-Lenz vector — although
Hermitean in its appropriate domain — does not act as a Hermitean operator in the
domain of the Hamiltonian. Hence, in retrospect the SU(2) commutation relations
46 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

of Eq.(3.4.6) are rather formal. In fact, they are satisfied for functions Ψ(r, ϕ) with
ϕ ∈ R, but not for the periodic functions in D[H] for which ϕ ∈ [0, 2π]. This is
another indication that the accidental “SU(2)” symmetry of Eq.(3.4.6) is rather
unusual.

3.4.4 Unusual Multiplets

In this subsection we study the multiplets by determining the Casimir spectrum,

and examine carefully the implication of the results of subsection (3.4.2) on the
multiplets.

As we saw in section (2.3) for the case of an SU(2) algebra, the spectrum of
the Casimir operator and the matrix elements of the raising and lowering operators
were derived under the assumption that the wave functions are normalizable and
form complete set. On the other hand, we have shown that this is not always the
case since R− or R+ may map the wave function onto another one that is divergent
at the origin and therefore outside D[H]. However, one can derive the spectrum of
the Casimir operator without having to worry about the normalizability of the wave
functions. This will help to further understand the puzzling result that the Casimir
e± on a
spin S is not always quantized in integer or half-integer units. Acting with R
2π-periodic wave function

hr, ϕ|nr , mi = ψnr ,m (r) exp(imϕ) (3.4.30)

one changes both nr ∈ N and m ∈ Z. For m > 0 one obtains

e+ |nr , mi ∝ |nr − 1, m + si, R

R e− |nr , mi ∝ |nr + 1, m − si, (3.4.31)

and for m < 0 one finds

e+ |nr , mi ∝ |nr + 1, m + si, R

R e− |nr , mi ∝ |nr − 1, m − si. (3.4.32)

Finally, for m = 0 we have

e+ |nr , 0i ∝ |nr − 1, si, R

R e− |nr , 0i ∝ |nr − 1, −si. (3.4.33)

These relations follow from the SU(2) algebra which implies that R e± are raising
and lowering operators for L e = L/s. Hence, by acting with R e± , the eigenvalue m
of L is shifted by ±s. Using the fact that the eigenvalue of the Casimir operator,
which is determined by S = nr + |m|/s, does not change under applications of
e± , one immediately obtains the effects of R
R e± on the radial quantum number nr .
Eqs.(3.4.31), (3.4.32), and (3.4.33) also follow directly by applying the explicit forms
3.4. RUNGE-LENZ VECTOR AT THE QUANTUM LEVEL 47

of R± = Rx ± iRy from Eq.(3.4.3) to the wave functions of Eq.(3.3.12) and using the
relations in section(3.4.2). Moreover, from the results of section(3.4.2) we obtain

e+
R nr +1 e+ |0, m + nr s ≥ 0i = 0,
|nr , m ≥ 0i ∝ R
e−
R nr +1 e− |0, m − nr s ≤ 0i = 0.
|nr , m ≤ 0i ∝ R (3.4.34)

e+ or with
Hence, depending on the sign of m, by acting nr + 1 times either with R
e− we reach zero, and thus the multiplet naturally terminates. This allows us to
R
confirm the value of the Casimir spin S = nr + |m|/s by evaluating
 
1 e e e− R
e+ ) + L e |0, m + nr s ≥ 0i
2
C|0, m + nr s ≥ 0i = (R+ R− + R
2
 
1 e e e e e 2
= ([R+ , R− ] + 2R− R+ ) + L |0, m + nr s ≥ 0i
2
= e+L
(L e2 )|0, m + nr s ≥ 0i
m  m 
= + nr + nr + 1 |0, m + nr s ≥ 0i
s s
= S(S + 1)|0, m + nr s ≥ 0i,
 
1 e e e e e 2
C|0, m − nr s ≤ 0i = (R+ R− + R− R+ ) + L |0, m − nr s ≤ 0i
2
 
1 e e e− , R
e− ]) + L e |0, m − nr s ≤ 0i
2
= (R+ R− + 2[R
2
= (−L e+L e2 )|0, m − nr s ≤ 0i
 m  m 
= − + nr − + nr + 1 |0, m − nr s ≤ 0i
s s
= S(S + 1)|0, m − nr s ≤ 0i. (3.4.35)

The multiplet of degenerate states with the same value of S can now be obtained
by n repeated applications of either R e+ or R e− . It is important to note that, if s
is not an integer, m ± ns may also not be an integer and thus the corresponding
state may be outside D[H]. Despite this, its radial wave function is still defined by
Eq.(3.3.12) and it still solves the radial Schrödinger equation.

Let us first consider the generic case of irrational s. In that case, the classical
orbits are not closed, there are no accidental degeneracies in the discrete spectrum
of the Hamiltonian, and the Casimir spin S = nr + |m|/s is irrational. Acting
with Re± on the 2π-periodic wave function |nr , m ∈ Zi an arbitrary number of
times, one generates functions which are not 2π-periodic and thus outside D[H].
As a consequence of parity symmetry, for m 6= 0 the two levels with the quantum
numbers m and −m are still degenerate. However, that two-fold degeneracy is not
accidental.
48 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

Next, let us discuss the case of rational s = p/q in which all classical orbits
are closed and there are accidental degeneracies in the discrete spectrum of the
Hamiltonian. First, we consider the case 2|m|/s = 2|m|q/p ∈ N for which the
Casimir spin S is an integer or a half-integer. Only in that case, the set of degenerate
wave functions terminates on both ends, i.e.
e2S+1 |0, m + nr s ≥ 0i = 0, R
R e2S+1 |0, m − nr s ≤ 0i = 0. (3.4.36)
− +

This follows by applying the operators of Eq.(3.4.3) to the wave function given in
Eq.(3.3.12) that we discussed in the previous subsection (3.4.2).

The multiplets are even more unusual in the case of rational s = p/q with the
Casimir spin S neither being an integer nor a half-integer. In that case, the set
of degenerate wave functions only terminates on one end, but not on the other.
In particular, while still R e+ |0, m + nr s ≥ 0i = 0, R ek |0, m + nr s ≥ 0i does not
−
vanish, even for arbitrarily large k (see the end of subsection (3.4.2)). Since an
infinite number of values m + (nr − k)s are integers, one might think that the
multiplet of degenerate states inside D[H] should contain an infinite number of
states. Interestingly, this is not the case for a rather unusual reason. For S neither
e−
being an integer nor a half-integer, the states R k
|0, m+nr s ≥ 0i with m+(nr −k)s <
0 are outside D[H] because the corresponding wave function is singular at the origin.
This again follows from applying the operators of Eq.(3.4.3) to the wave function
of Eq.(3.3.12). Although they do not qualify as physical states, the divergent wave
functions still are mathematical solutions of the Schrödinger differential equation
which take the form
αr 2|m|
ψ(r) = A exp(− )(αr)−|m|/s 1 F1 (−nr , − + 1, αr). (3.4.37)
2 s
The singularity may or may not make the wave function non-normalizable. Even if it
remains normalizable, the corresponding singular wave function does not belong to
D[H] because it does not obey the boundary condition of Eq.(3.2.20). For S neither
being an integer nor a half-integer, the states with positive and negative m have the
same energy as a consequence of parity symmetry, but they are not related to one
another by applications of the raising and lowering operators R e± . Remarkably, in
this case, by acting with a symmetry generator R ex or Rey on a wave function inside
D[H], one may generate a physically unacceptable wave function outside D[H]. A
sequence of physical and unphysical wave functions is illustrated in figure (3.4). To
summarize, for s 6= 1 different types of unusual multiplets arise. First, even for
integer or half-integer S = nr + |m|/s, the degeneracy of the physical multiplet is
not 2S + 1 because m ± ns may not be an integer in which case the corresponding
wave function is not 2π-periodic. When S = nr + |m|/s is neither an integer nor a
half-integer, there is an infinite number of degenerate solutions of the Schrödinger
equation. However, only a finite number of them obeys the boundary condition of
Eq.(3.2.20) and thus belongs to D[H].
3.4. RUNGE-LENZ VECTOR AT THE QUANTUM LEVEL 49

0 3 6 9 12 0 2 4 6 0 5 10 15

0 0 0 0

3 6 9 12 15 2 4 6 8 5 10 15 20
Αr Αr Αr

Figure 3.4: A sequence of wave functions for the 1/r potential with s = 3 obtained
e− . The quantum numbers are nr = 0, m = 4 (left),
from repeated applications of R
nr = 1, m = 4 − s = 1 (middle), and nr = 2, m = 4 − 2s = −2 (right). The
third state in the sequence is outside the domain of the Hamiltonian because the
corresponding wave function does not obey the boundary condition of Eq.(3.2.20)
and the state is thus unphysical.

3.4.5 Counting the Degeneracies in the Kepler Case

Our experience with SU(2) algebras would suggest that there are 2S + 1 degenerate
states. However, we should not forget that a single application of the raising and
lowering operators R e± may take us outside D[H], and only q applications of R e±
take us back into D[H]. As it is obvious, not all of the members of the multiplet
represent physical wave functions, because not all of them are 2π-periodic and are
regular at the origin. The ϕ-dependent part for each member of the multiplet is
exp(iϕ(mmax − sk)). One way to realize that is from studying the action of the
lowering operator R− on the wave function in Eq.(3.4.22) for k applications. For
the function to be 2π-periodic, (mmax − sk) must be an integer. As we proved
before, k could be finite for the case of |mmax |/s an integer or a half-integer. Then
k = 0, 1, ..., 2|mmax |/s or infinite when |mmax |/s is a fraction. A rule can be derived
by induction for the degeneracy g. When we begin with mmax as an integer, the
rule is    
2|mmax | 2S
g= +1= + 1. (3.4.38)
p q
We must note here that it is possible that |mmax | has a non-integer value. For
example, consider the case when S = nr + |m|/s = 4, with s = 1/3, m = 1, and
nr = 1. According to Eq.(3.4.11), applying the raising operator on the wave function
gives nr = 0, m = mmax = m + s = 1 + 1/3 = 4/3. For such case, the wave function
is unphysical although the multiplet has a finite number of members when |mmax |/s
is an integer or a half-integer. Counting the degeneracies needs more elaboration.
50 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

Let us assume that mmax = P/Q is not an integer. Now let us assume that i1
applications of R− are needed until one reaches the closest integer to P/Q say b.
One then obtains
P p
− i1 = b, (3.4.39)
Q q
which implies
q
i1 = (P − Qb). (3.4.40)
pQ
The rule of finding the degeneracy for b an integer is given by Eq.(3.4.38). Accord-
ingly we obtain  
2b
g= + 1. (3.4.41)
p
The previous argument shows that we found an unusual representation for the un-
conventional “SU(2)” algebra.

3.5 The r2 Potential on a Cone

Let us now turn to the problem of a particle moving on a cone and bound to its tip
by a harmonic oscillator potential
1
V (r) = Mω 2 r 2 . (3.5.1)
2
The Hamiltonian is then given by
 
1 L2 1
H =T +V = 2
pr + 2 2 + Mω 2 r 2 . (3.5.2)
2M r s 2

3.5.1 Classical Solutions

Using the corresponding classical equations of motion one obtains the classical orbits
1 MEs2
= [1 + f cos(2s(ϕ − ϕ0 ))], (3.5.3)
r2 L2
with E and L again denoting energy and angular momentum and with
r
ω 2 L2
f = 1− 2 2. (3.5.4)
E s
The radial component of the momentum is given by
pr MEs
= f sin(2s(ϕ − ϕ0 )). (3.5.5)
r L
3.5. THE R2 POTENTIAL ON A CONE 51

All classical orbits are closed as long as 2s = p/q is a rational number (with p, q ∈ N
again not sharing a common divisor). Some examples of classical orbits are shown
in figure (3.5).

Figure 3.5: Examples of bound classical orbits for the r 2 potential with s = 3 (left),
s = 12 (middle), and s = 1 (right). The latter case represents an elliptic orbit
of the standard harmonic oscillator. The orbits are shown in the x-y-plane with
(x, y) = r(cos ϕ, sin ϕ) where ϕ = χ/s ∈ [0, 2π] is the rescaled polar angle.

3.5.2 Semi-classical Bohr-Sommerfeld Quantization

As in the case of the 1/r potential, the semi-classical quantization condition for the
angular momentum is again given by L = m ∈ Z. For the harmonic oscillator the
quantization condition for the radial motion takes the form

I Z π/s    
|L|f 2 sin2 (2s(ϕ − ϕ0 )) E |L| 1
dr pr = dϕ =π − = 2π nr + ,
0 (1 + f cos(2s(ϕ − ϕ0 ))2 ω s 2
(3.5.6)
such that
 
|m|
E = ω 2nr + +1 . (3.5.7)
s

Again, it will turn out that the semi-classical result exactly reproduces the one of
the full quantum theory.
52 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

3.5.3 Solution of the Schrödinger Equation

For the particle on the cone with harmonic oscillator potential the radial Schrödinger
equation takes the form
   
1 2 1 m2 1 2 2
− ∂r + ∂r + + Mω r ψ(r) = Eψ(r). (3.5.8)
2M r 2Mr 2 s2 2

In this case, the solution is given by [8]

α2 r 2 |m| √
ψnr ,m (r) = A exp(− )(αr)|m|/s 1 F1 (−nr , + 1, α2r 2 ), α = Mω. (3.5.9)
2 s
The corresponding quantized energy values are given by Eq.(3.5.7). There are acci-
dental degeneracies if 2s = p/q is a rational number, which thus again arise exactly
when all classical orbits are closed. Some features of the energy spectrum are illus-
trated in figure (3.6).

Αr

Figure 3.6: The r 2 potential (solid curve) together with an effective potential includ-
ing the centrifugal barrier with m = ±1 (dashed curve) for s = 21 . The energies
of the ground state and the first two excited states are indicated by horizontal lines.
The numbers besides the lines specify the degree of degeneracy. The ground state
(with nr = 0, m = 0) is non-degenerate, while the first excited level (consisting of
the states with nr = 0, m = ±1 and nr = 1, m = 0) and the second excited level
(consisting of the states with nr = 0, m = ±2, nr = 1, m = ±1, and nr = 2, m = 0)
are accidentally three-fold, respectively five-fold degenerate.
3.5. THE R2 POTENTIAL ON A CONE 53

3.5.4 Runge-Lenz Vector

The accidental degeneracies for rational s again point to the existence of a conserved
Runge-Lenz vector. At the classical level, we can use Eqs.(3.5.3) and (3.5.5) to write

L2
Ef cos(2s(ϕ − ϕ0 )) = Ef [cos(2sϕ) cos(2sϕ0 ) + sin(2sϕ) sin(2sϕ0 )] = − H,
Mr 2 s
pr L
Ef sin(2s(ϕ − ϕ0 )) = Ef [sin(2sϕ) cos(2sϕ0 ) − cos(2sϕ) sin(2sϕ0 )] = ,(3.5.10)
Mrs
such that
 
L2 pr L
Rx = Ef cos(2sϕ0 ) = − H cos(2sϕ) + sin(2sϕ),
Mr 2 s2 Mrs
 
L2 pr L
Ry = Ef sin(2sϕ0 ) = 2 2
− H sin(2sϕ) − cos(2sϕ). (3.5.11)
Mr s Mrs

It should again be pointed out that Rx and Ry are proper conserved quantities only
if 2s is an integer. Otherwise the Runge-Lenz vector is not a 2π-periodic function
of the angle ϕ, and its value depends on the number of revolutions of the particle
around the tip of the cone. As before, it is useful to introduce the complex quantities
 
L2 pr L
R± = Rx ± iRy = −H ∓i exp(±2isϕ). (3.5.12)
Mr 2 s2 Mrs

For rational values 2s = p/q (with p, q ∈ N) the quantities

 q
q L2 pr L
R± = 2 2
−H ∓i exp(±ipϕ) (3.5.13)
Mr s Mrs

are again single-valued functions of x = r cos ϕ and y = r sin ϕ, and are hence proper
conserved quantities.

For the harmonic oscillator, the length of the Runge-Lenz vector is given by
 2 2 
2 L2 pr L
R = Rx2
+ Ry2
= + −H
Mr 2 s2
Mrs
 2 2
2  2  2
pr L 1 2 2 pr L 2 L
= − + Mω r + =H − ω (. 3.5.14)
2M 2Mr 2 s2 2 Mrs s

As in the case of the 1/r potential, it is useful to introduce rescaled variables which
now take the form
ex = 1 Rx , R
R ey = 1 Ry , L e= L. (3.5.15)
2ω 2ω 2s
54 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

We thus obtain
 2
H √
C= e2
R + e2
R e2
+L = ⇒ H = 2ω C. (3.5.16)
x y
2ω

Once again, it will turn out that the quantum analogue of C is the Casimir operator
of an accidental SU(2) symmetry.

3.5.5 The Runge-Lenz Vector as a Quantum Mechanical

Operator

We are using the same technique that has been used in the case of the 1/r potential.
The classical expression of the Runge-Lenz vector suggests the following ansatz

1 1 1 1
R= 2
A1 (ϕ)∂ϕ2 + 2 A2 (ϕ)∂ϕ + A3 (ϕ)∂ϕ ∂r + A4 (ϕ)∂r + A5 (ϕ)∂r2 + r 2 A6 (ϕ).
r r r r
(3.5.17)
As before the commutators of the above operator with the Hamiltonian give a system
of differential equations as follows

4n2 A6 (ϕ) + 2kMn2 A4 (ϕ) + 2kMn2 A5 (ϕ) + ∂ϕ2 A6 (ϕ) = 0,

4n2 A2 (ϕ) + ∂ϕ2 A2 (ϕ) = 0,
kMn2 A3 (ϕ) + ∂ϕ A6 (ϕ) = 0,
4n2 A1 (ϕ) + 2A4 (ϕ) − 6A5 (ϕ) + 2∂ϕ A2 (ϕ) + ∂ϕ A1 (ϕ) = 0,
A3 (ϕ) + ∂ϕ A1 (ϕ) = 0,
2n2 A4 (ϕ) − 2n2 A5 (ϕ) + ∂ϕ2 A4 (ϕ) = 0,
A6 (ϕ) + kMA5 (ϕ) = 0,
−4n2 A2 (ϕ) + 2n2 A3 (ϕ) + 2∂ϕ A4 (ϕ) + ∂ϕ2 A3 (ϕ) = 0,
−2n2 A1 (ϕ) + 2A5 (ϕ) + ∂ϕ A3 (ϕ) = 0,
−2n2 A4 (ϕ) + 2n2 A5 (ϕ) + ∂ϕ2 A5 (ϕ) = 0,
−n2 A3 (ϕ) + ∂ϕ A5 (ϕ) = 0. (3.5.18)

Solving the differential equations for this case is more complicated than in the case
of the 1/r potential. The differential Eqs.(3.5.18) give two solutions for which all
A’s are non-zero functions of ϕ. These two solutions represent the two components
of the Runge-Lenz operator. In addition to that, we have a third solution with all
the A’s equal to zero except A2 equal to a constant. This solution represents the
one-component angular momentum operator. At the quantum level the Runge-Lenz
3.5. THE R2 POTENTIAL ON A CONE 55

vector now takes the form

1 1 1
Rx = cos(2sϕ)∂r2 − 2 2
cos(2sϕ)∂ϕ2 + sin(2sϕ)∂ϕ
2M 2Mr s Mr 2 s
1 1 1
− Mω 2 r 2 cos(2sϕ) − sin(2sϕ)∂r ∂ϕ − cos(2sϕ)∂r ,
2 Mrs 2Mr
1 1 1
Ry = sin(2sϕ)∂r2 − 2 2
sin(2sϕ)∂ϕ2 − cos(2sϕ)∂ϕ
2M 2Mr s Mr 2 s
1 1 1
− Mω 2 r 2 sin(2sϕ) + cos(2sϕ)∂r ∂ϕ − sin(2sϕ)∂r . (3.5.19)
2 Mrs 2Mr
One can show that the Runge-Lenz vector as well as the angular momentum L
commute with the Hamiltonian, and that these operators obey the algebra
L
[Rx , Ry ] = 2iω , [Rx , L] = −2isRy , [Ry , L] = 2isRx . (3.5.20)
s
Applying the rescaling of Eq.(3.5.15), this leads to
ex , R
[R ey ] = iL,
e [R
ey , L]
e = iR
ex , [L,
e Rex ] = iR
ey , (3.5.21)

which again represents an SU(2) algebra.

3.5.6 Casimir Operator

The Casimir operator for the harmonic oscillator on the cone takes the form
 2
e 2 e 2 e 2 H 1
C = Rx + Ry + L = − , (3.5.22)
2ω 4
which implies r  
1 1
H = 2ω C + = 2ω S + . (3.5.23)
4 2
Comparing with Eq.(3.5.7) for the energy spectrum, we now identify
|m|
S = nr + . (3.5.24)
2s

3.5.7 The Raising and Lowering Operators Acting on the

Wave Function

The explicit form of the lowering operator can be found by substituting the expres-
sions for Rx and Ry from Eq.(3.5.19) into Eq.(3.4.7). After that one applies the
56 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

lowering operator to the explicit form of the wave function in Eq.(3.5.9). The result
depends on whether |m|/s is an integer or a half-integer or not. We will study both
cases and prove that the multiplet terminates under repeated applications of R−
when |m|/s is an integer or a half-integer, while it does not terminate otherwise.
One application of R− on the wave function with m ≥ 0 gives

2α2 |m|
−2 α2 r 2
R− ψnr ,m (r, ϕ) = A (αr) s exp(− ) exp(iϕ(m − 2s))
M( |m|
s
+ 1)( |m|
s
+ 2) 2
|m| |m| |m| |m| |m|
× ( + 1)( + 2)( − 1 − α2 r 2 )1 F1 (−nr , + 1, α2 r 2 )
s s s s s
 2

|m| |m|
× +nr − 2 2 + 2α2 r 2 + (−4 + α2 r 2 ) r 2 α2
s s
|m|
× 1 F1 (1 − nr , + 2, α2r 2 ) + nr (nr − 1)α4 r 4
s !
|m|
× 1 F1 (2 − nr , + 3, α2r 2 ) . (3.5.25)
s

Another important identity can be derived using the recurrence relations of the
confluent hypergeometric functions [31], and after a lengthy calculation we can prove
that

b(b2 − 1)(b − 2 − x)1 F1 (a, b, x) + (1 + a)ax2 1 F1 (a + 2, b + 2, x)

− ax(2 − 2b2 + x + bx)1 F1 (a + 1, b + 1, x)
= b(b2 − 1)(b − 2)1 F1 (a − 1, b − 2, x). (3.5.26)

Using the above identity, Eq.(3.5.25) can be written as

2α2 |m| α2 r 2
R− ψnr ,m (r, ϕ) = A (αr) s −2 exp(− ) exp(iϕ(m − 2s))
M 2
|m| |m| |m|
× ( − 1)1 F1 (−1 − nr , − 1, α2 r 2 )
s s s
= Bψnr +1,m−2s (r, ϕ), (3.5.27)

where B is a constant that results from one application of R− or R+ on the wave

function. The value of this constant is not important in our argument. We must
keep in mind that when B appears in different equations it may not have the same
value, but that will not affect the argument. A number of k applications of R− to
3.5. THE R2 POTENTIAL ON A CONE 57

the wave function with m ≥ 0 gives

k 2α2 k m|
−2k α2 r 2
R− ψnr ,m (r, ϕ) = A( ) (αr) s exp(− ) exp(iϕ(m − 2sk))
M 2
Yk   
|m| |m|
× + 2 − 2j + 1 − 2j
j=1
s s
|m|
× 1 F1 (−k − nr ,+ 1 − 2k, α2r 2 )
s
= Bψnr +k,m−2sk (r, ϕ). (3.5.28)

The lowering operator R− operating on the wave function with m < 0 gives

2α2 A |m|
+2 α2 r 2
R− ψnr ,−|m| (r, ϕ) = (αr) s exp(− )
M( |m|
s
+ 1)( |m|
s
+ 2) 2
|m| |m|
× exp(iϕ(m − 2s))nr ( + 2)1 F1 (1 − nr , + 2, α2 r 2 )
s s
!
|m|
+ (nr − 1)1 F1 (2 − nr , + 3, α2r 2 ) . (3.5.29)
s

Using identity (3.5.26), the above equation can be written as

2α2 |m|
+2 α2 r 2
R− ψnr ,m (r, ϕ) = A (αr) s exp(− )
M( |m|
s
+ 1)( |m|
s
+ 2) 2
|m| |m|
× exp(iϕ(m − 2s))nr (nr + + 1)1 F1 (1 − nr , + 3, α2 r 2 )
s s
= Bψnr −1,m−2s) (r, ϕ). (3.5.30)

The general rule for k applications of R− on the wave function is

k 2α2 k |m| α2 r 2
R− ψnr ,m (r, ϕ) = A( ) (αr) s +2k exp(− ) exp(iϕ(m − 2sk))
M 2 !
k
Y (nr − j + 1)(nr + |m| + j) |m|
× |m| |m|
s
1 F1 (k − nr , + 1 + 2k, α2 r 2 )
j=1 ( + 2j − 1)( + 2j) s
s s
= Bψnr −k,m−2sk (r, ϕ). (3.5.31)
58 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

The raising operator R+ operating on the wave function with m ≥ 0 gives

2α2 |m|
+2 α2 r 2
R+ ψnr ,m (r, ϕ) = A (αr) s exp(− ) exp(iϕ(m + 2s))
M( |m|
s
+ 1)( |m|
s
+ 2) 2
|m| |m|
× nr ( + 2)1 F1 (1 − nr , + 2, α2r 2 )
s s
!
|m|
+ (nr − 1)1 F1 (2 − nr , + 3, α2r 2 ) . (3.5.32)
s

Using the recurrence relations of the confluent hypergeometric function [31] the
above equation can be written as

2α2 |m|
+2 α2 r 2
R+ ψnr ,m (r, ϕ) = A (αr) s exp(− )
M( |m|
s
+ 1)( |m|
s
+ 2) 2
|m| |m|
× exp(iϕ(m + 2s))nr (nr + + 1)1 F1 (1 − nr , + 3, α2 r 2 )
s s
= Bψnr −1,m+2s (r, ϕ). (3.5.33)

The general rule for k applications of R+ on the wave function with m ≥ 0 is

k α2 k |m| α2 r 2
R+ ψnr ,m (r, ϕ) = A(2 ) (αr) s +2k exp(− ) exp(iϕ(m + 2sk))
M 2 !
k
Y (nr − j + 1)(nr + |m| + j) |m|
× |m| |m|
s
1 F1 (k − nr , + 1 + 2k, α2 r 2 )
j=1 ( + 2j − 1)( + 2j) s
s s
= Bψnr −k,m+2sk (r, ϕ). (3.5.34)

The general rule for k applications of R+ on the wave function with m < 0 is

k α2 |m| α2 r 2
R+ ψnr ,m (r, ϕ) = A(2 )k (αr) s −2k exp(− ) exp(iϕ(m + 2sk))
M 2
Yk   
|m| |m|
× + 2 − 2j + 1 − 2j
j=1
s s
|m|
× 1 F1 (−k + 1 − 2k, α2r 2 )
− nr ,
s
= Bψnr +k,m+2sk (r, ϕ). (3.5.35)

To count the number of wave functions in the multiplet we begin with nr = 0.

Then clearly Eq.(3.5.24) gives S = mmax /s. Consider first the case of m ≥ 0 with
m = mmax ≥ 0. The value of m = mmax can not be raised further because applying
3.5. THE R2 POTENTIAL ON A CONE 59

R+ gives zero according to Eq.(3.5.34). Now we apply the lowering operator R− a

number of k0 times, with k0 = mmax /s + 1/2. Using Eq.(3.5.28) we obtain

k0 2α2 k0 |mmax |
−2k0 α2 r 2
R− ψ0,mmax (r, ϕ) = A( ) (αr) s exp(− ) exp(iϕ(mmax − 2sk0 )
M 2
Yk0   
|mmax | |mmax |
× + 2 − 2j + 1 − 2j
j=1
s s
|mmax |
× 1 F1 (−k0 , + 1 − 2k0 , α2 r 2 ). (3.5.36)
s
Here we have two cases in which the multiplet terminates:

Half-integer mmax /2s ≥ 0

Applying the raising operator R+ once gives zero according to Eq.(3.5.34). On the
other hand, applying the lowering operator R− a number of k0 = (mmax /2s) + 1/2
times with mmax /s an odd number, and using the identity in Eq.(3.4.14), then
Eq.(3.5.36) can be written as

k0 α2 r 2 |mmax | 1
R− ψ0,mmax (r, ϕ) = C(αr)1 exp(− ) exp(−iϕs)1 F1 (− + , 2, α2r 2 )
2 2s 2
= Bψ |mmax | − 1 ,−s (r, ϕ), (3.5.37)
2s 2

where C is a constant. As we can see, the resulting wave function has m = −s <
′
0. The operator R− can be applied further k0 times. However, we must use the
appropriate relation Eq.(3.5.31), and we get

k′ k0 ′ α2 r 2
R−0 (R− ψ0,mmax (r, ϕ)) = C(αr)1+2k0 exp(− ) exp(−iϕs(1 + 2k0′ ))
2
k0′ !
Y ( |m2s
max |
+ 2 − j)( |m2s
1 max |
+ 21 + j)
×
j=1
2j(1 + 2j)
|mmax | 1
′
× 1 F1 (k0 − + , 2 + 2k0′ , α2 r 2 )
2s 2
= Bψ |mmax | − 1 −k′ ,−s−2sk′ (r, ϕ). (3.5.38)
2s 2 0 0

It is obvious from the above equation that applying R− a number of |mmax |/2s + 1/2
times gives zero. Therefore, the number of wave functions in the multiplet is

′ |mmax |
N = k0 + k0 = + 1. (3.5.39)
s
60 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

Integer mmax /2s ≥ 0

Applying the raising operator R+ once gives zero according to Eq.(3.5.34). On the
other hand applying the lowering R− operator a number of k0 = (mmax /2s)+1 times,
and after that using the identity in Eq.(3.4.15), then Eq.(3.5.36) can be written as

k0 2 α2 r 2 |mmax |
R− ψ0,mmax (r, ϕ) = C(αr) exp(− ) exp(−iϕ2s)1 F1 (− + 1, 3, α2r 2 )
2 2s
= Bψ |mmax | −1,−2s (r, ϕ). (3.5.40)
2s

As we can see, the resulting wave function has m = −2s < 0. The operator R−
can be applied further k0′ times. However, we must use the appropriate relation
Eq.(3.5.31), and we get

k′ k0 ′ α2 r 2
R−0 (R− ψ0,mmax (r, ϕ)) = C(αr)2+2k0 exp(− ) exp(−iϕs(2 + 2k0′ ))
2
k0′ !
Y ( |m2s
max |
− j)( |m2s
max |
+ 1 + j)
×
j=1
(2j + 2)(1 + 2j)

′ |mmax |
× 1 F1 (k0 − + 1, 3 + 2k0′ , α2 r 2 )
2s
= Bψ |mmax | −1−k′ ,−2s−2sk′ (r, ϕ). (3.5.41)
2s 0 0

It is obvious from the above equation that applying R− a number of |mmax |/2s times
gives zero. Therefore, the number of wave functions in the multiplet is
′ |mmax |
N = q0 + q0 = + 1. (3.5.42)
s

Not Integer or half integer mmax /s

Applying the lowering operator R− any number of times will not make either iden-
tity (3.4.14) or identity (3.4.15) applicable, since neither (|mmax |/s) + 1 − 2j nor
(|mmax |/s) + 2 − 2j in Eq.(3.5.28) is equal to 0 or −1 for any value of j ∈ Z. After
k0 > |mmax |/s applications of R− , (αr)(|mmax |/s)−2k0 is negative, and the resulting
wave functions diverge at the origin although it is still a solution of the Schrödinger
equation. The multiplet does not terminate for any number of applications of R− ,
and the number of wave functions in such a multiplet is infinite. The multiple ap-
plication of R− transforms a well-behaved wave function to one that diverges at the
origin after k0 > |mmax |/2s applications.

The same argument can be repeated for the case m = −|mmax |, this time by
using the raising operator R+ , as well as Eq.(3.5.34) and Eq.(3.5.35). The same
3.5. THE R2 POTENTIAL ON A CONE 61

result is reached regarding the relation between |mmax |/s and the number of wave
functions in the multiplet.

3.5.8 Unusual Multiplets

Let us now consider the unusual multiplets in case of the harmonic oscillator. The
discussion is similar to the one of the 1/r potential and will thus not be repeated in
all details. For m > 0 one now obtains
e+ |nr , mi ∝ |nr − 1, m + 2si, R
R e− |nr , mi ∝ |nr + 1, m − 2si, (3.5.43)
and for m < 0 one finds
e+ |nr , mi ∝ |nr + 1, m + 2si, R
R e− |nr , mi ∝ |nr − 1, m − 2si, (3.5.44)
while, for m = 0 we have
e+ |nr , 0i ∝ |nr − 1, 2si, R
R e− |nr , 0i ∝ |nr − 1, −2si. (3.5.45)
e±
As before, these relations follow from the SU(2) algebra which now implies that R
are raising and lowering operators for L e = L/2s. Hence, by acting with R e± the
eigenvalue m of L is now shifted by ±2s.

One now confirms the value of the Casimir spin S = nr + |m|/2s by evaluating
e+L
C|0, m + 2nr s ≥ 0i = (L e2 )|0, m + 2nr s ≥ 0i
m m 
= + nr + nr + 1 |0, m + 2nr s ≥ 0i
2s 2s
= S(S + 1)|0, m + 2nr s ≥ 0i,
C|0, m − 2nr s ≤ 0i = (−Le+L e2 )|0, m − 2nr s ≤ 0i
 m  m 
= − + nr − + nr + 1 |0, m − 2nr s ≤ 0i
2s 2s
= S(S + 1)|0, m − 2nr s ≤ 0i. (3.5.46)
The multiplet of degenerate states with the same value of S is again obtained by
e+ or R
repeated applications of R e− .

As in the case of the 1/r potential, for s 6= 1 different types of unusual multiplets
arise. Again, even for integer or half-integer S = nr + |m|/2s, the degeneracy of the
physical multiplet is not 2S + 1 because m ± 2ns may not be an integer in which
case the corresponding wave function is not 2π-periodic. When S = nr + |m|/2s is
neither an integer nor a half-integer, there is again an infinite number of degenerate
solutions of the Schrödinger equation. However, once more, only a finite number
of them obeys the boundary condition of Eq.(3.2.20) and thus belongs to D[H]. A
sequence of physical and unphysical wave functions is illustrated in figure (3.7).
62 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE

0 1 2 3 0 1 2 3 0 1 2 3

0 0
0 0

1 2 3 4 1 2 3 4 1 2 3
Αr Αr Αr

Figure 3.7: A sequence of wave functions for the r 2 potential with s = 3 obtained
e− . The quantum numbers are nr = 0, m = 7 (left),
from repeated applications of R
nr = 1, m = 7 − 2s = 1 (middle), and nr = 2, m = 7 − 4s = −5 (right). The
third state in the sequence is outside the domain of the Hamiltonian because the
corresponding wave function does not obey the boundary condition of Eq.(3.2.20)
and the state is thus unphysical.

3.5.9 Counting Degeneracies in the Case of the Isotropic

Harmonic Oscillator

The ϕ-dependent part of the wave function for each member of the multiplet is
exp(iϕ(mmax − 2sk)). For the function to be 2π-periodic (mmax − 2sk) must be
an integer. One way to realize that is by studying the application of the lowering
operator R− on the wave function a number of k times. The multiplet can be
finite for the case of |mmax |/2s being an integer or a half-integer. Then we have
k = 0, 1, ..., (|mmax |/s) + 1 or infinite when |mmax |/s is a fraction. A rule can be
derived by induction for the degeneracy g. When we begin with mmax as an integer,
the rule for even q is    
2|mmax | 2S
g= +1= + 1, (3.5.47)
p q
while the rule for odd q is
   
|mmax | S
g= +1= + 1. (3.5.48)
p q
For the case of non-integer mmax = P/Q, we use the same approach that was used
in the Kepler case. Let us assume that i1 applications of R− are needed to reach
the closest integer to P/Q, say b. One then obtains
P p
− 2i1 = b. (3.5.49)
Q q
3.5. THE R2 POTENTIAL ON A CONE 63

Evaluating i1 implies
q
i1 = (P − Qb) (3.5.50)
2pQ
Since the rule of finding the degeneracy for m = b being an integer is known and
given by Eq.(3.5.47), and Eq.(3.5.48), for even b, one obtains
 
2b
g= + 1, (3.5.51)
p

while for odd b one obtain  

b
g= + 1. (3.5.52)
p
64 CHAPTER 3. ACCIDENTAL SYMMETRY FOR MOTION ON A CONE
Chapter 4

Accidental Symmetries for a

Particle in a Constant Magnetic
Field on the Torus

4.1 Introduction to Cyclotron Motion

In this chapter we will discuss another case of accidental symmetry that emerges as
a result of classical orbits being closed. In cyclotron motion a particle moves on a
2-dimensional plane under the influence of a constant magnetic field perpendicular
to the plane. Then, the classical orbits are indeed closed and this gives rise to
an accidental symmetry [18, 19] as well as a conserved two-component Runge-Lenz
vector. When the plane of motion is infinite there is, in addition a conserved angular
momentum, as we explained in section (2.2). The angular momentum conservation
is due to rotational invariance. This symmetry is lost when the motion takes place on
a torus. This leads to less symmetry, and fewer degeneracies in the corresponding
quantum problem. The additional conserved quantities arise from the fact that
the vector pointing from the origin to the center of the circular cyclotron orbit has
conserved, analogous to the Runge-Lenz vector in the Kepler problem [19]. Also, the
radius of the cyclotron orbit is a conserved quantity directly related to the energy.
Interestingly, quantum mechanically, while the two coordinates of the center of the
circle are not simultaneously measurable, the radius of the circle has a sharp value
in an energy eigenstate. In the cyclotron problem, translation invariance disguises
itself as an “accidental” symmetry. As a consequence, the symmetry multiplets —
i.e. the Landau levels — are infinitely degenerate. In order to further investigate
the nature of the accidental symmetry, in [11] the charged particle in the magnetic
field was coupled to the origin by an r 2 harmonic oscillator potential. This explicitly

65
66 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

breaks translation invariance, and thus reduces the degeneracy to a finite amount,
while rotation invariance remains intact. In this chapter, we do the opposite, i.e. we
explicitly break rotation invariance, while leaving translation invariance (and hence
the accidental symmetry) intact by putting the problem on a torus. Remarkably,
the Polyakov loops of the Abelian vector potential, which are a consequence of the
non-trivial holonomies of the torus, give rise to non-trivial Aharonov-Bohm phases
which are observable at the quantum but not at the classical level. Analogous to
the quantum mechanical breaking of CP invariance due to the θ-vacuum angle in
non-Abelian gauge theories, here two self-adjoint extension parameters θx and θy
explicitly break the continuous translation invariance of the classical problem down
to a discrete magnetic translation group. This reduces the degeneracy to a finite
number, and allows us to further investigate the nature of the accidental symmetry.
In particular, just like for motion on a cone, symmetry manifests itself in a rather
unusual way in this quantum system. In particular, due to its relevance to the
quantum Hall effect, the Landau level problem has been studied very extensively (for
a recent review see [12]). For example, the problem has already been investigated
on a torus in [13, 32], however, without elaborating on the accidental symmetry
aspects. In this chapter, we concentrate exactly on those aspects, thus addressing
an old and rather well-studied problem from a somewhat unconventional point of
view.

4.2 Particle in the Infinite Volume

In this section we review more carefully what has been discussed in section (2.3.2),
and try to add new understanding to the standard knowledge about a non-relativistic
particle moving in a constant magnetic field in the infinite volume. We proceed from
a classical to a semi-classical, and finally to a fully quantum mechanical treatment.
In particular, we emphasize the symmetry aspects of the problem with a focus on
accidental symmetries. This section is a preparation for the case of a finite periodic
volume to be discussed in the next section.

4.2.1 Classical Treatment

Consider a non-relativistic electron of mass M and electric charge −e moving in a

~ = B~ez . The vector potential can be chosen as
constant magnetic field B
Ax (~x) = 0, Ay (~x) = Bx, Az (~x) = 0. (4.2.1)
The motion along the direction of the magnetic field is trivial because the velocity
component in the z−direction is just a constant with vanishing Poisson brackets with
4.2. PARTICLE IN THE INFINITE VOLUME 67

all the other conserved quantities. Therefore we restrict ourselves to 2-dimensional

motion in the x-y-plane. Obviously, this is just standard cyclotron motion. To
get started, in this subsection we treat the problem classically. The particle then
experiences the Lorentz force

F~ (t) = −e~v (t) × B~ez , (4.2.2)

which forces the particle on a circular orbit of some radius r. It moves along the
circle with an angular velocity ω, which implies the linear velocity v = ωr and the
acceleration a = ω 2 r. Hence, Newton’s equation takes the form
eB
Mω 2 r = eωrB ⇒ ω = , (4.2.3)
M
with the cyclotron frequency ω being independent of the radius r. Obviously, for
this system all classical orbits are closed. The same is true for a particle moving
in a 1/r or r 2 potential. In those cases, the fact that all bound classical orbits
are closed is related to the conservation of the Runge-Lenz vector which generates
a hidden accidental dynamical symmetry. Let us now investigate the question of
accidental symmetry for the particle in a constant magnetic field. The classical
Hamilton function takes the form
1 h i2 
~ x) = 1 p2 + (py + eBx)2 .

H= p~ + eA(~ x (4.2.4)
2M 2M
Let us define the following three quantities
   
eB eB
Px = px + eBy, Py = py , L = x py + x − y px + y . (4.2.5)
2 2

We are familiar with angular momentum section (2.2.2).It is straightforward to con-

vince oneself that H has vanishing Poisson brackets, {H, Px } = {H, Py } = {H, L} =
0. It is natural to think of Px , Py , and L as the gauge-covariant generators of trans-
lations and rotations. In particular, one obtains

{L, Px } = Py , {L, Py } = −Px , (4.2.6)

as one would expect for the rotation properties of the vector (Px , Py ). As is well-
known, however, in a magnetic field the two translations Px and Py do not commute,
i.e.
{Px , Py } = eB. (4.2.7)
How can these standard symmetry considerations be related to an accidental sym-
metry due to a Runge-Lenz vector? The Runge-Lenz vector is familiar from the
Kepler problem. It points from the center of force to the perihelion position, and
is conserved because all bound classical orbits are closed. Similarly, the orbit of
68 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

a charged particle in a constant magnetic field is a closed circle with a fixed cen-
ter. Indeed, in this case the position of this center plays the role of the conserved
Runge-Lenz vector and is given by
vy vy 1 py Py
Rx = x − r =x− =x− (py + eBx) = − =− ,
v ω Mω Mω eB
vx vx px Px
Ry = y + r = y + =y+ = . (4.2.8)
v ω Mω eB
Interestingly, the position (Rx , Ry ) of the center of the cyclotron circle is, at the
same time, proportional to (−Py , Px ), i.e. it is orthogonal to the generators of spatial
translations. Consequently, we can write
1
{Rx , Px } = − {Py , Px } = 1, (4.2.9)
eB
as well as
1 1
{Rx , Py } = − {Py , Py } = 0, {Ry , Px } = {Px , Px } = 0. (4.2.10)
eB eB
While Eqs.(4.2.9) and (4.2.10) look like the usual Poisson brackets of position and
momentum, one should not forget that Rx and Ry are just multiples of Py and Px ,
and should hence not be mistaken as independent variables. In particular, one also
obtains the relation
1
{Rx , Ry } = . (4.2.11)
eB
Hence, just like the two generators of translations, the x- and y-components of the
Runge-Lenz vector do not have a vanishing Poisson bracket. At the quantum level,
this will imply that the x- and y-components of the center of a cyclotron circle are
not simultaneously measurable with absolute precision. Another conserved quantity
is the radius r of the circular cyclotron orbit which can be expressed as
1 2 p2x 2H
r 2 = (x − Rx )2 + (y − Ry )2 = 2 2
(p y + eBx) + 2 2
= . (4.2.12)
M ω M ω Mω 2
This shouldn’t be a surprise since r 2 is proportional to the energy, therefore it should
be conserved.

4.2.2 Semi-classical Treatment

Next, we consider the same problem semi-classically, i.e. by using Bohr-Sommerfeld

quantization, which is equivalent to the quantization of angular momentum, i.e.
L = n. For a cyclotron orbit of radius r, it is easy to convince oneself that
r
eB 2 2n
L= r =n ⇒ r= . (4.2.13)
2 eB
4.2. PARTICLE IN THE INFINITE VOLUME 69

Consequently, in the semi-classical treatment the allowed radii of cyclotron orbits

are now quantized. Using Eq.(4.2.4) one finds for the energy
1
E = H = Mω 2 r 2 = nω. (4.2.14)
2
As is well-known, up to a constant ω/2, the semi-classically quantized energy values
are those of a harmonic oscillator with the cyclotron frequency ω.

4.2.3 Quantum Mechanical Treatment

Finally, we consider the problem fully quantum mechanically. In this case the
Schrödinger equation takes the form
1  2 
− ∂x + (∂y + ieBx)2 Ψ(~x) = EΨ(~x). (4.2.15)
2M
We now make the factorization ansatz

Ψ(~x) = ψ(x) exp(ipy y), (4.2.16)

and we obtain   
∂x2 1 2 py  2
− + Mω x + ψ(x) = Eψ(x). (4.2.17)
2M 2 Mω
This is nothing but the Schrödinger equation of a shifted harmonic oscillator. Hence,
the quantum mechanical energy spectrum takes the form
 
1
E =ω n+ . (4.2.18)
2
The energy of the charged particle is completely independent of the transverse mo-
mentum py . We can understand this from the classical solution of this problem.
The value of py = M ẏ + eAy depends on the choice of the gauge which is arbitrary.
While the total energy equal to M(ẋ2 + ẏ 2 )/2 is gauge-independent. The fact that
the energy is completely independent of the transverse momentum py leads to quan-
tized Landau levels with continuous infinite degeneracy. The energy eigenstates are
shifted one-dimensional harmonic oscillator wave functions ψn (x), i.e.
 py 
h~x|npy i = ψn x + exp(ipy y). (4.2.19)
Mω
Here ψn can be obtained by substituting the above Eq.(4.2.19) into Eq.(4.2.15), and
solving the differential equation. The solution is
  
py    py  α2  py  2
ψn x + = Hn α x + exp − x+ . (4.2.20)
Mω Mω 2 Mω
70 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

Similarly, one can construct eigenstates of the generator Px = −i∂x + eBy of in-
finitesimal translations (up to gauge transformations) in the x-direction
 px 
h~x|npx i = ψn y − exp(ipx x) exp(−ieBxy), (4.2.21)
Mω
where α2 = eB.
  
px    px  α2  px  2
ψn y − = Hn α y − exp − y− . (4.2.22)
Mω Mω 2 Mω

In the following we will prove that the two sets of eigenstates h~x|npy i and h~x|npx i
span the same subspace of the Hilbert space. For that to be the case, we must prove
that the following equation holds
Z ∞
h~x|npx i = dpy h~x|npy ihnpy |npx i. (4.2.23)
−∞

On the other hand

Z ∞ Z ∞
hnpy |npx i = dxdyhnpy |~xih~x|npx i. (4.2.24)
−∞ −∞

Substituting Eq.(4.2.19), Eq.(4.2.21), Eq.(4.2.20), and Eq.(4.2.22) in the above the

Eq.(4.2.23), and after using the following identity [33]
Z −∞ √
ξ2 η2
dξ exp(iξη) exp(− )Hn (ξ) = 2πin exp(− )Hn (η), (4.2.25)
∞ 2 2
we get
px py
hnpy |npx i = exp(−i ). (4.2.26)
α2
Substituting Eq.(4.2.26) in Eq.(4.2.23) and using Eq.(4.2.25) again gives
Z ∞   px 
dpy h~x|npy ihnpy |npx i = CHn α y − exp(ipx x) exp(−ieBxy),
−∞ Mω
(4.2.27)
which completes the proof.

Since all classical orbits are closed and the center of the cyclotron orbit plays the
role of a Runge-Lenz vector, it is natural to ask whether the degeneracy is caused by
an accidental symmetry. Of course, since the Runge-Lenz vector plays a dual role
and is also generating translations (up to gauge transformations), in this case the
“accidental” symmetry would just be translation invariance. Indeed, in complete
analogy to the classical case, it is easy to convince oneself that [H, Rx ] = [H, Ry ] =
4.2. PARTICLE IN THE INFINITE VOLUME 71

[H, L] = 0, with the Runge-Lenz vector and the angular momentum operator given
by

Py i∂y Px i∂x
Rx = − = , Ry = =y− ,
eB eB  eB  eB 
eBx eBy
L = x −i∂y + − y −i∂x + . (4.2.28)
2 2

As in the classical case, the radius of the cyclotron orbit squared is given by
 2
2 2 2 i∂y ∂x2 2H
r = (x − Rx ) + (y − Ry ) = x− − 2 2
= , (4.2.29)
eB eB Mω 2

and is thus again a conserved quantity. In particular, we can express the Hamiltonian
as
1
H = Mω 2 r 2 . (4.2.30)
2

Remarkably, although the two coordinates Rx and Ry of the center of the cy-
clotron circle are not simultaneously measurable, its radius r has a definite value in
an energy eigenstate. As it should, under spatial rotations the Runge-Lenz vector
(Rx , Ry ) indeed transforms as a vector, i.e.

[L, Rx ] = iRy , [L, Ry ] = −iRx . (4.2.31)

These relations suggest to introduce

R± = Rx ± iRy , (4.2.32)

which implies
[L, R± ] = ±R± . (4.2.33)

Hence, R+ and R− act as raising and lowering operators of angular momentum.

Still, it is important to note that Rx , Ry , and L do not form an SU(2) algebra. This
follows because, in analogy to the classical case

i
[Rx , Ry ] = , (4.2.34)
eB
i.e. Rx and Ry are generators of a Heisenberg algebra. From the above equation we
easily conclude that
2
[R+ , R− ] = . (4.2.35)
eB
72 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

4.2.4 Creation and Annihilation Operators

The particle in the magnetic field leads to the (infinitely degenerate) spectrum of a 1-
dimensional harmonic oscillator. This suggests that one can construct corresponding
creation and annihilation operators such that

 
†1
H =ω a a+ , [a, a† ] = 1. (4.2.36)
2
Interestingly, the creation and annihilation operators are closely related to the
Runge-Lenz vector, i.e. the vector that points to the center of the classical cyclotron
orbit. Since we have seen that
1 1  
H = Mω 2 r 2 = Mω 2 (x − Rx )2 + (y − Ry )2 , (4.2.37)
2 2
one is led to identify
r r
Mω † Mω
a= [x − Rx − i(y − Ry )] , a = [x − Rx + i(y − Ry )] , (4.2.38)
2 2
which indeed have the desired properties. One also finds that
[L, a] = −a, [L, a† ] = a† , (4.2.39)
which implies that a† and a also raise and lower the angular momentum. Interest-
ingly, we have seen before that
2 2
[L, R± ] = ±R± , [R+ , R− ] = = . (4.2.40)
eB Mω
Therefore, R+ and R− also act as raising and lowering operators of the angular
momentum. We can identify another set of creation and annihilation operators,
which are r r
Mω Mω
b= R+ , b† = R− , (4.2.41)
2 2
and obey
[L, b] = b, [L, b† ] = −b† . (4.2.42)
As a result, b raises and b† lowers the angular momentum by one unit. Introducing
r
2

z = x + iy = a† + b , (4.2.43)
Mω
it is straightforward to derive the commutation relations
[a, b] = [a† , b] = [a, b† ] = [a† , b† ] = 0, [b, b† ] = 1. (4.2.44)
Like in the case of a 2-dimensional harmonic oscillator, the particle in a magnetic
field is described by two sets of commuting creation and annihilation operators.
However, in contrast to the 2-dimensional harmonic oscillator, the Hamiltonian of
the particle in a magnetic field contains only a† a, but not b† b.
4.2. PARTICLE IN THE INFINITE VOLUME 73

4.2.5 Alternative Representation of the Hamiltonian

As we showed in Eq.(2.3.20), the Hamiltonian can be expressed as

1

H = Mω 2 Rx2 + Ry2 + ωL. (4.2.45)
2
Using the above equation as well as Eq.(4.2.41) gives
 
† 1
H = ω b b + + L = H0 + ωL. (4.2.46)
2
Here we have introduced the Hamiltonian of an ordinary 1-dimensional harmonic
oscillator  
† 1
H0 = ω b b + , (4.2.47)
2
and the angular momentum operator has been identified as

L = a† a − b† b. (4.2.48)

Interestingly, the creation and annihilation operators b† and b commute with the
total energy H because they raise (lower) H0 by ω, while they lower (raise) L by 1,
such that indeed

[H, b] = [H0 , b] + ω[L, b] = 0, [H, b† ] = [H0 , b† ] + ω[L, b† ] = 0. (4.2.49)

4.2.6 Energy Spectrum and Energy Eigenstates

In section (2.3.2) we discussed the explicit solution of the Schrödinger equation for
this problem, as well as an alternative method for finding the energy levels and
degeneracies from symmetry. Another interesting approach can be is also based also
on symmetry, and uses the fact that the algebraic structure of the problem (but not
the exact form of the Hamiltonian) is the same as for the 2-dimensional harmonic
oscillator. First of all, we construct a state |00i that is annihilated by both a and b,
i.e.
a|00i = b|00i = 0. (4.2.50)
Then we define states
n †
n′
′ a† b
|nn i = √ √ |00i, (4.2.51)
n! n′ !
which are eigenstates of the total energy
 
′ 1
H|nn i = ω n + |nn′ i, (4.2.52)
2
74 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

as well as of the angular momentum

L|nn′ i = (n − n′ )|nn′ i = m|nn′ i. (4.2.53)

It should be noted that the quantum number n ∈ {0, 1, 2, ...} (which determines
the energy) is non-negative, while the quantum number m = n − n′ ∈ Z (which
determines the angular momentum) is an arbitrary integer. The infinite degeneracy
of the Landau levels is now obvious because states with the same n but different
values of n′ have the same energy. One may wonder why in subsection (4.2.3) we
found an infinite degeneracy labeled by the continuous momentum py and now we
only find a countable variety of degenerate states (labeled by the integer m). This
apparent discrepancy is due to the implicit consideration of two different Hilbert
spaces. While the states in the discrete variety labeled by m are normalizable in the
usual sense, the continuous variety of plane wave states labeled by py is normalized
to δ-functions and thus belongs to an extended Hilbert space.

4.2.7 Coherent States

Coherent states are well-known from the harmonic oscillator, and have also been
constructed for the Landau level problem [34]. As usual, the coherent states are
constructed as eigenstates of the annihilation operators, i.e.

a|λλ′ i = λ|λλ′i, b|λλ′ i = λ′ |λλ′i, λ, λ′ ∈ C. (4.2.54)

Some expectation values in the coherent state |λλ′ i are given by

r r
2 ′ 2
hRx i = Reλ , ∆Rx = ,
Mω Mω
r r
2 2
hRy i = Imλ′ , ∆Ry = ,
Mω Mω
r
2 1
hx − Rx i = Reλ, ∆(x − Rx ) = √ ,
Mω 2Mω
r
2 1
hy − Ry i = − Imλ, ∆(y − Ry ) = √ ,
Mω 2Mω
r
√ Mω
hMvx i = hpx + eAx i = 2Mω Imλ, ∆(Mvx ) = ,
2
r
√ Mω
hMvy i = hpy + eAy i = 2Mω Reλ, ∆(Mvy ) = ,
  2
1
hHi = ω |λ|2 + , ∆H = ω|λ|. (4.2.55)
2
4.2. PARTICLE IN THE INFINITE VOLUME 75
p
Here ∆O = hO 2i − hOi2 describes the quantum uncertainty. In all cases ∆O/hOi
is proportional to 1/|λ| or 1/|λ′ |, which implies that the relative uncertainty goes
to zero in the classical limit. Just as for the ordinary harmonic oscillator, the time-
dependent Schrödinger equation i∂t |Ψ(t)i = H|Ψ(t)i with an initial coherent state
|Ψ(0)i = |λ(0)λ′ i is solved by
 
iωt
|Ψ(t)i = exp − |λ(t)λ′ i, λ(t) = λ(0) exp(−iωt), (4.2.56)
2

i.e. the state remains coherent during its time-evolution. In particular, this implies

|λ| √
hx − Rx i(t) = √ cos(ωt), hMvx i(t) = − 2Mω |λ| sin(ωt),
2Mω
|λ| √
hy − Ry i(t) = √ sin(ωt), hMvy i(t) = 2Mω |λ| cos(ωt). (4.2.57)
2Mω

Hence, the coherent state represents a wave packet moving around a circular cy-
clotron orbit just like a classical particle. This can be realized by solving the eigen-
value problem Eqs.(4.2.54). We denote h~x|λ(t)λ′i = Ψλ(t)λ′ (~x). Substituting for a
and b from Eq.(4.2.38) and Eq.(4.2.41) in Eqs.(4.2.54), then we will get the following
first order partial differential equations
r  
Mω i 1
∂y + ∂x + iy Ψλ(t)λ′ (~x) = λ′ Ψλ(t)λ′ (~x),
2 Mω Mω
r  
Mω 1 i
∂x − ∂y + x Ψλ(t)λ′ (~x) = λΨλ(t)λ′ (~x). (4.2.58)
2 Mω Mω

A solution of the above two differential equations is

r 
Mω ′
Ψλ(t)λ′ (~x) = A(t) exp (x − iy)λ + (x + iy)λ(t)
2
 !  ω 
Mω 2
− x + y 2 + 2ixy exp −i t (4.2.59)
4 2

The normalization constant is particularly important in this case because it is time-

dependent. From the normalization condition
Z ∞ Z ∞
dxdyΨλ(t)λ′ (~x)Ψ∗λ(t)λ′ (~x) = 1, (4.2.60)
−∞ −∞
76 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

we obtain the value of A(t). Accordingly the coherent state takes the following form
r r  
Mω Mω ′
Ψλ(t)λ′ (~x) = exp (x − iy)λ + (x + iy)λ(t)
2π 2
 !
Mω 2
− x + y 2 + ixy + (Reλ′ − Reλ(t))2 − (Imλ′ − Imλ(t))2
4
ω
× exp(−i t). (4.2.61)
2

From the above equation we get the probability density given by

 
2 Mω 2
|h~x|Ψ(t)i| = A exp − (x − hRx i − d cos(ωt − α))
2
 
Mω 2
× exp − (y − hRy i − d sin(ωt − α)) ,
2
r
2
d= |λ|. (4.2.62)
Mω

As it is obvious from the above equation, the coherent state is, in fact, a Gaussian
wave packet moving around a circular orbit.

4.3 Particle on a Torus

In this section we put the problem in a finite periodic volume. This breaks rotation
invariance, but leaves translation invariance intact (at least at the classical level),
and leads to an energy spectrum with finite degeneracy. In order to clarify some sub-
tle symmetry properties, we investigate issues of Hermiticity versus self-adjointness
of various operators.

4.3.1 Constant Magnetic Field on a Torus

In this subsection we impose a torus boundary condition over a rectangular region of

size Lx × Ly . As a result of this, there is a quantization condition for the magnetic
flux. Since the magnetic field is constant, it obviously is periodic. On the other
hand, the vector potential of the infinite volume theory Ax (x, y) = 0, Ay (x, y) = Bx
4.3. PARTICLE ON A TORUS 77

obeys the conditions

Ax (x + Lx , y) = Ax (x, y),
Ay (x + Lx , y) = Ay (x, y) + BLx = Ay (x, y) + ∂y (BLx y),
Ax (x, y + Ly ) = Ax (x, y),
Ay (x, y + Ly ) = Ay (x, y). (4.3.1)

As a gauge-dependent quantity, the vector potential is periodic only up to gauge

transformations, i.e.

Ai (x + Lx , y) = Ai (x, y) − ∂i ϕx (y), Ai (x, y + Ly ) = Ai (x, y) − ∂i ϕy (x). (4.3.2)

The gauge transformations ϕx (y) and ϕy (x) are transition functions in a fiber bundle
which specify the boundary condition. In our case the transition functions are given
by
θx θy
ϕx (y) = − BLx y, ϕy (x) = . (4.3.3)
e e
Besides the field strength, gauge theories on a periodic volume possess additional
gauge invariant quantities — the so-called Polyakov loops — which arise due to the
non-trivial holonomies of the torus. For an Abelian gauge theory the Polyakov loops
are defined as
Z Lx Z Ly
Φx (y) = dx Ax (x, y) − ϕx (y), Φy (x) = dy Ay (x, y) − ϕy (x). (4.3.4)
0 0

In our case, they are given by

θx θy
Φx (y) = BLx y − , Φy (x) = BLy x − . (4.3.5)
e e
In order to respect gauge invariance of the theory on the torus, under shifts the
wave function must also be gauge transformed accordingly

Ψ(x + Lx , y) = exp (ieϕx (y)) Ψ(x, y) = exp (iθx − ieBLx y) Ψ(x, y),
Ψ(x, y + Ly ) = exp (ieϕy (x)) Ψ(x, y) = exp (iθy ) Ψ(x, y). (4.3.6)

The angles θx and θy parameterize a family of self-adjoint extensions of the Hamil-

tonian on the torus. Applying the boundary conditions from above in two different
orders one obtains

Ψ(x + Lx , y + Ly ) = exp (iθx − ieBLx (y + Ly )) Ψ(x, y + Ly )

= exp (iθx + iθy − ieBLx (y + Ly )) Ψ(x, y),
Ψ(x + Lx , y + Ly ) = exp (iθy ) Ψ(x + Lx , y)
= exp (iθx + iθy − ieBLx y) Ψ(x, y). (4.3.7)
78 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

Hence, consistency of the boundary condition requires

2πnΦ
exp (−ieBLx Ly ) = 1 ⇒ B = , nΦ ∈ Z. (4.3.8)
eLx Ly
The total magnetic flux through the torus
2πnΦ
Φ = BLx Ly = , (4.3.9)
e
is hence quantized in integer units of the elementary magnetic flux quantum 2π/e.
Interestingly, the spectrum of a charged particle in a constant magnetic field is
discrete (but infinitely degenerate) already in the infinite volume. As we will see,
in the finite periodic volume it has only a finite |nΦ |-fold degeneracy determined by
the number of flux quanta. A quantum mechanical charged particle is sensitive to
the complex phases defined by the Polyakov loops
exp(ieΦx (y)) = exp(ieBLx y − iθx ), exp(ieΦy (x)) = exp(ieBLy x − iθy ), (4.3.10)
which are measurable in Aharonov-Bohm type experiments. Remarkably, the Polya-
kov loops explicitly break the translation invariance of the torus at the quantum
level. This is reminiscent of the quantum mechanical breaking of CP invariance due
to the θ-vacuum angle in non-Abelian gauge theories. The complex phases from
above are invariant under shifts by integer multiples of
2π Lx 2π Ly
ax = = , ay = = , (4.3.11)
eBLy nΦ eBLx nΦ
in the x- and y-directions, respectively. Hence, at the quantum level the continuous
translation group of the torus is reduced to a discrete subgroup which plays the
role of the accidental symmetry group. In this thesis, we treat the gauge field
as a classical background field, while only the charged particle is treated quantum
mechanically. It is interesting to note that, once the gauge field is also quantized, the
transition functions ϕx (y) and ϕy (x) become fluctuating physical degrees of freedom
of the gauge field. Still, as a consequence of
Ai (x + Lx , y + Ly ) = Ai (x, y + Ly ) − ∂i ϕx (y + Ly )
= Ai (x, y) − ∂i ϕx (y + Ly ) − ∂i ϕy (x),
Ai (x + Lx , y + Ly ) = Ai (x + Lx , y) − ∂i ϕy (x + Lx )
= Ai (x, y) − ∂i ϕy (x + Lx ) − ∂i ϕx (y), (4.3.12)
and of
Ψ(x + Lx , y + Ly ) = exp(ieϕx (y + Ly ))Ψ(x, y + Ly )
= exp(ieϕx (y + Ly ) + ieϕy (x))Ψ(x, y),
Ψ(x + Lx , y + Ly ) = exp(ieϕy (x + Lx ))Ψ(x + Ly , y)
= exp(ieϕy (x + Lx ) + ieϕx (y))Ψ(x, y), (4.3.13)
4.3. PARTICLE ON A TORUS 79

the transition functions must obey the cocycle consistency condition

2πnΦ
ϕy (x + Lx ) + ϕx (y) − ϕx (y + Ly ) − ϕy (x) = . (4.3.14)
e
In this case, the magnetic flux nΦ specifies a super-selection sector of the theory.
Analogous to the Z(N)d center symmetry of non-Abelian SU(N) gauge theories
on a d-dimensional torus [35], Abelian gauge theories coupled to charged matter
have a global Zd center symmetry. The self-adjoint extension parameters θx and θy
then turn into conserved quantities (analogous to Bloch momenta) of the global Z2
symmetry on the 2-dimensional torus. In this sense, θx and θy are analogous to the θ-
vacuum angle of non-Abelian gauge theories, which also distinguishes different super-
selection sectors of the theory. The θ-vacuum angle is a quantum mechanical source
of explicit CP violation. At the classical level, on the other hand, CP invariance
remains intact because θ does not affect the classical equations of motion. Similarly,
for a charged particle on the torus the angles θx and θy characterize the explicit
breaking of continuous translation invariance down to a discrete subgroup. Just
like CP invariance for a non-Abelian gauge theory, for a charged particle on the
torus the full continuous translation symmetry remains intact at the classical level,
because θx and θy do not appear in the classical equations of motion. In this thesis,
we treat the charged particle as a test charge which does not surround itself with its
own Coulomb field. This would change, once one would derive the charged particle
from its own quantum field. For example, if one considers full-fledged QED, a single
electron cannot even exist on the torus because the Coulomb field that surrounds
it is incompatible with periodic boundary conditions. Indeed, as a consequence of
the Gauss law, the total charge on a torus always vanishes. To cure this problem,
one could compensate the charge of the electron by a classical background charge
homogeneously spread out over the torus. In the present calculation this is not
necessary, because the charged particle is treated as a test charge without its own
surrounding Coulomb field.

4.3.2 Discrete Magnetic Translation Group

As we have seen, in order to respect gauge invariance, on the torus the wave function
must obey Eq.(4.3.6), which can be re-expressed as
 
2πinΦ y
Ψ(x + Lx , y) = exp iθx − Ψ(x, y), Ψ(x, y + Ly ) = exp(iθy )Ψ(x, y).
Ly
(4.3.15)
It is interesting to note that a factorization ansatz for the wave function as in
Eq.(4.2.16) is inconsistent with the boundary condition. Let us consider the unitary
shift operator generating translations by a distance ay in the y-direction
Ty = exp(iay Py ), (4.3.16)
80 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

which acts as
Ty Ψ(x, y) = Ψ(x, y + ay ). (4.3.17)

It is clear that Ty commutes with the Hamiltonian because Py does. Indeed, the
shifted wave function does obey the boundary condition Eq.(4.3.15), i.e.
 
2πinΦ (y + ay )
Ty Ψ(x + Lx , y) = Ψ(x + Lx , y + ay ) = exp iθx − Ψ(x, y + ay )
Ly
 
2πinΦ y
= exp iθx − Ty Ψ(x, y), (4.3.18)
Ly

which is the case only because

 
Ly 2πinΦ ay
ay = ⇒ exp − = 1. (4.3.19)
nΦ Ly

Furthermore, we also have

Ty Ψ(x, y + Ly ) = Ψ(x, y + ay + Ly ) = exp(iθx )Ψ(x, y + ay )

= exp(iθx )Ty Ψ(x, y). (4.3.20)

Hence, as we argued before, the translations in the y-direction are reduced to the
discrete group Z(nΦ ). In particular, all translations compatible with the boundary
n
conditions Ty y can be expressed as the ny -th power of the elementary translation
Ty . According to Eq.(4.2.8), Py = −eBRx , such that
     
iLy Py ieBLy Rx 2πiRx
Ty = exp = exp − = exp − . (4.3.21)
nΦ nΦ Lx

Since on the torus the Runge-Lenz vector component Rx , which determines the
x-coordinate of the center of the cyclotron orbit, is defined only modulo Lx , it
is indeed natural to consider the above translation operator. In fact, although it
formally commutes with the Hamiltonian, the operator Rx itself is no longer self-
adjoint in the Hilbert space of wave functions on the torus. The operator Ty , on
the other hand, does act as a unitary operator in the Hilbert space. Similarly, the
operator is given by
   
2πinΦ ax Ry 2πiRy
Tx = exp(iax Px ) = exp (iax eBRy ) = exp = exp .
Lx Ly Ly
(4.3.22)
As a consequence of the commutation relation [Rx , Ry ] = i/eB, one obtains
 
2πi
Ty Tx = exp Tx Ty . (4.3.23)
nΦ
4.3. PARTICLE ON A TORUS 81

This implies that  

2πiy Lx
Tx Ψ(x, y) = exp Ψ(x + , y), (4.3.24)
Ly nΦ
i.e., up to a periodic gauge transformation exp(2πiy/Ly ), Tx translates the wave
function by a distance Lx /nΦ . Remarkably, although at the classical level the torus
has two continuous translation symmetries, the corresponding infinitesimal gener-
ators Px and Py are not self-adjoint in the Hilbert space of wave functions on the
torus. Only the finite translations Tx and Ty are represented by unitary operators,
which, however, do not commute with each other. The two operators Tx and Ty
generate a discrete translation group G consisting of the elements
 
2πim
g(nx , ny , m) = exp Tyny Txnx ,
nΦ
nx , ny , m ∈ {0, 1, 2, ..., nΦ − 1}. (4.3.25)

The group multiplication rule takes the form

g(nx , ny , m)g(n′x , n′y , m′ ) = g(nx + n′x , ny + n′y , m + m′ − nx n′y ), (4.3.26)

with all summations being understood modulo nΦ . Obviously, the unit-element is

represented by
1 = g(0, 0, 0), (4.3.27)
while the elements  
2πim
zm = g(0, 0, m) = exp , (4.3.28)
nΦ
form the cyclic Abelian subgroup Z(nΦ ) ⊂ G. The inverse of a general group element
g(nx , ny , m) is given by

g(nx , ny , m)−1 = g(−nx , −ny , −m − nx ny ), (4.3.29)

because

g(nx , ny , m)g(−nx , −ny , −m − nx ny ) = g(0, 0, −nx ny + nx ny ) = g(0, 0, 0) = 1.

(4.3.30)
It is interesting to consider the conjugacy class of a group element g(nx , ny , m) which
consists of the elements

g(n′x , n′y , m′ )g(nx , ny , m)g(n′x , n′y , m′ )−1 =

g(n′x + nx , n′y + ny , m′ + m − n′x ny )g(−n′x , −n′y , −m′ − n′x n′y ) =
g(nx , ny , m − n′x (ny + n′y ) + (n′x + nx )n′y ) = g(nx , ny , m + nx n′y − n′x ny ).
(4.3.31)
82 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

Hence, as one would expect, the elements g(0, 0, m) = zm ∈ Z(nΦ ) are conjugate
only to themselves and thus form nΦ single-element conjugacy classes. In the case of
odd nΦ , the nΦ elements zm g(nx , ny , 0) with (nx , ny ) 6= (0, 0) form n2Φ − 1 additional
conjugacy classes. When nΦ is even, this is still true as long as not both nx and
ny are even. On the other hand, when nx , ny , and nΦ are all even, the elements
zm g(nx , ny , 0) with m even and with m odd form two distinct conjugacy classes,
each consisting of nΦ /2 elements. Obviously, multiplication by a phase zm is just a
global gauge transformation and thus leaves the physical state invariant. Hence, the
conjugacy classes correspond to gauge equivalence classes. The elements g(0, 0, m) =
zm commute with all other elements and thus form the center Z(nΦ ) of the group
G. Since the individual elements of the center form separate conjugacy classes,
the center is a normal subgroup and can hence be factored out. The center itself
represents global phase transformations of the wave function, and hence factoring it
out corresponds to identifying gauge equivalence classes. Physically speaking, the
quotient space G/Z(nΦ ) = Z(nΦ ) ⊗ Z(nΦ ) corresponds to discrete translations up to
gauge transformations. It should be pointed out that G is not simply given by the
direct product Z(nΦ ) ⊗ Z(nΦ ) ⊗ Z(nΦ ). In fact, the quotient space Z(nΦ ) ⊗ Z(nΦ ) is
not a subgroup of G, and hence G is also not the semi-direct product Z(nΦ )⊗Z(nΦ )×
Z(nΦ ). All we can say is that G is a particular central extension of Z(nΦ ) ⊗ Z(nΦ )
by the center subgroup Z(nΦ ).

4.3.3 Spectrum and Degeneracy on the Torus

It should be pointed out that on the torus the Hamiltonian is identically the same as
in the infinite volume. It now just acts on the restricted set of wave functions obeying
the boundary condition Eq.(4.3.15). In particular, the finite volume wave functions
are appropriate linear combinations of the infinitely many degenerate states of a
given Landau level. As a result, the energy spectrum remains unchanged, but the
degeneracy is substantially reduced. Let us use the fact that Ty commutes with
the Hamiltonian to construct simultaneous eigenstates of both H and Ty . Since
TynΦ = exp(iθy )1, the eigenvalues of Ty are given by exp(i(2πily + θy )/nφ ) with
ly ∈ {0, 1, ..., nΦ − 1}, while the eigenvalues of H are still given by En = ω(n + 12 ).
Hence, we can construct simultaneous eigenstates |nly i such that

   
1 2πily + iθy
H|nly i = ω n + |nly i, Ty |nly i = exp |nly i. (4.3.32)
2 nφ

The states |nly i are the finite-volume analog of the states |npy i of Eq.(4.2.19) with
py = (2πly + θy )/Ly . In coordinate representation these states are given by the wave
4.3. PARTICLE ON A TORUS 83

functions
X    
θy Lx
h~x|nly i = ψn x + nΦ nx + ly +
nx ∈Z
2π nΦ
   
2πiy θy
× exp nΦ nx + ly + − iθx nx . (4.3.33)
Ly 2π

As a consequence of Eq.(4.3.23) one obtains

   
2πi 2πi(ly + 1) + iθy
Ty Tx |nly i = exp Tx Ty |nly i = exp Tx |nly i, (4.3.34)
nφ nφ

from which we conclude that

Tx |nly i = |n(ly + 1)i. (4.3.35)

Since [Tx , H] = 0, the nΦ states |nly i with ly ∈ 0, 1, ..., nΦ − 1 thus form an

irreducible representation of the magnetic translation group. Using nΦ = 4 as a
concrete example, a matrix representation of the two generators of G is given by
   
0 0 0 1 1 0 0 0  
 1 0 0 0   0 i 0 
Tx =   , Ty =  0  exp iθy . (4.3.36)
 0 1 0 0   0 0 −1 0  4
0 0 1 0 0 0 0 −i

Similarly, one can construct simultaneous eigenstates |nlx i of H and Tx

   
1 2πilx + iθx
H|nlx i = ω n + |nlx i, Tx |nlx i = exp |nlx i. (4.3.37)
2 nφ

The states |nlx i are the finite-volume analog of the states |npx i of Eq.(4.2.21) with
px = (2πlx + θx )/Lx . In coordinate representation these states are given by the wave
functions
X    
θx Ly
h~x|nlx i = ψn y − nΦ ny + lx +
ny ∈Z
2π nΦ
   
2πix θx nΦ y
× exp nΦ ny + lx + − + iθy ny . (4.3.38)
Lx 2π Ly

It is worth noting that

Ty |nlx i = |n(lx − 1)i. (4.3.39)
Just as in the infinite volume, it is straightforward to show that the two sets of
eigenstates h~x|nly i and h~x|nlx i span the same subspace of the Hilbert space.
84 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

4.3.4 Coherent States on the Torus

The coherent state should also satisfy the boundary condition in Eq.(4.3.13). The
proposed form for the coherent state is
∞
X ∞
X
e λ(t)λ′ (x, y) = C
Ψ Ψλ(t)λ′ (x + nx Lx , y + ny Ly )
nx =−∞ ny =−∞
 
2πiy
× exp nΦ nx exp(−iθx nx ) exp(−iθy ny ), (4.3.40)
Ly
where C is a normalization constant that will be obtained in the next subsection.

Normalization of the Wave Function

The normalization condition can be obtained from the following equation

Z Lx Z Ly
dxdy Ψ e ∗ ′ (x, y) = 1.
e λ(t)λ′ (x, y)Ψ (4.3.41)
λ(t)λ
0 0

The coherent state for motion on the torus given by Eq.(4.3.40) can be written in
terms of the coherent state in the infinite volume with the help of the shift operators
(Tx )nΦ and (Ty )nΦ , and we have
∞
X ∞
X
f′ i = C
|λλ e−iθx nx e−iθy ny Teyny Texnx |λλ′ i, (4.3.42)
nx =−∞ ny =−∞

where Tex = (Tx )nΦ , and Tey = (Ty )nΦ .

In order to find the normalization constant, we write

∞
X ∞
X ∞
X ∞
X ′ ′
hλf f′i = |C|2
′ λ|λλ hλ′ λ|(Tex† )nx (Tey† )ny Teyny Texnx |λλ′ i
nx =−∞ ny =−∞ n′x =−∞ n′y =−∞
′ ′
× e−iθx (nx −nx ) e−iθy (ny −ny ) , (4.3.43)

where Z ∞ Z ∞
hλf f′ i =
′ λ|λλ dxdy Ψ e ∗ ′ (x, y).
e λ(t)λ′ (x, y)Ψ
λ(t)λ (4.3.44)
−∞ −∞

In the above equation, the intervals of the double integral in Eq.(4.3.41) were ex-
tended to (−∞, ∞). This will prove to be useful later in calculating the normal-
ization constant C from Eq.(4.3.43). This term diverge, but we should not worry
4.3. PARTICLE ON A TORUS 85

about that because it can be cancelled from both sides of Eq.(4.3.43) as we will see
later. Now using Eq.(4.3.41), Eq.(4.3.44) can be written as
X∞ ∞
X Z Lx Z Ly
hλf f′ i =
′ λ|λλ dxdy
nx =−∞ ny =−∞ 0 0

e λ(t)λ′ (x + nx Lx , y + ny Ly )Ψ
× Ψ e ∗ ′ (x + nx Lx , y + ny Ly )
λ(t)λ
∞
X X ∞
= 1. (4.3.45)
nx =−∞ ny =−∞

Using the definition of the shift operators in Eq.(4.3.22), and Eq.(4.3.21) it is easy
to show
Tex† = Tex−1 , Tey† = Tey−1 . (4.3.46)
Substituting for Tex−1 and Tey−1 from Eq.(4.3.46) into Eq.(4.3.43) then, the result is
am expression inside the summations that depends only on nx − n′x and ny − n′y .
This leads to
X X ∞
X ∞
X
hλf f′i = |C|2
′ λ|λλ hλ′ λ|Teymy Texmx |λλ′ i
nx −n′x =mx ny −n′y =my mx =−∞ my =−∞

× e−iθx mx e−iθy my , (4.3.47)

From the above equation and Eq.(4.3.45) we get

∞
X ∞
X
|C| −2
= hλ′ λ|Teymy Texmx |λλ′ ie−iθx mx e−iθy my , (4.3.48)
mx =−∞ my =−∞

In the following we will obtain the value of C. From the definition of the shift
operators in Eq.(4.3.22) and Eq.(4.3.21), and using Eqs.(4.2.41), we can write
Ry
Tex = exp(2πinΦ ) = exp(cx (b† − b)),
Ly
Rx
Tey = exp(−2πinΦ ) = exp(−icy (b† + b)), (4.3.49)
Ly
where r r
2 π 2 π
cx = nΦ , cy = nΦ . (4.3.50)
Mω Ly Mω Lx
Substituting Eqs.(4.3.49) in Eq.(4.3.48) gives
∞
X ∞
X
−2 † −b) † +b)
|C| = hλ′ λ|e−mx cx (b e−imy cy (b |λλ′ ie−iθx mx e−iθy my . (4.3.51)
mx =−∞ my =−∞
86 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

A well-known theorem [34] can be applied to find the value of C. If A and B are
two operators such that [A, B] commutes with both A and B then
1
eA+B = e− 2 [A,B] eA eB , eA eB = e[A,B] eB eA . (4.3.52)
Using the identities of Eqs.(4.3.52) one can shift all the exponentials with b† to the
left, and exponentials with b to the right. After some calculations, and using the fact
that the state |λλ′ i is an eigenstate of the operator b with eigenvalue λ′ , Eq.(4.3.51)
can be written as
X 1
|C|−2 = hλ′ λ|λλ′i exp(− c2x m2x + i(−2cx Imλ′ − θx )mx )
m ,m
2
x y

1
× exp(− c2y m2y − i(2cy Reλ′ + θy )my ) exp(icx cy mx my ). (4.3.53)
2
The summation over mx converges to a special function called the ϑ3 function. The
elliptic function plays an important role in the following discussions, therefore it is
useful to me define these functions and discuss some of their properties. It is defied
by the following equation
m=∞
X
√
ϑ2 (z, q) = q
4
q m(m+1) e(2m+1)iz , (4.3.54)
m=−∞
m=∞
X
m2 2miz
ϑ3 (z, q) = q e , (4.3.55)
m=−∞
m=∞
X 2
ϑ4 (z, q) = (−1)m q m e2miz ,
m=−∞
q = iπτ, |q| < 1. (4.3.56)
Accordingly, Eq.(4.3.55) can be written as
X 1 
−2 ′ − 21 c2x
|C| = ϑ3 (cx cy my − 2cx Imλ − θx ), e
my
2
 
1 2 2 ′
× exp − cy my − 2icy my Reλ − iθy my . (4.3.57)
2
The following properties of the ϑ3 function are useful to find the value of the above
summation
ϑ3 (z + nπ, q) = ϑ3 (z, q),
(
 π 
ϑ3 (z, q), for even k,
ϑ3 z + (2n + 1) k, q =
2 ϑ4 (z, q), for odd k,
q = iπτ, |q| < 1, n ∈ Z. (4.3.58)
4.3. PARTICLE ON A TORUS 87

It is easy to see from Eq.(4.3.50) that

cx cy = πnΦ . (4.3.59)

From the above equation and Eqs.(4.3.57), for even nΦ one obtains
 
−2 1 ′ − 21 c2x
|C| = ϑ3 −cx Imλ − θx , e
2
X

× exp −2c2y m2y − 4icy my Reλ′ − 2iθy my
my

X  !
1 2
+ exp − cy (2my + 1)2 − 2icy (2my + 1)Reλ′ − iθy (2my + 1) .
my
2
(4.3.60)

From the above equation and Eqs.(4.3.58), for odd nΦ one finds
 X
−2 ′ 1 − 21 c2x


|C| = ϑ3 −cx Imλ − θx , e exp −2c2y m2y − 4icy my Reλ′ − 2iθy my
2 my
 
1 1 2
+ ϑ4 −cx Imλ′ − θx , e− 2 cx
2
X  
1 2 2 ′
× exp − cy (2my + 1) − 2icy (2my + 1)Reλ − iθy (2my + 1) .
m
2
y

(4.3.61)

Using the definition of ϑ3 in Eq.(4.3.55) we can write Eq.(4.3.60) for even nΦ

   
−2 ′ 1 − 12 c2x 2
|C| = ϑ3 −cx Imλ − θx , e ϑ3 −2cy Reλ′ − θy , e−2cy
2
!
1  2

+ exp(−2icy Reλ′ − iθy − c2y )ϑ3 ic2y − 2cy Reλ′ − θy , e−2cy .
2
(4.3.62)

Using the following identity [31]

 
ln(q) 2 −n− 1
ϑ3 z + (2n + 1)i ,q = q −n 4 e(2n+1)iz ϑ2 (z, q),
2
q = iπτ, |q| < 1, (4.3.63)
88 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

for n = −1, Eq.(4.3.62) can be written as

   
−2 ′ 1 − 21 c2x ′ −2c2y
|C| = ϑ3 −cx Imλ − θx , e ϑ3 −2cy Reλ − θy , e
2
!
 
2 ′ −2c2y
+ exp(−cy )ϑ2 −2cy Reλ − θy , e . (4.3.64)

Finally, for even nΦ the normalization constant can be written as

   
iα ′ 1 − 12 c2x ′ −2c2y
C = e ϑ3 −cx Imλ − θx , e ϑ3 −2cy Reλ − θy , e
2
  !− 21
2
+ exp(−c2y )ϑ2 −2cy Reλ′ − θy , e−2cy , (4.3.65)

where α is a constant arbitrary phase. For odd nΦ the result is

   
iα 1
′ − 21 c2x 2
C = e ϑ3 −cx Imλ − θx , e ϑ3 −2cy Reλ′ − θy , e−2cy
2
   ! 1
1 1 2 2
 −2
+ exp(−c2y )ϑ4 −cx Imλ′ − θx , e− 2 cx ϑ2 −2cy Reλ′ − θy , e−2cy .
2
(4.3.66)

As it is obvious from Eq.(4.3.65) and Eq.(4.3.66), the first argument of ϑi , i =

2, 3, 4 either depend on cx Imλ′ shifted by θx , or on 2cy Reλ′ shifted by θy . The
theta functions are periodic functions with ϑi (z + nπ, q) = ϑi (z, q), i = 3, 4, and
ϑ2 (z + nπ, q) = (−1)n ϑ2 (z, q), with n ∈ Z. This means that the normalization
constant is a π-periodic function. That is because when we shift cx Imλ′ or cy Reλ′
or both by π, then C maintains the same value.

4.3.5 Calculating the Expectation Value of Txl and Tyl

The expectation value for the shift operator Tx raised to the power l for a coherent
state can be evaluated from the following equation
∞
X ∞
X ∞
X ∞
X ′ ′
hλf f′ i =
′ λ|T l |λλ
k e−iθx (nx −nx ) e−iθy (ny −ny )
nx =−∞ ny =−∞ n′x =−∞ n′y =−∞
′ ′
× hλ′ λ|(Tex† )nx (Tey† )ny Tkl Teyny Texnx |λλ′i, (4.3.67)
4.3. PARTICLE ON A TORUS 89

where k = x, y. The shift operators raised to the power l can be written as

Ry
Txl = exp(2inΦ π ) = exp(dx (b† − b)),
Ly
Rx
Tyl = exp(−2inΦ π ) = exp(−idy (b† + b)), (4.3.68)
Ly

where,
r r
2 π 2 π
dx = l, dy = l (4.3.69)
Mω Ly Mω Lx

We apply the same technique that was used for calculating the normalization con-
stant by shifting all the exponentials with b† to the left, and exponentials with b to
the right. Accordingly, for odd l the operator Txl takes the following form

′ 1 2
X
hλf f i = |C|2 hλ′ λ|λλ′ ie− 2 dx exp(−2idx Imλ′ )
′ λ|T l |λλ
x (−1)my
mx ,my
1
× exp(−dx cx mx − c2x m2x + i(cx cy my − 2cx Imλ′ − θx )mx )
2
1 2 2
× exp(− cy my − i(2cy Reλ′ + θy )my ), (4.3.70)
2

while for even l

hλf f ′ i = |C|2hλ′ λ|λλ′ ie− 21 d2x exp(−2idx Imλ′ )

′ λ|T l |λλ
x
X 1
× exp(−dx cx mx − c2x m2x + i(cx cy my − 2cx Imλ′ − θx )mx )
m ,m
2
x y

1
× exp(− c2y m2y − i(2cy Reλ′ + θy )my ). (4.3.71)
2

We can write Eq.(4.3.71) and Eq.(4.3.70) in terms of the ϑ-functions. For even nΦ
we get

hλf f ′ i = |C|2 e− 21 d2x exp(−2idx Imλ′ )

′ λ|T l |λλ
x
   
d x cx ′ 1 − 12 c2x ′ −2c2y
× ϑ3 i − cx Imλ − θx , e ϑ3 −2cy Reλ − θy , e
2 2
  !
2
+ (−1)l exp(−c2y )ϑ2 −2cy Reλ′ − θy , e−2cy , (4.3.72)
90 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD

and for odd nΦ we get

 
f ′i 2 − 12 d2x d x cx 1 1 2
hλf
′ λ|T l |λλ
x = |C| e ′
exp(−2idx Imλ ) ϑ3 i − cx Imλ′ − θx , e− 2 cx
2 2
 2

× ϑ3 −2cy Reλ′ − θy , e−2cy
 2

+ (−1)l exp(−c2y )ϑ2 −2cy Reλ′ − θy , e−2cy
 !
d x cx 1 1 2
× ϑ4 i − cx Imλ′ − θx , e− 2 cx . (4.3.73)
2 2

Similarly, we can show that

X
hλf f ′ i = |C|2 hλ′ λ|λλ′ie− 21 d2y exp(−2idy Reλ′ )
′ λ|T l |λλ (−1)mx
y
mx ,my
1
× exp(−dy cy my − c2x m2x + i(cx cy my − 2cx Imλ′ − θx )mx )
2
1 2 2
× exp(− cy my − i(2cy Reλ′ + θy )my ), (4.3.74)
2

and for even l

hλf f ′ i = |C|2 hλ′ λ|λλ′ ie− 21 d2y exp(−2idy Reλ′ )

′ λ|T l |λλ
y
X 1
× exp(−dy cy my − c2x m2x + i(cx cy my − 2cx Imλ′ − θx )mx )
m ,m
2
x y

1
× exp(− c2y m2y − i(2cy Reλ′ + θy )my ). (4.3.75)
2

We can write Eq.(4.3.75) in terms of the ϑ-functions. For even nΦ we get

hλf f ′ i = |C|2 e− 21 d2y exp(−2idy Reλ′ )

′ λ|T l |λλ
y
   
′ 1 − 21 c2x 2
× ϑ3 −cx Imλ − θx , e ϑ3 idy cy − 2cy Reλ′ − θy , e−2cy
2
 2

+ exp(−c2y )ϑ2 idy cy − 2cy Reλ′ − θy , e−2cy , (4.3.76)
4.3. PARTICLE ON A TORUS 91

and for odd nΦ we get

 
′
2 − 12 d2y 1 1 2
hλf y
f
′ λ|T l |λλ i = |C| e ′
exp(−2idy Reλ ) ϑ3 −cx Imλ − θx , e− 2 cx
′
2
 
′ −2c2y
× ϑ3 idy cy − 2cy Reλ − θy , e
 
2 ′ 1 − 12 c2x
+ exp(−cy )ϑ4 −cx Imλ − θx , e
2
!
 2

× ϑ2 idy cy − 2cy Reλ′ − θy , e−2cy . (4.3.77)

For odd l and even nΦ Eq.(4.3.70) gives

hλf f ′ i = |C|2 e− 21 d2y exp(−2idy Reλ′ )

′ λ|T l |λλ
y
   
′ 1 − 12 c2x 2
× ϑ4 −cx Imλ − θx , e ϑ3 idy cy − 2cy Reλ′ − θy , e−2cy
2
!
 2

− exp(−c2y )ϑ2 idy cy − 2cy Reλ′ − θy , e−2cy , (4.3.78)

while for odd nΦ , we get

hλf f ′ i = |C|2 e− 21 d2y exp(−2idy Reλ′ )

′ λ|T l |λλ
y
   
′ 1 − 21 c2x 2
× ϑ4 −cx Imλ − θx , e ϑ3 idy cy − 2cy Reλ′ − θy , e−2cy
2
 
2 ′ 1 − 21 c2x
− exp(−cy )ϑ3 −cx Imλ − θx , e
2
!
 2

× ϑ2 idy cy − 2cy Reλ′ − θy , e−2cy . (4.3.79)

Here we must keep in mind that the value of C depends on whether nΦ is odd or
even. This is obvious from Eq.(4.3.65) and Eq.(4.3.66).
92 CHAPTER 4. PARTICLE IN CONSTANT MAGNETIC FIELD
Chapter 5

Conclusions

We have considered the physics of a particle confined to the surface of a cone with
deficit angle δ and bound to its tip by a 1/r or an r 2 potential. In both cases, for
rational s = 1 − δ/2π, all bound classical orbits are closed and there are accidental
degeneracies in the discrete energy spectrum of the quantum system. There is an
accidental SU(2) symmetry generated by the Runge-Lenz vector and by the angular
momentum. However, the Runge-Lenz vector is not necessarily a physical operator.
For example, by acting with the Runge-Lenz vector on a physical state one may
generate an unphysical wave function outside the domain of the Hamiltonian. As
a result, the representations of the accidental SU(2) symmetry are larger than the
multiplets of degenerate physical states. In particular, some physical states are
contained in multiplets with an unusual value of the Casimir spin S which is neither
an integer nor a half-integer. Still, the fractional value of the spin yields the correct
value of the quantized energy.

In the study of the problem of a particle moving on a cone, a procedure was

developed to reduce the problem of finding the conserved components of the Runge-
Lenz vector to a problem of solving a system of differential equations (see section
(3.4) and subsection (3.5.5)). The ansatz was made with the help of the classical
expression of the Runge-Lenz vector. This procedure will be an important tool to
solve more complicated problems like the cone problem in higher dimensions.

The particle on a cone provides us with an interesting physical system in which

symmetries manifest themselves in a very unusual manner. Although the Hamilto-
nian commutes with the generators of an SU(2) symmetry, the multiplets of degen-
erate states do not always correspond to integer or half-integer Casimir spin. This is
because the application of the symmetry generators may lead us out of the domain
of the Hamiltonian. Only the states with square-integrable single-valued 2π-periodic
wave functions belong to the physical spectrum, and all other members of the corre-

93
94 CHAPTER 5. CONCLUSIONS

sponding “SU(2)” representation must be discarded. Mathematically speaking, the

symmetry generators — although Hermitean in their respective domain — do not
act as Hermitean operators in the domain of the Hamiltonian.

It contrast to many other quantum mechanics problems, in order to understand

motion on a cone it was necessary to address mathematical issues such as the do-
mains of operators as well as Hermiticity versus self-adjointness. Still, we have not
elaborated on some questions related to different possible self-adjoint extensions of
the Hamiltonian. For the particle on the cone, such issues seem worth investigat-
ing. In this work, we have limited ourselves to the standard Friedrichs extension
of the Hamiltonian. Alternative self-adjoint extensions correspond to an additional
δ-function potential located at the tip of the cone. This will modify the problem
in an interesting way. In particular, we expect that, in the presence of an addi-
tional δ-function potential, the accidental degeneracy will be partly lifted. However,
since the δ-function only affects states with m = 0, some accidental degeneracy will
remain. The particle on the cone provides us with another example for the deep
connection between the closedness of all bound classical orbits and accidental de-
generacies in the discrete spectrum of the Hamiltonian. Even if the classical system
has various quantum analogues (because there are different possible self-adjoint ex-
tensions) some accidental degeneracy still persists. It is also remarkable that, like
in other cases with accidental symmetries, for the particle on the cone semi-classical
Bohr-Sommerfeld quantization provides the exact quantum energy spectrum.

We are unaware of another system for which a similarly unusual symmetry be-
havior has been observed. It is interesting to ask if symmetry can manifest itself in
this unusual manner also in other quantum systems. For example, cones of graphene
may provide a motivation to study accidental degeneracies of the Dirac equation on
a cone. Also higher-dimensional spaces with conical singularities may be worth in-
vestigating. In any case, motion on a cone provides an illuminating example for a
rather unusual manifestation of symmetry in quantum mechanics.

We have also re-investigated an old and rather well-studied problem in quantum

mechanics — a charged particle in a constant magnetic field — from an unconven-
tional accidental symmetry perspective. While many aspects of this problem are
well-known, and some results of this thesis can be found in various places in the lit-
erature, we believe that we have painted a picture of cyclotron motion that reveals
new aspects of this fascinating system, which behaves in a unique and sometimes
counter-intuitive manner. The fact that all classical cyclotron orbits are closed
circles identifies the center of the circle as a conserved quantity analogous to the
Runge-Lenz vector of the Kepler problem. Remarkably, the corresponding “acci-
dental” symmetry is just translation invariance (up to gauge transformations). In
particular, the coordinates (Rx , Ry ) = (−Py , Px )/eB of the center of the cyclotron
circle simultaneously generate infinitesimal translations −Py and Px (up to gauge
 95

transformations) in the y- and x-directions, respectively. As is well-known, in a

constant magnetic field, translations in the x- and y-directions do not commute,
i.e. [Px , Py ] = ieB, and thus the two coordinates Rx and Ry of the center of the
cyclotron circle are also not simultaneously measurable at the quantum level. In
contrast, the radius of the cyclotron circle has a sharp value in an energy eigenstate.
The accidental symmetry leads to the infinite degeneracy of the Landau levels.

In order to further investigate the nature of the accidental symmetry, we have put
the system in a finite rectangular periodic volume. Obviously, this breaks rotation
invariance, but leaves translation invariance (and thus the accidental symmetry) in-
tact — at least at the classical level. Interestingly, at the quantum level continuous
translation invariance is explicitly broken down to a discrete magnetic translation
group, due to the existence of two angles θx and θy which parametrize a family
of self-adjoint extensions of the Hamiltonian on the torus. In a field theoretical
context, in which the gauge field is dynamical (and not just treated as a classical
background field), the parameters θx and θy characterize super-selection sectors. In
this sense, they are analogous to the vacuum angle θ of non-Abelian gauge theo-
ries. Just like the θ-vacuum angle explicitly breaks CP invariance at the quantum
level but is classically invisible, the angles θx and θy lead to quantum mechanical
explicit breaking of continuous translation invariance down to the discrete magnetic
translation group. The magnetic translation group G itself, which plays the role of
the accidental symmetry in the periodic volume, is a particular central extension of
Z(nΦ ) ⊗ Z(nΦ ) by the center subgroup Z(nΦ ), where nΦ is the number of magnetic
flux quanta trapped in the torus. We find it remarkable that the simple fact that all
classical cyclotron orbits are closed circles has such intricate effects at the quantum
level.

One of the most important results of studying the cyclotron motion on a torus
concerns the angles θx and θy which are parameters of self-adjoint extensions. Here
we must stress that not taking these parameters into account leads to a conceptual
difficulty. To understand this, let us ignore these parameters, and let us consider
the case when the flux quantum number nΦ = 1. Then there is no degeneracy. The
wave function in Eq.(4.3.33) then is a coherent state. Since there is no prefered
point on the torus, we expect that the probability density is constant. However,
this is not the case. Instead, the probability density is coordinate-dependent and
has a maximum at the origin. This means that we would have a prefered point
at the origin. The introduction of θx and θy solves this problem. For nΦ = 1 the
probability density is centered at the point (θx Lx /2π, θy Ly /2π). This is what one
expects since these parameters break the translation symmetry of the system.

The treatment that has been used in the cyclotron problem can be generalized to
include the effect of spin. Then we can repeat the whole treatment using the Pauli
equation or the Dirac equation. This will be a possible subject for future work.
96 CHAPTER 5. CONCLUSIONS

The observation that there is a deep relation between accidental symmetry and
bound classical orbits being closed has been supported by the construction of the
conserved quantities for systems with closed bound orbits. However, there is no
rigorous proof of this relation based on fundamental laws of classical mechanics.
The work in this thesis gives a motivation to find such a proof, and derive a general
procedure to find conserved quantities generated by accidental symmetry, or for any
other classical system.

It is worth mentioning that it is possible that a system with continuous fields

may have an accidental symmetry. In this case a closed orbit can not necessarily
be defined although the system possesses an accidental symmetry. For example,
Fröhlich and Studer proved [36] expanded the Dirac equation for a particle in an
electromagnetic field in powers of the inverse mass 1/M. If we assume that M is
sufficiently large, then we can limit ourselves to terms up to the power 1/M 3 . The
result is an equation with an accidental SU(2) gauge symmetry in addition to the
usual U(1) gauge symmetry of electromagnetism. The derivation of this result is not
based on a standard theorem applied to a Lagrangian of classical fields, but based
on observation. This indicates the potential of a very important problem in classical
mechanics of finding a general theorem that is applicable to continuous as well as
discrete systems, illuminating the relation between closed bound orbits and acci-
dental symmetry. In addition to that, such a theorem may provide a valuable tool
to extract conserved quantities from a complicated Lagrangian when mere intuition
fails to do so.

We find it remarkable that a well-studied subject such as quantum mechanics

still confronts us with interesting puzzles. In particular, accidental symmetries hold
the promise to further deepen our understanding of the essence of quantum physics.
Acknowledgements

I like to thank U.-J. Wiese for supervising this thesis. I am indebted to him for
his valuable explanation of important facts regarding the foundations of quantum
physics and for his endless patience in addressing concerns and answering questions
related to this thesis and beyond. My discussions with him improved my under-
standing of quantum physics significantly.

Further thanks go to Prof. Shailesh Chandrasekharan for several interesting dis-

cussions and, in particular, for his willingness to act as a co-referee of this thesis.

I am also indebted to F. Niedermayer for illuminating discussions.

Deep thanks go to Ottilia Haenni for her very valuable support in finding accom-
modation, dealing with administrative issues, and for her warm and kind treatment.

97
98 CHAPTER 5. CONCLUSIONS
Bibliography

[1] L. Schiff, Quantum Mechanics, Third edition, McGraw Hill. Inc., 1968.

[2] J. Bertrand, C. R. Acad. Sci. Paris 77 (1873) 849.

[3] H. Goldstein, Classical Mechanics, Second Edition, Addison Wesley Publishing

Company, Inc., 1980.

[4] A. Holas and N. H. March, J. Phys. A23 (1990) 735.

[5] Z.-B. Wu and J.-Y. Zeng, Phys. Rev. A62 (2002) 032509.

[6] R. C. O’Connell and K. Jagannathan, Am. J. Phys. 71 (3) (2002) 243.

[7] U.-J. Wiese, 1982, Solution of the Schrodinger equation for the Kepler Problem
modfied by an angular momentum dependent potential, (unpublished work).

[8] C. Furtado and F. Moraes, J. Phys. A: Math. Gen. 33 (2000) 5513.

[9] M. Al-Hashimi and U.-J. Wiese, Ann. Phys. 323 (2008) 82.

[10] G. B. Folland, Harmonic Analysis in Phase Space, Princeton University Press,

1989.

[11] V. A. Dulock and H. V. McIntosh, J. Math. Phys. 7 (8) (1966) 1401.

[12] A. Stern, Ann. Phys. 323 (2008) 204.

[13] G. Chen and L. Kuang, arXiv:cond-mat/9509091v1.

[14] X. Wen and Q. Niu, Phys. Rev. B 41 (13) (1990) 9377.

[15] V. Fock, Z. Physik 98 (1935) 145.

[16] V. Bargmann, Z. Physik 99 (1936) 576.

[17] W. Lenz, Z. Physik 24 (1924) 197.

[18] L. D. Landau, Z. Physik 64 (1930) 629.

99
100 BIBLIOGRAPHY

[19] M. H. Johnson and B. A. Lippmann, Phys. Rev. 76 (1949) 828.

[20] H. V. McIntosh, in Group Theory and its Applications, Vol. 2 (1971) 75, Aca-
demic Press, Inc. New York and London.

[21] M. Reed and B. Simon, Methods of modern mathematical physics, Vol. I. Func-
tional analysis, New York, London: Academic Press 1972; Vol. II. Fourier anal-
ysis, self-adjointness; New York, London: Academic Press 1975.

[22] J. Weidmann, Spectral Theory of Ordinary Differential Operators, Lecture

Notes in Mathematics, Springer, 1987.

[23] B. S. Kay and U. M. Studer, Commun. Math. Phys. 139 (1991) 103.

[24] G. ’t Hooft, Commun. Math. Phys. 117 (1988) 685.

[25] S. Deser and R. Jackiw, Commun. Math. Phys. 118 (1988) 495.

[26] P. de Sousa Gerbert and R. Jackiw, Commun. Math. Phys. 124 (1989) 229.

[27] P. de Sousa Gerbert, Phys. Rev. D40 (1989) 1346.

[28] G. de A. Marques and V. Bezerra, Class. Quantum Grav. 19 (2002) 985.

[29] S. Azevedo, C. Furtado, and F. Moraes, Phys. Stat. Sol. 207 (1998) 387.

[30] Y. A. Sitenko and N. D. Vlasii, arXiv:0706.2756 [cond-mat.str-el].

[31] Website: www.wolframfunctions.com.

[32] H. Zainuddin, Phys. Rev. D. 40 (2) (1989) 636.

[33] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products,

Fifth Edition, Academic Press 1996.

[34] J. R. Klauder, Coherent States, Bo-Sture Skagerstam Singapore: World Scien-

tific, 1985.

[35] G. ’t Hooft, Commun. Math. Phys 81 (1981) 267.

[36] J. Fröhlich and U. M. Studer, Rev. Mod. Phys. 65 (3) (1993) 733.