Topological Phases, Entanglement and Boson Condensation
Huan He
A Dissertation Presented to the Faculty of Princeton University in Candidacy for the Degree of Doctor of Philosophy
Recommended for Acceptance by the Department of Physics Adviser: Bogdan Andrei Bernevig
June 2019 c Copyright by Huan He, 2019. All rights reserved. Abstract
This dissertation investigates the boson condensation of topological phases and the entanglement entropies of exactly solvable models. First, the bosons in a “parent” (2+1)D topological phase can be condensed to obtain a “child” topological phase. We prove that the boson condensation formalism necessarily has a pair of modular matrix conditions: the modular matrices of the parent and the child topological phases are connected by an integer matrix. These two modular matrix conditions serve as a numerical tool to search for all possible boson condensation transitions from the parent topological phase, and predict the child topological phases. As applications of the modular matrix conditions, (1) we recover the Kitaev’s 16-fold way, which classifies 16 different chiral superconductors in (2+1)D; (2) we prove that in any layers of topological theories SO(3)k with odd k, there do not exist condensable bosons. Second, an Abelian boson is always condensable. The condensation formalism in this scenario can be easily implemented by introducing higher form gauge symmetry. As an application in (2+1)D, the higher form gauge symmetry formalism recovers the same results of previous studies: bosons and only bosons can be condensed in an Abelian topological phase, and the deconfined particles braid trivially with the condensed bosons while the confined ones braid nontrivially. We emphasize again that the there exist non-Abelian bosons that cannot be condensed. Third, the ground states of stabilizer codes can be written as tensor network states. The entanglement entropy of such tensor network states can be calculated exactly. The 3D fracton models, as exotic stabilizer codes, are known to have several features which exceed the 3D topological phases: (i) the ground state degeneracy generally increase with the system size; (ii) the gapped excitations are immobile or only mobile in certain sub-manifolds. In our work, we calculate, for the first time, the entanglement entropy for the fracton models, and show that the entanglement iii entropy has a topological term linear to the subregions’ sizes, whereas the topological phases only have constant topological entanglement entropies.
iv Acknowledgements
First and foremost, I would like to thank my advisor Prof. B. Andrei Bernevig for guidance and support. I have benefited enormously from his curiosity, dedication to physics and perseverance. He provides me with a complete freedom to work on the problems and topics that I am interested in, and more importantly, he never allows me to give up easily until we all find a satisfying answer. It is fair to say that rigorous and solid research is the trademark of our group, which I now completely agree is truly valuable. I will bear this spirit in mind for the rest of my career. During the journey, I am indebted to the help from many professors and friends who make me a better physicist. Among all of them, I owe the most to Prof Nicolas Regnault, Prof German Sierra and Yunqin Zheng. Many papers were not possible without their help and efforts. In fact, Nicolas and German are the two aspects of modern physicists. Nicolas’s numerical technologies are unbelievable. I still remember how I was amazed when he finished a new program and computation before I finished speaking my speculations. It turned out that I was wrong and he was correct. I am lucky to meet and learn from German. Every time I discussed with German, I would enjoy the aesthetic of his physics and formulas, which I have almost forgotten over the years of research. His formulas reminded me of my initial motivation of becoming a physicist: the simple and beautiful laws of nature. Yunqin has been my major collaborator since the end of my first year in Princeton. In fact, we have known each other since undergraduate (10 years ago from now). I would never forget all the time we spent together with chalks on blackboards, pens on papers, and Mathematica on screens. The PCTS program and our department certainly bring in brilliant and optimistic young physicists globally. I am grateful for the collaborations with the postdoctoral researchers in Princeton: Prof Curt von Keyserlingk, Prof Titus Neupert, Prof Barry Bradlyn, Prof Jennifer Cano, Prof Yi Li, Dr Abhinav Prem, Dr Lian Biao, Dr Yizhi v You and Dr Kiryl Pakrouski. I hope that they also enjoy discussing and working with me. It was joyful and also surprisingly efficient to work with Prof Jia-Wei Mei and Dr. Ji-Yao Chen. A powerful tensor network MATLAB program was delivered in only a few weeks over the summer. I need to thank Xue-Da Wen, Apoorv Tiwari, Peng Ye for analytically calculating entanglement entropies, and being reliable and helpful collaborators for this project. Particularly, Xue-Da is the person who I would like to discuss all the time because of his enthusiastic passion. It would be my fault if I did not mention my group members, who I have in- teractions with, for many enlightening and refreshing conversations and discussions: Yang-Le Wu, Aris Alexandradinata, Sanjay Moudgalya, Zhi-Jun Wang, Jian Li, Zhi- Da Song, Fang Xie. I also would like to thank the physicists that I met during confer- ences, visit, emails and Skype: Rui-Xing Zhang, Chen Fang, Hong Yao, Yuan-Ming Lu, Meng Cheng, Yang Qi, Yuan Wan, Xie Chen, Giuseppe Carleo, Alan Morningstar, Yi-Chen Huang, Juven Wang, Yidun Wan, Heidar Moradi, Wen Wei Ho, Tian Lan and many others. My life would be miserable if without my friends in Princeton: Jing-Yu Luo, Jie Wang, Zheng Ma, Jing-Jing Lin, Yu Shen, Wu-Di Wang, Zhen-Bin Yang, Jun Xiong, Liang-Sheng Zhang, Bin Xu, Zhao-Qi Leng, Tong Gao, Suerfu, Yin-Yu Liu, Jia-Qi Jiang, Si-Hang Liang, Xin-Ran Li, Xiao-Wen Chen, Bo Zhao, Jun-Yi Zhang, Xue Song, Xin-Wei Yu, Hao Zheng, Zi-Ming Ji, Ho Tat Lam, Wen-Li Zhao, Po-Shen Hsin, Xiao-Jie Shi, Chaney Lin and many many others. I appreciate Prof Duncan Haldane, Prof Waseem Bakr and Prof Silviu Pufu for be- ing my committee members, and Prof Robert Austin for leading me the experimental project.
vi Last but not the least, I appreciate the support from my parents during all these years. And most importantly, it is beyond my language to describe my gratitude for many years’ accompany with my love, Yina. This dissertation is dedicated to you.
vii To my love Yina
viii Contents
Abstract...... iii Acknowledgements...... v List of Tables...... xiii List of Figures...... xiv
1 Preliminaries1 1.1 Stabilizer Codes...... 2 1.2 Entanglement Entropy and Tensor Network States...... 4 1.3 Abelian Topological Field Theory...... 7 1.4 Modular Tensor Category...... 9
2 An Example: Toric Code Model 14 2.1 Stabilizer Code...... 14 2.2 Tensor Network State...... 16 2.3 Modular Tensor Category...... 18 2.4 BF Field Theory...... 20
3 Boson Condensation 23 3.1 Definitions and Assumptions...... 26
3.2 The Condensation Matrix Mab ...... 29 3.2.1 Proof that M commutes with T matrix of the A theory.... 31 3.2.2 Proof that M commutes with S matrix of the A theory.... 31
ix 3.3 The Modular Tensor Category after condensation...... 33 3.4 Simple currents...... 36 3.4.1 Introduction to simple currents...... 36 3.4.2 Simple current condensation...... 38 3.5 One confined particle theories...... 43 3.6 Formalism and implementation...... 44 3.6.1 Solutions for M ...... 45 3.6.2 Solutions for n ...... 48 3.6.3 The modular matrices of the new theory...... 51 3.7 Layer constructions and uncondensable bosons...... 54
3.7.1 Theories with Zm-graded condensations...... 55 3.7.2 Theories with Z-fold way: Fibonacci TQFT...... 60 3.8 Conclusions...... 63
4 No-Go Theorem for Boson Condensation in Topologically Ordered Quantum Liquids 65 4.1 First No-Go Theorem...... 66 4.2 Example (i): Multiple layers of Fibonacci...... 69
4.3 Example (ii): Single layer of SO(3)k ...... 71
4.4 Example (iii): Multiple layers of SO(3)k ...... 74 4.5 Summary...... 74
5 Abelian Boson Condensation in Field Theory 76 5.1 Abelian Boson Condensation Formalism...... 77 5.2 Condensations in K-Matrix Chern-Simons Theories...... 81 5.2.1 Condensable Condition...... 82 5.2.2 Confinement/Deconfinement...... 84
x 6 Fracton Models, Tensor Network States and Their Entanglement Entropies 87 6.1 Stabilizer Code Tensor Network States...... 91 6.1.1 Notations...... 91 6.1.2 Stabilizer Code and TNS Construction...... 94 6.1.3 TNS Norm...... 98 6.1.4 Transfer Matrix...... 100 6.2 Entanglement properties of the stabilizer code TNS...... 102 6.2.1 TNS as an exact SVD...... 102 6.2.2 Summary of the results...... 107 6.3 3D Toric Code...... 109 6.3.1 Hamiltonian of 3D Toric Code Model...... 109 6.3.2 TNS for 3D Toric Code...... 111 6.3.3 Concatenation Lemma...... 115 6.3.4 Entanglement...... 117 6.3.5 Transfer Matrix as a Projector...... 121 6.3.6 GSD and Transfer Matrix...... 124 6.4 X-cube Model...... 128 6.4.1 Hamiltonian of X-cube Model...... 129 6.4.2 TNS for X-cube Model...... 131 6.4.3 Concatenation Lemma...... 134 6.4.4 Entanglement...... 136 6.4.5 Transfer Matrix as a Projector...... 141 6.4.6 GSD and Transfer Matrix...... 142 6.5 Haah Code...... 146 6.5.1 Hamiltonian of Haah code...... 146 6.5.2 TNS for Haah Code...... 148
xi 6.5.3 Entanglement Entropy for SVD Cuts...... 154 6.5.4 Entanglement Entropy for Cubic Cuts...... 161 6.6 Conclusion and Discussion...... 169
A Appendix for Boson Condensation 172 A.1 Quantum dimensions of A and T ...... 172
P r A.1.1 Proof of da = r∈T nadr ...... 172 1 P r A.1.2 Proof of dr = q a∈A nada ...... 173 A.2 Chiral algebra...... 175 A.3 Condensations in SU(2) CFTs...... 181
A.3.1 SU(2)16 ...... 184
A.3.2 SU(2)28 ...... 186
B Appendix for Uncondensable Bosons 189 B.1 No-go theorem with Abelian sector...... 189
B.2 Proof for Example (iii), Multiple layers of SO(3)k ...... 191 B.3 General constraints on boson condensation...... 194
C Appendix for Tensor Network States and Fracton Models 197 C.1 Proof for the Concatenation Lemma for the 3D Toric Code Model.. 197 C.2 Proof for the Concatenation Lemma for the X-cube Model...... 200 C.3 GSD for the X-cube Model...... 204
Bibliography 208
xii List of Tables
xiii List of Figures
1.1 A link in (2+1)D space...... 8
2.1 The Hamiltonian terms of the 2D toric code model. Panel (a) is Av which is a product of 4 Z operators around the vertex v, and Panel
(b) is Bp which is a product of 4 X operators around the plaquette p. 15 2.2 The TNS for the 2D toric code model on a square lattice. On each bond, we associate a projector g tensor, and on each vertex, we as- sociate a local T tensor. The connected lines are contracted virtual indices. The lines with arrows are the physical indices...... 17 2.3 Left panel: The line operator in the first equation of Eq. (2.12). The blue line is the path for C. Right panel: The line operator in the second equation of Eq. (2.12). The blue line is the path for C˜ on the dual lattice. The red dots represent the excitation created. W (C) creates two excitations at the ends of C which have −1 eigenvalues of ˜ ˜ ˜ Av. W (C) creates two excitations at the ends of C which have −1
eigenvalues of Bp...... 19
xiv 4.1 Tunneling processes mediated by an anyon condensate. The gray re- gion is a phase in which a boson B is condensed. a) Vertex of a boson
B that localizes a zero mode of anyon ai. In the condensed phase, B can be converted into an identity particle world line (not shown). By the axioms of anyon condensation, processes a) and b) are equivalent,
i.e., B can be converted into ai by tunneling through the condensate. 67
4.2 Quantum dimensions and bosons (blue columns) for SO(3)k theories with a) k = 13 and b) k = 103. These are the smallest k, for which
SO(3)k contains two and four bosons, respectively. Indicated are also the ranges I–III defined in Eq. (4.10). The maximum quantum dimen- sion coincides with the boundary between range I and II in Eq. (4.10). For instance, to apply the no-go theorem to the j = 5 boson in a),
choose Ij=5 = {j = 2} and use that d5 ≈ 3.6 is smaller than d2 ≈ 4.2. 72
5.1 An illustration of integer current and Hodge dual in 3 dimensional
space. In Panel (a), jS1 is the integer which only has support on the loop S1. It is a 1-form represented by the red arrow. Its Hodge dual
1 ?jS1 is represented by the red square penetrated by S , which is a 2-
form perpendicular to jS1 . In Panel (b), jS is the integer 2-current
which only has support on the sphere S. In 3 dimensions, ?jS is then a 1-form going out S denoted by the red dot...... 79
6.1 Examples of TNS lattice wave functions in 1D and 2D. Each node is a tensor whose indices are the lines connecting to it. The physical indices - of the quantum Hilbert space - are the lines with arrows, while the lines without any arrows are the virtual indices. Connected lines means the corresponding indices are contracted. Panel (a) is an MPS for 1D systems. Panel (b) is a PEPS on a 2D square lattice...... 88
xv 6.2 An illustration of the TNS gauge in MPS. (a) A part of an MPS. A1
and A2 are two local tensors contracted together. (b) We insert the
identity operator I = UU −1 at the virtual level - it acts on the virtual
bonds. The tensor contraction of A1 and A2 does not change. (c) We −1 ˜ ˜ further multiply U with A1 and U with A2, resulting in A1 and A2
respectively in Panel (d). The tensor contraction of A1 and A2 is the ˜ ˜ same as the tensor contraction of A1 and A2. The TNS does not change as well. Similar TNS gauges also appear in other TNS such as PEPS. 93 6.3 (a) A plane of TNS on a cubic lattice. (b) TNS on a cube. The lines with arrows are the physical indices. The connected lines are the contracted virtual indices, while the open lines are not contracted. On each vertex, there lives a T tensor, and on each bond, we have a projector g tensor...... 94 6.4 Transfer matrix (red dashed square) of a 1D MPS. The connected lines are the contracted virtual indices. The connected arrow lines are the contracted physical indices. The MPS norm (or any other quantities) can be built using the transfer matrix. Higher dimensional transfer matrices are similarly defined for TNS on a cylinder or a torus, by contracting in all directions except one. This leads to a 1D MPS with a bond dimension exponentially larger than the TNS one...... 100
6.5 The Hamiltonian terms of the 3D toric code model. Panel (a) is Av
which is a product of 6 Z operators, and Panel (b) is Bp which is a product of 4 X operators. The circled X and Z represent the Pauli matrices acting on the spin-1/2’s. The toric code Hamiltonian includes
Av terms on all vertices v and Bp terms on all plaquettes p...... 110 6.6 Contraction of two local T tensors in the z-direction. We emphasize that there is no projector g tensor in this figure...... 115
xvi 6.7 The splitting of tensors near the entanglement cut...... 118
6.8 The Hamiltonian terms of the X-cube model: (a) Av,yz, (b) Av,xy, (c)
Av,xz and (d) Bc. The circled X and Z represent the Pauli matrices acting on the physical spin-1/2’s...... 128 6.9 Figures for several regions A for which we calculate the entanglement
entropies. (1) Region A is a cube of size lx × ly × lz. (2) Region A is a
0 0 0 cube of lx × ly × lz with a hole of size lx × ly × lz in it. (3) Region A is
0 a cube of size lx × ly × lz and a small cube of height lz on top of it. (4)
Region A is a cube of size lx × ly × lz carved on top of it a small cube
0 of height lz...... 136 6.10 Derivations for the first equation in Eq. (6.107). The rest two equations can be proved similarly. In the first step, the physical X operators can be transferred to the virtual level by using Eq. (6.36), and in the third step, all the virtual X operators are exactly canceled in pairs (dashed red rectangles in the third figure) due to Eq. (6.87)...... 141 6.11 Examples for the X-cube TNS in a xy-plane, obtained by acting ˜ ˜ WZ [Cz,x] and WZ [Cz,y] on the constructed TNS. The intersection of ˜ ˜ one WZ [Cz,x] operator and one WZ [Cz,y] operator with the xy-plane is only one Pauli Z operator, i.e., the circled blue Z in this figure.... 144 ˜ ˜ 6.12 We act WZ [Cx,y] and WZ [Cy,x] operators on Panel (a) in Fig. 6.11. Hence, we have four TNSs in this xy-plane that can be related to each other. See the text for the explanations...... 144 6.13 Tensor contraction for the Haah code TNS. (a) The lattice size is 2 × 3 × 3. (b) The lattice size is 3 × 3 × 3...... 148 6.14 Region A contains all the spins connecting with l − 1 T tensors which are contracted along z direction. The figure shows an example with l = 4...... 155
xvii 6.15 Region A contains all the spins connecting with T tensors which are contracted in a “tripod-like” shape, where three legs extend along x, y, z directions. There are three legs extending along x, y, z direc- tions respectively. In general, three legs can have different length, each
with lx − 1, ly − 1, lz − 1 cubes along three directions. This figure shows
an example where lx = ly = lz = 3...... 155
6.16 Transferring the Pauli X operators of the Bc operator from the region A (a) to the region A¯ (b)...... 162
xviii Chapter 1
Preliminaries
The topological phase has been one of the central topics in condensed matter physics for the past decades since the discovery of integer and fractional quantum Hall effects[1,2]. The definition of topological phases has been evolving over the years of research[3,4,5,6,7,8,9]. It generally refers to the gapped phases of matter which ground states are different due to the boundary conditions. For instance, the ground state degeneracy can vary when the system is on torus or sphere[4], or the boundary can support protected gapless modes when the system is on a disk[10, 11, 12, 13, 14, 15, 16]. Another significant signature is that the bulk of the topological phases supports excitations which have nontrivial braiding statistics and topological spins[17, 18, 19, 20] than those of bosons and fermions. They are dubbed as “anyons”. In this dissertation, we mostly focus on the topological phases with anyons. This dissertation studies two aspects of topological phases in 2 dimensions and 3 dimensions: boson condensations[21] and entanglement entropy[22, 23]. It will be based on several different languages for topological phases, namely, stabilizer codes, topological field theories and modular tensor categories. The stabilizer codes and topological field theories provide the exactly solvable lattice Hamiltonian and field
1 theory models respectively. The modular tensor category, on the other hand, “for- gets” the specific models and details such as the lattice or the couplings, but only “remembers” the universal data such as the fusion rules and the braiding statistics of the gapped excitations. In this chapter, each of the languages will be briefly re- viewed to provide the foundations for the rest of the dissertation. In Chapter2, the 2 dimensional toric code model will be described in each of the languages.
1.1 Stabilizer Codes
The stabilizer code refers to a class of exactly solvable Hamiltonian models specified as follows. The key ingredient is the notion of qubit, which is nothing but a 2-level
1 system. For instance, a qubit can be a spin- 2 . On the qubit, the operators can be generated by the Pauli X or Pauli Z operators satisfying anti-commutation relations:
0 1 1 0 XZ = −ZX,X = ,Z = . (1.1) 1 0 0 −1
Conventionally the Pauli Z basis is preferred:
1 0 |0i = , |1i = . (1.2) 0 1
Hence, Z|0i = |0i,Z|1i = −|1i,X|0i = |1i,X|1i = |0i. (1.3)
The stabilizer code is a many-qubit system, and the total Hilbert space is defined as a tensor product of the Hilbert spaces of each qubit. Hence, the Pauli operators on different sites commute:
[Xi,Zj] = 0, [Xi,Xj] = 0, [Zi,Zj] = 0, ∀ i 6= j. (1.4) 2 where the subscripts label different qubits. Each of the Hamiltonian terms is a product of Pauli X and Pauli Z operators. The Hamiltonian is exactly solvable because all local Hamiltonian terms are required to be commutative with each other. One simple example for the stabilizer codes is a 3-qubit system, and its Hamiltonian is:
H = −Z1Z2Z3 − X1X2 − X2X3 (1.5)
Note that all three terms are commutative:
[Z1Z2Z3,X1X2] = 0, [Z1Z2Z3,X2X3] = 0, [X1X2,X2X3] = 0. (1.6)
These commutation relations imply that:
[H,Z1Z2Z3] = 0, [H,X1X2] = 0, [H,X2X3] = 0. (1.7)
The eigenstates of the Hamiltonian are thus also the eigenstates of the local Hamil- tonian terms. Particularly, the ground state satisfies:
Z1Z2Z3|GSi = |GSi,X1X2|GSi = |GSi,X2X3|GSi = |GSi. (1.8)
The ground state wave function, in terms of the Pauli Z basis, contains all the con- figurations satisfying the first equality of Eq. (1.8):
1 |GSi = (|000i + |110i + |101i + |011i) (1.9) 2
The ground state wave function is much simplified if we use Pauli X basis instead:
1 |GSi = √ (|000ix + |111ix) , (1.10) 2
3 i.e. a GHZ state.[24] Another example for stabilizer codes, extremely useful in the topological phases, is the toric code model on a 2 dimensional lattice[19]. We defer the introduction of this model to Chapter2. There we will thoroughly describe this model in each of the languages.
1.2 Entanglement Entropy and Tensor Network
States
The entanglement (von Neumann) entropy for a wave function |ψi is defined as:
S(A) = TrA (ρA ln (ρA)) (1.11)
where ρA is the reduced density matrix for the subregion A by tracing out A’s com- plement A¯:
ρA = TrA¯|ψihψ|. (1.12)
The eigenvalues of the ρA are commonly referred as the entanglement spectrum. An equivalent method to calculate the entanglement entropy is to perform the singular value decomposition to the wave function coefficients M:
X |ψi = Mij|iiA ⊗ |jiA¯, ij (1.13) M = UΛV †.
¯ where |iiA and |jiA¯ form the orthonormal basis for the subregion A and A respectively. U, V are unitary matrices and Λ is a diagonal matrix containing the singular values. It is easy to show that the eigenvalues of the reduced density matrix are the square of the singular values.
4 A breakthrough was the conjecture of the area law[25]: the ground states of a local and gapped Hamiltonian has entanglement entropy linear to the subregion’s perimeter length/area: S(A) ∼ Area (∂A) (1.14) where S(A) is the entanglement for the subregion A and Area (∂A) refers to the A’s perimeter length/area. For instance, in 1 dimension, the entanglement entropy of the gapped local Hamiltonian is proved to be[26]:
S(A) ∼ Const. (1.15) since in 1 dimension, the subregion A is a line segment, and ∂A only contains two end points. Further, the entanglement entropy and spectrum, measured using the ground state wave functions, serve as the smoking gun to numerically and theoretically distinguish 2 dimensional topological phases. Two results need to be highlighted:
1. For a topological phase, the entanglement entropy scales as:
S(A) = α Area (∂A) − γ (1.16)
where the constant γ is a universal constant independent of the subregion’s size[23, 22]. “Universal” means that γ is the same at any point in the gapped phase. By computing γ from a microscopic model, most of the topological phases are eliminated, and only few of them can be the candidates which require further investigation.
2. In fractional quantum Hall systems, the entanglement spectrum degeneracy provides the “fingerprint” to identify the different quantum Hall states[27]. The
5 entanglement spectrum degeneracy is argued to be the same as the conformal field theories on the fractional quantum Hall edge[28].
Inspired by the area law, an efficient ansatz for the ground state wave function was proposed and widely applied both numerically and analytically. This set of ansatz is known as tensor network states. To name a few, they are matrix product state (MPS)[29], projected entangled pair states (PEPS)[30], multi-entanglement renor- malization ansatz[31, 32]. In this section, we briefly introduce MPS in 1 dimension. PEPS is the immediate generalization of MPS in higher dimensions and will be heav- ily used in Chapter6. Suppose we have an MPS with periodic boundary condition whose local tensor is A:
X X As1 As2 ...AsN |s , s , . . . , s i (1.17) i1,i2 i2,i3 iN ,i1 1 2 N s1,s2,...,sN i1,i2,...,iN
where the tensor A has one physical index sn of dimension d and two virtual indices
in of dimension D. The digram representation for the tensor A at the n-th site is:
(1.18)
where each line represents an index of the tensor. The physical index sn is dis- tinguished from the virtual indices by the arrows. Diagrammatically, the MPS is represented as:
(1.19)
6 The connected lines imply the contraction over the two indices. Note that the entan- glement of such an MPS is upper bounded by the virtual bond dimension D:
S ≤ 2 ln(D), (1.20)
because the rank of the reduced density matrix or the rank of the singular value matrix is upper bounded by D2 where the number 2 comes from the two cuts.
1.3 Abelian Topological Field Theory
In this section, we review a class of topological field theories, K-matrix Chern-Simons theories[33], which can describe any Abelian topological phases. The Lagrangian of a K-matrix Chern-Simons theory is:
K L = IJ µνλa ∂ a (1.21) 4π Iµ ν J,ρ
where gauge fields aI are U(1) gauge fields, and the matrix KIJ is symmetric and all elements are integers. In a differential geometry language, it can also be written as differential forms and exterior derivatives.
K L = IJ a ∧ da (1.22) 4π I J
When the diagonal elements of the matrix K are all even integers, this type of K- matrix Chern-Simons theories is called “bosonic”. For a K-matrix whose diagonal elements have odd integers, it is refereed as “fermionic”. In this dissertation, we mostly focus on the bosonic K-matrix Chern-Simons theories.
7 Figure 1.1: A link in (2+1)D space.
K-matrix Chern Simons theories have gauge invariant loop operators:
U (S1) = exp i n a , n ∈ (1.23) {n} ˛ I I I Z S1 where S1 is a closed loop in (2+1)D spacetime and is also the worldline of the corre- sponding pair of excitations. The linking correlation function of two loop operators can be calculated by the Green function method[34]:
1 1 −1 hU{n}(S1 )U{m}(S2 )i = exp i 2πm · K · n (1.24)
1 1 where loops S1 and S2 form a link as in Fig. 1.1, and “·” is the matrix multiplication. The physical interpretation of the linking correlation functions is that they are the
1 1 braiding statistics of the two excitations represented by U{n}(S1 ) and U{m}(S2 ). In the modular tensor category, this linking correlation is encoded in the modular S matrix. The topological spin for each loop operator is: