cond-mat0604506/SKW.tex
1: %version: Sato, June 3, 2006
2: %version: Sato, Apr 21, 2006
3: %version:   Wu, Apr 15, 2006
4: %version: Sato, Apr 03, 2006
5: %version:   Wu, Mar 19, 2006
6: %version: Sato, Mar 09, 2006
7: 
8: % ****** Start of file apssamp.tex ******
9: %
10: %   This file is part of the APS files in the REVTeX 4 distribution.
11: %   Version 4.0 of REVTeX, August 2001
12: %
13: %   Copyright (c) 2001 The American Physical Society.
14: %
15: %   See the REVTeX 4 README file for restrictions and more information.
16: %
17: % TeX'ing this file requires that you have AMS-LaTeX 2.0 installed
18: % as well as the rest of the prerequisites for REVTeX 4.0
19: %
20: % See the REVTeX 4 README file
21: % It also requires running BibTeX. The commands are as follows:
22: %
23: %  1)  latex apssamp.tex
24: %  2)  bibtex apssamp
25: %  3)  latex apssamp.tex
26: %  4)  latex apssamp.tex
27: %
28: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
29: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
30: % Some other (several out of many) possibilities
31: %\documentclass[preprint,aps]{revtex4}
32: %\documentclass[preprint,aps,draft]{revtex4}
33: %\documentclass[prb]{revtex4}% Physical Review B
34: \usepackage{graphicx}% Include figure files
35: \usepackage{dcolumn}% Align table columns on decimal point
36: \usepackage{bm}% bold math
37: %\nofiles
38: \begin{document}
39: \title{Braid Group, Gauge Invariance and Topological Order}
40: % Force line breaks with \\
41: \author{Masatoshi Sato
42: }%
43: \email{msato@issp.u-tokyo.ac.jp}
44: \author{Mahito Kohmoto}
45: \affiliation{%
46: Institute for Solid State Physics,
47: The University of Tokyo,
48: Kashiwanoha 5-1-5, Kashiwa, Chiba, 277-8581, Japan
49: }%
50: \author{Yong-Shi Wu}
51: \affiliation{
52: Department of Physics, University of Utah, Salt Lake City, Utah 84112
53: }%
54: \date{\today}% It is always \today, today,
55:              %  but any date may be explicitly specified
56: \begin{abstract}
57: Topological order in two-dimensional systems is studied 
58: by combining the braid group formalism with a gauge 
59: invariance analysis. We show that flux insertions (or 
60: large gauge transformations) pertinent to the toroidal 
61: topology induce automorphisms of the braid group, giving 
62: rise to a unified algebraic structure that characterizes 
63: the ground-state subspace and fractionally charged, anyonic 
64: quasiparticles. Minimal ground state degeneracy is derived 
65: without assuming any relation between quasiparticle charge 
66: and statistics. We also point out that noncommutativity 
67: between large gauge transformations is essential for the 
68: topological order in the fractional quantum Hall effect.  
69: \end{abstract}
70: 
71: \pacs{05.30.Pr, 71.10.Hf, 71.10.Pm, 73.43.-f}% PACS, the Physics and Astronomy
72:                              % Classification Scheme.
73: %\keywords{Suggested keywords}%Use showkeys class option if keyword
74:                               %display desired
75: \maketitle
76: 
77: In contrast to the expectation from classical thermodynamics,
78: physics at absolute zero temperature is very rich because of novel 
79: effects of quantum fluctuations. In recent years it has 
80: become increasingly clear that in a wide class of 
81: two-dimensional strongly correlated many-body systems,
82: transition driven by a non-thermal parameter may occur 
83: at zero temperature to a novel phase which can not be 
84: described by usual spontaneous symmetry breaking and 
85: order parameters. The characteristic signature of the 
86: novel phase is a finite ground state degeneracy that 
87: depends on the topology of the system; accompanying with 
88: it are charge fractionalization (with respect to that of 
89: the constituent particles) and/or fractional statistics
90: of the quasiparticles. The first known example is the 
91: Laughlin state \cite{Laughlin} for the fractional quantum 
92: Hall (FQH) effect, with electron filling factor $\nu=1/n$ 
93: with $n$ odd. Soon after, it was realized that in this 
94: phase the ground state is $n$-fold degenerate on a 
95: cylinder \cite{TW} or on a torus \cite{WN}, while is 
96: known nondegenerate on a sphere. Actually it is the 
97: ground state degeneracy that is responsible for the 
98: fractional quantization of Hall conductance \cite{TW,NTW} 
99: and dictates the fractional charge $e*=e/n$ \cite{WHK} 
100: and the anyon statistics $\theta=\pi/n$ \cite{WN,Einarsson} 
101: of the quasiparticles. This type of new order is dubbed 
102: as topological order \cite{Wen}. In recent years more 
103: systems, including bosonic ones or at zero magnetic 
104: field, are identified to possess topological order 
105: \cite{Wen4,WZ2,RS91,Sent00,Moes01,Misg02,Bal02,Motr02,Free04,DST}. 
