math0101219/ade.tex
1: % On a $q$-analog of the McKay correspondence and the ADE classification of
2: %   sl_2 conformal field theories
3: %  Alexander Kirillov, Jr. and Victor Ostrik
4: % LaTeX2e
5: % Version of December 2001 
6: % 
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8: \documentclass{amsart}
9: \usepackage{graphicx, amssymb}
10: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
11: \swapnumbers
12: \newtheorem*{theorem*}{Theorem}
13: \newtheorem{theorem}{Theorem}[section]
14: \newtheorem{lemma}[theorem]{Lemma}
15: \newtheorem{proposition}[theorem]{Proposition}
16: \newtheorem{corollary}[theorem]{Corollary}
17: \newtheorem{conjecture}[theorem]{Conjecture}
18: \newtheorem{assumption}[theorem]{Assumption}
19: 
20: \theoremstyle{definition}
21: \newtheorem{definition}[theorem]{Definition}
22: \newtheorem{example}[theorem]{Example}
23: \newtheorem{examples}[theorem]{Examples}
24: \newtheorem{xca}[theorem]{Exercise}
25: 
26: \theoremstyle{remark}
27: \newtheorem{remark}[theorem]{Remark}
28: \newtheorem{remarks}[theorem]{Remarks}
29: \newtheorem{claim}[theorem]{Claim}
30: 
31: \numberwithin{equation}{section}
32: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
33: \newcommand{\firef}[1]{Figure~{\rm\ref{#1}}}
34: \newcommand{\thref}[1]{Theorem~{\rm\ref{#1}}}
35: \newcommand{\prref}[1]{Proposition~{\rm\ref{#1}}}
36: \newcommand{\leref}[1]{Lemma~{\rm\ref{#1}}}
37: \newcommand{\coref}[1]{Corollary~{\rm\ref{#1}}}
38: \newcommand{\clref}[1]{Claim~{\rm\ref{#1}}}
39: \newcommand{\conref}[1]{Conjecture~{\rm\ref{#1}}}
40: \newcommand{\deref}[1]{Definition~{\rm\ref{#1}}}
41: \newcommand{\exref}[1]{Example~{\rm\ref{#1}}}
42: \newcommand{\ecref}[1]{Exercise~{\rm\ref{#1}}}
43: \newcommand{\reref}[1]{Remark~{\rm\ref{#1}}}
44: \newcommand{\quref}[1]{Question~{\rm\ref{#1}}}
45: \newcommand{\seref}[1]{Section~{\rm\ref{#1}}}
46: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
47: \newcommand\NB{\marginpar{\huge\bfseries ?!}}
48: \newcommand{\fig}[1]
49: {\raisebox{-0.5\height}%
50: {\includegraphics{#1}}}
51: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
52: \newcommand{\st}{\; | \;}                               %%  such that
53: \newcommand{\ttt}{\otimes}                               %% tensor product
54: \newcommand{\tta}{\otimes_A}                               %% tensor product
55: \newcommand{\tbox}{\boxtimes} 
56: \newcommand{\ot}{\stackrel{{\raisebox{-1pt}{\text{\bfseries\Large.}}}}
57:                            {\otimes}
58:                 }
59: \newcommand{\ti}{\tilde}
60: \newcommand{\ov}{\overline}
61: \renewcommand{\d}{\partial}
62: \newcommand{\<}{\langle}
63: \renewcommand{\>}{\rangle}
64: \newcommand{\surjto}{\twoheadrightarrow}      %   -->> surjection
65: \newcommand{\injto}{\hookrightarrow}          %  (--> injection
66: \newcommand{\isoto}{\xrightarrow{\sim}}       % isomor.
67: \newcommand{\xxto}{\xrightarrow}              % long arrow
68: 
69: %%%%%%%%%%%%%%%%%%%%%%%%%%%
70: \newcommand{\one}{\mathbf{1}}
71: \renewcommand{\i}{{\mathrm{i}}}   % imaginary unit
72: \newcommand{\Cset}{\mathbb{C}}    % complex numbers
73: \newcommand{\Ctimes}{\mathbb{C}^\times}       % nonzero complex numbers
74: \newcommand{\Z}{\mathbb{Z}}       % integers
75: \newcommand{\Rset}{\mathbb{R}}    % real numbers
76: \newcommand{\Q}{\mathbb{Q}}       % rational numbers
77: \newcommand{\N}{\mathbb{N}}       % natural numbers
78: \newcommand{\dd}{\mathbf{d}}      
79: \newcommand{\V}{{\mathcal{V}}}    % VOA  
80: \newcommand{\C}{\mathcal{C}}      % a category    
81: \newcommand{\A}{\mathcal{A}}      % another  category    
82: \newcommand{\Y}{\mathcal{Y}}      % intertwining operator for VOAs
83: \newcommand{\tY}{\tilde Y}
84: %%%%%%%%%%%%%%%%%%%%%%%%%%%
85: \newcommand{\al}{\alpha}
86: \newcommand{\be}{\beta}
87: \newcommand{\ga}{\gamma}
88: \newcommand{\Ga}{\Gamma}
89: \newcommand{\de}{\delta}
90: \newcommand{\De}{\Delta}
91: \newcommand{\la}{\lambda}
92: \newcommand{\ph}{\varphi}
93: \newcommand{\Ph}{\Phi}
94: \newcommand{\Si}{\Sigma}
95: \newcommand{\om}{\omega}
96: \newcommand{\eps}{\varepsilon}
97: \renewcommand{\th}{\theta}
98: %%%%%%%%%%%%%%%%%%%%%%%%%%%
99: \newcommand{\g}{\mathfrak{g}}
100: \newcommand{\ghat}{\widehat{\mathfrak{g}}}
101: \def\Uqg{U_q(\g)}                     %% quantum groups
102: \newcommand{\U}{U_q(\mathfrak{sl}_2)} %% U_q sl_2
103: \newcommand{\slt}{\mathfrak{sl}_2}    %% sl_2 
104: \newcommand{\slthat}{\widehat{\mathfrak{sl}}_2}    %% sl_2 
105: 
106: 
107: %%%%%%%%%%%%%%%%%%%%%%%%%%%
108: 
109: \DeclareMathOperator{\Res}{Res}
110: \DeclareMathOperator{\Rep}{Rep}
111: \DeclareMathOperator{\Ind}{Ind}
112: \DeclareMathOperator{\id}{id}
113: \DeclareMathOperator{\Aut}{Aut}
114: \DeclareMathOperator{\Hom}{Hom}
115: \DeclareMathOperator{\Ext}{Ext}
116: \DeclareMathOperator{\End}{End}
117: \DeclareMathOperator{\tr}{tr}
118: \DeclareMathOperator{\im}{Im}
119: 
120: 
121: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
122: 
123: \begin{document}
124: 
125: \title[On a $q$-analog of McKay correspondence] 
126:       {On a $q$-analog of the McKay correspondence and 
127:        the ADE classification of $\slthat$ conformal field theories}
128: 
129: \author{Alexander Kirillov, Jr.}
130: \address{Department of Mathematics, SUNY at Stony Brook, 
131:          Stony Brook, NY 11794, USA}
132: \email{kirillov@math.sunysb.edu}
133: \urladdr{http://www.math.sunysb.edu/\textasciitilde kirillov/}
134: \author{Viktor Ostrik}
135: \address{Department of Mathematics, MIT, 
136:          Cambridge, MA 02139}
137: \email{ostrik@math.mit.edu}
138: \thanks{The first author was supported in part by NSF Grant
139:   DMS--9970473.\\
140:     The second author was supported in part by NSF Grant DMS--0098830}
141:                            
142: \begin{abstract}
143:   The goal of this paper is to give a category theory based definition
144:   and classification of ``finite subgroups in $\U$'' where
145:   $q=e^{\pi\i/l}$ is a root of unity. We propose a definition of such a
146:   subgroup in terms of the category of representations of $\U$; we
147:   show that this definition is a natural generalization of the notion
148:   of a subgroup in a reductive group, and that it is also related with
149:   extensions of the chiral (vertex operator) algebra corresponding to
150:   $\slthat$ at level $k=l-2$.  We show that ``finite subgroups in
151:   $\U$'' are classified by Dynkin diagrams of types $A_n, D_{2n}, E_6,
152:   E_8$ with Coxeter number equal to $l$, give a description of this
153:   correspondence similar to the classical McKay correspondence, and
154:   discuss relation with modular invariants in $(\slthat)_k$ conformal
155:   field theory.
156:   
157:   The results we get are parallel to those known in the theory of von
158:   Neumann subfactors, but our proofs are independent of
159:   this theory. 
160: 
161:   
162: 
163: \end{abstract}
164: 
165: \maketitle
166: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
167: \section*{Introduction}
168: The goal of this paper is to describe a $q$-analogue of the McKay
169: correspondence. Recall that the usual McKay correspondence is a
170: bijection between finite subgroups $\Ga\subset SU(2)$ and affine
171: simply-laced Dynkin diagrams (i.e., affine ADE diagrams). Under this
172: correspondence, the vertices of Dynkin diagram correspond to
173: irreducible representations of $\Ga$ and the matrix of tensor product
174: with $\Cset^2$ is $2-A_{ij}$ where $A$ is the Cartan matrix of the ADE
175: diagram (see \cite{mckay}). 
176: 
177: There have been numerous results regarding generalization of McKay
178: correspondence with $SU(2)$ replaced by $\U$ with $q$ being a root of
179: unity, $q=e^{\pi\i/l}$.  Of course, since $\U$ is not a group, one
180: must first find a reasonable way of making sense of this question.
181: This is usually done by reformulating the problem in terms of
182: representation theory of $\U$. More precisely, it is known that the
183: category of finite-dimensional representations of $\U$ has a
184: semisimple subquotient category $\C$, with simple objects $V_0\dots,
185: V_k, k=l-2$ (definition of this category was suggested by Andersen and
186: his collaborators, see \cite{AP}; a review can be found, e.g., in
187: \cite{Finkelberg} or in \cite{BK}). As was shown by Finkelberg
188: \cite{Finkelberg}, using results of Kazhdan and Lusztig \cite{KL}`,
189: the category $\C$ is equivalent to the category of integrable
190: $\slthat$-modules of level $k$ with fusion tensor product. This latter
191: category plays a key role in the Wess--Zumino--Witten model of
192: conformal field theory.  Finally, as was shown by Wassermann and his
193: students, fusion in this category can also be described in the
194: language of von Neumann algebras (see, e.g.  \cite{was2},
195: \cite{laredo}).
196: 
197: Below is an overview of some of the known classification results
198: related to $\U$, or, more precisely, to category $\C$. 
199: 
200: 
201: 
202: 
203: \begin{enumerate}
204:   
205: \item {\em Ocneanu's classification of subfactors}.\footnote{We would
206:     like to thank the referee and M.~M\"uger for explaining to us the
207:     status of this classification.} It is known that to every
208:   inclusion of von Neumann factors $N\subset M$ of finite index one
209:   can associate a number of algebraic structures (index, principal
210:   graph, relative commutants, etc). In particular, such an inclusion
211:   defines a tensor category of $N-N$ bimodules. Ocneanu has suggested
212:   that a subfactor $N\subset M$ with the category of $N-N$ bimodules
213:   equivalent to $\C$ should be considered as a ``subgroup of $\U$'';
214:   thus, classification of subgroups reduces to classification of
215:   subfactors. He also gave \cite{ocneanu} a complete classification of
216:   such subfactors: they are classified by Dynkin diagrams of types $A$
217:   (which corresponds to trivial inclusion), $D_{2n}$, $E_6$, $E_8$,
218:   with Coxeter number equal to $l$. Full proof of this result has been
219:   given in the works of Popa \cite{popa}, Bion--Nadal \cite{BN},
220:   \cite{BN2}, Izumi \cite{I}, \cite{I2}.
221: 
222:   
223: \item {\em Cappelli-Itzykson-Zuber's classification of modular
224:     invariants} of conformal field theories based on integrable
225:   representations of $\slthat$ at level $k=l-2$ (see \cite{CIZ} or the
226:   review in \cite{CFT}). These modular invariants are classified by
227:   Dynkin diagrams of ADE type with Coxeter number equal to $l$. It is
228:   known, however, that modular invariants of types $A, D_{even},
229:   E_6,E_8$ can be obtained from extensions of the corresponding chiral
230:   (or vertex operator) algebra, while invariants of the type $E_7,
231:   D_{odd}$ can not be obtained in this way (see \cite{MST}). This
232:   classification is related to the previous one: it can be shown that
233:   every subfactor $N\subset M$ gives a modular invariant, see
234:   \cite{ocneanu2}, \cite{ocneanu3} and papers of
235:   B\"ockenhauer--Evans--Kawahigashi \cite{bek, bek2}.
236: 
237:   
238: \item {\em Etingof and Khovanov's classification of the ``integer''
239:   modules over the Gro\-then\-dieck ring} (``fusion algebra'') of $\C$ (see
240:   \cite{EK}). In this classification, all finite Dynkin diagrams and
241:   even diagrams with loops appear. 
242: 
243: \end{enumerate}
244: 
245: It should be noted that the classification of ``subgroups in $\U$''
246: given in the theory of subfactors requires good knowledge of von
247: Neumann algebras and subfactors. It is very different from the ideas
248: in the proof of the classical McKay correspondence.
249: 
250: The main goal of the present paper is to study an alternative
251: definition of a subgroup in $\U$ which uses nothuing but the
252: theory of tensor categories (which one has to use anyway to work in
253: $\C$). Namely, a subgroup in $\U$ is by definition a commutative
254: associative algebra in $\C$, i.e. an object $A\in \C$ with
255: multiplication morphism $\mu\colon A\ttt A\to A$ satisfying suitably
256: formulated commutativity, associativity and unit axioms and some mild
257: technical restrictions.\footnote{While we arrived at this definition
258:   independently, we are hardly the first to introduce it. This
259:   definition had also been suggested by Wassermann \cite{was} and
260:   M\"uger (unpublished).} We argue that this is the right definition for
261: the following reasons:
262: 
263: \begin{enumerate}
264: \item If we replace $\C$ by a category of representations of a
265:   reductive group $G$, then commutative associative algebras in $\C$
266:   correspond to subgroups of finite index in $G$.
267:   
268: \item If we replace $\C$ by a category of representations of some
269:   vertex operator algebra $\V$ (which is good enough so that $\C$ is a
270:   modular tensor category, as it happens for all VOA's appearing in
271:   conformal field theory), then associative commutative algebras in
272:   $\C$ (with some minor restrictions) exactly correspond to
273:   ``extensions'' $\V_e\supset \V$ of this VOA; in other words, in this
274:   way we recover the notion of extension of a conformal field theory.
275: 
276: \item Every subfactor $N\subset M$ defines such an algebra in the
277:   category of $N-N$ bimodules. 
278: 
279: \end{enumerate}
280: 
281: We show that for any modular category $\C$ a commutative associative
282: algebra $A\in\C$ gives rise to two different categories of modules
283: over $A$. One of these categories, $\Rep A$, comes with two natural
284: functors $F\colon \C\to \Rep A, G\colon \Rep A\to \C$; $F$ is a tensor
285: functor, so it defines on $\Rep A$ a structure of a module category
286: over $\C$. There is also a smaller category $\Rep^0 A$; if $A$ is
287: ``rigid'', then both $\Rep A$ and $\Rep^0 A$ are semisimple and
288: $\Rep^0A$ is modular. Both of these categories have appeared in the
289: physical literature in the language of extensions of chiral algebra:
290: in particular, $\Rep^0 A$ is the category of modules over the extended
291: VOA $\V_e$, and  $\Rep A$ is the category of ``twisted''
292: $\V_e$-modules. These modules appear as  possible boundary
293: conditions for extended CFT which preserve $\V$ (see \cite{FS},
294: \cite{PZ} and references therein); sometimes they are also called
295: ``solitonic sectors''.
296: 
297: 
298: Applying this general setup to $\C$ being the semisimple part of
299: category of representations of $\U$, we see that the fusion algebra of
300: $\Rep A$ is a module over the fusion algebra of $\C$, which gives a
301: relation with Etingof-Khovanov classification mentioned above. Using
302: their results, we get the following classification theorem which we
303: consider to be the $q$-analogue of McKay correspondence:
304: 
305: \begin{theorem*}
306:   Commutative associative algebras in $\C$ are classified by the
307:   \textup{(}finite\textup{)} Dynkin diagrams of the types $A_n,
308:   D_{2n}, E_6, E_8$ with Coxeter number equal to $l$. Under this
309:   correspondence, the vertices of the Dynkin diagram correspond to
310:   irreducible representations $X_i\in \Rep A$ and  the matrix of tensor
311:   product with $F(V_1)$ in this basis is $2-A$, where $A$ is the Cartan
312:   matrix of the Dynkin diagram and  $V_1$ is the fundamental
313:   \textup{(}2-dimensional\textup{)} representation of $\U$. 
314: \end{theorem*}
315: 
316: Since $\Rep^0 A$ is modular, each of these algebras gives a modular
317: invariant providing a relation with the ADE classification of
318: Cappelli-Itzykson-Zuber.
319: 
320: 
321: 
322: The first part of this theorem --- that is, that commutative
323: associative algebras are classified by Dynkin diagrams---is hardly
324: new; in the language of extensions of a chiral algebra, it has been
325: (mostly) known to physicists long ago (see, e.g.  \cite{MST}), and
326: these extensions have been studied in a number of papers.  However,
327: the second part of the theorem, which explicitly describes a
328: correspondence in a manner parallel to the classical McKay
329: correspondence to the best of our knowledge is new.
330: 
331: The main result of this theorem is parallel to the classification of
332: finite subgroups in $\U$ as defined in the theory of subfactors.  Not
333: being experts in this theory, we are not describing the precise
334: relation here\footnote{We were recently informed by M.~M\"uger that he
335:   is currently writing a series of papers, which, among other things,
336:   will give a detailed review of this connection.}. We just note that
337: our proofs are completely independent and are not based on the
338: subfactor theory, even though some of the methods we use (most
339: notably, the use of conformal embeddings) are parallel to those used
340: in the subfactor theory. 
341: 
342: 
343: Finally, it should be noted that the problem of finding $\C$ algebras
344: $A$ is closely related to the problem of finding all module categories
345: over $\C$ (such module categories also play important role in CFT; in
346: physical literature, they are usually described by a certain kind of
347: $6j$-symbols, see \cite{PZ}).  Indeed, for every $\C$-algebra $A$ the
348: category $\Rep A$ is a module category over $\C$. It is expected that
349: in the $\U$ case, all module categories over $\C$ are classified by
350: all ADE Dynkin diagrams with Coxeter number equal to $l$. Theorem
351: above gives construction of module categories of type $A_m, D_{even},
352: E_6, E_8$; it is easy to show that a module category of type
353: $D_{2n+1}$ can be constructed from representations of some associative
354: but not commutative $\C$-algebra.  We expect the same to hold for the
355: module category of type $E_7$.
356: 
357: 
358: 
359: 
360: 
361: 
362: 
363: 
364: 
365: 
366: \subsection*{Note} While working on this paper, we were informed by
367: A.~Wassermann and H.~Wenzl that they have obtained similar results
368: based on the subfactor theory. In fact, many of the results of this
369: paper coincide with those  announced by Wassermann in \cite{was}
370: (except that we do not use unitary structure), even though we arrived
371: at these results completely independently. 
372: 
373: 
374: \subsection*{Note} It was recently pointed out to us that some of the
375: results about algebras in a braided tensor category and modules over
376: them which we prove in Section 1 have been previously found in
377: \cite{Pa}. 
378: 
379: 
380: \subsection*{Acknowledgments} We would like to thank  Pavel Etingof for many
381: fruitful discussions and R.~Coquereaux, J.~Fuchs, Y.-Z.~Huang,
382: Y.~Kawahigashi, J.~Lepowsky, J.~McKay, M.~M\"uger, V.~Petkova,
383: C.~Schweigert, F.~Xu for comments on the first version of this paper.
384: We would also like to thank the referee for his detailed comments and
385: numerous suggestions.
386: 
387: 
388: 
389: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
390: \section{Algebras and modules}\label{sbasic}
391: Throughout the paper, we denote by $\C$ a semisimple abelian category
392: over $\Cset$ (most of the results are also valid for any base field
393: $k$ of characteristic zero). We denote by $I$ the set of isomorphism
394: classes of irreducible objects in $\C$ and fix some choice of
395: representative $V_i$ for every $i\in I$. We always assume that the
396: spaces of morphisms are finite-dimensional; since $\Cset$ is
397: algebraically closed, this implies that $\Hom_\C(V_i, V_i)=\Cset$. We
398: will denote $\langle X,Y\rangle= \dim\Hom_\C(X,Y)$; in particular,
399: $\langle V_i, X\rangle$ is the multiplicity of $V_i$ in $X$ which
400: shows that this multiplicity is finite. 
401: 
402: We assume that $\C$ is a rigid balanced braided tensor category (see,
403: e.g., \cite{BK} for a review of the theory of braided tensor
404: categories). The commutativity isomorphism will be denoted by
405: $R_{V,W}$; the associativity isomorphism and other canonical
406: identifications such as $(V\ttt W)^*\simeq W^*\ttt V^*$ will be
407: implicit in our formulas. Additionally, we require that $\one$ is a
408: simple object in $\C$. We will use the symbol $0$ to denote the
409: corresponding index in $I$ : $V_0=\one$.
410: 
411: We denote by $K(\C)$ the complexified Grothendieck ring (``fusion
412: algebra'') of the category $\C$; this is a commutative associative
413: algebra over $\Cset$ with a basis given by classes $[V_i]$ of simple
414: objects.
415: 
416: 
417: \begin{definition}\label{dalg}
418:   An associative commutative algebra $A$ in $\C$ (or $\C$-algebra for
419:   short) is an object $A\in \C$ along with morphisms $\mu\colon A\ttt
420:   A\to A$ and $\iota_A\colon \one \injto A$ such that the following
421:   conditions hold:
422: \begin{enumerate}
423: \item (Associativity) 
424: Compositions $\mu \circ(\mu \ttt \id), \mu\circ (\id\ttt \mu)\colon A^{\ttt
425:   3}\to A$ are equal. 
426: 
427: \item  (Commutativity) 
428: Composition $\mu \circ R_{AA} \colon A\ttt A \to A$ is equal to $\mu$.
429: 
430: \item (Unit) 
431: Composition $\mu \circ (\iota_A\ttt A)\colon A=\one\ttt A\to A$ is equal to
432: $\id_A$.
433: 
434: \item (Uniqueness of unit)
435: $\dim \Hom_{\C}(\one, A)=1$.
436: 
437: \end{enumerate}
438: \end{definition}
439: 
440: 
441: The notion of a $\C$-algebra is not new; it had been used in many
442: papers (for example, in \cite{Ros}). However, most authors only use
443: algebras in symmetric tensor categories. Algebras in braided
444: categories were studied in \cite{B}; however, the discussion there is
445: limited to the case where algebra $A$ is ``transparent'', i.e.
446: $R_{VA}R_{AV}=\id$ for every $V\in \C$. This setting is too
447: restrictive for us.  Most of results in \cite{B} generalize to
448: non-transparent case easily, others (mainly, the results regarding
449: distinction between two categories of modules, $\Rep A$ and $\Rep^0
450: A$) require significant work. 
451: 
452: Commutative algebras in a braided tensor category are also discussed
453: in \cite{Pa}. Unfortunately, this paper was only brought to our
454: attention after the first version of the current paper appeared in the
455: electronic archive. For this reason and for  for reader's
456: convenience, we give here complete proofs of most of the  results; still, we
457: would like to point out that most results of this section were first
458: obtained in \cite{Pa} and \cite{B}.
459: 
460: 
461: 
462: 
463: We will frequently use graphs to present morphisms in $\C$, as
464: suggested by Reshetikhin and Turaev. We will use the same conventions
465: as in \cite{BK}, namely, the morphisms act ``from bottom to top''. We
466: will use dashed line to represent $A$ and the graphs shown in \firef{fmuiota}
467: to represent $\mu$ and $\iota_A$. 
468: \begin{figure}[h]%% 30%
469: \begin{equation*}
470: \fig{muiota}
471: \end{equation*}
472: \caption{Morphisms $\mu$ and $\iota_A$}\label{fmuiota}
473: \end{figure}
474: 
475: With this notation, the axioms of a $\C$-algebra can be presented as
476: shown in \firef{fig1}. 
477: 
478: \begin{figure}[h]%% 30%
479: \begin{equation*}
480: \fig{dalgi.eps}\quad \fig{dalgii.eps}\quad \fig{dalgiii.eps}
481: \end{equation*}
482: \caption{Axioms of a commutative associative algebra.}\label{fig1}
483: \end{figure}
484: 
485: We leave it to the reader to define the notions of morphism of
486: algebras, subalgebras and ideals, quotient algebras etc. 
487: 
488: \begin{definition}\label{drepa}
489:   Let $\C$ be as above and $A$ --- a $\C$ algebra. Define the category
490:   $\Rep A$ as follows: objects are pairs $(V, \mu_V)$ where $V\in \C$
491:   and $\mu_V\colon A\ttt V\to V$ is a morphism in $\C$ satisfying the
492:   following properties: 
493: \begin{enumerate}
494: \item
495: $\mu_V \circ (\mu\ttt \id)=\mu_V \circ (\id \ttt
496:   \mu_V)\colon A\ttt A\ttt V\to V$
497: \item $\mu_V(\iota_A \ttt \id)=\id\colon \one\ttt V\to V$
498: \end{enumerate}
499: 
500: The morphisms are defined by 
501: \begin{multline}\label{amorph}
502: \Hom_{\Rep A}((V, \mu_V), (W, \mu_W))\\
503: =\{ \ph\in \Hom_\C(V, W)\st 
504:                    \mu_W\circ(\id \ttt \ph)=\ph\circ\mu_V\colon A\ttt V\to
505:                    W\} 
506: \end{multline}
507: (see \firef{fig2}). 
508: 
509: 
510: \end{definition}
511: 
512: 
513: \begin{figure}[h]%% 30%
514: \begin{equation*}
515: \fig{drepa.eps}
516: \end{equation*}
517: \caption{Definition of morphisms in $\Rep A$}\label{fig2}
518: \end{figure}
519: 
520: 
521: An instructive example of such a situation is when $G$ is a finite
522: group and $\C$ is the category of finite-dimensional complex
523: representations of $G$. In this case we will show that semisimple
524: $\C$-algebras correspond to subgroups in $G$ (see \seref{sgroup}).
525: 
526: \begin{remark}\label{rintuition}
527: Contrary to the usual intuition, typically the larger $A$, the
528: smaller is its category of representations. In the above mentioned
529: example $\C=\Rep G$, correspondence between subgroups $H\subset G$ and
530: $\C$-algebras is given by $A=F(G/H)$, so large $A$ corresponds to
531: small $H$ and thus, to small $\Rep A=\Rep H$.
532: \end{remark}
533: 
534: 
535: Let us study basic properties of $\Rep A$. For brevity, we will use
536: notation $\Hom_A$ instead of $\Hom_{\Rep A}$.
537: 
538: \begin{lemma}\label{labelian}\ \\
539: \begin{enumerate}
540: \item $\Rep A$ is an abelian category with finite-dimensional spaces
541:   of morphisms; every object in $\Rep A$ has finite length. 
542: \item $\Hom_A(A,A)=\Cset$. 
543: \end{enumerate}
544: \end{lemma}
545: \begin{proof}
546: Since $\C$ is an abelian category, it suffices to prove that for
547: $f\in \Hom_{A}(V, W)$, $\im f$ and $\ker f$ are actually
548: $A$-submodules in $W, V$ respectively. The check is straightforward
549: and is left to the reader. 
550: 
551: Let $\ph\in \Hom_A(A,A)$. By definition we have:
552: \begin{equation*}
553: \ph = \ph \mu (\id_A\ttt \iota_A)=\mu (\id \ttt \ph \iota_A)
554: \end{equation*}
555: But since $\one$ has multiplicity one in $A$, one has $\ph \iota_A=c \iota_A$
556: for some constant $c$. Thus, $\ph = c\mu (\id_A\ttt \iota_A)=c\id$. 
557: \end{proof}
558: 
559: \begin{theorem}\label{ttensor}
560: $\Rep A$ is a monoidal category with unit object $A$.
561: 
562: \end{theorem}
563: \begin{proof}
564: Let $V, W\in \Rep A$. Define $V\tta W=V\ttt W/\im(\mu_1-\mu_2)$ where
565: $\mu_1, \mu_2\colon A\ttt V\ttt W\to V\ttt W$ are defined by 
566: \begin{align*}
567: \mu_1&=\mu_V\ttt\id_W,\\
568: \mu_2&=(\id_V\ttt \mu_W)R_{AV}.
569: \end{align*}
570: This defines $V\tta W$ as an object of $\C$. Define $\mu_{V\tta W}$ to
571: be $\mu_1$ or $\mu_2$ which obviously give the same morphism. One
572: easily sees that this defines a structure of $A$-module on $V\ttt W$
573: and that so defined tensor product is associative. To
574: check that $A$ is the unit object, consider morphisms $\mu_V\colon
575: A\tta V\to V$ and $\iota_A\ttt\id_V\colon  V\to A\tta V$. Straightforward
576: check shows that they are well defined, commute with the action of $A$
577: (that is, satisfy \eqref{amorph}) and thus define morphisms in $\Rep
578: A$ and finally, that they are inverse to each other. 
579: \end{proof}
580: 
581: \begin{theorem}\label{tfunctors}
582: Define functors $F\colon \C\to \Rep A, G\colon \Rep A \to \C$ by 
583: $F(V)=A\ttt V, \mu_{F(V)}=\mu \ttt \id$ and $G(V, \mu_V)=V$. Then 
584: 
585: \begin{enumerate}
586: \item Both $F$ and $G$ are exact and injective on morphisms.
587: \item $F$ and $G$ are adjoint: one has canonical functorial
588:   isomorphisms 
589: $$
590: \Hom_A(F(V), X)=\Hom_\C(V, G(X)), \qquad V\in \C, X\in \Rep A.
591: $$
592: 
593: \item $F$ is a tensor functor: one has canonical isomorphisms $F(V\ttt
594:   W)= F(V)\tta F(W), F(\one)=A$. 
595: \item One has canonical isomorphisms $G(F(V))=A\ttt V$ and, more
596:   generally, $G(F(V)\tta X)=  V\ttt G(X)$. 
597: \end{enumerate}
598: \end{theorem}
599: 
600: \begin{proof}
601:  Part (1) is obvious; for part (2), define maps $\Hom_A(F(V), X)\to
602:  \Hom_\C(V, G(X))$ and $\Hom_\C(V, G(X))\to \Hom_A(F(V), X)$ as shown
603:  in \firef{fadjoint1}; it is easy to deduce from the axioms that these
604:  maps are inverse to each other. 
605: 
606: \begin{figure}[h]%% 30%
607: \begin{equation*}
608: \fig{adjoint1.eps}
609: \end{equation*}
610: \caption{Identifications $\Hom_A(F(V), X)=\Hom_\C(V, G(X))$.
611:         Here  $\ph\in\Hom_A(F(V), X), \ \Phi\in\Hom_\C(V, G(X))$.}
612:         \label{fadjoint1}
613: \end{figure}
614: 
615: 
616: 
617:   To prove that $F$ is a tensor functor, define functorial morphisms
618:   $f\colon F(V\ttt W)\to F(V)\tta F(W), g\colon F(V)\tta F(W)\to
619:   F(V\ttt W)$ by 
620: \begin{align*}
621: f&=\id_A\ttt\id_V\ttt \iota_A\ttt \id_W\colon A\ttt V\ttt
622:   W\to (A\ttt V)\tta (A\ttt W)\\
623: g&\colon (A\ttt V)\tta (A\ttt W)\xxto{R_{AV}^{-1}} A\tta
624:   A\ttt V \ttt V \xxto{\mu} A\ttt V\ttt W.
625: \end{align*}
626: It is immediate to check that they are well-defined and inverse to
627: each other. 
628: \end{proof}
629: 
630: \begin{corollary}\label{cmodule}
631: $\Rep A$ is a module category over $\C$, i.e. there is an additive  functor
632: $\tbox\colon \C\times \Rep A\to \Rep A$ and  isomorphisms 
633: \begin{align*}
634: &(V_1\ttt V_2)\tbox X\simeq V_1\tbox (V_2\tbox X), 
635:             \qquad V_1, V_2\in \C, X\in \Rep A\\
636: &\one\tbox X\simeq X, \quad X\in \Rep A
637: \end{align*}
638: satisfying usual compatibility conditions.
639: \end{corollary}
640: \begin{proof}
641: Suffices to take $V\tbox X=F(V)\tta X$. 
642: \end{proof}
643: 
644: 
645: In particular, this implies that the Grothendieck group $K(\Rep A)$ is
646: a module over the Grothendieck ring $K(\C)$. 
647: 
648: 
649: However, it is not true that $\Rep A$ is a braided tensor
650: category. In order to get a braided structure, we need to consider a
651: smaller category. 
652: 
653: 
654: \begin{definition}\label{drep0}
655: $\Rep^0A$ is the full subcategory in $\Rep A$ consisting of objects
656: $(V, \mu_V)$ such $\mu_V\circ R_{VA}R_{AV}=\mu_V$. 
657: \end{definition}
658: 
659: \begin{remark}
660: In \cite{Pa}, such $A$-modules are called ``dyslectic''. 
661: \end{remark}
662: 
663: If $\C$ is symmetric, then $\Rep^0 A=\Rep A$. More generally, the same
664: holds if  $A$ is ``transparent'',
665: or ``central'', in $\C$ (that is, $R_{VA}R_{AV}=\id$ for every $V\in
666: \C$); this is the situation considered in \cite{B}. However, in many
667: interesting cases $A$ is not central, and $\Rep A\ne \Rep^0 A$. 
668: 
669: Later we will justify this definition by showing that if $\C$ is a
670: category of representation of some vertex operator algebra, and $A$ is
671: an extended vertex operator algebra, then $\Rep^0A$ (and not $\Rep
672: A$!) is exactly the category of representations of the vertex operator
673: algebra $A$. 
674: 
675: 
676: 
677: 
678: \begin{theorem}[\cite{Pa}]
679: \label{tbraided}
680: The category $\Rep^0 A$ is a braided tensor category, with the
681: commutativity isomorphism inherited from $\C$.
682: \end{theorem}
683: \begin{proof}
684:   Let us first show that for $X,Y\in \Rep^0 A$, 
685:   $X\tta Y\in \Rep^0 A$. This follows from the sequence of identities
686:   shown in \firef{frtta}. The notation $f_1\equiv f_2$ for $f_1, f_2\colon
687:   A\ttt X\ttt Y\to X\ttt Y$ means that the induced operators $A\ttt
688:   (X\ttt Y)\to X\tta Y$ are equal, i.e.  $p\circ f_1=p\circ f_2$,
689:   where $p\colon X\ttt Y\to X\tta Y$ is the canonical projection.  
690: \begin{figure}[ht]% 30%
691: $$
692: \fig{tta}\quad=\quad\fig{ttb}\quad\equiv\quad\fig{ttc}
693: \quad=\quad\fig{ttd}\quad\equiv\quad\fig{tte}
694: $$
695: \caption{}\label{frtta}
696:   \end{figure}
697:  
698:   
699:   Next, we need to show that the commutativity isomorphism
700:   $R_{XY}\colon X\ttt Y\to Y\ttt X$ descends to isomorphism $ X\tta
701:   Y\to Y\tta X$.  This is equivalent to showing that $R_{XY}
702:   (I)\subset I$, where $I=\im (\mu_1-\mu_2)$ is the kernel of the
703:   canonical projection $X\ttt Y\to X\tta Y$.  To do so, let us rewrite
704:   composition $R_{XY}\circ (\mu_1-\mu_2)$ as shown in
705:   \firef{frcommut}.  Thus, $R_{XY}(\mu_1-\mu_2)\equiv 0$, or,
706:   equivalently, $R_{XY}(I)\subset I$.
707: \begin{figure}[ht]% 30%
708: \begin{align*}
709: & \fig{rcommuta}-\fig{rcommutb}
710:   =\fig{rcommutc}-\fig{rcommutd}\\
711: &\qquad
712:   \equiv \fig{rcommutc}-\fig{rcommute}\equiv 0.
713: \end{align*}
714: \caption{}\label{frcommut}
715:   \end{figure}
716:  
717: Abusing the language, we will also use the same notation $R_{X,Y}$ for
718: the descended morphisms $X\tta Y\to Y\tta X$. Then it is immediate
719: from the definition that these morphisms are $A$-morphisms. Finally,
720: since the commutativity isomorphism in $\C$ satisfies the hexagon
721: axioms, the same must hold for the descended operators; thus, the
722: descended operators $R_{X,Y}\colon X\tta Y\to Y\tta X$ define a
723: structure of a braided tensor category on $\Rep^0 A$. 
724: \end{proof}
725: 
726: Analysis of this proof also shows the reason why the larger category
727: $\Rep A$ is not braided: the last identity in \firef{frcommut} would
728: fail. 
729: 
730: 
731: 
732: We also need to know whether categories $\Rep A, \Rep^0 A$ are rigid.
733: Define $\eps_A\colon A\to \one$ so that $\eps_A \iota_A=\id$ (recall
734: that $\langle A,\one\rangle=1$, so this condition uniquely defines
735: $\eps_A$). We will use the graph shown in \firef{feps} to represent
736: $\eps_A$.
737: 
738: \begin{figure}[h]%% 30%
739: \begin{equation*}
740: \fig{eps}
741: \end{equation*}
742: \caption{Morphism $\eps_A$}\label{feps}
743: \end{figure}
744: 
745: 
746: 
747: 
748: We will say that a $\C$-morphism $f\colon V\ttt W\to \one$ defines a
749: {\em non-degenerate pairing} if $f$ defines an isomorphism $V\simeq
750: W^*$, i.e. there exists a map $g\colon \one\to W\ttt V$ such that
751: $f,g$ satisfy the rigidity axioms. 
752: 
753: 
754: \begin{definition} \label{drigid}
755: A $\C$-algebra $A$ is called {\em rigid} if the map 
756: \begin{equation}\label{erigid}
757: e_A\colon A\ttt A\xxto{\mu}A\xxto{\eps_A}\one
758: \end{equation}
759: is a non-degenerate pairing  and $\dim_\C A\ne 0$. 
760: \end{definition}
761: 
762: If $A$ is rigid, then there is a unique morphism $i_A\colon \one\to
763: A\ttt A$ such that $e_A, i_A$ satisfy the rigidity axioms: the
764: compositions $A\xxto{\id\ttt i_A}A\ttt A\ttt A\xxto{e_A\ttt \id}A$,
765: $A\xxto{i_A\ttt \id }A\ttt A\ttt A\xxto{\id \ttt e_A}A$ are equal to
766: identity (uniqueness follows from a well-known fact that the dual
767: object is unique up to a unique isomorphism).  We will frequently use
768: the following simple lemma.
769: 
770: \begin{lemma}\label{lsymmetry}
771: Both $e_A$, $i_A$ are symmetric: 
772: \begin{align*}
773: &e_A R_{AA}=e_A,\\
774: &R_{AA}i_A=i_A.
775: \end{align*}
776: \end{lemma}
777: \begin{proof} 
778:   The identity for $e_A$ immediately follows from commutativity of
779:   multiplication. For $i_A$, it suffices to prove that $R_{AA}i_A$
780:   satisfies the rigidity axioms, i.e.  both compositions below are
781:   equal to identity
782: \begin{align*}
783: &A\xxto{\id\ttt R_{AA}i_A}A\ttt A\ttt A\xxto{e_A\ttt \id}A,\\
784: &A\xxto{R_{AA}i_A\ttt \id }A\ttt A\ttt A\xxto{\id \ttt e_A}A.
785: \end{align*}
786: To prove the first identity, we represent the
787: composition by a graph and manipulate it as shown in \firef{ficommut}.
788: The second identity is proved in the same manner.
789: 
790: \begin{figure}[h]%% 30%
791: \begin{equation*}
792: \fig{icommuta}=
793: \fig{icommutb}=
794: \fig{icommutc}=\id_A
795: \end{equation*}
796: \caption{Proof of $R_{AA}i_A=i_A$.}\label{ficommut}
797: \end{figure}
798: \end{proof}
799: 
800: 
801: 
802: Finally, we also need to discuss the following subtle point. 
803: The map $e_A$ allows us to identify $A\simeq A^*$ and thus, we can
804: also identify $A\simeq A^{**}$. On the other hand, in any balanced
805: rigid braided category one has a canonical identification $\de_V\colon
806: V\simeq V^{**}$. It is natural to ask whether these two
807: identifications coincide. The answer is given by the following lemma.
808: 
809: \begin{lemma}\label{tha=1}
810:   Let $A$ be a rigid algebra. Then the morphism $A\simeq A^{**}$
811:   defined by $e_A$ coincides with $\de_A\colon A\to A^{**}$ iff
812:   $\th_A=\id$. 
813: \end{lemma} 
814: \begin{proof}
815: Recalling the relation between the twists $\th_V$ and $\de_V$ (see,
816: e.g., \cite[Section 2.2]{BK}), we see
817: that the statement of the lemma is equivalent to the following
818: equation:
819: $$
820: \fig{thA}=\th_A.
821: $$
822: But it easily follows from symmetry of $e_A$ that the right hand side
823: is the identity morphism. 
824: \end{proof}
825: This lemma shows that if $A$ is rigid, $\th_A=\id$, then  we can
826: identify $A\simeq A^*$ so that the canonical  morphisms $A\ttt
827: A^*\to \one, A^*\ttt A\to \one$ are both given by  $e_A$, and the
828: morphisms $\one\to A\ttt A^*, \one \to A^*\ttt A$ are both given by
829: $i_A$. As usual, we will use ``cap'' and ``cup'' to denote $e_A,
830: i_A$  in the figures. Then the statement of \leref{lsymmetry} can be
831: graphically presented  as follows: 
832: 
833: \begin{equation}\label{esymmetry}
834: \fig{frcap}=\fig{fcap}\qquad\qquad
835: \fig{frcup}=\fig{fcup}
836: \end{equation}
837: 
838: 
839: 
840: 
841: We will frequently use the following easy lemma.
842: \begin{lemma}\label{lnoose}
843: If $A$ is a rigid $\C$-algebra, $\th_A=\id$, then 
844: \begin{equation}
845: \fig{noose.eps}=\dim A \fig{one.eps}
846: \end{equation}
847: \end{lemma}
848: The proof is immediate if we note that both sides are morphisms $\one
849: \to A$ and by uniqueness of unit axiom must be proportional. 
850: 
851: 
852: 
853: 
854: 
855: \begin{theorem}\label{trigid}
856:   If $\C$ is a rigid category, $A$ --- a rigid $\C$-algebra,
857:   $\th_A=\id$, then the categories $\Rep A, \Rep^0 A$ are rigid.
858: \end{theorem}
859: \begin{proof}
860: Let $(V, \mu_V)\in \Rep A$. Define the dual object $(V^*, \mu_{V^*})$
861: as follows: $V^*$ is the dual of $V$ in $\C$ and $\mu_{V^*}$ is defined
862: by \firef{fvdual}.
863: 
864: \begin{figure}[h]%% 30%
865: \begin{equation*}
866: \fig{vdual}
867: \end{equation*}
868: \caption{Definition of dual object in $\Rep A$.}\label{fvdual}
869: \end{figure}
870: 
871: This definition implies the following identities:
872: 
873: \begin{figure}[h]%% 30%
874: \begin{equation*}
875: \fig{dualidenta}=\fig{dualidentb}
876: \qquad \fig{dualidentc}=\fig{dualidentd} 
877: \end{equation*}
878: \caption{}\label{fdualident}
879: \end{figure}
880: 
881: 
882: 
883: 
884: Define now the maps $\tilde i_V\in\Hom_{A}(A, V\tta V^*), \tilde e_V\in
885: \Hom_A(V^*\tta  V, A)$ by \firef{frigid} (we leave it to the reader to
886: check that these formulas indeed define morphisms in $\Rep A$).
887: 
888: 
889: 
890: \begin{figure}[h]%% 30%
891: \begin{equation*}
892: \tilde i_V=\frac{1}{\dim A}\fig{rigid1.eps}\qquad
893: \tilde e_V=\fig{rigid2.eps}
894: \end{equation*}
895: \caption{Rigidity maps in $\Rep A$}\label{frigid}
896: \end{figure}
897: 
898: 
899: 
900: It is easy to check by  using identities in \firef{fdualident} 
901: and isomorphisms $A\tta V\simeq V$ defined in the proof of \thref{ttensor}
902: that these two maps satisfy the rigidity axioms. 
903: \end{proof}
904: 
905: \begin{lemma}
906: Let $A$ be a rigid $\C$-algebra. Then
907: \begin{enumerate}
908: \item $F$ and $G$ are 2-sided adjoints of each other: in addition to
909:   results of \thref{tfunctors}, we also have canonical isomorphisms  
910: $$
911: \Hom_A(X, F(V))=\Hom_\C(G(X),V), \qquad V\in \C, X\in \Rep A.
912: $$
913: \item 
914: $F(V^*)\simeq (F(V))^*$.
915: \end{enumerate}
916: \end{lemma}
917: 
918: \begin{proof}
919: To prove part (1), we construct linear maps between $\Hom_A(X, F(V))$
920: and $\Hom_\C(G(X),V)$ as shown in \firef{fadjoint2}; we leave it to
921: the reader to check that these maps are inverse to each other. 
922: 
923: 
924: \begin{figure}[h]%% 30%
925: \begin{equation*}
926: \fig{adjoint2.eps}
927: \end{equation*}
928: \caption{Identifications $\Hom_A(X, F(V))=\Hom_\C(G(X), V)$.
929:         Here  $\ph\in\Hom_A(X, F(V)), \ \Phi\in\Hom_\C(G(X), V)$.}
930:         \label{fadjoint2}
931: \end{figure}
932: 
933: 
934: To prove (2), note that as object of $\C$, $(F(V))^*=V^*\ttt
935: A^*=V^*\ttt A$, where we used rigidity to identify $A=A^*$. Consider
936: the morphism $R_{AV^*}\colon A\ttt V^*\to V^*\ttt A$. Again, we leave
937: it to the reader to check that this morphism is actually a morphism of
938: $A$-modules $F(V^*)\to (F(V))^*$. 
939: 
940: This shows that $\Rep A$ is rigid. To prove rigidity of $\Rep^0 A$, it
941: suffices to show that for $X\in \Rep^0 A, X^*\in \Rep^0 A$ which
942: easily follows from the definition of $\Rep^0 A$ and the definition of
943: dual object given by \firef{fvdual}. 
944: \end{proof}   
945: 
946: 
947: 
948: Finally, we need to check discuss whether $\Rep A, \Rep^0 A$ are
949: balanced. Recall that 
950: balancing in a rigid braided tensor category is a system of functorial
951: isomorphisms $\de_V\colon V\simeq V^{**}$ satisfying conditions 
952: \begin{equation}\label{edev}
953: \begin{aligned}
954: {}&\de_{V\ttt W}=\de_V\ttt \de_W,\\
955: &\de_\one =\id,\\
956: &\de_{V^*}=(\de_V^*)^{-1},
957: \end{aligned}
958: \end{equation}
959: where we have used canonical identifications $(V\ttt W)^*=W^*\ttt V^*$
960: and for $F\colon X\to Y$, $f^*\colon Y^*\to X^*$ is the adjoint
961: morphism. 
962: 
963: This is equivalent to defining a system of functorial morphisms $\th_V
964: \colon V\to V$ (twists), satisfying 
965: \begin{equation}\label{ebalancing}
966: \begin{aligned}
967: {}&\th_{V\ttt W}=R_{WV}R_{VW}\th_V\ttt \th_W,\\
968: &\th_\one=\id,\\
969: &\th_{V^*}=(\th_V)^*.
970: \end{aligned}
971: \end{equation}
972: (see, for example,  \cite[Section 2.2]{BK}). 
973: 
974: 
975: 
976: 
977: 
978: \begin{theorem}\label{tbalanced}
979:   Let $\C$ be a rigid balanced braided category, and $A$---a rigid
980:   $\C$-algebra, $\th_A=\id$. Then
981: 
982: \begin{enumerate}
983: \item $\Rep^0 A=\{V\in \Rep A\st \th_V \text{ is an $A$-morphism }\}$.
984: \item $\Rep^0A$ is a rigid balanced braided category, with $\th$
985:   inherited from $\C$.
986: \item For any $V\in \Rep A$, the morphism $\de_V\colon V\to V^{**}$
987:   is an $A$-morphism.  
988: \end{enumerate}
989: \end{theorem}
990: \begin{proof}
991:   Part (1) follows from $\th_{A\ttt V}=R_{VA}R_{AV}\th_A\ttt \th_V$
992:   and $\th_A=\id$; (2) immediately follows from (1). 
993:   To prove (3), it suffices to prove that $\de_V^{-1}\colon V^{**}\to V$ is an
994:   $A$-morphism. We can rewrite $\de_V^{-1}$ in terms of $\th$ as
995:   follows (see \cite[Section 2.2]{BK}): 
996: $$
997: \de_V^{-1}=\fig{dev}
998: $$
999: 
1000: It now follows from the identities shown in \firef{fdev} 
1001:  (which uses \eqref{ebalancing} and
1002: identities from \firef{fdualident}) that $\de^{-1}$ is an
1003: $A$-morphism. 
1004: 
1005: \begin{figure}[ht]
1006: $$
1007: \fig{deva}=\fig{devb}=\fig{devc}=\fig{devd}
1008: $$
1009: \caption{Proof that $\de_V^{-1}$ is an $A$-morphism}\label{fdev}
1010: \end{figure}
1011: 
1012: \end{proof}
1013: 
1014: This theorem shows that the category $\Rep A$, which is a rigid
1015: monoidal category, while not braided, does have a system of functorial
1016: morphisms $\de_X\colon X\to X^{**}$ satisfying \eqref{edev}. Such
1017: categories are sometimes called ``pivotal''; in such a category, one
1018: can define for every object two numbers, its ``left'' and ``right''
1019: dimension   (see, e.g., \cite{BW}).  We will denote by 
1020: $\dim_A X$ the ``left'' dimension of an object $X\in \Rep A$. 
1021: 
1022: 
1023: 
1024: \begin{theorem}\label{tdim0}
1025:   Let $\C$ be a rigid balanced braided category, and $A$---a rigid
1026:   $\C$-algebra such that $\th_A=\id_A$. Then for every $X,Y \in \Rep
1027:   A$, $\dim_{A}(X\tta Y)=\dim_{A}(X)\dim_{A}(Y)$ and 
1028: \begin{align*}
1029: &\dim_{A}(X)=\frac{\dim_{\C}(X)}{\dim_{\C}A},\\
1030: &\dim_{A}(F(V))=\dim_\C(V).
1031: \end{align*}
1032: \end{theorem}
1033: \begin{proof}
1034: Formula $\dim_{A}(X\tta Y)=\dim_{A}(X)\dim_{A}(Y)$ holds in any
1035: pivotal category and can be easily deduced from \eqref{edev}.
1036: 
1037: Using definition of rigidity morphisms in $\Rep A$ shown in
1038: \firef{frigid},  we see that
1039: $\dim_A X$ is defined by the following identity:
1040: $$
1041: \frac{1}{\dim A}\fig{adim} =(\dim_A X) \id_A.
1042: $$
1043: Both sides are $A$-morphisms $A\to A$. Composing them with $\iota_A,
1044: \eps_A$,  we get $\frac{1}{\dim A}\dim_\C X=\dim_A X$. 
1045: 
1046: Applying this to $X=F(V)=V\ttt A$, we get 
1047: $$
1048: \dim_A F(V)=\frac{1}{\dim
1049:   A} \dim_\C F(V)= \frac{1}{\dim A} (\dim  A)\cdot(\dim V)=\dim V.
1050: $$ 
1051: \end{proof}
1052: 
1053: 
1054: As a useful corollary, we get the following result:
1055: \begin{equation}\label{dimA}
1056: \dim_\C(X\tta Y)=\frac{\dim_\C(X)\dim_\C(Y)}{\dim_\C(A)}.
1057: \end{equation}
1058: 
1059: \begin{remark}
1060: In the theorem above, we could have used ``right'' dimension instead of
1061: the ``left'' dimension (this would require minor change in the
1062: proof). Thus, we see that both left and right dimension of $X\in \Rep
1063: A$ are equal to  $\frac{\dim_{\C}(X)}{\dim_{\C}A}$; in particular,
1064: they are equal to each other. In a similar way, one could prove that
1065: for each $A$-morphism $F\colon X\to X$, its left and right dimension
1066: are equal; in terminology of \cite{BW}, $\Rep A$ is a {\em spherical}
1067: category. 
1068: \end{remark}
1069: 
1070: 
1071: 
1072: For future use, we note the following somewhat unusual result. 
1073: \begin{lemma}\label{lideals}
1074:   Let $\C$ be rigid and $A$ -- a $\C$-algebra such that $\th_A=\id,
1075:   \dim_\C A\ne 0$. Then $A$ is a rigid $\C$-algebra iff $A$ is simple
1076:   as an $A$-module.
1077: \end{lemma}
1078: \begin{proof}
1079:   Let $A$ be rigid; assume $I\subset A$ is a submodule. By rigidity,
1080:   $\one\subset \mu(A\ttt I)$. On the other hand, since $I$ is a
1081:   submodule, this implies that $\one \subset I$. By unit axiom, this
1082:   implies $I=A$.
1083: 
1084: Conversely, assume that $A$ is simple as $A$-module. Consider $A^*\in\C$
1085: and define on it the action of $A$ as in \thref{trigid}. Then one
1086: easily sees that the morphism 
1087: $$
1088: A\xxto{\id\ttt i_A} A\ttt A\ttt A^*\xxto{e_A\ttt \id}A^*,
1089: $$
1090: where $e_A$ is as in \eqref{erigid} is a morphism of $A$-modules.
1091: On the other hand, usual arguments show that if $A$ is a simple
1092: $A$-module, then so is $A^*$. Thus, such a map is either zero
1093: (impossible because of the  unit axiom) or an isomorphism. 
1094: \end{proof}
1095: 
1096: Finally, recall that we defined $X\tta Y$ as a quotient of $X\ttt
1097: Y$. It turns out that in the rigid case, $X\tta Y$ can also be
1098: described as a sub-object of $X\ttt Y$ and thus, as a direct summand. 
1099: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1100: \begin{lemma}\label{lsummand}
1101:   If $A$ is a rigid $\C$-algebra, $X,Y\in \Rep A$ and $Q\colon X\ttt
1102:   Y\to X\ttt Y$ is as shown in \firef{fQ}, then $Q^2=Q$ and $\ker
1103:   Q=\ker(X\ttt Y\to X\tta Y)$.
1104: \end{lemma}
1105: \begin{figure}[h]
1106: \begin{equation*}
1107: Q=\frac{1}{\dim A}\quad\fig{Q}
1108: \end{equation*}
1109: \caption{Projector on $X\tta Y\subset X\ttt Y$}\label{fQ}
1110: \end{figure}
1111: Proof of this lemma is left to the reader as an exercise. 
1112: 
1113: 
1114: \begin{corollary}
1115:   If $A$ is a rigid $\C$-algebra, then one has a canonical direct sum
1116:   decomposition $X\ttt Y=Z\oplus X\tta Y$ for some $Z\in \C$.
1117: \end{corollary}
1118: Indeed, it suffices to take $Z=\ker Q$ and identify $X\tta Y=\im Q$.
1119: 
1120: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1121: \section{Example: groups and subgroups}\label{sgroup}
1122: In this section we discuss an important example of the general setup
1123: discussed above. Namely, let $G$ be a group such that the category
1124: $\C$ of finite-dimensional complex representations of $G$ is
1125: semisimple (for example, $G$ is a finite group or $G$ is a reductive
1126: Lie group). Also, for a finite set $X$ let $F(X)$ be the space of
1127: complex-valued functions on $X$.
1128: 
1129: \begin{theorem} 
1130:   If $H\subset G$ is a subgroup of finite index, then
1131:   the space $A=F(G/H)$ of functions on $G/H$ is a semisimple
1132:   $\C$-algebra and $\Rep A$ is equivalent to the category $\Rep H$ of
1133:   representations of $H$; under this equivalence the functors $F$ and
1134:   $G$ are identified with the restriction and induction functor
1135:   respectively:
1136: \begin{align*}
1137: F=\Res^G_H\colon& \Rep G \to \Rep H,\\
1138: G=\Ind_H^G\colon& \Rep H\to \Rep G
1139: \end{align*}
1140: 
1141: 
1142: \end{theorem}
1143: \begin{proof}
1144:   By definition, an object of $\Rep A$ is a $G$-module $V$ with a
1145:   decomposition $V=\oplus_{x\in G/H} V_x$ such that $gV_x=V_{gx}$, and
1146:   tensor product in $\Rep A$ is given by $(V\tta W)_x=V_x\ttt W_x$.
1147:   Define functor $\Rep A\to \Rep H$ by $\oplus V_x\mapsto V_1$ and
1148:   $\Rep H\to \Rep A$ by $E\mapsto \Ind_H^G E$ (note that it follows
1149:   from definition of the induced module that $V=\Ind E$ has a natural
1150:   decomposition $V=\oplus_{x\in G/H}V_x$). It is trivial to check that
1151:   these functors preserve tensor product and  are inverse to each
1152:   other. 
1153: \end{proof}
1154: 
1155: \begin{theorem}
1156: For $\C=\Rep G$, any rigid $\C$-algebra is of the form $F(G/H)$
1157: for some subgroup $G$ of finite index. 
1158: \end{theorem}
1159: \begin{proof}
1160:   First, a $\C$-algebra is just a commutative associative algebra over
1161:   $\Cset$ on which $G$ acts by automorphisms. Next, if $A$ is rigid,
1162:   then $A$ is semisimple as a commutative associative algebra over
1163:   $\Cset$.  Indeed, let $N$ be the radical of $A$; then $N$ is
1164:   invariant under the action of $G$ and thus is an ideal in $A$ in the
1165:   sense of $\C$-algebras. By \leref{lideals}, $N=0$. 
1166:  
1167:   Thus, $A$ is the algebra of functions on a finite set $X$ (which can
1168:   be described as the set of primitive idempotents of $A$) and $G$
1169:   acts by permutations on $X$. Since $\Cset$ appears in decomposition
1170:   of $A$ as $G$-module with multiplicity one, this implies that the
1171:   action of $G$ on $X$ is transitive, so $X=G/H$.
1172: \end{proof}
1173: 
1174: % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1175: % \section{Braided categories}
1176: % In this sections, we consider the case when $\C$ is a braided tensor
1177: % category, i.e. when $R^2_{VW}$ does not have to be identity. In this
1178: % case, we still can define associative commutative algebra in $\C$,
1179: % replacing commutativity condition with the following one:
1180: % \begin{equation}
1181: % \mu \circ R_{AA}=\mu\circ R^{-1}_{AA}=\mu.
1182: % \end{equation}
1183: % We define the category $\Rep A$ in the same way as in the symmetric
1184: % case. Most properties can be trivially generalized if we also use the
1185: % following convention: in all the formulas where one needs to
1186: % interchange an object $V\in \Rep A$ with $A$, we use $R_{AV},
1187: % R_{AV}^{-1}$ but not $R_{VA}$. In the language of figures, this means
1188: % that we draw the strand corresponding to $A$ always on top of strands
1189: % corresponding to $V$. The same arguments as in the previous section
1190: % give the following theorem. 
1191: 
1192: 
1193: 
1194: % \begin{theorem}\label{tbraided}
1195: %   Let $\C$ be a semisimple braided tensor category, $A$---a
1196: %   $\C$-algebra. Then the category $\Rep A$ is an abelian monoidal
1197: %   category with unit object $A$; the functors $F\colon \C\to \Rep A$,
1198: %   $G\colon \Rep A\to \C$ have the same properties as in
1199: %   \thref{tfunctors}. If $\C$ is rigid and $A$ is a rigid algebra in
1200: %   $\C$ then $\Rep A$ is rigid. 
1201: % \end{theorem}
1202: 
1203: % However, there are also some notable differences between symmetric and
1204: % braided case. Most importantly, in the braided case it is impossible
1205: % to define a structure of a braided tensor category on $\Rep
1206: % A$. Instead, we have to consider a smaller category.
1207: 
1208: % \begin{definition}\label{drep0}
1209: % $\Rep^0A$ is the full subcategory in $\Rep A$ consisting of objects
1210: % $(V, \mu_V)$ such $\mu_V\circ R_{VA}R_{AV}=\mu_V$. 
1211: % \end{definition}
1212: 
1213: % Later we will justify this definition by showing that if $\C$ is a
1214: % category of representation of some vertex operator algebra, and $A$ is
1215: % an extended vertex operator algebra, then $\Rep^0A$ (and not $\Rep
1216: % A$!) is exactly the category of representations of the vertex operator
1217: % algebra $A$. 
1218: 
1219: 
1220: % \begin{theorem}\label{trigid0}
1221: % $\Rep^0A$ is a braided tensor category with unit object $A$; if $\C$
1222: % is rigid and $A$ is rigid, then $\Rep^0 A$ is rigid. 
1223: % \end{theorem}
1224: % \begin{proof}
1225: % Braiding is inherited from $\C$; it immediately follows from the
1226: % definitions that if $X,Y\in \Rep^0$ then $R_{XY}$ is well defined as
1227: % an operator $X\tta Y\to Y\tta X$ and commutes with the action of
1228: % $A$. Proof of rigidity is completely parallel to the symmetric case
1229: % (see proof of \thref{trigid}); the only new part is that we need to
1230: % show that for $V\in \Rep^0A$, $V^*$ is also in $\Rep^0A$. This is easy
1231: % to show by using the explicit formula for rigidity morphisms from
1232: % \firef{frigid}. 
1233: % \end{proof} 
1234: 
1235: % In the braided case,  one can not define the functor $F\colon \C\to \Rep^0A$;
1236: % however, it is possible to define $F(V)$ for $V\in \C$ such that
1237: % $R_{AV}^2=\id$. This functor will still have the property $F(V\ttt
1238: % W)=F(V)\tta F(W), F(\one)=A$. And of course, one still has the
1239: % ``forgetful'' functor $G\colon \Rep^0A \to \C$.  
1240: 
1241: % Finally, let us consider the balancing structure. Recall that
1242: % balancing in a rigid braided tensor category is a system of functorial
1243: % isomorphisms $\de_V\colon V\simeq V^{**}$ satisfying conditions 
1244: % \begin{equation}\label{edev}
1245: % \begin{aligned}
1246: % {}&\de_{V\ttt W}=\de_V\ttt \de W,\\
1247: % &\de_\one =\id,\\
1248: % &\de_{V^*}=(\de_V^*)^{-1},
1249: % \end{aligned}
1250: % \end{equation}
1251: % where we have used canonical identifications $(V\ttt W)^*=W^*\ttt V^*$
1252: % and for $F\colon X\to Y$, $f^*\colon Y^*\to X^*$ is the adjoint
1253: % morphism. 
1254: 
1255: % This is equivalent to defining a system of functorial morphisms $\th_V
1256: % \colon V\to V$ (twists), satisfying 
1257: % \begin{equation}\label{ebalancing}
1258: % \begin{aligned}
1259: % {}&\th_{V\ttt W}=R_{WV}R_{VW}\th_V\ttt \th_W,\\
1260: % &\th_\one=\id,\\
1261: % &\th_{V^*}=(\th_V)^*.
1262: % \end{aligned}
1263: % \end{equation}
1264: % (see, for example,  \cite[Section 2.2]{BK}). 
1265: 
1266: 
1267: % \begin{theorem}\label{tbalanced}
1268: %   Let $\C$ be a rigid balanced braided category, and $A$---a rigid
1269: %   $\C$-algebra. Assume in addition that $\th_A=\id$. Then
1270: 
1271: % \begin{enumerate}
1272: % \item $\Rep^0 A=\{V\in \Rep A\st \th_V \text{ is an $A$-morphism }\}$.
1273: % \item $\Rep^0A$ is a rigid balanced braided category, with $\th$
1274: %   inherited from $\C$.
1275: % \item For any $V\in \Rep A$, the morphism $\de_V\colon V\to V^{**}$
1276: %   is an $A$-morphism.  
1277: % \end{enumerate}
1278: % \end{theorem}
1279: % \begin{proof}
1280: %   Part (1) follows from $\th_{A\ttt V}=R_{VA}R_{AV}\th_A\ttt \th_V$
1281: %   and $\th_A=\id$; (2) immediately follows from (1), and (3) can be
1282: %   proved in the same way as (1) using the formula expressing $\de$ via
1283: %   $\th$ and $R$ (see \cite[Section 2.2]{BK}). 
1284: % \end{proof}
1285: 
1286: % This theorem shows that the category $\Rep A$, which is a rigid
1287: % monoidal category, while not braided, does have a system of functorial
1288: % morphisms $\de_X\colon X\to X^{**}$ satisfying \eqref{edev}. (Some
1289: % authors call such categories {\em spherical}.)  This allows, among
1290: % other things, to define the notion of dimension of an 
1291: % object in $\Rep A$; as before, we denote so defined dimension by
1292: % $\dim_A X$   The following theorem is an analogue of \coref{cdim}. 
1293: 
1294: % \begin{theorem}\label{tdim0}
1295: %   Let $\C$ be a rigid balanced braided category, and $A$---a rigid
1296: %   $\C$-algebra such that $\th_A=\id_A$. Then for every $X,Y \in \Rep
1297: %   A$, $\dim_{A}(X\tta Y)=\dim_{A}(X)\dim_{A}(Y)$ and 
1298: % \begin{align*}
1299: % &\dim_{A}(X)=\frac{\dim_{\C}(X)}{\dim_{\C}A},\\
1300: % &\dim_{A}(F(V))=\dim_\C(V).
1301: % \end{align*}
1302: % \end{theorem}
1303: % \begin{proof}
1304: % Formula $\dim_{A}(X\tta Y)=\dim_{A}(X)\dim_{A}(Y)$ holds in any
1305: % spherical category and can be easily deduced from \eqref{edev}. The
1306: % remianing parts of the proof are completely parallel to the proof of
1307: % \coref{cdim}.  
1308: % \end{proof}
1309: 
1310: % As a useful corollary, we get the following result:
1311: % \begin{equation}\label{dimA}
1312: % \dim_\C(X\tta Y)=\frac{\dim_\C(X)\dim_\C(Y)}{\dim_C(A)}.
1313: % \end{equation}
1314: 
1315: 
1316: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1317: \section{Semisimplicity}
1318: As before, we let $\C$ be a braided tensor category. 
1319: 
1320: \begin{definition}\label{dsemisimple}
1321: A $\C$-algebra is called semisimple if $\Rep A$ is semisimple.
1322: \end{definition}
1323: 
1324: % For future use, we note that because of the ``uniqueness of unit''
1325: % axiom, semisimplicity implies the following somewhat unusual result.
1326: % \begin{lemma}\label{lideals}
1327: % If $A$ is a semisimple $\C$-algebra, then it has no non-trivial
1328: % ideals: if $I\subset A$ is a sub-object such that $\mu(A\ttt
1329: % I)\subset I$, then $I=0$ or $I=A$. 
1330: % \end{lemma}
1331: % \begin{proof}
1332: % Consider $A$ as a left module over itself; then semisimplicity implies
1333: % that $A=I\oplus J$ for some ideal $J$. Since $\one$ appears in $A$
1334: % with multiplicity one, we either have $\one\subset I$ or $\one\subset
1335: % J$. In both cases, this is incompatible with the unit axiom. 
1336: % \end{proof}
1337: 
1338: We will be mostly interested in the case when $\C$ is rigid and
1339: balanced. In this case, semisimplicity of  $\Rep A$ implies
1340: semisimplicity of  $\Rep^0 A$. 
1341: 
1342: \begin{theorem}\label{tssrep0}
1343: Let $\C$ be rigid balanced, and $A$ --- a semisimple
1344: $\C$-algebra with $\th_A=\id$. Then 
1345: \begin{enumerate}
1346: \item If $X\in \Rep A$ is simple, then $X\in \Rep^0 A$ iff
1347:   $\th_X=c\cdot \id$. 
1348: 
1349: \item $\Rep^0 A$ is semisimple, with simple objects $X_{\pi}$ where
1350:   $X_\pi$ is a simple object in $\Rep A$ such that $\th_X=c\cdot \id$.
1351: \end{enumerate}
1352: \end{theorem}
1353: \begin{proof}
1354: Immediately follows from \thref{tbalanced} and the fact that for a
1355: simple object $X\in \Rep A$, $\Hom_A(X,X)=\Cset$. 
1356: \end{proof}
1357: 
1358: The main result of this section is the following theorem.
1359: 
1360: \begin{theorem}\label{trigidsemisimple}
1361: Let $\C$ be rigid, and $A$ --- a rigid $C$-algebra. Then $A$  is semisimple. 
1362: \end{theorem}
1363: 
1364: \begin{proof}
1365: The proof is based on the following lemma.
1366: \begin{lemma}[\cite{B}]
1367: If $A$ is rigid, then every $X\in \Rep A$ is a direct summand in
1368: $F(V)$ for some $V\in \C$.
1369: \end{lemma}
1370: \begin{proof}[Proof of the lemma]
1371: Consider the map $\mu\colon A\ttt A\to A$. It is surjective and
1372: is a morphism of $A$-modules. Moreover, both $A$ and $A\ttt A$ have
1373: canonical structures of $A$-bimodules, and $\mu$ is a morphism of
1374: $A$-bimodules (we leave the definition of $A$-bimodule as an exercise
1375: to the reader). This map has one-sided inverse: if we define
1376: $\ph\colon A\to A\ttt A$ by
1377: $$
1378: \ph=\frac{1}{\dim A}\fig{split}
1379: $$
1380: then $\ph$ is a morphism of $A$-bimodules and it immediately
1381: follows from \leref{lnoose} that $\mu\ph=\id_A$. Thus, $A\ttt A$
1382: splits: we can write 
1383: $$
1384: A\ttt A\simeq A\oplus Z
1385: $$ 
1386: for some $A$-bimodule $Z$ so that under this isomorphism, $\mu$ is the
1387: projection on the first summand. 
1388: 
1389: Therefore, $A\tta X\simeq X$ is a direct summand of 
1390: $(A\ttt A)\tta X= A\ttt (A\tta X)=A\ttt X=F(G(X))$. 
1391: \end{proof}
1392: 
1393: From this lemma, the proof is easy. Indeed, it easily follows from
1394: exactness of $G$ and adjointness of $F$ and $G$ (see
1395: \thref{tfunctors}) that for every $V\in \C$, $F(V)$ is a projective
1396: object in $\Rep A$. Since a direct summand of a projective object  is
1397: projective, the lemma implies that every $X\in \Rep A$ is projective
1398: and thus, $\Ext^1(X,Y)=0$ for every $X, Y\in \Rep A$.
1399: \end{proof}
1400: 
1401: \begin{remark}
1402:   Morally, this theorem is parallel to the following well known result
1403:   in Lie algebra theory: if the Killing form on $\g$ is non-degenerate,
1404:   then the category of finite-dimensional representations of $\g$ is
1405:   semisimple (this is combination of Cartan's criterion of
1406:   semisimplicity and Weyl's complete reducibility theorem). The proof,
1407:   of course, is completely different.
1408: \end{remark}
1409: 
1410: 
1411: 
1412: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1413: \section{Modularity}
1414: Recall that a semisimple balanced rigid braided category $\C$ is
1415: called modular if it has finitely many isomorphism classes of simple
1416: objects $V_i, i\in I, |I|<\infty$  and the matrix $\tilde s_{ij}$
1417: defined by \firef{fsij} is non-degenerate. (We will also use numbers
1418: $\tilde s_{VW}$ defined in the same way as $\tilde s_{ij}$ but with
1419: $V_i$ replaced by $V$, $V_j$ replaced by $W$). 
1420: 
1421: 
1422: \begin{figure}[h]%% 30%
1423: \begin{equation*}
1424: \fig{sij.eps}
1425: \end{equation*}
1426: \caption{Matrix $\tilde s_{ij}$}\label{fsij}
1427: \end{figure}
1428: 
1429: In this case, it is known that the matrices 
1430: \begin{equation}\label{st}
1431: \begin{aligned} 
1432: s_{ij}&=\frac{1}{D}\tilde s_{ij},\\
1433:  t_{ij}&=\frac{1}{\zeta} \de_{ij}\th_i,
1434: \end{aligned}
1435: \end{equation}
1436: where $D, \zeta$ are some
1437: non-zero numbers, satisfy the relations of $SL_2(\Z)$: $(st)^3=s^2,
1438: s^4=\id$. These matrices are naturally interpreted as matrices of some
1439: operators $s, t$ acting on $K(\C)$; for example, the operator
1440: $s=\frac{1}{D}\tilde s$ where 
1441: \begin{equation}\label{sv}
1442: \tilde s[V]=\sum \tilde s_{VV_j}[V_j]
1443: \end{equation}
1444: where $[V]$ is the class in $K$ of $V\in \C$. 
1445: 
1446: We also note that the numbers $D, \zeta$ appearing in \eqref{st} are
1447: determined uniquely up to a simultaneous change of sign. The number
1448: $D=(s_{00})^{-1}$ is sometimes called the {\em rank} of $\C$. If $\C$
1449: is Hermitian category, it is possible to choose $D$ to be a positive
1450: real number. In modular tensor categories coming from conformal field
1451: theory, the number $\zeta$ is given by $\zeta=e^{2\pi\i c/24}$, where
1452: $c$ is the (Virasoro) central charge of the theory. 
1453: 
1454: 
1455: In this section we assume that $\C$ is a modular tensor category and
1456: $A$ is a rigid $\C$-algebra, which satisfies $\th_A=\id$; by
1457: \thref{trigidsemisimple}, this implies that $A$ is semisimple.  We
1458: denote isomorphism classes of simple objects in $\Rep A$ by $X_\pi,
1459: \pi\in\Pi$, and let $K(A)$ be the fusion algebra of $\Rep A$.
1460: Similarly, set of simple objects in $\Rep^0 A$ is $\Pi^0\subset \Pi$
1461: (see \thref{tssrep0}) and the fusion algebra of $\Rep^0 A$ is
1462: $K^0(A)\subset K(A)$. We will denote by $P\colon K(A)\to K^0(A)$ the
1463: projection operator: $P([X_\pi])=[X_\pi]$ if $\pi\in \Pi^0$ and
1464: $P([X_\pi])=0$ otherwise. 
1465:  
1466: Define operator $\tilde s^A\colon K^0(A)\to K^0(A)$ in
1467: the same way as for $\C$ but replacing $V_j$ by
1468: $X_\pi$ and using rigidity morphisms in $\Rep^0A$ rather than in $\C$.
1469: 
1470: \begin{theorem}\label{tmodular} Let  $G\colon K^0(A)\to K$ be the map
1471:   induced by the functor $G$ from \thref{tfunctors}, and let
1472:   $F^0\colon K\to K^0(A)$ be the composition $PF$, where $P\colon
1473:   K(A)\to K^0(A)$ is the projection operator defined above. Then $G,
1474:   F^0$  commute with the action of $\tilde s, \tilde t$ up
1475: to a constant:
1476: \begin{alignat*}{2}
1477: G(\dim A) \tilde s^A&=\tilde s G, \qquad & 
1478:                   F^0 \tilde s&= (\dim A) \tilde s^AF^0  \\ 
1479: G\tilde t^A&=\tilde t G,          & 
1480:                   F^0\tilde t &=\tilde t^A F^0.
1481: \end{alignat*} 
1482: \end{theorem}
1483: 
1484: To prove this theorem, we will need several technical lemmas. 
1485: 
1486: \begin{lemma}\label{lstilde}
1487: For $X,Y\in \Rep^0 A$, the number $\tilde s^A_{XY}$ is given by 
1488: 
1489: \begin{equation}
1490: \tilde s^A_{XY}=\frac{1}{(\dim A)^2}\fig{stilde.eps}
1491: \end{equation}
1492: 
1493: \end{lemma}
1494: \begin{proof}
1495: Recalling the definition of rigidity isomorphisms in $\Rep A$ and
1496: isomorphisms $A\tta A\isoto A$, we see that  $\tilde s^A_{XY}$ is
1497: given by 
1498: \begin{equation}
1499: \frac{1}{(\dim A)^2}\fig{stilde2.eps}=\tilde s^A_{XY}\id_A
1500: \end{equation}
1501: 
1502: Restricting both sides to $\one\subset A$, we get 
1503: \begin{equation}
1504: \tilde s^A_{XY}=\frac{1}{(\dim A)^2}\fig{stilde3.eps}
1505: \end{equation}
1506: which is easily seen to be equivalent to the statement of the lemma.
1507: \end{proof}
1508: 
1509: 
1510: \begin{lemma}\label{lproj}
1511: Let $X_\pi\in \Rep A$ be simple. Define $P_\pi\colon X_\pi\to X_\pi$
1512: by 
1513: \begin{equation}
1514: P_\pi= \frac{1}{\dim A}\fig{proj1.eps}=\frac{1}{\dim A}\fig{proj2.eps}
1515: \end{equation}
1516: Then 
1517: \begin{equation}
1518: P_\pi=\begin{cases} \id_{X_\pi}, & \text{ if }X_\pi\in \Rep^0 A\\
1519:                     0,          &\text{ otherwise}
1520:       \end{cases}
1521: \end{equation}      
1522: \end{lemma}    
1523: \begin{proof}
1524:   If $X_\pi\in \Rep^0$, then the statement immediately follows from
1525:   \leref{lnoose}. Thus, let us assume that $X_\pi\notin \Rep^0 A$ and
1526:   prove that in this case, $P_\pi=0$.
1527: 
1528: 
1529: First, note that the composition  $\th^{-1}_\pi P_\pi$ can be rewritten
1530: as shown in \firef{fthP}. 
1531: 
1532: \begin{figure}[h]%% 30%
1533: \begin{equation*}
1534: \fig{thP.eps}
1535: \end{equation*}
1536: \caption{Presentation of $\th^{-1}_\pi P_\pi$}\label{fthP}
1537: \end{figure}
1538: 
1539: 
1540: From this presentation one easily sees that $\th^{-1}_\pi P_\pi$ is a
1541: morphism of $A$-modules; since $X_\pi$ is simple, this implies 
1542: \begin{equation}\label{thp=c}
1543: \th^{-1}_\pi P_\pi=c_\pi \id
1544: \end{equation}
1545: for some $c_\pi \in \Cset$. 
1546: 
1547: Next, let us calculate $P_\pi^2$:
1548: \begin{equation}
1549: \begin{aligned}
1550: P_\pi^2&=\frac{1}{(\dim A)^2}\fig{p2a.eps}
1551:        =\frac{1}{(\dim A)^2}\fig{p2b.eps}\\
1552:        &=\frac{1}{(\dim A)^2}\fig{p2c.eps}
1553:        =\frac{1}{\dim A}\fig{proj2.eps}\\
1554:        &=P_\pi.
1555: \end{aligned}
1556: \end{equation}
1557: Thus, $P_\pi$ is a projector. On the other hand, it follows from
1558: \eqref{thp=c} that $ P_\pi=c_\pi \th_{\pi}$. Combining these two
1559: results, we get
1560: $c_\pi^2\th_\pi=c_\pi$. If we assume that $c_\pi\ne 0$, then this
1561: implies that $\th_\pi=c^{-1}_\pi$; by \thref{tssrep0}, this is impossible if
1562: $X_\pi\notin \Rep^0 A$. Thus, $c_\pi=0$. 
1563: \end{proof}
1564: 
1565: \begin{lemma}\label{lsxy}
1566: For $X\in \Rep^0 A, Y\in \Rep A$, one has 
1567: \begin{equation}
1568: \langle\tilde s^A(X), Y\rangle
1569:  =\langle\tilde s^A(P(Y)), X\rangle
1570:  =\frac{1}{(\dim A)^2} \fig{stilde.eps}
1571: \end{equation}
1572: where $P\colon K(A)\to K^0(A)$ is as in \thref{tmodular}.
1573: \end{lemma}
1574: \begin{proof}
1575: Since both sides are linear in $Y$ it suffices to prove this
1576: formula when  $Y$ is simple. If $Y\in \Rep^0 A$, the
1577: statement immediately follows from \leref{lstilde}. Thus, we only need
1578: to prove that if $Y$ is simple, $Y\notin \Rep^0A$, then the
1579: right-hand side is zero. To prove this, let $C\in\Cset$ be defined by 
1580: $$
1581: C=\fig{sxy1.eps}
1582: $$
1583: 
1584: On one hand, it easily follows from \leref{lnoose} that 
1585: $$
1586: C=\dim A \fig{stilde.eps}
1587: $$
1588: On the other hand, we can deform the figure defining $C$ as shown
1589: below
1590: \begin{equation}
1591: \begin{aligned}
1592: %\fig{sxy1.eps}
1593: C&=\fig{sxy2.eps}=\fig{sxy3.eps}\\
1594:               &=\fig{sxy4.eps}=\fig{sxy5.eps}
1595: \end{aligned}
1596: \end{equation}
1597:           
1598: 
1599: By \leref{lproj}, this implies $C=0$.
1600: \end{proof}
1601: 
1602: Now we are ready to prove \thref{tmodular}.
1603: \begin{proof}[Proof of \thref{tmodular}]
1604:   Proof for $\tilde t$ is obvious from the definition. As for $\tilde
1605:   s$, it suffices to prove that $(\dim A)\langle G\tilde s^A (X), V\rangle
1606:   =\langle \tilde s (G (X)), V\rangle$ for any $X\in \Rep^0 A, V\in
1607:   \C$. Using adjointness of $G$ and $F$, this reduces to
1608: $$
1609: (\dim A)\langle \tilde s^A (X), F(V)\rangle=\tilde s_{G(X),V}.
1610: $$
1611:  (note that $F(V)\in \Rep A$, but in general, not in $\Rep^0 A$).
1612: Using \leref{lsxy} and definition of $F(V)$,  this can be rewritten as
1613: the following identity of figures:
1614: 
1615: \begin{equation}\label{keyeq} %30%
1616: \frac{1}{\dim A}\fig{sxfv}=\fig{sgxv}
1617: \end{equation}
1618: which can be proved by rewriting the graph in left hand side as shown in
1619: \firef{fsxfv} and using \leref{lproj}. 
1620: 
1621: \begin{figure}[ht] %30%
1622: $$
1623: \fig{sxfv}
1624:  =\fig{sxfv2}
1625:  =\fig{sxfv3}
1626: $$
1627: \caption{}\label{fsxfv}
1628: \end{figure}
1629: 
1630: 
1631: Similarly, the identity involving  $F^0$ is equivalent to 
1632: $$
1633: \langle(\dim A) \tilde s^A(F^0(V)), X\rangle=
1634: \langle\tilde s(V), G(X)\rangle
1635: $$
1636: which is also equivalent to \eqref{keyeq}.  This completes the
1637: proof of \thref{tmodular}.
1638: \end{proof} 
1639: 
1640: This theorem implies the following important result. 
1641: 
1642: \begin{theorem}\label{tmodular2}
1643:   If $\C$ is a modular category, $A$ is a rigid $\C$-algebra,
1644:   $\th_A=\id$, then $\Rep^0 A$ is modular and the numbers $D, \zeta$
1645:   appearing in \eqref{st} for $\Rep^0 A$ are related with the
1646:   corresponding numbers for $\C$ by
1647: \begin{equation}\label{dzeta}
1648: \begin{aligned}
1649: D(\Rep^0A)&=\frac{D(\C)}{\dim A}\\
1650: \zeta(\Rep^0 A)&=\zeta(\C).
1651: \end{aligned} 
1652: \end{equation}
1653: Also, the maps $G\colon K(\Rep^0 A)\to K(\C), F^0\colon K(\C)\to
1654: K(\Rep^0 A)$ commute with operators $s, t$.
1655: \end{theorem}
1656: \begin{proof}
1657: The proof is based on the following lemma. 
1658: \begin{lemma}
1659: Let $\A$ be a semisimple rigid braided balanced category over $\Cset$, with
1660: finitely many isomorphism classes of simple objects. Then $\A$ is
1661: modular iff the matrix $\tilde s$, defined by \firef{fsij}, satisfies 
1662: \begin{equation}\label{s^2}
1663: \tilde s^2\one =c\one
1664: \end{equation}
1665: for some $c\in \Cset, c\ne 0$. 
1666: \end{lemma}
1667: This lemma is not new; however, for the sake of completeness, we
1668: include its proof below. 
1669: 
1670: Thus, to prove that $\Rep^0A$ is modular, it suffices to prove
1671: $(\tilde s^A)^2 A=cA$ for some $c\ne 0$. But by \thref{tmodular},
1672: $\tilde s^A$ commutes with $F^0$ up to a constant; thus,
1673: \begin{align*}
1674: (\tilde s^A)^2 A&=(\tilde s^A)^2 F^0(\one)
1675:        =\frac{1}{(\dim A)^2}F^0(\tilde s^2\one)
1676:        =\frac{1}{(\dim A)^2}F^0(c\one)\\
1677:     &=\frac{c}{(\dim A)^2}A. 
1678: \end{align*}
1679: Thus, $\Rep^0A$ is modular; all other statements of the theorem
1680: immediately follow from \thref{tmodular}.
1681: 
1682: \begin{proof}[Proof of the lemma]
1683: If $\A$ is modular, the statement is well known and in fact $c=D^2$
1684: (see, e.g., \cite{BK}). Thus, let us assume that \eqref{s^2} holds and
1685: deduce from it non-degeneracy of $\tilde s$. 
1686: 
1687: First, note that $\tilde s\one =\dd=\sum d_i V_i$, where $V_i$ are simple
1688: objects in $\A$ and $d_i=\dim_\A V_i$. Thus, \eqref{s^2} implies
1689: $\langle \tilde s\dd, V_i\rangle =c\de_{i,0}$ which can be rewritten
1690: as 
1691: \begin{equation}\label{s^2a}
1692: \sum_{j}d_j \fig{s2.eps}=c\de_{i,0}\fig{idi.eps}
1693: \end{equation}
1694: 
1695: Let us now choose some $i,k\in I$ and let 
1696: $$
1697: C_{ijk}=\fig{s2ik.eps}
1698: $$
1699: On one hand, it is easy to show using the definition of $\tilde s$ that 
1700: $$
1701: C_{ijk}=\frac{\tilde s_{ij}\tilde s_{jk}}{d_j}
1702: $$
1703: (see, e.g., \cite[Lemma 3.1.4]{BK}) and thus, 
1704: $$
1705: \sum_j d_jC_{ijk}=\sum_j\tilde s_{ij}\tilde s_{jk}=(\tilde s ^2)_{ik}.
1706: $$
1707: On the other hand, decomposing $V^*_i\ttt V^*_k$ in a direct sum of
1708: irreducibles and using  \eqref{s^2a}, we get 
1709: $$
1710: \sum_j d_jC_{ijk}=c\langle V^*_i\ttt V^*_k,\one\rangle =c\de_{ik^*}
1711: $$
1712: which is a non-singular matrix. Therefore, $\tilde s^2$ is
1713: non-singular and thus $\A$ is modular. This completes the proof of
1714: the lemma and thus of \thref{tmodular2}.
1715: \end{proof}
1716: \renewcommand{\qed}{}
1717: \end{proof}
1718: \begin{remark}
1719: For modular tensor categories coming from conformal field theory, the
1720: identity $\zeta(\C)=\zeta(\Rep^0 A)$ can be interpreted as stating that
1721: an extended CFT has the same central charge as the original CFT,
1722: which, of course, should be expected. 
1723: \end{remark}
1724: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1725: 
1726: \section{Vertex operator algebras}\label{svoa}
1727: In this section, we give the example which was one of our main
1728: motivations for this work. Detailed proofs of the results given here
1729: will appear in a separate paper \cite{voaproof}; here we only outline
1730: the main ideas. We should also note that relation between extensions
1731: of vertex operator algebras and algebras in a category discussed here
1732: was independently found by A.~ Wassermann \cite{was}, who discussed
1733: his work with the second author (Ostrik) during his visit to MSRI in
1734: November 2000. At this time, we were finishing the first draft of this
1735: paper.
1736: 
1737: 
1738: 
1739: We assume that the reader is familiar with the notion of a vertex
1740: operator algebra (VOA); a review and list of references can be found,
1741: e.g., in \cite{Frenkel}, \cite{FHL}. To avoid ambiguity, we mention
1742: that we include the Virasoro element $\omega$ and $\Z$-grading in the
1743: definition of a VOA.  Similarly, when talking about modules over a
1744: VOA, we always assume that $L_0$ acts semisimply with
1745: finite-dimensional eigenspaces (this automatically gives
1746: $\Cset$-grading on a module). We only consider finite length modules.
1747: 
1748: 
1749: 
1750: 
1751: 
1752: 
1753:  Let $\V$ be a vertex operator algebra which
1754: is nice enough so that the following properties are satisfied:
1755: 
1756: \begin{enumerate}
1757:   
1758: \item For every simple $\V$-module $M$, its conformal dimension (i.e.
1759:   lowest eigenvalue of $L_0$) is real, $\ge 0$, with equality only for
1760:   $M=\V$, in which case $\dim \V_0=1$. 
1761: 
1762: \item The category of representations of $\V$ is semisimple, with only
1763:   finitely many simple objects, and all spaces of conformal blocks
1764:   (i.e., intertwining operators between tensor products of
1765:   representations) are finite-dimensional. Also, $\V$ is simple as a
1766:   $\V$-module. 
1767: 
1768: \item The category $\C$ of $\V$-modules is a rigid braided tensor
1769:   category. 
1770: \end{enumerate}
1771:   
1772: The first condition is technical; we will only need it to ensure
1773: uniqueness of vacuum vector (see proof of \thref{thvoa} below). The
1774: most important condition is the last one, which deserves detailed
1775: discussion. 
1776: 
1777: There are at least two ways to define the tensor product (usually
1778: called the fusion tensor product) structure on the category of
1779: $\V$-modules, both originating in the pioneering work of Moore and
1780: Seiberg.  The first construction, developed in a series of papers of
1781: Huang and Lepowsky \cite{tensor12}, \cite{tensor3}, \cite{tensor4}, is
1782: based on defining the tensor product via the space of intertwining
1783: operators. The second approach uses the vector spaces of coinvariants
1784: (see \cite{Frenkel}) which should give a modular functor, and then
1785: using this modular functor to define the structure of a tensor
1786: category (see \cite{BK}). This shows that for every VOA appearing in
1787: conformal field theory the category of modules has a structure of a
1788: rigid braided tensor category.  In fact, such a VOA has to satisfy a
1789: stronger restriction:
1790: \begin{enumerate}
1791: \item[($3'$)]
1792: The category of $\V$-modules is a modular tensor category. 
1793: \end{enumerate}
1794: Indeed, it follows from the axioms of a rational conformal field
1795: theory that the spaces of conformal blocks for such a VOA form a
1796: modular functor, and it is known that a modular functor allows one to
1797: define a structure of a modular tensor category (see, e.g.,
1798: \cite{BK}).
1799: 
1800: 
1801: These two approaches should give equivalent results; unfortunately, to
1802: the best of our knowledge, details of this equivalence are not
1803: available in the literature. In what follows, we will use the first
1804: approach, i.e. use the definition of the tensor structure given by
1805: Huang and Lepowsky.
1806: 
1807: In both approaches, it is relatively easy to give the definition of
1808: the tensor product, but it is extremely difficult to check that for a
1809: given VOA this tensor product is well-defined and defines a structure
1810: of a rigid balanced braided category (see \cite{tensor4} for the list
1811: of conditions that need to be checked).  So far, this has only been
1812: checked in very few examples.\footnote{Of course, as mentioned above,
1813:   this should hold for any VOA that comes from a rational conformal
1814:   field theory, but this does not help much: axioms of RCFT are even
1815:   more difficult to check.}
1816: 
1817: 
1818: Most important example of a VOA for which conditions (1)--(3) have
1819: been checked is the VOA coming from an  affine Lie algebra at positive
1820: integer level, discussed below.
1821: 
1822: \begin{example}\label{exwzw}
1823:   Let $\g$ be a simple Lie algebra, $\ghat$ --- corresponding affine
1824:   Lie algebra, and $k$ --- a non-negative integer (level). Let
1825:   $L_{0,k}$ be the integrable $\ghat$ module of level $k$ with highest
1826:   weight $0$ (the vacuum module). Then it is known that it has a
1827:   canonical structure of a VOA; we will denote this VOA by $\V(\g,
1828:   k)$. This VOA satisfies requirements (1)--($3'$) and thus, its
1829:   category of representations $\C(\g,k)$ is modular (see \cite{HL},
1830:   \cite{BK}). As an abelian category, $\C(\g,k)$ is just the category
1831:   of integrable $\ghat$ modules of level $k$. It is also known (see
1832:   \cite{Finkelberg}) that $\C(\g,k)$ is equivalent (as modular
1833:   category) to the ``semisimple part'' of the category of
1834:   representations of the quantum group $U_q\g$ with
1835:   $q=e^{\pi\i/m(k+h^\vee)}$, where $h^\vee$ is the dual Coxeter number
1836:   and $m=1$ for simply-laced algebras, $m=2$ for $B_n, C_n, F_4$ and
1837:   $m=3$ for $G_2$.
1838: \end{example}
1839: 
1840: 
1841: Let $\V\subset \V_e$ be a subalgebra (in the sense of VOA's). Assume
1842: in addition that $\V_e$ is finite length as a module over $\V$. Then
1843: we will call $\V_e$ an {\em extension} of $\V$.
1844: 
1845: \begin{theorem}\label{thvoa}
1846:   Let $\V$ be a VOA satisfying {\rm{(1)--(3)}} above, and let $\C$ be
1847:   the category of $\V$--modules. Then the following two notions are
1848:   equivalent:
1849: \begin{enumerate}
1850: \item
1851:  An extension $\V\subset \V_e$, where $\V_e$ is also a VOA satisfying
1852:  properties {\rm{(1)--(3)}} above 
1853: 
1854: \item A rigid $\C$-algebra $A$ with $\th_A=1$
1855: \end{enumerate}
1856: 
1857:   Under this correspondence, category of
1858:    $\V_e$-modules is identified with $\Rep^0A$.
1859: \end{theorem}
1860: \begin{proof}
1861:   We give a sketch of the proof; details will appear in the
1862:   forthcoming paper \cite{voaproof}.
1863:   
1864:   If $\V_e$ is a VOA, then for every $v\in \V_e$ we have the vertex
1865:   operator $Y(v,z)\colon \V_e\to \V_e[[z,z^{-1}]]$. Restricting it to
1866:   $v\in \V$, we get a structure of a $\V$-module on $\V_e$. It is
1867:   immediate from the definitions that the map $Y(\cdot, z)$ is an
1868:   intertwining operator of the type $\binom{\V_e}{\V_e\ \V_e}$ and
1869:   thus gives a morphism of $\V$-modules $Y\colon \V_e\stackrel{\bf
1870:     .}{\otimes}\V_e\to \V_e$, where $\stackrel{\bf .}{\otimes}$ is the
1871:   ``fusion'' tensor product. It follows from the usual commutativity
1872:   and associativity axioms for a VOA (see \cite{FHL}) that $Y$ defines
1873:   a structure of a commutative and associative algebra on $\V_e$.
1874:   Existence and uniqueness of unit follow from existence and
1875:   uniqueness of the vacuum vector in a VOA (see condition (1) above).
1876:   Condition $\th_A=1$ follows from the fact that eigenvalues of $L_0$
1877:   on $\V_e$ are integer. A straightforward check shows that the
1878:   arguments above can be reversed and that the category of
1879:   representations of $\V_e$ as a VOA coincides with $\Rep^0 A$.
1880: \end{proof}
1881: 
1882: 
1883: One of the general ways to construct extensions of the VOA $\V(\g,k)$
1884: is by using the notion of conformal embedding (note, however, that not
1885: all extensions can be obtained in this way). Let $\g\subset \g'$ be an
1886: embedding of Lie algebras; then it defines an embedding of affine Lie
1887: algebras $\ghat\subset \ghat'$. This embedding doesn't preserve the
1888: level --- a pullback of a $\ghat'$ module of level $k'$ will be a
1889: module of level $k=x_e k'$ for some integer $x_e$; we will
1890: symbolically write $(\ghat)_k\subset (\ghat')_{k'}$. It defines an
1891: embedding $\V(\g, k)\subset \V(\g', k')$ which preserves the operator
1892: product expansion (i.e., the algebra structure in $\V$) but in general
1893: not the Virasoro element. In some special cases, however, such an
1894: embedding preserves the Virasoro element as well and therefore defines
1895: an embedding of VOA's; they are called {\em conformal embeddings}.
1896: In this case it is easy to show (see, e.g., \cite[Chapter 17]{CFT})
1897: that $\V(\g', k')$ is automatically finite as $\ghat$-module, so
1898: $\V(\g', k')$ is an extension of $\V(\g,k)$. 
1899: 
1900: 
1901: 
1902: 
1903: \begin{example}\label{confembed}
1904:   Let $\C(\slt,k)$ be the category of integrable modules over
1905:   $\slthat$ of level $k$. Then it is known that for $k=10$, there is a
1906:   conformal embedding $(\slthat)_{10}\subset \widehat{sp(4)}_1$. The
1907:   easiest  way to describe this embedding is to note that the
1908:   irreducible 4-dimensional representation of $\slt$ has an invariant
1909:   non-degenerate skew-symmetric form, which gives an embedding
1910:   $\slt\subset sp(4)$.
1911:   
1912:   The decomposition of $\V(sp(4), 1)$ as $\V(\slt, 10)$ module is
1913:   given by $\V=L_{0,10}\oplus L_{6,10}$ (see \cite[Chapter 17]{CFT}).
1914:   Thus, this shows that the object $A=L_{0,10}\oplus L_{6,10}\in
1915:   \C(\slt, 10)$ has a structure of a rigid $\C$-algebra (later
1916:   we will show that such a structure is unique).
1917:   
1918:   Similarly, for $k=28$ there exists a conformal embedding
1919:   $(\slthat)_{28}\subset (\widehat{G}_2)_1$; the decomposition of
1920:   $\V(G_2,1)$ as $\V(\slt,28)$ module is given by $\V=L_{0,28}\oplus
1921:   L_{10, 28}\oplus L_{18, 28}\oplus L_{28,28}$.
1922: \end{example}
1923: 
1924: \begin{remark}
1925: The use of conformal embeddings to produce extensions of chiral
1926: algebras is, of course, well known in physics literature. Conformal
1927: embeddings can also be used in the subfactor theory --- see \cite{X}
1928: and references therein. 
1929: \end{remark}
1930: \begin{remark} 
1931:   It is very easy to prove rigorously that the conformal embedding
1932:   $\ghat_k\subset \ghat'_{k'}$ determines a structure of a rigid
1933:   $\C-$algebra on the vacuum module $V=L_{0,k'}$ over $\ghat'$
1934:   considered as a module over $\ghat$, even without referring to the
1935:   more general \thref{thvoa}. Let $\otimes_\g, \otimes_{\g'}$ denote
1936:   the fusion product over $\ghat, \ghat'$ respectively. this fusion
1937:   tensor product can be defined in terms of coinvariants for the
1938:   action of the algebra of rational $\g$ (respectively, $\g'$) valued
1939:   functions.  Namely, consider the rational curve ${\mathbb P}^1$ with
1940:   3 marked points and the representations $V, V, V^*$ assigned to
1941:   these points. The spaces of homomorphisms $V\otimes_{\g} V\to V,
1942:   V\otimes{\g'}$ are isomorphic to the dual of the spaces of
1943:   coinvariants for $\g, \g'$, see \cite{KL, Finkelberg}. Also, the space of
1944:   coinvariants for $\g'$ is canonically isomorphic to $\C$. By
1945:   definition, we have a surjection $Coinv(\g)\surjto Coinv(\g')$.
1946:   Thus, we have an embedding $\Hom (V\otimes_{\g'} V, V)\subset
1947:   \Hom(V\otimes_{\g} V, V)$. Thus, the canonical morphism
1948:   $V\otimes_{\g'} V\to  V$ defines a morphism $m:
1949:   V\otimes_h V\to V$. Let us prove that this morphism defines an
1950:   associative multiplication, that is $m(m\otimes_{\g} id)=m(id\otimes_{\g}
1951:   m)$. Both sides of this equality are represented by some
1952:   coinvariants (for ${\mathbb P}^1$ with 4 marked points and the
1953:   representations $V, V, V, V^*$ assigned to these points) and by the
1954:   construction of the fusion product these coinvariants actually come
1955:   from the coinvariants over $g$.  But the space of the coinvariants
1956:   over $\g'$ is one dimensional since $V\otimes_{g'} V\otimes_{\g'}V\simeq V$
1957:   and hence the LHS and the RHS are proportional. To compute the
1958:   proportionality coefficient it is enough to note that $m$ restricts
1959:   nontrivially on $V_0\subset V$ where $V_0$ is the vacuum module over
1960:   $\ghat$. The proof of commutativity is completely analogous.
1961:   Finally one can use the coinvariant above to identify $V$ and $V^*$
1962:   as $\ghat-$modules what implies that $m\colon V\otimes_h V\to V\to V_0$
1963:   can be used to identify $V$ and $V^*$ as $\ghat-$modules that is
1964:   $V$ is rigid $\C-$algebra over $\ghat$.
1965: \end{remark}
1966: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1967: \section{ADE classification for $\U$}
1968: In this section, we apply the general formalism developed above in a
1969: special case: when $\C$ is the semisimple part of the category of
1970: representations of $\U$ with $q=e^{\pi\i/l}, l\leq 2$ as defined by
1971: Andersen et al \cite{AP}. We assume that the reader is familiar with the
1972: definition and main properties of categories of representations of
1973: quantum groups at roots of unity; if not, we refer to the monograph
1974: \cite{BK} for a review.
1975: 
1976: It is known that the category $\C$ is semisimple, with simple objects
1977: $V_0, \dots, V_k$, where $k=l-2$ and $V_i$ is the usual
1978: $(i+1)$-dimensional irreducible representation of $\U$. Its
1979: Grothendieck ring $K$ is generated by one element, $V_1$. The quantum
1980: dimensions are given by
1981: $$
1982: \dim_\C V_n=[n+1]:=\frac{q^{n+1}-q^{-(n+1)}}{q-q^{-1}}
1983: $$
1984: which in particular implies that for any  non-zero object $V$,
1985: \begin{equation}\label{d>1}
1986: \dim_\C V\ge 1.
1987: \end{equation}
1988: It is also known that this category is modular and that the universal
1989: twist $\th$ is given by 
1990: \begin{equation}\label{thi}
1991: \th_n:=\th_{V_n}=q^{n(n+2)/2}=e^{2\pi\i\frac{n(n+2)}{4(k+2)}}.
1992: \end{equation}
1993: 
1994: Finally, we note that this category is equivalent to the category
1995: of integrable representations of affine Lie algebra $\widehat{\slt}$
1996: of level $k=l-2$, or, equivalently, the category of representations of
1997: the corresponding vertex operator algebra $\V(\slt, k)$ (see
1998: \cite{Finkelberg}). 
1999: 
2000: 
2001: 
2002: Our main goal is to classify all $\C$-algebras.
2003: 
2004: 
2005: % \begin{enumerate}
2006: % \item 
2007: %  $A$ is semisimple
2008: % \item $A$ is rigid
2009: % \item For all objects $X\in \Rep A$, 
2010: % \end{enumerate}
2011: 
2012: \begin{theorem} \label{tade}
2013:   There is a correspondence between rigid  $\C$-algebras with
2014:   $\th_A=\id$  and
2015:   Dynkin diagrams of types $A_n, D_{2n}, E_6, E_8$ with Coxeter number
2016:   equal to $l$. Under this correspondence, simple objects of $\Rep A$
2017:   correspond to vertices of the Dynkin diagram, and the matrix of
2018:   multiplication by $F(V_1)$ in $K(\Rep A)$ is $2-A$, where $A$ is the
2019:   Cartan matrix of the Dynkin diagram. 
2020: \end{theorem}
2021: 
2022: 
2023: \begin{proof}
2024:   Let $A$ be a rigid $\C$-algebra with $\th_A=\id$. In this case,
2025:   $\Rep A$ is a monoidal category and, by \coref{cmodule}, a module
2026:   category over $\C$. This implies that the Grothendieck ring $K(A)=K(\Rep
2027:   A)$ is a module over $K(\C)$. By \thref{trigidsemisimple}, $\Rep A$
2028:   is semisimple, so $K(A)$ has a distinguished basis (classes
2029:   $[X_\pi]$ of simple objects) so that in this basis, multiplication
2030:   by any $F(V), V\in \C$, has coefficients from $\Z_+$. In addition,
2031:   this module has the following properties:
2032: 
2033: \begin{enumerate}
2034: \item[(i)] The module $K(A)$ is indecomposable: it is impossible to
2035:   split the set of simple objects $\Pi$ as $\Pi=\Pi'\sqcup \Pi''$ so
2036:   that $K'=\oplus_{\Pi'}\Cset [X_\pi], K''=\oplus_{\Pi''}\Cset
2037:   [X_\pi]$ are $K(\C)$-submodules in $K(A)$. 
2038: 
2039:   Indeed, every simple module $X_\pi$ appears with non-zero
2040:   multiplicity in $F(V_i)\tta A=F(V_i)$ for some $V_i$. This follows
2041:   from $\langle F(V_i),X_\pi\rangle=\langle V_i,G(X_\pi)\rangle$. 
2042: 
2043: 
2044: \item[(ii)] There exists a map $d\colon K(A)\to \Cset$ such that
2045:   $d(X_\pi)\in\Rset_{>0}$ and $d(F(V)\tta X)=(\dim_\C V) d(X)$. 
2046:   
2047:   Indeed, it suffices to let $d(X)=\dim_{\Rep A}(X)$ and use
2048:   \thref{tdim0}. 
2049: 
2050: \item[(iii)] There exists a symmetric bilinear form $\langle, \rangle$
2051:   on $K(A)$ such that $\langle F(V)\tta X, Y\rangle=\langle X,
2052:   F(V)\tta Y \rangle$ for any $V\in \C, X,Y\in\Rep A$. 
2053:   
2054:   Indeed, we can let $\langle X, Y\rangle=\dim\Hom_A(X,Y)$ and use
2055:   rigidity and $F(V_i)^*\simeq F(V_i^*)\simeq F(V_i)$ (not
2056:   canonically).
2057: \end{enumerate}
2058: 
2059: 
2060: All modules $M$ over $K(\C)$ which have properties (i)--(iii) above
2061: were classified in \cite{EK}, where it is shown that they correspond
2062: to finite Dynkin diagrams with loops with Coxeter number equal to $l$.
2063: Under this correspondence, vertices of the Dynkin diagram correspond
2064: to the elements of distinguished basis of $M$, and the matrix of
2065: multiplication by $V_1\in \C$ is $2-A$, where $A$ is the Cartan matrix
2066: of the Dynkin diagram.
2067: 
2068: (Dynkin diagrams with loops, in addition to the usual Dynkin diagrams,
2069: include ``tadpole'' diagrams  $T_n$ shown in \firef{diagrL}; in
2070: \cite{EK}, this diagram is denoted by $L_n$. By
2071: definition, the Cartan matrix for such a diagram is the same as for
2072: $A_n$ but with $a_{11}=1$, and the Coxeter number for $T_n$ is equal
2073: to $2n+1$).
2074: 
2075: 
2076: 
2077: 
2078: 
2079: \begin{figure}[h]%% 30%
2080: \begin{equation*}
2081: \fig{diagrL.eps}
2082: \end{equation*}
2083: \caption{Dynkin diagram of type $T$.}\label{diagrL}
2084: \end{figure}
2085: 
2086: 
2087: 
2088: \begin{remark}  
2089:   In an interesting note \cite{mckay2}, it was shown that the dimension
2090:   vector $d$ can also be obtained from so-called ``semi-affine'' Dynkin
2091:   diagrams, which give $d$ both for finite and affine
2092:   Dynkin diagrams.
2093: \end{remark}
2094: 
2095: 
2096: 
2097: Now we have to check which of these modules can actually appear as
2098: Grothendieck ring $K(A)$ for some rigid $\C$-algebra $A$. 
2099: 
2100: First, note that if $K(A)$ is indeed the Grothendieck ring of a
2101: rigid $\C$-algebra $A$, then not only we have a
2102: distinguished basis $[X_\pi]$ and an inner product $\langle,\rangle$
2103: but in fact, the distinguished basis is orthonormal with respect to
2104: $\langle,\rangle$. This implies that the matrix of tensor product with
2105: $F(V_1)$ is symmetric in this basis. Thus, only simply-laced Dynkin
2106: diagrams can possibly come from $K(A)$. This leaves us with the ADET
2107: type diagrams.
2108: 
2109: 
2110: Next, we need to determine which vertex of the Dynkin diagram
2111: corresponds to the unit object, i.e. to $A$ itself. 
2112: 
2113: \begin{lemma}\label{lleg}
2114: If $A$ is a rigid $\C$-algebra, then $A$ corresponds to the
2115: end of the longest leg of the corresponding Dynkin diagram. 
2116: \end{lemma}
2117: \begin{remark}
2118: By an ``end'' we mean a vertex which is connected to exactly one
2119: vertex; in particular, the vertex with a loop in the diagram of type
2120: $T$ is not considered an end vertex.
2121: \end{remark}
2122: 
2123: \begin{proof}
2124:   Let $X\in \Rep A$ be the object corresponding to one of the
2125:   ends of legs of the Dynkin diagram. Then $F(V_1)\tta X$ is simple.
2126:   Since in a rigid category, tensor product of non-zero objects is
2127:   always non-zero, this implies that $F(V_1)\simeq F(V_1)\tta A$ is
2128:   simple. Thus, $A$ is connected to exactly one vertex, which
2129:   means that $A$ itself is an end of one of the legs. 
2130:  
2131:   To prove that $A$ is the end of the longest leg, note that if $X$ is
2132:   an end of the leg of length $m$ (that is, consisting of $m$ edges),
2133:   then $F(V_1)\tta X,\dots, F(V_m\tta X)$ are simple but
2134:   $F(V_{m+1})\tta X$  is not. This implies that $F(V_i)=F(V_i)\tta A,i=1,
2135:   \dots, m$ are simple, which means that the leg containing $A$
2136:   has length at least $m$. 
2137: \end{proof}
2138: 
2139: 
2140: 
2141: 
2142: % Next, let us make the following assumption. 
2143: % \begin{assumption}
2144: % $d(X)\ge 1$ for all non-zero $X\in \Rep A$. 
2145: % \end{assumption}
2146: 
2147: % This assumption automatically holds if $d(X\tta Y)=d(X)d(Y)$; indeed,
2148: % if $d(X)<1$ then $d(X^{\tta n})\to 0$ as $n\to \infty$, which is
2149: % impossible. 
2150: 
2151: % If so, then we can easily recognize which of the vertices of Dynkin
2152: % diagram corresponds to $A$. Indeed, the vector $d_\pi=d(X_\pi)$ is the
2153: % Frobenius--Perron vector for $2-A$, i.e., the eigenvector
2154: % corresponding to the maximal (in absolute value) eigenvalue; this
2155: % eigenvalue has multiplicity one, and thus $d_\pi$ is defined uniquely
2156: % up to a constant (see \cite{EK} for details). So if we require $\min
2157: % d_\pi=1$, this uniquely determines $d$. Therefore the vertex
2158: % corresponding to $A$ should be the one with minimal possible value of
2159: % $d$. Looking at the table of the eigenvectors $d$ given in \cite{}, we
2160: % see that in all cases, minimal $d$ corresponds to the end of the
2161: % longest leg of the Dynkin diagram; it is unique up to automorphism of
2162: % the diagram.
2163: 
2164: This determines the vertex corresponding to $A$ uniquely up to an
2165: automorphism of the Dynkin diagram. 
2166: 
2167: Once we know the vertex corresponding to $A$, we know the class of
2168: $F(V_1)$ in $K(A)$; since $F$ is a tensor functor and $V_1$ generates
2169: $K$, this uniquely determines the map $F$ at the level of Grothendieck
2170: rings, and thus, the adjoint map $G\colon K(A)\to K$. In other words,
2171: we can write for each vertex of the Dynkin diagram the structure of
2172: the corresponding object $X_\pi$ as an object of $\C$. In particular,
2173: this gives decomposition of $A$ itself as an object of $\C$.
2174: 
2175: Doing this explicitly for diagrams $A_n, D_n, E_n, T_n$ gives the
2176: answer shown in Table~\ref{table1} (no, it was not found using a
2177: computer --- it is done easily by hand), which agrees with the one
2178: given in Cappelli-Itzykson-Zuber classification.
2179: \begin{table}[h]
2180: \caption{Algebra $A$ for various Dynkin diagrams}\label{table1}
2181: \begin{tabular}{c|c|c}
2182: Diagram& $k=h-2$ & $A$\\
2183: \hline  
2184: $A_n$ & $n-1$ & $V_0$ \\
2185: $D_n$ & $2n-4$ & $V_0+V_k$ \\
2186: $T_n$ & $2n-1$ & $V_0+V_k$ \\
2187: $E_6$ & $10$ & $V_0+V_6$ \\
2188: $E_7$ & $16$ & $V_0+V_8+V_{16}$ \\
2189: $E_8$ & $28$ & $V_0+V_{10}+V_{18}+V_{28}$ \\
2190: \end{tabular}
2191: \end{table}
2192: 
2193: Next step is to find which of the possible $A$ given in this table do
2194: have a structure of a $\C$-algebra. 
2195: 
2196: {\bf Type $A$}: in this case, $A=\one$ obviously has a unique
2197: structure of commutative associative algebra, and $\Rep A=\C$. 
2198: 
2199: {\bf Type $D$}.  Let us introduce the notation 
2200: \begin{equation}\label{de}
2201: \de=V_k.
2202: \end{equation}
2203: It easily follows from explicit formulas that $\dim_\C \de=1$ and
2204: $\de\ttt V_n\simeq V_{k-n}$; in particular, $\de\ttt\de\simeq \one$. 
2205: 
2206: \begin{theorem}\label{tdtype}
2207: The object $A=\one\oplus\de$ in $\C$ has a structure of a rigid 
2208: $\C$-algebra iff $4|k$. In this case, the
2209: structure of an algebra is unique up to isomorphism, and this algebra
2210: satisfies $\th_A=\id$. 
2211: \end{theorem}
2212: \begin{proof}
2213:   Let $\mu$ be the multiplication map $\mu\colon
2214:   (\one\oplus\de)\ttt(\one\oplus\de)\to (\one\oplus\de)$. All
2215:   components of such a map are uniquely determined by the unit axiom,
2216:   except for $\mu_{\de\de}\colon \de\ttt\de\to \one$. Since
2217:   $\de\ttt\de\simeq \one$, such a map is unique up to a constant.
2218:   Rigidity implies that $\mu_{\de\de}\ne 0$. This proves
2219:   uniqueness. 
2220: 
2221:   To check existence, fix some non-zero $\mu_{\de\de}$. Then
2222:   associativity and commutativity are equivalent to 
2223: \begin{equation}
2224: \begin{aligned}
2225: \mu_{\de\de}\circ(\id \ttt\mu_{\de\de})
2226:  &=\mu_{\de\de}\circ(\id \ttt\mu_{\de\de})\colon \de\ttt\de\ttt\de\to\de\\
2227: \mu_{\de\de}\circ R_{\de\de}&=\mu_{\de\de}.
2228: \end{aligned}
2229: \end{equation}
2230: To check the second equation, we use the following lemma
2231: 
2232: 
2233: \begin{lemma}\label{lfR}
2234: For generic values of $q$, let $f\colon V_{a}\otimes V_{a}\to V_{2b}$
2235: be a nonzero homomorphism. Then 
2236: $$
2237: f\circ R_{V_a V_a}=(-1)^{a-b}\th_a^{-1} (\th_{2b})^{1/2}f
2238: $$
2239: where $\th_{a}=q^{a(a+2)/2}$ is the universal twist and
2240: $\th_{2b}^{1/2}=q^{2b(2b+2)/4}$. 
2241: \end{lemma}
2242: 
2243: To prove this lemma, note that it immediately follows from balancing
2244: axiom in $\C$ that $f\circ R^2=\th_a^{-2}\th_{2b}$, which gives the
2245: formula above up to a sign. To find the sign, it suffices to let
2246: $q=1$.
2247: 
2248: Since this formula works for generic values of $q$, it should also be
2249: valid for $q$ being a root of unity. In particular, applying this
2250: lemma to 
2251: $q=e^{\pi\i/(k+2)}$ and $\mu_{\de\de}\colon \de\ttt\de\to\de$, we get 
2252: $$
2253: \mu_{\de\de}\circ R_{\de\de}=(-1)^k \th^{-1}_{\de} \mu_{\de\de}.
2254: $$
2255: We have $\th_\de =q^{\frac{k(k+2)}{2}}=e^{\pi \i
2256:   k(k+2)/2(k+2)}=e^{2\pi \i k/4}=\i^k$. Thus,
2257: $(-1)^k\th^{-1}_\de=\i^k$ is equal to one iff $k$ is divisible by 4.
2258: Therefore, the map $\mu$ is commutative iff $k=4m$.
2259: 
2260: To check associativity, note that both sides are equal up to a
2261: constant (since $\dim \Hom(\de^{\ttt 3}, \de)=1$); to find the
2262: constant, take composition of  both sides with $(i_\de)\ttt \id\colon
2263: \de\to \de^{\ttt 3}$ and use $\dim_\C \de=1$.
2264: \end{proof}
2265: 
2266: The category of representations of this algebra is described in detail
2267: in \seref{sdtype}. It follows from the analysis there that the structure
2268: of $K(A)$ as $K(\C)$-module is described by the diagram $D_{2m+2}$.
2269: 
2270: 
2271: 
2272: {\bf Type $T$}.
2273: 
2274: The diagram $T_n$ can not appear as $K(A)$ for a
2275: commutative associative algebra $A$. Indeed, in this case $A$ must be
2276: isomorphic to $V_0\oplus V_k$, but it was proved in \thref{tdtype} that
2277: there is at most  one structure of a rigid $\C$-algebra on this
2278: object, and if it exists, $K(A)$ is described by $D_{n}$, not $T_n$. 
2279: 
2280: {\bf Type $E_7$}.
2281: 
2282: This diagram can not appear as $K(A)$ for a commutative associative
2283: algebra $A$. Indeed, in this case the table gives $A=V_0\oplus
2284: V_8\oplus V_{16}=(\one\oplus\de)\oplus V_8$. Obviously,
2285: $A'=\one\oplus\de$ is a subalgebra in $A$, and multiplication on $A$
2286: defines a structure of $A'$-module on $V_8$ and morphism of
2287: $A'$-modules $V_8\ttt V_8\to (\one\oplus\de)$. By rigidity, this morphism
2288: is non-zero, which also implies that the restriction of $\mu$ to
2289: $V_8\ttt V_8\to \de$ is non-zero. But it immediately follows from
2290: \leref{lfR} that such a morphism can not be symmetric.
2291: 
2292: {\bf Type $E_6$}. 
2293: 
2294: In this case, there is a unique up to isomorphism $\C$-algebra
2295: structure on $V_0\oplus V_6$. Existence follows from the discussion of
2296: the previous section and existence of a conformal embedding of affine
2297: Lie algebras $(\slthat)_{10}\injto \widehat {sp(4)}_{1}$ (see
2298: \exref{confembed}).  To prove uniqueness, note that the only
2299: non-trivial components of the multiplication map $\mu$ are $\mu'\colon
2300: V_6\ttt V_6\to \one$, $\mu''\colon V_6\ttt V_6\to V_6$. Both of them
2301: are unique up to a constant factor. We can fix some non-zero morphisms
2302: \begin{align*}
2303: e\colon &V_6\ttt V_6\to\one,\\ 
2304: f\colon &V_6\ttt V_6\to V_6.
2305: \end{align*}
2306: Then $\mu'=\al e, \mu''=\be f$ for some $\al, \be\in \Cset$. It
2307: follows from rigidity that $\al\ne 0$. Using isomorphism of
2308: $\C$-algebras $\ph\colon (\one\oplus V_6)\to (\one\oplus V_6)$ given
2309: by $\ph|_\one=\id, \ph|_{V_6}=\al^{1/2}\id$, we see that without loss
2310: of generality we can assume $\al=1$, so $\mu|_{V_6\ttt V_6}=e+\be f$.
2311: Condition that $\mu$ be associative gives the following quadratic
2312: equation on $\be$:
2313: $$
2314: \be^2\Ph_1=\Ph_2
2315: $$
2316: where $\Ph_1, \Ph_2$ are morphisms $V_{6}^{\ttt 3}\to V_6$ given by   
2317: \begin{align*}
2318: \Ph_1&= f\circ (\id \ttt f)-f\circ (f\ttt \id)\\
2319: \Ph_2&=e\ttt \id-\id \ttt e.
2320: \end{align*}
2321: 
2322: It is easy to see that $\Phi_2\ne 0$, so the equation
2323: $\be^2\Ph_1=\Ph_2$ is non-trivial. Thus, such an equation may either
2324: have no solutions at all or have exactly two solutions differing by
2325: sign: $\be=\pm \be_0$. These two solutions actually would give
2326: isomorphic algebras: the map $\ph\colon\one\oplus\de\to\one\oplus\de$ given
2327: by $\ph|_{\one}=1, \ph|_{\de}=-1$ gives the isomorphism.
2328: 
2329: {\bf Type $E_8$.}
2330: 
2331: In this case, there again exists a unique structure of a rigid
2332: $\C$-algebra on $A=V_0\oplus V_{10}\oplus V_{18}\oplus V_{28}$.
2333: Existence follows from existence of conformal embedding
2334: $(\slthat)_{28}\subset (\widehat{G}_2)_1$ (see \exref{confembed}). To
2335: prove uniqueness, let $A'\subset A$ be the subalgebra generated (as a
2336: $\C$-algebra) by $V_0\oplus V_{10}$. Let $\V_e$ be the vertex operator
2337: algebra corresponding to $A'$; by results of \seref{svoa}, it is an
2338: extension of the VOA $\V=\V(\slt, 28)$. From the definition, $\V_e$ is
2339: generated as a VOA by $\V$ and $L_{0,10}$. Since $L_{0,10}$ is an
2340: irreducible $\slthat$ module, it is generated (as $\slthat$ module) by
2341: its lowest degree component (degree stands for homogeneous degree,
2342: i.e. eigenvalue of $L_0$). This lowest degree is equal to
2343: $\Delta_{10}=\frac{10(10+2)/2}{2(28+2)}=1$.
2344: 
2345: Since it is well known that $\V(\g, k)$ is generated as a VOA by its
2346: degree one component $\V[1]\simeq \g$, we see that $\V_e$ is generated
2347: as a VOA by $\V[1]\oplus L_{10,28}[1]$. It is also easy to check that
2348: conformal dimensions (i.e., lowest eigenvalues of $L_0$) for
2349: $L_{18,28}$ and $L_{28,28}$ are greater than one, so
2350: $\V_e[1]=\V[1]\oplus L_{10, 28}[1]\simeq \slt\oplus L_{10}$, where
2351: $L_{10}=L_{10,28}[1]$ is an irreducible $\slt$-module with highest
2352: weight 10.
2353: 
2354: By \cite[Section 2.6]{Kac}, if $\V_e[0]=\Cset, \V_e[n]=0$ for $n<0$
2355: and $\V_e$ is generated as VOA by $\V_e[1]$, then $\V_e[1]=\g$ is a
2356: Lie algebra with an invariant bilinear form, and $\V_e$ is naturally a
2357: module over $\ghat$; moreover, $\V_e$ is a quotient of the Weyl module
2358: $V^{\g}_{0,k}$ over $\ghat$ for some $k$. Thus, we see that embedding
2359: $\V\subset \V_e$ defines an embedding $\slt\subset \g$. Rigidity of
2360: $A$ also implies that the multiplication map $V_{10}\ttt V_{10}\to
2361: V_0$ is non-zero, which implies that the restriction of the commutator
2362: in $\V_e[1]=\g$ to $L_{10}\ttt L_{10}\to \slt$ is non-zero. Now we can
2363: use the following lemma.
2364: 
2365: \begin{lemma}
2366: Let $\g$ be a finite-dimensional Lie algebra which contains a
2367: subalgebra isomorphic to $\slt$ and as a $\slt$-module,  
2368: $$
2369: \g\simeq \slt\oplus L_{10}. 
2370: $$
2371: If, in addition, restriction of the commutator map $[\, , \, ]\colon
2372: L_{10}\ttt L_{10}\to \slt$ is non-zero, then $\g\simeq G_2$.
2373: \end{lemma}
2374: \begin{proof}
2375: It is easy to see that in such a situation, $\g$ must be simple
2376: (indeed, the only possible ideals are $\slt$ and $L_{10}$, and none of
2377: them is an ideal). But the only 14-dimensional simple Lie algebra is
2378: $G_2$. 
2379: \end{proof}
2380: 
2381: 
2382: Therefore, embedding $\V\subset \V_e$ gives rise to an embedding
2383: $\slt\subset G_2$.  Since the Virasoro central charge is the same for
2384: $\V,\V_e$, this embedding extends to a conformal embedding
2385: $(\slthat)_{28}\subset (\ghat)_k$.  But it is well known (see, e.g,
2386: \cite[Chapter 17]{CFT}) that such a conformal embedding uinique,
2387: namely $(\slthat)_{28}\subset (\widehat{G}_2)_1$.
2388: 
2389: \end{proof}
2390: 
2391: 
2392: \begin{remark}
2393:   Note that the proof of \thref{tade} does not rely on
2394:   Itzykson-Cappelli-Zuber classification.
2395: \end{remark}
2396: 
2397: \begin{remark}
2398:   Explicit analysis shows that for all $\C$-algebras $A$ given by
2399:   \thref{tade}, for any $X,Y\in \Rep A$ one has $X\tta Y\simeq Y\tta
2400:   X$ (not canonically) even though there is no natural way to define
2401:   braiding on $\Rep A$; thus, the Grothendieck ring $K(A)$ is
2402:   commutative. Moreover, this ring coincides with the so-called
2403:   ``graph algebra'' of the Dynkin diagram (see \cite{CFT} for
2404:   discussion of graph algebras). In fact, many of the matrices and
2405:   constants which naturally appear in this theory (such as matrix of
2406:   $F\colon K\to K(A)$) can be calculated using only the Dynkin
2407:   diagram. This was first suggested by Ocneanu \cite{ocneanu} in
2408:   relation with the theory of subfactors; see,
2409:   e.g., \cite{C} for explicit calculations in $E_6$ case.  This
2410:   relation will be discussed in detail elsewhere.
2411: \end{remark} 
2412: 
2413: 
2414: 
2415: For future references, we give here some information about $K(A)$ for
2416: Dynkin diagrams of types $D_{even}, E_6$ and $E_8$. This information
2417: can be easily obtained by direct calculation outlined in the proof of
2418: \thref{tade}; checking which of simple $A$-modules lie in $\Rep^0 A$
2419: is trivial: explicit calculation shows that for each of these
2420: algebras, $\th_A=\id$ and thus we can use  \thref{tssrep0}.
2421: 
2422: 
2423: \begin{description}
2424: \item[$D_{2m+2}$] This algebra appears when the level $k=4m$; the
2425:   Coxeter number for $D_{2m+2}$ is $l=k+2=4m+2$. The diagram below shows,
2426:   for each of the simple $A$-modules, its structure as an object of
2427:   $\C$. For brevity, we write $i$ instead of $V_i$; thus, $0+(4m)$
2428:   stands for $V_0\oplus V_{4m}$, etc. Filled circles correspond to
2429:   simple objects which lie in $\Rep^0A$; empty circles are simple
2430:   objects in $\Rep A$ which are not in $\Rep^0 A$.  
2431: 
2432: \fig{D2n.eps}
2433: 
2434: \item[$E_6$] This algebra appears for $k=10$; the Coxeter number for
2435:   $E_6$ is $l=k+2=12$. All notations are same as before. 
2436: 
2437: \fig{E6.eps}
2438: 
2439: \item[$E_8$] This algebra appears for $k=28$; the Coxeter number for
2440:   $E_8$ is $l=k+2=30$. 
2441: \medskip
2442: 
2443: \fig{E8.eps}
2444: \end{description}
2445: 
2446: 
2447: % \begin{theorem}\label{tademod}
2448: % For each of rigid  $\C$-algebras $A$ described in \thref{tade},
2449: % the category $\Rep^0A$ is modular.
2450: % \end{theorem}
2451: % \begin{proof}
2452: % For type $A$ this is obvious, since $\Rep^0 A =\C$; for $E_6$ and $E_8$,
2453: % this follows from injectivity of the map 
2454: % $G\colon \Rep^0A\to \C$ (which is easily checked by direct
2455: % calculations) and \coref{cinjmod}.
2456: 
2457: % For type $D_{2m+2}$, the map $G$ is
2458: % not injective: it has one-dimensional kernel, spanned by
2459: % $\De=X^+_{2m}-X^-_{2m}$ (these are the objects corresponding to the
2460: % ends of the short legs; see for \seref{sdtype} for detailed
2461: % description). It is easy to deduce from \coref{cinjmod} that to prove
2462: % non-degeneracy of $\tilde s^A$, it suffices to show that $\tilde
2463: % s^A_{\De,\De}\ne 0$. 
2464: 
2465: 
2466: % Using the formula $\tilde s_{XY}=(\th_X\th_Y)^{-1}\sum_i N_i\th_i\dim
2467: % X_i$, where $X^*\tta Y=\sum N_i X_i$ (see, e.g., \cite[Chapter 3]{BK})
2468: % and explicit formulas for decomposition of $X_{2m}^\pm\tta
2469: % X_{2m}^\pm$ given in \seref{sdtype} we get 
2470: 
2471: % \begin{equation}
2472: % s^A_{\De,\De}=2\tilde s_{X_{2m}^+,\De}
2473: %     =2\th_{2m}^{-2} \sum_{j=0}^m(-1)^j\dim_A (X_{2j})\th_{X_{2j}} 
2474: % \end{equation}
2475: 
2476: % Since in this case $\dim_\C A=1+\dim \de=2$, by \thref{tdim0} we have
2477: % $\dim_A X=(\dim_\C X)/2$, which gives
2478:  
2479: % \begin{equation}
2480: % s^A_{\De,\De}=\th_{2m}^{-2}\sum_{j=0}^{2m}(-1)^j [2j+1]\th_{2j}.
2481: % \end{equation}
2482: 
2483: % Using explicit formula for $\th_{2j}$, this sum can be rewritten as
2484: 
2485: % \begin{equation}\label{thesum}
2486: % s^A_{\De,\De}=\th_{2m}^{-2}\frac{q^{-1}}{q-q^{-1}}
2487: %                \bigl(-a_0+2 a_1-2a_2+\dots-2 a_{2m}+a_{2m+1}
2488: %                \bigr)
2489: % \end{equation}
2490: % where $a_j=q^{2j^2}$.
2491: 
2492: % Since $q=e^{\pi\i/l}, l=4m+2$, explicit calculation shows
2493: % $\th_{2m}^{-2}q^{-1} =q^{-(2m+1)^2}=(-1)^m (-\i)$ and $q-q^{-1}=2\i
2494: % \sin (\pi/l)$, so 
2495: % $$
2496: % \th_{2m}^{-2}\frac{q^{-1}}{q-q^{-1}}=\frac{(-1)^{m+1}}{2\sin (\pi/l)}.
2497: % $$
2498: 
2499: % On the other hand, $a_j$ have the following symmetries:
2500: % $$
2501: % a_{j\pm l}=a_{l-j}=a_j,\quad a_{j+(l/2)}=-a_j
2502: % $$
2503: % which allows us to rewrite \eqref{thesum} as
2504: % \begin{equation}\label{thesum2}
2505: % s^A_{\De,\De}=\frac{(-1)^{m}}{2\sin (\pi/l)}
2506: %                 \sum_{j=1}^{l}e^{8\pi\i   j^2/l}
2507: %              =\frac{(-1)^{m}}{2\sin (\pi/l)} S(8,l)
2508: % \end{equation}
2509: % where we used the notation 
2510: % \begin{equation}\label{sal}
2511: % S(a,l)=\sum_{j=1}^le^{a\pi \i j^2/l}.
2512: % \end{equation}
2513: 
2514: % These sums are well-known in number theory; they generalize the
2515: % classical Gauss sums and have a number of remarkable properties. In
2516: % particular, they satisfy the following ``reciprocity law'' (see, e.g.,
2517: % \cite[Section  9.10]{apostol}): if $al$ is even, 
2518: % $$
2519: % S(a,l)=\sqrt{\frac{l}{a}}\left(\frac{1+\i}{\sqrt{2}}\right)\overline{S(l,a)}.
2520: % $$
2521: % One easily computes $S(l,8)=2\sqrt{2}(1+\i)\i^m$ which gives 
2522: % $$
2523: % S(8,l)=\sqrt{2l}(-\i)^m
2524: % $$
2525: % and 
2526: % \begin{equation}\label{thesum3}
2527: % \tilde s^A_{\De,\De}=\frac{\i^{m}}{2\sin (\pi/l)}\sqrt{2l}\ne 0
2528: % \end{equation}
2529: % which completes the proof of non-degeneracy of $\tilde s^A$ and thus, of
2530: % modularity of $\Rep^0 A$.
2531: % \end{proof}
2532: 
2533: Note that by \thref{tmodular2}, each of $\C$-algebras $A$ listed in
2534: \thref{tade} gives rise to a modular category $\Rep^0 A$ and thus, a
2535: modular invariant in the sense of conformal field theory. It is easily
2536: checked that these modular invariants coincide with those given by
2537: Cappelli-Itzykson-Zuber classification. Note, however, that our proofs
2538: are completely independent of Cappelli-Itzykson-Zuber classification.
2539: 
2540: 
2541: \begin{remark}
2542:   After publication of the first version of this paper, it was
2543:   pointed out  to us that the data given by the figures above had
2544:   previously appeared in the literature in other guises. Most
2545:   importantly, the map $F\colon K\to K(A)$ is a morphism of $K$-modules; in
2546:   particular, this implies that it is an ``intertwiner'' in the sense
2547:   of \cite{FZ}. The explicit formulas for $F$ given above coincide
2548:   with those in Table~1 of \cite{FZ}. However, in the construction in
2549:   \cite{FZ} this map is just one of many possible intertwiners; also,
2550:   they only consider this map at the level of fusion algebras. In our
2551:   approach, $F\colon K\to K(A)$ comes from a functor $F\colon
2552:   \C\to\Rep A$ which is completely determined by the algebra $A$.
2553: \end{remark}
2554: 
2555: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 
2556: \section{Algebra of type $D_{2n}$}\label{sdtype}
2557: 
2558: In this section we describe in detail the category of representations
2559: of the algebra $A=\one\oplus\de$ in $\C$, constructed in the previous
2560: section for $k=4m$. 
2561: 
2562: 
2563: \begin{theorem}\label{tdtype2}
2564: 
2565: {\ }
2566: \begin{enumerate}
2567: \item 
2568: Simple modules over $A$ are $X_i=V_i \oplus V_{k-i}= A\ttt (V_i),
2569: i=1, \dots, 2m-1$
2570: and two simple modules $X_{2m}^+, X_{2m}^-$, both isomorphic as
2571: objects of $\C$ to $V_{2m}$, with $\mu_{X^+}=\mu_{X^-}\circ p$, where
2572: $p\colon A\to A, p|_{\one}=1, p|_{\de}=-1$. 
2573: 
2574: \item Tensor product with $F(V_1)=X_1$ is given by 
2575: \begin{equation*}
2576: \begin{aligned}X_1&\tta X_0=X_1,\\
2577:       X_1&\tta X_i\simeq X_{i-1}\oplus X_{i+1},\quad i=1, \dots, 2m-2\\ 
2578:       X_1&\tta X_{2m-1}=X_{2m-2}\oplus X_{2m}^+\oplus X_{2m}^-,\qquad 
2579:       X_1\tta X_{2m}^{\pm}=X_{2m-1}.
2580:     \end{aligned}
2581:   \end{equation*}  
2582: \end{enumerate}
2583: \end{theorem}
2584: Proof is fairly straightforward if we notice that an $A$-module is the
2585: same as an object $V\in \C$ with an isomorphism $\mu\colon \de\ttt
2586: V\isoto V$ such $\mu^2\colon \de\ttt\de\ttt V\to V$ coincides with
2587: $\mu_{\de\de}\ttt \id_V$. 
2588: 
2589: 
2590:  We also note that formula $F(V)\tta
2591: F(W)\simeq F(V\ttt W)$ defines multiplication in the subring in
2592: $K(A)$ generated by $X_1, \dots, X_{2m-1}$, $(X^+_{2m}+X^-_{2m})$. However,
2593: it does not allow one to determine tensor products involving
2594: $X_{2m}^\pm$. To do so, we need to use the definition. 
2595: 
2596: 
2597: 
2598: 
2599: 
2600: 
2601: 
2602: \begin{theorem}
2603: For $8\mid k$, 
2604: 
2605: \begin{align*}
2606: X_{2m}^\pm\tta X_{2m}^\pm &\simeq 
2607: X_0\oplus X_4\oplus\dots \oplus X_{2m-4}\oplus X_{2m}^\pm,\\
2608: X_{2m}^\pm\tta X_{2m}^\mp &\simeq 
2609: X_2\oplus X_6\oplus\dots \oplus X_{2m-2}.
2610: \end{align*}
2611: For $k\equiv 4 \mod 8$, 
2612: 
2613: \begin{align*}
2614: X_{2m}^\pm\tta X_{2m}^\pm &\simeq 
2615: X_2\oplus X_6\oplus\dots \oplus X_{2m-4}\oplus X_{2m}^\mp,\\
2616: X_{2m}^\pm\tta X_{2m}^\mp &\simeq 
2617: X_0\oplus X_4\oplus\dots \oplus X_{2m-2}.
2618: \end{align*}
2619: 
2620: In particular, $(X^\pm)^*\simeq X^\pm$ for $8\mid k$, and
2621: $(X^\pm)^*\simeq X^\mp$ for $k\equiv 4 \mod 8$, 
2622: 
2623: 
2624: \end{theorem}      
2625: 
2626: 
2627: \begin{proof}
2628: By definition, $X\tta Y= (X\ttt Y)/\im (\mu_1-\mu_2)$. As an object of
2629: $\C$, 
2630: $$
2631: X_{2m}^\pm \ttt X_{2m}^\pm=V_{2m}\ttt V_{2m}=V_0\oplus V_2\oplus\dots
2632: \oplus V_k
2633: $$
2634: we need to check which of the modules $V_i$ are in the image of
2635: $\mu_1-\mu_2$. To do so, we use the following lemma.
2636: \begin{lemma}
2637:   Let $n$ be even, $n\le k$ and let $\mu_1, \mu_2\colon \de\ttt
2638:   V_{k-n}\to V_n$ be defined by the compositions
2639: \begin{align*}
2640: \mu_1\colon& \de\ttt V_{k-n}
2641:               \xxto{\id\ttt f} \de\ttt V_{2m}\ttt V_{2m} 
2642:               \xxto{\mu\ttt\id}V_{2m}\ttt V_{2m}
2643:               \xxto{g}V_n\\
2644: \mu_2\colon& \de\ttt V_{k-n}
2645:               \xxto{\id\ttt f} \de\ttt V_{2m}\ttt V_{2m} 
2646:               \xxto{R\ttt \id}V_{2m}\ttt \de\ttt V_{2m}
2647:               \xxto{\id\ttt\mu}V_{2m}\ttt V_{2m}
2648:               \xxto{g}V_n\\
2649: \end{align*}
2650: where $f\colon V_{k-n}\to V_{2m}\ttt V_{2m}$, $g\colon V_{2m}\ttt
2651: V_{2m}\to V_{n}$ and $\mu\colon \de\ttt V_{2m}\to V_{2m}$ are
2652: arbitrary non-zero morphisms. Then $\mu_1=(-1)^{(k-2n)/4}\mu_2$.
2653: \end{lemma}
2654: To prove the lemma, it suffices to consider the identity shown in
2655: \firef{lastfig} and then apply \leref{lfR} to both sides. This proves
2656: the lemma.
2657:  
2658: 
2659: \begin{figure}[h]%% 30%
2660: \begin{equation*}
2661: \fig{mu1mu2.eps}
2662: \end{equation*}
2663: \caption{}\label{lastfig}
2664: \end{figure}
2665: 
2666: 
2667: 
2668: This lemma  implies that for $X^\pm\ttt X^\pm$, $\im
2669: (\mu_1-\mu_2)$ consists of those $V_i$ with $i$ even and $k-2i\equiv 4\mod
2670: 8$, while for $X^\pm\ttt X^\mp$, $\im
2671: (\mu_1-\mu_2)$ consists of those $V_i$ with $i$ even and $k-2i\equiv 0\mod
2672: 8$. 
2673: 
2674: This determines the decomposition of $X^\pm\tta X^\pm, X^\pm\tta
2675: X^\mp$ as on object of $\C$. By \thref{tdtype2}, this determines this
2676: tensor product as a representation of $A$ uniquely except for
2677: ambiguity in the choice of the action of $A$ on $V_{2m}$; in other
2678: words, we do not know if $X_{2m}^+$ or $X_{2m}^-$ appears in
2679: decomposition of $X_{2m}^\pm\tta X_{2m}^\pm$. To answer this, note
2680: that we already know enough to deduce that for $8\mid k$,
2681: $(X_{2m}^\pm)^*\simeq X_{2m}^\pm$. Thus, using rigidity we find
2682: $$
2683: \langle X_{2m}^\pm\tta X_{2m}^\pm, X_{2m}^\mp\rangle
2684: =\langle X_{2m}^\pm, X_{2m}^\pm\tta X_{2m}^\mp\rangle=0
2685: $$
2686: since we already know decomposition of $X_{2m}^\pm\tta X_{2m}^\mp$. Similar
2687: arguments show that for $k\equiv 4 \mod 8$, $\langle X^\pm\tta X^\pm,
2688: X^\pm\rangle=0$. This completes the proof of the theorem. 
2689: \end{proof}
2690: 
2691: 
2692: % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2693: % \section{Technical Lemma}\label{stechnical}
2694: % The goal of this section is to prove the technical result used in the
2695: % proof of uniqueness of subgroup of type $E_8$. Namely, we let $k=28$,
2696: % $A'=V_0+V_{28}$ with the structure of $\C$-algebra defined in
2697: % \seref{sdtype}, and $X=V_{10}+V_{18}$ with the structure of
2698: % $A'$-module as defined in \seref{sdtype}. 
2699: 
2700: % \begin{lemma}\label{ltechnical0}
2701: % One has a canonical isomorphism of vector spaces 
2702: % $$
2703: % \Hom_{A'}(X\ttt_{A'}X\ttt\dots\ttt X, X)=\Hom_{\C}(V_{10}\ttt\dots\ttt
2704: % V_{10}, V_{10}\oplus V_{18}).
2705: % $$
2706: % \end{lemma}
2707: % Indeed, it suffices to note that $X=F(V_{10})$ and then use
2708: % \thref{tfunctors}. 
2709: 
2710: % In particular, this impies that the space $\Hom_{A'}(X\ttt_{A'}X, X)$
2711: % is 2-dimensional. Let us consider the following linear equation for
2712: % $\ph \in \Hom_{A'}(X\ttt_{A'}X, X)$:
2713: 
2714: % \begin{equation}\label{etechnical2}
2715: % e\circ(\id\ttt \ph)=e\circ(\ph\ttt\id)`\colon X\ttt X\ttt X\to A'
2716: % \end{equation}
2717: % where  $e$ is a fixed  a non-zero
2718: % morphism $X\ttt_{A'}X\to A'$ (since $X$ is irreducible module over
2719: % $A'$, this map is unique up to a constant).
2720: 
2721: % \begin{lemma}
2722: % Let $L=\{\ph\in \Hom_{A'}(X\ttt_{A'}X, X)\st \ph \text{ satisfies
2723: %   \eqref{etechnical2}}\}$. Then $\dim L\le 1$. 
2724: % \end{lemma}
2725: % \begin{proof}
2726: % Since the space $\Hom_{A'}(X\ttt_{A'}X, X)$ is 2-dimensional, it
2727: % suffices to show that $L\ne \Hom_{A'}(X\ttt_{A'}X, X)$, i.e., that
2728: % there are $\ph$ which do not satisfy \eqref{etechnical2}. Let
2729: % $\ph_0\colon V_{10}\ttt V_{10}\to V_{18}$ be a non-zero morphism (it
2730: % is unique up to a constant), and let $\ph$ be the corresponsding
2731: % morphism $X\ttt_{A'} X\to X$ as in \leref{ltechnical0}. Then our goal
2732: % is to show that 
2733: % $$
2734: % e\circ(\id\ttt \ph)\ne e\circ(\ph\ttt\id)
2735: % $$
2736: % Similar to \leref{ltechnical0}, we can identify 
2737: % \begin{equation}\label{notequal}
2738: % \Hom_{A'}(X\ttt X\ttt X, A)=\Hom_\C(V_{10}\ttt V_{10}\ttt V_{10},
2739: % V_0\oplus V_{28}) 
2740: % \end{equation}
2741: % Under this idenification, \eqref{} becomes 
2742: % \begin{equation}\label{notequal2}
2743: % e'\circ(\id\ttt \ph)\ne e''\circ(\ph\ttt\id)\colon `
2744: %  V_{10}\ttt V_{10}\ttt V_{10}\to V_{28}
2745: % \end{equation}
2746: % where $e'\colon V_{18}\ttt V_{10}\to V_{28}, e''\colon V_{10}\ttt
2747: % V_{18}\to V_{28}$ are restrictions of the morphism $e\colon X\ttt X\to
2748: % A'$ (note that there are no non-zero morphisms $V_{18}\ttt V_{10}\to
2749: % V_{0}$). Since $e$ satisfies $e\circ R=e$ (???), we have  $e''=e'\circ R$.  
2750: 
2751: % To prove \eqref{notequal2}, let us compose both sides with the
2752: % morphism $\Psi\colon V_{28}\to V_{10}\ttt V_{10}\ttt V_{10}$ defined
2753: % as the compostion
2754: % \begin{equation*}
2755: % V_{28}\to V_{10}\ttt V_{18}\xxto{\id\ttt \psi}V_{10}\ttt V_{10}\ttt
2756: % V_{10}
2757: % \end{equation*}
2758: % where $\psi\colon V_{18}\to V_{10}\ttt V_{10}$ is defined by 
2759: % $$
2760: % v_{18}\mapsto (fv_{10})\ttt v_{10}-q^{10}v_{10}\ttt (fv_{10})
2761: % $$
2762: % ($v_n$ is the highest weight vector in $V_n$, and $f$ is the generator of
2763: % $\U$. We use the same form of relations in $\U$ as in \cite{BK}.)  
2764: 
2765: % Explicit calculation which we omit here shows that 
2766: % $e'\circ(\id\ttt \ph)\circ \Psi =q^{10}\id$ and $e'\circ R\circ
2767: % (\ph\ttt\id)\circ \Psi= (1+q^{20})$. BUT THEY ARE EQUAL!!!!
2768: 
2769: 
2770: 
2771: 
2772: % \end{proof}
2773: 
2774: 
2775: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2776: % \bibliographystyle{amsalpha}
2777: \begin{thebibliography}{BEK2}
2778: 
2779: \bibitem[AP]{AP}
2780: Andersen, H.H. and Paradowski, J.,
2781: {\em{Fusion categories arising from semisimple Lie algebras}},
2782: Comm. Math. Phys.
2783: {\bf 169} (1995), 563--588.
2784: 
2785: \bibitem[B]{B}
2786: Brugui\`eres, A.
2787: {\em Cat\'egories pr\'emodulaires, modularisations et invariants des
2788:   vari\'et\'es de dimension 3},
2789: Math. Ann. {\bf 316} (2000), 215--236. 
2790: 
2791: \bibitem[BEK]{bek}
2792: B\"ockenhauer, J., Evans, D., Kawahigashi, Y., 
2793: {\em On $\alpha$-induction, chiral generators and modular invariants
2794:   for subfactors},  
2795: Comm. Math. Phys. {\bf 208} (1999), 429--487.
2796: 
2797: \bibitem[BEK2]{bek2}
2798: \bysame,
2799: {\em Chiral structure of modular invariants for  subfactors}, 
2800: Comm. Math. Phys. {\bf 210} (2000), 733--784.
2801: 
2802: 
2803: % \bibitem[BE2]{be2}
2804: % B\"ockenhauer, J., Evans, D.,
2805: % {\em Modular invariants and subfactors},
2806: % arXiv:math.OA/0008056
2807: 
2808: 
2809: 
2810: \bibitem[BK]{BK}
2811: Bakalov, B., Kirillov, A., Jr. ,
2812: {\em{Lectures on tensor categories and modular functors}},
2813: Amer. Math. Soc., 2000. 
2814: 
2815: % \bibitem[BF]{BF}
2816: % Ben-Zvi, D. and Frenkel, E.
2817: % {\em Chiral algebras and curves}, 
2818: % to be published by AMS in the Summer of 2001.
2819: 
2820: 
2821: \bibitem[BN1]{BN}
2822: Bion--Nadal, J. 
2823: {\em Subfactor of the hyperfinite $\mathrm{II}_1$ factor with Coxeter
2824:   graph $E_6$ as invariant}, 
2825:  J.~Operator Theory {\bf 28} (1992), pp. 27--50.
2826: 
2827: \bibitem[BN2]{BN2}
2828: \bysame, 
2829: {\em An example of a subfactor of the hyperfinite $\mathrm{II}_1$
2830:   factor whose pricipal grpah invariant is the Coxter graph $E_6$}, 
2831:  in {\em Curent Topics in Operator Algebras (Nara, 1990)},
2832:  pp. 104--113, World Sci. Publishing, River Edge, NJ, 1991. 
2833: 
2834: \bibitem[BW]{BW}
2835: J.~W.~Barrett and B.~W.~Westbury, 
2836: {\em Spherical categories},
2837: Adv. Math. {\bf 143} (1999), 357--375.
2838:  
2839: \bibitem[C]{C}
2840: Coquereaux, R.,
2841: {\em Notes on quantum tetrahedron},
2842: {\tt arXiv:math-ph/0011006}.
2843: 
2844: 
2845: \bibitem[CIZ]{CIZ}
2846: Cappelli, A., Itzykson, C., and Zuber, J.-B.,
2847: {\em The A-D-E classification of minimal and $A_1^{(1)}$ conformal
2848:   invariant theories}, 
2849: Commun. Math. Phys., {\bf 13}(1987), 1--26.
2850: 
2851: \bibitem[EK]{EK}
2852: Etingof, P.,  Khovanov, M.,
2853: {\em{Representations of tensor categories and Dynkin diagrams}}, 
2854: Internat. Math. Res. Notices 1995, no. {\bf 5}, 235--247.  
2855: 
2856: 
2857: \bibitem[Fr]{Frenkel}
2858: Frenkel, E.,
2859: {\em{Vertex Algebras and Algebraic Curves}},
2860: Seminaire Bourbaki, Exp. 875; available as electronic preprint
2861:     {\tt arXiv:math.QA/0007054}.  
2862: 
2863: \bibitem[Fi]{Finkelberg}
2864: Finkelberg, M.,
2865: {\em{An equivalence of fusion categories}},
2866: Geom. Funct. Anal. {\bf 6} (1996), 249--267.
2867: 
2868: \bibitem[FHL]{FHL}
2869: Frenkel, I., Huang, Y.-Z., and Lepowsky, J.
2870: {\em On axiomatic approach to vertex operator algebras and modules},
2871: Memoirs of AMS, Vol. 104, No. 494, Amer. Math. Soc., Providence, RI,
2872: 1993.  
2873: 
2874: \bibitem[FMS]{CFT} 
2875: Di Francesco, P., Mathieu, P., and  S\'en\'echal, D.,
2876: {\em{Conformal field theory}},
2877: Graduate Texts in Contemporary Physics, 
2878: Springer-Verlag, New York, 1997.
2879: 
2880: \bibitem[FS]{FS}
2881: Fuchs, J. and Schweigert, C.,
2882: {\em Lie algebra automorphisms in conformal field theory}, 
2883: {\tt arXiv:math.QA/0011160}.
2884: 
2885: \bibitem[FZ]{FZ}
2886: Di Francesco, P. and Zuber, J.-B.
2887: {\em SU(N) lattice integrable models associated with graphs},
2888: Nucl. Phys. {\bf B338}(1990), 602--646.
2889: 
2890: \bibitem[H]{tensor4}
2891: Huang, Y.-Z.
2892: {\em A theory of tensor products for module categories for a vertex
2893:   operator algebra}, IV, J. Pure Appl. Alg. {\bf 100} (1995), 173--216.
2894: 
2895: 
2896: \bibitem[HKL]{voaproof}
2897: Huang, Y.-Z., Kirillov, A., Lepowsky, J.,
2898: {\em Braided tensor categories and  extesnions of vertex operator
2899:   algebras}, in preparation.
2900: 
2901: 
2902: 
2903: 
2904: 
2905: \bibitem[HL1]{tensor12}
2906: Huang, Y.-Z.,  Lepowsky, J.,
2907: {\em A theory of tensor products for module categories for a vertex
2908:   operator algebra}, I, II, Selecta Math. (N.S.) {\bf 1} (1995),
2909: 699--756, 757--786 
2910: 
2911: 
2912: \bibitem[HL2]{tensor3}
2913: \bysame
2914: {\em A theory of tensor products for module categories for a vertex
2915:   operator algebra}, III, J. Pure Appl. Alg. {\bf 100} (1995), 141--171
2916: 
2917: \bibitem[HL3]{HL}
2918: \bysame,
2919: {\em Intertwining operator algebras and vertex tensor categories for
2920:   affine Lie algebras}, 
2921: Duke Math. J. {\bf 99} (1999),  113--134. 
2922: 
2923: \bibitem[I1]{I}
2924: Izumi, M., 
2925: {\em Subalgebras of infinite $C^*$--algebras with finite Watanati
2926:   indices. {\rm I.} Cuntz algebras}, 
2927: Comm. Math. Phys. {\bf 155} (1993), 157--182. 
2928: 
2929: 
2930: \bibitem[I2]{I2}
2931: \bysame,
2932: {\em On flatness of the Coxeter graph $E_8$}, 
2933: Pacific J. Math. {\bf 166} (1994), 305--327.
2934: 
2935: \bibitem[Kac]{Kac}
2936: Kac, V.G.,
2937: {\em{Vertex algebras for beginners}}, Second ed., 
2938: University Lecture Series, vol. 10, 
2939: American Mathematical Society, Providence, RI, 
2940: 1998.
2941: 
2942: 
2943: \bibitem[KL]{KL}
2944: Kazhdan, D. and Lusztig, G.,
2945: {\em{Tensor structures arising from affine Lie algebras}}.
2946: I, J. Amer. Math. Soc., {\bf 6} (1993), 905--947;
2947: II, J. AMS, {\bf 6} (1993), 949--1011;
2948: III, J. AMS, {\bf 7} (1994), 335--381;
2949: IV, J. AMS, {\bf 7} (1994), 383--453.
2950: 
2951: \bibitem[L]{laredo}
2952: Toledano Laredo,
2953: {\em Fusion of positive energy representations of $LSpin_{2n}$},
2954: Ph.~D. thesis, University of Cambridge, 1997.   
2955: 
2956: 
2957: \bibitem[M1]{mckay}
2958: McKay, J.,
2959: {\em Graphs, singularities, and finite groups},
2960:  in: {\em The Santa Cruz Conference on Finite Groups}
2961: (Univ. California, Santa Cruz, Calif., 1979), pp. 183--186, 
2962: Proc. Symp. Pure Math., 
2963: Amer. Math. Soc., Providence, R.I., 1980.
2964: 
2965: \bibitem[M2]{mckay2}
2966: \bysame,
2967: {\em Semi-Affine Coxeter-Dynkin Graphs and $G\subseteq SU_2(\Cset)$},
2968: Canad. Math. J. {\bf 51}(6), 1999, pp. 1223--1229.
2969: 
2970: \bibitem[MST]{MST}
2971:   Michel, L.; Stanev, Y. S.; Todorov, I. T., 
2972:  {\em  $D$-$E$ classification of the local
2973:   extensions of ${\rm SU}\sb 2$ current algebras},  
2974:   Teoret. Mat. Fiz. 92 (1992), no. 3, 507--521;
2975:   translation in Theoret. and Math. Phys. 92 (1992), no. 3, 1063--1074
2976:   (1993). 
2977: 
2978: 
2979: 
2980: \bibitem[O1]{ocneanu}
2981: Ocneanu, A.
2982: {\em Quantized groups, string algebras and Galois theory for
2983:   algebras}, in: {\em Operator algebras and applications}, Vol. 2,
2984: 119--172, London Math. Soc. Lecture Note Ser., 136, Cambridge
2985: Univ. Press, Cambridge, 1988. 
2986: 
2987: 
2988:  \bibitem[O2]{ocneanu2}
2989: \bysame,
2990:  {\em Paths on Coxeter diagrams: From Platonic solids and singularities
2991:    to minimal models and subfactors}, 
2992:  in {\em Lectures on operator theory}, the Fields Institute Monographs, 
2993:  Amer. Math. Soc., Providence, RI, 2000, pp. 243--323.
2994: 
2995:  \bibitem[O3]{ocneanu3}
2996: \bysame,
2997:  {\em Operator Algebras, Topology and Subgroups of Quantum
2998:    Symmetry. Construction of Subgroups of Quantum Groups},
2999:  Adv. Studies in Pure Math. {\bf 31}, 2001, Taniguchi Conference on
3000:  Mathematics, Nara '98, pp. 235--263.
3001: 
3002: \bibitem[Pa]{Pa}
3003: Pareigis, B.,
3004: {\em On braiding and dyslexia}, 
3005: J.~Alg. {\bf 171}, 413--425.
3006: 
3007: 
3008: \bibitem[Po]{popa}
3009: Popa, S.,
3010: {\em Classification of subfactors and their endomorphisms}, CBMS
3011: Regional Conference Series in Mathematics, {\bf 86}, AMS, Providence,
3012: RI, 1995.
3013: 
3014: 
3015: \bibitem[PZ]{PZ}
3016: Petkova, V. and Zuber, J.-B.,
3017: {\em The many faces of Ocneanu cells}, 
3018: arXiv:hep-th/0101151
3019: 
3020: \bibitem[R]{Ros}
3021:  Rosenberg, A.
3022:  {\em The existence of fiber functors}, in: {\em The Gelfand
3023:  Mathematical seminars, 1996--1999}, pp. 145--154, Birkh\"auser,
3024:  Boston, MA, 2000.
3025: 
3026: 
3027: 
3028: \bibitem[W1]{was2}
3029: Wassermann, A.
3030: {\em Operator algebras and conformal field theory. III. Fusion of
3031:   positive energy representations of $LSU(N)$ using bounded
3032:   operators}, Invent. Math. {\bf 133} (1998), 467--538.
3033: 
3034: \bibitem[W2]{was}
3035: \bysame,
3036: {\em Quantum subgroups and vertex algebras}, lectures given at
3037: MSRI in Dec. 2000. Available from 
3038: {\tt 
3039: http://www.msri.org/publications/ln/msri/2000/subfactors/wassermann/1/}
3040: 
3041: \bibitem[X]{X}
3042: Xu, F., 
3043: {\em New braided endomorphisms from conformal inclusions}, 
3044: Comm. Math. Phys. {\bf 192} (1998), 349--403.  
3045: 
3046: \bibitem[Z]{Z}
3047: Zuber, J.-B.
3048: {\em CFT, BCFT, ADE and all that}, 
3049: {\tt arXiv:hep-th/0006151}. 
3050: 
3051: 
3052: \end{thebibliography}
3053: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3054: 
3055: \end{document}
3056: 
3057: