1: \documentclass[12pt]{article}

2: \setlength{\textheight}{23cm}

3: \pagestyle{plain}

4: \usepackage[dvips]{graphicx}

5: \usepackage{amsmath,amsthm,amsfonts,amssymb,amscd,epic,eepic}

6: 

7: 

8: \newtheorem{thm}{Theorem}[section]

9: \newtheorem{cor}[thm]{Corollary}

10: \newtheorem{defn}[thm]{Definition}

11: \newtheorem{exm}[thm]{Example}

12: \newtheorem{lem}[thm]{Lemma}

13: \newtheorem{prop}[thm]{Proposition}

14: \newtheorem{ack}[thm]{Acknowledgments}

15: \newtheorem{property}[thm]{Property}

16: \newtheorem{rem}[thm]{Remark}

17: \newtheorem{rems}[thm]{Remarks}

18: \renewcommand{\theequation}  {\thesection.\arabic{equation}}

19: 

20: 

21: \title{$(2+1)$-dimensional topological quantum field theory 

22: from subfactors and Dehn surgery formula for 3-manifold invariants}

23: 

24: \author{

25: {\sc Yasuyuki Kawahigashi}\footnote{Supported in part by the

26: Grants-in-Aid for Scientific Research, JSPS.}\\

27: Department of Mathematical Sciences\\

28: University of Tokyo, Komaba, Tokyo, 153-8914, JAPAN\\

29: e-mail: {\tt yasuyuki@ms.u-tokyo.ac.jp}\\

30: \vphantom{X}\\

31: {\sc Nobuya Sato}$^*$ \\

32: %\footnote{Supported in part by the

33: %Grants-in-Aid for Scientific Research, JSPS.}\\

34: Department of Mathematics and Information Sciences\\

35: Osaka Prefecture University, Sakai, Osaka, 599-8531, JAPAN\\

36: e-mail: {\tt nobuya@mi.cias.osakafu-u.ac.jp}\\

37: \vphantom{X}\\

38: {\sc Michihisa Wakui}\\

39: Department of Mathematics\\

40: Osaka University, Toyonaka, Osaka, 560-0043, JAPAN \\

41: e-mail: {\tt wakui@math.sci.osaka-u.ac.jp}}

42: 

43: \begin{document}

44: 

45: \maketitle

46: 

47: \begin{abstract}

48: In this paper, we establish the general theory of $(2+1)$-dimensional 

49: topological quantum field theory (in short, TQFT) with a Verlinde basis. 

50: It is a consequence that we 

51: have a Dehn surgery formula for 3-manifold invariants for this kind of 

52: TQFT's. We will show that Turaev-Viro-Ocneanu unitary TQFT's obtained from 

53: subfactors satisfy the axioms of TQFT's with Verlinde bases. Hence, in a 

54: Turaev-Viro-Ocneanu TQFT, we have a Dehn surgery formula for 3-manifolds. 

55: It turns out that this Dehn surgery formula is nothing but the formula of 

56: the Reshetikhin-Turaev invariant constructed from a tube system, 

57: which is a modular category corresponding to the quantum double construction 

58: of a $C^*$-tensor category. In the forthcoming paper, we will exbit 

59: computations of Turaev-Viro-Ocneanu invariants for several \lq\lq basic 

60: 3-manifolds ''. In Appendix, we discuss the relationship 

61: between the system of $M_\infty$-$M_\infty$ bimodules arising from the 

62: asymptotic inclusion $M \vee M^{op} \subset M_\infty$ constructed from 

63: $N \subset M$ and the tube system obtained from a subfactor $N \subset M$.

64: \end{abstract}

65: 

66: 

67: \section{Introduction}

68: Since the discovery of the celebrated Jones polynomial \cite{J}, 

69: there have been a great deal of studies on quantum invariants of knots, 

70: links, and 3-manifolds.  Among them, we deal with the two constructions

71: in this paper; Ocneanu's generalization of the Turaev-Viro 

72: invariants \cite{TuraevViro} of 3-manifolds based on triangulation

73: and the Reshetikhin-Turaev invariants \cite{RT} of those based

74: on Dehn surgery.  The former uses quantum $6j$-symbols arising from

75: subfactors and the tensor categories we have are not necessarily braided,

76: while the the latter invariants require braiding, or more precisely,

77: modular tensor categories.  Methods to construct a modular tensor

78: category from a rational tensor category have been studied by

79: several authors and are often called ``quantum double'' constructions.

80: In the setting of subfactor theory, such a method was first found

81: by Ocneanu using his generalization of the Turaev-Viro topological

82: quantum field theory (TQFT) and Izumi \cite{Izumi1} later gave a formulation

83: based on sector theory.

84: 

85: We start with a (rational unitary) tensor category (arising from

86: a subfactor), which is not braided in general.  

87: Then we obtain two TQFT's with the methods above as follows.  One is

88: a Turaev-Viro-Ocneanu TQFT based on a state sum and triangulation which

89: directly uses $6j$-symbols of the tensor category.  The other is

90: a Reshetikhin-Turaev TQFT arising from the modular tensor category

91: we obtain with the quantum double construction of the original category.

92: It is natural to ask what relation we have between these two

93: TQFT's.  In order to make a general study on such TQFT's, we first

94: establish a Dehn surgery formula (Proposition \ref{Proposition1}) 

95: for a general TQFT, which represents an invariant of

96: a 3-manifold in terms of a weight sum of invariants of links that

97: we use for the Dehn surgery construction of the 3-manifold.  This

98: formula uses a basis of the Hilbert space for $S^1\times S^1$

99: arising from the TQFT.  We show that several nice properties for

100: such a basis are mutually equivalent (Theorem \ref{Theorem6}) and we say 

101: that a basis is a {\it Verlinde basis} when such properties hold 

102: (Definition \ref{vb}).  We next show that 

103: a Turaev-Viro-Ocneanu TQFT has a Verlinde basis in this sense 

104: (Theorem \ref{verlinde-basis}) and then as a corollary of the Dehn surgery 

105: formula, we conclude that the above two TQFT's are identical for 

106: any closed 3-manifold (Theorem \ref{main}). 

107: (Actually, the claim that a Turaev-Viro-Ocneanu TQFT has a Verlinde

108: basis has been announced by Ocneanu and a proof is presented in

109: \cite[Chapter 12]{EK}, but normalizations are inaccurate

110: there, so we include a proof for this claim for the sake of completeness

111: here, along the line of Ocneanu's tube algebra and braiding arising 

112: from the Turaev-Viro-Ocneanu TQFT.)

113: This identity result has been announced by two of us in \cite{SatoWakui}.

114: It has been also announced in page 244 of Ocneanu \cite{O},

115: but it seems to us that his line of

116: arguments are different from ours which relies on our

117: Dehn surgery formula.  This result also proves a conjecture

118: in \cite[Section 8.2]{Mueger} and gives an answer to a question

119: in \cite[page 546]{Kerler}. 

120: We refer to the book \cite{EK} for subfactor theory and its applications.

121: Throughout this paper, we only consider subfactors with 

122: finite depth and finite index. 

123: 

124: \bigskip

125: \noindent

126: Acknowledgement: A part of this work was done while the first two 

127: authors visited Mathematical Sciences Research Institute at Berkeley. 

128: We appreciate their financial supports and hospitality. 

129: 

130: 

131: \section{$(2+1)$-dimensional TQFT with Verlinde basis and Dehn surgery 

132: formula}

133: 

134: Giving a $(2+1)$-dimensional topological quantum field theory $Z$, we have 

135: a 3-manifold invariant $Z(M)$ in the canonical way. The purpose of this 

136: section is to describe $Z(M)$ in terms of the $S$-matrix and framed links, 

137: and to give a criterion for the Verlinde identity \cite{Verlinde} to hold in 

138: our framework. 

139: \par 

140: Throughout this section, we use the following notations. 

141: For an oriented manifold $M$, we denote by $-M$ the same manifold with the opposite orientation. For an orientation preserving diffeomorphism $f:M\longrightarrow N$, we denote by $-f$ the same diffeomorphism viewed as an orientation preserving diffeomorphism $-M\longrightarrow -N$. We regard the empty set $\emptyset $ as an oriented closed surface with the unique orientation. The dual vector space of a finite dimensional vector space $V$ over $\mathbb{C}$ is denoted by $V^{\ast }$, and the dual basis of a basis $\{ v_i\} _i$ is denoted by $\{ v_i^{\ast }\} _i$. 

142: The closure of a subspace $X$ in the $3$-sphere $S^3$ is denoted by $\overline{X}$. 

143: 

144: \par \medskip \noindent 

145: {\bf Convention of orientations for manifolds}.\ 

146: \par 

147: Throughout this section, we assume that the $3$-sphere $S^3$ is oriented, and assume that any $3$-dimensional submanifold of $S^3$ is oriented by the orientation induced from $S^3$. We also assume that the solid torus $D^2\times S^1$ equips with the orientation such that the diffeomorphism $h:D^2\times S^1\longrightarrow \mathbb{R}^3\subset S^3$ defined by

148: $$h((x,y),e^{i\theta })=((2+x)\cos \theta , (2+x)\sin \theta ,y)$$

149: is orientation preserving, and assume that the torus $S^1\times S^1$ is oriented by the orientation induced from $D^2\times S^1$ (see Figure \ref{Figure1}). 

150:  

151: \begin{figure}[hbtp]

152: 

153: \begin{center}

154: \includegraphics[width=5cm]{sw1.eps} 

155: \caption{the orientation of torus $S^1\times S^1$ \label{Figure1}}

156: \end{center}

157: \end{figure}

158: 

159: \par \medskip 

160: Let us recall the definition of $(2+1)$-dimensional topological quantum field theory due to Atiyah \cite{Atiyah}, \cite{FSS}. 

161: By a $3$-cobordism we mean a triple $(M; \Sigma _1,\Sigma _2)$ consisting of a compact oriented $3$-manifold $M$ and two closed oriented surfaces $\Sigma _1, \Sigma _2$ such that $\partial M=(-\Sigma _1)\cup \Sigma _2$ and $\Sigma _1\cap \Sigma _2=\emptyset $. For an oriented closed surface $\Sigma $, the identity cobordism is given by $Id_{\Sigma }=(\Sigma \times [0,1];\Sigma \times \{ 0\} ,\Sigma \times \{ 1\} )$. 

162: 

163: 

164: \par 

165: \bigskip 

166: \begin{defn}

167: Let $Z$ be a functor consisting of the following three functions. 

168: \begin{enumerate}

169: \item[(1)] 

170: To each closed oriented surface $\Sigma $, it assigns a finite dimensional $\mathbb{C}$-vector space $Z(\Sigma )$.

171: \item[(2)] 

172: To each cobordism $W$, it assigns a $\mathbb{C}$-linear map $Z_W$.

173: \item[(3)] 

174: To each orientation preserving diffeomorphism $f$ between closed oriented surfaces, it assigns a $\mathbb{C}$-linear isomorphism $Z(f)$. 

175: \end{enumerate}

176: If $Z$ has the following properties, it is called a {\rm $(2+1)$-dimensional 

177: topological quantum field theory} (TQFT in short).

178: \begin{enumerate}

179: \item[(i)] $Z$ is functorial with respect to the composition of orientation preserving diffeomorphisms. More precisely, 

180: \begin{enumerate} 

181: \item[(a)] $Z(g\circ f)=Z(g)\circ Z(f)$  

182: for orientation preserving diffeomorphisms $f$ and $g$. 

183: \item[(b)] $Z(id_{\Sigma })=id_{Z(\Sigma )}$ for an oriented closed surface $\Sigma $. 

184: \end{enumerate}

185: \item[(ii)] $Z$ is fuctorial with respect to the composition of $3$-cobordisms. More precisely, 

186: \begin{enumerate} 

187: \item[(a)] If two cobordisms $W_1=(M_1; \Sigma _1,\Sigma _2)$ and $W_2=(M_2; \Sigma _2,\Sigma _3)$ are obtained by cutting a cobordism $W=(M;\Sigma _1,\Sigma _3)$ along $\Sigma _2$ in $M=M_1\cup M_2$, then 

188: $$Z_W=Z_{W_2}\circ Z_{W_1}.$$

189: \item[(b)] $Z(i_1)^{-1}\circ Z_{Id_{\Sigma }}\circ Z(i_0)=id_{Z(\Sigma )}$ for the identity cobordism on $\Sigma $, where $i_t:\Sigma \longrightarrow \Sigma \times \{ t\} $ \ $(t=0,1)$ are orientation preserving diffeomorphisms defined by $i_t(x)=(x,t),\ x\in \Sigma $. 

190: \end{enumerate}

191: \item[(iii)] Let $W=(M,\Sigma _1,\Sigma _2),\ W'=(M',\Sigma _1',\Sigma 

192: _2')$ be two cobordisms. Suppose that there is an orientation preserving 

193: diffeomorphism $h:M\longrightarrow M'$ such that $h(\Sigma _i)=\Sigma 

194: _i'$ for $i=1,2$. Then, for $f_1:=-h\vert _{\Sigma _1}:\Sigma 

195: _1\longrightarrow \Sigma _1' \ and \ f_2:=h\vert _{\Sigma _2}:\Sigma _2\longrightarrow \Sigma _2'$, the following diagram commutes. 

196: $$\begin{CD}

197: Z(\Sigma _1)@>Z(f_1)>> Z(\Sigma _1') \\ 

198: @VZ_{W}VV @VVZ_{W'}V \\ 

199: Z(\Sigma _2) @>>Z(f_2)> Z(\Sigma _2')

200: \notag 

201: \end{CD}$$

202: \item[(iv)] Let $W_1=(M;\Sigma _1,\Sigma _2)$, $W_2=(N;\Sigma _2',\Sigma _3)$ be two cobordisms, and $f:\Sigma _2\longrightarrow \Sigma _2'$ an orientation preserving diffeomorphism. Then, for the cobordism $W:=(N\cup _fM; \Sigma _1, \Sigma _3)$ obtained by gluing of $W_1$ to $W_2$ along $f$, 

203: $$Z_W=Z_{W_2}\circ Z(f)\circ Z_{W_1}.$$

204: \item[(v)] There are natural isomorphisms 

205: \begin{enumerate}

206: \item[(a)] $Z(\emptyset )\cong \mathbb{C}$. 

207: \item[(b)] $Z(\Sigma _1\coprod \Sigma _2)\cong Z(\Sigma _1)\otimes Z(\Sigma _2)$ for oriented closed surfaces $\Sigma _1$ and $\Sigma _2$.

208: \item[(c)] $Z(-\Sigma )\cong Z(\Sigma )^{\ast }$ for an oriented closed surface $\Sigma $. 

209: \end{enumerate}

210: \end{enumerate}

211: \end{defn}

212: 

213: \par \medskip 

214: \begin{rems}

215: 1. For an oriented closed $3$-manifold $M$, we have a cobordism $W=(M;\emptyset ,\emptyset )$. This cobordism $W$ induces a linear map 

216: $$\mathbb{C}\cong Z(\emptyset )\xrightarrow {Z_W}Z(\emptyset )\cong \mathbb{C}.$$

217: We denote by $Z(M)$ the image of $1$ under the above map. By the condition (iii) we see that $Z(M)$ is a topological invariant of $M$. 

218: \par 

219: \noindent 

220: 2. Let $\varGamma_{\Sigma }$ denote the mapping class group of the oriented closed surface $\Sigma $. Then, we have a representation of $\varGamma_{\Sigma }$

221: $$\rho :\varGamma_{\Sigma }\longrightarrow GL(Z(\Sigma )),\ [f] \longmapsto Z(f),$$

222: where $[f]$ denotes the isotopy class of $f$. 

223: \par 

224: \noindent 

225: 3. For an oriented closed surface $\Sigma $, we have $\dim Z(\Sigma )=Z(\Sigma \times S^1)$. 

226: \par 

227: \noindent 

228: 4. By a $3$-cobordism with parametrized boundary we mean a triple $(M; j_1,j_2)$ consisting of a compact oriented $3$-manifold $M$, an orientation reversing embedding $j_1:\Sigma _1\longrightarrow \partial M$ and an orientation preserving embedding $j_2:\Sigma _2\longrightarrow \partial M$ such that $\partial M=(-j_1(\Sigma _1))\cup j_2(\Sigma _2)$ and $j_1(\Sigma _1)\cap j_2(\Sigma _2)=\emptyset $. Any cobordism with parametrized boundary ${\cal W}=(M; \Sigma _1\xrightarrow {\ j_1\ } \partial M,\Sigma _2 \xrightarrow {\ j_2\ } \partial M)$ induces a linear map $Z_{\cal W}:Z(\Sigma _1)\longrightarrow Z(\Sigma _2)$ such that the following diagram commutes.

229: 

230: \begin{equation*}

231: \begin{CD}

232: Z(\Sigma _1) @>Z_{\cal W}>> Z(\Sigma _2) \\ 

233: @VZ(j_1)VV @VVZ(j_2)V \\ 

234: Z(j_1(\Sigma _1)) @>Z_W>> Z(j_2(\Sigma _2)) 

235: \end{CD} 

236: \end{equation*}

237: 

238: Here, we set $W=(M;j_1(\Sigma _1),j_2(\Sigma _2))$. 

239: \end{rems}

240: 

241: \par \bigskip 

242: \par \bigskip 

243: Let $Z$ be a $(2+1)$-dimensional TQFT. We

244: consider the cobordism $W:=(Y\times S^1; \Sigma _1\sqcup \Sigma _2, \Sigma _3)$, where $Y$ is

245: the compact oriented surface in $\mathbb{R}^3$

246: depicted in Figure \ref{Figure5} and $\Sigma _i=C_i\times S^1$ for $i=1,2,3$. 

247: Then, $W$ induces a 

248: linear map $Z_W: Z(S^1\times S^1)\otimes Z(S^1\times S^1)\longrightarrow

249: Z(S^1\times S^1)$. It 

250: can be easily verified that the map $Z_W$ gives an associative algebra structure

251: on $Z(S^1\times S^1)$. The identity element of this algebra is given by

252: $Z_{W_0}(1)$, where $W_0:=(D^2\times S^1; \emptyset , S^1\times S^1)$. We

253: call this algebra the {\it fusion algebra associated with $Z$}.

254: 

255: \begin{figure}[hbtp]

256: \begin{center}

257: \scalebox{1.3}{\includegraphics[height=2.5cm]{sw5.eps} }

258: \caption{the compact oriented surface $Y$ \label{Figure5}}

259: \end{center}

260: \end{figure}

261: 

262: Let us introduce a Dehn surgery formula of $Z(M)$. 

263: Let $Z$ be a $(2+1)$-dimensional TQFT, and $S:S^1\times S^1\longrightarrow S^1\times S^1$ the orientation preserving diffeomorphism defined by $S(z,w)=(\bar{w},z),\ (z,w)\in S^1\times S^1$, where we regard $S^1$ as the set of complex numbers of absolute value $1$. Given a basis $\{ v_i\} _{i=0}^m$ of $Z(S^1\times S^1)$, we define $S_{ji}\in \mathbb{C}$ by $Z(S)v_i=\sum_{j=0}^mS_{ji}v_j$. 

264: \par 

265: Let $\{ v_i\} _{i=0}^m$ be a basis of $Z(S^1\times S^1)$. Then, we can define a framed link invariant as follows. Let $L=L_1\cup \cdots \cup L_r$ be a framed link with $r$-components in the 3-sphere $S^3$, and $h_i: D^2\times S^1\longrightarrow N(L_i)$ be the framing of $L_i$ for each $i\in \{ 1, \cdots , r\} $, where $N(L_i)$ denotes the tubular neighborhood of $L_i$. 

266: We fix an orientation for $\partial N(L_i)$ such that $j_i:=h_i\vert _{\partial D^2\times S^1}:S^1\times S^1\longrightarrow \partial N(L_i)$ is orientation preserving. Since the orientation for $N(L_i)$ is not compatible with the orientation for the link exterior $X:=\overline{S^3-N(L_1)\cup \cdots \cup N(L_r)}$, we can consider the cobordism with parametrized boundary ${\cal W}_L:=(X; \coprod \limits_{i=1}^r j_i ,\emptyset )$. 

267: This cobordism induces a $\mathbb{C}$-linear map $Z_{{\cal W}_L}: \bigotimes \limits_{i=1}^rZ(S^1\times S^1)\longrightarrow \mathbb{C}$. 

268: It is easy to see that for each $i_1,\cdots ,i_r=0,1,\cdots ,m$ the complex number $J(L; i_1, \cdots , i_r):=Z_{{\cal W}_L}(v_{i_1}\otimes \cdots \otimes v_{i_r})$  

269: is a framed link invariant of $L$. 

270: 

271: \par\bigskip 

272: \begin{prop}

273: \label{Proposition1}

274: Let $Z$ be a (2+1)-dimensional TQFT and  $\{ v_i\} _{i=0}^m$ a basis of $Z(S^1\times S^1)$ such that $v_0$ is the identity element in the fusion algebra. 

275: Let $M$ be a closed oriented 3-manifold obtained from $S^3$ by Dehn surgery along a framed link $L=L_1\cup \cdots \cup L_r$. 

276: Then, the 3-manifold invariant $Z(M)$ is given by the formula

277: $$Z(M)=\sum_{i_1, \cdots , i_r=0}^m S_{i_1,0}\cdots S_{i_r,0}J(L;i_1, \cdots , i_r).$$ 

278: \end{prop}

279: 

280: \medskip \noindent 

281: {Proof}.\ 

282: Let $X$ be the link exterior of $L$ in $S^3$, and $h_i:D^2\times S^1\longrightarrow \partial N(L_i)$ the framing of $L_i$ for each $i\in \{ 1,\cdots ,r\} $. 

283: Then, by using attaching maps $f_i: S^1\times S^1\longrightarrow \partial N(L_i)$ satisfying $f_i(S^1\times 1)=h_i(1\times S^1)$, $i=1,\cdots ,r$, we have 

284: $$M=X\bigcup \nolimits_{\amalg _{i=1}^rf_i}(\coprod^rD^2\times S^1).$$

285: Therefore, 

286: $$Z(M)=Z_{W_2}\circ Z(\coprod _{i=1}^rf_i)\circ Z_{W_1},$$

287: where $W_1:=(\coprod^r(D^2\times S^1); \emptyset , \coprod ^r(S^1\times S^1))$ and $W_2:=(X; -\partial X, \emptyset )$. This implies that 

288: $$Z(M)=Z_{\mathcal{W}_L}\circ (\bigotimes _{i=1}^rZ(j_i^{-1}))\circ (\bigotimes _{i=1}^rZ(f_i))\circ (\bigotimes ^rZ_{W_0}),$$

289: where $j_i:=h_i\vert _{\partial D^2\times S^1}:S^1\times S^1\longrightarrow \partial N(L_i)$ for $i=1,\cdots ,r$. 

290: \par 

291: For each $i$ we can choose $f_i$ satisfying $j_i^{-1}\circ f_i=S$ up to isotopy 

292: as for $f_i$ satisfying $f_i(S^1\times 1)=h_i(1\times S^1)$.    

293: Then, we have 

294: $$Z(M)=Z_{\mathcal{W}_L}\circ (\bigotimes _{i=1}^rZ(S))\circ (\bigotimes ^rZ_{W_0}).$$

295: This implies that 

296: $$Z(M)=\sum_{i_1, \cdots , i_r=0}^m S_{i_1,0}\cdots S_{i_r,0}J(L;i_1, \cdots , i_r).$$ 

297: This completes the proof. \hfill Q.E.D. 

298: \par \bigskip 

299: 

300: \begin{figure}[hbtp]

301: \begin{center}

302: \includegraphics[width=4cm]{sw2.eps} 

303: \caption{the Hopf link $H$ \label{Figure2}}

304: \end{center}

305: \end{figure}

306: 

307: \par \bigskip 

308: \begin{prop}

309: \label{Proposition2}  

310: Let $Z$ be a $(2+1)$-dimensional TQFT and $\{ v_i\} _{i=0}^m$ a basis of $Z(S^1\times S^1)$. 

311: Let $H$ be the Hopf link depicted as in Figure \ref{Figure2}, and $U$ the orientation preserving diffeomorphism from $S^1\times S^1$ to $-S^1\times S^1$ defined by $U(z,w)=(z,\bar{w})$. 

312: Then, the following are equivalent. 

313: \begin{enumerate}

314: \item[(i)] $J(H;i,j)=S_{ij}$ for all $i,j=0,1,\cdots ,m$. 

315: \item[(ii)] $Z(U)v_i=v_{i}^{\ast }$ for all $i=0,1,\cdots ,m$.

316: \end{enumerate} 

317: \end{prop}

318: 

319: \noindent 

320: {Proof}. \  

321: Let $h_i:D^2\times S^1\longrightarrow N(H_i)$ be the framing of $H_i$ for each $i=1,2$. Putting $j_i=h_i\vert _{S^1\times S^1}\ (i=1,2)$ and $X:=\overline{S^3-N(H_1)\cup N(H_2)}$, we have a cobordism with parametrized boundary ${\cal W}:=(X; j_1,-j_2)$. 

322: Then 

323: $$Z_{\cal W}(v_i)=\sum\limits_{i=0}^mJ(H;i,j)v_j^{\ast }.$$

324:  Let 

325: $V_1$ be the tubular neighborhood of $N(H_1)$ depicted  as in Figure \ref{Figure3}. 

326: 

327: \begin{figure}[htbp]

328: \begin{center}

329: \setlength{\unitlength}{1cm}

330: \includegraphics[height=4cm]{sw3.eps}

331: \caption{ \label{Figure3}}

332: \end{center}

333: \end{figure} 

334: 

335: Setting $V_2:=\overline{S^3-V_1}$, and defining orientation preserving diffeomorphisms $j_1':S^1\times S^1\longrightarrow \partial V_1$ and $j_2':S^1\times S^1\longrightarrow \partial V_2$ in parallel to $j_1$ and $j_2$ respectively, we see that 

336: \par 

337: $\bullet$\ $\partial V_1=-\partial V_2$,

338: \par 

339: $\bullet$\ $X=\overline{V_1-N(H_1)}\cup \overline{V_2-N(H_2)}$,

340: \par 

341: $\bullet$\ ${\cal W}_1:=(\overline{V_1-N(H_1)}; j_1, j_1')\cong Id_{S^1\times S^1}\quad \text{as cobordisms}$,

342: \par 

343: $\bullet$\ ${\cal W}_2:=(\overline{V_2-N(H_2)}; -j_2', -j_2)\cong Id_{-S^1\times S^1}\quad \text{as cobordisms}$. 

344: 

345: 

346: Since  

347: $$(-j_2')^{-1}\circ j_1':\ 

348: \begin{cases} 

349: m \longmapsto -l, \cr  

350: l \longmapsto -m. \end{cases}$$

351: for the meridian $m$ and the longitude $l$ (see Figure \ref{Figure4}), it follows that 

352: $$Z(-j_2'{^{-1}}\circ j_1')=Z(U\circ S).$$ 

353: 

354: \begin{figure}[htbp]

355: \begin{center}

356: \setlength{\unitlength}{1cm}

357: \includegraphics[width=5cm]{sw4.eps}

358: \caption{\label{Figure4}}

359: \end{center}

360: \end{figure}

361: Therefore, we obtain 

362: $$Z_{\cal W}=Z_{{\cal W}_2}\circ Z(U\circ S)\circ Z_{{\cal W}_1},$$

363: and whence 

364: $$\sum\limits_{j=0}^mJ(H;i,j)v_j^{\ast }=Z_{\cal W}(v_i)=(\theta \circ 

365: Z(U\circ S))(v_i),$$

366: where $\theta :Z(-S^1 \times S^1) \longrightarrow Z(S^1 \times S^1)^{\ast }$ 

367: denotes the natural isomorphism. 

368: Let $(U_{ij})_{i,j=0,1,\cdots ,m}$ be the presentation matrix of 

369: $\theta \circ Z(U):Z(S^1 \times S^1) \longrightarrow Z(S^1\times S^1)^{\ast }$ 

370: with respect to $\{ v_i \}_{i=0}^m$ and $\{ v_i^{\ast }\}_{i=0}^m$. 

371: Then, we have 

372: $$(J(H;i,j))_{i,j=0,1,\cdots ,m}=(U_{ij})(S_{ij}).$$

373: Hence, we have 

374: \begin{eqnarray*}

375: &(J(H;i,j))_{i,j=0,1,\cdots ,m}=(S_{ij})_{i,j=0,1,\cdots ,m} \notag \\ 

376: \Longleftrightarrow \ & (U_{ij})=I,\ \text{where $I$ is the identity matrix.} \notag \\ 

377: \Longleftrightarrow \ & (\theta \circ Z(U))(v_i)=v_{i}^{\ast } \ 

378: (i=0,1,\cdots ,m). \qquad \qquad \qquad \notag 

379: \end{eqnarray*}

380: This completes the proof. \hfill Q.E.D. 

381: 

382: Let us recall that the mapping class group $\varGamma_{S^1\times S^1}$ of the torus $S^1\times S^1$ is isomorphic to the group 

383: $SL_2(\mathbb{Z})$ of integral $2\times 2$-matrices with determinant $1$. It is well-known that this group is generated by $S=\begin{pmatrix} 0 & 1 \\ -1 & 0

384: \end{pmatrix}$ and $T=\begin{pmatrix} 1 & 0 \\ 1 & 1\end{pmatrix}$ with relations

385: $S^4=I,\ (ST)^3=S^2$. The matrices $S$ and $T$ correspond to the orientation preserving diffeomorphisms $S^1\times S^1$ to $S^1\times S^1$ which are defined by $S(z,w)=(\bar{w},z)$ and $T(z,w)=(zw,w),\ (z,w)\in S^1\times S^1$, respectively, where we regard $S^1$ as the set of complex numbers of absolute value $1$ . 

386: To define the Verlinde basis, we need one more orientation preserving diffeomorphism $U: S^1\times S^1\longrightarrow - S^1\times S^1$ defined by $U(z,w)=(z,\bar{w})$ for $(z,w)\in S^1\times S^1$ (see Figure \ref{Figure6}). 

387: 

388: \par \bigskip 

389: \begin{defn}\label{vb}\ \ 

390: Let $Z$ be a $(2+1)$-dimensional TQFT. A basis $\{ v_i\} _{i=0}^m$ of $Z(S^1\times S^1)$ is said to be a {\it Verlinde basis} if it has the following properties. 

391: \begin{enumerate} 

392: \item[(i)] $v_0$ is the identity element of the fusion algebra associated 

393: with $Z$. 

394: \item[(ii)] 

395: \begin{enumerate} 

396: \item[(a)] $Z(S)$ is presented by a unitary and symmetric matrix with respect to the basis $\{ v_i\} _{i=0}^m$.

397: \item[(b)] $Z(S)^2v_0=v_0$, and $Z(S)^2v_i\in \{ v_j\} _{j=0}^m$ for all $i$. 

398: \item[(c)] We define $S_{ji}\in \mathbb{C}$ by $Z(S)v_i=\sum_{i=0}^mS_{ji}v_j$. Then, 

399: \begin{enumerate} 

400: \item[1.] $S_{i0}\not= 0$ for all $i$. 

401: \item[2.] $N_{ij}^k:=\sum_{l=0}^m\frac{S_{il}S_{jl}\overline{S_{lk}}}{S_{0l}}\ (i,j,k=0,1,\cdots ,m)$ coincide with the structure constants of the fusion algebra with respect to $\{ v_i\} _{i=0}^m$. 

402: \end{enumerate}

403: \end{enumerate}

404: \item[(iii)] $Z(T)$ is presented by a diagonal matrix with respect to the basis $\{ v_i\} _{i=0}^m$. 

405: \item[(iv)] $Z(U)v_i=v_{i}^{\ast }$ for all $i$ under the identification $Z(-S^1\times S^1)\cong Z(S^1\times S^1)^{\ast }$. 

406: \end{enumerate}

407: We call a {\it pre-Verlinde basis} a basis $\{ v_i\} _{i=0}^m$ of $Z(S^1\times S^1)$ satisfying the four conditions (i), (ii.a), (ii.b) and $Z(U)v_0=v_0^{\ast }$. 

408: \end{defn}

409: 

410: \par \bigskip 

411: \begin{rems}

412: 1. If $\{ v_i\} _{i=0}^m$ is a pre-Verlinde basis, then $S_{i0}$ is a real number for all $i$.  

413: \par 

414: \noindent 

415: 2. A Verlinde basis is unique up to order of elements, since $w_i=S_{0i}\sum_{j=0}^m\overline{S_{ji}}v_j$\ $(i=0,1,\cdots ,m)$ are all orthogonal primitive idempotents in the fusion algebra satisfying $1=w_0+w_1+\cdots +w_m$. This fact follows from that the $S$-matrix diagonalizes the fusion rules in conformal field theory \cite{Verlinde}. 

416: \par 

417: \noindent 

418: 3. The map $\overline{\mathstrut \ \cdot \ }:\{ 0,1,\cdots ,m\} \longrightarrow \{ 0,1,\cdots ,m\} $ defined by $Z(S)^2v_i=v_{\bar{i}}$ is an involution satisfying $\bar{0}=0$. 

419: \par 

420: \noindent 

421: 4. The last condition (iv) was introduced in \cite{Wakui} and modified in \cite{SuzukiWakui}. 

422: \end{rems}

423: 

424: \begin{figure}[hbtp]

425: \begin{center}

426: \setlength{\unitlength}{1cm}

427: \includegraphics[height=6cm]{sw6.eps}

428: \caption{the action of $SL_2(\mathbb{Z})$ \label{Figure6}}

429: \end{center}

430: \end{figure}

431: 

432: \par \bigskip 

433: Let us describe some basic results on a $(2+1)$-dimensional TQFT with a 

434: pre-Verlinde basis. 

435: 

436: 

437: \par \bigskip 

438: \begin{lem}

439: \label{Lemma3}

440: Let $Z$ be a (2+1)-dimensional TQFT and $\{ v_i\} _{i=0}^m$ a pre-Verlinde basis. 

441: Then, for the cobordism $W=(-D^2\times S^1; S^1\times S^1, \emptyset )$, we have  

442: $Z_W=v_0^{\ast }: Z(S^1\times S^1)\longrightarrow \mathbb{C}$. In particular, $Z(S^2\times S^1)=1$. 

443: \end{lem}

444: 

445: \par \medskip \noindent 

446: {Proof}.\ \ 

447: By considering the orientation preserving diffeomorphism $\tilde{U}: D^2\times S^1\longrightarrow -D^2\times S^1$ defined by $\tilde{U}(z, w)=(z,\bar{w})$, we have $(Z(\tilde{U}\vert _{\partial D^2\times S^1}))(Z(D^2\times S^1))=Z(-D^2\times S^1)\in Z(-S^1\times S^1)$. Since $\tilde{U}\vert _{\partial D^2\times S^1}=U$ and $Z(D^2\times S^1)=v_0$, it follows that $v_0^{\ast }=Z(-D^2\times S^1)=Z_W$ as elements in $Z(S^1\times S^1)^{\ast }$. 

448: \par 

449: Next, we prove that $Z(S^2\times S^1)=1$. The $3$-manifold $S^2\times S^1$ is regarded as $(D_{+}^2\times S^1)\cup (-D_{-}^2\times S^1)$, where $D_{+}^2=\{ (x,y,z)\in S^2 \ \vert \ z\geq 0\} ,\ D_{-}^2=\{ (x,y,z)\in S^2 \ \vert \ z\leq 0\} $. Thus, for the cobordisms $W_1:=(D_{+}^2\times S^1; \emptyset, \partial D_{+}^2\times S^1)$ and $W_2:=(-D_{-}^2\times S^1; \partial D_{-}^2\times S^1,\emptyset )$, we have  

450: $$Z(S^2\times S^1)=Z_{W_2}\circ Z_{W_1}.$$ 

451: Since $W_1\cong (D^2\times S^1;\emptyset ,S^1\times S^1)=W_0$ and $W_2\cong (-D^2\times S^1;S^1\times S^1,\emptyset )=W$, it follows that 

452: $$Z(S^2\times S^1)=Z_{W}\circ Z_{W_0}=v_0^{\ast }(v_0)=1.$$ 

453: This completes the proof. \hfill Q.E.D. 

454: 

455: \par \bigskip 

456: \begin{lem} 

457: \label{Lemma4}

458: Let $Z$ be a (2+1)-dimensional TQFT and $\{ v_i\} _{i=0}^m$ a pre-Verlinde basis. Then, 

459: $J(\text{\LARGE $\bigcirc $};i)=S_{i0}$ for each $i\in \{ 0,1,\cdots ,m\} $.

460: \end{lem}

461: 

462: \par \medskip 

463: \noindent 

464: {Proof}. \ 

465: Let $K$ be the trivial knot with $0$-framing given by unit circle in $\mathbb{R}^3$, and $X=\overline{S^3-N(K)}$ the knot exterior of $K$. 

466: The cobordism $W_K=(X; \partial N(K), \emptyset )$ is isomorphic to the cobordism $W:=(-D^2\times S^1; S^1\times S^1, \emptyset )$ via the orientation preserving diffeomorphism $f:X\longrightarrow -D^2\times S^1$ such that  

467: $f(M)=l$ and $f(L)=-m$, where $M$ and $L$ are simple closed curves on $\partial N(K)$ depicted in Figure \ref{Figure7}. 

468: 

469: 

470: \begin{figure}[htbp]

471: \vspace{0.5cm}

472: \begin{center}

473: \setlength{\unitlength}{1cm}

474: \scalebox{0.9}[0.9]{\includegraphics[height=6cm]{sw7.eps}}

475: \caption{\label{Figure7}}

476: \end{center}

477: \end{figure}

478: 

479: \par 

480: Then, we have 

481: $$Z_{W_K}=Z_{W}\circ Z(f\vert _{\partial N(K)}).$$

482: Since 

483: $$Z(f\vert _{\partial N(K)})\circ Z(j)=Z(S)$$

484: for $j:=h\vert _{S^1\times S^1}$, where $h:D^2\times S^1\longrightarrow N(K)$ is the framing of $K$,   

485: we have 

486: $$J(K,i)=\langle v_0^{\ast },\ Z(S)(v_i)\rangle =S_{i0}.$$

487: This completes the proof.  \hfill Q.E.D. 

488: 

489: \par \bigskip 

490: The framed link invariants $J(L;i_1,\cdots ,i_r)$, $(i_1, \dots, i_r=0,1,\dots, 

491: m)$ have the following nice properties. 

492: 

493: \par 

494: \begin{figure}[hbtp]

495: \begin{center}

496: \setlength{\unitlength}{1cm}

497: \includegraphics[height=2cm]{sw8.eps} 

498: \caption{a positive curl and a negative curl \label{Figure8}}

499: \end{center}

500: \end{figure}

501: 

502: \par \bigskip 

503: \begin{lem} 

504: \label{Lemma5}

505: Let $Z$ be a $(2+1)$-dimensional TQFT and $\{ v_i\} _{i=0}^m$ a pre-Verlinde basis. 

506: For a framed link $L=L_1\cup L_2\cup \cdots \cup L_r$, the invariant $J(L; i_1, \cdots , i_r)$ has the following properties.  

507: \par 

508: (1) Let $L'=L_1\cup \cdots \cup (-L_k)\cup \cdots \cup L_r$ be the framed link obtained from $L$ by changing the orientation for the $k$-th component $L_k$. Then, 

509: $$J(L;i_1,\cdots ,i_k,\cdots ,i_r)=J(L';i_1, \cdots ,\bar{i}_k,\cdots ,i_r).$$

510: \indent 

511: (2) Suppose that $\{ v_i\} _{i=0}^m$ satisfies the condition (iii) in the definition of Verlinde basis. We define $t_i\in \mathbb{C}\ (i=0,1,\cdots ,r)$ by $Z(T)v_i=t_iv_i$. 

512: Let $L'$ be the framed link obtained from $L$ such that it is same as $L$ except for a small segment of the $k$-th component $L_k$ and the small segment is replaced by a positive or negative curl as shown in Figure \ref{Figure8}. Then, the following equations hold. 

513: \par 

514: If the small segment is replaced by a positive curl, then  

515: $$J(L'; i_1,\cdots ,i_r)=t_{i_k}^{-1}J(L;i_1,\cdots ,i_r).$$

516: If the small segment is replaced by a negative curl, then 

517: $$J(L'; i_1,\cdots ,i_r)=t_{i_k}J(L;i_1,\cdots ,i_r).$$

518: \end{lem}

519: 

520: \noindent 

521: {Proof}.\  

522: Without loss of generality, we may suppose that $k=1$. 

523: 

524: \noindent

525: (1) Let $h_i:D^2\times S^1\longrightarrow N(L_i)$ be the framing of $L_i$ for each $i=1,\cdots ,r$, and $h_1':D^2\times S^1\longrightarrow N(-L_1)$ the framing of $-L_1$. We set $j_i:=h_i\vert _{S^1\times S^1}$ for each $i=1,\cdots ,r$, and $j_1':=h_1'\vert _{S^1\times S^1}$. 

526: Then, 

527: $$j_1'=j_1\circ S^2,$$

528: since $h_1'=h_1\circ f$ for the orientation preserving diffeomorphism $f:D^2\times S^1\longrightarrow D^2\times S^1$ defined by $f(z,w)=(\bar{z},\bar{w})$ (see Figure \ref{Figure9}).

529: 

530: \begin{figure}[hbtp]

531: \begin{center}

532: \setlength{\unitlength}{1cm}

533: \includegraphics[height=5cm]{sw9.eps}

534: \caption{\label{Figure9}}

535: \end{center}

536: \end{figure}

537: 

538: Let $X$ be the link exterior of $L$. 

539: Since the two cobordisms 

540: $${\cal W}_L=(X; \coprod_{i=1}^rj_i, \emptyset ),\quad 

541: {\cal W}_{L'}=(X;j_1'\coprod (\coprod _{i=2}^rj_i), \emptyset )$$ 

542: are isomorphic via the identity map $id_X$, we see that 

543: $$Z_{{\cal W}_{L'}}\circ (Z(S^2)\circ id \otimes \cdots \otimes id)=Z_{{\cal W}_L}.$$

544: Since $Z(S^2)v_i=v_{\bar{i}}$, we have 

545: $$J(L;i_1,i_2,\cdots ,i_r)=J(L';\bar{i}_1, i_2,\cdots ,i_r).$$

546: \par 

547: (2) We suppose that $L'$ arises from $L$ by replacing a small segment of the first component $L_1$ by a positive curl. 

548: Let $h_1':D^2\times S^1\longrightarrow N(L_1')$ be the framing of $L_1'$. We set $j_1':=h_1'\vert _{S^1\times S^1}$. 

549: 

550: \begin{figure}[hbtp]

551: \begin{center}

552: \setlength{\unitlength}{1cm}

553: \includegraphics[height=6cm]{sw10.eps} 

554: \caption{\label{Figure10}}

555: \end{center}

556: \end{figure}

557: 

558: Let $X'$ be the link exterior of $L'$.

559: Then, there exists an orientation preserving diffeomorphism

560: $f:X'\longrightarrow X$

561: such that

562: \begin{enumerate}

563: \item[(i)] $j_1^{-1}\circ (-f\vert _{\partial N(L_1')})\circ j_1'=T^{-1}$ up

564: to isotopy,

565: \item [(ii)] $-f\vert _{\partial N(L_i)}=id_{\partial N(L_i)}$ for all

566: $i=2,\cdots ,r$.

567: \end{enumerate}

568: 

569: This map $f$ gives rise to the isomorphism between

570: the cobordisms ${\cal W}_{L'}=(X';\emptyset ,j_1'\coprod (\coprod

571: _{i=2}^rj_i))$ and ${\cal W}_L=(X;\emptyset ,\coprod _{i=1}^rj_i)$. Hence,

572: we have

573: $$Z_{{\cal W}_{L'}}\circ (Z({j_1'}^{-1})\otimes (\bigotimes

574: _{i=2}^rZ(j_i^{-1})))=Z_{{\cal W}_L} \circ (\bigotimes

575: _{i=1}^rZ(j_i^{-1})\circ Z(-f\vert _{\partial N(L_i)})).$$

576: It follows that

577: $$Z_{{\cal W}_{L'}}=Z_{{\cal W}_L} \circ (Z(T^{-1})\otimes id \cdots \otimes

578: id ).$$

579: This implies that

580: $$J(L'; i_1,\cdots ,i_r)=t_{i_1}^{-1}J(L;i_1,\cdots ,i_r).$$

581: 

582: Thus, the proof of the first equation of (2) is completed.

583: By a similar argument, the second equation of (2) can be proved.

584: This completes the proof. \hfill Q.E.D. 

585: 

586: \begin{figure}[htpb]

587: \begin{center}

588: \setlength{\unitlength}{1cm}

589: \includegraphics[height=5cm]{sw11.eps}

590: \caption{\label{Figure11}}

591: \end{center}

592: \end{figure} 

593: 

594: \par \bigskip 

595: Let us introduce a criterion for the Verlinde identity (ii.c.2). 

596: \par 

597: We suppose that a $(2+1)$-dimensional TQFT $Z$ arises from a semisimple ribbon $\mathbb{C}$-linear Ab-category in the sense of Turaev \cite{Turaev}. Let $\{ v_i\} _{i=0}^m$ be a pre-Verlinde basis satisfying the equivalent conditions (1) and (2) in Proposition \ref{Proposition2}. We suppose $S_{i0}\not= 0$ for all $i=0,1,\cdots ,m$.   

598: \par 

599: For the $3$-component framed link $L$ presented by the diagram as in Figure \ref{Figure11}, 

600: the framed link invariant $J(L;i,j,k)$ coincides with the quantum trace of $J(T_{i,l})\circ J(T_{j,l})$, where  $T_{i,l}$ is the colored framed tangle presented by the diagram as in Figure \ref{Figure12}. 

601: \par 

602: Since $J(T_{i,l})$ is a map from the simple object $l$ to $l$, the invariant $J(T_{i,l})$ is a scalar multiple by $id_l$. We define $a_{i}$ to be this 

603: scalar. Since the quantum trace of $T_{i,l}$ is the Hopf link with colors $i,l$, we see that 

604: $$a_i\cdot J(\text{\LARGE $\bigcirc $};l)=S_{il}.$$

605: This implies that 

606: 

607: \begin{equation}

608: J(L;i,j,l)=\dfrac{S_{il}S_{jl}}{S_{l0}}. \label{eq1} 

609: \end{equation}

610: 

611: \begin{figure}[hbtp]

612: \begin{center}

613: \setlength{\unitlength}{1cm}

614: \scalebox{0.9}[0.9]{\includegraphics[width=3.5cm]{sw12.eps}}

615: \caption{the colored $1-1$ tangle $T_{i,l}$ \label{Figure12}}

616: \end{center}

617: \end{figure} 

618: 

619: \par 

620: On the other hand, by fusing $L_i$ and $L_j$ we have 

621: 

622: \begin{equation}

623: J(L;i,j,l)=\sum\limits_{i=0}^mN_{ij}^kS_{kl},\label{eq2} 

624: \end{equation}

625: where $N_{ij}^k\ (i,j,k=0,1,\cdots ,m)$ are the structure constants of the fusion algebra with respect to $\{ v_i\}_{i=0}^m$. 

626: From (\ref{eq1}) and (\ref{eq2}), it follows that 

627: $$\sum\limits_{i=0}^mN_{ij}^kS_{kl}=\dfrac{S_{il}S_{jl}}{S_{l0}}.$$

628: Since the above equation induces the Verlinde identity 

629: $$N_{ij}^k=\sum_{l=0}^m\frac{S_{il}S_{jl}\overline{S_{lk}}}{S_{l0}},$$

630: the condition (2) in Proposition \ref{Proposition2} implies the condition (ii.c.2) in the definition of Verlinde basis (see \cite{Takata}, \cite{Witten} for  similar arguments). 

631: If $\{ v_i\} _{i=0}^m$ satisfies the condition (iii) in the definition of Verlinde basis, then we see that the converse is true by using the technique in 

632: \cite{Kohno} (see Figure \ref{Figure13}). 

633: 

634: \begin{figure}[htbp]

635: \begin{center}

636: \setlength{\unitlength}{1cm}

637: \scalebox{0.9}[0.9]{\includegraphics[height=3cm]{sw13.eps}}

638: \caption{\label{Figure13}}

639: \end{center}

640: \end{figure}

641: 

642: Thus, we conclude the following. 

643: 

644: \par \bigskip 

645: \begin{thm} 

646: \label{Theorem6}

647: Suppose that a $(2+1)$-dimensional TQFT $Z$ arises from a semisimple ribbon $\mathbb{C}$-linear Ab-category. 

648: Let $\{ v_i\} _{i=0}^m$ be a pre-Verlinde basis satisfying $S_{i0}\not= 0$ for all $i=0,1,\cdots ,m$. 

649: If $Z(U)v_i=v_{i}^{\ast }$ for all $i=0,1,\cdots ,m$, then 

650: $N_{ij}^k:=\sum_{l=0}^m\frac{S_{il}S_{jl}\overline{S_{lk}}}{S_{0l}}\ (i,j,k=0,1,\cdots ,m)$ coincide with the structure constants of the fusion algebra with respect to $\{ v_i\} _{i=0}^m$. 

651: \par 

652: Furthermore, if $\{ v_i\} _{i=0}^m$ satisfies the condition (iii) in the definition of Verlinde basis, then the converse is true. Therefore, the following are equivalent. 

653: \begin{enumerate}

654: \item[(i)] $J(H;i,j)=S_{ij}$ for all $i,j=0,1,\cdots ,m$. 

655: \item[(ii)] $Z(U)v_i=v_{i}^{\ast }$ for all $i=0,1,\cdots ,m$.

656: \item[(iii)] $N_{ij}^k:=\sum_{l=0}^m\frac{S_{il}S_{jl}\overline{S_{lk}}}{S_{0l}}\ (i,j,k=0,1,\cdots ,m)$ coincide with the structure constants of the fusion algebra with respect to $\{ v_i\} _{i=0}^m$. 

657: \end{enumerate} 

658: Here, $H$ is the Hopf link presented by the diagram as in Figure \ref{Figure2}. 

659: \end{thm}

660: 

661: 

662: \section{Turaev-Viro-Ocneanu $(2+1)$-dimensional TQFT (Review)}

663: 

664: \subsection{Sectors and finite system $\Delta$}

665: 

666: For a detailed exposition about sectors, see \cite{Iz}. 

667: Let $N \subset M$ be an inclusion of infinite factors. In this case, 

668: we also have a similar concept to Jones index named Kosaki 

669: index, and if it takes the minimal value, we write it $[M:N]_0$ as in 

670: the case of Jones index. In the sequel, we assmue subfactors have  

671: finite minimal indices. For the inclusion of $\rho(M) 

672: \subset M$, where $\rho \in {\rm End}(M)$, we call $[M:\rho(M)]_0^{1/2}$ 

673: the statistical dimension of $\rho$ and denote it by $d(\rho)$. 

674: We denote the set of $*$-homomorphisms from $N$ to $M$ with finite statistical 

675: dimensions by ${\rm Mor}(N, M)_0$.  

676: We say that $\rho_1$, $\rho_2 \in {\rm Mor}(N, M)_0$ 

677: are equivalent if there exisits a unitary $u$ in $M$ such that $u\rho_1(x) =

678: \rho_2(x) u$ for all $x \in N$.  This gives an equivalence relation in 

679: ${\rm Mor}(N, M)_0$. The set of the equivalencce classes is denoted by 

680: ${\rm Sect}(N,M)$ and its element is called an $M$-$N$ {\it sector}, which 

681: is denoted by $[\rho]$ for $\rho \in {\rm Mor}(N, M)_0$. In $M$-$M$ sectors 

682: ${\rm Sect}(M,M)=:{\rm Sect}(M)$, we have the product 

683: $[\rho_1] \cdot [\rho_2] = [\rho_1 \circ \rho_2]$ and the summation $[\rho_1] 

684: \oplus [\rho_2]$. (See \cite{Iz} for the definition of the summation of 

685: sectors.) These operations  define a semiring structure in 

686: ${\rm Sect}(M)$. For $[\rho_1], [\rho_2] \in {\rm Sect}(M)$, we define 

687: \[

688:  (\rho_1, \rho_2)=\{ V \in M| V\rho_1(x) = \rho_2(x) V, \forall x \in M  \}.

689: \]

690: It is called the {\it intertwiner space} between $\rho_1$ and $\rho_2$. 

691: If $(\rho, \rho) = \mathbb{C}1_M$, we say that $\rho$ is {\it irreducible}. 

692: The intertwiner space $(\rho_1, \rho_2)$ has an inner product $\langle V, W 

693: \rangle=W^* \cdot V$, $V, W \in (\rho_1, \rho_2)$, if $\rho_1$ is 

694: irreducible.  

695: 

696: Moreover, ${\rm Sect}(M)$ has a conjugation $\overline{[\rho]}=

697: [\bar{\rho}]$. Namely, for $[\rho] \in {\rm Sect}(M)$, there exist an 

698: endomorphism $\bar{\rho} \in {\rm End}(M)$ and a pair of intertwiners 

699: $R_\rho \in (id, \bar{\rho} \rho)$ and $\bar{R}_\rho \in 

700: (id, \rho \bar{\rho})$ such that $\bar{R}_\rho^* \rho(R_\rho)=

701: R_\rho^* \bar{\rho}(\bar{R}_\rho)=1/d(\rho)$. With these operations, 

702: ${\rm Sect}(M)$ becomes a $*$-semiring over $\mathbb{C}$. 

703: 

704: An important thing is that ${\rm Sect}(M)$ is closed under the operations such 

705: as product, direct sum, irreducible decommposition and conjugation. 

706: 

707: Let us introduce the notion of a finite system $\Delta$ of ${\rm End}(M)_0$, 

708: which is a basic data to describe a topological quantum field theory from 

709: subfactors. Since the embedding $\iota : N \hookrightarrow M$ is an element of 

710: ${\rm Mor}(N, M)_0$, we can consider the sector $[\iota] \in {\rm Sect}(N, M)$. 

711: We note that the conjugation $\overline{[\iota]}$ is an element of 

712: ${\rm Sect}(M, N)$, and  the product $[\iota] \overline{[\iota]}$ becomes an 

713: element of ${\rm Sect}(M)$. 

714: In a similar way, $\overline{[\iota]} [\iota]$ becomes an element of 

715: ${\rm Sect}(N)$. By decomposing  $([\iota] \overline{[\iota]})^n$, $([\iota] 

716: \overline{[\iota]})^n [\iota] $, $\overline{[\iota]} ([\iota] 

717: \overline{[\iota]})^n $ and $(\overline{[\iota]} [\iota])^n$ into irreducible 

718: sectors, we get $M$-$M$, $M$-$N$, $N$-$M$ and $N$-$N$ sectors 

719: responsibly. (A sector $[\rho]$ is said to be {\it irreducible} if 

720: $\rho$ is irreducible.) If the number of the irreducible sectors in the above 

721: decompositions is finite, then the subfactor $N \subset M$ is called of 

722: {\it finite depth}. Throughout this paper, we only consider subfactors with 

723: finite depth and finite index. For a finite depth subfactor, we 

724: get finitely many irrducible $M$-$M$ sectors. In other words, we have a 

725: representative set $\Delta=\{ \rho_\xi \}_{\xi \in \Delta_0}$ of finitely many 

726: irreducible $M$-$M$ sectors such that \\

727: (i) $[\rho_\xi]=[\rho_\eta]$ if and only if $\xi=\eta$ \\

728: (ii) There exists $e \in \Delta_0$ such that $\rho_e =id$ \\

729: (iii) For any $\xi \in \Delta_0$, there exists $\bar{\xi} \in \Delta_0$ 

730: such that $\overline{[\rho_\xi]}=[\rho_{\bar{\xi}}]$ \\

731: (iv) There exist non-negative intetgers $N_{\xi,\eta}^\zeta$ such that 

732: $[\rho_\xi][\rho_\eta]=\oplus_{\zeta \in \Delta_0} N_{\xi,\eta}^\zeta 

733: [\rho_\zeta]$. \\

734: 

735: \noindent

736: We call this $\Delta$ a finite system of ${\rm End}(M)_0$ or simply a 

737: {\it finite system}. 

738: 

739: \begin{rem}

740: We note that $\Delta$ can be seen as a $C^*$-tensor category in the following 

741: manner. The objects of the category are $\mathbb{C}$-linear span of elements of 

742: $\Delta$,  the morphisms of the category are intertwiners, and the 

743: tensor product structure is given by the compositions of endomorphisms 

744: in objects of the category. By an abuse of notation, we denote this 

745: category by $\Delta$.  See \cite{LR} for the details. 

746: \end{rem}

747: \subsection{Turaev-Viro-Ocneanu TQFT}

748: 

749: We need some preparations before constructing the Turaev-Viro-Ocneanu 

750: TQFT. In the sequel, we write $\xi$ instead of $\rho_\xi$ and so forth, 

751: for simplicity. 

752: 

753: Let $\Delta$ be the finite system of ${\rm End}(M)_0$. We consider the following 

754: diagram.

755: \[

756:  \begin{CD}

757: \xi \cdot  \alpha \cdot \eta 

758: @< \xi(T_1) << \xi \cdot \beta \\

759: @A{T_3}AA  @AA{T_2}A \\

760:  \gamma \cdot \eta 

761: @< T_4 << \delta

762:  \end{CD}

763: \]

764: Here, $\alpha, \beta, \gamma, \xi, \eta, \delta 

765: \in \Delta$ and $T_1 \in (\beta, \alpha \cdot \eta)$, 

766: $T_2 \in (\delta, \xi \cdot \beta)$, $T_3 \in (\gamma, 

767: \xi \cdot \alpha)$, $T_4 \in (\delta, \gamma \cdot \eta)$. 

768: Then, the composition of intertwiners $T_4^* \cdot T_3^* \cdot \xi(T_1) 

769: \cdot T_2$ belongs to $(\delta, \delta)$. Since $\delta$ is assumed 

770: to be irreducible, this composition of intertwiners is regarded as  

771: a complex number. We call this number a {\it quantum $6j$-symbol}.

772: 

773: The above diagram can be seen as a tetrahedron as in Figure \ref{6j}. 

774: \begin{figure}[htbp]

775:   \begin{center}

776: \scalebox{0.5}{

777: \includegraphics{6j.eps}}  

778:  \end{center}

779:  \caption{A tetrahedron as a diagram}

780:  \label{6j}

781: \end{figure}

782: 

783: We assign $d(\beta)^{-1/2}d(\gamma)^{-1/2} T_4^* \cdot T_3^* \cdot \xi(T_1) 

784: \cdot T_2$ to this tetrahedron. 

785: 

786: Let $V$ be an oriented closed 3-dimensional manifold. Choose a triangulation 

787: of $V$ and write it ${\cal T}$. To each vertex in ${\cal T}$, we assign  

788: the factor $M$, to each edge in ${\cal T}$ an element in $\Delta$ and to each face in 

789: ${\cal T}$ an intertwiner. 

790: Let $E$ be the set of the edges in ${\cal T}$, $e$ be an assignment of 

791: elements in $\Delta$ to the edges in ${\cal T}$, and $\varphi$ be an 

792: assignment of intertwiners to the faces in ${\cal T}$. 

793: A tetrahedron $\tau$ has the labeled edges by the assignment $e$ and 

794: the labeled faces by the assignment $\varphi$. Hence, to a tetrahedron 

795: $\tau$, we assign a complex number defined by using the $6j$-symbol as above 

796: or its complex conjugate depending on the orientation of $\tau$. 

797: We denote this complex number by $W(\tau;e,\varphi)$.  We multiply these 

798: $W$'s and the weights $\prod_E d(\xi)$. Then, sum up all these resulting 

799: values. Finally, we multiply some weights coming from the vertices 

800: in ${\cal T}$. 

801: \[

802:  Z_\Delta(V,{\cal T})= \lambda^{-a} \sum_e(\prod_E d(\xi))

803: \sum_\varphi \prod_\tau W(\tau;e,\varphi), 

804: \]

805: where $\lambda=\sum_{\xi \in \Delta_0} d(\xi)^2$ and $a$ is 

806: the number of the vertices in ${\cal T}$. This $Z_\Delta(V,{\cal T})$ is 

807: proven to be independent of any choice of triangulations because of 

808: the properties 

809: of quantum $6j$-symbols. (See \cite{EK} for a detailed account.) Namely, 

810: $Z_\Delta(V,{\cal T})$ turns out to be a topological invariant of $V$. So, we 

811: drop ${\cal T}$ off from $Z_\Delta(V,{\cal T})$ and denote this value 

812: by $Z_\Delta(V)$. We call $Z_\Delta(V)$ the {\it Turaev-Viro-Ocneanu invariant} 

813: of the 3-dimensional manifold $V$. 

814: When $V$ is an oriented, compact 3-dimensional manifold possibly with 

815: boundary, first we fix a triangulation  $\partial {\cal T}$ of the 

816: boundary of $V$ and extend it to the whole triangulation of $V$. Then, 

817: we assign an element in $\Delta$ to each edge in $\partial {\cal T}$ and 

818: assign an intrertwiner to each face. We fix these assignments to the end 

819: and denote them by $\partial e$ and $\partial \varphi$, respectively. In a similar 

820: way to the closed case, we assign $M$, an element in $\Delta$ and an 

821: intertwiner to each vertex, each edge and each face in the triangulation 

822: ${\cal T} \setminus \partial {\cal T}$. We make 

823: $Z_\Delta(V,{\cal T}, \partial e, \partial \varphi)$ as above:

824: \[

825:  Z_\Delta(V,{\cal T}, \partial e, \partial \varphi)= 

826:  \lambda^{-a+\partial a/2} \prod_{\partial e} 

827:  d(\xi)^{1/2} \sum_{e \setminus \partial e}(\prod_{E \setminus \partial E} 

828: d(\xi)) \sum_\varphi  \prod_\tau W(\tau;e,\varphi), 

829: \]

830: where $\partial a$ is the number of the vertices on the boundary. This 

831: value $Z_\Delta(V,{\cal T}, \partial e, \partial \varphi)$ does not 

832: depend on the assignments of the vertices, the edges and the faces in 

833: ${\cal T} \setminus \partial {\cal T}$. Such extended 

834: $Z_\Delta$ gives rise to a unitary TQFT because to an oriented closed surface, 

835: it assigns a finite dimensional Hilbert space with the inner product 

836: induced from the space of intertwiners. (See \cite{EK} for the detailed 

837: construction.) We call this TQFT the {\it $(2+1)$-dimensional 

838: Turaev-Viro-Ocneanu TQFT} and denote it by $Z_\Delta$, again. 

839: 

840: 

841: \section{Verlinde basis of $Z_{\Delta}(S^1 \times S^1)$}

842: 

843: Let $N \subset M$ be a subfactor of an infinite factor $M$ with finite 

844: index and finite depth, and let $\Delta$ be a finite system of irreducible 

845: $M$-$M$ endomorphisms arising from the subfactor. We write, for instance, 

846: $\xi$ instead of $\rho_\xi$ and so forth for the elements of $\Delta$. 

847: 

848: Based on $\Delta$, we construct a new finite dimensional $C^*$-algebra named 

849: the {\it tube algebra} ${\rm Tube} \Delta$ as in \cite{Ocneanu}. (Also see 

850: \cite{Izumi1}, but our normalization convention is different from that 

851: there.)  In this section, ${\rm Tube} \Delta$ plays a crutial role to find a 

852: nicely behaved basis of $Z_{\Delta}(S^1 \times S^1)$, which we call a Verlinde 

853: basis. It makes the Turaev-Viro-Ocneanu TQFT a rich theory. 

854: 

855: We sometimes use the simple notation $Z(S^1 \times S^1)$ instead of 

856: $Z_{\Delta}(S^1 \times S^1)$ in the sequel.

857: 

858: \subsection{Tube algebras}

859: 

860: A tube algebra ${\rm Tube}\Delta$, which was first introduced by Ocneanu 

861: in \cite{Ocneanu}, is defined by $\bigoplus_{\xi,\eta,\zeta} (\xi \cdot 

862: \zeta, \zeta \cdot \eta)$ as a vector space over $\mathbb{C}$. Its element 

863: is a linear combination of the composition  $T_2 \cdot T_1^*$ of the orthonormal 

864: bases of intertwiner spaces $T_1 \in (\delta, \xi \cdot \zeta)$, $T_2 \in (\delta, 

865: \zeta \cdot \eta)$, where $\xi, \eta, \zeta$ and $\delta$ run over $\Delta_0$. 

866: An element in ${\rm Tube}\Delta$ can be depicted as in the left-hand side of 

867: Figure \ref{tube}. We will define a product structure and a $*$-structure on it. 

868: 

869: 

870: \noindent

871: \underline{The product structure} \\

872: Let $X=X_2 X_1^* \in (\xi \cdot \zeta, \zeta \cdot \eta)$ for $X_1 \in 

873: (\delta, \xi \cdot \zeta)$, $X_2 \in (\delta, \zeta \cdot \eta)$ and 

874: $Y=Y_2 Y_1^* \in (\xi' \cdot \zeta', \zeta' \cdot \eta')$ for $Y_1 \in 

875: (\delta', \xi' \cdot \zeta')$, $Y_2 \in (\delta', \zeta' \cdot \eta')$. 

876: Then, the product of $X$ and $Y$ in ${\rm Tube} \Delta$ is defined by 

877: the following formula. 

878: \[

879:  X \cdot Y = \delta_{\eta, \xi'} \lambda \sum

880: d(\xi)^{-1/2} d(\eta')^{-1/2}  d(\eta) \lambda(X,Y;Z)Z.

881: \] 

882: Here, the summation is taken over $Z=Z_2 Z_1^*$, $Z_1 \in (\tau, \xi \cdot \nu)$, 

883: $Z_2 \in (\tau, \nu \cdot \eta')$, where $Z_1$ and $Z_2$ are orthonormal bases of 

884: the intertwiner spaces $(\tau, \xi \cdot \nu)$ and $(\tau, \nu \cdot 

885: \eta')$ respectively, and $\lambda(X,Y;Z)$ is the value of $Z_\Delta$ of 

886: the 3-manifold depicted in Figure \ref{tube-c}. 

887: For general elements $X$, $Y$ in ${\rm Tube} \; \Delta$, we 

888: define the product of them by linearlity since $X$ and $Y$ are linear 

889: combinations of the forms $X_2 X_1^*$ and $Y_2 Y_1^*$ as above. 

890: 

891: \begin{figure}[htbp]

892:   \begin{center}

893: \scalebox{0.35}{

894: \includegraphics{tube-c.eps}}  

895:  \end{center}

896:  \caption{A coeffcient of the product of $X \cdot Y$}

897:  \label{tube-c}

898: \end{figure}

899: 

900: \noindent

901: \underline{The $*$-structure} \\

902: Let $X$ be as above. Then, we can consider $X$ as a tube as in the right-hand 

903: side of Figure \ref{tube}. 

904: The $*$-operation is defined by the inversing the tubes inside out. We denote 

905: this $*$-operation by $X^*$. See Figure \ref{tube-r}.

906: 

907: \begin{figure}[htbp]

908:   \begin{center}

909: \scalebox{0.5}{

910: \includegraphics{tube.eps}}  

911:  \end{center}

912:  \caption{An element of the tube algebra as a tube}

913:  \label{tube}

914: \end{figure}

915: 

916: \begin{figure}[htbp]

917:   \begin{center}

918: \scalebox{0.4}{

919: \includegraphics{tube-r.eps}}  

920:  \end{center}

921:  \caption{An inversed tube}

922:  \label{tube-r}

923: \end{figure}

924: 

925: With the product and the $*$-structure defined as above, ${\rm Tube} \Delta$ 

926: becomes a finite dimensional $C^*$-algebra. Since any finite dimensional 

927: $C^*$-algebra is semisimple, we may assume that 

928: ${\rm Tube} \Delta \cong \bigoplus_{i=0}^r M_{n_i}({\mathbb C})$. 

929: 

930: We observe that the definition of tube algebras is compatible 

931: with the operations of Turaev-Viro-Ocneanu TQFT such as gluing, 

932: cutting and so forth.

933: 

934: \begin{rem}

935:  In \cite{Izumi1}, M. Izumi has introduced the tube algebra in the 

936:  setting of sectors, but it is slightly different from ours in the 

937: normalization coefficients. We followed the definition of the tube 

938: algebra in \cite[Chapter 12]{EK}. 

939: \end{rem}

940: 

941: Before we start the analysis of tube algebras, we list the notations that 

942: we will use frequently. 

943: 

944: For the solid torus $D^2 \times S^1$, we denote 

945: the value of $Z_\Delta$ of $D^2 \times S^1$ assigned a vector $\lambda$ 

946: on the boundary by $Z_\Delta(D^2 \times S^1;\lambda)$. 

947: 

948: For the 3-manifold $D^2 \times S^1 \setminus {\rm Int}D_0^2 \times S^1$, 

949: where $D_0$ is contained in ${\rm Int} D^2$, we denote the value of $Z_\Delta$ of 

950: $D^2 \times S^1 \setminus {\rm Int}D_0^2 \times S^1$ assigned vectors $\lambda$ and 

951: $\mu$, $\lambda$ for the boundary of $D_0^2 \times S^1$ and $\mu$ for the boundary of 

952: $D^2 \times S^1$ by $Z_\Delta(D^2 \times S^1 \setminus {\rm Int}D_0^2 \times S^1; 

953: \lambda,\mu)$. See Figure \ref{solid-a}. 

954: 

955: \begin{figure}[htbp]

956:   \begin{center}

957: \scalebox{0.4}{

958: \includegraphics{solid1.eps}}  

959:  \end{center}

960:  \caption{The labeled 3-manifold obtained by removing another solid torus from a solid torus}

961:  \label{solid-a}

962: \end{figure}

963: 

964: In a similar manner, for the 3-manifold $D^2 \times S^1 \setminus ( {\rm Int} D_1^2 

965: \times S^1 \cup {\rm Int} D_2^2 \times S^1)$, where $D_1^2$ and $D_2^2$ are two disjoint 

966: disks contained in ${\rm Int} D^2$, we denote the value of $Z_\Delta$ of $D^2 \times S^1 

967: \setminus ({\rm Int} D_1^2 \times S^1 \cup {\rm Int} D_2^2 \times S^1)$ 

968: assigned vectors $\lambda$, $\mu$ and $\nu$, $\lambda$ for the boundary of 

969: $D_1^2 \times S^1$, 

970: $\mu$ for the boundary of $D_2^2 \times S^1$ and $\nu$ for the the boundary of 

971: $D^2 \times S^1$, by $Z_\Delta(D^2 \times S^1 \setminus ({\rm Int}D_1^2 \times 

972: S^1 \cup {\rm Int}D_2^2 \times S^1); \lambda,\mu;\nu)$. 

973: See Figure \ref{solid-b}. 

974: 

975: \begin{figure}[htbp]

976:   \begin{center}

977: \scalebox{0.4}{

978: \includegraphics{solid2.eps}}  

979:  \end{center}

980:  \caption{The labeled 3-manifold obtained by removing two solid tori 

981:  from a solid torus}

982:  \label{solid-b}

983: \end{figure}

984: \bigskip

985: 

986: Let $\{ \pi_0, \cdots, \pi_r \}$ be the minimal central projections of 

987: ${\rm Tube} \Delta$. Then, we have the following lemma. 

988: \begin{lem}\label{innerproduct}

989: $\langle \pi_i, \pi_j \rangle_{Z(S^1 \times S^1)} = \delta_{i,j} n_i^2$ 

990: ($i,j=0, \dots, r$), where $\langle \cdot, \cdot \rangle_{Z(S^1 \times S^1)}$ is 

991: the inner product of $Z_\Delta(S^1 \times S^1)$ defined by the Turaev-Viro-Ocneanu 

992: TQFT. 

993: \end{lem}

994: 

995: \noindent

996: Proof. First, we consider the case $i=j$. We note that $\langle \pi_i, \pi_i 

997: \rangle_{Z(S^1 \times S^1)}=Z_\Delta(D^2 \times S^1 \setminus {\rm Int}D_0^2 

998: \times S^1;\pi_i, \pi_i)$. 

999: 

1000: Cut $D^2 \times S^1 \setminus {\rm Int}D_0^2 \times S^1$ along the meridian, 

1001: then $T ={\rm annulus} \times [0,1]$ is created. Let $A_{jk}$ be the value of 

1002: $Z_\Delta(T, \xi_j, \xi_k$), where $T$ is labeled by $\xi_j$ on one side of 

1003: the sections and by $\xi_k$ on the other side. Then, $Z_\Delta(D^2 \times S^1 

1004: \setminus {\rm Int}D_0^2 \times S^1;\pi_i, \pi_i)=\sum_{\xi_j} A_{jj}$. 

1005: There exists an operator $A$ such that $A_{jk}=\langle A 

1006: \xi_j, \xi_k \rangle$, where $\langle \cdot, \cdot \rangle$ is the inner 

1007: product on intertwiners on the sections of $T$ so that $\langle \xi_j, \xi_k 

1008: \rangle=\delta_{jk}$. 

1009: 

1010: It is easy to see that the operator $A$ is a projection. Hence, 

1011: $\langle \pi_i, \pi_i \rangle_{Z(S^1 \times S^1)}=\sum_{j=1}^m A_{jj} =

1012: {\rm Tr}(A) \in {\mathbb N}$. Namely, we have proved that 

1013: $\langle \pi_i, \pi_i \rangle_{Z(S^1 \times S^1)}={\rm dim}\; \pi_i({\rm Tube} 

1014: \Delta) = n_i^2$. 

1015: 

1016: When $i \ne j$, it is easy to see that $\langle \pi_i, \pi_j 

1017: \rangle_{Z(S^1 \times S^1)}=0$, because $\pi_i$ and $\pi_j$ are central 

1018: projections orthogonal to each other. This ends the proof. \hfill Q.E.D.

1019: 

1020: \begin{thm}\label{centerTube}

1021:  Let $\Delta$ be a finite system. Then, 

1022:  the center of ${\rm Tube}\Delta$ is naturally isomorphic to 

1023:  $Z_\Delta(S^1 \times S^1)$ as a vector space.

1024: \end{thm}

1025: 

1026: \noindent

1027: Proof.  Let $V(S^1 \times S^1)$ be the linear span of elements in ${\rm Tube}\; \Delta$ such 

1028: that the left and right labels in Figure \ref{tube} are equal. Since ${\rm Tube}\Delta 

1029: \supset V(S^1 \times S^1)$ and $V(S^1 \times S^1)$ contains the center of 

1030: ${\rm Tube}\Delta$, it is enough to consider $V(S^1 \times S^1)$ instead of whole 

1031: tube algebra. 

1032: 

1033: For $X \in V(S^1 \times S^1) $, we set  $\varphi_i(X)=Z_\Delta(D^2 \times S^1 \setminus 

1034: {\rm Int}D_0^2 \times S^1; \pi_i, X)$. If $X \perp V(S^1 \times S^1)$ 

1035: with respect to the colored inner product of ${\rm Tube} \Delta$, we set 

1036: $\varphi_i(X)=0$. Here, for $X=X_2 X_1^* \in (\xi \cdot \zeta, \zeta \cdot \eta)$, $X_1 \in 

1037: (\delta, \xi \cdot \zeta)$, $X_2 \in (\delta, \zeta \cdot \eta)$ and $Y=Y_2 Y_1^* \in 

1038: (\xi' \cdot \zeta', \zeta' \cdot \eta')$, $Y_1 \in (\delta', \xi' \cdot \zeta')$, 

1039: $Y_2 \in (\delta', \zeta' \cdot \eta')$,  the colored inner product 

1040: $\langle X, Y \rangle_{\rm color}$ of ${\rm Tube} \Delta$ is defined by 

1041: $\delta_{\xi,\xi'}\delta_{\eta,\eta'}\delta_{\zeta,\zeta'} \langle X_1, Y_1 \rangle

1042: \langle X_2, Y_2 \rangle$. The last two brackets stand for the inner 

1043: products of the intertwiner spaces $(\delta, \xi \cdot \zeta)$ and 

1044: $(\delta, \zeta \cdot \eta)$, respectively. Then, this $\varphi_i$ is a linear functional 

1045: defined on ${\rm Tube}\Delta$. It is easy to see that $\varphi_i$ is tracial. 

1046: 

1047: If $X$ is an element in $\pi_i({\rm Tube}\Delta)$, then we have

1048: \[

1049:  \varphi_i(X)=\langle \pi_i, \pi_i \rangle_{Z(S^1 \times S^1)} {\rm  tr}(X), 

1050: \]

1051: where ${\rm tr}$ is the normalized trace. In a similar way, for $X,Y \in 

1052: \pi_i({\rm Tube}\Delta)$, the value 

1053: $Z_\Delta(D^2 \times S^1 \setminus {\rm Int}D_0^2 \times S^1; Y, X)$ is 

1054: equal to 

1055: $\langle \pi_i, \pi_i \rangle_{Z(S^1 \times S^1)} {\rm tr}(X){\rm tr}(Y^*)$. 

1056: 

1057: When we write $X_j=\pi_jX$ for $X \in V(S^1 \times S^1)$, $X=\oplus_{j=0}^r X_j$.

1058: For $X \in {\rm Tube}\Delta$,  we put 

1059: \[

1060:  E(X)= \sum_{i=0}^r \frac{\varphi_i(X)}{\langle \pi_i, \pi_i 

1061:  \rangle_{Z(S^1 \times S^1)} } \pi_i=

1062:  \sum_{i=0}^r {\rm tr}(X_i) \pi_i. 

1063: \]

1064: Then, this is a conditional expectation from ${\rm Tube}\Delta$ to the 

1065: center of ${\rm Tube}\Delta$.  We have a description of the kernel of $E$ by 

1066: ${\rm Ker}E = {\rm the \; linear \; span \; of \; }

1067: \{ X \in V(S^1 \times S^1)|{\rm tr}(X_j)=0, j=0, \dots, r 

1068: \} \cup \{ X \in {\rm Tube}\Delta | \langle X, Y \rangle_{\rm color}=0 \; {\rm for \  

1069: all}\; Y \in  V(S^1 \times S^1) \}$. 

1070: 

1071: We have 

1072: \begin{eqnarray*}

1073:  \langle X_j, X_j \rangle_{Z(S^1 \times S^1)}

1074:  &=&\langle \pi_j, \pi_j 

1075:  \rangle_{Z(S^1 \times S^1)} {\rm tr}(X_j){\rm tr}(X_j^*) \\

1076:  &=& \langle \pi_j, \pi_j \rangle_{Z(S^1 \times S^1)} |{\rm tr}(X_j)|^2, 

1077: \end{eqnarray*}

1078: and we put $Q=\{  X \in V(S^1 \times S^1)|\langle X, X \rangle_{Z(S^1 \times 

1079: S^1)}=0 \}$.  

1080: Then, 

1081: $Z(S^1 \times S^1)=V(S^1 \times S^1)/Q = {\rm Tube}\Delta /{\rm Ker}E

1082: \cong {\rm Center}({\rm Tube}\Delta)$. 

1083: Precisely, denoting the embedding map from $Z_\Delta(S^1 \times S^1)$ into 

1084: Tube $\Delta$ by $\iota$, we have proved that Center(Tube $\Delta$) 

1085: $=\iota(Z_\Delta(S^1 \times S^1)) \subset {\rm Tube} \; \Delta$. 

1086: \hfill Q.E.D.

1087: 

1088: \begin{rem}

1089:  From this theorem and Lemma \ref{innerproduct}, 

1090:  $\{ \frac{\pi_i}{n_i} \}_{i=0}^r$ is an orthonormal basis of 

1091:  $Z(S^1 \times S^1)$. 

1092: \end{rem}

1093: 

1094: 

1095: \subsection{Verlinde basis}

1096: 

1097: In this subsection, we show the existence of a basis of 

1098: $Z_\Delta(S^1 \times S^1)$ nicely behaved under the action of 

1099: $SL_2(\mathbb{Z})$, which we call the {\it Verlinde basis}. It 

1100: deeply depends on the structure of the tube algebra. 

1101: 

1102: Before the proof of the existence of such basis, we need some preparations.

1103: 

1104: Let $p_i$ be a minimal projection in $\pi_i ({\rm Tube} \Delta)$ for 

1105: each $i=0,\dots,r$. By Lemma \ref{innerproduct}, we have 

1106: $\langle p_i, p_j \rangle_{Z(S^1 \times S^1)} = \delta_{ij}$. 

1107: (We use the same notation $p_i$ as an element of $Z_\Delta(S^1 \times S^1)$.) 

1108: Then, by the last remark in the previous subsection, we have 

1109: \begin{equation}\label{p_i-pi_i}

1110:  p_i=\frac{\pi_i}{n_i}\ (i=0,\dots, r)

1111: \end{equation}

1112: in $Z_\Delta(S^1 \times S^1)$.

1113: 

1114: Let us compute the value $Z_\Delta(D^2 \times S^1; p_i)$. For this, 

1115: we look at $Z_\Delta(D^2 \times S^1; \pi_i)$. It is a summation of 

1116: $Z_\Delta(D^2 \times [0,1], \xi_k,\xi_k)$ over the intertwiners $\xi_k$'s 

1117: since $D^2 \times [0,1]$ is created by cutting the solid torus along the 

1118: meridian, This is nothing but the summation of the dimension 

1119: of the partially labeled disks labeled by $p_{i_j}$ over $j$, $j=1,\cdots, 

1120: n_i$, where $p_{i_j}$'s are minimal projections such that $\sum_j p_{i_j}=

1121: \pi_i$. (See \cite[Chapter 12]{EK} for the definition of the partially labeled 

1122: surfaces.) See Figure \ref{label-pi}. 

1123: 

1124: \begin{figure}[htbp]

1125:   \begin{center}

1126: \scalebox{0.4}{

1127: \includegraphics{label-pi.eps}}  

1128:  \end{center}

1129:  \caption{The decomposition of a labeled disk}

1130:  \label{label-pi}

1131: \end{figure}

1132: 

1133: When we denote the dimension of the partially labeled disk with label 

1134: $p_{i_j}$ by $c_i$, the dimension of the partially labeled disk with label 

1135: $\pi_i$ becomes $n_i c_i$. 

1136: 

1137: Take a summation of $n_i c_i$ over all $i$'s, then it is equal to the 

1138: dimension of the partially labeled disk with label $\sum_i \pi_i=1$. Hence, 

1139: it is equal to the dimension of the triangulated disk in Figure 

1140: \ref{solid-d}, which edge $AB$ is glued together. 

1141: (The boudary element is a direct sum taken over arbitray $\rho$.)

1142: 

1143: \begin{figure}[htbp]

1144:   \begin{center}

1145: \scalebox{0.4}{

1146: \includegraphics{solid4.eps}}  

1147:  \end{center}

1148:  \caption{A triangulated disk}

1149:  \label{solid-d}

1150: \end{figure}

1151: 

1152: Then, this vector space is non-trivial only in the case of $\rho=id$ and then, 

1153: the dimension is one. It means that  $\sum_i n_i c_i =1$. Hence, 

1154: only one summand can survive. We may and do assume $n_0 c_0=1$. (Hence, 

1155: $n_0=1$, $c_0=1$.) In a similar manner, we denote $\pi_i$ by $\pi_0$ in this 

1156: case. 

1157: 

1158: Let us summerize the above argument as a lemma. 

1159: 

1160: \begin{lem}\label{value-solid}

1161: $Z_\Delta(D^2 \times S^1 \setminus {\rm Int} D_0^2 \times S^1;\pi_i)

1162: =Z_\Delta(D^2 \times S^1 \setminus {\rm Int} D_0^2 \times S^1;p_i)

1163: =\delta_{i0}$.

1164: \end{lem}

1165: 

1166: Let $N_{ij}^k$ be $Z_\Delta (D^2 \times S^1 \setminus ( {\rm Int}D_1^2 \times 

1167: S^1 \cup {\rm Int}D_2^2 \times S^1);p_i,p_j;p_k )$. See Figure \ref{solid-b}.  

1168: Cut $D^2 \times S^1 \setminus ( {\rm Int}D_1^2 \times S^1 \cup 

1169: {\rm Int}D_2^2 \times S^1)$ along the meridian, then we have $P=(D^2 \setminus 

1170: {\rm Int} D_1^2 \cup {\rm Int} D_2^2) \times [0,1]$, and as a value, $N_{ij}^k$ is 

1171: a summation of $Z_\Delta(P, \xi_k, \xi_k)$ over $\xi_k$'s, where $P$ is labeled 

1172: by $\xi_k$'s on the sections. This value can be written 

1173: $\langle A \xi_k, \xi_k \rangle$, where $\langle \cdot, \cdot \rangle$ is the inner 

1174: product of the Hilbert space of the section. Then, it is easy to see that $A$ is a 

1175: projection. Hence, $N_{ij}^k={\rm Tr}(A)  = {\rm dim } \; H_{\rm pants} \in {\mathbb N}$, 

1176: where $H_{\rm pants}$ is the Hilbert space associated with the 3-holed sphere in the 

1177: Turaev-Viro-Ocneanu TQFT. See Figure \ref{Nijk}. 

1178: 

1179: \begin{figure}[htbp]

1180:   \begin{center}

1181: \scalebox{0.4}{

1182: \includegraphics{Nijk.eps}}  

1183:  \end{center}

1184:  \caption{$H_{\rm pants}$}

1185:  \label{Nijk}

1186: \end{figure}

1187: 

1188: With these settings, we can prove the following theorem, which is one of 

1189: our main theorems in this paper. (Many of the contents have already appeared 

1190: in \cite{EK}, \cite{EK1} following several presentations of Ocneanu, but 

1191: some normalizations are missing or incorrect there, so we include a 

1192: complete proof here.)

1193: \begin{thm}\label{verlinde-basis}

1194:  Let $\Delta$ be a finite system and $\{ p_i \}_{i=0}^r$ minimal 

1195:  projections such that each $p_i$ belongs to $\pi_i ({\rm   Tube}(\Delta))$,  

1196:  $i=0,\dots, r$. Then, 

1197:  $\{ p_i \}_{i=0}^r$ is a Verlinde basis of $Z_\Delta(S^1 \times S^1)$ 

1198:  in the sense of Section 2. 

1199: \end{thm}

1200: 

1201: \noindent

1202: Proof. We check all the conditions for $\{ p_0, \dots, p_r \}$ to be a 

1203: Verlinde basis step by step. 

1204: 

1205: Condition (ii.c.1): 

1206: Look at the inner product $n_j^2=\langle \pi_j, \pi_j 

1207: \rangle_{Z(S^1 \times S^1)}$. This value is defined by 

1208: $Z_\Delta(D^2 \times S^1 \setminus {\rm Int}D_0^2 \times S^1;\pi_j,\pi_j)$. 

1209: Cut $D^2 \times S^1 \setminus {\rm Int}D_0^2 \times S^1$ along the meridian, 

1210: then we have $X_0={\rm annulus} \times [0,1]$ as topological object. 

1211: Since $\pi_j$ is central, $X_0$ can be depicted as in Figure \ref{solid-e}, 

1212: where $\xi_k$'s are intertwiners on the sections. 

1213: 

1214: \begin{figure}[htbp]

1215:   \begin{center}

1216: \scalebox{0.8}{

1217: \includegraphics{solid5.eps}}  

1218:  \end{center}

1219:  \caption{$X_0$ after changes}

1220:  \label{solid-e}

1221: \end{figure}

1222: 

1223: Let $X$ be the solid torus that we had at last in Figure \ref{solid-e}. 

1224: Then, we have 

1225: \begin{equation}

1226:  \langle \pi_j, \pi_j \rangle_{Z(S^1 \times S^1)}=\sum_{\xi_k} \alpha(\xi_k)

1227:  Z_\Delta (X),

1228: \end{equation}

1229: where $\alpha(\xi_k)'s$ are positive coefficients determined by 

1230: some products of statistical dimensions. 

1231: It is obvious that each $Z_\Delta (X)$ is non-negative. 

1232: 

1233: Since $S(p_j)=\sum_i S_{ij} p_i$, it follows from Lemma \ref{value-solid} that 

1234: $Z_\Delta (D^2\times S^1;S(p_j))= S_{0j}$. 

1235: 

1236: It is now easy to see  that the value of the solid torus labeled by 

1237: $S(p_j)$ has a similar expression of the summation when it is cut along the 

1238: meridian. Namely, $S_{0j}$ is written in the form $S_{0j}=\sum_{\xi_k} 

1239: \beta(\xi_k) Z_\Delta (X)$, where $\beta(\xi_k)$'s are  strictly positive 

1240: coefficients detemined by $n_j$ and some products of statistical dimensions. 

1241: As we saw, not all of the values of $Z_\Delta (X)$ can be zero, so 

1242: one of them must be strictly positive. Hence, $S_{0j} > 0$. 

1243: 

1244: Condition (ii.a), (iii): Since both $S$ and $T$-matrices are 

1245: unitary, we check that $S$ is symmetric and $T$ is diagonal with repsect to 

1246: $\{p_0, \dots, p_r \}$. 

1247: 

1248: First we prove that $S$ is symmetric. From a topological observation of 

1249: the $Z(S)$-action on $Z(S^1 \times S^1)$, 

1250: we have 

1251: \begin{eqnarray*}

1252: S_{ij}

1253: &=&\langle S(p_i), p_j \rangle_{Z(S^1 \times S^1)} 

1254: = \overline{\langle S(p_i)^*, p_j^* \rangle}_{Z(S^1 \times S^1)} 

1255: = \overline{\langle S(p_i)^*, p_j \rangle}_{Z(S^1 \times S^1)} \\

1256: &=& \overline{\langle S^*(p_i), p_j \rangle}_{Z(S^1 \times S^1)} 

1257: = \overline{\langle p_i, S(p_j) \rangle}_{Z(S^1 \times S^1)} 

1258: = \langle S(p_j), p_i \rangle_{Z(S^1 \times S^1)} \\

1259: &=&S_{ji}. 

1260: \end{eqnarray*}

1261: 

1262: Next, we prove that $T$ is diagonal. From a topological observation of 

1263: the $Z(T)$-action on $Z(S^1 \times S^1)$, we have 

1264: \[

1265:  \langle T(X) \cdot T^*(Y), Z \rangle_{Z(S^1 \times S^1)} = 

1266: \langle X \cdot Y, Z \rangle_{Z(S^1 \times S^1)}

1267: \] 

1268: for any $X,Y,Z \in V(S^1 \times S^1)$, where $\cdot$ stands for the multiplication 

1269: in the tube algebra. Hence, we have $T(X) \cdot T^*(Y)=X \cdot Y$ for any 

1270: $X, Y \in V(S^1 \times S^1)$. This implies that $T$ is diagonal.  

1271: 

1272: Condition (ii.c.2): (Verlinde identity) From the definition of the fusion 

1273: algebra associated with the Turaev-Viro-Ocneanu TQFT, $N_{ij}^k$ is a structure 

1274: constant of the fusion algebra. 

1275: 

1276: From the unitarity of $S$, it is enough to prove 

1277: \begin{equation}\label{verlinde-id}

1278:  \sum_{i,j} N_{ij}^k \; \overline{S_{im}}\; \overline{S_{jn}} = \delta_{mn} 

1279:  \frac{\overline{S_{mk}}}{S_{m0}}.

1280: \end{equation}

1281: The first term of the left-hand side of the equation (\ref{verlinde-id}) is 

1282: written as $Z_\Delta (D^2 \times S^1 \setminus ({\rm Int}D_1^2 

1283: \times S^1 \cup {\rm Int}D_2^2 \times S^1); p_i,p_j;p_k)$. The second and 

1284: the third terms of the left-hand side of the equation (\ref{verlinde-id}) can be 

1285: written as $Z_\Delta (D^2 \times S^1 \setminus {\rm Int }D_0^2 \times S^1;

1286: p_i, S(p_m))$, $Z_\Delta (D^2 \times S^1 \setminus {\rm Int}D_0^2 \times S^1;

1287: p_j, S(p_n))$, respectively. 

1288: Use gluing and embed the second and third terms into the first term, then 

1289: the left-hand side of the equation (\ref{verlinde-id}) is equal to 

1290: $Z_\Delta (D^2 \times S^1 \setminus ({\rm Int}D_1^2 \times S^1 \cup 

1291: {\rm Int}D_2^2 \times S^1); S(p_m),S(p_n);p_k)$. Then, cut $D^2 \times S^1 

1292: \setminus ({\rm Int}D_1^2 \times S^1 \cup {\rm Int}D_2^2 \times S^1$ along 

1293: the meridian. 

1294: 

1295: \begin{figure}[htbp]

1296:   \begin{center}

1297: \scalebox{0.45}{

1298: \includegraphics{pi_m-pi_n.eps}}  

1299:  \end{center}

1300:  \caption{}

1301:  \label{pi_m-pi_n}

1302: \end{figure}

1303: 

1304: By using the gluing axiom of TQFT, we separate the rectangular part of 

1305: this cut object by taking the summation of $Z_\Delta$ of the 3-manifold 

1306: depicted in Figure \ref{pi_m-pi_n} over orthonormal basis $\eta_l$'s 

1307: on the rectangle. Then, by the identity (\ref{p_i-pi_i}), we get the value 0 

1308: if $m \ne n$ because $\pi_m$ and $\pi_n$ are central orthogonal projections. 

1309: Hence, 

1310: \begin{eqnarray*}

1311: &&Z_\Delta (D^2 \times S^1 \setminus ({\rm Int}D_1^2 \times S^1 \cup 

1312: {\rm Int}D_2^2 \times S^1); S(p_m),S(p_n);p_k) \\

1313: &=& \delta_{mn} \frac{\sum_m Z_\Delta (D^2 \times S^1 \setminus ({\rm Int}D_1^2 

1314: \times S^1 \cup{\rm Int} D_2^2 \times S^1); S(p_m),S(p_n);p_k) \; 

1315: Z_\Delta (D^2 \times S^1;S(p_m))}{Z_\Delta (D^2 \times S^1;S(p_n))} \\

1316: &=& \delta_{mn} \frac{Z_\Delta (D^2 \times S^1 \setminus {\rm Int}D_0^2 \times 

1317: S^1;S(p_m), p_k)}{Z_\Delta (D^2 \times S^1;S(p_n))} \\

1318: &=& \delta_{mn} \frac{\overline{S_{mk}}}{S_{n0}}. 

1319: \end{eqnarray*}

1320: 

1321: Condition (i): From the Verlinde identity and the unitarity of $S$, we have 

1322: \[

1323:  N_{0j}^k=\sum_{l=0}^r \frac{S_{0l} S_{jl} \overline{S_{kl}}}{S_{0l}} 

1324:  = \sum_{l=0}^r S_{jl} \overline{S_{kl}} = \delta_{jk}.

1325: \]

1326: So $p_0$ is the identitiy element in the fusion algebra. 

1327: 

1328: Condition (ii.b): Since $S^2$ is $*$-antiisomorphism, it is clear that 

1329: $S^2(p_i)$'s are minimal projections again. 

1330: 

1331: We have $S^2(p_0)=p_0$ because the definition of $p_0$ is invariant 

1332: under the 180 degree rotation. 

1333: 

1334: Condition (iv): We note that $Z(U)$ is nothing but the $*$-operation of 

1335: ${\rm Tube}\; \Delta$. So $Z(U)(p_i)=p_i^*=p_i$. The canonical map $\theta$ maps 

1336: an orthonormal basis in $Z(-S^1 \times S^1)$ to the dual basis in $Z(S^1 

1337: \times S^1)^*$. So, $\theta(p_i)=\widehat{p_i}$, where 

1338: $\{ \widehat{p_i} \}_{i=0}^r$ is a dual basis of $\{ p_i \}_{i=0}^r$ in 

1339: $Z(S^1 \times S^1)^*$ such that $\widehat{p_i}(p_j)=\delta_{ij}$. Getting 

1340: together, we have $\theta \circ Z(U) (p_i) = \theta(p_i)=\widehat{p_i}$. 

1341: \hfill Q.E.D.

1342: 

1343: 

1344: 

1345: \section{Applications}

1346: 

1347: From the conclusion of Section 4, we know that there exists a Verlinde basis 

1348: of $Z_{\Delta}(S^1 \times S^1)$ in the sense of Section 2 in a  

1349: Turaev-Viro-Ocneanu TQFT. Hence, for a closed 3-manifold $M$, we have 

1350: the following Dehn surgery formula: 

1351: \[

1352:  Z_\Delta (M) = \sum_{i_1, \dots, i_m=0}^r S_{0i_1} \cdots S_{0i_m} 

1353:  J(L;i_1, \dots, i_m),

1354: \]

1355: where we have assumed that the manifold $M$ is obtained from $S^3$ by Dehn 

1356: surgery along a framed link $L=L_1 \cup \cdots \cup L_m$. 

1357: 

1358: The purpose in this section is to understand the formula of 

1359: right-hand side of the above equation by introducing a notion of 

1360: the {\it tube system} due to Ocneanu \cite{Ocneanu,O}, which gives a tensor 

1361: category. 

1362: 

1363: \subsection{Tube systems}

1364: 

1365: Let us start with the definition of a tube system. Let $\Delta$ be a 

1366: finite system. A {\it tube system} ${\cal D}(\Delta)$ is a tensor category 

1367: defined in the following way. 

1368: First of all, the objects of ${\cal D}(\Delta)$ are the ${\mathbb C}$-linear 

1369: span of all minimal projections in the tube algebra Tube\/$\Delta$. 

1370: For minimal projections $p_i$ and $p_j$, the hom-set ${\rm Hom}(p_i, p_j)$ 

1371: is the set of vectors in the labeled surface depicted in  Figure \ref{p_i-p_j}. 

1372: For general objects $X, Y \in {\cal D}(\Delta)$, we define ${\rm Hom} (X,Y)$ by 

1373: extending ${\rm Hom}(p_i, p_j)$ by linearity. 

1374: \begin{figure}[htbp]

1375:   \begin{center}

1376: \scalebox{0.47}{

1377: \includegraphics{p_i-p_j.eps}}  

1378:  \end{center}

1379:  \caption{Hom$(p_i,p_j)$}

1380:  \label{p_i-p_j}

1381: \end{figure}

1382: Let  $\sim$ be the 

1383: Murray-von Neumann equivalence relation between projections. (Namely, 

1384: two projections $p \; ,q \in {\rm Tube}\; \Delta$ are equivalent in the sense 

1385: of Murray-von Neumann if there exists a partial isometry $v \in {\rm Tube} \; 

1386: \Delta$ such that $p=v^* v$ and $q=v v^*$.) If $p_i \sim p_j$, then 

1387: ${\rm Hom}(p_i, p_j) \cong \mathbb{C}$ and if $p_i \nsim p_j$, then 

1388: ${\rm Hom}(p_i, p_j) =\{ 0 \}$.  Hence, minimal projections are  

1389: simple objects in ${\cal D}(\Delta)$. 

1390: We denote ${\rm Hom}(p_i, p_i)$ by ${\rm End}(p_i)$ for simplicity. 

1391: 

1392: For simple objects $p,q$ and $r$, the composition $x \cdot y$ of $x \in 

1393: {\rm Hom} (p,q)$ and $y \in {\rm Hom}(q,r)$ is defined by the concateneation 

1394: of $x$ followed by $y$. See Figure \ref{composition}. For general objects $p,q$ 

1395: and $r$, we define the composition of two morphisms as above by linearlity. 

1396: We also define $x^*$ for $x$ in ${\rm Hom}(p,q)$ by inversing $x$ inside 

1397: out. 

1398: \begin{figure}[htbp]

1399:   \begin{center}

1400: \scalebox{0.4}{

1401: \includegraphics{composition.eps}}  

1402:  \end{center}

1403:  \caption{The composition of $x \in {\rm Hom} (q,r)$ and $y \in {\rm Hom}(p,q)$}

1404:  \label{composition}

1405: \end{figure}

1406: 

1407: Let $p_j$ be a simple object and $q$ be an object in ${\cal D}(\Delta)$. We 

1408: define an inner product $\langle x, y \rangle$ of $x, y \in {\rm Hom}(q, p_i)$ by 

1409: $\langle x, y \rangle= x \cdot y^* $, as a composition of morphisms. Then, with this 

1410: inner product, ${\rm Hom}(q, p_i)$ becomes a Hilbert space. 

1411: 

1412: Next, we define the tensor product $p_i \otimes p_j$ of two simple objects 

1413: $p_i$ and $p_j$ by the fusion product $p_i * p_j$ which was defined in 

1414: Section 2. 

1415: Then, ${\rm Hom} (p_i \otimes p_j, p_k)$ consists of  vectors in the labeled 

1416: surface of Figure \ref{Nijk}. We define an inner product $\langle x, y 

1417: \rangle$ of $x, y \in {\rm Hom} (p_i \otimes p_j, p_k)$ by a composition of two 

1418: morphisms $\langle x, y \rangle= x \cdot y^*  \in {\rm End}(p_k) \cong \mathbb{C}$. 

1419: (Here, $y^*$ is defined by inversing $y$ (pants) inside out.) Note that we have 

1420: Frobenius reciprocities for morphisms of ${\rm Hom}(p_i \otimes p_j, p_k)$. 

1421: For instance, we obtain ${\rm Hom}(p_i \otimes p_j, p_k) \cong {\rm Hom}(

1422: p_i, p_k \otimes p_{\bar{j}})$. See Figure \ref{frobenius}. Frobenius 

1423: reciprocities are given by the graphical operations in the category 

1424: ${\cal D}(\Delta)$. 

1425: \begin{figure}[htbp]

1426:   \begin{center}

1427: \scalebox{0.65}{

1428: \includegraphics{frobenius.eps}}  

1429:  \end{center}

1430:  \caption{A Frobenius reciprocity map}

1431:  \label{frobenius}

1432: \end{figure}

1433: 

1434: We call the above defined semi-simple tensor category ${\cal D}(\Delta)$ 

1435: a {\it tube system}. 

1436: 

1437: To compute the quantum dimension of $p_i$, we make the composition of 

1438: morphisms $b_i^* \circ b_i$, where $b_i$ is the distinguished morphism 

1439: from $p_0$  to $p_i \otimes p_{\bar{i}}$ obtained from Frobenius 

1440: reciprocity of $id_{p_i} \in {\rm End}(p_i)$. It must be a scalar multiple 

1441: of $p_0$ since $p_0$ is a simple object. We put this value $c$. 

1442: (See Figure \ref{qdim}.) 

1443: \begin{figure}[htbp]

1444:   \begin{center}

1445: \scalebox{0.6}{

1446: \includegraphics{q-dim.eps}}  

1447:  \end{center}

1448:  \caption{Computing a quantum dimension}

1449:  \label{qdim}

1450: \end{figure}

1451: 

1452: To obtain the value $c$, we connect the upper $p_0$ and the lower $p_0$, 

1453: and embed it into $S^3$. Fill out the outside of the tube, and take the 

1454: Turaev-Viro-Ocneanu invariant of it. Then, we get the 

1455: $c=J(\text{\LARGE $\bigcirc $}; i)/S_{00}$. This is the quantum dimension of 

1456: $p_i$, denoted by ${\rm dim}\; p_i$.

1457: 

1458: We define the map $c_{p_i,p_j}$ from $p_i \otimes p_j$ to 

1459: $p_j \otimes p_i$ as in Figure \ref{braiding}.

1460: \begin{figure}[htbp]

1461:   \begin{center}

1462: \scalebox{0.8}{

1463: \includegraphics{braiding.eps}}  

1464:  \end{center}

1465:  \caption{A map $c_{p_i,p_j}$}

1466:  \label{braiding}

1467: \end{figure}

1468: It is easy to see that this map $c_{p_i,p_j}$ satisfies the axioms of 

1469: a braiding, since the map $c_{p_i,p_j}$ is defined in a topological 

1470: way. So, we now know the tensor category ${\cal D}(\Delta)$ is braided. 

1471: 

1472: We further define a map $\theta_{p_i} \in {\rm End}(p_i)$ by Figure 

1473: \ref{twist}, where $Z_\Delta$ is evaluated at the 3-ball removed a twisted 

1474: solid tube and the two tubes in the boudaries of it are both labeled by $p_i$. 

1475: \begin{figure}[htbp]

1476:   \begin{center}

1477: \scalebox{0.8}{

1478: \includegraphics{twist-tube.eps}}  

1479:  \end{center}

1480:  \caption{The twist map}

1481:  \label{twist}

1482: \end{figure}

1483: This map satisfies the following two equalities. (By the definition of 

1484: $\theta_{p_i}$, these equalities are proven by making pictures corresponding 

1485: to the formulas in both sides and  using some topological moves.)

1486: \begin{eqnarray*}

1487: & \theta_{p_i \otimes p_j} = c_{p_j,p_i} \circ c_{p_i,p_j} \circ (\theta_{p_i} 

1488:  \otimes \theta_{p_j}), \\

1489: & (\theta_{p_i} \otimes id_{p_{\bar{i}}}) \circ b_i = (id_{p_i} \otimes 

1490:  \theta_{p_{\bar{i}}} ) \circ b_i.

1491: \end{eqnarray*}

1492: Namely, $\theta_{p_i}$ defines a twist on ${\cal D}(\Delta)$ and this makes 

1493: ${\cal D}(\Delta)$ a ribbon category.  

1494: 

1495: Let us make the following compositions of morphisms in ${\cal D}(\Delta)$. 

1496: \begin{eqnarray*}

1497: &b_0& 

1498:  \circ (id_{p_0} \otimes  b_j) 

1499:  \circ (b_i \otimes id_{p_j} \otimes id_{b_{\bar{j}}}) 

1500:  \circ (id_{p_i} \otimes c_{p_{\bar{i}},p_{j}} \otimes id_{p_{\bar{j}}})

1501:  \circ (id_{p_i} \otimes c_{p_j, p_{\bar{i}}} \otimes id_{p_{\bar{j}}} ) \\

1502:  &\circ& (b_i^* \otimes id_{p_j} \otimes p_{\bar{j}}) 

1503:  \circ (id_{p_0} \otimes b_j^*) 

1504:  \circ b_0^*

1505: \end{eqnarray*}

1506: Then, it is a scalar multiple of $p_0$ and makes a Hopf link $H$ as a 

1507: diagram. (See Figure \ref{hopf}.) 

1508: \begin{figure}[htbp]

1509:   \begin{center}

1510: \scalebox{0.4}{

1511: \includegraphics{Sij.eps}}  

1512:  \end{center}

1513:  \caption{The Hopf link}

1514:  \label{hopf}

1515: \end{figure}

1516: Let us denote this scalar of the Hopf link $H$ by $s_{ij}$. Embed this 

1517: compositions of tubes into $S^3$ and 

1518: fill out the outside of tubes. By Proposition \ref{Proposition2} in Section 2, 

1519: this provides 

1520: \[

1521:  s_{ij}=\frac{J(H; i, j)}{S_{00}}=\frac{S_{ij}}{S_{00}}, 

1522: \]

1523: where $S=(S_{ij})_{i,j=0}^r$ is the $S$-matrix  with respect to our Verlinde basis 

1524: $\{ p_0, \dots, p_r \}$. 

1525: Since the $S$-matrix is unitary, the matrix $(s_{ij})_{i,j=0}^r$ is 

1526: invertible. It means that our category ${\cal D}(\Delta)$ is modular. 

1527: 

1528: \begin{rem}

1529:  The notions of the tube algebra and the tube system are also described by 

1530: the language of a II$_1$-subfactor $N \subset M$ with finite Jones index 

1531: and finite depth, although our exposition here uses  an infinite 

1532: subfactor. 

1533: \end{rem}

1534: 

1535: \subsection{Dehn surgery formula as a Reshetikhin-Turaev invariant}

1536: 

1537: Let $\cal C$ be a modular category and $\{ V_i \}_{i=0}^r$ its simple 

1538: objects. Put $\Delta=\Delta_+ = \sum_{i=0}^r t_i^{-1} ({\rm dim}\; V_i)^2$, 

1539: $\Delta_- = \sum_{i=0}^r t_i ({\rm dim}\; V_i)^2$ and 

1540: $D=(\sum_{i=0}^r ({\rm dim} V_i)^2)^{1/2}$. (Here, for $\Delta$, we 

1541: followed the notation in \cite{Turaev}.)

1542: Let $M$ be a closed 3-manifold obtained from $S^3$ by Dehn surgery along a 

1543: framed link $L=L_1 \cup \cdots \cup L_m$. Then, the Reshetikhin-Turaev invariant 

1544: of $M$ is given by the formula 

1545: \[

1546:  \tau(M) = \Delta^{\sigma(L)} D^{-\sigma(L)-m-1} \sum_{\lambda \in Col(L)} 

1547:  (\prod_{n=1}^m {\rm dim} V_{\lambda(n)}) F(L,\lambda), 

1548: \]

1549: where $\sigma(L)$ is the signature of $L$ and $F(L,\lambda)$ is 

1550: the invariant of the colored framed link  $(L, \lambda)$. 

1551: (See \cite{Turaev} for the details. Also see \cite{RT}.) 

1552: 

1553: We note that in the above formula ${\rm dim} V_{\lambda(n)} = 

1554: s_{0\lambda(n)}$. Hence the original Rehshtikhin-Turaev formula can be 

1555: rewritten with the $s$-matrix in the form 

1556: \[

1557:  \tau(M) = \Delta^{\sigma(L)} D^{-\sigma(L)-m-1} \sum_{\lambda \in Col(L)} 

1558:  (\prod_{n=1}^m s_{0 \lambda(n)})  F(L,\lambda).

1559: \]

1560: 

1561: Now, we start with the modular category ${\cal D}(\Delta)$ defined in 

1562: Section 5.1 instead of a 

1563: general modular category and make the 

1564: Rehsetikhin-Turaev formula. In our case, we already have $S$-matrix 

1565: $( S_{ij} )_{i,j=0}^r$ from the Turaev-Viro-Ocneanu TQFT, which is expressed 

1566: with respect to a Verlinde basis $\{ p_0, \dots, p_r \}$ in 

1567: $Z_\Delta(S^1 \times S^1)$, and since we know that 

1568: $\dim \; p_i=S_{0i}/S_{00}$, by using $D=1/S_{00}$, we can rewrite the 

1569: Reshetikhin-Turaev formula in the form

1570: \[

1571:  \tau(M) = \Delta^{\sigma(L)} D^{-\sigma(L)} \sum_{\lambda \in Col(L)} 

1572:  (\prod_{n=1}^m S_{0 \lambda(n)}) S_{00} F(L,\lambda).

1573: \]

1574: 

1575: We will prove that $D=\Delta$.

1576: First of all, from the equality $Z_\Delta (S^3)=Z_\Delta (L(1,1))

1577: =\sum_{i=0}^r t_i^{-1} S_{0i}^2$,  we have $\frac{1}{S_{00}}=

1578: \lambda=\sum_{i=0}^r t_i^{-1} S_{0i}^2 $ in our notation. From this, we get 

1579: \[

1580:  \frac{1}{S_{00}}=\lambda =\sum_{i=0}^r t_i^{-1} (\frac{S_{0i}}{S_{00}})^2

1581:  = \sum_{i=0}^r t_i^{-1}({\rm dim} \; p_i)^2=\Delta_+. 

1582: \]

1583: Taking the complex conjuegation of the above formula, we get 

1584: \[

1585:  \lambda = \sum_{i=0}^r t_i ({\rm dim} \; p_i)^2=\Delta_-. 

1586: \]

1587: Thus $\lambda^2 =\Delta_+ \Delta_- = \sum_{i=0}^r 

1588: ({\rm dim} \; p_i)^2$. Namely, we have $D=\Delta$. So, the Reshetikhin-Turaev 

1589: invariant constructed from ${\cal D}(\Delta)$ is given by 

1590: \[

1591:  \tau(M) = \sum_{\lambda \in Col(L)}  (\prod_{n=1}^m S_{0 \lambda(n)}) 

1592:  S_{00} F(L,\lambda). 

1593: \]

1594: 

1595: Since $F$ is uniquely determined by the category of the ribbon tangles 

1596: \cite[Part I, Chapter I]{Turaev}, taking the normalization into consideration, 

1597: we have $F(L,\lambda)=\frac{J(L,\lambda)}{S_{00}}$. Namely, in our 

1598: case, 

1599: \[

1600:  \tau(M) = \sum_{\lambda \in Col(L)}  (\prod_{n=1}^m S_{0 \lambda(n)}) 

1601:  J(L,\lambda), 

1602: \]

1603: which is nothing but our Dehn surgery formula Proposition \ref{Proposition1}. Let us 

1604: summerize this argument as a theorem. 

1605: 

1606: \begin{thm}\label{main}

1607:  Let $\Delta$ be a finite system. 

1608:  For a closed oriented 3-manifold $M$,  the Reshetikhin-Turaev invariant $\tau(M)$ 

1609:  constructed  from a tube system ${\cal D} (\Delta)$ coincides with the 

1610:  Turaev-Viro-Ocneanu invariant $Z_\Delta (M)$. 

1611: \end{thm}

1612: 

1613: The following corollary has been proven by several authors \cite{Turaev}, 

1614: \cite{Roberts} etc in various settings. 

1615: 

1616: \begin{cor}\label{split}

1617: Let $\Delta$ be a finite $C^*$-tensor category arising from a subfactor. 

1618: If  $\Delta$ is a modular category, then ${\cal D} (\Delta)$ is equivalent 

1619: to $\Delta \otimes \Delta^{\rm op}$ and this provides us with $\tau(M) \cdot 

1620: \overline{\tau(M)}=Z_{\Delta}(M)$ for a closed oriented 3-manifold $M$, 

1621: where $\tau$ is the Reshetikhin-Turaev invariant constructed from $\Delta$. 

1622: \end{cor}

1623: 

1624: \noindent

1625: Proof. In \cite{EK2}, it is proved that ${\cal D}(\Delta)$ is equivalent  

1626: to $\Delta \otimes \Delta^{\rm op}$ when $\Delta$ is a modular category. 

1627: (See Appendix too.) Hence, the rest is clear from Theorem \ref{main}. 

1628: \hfill Q.E.D.

1629: 

1630: \appendix 

1631: 

1632: 

1633: \section{Appendix}

1634: In this Appendix, we fix inaccuracies in \cite{EK} and \cite{EK1}.

1635: In \cite{EK} and \cite{EK1}, the authors have analyzed the structure of 

1636: $M_\infty$-$M_\infty$ bimodules obtained from the {\it asymptotic 

1637: inclusion} $M \vee M^{\rm op} \subset M_\infty$ starting from 

1638: the inclusion of AFD II$_1$ factors $N \subset M$ with finite Jones index 

1639: and finite depth, following Ocneanu. 

1640: 

1641: Let $\{p_0, \cdots, p_r \}$ be a representative set of the equivalence classes of 

1642: minimal projections of Tube$\Delta$ by the Murray-von Neumann equivalence relation. 

1643: 

1644: We present the $M_\infty$-$M_\infty$ bimodule $X(p_i)$ by Figure \ref{a1}, 

1645: where the circle at the middle is empty and the annulus around it is 

1646: labeled with the minimal projection $p_i$.

1647: (See Chapter 12 in \cite{EK} for more explanation of this kind of pictures.)

1648: \unitlength 0.5mm

1649: \thicklines

1650: \begin{figure}[htpb]

1651: \begin{center}\begin{picture}(160,100)

1652: \put(70,50){\circle{20}}

1653: \put(70,50){\circle{40}}

1654: \put(70,60){\circle*{1}}

1655: \put(70,70){\circle*{1}}

1656: \spline(70,70)(59,65)(53,58)(53.5,40.5)(71,32)(84,40)(83,57)(70,60)

1657: \put(72,34){\vector(1,0){0}}

1658: \put(70,67){\vector(0,1){0}}

1659: \put(70,60){\line(0,1){10}}

1660: \put(50,50){\vector(0,-1){0}}

1661: \put(50,45){\vector(0,-1){0}}

1662: \put(60,50){\vector(0,-1){0}}

1663: \put(60,45){\vector(0,-1){0}}

1664: \put(70,50){\ellipse{60}{80}}

1665: \put(70,50){\ellipse{120}{80}}

1666: \put(10,50){\vector(0,-1){0}}

1667: \put(40,50){\vector(0,-1){0}}

1668: \put(100,50){\vector(0,-1){0}}

1669: \put(130,50){\vector(0,-1){0}}

1670: \put(70,10){\vector(1,0){0}}

1671: \put(70,10){\vector(-1,0){0}}

1672: \put(70,10){\circle*{1}}

1673: \put(70,90){\circle*{1}}

1674: \put(11.9,40){\circle*{1}}

1675: \put(11.9,60){\circle*{1}}

1676: \put(128.1,40){\circle*{1}}

1677: \put(128.1,60){\circle*{1}}

1678: \put(18,30){\circle*{1}}

1679: \put(18,70){\circle*{1}}

1680: \put(122,30){\circle*{1}}

1681: \put(122,70){\circle*{1}}

1682: \put(30.3,20){\circle*{1}}

1683: \put(30.3,80){\circle*{1}}

1684: \put(109.7,20){\circle*{1}}

1685: \put(109.7,80){\circle*{1}}

1686: \spline(30.3,80)(40,86)(60,89)(70,90)(80,89)(100,86)(109.7,80)

1687: \spline(18,70)(40,84)(60,89)(70,90)(80,89)(100,84)(122,70)

1688: \spline(11.9,60)(40,84)(60,88)(70,90)(80,88)(100,84)(128.1,60)

1689: \spline(11.9,40)(40,83)(60,87)(70,90)(80,87)(100,83)(128.1,40)

1690: \spline(18,30)(40,80)(60,87)(70,90)(80,87)(100,80)(122,30)

1691: \spline(30.3,20)(40,78)(60,87)(70,90)(80,87)(100,78)(109.7,20)

1692: %\put(70,3){\makebox(0,0){Fig. 4.14}}

1693: \end{picture}

1694: \end{center}

1695: \caption{The bimodule $X(p_i)$}

1696: \label{a1}

1697: \end{figure}

1698: We can prove that each $M_\infty$-$M_\infty$ bimodule $X(p_i)$ $(i=0, 

1699: \cdots, r)$ is irreducible in a similar way to the proof of 

1700: Theorem 12.26 in \cite{EK}. 

1701: 

1702: Since $[{}_{M_\infty} X(p_i)_{M_\infty}]^{1/2}=\dim \; p_i$, 

1703: we have the equalities $\sum_{i=0}^r [X(p_i)]=\sum_{i=0}^r (\dim \; p_i)^2

1704: =\lambda^2$, where dim $p_i$ is the quantum dimension of $p_i$ as an object of 

1705: ${\cal D}(\Delta)$. On the other hand, the global index of the asymptotic 

1706: inclusion 

1707: is given by $\lambda^2$. This means that all the irreducible 

1708: $M_\infty$-$M_\infty$ bimodules obtained from the asymptotic inclusion are 

1709: given by $\{X(p_i) \}_{i=0}^r$. 

1710: 

1711: \begin{rem}

1712: In \cite{EK} and \cite{EK1}, irreducible bimodules are labeled by $\pi_i$'s, 

1713: i.e., minimal central projections of Tube$\Delta$, instead of the minimal 

1714: projections of it, which is incorrect.

1715: \end{rem}

1716: 

1717: We describe the fusion rule of irreducble $M_\infty$-$M_\infty$ 

1718: bimodules. The relative tensor product of two $M_\infty$-$M_\infty$ 

1719: bimodules $X(p_i) \otimes_{M_\infty} X(p_j)$ is decomposed into irreducible 

1720: bimodules as in Figure \ref{a5}.

1721: 

1722: \unitlength 0.45mm

1723: \thicklines

1724: \begin{figure}[htpb]

1725: \begin{center}\begin{picture}(240,100)

1726: \put(170,50){\circle{20}}

1727: \put(170,50){\circle{14}}

1728: \put(45,50){\circle{20}}

1729: \put(45,50){\circle{14}}

1730: \put(75,50){\circle{20}}

1731: \put(75,50){\circle{14}}

1732: \put(60,50){\circle{60}}

1733: \put(60,50){\circle{66}}

1734: \put(160,50){\vector(0,-1){0}}

1735: \put(163,50){\vector(0,-1){0}}

1736: \put(30,50){\vector(0,-1){0}}

1737: \put(27,50){\vector(0,-1){0}}

1738: \put(35,50){\vector(0,-1){0}}

1739: \put(38,50){\vector(0,-1){0}}

1740: \put(65,50){\vector(0,-1){0}}

1741: \put(68,50){\vector(0,-1){0}}

1742: \put(170,50){\ellipse{60}{80}}

1743: \put(170,50){\ellipse{120}{80}}

1744: \put(110,50){\vector(0,-1){0}}

1745: \put(140,50){\vector(0,-1){0}}

1746: \put(200,50){\vector(0,-1){0}}

1747: \put(230,50){\vector(0,-1){0}}

1748: \put(170,10){\vector(1,0){0}}

1749: \put(170,10){\vector(-1,0){0}}

1750: \put(170,10){\circle*{1}}

1751: \put(170,90){\circle*{1}}

1752: \put(111.9,40){\circle*{1}}

1753: \put(111.9,60){\circle*{1}}

1754: \put(228.1,40){\circle*{1}}

1755: \put(228.1,60){\circle*{1}}

1756: \put(118,30){\circle*{1}}

1757: \put(118,70){\circle*{1}}

1758: \put(222,30){\circle*{1}}

1759: \put(222,70){\circle*{1}}

1760: \put(130.3,20){\circle*{1}}

1761: \put(130.3,80){\circle*{1}}

1762: \put(209.7,20){\circle*{1}}

1763: \put(209.7,80){\circle*{1}}

1764: %+100

1765: \spline(130.3,80)(140,86)(160,89)(170,90)(180,89)(200,86)(209.7,80)

1766: \spline(118,70)(140,84)(160,89)(170,90)(180,89)(200,84)(222,70)

1767: \spline(111.9,60)(140,84)(160,88)(170,90)(180,88)(200,84)(228.1,60)

1768: \spline(111.9,40)(140,83)(160,87)(170,90)(180,87)(200,83)(228.1,40)

1769: \spline(118,30)(140,80)(160,87)(170,90)(180,87)(200,80)(222,30)

1770: \spline(130.3,20)(140,78)(160,87)(170,90)(180,87)(200,78)(209.7,20)

1771: \put(170,35){\makebox(0,0){$p_k$}}

1772: \put(45,35){\makebox(0,0){$p_i$}}

1773: \put(75,35){\makebox(0,0){$p_j$}}

1774: \put(60,10){\makebox(0,0){$p_k$}}

1775: \put(12,47){\makebox(0,0){$\displaystyle\sum_k$}}

1776: \put(100,50){\makebox(0,0){$\otimes$}}

1777: %\put(120,3){\makebox(0,0){Fig. 4.15}}

1778: \end{picture}

1779: \end{center}

1780: \caption{Decomposition of a bimodule}

1781: \label{a5}

1782: \end{figure}

1783: Hence, the fusion rule is given by the fusion rule of the fusion algebra 

1784: associated with the Turaev-Viro-Ocneanu TQFT. 

1785: 

1786: It is easy to see that $M_\infty$-$M_\infty$ bimodules arising from the 

1787: asymptotic inclusion $M \vee M^{\rm op} \subset M_\infty$ give rise to  

1788: a modular category with the same braiding and twist as ones in a tube system. 

1789: We denote this category of $M_\infty$-$M_\infty$ bimodules by ${\cal M}_\infty$. 

1790: To a simple object $X(p_i)$ in ${\cal M}_\infty$, we assign a simple object 

1791: $p_i$ in ${\cal D}(\Delta)$. 

1792: From Figure \ref{a5}, it is easy to see that we have a functor 

1793: $F: \; {\cal M}_\infty \longrightarrow {\cal D}(\Delta)$, when we look at 

1794: the morphisms in ${\cal M}_\infty$. 

1795: 

1796: Let us now consider the opposite direction to the functor $F$. 

1797: For a given morphism $x \in {\rm Hom} (p_i \otimes p_j, p_k)$ of 

1798: ${\cal D}(\Delta)$, we will construct a homomoprhism in ${\rm Hom} (X(p_i) 

1799: \otimes_{M_\infty} X(p_j), X(p_k))$. 

1800: 

1801: \unitlength 0.5mm

1802: \thicklines

1803: \begin{figure}[htpb]

1804: \begin{center}\begin{picture}(160,100)

1805: \put(85,50){\circle{20}}

1806: \put(55,50){\circle{20}}

1807: \put(85,50){\circle{14}}

1808: \put(55,50){\circle{14}}

1809: \put(45,50){\vector(0,-1){0}}

1810: \put(75,50){\vector(0,-1){0}}

1811: \put(48,50){\vector(0,-1){0}}

1812: \put(78,50){\vector(0,-1){0}}

1813: \put(70,50){\ellipse{60}{80}}

1814: \put(70,50){\ellipse{120}{80}}

1815: \put(10,50){\vector(0,-1){0}}

1816: \put(40,50){\vector(0,-1){0}}

1817: \put(100,50){\vector(0,-1){0}}

1818: \put(130,50){\vector(0,-1){0}}

1819: \put(70,10){\vector(1,0){0}}

1820: \put(70,10){\vector(-1,0){0}}

1821: \put(70,10){\circle*{1}}

1822: \put(70,90){\circle*{1}}

1823: \put(11.9,40){\circle*{1}}

1824: \put(11.9,60){\circle*{1}}

1825: \put(128.1,40){\circle*{1}}

1826: \put(128.1,60){\circle*{1}}

1827: \put(18,30){\circle*{1}}

1828: \put(18,70){\circle*{1}}

1829: \put(122,30){\circle*{1}}

1830: \put(122,70){\circle*{1}}

1831: \put(30.3,20){\circle*{1}}

1832: \put(30.3,80){\circle*{1}}

1833: \put(109.7,20){\circle*{1}}

1834: \put(109.7,80){\circle*{1}}

1835: \spline(30.3,80)(40,86)(60,89)(70,90)(80,89)(100,86)(109.7,80)

1836: \spline(18,70)(40,84)(60,89)(70,90)(80,89)(100,84)(122,70)

1837: \spline(11.9,60)(40,84)(60,88)(70,90)(80,88)(100,84)(128.1,60)

1838: \spline(11.9,40)(40,83)(60,87)(70,90)(80,87)(100,83)(128.1,40)

1839: \spline(18,30)(40,80)(60,87)(70,90)(80,87)(100,80)(122,30)

1840: \spline(30.3,20)(40,78)(60,87)(70,90)(80,87)(100,78)(109.7,20)

1841: \put(55,35){\makebox(0,0){$p_i$}}

1842: \put(85,35){\makebox(0,0){$p_{\bar{i}}$}}

1843: %\put(70,3){\makebox(0,0){Fig. 4.14}}

1844: \end{picture}

1845: \end{center}

1846: \caption{The bimodule $X(p_i) \otimes_{M_\infty} X(p_{\bar{i}})$}

1847: \label{a3}

1848: \end{figure}

1849: 

1850: Let $y_0$ be an arbitrary element of $X(p_i) \otimes_{M_\infty} X(p_j)$. This 

1851: $y_0$ is considered as an element in the bimodule in Figure \ref{a3}. We 

1852: denote it by $y_1$. Then, we attach $x \in {\rm Hom}(p_i \otimes p_j, p_k)$ to 

1853: this $y_1$ by using two tubes $p_i$, $p_j$. See Figure \ref{a6}. 

1854: \begin{figure}[htbp]

1855:   \begin{center}

1856: \scalebox{0.45}{

1857: \includegraphics{a6.eps}}  

1858:  \end{center}

1859:  \caption{}

1860:  \label{a6}

1861: \end{figure}

1862: \begin{figure}[htbp]

1863:   \begin{center}

1864: \scalebox{0.9}{

1865: \includegraphics{a7.eps}}  

1866:  \end{center}

1867:  \caption{}

1868:  \label{a7}

1869: \end{figure}

1870: The central part of Figure \ref{a6} can be viewed as in Figure \ref{a7}. 

1871: Hence, we get a morphism in ${\rm Hom}(X(p_i) \otimes_{M_\infty} X(p_j), 

1872: X(p_k))$. This induces a functor $G: \; {\cal D}(\Delta) \longrightarrow 

1873: {\cal M}_\infty$. It is not difficult to see that $F$ and $G$ are 

1874: functors, which preserve the operations in modular categories, and inverse to 

1875: each other. 

1876: 

1877: Hence, the category of $M_\infty$-$M_\infty$ bimodules obtained from the 

1878: asymptotic inclusion ${\cal M}_\infty$ is equivalent to the tube 

1879: system ${\cal D}(\Delta)$ constructed from $N \subset M$ as modular 

1880: categories. 

1881: 

1882: 

1883: \newpage

1884: \begin{thebibliography}{100}

1885: \bibitem{Atiyah}

1886: M. F. Atiyah,  {\it Topological quantum field theories}, 

1887: Publ. Math. I.H.E.S.  68 (1989) 175---186.

1888: \bibitem{EK}  D. E. Evans, Y. Kawahigashi, 

1889: {\it Quantum symmetries on operator algebras}, Oxford University Press, 1998.

1890: \bibitem{EK1} D. E. Evans, Y. Kawahigashi,

1891: {\it On Ocneanu's theory of asymptotic inclusions for subfactors, topological 

1892: quantum field theories and quantum doubles}, Internat. J. Math. 6 (1995) 

1893: 205--228. 

1894: \bibitem{EK2} D. E. Evans, Y. Kawahigashi,

1895: {\it Orbifold subfactors from Hecke algebras II}, Commun. Math. Phys. 196 

1896: (1998) 331--361. 

1897: \bibitem{FSS}

1898: L. Feher, A. Stipsicz and J. Szenthe ed., {\it Topological field theories and geometry of loop spaces}, 

1899: World Scientific, 1992.

1900: \bibitem{Iz}

1901: M. Izumi, {\it Applications of fusion rules to classification of subfactors}, 

1902: Publ. RIMS, 27, (1991), 953--994.

1903: \bibitem{Izumi1}

1904: M. Izumi, {\it The structures of sectors associated with the 

1905: Longo-Rehren inclusions I. General theory}, Commun. Math.Phys.  213 (2000) 

1906: 127---179. 

1907: \bibitem{Izumi2}

1908: M. Izumi, {\it The structures of sectors associated with the 

1909: Longo-Rehren inclusions II. Examples}, Rev. in Math.Phys. 13 (2001) 

1910: 603---674. 

1911: \bibitem{J}

1912: V. F. R. Jones, {\it A polynomial invariant for knots via von Neumann 

1913: algebras}, Bull. of the AMS, 12 (1985), 103--112.

1914: \bibitem{Kerler}

1915: T. Kerler, {\it Genealogy of non-perturbative quantum-invariants

1916: of 3-manifold: the surgical family}, in ``Geometry and Physics''

1917: (Aarhus 1995), 503--547. Lect. Notes in Pure and Appl. Math.,

1918: 184, Dekker, New York, 1997.

1919: \bibitem{Kohno}

1920: T. Kohno, {\it Topological invariants for $3$-manifolds using representations 

1921: of mapping class groups II: Estimating tunnel number of knots}, 

1922: Contemp. Math. 175 (1994) 193--217.

1923: \bibitem{LR}

1924: R. Longo and J. E. Roberts, {\it A theory of dimension}, $K$-theory 11 (1997), 

1925: 103--159. 

1926: \bibitem{Mueger}

1927: M. M\"uger, {\it From subfactors to categories and topology II.

1928: The quantum double of subfactors and categories}, preprint, math.CT/0111205, 

1929: 2001.

1930: \bibitem{Ocneanu}

1931: A. Ocneanu, {\it Chirality for operator algebras}, 39---63, in 

1932: ``Subfactors", ed. by H. Araki, et al., World Scientific, 1994.

1933: \bibitem{O} 

1934: A. Ocneanu, {\it Operator Algebras, Topology and Subgroups of Quantum 

1935: Symmetry--Construction of Subgroups of Quantum Groups--}, 235---263, 

1936: written by Satoshi Goto(Sophia University) and Nobuya Sato(Osaka Prefecture 

1937: University), in ``Advanced Studies in Pure and Mathematics'' 31, Taniguchi 

1938: Conference in Mathematics Nara '98, Mathematical Society of Japan, 2000.

1939: \bibitem{RT}

1940: N. Reshetikhin and V. G. Turaev, {\it Invariants of $3$-manifolds via link polynomials and quantum groups}, Invent. Math. 103 (1991), p. 547--597.

1941: \bibitem{Roberts}

1942: J. Roberts, {\it Skein theory and Turaev-Viro invariants}, Topology 34 (1995), 

1943: 771--787. 

1944: %\bibitem{Rolfsen}

1945: %D. Rolfsen, {\it Knots and links}, 

1946: %Publish or Perish, Berkeley 1976. 

1947: \bibitem{SatoWakui}

1948: N. Sato and M. Wakui, {\it (2+1)-dimensional topological quantum field theory 

1949: with Verlinde basis and Turaev-Viro-Ocneanu invariants of 3-manifolds}, 

1950: preprint 2000.

1951: \bibitem{SuzukiWakui}

1952: K. Suzuki and M. Wakui, 

1953: {\it On the Turaev-Viro-Ocneanu invariant of $3$-manifolds derived from the 

1954: $E_6$-subfactor}, Kyushu J. Math. 56 (2002) 59--81. 

1955: \bibitem{Takata}

1956: T. Takata, {\it Invariants of $3$-manifolds associated with quantum groups and Verlinde's formula}, 

1957: Publ. RIMS Kyoto Univ. 28 (1992) 139--167.

1958: \bibitem{Turaev}

1959: V. G. Turaev, {\it Quantum invariants of knots and 3-manifolds}, Walter 

1960: de Gruyter 1994.

1961: \bibitem{TuraevViro}

1962: V. G.  Turaev and O. Ya Viro, 

1963: {\it State sum invariants of 3-manifolds and quantum $6j$-symbols}, 

1964: Topology 31 (1992) 865---902. 

1965: \bibitem{Verlinde}

1966: E. Verlinde, {\it Fusion rules and modular transformations in 2D conformal field theory}

1967: , Nucl. Phys. B300 (1988) 360--376.

1968: \bibitem{Wakui}

1969: M. Wakui, {\it Fusion algebras for orbifold models (a survey)}, 

1970: in \lq\lq Topology, geometry and field theory" edited by K. Fukaya, M. Furuta, T. Kohno 

1971: and D. Kotschick, World Scientific, 1994, 225--235.

1972: \bibitem{Witten}

1973: E. Witten, {\it Quantum field theory and Jones polynomial}, 

1974: Commun. Math. Phys. 121 (1989) 351--399.

1975: 

1976: \end{thebibliography}

1977: 

1978: 

1979: \end{document}

1980: