1: %%% 

2: %%%	Paper: hep-th/ (LaTeX version)

3: %%%    

4: %%%

5: 

6: \documentclass[10pt]{article}

7: \usepackage{amssymb}

8: \usepackage{graphicx}

9: \usepackage{fancyhdr}

10: %\parindent=0pt

11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% COMMANDS %%%%%%%%%%%%%%%%%%%%%%%%%%%%

12: 

13: \newcommand{\beq}{\begin{equation}}

14: \newcommand{\eeq}{\end{equation}}

15: \newcommand{\bear}{\begin{eqnarray}}

16: \newcommand{\eear}{\end{eqnarray}}

17: \newcommand{\bearn}{\begin{eqnarray*}}

18: \newcommand{\eearn}{\end{eqnarray*}}

19: 

20: \renewcommand{\theequation}{{\rm \thesection.\arabic{equation}}}

21: 

22: 

23: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

24: 

25: \begin{document}

26: 

27: \input epsf

28: 

29: 

30: 

31: 

32: \begin{center}

33: {\bf{\Large{ Higher-dimensional Temperley-Lieb algebras }}}

34: \end{center}

35: 

36: \begin{center}

37: {\bf{\large{\bf Marcos Alvarez\qquad Paul P. Martin}}}

38: \end{center}

39: 

40: \begin{center} 

41: Centre for Mathematical Science, City University \\

42: Northampton Square, London EC1V 0HB, UK \\

43: {\tt{e-mail: m.alvarez@city.ac.uk\qquad p.p.martin@city.ac.uk}}

44: \end{center}

45: 

46: \begin{abstract}

47: 

48: A category which generalises to higher dimensions many of the features of 

49: the Temperley-Lieb category is introduced. 

50: 

51: \end{abstract}

52: 

53: {\hskip 10pt{MSC-class: 81R50 (primary), 82B20 (secondary).}}

54: 

55: %% FONTS %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

56: 

57: 

58: \def\bb#1{{\mathbb{#1}}}

59: 

60: \def\comment#1

61: {\bigskip\hrule\medskip{\begin{quotation}\noindent\bf{#1}\end{quotation}\medskip\hrule\bigskip}}

62: 

63: %% DEFINITIONs &etc NUMBERING MACROS%%%%%%%%%%%%%%%%%%%%%%%%

64: 

65: \newcount\defnumber

66: \newcount\propnumber

67: \newcount\theonumber

68: \newcount\notnumber

69: \newcount\lemmanumber

70: \newcount\fignumber

71: \newcount\anything

72: 

73: \def\definition#1#2{\medskip\global\advance\anything by 1 %%\defnumber by 1

74: \noindent{\bf Definition \the\anything\hskip 5pt} %%\thesection.\the\defnumber\hskip 5pt}%%{\sl (#1)}

75: {\it #2}\smallskip} %% delete after cross-references

76: 

77: \def\proposition#1#2{\medskip\global\advance\anything by 1 %%\propnumber by 1

78: \noindent{\bf Proposition \the\anything\hskip 5pt} %%\thesection.\the\propnumber\hskip 5pt}%{\sl (#1)}} 

79: {\it #2}\smallskip} 

80: 

81: \def\theorem#1#2{\medskip\global\advance\anything by 1 %%\theonumber by 1

82: \noindent{\bf Theorem \the\anything\hskip 5pt} %%\thesection.\the\theonumber\hskip 5pt}%{\sl (#1)}}

83: {\it #2}\smallskip} %done

84: 

85: %\def\notation#1{\medskip\global\advance\notnumber by 1

86: %{\bf Notation \thesection.\the\notnumber\hskip 5pt}%{\sl (#1)}} 

87: %} %done

88: 

89: \def\lemma#1#2{\medskip\global\advance\anything by 1 %%\lemmanumber by 1

90: \noindent{\bf Lemma \the\anything\hskip 5pt} %%\thesection.\the\lemmanumber\hskip 5pt}%{\sl (#1)} }

91: {\it #2}\smallskip} %done

92: 

93: 

94: 

95: 

96: 

97: %% DEFINITIONS &etc CROSS-REFERENCING %%%%%%%%%%%%%%%%%%%%%%%%%%%%

98: 

99: \def\num{(\thesection.\the\eqnum)}

100: \def\appnum{({\sl{A}}\the\appnumber.\the\aeqnum)}

101: \def\deflabel{\thesection.\the\defnumber}

102: \def\proplabel{\thesection.\the\propnumber}

103: \def\theolabel{\thesection.\the\theonumber}

104: \def\notlabel{\thesection.\the\notnumber}

105: \def\lemlabel{\thesection.\the\lemmanumber}

106: 

107: \def\anylabel{\the\anything}

108: 

109: \def\nlabel{\edef}

110: 

111: %% SYMBOLS %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

112: 

113: \def\ie{{\it{i.e.\/}}}

114: \def\eg{{\it{e.g.\/}}}

115: \def\N{{\bb{N}}}

116: \def\Z{{\bb{Z}}}

117: \def\E{{\bb{E}}}

118: \def\R{{\bb{R}}}

119: \def\G{{\cal{G}}}

120: \def\nobub{{\sf{R}}}

121: \def\mapd{{\bb{D}}}

122: \def\suchthat{{\,\,|\,\,}}

123: \def\at#1{{\big\vert_{#1}}}

124: 

125: \def\S{{S}}%\def\S{{\bf{S}}}

126: \def\C{{\cal{C}}}

127: \def\K{{\cal{K}}}

128: 

129: 

130: 

131: \def\block{{$\Box$}}%{\hfill\vrule height 1.6ex width 1ex depth .5ex}

132: 

133: 

134: \def\isotopy{{{\mathfrak{i}}}}

135: \def\striso{{{\mathfrak{si}}}}

136: \def\relrone{{{\mathfrak{r}_1}}}

137: \def\relr{{{\mathfrak{r}}}}

138: \def\het{{{\mathfrak{h}}}}

139: \def\strhet{{{\mathfrak{sh}}}}

140: 

141: 

142: 

143: \def\here{{\vskip 1cm\hrule\vskip .5cm REVISE HERE \vskip

144: .5cm\hrule\vskip 1cm}}

145: \def\incomplete{{\vskip 1cm\hrule\vskip .5cm INCOMPLETE \vskip

146: .5cm\hrule\vskip 1cm}}

147: 

148: 

149: \def\np{Nucl.~Phys.}

150: \def\pl{Phys.~Lett.}

151: \def\pr{Phys.~Rev.}

152: \def\cmp{Commun.~Math.~Phys.}

153: \def\prs{Proc.~Roy.~Soc.}

154: 

155: 

156: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

157: %%%%%          		   M A I N   T E X T     

158: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

159: 

160: %\tableofcontents

161: %\newpage

162: 

163: \section{Introduction}

164: \setcounter{equation}{0}

165: 

166: The Temperley-Lieb category provides a useful tool in computational 2d 

167: lattice statistical mechanics \cite{baxterbook}.

168: Its representation theory is the interlocutor between several

169: different `equivalent' lattice models (Potts, IRF, 6-vertex, etc.) \cite{baxterbook, ppm}, 

170: is universal among an important class of solutions to integrability

171: conditions, and provides the invariant theory for $U_qsl_2$ \cite{jimbo, kassel}.

172: Its categorical structure means that its representation theory can be

173: analysed in great generality \cite{ppm, adamovich} (it is universal among monoidal dual

174: categories with certain natural properties, and was the starting point

175: for Khovanov homology \cite{khovanov}). It is also important in a number of other

176: areas of Mathematics and Physics \cite{kaufman, koo}.

177: 

178: The generalisation to `higher dimensions' (in the mean-field sense of

179: every lattice point being a neighbour) is the partition category \cite{ppmtl},

180: but it is natural also to seek a 3d version. A 3d lattice subalgebra

181: of the partition algebra was studied by Dasmahapatra and Martin \cite{dasmahapatra}

182: but it has very few of the beautiful properties of the Temperley-Lieb and

183: partition categories (cf. Baxter-Bazhanov tetrahedra for example \cite{bb, bbtetra}),

184: and essentially no progress has been made in this direction. Here we

185: describe a generalisation of the full Temperley-Lieb category to 3d that is 

186: closer in spirit to the cobordism category of Topological Quantum Field Theory \cite{baez, laures}

187: and to Lattice Gauge Theories (\cite{ppm}, page 278).

188: The problem with such a construction is that there are significant 

189: initial problems with well-definedness, but the payoff is potential access to generalisations of several of

190: the structures mentioned above. In this note we solve the well-definedness problem, and the core 

191: enumeration problem in using the resultant algebras. Applications will be discussed elsewhere. 

192: 

193: 

194: 

195: \section{Concrete diagram categories}

196: \setcounter{equation}{0}

197: {\global\propnumber=0}

198: {\global\defnumber=0}

199: {\global\lemmanumber=0}

200: {\global\theonumber=0}

201: {\global\notnumber=0}

202: 

203: 

204: We begin by selecting a direction in $d$-dimensional Euclidean space $\E^d$ which 

205: we call ``time'', coordinatised by a real number $t$. For fixed $d\in\N$ and $t\geq 0$ we define 

206: $E_t=\R^{d-1}\times [0,t]$. If $D$ and $E$ are subsets of $\E^d$, we write $C_E(D)$ for the number of 

207: connected components of $E\setminus D$ in the Euclidean topology.

208: 

209: 

210: 

211: \definition{condiagppm}{

212: The set $S^{d,t}$ of {\rm concrete diagrams of duration $t$} is the set of compact subsets of $E_t$ such that

213: $D\in S^{d,t}$ if and only if, for all $D'$ obtained from $D$ by removing a single point, we have  

214: $C_{E_t}(D')=C_{E_t}(D)-1$. Then $E_t$ is called a {\rm universe} for $D$

215: Write $S^d$ for $\cup_{t\geq 0} S^{d,t}$ and $S^{d-1}_{[]}$ 

216: for the set of concrete diagrams in $S^d$ with $t=0$.}

217: \nlabel\concdiagppm{\anylabel} %%{\deflabel}

218: 

219: 

220: 

221: 

222: 

223: \begin{figure}

224:   \begin{center}

225:     \begin{minipage}[c]{.58\linewidth}

226:       \epsffile{tl-categor.eps}

227:     \end{minipage}\hfill

228:     \begin{minipage}[c]{0.4\linewidth} 

229: 		{\caption{A concrete diagram in $S^{2,1}$ having a bubble: \newline $b(D)=1$, $g(D)=0$, $|D|=6$,		

230: 						\newline $\chi(D)=|D|-b(D)=5$.}}

231:     \end{minipage}

232:   \end{center}

233: \end{figure}

234: 

235: 

236: 

237: For $d=2$, concrete diagrams coincide with concrete Temperley-Lieb diagrams, and in this sense our 

238: construction provides a ``higher-dimensional'' generalisation. 

239: For $d=3$, a concrete diagram can be thought of as consisting of a number of %non-intersecting surfaces, 

240: embedded submanifolds whose boundaries all lie on either of two parallel planes of constant time. 

241: See Figures 1 and 2 for examples. 

242: 

243: 

244: \begin{figure}

245: \centering

246: \epsffile{codiex.eps}

247: \caption{Representations of several concrete diagrams in $S^{3,1}$ whose boundaries are two pairs of concentric circles. One concrete diagram has its boundary loops labelled 1 ,2, 1' and 2' for a later purpose.}

248: \end{figure}

249: 

250: 

251: A {\it component} of $D\in S^{d,t}$ is any point $x\in D$ together with the maximal subset

252: $X\ni x$ such that $C_{E_t}(D\setminus X)=C_{E_t}(D\setminus \{x\})$.

253: A {\it bubble} is a component which does not intersect

254: the boundaries of $E_t$. A component $c$ has a {\it handle} if it properly contains a non-contractible loop 

255: not homologous in $c$ to a subset of the boundary of $E_t$. Write $|D|$ for the number of components of a

256: concrete diagram $D$, and $b(D)$ for the number of bubbles. 

257: 

258: 

259: The {\it number of handles} of a concrete diagram $D\in S^3$ is the genus of $D$, written 

260: $g(D)$. We will write $\chi(D)$ for the Euler number of $D$ \cite{armstrong, mendels}.

261: 

262: 

263: 

264: 

265: \definition{sffppm}{

266: Let $F,F'\subset \R^{d-1}$. Then $S^{d,t}[F,F']$ is the subset of $S^{d,t}$ of concrete diagrams

267: which intersects the t-boundary of $E_t$ in $F$ and the 0-boundary in $F'$. 

268: Set $S^d[F,F']=\cup_{t\geq 0} S^{d,t}[F,F']$. The {\rm boundary configuration} of 

269: $D\in S^d[F,F']$ is the ordered pair $(F,F')$. }

270: \nlabel\sffppm{\anylabel} %%{\deflabel}

271: 

272: (N.B.: $S^d[F,F']=\emptyset$ unless $F,F'\in S^{d-1}_{[]}$. In 

273: practice we shall restrict attention to cases in which $F$ and $F'$ are unions of $(d-2)$-spheres.)

274: 

275: For $T$ a set, let $\wp(T)$ be the set of partitions of $T$ and ${\cal P}(T)$ the power set of $T$.  

276: For $U$ and $V$ sets and $u\in\wp(U)$, $v\in\wp(V)$ we write $u\ast v$ for the partition

277: algebra composition of $u$ with $v$ (\cite{ppmsaleur} p.158, or more generally the $Ag$ product in 

278: \cite{ppmtl}, p.77).

279: 

280: %\definition{conn}{

281: For $D\in S^d[F,F'] $ define its {\it connectivity} $p(D)$ as the partition of the components 

282: (with respect to the obvious extension of the $\E^{d-1}$ topology) of $F\dot\cup F'$, such that $a,b$ are

283: in the same part if they are in the same component of $D$. %} \nlabel\conn{\deflabel}

284: For example, for $D$ the concrete diagram with labelled boundaries in Figure 2, the connectivity 

285: is $p(D)=\{\{1,1'\},\{2\},\{2'\}\}$.

286: 

287: For fixed $d\in\N$ define the map $\nobub:S^d\to S^d$

288: by $\nobub(D)$ being $D$ with all bubbles removed.  %\definition{minimal}{

289: Define $S^d_{min}\subset S^d$ as the subset of elements with no bubbles and no handles.

290: The elements of $S^d_{min}$ are called {\it minimal}. %}\nlabel\minimal{\deflabel}

291: For example, all the diagrams shown in Figure 2 are minimal, whereas the diagram of Figure 1 is not, as it 

292: contains a bubble. 

293: 

294: %\definition{syd}{

295: For any two sets $A$ and $B$, define the symmetric difference 

296: $A\wedge B:=(A\cup B)\setminus(A\cap B)$. For $A,B\in S^d$ we write 

297: $A\overline{\wedge}B$ for the closure of $A\wedge B$ in the Euclidean metric 

298: topology. Two concrete diagrams $D$ and $D'$ are {\it $\wedge$-composable} if they have a 

299: universe in common and $D\cap D'$ is a finite union of disjoint 

300: closed $(d-1)$-balls. %}\nlabel\wedgecomp{\deflabel}

301: It can then be shown that \cite{AM}

302: 

303: \lemma{wed}{

304: If $D$ and $D'$ are $\wedge$-composable then $D\overline{\wedge}D'\in S^d$. }\block

305: \nlabel\wed{\anylabel} %%{\lemlabel}

306: 

307: 

308: %\definition{deltab}{

309: For $B\in S^d$, define $dom B$ as the set of all $A\in S^d$ which are $\wedge$-composable

310: with $B$. Then we define the map $\delta_B: dom B \to S^d$ by $A \mapsto  A\overline{\wedge}B$.

311: %}\nlabel\deltab{\deflabel}

312: 

313: 

314: For any point $P\in\E^d$ with time coordinate $t_P$, the {\it time translate} $P_\tau$ is the point in $\E^d$

315: with the same projection on the $t=0$ subspace as $P$ but with time coordinate $t_P+\tau$. For any 

316: $Z\subset\E^d$, we define $Z_\tau$ by $P\in Z_\tau\Leftrightarrow P_{-\tau}\in Z$. The 

317: following diagram composition Lemma follows now from Definition \concdiagppm:

318: 

319: \lemma{conc}{

320: If $A\in S^{d,t}[F',F]$ and $B\in S^{d,\tau}[F,F'']$, then $A\circ B:=A_{\tau}\cup B$ is in $S^d[F',F'']$. }\block

321: \nlabel\conc{\anylabel} %%{\lemlabel}

322: 

323: Let us now define the triple $\C^d=(S_{[]}^{d-1}\>,\> \hom(-,-),\>\circ\>)$

324: consisting of the ``object set'' $S_{[]}^{d-1}$; 

325: and for each pair of objects $E,F\in S_{[]}^{d-1}$ the collection of ``morphisms'' 

326: $\hom(E,F)=S^d[E,F]$; and composition of

327: morphisms defined by $\circ$-composition of concrete diagrams. The morphism in $\hom(F,F)$ of duration 

328: one whose sections of constant time are time translates of $F$ will be denoted ${\mathfrak I}_F$. 

329: 

330: 

331: \theorem{cdcat}{

332: $\C^d$ is a category.}

333: 

334: {\sl Proof:} We require to show associativity of $\circ$, and existence of identity element in each

335: $hom(F,F)$. The former is clear, the latter is the concrete diagram of duration zero. \block

336: \nlabel\cdcat{\anylabel} %%{\theolabel}

337: 

338: Let $A\in S^3[F',F]$ and $B\in S^3[F,F'']$. Then

339: \beq

340: \chi(A\circ B)=\chi(A)+\chi(B),

341: \label{eq:euleradd}

342: \eeq

343: \nlabel\chicirc{\lemlabel}\noindent

344: (which follows directly from the definition of $\chi(A)$ as an alternating sum of

345: Betti numbers, or as a sum of Morse indices \cite{Milnor}, of $A$) and

346: \beq

347: g(A\circ B) =g(A)+g(B)+|A\circ B| -|A|-|B|+|F|.

348: \label{eq:genusformula}

349: \eeq

350: We will later need other results that follow easily from Eq. (\ref{eq:euleradd}) which we give

351: without proof. Here we assume that all $\circ$-compositions are defined; 

352: $D$ is a minimal concrete diagram with the same connectivity as $A\circ B$, and 

353: $G$ is a minimal concrete diagram with the same connectivity as $B\circ C$:

354: \beq

355: %b(A\circ B)&=&|A\circ B|-|D|,\\

356: %g(A\circ B\circ C)&=&|A\circ B\circ C| -|A|-|B|-|C|+2|F|,\\

357: %|A\circ B|-|D|+|D\circ G|&=& |B\circ C|-|G|+|A\circ G|,\\

358: \begin{array}{ccc}

359: &g(A\circ B)+g(D\circ C) = g(B\circ C)+g(A\circ G),& \\

360: &b(A\circ B)+b(D\circ C) = b(B\circ C)+b(A\circ G).&

361: \end{array}

362: \label{eq:bandg}

363: \eeq

364: 

365: 

366: 

367: 

368: 

369: \section{Equivalence relations on $S^3$}

370: \setcounter{equation}{0}

371: {\global\propnumber=0}

372: {\global\defnumber=0}

373: {\global\lemmanumber=0}

374: {\global\theonumber=0}

375: {\global\notnumber=0}

376: 

377: In this section we set $d=3$. We will write $\isotopy$ for the usual relation of isotopy 

378: (\ie, a continuous, one-parameter family of homeomorphisms {\cite{Moise}}). 

379: If $\mathfrak j$ is a specific isotopy with parameter $s\in[0,1]$ and $A$ 

380: is a concrete diagram, we write ${\mathfrak j}_u(A)$  for the image of $A$ under 

381: ${\mathfrak j}$ at $s=u$. In particular, ${\mathfrak j}_0(A)=A$ for all $A$.

382: Two concrete diagrams $A$ and $B$ are {\it isotopic} iff there is an isotopy $\mathfrak j$ 

383: such that ${\mathfrak j}_s(A)$ is a concrete diagram for all $s$, and

384: ${\mathfrak j}_1(A)=B$.

385: 

386: This defines the isotopy relation $\isotopy$ in the set of concrete diagrams. It is obviously an 

387: equivalence relation.

388: 

389: %\definition{strongiso}{

390: Let $A$ and $B$ be concrete diagrams. If $A$ and $B$ are isotopic by an isotopy ${\mathfrak j}$

391: such that the boundaries of ${\mathfrak j}_s(A)$ are time translates of the boundaries of $A$ for all $s$

392: we say that they are {\it strongly isotopic}. %%}\nlabel\strongiso{\deflabel}

393: 

394: For example, the isotopy whose action on a point in $\E^3$ with coordinates $(x,y,t)$ is

395: ${\mathfrak j}_s(x,y,t)=(x,y,(1+s)t)$ is strong. Strong isotopy is clearly an equivalence relation, which we 

396: denote $\striso$.

397: 

398: \definition{relations}{

399: \\ 1. For any $A,B\in S^3$, define relation $\relrone$ by $A\relrone B$ if there is a torus $T$ such that 

400: $A\cap T$ is a disk, and $B=\delta_T A$. Relation $\relr$ is the transitive closure of $\relrone$.

401: \\2. Relation $\strhet$, called {\rm strong heterotopy}, is the reflexive, symmetric and transitive closure of 

402: $\relrone$ and $\striso$.

403: \\3. Relation $\het$, called {\rm heterotopy}, is the reflexive, symmetric and transitive closure of 

404: $\relrone$ and $\isotopy$. }

405: \nlabel\relations{\anylabel} %%{\deflabel}

406: 

407: 

408: The cosets $S^3[F,F']/\strhet$ and $S^2[F,F']/\isotopy$ (Temperley-Lieb diagrams) are infinite owing to the 

409: possible presence of bubbles or handles. But $S^2_{min}[F,F']/\isotopy$ is finite. We will show in 

410: section~\ref{scon} that $S^3_{min}[F,F']/\strhet$ is finite.

411: 

412: The following Lemma follows immediately from the previous definitions:

413: 

414: \lemma{bubhet}{

415: If $A\het A'$ then $b(A)=b(A')$. If $A\strhet A'$ then $p(A)=p(A')$. }\block

416: \nlabel\bubhet{\anylabel} %%{\lemlabel}

417: 

418: \definition{hetclass}{

419: For $D\in S_{min}^3$, write $[D]_\het$ for the restriction of the $\het$-class of $D$ to

420: $S_{min}^3$, \ie, $[D]_\het=\{C\in S_{min}^3 \suchthat C\het D\}$. 

421: Write $\S_\het^3$ for the set of $\het$-classes in $S_{min}^3$ and $\S_\het^3[F,F']$ for the

422: set of $\het$-classes in $S_{min}^3[F,F']$. Similarly, write $[D]_\strhet$ for 

423: the restriction of the $\strhet$-class of $D$ to $S_{min}^3$, \ie,

424: $[D]_\strhet=\{C\in S_{min}^3 \suchthat C\strhet D\}$, and 

425: $\S_\strhet^3$ for the set of $\strhet$-classes in $S_{min}^3$, and $\S_\strhet^3[F,F']$ for the

426: set of $\strhet$-classes in $S_{min}^3[F,F']$.}

427: \nlabel\hetclass{\anylabel} %%{\deflabel}

428: 

429: By the handle decomposition theorem \cite{Milnor}:

430: 

431: \proposition{minrep}{

432: For every $A\in S^3$ there exists $D\in S_{min}^3$ such that $D\relr \nobub(A)$. \block}

433: \nlabel\minrep{\anylabel} %%{\proplabel}

434: 

435: \definition{mapdr}{

436: Define $\mapd_r:S^3 \to  {\cal P}(S_{min}^3)$ by 

437: $A \mapsto \{D\in S_{min}^3 \suchthat D\relr\nobub(A)\}$.}

438: \nlabel\mapdr{\anylabel} %%{\deflabel}

439: 

440: The following Proposition follows from Proposition \minrep\ and Definition \relations.

441: 

442: \proposition{mapdvsh}{

443: For each $A\in S^3$  there is $D\in S_{min}^3$ such that $\mapd_r(A)\subseteq[D]_\strhet$.

444: Moreover, if $\mapd_r(A)\subseteq[D]_\strhet$ and $\mapd_r(A)\subseteq[D']_\strhet$ then

445: $[D]_\strhet=[D']_\strhet$. }\block

446: \nlabel\mapdvsh{\anylabel} %%{\proplabel}

447: % GOES THROUGH JUST FINE FOR STRHET

448: %{\sl Proof: } Let $C\in \mapd_r(A)$ and suppose $\nobub(A)$ is minimal. Then $C$ is 

449: %$\relr$-related (hence $\het$-related) to $\nobub(A)$, and then $C\in [\nobub(A)]_\het$, 

450: %so we take $D=\nobub(A)$. If $\nobub(A)$ is non-minimal then there is $D$ minimal such that 

451: %$\nobub(A)\relr D$. By transitivity, $C$ is $\relr$-related to $D$, hence $\het$-related to $D$. Then 

452: %$C\in[D]_\het$ for all $C$, so  $\mapd_r(A)\subseteq [D]_\het$.

453: %

454: %Finally, let  $\mapd_r(A)\subseteq[D]_\het$ and $\mapd_r(A)\subseteq[D']_\het$. Then any

455: %$C\in\mapd_r(A)$ is in relation $\het$ to both $D$ and $D'$. By transitivity, $D\het D'$ and

456: %$[D]_\het=[D']_\het$\block

457: 

458: %\definition{themapd}{

459: Define $\mapd_\het(A)$ and $\mapd_\strhet(A)$ as, respectively, the $\het$- and $\strhet$-class 

460: in $S^3_{min}$ containing $\mapd_r(A)$. %} \nlabel\themapd{\deflabel}

461: 

462: \definition{redmap}{

463: For $\K$ a ring and $p,q\in\K$ define the {\rm reduction maps} $\mu$ and $\nu$ as

464: \bearn\centering

465: \begin{array}{rclccrcl}

466: \mu: S^3 &\to & \K \S_\het^3    &  &    &      \nu: S^3 &\to & \K \S_\strhet^3       \\

467: A &\mapsto & p^{g(A)}q^{b(A)}\mapd_\het(A)  &&& A &\mapsto & p^{g(A)}q^{b(A)}\mapd_\strhet(A)

468: \end{array}

469: \eearn

470: and extend linearly to $\K \S^3$ in each case.}

471: \nlabel\redmap{\anylabel} %%{\deflabel}

472: 

473: If $A$ and $B$ are $\circ$-composable, then any $A'\strhet A$ and $B'\strhet B$ are also 

474: $\circ$-composable. The next Lemma follows from noticing that the two strong heterotopies relating 

475: $A$ to $A'$ and $B$ to $B'$ combine into a single one that relates $A\circ B$ to $A'\circ B'$:

476: 

477: \lemma{shsh}{

478: Let $A$ and $B$ are $\circ$-composable. Let $A'\strhet A$ and $B'\strhet B$. Then

479: $(A\circ B)\strhet(A'\circ B')$. \block}

480: \nlabel\shsh{\anylabel} %%{\lemlabel}

481: 

482: Therefore we can extend $\circ$-composability to $\strhet$-classes 

483: in a well-defined way by $\circ$-composing representatives.

484: 

485: \definition{circsh}{

486: Let $A,B$ be $\circ$-composable. Then $[A]_\strhet \bullet [B]_\strhet := \nu(A\circ B)$.}

487: \nlabel\circsh{\anylabel} %%{\deflabel}

488: 

489: \theorem{catsh}{

490: The triple $\C_{sh}=(S^2_{[]}\>,\> \K\S^3_\strhet[-,-],\> \bullet\>)$

491: is a category whose morphisms are $\strhet$-classes of concrete diagrams.}

492: 

493: {\sl Proof: } i) Associativity of $\bullet$-composition of $\strhet$-classes follows from 

494: associativity of $\circ$-composition of concrete diagrams. ii) The unit in $\S^3_\strhet[F,F]$

495: is the $\strhet$-class of the diagram of zero duration in $S^3[F,F]$. \block

496: 

497: %\lemma{circh}{

498: %Let $A$ and $B$ be $\circ$-composable, and $B\het B'$. Then there exists $A'\isotopy A$ such that

499: %$(A\circ B) \het (A'\circ B')$.}

500: %

501: %{\sl Proof:} If $B\strhet B'$ it suffices to take $A'=A$. If $B\het B'$ but its upper boundary remains 

502: %fixed, then take $A'=A$ as well. Otherwise the heterotopy from $B$ to $B'$ includes an isotopy

503: %in a neighbourhood of the upper boundary of $B$. Then $A'$ is obtained from $A$ by extending that

504: %isotopy to a neighbourhood of the lower boundary of $A$. \block

505: %\nlabel\circh{\lemlabel}

506: 

507: \proposition{hettostrhet}{

508: Let $A,B\in S^d[F,F']$ and $A\het B$. Then there exist $L\in S^3_{min}[F,F]$ and $R\in S^3_{min}[F',F']$ 

509: such that $L\het {\mathfrak I}_F$, $R\het {\mathfrak I}_{F'}$, and $A\strhet(L\circ B\circ R)$.}

510: 

511: {\sl Proof:} If $A\strhet B$ then take $L={\mathfrak I}_F$ and $R={\mathfrak I}_{F'}$. Otherwise 

512: the $\het$-relation between 

513: $A$ and $B$ contains isotopies in a neighbourhood of the boundary of $A$. Extend those isotopies to

514: a neighbourhood of ${\mathfrak I}_F$ and ${\mathfrak I}_{F'}$ in 

515: ${\mathfrak I}_F\circ A\circ{\mathfrak I}_{F'}$

516: to obtain $L\circ B\circ R$. The $L$ and $R$ so defined are clearly $\het$-related to 

517: ${\mathfrak I}_F$ and ${\mathfrak I}_{F'}$ respectively (in fact, isotopic). \block

518: \nlabel\hettostrhet{\anylabel} %%{\proplabel}

519: 

520: By a routine check that the group axioms are satisfied, we have

521: 

522: \proposition{idclasses}{

523: The $\strhet$-classes inside $[{ \mathfrak I}_F]_\het$ in $S_{min}^3[F,F]$ form a group 

524: $\Pi_F$ under $\bullet$-composition, with unit $[{ \mathfrak I}_F]_\strhet$. \block}

525: \nlabel\idclasses{\anylabel} %%{\proplabel}

526: 

527: %{\sl Proof: } Let  $[Y]_\strhet$ and $[Z]_\strhet$ be two such $\strhet$-classes, represented 

528: %respectively by concrete diagrams $Y$ and $Z$.

529: %\bearn

530: %[Y]_\strhet \circ [Z]_\strhet= [Y\circ Z]_\strhet.

531: %\eearn

532: %Clearly, $Y\circ Z$ is minimal. We need to show that it is $\het$-related to ${\mathfrak I}_F$.

533: %But $Z\het {\mathfrak I}_F$, so by Lemma \circh\ there is $Y'\het Y$ such that 

534: %$(Y\circ Z) \het (Y'\circ{\mathfrak I}_F)$. But $(Y'\circ{\mathfrak I}_F)\het Y'$. Together with

535: %$Y'\het Y$ and $Y\het {\mathfrak I}_F$, and transitivity of $\het$, we find that

536: %$(Y\circ Z) \het {\mathfrak I}_F$. Therefore $[Y]_\strhet \circ [Z]_\strhet$ is a $\strhet$-class 

537: %inside $[{ \mathfrak I}_F]_\het$.

538: %

539: %That $[{ \mathfrak I}_F]_\strhet$ is the unit follows from $(Y\circ { \mathfrak I}_F)\strhet Y$ and 

540: %$({ \mathfrak I}_F\circ Y)\strhet Y$. Then

541: %\bearn

542: %[Y]_\strhet \circ [{ \mathfrak I}_F]_\strhet = [Y\circ { \mathfrak I}_F]_\strhet=[Y]_\strhet

543: %\eearn

544: %and similarly $[{ \mathfrak I}_F]_\strhet\circ [Y]_\strhet=[Y]_\strhet$. Associativity of

545: %$\circ$-composition of $\strhet$-classes has already been noted after Definition \circsh.

546: %Now we show that every $\strhet$-class inside $[{\mathfrak I}_F]_\het$ in 

547: %$S_{min}^3[F,F]$ has an inverse. Let $Y\in S_{min}^3[F,F]$ with $Y\het { \mathfrak I}_F$,

548: %and $Y^T$ be the image of $Y$ under time reversal $t\to -t$. We will prove first that 

549: %$(Y\circ Y^T)\strhet {\mathfrak I}_F$, thus proving that $[Y^T]_\strhet$ is the $\circ$-inverse

550: %of $[Y]_\strhet$.

551: %

552: %By proposition \hettostrhet\ there are $L,R\in [{ \mathfrak I}_F]_\het$ such that 

553: %$Y\strhet(L\circ { \mathfrak I}_F\circ R)$. By time inversion, we also have 

554: %$Y^T \strhet(R^T\circ { \mathfrak I}_F\circ L^T)$.

555: %Clearly $(L\circ L^T)\strhet { \mathfrak I}_F$ and $(R\circ R^T)\strhet { \mathfrak I}_F$ (in fact,

556: %strongly isotopic). Therefore

557: %\bearn

558: %(Y\circ Y^T)\strhet \left(

559: %(L\circ { \mathfrak I}_F\circ R)\circ (R^T\circ { \mathfrak I}_F\circ L^T)\right).

560: %\eearn

561: %Applying associativity of $\circ$-composition, it follows that $(Y\circ Y^T)\strhet {\mathfrak I}_F$,

562: %as required. It is clear that $(Y^T\circ Y)\strhet {\mathfrak I}_F$ also. Therefore all the group

563: %axioms are satisfied \block

564: 

565: 

566: \definition{sumx}{

567: For a finite $X\subset S^d_{min}$, define $\sigma_{\scriptscriptstyle X}=\sum_{A\in X} A\in\K S_{min}^d$.}

568: \nlabel\sumx{\anylabel} %%{\deflabel}

569: 

570: \section{$\het$-classes and connectivity.}\label{scon}

571: \setcounter{equation}{0}

572: {\global\propnumber=0}

573: {\global\defnumber=0}

574: {\global\lemmanumber=0}

575: {\global\theonumber=0}

576: {\global\notnumber=0}

577: 

578: In this Section we show that the $S^3[F,F']/\strhet$ is finite.

579: 

580: \definition{neighbours}{

581: Let $c$ and $d$ be components of $A\in S^d$. Then $c$ and $d$ are {\rm neighbours} if there is a

582: path connecting $c$ to $d$ not intersecting any other component of $A$. Let $c$ and $d$ be neighbours in 

583: $A$ with path $P$, and $e$ a surface obtained from $c$ and $d$ by removing a disk from each and joining the 

584: edges with a cylindrical thickening of $P$ not intersecting any other component of $A$. Then $e$ is a 

585: {\rm bridging} of $c$ and $d$ and the cylinder is a {\rm bridge} connecting $c$ and $d$.}

586: \nlabel\neighbours{\anylabel} %%{\deflabel}

587: 

588: \theorem{bigtheorem}{

589: Let $A,B\in S^3_{min}[F,F']$. Then $A\strhet B$ iff $p(A)=p(B)$. \block}

590: \nlabel\bigtheorem{\anylabel} %%{\theolabel}

591: 

592: Necessity follows from the definition of $\strhet$. A complete proof of sufficiency will be given in \cite{AM}. 

593: Here we present those ideas of the proof that are relevant to understanding the rest of this paper. The key 

594: observation is that any two concrete diagrams in $S^3_{min}[F,F']$ with $|F|+|F'|$ components are 

595: necessarily $\strhet$-related (in fact, $\striso$-related) and have equal connectivities. (Both statements follow 

596: from the fact that all components in any such concrete diagram are disks, each disk bounded by exactly one

597: loop in either $F$ or $F'$.) The proof then proceeds by induction in $k=|F|+|F'|-n$ where $n$ is the number of 

598: components of $A$ and $B$. Given $A,B\in S^3_{min}[F,F']$ with $p(A)=p(B)$ and $n<|F|+|F'|$ 

599: components (\ie, $k>0$), concrete diagrams $A',B'\in S^3_{min}[F,F']$ are constructed such that 

600: $p(A')=p(B')$, and both have $n+1$ components (\ie, $k$ reduced by one). The induction hypothesis is

601: that the Theorem is true for $k-1$, so that $A'\strhet B'$. It is then shown that $A$ and $B$ can be 

602: reconstructed from $A'$ and $B'$ by bridging, in a way that shows that $A$ and $B$ are also $\strhet$-related. 

603: This shows that the Theorem is true for $k$ if it is true for $k-1$. Being true for $k=0$ (concrete diagrams

604: in $S^3_{min}[F,F']$ with $|F|+|F'|$ components), it follows that it is always true.

605: 

606: \proposition{finpif}{

607: For any $F$, $\Pi_F$ is a finite group.}

608: 

609: {\sl Proof: } It was established in Proposition \idclasses\ that $\Pi_F$ is a group. Because $F$

610: is always a finite set, the partition set $\wp(F\dot\cup F)$ is also finite. Hence, by Theorem

611: \bigtheorem, there is a finite number of $\strhet$-classes in $[{\mathfrak I}_F]_\het$. \block

612: \nlabel\finpif{\anylabel} %%{\proplabel}

613: 

614: 

615: \section{Composition of $\het$-classes}

616: \setcounter{equation}{0}

617: {\global\propnumber=0}

618: {\global\defnumber=0}

619: {\global\lemmanumber=0}

620: {\global\theonumber=0}

621: {\global\notnumber=0}

622: \indent

623: 

624: Consequent to Proposition \hettostrhet, there are fewer $\het$-classes than

625: $\strhet$-classes in $S^3_{min}[F,F']$. We regard the extra symmetry $\het/\strhet$ as the 

626: higher-dimensional counterpart of order-preserving displacements of the endpoints of Temperley-Lieb

627: diagrams along the upper or lower edges. For that reason we regard $\het$-classes as the

628: natural generalisation of the notion of Temperley-Lieb diagram to higher dimensions, and they will be

629: the main object of study in the rest of this work.

630: 

631: Our goal, then, is to define a composition of $\het$-classes by which each $\S_\het^3[F,F]$ becomes a 

632: finite-dimensional, associative algebra with unit. 

633: 

634: \proposition{shmu}{

635: Let $A,A',B,B'\in S^3_{min}$ and $A\strhet A'$ and $B\strhet B'$. If $A$ and $B$ are 

636: $\circ$-composable, then $\mu(A\circ B)=\mu(A'\circ B')$.}

637: 

638: {\sl Proof:} From Lemma \shsh, $\mapd(A\circ B)=\mapd(A'\circ B')$. Therefore we only need to

639: prove that $b(A\circ B)=b(A'\circ B')$ and $g(A\circ B)=g(A'\circ B')$. 

640: By Lemma \shsh\ we know that $(A\circ B)\strhet(A'\circ B')$, and then by Lemma \bubhet\

641: $b(A\circ B)=b(A'\circ B')$. By Theorem \bigtheorem\

642: we know that $p(A)=p(A')$ and $p(B)=p(B')$, hence $p(A\circ B)=p(A'\circ B')$ and,

643: in particular, $A\circ B$ and $A'\circ B'$ have equal number of non-bubble 

644: components. Hence $|A\circ B|=|A'\circ B'|$. But it follows from

645: Eq. \ref{eq:genusformula} that $g(A\circ B)$ depends only on the 

646: number of components of $A\circ B$ if $A$ and $B$ are minimal. Therefore  

647: $g(A\circ B)=g(A'\circ B')$. \block

648: \nlabel\shmu{\anylabel} %%{\proplabel}

649: 

650: This Proposition can be extended to show that, if $A_i,A_i'\in S^3_{min}[F_i,F_{i+1}]$ with

651: $A_i\strhet A_i'$ for $i=1,\ldots,n$, then

652: \beq

653: \mu(A_1\circ\cdots\circ A_n)=\mu(A'_1\circ\cdots\circ A'_n).

654: \label{eq:muconc}

655: \eeq

656: 

657: 

658: \definition{complete}{

659: A subset of $[{ \mathfrak I}_F]_\het$ is {\rm complete} if it consists of exactly one element from 

660: each $\strhet$-class. }

661: \nlabel\complete{\anylabel} %%{\deflabel}

662: 

663: 

664: \lemma{redcomp}{

665: Let $A,A'\in S^d_{min}[F',F]$ and $B,B'\in S_{min}^d[F,F'']$, with $A\het A'$ and 

666: $B\het B'$. Let $X,X'$ be complete subsets of $[{ \mathfrak I}_F]_\het$. Then 

667: \bearn

668: \mu(A\circ \sigma_X\circ B)=\mu(A'\circ \sigma_{X'}\circ B').

669: \eearn}

670: \nlabel\redcomp{\anylabel} %%{\lemlabel}

671: {\sl Proof:} By Proposition \finpif, $X$ and $X'$ are finite sets and $\sigma_X$ and $\sigma_{X'}$ are defined.

672: By Proposition \hettostrhet\ there are $L_A\het {\mathfrak I}_{F'}$, 

673: $R_A\het {\mathfrak I}_F$ such that $A\strhet(L_A\circ A'\circ R_A)$, and similarly

674: $L_B\het {\mathfrak I}_F$, $R_B\het {\mathfrak I}_{F''}$ such that

675: $B\strhet(L_B\circ B' \circ R_B)$. Then, for each $Y\in\sigma_X$, Eq. (\ref{eq:muconc}) gives

676: $\mu(A\circ Y\circ B)=\mu(L_A\circ A'\circ R_A\circ Y \circ L_B\circ B' \circ R_B)$.

677: But, by completeness of $X$ and Proposition \idclasses\ there exists exactly one $Y'\in X'$

678: such that $(R_A\circ Y\circ L_B)\strhet Y'$, so $\mu(A\circ Y\circ B)=\mu(L_A\circ A'\circ Y'\circ B' \circ R_B)$.

679: Finally, $(L_A\circ A'\circ Y'\circ B' \circ R_B)\het (A'\circ Y'\circ B')$ and both have equal

680: number of bubbles and handles, so $\mu(A\circ Y\circ B)=\mu(A'\circ Y'\circ B')$.

681: The Lemma follows by summing over $Y\in X$. \block

682: 

683: 

684: 

685: That is, $\mu(A\circ \sigma_X\circ B)$ depends only on the $\het$-classes of $A$ and $B$ and not

686: on the choice of $X$. Therefore we have a well-defined composition of $\het$-classes into $\K\S^3_\het$:

687: \beq

688: [A]_\het \cdot' [B]_\het=\mu(A\circ \sigma_X\circ B),

689: \eeq

690: If $|X|$ has an inverse in $\K$ we define a new composition rule that has $[{\mathfrak I}_F]_\het$ 

691: as unit:

692: \beq

693: [A]_\het \cdot [B]_\het={1\over |X|}\mu(A\circ \sigma_X\circ B),

694: \label{eq:therule}

695: \eeq

696: As was done in Section 3 with $\strhet$-classes, the set $\S^3_\het[F,F']$ of $\het$-classes in $S^3[F,F']$

697: can be interpreted as the set of morphisms with object set $S^2_{[]}$ and composition rule 

698: given by \ref{eq:therule}. In the next subsection we show that this composition rule is associative. Then,

699: 

700: \theorem{cath}{

701: The triple $\C_\het=( S^2_{[]}\>,\> \K\S^3_\het[-,-], \>\cdot\>)$ 

702: defines a category which we will be called the {\it heterotopy category}. }\block

703: \nlabel\cath{\anylabel} %%{\theolabel}

704: 

705: The only non-trivial step in the proof of this Theorem is associativity:

706: 

707: \proposition{assoc}{

708: Let $[A]_\het, [B]_\het, [C]_\het\in\S^3$. Then 

709: \bearn

710: ([A]_\het \cdot [B]_\het)\cdot [C]_\het= [A]_\het \cdot ([B]_\het\cdot [C]_\het).

711: \eearn}

712: {\sl Proof:} Let $X$ be a complete subset of $[{\mathfrak I}_F]_\het$. For each $Y\in X$ define 

713: $[E_Y]=\mapd(A\circ Y\circ B)$ and $[H_Y]=\mapd(B\circ Y'\circ C)$. Then

714: \bearn

715: & &([A]_\het \cdot [B]_\het)\cdot [C]_\het =\\ 

716: &&\> ={1\over |X|^2}\sum_{Y,Y'\in X}p^{g(A\circ Y\circ B)}q^{b(A\circ Y\circ B)}

717: p^{g(E_Y\circ Y'\circ C)}q^{b(E_Y\circ Y'\circ C)}\mapd(E_Y\circ Y'\circ C),\\

718: &&{[A]_\het} \cdot ([B]_\het \cdot [C]_\het) =\\

719: &&\> ={1\over |X|^2}\sum_{Y,Y'\in X}p^{g(A\circ Y\circ H_{Y'})}q^{b(A\circ Y\circ H_{Y'})}

720: p^{g(B\circ Y'\circ C)}q^{b(B\circ Y'\circ C)}\mapd(A\circ Y\circ H_{Y'}).

721: \label{eq:assone}

722: \eearn

723: Because both $E_Y\circ Y'\circ C$ and $A\circ Y\circ H_{Y'}$ are in 

724: $\mapd(A\circ Y\circ B\circ Y'\circ C)$ we have 

725: $\mapd(E_Y\circ Y'\circ C)=\mapd(A\circ Y\circ H_{Y'})$. Therefore we only need to prove that

726: \begin{displaymath}

727: \begin{array}{c}

728: g(A\circ Y\circ B)+g(E_Y\circ Y'\circ C) =  g(A\circ Y\circ H_{Y'})+g(B\circ Y'\circ C),\\

729: b(A\circ Y\circ B)+b(E_Y\circ Y'\circ C) =  b(A\circ Y\circ H_{Y'})+b(B\circ Y'\circ C).

730: \end{array}

731: \end{displaymath}

732: Those equations follow from Eqs.(\ref{eq:bandg}) in page \pageref{eq:bandg}. \block

733: \nlabel\assoc{\anylabel} %%{\proplabel}

734: 

735: The rule (\ref{eq:therule}) does not provide a practical means of computing the 

736: composition of two given $\het$-classes, except in the simplest cases where representative concrete 

737: diagrams can be concatenated by hand. We address this question next.

738: 

739: 

740: \section{Computing the multiplication table}

741: \setcounter{equation}{0}

742: {\global\propnumber=0}

743: {\global\defnumber=0}

744: {\global\lemmanumber=0}

745: {\global\theonumber=0}

746: {\global\notnumber=0}

747: 

748: In order to generate examples, and to begin analysing the structure of the algebra, we need to be able to

749: compute compositions efficiently. If we wished to compute the multiplication table for the algebra

750: $S^3_\het[F,F]$ with $F$ two concentric circles, we could use as a concrete basis the concrete diagrams in 

751: Figure 2, concatenate them by hand and finally apply the reduction map $\mu$. This would result in the 

752: following multiplication table, in which $D_1,\cdots,D_9$ refers to the concrete diagrams shown in Figure 2, 

753: numbered from left to right.

754: 

755: 

756: 

757: {\scriptsize

758: \begin{displaymath}

759: \vbox{

760: {\offinterlineskip

761: \tabskip=0pt

762: \halign{

763: \hfil # \hfil & \vrule height2.75ex depth1.25ex width 0.6pt #

764: & \hskip 5pt\hfil # \hfil & \hfil # \hfil & \hfil # \hfil & \hfil # \hfil

765: & \hfil # \hfil & \hfil # \hfil & \hfil # \hfil & \hfil # \hfil

766: & \hfil # \hfil \cr

767: row $\times$ col

768: & & $D_1$ & $D_2$ & $D_3$ & $D_4$ & $D_5$ & $D_6$ & $D_7$ & $D_8$ & $D_9$ \cr

769: \multispan{11}\hrulefill\cr

770: $D_1$ & & $pD_1$ & $pD_2$ & $D_1$ & $pD_4$ & $D_1$ & $pD_6$ & $D_2$ & $D_4$

771: & $D_6$\cr

772: $D_2$ & & $D_1$ & $D_2$ & $qD_1$ & $D_4$ & $D_2$ & $D_6$ & $qD_2$ & $qD_4$

773: & $qD_6$\cr

774: $D_3$ & & $pD_3$ & $pD_7$ & $D_3$ & $pD_8$ & $D_3$ & $pD_9$ & $D_7$ & $D_8$

775: & $D_9$\cr

776: $D_4$ & & $pD_4$ & $pD_6$ & $D_4$ & $pqD_4$ & $D_4$ & $pqD_6$ & $D_6$ & $qD_4$

777: & $qD_6$\cr

778: $D_5$ & & $D_1$ & $D_2$ & $D_3$ & $D_4$ & $D_5$ & $D_6$ & $D_7$ & $D_8$

779: & $D_9$\cr

780: $D_6$ & & $D_4$ & $D_6$ & $qD_4$ & $qD_4$ & $D_6$ & $qD_6$ & $qD_6$ & $q^2D_4$

781: & $q^2D_6$\cr

782: $D_7$ & & $D_3$ & $D_7$ & $qD_3$ & $D_8$ & $D_7$ & $D_9$ & $qD_7$ & $qD_8$

783: & $qD_9$\cr

784: $D_8$ & & $pD_8$ & $pD_9$ & $D_8$ & $pqD_8$ & $D_8$ & $pqD_9$ & $D_9$ & $qD_8$

785: & $qD_9$\cr

786: $D_9$ & & $D_8$ & $D_9$ & $qD_8$ & $qD_8$ & $D_9$ & $qD_9$ & $qD_9$ & $q^2D_8$

787: & $q^2D_9$\cr}}

788: }

789: \end{displaymath}

790: }%

791: 

792: 

793: We now present a more efficient method of computing multiplication tables which makes use of the bijection 

794: established in theorem \bigtheorem\

795: between $\strhet$-classes in $S^3_{min}[F,F']$ and those partitions of $F\dot\cup F'$ which are connectivities 

796: of concrete diagrams in $S^3_{min}[F,F']$. A convenient way to represent a partition of $F\dot\cup F'$ 

797: for this purpose is by means of coloured graphs:

798: 

799: \definition{coldef}{

800: Let $G$ be a graph and $C$ a set. A {\rm colouring} of $G$ by $C$ is a map from the edge set of $G$

801: to $C$. Write $G^C$ for the set of all colourings of $G$ by $C$.}

802: \nlabel\coldef{\anylabel} %%{\deflabel}

803: 

804: Given $F$, let $\G(F)$ be the rooted undirected graph constructed as follows. The vertex set is the set of

805: connected components of $\R^2\setminus F$ (``regions''), the root being the vertex associated to the

806: unbounded region; there is an edge between two vertices if there is a component in $F$ which is a boundary

807: between the corresponding regions. 

808: 

809: We may associate a rooted tree to the graph $\G(F)$ by forgetting the labels on all the vertices

810: except the root. Now consider a concrete diagram $D\in S_{min}^3[F,F']$ and an injective map $f$ 

811: with domain the set of component of $D$. We will say that component $d$ has ``colour'' 

812: $f(d)$. Given the pair $(D,f)$ we define a colouring $\phi(D,f)$ of the edges of the ordered pair of graphs 

813: $(\G(F),\G(F'))$ as follows: if $l\in F\dot\cup F'$ is in the boundary of component $d$ in $D$, then the colour

814: of the edge associated to $l$ is $f(d)$.

815: 

816: We will regard $(\G(F),\G(F'))$ as the single tree $\G(F\dot\cup F')$ by identifying the roots.

817: Then $\phi(D,f)$ defines a partition of the set of edges of $\G(F\dot\cup F')$ which corresponds

818: to the partition $p(D)$ of the set of loops of $F\dot\cup F'$. By Theorem \bigtheorem, each such colouring

819: of $\G(F\dot\cup F')$ corresponds to a $\strhet$-class in $\S^3_\strhet[F,F']$. 

820: 

821: 

822: 

823: 

824: 

825: \begin{figure}

826: \centering

827: \leavevmode

828: \epsfbox{ labelloops.eps   }\hskip 1cm

829: \epsfbox{ colouredtrees.eps  }

830: \caption{(a) Construction of $\G(F\dot\cup F)$ from Figure 2. (b) Admissible colourings}

831: \label{fig:tb}

832: \end{figure}

833: 

834: 

835: 

836: 

837: 

838: 

839: 

840: \definition{admisscol}{

841: For any two edges $e$ and $e'$ in a tree let $ch(e,e')$ be the chain of edges connecting $e$ to $e'$ in the 

842: tree (excluding $e$ and $e'$). An element of $\G(F\dot\cup F')^C$ is {\rm admissible} iff for every pair of 

843: same-coloured edges $e$ and $e'$, either there is another edge in $ch(e,e')$ of the same colour, or else 

844: every colour in $ch(e,e')$ appears an even number of times.}

845: \nlabel\admisscol{\anylabel} %%{\deflabel}

846: 

847: \proposition{realis}{

848: The image under $\phi$ of the set of $n$-component elements of $S_{min}^3[F,F']$ is

849: the set of admissible colourings in $\G(F\dot\cup F')^{\{1,2,\cdots,n\}}$. }\block

850: \nlabel\colrule{\anylabel} %%{\proplabel}

851: 

852: We refer to \cite{AM} for a proof. This Proposition establishes a correspondence between 

853: admissibly-coloured graphs and $\strhet$-classes. Then a $\het$-class can be represented

854: by the coloured graph of any of the $\strhet$-classes of which it consists. 

855: 

856: The correspondence between $\het$-classes and coloured graphs gives us a practical way

857: of computing compositions of $\het$-classes. Let $A\in S^3_{min}[F',F]$ and $B\in S^3_{min}[F,F'']$ and

858: $f$ a map colouring the components of $A$ and $B$ so that no colour appears in both $A$ and $B$. To

859: compute $[A]_\het \cdot [B]_\het$ we first draw the coloured trees $\phi(A,f)$ and 

860: $\phi(B,f)$. If the group $\Pi_F$ has only one element (the identity) there is a well-defined 

861: correspondence between the $F$-edges of $\phi(A,f)$ and the $F$-edges of $\phi(B,f)$. 

862: We then identify the colours of every pair of edges that are in correspondence, and propagate this

863: identification to the $F'$-edges of $\phi(A,f)$ and the $F''$-edges of $\phi(B,f)$. The element of 

864: $\G(F'\dot\cup F'')^C$ obtained by joining $\phi(A,f)$ and $\phi(B,f)$ at the root after the identification 

865: and propagation of colours corresponds to $[A]_\het \cdot [B]_\het$. Therefore composition of

866: $\het$-classes coincides with the partition algebra composition of the connectivities if $\Pi_F$ is

867: trivial. This is not so if the group $\Pi_F$ has $n>1$ elements. Then there are $n$ ways in which 

868: the $F$-edges of $\phi(A,f)$ and the $F$-edges of $\phi(B,f)$ can be put in correspondence. The insertion 

869: of a complete set in Eq. (\ref{eq:therule}) corresponds to defining the composition to be the uniformly 

870: weighted sum of the $n$ possible outcomes. 

871: 

872: The power of $q$ in $[A]_\het\cdot[B]_\het$ is the number of colours 

873: after identification which do not propagate to the $F'$-edges of $\phi(A,f)$ or the $F'$-edges of $\phi(B,f)$. 

874: If the number of colours in $\phi(A,f)$ and $\phi(B,f)$ before and after identification are respectively $C_b$ and

875: $C_a$, then by Eq. (\ref{eq:genusformula}) the power of $p$ in $[A]_\het\cdot[B]_\het$ is $C_a-C_b+|F|$.

876: 

877: 

878: 

879: 

880: For example, consider again the concrete diagrams shown in Figure 2. Note first that $\Pi_F=1$, as the two 

881: concentric loops in $F$ cannot be interchanged by an isotopy. There are nine admissible colourings. Let 

882: $D_1,D_2,\cdots,D_9$ be the $\het$-classes defined by those admissible colourings, in the order shown in 

883: the right-hand side of Figure 3 (colours are indicated by letters r, y, b and g). The composition table for 

884: these heterotopy classes is then the one shown at the beginning of this section.

885: 

886: 

887: \begin{figure}

888: \centering

889: \leavevmode

890: \epsfbox{ symex.eps  }

891: \caption{Boundary configuration $(G,G)$ with $\Pi_G=\Z_2$, and two admissible colourings 

892: $D_1$ and $D_2$.}

893: \label{fig:tc}

894: \end{figure}

895: 

896: 

897: 

898: This example is atypical in that the loop configuration $(F,F)$ has $\Pi_F=1$. Let us now consider the

899: boundary configuration shown in Figure 4. There are now two different ways in which the coloured edges of 

900: two concrete diagrams can be put in correspondence. 

901: Taking the coloured trees shown in Figure 4, we find that the two elements of $\Pi_G$ give rise to

902: two different contributions to the composition $D_1\cdot D_1$, one of which is $qD_1$ and the other $D_2$.

903: Therefore we find that $D_1\cdot D_1={1\over 2}(qD_1+D_2)$.

904: 

905: \medskip

906: 

907: A more detailed analysis of the algebras defined here is given in \cite{AM}, together with a preliminary analysis 

908: of their representation theories, which are intriguingly much richer than the original Temperley-Lieb algebra

909: itself. 

910: 

911: \begin{thebibliography}{99}

912: 

913: \bibitem{adamovich} A. M. Adamovich, {\it Composition factors and submodules of Weyl modules}, 

914: Ph.D. Thesis, Moscow State University, 1992.

915: \bibitem{armstrong} M. A. Armstrong, {\it Basic Topology}, McGraw Hill, 1979.

916: \bibitem{baez} J. Baez and J. Dolan, {\it Higher-dimensional algebra and topological quantum field theory},

917: Jour. Math. Phys. 36 (1995), 6073-6105.

918: %\bibitem{baxter} R. J. Baxter, J. Stat. Phys. {\bf 8} (1973), 25.

919: \bibitem{baxterbook} R. J. Baxter, {\it Exactly solved models in statistical mechanics}, 

920: Academic Press, New York, 1982.

921: \bibitem{bb} ------ and V. V. Bazhanov, {\it  New solvable lattice models in three-dimensions}, 

922: J. Statist. Phys. 69 (1992), 453-585.

923: 

924: \bibitem{BN} H. W. Blote and M. P. Nightingale, {\it Critical behaviour of the two-dimensional Potts model with a 

925: continuous number of states; a finite size scaling analysis}, Physica 112{\bf A} (1982), 405-465.

926: \bibitem{dasmahapatra} S. Dasmahapatra and P. P. Martin, {\it On the algebraic approach to cubic lattice Potts 

927: model}, J. Phys. A {\bf 29} (1996), 263-278.

928: \bibitem{jimbo} M. Jimbo, {\it A q difference analog of U(g) and the Yang-Baxter equation}, 

929: Lett. Math. Phys. {\bf 10} (1985), 63-69.

930: \bibitem{bbtetra} R. M. Kashaev, V. V. Mangazeev and Yu. G. Stroganov, 

931: {\it Star square and tetrahedron equations in the Baxter-Bazhanov model}, 

932: Int. J. Mod. Phys. {\bf A} 8 (1993), 1399-1410.

933: \bibitem{kassel} C. Kassel, {\it Quantum Groups}, Springer-Verlag, 1995.

934: \bibitem{kaufman} L. Kaufman, {\it Knots and Physics}, World Scientific, 1993.

935: \bibitem{khovanov} M. Khovanov, {\it A categorification of the Jones polynomial}, arXiv:math/9908\\171; 

936: M. Khovanov and P. Seidel, {\it Quivers, Floer cohomology, and braid group actions}, arXiv:math/0006056.

937: %J. Bernstein, M. Khovanov and I. Frenkel, 

938: %{\it A categorification of the Temperley-Lieb algebra and Schur quotients of U(sl(2)) via projective and 

939: %Zuckerman functors}, Selecta Math. (N.S.) {\bf 5} (1999), no. 2, 199--241.

940: \bibitem{koo} W. M. T. Koo and H. Saleur, {\it Representations of the Virasoro algebra from lattice models},

941: Nucl. Phys. {\bf B} 426 (1994), 459-504.

942: \bibitem{laures} G. Laures, {\it On cobordism of manifolds with corners}, 

943: Trans. Am. Math. Soc. 352 no. 12, 5667-5688, 2000.

944: \bibitem{ppm} P. P. Martin, {\it Potts models and related problems in statistical mechanics}, World Scientific, 

945: Singapore, 1991.

946: \bibitem{ppmtl} ------, {\it Temperley-Lieb algebras for non-planar statistical mechanics --- the partition algebra 

947: construction}, Journal of Knot Theory and its Ramifications, {\bf 3} (1994), no. 1, 51-82.

948: \bibitem{ppmpart} ------, {\it The partition algebra and the Potts model transfer matrix spectrum in high

949: dimensions}, J. Phys. A {\bf 32} (2000), 3669-3695.

950: \bibitem{AM} ------ and M. Alvarez, {\it A Temperley-Lieb category for 2-manifolds}, in preparation.

951: \bibitem{ppmsaleur} ------ and H. Saleur, {\it On an algebraic approach to higher dimensional Statistical

952: Mechanics}, Commun. Math. Phys. {\bf 158} 155-190 (1993).

953: \bibitem{mendels} B. Mendelson, {\it Introduction to Topology}, Blackie, 1962.

954: \bibitem{Milnor} W. Milnor, {\it Morse Theory}, Princeton University Press, 1963.

955: \bibitem{Moise} E. E. Moise, {\it Geometric Topology in Dimensions 2 and 3}, Springer-Verlag, 1977.

956: %\bibitem{smartin} S. Martin, {\it Schur algebras and representation theory}, Cambridge University Press, 

957: %1993.

958: %\bibitem{nakano} S. Nakano and T. Uno, {\it Efficient generation of rooted trees}, NII Technical 

959: %Report (2003), http://research.nii.ac.jp/TechReports/03-005E.pdf

960: \bibitem{TL} H. N. V. Temperley and E. H. Lieb, {\it Relations between the 'percolation' and 'colouring' problem 

961: and other graph-theoretical problems associated with regular planar lattices: some exact results for the 

962: 'percolation' problem}, Proc. Roy. Soc. Lond. A 322 (1971), 251-280.

963: %\bibitem{wilf} H. S. Wilf, {\it Generatingfunctionology}, internet edition, Academic Press, 1994.

964: 

965: \end{thebibliography}

966: 

967: \end{document}

968: 

969: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

970: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

971: 

972: 

973: