1: \documentclass[11pt]{report} 

2: 

3: %\textwidth14cm

4: %

5: \textwidth16.3cm

6: \textheight23.3cm

7: \addtolength{\evensidemargin}{-1.9cm}

8: \addtolength{\oddsidemargin}{-1.9cm}

9: \addtolength{\topmargin}{-.9in}

10: %

11: %\def\baselinestretch{1.4}

12: %\pagestyle{empty}

13: \def \dfll {\leaders \hbox to 1em {\hss.\hss}\hfill}

14: \def\beq {\begin{equation}}

15: \def\eeq {\end{equation}}

16: \newtheorem{theorem}{Theorem}

17: \newtheorem{propos}[theorem]{Proposition}

18: \newtheorem{cor}[theorem]{Corollary}

19: \newtheorem{conjecture}[theorem]{Conjecture}

20: \newtheorem{rem}[theorem]{Remark}

21: \newtheorem{defi}{Definition}

22: \newtheorem{qes}{Question}

23: \newtheorem{lemma}{Lemma}

24: \def\blm{\begin{lemma}}

25: \def\elm{\end{lemma}}

26: \def\bdf{\begin{defi}}

27: \def\edf{\end{defi}}

28: \def\btm{\begin{theorem}}

29: \def\etm{\end{theorem}}

30: \def\bpp{\begin{propos}}

31: \def\epp{\end{propos}}

32: \def\bQ {\begin{qes}}

33: \def\eQ {\end{qes}}

34: \def\btm{\begin{theorem}}

35: \def\etm{\end{theorem}}

36: \def\ben{\begin{enumerate}}

37: \def\een{\end{enumerate}}

38: \def\into{\hookrightarrow}

39: %\def\bar{\begin{array}}

40: %\def\ear{\end{array}}

41: %\def\ep {\hfill{\rule{2.3mm}{2.3mm}}\medskip}

42: \def\ep{\ \hfill{\rule {2.5mm}{2.5mm}}\smallskip}

43: %\pagestyle{myheadings}

44: %\markboth{Thomas Kerler}{Homology TQFT's via Hopf algebras}

45: 

46: %\newcommand{\lbl}[1]{\label{#1}\mbox{[[{\bf\footnotesize#1}]]}}

47:  

48: \newcommand{\lbl}[1]{\label{#1}}

49: 

50: \def \isto {\widetilde{\longrightarrow}}

51: 

52: \def \Cob {{\cal C}ob_3}

53: 

54: \newcounter{cob}

55: 

56: \def\Sp{{\rm Sp}}

57: \def\SL{{\rm SL}}

58: 

59: 

60: \def\opc{{\scriptstyle =}}

61: 

62: \def\drulefill{$\mathord\opc\mkern-3mu\cleaders

63:                   \hbox{$\mkern-2mu\mathord\opc\mkern-2mu$}\hfill\mkern-2mu \mathord\opc$}

64: 

65: 

66: \newcommand{\dov}[1]{\vbox{\ialign{##\crcr\noalign

67:              {\kern-3pt\nointerlineskip}\drulefill \crcr\noalign

68:              {\kern0pt\nointerlineskip}

69:              $\hfil\displaystyle{#1}\hfil$\crcr}}}

70:  

71:  

72:  

73: \def\zer#1{\vbox{\ialign{##\crcr\noalign

74:              {\kern-3pt\nointerlineskip}

75:              {$\hfil\;\,\scriptscriptstyle\oslash\hfil\!$} \crcr\noalign

76:              {\kern.5pt\nointerlineskip}

77:              $\hfil\displaystyle{#1}\hfil$\crcr}}\!}

78: 

79: \def\ste#1{\vbox{\ialign{##\crcr\noalign

80:              {\kern-3pt\nointerlineskip}

81:              {$\hfil\;\,\scriptscriptstyle\bowtie\hfil\!$} \crcr\noalign

82:              {\kern.5pt\nointerlineskip}

83:              $\hfil\displaystyle{#1}\hfil$\crcr}}\!}

84: 

85: 

86: 

87: \newcommand{\TO}[2]{\stackrel {\mbox{#1}}{\hbox to #2pt{\rightarrowfill}}}

88: 

89: %\def\twoheadrightarrow{\to\mkern-20mu\to}

90: \def\thrafill{$\mathsurround=0pt \mathord- \mkern-6mu 

91: \cleaders\hbox{$\mkern-2mu

92: \mathord- \mkern-2mu$}\hfill \mkern-6mu\mathord\twoheadrightarrow$}

93: \newcommand {\onto} [1]{\hbox to #1pt{\thrafill}}

94: \newcommand {\ONTO} [2]{\stackrel{\mbox{#1}}{\onto {#2}}}

95: 

96: \newcommand{\emptystuff}[1]{}

97: 

98:  

99: %\def\ltimes {\mbox{\raisebox{.4ex}{$\scriptstyle \rhd\mkern-4mu<$}}}

100: %\def\ltimes {\mbox{$\rhd\!\!\!<$}}

101: % \def\rtimes {\mbox{\raisebox{.4ex}{$\scriptstyle >\mkern-4mu\lhd$}}}

102: 

103: 

104: \usepackage{amsfonts,

105:             amssymb}

106: %\usepackage{diagrams}      

107: 

108: \input{epsf.sty}    

109: 

110: 

111: \newcommand{\Z}{{\mathbb Z}}

112: \newcommand{\D}{{\mathbb D}}

113: \newcommand{\Ll}{{\mathbb L}}

114: \newcommand{\R}{{\mathbb R}}

115: \newcommand{\Cc}{{\mathbb C}}

116: \newcommand{\N}{{\mathbb N}}

117: \newcommand{\Q}{{\mathbb Q}}

118: \newcommand{\E}{{\mathbb E}}

119: \newcommand{\Ss}{{\mathbb S}}

120: \newcommand{\Aa}{{\mathbb A}}

121: \newcommand{\Pp}{{\mathbb P}}

122: 

123: \newcommand{\kk}{{\mathbb K}}

124: 

125: \newcommand{\posit}[2]

126: {\raise -1.4ex\hbox{${\textstyle #1}\atop {\stackrel{\uparrow}{#2}}$}}

127: 

128: 

129: \newcommand{\head}[1]{

130: \smallskip 

131: 

132: \begin{center}{\large \sc #1}\end{center}

133: 

134: \nopagebreak}      

135: 

136: \newcommand {\prg}[2]{\vspace*{.5cm}\subparagraph{ #1)\ #2\ }\lll{pg-#1}\ \newline}

137: 

138: \def\ub{\underline}

139: \def \A {\mbox{$\cal N\,$}}

140: \def \F {I\mkern -6.2mu  F}

141: \def \V {\mbox{$\cal V\,$}}

142: \def \C {\mbox{$\cal C\,$}}

143: \def \Am {\mbox{${\cal N}\!-\!mod\,$}}

144: \def\id{{1\mkern-5mu {\rm I}}}

145: \def \li {\mbox{$\cal L\,$}}

146: \def \lz  {\langle}

147: \def \rz  {\rangle}

148: 

149: \def \th {\theta}

150: \def \tb {\bar{\theta}}

151: \def \tx {{\theta^{\#}}}

152: 

153: \def \sempro {\mbox{$\bigcirc\mkern-14mu s\,\,$}}

154: 

155: \def \hopf {\mbox{$^0\bigcirc\mkern-13mu\bigcirc^0\,$}}

156: \def \nhopf {\mbox{$^0\bigcirc\mkern-13mu\bigcirc\,$}}

157: \def \twhopf {\mbox{$^0\bigcirc\mkern-13mu\bigcirc^1\,$}}

158: 

159: \def\B{{\cal B}}

160: \def\Bo{{\B}^{\circ}}

161: 

162: \def\Hh{\mbox{\bm${\cal H}$\ubm}}

163: 

164: \def\rcoa{\leftharpoondown}

165: \def\lcoa{\rightharpoonup}

166: \def\rac{\!\triangleleft\!}

167: \def\lac{\!\triangleright\!}

168: 

169: \newcommand{\ext}[1] {\mbox{\raisebox{.4ex}{$\bigwedge^{\!#1}$}}\mkern-1mu}

170: 

171: \def\Ii {\mbox{\raise .4 ex\hbox{$\int$}$\!\! I$}}

172: 

173: \def \tgl{\mbox{\boldmath${\cal T}\! gl$\unboldmath}}

174: \def \stA{\mbox{\boldmath$s\!{\cal T}$\unboldmath$({\cal N})$}}

175: 

176: \def \bm {\boldmath}

177: \def \ubm {\unboldmath}

178: \newcommand{\Qq}

179: {\mbox{\bm${\cal Q}$\ubm} \,}

180: 

181: 

182: 

183: \def \vc {Vect({\bf C})}

184: %\newcommand{\lll}[1]{\label{#1}\raisebox{1ex}{\mbox{\scriptsize [#1]$\;$}}}

185: \def\lll{\label}

186: 

187: 

188: \begin{document} 

189: 

190: \vspace*{1cm}

191: 

192: 

193:  

194: \begin{center}

195: 

196: 

197: \section*{Homology TQFT's and  the Alexander-Reidemeister 

198: Invariant of 3-Manifolds via Hopf Algebras and Skein Theory }

199: 

200: \bigskip

201: 

202: \medskip

203: 

204: 

205: {\large Thomas Kerler}\\

206: \medskip

207: 

208: May 2001

209:  \vspace*{1.2cm}

210: 

211: \bigskip

212: 

213:  

214: 

215: \end{center}

216: %\vspace*{1.7cm}

217: 

218: 

219: {\small \noindent{\bf Abstract :}

220: We develop an explicit  skein theoretical algorithm to compute 

221: the Alexander polynomial of a 3-manifold from a

222: surgery presentation employing the methods used in 

223: the construction of quantum invariants of 3-manifolds. 

224: As a prerequisite we establish and prove a rather unexpected 

225: equivalence between the topological quantum field

226: theory constructed by Frohman and Nicas using the

227: homology of $U(1)$-representation varieties

228: on the one side and the combinatorially constructed Hennings-TQFT based on the 

229: quasitriangular Hopf algebra ${\cal N}=\Z/2\ltimes \ext *\R^2$ on the

230: other side. We find that both TQFT's are $\SL(2,\R)$-equivariant functors and, 

231: as such, are isomorphic.  The

232: $\SL(2,\R)$-action in the Hennings construction comes from the natural

233: action on $\cal N$ and in the case of the Frohman-Nicas theory from the

234: Hard-Lefschetz decomposition of the $U(1)$-moduli spaces given that they 

235: are naturally

236: K\"ahler. The  irreducible components of this TQFT, corresponding  to simple 

237: representations of $\SL(2,\Z)$ and $\Sp(2g,\Z)$, thus yield a large family of 

238: homological TQFT's by taking sums and products. 

239: We give several examples of TQFT's and invariants

240: that appear to fit  into this family, such as Milnor and Reidemeister Torsion,  

241: Seiberg-Witten theories, Casson type theories for homology circles \'a la Donaldson, 

242: higher rank gauge theories following Frohman and Nicas, and the $\Z/p\Z$ reductions of

243: Reshetikhin-Turaev theories over the cyclotomic integers $\Z[\zeta_p]$. We also

244: conjecture that the Hennings TQFT for quantum-${\mathfrak s}{\mathfrak l}_2$   

245: is the product of the Reshetikhin-Turaev TQFT and such a homological TQFT. 

246:  \footnote{

247: 2000 Mathematics Subject Classification: Primary 57R56; Secondary 14D20, 16W30, 17B37, 18D35, 57M27.}

248: 

249: }

250: \medskip

251: 

252: \begin{center}

253: {\large \sc Contents}\

254: \smallskip

255: 

256: \parbox[t]{12cm}{

257: 1. Introduction\dotfill\pageref{S1}

258: 

259: 2. Topological Quantum Field Theory\dotfill\pageref{S2}

260: 

261: 3. The Frohman-Nicas TQFT for $U(1)$\dotfill\pageref{S3}

262: 

263: 4. The Mapping Class Groups and their Actions on Homology\dotfill\pageref{S4}

264: 

265: 5. Hennings TQFT's\dotfill\pageref{S5}

266: 

267: 6. The Algebra $\cal N$\dotfill\pageref{S6}

268: 

269: 7. The Hennings TQFT for $\cal N$\dotfill\pageref{S7}

270: 

271: 8. Skein theory for ${\cal V}_{\cal N}$\dotfill\pageref{S8}

272: 

273: 9. Equivalence of ${\cal V}_{\cal N}^{(2)}$ and ${\cal V}^{FN}$\dotfill\pageref{S9}

274: 

275: 10. Hard-Lefschetz Decomposition\dotfill\pageref{S10}

276: 

277: 11. Alexander-Conway Calculus for 3-Manifolds\dotfill\pageref{S11}

278: 

279: 12. Lefschetz compatible Hopf Algebra Structures on $H^*(J(\Sigma))$\dotfill\pageref{S12}

280: 

281: 13. More Examples of Homological TQFT's and Open Questions\dotfill\pageref{S13}

282: }

283: \end{center}

284: 

285: 

286: \head{1. Introduction}\lbl{S1}

287: 

288: 

289:  

290: 

291: In recent years much energy has been put into  finding new ways to

292:  describe and compute  classical invariants of 

293: 3-manifolds using the tools and structures developed in the relatively new area

294: of quantum topology. In this paper we will establish another such relation between

295: quantum and classical invariants, which emerged in quite different guises in recent

296: research in 3-dimensional topology. 

297: 

298: 

299: The classical invariant of a 3-manifold $M$ we are interested in here is its

300: Alexander polynomial $\Delta(M)\in\Z[H_1(M)]$. It is closely related and in most cases identical 

301: to the Reidemeister Milnor Torsion $r(M)$, see \cite{Mil61} and \cite{Tur76}. More recently,

302: Meng and Taubes \cite{MengTaub} show that this invariant is also equal to the 

303: Seiberg Witten invariant for 3-manifolds. Turaev  \cite{Tur98} proves a refined version of this

304: theorem by comparing the behavior of both invariants under surgery.

305: 

306: 

307: On the side of the quantum invariants we consider the formalism used for the  

308: Hennings invariant of 3-manifolds \cite{Hen96}. This invariant is motivated by 

309: and follows the same principles as the Witten-Reshetikhin-Turaev invariant,

310: which is  developed in  \cite{Wi89},  \cite{ResTur91} and \cite{TurVi92}, in the sense

311: that it 

312: assigns algebraic data to a surgery presentation for $M$. The innovation of the Hennings

313: approach is that  it starts

314: directly from a possibly non-semisimple  Hopf algebra ${\cal A}$ rather than its 

315: semisimple representation theory. This formalism is refined by Kauffman and Radford in

316:  \cite{KauRad95}. Also Kuperberg \cite{Kup96} gives a construction that  assigns data 

317: directly from a Hopf algebra to a Heegaard presentation of  $M$.

318:  

319: 

320: In this article we discover and explain in detail the relation between the Hennings theory 

321: for a certain 

322: 8-dimensional Hopf algebra $\cal N$ and the  (reduced) Alexander polynomial 

323: $\Delta_{\varphi}(M)\in \Z[t,t^{-1}]$ for the cyclic covering given by an epimorphism

324: $\varphi:\pi_1(M)\to\Z$. As a consequence we have at our disposal 

325: the entire combinatorial machinery of the Hennings formalism 

326:  in order to evaluate the Alexander polynomial from surgery

327: diagrams. Particularly, we are able to develop from this an efficient skein theoretical

328: algorithm. 

329: The method of relating these two very differently defined theories is based 

330: itself on a quite unexpected equivalence of more refined structures. 

331: 

332: More precisely, it turns out that underlying  both invariants is the structure of a 

333: topological quantum field theory (TQFT). The notion of a TQFT, which can be thought

334: of as a fiber functor on a category of cobordisms, was first cast into a mathematical 

335: axiomatic framework by Atiyah \cite{Ati88}. Typically (or by definition)

336:  all quantum invariants extend to TQFT's on 3-manifolds with boundaries. 

337: In the case of the semisimple theories generalizing the Witten-Reshetikhin-Turaev

338: invariant these TQFT's are described in great detail in \cite{Tur94}. In our

339: context we need the non-semisimple version as it is worked out for the Hennings invariant

340: in \cite{Ker96} and in full generality in \cite{KerLub00}. 

341:  

342: 

343: On the side of the classical invariants Frohman and Nicas \cite{FroNic92} 

344: managed to give an interpretation of the Alexander polynomial of knot complements  

345: in the setting of TQFT's. In particular, they 

346: construct a TQFT ${\cal V}^{FN}$, which  assigns to every surface $\Sigma$ as a vector

347: space the cohomology  ring $H^*(J(\Sigma))$

348: of the $U(1)$-representation variety $J(\Sigma)=Hom(\pi_1(\Sigma),U(1))$.

349: The morphisms are constructed in the style of the Casson invariant from the intersection

350: numbers of representation varieties for a given Heegaard splitting of a cobordism. The

351: Alexander polynomial is thus given as the Lefschetz trace over  ${\cal V}^{FN}(C_{\Sigma})$,

352: where $\Sigma$ is an arbitrary Seifert surface and $C_{\Sigma}$ is the 3-dimensional

353: cobordisms from $\Sigma$ to itself, obtained by cutting away a neighborhood of $\Sigma$.  

354:  

355: 

356:   The unexpected upshot is that this functor  ${\cal V}^{FN}$ is isomorphic to the  

357: Hennings TQFT ${\cal V}_{\cal N}$ for the non-semisimple Hopf algebra 

358: ${\cal N}\cong \Z/2\ltimes \ext *\R^2$. The realization of the abelian gauge    

359: field theory by a specific Hopf algebra 

360: is not at all obvious since ${\cal V}^{FN}$ and ${\cal V}_{\cal N}$ are

361: defined in entirely different ways. In fact the isomorphism between these functors

362: on the vectors spaces mixes up the degrees of exteriors algebras in still 

363: puzzling ways.  For these reason the proof is rather explicit and computational. 

364: 

365: Nonetheless, it can be seen quite easily that it is not possible to realize

366: ${\cal V}^{FN}$ as a semisimple theory.  

367:  Particularly,  ${\cal V}^{FN}$ represents  Dehn twists  by 

368: matrices of the form $1+N$ where $N$ is nilpotent. Furthermore, the invariant vanishes

369: on $S^1\times S^2$. Yet, in the semisimple

370: theories from \cite{Tur94} Dehn twists are represented by  

371: semisimple matrices $D$ with

372: $D^n=1$ and the invariant on $S^1\times S^2$ is never zero. 

373: 

374: 

375:  

376: 

377: Once ${\cal V}^{FN}$ and thus the Alexander polynomial $\Delta_{\varphi}$

378: are translated into the language of the Hennings formalism for the 

379: Hopf algebra ${\cal N}$ we are  in the position to develop a skein theory 

380: for the computation of $\Delta_{\varphi}$. The skein identities reflect 

381: algebraic relations in  ${\cal N}$. We derive from this a step by step recipe 

382: for the computation of the Alexander polynomial. 

383:  

384: 

385: Another intriguing feature of the two TQFT's is that both of them admit natural 

386: equivariant $\SL(2,\R)$-actions that have very different origins but are,

387: nevertheless, intertwined by the isomorphism between them. In the case of ${\cal V}^{FN}$ the 

388: $\SL(2,\R)$-action on  $H^*(J(\Sigma))$

389:  is given by the Hard Lefschetz decomposition of  the cohomology ring 

390: that arises from a K\"ahler structure on $J(\Sigma)$. For ${\cal V}_{\cal N}$

391: this action is derived from an $\SL(2,\R)$-actions on ${\cal N}$ as a Hopf algebra. 

392: As a consequence   $H^*(J(\Sigma))$

393: carries a nonstandard ring-structure induced by that of ${\cal N}^{\otimes g}$,

394: which, as opposed to the standard one,

395: is compatible with the Hard Lefschetz $\SL(2,\R)$-action. 

396: \medskip

397: 

398: Let us summarize the content and the

399: main results of this paper in better order and  detail. In Section~2  

400: we recall relevant 

401: notions that characterize 

402:  topological quantum field theories, such as (non)semisimplicity. Section~3 reviews

403: the  construction of the functor ${\cal V}^{FN}$

404: of Frohman and Nicas and its values  on basic cobordisms. In Section~4 we

405: describe a convenient set of

406: generators of the mapping class groups as combinations of Dehn twists and tangles,  

407: and determine 

408: their actions on homology. Section~5 introduces the basic rules for the construction of

409: a Hennings TQFT as well as a method that allows us to construct TQFT's even from

410: {\em non-modular}

411:  Hopf algebras or categories. In Section~6 we give the precise definition of  $\cal N$ as 

412: a quasi triangular Hopf algebra in the sense of Drinfel'd together with the 

413: $\SL(2,\R)$-action on it. The vector spaces and the basic morphisms of the associated Hennings 

414: TQFT are computed in Section~7 using standard tangle presentations. 

415: We prove $\SL(2,\R)$-covariance and single out an index 2

416:  subcategory of framed cobordisms that naturally

417: yields a {\em real} valued TQFT. For later applications

418: we also determine  the {\em categorical}

419:  Hopf algebra that is canonically associated to this TQFT. 

420: The nilpotent braided structure of $\cal N$ is then used in Section~8 to develop

421: a skein theory for the evaluation of tangle diagrams. The pivotal equivalence

422: of TQFT's that relates this theory to the Alexander polynomial is given by a natural 

423: isomorphism of  functors as follows. This is

424:  proven in Section~9 by explicit comparison of generating morphism. 

425:  

426: \begin{theorem}\lbl{thm-main}

427: There is an $\SL(2,\R)$-equivariant isomorphism  

428: $$

429: \xi\;:\;\;\; {\cal V}_{\cal N}^{(2)}

430: \;\;\stackrel{\bullet\,\,\cong}{-\!\!\!-\!\!\!-\-\!\!\!\!\!\longrightarrow}

431: \;\;{\cal V}^{FN}\;\;,

432: $$

433: where both TQFT's are ``non-semisimple'', 

434: $\Z/2\Z$-projective functors from the category $\Cob^{\bullet}$

435: of  surfaces with one boundary component and relative cobordisms to the category

436: of  real $\SL(2,\R)$-modules. 

437: \end{theorem}

438: 

439: The Hard Lefschetz $\SL(2,\R)$ action on the cohomology of the $U(1)$ moduli spaces and 

440: its covariance with ${\cal V}^{FN}$ are described more precisely in Section~10. The fact 

441: that $\xi$ is an $\SL(2,\R)$-equivariant transformation

442:  is proven. Moreover, we describe the canonical decompositions of the TQFT and the 

443: Alexander polynomial according to their dual $\SL(2,\R)$-representations. The summands

444: are irreducible TQFT's for which the mapping class groups are represented by  fundamental

445: weight representations of the symplectic groups $\Sp(2g,\Z)$. In 

446: Section~11 we use the equivalence from Section~9 and the skein theory for tangles

447: from Section~12 to lay out an explicit algorithm, based on a skein theory that extends

448: the Alexander-Conway calculus, for the computation of  $\Delta_{\varphi}(M)$. 

449: 

450: \begin{theorem}\label{them-skeintheory}

451:  Let $\cal L$ be a framed link and $\cal Z\subset \cal L$ a distinguished 

452: component that has zero framing and algebraic linking number zero with all other components.

453: Let $M_{\cal L}$

454: be the 3-manifold obtained by surgery along ${\cal L}$ and $\varphi_{\cal Z}:\pi_1(M)\to\Z$

455: the linking number with $\cal Z$. 

456: 

457: Then $\Delta_{\varphi_{\cal Z}}(M_{\cal L})\in\Z[t,t^{-1}]$ 

458: can be computed systematically as follows:

459: \begin{itemize}

460: \item Use the skein relations from Proposition~\ref{propos-Yskein} to unknot the special

461: strand $\cal Z$.

462: \item Put the new configuration into a standard form as depicted in Figure~\ref{fig-stan},

463: yielding a  tangle $\cal T$.

464: \item Use the skein relations from Theorem~\ref{thm-skein} and framing relations from

465: Figure~\ref{fig-frame} to decompose ${\cal T}^{\#}$ into elementary diagrams as

466: described in in Theorem~\ref{thm-solve}.

467: \item Translate the elementary tangle diagrams into Hopf algebra diagrams as in 

468: (\ref{eq-symhopfeva}). 

469: \item Go through the steps of Proposition~\ref{prop-HopfAlexeval} to assign 

470: polynomials to each component of a diagram. 

471: \item Take products over components and sums over elementary diagrams.

472: \end{itemize}

473: 

474: \end{theorem}

475: 

476: The calculus described here for the evaluation of tangle diagrams is precisely the

477: one used to compute the morphisms for the TQFT functors from Theorem~\ref{thm-main}

478: via tangle surgery presentations of cobordisms. 

479:  

480: 

481: Another application of the equivalence  established in

482: Theorem~\ref{thm-main} arises from the observation 

483: that every TQFT ${\cal V}$ on $\Cob^{\bullet}$ naturally implies a braided

484: Hopf algebra structure ${\cal H}_{\cal V}$ on ${\cal N}_0:={\cal V}(\Sigma_{1,1})$. 

485: Now, the cohomology ring $H^*(J(\Sigma_g, U(1)))\cong\ext *H_1(\Sigma_g)$ already has

486: a canonical structure ${\cal H}_{ext}$  of a $\Z/2$-graded Hopf algebra induced by

487: the group structure on $J(\Sigma_g, U(1))$. It is easy to see that 

488: ${\cal H}_{ext}$ is {\em not} compatible with the Lefschetz $\SL(2,\R)$-action.

489: However, the  braided Hopf algebra  

490: structure ${\cal H}_{{\cal V}^{FN}}$ inherited from the TQFT's in 

491: Theorem~\ref{thm-main} is naturally $\SL(2,\R)$-variant, and, furthermore,

492:  equivalent to ${\cal H}_{ext}$: 

493: 

494: \begin{theorem}\lbl{thm-structure}

495: For any choice of an integral Lagrangian decomposition, $H_1(\Sigma_g,\Z)=\Lambda\oplus\Lambda^*$,

496: and volume forms, $\omega_{\Lambda}\in\ext g\Lambda$ and $\omega_{\Lambda^*}\in\ext g\Lambda^*$, 

497: the space

498: $H^*(J(\Sigma_g))$ admits a canonical structure ${\cal H}_{\Lambda}$ of a $\Z/2$-graded

499: Hopf algebra. It coincides with the braided Hopf algebra structure induced

500: by ${\cal V}^{FN}$ and is isomorphic to the canonical structure ${\cal H}_{ext}$. 

501: 

502: In particular,

503: $(H^*(J(\Sigma_g)),{\cal H}_{\Lambda})$ is commutative and cocommutative in the graded

504: sense, with unit $\omega_{\Lambda^*}$, integral $\omega_{\Lambda}$, and primitive elements given by

505: $a\wedge\omega_{\Lambda^*}$ and $i^*_z\omega_{\Lambda^*}$ for $a\in H_1(\Sigma)$ and $z\in H^1(\Sigma)$.

506: 

507: The   structure  ${\cal H}_{\Lambda}$ is, furthermore, compatible with the 

508: Hard-Lefschetz $\SL(2,\R)$-action. Specifically, this action is the Howe dual to 

509: the action of $\SL(g,\Z)$ on the Lagrangian subspace in the group of Hopf

510: automorphisms:

511: $$

512: \SL(2,\R)_{Lefsch.}\times \SL(\Lambda)\,\;\subset \;\, {\rm GL}(2g,\R)= Aut(H^*(J(\Sigma_g)),{\cal H}_{\Lambda})

513: $$ 

514: \end{theorem} 

515: 

516: In Section~13 we   discuss the appearance of these TQFT's in other

517: contexts. To this end let us denote by  

518: ${\cal V}^{(j)}$ the irreducible component of ${\cal V}^{FN}$ dual to the $j$-dimensional

519: $\SL(2,\R)$-representation. A detailed  description of it is given in Theorem~\ref{cor-dec}.

520: Choose for a closed 3-manifold $M$ with Betti number

521: $\beta_1(M)\geq 1$ a surjection $\varphi:H_1(M)\onto{13}\Z$

522: (which would be canonical for homology circles as given by 0-surgeries on knots). A series

523: of  invariants for the pair $(M,\varphi)$ can now be constructed  by choosing any two-sided,

524: embedded surface $\Sigma \subset M$ that is dual to $\varphi$, and considering the cobordism

525: $C_{\Sigma}:\,\Sigma\to\Sigma$ obtained by removing an open tubular neighborhood of $\Sigma$

526: from $M$. The {\em $j$-th (fundamental) Alexander Character} is now defined to be  the integer

527: \beq\label{eq-AlexMom}

528: \Delta^{(j)}_{\varphi}(M)\;\;=\;\;trace\Bigl({\cal V}^{(j)}(C_{\Sigma})\Bigr)\;,

529: \eeq

530: which is easily seen to depend only on $\varphi$ but not the choice of $\Sigma$. 

531: Besides the Alexander Polynomial also two other invariants invariant $I^{SW}$ and

532: $I^{DC}$ depending this data have been constructed by Donaldson in \cite{Don99}

533: from a Seiberg-Witten Theory and an $SO(3)$-Casson-type gauge theory respectively.

534: Let us also denote by $\lambda_L$ the Lescop Invariant \cite{Les}. As specified in

535: the next theorem all of these invariants are in fact linear combinations

536: of the (fundamental) Alexander Characters. 

537: 

538: \begin{theorem}[{\rm\small mostly corollaries to} {\cite{FroNic92},  

539: \cite{Don99},\cite{Les},\cite{KerKyoto}}]\label{thm-relations}

540: \begin{eqnarray}

541: \Delta_{\varphi}(M)\;\;&=&\;\quad\sum_{j\geq 1}\,[j]_{-t}\cdot \Delta^{(j)}_{\varphi}(M)\label{eq-AlexChar}\;

542: \\

543: &&\nonumber\\ 

544: I^{DC}_{\varphi}(M)\;\;&=&\;\;\;\sum_{j\geq 2}\,{{j+1}\choose 3}\cdot \Delta^{(j)}_{\varphi}(M)\label{eq-DCChar}\;

545: \\

546: &&\nonumber\\

547: I^{SW}_{d,\varphi}(M)\;\;&=&\;\;\sum_{j\geq d+2}\,\left[\!\!{\left[

548: \Bigl({\frac{j-d}2}\Bigr)^2\right]}\!\!\right]\cdot

549: \Delta^{(j)}_{\varphi}(M)\label{eq-SWChar}

550: \\

551: &&\nonumber\\

552: \lambda_{L}(M)\;\;&=&\;\;\sum_{j\geq 1} (-1)^{j-1}\frac {j(2j^2-3)}{12}\cdot

553: \Delta^{(j)}_{\varphi}(M)\;\label{eq-LesChar}

554: \end{eqnarray}

555: \end{theorem}

556: Here we denoted $[j]_q=\frac {q^j-q^{-j}}{q-q^{-1}}$ and by $\bigl[\![x]\!\bigr]$ the

557: largest integer $\leq x$. We further review in how far the higher $PSU(n)$ knot invariants 

558: $I_{k,n,\varphi}^{FN}$

559: of Frohman and Nicas \cite{FroNic94} come out to be 

560: polynomial expressions in the Alexander Characters.

561: As products of characters are associated to tensor products of TQFT's and their decompositions

562: into irreducible components it is natural to  consider the corresponding higher,

563: irreducible Alexander Characters $\Delta^{(\gamma)}$. We conjecture that the  

564: $I_{k,n,\varphi}^{FN}$ are linear combinations of the $\Delta^{(\gamma)}$ with 

565: coefficients in ${\mathbb N}\cup\{0\}$ as it is the case for $I^{DC}$ and $I^{SW}$. 

566: 

567: %%%%%%%%%%%%%%%%  BEGIN CUT %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

568: 

569: \emptystuff{

570:  From this sequence of TQFT's  we are able to construct a large family of homological 

571: TQFT functors ${\cal V}^{(P)}$ for

572: suitable polynomials $P$ in variables $x_0, x_1, \ldots$ by taking tensor products and direct sums

573: accordingly. 

574: 

575:  One example of such a TQFT ${\cal V}^{(Y)}$ with $Y=x_0+x_3-x_5-x_8+x_{10}+\ldots\,$

576: taken over $\Z/5$ turns out describe the lowest order contribution

577: of the Reshetikhin Turaev invariant taken over the cyclotomic integers $\Z[\zeta_5]$ at least 

578: in small genera and very likely in general \cite{Kerr=5}. A rather interesting application that

579: emerges from that is found in in joint work with  Gilmer \cite{GilKer}. Namely, that non triviality

580: of the Alexander polynomial evaluated at a 5-th root of unity implies that a 3-manifold disconnects

581: if a surface with more than one component is removed from it. 

582:  

583:  The polynomial homological TQFT's appear to be also isomorphic to ones

584:  constructed by gauge theoretical means

585: using an approach similar to the Casson invariant or extending the methods

586: of Seiberg-Witten theory as described by Donaldson. Furthermore,  Frohman and Nicas consider

587: generalizations to higher rank Lie groups. In all cases the Alexander polynomial appears

588: as the dominant invariant, suggesting the corresponding decomposition of the TQFT into

589:  the basic functors ${\cal V}^{(j)}$. 

590: }

591: 

592: %%%%%%%%%%%%%%%%   END CUT %%%%%%%%%%%%%%%%%%%%%%%%

593: 

594: Moreover, we explain the irreducible  $p$-modular reductions  $\dov{\cal V}^{(j)}_p$

595: over ${\mathbb F}_p=\Z/p\Z$ of the ${\cal V}^{(j)}$ relate to the irreducible factors

596: of the $\Z[\zeta_p]\to{\mathbb F}_p$ of the

597: Reshetikhin Turaev TQFT's at a $p$-th root of unity $\zeta_p$. 

598:  We finally give evidence that  the TQFT from Theorem~\ref{thm-main}

599: is essentially the missing tensor factor that relates the semisimple and the non-semisimple

600: TQFT constructions for $U_q({\mathfrak s}{\mathfrak l}_2)$

601: following Reshetikhin Turaev and Hennings respectively. 

602: 

603: \medskip

604: 

605: 

606: \paragraph{Acknowledgements:} I am  indebted to Charlie Frohman for making me aware 

607: of \cite{FroNic92}  and explaining  \cite{FroNic94} to me. I thank 

608: Bernhard Kr\"otz for discussions about Howe pairs, and David Radford for helping me find

609: his example in \cite{Rad76}. Thanks also to Andrew Nicas and Hans Boden for their

610: interest and  Pierre Deligne, Daniel Huybrechts, and Manfred Lehn

611: for discussions about Lefschetz decompositions in the higher rank case.

612: Finally, I want to thank Razvan Gelca, Pat Gilmer, Jozef Przytycki 

613: for opportunities to speak about this paper, when it was still in its

614: early stages. 

615: \newpage 

616: 

617: \head{2. Topological Quantum Field Theory}\lbl{S2}

618: We start with the definition of  a TQFT as a functor as proposed by  

619: Atiyah \cite{Ati88}, largely suppressing a more detailed discussion 

620: of the tensor structures. 

621: 

622: 

623: For every integer, $g\geq 0$, choose a compact, oriented

624:  model surface, $\Sigma_g$, of genus $g\,$, and to a tuple of 

625: integers $\underline g=(g_1,\ldots,g_n)$ associate the ordered union

626: $\Sigma_{\underline g}:=\Sigma_{g_1}\sqcup\ldots\sqcup\Sigma_{g_n}$.

627: A {\em cobordism}  

628: is a collection, ${\bf M}=(M,\phi_{\#},\Sigma_{g_{\#}})$, of the following:

629: 

630: 

631:  A compact, oriented 3-manifold, $M$, whose boundary  is divided into 

632: two components 

633: $\partial M=-\partial_{in} M\sqcup\partial_{out} M$,  

634: two standard surfaces $\Sigma_{\underline g_{in}}$ and 

635: $\Sigma_{\underline g_{out}}$,   and two orientation preserving 

636: homeomorphisms $\phi_{in}:\Sigma_{\underline g_{in}}

637: \,\isto\, \partial_{in} M\,$  and $\phi_{out}:\Sigma_{\underline g_{out}}

638: \,\isto\, \partial_{out} M\,$. 

639:  

640: We say two cobordisms, ${\bf M}$ and ${\bf M}'$, are equivalent if

641: they have the same "in" and "out" standard surfaces, and  

642: there is a homeomorphism $h:\,M\,\isto \, M'\,$, such that 

643:  $h\circ \phi_{\#}=\phi_{\#}'\,$. 

644: 

645: Let $\Cob$ be the category of cobordisms in dimension 2+1, which has

646: the standard surfaces as objects and equivalence classes of cobordisms

647: as morphism. The composition of morphisms is defined via gluing over

648: boundary components using the coordinate maps to the same standard

649: surfaces.  In addition, $\Cob$ has a tensor

650: product given by disjoint unions of surfaces and cobordisms. 

651: 

652: 

653: A {\em Topological Quantum Field Theory} (TQFT)   is  a functor,

654: ${\cal V}:\Cob\longrightarrow{\rm Vect}(\kk)$, from the category of cobordisms

655: to the category of vector spaces over a field $\kk\,$. 

656: \medskip

657: 

658:   Let us recall next some generalizations of the definition given in

659: \cite{Ati88} that will be relevant for our purposes. 

660: By $\Cob^{2fr}$ we denote the category of 2-framed cobordisms, where we 

661: fixed some standard framings on the model surfaces $\Sigma_g\,$, see \cite{Ker99}. A 

662: {\em 2-framed TQFT} is now a functor

663:  ${\cal V}:\Cob^{2fr}\longrightarrow{\rm Vect}(\kk)$. The category of 2-framed  

664: cobordisms can be understood as a central extensions 

665: \beq\label{eq-cobext}

666: \,1\;\to\;\Z\;\longrightarrow\;\Cob^{2fr}\;\longrightarrow\;\Cob\;\to \;1\,

667: \eeq 

668: of the 

669: ordinary cobordism category, if restricted to connected cobordisms. Hence, an

670: irreducible 2-framed TQFT yields a {\em projective TQFT} 

671: since $\Z$ is presented as a scalar. 

672: See \cite{Ker99} for further descriptions of this extension in terms of signatures of

673: bounding 4-manifolds. 

674: 

675: 

676: For a group, $G$, we introduce the notion of a {\em $G$-equivariant TQFT}.

677: It is a functor, 

678: ${\cal V}:\Cob\longrightarrow G-{\rm mod}_{\kk}$, from the category of 

679: cobordisms to the category of finite dimensional $G$-modules over a field $\kk$. This means

680: that the linear map associated to any cobordism commutes with 

681: the action of $G$ on the vector spaces of the respective boundary

682: components. 

683: 

684: Recall also from \cite{Ker98} that a {\em half-projective} or {\em non-semisimple} 

685: TQFT is one in which functoriality is weakened and replaced by 

686: the composition law ${\cal V}(MN)=0^{\mu(M,N)}{\cal V}(M){\cal V}(N)\,$.

687: Here $\mu(M,N)=b(MN)-b(M)-b(N)\in\Z^{+,0}\,$, where $b(M)$ is the 

688: number of components of $M$ minus half the number of components of

689: $\partial M$. Note that $0^0=1$. 

690: 

691: We often call a cobordism for which  all (rational) homology comes from the homology

692: of the boundary {\em (rationally) homologically trivial} (r.h.t). More precisely, we mean by this 

693: that $i_*:H_1(\partial M,\Q)\to H_1(M,\Q)$ is onto. Typical examples of r.h.t.

694: cobordisms are the ones in (\ref{eq-defpsi}) and (\ref{eq-hdlcob})

695: below and closed, rational homology spheres. Examples of cobordisms that are not 

696: r.h.t. are any connected sums with closed manifolds $M$ with $\beta_1(M)\geq 1$. 

697: We find the following vanishing property:

698: \begin{lemma}[\cite{Ker98}]\lbl{lm-vanish}

699: If ${\cal V}$ is a non-semisimple TQFT, then for any cobordism $M$, 

700: $$

701: {\rm if}\quad {\cal V}(M)\, \neq \, 0\qquad{\rm then}\qquad 

702: M\;\;\;\mbox{is r.h.t.}

703: $$

704: \end{lemma} 

705: 

706: We further introduce $\Cob^{\bullet}$, the  category of cobordisms, for which 

707: the surfaces are connected and have exactly one boundary component. As objects we

708: thus use model surfaces $\Sigma_{g,1}$, such that $\Sigma_{g+1,1}$ is obtained

709: from $\Sigma_{g,1}$ by gluing in a torus, $\Sigma_{1,2}$, with two boundary 

710: components. Thus, we have a presentation  

711: \begin{equation}\lbl{eq-modelsurf}

712: \Sigma_{g,1}\;=\;\underbrace{\Sigma_1\#\ldots\#\Sigma_1\#\Sigma_{1,1}}_g

713: \qquad\qquad

714: \mbox{with inclusions}\qquad\quad \Sigma_{g,1}\;\subset\;\Sigma_{g+1,1}\;.

715: \end{equation}

716:  Instead of ordinary cobordisms we then consider {\em relative} ones. We finally introduce 

717: categories of cobordisms with 

718: combinations of these properties such as  $\Cob^{2fr, \bullet}$, the category of

719: 2-framed, relative cobordisms. 

720: \medskip

721: 

722:  For any homeomorphsim,  $\psi\in Homeo^+(\Sigma_g)$, of a surface to itself 

723: we define the cobordism 

724: \beq\lbl{eq-defpsi}

725: {\bf I}_\psi\;\;=\;\;(\Sigma_g\times[0,1], id\sqcup \psi, \Sigma_g\sqcup \Sigma_g)\quad. 

726: \eeq

727: The morphism $[{\bf I}_{\psi}]$

728: depends only on the isotopy class $\{\psi\}$ of $\psi$,  and the resulting map

729: $\Gamma_g\to Aut(\Sigma_g):\{\psi\}\mapsto [{\bf I}_{\psi}]$ from the mapping class 

730: group to the group of invertible

731: cobordisms on $\Sigma_g$ is an isomorphism, see \cite{KerLub00}. Consequently,

732: every TQFT defines a representation  of the mapping class group 

733: $\Gamma_g\to GL({\cal V}(\Sigma_g)):\,

734: \{\psi\}\mapsto {\cal V}([{\bf I}_{\psi}])\,$.

735: 

736: 

737: Moreover, let us introduce  special cobordisms

738: \beq\lbl{eq-hdlcob}

739: {\bf H_{g}^+}\;:=\;=( H_{g}^+,id\sqcup id, \Sigma_g\sqcup \Sigma_{g+1})\;,

740: \eeq

741: where $H^+_g$ is obtained by  adding a 

742: full 1-handle to the cylinder $\Sigma_g\times[0,1]$  

743: at two discs in $\Sigma_g\times 1$.

744: This is done in a way compatible with the choice of the model surfaces in

745: equation~(\ref{eq-modelsurf}). Another cobordism $H_{g}^-$ is built by gluing in

746: a 2-handle into the thickened surface $\Sigma_{g+1}\times[0,1]$ along a

747: curve  $b_{g+1}$, which lies in the added torus from (\ref{eq-modelsurf})

748: and  has geometric intersection number one with the meridian

749: of the 1-handle added by  $H_g^+$. From this we obtain a cobordism 

750: ${\bf H_g^-}=( H_g^-,\Sigma_{g+1}\sqcup\Sigma_g)$ in opposite 

751: direction, with the property that ${\bf H_{g}^-}\circ {\bf H_{g}^+}$ is 

752: equivalent to the identity. 

753: 

754: 

755: Basic Morse theory implies a Heegaard decomposition as follows for any cobordism 

756: \beq\lbl{eq-Heegaard}

757: {\bf M}\;\;\;\cong\;\;\;{\bf H_{g_2}^-}\circ {\bf H_{g_2+1}^-}\circ\ldots

758: \circ {\bf H_{N-1}^-}\circ {\bf I}_{\psi}\circ {\bf H_{N-1}^+}\circ\ldots\circ 

759: {\bf H_{g_1+1}^+}

760: \circ {\bf H_{g_1}^+}\;,

761: \eeq

762: where $\psi\in Homeo^+(\Sigma_N)$. Hence, a  TQFT is completely determined by

763: the induced representations of the mapping class groups and the maps 

764: ${\cal V}([{\bf H}^+_g])$ and ${\cal V}([{\bf H}^-_g])$. Therefore, any two TQFT's

765: coinciding on  the basic generators from (\ref{eq-defpsi}) and  (\ref{eq-hdlcob})

766: have to be equal. 

767: 

768: \head{3. The Frohman-Nicas TQFT for $U(1)$}\lbl{S3}

769: 

770: Let us review the basic steps in the construction of the topological

771: quantum field theory ${\cal V}^{FN}$ as given in \cite{FroNic92}

772: via intersection theory of $U(1)$-representation varieties:

773: 

774: For a compact, connected manifold $X$  its  $U(1)$-representation variety is

775: defined as

776: \beq\lbl{eq-defJX}

777: J(X)\quad:=\quad {\rm Hom}(\pi_1(X), U(1))\quad \cong \quad H^1(X,U(1))\;.

778: \eeq

779: Observe that $J(X)$ is  a manifold of dimension $\beta_1(X)$. Specifically, it is

780: a torus if $H_1(X,\Z)$ is torsion free, and a discrete group if $\beta_1(X)=0$. 

781: 

782: The vector space associated  to a surface $\Sigma_{\underline g}$ is given

783: by ${\cal V}^{FN}(\Sigma_{\underline g})

784: = H^*(J(\Sigma_{g_1})\times\ldots\times J(\Sigma_{g_N}) ,\R)$. 

785: 

786: We consider first  cobordisms, $M$, between  surfaces, $\partial_{in}M$ 

787: and $\partial_{out}M$,  that are  rationally homologically trivial in the

788: sense of Section~2. 

789: In this case the map $j:\,J(M)\to J(\partial_{in}M)\times J(\partial_{out}M)$ is a half

790: dimensional immersion. Thus the top form $\pm[J(M)]$ defines (up to sign)

791: a middle dimensional 

792: homology class in $H_*(J(\partial_{in}M),\R)\otimes H_*(J(\partial_{out}M),\R)$.

793: Using Poicar\'e Duality and the coordinate maps of the cobordism, 

794: the latter space is isomorphic to the space of linear maps

795: from ${\cal V}^{FN}(\Sigma_{\underline g_{in}})$ to 

796: ${\cal V}^{FN}(\Sigma_{\underline g_{out}})$. ${\cal V}^{FN}(M)$, for a  

797:  homologically  trivial cobordism $M$, is now the linear map  associated to 

798:  $j_*(\pm[J(M)])$. 

799: 

800: 

801: In the general case Frohman and Nicas define   ${\cal V}^{FN}(M)$ via

802: a Heegaard splitting of $M$ as in (\ref{eq-Heegaard}), and consider the intersection 

803: number of representation

804: varieties of the elementary thick surfaces with handles separated by the Heegaard surface. 

805: In the case where $H_1(\partial M,\R)\to H_1(M,\R)$ is not onto, i.e., $M$ is

806: not homologically trivial, these varieties

807: no longer transversely intersect so that ${\cal V}^{FN}(M)=0$. 

808:  

809: Regarding the composition structure ${\cal V}^{FN}$ has a couple of nonstandard

810: properties. For one, functoriality fails to hold when $M$ and $N$ are homologically

811: trivial but $M\circ N$ is not. Moreover, the orientation of the classes $\pm[J(M)]$ and cycles

812: cannot be chosen  consistently with composition so that a sign-projectivity persists.

813: Recall, however, that a 2-framed TQFT is really defined on the $\Z$-extensions of cobordisms

814: given in (\ref{eq-cobext}).

815: 

816: \begin{lemma}\lbl{lm-FNpart}

817: ${\cal V}^{FN}$ is  a non-semisimple, $\Z/2\Z$-projective TQFT in the sense of Section~2.

818: \end{lemma} 

819: 

820: The mechanism by which the universal $\Z$-extension is factored into  a $\Z/2\Z$-extension

821: is explained further for the quantum theory in Lemma~\ref{lm-mod2prop} and 

822: Proposition~\ref{pp-ATQFT} of Section~7. At least indirectly, we have thus related the

823: orientation ambiguities in \cite{FroNic92} to the usual framing ambiguities of quantum

824: theories. 

825: 

826: Now, in the $U(1)$ case $J(X)$ has a group structure itself, which induces a 

827: coalgebra structure  on the cohomology ring so that $H^*(J(X))$ is endowed

828: with a canonical Hopf algebra structure ${\cal H}_{ext}$. If $H_1(X)$ is torsion free 

829: then $H^*(J(X))$ is connected and we obtain a natural isomorphism

830: $H^*(J(X))\cong\ext * H_1(X)$ of $\Z/2$-graded Hopf algebras, and

831: $H_1(X)$ is the space of primitive elements. Hence, we can write for the vector spaces:

832: \begin{equation}

833: {\cal V}^{FN}(\Sigma_{\underline g})\;=\;\ext * H_1(\Sigma_{\underline g})\;.

834: \end{equation}

835: The representation of the mapping class group $\Gamma_{\underline g}$

836: on this space is given by the obvious action 

837: \beq\lbl{eq-FNmapcg}

838: {\cal V}^{FN}([{\bf I_{\psi}}])\;\;=\;\;\ext * [\psi]\qquad

839: \forall \{\psi\}\in\Gamma_g\;.

840: \eeq

841: Here, $[\psi]\in \Sp(H_1(\Sigma_{\underline g}))$ is the natural, induced action on homology.

842: For a connected surface $\Sigma_g$ we have the associated short exact sequence

843: \beq\lbl{eq-TGSseq}

844: 1\;\to\;{\cal J}_g\;\longrightarrow\;\Gamma_g\;\stackrel{\psi\mapsto[\psi]}

845: {-\!\!\!-\!\!\!-\!\!\!-\!\!\!\longrightarrow}\;\Sp(2g,\Z)\;\to\;1\;,

846: \eeq

847: where ${\cal J}_g$ is the Torelli group. 

848: 

849: Let  

850: ${\bf H_{g}^+}$ be the cobordism as defined in (\ref{eq-hdlcob}), and 

851: let $[a_{g+1}]$ be a generator of  

852: $ker(H_1(\Sigma_{g+1},\Z)\to H_1(H_{g+},\Z))$ seen as an element 

853: of $H_1(\Sigma_{g+1},\R)$. It is represented by the meridian $a_{g+1}$ of

854: the added handle. 

855: In a slight variation of the Frohman Nicas formalism we see that the associated

856: linear map is given as 

857: \beq\lbl{eq-FNmcg}

858: {\cal V}^{FN}({\bf H_{g}^+}):\ext * H_1(\Sigma_g)\longrightarrow\ext * H_1(\Sigma_{g+1})\;:

859: \quad \alpha\;\mapsto\;i_*(\alpha)\wedge [a_{g+1}]\;.

860: \eeq

861: Here we use the fact that  $H_1(\Sigma_{g,1})=H_1(\Sigma_g)$ so that the 

862: inclusion of surfaces in (\ref{eq-modelsurf}) implies also an inclusion

863: $i_*:H_1(\Sigma_g)\subset H_1(\Sigma_{g+1})$. 

864: 

865: 

866:  

867: Let ${\bf H_{g}^-}$ be the cobordism obtained by gluing a 2-handle along $b_{g+1}$

868: as defined above. We note that 

869: $H_1(\Sigma_{g+1})=H_1(\Sigma_{g})\oplus\lz[a_{g+1}],[b_{g+1}]\rz$ so that 

870: $\ext * H_1(\Sigma_{g+1})$ is the direct sum of spaces 

871: $V_1\oplus V_a\oplus V_b\oplus V_{a\wedge b}$ where 

872: $V_{x}=[x_{g+1}]\wedge \ext *H_1(\Sigma_{g})$. The linear map associated in \cite{FroNic92}

873: to ${\bf H_{g}^-}$ acts on $V_a$ as 

874: \beq\lbl{eq-FNgcm}

875: {\cal V}^{FN}({\bf H_{g}^-}): V_a\longrightarrow\ext * H_1(\Sigma_{g})\;:

876: \quad i_*(\alpha)\wedge [a_{g+1}]\;\mapsto\;\alpha

877: \eeq

878: and is zero on all other summands. 

879: 

880: 

881: % Furthermore, let us assume that we have chose an inner product 

882: % $\lz .,.\rz$ on $H_1(\Sigma_g)$ 

883: % such that $i_*$ is an isometrie, and $\|[a_{g}]\|=1$. 

884: % Together with the symplectic form on $H_1(\Sigma_g)$ this

885: % yields an almost complex structure $J$ with $J^2=-1$. 

886: % If we set $[b_{g+1}]:=J[a_{g+1}]$   

887: % we see that $a_g\cdot b_g=\pm 1$ and intersection 

888: % numbers with cycles from $i_*(H_1(\Sigma_g))$ are all zero. 

889: 

890: % We assume that the attaching curve $b_{g+1}$ for the 2-handle in the 

891: % construction of $H_{g-}$ is a representative of  $[b_{g+1}]$.

892: % The associated  linear map comes out to be

893: % $$

894: % {\cal V}^{FN}({\bf H_{g}^-})\quad=\quad {\cal V}^{FN}({\bf H_{g}^+})^T

895: % $$

896: % where $.^T$ denotes the transpose with respect to $\lz .,.\rz$. 

897: 

898: 

899: \head{4. The Mapping Class Groups and their Actions on Homology}\lbl{S4}

900: The mapping class group

901: $\Gamma_{g,1}=\pi_0(Homeo^+(\Sigma_{g,1}))$ on a model surface

902: $\Sigma_{g,1}$ is generated by the right handed 

903: Dehn twists along oriented curves $a_j$, $b_j$, 

904: and $c_j$,   as depicted in

905: Figure~\ref{fig-modsurf}. We denote them by capital letters

906:  $A_j,\,B_j,\,C_j\, \in\Gamma_{g,1}$ respectively.

907: In fact we only need the $A_2$ of the $A_j$'s to generate 

908: $\Gamma_{g,1}$. A presentation of $\Gamma_{g,1}$ in 

909: these generators is given by Wajnryb \cite{Waj83}. 

910: \begin{figure}[ht]

911: \begin{center}

912: \leavevmode

913: \epsfxsize=7cm

914: \epsfbox{mcg-gen.eps}

915: \end{center}

916: \caption{Curves on $\Sigma_{g,1}$}\lbl{fig-modsurf}

917: \end{figure}

918: For our purposes we prefer the  set $\{A_j, D_j, S_j\}$ of generators

919: defined as follows: 

920: \begin{equation}\lbl{eq-newgen}

921: D_j\;:=\;A_j^{-1}A_{j+1}^{-1}C_j\qquad\mbox{and}\qquad S_j\;:=\;A_jB_jA_j\qquad\quad

922: \mbox{for}\;\;j=1,\ldots,g \;.

923: \end{equation}

924: In \cite{MatPol94} a tangle presentation of $\Gamma_{g,1}$ is given using the results

925: in \cite{Waj83}. The same presentation results from the tangle presentation of 

926: $\Cob^{2fr,\bullet}$ in \cite[Proposition 14]{Ker99}, which extends to the central

927: extension $1\to\Z\to\Gamma_{g,1}^{2fr}\to\Gamma_{g,1}\to 1$ that stems from the 

928: 2-framing of cobordisms. The framed tangles associated to our preferred generators

929: are given in Figures~\ref{fig-A-tgl}, \ref{fig-D-tgl}, and \ref{fig-S-tgl}.

930: We use an empty circle to  indicate a right handed $2\pi$-twist on the framing 

931: of a strand as in Figure~\ref{fig-A-tgl}, and a full circle for a left handed one as in 

932: Figure~\ref{fig-Rigg-tgl}. Note, that the extra 1-framed circle in

933: Figure~\ref{fig-S-tgl} does not change the 3-cobordism in $\Cob^{\bullet}$  but 

934: shifts its 2-framing in $\Cob^{2fr, \bullet}$  by one. 

935: 

936: \begin{figure}[ht]

937: \begin{center}

938: \leavevmode

939: \epsfysize=2.2cm

940: \epsfbox{A-tgl.eps}

941: \end{center}

942: \caption{Tangle for $A_j$}\lbl{fig-A-tgl}

943: \end{figure}

944: 

945: \begin{figure}[ht]

946: \begin{center}

947: \leavevmode

948: \epsfysize=2.2cm

949: \epsfbox{D-tgl.eps}

950: \end{center}

951: \caption{Tangle for $D_j$}\lbl{fig-D-tgl}

952: \end{figure}

953: 

954: 

955: \begin{figure}[ht]

956: \begin{center}

957: \leavevmode

958: \epsfysize=2.2cm

959: \epsfbox{S-tgl.eps}

960: \end{center}

961: \caption{Tangle for $S_j$}\lbl{fig-S-tgl}

962: \end{figure}

963: $\Gamma_{g,1}^{2fr}$ can then be thought of as the sub-group of tangles generated

964: by these diagrams, modulo isotopies, 2-handle slides, the $\sigma$-move and the

965: Hopf link move, see \cite{Ker99}.  

966: \medskip

967: 

968: For later purposes we give the explicit action of these generators on

969: $H_1(\Sigma_g,\Z)=H_1(\Sigma_{g,1},\Z)$ in the sense of (\ref{eq-TGSseq}). 

970: Suppose $p,\,f\,\subset\Sigma_{g,1}$ 

971: are two transverse, oriented curves. We denote by $P$ the Dehn twist along $p$,

972: by $[P]\in \Sp(2g,\Z)$ its action on homology, 

973: and by $[p]$ and $[f]$ the respective homology classes. We have

974: \begin{equation}\lbl{eq-DehnHom}

975: [P].[f]\quad=\quad[f]\,+\,([p]\cdot[f])[p]\;.

976: \end{equation}

977: Here $([p]\cdot[f])\in\Z$ is the algebraic intersection number of $p$ with $f$,

978: counting $+1$ for  a crossing if the tangent vectors of $p,f$ form an oriented

979: basis and $-1$ if the basis has opposite orientation. 

980: 

981: A basis for $H_1(\Sigma_g)$ is given by 

982: $\{[a_1],\ldots, [a_g],[b_1],\ldots, [b_g]\}$,

983:  and intersection numbers can be read

984: off Figure~\ref{fig-modsurf}. 

985: For example $a_j$ intersects $b_j$ in only one point, where 

986: $[a_j]\cdot[b_j]= +1$ since $b_j$ follows $a_j$ counter clockwise at the 

987: crossing. Hence 

988: \begin{equation}\lbl{eq-A-hom}

989: [A_j].[b_j]=[b_j]+[a_j]\qquad \mbox{and}\quad [A_j].[x]=[x]\quad  

990: \mbox{for all other basis vectors.}

991: \end{equation}

992: Similarly, we have that $[C_j]$ only acts on $[b_j]$ and $[b_{j+1}]$

993: with $[C_j].[b_j]=[b_j]+[c_j]$ and $[C_j].[b_{j+1}]=[b_{j+1}]-[c_j]$. 

994: Substituting $[c_j]=[a_j]-[a_{j+1}]$, and using the definition  of $D_j$ in

995: (\ref{eq-newgen}) and  (\ref{eq-A-hom}) we compute

996: \begin{equation}\lbl{eq-D-hom}

997: [D_j].[b_j]=[b_j]-[a_{j+1}]\quad \mbox{and}\quad [D_j].[b_{j+1}]=[b_{j+1}]-[a_{j}] \;,

998: \end{equation}

999: and, again,  $[D_j].[x]=[x]$ for all other basis vectors $[x]$ of $H_1(\Sigma_1,\Z)$.

1000: Finally, we find $[B_j].[a_j]=[a_j]-[b_j]$ so that

1001: \begin{equation}\lbl{eq-S-hom}

1002: [S_j].[a_j]=-[b_j] \quad \mbox{and}\quad [S_j].[b_j]=[a_{j}] 

1003: \end{equation}

1004: and $[S_j].[x]=[x]$ elsewise. 

1005: 

1006: The above action can be identified with specific generators  of

1007: the Lie algebra ${\mathfrak s}{\mathfrak p}(2g,\R)$ as follows:

1008: $$

1009: [A_j]\;

1010: =\;I_{2g}\;+\;E_{j,-j}\;=\;I_{2g}\;+\;e_{2\epsilon_j}\;=\;\exp(e_{2\epsilon_j})

1011: $$

1012: \begin{equation}\lbl{eq-sproots}

1013: [B_j]\;=\;I_{2g}\;-\;E_{-j,j}\;=\;I_{2g}\;-\;f_{2\epsilon_j}\;=\;\exp(-f_{2\epsilon_j})

1014: \end{equation}

1015: $$

1016: [D_j]\;=\;I_{2g}\;-\;E_{j,-(j+1)}-E_{j+1,-j}\;=

1017: \;I_{2g}\;-\;e_{\epsilon_j+\epsilon_{j+1}}\;=\;\exp(-e_{\epsilon_j+\epsilon_{j+1}})\\

1018: $$

1019: The conventions and notations for the weights $\epsilon_j$ and the 

1020: matrices $E_{i,j}$ are taken from \cite[Chapter 2.3]{GooWal98}. Hence,

1021: the natural representation on ${\rm Sp}(2g,\Z)$ clearly lifts   to the

1022: fundamental representation of ${\rm Sp}(2g,\R)$. 

1023: 

1024: 

1025: Finally, there is an 

1026:  ${\rm Sp}(2g,\Z)$-invariant 2-form, which is unique up to signs

1027: and given in our basis as:

1028: \begin{equation}\lbl{eq-defsymform}

1029: \omega_g\quad:=\quad \sum_{j=1}^g[a_j]\wedge[b_j]\qquad\in\;

1030: \ext 2 H_1(\Sigma_g)\,=\,H^2(J(\Sigma_g))\;.

1031: \end{equation}

1032: It is identical to twice the K\"ahler metric form in

1033: $H^2(J(\Sigma_g))$, see Section~10 and \cite{GriHar78}. 

1034: 

1035: 

1036: 

1037: \head{5. Hennings TQFT's}\lbl{S5}

1038: In \cite{Hen96} Hennings describes a calculus

1039: that allows us to compute an invariant, ${\cal V}_{\cal A}^H(M)$,

1040: for a closed 3-manifold, $M$, starting from a

1041: surgery presentation, $M=S^3_{\cal L}$,  by a framed link, ${\cal L}\subset S^3$, and 

1042:  a quasitriangular Hopf algebra $\cal A$. It is obtained by inserting and moving elements

1043: of $\cal A$ along the strands of a projection of $\cal L$ and evaluating them against

1044: integrals. This procedure was  refined by Kauffman and 

1045: Radford \cite{KauRad95} permitting

1046:  unoriented links and simplifying the evaluation and proofs substantially. 

1047: ${\cal V}^H_{\cal A}$ turns out to be a special case of the invariant given by  

1048: Lyubashenko \cite{Lub95}, which is constructed from general abelian categories. 

1049: In \cite[Theorem 14]{Ker96} we generalize the Hennings procedure to 

1050: tangles and cobordisms and thus construct a topological 

1051: quantum field theory ${\cal V}^H_{\cal A}$ for any modular Hopf algebra $\cal A$. 

1052: In turn ${\cal V}^H_{\cal A}$ is derived as a special case of

1053: the general TQFT construction by Lyubashenko and 

1054: the author in \cite{KerLub00}. 

1055: 

1056: The TQFT in  \cite{Ker96} was formulated as

1057: a contravariant functor, ${\cal V}_{\cal A}^{*}:\,Cob_3^{\bullet}\to Vect(\kk)$,

1058: where ${\cal V}_{\cal A}^{*}(\Sigma_{g,1})={\cal A}^{\otimes g}$. In this section we

1059: will  give  the rules for construction for the covariant version,  defined by 

1060: ${\cal V}_{\cal A}(M)=(f^{\otimes g})^{-1}({\cal V}^*_{\cal A}(M))^*f^{\otimes g}\,$,  

1061: where  $f:{\cal A}\to{\cal A}^*:\, x\mapsto \mu(S(x)\_\_)$.  We generalize 

1062: \cite{Ker96} further by allowing Hopf algebras, $\cal A$, that are not modular,

1063: at the expense of reducing the vector space by a canonical projection. 

1064: 

1065: 

1066: Let $M$ be a 2-framed cobordism between two model surfaces,

1067: $\Sigma_{g_1}$ and $\Sigma_{g_2}$.

1068: As in \cite{Ker99} we associate to the homeomorphism class of $M$  an equivalence 

1069: class of framed tangle diagrams. The projection of a  representative tangle,  $T_M$, in

1070: $\R\times [0,1]$  has $2g_1$  endpoints $\,1^-<1^+<2^-<\ldots<g_1^-<g_1^+\,$

1071: in the  top line $\R\times 1$ and $2g_2$ endpoints  $\,1^-<1^+<2^-<\ldots<g_2^-<g_2^+\,$

1072: in the bottom line  $\R\times 0$.  Besides closed components ($\cong S^1$) the tangle can

1073: have components with boundary ($\cong[0,1]$). An interval component, $J$, of the tangle 

1074: can either run between points $j^-$ and $j^+$ at the top line or between 

1075:  $j^-$ and $j^+$ at the bottom line.

1076:  As a forth possibility we admit pairs of components, $I$ and $J$, of

1077: which each starts at the top line and ends at the bottom line and cobords a pair

1078: $\{j^-,j^+\}$ to a pair $\{k^-,k^+\}$. The equivalences of tangles are generated 

1079: by isotopies, 2-handle slides (second Kirby move) over closed components,  the 

1080: addition and removal of an isolated Hopf link, in which one component has 0-framing, and

1081: additional boundary moves, called $\sigma$- and $\tau$-Moves, see  \cite{Ker99}. 

1082: For later purposes we also depict here the  $\sigma$-Move: 

1083: 

1084: \beq\label{eq-boundmove}

1085: \epsfbox{boundmove.eps}

1086: \eeq

1087: 

1088: 

1089: The next ingredient is a unimodular, ribbon Hopf algebra, ${\cal A}$,  

1090: in the sense of \cite{Rad94}, over a

1091: perfect  field $\kk$ with $char(\kk)=0$. 

1092: In particular,  ${\cal A}$ is a {\em quasitriangular} Hopf algebra

1093: as introduced by Drinfel'd \cite{Dri90}. This means there exists   an element 

1094: ${\cal R}=\sum_je_j\otimes f_j\,\in{\cal A}^{\otimes 2}$, called the {\em R-matrix},

1095: which fulfills several natural conditions. As in \cite{Dri90} we define the element

1096: $u=\sum_jS(f_j)e_j$, which implements the square of the antipode $S$ by $S^2(x)=uxu^{-1}$.

1097: A {\em ribbon} Hopf algebra is now a quasitriangular Hopf algebra with a group like

1098: element, $G$, such that $G$ also implements $S^2$ and  $G^2=uS(u)^{-1}$. 

1099: From this we define the ribbon element $v:=u^{-1}G$, which is central in

1100: ${\cal A}$. Furthermore, it satisfies the equation 

1101: \begin{equation}\lbl{eq-M-matrix}

1102: \,{\cal M}\;\;=\;\;{\cal R}^{\dagger}{\cal R}\;\;=\;\;\Delta(v^{-1})v\otimes v\;\;,

1103: \end{equation}

1104: where $(a\otimes b)^{\dagger}=b\otimes a$ is the transposition of tensor factors.

1105: 

1106:  Now, any finite dimensional Hopf algebra contains a {\em right integral},

1107: which is an element $\mu \in {\cal A}^*$ characterized by the equation:

1108: \beq\lbl{eq-int-def}

1109: ( \mu\otimes id_{\cal A})(\Delta(x))\;=\;1\cdot\mu(x)

1110: \eeq

1111: Its existence and uniqueness (up to scalar multiplication) has been proven

1112: in \cite{LarSwe69}. The adjective  ``unimodular''

1113: implies that 

1114: \beq\lbl{eq-muSinvar}

1115: \mu(xy)\;\;=\;\;\mu(S^2(y)x)\qquad\qquad\mbox{and}\qquad\qquad\mu(S(x))\;\;=\;\;\mu(G^2x)\;,  

1116: \eeq

1117: see \cite{Rad94}. For the remainder of this article  we  will also

1118:  assume  the following normalizations:

1119: \beq\lbl{eq-balnorm}

1120: \mu\otimes\mu({\cal M})\;\;=\;\;1\qquad\mbox{and}\qquad\;\;\mu(v)\mu(v^{-1})=1

1121: \eeq

1122:  

1123:  

1124: The next step in the Hennings procedure is to replace the tangle projection $T_M$ 

1125: with distinguished over and under crossings by a formal linear combination of copies of 

1126: the  projection $T_M$ in which we do not distinguish between over and under crossings 

1127: but decorate 

1128: segments of the resulting planar curve with elements of $\,{\cal A}\,$. Specifically, 

1129: we replace an over crossing by an indefinite crossing and insert at the two incoming

1130: pieces the elements occurring in the $R$-matrix, and similarly for an under crossing,

1131:   as indicated in the following diagrams.  

1132: 

1133: 

1134: \beq\lbl{fig-Dec-cross}

1135: \setlength{\unitlength}{0.25mm}

1136: \begin{picture}(470,50)

1137: \put(20,5){\line(1,1){40}}

1138: \put(20,45){\line(1,-1){17}}

1139: \put(60,5){\line(-1,1){17}}

1140: 

1141: \put(85,30){}

1142: \put(80,25){\vector(1,0){26}}

1143: 

1144: \put(125,20){$\displaystyle \sum_j$}

1145: 

1146: \put(165,50){\line(1,-1){50}}

1147: \put(165,0){\line(1,1){50}}

1148: \put(180,44){$\scriptscriptstyle e_j$}

1149: \put(175,40){\circle*{10}}

1150: \put(205,40){\circle*{10}}

1151: \put(214,38){$\scriptscriptstyle f_j$}

1152: 

1153: 

1154: %\put(275,50){\line(1,-1){50}}

1155: %\put(275,0){\line(1,1){50}}

1156: %\put(285,40){\circle*{10}}

1157: %\put(260,44){$\scriptscriptstyle  e_j$}

1158: %\put(315,40){\circle*{10}}

1159: %\put(294,38){$\scriptscriptstyle f_j$}

1160: 

1161: \put(280,5){\line(1,1){17}}

1162: \put(320,45){\line(-1,-1){17}}

1163: \put(320,5){\line(-1,1){40}}

1164: 

1165: 

1166: %\put(308,30){$\scriptscriptstyle {\cal  D}ec_{\cal A}$}

1167: \put(333,25){\vector(1,0){27}}

1168: 

1169: \put(375,20){$\displaystyle \sum_j$}

1170: 

1171: 

1172: \put(425,50){\line(1,-1){50}}

1173: \put(425,0){\line(1,1){50}}

1174: \put(435,10){\circle*{10}}

1175: \put(405,15){$\scriptscriptstyle S(e_j)$}

1176: \put(465,10){\circle*{10}}

1177: \put(482,13){$\scriptscriptstyle f_j$}

1178: \end{picture}

1179: \eeq

1180: 

1181: 

1182: The elements on the segments of 

1183: the planar diagram can then be moved along the connected components according to

1184: the following rules. 

1185: 

1186: \beq\lbl{eq-elem-mv}

1187: \setlength{\unitlength}{0.25mm}

1188: \begin{picture}(440,85)(0,-15)

1189: \put(20,5){\line(0,1){60}}

1190: \put(20,20){\circle*{10}}

1191: \put(27,21){$x$}

1192: \put(20,50){\circle*{10}}

1193: \put(27,51){$y$}

1194: 

1195: \put(37,35){$=$}

1196: 

1197: \put(60,5){\line(0,1){60}}

1198: 

1199: \put(60,35){\circle*{10}}

1200: \put(67,36){$xy$}

1201: 

1202: 

1203: \put(135,35){\oval(40,40)[b]}

1204: \put(115,35){\line(0,1){20}}

1205: \put(155,35){\line(0,1){20}}

1206: \put(155,42){\circle*{10}}

1207: \put(162,43){$S(x)$}

1208: 

1209: \put(190,33){$=$}

1210: 

1211: \put(235,35){\oval(40,40)[b]}

1212: \put(215,35){\line(0,1){20}}

1213: \put(255,35){\line(0,1){20}}

1214: \put(215,42){\circle*{10}}

1215: \put(222,43){$x$}

1216: 

1217: 

1218: \put(300,55){\line(1,-1){40}}

1219: \put(300,15){\line(1,1){40}}

1220: \put(330,25){\circle*{10}}

1221: \put(340,25){$x$}

1222: 

1223: \put(358,35){$=$}

1224: 

1225: \put(380,55){\line(1,-1){40}}

1226: \put(380,15){\line(1,1){40}}

1227: \put(390,45){\circle*{10}}

1228: \put(397,48){$x$}

1229: 

1230: \end{picture}

1231: \eeq

1232: 

1233: 

1234: Finally, every diagram can be untangled using the local moves given below,

1235: and the usual planar third Reidemeister move. In particular, undoing a closed

1236: curve in the diagram  

1237: yields an extra overall  factor $G^d$, where $G$ is the  group like element defined

1238: above and $d$ the Whitney number of the curve. 

1239: 

1240: \beq\lbl{fig-st-isom}

1241: \setlength{\unitlength}{0.25mm}

1242: \begin{picture}(400,100)(0,-15)

1243: \put(30,30){\oval(40,40)[b]}

1244: \put(10,30){\line(2,5){6}}

1245: \put(50,30){\line(-2,5){6}}

1246: \put(16,45){\line(1,1){34}}

1247: \put(44,45){\line(-1,1){34}}

1248: 

1249: \put(72,43){$=$}

1250: 

1251: \put(120,30){\oval(40,40)[b]}

1252: \put(100,30){\line(0,1){46}}

1253: \put(140,30){\line(0,1){46}}

1254: \put(140,55){\circle*{10}}

1255: \put(147,56){$G$}

1256: 

1257: %\put(60,-15){\em Rm1}

1258: 

1259: \put(220,40){\line(0,1){20}}

1260: \put(250,40){\line(0,1){20}}

1261: \put(220,60){\line(2,5){6}}

1262: \put(250,60){\line(-2,5){6}}

1263: \put(220,40){\line(2,-5){6}}

1264: \put(250,40){\line(-2,-5){6}}

1265: \put(250,99){\line(-1,-1){24}}

1266: \put(220,99){\line(1,-1){24}}

1267: \put(250,1){\line(-1,1){24}}

1268: \put(220,1){\line(1,1){24}}

1269: 

1270: \put(270,50){$=$}

1271: 

1272: \put(290,4){\line(0,1){92}}

1273: \put(320,4){\line(0,1){92}}

1274: 

1275: %\put(263,-15){\em Rm2}

1276: 

1277: \end{picture} 

1278: \eeq 

1279: 

1280: The assignments that result from this for the left and right ribbon $2\pi$-twists are 

1281: summarized in Figure~\ref{fig-Rigg-tgl}. Note, that in the assignment on the 

1282: right hand side the full circle on the left side stands for a left handed twist

1283: for the framing, while the fat dot on the right hand side indicates a decoration of the strand

1284: by the element $v^{-1}$. 

1285: 

1286: 

1287: \begin{figure}[ht]

1288: \begin{center}

1289: \leavevmode

1290: \epsfxsize=13cm

1291: \epsfbox{Rigg-tgl.eps}

1292: \end{center}

1293: \caption{Twist Assignments}\lbl{fig-Rigg-tgl}

1294: \end{figure}

1295: 

1296: 

1297: 

1298: 

1299: 

1300: 

1301:  

1302: It is clear that after application of these types of manipulations to any decorated diagram 

1303: we eventually obtain a set of disjoint, planar curves  which can be one of four types.

1304: For each of these types we describe next the evaluation rule that leads to the definition

1305: of a linear map ${\cal V}^{\#}(T_M)$:

1306: 

1307: Components of the first type are closed circles decorated with one element $a_i\in{\cal A}\,$ on the

1308: right side. To this we associate the number $\mu(a_i)\in\kk\,$. 

1309: 

1310: Next,  we may have an arc at the bottom line of the diagram connecting points

1311: $p_k'$ and $q_k'$ with one decoration 

1312: $\,b_k\in{\cal A}\,$ at the left strand. To this to we associate 

1313: the vector $\,b_k\in{\cal A}^{(k)}\,$ in the $k$-th copy of 

1314: the tensor product $\,{\cal A}^{\otimes g_2}\,$.

1315: 

1316: Thirdly, for an arc at the top line between points $p_j$ and $q_j$ with decoration

1317: $c_j\in{\cal A}\,$ on the right we assign the linear form $l_{c_j}:{\cal A}^{(j)}\to\kk\,$

1318: given by $l_{c_j}(x)=\mu(S(x)c_j)$

1319:  on the $j$-th copy of 

1320: the tensor product $\,{\cal A}^{\otimes g_1}\,$. 

1321: 

1322: Finally, we may have pairs of straight strands that connect a pair $\,\{p_j,q_j\}\,$ to the pair

1323: $\,\{p_k',q_k'\}\,$,  carrying decorations, $a$ and $b$. In case the strands are parallel,

1324: that is, one connects $p_j$ to $p'_k$ and the other $q_j$ to $q'_k$, we assign a linear

1325: map $T_{a,b}\,:{\cal A}^{(j)}\to {\cal A}^{(k)}\,$ between the $j$-th copy of 

1326: $\,{\cal A}^{\otimes g_1}\,$ to the $k$-th copy of $\,{\cal A}^{\otimes g_2}\,$,

1327: by $T_{a,b}(x)=axS(b)\,$. 

1328: 

1329: If the connecting strands cross over we apply in addition the endomorphism $K(x)=G^{-1}S(x)$ on the

1330: $k$-th copy ${\cal A}^{(k)}$ for a crossing right at the bottom line. It is quite useful to summarize 

1331: these rules also pictorially as follows: 

1332: \beq\lbl{pic-Rclos}

1333: \setlength{\unitlength}{0.25mm}

1334: \begin{picture}(250,70)(0,-5)

1335: \put(35,35){\circle{50}}

1336: \put(60,35){\circle*{8}}

1337: \put(65,37){$a_i$}

1338: 

1339: \put(130,37){\vector(1,0){20}}

1340: 

1341: \put(200,35)

1342: {$\;\mu(a_i)\;$}

1343: \end{picture}

1344: \eeq

1345: \beq\lbl{pic-Rbot}

1346: \setlength{\unitlength}{0.25mm}

1347: \begin{picture}(400,80)(0,-15)

1348: \put(5,10){\rule{15mm}{1mm}}

1349: \put(12,-10){$p_k'$}

1350: \put(52,-10){$q_k'$}

1351: \put(35,35){\oval(40,40)[t]}

1352: \put(15,35){\line(0,-1){25}}

1353: \put(55,35){\line(0,-1){25}}

1354: \put(15,24){\circle*{10}}

1355: \put(21,25){$b_k$}

1356: 

1357: \put(130,30){\vector(1,0){20}}

1358: 

1359: \put(200,25)

1360: {$b\,:\kk\longrightarrow {\cal A}^{(k)} \,\,\,:\;\,\,\;1\,\mapsto\,\,b_k\,$}

1361: \end{picture}

1362: \eeq

1363: \beq\lbl{pic-Rtop}

1364: \setlength{\unitlength}{0.25mm}

1365: \begin{picture}(500,93)(0,-5)

1366: \put(5,60){\rule{15mm}{1mm}}

1367: \put(12,80){$p_j$}

1368: \put(52,80){$q_j$}

1369: \put(35,35){\oval(40,40)[b]}

1370: \put(15,35){\line(0,1){25}}

1371: \put(55,35){\line(0,1){25}}

1372: \put(55,42){\circle*{10}}

1373: \put(62,43){$c_j$}

1374: 

1375: \put(130,53){\vector(1,0){20}}

1376: 

1377: \put(200,40)

1378: {$l_{c_j}\,:{\cal A}^{(j)}\longrightarrow \kk\,\,\,:\;\,\,\;x\,\mapsto\,\,\mu(S(x)c_j)\,$}

1379: \end{picture}

1380: \eeq

1381: \beq\lbl{pic-Rthru}

1382: \setlength{\unitlength}{0.25mm}

1383: \begin{picture}(500,112)(0,-15)

1384: \put(20,5){\line(0,1){60}}

1385: \put(20,35){\circle*{10}}

1386: \put(27,36){$a$}

1387: 

1388: \put(12,80){$p_j$}

1389: \put(52,80){$q_j$}

1390: 

1391: \put(12,-10){$p_k'$}

1392: \put(52,-10){$q_k'$}

1393: \put(60,5){\line(0,1){60}}

1394: 

1395: \put(60,35){\circle*{10}}

1396: \put(67,36){$b$}

1397: 

1398: \put(130,33){\vector(1,0){20}}

1399: 

1400: 

1401: \put(200,27){$T_{a,b}\,:{\cal A}^{(j)}\longrightarrow {\cal A}^{(k)}\,\,\,:\;\,\,\;x\,\mapsto\,\,

1402: a x S(b)\,$}

1403: \end{picture}

1404: \eeq

1405: 

1406: \setlength{\unitlength}{0.25mm}

1407: \begin{picture}(400,80)(0,-15)

1408: \put(5,10){\rule{15mm}{1mm}}

1409: \put(12,-10){$p_k'$}

1410: \put(52,-10){$q_k'$}

1411: \put(15,48){\line(1,-1){37}}

1412: \put(55,48){\line(-1,-1){37}}

1413: 

1414: 

1415: \put(130,30){\vector(1,0){20}}

1416: 

1417: \put(200,25){$K\,:{\cal A}^{(k)}\longrightarrow {\cal A}^{(k)}

1418: \,\,\,:\;\,\,\;x\,\mapsto\,\, G^{-1}S(x)\,$}

1419: 

1420: \end{picture}

1421: 

1422: 

1423: 

1424: 

1425: From these rules for evaluating diagrams 

1426:  we obtain a linear map $\,{\cal A}^{\otimes g_1}\to{\cal A}^{\otimes g_2}\,$

1427: for any decorated planar tangle. For a given tangle $T_M$ we denote by $\,{\cal V}^{\#}(T_M)\,$ the 

1428: sum of all of these maps associated to the sum of decorated diagrams for $T_M$. 

1429: Thus,  if we consider, for simplicity, a tangle $T_M$ without

1430: components of the fourth type, and denote by $a^{\nu}_i\,$,

1431: $b^{\nu}_j\,$ and $c^{\nu}_k\,$ the respective elements of the $\nu$-th summand of 

1432: the same untangled curve of $T_M$, this linear map can be expressed as 

1433: $$

1434: {\cal V}^{\#}(T_M)\;\;:=\;\;\sum_{\nu\, }\mu(a_1^{\nu})\ldots \mu(a_N^{\nu})\, 

1435: b_1^{\nu}\otimes\ldots\otimes b_{g_2}^{\nu}\, l_{a_1^{\nu}}\otimes\ldots\otimes 

1436: l_{a_{g_1}^{\nu}}\;\;. 

1437: $$

1438: For tangles with strand pairs that connect top and bottom pairs we insert the operators

1439: $T_{a,b}\,$ in the respective positions.

1440: \medskip

1441: 

1442: 

1443: \begin{lemma}\lbl{lm-Vprop}

1444: The linear maps ${\cal V}^{\#}(T_M)$ are well defined, (covariantly)

1445: functorial under the composition of tangles, and they commute with   

1446:  the adjoint action of ${\cal A}$ on ${\cal A}^{\otimes g}$. They are also invariant under isotopies

1447: and the following moves: 

1448: \begin{enumerate}

1449: \item 2-handle slides of any type of strand over a closed component of  $\,T_M\,$

1450: \item Adding/removing an isolated Hopf link for which one component has 0-framing 

1451: and the other framing 0 or 1. 

1452: \end{enumerate}

1453: \end{lemma}

1454: 

1455: {\em Proof:} The fact that the construction procedure for a given diagram is unambiguous

1456: is almost straight forward, except that one has to pay attention to the positioning of the 

1457: resulting elements. Details for closed links can be found in \cite{Kau94}. 

1458:  Functoriality is easily checked from the rules of construction. The fact that the maps

1459: are ${\cal A}$-equivariant follows from the fact that it is a special

1460: case of the categorical construction in \cite{KerLub00}, and the fact that 

1461: $f:{\cal A}\to{\cal A}^*$ intertwines the adjoint with the  coadjoint action. 

1462:  Invariance under isotopies follows, as in \cite{Hen96} or \cite{KauRad95}, from the

1463: properties of the $R$-matrix of a quasitriangular Hopf algebra. In the same articles

1464: the 2-handle slide is directly related to the defining equation (\ref{eq-int-def})

1465: of the right integral, see also \cite{Lub95} for the categorical version of the 

1466: argument. Invariance under the Hopf link moves is a direct consequence of the normalizations

1467: in (\ref{eq-balnorm}), since they imply that  the Hennings invariants on the Hopf links are all one. 

1468: \ep

1469: \medskip

1470: 

1471: In order to describe the reduction procedure that allows us to define a TQFT also

1472: for non-modular Hopf algebras 

1473: we introduce the operators associated to the diagrams in Figure~\ref{fig-Smat},

1474: the left being isotopic to the one in Figure~\ref{fig-S-tgl}.

1475: \begin{figure}[ht]

1476: \begin{center}

1477: \leavevmode

1478: \epsfysize=2.2cm

1479: \epsfbox{Smat.eps}

1480: \caption{$S^{\pm}$-Transformations}\lbl{fig-Smat}

1481: \end{center} 

1482: \end{figure}

1483: The double crossing is replaced by the elements $m^{+}_j, n^{+}_j$ from 

1484: ${\cal M}\,=\,\sum_jm_j^+\otimes n_j^+\,$, as defined in (\ref{eq-M-matrix}). 

1485: The transformation $S^+:{\cal A}\to{\cal A}$ is readily worked out to be

1486: \begin{equation}\lbl{eq-S-matrix}

1487: S^+(x)\;=\;\sum_j\mu(S(x) m_j^+) n_j^+\qquad.

1488: \end{equation}

1489: The formula for $S^-$ follows analogously, substituting ${\cal M}$ for

1490: ${\cal M}^{-1}=\sum_jm_j^-\otimes n_j^-$. We consider next the result $\Pi$ 

1491: of stacking the two tangles in Figure~\ref{fig-Smat} on top of each other:

1492: \begin{lemma}\lbl{lm-Pi}

1493: Let  $\Pi:=S^+\circ S^-=S^-\circ S^+$, and denote

1494: $\Pi^{(j)}=1\otimes\ldots1\otimes \Pi\otimes 1\ldots\otimes 1$, with $\Pi$

1495: occurring in the $j$-th tensor position.  

1496: \begin{enumerate}

1497: \item $\Pi$ is an idempotent that commutes with the adjoint action of $\cal A$.

1498: \item ${\cal V}^{\#}(T_M)\Pi^{(j)}={\cal V}^{\#}(T_M)$ if the $j$-th top index pair in 

1499: $T_M$ is attached to a top ribbon in $T_M$. (Analogously for bottom ribbons).

1500: \item $\Pi^{(k)}{\cal V}^{\#}(T_M)={\cal V}^{\#}(T_M)\Pi^{(j)}$ if $T_M$ has 

1501: a through pair

1502: connecting the $j$-th top pair to the $k$-th bottom pair. 

1503: \end{enumerate}

1504: \end{lemma}

1505: 

1506: {\em Proof:} For {\em 1.} note that the picture for $\Pi$ consists of two arcs

1507: that are connected by a circle. Stacking $\Pi$ on top of itself we obtain the picture 

1508: for $\Pi^2$ by functoriality in 

1509: Lemma~\ref{lm-Vprop}. The resulting tangle is the chain of circles $C_j$ and 

1510: arcs $A_{t/b}$ depicted on the left of Figure~\ref{fig-pi-idem}. By  

1511: {\em 1.} of Lemma~\ref{lm-Vprop} we may use 2-handle slides to manipulate

1512: this picture. We first slide $C_1$ over $C_3$, and then $A_b$ over $C_2$.

1513: The result is the tangle for $\Pi$ and a separate Hopf link. The value of

1514: the latter, however, is 1 by (\ref{eq-balnorm}). Hence, $\Pi^2=\Pi$. 

1515: \begin{figure}[ht]

1516: \begin{center}

1517: \leavevmode

1518: \epsfysize=2.2cm

1519: \epsfbox{pi-idem.eps}

1520: \caption{$\Pi$ is idempotent}\lbl{fig-pi-idem}

1521: \end{center} 

1522: \end{figure}

1523: 

1524: Equivariance with respect to the action of ${\cal A}$ is immediate from 

1525: Lemma~\ref{lm-Vprop}. 

1526: 

1527: For {\em 2.} we repeat an argument from \cite{KerLub00}. Suppose $\tau$ is a top

1528: component and $\eta$ any band connecting two intervals $I_i$ in $\tau$ in an

1529: orientation preserving way. To this we associated the surgered diagram in which

1530: the component $\tau$ is replaced by the union $\tau_{\eta}$ of three components.

1531: They are obtained by cutting away the intervals $I_i$ from $\tau$ and inserting

1532: the other two edges of $\eta$ at the endpoints  $\partial I_i$ as indicated in

1533: Figure~\ref{fig-toppath}. Furthermore,  we insert a 0-framed annulus $A$ around $\eta$. 

1534: \begin{figure}[ht]

1535: \begin{center}

1536: \leavevmode

1537: \epsfysize=2.2cm

1538: \epsfbox{toppath.eps}

1539: \caption{$\eta$-Surgery}\lbl{fig-toppath}

1540: \end{center} 

1541: \end{figure}

1542: Sliding any other component over $A$ at an arbitrary point along $\eta$

1543: has the effect of just moving it through $\eta$ at this point. Moreover, we

1544: can slide a $\pm 1$-framed annulus $K$ over $A$ so that it surround the two 

1545: parallel strands in $\tau_{\eta}\,$, and then slide the two strands over $K$. The 

1546: effect is the same as putting a $2\pi$-twist into $\eta$. These two operation

1547: allow us to move any band $\eta$ to any other band $\eta'$ such that $\tau_{\eta}$

1548: and $\tau_{\eta'}$ are related by a sequence of two handle slides. 

1549: 

1550: Now,  adding the picture of $\Pi$ to the top-component $\tau$ of a  tangle $T_M$ 

1551: is the same as surgering $\tau$ along a straight band parallel and close to

1552: the interval between the attaching points of $\tau$ at the top line. We replace

1553: this $\eta$ by a small planar arc at $\tau$ separate from the rest of the 

1554: tangle. Surgery along this corresponds to linking  a Hopf link to $\tau$, 

1555: as $C_2\cup C_3$ is linked to $A_b$ in the middle of Figure~\ref{fig-pi-idem}, 

1556: and consequently  can be removed by the same argument. 

1557: The proofs for the formulas for bottom and through strands are entirely analogous. 

1558: \ep 

1559: 

1560: Set $\Pi^{\#}=\Pi^{\otimes g}$, when acting on ${\cal A}^{\otimes g}$. It follows

1561: now easily from Lemma~\ref{lm-Pi} that 

1562: ${\cal V}^{\#}(T_M)\Pi^{\#}=\Pi^{\#}{\cal V}^{\#}(T_M)$ for all $T_M$. Thus 

1563: each ${\cal V}^{\#}(T_M)$ maps the image of $\Pi^{\#}$ to itself so that we can 

1564: define the restriction 

1565: \begin{equation}\lbl{eq-redtqft}

1566: {\cal V}(T_M):={\cal V}^{\#}(T_M)\Bigl |_{im(\Pi^{\#})}:\;\;

1567: {\cal V}_{\cal A}(\Sigma_{g_1,1})\;\longrightarrow\;{\cal V}_{\cal A}(\Sigma_{g_2,1})\;\;,

1568: \end{equation}

1569: where the vector spaces are given as 

1570: \beq\lbl{eq-A0def}

1571: {\cal V}_{\cal A}(\Sigma_{g,1})=\Pi^{\#}({\cal V}^{\#}(\Sigma_{g}))={\cal A}_0^{\otimes g}

1572: \qquad\mbox{with} \qquad {\cal A}_0=\Pi({\cal A})\;\;\;.

1573: \eeq

1574: \begin{theorem}\lbl{thm-henn}

1575: The assignment $\cal V$ as given in (\ref{eq-redtqft}) yields a 

1576: well defined, 2-framed, relative, ${\cal A}-equivariant$ 

1577: topological quantum field theory

1578: $$

1579: {\cal V}_{\cal A}\;:\quad \Cob^{2fr,\bullet}\;\longrightarrow\;{\cal A}{\rm -mod}_{\kk}\quad\subset

1580: \quad Vect(\kk)\quad. 

1581: $$

1582: Using the invariance functor ${\rm Inv}={\rm Hom}(1,\_):{\cal A}{\rm -mod}\to

1583: Vect(\kk)$ we  obtain an ordinary  2-framed TQFT for closed surfaces as

1584: $$

1585: {\cal V}_{\cal A}^0:={\rm Inv}\circ{\cal V}_{\cal A}\;:

1586: \quad \Cob^{2fr}\;\longrightarrow\; 

1587: \quad Vect(\kk)\quad. 

1588: $$

1589: \end{theorem}

1590: 

1591: {\em Proof:}

1592: We recall from \cite[Proposition 12]{Ker99} that two presentations, $T_M$ and $T'_M$,

1593: of a framed, relative cobordism $M\in \Cob^{2fr,\bullet}$ are related by the 

1594: moves described in Lemma~\ref{lm-Vprop} and the so called $\sigma$-moves, which

1595: consist of adding the picture of $\Pi$ to a pair of points at the top or bottom

1596: line of the diagram. From ${\cal V}(T_M)\Pi^{(j)}={\cal V}^{\#}(T_M)\Pi^{\#}\Pi^{(j)}

1597: ={\cal V}^{\#}(T_M)\Pi^{\#}$ we see that ${\cal V}(T_M)$ is invariant under this

1598: move. Hence, ${\cal V}(T_M)$ only depends on the cobordism represented by $T_M$

1599: and we can write ${\cal V}_{\cal A}(M):={\cal V}(T_M)$.  

1600: 

1601: Due to the equivariance of  $\Pi$ also ${\cal A}_0$ from (\ref{eq-A0def}) is invariant 

1602: under the adjoint action of ${\cal A}$, and the restricted maps commute with

1603: the action of ${\cal A}$ as well. Functoriality of ${\cal V}$ follows from

1604: functoriality of ${\cal V}^{\#}$ and the fact that $\Pi^{\#}$ commutes with 

1605: ${\cal V}^{\#}$.

1606: 

1607: Since each ${\cal V}(M)$ commutes with the action of ${\cal A}$ they also

1608: map the ${\cal A}$-invariant subspaces 

1609: ${\cal V}^0(\Sigma_{g}):={\rm Inv}({\cal V}(\Sigma_{g,1}))$ to themselves. 

1610: This implements the additional $\tau$-move \cite{Ker99} needed to represent

1611: cobordisms between closed surfaces. 

1612: \ep

1613: 

1614: 

1615: \head{6. The Algebra $\cal N$}\lbl{S6}

1616: The Hopf algebra $\cal N$ we will define in this section is  the same 

1617: as the algebra  $A_2$  described by Radford in Example~1 of Section~4.1 

1618: in \cite{Rad76}.  The quasitriangular structure that we endow  $\cal N$ 

1619: with is essentially distilled from the one of $U_{-1}({\mathfrak s}{\mathfrak l}_2)$.

1620: 

1621: Let $\E\cong\R^2$ be the Euclidean plane, and consider 

1622: the 8-dimensional algebra 

1623: \beq\lbl{eq-Adef}

1624: {\cal N}\;:=\;\Z/2\ltimes \ext * \E\;.

1625: \eeq

1626: The generator of $\Z/2$ is denoted by $K$, with $K^2=1$, and we write $x^K=KxK$

1627: for any $x\in{\cal N}$. We thus have relations $w'w=-ww'$ and $w^K:=KwK=-w$ 

1628: for all $w,w'\in\E$.

1629: 

1630: \blm\lbl{lm-Ahopf}

1631: $\cal N$ is a Hopf algebra with coproducts 

1632: \beq\lbl{eq-coprod}

1633: \Delta(K)=K\otimes K\quad\mbox{ and}\qquad \Delta(w)\,=\,w\otimes 1\,+\,K\otimes w\;\;

1634: \forall w\in\E\;.

1635: \eeq

1636: \elm 

1637: {\em Proof:} The fact that $\Delta:{\cal N}\to{\cal N}^{\otimes 2}$ is a 

1638: coassociative homomorphism

1639: is readily verified. The antipode is given by 

1640: \beq\lbl{eq-antipode}

1641: S(K)\,=\,K\qquad\mbox{and}\qquad S(w)=-Kw,\;\;\forall w\in\E\;. 

1642: \eeq

1643: \ep

1644: 

1645: We note the following formulas for the adjoint action and antipode: 

1646: \beq\lbl{eq-somefor}

1647: ad(w)(x)=wx-x^Kw\qquad, \qquad S^2(x)=x^K\qquad\forall x\in{\cal N}, w\in\E

1648: \eeq

1649:  

1650: Let us pick a non-zero element $\rho\in\ext 2\E\subset{\cal N}$, and for this define

1651: a form $\mu_0\in{\cal N}^*$ as follows:

1652: \beq\lbl{eq-Ainteg}

1653: \mu_0(\rho)\;=\;1\;\,,\qquad\quad \mu_0(K\rho)=0\;,\qquad\mbox{and}

1654: \eeq

1655: $$

1656: \mu_0(K^{\delta}x)=0\;,\qquad\forall x\in\ext j\E\;\;,\;\;\mbox{whenever}

1657: \;\;j,\delta\in\{0,1\}\;\;.

1658: $$

1659: \blm\lbl{lm-Ainteg}\ \ \ 

1660: $\mu_0$ is a right (and left) integral on ${\cal N}$. Moreover,  

1661: \beq\lbl{eq-intIN}

1662: \lambda_0:=  (1+K)\rho\qquad\mbox{with}\quad\mu_0(\lambda_0)=1

1663: \eeq 

1664: is a two sided integral in ${\cal N}$.

1665: \elm

1666:  

1667: {\em Proof:} Straightforward verification of (\ref{eq-int-def}). The defining

1668: equation for a two sided integral in ${\cal N}$ is 

1669: $x\lambda_0=\lambda_0 x=\epsilon(x)\lambda_0$, which is also readily found.

1670: \ep

1671: 

1672: Next, we fix a basis $\{\th,\tb\}$ for $\E$.

1673: We define an $R$-matrix, ${\cal R}\in{\cal N}\otimes{\cal N}$, by the formula 

1674: \beq\lbl{eq-ARmat}

1675: {\cal R}\quad:=\quad \Bigl(1\otimes 1\,+\,\th\otimes K\tb\Bigr)\cdot {\cal Z}\;, 

1676: \qquad\mbox{where}\quad {\cal Z}:=\frac 12 \sum_{i,j=0}^1(-1)^{ij}K^i\otimes K^j

1677: %\quad =\quad \cdot {\cal Z}\Bigl(1\otimes 1\,+\,\th K\otimes \tb\Bigr)

1678: \eeq

1679:  

1680: \blm\lbl{lm-Rmat}

1681: The element ${\cal R}$ makes ${\cal N}$ into a quasitriangular Hopf algebra.

1682: 

1683: Moreover, ${\cal N}$ is a ribbon Hopf algebra with unique balancing element $G=K$. 

1684: 

1685: \elm

1686: {\em Proof:} Quasitriangularity follows from a  

1687: straightforward verification of the axioms in \cite{Dri90}.

1688: We compute the special element  $u^{-1}=\sum_jf_jS^2(e_j)=K(1+\tb\th)$ for which

1689: $uS(u)^{-1}=uu^{-1}=1$ so that $G=K$ is a valid and unique choice. The ribbon element

1690: is then given by 

1691: \beq\lbl{lm-Aribb}

1692: v\;:=\;1+\rho\qquad\mbox{with}\qquad \rho:=\tb\th

1693: \eeq

1694: \ep

1695: 

1696: % For the computations involved in the proof and later purposes we note the following 

1697: % formulas for ${\cal Z}$:

1698: % \beq\lbl{eq-AZrel}

1699: % \begin{array}{lll}

1700: % a)\;{\cal Z}^2=1\;\;\; & c)\;{\cal Z}(1\otimes w)=(K\otimes w){\cal Z} \;\;\;& e)\;

1701: % 1\otimes \Delta({\cal Z})={\cal Z}_{13}{\cal Z}_{12}\\

1702: % b)\;{\cal Z}^{\dagger}={\cal Z} & d)\;{\cal Z}(w\otimes 1)=(w\otimes K){\cal Z} & f)\;

1703: % \Delta\otimes 1({\cal Z})={\cal Z}_{13}{\cal Z}_{23}\\

1704: % \end{array}

1705: % \eeq

1706: % Here $w\in\E$ and the convention for ${\cal Z}_{ij}\in{\cal N}^{\otimes 3}$ is as in  \cite% {Dri90}.

1707: 

1708: 

1709: For the monodromy matrix, as defined in (\ref{eq-M-matrix}),  we obtain:

1710: \beq\lbl{eq-Amono}

1711: {\cal M} \;= \; 1\,+\,K\tb\otimes \th \,+\,\th K\otimes \tb-

1712: \rho\otimes\rho\;\;.

1713: \eeq

1714: Setting $T=K\tb\otimes \th \,+\,\th K\otimes \tb$ we compute $T^2=-2\rho\otimes\rho\,$ and 

1715: $T^3=0\,$ 

1716: so that ${\cal M}=\exp(T)$. Hence we can also compute $p$-th powers of the monodromy matrix:

1717: \beq\lbl{eq-qAmono}

1718: {\cal M}^p\;=\;\exp(pT)\;=\;1\,+\,pT\,+\,\frac {p^2}2 T^2\;.

1719: \eeq

1720: 

1721: With $\mu_0$ as defined in (\ref{eq-Ainteg}), and  for $\rho$ as in (\ref{lm-Aribb})

1722: we find $\mu_0\otimes\mu_0({\cal M})=

1723: \mu_0(v)\mu_0(v^{-1})=-1$. Hence, in order to fulfill (\ref{eq-balnorm}) we need to use the 

1724: renormalized integrals

1725: \beq\lbl{eq-renormint}

1726: \mu\;=\;i\mu_0\;\;, \qquad\;\;\lambda=\frac 1 i\lambda_0\;\;,

1727: \qquad\;\; \mbox{with}\;\;\;i=\sqrt{-1}\;\;.

1728: \eeq

1729: For these choices we compute the $S^{\pm}$-transformations assigned to (\ref{eq-S-matrix}) as follows:

1730: \beq\lbl{eq-ASmat}

1731: \begin{array}{lll}

1732: \frac 1 iS^{\pm}(w)=\mp w\;\;\;\forall w\in\E &\qquad \frac 1 i S^{\pm}(\rho)=1\\

1733: \frac 1 iS^{\pm}(Kx)=\,0\;\;\;\forall x\in\ext * \E &\qquad \frac 1 i S^{\pm}(1)=-\rho\;.\\

1734: \end{array}

1735: \eeq

1736: This implies that the projector $\Pi$ from Lemma~\ref{lm-Pi} has   

1737: kernel $ker(\Pi)=\{Kw\, :\, w\in\ext *\E\}$ and image

1738: \beq\lbl{eq-APi}

1739: {\cal N}_0\;\;=\;\;im(\Pi)\;\;\;=\;\;\ext * \E\;.

1740: \eeq

1741: From (\ref{eq-somefor}) we see that ${\cal N}_0$ acts trivially

1742: on itself  so that the action of ${\cal N}$ factors through the obvious 

1743: $\Z/2\Z={\cal N}/{\cal N}_0$-action.   

1744: 

1745: 

1746: Finally, we note that $\SL(2,\R)$ acts on  $\E$ and, hence, also on $\cal N$,

1747: assuming $K$ is $\SL(2,\R)$-invariant.  

1748: \begin{lemma}\lbl{lm-ASLinv}  

1749: $\SL(2,\R)$ acts on $\cal N$ 

1750:  by Hopf algebra automorphisms. 

1751: 

1752: The ribbon element $v$, the monodromy $\cal M$, and the two integrals

1753:  are invariant under this action. 

1754: \end{lemma}

1755: 

1756: {\em Proof:} The fact that  $\SL(2,\R)$ yields algebra automorphsims is 

1757: obvious by construction. Linearity of coproduct and antipode in $w$ 

1758: in (\ref{eq-coprod}) and (\ref{eq-antipode}) imply that this is, in fact, 

1759: a Hopf algebra homomorphism. $v$ and $\lambda$ are invariant

1760: since $\SL(2,\R)$ acts trivially on $\E\wedge\E\,$. Invariance of $\cal M$

1761: follows then from (\ref{eq-M-matrix}). 

1762: \ep

1763: 

1764: Note,  that $\cal R$ itself is {\em not} $\SL(2,\R)$-invariant. 

1765: 

1766:  

1767: \bigskip

1768: 

1769: \head{7. The Hennings TQFT for $\cal N$}\lbl{S7}

1770: 

1771: 

1772: From (\ref{eq-APi}) and (\ref{eq-redtqft}) we see that the vector spaces of the 

1773: Hennings TQFT for the algebra from  (\ref{eq-Adef}) are given as

1774: \beq\lbl{eq-AVvect}

1775: {\cal V}_{\cal N}(\Sigma_g)\;\;:=\;\;\Bigl(\ext *\E\Bigr)^{\otimes g}

1776: \qquad\mbox{with}\qquad {\rm dim}({\cal V}_{\cal N}(\Sigma_g))=4^g\;.

1777: \eeq 

1778: We now compute the action of the mapping class group generators from the 

1779: tangles in Figures~\ref{fig-A-tgl}, \ref{fig-D-tgl}, and \ref{fig-S-tgl}. 

1780: From the extended Hennings rules it is clear that

1781:  the pictures for both $A_j$ and $S_j$ result in actions only on the  $j$-th

1782: factor in the tensor product in (\ref{eq-AVvect}). For $A_j$ we use the presentation

1783: from Figure~\ref{fig-A-tgl} and the rules from Figure~\ref{fig-Rigg-tgl} and

1784: (\ref{pic-Rthru}) to obtain the linear map $\Aa(x):=x\cdot v$.

1785: 

1786: The extra 1-framed circle in Figure~\ref{fig-S-tgl} results in an extra factor

1787: $\mu(v)=i$, since an empty circle corresponds to an insertion of $v$. The action on 

1788: the $j$-th factor is thus given by an application of $\Ss:=iS^{+}\bigl |_{{\cal N}_0}$

1789: so that

1790: \beq\lbl{eq-SSform}

1791: \Ss(\rho)=-1\;,\qquad\quad \Ss(1)=\rho\;,\qquad{\rm and}\qquad \Ss(w)=w\;,\quad\forall w\in\E\;\;.

1792: \eeq 

1793: 

1794: Similarly,  $D_j$ acts only on the $j$-th and the $(j+1)$-st

1795: factors of ${\cal N}_0^{\otimes g}$. From  (\ref{pic-Rthru}) and the formula 

1796: for ${\cal M}^{-1}$ in (\ref{eq-Amono}) we compute for the action on these two factors

1797: \beq\lbl{eq-D-2act}

1798: \D\;:\;{\cal N}_0^{\otimes 2}\,\to\,{\cal N}_0^{\otimes 2}, \qquad

1799: x\otimes y \;\mapsto\; x\otimes y +  x\th\otimes \tb y -

1800: x\tb\otimes \th y -  x\rho \otimes \rho y\;. 

1801: \eeq

1802: The generators of the mapping class group $\Gamma_g$ are thus represented as follows:

1803: \beq\lbl{eq-AVgen}

1804: \begin{array}{ll}

1805: {\cal V}_{\cal N}({\bf I}_{A_j})\;\;=\;\;I^{\otimes j-1}\otimes \Aa\otimes I^{\otimes g-j}\;,

1806:  &\qquad \quad {\cal V}_{\cal N}({\bf I}_{S_j})\;\;=\;\;I^{\otimes j-1}\otimes \Ss\otimes I^{\otimes g-j}\\

1807: &\\

1808: \qquad\qquad\mbox{and}&

1809: {\cal V}_{\cal N}({\bf I}_{D_j})\;\;=\;\;I^{\otimes j-1}\otimes \D\otimes I^{\otimes g-j-1}\;.\\

1810: \end{array}

1811: \eeq

1812: Let us also compute the linear  maps associated to the cobordisms ${\bf H}_g^{\pm}$

1813: from (\ref{eq-hdlcob}). Their tangle presentations follow from \cite{Ker99} and 

1814: have the forms given in Figure~\ref{fig-tglhdl}. 

1815: 

1816: \begin{figure}[ht]

1817: \begin{center}

1818: \leavevmode

1819: \epsfysize=3.2cm

1820: \epsfbox{HG-tgls.eps}

1821: \caption{Tangles for Handle additions}\lbl{fig-tglhdl}

1822: \end{center} 

1823: \end{figure}

1824: 

1825: We included $\pm 1$-framed circles to adjust the 2-framings of ${\bf H}_g^{\pm}$. 

1826: A 0-framed circle around a

1827: strand has the effect of inserting $\lambda= S^+(1)=\frac 1 i\rho$. In this

1828: normalization we find with $\rho=i\Pi\lambda$ and (\ref{pic-Rbot}) that 

1829: \beq\lbl{eq-Ahdl+map}

1830: {\cal V}_{\cal N}({\bf H}_g^{+})\;:\;\;\alpha\;\mapsto\;\;\alpha\otimes \rho

1831: \qquad \quad \forall \alpha \in{\cal N}_0^{\otimes g}\;.

1832: \eeq

1833: Similarly, we obtain from (\ref{pic-Rtop}) that 

1834: \beq\lbl{eq-Ahdl-map}

1835: {\cal V}_{\cal N}({\bf H}_g^{-})\;:\;\;\alpha\otimes x\;\mapsto\;\;\mu_0(x)\alpha

1836: \qquad \quad \forall \alpha \in{\cal N}_0^{\otimes g}, \; x\in {\cal N}_0\;,

1837: \eeq

1838: where $\mu_0$ is as in (\ref{eq-Ainteg}). We note the following:

1839: 

1840: \begin{lemma}\lbl{lm-ASL2inv}

1841: The generators in (\ref{eq-AVgen}), (\ref{eq-Ahdl+map}), and (\ref{eq-Ahdl-map})

1842: intertwine the $\SL(2,\R)$-action on ${\cal N}_0^{\otimes g}$.

1843: \end{lemma}

1844: 

1845: {\em Proof:} The fact that $\Aa$ and $\D$ commute with the $\SL(2,\R)$-action follows from

1846: invariance of $v$ and $\cal M$. From (\ref{eq-ASmat}) we see that $\Ss$ is scalar on the 

1847: non-invariant part, and thus commutes as well. Finally, $\rho$ and $\mu_0$ are clearly invariant.

1848: 

1849: \ep

1850: 

1851: For $g\geq 0$ set $\chi_g:=S_g\circ\ldots\circ S_1$, $h^+_g:= {\bf H}_{g-1}^+\circ \ldots 

1852: \circ {\bf H}_{0}^+$, and $h^-_g:={\bf H}_{0}^-\circ \ldots 

1853: \circ {\bf H}_{g-1}^-$. We define a standard closure of a 2-framed 3-cobordism as the closed

1854: 3-manifold 

1855: \beq\lbl{eq-closM}

1856: \lz M\rz\;\;:=\;\;h^-_{g_2}\circ\chi_{g_2}\circ M\circ \chi^{-1}_{g_1}\circ h^+_{g_1}\;\cup\,D^3\;.

1857: \eeq

1858: If $M$ is represented by a tangle $T$ we obtain, similarly, a link $\lz T\rz$. We introduce 

1859: the following function from the class of 2-framed cobordisms into $\Z/2$:

1860: \beq\lbl{eq-number}

1861: \varphi(M)\quad:=\quad \beta_1(\lz M\rz)\;+\;{\rm sign}(\lz T\rz)\;\;{\rm mod}\; 2\;,

1862: \eeq

1863: where $\beta_j$ denotes the $j$-th Betti number. 

1864: We further denote by $\Cob^{22fr,*}\subset \Cob^{2fr,*}$ the subset of  all cobordisms $M$

1865: with $\varphi(M)=0$, which we will call {\em evenly} 2-framed. 

1866: 

1867: \begin{lemma}\lbl{lm-mod2prop}

1868: \begin{enumerate}

1869: \item $\varphi(M)\;=\;|\lz T\rz|\;{\rm mod}\, 2$, where $|\lz T\rz|\,:=$ $\#$ components of $\lz T\rz$.

1870: \item  $\varphi(M)\;=\;\#$ components of $ T$ not connected to the bottom line. 

1871: \item ${\cal V}_{\cal N}(M)$ is real  if $\varphi(M)=0$ and imaginary for $\varphi(M)=1$.  

1872: \item $\Cob^{22fr,*}$ is a subcategory.

1873: \end{enumerate}

1874: \end{lemma}

1875: 

1876: 

1877: {\em Proof:} Let $W$ be the 4-manifold given by adding 2-handles to $D^4$ along 

1878: $\lz T\rz\subset S^3$ so that $\lz M\rz=\partial W$, and let $L_T$ be the linking matrix of 

1879: $\lz T\rz$. We have $\beta_2(W)= |\lz T\rz |=d_++d_-+d_0$, where $d_+$, $d_-$, and $d_0$ are

1880: the number of eigenvalues of $L_T$ that are $>0$, $<0$, and $=0$ respectively. 

1881: From the exact sequence 

1882: $0\to H_2(\lz M\rz )\to H_2(W) \stackrel {L_T}{\longrightarrow} H^2(W)\to H_1(\lz M\rz)\to 0$

1883: we find that $\beta_1(\lz M\rz)=d_0$, which implies {\em 1.} using ${\rm sign}(W)=d_+-d_-$.

1884: {\em 2.} follows immediately from the respective tangle compositions. 

1885: 

1886: The possible components

1887: not connected to the bottom line are strands connecting point pairs at  the top line

1888: or closed components. From the rules (\ref{pic-Rclos}) through (\ref{pic-Rthru}) we see 

1889: that these are just the types of components that involve an evaluation against $\mu=i\mu_0$.

1890: All other parts of the Hennings procedure involve only real maps. Finally, {\em 4.} follows

1891: from counting tangle components under composition. 

1892: \ep

1893: 

1894: \begin{propos}\lbl{pp-ATQFT}

1895: The Hennings procedure yields a relative, 2-framed, $\SL(2,\R)$-equivariant, half-projective TQFT

1896: $$

1897: {\cal V}_{\cal N}\;:\;\;\Cob^{2fr,\bullet}\;\longrightarrow\;\SL(2,\R)-mod_{\Cc}\;,

1898: $$

1899: which is $\Z/4$-projective on $\Cob^{\bullet}$. We have a restriction

1900: $$

1901: {\cal V}_{\cal N}^{(2)}\;:\;\;\Cob^{22fr,\bullet}\;\longrightarrow\;\SL(2,\R)-mod_{\R}\;,

1902: $$

1903: which is $\Z/2$-projective on $\Cob^{\bullet}$.

1904: \end{propos}

1905: 

1906: 

1907: {\em Proof:} From Lemma~\ref{lm-ASL2inv} we know that the generators of $\Gamma_g$ are

1908: represented $\SL(2,\R)$-equivariantly, hence also $\Gamma_g$ itself. The decomposition in 

1909: (\ref{eq-Heegaard}) and equivariace of the maps in (\ref{eq-Ahdl+map}) and  

1910: (\ref{eq-Ahdl-map}) implies the same for  general cobordisms. That this TQFT is

1911: half-projective follows from the fact that $\cal N$ is non-semisimple, or, equivalently,

1912: that ${\cal V}_{\cal N}(S^1\times S^2)=\mu(1)=\varepsilon(\lambda)=0$, see \cite{Ker98}.

1913: The projective phase of the TQFT is determined by the value $\mu(v)=i$ on the 1-framed

1914: circle.

1915: 

1916: Lemma~\ref{lm-mod2prop}, {\em 3.} implies that ${\cal V}_{\cal N}^{(2)}$ maps into

1917: the {\em real}  $\SL(2,\R)$-equivariant maps and modules. This reduces the 

1918: ambiguity of multiplication with $i$ to a  sign ambiguity. 

1919: \ep 

1920: 

1921: \medskip

1922: 

1923:  An important point of view in the TQFT constructions in \cite{KerLub00} 

1924: is the existence of a categorical Hopf algebra, which can be understood as

1925: the TQFT image of a topological Hopf algebra given as an object in $\Cob^{\bullet}$. 

1926: 

1927: To be more precise, in 

1928: \cite{Yet97} and \cite{Ker96} $\Cob^{\bullet}$ is described as a braided tensor

1929: category, and it is found 

1930: that the object $\Sigma_{1,1}\in \Cob^{\bullet}$ is naturally identified as a

1931: braided Hopf algebra in this category in the sense of \cite{Maj93} and \cite{LubMaj94}. 

1932: Particularly, $\Sigma_{2,1}$  is identified with   $\Sigma_{1,1}\otimes\Sigma_{1,1}$

1933: since the tensor product on $\Cob^{\bullet}$  is defined by sewing two 

1934: surfaces together along a pair of pants. The 

1935: multiplication and comultiplication are thus given by 

1936: elementary cobordisms ${\bf M}:\Sigma_{2,1}\to\Sigma_{1,1}$ and 

1937: ${\bf \Delta} :\Sigma_{1,1}\to \Sigma_{2,1}$. Their tangle diagrams are worked out

1938: explicitly in \cite{BKLT00}, and depicted in Figure~\ref{fig-MC-tgl}

1939: with minor modifications in the conventions:

1940: 

1941: \begin{figure}[ht]

1942: \begin{center}

1943: \leavevmode

1944: \epsfysize=2.2cm

1945: \epsfbox{MC-tgl.eps}

1946: \caption{Tangles for Mulitplications}\lbl{fig-MC-tgl}

1947: \end{center} 

1948: \end{figure}

1949: 

1950: Here ${\bf c}:\Sigma_{2,1}\to\Sigma_{2,1}$ is the braid isomorphism. 

1951: The braided antipode is given by the tangle 

1952: $\Gamma=(S^+)^2$, with $S^+$ as in Figure~\ref{fig-Smat}. 

1953: 

1954: \begin{lemma}

1955: \lbl{lm-multHeeg}

1956: The cobordisms {\bf M} and ${\bf \Delta}$ have the following 

1957: Heegaard decompositions.

1958: $$

1959: {\bf M}\;=\;{\bf H}_2^-\circ {\bf I}_{D_1 \circ S_2} \qquad\mbox{and}

1960: \qquad {\bf \Delta}\;=\;

1961:  {\bf I}_{S_1\circ  D_1^{-1} \circ S_1^{-1}\circ S_2^{-1}}

1962: \circ{\bf H}_2^+

1963: $$

1964: \end{lemma}

1965: 

1966: {\em Proof:} Verification by composition of the associated tangles.\ep

1967: 

1968: The explicit formulae for the linear maps associated to the generators

1969: of the mapping class group and the handle attachments in Section~7 allow us now to 

1970: compute the braided Hopf algebra structure induced on 

1971: ${\cal N}_0={\cal V}_{\cal N}(\Sigma_{1,1})$. We write 

1972: $M_0:={\cal V}_{\cal N}({\bf M})$, $\Delta_0:={\cal V}_{\cal N}({\bf \Delta})$, 

1973: $S_0:={\cal V}_{\cal N}(S_1^2)$, and $c_0:={\cal V}_{\cal N}({\bf c})$ for

1974: the braided multiplication, comultiplication, antipode and braid isomorphism

1975: respectively. 

1976: 

1977: \begin{lemma}\lbl{lm-super-Hopf}

1978: The induced braided Hopf algebra structure on ${\cal N}_0$ is the canonical $\Z/2$-graded 

1979: Hopf algebra with:

1980: $$ 

1981: M_0(x\otimes y)= xy\qquad\quad c_0(x\otimes y)=(-1)^{d(x)d(y)}y\otimes x\qquad\quad

1982: \forall x,y\in {\cal N}_0

1983: $$

1984: $$

1985: \mbox{and} \qquad \Delta_0(w)=w\otimes 1 + 1\otimes w \qquad\quad 

1986: \Gamma_0(w)=-w \qquad\forall w\in\E\quad.

1987: $$

1988: In particular, ${\cal N}_0$ is commutative and cocommutative in the graded and braided sense,

1989: ${\cal N}_0\cong{\cal N}_0^*$ is self dual, $\SL(2,\R)$ still acts by Hopf automorphisms 

1990: on ${\cal N}_0$, and $S_0$ is an  involutory homomorphism on ${\cal N}_0$.

1991: \end{lemma}

1992:  

1993: {\em Proof:} For {\bf M} and ${\bf \Delta }$ insert the morphism associated to 

1994: the generators in Lemma~\ref{lm-multHeeg}. The braid isomorphism is given via the Hennings

1995: rules by acting with the operator $ad\otimes ad({\cal R})$ on 

1996: ${\cal N}_0^{\otimes 2}$ and then permuting the factors. It is easy to see that

1997: $ad\otimes ad({\cal Z})$ acts on $x\otimes y$ by multiplying $(-1)^{d(x)d(y)}$,

1998: where $d(x)$ is the $\Z/2$-degeree of $x$ in ${\cal N}_0$. Moreover, we we know 

1999: that the adjoint action of ${\cal N}_0$ on itself is trivial so that

2000: the term $\th\otimes K\tb$ in the second factor of ${\cal R}$ in (\ref{eq-ARmat})

2001: does not contribute.

2002: \ep

2003: 

2004: 

2005: \smallskip

2006: 

2007: 

2008: \medskip

2009: 

2010: 

2011: \head{8. Skein theory for ${\cal V}_{\cal N}$}\lbl{S8}

2012: 

2013: The skein theory of the Hennings calculus over ${\cal N}$ is mostly a consequence

2014:  on the form 

2015: $v=1+\rho$ of the ribbon element 

2016: as in (\ref{lm-Aribb}). In the Hennings procedure we substitute a strand 

2017: with decoration $\frac 1 i\rho$ by a dotted strand (with possibly more decorations)

2018: as shown on the left of Figure~\ref{fig-rhostrand}. Observe from (\ref{eq-Amono}) that 

2019: $$

2020: {\cal M}^{\pm 1}(1\otimes \rho)=(1\otimes \rho)\qquad\mbox{and}

2021: \qquad {\cal M}^{\pm 1}(\rho\otimes 1)=(\rho\otimes 1).

2022: $$

2023: This means that for a dotted strand we do not have to distinguish between over

2024: and undercrossing with other strands as indicated on the right of Figure~\ref{fig-rhostrand}.

2025: As a result such a strand can be 

2026: disentangled from the rest of the diagram.  

2027: 

2028: \begin{figure}[ht]

2029: \begin{center}

2030: \leavevmode

2031: \epsfbox{rhostrand.eps}

2032: \end{center}

2033: \caption{Transparent $\rho$-decorated strand}\lbl{fig-rhostrand}

2034: \end{figure}

2035: 

2036: The next additional ingredient in the calculus are symbols for 1-handles. They are used

2037: in the bridged link calculus as described in \cite{Ker99} and \cite{KerLub00}. We indicate 

2038: a pair of 1-surgery balls by pairs of coupons. The defining relation is the modification

2039: move depicted on the left of Figure~\ref{fig-modif}. The move indicated on the right 

2040: of Figure~\ref{fig-modif} and its reflections  is a standard consequence of the boundary

2041: move from (\ref{eq-boundmove}). 

2042: 

2043: \begin{figure}[h]

2044: \begin{center}

2045: \leavevmode

2046: \epsfbox{modif.eps}

2047: \end{center}

2048: \caption{Coupons for 1-handles}\lbl{fig-modif}

2049: \end{figure}

2050: 

2051: Since $v^{k}=1+k\rho$ for $k\in\Z$ we find that the framing of any component can be changed

2052: at the expense of introducing dotted lines. This translates to the diagrams in 

2053: Figure~\ref{fig-frame}. 

2054: 

2055: \begin{figure}[h]

2056: \begin{center}

2057: \leavevmode

2058: \epsfbox{frame.eps}

2059: \end{center}

2060: \caption{Framing shift}\lbl{fig-frame}

2061: \end{figure}

2062: 

2063: The skein relation is now obtained by applying  Figure~\ref{fig-frame} to

2064: the Fenn Rourke move as in Figure~\ref{fig-FR}, see also\cite{MatPol94}. 

2065:            

2066: \begin{figure}[h]

2067: \begin{center}

2068: \leavevmode

2069: \epsfysize=2.6cm

2070: \epsfbox{FRmove.eps}

2071: \end{center}

2072: \caption{Fenn Rourke Move}\lbl{fig-FR}

2073: \end{figure}

2074: 

2075: \begin{lemma}\label{lm-skein}  For two strands belonging to two different components of a tangle

2076: diagram we have the relation

2077: \begin{center}

2078: \leavevmode

2079: \epsfbox{skein.eps}

2080: \end{center}

2081: For strands belonging to the same component of the tangle the relation is 

2082: \begin{center}

2083: \leavevmode

2084: \epsfbox{skeinsame.eps}

2085: \end{center}

2086: \end{lemma} 

2087: 

2088: At this point it is convenient to extend the tangle presentations to general diagrams,

2089: dropping the condition that a strand starting at a point $j^-$ has to end at a point

2090: $j^+$ (or the corresponding condition for through strands). From such a general tangle 

2091: diagram we can get to an admissible one by applying boundary moves (\ref{eq-boundmove})

2092: at all intervals $[j^-,j^+]$. (This is in fact the original definition used in 

2093: \cite{Ker99}.) We shall allow the occurrence of coupons but restrict ourselves to the cases

2094: where exactly two strands enter (or exit) a coupon as in Lemma~\ref{lm-skein}. 

2095: 

2096: We also introduce two notions of components:

2097: The first  is that of a {\em diagram component} $\cal X$  of a generalized tangle diagram.

2098: It is given by a concatenation of curve segments, coupons that have two strands going in

2099: on one side, and intervals  $[j^-,j^+]$ connecting a strand  ending in $j^-$ with the one

2100: ending in $j^+$. 

2101: 

2102: The second is a  {\em strand component}, which is also a collection of curves that can be 

2103: joined in two ways.  As before curves that end in two sides of the same interval 

2104: $[j^-,j^+]$ belong to the same strand component, as well as  curves exiting and entering a coupon

2105: pair that would be connected under application of Figure~\ref{fig-modif}. 

2106: 

2107: 

2108: 

2109: We have the following rules for manipulating the coupons:

2110: 

2111: \begin{lemma} \label{lm-1handlemoves}

2112: In the following equivalences the labels $A, B,\ldots$ indicate which coupons form a pair. 

2113: \begin{enumerate}

2114: \item 1-handles can be slid over other 1-handles, through a boundary interval, and hence 

2115: anywhere along a strand component. 

2116: \beq\label{eq-1handlemove}

2117: \epsfbox{1handlemoves.eps}

2118: \eeq

2119: \item If in a diagram the coupons of a pair belong to different diagram components the entire

2120: diagram does not contribute, i.e., is evaluated as zero.  Hence only diagrams contribute in which

2121: the diagram components coincide with strand components. 

2122: \beq\label{eq-pairvanish}

2123: \epsfbox{pairvanish.eps}

2124: \eeq

2125: \item Direct 1-handle cancellation: If coupons with the same label are adjacent on the same 

2126: side of a strand they can be canceled: 

2127: \beq\label{eq-1handlecanc}

2128: \epsfbox{1handlecanc.eps}

2129: \eeq

2130: \item Opposite 1-handle cancellation: If coupons with the same label are adjacent on opposite

2131: sides of a strand the strand is replaced by a dotted strand and the evaluation gains a 

2132: factor of 4. 

2133: \beq\label{eq-1handleoppcanc}

2134: \epsfbox{1handleoppcanc.eps}

2135: \eeq

2136: \item If a generalized tangle diagram contains a coupon configuration as indicated the entire 

2137: diagram is evaluated as zero.

2138: \beq\label{eq-coupontorus}

2139: \epsfbox{coupontorus.eps}

2140: \eeq

2141: \end{enumerate} 

2142: \end{lemma}

2143: 

2144: 

2145: {\em Proof:} The slide of $B$ over the pair $A$ in (\ref{eq-1handlemove}) translates to

2146: a simple isotopy if we apply the move in 

2147: Figure~\ref{fig-modif} to the $A$-pair. Similarly, the slide

2148: through a boundary interval is given by an isotopy conjugated by a $\sigma$-move as in

2149: (\ref{eq-boundmove}). 

2150: 

2151: 

2152: For {\em b)}

2153: let  $\cal X$ be a diagram component that contains coupons $A_1,\ldots, A_n$ whose partner

2154: lie on different diagram components. Performing boundary moves we can make $\cal X$ to be a true inner

2155: component. Furthermore, we can eliminate the other coupons on $\cal X$ that occur in pairs

2156: by undoing the modification from Figure~\ref{fig-modif}. The component $\cal X$ is now a 

2157: closed curve interrupted only by coupons $A_1,\ldots, A_n$. We undo the modification also

2158: for these and the corresponding annuli added in the move 

2159:  bound discs that we denote by $D_1,\ldots, D_n$. Note, that the arcs of

2160: $\cal X$ all end in only one side of a disc $D_j$ since the strands emerging from the other

2161: side belong to a different component. We can thus surger the discs along the arcs, as shown in 

2162: (\ref{eq-pairvanish}), so that we 

2163: obtain a torus $T$ with $n$  holes 

2164: $\partial T=\partial D_1\sqcup\ldots\sqcup\partial D_n$

2165: which misses all other parts of the tangle. After surgery along the annuli the

2166: torus $T$ can be capped off so that we have found a non-separating surface 

2167: inside the represented cobordism. Since we are dealing with a non-semisimple

2168: TQFT this implies that the associated linear map is zero.

2169: 

2170: The direct cancellation in (\ref{eq-1handlecanc}) follows by applying Figure~\ref{fig-modif}.

2171: In the resulting configuration in the middle of (\ref{eq-1handlecanc}) the Hopf link can 

2172: be slid off and removed. 

2173: 

2174: The opposite cancellation in (\ref{eq-1handleoppcanc}) and the remodification from 

2175: Figure~\ref{fig-modif} give the tangle in the middle. Now consider in general a 

2176: straight strand that is entangled with an annulus with $2p$ positive crossings as

2177: in (\ref{eq-annuluswinding}). 

2178: \beq\label{eq-annuluswinding}

2179: \epsfbox{annuluswinding.eps}

2180: \eeq

2181: Using  the formula in (\ref{eq-qAmono}) we find by applying the Hennings procedure and

2182:  evaluating the elements on the annulus against the integral

2183: that the resulting element on the open strand

2184: is 

2185: $$

2186: \mu\otimes id({\cal M}^p)\;\;=\;\;\frac {p^2} i \,\rho

2187: $$

2188: which with Figure~\ref{fig-rhostrand} implies the claim. 

2189: 

2190: Finally, we also 

2191: reexpress the  coupons in 

2192: in (\ref{eq-coupontorus}) by a tangle. As before non-semisimplicity 

2193: of the TQFT implies that a diagram containing such a subdiagram is always zero. For example the

2194: 0-framed annulus clearly bounds a surface disjoint from the rest of the link so that the 

2195: cobordism contains a non separating surface. 

2196: 

2197: \ep

2198: 

2199: 

2200: 

2201: We  now combine the previous two lemmas in the following skein relations without coupons. 

2202: 

2203: \begin{theorem} \label{thm-skein}

2204: For generalized tangle diagrams we have the following skein relations:

2205: 

2206: For crossings of strands of different components: 

2207: 

2208: \beq\label{eq-SKEIN}

2209: \epsfbox{SKEIN.eps}

2210: \eeq

2211: 

2212: For crossing of strands of the same component we need to introduce an orientation on the component. 

2213: 

2214: \beq\label{eq-SKEINsame}

2215: \epsfbox{SKEINsame.eps}

2216: \eeq

2217: 

2218: \end{theorem} 

2219: 

2220:   {\em Proof:}   The proof is given by moving the coupons in the skein relations of 

2221: Lemma~\ref{lm-skein}  through the components using   Lemma~\ref{lm-1handlemoves}. 

2222:   \ep

2223: 

2224: Note, that relation (\ref{eq-SKEINsame}) implies the relation for the Kauffman 

2225: polynomial for $z=\frac 12$. However, the framing relations are quite different. 

2226: 

2227: Let $\widehat {\cal B}_g$ be  the group  of tangles in $2g$ strands generated 

2228: by the braidings $\bf c$ of double strands  and the 

2229: braided antipodes $\Gamma$ as in Figure~\ref{fig-MC-tgl} 

2230: acting in different positions.  It is thus the image of the abelian extension  $B_g\ltimes {\Z/2}^g $

2231: of the braid group. 

2232: 

2233: Moreover, let us introduce a few elementary generalized tangles $M_k: k\to 0$, $\varepsilon:1\to 0$

2234: and $X_n:0\to 2n$ as depicted below. 

2235: 

2236: \beq\label{eq-planelem}

2237: \epsfbox{planelem.eps}

2238: \eeq

2239: 

2240: \begin{theorem} \label{thm-solve}

2241: Every tangle $T:G\to 0$   with $2G$ starting (top)  points and {\em no} endpoints 

2242: can be resolved via the skein relations in Theorem~\ref{thm-skein} into a combination

2243: of tangles of the form 

2244: $$

2245: T\;=\;(M_{k_1}\otimes\ldots M_{k_r}\otimes \epsilon^{\otimes N})\circ B\;,

2246: $$

2247: with $B\in \widehat {\cal B}_G$ and $\sum_{i=1}^rk_i=G-N$. 

2248: \end{theorem}

2249: 

2250: {\em Proof:}    We consider generalized tangles without coupons.  We proceed by induction on 

2251: the number $m$ of connected components of $T$. We only count components that involve solid lines,

2252: those with dotted lines reduce to a collection of $\varepsilon$-diagrams at the intervals belonging to

2253: that component or closed dotted circles that do not contribute.

2254:  Suppose now $T$ has only one component, which we equip with some orientation. 

2255: Applying $\Gamma$'s  to the intervals we can arrange it that the strands enter an interval

2256: $[j^-,j^+]$  at the left point $j^-$ and leave at the right one $j^+$. Furthermore, we can find a 

2257: permutation of intervals so that the strand exiting $j^+$ enters at $(j+1)^-$, except for 

2258: $G^+$, which is connected to $1^-$. Hence, by multiplying an element of $\widehat {\cal B}_G$ 

2259: to $T$ we can assume that the endpoints of the intervals are connected to each other by strands

2260: as they are for $M_G$. 

2261: 

2262: Next we note that the skein relation (\ref{eq-SKEINsame}) from Theorem~\ref{thm-skein} does not

2263: change this connectivity property for the solid lines and any diagram with dotted lines

2264: collapses to $\varepsilon$-diagrams.

2265: 

2266: For diagrams where equally labeled coupons are on the same components

2267: there are three planar moves that allow us to

2268: manipulate the arrangement of coupons. They are the 1-handle slide and the

2269: 1-handle cancellation depicted below, and the boundary flip as in Figure~\ref{fig-modif}. 

2270: In fact it is easy to see that 

2271: we have the skein relation $T=M_G+iw(T)\varepsilon^{\otimes G}$, where $w(T)$ is the generalization 

2272: of the writhe number of the diagram as defined, for example, in  \cite{Lik97}. In case $G=0$ the diagram

2273: $M_0$ is a closed solid circle which therefore makes the entire diagram zero. 

2274: 

2275: Assume now $T$ has $m$ components and the claim is true for all diagrams with $m-1$ components.

2276: Pick one component  $C$ and apply an element of $\widehat {\cal B}_G$ such that the intervals included in 

2277: this component are all to the left of the other intervals. Note,

2278:  that the set of intervals that belongs to

2279: $C$ may also be empty. Next apply the skein relations (\ref{eq-SKEIN})

2280: from Theorem~\ref{thm-skein} to untangle $C$ from the other components. In each step of changing 

2281: crossings of a strand of $C$ with the strand of another component $D$ we can choose the relation 

2282: for  which the tangle that belongs to the first local diagram on the right side of the equation has

2283: one component less since $C$ and $D$ are connected. The other diagrams on the right side also

2284: have one less component since we do not count dotted lines. Hence, by induction, the error of changing 

2285: a crossing between $C$ and another component can be resolved into elementary diagrams as claimed. 

2286: After $C$ is untangled  we have expressed $T$, modulo elementary diagrams,

2287:  in the form $C\otimes T'$ (juxtaposition) where $T'$

2288: has $m-1$ components. Again each factor can be resolved independently by induction, and, hence, 

2289: the whole

2290: diagram since $\otimes$-products of elementary diagrams are again elementary. 

2291: 

2292: \ep 

2293: 

2294: 

2295: Next note that every tangle $R:\,g_1\to g_2$ is in fact of the form 

2296: \beq\label{eq-RT}

2297: R\;=\; (T\otimes id_{g_2})\circ (id_{g_1}\otimes X_{g_2})

2298: \eeq

2299: for some $T:\,g_1+g_2\to 0$.   Thus, in order to evaluate a general tangle diagram it suffices

2300: by Theorem~\ref{thm-solve}  to specify the evaluations of  the  elementary tangles in (\ref{eq-planelem}). To this end we define the tensor

2301: \beq\label{eq-defAtens}

2302: A\;=\;\frac 1 i S\otimes 1\Delta(\rho)\;=\;\frac 1 i \Bigl( \rho\otimes 1\,+\,1\otimes\rho\,-\,\tb\otimes\th\,+\,\th\otimes\tb\Bigr)\;\;\in\,{\cal N}_0^{\otimes 2}\;.

2303: \eeq

2304: 

2305: \begin{cor} \label{cor-eval}

2306: Every diagram can be resolved into a sum of composites of diagrams in 

2307: (\ref{eq-planelem}). The linear maps associated to them are 

2308: \beq\label{eq-X1}

2309: {\cal V}_{\cal N}(X_1)\,:\,\Cc\to {\cal N}_0^{\otimes 2}\,:\;1\,\mapsto

2310: A\,=\,\sum_{\nu}x_{\nu}\otimes y_{\nu}

2311: \eeq

2312: \begin{eqnarray}\label{eq-Xn}

2313: {\cal V}_{\cal N}(X_n)\;&=&\;

2314: (1^{\otimes (n-1)}\otimes {\cal V}_{\cal N}(X_1)\otimes1^{\otimes (n-1)})\circ

2315: {\cal V}_{\cal N}(X_{n-1})\,

2316: \;

2317: :\, \Cc\to {\cal N}_0^{\otimes 2n}\,

2318: \\

2319: &:&\;

2320: 1\,\mapsto

2321: A_{\{n\}}\,

2322: =\,\sum_{\nu_1,\ldots,\nu_n}x_{\nu_1}\otimes x_{\nu_2}\otimes \ldots\otimes x_{\nu_n}\otimes y_{\nu_n}

2323: \otimes \ldots\otimes y_{\nu_2} \otimes y_{\nu_1}\nonumber

2324: \end{eqnarray}

2325: \beq\label{eq-Mn}

2326: {\cal V}_{\cal N}(M_n):\;\; {\cal N}_0^{\otimes n}\to \Cc\;\;:\;\;\;\;a_1\otimes\ldots\otimes a_n\;\mapsto\;\mu(a_1\cdot\ldots\cdot a_n)

2327: \eeq

2328: 

2329: Dotted circles can be removed and diagrams with solid circles do not contribute. 

2330: \end{cor}

2331: 

2332: {\em Proof:} The formulae follow easily from the pictures in Figure~\ref{fig-MC-tgl} to which we 

2333: assigned linear maps in Lemma~\ref{lm-super-Hopf}. Particularly, we find that the upside down

2334: reflection of the multiplication tangle {\bf M} is mapped to the S-conjugate coproduct

2335: \beq\lbl{eq-Scoprod}

2336: \widetilde \Delta= i \Ss^{-1}\otimes\Ss^{-1}\Delta_0\Ss\,:\;{\cal N}_0\otimes{\cal N}_0\to

2337: {\cal N}_0\;.

2338: \eeq

2339: The tangle $X_1$ is obtained by capping this off with an arc at the top, which corresponds to

2340: the insertion of the unit. Hence, $A=\widetilde \Delta(1)$. The diagrams $M_p$ are easily 

2341: identified as composites $M^p=(M\otimes 1^{\otimes(p-1)})\circ M^{p-1}$ capped off with

2342: an arc at the  bottom, which is hence assigned to the $p$-fold multiplication followed

2343: by an evaluation against the integral $\mu\in{\cal N}^*$. 

2344: \ep

2345: \medskip

2346: 

2347: Let us consider a few examples. One useful case is when the braid $B\in\widehat{\cal B}_n$ can 

2348: be chosen trivially. Hence the contribution to the linear map for a tangle $R:g_1\to g_2$ is

2349: given by a union of planar diagrams as depicted in (\ref{eq-plancomp}):

2350: \beq\label{eq-plancomp}

2351: \epsfbox{plancomp.eps}

2352: \eeq

2353: Define the map

2354: \beq\label{eq-defCpq}

2355: C_p^q\,=\,\widetilde\Delta^{q-1}\circ M_0^{p-1}\;:

2356: \quad {\cal N}_0^{\otimes p}\longrightarrow {\cal N}_0^{\otimes q}\;,

2357: \eeq

2358: where the exponents denote the usual multiple products and coproducts. 

2359: The linear map associated to a planar diagram is now the tensor product

2360: of maps associated to the individual components of the diagram. For example, 

2361: if we want to evaluate the linear map on a homogeneous vector $x_1\otimes\ldots\otimes x_{g_1}$   

2362: an the diagram has 

2363: a component with solid lines as in (\ref{eq-plancomp}) containing

2364: top intervals $[i_1^-,i_1^+], \ldots, [i_p^-,i_p^+]$ and bottom intervals

2365: $[j_1^-,j_1^+], \ldots, [j_q^-,j_q^+]$ we compute the vector 

2366: $C_p^q(x_{i_1}\otimes\ldots \otimes x_{i_p})\in {\cal N}_0^{\otimes q}$ and insert the

2367: entries in order into the positions $j_1,\ldots, j_q$ in ${\cal N}_0^{\otimes g_2}$.

2368:  

2369:  

2370: 

2371: With these rules the computation of the maps associated to the generators of the

2372: mapping class group are readily carried out. For

2373:  example we can evaluate the diagram for the $S$-transformation from 

2374: Figure~\ref{fig-S-tgl}. We resolve the right most crossing by taking the 

2375: skein relation in the first row in Proposition~\ref{thm-skein} but with every

2376: diagram rotated  clockwise by $\frac \pi 2$. The result is 

2377: $$

2378: \Ss\;=\;id - \rho\otimes\mu_0 - 1\otimes \epsilon - 1\otimes \mu_0 + \rho\otimes \epsilon

2379: $$

2380: This  yields exactly  the formula from (\ref{eq-SSform}).

2381: 

2382:  

2383: As another example we may consider the $C_1$ waist cycle in $\Sigma_2$. The diagram consists

2384: of four parallel strands with a 1-framed annulus around the second and third. We apply 

2385: Figure~\ref{fig-frame} and then  Figure~\ref{fig-modif} to this annulus. The resulting coupons

2386: can be canceled. We find 

2387: $${\cal V}_{\cal N}({\bf I}_{C_1})=id - iC_1^1\,.

2388: $$

2389: This implies the formula for the $D$-transformation from (\ref{eq-D-2act}). 

2390: \smallskip

2391:  

2392: 

2393: Finally, let us show how to use the skein calculus to find

2394:  the precise formula for the invariant of a 2-framed

2395: closed 3-manifold presented by a link ${\cal L}\subset S^3$. It is 

2396: basically given by the order of the first integral homology. More precisely,

2397: let

2398: \beq\label{eq-defeta}

2399: \eta(M)\;:=\;\left\{\begin{array}{cl}

2400: \Bigl| H_1(M,\Z)\Bigr| & \mbox{for}\; \beta_1(M)=0\\

2401: 0 & \mbox{for}\;\beta_1(M)>0\end{array}\right.

2402: \eeq

2403: 

2404: \begin{lemma}\label{lm-defeta} For a given framed link ${\cal L}\subset S^3$ and 

2405: $\eta$ as in (\ref{eq-defeta}) we have 

2406: $$

2407: {\cal V}_{\cal N}(M_{\cal L})\;=\; i^{|{\cal L}|}det({\cal L}\cdot{\cal L})

2408: =\;\pm i^{|{\cal L}|} \eta(M)

2409: $$

2410: \end{lemma}

2411: 

2412: {\em Proof:}

2413: By 2-handle slides we can move ${\cal L}$ into a link ${\cal L}^{\delta}$ 

2414: so that the linking form ${\cal L}^{\delta}\cdot {\cal L}^{\delta}$ is

2415: diagonal and equivalent to the original one ${\cal L}\cdot{\cal L}$. Suppose 

2416: $f_j$ is the framing number of the $j$-th component ${\cal L}_j^{\delta}$. 

2417: From Figure~\ref{fig-frame} we see that 

2418: $$

2419: {\cal V}_{\cal N}({\cal L}^{\delta})= 

2420: {\cal V}_{\cal N}({\cal L}^{\delta, -f_j}) + if_j

2421: {\cal V}_{\cal N}({\cal L}^{\delta}-{\cal L}_j^{\delta})

2422: $$

2423: Here, ${\cal L}^{\delta, -f_j}$ is the link in which the framing of the $j$-th

2424: component is shifted to zero. As a result the manifold represented by this

2425: link has non-trivial rational homology. Since ${\cal V}_{\cal N}$ is  a 

2426: non-semisimple theory this implies that 

2427: ${\cal V}_{\cal N}({\cal L}^{\delta, -f_j})=0$. Iterating the above identity

2428: we find 

2429: ${\cal V}_{\cal N}({\cal L}^{\delta})=\prod_{j=1}^{|{\cal L}|}(if_j)

2430: {\cal V}_{\cal N}(\emptyset)$. Clearly, $\prod_{j=1}^{|{\cal L}|}(f_j)$

2431: is the determinant of the linking form of ${\cal L}^{\delta}$ and 

2432: hence also the one of ${\cal L}$. 

2433: 

2434: \ep

2435: 

2436: 

2437: \head{9. Equivalence of ${\cal V}_{\cal N}^{(2)}$ and ${\cal V}^{FN}$}\lbl{S9}

2438: 

2439:  In this section we compare the two topological quantum field theories ${\cal V}^{FN}$

2440: described in Section~3 and ${\cal V}_{\cal N}^{(2)}$ constructed in Section~7. We already

2441: found  a number of general properties that are shared by both theories:

2442: 

2443: By Lemma~\ref{lm-FNpart} and Proposition~\ref{pp-ATQFT} both theories are $\Z/2$-projective 

2444: on $\Cob^{\bullet}$ and non-semisimple, fulfilling the property of Lemma~\ref{lm-vanish}. 

2445: The $\Z/2$-projectivity is due to ambiguities of even 2-framings in the case of 

2446: ${\cal V}_{\cal N}^{(2)}$ and ambiguities of orientations in the case of ${\cal V}^{FN}$.

2447: The non-semisimple half-projective property results in the case of ${\cal V}^{FN}$ from

2448: representation varieties that are transversely disjoint, and in the case of 

2449: ${\cal V}_{\cal N}^{(2)}$  from the  nilpotency

2450: of the integral $\lambda\in\cal N$. 

2451: Further common features are the dimensions of vector spaces $(=4^g)$,

2452: actions of $\SL(2,\R)$, see Section~9, and the fact that ${\cal J}_g$ lies in the kernel

2453: of the mapping class group representations. 

2454: 

2455: We construct now an explicit isomorphism between ${\cal V}^{FN}$ and ${\cal V}_{\cal N}^{(2)}$. 

2456: Let ${\cal Q}=\ext *\lz a,b\rz$ be the  exterior algebra over $\R^2$ with basis $a,b\in\R^2$.

2457: We obtain a canonical isomorphism, which is defined on monomial elements as follows: 

2458: \begin{equation}\lbl{eq-defi}

2459: i_*\;:\quad {\cal Q}^{\otimes g}\;\isto\;\ext * H_1(\Sigma_g)\;\;:\qquad 

2460: q_1\otimes \ldots\otimes q_g\mapsto i_1(q_1)\wedge\ldots\wedge i_g(q_g) \;, 

2461: \end{equation}

2462: where $i_j:{\cal Q}\,\isto\,\ext *\lz [a_j], [b_j]\rz$ is the canonical map sending $a$ and $b$ 

2463: to $[a_j]$ and $[b_j]$ respectively. Next, 

2464: we define an isomorphism between ${\cal Q}$ and ${\cal N}_0$, seen as linear spaces, 

2465: by the following assignment of basis vectors: 

2466: %$\,\,\phi_j\,:\;\mbox{$\bigwedge^*$} (\lz \th,\tb\rz)\;

2467: %\longrightarrow\;\mbox{$\bigwedge^*$} (\lz a_j,b_j\rz)\;$

2468: \begin{equation}\lbl{eq-phidef}

2469: \begin{array}{lcc}

2470: &\quad \phi(1)=b \quad

2471: &\quad\phi(\tb\th)=a\quad\\

2472: \phi\;:\quad{\cal N}_0\;\isto\;{\cal Q}\qquad\;\;\mbox{with}\;\;&&\\

2473:  &\qquad \phi(\th)=a\wedge b \quad &\quad \phi(\tb)=1\quad \\

2474: \end{array}\;.

2475: \end{equation}

2476: Note, that this map has odd $\Z/2$-degree and is, in particular,

2477:  not an algebra homomorphism. From (\ref{eq-phidef}) we infer 

2478: directly the following identities:

2479: \beq\lbl{eq-}

2480: \phi(\th x)=-\phi(x)\wedge a \qquad\qquad \phi(x \th )=a\wedge  \phi(x)\\

2481: \eeq

2482: \beq\lbl{eq-AS1phi}

2483: \phi(\Aa x)\;=\;[A_1]\phi(x)\qquad\qquad \phi(\Ss x)\;=\;[S_1]\phi(x)\\

2484: \eeq

2485: Here, $\Aa$ and $\Ss$ are as in (\ref{eq-AVgen}), and $[A_1]$ and $[S_1]$ are the 

2486: maps on $H_1(\Sigma_1)$ as in (\ref{eq-A-hom}) and (\ref{eq-S-hom}). 

2487: 

2488: 

2489: Moreover, let us introduce a sign-operator $(-1)^\Lambda$ on ${\cal Q}^{\otimes g}$ 

2490: defined on monomials by 

2491: \begin{equation}\lbl{eq-Lambda}

2492: (-1)^{\Lambda_g}(q_1\otimes \ldots\otimes q_g)\quad=\quad

2493: (-1)^{\lambda_g(d_1,\ldots,d_g)}q_1\otimes \ldots\otimes q_g\;.

2494: \end{equation}

2495: The function $\lambda_N$ is defined in the $N$-fold  product of 

2496:  $\Z/2$'s as follows:

2497: \begin{equation}\lbl{eq-lambdadef}

2498: \lambda_N:\;(\Z/2)^N\,\to\,\Z/2\qquad\mbox{with}\quad

2499: \lambda_N(d_1,\ldots,d_N)\;=\;\sum_{i<j}d_i(1-d_j)\;,

2500: \end{equation}

2501: where $d_j={\rm deg}(q_j)\,{\rm mod}\, 2\,$. Consider now the following isomorphism

2502: of vector spaces. 

2503: \begin{equation}\lbl{eq-defxi}

2504:  \xi_g:=\;i_*\circ(-1)^{\Lambda_g}\circ\phi^{\otimes g}\quad:

2505: \qquad{\cal N}_0^{\otimes g}\,\;\isto\,\;\ext *H_1

2506: \end{equation}

2507: Given a linear map, $F:{\cal N}^{\otimes g_1}\to {\cal N}^{\otimes g_2}$, we write 

2508: $(F)^{\xi}:=\xi_{g_2}\circ F\circ \xi_{g_1}^{-1}$ for the respective map on homology.

2509: Moreover, we denote by  ${\bf L}^{(k)}_x$ the operator on ${\cal N}^{\otimes g}$  

2510: that multiplies the $k$-th factor in the tensor product by $x$ from the left,

2511: and by ${\bf R}^{(k)}_x$ the respective operator for multiplication from the right. We

2512: compute: 

2513: \beq\lbl{eq-LRdef}

2514: \begin{array}{cl}

2515: \quad &({\bf L}^{(k)}_{\th})^{\xi}(\alpha\wedge u_k\wedge\beta)\;\;=\;\;(-1)^{g-k+s+1}\,\,

2516: \alpha\wedge a_k\wedge u_k\wedge\beta\;,\\

2517: \\

2518: \mbox{and}\qquad &({\bf R}^{(k)}_{\th})^{\xi}(\alpha\wedge u_k\wedge\beta)\;\;=\;\;(-1)^{g-k+s}\,\,

2519: \alpha\wedge u_k\wedge a_k \wedge\beta\;,

2520: \end{array}

2521: \eeq

2522: where $s=\sum_{j=1}^g d_j$ is the total degree of $\alpha\wedge u_k\wedge\beta$, 

2523: $\alpha\in\ext *\lz a_1,\ldots, b_{k-1}\rz$,  and $\beta\in\ext *\lz a_{k+1},\ldots, b_{g}\rz$. 

2524: 

2525: \begin{lemma}\lbl{lm-commut}

2526: For every  standard generator $G\in \{A_j, D_j, S_j\}$, we have 

2527: $$

2528: ({\cal V}_{\cal N}({\bf I}_{G}))^{\xi}\;=\;\ext * [G]\;\;,

2529: $$

2530: where $[G]$ denotes as before the action on homology. 

2531: \end{lemma}

2532: 

2533: {\em Proof:} For the $A_j$ and $S_j$ this follows readily from (\ref{eq-AS1phi}), and 

2534: the fact that $[A_j]$ and $[S_j]$ do not change the  degrees $d_j$ and hence commute with 

2535: $(-1)^{\Lambda_g}$. 

2536: 

2537: The operator in (\ref{eq-D-2act}) decomposes into $\D=\D^0+\D^1$, where 

2538: $\D^0=id - {\bf R}_{\rho}\otimes {\bf L}_{\rho}$ and 

2539: $\D^1={\bf R}_{\th}\otimes {\bf L}_{\tb}-{\bf R}_{\tb}\otimes {\bf L}_{\th}$.

2540: Now $\D^0$ does not change the $\Z/2$-degree of both factors, and $\D^1$ flips the

2541: degree of both factors. One readily verifies that

2542: $$

2543: \lambda_g(\ldots, 1-d_j, 1-d_{j+1},\ldots)-\lambda_g(\ldots, d_j, d_{j+1},\ldots)=d_j + d_{j+1}\qquad

2544: {\rm mod}\,2

2545: $$

2546: $$

2547: \mbox{so that}\qquad

2548: {\cal V}_{\cal N}({\bf I}_{D_j})^{\xi}\;\;=\;\;({\cal V}_{\cal N}^0({\bf I}_{D_j}))^{\zeta}\,+\,

2549: (-1)^{d_j+d_{j+1}}({\cal V}_{\cal N}^1({\bf I}_{D_j}))^{\zeta}\quad

2550: %\mbox{with}\;\;\zeta_g=i_*\circ \phi^{\otimes g}

2551: $$

2552: $$\qquad\quad=\;\;(I^{\otimes j-1}\otimes (\D^0)^{\phi^{\otimes 2}}\otimes I^{\otimes g-j-1})^{i_*}\;+\;(-1)^{d_j+d_{j+1}}(I^{\otimes j-1}\otimes (\D^1)^{\phi^{\otimes 2}}

2553: \otimes I^{\otimes g-j-1})^{i_*}

2554: $$

2555: 

2556: Here, $\zeta_g=i_*\circ \phi^{\otimes g}$ and 

2557: ${\cal V}_{\cal N}^i({\bf I}_{D_j})$ is the operator with $\D^i$ in $j$-th position.

2558: Since $\zeta_g=\zeta^{\otimes g}_1$ the $\zeta$-conjugate maps only act on the generators

2559: $\{a_j, b_j, a_{j+1}, b_{j+1}\}$ the action is the same for all positions $j\,$.

2560: Observe that also $[D_j]$ acts only on the homology generators $\{a_j, b_j, a_{j+1}, b_{j+1}\}$.

2561: It is, therefore, enough to prove the relation for $g=2$ and ${\cal V}_{\cal N}({\bf I}_{D_1})

2562: =\D$. 

2563: 

2564: Now, from (\ref{eq-D-2act}) it is obvious

2565: that ${\cal V}_{\cal N}({\bf I}_{D_j})$ commutes

2566: with ${\bf L}^{(j)}_{\th}$ and ${\bf R}^{(j+1)}_{\th}$. Moreover, it is easy to see that

2567: $\ext *[D_j]$, as given in (\ref{eq-D-hom}), commutes with 

2568: $({\bf L}^{(j)}_{\th})^{\xi}$ and $({\bf R}^{(j+1)}_{\th})^{\xi}$ from (\ref{eq-LRdef}). 

2569: Specifically, we use that $\ext *[D_j]$ does not change the total degree, and acts trivially on

2570: $a_j$ and $a_{j+1}$. It thus suffices to check 

2571: \beq\lbl{eq-12Drel}

2572: \ext 2[D_1]\circ\zeta_2(x_1\otimes x_2)=\zeta_2\circ \D^0(x_1\otimes x_2) +(-1)^{d_1+d_2}

2573: \zeta_2\circ\D^1 (x_1\otimes x_2)

2574: \eeq

2575:  with $d_i={\rm deg}(\phi(x_i))$, and only for $x_i\in\{1,\tb\}$. For example, for $x_1=x_2=1$,

2576: with $d_1+d_2=0$, we find from (\ref{eq-D-2act}) and (\ref{eq-D-hom}) that

2577: $$

2578: \begin{array}{lcl}

2579: \zeta_2 \circ \D(1\otimes 1)&=&\zeta_2(

2580: 1\otimes 1  +  \th\otimes \tb  -

2581: \tb\otimes \th  -  \rho \otimes \rho)\;\\

2582: &=& b_1\wedge b_2 + a_1\wedge b_1 - a_2\wedge b_2 - a_1\wedge a_2\\

2583: &=& (b_1-a_2)\wedge (b_2-a_1)  \;=\;\ext 2[D_1](b_1\wedge b_2)\;=\;\ext 2[D_1](\zeta_2(1\otimes 1))\\

2584: \end{array}

2585: $$

2586: We also compute for the case $x_1=\tb$ and $x_2=1$, with $d_1+d_2=1$:

2587: $$

2588: \begin{array}{lcl}

2589: \zeta_2 \circ (\D^0-\D^1)(\tb\otimes 1)&=&\zeta_2(\tb\otimes 1 -\tb\th\otimes\tb)

2590: \;=\;b_2 - a_1\\

2591: &=& \ext 2[D_1](b_2)\;\;\;=\;\;\;\ext 2[D_1](\zeta_2(\tb\otimes 1))

2592: \end{array}\;.

2593: $$ 

2594: The other two cases follow similarly.

2595: \ep

2596: 

2597: As the $\{A_j, D_j, S_j\}$ generate $\Gamma_g$ we conclude from Lemma~\ref{lm-commut} and

2598: (\ref{eq-FNmapcg})  that $({\cal V}_{\cal N}({\bf I}_{\psi}))^{\xi}

2599: ={\cal V}^{FN}({\bf I}_{\psi})$ for all $\psi\in\Gamma_g\,$.

2600: 

2601: Let us also consider the maps associated by both functors to the

2602: handle additions  ${\bf H}_g^{\pm}$. We note that

2603: $$

2604: \lambda_{g+1}(d_1,\ldots,d_g,1)=\lambda_g(d_1,\ldots,d_g)

2605: $$

2606: so that we find from (\ref{eq-Ahdl+map}), (\ref{eq-FNmcg}) and (\ref{eq-phidef}) that

2607: $({\cal V}_{\cal N}({\bf H}_g^+))^{\xi}={\cal V}^{FN}({\bf H}_g^+)$. Similarly,

2608: (\ref{eq-Ahdl-map}), (\ref{eq-FNgcm}) and (\ref{eq-Ainteg}) imply 

2609: $({\cal V}_{\cal N}({\bf H}_g^-))^{\xi}={\cal V}^{FN}({\bf H}_g^-)$. 

2610: Using the Heegaard 

2611: decomposition (\ref{eq-Heegaard}) we finally infer equivalence:

2612: \begin{propos}\lbl{pp-Vequiv}

2613: The maps $\xi_g$ defined in (\ref{eq-defxi}) give rise to an isomorphism

2614: $$

2615: \xi\;:\; {\cal V}_{\cal N}\;\;\;\;\stackrel{\bullet\,\cong}

2616: {-\!\!\!-\!\!\!-\!\!\!\longrightarrow}\;

2617: \;\;\;{\cal V}^{FN}\;\;

2618: $$

2619: of relative, non-semisimple, $\Z/2$-projective functors from $\Cob^{\bullet}$ to $Vect(\kk)$. 

2620:  \end{propos}

2621: 

2622: 

2623: 

2624: \head{10. Hard-Lefschetz decomposition and Invariants}\lbl{S10}

2625:  

2626: The tangent bundle over the  moduli space $J(\Sigma_g)$ is trivial with fiber

2627: $H^*(\Sigma_g,\R)$ so that its cohomology ring is naturally $\ext *H_1(\Sigma_g,\R)$. 

2628: There is an almost complex structure on $J(\Sigma_g)$ given by a map ${\sf J}$ with

2629:  ${\sf J}^2=-1$ in the cohomology. It is given by ${\sf J}.[a_j]=-[b_j]$ and 

2630: ${\sf J}.[b_j]=[a_j]$.

2631: With the  K\"ahler form

2632: $\omega_g\in H^2(J(\Sigma_g))$ defined in (\ref{eq-defsymform}) 

2633: it is also a K\"ahler manifold. The dual K\"ahler metric provides us with a Hodge star

2634: $\star:\ext jH_1(\Sigma_g)\to\ext {2g-j}H_1(\Sigma_g)$ for a given volume form

2635: $\Omega\in\ext {2g} H_1(\Sigma_g)$ by the equation 

2636: $\alpha\wedge\star\beta=\lz\alpha,\beta\rz\Omega$. 

2637: Specifically, the $2g$ generators  $\{[a_1],\ldots, [b_g]\}$ of $H_1(\Sigma_g)$, 

2638: with volume form $\Omega=[a_1]\wedge\ldots\wedge [b_g]$ the Hodge star is given by  

2639: $\star(a_1^{1-\epsilon_1}\wedge\ldots\wedge b_g^{1-\epsilon_{2g}})

2640: =(-1)^{\lambda_{2g}(\epsilon_1,\ldots,\epsilon_{2g})}a_1^{\epsilon_1}\wedge\ldots\wedge 

2641: b_g^{\epsilon_{2g}}$, where $\lambda_{2g}$ is as in (\ref{eq-lambdadef}). 

2642: 

2643: 

2644: As a K\"ahler manifold $H^*(J(\Sigma_g))$ admits an $\SL(2,\R)$-action,  

2645: see for example \cite{GriHar78},  given for the standard generators  

2646: $E, F, H\,\in\,{\mathfrak s}{\mathfrak l}_2(\R)$ by 

2647: \beq\lbl{eq-defsl2act}

2648: H\alpha:=(j-g)\alpha\quad\; \forall \alpha\in \ext jH_1(\Sigma_g)\;,

2649: \qquad\quad E\alpha:=\alpha\wedge\omega_g\;,\qquad\quad F:=\star\circ E\circ\star^{-1}

2650: \eeq

2651: \begin{lemma}\lbl{lm-FN-equiv}

2652: The functor ${\cal V}^{FN}$ is $\SL(2,\R)$-equivariant with respect to the action

2653: in (\ref{eq-defsl2act}).

2654: \end{lemma}

2655: 

2656: {\em Proof:}  Commutation with $H$ follows from counting degrees. Since $\omega_g$ is invariant

2657: under the $\Sp(2g,\R)$-action, $E$ commutes with the maps in (\ref{eq-FNmapcg}),

2658: and since $\omega_g\wedge [a_{g+1}]=[a_{g+1}]\wedge \omega_{g+1}$ also with the ones 

2659: in  (\ref{eq-FNmcg}) and  (\ref{eq-FNgcm}). Finally, as all maps ${\cal V}^{FN}(M)$

2660: are isometries with respect to $\lz.,.\rz$ they also commute with $F$. \ep

2661: 

2662: In order to finish the proof of Theorem~\ref{thm-main} we still need to show that

2663: the $\xi_g$ are $\SL(2,\R)$-equivariant as well. The fact that $H$ commutes with $\xi_g$ 

2664: is again a matter of counting  degrees. We have $E=\sum (E_1^{(i)})^{i_*}$,

2665:  where $E_1^{(i)}$ acts

2666: on the $i$-th factor of ${\cal Q}^{\otimes g}$ 

2667: by $q\mapsto E_1(q)= q\wedge a\wedge b$. Since $E$ does not change

2668: degrees we find that $E^{\xi}=\sum (E^{(i)})^{\phi^{(i)}}$, where

2669: $(E^{(i)})^{\phi^{(i)}}$ acts on the $i$-th factor by $E_1^{\phi}$.

2670: We find $E_1^{\phi}(\tb)=\th$, and 

2671: $E_1^{\phi}(1)=E_1^{\phi}(\th)=E_1^{\phi}(\tb\th)=0$, which yields precisely

2672:  the desired action of $E$ 

2673: on ${\cal N}_0$. The conjugate action of $\star$ on ${\cal N}_0^g$ is as follows: 

2674: \beq\lbl{eq-xistar}

2675: \star^{\xi}\;:\;\; x_1\otimes\ldots x_g\;\;\mapsto \; (-1)^{\sum_{i<j}d_id_j}(\star x_1)

2676: \otimes\ldots\otimes (\star x_g)\qquad\;\forall x_j\in{\cal N}_0\;, 

2677: \eeq	

2678: where $\star \th=\tb$, $\star\tb=\th$, $\star\tb\th=1$, and $\star 1=-\tb\th$. 

2679: From this we see  that $F^{\xi}$ acts on each factor by 

2680: $F_1^{\phi}(\th)=\tb$, and 

2681: $F_1^{\phi}(1)=F_1^{\phi}(\tb)=F_1^{\phi}(\tb\th)=0$, as required.

2682: 

2683: With Lemma~\ref{lm-FN-equiv} and equivariance of $\xi_g$ 

2684: we have thus completed the proof of Theorem~\ref{thm-main}.

2685: Henceforth, we will use the simpler notation ${\cal V}={\cal V}^{FN}={\cal V}_{\cal N}$

2686:  \ep

2687: 

2688: \medskip

2689: 

2690: The $\SL(2,\R)$-action implies a Hard-Lefschetz decomposition \cite{GriHar78} as follows

2691: \beq\lbl{eq-Lefsch}

2692: H^*(J(\Sigma_g))\quad\cong\quad \bigoplus_{j=1}^{g+1} V_j\otimes W_{g,j}\;.

2693: \eeq

2694: Here, $V_j$ is the irreducible ${\mathfrak s}{\mathfrak l}_2$-module 

2695: with ${\rm dim}(V_j)=j\,$, and 

2696: \beq\lbl{eq-W}

2697: W_{g,j}\;:=\;\{u\in\ext {g-j+1}H_1(\Sigma_g)\,:\;\omega_g\wedge u=0\}

2698: \eeq	

2699: is the space of {\em isotropic} vectors of degree $(g-j+1)$, or, equivalently, the

2700: space of ${\mathfrak s}{\mathfrak l}_2$-highest weight vectors of weight $(j-1)$. On each of these

2701: spaces we have an action of the mapping class groups from (\ref{eq-FNmapcg})

2702: factoring through $\Sp(2g,\R)$.

2703: \begin{theorem}[\cite{GooWal98} Chapter 5.1.8]\lbl{thm-GW}

2704: Each $W_{g,j}$ is an irreducible $\Sp(2g,\R)$-module with fundamental highest weight

2705: $\varpi_{g-j+1}$ and dimension 

2706: $$

2707: {\rm dim}(W_{g,j})={2g\choose g-j+1}\,-\, {2g\choose g-j-1}

2708: $$

2709: In particular, the pair of subgroups 

2710: $$

2711: \SL(2,\R)\,\times\,\Sp(2g,\R)\;\;\subset\;\;{\rm GL}(H^*(J(\Sigma_g)))

2712: $$ 

2713: forms a Howe pair, that is, the two subgroups are exact commutants of each other.

2714: \end{theorem}

2715: The fundamental weights are given as in \cite{GooWal98} by 

2716: $\varpi_{k}=\epsilon_1+\ldots +\epsilon_{k}$ with $\epsilon_j$ as

2717: in  (\ref{eq-sproots}). 

2718: 

2719:  

2720: 

2721: 

2722: 

2723: In the decomposition into irreducible TQFT's the one for $j=1$ associated to the

2724: trivial $\SL(2,\Cc)$ representation plays a special role for invariants of closed

2725: manifolds. 

2726: 

2727: For any invariant, $\tau$, of closed 3-manifolds there is a standard ``reconstruction'' 

2728: of TQFT vector spaces 

2729: as follows. We take the formal $\kk$-linear span ${\mathfrak C}_g^+$ of 

2730: cobordisms $M:\emptyset\to\Sigma_g$

2731: and ${\mathfrak C}_g^-$ of cobordisms $N:\Sigma_g\to\emptyset$. We obtain a pairing

2732: ${\mathfrak C}_g^-\times {\mathfrak C}_g^+\to\kk:\,(N,M)\to \tau(N\circ M)$. If 

2733: ${\mathfrak N}^+_g\subset {\mathfrak C}_g^+$ is the null space of this pairing

2734: we define ${\cal V}^{\tau-rec}(\Sigma_g)={\mathfrak C}_g^+/{\mathfrak N}^+_g$.

2735:  For generic $\tau$ these vector spaces are infinite dimensional. The exception is when $\tau$

2736: stems from a TQFT. In this case 

2737: ${\cal V}^{\tau-rec}(\Sigma_g)^*={\mathfrak C}_g^-/{\mathfrak N}^-_g$, and the linear map 

2738: ${\cal V}^{\tau-rec}(P)$ associated to a cobordism $P$ is reconstructed from its

2739: matrix elements $\tau(N\circ P\circ M)$. This construction, which basically imitates the

2740: GNS construction of operator algebras, is folklore since the emergence of TQFT's and 

2741: appears, for example, in \cite{Tur94}.

2742: 

2743: \begin{theorem}\lbl{cor-dec}

2744: \ben

2745: \item

2746: The TQFT functor from Theorem~\ref{thm-main} decomposes into a direct sum

2747: $$

2748: {\cal V}\;=\;\;\bigoplus \R^{j}\otimes {\cal V}^{(j)}\;\;\;=\;\;\;

2749: {\cal V}^{(1)}\,\oplus\, \R^2\otimes {\cal V}^{(2)}

2750: \,\oplus\, \R^3\otimes{\cal V}^{(3)}\,\ldots \,

2751: $$

2752: of irreducible TQFT's with multiplicities. 

2753: \item 

2754: The associated vector space for each TQFT is ${\cal V}^{(j)}(\Sigma_g)=W_{g,j}$

2755: so that ${\cal V}^{(j)}(\Sigma_g)=0$ whenever $j>g+1$. 

2756: In particular, for any closed 3-manifold $M$ and $j>1$ we have ${\cal V}^{(j)}(M)=0$ so that

2757: ${\cal V}(M)={\cal V}^{(1)}(M)$. 

2758: \item 

2759: The vector spaces associated to the invariant $\pm\eta$ from (\ref{eq-defeta}) 

2760: are finite dimensional. The 

2761: reconstructed $\Z/2$-projective TQFT is ${\cal V}^{\eta-rec}={\cal V}^{(1)}$ with dimensions

2762: ${\rm dim}({\cal V}^{\eta-rec}(\Sigma_g))=

2763: {\rm dim}(W_{g,1})=\frac 2{g+2}{2g+1\choose g}\,$.

2764: \een

2765: \end{theorem}

2766: 

2767: 

2768: {\em Proof:} The fact that the TQFT's decompose in the prescribed manner follows 

2769: from the $\SL(2,\R)$-covariance. Irreducibility of each ${\cal V}^{(j)}$, meaning there

2770: are no proper sub-TQFT's, results from the fact that each $\Sp(2g,\Z)$ representation

2771: is irreducible so that in a sub-TQFT the vector spaces for each $g$ are either 

2772: ${\cal V}^{(j)}(\Sigma_g)$ or 0. Since the handle maps yield non-zero maps between 

2773: these vector spaces if one space is non-zero none of them can be. The reconstructed TQFT

2774: must be a quotient TQFT of ${\cal V}^{(1)}$, which is, however, irreducible. 

2775: Hence, they are equal.

2776: 

2777: \ep 

2778: 

2779: %%%%%%%%%%%%%%%%%  BEGIN CUT  %%%%%%%%%%%%%%%%%%%%%%%%%%%

2780: 

2781: \emptystuff{

2782: From the irreducible TQFT's in Theorem~\ref{cor-dec} we can construct a much larger class of TQFT's,

2783: which appear to be  related to higher rank gauge theories, 

2784: as follows. Let ${\cal P}^+

2785: \subset {\Z}^{0,+}[x_1,x_2,\ldots]$ be the set of formal power series 

2786: $$

2787: P(x_1,x_2,\ldots)=\sum_{k=1}^{\infty}\,\sum_{n_1, \ldots , n_k=1}^{\infty}c_{n_1,n_2,\ldots,n_k}

2788: x_1^{n_1}x_2^{n_2}\ldots x_k^{n_k}\;,

2789: $$

2790: such that all $c_{n_1,n_2,\ldots,n_k}\in {\Z}^{0,+}$ are non-negative integers, and for fixed $k$

2791: only finitely many $c_{n_1,n_2,\ldots,n_k}$ are non-zero. To every such $P$ we associate a

2792: TQFT by the formula

2793: \beq\label{eq-defpolyTQFT}

2794: {\cal V}^{(P)}\;=\;\bigoplus_{k=1}^{\infty}\bigoplus_{n_1, \ldots , n_k=1}^{\infty}

2795: \, \R^{c_{n_1,n_2,\ldots,n_k}}\otimes ({\cal V}^{(1)})^{\otimes n_1}\otimes 

2796: ({\cal V}^{(2)})^{\otimes n_2}\otimes \ldots\otimes ({\cal V}^{(k)})^{\otimes n_k}

2797: \;.

2798: \eeq

2799: For example ${\cal V}^{FN}={\cal V}^{(F)}$, where $F(x_1,x_2,\ldots)=\sum_j j x_j$. 

2800: The restriction on the coefficients together with the second part of Theorem~\ref{cor-dec}

2801: implies that all vector spaces are finite dimensional. 

2802: \begin{lemma}\label{lm-polyTQFT}

2803: The TQFT functor  ${\cal V}^{(P)}$ is well defined for every $P\in {\cal P}^+$ . 

2804: \end{lemma} 

2805: }

2806: 

2807: %%%%%%%%%%%%%%%%%%%  END CUT %%%%%%%%%%%%%%%%%%%%%%

2808: 

2809: 

2810: Let us finally give an alternative proof of Lemma~\ref{lm-defeta} using the language in

2811: which the Frohman Nicas invariant is constructed. 

2812: 

2813: We present $M$  by a Heegaard splitting 

2814: $M_{\psi}=h^-_g\circ{\bf I}_{\psi}\circ h^+_g$, as defined in 

2815: (\ref{eq-Heegaard}) and (\ref{eq-closM}).

2816:  The invariant is given as

2817: the matrix coefficient of $\ext g[\psi]$ for the basis vector

2818: ${\cal V}(h^+_g)=[a_1]\wedge[a_2]\wedge\ldots\wedge[a_g]$. If we denote by

2819: $[\psi]_{aa}$ the $g\times g$-block of $[\psi]$ acting on the Lagrangian

2820: subspace spanned by the $[a_i]$'s this number is just ${\rm det}([\psi]_{aa})$.

2821: At the same time, the Mayer-Vietoris sequence for $M_{\psi}$ shows that

2822: $[\psi]_{aa}$ is a presentation matrix for the group $H_1(M_{\psi},\Z)$

2823: so that the order of $H_1(M_{\psi},\Z)$ is, indeed, given by $\pm{\rm det}([\psi]_{aa})$.

2824:   

2825: \ep 

2826: 

2827: 

2828: \head{11. Alexander-Conway Calculus for 3-Manifolds}\lbl{S11}

2829: 

2830: Let $M$ be a 3-manifold   with an  epimorphism

2831: $\varphi:H_1(M,\Z)\twoheadrightarrow \Z$.  We recall the definition of

2832: the  {\em (reduced) Alexander polynomial}

2833: $\Delta_{\varphi}(M)$, as it is given in the case of 

2834: knot and link complements for example in \cite{BurZie}. 

2835: 

2836: 

2837: Let $\widetilde M\to M$

2838: be the cyclic cover associated to $\varphi$ and view $H_1(\widetilde M)$  as

2839: a $\Z[t,t^{-1}]$-module with $t$ acting by Decktransformation. Let 

2840: $E_1\subset \Z[t,t^{-1}]$ be the first elementary ideal generated by the 

2841: $n\times n$ minors of an $n\times m$ presentation matrix $A(t)$ of $H_1(\widetilde M)$. 

2842: Then $\Delta_{\varphi}(M)$ is the generator of the smallest principal idea containing

2843: $E_1$, or, equivalently, the g.c.d. of the $n\times n$ minors of a presentation

2844: matrix. Particularly, if $A(t)$ is a square matrix $\Delta_{\varphi}(M)=det(A(t))$ and

2845: if $n>m$, i.e., there are more rows than columns,  $\Delta_{\varphi}(M)=0$.  

2846:  

2847: 

2848: Another important invariant of a 3-manifold is its Reidemeister Torsion, which is

2849: obtained as the torsion of a  chain complex over $\Q [t,t^{-1}]$ obtained from a 

2850: cell decomposition of $\widetilde M$. 

2851: The Alexander polynomial turns out to be

2852:  almost the same as the Reidemeister Torsion of a 3-manifold.

2853: The relation described in the next theorem was first proven for homology circles by 

2854: Milnor and in the general case by Turaev.

2855: \begin{theorem}[\cite{Mil61}\cite{Tur76}]\label{thm-ReidAlex} 

2856: Let $M$ be a compact, oriented 3-manifolds,

2857: $\varphi: H_1(M)\to\Z$ an epimorphism as above, 

2858: $r_{\varphi}(M)$ its Reidemeister Torsion, and $\Delta_{\varphi}(M)$ its Alexander polynomial. 

2859: \ben

2860: \item If $\partial M\neq \emptyset$ then 

2861: $\displaystyle r_{\varphi}(M)\,=\,\frac 1 {(t-1)}\Delta_{\varphi}(M)$

2862: \item If $\partial M = \emptyset$ then 

2863: $\displaystyle r_{\varphi}(M)\,=\,\frac 1 {(t-1)^2}\Delta_{\varphi}(M)$

2864: \een

2865: \end{theorem}

2866: 

2867: 

2868: For a 3-manifold given by surgery along a framed link we will now give a

2869: procedure to compute the Alexander polynomial (and thus also Reidemeister Torsion). 

2870: 

2871:   

2872: Let ${\cal Z}\sqcup{\cal L}\subset S^3$ be a framed link consisting of a 

2873:  framed link ${\cal L}$ and 

2874:  a curve ${\cal Z}$ which has

2875: trivial linking number of all components of ${\cal L}$, i.e.,  

2876: with ${\cal L}\cdot{\cal Z}=0$.  

2877: We denote by $M_{{\cal Z},{\cal L}}^{\bullet}$ the manifold obtained by cutting

2878: out a tubular neighborhood of ${\cal Z}$ and doing surgery along ${\cal L}$. 

2879: Hence, $\partial  M_{{\cal Z},{\cal L}}^{\bullet}=S^1\times S^1$, with canonical

2880: meridian and longitude (given by 0-framing).  

2881: Also let $M_{{\cal Z},{\cal L}}$ be the closed manifold obtained 

2882: by doing 0-surgery  along ${\cal Z}$ so that 

2883: $M_{{\cal Z},{\cal L}}=M_{{\cal Z},{\cal L}}^{\bullet}\cup D^2\times S^1$. 

2884: The special component $\cal Z$ defines an epimorphism 

2885: $\varphi_{\cal Z}: H_1(M^{(\bullet)})\to\Z$, for example via intersection

2886: numbers with a Seifert surface. We write  

2887: $\Delta_{{\cal Z},{\cal L}}=\Delta_{\varphi_{\cal Z}}(M_{{\cal Z},{\cal L}})=

2888: \Delta_{\varphi_{\cal Z}}(M_{{\cal Z},{\cal L}}^{\bullet})$ for the associated 

2889: reduced Alexander polynomial,  which is the same in both cases. 

2890: 

2891: Consider a general Seifert surface $\Sigma^{\bullet}\subset S^3$ with 

2892: $\partial \Sigma^{\bullet} ={\cal Z}$ and 

2893: $\Sigma^{\bullet}\cap{\cal L}=\emptyset$.  By removing a neighborhood

2894: of the surface we obtain a relative cobordism 

2895: $C_{\Sigma}^{\bullet}=M_{{\cal Z},{\cal L}}^{\bullet}-\Sigma^{\bullet}

2896: \times(-\epsilon,\epsilon)$ from $\Sigma^{\bullet}$ to itself. 

2897: Similarly, $C_{\Sigma}=M_{{\cal Z},{\cal L}}-\Sigma

2898: \times(-\epsilon,\epsilon)$, where $\Sigma$ is the closed capped off

2899: surface $\Sigma^{\bullet}\cup D^2$. The cobordism $C_{\Sigma}$ is obtained

2900: from $C_{\Sigma}^{\bullet}$ by gluing in a full cylinder $D^2\times[0,1]$. 

2901: 

2902: 

2903: Denote by 

2904: $\psi^{(\bullet)}_{\pm}:\Sigma^{\bullet}_{\pm}\hookrightarrow  C_{\Sigma}$

2905: the inclusion maps of the bounding surfaces, and by 

2906: $$

2907: A_{\pm}=H_1(\psi^{(\bullet)}_{\pm}):\,H_1(\Sigma)\to 

2908: H_1(C_{\Sigma}^{(\bullet)})\to H_1^{\it free}(C_{\Sigma}^{(\bullet)})\;, 

2909: $$

2910: the maps on the free part of homology, where the free part is

2911: $G^{\it free}=\frac G {Tors(G)}\,$. As 

2912: $H_1(\widetilde M)\cong 

2913: H_1^{\it free}(\widetilde M)\oplus\,Tors(H_1(M))\otimes\Z[t,t^{-1}]$

2914: we will consider  the first elementary ideal for the free part, which differs only by

2915: a factor of $|Tors(H_1(M))|$. 

2916: 

2917: 

2918: Suppose first that $C$ does not have interior homology. This means the $A_{\pm}$

2919: can be presented as square matrices, and $A_+-tA_-$ is  a presentation matrix.  

2920: Consequently 

2921: $\Delta_{{\cal Z},{\cal L}}=\pm t^p det(A_+-tA_-)$. By some linear algebra

2922: \cite{FroNic92}

2923: this is the same as the Lefschetz polynomial 

2924: $$

2925: det(A_+-tA_-)\;=\;\sum_{k=0}^{2g}(-t)^{2g-k}

2926: trace\Bigl((\ext k A_+)\circ \star^{-1}\circ

2927:  (\ext {2g-k}A_-^*)\circ\star\Bigr)

2928: $$

2929: In  \cite{FroNic92} it is also shown that the expression inside the trace is

2930: the same as ${\cal V}^{FN}(C_{\Sigma}^{\bullet})_k$ or 

2931: ${\cal V}^{FN}(C_{\Sigma})_k$ depending on context. Hence, we have (multiplying by a unit $(-t)^{-g}$)

2932: that 

2933: %$$

2934: \begin{eqnarray}

2935: \Delta_{{\cal Z},{\cal L}} & = &\sum_{k=0}^{2g}(-t)^{g-k} trace({\cal V}^{FN}(C_{\Sigma})_k)

2936: \label{eq-leftrace}

2937: \\

2938: &=& trace((-t)^{-H}{\cal V}^{FN}(C_{\Sigma}))

2939: \label{eq-leftrace2}

2940: \\

2941: &=& \sum_{j=1}\;[j]_{-t}\,trace({\cal V}^{(j)}(C_{\Sigma}))\;=\;

2942: \sum_{j=1}\;[j]_{-t}\,\Delta_{{\cal Z},{\cal L}}^{(j)}\;,

2943: \label{eq-leftrace3}

2944: \end{eqnarray}

2945: where $[n]_q=\frac{q^n-q^{-n}}{q-q^{-1}}$. In (\ref{eq-leftrace2}) we used the generator $H$ of

2946: the $\SL(2,\R)$-Lefschetz action. Formula (\ref{eq-leftrace3}) is a consequence of the 

2947: Hard-Lefschetz decomposition from (\ref{eq-Lefsch}). We call the invariant

2948: $\Delta_{{\cal Z},{\cal L}}^{(j)}$ the {\em $j$-th Alexander Character} of the 

2949: Alexander polynomial. 

2950: 

2951: 

2952: In case $C$ does have interior rational homology the dimension of 

2953: $H_1^{\it free}(C_{\Sigma}^{(\bullet)})$ is bigger than $H_1(\Sigma)$ so that

2954: $H_1(\widetilde M)$ has $\Z[t,t^{-1}]$ as a direct summand. Consequently, the Alexander

2955: polynomial vanishes. At the same time ${\cal V}^{FN}(C_{\Sigma})$ is zero since it is

2956: a non-semisimple TQFT. Hence, (\ref{eq-leftrace3}) holds for all cases. 

2957: 

2958: 

2959: Suppose that in our presentation ${\cal Y}\subset S^3$ is the unknot. In this case we 

2960: can isotop the diagram ${\cal L}\sqcup {\cal Y}\subset S^3$ into the  form shown on the 

2961: right side of Figure~\ref{fig-stan}. Specifically, we arrange 

2962: it that the strands of one link component alternate orientations

2963: as we go from left to right. By application of the connecting annulus 

2964: moves, see for example \cite{Ker99}, we can modify the link further such that the resulting

2965: tangle ${\cal T}$ in the indicated box is admissible without through pairs

2966:  as described in the beginning of 

2967: Section~5 or, again, \cite{Ker99}. There is a canonical Seifert surface $\Sigma_{\cal T}$

2968: associated to a diagram as in Figure~\ref{fig-stan} obtained by surgering the disc

2969: bounded by ${\cal Z}$ along the framed 

2970: components of ${\cal L}$ emerging at the bottom side. By construction ${\cal T}$ is

2971: then a tangle presentation of $C_{\Sigma_{\cal T}}$. 

2972: 

2973: 

2974: \begin{figure}[ht]

2975: \begin{center}

2976: \leavevmode

2977: \epsfbox{stan.eps}

2978: \end{center}

2979: \caption{Standard Presentation }\lbl{fig-stan}

2980: \end{figure}

2981: 

2982: For the evaluation of this diagram it

2983:  is convenient to introduce an extension of ${\cal N}$ over $\Z[t,t^{-1}]$,

2984: given by $\Z[\gamma^{\pm 1}]\ltimes {\cal N}$. The extra generator $\gamma$ is

2985: group like with $S(\gamma)=\gamma^{-1}$ and it acts on ${\cal N}$ by

2986: $\gamma x\gamma^{-1}=t^Hx=t^{deg(x)}x$ for $x\in{\cal N}$ and $deg(x)$ the degree

2987: for homogenous elements. 

2988: 

2989: In order to evaluate the diagram we apply the Hennings substitutions for crossing 

2990: (\ref{fig-Dec-cross})  and rules (\ref{eq-elem-mv})   through (\ref{pic-Rclos}) 

2991:  to the  ${\cal T}$ part to obtain a combination of ${\cal N}$-decorated 

2992: arcs as in (\ref{pic-Rbot}) and (\ref{pic-Rtop}). Furthermore, we

2993: remove the circle ${\cal Y}$ at the expense of introducing a $\gamma$-decoration

2994: on each strand. The Hennings procedure is continued with the extended algebra over

2995: $\Z[t,t^{-1}]$. It is easy to see that the elements that have to be evaluated against

2996: the integral all lie in $\Z[t,t^{-1}]\otimes{\cal N}$ and that $\mu$ is cyclic also 

2997: with respect to $\gamma$. Hence, the evaluation is well defined. 

2998: 

2999: \begin{lemma}

3000: The evaluation procedure for a diagram as in Figure~\ref{fig-stan} yields the 

3001: Alexander polynomial.

3002: \end{lemma}

3003: 

3004: {\em Proof:} The standard evaluation of ${\cal T}$ yields a sum of diagrams with top and bottom 

3005: arcs, where the $j$-th bottom arc is decorated by $b_j$ and the $j$-th top arc by $c_j$ 

3006: as in (\ref{pic-Rbot}) and (\ref{pic-Rtop}). Hence, ${\cal V}_{\cal N}(C_{\Sigma})$ is the sum 

3007: over all diagrams of linear maps

3008: $\bigotimes_j^g (b_j\otimes \mu(S(\_)c_j))$.

3009:   The extended evaluation yields closed curves, each of

3010: which is decorated with four elements $b_j$, $c_j$, $\gamma$, and $\gamma^{-1}$. Using the 

3011: antipodal sliding rule from (\ref{pic-Rclos}) we collect them at one side of a circle so that the 

3012: evaluation becomes

3013: $$

3014: \mu(S^{-1}(b_j)\gamma c_j\gamma^{-1})\;=\;(-1)^{deg(b_j)}t^{deg(c_j)}\mu(S(b_j)c_j)

3015: \;=\; (-t)^{-deg(b_j)}trace(b_j\otimes \mu(S(\_)c_j))\;.

3016: $$

3017: Note here, that $S^2(b_j)=(-1)^{deg(b_j)}$ and that the evaluation is non zero only if 

3018: $deg(c_j)+deg(b_j)=0$. The sum (over all decorations) of the products (over $j$)

3019: of these individual traces is thus just the trace of $(-t)^{-H}{\cal V}_{\cal N}(C_{\Sigma})$. 

3020: Since this is (up to sign) identical with   $(-t)^{-H}{\cal V}^{FN}(C_{\Sigma})$ it follows

3021: from (\ref{eq-leftrace2}) that the evaluation gives the Alexander polynomial.

3022: 

3023: \ep

3024: 

3025: 

3026: The evaluation of a standard diagram can be described also more explicitly without the use

3027: of the $\Z[\gamma]$ extension. Let ${\cal T}^{\#}: 2g\to 0$ be the diagram consisting of 

3028: the tangle ${\cal T}: g\to g$ and the lower arcs. That is, 

3029: ${\cal T}=(1^g\otimes {\cal T}^{\#})\circ (X_g\otimes 1^g)$ and 

3030: ${\cal T}^{\#}= (X_g^{\dagger})\circ(1^g \otimes {\cal T})$, where $X_g^{\dagger}$ is the

3031: upside down reflection of $X_g\,$. We define $A^{\gamma}\in

3032: {\cal N}_0^{\otimes 2}\otimes\Z[t,t^{-1}]$ as 

3033: \beq\label{eq-defAg}

3034: A^{\gamma}\;=\;({\gamma}\otimes 1)A({\gamma^{-1}}\otimes 1)\;=\;

3035: \frac 1 i \Bigl( \rho\otimes 1\,+\,1\otimes\rho\,-\,t^{-1}\tb\otimes\th\,+\,t

3036: \th\otimes\tb\Bigr)\;\;\;.

3037: \eeq

3038: Moreover, we define 

3039: $A_{\{g\}}^{\gamma}\in{\cal N}_0^{\otimes 2g}\otimes\Z[t,t^{-1}]$  from $A^{\gamma}$

3040: as $A_{\{g\}}$ in (\ref{eq-Xn}) is defined from $A$ in (\ref{eq-defAtens}) and (\ref{eq-X1}), or,

3041: equivalently, by

3042: $$

3043: A_g^{\gamma}\;=\;({\gamma}^{\otimes g}\otimes 1^{\otimes g})\circ A_{\{g\}}\circ 

3044: ({(\gamma^{-1})}^{\otimes g}\otimes 1^{\otimes g})\;. 

3045: $$

3046: This tensor is assigned to the upper arcs and the $\gamma$ elements in the standard diagram.

3047: Hence, by the extended Hennings evaluation procedure the  Alexander polynomial is given by the

3048: composition 

3049: $$

3050: \Delta_{{\cal Z},{\cal L}}\;=\;{\cal V}^{FN}({\cal T}^{\#}) (A^{\gamma}_g)\;,

3051: $$

3052: where we think of ${\cal V}^{FN}({\cal T}^{\#}):{\cal N}_0^{\otimes 2g}\to \Cc$ as being

3053: naturally extended to a $\Z[t,t^{-1}]$-map from ${\cal N}_0^{\otimes 2g}\otimes\Z[t,t^{-1}]

3054: \,\to\, \Cc[t,t^{-1}]\,$. 

3055: 

3056: 

3057: For further evaluation we use  Theorem~\ref{thm-solve} to write 

3058: ${\cal V}^{FN}({\cal T}^{\#})=\sum_{\nu}{\cal V}^{FN}(E_{\nu})$

3059: as a combination of elementary tangles 

3060: $E_{\nu}=(M_{k_1}\otimes\ldots M_{k_r}\otimes \epsilon^{\otimes N})\circ B\,$ so that

3061: the Alexander polynomial is the sum of polynomials $E_{\nu}(A^{\gamma}_g)$. 

3062: For the computation of these elementary polynomials it is convenient to use 

3063: the following

3064: graphical notation. As shown in (\ref{eq-symhopfeva}) we indicate the morphism $M_k$

3065: by a tree with $k$ incoming branches. The  morphism  $X_1$ is drawn as an arc and

3066: $X_g$ as  $g$ concentric arcs. 

3067: \beq\label{eq-symhopfeva}

3068: \epsfbox{symhopfeval.eps}

3069: \eeq

3070: For $E=(M_1^{\otimes 3}\otimes M_2 \otimes M_4\otimes \epsilon)\circ B$ we obtain 

3071: the composite shown on the right of (\ref{eq-symhopfeva}). Using relations 

3072: $(\mu\otimes 1)A^{\gamma}=( 1\otimes \mu)A^{\gamma}= 1$, 

3073: $(\varepsilon\otimes 1)A^{\gamma}=( 1\otimes\varepsilon)A^{\gamma}= \frac 1 i\rho$,

3074: and $\mu(x\frac 1i \rho)=\varepsilon(x)$ we find the  graphical relations

3075: depicted in (\ref{eq-elimeval}). 

3076: \beq\label{eq-elimeval}

3077: \epsfbox{elimeval.eps}

3078: \eeq

3079:  Now, to each of the arcs 

3080: the tensor $A^{\gamma}$ is associated containing the four terms $\rho\otimes 1$,

3081: $1\otimes \rho$, $\tb\otimes\th$, and $\th\otimes\tb$ with coefficients of the form

3082: $\pm it^m$. We represent the elementary

3083: polynomial thus as a sum over all combinations of these terms, i.e., $4^g$ terms 

3084: for $A^{\gamma}_{\{g\}}$. We indicate a combination in a diagram by drawing a line

3085: with a down arrow for $\tb$, a line with an up arrow for $\th$, a line with arrows for

3086: $\rho$ and a dashed line for $1$. Hence, (\ref{eq-defAg}) becomes the first line in 

3087: (\ref{eq-termsarr}).

3088: \beq\label{eq-termsarr}

3089: \epsfbox{termsarr.eps}

3090: \eeq

3091: The tensors associated to the $M_k$ are non zero only in two cases. Namely, if one

3092: element is $\th$, another $\tb$ and all other $1$, or if one element is $\rho$

3093: and all others 1. In diagrams we obtain the evaluation rules as depicted. All

3094: other configurations are evaluated to zero. 

3095: 

3096: For an elementary diagram let $N_x(=g)$ be the number of arcs at the top, 

3097: $N_0$ the number of $\varepsilon$'s, and $N_k$ the number of $M_k$'s at 

3098:  the bottom of the diagram for $k\geq 1$. Let us also call an elementary

3099: diagram {\em reduced} if $N_0=N_1=0$.  We can now give the recipe for evaluating

3100: elementary diagrams:

3101: 

3102: \begin{propos}\label{prop-HopfAlexeval}

3103: \ 

3104: 

3105: \ben

3106: \item We have the  relations \ \  \ \ 

3107: $\displaystyle 

3108: 2N_x\,=\, N_0\,+\,\sum_{k\geq 1} kN_k\;,$  \ \  and  \ \ 

3109: $\displaystyle  N_x\;=\;\sum_{k\geq 1} N_k\;.$

3110: \item Every elementary diagram is zero or equivalent to a reduced one by application

3111: of the moves in (\ref{eq-elimeval}). 

3112: \item A reduced diagram is non zero only of $N_j=0$ for $j\geq 3$. That is, if

3113: the diagram is of the form $D=M_2^{\otimes g}\circ B\circ X_g$.

3114: \item A contributing reduced diagram $D=P_1\sqcup \ldots\sqcup P_n$ is the union of closed 

3115: paths $P_j$, and the

3116: polynomial $\Delta_D=\prod_j\Delta_{P_j}$ assigned to $D$ is the product 

3117: of the polynomials assigned to the the components $P_j$. 

3118: \item The polynomial associated to a connected component is 

3119: $$

3120: \Delta_P\;=\;2\,-(-1)^b\,(t^p+t^{-p})\;,

3121: $$

3122: where $p$ is the algebraic intersection number of the closed 

3123: path $P$ with  a radial line segment $\Xi$ as in (\ref{eq-symhopfeva}), and $b$

3124: is the total  number of half twists (or antipode insertions) in $B$. 

3125: \een 

3126: \end{propos}

3127: 

3128: {\em Proof:} {\em 1.}

3129: In a diagram as in (\ref{eq-symhopfeva}) the number of strands entering from

3130: the top is $2N_x$, two for each arc, and the number of strands entering from the bottom 

3131: is $N_0+\sum_{k\geq 1}kN_k$. Obviously, 

3132: both  numbers have to be equal. For an admissible configuration

3133: of a contributing diagram we can also call weighted edges, where the dashed ones are

3134: weighted 0, the ones with one arrow as 1, and those with double arrows as 2. The

3135: top part of the diagram shows that the total weight has to be $2N_x$ since every 

3136: admissible arc has weight 2. Also every tree has weight 2 and the $\epsilon$'s have

3137:  weight 0 so that the total weight must also be given by $\sum_{k\geq 1} 2N_k$. 

3138: 

3139: {\em 2.} This is clear since every non-reduced one allows the application of a 

3140: move that reduces the number of edges. 

3141: 

3142: {\em 3.} If we subtract twice the second identity in {\em i)} from the first we

3143: find $0=N_0-N_1+N_3+2N_4+3N_5+\ldots\,$. In the reduced case with $N_0=N_1=0$ this 

3144: implies $0=N_3=N_4=N_5=\ldots$ since these are all non negative integers. 

3145: 

3146: {\em 4.} Any graph where all vertices have valency 2 is the union of closed 

3147: paths. Since we have a symmetric commutativity constraint we can untangle components

3148: from each other and move them apart. The evaluation of disjoint unions of diagrams

3149: is given by their products.

3150: 

3151: {\em 5.} There are four configurations that contribute to $\Delta_P$ for a 

3152: closed path. Two if them are given by dashed lines alternating with double arrow

3153: lines. This corresponds to paring factors $\frac 1 i\rho$ with integrals $\mu$ in

3154: two different ways each evaluated as 1. Thus these two cases contribute the 2 in the 

3155: expression. The other two configurations are given by two orientations of $P$ with

3156: single arrows everywhere. For one given orientation we get from 

3157: (\ref{eq-termsarr}) a factor $\frac 1 i t$ if

3158: $P$ crosses $\Xi$ left to right and a factor $\frac 1 i (-t^{-1})$ if $P$ crosses

3159: right to left. Thus the arcs yield a tensor $\pm (\frac 1 i)^g t^b(x_1\otimes\ldots\otimes x_{2g})$,

3160: where each $x_i$ is either $\th$ or $\tb$. Application of $B$ yields a tensor

3161: $\pm (\frac 1 i)^g t^b(y_1\otimes\ldots\otimes y_{g})$ where each $y_j$ is either $\th\otimes\tb$

3162: or $\tb\otimes\th$ depending on which way the path runs through the $M_2$ piece. The 

3163: pairwise multiplication thus yields the tensor $\pm t^b(\frac 1i\rho)^{\otimes g}$ and 

3164: evaluation against $\mu$ the factor $\pm t^b$. For the opposite orientation the tensor

3165: for the arcs is obtained by exchanging $t$ for $t^{-1}$ and multiplying a factor $(-1)^g$. 

3166: The factor picked up by 

3167: application of $B$ is unchanged, and in the evaluation against the $\mu$ we pick up a factor 

3168: $(-1)^g$ because the orders of $\th$ and $\tb$ are exchanged canceling the one from the top. 

3169: Hence the contribution for the opposite orientation is the same with $t$ and $t^{-1}$ exchanged.

3170: Thus $\Delta_P=2\pm(t^b+t^{-b})$. The sign can be determined by evaluating the polynomial

3171: at $t=1$. This is identical with the usual Hennings invariant of the 3-manifold given by 

3172: surgery along a link associated to the connected diagram $P$ as follows. 

3173: 

3174: First choose 

3175: over and under crossing for $P$ pushing it slightly outside the plane of projection into

3176: a knot $P^*$. This knot is thickened to a band $N(P^*)$, which is  parallel to the

3177: plane of projection except for half twists that are

3178: introduced at the points where $B\subset P$ has antipodes

3179: inserted. 

3180: 

3181: Consider the link $\partial N(P^*)$ given by the edges of the band. Generically this

3182: link consists of  parallel strands that double cross as in Figure~\ref{fig-MC-tgl} at

3183: simple crossings of $P^*$ and has $\Gamma$-diagram  also as in Figure~\ref{fig-MC-tgl}

3184: for every half twist. We further modify this link  at some generic point in the band

3185: by replacing  the parallel strands 

3186: by a configuration with a connecting annulus as in the $\sigma$-Move of (\ref{eq-boundmove}).

3187: We obtain a two component link ${\cal L}_P={\cal A}_P\sqcup{\cal C}_P$, where  

3188: ${\cal A}_P$ is the 0-framed annulus. The other part ${\cal C}_P$ bounds the disc obtained

3189: by removing the small piece from the band where we applied the $\sigma$-Move and thus 

3190: carries a natural framing.  We have by construction that 

3191: $\Delta_P(1)=\pm\eta(M_{{\cal L}_P})$ with $\eta$ as in (\ref{eq-defeta}). For self intersection

3192: numbers we clearly have ${\cal A}_P\cdot {\cal A}_P=0$ and 

3193: ${\cal C}_P\cdot {\cal C}_P=0$. For an even number of twists in the band $N(P^*)$ we

3194: obtain also ${\cal A}_P\cdot {\cal C}_P=0$ and for an odd number of twists we have 

3195: ${\cal A}_P\cdot {\cal C}_P=\pm 2$. Hence $\eta(M_{{\cal L}_P})=0$ in the first case

3196: and $\eta(M_{{\cal L}_P})=4$ in the second. 

3197: \ep 

3198: 

3199: 

3200: Note, that the form of the $\Delta_P$ implies again the symmetry $\Delta(t)=\Delta(t^{-1})$

3201: of the Alexander polynomial. In order to instill some confidence in our procedure let us

3202: recalculate the familiar formula for the left-handed trefoil in this setting. Using the Fenn Rourke

3203: move from  Figure~\ref{fig-FR} we  present the trefoil as an unknotted curve $\cal Z$ 

3204:  in   a surgery diagram of Borromean rings as in (\ref{eq-trefoil}). 

3205: \beq\label{eq-trefoil}

3206: \epsfbox{trefoil.eps}

3207: \eeq

3208: The standard form is obtained by moving ${\cal C}_1$ to the right off $\cal Z$ and 

3209: letting ${\cal C}_2$ follow at the ends. The tangle ${\cal T}^{\#}$ is then as depicted on

3210: the left of (\ref{eq-trefoiltgl}) below.  Using the framing moves from Figure~\ref{fig-frame} we 

3211: expand it into elementary diagrams as on the right of  (\ref{eq-trefoiltgl}).

3212: \beq\label{eq-trefoiltgl}

3213: \epsfbox{trefoiltgl.eps}

3214: \eeq

3215: The translation into Hopf algebra diagrams and subsequently polynomials is indicated 

3216: next in (\ref{eq-trefoilhopf}). 

3217: \beq\label{eq-trefoilhopf}

3218: \epsfbox{trefoilhopf.eps}

3219: \eeq

3220: Thus the polynomial comes out to be  $t+t^{-1}-1$ as it had to be. 

3221: The same calculation carries through

3222: if we change the framings $f_j$ of the components ${\cal C}_j$ in (\ref{eq-trefoil}). The 

3223: difference is the sign of the first summand, that is 

3224: $\Delta_{\cal Z}=f_1 f_2(t+t^{-1}-2)+1$. Thus, if we flip both framings we obtain the right-handed

3225: trefoil with the same polynomial. If we flip only one framing so that $f_1=-f_2$ we obtain one

3226: of two figure-eight knots with polynomial $-t-t^{-1}+3$. Many other Alexander polynomials with

3227: multiple twists as for example $(p,q,r)$-pretzel knots  can be computed quite conveniently

3228: in this fashion using Fenn Rourke moves and the nilpotency of the ribbon element $v^k=1+k\rho$. 

3229: Thus, our method proves to be 

3230:  quite useful in the calculation of the Alexander Polynomial for knots although its primary 

3231: application is the generalization to 3-manifolds. 

3232: 

3233: 

3234: We describe next a more systematic way to unknot the special strand $\cal Z$ in a general diagram 

3235: more akin the traditional skein theory. The additional 

3236: relations that allow us to put any diagram ${\cal L}\sqcup {\cal Z}$ into a standard form

3237: are as follows.    

3238: 

3239: \begin{propos}\label{propos-Yskein}

3240:  We have the following two skein relations for the special strand ${\cal Z}$

3241: \beq\label{eq-Yskein}

3242: \epsfbox{Yskein.eps}

3243: \eeq

3244: and

3245: \beq\label{eq-Ycoupon}

3246: \epsfbox{Ycoupon.eps}

3247: \eeq

3248: as well as the slide and cancellations moves analogous to (\ref{eq-1handlemove}),

3249: and a vanishing property as in (\ref{eq-pairvanish}). %  Lemma~\ref{lm-1handlemoves}

3250:   %  (\ref{eq-1handlecanc})   (\ref{eq-1handleoppcanc}).

3251:  

3252: These equivalences allow us to express the Alexander polynomial of

3253:  any diagram ${\cal Y}\sqcup {\cal L}\subset S^3$ as a combination of the 

3254: evaluations of diagrams in standard form. 

3255: \end{propos}

3256: 

3257: {\em Proof:}  As before we change a self crossing of ${\cal Y}$ by sliding a 1-framed annulus 

3258: ${\cal A}$ over the crossing. Note, that we do not have to keep track of the framing of 

3259: ${\cal Y}$ as it is unchanged and by convention zero. 

3260:   Using the orientation of ${\cal Y}$ we can  do this 

3261: such that the linking numbers of ${\cal Y}$ and ${\cal A}$ remain zero. It is easy to see

3262: that we can bring a diagram into the standard position as in  Figure~\ref{fig-stan} without

3263: ever sliding a strand over the new component ${\cal A}$. The evaluation is obtained as the

3264: weighted trace over the linear map associated by  ${\cal V}_{\cal N}$ to the

3265:  cobordism represented by the tangle,

3266: which contains ${\cal A}$. Inserting 

3267: the relation from Figure~\ref{fig-frame} we see that this linear map, and hence

3268: the associated polynomial, is the 

3269: combination of the one for which ${\cal A}$ has been removed and the one for which the framing

3270: of ${\cal A}$ has been shifted by one. In both cases the unknotting procedure can be reversed

3271: so that we obtain the original pictures with ${\cal A}$ removed or its framing shifted by one. 

3272: The situation in which ${\cal A}$ is removed corresponds to the opposite crossing. In the other

3273: contribution we have a 0-framed annulus around the crossing which can be rewritten as an index-1

3274: surgery represented by a pair of coupons. This yields (\ref{eq-Yskein}). 

3275: 

3276: The coupon combination in (\ref{eq-Ycoupon}) can be reexpressed by a tangle as in 

3277: (\ref{eq-coupontorus}), can be isotoped into the position shown in (\ref{eq-Yext}). 

3278: \beq\label{eq-Yext}

3279: \epsfbox{Yext.eps}

3280: \eeq

3281: The extra tangle piece ${\cal Q}$ maps to the identity on a torus block. More precisely,

3282: ${\cal V}_{\cal N}({\cal Q}\sqcup{\cal T})=id_{{\cal N}_0}\otimes 

3283: {\cal V}_{\cal N}({\cal T})$. The weighted traces thus differ by a factor

3284:  $trace_{{\cal N}_0}((-t)^{-H})=-t+2-t^{-1}=-(t^{\frac 12}-t^{-\frac 12})^2$. 

3285: \ep

3286: 

3287: For ordinary link and knot complements there are well known skein relations 

3288: that uniquely characterize the Alexander-Conway polynomial of the knot,

3289: see for example \cite{BurZie} Chapter 12.C. 

3290: 

3291: \begin{cor}

3292: For ordinary knot complements (that is if ${\cal L}=\emptyset$) the relations 

3293: Proposition~\ref{propos-Yskein} reduce to the ordinary Alexander-Conway skein

3294: relations. 

3295: \end{cor}

3296: 

3297: {\em Proof:} It is clear that with Proposition~\ref{propos-Yskein} we can resolve every diagram

3298: into disjoint circles in the plane with coupons on them in exactly the same way as 

3299: for the Alexander-Conway polynomial. The difference is that wherever we pick up a 

3300: factor $(t^{\frac 12}-t^{-\frac 12})$ from the smoothening in the traditional calculus

3301: we obtain a factor $\frac 1 i$ and a pair of coupons in our case, but all other 

3302: numbers are the same. 

3303: 

3304: 

3305: Suppose now after resolving the crossings 

3306: we have more than one circle. Since the strand $\cal Z$ has to run

3307: though all of these components we must have coupons that are paired but on different 

3308: circles. By (\ref{eq-pairvanish}) of Lemma~\ref{lm-1handlemoves} it follows that such a

3309: configuration must vanish. In the Alexander-Conway calculus we also have the rule

3310: that the link invariant for the unlinked union of an unknot with a non-trivial link

3311: is zero. Hence we only need to compare the contributions that come from single circles.

3312: If in the process of applying the skein relations we carried out $N$ smoothenings of

3313: crossings the circle will carry $2N$ coupons. 

3314: 

3315: Next we claim that it is not possible to slide two paired coupons in adjacent position.

3316: To this end note that  the coupons in the resolution of 

3317: Proposition~\ref{propos-Yskein} stay all on one side of the

3318: special strand. I.e., in the depicted orientation of $\cal Z$

3319: the coupons are always on the left of $\cal Z$. Thus, 

3320: if they become adjacent 

3321: we would have a situation as in (\ref{eq-1handlecanc}) of  Lemma~\ref{lm-1handlemoves}. 

3322: This is not possible since then $\cal Z$ would have at least two components. 

3323: Thus the number $2N$  of coupons will remain the same under handle slides.

3324: 

3325: We next observe that a circle with edges that are labeled in pairs and subject to 

3326: handle slides also occurs in the classification of compact, oriented surfaces via

3327: their triangulations as in \cite{massey} Chapter 1. It is shown there that any

3328: such configuration is under application of handle slides and cancellation

3329: moves as in (\ref{eq-1handlecanc}) equivalent to a sequence of blocks as in

3330: (\ref{eq-Ycoupon}). As before we may assume that all coupons lie on one side of 

3331: the circle. In fact, as $\cal Z$ is connected we see from \cite{massey} that 

3332: we can move to the configuration in standard block form without the use of 

3333: cancellations.

3334: 

3335: Thus, we have $\frac N 2$ 4-coupon (torus) blocks as in  (\ref{eq-Ycoupon})

3336: contributing a factor of 

3337: $(-(t^{\frac 12}-t^{-\frac 12})^2)^{\frac N 2}=(i)^N(t^{\frac 12}-t^{-\frac 12})^N$.

3338: Recall that in each resolution we also had a factor $\frac 1 i$ so that the 

3339: total factor for the circle is just $(t^{\frac 12}-t^{-\frac 12})^N$ and $N$ is the

3340: number of smoothenings. But  $(t^{\frac 12}-t^{-\frac 12})$ is precisely the factor

3341: assigned to each smoothening by the usual Alexander-Conway calculus.

3342: 

3343: \ep 

3344: 

3345: Although we now have a systematic procedure for computing the Alexander polynomial

3346: of a 3-manifold  it is often times efficient to use the skein relations leading up

3347: to it directly. We illustrate this by  computing $\Delta_{{\cal C}_{k,l},{\cal Z}}$,

3348: where ${\cal C}_{k,l}$ is the component depicted in (\ref{eq-exampleknot})

3349: \beq\label{eq-exampleknot}

3350: \epsfbox{exampleknot.eps}

3351: \eeq

3352: The two middle strands are twisted with each other $k$ times generating $2k$ crossings,

3353: and we have $q$ full circles on the upper strand indicating shifts in the framing by -1.

3354: The definition for $k<0$ or $q<0$ is given by choosing the opposite twistings. 

3355: 

3356: \begin{lemma} The Alexander Polynomial of $M_{{\cal C}_{k,l},{\cal Z}}$ is given  by 

3357: the ordinary  Alexander polynomial of the knot as follows:

3358: $$

3359: \Delta_{{\cal C}_{k,l},{\cal Z}}\;=\;i(k(t+t^{-1})-q)\Delta_{{\cal Z}}

3360: $$

3361: \end{lemma}

3362:  

3363: 

3364: {\em Proof:} We combine every twist with two circles so that we have $k$ twist

3365: configurations as in Figure~\ref{fig-FR} and $l=q-2k$ remaining circles. Applying

3366: the Fenn Rourke move to each of these we obtain a configuration in which we have a 

3367: parallel instead of twisted pair of strands in the middle surrounded by $k$

3368: annuli with an empty circle on them. In addition, we have $k$ separate annuli with 

3369: full circles. Denote by $\Delta_{k,l}$ the associated Alexander Polynomial. For $k>0$

3370: we choose one of the first annuli and apply the framing shift relation

3371: (\ref{fig-frame}) to the empty circle on it. In the second contribution we omit the

3372: dotted line so that we obtain the same configuration with one less annulus around

3373: the double strands. The factor $i$ in  (\ref{fig-frame}) is canceled against one

3374: of the separate annuli with a full circle so that the second contribution is exactly 

3375: $\Delta_{l,k-1}$. In the first contribution we have a 0-framed annulus which, by 

3376: Figure~\ref{fig-modif}, can be turned into a pair of coupons. The other $k-1$ coupons

3377: can thus be slid off and canceled against $k-1$ annuli with full circles. Moreover,

3378: the remaining $l$ full circles on the upper strand can be removed since inserting a

3379: dotted line leaves two isolated coupons, which yields zero. The resulting 

3380: configuration is the knot $\cal Z$ with a tangle piece ${\cal Q}$ as in 

3381: (\ref{eq-Yext}), contributing an

3382: extra factor $-(t^{\frac 12}-t^{-\frac 12})^2$, and an 

3383: extra annulus with full circle with a factor $-i$. We thus obtain the recursion

3384: relation  $P_{k,l}\,=\,i(t^{\frac 12}-t^{-\frac 12})^2\,+\,P_{k-1,l}$ so that

3385: $P_{k,l}\,=\,ik(t^{\frac 12}-t^{-\frac 12})^2\Delta_{\cal Z}\,+\,P_{0,l}$. 

3386: But the configuration

3387: for $k=0$ is the separate union of $\cal Z$ and an annulus with $l$ full circles.

3388: The latter yields a factor $-il$ so that 

3389: $P_{k,l}\,=\,i(k(t^{\frac 12}-t^{-\frac 12})^2-l)\Delta_{\cal Z}$, which computes

3390: to the desired formula.

3391: 

3392: \ep 

3393: 

3394: 

3395:   

3396: \head{12. Lefschetz compatible Hopf algebra structures on $H^*(J(\Sigma))$}\lbl{S12}

3397: 

3398: 

3399: It is easy to see that the natural ring structure on the cohomology 

3400: $H^*(J(\Sigma))\cong\ext * H_1(\Sigma)$ is not compatible with the $\SL(2,\R)$

3401: Lefschetz action as described in Section~10. For example $E(x\wedge y)=x\wedge y\wedge \omega$

3402: but $(Ex)\wedge y +x\wedge (Ey) =2x\wedge y\wedge \omega$. The isomorphism with 

3403: ${\cal N}_0^{\otimes g}$ however induces another multiplication structure compatible 

3404: with the $\SL(2,\R)$ action. In this section we will describe it explicitly. 

3405: 

3406: The $\Z/2$-graded Hopf algebra structure on ${\cal N}_0$ given in 

3407: Lemma~\ref{lm-super-Hopf}  extends to a $\Z/2$-graded Hopf algebra 

3408: structure ${\cal H}_{\cal N}$ on ${\cal N}_0^{\otimes g}$ with 

3409: $$

3410: (x_1\otimes \ldots\otimes x_g)(y_1\otimes \ldots\otimes y_g)

3411: \;\;=\;\;(-1)^{\sum_{i<j} d(x_j)d(y_i)}\,x_1y_1\otimes \ldots\otimes x_gy_g\;.

3412: $$

3413: The formula for ${\bf \Delta}$ is the dual analog.

3414: The precise form of ${\cal H}_{\cal N}$ is given as follows:

3415: 

3416: 

3417: \begin{lemma}\lbl{lm-StrucN}  For a choice of basis of $\R^g$ there 

3418: is  a natural isomorphism of Hopf algebras

3419: $$

3420: \varrho\;:\;\;\ext *(\E\otimes \R^g)\;\;\isto\;\;{\cal N}_0^{\otimes g}\;\;

3421: $$

3422: 

3423: so that \ \ \  ${\rm Aut}({\cal N}_0^{\otimes g}, 

3424: {\cal H}_{\cal N})\;\cong\;{\rm GL}(\E\otimes \R^g)$. 

3425: \end{lemma}

3426: 

3427: {\em Proof:} Let $\{e_j\}$ be a basis of $\R^g$. The generating set of primitive vectors

3428: of  $\ext *(\E\otimes \R^g)$ is given by $\E\otimes \R^g$. On this subspace we set 

3429: $\varrho(w\otimes e_j)=1\otimes\ldots 1\otimes w \otimes 1\ldots\otimes 1$, with $w$ in

3430: $j$-th position. We easily see  that the vectors in $\varrho(\E\otimes \R^g)$ form 

3431: again a generating set of anticommuting, primitive vectors of ${\cal N}_0^{\otimes g}$ 

3432: so that $\varrho$ extends to a Hopf algebra epimorphism. Equality of dimensions thus implies

3433: that $\varrho$ is an isomorphism.\ep

3434: 

3435:  The canonical $\SL(2,\R)$-action on ${\cal N}_0^{\otimes g}$ is still compatible 

3436: with ${\cal H}_{\cal N}$ since it preserves the degrees and factors. Under the isomorphism 

3437: in Lemma~\ref{lm-StrucN} it is readily identified as the $\SL(2,\R)$-action on the $\E$-factor. 

3438: The remaining action on the $\R^g$-part can be understood geometrically. Specifically, 

3439: $\Sp(2g,\Z)$ acts on ${\cal N}_0^{\otimes g}$ since the ${\cal V}$-representation of the 

3440: mapping class group factors through a the symplectic group with representation 

3441: ${\cal V}^{Sp}:\,\Sp(2g,\Z)\to {\rm GL}({\cal N}_0^{\otimes g})\,:\,[\psi]\mapsto 

3442:  {\cal V}^{Sp}([\psi]):={\cal V}({\bf I}_{\psi})$. 

3443: For a given decomposition into Lagrangian subspaces we denote the standard inclusion

3444: \beq\lbl{eq-defkappa}

3445: \kappa\,:\;

3446: \SL(g,\Z)\,\into {\rm }GL(g,\Z)\,\into \,\Sp(2g,\Z)\;:

3447: \;\;A\,\mapsto\, \kappa(A):=\,A\oplus (A^{-1})^T\;.

3448: \eeq

3449: 

3450:  

3451: \begin{lemma}\lbl{lm-slgz-invar}

3452: The action of $\SL(g,\Z)$ on ${\cal N}_0^{\otimes g}$ induced by ${\cal V}^{Sp}\circ\kappa$

3453: is compatible with ${\cal H}_{\cal N}$, and under the isomorphism  $\varrho$ from 

3454: Lemma~\ref{lm-StrucN} it is identical with the $\SL(g,\Z)$-action on $\R^g$ for the given basis. 

3455: In particular,  it commutes with the $\SL(2,\R)$-action so that we have the following

3456: natural inclusion of the Howe pairs:

3457: $$

3458: \SL(2,\R)\times \SL(g,\Z) \;\;\subset\;\; 

3459: {\rm GL}(\E\otimes \R^g)={\rm Aut}({\cal N}_0^{\otimes g},{\cal H}_{\cal N} )\;\;. 

3460: $$

3461: \end{lemma}

3462: 

3463: {\em Proof:} Consider the  elements

3464: $P_j:=S_j\circ D_j^{-1}\circ S_j^{-1}$ and 

3465: $Q_j:=S_{j+1}\circ D_j^{-1}\circ S_{j+1}^{-1}$  of $\Gamma_{g,1}$. 

3466: From (\ref{eq-D-hom}) and (\ref{eq-S-hom}) we compute the homological

3467: action as $[R_j]=\kappa(I_g+E_{j+1, j})$ and $[Q_j]=\kappa(I_g+E_{j, j+1})$,

3468: with conventions again as in \cite{GooWal98}. 

3469: The matrices $I_g+E_{j+1, j}$ and $I_g+E_{j, j+1}$ generate $\SL(g,\Z)$, and hence 

3470: $[P_j]$ and $[Q_j]$ generate $\kappa(\SL(g,\Z))\subset \Sp(2g,\Z)\,$. 

3471: The actions of ${\cal V}({\bf I}_{P_j})$ and 

3472: ${\cal V}({\bf I}_{Q_j})$ on ${\cal N}_0^{\otimes g}$ are given by 

3473: placing the maps $\Pp:=(\Ss\otimes 1)\D^{-1}(\Ss^{-1}\otimes 1)$ and

3474: $\Q:=(1 \otimes\Ss )\D^{-1}(1 \otimes\Ss^{-1})$ in the $j$-th and $j+1$-st tensor 

3475: positions. In order to show that the actions of $P_j$ and $Q_j$ on ${\cal N}_0^{\otimes g}$

3476: yield Hopf algebra automorphisms it thus suffices to prove this for the maps 

3477: $\Pp$ and $\Q$ in the case $g=2$. From the tangle presentations  we find 

3478: identities ${\bf I}_{Q_1}= ({\bf M}\otimes 1)\circ (1\otimes {\bf \Delta})$

3479: and  ${\bf I}_{P_1}= (1\otimes {\bf M})\circ({\bf \Delta}\otimes 1)$. 

3480: It follows that $\Pp(x\otimes y)=\Delta_0(x)(1\otimes y)$ and 

3481: $\Q(x\otimes y)=(x\otimes 1)\Delta_0(y)$. The fact that these are Hopf automorphisms

3482: on ${\cal N}_0\otimes {\cal N}_0$ can be verified by direct computations. For the 

3483: multiplication this amounts to verification of equations such as 

3484: $\Delta(w)1\otimes v=-1\otimes v\Delta(w), \forall v,w\in\E\,$, and for the 

3485: comultiplication we use the

3486: fact that ${\cal N}_0$ is self dual. 

3487: 

3488: From the above identities we have that  

3489: ${\cal V}({\bf I}_{Q_1})=(M_0\otimes 1)\circ (1\otimes  \Delta_0)$ so that 

3490: ${\cal V}({\bf I}_{Q_j})$ is given on a monomial by taking the coproduct of the element in

3491: $(j+1)$-st position, multiplying the first factor of that to the element in $j$-th position 

3492: and placing the second factor into $(j+1)$-st position. We readily infer for every $w\in\E$ that

3493: ${\cal V}({\bf I}_{Q_j})(\varrho(w\otimes e_k))\,=\,

3494: \varrho(w\otimes e_k+\delta_{j+1,k}w\otimes e_j)\,=\,\varrho(w\otimes (I_g+E_{j+1,j})e_k)$.

3495: The analogous relation holds for $[P_j]$ so that 

3496: $$

3497: {\cal V}^{Sp}(\kappa(A))(w\otimes x)\quad=\quad w\otimes (Ax)\;\;\;

3498: \forall A\in \SL(g,\Z). 

3499: $$

3500: This is precisely the claim made in Lemma~\ref{lm-slgz-invar}. 

3501: \ep

3502: 

3503: 

3504: The structure ${\cal H}_{\cal N}$ is mapped by the isomorphism $\xi_g$ from (\ref{eq-defxi}) 

3505: to a $\Z/2$-graded Hopf algebra structure ${\cal H}_{\Lambda}$ on $H^*(J(\Sigma_g))$. 

3506: A-priori the isomorphism $\xi_g$ and thus also  ${\cal H}_{\Lambda}$ depend on the choice of

3507: a basis of $H_1(\Sigma_g)$. However, the $\SL(g,\Z)$-invariance determined in Lemma~\ref{lm-slgz-invar}

3508: translates to the $\SL(g,\Z)$-invariance of ${\cal H}_{\Lambda}$, where 

3509: $\kappa(\SL(g,\Z))\subset \Sp(2g,\Z)$ acts in the canonical way on $H^*(J(\Sigma_g))$. 

3510: Hence, ${\cal H}_{\Lambda}$ only depends on the oriented subspaces 

3511: $\Lambda=\lz [a_1],\ldots,[a_g]\rz\subset H_1(\Sigma_g,\Z)$

3512: and $\Lambda^*=\lz [b_1],\ldots,[b_g]\rz\subset H_1(\Sigma_g,\Z)$, 

3513: but not the specific choice of basis within them. 

3514: The orientations can be given by volume forms $\omega_{\Lambda}:=[a_1]\wedge\ldots\wedge[a_g]$

3515: and $\omega_{\Lambda^*}:=[b_1]\wedge\ldots\wedge[b_g]$. 

3516: The primitive elements $\varrho(\th\otimes e_j)$ and $\varrho(\tb\otimes e_j)$

3517: of ${\cal N}_g^{\otimes g}$  are mapped by $\xi_g$ to

3518: \beq\lbl{eq-primit}

3519: \pm[a_j]\wedge\omega_{\Lambda^*}\in\ext {g+1}H_1(\Sigma_g)\qquad\mbox{and}\qquad

3520: \pm i^*_{z_j}(\omega_{\Lambda^*})\in\ext {g-1}H_1(\Sigma_g)

3521: \eeq

3522: respectively, where $[a_j]\in H_1(\Sigma_g)$ and $z_j\in H^1(\Sigma_g)$, with $z_j([b_j])=1$ and

3523: $z_j([x])=0$ on all other basis vectors. We also have $\xi_g(1)=\omega_{\Lambda^*}$ and 

3524: $\xi_g(\rho^{\otimes g})=\omega_{\Lambda}$. 

3525: 

3526: This completes the proof of Theorem~\ref{thm-structure}. \ep

3527: \medskip

3528: 

3529: In the remainder of this section we give a more explicit description of the 

3530: structure ${\cal H}_{\Lambda}$ on $H^*(J(\Sigma_g))$, and relate it to an involution,

3531: $\tau$,

3532:  on $H^*(J(\Sigma_g))$, which acts as identity on the $\Lambda$-factor and, modulo signs,

3533: as a Hodge star on the opposite $\Lambda^*$-factor. 

3534: 

3535: The product $\diamond$ on $(H^*(J(\Sigma_g)),{\cal H}_{\Lambda})$ 

3536: is given on a genus one block, $\ext * \lz[a],[b]\rz$, as follows:

3537: \beq\lbl{eq-diamond}

3538: \begin{array}{c}

3539: \mbox{Table for}\\

3540: \\

3541: \quad u\,\diamond\, t\;:=\;\phi(\phi^{-1}(u)\phi^{-1}(t))\\

3542: \end{array}

3543: \quad\quad

3544: \begin{tabular}{ | c || c|c|c|c|}

3545: \hline

3546: ${}_u \,\backslash \,{}^t$\ \  

3547: &\ \  1\ \  &\ \  $[a]$\ \   &\ \  $[b]$\ \  &\ \  $[a]\wedge [b]$ \ \ \\

3548: \hline\hline

3549: 1 & 0 &  0 & 1 & $[a]$ \\

3550: \hline 

3551: $[a]$ & 0 & 0 & $a$ & 0 \\

3552: \hline 

3553: $[b]$ & 1 & $[a]$ & $[b]$ & $[a]\wedge [b]$ \\

3554: \hline 

3555: $[a]\wedge [b]$ & $-[a]$ & 0 &  $[a]\wedge [b]$ & 0\\

3556: \hline 

3557: \end{tabular}

3558: \qquad\qquad

3559: \eeq

3560: It extends to $\ext * H_1(\Sigma_g)$ via the formula 

3561: \beq\lbl{eq-totprod}

3562: (u_1\wedge\ldots\wedge u_g)\diamond (t_1\wedge\ldots\wedge t_g)

3563: \quad=\quad (-1)^{\sum_{i<j}d_il_j}\,\,(u_1\diamond t_1)

3564: \wedge \ldots\wedge(u_g\diamond t_g)\;,\quad

3565: \eeq

3566: where $u_i, t_i\in\ext *\lz [a_i],[b_i]\rz$, $d_i=1-deg(u_i)$ and  $l_j=1-deg(t_j)$. 

3567: In particular, we have $\,u\diamond t=(-1)^{d l}\,t\diamond u\,$, 

3568: with $d=\sum_id_i=g-deg(u)$ and $l=\sum_il_i=g-deg(t)$, which reflects the

3569: $\Z/2$-commutativity of $H^*(J(\Sigma_g))$. 

3570:  

3571: 

3572: The product structure and another proof of Lemma~\ref{lm-slgz-invar} can be also

3573: found from an involution, $\tau$, defined as follows:

3574:  

3575: Every cohomology

3576: class $x\in H^*(J(\Sigma_g))$ is uniquely written as $x=\alpha\wedge\beta$, where

3577: $\alpha\in\ext *\Lambda$ and $\beta\in \ext *\Lambda^*$. For $x$ in this form the map $\tau$

3578: is uniquely determined by the  relations  

3579: \beq\lbl{eq-deftau}

3580: \tau(\alpha\wedge\beta)=\alpha\wedge\tau(\beta)\quad\mbox{and}\quad

3581: \tau(b_1^{\epsilon_1}\wedge\ldots\wedge b_g^{\epsilon_g})=

3582: b_1^{1-\epsilon_1}\wedge\ldots\wedge b_g^{1-\epsilon_g}\;.

3583: \eeq

3584: From the formulae in (\ref{eq-diamond}) and (\ref{eq-totprod}) we find that $\tau^2=1\,$, and

3585: \beq\lbl{eq-tauprop}

3586: \tau(u\diamond t)\;\;=\;\;\tau(t)\wedge\tau(u)\;,\quad\quad\;

3587: \eeq

3588: and that $\tau$ maps $\ext *\Lambda$ as well as  $\ext *\Lambda^*$ to itself. It is clear from  

3589: (\ref{eq-deftau}) and (\ref{eq-tauprop}) that $\SL(g,\Z)$-variance

3590: of $\diamond$ on $H^*(J(\Sigma_g))$ is equivalent to $\SL(g,\Z)$-variance

3591: of $\diamond$ on $\ext * \Lambda^*$. Now,  for any $A\in \SL(\Lambda^*)$ the following

3592: identity holds:

3593: \beq\lbl{eq-tauAid}

3594: \tau\circ(\ext * A)\circ \tau\;\;=\;\;\ext* \iota(A)\;,

3595: \eeq

3596: where $\iota$ is the involution on $\SL(\Lambda^*)$ defined by 

3597: $$

3598: \iota(A)\;:=\;D\circ(A^{-1})^T\circ D\,,\qquad\mbox{with}\;\; D[b_j]=(-1)^j[b_j]\;.

3599: $$

3600: This can be proven either by considering again generators of $\SL(\Lambda^*)$,

3601: or by applying the generalized Leibniz formula for the expansion of the determinant

3602: of a $g\times g$-matrix into products of determinants of $k\times k$ and 

3603: $(g-k)\times(g-k)$-submatrices. See also Lemma~5.2 in \cite{FroNic94}. (\ref{eq-tauprop})

3604: together with  (\ref{eq-tauAid}) implies now that $\diamond$ depends only on the

3605: decomposition $H_1(\Sigma_g,\Z)=\Lambda\oplus\Lambda^*$. 

3606: 

3607: In summary, we have the following isomorphism of $\Z/2$-graded Hopf algebras:

3608: $$

3609: \tau':=\ext *D\circ\tau\,:\,(H^*(J(\Sigma_g)),{\cal H}_{\Lambda})\;\isto\;

3610: (H^*(J(\Sigma_g)),{\cal H}_{ext})\;\;,

3611: $$

3612: The Howe pair 

3613: $\SL(2,\R)\times \SL(g,\R)\subset GL(H_1(\Sigma_g))=

3614: {\rm Aut}(H^*(J(\Sigma_g)),{\cal H}_{ext})$, with 

3615: $H_1(\Sigma_g)=\E\otimes \Lambda$, is conjugated by $\tau'$ to the pair 

3616: $\SL(2,\R)_{Lefsch.}\times \kappa(\SL(g,\R))\subset 

3617: {\rm Aut}(H^*(J(\Sigma_g)),{\cal H}_{\Lambda})$.  

3618: \medskip

3619: 

3620: \newpage

3621: 

3622: 

3623: \head{13. More Examples of Homological TQFT's and Open Questions}\lbl{S13}

3624: 

3625: 

3626: %%%%%%%%%%%%%%%%%   BEGIN CUT   %%%%%%%%%%%%%%%%%%%%%%%%%%%

3627: 

3628: \emptystuff{

3629: \paragraph{A. Homology TQFT's over $\Z/r$ and cut numbers.}\ 

3630: 

3631: 

3632: Although the TQFT's  of Reshetikhin and Turaev are semisimple and 

3633: non trivial on the Torelli groups they contain homological TQFT's in an 

3634: indirect manner. Specifically, if we consider the TQFT for $U_q({\mathfrak s}{\mathfrak l}_2)$

3635: for $q$ a primitive $r$-th root of unity and $r$ is an odd prime 

3636: Gilmer \cite{Gil01} shows that it can be defined essentially as a 

3637: theory ${\cal W}^*_r$

3638: over the ring over cyclotomic integers $\Z[q]$. This generalizes

3639: the integrality results in \cite{MasRob97} and \cite{Mur94} for the 

3640: invariants of closed manifolds. 

3641: 

3642: Of particular interest are expansions in $(q-1)$ which on the level of

3643: invariants of closed manifolds lead to the Ohtsuki invariants in $\Z/r$ 

3644: \cite{Oh96} which in 0-th order coincides with the invariant from (\ref{eq-defeta})

3645: and in next order is identical to the Casson invariant \cite{Mur94}. 

3646: 

3647: A candidate for a useful homological TQFT is the lowest order of the TQFT

3648: over the cyclotomic integers. It is given by the extending the

3649:  trace function $\Z[q]\to\Z/r$ to a transformation  

3650: ${\cal W}^*_r\,\to\,{\cal W}_r$, where ${\cal W}_r$ is the respective TQFT

3651: defined over the finite field $\Z/r$. 

3652: 

3653: 

3654: In \cite{Kerr=5} we consider the first non-trivial prime $r=5$ and

3655: find an explicit basis for ${\cal W}^*_5$ and hence a description of

3656: ${\cal W}_5$. We find that the Torelli group is not entirely in the

3657: kernel of ${\cal W}_5$ but factors through the Johnson homomorphism.

3658: It does, however, contain a sub-TQFT $ {\cal U}_5\subset {\cal W}_5$

3659: which is homological, meaning does not see the Torelli group, 

3660: such that also the quotient TQFT 

3661: ${\cal Q}_5={\cal W}_5/ {\cal U}_5$ is homological. 

3662: 

3663: Explicit computations strongly suggest that these homological TQFT's are  

3664: also of the general form (\ref{eq-defpolyTQFT}). More precisely,

3665: we define the following (linear)

3666: polynomials in $\Z[x_0,x_1,\ldots]$. 

3667: $$

3668: Q\,=\,x_0-x_8+x_{10}-x_{18}+x_{20}-x_{28}+\ldots

3669: $$

3670: $$\mbox{and} 

3671: \qquad\qquad 

3672: U\,=\,x_3-x_5+x_{13}-x_{15}+x_{23}-x_{25}+\ldots \;\;. 

3673: \qquad\qquad \qquad

3674: $$

3675: Explicit verification for low genera and comparison of dimensions leads

3676: us to the following. 

3677: \begin{conjecture}[see\cite{Kerr=5}]\label{conj-5}

3678:  We have isomorphisms of TQFT's defined 

3679: over $\Z/5$:

3680: $$

3681: {\cal Q}_5\,\,\cong\,\,{\cal V}^{(Q)}

3682: \qquad\qquad \mbox{and}\qquad \qquad 

3683: {\cal U}_5\,\,\cong\,\,{\cal V}^{(U)}

3684: $$

3685: \end{conjecture}

3686: Note that the polynomials $Q$ and $U$ also have negative integers so that

3687: we need to make sense of subtracting vector spaces or TQFT's. To this end note that

3688: the $\Sp(2g,\Z)$-representations $W_{g,j}$ are irreducible over $\Z$ but

3689: become reducible if we take them, for a given basis, over $\Z/5$. For instance the 

3690: $\Z/5$-reduction of $W_{6,3}$ contains a subrepresentation isomorphic to

3691: the $\Z/5$-reduction of  $W_{6,5}$. This explains the meaning of the 

3692: difference $x_3-x_5$ in the expression for $U$.

3693: 

3694: A fascinating topological application is the determination of so called

3695: {\em cut numbers}, which is investigated in joint work with Gilmer \cite{GilKer}. 

3696: Let us denote by $cut(M)$ the maximal number rank $n$ of a (non-abelian) free group

3697: $F_n=\Z*\ldots*\Z$ such that there is an epimorphism $\phi:\pi_1(M)\to F_n$. 

3698: This is also the maximal number of surfaces that can be removed from $M$ without

3699: disconnecting the manifold. For a given epimorphism $\varphi:H_1(M)\to \Z$ we

3700: also define the relative cut number $cut(M,\varphi)$ as the maximal $n$ such that

3701: there is an epimorhpism $\phi:\pi_1(M)\to F_n$ which factors through $\varphi$,

3702: meaning there is a map $\tau:F_m\to\Z$ such that $\varphi=\tau\circ\phi$. This 

3703: counts non separating surfaces with the constraint that one represents $\varphi$, \cite{GilKer}. 

3704: 

3705: Clearly we  have $cut(M)\geq cut(M,\varphi)\geq 1$ if defined. Aside from these 

3706: constraints the

3707: absolute and relative cut number are independent.  For example

3708: let $M=S^1\times \Sigma_g$ with canonical projection $\pi:M\to S^1$. Then 

3709: $cut(M,\pi_*)=1$ but $cut(M)\geq g$. 

3710: 

3711: 

3712: As we remarked in the beginning of Section~11 an additional non-separating

3713: surface in the cut cobordism $C$ used to define the Alexander polynomial

3714: implies ${\cal V}^{FN}(C)=0$ by non-semisimplicity. Thus for a 3-manifold $M$ 

3715: with epimorphism $\varphi:H_1(M)\to\Z$  as before we have the implication: 

3716: \beq\label{eq-cutrel}

3717: \mbox{If}\qquad\quad cut(M,\varphi) >1\qquad\quad{\rm then}

3718: \qquad\quad \Delta_{M,\varphi}\,=\,0.

3719: \eeq

3720: In \cite{GilKer} we manage to obtain  a criterium on the bare cut number

3721: {\em independent} of  a choice of $\varphi$:  

3722: \beq\label{eq-cut}

3723: \mbox{If}\qquad\quad cut(M) >1\qquad\quad{\rm then}

3724: \qquad\quad trace({\cal W}_5(C))\,\equiv\,0\;{\rm mod}\,5.

3725: \eeq

3726: Note that the expression on the right only depends on the homological

3727: functors $\cal Q$ and $\cal U$. It turns out that  under the assumption of 

3728: Conjecture~\ref{conj-5} the respective traces are easily computed form the Alexander polynomial.

3729: In fact, under this assumption,

3730:  the trace expression

3731: in (\ref{eq-cut}), which is just the sum of the traces for $\cal Q$ and $\cal U$,  

3732:   comes out to be equal to 

3733: the unique coefficient $T_{M,\varphi}$

3734: of the Alexander polynomial when written as follows. 

3735: $$

3736: \Delta_{M,\varphi}(q)\;=\;T_{M,\varphi}\;+\;B(q+q^{-1})\;\;\;\in\;\;\;\Z[q]\qquad\mbox{with}\quad 

3737: T_{M,\varphi}, B\in \Z\;\;.

3738: $$

3739: The contrapositive of (\ref{eq-cut}) under the assumption of the conjecture thus becomes

3740: \beq\label{eq-contracut}

3741: \mbox{If}\qquad\quad T_{M,\varphi}\neq 0 \,\, {\rm mod}\,5\qquad\quad{\rm then}

3742: \qquad\quad cut(M)=1\;.\qquad\mbox{(for any choice of $\varphi$)}

3743: \eeq

3744: See \cite{GilKer} for more details and applications. 

3745: }

3746: 

3747: %%%%%%%%%%%%%%%%%%%%%%%%%%    END   CUT    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3748: 

3749: 

3750: \paragraph{A. Relations to Gauge Theories and the TQFT-Ring \Qq Generated by  ${\cal V}$.}\ 

3751: 

3752: We begin by collecting the ingredients that imply Theorem~\ref{thm-relations}.

3753: The first identity (\ref{eq-AlexChar}) has already been computed in 

3754: (\ref{eq-leftrace3}). 

3755: 

3756: The invariants $I^{DC}$ and $I^{SW}_d$ are obtained by Donaldson in \cite{Don99} 

3757: from TQFT's  ${\cal V}^{DC}$ and ${\cal V}^{SW}_d$ respectively. For both TQFT's the 

3758: vector spaces associated to a surface $\Sigma$ are the homologies of  natural 

3759: moduli spaces. In the case of ${\cal V}^{DC}$ this is the  moduli space ${\cal M}(\Sigma)$ 

3760: of flat connections on a non-trivial $SO(3)$ bundle. For  ${\cal V}^{SW}_d$

3761: the moduli space of solutions to certain vortex equations is considered, which turn

3762: identified with the  symmetric products of the surface. The action of the mapping

3763: class group on the resulting homologies also factors through the symplectic groups (with 

3764: the familiar ${\mathbb F}_2$-ambiguity). Donaldson thus derives the following 

3765: isomorphisms between $\Sp(2g,\Z)$-modules.

3766: \beq\label{eq-Donaldson}

3767: \begin{array}{rcccl}

3768: \displaystyle {\cal V}^{DC}(\Sigma_g)&\;\;=\;\;&H_*\bigl({\cal M}(\Sigma_g)\bigr)

3769: &\;\;\cong\;\;&\displaystyle 

3770: \bigoplus_{j=0}^g\,{\mathbb Q}^{j^2}\otimes \ext {g-j}H_1(\Sigma_g)\;\;\mbox{and}\\

3771: %&&&\\

3772: \displaystyle {\cal V}^{SW}_d(\Sigma_g)&\;\;=\;\;&H_*\bigl(Sym^k(\Sigma_g)\bigr)

3773: &\;\;\cong\;\;&\displaystyle 

3774: \bigoplus_{j=1}^{g-d}\,{\mathbb Q}^{j}\otimes \ext {g-d-j}H_1(\Sigma_g)\;,\\

3775: \end{array}

3776: \eeq

3777: where $k=g-1-d$ is the degree the holomorphic line bundle of which the vortex solutions

3778: are sections. The $\Sp(2g,\Z)$-representations can be further identified with the 

3779: irreducible parts, which takes in our notation the form

3780: \beq\label{eq-extexpand}

3781: \ext {g-j}H_1(\Sigma_g)\;\;=\;\;{\cal V}^{(j+1)}(\Sigma)\;\oplus\;

3782: {\cal V}^{(j+3)}(\Sigma)\;\oplus\;{\cal V}^{(j+5)}(\Sigma)\;\oplus\;\ldots

3783: \eeq

3784: Inserting (\ref{eq-extexpand}) into the isomorphisms in (\ref{eq-Donaldson}) we

3785: find  that the ${\cal V}^{DC}(\Sigma_g)$ and ${\cal V}^{SW}(\Sigma_g)$ are direct

3786: sums of the ${\cal V}^{(j)}(\Sigma_g)$ with $g$-independent 

3787: multiplicities given by precisely the

3788: non-negative coefficients in (\ref{eq-DCChar}) and (\ref{eq-SWChar}). In Chapter~5 of

3789: \cite{Don99} 

3790: Donaldson exploits this fact to show that the decomposition thus extends to the 

3791: entire TQFT's, meaning that cobordisms act trivially on the mulitiplicity spaces and

3792: have block-wise actions on the  ${\cal V}^{(j)}$ components equivalent to those in 

3793: the ${\cal V}^{FN}$ case. Summarily, we have the following isomorphisms 

3794: {\em of TQFT's}.

3795: \beq\label{eq-TQFTidentities}

3796: \begin{array}{rcl}

3797: \displaystyle {\cal V}^{DC}

3798: &\;\;\cong\;\;&\displaystyle 

3799: \bigoplus_{j\geq 2}\,{\mathbb Q}^{{j+1}\choose 3}\otimes {\cal V}^{(j)}\;\\

3800: %&&&\\

3801: \displaystyle {\cal V}^{SW}_d

3802: &\;\;\cong\;\;&\displaystyle 

3803: \bigoplus_{j\geq d+2}\,{\mathbb Q}^{\left[\!\!{\left[

3804: \Bigl({\frac{j-d}2}\Bigr)^2\right]}\!\!\right]}\otimes {\cal V}^{(j)}\;.\\

3805: \end{array}

3806: \eeq

3807: Identities (\ref{eq-DCChar}) and (\ref{eq-SWChar}) are now immediate. For the

3808: last equation  (\ref{eq-LesChar}) in Theorem~\ref{thm-relations}

3809: we refer to \cite{KerKyoto}.

3810: 

3811: \ep

3812: 

3813: In an effort  to find new knot invariants Frohman and Nicas generalized their

3814: approach in  \cite{FroNic94} to higher rank Lie algebras. They construct a TQFT

3815: ${\cal V}^{PSU(n)}_k$, 

3816: whose vector spaces are

3817: given as intersection homology groups of certain restricted moduli spaces 

3818: of $PSU(n)$-representations and derive from these by similar trace formulae 

3819:  invariants $\lambda_{n,k}$ depending  on the 

3820: rank $n$ and weight $k$. In \cite{Fro93} Frohman finds a recursive 

3821: procedure to compute the invariants $\lambda_{n,k}$ and shows that they 

3822: are determined by the polynomial expressions in the

3823: coefficients of the Alexander polynomial. Consequently, they are also

3824: polynomial in the  Alexander Characters so that we can write

3825: \beq\label{eq-laRnk}

3826: \lambda_{n,k}\;=\; R_{n,k}(\Delta^{(1)}, \Delta^{(2)}, \ldots)\;,

3827: \eeq

3828: with $R_{n,k}\in{\mathbb Z}[x_1,x_2,\ldots]$. A general, closed

3829: formula and some integrality issues for the $R_{n,k}$ are still unresolved 

3830: though, see also \cite{BodNic00}. This relation in (\ref{eq-laRnk}) is more 

3831: general than those expressed in  Theorem~\ref{thm-relations} as it is no longer

3832: linear.

3833: 

3834: More precisely, define the space of invariants 

3835: $\Qq^{[0]}\;=\;\left\{{n_1\Delta^{(1)}\,+\,n_2\Delta^{(2)}\,+\,\ldots\,|\,

3836: n_i\in{\mathbb Z}^{+,0}}\right\}$. Then is clear that invariant that descends

3837: from a TQFT that is homological must be in $\Qq^{[0]}$ where the $n_i\geq 0$

3838: are the multiplicities of the irreducible summands. Thus $I^{DC},\,I^{SW}_d\,

3839: \in\,\Qq^{[0]}$, but we also have ${\lambda}_L\not\in\Qq^{[0]}$ since some of

3840: the coefficients are negative. ${\lambda}_L$ is, nevertheless, related to the

3841: quantum TQFT's, but the derivations use slightly more subtle $p$-modular

3842: interpretations, see \cite{KerKyoto}. 

3843: 

3844: \bigskip

3845: 

3846: Similar, to sums we can derive the 

3847: invariant given by the product 

3848: of two Alexander Characters, say $\Delta^{(i)}\cdot \Delta^{(j)}$,

3849: from the tensor product of the corresponding TQFT's, namely

3850: ${\cal V}^{(i)}\otimes {\cal V}^{(j)}$. Thus is the coefficients of all

3851: the $R_{n,k}$ were non-negative integers we could easily produce a 

3852: homological TQFT by taking corresponding direct sums and tensor products

3853: of the ${\cal V}^{(i)}$ in order to reproduce $\lambda_{n,k}$. This 

3854: invariant, indeed, descends from the TQFT ${\cal V}^{PSU(n)}_k$, however, 

3855: the coefficients of the $R_{n,k}$. The point to observe here is that, for

3856: example, ${\cal V}^{(i)}\otimes {\cal V}^{(j)}$ is generally not an 

3857: irreducible TQFT and can be decomposed. 

3858: 

3859: Denote by ${\cal V}^{(\vec \lambda)}$ the irreducible TQFT's obtained as summands of

3860: quotients of multiple tensor products of the ${\cal V}^{(i)}$. The superscript label,

3861: $\vec\lambda\in\vec\Lambda$, may be roughly thought of as a semi-infinite branching

3862: path for $\Sp(2)\subset \Sp(4)\subset \Sp(6)\subset\ldots\,$. The space of TQFT's

3863: \beq

3864: \Qq^{[+]}\;=\;

3865: \Biggl\{{\,\bigoplus_{{\vec\lambda}\in{\vec\Lambda}} \,{\mathbb Q}^{n_{\vec\lambda}}

3866: \otimes {\cal V}^{(\vec\lambda)}\,\,\Bigg |\,\,\,n_{\vec\lambda}\in {\mathbb N}\cup

3867: \{0\}}\Biggr\}\; 

3868: \eeq

3869: thus has a natural ring structure with operations $\oplus$ and $\otimes$ and can be 

3870: thought of as a type of  Grothendieck $K_0$-ring for a homological subquotient of $\Cob$. 

3871: We denote the corresponding set of higher Alexander Characters abusively in the same

3872: way, since it possesses the same ring structure under usual addition and multiplication.

3873: Clearly, $\Qq^{[0]}\subset\Qq^{[+]}$. The following conjecture together with an understanding

3874: of the ring structure of $\Qq^{[+]}$ should shed light

3875: onto the general structure of the polynomials $R_{n,k}$. 

3876: \begin{conjecture}

3877: $$

3878: {\cal  V}^{PSU(n)}_k\;\;\;\in\;\;\;\Qq^{[+]}

3879: $$

3880: \end{conjecture}

3881: 

3882: \emptystuff{

3883: 

3884: 

3885: 

3886: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%  BEGIN CUT   %%%%%%%%%%%%%%%%%%%%%%%%%%

3887: 

3888: 

3889: 

3890: 

3891: 

3892: \bigskip

3893: 

3894: \bigskip

3895: 

3896: 

3897: Using the coefficients of

3898: the Conway polynomial $\nabla (z)=\sum_j c_j z^{2j}$ with 

3899: $z=t^{\frac 12}-t^{-\frac 12}$ this yields polynomials

3900: $\lambda_{n,k}=q_{n,k}(c_1,c_2,\ldots)$, which appear to

3901: have  non-negative

3902: integer coefficients, that is $q_{n,k}\in{\cal P}^+$ as defined in

3903: Section~10. See also \cite{BodNic00} for more explicit formulae.

3904: Changing the basis of the polynomial ring from $z^{2j}$ to

3905: the $[j]_{-t}$ we are similarly able to express the higher 

3906: rank invariants in terms of the Alexander Characters $\Delta^{(j)}$ defined

3907: in (\ref{eq-leftrace3}). We can thus write

3908: $$

3909: \lambda_{n,k}\;=\; R_{n,k}(\Delta^{(1)}, \Delta^{(2)}, \ldots)

3910: $$

3911: for some polynomial $R_{n,k}$ with integral coefficients.

3912: Note further that if $C$ is the cobordism on a Seifert surface of a knot 

3913: and $P\in{\cal P}^+$ then 

3914: $trace({\cal V}^{(P)}(C))=P(\Delta^{(1)}, \Delta^{(2)}, \ldots)$

3915: 

3916: The natural

3917: question tied to these observations is whether the TQFT's constructed 

3918: in  \cite{FroNic94} for general gauge groups are related or  equivalent

3919: to a TQFT of the polynomial form ${\cal V}^{(R_{n,k})}$ as defined in 

3920: (\ref{eq-defpolyTQFT}). If the coefficients of the $R_{n,k}$ are not

3921: all non-negative we may have to consider two theories 

3922: ${\cal V}^{(R_{n,k}^{\pm})}$ with $R_{n,k}=R_{n,k}^+-R_{n,k}^-$ and

3923: $R_{n,k}^{\pm}\in{\cal P}^+$ and make sense of their difference. 

3924: 

3925: 

3926:  

3927: 

3928: 

3929: 

3930: In \cite{Don99} Donaldson describes a slightly different TQFT, ${\cal V}^{DF}$, 

3931: modeled on moduli spaces ${\cal M}_g$ of connections on a non-trivial $SO(3)$ bundle. 

3932: This TQFT leads up to a Casson type invariant for homology circles $Y$, which is

3933: determined by the coefficients of the   Alexander polynomial $\Delta_Y$. 

3934: The vector spaces are given as 

3935: \beq\label{eq-Donaldson}

3936: {\cal V}^{DF}(\Sigma_g)\;\;=\;\;H_*\bigl({\cal M}_g\bigr)

3937: \;\;\cong\;\;\bigoplus_{j=0}^g\,\R^{j^2}\otimes \ext {g-j}H_1(\Sigma_g)\;.

3938: \eeq

3939: The morphisms ${\cal V}^{DF}(M)$ are similarly constructed via the intersection

3940: theory of representation varieties, using also  a dimension

3941: reduction of the Floer-cohomology on $\hat M\times S^1$.  

3942: 

3943: Now from Corollary~5.1.9 in \cite{GooWal98} we see that

3944:  $\ext {g-j}H_1(\Sigma_g)\,\cong\,W_{g,j+1}\oplus W_{g,j+3}\oplus W_{g,j+5}\ldots$

3945: as $\Sp(2g,\Z)$ modules. Inserting this decomposition into (\ref{eq-Donaldson}) we

3946: obtain the multiplicities stated in the following conjecture.

3947: \begin{conjecture} Let $\displaystyle D=\sum_{k\geq 1}

3948: {\scriptstyle { k+1 \choose  3}} x_k\;\in\;{\cal P}^+$. Then 

3949: $$

3950: {\cal V}^{DF}\;\cong\;{\cal V}^{(D)}\;.

3951: $$

3952: \end{conjecture}

3953: Note that on the level of vector spaces and invariants we do in fact have equality. 

3954: 

3955: 

3956: The theories in \cite{FroNic94} and \cite{Don99} are all inherently

3957: $\Z/2$-projective, and have the vanishing properties from Lemma~\ref{lm-vanish}.

3958: This indicates that they also belong into the class  of

3959: half-projective or non-semisimple TQFT's. 

3960: 

3961: 

3962: Another conjecture that is independent of a particular gauge theory may be stated

3963: as follows.

3964: \begin{conjecture} Suppose ${\cal V}$ is a non-semisimple TQFT in which the kernel

3965: of the mapping class group representations are precisely the Torelli groups. The

3966:  ${\cal V}$ is isomorphic to a sub-TQFT of some ${\cal V}^{(P)}$ for some 

3967: $P\in{\cal P}^+$.

3968: \end{conjecture}

3969: To say that the Torelli group is {\em precisely} the kernel implies that we

3970: have faithful $\Sp(2g,\Z)$-representations so that by Margulis' rigidity

3971: these lift to algebraic  $\Sp(2g,\R)$-representations. Classifying homological 

3972: TQFT's as described will thus involve exercises in branching rules as given

3973: for example in Section~8.3.4 of  \cite{GooWal98}. 

3974: 

3975: 

3976: Another approach to describing TQFT's behind the higher rank Frohman Nicas

3977: $PSU(n)$-theories or the Donaldson $SO(3)$-construction is to try to extract

3978: a categorical Hopf algebra ${\cal A}_{\cal V}$ for the given TQFT by evaluation

3979: of the cobordisms in Figure~\ref{fig-MC-tgl}. The problem with this approach,

3980: however, is that in the non-abelian TQFT's we do not seem to have a nicely

3981: defined tensor structure arising from gluing two one holed surfaces over a 

3982: pair of pants. Specifically we need an isomorphism 

3983: ${\cal V} (\Sigma_{2,1})=

3984: {\cal V} (\Sigma_{1,1})\otimes{\cal V} (\Sigma_{1,1})$ which is generally

3985: not true because of gauge constraints over the pair of pants. 

3986: 

3987: It is also not clear whether higher rank theories such as those in \cite{FroNic94}

3988: exhibit symmetries similar to the $\SL(2,\R)$-equivariance

3989:  that yields a type of Lefschetz

3990: decomposition. Particularly,  the non-abelian moduli spaces have

3991: no canonical K\"ahler structure. They do, however, admit useful Poisson structures

3992: \cite{FocRos97}.

3993: }

3994: 

3995: 

3996: %%%%%%%%%%%%%%%%%%%%%%   END CUT   %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 

3997: 

3998: 

3999: \paragraph{B. Homology TQFT's from the  Reshetikhin-Turaev Theory:}\ 

4000: 

4001: 

4002: Recall that the TQFT ${\cal V}^{(j)}$ is in fact a functor to 

4003: the category of free  $\Z$-modules rather than just the category

4004: of vector spaces over ${\mathbb Q}$. Now, for any prime $p\geq 3$, 

4005: by taking all lattices modulo $p$ 

4006: this in turn maps to the category of vector spaces over the finite

4007: field ${\mathbb F}_p=\Z/p\Z$. The resulting TQFT  ${\cal V}^{(j)}_p$

4008: over ${\mathbb F}_p$

4009: is now no longer irreducible, but it has a {\em unique}

4010: irreducible subquotient, which we denote by $\dov {\cal V}^{(j)}_p$,

4011: see \cite{KerRes}.

4012: 

4013: Another way of generating TQFT's over ${\mathbb F}_p$ is to consider

4014: the Reshetikhin Turaev Theory for quantum-$SO(3)$ at a primitive

4015: $p$-th root of unity $\zeta_p$. As shown in \cite{Gil01} this can

4016: be regarded as a TQFT over  the ring of cyclotomic integers

4017: $\Z[\zeta_p]$.  The TQFT obtained from the ring reduction

4018: $\Z[\zeta_p]\onto{12}{\mathbb F}_p\,:\,\zeta_p\mapsto 1$ is denoted

4019: ${\cal V}^{RT}_p$. The example $p=5$, which is in some sense a

4020: fundamental case, is analyzed in \cite{Kerr=5}. We obtain an exact,

4021: but non-split sequence of TQFT's as follows

4022: \beq\label{eq-ext}

4023: 0\;\;\to\;\;\dov {\cal V}^{(4)}_5

4024: \;\;\longrightarrow\;\;

4025: {\cal V}^{RT}_5

4026: \;\;\longrightarrow\;\;

4027: \dov {\cal V}^{(1)}_5

4028: \;\;\to\;\;0

4029: \eeq

4030: As an extension of the mapping class group $\Gamma_g$ (\ref{eq-ext})

4031: involves a Johnson-Morita subquotient of $\Gamma_g$. The precise modular

4032: structure of the ${\cal V}^{(j)}_p$ and $\dov {\cal V}^{(j)}_p$

4033: TQFT's is unraveled in \cite{KerRes}. There we find resolutions

4034: of the $\dov {\cal V}^{(j)}_p$ in terms of the ${\cal V}^{(j)}_p$,

4035: which lead to important identities between the $p$-modular versions

4036: of the invariants from Theorem~\ref{thm-relations} and the 

4037: Reshetikhin Turaev Invariants.

4038: 

4039: It is easy to see that the irreducible factors of ${\cal V}^{RT}_p$ for

4040: $p\geq 7$ can no longer be reductions of the ${\cal V}^{(j)}$. There is,

4041: however, evidence that suggests that the irreducible factors are 

4042: reductions of summands in the symmetric powers of the fundamental ones.

4043: That is, TQFT's of the form

4044: \beq

4045: {\cal V}^{\vec\lambda}

4046: \;\subseteq\;S^{\frac {p-3}2}{\cal V}^{FN}\,\;\in\,\;\Qq^{[+]}\,.

4047: \eeq 

4048: This is closely related to the conjecture that the Lescop invariant

4049: for a closed 3-manifold $M$ with $\beta_1(M)\geq 1$ relates to the 

4050: Reshetikhin Turaev Invariant as follows.

4051: \beq

4052: {\cal V}^{RT}_{\zeta_p}(M)\;\;=\;\;C_p\cdot 

4053: \Bigl((\zeta_p-1)\lambda_L(M)\Bigr)^{\frac {p-3}2}\;\;+\;\;{\cal O}

4054: \bigl((\zeta_p-1)^{\frac {p-1}2}\bigr)\;\;.

4055: \eeq

4056: This has been verified for $p=5$ in \cite{KerKyoto}. 

4057: 

4058: 

4059: \paragraph{C. Relation of Reshetikhin-Turaev and Hennings Theory:}\ 

4060: 

4061: Given a quasitriangular Hopf algebra, ${\cal A}$, we have described in 

4062: Section~5 a procedure to construct a topological quantum field theory,

4063: ${\cal V}_{\cal A}^{H}$. In \cite{ResTur91} and \cite{Tur94} Reshetikhin and

4064: Turaev give another procedure to construct a TQFT, ${\cal V}^{RT}_{\cal S}$, 

4065: from a {\em semisimple} modular  category, ${\cal S}$. A more general

4066: construction in \cite{KerLub00} allows us to construct a TQFT, 

4067: ${\cal V}_{\cal C}^{KL}$, also for modular categories, ${\cal C}$,

4068: that are not semisimple, and we show in \cite{Ker96} that 

4069: ${\cal V}_{\cal A}^{H}= {\cal V}_{{\cal A}-mod}^{KL}$ and 

4070: ${\cal V}^{RT}_{\cal S}={\cal V}_{\cal S}^{KL}$ for semisimple

4071:  ${\cal S}$. For a non-semisimple, quasitriangular algebra, ${\cal A}$, 

4072: the semisimple category used in \cite{ResTur91} and \cite{Tur94}

4073: is given as the semisimple trace-quotient 

4074: ${\cal S}({\cal A})=\overline{{\cal A}-mod}$ of the representation

4075: category of $\cal A$. The relation between ${\cal V}_{\cal A}^{H}$ 

4076: and ${\cal V}^{RT}_{{\cal S}({\cal A})}$ is generally unknown. 

4077: We make the following conjecture in the case of quantum ${\mathfrak s}{\mathfrak l}_2$.

4078: 

4079: \begin{conjecture}

4080: Let ${\cal A}=U_q({\mathfrak s}{\mathfrak l}_2)^{red}$, with $q$ an odd $p$-th 

4081: root of unity, and  relations $E^p=F^p=0$ and $K^{2p}=1$ for the standard generators. Then there is monomorphic, natural transformation 

4082: \beq\label{eq-HRT}

4083: {\cal V}^{FN}\otimes 

4084: {\cal V}^{RT}_{{\cal S}({\cal A})}\;\;\into\;\;{\cal V}_{\cal A}^H\;.

4085: \eeq

4086: \end{conjecture}

4087: 

4088: In the genus one case we have shown in  \cite{Ker94} and \cite{Ker96}

4089: that the mapping class group representations and invariants of 

4090: lens spaces of both theories in ({\ref{eq-HRT}) are in fact equal. 

4091: The above inclusion of TQFT functors can also be phrased in the form

4092: ${\cal V}_{{\cal C}^{\#}}^{KL}\into{\cal V}^{KL}_{{\cal C}}$, 

4093: where ${\cal C}:=U_q({\mathfrak s}{\mathfrak l}_2)^{red}-mod$ and 

4094: ${\cal C}^{\#}:=({\cal N}-mod)\otimes \overline{\cal C}$. 

4095: The categories ${\cal C}$ and ${\cal C}^{\#}$ are in fact rather

4096: similar as linear abelian categories. From \cite{Ker98} it follows 

4097: that there an isomorphism of {\em abelian} categories

4098: \beq

4099: %\mbox{\bm${\cal H}$\ubm}

4100: \Hh\;\;:\;\;

4101: {\cal C}^{\#}\;\oplus\;2\cdot {\rm Vect}({\mathbb C}) \;\;\;

4102: \stackrel{\cong}{\longrightarrow}\;\;\;{\cal C}\;,

4103: \eeq

4104: where the two extra ${\rm Vect}({\mathbb C})$'s account for the two $p$-dimensional,

4105: irreducible Steinberg modules. This, however, is {\em not} a 

4106: monoidal functor. Instead we have a natural set of monomorphisms of

4107: the form $\Hh(X)\otimes\Hh(Y)\,\into\,\Hh(X\otimes Y)$. As a result

4108: the braidings, integrals, and coends, that enter in a crucial way

4109: the construction of the TQFT's  \cite{KerLub00}

4110: can no longer be na\"\i vely

4111: identified. Strategies of proof would include a basis 

4112: of ${\cal A}$ as worked out in \cite{Ker94} and the use of the special

4113: central, nilpotent element ${\sf Q}$ defined in \cite{Ker96}. 

4114: 

4115: %%%%%%%%%%%%%%%%%%%   BEGIN CUT %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4116: 

4117: \emptystuff{

4118: 

4119: 

4120: Specifically, we know the following: 

4121: \begin{theorem}[\cite{Ker89}]\ Let ${\cal A}=U_q({\mathfrak s}{\mathfrak l}_2)^{red}$ and 

4122: $\cal N$ as in Section~6. 

4123: \begin{enumerate}

4124: \item 

4125: For any generic Casimir value, $c\in ({\mathfrak z}({\cal A}))^*$, 

4126: the corresponding subcategory ${\cal C}_c\subset{\cal A}-mod$ 

4127: of representations is isomorphic to ${\cal N}-mod$. 

4128: \item The representations with non-generic Casimir values are 

4129: sums of the two irreducible Steinberg modules of dimension $r$

4130: and quantum dimension 0. 

4131: \item An indecomposable representation of ${\cal N}$ is either 

4132:  one of the two 4-dim

4133: projective representations in  ${\cal N}={\cal N}^+\oplus{\cal N}^-$,

4134: or an indecomposable representation of one of the two Kronecker quivers 

4135: $\bullet \raisebox{-.5ex}{\hbox{$\stackrel{\mbox{$\longrightarrow$}}{\longrightarrow}$}}\bullet$ and 

4136: $\bullet \raisebox{-.5ex}{\hbox{$\stackrel{\mbox{$\longleftarrow$}}{\longleftarrow}$}}\bullet$,

4137: where the $\bullet$'s stand for an eigenspaces of $K$.  

4138: \end{enumerate}

4139: \end{theorem}

4140: The generic Casimir values are in a two to one correspondence with the

4141: admissible irreducible representations, and we have 

4142: ${\cal C}=\bigoplus_c{\cal C}_c$ and  

4143: ${\cal C}^{\#}=\bigoplus_j {\cal N}-mod$, where $j$ runs over  

4144: irreducible representations. Thus we have a close correspondence between

4145: the modules in both categories. They differ, however, more strongly 

4146: as tensor categories. Strategies of proof would include a basis 

4147: of ${\cal A}$ as worked out in \cite{Ker94} and the use of the special

4148: central, nilpotent element ${\sf Q}$ defined in \cite{Ker96}. 

4149: 

4150: }

4151: 

4152: %%%%%%%%%%%%%%%%%%%%%%%%%     BEGIN   CUT    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4153: 

4154: \emptystuff{

4155: 

4156: \paragraph{C. Universality of ${\cal V}$ and Casson Type Gauge Theories.}\ 

4157: 

4158: In order to find new knot invariants Frohman and Nicas generalized their

4159: approach in  \cite{FroNic94} to higher rank Lie algebras. They construct a TQFT

4160: whose vector spaces are

4161: given as intersection homology groups of certain restricted moduli spaces 

4162: of $PSU(n)$-representations and derive from these by similar trace formulae 

4163:  invariants $\lambda_{n,k}$ depending  on the 

4164: rank $n$ and weight $k$. In \cite{Fro93} Frohman finds a recursive 

4165: procedure to compute the invariants $\lambda_{n,k}$ and shows that they 

4166: are determined by the polynomial expressions in the

4167: coefficients of the Alexander polynomial. Using the coefficients of

4168: the Conway polynomial $\nabla (z)=\sum_j c_j z^{2j}$ with 

4169: $z=t^{\frac 12}-t^{-\frac 12}$ this yields polynomials

4170: $\lambda_{n,k}=q_{n,k}(c_1,c_2,\ldots)$, which appear to

4171: have  non-negative

4172: integer coefficients, that is $q_{n,k}\in{\cal P}^+$ as defined in

4173: Section~10. See also \cite{BodNic00} for more explicit formulae.

4174: Changing the basis of the polynomial ring from $z^{2j}$ to

4175: the $[j]_{-t}$ we are similarly able to express the higher 

4176: rank invariants in terms of the Alexander Characters $\Delta^{(j)}$ defined

4177: in (\ref{eq-leftrace3}). We can thus write

4178: $$

4179: \lambda_{n,k}\;=\; R_{n,k}(\Delta^{(1)}, \Delta^{(2)}, \ldots)

4180: $$

4181: for some polynomial $R_{n,k}$ with integral coefficients.

4182: Note further that if $C$ is the cobordism on a Seifert surface of a knot 

4183: and $P\in{\cal P}^+$ then 

4184: $trace({\cal V}^{(P)}(C))=P(\Delta^{(1)}, \Delta^{(2)}, \ldots)$

4185: 

4186: The natural

4187: question tied to these observations is whether the TQFT's constructed 

4188: in  \cite{FroNic94} for general gauge groups are related or  equivalent

4189: to a TQFT of the polynomial form ${\cal V}^{(R_{n,k})}$ as defined in 

4190: (\ref{eq-defpolyTQFT}). If the coefficients of the $R_{n,k}$ are not

4191: all non-negative we may have to consider two theories 

4192: ${\cal V}^{(R_{n,k}^{\pm})}$ with $R_{n,k}=R_{n,k}^+-R_{n,k}^-$ and

4193: $R_{n,k}^{\pm}\in{\cal P}^+$ and make sense of their difference. 

4194: 

4195: 

4196:  

4197: 

4198: 

4199: 

4200: In \cite{Don99} Donaldson describes a slightly different TQFT, ${\cal V}^{DF}$, 

4201: modeled on moduli spaces ${\cal M}_g$ of connections on a non-trivial $SO(3)$ bundle. 

4202: This TQFT leads up to a Casson type invariant for homology circles $Y$, which is

4203: determined by the coefficients of the   Alexander polynomial $\Delta_Y$. 

4204: The vector spaces are given as 

4205: \beq\label{eq-Donaldson}

4206: {\cal V}^{DF}(\Sigma_g)\;\;=\;\;H_*\bigl({\cal M}_g\bigr)

4207: \;\;\cong\;\;\bigoplus_{j=0}^g\,\R^{j^2}\otimes \ext {g-j}H_1(\Sigma_g)\;.

4208: \eeq

4209: The morphisms ${\cal V}^{DF}(M)$ are similarly constructed via the intersection

4210: theory of representation varieties, using also  a dimension

4211: reduction of the Floer-cohomology on $\hat M\times S^1$.  

4212: 

4213: Now from Corollary~5.1.9 in \cite{GooWal98} we see that

4214:  $\ext {g-j}H_1(\Sigma_g)\,\cong\,W_{g,j+1}\oplus W_{g,j+3}\oplus W_{g,j+5}\ldots$

4215: as $\Sp(2g,\Z)$ modules. Inserting this decomposition into (\ref{eq-Donaldson}) we

4216: obtain the multiplicities stated in the following conjecture.

4217: \begin{conjecture} Let $\displaystyle D=\sum_{k\geq 1}

4218: {\scriptstyle { k+1 \choose  3}} x_k\;\in\;{\cal P}^+$. Then 

4219: $$

4220: {\cal V}^{DF}\;\cong\;{\cal V}^{(D)}\;.

4221: $$

4222: \end{conjecture}

4223: Note that on the level of vector spaces and invariants we do in fact have equality. 

4224: 

4225: 

4226: The theories in \cite{FroNic94} and \cite{Don99} are all inherently

4227: $\Z/2$-projective, and have the vanishing properties from Lemma~\ref{lm-vanish}.

4228: This indicates that they also belong into the class  of

4229: half-projective or non-semisimple TQFT's. 

4230: 

4231: 

4232: Another conjecture that is independent of a particular gauge theory may be stated

4233: as follows.

4234: \begin{conjecture} Suppose ${\cal V}$ is a non-semisimple TQFT in which the kernel

4235: of the mapping class group representations are precisely the Torelli groups. The

4236:  ${\cal V}$ is isomorphic to a sub-TQFT of some ${\cal V}^{(P)}$ for some 

4237: $P\in{\cal P}^+$.

4238: \end{conjecture}

4239: To say that the Torelli group is {\em precisely} the kernel implies that we

4240: have faithful $\Sp(2g,\Z)$-representations so that by Margulis' rigidity

4241: these lift to algebraic  $\Sp(2g,\R)$-representations. Classifying homological 

4242: TQFT's as described will thus involve exercises in branching rules as given

4243: for example in Section~8.3.4 of  \cite{GooWal98}. 

4244: 

4245: 

4246: Another approach to describing TQFT's behind the higher rank Frohman Nicas

4247: $PSU(n)$-theories or the Donaldson $SO(3)$-construction is to try to extract

4248: a categorical Hopf algebra ${\cal A}_{\cal V}$ for the given TQFT by evaluation

4249: of the cobordisms in Figure~\ref{fig-MC-tgl}. The problem with this approach,

4250: however, is that in the non-abelian TQFT's we do not seem to have a nicely

4251: defined tensor structure arising from gluing two one holed surfaces over a 

4252: pair of pants. Specifically we need an isomorphism 

4253: ${\cal V} (\Sigma_{2,1})=

4254: {\cal V} (\Sigma_{1,1})\otimes{\cal V} (\Sigma_{1,1})$ which is generally

4255: not true because of gauge constraints over the pair of pants. 

4256: 

4257: It is also not clear whether higher rank theories such as those in \cite{FroNic94}

4258: exhibit symmetries similar to the $\SL(2,\R)$-equivariance

4259:  that yields a type of Lefschetz

4260: decomposition. Particularly,  the non-abelian moduli spaces have

4261: no canonical K\"ahler structure. They do, however, admit useful Poisson structures

4262: \cite{FocRos97}.

4263:  

4264: 

4265: \paragraph{D. Milnor Torsion and Seiberg Witten invariants}\ 

4266: 

4267: 

4268: The Milnor Torsion of a 3-manifold $M$ is defined from a cell complex of a 

4269: simplicial representation of the cyclic covering space $\widetilde M$. 

4270: The relation between the Alexander polynomial and Milnor Torsion as stated

4271: in Theorem~\ref{thm-ReidAlex} suggests that there should be a quantum topological

4272: description of the invariants obtained from a simplicial complex. In fact 

4273: the Turaev Viro and Kuperberg invariant as examples of quantum invariants 

4274: that start from a cell decomposition of $M$. Given the non-semisimple nature

4275: of our theory the Kuperberg invariant \cite{Kup96} is a more natural candidate.  

4276: The basic Hopf algebra is likely to be similar to the Borel subalgebra 

4277: generated by $K$ and $\th$ but not $\tb$ following a conjecture that the

4278: Hennings and Kuperberg are related by Drinfel'd double construction. The

4279: difficulties, however, consist in describing a cell decomposition of the

4280: cyclic covering space $\widetilde M$ from a Heegaard diagram for $M$. 

4281: The main problem being that the Kuperberg theory has no easy extension 

4282: to a TQFT. 

4283: 

4284: Nevertheless, we propose as a problem to find a direct description of

4285: Milnor-Reidemeister torsion via the picture developed by Kuperberg for

4286: the construction of 3-manifold invariants. 

4287: 

4288: In \cite{Tur98} Turaev shows that an extensions of

4289: the  Milnor torsion is equal to the Seiberg Witten invariant 

4290: for 3-manifolds equipped with $Spin^C$-structures or, equivalently, Euler

4291: structures and with $\beta_1(M)>0$. A weaker version of such an equivalence 

4292: without additional structures was shown by Meng and Taubes 

4293: \cite{MengTaub}. 

4294:  The proof in \cite{Tur98} uses the fact that

4295: both invariants follows the same recursion formulae under surgery.

4296: It should be interesting to relate these formulae to the skein theory

4297: developed here and find ways of including the additional structures

4298: in our context.  

4299: 

4300: 

4301: In general our procedure is also limited to either the reduced Torsion

4302: or Alexander polynomial if $\beta_1(M)\geq 2$. Additional generators 

4303: of homology can be represented as additional  0-framed surgery links with

4304: zero linking numbers. It is, however, not as obvious in this case how 

4305: to generate a TQFT picture that would allow us to describe the full

4306: invariants with values in $\Z[H_1^{(free)}(M)]$. 

4307: 

4308:  

4309: \paragraph{E. Relations to quantum field theories:}\ 

4310: 

4311: Let us mention here only briefly interpretations of the homological TQFT's in

4312: a physics context. For small genera Rozansky and Saleur

4313: \cite{RosSal93} find the same vector spaces and representations of the

4314: mapping class groups from the $U(1,1)$ Wess-Zumino-Witten theory. 

4315: 

4316: The exterior product spaces may also be interpreted as fermionic Fock spaces.

4317: Ideas of constructing such ferminoc topological $U(1)$-theories in general

4318: have been suggested for example by Louis Crane.  

4319: }

4320: 

4321: 

4322: %%%%%%%%%%%%%%%%%%%%%%%%%%        END     CUT    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 

4323: 

4324: {\small 

4325: 

4326: 

4327: \begin{thebibliography}{99}

4328: 

4329: %\bibitem[Dr]{Dri:QG}

4330: %Drinfeld, V.G.: {Quantum groups},

4331: %Proceedings of the ICM, Amer. Math. Soc.,

4332: %Providence, Rhode Island, {\bf 1} (1987), 798--820.

4333: \bibitem{Ati88}

4334: M. Atiyah: {Topological quantum field theories}, Inst. Hautes

4335: \'Etudes Sci. Publ. Math. \textbf{68} (1988), 175--186.

4336: 

4337: \bibitem{BodNic00} H. Boden, A. Nicas:  

4338: Universal formulae for ${\rm SU}(n)$ Casson invariants of knots.  

4339: {\em Trans. Amer. Math. Soc.}   \textbf{352} (2000), no.~7, 3149--3187. 

4340: 

4341: \bibitem{BKLT00} Yu. Bespalov , T. Kerler, V. Lyubashenko,  V. Turaev:

4342: Integrals for braided Hopf algebras. {\em J. Pure Appl. Algebra}

4343:   \textbf{148} (2000)  no.~2, 113--164. 

4344: 

4345: \bibitem{BurZie} G. Burde, H. Zieschang: Knots. 

4346: {\em de Gruyter Stud. Math.} \textbf{5} Walter de Gruyter 1985. 

4347: 

4348: \bibitem{Don99}

4349: S. K. Donaldson:

4350: Topological field theories and formulae of Casson and Meng-Taubes. 

4351: Proceedings of the Kirbyfest, {\em Geom. Topol.} {\bf 2} (1999) 87--102

4352: (electronic). 

4353: \bibitem{Dri90}

4354: V.G. Drinfeld: \emph{On almost cocommutative {H}opf algebras},

4355: {\em Leningrad Math. J.} \textbf{1} (1990) no.~2, 321--342.

4356: \bibitem{Fro93} C. Frohman:  

4357: Unitary representations of knot groups. 

4358: {\em Topology} \textbf{32} (1993), no.~1, 121--144. 

4359: 

4360: \bibitem{FroNic92} C. Frohman, A. Nicas: The Alexander Polynomial via topological quantum field

4361: theory, {\em Differential Geometry, Global Analysis, and Topology, Canadian Math. Soc. Conf.

4362: Proc.} Vol. 12, Amer. Math. Soc. Providence, RI, (1992)  27--40. 

4363:  \bibitem{FocRos97}V. V. Fock,  A. A. Rosly:

4364:  Moduli space of flat connections as a Poisson manifold. 

4365:  Advances in quantum field theory and statistical mechanics: 

4366:  2nd Italian-Russian collaboration (Como, 1996). 

4367:  {\em Internat. J. Modern Phys.} B {\bf 11} (1997) no.~26-27, 3195--3206.   

4368: \bibitem{FroNic94} C. Frohman, A. Nicas: 

4369: An intersection homology invariant for knots in a rational

4370: homology 3-sphere, {\em Topology} {\bf 33} (1994) no.~1,  123-158.

4371: 

4372: \bibitem{Gil01} P. Gilmer: Integrality for TQFTs. Preprint 2001 

4373: {\tt math.QA/0105059}. 

4374: 

4375: %\bibitem{GilKer} P. Gilmer, T. Kerler: in preparation. 

4376: 

4377: \bibitem{GooWal98} 

4378: R. Goodman, N.R. Wallach: 

4379: Representations and invariants of the classical groups. 

4380: {\em Encyclopedia of Mathematics and its Applications} {\bf 68} Cambridge University

4381: Press, 1998. 

4382: \bibitem{GriHar78} P. Griffith, J. Harris: Principles of algebraic geometry. 

4383: {\em  Wiley Classics Library}. John Wiley \& Sons, Inc., 

4384: New York (1978/1994) xiv+813 pp. ISBN: 0-471-05059-8. 

4385: \bibitem{Hen96} 

4386: M. Hennings:

4387: Invariants of links and $3$-manifolds obtained from Hopf algebras,

4388: {\em J. London Math. Soc. (2)} \textbf{54} (1996), no.~3, 594--624.

4389: \bibitem{KauRad95} 

4390: L.H. Kauffman,  D.E. Radford: Invariants of $3$-manifolds derived

4391: from finite-dimensional Hopf algebras,

4392: {\em J. Knot Theory Ramifications} \textbf{4} (1995) no.~1, 131--162.

4393: \bibitem{Kau94} L.H. Kauffman: 

4394: Gauss codes, quantum groups and ribbon Hopf algebras, 

4395: {\em Rev. Math. Phys.} {\bf  5} (1993) no.~4, 735--773. 

4396: \bibitem{Ker89}

4397: T. Kerler: Darstellungen der Quantengruppen und Anwendungen. 

4398: Diploma thesis,  ETH-Z\"urich, 1989, unpublished. 

4399: \bibitem{Ker94}

4400: T. Kerler: Mapping class group actions on quantum doubles,

4401: {\em Commun. Math. Phys.} {\bf 168} (1994) 353-388.

4402: \bibitem{Ker96}

4403: T. Kerler : Genealogy of nonperturbative quantum-invariants of

4404: 3-manifolds: The surgical family. In `Geometry and Physics', {\em Lecture

4405: Notes in Pure and Applied Physics} {\bf 184}, Marcel Dekker (1997)

4406: 503-547.

4407: \bibitem{Ker98}

4408: T. Kerler: On the connectivity of cobordisms and half-projective

4409: TQFT's, 

4410: \emph{Commun. Math. Phys.} {\bf 198} No. 3 (1998) 535-590.

4411: \bibitem{Ker99}

4412: T. Kerler: Bridged links and tangle presentations

4413: of cobordism categories,  {\em Adv. Math.} {\bf 141} (1999) 207-281.

4414: \bibitem{KerRes} T. Kerler: Resolutions of $p$-Modular TQFT's and Representations of Symmetric Groups.

4415: {\tt  math.GT/0110006}

4416: 

4417: \bibitem{Kerr=5} 

4418: T. Kerler: The Structure of the Fibonacci TQFT. In preparation. 

4419: 

4420: 

4421: \bibitem{KerKyoto} T. Kerler: $p$-Modular TQFT's, 

4422: Milnor-Torsion, and the Casson-Lescop Invariant.

4423: {\tt  math.GT/0203256}. {\em Geom. Topolo. Monogr.} (in press).

4424: 

4425: \bibitem{KerLub00} T. Kerler, V.V. Lyubashenko: Non-semisimple topological quantum field theories

4426: for 3-manifolds with corners.   

4427: {\em Lecture Notes in Mathematics} 2001 Springer-Verlag (to appear)

4428: 380 pages,  200 illustrations. 

4429: \bibitem{Kir78}

4430: R. Kirby: A calculus for framed links in $S^3$,

4431: {\em Invent. Math.} {\bf 45} (1978), 35-56.

4432: 

4433: 

4434: \bibitem{Kup96} G. Kuperberg: 

4435: Noninvolutory Hopf algebras and $3$-manifold invariants. 

4436: Duke Math. J. {\bf 84}  (1996), no. 1, 83--129. 

4437: 

4438:  

4439: Involutory Hopf algebras and $3$-manifold invariants. 

4440: Internat. J. Math. {\bf 2} (1991), no. 1, 41--66. 

4441: 

4442: \bibitem{LarSwe69}

4443:  R.G. Larson,  M.E. Sweedler: An associative orthogonal

4444: bilinear form for Hopf algebras {\em  Amer. J. Math.} {\bf 91} (1969) 75--94.

4445: 

4446: \bibitem{Les} {Lescop, C.}: 

4447: Global surgery formula for the Casson-Walker invariant. 

4448: {\em Annals of Mathematics Studies}, {\bf 140}. 

4449: Princeton University Press, Princeton, NJ, 1996. 

4450: 

4451: 

4452: \bibitem{Lik97} W.B.R. Lickorish: An Introduction to Knot Theory. 

4453: {\em Graduate Texts in Mathematics} {\bf 127}. Springer Verlag, 1997. 

4454: 

4455: \bibitem{Lub95}

4456: V.V. Lyubashenko : Invariants of 3-manifolds and projective

4457: representations of mapping class groups via quantum groups at roots

4458: of unity, {\em Commun. Math. Phys. }\textbf{172} (1995) 467--516.

4459: 

4460: \bibitem{LubMaj94}

4461: V.V. Lyubashenko, S. Majid: {Braided groups and quantum

4462: Fourier transform}, J. Algebra {\bf 166} (1994), n.~3, 506--528.

4463: 

4464: \bibitem{Maj93}

4465: S. Majid: {Braided groups},

4466: J. Pure Appl. Algebra {\bf 86} (1993) n.~2, 187--221.

4467: 

4468: \bibitem{massey} W. S. Massey: A Basic Course in Algebraic Topology. 

4469: {\em Graduate Texts in Mathematics} {\bf 127}. Springer Verlag, 1991. 

4470: 

4471: \bibitem{MasRob97} G. Masbaum, J.D. Roberts:

4472: A simple proof of integrality of quantum invariants at prime roots of unity. 

4473: Math. Proc. Cambridge Philos. Soc. {\bf 121} (1997), no. 3, 443--454.

4474: 

4475: 

4476: 

4477: \bibitem{MatPol94}

4478: S. Matveev, M.  Polyak: A geometrical presentation of the surface

4479: mapping class group and surgery,

4480: {\em Commun. Math. Phys.}  {\bf 160} (1994) 537--550.

4481: 

4482: \bibitem{MengTaub} G. Meng, C. Taubes:

4483: $\underline{\rm SW}=$ Milnor torsion.  

4484: {\em Math. Res. Lett.} {\bf 3} (1996), no. 5, 661--674. 

4485: 

4486: \bibitem{Mil61}

4487: J. Milnor: A duality theorem for Reidemeister torsion. 

4488: {\em Ann. of Math.} (2) {\bf 76} (1962) 137--147. 

4489: 

4490: \bibitem{Mur94} H. Murakami: Quantum $\,SU(2)$-invariants dominate

4491: Casson's $\,SU(2)$-invariant. Math. Proc. Camb. Phil. Soc. {\bf 115} (1994),

4492: 83--103.

4493: 

4494: \bibitem{Oh96}

4495: T. Ohtsuki: A polynomial invariant of rational homology $3$-spheres. Invent.

4496: Math. 123 (1996), no.~2, 241--257. 

4497: 

4498: 

4499: \bibitem{Rad76} D. Radford: The Order of the Antipode of a Finite 

4500: Dimensional Hopf Algebra is Finite.

4501: {\em American J. Math.} (1976) {\bf 98}, no.~2, 333-355.

4502: 

4503: 

4504: \bibitem{Rad94}

4505: D. Radford: The trace function and Hopf algebras. {\em J. Algebra} {\bf 163} (1994) 583--622. 

4506: 

4507: 

4508: 

4509: 

4510: \bibitem{ResTur91}

4511: N.Yu. Reshetikhin,  V.G.  Turaev:

4512: {Invariants of 3-manifolds via link polynomials and quantum groups},

4513: Invent. Math. {\bf 103} (1991), 547--597.

4514: \bibitem{RosSal93}

4515: L. Rozansky, H. Saleur:

4516: $S$- and $T$-matrices for the super ${\rm U}(1,1) {\rm WZW}$ model. Application to surgery

4517: and $3$-manifolds invariants based on the Alexander-Conway polynomial.  

4518: {\em Nuclear Phys. B} {\bf 389} (1993) no. 2, 365--423. 

4519: 

4520: 

4521: \bibitem{Tur76} Turaev, V. G. 

4522: Reidemeister torsion and the Alexander polynomial. (Russian) 

4523: Mat. Sb. (N.S.) {\bf 18(66)} (1976), no. 2, 252--270.

4524: 

4525: \bibitem{Tur94}

4526: V. Turaev: {Quantum invariants of knots and 3-manifolds}, de Gruyter

4527:   Stud. Math. {\bf 18} Walter de Gruyter \& Co., Berlin (1994) 588 pp.

4528: 

4529: \bibitem{Tur98} V. Turaev:

4530: A combinatorial formulation for the Seiberg-Witten invariants of $3$-manifolds. 

4531: Math. Res. Lett. {\bf 5} (1998), no. 5, 583--598.

4532: 

4533: \bibitem{TurVi92} V. Turaev, O. Viro:

4534: State sum invariants of $3$-manifolds and quantum $6j$-symbols. 

4535: Topology {\bf 31} (1992), no. 4, 865--902.

4536: 

4537: \bibitem{Waj83}B. Wajnryb: A simple presentation for the mapping class group of an orientable surface, {\em Israel J. Math.}{\bf 45 }(1983) no. 2-3, 157--174. 

4538: B. Wajnryb: An elementary approach to the mapping class group of a surface,

4539: {\em Geom. Topol.}{\bf 3} (1999) 405--466 (electronic). 

4540: 

4541: \bibitem{Wi89} E. Witten:  

4542: Quantum field theory and the Jones polynomial. 

4543: {\em Comm. Math. Phys.}   {\bf 121} (1989), no.~3, 351--399. 

4544: 

4545: \bibitem{Yet97}

4546: D. Yetter:

4547: Portrait of the handle as a Hopf algebra. 

4548:  Geometry and physics. (Aarhus, 1995),

4549: {\em Lecture Notes in Pure and Appl. Math.} {\bf  184}, 

4550: Dekker, New York, 1997,  pp. 481--502.  

4551: 

4552: \end{thebibliography} 

4553: }

4554: 

4555: {\sc\footnotesize  The Ohio State University, 

4556: 

4557: Department of Mathematics,

4558:          

4559: 

4560: 231 West 18th Avenue,

4561: 

4562:          Columbus, OH 43210, U.S.A. }

4563: 

4564:  {\em E-mail: }{ \tt kerler@math.ohio-state.edu}

4565: \end{document}

4566: \newpage

4567:  

4568: \begin{comment}

4569: The combinatorial description of ${\cal V}_{\cal N}$ can be 

4570: greatly simplified with a few modifications of the Hennings calculus. 

4571: For example, in the substitution in (\ref{fig-Dec-cross}) we want to 

4572: reinterpret the indifferent crossing as the ordinary, original crossing with 

4573: ${\cal Z}={\cal Z}^{\dagger}$ still inserted. Hence, the sum $\sum_j$ in (\ref{fig-Dec-cross})

4574: contains only the two terms $1\otimes 1$ and $\th\otimes K\tb$. One easily verifies

4575: that the new crossings with ${\cal Z}$ still fulfill the strict Reidemeister 

4576: moves of curves in the plane without the extra element $G=K$ for the first move.

4577: Moreover,  we can move $\th$ and $\tb$ through a 

4578: crossing at the expense of adding a $K$ to the opposing strand. Other useful rules 

4579: are that we only need to consider summands for which every closed

4580: component has exactly one $\th$ and one $\tb$, and a component that contains a 

4581: decoration $\rho$ can be removed after multiplication with $i$. Further simplifications

4582: include  rules for the insertion of the tensor $\tb\otimes\th-\th\otimes\tb$ on

4583: different strands. Summarily, we have an effective calculus at our hands that allows us to compute 

4584: the relevant  homological data for ${\cal V}^{FN}$ from a surgery diagram.  

4585: 

4586: %$\,B_1(x\otimes y)\;=\;\;\;{\sf T}(x)\otimes y\;=\;

4587: % \;\;\;x\otimes y +\mu(x)1\otimes y\;$

4588: %\medskip

4589: \end{comment}

4590: 

4591: 

4592: 

4593: +++++++++++++++++++++++

4594: 

4595: In the constructions for the $U(1)$-case are generalized to 

4596: $PU(n)$-representations, yielding  TQFT's,  

4597: ${\cal V}^{FN}_{n,k}$,  and knot invariants,  . The vector spaces

4598: for connected surfaces with one boundary component are given as 

4599: $$

4600: {\cal V}^{FN}_{n,k}(\Sigma_{g,1})\;=\;

4601: IH^m_*\Biggl(p_*\Bigl({\rm Hom}_k(\pi_1(\Sigma_{g,1}),SU(n))/SU(n)\Bigr)\Biggr)\;.

4602: $$

4603: Here, ${\rm Hom}_k$ stands for all those representation 

4604: $\rho: \pi_1(\Sigma_{g,1})\to SU(n)$, which map the class $z\in \pi_1$

4605: of the loop around the hole to the central element

4606: $e^{2\pi i\frac kn}\in SU(n)$. $\ldots/SU(n)$ denotes conjugacy classes with

4607: respect to the adjoint action on $SU(n)$, and $p_*$ is the map induced

4608: by $p:SU(n)\to PU(n)$. Finally, $IH^m_*$ denotes the Goresky-MacPherson

4609: intersection homology in middle perversity $m$. 

4610: 

4611: 

4612: +++++++++++++++++++++++++

4613: 

4614: Here corresponding coupons are labeled with the same letter. 

4615: Given orientations on the strands we thus have the same situation as in the classification

4616: of 2-dimensional surfaces, see \cite{massey}. The coupons can be slid into the standard

4617: form with torus blocs as in the picture below. 

4618: 

4619: 

4620:  

4621: \begin{center}

4622: \leavevmode

4623: \epsfbox{skeineval.eps}

4624: \end{center}

4625: %\caption{Skein Evaluation}\lbl{fig-skeineval}

4626: %\end{figure}

4627: 

4628: