1: %Last changed by DK on 8 September 2004; minor changes on 14 January 2006

2: 

3: \documentclass[12pt]{amsart}

4: \usepackage{epsfig,amscd,amssymb,times}

5: %\usepackage{backref}

6: %\usepackage{graphics}

7: \setlength{\headheight}{6.15pt}

8: 

9: \newtheorem{theorem}{Theorem}%[section]

10: \newtheorem{lemma}[theorem]{Lemma}

11: \newtheorem{proposition}[theorem]{Proposition}

12: \newtheorem{corollary}[theorem]{Corollary}

13: \theoremstyle{definition}

14: \newtheorem{definition}[theorem]{Definition}

15: \newtheorem{problem}[theorem]{Problem}

16: \newtheorem{example}[theorem]{Example}

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

18: \theoremstyle{remark}

19: \newtheorem{remark}[theorem]{Remark}

20: %\numberwithin{figure}{section}

21: %\numberwithin{table}{section}

22: 

23: \def\varph{{\varphi}}

24: \def\ra{{\rightarrow}}

25: \def\lra{{\longrightarrow}}

26: \def\om{{\omega}}

27: \def\Om{{\Omega}}

28: \def\la{{\lambda}}

29: \def\al{{\alpha}}

30: 

31: \def\La{{\Lambda}}

32: \def\ga{{\gamma}}

33: \def\Ga{{\Gamma}}

34: \def\vGa{{\varGamma}}

35: \def\bC{{\mathbb C}}

36: \def\bZ{{\mathbb Z}}

37: \def\bQ{{\mathbb Q}}

38: \def\bR{{\mathbb R}}

39: \def\inv{{^{-1}}}

40: \def\Sg{{\Sigma_g}}

41: \def\Mg{{\mathcal M}_g}

42: \def\Msg{{\mathcal M}_{g,*}}

43: \def\M1g{{\mathcal M}_{g,1}}

44: \def\I1g{{\mathcal I}_{g,1}}

45: 

46: \newcommand\Hom{\operatorname{Hom}}

47: \newcommand\End{\operatorname{End}}

48: \newcommand\Aut{\operatorname{Aut}}

49: \newcommand\Der{\operatorname{Der}}

50: \newcommand\Out{\operatorname{Out}}

51: \newcommand\vol{\operatorname{vol}}

52: \newcommand\sgn{\operatorname{sgn}}

53: \newcommand\Flux{\operatorname{Flux}}

54: \newcommand\Symp{\operatorname{Symp}}

55: \newcommand\Ham{\operatorname{Ham}}

56: \newcommand\eFlux{\operatorname{\widetilde{Flux}}}

57: \newcommand\Cal{\operatorname{Cal}}

58: \newcommand\Ker{\operatorname{Ker}}

59: \newcommand\Cok{\operatorname{Cok}}

60: \newcommand\im{\operatorname{Im}}

61: \newcommand\Int{\operatorname{Int}}

62: \newcommand\sign{\operatorname{sign}}

63: \newcommand\BS{\operatorname{BSymp}}

64: \newcommand\ES{\operatorname{ESymp}}

65: \newcommand\BD{\operatorname{BDiff}}

66: \newcommand\BG{\operatorname{B\bar\Gamma}}

67: \newcommand\Diff{\operatorname{Diff}}

68: \newcommand\id{\operatorname{id}}

69: \catcode`\@=\active % make @ other character ( is default)

70: 

71: \begin{document}

72: \title[Characteristic classes of foliated surface bundles]

73: {Characteristic classes of foliated surface bundles with area-preserving 

74: holonomy}

75: \author{D.~Kotschick}

76: \address{Mathematisches Institut, Ludwig-Maxi\-mi\-lians-Universit\"at 

77: M\"unchen,

78: Theresienstr.~39, 80333 M\"unchen, Germany}

79: \email{dieter{\char'100}member.ams.org}

80: \author{S.~Morita}

81: \address{Department of Mathematical Sciences\\

82: University of Tokyo \\Komaba, Tokyo 153-8914\\

83: Japan}

84: \email{morita{\char'100}ms.u-tokyo.ac.jp}

85: 

86: \keywords{symplectomorphism, area-preserving diffeomorphism,

87: foliated surface bundle, Hamiltonian symplectomorphism}

88: \thanks{The first author is grateful to the {\sl Deutsche 

89: Forschungsgemeinschaft}

90: for support of this work. The second author is partially supported by JSPS 

91: Grant

92: 16204005}

93: \subjclass{Primary 57R17, 57R50, 57M99; secondary 57R50, 58H10}

94: 

95: \begin{abstract}

96: Making use of the {\it extended} flux homomorphism defined in~\cite{KM03}

97: on the group $\Symp\Sg$ of symplectomorphisms of a closed oriented surface

98: $\Sg$ of genus $g\geq 2$, we introduce new characteristic classes of foliated

99: surface bundles with symplectic, equivalently area-preserving, total holonomy.

100: These characteristic classes are stable with respect to $g$ and we show that

101: they are highly non-trivial. We also prove that the second homology of

102: the group $\Ham\Sg$ of Hamiltonian symplectomorphisms of $\Sg$, equipped with

103: the discrete topology, is very large for all $g\geq 2$.

104: \end{abstract}

105: 

106: \maketitle

107: 

108: \section{Introduction}\label{s:intro}

109: 

110: In this paper we study the homology of symplectomorphism groups of

111: surfaces considered as discrete groups. We shall prove that certain

112: homology groups are highly non-trivial by constructing characteristic

113: classes of foliated surface bundles with area-preserving holonomy, and

114: proving non-vanishing results for them.

115: 

116: Let $\Sg$ be a closed oriented surface of genus $g\geq 2$, and $\Diff_+\Sg$

117: its group of orientation preserving selfdiffeomorphisms. We fix an area

118: form $\omega$ on $\Sg$, which, for dimension reasons, we can also think of

119: as a symplectic form. We denote by $\Symp\Sg$ the subgroup of $\Diff_+\Sg$

120: preserving the form $\om$. The classifying space $\BS ^\delta\Sg$ for the group

121: $\Symp\Sg$ with the {\it discrete} topology is an Eilenberg-MacLane space

122: $K(\Symp^\delta\Sg,1)$ which classifies foliated $\Sg$-bundles with

123: area-preserving total holonomy groups.

124: 

125: Our construction of characteristic classes proceeds as follows.

126: Let $\Symp_0\Sg$ be the identity component of $\Symp\Sg$.

127: A well-known theorem of Moser~\cite{Moser} concerning volume-preserving

128: diffeomorphisms implies that the quotient $\Symp\Sg/\Symp_0\Sg$ can be

129: naturally identified with the mapping class group $\Mg$, so that

130: we have an extension

131: $$

132: 1\lra\Symp_0\Sg\lra\Symp\Sg\overset{p}{\lra}\Mg\lra 1 \ .

133: $$

134: There is a surjective homomorphism $\Flux\colon\Symp_0\Sg\ra H^1(\Sg;\bR)$,

135: called the flux homomorphism. In~\cite{KM03} we proved that this homomorphism

136: can be extended to a crossed homomorphism

137: $$

138: \widetilde\Flux\colon\Symp\Sg\lra H^1(\Sg;\bR) \ ,

139: $$

140: which we call the {\it extended} flux homomorphism. This extension is 

141: essentially

142: unique in the sense that the associated cohomology class

143: $$

144: [\widetilde\Flux]\in H^1(\Symp^\delta\Sg;H^1(\Sg;\bR))

145: $$

146: with twisted coefficients is uniquely defined. Now we consider the powers

147: $$

148: [\widetilde\Flux]^k\in H^k(\Symp^\delta\Sg;H^1(\Sg;\bR)^{\otimes k})\quad

149: (k=2,3,\cdots) \ ,

150: $$

151: and apply $\Mg$-invariant homomorphisms

152: $$

153: \la\colon H^1(\Sg;\bR)^{\otimes k}\lra \bR

154: $$

155: to obtain cohomology classes

156: $$

157: \la([\widetilde\Flux]^k)\in H^{k}(\BS^\delta\Sg;\bR)

158: $$

159: with constant coefficients.

160: The usual cup-product pairing $H^1(\Sg;\bR)^{\otimes 2}\ra\bR$

161: is the main example of such a homomorphism $\la$.

162: 

163: This method of constructing constant cohomology classes out of twisted ones

164: was already used in~\cite{Morita96} in the case of the mapping class group,

165: where the Torelli group (respectively the Johnson homomorphism) played the

166: role of $\Symp_0\Sg$ (respectively of the flux homomorphism) here. In that

167: case, it was proved in {\it loc.~cit.}~that all the Mumford--Morita--Miller

168: classes can be obtained in this way. The precise formulae were given

169: in~\cite{KM96,KM01}, with the important implication that no other classes

170: appear. In our context here, we can go further by enhancing the coefficients

171: $\bR$ to associated $\bQ$-vector spaces which appear as the targets of various

172: multiples of the {\it discontinuous} cup-product pairing

173: $$

174: H^1(\Sg;\bR)\otimes_\bZ H^1(\Sg;\bR)\lra S^2_\bQ\bR \ ,

175: $$

176: where $S^2_\bQ\bR$ denotes the second symmetric power of $\bR$ over

177: $\bQ$, see Section~\ref{s:main} for the details. In this way, we obtain

178: many new characteristic classes in

179: $$

180: H^*(\Symp^\delta\Sg;S^*(S^2_\bQ\bR)) \ ,

181: $$

182: where

183: $$

184: S^*(S^2_\bQ\bR)=\bigoplus_{k=1}^\infty S^k(S^2_\bQ\bR)

185: $$

186: denotes the symmetric algebra of $S^2_\bQ\bR$.

187: On the other hand, we proved in~\cite{KM03} that any power

188: $$

189: e_1^k\in H^{2k}(\Symp^\delta\Sg;\bQ)

190: $$

191: of the first Mumford--Morita--Miller class $e_1$ is non-trivial for

192: $k\leq \frac{g}{3}$. Now we can consider the cup products of $e_1^k$ with

193: the new characteristic classes defined above. The main purpose of the present

194: paper is to prove that these characteristic classes are all non-trivial in a

195: suitable stable range.

196: 

197: The contents of this paper is as follows. In Section~\ref{s:main} precise

198: statements of the main results are given. In Section~\ref{s:trans} we study,

199: in detail, the transverse symplectic class of foliated $\Sg$-bundles with

200: area-preserving total holonomy groups. In Section~\ref{s:open} we construct

201: two kinds of {\it extended} flux homomorphisms for open surfaces

202: $\Sigma_g^0=\Sg\setminus D^2$. We compare these extended flux homomorphisms

203: with the one obtained in the case of closed surfaces. This is used in

204: Section~\ref{s:secondopen} to generalize our result on the second homology of

205: the symplectomorphism group to the case of open surfaces. Sections~\ref{s:open}

206: and~\ref{s:secondopen} are the heart of this paper. Then in Section~\ref{s:proofs} we use the results of the previous sections to show the non-t 

207: riviality of cup

208: products of various characteristic classes, thus yielding proofs of the main

209: results about the homology of symplectomorphism groups as discrete groups. In

210: the final Section~\ref{s:misc} we give definitions of yet more characteristic

211: classes, other than the ones given in Section~\ref{s:main}. We propose several

212: conjectures and problems about them.

213: 

214: \section{Statement of the main results}\label{s:main}

215: 

216: Consider the usual cup-product pairing

217: $$

218: \iota\colon H^1(\Sg;\bR)\otimes H^1(\Sg;\bR)\lra \bR

219: $$

220: in cohomology, dual to the intersection pairing in homology. For simplicity,

221: we denote $\iota(u,v)$ by $u\cdot v$, where $u,v\in H^1(\Sg;\bR)$.

222: We first lift this pairing as follows.

223: 

224: As in Section~\ref{s:intro}, let $S^2_{\bQ}\bR$ denote the second symmetric

225: power of $\bR$ over $\bQ$. In other words, this is a vector space over $\bQ$,

226: consisting of the homogeneous polynomials of degree two generated by the

227: elements of $\bR$ considered as a vector space over $\bQ$. For each element

228: $a\in\bR$, we denote by $\hat a$ the corresponding element in $S^1_\bQ\bR$.

229: Thus any element in $S^2_\bQ\bR$ can be expressed as a finite sum

230: $$

231: \hat a_1\hat b_1+\cdots +\hat a_k\hat b_k

232: $$

233: with $a_i, b_i\in\bR$. We have a natural projection

234: $$

235: S^2_\bQ\bR\lra \bR

236: $$

237: given by the correspondence $\hat a\mapsto a\ (a\in\bR)$.

238: 

239: With this terminology we make the following definition:

240: \begin{definition}[{\it Discontinuous intersection pairing}]

241: Define a pairing

242: $$

243: \tilde\iota\colon H^1(\Sg;\bR)\times H^1(\Sg;\bR)\lra S^2_\bQ\bR

244: $$

245: as follows. Choose a basis $x_1,\cdots , x_{2g}$ of $H^1(\Sg;\bQ)$.

246: For any two elements, $u,v\in H^1(\Sg;\bR)$, write

247: $$

248: u=\sum_i a_i x_i,\quad v=\sum_i b_i x_i\quad (a_i, b_i\in\bR) \ .

249: $$

250: Then we set

251: $$

252: \tilde\iota (u,v)=\sum_{i,j} \iota(x_i,x_j) \hat a_i \hat b_j \in S^2_\bQ

253: \bR \ .

254: $$

255: \end{definition}

256: Clearly $\tilde\iota$ followed by the projection $S^2_\bQ\bR\lra \bR$ is

257: nothing but the usual intersection or cup-product pairing $\iota$.

258: Henceforth we simply write $u\odot v$ for $\tilde\iota (u,v)$.

259: 

260: It is easy to see that $\tilde\iota$ is well defined independently of the

261: choice of basis in $H^1(\Sg;\bQ)$. We can see from the following proposition

262: that $\tilde\iota$ enumerates all $\bZ$-multilinear skew-symmetric pairings

263: on $H^1(\Sg;\bR)$ which are $\Mg$-invariant. Let $\La^2_{\bZ} H^1(\Sg;\bR)$

264: denote the second exterior power, over $\bZ$, of $H^1(\Sg;\bR)$ considered as

265: an abelian group, rather than as a vector space over $\bR$. Also let

266: $\left(\La^2_{\bZ} H^1(\Sg;\bR)\right)_{\Mg}$ denote the abelian group of

267: coinvariants of $\La^2_{\bZ} H^1(\Sg;\bR)$ with respect to the natural action

268: of $\Mg$.

269: \begin{proposition}\label{prop:la2}

270: There exists a canonical isomorphism

271: $$

272: \left(\La^2_{\bZ} H^1(\Sg;\bR)\right)_{\Mg}\cong S^2_{\bQ} \bR

273: $$

274: given by the correspondence

275: $$

276: \left(\sum_i a_i u_i\right)\land

277: \left(\sum_j b_j v_j\right)\longmapsto

278: \sum_{i,j} \iota(u_i,v_j)\ \hat a_i \hat b_j\ \ ,

279: $$

280: where $a_i, b_j\in\bR$, $u_i,v_j\in H^1(\Sg;\bQ)$.

281: \end{proposition}

282: In order not to digress, we refer the reader to the appendix for a proof.

283: 

284: Before we can define some new cocycles on the group $\Symp\Sg$, we have to

285: recall some facts from~\cite{KM03}. The symplectomorphism group $\Symp\Sg$

286: acts on its identity component by conjugation, and acts on $H^{1}(\Sg;\bR)$

287: from the left via $\varph(w)=(\varph^{-1})^*(w)$.

288: The flux homomorphism $\Flux\colon\Symp_{0}\Sg\rightarrow H^1(\Sg;\bR)$

289: is equivariant with respect to these actions by Lemma~6 of~\cite{KM03}.

290: Its extension $\widetilde{\Flux}\colon\Symp\Sg\rightarrow H^1(\Sg;\bR)$

291: is a crossed homomorphism for the above action in the sense that

292: \begin{equation}\label{eq:eflux}

293: \widetilde{\Flux}(\varph\psi)=\widetilde{\Flux}(\varph)+

294: (\varph\inv)^*\widetilde{\Flux}(\psi) \ .

295: \end{equation}

296: \begin{definition}\label{def:al}

297:      Let $\varph_1,\ldots,\varph_{2k}\in\Symp\Sg$, and

298:      $$

299: \xi_i=((\varph_1 \ldots\varph_{i-1})^{-1})^*\widetilde{\Flux}(\varph_i) \ .

300: $$

301: Define a $2k$-cocycle $\tilde\al^{(k)}$ with values in $S^{k}(S^2_\bQ\bR)$ by

302: \begin{align*}

303: \tilde\al^{(k)}&(\varph_1,\ldots,\varph_{2k})\\

304: &=

305: \frac{1}{(2k)!}\sum_{\sigma\in\mathfrak{S}_{2k}}

306: \sgn\sigma\ (\xi_{\sigma(1)}\odot\xi_{\sigma(2)})\ldots

307: (\xi_{\sigma(2k-1)}\odot\xi_{\sigma(2k)})\in S^{k}(S^2_\bQ\bR) \ ,

308: \end{align*}

309: where the sum is over permutations in the symmetric group $\mathfrak{S}_{2k}$.

310: \end{definition}

311: 

312: That $\tilde\al^{(k)}$ is indeed a cocycle is easy to check by a standard

313: argument in the theory of cohomology of groups using~\eqref{eq:eflux}. Thus

314: we have the corresponding cohomology classes

315: $$

316: \tilde\al^{(k)}\in H^{2k}(\Symp^\delta\Sg;S^{k}(S^2_\bQ\bR)) \ ,

317: $$

318: denoted by the same letters. If we apply the canonical projection

319: $S^{k}(S^2_\bQ\bR)\ra\bR$ to these classes, we obtain real cohomology classes

320: $$

321: \al^k\in H^{2k}(\Symp^\delta\Sg;\bR) \ ,

322: $$

323: which are the usual cup products of the first one $\al\in 

324: H^2(\Symp^\delta\Sg;\bR)$.

325: The refined classes $\tilde\al^{(k)}$ can be considered as a twisted version of

326: {\it discontinuous invariants} in the sense of~\cite{Morita85} arising from the

327: flux homomorphism.

328: 

329: Now we can state our first main result.

330: \begin{theorem}\label{th:hev}

331: For any $k\geq 1$ and $g\geq 3k$, the characteristic classes

332: $$

333: e_1^k, e_1^{k-1}\tilde\al, \ldots, e_1\tilde\al^{(k-1)}, \tilde\al^{(k)}

334: $$

335: induce a surjective homomorphism

336: $$

337: H_{2k}(\Symp^{\delta}\Sg;\bZ)\lra

338: \bZ\oplus S^{2}_{\bQ}\bR\oplus\cdots\oplus S^{k}(S^2_\bQ\bR) \ .

339: $$

340: \end{theorem}

341: For $k=1$ this is not hard to see, so we give the proof right away.

342: For $k>1$ the proof is given in Section~\ref{s:proofs} below and requires

343: the technical results developed in the body of this paper.

344: 

345: Consider the subgroup $\Ham\Sg$ of $\Symp_0\Sg$ consisting of all

346: Hamiltonian symplectomorphisms of $\Sg$. As is well known

347: (see~\cite{Banyaga,MS}), we have an extension

348: \begin{equation}\label{Ham}

349: 1\lra\Ham\Sg\lra\Symp_0\Sg\ \overset{\Flux}{\lra}\

350: H^1(\Sg;\bR)\lra 1 \ .

351: \end{equation}

352: This gives rise to a $5$-term exact sequence in cohomology:

353: \begin{align*}

354: 0 \lra H^{1}(H^1(\Sg;\bR)^\delta;\bZ) &\stackrel{\Flux^{*}}{\lra}

355: H^{1}(\Symp_0^\delta\Sg;\bZ)\lra

356: H^{1}(\Ham^\delta\Sg;\bZ)^{H^1_{\bR}}\\

357: &\lra H^{2}(H^1(\Sg;\bR)^\delta;\bZ)\stackrel{\Flux^{*}}{\lra}

358: H^{2}(\Symp_0^\delta\Sg;\bZ) \ ,

359: \end{align*}

360: where we have written $H^1_{\bR}$ for $H^1(\Sg;\bR)$. Now $\Ham\Sg$ is

361: a perfect group by a result of Thurston~\cite{Thurston0}, see also

362: Banyaga~\cite{Banyaga}. Therefore, $\Flux^{*}$ injects the second

363: cohomology of $H^1(\Sg;\bR)$ as a discrete group into that of

364: $\Symp_0^\delta\Sg$. By definition, the class $\tilde\alpha$ is the image

365: of the class of $\tilde\iota$ under $\Flux^{*}$. So $\tilde\alpha$ is

366: nontrivial on $\Symp_{0}^{\delta}\Sg$, and is defined on the whole

367: $\Symp^{\delta}\Sg$. We conclude that $\tilde\alpha$ defines a

368: surjective homomorphism

369: $$

370: H_{2}(\Symp^{\delta}\Sg;\bZ)\rightarrow S^{2}_{\bQ}\bR \ ,

371: $$

372: for any $g\geq 2$. We already proved in~\cite{KM03} that the first

373: Mumford--Morita--Miller class $e_{1}$ defines a surjection

374: $H_{2}(\Symp^{\delta}\Sg;\bZ)\longrightarrow \bZ$ for all $g\geq 3$.

375: Clearly the two classes are linearly independent because $\tilde\alpha$

376: is nonzero on $\Symp_{0}^{\delta}\Sg$, to which $e_{1}$ restricts trivially.

377: This proves Theorem~\ref{th:hev} in the easy case when $k=1$.

378: 

379: We can restrict the homomorphism

380: $$

381: \Flux^*\colon H^*(H^1(\Sg;\bR)^\delta;\bR)\lra

382: H^*(\Symp_0^\delta\Sg;\bR)

383: $$

384: to the {\it continuous cohomology}

385: $$

386: H^*_{ct}(H^1(\Sg;\bR)^\delta;\bR)\cong H_*(T^{2g};\bR)

387: \subset H^*(H^1(\Sg;\bR)^\delta;\bR) \ ,

388: $$

389: see Section~\ref{s:trans} for the precise definition. Thereby we obtain a ring

390: homomorphism

391: $$

392: \Flux^*\colon H_*(T^{2g};\bR)\longrightarrow H^*(\Symp_0^\delta\Sg;\bR) \ ,

393: $$

394: where $T^{2g}=K(\pi_1\Sg,1)$ is the Jacobian manifold of $\Sg$, and the ring

395: structure on the homology of $T^{2g}$ is induced by the Pontrjagin product.

396: Let $\om_0\in H_2(T^{2g};\bR)$ be the homology class represented by the dual

397: of the standard symplectic form on $T^{2g}$. We decompose the

398: $Sp(2g,\bR)$-module $H_k(T^{2g};\bR)$ into irreducible components.

399: For this, consider the homomorphism

400: $$

401: \om_0\wedge \ \colon H_{k-2}(T^{2g};\bR)\longrightarrow H_k(T^{2g};\bR)

402: $$

403: induced by the wedge product with $\om_0$. On the one hand, it is easy

404: to see using Poincar\'e duality on $T^{2g}$, that the above homomorphism is

405: surjective for any $k\geq g+1$. On the other hand, it is well-known

406: (see~\cite{FH}), that the kernel of the contraction homomorphism

407: $$

408: C\colon H_k(T^{2g};\bQ)\longrightarrow H_{k-2}(T^{2g};\bQ)

409: $$

410: induced by the intersection pairing

411: $H_2(T^{2g};\bQ)\cong \La^2 H_1(\Sg;\bQ)\ra\bQ$

412: is the irreducible representation of the algebraic group $Sp(2g,\bQ)$

413: corresponding to the Young diagram $[1^{k}]$ for any $k\leq g$.

414: Let $[1^k]_\bR=[1^k]\otimes\bR$ denote the real form of this representation.

415: Then we have a direct sum decomposition

416: \begin{equation}\label{eq:irrep}

417: H_k(T^{2g};\bR)= [1^k]_\bR\oplus \om_0\land H_{k-2}(T^{2g};\bR)

418: \quad (k\leq g) \ .

419: \end{equation}

420: \begin{theorem}\label{th:om2}

421: The kernel of the homomorphism

422: $$

423: \Flux^*\colon H_*(T^{2g};\bR)\lra H^*(\Symp_0^\delta\Sg;\bR)

424: $$

425: induced by the flux homomorphism is the ideal generated by the subspace

426: $\om_0\land H_1(T^{2g};\bR)\subset H_3(T^{2g};\bR)$, and the image of this

427: homomorphism can be described as

428: $$

429: \im \Flux^*\cong \bR\oplus \bigoplus_{k=1}^{g} [1^k]_\bR \ ,

430: $$

431: where $\bR$ denotes the image of the subspace of $H_2(T^{2g};\bR)$ spanned

432: by $\om_0$.

433: \end{theorem}

434: %\begin{theorem}\label{th:om2}

435: %Let $\om_0\in H_2(T^{2g};\bR)$ be the homology class represented by the dual

436: %of the standard symplectic form on $T^{2g}$. The kernel of the homomorphism

437: %$$

438: %\Flux^*\colon H_*(T^{2g};\bR)\lra H^*(\Symp_0^\delta\Sg;\bR)

439: %$$

440: %induced by the flux homomorphism is the ideal generated by the subspace

441: %$\om_0\land H_1(T^{2g};\bR)\subset H_3(T^{2g};\bR)$.

442: %\end{theorem}

443: 

444: The group $H^1(\Sg;\bR)$ acts on $\Ham\Sg$ by outer automorphisms.

445: In particular, it acts on the homology $H_*(\Ham^{\delta}\Sg;\bZ)$

446: of the discrete group $\Ham^{\delta}\Sg$ so that we can consider the

447: coinvariants $H_*(\Ham^{\delta}\Sg;\bZ)_{H^1_{\bR}}$, where for simplicity

448: we have written $H^1_{\bR}$ instead of $H^1(\Sg;\bR)$.

449: The following result shows that the second homology group

450: $H_2(\Ham^{\delta}\Sg;\bZ)$ is highly non-trivial.

451: \begin{theorem}\label{th:ham}

452: For any $g\geq 2$, there exists a natural injection

453: $$

454: H^1(\Sg;\bR)\subset H_2(\Ham^{\delta}\Sg;\bZ)_{H^1_{\bR}} \ .

455: $$

456: \end{theorem}

457: 

458: \section{The transverse symplectic class}\label{s:trans}

459: 

460: Since $\Sg$ is an Eilenberg-MacLane space, the total space

461: of the universal foliated $\Sg$-bundle over the classifying

462: space $\BS^\delta\Sg$ is again a $K(\pi,1)$ space.

463: Hence if we denote by $\ES^{\delta}\Sg$ the fundamental group of

464: this total space, then we obtain a short

465: exact sequence

466: \begin{equation}\label{eq:esymp}

467: 1\lra\pi_1\Sg\lra\ES^{\delta}\Sg\lra\Symp^{\delta}\Sg\lra 1

468: \end{equation}

469: and any cohomology class of the total space can be considered as

470: an element in the group cohomology of $\ES^{\delta}\Sg$.

471: Now on the total space of any foliated $\Sg$-bundle with total holonomy

472: group contained in $\Symp\Sg$ there is a closed $2$-form $\tilde\om$

473: which restricts to the symplectic form $\om$ on each fiber. At the

474: universal space level, the de Rham cohomology class of $\tilde\om$

475: defines a class $v\in H^2(\ES^\delta;\bR)$ which we call the transverse

476: symplectic class. We normalize the symplectic form $\om$ on $\Sg$

477: so that its total area is equal to $2g-2$. It follows that the

478: restriction of $v$ to a fiber is the same as the negative of the Euler

479: class $e\in H^2(\ES^\delta;\bR)$ of the vertical tangent bundle.

480: 

481: Let $\ES_0\Sg$ denote the subgroup of $\ES^{\delta}\Sg$ obtained by

482: restricting the extension~\eqref{eq:esymp} to $\Symp_0\Sg\subset\Symp\Sg$.

483: Since any foliated $\Sg$-bundle with total holonomy in $\Symp_0\Sg$ is

484: trivial as a differentiable $\Sg$-bundle, there exists an isomorphism

485: $$

486: \ES_0\Sg\cong \pi_1\Sg\times\Symp_0\Sg \ .

487: $$

488: Henceforth we identify the above two groups.

489: By the K\"unneth decomposition, we have an isomorphism

490: \begin{equation}\label{eq:kunneth}

491: \begin{split}

492: H^2(&\ES_0^\delta\Sg;\bR)\cong H^2(\Sg;\bR)\oplus\\

493: &\oplus \left(H^1(\Sg;\bR)\otimes

494: H^1(\Symp_0^\delta\Sg;\bR)\right)\oplus H^2(\Symp_0^\delta\Sg;\bR) \ ,

495: \end{split}

496: \end{equation}

497: where we identify $H^*(\pi_1\Sg;\bR)$ with $H^*(\Sg;\bR)$.

498: %which is finite dimensional.

499: Let $\mu\in H^2(\Sg;\bZ)$ be the fundamental cohomology class

500: of $\Sg$. Clearly the Euler class $e\in H^2(\ES_0^\delta\Sg;\bR)$

501: is equal to $(2-2g)\mu$.

502: The flux homomorphism %$\Flux\colon\Symp_0\Sg\ra H^1(\Sg;\bR)$

503: gives rise to an element

504: \begin{align*}

505: [\Flux]\in \hspace{3mm} &\Hom_\bZ (H_1(\Symp_0^\delta\Sg;\bZ),H^1(\Sg;\bR))\\

506: \cong & \Hom_\bR(H_1(\Symp_0^\delta\Sg;\bR),H^1(\Sg;\bR))\\

507: \cong & H^1(\Sg;\bR)\otimes H^1(\Symp_0^\delta\Sg;\bR) \ ,

508: \end{align*}

509: where the last isomorphism exists because $H^1(\Sg;\bR)$

510: is finite dimensional. Choose a symplectic basis

511: $x_1,\ldots,x_g$, $y_1,\ldots,y_g$

512: of $H_1(\Sg;\bR)$ and denote by $x_1^*,\ldots,x_g^*$, $y_1^*,\ldots,y_g^*$

513: the dual basis of $H^1(\Sg;\bR)$. Then Poincar\'e duality

514: $H_1(\Sg;\bR)\cong H^1(\Sg;\bR)$ is given by the

515: correspondence $x_i\mapsto -y_i^*, y_i\mapsto x_i^*$.

516: The element $[\Flux]$ can be described explicitly as

517: \begin{equation}\label{eqn:flux}

518: [\Flux]=\sum_{i=1}^{g} (x_i^*\otimes \tilde x _i+y_i^*\otimes \tilde y_i)

519: \in H^1(\Sg;\bR)\otimes H^1(\Symp_0^\delta\Sg;\bR)

520: \end{equation}

521: where $\tilde x_i,\tilde y_i\in H^1(\Symp_0^\delta\Sg;\bR)\cong

522: \Hom (H_1(\Symp_0\Sg;\bZ);\bR)$ is defined by the

523: equality

524: $$

525: \Flux(\varph)=\sum_{i=1}^{g}

526: (\tilde x_i(\varph) x_i+\tilde y_i(\varph) y_i)\quad (\varph\in \Symp_0\Sg).

527: $$

528: The elements $\tilde x_i,\tilde y_i$ can be also interpreted

529: as follows.

530: The flux homomorphism induces a homomorphism in cohomology

531: \begin{equation}\label{eq:flux*}

532: \Flux^*\colon H^*(H^1(\Sg;\bR)^\delta;\bR)\lra H^*(\Symp_0^\delta\Sg;\bR)

533: \end{equation}

534: where the domain

535: $$

536: H^*(H^1(\Sg;\bR)^\delta;\bR)\cong \Hom_{\bZ}(\La_{\bZ}^*(H^1(\Sg;\bR)),\bR)

537: $$

538: is the cohomology group of $H^1(\Sg;\bR)$ considered as a {\it discrete

539: abelian group}, rather than as a vector space over $\bR$, so that it is

540: a very large group. Its {\it continuous part} is defined as

541: \begin{align*}

542: H^*_{ct}(H^1(\Sg;\bR)^\delta;\bR)=&\Hom_{\bR}(\La_{\bR}^*(H^1(\Sg;\bR)),\bR)\\

543: \subset &\Hom_{\bZ}(\La_{\bR}^*(H^1(\Sg;\bR)),\bR)\\

544: \subset &\Hom_{\bZ}(\La_{\bZ}^*(H^1(\Sg;\bR)),\bR)\cong

545: H^*(H^1(\Sg;\bR)^\delta;\bR) \ ,

546: \end{align*}

547: where the second inclusion is induced by the natural projection

548: $$

549: \La_{\bZ}^*(H^1(\Sg;\bR))\longrightarrow \La_{\bR}^*(H^1(\Sg;\bR)) \ .

550: $$

551: Denoting by $T^{2g}=K(\pi_1\Sg,1)$ the Jacobian torus of $\Sg$, there is

552: a canonical isomorphism

553: $$

554: \La^*_{\bR}H^1(\Sg;\bR)\cong H^*(T^{2g};\bR) \ ,

555: $$

556: so that we can identify

557: $$

558: H^*_{ct}(H^1(\Sg;\bR)^\delta;\bR)\cong \Hom_\bR(H^*(T^{2g};\bR);\bR)

559: \cong H_*(T^{2g};\bR) \ .

560: $$

561: Thus, by restricting the homomorphism $\Flux^{*}$ in~\eqref{eq:flux*}

562: to the continuous cohomology, we obtain a homomorphism

563: $$

564: \Flux^*\colon H_*(T^{2g};\bR)\longrightarrow H^*(\Symp_0^\delta\Sg;\bR) \ .

565: $$

566: It is easy to see that under this homomorphism we have

567: $$

568: \tilde x_i=\Flux^*(x_i) \ ,\quad \tilde y_i=\Flux^*(y_i) \ .

569: $$

570: Let $\om_0\in \La^2_{\bR}H_1(\Sg;\bR)$ be the symplectic class

571: defined by

572: $$

573: \om_0=\sum_{i=1}^g x_i\land y_i

574: $$

575: and set

576: $$

577: \tilde\om_0=\Flux^*(\om_0)

578: =\sum_{i=1}^{g} \tilde x_i \tilde y_i\in H^2(\Symp_0^\delta\Sg;\bR) \ .

579: $$

580: 

581: \begin{lemma}\label{lem:flux2}

582: We have the equality

583: $$

584: [\Flux]^2=-2\mu\otimes \tilde\om_0\in

585: H^2(\Sg;\bR)\otimes H^2(\Symp_0^\delta\Sg;\bR) \ .

586: $$

587: \end{lemma}

588: \begin{proof}

589: A direct calculation using the expression~\eqref{eqn:flux} yields

590: $$

591: [\Flux]^2=-\sum_{i=1}^{g} (x_i^* y_i^*\otimes\tilde x_i\tilde y_i

592: +y_i^* x_i^*\otimes \tilde y_i\tilde x_i) \ .

593: $$

594: Since $x_i^*y_i^*=-y_i^*x_i^*=\mu$, we obtain

595: %\begin{align*}

596: $$

597: [\Flux]^2=-2\mu\otimes\sum_{i=1}^{g} \tilde x_i\tilde y_i

598: =-2\mu\otimes\tilde\om_0

599: $$

600: %\end{align*}

601: as required.

602: \end{proof}

603: 

604: Now we can completely determine the transverse symplectic class $v$ of

605: foliated $\Sg$-bundles whose total holonomy groups are contained

606: in the identity component $\Symp_0\Sg$ of $\Symp\Sg$ as follows.

607: \begin{proposition}\label{prop:v}

608: On the subgroup $\ES_0^\delta\Sg$ the transverse symplectic class

609: $v\in H^2(\ES_0^\delta\Sg;\bR)$ is given by

610: $$

611: v=(2g-2)\mu+[\Flux]+\frac{1}{2g-2}\tilde\om_0

612: $$

613: under the isomorphism \eqref{eq:kunneth}.

614: Furthermore, the homomorphism

615: $$

616: \tilde\om_0\otimes H^1_{ct}(\Symp_0^\delta\Sg;\bR)\lra

617: H^3(\Symp_0^\delta\Sg;\bR)

618: $$

619: induced by the cup product is trivial, where

620: $H^1_{ct}(\Symp_0^\delta\Sg;\bR)\cong H_1(\Sg;\bR)$

621: denotes the subgroup of $H^1(\Symp_0^\delta\Sg;\bR)$

622: generated by the continuous cohomology classes

623: $\tilde x_i, \tilde y_i$.

624: In particular, $\tilde\om_0^2=0$.

625: \end{proposition}

626: \begin{proof}

627: Since the restriction of $v$ to each fiber is equal to the negative of

628: that of the Euler class $e$ by our normalization, $v$ restricts to

629: $(2g-2)\mu$ on the fiber. This gives the first component of the formula.

630: The second component follows from Lemma~8 of~\cite{KM03}.

631: Thus we can write

632: $$

633: v=(2g-2)\mu+[\Flux]+ \ga \in H^*(\Sg;\bR)\otimes H^*(\Symp_0^\delta\Sg;\bR)

634: $$

635: for some $\ga\in H^2(\Symp_0^\delta\Sg;\bR)$.

636: Now observe that $v^2=0$ because $\tilde\om^2=0$.

637: Also, because we have restricted to $\Symp_{0}\Sg$, we have

638: $\mu^2=0$, $\mu [\Flux]=0$. Hence we obtain

639: $$

640: [\Flux]^2+\ga^2+ 2(2g-2)\mu \ga+2 [\Flux] \ga=0 \ .

641: $$

642: It follows that

643: \begin{align*}

644: [\Flux]^2+2(2g-2)\mu \ga&=0\\

645: [\Flux] \ga&=0\\

646: \ga^2&=0

647: \end{align*}

648: because these three elements belong to different summands in the

649: K\"unneth decomposition

650: of $H^*(\Sg;\bR)\otimes H^*(\Symp_0^\delta;\bR)$.

651: If we combine the first equality above and Lemma~\ref{lem:flux2},

652: then we can conclude that

653: \begin{equation}\label{eqn:ga}

654: \ga=\frac{1}{2g-2}\tilde\om_0 \ .

655: \end{equation}

656: This proves the first claim of the proposition.

657: If we substitute~\eqref{eqn:flux} and~\eqref{eqn:ga} in

658: the second equality above, then we see that

659: $\tilde\om_0\tilde x_i=\tilde\om_0\tilde y_i=0$ for any $i$,

660: whence the second claim.

661: Observe that the third equality $\ga^2=0$, which is equivalent to

662: $\tilde\om_0^2=0$ by the above, is a consequence of the second claim.

663: \end{proof}

664: 

665: Now we can calculate the restriction of the cocycle $\alpha$ defined

666: in Section~\ref{s:main} to the identity component $\Symp_{0}\Sg$.

667: \begin{proposition}\label{prop:alom}

668: Let $i\colon\Symp_0^\delta\Sg\ra\Symp^\delta\Sg$ be the inclusion.

669: Then %we have

670: $$

671: i^*\al = 2 \tilde\om_0\in H^2(\Symp_0^\delta\Sg;\bR)

672: $$

673: and

674: $$

675: i^*\al^2=0 \ .

676: $$

677: \end{proposition}

678: \begin{proof}

679: Let $\varph, \psi\in \Symp_0\Sg$ be any two elements. Then, by the

680: definition of $\al$, see Definition~\ref{def:al} and the subsequent

681: discussion, we have

682: $$

683: \al(\varph,\psi)=\iota(\Flux(\varph), \Flux(\psi)) \ .

684: $$

685: By the definition of the cohomology classes $\tilde x_i, \tilde y_i$, we have

686: $$

687: \Flux(\varph)=\sum_{i=1}^{g}

688: (\tilde x_i(\varph) x_i+\tilde y_i(\varph) y_i)\ ,\quad

689: \Flux(\psi)=\sum_{i=1}^{g}

690: (\tilde x_i(\psi) x_i+\tilde y_i(\psi) y_i)\ .

691: $$

692: Hence we obtain

693: $$

694: \al(\varph,\psi)=\sum_{i=1}^g \left\{\tilde x_i(\varph) \tilde y_i(\psi)

695: -\tilde y_i(\varph) \tilde x_i(\psi)\right\}\ .

696: $$

697: Using the Alexander--Whitney cup product, we also have

698: $$

699: \tilde\om_0(\varph,\psi)=\sum_{i=1}^g \tilde x_i(\varph) \tilde y_i(\psi)

700: =-\sum_{i=1}^g \tilde x_i(\psi) \tilde y_i(\varph) \ .

701: $$

702: Thus $\al=2 \tilde\om_0$ as required. The last statement follows from

703: Proposition~\ref{prop:v}.

704: \end{proof}

705: We see from this proposition that $i^*\al^2=0$ in

706: $H^4(\Symp_0^\delta\Sg;\bR)$, while we will prove the

707: non-triviality of $\al^2\in H^4(\Symp^\delta\Sg;\bR)$, which is a special

708: case of Theorem~\ref{th:hev}.

709: One could say that the non-triviality of $\al^2$ is realized

710: by an interaction of the two groups $\Symp_0\Sg$ and $\Mg$.

711: 

712: %\section{The second homology group of $\Symp^\delta\Sg$}\label{s:second}

713: 

714: %In this section, we prove Theorem \ref{th:hev} for the special

715: %case $k=1$. In fact we prove the following slightly stronger statement.

716: %\begin{theorem}\label{th:h2}

717: %For any $g\geq 3$ the characteristic classes $e_1$ and $\tilde \al$ induce

718: %a surjective homomorphism

719: %$$

720: %H_2(\Symp^\delta\Sg;\bZ)\lra \bZ\oplus S^2_\bQ\bR \ .

721: %$$

722: %For $g=2$, $\tilde \al$ induces a surjection

723: %$$

724: %H_2(\Symp^\delta\Sigma_2;\bZ)\ra S^2_\bQ\bR \ .

725: %$$

726: %\end{theorem}

727: 

728: %Recall that we have the following two group extensions

729: %\begin{equation}\label{eq:hss}

730: %1\lra\Ham\Sg\lra\Symp_0\Sg\ \overset{\Flux}{\lra}\ H^1(\Sg;\bR)\lra 1 \ ,

731: %\end{equation}

732: %\begin{equation}\label{eq:ssm}

733: %1\lra\Symp_0\Sg\lra\Symp\Sg\lra\Mg\lra 1 \ .

734: %\end{equation}

735: %\begin{proposition}\label{prop:fl2}

736: %The homomorphism

737: %$$

738: %\Flux_*\colon H_2(\Symp_0^\delta\Sg;\bZ)\lra H_2(H^1(\Sg;\bR)^\delta;\bZ)

739: %\cong \La^2_{\bZ}H^1(\Sg;\bR)

740: %$$

741: %induced by the flux homomorphism is surjective for any $g\geq 2$.

742: %\end{proposition}

743: %\begin{proof}

744: %We consider the Hochschild-Serre exact sequence

745: %\begin{align*}

746: %H_2&(\Symp_0^\delta\Sg;\bZ)\ \overset{\Flux_*}{\lra}\

747: %H_2(H^1(\Sg;\bR)^\delta;\bZ)\lra H_1(\Ham^\delta\Sg;\bZ)_{H^1_{\bR}}\\

748: %&\lra H_1(\Symp_0^\delta\Sg;\bZ)\lra

749: %H_1(H^1(\Sg;\bR)^\delta;\bZ)\lra 0

750: %\end{align*}

751: %of the group extension~\eqref{eq:hss}. It is known that the group $\Ham\Sg$

752: %is perfect (see Banyaga~\cite{Banyaga}). Hence $H_1(\Ham^\delta\Sg;\bZ)=0$.

753: %The result follows.

754: %\end{proof}

755: 

756: %\begin{proof}[Proof of Theorem~\ref{th:h2}]

757: %First of all, we proved in~\cite{KM03} that for $g\geq 3$ the

758: %first Mumford-Morita-Miller class $e_1$, more precisely $\frac{1}{12}

759: %e_1$, induces a surjection $H_2(\Symp^\delta\Sg;\bZ)\ra \bZ$.

760: 

761: %Next we consider the group extension~\eqref{eq:ssm} and the spectral

762: %sequence $\{E^r_{p,q},d^r\}$ for its integral homology. Then we have

763: %$$

764: %E^2_{0,2}\cong H_0(\Mg;H_2(\Symp_0^\delta\Sg;\bZ))

765: %\cong H_2(\Symp_0^\delta\Sg;\bZ)_{\Mg}  \ .

766: %$$

767: %By Proposition~\ref{prop:fl2}, the homomorphism between the

768: %spaces of $\Mg$-coinvariants

769: %$$

770: %H_2(\Symp_0^\delta\Sg;\bZ)_{\Mg}\lra H_2(H^1(\Sg;\bR)^\delta;\bZ)_{\Mg}

771: %$$

772: %induced by the flux homomorphism is surjective.

773: %Since by Proposition~\ref{prop:la2} the target of this surjection can be

774: %identified with $S^2_\bQ\bR$, we obtain a surjection

775: %$$

776: %E^2_{0,2}\lra S^2_\bQ\bR \ .

777: %$$

778: %The restriction of the characteristic class $\tilde\al$ to the subgroup

779: %$\Symp_0\Sg\subset\Symp\Sg$ clearly detects this surjection.

780: %Hence it survives to the $E^\infty$-term and

781: %there exists a surjection

782: %$$

783: %H_2(\Symp^\delta\Sg;\bZ)\supset E^\infty_{0,2}\lra S^2_\bQ\bR \ .

784: %$$

785: %Again because $\tilde\al$ is defined on the whole group $\Symp\Sg$

786: %and not just on $\Symp_0\Sg$, we conclude that the homomorphism

787: %$H_2(\Symp^\delta\Sg;\bZ)\ra S^2_\bQ\bR$ is surjective.

788: 

789: %Finally, we have to prove that the two characteristic classes $e_1$

790: %and $\tilde \al$ are linearly independent. But this follows easily

791: %from the fact that the non-triviality of the $S^2_\bQ\bR$ factor

792: %can be realized on cycles whose supports lie in the subgroup $\Symp_0\Sg$,

793: %while the characteristic class $e_1$ is trivial on these cycles because

794: %topologically they are trivial $\Sg$-bundles. This completes the proof.

795: %\end{proof}

796: 

797: \section{Extended flux homomorphisms for open surfaces}\label{s:open}

798: 

799: We consider the open surface $\Sigma_g^0=\Sigma_g\setminus D$ obtained

800: from $\Sg$ by removing a closed embedded disk $D\subset\Sg$. Let

801: $j\colon\Sigma_g^0\ra \Sg$ be the inclusion. We denote by $\Symp^c\Sigma_g^0$

802: the symplectomorphism group of $(\Sigma_g^0,j^*\om)$ with {\it compact 

803: supports}.

804: Hence the group $\Symp^c\Sigma_g^0$ can be considered as a subgroup of 

805: $\Symp\Sg$

806: with inclusion $j\colon\Symp^c\Sigma_g^0\ra\Symp\Sg$. Let $\Symp_0^c\Sigma_g^0$

807: be the identity component of $\Symp^c\Sigma_g^0$. Clearly $\Symp_0^c\Sigma_g^0$

808: is a subgroup of $\Symp_0\Sg$.

809: 

810: Let $\M1g$ denote the mapping class group of $\Sg$ {\it relative to}

811: the embedded disk $D^2\subset\Sg$, equivalently, the mapping class

812: group of the compact surface $\overline{\Sigma_g^0}$ with boundary.

813: We have a natural homomorphism $p\colon\Symp^c\Sigma_g^0\ra\M1g$

814: which is easily seen to be surjective. Moreover, Moser's theorem~\cite{Moser}

815: adapted to the present case (see~\cite{Tsuboi} for a general statement),

816: implies that the kernel of this surjection is precisely the group

817: $\Symp_0^c\Sigma_g^0$. We summarize the situation in the following

818: diagram:

819: \begin{equation}

820: \begin{CD}

821: 1@>>> \Symp^{c}_{0}\Sigma_g^{0} @>{i^c}>>

822: \Symp^c\Sigma_g^{0}@>{p}>> \M1g @>>> 1\\

823: @. @V{j_0}VV @V{j}VV @V{q}VV  @.\\

824: 1@>>> \Symp_{0}\Sigma_g @>{i}>> \Symp\Sigma_g @>>> \Mg @>>> 1

825: \end{CD}

826: \label{eqn:GG}

827: \end{equation}

828: where $q\colon\M1g\ra\Mg$ denotes the natural projection.

829: 

830: The restriction of the flux homomorphism

831: \begin{equation}\label{eqn:Flux}

832: \Flux\colon\Symp_0\Sg\lra H^1(\Sg;\bR)

833: \end{equation}

834: to the subgroup $\Symp_0^c\Sigma_g^0$, denoted $j^*\Flux$, can be described

835: as follows. The restriction $j^{*}\omega$ of the area form $\omega$

836: to the open surface $\Sigma_{g}^{0}$ is exact. Choose a $1$-form $\la$ %on 

837: $\Sigma_g^0$

838: such that $d\la=-j^*\om$. Then, for any element $\varph\in\Symp_0^c\Sigma_g^0$,

839: the $1$-form $\la-\varph^*\la$ is a closed form with compact support.

840: Hence the corresponding de Rham cohomology class $[\la-\varph^*\la]$,

841: which can be shown to be independent of the choice of $\la$,

842: is an element of the first cohomology group $H^1_c(\Sigma_g^0;\bR)$

843: of $\Sigma_g^0$ with compact support. It is easy to see that

844: $H^1_c(\Sigma_g^0;\bR)$ is canonically isomorphic to $H^1(\Sg;\bR)$,

845: and that under this isomorphism

846: \begin{equation}\label{eq:Fluxc}

847: (j^*\Flux)(\varph)=[\la-\varph^*\la]\in H^1_c(\Sigma_g^0;\bR)

848: \cong H^1(\Sg;\bR) %\quad (\varph\in\Symp_0^c\Sigma_g^0)

849: \end{equation}

850: for all $\varph\in\Symp_0^c\Sigma_g^0$, see Lemma~10.14 of~\cite{MS}.

851: We obtain the following commutative diagram:

852: \begin{equation}

853: \begin{CD}

854: \Symp^{c}_{0}\Sigma_g^{0} @>{j^*\Flux}>> H^1_{c}(\Sigma_g^{0};\bR)\\

855: @VVV @VV{\cong}V\\

856: \Symp_{0}\Sigma_g @>>{\Flux}> H^1(\Sg;\bR).

857: \end{CD}

858: \label{eqn:FFlux}

859: \end{equation}

860: From now on we identify $H^1_{c}(\Sigma_g^{0};\bR)$ with $H^1(\Sg;\bR)$.

861: 

862: As was already mentioned in the Introduction, we proved in~\cite{KM03} that

863: the flux homomorphism~\eqref{eqn:Flux} can be extended to a crossed 

864: homomorphism

865: \begin{equation}\label{eqn:TFlux}

866: \widetilde{\Flux}\colon\Symp\Sg\lra H^1(\Sg;\bR) \ ,

867: \end{equation}

868: and that the extension is unique up to the addition of coboundaries.

869: The restriction of such a crossed homomorphism to the subgroup

870: $\Symp^c\Sigma_g^0\subset \Symp\Sg$, denoted $j^*\widetilde{\Flux}$,

871: is of course an extension of the flux homomorphism $j^*\Flux$.

872: However, we also have another extension of the same flux homomorphism

873: $j^*\Flux$ to the group $\Symp^c\Sigma_g^0$ as follows.

874: \begin{proposition}\label{prop:Fluxb}

875: The map

876: $$

877: \widetilde{\Flux}_c\colon\Symp^c\Sigma_g^0\lra H^1_c(\Sigma_g^0;\bR)

878: \cong H^1(\Sg;\bR) %\quad (\varph\in\Symp^c\Sigma_g^0)

879: $$

880: defined by

881: $\widetilde{\Flux}_c(\varph)=[(\varph\inv)^*\la-\la]\in H^1_c(\Sigma_g^0;\bR)$

882: is a crossed homomorphism which extends the flux homomorphism $j^*\Flux$.

883: Its cohomology class

884: $[\widetilde{\Flux}_c]\in H^1(\Symp^c\Sigma_g^0;H^1(\Sg;\bR))$

885: is uniquely determined independently of the choice of the

886: $1$-form $\la$ such that $d\la=-j^*\om$.

887: \end{proposition}

888: \begin{proof}

889: First observe that, for any $\varph\in \Symp_0^c\Sigma_g^0$, we have

890: $$

891: [\la-\varph^*\la]=(\varph\inv)^*[\la-\varph^*\la]=[(\varph\inv)^*\la-\la]

892: $$

893: because $\varph$ acts trivially on $H^1(\Sg;\bR)$. Hence

894: by~\eqref{eq:Fluxc} we have

895: $$

896: \widetilde{\Flux}_c(\varph)=(j^*\Flux)(\varph) \ .

897: $$

898: Next, for any two elements $\varph, \psi\in \Symp^c\Sigma_g^0$, we have

899: \begin{align*}

900: \widetilde{\Flux}_c(\varph\psi)&=[((\varph\psi)\inv)^*\la-\la]\\

901: &=[(\varph\inv)^*\la-\la +(\varph\inv)^*(\psi\inv)^*\la-(\varph\inv)^*\la]\\

902: &=[(\varph\inv)^*\la-\la]+[(\varph\inv)^*((\psi\inv)^*\la-\la)]\\

903: &=\widetilde{\Flux}_c(\varph)+(\varph\inv)^*\widetilde{\Flux}_c(\psi).

904: \end{align*}

905: Therefore $\widetilde{\Flux}_c$ is a crossed homomorphism which extends 

906: $j^*\Flux$.

907: 

908: Finally, let $\la'$ be another $1$-form on $\Sigma_g^0$

909: satisfying $d\la'=-j^*\om$, and let $\widetilde{\Flux}'_c$ be the corresponding

910: crossed homomorphism. Then $a=\la'-\la$ is a closed $1$-form defining a de

911: Rham cohomology class $[a]\in H^1(\Sigma_g^0;\bR)\cong H^1(\Sg;\bR)$.

912: Now

913: \begin{align*}

914: \widetilde{\Flux}'_c(\varph)&=[(\varph\inv)^*\la'-\la']\\

915: &=[(\varph\inv)^*(\la+a)-(\la+a)]\\

916: &=[(\varph\inv)^*\la-\la]+[(\varph\inv)^*a-a]\\

917: &=\widetilde{\Flux}_c(\varph)+(\varph\inv)^*[a]-[a] \in H^1(\Sg;\bR)

918: \ .

919: \end{align*}

920: This shows that the difference $\widetilde{\Flux}'_c-\widetilde{\Flux}_c$

921: is a coboundary, completing the proof of the proposition.

922: \end{proof}

923: 

924: We have proved that the restriction $j^*\Flux$ of the flux

925: homomorphism~\eqref{eqn:Flux} to the subgroup $\Symp_0^c\Sigma_g^0$ has

926: two extensions to the group $\Symp^c\Sigma_g^0$ as a crossed homomorphism.

927: One is the restriction $j^*\widetilde{\Flux}$ of $\widetilde{\Flux}$, and

928: the other is $\widetilde{\Flux}_c$. We will show that these two crossed

929: homomorphisms are essentially different. More precisely, we will show that

930: the difference of these two crossed homomorphisms can be expressed by an

931: element of the cohomology group

932: \begin{equation}\label{eqn:hm1g}

933: H^1(\M1g;H^1(\Sg;\bR)) \ .

934: \end{equation}

935: It was proved in~\cite{Morita89} that $H^1(\M1g;H^1(\Sg;\bZ))$ is

936: isomorphic to $\bZ$ for all $g\geq 2$. It follows that the above

937: group~\eqref{eqn:hm1g} is isomorphic to $\bR$ and if

938: $$

939: k\colon \M1g \lra  H^1(\Sg;\bZ)

940: $$

941: is any crossed homomorphism whose cohomology class is a

942: generator of $H^1(\M1g;H^1(\Sg;\bZ))$, then the associated

943: crossed homomorphism

944: $$

945: k_\bR\colon\M1g \lra  H^1(\Sg;\bR)

946: $$

947: represents the element $1\in H^1(\M1g;H^1(\Sg;\bR))\cong\bR$.

948: Let $p^*k_\bR\in H^1(\Symp^c\Sg;H^1(\Sg;\bR))$ be the

949: class induced from $k_\bR$ by the projection $p\colon\Symp^c\Sg\ra\M1g$.

950: 

951: \begin{theorem}\label{th:h1sc}

952: For any $g\geq 2$, there exists an isomorphism

953: $$

954: H^1(\Symp^c\Sigma_g^0;H^1(\Sg;\bR))\cong \bR\oplus\Hom_{\bZ}(\bR,\bR)

955: $$

956: such that $p^*k_\bR$ and $[j^*\widetilde{\Flux}]$ represent

957: the classes $1\in \bR$ and $\id\in \Hom_{\bZ}(\bR,\bR)$ respectively.

958: \end{theorem}

959: \begin{proof}

960: The top extension in~\eqref{eqn:GG} gives rise to an exact sequence

961: \begin{align*}

962: & 0\ \lra \ H^1(\M1g;H^1(\Sg;\bR))\ \cong\bR\ \overset{p^*}{\lra} \\

963:    H^1(&\Symp^c\Sg;H^1(\Sg;\bR))\lra

964: H^1(\Symp^c_0\Sg;H^1(\Sg;\bR))^{\M1g}\lra\cdots .

965: \end{align*}

966: It was proved in~\cite{KM03} that $j^*\Flux$ induces an isomorphism

967: $$

968: j^*\Flux\colon H_1(\Symp^c_0\Sigma_g^0;\bZ)\cong

969: H^1_c(\Sigma_g^0;\bR)=H^1(\Sg;\bR) \ .

970: $$

971: Hence, by an argument similar to  the one given for $\Symp_0\Sg$ 

972: in~\cite{KM03},

973: we have an isomorphism

974: $$

975: H^1(\Symp^c_0\Sigma_g^0;H^1(\Sg;\bR))^{\M1g}\cong \Hom_\bZ(\bR,\bR) \ ,

976: $$

977: and clearly $j^*[\widetilde{\Flux}]$ corresponds to $\id\in \Hom_\bZ(\bR,\bR)$.

978: The result follows from this.

979: \end{proof}

980: 

981: \begin{theorem}\label{th:ffk}

982: We have the identity

983: $$

984: [\widetilde{\Flux}_c]=j^*[\widetilde{\Flux}]- p^*k_\bR

985: $$

986: in $H^1(\Symp^c\Sigma_g^0;H^1(\Sg;\bR))$.

987: \end{theorem}

988: \begin{proof}

989: Since both crossed homomorphisms $\widetilde{\Flux}_c$ and 

990: $j^*\widetilde{\Flux}$

991: are extensions of the flux homomorphism $j^*\Flux$, the proof of

992: Theorem~\ref{th:h1sc} implies that

993: $$

994: [\widetilde{\Flux}_c]=j^*[\widetilde{\Flux}]+a\ p^*k_\bR

995: $$

996: for some constant $a\in\bR$. Let $\I1g\subset\M1g$ denote the Torelli

997: subgroup consisting of mapping classes which act trivially on homology.

998: We set ${\mathcal I}\Symp^c\Sigma_g^0=p^{-1}(\I1g)\subset\Symp^c\Sigma_g^0$.

999: If we restrict the crossed homomorphims $\widetilde{\Flux}_c$ and

1000: $\widetilde{\Flux}$ to this subgroup ${\mathcal I}\Symp^c\Sigma_g^0$, then

1001: they become homomorphisms which depend only on the cohomology classes

1002: $[\widetilde{\Flux}_c]$ and $j^*[\widetilde{\Flux}]$, and not on the

1003: particular crossed homomorphisms representing these cohomology classes.

1004: This is because any crossed homomorphism which is a coboundary is trivial

1005: on ${\mathcal I}\Symp^c\Sigma_g^0$. It was proved in~\cite{Morita89}

1006: that a generator of the group $H^1(\M1g;H^1(\Sg;\bZ))\cong \bZ$ is

1007: characterized by the fact that the {\it Poincar\'e dual} of its value on

1008: a single non-trivial element $\varph \in \I1g$ is equal to $\pm C 

1009: \tau(\varph)$,

1010: where $\tau\colon\I1g\ra \La^3 H_1(\Sg;\bZ)$ denotes the

1011: (first) Johnson homomorphism and $C\colon\La^3 H_1(\Sg;\bZ)\ra H_1(\Sg;\bZ)$

1012: denotes the contraction. Hence we only have to compute

1013: the values of $\widetilde{\Flux}_c$ and $\widetilde{\Flux}$

1014: on some particular element $\tilde\varph\in {\mathcal I}\Symp^c\Sigma_g^0$.

1015: We choose such element as follows.

1016: 

1017: \begin{figure}

1018: \epsfig{file=figure1.eps,width=12cm}

1019: %\epsfig{file=figure1.pdf,width=12cm}

1020: \caption{ }\label{fig:torelli}

1021: \end{figure}

1022: 

1023: Consider an embedded disk $D'$ in $\Sigma_g^0=\Sg\setminus D$ as depicted

1024: in Figure~\ref{fig:torelli} and set $\Sigma_g^{00}=\Sg\setminus (D\cup D')$.

1025: We also consider two disjoint simple closed curves $C_1, C_2\subset 

1026: \Sigma_g^{00}$

1027: as shown in Figure~\ref{fig:torelli}.

1028: Let $\tau_i\ (i=1,2)$ denote the Dehn twist along $C_i$

1029: and choose lifts $\tilde\tau_i\in\Symp^c\Sigma_g^0$ of $\tau_i$ with

1030: supports in small neighborhoods of the $C_{i}$.

1031: Now we set

1032: $$

1033: \tilde\varph=\tilde\tau_1 \tilde\tau_2^{-1} \ .

1034: $$

1035: Since $\varph=\tau_1\tau_2^{-1}$ belongs to the Torelli group $\I1g\subset 

1036: \M1g$,

1037: its lift $\tilde\varph$ is an element of the subgroup

1038: ${\mathcal I}\Symp^c\Sigma_g^0$. We would like to compute the difference

1039: \begin{equation}\label{eq:difference}

1040: \begin{split}

1041: \widetilde{\Flux}_c(\tilde\varph)-\widetilde{\Flux}(\tilde\varph)

1042: &= a\ p^*k_\bR(\tilde\varph)\\

1043: &=a\ k(\varph)\in H^1_c(\Sigma_g^0;\bR)\cong H^1(\Sg;\bR) \ .

1044: \end{split}

1045: \end{equation}

1046: 

1047: Now we consider the long exact sequence

1048: \begin{equation}\label{eq:s0}

1049: \begin{split}

1050: 0\lra H^1(\Sg;\bR)&\overset{i^*}{\lra}H^1(\Sigma_g^{00};\bR)

1051: \lra \\

1052: & H^2(\Sg,\Sigma_g^{00};\bR)\cong\bR^2\lra

1053: H^2(\Sg;\bR)\lra 0

1054: \end{split}

1055: \end{equation}

1056: of the pair $(\Sg, \Sigma_g^{00})$.

1057: Also consider the following short exact sequences

1058: \begin{equation}\label{eq:s0'}

1059: \begin{split}

1060: 0\lra H^1_c(\Sigma_g^0;\bR)\cong H^1(\Sg;\bR)&\overset{j^*}{\lra}

1061: H^1(\Sigma_g^{00};\bR)\lra \bR\lra 0\\

1062: 0\lra H^1_c(\Sigma_g\setminus D';\bR)\cong H^1(\Sg;\bR)&\overset{(j')^*}{\lra}

1063: H^1(\Sigma_g^{00};\bR)\lra \bR\lra 0 \ .

1064: \end{split}

1065: \end{equation}

1066: In view of the above exact sequences~\eqref{eq:s0} and~\eqref{eq:s0'},

1067: we can determine the value of~\eqref{eq:difference} in the

1068: group $H^1(\Sigma_g^{00};\bR)$ because both groups

1069: $H^1_c(\Sigma_g^0;\bR)$ and $H^1(\Sg;\bR)$ are embedded in it.

1070: Choose $1$-forms

1071: $\la$ and $\la'$ on $\Sigma_g^0=\Sg\setminus D$ and

1072: $\Sg\setminus D'$ respectively, such that

1073: $$

1074: d\la=-j^*\om,\quad d\la'=-(j')^*\om

1075: $$

1076: where $j\colon\Sigma_g^0\ra\Sg$ and $j'\colon\Sg\setminus D'\ra\Sg$ are

1077: the inclusions. Then $d(\la-\la')=0$ so that $\nu=\la-\la'$ is a

1078: closed $1$-form on $\Sigma_g^{00}$. We have the identity

1079: \begin{align*}

1080: (\tilde\varph\inv)^*\la-\la&=(\tilde\varph\inv)^*(\la'+\nu)-(\la'+\nu)\\

1081: &=(\tilde\varph\inv)^*\la'-\la'+(\tilde\varph\inv)^*\nu-\nu \ .

1082: \end{align*}

1083: Hence we have the equality

1084: \begin{equation}\label{eq:[]}

1085: [(\tilde\varph\inv)^*\la-\la]=[(\tilde\varph\inv)^*\la'-\la']+

1086: [(\tilde\varph\inv)^*\nu-\nu]

1087: \end{equation}

1088: in the group $H^1(\Sigma_g^{00};\bR)$.

1089: By the definition of $\widetilde{\Flux}_c$, we have

1090: \begin{equation}\label{eq:fla}

1091: \widetilde\Flux_c(\tilde\varph)=[(\tilde\varph\inv)^*\la-\la]\in

1092: H^1_c(\Sigma_g^0;\bR)\subset H^1(\Sigma_g^{00};\bR) \ .

1093: \end{equation}

1094: Next we compute $\widetilde{\Flux}(\tilde\varph)$. For this, observe

1095: that $\tilde\varph$ is isotopic to the identity as an element of

1096: $\Symp^c(\Sg\setminus D')\subset \Symp\Sg$, although it is {\it not}

1097: isotopic to the identity as an element of $\Symp^c(\Sg\setminus D)$.

1098: Hence $\tilde\varph\in \Symp^c_0(\Sg\setminus D')$ and

1099: $\widetilde{\Flux}(\tilde\varph)=((j')^*\Flux)(\tilde\varph)$.

1100: By replacing $D$ with $D'$ in equation~\eqref{eq:Fluxc}, we obtain

1101: \begin{equation}\label{eq:Fluxc'}

1102: \begin{split}

1103: ((j')^*\Flux) &(\tilde\varph)=[\la'-\tilde\varph^*\la']\\

1104: = &[(\tilde\varph\inv)^*\la'-\la']\in

1105: H^1_c(\Sigma_g\setminus D';\bR)\subset H^1(\Sigma_g^{00};\bR) \ .

1106: \end{split}

1107: \end{equation}

1108: By combining the equations~\eqref{eq:difference}, \eqref{eq:[]},

1109: \eqref{eq:fla} and~\eqref{eq:Fluxc'}, we shall prove

1110: \begin{equation}\label{eq:nutau}

1111: [(\tilde\varph\inv)^*\nu-\nu]=

1112: a k(\tilde\varph)=a k(\varph)=a PD\circ C \tau(\varph) \ ,

1113: \end{equation}

1114: where $PD$ denotes the Poincar\'e duality isomorphism.

1115: 

1116: Clearly $(\tilde\varph\inv)^*\nu-\nu$ is a closed $1$-form on $\Sg$ whose

1117: support is contained in a neighborhood of $C_1\cup C_2$.

1118: Hence the Poincar\'e dual of the de Rham cohomology class

1119: $[(\tilde\varph\inv)^*\nu-\nu]\in H^1(\Sg;\bR)$ is a

1120: multiple of the homology class

1121: $[C_1]\in H_1(\Sg;\bZ)$ represented by the simple closed curve

1122: $C_1$ with a fixed orientation depicted in Figure~\ref{fig:torelli}.

1123: However, according to Johnson~\cite{Johnson80}, we have

1124: $$

1125: \tau(\varph)=(x_1\land y_1+\ldots +x_{g-1}\land y_{g-1})\land [C_1]

1126: $$

1127: where $x_1. y_1,\ldots ,x_{g-1}, y_{g-1}$ is a symplectic basis

1128: of the homology group of the left hand subsurface of $\Sg$

1129: obtained by cutting $\Sg$ along the simple closed curve $F_2$

1130: depicted in Figure~\ref{fig:torelli}. It follows that

1131: $$

1132: C \tau(\varph)=2(g-1)[C_1] \ .

1133: $$

1134: This checks the equality~\eqref{eq:nutau}, for some constant $a$.

1135: 

1136: To determine this constant, it is enough to compute the

1137: value of the cohomology class $[(\tilde\varph\inv)^*\nu-\nu]$

1138: on the homology class represented by the oriented simple

1139: closed curve $E$ depicted in Figure~\ref{fig:torelli}.

1140: Observe that the homology class

1141: $\tilde\varph\inv_*[E]-[E]\in H_1(\Sigma_g^{00};\bZ)$

1142: can be represented by the oriented simple closed curve

1143: $F_1$ also depicted in Figure~\ref{fig:torelli}.

1144: Let $\widetilde D$ denote the right hand compact subsurface

1145: of $\Sg$ obtained by cutting along $F_1$.

1146: Thus $\widetilde D$ is diffeomorphic to a disk which contains

1147: the original embedded disk $D$ in its interior.

1148: Also let $\Sigma =\Sg\setminus\textrm{Int}\tilde D$.

1149: 

1150: Now we compute

1151: \begin{align*}

1152: [(\tilde\varph\inv)_*\nu-\nu]([E])&=\nu((\tilde\varph\inv)_*[E])-\nu([E])\\

1153: &=\nu([F_1])

1154: =\int_{F_1} \la-\la'\\

1155: &=\int_{\partial \Sigma}\la-\int_{-\partial \widetilde D}\la'\\

1156: &=\int_{\Sigma} d\la +\int_{\widetilde D}d\la'\\

1157: &=\int_{\Sg} -\om =2-2g \ .

1158: \end{align*}

1159: Thus we can conclude that $a=-1$, completing the proof.

1160: \end{proof}

1161: 

1162: \section{The second homology group of

1163: $\Symp^{c,\delta}\Sigma_g^0$}\label{s:secondopen}

1164: 

1165: In this section, we generalize the claim of Theorem~\ref{th:hev}

1166: about the second homology, i.~e.~for $k=1$, to the case of the

1167: open surface $\Sigma_g^0$. We have the cohomology class

1168: $$

1169: j^*\tilde \al\in H^2(\Symp^{c,\delta}\Sigma_g^0;S^2_\bQ\bR)

1170: $$

1171: induced by the inclusion $j\colon\Sigma_g^0\ra\Sg$.

1172: \begin{theorem}\label{th:h20}

1173: The characteristic classes $e_1$ and $j^*\tilde \al$ induce a surjective

1174: homomorphism

1175: $$

1176: H_2(\Symp^{c,\delta}\Sigma_g^0;\bZ)\lra \bZ\oplus S^2_\bQ\bR

1177: $$

1178: for any $g\geq 3$. For $g=2$, the class $j^*\tilde \al$ induces

1179: a surjection

1180: $$

1181: H_2(\Symp^{c,\delta}\Sigma_2^0;\bZ)\lra S^2_\bQ\bR \ .

1182: $$

1183: \end{theorem}

1184: The proof of this theorem, which occupies the rest of this section,

1185: consists of a rather long and delicate argument. We first describe the

1186: reason why the easy proof of Theorem~\ref{th:hev} in the case $k=1$,

1187: which treated the case of

1188: closed surfaces, does not work for Theorem~\ref{th:h20}, thereby

1189: making the difficulty in the case of open surfaces explicit.

1190: 

1191: Consider the top extension

1192: \begin{equation}\label{eq:scm}

1193: 1 \lra \Symp^{c}_{0}\Sigma_g^{0} \overset{i^c}{\lra}

1194: \Symp^c\Sigma_g^{0} \overset{p}{\lra} \M1g \lra 1

1195: \end{equation}

1196: in the commutative diagram~\eqref{eqn:GG}. We have the following proposition,

1197: contrasting with our discussion of the closed case in Section~\ref{s:main}.

1198: \begin{proposition}\label{prop:fl2c}

1199: For any $g\geq 2$ the image of the homomorphism

1200: $$

1201: (j^*\Flux)_*\colon H_2(\Symp_0^{c,\delta}\Sigma_g^0;\bZ)\lra

1202: H_2(H^1(\Sg;\bR)^\delta;\bZ)\cong \La^2_{\bZ}H^1(\Sg;\bR)

1203: $$

1204: induced by the restriction $j^*\Flux$ of the flux homomorphism

1205: to the subgroup $\Symp_0^{c}\Sigma_g^0$

1206: is equal to the kernel of the natural intersection pairing

1207: $$

1208: \La^2_{\bZ}H^1(\Sg;\bR)\lra \bR \ .

1209: $$

1210: \end{proposition}

1211: \begin{proof}

1212: Recall that we have an extension

1213: \begin{equation}\label{eq:hcss}

1214: 1\lra\Ham^c\Sigma_g^0\lra\Symp_0^c\Sigma_g^0\ \overset{j^*\Flux}{\lra}\

1215: H^1(\Sg;\bR)\lra 1 \ ,

1216: \end{equation}

1217: where $\Ham^c\Sigma_g^0$ denotes the subgroup consisting of

1218: Hamiltonian symplectomorphisms with compact supports

1219: (see~\cite{MS} as well as~\cite{KM03}).

1220: Also recall that there is a surjective homomorphism

1221: $$

1222: {\rm Cal}\colon \Ham^c \Sigma_g^0 \lra\bR

1223: $$

1224: called the (second) Calabi homomorphism (see~\cite{Calabi}).

1225: Banyaga~\cite{Banyaga} proved that the kernel of this

1226: homomorphism is perfect. Hence we have an isomorphism

1227: $$

1228: H_1(\Ham^{c,\delta}\Sigma_g^0;\bZ)\cong \bR \ .

1229: $$

1230: Now we consider the Hochschild--Serre exact sequence

1231: \begin{align*}

1232: H_2(\Symp_0^{c,\delta}\Sigma_g^0;\bZ)&\ \overset{j^*\Flux_*}{\lra}\

1233: H_2(H^1(\Sg;\bR)^\delta;\bZ)\overset{\partial}{\lra}

1234: H_1(\Ham^{c,\delta}\Sigma_g^0;\bZ)_{H^1_{\bR}}\cong\bR\\

1235: &\lra H_1(\Symp_0^{c,\delta}\Sigma_g^0;\bZ)\

1236: \overset{j^*\Flux_*}{\lra}\

1237: H_1(H^1(\Sg;\bR)^\delta;\bZ)\lra 0

1238: \end{align*}

1239: of the group extension~\eqref{eq:hcss}. We proved in~\cite{KM03}

1240: (Proposition~11 and Corollary~12) that the last homomorphism

1241: $j^*\Flux_*$ in the above sequence is an isomorphism, and that

1242: the boundary operator $\partial$ coincides with the intersection

1243: pairing. The result follows.

1244: \end{proof}

1245: 

1246: \begin{corollary}\label{cor:trivial}

1247: The restriction of $\al\in H^2(\Symp^\delta\Sg;\bR)$ to the subgroup

1248: $\Symp_0^{c,\delta}\Sigma_g^0$ is trivial.

1249: \end{corollary}

1250: \begin{proof}

1251: This follows from Proposition~\ref{prop:fl2c} and the definition of

1252: the cohomology class $\al$.

1253: \end{proof}

1254: Thus, in order to prove the non-triviality of $\al$ on the group

1255: $\Symp^c\Sigma_g^0$, we must combine the roles of the two groups $\M1g$

1256: and $\Symp_0^{c,\delta}\Sigma_g^0$. This contrasts sharply

1257: with the case of closed surfaces treated in Section~\ref{s:main}.

1258: 

1259: Let $\{E_{p,q}^r\}$ be the Hochschild--Serre spectral sequence for the

1260: integral homology of the extension~\eqref{eq:scm}.

1261: This gives rise to two short exact sequences

1262: \begin{equation}\label{eq:ss}

1263: \begin{split}

1264: 0\lra \Ker& \lra H_2(\Symp^{c,\delta}\Sigma_g^0;\bZ)\lra

1265: E^\infty_{2,0} \lra 0\\

1266: & 0\lra E^\infty_{0,2}\lra\Ker \lra E^\infty_{1,1}\lra 0 \ ,

1267: \end{split}

1268: \end{equation}

1269: where $E^\infty_{2,0} \subset H_2(\M1g;\bZ)$ concerns the first

1270: Mumford-Morita-Miller class already discussed in~\cite{KM03}.

1271: Proposition~\ref{prop:fl2c} shows that the image of the map

1272: $$

1273: E^\infty_{0,2}=\im\left(H_2(\Symp_0^{c,\delta}\Sigma_g^0;\bZ)

1274: \ra H_2(\Symp^{c,\delta}\Sigma_g^0;\bZ)\right)\lra S^2_\bQ\bR

1275: $$

1276: defined by the Kronecker product with $\tilde\al$ is

1277: precisely the kernel of the natural map $S^2_\bQ\bR\ra\bR$.

1278: It remains to determine the $E^\infty_{1,1}$-term.

1279: We have

1280: \begin{equation}\label{eq:E1}

1281: E^2_{1,1}=H_1(\M1g;H_1(\Symp_0^{c,\delta}\Sigma_g^0;\bZ))

1282: \cong H_1(\M1g;H_1(\Sg;\bR))

1283: \end{equation}

1284: because, as was already mentioned above, it was proved in~\cite{KM03}

1285: that $j^*\Flux$ induces an isomorphism

1286: $H_1(\Symp_0^{c,\delta}\Sigma_g^0;\bZ)\cong H^1(\Sg;\bR)$.

1287: 

1288: We now recall a result which was essentially proved in~\cite{Morita89}.

1289: %Here we describe it because we need a precise form of the result.

1290: Without repeating everything done in~\cite{Morita89}, we  want to give

1291: a precise statement and proof of what is needed in the sequel.

1292: We use the Lickorish generators for the mapping class group $\M1g$,

1293: denoted $\la_i, \mu_i, \nu_i$ as in Figure~1 of~\cite{Morita89}, and a

1294: symplectic basis $x_1,\cdots,x_g,y_1,\cdots,y_g$ of $H_1(\Sg;\bZ)$.

1295: In particular

1296: \begin{equation}\label{eq:action}

1297: \nu_i(x_i)=x_i-y_i+y_{i+1},\quad \mu_i(x_i)=x_i-y_i

1298: \end{equation}

1299: which we record here for later use.

1300: \begin{proposition}\label{prop:h1m1}

1301: For any $g\geq 2$, we have an isomorphim

1302: $$

1303: H_1(\M1g;H_1(\Sg;\bZ))\cong \bZ \ .

1304: $$

1305: Furthermore, for any $i=1,2,\cdots,g-1$, the element

1306: $$

1307: c=\nu_i\otimes x_i-\mu_i\otimes x_i+\mu_{i+1}\otimes x_{i+1}

1308: $$

1309: is a $1$-cycle of $\M1g$ with twisted coefficients in $H_1(\Sg;\bZ)$

1310: and it represents a generator of the above infinite cyclic group.

1311: \end{proposition}

1312: \begin{proof}

1313: The group extension

1314: $$

1315: 1\lra \pi_1\Sg\lra \Msg\lra \Mg\lra 1 \ ,

1316: $$

1317: where $\Msg$ denotes the mapping class group of $\Sg$ {\it relative}

1318: to a basis point, yields the Hochschild--Serre exact sequence

1319: \begin{equation}\label{eq:msg}

1320: H_2(\Mg;H)\lra (H\otimes H)_{\Mg}\lra H_1(\Msg;H)

1321: \lra H_1(\Mg;H)\lra 0 \ .

1322: \end{equation}

1323: Here and henceforth $H$ is a shorthand for $H_1(\Sg;\bZ)$.

1324: It is easy to see that the intersection pairing induces an isomorphism

1325: $(H\otimes H)_{\Mg}\cong\bZ$ and the element $x_1\otimes y_1$, for

1326: example, represents a generator. It was proved in~\cite{Morita89}

1327: that

1328: $$

1329: H^1(\Msg;H^1(\Sg;\bZ))\cong Z,\quad

1330: H_1(\Mg;H^1(\Sg;\bZ))\cong \bZ/(2g-2)\bZ \ .

1331: $$

1332: If we apply the crossed homomorphism $f\colon\Msg\ra H^1(\Sg;\bZ)$

1333: given in the above cited paper, which detects a generator of the

1334: above infinite cyclic group, to the element $x_1\otimes y_1$

1335: considered as a cycle of $\Msg$ with coefficients in $H$, we obtain

1336: $$

1337: f(x_1)(y_1)=2-2g \ .

1338: $$

1339: On the other hand, $c$ is a cycle because

1340: $$

1341: \partial c=x_i-\nu_i(x_i)-x_i+\mu_i(x_i)+x_{i+1}-\mu_{i+1}(x_{i+1})

1342: =0

1343: $$

1344: by~\eqref{eq:action}, and it was shown that

1345: \begin{equation}\label{eq:k}

1346: f(c)=f(\nu_i)(x_i)-f(\mu_i)(x_i)+f(\mu_{i+1})(x_{i+1})=1 \ ,

1347: \end{equation}

1348: see~\cite{Morita89} for details. In view of the exact sequence~\eqref{eq:msg},

1349: we can conclude that $H_1(\Msg;H)\cong \bZ$. Finally, it is easy to deduce

1350: from the central group extension $0\ra \bZ\ra \M1g\ra\Msg\ra 1$ that we have an

1351: isomorphism $H_1(\M1g;H)\cong H_1(\Msg;H)$. This finishes the proof.

1352: \end{proof}

1353: 

1354: Going back to the $E^2_{1,1}$-term in~\eqref{eq:E1}, we have

1355: $$

1356: E^2_{1,1}\cong H_1(\M1g;H\otimes\bR)\cong\bR

1357: $$

1358: by Proposition~\ref{prop:h1m1}. Now consider the exact sequence

1359: $$

1360: E^2_{3,0}\cong H_3(\M1g;\bZ)\ \overset{d^2}{\lra} \

1361: E^2_{1,1}\cong \bR\lra E^3_{1,1}=E^\infty_{1,1}\lra 0 \ .

1362: $$

1363: Harer~\cite{Harer92} determined the third stable

1364: rational cohomology group

1365: $$

1366: \lim_{g\to\infty} H^3(\M1g;\bQ)

1367: $$

1368: of the mapping class group to be trivial\footnote{There is now a final

1369: result on the stable cohomology of $\Mg$ due to Madsen and Weiss~\cite{MW}.}.

1370: It follows that $E^2_{3,0}$ is a finite group for all sufficiently large $g$.

1371: Hence we can conclude that $E^\infty_{1,1}\cong \bR$ for such $g$.

1372: It is natural to expect that this $\bR$ will recover the missing $\bR$ in

1373: $E^\infty_{0,2}$ so that we obtain the surjectivity of the $\tilde\al$-factor

1374: in Theorem~\ref{th:h20}. It turns out that this is indeed the case, and below

1375: we shall give a proof of this fact which does not use Harer's result mentioned

1376: above. Before doing so, we have to prepare some general facts concerning

1377: group (co)homology in small degrees.

1378: 

1379: Consider a group extension

1380: \begin{equation}\label{eq:ext}

1381: 1\lra K\lra G\lra Q\lra 1 \ ,

1382: \end{equation}

1383: and

1384: %let $\{E^r_{p,q}\}$ be the Hochschild--Serre spectral

1385: %sequence for the integral homology of $G$.

1386: suppose we are given a $1$-cycle $c=\sum_i q_i\otimes u_i\in Z_1(Q;H_1(K))$

1387: of the group $Q$ with coefficients in the abelianization $H_1(K)$ of $K$

1388: considered as a natural $Q$-module, where $q_i\in Q$ and $u_i\in H_1(K)$.

1389: \begin{lemma}\label{lem:cc}

1390: For any choices of lifts $\tilde q_i\in G$ of the $q_i$

1391: and representatives $k_i\in K$ with $[k_i]=u_i$, the element

1392: $$

1393: \tilde c=\sum_{i} \left\{(\tilde q_i, k_i)+(\tilde q_i k_i,\tilde q_i^{-1})

1394: -(\tilde q_i,\tilde q_i^{-1})-(\id,\id)\right\}+ d

1395: $$

1396: is a $2$-cycle of $G$, where $d$ is a $2$-chain of the group $K$ such that

1397: $$

1398: \partial d=\sum_i\left\{(\tilde q_i k_i\tilde q_i^{-1})-(k_i)\right\}

1399: \ .

1400: $$

1401: Furthermore, $p_*([\tilde c])=0\in H_2(Q;\bZ)$, where $p\colon G\ra Q$ denotes

1402: the projection, and in the short exact sequence

1403: \begin{align*}

1404: 0\lra &E^\infty_{0,2}\lra

1405: \Ker\left(H_2(G;\bZ)\overset{p_*}{\lra} H_2(Q;\bZ)\right)\\

1406: & \lra E^\infty_{1,1}\

1407: \left(\cong H_1(Q;H_1(K))/d^2(E^2_{3,0})\right)\lra 0

1408: \end{align*}

1409: arising from the Hochschild--Serre spectral sequence of~\eqref{eq:ext} the

1410: class $[\tilde c]\in \Ker$ is a lift of $[c]\in H_1(Q;H_1(K))$.

1411: \end{lemma}

1412: \begin{proof}

1413: Since $c$ is a cycle by the assumption, we have

1414: $$

1415: \partial c=\sum_i (q_i(u_i)-u_i)=0 \in H_1(K) \ .

1416: $$

1417: It follows that there exists a $2$-chain $d\in C_2(K;\bZ)$

1418: with the property described in the statement of the lemma.

1419: Then a direct computation shows that $\partial \tilde c=0$.

1420: Clearly $p_*(\tilde c)=0$ and it is easy to check

1421: the rest of the required assertions.

1422: \end{proof}

1423: 

1424: The following can be proved by a standard argument in the cohomology theory

1425: of groups, see~\cite{Brown}.

1426: \begin{lemma}\label{lem:ch}

1427: Let $G$ be a group and $M$ a $G$-module. Assume we have a

1428: $G$-invariant skew-symmetric bilinear pairing

1429: $$

1430: \iota\colon M\times M\lra A \ ,

1431: $$

1432: where $A$ is an abelian group with trivial $G$-action, and

1433: we are given two crossed homomorphisms

1434: $$

1435: f_i\colon G\lra M\quad (i=1,2)

1436: $$

1437: so that $f_i(gh)=f_i(g)+ g_* f_i(h)$ for all $g,h\in G$.

1438: Then the assignment

1439: $$

1440: G\times G\ni (g,h)\longmapsto\iota(f_1(g),g_*f_2(h))\in A \ ,

1441: $$

1442: which we denote by $f_1\cdot f_2$, is a $2$-cocycle of $G$ with values

1443: in $A$ and its cohomology class in $H^2(G;A)$ depends only on the cohomology

1444: classes $[f_i]\in H^1(G;M)$ of the crossed homomorphisms $f_i$.

1445: Furthermore, $f_2\cdot f_1$ is cohomologous to $f_1\cdot f_2$,

1446: so that $[f_2\cdot f_1]=[f_1\cdot f_2]\in H^2(G;A)$.

1447: \end{lemma}

1448: %\begin{proof}

1449: %In fact, the assignment

1450: %$(g,h)\mapsto f_1(g)\otimes g_*f_2(h)\in M\otimes M$ is nothing

1451: %other than the Alexander-Whitney cup product $f_1\cup f_2$ of the

1452: %two $1$-cocycles $f_i$ of $G$ with values in $M$. The cocycle

1453: %$f_1\cdot f_2$ is obtained from this by applying the

1454: %$G$-invariant skew-symmetric pairing $M\otimes M\ra A$

1455: %to the usual cup product $[f_1][f_2]$,

1456: %namely $[f_1\cdot f_2]=\iota_*([f_1][f_2])$, so that

1457: %its cohomology class depends only on those of $f_i$.

1458: %As is well-known, $[f_2][f_1]=- [f_1][f_2]$ under the

1459: %isomorphism $t$ of $M\otimes M$ given by $t(m,n)=(n,m)$.

1460: %Since $\iota$ is skew-symmetric by assumption, we obtain

1461: %the equality $[f_2\cdot f_1]=[f_1\cdot f_2]$ as required.

1462: %This completes the proof.

1463: %\end{proof}

1464: 

1465: Now in the situation of Lemma~\ref{lem:ch}, we consider the case where

1466: $G=\Symp^c\Sigma_g^0$, $M=H^1(\Sg;\bR)$ and $\iota$ is the intersection

1467: pairing. Then we have three crossed homomorphisms

1468: $$

1469: j^*\widetilde{\Flux},\ \widetilde{\Flux}_c,\ p^*k_\bR:

1470: \Symp^c\Sigma_g^0\lra H^1(\Sg;\bR)

1471: $$

1472: and, by the definition of $\al$, we have

1473: $j^*\al=[j^*\widetilde{\Flux}\cdot j^*\widetilde{\Flux}]$.

1474: \begin{proposition}\label{prop:fluxc}

1475: In the above notation, we have

1476: $[\widetilde{\Flux}_c\cdot\widetilde{\Flux}_c]=0$

1477: and

1478: $$

1479: j^*\al=2 [p^*k_\bR\cdot\widetilde{\Flux}_c]-p^* e_1 \ .

1480: $$

1481: \end{proposition}

1482: \begin{proof}

1483: Define a map

1484: $$

1485: \widetilde{\Cal}\colon\Symp^c\Sigma_g^0\lra \bR

1486: $$

1487: by the formula

1488: $$

1489: \widetilde{\Cal}(\varph)=\int_{\Sigma_g^0}(\varph\inv)^*\la\land \la

1490: %\quad (\varph\in \Symp^c\Sigma_g^0).

1491: \ .

1492: $$

1493: The restriction of this map to the subgroup $\Ham^c\Sigma_g^0$ is a

1494: homomorphism, which, suitably normalized, is called the (second) Calabi

1495: homomorphism (see~\cite{Calabi}). In our previous paper~\cite{KM03}, we

1496: examined how the restriction of $\widetilde{\Cal}$ to the subgroup

1497: $\Symp^c_0\Sigma_g^0$ fails to be a homomorphism. Here we extend this

1498: discussion to the whole group $\Symp^c\Sigma_g^0$. For any two elements

1499: $\varph, \psi\in \Symp^c\Sigma_g^0$, we claim that

1500: \begin{equation}\label{eq:cal}

1501: \widetilde{\Cal}(\varph\psi)

1502: = \widetilde{\Cal}(\varph)+\widetilde{\Cal}(\psi)+

1503: \widetilde{\Flux}_c(\varph)\cdot (\varph\inv)^*\widetilde{\Flux}_c(\psi)

1504: \ .

1505: \end{equation}

1506: This is because

1507: \begin{align*}

1508: \widetilde{\Cal}(\varph\psi)=&\int_{\Sigma_g^0}

1509: ((\varph\psi)\inv)^*\la\land\la\\

1510: =& \int_{\Sigma_g^0} (\varph\inv)^*((\psi\inv)^*\la-\la)\land\la

1511: +(\varph\inv)^*\la\land\la\\

1512: =& \int_{\Sigma_g^0}(\varph\inv)^*\widetilde{\Flux}_c(\psi)\land

1513: ((\varph\inv)^*\la-\widetilde{\Flux}_c(\varph))+\widetilde{\Cal}(\varph)\\

1514: =& \widetilde{\Flux}_c(\varph)\cdot (\varph\inv)^*\widetilde{\Flux}_c(\psi)

1515: +\int_{\Sigma_g^0} ((\psi\inv)^*\la-\la)\land\la +\widetilde{\Cal}(\varph)\\

1516: =& \widetilde{\Flux}_c(\varph)\cdot (\varph\inv)^*\widetilde{\Flux}_c(\psi)

1517: +\widetilde{\Cal}(\psi)+\widetilde{\Cal}(\varph) \ .

1518: \end{align*}

1519: Equation~\eqref{eq:cal} implies that the $2$-cocycle

1520: $\widetilde{\Flux}_c\cdot \widetilde{\Flux}_c$ of the group $\Symp^c\Sigma_g^0$

1521: is a coboundary. Hence $[\widetilde{\Flux}_c\cdot

1522: \widetilde{\Flux}_c]=0$ as claimed.

1523: 

1524: The definition of $\al$ implies

1525: $$

1526: j^*\al= [j^*\widetilde{\Flux}\cdot j^*\widetilde{\Flux}]

1527: $$

1528: so that, by Theorem~\ref{th:ffk}, we can write

1529: \begin{equation}\label{eq:all}

1530: j^*\al=

1531: [\widetilde{\Flux}_c\cdot\widetilde{\Flux}_c]+[\widetilde{\Flux}_c\cdot

1532: p^*k_\bR]+[p^*k_\bR\cdot \widetilde{\Flux}_c]+p^*[k_\bR\cdot k_\bR] \ .

1533: \end{equation}

1534: By Lemma~\ref{lem:ch} we have

1535: $[\widetilde{\Flux}_c\cdot p^*k_\bR]=[p^*k_\bR\cdot

1536: \widetilde{\Flux}_c]$,

1537: and it was proved in~\cite{Morita89b} that $[k_\bR\cdot k_\bR]=-e_1$.

1538: If we substitute these relations in \eqref{eq:all}, we obtain the desired

1539: identity.

1540: \end{proof}

1541: 

1542: \begin{proof}[Proof of Theorem~\ref{th:h20}]

1543: In view of Proposition~\ref{prop:fl2c} and the discussion following

1544: Corollary~\ref{cor:trivial}, it suffices to show that there exist $2$-cycles

1545: of the group $\Symp^c\Sigma_g^0$ such that the evaluations of $j^*\al$ on

1546: them have as values any real number. To construct such $2$-cycles, we use

1547: the $1$-cycle $c$ of $\M1g$ with coefficients in $H$ described in

1548: Proposition~\ref{prop:h1m1}, which represents a generator of 

1549: $H_1(\M1g;H)\cong\bZ$.

1550: In order to adapt to the situation here, we consider the dual cycle

1551: $$

1552: c^*=\nu_i\otimes y^*_i-\mu_i\otimes y^*_i+\mu_{i+1}\otimes y^*_{i+1}

1553: $$

1554: with coefficients in $H^1(\Sg;\bZ)$, where the

1555: $x^*_1,\cdots,x^*_g,y^*_1,\cdots,y^*_g$ denote the dual basis of 

1556: $H^1(\Sg;\bZ)$.

1557: If $k\colon\M1g\ra H^1(\Sg;\bZ)$ is a crossed homomorphism which represents

1558: a generator of $H^1(\M1g;H^1(\Sg;\bZ))$, then we have

1559: \begin{equation}\label{eq:kc}

1560: k(c^*)=k(\nu_i)\cdot y^*_i-k(\mu_i)\cdot y^*_i+k(\mu_{i+1})\cdot

1561: y^*_{i+1}=1 \ .

1562: \end{equation}

1563: For any real number $r\in\bR$, we consider the $1$-cycle

1564: $$

1565: c^*_r=\nu_i\otimes r y^*_i-\mu_i\otimes r y^*_i+\mu_{i+1}\otimes r y^*_{i+1}

1566: $$

1567: of $\M1g$ with coefficients in $H^1(\Sg;\bR)$.

1568: Now we apply Lemma~\ref{lem:cc} in the case where

1569: $G=\Symp^c\Sigma_g^0$, $K=\Symp^c_0\Sigma_g^0$, and $Q=\M1g$.

1570: We proved in~\cite{KM03} that the flux homomorphism induces an isomorphism

1571: $$

1572: \Flux\colon H_1(\Symp^c_0\Sigma_g^0;\bZ)\cong H^1(\Sg;\bR).

1573: $$

1574: Therefore,

1575: $c^*_r$ can be considered as a $1$-cycle of $\M1g$ with coefficients

1576: in the abelianization of $\Symp^c_0\Sigma_g^0$.

1577: Hence, if we choose elements

1578: $$

1579: \tilde{\nu}_i, \tilde{\mu}_i\in \Symp^c\Sigma_g^0,

1580: \quad \varph^r_i\in \Symp^c_0\Sigma_g^0

1581: $$

1582: such that

1583: \begin{equation}\label{eq:kf}

1584: p(\tilde{\nu}_i)=\nu_i,\quad p(\tilde{\mu}_i)=\mu_i,

1585: \quad \Flux(\varph^r_i)= \Flux_c(\varph^r_i)=r y^*_i \ ,

1586: \end{equation}

1587: where $p\colon\Symp^c\Sigma_g^0\ra\M1g$ denotes the projection as before,

1588: then

1589: \begin{align*}

1590: \tilde{c}^*_r=& (\tilde{\nu}_i, \varph^r_i)+

1591: (\tilde{\nu}_i\varph^r_i,\tilde{\nu}^{-1}_i)-

1592: (\tilde{\nu}_i,\tilde{\nu}^{-1}_i)-(\id,\id)+\\

1593: & (\tilde{\mu}_i, \varph^r_i)+

1594: (\tilde{\mu}_i\varph^r_i,\tilde{\mu}^{-1}_i)-

1595: (\tilde{\mu}_i,\tilde{\mu}^{-1}_i)-(\id,\id)+\\

1596: &  (\tilde{\mu}_{i+1}, \varph^r_{i+1})+

1597: (\tilde{\mu}_{i+1}\varph^r_{i+1},\tilde{\mu}^{-1}_{i+1})-

1598: (\tilde{\mu}_{i+1},\tilde{\mu}^{-1}_{i+1})-(\id,\id)+d

1599: \end{align*}

1600: is a $2$-cycle of $\Symp^c\Sigma_g^0$, where $d$ is a $2$-chain of the

1601: group $\Symp^c_0\Sigma_g^0$ such that

1602: $$

1603: \partial d=(\tilde{\nu}_i\varph^r_i\tilde{\nu}^{-1}_i)-(\varph^r_i)

1604: +(\tilde{\mu}_i\varph^r_i\tilde{\mu}^{-1}_i)-(\varph^r_i)

1605: +(\tilde{\mu}_{i+1}\varph^r_{i+1}\tilde{\mu}^{-1}_{i+1})-(\varph^r_{i+1}) \ .

1606: $$

1607: Now we claim that

1608: \begin{equation}\label{eq:cr}

1609: j^*\al(\tilde{c}^*_r)=2r \ ,

1610: \end{equation}

1611: which will finish the proof of the theorem.

1612: To show this, observe first that $p^*e_1(\tilde{c}^*_r)=0$ because

1613: clearly $p_*(\tilde{c}^*_r)=0$. Hence, by Proposition~\ref{prop:fluxc},

1614: $$

1615: j^*\al(\tilde{c}^*_r)=2 [p^*k\cdot\widetilde{\Flux}_c](\tilde{c}^*_r)

1616: \ .

1617: $$

1618: Observe that

1619: $$

1620: [p^*k\cdot\widetilde{\Flux}_c]\left((\tilde{\nu}_i\varph^r_i,

1621: \tilde{\nu}^{-1}_i)-(\tilde{\nu}_i,\tilde{\nu}^{-1}_i)\right)=0 \ ,

1622: $$

1623: because $p^*k(\tilde{\nu}_i\varph^r_i)=p^*k(\tilde{\nu}_i)$ and

1624: $\varph^r_i$ acts trivially on the homology of $\Sigma_g^0$.

1625: The same is true for two other similar terms. Since $d$ is a $2$-chain

1626: of $\Symp^c_0\Sigma_g^0$, the evaluation of $[p^*k\cdot\widetilde{\Flux}_c]$

1627: on it vanishes. Keeping in mind equations~\eqref{eq:kf} and~\eqref{eq:kc},

1628: we can now conclude that

1629: $$

1630: j^*\al(\tilde{c}^*_r)=2 [p^*k\cdot\widetilde{\Flux}_c](\tilde{c}^*_r)

1631: =2r p^*k(\tilde{c}^*)=2r \ .

1632: $$

1633: This proves~\eqref{eq:cr} and hence the theorem.

1634: \end{proof}

1635: 

1636: \section{Proof of the main results}\label{s:proofs}

1637: 

1638: In this section we give the proofs of the main results described

1639: in Section~\ref{s:main}.

1640: 

1641: \begin{proof}[Proof of Theorem~\ref{th:hev}]

1642: Here we follow the argument of~\cite{Morita2} and of our previous

1643: paper~\cite{KM03} to prove the non-triviality of the cup products

1644: $e_1^k\tilde\al^{(\ell)}$. For this we first observe that, similar

1645: to the class $e_1$, the class $\tilde\al$ is stable, with respect to

1646: $g$, and also that it is primitive in the following sense. For each

1647: $k$, consider the genus $kg$ surface

1648: $\Sigma_{kg,1}=\Sigma_{kg}\setminus \Int D^2$ with one boundary component

1649: as the boundary connected sum

1650: $$

1651: \Sigma_{kg,1}=\Sigma_{g,1}\ \natural \cdots\natural\ \Sigma_{g,1}

1652: $$

1653: of $k$ copies of $\Sigma_{g,1}=\Sg\setminus \Int D^2$.

1654: This induces a homomorphism

1655: \begin{equation}\label{eqn:fk}

1656: f_k\colon \Symp^c \Sigma_g^0\times\cdots\times \Symp^c \Sigma_g^0

1657: \lra \Symp^c \Sigma_{kg}^0

1658: \end{equation}

1659: from the direct product of $k$ copies of the group $\Symp^c \Sigma_g^0$ to

1660: $\Symp^c \Sigma_{kg}^0$.

1661: %Namely we set the supports of the images of the $k$-components of the

1662: %former group to be disjoint in the target group.

1663: Under this homomorphism we have the equality

1664: $$

1665: f_{k}^*(\tilde\al)=\tilde\al\times 1\times\cdots\times 1+\cdots

1666: +1\times\cdots\times 1\times \tilde\al \ ,

1667: $$

1668: which follows easily from the definition of $\tilde\al$. Now we can combine

1669: Theorem~\ref{th:h20} with the above property to show the required assertion

1670: in the theorem. This finishes the proof.

1671: \end{proof}

1672: 

1673: \begin{proof}[Proof of Theorem~\ref{th:om2}]

1674: The fact that the ideal generated by $\om_0\land H_1(T^{2g};\bR)$

1675: is contained in the kernel of $\Flux^*$ has already been proved in

1676: Proposition~\ref{prop:v}. To show that $\Ker \Flux^*$ is precisely

1677: this ideal, we use the decomposition~\eqref{eq:irrep} of $H^k(T^{2g};\bR)$

1678: into irreducible summands given in Section~\ref{s:main}. It is easy to

1679: see that the quotient of this module divided by the ideal generated by

1680: $\om_0\land H_1(T^{2g};\bR)\subset H_3(T^{2g};\bR)$ is precisely

1681: $$

1682: \bR\oplus \bigoplus_{k=1}^{g} [1^k]_\bR \ .

1683: $$

1684: The $\bR$-summand in degree $2$ corresponds to the class $\alpha$

1685: and its non-triviality has already been shown in Theorem~\ref{th:hev}.

1686: Hence, to prove the assertion, it remains to show that $\Flux^*([1^k])$

1687: is non-trivial for any $k\leq g$. The highest weight vector of the

1688: irreducible representation $[1^k]$ is $x_1\land\cdots\land x_k$, where

1689: $x_1,\cdots,x_g,y_1,\cdots,y_g$ is a symplectic basis of $H_1(\Sg;\bZ)$

1690: as before. Now the definition of the flux homomorphism implies that, for

1691: any $i$, there exists an element $\varph_i\in\Symp_0\Sg$ such that

1692: $\Flux(\varph_i)=x_i^*$. Here we can choose the support of $\varph_i$ to

1693: be contained in an arbitrarily small neighbourhood of a simple closed curve

1694: which represents the homology class $x_i$. Then the $k$ elements

1695: $\varph_1,\cdots,\varph_k$ mutually commute because their supports are

1696: disjoint. Hence they form a cycle of $\Symp_0\Sg$ supported on a

1697: $k$-dimensional torus, and the cohomology class

1698: $\Flux^*(x_1\land\cdots\land x_k)\in H^{k}(\Symp_0^\delta\Sg;\bR)$

1699: takes a non-zero value (namely $1$) on this cycle. This completes the proof.

1700: \end{proof}

1701: 

1702: \begin{proof}[Proof of Theorem~\ref{th:ham}]

1703: We consider the extension

1704: $$

1705: 1\lra \Ham\Sg\lra \Symp_0\Sg\overset{\Flux}{\lra}H^1(\Sg;\bR)\lra 1 \ .

1706: $$

1707: Let $\{E^r_{p,q}\}$ be the Hochschild--Serre spectral sequence for its

1708: homology. Since $\Ham\Sg$ is perfect by Thurston~\cite{Thurston0}, see

1709: also~\cite{Banyaga}, we have

1710: $$

1711: E^2_{1,1}\cong H_1(H^1(\Sg;\bR)^\delta;H_1(\Ham^\delta\Sg;\bZ))=0 \ .

1712: $$

1713: Hence the differential $d^2\colon E^2_{3,0}\ra E^2_{1,1}$ vanishes, so that

1714: $$

1715: E^3_{3,0}\cong E^2_{3,0}\cong H_3(H^1(\Sg;\bR)^\delta;\bZ)

1716: \cong \La_\bZ^3 H^1(\Sg;\bR) \ .

1717: $$

1718: Similarly $E^2_{2,1}=0$, so that the differential $d^2\colon E^2_{2,1}\ra 

1719: E^2_{0,2}$

1720: vanishes and

1721: $$

1722: E^3_{0,2}\cong E^2_{0,2}\cong

1723: H_0(H^1(\Sg;\bR)^\delta;H_2(\Ham^\delta\Sg;\bZ))

1724: \cong H_2(\Ham^\delta\Sg;\bZ)_{H^1_\bR} \ .

1725: $$

1726: However, $E^\infty_{3,0}=E^4_{3,0}$ is equal to the image of the homomorphism

1727: $$

1728: \Flux_*\colon H_3(\Symp_0^\delta\Sg;\bZ)\lra

1729: H_3(H^1(\Sg;\bR)^\delta;\bZ) \ .

1730: $$

1731: Now we can conclude that the exact sequence

1732: $$

1733: E^\infty_{3,0}=E^4_{3,0}\lra E^3_{3,0}\lra E^3_{0,2}=E^2_{0,2}

1734: $$

1735: yields an exact sequence

1736: \begin{equation}

1737: H_3(\Symp_0^\delta\Sg;\bZ)\lra H_3(H^1(\Sg;\bR)^\delta;\bZ)\lra

1738: H_2(\Ham^\delta\Sg;\bZ)_{H^1_\bR} \ .

1739: \label{eq:eh}

1740: \end{equation}

1741: By the same computation as above in the dual context of cohomology,

1742: we obtain an exact sequence

1743: \begin{equation}

1744: H^2(\Ham^\delta\Sg;\bR)^{H^1_\bR}\lra H^3(H^1(\Sg:\bR)^\delta;\bR)

1745: \overset{\Flux^*}{\lra} H^3(\Symp_0^\delta\Sg;\bR) \ .

1746: \label{eq:ech}

1747: \end{equation}

1748: Proposition~\ref{prop:v} (see also Theorem~\ref{th:om2})

1749: implies that the continuous cohomology classes

1750: $$

1751: \tilde\om_0\land H_1(\Sg;\bR)\subset H^3(H^1(\Sg:\bR)^\delta;\bR)

1752: $$

1753: vanish under the homomorphism $\Flux^*$ so that they can be

1754: lifted to elements of $H^2(\Ham^\delta\Sg;\bR)^{H^1_\bR}$

1755: by~\eqref{eq:ech}. Now consider the cycles

1756: $$

1757: \om_0 \land H^1(\Sg;\bR)\subset \La^3_\bZ H^1(\Sg;\bR)\cong

1758: H_3(H^1(\Sg;\bR)^\delta;\bZ) \ ,

1759: $$

1760: where $\om_0\in\La^2_\bZ H^1(\Sg;\bR)$, and also their images

1761: in $H_2(\Ham^\delta\Sg;\bZ)_{H^1_\bR}$ in the exact sequence~\eqref{eq:eh}.

1762: If we consider the Kronecker products of these cycles with the above lifted

1763: cohomology classes, we can conclude that $\om_0 \land H^1(\Sg;\bR)$ maps

1764: injectively into $H_2(\Ham^\delta\Sg;\bZ)_{H^1_\bR}$. This completes the proof.

1765: \end{proof}

1766: 

1767: \begin{remark}

1768: It would be interesting to obtain explicit group cocycles of $\Ham\Sg$ which

1769: represent the above degree two cohomology classes. There should also be a

1770: relation to the work of Ismagilov~\cite{Is}. We shall pursue this elsewhere.

1771: \end{remark}

1772: 

1773: \section{Further discussion}\label{s:misc}

1774: 

1775: As in Section~\ref{s:trans}, let $e, v\in H^2(\ES^{\delta}\Sg;\bR)$ be

1776: the Euler class and the transverse symplectic class, respectively.

1777: By analogy with the definition $e_1=\pi_*(e^2)$, where

1778: $$

1779: \pi_*\colon H^4(\ES^{\delta}\Sg;\bR)\lra H^2(\Symp^\delta\Sg;\bR)

1780: $$

1781: denotes the integration over the fiber, we can define a cohomology class

1782: $$

1783: v_1 \in H^2(\Symp^\delta\Sg;\bR)

1784: $$

1785: by setting $v_1=\pi_*(ev)$. After we conjectured that $v_1$ is a linear

1786: combination of the two classes $\al$ and $e_1$, Kawazumi~\cite{Kawazumi}

1787: kindly provided a proof. More precisely, he pointed out that the

1788: contraction formula, Theorem~6.2 of~\cite{KM01}, can be adapted to the

1789: case of the cohomology class $\al$ of the group $\ES^{\delta}\Sg$, and

1790: that the following equality holds:

1791: \begin{equation}

1792: \al=-\pi_*\left((e+v)^2\right)=-e_1-2 v_1 \ .

1793: \label{v1}

1794: \end{equation}

1795: %We now interpret this relation on the subgroup

1796: %$\Symp_0^\delta\Sg\subset \Symp^\delta\Sg$ from the point of view of the

1797: %results of Section~\ref{s:trans}.

1798: %Consider the following commutative diagram, where $\Msg$

1799: %denotes the mapping class group of $\Sg$ relative to a base point:

1800: %\begin{equation}

1801: %\begin{CD}

1802: %1@>>> \ES^\delta_0\Sg  @>{\tilde i}>> \ES^\delta\Sg @>>> \Msg @>>> 1 \\

1803: %@. @V{\pi}VV @V{\pi}VV @VVV  @.\\

1804: %1@>>> \Symp^\delta_{0}\Sg @>{i}>> \Symp^\delta\Sg @>>> \Mg @>>> 1 \ .

1805: %\end{CD}

1806: %\label{eqn:ES}

1807: %\end{equation}

1808: 

1809: %If we restrict $ev$ to the

1810: %subgroup $\ES^\delta_0\Sg$, then Proposition~\ref{prop:v} implies that

1811: %$$

1812: %\tilde i^*(ev)=\{(2-2g)\mu\}\{(2g-2)\mu+[\Flux]+\frac{1}{2g-2}\tilde\om_0\}

1813: %=-\mu \tilde\om_0 \ .

1814: %$$

1815: %Hence

1816: %\begin{equation}

1817: %\pi_*(\tilde i^*(ev))=-\tilde\om_0=-\frac{1}{2}i^*\al \ ,

1818: %\label{eq:ev}

1819: %\end{equation}

1820: %where the second equality follows from Proposition~\ref{prop:alom}.

1821: %By the naturality of the integration over the fiber, we also have

1822: %$$

1823: %\pi_*(\tilde i^*(ev))=i^*\pi_*(ev)=i^* v_1 \ ,

1824: %$$

1825: %so that we can conclude

1826: %$$

1827: %i^* v_1=-\frac{1}{2}i^*\al \ .

1828: %$$

1829: %Now the Hochschild--Serre exact sequence of the bottom extension

1830: %of~\eqref{eqn:ES} implies that

1831: %$

1832: %v_1=-\frac{1}{2} \al + c e_1

1833: %$

1834: %for some constant $c\in\bR$. Kawazumi's argument mentioned above

1835: %implies that the constant $c$ equals $-\frac{1}{2}$.

1836: 

1837: Since we know by~\cite{KM03} that $e^2\not=0$, we could also apply

1838: integration over the fiber to the cohomology class $e^2v$ in order to

1839: obtain some more cohomology. However, $e^2 v$ vanishes in

1840: $H^6(\ES^\delta\Sg;\bR)$ by the Bott vanishing theorem.

1841: %The class $e^2$ is equal to the first Pontrjagin class $p_1$

1842: %of the normal bundle of the transverse foliation on the foliated

1843: %surface bundle so that, at the level of a suitable associated

1844: %jet bundle, it is contained in the ideal generated by the canonical

1845: %forms $\theta^1, \theta^2$ with respect to the fiber $\Sg$.

1846: %On the other hand, $v$ is represented by a non-zero constant times

1847: %the closed $2$ form $\theta^1\theta^2$ which is the (pull back of

1848: %the) area from of $\Sg$. Hence the form representing $p_1 v$ vanishes

1849: %identically.

1850: 

1851: 

1852: %\subsection{Further classes}

1853: 

1854: %In this subsection, we introduce some further characteristic classes

1855: %for foliated surface bundles with area-preserving holonomy.

1856: 

1857: A more promising approach to find more cohomology for the

1858: symplectomorphism groups is the following.

1859: Consider the extension

1860: \begin{equation}

1861: 1\lra\Symp_0\Sg\lra\Symp\Sg\overset{p}{\lra}\Mg\lra 1 \ .

1862: \label{eq:ssm}

1863: \end{equation}

1864: On the one hand, Theorem~\ref{th:om2} shows that we have an injection

1865: \begin{equation}

1866: \bigoplus_{k=1}^g [1^k]_\bR \subset H^*(\Symp_0^\delta\Sg;\bR) \ .

1867: \label{eq:1k}

1868: \end{equation}

1869: On the other hand, Looijenga~\cite{Looijenga} determined the stable

1870: cohomology $H^*(\Mg;V)$ of the mapping class group with coefficients

1871: in any irreducible representation $V$ of the algebraic group $Sp(2g;\bQ)$.

1872: In particular, the cohomology groups $H^*(\Mg;[1^k])$ are highly non-trivial.

1873: In the spectral sequence for the cohomology of the extension~\eqref{eq:ssm},

1874: there are many non-trivial $E_2$-terms

1875: $$

1876: E^{p,q}_2=H^p(\Mg;H^q(\Symp_0^\delta\Sg;\bR)) \ .

1877: $$

1878: For example, it was proved in~\cite{Morita93} that 

1879: $H^1(\Mg;[1^3]_\bR)\cong\bR$,

1880: and an explicit computation using Looijenga's formula shows

1881: $H^2(\Mg;[1^2]_\bR)\cong\bR$. Hence we have injections

1882: \begin{align*}

1883: &\bR\subset E^{1,3}_2=H^1(\Mg;H^3(\Symp_0^\delta\Sg;\bR)) \\

1884: &\bR\subset E^{2,2}_2=H^2(\Mg;H^2(\Symp_0^\delta\Sg;\bR))

1885: \end{align*}

1886: for all sufficiently large $g$. It seems likely that these

1887: two copies of $\bR$ survive to the $E_\infty$ term, so that they

1888: define certain cohomology classes in $H^4(\Symp^\delta\Sg;\bR)$.

1889: More generally, the summands~\eqref{eq:1k} and the non-trivial

1890: cohomology groups $H^*(\Mg;[1^k]_\bR)$ should give rise to

1891: infinitely many cohomology classes in $H^*(\Symp^\delta\Sg;\bR)$.

1892: \begin{problem}

1893: Prove that these cohomology classes are non-trivial.

1894: \end{problem}

1895: 

1896: Next, there are completely different candidates for possible classes

1897: in $H^*(\Symp^\delta\Sg;\bR)$ coming from the cohomology of the Lie

1898: algebra $V_n$ of formal Hamiltonian vector fields on $\bR^{2n}$

1899: first studied by Gelfand, Kalinin and Fuks in~\cite{GKF}. This Lie

1900: algebra $V_n$ contains ${\mathfrak sp}(2n,\bR)$ as a subalgebra

1901: consisting of vector fields corresponding to linear symplectomorphisms.

1902: Let $\BG_2^\om$ be the Haefliger classifying space for the pseudogroup

1903: of local symplectomorphisms of $\bR^2$ with respect to the standard

1904: symplectic form. Then there is a natural homomorphism

1905: \begin{equation}

1906: H^*_c(V_1;{\frak sp}(2,\bR))\lra H^*(\BG_2^\om;\bR)

1907: \label{eq:vbg}

1908: \end{equation}

1909: from the continuous cohomology of $V_1$ relative to the subalgebra

1910: ${\frak sp}(2,\bR)$ to the real cohomology group of $\BG_2^\om$.

1911: 

1912: There is also an obvious continuous mapping

1913: $$

1914: K(\ES^\delta\Sg,1)\lra \BG_2^\om

1915: $$

1916: which classifies the transversely symplectic codimension $2$

1917: foliation on the classifying space for the group $\ES^\delta\Sg$,

1918: that is the total space of the universal foliated $\Sg$-bundle over

1919: $\BS^\delta\Sg$. This induces homomorphisms

1920: \begin{equation}

1921: H^*(\BG_2^\om;\bR)\lra H^*(\ES^\delta\Sg;\bR)

1922: \overset{\pi_*}{\lra} H^{*-2}(\Symp^\delta\Sg;\bR) \ ,

1923: \label{eq:bgs}

1924: \end{equation}

1925: where the last homomorphism is the integration along the fibre.

1926: Combining~\eqref{eq:vbg} and~\eqref{eq:bgs} we obtain a homomorphism

1927: \begin{equation}

1928: H^*_c(V_1;{\frak sp}(2,\bR))\lra H^{*-2}(\Symp^\delta\Sg;\bR) \ .

1929: \label{eq:}

1930: \end{equation}

1931: 

1932: Now, Gelfand, Kalinin and Fuks~\cite{GKF} found a new cohomology class

1933: in $H^7_c(V_1;{\frak sp}(2,\bR))$.

1934: %besides standard elements, e.g. the transverse

1935: %symplectic class as well as the first Pontrjagin class.

1936: Later, Metoki~\cite{Metoki} extended their computation and found another

1937: exotic class in $H^9_c(V_1;{\frak sp}(2,\bR))$. It seems to be widely

1938: believed that there should exist infinitely many such exotic classes.

1939: \begin{problem}

1940: Study the cohomology classes in $H^*(\Symp^\delta\Sg;\bR)$ induced from

1941: exotic classes in $H^{*+2}_c(V_1;{\frak sp}(2,\bR))$. In particular, prove

1942: that the two elements in $H^5(\Symp^\delta\Sg;\bR)$ and 

1943: $H^7(\Symp^\delta\Sg;\bR)$

1944: induced from the classes found by Gelfand, Kalinin and Fuks and by Metoki,

1945: are non-trivial.

1946: \end{problem}

1947: 

1948: Recall that Harer~\cite{Harer85} proved that the homology groups of the mapping

1949: class groups $\Mg$ stabilize with respect to the genus $g$ in a certain stable

1950: range. In view of the fact that all the characteristic classes we introduced in

1951: this paper are stable with respect to $g$, we would like to propose the 

1952: following

1953: problem, although it appears to be beyond the range of available techniques 

1954: at the

1955: moment:

1956: \begin{problem}

1957: Determine whether the homology groups of $\Symp^\delta \Sg$ stabilize

1958: with respect to $g$, or not. In particular, is it true that

1959: $$

1960: H_2(\Symp^\delta\Sg;\bZ)\cong \bZ\oplus S^2_\bQ\bR

1961: $$

1962: for all $g\geq 3$ ?

1963: \end{problem}

1964: 

1965: \section*{Appendix: Proof of Proposition~\ref{prop:la2}}

1966: 

1967: To prove Proposition~\ref{prop:la2}, observe first that the second exterior

1968: power $\La^2_{\bZ}H^1(\Sg;\bR)$ over $\bZ$ is naturally isomorphic to the same

1969: power $\La^2_{\bQ}H^1(\Sg;\bR)$ over $\bQ$ because $H^1(\Sg;\bR)$ is a uniquely

1970: divisible abelian group. Choose a Hamel basis $a_\la\ (\la\in A)$ of $\bR$ as a

1971: vector space over $\bQ$. Then we can write

1972: $$

1973: H^1(\Sg;\bR)=\sum_{\la} a_\la H^1(\Sg;\bQ) \ .

1974: $$

1975: Hence

1976: $$

1977: \La^2_{\bQ} H^1(\Sg;\bR)=\sum_{\la} a_\la \La^2_{\bQ}H^1(\Sg;\bQ)\oplus

1978: \sum_{\la<\mu} a_\la H^1(\Sg;\bQ)\otimes a_\mu H^1(\Sg;\bQ) \ ,

1979: $$

1980: where we choose a total order in the index set $A$. Clearly, this is a

1981: decomposition of $\Mg$-modules. It is easy to see that the intersection

1982: pairing gives rise to an isomorphism

1983: $$

1984: \left(\La^2_{\bQ}H^1(\Sg;\bQ)\right)_{\Mg}\cong \bQ \ .

1985: $$

1986: We also have

1987: %\begin{align*}

1988: $$

1989: \left(H^1(\Sg;\bQ)\otimes H^1(\Sg;\bQ)\right)_{\Mg}

1990: \cong

1991: \left(S^2H^1(\Sg;\bQ)\oplus \La^2H^1(\Sg;\bQ)\right)_{\Mg}\\

1992: \cong \bQ \ .

1993: $$

1994: %\end{align*}

1995: Here we have used the fact that

1996: $$

1997: \left(S^2H^1(\Sg;\bQ)\right)_{\Mg}=0 \ ,

1998: $$

1999: which is true because the action of $\Mg$ on $S^2H^1(\Sg;\bQ)$ factors

2000: through that of the algebraic group $Sp(2g,\bQ)$, and $S^2H^1(\Sg;\bQ)$

2001: is a non-trivial irreducible $Sp(2g,\bQ)$-module.

2002: 

2003: Thus we obtain an isomorphism

2004: $$

2005: \left(\La^2_{\bZ} H^1(\Sg;\bR)\right)_{\Mg}\cong

2006: \left(\sum_{\la} a_\la\otimes a_\la \bQ\right)\oplus

2007: \left(\sum_{\la<\mu} a_\la\otimes a_\mu\bQ\right) \ .

2008: $$

2009: It is easy to see that the right-hand side of the above expression

2010: can be naturally identified with $S^2_\bQ\bR$. This completes the proof.

2011: 

2012: 

2013: 

2014: \bibliographystyle{amsplain}

2015: 

2016: \begin{thebibliography}{999}

2017: 

2018: \bibitem{Banyaga}

2019: A.~Banyaga,

2020: {\it Sur la structure du groupe des diff\'eomorphismes qui

2021: pr\`eservent une forme symplectique},

2022: Comment.~Math.~Helv.~{\bf 53}

2023: (1978), 174--227.

2024: 

2025: \bibitem{Brown}

2026: K. S.  Brown,

2027: {\sl Cohomology of Groups},

2028: Graduate Texts in Mathematics vol.~87,

2029: Springer Verlag,

2030: 1982.

2031: 

2032: \bibitem{Calabi}

2033: E.~Calabi,

2034: {\it On the group of automorphisms of a symplectic manifold},

2035: in {\sl Problems in Analysis},

2036: ed.~R.~Gunning, Princeton University Press, 1970, 1--26.

2037: 

2038: \bibitem{FH}

2039: W. Fulton, J. Harris,

2040: {\sl Representation Theory},

2041: Graduate Texts in Mathematics vol.~129,

2042: Springer Verlag 1991.

2043: 

2044: \bibitem{GKF}

2045: I.~M.~Gelfand, D.~I.~Kalinin, D.~B.~Fuks,

2046: {\it The cohomology of the Lie algebra of Hamiltonian formal vector fields},

2047: (Russian) Funkcional.~Anal.~i Prilozen.~{\bf 6} (1972), 25--29;

2048: English trans.~in Functional Anal.~Appl.~{\bf 6} (1972), 193--196.

2049: 

2050: \bibitem{Harer85}

2051: J.~Harer, {\it Stability of the homology of the mapping class group of

2052: an orientable surface},

2053: Ann.~of Math.~{\bf 121} (1985), 215--249.

2054: 

2055: \bibitem{Harer92}

2056: J.~Harer,

2057: {\it The third homology group of the moduli space of curves},

2058: Duke Math.~J.~{\bf 63}

2059: (1992), 25--55.

2060: 

2061: \bibitem{Is}

2062: R.~S.~Ismagilov,

2063: {\it Inductive limits of area-preserving diffeomorphism groups},

2064: (Russian) Funktsional.~Anal.~i Prilozhen.~{\bf 37} (2003), 36--50, 95;

2065: Engl.~trans.~in Funct.~Anal.~Appl.~{\bf 37} (2003), 191--202.

2066:  

2067: \bibitem{Johnson80}

2068: D.~Johnson,

2069: {\it An abelian quotient of the mapping class group ${\mathcal I}_g$},

2070: Math.~Ann.~{\bf 249}

2071: (1980), 225--242.

2072: 

2073: \bibitem{Kawazumi}

2074: N.~Kawazumi,

2075: private communication.

2076: 

2077: \bibitem{KM96}

2078: N.~Kawazumi, S.~Morita,

2079: {\it The primary approximation to the cohomology of the moduli

2080: space of curves and cocycles for the stable characteristic classes},

2081: Math.~Research Letters {\bf 3}

2082: (1996), 629--641.

2083: 

2084: \bibitem{KM01}

2085: N.~Kawazumi, S.~Morita,

2086: {\it The primary approximation to the cohomology of the moduli

2087: space of curves and cocycles for the Mumford-Morita-Miller classes},

2088: preprint.

2089: 

2090: \bibitem{KM03}

2091: D.~Kotschick, S.~Morita,

2092: {\it Signatures of foliated surface bundles and the symplectomorphism

2093: groups of surfaces}, Topology {\bf 44} (2005), 131--149.

2094: 

2095: \bibitem{Looijenga}

2096: E.~Looijenga,

2097: {\it Stable cohomology of the mapping class group with symplectic

2098: coefficients and of the universal Abel-Jacobi map},

2099: J.~Algebraic Geometry {\bf 5}

2100: (1996), 135--150.

2101: 

2102: \bibitem{MW}

2103: I.~Madsen, M.~Weiss,

2104: {\it The stable moduli space of Riemann surfaces: Mumford's conjecture},

2105: preprint math.AT/0212321 v3 14Jul2004.

2106: 

2107: \bibitem{MS}

2108: D.~McDuff, D.~Salamon,

2109: {\sl Introduction to Symplectic Topology},

2110: second edition, Oxford University Press,

2111: 1998.

2112: 

2113: \bibitem{Metoki}

2114: S.~Metoki,

2115: {\sl Non-trivial cohomology classes of Lie algebras of volume

2116: preserving formal vector fields},

2117: thesis, University of Tokyo, 2000.

2118: 

2119: \bibitem{Morita85}

2120: S.~Morita,

2121: {\it Discontinuous invariants of foliations},

2122: Adv.~Stud.~Pure Math.~{\bf 5}

2123: (1985), 169--193.

2124:  

2125: \bibitem{Morita2}

2126: S.~Morita,

2127: {\it Characteristic classes of surface bundles},

2128: Invent.~Math.~{\bf 90}

2129: (1987), 551--577.

2130: 

2131: \bibitem{Morita89}

2132: S.~Morita,

2133: {\it Families of Jacobian manifolds and characteristic classes of surface 

2134: bundles I},

2135: Ann.~Inst.~Fourier {\bf 39}

2136: (1989), 777--810.

2137: 

2138: \bibitem{Morita89b}

2139: S.~Morita,

2140: {\it Families of Jacobian manifolds and characteristic classes of surface 

2141: bundles II},

2142: Math.~Proc.~Camb.~Phil.~Soc.~{\bf 105}

2143: (1989), 79--101.

2144: 

2145: \bibitem{Morita93}

2146: S.~Morita, {\it The extension of Johnson's homomorphism from the Torelli

2147: group to the mapping class group}, Invent.~Math.~{\bf 111} (1993), 197--224.

2148: 

2149: \bibitem{Morita96}

2150: S.~Morita,

2151: {\it A linear representation of the mapping class group

2152: of orientable surfaces and characteristic classes of surface bundles},

2153: in {\sl Proceedings of the Taniguchi Symposium on Topology and Teichm\"uller

2154: Spaces} held in Finland, July 1995,

2155: World Scientific

2156: 1996, 159--186.

2157: 

2158: \bibitem{Moser}

2159: J.~Moser,

2160: {\it On the volume elements on a manifold},

2161: Trans.~Amer.~Math.~Soc.~{\bf 120}

2162: (1965), 286--294.

2163: 

2164: \bibitem{Thurston0}

2165: W. Thurston,

2166: {\it On the structure of volume-preserving diffeomorphisms},

2167: unpublished manuscript.

2168: 

2169: \bibitem{Tsuboi}

2170: T. Tsuboi,

2171: {\it The Calabi invariant and the Euler class},

2172: Trans.~Amer.~Math.~Soc.~{\bf 352} (2000), 515--524.

2173: 

2174: 

2175: \end{thebibliography}

2176: 

2177: \end{document}

2178: