in partial satisfaction of mevsem ’17

Physical Applications of Topological Quantum Field Theory

Introduction to Chern-Simons Theory

Ademola Adeifeoba†
†
Institute for Theoretical Physics,
Heidelberg, Germany.

E-mail: [email protected]

Abstract: The 2 + 1 Yang-Mills theory allows for an interaction term called the Chern-Simons
term. This topological term plays a useful role in understanding the field theoretic description of
the excitation of the quantum hall system such as Anyons. While solving the non-Abelian Chern-
simons theory is rather complicated, its knotty world allows for a framework for solving it. In the
framework, the idea was to relate physical observables with the Jones polynomials. In this note, I
will summarize the basic idea leading up to this framework.
Contents

0 Prologue 2

1 Introduction 2

2 Elements of Anyon Models 4

2.1 Particle Type 4
2.2 Fusion Rule 4
2.3 Artin’s braid group 6

3 Chern-Simons Theory 8
3.1 Classical CS Invariant 8
3.2 CS Model of Anyons 8

4 Wilson Loops and Anyonic Statistics 12

4.1 Abelian and non-Abelian 12
4.2 Expectation Value 13

5 The Jones Polynomial 16

5.1 Knots and Links 16

6 Non-Abelian CS theory 19
6.1 Quantization on Σ × R 19
6.1.1 Quantization on Σ with punctures 20
6.2 Evaluation by TQFT 21
6.2.1 Unknotting Procedure 22

–1–
0 Prologue

This note is a report in partial satisfaction of my study and seminar work on “Physical Applications
of Topological Quantum Field Theory” at the University of Heidelberg. It is fair to declare forward
that, while I have made efforts to carefully craft the basic ideas I have tried understood about a
very fascinating topic within a very limited time, there are chances of omission or partial coverage
of some concepts. I tried as much to cite important references where I consider it needful.

We shall study Witten’s original paper on “Quantum field theory and the Jones polynomial.” [1].
Elementary readers like me will agree that without some formal background understanding, this
article belong to a class usually tagged “hard to read”, more because it assumes many important
advances, not only in physics but also in mathematics, as prerequisite. As such, I consider a short
introduction about few background stories as useful and motivating towards our study. Following
that, I will discuss the important elements of model description of anyons. The features introduced
will play important roles throughout the remaining sections. I will follow similar structure of the
lecture, but more explicit in this note. Our final destination is the evaluation of Wilson link invariant
of the non-Abelian Chern-simon theory. The aim is to derive the rule for unknotting the Wilson
link and compute the expectation value by relating physical observables to Jones polynomials which
are knots invariant. As such, the roads to our final destination demands few important branches
to transit towards an expected end.

1 Introduction

One of the ideas that revolutionized 20th century physics is Quantum field theory – a theoretical
framework for modeling quantum mechanical system of elementary particles. Basically, the world
is quantum, classicality is a certain limit of it and fields are nothing but part of reality. The
emergent particles are excitation of fields. While in classical mechanics, particles are considered
as distinguishable, new features arise on transition into Quantum theory. Among them is the
fact that identical particles are indistinguishable. As a consequence, interchanging particles in a
multiparticle state does not lead to a new configuration of the system, so that all probabilities is
the same under such operation, i.e.

|ψ(π(r1 , ...., rn ))|2 = |ψ(r1 , ...., rn )|2 , (1.1)

where π is the permutation operation of n particle co-ordinates. More precisely, the wavefunction
is left invariant by the interchanging operation up to a certain phase. This understanding led
to quantum statistical decription of particles, where interchanging operation in three or higher
spatial dimension leads to an interpretation of wavefunction as symmetric or antisymmetric under
exchange depending on whether it transforms as boson or fermion in the following sense:
(
+1, boson
ψ(r1 , r2 ) −→ η ψ(r2 , r1 ) −→ η 2 ψ(r1 , r2 ), η= (1.2)
−1, f ermion.
The study of planar physics is not a new thing in physical study. In particular, the 2D physics
received some attention and rapid progress in the 70–90s towards understanding of possible particle
statistics and their behaviour. On the first hand is an experimental physics leading to the discovery
of fractional quantum Hall effect[2]. Way back in 1879, Edwin Hall had discovered the classical

–2–
Hall effect as a result of the motion of charged particles in a magnetic field[3]. An interesting fact
about the very fractional quantum Hall effect is that, the hall conductivity

e2
σxy = ν, ν∈Q (1.3)
2π~
is surprisingly quantized. On the other hand is a theoretical curiousity leading to the prediction of
quasiparticles and quasiholes whose statistics are different from those of bosons and fermions. The
major stages include the following:

• Laughlin Wavefunction: This wavefunction was proposed as an ansatz to the lowest Lan-
dau levels at filling fraction ν = m
1
for m ∈ Z, with the excitations taking fractional charge
e
m [4].

• Statistics: In a system restricted to two spatial dimensions, some emergent phenomena

could be observed [5–8], namely, emergent particles turns out to obey statistics interpolating
between Fermi–Dirac and Bose–Einstein. These fractionally charged quasiparticle excitations
of the Laughlin state are called Abelian Anyons or simply Anyons. The excitations emerge
with fractional statistical angle θ = m
π
as their wavefunctions acquire phase factor of η =
e when identical quasiparticles undergo an exchange operation similar to (1.2). When
iθ

exchange operation which transforms the system’s quantum state is not commutative, non-
Abelian anyons arise. Examples of non-Abelian quasiparticle statistics are found in system
of Fibonacci anyons and Ising anyons.

• Hierarchy states: The idea that a more generic category of fractional quantum hall effect
defined by infinite continued filling fraction
1
ν= 1 (1.4)
m± e 1± m
m 1
e 2 ±...

was proposed and later validated quantitatively. [6, 9]

The field theoretic framework underlying this planar physics phenomena is known as Chern–simons
theory (CS). It is a topological quantum field theory (TQFT) featuring topological invariant ob-
servables. The field theory enjoys general covariance since there is no a priori choice of metric on
the manifold on which the theory is formulated. As we shall see furthermore, the anyonic phenom-
ena does not emerge from this theory as a result of gauge invariance, but rather as a consequence of
the theory being quasi-invariant, i.e. while the action is not fully gauge invariant, the observables
and therefore the partition function are fully gauge invariant.

Remarkably, the nature of anyons has inspired some practical purposes such as in topological quan-
tum computation. The very features of Chern-Simons invariants have lead to new developments in
mathematics. One of our goals in this study is to understand how quantum Chern-Simons invariant
is related to the Jones invariants of link which have found useful applications in knot theory.

–3–
2 Elements of Anyon Models

“Since interchange of two of these particles can give any phase,

I will call them generically anyons.” – Frank Wilczek in [8]

There is a general structure that all non–Abelian anyon models are required to feature. In this
section, our goal is to discuss these basic features characterizing a model description of anyon. The
features arise as a consequence of the following insights:

• Anyons can be created or annihilated pairwisely;

• Anyons can be fused to form composite anyons;

• Anyons has braiding rules.

These features will play roles in the model description of anyons by the planar Chern–Simons
theory.

2.1 Particle Type

A non-trivial anyonic system generally consists of multiple types of anyon. Therefore, to define an
anyon model, we must declare all the distinct type of anyon in the model. A given anyon model is
expected to have:

• a finite set of elementary particle specie, say, {φa , φb , φc , ...}. Each particle is locally dis-
tinguished by its label a, b, c, ... which we think of as topological charges. These topological
charges are conserved quantum numbers.

• trivial particle φ1 which corresponds to a unique vacuum.

• antiparticle φā such that φa × φā = φ1 = φ1̄ . This implies that particles are only created in
pair from the vacuum.

The simplest non-trivial anyon model is therefore spanned by the particle set

Ξ = {φ1 , φa , φb , φc , ...}. (2.1)

2.2 Fusion Rule

Given that anyons come in multiple, bound states could exist in principle. Even when no sta-
ble bound state exists, composite anyon can be formed by bringing two anyons close together.
The fusion rule of Abelian anyon and the quantum number of the eventual composite particles
is straightforward. There is only one possible fusion channel. For example, fusion of two given
2 2
Abelian anyons with statistic αmπ and βmπ is given as

α2 π β 2 π (α + β)2 π
× = . (2.2)
m m m
What is the statistical behaviour of composite non-Abelian anyon? Tensor product doesn’t come
in handy. The answer to this arise by recognizing that topological quantum numbers combine in
non-unique manner. This means that there are different possible fusion channels. Fusion rule is
given as (
1, Abelian anyons
φa × φb = Nab φc , Z 3 Nab =
c c
(2.3)
≥ 2, non − Abelian anyons.

–4–
Fusion is depicted in figure 1.The integer number Nabc is the number of distinguishable fusion

channels. These fusion channels can be understood to be orthonormal basis states

c
{|ab, c; µi , µ = 1, 2, ..., Nab } (2.4)
of the fusion Hilbert space Vab
c with the following properties:

• orthonormality: hab, c; µ |ab, d; νi = δcd δµν

• completeness: X
|ab, d; µi hab, c; µ| = Iab ,
µ,c
where Iab is the trivial fusion channel.

• For non-Abelian anyons, !

M X
c c
dim Vab = Nab ≥2
c c

• fusion is associative, i.e.

(
φa × φb = φb × φa
(φa × φb ) × φc = φa × (φb × φc ).
This basically insist that the order of tensor product is irrelevant, so that one order of fusing
three anyons can be rotated into another in the fusion space by the Fusion matrix Fabc d as

described in figure 2.

Figure 1. Worldlines of anyons. The first diagram features anyonic fundamental braid. Second diagram
depicts pair-creation from the vacuum of a particle a and its antiparticle ā. The third describes the fusion
of two particles a and b by two lines fusion into composite c

The simplest model of anyonic system is Fibonacci anyons[11] . It is an anyonic system having
only two particle types: the vacuum φ1 and the nontrivial particle type φτ . Fibonacci anyons are
self-dual, i.e. particle τ is the same as its anti-particle. Fusion rules just go by
(
φτ × φ1 = φ1 × φτ = φτ
φτ × φτ = φ1 ⊕ φτ ,

implying that the dimension of fusion Hilbert space is 2.

–5–
Figure 2. Order of fusion mapped by sufficient number of moves implemented by Fusion matrix Fabc d
.
Each order of fusion corresponds to a choice of basis with F matrix as a transformation between different
bases. That i and j are different is consistent with having a fixed composite fusion outcome d.

2.3 Artin’s braid group

In the introduction, we recognised in (1.1) that under the exchange of particles in 3 + 1 spacetime
dimensions, wavefunction describing a system of identical particles acquires ± phase and therefore
exhibits two possible types of symmetry depending on whether it is a boson or a fermion. In path
integral interpretation, twice operation of particle exchange is equivalent to a process in which one
of the particles is taken along a trajectory that wrap around the other. Different trajectories belong
to topological class corresponding to the elements of the permutation group SN specifying how the
initial positions are permuted into the final positions. These trajectories in three spatial dimensions
can be continuously deformed into straight in time direction. As such, the wave function is left
invariant under exchange operation.

We further learnt that new statistical behaviour emerges in 2 + 1 spacetime dimensions, and anyons
can exist in principle. The precise statement is that, under twice operation of particle exchange,
η 2 isn’t necessarily equals 1, so that the wavefunction need not return to the initial state, i.e. it
acquires a non–trivial phase such that

ψ(r1 , r2 ) −→ eiθ ψ(r2 , r1 ) −→ ei2θ ψ(r1 , r2 ), (2.5)

and the special case of boson and fermions are realized for θ = 0, π respectively. Particles with sta-
tistical angle θ interpolating between 0 and π are called anyons. In general case of n-particle state
ψ(r1 , ...., rn ), the topological classes of trajectories which take these particles from initial positions
(ti , r1 , ...., rn ) to final position (tf , r1 , ...., rn ) is in one-to-one correspondence with the elements of
the braid group Bn . In figure 1 is a description of anyon worldlines originating from a plane with
a characteristic fundamental braid.

Braid is an everyday concept but the formal definition of an n-braid as a topological object was
given by Artin[10]. Consider two parallel frames with each frame as a plane in euclidean 3-space.
Let pi (i = 1, ..., n) be distinct mark ponts on the lower frame ΣL and denote their orthogonal
projections onto the upper frame ΣU as qi . Each pi can be joined with qj by means of strings
which intersect any parrallel plane between ΣL and ΣU exactly once.

–6–
A braid is an intertwining of some number of strings fi : I = [0, 1] 7−→ Σ such that:

1. fi (0) = pi ,

2. fi (1) = pπ(i) = qi for some for some permutation π,

3. Given that i 6= j, then fi (t) 6= fj (t) for each t ∈ I.

The braid group Bn therefore consists of the set of all braids with n strings under multiplication
operation given by glueing. The braid group is furnished in a presentation provided in the following
theorem:

Theorem (Artin[10]): Let Bn be a group generated by σ1 , σ2 , ..., σn−1 modulo the following rela-
tions:

1. σi σj = σj σi for |i − j| ≥ 2.

2. σi σi+1 σi = σi+1 σi σi+1 for i = 1, ..., n − 2,

There is an isomorphism ρ : Bn 7−→ Bn determined by σi .

The elementary braids are those generated by σi with just one crossing. The group Bn is different
from permutation group in that σi2 6= 1. As a consequence, while the permutation group is finite
with |SN | = N !, the braid group is infinite. The two relations in the Artin’s theorem are better
revealed in figure 3 which notes that one composition can be smoothly deformed into the other.
The first relation implies that the order of composition is important because the braid group is non-
Abelian, e.g. σ1 σ2 6= σ2 σ1 . The second relation is nothing but Yang–Baxter equation characterizing
quantum integrable system. The Yang–Baxter equation is a statement that the two possible paths
that three strings can take are different but equivalent.

Figure 3. Relations in Artin’s braiding.

–7–
3 Chern-Simons Theory

We have recognized Quantum field theory as a framework of model description of elementary par-
ticles. As we shall now demonstrate, the low energy behaviour of the fractional quantum Hall
effect can be described by a certain planar Yang–Mill theory in the presence of Chern-Simons(CS)
term [12]. In the following, we shall use both classical and quantum mechanical understanding
to demonstrate that the source of Abelian and non-Abelian Chern-Simons theories have relevant
anyonic features.

3.1 Classical CS Invariant

Let Σ be a surface embedded in 3D Euclidean space with a compact boundary ∂Σ. The Riemannian
metric determines a measure dµΣ on Σ and defines the curvature of Σ as a function K : Σ → R.
The curvature is intrinsic in that it depends only on the induced metric on Σ, and not on the
embedding of Σ into space. The generalized Gauss-Bonnet formula is
Z Z
K dµΣ + κ∂Σ dµ∂Σ = 2πχ (Σ) (3.1)
Σ ∂Σ

where χ (Σ) is the Euler characteristic of Σ and κ is the total geodesic curvature of the boundary.
The classical CS invariant[13] is a generalization of the total geodesic curvature.

Let G be a finite Lie group and t be its Lie algebra. We fix a certain closed oriented 3-manifold
M with m-form Ωm M (t) living on it. A connection Θ = Θµ ta dx on the G-bundle over M is
a µ

a skew-Hermitian matrix of 1-forms having trace zero. The curvature of Θ is a t-valued 2-form
Ω2 (Θ) = dΘ + 21 [Θ ∧ Θ] satisfying the Bianchi identity

dΩ + [Θ ∧ Ω] = 0.

The Chern-Simons 3-form is given as

2
Ω3 (Θ) = Θ ∧ dΘ + [Θ ∧ Θ ∧ Θ] , (3.2)
3
and the corresponding classical CS invariant is
Z  
k 2
SCS (Θ) = Tr Θ ∧ dΘ + [Θ ∧ Θ ∧ Θ] , (3.3)
4π M 3
with k being the level of the theory. CS field theories exist for all odd and higher dimensions. The
generalized CS form in 2n + 1 dimensions reads as

Ω2n+1 (Θ) = Θ ∧ (dΘ)n + α1 Θ3 ∧ (dΘ)n−1 + ..... + αn Θ2n+1 (3.4)

and αi ∈ Q is fixed.

3.2 CS Model of Anyons

Let aµ be an emergent gauge fields of a certain Yang-Mill theory. In QFT, all physical information
are encoded in the n–point correlation functions which can be extracted from functional differential
of the partition function Z
Z[aµ ] = DΨeiS[Ψ,aµ ] = eiSef f [aµ ] , (3.5)

–8–
where Ψ represent all dynamical fields which are integrated out towards arriving at the effective
field action Sef f . The effective action describes some physics, accurate at the very low energy. Our
first goal will be to identify the Chern-Simons action which describes the relevant physics of interest.

Abelian Chern-Simons theory emerge as a possible generalization of 3+1 electromagnetic U (1)

gauge theory to 2+1 dimensions. Locally in four spacetime dimensions, gauge connection aµ is
described by massless degrees of freedom, whose dynamics are encoded in the Maxwell’s equations
obtained by extremizing the action
Z
1
Smaxwell [aµ ] = − 2 d4 x fµν f µν , (3.6)
4e M
where fµν = ∂µ aν − ∂ν aµ is the curvature known as the electromagnetic field strength. We adopt a
manifold M as a space on which the theory is formulated. We assume M has trivial topology with
boundary ∂M arbitrarily far from particle worldlines, so that all field vanish on the boundary.
The action is the most general Lorentz invariant action compatible with gauge transformations
aµ → aµ + ∂µ α. The massless gauge field is in turn a consequence of gauge invariance which forbids
some forms of interaction including a mass term. However in three spacetime dimensions, new
interaction terms is allowed in the full action

S = Smaxwell + SCS ,

which can potentially change the dynamics of the system. This term establishes the Abelian
Chern-Simons action Z
k
SCS [a] = d3 x µνρ aµ ∂ν aρ + (......) (3.7)
4π M
where (......) may include some higher derivative terms. Before we explain why this action encodes
an effective field theory describing the quantum hall system, it might be useful to generalize the
Chern-Simons action into non-Abelian just as in the usual QFT formalism. We recognize the
relevant Lie gauge group G = SU (N ) with aµ = acµ tc considered as Lie(G)-valued gauge potential.
The killing form κab = tr ta tb = 21 δ ab can be used to raise and lower indices of the Lie-algebra.

The curvature associated with the connection is

fµν = ∂µ aν − ∂ν aµ − i[aµ , aν ].

The full action in 2+1 dimension is therefore nothing but a quantum Yang-Mills theory, with an
action consisting of the Chern-Simons term that has additional 3-point gauge structure, i.e.

S = SY M + SCS

with Z  
k 2i
SCS [a] = 3
d x  µνρ
Tr aµ ∂ν aρ − aµ aν aρ . (3.8)
4π M 3
This is the action of level k non-Abelian Chern-Simons theory with gauge group Gk . It takes the
same structure as the Chern-Simons invariant (3.3). The most important gauge group for the
purpose of our study will be SU (2)k .

So, what are the features of Chern-simons theory that makes it a good effective field theory of
the quantum hall system? In the following, we shall highlight both the features and some physics
associated with the hall system.

–9–
• The Chern-Simons action has no metric notion, instead, the Levi-Civita µνρ plays the role
of raising and lowering. As such, the CS action is invariant under all diffeomorphism. The
consequence of this is that, all n-point correlation functions are independent of the metric of
spacetime, i.e. δgδµν hO1 (x1 )....On (xn )i = 0. This make the correlation functions topological
invariants and Chern-Simons theory a good example of topological quantum field theory.

• On dimensional ground, the Chern-Simons term dominates over the Yang-Mill and (......)
term in 2+1 dimension as it contains only one derivative relative to higher derivatives in the
other terms. With Yang-Mill and other terms suppressed, the relevant long distance physics
of the hall system is completely described by the Chern-Simons effective local Lagrangian,
provided that k 6= 0.

• The Chern-simons action is invariant under rotation but violate parity (x → −x) and time
reversal (t → −t) invariance. This is exactly the feature exhibited by quantum hall system
where parity and time reversal are broken due to the external or background magnetic field
as particles are restricted to a plane.

• Conservation of current plus the fact that vector theory demands the gauge field couples with
the current implies that the current is the curl of the vector potential. Extremizing the CS
action (3.7), we have the current as

δSCS [a] k µνρ k

Jµ = =  ∂ν aρ =⇒ Ji = − ij E j .
δaµ 2π 2π

This means that, Chern-Simons action describes a system with conductivity

k
σxy = . (3.9)
2π
We shall soon understand that, indeed, this is the hall conductivity of the very quantum hall
effect if we identify the Chern-Simons level k with ν landau level of the quantum hall system.

• Under a gauge transformation

aµ −→ auµ = u−1 aµ u + iu−1 ∂µ u (3.10)

with u(x) ∈ G, the Chern-Simons action (3.8) transforms as

Z
k
SCS [a] −→ SCS [au ] = SCS [a] + d3 x µνρ ∂µ Tr ∂ν uu−1 aρ


4π M
Z (3.11)
k
d3 x µνρ Tr u−1 ∂µ u u−1 ∂ν u u−1 ∂ρ u .


+
12π M
This means that the action changes by a total derivative and an additional term
Z
1
Z 3 ς(u) = d3 x µνρ Tr u−1 ∂µ u u−1 ∂ν u u−1 ∂ρ u (3.12)


2
24π M

Recognizing the total derivative in (3.11), the purely surface integral vanishes on the bound-
ary. This follows because we have demanded all field to vanish on the boundary. However,
the term involving ς(u) does not necessarily vanish if we demand u(x) → 1 at infinity over
the entire argument. Since the gauge transformation u(x) is defined throughout R3 and its
argument is equivalent to S3 , the local gauge transformation is a map induced by a change

– 10 –
 of coordinates, i.e. u : S3 → G, and ς(u) measures the number of time the mapping winds
around the spacetime. This therefore implies that the Chern-Simons action transformaming
as
SCS [a] −→ SCS [au ] = SCS [a] + 2πk ς(u)
is not invariant under gauge transformation. At this point, it appears gauge invariance is an
obstacle towards having CS as a good model description of anyon, however, this violation is
precisely all that is left to produce exact same scenario of the hall conductivity. In QFT, the
object that must actually demand full compliance of gauge invariance isn’t the action itself,
but the very partition function (3.5) that encodes all information of the theory. Indeed, if we
demand the Chern-Simons level k to be an integer, so that on reinstating the natural units
e = 1 = ~, we can achieve
∈ Z,
~k
(3.13)
e2
then the partition function Z[aµ ] is always gauge invariant. In fact, by associating ~k
e2
with the
filling fraction ν in (3.9) i.e. ν = e2 , we precisely reproduce the quantized hall conductivity
~k

e2
σxy = 2π~ ν in (1.3). This is an important result arising from CS theory being quasi-invariant.

• It is worth noting that the hierarchy state filling fraction in (1.4) can be constructed from
this formalism by introducing N emergent gauge fields aiµ with i = 1, ..., N. The full effective
theory with quasiparticle excitations is given by[14]
1 1 µνρ
L= Kij µνρ aiµ ∂ν ajρ + ti  Aµ ∂ν aiρ (3.14)
4π 2π
where the theory is completely specified by the CS couplings Kij and the charge vector ti .
The K-matrix associated with the hierarchy in (1.4) is given as
 
m −1 0 . . .
−1 m e 1 −1
 
(3.15)

Kij =  0 −1 m e2


.. ..
 
. .

• A quick check into the field equation of the full theory shows that the gauge fields have
acquired mass which decay exponentially according as e−mr , thereby rendering them as short
ranged. This shows that, just like in the Aharonov-Bohm effect, the field can take large
values away from the sources.

By now, we have understood how Chern-Simons theory makes a good model description of certain
quasiparticles similar to those arising from the quantum hall system. Our next goal is to compute
the expectation values of CS observables and then study how it features the anyonic properties of
the source.

– 11 –
4 Wilson Loops and Anyonic Statistics

Solving a QFT is up to determination of the n-point correlation functions. For Chern-Simons

theory, we would like to evaluate the topological invariants of the theory. These are essentially the
expectation value of a certain observables Oi on the manifold M, i.e.
Z
1
hO1 (x1 )....On (xn )i = Da eiSCS [a] O1 (x1 )....On (xn ). (4.1)
Z M
In this evaluation, anyonic statistics will depends on the evolution of wavefunction under the
braiding operation of worldlines of anyons. We already saw braiding as a topological property of
anyons. We will now study how this arise from CS theoretic description.

4.1 Abelian and non-Abelian

Our first goal is to identify an appropriate observable on the manifold. Define the Chern-Simons
source as a set of particles on worldline in 2+1 spacetime dimensions. We would like to find a
physical description for space of trajectory of charged particles living inside a Manifold M with
boundary ∂M far away from worldlines, so that fields vanish on the boundary. Lets assign particle
a spin j in the representation of CS gauge group Gk . We demand that any appropriate observable
respect gauge invariance as the symmetry of the theory. There is a certain set of gauge invariant
functional integral that satisfy this purpose. It is called the Wilson loop.

Starting with the Abelian. Given a 1-form aµ as the gauge connection, an important concept is
holonomy HC,j , defined as the parallel transport along a closed curve C:
 I 
HC,j [a] = P exp iq µ
aµ dx , (4.2)
C
where P is the path-ordering operator and q is the charge associated with the particle source with
label j. Holonomies are important concept because any gauge invariant quantity involving the
connection aµ is a traces of holonomies. This gauge invariant object is the Abelian Wilson loop
 I 
WC,j [a] = Tr P exp iq µ
aµ dx . (4.3)
C
A simple interpretation will be to consider a source of quasiparticle-quasihole pair of type j, created
from the vacuum, evolves along Wilson loop and then fused back to the vacuum. This process is
described in figure 4. We shall soon see that the Chern-Simons level k restricts the allowed set of
spin representation j propagating in the loop.

Loosely speaking, a collection of disjoint loops (Ci , i = 1, ...N ) is called link (L) , so that we can
define the product of Wilson loops as
N I !
Y
WL,R [a] = Tr P exp iqi aµ dxµ . (4.4)
i=1 Ci

We can generalize the Wilson loop into non-Abelian by recognising that the gauge connection takes
its value from the Lie algebra spanned by tc . We can set the charge as q = 2j and define tc(j) = 2j tc .
The the non-Abelian Wilson loops in the representation R of the gauge group Gk is defined as
N I !
Y
WL,R [a] = TrR P exp i c c µ
t(j) aµ dx . (4.5)
i=1 Ci

– 12 –
Figure 4. Worldlines of particles in a 3D Euclidean space M. Wilson lines are defined by loops C1 and
C2 . The two loops together form a link. The link intersect disk Σ0 at points Pi . The diagram depicts four
particle sources pairwisely created from the vacuum; the source braided and then fused back to the vacuum

4.2 Expectation Value

I. Abelian CS
As we have noted in (4.1), the expectation value of the Wilson loop is given as

Da WL,j eiSCS [a]

R
hWL,j i = MR iSCS [a]
. (4.6)
M Da e

Consider an Abelian Chern-Simons theory (3.7) and a link of disjoint loops Ci associated with
charges qi living in the worldvolume M. Figure 4 depicts the case of two loops link. We are left
with a process in a link configuration such that, particles are created pairwisely from the vacuum,
propagate through the worldlines Ci and then fuse back to the vacuum. It is not hard to show in
the case of Abelian CS, that the expectation value of the Wilson loops is given by[15]
 
i2π X
hWL,j i = exp  qi qj Φ(Ci , Cj ) , (4.7)
k
i,j

with Gauss linking number

(x − y)ρ
I I
1
Z 3 Φ(Ci , Cj ) = dxµ dy ν µνρ , f or i 6= j. (4.8)
4π Ci Cj |x − y|3

Φ(Ci , Cj ) and hWL,j i are topological invariants ( of link. Φ(Ci , Cj ) measures the number of times
one loop Ci winds around another loop Cj . It may be positive or negative depending on the orien-
tation of the loops involved.

Comments:
1. The Gauss linking number is divergent for i = j. This can be resolved by a framing prescrip-
tion that displace one of the loops, say Ci , slightly in a given direction to a loop Ci0 . As such,
a ribbon structure with boundaries Ci and Ci0 is formed. This results in a well defined Gauss

– 13 –
 linking number.

2. To display the anyonic nature of the source, we consider the trivial loops L0 corresponding
to disentangled loops. We note that an entagled loop corresponds to a braided sources
characterizing anyons. Consider two loops C1 and C2 winding one another once according
as Φ(C1 , C2 ) = ±1 so that their trivial link corresponds to two disentangled loops. Then
braiding operation according to the expectation value (4.7) is given as
 2
4q
hWL,j i = exp i hWL0 ,j i. (4.9)
k
Indeed, the braiding operation left a phase. This implies that the Chern-Simons sources have
anyonic statistics.

3. The generalization of the expectation value to non-Abelian CS theory is complicated by large

class of links having the same Gauss linking number Φ(Ci , Cj ). We need a different algorithm
to resolve this situation. This algorithm aim to evaluate a certain knot invariant called the
Jones polynomials. However for now, consider two sources of a non-Abelian SU (2)k CS gauge
group characterized by a label j. Fusion of j1 and j2 is dictated by vector composition
λ(j1 ,j2 )
X
j1 ⊗ j2 = j. (4.10)
|j1 −j2 |

Are all j allowed by fusion rules in CS theory? The answer is no as not all j gives distinct
source . Indeed, expectation value of the non-Abelian Wilson loops for jmax > k2 produce the
same result as those of the representation with 0 ≤ j ≤ k2 . Therefore, there are only k + 1
allowed representations: 0, 21 , ..., k2 – with
(
j1 + j2 , j1 + j2 ≤ k2
λ(j1 , j2 ) = (4.11)
k − j1 − j2 , j1 + j2 > k2 .

II. Non-Abelian CS
We noted earlier that the generalization of the expectation value to non-Abelian CS theory is not
straightforward. The complication arises due to a large class of links having the same Gauss linking
number Φ(Ci , Cj ). We state forward at this point that, we are yet to have enough technique to
confront the problem. However, we shall lay down the formal approach towards the computation
of the partition function. Here, we shall sketch out Witten’s ideas. This idea will be used later to
resolve through the problem at hand.

Witten’s Procedure
Consider the three manifold M (picture it as S 3 ). Inside this manifold is our theory, i.e. Wilson
loop WC,j is living therein as shown in figure 5.

• Cut the 3-manifold M along along a Riemann surface Σ, into two 3-manifolds M1 and M2 .
Obviously, M1 and M2 contain boundary surfaces, say Σ1 and Σ2 , so that after cutting, the
boundary of the pieces become ∂M1 = Σ1 ∪ Σ and ∂M2 = Σ2 ∪ Σ∗ . Σ∗ is essentially Σ with
opposite orientation. The surface of the cut would appear like Σ × R, where Σ is the spatial
part and R is the time direction in a similar way to figure 4.

– 14 –
Figure 5. Witten’s Quantization procedure. Left: (a) depicts the cutting procedure of a general manifold
M into M1 and M2 . One can do similar cutting procedure for a three sphere S 3 in (b). Right: (a) shows
the inclusion of Wilson loop W and the cutting of the three manifold M. Wilson loop carrying representation
R of the gauge group pierce through Riemann surface Σ and leaves marked points on it. (b) shows that,
R
near the cut, M looks locally like Σ × .

• Canonically quantize CS theory on Σ and construct the physical Hilbert space HΣ . A path
integral over M1 is a quantum state |ψM1 i ∈ HΣ . Since the surface Σ∗ of ∂M2 has opposite
orientation of Σ, the physical Hilbert space on ∂M2 is just a dual space HΣ∗ and the path

integral over it is denoted as hψM2 | ∈ HΣ

∗.

• Heegaard splitting and Gluing: Let Every M, M1 , and M2 be compact, closed, con-
nected, orientable 3-manifolds and ∂M1 = ∂M2 = Σ = M1 ∩ M2 be a Riemann surface.
M admits Heegaard splitting (Σ, M1 , M2 ) such that, given an orientation reversing homeo-
morphism f : Σ → Σ, there is a gluing procedure leading to M = M1 ∪f M2 [17].

• Atiyah’s Axioms: A TQFT is a functor Z : nCob → H with respect to orientation preserv-

ing diffeomorphisms of Σ and M. This statement basically means that, we assign a Hilbert
spaces HΣ and HM to Rieman surface Σ and 3-manifolds M respectively. Functor Z is
involutory, i.e. Z(Σ∗ ) = Z(Σ)∗ and multiplicative, i.e. Z(Σ1 ∪ Σ2 ) = Z(Σ1 ) ⊗ Z(Σ2 ) [18].

• The homeomorphism f : Σ → Σ induces a homeomorphism Uf : HΣ → HΣ ∗ . U is a unitary

f
operator so that we can compute the partition function Z(M) = hψM2 | Uf |ψM1 i on M.

• Now insert the Wilson loops in M. On Σ is left mark points or punctures Pi with each
assigned a representation Ri of the gauge group Gk . This therefore allows us to quantize the
theory on Σ × R, and the theory is solved on evaluating the expectation value
Z
1
hWC1 ,R1 ...WCn ,Rn i = Da eiSCS [a] WC1 ,R1 ...WCn ,Rn . (4.12)
Z M

– 15 –
5 The Jones Polynomial

In this section, our goal is to get accustomed to the machinery we shall use in evaluating the
expectation value of the non-Abelian Wilson loops. While the concept associated with the topic in
this section is broad as an area of interest, all we need for our purpose is a mathematical concept
called the Jones polynomial. Our knot excursion shall therefore be brief.

5.1 Knots and Links

Here, we start with a basic overview of knot theory leading to the Jones polynomial.

• A knot is an embedding of a closed curve in R3 ; a circle is a trivial knot. A link is a finite

family of disjoint union of knots. This implies that a knot is a link with a single closed curve.
It is worth mentioning that every link is the closure of some braid.

• Knots are 3D objects but mostly studied through knot diagram obtained by projecting the
3D object onto a plane R2 just as depicted in figure 6. The non-trivial knots in figure 6 are
examples showing that there is no unique way of embedding a closed curve in R3 . Any closed
curve can be knotted in different ways having different number of crossings. This leaves a
question in knot theory: when are two knots, or perhaps two knot diagrams, equivalent?

Figure 6. The first two diagrams are non-trivial knots with 3 and 4 crossings respectively. The third
diagram is the Hopf link.

• Existence of orientation preserving homeomorphism between two knot diagrams may provide
equivalence of two knots. However, this is not always true and such homeomorphism is hard
to find in practice.

• Two links are equivalent or isotopic if they differ by a finite sequence of local sequence of
Reidemeister moves and an orientation- preserving homeomorphism of the plane. There are
three types of Reidemeister moves as shown in figure 7. Any homeomorphism of the plane
must preserve all crossing information.

Figure 7. Three types of Reidemeister moves

– 16 –
 • By orientation of a link, we means a choice of direction of trajectory around each component
of the link. A cross change can be done in a diagram of an oriented link. Such crossing is a
local modification of the type in figure 8. They are completely different from Reidemeister
moves.

Figure 8. Two types of crossing change

• Any link can be unknotted by a finite number of cross changes. This cross changes allows
for construction of link invariants. A link invariant assigns same object to links in the same
isotopy class. An example of link invariant is the Jones polynomial.

Basically, the Jones polynomial is an invariant of link. Of all different model, Jones polynomial[16]
is simpler to define in the light of Kauffman bracket polynomial. There are The Kauffman bracket
is easier computed through skein relations.

Definition. It is a function from unoriented link diagrams to Laurent polynomials in variable A

using the following rules:
D E
1. = 1, ; = trivial link.
D E
2. L ∪ = (−A2 − A−2 )hLi ; L = link diagram.
D E D E D E
3. =A + A−1 .

As an example, the computation of the bracket polynomial of Trefoil knot using skein relation is
shown in figure 9. The bracket polynomial is invariant under type II and type III Reidemeister
moves. Under type I move, the invariance is up to some pre-factors. This prefactor anomaly is as
a result writhe.

Definition. The writhe w(L) of a diagram L of an oriented link is the sum of the signs of the
crossing change defined according as the convention in figure 8.

w(L) is invariant under a type II or III Reidemeister move in L, but it does change by +1 or
−1 under type I Reidemeister move in L. This is in fact the reason why bracket polynomial is
not invariant due to some prefactors. It turns out that multiplication of bracket polynomial by
(−A)−3w(L) cancels out all the prefactors left by Type I Reidemeister moves. This leads to the
definition that the polynomial
X(L) = (−A)−3w(L) hLi, (5.1)
is an invariant of the oriented link L. The Jones polynomials is a special case with variable
A = q −1/4 .

– 17 –
 Figure 9. Bracket polynomial of Trefoil knot

Definition. The Jones polynomial V (L) of an oriented link L is a Laurent polynomial in variable
q 1/2 assigned to an oriented link L defined by

V (L) = (−A)−3w(L) hLiA−2 =q1/2 (5.2)

satisfying the following properties:

1. V(trivial knot) =1

2. if L is isotopic to L0 , then V (L) = V (L0 ); the converse is not necessarily true.

3. q −1 V (L+ ) − qV (L− ) + (q −1/2 − q 1/2 )V (L0 ) = 0, where L0 , L− and L+ are Skein relation in
Figure 10 showing identical oriented link diagrams but differ only at one crossing.

Figure 10. Skein relation

The Jones polynomial was discovered through a new representation of the Artin braid group. It
has proved very useful in that, it has strong feature of distinguishing one link from another. This
includes mirror images of link, which can be obtained for the Jones polynomial by replacing q with
q −1 . Thus far, we have learn that any link can be unknotted by a finite number of crossing and
that, the Jones polynomial allows us to evaluate the link invariant from the skein diagram.

– 18 –
6 Non-Abelian CS theory

Now, we shall return to Physics and embark on studying the Witten’s approach to solving non-
Abelian Chern-Simons theory. At the concluding part of section 4, we already highlighted the the
basic procedure for this section. We shall now discuss them step-by-step. The first non-trivial
step is the Quantization of CS theory on Σ × R. As we shall realize, the canonical approach for
evaluating the Wilson loops will be based on constructing the Hilbert space on a Riemann surface
Σ with punctures.

6.1 Quantization on Σ × R
We return back to the non-Abelian CS theory (3.8). In the euclidean theory, we can well write the
Lagrangian density as
Z  
k 2
L= ijk
 Tr ai (∂j ak − ∂k aj ) + ai [aj , ak ] (6.1)
8π M 3

For now, we shall ignore the Wilson loop, and canonically quantize CS theory. Let us denotes the
connection in the Σ-direction as ai=1,2 and a0 as the gauge connection in the R-direction. Towards
quantization, we recognize that there exist a natural gauge a0 = 0 on Σ × R. With this gauge, the
nonlinearity is tractable and the Lagrangian density reads
Z Z  
k ∂
L= ij
dt  Tr ai aj + a0 fij . (6.2)
4π R Σ ∂t

In this way, ai is a canonical conjugate of 4π  aj . This implies that, in CS theory without the Wilson
k ij

loops, components of gauge connection are canonical conjugate to one another. Furthermore, we
recognize that a0 is a Lagrangian multiplier that enforces a Gauss constraint fij = 0; as such, we
are left with the quantization of contained system. Constrained systems can be quantized in two
different but equivalent ways, namely,

• first canonically quantize and then enforces the Gauss constraint at the quantum level. In
this way, we have the commutation relation as

2πi
[abi (x), acj (y)] = ij δ bc δ 2 (x − y),
k

subject to the Gauss constraint ij fija = 0 on the wave solution.

• Impose the Gauss constraint ij fija = 0 classically and then quantize the resulting degrees of
freedom. So, we need first solve the constraint on Σ. This is equivalent to a problem of solving
for the moduli space of flat connections on Σ for which curvature vanishes. A connection is
expected to detail a way to do parallel transport in a principal bundle over Σ with structure
group Gk . Therefore, for any closed loop in Σ, a connection determines a holonomy, i.e.
the solutions of the constraint are parametrized by the holonomies of the gauge connection
around a cycle in Σ. For a genus g surface Σ, and a gauge group Gk , the dimension of the
moduli space of flat connection is (2g − 2) dim Gk .

• We are particularly interested in Σ = S 2 . However, there are no flat connection on S 2 , thereby

rendering quantization trivial with with just a unique state. As such, dim HΣ = 1.

– 19 –
 • For a genus g ≥ 1 surface Σg and gauge group SU (2)k , the dimension of the Hilbert space is
given by
k 
k + 2 g−1 X (j + 1)π 2−2g
  
dim HΣg = sin (6.3)
2 k+2
j=0

This is the Verlinde formula[19] arising as a consequence of the fact that, the modular S-
Matrix implementing modular transformation τ → − τ1 diagonalizes the fusion rule. A very
important feature of this formula is that, it is an integer and a finite polynomial in k.

6.1.1 Quantization on Σ with punctures

So far, we have quantized the CS theory on Σ without the Wilson loop. On inclusion of the
Wilson loop, the key insight is to recognise that the worldlines of particles will leave markpoints
(punctures) of static charges at points ξi on the Riemann surface Σ as the loops pierce through it.
Each point ξi corresponds to a representation Ri of the gauge group Gk . Quantization of CS theory
in the presence of Wilson loop is therefore up to finding the Hilbert space HΣ,ξi ,Ri associated to
the surface Σ with the choice of markpoint ξi corresponding to Ri . As the Wilson lines correspond
to static non-abelian charges, the Gauss constraint is therefore modified as
p
k ij c X
 fij (x) = δ 2 (x − ξn ) tc(n) , (6.4)
8π
n=1

where tc(r) , c = 1, ..., dim G are the generator of the gauge group associated with static external
charges placed at ξr . Following either of the earlier quantization approach result into rather quan-
tization issues resolved as follows:

• the large k limit, i.e. k → ∞ corresponds to the weak coupling limit of the CS theory, so
that we can do calssical analysis.

• Dirac quantization criterion applies to each fijc .

• for large k, Dirac quantisation cannot be applied alongside with non-trivial charge on the
right hand side of the Gauss constraint.

• possible resolution is for all charges to sum up to zero. This is indeed the case as equal
quantities of positive and negative charge is expected no matter how weak the coupling is.
As such, the net charge is zero. In non-Abelian theory, this implies that all of the charges
must be coupled with trivial representation of the gauge group.
Nn
• i.e. these are all trivial representation arising from the decomposition of i=1 Ri in the large
k limit and it is restricted to this.

• In the light of this resolution, the physical Hilbert space on Σ = S 2 with three marked points
corresponding to representations Ri , Rj and R̄k is spanned by the orthonormal states of all
possible fusion channels Vijk whose dimension is k Nijk .
P

In the spirit of quantization, there is a geometric construction of the Hilbert space according as the
Borel-Weil-Bott theorem. They constructed the Hilbert space as the space of holomorphic sections
of line bundle leading to a linear representation of G. In this formalism, all irreducible representation

– 20 –
arise from such kind of construction, an understanding that become useful in quantization of Chern-
Sirnons theory. The Hilbert space is given as
n
O
H(Σ;ξ1 ,R1 ;...;ξn ,Rn ) = RiG = (R1 ⊗ ... ⊗ Rn )G . (6.5)
i=1

• For n = 0, there is only the trivial representation on Σ and dimHΣ = 1.

• For n = 1, the Hilbert space is HΣ;ξ1 ,R1 = R1G , so that HΣ;ξ1 ,R1 = 1 if R is trivial and 0
otherwise.

• For n = 2, dim HΣ;ξ1 ,R1 ,ξ2 ,R2 = dim(R1 ⊗ R2 )G = 1 if R2 is dual R1 and 0 otherwise.

• For n = 3, the dim HΣ;ξi ,Ri ,ξj ,Rj ,ξk ,Rk = dim(Ri ⊗ Rj ⊗ Rk )G = Nijk . The Hilbert space is
spanned by the orthonormal states of all possible fusion channels.

• Let R be the representation of the gauge group G = SU (N ). With two Wilson loops, there
are four marked points, i.e. n = 4, corresponding to representations R, R, R̄ and R̄. There
are two distinct irreducible representations arising from decomposition R ⊗ R = E1 ⊕ E2 . E1
is the symmetric representation generated by 21 (ei ⊗ ej + ej ⊗ ei ), and E2 is the antisymmetric
representation generated by 21 (ei ⊗ ej − ej ⊗ ei ), so that we have

R ⊗ R ⊗ R̄ ⊗ R̄ = (E1 ⊕ E2 ) ⊗ (Ē1 ⊕ Ē2 ).

For SU(2), this implies that

dim HΣ;ξ1 ,R1 ,ξ2 ,R2 ,ξ3 ,R3 ,ξ4 ,R4 = dim (R ⊗ R ⊗ R̄ ⊗ R̄)G = 2, (6.6)

i.e. SU (2) is generated by E1 ⊗ Ē1 and there are only two possible fusion outcomes. This is
the central idea we shall use to explain skein relations in Jones polynomials.

6.2 Evaluation by TQFT

Before we embark on computation, we first establish some factorization properties which has po-
tential of reducing evaluation of many loops to that of a single loop. This will follow from our
earlier observation depicted in figure 5 and the TQFT axioms that follows it.

• Factorization I: Let M1 and M2 be two three manifolds and M = M1 #M2 be the

connected sum (#) of two M1 and M2 along S 2 . Following Atiyah’s axiom, we deduce a
factorization
Z(M ) · Z(S 3 ) = Z(M1 ) · Z(M1 ), (6.7)
where Z(X) denotes the partition function over any three manifold X inside which knots
may live. S 3 is the three sphere filling in S 2 ; it carries no knot and the partition function
over it is Z(S 3 ). (6.7) implies that
Z(M ) Z(M1 ) Z(M1 )
= · . (6.8)
Z(S 3 ) Z(S 3 ) Z(S 3 )

• Factorization II: Given any link L, living in a three manifold M, the expectation value of
L is given as
Z(M, L)
hLi = . (6.9)
Z(M )

– 21 –
 • Factorization III: Let C1 , ...., Cn be n unlinked, unknotted loops living on the three sphere
S 3 . Following (6.8) and (6.9), then
n
Y
hC1 , ..., Cn i = hCk i. (6.10)
k=1

6.2.1 Unknotting Procedure

We shall now present SU(2) Chern-Simons theory and its connection with the Skein relation. We
shall not repeat the entire story associating with figure 5 again, rather, we remind ourselves that
our goal is to evaluate the partition function Z(L) of a link L living on a three manifold M, from
which the expectation value hWL i can be computed. Witten derived the rule for unknotting the
link as follows:

• Embed a link L in a general three manifold M and let the link components (i.e. each Wilson
loop) carry fundamental representation R of the SU (2) gauge group. The link features a
crossing configuration (see figure 11a). Imagine inserting a small sphere around the crossing
configuration. The sphere would intersect the link at four marked points, so that it cuts M
into two pieces corresponding to ML and MR in Figure 11b.

Figure 11. Cutting procedure of three manifold M

• In this way, a part of the link L inside MR has a single crossing and the other part is contained
in ML . However, ML and MR shares same boundary Σ = S 2 .

• We already quantized the theory on the boundary Σ = S 2 . We therefore associate physical

Hilbert spaces HL and HR with the boundaries of ML and MR respectively with HL = HR ∗.

We know the Hilbert spaces are two dimensional according as (6.6). Feynman path integral
over ML and MR yields vectors hψML | ∈ HL and |ψMR i ∈ HR respectively.

• The partition function then gives Z(L) = hψML | Uf |ψMR i , where Uf : HL → HR is a unitary
operator induced by orientation preserving homeomorphism f used in the gluing procedure.

– 22 –
• Again, HL and HR are two dimensional. Since a linear combination of three vectors in 2D
vector space vanishes, then given |χMR i , |ϕMR i ∈ HR , there is a linear relation

α |ψMR i + β |χMR i + γ |ϕMR i = 0, (6.11)

and as a consequence,

α hψML | Uf |ψMR i + β hψML | Uf |χMR i + γ hψML | Uf |ϕMR i = 0,

αZ(L) + βZ(L0 ) + γZ(L00 ) = 0, (6.12)

where L0 and L00 are links associated with respective three manifolds M0 and M00 . While we
shall take M0 = M = M00 , both L0 and L00 carry worldline braiding different from that of L
as a consequence of replacing three MR with M0R and M00R respectively and application of
braiding rule.

Figure 12. Braiding operation of a section of Wilson link living in MR

• As shown in Figure 12, clockwise exchange of the first two marked points on the boundary of
M0R yield the link configuration LR , while the anticlockwise exchange yield the configuration
L00R . This clockwise and anticlockwise exchange are respectively implemented by braiding
unitary operator B −1 and B, so that state corresponding to LR and L00R are respectively given
as (
|ψMR i = B −1 |χM0R i = B −1 |χMR i
(6.13)
|ϕMR i = B |χM0R i = B |χMR i .

• (6.12) is further represented in Figure 13. Indeed, this diagram represents an expression
related to the Skein relation encountered in section 5 as a property of the Jones polynomial.

Figure 13. Skein-Recursion relation for links

– 23 –
• (6.12) is interpreted as follows. Considers three topologically equivalent closed contours
corresponding to three links whose plane projections are identical outside a disc, and respec-
tively have Skein relation inside the disc (see Figure 10 and 14). The expectation values
of the three topologically equivalent closed contours weighted with α, β and γ is zero, i.e.
αZ(L) + βZ(L0 ) + γZ(L00 ) = 0 for three links L, L0 and L00 .

Figure 14. Skein relation for full links plane projection

• Now take M = S 3 and denote the partition function of n unlinked, unknotted loops living
on the three sphere S 3 as Z(S 3 ; C1 , ..., Cn ). Following Figure 14, then (6.12) sum up to
αZ(S 3 ; C) + βZ(S 3 ; C 2 ) + γZ(S 3 ; C) = 0.
Using the factorization properties earlier outlined, then
Z(S 3 ; C) Z(S 3 ; C 2 ) Z(S 3 ; C)
α 3
+β 3
+γ =0
Z(S ; C) Z(S ; C) Z(S 3 ; C)
implies
α+γ
hCi = − . (6.14)
β
• For SU (N )k , a braiding operator B whose eigenvalues are
   
iπ(−N + 1) iπ(N + 1)
λ1 = exp , λ2 = − exp
N (N + k) N (N + k)
can be used to evaluate expectation value hCi with
 
2πi
α = − exp ,
N (N + k)
iπ(2 − N − N 2 ) iπ(2 + N − N 2 )
   
β = − exp + exp ,
N (N + k) N (N + k)
2πi(1 − N 2 )
 
γ = exp .
N (N + k)
 
• In fact, by introducing variable q = exp N2πi +k , the very property of Jones polynomial
 
−1/2
N/2
−q L+ q −q 1/2
L0 + q −N/2 L00 = 0, (6.15)
realizes the expectation value for the single loop as
α+γ q N/2 − q −N/2
hCi = − = 1/2 , (6.16)
β q − q −1/2
• While this computation is for a single Wilson loop, the same procedure generalizes to more
many complicated loops. This immediately suggest the interpretation of the partition func-
tion Z(M ; L, R) of the  )k Chern-Simons theory as nothing but the Jones polynomial,
 SU (N
evaluated at q = exp N +k , and the problem is solved.
2πi

– 24 –
Conclusion

Our study cut across many interesting physical ideas, each of which are very enlightening and
may worth studying independently. These include some progresses in planar condensed matter
physics leading to non-Abelian anyons, and then to finding a good topological quantum field the-
ory (TQFT) to describe those physics. While the studies in relation to understanding these rich
concepts is inexhaustive, we have been able to, at least, achieve a primary aim of recognizing
the 2 + 1-dimensional Chern-Simons gauge theory as a good example TQFT exhibiting most im-
portant features of the underlying physics of the quantum hall system and the non-Abelian anyons.

The study also exposed us to the realization that, there is really lots of stories and ideas about
quantum Chern-Simons invariants in topology, and physics. Indeed, we have learnt a very important
lesson, in that, given a link L = ∪Ci ⊂ S 3 in the representation Ri of the SU (N ) Chern-Simons
level k theory, the partition function,
Z  Z Y Z
3 ik 2
Z(S ; L, Ri ) = Da exp Tr(a × da + a × a × a) TrRi P exp a,
4π S 3 3 Ci
i
 
+k . This is Witten’s novel achievement.
is the Jones polynomial evaluated at variable q = exp N2πi
As such, it has provided another exchange between mathematics and physical studies.

Suffices to mention here that the Chern-Simons terms have been accommodated into some other
theories. Due to the gauge principle it exhibits, CS effective modification connects interesting areas
of physics studies such as particle physics, String Theory Loop Quantum Gravity and Cosmology.
There is so much more to groak.

Acknowledgments

I thank Richard Eager for making this topic available for my study and seminar work. My knowl-
edge of the topic improves greatly through some useful discussion with Ingmar Saberi.

References
[1] Edward Witten, Quantum field theory and the Jones polynomial, Comm. Math. Phys. 121(3) (1989)
351–399
[2] D. C. Tsui, H. L. Stormer, and A. C. Gossard, Two-Dimensional Magnetotransport in the Extreme
Quantum Limit, Phys.Rev.Lett. 48 (1982) 1559-1562
[3] Edwin Hall, On a New Action of the Magnet on Electric Currents, American Journal of Mathematics
2(3) (1879) 287–292
[4] R. B. Laughlin, Anomalous Quantum Hall Effect: An Incompressible Quantum Fluid with
Fractionally Charged Excitations, Phys. Rev. Lett. 50 ( 1983) 1395–1398
[5] J.M. Leinaas and J. Myrheim, On the theory of identical particles, Nuovo Cim B (1977) 37
[6] B. I. Halperin, Statistics of Quasiparticles and the Hierarchy of Fractional Quantized Hall States,
Phys. Rev. Lett. 52 ( 1984) 1583–1586
[7] Frank Wilczek, Magnetic Flux, Angular Momentum, and Statistics, Phys.Rev.Lett. 48 (1982)
1144–1146

– 25 –
 [8] Frank Wilczek, Quantum Mechanics of Fractional-Spin Particles, Phys. Rev. Lett. 49 (1982) 957–959
[9] F. D. M. Haldane, Fractional Quantization of the Hall Effect: A Hierarchy of Incompressible
Quantum Fluid States, Phys. Rev. Lett. 51 (1983) 605–608; Martin Greiter, Microscopic formulation
of the hierarchy of quantized Hall states, Physics Letters B 336 (1994) 48-53 .
[10] E. Artin, Theorie der Zöpfe, Abh. Math. Sem. Univ. Hamburg 4 (1926), 47-72, FdM51, 450; E. Artin,
Theory of Braids, Ann. of Math. 48, no. 1 (1947a) 101–126
[11] J.K. Slingerland and F.A.Bais, Quantum groups and non-Abelian braiding in quantum Hall systems,
Nuclear Physics B 612 (2001) 229–290 ; Simon Trebst et. al. A short introduction to Fibonacci anyon
models, Prog Theor Phys (2008) 176: 384-407
[12] A. Zee, Quantum hall fluids, Field Theory, Topology and Condensed Matter Physics, LNP 456 (1994)
99-153
[13] S. Chern, J. Simons, Characteristic forms and geometric invariants, Ann. of Math. 99 (1974) 48–69;
Daniel S. Freed, Remarks on Chern-Simons theory, Bull. Amer. Math. Soc. 46 (2009), 221
[14] Xiao-Gang Wen, Topological orders and edge excitations in fractional quantum Hall states, Advances
in Physics 44 (1995) 405-473
[15] E.Guadagnini et. al., Wilson lines in Chern-Simons theory and link invariants, Nuclear Physics B
330 (1990) 575
[16] V.F.R Jones, Polynomial invariant for knots via von Neumann algebras, Bull. Amer. Math. Soc.
(N.S.) 12 (1985) 1, 103-111 ; L. H. Kauffman State models and the Jones polynomial, Topology 26
(1987) 3, 395-407 ; L. H. Kauffman, Knots and Physics , World Scientific, 3 (2001)
[17] Poul Heegaard, Forstudier til en topologisk Teori for de algebraiske Fladers Sammenhang, Thesis
(1898) JFM 29.0417.02 ; Jesse Johnson, Notes on Heegaard Splittings, JesseJ
[18] Michael Atiyah, Topological quantum field theory, Publ. Math. IHES 68 (1989) 175-186; John C.
Baez, Quantum Quandaries: a Category-Theoretic Perspective, arXiv: quant-ph/0404040
[19] Erik Verlinde, Fusion rules and modular transformations in 2D conformal field theory, Nucl. Phys.
B300 (1988) 360; R. Dijkgraaf and E. Verlinde, Modular Invariance and the Fusion Algebra, Nucl.
Phys. B (Proc. Suppl.) 5B (1988) 87; M. Blau and G. Thompson, Derivation of the Verlinde Formula
from Chern-Simons Theory and the G/G model, Nucl.Phys. B408 (1993) 345

Complementary Reading
• David Tong, The Quantum Hall Effect, TIFR Infosys Lectures, arXiv:1606.06687
• Jiannis K. Pachos, Introduction to Topological Quantum Computation, CUP (2012)
• Chetan Nayak et. al., Non-Abelian Anyons and Topological Quantum Computation, Rev. Mod.
Phys. 80, 1083 (2008)
• W. B. Raymond Lickorish, An Introduction to Knot Theory, Springer (1997)

– 26 –