106:  
107: In the study of topological order, a central issue is how 
108: to characterize or classify topological orders. Previously
109: there has been the idea \cite{TW,WHK} that the topology 
110: dependent ground state degeneracy seems to be dictated by 
111: an (emergent) discrete symmetry. But the latter was never 
112: identified explicitly, except being $Z_n$ for the Laughlin 
113: states. Another important issue is how to understand the 
114: relationship between ground state degeneracy and charge 
115: fractionalization and/or quasiparticle statistics. A 
116: puzzling fact is that different patterns have 
117: appeared in investigations of various systems. For example, 
118: it was concluded \cite{WN} for the FQH systems that on a 
119: surface with non-zero genus $g$, the appearance of a 
120: fractional $\theta=\pi m/n$ statistics, with $m$ and $n$ 
121: co-primes, requires $n^g$-fold degenerate ground states, 
122: confirmed by the braid group analysis \cite{Einarsson,HKW} 
123: and by the effective Chern-Simons theory as well 
124: \cite{Wen2,WDF}. On the other hand, in a recent paper \cite{OS} 
125: it was shown, by using a gauge invariance argument \cite{WHK}, 
126: that charge fractionalization with $e^{*}=ep/q$, with $p$ 
127: and $q$ co-primes, requires a ground state degeneracy $q^{2g}$ 
128: if the quasiparticles are ordinary bosons and fermions, 
129: while the FQH ground state degeneracy is known to be only 
130: $q^g$-fold \cite{WN,HKW}.  
131: 
132: In this letter, we start with a reexamination of the 
133: interplay between charge fractionalization and quasiparticle 
134: anyon statistics, if they coexist, in constraining ground 
135: state degeneracy. This has been studied only for the Laughlin 
136: states and their variants \cite{WN,HKW}, where fractional 
137: charge and anyon statistics are known to be closely related 
138: to each other \cite{ASW}. Below we shall derive minimal 
139: ground state degeneracy without assuming any relation 
140: between quasiparticle charge and statistics. A bonus of 
141: our reexamination is the identification of the discrete 
142: topological symmetry algebra that underlies the ground 
143: state degeneracy, which can be used to classify topological 
144: orders that support Abelian anyonic quasiparticle excitations.  
145:  
146: We will start with the braid group formalism 
147: \cite{Wu,Einarsson,HKW2} for fractional statistics. Consider 
148: for $N$ quasiparticles in a toroidal system with size 
149: $L_x\times L_y$. The braid group generators \cite{Birman}
150: consist of $\sigma_i$ $(i=1,\cdots, N-1)$, which exchanges the 
151: $i$th and $(i+1)$th particles clockwise without 
152: enclosing any other quasiparticle, and of $\tau_i$ and 
153: $\rho_i$ $(i=1,\cdots, N)$, which represent moving the 
154: $i$-th quasiparticle along a loop on the torus in 
155: $x$- and $y$-direction, respectively. (See 
156: Fig.\ref{fig:braidloop}.) Define operators 
157: $A_{i,j}$ and $C_{i,j}$ as 
158: \begin{eqnarray}
159: A_{j,i}=\tau_j^{-1}\rho_i\tau_j\rho_i^{-1},
160: \quad
161: C_{j,i}=\rho_j^{-1}\tau_i\rho_j\tau_i^{-1},
162: \end{eqnarray}
163: where $1\le i<j\le N$. The exchange operators 
164: $\sigma_i$ satisfy the 
165: following relations,
166: \begin{eqnarray}
167: &&\sigma_k\sigma_l=\sigma_l\sigma_k,
168: \quad 1\le k\le N-3, \quad |l-k|\ge 2,
169: \nonumber\\ 
170: &&\sigma_k\sigma_{k+1}\sigma_k
171: =\sigma_{k+1}\sigma_k\sigma_{k+1}, 
172: \quad 1\le k \le N-2,  
173: \nonumber\\
174: && \tau_{i+1}=\sigma_{i}^{-1} \tau_i \sigma_i^{-1},
175: \quad 
176: \rho_{i+1}=\sigma_i \rho_i \sigma_i,
177: \nonumber\\
178: && \tau_1\sigma_j=\sigma_j\tau_1, 
179: \quad 
180: \rho_1\sigma_j=\sigma_j\rho_1, \quad 
181: \sigma_i^2=A_{i+1,i}, 
182: \label{eq:braid1}
183: \end{eqnarray}
184: where $1\le i\le N-1$ and $2\le j\le N-1$.
185: And $\tau_i$'s and $\rho_i$'s satisfy
186: \begin{eqnarray}
187: &&A_{m,l}\tau_k=\tau_k A_{m,l},
188: \quad
189: A_{m,l}\rho_k=\rho_k A_{m,l}
190: \nonumber\\
191: &&\tau_i \tau_j=\tau_j \tau_i, 
192: \quad
193: \rho_i\rho_j=\rho_j\rho_i,
194: \nonumber\\
195: &&C_{j,i}=(\tau_i\tau_j)A_{j,i}^{-1}(\tau_j^{-1}\tau_i^{-1}),
196: \quad
197: A_{j,i}=(\rho_i\rho_j)C_{j,i}^{-1}(\rho_j^{-1}\rho_i^{-1}),
198: \nonumber\\
199: &&C_{j,i}=(A_{j,j-1}^{-1}\cdots A_{j,i+1}^{-1})A_{j,i}^{-1}
200: (A_{j,i+1}\cdots A_{j,j-1}),
201: \nonumber\\
202: &&\tau_1\rho_1\tau_1^{-1}\rho_1^{-1}
203: =A_{2,1}A_{3,1}\cdots A_{N-1,1}A_{N,1},
204: \label{eq:braid2}
205: \end{eqnarray}
206: where $1\le k <l<m\le N$ and $1\le i<j\le N$.
207: \begin{figure}[h]
208: \begin{center}
209: \includegraphics[width=7cm]{braidloop.eps}
210: \caption{a) An exchange of quasiparticles; 
211: b) Translations along non-shrinkable loops on 
212: the torus.}
213: \label{fig:braidloop}
214: \end{center}
215: \end{figure}
216: 
217: Let us now assume that the quasiparticles have a fractional 
218: charge $e^{*}=(p/q)e$, where $p$ and $q$ are mutually prime 
219: integers, and consider an adiabatic insertion of flux 
220: $2\pi/e$ through one of the holes of the torus. If the 
221: adiabatic flux insertion induces an infinitesimal electric 
222: field in $x$-direction, the process can be realized by 
223: a large gauge transformation $U_x$, in which the 
224: $x$-component of the gauge field changes from $A_{x}=0$ 
225: to $A_{x}=2\pi/eL_x$. After the large gauge transformation, 
226: the gauge potential $A_x=2\pi/eL_x$ will give rise to an 
227: Aharanov-Bohm phase $e^{-2\pi i p/q}$ when we apply 
228: $\tau_i$. Therefore, we obtain
229: \begin{eqnarray}
230: U_x\tau_i=e^{-2\pi ip/q}\tau_i U_x. 
231: \label{eq:ux1}
232: \end{eqnarray}
233: On the other hand, because $\sigma_i$ and $\rho_i$ do not 
234: encircle the flux, we have
235: \begin{eqnarray}
236: U_x\rho_i =\rho_i U_x,
237: \quad
238: U_x\sigma_i =\sigma_i U_x.
239: \label{eq:ux2}
240: \end{eqnarray}
241: Similarly using the adiabatic flux insertion, we can define 
242: another large gauge transformation $U_y$, in which the 
243: $y$-component of the gauge potential changes from $A_y=0$ to 
244: $A_y=2\pi/eL_y$; and we have 
245: \begin{eqnarray}
246: U_y \tau_i =\tau_i U_y,
247: \quad
248: U_y \rho_i =e^{-2\pi i p/q}\rho_i U_y, 
249: \quad
250: U_y \sigma_i =\sigma_i U_y. 
251: \label{eq:uy}
252: \end{eqnarray}
253: \begin{figure}[h]
254: \begin{center}
255: \includegraphics[width=3cm]{flux.eps}
256: \caption{Two possible insertions of a unit flux $\Phi_0$.}
257: \label{fig:flux}
258: \end{center}
259: \end{figure}
260: 
261: 
262: We notice that the relations (\ref{eq:ux1}), (\ref{eq:ux2}) 
263: and (\ref{eq:uy}) are compatible with the braid group algebra 
264: (\ref{eq:braid1}) and (\ref{eq:braid2}). Thus the large gauge 
265: transformations $U_a$ ($a=x,y$) are (outer) automorphism 
266: of the braid group operators:
267: \begin{eqnarray}
268: \sigma'_i=U_a\sigma_i U_a^{-1},
269: \quad
270: \tau'_i=U_a\tau_i U_a^{-1},
271: \quad
272: \rho'_i=U_a\rho_i U_a^{-1}.
273: \end{eqnarray}  
274: Namely by using relations (\ref{eq:ux1}), (\ref{eq:ux2}) and 
275: (\ref{eq:uy}), one can check that the new operators $\sigma_i'$, 
276: $\tau'_i$ and $\rho'_i$ also satisfy the same braid group 
277: algebra as $\sigma_i$, $\tau_i$ and $\rho_i$.
278: 
279: 
280: It is easy to verify that $U_xU_yU_x^{-1}U_y^{-1}$ 
281: commutes with all $\sigma_i$, $\tau_i$ and $\rho_i$. 
282: Therefore, by Schur's lemma, for any irreducible 
283: representation, $U_xU_yU_x^{-1}U_y^{-1}$ is a 
284: (unimodular) c-number; namely
285: \begin{eqnarray}
286: U_xU_y=e^{2\pi i\lambda} U_y U_x.
287: \label{eq:uxuy}
288: \end{eqnarray}
289: As we will see, $\lambda$ is rational and can be fixed
290: by the requirement of a finite minimal ground state 
291: degeneracy. It is a new many-body quantum number, also 
292: characterizing the topological order of the system, 
293: and is closely related to the fractional quantization 
294: of Hall conductance (see below).   
295: 
296: Assume that the quasiparticles are Abelian anyons:
297: \begin{eqnarray}
298: \sigma_j=e^{i\theta}{\bm 1}, 
299: \end{eqnarray}
300: where ${\bm 1}$ is the unit matrix. Then the braid 
301: group representation is uniquely determined as 
302: \begin{eqnarray}
303: \tau_j=e^{-2i\theta(j-1)}T_x,
304: \quad 
305: \rho_j=e^{2 i\theta(j-1)}T_y, 
306: \label{eq:rep_tau_rho}
307: \end{eqnarray}
308: with $T_x$ and $T_y$ matrices satisfying
309: \begin{eqnarray}
310: T_x T_y=e^{-2i\theta}T_y T_x.
311: \label{eq:exchangeT}
312: \end{eqnarray}
313: On a torus we also have the constraint on $N$ and $\theta$
314: \begin{eqnarray}
315: e^{2iN\theta}=1.
316: \end{eqnarray}
317: Assuming $N\ge 2$, $\theta/\pi$ must be a rational number, 
318: $\theta=\pi m/n$, where $m$ and $n$ are mutually prime 
319: integers. Thus, $T_x$ and $T_y$ satisfy 
320: \begin{eqnarray}
321: T_x T_y=e^{-2\pi im/n}T_y T_x. 
322: \label{eq:exchangeT2}
323: \end{eqnarray}
324: The linear automorphisms induced by $U_x$ and $U_y$ now 
325: reduces to 
326: \begin{eqnarray}
327: &&U_x T_x U_x^{-1}=e^{-2\pi i p/q}T_x,
328: \quad 
329: U_y T_x U_y^{-1}=T_x,  
330: \nonumber\\
331: &&U_x T_y U_x^{-1}=T_y,
332: \quad
333: U_y T_y U_y^{-1}=e^{-2\pi i p/q}T_y.
334: \label{eq:tu}
335: \end{eqnarray}
336: 
337: To count the ground state degeneracy, we consider the following 
338: process. First create $N$ pairs of quasiparticle and quasiholes 
339: out of the ground state, then move the $i$-th quasiparticle by 
340: $\tau_i$. After it returns to the original position, we pair 
341: annihilate all quasiparticles and quasiholes. This process 
342: defines an operation of $\tau_i$ to the ground states. 
343: Similarly, we define the operation of $\rho_i$ and $\sigma_i$ 
344: to the ground states. Throughout this letter, we assume that 
345: Fermi level lies in a gap and the gap remains finite in the 
346: operations above. Since for a system with Abelian anyonic 
347: excitations, the operations $\sigma_i$'s on the ground state 
348: generate merely a phase. Thus we concentrate on the operations 
349: $\tau_1=T_x$ and $\rho_1=T_y$ on the ground state; from 
350: Eq. (\ref{eq:rep_tau_rho}) all other $\tau_i$ and $\rho_i$ 
351: can be expressed in terms of them.
352: 
353: First, we reproduce the degeneracy due to the fractional 
354: statistics \cite{Birman, Einarsson}. Let us take the basis 
355: of the ground state to be an eigenstate of $T_x$,
356: \begin{eqnarray}
357: T_x |\eta\rangle =e^{i\eta}|\eta\rangle. 
358: \end{eqnarray}
359: By applying $T_y$ to $|\eta\rangle$ and using 
360: (\ref{eq:exchangeT2}), the following new states are obtained: 
361: \begin{eqnarray}
362: T_x \left(
363: T_y^{s}|\eta\rangle 
364: \right)
365: =e^{i(\eta-2\pi sm/n)}T_y^{s}|\eta\rangle,
366: \end{eqnarray}
367: where $s$ is an integer. Since the new states have $n$ 
368: different eigenvalues of $T_x$, the ground state has $n$-fold 
369: degeneracy at least. 
370: 
371: The charge fractionalization gives another constraint for 
372: the degeneracy. From Eq.(\ref{eq:exchangeT2}), 
373: $T_x$ and $T_y^n$ commute with each other,
374: \begin{eqnarray}
375: T_x T_y^n=T_y^n T_x. 
376: \end{eqnarray}
377: Therefore, we can take the basis of the ground states which 
378: diagonalize $T_x$ and $T_y^n$ simultaneously, 
379: \begin{eqnarray}
380: T_x|\eta_1, \eta_2\rangle
381: =e^{i \eta_1} |\eta_1,\eta_2\rangle,
382: \quad  
383: T_y^n|\eta_1, \eta_2\rangle
384: =e^{i \eta_2} |\eta_1,\eta_2\rangle.
385: \end{eqnarray}
386: By applying $U_x$ and $U_y$ to this and using Eq.(\ref{eq:tu}), 
387: we have
388: \begin{eqnarray}
389: &&T_x \left(
390: U_x^{s}U_y^{t}|\eta_1,\eta_2\rangle
391: \right)
392: =e^{i(\eta_1+2\pi sp/q)}U_x^{s}U_y^{t}
393: |\eta_1,\eta_2\rangle,
394: \nonumber\\ 
395: &&T_y^n \left(
396: U_x^{s}U_y^{t}|\eta_1,\eta_2\rangle \right)
397: =e^{i(\eta_2+2\pi tnp/q)}U_x^{s}U_y^{t}
398: |\eta_1,\eta_2\rangle, 
399: \nonumber\\
400: \end{eqnarray}
401: where $s$ and $t$ are integers. If $n/q={\cal N}/{\cal Q}$ 
402: where ${\cal N}$ and ${\cal Q}$ are mutually prime integers, it is 
403: found that there are $q{\cal Q}$ sets of eigenvalues of $T_x$ and 
404: $T_y^n$. This implies that the ground state has $q{\cal Q}$-fold 
405: degeneracy at least.
406: 
407: By combining the results above, we find that the minimal 
408: degeneracy of the ground state should be the least common 
409: multiplet of $n$ and $q{\cal Q}=n{\cal Q}^2/{\cal N}$. 
410: Namely, the system has {\it $n{\cal Q}^2$-fold} ground 
411: state degeneracy. This indicates clearly that the fractional 
412: statistics and the charge fractionalization are {\it both 
413: responsible} for the ground state degeneracy. The minimal 
414: degeneracy obtained here includes both the results in 
415: Ref.\cite{WN} and Ref.\cite{OS} as special cases. More 
416: possibilities are predicted.  
417: 
418: Up to now, we have not used non-commutating relation 
419: (\ref{eq:uxuy}) between $U_x$ and $U_y$. This relation
420: contains an additional parameter $\lambda$, which is 
421: not fixed uniquely by $e^{*}$ and $\theta$. We believe 
422: this parameter $\lambda$ could be determined by the 
423: low-energy effective Lagrangian, which we do not discuss 
424: here. In order for the degeneracy to be finite, $\lambda$ 
425: has to be a rational number $\lambda=k/l$, where $k$ and 
426: $l$ are co-primes. At least in the following examples we 
427: find that the integers $k$ and $l$ can be determined by 
428: requiring the degeneracy be minimal given $e^*$ and 
429: $\theta$. In any case, the degeneracy is given by a 
430: multiple of $n{\cal Q}^2$.
431: 
432: Now we present some explicit representations of $T_x$, 
433: $T_y$, $U_x$ and $U_y$ and corresponding degeneracy.   
434: 
435: (1) $\theta=\pi/n$ and $e^{*}=e/n$. -
436: This corresponds to the Laughlin state with $\nu=1/n$.
437: Because ${\cal N}={\cal Q}=1$, the minimum degeneracy 
438: is $n$. If we assume that $U_x$ and $U_y$ satisfy
439: \begin{eqnarray}
440: U_x U_y=e^{-2\pi i/n}U_y U_x, 
441: \end{eqnarray}
442: we can construct $T_x$, $T_y$, $U_x$ and $U_y$ without 
443: increasing the degeneracy,
444: \begin{eqnarray}
445: &&T_x=S_{n\times n},
446: \quad
447: T_y=R_{n\times n},
448: \nonumber\\
449: &&U_x=R^{-1}_{n\times n},
450: \quad
451: U_y=S_{n\times n}. 
452: \end{eqnarray}
453: Here $S_{n\times n}={\rm diag}\{1,e^{i2\pi/n},\cdots,
454: e^{i2\pi(n-1)/n}\}$ and
455: \begin{eqnarray}
456: R_{n\times n}=
457: \left(
458: \begin{array}{ccccc}
459: 0&1& 0& \cdots & 0\\
460: \vdots &0&\ddots&\ddots&\vdots\\
461: \vdots &\vdots &\ddots &\ddots & 0\\
462: 0&\cdots  &\cdots &0 &1 \\
463: 1&0&\cdots&\cdots&0
464: \end{array} 
465: \right).
466: \end{eqnarray}
467: They satisfy $S_{n\times n}R_{n\times n}
468: =e^{-2\pi i/n}R_{n\times n}S_{n \times n}$. This result 
469: reproduces the degeneracy given in Ref.\cite{WN}.
470: 
471: (2) $\theta=0$ or $\theta=\pi$. -
472: The quasiparticles are bosons or fermions.
473: Since $n=1$, we obtain ${\cal N}=1$ and ${\cal Q}=q$. 
474: Thus the minimal degeneracy is $q^{2}$ \cite{OS}.  
475: We find that if $U_x$ and $U_y$ commutes with each other,
476: the minimal degeneracy is realized as
477: \begin{eqnarray}
478: &&T_x=R_{q\times q}\otimes 1_{q\times q},
479: \quad
480: T_y=1_{q\times q}\otimes R_{q\times q}, 
481: \nonumber\\
482: &&U_x=S_{q\times q}^{p}\otimes 1_{q\times q}, 
483: \quad 
484: U_y=1_{q\times q}\otimes S_{q\times q}^{p}.
485: \end{eqnarray}
486: 
487: 
488: (3) $q$ and $n$ are mutually prime. -
489: The degeneracy is $n q^{2}$. We can construct the following 
490: representation for $T_x$, $T_y$, $U_x$ and $U_y$: 
491: \begin{eqnarray}
492: T_x=
493: 1_{q\times q}\otimes R_{q\times q}\otimes S^{m}_{n\times n},
494: \nonumber\\ 
495: T_y=
496: R_{q\times q}\otimes 1_{q\times q}\otimes R_{n\times n},
497: \nonumber\\
498: U_x=1_{q\times q}\otimes S_{q\times q}^{p}\otimes 1_{n\times n},
499: \nonumber\\
500: U_y=S_{q\times q}^{p}\otimes 1_{q\times q}\otimes 1_{n\times n},
501: \end{eqnarray}
502: where the minimal degeneracy is realized. $U_x$ and $U_y$ 
503: commute with each other in this representation.
504: 
505: (4) $n={\cal N}q$ and $m=1$. -
506: Because of ${\cal Q}=1$, the minimum degeneracy is $n$.
507: A representation is given by
508: \begin{eqnarray}
509: &&T_x=S_{n\times n},
510: \quad
511: T_y=R_{n\times n}, 
512: \nonumber\\
513: &&U_x=R_{n\times n}^{-{\cal N}p}, 
514: \quad 
515: U_y=S_{n\times n}^{{\cal N}p}.
516: \end{eqnarray}
517: $U_x$ and $U_y$ satisfy
518: $U_x U_y=e^{-2\pi i ({\cal N}p^2/q)}U_y U_x$. 
519: 
520: 
521: (5) $q={\cal Q}n$. - 
522: In this case, ${\cal N}=1$, thus the least degeneracy is 
523: $n{\cal Q}^2$. When ${\cal Q}$ and $n$ are mutually prime 
524: and $p=m=1$, we can construct the following representation:
525: \begin{eqnarray}
526: &&T_x=S_{{\cal Q}\times {\cal Q}}\otimes S_{n\times n}
527: \otimes 1_{{\cal Q}\times {\cal Q}},
528: \nonumber\\
529: &&T_y=1_{{\cal Q}\times {\cal Q}}\otimes R_{n\times n}
530: \otimes S_{{\cal Q}\times {\cal Q}}, 
531: \nonumber\\
532: &&U_x=R_{{\cal Q}\times {\cal Q}}^{-l}\otimes R_{n\times n}^{-k}
533: \otimes 1_{{\cal Q}\times {\cal Q}}, 
534: \nonumber\\
535: &&U_y=1_{{\cal Q}\times {\cal Q}}\otimes S_{n\times n}^{k}
536: \otimes R_{{\cal Q}\times {\cal Q}}^{-l},
537: \end{eqnarray}
538: where $k/n+l/{\cal Q}=1/{\cal Q}n \, ({\rm mod.} 1)$.
539: In this case, $U_x U_y =e^{-2\pi i k^2/n} U_y U_x$. 
540: 
541: Here we would like to mention that the noncommutativity of the 
542: large gauge transformation $U_xU_y=e^{2\pi i k/l}U_y U_x$ is 
543: closely related to the topological order in the fractional 
544: quantum Hall effect. To see this, consider the degenerate 
545: ground states $\phi_K$ $(K=1,\cdots,d)$, on a torus with 
546: boundary conditions parametrized by twisted phases 
547: $\theta$ and $\varphi$ \cite{NTW}, satisfying 
548: $U_x|\theta,\varphi \rangle_K =|\theta+2\pi,\varphi \rangle_K$ 
549: and 
550: $U_y|\theta,\varphi \rangle_K =|\theta,\varphi+2\pi \rangle_K$.
551: The Hall conductance is
552: \begin{eqnarray}
553: \frac{e^2}{h d} \sum_{K=1}^{d}\int_0^{2\pi}
554: \int_0^{2\pi} \frac {d\theta d\varphi}{2\pi i}
555: \left[\left\langle \frac{\partial\phi_K}{\partial\varphi}|
556: \frac{\partial\phi_K}{\partial \theta}\right\rangle
557: -(\theta\leftrightarrow\varphi) \right].\nonumber
558: \end{eqnarray} 
559: Because of $U_x^{l}U_y=U_y U_x^l$, we can take the basis which
560: diagonalizes both $U_x^{l}$ and $U_y$. In this basis a change 
561: in $\theta$ by $2\pi l$ or in $\varphi$ by $2\pi$ leads the 
562: state back to itself. Therefore, we have a torus with
563: $0\le\theta <2\pi l$ and $0\le\varphi <2\pi$. The above 
564: integral can be recast into
565: \begin{eqnarray}
566: \frac{e^2}{h d} \sum_{r=0}^{d/l-1}
567: \int_0^{2\pi l}\int_0^{2\pi} \frac{d\theta d\varphi}{2\pi i}
568: \left[\left\langle \frac{\partial\phi_{rl+1}}{\partial\varphi}|
569: \frac{\partial\phi_{rl+1}}{\partial \theta}\right\rangle
570: -(\theta\leftrightarrow\varphi) \right], 
571: \nonumber
572: \end{eqnarray}
573: since the degenerate ground states $\phi_K$ satisfy 
574: $\phi_{rl+m}(\theta+2\pi,\varphi)=\phi_{rl+m+1}(\theta,\varphi)$
575: $(r=0,\cdots, d/l-1, m=1,\cdots l-1)$.
576: Therefore we have the following formula for the Hall 
577: conductance, generalizing the result of Ref.\cite{NTW}: 
578: \begin{eqnarray} 
579: \sigma_{xy}=\frac{e^2}{hd}\sum_{r=0}^{d/l-1}I_r
580: =\frac{e^2}{h}\frac{I}{l}.
581: \end{eqnarray}
582: Here $I_r$ is a generalized TKNN integer \cite{TKNN} defined 
583: by $\phi_{rl+1}$. For the second equality, we have noted  
584: that all $I_r$'s take the same value $I$, since the 
585: degenerate ground states are related to each other by 
586: symmetry operations $T_x$ and $T_y$. This indicates clearly 
587: that the noncommutativity $\lambda=k/l$ of the large gauge 
588: transformations is essential to fractional quantization of 
589: the Hall conductance.
590: 
591: The fractional charge we discussed above can be any conserved 
592: $U(1)$ quantum number, with the flux threading understood as 
593: twisted boundary conditions. The minimal degeneracy obtained 
594: can be generalized to a high-genus Riemann surface: If 
595: the genus is $g$, for Abelian topological orders we find $g$ commuting
596: copies of the discrete algebra presented above, so the minimal degeneracy is 
597: $n^{g}{\cal Q}^{2g}$.
598: 
599: To conclude, in this letter we have proposed a discrete 
600: symmetry algebra, Eqs.(\ref{eq:uxuy}), (\ref{eq:exchangeT2}) and 
601: (\ref{eq:tu}), of the operations $T_x$, $T_y$, $U_x$ and $U_y$, 
602: that completely characterizes the ground state subspace of a 
603: generic {\it Abelian} topological order, i.e. that supports
604: Abelian anyonic excitations, on a torus. The identification 
605: verifies the old idea that the ground state degeneracy in a 
606: topological phase is due to the emergence of a discrete 
607: symmetry \cite{TW,WHK}. We note that the algebra identified is indeed 
608: of topological origin and contains only three fractional 
609: parameters: quasiparticle charge $e^*/e$, anyon statistics 
610: $\theta/2\pi$ and flux noncommutativity $\lambda$. Ground 
611: state degeneracy is determined by the representations of 
612: this symmetry algebra.  
613: 
614: {\it Acknowledgement} The work was supported in part by the U.S. NSF
615: through grant No. PHY-0407187 (YSW). This work was begun in summer
616: 2005, when YSW visited the ISSP, University of Tokyo. He thanks the 
617: financial support and warm hospitality from the host institution.   
618: 
619: %\appendix
620: 
621: \bibliography{SKW}% Produces the bibliography via BibTeX.
622: 
623: \end{document}
624: %
625: % ****** End of file apssamp.tex ******
626: 
627: 
628: 
629: 
630: