hep-th0412290/IV.tex
1: 
2: % v 8.4  8.6.05  hep-th/0412290
3: 
4: \catcode`\@=11
5: \def\citen#1{\if@filesw \immediate\write \@auxout {\string\citation{#1}}\fi%
6: \@tempcntb\m@ne \let\@h@ld\relax \def\@citea{}%
7: \@for \@citeb:=#1\do {\@ifundefined {b@\@citeb}%
8:     {\@h@ld\@citea\@tempcntb\m@ne{\bf ?}%
9:     \@warning {Citation `\@citeb ' on page \thepage \space undefined}}%
10:     {\@tempcnta\@tempcntb \advance\@tempcnta\@ne
11:     \setbox\z@\hbox\bgroup\ifcat0\csname b@\@citeb \endcsname \relax
12:     \egroup \@tempcntb\number\csname b@\@citeb \endcsname \relax
13:     \else \egroup \@tempcntb\m@ne \fi \ifnum\@tempcnta=\@tempcntb
14:     \ifx\@h@ld\relax \edef \@h@ld{\@citea\csname b@\@citeb\endcsname}%
15:     \else \edef\@h@ld{\hbox{--}\penalty\@highpenalty
16:     \csname b@\@citeb\endcsname}\fi
17:     \else \@h@ld\@citea\csname b@\@citeb \endcsname \let\@h@ld\relax \fi}%
18: \def\@citea{,\penalty\@highpenalty\hskip.13em plus.13em minus.13em}}\@h@ld}
19: \def\@citex[#1]#2{\@cite{\citen{#2}}{#1}}%
20: \def\@cite#1#2{\leavevmode\unskip\ifnum\lastpenalty=\z@\penalty\@highpenalty\fi%
21:   \ [{\multiply\@highpenalty 3 #1%
22:   \if@tempswa,\penalty\@highpenalty\ #2\fi}]}   %
23: \makeatother % end of CITE.STY
24: \catcode`\@=12
25: 
26: \newcounter{defthm}
27: 
28: \def\AA            {{\!A}}
29: \def\alg           {algebra}
30: \def\Alpha         {\aleph}
31: \newcommand\Ann[2] {{{\rm A}_{#1}}^{\!\!#2}}
32: \def\AutO          {{{\rm Aut}\,\mathcal{O}}}
33: \def\Bb            {\ensuremath{\Langle\vhi\vsi\Rangle}}
34: \def\bbb           {\ensuremath{\Langle\vsi\vsi\vsi\Rangle}}
35: \def\BBB           {\ensuremath{\Langle\vhi\vhi\vhi\Rangle}}
36: \def\BBmat         {{\sf B}}
37: \newcommand\Bmat[9]{{\sf B}_{\!{\sss#6}#4{\sss#7},{\sss#8}#5{\sss#9}}
38:                    ^{\,({#1}\,{#2})\,{#3}}}
39: \newcommand\Bsmat[6]{{\sf B}_{\,{#5}\,{#6}}^{\,({#1}\,{#2}\,{#3})\,{#4}}} 
40: \def\bc            {boundary condition}
41: \def\Bc            {Boundary condition}
42: \def\be            {\begin{equation}}
43: \def\bea           {\begin{equation}\begin{array}l}
44: \def\beaa          {\begin{equation}\begin{array}{ll}}
45: \def\bearl         {\begin{array}{l}}
46: \def\bearll        {\begin{array}{ll}}
47: \def\bndb          {\mathrm{b}}
48: \def\BX            {\ensuremath{\Langle\vhi\Rangle_\times}}
49: \def\CA            {{\mathfrak V}}
50: \def\calb          {\mathcal{B}}
51: \def\calc          {\ensuremath{\mathcal C}}
52: \def\calca         {\ensuremath{{\mathcal C}_{\!A}}}
53: \def\calcaa        {\ensuremath{{}_A\!{\,}{\mathcal C}_{\!A}}}
54: \def\cald          {{\mathcal D}}
55: \def\calh          {\ensuremath{\mathcal H}}
56: \def\calH          {{\mathcal H}}
57: \def\calm          {\mathcal{M}}
58: \def\calp          {\mathcal{P}}
59: \def\cats          {categories}
60: \def\Cft           {Conformal field theory}
61: \def\cft           {conformal field the\-o\-ry}
62: \newcommand\cbs    {{\hat\calb}}
63: \def\cfts          {conformal field theories}
64: \def\Cfts          {Conformal field theories}
65: \def\chii          {\raisebox{.15em}{$\chi$}}
66: \def\chiI          {\raisebox{.15em}{$\scriptstyle\chi$}}
67: \def\cir           {\,{\circ}\,}
68: \def\coborc        {\mbox{{\sf 3}\hy{\sf cobord}$(\calc)$}}
69: \def\cobord        {\mbox{{\sf 3}\hy{\sf cobord}}}
70: \def\compac        {compactification}
71: \def\complex       {\mathbbm C}
72: \def\con           {conformal }
73: \def\Con           {Conformal }
74: \def\corfu         {correlation function}
75: \def\Corfu         {Correlation function}
76: \def\cua           {current algebra}
77: \def\DDD           {\ensuremath{\Langle\Thetha\Thetha\Thetha\Rangle}}
78: \def\dftd          {\mathrm{d}}
79: \def\dim           {{\rm dim}}
80: \def\dimA          {{\rm dim}(A)}
81: \def\dimc          {{\rm dim}_{\scriptscriptstyle\complex}}
82: \def\dimM          {{\rm dim}(\dot M)}
83: \def\dsty          {\displaystyle }
84: \def\DX            {\ensuremath{\Langle\Thetha\Rangle_\times}}
85: \def\E             {\mbox{E$_7$} }
86: \def\eaa           {\jmath_{\,\One}^{A}}
87: \def\eE            {{\rm e}}
88: \def\ee            {\end{equation}}
89: \def\eear          {\end{array}}
90: \newcommand\eev[1] {{{}^{\vee\!}}{#1}} 
91: \newcommand\Eev[1] {{{}^{\vee\!}}\!{#1}} 
92: \def\End           {{\rm End}}
93: \newcommand\epicture[2] {\end{picture}\\{}\\[#1.#2em]\end{array}}
94: \def\eps           {\varepsilon}
95: \def\Eps           {\epsilon}
96: \def\epsnat        {\eps_{\ell}}
97: \def\epsnat        {\eps_{\natural}}
98: \def\epsnatt       {\eps_{r}}
99: \def\epsnatt       {\eps_{\rho}}
100: \def\epsnattop     {\eps_{\natural,{\rm top}}}
101: \def\eq            {\,{=}\,}
102: \def\equi          {equivalent}
103: \newcommand\erf[1] {(\ref{#1})}
104: \newcommand\erp[2] {(\ref{pic4-#1#2})}
105: \newcommand\Erp[3] {(\ref{pic4-#1#2#3})}
106: \newcommand\F[9]   {{\sf F}_{\!{\sss#6}#4{\sss#7},{\sss#8}#5{\sss#9}}
107:                    ^{\,({#1}\,{#2})\,{#3}}}
108: \newcommand\Fbox[1]{\shadowbox{#1}}
109: \def\ff            {{\mathscr F}}
110: \def\FF            {{\sf F}}
111: \def\findim        {fini\-te-di\-men\-si\-o\-nal}
112: \newcommand\foodnode[1] {\,\footnote{~#1}}
113: \newcommand{\fose}[1] {{[\![#1]\!]}}
114: \newcommand\Frac[2]{\mbox{\large$\frac{#1}{#2}$}}
115: \def\frob          {Fro\-be\-ni\-us algebra}
116: \newcommand\Fs[6]  {{\sf F}_{\,{#5}\,{#6}}^{\,({#1}\,{#2}\,{#3})\,{#4}}}
117: \def\fsi           {Fro\-be\-ni\-us\hy Schur indicator}
118: \def\ft            {field theory}
119: \def\fts           {field theories}
120: \def\furu          {fusion rule}
121: \newcommand\G[9]   {{\sf G}_{\,{\sss#6}#4{\sss#7},{\sss#8}#5{\sss#9}}
122:                    ^{\,({#1}\,{#2})\,{#3}}}
123: \def\gemo          {Gepner model}
124: \def\GG            {{\sf G}}
125: \newcommand\glue[1]{\stackrel{#1}{\leftarrowtail}}
126: \newcommand\Gs[6]  {{\sf G}_{\,{#5}\,{#6}}^{\,({#1}\,{#2}\,{#3})\,{#4}}}
127: \def\Gz            {Generalised }
128: \def\hatmu         {\Thetha}  %% defect field 
129: \def\hatphi        {[\vPhi]}
130: \def\hatpsi        {\varPsi}
131: \def\Hom           {{\rm Hom}}
132: \def\HomA          {{\rm Hom}_{\!A}}
133: \def\HomAA         {{\rm Hom}_{\!A|A}}
134: \def\hopf          {Hopf algebra}
135: \newcommand\hsp[1] {\mbox{\hspace{#1 em}}}
136: \def\hy            {$\mbox{-\hspace{-.66 mm}-}$}
137: \def\I             {\mbox{$\II$}}
138: \newcommand\iaa[2] {\imath_{#1#2}^{\,A}}
139: \def\id            {\mbox{\sl id}}
140: \def\idsmall       {\mbox{\scriptsize\sl id}}
141: \def\ide           {identification}  
142: \def\idM           {\id_{\dot M}}
143: \def\idN           {\id_{\dot N}}
144: \def\ii            {{\rm i}}
145: \def\II            {\mathcal I}
146: \def\Im            {{\rm Im}\,}
147: \def\iN            {\,{\in}\,}
148: \def\In            {\prec}
149: \def\IN            {\,{\In}\,}
150: \newcommand\includeourbeautifulpicture[2] {{\begin{picture}(0,0)(0,0)
151:             \scalebox{.38}{\includegraphics{pic4_#1#2.eps}} \end{picture}}}
152: \newcommand\Includeourbeautifulpicture[3] {{\begin{picture}(0,0)(0,0)
153:             \scalebox{.38}{\includegraphics{pic4_#1#2#3.eps}} \end{picture}}}
154: \newcommand\INcludeourbeautifulpicture[4] {{\begin{picture}(0,0)(0,0)
155:             \scalebox{.38}{\includegraphics{pic4_#1#2#3#4.eps}} \end{picture}}}
156: \newcommand\includeournicelargepicture[2] {{\begin{picture}(0,0)(0,0)
157:             \scalebox{.28}{\includegraphics{pic4_#1#2.eps}} \end{picture}}}
158: \newcommand\Includeournicelargepicture[3] {{\begin{picture}(0,0)(0,0)
159:             \scalebox{.28}{\includegraphics{pic4_#1#2#3.eps}} \end{picture}}}
160: \newcommand\includeourniceLargepicture[2] {{\begin{picture}(0,0)(0,0)
161:             \scalebox{.14}{\includegraphics{pic4_#1#2.eps}} \end{picture}}}
162: \newcommand\includeournicehugepicture[2] {{\begin{picture}(0,0)(0,0)
163:             \scalebox{.10}{\includegraphics{pic4_#1#2.eps}} \end{picture}}}
164: \newcommand\Includeournicehugepicture[3] {{\begin{picture}(0,0)(0,0)
165:             \scalebox{.10}{\includegraphics{pic4_#1#2#3.eps}} \end{picture}}}
166: \def\inda          {{\rm Ind}_{A\!}}
167: \def\indar         {\widetilde{{\rm Ind}}_{A\!}}
168: \def\indaz         {{\rm Ind}_{A_{2r}\!}}
169: \def\Intro         {Introduction }
170: \newcommand\Ising  {{\mathcal{C}_{\sf Is}}}
171: \def\Itemize       {\def\leftmargini{1.8em}\begin{itemize}\addtolength\itemsep{-3pt}} 
172: \def\ITemize       {\def\leftmargini{1.0em}\begin{itemize}\addtolength\itemsep{-3pt}} 
173: \def\J             {\mbox{$\JJ$}}
174: \def\JJ            {\mathcal J}
175: \def\K             {{\rm K}}
176: \def\kappai        {\kappa}
177: \def\kappaj        {{\kappa'}}
178: \def\KK            {\mathcal{K}} 
179: \def\ksb           {\Xi}
180: \def\KZ            {Knizh\-nik\hy Za\-mo\-lod\-chi\-kov }
181: \newcommand\labl[1]{\label{#1}\ee}
182: \newcommand\labp[2]{\label{pic4-#1#2}\ee}
183: \newcommand\Labp[3]{\label{pic4-#1#2#3}\ee}
184: \def\Langle        {\langle}
185: \def\lhs           {left hand side}
186: \def\llb           {\mbox{\large[}}
187: \def\Llb           {\mbox{\Large(}}
188: \def\Loc           {{\rm Hom}_{\rm loc}}
189: \def\LocUp         {{\rm Hom}^{\rm loc}}
190: \newcommand\lr[2]  {\langle #1{,}#2\rangle}
191: \newcommand\lra[2] {\langle #1{,}#2\rangle_{\!A}^{}}
192: \def\lrb           {\mbox{\large]}}
193: \def\Lrb           {\mbox{\Large)}}
194: \def\map           {morphism}
195: \newcommand\m[7]   {m_{#1#2,#3#4}^{\;#5#6;\,#7}}
196: \def\M             {{\dot M}}
197: \def\mathcalD      {{\mathscr{D}}}
198: \newcommand\mC[7]  {C_{#1#2,#3#4}^{\;#5#6;\,#7}}
199: \newcommand\mCb[8] {C_{#1#2,#3#4}^{(#8)#5#6;\,#7}}
200: \def\mimo          {minimal model}
201: \def\mocla         {\mbox{$\mathcal S$}}
202: \def\modinv        {modular invarian}
203: \def\mtop          {{m_{\rm top}}}
204: \def\M             {{\rm M}}
205: \def\MX            {{\rm M}_{\rm X}}
206: \def\n             {{\dot N}}
207: \newcommand\N[3]   {{N_{#1#2}}^{\!\!#3}}
208: \def\nxt           {\raisebox{.08em}{\rule{.44em}{.44em}}\hsp{.4}}
209: \def\Nxt           {\raisebox{.08em}{\rule{.44em}{.44em}}}
210: \def\Nu            {{\mathcal V}}
211: \def\oa            {operator algebra}
212: \def\obj           {{\mathcal O}bj}
213: \def\objc          {{\mathcal O}bj(\calc)}
214: \def\objca         {{\mathcal O}bj(\calca)}
215: \def\one           {{\bf1}}
216: \def\One           {0}
217: \def\onemat        {{\mbox{\small1}\!\!{1}}}
218: \def\ope           {operator product expansion}
219: \def\opl           {\,{\oplus}\,}
220: \def\Opl           {\!\!\!\!{\oplus}\!\!\!\!}
221: \def\orr           {{\rm or}}
222: \newcommand\ot[1]  {\,{\otimes^{#1}}\,}
223: \newcommand\oT[1]  {\,{\otimes^{#1}}}
224: \def\ota           {\,{\otimes}_{\!A}^{}\,}
225: \def\otA           {\,{\otimes}_{\!A}^{}\,}
226: \def\OtA           {{\otimes}_{\!A}^{}}
227: \def\otB           {\,{\otimes}_{B}^{}\,}
228: \def\OtB           {{\otimes}_{B}^{}}
229: \def\oti           {\,{\otimes}\,}
230: \def\Oti           {{\otimes}}
231: \def\otim          {\,{\otimes^-}\,}
232: \def\otip          {\,{\otimes^+}\,}
233: \def\otiP          {\,{\otimes^+}}
234: \def\parfu         {partition function}
235: \def\PHI           {{[\vPhi]}}
236: \newcommand\ppmatrix[4]{\mbox{{\Large(}$\!\!\begin{array}{cc}{}\\[-1.8em]
237:                    \scs #1\!&\!\!\scs #2\\[-.39em]\scs #3\!&\!\!\scs #4
238:                    \\[-.19em]\eear\!\!${\Large)}}}
239: \newcommand\ppvector[2]{\mbox{{\Large(}$\!\!\begin{array}{cc}{}\\[-1.8em]
240:                    \!\scs #1\!\\[-.39em]\!\scs #2\!\\[-.19em]\eear\!\!${\Large)}}}
241: \def\q             {quantum }
242: \def\Q             {Quantum }
243: \def\qed           {\hfill\checkmark\medskip}
244: \def\qft           {quantum field theory}
245: \def\qg            {quantum group}
246: \long\def\query#1{\hskip 0pt{\vadjust{\everypar={}\small\vtop to 0pt{\hbox{}%
247:      \vskip -13pt\rlap{\hbox to 47.0pc{\hfil{\vtop{\hsize=8pc\tolerance=6000%
248:      \hfuzz=.5pc\rightskip=0pt plus 3em\noindent#1}}}}\vss}}}}%
249: \newcommand\R[5]   {{\sf R}^{(#1\,#2)\,#3}_{#4\,#5}}
250: \def\Rangle        {\rangle}
251: \def\RCA           {{\mathrm{R}}} % representation of the chiral algebra \CA
252: \def\rcfts         {rational conformal field theories}
253: \def\rep           {representation}
254: \def\Rep           {{\mathcal R}ep}
255: \newcommand\rhobas[8]{\rho_{(#1#2)(#3#4)}^{#8\,(#5#6);#7}}
256: \def\rhs           {right hand side}
257: \def\RIsu          {Riemann surface}
258: \newcommand\Rm[5]  {{\sf R}^{-\,(#1\,#2)\,#3}_{\;#4\,#5}}
259: \def\rmr           {{\rm r}}
260: \newcommand\Rp[5]  {{\sf R}^{+\,(#1\,#2)\,#3}_{\;#4\,#5}}
261: \newcommand\Rpm[5] {{\sf R}^{\pm\,(#1\,#2)\,#3}_{\;#4\,#5}}
262: \newcommand\rbas[4]{\rho_{#2\,,\,#3}^{#1\, #4}}
263: \def\reals         {\mbox{$\mathbb R$}}
264: \def\rmm           {{\rm m\ddot o}}
265: \def\rmM           {{\rm M\ddot o}}
266: \def\rmT           {{\rm T}}
267: \newcommand\RP     {\mathbb{RP}^3}
268: \def\RR            {{\sf R}}
269: \newcommand\Rs[4]  {{\sf R}^{#1\,(#2\,#3)\,#4}}
270: \newcommand\Rss[3] {{\sf R}^{(#1\,#2)\,#3}}
271: \def\SCA           {{\mathrm{S}}} % representation of the chiral algebra \CA
272: \def\scs           {\scriptstyle}
273: \newcommand\sect[1]{\section{#1}\setcounter{equation}0\setcounter{defthm}0}
274: \def\sigrp         {\sigma_{\mathbb{RP}^2}}
275: \def\slz           {\mbox{$\mathfrak{sl}(2)$}}
276: \def\smallone      {{\bf1}}
277: \def\ssFA          {symmetric special Frobenius algebra}
278: \def\sss           {\scriptscriptstyle}
279: \def\Stn           {S^3{}_{\!\!\!\! {\rm n}.}}
280: \def\stt           {string theory}
281: \def\stts          {string theories}
282: \def\su            {\mathfrak{su}}
283: \def\susic         {supersymmetric }
284: \def\sym           {symmetry}
285: \def\tcs           {tensor categories}
286: \newcommand\Tfrac[2]{\textstyle{\frac{#1}{#2}}}
287: \def\tft           {topological field theory}
288: \def\theequation   {\thesection.\arabic{equation}}
289: \def\Thetha        {\varTheta}
290: \def\tildeF        {\tilde{\ff}}
291: \def\Times         {\,{\times}\,}
292: \def\To            {\,{\to}\,}
293: \def\torus         {{\rm T}}
294:    %%%  \def\top        {_{\rm top}}  % already used.....
295: \def\tr            {{\rm tr}\,}
296: \newcommand\Tr[1]  {{\rm tr}_{#1}\,}
297: \def\twodim        {two-di\-men\-si\-o\-nal}
298: \def\U             {\,}
299: \def\V             {\mbox{$\mathcal V$}}
300: \def\Vect          {{\mathcal V}\mbox{\sl ect}}
301: \def\Vee           {^{\vee}}
302: \def\vheta         {\vartheta}
303: \def\vhi           {\vPhi}
304: \def\Vir           {{\mathrm{Vir}}}
305: \newcommand\VO[5]  {{\mathrm V}^{#1}_{#2#3}(#4;#5)}
306: \def\voa           {vertex operator algebra}
307: \newcommand\VOb[6] {{\mathrm V}^{#1,#4}_{#2#3}(#5;#6)}
308:  %\newcommand\void[1]{} %%
309: \def\vop           {vertex operator}
310: \def\VOs           {{\mathrm V}}
311: \def\vPhi          {\varPhi}
312: \def\vsi           {\hatpsi}
313: \def\WZ            {Wess\hy Zu\-mi\-no }
314: \def\WZW           {Wess\hy Zu\-mi\-no\hy Wit\-ten }
315: \def\wzwm          {WZW model}
316: \def\wzwt          {WZW theory}
317: \newcommand\x[4]   {\lambda_{(#1,#2) #3}^{\,#4}}
318: \def\X             {{\rm X}}
319: \newcommand\xlabp[2]{}
320: \newcommand\Xlabp[3]{}
321: \newcommand\xx[4]  {\lambda_{(#1,#2) #3}^{}}
322: \newcommand\y[4]   {\Upsilon^{(#1,#2) #3}_{\bar{#4}}}
323: \newcommand\yy[4]  {\Upsilon^{(#1,#2) #3}_{}}
324: \newcommand\Zdfct[4]{Z^{#1|#2}_{#3 #4}}
325: \def\zet           {\ensuremath{\mathbb Z}}
326: 
327: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
328: 
329: \documentclass[12pt]{article}
330: \renewcommand\theequation{\thesection.\arabic{equation}}
331: \usepackage{latexsym,amsmath,amssymb,amsfonts,epsf,color,colordvi}
332: \usepackage{bbm}
333: \newcommand{\shadowbox}[1]{\fbox{#1}}
334: \usepackage[dvips]{graphics} 
335: \usepackage[mathscr]{eucal}
336: \usepackage{rotating}
337: 
338: \setlength{\textwidth}{17cm} \hoffset -20mm
339: \setlength{\textheight}{23.3cm} \topmargin= -14mm
340: \newtheorem{defthm}{\whattheorem}[section]
341: \newcommand\dt[1]  {\noindent\def\whattheorem{#1}\pagebreak[0]\begin{defthm}{}%
342:                    \samepage{$\!\!${\rm:}\nopagebreak\\[-1.91em]{}}\end{defthm}}
343: \newcommand\dtl[2] {\noindent\def\whattheorem{#1}\pagebreak[0]\begin{defthm}{}%
344:                    \samepage{$\!\!${\rm:}\label{#2}\nopagebreak\\[-1.91em]{}}%
345:                    \end{defthm}}
346: 
347: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
348: 
349: \begin{document}
350: 
351: \begin{flushright}  {~} \\[-12mm]
352: {\sf hep-th/0412290}\\[1mm]{\sf AEI-2004-124}\\[1mm]
353: {\sf Hamburger$\;$Beitr\"age$\;$zur$\;$Mathematik$\;$Nr.$\;$209}\\[1mm]
354: {\sf December 2004} \end{flushright}
355: \begin{center} \vskip 13mm
356: {\Large\bf TFT CONSTRUCTION OF RCFT CORRELATORS}\\[4mm]
357: {\Large\bf IV: STRUCTURE CONSTANTS}\\[4mm]
358: {\Large\bf AND CORRELATION FUNCTIONS}\\[18mm]
359: {\large
360: J\"urgen Fuchs$\;^1$ \ \ \ Ingo Runkel$\;^2$ \ \ \ Christoph Schweigert$\;^{3}$}
361: \\[8mm]
362: $^1\;$ Institutionen f\"or fysik, \ Karlstads Universitet\\
363: Universitetsgatan 5, \ S\,--\,651\,88\, Karlstad\\[5mm]
364: $^2\;$ Max-Planck-Institut f\"ur Gravitationsphysik,
365: Albert-Einstein-Institut\\
366: Am M\"uhlenberg 1, \ D\,--\,14\,476\, Golm\\[5mm]
367: $^3\;$ Fachbereich Mathematik, \ Universit\"at Hamburg\\
368: Schwerpunkt Algebra und Zahlentheorie\\
369: Bundesstra\ss e 55, \ D\,--\,20\,146\, Hamburg
370: \end{center}
371: \vskip 12mm
372: 
373: \begin{quote}{\bf Abstract}\\[1mm]
374: We compute the fundamental correlation functions in two-dimensional rational
375: conformal field theory, from which all other correlators can be obtained by 
376: sewing: the correlators of three bulk fields on the sphere,
377: one bulk and one boundary field on the disk, three boundary fields on the disk,
378: and one bulk field on the cross cap. We also consider conformal defects and
379: calculate the correlators
380: of three defect fields on the sphere and of one defect field on the cross cap.
381: \\
382: Each of these correlators is presented as the product of a structure constant 
383: and the appropriate conformal two- or three-point block. The structure 
384: constants are expressed as invariants of ribbon graphs in three-manifolds.
385: 
386: \end{quote} \vfill \newpage
387: 
388: \tableofcontents
389: 
390: % \newpage
391: 
392: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
393: 
394: \sect{Introduction and summary}
395: 
396: 
397: This paper constitutes part IV of the series announced in \cite{fuRs}, in which
398: the relation \cite{witt27,frki2,fffs3} between rational two-dimensional \cft\
399: (RCFT) and three-dimensional topological quantum field theory (TFT) is 
400: combined with non-commutative algebra in modular tensor categories to obtain 
401: a universal, model independent, construction of CFT correlation functions.
402: In the previous parts of the series, we concentrated on the description of 
403: correlators without field insertions, i.e.\ partition functions \cite{fuRs4},
404: on the special features that arise when the world sheet is unoriented
405: \cite{fuRs8}, and on the case that the CFT is obtained as a simple current
406: construction \cite{fuRs9}. In the present part we give the fundamental 
407: correlation functions of the CFT, i.e.\ a finite set of correlators from 
408: which all others can be obtained \cite{sono2,lewe3,fips} via sewing
409: and chiral Ward identities.
410: 
411: 
412: \subsection{Correlators from sewing of world sheets}\label{sec:in-1}
413: 
414: The correlators of a CFT should be single-valued functions of the world sheet 
415: moduli, like the positions of the field insertions or, more generally, the 
416: world sheet metric. Further, the correlators are subject to factorisation, or 
417: sewing, constraints. These constraints arise when a world sheet is cut along a
418: circle or an interval and a sum over intermediate states is inserted on the cut 
419: boundaries. This procedure allows one to express a correlator on a complicated 
420: world sheet (e.g.\ of higher genus) in terms of simpler building blocks. 
421: The fundamental building blocks are \cite{sono2,lewe3,fips} the correlators 
422: \BBB\ for three bulk fields on the sphere,
423: \Bb\ for one bulk field and one boundary field on the disk,
424: \bbb\ for three boundary fields on the disk, and
425: \BX\ for one bulk field on the cross cap. Actually, the requirement of 
426: covariance under global conformal transformations determines each of these 
427: correlators up to the choice of a non-zero vector in the relevant coupling 
428: space. The components of these vectors are called {\em structure constants\/}.
429: 
430: Depending on what class of surfaces one allows as world sheets for the CFT,
431: different sets of these
432: fundamental correlators (or, equivalently, structure constants) are required:
433: \\[-1.1em]
434: \begin{center}
435: \begin{tabular}{rll}
436:        \multicolumn2l {class of world sheets}
437:      & \multicolumn1l {fundamental correlators} \\[4pt]
438:  (1) & oriented, with empty boundary                 & \BBB            \\[2pt]
439:  (2) & unoriented, with empty boundary               & \BBB, \BX       \\[2pt]
440:  (3) & oriented, with empty or non-empty boundary    & \BBB, \Bb, \bbb \\[2pt]
441:  (4) & unoriented, with empty or non-empty boundary  & \BBB, \Bb, \bbb, \BX
442: \end{tabular}  
443: \end{center}
444: 
445: \smallskip
446: 
447: The factorisation of a given correlator into the fundamental building blocks
448: is not unique. Requiring that the different ways to express a correlator
449: in terms of sums over intermediate states agree gives rise to an infinite
450: system of constraints for the data \BBB, \Bb, \bbb, \BX. 
451:  %
452: One way to define a CFT is thus to provide a collection of candidates for the 
453: fundamental correlators that are required in (1)\,--\,(4) and show that they 
454: solve these factorisation constraints and lead to single-valued correlators. 
455: 
456: \medskip
457: 
458: Holomorphic bulk fields (like the component $T$ of the stress tensor) lead
459: to conserved charges and chiral Ward identities, which constrain the form of the
460: correlation functions. Solutions to the chiral Ward identities are called
461: conformal blocks. In rational CFT the space of conformal blocks for a given 
462: correlator is finite-dimensional. 
463: Another important consequence of rationality is that the fundamental correlators
464: have to be given only for a finite collection of fields; for all other
465: fields they are then determined by the chiral Ward identities.
466: 
467: The conformal blocks that are relevant for the correlators on a
468: given world sheet \X\ are actually to be constructed on the double $\hat\X$ of
469: \X. The surface $\hat\X$ is obtained as the orientation bundle over $\X$, 
470: divided by the equivalence relation that identifies the two possible 
471: orientations above points on the boundary of $\X$: 
472:   \be
473:   \hat\X = {\rm Or}(\X){/}{\sim} \qquad {\rm with}
474:   \qquad (x,{\rm or}) \sim (x,-{\rm or}) ~{\rm for}~ x \iN \partial\X \,.
475:   \labl{eq:double-def}
476: For example, if $\X$ is orientable and has empty boundary, then $\hat \X$
477: just consists of two copies of $\X$ with opposite orientation.
478: A correlator $C(\X)$ on the world sheet $\X$ is an element in the space
479: of conformal blocks on the double,
480:   \be 
481:   C(\X) \in \calh(\hat \X) \,.
482:   \labl{eq:CX-HX}
483: This is known as the principle of {\em holomorphic factorisation} \cite{witt39}.
484: 
485: The conformal blocks on a surface $\Sigma$ are generically multivalued functions
486: of the moduli of $\Sigma$. Upon choosing bases in the spaces $\calh(\Sigma)$, 
487: the behaviour of the blocks under analytic continuation and factorisation can 
488: be expressed through braiding and fusing matrices $\BBmat$ and $\FF$ 
489: \cite{Mose,frki,bakI}.
490: These matrices are in fact already determined by the four-point blocks on the 
491: Riemann sphere. One can then express the constraints on the fundamental
492: correlators coming from factorisation and single-valuedness in terms of 
493: fusing and braiding matrices. It turns out that there is a {\em finite\/} set of
494: equations which the structure constants must obey in order to
495: yield a consistent CFT. These are the so-called {\em sewing constraints\/} 
496: \cite{sono2,lewe3}.
497: 
498: \medskip
499: 
500: The first class of CFTs for which explicit solutions for the structure
501: constants have been calculated are the A-series Virasoro minimal models.
502: The constants for \BBB, \Bb, \bbb\ and \BX\ are given in 
503: \cite{dofa2,cale,runk,fips}, respectively. By now many more solutions are 
504: known; pertinent references will be given at the end of sections 
505: \ref{sec:rc-bnd3pt}\,--\,\ref{sec:rc-1bulk-xcap}, in which the calculation 
506: of the individual structure constants in the TFT framework is carried out.
507: 
508: Remarkably, one can also obtain exact expressions for some of the fundamental 
509: correlation functions in a few examples of non-rational CFTs in which the 
510: chiral algebra $\CA$ is too small to make all spaces of conformal blocks 
511: finite-dimensional. The most complete answers are known for Liouville theory, 
512: see e.g.\ \cite{pons3,tesc12,naKA'} and references cited therein. The structure 
513: of the expressions found in this case is very similar to the Cardy case of a 
514: rational CFT. This similarity might serve as a guideline when trying to extend 
515: aspects of the TFT-based construction beyond the rational case.
516: 
517: 
518: \subsection{Correlators from 3-d TFT and algebras in tensor categories}
519: 
520: The sewing constraints form a highly overdetermined system of polynomial
521: equations, which is difficult to solve directly. Instead, one can try to
522: identify an underlying algebraic structure that encodes all information
523: about a solution. In our TFT-based construction, this underlying structure
524: is an algebra $A$ in a certain tensor category.
525: The associativity of this algebra is in fact a sewing constraint, applied
526: to a disk with four insertions on the boundary. This turns out to be the
527: only non-linear constraint to be solved in the TFT approach; all other
528: relevant equations are linear.
529: 
530: We will consider CFTs that can be defined on world sheets of type (3)
531: or type (4). Each of these two cases requires slightly different properties of
532: the algebra $A$, but they can, and will, be treated largely in parallel.
533: 
534: \medskip\noindent
535: 
536: Our construction of CFT correlators takes two pieces of data as an input:
537: 
538: \Itemize
539: \item[(i)~] The {\em chiral algebra} $\CA$. 
540: \\
541: This fixes the minimal amount of symmetry present in the CFT to be constructed. 
542: More specifically, the holomorphic (and anti-holomorphic) bulk fields contain 
543: as a subset the fields in $\CA$, and all boundary conditions and defect lines 
544: preserve $\CA$ (with trivial gluing automorphism). 
545: \end{itemize}
546: %
547: We demand $\CA$ to be a rational conformal vertex algebra and denote by 
548: ${\Rep}(\CA)$ its representation category. The properties\,\footnote{%
549:   ~To be precise, $\CA$ must obey certain conditions on its
550:   homogeneous subspaces, be self-dual, have a semi-simple representation
551:   category, and fulfill Zhu's $C_2$ cofiniteness condition \cite{huan21}.} 
552: of $\CA$ then imply that ${\Rep}(\CA)$ is a modular tensor category
553: (see section I:2.1 for definition, references and our notational conventions).
554:  %
555: To any modular tensor category \calc\ one can associate a three-dimensional 
556: topological field theory TFT($\calc$), see e.g.\ \cite{TUra,BAki} as well as
557: section I:2.4 and section 3.1 below. The TFT supplies two assignments.
558: The first, $E \,{\mapsto}\, \calh(E)$ associates to an (extended) surface
559: $E$ the space $\calh(E)$ of states, a finite-dimensional vector space. The 
560: second assignment takes
561: a cobordism $E{\overset{\M}{\longrightarrow}}E'$, i.e.\ a three-manifold
562: with `in-going' boundary $E$ and `out-going' boundary $E'$ together with
563: an embedded ribbon graph, and assigns to it a linear map
564: $Z(\M){:}\; \calh(E)\To\calh(E')$. In the case of Chern\hy Simons theory, the 
565: state spaces of the TFT can be identified with the spaces of conformal blocks 
566: of the corresponding WZW model \cite{witt27}. 
567: In the case of three points on the sphere, for a general RCFT the
568: identification of the state spaces of the TFT with spaces of conformal
569: blocks is a consequence of the definition of the tensor product;
570: for more general situations it follows by the hypothesis of factorisation.
571: 
572: According to the principle \erf{eq:CX-HX}, in order to specify a CFT we must assign
573: to every world sheet $\X$ a vector $C(\X)$ in $\calh(\hat \X)$. This will be
574: done with the help of TFT($\calc$), by specifying for every $\X$ a cobordism 
575: $\emptyset\,{\overset{\MX\,}{\longrightarrow}}\,\hat \X$ and setting 
576: $C(\X)\,{:=}\, Z(\MX)$. Since $\calc \eq {\Rep}(\CA)$ encodes all the monodromy
577: properties of the conformal blocks, the question whether the collection $\{C(\X)\}$
578: of candidate correlators satisfies the requirements
579: of single-valuedness and factorisation can be addressed solely at the level of 
580: the category \calc. In particular, the analysis does not require the knowledge 
581: of the explicit form of $\CA$ or of the conformal blocks. The first input
582: in the TFT construction is thus
583:    \def\leftmargini{2.3em}
584:    \begin{itemize}
585:    \item
586: a modular tensor category $\calc$ 
587:    \end{itemize}
588: which determines in particular also TFT($\calc$). 
589: 
590: For the construction 
591: of the vector $C(\X) \iN \calh(\hat \X)$ it is actually not necessary that
592: \calc\ is the representation category of a vertex algebra; only
593: the defining properties of a modular tensor category are used. 
594: The vertex algebra and its conformal blocks are relevant only when
595: one wishes to determine the correlators as actual {\em functions\/}.
596: In particular, in those cases where non-isomorphic vertex algebras have
597: equivalent representation categories (as it happens e.g.\ for WZW theories
598: based on $\mathfrak{so}(2n{+}1)$ at level 1 with $n$ taking different values 
599: congruent modulo 8), the TFT approach yields the
600: same link invariants in the description of the correlators. It is only after
601: substituting multivalued functions for the conformal blocks that two
602: such models have different correlators as actual functions of moduli and
603: insertion points.
604: 
605: \Itemize
606: \item[(ii)] An {\em algebra} $A$ in \calc.
607: \\
608: There are generically several distinct CFTs which share the same chiral algebra
609: $\CA$, the best known example being the A-D-E-classification of minimal models 
610: \cite{caiz2}. The additional datum we need to specify the CFT uniquely is an 
611: algebra $A$ in the tensor category $\calc$. 
612: \end{itemize}
613: More precisely, what is needed is
614:    \def\leftmargini{2.3em} \begin{itemize}
615:    \item
616: a {\em symmetric special Frobenius algebra\/} (see section I:3)
617: for obtaining a CFT defined on world sheets of type (3);
618:    \item
619: a {\em Jandl algebra\/}, i.e.\ a symmetric special Frobenius algebra
620: together with a {\em reversion\/}\\
621: (see section II:2), for obtaining a CFT defined on world sheets of type (4).
622:    \end{itemize}
623: At the level of the CFT, specifying a symmetric special Frobenius algebra
624: corresponds to providing
625: the structure constants for three boundary fields on the disk (denoted \bbb\
626: above) for a single boundary condition. To obtain a Jandl algebra, one 
627: must \cite{fuRs8} in addition provide a reversion (a braided analogue of an 
628: involution) on the boundary fields that preserve the given boundary condition.
629: 
630: \medskip
631: 
632: Suppose now we are given such a pair $(\calc , A)$ of data. The correlators
633: $C_\AA$ of the conformal field theory CFT($A$) associated to this pair
634: are obtained as follows. Given a world sheet $\X$, one constructs a particular
635: cobordism $\MX$, the {\em connecting manifold\/} \cite{fffs3}
636: together with a ribbon graph $R_\X$ embedded in $\MX$. The different
637: ingredients of $\X$, like field insertions, boundary conditions and
638: defect lines, are represented by specific parts of the ribbon
639: graph $R_\X$, and also the algebra $A$ enters in the definition of $R_\X$; 
640: this will be described in detail in section \ref{sec:fi}. 
641: Next one uses TFT($\calc$) to define the correlator of CFT($A$) on $\X$
642: as $C_\AA(\X) \,{:=}\, Z(\MX)$.
643: 
644: One can prove \cite{fjfrs} that the correlators supplied by the assignment
645: $\X \,{\mapsto}\, C_\AA(\X)$ are single-valued and consistent with 
646: factorisation or, equivalently, with its inverse operation, sewing. This 
647: implies in particular that the fundamental correlators
648: \BBB, \Bb, \bbb\ and \BX\ obtained in this manner solve the sewing constraints.
649: 
650: We conjecture that, conversely, {\em every\/} CFT build from conformal blocks
651: of the rational vertex algebra $\CA$ defined on world sheets of types (3) and 
652: (4) can be obtained from a suitable pair $(\calc,A)$ with $\calc\eq{\Rep}(\CA)$.
653: This converse assignment is not unique: several different algebras lead to one
654: and the same CFT. This corresponds to the freedom of choosing the particular 
655: boundary condition whose structure constants are used to define the algebra
656: structure. All such different algebras are, however, Morita equivalent.
657: 
658: \medskip
659: 
660: For theories defined on world sheets of type (3), it has been advocated in
661: \cite{bppz2,pezu6,pezu7,pezu8} that aspects of the
662: fundamental correlators are captured by the structure of a
663: weak Hopf algebra, respectively an Ocneanu double triangle algebra. Indeed,
664: a weak Hopf algebra furnishes (see \cite{ostr}, and e.g.\ \cite{ffrs4} for a 
665: review) a non-canonical description of a module category over a tensor category,
666: i.e.\ of the situation that in our TFT-based construction is
667: described by the algebra $A$ in the tensor category ${\Rep}(\CA)$.
668: 
669: 
670: \subsection{Plan of the paper}
671: 
672: In section \ref{sec:mb} we introduce the fusing matrices for the categories 
673: \calca\ of $A$-modules and \calcaa\ of $A$-bimodules. These are useful for
674: obtaining compact expressions for the structure constants. In section
675: \ref{sec:fi} the ribbon graph representation of field insertions
676: is described. We will consider three kinds of fields: boundary fields,
677: bulk fields, and defect fields. By a boundary field we mean either
678: a field that lives on a boundary with a given boundary condition or
679: a field that changes the boundary condition. Likewise, a defect field
680: can either live on a defect line of a given type, or change the defect type. 
681: A bulk field can thus be regarded as a special type of defect field,
682: namely one that connects the invisible defect to the invisible defect.
683: 
684: In section \ref{sec:rc} the ribbon invariants for the fundamental
685: correlators \BBB, \Bb, \bbb\ and \BX\ are computed, as well as the correlators
686: \DDD\ of three defect fields on the sphere, and \DX\ of one defect field on the
687: cross cap.  Up to this point the discussion proceeds entirely at the level of 
688: the tensor category \calc, i.e.\ of the monodromy data of the conformal blocks. 
689: 
690: We would also like to obtain the fundamental correlators as explicit functions 
691: of the field insertion points. The necessary calculations, which are presented 
692: in section \ref{sec:cf}, require additional input beyond the level of \calc, 
693: namely the vertex algebra $\CA$ with representation category ${\Rep}
694: (\CA)$ and the notions of intertwiners between $\CA$-\rep s and of conformal 
695: blocks, as well as the isomorphism between the state spaces of the TFT and 
696: the spaces of conformal blocks. These additional ingredients are the subject 
697: of section \ref{sec:cb}. Unlike the mathematical prerequisites employed in the 
698: earlier sections \ref{sec:mb}\,--\,\ref{sec:rc}, some of this machinery is not 
699: yet fully developed in the literature. However, in section \ref{sec:cf} we 
700: only need the two- and three-point conformal blocks, which are well understood.
701: 
702: \medskip
703: 
704: The formulas which give our results for correlators of boundary fields
705: $\varPsi$, bulk fields $\vhi$ and defect fields $\Thetha$
706: are listed in the following table.  
707: 
708: \medskip
709: 
710: {}~~~\begin{tabular}{ccccc}
711:            &   Ribbon  & Ribbon invariant & $\!\!$Ribbon invariant$\!\!$
712:            & Correlation \\[-2pt]
713: Correlator &~~invariant~~& in a basis & $\!\!$in Cardy case $A\eq\one\!\!$
714:            & function \\[3pt]
715: \hline \\[-3pt]
716: \bbb  & eq.~~\erf{eq:uhp-cobord}         & eq.~\erf{eq:3bnd-inv-basis} & 
717:         eq.~\erf{eq:3bnd-basis-cardy}    & eq.~\erf{eq:3bnd-corr} \\[4pt]
718: %       
719: \Bb   & eq.~\erf{eq:1bulk1bnd-rib-2}     & eq.~\erf{eq:1B1b-basis} & 
720:         eq.~\erf{eq:1B1b-bas-Cardy}      & eq.~\erf{eq:1bulk-1bnd} \\[4pt]
721: %
722: \BBB  & eq.~\erf{eq:3bulk-rib}           & eq.~\erf{eq:3bulk-coeff} & 
723:         eq.~\erf{eq:3bulk-basis-Cardy}   & eq.~\erf{eq:3bulk}      \\[4pt]
724: %       
725: \DDD  & eq.~\erf{eq:3defect-rib}         & eq.~\erf{eq:3def-basis} &
726:         eq.~\erf{eq:3def-basis-Cardy}    & eq.~\erf{eq:3defect}    \\[4pt]
727: %
728: \BX   & eq.~\erf{eq:xcap-ribbon-emb}     & eq.~\erf{chatphic} &
729:         eq.~\erf{eq:cardy-1bulk-xcap}    & eq.~\erf{eq:1B-xcap}    \\[4pt]
730: %
731: \DX   & eq.~\erf{eq:xcap-def-rib}        & eq.~\erf{eq:1D-Xcap-basis} &
732:         eq.~\erf{eq:1def-Xcap-bas-Cardy} & eq.~\erf{eq:1D-xcap} 
733: \\{}
734: \end{tabular}
735: 
736: \medskip
737: 
738: \noindent
739: {\bf Acknowledgements.}
740: We are indebted to N.~Potylitsina-Kube for her skillful help with the numerous 
741: illustrations. J.F.\ is supported by VR under project no.\ 621--2003--2385,
742: and C.S.\ is supported by the DFG project SCHW 1162/1-1.
743: 
744: 
745: \newpage
746: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
747: 
748: \sect{Fusing matrices for modules and bimodules}\label{sec:mb}
749: 
750: 
751: To exhibit the mathematical structure encoded by the bulk and  boundary
752: structure constants, we make use of the concept of module 
753: \cite{pare11,pare16,beRn,ostr} and bimodule categories. As the main focus
754: of this paper lies on the computation of correlators,  we do not develop
755: the mathematical formalism in all detail.
756: 
757: Given a tensor category $\calc$, a {\em right module category\/} over 
758: $\calc$ comes by definition with a prescription
759: for tensoring objects of $\calm$ with objects of $\calc$ from the right,
760:   \be
761:   \otimes_\calm :\quad \calm \times \calc \rightarrow \calm \,.  \ee
762: The tensor product $\otimes_\calm$ must be compatible with the tensor product 
763: $\otimes$ of $\calc$ in the sense that there are associativity isomorphisms
764:   \be
765:   (M\otimes_\calm U) \otimes_\calm V
766:   \,\cong\, M\otimes_\calm (U \oti   
767:   V)
768:   \ee
769: for $U,V \iN \objc$ and $M\iN\obj(\calm)$, such that the corresponding 
770: pentagon identity is satisfied, as well as unitality isomorphisms 
771: $M\,{\otimes_\calm}\,\one\,{\cong}\, M$ obeying appropriate triangle identities.
772: 
773: A {\em bimodule category\/} $\calb$ is a straightforward extension 
774: of this concept, allowing both for a left action of a tensor category $\calc$
775: and a right action of a possibly different tensor category $\cald$,
776:   \be
777:   \otimes_\calb^l :\quad \calc \times \calb \rightarrow \calb 
778:   \qquad {\rm and} \qquad
779:   \otimes_\calb^r :\quad \calb \times \cald \rightarrow \calb  \,.
780:   \ee
781: The compatibility conditions of the different tensor products are
782: given by associativity isomorphisms
783:   \be\begin{array}{ll}
784:   U \otimes_\calb^l ( V \otimes_\calb^l X) \,\cong\,
785:     ( U \,{\otimes_{\mathcal{C}}}\, V) \otimes_\calb^l X
786:     \qquad & {\rm (left~action)} \,,\\[5pt]
787:   (X\otimes_\calb^r R) \otimes_\calb^r S \,\cong\,
788:     X\otimes_\calb^r (R \,{\otimes_\cald}\, S)
789:     \qquad & {\rm (right~action)} \,,\\[5pt]
790:   U \otimes_\calb^l ( X \otimes_\calb^r R) \,\cong\,
791:     ( U \otimes_\calb^l X) \otimes_\calb^r R 
792:     \qquad & {\rm (left~and~right~action~commute)} 
793:   \eear\ee
794: for $U,V \iN \objc$, $R,S \iN \obj(\cald)$ and $X \iN \obj(\calb)$. Again the 
795: various associators must fulfill the corresponding pentagon identities,
796: and again there are also unitality isomorphisms satisfying triangle identities.
797: In our application it 
798: is natural to take the category $\calb$ to be a tensor category as well; then 
799: there are also the corresponding additional associators and pentagon identities.
800: 
801: In section \ref{sec:rc} we will see that the boundary structure constants 
802: encode the associator of a module category, while the structure constants of 
803: bulk and defect fields give the associator of a bimodule category. In the 
804: present section we define the expansion of these associators in a basis. 
805: This gives the fusing matrices for modules and bimodules.
806: 
807: 
808: \subsection{Fusing matrices for modules}
809: 
810: The right module category relevant for our purposes is the
811: category $\calca$ of left $A$-modules, where $A$ is an
812: algebra in a tensor category $\calc$ (see e.g.\ section I:4, or section 2.3 
813: of \cite{ffrs}; we take $\calc$ to be strict.) Given a left $A$-module
814: $M \eq (\dot M, \rho)$, we can tensor with any object $U$ of $\calc$ from 
815: the right so as to obtain again a (generically different) left $A$-module
816: $M\oti U \eq (\dot M \oti U, \rho \oti \id_U)$ (with all tensor
817: products taken in $\calc$). This defines a right action
818: $\otimes{:}\; \calca \,{\times}\, \calc \To \calca$.
819: 
820: We take $A$ to be a symmetric special Frobenius algebra and $\calc$ to be a
821: modular tensor category. Then also $\calca$ is semisimple and has a finite 
822: number of non-isomorphic simple objects \cite{fuSc16,kios}.
823: Let $\{ U_i \,|\, i\iN\II \}$ be a set of representatives of isomorphism
824: classes of simple objects of $\calc$ and $\{ M_\kappa \,|\, \kappa\iN\JJ \}$ 
825: a set of representatives of isomorphism classes of simple left $A$-modules. 
826: Thus lower case roman letters $i,j,k,...$ appear as labels for simple
827: objects $U_i, U_j, U_k, ... $ of $\calc$, while lower case greek letters 
828: $\mu,\nu,\kappa, ... $ from the upper range of the alphabet 
829: appear as labels for simple $A$-modules $M_\mu, M_\nu, M_\kappa, ... $ 
830: (later on, we will use such greek letters also for simple bimodules).
831: 
832: As a first step towards defining the fusing matrices for $\calca$
833: we choose bases in the spaces $\Hom_A(M_\mu \oti U_i, M_\nu)$,
834:   %  [ pic 4_59 ]
835:   \bea \begin{picture}(190,60)(0,33)
836:   \put(60,0)     {\includeourbeautifulpicture 59 }
837:   \put(0,41)     {$ \psi_{(\mu i)\nu}^{\alpha} \;:= $}
838:   \put(112,41)   {$ \in \Hom_A(M_\mu \oti U_i, M_\nu)  \;. $}
839:   \put(59,-8.2)  {\scriptsize$ M_\mu $} 
840:   \put(59,91.2)  {\scriptsize$ M_\nu $} 
841:   \put(68,44.8)  {\scriptsize$ \alpha $} 
842:   \put(80.6,-8.2){\scriptsize$ U_i $} 
843:   \epicture25 \labl{eq:MUM-basis} \xlabp 59
844: We do not restrict this choice of basis in any way,
845: except for $i\eq0$, i.e.\ when $U_i\eq\one$. In this
846: case we demand the basis element to be $\psi_{(\mu 0)\mu}\eq\id_{M_\mu}$.
847:     %  needs module-strictness
848: Both the basis morphisms \erf{eq:MUM-basis} and the basis morphisms 
849: $\lambda_{(ij)k}^{\alpha}\iN\Hom(U_i\oti U_j,U_k)$ that were chosen in 
850: (I:2.29) will be labelled by lower case greek letters 
851: $\alpha, \beta, \gamma, ... $, chosen from the beginning of the alphabet. 
852: 
853: We also need bases in the spaces $\Hom_A(M_\nu, M_\mu \oti U_i)$, which we 
854: denote by $\{ \bar\psi_{\alpha}^{(\mu i)\nu} \}$. We choose these bases 
855: such that they are dual to the $\psi_{(\mu i)\nu}^{\alpha}$ in the sense
856: that $\psi_{(\mu i)\nu}^{\alpha} \circ \bar\psi_{\beta}^{(\mu i)\nu}
857: {=}\, \delta_{\alpha,\beta}\, \id_{M_\nu}$. Just as for the morphisms
858: $\lambda_{(ij)k}^{\alpha}$ in (I:2.31), owing to the semisimplicity of $\calca$
859: the morphisms $\psi_{(\mu i)\nu}^{\alpha}$ obey the completeness relation
860:   %  [ pic 4_61 ]
861:   \bea \begin{picture}(190,60)(0,40)
862:   \put(55,0)     {\Includeourbeautifulpicture 61a }
863:   \put(130,0)    {\Includeourbeautifulpicture 61b }
864:   \put(0,47)     {$ \dsty\sum_{\kappa\in\JJ}\;\sum_{\alpha=1}^{\Ann{i\mu}{\kappa}} $}
865:   \put(54,-8.2)  {\scriptsize$ M_\mu $} 
866:   \put(54,102.5) {\scriptsize$ M_\mu $} 
867:   \put(61,46.5)  {\scriptsize$ M_\kappa $} 
868:   \put(63,37.9)  {\scriptsize$ \alpha $} 
869:   \put(63,57.3)  {\scriptsize$ \bar\alpha $} 
870:   \put(75.5,-8.2){\scriptsize$ U_i $} 
871:   \put(75.5,102.5){\scriptsize$ U_i $} 
872:   \put(100,47)   {$ = $}
873:   \put(126,-8.2) {\scriptsize$ M_\mu $} 
874:   \put(126,102.5){\scriptsize$ M_\mu $} 
875:   \put(144.3,-8.2){\scriptsize$ U_i $} 
876:   \put(144.3,102.5){\scriptsize$ U_i $} 
877:   \epicture26 \labl{pic61} \xlabp 61
878: Here we abbreviate the dimension of the morphism space $\Hom_A(M_\mu \Oti U_i,
879: M_\kappa)$ by $\Ann{i\mu}{\kappa}$; according to proposition I:5.22, 
880: these non-negative integers give the annulus coefficients of the CFT.
881: 
882: The fusing matrices $\GG[A]$ relate two different bases of the space
883: $\Hom_A(M_\mu\Oti U_i \Oti U_j,M_\nu)$. We denote these bases
884: by $\{ b^{(1)}_{\alpha\kappa\beta} \}$ and $\{ b^{(2)}_{\gamma k\delta} \}$, 
885: with the individual vectors given by
886:   %  [ pic 4_60 ]
887:   \bea \begin{picture}(330,60)(0,48)
888:                  \put(56,0){ \begin{picture}(0,0)
889:   \put(0,0)      {\Includeourbeautifulpicture 60a }
890:   \put(-10.4,46.5){\scriptsize$ M_{\kappa}^{} $} 
891:   \put(-1,-8.2)  {\scriptsize$ M_\mu $} 
892:   \put(-1,102.5) {\scriptsize$ M_\nu $} 
893:   \put(12,32.8)  {\scriptsize$ \alpha $} 
894:   \put(12,60)    {\scriptsize$ \beta $} 
895:   \put(20.5,-8.2){\scriptsize$ U_i $} 
896:   \put(40.8,-8.2){\scriptsize$ U_j $} 
897:                  \end{picture} }
898:                  \put(246,0){ \begin{picture}(0,0)
899:   \put(0,0)      {\Includeourbeautifulpicture 60b }
900:   \put(-1,-8.2)  {\scriptsize$ M_\mu $} 
901:   \put(-1,102.5) {\scriptsize$ M_\nu $} 
902:   \put(12,68.3)  {\scriptsize$ \gamma $} 
903:   \put(31.9,48.9){\scriptsize$ \delta $} 
904:   \put(16.5,-8.2){\scriptsize$ U_i $} 
905:   \put(19.2,57.5){\scriptsize$ U_k $} 
906:   \put(36.5,-8.2){\scriptsize$ U_j $} 
907:                  \end{picture} }
908:   \put(0,49)     {$ b^{(1)}_{\alpha\kappa\beta} \;:= $}
909:   \put(148,49)   {and $\qquad b^{(2)}_{\gamma k\delta} \;:= $}
910:   \epicture35 \labl{pic60} \xlabp 60
911: Let us verify that these are indeed bases. Linear independence can be seen by 
912: composing with the corresponding dual basis elements. Completeness of 
913: $\{ b^{(1)}_{\alpha\kappa\beta} \}$ is seen by applying \erf{pic61} twice.  
914: A similar reasoning establishes completeness of $\{b^{(2)}_{\gamma k\delta}\}$.
915: Also note that there are $\sum_\kappa \Ann{i \mu}{\kappa} \Ann{j\kappa}{\nu}$
916: basis elements of type $b^{(1)}$ and $\sum_k \N ijk \Ann{k \mu}{\nu}$ basis 
917: elements of type $b^{(2)}$. That these two numbers are indeed equal has
918: been established in theorem I:5.20(v). In CFT, equality of the two sums means 
919: that the annulus coefficients furnish a NIMrep of the fusion rules 
920: \cite{prss3,bppz2,gann17}.
921: 
922: The fusing matrix of the module category $\calca$ describes a change of basis
923: from $b^{(1)}$ to $b^{(2)}$. There exists a unique set of numbers 
924: $\GG[A]^{(\mu ij)\,\nu}_{\alpha\kappa\beta,\gamma k\delta}$ such that
925:   %  [ pic 4_60 (zweite version) ]
926:   \bea \begin{picture}(280,59)(0,46)
927:                  \put(0,0){ \begin{picture}(0,0)
928:   \put(0,0)      {\Includeourbeautifulpicture 60a }
929:   \put(-10.4,46.5){\scriptsize$ M_{\kappa}^{} $}
930:   \put(-1,-8.2)  {\scriptsize$ M_\mu $}
931:   \put(-1,102.5) {\scriptsize$ M_\nu $}
932:   \put(12,32.8)  {\scriptsize$ \alpha $}
933:   \put(12,60)    {\scriptsize$ \beta $}
934:   \put(20.5,-8.2){\scriptsize$ U_i $}
935:   \put(40.8,-8.2){\scriptsize$ U_j $}
936:                  \end{picture} }
937:                  \put(220,0){ \begin{picture}(0,0)
938:   \put(0,0)      {\Includeourbeautifulpicture 60b }
939:   \put(-1,-8.2)  {\scriptsize$ M_\mu $}
940:   \put(-1,102.5) {\scriptsize$ M_\nu $}
941:   \put(12,68.3)  {\scriptsize$ \gamma $}
942:   \put(31.9,48.9){\scriptsize$ \delta $}
943:   \put(16.5,-8.2){\scriptsize$ U_i $}
944:   \put(19.2,57.5){\scriptsize$ U_k $}
945:   \put(36.5,-8.2){\scriptsize$ U_j $}
946:                  \end{picture} } 
947:   \put(65,46)    {$ =\; \dsty\sum_{k\in\II}\;
948:                  \sum_{\gamma=1}^{\Ann{k \mu}{\nu}}\; \sum_{\delta=1}^{\N ijk}\,
949:                  \GG[A]^{(\mu ij)\,\nu}_{\alpha\kappa\beta,\gamma k\delta} $}
950:   \epicture33 \labl{eq:GA-def} \xlabp 60
951: We denote the fusing matrix arising here by $\GG[A]$ rather than by
952: $\FF[A]$, because in the Cardy case $A\eq\one$ it reduces
953: to the inverse fusing matrix $\GG$ of $\calc$, see (I:2.40). By composing
954: \erf{eq:GA-def} with the basis of $\Hom_A(M_\nu,M_\mu\Oti U_i \Oti U_j)$ 
955: dual to $\{b^{(2)}_{\gamma k\delta}\}$,
956: one arrives at the following ribbon invariant for $\GG[A]$:
957:   %  [ pic 4_63 ]
958:   \bea \begin{picture}(200,56)(0,32)
959:   \put(66,0)     {\Includeourbeautifulpicture 63a }
960:   \put(120,0)    {\Includeourbeautifulpicture 63b }
961:   \put(0,38)     {$ \GG[A]^{(\mu ij)\,\nu}_{\alpha\kappa\beta,\gamma k\delta} $}
962:   \put(62,-8.2)  {\scriptsize$ M_\nu $}
963:   \put(62,85.2)  {\scriptsize$ M_\nu $}
964:   \put(90,38)    {$ = $}
965:   \put(110,31.5) {\scriptsize$ M_\mu $}
966:   \put(110,59.1) {\scriptsize$ M_\kappa $}
967:   \put(113.5,50.9){\scriptsize$ \alpha $}
968:   \put(113.8,67.9){\scriptsize$ \beta $}
969:   \put(113.8,11.3){\scriptsize$ \bar\gamma $}
970:   \put(119,-8.2) {\scriptsize$ M_\nu $}
971:   \put(119,85.2) {\scriptsize$ M_\nu $}
972:   \put(131,36.2) {\scriptsize$ U_i $}
973:   \put(136,22.5) {\scriptsize$ U_k $}
974:   \put(146,57.3) {\scriptsize$ U_j $}
975:   \put(152.3,31.3){\tiny$ \bar\delta $}
976:   \epicture26 \labl{eq:GA-rib-inv} \xlabp 63
977: A possibility to compute these numbers is to first find
978: all simple $A$-modules $M_\mu$ by decomposing induced modules (see
979: section I:4.3), which is a linear (and finite) problem. 
980: Next one needs to identify the spaces $\Hom_A(M_\mu \Oti U_i, M_\nu)$ as 
981: subspaces of $\Hom(\dot M_\mu \Oti U_i , \dot M_\nu)$. This can be done
982: by finding the eigenspaces of the projector $f \,{\mapsto}\, \overline f$
983: that takes a morphism $f \iN \Hom(\dot M_\mu \Oti U_i , \dot M_\nu)$
984: to its $A$-averaged morphism $\overline f$, which is known to lie
985: in the subspace $\Hom_A(M_\mu \Oti U_i, M_\nu)$, see definition I:4.3
986: and lemma I:4.4. This is again a linear problem. 
987: These ingredients suffice to evaluate \erf{eq:GA-rib-inv}.
988: 
989: By construction, the matrices $\GG[A]$ solve the pentagon relation
990:   %  [ pic 4_64 ]
991:   \bea \begin{picture}(420,150)(0,60)
992:   \put(0,55)     {\Includeourbeautifulpicture 64a }
993:   \put(0,55)     {\setlength{\unitlength}{.38pt}
994:                   \put(-5,12){\scriptsize$ \mu $} 
995:                   \put( 63, 5){\scriptsize$ i $} 
996:                   \put(152, 5){\scriptsize$ j $} 
997:                   \put(240, 5){\scriptsize$ k $} 
998:                   \put(120,203){\scriptsize$ \nu $} 
999:                   \put(55,71){\scriptsize$ \kappa $} 
1000:                   \put(99,115){\scriptsize$ \lambda $} 
1001:                   \setlength{\unitlength}{1pt}}
1002:   \put(150,125)  {\Includeourbeautifulpicture 64b }
1003:   \put(150,125)  {\setlength{\unitlength}{.38pt}
1004:                   \put(-5,12){\scriptsize$ \mu $} 
1005:                   \put( 80, 5){\scriptsize$ i $} 
1006:                   \put(152, 5){\scriptsize$ j $} 
1007:                   \put(240, 5){\scriptsize$ k $} 
1008:                   \put(120,203){\scriptsize$ \nu $} 
1009:                   \put(121,71){\scriptsize$ p $} 
1010:                   \put(99,115){\scriptsize$ \lambda $} 
1011:                   \setlength{\unitlength}{1pt}}
1012:   \put(310,125)  {\Includeourbeautifulpicture 64c }
1013:   \put(310,125)  {\setlength{\unitlength}{.38pt}
1014:                   \put(-5,12){\scriptsize$ \mu $} 
1015:                   \put( 80, 5){\scriptsize$ i $} 
1016:                   \put(152, 5){\scriptsize$ j $} 
1017:                   \put(240, 5){\scriptsize$ k $} 
1018:                   \put(120,203){\scriptsize$ \nu $} 
1019:                   \put(141,70){\scriptsize$ p $} 
1020:                   \put(160,120){\scriptsize$ q $} 
1021:                   \setlength{\unitlength}{1pt}}
1022:   \put(150,0)    {\Includeourbeautifulpicture 64d }
1023:   \put(150,0)    {\setlength{\unitlength}{.38pt}
1024:                   \put(-7,12){\scriptsize$ \mu $} 
1025:                   \put( 72, 5){\scriptsize$ i $} 
1026:                   \put(157, 5){\scriptsize$ j $} 
1027:                   \put(240, 5){\scriptsize$ k $} 
1028:                   \put(120,203){\scriptsize$ \nu $} 
1029:                   \put(81, 99){\scriptsize$ \kappa $} 
1030:                   \put(181,99){\scriptsize$ r $} 
1031:                   \setlength{\unitlength}{1pt}}
1032:   \put(310,0)    {\Includeourbeautifulpicture 64e }
1033:   \put(310,0)    {\setlength{\unitlength}{.38pt}
1034:                   \put(-5,12){\scriptsize$ \mu $} 
1035:                   \put( 80, 5){\scriptsize$ i $} 
1036:                   \put(170, 5){\scriptsize$ j $} 
1037:                   \put(240, 5){\scriptsize$ k $} 
1038:                   \put(120,203){\scriptsize$ \nu $} 
1039:                   \put(160,120){\scriptsize$ q $} 
1040:                   \put(207,73){\scriptsize$ r $} 
1041:                   \setlength{\unitlength}{1pt}}
1042:   \put(90,110)   {\Includeourbeautifulpicture 64f }
1043:     \put(92,128)   {\scriptsize$ \GG[A] $}
1044:   \put(120,40)   {\Includeourbeautifulpicture 64g }
1045:     \put(125,45)   {\scriptsize$ \GG[A] $}
1046:   \put(256,40)   {\Includeourbeautifulpicture 64h }
1047:     \put(270,33)   {\scriptsize$ \GG[A] $}
1048:   \put(256,165)  {\Includeourbeautifulpicture 64h }
1049:     \put(270,175)   {\scriptsize$ \GG[A] $}
1050:   \put(341,81)   {\Includeourbeautifulpicture 64i }
1051:     \put(390,100)   {\scriptsize$ \GG $}
1052:     \put(334,100)   {\scriptsize$ \FF $}
1053:   \epicture44 \labp 64
1054: Explicitly, this pentagon reads
1055:   \bea\displaystyle
1056:   \sum_\eps 
1057:   \GG[A]^{(\kappa j k)\,\nu}_{\alpha_2 \lambda \alpha_3 \, , \, \eps r \delta_2}\,
1058:   \GG[A]^{(\mu i r)\,\nu}_{\alpha_1 \kappa \eps \, , \, \gamma_1 q \delta_1}
1059:   \\[5pt] \displaystyle \hspace*{7em} = 
1060:   \sum_p \sum_{\beta_1,\beta_2,\gamma_1}
1061:   \GG[A]^{(\mu i j)\,\lambda}_{\alpha_1 \kappa \alpha_2 \, , \, \beta_1 p \beta_2}\,
1062:   \GG[A]^{(\mu p k)\,\nu}_{\beta_1 \lambda \alpha_3 \, , \, \gamma_1 q \gamma_2}\,
1063:   \GG^{(ijk)\,q}_{\beta_2 p \gamma_2 \, , \, \delta_1 r \delta_2} \,.
1064:   \eear\labl{eq:module-pent-G}
1065: If one prefers the fusing matrix $\FF$ of $\calc$ to appear directly,
1066: rather than its inverse $\GG$, one may rewrite this in the form
1067:   \bea\displaystyle
1068:   \sum_{\beta_1} 
1069:   \GG[A]^{(\mu i j)\,\lambda}_{\alpha_1 \kappa \alpha_2 \, , \, \beta_1 p \beta_2}\,
1070:   \GG[A]^{(\mu p k)\,\nu}_{\beta_1 \lambda \alpha_3 \, , \, \gamma_1 q \gamma_2}
1071:   \\[5pt] \displaystyle \hspace*{7em} = 
1072:   \sum_r \sum_{\delta_1,\delta_2,\eps}
1073:   \GG[A]^{(\kappa j k)\,\nu}_{\alpha_2 \lambda \alpha_3 \, , \, \eps r \delta_2}\,
1074:   \GG[A]^{(\mu i r)\,\nu}_{\alpha_1 \kappa \eps \, , \, \gamma_1 q \delta_1}\,
1075:   \FF^{(ijk)\,q}_{\delta_1 r \delta_2 \, , \, \beta_2 p \gamma_2} \,.
1076:   \eear \hspace*{1.6em} \labl{eq:module-pent-F}
1077: When all morphism spaces involved are one-dimensional, these relations
1078: simplify considerably, e.g.\ \erf{eq:module-pent-G} reads
1079:   \be
1080:   \GG[A]^{(\kappa j k)\,\nu}_{\lambda \,,\, r}\,
1081:   \GG[A]^{(\mu i r)\,\nu}_{\kappa \,,\, q}
1082:   = \sum_p \GG[A]^{(\mu i j)\,\lambda}_{\kappa \,,\, p}\,
1083:   \GG[A]^{(\mu p k)\,\nu}_{\lambda \,,\, q}\,
1084:   \GG^{(ijk)\,q}_{p \,,\, r} \,.  \ee
1085: 
1086: \dtl{Remark}{rem:bnd-6j}
1087: (i)~\,The fusing matrices $\GG[A]$ (or rather, their inverses $\FF[A]$)
1088: also appear in an approach to CFT based on graphs and cells
1089: \cite{bppz2,pezu6}, where the notation $^{(1)\!}F$ is used. The $^{(1)\!}F$ 
1090: are required to solve a pentagon relation; the corresponding equation 
1091: for $\GG[A]$ is \erf{eq:module-pent-G}. In fact, the associativity conditions
1092: between the categories $\calc$, $\calca$ and $\calcaa$ give rise to a number of
1093: pentagon relations, called the ``Big Pentagon'', for the corresponding 
1094: associators. The associators and the pentagon relations can be visualised using
1095: cells (known as Ocneanu cells) \cite{ocne10,bosz},
1096: and can also be interpreted in terms a 2-category with two objects,
1097: see e.g.\ \cite{ffrs4} for more details. 
1098: \\[.3em]
1099: (ii)~As observed for the Cardy case $A\eq\one$ in \cite{runk,fffs}, and for the
1100: general case in \cite{bppz}, the pentagon for $\GG[A]$ is equivalent to the 
1101: sewing (or factorisation) constraint \cite{lewe3} for four boundary fields 
1102: on the upper half plane. If $\hatpsi^{\mu\nu}_a$ denotes a boundary field
1103: which changes a boundary condition described by the $A$-module $M_\nu$
1104: to $M_\mu$ (looking along the real axis towards $+\infty$) and has chiral 
1105: representation label $a$, the boundary OPE symbolically takes the form
1106:   \be
1107:   \hatpsi^{\mu\nu}_a\, \hatpsi^{\nu\kappa}_a =~\sum_{c} c^{\mu\nu\kappa}_{abc} 
1108:   \, \hatpsi^{\mu\kappa}_c \,,
1109:   \ee
1110: where the constants $c^{\mu\nu\kappa}_{abc}$ are the OPE coefficients\,%
1111:   \footnote{~With all multiplicity labels and coordinate dependence
1112:   in place, one obtains an equation of the form (I:3.11).}. 
1113: Factorisation of the correlator of four boundary fields leads to
1114: the constraint \cite{lewe3}
1115:   \be
1116:   c^{\mu\nu\kappa}_{abf} \, c^{\mu\kappa\rho}_{fcd} = 
1117:   \sum_c c^{\nu\kappa\rho}_{bce} \, c^{\mu\nu\rho}_{aed}\, \Fs abcdef \,.
1118:   \ee
1119: By relation \erf{eq:module-pent-F}, this equation in the unknowns
1120: $c$ is solved by
1121:   \be
1122:   c^{\mu\nu\kappa}_{abc} = \GG[A]^{(\mu a b)\,\kappa}_{\nu \,,\, c} \,,
1123:   \labl{eq:bnd-GA-id}
1124: where again all multiplicity labels are omitted.
1125: \\
1126: And indeed, when evaluating the
1127: corresponding ribbon invariant in \erf{eq:bnd-3pt-rib} below, one finds
1128: the explicit expression \erf{eq:3bnd-inv-basis} for the coefficient
1129: occurring in the correlator of three boundary fields. This expression
1130: implies\,\footnote{~%
1131:   In \erf{eq:3bnd-inv-basis} the combination
1132:   $c^{\kappa \nu \mu}_{ji \bar k} c^{\mu \kappa \mu}_{k \bar k 0}$
1133:   appears because it describes a three-point correlator, rather than an
1134:   OPE coefficient.}
1135: \erf{eq:bnd-GA-id}. We thus arrive at the conclusion that the boundary OPE 
1136: coefficients are given by the 6j-symbols of the module category $\calca$.
1137: 
1138: 
1139: \subsection{Fusing matrices for bimodules}
1140: 
1141: The bimodule category that is needed for the description of the bulk
1142: structure constants is of the form\,\footnote{%
1143:   ~As mentioned in section \ref{sec:in-1}, we treat here only theories
1144:   defined on surfaces of types (3) and (4). When one does not include
1145:   surfaces with boundary, one can also consider heterotic theories,
1146:   in which case more general bimodule categories
1147:   $\calc \Times \calb \Times \overline{\cald} \To \calb$ occur.}
1148: $\calc \Times \calb \Times \overline{\calc} \To \calb$. Here 
1149: $\overline{\calc}$ denotes the tensor category dual\,\footnote{%
1150:   ~The dual of a tensor category $(\calc,\otimes)$ is e.g.\ defined in section
1151:   6.2 of \cite{ffrs}, where it was taken to be $(\calc^{\rm opp}, \otimes)$
1152:   rather than $(\calc,\otimes^{\rm opp})$. These two categories
1153:   are tensor equivalent via the functor $?^\vee$, 
1154:   i.e.\ by taking the duals of objects and morphisms.}
1155: to $\calc$ in the sense that if $\calc \eq (\calc , {\otimes})$ then 
1156: $\overline{\calc} \eq (\calc,{\otimes}^{\rm opp})$. Denoting the quantities in 
1157: $\overline{\calc}$ with a bar, this means that $\overline U \eq U$ and 
1158: $\overline U \,\overline{\otimes}\,\overline V \eq V \Oti\, U$. The 
1159: bimodule category $\calb$ of interest here
1160: is the category $\calcaa$ of $A$-bimodules, see definition I:4.5. 
1161: 
1162: For the action of $\calc$ and $\overline{\calc}$ on $\calcaa$
1163: there are, in each case, two possibilities. Let us start with 
1164: $\otimes_{\calb}^l{:}\; \calc \Times \calcaa \To \calcaa$. Given an object $U$ 
1165: of $\calc$ and an $A$-bimodule $X \eq (\dot X, \rho_l, \rho_r)$, we define 
1166: the bimodules $U \,{\otimes^\pm} X$ as
1167:   \bea
1168:   U \,{\otimes^+}\, X := (U \Oti \dot X , 
1169:     (\id_U\Oti\rho_l) \cir ( c_{U,A}^{~-1} \Oti \id_X), \id_U\Oti\rho_r )
1170:   \qquad{\rm and} \\{}\\[-.7em]
1171:   U \,{\otimes^-}\, X := (U \Oti \dot X , 
1172:     (\id_U\Oti\rho_l) \cir ( c_{\!A,U}^{} \Oti \id_X), \id_U\Oti\rho_r ) \,,
1173:   \eear\labl{UXUX}
1174: respectively. (In graphical notation, for $\otimes^+$ the $U$-ribbon passes 
1175: above the $A$-ribbon, while for $\otimes^-$ it passes below.)
1176:     %  compare section III:4.3.
1177: We take the functor $\otimes_{\calb}^l$ to be given by $\otimes^+$.
1178: %
1179: Similarly we can define bimodules $X \,{\otimes^\pm} U$ when
1180: tensoring with $U$ from the right as
1181:   \bea
1182:   X \,{\otimes^+}\, U := (\dot X \Oti U , \rho_l \Oti \id_U,
1183:   (\rho_r \Oti \id_U) \cir (\id_X\Oti c_{U,A}^{}) ) 
1184:   \qquad{\rm and} \\{}\\[-.8em]
1185:   X \,{\otimes^-}\, U := (\dot X \Oti U , \rho_l \Oti \id_U,
1186:   (\rho_r \Oti \id_U) \cir (\id_X\Oti c_{\!A,U}^{\,-1}) ) \,,
1187:   \eear\labl{XUXU}
1188: respectively. (In graphical notation, again for $\otimes^+$, $U$ passes
1189: above $A$, while for $\otimes^-$, it passes below.)
1190: We take the functor $\otimes_{\calb}^r$ to be given by $\otimes^-$.
1191: 
1192: The reason for choosing the actions of $\calc$ and $\overline{\calc}$
1193: on $\calcaa$ in this particular way is that defect fields are to be labelled 
1194: by elements of the space $\HomAA( U \,{\otimes^+} X \,{\otimes^-}\, V , Y)$
1195: of bimodule morphisms between the $A$-bimodules
1196: $U {\otimes^+} X {\otimes^-} V$ and $Y$. In \cite{fuRs4,fuRs8} instead
1197: the space $\HomAA( X {\otimes^-} V, Y {\otimes^+} U^\vee )$
1198: of bimodule morphisms was used. These two spaces are canonically
1199: isomorphic; an isomorphism can be given as follows. 
1200: \\[-2.3em]
1201: 
1202: \dtl{Lemma}{lem:defect-iso}
1203: The mapping 
1204:   \be
1205:   \varphi :\quad \HomAA( X \,{\otimes^-}\, V, Y \,{\otimes^+}\, U^\vee )
1206:   \rightarrow \HomAA( U \,{\otimes^+} X \,{\otimes^-}\, V , Y)
1207:   \ee  
1208: given by
1209:   \be
1210:   \varphi(f) := (\tilde d_U \oti\id_Y) \cir (\id_U \oti [c_{Y,U^\vee}^{}\cir f])
1211:   \ee
1212: is an isomorphism.
1213: 
1214: \medskip\noindent
1215: Proof:\\
1216: That $\varphi(f)$ commutes with the left and right action of $A$ follows by a 
1217: straightforward use of the definitions. Further, one
1218: readily checks that the map $\tilde\varphi$ acting as
1219:   \be
1220:   \tilde\varphi(g) := c_{Y,U^\vee}^{~-1}
1221:   \cir (\id_{U^\vee} \oti g) \cir (\tilde b_U \oti \id_X \oti \id_V) 
1222:   \ee
1223: for $g\iN\HomAA( U{\otimes^+}X{\otimes^-}V , Y)$ is left and right
1224: inverse to $\varphi$.
1225: \qed
1226: 
1227: \medskip
1228: 
1229: It follows from proposition I:4.6 that the category of $A$-bimodules
1230: is isomorphic to the category of left modules over the algebra
1231: $A\oti A_{\rm op}$. Since $A$ is a symmetric special Frobenius algebra, so is
1232: $A\oti A_{\rm op}$. It follows that the category of 
1233: $A\oti A_{\rm op}$-modules is semisimple, and hence $\calcaa$ is semisimple,
1234: too. Let us choose a set of representatives $\{ X_\kappa\,|\,\kappa\iN\KK\}$
1235: of isomorphism classes of simple $A$-bimodules. 
1236: 
1237: The algebra $A$ is called {\em simple} iff it is a simple bimodule over itself 
1238: or, equivalently, iff $Z(A)_{00}\eq1$, see definition 2.26 and remark 2.28(i) 
1239: of \cite{ffrs} for details. If $A$ is simple, then we choose the 
1240: representatives of simple modules such that $X_0 \eq A$.
1241: 
1242: Further, we choose bases $\{ \xi_{(i\mu j)\nu}^{\alpha} \}$ of the spaces
1243: $\HomAA(U_i{\otimes^+}X_\mu{\otimes^-}U_j,X_\nu)$, with graphical representation
1244:   %  [ pic 4_65 ]
1245:   \bea \begin{picture}(130,56)(0,38)
1246:   \put(60,0)     {\includeourbeautifulpicture 65 }
1247:   \put(0,41)     {$ \xi_{(i\mu j)\nu}^{\alpha} \;=: $}
1248:   \put(57.2,-8.2){\scriptsize$ U_i $}
1249:   \put(76.5,-8.2){\scriptsize$ X_\mu $}
1250:   \put(76.5,90.8){\scriptsize$ X_\nu $}
1251:   \put(89.8,40.2){\scriptsize$ \alpha $}
1252:   \put(98.5,-8.2){\scriptsize$ U_j $}
1253:   \epicture28 \labl{eq:xi-basis} \xlabp 65
1254: For the special case $i\eq j\eq 0$ 
1255:    %  by strictness
1256: we choose $\xi_{(0\mu 0)\mu}\eq \id_{X_\mu}$.
1257: The number of elements in the basis $\{ \xi_{(i\mu j)\nu}^{\alpha} \}$ of 
1258: $\HomAA( U_i{\otimes^+}X_\mu{\otimes^-}U_j , X_\nu)$ is denoted by
1259: $Z_{ij}^{X_\mu|X_\nu}$; more generally, for any two $A$-bimodules $X$ and $Y$ 
1260: we set
1261:   \be
1262:   Z_{ij}^{X|Y} 
1263:   := \dim \HomAA( U_i{\otimes^+}X{\otimes^-}U_j , Y)
1264:   = \dim \HomAA( X{\otimes^-}U_j, Y{\otimes^+}U_{\bar\imath} ) \,;
1265:   \labl{ZijXY}
1266: the second equality follows from lemma \ref{lem:defect-iso} together with 
1267: $U_i^\vee \,{\cong}\, U_{\bar\imath}$. We denote the basis elements 
1268: dual to \erf{eq:xi-basis} by $\bar\xi_{\alpha}^{(i\mu j)\nu}{\in}\,
1269: \HomAA(X_\nu,U_i{\otimes^+}X_\mu{\otimes^-}U_j)$; they obey
1270:   \be
1271:   \xi_{(i\mu j)\nu}^{\alpha} \cir
1272:   \bar\xi_{\beta}^{(i\mu j)\nu} = \delta_{\alpha,\beta}\, \id_{X_\nu}
1273:   \quad\ {\rm and} \quad\
1274:   \sum_{\nu\in\KK} \sum_{\alpha=1}^{Z_{ij}^{X_\mu|X_\nu}}\!\!
1275:   \bar\xi_{\alpha}^{(i\mu j)\nu} \cir
1276:   \xi_{(i\mu j)\nu}^{\alpha} = \id_{U_i} \oti \id_{X_\mu} \oti \id_{U_j}
1277:   \labl{eq:xi-bas-prop}
1278: The definition \erf{ZijXY} of $Z_{ij}^{X|Y}$ is in accordance with the
1279: definition via the ribbon invariant (I:5.151), as can be seen by
1280: inserting the second identity of \erf{eq:xi-bas-prop} into
1281: the ribbon graph (I:5.151) and using $Z_{00}^{X|Y} {=}\, \dim \HomAA(X,Y)$.
1282: The numbers $Z_{ij}^{X|Y}$ also have a physical interpretation; they
1283: describe the torus partition function with the insertion of two
1284: parallel defect lines \cite{pezu5,pezu7}; see section I:5.10 for details.
1285: 
1286: To motivate the definition of the fusing matrices for $\calb\eq\calcaa$,
1287: let us have a look at the left action of $\overline{\calc}$ on $\calb$. 
1288: By semisimplicity,
1289: the isomorphism $(X_\mu{\otimes_{\calb}^r}\overline{U}_i) \,{\otimes_{\calb}^r}\, 
1290: \overline{U}_j \,{\cong}\, X_\mu \,{\otimes_{\calb}^r}\, (\overline{U}_i \, 
1291: \overline{\otimes} \, \overline{U}_j)$ on objects gives rise to an isomorphism
1292:   \bea\displaystyle
1293:   \bigoplus_\kappa
1294:   \Hom^{\calb}(X_\mu \,{\otimes_{\calb}^r}\, \overline{U}_i, X_\kappa) 
1295:   \otimes
1296:   \Hom^{\calb}(X_\kappa {\otimes_{\calb}^r}\, \overline{U}_j, X_\nu) 
1297:   \\[5pt]\displaystyle
1298:   \hspace*{6em} \stackrel{\cong}{\longrightarrow}\;
1299:   \bigoplus_k \Hom^{\calb}(X_\mu \,{\otimes_{\calb}^r}\, \overline{U}_k, X_\nu) 
1300:   \otimes \Hom^{\overline{\mathcal{C}}}(
1301:   \overline{U}_i \, \overline{\otimes} \,\overline{U}_j , \overline{U}_k ) 
1302:   \eear\labl{eq:right-action-cob}
1303: of morphism spaces. By definition, the space $\Hom^{\overline{\mathcal{C}}}(
1304: \overline{U}_i \, \overline{\otimes} \,\overline{U}_j , \overline{U}_k )$
1305: of morphisms in $\overline{\calc}$ is equal to the
1306: morphism space $\Hom^{\mathcal{C}}(U_j{\otimes}U_i,U_k)$ in $\calc$, while 
1307: $\Hom^{\calb}(X_\mu{\otimes_{\calb}^r}\overline{U}_i, X_\kappa) 
1308: \eq \HomAA(X_\mu{\otimes^{-}} U_i, X_\kappa)$ etc. 
1309: We would still like to describe the isomorphism
1310: \erf{eq:right-action-cob} with the help of the graphical rep\-resentation
1311: of morphisms in the ribbon category $\calc$. To this end we need a 
1312: visualisation for morphisms in $\overline{\calc}$. 
1313: A natural possibility is to represent them graphically by
1314: coupons that face the reader with their back side. Thus, using dashed lines
1315: for ribbons with back side facing the reader (see the beginning of
1316: section II:3 for conventions), we have
1317:   %  [ pic 4_66 ]
1318:   \bea \begin{picture}(120,64)(0,43)
1319:                  \put(0,0){ \begin{picture}(0,0)
1320:   \put(0,0)      {\Includeourbeautifulpicture 66a }
1321:   \put(-.6,-8.2) {\scriptsize$ U $}
1322:   \put(14.8,106) {\scriptsize$ W $}
1323:   \put(16.1,57.4){\tiny$ f $}
1324:   \put(29.5,-8.2){\scriptsize$ V $}
1325:                  \end{picture} }
1326:   \put(56,50)    {$ = $}
1327:                  \put(78,0){ \begin{picture}(0,0)
1328:   \put(2,0)      {\Includeourbeautifulpicture 66b }
1329:   \put(-.6,-8.2) {\scriptsize$ U $}
1330:   \put(14.8,106) {\scriptsize$ W $}
1331:   \put(16.1,57.4){\tiny$ f $}
1332:   \put(29.5,-8.2){\scriptsize$ V $}
1333:                  \end{picture} }
1334:   \epicture26 \labp 66
1335: for $f \iN \Hom^{\overline{\mathcal{C}}}(\overline{U}\,\overline{\otimes}\,
1336: \overline{V},\overline{W}) \eq \Hom^{\mathcal{C}}(V{\otimes}U,W)$, 
1337: where on the left hand side the coupon labelled by $f$ is showing the black
1338: side to the reader while on the right hand side it is showing its white side.
1339: 
1340: \medskip
1341: 
1342: After these preliminaries we are ready to give the definition of the
1343: fusing matrices $\FF[A|A]$ of the bimodule category $\calcaa$. We have
1344:   %  [ pic 4_67 ]
1345:   \bea \begin{picture}(410,67)(0,46)
1346:                  \put(0,0){ \begin{picture}(0,0)
1347:   \put(0,0)      {\Includeourbeautifulpicture 67a }
1348:   \put(-2.8,-8.2){\scriptsize$ U_i $}
1349:   \put(17.5,-8.2){\scriptsize$ U_j $}
1350:   \put(29.5,56.5){\scriptsize$ X_{\!\kappa} $}
1351:   \put(37.5,-8.2){\scriptsize$ X_\mu $}
1352:   \put(37.5,110) {\scriptsize$ X_\nu $}
1353:   \put(49.8,49.6){\scriptsize$ \alpha $}
1354:   \put(49.8,74.6){\scriptsize$ \beta $}
1355:   \put(59.2,-8.2){\scriptsize$ U_k $}
1356:   \put(79.5,-8.2){\scriptsize$ U_l $} 
1357:                  \end{picture} }
1358:                  \put(311,0){ \begin{picture}(0,0)
1359:   \put(0,0)      {\Includeourbeautifulpicture 67b }
1360:   \put(-2.8,-8.2){\scriptsize$ U_i $}
1361:   \put(22.7,-8.2){\scriptsize$ U_j $}
1362:   \put(24.7,66.2){\scriptsize$ U_{q_1} $}
1363:   \put(41.8,-8.2){\scriptsize$ X_\mu $}
1364:   \put(41.8,110) {\scriptsize$ X_\nu $}
1365:   \put(54.8,75.9){\scriptsize$ \gamma $}
1366:   \put(59.8,62.2){\scriptsize$ U_{q_2} $}
1367:   \put(63.2,-8.2){\scriptsize$ U_k $}
1368:   \put(17.2,46.3){\tiny$ \delta_1 $}
1369:   \put(82.5,46.3){\tiny$ \delta_2 $}
1370:   \put(89.1,-8.2){\scriptsize$ U_l $} 
1371:                  \end{picture} }
1372:   \put(97,50)    {$ = \dsty\sum_{q_1,q_2\in\II}\!
1373:                  \sum_{\gamma=1}^{Z_{q_1 q_2}^{X_\mu|X_\nu}}
1374:                  \sum_{\delta_1=1}^{\N ij{q_1}}\;\sum_{\delta_2=1}^{\N ij{q_2}}\,
1375:      \FF[A|A]^{(ij\mu kl)\,\nu}_{\alpha\kappa\beta,\gamma q_1q_2\delta_1\delta_2} $}
1376:   \epicture32 \labl{eq:FAA-def} \xlabp 67
1377: Note that this encodes both the left action of $\calc$
1378: and the right action of $\overline{\calc}$; the individual actions can be 
1379: extracted by setting $i\eq j\eq 0$ or $k\eq l\eq 0$, respectively.
1380: 
1381: An expression for the bimodule fusing matrix $\FF[A|A]$ 
1382: is obtained from \erf{eq:FAA-def} by composing with the dual basis. One finds
1383:   %  [ pic 4_68 ]
1384:   \bea \begin{picture}(400,140)(0,51)
1385:   \put(102,0)    {\Includeourbeautifulpicture 68a }
1386:   \put(150,0)    {\Includeourbeautifulpicture 68b }
1387:   \put(287,0)    {\Includeourbeautifulpicture 68c }
1388:   \put(0,87)     {$ \FF[A|A]^{(ij\mu kl)\, \nu}_{\alpha \pi \beta, 
1389:                  \gamma q_1 q_2 \delta_1 \delta_2} $}
1390:   \put(98,-8)    {\scriptsize$ X_\nu $}
1391:   \put(98,183)   {\scriptsize$ X_\nu $} 
1392:   \put(121,87)   {$ = $}
1393: %
1394:   \put(192,-8)   {\scriptsize$ X_\nu $}
1395:   \put(192,183)  {\scriptsize$ X_\nu $} 
1396:   \put(182,146)  {\scriptsize$ \beta $}
1397:   \put(184,129)  {\scriptsize$ X_\pi $}
1398:   \put(182,122)  {\scriptsize$ \alpha $}
1399:   \put(184,85)   {\scriptsize$ X_\mu $}
1400:   \put(182,29)   {\scriptsize$ \overline{\gamma} $}
1401:   \put(166.5,102){\scriptsize$ U_j $}
1402:   \put(219,102)  {\scriptsize$ U_k $} 
1403:   \put(148,122)  {\scriptsize$ U_i $}
1404:   \put(237,120)  {\scriptsize$ U_l $} 
1405:   \put(153,52)   {\scriptsize$ \overline{\delta_1} $}
1406:   \put(233,52)   {\scriptsize$ \overline{\delta_2} $} 
1407:   \put(170,42.1) {\scriptsize$ U_{q_1} $}
1408:   \put(216,42.1) {\scriptsize$ U_{q_2} $} 
1409: %
1410:   \put(260,87)   {$ = $} 
1411: %
1412:   \put(329,-8)   {\scriptsize$ X_\nu $}
1413:   \put(329,183)  {\scriptsize$ X_\nu $} 
1414:   \put(320,146)  {\scriptsize$ \beta $} 
1415:   \put(320,122)  {\scriptsize$ \alpha $}
1416:   \put(320,85)   {\scriptsize$ X_\mu $}
1417:   \put(320,29)   {\scriptsize$ \overline\gamma $} 
1418:   \put(321.5,129){\scriptsize$ X_\pi $}
1419:   \put(285,120)  {\scriptsize$ U_i $}
1420:   \put(375,120)  {\scriptsize$ U_l $} 
1421:   \put(290,52)   {\scriptsize$ \overline{\delta_1} $}
1422:   \put(368,51.6) {\scriptsize$ \overline{\delta_2} $} 
1423:   \put(303,102)  {\scriptsize$ U_j $}
1424:   \put(354,102)  {\scriptsize$ U_k $} 
1425:   \put(303,42)   {\scriptsize$ U_{q_1} $}
1426:   \put(354,42)   {\scriptsize$ U_{q_2} $} 
1427:   \epicture34 \labl{eq:FAA-rib-inv} \xlabp 68
1428: Evaluating this expression is again a linear
1429: problem, but it is more complicated than the procedure for evaluating 
1430: \erf{eq:GA-rib-inv}. First one must find all simple $A$-bimodules. 
1431: By proposition I:4.6, this is the same as finding all
1432: simple $A{\otimes}A_{\rm op}$-modules, which (via the method
1433: of induced modules) is a finite, linear problem. Next, the subspaces
1434: $\HomAA( U_i\,{\otimes^+}\,X\,{\otimes^- U_j}, Y)$ of
1435: $\Hom( U_i\oti\dot X\oti U_j, \dot Y)$
1436: can be determined as images of the projection operators 
1437:   %  [ pic 4_69 ]
1438:   \bea \begin{picture}(180,93)(0,51)
1439:   \put(62,0)    {\includeourbeautifulpicture 69 }
1440:   \put(0,63)      {$ P(f) \;:= $}
1441:   \put(59,-9.5)   {\scriptsize$ U_i $}
1442:   \put(154,-9.5)  {\scriptsize$ U_j $}
1443:   \put(106,-9.5)  {\scriptsize$ X $}
1444:   \put(59,78)     {\scriptsize$ A $}
1445:   \put(95,58)     {\scriptsize$ A $}
1446:   \put(154.8,60)  {\scriptsize$ A $}
1447:   \put(108,87)    {\scriptsize $f$}
1448:   \put(108,140)   {\scriptsize$ Y $} 
1449:   \epicture35 \labp 69 
1450: for $f\iN\Hom( U_i\oti\dot X\oti U_j, \dot Y)$.
1451: This is a linear problem, after the solution of which  
1452: one has all the ingredients at hand that are needed to evaluate 
1453: \erf{eq:FAA-rib-inv}.
1454: 
1455: By construction, the $\FF[A|A]$ satisfy a pentagon relation; schematically, this
1456: relation looks like
1457:   %  [ pic 4_70 ]
1458:   \bea \begin{picture}(420,280)(10,85)
1459:   \put(0,110)    {\Includeourbeautifulpicture 70a }
1460:   \put(0,110)     {\setlength{\unitlength}{.38pt}
1461:                   \put(144.3,23){\scriptsize$ \mu $} 
1462:                   \put(144.3,150){\scriptsize$ \kappa $} 
1463:                   \put(144.3,270){\scriptsize$ \lambda $} 
1464:                   \put(144.3,375){\scriptsize$ \nu $} 
1465:                   \put(  1,23){\scriptsize$ i $} 
1466:                   \put( 52,23){\scriptsize$ j $} 
1467:                   \put(103,23){\scriptsize$ k $} 
1468:                   \put(213,23){\scriptsize$ l $} 
1469:                   \put(266,23){\scriptsize$ m $} 
1470:                   \put(319,23){\scriptsize$ n $} 
1471:                   \setlength{\unitlength}{1pt}}
1472:   \put(150,220)  {\Includeourbeautifulpicture 70b }
1473:   \put(310,220)  {\Includeourbeautifulpicture 70c }
1474:   \put(150,0)    {\Includeourbeautifulpicture 70d }
1475:   \put(310,0)    {\Includeourbeautifulpicture 70e }
1476:   \put(310,0)     {\setlength{\unitlength}{.38pt}
1477:                   \put(144.3,23){\scriptsize$ \mu $} 
1478:                   \put(144.3,375){\scriptsize$ \nu $} 
1479:                   \put(  1,23){\scriptsize$ i $} 
1480:                   \put( 52,23){\scriptsize$ j $} 
1481:                   \put(102,23){\scriptsize$ k $} 
1482:                   \put(213,23){\scriptsize$ l $} 
1483:                   \put(264,23){\scriptsize$ m $} 
1484:                   \put(319,23){\scriptsize$ n $} 
1485:                   \put(104,251){\scriptsize$ r $} 
1486:                   \put(47,126){\scriptsize$ t $} 
1487:                   \put(217,251){\scriptsize$ s $} 
1488:                   \put(271,126){\scriptsize$ u $} 
1489:                   \setlength{\unitlength}{1pt}}
1490:   \put(92,228)   {\scriptsize$ \FF[A|A] $}
1491:   \put(122,80)   {\scriptsize$ \FF[A|A] $}
1492:   \put(276,300)  {\scriptsize$ \FF[A|A] $}
1493:   \put(276,80)   {\scriptsize$ \FF[A|A] $}
1494:   \put(376,184)  {\scriptsize$ \FF \cdot \FF $}
1495:   \put(100,210)  {\Includeourbeautifulpicture 64f }
1496:   \put(130,72)   {\Includeourbeautifulpicture 64g }
1497:   \put(265,70)   {\Includeourbeautifulpicture 64h }
1498:   \put(265,290)  {\Includeourbeautifulpicture 64h }
1499:   \put(368,158)  {\Includeourbeautifulpicture 64j }
1500:   \epicture61 \labp 70
1501: and explicitly it reads
1502:   \bea\displaystyle \hsp{-.7}
1503:   \sum_\delta
1504:   \FF[A|A]^{(ij \kappa mn)\, \nu}
1505:           _{\alpha_2 \lambda \alpha_3 \,,\, \delta t u \eps_1 \varphi_1} \,
1506:   \FF[A|A]^{(tk \mu lu)\,\nu}
1507:           _{\alpha_1 \kappa \delta \,,\,  \gamma_1 r s \eps_2 \varphi_2}
1508:   \\[5pt] \hsp{2.2} \displaystyle 
1509:   = \sum_{p,q} \sum_{\beta_1,\beta_2,\beta_3,\gamma_2,\gamma_3} \hspace*{-1em}
1510:   \FF[A|A]^{(jk \mu lm)\,\lambda}
1511:           _{\alpha_1 \kappa \alpha_2 \,,\, \beta_1 p q \beta_2 \beta_3} \,
1512:   \FF[A|A]^{(ip \mu qn)\,\nu}
1513:           _{\beta_1 \lambda \alpha_3 \,,\, \gamma_1 r s \gamma_2 \gamma_3} \,
1514:   \FF^{(ijk)\,r}_{\gamma_2 p \beta_2 \,,\, \eps_1 t \eps_2} \,
1515:   \FF^{(nml)\,s}_{\gamma_3 q \beta_3 \,,\, \varphi_1 u \varphi_2} .
1516:   \eear\labl{eq:FAA-pentagon}
1517: 
1518: \dtl{Remark}{rem:bulk-6j}
1519: (i)~\,In remark \ref{rem:bnd-6j}(ii) we have seen that the boundary OPE is
1520: given by the 6j-symbols of the module category $\calca$. In fact, similar 
1521: considerations apply to bulk and defect fields.\\
1522: Let $\vPhi_\alpha$ be a bulk field with chiral/anti-chiral representation
1523: labels $\alpha_l$, $\alpha_r$. Omitting all position dependence and
1524: multiplicity indices, the OPE of two bulk fields can symbolically we written as
1525:   \be
1526:   \vPhi_\alpha \, \vPhi_\beta = \sum_{\gamma} C_{\alpha\beta}^{~\;\gamma} \,
1527:   \vPhi_\gamma \,.
1528:   \ee
1529: The OPE coefficients $C_{\alpha\beta}^{~~\gamma}$ must solve a factorisation
1530: constraint coming from the four-point correlator on the sphere 
1531: \cite{sono2,lewe3,dofa},
1532:   \be
1533:   C_{\alpha\beta}^{~\;\varphi}\, C_{\varphi\gamma}^{~\;\delta} 
1534:   = \sum_{\eps}
1535:   C_{\alpha\eps}^{~\;\delta} \, C_{\beta\gamma}^{~\;\eps} \,
1536:   \Fs{\alpha_l}{\beta_l}{\gamma_l}{\delta_l}{\eps_l}{\varphi_l} \,
1537:   \Fs{\alpha_r}{\beta_r}{\gamma_r}{\delta_r}{\eps_r}{\varphi_r} \,,
1538:   \labl{eq:4bulk-fact}
1539: where again all multiplicity indices and summations are omitted. If we identify
1540:   \be
1541:   C_{\alpha\beta}^{~~\gamma} =
1542:   \FF[A|A]^{(\alpha_l \beta_l 0 \beta_r \alpha_r) 0}_{\cdot 0 \cdot , 
1543:   \cdot \gamma_l \gamma_r \cdot \cdot } \,,
1544:   \labl{eq:bulk-OPE-FAA}
1545: then \erf{eq:4bulk-fact} is equivalent to the pentagon relation 
1546: \erf{eq:FAA-pentagon} for the $\FF[A|A]$-matrices. Recall that
1547: the bimodule index $0$ refers to the algebra $A$ itself (we take
1548: $A$ to be simple).
1549: \\
1550: In the TFT approach, the factorisation constraints hold by construction,
1551: and indeed the ribbon invariant \erf{eq:3bulk-rib} for the
1552: correlator of three bulk fields on the sphere (to be constructed
1553: in section \ref{sec:rc-3bulk} below) leads to the coefficient
1554: \erf{eq:3bulk-coeff}, which implies \erf{eq:bulk-OPE-FAA}. Note that
1555: in \erf{eq:3bulk-coeff} two factors $\FF[A|A]$ appear because this
1556: equation describes the coefficient for a three-point correlator, which
1557: is a product of two OPE coefficients.
1558: \\
1559: Let us now turn to defect fields. Denote by $\Thetha^{\mu\nu}_\alpha$
1560: a defect field which changes a defect described by a bimodule $X_\nu$
1561: into a defect described by $X_\mu$. 
1562: One can think of a bulk field $\vPhi_\alpha$ as a special defect field
1563: $\Thetha^{\mu\nu}_\alpha$ which changes the invisible defect 
1564: (labelled by $A$ itself)
1565: to the invisible defect, i.e.\ with $\mu$ and $\nu$ set to zero. 
1566: One can convince oneself that the OPE-coefficients of defect fields
1567: have to fulfill a factorisation condition similar to 
1568: \erf{eq:4bulk-fact} and accordingly we would expect that these OPE-coefficients
1569: are also given by matrix elements of $\FF[A|A]$. As before, this does indeed
1570: result from the TFT-computation, see \erf{eq:3def-basis} below.
1571: Thus, just like is the case of boundary fields, the OPE coefficients of
1572: defect fields can be seen as the coefficients of a suitable associator
1573: morphism.
1574: \\[5pt]
1575: (ii)~It is also worth recalling a well-known geometric interpretation of the 
1576: pentagon identity for the $\FF$-matrices (see e.g.\ \cite{dujn}).
1577: This uses a description of the 6j-symbols $\FF$ of a semisimple tensor category
1578: as tetrahedra, with edges labelled by simple objects $U_i$ and faces by 
1579: morphisms $\Hom(U_i\oti U_j,U_k)$, i.e.\ schematically
1580:   %  [ pic 4_41 ]
1581:   \bea \begin{picture}(140,40)(0,18)
1582:   \put(40,0)     {\includeourbeautifulpicture 41 }
1583:   \put(0,23)     {$ \FF \;= $}
1584:   \epicture06 \labp 41
1585: In this description, the pentagon identity corresponds to the equality between
1586: two ways to obtain the body of a `double tetrahedron': first, by gluing two 
1587: tetrahedra along a common face, and second, by gluing three tetrahedra along 
1588: an edge common to all three of them (namely the edge that, in the first 
1589: description, connects the two vertices not belonging to the distinguished face)
1590: and along three faces each of which is common to two of the tetrahedra. In the 
1591: resulting equation, the labels of the common faces and the internal edge are 
1592: to be summed over. 
1593: 
1594: To obtain an analogous interpretation of the pentagon relation
1595: \erf{eq:FAA-pentagon} for bimodules, one can represent the $\FF[A|A]$ as prisms
1596:   %  [ pic 4_42 ]
1597:   \bea \begin{picture}(200,41)(0,21)
1598:   \put(70,0)     {\includeourbeautifulpicture 42 }
1599:   \put(0,24)     {$ \FF[A|A] \;= $}
1600:   \epicture11 \labp 42
1601: in which the edges of the two triangular faces are labelled by simple
1602: objects of \calc\ and the remaining edges by simple $A$-bimodules.
1603: The pentagon \erf{eq:FAA-pentagon} then describes again the equality between
1604: two ways to obtain a certain geometric body, this time a `double prism':
1605: on one side (two $\FF[A|A]$s) two prisms are glued along a common
1606: quadrangle; on the other side, they are instead glued along a quadrangle of 
1607: which two edges are those edges of the original outer quadrangles that were 
1608: not involved in the previous gluing; at the remaining two edges the quadrangle 
1609: must touch triangles, and this is achieved by gluing a tetrahedron to each of 
1610: the original pairs of triangles. Again, the labels of common faces and internal 
1611: edges are to be summed over. Pictorially,
1612:   %  [ pic 4_44 ]
1613:   \bea \begin{picture}(335,50)(0,46)
1614:   \put(0,0)      {\Includeourbeautifulpicture 44a }
1615:   \put(194,0)    {\Includeourbeautifulpicture 44b }
1616:   \put(151,48)   {$ = $}
1617:   \epicture19 \labp 44
1618: 
1619: \bigskip{}
1620: 
1621: 
1622: \subsection{Expressions in a basis} \label{sec:mb-basis}
1623: 
1624: As already pointed out, the fusing matrices $\GG[A]$ and $\FF[A|A]$
1625: are completely determined by the modular tensor category $\calc$, the
1626: symmetric special Frobenius algebra $A$, and a choice of bases in the 
1627: relevant morphism spaces. While the explicit expressions
1628: are not very illuminating, we still quickly go through the calculations,
1629: because later on we will need some of the quantities entering these computations.
1630: 
1631: We can describe simple subobjects of an object $V$ of $\calc$ by
1632: bases $\{b_{(i,\alpha)}^V\}$ of the morphism spaces $\Hom(U_i,V)$
1633: and the dual bases $\{b^{(i,\alpha)}_V\}$ of $\Hom(V,U_i)$, satisfying 
1634: $b_{V}^{(j,\beta)} \cir b^V_{(i,\alpha)} \eq\delta_i^j\, \delta_\alpha^\beta 
1635: \,\id_{U_i}^{}$ and the completeness property $\sum_{i\in\II}\sum_\alpha 
1636: b^V_{(i,\alpha)} \cir b_{V}^{(i,\alpha)}\eq\id_V$. Their graphical notation is
1637:  %   [pic~77] 
1638:   \bea \begin{picture}(180,34)(0,21)
1639:   \put(45,0) {\Includeourbeautifulpicture 77a }
1640:   \put(-5,20)     {$ b^V_{(i,\alpha)}=: $}
1641:   \put(46.1,-7.8) {\scriptsize$ U_i $}
1642:   \put(46.1,46.9) {\scriptsize$ V $}
1643:   \put(53.1,21)   {\tiny$ \alpha $}
1644:   \put(173,0){\Includeourbeautifulpicture 77b }
1645:   \put(77,20)     {$ , $}
1646:   \put(122,20)    {$ b_V^{(j,\beta)}=: $}
1647:   \put(172.9,-8.4){\scriptsize$ V $}
1648:   \put(174.2,48.4){\scriptsize$ U_j $}
1649:   \put(181.3,21.2){\tiny$\bar\beta $}
1650:   \epicture11 \labl{pic77} \xlabp 77
1651: For modules, this amounts to the choice of morphisms introduced in (I:4.21).
1652: For bimodules we take analogously
1653: $b_{(i,\alpha)}^X\iN\Hom(U_i,\dot X)$ and $b^{(i,\alpha)}_X\iN\Hom(\dot X,U_i)$.
1654: In terms of these embedding and restriction morphisms we can now expand the 
1655: morphisms $\psi_{(\mu k)\nu}^{\alpha}$ and $\xi_{(i\mu j)\nu}^{\alpha}$ 
1656: described before, as well as their duals. We choose the conventions
1657:   %  [ pic 4_78 pic 4_79 ]
1658:   \bea \begin{picture}(360,169)(0,46)
1659:                  \put(0,115){\begin{picture}(0,0)
1660:   \put(60,0)     {\Includeourbeautifulpicture 78a }
1661:   \put(312,0)    {\Includeourbeautifulpicture 78b }
1662:   \put(0,42)     {$ \psi^\alpha_{(\mu k)\nu} \;\equiv $}
1663:   \put(105,42)   {$\dsty := \sum_{m,n\in\II}
1664:                  \sum_{\gamma_1=1}^{\langle U_m, \dot M_\mu\rangle}
1665:                  \sum_{\gamma_2=1}^{\langle U_n, \dot M_\nu\rangle}
1666:                  \sum_{\delta=1}^{\N mkn}\,
1667:                  [\psi^\alpha_{(\mu k)\nu}]^{mn}_{\gamma_1 \gamma_2 \delta} $}
1668:  %%label for 78a
1669:   \put(58,-7)    {\scriptsize$ M_\mu $}
1670:   \put(58,91)    {\scriptsize$ M_\nu $}
1671:   \put(81,-7)    {\scriptsize$ U_k $}
1672:   \put(73,44.5)  {\scriptsize$ \alpha $} 
1673:  %%label for 78b
1674:   \put(310,-7)   {\scriptsize$ M_\mu $}
1675:   \put(310,91)   {\scriptsize$ M_\nu $}
1676:   \put(304,20)   {\scriptsize$ \overline\gamma_1 $}
1677:   \put(304,67)   {\scriptsize$ \gamma_2 $}
1678:   \put(331,-7)   {\scriptsize$ U_k $}
1679:   \put(328,45)   {\scriptsize$ \delta $}
1680:   \put(307,30)   {\scriptsize$ m $}
1681:   \put(307,56)   {\scriptsize$ n $} 
1682:                  \end{picture}}
1683:                  \put(0,0){\begin{picture}(0,0)
1684:   \put(60,0)     {\Includeourbeautifulpicture 79a }
1685:   \put(312,0)    {\Includeourbeautifulpicture 79b }
1686:   \put(0,42)     {$ \bar\psi_\alpha^{(\mu k)\nu} \;\equiv $}
1687:   \put(105,42)   {$\dsty := \sum_{m,n\in\II} 
1688:                  \sum_{\gamma_1=1}^{\langle U_m, \dot M_\mu\rangle}
1689:                  \sum_{\gamma_2=1}^{\langle U_n, \dot M_\nu\rangle}
1690:                  \sum_{\delta=1}^{\N mkn}\,
1691:                  [\bar\psi_\alpha^{(\mu k)\nu}]^{mn}_{\gamma_1 \gamma_2 \delta} $}
1692:                  \end{picture}}
1693:  %%label for 79a
1694:   \put(58,91)    {\scriptsize$ M_\mu $}
1695:   \put(58,-7)    {\scriptsize$ M_\nu $}
1696:   \put(80,91)    {\scriptsize$ U_k $}
1697:   \put(73,41)    {\scriptsize$ \overline{\alpha} $}
1698:  %%label for 79b
1699:   \put(310,91)   {\scriptsize$ M_\mu $}
1700:   \put(310,-7)   {\scriptsize$ M_\nu $}
1701:   \put(304,67)   {\scriptsize$ \gamma_1 $}
1702:   \put(304,20)   {\scriptsize$ \overline\gamma_2 $}
1703:   \put(331,91)   {\scriptsize$ U_k $}
1704:   \put(324.5,49.5) {\scriptsize$ \overline\delta $}
1705:   \put(308.7,28) {\scriptsize$ n $}
1706:   \put(307,55)   {\scriptsize$ m $} 
1707:   \epicture33 \labl{eq:psi-expansion} \xlabp 78 \xlabp 79
1708: for the module basis; the expansion coefficients
1709: $[\psi^\alpha_{(\mu k)\nu}]^{mn}_{\gamma_1 \gamma_2 \delta}$ are
1710: complex numbers. Analogously, for bimodules we choose
1711:   %  [ pic 4_80 pic 4_81 ]
1712:   \bea \begin{picture}(370,246)(0,39)
1713:                  \put(0,155){\begin{picture}(0,0)
1714:   \put(64,0)     {\Includeourbeautifulpicture 80a }
1715:   \put(300,0)    {\Includeourbeautifulpicture 80b }
1716:   \put(0,59)     {$ \xi^\alpha_{(i\mu j)\nu} \;\equiv $}
1717:   \put(127,59)   {$\dsty := \sum_{l,m,n\in\II}\,
1718:                  \sum_{\gamma_1,\gamma_2,\delta_1,\delta_2}
1719:                  [\xi^\alpha_{(i\mu j)\nu}]^{lmn}_{\gamma_1 \gamma_2 \delta_1 \delta_2} $}
1720:   \put(32,33)  {} 
1721:  %%label for 80a
1722:   \put(81,-7)    {\scriptsize$ X_\mu $}
1723:   \put(82,122)   {\scriptsize$ X_\nu $}
1724:   \put(61,-7)    {\scriptsize$ U_i $}
1725:   \put(103,-7)   {\scriptsize$ U_j $}
1726:   \put(95,59)    {\scriptsize$ \alpha $}
1727:  %%label for 80b
1728:   \put(315,-7)   {\scriptsize$ X_\mu $}
1729:   \put(315,122)  {\scriptsize$ X_\nu $}
1730:   \put(307,20)   {\scriptsize$ \overline\gamma_1 $}
1731:   \put(307,95)   {\scriptsize$ \gamma_2 $}
1732:   \put(330,47)   {\scriptsize$ \delta_1 $}
1733:   \put(300,77)   {\scriptsize$ \delta_2 $}
1734:   \put(321,84)   {\scriptsize$ n $}
1735:   \put(321,63)   {\scriptsize$ m $}
1736:   \put(314.5,32) {\scriptsize$ l $}
1737:   \put(297,-7)   {\scriptsize$ U_i $}
1738:   \put(334,-7)   {\scriptsize$ U_j $}
1739:                  \end{picture}}
1740:                  \put(0,0){\begin{picture}(0,0)
1741:   \put(64,0)     {\Includeourbeautifulpicture 81a }
1742:   \put(300,0)    {\Includeourbeautifulpicture 81b }
1743:   \put(0,59)     {$ \bar\xi_\alpha^{(i\mu j)\nu} \;\equiv $}
1744:   \put(127,59)   {$\dsty := \sum_{l,m,n\in\II}\,
1745:                  \sum_{\gamma_1,\gamma_2,\delta_1,\delta_2}
1746:                  [\bar\xi_\alpha^{(i\mu j)\nu}]^{lmn}_{\gamma_1 \gamma_2 \delta_1 \delta_2} $}
1747:                  \end{picture}}
1748:  %%label for 81a
1749:   \put(82,122.8) {\scriptsize$ X_\mu $}
1750:   \put(80,-7)    {\scriptsize$ X_\nu $}
1751:   \put(61,122.8) {\scriptsize$ U_i $}
1752:   \put(103,122.8){\scriptsize$ U_j $}
1753:   \put(95,57)    {\scriptsize$ \overline{\alpha} $}
1754:  %%label for 81b
1755:   \put(315,122)  {\scriptsize$ X_\mu $}
1756:   \put(313.5,-7) {\scriptsize$ X_\nu $}
1757:   \put(307,20)   {\scriptsize$ \overline\gamma_2 $}
1758:   \put(307,95)   {\scriptsize$ \gamma_1 $}
1759:   \put(307,47)   {\scriptsize$ \overline{\delta_2} $}
1760:   \put(325,79)   {\scriptsize$ \overline{\delta_1} $}
1761:   \put(313,84)   {\scriptsize$ l $}
1762:   \put(321,54)   {\scriptsize$ m $}
1763:   \put(312,28)   {\scriptsize$ n $}
1764:   \put(297,122)  {\scriptsize$ U_i $}
1765:   \put(334,122)  {\scriptsize$ U_j $}
1766:   \epicture33 \labl{eq:xi-expansion} \xlabp 80 \xlabp 81
1767: Substituting the expansion \erf{eq:psi-expansion}
1768: into the ribbon invariant \erf{eq:GA-rib-inv} for $\GG[A]$ and
1769: taking the trace on both sides of \erf{eq:GA-rib-inv} immediately leads to
1770:   \be
1771:   \GG[A]^{(\mu i j)\, \nu}_{\alpha\kappa\beta , \gamma k \delta}
1772:   = \sum_{m,n,r}
1773:   \Frac{\dim(U_n)}{\dim(\dot M_\nu)}
1774:   \sum_{\stackrel{\gamma_1, \gamma_2, \gamma_3, \gamma_4,}{\eps_1,\eps_2,\eps_3}}
1775:   [\psi^\alpha_{(\mu i)\kappa}]^{mr}_{\gamma_2 \gamma_3 \eps_2}
1776:   [\psi^\beta_{(\kappa j)\nu}]^{rn}_{\gamma_3 \gamma_1 \eps_3}
1777:   [\bar\psi_\gamma^{(\mu k)\nu}]^{mn}_{\gamma_2 \gamma_1 \eps_1}
1778:   \GG^{(mij)\,n}_{\eps_2 r \eps_3 , \eps_1 k \delta} \,.
1779:   \labl{eq:GA-basis}
1780: For the fusion matrices of the bimodule category one must substitute 
1781: \erf{eq:xi-expansion} into \erf{eq:FAA-rib-inv}. 
1782: Using also the braiding relation (I:2.41), this results in 
1783:   %  [ pic 4_82 ]
1784:   \bea \begin{picture}(420,191)(0,23)
1785:   \put(305,-2)   {\includeourbeautifulpicture 82 }
1786:   \put(-14,130)  {$\dsty \FF[A|A]^{(ij\mu kl)\,\nu}_{\alpha\pi\beta,
1787:                  \gamma q_1 q_2 \delta_1 \delta_2} 
1788:                  \,=\, \Frac{1}{\dim(\dot X_\nu^{\phantom|})} 
1789:                  \sum_{x_1,\dots,x_6} \sum_{\varphi_1,\varphi_2,\varphi_3}
1790:                  \sum_{\eps_1,\dots,\eps_6} $}
1791:   \put(17,97)    {$\dsty [\xi^\alpha_{(j\mu k)\pi}]^{x_5 x_4 x_3}
1792:                  _{\varphi_3\varphi_2\eps_4\eps_3}\,
1793:                  [\xi^\beta_{(i\pi l)\nu}]^{x_3 x_2 x_1}
1794:                  _{\varphi_2\varphi_1\eps_2\eps_1}\,
1795:                  [\bar\xi_\gamma^{(q_1 \mu q_2)\nu}]^{x_5 x_6 x_1}
1796:                  _{\varphi_3\varphi_1\eps_5\eps_6}\!
1797:                  \sum_{\rho_1} \RR^{-(lk)\,q_2}_{\rho_1 \delta_2} $}
1798:   \put(323.5,210){\scriptsize$ x_1 $}
1799:   \put(334.5,202){\scriptsize$ \epsilon_1 $}
1800:   \put(332,183)  {\scriptsize$ x_2 $}
1801:   \put(353,172)  {\scriptsize$ \epsilon_2 $}
1802:   \put(300,160)  {\scriptsize$ i $}
1803:   \put(312,118.5){\scriptsize$ \delta_1 $}
1804:   \put(316,70)   {\scriptsize$ q_1 $}
1805:   \put(329,25.5) {\scriptsize$ \epsilon_6 $}
1806:   \put(323.5,9)  {\scriptsize$ x_1 $}
1807:   \put(343,33)   {\scriptsize$ x_6 $}
1808:   \put(348,56)   {\scriptsize$ \epsilon_5 $}
1809:   \put(366,88.5) {\scriptsize$ \rho_1 $}
1810:   \put(362,64)   {\scriptsize$ q_2 $}
1811:   \put(381,120)  {\scriptsize$ l $}
1812:   \put(343,80)   {\scriptsize$ x_5 $}
1813:   \put(355,112)  {\scriptsize$ \epsilon_4 $}
1814:   \put(362,98)   {\scriptsize$ k $}
1815:   \put(336,145)  {\scriptsize$ \epsilon_3 $}
1816:   \put(318,130)  {\scriptsize$ j $}
1817:    \put(332,160) {\scriptsize$ x_3 $}
1818:   \put(344,130)  {\scriptsize$ x_4 $} 
1819:   \epicture13 \labl{eq:FAA-calc-rib} \xlabp 82
1820: Applying the fusing relation (I:2.40) first to the vertex labelled $\epsilon_3$
1821: and then to the vertex labelled $\epsilon_4$ finally yields
1822:   \be  \bearll \displaystyle
1823:   \FF[A|A]^{(ij\mu kl)\,\nu}_{\alpha\pi\beta, \gamma q_1 q_2 \delta_1 \delta_2}
1824:   = \!\!& \displaystyle
1825:   \sum_{n_1,\dots,n_6} 
1826:   \Frac{\dim(U_{n_1})}{\dim(\dot X_\nu)} \sum_{\varphi_1,\varphi_2,\varphi_3}
1827:   \sum_{\eps_1,\dots,\eps_6}
1828:   \\{}\\[-.6em]& \displaystyle \hspace*{2.5em}
1829:   [\xi^\alpha_{(j\mu k)\pi}]^{n_5 n_4 n_3}_{\varphi_3\varphi_2\eps_4\eps_3}
1830:   [\xi^\beta_{(i\pi l)\nu}]^{n_3 n_2 n_1}_{\varphi_2\varphi_1\eps_2\eps_1}
1831:   [\bar\xi_\gamma^{(q_1 \mu q_2)\nu}]^{
1832:      n_5 n_6 n_1}_{\varphi_3\varphi_1\eps_5\eps_6}
1833:   \\[10pt]& \displaystyle \hspace*{2.5em}
1834:   \sum_{\rho_1,\rho_2,\rho_3} \RR^{-(lk)\,q_2}_{\rho_1 \delta_2} \,
1835:   \GG^{(j n_4 l)\, n_2}_{\eps_3 n_3 \eps_2\,,\,\rho_2 n_6 \rho_3}
1836:   \GG^{(n_5 k l)\,n_6}_{\eps_4 n_4 \rho_3 \,,\,\eps_5 q_2 \rho_1}
1837:   \FF^{(ij n_6)\,n_1}_{\eps_1 n_2 \rho_2\,,\,\delta_1 q_1 \eps_6} \,.
1838:   \eear\labl{eq:FAA-basis}
1839: In writing the expansions \erf{eq:psi-expansion} and \erf{eq:xi-expansion}, as 
1840: well as the expressions \erf{eq:GA-basis} and \erf{eq:FAA-basis} for the fusing 
1841: matrices, we have included all multiplicity indices. In the Cardy case 
1842: $A\eq\one$ treated below, as well as later on in section \ref{sec:rc-Cardy}, 
1843: we make the simplifying assumption that the fusion rules in $\calc$ obey 
1844: $\N ijk \iN \{0,1\}$ so as to somewhat reduce the notational burden.
1845: 
1846: 
1847: \subsection{Example: The Cardy case}
1848: 
1849: Let us now work out the fusion matrices for the module and bimodule categories
1850: in the case $A\eq\one$. This choice for the algebra $A$ corresponds to having 
1851: the charge conjugation modular invariant as the partition function on the
1852: torus. As already mentioned, for further simplification we assume in the
1853: sequel that the fusion rules of the modular tensor category $\calc$ satisfy 
1854: $\N ijk\iN\{0,1\}$; this allows us to suppress many of the multiplicity labels.
1855: 
1856: The simple $\one$-modules are just the simple objects of $\calc$,
1857: so that $\calca \,{\cong}\, \calc$ as categories. We choose the 
1858: representatives of simple modules as $M_\mu\eq U_\mu$ with $\mu\iN\II$. 
1859: Similarly, for the bimodules we have $\calcaa \,{\cong}\,\calc$, this time
1860: even as tensor categories, and we choose the representatives of simple bimodules
1861: as $X_\mu\eq U_\mu$, $\mu\iN\II$. Note that we continue to use greek letters to 
1862: label simple modules and bimodules to avoid confusion, even though now the 
1863: three index sets $\II$, $\JJ$ and $\KK$ are identical.
1864: 
1865: Next we must specify the basis elements $\psi^\alpha_{(\mu k)\nu}$ and 
1866: $\xi^\gamma_{(i\mu j)\nu}$ that we will use. We make the choices
1867:   %  [ pic 4_72 4_73 ]
1868:   \bea \begin{picture}(365,72)(0,28)
1869:   \put(80,8)     {\includeourbeautifulpicture 72 }
1870:   \put(300,0)    {\includeourbeautifulpicture 73 }
1871:   \put(0,40)     {$ \psi_{(\mu k)\nu} \,:=\, \beta^{\mu,\nu}_k $}
1872:   \put(150,40)   {and $\qquad \xi^\gamma_{(i \mu j) \nu} \,:=\,
1873:                  t_j^{\,-1}\, {\alpha^{\mu|\nu}_{ij,\gamma}} $}
1874:   \put(77,-1)    {\scriptsize$ U_\mu $}
1875:   \put(98,-1)    {\scriptsize$ U_k $}
1876:   \put(86.5,78)  {\scriptsize$ U_\nu $}
1877:   \put(316,-7)   {\scriptsize$ U_\mu $}
1878:   \put(297,-7)   {\scriptsize$ U_i $}
1879:   \put(335,-7)   {\scriptsize$ U_j $}
1880:   \put(320,43.5) {\scriptsize$ U_\gamma $}
1881:   \put(316,94)   {\scriptsize$ U_\nu $} 
1882:   \epicture15 \labl{eq:cardy-bases} \xlabp 72 \xlabp 73
1883: where the (non-zero) constants $\beta^{\mu,\nu}_k$ and $\alpha^{\mu|\nu}_{ij,
1884: \gamma}$ can be chosen at will; later on they will serve as normalisations of 
1885: the bulk, boundary and defect fields. The phases $t_j$ appearing in 
1886: \erf{eq:cardy-bases} have been defined in (II:3.43). If $\calc$ is the \rep\
1887: category of a conformal vertex algebra we take $t_j\eq\exp(-\pi\ii\Delta_j)$
1888: with $\Delta_j$ the conformal highest weight of the \rep\ labelled by $j$.
1889: The phases $t_j$ are convenient because they will lead to simpler expressions 
1890: for the bulk two-point function on the sphere and for the bulk one-point 
1891: function on the disk.
1892: 
1893: Because of $\psi_{(\mu k)\nu} \iN \HomA(M_\mu \Oti U_k, M_\nu) \,{\cong}\, 
1894: \Hom(U_\mu\Oti U_k, U_\nu)$, there is no need for a multiplicity label in 
1895: the first formula in \erf{eq:cardy-bases}. In contrast, the dimension of the
1896: morphism space $\HomAA(U_i{\otimes}^+ X_\mu {\otimes}^- U_j,X_\nu)
1897: \eq \Hom(U_i \oti U_\mu \oti U_j, U_\nu)$ can be larger than one, and
1898: accordingly the basis elements $\xi^\gamma_{(i \mu j) \nu}$ are labelled by 
1899: those simple objects $U_\gamma$ that occur in the fusion 
1900: $U_\mu \oti U_j$ and for which $\N i\gamma\nu \,{\ne}\, 0$.
1901: In the notation introduced in \erf{eq:psi-expansion}
1902: and \erf{eq:xi-expansion} the basis choice \erf{eq:cardy-bases} reads
1903:   \be
1904:   [\psi_{(\mu k)\nu}]^{mn}_{\cdot\cdot\cdot}
1905:   = \beta^{\mu,\nu}_k \, \delta_{m\mu} \,\delta_{n\nu}
1906:   \qquad{\rm and}\qquad
1907:   [\xi^\gamma_{(i\mu j)\nu}]^{xyz}_{\cdot\cdot\cdot\cdot}
1908:   = t_j^{\,-1} \, \alpha^{\mu|\nu}_{ij,\gamma}\, \delta_{\mu x}\, \delta_{\gamma y}\,
1909:   \delta_{\nu z} \,.
1910:   \labl{eq:cardy-bas-rel}
1911: The dot `$\,\cdot\,$' stands for a multiplicity label that can take only a
1912: single value, the corresponding morphism space being one-dimensional.
1913: 
1914: We also need the duals of the bases \erf{eq:cardy-bases}. One
1915: easily checks that these are given by
1916:   %  [ pic 4_74 4_75 ]
1917:   \bea \begin{picture}(350,72)(0,28)
1918:   \put(83,8)     {\includeourbeautifulpicture 74 }
1919:   \put(280,0)    {\includeourbeautifulpicture 75 }
1920:   \put(0,40)     {$\dsty \bar\psi^{(\mu k)\nu}\,=\,\frac{1}{\beta^{\mu,\nu}_k} $}
1921:   \put(150,40)   {and $\qquad\dsty \bar\xi_\gamma^{(i \mu j) \nu} \,=\, 
1922:                  \frac{t_j}{\alpha^{\mu|\nu}_{ij,\gamma}} $}
1923:   \put(79,78.5)  {\scriptsize$ U_\mu $}
1924:   \put(100,78.5) {\scriptsize$ U_k $}
1925:   \put(90.2,0)   {\scriptsize$ U_\nu $} 
1926:   \put(296,93)   {\scriptsize$ U_\mu $}
1927:   \put(277,93)   {\scriptsize$ U_i $}
1928:   \put(314,93)   {\scriptsize$ U_j $}
1929:   \put(300.8,40) {\scriptsize$ U_\gamma $} 
1930:   \put(295,-8)   {\scriptsize$ U_\nu $} 
1931:   \epicture15 \labp 74 \xlabp 75 
1932: Substituting the choice of basis \erf{eq:cardy-bases} into 
1933: \erf{eq:GA-rib-inv} immediately yields 
1934:   \be
1935:   \GG[\one]^{(\mu i j)\,\nu}_{\kappa \,,\, k} =
1936:   \frac{\beta^{\kappa,\nu}_j \beta^{\mu,\kappa}_i}{\beta^{\mu,\nu}_k}
1937:   \,\GG^{(\mu i j)\,\nu}_{\kappa\,k} 
1938:   \labl{eq:GA-bas-Cardy}
1939: for the fusion matrices of $\calca$. 
1940: 
1941: For the fusion matrices of the 
1942: bimodule category we deduce from \erf{eq:FAA-rib-inv} that
1943:   %  [ pic 4_76 ]
1944:   \bea \begin{picture}(330,178)(0,50)
1945:   \put(75,0)     {\Includeourbeautifulpicture 76a }
1946:   \put(239,0)    {\Includeourbeautifulpicture 76b }
1947:   \put(0,106)    {$ \FF[\one|\one]^{(ij\mu kl)\,\nu}_{\alpha\pi\beta,\gamma pq} $}
1948:   \put(99,106)   {$\dsty =\ \frac{
1949:                    \alpha^{\pi|\nu}_{il,\beta}\, \alpha^{\mu|\pi}_{jk,\alpha}}
1950:                    {\alpha^{\mu|\nu}_{pq,\gamma}}\,
1951:                    \Rs{-}lk{\,q}\,\frac{t_q}{t_l\,t_k} $}
1952:   \put(72,-7.7)  {\scriptsize$ U_\nu $}
1953:   \put(72,224)   {\scriptsize$ U_\nu $}
1954:   \put(264,224)  {\scriptsize$ U_\nu $}
1955:   \put(264,-7.7) {\scriptsize$ U_\nu $}
1956:   \put(278,35)   {\scriptsize$ \gamma $}
1957:   \put(297,65)   {\scriptsize$ q $}
1958:   \put(251,60)   {\scriptsize$ p $}
1959:   \put(269,80)   {\scriptsize$ \mu $}
1960:   \put(235,160)  {\scriptsize$ i $}
1961:   \put(270,185)  {\scriptsize$ \beta $}
1962:   \put(270,160)  {\scriptsize$ \pi $}
1963:   \put(253,130)  {\scriptsize$ j $}
1964:   \put(316,130)  {\scriptsize$ l $}
1965:   \put(279,130)  {\scriptsize$ \alpha $}
1966:   \put(297,98)   {\scriptsize$ k $} 
1967:   \epicture36 \labp 76
1968: Similarly to the calculation in \erf{eq:FAA-calc-rib}, applying the fusion 
1969: relation (I:2.40) first to the vertex for $\Hom(U_j\oti U_\alpha,U_\pi)$ and
1970: then to the one for $\Hom(U_\mu\oti U_k,U_\alpha)$ results in the expression
1971:   \be
1972:   \FF[\one|\one]^{(ij\mu kl)\, \nu}_{\alpha \pi \beta, \gamma p q} \,=\, 
1973:   \frac{ \alpha^{\pi|\nu}_{il,\beta} \, \alpha^{\mu|\pi}_{jk,\alpha}}
1974:   {\alpha^{\mu|\nu}_{pq,\gamma}} \,
1975:   \Rs{-}lk{\,q} \frac{t_q}{t_l\,t_k} \, \Gs j\alpha l\beta\pi\gamma \,
1976:   \Gs \mu kl\gamma \alpha q \, \Fs ij\gamma\nu\beta p .
1977:   \labl{eq:FAA-bas-Cardy}
1978: 
1979: 
1980: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1981: 
1982: \sect{Ribbon graph representation of field insertions}\label{sec:fi}
1983: 
1984: 
1985: In this section we present the ribbon graph representation
1986: for field insertions on boundaries, in the bulk and on defect lines.
1987: This will be used in section \ref{sec:rc} to express structure 
1988: constants as invariants of ribbon graphs in three-manifolds.
1989: 
1990: \medskip
1991: 
1992: The ribbon graph representations will be given explicitly
1993: only for the case of unoriented world sheets. Recall from \cite{fuRs8} that
1994: in this case the relevant algebraic structure to describe
1995: the full CFT is a Jandl algebra. 
1996: 
1997: It is then straightforward to obtain the oriented case as well. Of course, 
1998: given an oriented world sheet, one can just forget the orientation to obtain 
1999: an unoriented world sheet, for which one can apply the construction below. 
2000: However, this is not the correct procedure to use, since in the two cases 
2001: different algebraic structures are relevant:
2002: 
2003: \begin{center}\begin{tabular}{lcl}
2004:   oriented world sheets & $\rightarrow$ & 
2005:     symmetric special Frobenius algebra \\[3pt]
2006:   unoriented world sheets  & $\rightarrow$ &  Jandl algebra
2007: \end{tabular}\end{center}
2008: 
2009: \noindent
2010: Thus for oriented world sheets one must formulate 
2011: the construction of the ribbon graphs in a way that is applicable for 
2012: {\em any\/} symmetric special Frobenius algebra, not just for those
2013: admitting a reversion. This can be done as follows.
2014: 
2015: First, as already emphasised, instead of a Jandl algebra, just a symmetric 
2016: special Frobenius  algebra is to be used.  Second, whenever in the unoriented 
2017: case an orientation $\orr_1$ of a boundary component, respectively $\orr_2$ 
2018: of a surface, enters the construction, then in the corresponding oriented
2019: case a canonical choice is provided by the orientation of the world sheet.
2020: In the un\-ori\-en\-ted case there are equivalence relations
2021: linking the two possible choices for $\orr_{1,2}$, and this
2022: is in fact the only place where the reversion of the Jandl algebra 
2023: enters. In the oriented case these equivalence relations
2024: are not needed and hence only the symmetric special Frobenius
2025: structure is used.
2026: 
2027: For example, recall from section II:3 that in the unoriented case
2028: boundary conditions are labelled by equivalence classes
2029: $[M,\orr_1]$, where $M$ is a left $A$-module and $\orr_1$ is
2030: an orientation of the boundary component. The equivalence relation (II:3.1) 
2031: is that $(M,\orr_1) \,{\sim}\, (M',\orr_1')$
2032: if either $\orr_1'\eq\orr_1$ and $M'\,{\cong}\, M$
2033: or if $\orr_1'\eq{-}\orr_1$ and $M'\,{\cong}\, M^\sigma$, where
2034: $M^\sigma$ is the module conjugate to $M$ (see section II:2.3).
2035: To obtain the prescription in the oriented case, we use the
2036: orientation $\orr_{\partial \X}$ of the boundary that is induced by the world
2037: sheet orientation $\orr_\X$ to select the pair $(M,\orr_{\partial \X})$. 
2038: The second alternative in the equivalence relation is then obsolete.
2039: 
2040: 
2041: \subsection{A note on conventions}\label{sec:fi-conv}
2042: 
2043: The construction of the ribbon graph described in this section departs slightly
2044: from the one used in \cite{fuRs4} and \cite{fuRs8}, resulting from a number 
2045: of choices of conventions that have to be made. However, in the absence of 
2046: field insertions the final result, i.e.\ the ribbon 
2047: graph embedded in the connecting manifold, is still the
2048: same as in \cite{fuRs4,fuRs8}. To see this let us summarise the conventions
2049: we choose and point out how they differ from those in \cite{fuRs4,fuRs8}.
2050: 
2051: 
2052: \subsubsection*{Category of three-dimensional cobordisms}
2053: 
2054: Recall from the summary in sections I:2.3 and I:2.4 that a three-dimensional
2055: topological field theory furnishes a functor $(Z,\calh)$ from the cobordism 
2056: category $\coborc$ to the category of finite-dimensional
2057: complex vector spaces; such a functor can be constructed from any modular 
2058: tensor category $\calc$. It turns out to be convenient to give a
2059: definition of the objects and morphisms of $\coborc$ slightly
2060: different from the one used in section I:2.4.
2061:   
2062: Objects of $\cobord$ are extended surfaces and morphisms 
2063: of $\cobord$ are cobordisms between extended surfaces.
2064: An {\em extended surface\/} $E$ consists of the following data:
2065: \ITemize
2066: \item[\Nxt] 
2067: A compact oriented two-dimensional manifold without
2068: boundary, also denoted by $E$.
2069: \item[\Nxt]
2070: A finite set of marked points -- that is, of quadruples
2071: $(p_i,[\gamma_i],V_i,\eps_i)$, where
2072: the $p_i\iN E$ are mutually distinct points of the surface $E$
2073: and $[\gamma_i]$ is a germ of arcs\,%
2074:   \footnote{~By a {\em germ of arcs\/} we mean an equivalence class $[\gamma]$
2075:   of continuous embeddings $\gamma$ of intervals 
2076:   $[-\delta,\delta] \,{\subset}\, \reals$ into the extended surface $E$. 
2077:   Two embeddings $\gamma{:}~[-\delta,\delta] \To E$ and
2078:   $\gamma'{:}~[-\delta',\delta'] \To E$ are equivalent if there is a 
2079:   positive $\eps\,{<}\,\delta,\delta'$ such that $\gamma$ and $\gamma'$
2080:   are equal when restricted to the interval $[-\eps,\eps]$.
2081:   }
2082: $\gamma_i{:}\; [-\delta,\delta] \To E$ with $\gamma_i(0)\eq p_i$.
2083: Furthermore, $V_i \iN \obj(\calc)$, and $\eps_i\iN\{\pm 1\}$ is a sign.
2084: \item[\Nxt]
2085: A Lagrangian subspace $\lambda \,{\subset}\, H_1(E,\reals)$.
2086: \end{itemize}
2087: A morphism $E \To E'$ is a {\em cobordism\/} $\M$, consisting of
2088: the following data:
2089: \begin{itemize}
2090: %
2091: \item[\Nxt]
2092: A compact oriented three-manifold, also denoted by $\M$, such that
2093: $\partial \M \eq(-E){\sqcup}E'$.
2094: \\
2095: Here $-E$ is obtained from $E$ by reversing its $2$-orientation
2096: and replacing any marked point $(p,[\gamma],U,\eps)$ by
2097: $(p,[\tilde\gamma],U,-\eps)$ with $\tilde\gamma(t)\eq\gamma(-t)$,
2098: see section I:2.4. The boundary $\partial \M$ of a cobordism is 
2099: oriented according to the inward pointing normal.
2100: %
2101: \item[\Nxt]
2102: A ribbon graph $R$ inside $\M$ such that for each marked point
2103: $(p,[\gamma],U,\eps)$ of $(-E){\sqcup}E'$ there is a ribbon ending on
2104: $(-E){\sqcup}E'$ in the following way.
2105: Let $\varphi{:}\; [0,1]\Times[-\frac1{10},\frac1{10}]\To \M$ be the
2106: parametrisation\,\footnote{%
2107:   ~By a {\em parametrisation\/} of a ribbon $S$ we mean the following.
2108:   $S$ has a $2$-orientation of its surface and a $1$-orientation of its core.
2109:   The rectangle $[0,1]\Times[-\frac1{10},\frac1{10}]$ carries a natural
2110:   $2$-orientation as a subset of $\reals^2$, and the interval
2111:   $[0,1]\Times\{0\}$ carries a natural $1$-orientation as a subset
2112:   of the $x$-axis. A parametrisation $\varphi$ of $S$ is
2113:   required to preserve both of these orientations.
2114:   }
2115: of a ribbon $S$ embedded in $\M$ and labelled by $U\iN\objc$. If $\eps\eq{+}1$,
2116: then the core of $S$ must point away from $\partial \M$ and the end of $S$
2117: must induce the germ $[\gamma]$, i.e.\ $\varphi(0,0)\eq p$ and
2118: $[\varphi(0,t)]\eq [\gamma(t)]$. If $\eps\eq{-}1$, then the core of $S$
2119: must point towards $\partial \M$, and $\varphi(1,0)\eq p$ as well as
2120: $[\varphi(1,-t)] \eq [\gamma(t)]$, the presence of the minus sign implying
2121: that the orientation of the arc is reversed.
2122: \end{itemize}
2123: The difference to the definition used in section I:2.4 is that we do not
2124: take a surface with embedded arcs, but rather with germs of arcs.\,%
2125:   \footnote{~This convention should be compared to those of \cite{TUra}, which
2126:   uses the arcs themselves, and of \cite{BAki}, which uses tangential vectors
2127:   at the insertion points $p$.
2128:                       %see section 4.4 there
2129:   The present convention is a compromise between the two. Only the
2130:   behaviour in an arbitrarily small neighbourhood of $p$ is relevant, but it
2131:   does not need a differential structure on $E$.
2132:   }
2133: The reason for using this definition of extended surface will become clear in
2134: section \ref{sec:cb-link} where we assign a germ of arcs to a germ of local
2135: coordinates.
2136: 
2137: Note also that the allowed ways for a ribbon 
2138: to end on an arc-germ of an extended surface depend on the orientations
2139: of arc, ribbon-surface and ribbon-core, but not on the 
2140: orientation of the three-manifold $\M$ or of the extended surface $E$.
2141: The two allowed possibilities are -- looked at from `above' and `below'
2142: $E$, as well as from `in front of' and `behind' the ribbon -- the following:
2143:   %  [ pic 4_08 ]
2144:   \bea \begin{picture}(370,85)(0,41)
2145:   \put(0,75)    {\Includeourbeautifulpicture 08a }
2146:   \put(100,75)  {\Includeourbeautifulpicture 08b }
2147:   \put(200,75)  {\Includeourbeautifulpicture 08c }
2148:   \put(300,75)  {\Includeourbeautifulpicture 08d } 
2149:   \put(0,0)     {\Includeourbeautifulpicture 08e }
2150:   \put(100,0)   {\Includeourbeautifulpicture 08f }
2151:   \put(200,0)   {\Includeourbeautifulpicture 08g }
2152:   \put(300,0)   {\Includeourbeautifulpicture 08h }
2153:   \put(21,5.6)    {\scriptsize$(U,+) $}
2154:   \put(121,5.6)   {\scriptsize$(U,-) $}
2155:   \put(221,5.6)   {\scriptsize$(U,-) $}
2156:   \put(321,5.6)   {\scriptsize$(U,+) $}
2157:   \put(21,111)    {\scriptsize$(U,-) $}
2158:   \put(121,111)   {\scriptsize$(U,+) $}
2159:   \put(221,111)   {\scriptsize$(U,+) $}
2160:   \put(321,111)   {\scriptsize$(U,-) $}
2161:   \put(39,37)     {\scriptsize$ U $}
2162:   \put(139,37)    {\scriptsize$ U $}
2163:   \put(239,37)    {\scriptsize$ U $}
2164:   \put(339,37)    {\scriptsize$ U $} 
2165:   \put(39,77)     {\scriptsize$ U $}
2166:   \put(139,77)    {\scriptsize$ U $}
2167:   \put(239,77)    {\scriptsize$ U $}
2168:   \put(339,77)    {\scriptsize$ U $}
2169:   \epicture24 \labl{eq:ribbon-arc-end} \xlabp 08 
2170:       %  compare (I:2.55)
2171: (For instance, the first picture in the first row is the same as the fourth 
2172: picture in the first row, as well as the second and third picture in the 
2173: second row.)
2174: 
2175: 
2176: \subsubsection*{Ribbon graph embedded in the connecting manifold} 
2177: 
2178: In the TFT description of
2179: correlators for oriented world sheets, we have the following conventions:
2180: \begin{itemize} 
2181: \item[\Nxt] 
2182:   The correlator for a world sheet $X$ is described by a cobordism $\emptyset 
2183:   \,{\overset{\MX\,}{\longrightarrow}}\,\hat\X$, where $\hat\X$ is the double of
2184:   $\X$. In particular, $\hat\X$ is the {\em outgoing\/} part of the boundary
2185:   $\partial\MX$.
2186: \item[\Nxt] 
2187:   Field insertions on the world sheet $\X$ lead to marked points on the
2188:   double $\hat\X$. For the ribbon graph inside the cobordism $\emptyset 
2189:   \,{\overset{\MX\,}{\longrightarrow}}\, \hat\X$ we take all ribbons ending on 
2190:   $\partial\MX$ to have cores pointing {\em away\/} from the boundary, 
2191:   i.e.\ $\eps\eq{+}1$ for all marked points on $\hat\X$.
2192: \item[\Nxt] 
2193: The non-negative integers (compare formulas (I:5.30) and (I:5.65), and lemma 
2194: \ref{lem:defect-iso}) 
2195:   \be
2196:   Z(A)_{ij} := \dim\,\HomAA(\alpha^-_\AA(U_j),\alpha^+_\AA(U_{\bar\imath}))
2197:   = \dim\,\HomAA(U_i\oT+A\ot-U_j,A)
2198:   \,, \ee
2199: with $\alpha^\pm_\AA(U)$ denoting $\alpha$-induced\,%
2200:   \footnote{~For the notion of $\alpha$-induction, which is due to \cite{lore},
2201:   see e.g.\ definition 2.21 of \cite{ffrs}. The definition, as well as a list
2202:   of references, is also given in section I:5.4 where, however, the notation
2203:   $\alpha^{(\pm)}$ is used in place of $\alpha^\pm_\AA$. \label{alphafootnote} }
2204: bimodules, count the number of linearly independent primary bulk fields 
2205: $\vPhi_{ij}(z,\bar z)$ that have chiral representation label $i$ and
2206: anti-chiral representation label $j$.
2207: \item[\Nxt] 
2208: Boundary conditions are labelled by left $A$-modules.
2209: \end{itemize}
2210: As a consequence (compare (I:5.30) or \erf{eq:bulk-insert} below),
2211: when constructing the ribbon graph in $\MX$ for an oriented world sheet $\X$,
2212: the ribbons embedded in $\X$ must be inserted in such a manner that their 
2213: `white' side faces the boundary of $\MX$ which has the same orientation as $\X$. 
2214: In particular, the orientation of the ribbon-surfaces is {\em opposite\/} to 
2215: that of the world sheet $\X$.
2216: 
2217: This clashes with the prescription given in section I:5.1, 
2218:   %more precisely:  the second point in the list below equation (5.5)
2219: according to which orientation of the world sheet and the ribbons agree. The
2220: reason is that in \cite{fuRs4} implicitly, the boundary of a cobordism was taken to be
2221: oriented according to the {\em outward\/} pointing normal. However due to
2222: the conventions above we should use the {\em inward\/} pointing normal. 
2223: 
2224: It is not difficult to convince oneself that {\em all pictures and formulas
2225: in\/} \cite{fuRs4} {\em and\/} \cite{fuRs8} {\em remain unchanged\/} provided that 
2226: \begin{itemize}
2227: \item[--] one uses the
2228: {\em inward} pointing normal to orient the boundary of a cobordism,
2229: and that
2230: \item[--] the prescription in sections I:5 and II:3.1 is modified 
2231: such that ribbons embedded in $\X\,{\subset}\, \MX$ have surface orientation
2232: {\em opposite} to the (local) orientation of $\X$, and similarly
2233: their core has {\em opposite} orientation to $\partial \X$ if they
2234: lie on the world sheet boundary.
2235: \end{itemize}
2236: %
2237: {}From hereon we 
2238: use this modified prescription both in the oriented and in the unoriented case.
2239: 
2240: 
2241: \subsection{Boundary fields}\label{sec:fi-bnd}
2242: 
2243: A connected component $\bndb$ of the boundary $\partial \X$
2244: of the world sheet has the topology of a circle. On $\bndb$
2245: there may be several insertions of boundary fields, and accordingly the
2246: intervals between the field insertions may be labelled by
2247: different boundary conditions. Below we will see that the
2248: configurations of fields and boundary conditions on such a 
2249: boundary component are again labelled
2250: by equivalence classes, and that this labelling generalises
2251: the one for boundaries without field insertions.
2252: 
2253: In the sequel we say that two curves $\varpi$ and $\mu$ are {\em aligned} 
2254: in a point $p$ of the world sheet iff there exist parametrisations 
2255: $\varpi'(t)$ of $\varpi$ and 
2256: $\mu'(t)$ of $\mu$ such that $\varpi'(0) \eq p \eq \mu'(0)$ and 
2257: $[\varpi'] \eq [\mu']$.
2258: 
2259: 
2260: \subsubsection*{Labelling of boundary insertions}
2261: 
2262: A {\em boundary field} $\hatpsi$ is a collection 
2263:   \be
2264:   \hatpsi = ( M , N , V , \psi , p , [\gamma] )
2265:   \ee
2266: of the following data:
2267: $M, N$ are left $A$-modules, $V$ is an object of $\calc$,
2268: the morphism $\psi$ is an element of $\HomA(M\oti V,N)$,
2269: $p \iN \partial \X$ is a point on the boundary of the world sheet
2270: $\X$, and finally $[\gamma]$ is an arc germ such that $\gamma(0)\eq p$ and
2271: that $\gamma$ is aligned to $\partial \X$ in $p$. 
2272: 
2273: Denote by $\bndb([\gamma_1],...\,,[\gamma_n])$ a connected
2274: component of $\partial \X$ together with $n$ arc germs 
2275: $[\gamma_k]$ such that $\gamma_k$ is aligned to $\partial \X$ in
2276: $\gamma_k(0)$. The points $\gamma_k(0)$ mark the insertion points of the
2277: boundary fields. A boundary component 
2278: $\bndb([\gamma_1],...\,,[\gamma_n])$ with $n$ boundary field 
2279: insertions is labelled by equivalence classes of tuples
2280:   \be
2281:   (M_1, ...\, , M_{n'} ; \hatpsi_1, ...\, , \hatpsi_n ; \orr_1) 
2282:   \labl{eq:bnd-tupel}
2283: subject to the following conditions.
2284: \\
2285: \nxt $\orr_1$ is an orientation of the boundary circle $\bndb$.
2286: \\
2287: \nxt The points $p_k \eq \gamma_k(0)$ are ordered
2288: such that when passing along the boundary circle $\bndb$ 
2289: opposite to the direction 
2290: given by $\orr_1$, one passes from $p_k$ to $p_{k+1}$. (Here
2291: and below it is understood that $p_0 \eq p_n$ and 
2292: $p_{n+1} \eq p_1$, and analogously for $M_k$ and $\hatpsi_k$.)
2293: \\
2294: \nxt The $M_1,...\,,M_{n'}$ are left $A$-modules. $M_k$ 
2295: labels the boundary condition on the stretch of
2296: boundary between $p_{k-1}$ and $p_k$. Thus if there
2297: are no field insertions, then $n'\eq1$, and otherwise $n'\eq n$.
2298: \\
2299: \nxt The $\hatpsi_1,...\,,\hatpsi_n$ are boundary fields with
2300: defining data
2301:   \be
2302:   \hatpsi_k = ( M_k, M_{k+1} , V_k, \psi_k, p_k, [\gamma_k] ) \,.
2303:   \labl{eq:bnd-psik}
2304: 
2305: To define the equivalence relation we need 
2306: \\[-2.3em]
2307: 
2308: \dtl{Definition}{def:psi-sig}
2309: Let $A$ be a Jandl algebra in $\calc$.
2310: Given an object $V$ of $\calc$ and two left $A$-modules $M$ and $N$, 
2311: $s\,{\equiv}\,s_{M,N,V}$ is the map $s{:}\ \Hom(\dot M\oti V, \dot N) \To
2312: \Hom(\dot N^\vee \oti V , \dot M^\vee)$ defined by
2313:   %  [ pic 4_07 ]
2314:   \bea \begin{picture}(150,89)(0,50)
2315:   \put(56,0)     {\includeourbeautifulpicture 07 }
2316:   \put(0,56)     {$ s(\psi) \;:= $}
2317:   \put(87,130)   {\scriptsize$ \dot{M}^\vee $}
2318:   \put(92.5,-9)  {\scriptsize$ \dot{N}^\vee $}
2319:   \put(122.8,-7) {\scriptsize$ V $}
2320:   \put(93.4,72.5)  {\scriptsize$ \psi $}
2321:   \epicture27 \labl{eq:bnd-s-def} \xlabp 07
2322: for $\psi\iN\Hom(\dot M\oti V, \dot N)$.
2323: 
2324: \medskip
2325: 
2326: \dtl{Lemma}{lem:psi-s}
2327: Let $A$ and $s$ be as in definition \ref{def:psi-sig}. If
2328: $\psi\iN\HomA(M\oti V,N)$, then $s(\psi) \iN \HomA(N^\sigma\oti V,M^\sigma)$.
2329: 
2330: \medskip\noindent
2331: Proof:\\
2332: The assertion follows by direct computation from the defining properties
2333: of conjugate left modules, as given in definition II:2.6.
2334: \qed
2335: 
2336: We can now formulate the equivalence relation on labellings
2337: \erf{eq:bnd-tupel} of a boundary component $\bndb$ with
2338: (or without) field insertions. Consider two tuples
2339:   \be
2340:   (M_1, ...\, , M_{n'} ; \hatpsi_1, ...\, , \hatpsi_n ; \orr_1) 
2341:   \quad\ {\rm and} \quad\
2342:   (M'_1, ...\, , M'_{n'} ; \hatpsi'_1, ...\, , \hatpsi'_n ; \orr'_1) \,,
2343:   \labl{eq:bnd-equiv-def}
2344: where $\hatpsi_k\eq( M_k, M_{k+1} , V_k, \psi_k, p_k, [\gamma_k] )$ and
2345: $\hatpsi_k'\eq( M_k', M_{k+1}' , V'_k, \psi_k', p_k', [\gamma_k'] )$.
2346: We assume that the numbering of boundary insertions is done such that 
2347: either $p_k \eq p_k'$ or $p_k \eq p_{n-k}'$ (otherwise the numbering 
2348: must be shifted appropriately). 
2349: 
2350: \dtl{Definition}{bnd-equiv-def}
2351: The set of labels for the boundary component $\bndb([\gamma_1],...\,,[\gamma_n])$
2352: is the set
2353:   \be
2354:   \mathcal{B}\big(\bndb([\gamma_1],...\,,[\gamma_n]) \big)
2355:   := \big\{ (M_1, ...\, , M_{n'} ; \hatpsi_1, ...\, , 
2356:   \hatpsi_n ; \orr_1) \big\} /{\sim}
2357:   \ee
2358: of equivalence classes with respect to the following
2359: equivalence relation $\sim$\,: The two tuples \erf{eq:bnd-equiv-def} 
2360: are equivalent iff one of the following two conditions is satisfied:
2361: \\[3pt]
2362: (i) We have $\orr_1'\eq\orr_1$, $V_k'\eq V_k$, $p_k'\eq p_{k}$, 
2363: $[\gamma_k'] \eq [\gamma_{k}]$, and for $k\eq1,...\,,n'$ there exist isomorphisms 
2364: $\varphi_k \iN \HomA(M_k', M_k)$ such that 
2365:   \be
2366:   \psi_k = \varphi_{k+1} \cir \psi'_k \cir (\varphi_k^{-1} \oti \id_{V_k})\,.
2367:   \labl{eq:bnd-equiv-def-i}
2368: %\noindent
2369: (ii) $\orr_1' \eq {-}\orr_1$, $V_k' \eq V_{n-k}$, $p_k' \eq p_{n-k}$, 
2370: $[\gamma_k'] \eq [\gamma_{n-k}]$, and for $k\eq1,...\,,n'$ there exist isomorphisms
2371: $\varphi_k \iN \HomA(M_k', M_{n'-k+1}^\sigma)$ such that 
2372: \be
2373:   s(\psi_{n-k}) = \left\{ \begin{array}{ll}
2374:   \varphi_{k+1} \cir \psi'_k \cir 
2375:     (\varphi_{k}^{-1} \otimes \id_{V_k'})
2376:   ~ & {\rm if~~} \orr_1' \eq \orr(\gamma_k') \,, \\{}\\[-.8em]
2377:   \varphi_{k+1} \cir \psi'_k \cir 
2378:     (\varphi_{k}^{-1} \otimes \theta_{V_k'})
2379:   ~ & {\rm if~~} \orr_1' \eq {-}\orr(\gamma_k') \,,  \eear\right.
2380:   \labl{eq:bnd-equiv-def-ii}
2381: where $\orr(\gamma_k')$ denotes the local orientation of $\bndb$
2382: induced by the arc-germ $[\gamma_k']$.
2383: 
2384: 
2385: \subsubsection*{Marked points on the double}
2386: 
2387: Every boundary field insertion $\hatpsi\eq(M,N,V,\psi,p,[\gamma])$ on $\X$ gives
2388: rise to one marked point on the double $\hat\X$. This marked point is given by
2389:   \be
2390:   (\tilde p, [\tilde \gamma], V, +)
2391:   \ee
2392: with $\tilde p\eq [p,\pm \orr_2(p)]$. Recall from \erf{eq:double-def}
2393: that $\hat\X$ is the orientation bundle over $\X$ with the two
2394: orientations identified over the boundary $\partial \X$. To obtain the germ 
2395: $[\tilde\gamma]$, first choose a representative $\gamma$ of $[\gamma]$. Since 
2396: $\gamma$ is aligned with $\partial\X$ at $p$, by definition there exists a
2397: $\delta\,{>}\,0$ such that $\gamma {:}\; [-\delta,\delta] \To \partial \X$.
2398: A representative $\tilde\gamma{:}\; [-\delta,\delta] \To \hat \X$ of
2399: $[\tilde\gamma]$ is then given by $\tilde\gamma(t)\,{:=}\, [\gamma(t),\pm\orr_2]$.
2400: 
2401: Given a labelling $B\iN\mathcal{B}\big(\bndb([\gamma_1],...\,,[\gamma_n]) \big)$,
2402: one can verify that the set of marked points on $\hat \X$ obtained in the
2403: way described above is indeed independent of the choice of representative of $B$.
2404: 
2405: 
2406: \subsubsection*{Ribbon graph embedded in the connecting manifold}
2407: 
2408: Note that in a small neighbourhood of a point $p$ on the boundary
2409: of the world sheet $\X$, the connecting manifold $\MX$ from (II:3.8)
2410: looks as follows:
2411:   %% [pic~09]
2412:   \bea  \begin{picture}(280,66)(0,32)
2413:   \put(2,0)         {\includeourbeautifulpicture 09 }
2414:   \put(202,39)         {\scriptsize $p$}
2415:   \epicture20 \labl{eq:con-mf-bnd} \xlabp 09
2416: Indicated are also the intervals $[-1,1]$, which according to the identification
2417: rule of the general prescription degenerate into an interval $[0,1]$ above each
2418: boundary point. The ribbon graph representation of the boundary component
2419: $\bndb([\gamma_1],...\,,[\gamma_n])$ labelled by an equivalence class 
2420: $B \eq [M_1, ...\, , M_{n'} ; \hatpsi_1, ...\, , \hatpsi_n ; \orr_1]$ in 
2421: $\mathcal{B}\big(\bndb([\gamma_1],...\,,[\gamma_n]) \big)$
2422: is now constructed as follows.
2423: 
2424: \ITemize
2425: \item[\Nxt] 
2426: Choose  a representative 
2427: $(M_1, ...\, , M_{n'} ; \hatpsi_1, ...\, , \hatpsi_n ; \orr_1)$ of 
2428: the equivalence class $B$ -- \underline{Choice \#1}.
2429: 
2430: \item[\Nxt] 
2431: Choose a dual triangulation of $\X$ such that the insertion points of $\bndb$ 
2432: lie on edges of the triangulation -- \underline{Choice \#2}.
2433: 
2434: \item[\Nxt] 
2435: The orientation $\orr_1$ of the boundary circle $\bndb$ together with the 
2436: inward pointing normal induces an orientation of its neighbourhood in the 
2437: world sheet $\X$. This fixes a local orientation of $\X$ close to $\bndb$, which
2438: agrees with both the bulk and the boundary orientation of the upper half plane.
2439: 
2440: \item[\Nxt] At each vertex of the triangulation which lies
2441: on the boundary $\bndb$ place the element\,\footnote{%
2442:   ~Note that the orientations in this picture are opposite to those in
2443:   (II:3.12). This is due to the change in convention
2444:   explained in section \ref{sec:fi-conv}.} 
2445:   %% [pic~43]
2446:   \bea  \begin{picture}(195,87)(0,37)
2447:   \put(2,0)         {\includeourbeautifulpicture 43 }
2448:                 \put(-1.5,19){
2449:   \put(20,89.5)     {\tiny$y$}
2450:   \put(-1,67.3)     {\tiny$x$}
2451:   \put(12.3,24.8)   {\scriptsize$ \dot M $}
2452:   \put(92,94.3)     {\scriptsize$ A $}
2453:   \put(148.3,24.8)  {\scriptsize$ \dot M $}
2454:                            } 
2455:   \epicture24 \labl{eq:bnd-vertex} \xlabp 43
2456: in such a way that the orientation of bulk and boundary agree with that
2457: around the vertex.
2458: 
2459: \item[\Nxt] 
2460: At each insertion point $\hatpsi_k \eq (M_k, M_{k+1}, V_k, \psi_k, p_k , 
2461: [\gamma_k])$ place one of the two elements
2462:   %  [ pic 4_10 ]
2463:   \bea \begin{picture}(360,57)(0,45)
2464:   \put(0,0)       {\Includeourbeautifulpicture 10a }
2465:   \put(200,0)     {\Includeourbeautifulpicture 10b }
2466:   \put(37,64)     {\scriptsize$ M_k $}
2467:   \put(129,65)    {\scriptsize$ M_{k+1} $}
2468:   \put(88,54.5)   {\scriptsize$ \psi_k $}
2469:   \put(67,23)     {\scriptsize$ V_k $} 
2470:   \put(244,64)    {\scriptsize$ M_k $}
2471:   \put(328,65)    {\scriptsize$ M_{k+1} $}
2472:   \put(288.1,54.5){\scriptsize$ \psi_k $}
2473:   \put(267,38)    {\scriptsize$ V_k $} 
2474:   \put(17,96.5)   {\tiny$y$}
2475:   \put(-5,74.3)   {\tiny$x$}
2476:   \put(217,96.5)  {\tiny$y$}
2477:   \put(195,74.3)  {\tiny$x$}
2478:   \epicture26 \labl{eq:bnd-ribbon} \xlabp 10
2479: depending on the orientation of the arc-germ $[\gamma_k]$ relative to that of 
2480: the boundary (recall also the convention in \erf{eq:ribbon-arc-end}). Shown in 
2481: \erf{eq:bnd-ribbon} is a horizontal section of the connecting manifold 
2482: \erf{eq:con-mf-bnd}. Correspondingly the lower boundary in \erf{eq:bnd-ribbon} 
2483: is that of $\MX$ while the ribbons $M_k$ and $M_{k+1}$ are placed on the 
2484: boundary of $\X$ as embedded in $\MX$. The arrow on the boundary in 
2485: \erf{eq:bnd-ribbon} indicates the orientation of $\partial\X$ (transported 
2486: to $\partial\MX$ along the preferred intervals). The ribbon graph must be 
2487: placed in the plane of the world sheet $\X$ (embedded in $\MX$) in such a 
2488: way that the bulk and boundary orientation of \erf{eq:bnd-ribbon} agree 
2489: with the local orientations at the insertion point $p_k\iN\X$. 
2490: \end{itemize}
2491: 
2492: \noindent
2493: The prescription involves two choices, and we proceed to
2494: show that different choices lead to equivalent ribbon graphs.
2495: \\[5pt]
2496: \nxt\underline{Choice \#2}:\\[2pt]
2497: As in sections I:5.1 and II:3.1, independence of the triangulation follows from 
2498: the identities (I:5.11) and (I:5.12). However, if there are field insertions on 
2499: the boundary we need an additional move to relate any one dual triangulation to 
2500: any other dual triangulation. It consists of taking a vertex of the 
2501: triangulation past a boundary insertion,
2502:   %  [ pic 4_11 ]
2503:   \bea \begin{picture}(352,57)(0,44)
2504:   \put(0,0)     {\Includeourbeautifulpicture 11a }
2505:   \put(184,0)   {\Includeourbeautifulpicture 11b }
2506:   \put(164,33)  {$ = $}
2507:   \put(10,63)   {\scriptsize$ M $}
2508:   \put(67,63)   {\scriptsize$ M $}
2509:   \put(140,63)  {\scriptsize$ N $}
2510:   \put(195,63)  {\scriptsize$ M $}
2511:   \put(329,63)  {\scriptsize$ N $}
2512:   \put(51,84)   {\scriptsize$ A $}
2513:   \put(295,98)  {\scriptsize$ A $}
2514:   \put(87,54)   {\scriptsize$ \psi $}
2515:   \put(246,54)  {\scriptsize$ \psi $}
2516:   \put(65,34)   {\scriptsize$ V $}
2517:   \put(225,34)  {\scriptsize$ V $}
2518:   \put(267,46)  {\scriptsize$ N $}
2519:   \epicture24 \labl{eq:bnd-ch2} \xlabp 11
2520: together with a similar identity for an insertion of (\ref{eq:bnd-ribbon}\,b).
2521: The dashed circle in \erf{eq:bnd-ch2} represents the piece of ribbon graph 
2522: inside the dashed circle in \erf{eq:bnd-vertex}. This identity is a direct 
2523: consequence of the fact that the morphism $\psi$ in $\hatpsi \eq (M,N,V,\psi,
2524: p,[\gamma])$ is an intertwiner of $A$-modules, $\psi \iN \HomA(M \Oti V, N)$.
2525: \\[5pt]
2526: %
2527: \nxt\underline{Choice \#1}:
2528: \\[2pt]
2529: Denote by 
2530: $R \eq (M_1, ...\, , M_{n'} ; \hatpsi_1, ...\, , \hatpsi_n ; \orr_1)$ and
2531: $R' \eq (M'_1, ...\, , M'_{n'} ; \hatpsi'_1, ...\, , \hatpsi'_n ; \orr'_1)$
2532: two representatives of $B$. 
2533: We will apply the above construction to the representative $R'$ and
2534: show that the resulting ribbon graph can be turned into the one obtained from
2535: $R$. According to definition \ref{bnd-equiv-def} we must consider two cases.
2536: \\[3pt]
2537: (i)~\,$\orr_1' \eq \orr_1$. Then also $V_k' \eq V_k$, $p_k' \eq p_{k}$ and
2538: $[\gamma_k'] \eq [\gamma_{k}]$, and there exists a choice of isomorphisms 
2539: $\varphi_k \iN \HomA(M_k', M_k)$ fulfilling \erf{eq:bnd-equiv-def-i}.
2540: Starting with the ribbon graph constructed from $R'$, we insert on
2541: each ribbon $M_k'$ the identity morphism in the form
2542: $\id_{M_k'} \eq \varphi_k^{-1} \cir \varphi_k$, in such a way that
2543: $\varphi_k^{-1}$ comes next to the field insertion $\hatpsi_k'$.
2544: Then we move the morphism $\varphi_k$ to the other end of the $M_k'$-ribbon. 
2545: As in (II:3.17), $\varphi_k$ can be taken past vertices of the
2546: triangulation, which may lie on the boundary interval labelled by
2547: $M_k'$. After these manipulations, every field insertion
2548: $\hatpsi_k'$ comes with the morphism $\varphi_k^{-1}$ to one side and 
2549: with $\varphi_{k+1}$ to the other. In other words, for each $k$ the morphism
2550: $\psi_k'$ appearing in $\hatpsi_k'$ is replaced by
2551: $\varphi_{k+1} \cir \psi_k' \cir (\varphi_k^{-1} \oti \id_{V_k})
2552: \,{\equiv}\,\psi_k$. We thus remain with the ribbon graph constructed from $R$.
2553: \\[4pt]
2554: (ii)~$\orr_1' \eq {-}\orr_1$. Then also $V_k' \eq V_{n-k}$, $p_k' \eq p_{n-k}$
2555: and $[\gamma_k'] \eq  [\gamma_{n-k}]$, and there exist isomorphisms 
2556: $\varphi_k \iN \HomA(M_k', M_{n'-k+1}^\sigma)$, for $k \eq 1,...\,,n'$,
2557: fulfilling \erf{eq:bnd-equiv-def-ii}. We shall demonstrate the equivalence of 
2558: ribbon graphs in some detail for the second case in \erf{eq:bnd-equiv-def-ii}; 
2559: the first case can be seen similarly. Around the field insertion $\hatpsi_k'$ 
2560: the ribbon graph obtained from $R'$, which we denote by $[R']$, looks as follows:
2561:   %  [ pic 4_12 ]
2562:   \bea \begin{picture}(220,39)(0,30)
2563:   \put(55,0)     {\includeourbeautifulpicture 12 }
2564:   \put(0,34)     {$ [R'] \ = $} 
2565:   \put(165,66)   {\scriptsize$ M'_k $}
2566:   \put(70,66)    {\scriptsize$ M'_{k+1} $}
2567:   \put(142,40)   {\scriptsize$ V'_k $}
2568:   \put(136.8,4)  {\scriptsize$ p'_k $}
2569:   \put(74,-2)    {\scriptsize$ \orr'_1 $}
2570:   \put(76,23)    {\scriptsize$ \orr_1 $}
2571:   \put(108.4,54) {\scriptsize$ \psi'_k $}
2572:   \epicture15 \labl{eq:bnd-equiv-def-iii} \xlabp 12
2573: The two opposite orientations of the boundary are indicated. We are facing the 
2574: black side of the ribbons because in order to make the local orientations of 
2575: \erf{eq:bnd-equiv-def-iii} induced by $\orr_1'$ agree with those in 
2576: (\ref{eq:bnd-ribbon}\,b), the element (\ref{eq:bnd-ribbon}\,b) must be turned 
2577: `upside down' before it is inserted. As in case (i) we insert 
2578: $\id_{M_l'} \eq \varphi_l^{-1}{\circ}\,\varphi_l$
2579: on each $M_l'$, close to $\hatpsi_l'$. After applying in addition
2580: a half-twist and using that, by definition, $\dot M^\sigma \eq \dot M^\vee$ as
2581: an object, this leads to
2582:   %  [ pic 4_13 ]
2583:   \bea \begin{picture}(310,44)(0,34)
2584:   \put(55,0)     {\includeourbeautifulpicture 13 }
2585:   \put(0,34)     {$ [R'] \ = $}    
2586:   \put(140,66)   {\scriptsize$ M'_k $}
2587:   \put(70,66)    {\scriptsize$ M'_{k+1} $}
2588:   \put(197,67)   {\scriptsize$ M_{n-k+1} $}
2589:   \put(272,66)   {\scriptsize$ M'_k $}  
2590:   \put(142,40)   {\scriptsize$ V'_k $}
2591:   \put(136.2,4)  {\scriptsize$ p'_k $}
2592:   \put(73,-2)    {\scriptsize$ \orr'_1 $}
2593:   \put(77,23)    {\scriptsize$ \orr_1 $}
2594:   \put(108.6,55) {\scriptsize$ \psi'_k $}
2595:   \put(166,58.6) {\scriptsize$ \varphi^{-1}_k $}
2596:   \put(247.8,59) {\scriptsize$ \varphi_k $} 
2597:   \epicture21 \labp 13
2598: By a reasoning similar to that of (II:3.18) this specific combination
2599: of half-twist and $\varphi_k$ can be taken past vertices of the
2600: triangulation that lie on the boundary, i.e.\ past locations at which
2601: $A$-ribbons arriving from the bulk are attached to the module-ribbons
2602: on the boundary. Taking all of the $\varphi_l$ to the neighbouring
2603: field insertion leads to a ribbon graph which close to $\hatpsi_k'$ looks as 
2604:   %  [ pic 4_14 ]
2605:   \bea \begin{picture}(300,48)(0,39)
2606:   \put(55,0)     {\includeourbeautifulpicture 14 }
2607:   \put(0,39)     {$ [R'] \ = $}    
2608:   \put(125,64)   {\scriptsize$ M'_{k+1} $}
2609:   \put(61,65)    {\scriptsize$ M_{n-k} $}
2610:   \put(258,65)   {\scriptsize$ M_{n-k+1} $}
2611:   \put(187,65.7) {\scriptsize$ M'_k $} 
2612:   \put(188,40)   {\scriptsize$ V'_k $}
2613:   \put(182.4,4)  {\scriptsize$ p'_k $}
2614:   \put(73,-2)    {\scriptsize$ \orr'_1 $}
2615:   \put(77,23)    {\scriptsize$ \orr_1 $}
2616:   \put(153.5,54) {\scriptsize$ \psi'_k $}
2617:   \put(212,59)   {\tiny$ \varphi^{-1}_k $}
2618:   \put(105,59)   {\tiny$ \varphi_{k+1} $}
2619:   \epicture29 \labp 14
2620: The morphism $\varphi_{k+1}$ has arrived from the insertion
2621: $\hatpsi_{k+1}'$ which is not visible in the section of the
2622: full ribbon graph $[R']$ that we display. Next rotate
2623: the coupons labelled by $\varphi_{k+1}$, $\psi_k'$ and
2624: $\varphi_{k}^{-1}$ by $180^{\circ}$ in the manner already indicated in
2625: figure \erp 14, and deform the resulting ribbon graph slightly,
2626:  % (drawing out the ribbons explicitly helps), so as to obtain
2627:   %  [ pic 4_57 ]
2628:   \bea \begin{picture}(370,62)(0,47)
2629:   \put(55,0)     {\includeourbeautifulpicture 57 }
2630:   \put(0,53)     {$ [R'] \ = $}    
2631:   \put(57,79)    {\scriptsize$ M_{n-k} $}
2632:   \put(327,80)   {\scriptsize$ M_{n-k+1} $}
2633:   \put(217,72)   {\scriptsize$ M'_{k+1} $}
2634:   \put(176,78)   {\scriptsize$ M'_k $} 
2635:   \put(108.3,34) {\scriptsize$ V'_k\eq V_{n-k} $}
2636:   \put(102,5)    {\scriptsize$ p'_k=p_{n-k} $}
2637:   \put(71,-2)    {\scriptsize$ \orr'_1 $}
2638:   \put(76,23)    {\scriptsize$ \orr_1 $} 
2639:   \put(202,63.8) {\tiny{$ \psi'_k $}}
2640:   \put(160,73)   {\tiny{$ \varphi^{-1}_k $}}
2641:   \put(240.9,65) {\tiny{$ \varphi_{k+1} $}}
2642:   \epicture33 \labp 57
2643: By the second alternative in \erf{eq:bnd-equiv-def-ii}, the morphism in the 
2644: dashed box is equal to $s(\psi_{n-k})$.  After replacing the dashed box by 
2645: the explicit form of $s(\psi_{n-k})$ in \erf{eq:bnd-s-def},
2646: the ribbon graph can be deformed so as to obtain the 
2647: ribbon graph constructed from the representative $R$ of $B$.
2648: 
2649: \medskip
2650: \noindent
2651: We have thus established that the ribbon graph that is obtained
2652: by our prescription for a boundary component 
2653: $\bndb([\gamma_1],...\,,[\gamma_n])$ labelled by an equivalence class in 
2654: $\mathcal{B}\big(\bndb([\gamma_1],...\,,[\gamma_n])\big)$ 
2655: is independent of the choices involved.
2656: 
2657: 
2658: \subsection{Bulk fields} \label{sec:fi-bulk}
2659: 
2660: A bulk field insertion is described by a point $p$ in the
2661: interior of $\X$, together with an arc-germ $[\gamma]$ at $p$
2662: and a label $\PHI$. The set of labels for bulk insertions
2663: at $[\gamma]$ will be denoted by $\mathcalD([\gamma])$. Similarly as for 
2664: boundary fields, this set consists of equivalence classes of certain tuples.
2665: 
2666: 
2667: \subsubsection*{Labelling of bulk fields}
2668: 
2669: The equivalence relation will be formulated in terms of a linear
2670: map $\omega^A_{UV}$ that is defined as follows.
2671: \\[-2.3em]
2672: 
2673: \dtl{Definition}{def:bulk-omega}
2674: For $A$ a Jandl algebra in $\calc$, the morphism $\omega^A_{UV}{:}\; 
2675: \Hom(U\oti A\oti V, A) \To \Hom(V\oti A\oti U, A)$ is defined by
2676:   %  [ pic 4_92 ]
2677:   \bea \begin{picture}(170,90)(0,43)
2678:   \put(75,0)     {\includeourbeautifulpicture 92 }
2679:   \put(0,60)     {$ \omega^A_{UV}(\phi) \ := $}    
2680:   \put(99.3,46.4) {\scriptsize$ \sigma^{\!-\!1} $}
2681:   \put(102.4,87.9){\scriptsize$ \sigma $}
2682:   \put(101,125.7){\scriptsize$ A $}
2683:   \put(87,-8)    {\scriptsize$ V $}
2684:   \put(100.5,-8) {\scriptsize$ A $}
2685:   \put(115,-8)   {\scriptsize$ U $}
2686:   \put(102,68.3) {\scriptsize$ \phi $}
2687:   \epicture27 \labl{eq:bulk-om-def} \xlabp 92
2688: for $\phi\iN\Hom(U\oti A\oti V, A)$.
2689: 
2690: \medskip
2691: 
2692: Some properties of the map $\omega^A_{UV}$ that will be needed below 
2693: are listed in the following lemma. 
2694: \\[-2.2em]
2695: 
2696: \dtl{Lemma}{lem:omega-prop}
2697: With $A$ and $\omega^A_{UV}$ as in definition \ref{def:bulk-omega}, we have:
2698: \\[.2em]
2699: (i)~\,For any $\phi\iN\Hom(U\oti A\oti V, A)$ one
2700: has $\omega^A_{VU}\big(\omega^A_{UV}(\phi)\big) \eq \phi$.
2701: \\[.3em]
2702: (ii)~If $\phi\iN\HomAA(U\oT+A\ot-V,A)$,
2703: then $\omega^A_{UV}(\phi) \iN \HomAA(V\oT+A\ot-U,A)$.
2704: 
2705: \medskip\noindent
2706: Proof:\\
2707: (i)~\,Writing out the ribbon graph for the morphism $\omega^A_{VU}\big(
2708: \omega^A_{UV}(\phi)\big)$ and using $\sigma\cir\sigma\eq\theta_A$,
2709: one can deform the resulting ribbon graph (which amounts to using that
2710: $\calc$ is sovereign) so as to be left with the graph for $\phi$.
2711: \\[.3em]
2712: (ii)~One must verify that the morphism $\omega^A_{UV}(\phi)$ commutes
2713: with the left and right action of $A$ if $\phi$ itself does.
2714: This can be done by a straightforward computation using the
2715: definitions \erf{UXUX} and \erf{XUXU} of the left and right action of $A$
2716: as well as the defining properties (II:2.6) of the reversion $\sigma$ on $A$.
2717: \qed
2718: 
2719: \dtl{Definition}{def:bulk-PHI}
2720: Let $[\gamma]$ be an arc-germ around a point $p\eq\gamma(0)$ in
2721: the interior of $\X$. The possible bulk fields that can be
2722: inserted at $p$ are labelled by equivalence classes 
2723:   \be
2724:   \mathcalD([\gamma]) = \big\{
2725:   (i,j,\phi,p,[\gamma],\orr_2(p))\big\} /{\sim} \,.
2726:   \ee
2727: Here $i,j \iN \II$ label simple objects,
2728: $\phi$ is a morphism in $\HomAA(U_i\,{\otimes^+}A\,{\otimes^-}U_j,A)$,
2729: the insertion point $p$ is at $\gamma(0)$, and $\orr_2(p)$ is a local
2730: orientation of the world sheet $\X$ around $p$. The
2731: equivalence relation is defined as
2732:   \be
2733:   \big(i,j,\phi,p,[\gamma],\orr_2(p)\big) \sim 
2734:   \big(j,i,\omega^A_{U_iU_j}(\phi),p,[\gamma],-\orr_2(p)\big)\,.
2735:   \ee
2736: 
2737: 
2738: \subsubsection*{Marked points on the double}
2739: 
2740: Each bulk field insertion $\PHI$ on the world sheet $\X$ gives
2741: rise to {\em two\/} marked points on the double $\hat \X$. If
2742: $\big(i,j,\phi,p,[\gamma],\orr_2(p)\big)$ is a representative
2743: of $\PHI$, then these two points are
2744:   \be
2745:   (\tilde p_i, [\tilde\gamma_i] , U_i , +) \qquad {\rm and} \qquad
2746:   (\tilde p_j, [\tilde\gamma_j] , U_j , +) \,;
2747:   \labl{eq:bulk-marked-Xh}
2748: the terms appearing in these tuples are given by $\tilde p_i \eq [p,\orr_2(p)]$, 
2749: $\tilde p_j \eq [p,-\orr_2(p)]$,
2750: as well as $\tilde\gamma_i(t) \eq [\gamma(t),\orr_2(\gamma(t))]$ and
2751: $\tilde\gamma_j(t) \eq [\gamma(t),-\orr_2(\gamma(t))]$. Here $\orr_2(\gamma(t))$ 
2752: is obtained by the extension of $\orr_2(p)$ to a neighbourhood of $p$.
2753: 
2754: Just like in the case of boundary fields, one verifies that the marked points 
2755: \erf{eq:bulk-marked-Xh} on $\hat \X$ are independent of the choice of
2756: representative of $\PHI$.
2757: 
2758: 
2759: \subsubsection*{Ribbon graph embedded in the connecting manifold}
2760: 
2761: The ribbon graph representation of a bulk field $\PHI$ is
2762: constructed as follows.
2763: 
2764: \ITemize
2765: \item[\Nxt] 
2766: Choose  a representative $(i,j,\phi,p,[\gamma],\orr_2(p))$
2767: of $\PHI$ -- \underline{Choice \#1}.
2768: 
2769: \item[\Nxt] Choose a dual triangulation of $\X$ such that the insertion point 
2770: $p$ lies on an edge and such that the arc $\gamma$ is aligned with this edge 
2771: at $p\eq\gamma(0)$ -- \underline{Choice \#2}.
2772: 
2773: \item[\Nxt] The bulk insertion is treated as a two-valent vertex that is to
2774: be included in the dual triangulation. Locally around the insertion
2775: point $p$, the world sheet has the orientation $\orr_2$. In
2776: $\MX$ place the following ribbon graph:
2777:   %  [ pic 4_93 + 4_94 ]
2778:   \begin{eqnarray}
2779:      \begin{picture}(420,240)(5,0)
2780:   \put(-20,100)  {\Includeourbeautifulpicture 94a }
2781:   \put(130,0)    {\Includeournicelargepicture 94b }
2782:   \put(350,227)    {$\partial\MX$}
2783:   \put(341,19)     {$\partial\MX$}
2784:   \put(291,234)    {\tiny $x$}
2785:   \put(284,221)    {\tiny $y$}
2786:   \put(26,105.5)   {\tiny $x$}
2787:   \put(-4.4,119)   {\tiny $y$}
2788:   \put(56,127)     {\scriptsize $\gamma$}
2789:   \put(36,115)     {\scriptsize $p$}
2790:   \put(316,192)    {\scriptsize$ U_j$}
2791:   \put(310,41)     {\scriptsize$ U_i$}
2792:   \put(175,128)    {\scriptsize$A$}
2793:   \put(237,128)    {\scriptsize$A$}
2794:   \put(349,128)    {\scriptsize$A$}
2795:   \put(295,120)    {$\phi$}
2796:   \put(330,168.1)     {$t\eq0$-plane}
2797:   \put(220,63)     {\tiny $x$}
2798:   \put(199,77)     {\tiny $y$}
2799:   \put(390,30)     {\tiny $x$}
2800:   \put(374.3,50)   {\tiny $z$}
2801:   \put(363,46)     {\tiny $y$}
2802:   \put(279,16)     {\tiny $y$}
2803:   \put(297.8,4)    {\tiny $x$}
2804:   \put(112,120)  {$ \longmapsto $}
2805:   \end{picture} \nonumber
2806:   \\                 \begin{picture}(420,265)(5,0)
2807:   \put(130,0)    {\includeournicelargepicture 93 }
2808:   \put(110,120)  {$ = $}
2809:   \put(350,227)    {$\partial\MX$}
2810:   \put(343,14)     {$\partial\MX$}
2811:   \put(291,234)    {\tiny $x$}
2812:   \put(284,221)    {\tiny $y$}
2813:   \put(316,192)    {\scriptsize$ U_i$}
2814:   \put(310,40)     {\scriptsize$ U_j$}
2815:   \put(190,111)    {\scriptsize$A$}
2816:   \put(248,111)    {\scriptsize$A$}
2817:   \put(349,111)    {\scriptsize$A$}
2818:   \put(295,120)    {$\phi$}
2819:   \put(330,168.1)     {$t\eq0$-plane}
2820:   \put(422,65)     {\tiny $x$}
2821:   \put(428,77)     {\tiny $y$}
2822:   \put(390,30)     {\tiny $x$}
2823:   \put(374.3,50)   {\tiny $z$}
2824:   \put(363,46)     {\tiny $y$}
2825:   \put(279,16)     {\tiny $y$}
2826:   \put(297.8,4)    {\tiny $x$}
2827:   \end{picture} \nonumber 
2828:   \\[-1.5em]{} \label{eq:bulk-insert}
2829:   \end{eqnarray} \xlabp 93 \xlabp 94
2830: In this figure we display the ribbon graph twice, once looked upon from 
2831: `above' the connecting manifold and once looked at from `below'. In any case, 
2832: the orientation of the $A$-ribbons is opposite to that of the world sheet and 
2833: their white side faces the insertion point of the chiral field label $U_i$.
2834: The ribbon graph \erf{eq:bulk-insert} must be placed in $\MX$ in such a manner 
2835: that the embedding respects the orientation of the three-manifold $\MX$, the 
2836: local orientation $\orr_2$ of the world sheet $\X$, as well as the orientation 
2837: of the arcs on the boundary of $\MX$. There is precisely one way to do this; 
2838: no further choice needs to be made.
2839: \end{itemize}
2840: 
2841: \noindent
2842: We proceed to establish independence of the two choices.
2843: \\[5pt]
2844: \nxt\underline{Choice \#2}:\\[2pt]
2845: Independence of the dual triangulation follows from the identities (I:5.11)
2846: together with a move that allows to take three-valent vertices of the 
2847: triangulation past the two-valent vertex formed by the bulk field insertion
2848:   %  [ pic 4_18 ] 
2849:   \bea \begin{picture}(420,48)(0,4) 
2850:   \put(0,0)     {\Includeourbeautifulpicture 18a 
2851:                  \put(105,8){ $\phi$ }
2852:                 }
2853:   \put(230,0)   {\Includeourbeautifulpicture 18b 
2854:                  \put(62,8){ $\phi$ }
2855:                 }
2856:   \put(202,11)  {$ = $}
2857:   \epicture-4 \labl{eq:bulk-ch2} \xlabp 18
2858: and together with similar identities for different local orientations and 
2859: for situations where the $A$-ribbon arrives from the other side. The empty 
2860: dashed circle in picture \erf{eq:bulk-ch2} stands for the element (II:3.11), 
2861: the dashed rectangular box for (II:3.13), and the dashed circle with
2862: inscribed $\phi$ for the element \erf{eq:bulk-insert}.
2863: Analogously as for (I:5.11) and (I:5.12),\,% 
2864:     \footnote{~For detailed derivations, see \cite{fjfrs}.}
2865: the equality follows in a straightforward way upon substituting the definitions
2866: and using that the morphism $\phi$ is an intertwiner of $A$-bimodules.
2867: \\[5pt]
2868: \nxt\underline{Choice \#1}:\\[2pt]
2869: Let $(i,j,\phi,p,[\gamma],\orr_2)$ and $(j,i,\phi',p,[\gamma],-\orr_2)$ 
2870: be the two representatives of $\PHI$, i.e.\ we have 
2871: $\phi'\eq\omega^A_{U_iU_j}(\phi)$. We must show that
2872:   %  [ pic 4_19 ]
2873:   \bea \begin{picture}(350,15)(0,12) 
2874:   \put(0,0)     {\Includeourbeautifulpicture 19a 
2875:                  \put(62,8){ $\phi$ }
2876:                 }
2877:   \put(199,0)   {\Includeourbeautifulpicture 19b 
2878:                  \put(60,8){ $\phi'$ }
2879:                 }
2880:   \put(166,9)   {$ = $}
2881:   \epicture01 \labl{eq:bulk-ch1} \xlabp 19
2882: The dashed boxes stand for (II:3.13) and (II:3.14), respectively, 
2883: while the dashed circle with $\phi$ or $\phi'$ inscribed stands for the 
2884: element \erf{eq:bulk-insert}, inserted with the appropriate orientation.
2885: 
2886: To facilitate drawing, let us rotate the middle plane of \erf{eq:bulk-insert} 
2887: into the paper plane and use the blackboard framing convention. Equation 
2888: \erf{eq:bulk-ch1} then looks as follows.
2889:   %  [ pic 4_95 ]
2890:   \bea \begin{picture}(420,75)(8,43)
2891:   \put(0,0)      {\Includeourbeautifulpicture 95a }
2892:   \put(215,0)    {\Includeourbeautifulpicture 95b }
2893:   \put(190,57)   {$ = $}
2894:   \put(30,62.8)  {\scriptsize$ A $}
2895:   \put(115,63.5) {\scriptsize$ A $}
2896:   \put(150,63.5) {\scriptsize$ A $}
2897:   \put(215,63.5) {\scriptsize$ A $}
2898:   \put(390,63.5) {\scriptsize$ A $}
2899:   \put(80,59)    {\scriptsize$ \phi $}
2900:   \put(294,59)   {\scriptsize$ \phi' $}
2901:   \put(47.7,94)  {\scriptsize$ i $}
2902:   \put(269.5,95) {\scriptsize$ i $}
2903:   \put(270,35)   {\scriptsize$ j $}
2904:   \put(46.7,38)  {\scriptsize$ j $}
2905:   \put(235,58.8) {\scriptsize$ \sigma $}
2906:   \put(344.5,58.8){\scriptsize$ \sigma $} 
2907:   \epicture22 \labl{eq:bulk-ch1-b} \xlabp 95
2908: In the right figure one must insert the element \erf{eq:bulk-insert} `upside 
2909: down' because the local orientation of the world sheet is inverted with 
2910: respect to the left figure. As compared to \erf{eq:bulk-ch1}, on both sides 
2911: of the equation one coproduct has been removed against one product using
2912: the Frobenius property. In the right figure of \erf{eq:bulk-ch1-b} the central 
2913: part of the ribbon graph can be rotated by $180^\circ$ as indicated. Using 
2914: also $\theta_A^{-1}\cir\sigma\eq\sigma^{-1}$ this leads to
2915:   %  [ pic 4_96 ]
2916:   \bea \begin{picture}(320,77)(8,46)
2917:   \put(65,0)     {\includeourbeautifulpicture 96 }
2918:   \put(0,57)     {$ {\rm \erf{eq:bulk-ch1-b}}\ = $}
2919:   \put(64,63.2)  {\scriptsize$ A $}
2920:   \put(239.4,66) {\scriptsize$ A $}
2921:   \put(271,66)   {\scriptsize$ A $} 
2922:   \put(175.5,59) {\scriptsize$ \phi' $} 
2923:   \put(108,93)   {\scriptsize$ i $}
2924:   \put(112,38)   {\scriptsize$ j $}
2925:   \put(207,80)   {\scriptsize$ j $} 
2926:   \put(155.5,59) {\scriptsize$ \sigma^{-1} $}
2927:   \put(196.5,59) {\scriptsize$ \sigma $}
2928:   \epicture26 \labl{eq:bulk-ch1-c} \xlabp 96
2929: Comparing with \erf{eq:bulk-om-def} we see that
2930: the part of the ribbon graph inside the dashed box amounts to the morphism
2931:   \be
2932:   { \raisebox{-3pt}{\dashbox{3.5}(63,14){{\rm dashed~box}}} }_{\phantom|} \;=
2933:   \omega^A_{U_{j_{\phantom|}^{}}\!U_i}(\phi') = \phi \,,
2934:   \ee
2935: where the last equality follows from the equivalence relation for bulk labels 
2936: which forces $\phi'\eq\omega^A_{U_iU_j}(\phi)$, together with
2937: lemma \ref{lem:omega-prop}. Thus \erf{eq:bulk-ch1-c} is equal to the left hand 
2938: side of \erf{eq:bulk-ch1-b}, establishing the claimed equality between the 
2939: left hand side and the right hand side of \erf{eq:bulk-ch1-b}.
2940: 
2941: \medskip\noindent
2942: This finishes the demonstration that the ribbon graph for a bulk field 
2943: insertion is independent of the choices involved in its construction.
2944: 
2945: 
2946: \subsection{Defect fields}\label{sec:fi-defect}
2947: 
2948: Together with the defect fields treated in this
2949: section, the three types of fields we consider are
2950:   %  [ pic 4_22 ]
2951:   \bea \begin{picture}(340,65)(0,8)
2952:   \put(22,18)    {\Includeourbeautifulpicture 22a }
2953:   \put(125,0)    {\Includeourbeautifulpicture 22b }
2954:   \put(260,14)   {\Includeourbeautifulpicture 22c }
2955:   \put(0,55)     { {\em bulk field} }
2956:   \put(113,55)   { {\em boundary field} }
2957:   \put(252,55)   { {\em defect field} }
2958:   \epicture-5 \labl{eq:fields-consd} \xlabp 22
2959: Note, however, that this is not the most general way to treat defects. 
2960: In fact, upon forming operator products, the first two kinds of fields in
2961: \erf{eq:fields-consd} close amongst each other. But once we include also 
2962: the defect field, we can generate the following more general
2963: class of fields by taking operator products,
2964:   %  [ pic 4_23 ] 
2965:   \bea \begin{picture}(220,49)(0,7)
2966:   \put(9,4)      {\Includeourbeautifulpicture 23a }
2967:   \put(155,0)    {\Includeourbeautifulpicture 23b }
2968:   \epicture-1 \labl{eq:fields-not-consd} \xlabp 23
2969: i.e.\ a boundary field with several defect lines attached and a defect field 
2970: with more than two defect lines ending on it. For example, a field of the 
2971: first kind can be obtained by taking the operator product of a boundary 
2972: field and a defect field of the types listed in \erf{eq:fields-consd}. 
2973: While fields of the type \erf{eq:fields-not-consd} and their operator 
2974: products can be treated in the TFT-formalism in a natural way, in this paper 
2975: we will nonetheless only consider fields of the types \erf{eq:fields-consd}.
2976: 
2977: 
2978: \subsubsection*{Labelling of defect fields}
2979: 
2980: Recall from section II:3.8 that a circular defect $\dftd$ without
2981: field insertions and with orientable neighbourhood is labelled
2982: by equivalence classes of triples $[X,\orr_1,\orr_2]$, where
2983: $X$ is an $A$-bimodule, $\orr_1$ is an orientation of the
2984: defect circle and $\orr_2$ is an orientation of its neighbourhood.
2985: The equivalence relation is given in (II:3.150).
2986: 
2987: In order to describe the labelling of a defect circle with field insertions 
2988: $\dftd([\gamma_1],...\,,[\gamma_n])$ we first need the following definition.
2989: \\[-2.3em]
2990: 
2991: \dtl{Definition}{def:defect-equiv}
2992: The linear maps
2993:   \beaa\displaystyle
2994:   & u :\quad \Hom(U_i \oti X\oti U_j, Y) \to \Hom(U_j \oti X\oti U_i, Y) 
2995:   \\{}\\[-.8em]  {\rm and} \quad& \displaystyle
2996:   \tilde u :\quad \Hom(U_i \oti X\oti U_j, Y) \to
2997:       \Hom(U_i \oti Y^\vee{\otimes}\, U_j, X^\vee) 
2998:   \eear\ee
2999: are defined by
3000:   %  [ pic 4_97 ]   
3001:   \bea \begin{picture}(360,93)(0,50)
3002:   \put(290,0)    {\Includeourbeautifulpicture 97a }
3003:   \put(60,0)     {\Includeourbeautifulpicture 97b }
3004:   \put(230,0){
3005:    \put(67,-9.5)   {\scriptsize$U_i$}
3006:    \put(114,-9.5)  {\scriptsize$U_j$}
3007:    \put(90,-9.5)   {\scriptsize$ Y^\vee $}
3008:    \put(91,71)     {\small $\vartheta$}
3009:    \put(90,144)    {\scriptsize$ X^\vee $}
3010:   }
3011:   \put(-230,0){
3012:    \put(334,-9.5)  {\scriptsize$U_i$} 
3013:    \put(295,-9.5)  {\scriptsize$U_j$}
3014:    \put(315,-9.5)  {\scriptsize$X$}
3015:    \put(316.5,71)  {\small $\vartheta$}
3016:    \put(316,144.5) {\scriptsize$Y$}
3017:   }
3018:   \put(0,68)       {$ u(\vartheta) \;:= $}
3019:   \put(170,64)     { and }
3020:   \put(230,64)     {$ \tilde u(\vartheta) \;:= $}
3021:   \epicture39 \labl{eq:uv-def} \xlabp 97
3022: for $\vartheta\iN\Hom(U_i \oti X\oti U_j, Y)$.
3023: 
3024: \medskip
3025: 
3026: The maps $u$ and $\tilde u$ will enter into the formulation of the
3027: equivalence relation for defect labels. This also requires 
3028: the following lemma, which can be verified by direct calculation.
3029: \\[-2.3em]
3030: 
3031: \dtl{Lemma}{lem:defect-equiv}
3032: Let $A$ be an algebra and let $X,Y$ be $A$-bimodules. 
3033: Then $\tilde u$ restricts to an isomorphism
3034:   \be
3035:   \tilde u :\quad \HomAA(U_i\,{\otimes^+}\,X\,{\otimes^-}\,U_j,Y) \to
3036:   \HomAA(U_i\,{\otimes^+}\,Y^v\,{\otimes^-}\,U_j,X^v) \,.
3037:   \ee
3038: If $A$ is a Jandl algebra, then in addition $u$ restricts to an isomorphism
3039:   \be
3040:   u :\quad \HomAA(U_i\,{\otimes^+}\,X\,{\otimes^-}\,U_j,Y) \to
3041:   \HomAA(U_j\,{\otimes^+}\,X^s\,{\otimes^-}\,U_i,Y^s) \,.
3042:   \ee
3043: Here $X^v,Y^v$ and $X^s,Y^s$ denote two of the three 
3044: duals one has for the bimodules $X,Y$, as defined in (II:2.40).
3045: 
3046: \medskip
3047: 
3048: After these preliminaries we turn to the description of defect labels.
3049: Let $A$ be a Jandl algebra. A {\em defect field\/} $\Thetha$ 
3050: is a collection of data
3051:   \be
3052:   \Thetha = \big(\,X,\orr_2(X),Y,\orr_2(Y),i,j,\vartheta,p,[\gamma],
3053:   \orr_2(p)\,\big) \,.
3054:   \ee
3055: Here $i,j\iN\II$, $X,Y$ are $A$-bimodules, and $\orr_2(X)$ and $\orr_2(Y)$ 
3056: are local orientations of the world sheet on a neighbourhood of the
3057: defect line labelled by $X$ and $Y$, respectively; $\vartheta$ is a morphism 
3058:   \be
3059:   \vartheta \in \HomAA(U_i\,{\otimes^+}\,Z\,{\otimes^-}\,U_j,Z') \,,
3060:   \ee
3061: where
3062:   \be
3063:   Z =   \left\{ \bearll
3064:     X~&  {\rm if}~~ \orr_2(X)=\orr_2(p) \,, \\[.2em]
3065:     X^s~&  {\rm if}~~ \orr_2(X)=-\orr_2(p) 
3066:   \eear\right.
3067:   \quad\ {\rm and} \quad\
3068:   Z' =   
3069:   \left\{ \bearll
3070:     Y~& {\rm if}~~ \orr_2(Y)=\orr_2(p) \,, \\[.2em]
3071:     Y^s~& {\rm if}~~ \orr_2(Y)=-\orr_2(p) \,.  
3072:   \eear\right.
3073:   \labl{eq:defect-ZZ-XY}
3074: Finally $p\eq\gamma(0)$ is the insertion point of the defect field,
3075: and $[\gamma]$ is a germ of arcs aligned with the defect line at $p$.  
3076: 
3077: \medskip
3078: 
3079: A defect circle $\dftd([\gamma_1],...\,,[\gamma_n])$ on the world
3080: sheet is labelled by equivalence classes of tuples
3081:   \be
3082:   \big(\, X_1,\orr_2(X_1),...\,,X_{n'},\orr_2(X_{n'}) \,;\,
3083:   \Thetha_1, ...\, , \Thetha_n \,;~ \orr_1(\dftd) \,\big)  \,,
3084:   \labl{eq:defect-tupel}
3085: subject to the following conditions.
3086: \ITemize
3087: \item[\Nxt] 
3088: $\orr_1(\dftd)$ is an orientation of the defect circle $\dftd$.
3089: \item[\Nxt]
3090: The points $p_k\eq\gamma_k(0)$ are ordered
3091: such that when passing along the defect circle $\dftd$ in the direction 
3092: opposite to $\orr_1(\dftd)$, one passes from $p_k$ to $p_{k+1}$. Here
3093: and below it is understood that $p_0 \,{\equiv}\, p_n$ and 
3094: $p_{n+1} \,{\equiv}\, p_1$, and similarly for $X_k$ and $\Thetha_k$.
3095: \item[\Nxt]
3096: The $X_1,...\,,X_{n'}$ are $A$-bimodules. $X_k$ 
3097: labels the defect interval between $p_{k-1}$ and $p_k$. Thus if there
3098: are no field insertions, then $n'\eq 1$, and otherwise $n'\eq n$.
3099: \item[\Nxt]
3100: $\orr_2(X_k)$ is a local orientation\,\footnote{%
3101:   ~If the neighbourhood of the whole defect circle is non-orientable
3102:   we require that there is at least one field insertion.  
3103:   This field insertion can transform like an identity field, though, in
3104:   which case correlators are independent of the insertion
3105:   point of this field (recall the corresponding discussion in the
3106:   second half of section II:3.8).}
3107: of the world sheet on a
3108: neighbourhood of the defect interval between $p_{k-1}$ and $p_k$.
3109: \item[\Nxt]
3110: $\Thetha_k$ is a defect field with defining data
3111:   \be
3112:   \Thetha_k = \big(\, X_{k},\orr_2(X_{k}),X_{k+1},\orr_2(X_{k+1}),i_k,j_k,
3113:   \vartheta_k,p_k,[\gamma_k], \orr_2(p_k)\,\big)  \,.
3114:   \ee
3115: \end{itemize}
3116: 
3117: \medskip
3118: 
3119: Because of the large number of orientations entering the labelling of a defect 
3120: circle, the equivalence relation on the tuples \erf{eq:defect-tupel} becomes 
3121: somewhat technical; we include it merely for completeness of the presentation.
3122: 
3123: \medskip
3124: 
3125: The relation is conveniently formulated in terms of five basic relations. 
3126: The full equivalence class is obtained by completing the relations generated 
3127: by the basic relations with respect to transitivity and reflexivity. Consider 
3128: the two tuples
3129:   \bea\displaystyle
3130:   T = \big(\, X_1,\orr_2(X_1),...\,,X_{n'},\orr_2(X_{n'}) \,;\,
3131:   \Thetha_1, ...\, , \Thetha_n \,;\, \orr_1(\dftd) \,\big) \,,
3132:       \\[7pt]\displaystyle
3133:   T' = \big(\, X'_1,\orr_2'(X'_1),...\,,X'_{n'},\orr_2'(X'_{n'}) \,;
3134:   \Thetha_1', ...\, , \Thetha_n' \,;\, \orr_1'(\dftd) \,\big) \,.
3135:   \eear\ee
3136: The basic relations are:
3137: \\[4pt]
3138: \nxt \underline{Relation 1}: Shift of labels.
3139: \\[.2em]
3140: The tuples $T$ and $T'$ are equivalent if their data are related by
3141: $\orr_1'(\dftd) \eq \orr_1^{}(\dftd)$ and, for every 
3142: $k$, $X'_k \eq X_{k+1}^{}$, $\orr_2'(X_k') \eq \orr_2(X_{k+1}^{})$ 
3143: and $\Thetha_k'\eq \Thetha_{k+1}^{}$.
3144: \\[4pt]
3145: %
3146: \nxt \underline{Relation 2}: 
3147: Reverse orientation of the defect circle, $\orr_1'(\dftd)\eq{-}\orr_1(\dftd)^{}$. 
3148: \\[.2em]
3149: Suppose $\orr_1'(\dftd) \eq {-} \orr_1(\dftd)$. Then $T$ and $T'$ are
3150: equivalent if the following conditions hold.
3151: First, the ordering of the points is reversed in both labellings,
3152: $[\gamma_k'] \eq [\gamma_{n-k}]$. Second, the bimodules labelling the defect
3153: intervals are related by $X_{k} \eq (X_{n-k+1}')^v$.
3154: Finally, the defect fields $\Thetha_{k}$ are related to the fields
3155:   \be
3156:   \Thetha_{\ell}' = 
3157:   \big(\, X_{\ell}',\orr_2'(X_{\ell}'),X_{\ell+1}',\orr_2'(X_{\ell+1}'),
3158:   i_{\ell}',j_{\ell}',\vartheta_{\ell}',p_{\ell}',[\gamma_{\ell}'],
3159:         \orr_2'(p_{\ell}')\, \big) 
3160:   \ee
3161: via
3162:   \bea\displaystyle
3163:   \Thetha_{k} = 
3164:   \big(\, (X_{n-k+1}')^v,\orr_2'(X_{n-k+1}'), (X_{n-k}')^v,\orr_2'(X_{n-k}'),
3165:   \\[5pt]\displaystyle\hspace{8em}
3166:   i_{n-k}',j_{n-k}', \tilde \beta \cir \tilde u(\vartheta_{n-k}') 
3167:   \cir (\alpha \oti \beta \oti \tilde \alpha) , p_{n-k}',[\gamma_{n-k}'],
3168:   \orr_2'(p_{n-k}') \,\big) \,, 
3169:   \eear\labl{eq:muk-rel5}
3170: where $\tilde u$ is the map defined in \erf{eq:uv-def}, and
3171:   \beaa
3172:   \displaystyle
3173:   \alpha = \begin{cases}
3174:     \id_{U_{i_{n-k}'}}  & {\rm if}~ \orr_1(\gamma_k) \eq \orr_1(\dftd) \,, \\
3175:     \theta_{U_{i_{n-k}'}}  & {\rm if}~ \orr_1(\gamma_k) \eq {-}\orr_1(\dftd)\,,    
3176:   \end{cases}
3177:   &
3178:   \beta = \begin{cases}
3179:     \id_{(X_{n-k+1}')^\vee} & \!\!{\rm if}~ \orr_2(p_k) \eq \orr_2(X_k) \,, \\
3180:     \theta_{(X_{n-k+1}')^\vee}^{-1} &\!\!{\rm if}~\orr_2(p_k) \eq {-}\orr_2(X_k)
3181:     \,, \end{cases}
3182:   \\[2em] \displaystyle
3183:   \tilde\alpha = \begin{cases}
3184:     \id_{U_{j_{n-k}'}} & {\rm if}~ \orr_1(\gamma_k) \eq \orr_1(\dftd) \,, \\
3185:     \theta_{U_{j_{n-k}'}}^{-1} & {\rm if}~ \orr_1(\gamma_k) \eq {-}\orr_1(\dftd) \,,   
3186:   \end{cases}
3187:   ~~ &
3188:   \tilde\beta = \begin{cases}
3189:     \id_{(X_{n-k}')^\vee} & \!\!{\rm if}~ \orr_2(p_k) \eq \orr_2(X_{k+1}) \,, \\
3190:     \theta_{(X_{n-k}')^\vee} & \!\!{\rm if}~ \orr_2(p_k) \eq {-}\orr_2(X_{k+1}) \,. 
3191:   \end{cases}
3192:   \eear\ee
3193: \\[4pt]
3194: %
3195: For the next relations, select any $k_0\iN\{1,2,...\,,n'\}$.
3196: \\[4pt]
3197: %
3198: \nxt \underline{Relation 3}: Isomorphism of bimodules,
3199: $X_{k_0}' \,{\cong}\, X_{k_0}$.
3200: \\[.2em]
3201: The tuples $T$ and $T'$ are equivalent if all data of $T'$ coincide with 
3202: the corresponding data of $T$, possibly with the exception of
3203: $X'_{k_0}$, $\Thetha'_{k_0-1}$ and $\Thetha'_{k_0}$. 
3204: In the latter case we require the existence of an isomorphism
3205: $\varphi\iN\HomAA(X_{k_0}',X_{k_0})$ such that 
3206:   \be
3207:   \vartheta'_{k_0-1} = \varphi^{-1} \cir \vartheta_{k_0-1}
3208:   \quad\ {\rm and} \quad\
3209:   \vartheta'_{k_0} = \vartheta_{k_0} \cir
3210:   (\id_{U_{i_{k_0}}}\oti\varphi\oti\id_{U_{j_{k_0}}}) \,,
3211:   \ee
3212: where $\vartheta'$ and $\vartheta$ are the morphisms in the defining data
3213: of $\Thetha'$ and $\Thetha$.
3214: \\[4pt]
3215: %
3216: \nxt \underline{Relation 4}: Reverse orientation at a field insertion,
3217: $\orr_2'(p_{k_0}) \eq {-} \orr_2(p_{k_0})$. 
3218: \\[.2em]
3219: The tuples $T$ and $T'$ are equivalent if
3220: all data of $T'$ and $T$ agree, with the exception of
3221: $\Thetha'_{k_0}$, for which it is instead required that if
3222:   \be
3223:   \Thetha_{k_0} = 
3224:   \big(\, X_{k_0},\orr_2(X_{k_0}),X_{k_0+1},\orr_2(X_{k_0+1}),
3225:   i_{k_0},j_{k_0},\vartheta_{k_0},p_{k_0},[\gamma_{k_0}],\orr_2(p_{k_0})\,\big) \,, 
3226:   \ee
3227: then $\Thetha'_{k_0}$ is given by
3228:   \bea\displaystyle
3229:   \Thetha_{k_0}' = 
3230:   \big(\, X_{k_0},\orr_2(X_{k_0}),X_{k_0+1},\orr_2(X_{k_0+1}),
3231:   \\[5pt]\displaystyle\hspace{9em}
3232:   j_{k_0},i_{k_0}, \tilde \beta \cir u(\vartheta_{k_0})
3233:   \cir (\alpha \oti \beta \oti \tilde \alpha)
3234:   ,p_{k_0},[\gamma_{k_0}], -\orr_2(p_{k_0})\, \big) \,,
3235:   \eear\ee
3236: where $u$ is the map defined in \erf{eq:uv-def}, and
3237:   \beaa
3238:   \displaystyle
3239:   \alpha = \begin{cases}
3240:     \theta_{U_{j_{k_0}}}^{-1}  & {\rm if}~ \orr_1(\gamma_{k_0}) \eq \orr_1(\dftd) \,, \\
3241:     \id_{U_{j_{k_0}}}  & {\rm if}~ \orr_1(\gamma_{k_0}) \eq {-}\orr_1(\dftd)  \,,  
3242:   \end{cases}
3243:   \quad &
3244:   \beta = \begin{cases}
3245:     \id_{X_{k_0}} & {\rm if}~ \orr_2(p_{k_0}) \eq \orr_2(X_{k_0}) \,, \\
3246:     \theta_{X_{k_0}} & {\rm if}~ \orr_2(p_{k_0}) \eq {-}\orr_2(X_{k_0})\,,  
3247:   \end{cases}
3248:   \\[2em] \displaystyle
3249:   \tilde\alpha = \begin{cases}
3250:     \theta_{U_{i_{k_0}}} & {\rm if}~ \orr_1(\gamma_{k_0}) \eq \orr_1(\dftd) \,, \\
3251:     \id_{U_{i_{k_0}}} & {\rm if}~ \orr_1(\gamma_{k_0}) \eq {-}\orr_1(\dftd) \,,   
3252:   \end{cases}
3253:   \quad &
3254:   \tilde\beta = \begin{cases}
3255:     \id_{X_{k_0+1}} & {\rm if}~ \orr_2(p_{k_0}) \eq \orr_2(X_{k_0+1})\,,  \\
3256:     \theta_{X_{k_0+1}}^{-1} & {\rm if}~ \orr_2(p_{k_0}) \eq {-}\orr_2(X_{k_0+1})\,.  
3257:   \end{cases}
3258:   \eear\ee
3259: \\[3pt]
3260: %
3261: \nxt \underline{Relation 5}: Reverse orientation at a defect segment,
3262: $\orr_2'(X_{k_0}') \eq {-} \orr_2(X_{k_0})$. 
3263: \\[.2em]
3264: The tuples $T$ and $T'$ are equivalent if all data of $T'$ and $T$ agree, with 
3265: the exception of $X_{k_0}$ and $\orr_2(X_{k_0})$, as well as 
3266: $\Thetha_{k_0-1}$ and $\Thetha_{k_0}$. For the latter we demand
3267:   \be
3268:   (X_{k_0}')^s = X_{k_0}
3269:   \qquad {\rm and} \qquad \orr_2'(X_{k_0}') = - \orr_2(X_{k_0}) \,,
3270:   \ee
3271: as well as
3272:   \bea\displaystyle
3273:   \Thetha_{k_0-1} = 
3274:   \big(\, X_{k_0-1}',\orr_2'(X_{k_0-1}'),(X_{k_0}')^s,-\orr_2'(X_{k_0}'),
3275:   \\[5pt]\displaystyle\hspace{9em}
3276:     i_{k_0-1}',j_{k_0-1}', \tilde\beta \cir \vartheta_{k_0-1}',
3277:     p_{k_0-1}',[\gamma_{k_0-1}'], \orr_2'(p_{k_0-1}')\, \big) \,,  
3278:   \\[10pt]\displaystyle
3279:   \Thetha_{k_0} = 
3280:   \big(\, (X_{k_0}')^s,-\orr_2'(X_{k_0}'),X_{k_0+1}',\orr_2'(X_{k_0+1}'),
3281:   \\[5pt]\displaystyle\hspace{9em}
3282:     i_{k_0}',j_{k_0}', \vartheta_{k_0}' \cir (\id_{U_{i_{k_0}'}} {\otimes}\, 
3283:     \beta \oti \id_{U_{j_{k_0}'}}) , p_{k_0}',[\gamma_{k_0}'],
3284:     \orr_2'(p_{k_0}')\, \big)  \,,  
3285:   \eear\ee
3286: where
3287:   \be
3288:   \beta = \begin{cases}
3289:     \id_{X_{k_0}'} & {\rm if}~ \orr_2(p_{k_0}) \eq \orr_2(X_{k_0}) \,, \\
3290:     \theta_{X_{k_0}'}^{-1} & {\rm if}~ \orr_2(p_{k_0}) \eq {-}\orr_2(X_{k_0})\,,  
3291:   \end{cases}
3292:   \qquad
3293:   \tilde\beta = \begin{cases}
3294:     \id_{X_{k_0}'} & {\rm if}~ \orr_2(p_{k_0-1}) \eq \orr_2(X_{k_0}) \,, \\
3295:     \theta_{X_{k_0}'} & {\rm if}~ \orr_2(p_{k_0-1}) \eq {-}\orr_2(X_{k_0})\,.  
3296:   \end{cases}
3297: \ee
3298: 
3299: \medskip
3300: 
3301: The set of labels for the defect circle $\dftd([\gamma_1],...\,,[\gamma_n])$ is 
3302: given by the equivalence classes
3303:   \be
3304:   \mathcalD(\dftd([\gamma_1],...\,,[\gamma_n]))
3305:   = \big\{\,\big(\, X_1,\orr_2(X_1),...\,,X_{n'},\orr_2(X_{n'}) \,;\,
3306:   \Thetha_1, ...\, , \Thetha_n \,;\,\orr_1(\dftd) \,\big) \,
3307:   \big\} / {\sim} \,,
3308:   \ee
3309: where $\sim$ is the equivalence relation generated
3310: by the five basic relations given above.
3311: 
3312: 
3313: \subsubsection*{Marked points on the double}
3314: 
3315: For a defect insertion on the world sheet $\X$
3316: the marked points on the double $\hat \X$ are the same as 
3317: for bulk fields. Each insertion
3318:   \be
3319:   \Thetha = \big(\, X,\orr_2(X),Y,\orr_2(Y),i,j,\vartheta,p,[\gamma],
3320:   \orr_2(p)\, \big) 
3321:   \ee
3322: on the world sheet $\X$ gives rise to two marked points
3323:   \be
3324:   (\tilde p_i, [\tilde\gamma_i] , U_i , +) 
3325:   \qquad {\rm and} \qquad
3326:   (\tilde p_j, [\tilde\gamma_j] , U_j , +) 
3327: \ee
3328: on the double $\hat \X$. Here again
3329: $\tilde p_i \eq [p,\orr_2(p)]$, $\tilde p_j \eq [p,-\orr_2(p)]$,
3330: as well as $\tilde\gamma_i(t) \eq [\gamma(t),\orr_2(\gamma(t))]$ and
3331: $\tilde\gamma_j(t) \eq [\gamma(t),-\orr_2(\gamma(t))]$. 
3332: 
3333: 
3334: \subsubsection*{Ribbon graph embedded in the connecting manifold}
3335: 
3336: The ribbon graph representation of a 
3337: defect circle $\dftd([\gamma_1],...\,,[\gamma_n])$ labelled
3338: by an equivalence class 
3339: $D \iN \mathcalD(\dftd([\gamma_1],...\,,[\gamma_n]))$ is 
3340: constructed as follows.
3341: %
3342: \ITemize
3343: \item[\Nxt] 
3344: Choose  a representative $\big( X_1,\orr_2(X_1),...\,,X_{n'},\orr_2(X_{n'}) ;
3345: \Thetha_1, ...\, , \Thetha_n ; \orr_1(\dftd) \big)$
3346: of $D$ -- \underline{Choice \#1}.
3347: %
3348: \item[\Nxt] Choose a dual triangulation of $\X$ which contains
3349: the defect circle as a subset, such that each defect insertion lies on an
3350: edge of the dual triangulation -- \underline{Choice \#2}.
3351: %
3352: \item[\Nxt] At each insertion of a defect field
3353: $\Thetha \eq \big(X,\orr_2(X),Y,\orr_2(Y),i,j,\vartheta,p,[\gamma],
3354: \orr_2(p)\big)$ place -- depending on the various orientations -- one of the 
3355: following eight ribbon graphs:
3356:   %  [ pic 4_98 ]
3357:   \begin{eqnarray}  \begin{picture}(400,250)(0,0)
3358:   \put(65,0)     {\Includeournicelargepicture 98a }
3359:   \put(325,30)     {\tiny $x$}
3360:   \put(310,50)     {\tiny $z$}
3361:   \put(302,47)     {\tiny $y$}
3362:   \put(234,4)      {\tiny $x$}
3363:   \put(214,17)     {\tiny $y$}
3364:   \put(93,170)     {\tiny $x$}
3365:   \put(86,160)     {\tiny $y$}
3366:   \put(226,235)    {\tiny $x$}
3367:   \put(220,225)    {\tiny $y$}
3368:   \put(250,192)    {\scriptsize $U_i$}
3369:   \put(244,41)     {\scriptsize $U_j$}
3370:   \put(172,129)    {\scriptsize $Z$}
3371:   \put(277,129)    {\scriptsize $Z'$}
3372:   \put(186,106)    {\small $\rm d $}
3373:   \put(266,106)    {\small $\rm d $}
3374:   \put(112,105)    {\scriptsize $\orr_1(\rm d)$}
3375:   \put(316,105)    {\scriptsize $\orr_1(\rm d)$}
3376:   \put(229.6,126)  {\begin{turn}{280} $\vartheta$\end{turn}}
3377:   \put(-29,228)  { Cases (1)\,--\,(4)\,: }
3378:   \end{picture} \nonumber
3379:   \\                 \begin{picture}(400,260)(0,0)
3380:   \put(65,0)     {\Includeournicelargepicture 98b }
3381:   \put(325,30)     {\tiny $x$}
3382:   \put(310,50)     {\tiny $z$}
3383:   \put(302,47)     {\tiny $y$}
3384:   \put(234,4)      {\tiny $x$}
3385:   \put(214,17)     {\tiny $y$}
3386:   \put(93,170)     {\tiny $x$}
3387:   \put(86,160)     {\tiny $y$}
3388:   \put(226,235)    {\tiny $x$}
3389:   \put(220,225)    {\tiny $y$}
3390:   \put(250,192)    {\scriptsize $U_i$}
3391:   \put(244,40)     {\scriptsize $U_j$}
3392:   \put(172,129)    {\scriptsize $Z$}
3393:   \put(277,129)    {\scriptsize $Z'$}
3394:   \put(112,105)    {\scriptsize $\orr_1(\rm d)$}
3395:   \put(316,105)    {\scriptsize $\orr_1(\rm d)$}
3396:   \put(229.6,126)  {\begin{turn}{280} $\vartheta$\end{turn}}
3397:   \put(-29,228)  { Cases (5)\,--\,(8)\,: } 
3398:   \end{picture} \nonumber
3399:   \\[-1.6em]{} \label{eq:pic41}
3400:   \end{eqnarray} \xlabp 98
3401: Here in all cases the coupon labelled by $\vartheta$ is placed in such a way that 
3402: its orientation is opposite to $\orr_2(p)$. The bimodule ribbons are always 
3403: placed such that the orientation of their cores is opposite to $\orr_1(\dftd)$.
3404: The bimodules $Z$ and $Z'$ are related to $X$ and $Y$ as in 
3405: \erf{eq:defect-ZZ-XY}. In cases (1)\,--\,(4) the orientation of
3406: the arc-germ $[\gamma]$ is equal to $-\orr_1(\dftd)$. Those parts 
3407: of the ribbon graphs that we have put in brackets are to be inserted only if
3408: $\orr_2(X) \eq {-}\orr_2(p)$ and $\orr_2(Y) \eq {-}\orr_2(p)$, respectively.
3409: In cases (5)\,--\,(8) the orientation of the arc-germ $[\gamma]$ is equal to 
3410: $+\orr_1(\dftd)$; the parts of the ribbon graph in brackets are to be 
3411: understood in the same way as in cases (1)\,--\,(4).
3412: \end{itemize}
3413: 
3414: \noindent
3415: We only briefly sketch how to establish independence of these
3416: choices.
3417: \\[5pt]
3418: \nxt\underline{Choice \#2}:\\[2pt]
3419: As for bulk fields the fact that the morphism $\vartheta$ in a
3420: defect field $\Thetha$ is an intertwiner of $A$-bimodules
3421: allows one to move vertices of the dual triangulation past
3422: insertion points of defect fields.
3423: \\[5pt]
3424: \nxt\underline{Choice \#1}:\\[2pt]
3425: Suppose $T$ and $T'$ are two representatives of the
3426: equivalence class $D$ labelling the defect circle. We have
3427: to establish that if $T$ and $T'$ are linked by one of
3428: the five basic relations above, then the construction leads
3429: to equivalent ribbon graphs. This can be done by straightforward, though
3430: tedious, computations. We refrain from presenting them in any detail, except
3431: for the example that $T$ and $T'$ are linked by relation~2.
3432: 
3433: If $T$ and $T'$ are linked by relation~2, then in particular
3434: $p_k \eq p'_{n-k}$. Select a point $p\eq p_{k_0}\eq p'_{n-k_0}$ and
3435: consider only the fragment of the ribbon graph for $T$ and $T'$ close to $p$. 
3436: Suppose further that $\orr_1(\gamma_{k_0}) \eq \orr_1(\dftd)$, $\orr_2(p_{k_0})
3437: \eq\orr_2(X_{k_0})$ and $\orr_2(p_{k_0})\eq{-}\orr_2(X_{k_0+1})$. The other 
3438: instances of relation~2 follow by similar considerations. Denote the fragments 
3439: close to $p$ by $[T]$ and $[T']$. Applying the construction above gives 
3440:   %  [ pic 4_110 ]
3441:   \begin{eqnarray} \begin{picture}(420,130)(21,0)
3442:   \put(37,0)    {\INcludeourbeautifulpicture 110a }
3443:   \put(293,0)   {\INcludeourbeautifulpicture 110b }
3444:   \put(50,65)     {\scriptsize$X_{k_0}$}
3445:   \put(188,65)    {\scriptsize$X_{k_0+1}$}
3446:   \put(103,35)    {\scriptsize$j_{k_0}$}
3447:   \put(103,79)    {\scriptsize$i_{k_0}$}
3448:   \put(61,42)     {\scriptsize$\orr_1(\rm d)$}
3449:   \put(125.5,66)  {\begin{turn}{270}\small$\vartheta_{k_0}$\end{turn}}
3450:   \put(301,65)    {\scriptsize$X'_{n-k_0+1}$}
3451:   \put(451,65)    {\scriptsize$X'_{n-k_0}$}
3452:   \put(392,32)    {\scriptsize$j'_{n-k_0}$}
3453:   \put(392,82)    {\scriptsize$i'_{n-k_0}$}
3454:   \put(307,42)    {\scriptsize$\orr_1(\rm d)'$}
3455:   \put(356.9,46.9){\begin{turn}{90}\small$\vartheta'_{\!n-k_0}$\end{turn}}
3456:   \put(0,56)    {$ [T] \;= $}
3457:   \put(252,56)  {$ [T'] \;= $}
3458:   \end{picture} \nonumber
3459:   \\[.1em]{} \label{eq:dft-choice-TT}
3460:   \end{eqnarray} \Xlabp 110
3461: Here we displayed the graphs \erf{eq:pic41} using blackboard framing. 
3462: The upper and lower boundary in these figures represent the upper
3463: and lower boundary of the connecting manifold $\M_X$.
3464: According to \erf{eq:muk-rel5} in the graph $[T]$ we should substitute
3465: $X_{k_0} \eq (X_{n-k_0+1}')^v$, $X_{k_0+1} \eq (X_{n-k_0}')^v$,
3466: $i_{k_0} \eq i_{n-k_0}'$, $j_{k_0} \eq j_{n-k_0}'$ and
3467: $\vartheta_{k_0} \eq \theta_{(X_{n-k_0}')^\vee} \cir \tilde u(\vartheta_{n-k_0}')$.
3468: This results in
3469:   %  [ pic 4_111 ]
3470:   \bea \begin{picture}(340,84)(0,40)
3471:   \put(50,0)    {\Includeourbeautifulpicture 111 }
3472:   \put(0,56)    {$ [T] \;= $}
3473:   \put(48,65)     {\scriptsize$X'_{n-k_0+1}$}
3474:   \put(273,65)    {\scriptsize$X'_{n-k_0}$}
3475:   \put(112,22)    {\scriptsize$j'_{k_0}$}
3476:   \put(113,95)    {\scriptsize$i'_{k_0}$}
3477:   \put(132.8,71.8){\begin{turn}{270}\small$\vartheta'_{\!n-k_0}$\end{turn}}
3478:   \epicture21 \labl{eq:dft-choice-or1} \Xlabp 111
3479: One can now verify that this ribbon graph fragment can be
3480: deformed into $[T']$ in \erf{eq:dft-choice-TT}.
3481: 
3482: 
3483: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3484: 
3485: \sect{Ribbon graphs for structure constants}\label{sec:rc}
3486: 
3487: 
3488: The correlation function for an arbitrary world sheet $\X$, which may have 
3489: boundaries and field insertions, and which may or may not be orientable,
3490: can be computed by cutting $\X$ into
3491: smaller pieces and summing over intermediate states. This procedure
3492: is known as `sewing'. In principle, every correlator can be obtained in this 
3493: way from a small number of basic building blocks, namely the correlators of
3494: \\[.5em]
3495: \hspace*{3.5em}--- three boundary fields on the disk;
3496: \\[.4em]
3497: \hspace*{3.5em}--- one bulk field and one boundary field on the disk;
3498: \\[.4em]
3499: \hspace*{3.5em}--- three bulk fields on the sphere;
3500: \\[.4em]
3501: \hspace*{3.5em}--- one bulk field on the cross cap.
3502: \\[.5em]
3503: (The last of these is only needed if non-orientable world sheets are admitted.)
3504: %\\[.5em] 
3505: Some correlators involving defect fields will be presented as well, namely
3506: \\[.5em]
3507: \hspace*{3.5em}--- three defect fields on the sphere;
3508: \\[.4em]
3509: \hspace*{3.5em}--- one defect field on the cross cap.
3510: \\[.5em]
3511: However, recall from the discussion in section \ref{sec:fi-defect}
3512: that this does not correspond to the most general way to treat defects. When 
3513: cutting a world sheet involving defect lines into smaller pieces, 
3514: one is forced to also consider fields of
3515: the type \erf{eq:fields-not-consd}, which we will not do in this paper.
3516: 
3517: \medskip
3518: 
3519: The aim of this and the next two sections is to derive expressions for the 
3520: building blocks listed above in the form 
3521:   \be
3522:   \mbox{(\,correlator\,)}
3523:   ~=~ \mbox{(\,const\,)} ~\times~ \mbox{(\,conformal block\,)} \,. 
3524:   \labl{corr = c x block}
3525: The constants appearing here, to be referred to as {\em structure constants\/},
3526: will be obtained in sections \ref{sec:rc-bnd3pt}\,--\,\ref{sec:rc-1defect-xcap} 
3527: as ribbon invariants in the 3-d TFT. The conformal blocks are then introduced 
3528: in section \ref{sec:cb}, and finally the connection to correlation functions
3529: is made in section \ref{sec:cf}.
3530: 
3531: Since a given world sheet $\X$ can be reduced to the fundamental building 
3532: blocks in several different ways, the latter must fulfill a set of consistency 
3533: conditions, the so-called sewing or factorisation constraints 
3534: \cite{bepz,sono2,sono3,cale,lewe3,fips,bppz2}. The correlators computed in 
3535: the TFT approach satisfy these constraints by construction \cite{fffs3,fjfrs}.
3536: 
3537: \medskip
3538: 
3539: When giving the correlators on the disk and on the sphere in sections 
3540: \ref{sec:rc-bnd3pt}\,--\,\ref{sec:rc-3defect}, we will fix a (bulk and boundary)
3541: orientation for the world sheet from the start. In this way the expressions
3542: we obtain are valid for oriented CFTs as well. In the unoriented
3543: case we can change the orientation without affecting the result for the 
3544: correlator, by using the equivalence relations of section \ref{sec:fi}.
3545: 
3546: 
3547: \subsection{Standard bases of conformal blocks on the sphere}
3548: \label{sec:standbas}
3549: 
3550: Let the world sheet $\X$ be one of the building blocks listed above. To compute 
3551: the constants in \erf{corr = c x block} we first construct the ribbon graph
3552: representation in $\MX$ of the correlator on $\X$ according to the rules
3553: given in section \ref{sec:fi}. The TFT assigns to $\MX$ an element $C(\X)$
3554: in the space $\calh(\hat \X)$ of blocks on the double $\hat \X$. In the cases
3555: considered here, $\hat\X$ is either the sphere $S^2$ or the disjoint
3556: union $S^2{\sqcup}(-S^2)$.
3557: The constants are the coefficients occurring in the expansion of $C(\X)$ 
3558: in a standard basis of blocks in $\calh(\hat\X)$. In the present section we 
3559: describe our convention for the standard basis in $\calh(S^2)$. In section 
3560: \ref{sec:cb} this basis will be related to products of chiral vertex operators.
3561: 
3562: \medskip
3563: 
3564: Take the sphere $S^2$ to be parametrised as $\reals^2 \,{\cup}\, \{\infty\}$.
3565: Define the standard extended surface $E_{0,n}[i_1,\cdots i_n]$ to be $\reals^2 
3566: \,{\cup}\, \{\infty\}$ with the standard orientation, together with the marked 
3567: points $\{((k,0),[\gamma_k],U_{i_k},+) \,|\, k\eq1,...\,,n \}$ and 
3568: $\gamma_k(t)\eq (k{+}t,0)$. If it is clear from the context what 
3569: representations occur at the marked points, we use the shorthand
3570: notation $E_{0,n}$ for $E_{0,n}[i_1,\cdots i_n]$. Let the cobordism
3571: $B_{0,n} \,{\equiv}\, B_{0,n}[i_1,...\,,i_n;
3572: p_1,...\,, p_{n-3};\delta_1 ,...\,,\delta_{n-2}]$ be given by the
3573: three-manifold $\{ (x,y,z)\iN\reals^3\,|\,z\,{\ge}\,0 \}\,{\cup}\,\{\infty\}$
3574: together with the ribbon graph
3575:   %  [ pic 4_02 ]
3576:   \bea \begin{picture}(360,200)(0,44)
3577:   \put(40,0)     {\includeourbeautifulpicture 02 }
3578:   \put(0,124)    {$ B_{0,n}^{} \ = $}
3579:   \put(73.4,68)  {\scriptsize$ z $}
3580:   \put(92,43.7)  {\scriptsize$ y $}
3581:   \put(335,21)   {\scriptsize$ x $}
3582:   \put(108,13)   {\scriptsize$ 1 $}
3583:   \put(170,13)   {\scriptsize$ 2 $}
3584:   \put(216,13)   {\scriptsize$ 3 $}
3585:   \put(306,14)   {\scriptsize$ n $}
3586:   \put(111,62)   {$ U_{\!i_{1}}^{} $}
3587:   \put(158.2,62) {$ U_{\!i_{2}}^{} $}
3588:   \put(205.2,62) {$ U_{\!i_{3}}^{} $}
3589:   \put(295.2,62) {$ U_{\!i_{n}}^{} $}
3590:   \put(139,77)   {\scriptsize$ \delta_{1}^{} $}
3591:   \put(162,123)  {\scriptsize$ \delta_{2}^{} $}
3592:   \put(208.8,213){\scriptsize$ . $}
3593:   \put(131,106)  {$ U_{\!p_{1}^{}}^{} $}
3594:   \put(150,144)  {$ U_{\!p_{2}^{}}^{} $}
3595:   \put(177,197)  {$ U_{\!\overline{\imath_{n}^{}}}^{} $}
3596:   \epicture32 \labl{eq:standard-cobord} \xlabp 02
3597: where the $\delta_m$ label the basis elements (I:2.29) in the spaces of
3598: three-point couplings. For example $\delta_1\eq1,...\,,\N{i_2}{i_1}{p_1}$
3599: labels the basis of $\Hom(U_{i_{2}} \oti U_{i_{1}} , U_{p_{1}} )$.
3600: Note that in \erf{eq:standard-cobord} all ribbons are oriented in such a way 
3601: that their black side is facing the reader. The linear map $Z(B_{0,n},
3602: \emptyset,E_{0,n}) {:}\; \complex \To\calh(E_{0,n})$, applied to $1\iN\complex$,
3603: and with all possible choices of the labels $p_1,...\,,p_{n-3}$ and 
3604: $\delta_1,...\,,\delta_{n-2}$, gives the standard basis in the space 
3605: $\calh(E_{0,n})$. For a proof that this is indeed a basis, see e.g.\ 
3606: lemma 2.1.3 in chapter IV of \cite{TUra} or section 4.4 of \cite{BAki}. 
3607: Situations in which the insertion points are at different positions
3608: or in which the arcs are oriented in a different way can be obtained
3609: from \erf{eq:standard-cobord} by continuous deformation.
3610: 
3611: 
3612: \subsection{Boundary three-point function on the disk}\label{sec:rc-bnd3pt}
3613: 
3614: Consider the correlator of three boundary fields on a disk.  
3615: For future convenience we regard the disk as the upper half plane plus a 
3616: point. What is shown in the picture below is a piece of the upper half plane:
3617:   %  [ pic 4_24 ]
3618:   \bea \begin{picture}(160,43)(0,33)
3619:   \put(0,0)      {\Includeourbeautifulpicture 24b }
3620:   \put(11,16)    {\scriptsize$ M $}
3621:   \put(25.6,69)  {\scriptsize$ y $}
3622:   \put(162,12.6) {\scriptsize$ x $}
3623:   \put(49,5)     {\scriptsize$ x_1 $}
3624:   \put(81,5)     {\scriptsize$ x_2 $}
3625:   \put(118,5)    {\scriptsize$ x_3 $}
3626:   \put(48,19)    {\scriptsize$ \hatpsi_1 $}
3627:   \put(80,19)    {\scriptsize$ \hatpsi_2 $}
3628:   \put(117,19)   {\scriptsize$ \hatpsi_3 $}
3629:   \put(134,16)   {\scriptsize$ M $}
3630:   \put(99,16)    {\scriptsize$ K $}
3631:   \put(65,16)    {\scriptsize$ N $}
3632:   \epicture22 \labp 24
3633: On the world sheet we choose the standard orientation for bulk and
3634: boundary, as indicated. The three boundary fields are given by
3635:   \be \bearl
3636:   \hatpsi_1 = (N,M,U_i,\psi_1,x_1,[\gamma_1])\,, \qquad
3637:   \hatpsi_2 = (K,N,U_j,\psi_2,x_2,[\gamma_2])\,, \\{}\\[-.7em]
3638:   \hatpsi_3 = (M,K,U_k,\psi_3,x_3,[\gamma_3])\,.  \eear \ee
3639: Here the coordinates of the
3640: insertion points satisfy $0\,{<}\,x_1\,{<}\,x_2\,{<}\,x_3$, and the arc germs 
3641: $[\gamma_\ell]$ at these points are given by $\gamma_\ell(t) \eq (x_\ell{+}t,0)$
3642: for $\ell\iN\{1,2,3\}$. Further, $U_i$, $U_j$, $U_k$ are simple objects, while
3643: $M, N, K$ are left $A$-modules.
3644: 
3645: Following the procedure for boundary fields described in section 
3646: \ref{sec:fi-bnd}, one finds that the correlator is determined by the cobordism
3647:   %  [ pic 4_25 ]
3648:   \bea \begin{picture}(380,95)(0,31)
3649:   \put(45,0)     {\includeourbeautifulpicture 25 }
3650:   \put(0,64)     {$ \MX \;= $}
3651:   \put(87.8,68)  {\scriptsize$ z $}
3652:   \put(102,43.7){\scriptsize$ y $}
3653:   \put(327,21)   {\scriptsize$ x $}
3654:   \put(168,35)   {\scriptsize$ i $}
3655:   \put(225.6,35) {\scriptsize$ j $}
3656:   \put(283,35)   {\scriptsize$ k $}
3657:   \put(164,15)   {\scriptsize$ x_1 $}
3658:   \put(222,15)   {\scriptsize$ x_2 $}
3659:   \put(279,15)   {\scriptsize$ x_3 $}
3660:   \put(168,76)   {\scriptsize$ N $}
3661:   \put(184,118)  {\scriptsize$ M $}
3662:   \put(224,76)   {\scriptsize$ K $}
3663:   \put(131,68)   {\scriptsize$ \psi_1 $}
3664:   \put(189,68)   {\scriptsize$ \psi_2 $}
3665:   \put(247,68)   {\scriptsize$ \psi_3 $}
3666:   \epicture19 \labl{eq:uhp-cobord} \xlabp 25
3667: In more detail, this cobordism is obtained as follows. First note that in
3668: the situation at hand any $A$-ribbons resulting from a dual triangulation 
3669: of the world sheet $\X$ can be eliminated by repeatedly using the moves 
3670: (I:5.11) and (I:5.12). Thus only the module ribbons with the field 
3671: insertions as in figure \erf{eq:bnd-ribbon} remain. As explained in section 
3672: \ref{sec:fi-conv}, all ribbons are inserted with orientation opposite to 
3673: that of the world sheet. Also note that the three-manifold displayed in 
3674: \erf{eq:uhp-cobord} is topologically equivalent to the
3675: one indicated in \erf{eq:con-mf-bnd}.
3676: 
3677: According to section \ref{sec:standbas}, the structure
3678: constants $c(M \hatpsi_1 N \hatpsi_2 K \hatpsi_3 M)_\delta$
3679: are defined to be the coefficients occurring when expanding
3680: \erf{eq:uhp-cobord} in terms of a standard basis for the space
3681: of blocks $\calh(\hat X)$. In the case under consideration, $\hat X$ is the 
3682: sphere with field insertions at $x_1,\, x_2,\, x_3$. The standard basis
3683: is given by the cobordisms
3684:   %  [ pic 4_26 ]
3685:   \bea \begin{picture}(400,77)(0,46)
3686:   \put(90,0)      {\includeourbeautifulpicture 26 }
3687:   \put(0,69)      {$ B(x_1,x_2,x_3)_\delta^{} \;= $}
3688:   \put(133.4,68)  {\scriptsize$ z $}
3689:   \put(148,43)    {\scriptsize$ y $} 
3690:   \put(372,21)    {\scriptsize$ x $}
3691:   \put(213,56)    {\scriptsize$ i $}
3692:   \put(262.7,56)  {\scriptsize$ j $}
3693:   \put(328.6,56)  {\scriptsize$ k $}
3694:   \put(208,15)    {\scriptsize$ x_1 $}
3695:   \put(263,15)    {\scriptsize$ x_2 $}
3696:   \put(323,15)    {\scriptsize$ x_3 $}
3697:   \put(260,109)   {\scriptsize$ \overline{k} $}
3698:   \put(231,89)    {\scriptsize$ \delta $}
3699:   \epicture27 \labp 26
3700: which are obtained from \erf{eq:standard-cobord} by shifting the insertion 
3701: points appropriately. The defining equation for the structure constants thus 
3702: reads
3703:   \be
3704:   Z(\MX,\emptyset,\hat \X) = \sum_{\delta=1}^{\N ij{\bar k}} 
3705:   c(M \hatpsi_1 N \hatpsi_2 K \hatpsi_3 M)_\delta\,
3706:   Z(B(x_1,x_2,x_3)_\delta , \emptyset , \hat \X) \,.
3707:   \labl{eq:uhp-3pt-cobord}
3708: To determine the coefficients $c_\delta$ we compose both sides of the equality 
3709: with $Z(\overline{B}(x_1,x_2,x_3)_\delta,\hat \X,\emptyset)$ from the left,
3710: where $\overline{B}_\delta$ is the cobordism dual to $B_\delta$. This removes 
3711: the summation and results in the desired ribbon invariant expressing $c_\delta$:
3712:   %  [ pic 4_27 ] 
3713:   \bea \begin{picture}(420,288)(0,0)
3714:   \put(140,150)     {\begin{picture}(0,0)
3715:   \put(0,0)       {\Includeourbeautifulpicture 27a }
3716:                     \end{picture} }
3717:   \put(140,0)       {\begin{picture}(0,0)
3718:   \put(0,0)       {\Includeourbeautifulpicture 27b }
3719:                     \end{picture} }
3720:   \put(0,219)     {$ c(M \hatpsi_1 N \hatpsi_2 K \hatpsi_3 M)_\delta \;= $}
3721:   \put(110,69)    {$ = $}
3722:   \put(290,87)    {\scriptsize$ N $}
3723:   \put(230,129)   {\scriptsize$ M $}
3724:   \put(230,86)    {\scriptsize$ K $} 
3725:   \put(294,64)    {\scriptsize$ i $}
3726:   \put(245,64)    {\scriptsize$ j $}
3727:   \put(185,44)    {\scriptsize$ k $} 
3728:   \put(250,23)    {\scriptsize$ \overline{k} $} 
3729:   \put(270,38)    {\scriptsize$ \overline{\delta} $} 
3730:   \put(324.3,80)  {\scriptsize$ \psi_1 $}
3731:   \put(266.6,80)  {\scriptsize$ \psi_2 $}
3732:   \put(208.6,80)  {\scriptsize$ \psi_3 $}
3733:   \put(393.7,123.5) {$ \Stn $} 
3734:   \put(206,238)   {\scriptsize$ N $}
3735:   \put(222,280)   {\scriptsize$ M $}
3736:   \put(265,237)   {\scriptsize$ K $} 
3737:   \put(209,215)   {\scriptsize$ i $}
3738:   \put(257,214)   {\scriptsize$ j $}
3739:   \put(324,208)   {\scriptsize$ k $}
3740:   \put(250,160)   {\scriptsize$ \overline{k} $} 
3741:   \put(231.7,188) {\scriptsize$ \overline{\delta} $} 
3742:   \put(172.2,230) {\scriptsize$ \psi_1 $}
3743:   \put(229.7,230) {\scriptsize$ \psi_2 $}
3744:   \put(286.9,230) {\scriptsize$ \psi_3 $}
3745:   \put(394.5,273.7) {$ \Stn $} 
3746:   \epicture{-1}1 \labl{eq:bnd-3pt-rib} \xlabp 27
3747: Here the second equality amounts to rotating the ribbon graph in $S^3$ by 
3748: $180^\circ$ degrees, so that one is looking at the white sides of the ribbons, 
3749: rather than their black sides. 
3750: Also, we introduced the notation $\Stn$ to indicate the graphical representation
3751: of the {\em normalised\/} invariant of a ribbon graph in $S^3$. Recall from 
3752: the end of section I:5.1 that when the picture of a ribbon graph in a 
3753: three-manifold $\M$ is drawn, it represents the invariant $Z(\M)$ 
3754: {\em except\/} when $\M\eq S^3$, in which case it rather represents 
3755: $Z(\M)/S_{0,0}$. Since $S_{0,0}\eq Z(S^3)$, this amounts to assigning the 
3756: value $1$ to the empty graph in $S^3$. To stress this difference,
3757: we label figures representing normalised invariants in $S^3$ by $\Stn$,
3758: rather than by $S^3$. (This notational distinction was not made in 
3759: \cite{fuRs4,fuRs8,fuRs9}, where the normalisation convention for $S^3$ is used 
3760: implicitly.) Recall also that, by definition, we have
3761:   \be 
3762:   S_{0,0}^{~\,-2} = \sum_{i \in \II} \dim(U_i)^2_{} \,.
3763:   \ee  
3764:        
3765: \medskip
3766: 
3767: Let us express the constant $c(M \hatpsi_1 N \hatpsi_2 K \hatpsi_3 M)_\delta$
3768: in terms of the basis introduced in \erf{eq:MUM-basis}. To this end we take 
3769: the modules $M, N, K$ to be simple. A correlator labelled by reducible
3770: modules can be expressed as a sum of correlators labelled by
3771: simple modules. We set
3772:   \be
3773:   M=M_\mu \,, \quad N=M_\nu \,, \quad K=M_\kappa \ee
3774: and
3775:   \be
3776:   \psi_1 = \psi_{(\nu i)\mu}^{\alpha_1} \,, \quad
3777:   \psi_2 = \psi_{(\kappa j)\nu}^{\alpha_2} \,, \quad
3778:   \psi_3 = \psi_{(\mu k)\kappa}^{\alpha_3} \,.  \ee
3779: It is then straightforward to compute the invariant \erf{eq:bnd-3pt-rib} 
3780: from relation \erf{eq:GA-def} for the fusing matrix of the module category 
3781: $\calca$. The result is
3782:   \be
3783:   c(M \hatpsi_1 N \hatpsi_2 K \hatpsi_3 M)_\delta
3784:   = \dim(\dot M_\mu) \!\! \sum_{\beta=1}^{\Ann{\bar k M_\kappa}{M_\mu}} \!\!\!\!
3785:   \GG[A]^{(\kappa ji)\mu}_{\alpha_2 \nu \alpha_1 \,,\, \beta \bar k \delta}\,
3786:   \GG[A]^{(\mu k \bar k)\mu}_{\alpha_3 \kappa \beta \,,\, \cdot 0 \cdot} \,.
3787:   \labl{eq:3bnd-inv-basis}
3788: The dot `$\,\cdot\,$' stands again for a multiplicity label that can only 
3789: take a single value, the corresponding morphism space being one-dimensional.
3790: 
3791: Formula \erf{eq:3bnd-inv-basis} is in agreement with equation (4.16) of
3792: \cite{bppz2}. Note that in \cite{bppz2}, the quantities
3793: $^{(1)\!}F$ are taken as an input, whereas the corresponding 
3794: quantities $\GG[A]$ can be computed (with some effort) in terms of 
3795: the multiplication of $A$ and the fusing matrices $\FF$ of $\calc$,
3796: see section \ref{sec:mb-basis}.
3797: 
3798: Expressions for the structure constants of three boundary fields in rational 
3799: conformal field theory have been obtained previously in various cases
3800: for models with charge conjugation invariant \cite{runk,bppz2,fffs2,fffs3} 
3801: and non-diagonal models \cite{runk2,brSc2}.
3802: 
3803: 
3804: \subsection{One bulk and one boundary field on the disk}
3805: \label{sec:rc-1bulk1bnd}
3806: 
3807: The next correlator we consider is that of a bulk field $\vhi$ and a boundary 
3808: field $\vsi$ on the disk, which we again describe as the upper half plane:
3809:   %  [ pic 4_29 ]
3810:   \bea \begin{picture}(160,32)(0,43)
3811:   \put(0,0)       {\includeourbeautifulpicture 29 }
3812:   \put(150,12)    {\scriptsize$ x $}
3813:   \put(26,68.6)   {\scriptsize$ y $}
3814:   \put(9,17)      {\scriptsize$ M $}
3815:   \put(126,17)    {\scriptsize$ M $}
3816:   \put(105,17)    {\scriptsize$ \vsi $}
3817:   \put(59,49)     {\scriptsize$ \vhi $}
3818:   \put(104,6)     {\scriptsize$ s $}
3819:   \put(60,38.5)   {\scriptsize$ p $}
3820:   \epicture24 \labl{eq:1bulk-1bnd-geom} \xlabp 29
3821: The dotted line indicates the dual triangulation that we will
3822: use later on in the construction of the ribbon graph.
3823: The bulk and boundary fields are the tuples
3824:   \be
3825:   \vhi = (i,j,\phi,p,[\gamma_\phi],\orr_2) \qquad {\rm and} \qquad
3826:   \vsi = (M,M,U_k,\psi,s,[\gamma_\psi]) \,,
3827:   \ee
3828: where $i,j\iN\II$ label simple objects, $\phi\iN\HomAA(U_i\otiP A\otim U_j,A)$ 
3829: and $\psi\iN\HomA(M\oti U_k,M)$. Further, the insertion points are 
3830: $p \,{\equiv}\, (x,y)$ and $s \,{\equiv}\, (s,0)$ with
3831: $y\,{>}\,0$ and $s\,{>}\,x\,{>}\,0$, and the two arc germs
3832: are given by $\gamma_\phi(t)\eq(x{+}t,y)$ and $\gamma_\psi(t)\eq(s{+}t,0)$,
3833: respectively. The orientation $\orr_2$ is the standard
3834: orientation of the upper half plane at the insertion point of the bulk field.
3835: 
3836: \medskip
3837: 
3838: The construction of section \ref{sec:fi} yields the following
3839: ribbon graph inscribed in the connecting manifold:
3840:   %  [ pic 4_99 ]
3841:   \begin{eqnarray} \begin{picture}(420,225)(15,0)
3842:   \put(0,0)      {\includeourniceLargepicture 99 }
3843:   \put(369,55)     {\scriptsize $U_k$}
3844:   \put(243.5,80)   {\scriptsize $U_i$}
3845:   \put(249.2,154.5){\scriptsize $U_j$}
3846:   \put(165,39.8)   {\small ${\rho_M}$}
3847:   \put(273,58.8)   {\small ${\rho_M}$}
3848:   \put(215,60)     {\begin{turn}{12}\small $M$\end{turn}}
3849:   \put(361,86)     {\begin{turn}{12}\small $M$\end{turn}}
3850:   \put(239,124)    {$\phi$}
3851:   \put(338,89)     {$\psi$}
3852:   \put(195.5,80)   {\scriptsize $A$}
3853:   \put(292.5,100)  {\scriptsize $A$}
3854:   \put(0,195)      {$ \MX~ = $}
3855:   \end{picture} \nonumber
3856:   \\[-1.7em]{} \label{eq:1bulk1bnd-rib}
3857:   \end{eqnarray} \xlabp 99
3858: Here the three-manifold has been drawn as in \erf{eq:con-mf-bnd}.
3859: Also, in \erf{eq:1bulk1bnd-rib} it is understood that the $M$-ribbon
3860: that leaves the picture to the left and to the right is closed to a large 
3861: loop. One may now `unfold' the boundary $\partial\MX$ so as to arrive at 
3862: the equivalent representation of the three-manifold used also in 
3863: \erf{eq:uhp-cobord}. This yields
3864:   %  [ pic 4_100 ]
3865:   \bea \begin{picture}(400,139)(0,50)
3866:   \put(36,0)     {\Includeourbeautifulpicture 100 }
3867:   \put(96,101)     {\scriptsize$ z $}
3868:   \put(345.4,43.2) {\scriptsize$ x $}
3869:   \put(117,74)     {\scriptsize$ y $} 
3870:   \put(181,68)     {\scriptsize$ i $}
3871:   \put(153,68)     {\scriptsize$ j $}
3872:   \put(312,68)     {\scriptsize$ k $} 
3873:   \put(276,118)    {\scriptsize$ \psi $}
3874:   \put(200,111)    {\scriptsize$ M $}
3875:   \put(108,145)    {\scriptsize$ A $}
3876:   \put(157,138)    {\scriptsize$ A $}
3877:   \put(252,145)    {\scriptsize$ A $} 
3878:   \put(200,185)    {\scriptsize$ M $}
3879:   \put(206,144)    {\scriptsize$ \phi $} 
3880:   \put(0,90)       {$ \MX \;= $}
3881:   \epicture33 \labl{eq:1bulk1bnd-rib-2} \Xlabp 100
3882: To verify that this cobordism is indeed the same as \erf{eq:1bulk1bnd-rib},
3883: one best works in the full ribbon representation and translates
3884: the picture to blackboard framing only at the very end.
3885: 
3886: The basis of blocks in $\calh(\hat\X)$ in which we expand
3887: $Z(\MX,\emptyset,\hat\X)$ is given by
3888:   %  [ pic 4_32 ]
3889:   \bea \begin{picture}(410,130)(0,43)
3890:   \put(55,0)     {\includeourbeautifulpicture 32 }
3891:   \put(329,34)     {\small $s$}
3892:   \put(249.5,31.4) {\small $3$}
3893:   \put(210,31.4)   {\small $2$}
3894:   \put(171,31.4)   {\small $1$}
3895:   \put(165,14)     {\scriptsize $\bar{p}$}
3896:   \put(207.5,70)   {\scriptsize $p$}
3897:   \put(150,73)     {\scriptsize $y$}
3898:   \put(132,102)    {\scriptsize $z$}
3899:   \put(363,38)     {\scriptsize $x$}
3900:   \put(196,122)    {\scriptsize $U_i$}
3901:   \put(165,122)    {\scriptsize $U_j$}
3902:   \put(180,141)    {\scriptsize $\delta$}
3903:   \put(208,156)    {\scriptsize $U_{\bar k}$}
3904:   \put(334,122)    {\scriptsize $U_k$}
3905:   \put(0,83)       {$ B(x,y,s)_\delta \;= $}
3906:   \epicture29 \labl{eq:Bxys} \xlabp 32
3907: Also indicated in the figure is the continuation path that one has to
3908: choose starting from the standard situation \erf{eq:standard-cobord}.
3909: The structure constants $c(\vhi; M \vsi)_\delta$ are the expansion 
3910: coefficients of \erf{eq:1bulk1bnd-rib-2} in terms of \erf{eq:Bxys},
3911:   \be
3912:   Z(\MX , \emptyset , \hat \X) = 
3913:   \sum_{\delta} c(\vhi; M \vsi)_\delta\,
3914:   Z(B(x,y,s)_\delta , \emptyset , \hat \X) \,.  \labl{eq:1B1b-TFT}
3915: Gluing the cobordism dual to $B(x,y,s)_\delta$ selects one term in the 
3916: sum on the \rhs\ and thus yields, after a rotation by $180^\circ$,
3917: the ribbon graph 
3918:   %  [ pic 4_101 ]
3919:   \bea \begin{picture}(370,133)(0,53)
3920:   \put(86,0)     {\Includeourbeautifulpicture 101 } 
3921:   \put(234,88)   {\scriptsize$ i $}
3922:   \put(270,88)   {\scriptsize$ j $} 
3923:   \put(149,10)   {\scriptsize$ k $}
3924:   \put(205,8)    {\scriptsize$ \overline{k} $} 
3925:   \put(147,108)  {\scriptsize$ \psi $} 
3926:   \put(295,103)  {\scriptsize$ M $}
3927:   \put(305,124)  {\scriptsize$ A $}
3928:   \put(253,129)  {\scriptsize$ A $}
3929:   \put(161,134)  {\scriptsize$ A $} 
3930:   \put(217,175)  {\scriptsize$ M $}
3931:   \put(207,134)  {\scriptsize$ \phi $}
3932:   \put(368.5,172.2){$ \Stn $}
3933:   \put(0,88)     {$ c(\vhi; M \vsi)_\delta \;= $}
3934:   \epicture20 \labl{eq:1bulk-1bnd-sc} \Xlabp 101
3935: 
3936: \dtl{Remark}{rem:1bulk-uhp}
3937: An interesting special case is the correlator of one bulk
3938: field on the upper half plane, without any boundary field insertion.
3939: The corresponding structure constant $c(\vhi;M)$ can be obtained 
3940: from \erf{eq:1bulk-1bnd-sc} by taking the boundary field to be the 
3941: identity, $\vsi \eq (M,M,\one,\id_{\dot M},[\gamma])$. Doing so, one gets
3942:   %  [ pic 4_102 ]
3943:   \bea \begin{picture}(290,152)(0,46)
3944:   \put(76,0)     {\Includeourbeautifulpicture 102 }
3945:   \put(0,92)     {$ c(\vhi; M) \;= $}
3946:   \put(157,58)   {\scriptsize$ i $}
3947:   \put(178,58)   {\scriptsize$ \overline{i} $}
3948:   \put(175,108)  {\scriptsize$ i $}
3949:   \put(195,108)  {\scriptsize$ \overline{i} $} 
3950:   \put(73,108)   {\scriptsize$ i $} 
3951:   \put(167,80)   {\scriptsize$ \phi $} 
3952:   \put(170.3,39.6) {\scriptsize$ A $}
3953:   \put(160,101)  {\scriptsize$ A $}
3954:   \put(148,130)  {\scriptsize$ A $} 
3955:   \put(190,193)  {\scriptsize$ M $}
3956:   \put(240,66)   {\scriptsize$ M $}
3957:   \put(279,164.5){$ \Stn $}
3958:   \epicture32 \labl{eq:1bulk-sc} \Xlabp 102
3959: To obtain \erf{eq:1bulk-sc} from \erf{eq:1bulk-1bnd-sc}, one needs to get 
3960: rid of one of the $A$-ribbons that are attached to the 
3961: $M$-ribbon, by first moving the two representation morphisms close to each 
3962: other, and then using the representation property (I:4.2), taking the 
3963: resulting multiplication morphism past $\phi$ (which is allowed because 
3964: $\phi$ commutes in particular with the left action of $A$ on itself). The 
3965: resulting $A$-loop can be omitted owing to specialness, so that one
3966: is left with the unit $\eta$. To arrive at \erf{eq:1bulk-sc} it then only 
3967: remains to deform the resulting ribbon graph a bit.
3968: \\
3969: A ribbon graph of the form \erf{eq:1bulk-sc} was already encountered in
3970: the definition of the quantity $\tilde S^A$ in (II:3.89). It follows that
3971:   \be
3972:   c(\vhi; M ) = \frac{\tilde S^A(M;U_i,\tilde\phi)}{S_{0,0}}
3973:   \,, \ee
3974: where $\tilde\phi$ is the morphism enclosed in a dashed box
3975: in \erf{eq:1bulk-sc}. 
3976: 
3977: \medskip
3978: 
3979: Next we express the invariant \erf{eq:1bulk-1bnd-sc} in a basis, making the 
3980: simplifying assumption that $\N ijk \iN \{0,1\}$ and that $\langle U_i,A 
3981: \rangle \iN \{0,1\}$. This implies in particular that $A$ is haploid and thus 
3982: simple. By convention, the bimodules are then labelled in such a way that 
3983: $X_0\eq A$. Let us choose the boundary condition and the morphisms for 
3984: the bulk and boundary field in $c(\vhi;M \hatpsi)$ as
3985:   \be
3986:   M = M_\mu \,, \qquad
3987:   \psi = \psi^\alpha_{(\mu k)\mu} \,, \qquad
3988:   \phi = \xi^\beta_{(i0j)0} \,.
3989:   \labl{eq:1bulk1bnd-basis-def}
3990: Note that \erf{eq:1bulk-1bnd-sc} can be slightly simplified by 
3991: removing the $A$-ribbon that arrives at $\phi$ in the same way as explained 
3992: in remark \ref{rem:1bulk-uhp}. Substituting \erf{eq:1bulk1bnd-basis-def} 
3993: into \erf{eq:1bulk-1bnd-sc} leads to the ribbon graph
3994:   %  [ pic 4_83 ]
3995:   \bea \begin{picture}(340,260)(0,0)
3996:   \put(90,30)   {\includeourbeautifulpicture 83}
3997:   \put(206,90)     {\scriptsize $U_i$}
3998:   \put(233,90)     {\scriptsize $U_j$}
3999:   \put(232.4,168.3){\scriptsize $\beta$}
4000:   \put(141.4,78)   {\scriptsize $\alpha$}
4001:   \put(187.4,46)   {\scriptsize $U_{\bar k}$}
4002:   \put(163,46)     {\scriptsize $U_k$}
4003:   \put(224.7,134)  {\scriptsize $A$}
4004:   \put(230,204.3)  {\scriptsize $A$}
4005:   \put(279.6,170)  {\scriptsize $M_{\mu}$}
4006:   \put(94.7,140)   {\scriptsize $M_{\mu}$}
4007:   \put(323,230.3)  {$ \Stn $}
4008:   \put(0,140)      {$ c(\phi;M \hatpsi) ~= $}
4009:   \put(58,0)       {$ \dsty 
4010:             =~ \sum_{m,n,a} \sum_{\rho,\sigma} [\xi^\beta_{(i0j)0}]^{0ja}
4011:             \rho^{{M_\mu}\,(m\rho)}_{a\,(n\sigma)} 
4012:             [\psi^\alpha_{(\mu k)\mu}]^{mn}_{\rho\sigma}\,I(i,j,k,a,m,n) \,,$}
4013:   \epicture-5 \labl{eq:1B1b-basis} \xlabp 83
4014: where in the second step we have substituted the expansions
4015: of $\psi^\alpha_{(\mu k)\mu}$ and $\xi^\beta_{(i0j)0}$
4016: given in \erf{eq:psi-expansion} and \erf{eq:xi-expansion}.
4017: The coefficients $\rho^M$, introduced in (I:4.56), express a representation 
4018: morphism in a basis. The function $I(i,j,k,a,m,n)$ is the ribbon invariant
4019:   %  [ pic 4_84 ]
4020:   \bea \begin{picture}(365,220)(0,0)
4021:   \put(120,0)   {\includeourbeautifulpicture 84}
4022:   \put(0,100)     {$ \dsty I(i,j,k,a,m,n) ~= $}
4023:   \put(220,90)    {\scriptsize $i$}
4024:   \put(282,82)    {\scriptsize $j$}
4025:   \put(196,11.5)  {\scriptsize $k$}
4026:   \put(223,11.5)  {\scriptsize $\bar{k}$}
4027:   \put(251,160)   {\scriptsize $a$}
4028:   \put(254,86.5)  {\scriptsize $n$}
4029:   \put(125,107)   {\scriptsize $m$}
4030:   \put(134,0)     {\small \dashbox{2.2}(11,11){1} }
4031:   \put(201,84)    {\small \dashbox{2.2}(11,11){2} }
4032:   \put(265,121.2) {\small \dashbox{2.2}(11,11){3} }
4033:   \put(320.7,192.5){$ \Stn $}
4034:   \epicture-7 \labl{eq:1B1b-I-def} \xlabp 84
4035: To evaluate this invariant one first applies the appropriate relation from 
4036: appendix II:A.1 to the dashed box marked {\small\dashbox{2.2}(11,11){1}}\,.
4037: Next one inserts a complete basis (I:2.31) at {\small\dashbox{2.2}(11,11){2}}
4038: and removes all braidings with the help of (I:2.41). Finally one applies an 
4039: $\FF$-move (II:2.36) to {\small\dashbox{2.2}(11,11){3}}\,, so as to obtain
4040:   \be
4041:   I(i,j,k,a,m,n) = 
4042:   \GG^{(mk\bar k)m}_{n0}\, \Rs{}nam \dim(U_m)
4043:   \sum_p \frac{\theta_n \theta_i}{\theta_p}\, \Fs nijmap\, \Gs nijmp{\bar k} \,.
4044:   \ee
4045: 
4046: The correlators for one bulk field and one boundary field on the upper half 
4047: plane, i.e.\ the structure 
4048: constants describing the coupling of bulk to boundary fields, have first been 
4049: considered in \cite{cale}. There, a constraint is derived which these 
4050: coefficients have to solve and an explicit expression for their square is found 
4051: for the case of the charge conjugation invariant. Since then, solutions for the 
4052: coefficients themselves have been found, both for models with charge conjugation
4053: invariant and for more general cases \cite{runk,bppz2,runk2,fgrs,fffs3}.
4054: Also, in equation (5.6) of \cite{pezu6} the bulk-boundary coefficients of 
4055: spinless bulk fields are expressed through $^{(1)\!}F$.  
4056: 
4057: The coupling of a bulk field to the identity on the boundary, i.e.\ the 
4058: one-point function of a bulk field on the disk as in remark \ref{rem:1bulk-uhp},
4059: is easier to obtain than the general bulk-boundary structure constants, as it 
4060: can already be recovered from the NIM-rep data (multiplicities in the tensor 
4061: products of $\calc$ and $\calca$) and does not require knowledge of the 
4062: 6j-symbols \cite{card9,cale}. These coefficients, which determine the boundary 
4063: state, have been calculated for many models.
4064: 
4065: 
4066: \subsection{Three bulk fields on the sphere} \label{sec:rc-3bulk}
4067: 
4068: We now turn to the correlator of three bulk fields 
4069: on the sphere. Analogously as in the discussion of the disk, we describe the 
4070: sphere as $\reals^2$ plus a point. The correlator then takes the form
4071:   %  [ pic 4_35 ]
4072:   \bea \begin{picture}(150,76)(0,27)
4073:   \put(0,0)     {\includeourbeautifulpicture 35 }
4074:   \put(2,100)    {\scriptsize $y$}
4075:   \put(151,27)   {\small $x$}
4076:   \put(39,38)    {\scriptsize $p_1$}
4077:   \put(75,3)     {\scriptsize $p_2$}
4078:   \put(111,55)   {\scriptsize $p_3$}
4079:   \put(38,51)    {\small $\vhi_1$}
4080:   \put(74,17)    {\small $\vhi_2$}
4081:   \put(109,68)   {\small $\vhi_3$}
4082:   \epicture14 \labl{eq:3bulk-geom} \xlabp 35
4083: The dotted line again indicates the dual triangulation that will be used 
4084: in the construction of the ribbon graph. The three bulk fields are tuples
4085:   \be
4086:   \vhi_1 = (i,j,\phi_1,p_1,[\gamma_1],\orr_2) \,, \quad
4087:   \vhi_2 = (k,l,\phi_2,p_2,[\gamma_2],\orr_2) \,, \quad
4088:   \vhi_3 = (m,n,\phi_3,p_3,[\gamma_3],\orr_2) \,. \ee
4089: Here $i,j,k,l,m,n \iN \II$ label simple objects. $\phi_1$ is a morphism in
4090: $\HomAA(U_i\,{\otimes^+}\,A\,{\otimes^-}\,U_j,A)$, and similarly one has
4091: $\phi_2\iN\HomAA(U_k\,{\otimes^+}\,A\,{\otimes^-}\,U_l,A)$ and
4092: $\phi_3\iN\HomAA(U_m\,{\otimes^+}\,A\,{\otimes^-}\,U_n,A)$.  
4093: The three fields are inserted at $p_1$, $p_2$ and $p_3$,
4094: respectively, with arc germs $[\gamma_1]$, $[\gamma_2]$, $[\gamma_3]$ given by
4095: $\gamma_{1,2,3}(t)\eq p_{1,2,3} \,{+}\, (t,0)$. The orientations
4096: $\orr_2$ around the insertion points of the three bulk fields
4097: are given by the standard orientation of $\reals^2$.
4098: 
4099:     \newpage
4100: 
4101: The construction of section \ref{sec:fi} gives the following
4102: ribbon graph inscribed in the connecting manifold:
4103:   %  [ pic 4_103 ]
4104:   \begin{eqnarray} \begin{picture}(420,333)(25,0)
4105:   \put(0,0)      {\Includeournicehugepicture 103 }
4106:   \put(153,5)       {\tiny $x$}
4107:   \put(137,15)      {\tiny $y$}
4108:   \put(153,51)      {\tiny $x$}
4109:   \put(137,60)      {\tiny $y$}
4110:   \put(135,114)     {\tiny $x$}
4111:   \put(128,105)     {\tiny $y$}
4112:   \put(229,181)     {\tiny $x$}
4113:   \put(207,192)     {\tiny $y$}
4114:   \put(214,195)     {\tiny $z$}
4115:   \put(152,110)     {\scriptsize $i$}
4116:   \put(144,171)     {\scriptsize $j$}
4117:   \put(251,135)     {\scriptsize $l$}
4118:   \put(248,69)      {\scriptsize $k$}
4119:   \put(345,181)     {\scriptsize $n$}
4120:   \put(343.2,123)   {\scriptsize $m$}
4121:   \put(276,140)     {\scriptsize $A$}
4122:   \put(188,130)     {\scriptsize $A$}
4123:   \put(209,250)     {\scriptsize $A$}
4124:   \put(148,150)     {\small $\phi_1$}
4125:   \put(245,116)     {\small $\phi_2$}
4126:   \put(339,161)     {\small $\phi_3$}
4127:   \put(5,320)       {$ \MX~ = $}
4128:   \end{picture} \nonumber
4129:   \\[.5em]{} \label{eq:3bulk-rib}
4130:   \end{eqnarray} \Xlabp 103
4131: As a basis for the space $\calh(\hat\X)$ we take the cobordisms
4132:   %  [ pic 4_37 ]
4133:   \begin{eqnarray} \begin{picture}(420,434)(0,0)
4134:   \put(102,220)  {\Includeourbeautifulpicture 37a }
4135:   \put(175,0)    {\Includeourbeautifulpicture 37b }
4136:   \put(202,371)     {\scriptsize $p_1$}
4137:   \put(200,335)     {\scriptsize $j$}
4138:   \put(217,286)     {\scriptsize $\nu$}
4139:   \put(240,407.5)   {\scriptsize $p_2$}
4140:   \put(246,335)     {\scriptsize $l$}
4141:   \put(307.8,388.5) {\scriptsize $p_3$}
4142:   \put(313.5,322)   {\scriptsize $n$}
4143:   \put(243,267)     {\scriptsize $\bar{n}$}
4144:   \put(360,423)     { $B^3$}
4145:   \put(132,359)     {\tiny $y$}
4146:   \put(143,341)     {\tiny $x$}
4147:   \put(119,364)     {\tiny $z$}
4148:   \put(275,58.5)    {\scriptsize $p_1$}
4149:   \put(273,89)      {\scriptsize $i$}
4150:   \put(288.7,140)   {\scriptsize $\mu$}
4151:   \put(312.5,17.3)  {\scriptsize $p_2$}
4152:   \put(319,90)      {\scriptsize $k$}
4153:   \put(381,38.8)    {\scriptsize $p_3$}
4154:   \put(387,129)     {\scriptsize $m$}
4155:   \put(307,155)     {\scriptsize $\bar{m}$}
4156:   \put(426,198.5)   { $B^3$}
4157:   \put(233,143)     {\tiny $y$}
4158:   \put(242,127)     {\tiny $x$}
4159:   \put(219,151)     {\tiny $z$}
4160:   \put(0,330)      {$ B(p_1,p_2,p_3)_{\mu\nu} \;= $}
4161:   \put(148,90)     {$ \sqcup $}
4162:   \end{picture} \nonumber
4163:   \\[-1.4em]{} \label{eq:3bulk-bas}
4164:   \end{eqnarray} \xlabp 37
4165: where the symbol `$B^3$' indicates that the manifolds are solid three-balls,
4166: with orientations as indicated, while `$\sqcup$'
4167: denotes the disjoint union of the two three-manifolds. 
4168: Expanding \erf{eq:3bulk-rib} in terms of \erf{eq:3bulk-bas} defines the bulk 
4169: structure constants $c(\vhi_1 \vhi_2 \vhi_3)_{\mu\nu}$ via
4170:   \be
4171:   Z(\MX , \emptyset , \hat \X) = 
4172:   \sum_{\mu\nu} c(\vhi_1 \vhi_2 \vhi_3)_{\mu\nu} \,
4173:   Z(B(p_1,p_2,p_3)_{\mu\nu},\emptyset,\hat \X) \,. 
4174:   \labl{Z3bul}
4175: By gluing the cobordism dual to $B(p_1,p_2,p_3)_{\mu\nu}$ we get rid of the 
4176: summation on the \rhs, thus obtaining the ribbon graph 
4177:   %  [ pic 4_104 ]
4178:   \bea \begin{picture}(360,139)(0,40)
4179:   \put(105,0)     {\Includeourbeautifulpicture 104 }
4180:   \put(-65,0){
4181:   \put(218,61)    {\scriptsize $i$}
4182:   \put(262,61)    {\scriptsize $k$}
4183:   \put(267,8)     {\scriptsize $\bar{m}$}
4184:   \put(300,8)     {\scriptsize ${m}$}
4185:   \put(249,32)    {\footnotesize $\bar{\mu}$}
4186:   \put(237.9,80)  {\footnotesize $\phi_1$}
4187:   \put(295.3,80)  {\footnotesize $\phi_2$}
4188:   \put(353.1,80)  {\footnotesize $\phi_3$}
4189:   \put(253.4,85)  {\scriptsize $A$}
4190:   \put(312,85)    {\scriptsize $A$}
4191:   \put(330,173)   {\scriptsize $A$}
4192:   \put(235,130)   {\footnotesize $\bar{\nu}$}
4193:   \put(420.6,165.5){ \small $\Stn$}
4194:   \put(217,102)   {\scriptsize $j$}
4195:   \put(264,101)   {\scriptsize $l$}
4196:   \put(261,150)   {\scriptsize $\bar{n}$}
4197:   \put(310,150)   {\scriptsize ${n}$}
4198:   }
4199:   \put(-10,86)      {$ \dsty c(\vhi_1\vhi_2\vhi_3)_{\mu\nu} \;=\;\frac{1}{S_{0,0}} $}
4200:   \epicture20 \labl{eq:3bulk-sc-aux} \Xlabp 104
4201: The additional factor $1/S_{0,0}$ comes from the normalisation of the cobordism
4202: dual to $B(p_1,p_2,p_3)_{\mu\nu}$, which consists of two copies of $B^3$.
4203: 
4204: \medskip
4205: 
4206: Let us assume that the algebra $A$ is simple, and thus by convention 
4207: $X_0\eq A$. Also, we choose the morphisms entering the three bulk fields 
4208: in $c(\vhi_1 \vhi_2 \vhi_3)_{\eps\varphi}$ as
4209:   \be
4210:   \phi_1 = \xi^{\alpha_1}_{(i0j)0} \, , \qquad
4211:   \phi_2 = \xi^{\alpha_2}_{(k0l)0} \, , \qquad
4212:   \phi_3 = \xi^{\alpha_3}_{(m0n)0} \,.
4213:   \ee
4214: Inserting these expressions in \erf{eq:3bulk-sc-aux} and rotating the
4215: resulting graph in such a way that the white side of the $A$-ribbons
4216: faces the reader results in
4217:   %  [ pic 4_85 ]
4218:   \bea \begin{picture}(315,147)(0,52)
4219:   \put(99,0)     {\includeourbeautifulpicture 85 }
4220:   \put(144,75)    {\scriptsize $i$}
4221:   \put(129,75)    {\scriptsize $k$}
4222:   \put(91.5,60)   {\scriptsize $m$} 
4223:   \put(140,32)    {\scriptsize $\bar{m}$}
4224:   \put(153,115)   {\footnotesize $\alpha_1$}
4225:   \put(153,140)   {\footnotesize $\alpha_2$}
4226:   \put(153,163)   {\footnotesize $\alpha_3$}
4227:   \put(173,21)    {\scriptsize $X_0$}
4228:   \put(173.6,121) {\scriptsize $X_0$}
4229:   \put(173.6,148) {\scriptsize $X_0$}
4230:   \put(132,46)    {\footnotesize $\bar{\varepsilon}$}
4231:   \put(301,185)   { \small $\Stn$}
4232:   \put(195,74)    {\scriptsize $j$}
4233:   \put(210.9,74)  {\scriptsize $l$}
4234:   \put(197.7,32)  {\scriptsize $\bar{n}$}
4235:   \put(242,60)    {\scriptsize ${n}$}
4236:   \put(207,49)    {\footnotesize $\bar{\varphi}$}
4237:   \put(-19,94) {$ \dsty c(\vhi_1\vhi_2\vhi_3)_{\eps\varphi}~=~\frac{1}{S_{0,0}} $}
4238:   \epicture36 \labl{eq:3bulk-basis-aux} \xlabp 85
4239: Applying now the definition \erf{eq:FAA-def} of the $\FF[A|A]$-matrices
4240: twice, first to the pair $\alpha_1$, $\alpha_2$ of morphisms, and then to the 
4241: remaining two bimodule morphisms, the invariant \erf{eq:3bulk-basis-aux} becomes
4242:   \be
4243:   c(\vhi_1 \vhi_2 \vhi_3)_{\eps\varphi} = \frac{\dim(A)}{S_{0,0}}
4244:   \sum_{\beta} \FF[A|A]^{(ki0jl)0}_{\alpha_1 0 \alpha_2\,,\,
4245:     \beta \bar m \bar n \eps \varphi} \,
4246:   \FF[A|A]^{(m \bar m 0 \bar n n)0}_{\beta 0 \alpha_3 \,,\,
4247:   \cdot 0 0 \cdot \cdot} \,.
4248:   \labl{eq:3bulk-coeff}
4249: Multiplying this with the inverse of the two-point correlator, one finds that 
4250: the bulk OPE coefficients are given precisely by entries of the matrix 
4251: $\FF[A|A]$, as announced in remark \ref{rem:bulk-6j}.
4252: 
4253: Of the various types of structure constants considered in this paper, the 
4254: coefficient for the correlator of three bulk fields has the longest history. 
4255: Explicit expressions have first been obtained for the A-series Virasoro minimal 
4256: models \cite{dofa2,dofa3}, and subsequently for many models with charge 
4257: conjugation invariant \cite{knza,zafa2,chfl,musS2,kikkmo} and more general 
4258: theories \cite{dmfs,jf12,fukl,petk2,dotr,kaki,fuks,rest,pezu,bppz2}.
4259: 
4260: 
4261: \subsection{Three defect fields on the sphere} \label{sec:rc-3defect}
4262: 
4263: The calculation for three defect fields is almost the same as for three bulk 
4264: fields. This is not surprising, as a bulk field is just a special kind
4265: of defect field, namely the one connecting the invisible defect to the 
4266: invisible defect. We consider three defect fields on a defect circle,
4267:   %  [ pic 4_46 ]
4268:   \bea \begin{picture}(160,76)(0,30)
4269:   \put(0,0)     {\includeourbeautifulpicture 46 }
4270:   \put(2,100)     {\scriptsize $y$}
4271:   \put(151,27.7)  {\scriptsize $x$}
4272:   \put(39,38)     {\scriptsize $p_1$}
4273:   \put(75,3)      {\scriptsize $p_2$}
4274:   \put(112,55)    {\scriptsize $p_3$}
4275:   \put(37,51)     {\small $\Thetha_1$}
4276:   \put(74,16)     {\small $\Thetha_2$}
4277:   \put(110,68)    {\small $\Thetha_3$}
4278:   \put(65,33)     {\small $Y$}
4279:   \put(100.8,33)  {\small $Z$}
4280:   \put(85,92)     {\small $X$} 
4281:   \epicture14 \labl{eq:3defect-geom} \xlabp 46
4282: The three defect fields are tuples
4283:   \bea
4284:   \Thetha_1=(X,\orr_2,Y,\orr_2,i,j,\vartheta_1,p_1,[\gamma_1],\orr_2)\,,\\[.6em]
4285:   \Thetha_2=(Y,\orr_2,Z,\orr_2,k,l,\vartheta_2,p_2,[\gamma_2],\orr_2)\,,\\[.6em]
4286:   \Thetha_3=(Z,\orr_2,X,\orr_2,m,n,\vartheta_3,p_3,[\gamma_3],\orr_2)\,.
4287:   \eear\ee
4288: Here $\orr_2$ is the standard orientation of $\reals^2$, $i,j,k,l,m,n\in\II$ 
4289: label simple objects, and $\vartheta_{1,2,3}$ are morphisms
4290:   \bea
4291:   \vartheta_1 \in \HomAA(U_i\,{\otimes^+}\,X\,{\otimes^-}\,U_j,Y) \,,\\[.6em]
4292:   \vartheta_2 \in \HomAA(U_k\,{\otimes^+}\,Y\,{\otimes^-}\,U_l,Z) \,,\\[.6em]
4293:   \vartheta_3 \in \HomAA(U_m\,{\otimes^+}\,Z\,{\otimes^-}\,U_n,X) \,.
4294:   \eear\ee
4295: The three fields are inserted at $p_1$, $p_2$, $p_3$, with arc germs 
4296: $[\gamma_{1,2,3}]$ given by $\gamma_{1,2,3}(t)\eq p_{1,2,3} \,{+}\, (t,0)$. 
4297: 
4298: \medskip
4299: 
4300: The ribbon graph differs from the one for three bulk fields only in that the 
4301: annular $A$-ribbon is replaced by the appropriate bimodule ribbons. Thus we 
4302: obtain
4303:   %  [ pic 4_47 ]
4304:   \begin{eqnarray} \begin{picture}(420,333)(25,0)
4305:   \put(0,0)      {\includeournicehugepicture 47 }
4306:   \put(153,5)       {\tiny $x$}
4307:   \put(137,15)      {\tiny $y$}
4308:   \put(153,51)      {\tiny $x$}
4309:   \put(137,60)      {\tiny $y$}
4310:   \put(135,114)     {\tiny $x$}
4311:   \put(128,105)     {\tiny $y$}
4312:   \put(229,181)     {\tiny $x$}
4313:   \put(207,192)     {\tiny $y$}
4314:   \put(214,195)     {\tiny $z$}
4315:   \put(152,110)     {\scriptsize $i$}
4316:   \put(144,170)     {\scriptsize $j$}
4317:   \put(250,134)     {\scriptsize $l$}
4318:   \put(248,69)      {\scriptsize $k$}
4319:   \put(345,180)     {\scriptsize $n$}
4320:   \put(343,123)     {\scriptsize $m$}
4321:   \put(277,140)     {\scriptsize $Z$}
4322:   \put(188,129)     {\scriptsize $Y$}
4323:   \put(205,250)     {\scriptsize $X$}
4324:   \put(148,149)     {\small $\vartheta_1$}
4325:   \put(244,115.7)   {\small $\vartheta_2$}
4326:   \put(339,159.7)   {\small $\vartheta_3$}
4327:   \put(5,320)       {$ \MX~ = $}
4328:   \end{picture} \nonumber
4329:   \\[.5em]{} \label{eq:3defect-rib}
4330:   \end{eqnarray} \xlabp 47
4331: Expanding this in the basis \erf{eq:3bulk-bas} defines
4332: the structure constants of three defect fields via
4333:   \be
4334:   Z(\MX , \emptyset , \hat \X) = 
4335:   \sum_{\mu\nu} c(X,\Thetha_1, Y ,\Thetha_2 ,Z, \Thetha_3, X)_{\mu\nu} \,
4336:   Z(B(p_1,p_2,p_3)_{\mu\nu},\emptyset,\hat \X) \,. 
4337:   \labl{Z3def}
4338: Gluing the cobordism dual to $B(p_1,p_2,p_3)_{\mu\nu}$ yields a ribbon 
4339: invariant very similar to \erf{eq:3bulk-sc-aux}:
4340:   %  [ pic 4_86 ]
4341:   \bea \begin{picture}(400,185)(0,0)
4342:   \put(170,0)   {\includeourbeautifulpicture 86}
4343:   \put(218,61)    {\scriptsize $i$}
4344:   \put(262,61)    {\scriptsize $k$}
4345:   \put(267,8)     {\scriptsize $\bar{m}$}
4346:   \put(300,8)     {\scriptsize ${m}$}
4347:   \put(249,32)    {\footnotesize $\bar{\mu}$}
4348:   \put(237.9,80)  {\footnotesize $\vartheta_1$}
4349:   \put(295.3,80)  {\footnotesize $\vartheta_2$}
4350:   \put(353.1,80)  {\footnotesize $\vartheta_3$}
4351:   \put(256,85)    {\scriptsize $Y$}
4352:   \put(312,85)    {\scriptsize $Z$}
4353:   \put(330,176)   {\scriptsize $X$}
4354:   \put(235,130)   {\footnotesize $\bar{\nu}$}
4355:   \put(421,165.4) { \small $\Stn$}
4356:   \put(218,101)   {\scriptsize $j$}
4357:   \put(264,101)   {\scriptsize $l$}
4358:   \put(261,150)   {\scriptsize $\bar{n}$}
4359:   \put(310,150)   {\scriptsize ${n}$}
4360:   \put(-20,80)    {$ \dsty c(X,\Thetha_1, Y ,\Thetha_2 ,Z, \Thetha_3, X)
4361:                      _{\mu\nu}^{} \;=~ \frac{1}{S_{0,0}} $}
4362:   \epicture{-1}4 \labl{eq:3defect-sc} \xlabp 86
4363: The calculation of the invariant \erf{eq:3defect-sc} runs along the 
4364: same line as in the case of three bulk fields. Let us choose
4365:   \be
4366:   X = X_\mu \,, ~~ Y = X_\nu \,, ~~ Z = X_\kappa \,, \quad
4367:   \vartheta_1 = \xi^{\alpha_1}_{(i \mu j)\nu} \,, ~~
4368:   \vartheta_2 = \xi^{\alpha_2}_{(k \nu l)\kappa} \,, ~~
4369:   \vartheta_3 = \xi^{\alpha_3}_{(m \kappa n)\mu} \,.
4370:   \ee
4371: Inserting this in \erf{eq:3defect-sc} leads to a ribbon invariant similar
4372: to \erf{eq:3bulk-basis-aux}, which can again be evaluated with the help
4373: of the definition \erf{eq:FAA-def}. We find
4374:   \be
4375:   c(X_\mu,\Thetha_1, X_\nu ,\Thetha_2 ,X_\kappa, \Thetha_3, X_\mu)_{\eps\varphi} 
4376:   = \frac{\dim(\dot X_\mu)}{S_{0,0}}
4377:   \sum_{\beta} \FF[A|A]^{(ki\mu jl)\kappa}_{\alpha_1 \nu \alpha_2\,,\,
4378:     \beta \bar m \bar n \eps \varphi} \,
4379:   \FF[A|A]^{(m \bar m \mu \bar n n)\mu}_{\beta \kappa \alpha_3 \,,\,
4380:   \cdot 0 0 \cdot \cdot} \,.  \labl{eq:3def-basis}
4381: Just like for three bulk fields, multiplying this expression with the inverse 
4382: of the two-point correlator one finds that the OPE coefficients for defect 
4383: fields are entries of the matrix $\FF[A|A]$.
4384: 
4385: 
4386: \subsection{One bulk field on the cross cap} \label{sec:rc-1bulk-xcap}
4387: 
4388: After dealing with the fundamental correlators allowing to obtain all
4389: orientable world sheets by sewing, we finally consider the additional
4390: world sheet that is needed to construct all non-orientable surfaces as well,
4391: i.e.\ the cross cap. Correspondingly, the algebra $A$ is now a Jandl algebra.
4392: 
4393: We present the cross cap as $\reals^2{/}{\sim}$, with the equivalence relation 
4394: $\sim$ identifying points according to the rule $u \,{\sim}\,\sigrp(u)$,
4395: where in standard coordinates $x,y$ on $\reals^2$, $\sigrp$ acts as
4396:   \be
4397:   \sigrp(x,y) := \Frac{-1}{x^2+y^2}\, (x,y) \,.
4398:   \ee 
4399: Consider now the correlator of one bulk field on the cross cap,
4400: with the bulk field given by $\vhi\eq (i,j,\phi,p,[\gamma],\orr_2)$:
4401:   %  [ pic 4_105 ]
4402:   \bea \begin{picture}(160,170)(0,0)
4403:   \put(0,0)   {\Includeourbeautifulpicture 105}
4404:   \put(72,159)     {\small $y$}
4405:   \put(156,72.5)   {\small $x$}
4406:   \put(121,84.5)   {\tiny $x$}
4407:   \put(100,106.4)  {\tiny $y$} 
4408:   \put(104,65)   {\small $p$}
4409:   \put(104,76)   {\small $\vhi$}
4410:   \epicture-4 \labl{eq:1bulk-xcap-geom} \Xlabp 105
4411: As before, $i,j\iN\II$, $\phi\iN\HomAA(U_i\otiP A\otim U_j,A)$
4412: and $\gamma(t)\eq p \,{+}\, (t,0)$, where $p$ is the insertion point
4413: of the bulk field. The local orientation $\orr_2$ around $p$
4414: is chosen as indicated in figure \erf{eq:1bulk-xcap-geom}.
4415: The interior of the dashed circle is a fundamental domain of the equivalence
4416: relation $u \,{\sim}\,\sigrp(u)$. On its boundary, i.e.\ for $|u|\eq 1$,
4417: points are identified according to $u \,{\sim}\,{-}u$.
4418: 
4419: \medskip
4420: 
4421: Similarly as in (II:3.10), one can write the connecting manifold as
4422:   \be
4423:   \MX = \hat \X \times [-1,0] \,{/} \sim
4424:   \qquad {\rm with} \quad (u,0) \,{\sim}\, (\sigrp(u),0) 
4425:   {\rm ~~for~all~} u \iN \hat\X \,.  \ee
4426: For the cross cap we obtain in this way the three-manifold
4427:   %  [ pic 4_106 ]
4428:   \bea \begin{picture}(280,180)(0,0)
4429:   \put(0,0)   {\Includeourbeautifulpicture 106}
4430:   \put(26,34)      {\tiny $y$}
4431:   \put(40,8.5)     {\tiny $x$}
4432:   \put(42.6,86.7)  {\tiny $x$}
4433:   \put(34,108)     {\tiny $y$}
4434:   \put(24,112.3)   {\tiny $z$}
4435:   \put(251,95)     {\ $t=-1$}
4436:   \put(251,165)    {\ $t=0$}
4437:   \epicture-4 \labl{eq:xcap-3mfld}  \Xlabp 106
4438: with orientations as indicated, and with points in the $t{=}0$\,-plane
4439: to be identified according to $u \,{\sim}\, \sigrp(u)$. 
4440: Next we draw the ribbon graph embedded in the $t{=}0$\,-plane:
4441:   %  [ pic 4_107 ]
4442:   \bea \begin{picture}(260,160)(0,20)
4443:   \put(0,0)   {\Includeourbeautifulpicture 107}
4444:   \put(17,101)     {\scriptsize $i$}
4445:   \put(16.6,74)    {\scriptsize $j$}
4446:   \put(161.5,104.5){\scriptsize $j$}
4447:   \put(160.9,71.9) {\scriptsize $i$}
4448:   \put(182.4,87)   {\small $\phi$}
4449:   \put(48.5,87)    {\small $\phi$}
4450:   \put(0,47)       {\scriptsize $A$}
4451:   \put(65,93)      {\scriptsize $A$}
4452:   \put(120,130)    {\scriptsize $A$}
4453:   \put(233,55)     {\scriptsize $A$}
4454:   \put(137,174)    {\scriptsize $y$}
4455:   \put(238,88.2)   {\scriptsize $x$}
4456:   \epicture07 \labl{eq:xcap-ribgr}  \Xlabp 107
4457: Note that there is only a single field insertion, but as a consequence of the 
4458: identification $u \,{\sim}\,\sigrp(u)$ the ribbon graph is duplicated in the 
4459: picture. In \erf{eq:xcap-ribgr} the $U_i$- and $U_j$-ribbons end in dashed 
4460: circles; this way we indicate the points at which they leave the $t{=}0$-plane 
4461: so as to connect to the marked points on the boundary of $\MX$. 
4462: Recall also the notation (II:3.4) for the element connecting
4463: two $A$-ribbons with opposite orientation.
4464: 
4465: The complete ribbon graph, lifted out of the $t{=}0$-plane to $t\,{<}\,0$ and 
4466: rotated into blackboard framing, looks as follows
4467: (again one better uses full ribbon notation to verify this move):
4468:   %  [ pic 4_108 ]
4469:   \bea \begin{picture}(360,226)(0,41)
4470:   \put(58,0)   {\Includeourbeautifulpicture 108}
4471:   \put(79.6,34)    {\tiny $y$}
4472:   \put(94.6,6)     {\tiny $x$}
4473:   \put(80,102.2)   {\tiny $x$}
4474:   \put(71,124)     {\tiny $y$}
4475:   \put(61.6,128)   {\tiny $z$}
4476:   \put(310,93)     {\ $t\eq{-}1$}
4477:   \put(310,254)    {\ $t\eq0$}
4478:   \put(204,106)    {\scriptsize $i$}
4479:   \put(126,108)    {\scriptsize $j$}
4480:   \put(204,150)    {\scriptsize $j$}
4481:   \put(227.5,126)  {\small $\phi$}
4482:   \put(183,131)    {\scriptsize $A$}
4483:   \put(282,131)    {\scriptsize $A$}
4484:   \put(0,123)  {$ \MX \;= $}
4485:   \epicture26 \labl{eq:xcap-ribbon-emb}  \Xlabp 108 
4486: The space $\calh(\hat X)$ in this case is one-dimensional if 
4487: $j\eq\bar\imath$ and zero-dimensional otherwise. For $j\eq\bar \imath$ we select
4488:   %  [ pic 4_52 ]
4489:   \begin{eqnarray} \begin{picture}(420,163)(10,0)
4490:   \put(44,0)  {\Includeourbeautifulpicture 52a}
4491:   \put(81,25)      {\tiny $y$}
4492:   \put(87,4)       {\tiny $x$}
4493:   \put(116,70)     {\scriptsize $\bar\imath$}
4494:   \put(211,70)     {\scriptsize $i$}
4495:   \put(242,-70)    {\scriptsize $\bar\imath$}
4496:   \put(338,-70)    {\scriptsize $i$}
4497:   \put(274,159)    {\footnotesize $B^3$}
4498:   \put(452,6)      {\footnotesize $B^3$}
4499:   \put(0,80)       {$ B_2 \;= $}
4500:               \end{picture}
4501:   \nonumber\\  \begin{picture}(420,150)(0,0)
4502:   \put(190,0)  {\Includeourbeautifulpicture 52b}
4503:   \put(221.5,132)  {\tiny $y$}
4504:   \put(260,152.5)  {\tiny $x$}
4505:   \put(160,72)     {$ = $}
4506:   \end{picture}
4507:   \nonumber\\[-2.2em] \label{eq:xcap-aux-1}
4508:   \end{eqnarray} \xlabp 52
4509: as the cobordism describing the basis element in the space of blocks.
4510: The structure constant appearing in
4511:   \be
4512:   Z(\MX , \emptyset , \hat \X) = c(\vhi) \, Z(B_2,\emptyset,\hat \X) 
4513:   \labl{eq:xcap-1bulk-expand}
4514: is again found by composing $Z(\MX,\emptyset,\hat \X)$ with the cobordism dual 
4515: to \erf{eq:xcap-aux-1}. After a suitable rotation of the 
4516: three-manifold, this results in
4517:   %  [ pic 4_109 ]
4518:   \bea \begin{picture}(420,210)(10,40)
4519:   \put(50,0)    {\INcludeourbeautifulpicture 109a}
4520:   \put(282,0)   {\INcludeourbeautifulpicture 109b}
4521:   \put(78,154)     {\scriptsize $\bar\imath$}
4522:   \put(164,159)    {\scriptsize $i$}
4523:   \put(187,118)    {\begin{turn}{90}\small $\phi$\end{turn}}
4524:   \put(146,123)    {\scriptsize $A$}
4525:   \put(165,85)     {\scriptsize $\bar\imath$}
4526:   \put(202.3,73.3) {\tiny $\sigma$}
4527:   \put(206,46)     {\scriptsize $A$}
4528:   \put(238.1,194)  {\footnotesize $\mathbb{RP}^3$}
4529:   \put(317,70)     {\scriptsize $\bar\imath$}
4530:   \put(367,165)    {\scriptsize $i$}
4531:   \put(398,133.4)  {\small $\phi$}
4532:   \put(339,49)     {\scriptsize $A$}
4533:   \put(389,109)    {\scriptsize $A$}
4534:   \put(403,146)    {\scriptsize $A$}
4535:   \put(417,90)     {\scriptsize $\bar\imath$}
4536:   \put(417,222)    {\scriptsize $\bar\imath$}
4537:   \put(367,96.5)   {\tiny $\sigma$}
4538:   \put(437,45)     {\scriptsize $A$}
4539:   \put(454.4,246.5){\footnotesize $\mathbb{RP}^3$} 
4540:   \put(-3,127)     {$\dsty c(\vhi) \;=\; \frac{1}{S_{0,0}} $}
4541:   \put(262,127)    {$ =~ \dsty \frac{1}{S_{0,0}} $}
4542:   \epicture18 \labl{eq:xcap-cobord}  \Xlabp 109
4543: 
4544: In the second equality, after deforming the ribbon graph, we use that $\phi$
4545: is a morphism of bimodules, $\phi\iN\HomAA(U_i\otiP A\otim U_j,A)$.
4546: Comparing to (II:3.91), we conclude that the resulting ribbon graph is just
4547:   \be
4548:   c(\vhi) = \Gamma^\sigma(U_{\bar\imath},\phi') \,/\, S_{0,0} \,,
4549:   \labl{chatphic}
4550: where $\phi'$ is the morphism inside the dashed box on the \rhs\ of 
4551: \erf{eq:xcap-cobord}. $\Gamma^\sigma$ is the ribbon invariant that
4552: was introduced in (II:3.90), which was computed in a basis in (II:3.110).  
4553: 
4554: \medskip
4555: 
4556: One-point functions on a cross cap have been first obtained for free
4557: bosons, fermions and ghosts in \cite{clny2,poca} and for the Ising model in
4558: \cite{fips}. The cross cap Ishibashi states were discussed in \cite{ishi}. 
4559: A set of cross cap one-point functions for the Cardy case in rational CFTs 
4560: was given in \cite{prss,fffs3}, and for the D-series of su(2) in \cite{prss2}. 
4561: Simple current techniques allow one to find more solutions in the Cardy case, as well 
4562: as for more general simple current invariants \cite{huss,huss2,husc,fhssw,brho}. 
4563: In particular, Gepner (and related) models have recently received a lot of 
4564: attention, see e.g.\ \cite{bhhw,blwe,dihs2}.
4565: 
4566: 
4567: \subsection{One defect field on the cross cap} \label{sec:rc-1defect-xcap}
4568: 
4569: As already noted in section \ref{sec:rc-3defect}, replacing bulk fields by 
4570: defect fields changes very little in the calculation of a correlator. Thus we
4571: discuss the correlator of one defect field on the cross cap only very briefly.
4572: The geometric setup for the correlator is 
4573:   %  [ pic 4_112 ]
4574:   \bea \begin{picture}(180,166)(0,0)
4575:   \put(0,0)   {\Includeourbeautifulpicture 112}
4576:   \put(205,82.4)   {\tiny $x$}
4577:   \put(184.4,104.5){\tiny $y$} 
4578:   \put(205.6,46)   {\tiny $x$}
4579:   \put(184.4,68)   {\tiny $y$} 
4580:   \put(186,26.4)   {\tiny $x$} 
4581:   \put(164.6,5)    {\tiny $y$}
4582:   \put(155.6,72.7) {\small $x$}
4583:   \put(71.9,158.6) {\small $y$}
4584:   \put(59,121)     {\small $X$} 
4585:   \put(151,89.3)   {\tiny $\orr_2(X)=$}
4586:   \put(151,53)     {\tiny $\orr_2(p)\,=$}
4587:   \put(132,18)     {\tiny $\orr_2(X)=$} 
4588:   \epicture-9 \Labp 112
4589: where the defect field is given by 
4590: $\Thetha \eq (X,-\orr_2,X,\orr_2,i,j,\vartheta,p,[\gamma],\orr_2)$ 
4591: with $\gamma(t) \eq p{+}(t,0)$ and $\vartheta\iN\HomAA(U_i\otip X^s\otim U_j,X)$.
4592: The choice of $\gamma$ implies that at the insertion point of the defect field
4593: the defect runs parallel to the real axis. The orientations $\orr_2(p)$ around 
4594: the insertion point and $\orr_2(X)$ around the defect line are chosen as 
4595: indicated.  When giving $\orr_2(X)$ one must take the orientation-reversing 
4596: identification of points on $\reals^2$ into account.
4597: 
4598: The connecting manifold $\M_X$ can be found by the same method that led to 
4599: \erf{eq:xcap-ribbon-emb}. The only difference is that the $A$-ribbon gets 
4600: replaced by an $X$-ribbon, and that there is a half-twist in accordance with 
4601: \erf{eq:pic41}. The resulting ribbon graph is, in blackboard framing,
4602:   %  [ pic 4_113 ]
4603:   \bea \begin{picture}(370,224)(0,40)
4604:   \put(70,0)    {\Includeourbeautifulpicture 113}
4605:   \put(92.6,102)   {\tiny $x$}
4606:   \put(84,124)     {\tiny $y$}
4607:   \put(73.6,128.2) {\tiny $z$}
4608:   \put(106,6)      {\tiny $x$}
4609:   \put(92,33)      {\tiny $y$}
4610:   \put(322,93)     {\ $t\eq{-}1$}
4611:   \put(322,256)    {\ $t\eq0$}
4612:   \put(216,105)    {\scriptsize $i$}
4613:   \put(129,105)    {\scriptsize $j$}
4614:   \put(216,148)    {\scriptsize $j$}
4615:   \put(174,133)    {\scriptsize $X$}
4616:   \put(296,131)    {\scriptsize $X$}
4617:   \put(239.5,132)  {\begin{turn}{270}\small $\vartheta$\end{turn}}
4618:   \put(0,127)   {$ \M_X \;= $}
4619:   \epicture26 \labl{eq:xcap-def-rib}  \Xlabp 113
4620: Composing with the dual of the cobordism \erf{eq:xcap-aux-1}
4621: one determines the constant $c(X,\Thetha)$ appearing in the relation 
4622: $Z(\MX,\emptyset,\hat\X)\eq c(X,\Thetha)\,Z(B_2,\emptyset,\hat \X)$ to be
4623:   %  [ pic 4_114 ]
4624:   \bea \begin{picture}(300,170)(0,36)
4625:   \put(70,0)    {\Includeourbeautifulpicture 114}
4626:   \put(147,182)    {\scriptsize $i$}
4627:   \put(256,115)    {\scriptsize $X$}
4628:   \put(150,105)    {\scriptsize $X^s_{}$}
4629:   \put(115,182)    {\scriptsize $\bar\imath$}
4630:   \put(204,105)    {\scriptsize $\bar\imath$}
4631:   \put(193,139)    {\small $\vartheta$}
4632:   \put(277.7,193.4){\footnotesize $\mathbb{RP}^3$}
4633:   \put(-15,112)    {$\dsty c(X,\Thetha) ~=~ \frac{1}{S_{0,0}} $}
4634:   \epicture11 \labl{eq:xcap-def-inv}  \Xlabp 114
4635: 
4636: \medskip
4637: 
4638: To work out the invariant \erf{eq:xcap-def-inv} we need to introduce 
4639: a bit more notation. Note that the morphism $\vartheta$ entering the data for 
4640: the defect field $\Thetha$ is an element of 
4641: $\HomAA(U_i\,{\otimes^+}\,X^s\,{\otimes^-}\,U_{\bar\imath},X)$.
4642: If we choose $X$ to be the simple bimodule $X_\nu$, then
4643: also $X^s \,{\equiv}\, (X_\nu)^s$ is a simple bimodule. Define a map
4644:   \be
4645:   s:\quad \KK \longrightarrow \KK
4646:   \ee
4647: via $(X_\nu)^s \,{\cong}\, X_{s(\nu)}$. We explicitly choose, once
4648: and for all, a set of isomorphisms
4649:   \be
4650:   \Pi^s_\nu \in \HomAA(X_\nu , (X_{s(\nu)})^s) \,.  \ee
4651: This is similar to the choice of $\pi_i \iN \Hom(U_i,(U_{\bar\imath})^\vee)$ 
4652: made in (I:2.23). For our purposes we need to expand the inverse of 
4653: $\Pi^s_\nu$ in a basis,
4654:   %  [ pic 4_87 ]
4655:   \bea \begin{picture}(200,55)(0,38)
4656:   \put(0,0)    {\Includeourbeautifulpicture 87a}
4657:   \put(190,0)  {\Includeourbeautifulpicture 87b}
4658:   \put(7,91)   {\scriptsize $\dot{X}_{\nu}$}
4659:   \put(7,-9)   {\scriptsize $\dot{X}_{\nu}$}
4660:   \put(0.3,43.4){\tiny $\Bigl(\Pi_{\nu}^s\Bigr)^{\hspace{-0.1cm}-1\!}$}
4661:   \put(189,91) {\scriptsize $\dot X_{\nu}$}
4662:   \put(189,-9) {\scriptsize $\dot X_{\nu}$}
4663:   \put(199,60) {\scriptsize $\beta$}
4664:   \put(198,23) {\scriptsize $\alpha$}
4665:   \put(196.4,41) {\scriptsize $x$}
4666:   \put(50,43)  {$ =~\dsty\sum_{x\in\II}
4667:                    \sum_{\alpha,\beta=1}^{\langle U_x,\dot X_\nu \rangle}
4668:                    \big[ (\Pi^s_\nu)^{-1} \big]^x_{\alpha\beta} $}
4669:   \epicture24 \labl{eq:Pis-expansion} \xlabp 87
4670: In writing this relation we used that as objects in $\calc$, $X_\nu$ and
4671: $(X_{s(\nu)})^s$ are both equal to $\dot X_\nu$. They only differ
4672: in the left and right actions of $A$.
4673: 
4674: After these preliminaries, let us evaluate the invariant \erf{eq:xcap-def-inv}.
4675: We make again the simplifying assumption that $\N ijk\iN\{0,1\}$.  Let us choose
4676:   \be
4677:   X = X_\nu
4678:   \qquad {\rm and} \qquad
4679:   \vartheta = \xi^\alpha_{(i s(\nu) \bar\imath) \nu}
4680:   \cir \big(\id_{U_i}\Oti\,(\Pi^s_{s(\nu)})^{-1}\Oti\, \id_{U_{\bar\imath}}\big)
4681:   \,. \ee
4682: Substituting this into \erf{eq:xcap-def-inv} results in
4683:   %  [ pic 4_88 ]
4684:   \begin{eqnarray} \begin{picture}(410,220)(0,0)
4685:   \put(75,0)    {\includeourbeautifulpicture 88}
4686:   \put(-5,115)      {$\dsty c(X,\Thetha) ~=~ \frac{1}{S_{0,0}} $}
4687:   \put(152,205)    {\scriptsize $i$}
4688:   \put(200,137)    {\scriptsize $X_{\!s(\nu)}$}
4689:   \put(262,169)    {\scriptsize $X_{\nu}$}
4690:   \put(147,107)    {\scriptsize $(X_{\nu})^s_{}$}
4691:   \put(121,205)    {\scriptsize $\bar\imath$}
4692:   \put(221,115)    {\scriptsize $\bar\imath$}
4693:   \put(210.5,162.9){\scriptsize $\alpha$}
4694:   \put(190.4,124)  {\tiny $\Pi^{s~-1}_{\!s(\nu)}$}
4695:   \put(282.7,217)  {\footnotesize $\mathbb{RP}^3$}
4696:   \end{picture} \nonumber \\[.4em] 
4697:   =\; \dsty \frac{1}{S_{0,0}} \sum_{x,y} \sum_{\gamma_1,\gamma_2}
4698:            [\xi^\alpha_{(i s(\nu) \bar\imath) \nu}]^{xyx}_{\gamma_1 \gamma_2}\,
4699:            [(\Pi^s_{s(\nu)})^{-1}]^x_{\gamma_2 \gamma_1} I(i,x,y) \,. ~~~
4700:   \label{eq:1D-Xcap-basis}
4701:   \end{eqnarray}  \xlabp 88
4702: In the second equality we have replaced $\xi^\alpha_{(i s(\nu) 
4703: \bar\imath) \nu}$ and $(\Pi^s_{s(\nu)})^{-1}$ by their expansions 
4704: \erf{eq:xi-expansion} and \erf{eq:Pis-expansion}, respectively. If one
4705: in addition drags the points where the $U_{\bar\imath}$-ribbon touches the
4706: identification plane along the paths indicated\,\footnote{%
4707:   ~When moving the $U_{\bar\imath}$-ribbon one must
4708:   carefully take into account the identification. As discussed
4709:   at the end of section II:3.5 this introduces additional twists.}
4710: one obtains the following representation of the invariant $I(i,x,y)$:
4711:   %  [ pic 4_89 ]
4712:   \bea \begin{picture}(385,150)(0,75)
4713:   \put(85,0)    {\includeourbeautifulpicture 89}
4714:   \put(94,138)     {\scriptsize $\bar\imath$}
4715:   \put(148,133)    {\scriptsize $i$}
4716:   \put(227,58)     {\scriptsize $\bar\imath$}
4717:   \put(177,54)     {\scriptsize $x$}
4718:   \put(203.8,134)  {\scriptsize $y$}
4719:   \put(281.8,152)  {\scriptsize $\bar\imath$}
4720:   \put(326,152)    {\scriptsize $x$}
4721:   \put(362.7,217)  {\footnotesize $\mathbb{RP}^3$}
4722:   \put(0,123)      {$ I(i,x,y) \;= $}
4723:   \epicture44 \labp 89
4724: To compute $I(i,x,y)$, we apply the relation (II:3.73) to the morphism that is
4725: indicated by the dashed box. Thereby $I(i,x,y)$, which according to \erp 89 is
4726: the invariant of a graph in $\RP$, is expressed as the invariant of a graph in 
4727: $S^3$. In the latter form, $I(i,x,y)$ is straightforward to evaluate; we find
4728:   \be
4729:   I(i,x,y) = S_{0,0}\, \kappa^{-1} \sum_{j\in\II} S_{0,j}\, \theta_j^{\,-2}
4730:   \,\frac{\theta_x}{\theta_y} \, \frac{1}{\Rs{}iyx} \, 
4731:   \Gs yi{\bar\imath}yx0  s_{j,y}
4732:   = P_{0,y} \, \frac{ \theta_x\, \Gs yi{\bar\imath}yx0}{t_y\, \Rs{}iyx} \,.
4733:   \ee
4734: In the second step we made use of the definition of the $P$-matrix 
4735: \cite{bedu2,bisa3}, see (II:3.122), and used the notations
4736:   \be
4737:   P_{i,j} = \kappa^{-1} \hat P_{i,j}\,,\quad
4738:   \hat P_{i,j} = (\hat T^{1/2}  S \hat T^2 S \hat T^{1/2})_{ij}\,,\quad
4739:   \hat T_{ij} = \theta_i^{\,-1} \delta_{ij}\,,\quad
4740:   \hat T_{ij}^{\,1/2} = t_i^{\,-1} \delta_{ij}\,;
4741:   \labl{eq:P-def}
4742: finally, $\kappa \eq S_{0,0} \sum_k\! \theta_k^{\,-1} (\dim\,U_k)^2$
4743: is the charge of the modular category $\calc$. If $\calc$ is the
4744: representation category of a RVOA, it is given by
4745: $\kappa \eq \eE^{\pi \ii c/4}$, with $c$ the central charge of the chiral CFT.
4746: 
4747: 
4748: \subsection{Example: The Cardy case} \label{sec:rc-Cardy}
4749: 
4750: In this section we apply the TFT formalism to the simplest situation,
4751: i.e.\ to the algebra $A\eq\one$. This case has already been treated\,\footnote{%
4752:   ~In \cite{fffs2,fffs3} the structure constants were derived for a specific 
4753:   normalisation of the bulk and boundary fields. We find it helpful to 
4754:   reproduce the results with normalisations kept arbitrary. Also, structure 
4755:   constants of defect fields where not considered in \cite{fffs2,fffs3}.
4756:   }
4757: in \cite{fffs2,fffs3}. Furthermore we make again the
4758: simplifying assumption that $\N ijk\iN\{0,1\}$.
4759: 
4760: With the exception of the structure constants for three defect
4761: fields on the sphere \erf{eq:3def-basis-Cardy}
4762: and one defect field on the cross cap
4763: \erf{eq:1def-Xcap-bas-Cardy}, the
4764: results of this section are not new and merely serve to illustrate
4765: the general formalism developed above. References to previous
4766: results in the literature are listed at the end of sections
4767: \ref{sec:rc-bnd3pt}\,--\,\ref{sec:rc-1bulk-xcap} above.
4768:     
4769: 
4770: \subsubsection*{Field content}
4771: 
4772: As is immediate from (I:5.30) and (I:5.19), for $A\eq\one$ the torus
4773: partition function is given by the charge conjugation modular invariant, 
4774: $Z(\one)_{ij} \eq \delta_{i,\bar\jmath}$. Thus there is a bulk field $\vhi_i$ 
4775: for each simple object. In the notation of section \ref{sec:fi-bulk}, we choose 
4776:   \be
4777:   \vhi_i := \big(i,\bar\imath, \phi_i, p, [\gamma], \orr_2(p) \big)\,,
4778:   \ee
4779: where the morphism $\phi_i\eq\xi^{\bar\imath}_{(i0\bar\imath)0}$ 
4780: is given by the basis element introduced in \erf{eq:cardy-bases}.
4781: 
4782: The boundary conditions are in one-to-one correspondence with the simple 
4783: objects $U_i$, too, and the annulus coefficients are given by the fusion rules 
4784: \cite{card9}. Thus we have $\JJ\eq\II$ and $\Ann{i\mu}{\nu} \eq \N \mu i\nu$, 
4785: compare remark I:5.21(i). Nonetheless we continue use greek letters to label 
4786: boundary conditions. Let us denote the field that changes a boundary condition 
4787: $\nu$ to $\mu$ and is labelled by the simple object $U_k$ by $\vsi^{\mu\nu}_k$. 
4788: Thus in terms of the notation of section \ref{sec:fi-bnd}, we choose
4789:   \be
4790:   \vsi^{\mu\nu}_k := (M_\mu, M_\nu, U_k, \psi, p, [\gamma] ) \,,
4791:   \ee
4792: where the morphism $\psi \equiv \psi_{(\mu k)\nu}$ is the basis element
4793: defined in \erf{eq:cardy-bases}.
4794: 
4795: The defects are in one-to-one correspondence with the simple objects 
4796: $U_i$ as well, so that $\KK\eq\II$. Moreover, the twisted partition 
4797: functions are again expressed in terms of the fusion rules \cite{pezu5};
4798: as follows from the ribbon representation (I:5.151), we have 
4799: $Z^{\mu|\nu}_{kl} \eq \sum_{k\in\II} \N klm \N m\mu\nu$. The number
4800: $Z^{\mu|\nu}_{kl}$ gives the multiplicity of defect fields carrying 
4801: chiral and antichiral labels $U_k$ and $U_l$, respectively, that change a 
4802: $\mu$-defect into
4803: a $\nu$-defect. Let us denote such fields by $\Thetha^{\mu|\nu}_{kl,\alpha}$.
4804: To be specific, in the notation of \ref{sec:fi-defect} we choose
4805:   \be
4806:   \Thetha^{\mu|\nu}_{kl,\alpha} := (X_\mu, \orr_2(p), X_\nu, \orr_2(p), 
4807:   k,l,\xi, p, [\gamma], \orr_2(p))\,,  \ee
4808: where $\xi \equiv \xi^\alpha_{(k\mu l)\nu}$. The multiplicity label $\alpha$ 
4809: takes all values in $\II$ for which $\N\mu l\gamma \N k\gamma\nu \,{\ne}\, 0$.
4810: 
4811: 
4812: \subsubsection*{Three boundary fields}
4813: 
4814: Let us first consider the case of three boundary fields
4815: $\vsi^{\nu\mu}_i$, $\vsi^{\kappa\nu}_j$ and 
4816: $\vsi^{\mu\kappa}_k$ on the disk, changing boundary conditions from $\mu$ 
4817: to $\nu$ to $\kappa$ to $\mu$ as we move along the real axis. Substituting 
4818: the expression \erf{eq:GA-bas-Cardy} for the fusion matrices of $\calca$
4819: into the general expression \erf{eq:3bnd-inv-basis} immediately results in
4820:   \be
4821:   c(M_\mu \vsi^{\nu\mu}_i M_\nu \vsi^{\kappa\nu}_j M_\kappa 
4822:   \vsi^{\mu\kappa}_k M_\mu)
4823:   = \beta^{\nu\mu}_i \beta^{\kappa\nu}_j \beta^{\mu\kappa}_k \,
4824:   \Gs\kappa ji\mu\nu{\bar k}\,\Gs\mu k{\bar k}\mu\kappa 0\, \dim(U_\mu)
4825:   \,.  \labl{eq:3bnd-basis-cardy}
4826: Recall that the constants $\beta^{\nu\mu}_i$ etc.\ give the normalisation
4827: of the boundary fields. 
4828: 
4829: \dt{Remark}
4830: In view of the results in \cite{runk,bppz2,fffs2,fffs3,pezu6}
4831: one might have expected that the $\FF$-matrices appear in
4832: \erf{eq:3bnd-basis-cardy}, rather than their inverses $\GG$.
4833: This is, however, merely an issue of normalisation and of labelling of 
4834: the boundary conditions, and illustrates the fact that the labelling
4835: of boundary conditions by simple objects is not canonical. Indeed,
4836: at the end of this section we will see that by suitably changing
4837: the normalisation of the boundary fields, in addition
4838: to relabelling the boundary conditions as $\mu \,{\mapsto}\, \bar\mu$,
4839: the structure constants \erf{eq:3bnd-basis-cardy} can equivalently
4840: be expressed in terms of $\FF$-matrices.
4841: 
4842: 
4843: \subsubsection*{Three defect fields}
4844: 
4845: Next we consider three defect fields. The structure constant of three bulk 
4846: fields can then be obtained as a special case thereof. We consider a circular 
4847: defect consisting of three segments of defect types $\mu$, $\nu$ and $\kappa$, 
4848: respectively, see \erf{eq:3defect-geom}. The change of defect type is effected
4849: by three defect fields $\Thetha^{\mu|\nu}_{ij,\alpha_1}$, 
4850: $\Thetha^{\nu|\kappa}_{kl,\alpha_2}$ and $\Thetha^{\kappa|\mu}_{mn,\alpha_3}$.
4851: The corresponding structure constant is obtained by substituting
4852: \erf{eq:FAA-bas-Cardy} into \erf{eq:3def-basis}, resulting in
4853:   \bea\displaystyle
4854:   c(X_\mu, \Thetha^{\mu|\nu}_{ij,\alpha_1}, 
4855:   X_\nu , \Thetha^{\nu|\kappa}_{kl,\alpha_2} ,
4856:   X_\kappa, \Thetha^{\kappa|\mu}_{mn,\alpha_3} , X_\mu)
4857:   \\{}\\[-.7em] \displaystyle \hspace*{7em}
4858:   = 
4859:   \alpha^{\mu|\nu}_{ij,\alpha_1} \alpha^{\nu|\kappa}_{kl,\alpha_2} 
4860:   \alpha^{\kappa|\mu}_{mn,\alpha_3}
4861:   \frac{ \Rs-lj{\bar n}\Rs-n{\bar n}0 }{ t_l\,t_j\,t_n }\, 
4862:   \frac{ \dim(U_\mu) }{S_{0,0}}
4863:   \\{}\\[-.7em] \displaystyle \hspace*{7em}\quad \sum_\beta
4864:   \Gs i{\alpha_1}l{\alpha_2}\nu\beta
4865:   \Gs \mu jl\beta{\alpha_1}{\bar n}
4866:   \Fs ki\beta\kappa{\alpha_2}{\bar m}
4867:   \Gs{\bar m}\beta n{\alpha_3}\kappa\mu
4868:   \Gs\mu{\bar n}n\mu\beta 0
4869:   \Fs m{\bar m}\mu\mu{\alpha_3}0 \,.
4870:   \eear\labl{eq:3def-basis-Cardy}
4871: The constants $\alpha^{\mu|\nu}_{ij,\alpha_1}$ etc.\ determine
4872: the normalisation of the defect fields. 
4873: 
4874: 
4875: \subsubsection*{Three bulk fields}
4876: 
4877: The structure constants of three bulk fields $\vhi_i$, $\vhi_j$ and $\vhi_k$
4878: on the sphere are obtained by restricting formula \erf{eq:3def-basis-Cardy} 
4879: to the case $\mu\eq\nu\eq\kappa\eq0$.
4880: The expression then simplifies a lot, leading to
4881:   \be
4882:   c(\vhi_i,\vhi_j,\vhi_k)
4883:   = \alpha_i \, \alpha_j \, \alpha_k \,
4884:   \Rs-{\bar\jmath}{\bar\imath}k \, \frac{t_k}{t_i\,t_j} \,
4885:   \frac{\Gs i{\bar\imath}{\bar\jmath}{\bar\jmath}0k \,
4886:   \Fs jik0{\bar\jmath}{\bar k}} {S_{0,0} \, \dim(U_k)} \,.
4887:   \labl{eq:3bulk-basis-Cardy}
4888: Here $\alpha_i \,{\equiv}\, \alpha^{0|0}_{i \bar\imath, \bar\imath}$
4889: etc.\ give the normalisation of the bulk fields. In deriving
4890: this formula we made use of the relation
4891:   \be
4892:   \Gs p{\bar p}pp00 = \Rs{}{\bar p}p0\, \theta_p\, \dim(U_p)^{-1}\,,
4893:   \labl{eq:G=Rthdim}
4894: which can be verified analogously as the corresponding equation
4895: for $\FF$ in (I:2.45). 
4896: 
4897: The normalisation of the bulk fields can be read off from the two-point 
4898: structure constant, divided by the squared norm of the vacuum (expressed 
4899: as the two-point function of the identity field).
4900: Setting $m\eq 0$ and $j\eq \bar\imath$ in 
4901: \erf{eq:3bulk-basis-Cardy} and using once more \erf{eq:G=Rthdim} gives
4902:   \be
4903:   \frac{c(\vhi_i,\vhi_{\bar\imath})}{c(\one,\one)}
4904:   = \frac{\alpha_i \, \alpha_{\bar\imath}}{\dim(U_i)} \,.  \ee
4905: 
4906: 
4907: \subsubsection*{One bulk and one boundary field}
4908: 
4909: Consider now the correlator for one bulk field $\vhi_i$ and one boundary
4910: field $\vsi^{\mu\mu}_k$ on the disk, with boundary condition labelled by 
4911: $\mu$. To evaluate the general expression \erf{eq:1B1b-basis}, first note 
4912: that the expansion coefficients of the representation morphism are now 
4913: trivial, $\rho^{M \,(m)}_{0,(m)}\eq 1$. Also recall relation
4914: \erf{eq:cardy-bas-rel} for the expansion coefficients of $\xi$
4915: and $\psi$. We then find
4916:   \be
4917:   c(\vhi_i,M_\mu,\vsi^{\mu\mu}_k) = t_i^{\,-1}\,
4918:   \alpha_i\,\beta^{\mu\mu}_k\,
4919:   \GG^{(\mu k\bar k)\mu}_{\mu 0} \, \dim(U_\mu)
4920:   \sum_p \frac{\theta_\mu \theta_i}{\theta_p} \,
4921:   \Fs \mu i{\bar\imath}\mu 0p \Gs \mu i{\bar\imath}\mu p{\bar k} .
4922:   \labl{eq:1B1b-bas-Cardy}
4923: Of particular interest is the correlator of just one bulk field on the disk,
4924: which is obtained from
4925: \erf{eq:1B1b-bas-Cardy} by setting $k\eq1$. Using the identities
4926:   \be
4927:   \Fs\mu i{\bar\imath}\mu 0p\, \Gs\mu i{\bar\imath}\mu p0
4928:   = \frac{\dim(U_p)}{\dim(U_i)\, \dim(U_\mu)}
4929:   \qquad {\rm and} \qquad
4930:   s_{i,\bar\jmath}
4931:   = \sum_{k\in\II} \frac{\theta_i \, \theta_j}{\theta_k}\, \N ijk \dim(U_k)
4932:   \,, \ee
4933: one finds
4934:   \be
4935:   c(\vhi_i,M_\mu) = 
4936:   \frac{1}{t_i}\, \frac{\alpha_i}{\dim(U_i)}\,s_{i,\bar\mu}\,.
4937:   \labl{eq:Cardy-1disc}
4938: The phase $t_i^{\,-1}$ will cancel a corresponding phase in the correlator, 
4939: see equation \erf{eq:1bulk} below. Also note that here one might have 
4940: expected $s_{i,\mu}$ instead of $s_{i,\bar\mu}$. This difference amounts to 
4941: the freedom to choose the labelling of the boundary conditions at will. 
4942: Indeed, our labelling is related to the one leading to $s_{i,\mu}$ 
4943: by $\mu\,{\leftrightarrow}\,\bar\mu$;
4944: we will return to this point at the end of the section.
4945: 
4946: 
4947: \subsubsection*{One defect field on the cross cap}
4948: 
4949: Consider a defect of type $\nu$ on a cross cap, together with
4950: a defect field insertion $\theta^{s(\nu)|\nu}_{i\bar\imath,\alpha}$.
4951: To evaluate \erf{eq:1D-Xcap-basis} we need to make a definite choice for
4952: the isomorphisms $\Pi^s_\nu$. First note that for $A\eq\one$ one has
4953: $s(\nu)\eq\nu$ for all $\nu\iN\KK$. It is convenient to 
4954: choose $\Pi^s_\nu \eq t_\nu \id_{U_\nu}$. One then finds
4955:   \be
4956:   c(X_\nu,\Theta^{\nu|\nu}_{i\bar\imath,\alpha}) = 
4957:   \alpha^{\nu|\nu}_{i\bar\imath,\alpha}\,
4958:   \Rs-\alpha i\nu \frac{t_\nu}{t_i\,t_\alpha}\,
4959:   \Gs\alpha i{\bar\imath}\alpha\nu 0
4960:   \frac{P_{0,\alpha}}{S_{0,0}} \,.
4961:   \labl{eq:1def-Xcap-bas-Cardy}
4962: 
4963: 
4964: \subsubsection*{One bulk field on the cross cap}
4965: 
4966: The structure constants of one bulk field $\vhi_i$ on the cross cap
4967: are obtained by setting $\nu\eq 0$ in \erf{eq:1def-Xcap-bas-Cardy}:
4968:   \be
4969:   c(\vhi_i) = \frac{\alpha_i}{\dim(U_i)}\, 
4970:   \frac{P_{0,\bar\imath}}{S_{0,0}} \,.
4971:   \labl{eq:cardy-1bulk-xcap}
4972: 
4973: 
4974: \subsubsection*{Change of normalisation for boundary fields}
4975: 
4976: Let us now return to the question of expressing the structure constant 
4977: \erf{eq:3bnd-basis-cardy} of three boundary fields in terms of $\FF$-matrices 
4978: instead of their inverses $\GG$. We would like to demonstrate that there 
4979: is a normalisation of the boundary fields in which \erf{eq:3bnd-basis-cardy}
4980: is expressed in terms of $\FF$, but with all boundary labels
4981: $\mu$, $\nu$, $\kappa$ replaced by their conjugates.
4982: 
4983: Introduce numbers $\gamma_{\mu i}^\nu$ by relating two different basis 
4984: vectors of $\Hom(U_\mu \oti U_i, U_\nu)$, according to
4985:   %  [ pic 4_90 ]
4986:   \bea \begin{picture}(265,140)(0,7)
4987:   \put(0,0)   {\Includeourbeautifulpicture 90a}
4988:   \put(125,0)   {\Includeourbeautifulpicture 90b}
4989:   \put(-2.3,-9.2)  {\scriptsize $\mu$}
4990:   \put(51,-9.2)    {\scriptsize $i$}
4991:   \put(25,140)     {\scriptsize $\nu$}
4992:   \put(131,-9.2)   {\scriptsize $\mu$}
4993:   \put(165,-9.2)   {\scriptsize $i$}
4994:   \put(126,86)     {\scriptsize $\bar{\mu}$}
4995:   \put(128,37)     {\small $\pi_{\mu}$}
4996:   \put(253,44)     {\scriptsize $\bar{\nu}$}
4997:   \put(242.5,88)   {\small $\pi_{\nu}^{\!-1}$}
4998:   \put(248.5,143)  {\scriptsize $\nu$}
4999:   \put(80,66)      {$ \dsty  =~ \gamma_{\mu i}^\nu $}
5000:   \epicture07 \labl{eq:G-F-aux1} \xlabp 90
5001: A short computation shows that this implies the relation 
5002:   %  [ pic 4_91 ]
5003:   \bea \begin{picture}(325,140)(0,7)
5004:   \put(0,0)   {\Includeourbeautifulpicture 91a}
5005:   \put(170,0)   {\Includeourbeautifulpicture 91b}
5006:   \put(-1.6,141)   {\scriptsize $\mu$}
5007:   \put(51,140)     {\scriptsize $i$}
5008:   \put(24.5,-8.6)  {\scriptsize $\nu$}
5009:   \put(177,145)    {\scriptsize $\mu$}
5010:   \put(210,144)    {\scriptsize $i$}
5011:   \put(171,52)     {\scriptsize $\bar{\mu}$}
5012:   \put(172,100.3)  {\footnotesize $\pi_{\mu}^{\!-1}$}
5013:   \put(298,97)     {\scriptsize $\bar{\nu}$}
5014:   \put(290,47)     {\small $\pi_{\nu}$}
5015:   \put(293,-8.6)   {\scriptsize $\nu$}
5016:   \put(75,66)      {$ \dsty  =~ \frac{\dim(U_\nu)}{\dim(U_\mu)} \,
5017:                    \frac{1}{\gamma_{\mu i}^\nu}  $}
5018:   \epicture00 \labl{eq:G-F-aux2} \xlabp 91
5019: between the dual basis vectors. Consider now the ribbon representation of
5020: $\Gs\kappa ji\mu\nu{\bar k}$ as in (I:2.40). Take the trace of that 
5021: relation and replace the three basis vectors $\lambda_{(\nu i)\mu}$, 
5022: $\lambda_{(\kappa j)\nu}$ and $\Upsilon^{(\kappa \bar k) \mu}$ by the 
5023: corresponding right hand sides of \erf{eq:G-F-aux1} and \erf{eq:G-F-aux2}.
5024: It is then not difficult to see that the resulting ribbon graph can be 
5025: deformed to equal the trace of the left hand side of
5026: (I:2.37). Altogether we obtain
5027:   \be
5028:   \Gs\kappa ji\mu\nu{\bar k}
5029:   = \frac{ \gamma_{\nu i}^\mu \, \gamma_{\kappa j}^\nu}{
5030:     \gamma_{\kappa \bar k}^\mu} \, 
5031:   \Fs ji{\bar \mu}{\bar\kappa}{\bar\nu}{\bar k}\,.  \ee
5032: Substituting this relation into \erf{eq:3bnd-basis-cardy}
5033: results in the equality
5034:   \be
5035:   c(M_\mu \vsi^{\nu\mu}_i M_\nu \vsi^{\kappa\nu}_j M_\kappa 
5036:   \vsi^{\mu\kappa}_k M_\mu)
5037:   = 
5038:   (\gamma_{\nu i}^\mu \, \beta^{\nu\mu}_i) \,
5039:   (\gamma_{\kappa j}^\nu \, \beta^{\kappa\nu}_j) \,
5040:   (\gamma_{\mu k}^\kappa \, \beta^{\mu\kappa}_k) \,
5041:   \Fs ji{\bar\mu}{\bar\kappa}{\bar\nu}{\bar k}\,
5042:   \Fs k{\bar k}{\bar\mu}{\bar\mu}{\bar\kappa}0\, \dim(U_\mu) \,.
5043:   \ee
5044: We conclude that by changing the normalisation of the boundary fields and 
5045: renaming the boundary conditions as $\mu \,{\to}\, \bar\mu$, one 
5046: arrives at the usual expression of the boundary structure constants in 
5047: terms of $\FF$-matrix elements.
5048: 
5049: 
5050: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5051: 
5052: \sect{Conformal blocks on the complex plane}\label{sec:cb}
5053: 
5054: 
5055: In the study of CFT correlators there are two different aspects,
5056: a complex-analytic and a purely topological one.
5057: The construction of the correlators, as presented in detail in sections 
5058: \ref{sec:mb}\,--\,\ref{sec:rc} for the fundamental correlators \BBB, \Bb, 
5059: \bbb\ and \BX, as well as for the defect correlators \DDD\ and \DX, 
5060: proceeds entirely at the topological level.
5061: Given this construction, one would next like to address the complex-analytic
5062: aspect of the correlators, thereby obtaining them as actual functions of
5063: the insertion points. This aspect is treated in the present section and
5064: in section \ref{sec:cf}. We start in sections \ref{sec:cb-cvo} and 
5065: \ref{sec:cb-spaces} by recalling the concepts of
5066: vertex algebras and bundles of conformal blocks. 
5067: 
5068: The relation between the 3-d TFT formulation of the correlators
5069: and the complex-analytic description relies on a number of assumptions,
5070: thus restricting the class of vertex algebras $\CA$ to
5071: which our construction can be applied. First, we require the category 
5072: ${\Rep}(\CA)$ of representations of $\CA$ to be a modular tensor category. 
5073: Sufficient conditions on $\CA$ which 
5074: ensure this property of ${\Rep}(\CA)$ have been given in \cite{huan21}.
5075: 
5076: Given $\CA$ we can define bundles of conformal blocks for Riemann surfaces 
5077: with marked points.
5078: Our second condition is to demand that these spaces are finite-dimensional, and 
5079: that for surfaces of arbitrary genus the conformal blocks give rise to the same 
5080: monodromy and factorisation data as the 3-d TFT constructed from ${\Rep}(\CA)$.
5081: So far there is no complete treatment of this issue in the literature. However, 
5082: many results have been obtained, see e.g.~\cite{Mose,HUan,FRbe,BAki,LEli,huan14}.  
5083: For the purposes of this paper we do not need a complex-analytic construction
5084: of correlators in its full generality, though, as we only consider particularly 
5085: simple cases, namely the fundamental correlators listed above. For these, at 
5086: most conformal three-point blocks on the Riemann sphere are needed.
5087: 
5088: For the purposes of this paper, we will call the class of vertex 
5089: algebras satisfying these conditions rational.
5090: {}From now on we take $\CA$ to be rational in this sense.
5091: 
5092: Another important issue is the metric dependence of the correlation functions 
5093: through the Weyl anomaly, which arises whenever the Virasoro central charge of 
5094: $\CA$ is nonzero. Here again we do not need a general treatment, because we will
5095: be interested in the ratio of a correlator and of the corresponding correlator
5096: with all field insertions removed (see section \ref{sec:ws-db-cmf} below).
5097: The presence of field insertions does not affect the metric-dependence
5098: of a correlator, and hence the Weyl anomaly cancels in such ratios.
5099: 
5100: 
5101: \subsection{Chiral vertex operators}\label{sec:cb-cvo}
5102: 
5103: In this section we describe the space of conformal
5104: blocks on the Riemann sphere $\mathbb{P}^1$, realised as
5105: $\mathbb{P}^1\eq\complex\,{\cup}\,\{\infty\}$. To this end we collect a 
5106: few notions from the theory of vertex algebras, in particular that
5107: of an intertwiner between representations of vertex algebras.
5108: However, this section neither is intended as a review of vertex algebras, nor 
5109: is it as general as it could be in a vertex algebra setting. Rather, the main
5110: purpose is to specify some notations and conventions needed in the description 
5111: of the CFT correlation functions. For a much more detailed exposition of
5112: vertex algebras the reader may consult e.g.\ \cite{FRbe,LEli,hule3.5}.
5113: 
5114: 
5115: \subsubsection*{Vertex algebras and intertwiners}
5116: 
5117: One formalisation of the physical concept of a chiral algebra for a \cft\ is 
5118: the structure of a conformal vertex algebra $\CA$. This is a vector space 
5119: $\RCA_\Omega$, the space of states, which is $\zet_{\ge 0}$-graded with  
5120: finite-dimensional homogeneous subspaces, equipped 
5121:   with some additional structure. In particular, there is a state-field
5122:   correspondence, assigning to every $W{\in}\, \RCA_\Omega$ a 
5123: %
5124: field $W(z)\eq Y(W;z) \iN \End(\RCA_\Omega)\fose{z,z^{-1}}$.
5125: Here $\End({\rm V})\fose{z}$, for $\rm V$ a vector space, denotes the space 
5126: of formal power series in the indeterminate $z$ with coefficients in 
5127: $\End({\rm V})$. The state space $\RCA_{\Omega}$, also called the vacuum 
5128: module, contains a distinguished vector $v_\Omega$, the vacuum state, 
5129: for which the associated field is the identity, $Y(v_\Omega;z) \eq 
5130: \id_{\RCA_{\Omega}}$, and for which $Y(u;z)v_\Omega \eq u + O(z)$ for all 
5131: $u\iN\RCA_\Omega$. The grading of the state space $\RCA_{\Omega}$ is given by 
5132: the conformal weight. The vacuum $v_\Omega$ is a non-zero vector in the
5133: component of weight zero, which we assume to be one-di\-men\-si\-o\-nal.
5134: The component of weight two contains the Virasoro vector $v_{\rm vir}$,
5135: whose associated field $Y(v_{\rm vir};z) \eq T(z)$ is the
5136: (holomorphic component of the) stress tensor.
5137: 
5138: We will be interested in representations of the vertex algebra $\CA$ and 
5139: intertwiners between them. A representation of $\CA$ of lowest conformal weight
5140: $\Delta$ consists of a $(\Delta{+}\zet_{\ge 0})$-graded vector space $\RCA$ and 
5141: a map $\rho{:}~\RCA_{\Omega} \To \End(\RCA)\fose{z^{\pm1}}$, subject to some 
5142: additional consistency conditions. An intertwining operator between
5143: three modules $\RCA_i$, $\RCA_j$, $\RCA_k$ with 
5144: conformal highest weights $\Delta_{i},\, \Delta_{j},\, \Delta_{k}$ is a map
5145:   \be
5146:   \VO ijk{\cdot\,}z :\quad \RCA_j \longrightarrow
5147:   z^{\Delta_i-\Delta_j-\Delta_k} \Hom(\RCA_k,\RCA_i)\fose{z^{\pm1}} \,,
5148:   \ee
5149: satisfying a condition to be described in a moment.
5150: 
5151: In the sequel we make the simplifying assumption that certain power series 
5152: of our interest possess a non-empty domain of convergence when the indeterminate
5153: $z$ is replaced by a complex variable, again denoted by $z$; 
5154: they can then be regarded as complex-valued functions analytic in that domain.
5155: For instance, suppose we are given some fields $\varphi_i(z) \iN z^{\alpha_i}
5156: \Hom(\RCA_i,\RCA_{i+1})\fose{z^{\pm1 }}$, which may be either fields $W(z)$ in 
5157: the chiral algebra (in which case $\RCA_{i+1}\eq\RCA_i$) or intertwining 
5158: operators $\VO ip{,i{+}1}vz$ for some $\CA$-\rep\ $\RCA_p$ and some 
5159: $v \iN \RCA_p$.
5160: For any two homogeneous vectors $a \iN \RCA_1$ and\,\footnote{~%
5161:   Here $\RCA_i^*$ is the restricted dual of 
5162:   the vector space $\RCA_i$, which by definition is the direct sum 
5163:   of the duals of the (finite-dimensional) homogeneous subspaces of $\RCA_i$,
5164:   i.e.\ $\RCA_i^* \eq \bigoplus_{n=0\!}^{\infty} \RCA_{(n)}^*$ for
5165:   $\RCA_i\eq\bigoplus_{n=0\!}^{\infty} \RCA_{(n)}$.}
5166: $b^* \iN \RCA_{n+1}^*$ we have
5167:   \be
5168:   (b^* , \varphi_n(z_n) \cdots \varphi_1(z_1)\, a)
5169:   \in z_1^{\alpha_1} \cdots z_n^{\alpha_n}\,
5170:   \complex\fose{z_1^{\pm 1},...\,, z_n^{\pm 1}} 
5171:   \labl{bphia}
5172: with suitable numbers $\alpha_i$. It is then assumed that, when taking the 
5173: $z_j$ in the power series \erf{bphia} to be complex variables (and thereby 
5174: interpreting them as field insertion points in the complex plane),
5175: this series converges for all $(z_1,z_2,...\,, z_n)\iN\complex^n$ 
5176: satisfying $|z_n| \,{>}\, \dots \,{>}\, |z_1| \,{>}\, 0$.
5177: 
5178: As another example where we assume convergence, let us formulate the condition 
5179: for $\VO ijk{\cdot\,}z$ to be an intertwining operator.
5180: For $a^* \iN \RCA_i^*$, $b \iN \RCA_j$ and $c\iN\RCA_k$,
5181: consider the three formal power series
5182:   \bea
5183:   f_1(z) = (\,a^* ,\, \rho_i(W(z))\, \VO ijkbw \,c\,)  \, ,\\{}\\[-.8em]
5184:   f_2(z) = (\,a^* ,\, \VO ijk{\rho_j(W(z{-}w))\,b}w \,c\,) \, ,\\{}\\[-.8em]
5185:   f_3(z) = (\,a^* ,\, \VO ijkbw \, \rho_k(W(z)) \,c\,) \,,
5186:   \eear\ee
5187: which are elements of $w^{\Delta_i-\Delta_j-\Delta_k} \complex
5188: \fose{z^{\pm 1}{,} w^{\pm 1}{,} (z{-}w)^{-1}}$. What we assume is that, with 
5189: $z$ and $w$ interpreted as complex variables, each of these three series has a 
5190: non-zero domain of convergence. Usually $f_1(z)$ converges
5191: for $|z|\,{>}\,|w|$, $f_2(z)$ for $|z{-}w|\,{<}\,|w|$, and $f_3(z)$
5192: for $|w|\,{>}\,|z|\,{>}\,0$. Then the condition for
5193: $\VO ijk{\cdot\,}w$ to be an intertwining operator is that the
5194: three functions $f_{1,2,3}(z)$ coincide upon analytic continuation and 
5195: can have poles only at $z\eq0$, $w$ and $\infty$.
5196: 
5197: \medskip
5198: 
5199: 
5200: As already mentioned, we demand that ${\Rep}(\CA)$ is a modular tensor category.
5201: In particular, $\CA$ has only finitely many inequivalent irreducible 
5202: representations,
5203: and every $\CA$-\rep\ is fully reducible. We choose, once and for all, a set
5204:   \be
5205:   \{ \SCA_i \,|\, i \iN \II \}
5206:   \labl{eq:CA-irreps}
5207: of representatives of isomorphism classes of irreducible $\CA$-modules. By 
5208: $\SCA_{\bar k}$ we denote the element in this list that is isomorphic to the 
5209: dual module ${(\SCA_k)}^\vee_{}$. We also take $\SCA_0\eq\RCA_\Omega$.
5210: 
5211: The morphisms of the tensor category $\CA$ are intertwiners of $\CA$-modules. 
5212: In particular, $\Hom_\CA(\SCA_j \oti \SCA_k, \SCA_i)$ is the vector space of 
5213: intertwiners $\VO ijk{\cdot\,}z$ between $\SCA_j$, $\SCA_k$ and $\SCA_i$.
5214: As $\CA$ is rational, the dimension $\N jki$ of this space is finite.
5215: 
5216: 
5217: \subsubsection*{Choice of basis for intertwining operators}
5218: 
5219: We choose bases
5220:   \be
5221:   \{ \VOb ijk{\delta}{\cdot\,}z \,|\, \delta\eq1,...\,,\N jki \}
5222:   \subset \Hom_\CA(\SCA_j \oti \SCA_k, \SCA_i) \ee
5223: in the spaces of intertwiners $\VO ijk{\cdot\,}z$. We do not restrict this 
5224: choice, except when $\SCA_j$ or $\SCA_k$ is the vacuum module. Take first 
5225: $j\eq0$; then $\N jki\eq\delta_{ki}$. Thus $\VO i0i{\,\cdot\,}z{:}~
5226: \SCA_0 \,{\to}\,\End(\SCA_i)\fose{z^{\pm1}}$ is unique up to multiplication 
5227: by a scalar. As a basis, we take the \rep\ morphism itself, ${\mathrm V}^i_{0i}
5228: \eq\rho_i$. Then in particular the field corresponding to the vacuum $v_\Omega$ 
5229: is the identity,
5230:   \be
5231:   \VO i0i{v_\Omega}z = \id_{\SCA_i} \,.  \ee
5232: Similarly, for $k\eq0$ we normalise 
5233: $\VO ii0{\,\cdot\,}z{:}~\SCA_i \,{\to}\,\Hom(\SCA_0,\SCA_i)\fose{z^{\pm1}}$
5234: via the condition that 
5235:   \be
5236:   \VO ii0uz v_\Omega = u + O(z)  \ee
5237: for all $u\iN\SCA_i$.
5238: 
5239: 
5240: \subsubsection*{Two- and three-point conformal blocks}
5241: 
5242: Let $v_\Omega^*$ be the unique vector of weight zero in $\RCA_\Omega^*$
5243: satisfying $v_\Omega^*(v_\Omega) \eq 1$. For $X$ a product of
5244: intertwiners, we write
5245: $\langle 0|\, X \,|0\rangle$ for the number $v_\Omega^*(X v_\Omega)$.
5246: 
5247: For products of at most three intertwiners, the functional dependence 
5248: of $\langle 0|\, X \,|0\rangle$ on the insertion points 
5249: is fixed by conformal invariance.
5250: For the case of two intertwiners one finds explicitly
5251:   \bea\displaystyle
5252:   \langle 0|\, \VO 0{\bar k}kv{z_2}\, \VO kk0u{z_1}\, |0\rangle
5253:   = (z_2{-}z_1)^{-2\Delta_k(u)} B_{\bar k k}(v,u)\,,
5254:   \\{}\\[-.7em]\displaystyle  {\rm with} \qquad
5255:   B_{\bar k k}(v,u) := \langle 0|\,\VO 0{\bar k}kv{2}\,\VO kk0u{1}\,|0\rangle
5256:   \,. \eear\labl{eq:2pt-block}
5257: Here we take $u \iN \SCA_k$ and $v \iN \SCA_{\bar k}$ to be 
5258: homogeneous of conformal weight $\Delta_k(u)$ and $\Delta_{\bar k}(v)$,
5259: respectively. The two-point conformal block vanishes unless 
5260: $\Delta_k(u)\eq\Delta_{\bar k}(v)$. 
5261: 
5262: For three intertwiners one gets
5263:   $$ \begin{array}{l} 
5264:   \langle 0|\, \VO 0k{\bar k}w{z_3}
5265:             \, \VO {\bar k,\delta}jiv{z_2}
5266: 	    \, \VO ii0u{z_1} \,|0\rangle
5267:   \\{}\\[-.7em] \quad\; =
5268:   (z_3{-}z_2)^{\Delta_i(u)-\Delta_j(v)-\Delta_k(w)} 
5269:   (z_3{-}z_1)^{\Delta_j(v)-\Delta_i(u)-\Delta_k(w)} 
5270:   (z_2{-}z_1)^{\Delta_k(w)-\Delta_i(u)-\Delta_j(v)} B^\delta_{kji}(w,v,u)\,,
5271:   \eear $$   \be  {\rm with} \qquad 
5272:   B^\delta_{kji}(w,v,u) := 2^{-\Delta_j(v)+\Delta_i(u)+\Delta_k(w)} \,
5273:   \langle 0|\, \VO 0k{\bar k}w{3}\, \VO {\bar k,\delta}jiv{2}\,
5274: 	     \VO ii0u{1} \,|0\rangle \,, \qquad \qquad
5275:   \labl{eq:3pt-block}
5276: where $u\iN\SCA_i$, $v\iN\SCA_j$ and $w\iN\SCA_k$ and, as before,
5277: $\Delta_i(u)$, $\Delta_j(v)$ and $\Delta_k(w)$ are the conformal weights of 
5278: the respective vectors.
5279: 
5280: 
5281: \subsubsection*{Local coordinate changes}
5282: 
5283: For discussing the relationship to three-dimensional topological field theory 
5284: in the next section, one needs to introduce the group $\AutO$ of local 
5285: coordinate changes or, equivalently, locally invertible function germs. We 
5286: describe $\AutO$ briefly; for details see e.g.\ section 6.3.1 of \cite{FRbe}.
5287: 
5288: The group $\AutO$ can be identified with a subset of the formal power 
5289: series $\complex\fose{z}$ consisting of those $f \iN \complex\fose{z}$ 
5290: that are of the form $f(z)\eq a_1 z + a_2 z^2 + \dots$ with $a_1\,{\ne}\,0$.
5291: Since a $\CA$-module $\RCA$ is in particular a $\Vir$-module,
5292: there is an action $R$ of $(\AutO)_{\rm opp}$ on the module $\RCA$,
5293:   \be
5294:   R\,{:}\quad \AutO \to \End(\RCA) \qquad {\rm such~that} \qquad
5295:   R(\id) = \id \quad{\rm and}\quad R(f\cir g) = R(g) R(f) \,.
5296:   \ee
5297: This action $R$ is constructed as follows. For the coordinate change
5298: $f\iN \AutO$ that is given by $f(z)\eq\sum_{k=1}^\infty a_k z^k$,
5299: set $D \,{:=}\, \sum_{j=1}^\infty\! v_j t^{j+1} \frac\partial{\partial t}$,
5300: with the coefficients $v_j$ determined by solving
5301: $v_0\exp(D)t\eq f(t)$ order by order in $t$. The first few coefficients are
5302:   \be
5303:   v_0 = a_1 \, , \qquad
5304:   v_1 = \frac{a_2}{a_1} \, ,\qquad
5305:   v_2 = \frac{a_3}{a_1} - \Big(\frac{a_2}{a_1}\Big)^2 .
5306:   \ee
5307: The operator $R$ is then given by
5308: $R(f) \,{:=}\, \exp(-\sum_{j=1}^\infty v_j L_j) {v_0}^{\!-L_0}$.
5309: Actually we will later rather need the inverse of $R(f)$, which takes the form
5310:   \be
5311:   R(f)^{-1} = {v_0}^{\!L_0}\, \exp\!\Big(\sum_{j=1}^\infty v_j L_j\Big) \,.
5312:   \labl{eq:Rinv-op}
5313: For example, if $|\Delta\rangle$ is a $\Vir$-highest weight vector of
5314: $L_0$-eigenvalue $\Delta$, then for the first two levels one finds
5315:   \be\bearll
5316:   R(f)^{-1} |\Delta\rangle & = (f')^\Delta\, |\Delta\rangle \qquad{\rm and}
5317:   \\{}\\[-.5em]
5318:   R(f)^{-1} L_{-1} |\Delta\rangle \!\!\!&=  (f')^\Delta\, \big(\,
5319:   f' L_{-1} \, |\Delta\rangle + \Delta f''{/}f'\, |\Delta\rangle \,\big) \,,
5320:   \eear\labl{eq:Rinv-action}
5321: with $f'$ and $f''$ denoting derivatives of $f$ at $z\eq0$,
5322: i.e.\ $f'\eq a_1$ and $f''\eq 2 a_2$.  Define a functional $\kappa_t$ via
5323:   \be
5324:   \kappa_t(f)(z) := f(z{+}t)-f(t) \,.
5325:   \ee
5326: Then for vertex operators we have the identity \cite[section~5.4]{HUan} 
5327:   \be
5328:   Y(v;f(z)) = R(f)^{-1}\, Y(R(\kappa_z(f))v;z) R(f) \,,
5329:   \labl{eq:Y-xfer}
5330: where again $f$ satisfies $f(0)\eq0$ and $f'(0) \neq 0$.
5331: 
5332: 
5333: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5334: 
5335: 
5336: \subsection{Spaces of conformal blocks}\label{sec:cb-spaces}
5337: 
5338: \subsubsection*{Extended Riemann surfaces}
5339: 
5340: Recall from section \ref{sec:fi-conv} that the 3-d TFT assigns to every extended
5341: surface $E$ a vector space $\calh(E)$. On the other hand, for the (chiral) 
5342: 2-d CFT we need surfaces with a complex structure. The appropriate objects
5343: here are {\em extended Riemann surfaces\/}. An extended Riemann surface
5344: $E^c$ is given by the following data:
5345: \def\leftmargini{1em}
5346: \begin{itemize}
5347:     \item[\Nxt]  
5348: A compact Riemann surface without boundary, also denoted by $E^c$.
5349:     \item[\Nxt]  
5350: A finite ordered set of {\em marked points\/}, i.e.\ triples 
5351: $(p_i,[\varphi_i],\RCA_i)$,
5352: where the $p_i \iN E^c$ are mutually distinct points, $[\varphi_i]$ is
5353: a germ of injective holomorphic functions from a small disk
5354: $D_\delta \,{\subset}\, \complex$ around 0 to $E^c$ such that 
5355: $\varphi_i(0)\eq p_i$, and $\RCA_i$ is a module of the vertex algebra $\CA$.
5356:     \item[\Nxt] 
5357: A Lagrangian submodule $\lambda^c \,{\subset}\, H_1(E^c,\zet)$ --
5358: that is, a $\zet$-submodule $\lambda^c$ of $H_1(E^c,\zet)$ such
5359: that the intersection form $\langle \,\cdot\, , \cdot\, \rangle$
5360: vanishes on $\lambda^c$ and that any $x \iN H_1(E^c,\zet)$ obeying
5361: $\langle x , y \rangle \eq 0$ for all $y\iN\lambda^c$ lies in $\lambda^c$.
5362: \end{itemize}
5363: Given extended Riemann surfaces $E^c$ and $\tilde E^c$ with marked
5364: points $(p_i,[\varphi_i],\RCA_i)$ and $(\tilde p_i,[\tilde \varphi_i],\RCA_i)$, 
5365: respectively, an {\em isomorphism\/} $f{:}~ E^c \To \tilde E^c$ of extended 
5366: Riemann surfaces is a holomorphic bijection of the underlying Riemann surfaces 
5367: that is compatible with the (ordered) marked points in the sense that
5368:   \be
5369:   (\tilde p_i,[\tilde \varphi_i],\RCA_i) = (f(p_i),[f{\circ}\varphi_i],\RCA_i)
5370:   \,. \ee
5371: 
5372: 
5373: \subsubsection*{Bundles of conformal blocks}
5374: 
5375: Let $E^c$ be an extended Riemann surface with marked points
5376: $(p_i,[\varphi_i],\RCA_i)$, $i\eq1,2,...\,,n$. Denote by
5377: ${(\RCA_1 \oti \cdots \oti \RCA_n)}^*_{}$ the space of linear functions\,%
5378:   \footnote{~Here ${\otimes}\equiv\otimes_\complex$ denotes the tensor
5379:   product of complex vector spaces, not the tensor product in $\Rep(\CA)$.}
5380: $\RCA_1 \oti \cdots \oti \RCA_n \To \complex$. To an extended Riemann surface 
5381: $E^c$ the chiral CFT assigns the space of conformal blocks on $E^c$, denoted 
5382: by $\calh^c(E^c)$. It is defined as a subspace\,%
5383:   \footnote{~The subspace $\calh^c(E^c)\,{\subset}\,{(\RCA_1\oti\cdots\oti 
5384:   \RCA_n)}^*_{}$ is described via a compatibility condition with the action of 
5385:   $\CA$. For details we refer e.g.\ to
5386:   \cite[section 4]{Scfw} and \cite[chapter 9]{FRbe}.}
5387: of $(\RCA_1 \oti \cdots \oti \RCA_n)^*_{}$. As mentioned at the beginning 
5388: of section \ref{sec:cb} we assume all $\calh^c(E^c)$ to be finite-dimensional.
5389: One can show that if
5390: $E^c$ and $\tilde E^c$ are isomorphic as extended Riemann surfaces, then 
5391: $\calh^c(E^c)\eq\calh^c(\tilde E^c)$. Note that this is an {\em equality\/}
5392: of subspaces of ${(\RCA_1 \oti \cdots \oti \RCA_n)}^*_{}$, not just an 
5393: isomorphism. In other words, there exists a well-defined assignment 
5394: $[E^c] \,{\mapsto}\,\calh^c(E^c)$
5395: on isomorphism classes $[E^c]$ of extended Riemann surfaces.
5396: 
5397: Denote by $\hat\calm_{g,n}(\RCA_1,\dots,\RCA_n)$ the moduli space of extended 
5398: Riemann surfaces of genus $g$ with $n$ marked points, such that the $k$th 
5399: point is labelled by the $\CA$-module $\RCA_k$. 
5400: If the choice of representations is obvious from the context, or not important
5401: to the argument, we will write $\hat\calm_{g,n}$ in place of
5402: $\hat\calm_{g,n}(\RCA_1,\dots,\RCA_n)$. The same convention will be applied to 
5403: related objects defined below.
5404: Consider the trivial vector bundle
5405:   \be
5406:   \hat\calm_{g,n}(\RCA_1,\dots,\RCA_n)\times {(\RCA_1\oti\cdots\oti\RCA_n)}^*_{}
5407:   \rightarrow \hat\calm_{g,n}(\RCA_1,\dots,\RCA_n) \,.
5408:   \labl{eq:cb-triv-bundle}
5409: Each point in $\hat\calm_{g,n}$ is an isomorphism class
5410: $[E^c]$ of extended Riemann surfaces, to which we can assign the subspace
5411: $\calh^c(E^c) \,{\subset}\,{(\RCA_1 \oti \cdots \oti \RCA_n)}^*_{}$. This 
5412: defines a subbundle of \erf{eq:cb-triv-bundle}, the {\em bundle of conformal 
5413: blocks\/} on $\hat\calm_{g,n}(\RCA_1,\dots,\RCA_n)$; we denote it by
5414:   \be
5415:   \cbs_{g,n}(\RCA_1,\dots,\RCA_n)
5416:   \rightarrow \hat\calm_{g,n}(\RCA_1,\dots,\RCA_n) \,.
5417:   \ee
5418: 
5419: \medskip
5420: 
5421: The bundle $\cbs_{g,n}$ comes equipped with a projectively flat connection, 
5422: the \KZ connection, see \cite{frsh2} and e.g.\ \cite[chapter~17]{FRbe}.
5423: Via parallel transport, this connection allows us to assign an
5424: isomorphism $U_\gamma{:}~\calh^c(E^c) \To \calh^c(\tilde E^c)$ to 
5425: every curve $\gamma{:}~[0,1] \To \hat\calm_{g,n}$
5426: with $\gamma(0) \eq [E^c]$ and $\gamma(1) \eq [\tilde E^c]$. 
5427: 
5428: 
5429: \subsubsection*{Two examples of the map $U_\gamma$}
5430: 
5431: We will later use two instances of the map $U_\gamma$, the case that $\gamma$ 
5432: corresponds to a change of local coordinates and the case that $\gamma$ amounts 
5433: to moving the insertion points. 
5434: Restricting to these two directions in $\hat\calm_{g,n}$ -- moving
5435: insertion points and changing local coordinates while the complex
5436: structure of the underlying Riemann surface remains unaltered -- 
5437: results in a flat (not only projectively flat) connection.
5438: $U_\gamma$ then depends only on the homotopy class of $\gamma$.
5439: 
5440: In both examples considered below, $E^c$ denotes an extended 
5441: Riemann surface with marked points $(p_i,[\varphi_i],\RCA_i)$, 
5442: $i\eq1,2,...\,,n$. We will consider curves with $\gamma(0)\eq[E^c]$.
5443: 
5444: \medskip
5445: 
5446: For the first example, define a family $E^c(t)$ of extended Riemann surfaces 
5447: by taking $E^c(t)$ to be given by $E^c$, with the local coordinate at the
5448: $k$th marked point replaced by $[\varphi_k \cir \eta_t]$ 
5449: with $\eta_t{:}~ D_\eps \To D_\eps$ such that 
5450: $\eta_t(0)\eq0$ and $\eta_0 \eq \id$. This defines a curve in $\hat\calm_{g,n}$
5451: by setting $\gamma(t)\,{:=}\,[E^c(t)]$. For this curve the map $U_\gamma$ 
5452: takes the form, for $\beta\iN\calh^c(E^c)$,
5453:   \be
5454:   (U_\gamma \beta)(v_1,...\,,v_n) =
5455:   \beta(v_1,...\,,\hat R(\eta)^{-1}v_k,...\,,v_n) \in \calh^c(\tilde E^c)
5456:   \,, \labl{eq:U-coord-change}
5457: where $\tilde E^c \eq E^c(1)$. Note that we cannot directly take the map
5458: $R(\eta_1)^{-1}$ as defined in \erf{eq:Rinv-op} because if $\RCA_k$ 
5459: has non-integral conformal weight, then the expression $(\eta_1')^{\Delta_k}$ 
5460: in \erf{eq:Rinv-action} is only defined up to a phase. Instead, we set
5461:   \be
5462:   \hat R(\eta)^{-1} := \hat r(1) \,,
5463:   \ee
5464: where
5465:   \be \begin{array}{rcl}
5466:   \hat r :\quad [0,1] &\longrightarrow& \End(\RCA_k)\\[2pt]
5467:   t &\longmapsto& R(\eta_t)^{-1} \,,
5468:   \eear\labl{eq:U-coord-Rhat-def}
5469: and where now $R(\eta_t)^{-1}$ is the map defined in \erf{eq:Rinv-op}, with
5470: the phase ambiguity resolved by demanding $\hat r$ to be continuous and to obey
5471: $\hat r(0) \eq \id$. 
5472: 
5473: \medskip
5474: 
5475: For the second example consider the curves $\gamma_\delta(t) \,{:=}\, [E^c(t)]$
5476: with $\delta\iN\complex$. Here $E^c(t)$ is again equal to $E^c$, but with the 
5477: $k$th marked point replaced this time by $(p^t_k,[\varphi_k\cir\eta_t],\RCA_k)$, 
5478: where $\eta_t(\zeta)\eq\zeta\,{+}\,t\delta$ and $p_k^t\eq\varphi_k(\eta_t(0))$. 
5479: If $\varphi_k$ is defined on $D_\eps$, then $\varphi_k \cir \eta_t$ is a
5480: well-defined local coordinate around $p^t_k$ as long as $\delta\,{<}\,\eps$.
5481: With these definitions, in $\tilde E^c \eq E^c(1)$ the point $p_k\eq\varphi_k(0)$
5482: has been moved to $\tilde p_k\eq\varphi_k(\delta)$. The map $U_{\gamma_\delta}$
5483: must obey, for each $\beta \iN \calh^c(E^c)$,
5484:   \be
5485:   \frac{\rm d}{{\rm d}\delta} (U_{\gamma_\delta} \beta)(v_1,...\,,v_n)\big|
5486:   _{\delta=0} = \beta(v_1,...\,,L_{-1}v_k,...\,,v_n) \,.
5487:   \labl{eq:KZ-shift-coord}
5488: 
5489: 
5490: \subsubsection*{Renumbering marked points}
5491: 
5492: Given an extended Riemann surface $E^c$ with marked points
5493: $(p_i,[\varphi_i],\RCA_i)$, $i\eq1,2,...\,,n$, and a permutation
5494: $\pi$ of $\{1,2,...\,,n\}$, one can define a new extended Riemann
5495: surface $\tilde E^c \eq P_\pi(E^c)$ by taking the same underlying
5496: Riemann surface, but now ordering the marked points as
5497: $(\tilde p_i,[\tilde \varphi_i],\tilde \RCA_i) \eq 
5498: (p_{\pi^{-1}(i)},[\varphi_{\pi^{-1}(i)}],\RCA_{\pi^{-1}(i)})$.
5499: 
5500: In the same spirit, one can define an isomorphism
5501: $\calp_\pi {:}~ \calh^c(E^c) \To \calh^c(P_\pi(E^c))$ of conformal block
5502: spaces via $(\calp_\pi \beta)(\tilde v_1,...\,,\tilde v_n)
5503: \,{:=}\, \beta(\tilde v_{\pi(1)},...\,,\tilde v_{\pi(n)})$,
5504: where $\tilde v_k \iN \tilde\RCA_k \eq \RCA_{\pi^{-1}(k)}$. This extends
5505: to a bundle isomorphism, which we will also denote by $\calp_\pi$,
5506:   \be
5507:   \calp_\pi :\quad \cbs_{g,n}(\RCA_1,...\,,\RCA_n)
5508:   \longrightarrow \cbs_{g,n}(\RCA_{\pi^{-1}(1)},...\,,\RCA_{\pi^{-1}(n)}) \,.
5509:   \ee
5510: 
5511: 
5512: \subsubsection*{Conformal blocks on the Riemann sphere}
5513: 
5514: Let us describe the bundle $\cbs_{0,n}$ more explicitly. Define the
5515: standard extended Riemann surface $E^c_{0,n}[i_1,i_2,...\,,i_n]$
5516: to be the Riemann sphere, realised as $\complex \,{\cup}\, \{\infty\}$, 
5517: with marked points $\{(k,[\varphi_k],\SCA_{i_k}) \,|\, k\eq1,2,...\,,n \}$,
5518: with $\varphi_k$ given by
5519: $\varphi_k(z)\eq z{+}k$. When it is clear from the context what 
5520: representations occur at the marked points, we use again the shorthand
5521: notation $E^c_{0,n}$ for $E^c_{0,n}[i_1,i_2,...\,,i_n]$.
5522: A basis for $\calh^c(E^c_{0,n})$ is provided by the conformal blocks
5523:   \be
5524:   \beta_{0,n}[i_1,...\,,i_n \,;\, p_1,...\,,p_{n-3} \,;\,
5525:   \delta_1,...\,,\delta_{n-2}] \,\equiv\, \beta_{0,n} \,,
5526:   \ee
5527: given by a product of intertwiners,
5528:   \be
5529:   \beta_{0,n} :\quad
5530:   v_1{\otimes}\cdots{\otimes}v_n \, \mapsto\,
5531:   \langle 0 |\, \VOs^0_{i_n,\overline{i_n}}\big(v_n,n\big)  \cdots
5532:   \VOs^{p_1,\delta_1}_{i_2,i_1}\big(v_2,2\big)\,
5533:   \VOs^{i_1}_{i_1,0}\big(v_1,1\big) \,|0\rangle \,,
5534:   \labl{eq:standard-block}
5535: where $i_1,...\,,p_{n-3} \iN \II$ and 
5536: $\delta_1,...\,,\delta_{n-2}$ are multiplicity labels.
5537: 
5538: The moduli space $\hat\calm_{0,n}$ is connected,
5539: and the \KZ connection on $\hat\calm_{0,n}$ is flat. 
5540: To define the bundle $\cbs_{0,n}$, pick a curve $\gamma$ from 
5541: $[E^c_{0,n}]$ to some $[\tilde E^c] \iN \hat\calm_{0,n}$. Then we have
5542: $\calh^c(\tilde E^c) \eq U_\gamma(\calh^c(E^c_{0,n}))$, as a subspace
5543: of ${(\SCA_{i_1} \oti \cdots \oti \SCA_{i_n})}^*_{}$. 
5544: (Note that while this defines the subspace, it does not provide a canonical
5545: basis, since $U_\gamma$ itself depends on the homotopy class of $\gamma$,
5546: whereas the image of $U_\gamma$ does not.) 
5547: In fact, if $E^c$ is given by $\complex \,{\cup}\, \{\infty\}$ with
5548: marked points $(z_k,[\varphi_k],\SCA_{i_k})$, then a basis of $\calh^c(E^c)$
5549: can be given in terms of products of intertwiners,
5550:   \bea\displaystyle
5551:   \!\!\!\!\beta_{p_1,...\,,p_{n-3} \,;\, \delta_1 \dots \delta_{n-2}} : 
5552:   \quad v_1{\otimes}\cdots{\otimes}v_n 
5553:     \\[3pt]\displaystyle
5554:   \qquad\ \longmapsto\; \langle 0 |\, 
5555:   \VOs^0_{i_n,\overline{i_n}}
5556:     \!\Big(R\big(\kappa_0(\varphi_n)\big)^{-1} v_n \,,\, z_n\Big)\,
5557:   \VOs^{\overline{i_n},\delta_{n-2}}_{i_{n-1},p_{n-3}}
5558:     \!\Big(R\big(\kappa_0(\varphi_{n-1})\big)^{-1} v_{n-1}\,,\,z_{n-1}\Big)\,
5559:   \cdots
5560:   \\{}\\[-.7em]\displaystyle
5561:   \hsp{9.9} \cdots\,
5562:   \VOs^{p_1,\delta_1}_{i_2,i_1}
5563:     \!\Big(R\big(\kappa_0(\varphi_2)\big)^{-1} v_2\,,\,z_2\Big)\,
5564:   \VOs^{i_1}_{i_1,0}
5565:     \!\Big(R\big(\kappa_0(\varphi_1)\big)^{-1} v_1\,,\,z_1\Big)\,
5566:   |0\rangle \,.
5567:   \eear\labl{eq:H0n-conf-block}
5568: The choice of branch cuts that are required when evaluating 
5569: \erf{eq:H0n-conf-block} is related to the path in $\hat\calm_{0,n}(i_1,...\,,
5570: i_n)$ chosen to obtain \erf{eq:H0n-conf-block} from \erf{eq:standard-block}.
5571: 
5572: 
5573: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5574: 
5575: 
5576: \subsection{Relation to TFT state spaces}\label{sec:cb-link}
5577: 
5578: In the three-dimensional topological field theory
5579: given by the Chern\hy Simons path integral, the space of states on a
5580: component of the boundary of a three-manifold
5581: can be thought of as the space of conformal blocks of the corresponding 
5582: WZW model on that same surface \cite{witt27}. One can in fact 
5583: extract the \KZ equation for field insertions
5584: on the complex plane from the Chern\hy Simons path integral \cite{frki2}.
5585:   %  eq. (2.7)
5586: 
5587: In our approach to topological and conformal
5588: field theory the path integral does not play any role. Instead, the space of
5589: states for the 3-d TFT and the space of conformal blocks for the 2-d CFT 
5590: are constructed independently, and it is a separate task to find an isomorphism
5591: between the two spaces. In this section we make such an isomorphism explicit
5592: in the case that the surface is the Riemann sphere.
5593: The TFT that is used in the present and the following section
5594: is the one obtained from the modular tensor category $\calc \eq {\rm Rep}(\CA)$
5595: of representations of the chiral algebra $\CA$.
5596: 
5597: 
5598: \subsubsection*{Extended surfaces from extended Riemann surfaces}
5599: 
5600: To an extended Riemann surface $E^c$ one can assign an extended surface  
5601:   \be
5602:   E = \ff(E^c) 
5603:   \ee
5604: as follows. As a two-manifold, $E$ coincides with $E^c$. Further, a marked point
5605: $(p,[\varphi],\RCA)$ on $E^c$ becomes the marked point $(p,[\gamma],\RCA,+)$ on 
5606: $E$, where the arc-germ $[\gamma]$ is defined by $\gamma(t)\,{:=}\,\varphi(t)$, 
5607: that is, by the real axis of the local coordinate system around $p$. 
5608: The Lagrangian subspace of $H_1(E,\reals)$ is obtained as $\lambda\eq \lambda^c 
5609: \,{\otimes}_{\zet}\, \reals$ from $\lambda^c \,{\subset}\, H_1(E^c,\zet)$.
5610: %
5611: The assignment $\ff$ thus acts on extended Riemann surfaces by forgetting the 
5612: complex structure and by replacing germs of local coordinates by germs of arcs.%
5613:   \,\footnote{~This also constitutes the main reason for using a slightly
5614:   different definition of extended surface as compared to section I:2.4. While
5615:   there is no canonical way to associate an {\em arc\/} to a germ of local
5616:   coordinates, there
5617:   is no problem in assigning a {\em germ of arcs\/} in the way described.}
5618: For instance, recalling the definition of the standard extended surfaces 
5619: $E_{0,n}[i_1,...\,,i_n]$ in section \ref{sec:standbas} and of the standard 
5620: extended Riemann surfaces $E^c_{0,n}[i_1,...\,,i_n]$ above, we see that
5621:   \be
5622:   \ff(E^c_{0,n}) = E_{0,n} \,.  \ee
5623: 
5624: 
5625: \subsubsection*{The topological counterparts of $P_\pi$ and $U_\gamma$} 
5626: 
5627: Recall that for an extended Riemann surface $E^c$,
5628: $P_\pi(E^c)$ describes a permutation of the order of the 
5629: marked points on $E^c$. Since the set of marked points on an 
5630: extended surface $E$ is not ordered, we simply have 
5631:   \be
5632:   \ff \cir P_\pi = \ff \,.  \ee
5633: 
5634: To be able to formulate the analogue of $U_\gamma$ in the two examples 
5635: discussed in section \ref{sec:cb-spaces} above, we need the following concepts.
5636: By a family $E^c(t)$ of extended Riemann surfaces over
5637: a Riemann surface $\Sigma^c$ we mean that each extended surface 
5638: $E^c(t)$ is given by $\Sigma^c$ together with a set of marked points
5639: $(p_k^t, [\varphi_k^t], \RCA_k)$ such that $p_k^t$ and $\varphi_k^t$ depend
5640: smoothly on $t$. A family $E(t)$ of extended surfaces over
5641: a two-manifold $\Sigma$ is defined in the same way. 
5642: If $E^c(t)$ is a family of extended Riemann surfaces over
5643: $\Sigma^c$, then $\ff(E^c(t))$ is a family of extended surfaces
5644: over $\ff(\Sigma^c)$.
5645: 
5646: To a family $E^c(t)$ with $t \iN [0,1]$ we can assign a curve in the
5647: moduli space $\hat\calm_{g,n}$ by setting $\gamma_{E^c}(t) 
5648: \,{:=}\, [E^c(t)]$.  We then also get the isomorphism 
5649: $U_{\gamma_{E^c}^{}}{:}~ \calh^c(E^c(0)) \To \calh^c(E^c(1))$.  
5650: 
5651: On the topological side, from a family $E(t)$ over $\Sigma$, with
5652: $t \iN [0,1]$, we can construct a cobordism $\M_E {:}~ E(0) \To
5653: E(1)$. As a three-manifold, $\M_E$ is given by the cylinder 
5654: $\Sigma \Times [0,1]$. The ribbon graph in $\M_E$ is fixed by demanding
5655: that the ribbons intersect the slice of $\M_E$ at `time' $t$ 
5656: as prescribed by the marked points on $E(t)$. The topological
5657: counterpart of $U_{\gamma_{E^c}^{}}$ is then the linear map
5658: $Z(\M_E) {:}~ \calh(E(0)) \To \calh(E(1))$ that is supplied by the TFT.
5659: 
5660: \medskip
5661: 
5662: In this context the Riemann sphere with marked points is special:
5663: any curve in $\hat\calm_{0,n}$ can be obtained as
5664: $\gamma_{E^c}^{}$ for a family of extended Riemann surfaces over
5665: a single fixed base Riemann surface, which we take to be
5666: $\complex \,{\cup}\, \{ \infty \}$. For surfaces of higher genus one 
5667: needs to vary the complex structure moduli of the underlying Riemann
5668: surfaces as well. {}From here on we restrict our attention to the 
5669: Riemann sphere with marked points.
5670: 
5671: 
5672: \subsubsection*{The isomorphism in the case of the Riemann sphere}
5673: 
5674: Let $[E^c] \iN \hat\calm_{0,n}$. We would like to describe an isomorphism 
5675:   \be
5676:   \Alpha :\quad \calh^c(E^c) \to \calh(\ff(E^c)) 
5677:   \ee
5678: between the space of
5679: conformal blocks associated to the extended Riemann surface $E^c$ and the
5680: space of states of the TFT on the extended surface $\ff(E^c)$.
5681: 
5682: The TFT encodes the monodromy data of the conformal blocks. This must
5683: be reflected in the isomorphism $\Alpha$. More precisely, we require
5684: $\Alpha$ to fulfill 
5685:   \be
5686:   \Alpha \cir \calp_\pi = \Alpha \qquad {\rm and} \qquad
5687:   \Alpha \cir U_{\gamma_{E^c}^{}} = Z(\M_{\ff(E^c)}) \cir \Alpha 
5688:   \labl{eq:Alpha-prop}
5689: for any permutation $\pi$ and any family $E^c(t)$ over $\complex \,{\cup}\, 
5690: \{ \infty \}$. In addition to \erf{eq:Alpha-prop}, the isomorphism $\Alpha$ 
5691: must be consistent with cutting and gluing of extended Riemann surfaces. 
5692: For extended surfaces of higher genus, there are further conditions arising
5693: from varying the complex structure moduli of the underlying Riemann surface. 
5694: We will not analyse the latter two types of conditions in this paper. Instead 
5695: we assume that $\Alpha$ does satisfy these conditions, and
5696: content ourselves with describing $\Alpha$ for
5697: points in $\hat\calm_{0,n}$ and illustrating its consistency with
5698: \erf{eq:Alpha-prop} in some examples.
5699: 
5700: \medskip
5701: 
5702: For the Riemann sphere we fix the isomorphism $\Alpha$ by demanding that the 
5703: standard conformal block $\beta_{0,n} \iN \calh(E^c_{0,n})$ from 
5704: \erf{eq:standard-block} gets mapped to the state described by the cobordism 
5705: $B_{0,n}$ from \erf{eq:standard-cobord},
5706:   \bea
5707:   \Alpha( \beta_{0,n}[i_1,...\,,i_n \,;\, p_1,...\,,p_{n-3} \,;\,
5708:   \delta_1,...\,,\delta_{n-2}])
5709:   \\[7pt] \hsp6
5710:   = Z( B_{0,n}[i_1,...\,,i_n \,;\, p_1,...\,,p_{n-3} \,;\,
5711:   \delta_1,...\,,\delta_{n-2}] , \emptyset, E_{0,n}[i_1,...\,,i_n]
5712:   )\,1 \,. 
5713:   \eear\labl{eq:Alpha-on-seRS}
5714: Since $\hat\calm_{0,n}$ is connected, relation \erf{eq:Alpha-prop} can be used 
5715: to determine $\Alpha$ for all points in $\hat\calm_{0,n}$, starting from its 
5716: value at the point $[E^c_{0,n}]$ as given by \erf{eq:Alpha-on-seRS}.
5717: This way of defining $\Alpha$ is consistent iff \erf{eq:Alpha-prop} holds for
5718: every closed loop in $\hat\calm_{0,n}$. The first property in 
5719: \erf{eq:Alpha-prop} just means that $\Alpha$ is independent of the ordering of 
5720: the marked points of the extended Riemann surface.
5721: The second property in \erf{eq:Alpha-prop} can be reformulated as follows. Two 
5722: representations $\rho^c_{0,n}$ and $\rho_{0,n}^{}$ of $\pi_1(\hat\calm_{0,n})$
5723: on $\calh^c(E_{0,n}^c)$ and on $\calh(E_{0,n})$, respectively, are given by
5724:   \be
5725:   \rho^c_{0,n}([\gamma^c]) :=  U_{\gamma^c}  \qquad {\rm and} \qquad 
5726:   \rho_{0,n}^{}([\gamma^c]) := Z(\M_{\ff(E^c)}) \,,
5727:   \ee
5728: were the $\gamma^c$
5729: are paths starting and ending on $[E^c_{0,n}]$ and $E^c$ is a family over
5730: $\complex \,{\cup}\, \{ \infty \}$ such that $\gamma_{E^c}^{}\eq\gamma^c$.
5731: The isomorphism $\Alpha$ must intertwine these two representations,
5732:   \be
5733:   \Alpha \cir \rho^c_{0,n} = \rho_{0,n}^{} \cir \Alpha \,.
5734:   \labl{eq:Alpha-inter}
5735: Below we will verify the intertwining property in two examples, and sketch 
5736: an argument for it to hold in general at the end of section \ref{sec:cb-braid}.
5737: 
5738: 
5739: \subsubsection*{Example: Closed path resulting in a twist}
5740: 
5741: Define a family of extended Riemann surfaces $E^c(t)$ over 
5742: $\complex \,{\cup}\, \{ \infty \}$ by taking $E^c(t)$ to be equal to 
5743: $E^c_{0,n}[i_1,...\,,i_n]$, except that the $k$th marked point gets replaced by 
5744: $(k,\varphi_k \cir \eta_t, \SCA_{i_k})$ with $\eta_t(z) \eq \eE^{2\pi \ii t} z$.
5745: Using $U$ to compute the transport of a block $\beta_{0,n}$ along the closed 
5746: path $\gamma_{E^c}^{}$ gives
5747:   \be
5748:   (U_{\gamma_{E^c}^{}} \beta_{0,n})(v_1,...\,,v_n)
5749:   = \eE^{2 \pi \ii \Delta_{i_k}} \beta_{0,n}(v_1,...\,,v_n) \,.
5750:   \ee
5751: To see this, start from formula \erf{eq:U-coord-change} and use
5752: \erf{eq:Rinv-action} and \erf{eq:U-coord-Rhat-def} so as to deduce
5753: first that $R(\eta_t)^{-1}v_k \eq \eE^{2\pi \ii t \Delta(v_k)} v_k$ and
5754: then, also using that $\Delta_{i_k} \eq \Delta(v_k) \bmod \zet$,
5755: that $\hat R(\eta)v_k \eq \eE^{2 \pi \ii \Delta_{i_k}} v_k$. We thus find 
5756:   \be
5757:   \rho^c_{0,n}([\gamma_{E^c}]) = \theta_{i_k}^{-1}\, \id_{\calh^c(E^c_{0,n})}\,.
5758:   \ee
5759: Further, let $E(t) \,{:=}\, \ff(E^c(t))$; then the cobordism
5760: $\M_E{:}~ E(0) \To E(1)$ is given by
5761:   %  [ pic 4_115 ]
5762:   \bea \begin{picture}(330,155)(10,40)
5763:   \put(60,0)    {\Includeourbeautifulpicture 115 }
5764:   \put(85,66)     {\tiny $x$}
5765:   \put(78,90)     {\tiny $y$}
5766:   \put(65.3,99.3) {\tiny $z$}
5767:   \put(100,8)     {\tiny $x$}
5768:   \put(87,33)     {\tiny $y$}
5769:   \put(138,86)    {\small $S_{i_{k-1}}$}
5770:   \put(191.4,86)  {\small $S_{i_{k}}$}
5771:   \put(244,86)    {\small $S_{i_{k+1}}$}
5772:   \put(312,183)   { $t\eq0$}
5773:   \put(311,44)    { $t\eq1$}
5774:   \put(0,89)      {$ \M_E \;= $}
5775:   \epicture25 \labl{eq:twist-ex-MU} \Xlabp 115
5776: with the orientations of three-manifold and ribbons as indicated.
5777: Once oriented by the inward pointing normal, the boundary at
5778: $t\eq0$ is given by $-E(0)\eq{-}E_{0,n}$ and the boundary at $t\eq1$
5779: by $E(1) \eq E_{0,n}$, as it should be. By definition of the twist
5780: (see (I:2.9)), the middle ribbon in \erf{eq:twist-ex-MU}
5781: can be replaced by $\theta_{i_k}^{-1}$ times a straight ribbon, so that
5782:   \be
5783:   Z(\M_E) = \theta_{i_k}^{-1} Z( E_{0,n} \Times [0,1] ) \,.
5784:   \ee
5785: Since $Z$ applied to the cylinder over $E_{0,n}$ gives the
5786: identity on $\calh(E_{0,n})$ we obtain
5787:   \be
5788:   \rho_{0,n}([\gamma_{E^c}]) = \theta_{i_k}^{-1} \, \id_{\calh(E_{0,n})} \,.
5789:   \ee
5790: For the closed path considered in this example, the desired identity 
5791: \erf{eq:Alpha-inter} thus even holds independent of $\Alpha$ (and hence provides
5792: a consistency check of the conventions described in section \ref{sec:fi-conv}).
5793: 
5794: 
5795: \subsubsection*{Example: Closed path resulting in a double braiding}
5796: 
5797: In this example we deal again with a family $E^c(t)$ over
5798: $\complex \,{\cup}\, \{ \infty \}$. We consider the situation that each of the 
5799: extended Riemann surfaces $E^c(t)$ has three marked points, 
5800: $(1,\varphi_1(z),\SCA_{i})$, $(p_2^t,\varphi_2^t(z),\SCA_{j})$ and
5801: $(3,\varphi_3(z),\SCA_{k})$, with the functions $\varphi_i$ given by
5802: $\varphi_1(z) \eq z{+}1$, $\varphi_2^t(z) \eq z{+}1{+}\eE^{2 \pi \ii t}$, 
5803: $\varphi_3(z) \eq z{+}3$ and $p_2^t\eq \varphi_2^t(0)$.
5804: Note that $E^c(0)\eq E^c(1)\eq E^c_{0,3}[i,j,k]$. 
5805: A basis $\beta^\delta$ for the conformal blocks in $\calh^c(E^c_{0,3})$ 
5806: is given by \erf{eq:3pt-block} with $z_1\eq1$, $z_2\eq2$, $z_3\eq3$. 
5807: Transporting $\beta^\delta$ along $\gamma_{E^c}^{}$ from $0$ to $t$
5808: results in the conformal block $\beta^\delta_t \iN\calh(E^c(t))$ given by
5809:   \bea\displaystyle
5810:   \beta^\delta_t(u,v,w) = 
5811:   \langle 0 |\, \VO 0k{\bar k}w3\, \VO{\bar k}jiv{1{+}\eE^{2\pi \ii t}}\,
5812:   \VO ii0u1 \,|0\rangle
5813:   \\[8pt]\displaystyle
5814:   \hspace{2.5em} = B^\delta_{kji}(w,v,u) \,
5815:   (2{-}\eE^{2\pi \ii t})^{\Delta(u)-\Delta(v)-\Delta(w)} 
5816:   2^{\Delta(v)-\Delta(u)-\Delta(w)} 
5817:   \eE^{2\pi \ii t(\Delta(w)-\Delta(u)-\Delta(v))} \,.
5818:   \eear\labl{eq:dbraid-ex1}
5819: This block is obtained by setting $z_1\eq1$, $z_2\eq1{+}\eE^{2\pi \ii t}$
5820: and $z_3\eq3$ in \erf{eq:3pt-block}, as follows from \erf{eq:KZ-shift-coord}
5821: together with the property $\VO ijk{L_{-1}v}{z} \eq \frac{\rm d}{{\rm d}z}
5822: \VO ijk{v}{z}$ of the intertwiners. Setting $t\eq1$ we obtain
5823:   \be
5824:   \beta^\delta_1(u,v,w) = 
5825:   \eE^{2\pi \ii (\Delta_k-\Delta_i-\Delta_j)} \, \beta^\delta(u,v,w) \,,
5826:   \ee
5827: where we also used $\Delta(u) \eq \Delta_i \bmod \zet$ etc. Thus
5828:   \be
5829:   \rho^c_{0,3}([\gamma_{E^c}^{}]) = \frac{\theta_i \theta_j}{\theta_k} \,
5830:   \id_{\calh^c(E^c_{0,3})} \,.
5831:   \ee
5832: Regarding the topological side, the cobordism $\M_E$ is easily found to be
5833:   %  [ pic 4_116 ]
5834:   \bea \begin{picture}(375,130)(10,50)
5835:   \put(60,0)    {\Includeourbeautifulpicture 116 }
5836:   \put(90.3,70.5) {\tiny $x$}
5837:   \put(77,96)     {\tiny $y$}
5838:   \put(64,105)    {\tiny $z$}
5839:   \put(325,21)    {\tiny $x$}
5840:   \put(118.5,46)  {\tiny $y$}
5841:   \put(325,148)   {\tiny $x$}
5842:   \put(114,164.5) {\tiny $y$}
5843:   \put(336,164)   { $t\eq0$}
5844:   \put(336,42)    { $t\eq1$}
5845:   \put(157,12)    {\small $1$}
5846:   \put(214,12)    {\small $2$}
5847:   \put(273,12)    {\small $3$}
5848:   \put(157,152)   {\small $1$}
5849:   \put(214,152)   {\small $2$}
5850:   \put(273,152)   {\small $3$}
5851:   \put(110,84)    {\scriptsize $j$}
5852:   \put(162,83)    {\scriptsize $i$}
5853:   \put(277.2,83)  {\scriptsize $k$}
5854:   \put(0,81)      {$ \M_E \;= $}
5855:   \epicture36 \Labp 116
5856: Composing this cobordism with $B_{0,3}$ results in
5857:   %  [ pic 4_117 ]
5858:   \begin{eqnarray} \begin{picture}(420,170)(10,0)
5859:   \put(80,0)    {\Includeourbeautifulpicture 117 }
5860:   \put(0,86)    {$ Z(\M_E) \cir Z(B_{0,3}) ~= $}
5861:   \put(375,86)  {$ =~ \dsty\frac{\theta_i\theta_j}{\theta_k}\,Z(B_{0,3}) \;, $}
5862:   \put(256,81)  {\scriptsize$ i $}
5863:   \put(199.5,81){\scriptsize$ j $} 
5864:   \put(318.4,81){\scriptsize$ k $}
5865:   \put(251,158) {\scriptsize$ \overline{k} $}
5866:   \put(198,13)  {\scriptsize$ 1 $}
5867:   \put(256,13)  {\scriptsize$ 2 $}
5868:   \put(314,13)  {\scriptsize$ 3 $}
5869:   \put(221,138) {\scriptsize$ \delta $} 
5870:   \put(123.2,67){\tiny$ z $}
5871:   \put(138.4,45){\tiny$ y $}
5872:   \put(361.2,21){\tiny$ x $}
5873:   \end{picture} \nonumber \\[-1.6em]{}
5874:   \label{pic4-117}  
5875:   \end{eqnarray} \Xlabp 117
5876: where we applied (I:2.41) twice and then substituted (I:2.43) (or rather the 
5877: corresponding relations for $\RR^{-}$ instead of $\RR$). Thus for $\rho_{0,3}$ 
5878: we get
5879:   \be
5880:   \rho_{0,3}([\gamma_{E^c}^{}]) = \frac{\theta_i \theta_j}{\theta_k} \,
5881:   \id_{\calh(E_{0,3})} \,.
5882:   \ee
5883: This shows that also in this example the identity 
5884: \erf{eq:Alpha-inter} holds independently of $\Alpha$, giving
5885: another consistency check of the conventions in section  \ref{sec:fi-conv}. 
5886: 
5887: 
5888: \subsection{The braid matrices}\label{sec:cb-braid}
5889: 
5890: It is common to express various quantities of interest in terms of
5891: matrices $\Bsmat ijkl{}{} \,{\equiv}\, (\Bsmat ijklpq)$ which are certain
5892: specific combinations of fusing matrices $\FF$ and braiding matrices $\RR$
5893: \cite{frki,Mose}. Conversely, one can recover $\FF$ and $\RR$ from $\BBmat$.
5894: In terms of conformal blocks, the braid matrices $\BBmat$ can be deduced 
5895: from the exchange of two insertion points in a four-point block on the 
5896: sphere. Using the map $\Alpha$ one then verifies that this definition of 
5897: $\BBmat$ is equivalent to the one used in the 3-d TFT context.
5898: 
5899: 
5900: \subsubsection*{Braid matrices in modular tensor categories}
5901: 
5902: Similarly to the definition of the fusing matrices in (I:2.36), for a modular 
5903: tensor category $\calc$ one defines the {\em braid matrices\/} $\Bsmat ijkl
5904: {}{}$ to be the matrices whose entries appear as coefficients in the relation
5905:   %  [ pic 4_05 ]
5906:   \bea \begin{picture}(275,89)(0,32) 
5907:   \put(0,0)     {\Includeourbeautifulpicture 05a }
5908:   \put(200,0)   {\Includeourbeautifulpicture 05b }
5909:   \put(85,50)   {$ \dsty =~\sum_q \sum_{\gamma,\delta}
5910:                  \Bmat{i\,j}klpq\alpha\beta\gamma\delta $}
5911:   \put(-1,-8)   {\scriptsize$ i $}
5912:   \put(28,-8)   {\scriptsize$ j $}
5913:   \put(58,-8)   {\scriptsize$ k $} 
5914:   \put(199,-8)  {\scriptsize$ i $}
5915:   \put(228,-8)  {\scriptsize$ j $}
5916:   \put(258,-8)  {\scriptsize$ k $} 
5917:   \put(248,55)  {\scriptsize$ \delta $}
5918:   \put(226.5,85){\scriptsize$ \gamma $}
5919:   \put(47,55.5) {\scriptsize$ \beta $}
5920:   \put(26,85)   {\scriptsize$ \alpha $} 
5921:   \put(28,71)   {\scriptsize$ p $}
5922:   \put(23,110)  {\scriptsize$ l$}
5923:   \put(230,71)  {\scriptsize$ q $}
5924:   \put(223,110) {\scriptsize$ l$}
5925:   \epicture22 \labl{eq:Bmat-def} \xlabp 05 
5926: If instead the inverse braiding is used one obtains the inverse
5927: braid matrices $\BBmat^{-(ijk)l}_{}$.
5928: 
5929: The braid matrices can be expressed in terms of the fusing
5930: matrices $\FF$ from (I:2.36) and the braid matrices $\RR$ from (I:2.41) as
5931:   \be
5932:   \Bmat{i\,j}klpq\alpha\beta\gamma\delta = \sum_{\mu,\nu}
5933:   \Rm kip\beta\mu\, \F{j\,k}ilpq\alpha\mu\delta\nu\, \R iql\nu\gamma\, .
5934:   \labl{eq:B=RFR}
5935: This can be verified by first applying an inverse $\RR$-move to the morphism 
5936: $\beta\iN\Hom(U_i\oti U_k,U_p)$ on the left hand side of \erf{eq:Bmat-def},
5937: then performing an $\FF$-move, and finally an $\RR$-move applied to the 
5938: resulting morphism in $\Hom(U_q\oti U_i,U_l)$.
5939: 
5940: Setting $k\eq0$ in \erf{eq:Bmat-def} and comparing to (I:2.41),
5941: and recalling also convention (I:2.33), one arrives at the relation
5942:   \be
5943:   \R ijk\alpha\beta = \Bmat{i\,j}0kij\alpha{}\beta{} .
5944:   \labl{eq:R=B}
5945: Note that by convention the couplings in the one-dimensional morphism
5946: spaces involving the tensor unit $U_0\eq\one$ are not spelt out explicitly.
5947: 
5948: Furthermore, using (I:2.42) the relation \erf{eq:B=RFR} is easily inverted:
5949:   \be
5950:   \F{i\,j}klpq\alpha\beta\gamma\delta = \sum_{\mu,\nu}
5951:   \R kjp\beta\mu\, \Bmat{k\,i}jlpq\alpha\mu\nu\gamma\, \Rm qkl\nu\delta .
5952:   \labl{eq:F=RBR}
5953: Thus one can obtain both $\RR$ and $\FF$ from the braid matrices $\BBmat$. One 
5954: may wonder whether it is possible to recover the twists $\theta_k$ and the 
5955: quantum dimensions $\dim(U_k)$ as well. For this to be the case, we need to 
5956: assume that the quantum dimensions are all positive. Then we can take 
5957: the absolute value of (I:2.45) so as to find
5958:   \be
5959:   \dim(U_k) = \big| \Fs k{\bar k}kk00\, \Rs -{\bar k}k0 \big|^{-1} .
5960:   \labl{eq:dimk=FR}
5961: The twist is then directly given by (I:2.45),
5962:   \be
5963:   \theta_k = \dim(U_k)\, \Fs k{\bar k}kk00\, \Rs -{\bar k}k0 .
5964:   \labl{eq:thetak=dimFR}
5965: 
5966: Up to equivalence, a modular tensor category is determined by the
5967: list of isomorphism classes of simple objects $\{ U_k \,|\,k{\in}\II\}$,
5968: the dimensions $\N ijk$ of the morphism spaces
5969: $\Hom(U_i{\otimes}U_j,U_k)$, and the numbers
5970:   \be
5971:   \dim(U_k) \,,\quad
5972:   \theta_k \,,\quad \R ijk\alpha\beta \,,\quad
5973:   \F{i\,j}klpq\alpha\beta\gamma\delta \,.
5974:   \labl{eq:modcat-data}
5975: The considerations above show that, upon the assumption that the
5976: quantum dimensions are positive, instead of \erf{eq:modcat-data} it is 
5977: sufficient to give the numbers
5978:   \be
5979:   \Bmat{i\,j}klpq\alpha\beta\gamma\delta \,.
5980:   \ee
5981: 
5982: 
5983: \subsubsection*{Braid matrices from four-point conformal blocks}
5984: 
5985: The braid matrices also appear upon analytic continuation of the standard 
5986: four-point conformal blocks. In short, one takes the standard four-point 
5987: block and exchanges the field insertions at the points 2 and 3 via analytic 
5988: continuation. The result can be expanded in terms of standard blocks, and 
5989: the coefficients occurring in that expansion are the braid matrices. 
5990: A more precise description is as follows.
5991: 
5992: Define a family $E^c(t)$ of extended Riemann surfaces over $\complex 
5993: \,{\cup}\, \{ \infty \}$ by taking $E^c(t)$ to have field insertions
5994:   \be
5995:   (1,[\varphi_1],\SCA_i) \,,\qquad
5996:   (w_t,[\varphi_2^t],\SCA_j) \,,\qquad
5997:   (z_t,[\varphi_3^t],\SCA_k) \,,\qquad
5998:   (4,[\varphi_4],\SCA_{\bar l}) ) \,,  \ee
5999: where $\varphi_1(\zeta)\eq \zeta{+}1$, $\varphi_4(\zeta)\eq \zeta{+}4$,
6000:   \be
6001:   \varphi_2^t(\zeta) = \zeta - \Tfrac12\, \eE^{-\pi\ii t} + \Tfrac52 \,, \qquad
6002:   \varphi_3^t(\zeta) = \zeta + \Tfrac12\, \eE^{-\pi\ii t} + \Tfrac52 \,,
6003:   \ee
6004: and $w_t \eq \varphi_2^t(0)$, $z_t \eq \varphi_3^t(0)$. It follows in particular 
6005: that $E^c(0) \eq E^c_{0,4}[i,j,k,\bar l]$. For $E^c(1)$ we must also
6006: take into account the ordering of the marked points, resulting in
6007: $P_{(23)}(E^c(1)) \eq E^c_{0,4}[i,k,j,\bar l]$, where $(23)$ denotes
6008: the transposition that exchanges $2$ and $3$. Transporting the conformal block
6009: $\beta^{\mu\nu}[i,j,k,\bar l;p;\mu,\nu] \iN \calh^c(E^c_{0,4})$ along
6010: $\gamma_{E^c}^{}$ from $0$ to $t$ results in the block
6011: $\beta^{\mu\nu}_t \iN \calh^c(E^c(t))$ given by
6012:   \be
6013:   \beta^{\mu\nu}_t(v_1,v_2,v_3,v_4)
6014:   = \langle 0 |\, \VO 0{\bar l}l{v_4}4 \,
6015:   \VO {l,\nu}kp{v_3}{z_t} \, \VO {p,\mu}ji{v_2}{w_t} \,
6016:   \VO {i}i0{v_1}{1} \, |0\rangle  \ee
6017: with $z_t \eq \Tfrac12\, \eE^{-\pi\ii t} + \Tfrac52$ and $w_t \eq{-}\Tfrac12\, 
6018: \eE^{-\pi\ii t} z + \Tfrac52$. Since $\beta^{\mu\nu}_1$ is in the space 
6019: $\calh^c(E^c(1))$ we can apply to it the map $\calp_{(23)}{:}~ 
6020: \calh^c(E^c(1)) \To \calh^c(E^c_{0,4}[i,k,j,\bar l])$ and expand the result
6021: in the standard basis for the latter space,
6022:   \beaa
6023:   (\calp_{(23)} \beta^{\mu\nu}_1)(u_1,u_2,u_3,u_4) \!\!
6024:   &= \beta^{\mu\nu}_1(u_1,u_3,u_2,u_4)
6025:   \\{}\\[-.8em] &\displaystyle
6026:   = \sum_q \sum_{\rho,\sigma} \Bmat{jk}ilpq\nu\mu\rho\sigma\,
6027:   \beta_{0,4}[i,k,j,\bar l;q;\sigma\rho](u_1,u_2,u_3,u_4)\,.
6028:   \eear \labl{eq:Bmat-conf}
6029: By definition of the braiding in the representation category
6030: $\calc \eq {\rm Rep}(\CA)$ of the chiral algebra, the matrices 
6031: $\Bsmat ijkl{}{}$ which appear in \erf{eq:Bmat-conf}
6032: are the same as those defined in \erf{eq:Bmat-def} through a
6033: relation between morphisms in $\calc$, see e.g.\ \cite{huan24} for details.
6034: 
6035: 
6036: \subsubsection*{Compatibility of $\Alpha$ and braiding}
6037: 
6038: Again one obtains a family $E(t)$ of extended surfaces by applying $\ff$ to the 
6039: family $E^c(t)$. For the associated cobordism $\M_E$, composed with a cobordism 
6040: $B_{0,4}$ describing the standard basis of $\calh(E_{0,4})$, one finds
6041:   %  [ pic 4_118 ]
6042:   \begin{eqnarray} \begin{picture}(420,175)(10,0)
6043:   \put(140,0)   {\Includeourbeautifulpicture 118 }
6044:   \put(140,0)     {\setlength{\unitlength}{.38pt}
6045:                   \put(243,262){\scriptsize$ i $} 
6046:                   \put(381,262){\scriptsize$ j $} 
6047:                   \put(527,262){\scriptsize$ k $} 
6048:                   \put(671,262){\scriptsize$ \overline l $} 
6049:                   \put(360,406){\scriptsize$ p $} 
6050:                   \put(438,459){\scriptsize$ l $} 
6051:                   \put(317,373){\scriptsize$ \mu $} 
6052:                   \put(402,437){\scriptsize$ \nu $} 
6053:                   \put(254,39){\scriptsize$ 1 $} 
6054:                   \put(396,39){\scriptsize$ 2 $} 
6055:                   \put(539,39){\scriptsize$ 3 $} 
6056:                   \put(680,39){\scriptsize$ 4 $} 
6057:                   \setlength{\unitlength}{1pt}}
6058:   \put(0,125)   {$ Z(\M_E) \cir Z(B_{0,4}[i,j,k,{\bar l};p;\mu,\nu]) ~= $}
6059:   \end{picture} \nonumber
6060:   \\[1.2em] \dsty
6061:   =\; \sum_q \sum_{\rho,\sigma} \Bmat{jk}ilpq\nu\mu\rho\sigma \,
6062:   Z(B_{0,4}[i,k,j,\bar l;q;\sigma,\rho]) \;.~~~
6063:   \label{eq:Al-braid-cobord}  
6064:   \end{eqnarray} \Xlabp 118
6065: We can now verify the relation \erf{eq:Alpha-inter} for the braiding on a basis 
6066: $\beta_{0,4}[i,j,k,\bar l;p;\mu,\nu]$ of the space $\calh^c(E^c_{0,4}[i,j,k,\bar l])$ 
6067: as follows:
6068:   \beaa
6069:   Z(\M_E) \cir \Alpha( \beta_{0,4}[i,j,k,\bar l;p;\mu,\nu] ) \!\!
6070:   &= Z(\M_E) \cir Z(B_{0,4}[i,j,k,{\bar l};p;\mu,\nu]) 1
6071:   \\{}\\[-.6em] &\displaystyle
6072:   = \sum_q \sum_{\rho,\sigma} \Bmat{jk}ilpq\nu\mu\rho\sigma \,
6073:       Z(B_{0,4}[i,k,j,\bar l;q;\sigma,\rho ] ) 1
6074:   \\{}\\[-.8em] &\displaystyle
6075:   = \sum_q \sum_{\rho,\sigma}
6076:     \Bmat{jk}ilpq\nu\mu\rho\sigma \,
6077:       \Alpha( \beta_{0,4}[i,k,j,\bar l;q;\sigma,\rho] )
6078:   \\{}\\[-.8em] 
6079:   &= \Alpha \cir \calp_{(23)} ( \beta^{\mu\nu}_1 )
6080:   = \Alpha \cir U_{\gamma_{E^c}^{}} ( \beta_{0,4}[i,j,k,\bar l;p;\mu,\nu] )\,.
6081:   \eear\ee
6082: Here the first and third step just use the definition of
6083: $\Alpha$ in \erf{eq:Alpha-on-seRS}, the second and fourth step 
6084: are the equalities in \erf{eq:Al-braid-cobord} and \erf{eq:Bmat-conf},
6085: respectively, and the final step amounts to the definition 
6086: $\beta^{\mu\nu}_1 \eq U_{\gamma_{E^c}^{}} \beta_{0,4}[i,j,k,\bar l;p;\mu,\nu]$
6087: together with the first of the properties \erf{eq:Alpha-prop}.
6088: 
6089: 
6090: \subsubsection*{Consistency of $\Alpha$} 
6091: 
6092: Let us now sketch how compatibility with the braiding implies
6093: \erf{eq:Alpha-inter}. Let $\gamma$ be a closed path in $\hat\calm_{0,n}$ 
6094: starting and ending at $[E^c_{0,n}]$. $\gamma^c$ can always be written as the 
6095: composition of two paths $\gamma_1$ and $\gamma_2$, of which $\gamma_1$ only 
6096: changes the local coordinates but leaves the insertion points fixed, while 
6097: $\gamma_2$ only moves the insertion points, with all local coordinates 
6098: remaining of the form $\varphi_k(z) \eq z{+}p_k$.
6099: 
6100: The paths $\gamma_1$ and $\gamma_2$ are closed already by themselves, and 
6101: hence can be treated separately. For paths of type $\gamma_1$ relation 
6102: \erf{eq:Alpha-inter} holds by an argument similar to the one used in the 
6103: first example above. A path of type $\gamma_2$ furnishes an element of the 
6104: braid group with $n$ strands, which has the special property that the strand 
6105: starting at position $k$ also ends at position $k$. The braid group is generated
6106: by moves $g_{k,k+1}$ that exchange two neighbouring strands. It is thus 
6107: sufficient to verify the relation $\Alpha \cir U_{\gamma_{E^c}^{}} \eq Z(\M_E) 
6108: \cir \Alpha$ for such elements. This in turn follows by a similar argument as 
6109: used for the four-point conformal block above, taking also into account that 
6110: the braid matrices $\Bsmat ijkl{}{}$ control the behaviour of the intertwiners 
6111: themselves under analytic continuation, not only that of the four-point block,
6112:   \be
6113:   \VO {l,\nu}kp{v_3}{z} \, \VO {p,\mu}ji{v_2}{w} 
6114:   = \sum_q \sum_{\rho,\sigma} \Bmat{j\,k}ilpq\nu\mu\rho\sigma\,
6115:   \VO {l,\rho}jq{v_2}{w} \, \VO {q,\sigma}ki{v_3}{z} \,,
6116:   \ee
6117: where the \lhs\ is understood as analytic continuation from $|w|\,{<}\,|z|$ 
6118: to $|w|\,{>}\,|z|$, taking $w$ clockwise around $z$.
6119: 
6120: 
6121: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6122: 
6123: \sect{The fundamental correlation functions}\label{sec:cf}
6124: 
6125: Ultimately, we would like to obtain the correlators of a CFT as functions
6126: of the world sheet moduli and of the field insertions. To this end we must
6127: combine the results from the previous sections with additional information
6128: about the world sheet.
6129: In particular, for non-zero central charge the correlators $C$ do depend
6130: explicitly on the world sheet metric $g$ and not only on the conformal
6131: equivalence class of metrics, see e.g.\ \cite{gawe13}. Thus, to find the
6132: value of a correlator it is not enough to know the conformal structure of
6133: the world sheet, but rather the metric itself is relevant.
6134: In sections \ref{sec:cf-bnd3pt}\,--\,\ref{sec:cf-1defect-xcap} we will treat 
6135: certain ratios of correlators in which the Weyl anomaly cancels.\,%
6136:   \footnote{~These ratios, in turn, are obtained from {\em unnormalised
6137:   correlators\/} $C(\,\cdots)$, as opposed to the (normalised) expectation
6138:   values $\langle\, \cdots \rangle$ which obey $\langle\, \one \,\rangle \eq
6139:   1$. It is the correlators $C$ which appear in factorisation constraints,
6140:   and which give, e.g., the torus and annulus partition functions.}
6141: 
6142: 
6143: \subsection{Riemannian world sheets}\label{sec:ws-db-cmf}
6144: 
6145: In sections \ref{sec:fi} and \ref{sec:rc} we were concerned
6146: with the 3-d TFT, and accordingly only needed the world sheet as
6147: a topological manifold. To obtain the correlator as a function,
6148: rather than as a vector in the TFT state space, we need the
6149: world sheet metric. To distinguish the two settings, we will use the terms 
6150: {\em topological world sheet\/} and {\em Riemannian world sheet\/}.
6151: 
6152: 
6153: \subsubsection*{Definition}
6154: 
6155: By a Riemannian word sheet $\X^g$ we mean a compact,
6156: not necessarily orientable, two-manifold with
6157: metric $g$ which may have non-empty boundary, defects and field insertions.
6158: 
6159: In section \ref{sec:fi} we found that on the topological world sheet the fields,
6160: boundaries and defects are labelled by equivalence classes of certain tuples.
6161: The same holds true for insertions on the Riemannian world
6162: sheet, but the quantities entering the tuples are slightly
6163: different. We will only describe the tuples and refrain from
6164: giving the equivalence relations explicitly. They take a similar form as in 
6165: the topological case, and will not be needed for the applications below.
6166: 
6167: Define a {\em local bulk coordinate germ\/} $[f]$ to be a germ of injective 
6168: functions $f{:}\; D_\delta \To \X^g$ of an arbitrarily small disk $D_\delta\eq
6169: \{ z\iN\complex \,\big|\,|z|\,{<}\,\delta \}$ to $\X^g$ such that $f(0)$ lies 
6170: in the interior of $\X^g$. On the disk $D_\delta$ we take the standard metric 
6171: induced by $\complex \cong \reals^2$. The maps $f$ are required to be conformal, 
6172: i.e.\ satisfy $f^* g \eq \Omega(x,y) ({\rm d}x^2 \,{+}\, {\rm d}y^2)$.
6173:  %
6174: Define further a {\em local boundary coordinate germ} $[d]$ 
6175: to be a germ of injective functions $d{:}\; H_\delta \To \X^g$ 
6176: of an arbitrarily small semidisk $H_\delta \eq 
6177: \{ z\iN\complex \,\big|\,|z|\,{<}\,\delta\;{\rm and}\;{\rm Im}(z)\,{\ge}\,0 \}$ 
6178: to $\X^g$ such that $d(x)\iN\partial \X^g$ iff $x\iN\reals$. Again on $H_\delta$
6179: we take the standard metric and require the map $d$ to be conformal.
6180: 
6181: Since the definition of local coordinate germs depends only on the
6182: conformal structure on $\X^g$, we denote fields on $\X^g$ with
6183: a superscript $c$ to distinguish them from their topological counterparts in 
6184: section \ref{sec:fi}. The possible field insertions on $\X^g$ are as follows.
6185: \ITemize
6186: \item[\Nxt]
6187: A {\em bulk field\/} $\vPhi^c$ is a tuple
6188:   \be
6189:   \vPhi^c = (i,j,\phi,[f])\,,
6190:   \ee
6191: where $i,j\iN\II$ label simple objects, $\phi$ is a morphism
6192: in $\HomAA(U_i\,{\otimes^+}A\,{\otimes^-}\,U_j,A)$,
6193: and $[f]$ is a local bulk coordinate germ.
6194: %
6195: \item[\Nxt]
6196: A {\em boundary field\/} $\hatpsi^c$ is a tuple
6197:   \be
6198:   \hatpsi^c = (M,N,\RCA,\psi,[d])\,,
6199:   \ee
6200: where $M$ and $N$ are left $A$-modules, $\RCA$ is an object of $\calc$, 
6201: $\psi\iN\HomA(M\Oti \RCA,N)$, and $[d]$ is a local boundary coordinate germ.
6202: %
6203: \item[\Nxt]
6204: A {\em defect field\/} $\Thetha^c$ is a tuple
6205:   \be
6206:   \Thetha^c = (\X,\orr_2(X),Y,\orr_2(Y),i,j,\vartheta,[f])\,,
6207:   \ee
6208: where $X$, $Y$ are $A$-bimodules, $\orr_2(X)$ is
6209: a local orientation of $\X^g$ around the defect segment
6210: labelled by $X$, and similarly for $\orr_2(Y)$. Further, $i,j\iN\II$ and
6211: $\vartheta \iN \HomAA(U_i\,{\otimes^+}\,Z\,{\otimes^-}\,U_j,Z')$,
6212: where $Z, Z'$ are defined as in \erf{eq:defect-ZZ-XY}. $[f]$ is a local bulk 
6213: coordinate germ which must fulfill the condition that $f(x)$ lies on the 
6214: defect circle if $x$ lies on the real axis.
6215: \end{itemize}
6216: %
6217: As an additional datum, the boundary and bulk insertions are ordered, i.e.\ 
6218: if there are $m$ boundary fields and altogether $n$ bulk and defect fields,
6219: we have to pick maps from $\{1,2,\dots,m\}$ to the set of boundary
6220: fields and from $\{1,2,\dots,n\}$ to the set of bulk/defect fields.
6221: Thinking of bulk fields as a special type of defect fields, we write
6222:   \be
6223:   \hatpsi^c_k = (M_k,N_k,\RCA_k,\psi_k,[d_k]) \qquad {\rm and} \qquad
6224:   \Thetha^c_k = (\X_k,\orr_2(X)_k,Y_k,\orr_2(Y)_k,i_k,j_k,\vartheta_k,[f_k])
6225:   \ee
6226: for the $k$th boundary and defect field, respectively, and their defining data.
6227: The ordering is needed because below we construct an extended Riemann surface
6228: from $\X^g$, and in that situation the marked points from an ordered set.
6229: 
6230: 
6231: \subsubsection*{The complex double of $\X^g$}
6232: 
6233: From a Riemannian world sheet $\X^g$ we can obtain an
6234: extended Riemann surface $\hat\X^g$, the {\em complex double of} $\X^g$.
6235: It is constructed as follows.
6236: 
6237: If there are no field insertions on $\X^g$, then $\hat \X^g$ is
6238: simply the orientation bundle over $\X^g$, divided by the usual
6239: equivalence relation, i.e.
6240:   \be
6241:   \hat \X^g = {\rm Or}(\X^g) /{\sim}
6242:   \qquad {\rm with}~~ (x,\orr_2) \sim (x,-\orr_2) 
6243:   \quad{\rm for}~~ x\iN\partial \X^g \,.
6244:   \ee
6245: Since a metric (or just a conformal structure) together with an orientation 
6246: defines a complex structure, we obtain a complex structure
6247: on ${\rm Or}(\X^g)$. The equivalence relation leading from ${\rm Or}(\X^g)$ to 
6248: $\hat \X^g$ respects this complex structure. In terms of
6249: charts it amounts to identifying a subset of the upper 
6250: half plane with the conjugate subset of the lower half plane along 
6251: the real axis. Thus $\hat \X^g$ is a compact Riemann surface without boundary.
6252: 
6253: If in addition there are field insertions on $\X^g$, one obtains an
6254: ordered set of marked points on $\hat \X^g$ as follows.
6255: \ITemize
6256: \item[\Nxt] 
6257: Each boundary field $\hatpsi^c \eq (M,N,\RCA,\psi,[d])$ gives rise to a single
6258: marked point $(p,[\varphi],\RCA)$ on $\hat \X^g$, where
6259:   \be
6260:   \varphi(z) = \begin{cases}
6261:   [d(z),\orr_2(d)] & {\rm for}~ {\rm Im}(z) \,{\ge}\, 0 \,, \\[.3em]
6262:   [d(z^*),-\orr_2(d)] & {\rm for}~ {\rm Im}(z) \,{\le}\, 0 \,. \end{cases}
6263:   \labl{eq:com-double-bnd}
6264: Here $\orr_2(d)$ denotes the local orientation of $\X^g$ obtained
6265: form the push-forward of the standard orientation on $\complex$ via $d$. 
6266: One can verify that $\varphi$ is an injective analytic function from 
6267: a small disk in $\complex$ to $\hat \X^g$.
6268: The insertion point $p$ is equal to $\varphi(0)$. Finally, if $\hatpsi^c$
6269: has label $k$ from the ordering of fields, we also assign the number $k$ 
6270: to the marked point \erf{eq:com-double-bnd}.
6271: %
6272: \item[\Nxt] 
6273: Each bulk field $\vPhi^c \eq (i,j,\phi,[f])$ or defect field
6274: $\Thetha^c \eq (X,\orr_2(X),Y,\orr_2(Y),i,j,\vartheta,[f])$ 
6275: gives rise to two marked points $(p_i,[\varphi_i],\SCA_i)$ and 
6276: $(p_j,[\varphi_j],\SCA_j)$ on $\hat \X^g$. Here 
6277:   \be
6278:   \varphi_i(z) = [f(z),\orr_2(f)]  \qquad{\rm and}\qquad
6279:   \varphi_j(z) = [f(z^*),-\orr_2(f)]
6280:   \labl{eq:com-double-bulk}
6281: are analytic functions, and $p_{i,j} \eq \varphi_{i,j}(0)$. If the 
6282: bulk\,/\,defect field has label $k$  from the ordering on $\X^g$, the two marked
6283: points \erf{eq:com-double-bulk} are assigned the numbers $m\,{+}\,2k\,{-}\,1$ 
6284: and $m\,{+}\,2k$, respectively, with $m$ the number of boundary fields.
6285: \end{itemize}
6286: %
6287: We also need to fix a Lagrangian submodule $\lambda^c\,{\subset}\,H_1(\hat\X^g,
6288: \zet)$. This is again done via the inclusion $\iota{:}~X^g \To \MX$ (here $\MX$ 
6289: is the connecting manifold for $\X^g$, considered as a topological manifold), 
6290: which gives rise to a map $\iota_*{:}~ H_1(\hat\X^g,\zet) \To H_1(\MX,\zet)$.
6291: We take $\lambda^c$ to be the unique Lagrangian submodule that contains 
6292: ${\rm ker}(\iota_*)$.
6293: 
6294: 
6295: \subsubsection*{Holomorphic factorisation}
6296: 
6297: Given a Riemannian world sheet $\X^g$, the correlation function
6298: $C(\X^g)$ of the CFT for that world sheet is a linear function
6299:   \be
6300:   C(\X^g) : \quad
6301:   \bigotimes_{k=1}^m \RCA_k \otimes
6302:   \bigotimes_{l=1}^n\, (\SCA_{i_l} \oti \SCA_{j_l})
6303:   \,\longrightarrow\, \complex \,.
6304:   \labl{eq:C-function}
6305: Any correlation function must solve the chiral Ward identities coming from the 
6306: holomorphic (and anti-holomorphic) fields in the theory, in particular those 
6307: coming from fields in the chiral algebra. This leads to the principle of 
6308: holomorphic factorisation \cite{witt39}, which can be simply formulated 
6309: as the statement that
6310:   \be
6311:   C(\X^g) \in \calh^c( \hat \X^g ) \,,
6312:   \ee
6313: i.e.\ that the correlator for $\X^g$ is an element of the space of
6314: conformal blocks on the complex double of $\X^g$.
6315: 
6316: 
6317: \subsubsection*{Metric dependence}
6318: 
6319: Consider a Riemannian world sheet $\X^g$. 
6320: Under a Weyl transformation $g \,{\mapsto}\, g' \eq \eE^{\gamma} g$
6321: for some $\gamma{:}~ \X^g \To \reals$, the correlator changes by a factor:
6322:   \be
6323:   C(\X^{g'}) = \eE^{c S[\gamma]}\, C(\X^g) \,,
6324:   \labl{eq:Weyl-anom}
6325: where $c$ is the central charge and $S[\gamma]$ is the 
6326: (suitably normalised) Liouville action, see e.g.\ \cite{gawe13} for details. 
6327: 
6328: If one works with ratios of correlators, the prefactor can be cancelled 
6329: in the following manner. Let $\dot\X^g$ be the Riemannian world
6330: sheet obtained from $\X^g$ by removing all field insertions and
6331: labelling all boundaries and defects by the algebra $A$. Then
6332: $C(\dot\X^g)$ is just a complex number, rather than a function
6333: as in \erf{eq:C-function}, but it shows the same behaviour 
6334: \erf{eq:Weyl-anom} under
6335: Weyl transformations of the metric as $C(\X^g)$. It follows that the ratio
6336:   \be
6337:   C(\X^g) / C(\dot\X^g) 
6338:   \labl{eq:corr-ratio}
6339: only depends on the conformal equivalence class of the metric $g$.
6340: It is these ratios, rather than the correlators themselves, that we will 
6341: present below for the fundamental correlators.
6342: 
6343: 
6344: \subsubsection*{The topological world sheet and its double}
6345: 
6346: Let us define an operation $\tildeF$ which assigns to a Riemannian world
6347: sheet $\X^g$ a topological world sheet $\X \eq \tildeF(\X^g)$ as follows.
6348: 
6349: If there are no field insertions on $\X^g$, then $\X$ is obtained from
6350: $\X^g$ by just forgetting the metric. If in addition there
6351: are field insertions on $\X^g$, then the corresponding insertions on the
6352: topological world sheet $\X$ are obtained as follows.
6353:   \ITemize
6354:   \item[\Nxt] 
6355: A boundary field  $\hatpsi^c \eq (M,N,V,\psi,[d])$ on $\X^g$ gives an insertion
6356: $\hatpsi \eq (M,N,V,\psi,p,[\gamma])$ on $\X$, where
6357: $p \eq d(0)$ and $\gamma(t) \eq d(t)$.
6358:   \item[\Nxt] 
6359: A bulk field $\vPhi^c \eq (i,j,\phi,[f])$ on $\X^g$ gives an insertion
6360: $\vPhi \eq (i,j,\phi,p,[\gamma],\orr_2(p))$ on $\X$, where $p \eq f(0)$ and 
6361: $\gamma(t) \eq f(t)$.  The local orientation $\orr_2(p)$ of $\X^g$ is given 
6362: by $\orr_2(f)$, i.e.\ the push-forward of the standard orientation
6363: of $\complex$ via $f$.
6364: \\[3pt]
6365: Similarly, for a defect field $\Thetha^c \eq (X,\orr_2(X),Y,\orr_2(Y),i,j,
6366: \vartheta,[f])$  on $\X^g$ one gets an insertion $\Thetha^c \eq 
6367: (X,\orr_2(X),Y,\orr_2(Y),i,j,\vartheta,p,[\gamma],\orr_2(p))$ on $\X$.
6368:   \end{itemize}
6369: With these definitions one can verify that the diagram
6370:   \be
6371:   %% [pic~45]
6372:   \begin{array}{rcccl}
6373:   & \X^g & \stackrel\tildeF\longmapsto & \X
6374:   \\[1pt]
6375:   &\top&&\top \\[-8pt] &\downarrow&&\downarrow
6376:   \\[2pt]
6377:   & \hat\X^g & \stackrel\ff\longmapsto & \hat\X
6378:   \end{array}
6379:   \ee
6380: with $\ff$ the map from extended Riemann surfaces to extended surfaces 
6381: defined in section \ref{sec:cb-link}, commutes.
6382: 
6383: The TFT construction describes the correlator as a state of the TFT,
6384: $C(\X) \iN \calh(\hat \X)$. In order to obtain the correlator 
6385: $C(\X^g) \iN \calh^c( \hat \X^g )$ we thus need an isomorphism
6386:   \be \Alpha :\quad \calh^c( \hat \X^g ) \stackrel \cong \to
6387:   \calh(\hat \X) \,.
6388:   \ee
6389: Since the metric $g$ does not appear in the topological construction, 
6390: $\Alpha$ should have some dependence on $g$. 
6391: Here we are interested in correlators (or rather, the ratios
6392: \erf{eq:corr-ratio}) on the upper half plane and on the complex plane
6393: (together with the point at infinity). In both cases we consider
6394: the standard metric\,\footnote{% 
6395:   ~This makes both world sheets non-compact, but they are still conformally
6396:   equivalent to compact world sheets. We could have chosen
6397:   a compact metric instead, but since the ratios \erf{eq:corr-ratio}
6398:   are invariant under Weyl-transformations this makes no difference.}
6399: ${\rm d}x^2+{\rm d}y^2$. For this metric we take the isomorphism $\Alpha$ to be
6400: the one defined in section \ref{sec:cb-link}.
6401: 
6402: 
6403: \subsection{Boundary three-point function on the upper half plane}
6404: \label{sec:cf-bnd3pt}
6405: 
6406: Let $A$ be a symmetric special Frobenius algebra in $\calc$. Take the 
6407: Riemannian world sheet $\X^g$ to be the upper half plane together with three
6408: field insertions
6409:   \be
6410:   \hatpsi_1 = (N,M,\SCA_i,\psi_1,[d_1])\,,\quad
6411:   \hatpsi_2 = (K,N,\SCA_j,\psi_2,[d_2])\,,\quad
6412:   \hatpsi_3 = (M,K,\SCA_k,\psi_3,[d_3])
6413:   \ee
6414: on its boundary. Here $M, N, K$ are left $A$-modules, the three fields
6415: are inserted at positions $x_i$ satisfying $0\,{<}\,x_1\,{<}\,x_2\,{<}\,x_3$, 
6416: and the local boundary coordinate germs $[d_n]$ around these points are given 
6417: by arcs $d_n(z) \eq x_n{+}z$.  
6418: Note that $\tildeF(\X^g)$ gives precisely the 
6419: topological world sheet displayed in figure \erp 24.
6420: 
6421: The complex double of $\X^g$ is the extended Riemann surface consisting of
6422: the Riemann sphere $\hat \X^g\eq\mathbb{P}^1$, parametrised
6423: as $\complex\,{\cup}\,\{\infty\}$, and the three marked points
6424: $(x_1, [\varphi_1] , \SCA_i)$, $(x_2, [\varphi_2] , \SCA_j)$ and 
6425: $(x_3, [\varphi_3] , \SCA_k)$, with the local coordinates given by
6426: $\varphi_n(z) \eq x_n{+}z$, $n\eq1,2,3$. The correlator $C(\X^g)$ is an element 
6427: of the space of conformal blocks $\calh^c(\hat \X^g)$. A basis of this space 
6428: is provided by the three-point blocks \erf{eq:3pt-block}. One can thus 
6429: write the correlator as a linear combination
6430:   \be
6431:   C(\X^g)(u,v,w) = \sum_{\delta=1}^{\N ij{\bar k}} 
6432:   c(M \hatpsi_1 N \hatpsi_2 K \hatpsi_3 M)_\delta \,
6433:      \langle 0|\, \VO 0k{\bar k}w{x_3}\,
6434:              \VO {\bar k,\delta}jiv{x_2}\, \VO ii0u{x_1} \,|0\rangle \,,
6435:   \labl{eq:uhp-3pt-block}
6436: where $u\iN\SCA_i$, $v\iN\SCA_j$ and $w\iN\SCA_k$.
6437: The constants $c(M \hatpsi_1 N \hatpsi_2 K \hatpsi_3 M)_\delta$
6438: are precisely those determined by the TFT-analysis in \erf{eq:bnd-3pt-rib}.
6439: This follows from applying the map $\Alpha$ to both sides of 
6440: \erf{eq:uhp-3pt-block}, which results in \erf{eq:uhp-3pt-cobord}.
6441: Altogether, for the ratio \erf{eq:corr-ratio} we find, 
6442: using the explicit form \erf{eq:3pt-block} of the three-point block, 
6443:   \be  \!\begin{array}{lr}\dsty
6444:   \frac{C(\X^g)(u,v,w)}{C(\dot\X^g)} \!\! &\dsty
6445:   = \Big( \sum_{\delta=1}^{\N ij{\bar k}} 
6446:   \frac{c(M \hatpsi_1 N \hatpsi_2 K \hatpsi_3 M)_\delta}{
6447:   \dim(A)}\, B^\delta_{kji}(w,v,u) \Big) \,
6448:   (x_3{-}x_2)^{\Delta_i(u)-\Delta_j(v)-\Delta_k(w)} ~~
6449:   \\{}\\[-.8em] &\dsty
6450:   (x_3{-}x_1)^{\Delta_j(v)-\Delta_i(u)-\Delta_k(w)}\,
6451:   (x_2{-}x_1)^{\Delta_k(w)-\Delta_i(u)-\Delta_j(v)} .
6452:   \eear\labl{eq:3bnd-corr}
6453: The vectors $u\iN\SCA_i$, $v\iN\SCA_j$ and $w\iN\SCA_k$ tell 
6454: us which fields of the infinite-dimensional representation spaces of the
6455: chiral algebra are inserted at the points $x_3\,{>}\,x_2\,{>}\,x_1$, and
6456: $\Delta_i(u)$, $\Delta_j(v)$, $\Delta_k(w)$ are their conformal weights.
6457: Note also that $C(\dot\X^g) \eq \dim(A)$. The constants $c_\delta$ have
6458: been expressed in a basis in \erf{eq:3bnd-inv-basis};
6459: for the Cardy case they are given by \erf{eq:3bnd-basis-cardy}.
6460: 
6461: \dt{Remark}
6462: Even though the correlator of three boundary
6463: fields is built directly from a single conformal block, rather than from
6464: a bilinear combination of blocks like the correlator of three bulk
6465: fields, it is {\em not\/}, in general, an analytic function of the
6466: insertion points. As an example, take the correlator of two
6467: equal boundary fields\,\footnote{%
6468:   ~This is zero in general, in particular if $i\,{\neq}\,\bar\imath$.
6469:   But there are many examples, like the Virasoro minimal models, where it
6470:   can be non-zero.}
6471: $\hatpsi^c$, which for $x_2\,{>}\,x_1$ is given by
6472:   \be
6473:   C = c(M\hatpsi M\hatpsi M)\, B_{ii}(u,u)\, (x_2\,{-}\,x_1)^{-2\Delta_i(u)}
6474:   .  \ee
6475: For $x_1\,{>}\,x_2$ one finds the same answer up to an exchange 
6476: $x_1\,{\leftrightarrow}\, x_2$, so that for all values of $x_1, x_2$ we can
6477: write
6478:   \be
6479:   C = c(M\hatpsi M\hatpsi M)\, B_{ii}(u,u)\,
6480:   \big|x_2\,{-}\,x_1\big|^{-2\Delta_i(u)} , \ee
6481: which is not analytic in $x_1$ and $x_2$ (unless $\Delta_i\iN\zet$, which is 
6482: a very special case). It is thus not quite appropriate to say that
6483: the theory on the boundary is chiral.
6484: 
6485: \dtl{Remark}{MNKA}
6486: Consider the case $M\eq N\eq K\eq A$, i.e.\ on each boundary segment the 
6487: boundary condition is given by $A$, regarded as a left module over itself. 
6488: Using the Frobenius reciprocity relation $\HomA(A\oti U, A) \,{\cong}\,
6489: \Hom(U,A)$ (see e.g.\ proposition I:4.12), the elements of
6490: $\HomA(A\oti \SCA_k, A)$ can be expressed as
6491:   \be
6492:   \psi = m \circ (\id_A \oti b^A_{k,\alpha})\,,
6493:   \ee
6494: where $m$ is the multiplication morphism of $A$ and
6495: $b^A_{k,\alpha}$ denotes the basis \erf{pic77} of $\Hom(\SCA_k,A)$. 
6496: The ribbon graph \erf{eq:bnd-3pt-rib} then takes the form
6497:   %  [ pic 4_28 ]
6498:   \bea \begin{picture}(380,133)(10,46)
6499:   \put(130,0)   {\Includeourbeautifulpicture 28a }
6500:   \put(310,0)   {\Includeourbeautifulpicture 28b }
6501:   \put(163,144)    {\scriptsize $A$}
6502:   \put(180,104)    {\scriptsize $A$}
6503:   \put(195,123)    {\scriptsize $A$}
6504:   \put(249,60)     {\scriptsize $i$}
6505:   \put(206,59)     {\scriptsize $j$}
6506:   \put(217,17.3)   {\scriptsize $\bar k$}
6507:   \put(190,27)     {\scriptsize $k$}
6508:   \put(230.8,41)   {\scriptsize $\delta$}
6509:   \put(253,77)     {\scriptsize $\alpha$}
6510:   \put(213,77)     {\scriptsize $\beta$}
6511:   \put(189,65)     {\scriptsize $\gamma$}
6512:   \put(349,143)    {\scriptsize $A$}
6513:   \put(330,100)    {\scriptsize $A$}
6514:   \put(380,100)    {\scriptsize $A$}
6515:   \put(307,116)    {\scriptsize $A$}
6516:   \put(382.4,84)   {\scriptsize $\alpha$}
6517:   \put(343,84)     {\scriptsize $\beta$}
6518:   \put(318.2,84)   {\scriptsize $\gamma$}
6519:   \put(373.5,67)   {\scriptsize $i$}
6520:   \put(341.5,67)   {\scriptsize $j$}
6521:   \put(346,17)     {\scriptsize $\bar k$}
6522:   \put(319,27)     {\scriptsize $k$}
6523:   \put(359.5,41)   {\scriptsize $\delta$}
6524:   \put(0,86)       {$ c(A \hatpsi_1 A \hatpsi_2 A \hatpsi_3 A)_\delta \;= $} 
6525:   \put(275,86)     {$ = $} 
6526:   \epicture29 \labl{eq:bnd-3pt-rib-A} \xlabp 28
6527: In the second equality one uses associativity of the multiplication, the
6528: transformations (I:3.49), as well as the specialness relation 
6529: $m\cir\Delta\eq\id_A$. One can now substitute the expression of the
6530: multiplication in a basis as in (I:3.7) to evaluate the ribbon invariant
6531: \erf{eq:bnd-3pt-rib-A}. This yields
6532:   \be
6533:   c(A \hatpsi_1 A \hatpsi_2 A \hatpsi_3 A)_\delta
6534:   = \sum_{\mu,\nu} m_{j\beta,i\alpha}^{\bar k \mu; \delta}\,
6535:   m_{k\gamma,\bar k\mu}^{0\nu}\, \big(\eps \cir b^A_{0,\nu}\big) \,.
6536:   \labl{eq:C-m-OPE}
6537: Using the OPE of boundary fields as defined in 
6538: (I:3.11) to evaluate the correlator
6539: $\langle 0|\hatpsi_3 \hatpsi_2 \hatpsi_1 |0\rangle$
6540: and comparing the result to \erf{eq:C-m-OPE} gives
6541:   \be
6542:   m_{j\beta,i\alpha}^{\bar k \mu; \delta} =
6543:   C_{j\beta,i\alpha}^{\bar k \mu; \delta} \,.  \ee
6544: This is nothing but the relation (I:3.14) that was already established
6545: in section I:3.2. In words it says that the boundary OPE on the boundary
6546: condition labelled by $A$ is equal to the multiplication on the algebra $A$.
6547: 
6548: 
6549: \subsection{One bulk and one boundary field on the upper half plane}
6550: \label{sec:cf-1bulk1bnd}
6551: 
6552: The next correlator we consider is that of one bulk field and
6553: one boundary field on the upper half plane. 
6554: Let $\X^g$ be the upper half plane with field insertions
6555:   \be
6556:   \vPhi^c = (i,j,\phi,[f]) \qquad {\rm and} \qquad
6557:   \hatpsi^c = (M,M,\SCA_k,\psi,[d]) \,,
6558:   \ee
6559: where $i,j,k\iN\II$ label simple objects, $\phi\iN\HomAA(\SCA_i\otiP A\otim 
6560: \SCA_j,A)$ and $\psi\iN\HomA(M\oti \SCA_k,M)$. Further, $[f]$ is a local 
6561: bulk coordinate germ and $[d]$ a local boundary coordinate germ, with
6562: $f(\zeta)\eq z\,{+}\,\zeta$ and $d(\zeta) \eq s\,{+}\,\zeta$ for
6563: $z$ in the upper half plane and $s$ on the real axis.
6564: In fact we take $z\eq x\,{+}\,\ii y$ and $s\,{>}\,x\,{>}\,0$.
6565: 
6566: The complex double of $\X^g$ is the extended Riemann surface $\hat \X^g$ 
6567: consisting of the Riemann sphere $\complex\,{\cup}\,\{\infty\}$ and the three 
6568: marked points $(s, [\varphi_1] , \SCA_k)$, $(x{+}\ii y, [\varphi_2] , \SCA_i)$ 
6569: and $(x{-}\ii y, [\varphi_3], \SCA_j)$, with the local coordinates given by 
6570: $\varphi_1(\zeta) \eq s\,{+}\,\zeta$, $\varphi_2(\zeta) \eq x{+}\ii y\,{+}\,
6571: \zeta$ and $\varphi_3(\zeta) \eq x{-}\ii y\,{+}\,\zeta$. We select a basis 
6572: $\{ \beta_\delta \}$ in the space of conformal blocks $\calh^c(\hat \X^g)$ as
6573:   \beaa
6574:   \beta_\delta \!\!& := \langle 0|\, \VO 0k{\bar k}w{z_3}\,
6575:              \VO {\bar k,\delta}iju{z_2}\, \VO jj0v{z_1} \,|0\rangle 
6576:   \,\Big|_{z_1=x{-}\ii y , z_2=x{+}\ii y , z_3=s}~~
6577:   \\[15pt]\displaystyle
6578:   & \;= B^\delta_{kij}(w,u,v)\,
6579:   \eE^{\frac{\pi \ii}{2}(\Delta_k(w)-\Delta_i(u)-\Delta_j(v))}\,
6580:   \big(s{-}x{+}\ii y\big)^{2(\Delta_i(u)-\Delta_j(v))}
6581:   \\[7pt]\displaystyle&\hspace*{10em}\times
6582:   \big( (s{-}x)^2+y^2 \big)^{-\Delta_i(u)+\Delta_j(v)-\Delta_k(u)}
6583:   \big(2y\big)^{\Delta_k(w)-\Delta_i(u)-\Delta_j(v)} ,
6584:   \eear\ee
6585: where in the first line it is understood that the points $z_{1,2,3}$ are taken 
6586: to their present position from the standard block \erf{eq:standard-block} with 
6587: $z_n\eq n$ via continuation along the contours indicated in figure 
6588: \erf{eq:Bxys}.  In particular, the connection to $B_\delta$ in \erf{eq:Bxys}
6589: is $\Alpha(\beta_\delta) \eq Z(B(x,y,s)_\delta, \emptyset,\hat\X)$.
6590: Applying $\Alpha^{-1}$ to both sides of \erf{eq:1B1b-TFT} thus yields
6591:   \be
6592:   C(\X^g) = \sum_{\delta} c(\vPhi ; M \hatpsi)_\delta \,\beta_\delta \,,
6593:   \labl{eq:blk-bnd-corr-const}
6594: where the constants $c(\vPhi; M \hatpsi)_\delta$ 
6595: are given by the ribbon invariant \erf{eq:1bulk-1bnd-sc}.
6596: 
6597: We conclude that the correlator ratio for one bulk field 
6598: and one boundary field on the upper half plane is of the form
6599:   \bea  
6600:   \!\!\!\! \dsty \frac{C(\X^g)(w,u,v)}{C(\dot\X^g)} = \dsty 
6601:   \Big( \sum_{\delta=1}^{\N ij{\bar k}}
6602:   \frac{c(\phi; M \hatpsi)_\delta}{\dim(A)} \, 
6603:   B^\delta_{kij}(w,u,v)\,
6604:   \eE^{\frac{\pi \ii}{2}(\Delta_k(w)-\Delta_i(u)-\Delta_j(v))} \Big)
6605:   \\{}\\[-.7em]  \hspace*{3.9em} \dsty
6606:   \big(s{-}x{+}\ii y\big)^{2(\Delta_i(u)-\Delta_j(v))} 
6607:   \big( (s{-}x)^2\,{+}\,y^2 \big)^{-\Delta_i(u)+\Delta_j(v)-\Delta_k(u)}
6608:   \big(2y\big)^{\Delta_k(w)-\Delta_i(u)-\Delta_j(v)} ,
6609:   \eear\labl{eq:1bulk-1bnd}
6610: where $u \oti v \iN \SCA_i \oti \SCA_j$ gives the bulk field and
6611: $w \iN \SCA_k$ the boundary field.
6612: 
6613: \dtl{Remark}{rem:1bulk-uhp'}
6614: The special case of one bulk field on the upper half plane without any boundary 
6615: field insertion (compare remark \ref{rem:1bulk-uhp}) is obtained by setting 
6616: $\hatpsi \eq (M,M,\one,\id_M,[g])$ as well as $w\eq |0\rangle \iN \SCA_0$. 
6617: Then \erf{eq:1bulk-1bnd} becomes
6618:   \be
6619:   \frac{C(\X^g)(u,v)}{C(\dot\X^g)} = \frac{c(\phi; M )}{\dim(A)} \, 
6620:   B_{ij}(u,v)\, \eE^{- \pi \ii \Delta_i(u)} \big(2y\big)^{-2\Delta_i(u)} .
6621:   \labl{eq:1bulk}
6622: In \erf{eq:1bulk} the two-point block \erf{eq:2pt-block} appears,
6623: which vanishes unless $j\eq\bar\imath$ and $\Delta_i(u)\eq\Delta_j(v)$. Note 
6624: that in the Cardy case, the phase in \erf{eq:1bulk} combines with the factor 
6625: $t_i$ in \erf{eq:Cardy-1disc} to a sign $(-1)^{\Delta_i(u)-\Delta_i}$.
6626: 
6627: 
6628: \subsection{Three bulk fields on the complex plane} \label{sec:cf-3bulk}
6629: 
6630: We now turn to the correlator of three bulk fields 
6631: on the complex plane, i.e.\ $\X^g$ is $\complex \,{\cup}\, \{\infty\}$ together
6632: with three field insertions
6633:   \be
6634:   \vPhi_1^c = (i,j,\phi_1,[f_1]) \,,\qquad
6635:   \vPhi_2^c = (k,l,\phi_2,[f_2]) \,,\qquad
6636:   \vPhi_3^c = (m,n,\phi_3,[f_3]) \,.
6637:   \ee
6638: Here $i,j,k,l,m,n \iN \II$ label simple objects; $\phi_1$ is an element of
6639: $\HomAA(\SCA_i\otiP A\otim \SCA_j,A)$,
6640: and similarly one has $\phi_2\iN\HomAA(\SCA_k\otiP A\otim \SCA_l,A)$
6641: and $\phi_3\iN\HomAA(\SCA_m\otiP A\otim \SCA_n,A)$.
6642: The three fields are inserted at the positions $z_1$, $z_2$, $z_3$
6643: with local bulk coordinate germs $[f_{1,2,3}]$ 
6644: given by $f_{1,2,3}(\zeta) \eq z_{1,2,3} \,{+}\, \zeta$. 
6645: 
6646: The complex double of $\X^g$ is the extended Riemann surface $\hat \X^g$ 
6647: given by two copies of the Riemann sphere $\complex\,{\cup}\,\{\infty\}$
6648: together with six marked points, three on each connected component.
6649: Recall that we have chosen an orientation of $\X^g$ from the outset.
6650: Let $\hat \X^g_+$ be the connected component of $\hat \X^g$ that has the
6651: same orientation as $\X^g$, and $\hat \X^g_-$ be the component with
6652: opposite orientation. We have $\hat \X^g_\pm \,{\cong}\, \mathbb{P}^1$.
6653: The three marked points on the component $\hat \X^g_+$ 
6654: are $(z_1, [\varphi_1] , \RCA_i)$, $(z_2, [\varphi_2] , \RCA_k)$ and
6655: $(z_3, [\varphi_3] , \RCA_m)$, while on $\hat \X^g_-$ the marked points are
6656: $(z^*_1, [\tilde\varphi_1] , \RCA_j)$, $(z^*_2, [\tilde\varphi_2] , \RCA_l)$,
6657: $(z^*_3, [\tilde\varphi_3] , \RCA_n)$. Here $z^*\eq x\,{-}\,\ii y$ 
6658: is the point complex conjugate to $z\eq x\,{+}\,\ii y$.
6659: The local coordinates are given by $\varphi_i(\zeta) \eq z_i\,{+}\,\zeta$
6660: and $\tilde\varphi_i(\zeta) \eq z_i^*\,{+}\,\zeta$, for $i\eq1,2,3$.
6661: We select a basis $\{ \beta_{\mu\nu} \}$ 
6662: in $\calh^c(\hat \X^g)$ via the cobordism \erf{eq:3bulk-bas},
6663:   \be
6664:   \Alpha(\beta_{\mu\nu}) = Z(B(z_1,z_2,z_3)_{\mu\nu},\emptyset,\hat \X)\,, 
6665:   \labl{eq:3bulk-block}
6666: i.e.\ $\beta_{\mu\nu}$ is a product of two three-point blocks 
6667: \erf{eq:3pt-block}. Applying $\Alpha^{-1}$ to both sides
6668: of \erf{Z3bul} results in
6669:   \be
6670:   C(\X^g) = \sum_{\mu,\nu} c(\vPhi_1 \vPhi_2 \vPhi_3)_{\mu\nu}\,
6671:   \beta_{\mu\nu} \,,
6672:   \labl{eq:3bulk-corr-const}
6673: with $c(\vPhi_1 \vPhi_2 \vPhi_3)_{\mu\nu}$ given by the ribbon invariant
6674: \erf{eq:3bulk-sc-aux}.
6675: 
6676: To find the explicit functional dependence of the correlator on the insertion 
6677: points, we use that the conformal block \erf{eq:3bulk-block} is a product of 
6678: two three-point blocks. Rearranging the individual factors
6679: in a suitable way we arrive at the expression
6680:   \beaa \displaystyle
6681:   \frac{C(\X^g)(v_i,...\,,v_n)}{C(\dot\X^g)} \!\!& \dsty
6682:   = \sum_{\mu=1}^{\N ik{\bar m}} \sum_{\nu=1}^{\N jl{\bar n}} 
6683:   \frac{S_{0,0} \, c(\vPhi_1 \vPhi_2 \vPhi_3)_{\mu\nu}}{  \dim(A)}\,
6684:   B^\mu_{mki}(v_m,v_k,v_i)~B^\nu_{nlj}(v_n,v_l,v_j) 
6685:   \\{}\\[-.7em] & \displaystyle~~ \times\,
6686:   |z_3-z_2|^{w_1-w_2-w_3}\, |z_3-z_1|^{w_2-w_1-w_3}\, |z_2-z_1|^{w_3-w_1-w_2}
6687:   \\[7pt]  & \displaystyle~~ \times\,
6688:   \exp\!\big(\, \ii \varphi_{32} (s_1{-}s_2{-}s_3) +
6689:     \ii \varphi_{31} (s_2{-}s_1{-}s_3) +
6690:     \ii \varphi_{21} (s_3{-}s_1{-}s_2) \,\big)
6691:   \eear\labl{eq:3bulk}
6692: for the correlator ratio for three bulk fields.
6693: Here $\varphi_{ij} \eq {\rm arg}(z_i{-}z_j)$, and we abbreviated
6694:   \be\begin{array}{llll}
6695:   w_1 \!\!\!&= \Delta_i(v_i)+\Delta_j(v_j) \,,\qquad
6696:   w_2 \!\!\!&= \Delta_k(v_k)+\Delta_l(v_l) \,,\qquad
6697:   w_3 \!\!\!&= \Delta_m(v_m)+\Delta_n(v_n) \,,~~ \\[7pt]
6698:   s_1 \!\!\!&= \Delta_i(v_i)-\Delta_j(v_j) \,,\qquad
6699:   s_2 \!\!\!&= \Delta_k(v_k)-\Delta_l(v_l) \,,\qquad
6700:   s_3 \!\!\!&= \Delta_m(v_m)-\Delta_n(v_n) \,.
6701:   \eear\ee
6702: The function \erf{eq:3bulk} is a continuous (but not analytic)
6703: function of the insertion points $z_1$, $z_2$, $z_3$. It does
6704: not have branch cuts, because $s_{1,2,3} \iN \zet$ which, in turn,
6705: follows from the fact that a bulk field transforming in the
6706: chiral/antichiral representation $i,j$ exists only if
6707: $Z(A)_{ij} \,{\neq}\, 0$. The matrix $Z(A)_{ij}$ commutes with
6708: the matrix $\hat T_{kl} \eq \delta_{k,l} \theta_k^{-1}$ (this is part
6709: of modular invariance, see theorem I:5.1), implying that
6710: $Z(A)_{ij}\eq (\theta_i / \theta_j)\, Z(A)_{ij}$, i.e.\ 
6711: $\theta_i\eq\theta_j$ for $Z(A)_{ij} \,{\neq}\, 0$.
6712: 
6713: 
6714: \subsection{Three defect fields on the complex plane}
6715: \label{sec:cf-3defect}
6716: 
6717: As already seen in section \ref{sec:rc-3defect}, the
6718: calculation for three defect fields differs only slightly from the one
6719: for three bulk fields. $\X^g$ is now the complex plane with field insertions
6720:   \bea
6721:   \Thetha^c_1 = (X,\orr_2,Y,\orr_2,i,j,\vartheta_1,[f_1])\,,
6722:   \qquad
6723:   \Thetha^c_2 = (Y,\orr_2,Z,\orr_2,k,l,\vartheta_2,[f_2])\,,
6724:   \\{}\\[-.7em]
6725:   \Thetha^c_3 = (Z,\orr_2,X,\orr_2,m,n,\vartheta_3,[f_3])\,.
6726:   \eear\ee
6727: Here $\orr_2$ is the standard orientation of the complex plane, $i,j,k,l,m,n 
6728: \iN \II$ label simple objects, and the morphisms $\vartheta_{1,2,3}$ are elements 
6729: of the relevant spaces of bimodule morphisms:
6730:   \bea
6731:   \vartheta_1 \in \HomAA(\SCA_i\otiP X\otim \SCA_j,Y) \,,
6732:   \qquad
6733:   \vartheta_2 \in \HomAA(\SCA_k\otip Y\otim \SCA_l,Z) \,,~~
6734:   \\{}\\[-.7em]
6735:   \vartheta_3 \in \HomAA(\SCA_m\otiP Z\otim \SCA_n,X) \,.
6736:   \eear\ee
6737: The three fields are inserted at $z_1$, $z_2$ and $z_3$, with local bulk 
6738: coordinate germs $[f_{1,2,3}]$ given by $f_{1,2,3}(\zeta) \eq 
6739: z_{1,2,3}\,{+}\,\zeta$. 
6740: 
6741: Here it is important to note that in \erf{eq:3defect-geom}
6742: we have chosen the defect circle to run parallel to the
6743: real axis in a neighbourhood of the defect field insertions
6744: $z_i$. Otherwise the local bulk coordinate germs
6745: $[f_i]$ must be modified so as to assure the condition that
6746: $f_i(x)$ lies on the defect if $x$ lies on the real axis.
6747: We further take $|z_3|\,{>}\,|z_2|\,{>}\,|z_1|$ as indicated in figure
6748: \erf{eq:3defect-geom}.
6749: 
6750: The complex double of $\X^g$ is the same as for the case of three
6751: bulk fields. Correspondingly the conformal blocks needed to express the 
6752: correlator \erf{eq:3defect-geom} are the same, too, and we can expand
6753:   \be
6754:   C(\X^g) = \sum_{\mu,\nu} c(X,\Thetha_1,Y,\Thetha_2,Z,\Thetha_3,X)_{\mu\nu}\,
6755:   \beta_{\mu\nu} \,,
6756:   \labl{eq:3defect-corr-const}
6757: with blocks $\beta_{\mu\nu}$ as defined in \erf{eq:3bulk-block}.
6758: The ribbon invariant for the coefficient is given by \erf{eq:3defect-rib},
6759: as calculated in section \ref{sec:rc-3defect}.
6760: In the same manner as we arrived at the explicit form of the correlator 
6761: \erf{eq:3bulk} of three bulk fields, for three defect fields we then find
6762:   \beaa\dsty
6763:   \frac{C(\X^g)(v_i,{...},v_n)}{C(\dot\X^g)} \!\!\!&\dsty
6764:   = \sum_{\mu=1}^{\N ik{\bar m}} \sum_{\nu=1}^{\N jl{\bar n}}
6765:   \frac{S_{0,0} \, c(X,\Thetha_1,Y,\Thetha_2,Z,\Thetha_3, X)_{\!\mu\nu\,}^{}}
6766:      { \dim(A) } 
6767:   B^\mu_{mki}(v_m,v_k,v_i)\, B^\nu_{nlj}(v_n,v_l,v_j)  
6768:   \\{}\\[-.6em] &\displaystyle~~\times\,
6769:   |z_3{-}z_2|^{w_1-w_2-w_3}\, |z_3{-}z_1|^{w_2-w_1-w_3}\,
6770:   |z_2{-}z_1|^{w_3-w_1-w_2}
6771:   \\[7pt]&  \displaystyle~~\times
6772:   \exp\!\big(\, \ii \varphi_{32} (s_1{-}s_2{-}s_3) +
6773:     \ii \varphi_{31} (s_2{-}s_1{-}s_3) +
6774:     \ii \varphi_{21} (s_3{-}s_1{-}s_2) \,\big) \,,
6775:   \\[-1.7em]{}
6776:   \eear\labl{eq:3defect}
6777: where $\varphi_{ij}\eq {\rm arg}(z_i{-}z_j)$ and
6778:   \be\begin{array}{llll}
6779:   w_1 \!\!&= \Delta_i(v_i)+\Delta_j(v_j) \,,\quad
6780:   w_2 \!\!&= \Delta_k(v_k)+\Delta_l(v_l) \,,\quad
6781:   w_3 \!\!&= \Delta_m(v_m)+\Delta_n(v_n) \,,\\[7pt]
6782:   s_1 \!\!&= \Delta_i(v_i)-\Delta_j(v_j) \,,\quad
6783:   s_2 \!\!&= \Delta_k(v_k)-\Delta_l(v_l) \,,\quad
6784:   s_3 \!\!&= \Delta_m(v_m)-\Delta_n(v_n) \,.
6785:   \eear\ee
6786: In contrast to the correlator of three bulk fields,
6787: the expression \erf{eq:3defect} is {\em not\/} necessarily
6788: single valued in the coordinates $z_1 , z_2 , z_3$, since the numbers
6789: $s_k$ do not have to be integers. This is to be expected. Indeed,
6790: if, for example, in the setup \erf{eq:3defect-geom} we take
6791: the field at $z_2$ around the field at $z_1$ we must deform
6792: the defect circle in order to avoid taking the field $\Thetha^c_2$ across
6793: the defect. In this way we end up with an arrangement of
6794: defect lines different from the one we started with.
6795: 
6796: 
6797: \subsection{One bulk field on the cross cap}
6798: \label{sec:cf-1bulk-xcap}
6799: 
6800: Next we consider the correlator of one bulk field on the cross cap. This 
6801: surface is non-orientable; accordingly we take $A$ to be a Jandl algebra. 
6802: Analogously as in section \ref{sec:rc-1bulk-xcap} the world sheet
6803: $\X^g$ is given by $\complex / \sigrp$ with $\sigrp$ the anti-holomorphic
6804: involution $\sigrp(\zeta) \eq {-}1/{\zeta^*}$. The bulk insertion is
6805: $\vPhi^c \eq (i,j,\phi,[f])$ with $f(\zeta) \eq z \,{+}\, \zeta$,
6806: $i,j\iN\II$, and $\phi\iN\HomAA(\SCA_i\otiP A\otim \SCA_j,A)$.
6807:  %
6808: The complex double of $\X^g$ is the Riemann sphere
6809: $\complex \cup \{\infty\}$ together with the two marked points
6810: $(z,[\varphi],\RCA_i)$ and $(\sigrp(z),[\tilde\varphi],\RCA_j)$.
6811: The holomorphic coordinate germs are given by
6812: $\varphi(\zeta) \eq z\,{+}\,\zeta$ and 
6813: $\tilde\varphi(\zeta) \eq \sigrp(\varphi(\zeta^*)) \eq {-}(z^*{+}\zeta)^{-1}$.
6814: 
6815: The space $\calh^c(\hat \X^g)$ is the space of two-point blocks
6816: on the Riemann sphere and hence one-dimensional, provided
6817: $j\eq\bar\imath$. It is spanned by the block
6818:   \be
6819:   \beta_2 = \langle 0 |  \,
6820:   \VOs^0_{i,\bar\imath}
6821:     \!\big(R\big(\kappa_0(\varphi)\big)^{-1} v_i \,;\, z\big)\,
6822:   \VOs^{\bar\imath}_{\bar\imath,0}
6823:     \!\big(R\big(\kappa_0(\tilde\varphi)\big)^{-1} 
6824:     v_{\bar\imath} \,;\,-\Frac1{z^*}\big)\,
6825:   |0\rangle\,, \labl{eq:xcap-block}
6826: where we also made explicit the dependence on the local
6827: coordinates as in \erf{eq:H0n-conf-block}. In the correlators treated in the 
6828: previous sections, all operators $R(\,\cdots)^{-1}$ were just the identity,
6829: owing to the local coordinates taking the simple form 
6830: $\zeta\,{ \mapsto}\, p{+}\zeta$. In the present case this is only true
6831: for $R\big(\kappa_0(\varphi)\big)^{-1}$. The correlator is thus given by
6832:   \be
6833:   C(\X^g)(v_i,v_{\bar\imath}) = c(\vPhi^c) \, \langle 0 |\,
6834:   \VOs^0_{i,\bar\imath} \!\big( v_i \,;\, z\big)\,
6835:   \VOs^{\bar\imath}_{\bar\imath,0}
6836:     \!\big(R\big(\kappa_0(\tilde\varphi)\big)^{-1} 
6837:     v_{\bar\imath} \,;\,-\Frac1{z^*}\big) 
6838:   |0\rangle \,. \labl{eq:xcap-corr-pre}
6839: The constant $c(\vPhi^c)$ is given by the ribbon invariant \erf{eq:xcap-cobord},
6840: as follows from applying $\Alpha$ to both sides of \erf{eq:xcap-corr-pre}
6841: and comparing with \erf{eq:xcap-1bulk-expand}. This result also uses the
6842: equality $\Alpha(\beta_2) \eq Z(B_2,\emptyset,\hat \X)1$, with $B_2$ given by
6843: \erf{eq:xcap-aux-1}.
6844: 
6845: We conclude that the correlator ratio for one bulk field on the cross cap is
6846:   \be
6847:   \frac{C(\X^g)(v_i,v_{\bar\imath})}{C(\dot\X^g)} 
6848:   = \frac{S_{0,0}\,c(\vPhi^c)}{
6849:   {\Gamma^{\phantom I\!\!\!}}^\sigma(\one,\eps)}\,
6850:   B_{i\bar\imath}(v_i, R\big(\kappa_0(\tilde\varphi)\big)^{-1}
6851:     v_{\bar\imath})\, \big(z\,{+}\,\Frac1{z^*}\big)^{-2\Delta_i(v_i)} .
6852:   \ee
6853: Here $v_i \oti v_{\bar\imath} \iN \SCA_i \oti \SCA_{\bar\imath}$ specifies the 
6854: state representing the bulk field. 
6855: Note that for the identity bulk insertion, in \erf{chatphic} we have
6856: $\phi'\eq\eps$, the counit of $A$, so that $C(\dot\X^g)\eq
6857: {\Gamma^{\phantom I\!\!\!}}^\sigma(\one,\eps) / S_{0,0}$. Using
6858: \erf{eq:P-def} and (II:3.90) leads to the expression
6859:   \be
6860:   \Gamma^\sigma(\one,\eps) = \sum_{a \In A} \frac{t_a P_{0,a}}{\sigma(a)} \,.
6861:   \ee
6862: Suppose now further that $v_{\bar\imath}$ is a Virasoro-primary state. Then
6863: $R\big(\kappa_0(\tilde\varphi)\big)^{-1} v_{\bar\imath} \eq (z^*)^{-2\Delta_
6864: {\bar\imath}(v_{\bar\imath})}v_{\bar\imath}$, so that in this case
6865:   \be
6866:   \frac{C(\X^g)(v_i,v_{\bar\imath})}{C(\dot\X^g)} 
6867:   = \frac{S_{0,0}\,c(\vPhi^c)}{{\Gamma^{\phantom I\!\!\!}}^\sigma(\one,\eps)}\,
6868:   B_{i\bar\imath}(v_i,v_{\bar\imath})\,\big(1\,{+}\,|z|^2\big)^{-2\Delta_i(v_i)}
6869:   .  \labl{eq:1B-xcap}
6870: The constant $c(\vPhi^c)$ is evaluated in \erf{chatphic} and (II:3.110);
6871: in the Cardy case it takes the from \erf{eq:cardy-1bulk-xcap}.
6872: 
6873: 
6874: \subsection{One defect field on the cross cap}
6875: \label{sec:cf-1defect-xcap}
6876: 
6877: To obtain the correlator of a defect field on the cross cap one 
6878: essentially repeats the calculation in the previous section. Again,
6879: $\X^g \eq \complex / \sigrp$, this time with the insertion of a defect
6880: field, given by $\Thetha^c \eq (X,-\orr_2,X,\orr_2,i,j,\vartheta,[f])$ with 
6881: $f(\zeta) \eq z \,{+}\, \zeta$ and $\vartheta \iN \HomAA(\SCA_i\otiP X^s 
6882: \otim \SCA_j,X)$. The choice of $f$ implies that the defect runs parallel 
6883: to the real axis at the insertion point of the defect field.
6884: The orientations $\pm\orr_2$ of the neighbourhood of $X$
6885: are obtained as follows. Let $\orr_2$ be the local orientation around
6886: $f(0)$ induced by $[f]$. The orientation of $X$ is obtained by transporting
6887: $\orr_2$ along the defect to the right. When passing through the identification
6888: circle, the orientation gets reversed, so that one arrives at the
6889: insertion point $f(0)$ with $-\orr_2$ from the left, see figure \Erp 112.
6890: 
6891: The space $\calh^c(\hat \X^g)$ is again spanned by the single
6892: block \erf{eq:xcap-block}; the correlator is 
6893:   \be
6894:   C(\X^g) = c(X,\Thetha^c)\, 
6895:   \langle 0 |  \VOs^0_{i,\bar\imath\,} \!\big( v_i \,; z\big)
6896:   \VOs^{\bar\imath}_{\bar\imath,0\,}
6897:     \!\big(R\big(\kappa_0(\tilde\varphi)\big)^{-1} 
6898:     v_{\bar\imath} \,;\,-\Frac1{z^*}\big) |0\rangle \,,
6899:   \labl{eq:xcap-defect-corr-pre}
6900: where $c(X,\Thetha^c)$ is given by the ribbon invariant \erf{eq:xcap-def-inv}, 
6901: as calculated in section \ref{sec:rc-1defect-xcap}.
6902: Taking the states $v_i\iN\SCA_i$ and $v_{\bar\imath}\iN\SCA_{\bar\imath}$
6903: describing the defect field to be Virasoro-primaries, altogether the ratio 
6904: of correlators for one defect field on the cross cap becomes
6905:   \be
6906:   \frac{C(\X^g)(v_i,v_{\bar\imath})}{C(\dot\X^g)} 
6907:   = \frac{S_{0,0} \, c(X,\Thetha^c)}{\Gamma^\sigma_{}(\one,\eps)}\,
6908:   B_{i\bar\imath}(v_i,v_{\bar\imath}) \,
6909:   \big(1\,{+}\,|z|^2\big)^{-2\Delta_i(v_i)} .
6910:   \labl{eq:1D-xcap}
6911: In the Cardy case, the constant
6912: $c(X,\Thetha^c)$ is given by \erf{eq:1def-Xcap-bas-Cardy}.
6913:  
6914: 
6915:  \newpage
6916: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6917: 
6918: \newcommand\wb{\,\linebreak[0]} \def\wB {$\,$\wb}
6919:  \newcommand\Bi       {\bibitem}
6920:  \renewcommand\J[5]     {{\sl #5\/}, {#1} {#2} ({#3}) {#4} }
6921:  \renewcommand\K[6]     {{\sl #6\/}, {#1} {#2} ({#3}) {#4}~\,[#5]}
6922:  \newcommand\PhD[2]   {{\sl #2\/}, Ph.D.\ thesis (#1)}
6923:  \newcommand\Mast[2]  {{\sl #2\/}, Master's thesis (#1)}
6924:  \newcommand\Prep[2]  {{\sl #2\/}, pre\-print {#1}}
6925:  \newcommand\BOOK[4]  {{\sl #1\/} ({#2}, {#3} {#4})}
6926:  \newcommand\inBO[7]  {{\sl #7\/}, in:\ {\sl #1}, {#2}\ ({#3}, {#4} {#5}), p.\ {#6}}
6927:  \newcommand\iNBO[7]  {{\sl #7\/}, in:\ {\sl #1} ({#3}, {#4} {#5}) }
6928:  \newcommand\Erra[3]  {\,[{\em ibid.}\ {#1} ({#2}) {#3}, {\em Erratum}]}
6929:  \def\jf    {J.\ Fuchs}
6930:  \def\dim   {dimension}  
6931:  \def\adma  {Adv.\wb Math.}
6932:  \def\anop  {Ann.\wb Phys.}
6933:  \def\aspm  {Adv.\wb Stu\-dies\wB in\wB Pure\wB Math.}
6934:  \def\coia  {Com\-mun.\wB in\wB Algebra}
6935:  \def\comp  {Com\-mun.\wb Math.\wb Phys.}
6936:  \def\cpma  {Com\-pos.\wb Math.}
6937:  \def\fiic  {Fields\wB Institute\wB Commun.}
6938:  \def\foph  {Fortschritte\wB d.\wb Phys.}
6939:  \def\ijmp  {Int.\wb J.\wb Mod.\wb Phys.\ A}
6940:  \def\jams  {J.\wb Amer.\wb Math.\wb Soc.}
6941:  \def\jgap  {J.\wb Geom.\wB and\wB Phys.}
6942:  \def\jhep  {J.\wb High\wB Energy\wB Phys.}
6943:  \def\jomp  {J.\wb Math.\wb Phys.}
6944:  \def\jopa  {J.\wb Phys.\ A}
6945:  \def\josp  {J.\wb Stat.\wb Phys.}
6946:  \def\lemp  {Lett.\wb Math.\wb Phys.} 
6947:  \def\maan  {Math.\wb Annal.}
6948:  \def\mpla  {Mod.\wb Phys.\wb Lett.\ A}
6949:  \newcommand\nqma[2] {\inBO{Non-perturbative QFT Methods and Their Applications}
6950:             {Z.\ Horv\'ath and L.\ Palla, eds.} \WS\Si{2001} {{#1}}{{#2}} }
6951:  \newcommand\nqft[2] {\inBO{Nonperturbative Quantum Field Theory}
6952:             {G.\ 't Hooft et al., eds.} \PL\NY{1988} {{#1}}{{#2}} }
6953:  \def\nuci  {Nuovo\wB Cim.}
6954:  \def\nupb  {Nucl.\wb Phys.\ B}
6955:  \newcommand\phgt[2] {\inBO{Physics, Geometry, and Topology}
6956:             {H.C.\ Lee, ed.} \PL\NY{1990} {{#1}}{{#2}} }
6957:  \def\phla  {Phys.\wb Lett.\ A}
6958:  \def\phlb  {Phys.\wb Lett.\ B}
6959:  \def\phrl  {Phys.\wb Rev.\wb Lett.}
6960:  \def\phrp  {Phys.\wb Rep.}
6961:  \def\phrd  {Phys.\wb Rev.\ D}
6962:  \def\phre  {Phys.\wb Rev.\ E}
6963:  \def\pnas  {Proc.\wb Natl.\wb Acad.\wb Sci.\wb USA}
6964:  \def\prtp  {Progr.\wb Theor.\wb Phys.}
6965:  \newcommand\qfsm[2] {\inBO{\Q Fields and Strings: A Course for Mathematicians}
6966:             {P.\ Deligne et al., eds.} \AMS\PR{1999} {{#1}}{{#2}} }
6967:  \def\rvmp  {Rev.\wb Math.\wb Phys.}
6968:  \def\sebo  {S\'emi\-naire\wB Bour\-baki}
6969:  \def\sjnp  {Sov.\wb J.\wb Nucl.\wb Phys.}
6970:  \def\trgr  {Trans\-form.\wB Groups}
6971:  \def\zfpc  {Z.\wb Phy\-sik C}
6972:  \def\AMS    {{American Mathematical Society}}
6973:  \def\BIR    {{Birk\-h\"au\-ser}}
6974:  \def\PL     {{Plenum Press}}
6975:  \def\SV     {{Sprin\-ger Ver\-lag}}
6976:  \def\WS     {{World Scientific}}
6977:  \def\Bo     {{Boston}} 
6978:  \def\PR     {{Providence}}
6979:  \def\Si     {{Singapore}}
6980:  \def\NY     {{New York}}
6981: 
6982: \small
6983: \begin{thebibliography}{XX} \addtolength{\itemsep}{-6pt}
6984: \bibitem[{\,I\,}]{fuRs4} {\jf, I.\ Runkel, and C.\ Schweigert,
6985:            \K\nupb{646}{2002}{353} {hep-th/0204148}
6986:            {TFT construction of RCFT correlators I: Partition functions}}
6987: \bibitem[{\,II\,}]{fuRs8} {\jf, I.\ Runkel, and C.\ Schweigert,
6988:            \K\nupb{678}{2004}{511} {hep-th/0306164}
6989:            {TFT construction of RCFT correlators II: Unoriented surfaces}}
6990: \bibitem[{\,III\,}]{fuRs9} {\jf, I.\ Runkel, and C.\ Schweigert,
6991:            \K\nupb{694}{2004}{277} {hep-th/0403158}
6992:            {TFT construction of RCFT correlators III: Simple currents}} 
6993: \bibitem[{\,V\,}]{fjfrs} {J.\ Fjelstad, \jf, I.\ Runkel, and C.\ Schweigert,
6994:            \Prep{hep-th/0503194}
6995:            {TFT construction of RCFT correlators V: Proof of modular
6996:            invariance and factorization}}
6997: \Bi{fuRs}  {\jf, I.\ Runkel, and C.\ Schweigert, \K\nupb{624}{2002}{452}
6998:            {hep-th/0110133} {\Con \corfu s, \frob s and triangulations}}
6999: \Bi{witt27}{E.\ Witten, \J\comp{121}{1989}{351} 
7000:            {Quantum field theory and the Jones polynomial}}
7001: \Bi{frki2} {J.\ Fr\"ohlich and C.\ King, \J\comp{126}{1989}{167}
7002:            {The Chern\hy Si\-mons theory and knot polynomials}}
7003: \Bi{fffs3} {G.\ Felder, J.\ Fr\"ohlich, \jf, and C.\ Schweigert,
7004:            \K\cpma{131}{2002}{189} {hep-th/9912239}
7005:            {\Corfu s and \bc s in RCFT and three-\dim al topology}}
7006: \Bi{sono2} {H.\ Sonoda, \J\nupb{311}{1988}{401} {Sewing \cfts}}
7007: \Bi{lewe3} {D.C.\ Lewellen, \J\nupb{372}{1992}{654}
7008:            {Sewing constraints for \cfts\ on surfaces with boundaries}}
7009: \Bi{fips}  {D.\ Fioravanti, G.\ Pradisi, and A.\ Sagnotti, \K\phlb{321}{1994}{349}
7010:            {hep-th/9311183} {Sewing constraints and non-orientable strings}}
7011: \Bi{witt39}{E.\ Witten, \J\comp{144}{1992}{189}
7012:            {On holomorphic factorization of WZW and coset models}}
7013: \Bi{Mose}  {G.\ Moore and N.\ Seiberg, \phgt{263} {Lectures on RCFT}}
7014: \Bi{frki}  {J.\ Fr\"ohlich and C.\ King,
7015:            \J\ijmp4{1989}{5321} {Two-\dim al \cft\ and three-\dim al topology}}
7016: \Bi{bakI}  {B.\ Bakalov and A.N.\ Kirillov, \K\trgr5{2000}{207}
7017:            {math.GT/9809057} {On the Lego-Teichm\"uller game}}
7018: \Bi{dofa2} {Vl.S.\ Dotsenko and V.A.\ Fateev,
7019:            \J\nupb{251}{1985}{691} {Four-point \corfu s
7020:            and \oa\ in 2D \con invariant theories with central charge $c\le1$}}
7021: \Bi{cale}  {J.L.\ Cardy and D.C.\ Lewellen,
7022:            \J\phlb{259}{1991}{274} {Bulk and boundary operators in \cft}}  
7023: \Bi{runk}  {I.\ Runkel, \K\nupb{549}{1999}{563} {hep-th/9811178}
7024:            {Boundary structure constants for the A-series Virasoro \mimo s}}
7025: \Bi{pons3} {B.\ Ponsot,
7026:            \Prep{hep-th/0301193} {Recent progress on Liouville \ft}}
7027: \Bi{tesc12}{J.A.\ Teschner,
7028:            \Prep{hep-th/0303150} {A lecture on the Liouville \vop s}}
7029: \Bi{naKA'} {Y.\ Nakayama, \Mast{Tokyo 2004, hep-th/0402009}
7030:            {Liouville \ft\ -- A decade after the revolution}}
7031: \Bi{huan21}{Y.-Z.\ Huang, \K\pnas{102}{2005}{5352} {math.QA/0412261} 
7032:            {Vertex \oa s, the Verlinde conjecture and modular \tcs}}
7033: \Bi{TUra}  {V.G.\ Turaev, \BOOK{\Q
7034:            Invariants of Knots and $3$-Manifolds} {de Gruyter}\NY{1994}}
7035: \Bi{BAki}  {B.\ Bakalov and A.A.\ Kirillov, \BOOK{Lectures on
7036:            Tensor Categories and Modular Functors} \AMS\PR{2001}}
7037: \Bi{caiz2} {A.\ Cappelli, C.\ Itzykson, and J.-B.\ Zuber, \J\comp{113}{1987}1
7038:            {The A-D-E classification of minimal and $A_1^{(1)}$ conformal
7039:            invariant theories}}
7040: \Bi{bppz2} {R.E.\ Behrend, P.A.\ Pearce, V.B.\ Petkova, and J.-B.\ Zuber,
7041:            \K\nupb{579}{2000}{707} {hep-th/9908036} {\Bc s in rational \cfts}}
7042: \Bi{pezu6} {V.B.\ Petkova and J.-B.\ Zuber, \K\nupb{603}{2001}{449}
7043:            {hep-th/0101151} {The many faces of Ocneanu cells}}
7044: \Bi{pezu7} {V.B.\ Petkova and J.-B.\ Zuber, \nqma{1~\,[hep-th/0103007]}
7045:            {\Con \bc s and what they teach us}}
7046: \Bi{pezu8} {V.B.\ Petkova and J.-B.\ Zuber,
7047:            \inBO{MathPhys Odyssey 2001 -- Integrable Models and Beyond}
7048:            {M.\ Kashiwara and T.\ Miwa, eds.} \BIR\Bo{2002}
7049:            {415~\,[hep-th/0108236]} {\Cfts, graphs and \q \alg s}}
7050: \Bi{ostr}  {V.\ Ostrik, \K\trgr8{2003}{177} {math.QA/0111139}
7051:            {Module \cats, weak \hopf s and modular invariants}}
7052: \Bi{ffrs4} {J.\ Fr\"ohlich, \jf, I.\ Runkel, and C.\ Schweigert, \K\foph{52}
7053:            {2004}{672} {math.CT/0411507} {Picard groups in rational \cft}}
7054: \Bi{pare11}{B.\ Pareigis, \J{Publ.\wb Math.\wB Debrecen}{24}{1977}{189}
7055:            {Non-additive ring and module theory I.\ General theory of monoids}}
7056: \Bi{pare16}{B.\ Pareigis, \J\coia9{1981}{1455}
7057:            {Morita equivalence of module \cats\ with tensor products}}
7058: \Bi{beRn}  {J.\ Bernstein, \Prep{q-alg/9501032 (Sackler Lectures)}
7059:            {Tensor \cats}}
7060: \Bi{ffrs}  {J.\ Fr\"ohlich, \jf, I.\ Runkel, and C.\ Schweigert,
7061:            \Prep{math.CT/0309465, \adma, in press}
7062:            {Correspondences of ribbon \cats}}
7063: \Bi{fuSc16}{\jf\ and C.\ Schweigert, \K\fiic{39}{2003}{25}
7064:            {math.CT/0106050} {Category theory for \con \bc s}}
7065: \Bi{kios}  {A.A.\ Kirillov and V.\ Ostrik, \K\adma{171}{2002}{183}
7066:            {math.QA/0101219} {On q-analog of McKay correspondence
7067:            and ADE classification of $\widehat{\mathfrak{sl}}(2)$ \cfts}}
7068: \Bi{prss3} {G.\ Pradisi, A.\ Sagnotti, and Ya.S.\ Stanev,
7069:            \K\phlb{381}{1996}{97} {hep-th/9603097}
7070:            {Completeness conditions for boundary operators in 2D \cft}}
7071: \Bi{gann17}{T.\ Gannon, \K\nupb{627}{2002}{506}
7072:            {hep-th/0106105} {Boundary \cft\ and fusion ring \rep s}}
7073: \Bi{ocne10}{A.\ Ocneanu, \J\aspm{31}{2001}{235} {Operator \alg s, topology
7074:            and subgroups of \q \sym\ -- construction of subgroups of \qg s}}
7075: \Bi{bosz}  {G.\ B\"ohm and K.\ Szlach\'anyi, \K\lemp{38}{1996}{437}{q-alg/9509008}
7076:            {A coassociative $C^*$-quantum group with non-integral \dim s}}
7077: \Bi{fffs}  {G.\ Felder, J.\ Fr\"ohlich, \jf, and C.\ Schweigert,
7078:            \K\jgap{34}{2000}{162} {hep-th/9909030} {The geometry of WZW branes}}
7079: \Bi{bppz}  {R.E.\ Behrend, P.A.\ Pearce, V.B.\ Petkova, and J.-B.\ Zuber,
7080:            \K\phlb{444}{1998}{163}
7081:            {hep-th/9809097} {On the classification of bulk and boundary \cfts}}
7082: \Bi{pezu5} {V.\ Petkova and J.-B.\ Zuber, \K\phlb{504}{2001}{157}
7083:            {hep-th/0011021} {Generalized twisted partition functions}}
7084: \Bi{dofa}  {Vl.S.\ Dotsenko and V.A.\ Fateev, \J\nupb{240}{1984}{312}
7085:            {\Con \alg\ and multipoint \corfu s in 2D statistical models}}
7086: \Bi{dujn}  {B.\ Durhuus, H.P.\ Jakobsen, and R.\ Nest, \J\rvmp5{1993}1
7087:            {Topological quantum field theories from generalized $6j$-symbols}}
7088: \Bi{lore}  {R.\ Longo and K.-H.\ Rehren,
7089:            \K\rvmp7{1995}{567} {hep-th/9411077} {Nets of subfactors}}
7090: \Bi{bepz}  {A.A.\ Belavin, A.M.\ Polyakov, and A.B.\ Zamolodchikov,
7091:            \J\nupb{241}{1984}{333} {Infinite \con \sym\ in \twodim\ \qft}}
7092: \Bi{sono3} {H.\ Sonoda, \J\nupb{311}{1988}{417} {Sewing \cfts\ 2}}
7093: \Bi{fffs2} {G.\ Felder, J.\ Fr\"ohlich, \jf, and C.\ Schweigert, \K\phrl{84}
7094:            {2000}{1659} {hep-th/9909140} {\Con \bc s and three-\dim al \tft}}
7095: \Bi{runk2} {I.\ Runkel, \K\nupb{579}{2000}{561} {hep-th/9908046}
7096:            {Structure constants for the D-series Virasoro \mimo s}}
7097: \Bi{brSc2} {I.\ Brunner and V.\ Schomerus, \K\jhep{0010}{2000}{016}
7098:            {hep-th/0008194} {On superpotentials for D-branes in \gemo s}}
7099: \Bi{fgrs}  {J.\ Fr\"ohlich, O.\ Grandjean, A.\ Recknagel, and  V.\ Schomerus,
7100:            \K\nupb{583}{2000}{381}
7101:            {hep-th/9912079} {Fundamental strings in Dp-Dq brane systems}}
7102: \Bi{card9} {J.L.\ Cardy, \J\nupb{324}{1989}{581}
7103:            {Boundary conditions, fusion rules and the Verlinde formula}}
7104: \Bi{dofa3} {Vl.S.\ Dotsenko and V.A.\ Fateev, \J\phlb{154}{1985}{291} {Operator
7105:            \alg\ of \twodim\ \con\ theories with central charge $c\le1$}}
7106: \Bi{knza}  {V.G.\ Knizhnik and A.B.\ Zamolodchikov,
7107:            \J\nupb{247}{1984}{83} {Current \alg\ and \WZ model in two \dim s}}
7108: \Bi{zafa2} {A.B.\ Zamolodchikov and V.A.\ Fateev,
7109:            \J\sjnp{43}{1986}{657} {Operator \alg\ and
7110:            \corfu s in the \twodim\ SU$(2)\times$SU$(2)$ chiral \WZ model}}
7111: \Bi{chfl}  {P.\ Christe and R.\ Flume, \J\nupb{282}{1987}{466}
7112:            {The four point correlations of all primary operators of the
7113:            $d\eq2$ con\-formally invariant SU(2) $\sigma$-model with \WZ term}}
7114: \Bi{musS2} {G.\ Mussardo, G.\ Sotkov, and M.\ Stanishkov,
7115:            \J\nupb{305}{1988}{69} {Fine structure of the \susic \ope\ \alg s}}
7116: \Bi{kikkmo}{Y.\ Kitazawa, N.\ Ishibashi, A.\ Kato, K.\ Kobayashi, Y.\ Matsuo,
7117:            and S.\ Okade, \J\nupb{306}{1988}{425} {Operator product expansion
7118:            coefficients in $N=1$ super\cft\ and slightly relevant perturbation}}
7119: \Bi{dmfs}  {L.J.\ Dixon, E.J.\ Martinec, D.H.\ Friedan, and S.H.\ Shenker,
7120:            \J\nupb{282}{1987}{13} {The \cft\ of orbifolds}}
7121: \Bi{jf12}  {\jf, \J\phrl{62}{1989}{1705}
7122:            {Operator product coefficients in nondiagonal \cfts}}
7123: \Bi{fukl}  {\jf\ and A.\ Klemm, \J\anop{194}{1989}{303}
7124:            {The computation of the operator \alg\ in nondiagonal \cfts}}
7125: \Bi{petk2} {V.B.\ Petkova, \J\phlb{225}{1989}{357}
7126:            {Structure constants of the $(A,D)$ minimal $c<1$ \con models}}
7127: \Bi{dotr}  {M.\ Douglas and S.\ Trivedi, \J\nupb{320}{1989}{461}
7128:            {Operator product coefficients in nonstandard SU$(2)$ \WZW models}}
7129: \Bi{kaki}  {A.\ Kato and Y.\ Kitazawa, \J\nupb{319}{1989}{474}
7130:            {E$_7$ type \modinv t \WZ\ theory and Gepner's string \compac}}
7131: \Bi{fuks}  {\jf, A.\ Klemm, and C.\ Scheich, \J\zfpc{46}{1990}{71}
7132:            {The operator \alg\ of the $E_8$ type\/ {\rm su(2)} \wzwt}}
7133: \Bi{rest}  {K.-H.\ Rehren, Ya.S.\ Stanev, and I.T.\ Todorov,
7134:            \K\comp{174}{1995-6}{605} {hep-th/9409165}
7135:            {Characterizing invariants for local extensions of \cua s}}
7136: \Bi{pezu}  {V.B.\ Petkova and J.-B.\ Zuber, \K\nupb{438}{1995}{347}
7137:            {hep-th/9410209} {On structure constants of $sl(2)$ theories}}
7138: \Bi{clny2} {C.G.\ Callan, C.\ Lovelace, C.R.\ Nappi, and S.A.\ Yost,
7139:            \J\nupb{293}{1987}{83} {Adding holes and crosscaps to the superstring}}
7140: \Bi{poca}  {J.\ Polchinski and Y.\ Cai,
7141:            \J\nupb{296}{1988}{91} {Consistency of open super\stts}}
7142: \Bi{ishi}  {N.\ Ishibashi,
7143:            \J\mpla4{1989}{251} {The boundary and crosscap states in \cfts}}
7144: \Bi{prss}  {G.\ Pradisi, A.\ Sagnotti, and Ya.S.\ Stanev, \K\phlb{354}{1995}
7145:            {279} {hep-th/9503207} {Planar duality in $SU(2)$ \wzwm s}}
7146: \Bi{prss2} {G.\ Pradisi, A.\ Sagnotti, and Ya.S.\ Stanev,
7147:            \K\phlb{356}{1995}{230} {hep-th/9506014}
7148:            {The open descendants of non-diagonal $SU(2)$ \wzwm s}}
7149: \Bi{huss}  {L.R.\ Huiszoon, A.N.\ Schellekens, and N.\ Sousa, \K\phlb{470}
7150:            {1999}{95} {hep-th/9909114} {Klein bottles and simple currents}}
7151: \Bi{huss2} {L.R.\ Huiszoon, A.N.\ Schellekens, and N.\ Sousa,
7152:            \K\nupb{575}{2000}{401}
7153:            {hep-th/9911229} {Open descendants of non-diagonal invariants}}
7154: \Bi{husc}  {L.R.\ Huiszoon and A.N.\ Schellekens, \K\nupb{583}{2000}{705}
7155:            {hep-th/0004100} {Crosscaps, boundaries and T-Duality}}
7156: \Bi{fhssw} {\jf, L.R.\ Huiszoon, A.N.\ Schellekens, C.\ Schweigert, and J.\
7157:            Walcher, \K\phlb{495}{2000}{427}
7158:            {hep-th/0007174} {Boundaries, crosscaps and simple currents}}
7159: \Bi{brho}  {I.\ Brunner and K.\ Hori, \K\jhep{0407}{2004}{023}
7160:            {hep-th/0208141} {Notes on orientifolds of \rcfts}}
7161: \Bi{bhhw}  {I.\ Brunner, K.\ Hori, K.\ Hosomichi, and J.\ Walcher,
7162:            \Prep{hep-th/0401137} {Orientifolds of \gemo s}}
7163: \Bi{blwe}  {R.\ Blumenhagen and T.\ Weigand, \K\jhep{0402}{2004}{041}
7164:            {hep-th/0401148} {Chiral \susic \gemo\ orientifolds}}
7165: \Bi{dihs2} {T.P.T.\ Dijkstra, L.R.\ Huiszoon, and A.N.\ Schellekens,
7166:            \K\nupb{710}{2005}{525} {hep-th/0411129}
7167:            {Supersymmetric standard model spectra from RCFT orientifolds}}
7168: \Bi{bedu2} {Z.\ Bern and D.C.\ Dunbar, \J\ijmp5{1990}{4629}
7169:            {\Cft\ on surfaces with boundaries and nondiagonal \modinv ts}}
7170: \Bi{bisa3} {M.\ Bianchi and A.\ Sagnotti, \J\nupb{361}{1991}{519}
7171:            {Twist symmetry and open-string Wilson lines}}
7172: \Bi{HUan}  {Y.-Z.\ Huang, \BOOK{Two-\dim al \Con Geometry and Vertex 
7173:            Operator Algebras} \BIR\Bo{1997}}
7174: \Bi{FRbe}  {E.\ Frenkel and D.\ Ben-Zvi, \BOOK{Vertex Algebras and Algebraic 
7175:            Curves{\rm, second edition}}\AMS{Providence,}{in press}
7176:            \,[www.math.berkeley.edu/\~{}frenkel/BOOK] }
7177: \Bi{LEli}  {J.\ Lepowsky and H.\ Li, \BOOK{\Intro to Vertex Operator 
7178:            Algebras and their Representations} \BIR\Bo{2004}}
7179: \Bi{huan14}{Y.-Z.\ Huang, \K\fiic{39}{2003}{109}
7180:            {math.QA/0212308} {\RIsu s with boundaries and the theory of \voa s}}
7181: \Bi{hule3.5}{Y.-Z.\ Huang and J.\ Lepowsky, 
7182:            \inBO{Lie Theory and Geometry} {R.\ Brylinski, J.-L.\ Brylinski, V.\ 
7183:            Guillemin, and V.G.\ Kac, eds.} \BIR\Bo{1994} {349~\,[hep-th/9401119]}
7184:            {Tensor products of modules for a \voa\ and vertex tensor \cats}}
7185: \Bi{Scfw}  {C.\ Schweigert, \jf, and J.\ Walcher,
7186:            \nqma{37~\,[hep-th/0011109]} {\Cft, \bc s and applications to \stt}} 
7187: \Bi{frsh2} {D.H.\ Friedan and S.H.\ Shenker,
7188:            \J\nupb{281}{1987}{509} {The analytic geometry of \twodim\ \cft}}
7189: \Bi{huan24}{Y.-Z.\ Huang, \Prep{(2004)}  
7190:            {Vertex \oa s, the Verlinde conjecture and modular transformations}}
7191: \Bi{gawe13}{K.\ Gaw\c edzki, \qfsm{727} {Lectures on \cft}
7192:            \,[www.math.ias.edu/QFT/fall/index.html] }
7193: 
7194: \end{thebibliography}
7195: \end{document}
7196: 
7197: