1: \documentclass[letterpaper]{article}

2: \usepackage[margin=2cm]{geometry}

3: \usepackage{array}

4: %%\usepackage{rotating}

5: \usepackage[leqno]{amsmath}

6: \usepackage{amsfonts,amssymb,amsthm}

7: \usepackage{amscd,amsxtra}

8: %%\usepackage{diagrams}

9: %%\diagramstyle[amstex]

10: 

11: 

12: %%%%%%%%%%%%%%%%%%%%%%%

13: %% euler script for mathcal

14: \usepackage{eucal}

15: %%\usepackage[mathscr]{eucal}

16: 

17: %%\usepackage{stmaryrd}

18: %%\usepackage{mathbbol}

19: \usepackage{mathrsfs} %provides \mathscr

20: 

21: %%%%%%%%%%%%%%%%%%%%%%% concrete fonts and math

22: %%\usepackage{beton}

23: %%\usepackage{euler}

24: %%\usepackage{concmath}

25: %%\usepackage{concrete}

26: 

27: %%%%%%%%%%%%%%%%%%%%%%% postscript fonts

28: %%\usepackage{times}

29: %%\usepackage{mathtime}

30: %%\usepackage{pslatex}

31: %%\usepackage{mathptm}

32: %%\usepackage[scaled=.92]{helvet}

33: %%\renewcommand{\familydefault}{\sfdefault}

34: 

35: %%%%%%%%%%%%%%%%%%%%%%% other

36: %%\usepackage{comment}

37: 

38: %%XyPic

39: \usepackage[all]{xy}

40: %%\CompileMatrices %xypic: precompile diagrams

41: \SelectTips{cm}{} %xypic: arrowheads with cm fonts (like AmsCD)

42: 

43: %%\usepackage{color,graphicx}

44: 

45: %%\usepackage[colorlinks=true]{hyperref}

46: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

47: \newtheorem{theorem}{Theorem}[section]

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

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

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

51: %%

52: %%\newtheorem{alphthm}{Theorem}

53: %%\renewcommand{\thealphthm}{\Alph{alphthm}}

54: %%

55: \theoremstyle{definition}

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

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

58: %%\newtheorem*{assumption}{Assumption}

59: %%

60: \theoremstyle{remark} \newtheorem{remark}[theorem]{Remark}

61: 

62: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

63: \numberwithin{equation}{section}

64: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

65: 

66: %%  Curly (=cal in the sense of mathrsfs) symbol

67: \newcommand{\curly}[1]{\mathscr{#1}}

68: \newcommand{\cD}{\curly{D}}

69: \newcommand{\cL}{\curly{L}}

70: 

71: 

72: %% fields and other notable objects requiring a blackboard

73: %% boldface symbol

74: \newcommand{\field}[1]{\ensuremath{\mathbb{#1}}}

75: \newcommand{\ZZ}{\field{Z}}

76: \newcommand{\RR}{\field{R}}

77: \newcommand{\CC}{\field{C}}

78: \newcommand{\HH}{\field{H}}

79: \newcommand{\PP}{\field{P}}

80: \newcommand{\TT}{\field{T}}

81: 

82: %% objects related to complex analysis and alg. curves

83: \newcommand{\projspace}[1]{\PP^{#1}}

84: \newcommand{\pione}{\projspace{1}}

85: 

86: %% cross-ratio

87: \newcommand{\CR}[4]{[#1\colon #2\colon #3\colon #4]}

88: 

89: %% differential complexes

90: \newcommand{\complex}[1]{\mathsf{#1}} %%

91: \newcommand{\CCC}{\complex{C}}

92: %% \newcommand{\complex}[2][\bullet]{{#2}^{#1}}

93: %% \newcommand{\CCC}{{\complex[{}]{C}}}

94: %% Cech complexes

95: \newcommand{\sheafcech}{\Check{\curly{C}}}

96: 

97: %% Categories

98: \newcommand{\cat}[1]{\mathsf{#1}}

99: 

100: %% Sheaves

101: \newcommand{\sheaf}[1]{\underline{\mathnormal{#1}}}

102: %%\newcommand{\locsys}[1]{\mathbb{#1}}

103: %%\newcommand{\locsys}[1]{\mathrm{#1}}

104: \newcommand{\sha}[2][\bullet]{\sheaf{A}_{#2}^{#1}}

105: \newcommand{\shomega}[2][\bullet]{\sheaf{\Omega}_{#2}^{#1}}

106: \newcommand{\she}[2][\bullet]{\sheaf{\mathcal{E}}_{#2}^{#1}}

107: \newcommand{\sho}[1]{\mathcal{O}_{#1}}

108: \newcommand{\deligne}[3][\bullet]{#2(#3)^{#1}_\mathcal{D}}

109: \newcommand{\deltilde}[3][\bullet]{%

110:   \smash[t]{\widetilde{#2(#3)}}^{#1}_\mathcal{D}}

111: \newcommand{\delub}[3][\bullet]{%

112:   \smash[b]{\underline{#2(#3)}}^{#1}_\mathcal{D}}

113: \newcommand{\dhh}[2][\bullet]{D(#2)_\mathit{h.h.}^{#1}}

114: \newcommand{\delH}[4][\bullet]{H^{#1}_\mathcal{D}(#2, #3(#4))}

115: \newcommand{\dhhH}[3][\bullet]{%

116:   H^{#1}_{\mathcal{D}_\mathit{h.h.}}(#2,#3)}

117: 

118: %% quasi-isomorphism

119: \newcommand{\qi}{\xrightarrow{\simeq}}

120: \newcommand{\lqi}{\overset{\simeq}{\longrightarrow}}

121: 

122: %% hypercohomology

123: \newcommand{\hyper}[1]{\mathbf{#1}}

124: \newcommand{\HHH}{\hyper{H}}

125: \newcommand{\RRR}{\hyper{R}}

126: \newcommand{\LLL}{\hyper{L}}

127: 

128: %% Cech coverings

129: \newcommand{\cover}[1]{\mathcal{#1}}

130: 

131: %% differentials

132: \newcommand{\del}{\partial} %usual partial

133: \newcommand{\delb}{\Bar\partial} %same with bar

134: \newcommand{\deltacheck}{\Check\delta} %for Cech complexes

135: %% double complexes

136: \newcommand{\delp}{\partial^\prime}

137: \newcommand{\delpp}{\partial^{\prime\prime}}

138: \newcommand{\deltap}{\d}%

139: \newcommand{\deltapp}{\delta}

140: 

141: %% operator names

142: %% \renewcommand{\d}{\operatorname{d}}

143: \DeclareMathOperator{\dd}{d} \DeclareMathOperator{\id}{id}

144: \DeclareMathOperator{\I}{I} \DeclareMathOperator{\Tot}{Tot}

145: \DeclareMathOperator{\Hom}{Hom} \DeclareMathOperator{\Tor}{Tor}

146: \DeclareMathOperator{\Ext}{Ext} \DeclareMathOperator{\Ker}{Ker}

147: \DeclareMathOperator{\cone}{Cone} \DeclareMathOperator{\Coker}{Coker}

148: \DeclareMathOperator{\im}{Im} \DeclareMathOperator{\SL}{SL}

149: \DeclareMathOperator{\PSL}{PSL} \DeclareMathOperator{\SO}{SO}

150: \DeclareMathOperator{\Ad}{Ad} \DeclareMathOperator{\ad}{ad}

151: \DeclareMathOperator{\deck}{Deck}

152: 

153: %% miscellaneous

154: \newcommand{\tame}[2]{\bigl(#1,#2\bigr]}

155: \newcommand{\tate}{2\pi\sqrt{-1}}

156: \newcommand{\bil}[2]{\bigl(#1\,\vert\,#2\bigr)}

157: \newcommand{\dual}[2]{\langle#1\,,\,#2\rangle}

158: \newcommand{\abs}[1]{\lvert#1\rvert}

159: \newcommand{\norm}[1]{\lVert#1\rVert}

160: \newcommand{\interior}{\lrcorner}

161: \newcommand{\onehalf}{\frac{1}{2}}

162: \newcommand{\onefourth}{\frac{1}{4}}

163: \newcommand{\ihalf}{\frac{\sqrt{-1}}{2}}

164: \newcommand{\eqdef}{\overset{\mathrm{def}}{=}}

165: \newcommand{\pic}[1]{\mathrm{Pic}(#1)}

166: \newcommand{\cm}[1]{\mathscr{CM}(#1)}

167: \newcommand{\li}{\mathit{Li}_2}

168: \newcommand{\bwli}{{\cL}_2}

169: \newcommand{\bwd}{{\cD}_2}

170: 

171: %% other

172: \newcommand{\bei}{Be\u\i{}linson}

173: \newcommand{\cech}{\v{C}ech}

174: 

175: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

176: \setcounter{tocdepth}{2}

177: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

178: 

179: \title{On hermitian-holomorphic classes related to uniformization, the

180:   dilogarithm, and the Liouville Action}

181: 

182: 

183: \author{Ettore Aldrovandi\\

184:   Department of Mathematics\\

185:   Florida State University\\

186:   Tallahassee, FL 32306-4510, USA\\

187:   \texttt{aldrovandi@math.fsu.edu} }

188: 

189: %%\author{Ettore Aldrovandi}

190: %%\address{Department of Mathematics\\

191: %%  Florida State University\\

192: %%  Tallahassee, FL 32306-4510, USA}

193: %%\email{aldrovandi@math.fsu.edu}

194: 

195: \date{}

196: 

197: \begin{document}

198: 

199: \maketitle

200: 

201: \begin{abstract}

202:   Metrics of constant negative curvature on a compact Riemann surface

203:   are critical points of the Liouville action functional, which in

204:   recent constructions is rigorously defined as a class in a

205:   \cech-de~Rham complex with respect to a suitable covering of the

206:   surface.

207:   

208:   We show that this class is the square of the metrized holomorphic

209:   tangent bundle in hermitian-holomorphic Deligne cohomology. We achieve

210:   this by introducing a different version of the hermitian-holomorphic

211:   Deligne complex which is nevertheless quasi-isomorphic to the one

212:   introduced by Brylinski in his construction of Quillen line bundles. We

213:   reprove the relation with the determinant of cohomology construction.

214:   

215:   Furthermore, if we specialize the covering to the one provided by a

216:   Kleinian uniformization (thereby allowing possibly disconnected

217:   surfaces) the same class can be reinterpreted as the transgression of

218:   the regulator class expressed by the Bloch-Wigner dilogarithm.

219: \end{abstract}

220: 

221: \tableofcontents

222: 

223: \section{Introduction}

224: \label{sec:introduction}

225: 

226: Metrics of constant negative curvature play a very important role in

227: uniformization problems for compact Riemann surfaces of genus $g>1$. The

228: condition that the scalar curvature associated to a conformal metric on

229: a Riemann surface $X$ be equal to $-1$ is equivalent to the fact that

230: the associated conformal factor satisfies a nonlinear partial

231: differential equation known as the Liouville equation.

232: 

233: The Liouville equation appears as early as in one of the approaches

234: considered by Poincar\'e to attack the uniformization theorem

235: \cite{poincare:liou}. In relatively recent times, it has received

236: considerable attention in Theoretical and Mathematical Physics due to

237: the key role it plays in Polyakov's approach to String Theory

238: \cite{polyakov:bosonic}, especially from the point of view of

239: non-critical strings and two-dimensional quantum gravity. In this

240: context one refers to the conformal factor of the metric as the

241: Liouville ``field.''

242: 

243: As usual in the context of differential equations with a physical

244: motivation, one would normally like to formulate a variational principle

245: to express the Liouville equation as an extremum condition. Namely,

246: given a Riemann surface $X$ and the space $\cm{X}$ of all conformal

247: metrics on it, the metric of constant negative curvature should be a

248: critical point of a functional defined over $\cm{X}$. This functional is

249: the Liouville action. As it happens, action functionals may turn out to

250: be even more relevant than the equations they are associated to. The

251: Liouville action is no exception in this sense: it has deep connection

252: with the geometry of Teichm\"uller spaces

253: \cite{zogtak1987-1,zogtak1987-2}, and in Physics it describes the

254: conformal anomaly in String Theory.

255: 

256: Providing a rigorous mathematical definition of the Liouville action

257: functional is however far from trivial. The very geometric properties of

258: the Liouville field itself prevent expressing the corresponding

259: functional as a plain integral of a $2$-form on a Riemann surface.

260: Correction terms are required, typically in the form of integration of

261: lower degree forms over the $1$-skeleton of an appropriate simplicial

262: realization of $X$. (One should notice that this behavior is not

263: specific to the Liouville equation, and it is by now possible to give a

264: characterization, in terms of homological algebra, of these type of

265: functionals, see ref.\ \cite{MR1908413}.)

266: 

267: It is possible to directly determine the necessary correction terms by

268: requiring that the variational problem be well defined. This, however,

269: is not completely satisfactory from the point of view of certain

270: applications to deformation theory, where a consistent definition across

271: a \emph{family} of surfaces is required. Quite recently, a more

272: systematic construction, based on the homological algebra techniques

273: developed by the author and L.\ A.\ Takhtajan in \cite{aldtak1997}, was

274: carried out by L.\ A.\ Takhtajan and L.-P.\ Teo in ref.\ 

275: \cite{math.CV/0204318}, generalizing the earlier results of

276: \cite{zogtak1987-1,zogtak1987-2}.  The authors of ref.\ 

277: \cite{math.CV/0204318} constructed a \cech\ cocycle with respect to the

278: \'etale cover of $X$ associated to a quasi-Fuchsian (and more generally

279: Kleinian) uniformization. Since their construction works across

280: (Kleinian) deformations, it could be exploited to obtain results of

281: global nature on the analytic geometry of Kleinian deformation spaces.

282: As a further result, the authors of loc.\ cit.\ were able to rigorously

283: prove the validity of the ``holography principle'' for the Liouville

284: action corresponding to a large class of Kleinian (in particular

285: Fuchsian and quasi-Fuchsian) uniformizations.  Specifically, they proved

286: that given a second kind Kleinian group, the corresponding Liouville

287: action can be obtained as the regularized limit of the hyperbolic volume

288: of the corresponding associated $3$-manifold.  This extends to the

289: general Kleinian case a previous formula obtained by Krasnov

290: \cite{MR2002k:81230} for classical Schottky groups.

291: 

292: Our interest in this matter is two-fold. From the perspective of the

293: newer methods adopted in \cite{aldtak2000}, the covering map $U\to X$

294: associated to the uniformization by a discrete group $\Gamma=\deck

295: (U/X)$ is but one of the many possible covers comprising an appropriate

296: category $\cat{C}$ of, say, local diffeomorphisms $U\to X$---the most

297: obvious choice being that of standard open cover $\cover{U}=\{ U_i\}$

298: with associated space $U=\coprod_i U_i$. In particular one expects to be

299: able to apply the methods of \cite{aldtak1997} and \cite{aldtak2000}

300: uniformly on a class of reasonably behaved covers of $X$.

301: 

302: Second, the focus of ref.\ \cite{aldtak2000} was on the rigorous

303: definition of a functional for quasi-conformal deformation of the

304: Riemann surface $X$ and its application to the study of projective

305: structures. A main result is that the construction of the action is

306: possible thanks to the vanishing of the ``tame symbol'' (see refs.\ 

307: \cite{del:symbole} and \cite{brymcl:deg4_II} for the relevant

308: definitions) $\tame{T_X}{T_X}$, where $T_X$ is the holomorphic tangent

309: line bundle of $X$. The vanishing determines local choices (with respect

310: to a cover) of a Bloch-type dilogarithm which then allow for a

311: cohomological construction of the action. There are many indication that

312: the Liouville action ought to be the hermitian square of a functional of

313: the type studied in \cite{aldtak2000}.\footnote{From a physical point of

314:   view this originates in the modular geometry approach to Conformal

315:   Field Theory advocated by Friedan and Shenker in \cite{MR88b:81146}.

316:   Mathematically speaking, it is one of the many proposed forms of the

317:   holomorphic factorization property for determinant line bundles.} Thus

318: it is natural to ask whether there is an analogous mechanism as the one

319: in loc.\ cit.\ to obtain a general construction of the Liouville action

320: by replacing the holomorphic symbol maps and dilogarithms with

321: corresponding real objects.

322: 

323: In this paper we answer this question in the affirmative. More

324: precisely, we show that the Liouville action (up to the area term

325: which is given by an ordinary $2$-form) can be computed as a

326: symbol map taking values in hermitian holomorphic Deligne

327: cohomology, first introduced by Brylinski and McLaughlin in their

328: study of degree four characteristic classes

329: \cite{brymcl:deg4_II}. (By way of comparison, the tame symbols

330: used in ref.\ \cite{aldtak2000} used holomorphic and smooth

331: Deligne cohomology.) In particular we show that the dilogarithm

332: type terms are replaced here by the Bloch-Wigner function, the

333: real valued counterpart of the dilogarithm (see refs.

334: \cite{MR2001i:11082}, and \cite{MR94k:19002,MR2002g:52013} for a

335: review.)

336: 

337: The appearance of the Bloch-Wigner function ties very well with

338: the holography property of the Liouville function proved in

339: \cite{math.CV/0204318} in the following sense. As mentioned

340: before, the Liouville action (up to the area term) relative to a

341: Kleinian uniformization\footnote{Note that $X$ is allowed to be

342:   disconnected.}  $U\to X$ can also be computed as the

343: ``regularized volume'' of the associated $3$-manifold $N=\Gamma

344: \backslash (U \cup \HH^3)$, where $\Gamma =\deck (U/X)$ as

345: before, $U\subset \PP^1$ is the domain of discontinuity for

346: $\Gamma$, and $\HH^3$ is the standard hyperbolic $3$-space. (To

347: define the regularized volume would lead us too far afield. It

348: suffices to mention that the conformal factor of a metric on

349: $X=\partial N$ can be used to select a compact submanifold

350: $N_\epsilon$ whose volume is finite. One then subtracts from the

351: volume of $N_\epsilon$ the areas of the boundary components and

352: other carefully chosen constants independent of the metric

353: structure, so that the resulting quantity will have a finite

354: limit as $\epsilon \to 0$.)  On the other hand, the hyperbolic

355: volume in three dimensions corresponds to a three dimensional

356: (purely imaginary) class on $\PSL_2(\CC)$ expressible through the

357: Bloch-Wigner dilogarithm, the so-called \emph{regulator class.}

358: We show that the regulator is precisely the class that needs to

359: be killed in order to close the cohomological descent conditions

360: required to calculate the Liouville action for a covering map

361: $U\to X$ with covering group a Kleinian group $\Gamma$.  This is

362: possible, since for a second kind Kleinian group the quotient

363: $\HH^3/\Gamma$ is non-compact, hence it carries no cohomology in

364: dimension three, so the class represented by the Bloch-Wigner

365: function, pulled back to $\Gamma$ via the imbedding $\Gamma

366: \hookrightarrow \PSL_2(\CC)$, vanishes.

367: 

368: Returning to the cohomological interpretation of the construction

369: of the Liouville action, it should also be noted that leaving

370: aside the area term, our results show that the cohomologically

371: non trivial part is indeed a square. Namely, for a conformal

372: metric $\rho \in \cm{X}$ we consider the pair $(T_X,\rho)$ as a

373: holomorphic line bundle equipped with an hermitian metric. Then,

374: using that hermitian holomorphic Deligne cohomology has a cup

375: product, we show that the Liouville action is just the square of

376: the class of $(T_X,\rho)$. In fact this identification holds at

377: the level of cocycles, rather than only for the corresponding

378: classes.

379: 

380: Again leaving aside the area term, it immediately follows from

381: the properties of hermitian holomorphic Deligne cohomology that

382: most of the story carries over to the case of a \emph{pair} of

383: holomorphic line bundles $L$ and $L'$ equipped with hermitian

384: metrics $\rho$ and $\rho'$, respectively. Furthermore, Brylinski

385: shows in \cite{bry:quillen} that the pairing of two such

386: holomorphic line bundles with metrics corresponds to the pairing

387: defined by Deligne on the determinant of cohomology in

388: \cite{MR89b:32038}. Without introducing the machinery of

389: $2$-gerbes, we reobtain this result in our setting. Specifically,

390: we directly obtain Gabber's formula for the hermitian metric on

391: the determinant line from the explicit cocycle for the cup

392: product of two metrized line bundles. In turn this shows that the

393: Liouville action is a multiple of the determinant of cohomology,

394: thereby generalizing earlier results (cf.\ 

395: \cite{zograf1990})---without assuming criticality.

396: 

397: 

398: \subsection{Organization of the paper}

399: \label{sec:Organization-paper}

400: 

401: This paper is organized as follows.

402: Sections~\ref{sec:deligne-complexes}

403: and~\ref{sec:herm-holom-deligne} are devoted to expounding some

404: background material for the sake of keeping this paper

405: self-contained and to put the reader in position of reproducing

406: the necessary calculations. Section~\ref{sec:deligne-complexes}

407: contains background facts on Deligne cohomology, paying special

408: attention to the product structures and the cone constructions.

409: We provide some examples and collect some facts about the

410: dilogarithm from the point of view of Deligne cohomology.  The

411: particular model of hermitian holomorphic Deligne cohomology we

412: use later in the paper requires certain constructions available

413: in the literature, and recalled in

414: section~\ref{sec:deligne-complexes}, to be slightly modified in

415: order to obtain a (graded) commutative product. The necessary

416: arguments, being somewhat outside the line of development of the

417: paper are presented in Appendix~\ref{sec:Cones}. Hermitian

418: holomorphic Deligne cohomology is introduced in

419: section~\ref{sec:herm-holom-deligne}. We give the definition as

420: in refs.\ \cite{brymcl:deg4_II} and \cite{bry:quillen}, and then

421: introduce another model which, albeit more complex, has the

422: advantage for us of keeping the metric structure explicit. We

423: have explicitly proved the isomorphism in

424: Lemma~\ref{lemma:qi-dhh-bry}. For the sake of completeness, we

425: give an explicit description of the cocycle determined by a

426: holomorphic line bundle with hermitian metric, and in

427: sect.~\ref{sec:Cup-prod-herm} we explicitly compute the cup

428: product of two metrized line bundles for later usage. Results

429: about the existence of a fiber integration map are mentioned in

430: the paper, so some background material is provided in

431: sect.\ \ref{sec:integr-along-fiber}.

432: 

433: Sections~\ref{sec:Conf-metr-Liouv} and~\ref{sec:conf-metr-herm}

434: form the core of the paper. The direct construction of the

435: Liouville action according to the techniques of refs.\ 

436: \cite{aldtak1997,aldtak2000,math.CV/0204318} is presented in

437: section~\ref{sec:Conf-metr-Liouv}. Since explicit calculations

438: have been presented in great detail in ref.\ 

439: \cite{math.CV/0204318}, and the calculations we need are quite

440: straightforward, we keep details to a minimum. In

441: Proposition~\ref{prop:2} and Corollary~\ref{cor:1} we show that

442: the Liouville action functional computed via descent theory does

443: solve the variational problem. These results have appeared also

444: in ref.\ \cite{math.CV/0204318} and are presented for here

445: completeness, although our choice for the cover is different. The

446: framework of loc.\ cit.\ is that of a Kleinian cover $U\to X\cong

447: U/\Gamma$, where $\Gamma$ is a second kind purely loxodromic

448: geometrically finite Kleinian group, which then we treat in some

449: detail in sect.\ \ref{sec:liouv-funct-bloch}. We illustrate how

450: the genuine Bloch-Wigner function of sect.\ 

451: \ref{sec:remarks-dilogarithm} appears in the descent equations

452: relative to the Kleinian cover. Having observed descent equations

453: close on general cohomological grounds in sect.\ 

454: \ref{sec:solut-vari-probl}, we now point out that for the case of

455: a Kleinian cover this is due to the vanishing of the regulator

456: class for the non-compact $3$-manifold $\HH^3/\Gamma$.

457: 

458: A true geometric construction of the Liouville action, which does

459: not rely on the arguments of sect.\ \ref{sec:Conf-metr-Liouv} to

460: close the descent equations, is carried out in

461: section~\ref{sec:conf-metr-herm}, which contains our main result:

462: We compare the descent calculations with the cup products

463: computed in section~\ref{sec:herm-holom-deligne} and conclude

464: that the quadratic part of the Liouville action is in fact

465: (modulo an area term) the cup square of the metrized holomorphic

466: tangent bundle in Hermitian holomorphic Deligne cohomology, see

467: Theorem~\ref{thm:2}, Corollary~\ref{corollary:1}, and

468: Proposition~\ref{prop:3}. It follows that from this point of view

469: the descent equations close automatically, without the need for

470: special arguments.  We then prove that the cup square is

471: identified with the determinant of cohomology construction in

472: Theorem~\ref{thm:3} and Corollary~\ref{corollary:2}. Auxiliary

473: facts about the homological algebra of cones and conventions on

474: Kleinian groups are stored in the appendices.

475: 

476: Finally, we draw our conclusions in sect.\ 

477: \ref{sec:conclusions-outlook}.

478: 

479: \subsection{Notation and conventions}

480: \label{sec:notation}

481: 

482: If $z$ is a complex number, then $\pi\sb{p} (z) \eqdef \onehalf ( z +

483: (-1)\sp p \Bar z)$, and similarly for any other complex quantity, e.g.

484: complex valued differential forms. If $A$ is a subring of $\RR$, we will

485: use the ``twist'' $A(j) = (2\pi \sqrt{-1})^j\,A$.

486: 

487: If $X$ is a complex manifold, $\sha{X}$ and $\shomega{X}$ denote

488: the de~Rham complexes of smooth $\CC$-valued and holomorphic

489: forms, respectively. We denote by $\she{X}$ the de~Rham complex

490: of sheaves of real valued differential forms and by $\she{X} (j)$

491: the twist $\she{X} \otimes_\RR \RR(j)$. We set $\sho{X} \equiv

492: \shomega[0]{X}$ as usual.  When needed, $\sha[{p,q}]{X}$ will

493: denote the sheaf of smooth $(p,q)$-forms. We use the standard

494: decomposition $d=\del + \delb$ according to types. Furthermore,

495: we introduce the differential operator $d^c = \del -\delb$

496: (contrary to the convention, see, e.g. \cite{lang:arakelov}, we

497: omit the factor $1/(4\pi \sqrt{-1})$). We have $2\del\delb =

498: d^cd$. The operator $d^c$ is an imaginary one, and accordingly,

499: we have the rules

500: \begin{gather*}

501:   d\pi_p(\omega) = \pi_p(d\omega)\\

502:   d^c\pi_p(\omega) = \pi_{p+1}(d^c\omega)

503: \end{gather*}

504: for any complex form $\omega$.

505: 

506: An open cover of $X$ will be denoted by $\cover{U}_X$. If $\{U_i\}_{i\in

507:   I}$ is the corresponding collection of open sets, we denote $U_{ij} =

508: U_i\cap U_j$, $U_{ijk} = U_i\cap U_j\cap U_k$, and so on. We can also

509: consider more general covers $\cover{U}_X = \{ U_i \to X\}_{i\in I}$

510: where the maps are regular coverings in the appropriate category. In

511: this case intersections are replaced by $(n+1)$-fold fibered products

512: \begin{math}

513:   U_{i_0 i_1\dots i_n}= U_{i_0} \times_X\dots \times_X U_{i_n}\,.

514: \end{math}

515: Open coverings fit this more general description, since if $U_i$ and

516: $U_j$ are two open sets, then $U_i\cap U_j = U_i\times_X U_j$, where the

517: fiber product is taken with respect to the inclusion maps. As another

518: example, one can consider regular covering maps $U\to X$ with $\Gamma =

519: \mathrm{Deck}(U/X)$, and in this case, taking the fiber product over $X$

520: $(n+1)$-times, one gets

521: \begin{math}

522:   U\times_X\dots\times_X U = U\times \Gamma \times \dots \times \Gamma\,,

523: \end{math}

524: where the group factor is repeated $n$-times. This includes the cases of

525: Kleinian (and in particular Fuchsian) covers of Riemann surfaces.

526: 

527: The \emph{nerve} of the cover $\cover{U}_X$ is

528: the simplicial object

529: \begin{math}

530:   n\mapsto N_n(\cover{U}_X) = \coprod U_{i_0} \times_X\dots \times_X

531:   U_{i_n}

532: \end{math}

533: where $N_n(\cover{U}_X)$ maps into $N_{n-1}(\cover{U}_X)$ in $(n+1)$

534: ways by forgetting in turn each factor. For open covers this just yields

535: the expected inclusion maps.

536: 

537: If $\sheaf{F}^\bullet$ is a complex of sheaves on $X$, its \cech\ 

538: resolution with respect to a covering $\cover{U}_X\to X$ is the double

539: complex

540: \begin{math}

541:   \CCC^{p,q} (\sheaf{F}) \eqdef \check{C}^q(\cover{U}_X,\sheaf{F}^p)\,,

542: \end{math}

543: where the $q$-cochains with values in $\sheaf{F}^p$ are given by

544: \begin{math}

545: \prod \sheaf{F}^p (U_{i_0} \times_X\dots \times_X U_{i_n})\,.

546: \end{math}

547: The \cech\ coboundary operator is denoted $\deltacheck$. The sign

548: convention we are going to use is that the index along the \cech\ 

549: resolution is the \emph{second} one, so if we denote by $d$ the

550: differential in the complex $\sheaf{F}^\bullet$, the total differential

551: in the total simple complex of $\check{C}^q(\cover{U}_X,\sheaf{F}^p)$

552: will be $D=d\pm \deltacheck$. For open covers we just get the familiar

553: \cech\ (hyper)cohomology. The other interesting example is that of a

554: regular covering map $U\to X$: \cech\ cohomology with respect to this

555: cover is the same as group cohomology for $\Gamma = \mathrm{Deck}(U/X)$

556: with coefficients in the $\Gamma$-module $\sheaf{F}^p(U)$.

557: 

558: The Koszul sign rule that results in a sign being picked whenever two

559: degree indices are formally exchanged is applied. In particular, for

560: \cech\ resolutions of complexes of sheaves, it leads to the following

561: conventions. If $\sheaf{G}^\bullet$ is a second complex of sheaves on

562: $X$, then one defines the cup product

563: \begin{equation*}

564:   \cup : \CCC^{p,q}(\sheaf{F}) \otimes \CCC^{r,s}(\sheaf{G})

565:   \longrightarrow \Check{C}^{q+s}(\cover{U}_X,\sheaf{F}^p\otimes

566:   \sheaf{G}^r) \subset \CCC^{p+r,q+s}(\sheaf{F}\otimes\sheaf{G})

567: \end{equation*}

568: of two elements $\{f_{i_0,\dots,i_q}\}\in \CCC^{p,q}(\sheaf{F})$

569: and $\{g_{j_0,\dots,j_s}\} \in \CCC^{r,s}(\sheaf{G})$ by

570: \begin{equation*}

571:   (-1)^{qr}\,f_{i_0,\dots,i_q}\otimes

572:   g_{i_q,i_{q+1},\dots,i_{q+s}} \,.

573: \end{equation*}

574: 

575: \section*{Acknowledgments}

576: Parts of this work were completed during visits at the

577: International School for Advanced Studies (SISSA) in Trieste,

578: Italy, and at the Department of Mathematics, Instituto Superior

579: T\'ecnico in Lisbon, Portugal. I would like to thank both

580: institutions for support and for creating a friendly and

581: stimulating research environment. I would also like to thank

582: Paolo Aluffi, Phil Bowers, Ugo Bruzzo, Johan Dupont, Leon

583: Takhtajan for illuminating discussions and/or patiently answering

584: my many questions. Also, special thanks are due to the referee

585: for his or her thoroughness and for providing very detailed

586: comments.

587: 

588: \section{Deligne complexes}

589: \label{sec:deligne-complexes}

590: 

591: \subsection{Cup products on cones}

592: \label{sec:cup-products-cones}

593: 

594: Recall that the cone of a map $f: X^\bullet \to Y^\bullet$

595: between two complexes is the complex $C^\bullet (f) =

596: X^\bullet[1]\oplus Y^\bullet$ with differential $d_f(x,y) =

597: (-d\,x, f(x) + d\,y)$, where $[k]$ denotes the shift functor. The

598: cone fits into the exact sequence

599: \begin{equation*}

600:   0 \longrightarrow

601:   Y^\bullet \longrightarrow C^\bullet (f) \longrightarrow

602:   X^\bullet[1] \longrightarrow 0\,.

603: \end{equation*}

604: The following constructions are a special case of those considered by

605: \bei\ in ref.\ \cite{bei:hodge_coho}. Suppose we are given complexes

606: $X\sp\bullet\sb i$, $Y\sp\bullet\sb i$, and $Z\sp\bullet\sb i$ and maps

607: $f\sb i : X\sp\bullet\sb i \rightarrow Z\sp\bullet \sb i$, $g\sb i :

608: Y\sp\bullet\sb i \rightarrow Z\sp\bullet \sb i$, for $i=1,2,3$.  Suppose

609: also that we have product maps $X^\bullet_1\otimes X^\bullet_2

610: \xrightarrow{\cup} X^\bullet_3$, and similarly for $Y^\bullet_i$, and

611: $Z^\bullet_i$, strictly compatible with the $f_i$, $g_i$ in the obvious

612: sense. Then we can consider the cones

613: \begin{equation*}

614:   \cone \big( f_i -g_i \big)[-1] \equiv

615:   \cone \big( X^\bullet_i\oplus Y^\bullet_i \xrightarrow{f_i -

616:   g_i} Z^\bullet_i \big)[-1]\,.

617: \end{equation*}

618: For a real parameter $\alpha$, there is a family of products

619: \begin{equation}

620:   \label{eq:1}

621:   \cone \big( f_1 - g_1 \big)[-1] \otimes

622:   \cone \big( f_2 - g_2 \big)[-1] \xrightarrow{\cup_\alpha}

623:   \cone \big( f_3 - g_3 \big)[-1]

624: \end{equation}

625: determined as follows. For

626: \begin{math}

627:   (x_i,y_i,z_i) \in X^\bullet_i \oplus Y^\bullet_i \oplus

628:   Z^{\bullet -1}_i\,,\;i=1,2\,,

629: \end{math}

630: one defines

631: \begin{equation}

632:   \label{eq:2}

633:   \begin{split}

634:     (x_1,y_1,z_1) \cup_\alpha (x_2,y_2,z_2) =

635:     \Big(&x_1\cup x_2, y_1\cup y_2, \\

636:     &(-1)^{\deg (x_1)}

637:     \big((1-\alpha )f_1(x_1) + \alpha g_1(y_1) \big) \cup z_2 \\

638:     &\quad +z_1\cup \big( \alpha f_2(x_2) +

639:     (1-\alpha)g_2(y_2)\big) \Big)\,.

640:   \end{split}

641: \end{equation}

642: Note that $\deg (x_1)=\deg (x_1,y_1,z_1)$. Checking that

643: $\cup_\alpha$ is a map of complexes is a straightforward routine

644: calculation. Different products for different values $\alpha$ and

645: $\beta$ of the real parameter are homotopic. Explicitly, we have

646: \begin{equation*}

647:   (x_1,y_1,z_1) \cup_\alpha (x_2,y_2,z_2) -

648:   (x_1,y_1,z_1) \cup_\beta (x_2,y_2,z_2)  =

649:   \big( d \, h_{\alpha,\beta} + h_{\alpha,\beta} d \big)

650:   ((x_1,y_1,z_1) \otimes (x_2,y_2,z_2)) \,,

651: \end{equation*}

652: where the homotopy

653: \begin{equation*}

654:   h_{\alpha,\beta} : \Tot \Big( \cone \big( f_1 - g_1\big)[-1]

655:   \otimes \cone \big( f_2 - g_2\big)[-1] \Big)^\bullet

656:   \longrightarrow \cone \big( f_3 - g_3\big)[-1]^{\bullet -1}

657: \end{equation*}

658: is given by the formula

659: \begin{equation}

660:   \label{eq:3}

661:   h_{\alpha,\beta} ((x_1,y_1,z_1) \otimes (x_2,y_2,z_2)) =

662:   (\alpha -\beta) \, (-1)^{\deg (x_1) -1} (0,0,z_1\cup

663:   z_2) \,.

664: \end{equation}

665: If the products

666: \begin{math}

667:   X^\bullet_1 \otimes X^\bullet_2

668:   \xrightarrow{\cup} X^\bullet_3\,,

669: \end{math}

670: etc., are graded commutative, then the swap functor on the tensor

671: product maps the $\cup_\alpha$ product structure on the cones into the

672: $\cup_{1-\alpha}$ structure. Using the homotopy~\eqref{eq:3} it follows

673: at once that there is a well defined graded commutative product in

674: cohomology.

675: 

676: If we do not assume the product structures $X^\bullet_1\otimes

677: X^\bullet_2 \xrightarrow{\cup} X^\bullet_3$, etc., are strictly

678: compatible with the maps $f_i$, $g_i$, some of the preceding

679: constructions must be slightly modified. With an eye toward

680: certain constructions to be carried out later in this paper, let

681: us assume we have compatibility up to homotopy, namely there

682: exist maps

683: \begin{gather*}

684:   h \colon \bigl( X_1\otimes X_2 \bigr)^\bullet

685:   \longrightarrow Z_3^{\bullet -1}\\

686:   k \colon \bigl( Y_1\otimes Y_2 \bigr)^\bullet

687:   \longrightarrow Z_3^{\bullet -1}

688: \end{gather*}

689: such that 

690: \begin{equation}

691:   \label{eq:4}

692:   \begin{gathered}

693:     f_3\circ \cup - \cup \circ (f_1\otimes f_2)

694:     = d\, h + h\,d \\

695:     g_3\circ \cup - \cup \circ (g_1\otimes g_2)

696:     = d\, k + k\,d\,,

697:   \end{gathered}

698: \end{equation}

699: with obvious meaning of the symbols.

700: \begin{lemma}

701:   \label{lemma:mod-cup-prod}

702:   Let $X_i$, $Y_i$, $Z_i$ and the maps $f_i$, $g_i$ be as above. Let

703:   $\alpha$ be a real parameter. We have a product of type~\eqref{eq:1}

704:   for the cones $\cone(f_i-g_i)[-1]$ defined by the following

705:   modification of formula~\eqref{eq:2}:

706:   \begin{equation}

707:     \label{eq:5}

708:     \begin{split}

709:       (x_1,y_1,z_1) \cup_\alpha (x_2,y_2,z_2) =

710:       \Big(&x_1\cup x_2, y_1\cup y_2, \\

711:       &(-1)^{\deg (x_1)}

712:       \big((1-\alpha )f_1(x_1) + \alpha g_1(y_1) \big) \cup z_2 \\

713:       &\quad +z_1\cup \big( \alpha f_2(x_2) +

714:       (1-\alpha)g_2(y_2)\big)\\

715:       &\qquad -h(x_1\otimes x_2) +k(y_1\otimes y_2)

716:       \Big)\,.

717:     \end{split}   

718:   \end{equation}

719:   The product~\eqref{eq:5} is a map of complexes and two products

720:   $\cup_\alpha$ and $\cup_\beta$ are related by the same homotopy

721:   formula~\eqref{eq:3}.

722: \end{lemma}

723: \begin{proof}

724:   Direct verification.

725: \end{proof}

726: This modified framework carries over to the full structure considered by 

727: \bei\ in ref.\ \cite{bei:hodge_coho}. We will still refer to this

728: modified product as the \bei\ product. It is also necessary to relax the 

729: assumption that the products

730: \begin{math}

731:   X^\bullet_1 \otimes X^\bullet_2

732:   \xrightarrow{\cup} X^\bullet_3\,,

733: \end{math}

734: etc., be graded commutative. It is possible to complete all the diagrams

735: so that the permutation of factors in the tensor products still yields a

736: homotopy commutative product~\eqref{eq:5} for the cones. As a

737: consequence the induced product in cohomology will still be graded

738: commutative. Explicit formulas are not needed except to ensure this

739: latter fact, therefore we shall not discuss this matter any further and

740: refer the reader to the appendix, where a brief but explicit treatment

741: can be found.

742: 

743: \subsection{Deligne complexes}

744: \label{sec:deligne-compl}

745: 

746: Let $X$ be a complex manifold. Recall the standard Hodge

747: filtration of $\shomega{X}$:

748: \begin{equation}

749:   \label{eq:6}

750:   F^p\shomega{X} : 0 \longrightarrow \cdots \longrightarrow

751:   \shomega[p]{X} \longrightarrow \cdots \longrightarrow

752:   \shomega[n]{X}\,,

753: \end{equation}

754: where $n=\dim_\CC X$.

755: 

756: The corresponding filtration for the complex of smooth

757: $\CC$-valued forms is defined as follows: denote by $F^p\sha{X}$

758: the subcomplex of $\sha{X}$ comprising forms of type $(r,s)$

759: where $r$ is at least $p$, so that $F^p\sha[k]{X} = \oplus_{r\geq

760:   p} \sha[{r,k-r}]{X}$. Then (cf. \cite{del:hodge_II}) the

761: inclusion $\shomega{X} \hookrightarrow \sha{X}$ is a

762: quasi-isomorphism respecting the filtrations, namely

763: $F^p\shomega{X} \hookrightarrow F^p\sha{X}$, and the latter

764: inclusion induces an isomorphism in cohomology.

765: 

766: If $A$ is a subring of $\RR$, and $\imath$ and $\jmath$ denote

767: the inclusions of $A(p)$ and $F^p\shomega{X}$ into $\shomega{X}$

768: respectively, the $p$-th Deligne complex of sheaves is

769: defined by

770: \begin{equation}

771:   \label{eq:7}

772:   \deligne{A}{p} = \cone \bigl( A(p)_X \oplus F^p\shomega{X}

773:   \xrightarrow{\imath -\jmath} \shomega{X} \bigr)[-1]\,.

774: \end{equation}

775: It is quasi-isomorphic to the complex:

776: \begin{equation}

777:   \label{eq:8}

778:   \cone \bigl( A(p)_X \oplus F^p\sha{X}

779:   \xrightarrow{\imath -\jmath} \sha{X} \bigr)[-1]\,,

780: \end{equation}

781: where $\imath$ and $\jmath$ have the same meaning. We also notice

782: the quasi-isomorphism

783: \begin{equation}

784:   \label{eq:9}

785:   \deligne{A}{p}\lqi

786:   \bigl( A(p)_X 

787:   \overset{\imath}{\longrightarrow} \sho{X}

788:   \overset{d}{\longrightarrow} \shomega[1]{X}

789:   \overset{d}{\longrightarrow} \dotsm

790:   \overset{d}{\longrightarrow} \shomega[{p-1}]{X}\bigr)\,.

791: \end{equation}

792: When $A=\RR$ there are further quasi-isomorphisms, namely

793: \begin{equation*}

794:   \deligne{\RR}{p} \xrightarrow{\simeq} \cone \big(

795:   F^p\shomega{X} \rightarrow \she{X} (p-1)\big) [-1]

796:   \xrightarrow{\simeq} \cone \big( F^p\sha{X} \rightarrow \she{X}

797:   (p-1)\big) [-1]

798: \end{equation*}

799: since the maps

800: \begin{equation*}

801:   \bigl( \RR(p) \rightarrow \shomega{X}\bigr)

802:   \xrightarrow{\simeq} \bigl( \RR(p)

803:   \rightarrow \CC \bigr) \xrightarrow{\simeq} \RR(p-1)

804:   \xrightarrow{\simeq} \she{X}(p-1)

805: \end{equation*}

806: are all quasi-isomorphisms in the derived category, cf.

807: \cite{esn-vie:del}. Here we have used $\CC \cong \RR(p) \oplus

808: \RR(p-1)$. Following op.~cit., we set:

809: \begin{equation}

810:   \label{eq:11}

811:   \deltilde{\RR}{p} \eqdef

812:   \cone \big( F^p\sha{X} \xrightarrow{-\pi\sb{p-1}} \she{X} (p-1)\big)

813:   [-1] \, .  

814: \end{equation}

815: Again, there is an explicit quasi-isomorphism

816: (\cite{MR86h:11103,esn-vie:del}):

817: \begin{equation}

818:   \label{eq:12}

819:   \begin{gathered}

820:     \rho_p : \deligne{\RR}{p} \overset{\simeq}{\longrightarrow}

821:     \deltilde{\RR}{p} \\

822:     \rho_p \lvert_{\RR(p)} = 0\,,\quad \rho_p

823:     \lvert_{F^p\shomega{X}} = \mathit{incl} \,,\quad \rho_p

824:     \lvert_{\shomega{X}} = \pi_{p-1}

825:   \end{gathered}

826: \end{equation}

827: 

828: The \emph{Deligne cohomology groups} of $X$ with coefficients in

829: $A(p)$ are the hypercohomology groups

830: \begin{equation*}

831:   \delH{X}{A}{p} = \HHH^\bullet (X, \deligne{A}{p})\,. 

832: \end{equation*}

833: and clearly, any complex quasi-isomorphic to $\deligne{A}{p}$

834: would do. In order to perform calculations with these cohomology

835: groups we shall normally resort to a \cech\ resolution with

836: respect to an open cover $\cover{U}_X$ of $X$ or an \'etale map

837: $\cover{U}_X \to X$, e.g. a regular cover with deck group

838: $\Gamma$.

839: 

840: One of the important properties of Deligne cohomology is the existence

841: of a graded commutative cup product

842: \begin{equation}

843:   \label{eq:13}

844:   \delH[i]{X}{A}{p} \otimes \delH[j]{X}{A}{q}

845:   \xrightarrow{\cup} \delH[i+j]{X}{A}{p+q}\,,

846: \end{equation}

847: which follows from the existence of the \bei{} cup product at the

848: level of Deligne complexes whose construction was recalled above.

849: There are products $A(p) \otimes A(q) \to A(p+q)$ and

850: $F^p\shomega{X} \otimes F^q\shomega{X} \to F^{p+q}\shomega{X}$,

851: plus the obvious (wedge) product on $\shomega{X}$, thus it

852: follows from the cone version~\eqref{eq:7} that the Deligne

853: complexes come equipped with the \bei{} product, and therefore

854: the Deligne cohomology groups inherit the graded commutative cup

855: product~\eqref{eq:13}. The explicit form, that is, the

856: translation of~\eqref{eq:2} to the case at hand can be found

857: in~\cite{esn-vie:del}. The explicit form of the cup product for

858: the complex~\eqref{eq:11} as computed in \cite{MR86h:11103} (see

859: also \cite{esn-vie:del}) will be needed in the sequel. Let

860: $(\omega\sb 1, \eta\sb 1)$ be an element of degree $n$ in

861: $\deltilde{\RR}{p}$---this means that $\omega\sb 1\in

862: F^p\shomega[n]{X}$ and $\eta\sb 1\in \she[n-1]{X}(p-1)$---and

863: $(\omega\sb 2, \eta\sb 2)$ any element in $\deltilde{\RR}{q}$.

864: The product is defined by the formula:

865: \begin{equation}

866:   \label{eq:15}

867:   (\omega_1,\eta_1) \,\Tilde\cup\, (\omega_2,\eta_2) = \bigl(

868:   \omega\sb 1 \wedge \omega\sb 2 ,

869:   (-1)^n\, \pi\sb p\omega\sb 1 \wedge \eta\sb 2

870:   +\eta\sb 1\wedge \pi\sb q \omega\sb 2 \bigr)\,.

871: \end{equation}

872: The product $\Tilde\cup$ is a morphism of complexes and (modulo

873: the quasi-isomorphisms $\rho_p$) is homotopic to the \bei\ 

874: products $\cup_\alpha$ on the complexes $\deligne{\RR}{p}$.

875: Specifically, if we denote an element of $\deligne{\RR}{p}$ by

876: the triple $(r,f,\omega)$, where $r\in \RR(p)_X$, $f\in

877: F^{p}\!\shomega{X}$, and $\omega \in \shomega{X}$, the homotopy

878: between $\Tilde\cup \circ (\rho_p\otimes \rho_q)$ and

879: $\rho_{p+q}\circ \cup_\alpha$ is given by

880: \begin{equation}

881:   \label{eq:16}

882:   \Tilde h ((r,f,\omega)\otimes (r',f',\omega'))

883:   = (-1)^{\deg \omega} \bigl(0,

884:   (1-\alpha)\, \pi_p\omega\wedge\pi_{q-1}\omega'

885:   -\alpha\, \pi_{p-1}\omega\wedge\pi_{q}\omega'

886:   \bigr)\,.

887: \end{equation}

888: 

889: 

890: \subsection{Examples}

891: \label{sec:examples}

892: 

893: \subsubsection{}

894: Let $A=\ZZ$. From~\eqref{eq:9} we have $\deligne{\ZZ}{1} \qi

895: \sho{X}^\times[-1]$ via the standard exponential sequence, so

896: that

897: \begin{math}

898:   \delH[k]{X}{\ZZ}{1} \cong H^{k -1}(X, \sho{X}^\times)\,.

899: \end{math}

900: In particular $\delH[1]{X}{\ZZ}{1}\cong H^0(X,\sho{X}^\times)$,

901: the global invertibles on $X$, and $\delH[2]{X}{\ZZ}{1} \cong

902: \pic{X}$, the Picard group of line bundles over $X$.

903:   

904: If an open cover $\{U_i\}_{i\in I}$ of $X$ is chosen, the class

905: of a line bundle $L$ in $\delH[2]{X}{\ZZ}{1}$ can be represented

906: via a \cech\ resolution by the cocycle $(f_{ij},c_{ijk})$, where

907: $f_{ij}\in \sho{X}(U_{ij})$, and $c_{ijk}\in \ZZ(1)_X(U_{ijk})$.

908: Thus the functions $f_{ij}$ should be interpreted as

909: \emph{logarithms} of the corresponding transition functions for

910: $L$. Then, the collection $c_{ijk}=(\deltacheck f)_{ijk}$

911: provides a representative for the first Chern Class $c_1(L)$.

912: Similarly, an invertible function $f$ would be described by the

913: collection $f_i$ of its logarithms on each open $U_i$, plus the

914: ``integers'' $m_{ij} = f_j -f_i\in \ZZ(1)$ on each $U_{ij}$.

915:   

916: \subsubsection{}

917: Still using the exponential sequence, $\deligne{\ZZ}{2} \qi

918: \bigl( \sho{X}^\times \xrightarrow{d\log}

919: \shomega[1]{X}\bigr)[-1]$. Thus $\delH[2]{X}{\ZZ}{2}$ is the

920: group of isomorphism classes of holomorphic line bundles with

921: (holomorphic) connection. Using the (in fact, any) product

922: $\deligne{\ZZ}{1} \otimes \deligne{\ZZ}{1} \to \deligne{\ZZ}{2}$,

923: the cup product of two global invertible holomorphic functions

924: $f$ and $g$ on $X$ yields a line bundle with connection---the

925: tame symbol---denoted by $\tame{f}{g}$ whose class is in

926: $\delH[2]{X}{\ZZ}{2}$,

927: see~\cite{del:symbole,bloch:dilog_lie,rama:reg_hei}. Higher cup

928: products in this spirit have been studied

929: in~\cite{brymcl:deg4_II}.

930:   

931: \subsubsection{}

932: If $A=\RR$, we have

933: \begin{math}

934:   \delH[2p]{X}{\RR}{p} = H^{2p}(X,\RR(p)) \cap H^{p,p}(X)\,.

935: \end{math}

936: The \cech\ representative $(c_{ijk},f_{ij})$ of a class in

937: $\delH[2]{X}{\ZZ}{1}$ mentioned above maps to the cocycle $(-d{}

938: f_{ij}, -\abs{f_{ij}})$ under~\eqref{eq:12}. Taking into account

939: that the $f_{ij}$ are the logarithms of the transition functions,

940: the corresponding $(1,1)$ class would be given by the associated

941: canonical connection, see sec.~\ref{sec:line-bun}.

942:   

943: \subsubsection{}\label{sec:ex_4}

944: $\delH[1]{X}{\RR}{1}$ is the group of real valued functions $f$ on $X$

945: such that there exists a holomorphic one-form $\omega$ such that $\pi\sb

946: 0\omega = df$. In other words it is the group of those smooth functions

947: $f$ such that $\del f$ is holomorphic, which amounts to say that such an

948: $f$ itself is harmonic.

949: 

950: 

951: \subsection{Remarks on the cup product $f\cup g$}

952: \label{sec:remarks-dilogarithm}

953: 

954: It is convenient to consider the case of the cup product of two

955: invertible functions $f$ and $g$ in various complexes in more detail,

956: and to introduce some related notions we shall need later.

957: 

958: As observed, $\deligne{\ZZ}{1} \qi \sho{X}^\times[-1]$ and an

959: invertible function $f$ can be considered as an element of

960: $\delH[1]{X}{\ZZ}{1}$. Therefore, via~\eqref{eq:12}, it induces

961: $\rho_1(f) \in \delH[1]{X}{\RR}{1}$ represented by $(d\log f,

962: \log\,\abs{f})$. (Note that $\log \abs{f}$ fits the description

963: of $\delH[1]{X}{\RR}{1}$ in \ref{sec:ex_4}.) Given two such $f$

964: and $g$, the expression for the cup product~\eqref{eq:15} gives

965: the following element of $\delH[2]{X}{\RR}{2}$:

966: \begin{equation}

967:   \label{eq:17}

968:   \rho_1(f)\,\Tilde\cup\, \rho_1(g) =

969:   \bigl( d\log f \wedge d\log g,

970:   -\pi_1(d\log f)\, \log\,\abs{g}

971:   +\log\,\abs{f}\, \pi_1(d\log g) \bigr)\,.

972: \end{equation}

973: The first term is obviously zero when $X$ is a curve. Given $f$

974: and $g$, invertible on $X$, let us define the imaginary $1$-form:

975: \begin{equation}

976:   \label{eq:18}

977:   r_2(f,g) = \pi_1(d\log f)\, \log\,\abs{g}

978:   -\log\,\abs{f}\, \pi_1(d\log g)\,.

979: \end{equation}

980: On the other hand, the cup product of $f$ and $g$ as elements of

981: $\delH[1]{X}{\ZZ}{1}$ yields an element $f\cup g$ of

982: $\delH[2]{X}{\ZZ}{2}$ represented by

983: \begin{math}

984:   \bigl( d\log f \wedge d\log g, \log f\; d\log g \bigr)

985: \end{math}

986: (if we use the $\cup_0$ product) and this maps via $\rho_2$ to

987: the element

988: \begin{equation*}

989:   \bigl( d\log f \wedge d\log g, -\pi_1(\log f\; d\log g)

990:   \bigr)\,.

991: \end{equation*}

992: This is equal to~\eqref{eq:18} only up to homotopy. Indeed, using

993: \begin{math}

994:   \pi_{p+q-1}(a\wedge b) = \pi_p(a)\wedge \pi_{q-1}(b) +

995:   \pi_{p-1}(a)\wedge \pi_q(b)\,,

996: \end{math}

997: we find

998: \begin{equation}

999:   \label{eq:19}

1000:   r_2(f,g) = d \bigl( \pi_1(\log f)\, \log\,\abs{g} \bigr)

1001:   -\pi_1 \bigl( \log f\; d\log g \bigr)\,,

1002: \end{equation}

1003: where the first term is just the explicit homotopy as computed

1004: from~\eqref{eq:16}.

1005: 

1006: Recall that the \emph{tame symbol}

1007: (\cite{del:symbole,bloch:dilog_lie}) $\tame{f}{g}$ associated to

1008: $f$ and $g$ is the line bundle with connection determined (up to

1009: isomorphism) by the class $f\cup g$. A ``Bloch dilogarithm''

1010: (\cite{esn-vie:del}) is (the logarithm of) a horizontal

1011: trivializing section, namely a function $L$ on $U\subset X$

1012: satisfying the equation

1013: \begin{equation*}

1014:   dL = -\log f\; d\log g\,.

1015: \end{equation*}

1016: Thus $L$ realizes the isomorphism $\tame{f}{g}\cong \sho{X}$ over

1017: $U\subset X$.

1018: 

1019: Thus a Bloch dilogarithm will only locally be available, in

1020: general. However, when $g=1-f$ then $\tame{1-f}{f}$ is

1021: \emph{globally} trivial~\cite{bloch:dilog_lie,esn-vie:del}, i.e.

1022: $\tame{1-f}{f}\cong \sho{X}$. This is the the \emph{Steinberg

1023:   relation} satisfied by the Tame symbol. It can be deduced from

1024: the following universal case. Set $f=z$, $X=\PP^1 \setminus

1025: \{0,1,\infty\}$, consequently $g=1-z$. Then $L$ is identified

1026: with the classical Euler dilogarithm $\li$, namely

1027: \begin{equation*}

1028:   \li (z) = -\int_0^z \log (1-t) \frac{dt}{t}\,,

1029: \end{equation*}

1030: see~\cite{del:symbole} and~\cite{MR94k:19002} for details.

1031: 

1032: On $\PP^1 \setminus \lbrace 0,1,\infty\rbrace$ the classical

1033: dilogarithm has a single valued parter, denoted $\bwd$,

1034: introduced by Bloch and Wigner:

1035: \begin{equation}

1036:   \label{eq:20}

1037:   \bwd (z) = \arg (1-z) \log\,\abs{z} + \im\li (z)\,.

1038: \end{equation}

1039: $\bwd$ is real-analytic on $\PP^1 \setminus \{0,1,\infty\}$ and

1040: extends continuously to $\PP^1$. That it is single-valued can be

1041: verified as follows. Choose a branch of the logarithm, say the

1042: principal one, to define $\li$ (and $\bwd$) on the cut plane $\CC

1043: \setminus (-\infty,0] \cup [1,\infty)$. Then one shows that the

1044: expression~\eqref{eq:20} is in fact single-valued by analytic

1045: continuation along paths based, say, at $1/2\in \CC$, and winding

1046: around the points $0$ and $1\in\CC$. Explicit computations can be

1047: found in ref.~\cite{MR2001i:11082}.

1048: 

1049: It is convenient to introduce

1050: \begin{equation*}

1051:   \bwli (z) = \sqrt{-1}\, \bwd (z) \,,

1052: \end{equation*}

1053: so that

1054: \begin{equation*}

1055:   d\bwli = r_2(1-z,z)\,.

1056: \end{equation*}

1057: More generally, if $L$ trivializes $\tame{f}{g}$ over $U$ in the

1058: sense explained above, we can associate a function $\bwli (f,g)$

1059: over $U$ such that

1060: \begin{equation*}

1061:   d \bwli(f,g) = r_2(f,g)

1062: \end{equation*}

1063: via the position

1064: \begin{equation}

1065:   \label{eq:21}

1066:   \bwli (f,g) = \pi_1(\log f)\, \log\,\abs{g} +\im L\,.

1067: \end{equation}

1068: 

1069: 

1070: \section{Constructions in hermitian holomorphic Deligne cohomology}

1071: \label{sec:herm-holom-deligne}

1072: 

1073: In this section we recall the definition of hermitian holomorphic

1074: Deligne cohomology. In ref.~\cite{bry:quillen} Brylinski

1075: introduced certain complexes $C(l)^\bullet$, for a positive

1076: integer $l$, in order to compare the \bei{}-Chern classes of a

1077: holomorphic vector bundle $E$ with the Cheeger-Chern-Simons

1078: classes determined by $(E,\nabla)$, where $\nabla$ is the

1079: \emph{canonical connection,} namely the unique connection

1080: compatible with both the holomorphic and hermitian structures.

1081: The cohomology groups determined by these complexes are aptly

1082: named \emph{Hermitian Holomorphic Deligne cohomology groups.} For

1083: a holomorphic \emph{line} bundle equipped with the canonical

1084: connection, the complex $C(1)^\bullet$ encodes the reduction of

1085: the structure group from $\CC$ to $\TT$ afforded by the hermitian

1086: fiber metric.

1087: 

1088: In the following we will need to compute Hermitian Holomorphic

1089: Deligne cohomology by means of different---but

1090: quasi-isomorphic---sheaf complexes we denote $\dhh{l}$. These

1091: complexes are tailored to a direct description of a metrized line

1092: bundle in terms of local representatives of the hermitian fiber

1093: metric. Since the two constructions are related by a

1094: quasi-isomorphism, the resulting cohomology groups are the same.

1095: 

1096: \subsection{Hermitian holomorphic Deligne cohomology}

1097: \label{sec:herm-holom-deligne-1}

1098: 

1099: In ref.~\cite{bry:quillen}, where Brylinski introduces the

1100: complexes:

1101: \begin{equation}

1102:   \label{eq:24}

1103:   C(l)^\bullet = \cone \bigl(

1104:   \ZZ(l)_X \oplus (F^l\!\sha{X}\cap \sigma^{2l}\she{X}(l))

1105:   \longrightarrow \she{X}(l)

1106:   \bigr)[-1]\,,

1107: \end{equation}

1108: where $\sigma^p$ denotes the (sharp) truncation in degree $p$,

1109: namely for a complex $\sheaf{F}^\bullet$, $\sigma^p\sheaf{F}^k$

1110: is zero for $k<p$ and equal to $\sheaf{F}^k$ for $k\geq p$.

1111: In~\eqref{eq:24} we take the cone of the difference between the

1112: two inclusions. We have the following:

1113: \begin{definition}[\cite{bry:quillen}]

1114:   \label{def:1}

1115:   The hermitian holomorphic Deligne cohomology groups are the

1116:   hypercohomology groups of the complexes~\eqref{eq:24}:

1117:   \begin{equation}

1118:     \label{eq:23}

1119:     \dhhH[p]{X}{l} \eqdef \HHH^{\,p}(X,C(l)^\bullet)\,.

1120:   \end{equation}

1121: \end{definition}

1122: The complexes~\eqref{eq:24} are expressed as cones, and therefore

1123: admit a (standard) \bei\ product~\eqref{eq:2}. (The wedge product

1124: induces cup products on both $\she{X}(l)$ and $F^l\!\sha{X}\cap

1125: \sigma^{2l}\she{X}(l)$.)\footnote{It appears signs should be

1126:   adjusted in the product formula quoted in

1127:   ref.~\cite{bry:quillen}, and that using~\eqref{eq:2} is more

1128:   appropriate.} It follows there is a graded commutative product

1129: on cohomology:

1130: \begin{equation}

1131:   \label{eq:27}

1132:   \dhhH[i]{X}{l} \otimes \dhhH[j]{X}{k}

1133:   \overset{\cup}{\longrightarrow}

1134:   \dhhH[i+j]{X}{l+k}\,.

1135: \end{equation}

1136: Also, from the standard cone exact sequences we get

1137: (cf.\ ref.~\cite{bry:quillen}):

1138: \begin{equation}

1139:   \label{eq:84}

1140:   \dotsm \longrightarrow H^{2l-1}(X,\RR (l))

1141:   \longrightarrow \dhhH[2l]{X}{l} \longrightarrow

1142:   H^{2l}(X,\ZZ (l)) \oplus A^{(l,l)}(X)_{\RR(l)}

1143:   \longrightarrow H^{2l}(X,\RR (l))

1144:   \longrightarrow \dotsm

1145: \end{equation}

1146: where $A^{(l,l)}(X)_{\RR(l)}$ denotes the space of smooth

1147: $\RR(l)$-valued global $(l,l)$-forms on $X$. Thus we see

1148: hermitian holormorphic classes are $\ZZ(l)$-valued classes

1149: represented by (necessarily) closed $\RR(l)$-valued $2l$-forms of

1150: pure type $(l,l)$.  For a line bundle this corresponds to a

1151: structure group reduction from $\sho{X}$ to $\she[0]{X}$, namely

1152: $\TT$-valued sections, at the same time controlling the Hodge

1153: type of the resulting class, cf.\ ref.~\cite{bry:quillen}

1154: 

1155: Later (cf. sec.~\ref{sec:line-bun}) we will want to work with the

1156: hermitian structure on a holomorphic line bundle, together with

1157: the imaginary $(1,1)$-form built from the canonical connection,

1158: directly in the holomorphic frame. To carry this out in general

1159: for $(l,l)$ classes, we introduce the complex

1160: \begin{equation}

1161:   \label{eq:22}

1162:   \dhh{l} = \cone \bigl(

1163:   \deligne{\ZZ}{l}\oplus (F^l\!\sha{X}\cap \sigma^{2l}\she{X}(l))

1164:   \longrightarrow \deltilde{\RR}{l}

1165:   \bigr)[-1]\,.

1166: \end{equation}

1167: The map

1168: \begin{math}

1169:   \deligne{\ZZ}{l} \to \deltilde{\RR}{l}

1170: \end{math}

1171: is the composite of the obvious map

1172: \begin{math}

1173:   \deligne{\ZZ}{l} \to \deligne{\RR}{l}

1174: \end{math}

1175: with the quasi-isomorphism $\rho_{\,l}$ defined by~\eqref{eq:12}. We

1176: will simply denote it by $\rho_l$ in the sequel, suppressing the first

1177: morphism in the notation. The map

1178: \begin{math}

1179:   (F^l\!\sha{X}\cap \sigma^{2l}\she{X}(l)) \to \deltilde{\RR}{l}

1180:   = \cone \bigl( F^l\!\sha{X} \to \she{X}(l-1) \bigr)[-1]

1181: \end{math}

1182: is induced by the inclusion of $(F^l\!\sha{X}\cap

1183: \sigma^{2l}\she{X}(l))$ into $F^l\!\sha{X}$. In~\eqref{eq:22} we

1184: take the cone of the difference between these two maps.  The

1185: complex~\eqref{eq:22} offers another model for Hermitian

1186: holomorphic Deligne cohomology. Indeed we have:

1187: \begin{lemma}

1188:   \label{lemma:qi-dhh-bry}

1189:   The complexes $C(l)^\bullet$ and $\dhh{l}$ are

1190:   quasi-isomorphic.

1191: \end{lemma}

1192: \begin{proof}

1193:   By elementary manipulation of cones

1194:   \begin{equation*}

1195:     \dhh{l}= \cone \Bigl(

1196:     F^l\!\sha{X}\cap \sigma^{2l}\she{X}(l) \to

1197:     \cone \bigl( \deligne{\ZZ}{l} \to

1198:     \deltilde{\RR}{l} \bigr) \Bigr)[-1]

1199:   \end{equation*}

1200:   and clearly:

1201:   \begin{math}

1202:     \cone \bigl( \deligne{\ZZ}{l} \to \deltilde{\RR}{l} \bigr)

1203:     \qi \cone \bigl( \ZZ(l)_X \to \RR(l)_X \bigr) \qi \cone

1204:     \bigl( \ZZ(l)_X \to \she{X}(l) \bigr)\,,

1205:   \end{math}

1206:   where all arrows are quasi-isomorphisms. Thus

1207:   \begin{align*}

1208:     \dhh{l} &\lqi \cone \Bigl( F^l\!\sha{X}\cap

1209:     \sigma^{2l}\she{X}(l) \to

1210:     \cone \bigl( \ZZ(l)_X \to \she{X}(l) \bigr) \Bigr)[-1] \\

1211:     &= \cone \bigl( \ZZ(l)_X \oplus (F^l\!\sha{X}\cap

1212:     \sigma^{2l}\she{X}(l)) \to \she{X}(l) \bigr)[-1] \\ &\equiv

1213:     C(l)\,,

1214:   \end{align*}

1215:   as wanted.

1216: \end{proof}

1217: It follows from Lemma~\ref{lemma:qi-dhh-bry} that

1218: \begin{math}

1219:   \HHH^p(X,\dhh{l}) \cong \dhhH[p]{X}{l}\,,

1220: \end{math}

1221: so we can use either complex to compute the Hermitian-holomorphic

1222: Deligne cohomology groups. 

1223: 

1224: Again from the cone exact sequence applied to~\eqref{eq:22}, we

1225: see the groups $\dhhH{X}{l}$ also satisfy the exact sequence

1226: \begin{equation}

1227:   \label{eq:25}

1228:   \dotsm \longrightarrow \delH[2l-1]{X}{\RR}{l}

1229:   \longrightarrow

1230:   \dhhH[2l]{X}{l} \longrightarrow

1231:   \delH[{2l}]{X}{\ZZ}{l} \oplus A^{(l,l)}(X)_{\RR(l)}

1232:   \longrightarrow \delH[{2l}]{X}{\RR}{l}

1233:   \longrightarrow \dotsm

1234: \end{equation}

1235: which we can rewrite as

1236: \begin{equation*}

1237:   \dotsm \longrightarrow \delH[2l-1]{X}{\RR}{l}

1238:   \longrightarrow \dhhH[2l]{X}{l} 

1239:   \longrightarrow \delH[{2l}]{X}{\ZZ}{l} \oplus

1240:   A^{(l,l)}(X)_{\RR(l)} \longrightarrow H^{2l}(X,\RR(l))\cap

1241:   H^{l,l}(X) \longrightarrow \dotsm

1242: \end{equation*}

1243: Thus the elements of $\dhhH[2l]{X}{l}$ map onto those

1244: $(l,l)$-forms representing the Hodge classes corresponding to

1245: $\delH[2l]{X}{\ZZ}{l}$.

1246: 

1247: The complexes $\deligne{\ZZ}{l}$, $\deltilde{\RR}{l}$, and

1248: $F^l\!\sha{X}\cap \sigma^{2l}\she{X}(l)$ appearing in the

1249: cone~\eqref{eq:22}, all have cup products. It follows that we

1250: have the \bei{} family of products

1251: \begin{equation}

1252:   \label{eq:26}

1253:   \dhh{l} \otimes \dhh{k} \overset{\cup_\alpha}{\longrightarrow} 

1254:   \dhh{l+k} \,.

1255: \end{equation}

1256: \begin{remark}

1257:   The product $\cup_\alpha$ in eq.\eqref{eq:26} should be

1258:   intended in the modified version provided by~\eqref{eq:5},

1259:   since the complexes $\deligne{\ZZ}{l}$ and $\deltilde{\RR}{l}$

1260:   have product structures that are compatible with the map

1261:   $\rho_l$ only up to homotopy (given by formula~\eqref{eq:16}).

1262:   Moreover, these complexes have product structures that are

1263:   themselves graded commutative up to homotopy: that the

1264:   product~\eqref{eq:26} is graded commutative up to homotopy

1265:   follows from Proposition~\ref{prop:4} in the appendix.

1266: \end{remark}

1267: 

1268: \subsection{Hermitian holomorphic line bundles}

1269: \label{sec:line-bun}

1270: 

1271: A hermitian holomorphic line bundle or, equivalently, a metrized

1272: line bundle, cf. \cite{lang:arakelov}, is a holomorphic line

1273: bundle $L$ over $X$ together with a hermitian fiber metric $\rho

1274: : L \rightarrow \RR_{\geq 0}$. As a rule, we will not distinguish

1275: $L$ and its sheaf of holomorphic sections. We will also use the

1276: alternate notation $\abs{s}_\rho$ to denote the length of a local

1277: section $s$ of $L$ with respect to $\rho$. Metrized line bundles

1278: can be tensor multiplied and an inverse is defined, see. op. cit.

1279: An isomorphism of metrized line bundles $(L,\rho)$ and

1280: $(L^\prime,\rho^\prime)$ is defined in the obvious way, namely it

1281: is a map $\phi : L \to L^\prime$ such that $\abs{s}_\rho =

1282: \abs{\phi(s)}_{\rho^\prime}$ for some local section $s$ of $L$.

1283: We denote by $\widehat{\pic{X}}$ the group of isomorphism classes

1284: of metrized line bundles.

1285: 

1286: If $L$ is trivialized over a \cech\ cover $\cover{U}_X =

1287: \{U_i\}_{i\in I}$ by sections $s_i$, then as usual we obtain the

1288: cocycle of transition functions $g_{ij}\in \sho{X}^\times

1289: (U_i\cap U_j)$ by writing $s_j = s_i\, g_{ij}$.  Then if

1290: $(L,\rho)$ is a metrized line bundle, we can define the positive

1291: function $\rho_i = \abs{s_i}^2_\rho$, namely the local

1292: representative of the hermitian structure with respect to the

1293: given trivialization. It follows that the various local

1294: representatives satisfy

1295: \begin{equation}

1296:   \label{eq:29}

1297:   \rho_j = \rho_i \, \abs{g_{ij}}^2\,.

1298: \end{equation}

1299: Let us work out the local version of the isomorphism introduced

1300: above. Let $s'_i$ be a local section of $L'$ over $U_i$.

1301: Introduce analogous (primed) quantities for $L'$ as we just did

1302: for $L$. Given the isomorphism $\phi: L\to L'$ we have $\phi(s_i)

1303: = s'_i\, f_i$, for some $f_i\in \sho{X}^\times(U_i)$. Then we

1304: find $f_i\, g_{ij} = g'_{ij}\, f_j$ and $\rho_i = \rho'_i\,

1305: \abs{f_i}^2$.

1306: 

1307: Still working with respect to the chosen cover $\cover{U}_X$, a

1308: \emph{connection} compatible with the holomorphic structure is

1309: the datum of a collection of $(1,0)$-forms $\xi_i \in

1310: \sha[{1,0}]{X}(U_i)$ satisfying

1311: \begin{equation}

1312:   \label{eq:30}

1313:   \xi_j - \xi_i = d\log g_{ij}\,.

1314: \end{equation}

1315: Note for future reference that $\sha[{1,0}]{X} =

1316: F^1\!\sha[1]{X}$. The connection is compatible with the hermitian

1317: metric if

1318: \begin{equation}

1319:   \label{eq:31}

1320:   \pi_0(\xi_i) = \onehalf d\log \rho_i\,.

1321: \end{equation}

1322: Using $d=\del +\delb$ and decomposition with respect to

1323: $(p,q)$-types, we find the familiar relation

1324: \begin{equation}

1325:   \label{eq:32}

1326:   \xi_i = \del \log \rho_i

1327: \end{equation}

1328: for the \emph{unique} connection compatible with both the complex

1329: and hermitian structures \cite{gh:alg_geom}.

1330: 

1331: The global $2$-form

1332: \begin{equation}

1333:   \label{eq:33}

1334:   c_1(\rho) = \delb\del\log \rho_i

1335: \end{equation}

1336: represents the first Chern class of $L$ in $H^2(X,\RR(1))$.

1337: Actually, the class of $c_1(\rho)$ is a pure Hodge class in

1338: $H^{1,1}(X)$ and, according to the examples, it is the image of

1339: the first Chern class of $L$ under the map $\delH[2]{X}{\ZZ}{1}

1340: \to H^2_\mathcal{D}(X,\RR(1))$ induced by $\ZZ(1) \to \RR(1)$.

1341: Observe that $c_1(\rho) = c_1(\rho')$ under the isomorphism

1342: considered above.

1343: 

1344: The following proposition can be found, for example, in

1345: refs.~\cite{brymcl:deg4_II,bry:quillen}. (It apparently was

1346: observed first by Deligne, cf.~\cite{esn:char}.) The proof is

1347: based on writing out an explicit cocycle in a \cech\ resolution

1348: of $C(1)^\bullet$ or $\dhh{1}$. This calculation will be needed

1349: later on with the complex $\dhh{1}$, so we provide a proof here.

1350: \begin{proposition}

1351:   \label{prop:1}

1352:   \begin{equation*}

1353:       \widehat{\pic{X}} \cong \dhhH[2]{X}{1}

1354:   \end{equation*}

1355: \end{proposition}

1356: \begin{proof}

1357:   Recall that $\dhh{1}$ is the cone of the map $\rho_1 -\jmath$,

1358:   where $\rho_1: \deligne{\ZZ}{1} \to \deltilde{\RR}{1}$, and

1359:   \begin{math}

1360:     \jmath : F^1\!\sha{X}\cap \sigma^{2}\she{X}(1) \to

1361:     \deltilde{\RR}{1}\,.

1362:   \end{math}

1363:   By unraveling the structure of all the cones involved we have:

1364:   \begin{equation*}

1365:     \begin{CD}

1366:       \ZZ(1)_X @>>> \shomega[1]{X}\oplus\sho{X} @>>>

1367:       \shomega[2]{X}\oplus \shomega[1]{X} @>>>

1368:       \shomega[3]{X}\oplus \shomega[2]{X} @>>> \dotsm \\

1369:       & & @VV{\imath\oplus\pi_0}V @VV{\imath\oplus\pi_0}V

1370:       @VV{\imath\oplus\pi_0}V \\

1371:       & & F^1\!\sha[1]{X} \oplus \she[0]{X} @>>> F^1\!\sha[2]{X}

1372:       \oplus \she[1]{X} @>>>

1373:       F^1\!\sha[3]{X} \oplus \she[2]{X} @>>> \dotsm \\

1374:       & & & & @AA{\jmath\oplus 0}A @AA{\jmath\oplus 0}A \\

1375:       & & & & F^1\!\sha[2]{X}\cap \she[2]{X}(1) @>>>

1376:       F^1\!\sha[3]{X}\cap \she[3]{X}(1) @>>> \dotsm

1377:     \end{CD}

1378:   \end{equation*}

1379:   With respect to this diagram, an element of total degree $2$

1380:   can be written in the form:

1381:   \begin{equation}

1382:     \label{eq:37}

1383:     \setlength{\extrarowheight}{4pt}

1384:     \begin{array}{c|c|c}

1385:       c_{ijk} & -\;d \log\,g_{ij}\oplus \log\,g_{ij} & 0 \\

1386:       \hline

1387:       0 & \xi_i \oplus \sigma_i & \mathsf{X} \\

1388:       \hline

1389:       0 & 0 & \eta_i

1390:     \end{array}

1391:   \end{equation}

1392:   for $\xi_i\in F^1\!\sha[1]{X}(U_i)$, $\sigma_i \in

1393:   \she[0]{X}(U_i)$ and $\eta_i\in (F^1\!\sha[2]{X}\cap

1394:   \she[2]{X}(1))(U_i)$. To make sense out of~\eqref{eq:37}, note

1395:   that each entry is an element of the object in the

1396:   corresponding position in the left $3\times 3$ part of the

1397:   previous diagram. Then since the total degree is $2$, the

1398:   degree of each element \emph{in the complex} $\dhh{1}$ is $2$

1399:   minus the \cech\ degree as found in~\eqref{eq:37}. The top line

1400:   is the class of $L$ in $\deligne[2]{\ZZ}{1}$. Finally, the

1401:   entry marked $\mathsf{X}$ means there is no applicable

1402:   element---it would have degree $3$.

1403:   

1404:   A totally routine calculation shows that~\eqref{eq:37} is a

1405:   degree $2$ cocycle if and only if

1406:   relations~\eqref{eq:30},~\eqref{eq:31} and $\eta_i = d\xi_i =

1407:   2\,\delb\del \sigma_i$ are satisfied (with $\sigma_i = \onehalf

1408:   \log \rho_i$). Thus $\eta_i = c_1(\rho)\vert_{U_i}$ and all the

1409:   relations defining a metrized line bundle with its canonical

1410:   connection are satisfied.  The verification that adding an

1411:   appropriate coboundary to the cocycle leads to an isomorphic

1412:   metrized bundle in the sense explained above is also routine.

1413:   Finally, the correspondence between component-wise addition of

1414:   cocycles modulo coboundaries and the group operation in

1415:   $\widehat{\pic{X}}$ is again a direct verification.

1416: \end{proof}

1417: From the proof of Proposition~\ref{prop:1} we have that the

1418: cocycle representing $(L,\rho)$ in the \cech\ resolution of

1419: $\dhh{1}$ has the expression:

1420: \begin{equation}

1421:   \label{eq:38}

1422:   \setlength{\extrarowheight}{4pt}

1423:   \begin{array}{c|c|c}

1424:     c_{ijk} & -\;d \log\,g_{ij}\oplus \log\,g_{ij} & 0 \\

1425:     \hline

1426:     0 & \del \log\rho_i \oplus \onehalf \log \rho_i & \mathsf{X} \\

1427:     \hline

1428:     0 & 0 & \delb\del\,\log \rho_i 

1429:   \end{array}

1430: \end{equation}

1431: This cocycle is expressed purely in terms of holomorphic and

1432: metric data (the local expression for the fiber metric $\rho$),

1433: and it also explicitly encodes the canonical connection and its

1434: curvature.

1435: 

1436: By way of comparison, a degree two cocycle in the \cech\ 

1437: resolution of $C(1)^\bullet$ with respect to the same cover would

1438: be given by a triple

1439: \begin{equation*}

1440:   \bigl( \eta_i\oplus \sigma_i, \log u_{ij}, c_{ijk} \bigr)\,,

1441: \end{equation*}

1442: where $\eta_i\in \sha[(1,1)]{X}(U_i)_{\RR(1)}$, $\sigma_i \in

1443: \she[1]{X}(1)(U_i)$, $\log u_{ij}\in \she[0]{X}(1)(U_{ij})$ and

1444: $c_{ijk}\in \ZZ(1)$, satisfying the following relations: aside

1445: from the standard cocycle condition for the $\TT$-valued

1446: functions $u_{ij}$, we must have

1447: \begin{equation}

1448:   \label{eq:86}

1449:   \begin{gathered}

1450:     d\eta_i = 0\,,\\

1451:     d\sigma_i=\eta_i\,,\\

1452:     \sigma_j -\sigma_i = d \log u_{ij}\,.

1453:   \end{gathered}

1454: \end{equation}

1455: Given a holomorphic trivialization of $(L,\rho)$ as above, the

1456: position

1457: \begin{equation}

1458:   \label{eq:87}

1459:   \eta_i = \delb\del\log \rho_i\,, \quad

1460:   \sigma_i = \onehalf d^c \log \rho_i\,, \quad

1461:   u_{ij} = g_{ij}/\abs{g_{ij}}

1462: \end{equation}

1463: satisfies relations~\eqref{eq:86}. From here, the canonical

1464: connection is recovered. It is also easily seen that another

1465: solution of~\eqref{eq:86} is equivalent to the one above. The

1466: following is evident:

1467: \begin{lemma}

1468:   \label{lem:1}

1469:   Relations~\eqref{eq:87} provide an explicit quasi-isomorphism

1470:   between the \cech\ resolutions of $\dhh{1}$ and $C(1)^\bullet$.

1471: \end{lemma}

1472: 

1473: On the

1474: other hand, in~\eqref{eq:38} the relations defining the canonical

1475: connection are already enforced at the level of the complex.

1476: 

1477: Finally, if $[L,\rho]$ and $[L',\rho']$ are the classes

1478: corresponding to the metrized bundles $(L,\rho)$ and

1479: $(L',\rho')$, then we write

1480: \begin{math}

1481:   [ L\otimes L', \rho\rho'] = [L,\rho] + [L',\rho']\,.

1482: \end{math}

1483: 

1484: \subsection{Cup product of hermitian holomorphic line bundles}

1485: \label{sec:Cup-prod-herm}

1486: 

1487: If $L$ and $L'$ are two line bundles on $X$, their cup product in

1488: Deligne cohomology would produce a class in $\delH[4]{X}{\ZZ}{2}$

1489: denoted $\tame{L}{L'}$. Calculations were carried out

1490: in~\cite{brymcl:deg4_II} where a geometric interpretation of

1491: $\tame{L}{L'}$ as a 2-gerbe and its connection with the

1492: determinant of cohomology (when $X$ is the total space of a

1493: family of Riemann surfaces) were put forward. The structure

1494: ensuing from the generalization to line bundles equipped with a

1495: hermitian metric was further analyzed in ref.~\cite{bry:quillen}

1496: by means of hermitian holomorphic Deligne cohomology.

1497: 

1498: The cup product of two metrized line bundles $(L,\rho)$ and

1499: $(L',\rho')$ in hermitian holomorphic Deligne cohomology produces

1500: a class in $\dhhH[4]{X}{2}$. Despite the more involved definition

1501: of $\dhh{l}$ as opposed to that of $C(l)^\bullet$, it will

1502: nonetheless be more advantageous from the perspective of

1503: sect.~\ref{sec:conf-metr-herm} to use the former to calculate the

1504: desired class. Thus let us explicitly compute a representative of

1505: this class in terms of the expression~\eqref{eq:38} for the class

1506: of a metrized line bundle and the modified \bei\ 

1507: product~\eqref{eq:5} for

1508: \begin{equation*}

1509:   \dhh{1} \otimes \dhh{1} \xrightarrow{\cup_{\alpha}}

1510:   \dhh{2}  \,. 

1511: \end{equation*}

1512: A diagram for the complex $\dhh{2}$ analogous to that for

1513: $\dhh{1}$ displayed in the proof of Proposition~\ref{prop:1} is:

1514: \begin{equation}

1515:   \label{eq:39}

1516:   \begin{CD}

1517:     \ZZ(2)_X @>>> \sho{X} @>>> \shomega[2]{X}\oplus

1518:     \shomega[1]{X} @>>> \shomega[3]{X}\oplus \shomega[2]{X} @>>>

1519:     \shomega[4]{X}\oplus \shomega[3]{X} @>>>

1520:     \dotsm \\

1521:     & & @VV{\pi_1}V @VV{\imath\oplus\pi_1}V

1522:     @VV{\imath\oplus\pi_1}V @VV{\imath\oplus\pi_1}V \\

1523:     & & \she[0]{X}(1) @>>> F^2\!\sha[2]{X} \oplus \she[1]{X}(1)

1524:     @>>> F^2\!\sha[3]{X} \oplus \she[2]{X}(1) @>>>

1525:     F^2\!\sha[4]{X} \oplus \she[3]{X}(1) @>>>

1526:     \dotsm \\

1527:     & & & & & & & & @AA{\jmath\oplus 0}A \\

1528:     & & & & & & & & F^1\!\sha[4]{X}\cap \she[4]{X}(2) @>>> \dotsm

1529:   \end{CD}

1530: \end{equation}

1531: The cup product of two elements of the form~\eqref{eq:38} in the

1532: \cech\ resolution of $\dhh{1}$ would result in an analogous

1533: $5\times 3$ table. The degrees (in the cone) start from $0$ in

1534: the leftmost entry in the top and bottom rows, and from $1$ in

1535: the center one. The overall signs are determined by those in the

1536: \bei\ product~\eqref{eq:5} plus those arising from the \cech\ 

1537: resolution as explained in sect.~\ref{sec:notation}. Actually, it

1538: is visually less cumbersome to display the resulting expression

1539: as in~\eqref{tab:big-cup} below, where the corresponding

1540: bidegrees are explicitly indicated: the first degree is the

1541: overall degree in the cone and the second is the \cech\ degree.

1542: (We have explicitly written only the nonzero terms.)

1543: \begin{small}

1544:   \begin{equation}

1545:     \renewcommand{\onehalf}{\tfrac{1}{2}}

1546:     \renewcommand{\arraystretch}{1.3}

1547:     %%\setlength{\extrarowheight}{2pt}

1548:     \begin{array}{c|c}

1549:       (0,4) & c_{ijk}c'_{klm} \\ \hline

1550:       (1,3) & -c_{ijk} \log \, g'_{kl} \\ \hline

1551:       (2,2) & \bigl( -d \log \, g_{ij} \wedge d \log\, g'_{jk} \oplus

1552:       -\log \, g_{ij} \, d\log\,g'_{jk} \bigr) \bigoplus

1553:       -\pi_1\log\,g_{ij}\, \log\,\abs{g'_{jk}} \\ \hline

1554:       (3,1) & 

1555:       \begin{aligned}

1556:         \bigl( -(1-\alpha) \, d\log g_{ij} \wedge \del \log

1557:         \rho'_j +\alpha \, \del \log \rho_i \wedge d \log g'_{ij}

1558:         \bigr)  

1559:         \oplus 

1560:         \Bigl\{ (1-\alpha) \bigl( d^c\log\,\abs{g_{ij}}\;

1561:         \onehalf \log\rho'_j &-\log\,\abs{g_{ij}}\; \onehalf d^c

1562:         \log\,\rho'_j

1563:         \bigr) \\

1564:          +\alpha \bigl( -\onehalf d^c

1565:         \log\rho_i\;\log\,\abs{g'_{ij}} &+\onehalf \log\rho_i\;

1566:         d^c\log\,\abs{g'_{ij}} \bigr) \Bigr\}

1567:       \end{aligned} \\ \hline

1568:       (4,0) & 

1569:       \begin{aligned}

1570:         \delb\del \log\rho_i \wedge

1571:         \delb\del \log \rho'_i

1572:         \bigoplus

1573:         \bigl( & \alpha\, \delb\del\log\rho_i \wedge \del \log

1574:         \rho'_i +(1-\alpha)\, \del\log \rho_i \wedge

1575:         \delb\del\log\rho'_i

1576:         \bigr) \\

1577:         & \oplus \bigl( \alpha\, \delb\del \log\rho_i\;

1578:         \onehalf\log\rho'_i +(1-\alpha)\, \onehalf \log\rho_i \;

1579:         \delb\del\log\rho'_i \bigr)

1580:       \end{aligned}      

1581:     \end{array}

1582:     \label{tab:big-cup}

1583:   \end{equation}

1584: \end{small}

1585: 

1586: 

1587: \subsection{Integration along the fiber}

1588: \label{sec:integr-along-fiber}

1589: 

1590: To conclude this introduction, let us quickly mention that

1591: Hermitian holomorphic Deligne cohomology has an integration along

1592: the fiber map. Namely, if $\pi\colon X\rightarrow S$ is a proper

1593: submersion of complex manifolds, it follows from

1594: \cite{MR91j:14017} that Deligne cohomology has an integration

1595: along the fiber $\int_\pi$, hence by

1596: \cite{brymcl:deg4_II,bry:quillen} there is a map

1597: \begin{equation}

1598:   \label{eq:28}

1599:   \dhhH[i]{X}{l} \longrightarrow \dhhH[i-2d]{S}{l-d}

1600: \end{equation}

1601: where $d$ is the complex dimension of the fiber, and commutative

1602: diagrams analogous to \cite[Theorem 5.1]{bry:quillen}.

1603: 

1604: We will be interested in the case of complex relative dimension

1605: $1$, namely $\pi \colon X\rightarrow S$ is a holomorphic

1606: fibration with compact connected Riemann surfaces as fibers. From

1607: \eqref{eq:28} in degree $4$ we have the map:

1608: \begin{math}

1609:   \dhhH[4]{X}{2} \rightarrow \dhhH[2]{S}{1}

1610: \end{math}

1611: into the group of complex hermitian line bundles on $S$ (cf.\ 

1612: Proposition \ref{prop:1}).

1613: 

1614: We refer to refs.\ \cite{brymcl:deg4_II,bry:quillen} for

1615: a complete treatment of the map~\eqref{eq:28}, in particular for

1616: the ``trace'' map

1617: \begin{equation*}

1618:   \RRR^\bullet\pi_* (\dhh{l}) \longrightarrow

1619:   \dhh[\bullet-2d]{l-d}\,,

1620: \end{equation*}

1621: which induces~\eqref{eq:28} at the level of cohomology. (Clearly

1622: any quasi-isomorphic model for Hermitian-holomorphic cohomology

1623: will do.) We will limit ourselves to observe the following: if

1624: $\cover{U}_S$ is a good cover of $S$, and $\cover{U}_X$ a good

1625: cover of $X$ which refines $\pi^{-1}\cover{U}_S$, the $i$-{th}

1626: direct image $\RRR^i\pi_* \sheaf{F}^\bullet$ of any complex

1627: $\sheaf{F}^\bullet$ on $X$ is the sheaf on $S$ associated to the

1628: presheaf

1629: \begin{equation*}

1630:   V \longmapsto H^i \bigl(\Tot

1631:   \Check{C}^\bullet ( \cover{U}_X\cap \pi^{-1}(V),\sheaf{F}^\bullet)

1632:   \bigr)\,. 

1633: \end{equation*}

1634: Then unraveling the cone structure of ${\dhh{l}}_X$ (or of

1635: $C(l)^\bullet_X$) reduces to computing the direct images of

1636: $\ZZ(l)_X$ and $\sha{X}$, the de~Rham complex. The trace map is

1637: then obtained by capping a total cocycle, say in

1638: \begin{equation}

1639:   \label{eq:89}

1640:   \bigoplus_{p+q=i}

1641:   \Check{C}^q ( \cover{U}_X\cap \pi^{-1}(V),\sha[p]{X})\,,

1642: \end{equation}

1643: with a representative of the fundamental class $[M]$ of the

1644: smooth model $M$ of the fiber of $\pi^{-1}(V)\cong V\times M$ (as

1645: smooth manifolds). Explicit representatives for $[M]$ for a

1646: triangulation subordinated to the nerve of $\cover{U}_X\cap

1647: \pi^{-1}(V)$ are computed in ref.\ \cite{aldtak2000}.  If

1648: $\omega^p_{i_0,\dots,i_q}$ are the components of a cocycle

1649: $\omega^i_V$ in~\eqref{eq:89}, then

1650: \begin{equation*}

1651:   \omega^i_V \cap [M]

1652:   =\sum \omega^p_{i_0,\dots,i_q} \cap \Delta^{2d-q}_{i_0,\dots,i_q}

1653: \end{equation*}

1654: where the $\Delta^{2d-q}_{i_0,\dots,i_q}$ are \emph{signed}

1655: generators of the nerve of dimension $2d -q$. The exact

1656: combinatorics can be found in ref.\ \cite{aldtak2000}, so we will

1657: not further pursue the matter here.

1658: 

1659: Let us conclude by noticing that for $\pi\colon X\to S$ the cup

1660: product of $(L,\rho)$ and $(L',\rho')$ induces a metrized line

1661: bundle on $S$. {F}rom the preceding discussion, the connection

1662: component, for example, will be obtained by following the (two)

1663: components $F^2\!\sha{X}\cap \she[\bullet \geq 4]{X}(2)$

1664: in~\eqref{tab:big-cup}. The one in degree $(4,0)$ will be capped

1665: with generators $\Delta^{2}$ of dimension $2$ (hence integrated

1666: over $2$-simplices), those in degree $(3,1)$ with generators

1667: $\Delta^{1}$ of dimension $1$.

1668: 

1669: 

1670: From \cite[Theorem 6.1]{brymcl:deg4_II} there is a

1671: corresponding map

1672: \begin{math}

1673:   A^{(2,2)}(X)_{\RR(2)} \rightarrow A^{(1,1)}(S)_{\RR(1)}\,.

1674: \end{math}

1675: {F}rom the explicit cocycles we have computed the corresponding

1676: representative in $H^2(S,\RR(1))\cap H^{1,1}(S)$ of the first

1677: Chern class of the resulting metrized line bundle is

1678: \begin{equation*}

1679:   \int_\pi c_1(\rho) \wedge c_1(\rho')\,,

1680: \end{equation*}

1681: as in \cite[Proposition 6.6.1]{MR89b:32038}.

1682: 

1683: 

1684: \section{Conformal metrics and the Liouville functional}

1685: \label{sec:Conf-metr-Liouv}

1686: 

1687: Let $X$ be a compact Riemann surface of genus $g\geq 2.$ For simplicity

1688: we can assume $X$ to be connected, although this is not necessary, and

1689: in fact this assumption will be dropped when dealing with Kleinian

1690: groups.

1691: 

1692: Let $\cm{X}$ be the space of conformal metrics on $X$. Locally on $X$,

1693: if $z$ is a local analytic coordinate defined on an open set $U$, any

1694: metric $ds^2$ can be represented as

1695: \begin{equation*}

1696:   ds^2 = \rho \, \abs{dz}^2

1697: \end{equation*}

1698: for a positive function $\rho : U \to \RR_{>0}$. According to

1699: sec.~\ref{sec:line-bun}, a conformal metric corresponds to

1700: considering the metrized line bundle $(T_X,\rho)$, where $T_X$ is

1701: the holomorphic tangent bundle of $X$. With respect to a cover

1702: $\cover{U}_X$ of $X$, the conformal factors $\rho_i$ and

1703: $\rho_j$, associated to $U_i$ and $U_j$ respectively, satisfy the

1704: relation

1705: \begin{equation}

1706:   \label{eq:41}

1707:   \rho_j = \rho_i\, \abs{z'_{ij}}^2\,,

1708: \end{equation}

1709: where $z'_{ij} = dz_i/dz_j$ and $z_i$ is a local coordinate

1710: defined over $U_i$.

1711: 

1712: It follows from the uniformization theorem that there exists a

1713: unique conformal metric of scalar curvature equal to $-1$, the

1714: Poincar\'e metric. Locally on $U_i\subset X$, the condition for

1715: the metric to have curvature $-1$ is equivalent to the nonlinear

1716: PDE

1717: \begin{equation}

1718:   \label{eq:42}

1719:   \frac{\del^2}{\del z_i\del\Bar z_i} \phi_i = \onehalf

1720:   \exp {\phi_i} \,,

1721: \end{equation}

1722: known as the Liouville equation, for the smooth function $\phi_i

1723: = \log \rho_i$. Observe that equation~\eqref{eq:42} can be

1724: written in the form

1725: \begin{equation*}

1726:   c_1(\rho) = \sqrt{-1}\,\omega_\rho\,,

1727: \end{equation*}

1728: where we have used the K\"ahler form associated to the metric:

1729: \begin{equation*}

1730:   \omega_\rho\vert_{U_i} = \ihalf \rho_i \,

1731:   dz_i\wedge d\Bar z_i\,.  

1732: \end{equation*}

1733: This representation makes it apparent that the Liouville equation

1734: is independent of the choice of the coordinate system. On the

1735: other hand, a direct verification of this fact is immediate using

1736: $\phi_j -\phi_i = \log \abs{z'_{ij}}^2$.

1737: 

1738: 

1739: \subsection{Variational problem for conformal metrics}

1740: \label{sec:vari-probl-conf}

1741: 

1742: It is well known that the Liouville equation has a local

1743: variational principle in the following sense. Let $D$ be a region

1744: in the complex plane. Then equation~\eqref{eq:42} is the

1745: Euler-Lagrange equation for the variational problem defined by

1746: the action functional

1747: \begin{equation}

1748:   \label{eq:43}

1749:   S[\phi] = \ihalf \int_D \bigl( \del \phi \wedge \delb \phi

1750:   + e^\phi\,dz\wedge d\Bar z \,\bigr) \,,

1751: \end{equation}

1752: defined on smooth functions $\phi :D \to \RR$, with the condition

1753: that variations

1754: \begin{math}

1755:   \left.\frac{d}{d\alpha} (\phi_\alpha) \right\vert_{\alpha=0}

1756: \end{math}

1757: of $\phi$ be zero on $\del D$. However, it is easily seen that

1758: the functional~\eqref{eq:43} cannot be defined globally on $X$,

1759: since, as a consequence of~\eqref{eq:41}, the first term under

1760: the integral sign would not yield a well-defined $2$-form on $X$.

1761: (The second one would of course present no problems, it would

1762: just give the area $A_X(\rho)$ of $X$ with respect to the given

1763: metric $\rho\,\abs{dz}^2 = e^\phi\,\abs{dz}^2$.)

1764: Accordingly, it is convenient to write the integrand in~\eqref{eq:43} as 

1765: \begin{equation*}

1766:   \sqrt{-1}\,\omega[\phi] + \omega_\rho

1767: \end{equation*}

1768: where we have defined the $2$-form\footnote{Note that

1769:   equation~\eqref{eq:44} defines an imaginary form. The reason for this

1770:   choice will be apparent later.}

1771: \begin{equation}

1772:   \label{eq:44}

1773:   \omega^{0}[\phi] = \onehalf \del \phi \wedge \delb \phi\,,

1774: \end{equation}

1775: and restrict our considerations

1776: to the first term of~\eqref{eq:43} which we denote

1777: \begin{equation}

1778:   \label{eq:45}

1779:   \Check{S}[\phi] = \sqrt{-1}\,\int_D \omega^{0}[\phi]\,.

1780: \end{equation}

1781: 

1782: There is by now an established procedure on how to address the problem

1783: caused by the fact that~\eqref{eq:44} is not globally defined.  In

1784: general terms, given the choice of a conformal metric $ds^2\in\cm{X}$

1785: and a cover $\cover{U}_X$, one works with the full \cech-de~Rham complex

1786: $\Check{C}^\bullet(\cover{U}_X,\she{X}(1))$ with respect to

1787: $\cover{U}_X$, rather than with just differential forms. The

1788: $2$-form~\eqref{eq:44} is then completed to a total degree

1789: $2$-cocycle---to be denoted $\Omega[\phi]$. This results in a class in

1790: $H^2(X,\RR(1))$ after taking cohomology. (Whether or not there also is a

1791: de~Rham type theorem will depend on the acyclicity properties of

1792: $\cover{U}_X$.)

1793: 

1794: This scheme has been previously carried out not quite for covers

1795: of $X$ by open sets, but actually for different choices of planar

1796: coverings. For the covering associated to a Schottky

1797: uniformization of $X$ a generalization of eq.~\eqref{eq:43} was

1798: written in ref.\ \cite{zogtak1987-2}. More recently, a detailed

1799: calculation of the cocycle for the general case of a covering

1800: associated to a Kleinian uniformization was carried out in ref.\

1801: \cite{math.CV/0204318} by exploiting the homological methods

1802: developed in ref.\ \cite{aldtak1997}. Note, however, that from the

1803: point of view of ref.\ \cite{aldtak2000} these planar coverings

1804: are ``\'etale'' coordinates on $X$, so the group cohomology

1805: constructions required to work with the various kinds of

1806: uniformization coverings just follow from specializing the \cech\ 

1807: formalism to the coverings at hand.

1808: 

1809: Finally, the integration in eqs.~\eqref{eq:43} or~\eqref{eq:45} should

1810: be replaced by the evaluation of $\Omega[\phi]$ against an appropriate

1811: representative $\Sigma$ of the fundamental class of $X$. The

1812: ``appropriate'' form for both $\Sigma$ and the evaluation will be

1813: dictated by the chosen cover $\cover{U}_X$ \emph{and} the cohomology

1814: theory being used. Typically $\Sigma$ will be a cycle in a double

1815: complex of $\cover{U}_X$-small simplices, where the differentials are

1816: the singular one and the one determined by the face maps induced by the

1817: cover.  Thus, in the case of a \cech\ cover, it will be the complex

1818: determined by a triangulation of $X$ subordinated to the open

1819: $\cover{U}_X$, and in the same way, for a planar cover the singular

1820: complex of the planar domain $\cover{U}_X$ tensored with an appropriate

1821: bar resolution of the group of deck transformations. These issues have

1822: been discussed at length in refs.\ 

1823: \cite{aldtak1997,aldtak2000,math.CV/0204318}, so we will not repeat the

1824: discussion here. Whenever we have a cocycle extending~\eqref{eq:44} and

1825: a cycle $\Sigma$ representing $X$ we state

1826: \begin{definition}

1827:   \label{def:2}

1828:   The Liouville functional (without the area term) is given by

1829:   the evaluation:

1830:   \begin{equation}

1831:     \label{eq:46}

1832:     \Check{S}[\phi] = -\frac{1}{\tate}

1833:     \dual{\Omega[\phi]}{\Sigma}\,. 

1834:   \end{equation}

1835:   For the complete functional we add the area term

1836:   \begin{equation}

1837:     \label{eq:47}

1838:     S[\phi] = \Check{S}[\phi] +\frac{1}{2\pi}

1839:     \int_X\omega_\rho\,. 

1840:   \end{equation}

1841: \end{definition}

1842: \begin{remark}

1843:   $\dual{\Omega[\phi]}{\Sigma}\in \RR(1)$, and $\Check{S}[\phi]$ (or

1844:   $S[\phi]$) is real. Division by $\tate$ is conventional, but note that

1845:   $\RR(1) \qi \RR_+$ via $\exp(\cdot / \tate )$. In the sequel it will

1846:   be more convenient to work with imaginary classes. (See also

1847:   section~\ref{sec:conf-metr-herm}.)

1848: \end{remark}

1849: In next two subsections we examine these constructions in some

1850: detail. For definiteness, we initially make use of an ordinary

1851: open cover. First, we recall the direct construction of a cocycle

1852: generalizing~\eqref{eq:44}, and we show that this way

1853: eq.~\eqref{eq:42} indeed is the resulting extremum condition.

1854: Then we emphasize the role played by Deligne cohomology and the

1855: tame symbol. These aspects will become important when introducing

1856: a Kleinian uniformization later in the paper, when we discuss

1857: connections with the dilogarithm function.  In a later section we

1858: shall tackle the question of its geometrical significance by

1859: making full use of the hermitian-holomorphic version of Deligne

1860: cohomology presented in section~\ref{sec:herm-holom-deligne}, and

1861: we show that the cocycle constructed following refs.

1862: \cite{zogtak1987-2,math.CV/0204318} corresponds to the square of

1863: $(T_X,\rho)$ in hermitian holomorphic Deligne cohomology.

1864: 

1865: 

1866: \subsection{Direct construction of the Liouville cocycle}

1867: \label{sec:Direct-constr-Liouv}

1868: 

1869: \subsubsection{Initial setup}

1870: \label{sec:Initial-setup}

1871: 

1872: Let $X$ a compact Riemann surface of genus greater than $2$. We

1873: shall not include the area term in our explicit calculations,

1874: therefore it makes sense to extend our considerations to a

1875: general metrized line bundle $(L,\rho)$. Of course, whenever

1876: referring to a conformal metric or to the variational problem for

1877: the Liouville equation, it will be be assumed that $L=T_X$ and

1878: that $g_{ij}= z'_{ij}$. Upon choosing a cover $\cover{U}_X$,

1879: which for now we assume to be a \cech\ cover by open sets, the

1880: pair $(L,\rho)$ is described in terms of the data expounded in

1881: section~\ref{sec:line-bun}. Our starting point will be the

1882: $0$-cochain

1883: \begin{equation}

1884:   \label{eq:48}

1885:   \omega^{0}_i[\log\rho_i]

1886:   =\onehalf\,\del \log \rho_i \wedge \delb \log \rho_i

1887:   = -\onehalf d \log \rho_i \wedge \onehalf d^c \log \rho_i

1888: \end{equation}

1889: with values in $\she[2]{X}(1)(U_i)$.

1890: \begin{remark}

1891:   \label{rem:1}

1892:   A generalization for eq.~\eqref{eq:48} would be to consider a

1893:   \emph{pair} of metrized line bundles $(L,\rho)$ and $(L',\rho')$, and

1894:   then the analog of~\eqref{eq:48} would be

1895:   \begin{equation}

1896:     \label{eq:49}

1897:     \omega^{0}_i[\log\rho_i,\log\rho'_i] = \onehalf \Bigl(

1898:     -\onehalf d \log \rho_i \wedge \onehalf d^c \log \rho'_i

1899:     +\onehalf d^c \log \rho_i \wedge \onehalf d \log \rho'_i

1900:     \Bigr)

1901:   \end{equation}

1902:   Note, however, that the expressions are quadratic. Moreover,

1903:   $L\otimes L'$ has metric $\rho\rho'$, so that there is a

1904:   natural ``polarization identity''

1905:   \begin{equation}

1906:     \label{eq:50}

1907:     \omega[\log\rho,\log\rho'] =

1908:     \onefourth \omega[\log \rho\rho']

1909:     - \onefourth \omega[\log \frac{\rho}{\rho'}]

1910:   \end{equation}

1911:   where we have omitted the indexes for simplicity of notation.  We

1912:   shall comment later on the significance of eq.~\eqref{eq:50}.

1913: \end{remark}

1914: 

1915: 

1916: \subsubsection{Computation}

1917: \label{sec:computation}

1918: 

1919: Let us extend~\eqref{eq:48} to a degree $2$ cocycle in the total

1920: simple complex associated to the double complex

1921: \begin{math}

1922:   \Check{C}^\bullet(\cover{U}_X,\she{X}(1))

1923: \end{math}

1924: of \cech\ cochains with values in the de~Rham complex of imaginary

1925: smooth forms. This is accomplished in the usual fashion (see

1926: e.g.~\cite{bott1982}) by finding a $1$-cochain of $1$-forms

1927: $\omega^{1}_{ij}[\log \rho]$ on $U_{ij}$ and a $2$-cochain of $0$-forms

1928: $\omega^{2}_{ijk}[\log \rho]$ on $U_{ijk}$ such that the relations

1929: \begin{equation*}

1930:   \deltacheck \omega^{0} = -d \omega^{1}\,, \qquad

1931:   \deltacheck \omega^{1} = d \omega^{2}\,, \qquad

1932:   \deltacheck \omega^{2} = 0

1933: \end{equation*}

1934: are satisfied. Of course, the remaining one, namely $d\omega =0$

1935: is automatically satisfied for dimensional reasons. It turns out

1936: that to a great extent these relations are explicitly computable

1937: without further assumptions, such as that the cover $\cover{U}_X$

1938: be good.  The needed calculations are fairly standard, and they

1939: are presented in great detail in ref.~\cite{math.CV/0204318}, so

1940: we shall be brief. (The observation in \cite{lang:arakelov} that

1941: on a Riemann surface for two smooth functions $f$ and $g$ one has

1942: \begin{math}

1943:   df \wedge d^c g = dg\wedge d^c f

1944: \end{math}

1945: is useful in carrying out the calculations.) The first two steps

1946: are as follows.

1947: 

1948: First, one has:

1949: \begin{equation}

1950:   \label{eq:51}

1951:   \begin{gathered}

1952:     \omega^{0}_j[\rho] - \omega^{0}_i[\rho]

1953:     = -d\,\omega^{1}_{ij} [\rho] \\

1954:     \omega^{1}_{ij} [\rho]

1955:     = \onehalf\log\rho_i\;d^c \log\,\abs{g_{ij}}

1956:     +d^c \log\,\abs{g_{ij}}\; \onehalf \log \rho_j

1957:   \end{gathered}

1958: \end{equation}

1959: The next step yields:

1960: \begin{equation*}

1961:   \begin{aligned}

1962:     \deltacheck\bigl( \omega^{1}[\log\rho] \bigr)_{ijk} &=

1963:     \omega^{1}_{jk} [\rho] -\omega^{1}_{ik} [\rho]

1964:     +\omega^{1}_{ij} [\rho] \\

1965:     &= \log\,\abs{g_{ij}} \; d^c \log\,\abs{g_{jk}}

1966:     -d^c \log\,\abs{g_{ij}} \; \log\,\abs{g_{jk}}

1967:   \end{aligned}

1968: \end{equation*}

1969: and notice that

1970: \begin{math}

1971:   d^c \log\,\abs{g_{ij}} = \pi_1 d\,\log \,g_{ij}\,,

1972: \end{math}

1973: and

1974: \begin{math}

1975:   \pi_{p+q-1}(a\wedge b) = \pi_p(a)\wedge \pi_{q-1}(b) +

1976:   \pi_{p-1}(a)\wedge \pi_q(b)\,,

1977: \end{math}

1978: so we have

1979: \begin{equation}

1980:   \label{eq:52}

1981:   \deltacheck\bigl( \omega^{1}[\log\rho] \bigr)_{ijk} =

1982:   \pi_1\bigl( \log\,g_{ij} \; d\,\log\, g_{jk} \bigr)

1983:   -d \bigl( \pi_1(\log\,g_{ij})\; \log\,\abs{g_{jk}} \bigr) \,.

1984: \end{equation}

1985: Observe that now the problem of continuing the descent becomes

1986: independent of the chosen metric $\rho$. The most direct way of

1987: proceeding is the following. If we assume the cover $\cover{U}_X$ to be

1988: acyclic for the de~Rham complex $\she{X}(1)$, then there exists

1989: $\omega^{2}_{ijk} \in \she[0]{X}(1)(U_{ijk})$ such that

1990: $d\omega^{2}_{ijk} = \deltacheck (\omega^{1}[\log\rho])_{ijk}$.

1991: Furthermore, consistency on a quadruple intersection requires that

1992: $\deltacheck\omega^{2}$ be a $3$-cocycle with values in $\RR(1)_X$.

1993: Since $H^3(X,\RR(1))=0$, this cocycle must be a coboundary, therefore,

1994: up to readjusting the constants, there exists a choice of

1995: $\omega^{2}_{ijk}$ such that $\deltacheck\omega^{2} = 0$, and

1996: furthermore, $\omega^{2}_{ijk}$ does not depend on the metric structure.

1997: 

1998: 

1999: \subsubsection{Solution to the variational problem}

2000: \label{sec:solut-vari-probl}

2001: 

2002: The previous preliminary calculation is sufficient from the point of

2003: view of finding the extrema. To this effect, we set $L = T_X$, for

2004: $\rho$ a conformal metric in $\cm{X}$. Notice that the space of

2005: conformal metrics on $X$ is affine over $C^\infty(X,\RR) \equiv

2006: \mathcal{E}^0(X)$: if $ds^2$ and ${ds'}^2$ are two conformal metrics

2007: with local expressions $\rho_i\,\abs{dz_i}^2$ and

2008: $\rho'_i\,\abs{dz_i}^2$ respectively, then there exists $\sigma \in

2009: C^\infty(X,\RR)$ such that $\log\rho'_i =\sigma\vert_{U_i} +

2010: \log\rho_i$. The change from $\Omega[\log\rho]$ to $\Omega[\log\rho

2011: +\sigma]$ can be exactly computed thanks to the fact that the last step

2012: in the determination of $\Omega[\log\rho]$ is independent of $\rho$ and

2013: the quadratic character of~\eqref{eq:44}. Indeed we have:

2014: \begin{proposition}

2015:   \label{prop:2}

2016:   \begin{equation*}

2017:     S[\log\rho +\sigma] -S[\log\rho] =

2018:     \frac{1}{\tate} \int_X \bigl(

2019:     \onehalf \del\sigma \wedge \delb\sigma

2020:     +\sigma\, c_1(\rho) -\sqrt{-1}(e^\sigma -1)

2021:     \omega_\rho\bigr)

2022:   \end{equation*}

2023: \end{proposition}

2024: \begin{proof}

2025:   The change in $\omega^{0}_i[\log \rho]$ is computed as

2026:   \begin{equation*}

2027:     \omega^{0}_i[\log \rho + \sigma] -\omega^{0}_i[\log \rho] =

2028:     -\onehalf d\sigma_i \wedge \onehalf d^c\sigma_i

2029:     + \sigma_i\, \onehalf dd^c \log \rho_i

2030:     - \onehalf d(\sigma_i \, d^c \log\rho_i)\,,

2031:   \end{equation*}

2032:   where we set $\sigma_i \equiv \sigma\rvert_{U_i}$. Note that

2033:   the first two terms on the right hand side are globally

2034:   well-defined $2$-forms. On the other hand

2035:   \begin{equation*}

2036:     \omega^{1}_{ij}[\log \rho +\sigma]

2037:     -\omega^{1}_{ij}[\log\rho] =

2038:      \sigma_i\,d^c \log\,\abs{g_{ij}}\,.

2039:   \end{equation*}

2040:   Letting $\chi_i = \sigma_i \, d^c \log\rho_i$, we see that

2041:   \begin{equation*}

2042:     \Omega[\log \rho +\sigma] - \Omega[\log \rho] =

2043:     -\onehalf d\sigma \wedge \onehalf d^c\sigma

2044:     + \sigma\, \onehalf dd^c \log \rho -D\chi\,,

2045:   \end{equation*}

2046:   and taking the area terms into account, establishes the

2047:   formula.

2048: \end{proof}

2049: As a consequence, we obtain

2050: \begin{corollary}

2051:   \label{cor:1}

2052:   The Liouville equation~\eqref{eq:42} is the Euler-Lagrange

2053:   equation for the Liouville functional~\eqref{eq:47} introduced

2054:   in definition~\ref{def:2}. The critical point is

2055:   non-degenerate.

2056: \end{corollary}

2057: \begin{proof}

2058:   Replacing $\sigma$ with $t\,\sigma$, $t\in\RR$, in the previous

2059:   proposition we find the infinitesimal change in $\Omega[\log

2060:   \rho]$ to be

2061:   \begin{equation*}

2062:     \left.\frac{d}{dt}\right\rvert_{t=0}

2063:     S[\log \rho + t\sigma] = -\frac{1}{\tate}\, \int_X \sigma

2064:     \bigl( c_1(\rho) - \sqrt{-1}\,\omega_\rho \bigr)\,,

2065:   \end{equation*}

2066:   and it follows that $S[\log\rho]$ has an extremum if and only

2067:   if the Liouville equation is satisfied. Non degeneracy follows

2068:   from the quadratic part in the exact change formula in

2069:   proposition~\ref{prop:2}.

2070: \end{proof}

2071: \begin{remark}

2072:   The fact that the change in the cocycle is given by a pure $2$-form

2073:   term up to total coboundary can also be analyzed in terms of gluing

2074:   properties of variational bicomplexes, cf.  Theorem 1.2 in ref.\

2075:   \cite{MR1908413} and the proof of Theorem 1 in ref.\ \cite{aldtak2000}.

2076:   From this perspective, Corollary~\ref{cor:1} is a direct consequence

2077:   of the affine structure of the space $\cm{X}$ of conformal metrics.

2078: \end{remark}

2079: 

2080: 

2081: \subsection{A cup product}

2082: \label{sec:cup-product}

2083: Formula~\eqref{eq:52} can be handled in a more geometric fashion

2084: as follows. From sec.~\ref{sec:remarks-dilogarithm} we can

2085: rewrite~\eqref{eq:52} as

2086: \begin{equation}

2087:   \label{eq:53}

2088:   \deltacheck\bigl( \omega^{1}[\log\rho] \bigr)_{ijk} =

2089:    -r_2(g_{ij}, g_{jk})\,,

2090: \end{equation}

2091: and we have the collection of tame symbols

2092: $\tame{g_{ij}}{g_{jk}}$ associated any triple intersection

2093: $U_{ijk}$ in the cover $\cover{U}_X$, \cite{brymcl:deg4_II}.

2094: These symbols glue to form a global symbol $\tame{L}{L}$. As a

2095: cohomology class on a curve $X$, however, $\tame{L}{L}$ will be

2096: zero (that is, there is a global object in the associated

2097: $2$-stack, cf. \cite{bry:quillen}), so that it will be possible

2098: to choose local functions $L_{ijk}$ such that

2099: \begin{equation*}

2100:   d L_{ijk} = -\log\,g_{ij} \; d\,\log\, g_{jk}\,,

2101: \end{equation*}

2102: as explained in detail in \cite{aldtak2000}. Note that we still need the

2103: cover $\cover{U}_X$ to be fine enough. Moreover, cf. loc.~cit., the

2104: collection $L_{ijk}$ can be chosen in a way that

2105: \begin{equation*}

2106:   \deltacheck L _{ijkl} = -c_{ijk}  \log \, g_{kl} + n_{ijkl}\,,

2107: \end{equation*}

2108: where $n_{ijkl} \in \ZZ(2)$. Therefore

2109: from~\eqref{eq:21},~\eqref{eq:52} and~\eqref{eq:53}, we can set

2110: \begin{equation}

2111:   \label{eq:54}

2112:   \omega^{2}_{ijk} = 

2113:   -\bigl( \pi_1(\log\,g_{ij})\; \log\,\abs{g_{jk}} \bigr)

2114:   -\pi_1 L_{ijk}\,,

2115: \end{equation}

2116: namely according to sec.~\ref{sec:remarks-dilogarithm} we have set

2117: $\omega^{2}_{ijk} = -\bwli (g_{ij},g_{jk})$. Now, the last compatibility

2118: condition is satisfied, indeed we have:

2119: \begin{align*}

2120:   \deltacheck \omega^{2}_{ijkl} &= 

2121:   -c_{ijk}\, \log \,\abs{g_{kl}} -\pi_1 (\deltacheck L_{ijkl}) \\

2122:   &= -c_{ijk}\, \log \,\abs{g_{kl}}

2123:   +\pi_1 (c_{ijk} \log g_{kl} ) \\

2124:   &= 0\,.

2125: \end{align*}

2126: As a result, we obtain a $2$-cocycle in the the total complex associated

2127: to $\Check{C}^\bullet (\cover{U}_X,\she{X}(1))$ as before, with a more

2128: geometric interpretation of $\omega^{2}_{ijk}$ in terms of the

2129: trivialization of the symbol $\tame{L}{L}$. Moreover, notice that if $X$

2130: is a curve then $\she{X}(1)[1] \qi \deltilde{\RR}{2}$ thus we may

2131: interpret the class so determined by $\Omega[\log \rho]=

2132: \omega^{0}[\log\rho] +\omega^{1}[\log\rho] +\omega^{2}$ as a (degree

2133: $3$) class in $\delH[3]{X}{\RR}{2}$.

2134: 

2135: 

2136: \subsection{Two line bundles}

2137: \label{sec:two-line-bundles}

2138: 

2139: For a pair $(L,\rho)$, $(L',\rho')$ of metrized line bundles we can

2140: complete~\eqref{eq:49} to a cocycle $\Omega[\log\rho,\log\rho']$ via an

2141: analogous procedure to the one presented in

2142: sects.~\ref{sec:Direct-constr-Liouv} and~\ref{sec:cup-product}.  The

2143: relevant calculations being entirely similar, we limit ourselves to

2144: quoting the relevant expressions. Starting from~\eqref{eq:49}, which we

2145: rewrite in the form

2146: \begin{equation}

2147:   \label{eq:55}

2148:   \omega^{0}_i[\log\rho_i,\log\rho'_i] =

2149:   \onehalf d^c \log \rho_i \wedge \onehalf d \log \rho'_i\,,

2150: \end{equation}

2151: the corresponding expression for the degree $(1,1)$ term is:

2152: \begin{equation}

2153:   \label{eq:56}

2154:   \omega^{1}_{ij}[\log\rho,\log\rho'] =

2155:   \onehalf \log\rho_i\;d^c \log\,\abs{g'_{ij}} 

2156:   +\onehalf d^c \log\,\abs{g_{ij}}\; \log \rho'_j\,.

2157: \end{equation}

2158: Computing the \cech\ coboundary we find:

2159: \begin{equation}

2160:   \label{eq:57}

2161:   \begin{aligned}

2162:     \bigl(\deltacheck \omega^{1}[\log\rho,\log\rho']\bigr)_{ijk} &=

2163:     \log\,\abs{g_{ij}} \; d^c \log\,\abs{g'_{jk}}

2164:     -d^c \log\,\abs{g_{ij}} \; \log\,\abs{g'_{jk}} \\

2165:     &= -r_2(g_{ij},g'_{jk})\,,

2166:   \end{aligned}

2167: \end{equation}

2168: from which $\omega^{2}_{ijk}$ (now independent of $\rho$ and $\rho'$)

2169: can be obtained as $-\bwli (g_{ij},g'_{jk})$ by looking at a collection

2170: $L_{ijk}$ such that

2171: \begin{equation*}

2172:   d L_{ijk} = -\log\,g_{ij} \; d\,\log\, g'_{jk}\,,

2173: \end{equation*}

2174: from the triviality of the symbol $\tame{L}{L'}$.

2175: 

2176: \subsection{Additional remarks on the Liouville functional and the

2177:   Bloch-Wigner dilogarithm}

2178: \label{sec:liouv-funct-bloch}

2179: 

2180: We wish to compare the previous constructions to those of ref.\ 

2181: \cite{math.CV/0204318}. To this end we need to specifically

2182: consider the case of a cover of $X$ provided by a Kleinian

2183: uniformization. (The reader should consult loc.\ cit.\ for

2184: reference and complete details.) The comparison offers a better

2185: perspective on the absence of cohomological obstructions in the

2186: calculations of ref.\ \cite{math.CV/0204318}, and on the

2187: relations with three-dimensional hyperbolic geometry.

2188: 

2189: Let $\Gamma$ be a purely loxodromic second kind Kleinian group

2190: satisfying all the conventions spelled out in

2191: Appendix~\ref{sec:kleinian-groups}, to which we refer for the

2192: notation.  Let $U_\Gamma\subset \pione$ be the region of

2193: discontinuity, and $X=U_\Gamma/\Gamma$ the resulting (possibly

2194: disconnected) Riemann surface.

2195: 

2196: A conformal metric $\rho$ on $X=X_1\sqcup \dots \sqcup X_n$

2197: appears as an automorphic function on $U_\Gamma$:

2198: \begin{equation}

2199:   \label{eq:62}

2200:   \onehalf \log \rho - \onehalf \log \rho\circ \gamma

2201:   = \log\, \abs{\gamma'}\,,\quad \gamma\in \Gamma\,.

2202: \end{equation}

2203: Eq.~\eqref{eq:62} is the direct translation of \eqref{eq:29}

2204: following the principles of \cite{aldtak2000}.  Accordingly, the

2205: first two terms of the Liouville cocycle computed by applying the

2206: procedure explained in sect.~\ref{sec:Conf-metr-Liouv} are:

2207: \begin{gather}

2208:   \omega^{0}[\log\rho]

2209:   = -\onehalf d \log \rho \wedge \onehalf d^c \log \rho \label{eq:63} \\

2210:   \omega^{1}_{\gamma} [\log\rho]

2211:   = \bigl(\onehalf\log\rho +\onehalf \log \rho

2212:   \circ \gamma \bigr)\; d^c \log\,\abs{\gamma'}\label{eq:64}

2213: \end{gather}

2214: and computing the coboundary of~\eqref{eq:64} according to the

2215: prescription in Appendix~\ref{sec:kleinian-groups} yields

2216: \begin{equation}

2217:   \label{eq:65}

2218:   \deltacheck\bigl( \omega^{1}[\log\rho]

2219:   \bigr)_{\gamma_1,\gamma_2}

2220:   =-r_2((\gamma_1\gamma_2)', \gamma'_2)\,,

2221: \end{equation}

2222: where $r_2$ has been introduced in eqs.~\eqref{eq:18}

2223: and~\eqref{eq:19}. Note that, as in sect.~\ref{sec:cup-product},

2224: the coboundary of $\omega^{1}$ is a cup product in real Deligne

2225: cohomology:

2226: \begin{equation*}

2227:   r_2((\gamma_1\gamma_2)', \gamma'_2)

2228:   = (d\log (\gamma_1\gamma_2)', \log\, \abs{(\gamma_1\gamma_2)'})

2229:   \cup (d\log \gamma'_2, \log\, \abs{\gamma'_2})\,.

2230: \end{equation*}

2231: where the two classes are associated to the rational functions

2232: $(\gamma_1\gamma_2)'$ and $\gamma'_2$, respectively (cf.\ 

2233: sect.~\ref{sec:remarks-dilogarithm}). Hence, we can work with the

2234: double complex of group cochains on $\Gamma$ with values in the

2235: real Deligne complex $\deltilde{\RR}{2}\qi \she{U_\Gamma}(1)$ on

2236: the region of discontinuity $U_\Gamma$. Since from this point on

2237: only rational functions with singularities at certain prescribed

2238: points will appear, following ref.\ \cite{math.AG/0003086} we

2239: will consider the Deligne complex on the generic point

2240: $\eta_{\pione}$ of $\pione$.

2241: 

2242: For any two elements $\gamma_i, \gamma_j \in \Gamma$ define

2243: $T_{ij}\in \PSL_2(\CC)$ by

2244: \begin{equation}

2245:   \label{eq:66}

2246:   z \longmapsto T_{ij} (z) = \CR{z}{z_{ij}}{z_j}{\infty}

2247:   = \frac{z-z_{ij}}{z-z_j}\,.

2248: \end{equation}

2249: where $z_j = \gamma_j^{-1}(\infty)$ and $z_{ij} =

2250: \gamma_j^{-1}(z_i)$. Following ref.\ \cite{math.CV/0204318}, we

2251: introduce the $1$-cochain on $\Gamma$ with values in

2252: $\she[1]{\pione}(1)(\eta_{\pione})$:

2253: \begin{equation}

2254:   \label{eq:68}

2255:   \kappa_\gamma = \log\,\abs{c_\gamma} \, \pi_1 d\log \gamma'\,.

2256: \end{equation}

2257: We have:

2258: \begin{lemma}

2259:   \label{lemma:kappa}

2260:   \begin{math}

2261:     r_2((\gamma_1\gamma_2)', \gamma'_2) =

2262:     4\, d \bigl( \bwli \circ T_{12} \bigr)

2263:     +\deltacheck\kappa_{\gamma_1,\gamma_2}\,,

2264:   \end{math}

2265:   where $\bwli = \sqrt{-1} \bwd$, and $\bwd$ is the standard

2266:   Bloch-Wigner dilogarithm, cf.\ \ref{sec:remarks-dilogarithm}.

2267: \end{lemma}

2268: \begin{proof}

2269:   A straightforward calculation exploiting relation~\eqref{eq:60}.

2270: \end{proof}

2271: Since obviously $d\kappa_\gamma=0$, we can redefine

2272: \begin{math}

2273:   \omega^{1}_\gamma \to \omega^{1}_\gamma + \kappa_\gamma\,,

2274: \end{math}

2275: so that using the lemma, from eq.~\eqref{eq:65} we have:

2276: \begin{equation}

2277:   \label{eq:69}

2278:   \deltacheck\bigl( \omega^{1}[\log\rho]

2279:   \bigr)_{\gamma_1,\gamma_2} 

2280:   = -4\, d\,\bigl(\bwli \circ T_{12} \bigr)\,.

2281: \end{equation}

2282: For convenience of notation, let us temporarily set

2283: \begin{math}

2284:   \Hat\omega^{2}_{\gamma_1,\gamma_2} := -4\, \bigl(\bwli \circ

2285:   T_{12} \bigr)\,.

2286: \end{math}

2287: We then have the following

2288: \begin{lemma}

2289:   \label{lem:3}

2290:   \begin{equation}

2291:     \label{eq:71}

2292:     D(\omega^{0}+\omega^{1}+\Hat\omega^{2}) =

2293:     -4 \,\bwli (

2294:     \CR{\infty}%

2295:     {\gamma_1(\infty)}{\gamma_1\gamma_2(\infty)}%

2296:     {\gamma_1\gamma_2\gamma_3(\infty)}) \,. 

2297:   \end{equation}

2298: \end{lemma}

2299: \begin{proof}

2300:   By construction,

2301:   \begin{math}

2302:     D(\omega^{0}+\omega^{1}+\Hat\omega^{2}) =

2303:     \deltacheck\Hat\omega^{2}\,,

2304:   \end{math}

2305:   and the latter \cech\ coboundary is computed as

2306:   \begin{equation*}

2307:     \begin{split}

2308:       \deltacheck\Hat\omega^{2}_{\gamma_1,\gamma_2, \gamma_3}

2309:       &= -4 \Bigl(\bwli (\CR{z}{z_{23}}{z_3}{\infty})

2310:       -\bwli (\CR{z}{z_{123}}{z_3}{\infty})\\

2311:       &\qquad +\bwli (\CR{z}{z_{123}}{z_{23}}{\infty})

2312:       -\bwli (\CR{\gamma_3(z)}{z_{12}}{z_2}{\infty}) \Bigr)

2313:     \end{split}

2314:   \end{equation*}

2315:   In the last term in the previous relation we have

2316: \begin{math}

2317:   \CR{\gamma_3(z)}{z_{12}}{z_2}{\infty}

2318:   =\CR{z}{z_{123}}{z_{23}}{z_3}\,.

2319: \end{math}

2320: The Bloch-Wigner dilogarithm satisfies the $5$-term

2321: relation~\cite{MR2002g:52013,MR94k:19002}:

2322: \begin{equation}

2323:   \label{eq:70}

2324:   \sum_{i=0}^4 (-1)^i \bwd ([a_0:\dots :\Hat{a_i}:\dots :a_4]) = 0\,,

2325: \end{equation}

2326: where $a_0,\dots, a_4 \in \pione$ and the hat sign denotes

2327: omission. As a consequence we have:

2328: \begin{equation*}

2329:   \deltacheck\Hat\omega^{2}_{\gamma_1,\gamma_2, \gamma_3} =

2330:   -4 \,\bwli (\CR{z_{123}}{z_{23}}{z_3}{\infty}) \,,

2331: \end{equation*}

2332: and again by the invariance of the cross ratio, we

2333: obtain~\eqref{eq:71}.

2334: \end{proof}

2335: \begin{remark}

2336:   \begin{math}

2337:     \bwd (\CR{a}{b}{c}{d} )

2338:   \end{math}

2339:   is the hyperbolic volume of the ideal hyperbolic tetrahedron

2340:   with vertices at the points

2341:   \begin{math}

2342:     a,b,c,d \in \pione\,,

2343:   \end{math}

2344:   see, e.g.\ refs.\ \cite{MR84b:53062b,MR88k:57032,MR94k:19002}.

2345: \end{remark}

2346: It follows from the five-term relation~\eqref{eq:70} that the

2347: right hand side of eq.~\eqref{eq:71} defines an $\RR(1)$-valued

2348: $3$-cocycle on $\Gamma$. Moreover, this cocycle is already defined

2349: on $\PSL_2(\CC)$, where its class is known to generate the

2350: Eilenber-Mac~Lane cohomology group

2351: \begin{math}

2352:   H^3(\PSL_2(\CC),\RR(1))\,,

2353: \end{math}

2354: \cite{MR2001i:11082,MR88k:57032,MR94k:19002}.  It is also known

2355: that up to a factor $2$ it agrees with the imaginary part of the

2356: second Cheeger-Simons universal secondary class $\Hat C_2$.

2357: 

2358: Thus the complete Liouville cocycle

2359: \begin{math}

2360:   \Omega = \omega^{0}+\omega^{1}+\omega^{2}

2361: \end{math}

2362: subordinated to the cover $U_\Gamma \to X$ is found as follows.

2363: The pullback of $\Hat C_2$ along the inclusion map

2364: \begin{math}

2365:   \Gamma \hookrightarrow \PSL_2(\CC)

2366: \end{math}

2367: is zero, since ($\Gamma$ being of second kind) the

2368: $3$-manifold $M_\Gamma = \HH^3/\Gamma$ is non-compact, and

2369: $H^\bullet(\Gamma,\RR(1)) \cong H^\bullet(M_\Gamma,\RR(1))$. It

2370: follows that the restriction of the cocycle given by the

2371: Bloch-Wigner dilogarithm to $\Gamma$ must be a coboundary, hence

2372: there exists a group $2$-cochain $c$ on $\Gamma$ with values in

2373: $\RR(1)$ such that

2374: \begin{equation}\label{eq:67}

2375:  4 \,\bwli (

2376:   \CR{\infty}%

2377:   {\gamma_1(\infty)}{\gamma_1\gamma_2(\infty)}%

2378:   {\gamma_1\gamma_2\gamma_3(\infty)})

2379:   = (\deltacheck c)_{\gamma_1,\gamma_2,\gamma_3}\,.

2380: \end{equation}

2381: It follows that the cochain $c$ provides the necessary

2382: ``integration constants,'' namely the required $2$-cochain on

2383: $\Gamma$ with values in $\she[0]{U_\Gamma}$ to complete the

2384: Liouville cocycle is

2385: \begin{math}

2386:   \omega^{2}_{\gamma_1,\gamma_2}

2387:   = -4\, \bigl(\bwli \circ

2388:   T_{12} \bigr)

2389:   + c_{\gamma_1,\gamma_2}\,.

2390: \end{math}

2391: 

2392: 

2393: \section{Conformal metrics and hermitian holomorphic cohomology}

2394: \label{sec:conf-metr-herm}

2395: 

2396: In sect.~\ref{sec:Conf-metr-Liouv} we presented a construction of a

2397: degree $2$, $\RR(1)$-valued class corresponding to a conformal metric

2398: $\rho\in \cm{X}$, represented by the cocycle $\Omega[\log\rho]$.

2399: Supplemented by the area of $X$ computed with respect to $\rho$, it

2400: provides a global functional for the variational problem associated to

2401: the Liouville equation~\eqref{eq:42}. We now show that it coincides

2402: with the square of the class of $(T_X,\rho)$ in hermitian holomorphic

2403: Deligne cohomology introduced in sect.~\ref{sec:herm-holom-deligne}.

2404: Moreover, we show that this equality holds at the cocycle level. More

2405: generally, without considering the area term, we show that the cup

2406: product of $(L,\rho)$ and $(L',\rho')$ in hermitian holomorphic Deligne

2407: cohomology coincides with the class of $\Omega[\log\rho,\log\rho']$, and

2408: again the equality in fact holds at the cocycle level.

2409: 

2410: \subsection{Comparison on a curve}

2411: \label{sec:comparison-curve}

2412: 

2413: Let $X$ be a compact Riemann surface. From the results of

2414: sect.~\ref{sec:Cup-prod-herm}, the cup product of the classes of

2415: $(L,\rho)$ and $(L',\rho')$ in hermitian holomorphic Deligne cohomology

2416: yields a class in $\dhhH[4]{X}{2}$, and on a curve we only capture the

2417: $2$-dimensional part of this class. Indeed, in the exact

2418: sequence~\eqref{eq:25}, the cohomology class corresponding to the symbol

2419: $\tame{L}{L'}\in \delH[4]{X}{\ZZ}{2}$ is zero, and

2420: $A^{(2,2)}(X)_{\RR(2)}$ is also zero for obvious dimensional reasons, so

2421: we have:

2422: \begin{equation*}

2423:   \dotsm \longrightarrow \delH[3]{X}{\RR}{2}

2424:   \longrightarrow

2425:   \dhhH[4]{X}{2} \longrightarrow 0\,.

2426: \end{equation*}

2427: It follows that the class $[L,\rho]\cup [L',\rho']$ must come from an

2428: element in $\delH[3]{X}{\RR}{2}$. As already remarked, on a curve we

2429: have $\deltilde{\RR}{2} \qi \she{X}(1)[-1]$, thus

2430: $\delH[3]{X}{\RR}{2}\cong \HHH^3(X,\she{X}(1)[-1]) \cong

2431: \HHH^2(X,\she{X}(1)) \cong H^2(X,\RR(1))$, in agreement with the

2432: calculations performed in sect.~\ref{sec:Conf-metr-Liouv}.

2433: 

2434: More in detail, in complex dimension $1$ the second hermitian

2435: holomorphic Deligne complex $\dhh{1}$ simplifies considerably and

2436: diagram~\eqref{eq:39} reduces to

2437: \begin{equation}

2438:   \label{eq:72}

2439:   \begin{CD}

2440:     \ZZ(2) @>-\imath>> \sho{X} @>-d>> \shomega[1]{X} \\

2441:     & & @VV\pi_1V @VV\pi_1V \\

2442:     & & \she[0]{X}(1) @>-d>> \she[1]{X}(1) @>-d>> \she[2]{X}(1)

2443:   \end{CD}

2444: \end{equation}

2445: so that $\dhh{2}$ becomes just the cone of the morphism

2446: $\deligne{\ZZ}{2} \overset{\pi_1}{\longrightarrow}

2447: \she{X}(1)[-1]$. In other words, on a curve $X$ we have that

2448: $\dhh{2}$ is given by the complex

2449: \begin{equation}

2450:   \label{eq:73}

2451:   \begin{CD}

2452:     \ZZ(2)_X @>{-\imath}>>

2453:     \sho{X}  @>{(-d,-\pi_1)}>>

2454:     \shomega[1]{X}\oplus \she[0]{X}(1) @>{-\pi_1+d}>>

2455:     \she[1]{X}(1) @>{d}>>

2456:     \she[2]{X}(1)\,,

2457:   \end{CD}

2458: \end{equation}

2459: where the differentials have been written explicitly. We can see

2460: the complex $\she{X}$ appears as a subcomplex in~\eqref{eq:73}

2461: and the shift of two positions to the right clearly accounts for

2462: the cohomology degree shift from $2$ to $4$.

2463: 

2464: Our main result is the following comparison

2465: \begin{theorem}

2466:   \label{thm:2}

2467:   Let $X$ be a compact Riemann surface of genus $g>1$. Let $(L,\rho)$

2468:   and $(L',\rho')$ be two hermitian holomorphic line bundles. The class

2469:   of $[L,\rho]\cup [L',\rho']$ in $\dhhH[4]{X}{2} \cong H^2(X,\RR(1))$

2470:   coincides with the one represented by the cocycle $\Omega[\log

2471:   \rho,\log \rho']$ constructed in section~\ref{sec:Conf-metr-Liouv}.

2472: \end{theorem}

2473: \begin{proof}

2474:   We have observed above that $\dhhH[4]{X}{2} \cong

2475:   H^2(X,\RR(1))$, and by construction the class of

2476:   \begin{math}

2477:     \Omega[\log\rho,\log\rho']

2478:   \end{math}

2479:   is in $\HHH^2(X,\she{X}(1)) \cong H^2(X,\RR(1))$. Note that for $X$

2480:   connected they must coincide up to a proportionality factor, since

2481:   $H^2(X,\RR(1))\cong \RR(1)$ in this case. In general, we compute the

2482:   proportionality factor from the explicit cocycles from

2483:   sects.~\ref{sec:Cup-prod-herm} and~\ref{sec:Direct-constr-Liouv}

2484:   to~\ref{sec:two-line-bundles}. (Since $\Omega[\log\rho,\log\rho']$ is

2485:   computed under suitable acyclicity assumptions on the cover, so we

2486:   will use such a cover to establish the comparison.)

2487:   

2488:   Let us assume $L$ and $L'$ and their respective hermitian metric

2489:   structures are represented by cocycles of type~\eqref{eq:38} with

2490:   respect to the chosen cover $\cover{U}_X$. Specializing the general

2491:   expression in table~\ref{tab:big-cup} in sect.~\ref{sec:Cup-prod-herm}

2492:   to the case at hand we obtain, with reference to~\eqref{eq:73}:

2493:   \begin{equation}

2494:     \label{eq:74}

2495:     \renewcommand{\arraystretch}{1.3}

2496:     \begin{array}{cc}

2497:       (0,4) & c_{ijk}c'_{klm} \\ \hline

2498:       (1,3) & -c_{ijk} \log \, g'_{kl} \\ \hline

2499:       (2,2) & -\log \, g_{ij} \, d\log\,g'_{jk} \oplus

2500:       -\pi_1\log\,g_{ij}\, \log\,\abs{g'_{jk}} \\ \hline

2501:       (3,1) &

2502:         (1-\alpha) \bigl( d^c\log\,\abs{g_{ij}}\; \onehalf \log\rho'_j

2503:         -\log\,\abs{g_{ij}}\; \onehalf d^c \log\,\rho'_j

2504:         \bigr) 

2505:         +\alpha \bigl( -\onehalf d^c \log\rho_i\;\log\,\abs{g'_{ij}}

2506:         +\onehalf \log\rho_i\; d^c\log\,\abs{g'_{ij}} \bigr) \\ \hline

2507:       (4,0) &

2508:       \alpha\, \delb\del \log\rho_i\;

2509:         \onehalf\log\rho'_i +(1-\alpha)\, \onehalf \log\rho_i \;

2510:         \delb\del\log\rho'_i

2511:     \end{array}

2512:   \end{equation}

2513:   where we have followed the convention explained in the introduction

2514:   for the bidegrees. Let us denote by $\theta^{i}$ the term of bidegree

2515:   $(4-i,i)$ in~\eqref{eq:74} and by $\Theta$ the total cocycle. (For

2516:   simplicity, we suppress $\rho$ and $\rho'$ from the notation.) A

2517:   direct calculation shows that

2518:   \begin{equation*}

2519:     \theta^{0}_i = \omega^{0}_i + d\lambda^{0}_i\,,\qquad

2520:     \theta^{1}_{ij} = \omega^{1}_{ij} -\deltacheck\lambda^{0}_{ij}\,,

2521:   \end{equation*}

2522:   where $\omega^{0}_i$ is given by eq.~\eqref{eq:55}, $\omega^{1}$

2523:   is given by eq.~\eqref{eq:56}, and

2524:   \begin{equation*}

2525:     \lambda^0_i = \alpha\,

2526:     \onehalf d^c \log\rho_i \, \log\rho'_i

2527:     +(1-\alpha)\, \onehalf \log\rho_i \, d^c \log\rho'_i\,.

2528:   \end{equation*}

2529:   Note that at this point we could simply \emph{define} $\Omega = \Theta

2530:   -D\lambda^0$. Furthermore, note that $\Omega$ does not explicitly

2531:   depend on the parameter $\alpha$ from the \bei\ product~\eqref{eq:5}.

2532:   To finish the comparison, let us assume the cover $\cover{U}_X$ allows

2533:   us to find a collection $L_{ijk}\in \sho{X}(U_{ijk})$ such that

2534:   \begin{equation}

2535:     \label{eq:75}

2536:     \begin{aligned}

2537:       d L_{ijk} &= -\log\,g_{ij} \; d\,\log\, g'_{jk}\\

2538:       \deltacheck L _{ijkl} &= -c_{ijk}  \log \, g'_{kl} + n_{ijkl}\,,

2539:     \end{aligned}

2540:   \end{equation}

2541:   as in sect.~\ref{sec:cup-product}. In this way we have 

2542:   \begin{math}

2543:     \Omega =

2544:     (\omega^{0},\omega^{1}, \omega^{2})

2545:   \end{math}

2546:   with $\omega^{2}_{ijk} = -\bwli (g_{ij},g'_{jk})$. $\Omega$ is a

2547:   cocycle of total degree $2$ in $\Tot \Check{C}^\bullet

2548:   (\cover{U}_X,\she{X})$, and it injects (via the exact sequence of the

2549:   cone) into $\Tot\Check{C}^\bullet (\cover{U}_X,\dhh{2})$ as

2550:   \begin{equation*}

2551:     (\omega^{0},\omega^{1}, 0\oplus\omega^{2})\,.

2552:   \end{equation*}

2553:   Then via equations~\eqref{eq:75} it is easily seen that

2554:   \begin{equation*}

2555:     \theta^{2}_{ijk}=d_{\dhh{2}} (-L_{ijk})

2556:     + 0\oplus \omega^{2}_{ijk}\,,

2557:   \end{equation*}

2558:   where $d_{\dhh{2}}$ is the differential in $\dhh{2}$, and therefore

2559:   \begin{equation*}

2560:     \Theta = (\omega^{0},\omega^{1}, 0\oplus \omega^{2})

2561:     +D\lambda^{0} +D(-L,n)\,,

2562:   \end{equation*}

2563:   where we have put $D=d_{\dhh{2}}\pm \deltacheck$ for the total

2564:   differential. Thus the two cocycles constructed via the direct method

2565:   of sect.~\ref{sec:Direct-constr-Liouv} and the cup product of metrized

2566:   bundles define the same class. By direct comparison, the

2567:   proportionality factor is $1$.

2568: \end{proof}

2569: In light of the previous theorem, the polarization identity in

2570: Remark~\ref{rem:1} is now easily explained. Using

2571: \begin{math}

2572:   [L\otimes L',\rho\rho'] = [L,\rho] + [L',\rho']

2573: \end{math}

2574: and

2575: \begin{math}

2576:   [L\otimes {L'}^\vee,\rho/\rho'] = [L,\rho] - [L',\rho'] \,,

2577: \end{math}

2578: and the (graded) commutativity of the cup product

2579: \begin{equation*}

2580:     \dhhH[2]{X}{1} \otimes \dhhH[2]{X}{1}

2581:   \overset{\cup}{\longrightarrow}

2582:   \dhhH[4]{X}{2}\,,

2583: \end{equation*}

2584: we obtain the polarization identity

2585: \begin{equation*}

2586:   4 [L,\rho]\cup [L',\rho'] = [L\otimes L',\rho\rho']^2

2587:   -[L\otimes {L'}^\vee,\rho/\rho']^2\,,

2588: \end{equation*}

2589: where the squares in the right hand side refer to $\cup$. A polarization

2590: identity at the level of representative cocycles, and hence the one in

2591: Remark~\ref{rem:1}, follow by applying Thm.~\ref{thm:2} to the latter

2592: identity.

2593: 

2594: By choosing $L=L'=T_X$, the holomorphic tangent

2595: bundle of $X$, we immediately obtain:

2596: \begin{corollary}

2597:   \label{corollary:1}

2598:   Let $\rho\in\cm{X}$ be a conformal metric. The Liouville functional

2599:   without area term~\eqref{eq:46} is given by the (evaluation of) the

2600:   square $[T_X,\rho]\cup [T_X,\rho]$ with respect to the cup product in

2601:   hermitian holomorphic Deligne cohomology. The full-fledged Liouville

2602:   functional is obtained by adding the area term

2603:   \begin{math}

2604:     \frac{1}{2\pi}\int_X \omega_\rho

2605:   \end{math}

2606:   to~\eqref{eq:46}.

2607: \end{corollary}

2608: \begin{remark}

2609:   Due to the specific form of the differential in the complex $\dhh{2}$

2610:   the descent equations are explicit and close automatically. Therefore

2611:   the cocycle resulting from the calculation of the cup product

2612:   sidesteps the problem of the explicit calculation of the last term,

2613:   unlike the more direct version from sect.~\ref{sec:Conf-metr-Liouv}.

2614:   Thus, thanks to the explicit character of the calculation, specific

2615:   assumptions on the nature of the cover $\cover{U}_X$ are not required.

2616: \end{remark}

2617: It follows from Thm~\ref{thm:2}, corollary~\ref{corollary:1} and

2618: the previous remark that definition~\ref{def:2} applies to any

2619: (\'etale) cover $\cover{U}_X \to X$. Indeed, proposition~\ref{prop:2}

2620: from sect.~\ref{sec:Direct-constr-Liouv} can be reformulated at the

2621: cocycle level as follows:

2622: \begin{proposition}

2623:   \label{prop:3}

2624:   Let $X$ be a compact, genus $g>1$ Riemann surface and let

2625:   $\cover{U}_X\to X$ be a cover. For a conformal metric $\rho \in

2626:   \cm{X}$ and $\sigma \in C^\infty(X,\RR)$, there is a cocycle

2627:   $\Hat{\Omega}_{\cover{U}_X} [\log\rho]$ solving the variational

2628:   problem for the Liouville equation.

2629: \end{proposition}

2630: \begin{proof}

2631:   If $\rho$ is a conformal metric, let the pair $(T_X,\rho)$ be

2632:   represented, as an hermitian line bundle, by a cocycle $c(T_X,\rho)$

2633:   with respect to the cover $\cover{U}_X$. We set

2634:   \begin{equation*}

2635:     \Omega[\log\rho] = c(T_X,\rho)\cup c(T_X,\rho)\,,

2636:   \end{equation*}

2637:   and a simple calculation starting from eq.~\eqref{eq:74} yields

2638:   \begin{equation*}

2639:     \Omega[\log\rho +\sigma] - \Omega[\log\rho] =

2640:     \sigma\,c_1(\rho) + \onehalf\,\sigma\, \del\delb\sigma +

2641:     D\chi\,,

2642:   \end{equation*}

2643:   where

2644:   \begin{math}

2645:     \chi_i = \onehalf \sigma \, \onehalf d^c\log \rho_i

2646:     -\onehalf d^c\sigma\, \onehalf \log\rho_i\,.

2647:   \end{math}

2648:   Now define

2649:   \begin{equation*}

2650:     \Hat{\Omega}_{\cover{U}_X}[\log\rho] = \Omega[\log\rho]

2651:     -\sqrt{-1}\omega_\rho\,. 

2652:   \end{equation*}

2653:   We see that it yields the formula in Proposition~\ref{prop:2}. In

2654:   particular we have that 

2655:   \begin{equation*}

2656:     \left.\frac{d}{dt}\right\rvert_{t=0}

2657:     \Hat{\Omega}_{\cover{U}_X}[\log\rho +t\sigma]

2658:     \equiv \sigma ( c_1(\rho) -\sqrt{-1}\omega_\rho )\,,

2659:   \end{equation*}

2660:   where $\equiv$ means ``up to total coboundary.''

2661: \end{proof}

2662: 

2663: \subsection{Determinant of cohomology}

2664: \label{sec:determ-cohom}

2665: 

2666: Let again $L$ and $L'$ be two holomorphic line bundles with

2667: hermitian metrics $\rho$ and $\rho'$, respectively, on the

2668: compact Riemann surface $X$. Brylinski proves in

2669: \cite{bry:quillen} that the cup product of $L$ and $L'$ in

2670: hermitian holomorphic Deligne cohomology yields the (logarithm)

2671: of the metric $\norm{\cdot}$ on the Deligne pairing $\dual{L}{L'}$

2672: defined in \cite{MR89b:32038}.

2673: 

2674: It follows, via Thm~\ref{thm:2} and the

2675: isomorphism~\ref{lemma:qi-dhh-bry} between our version of hermitian

2676: holomorphic cohomology and Brylinski's, that the class of

2677: $\Omega[\log\rho,\log\rho']$ is also equal to $\log

2678: \norm{\dual{L}{L'}}$. It is worthwhile to provide a direct proof of this

2679: fact starting from the explicit cocycle given in \eqref{eq:74}.

2680: 

2681: First, we need to recall a few definitions from \cite{MR89b:32038}. A

2682: complex line $\dual{L}{L'}$ is assigned to the pair $(L, L')$ as

2683: follows.  Let $D$ and $D'$ be divisors on $X$ corresponding to $L$ and

2684: $L'$, and assume they have disjoint supports. Consider two rational

2685: sections, $s$ and $s'$ such that $(s)=D$ and $(s')=D'$. To this datum

2686: one assigns a copy of the complex line generated by the symbol

2687: $\dual{s}{s'}$ subject to the relations:

2688: \begin{equation}

2689:   \label{eq:76}

2690:   \begin{aligned}

2691:     \dual{fs}{s'} &= f(D') \dual{s}{s'}\\

2692:     \dual{s}{gs'} &= g(D) \dual{s}{s'}

2693:   \end{aligned}

2694: \end{equation}

2695: whenever $f$ is a rational function with divisor $(f)$ disjoint from

2696: $D'$, and similarly for $g$. The Weil reciprocity relation

2697: $f(\mathit{div}(g)) = g(\mathit{div}(f))$ (cf. ref.\ \cite{gh:alg_geom})

2698: for two rational functions $f$ and $g$ with disjoint divisors implies

2699: that the relations~\eqref{eq:76} are consistent and the complex line

2700: depends only on the pair $L$, $L'$. When the line bundles are equipped

2701: with hermitian metrics, generically denoted by $\norm{\cdot}$, the

2702: assignment\footnote{We write the square explicitly, whereas the symbol

2703:   $\norm{\cdot}$ used in ref.~\cite{MR89b:32038} denotes the

2704:   \emph{square} of the norm.}

2705: \begin{equation}

2706:   \label{eq:77}

2707:   \log\, \norm{\dual{s}{s'}}^2 = \frac{1}{\tate} 

2708:   \int_X \del\delb\log \norm{s}^2 \,\log\norm{s'}^2

2709:   + \log\norm{s}^2[D'] +\log\norm{s'}^2[D]

2710: \end{equation}

2711: is compatible with the relations~\eqref{eq:76} and defines an hermitian

2712: metric on the complex line $\dual{L}{L'}$. In formula~\eqref{eq:77} the

2713: operator $\del\delb$ is to be computed in the sense of distributions.

2714: 

2715: Having covered the main definitions, we can now state

2716: \begin{theorem}

2717:   \label{thm:3}

2718:   The cup product of $(L,\rho)$ with $(L',\rho')$ in hermitian

2719:   holomorphic Deligne cohomology corresponds to the logarithm of the

2720:   norm \eqref{eq:77} on the Deligne pairing $\dual{L}{L'}$. The

2721:   proportionality factor is $-\pi\,\sqrt{-1}$.

2722: \end{theorem}

2723: \begin{proof}

2724: Let $D$ and $D'$ be divisors with disjoint support on $X$ corresponding

2725: to $L$ and $L'$, respectively. Using the same technique as in refs.

2726: \cite{bry:quillen,MR89b:32038}, consider two $C^\infty$ positive real

2727: functions $f_1$ and $f_2$ such that $f_1+f_2=1$ and $f_1$ (resp. $f_2$)

2728: vanishes in a neighborhood of the support of $D'$ (resp. $D$). Also, set

2729: $U_1 = X\setminus \mathrm{supp}(D')$ and $U_2 = X\setminus

2730: \mathrm{supp}(D)$. Thus $\{f_1,f_2\}$ is just a partition of unity

2731: subordinated to the cover $\cover{U}_X = \{ U_1, U_2\}$.

2732: 

2733: The only two terms different from zero in the cocycle $\Theta$

2734: in~\eqref{eq:74} representing the class $[L,\rho]\cup [L',\rho']$ with

2735: respect to this cover are $\theta^{0}_i$ and $\theta^{1}_i$, with the

2736: \cech\ index $i \in \{ 1,2 \}$. Thus the class we are after is

2737: equivalently given by the integral

2738: \begin{equation}

2739:   \label{eq:78}

2740:   \int_X f_1\theta^{0}_1 +f_2\theta^{1}_2

2741:   + df_2 \wedge \theta^{1}_{21} \,,

2742: \end{equation}

2743: which is arrived at by applying in the standard homotopy operator based

2744: on the partition of unity $\{f_1,f_2\}$: from

2745: $\deltacheck\theta^{1}_{12}=0$ we obtain that $\theta^{1}_{12}$ is the

2746: coboundary of the cochain $j\to \sum_{i=1,2} f_i\theta^{1}_{ij}$ and

2747: then we use $\theta^{0}_2-\theta^{0}_2=-d\theta^{1}_{12}$. Observe that

2748: the $2$-form in~\eqref{eq:78} is globally well defined over $X$:

2749: $\theta^{1}_{21}$ is defined only on $U_1\cap U_2$, but $df_2$ has

2750: support on $U_1\cap U_2$, so their wedge product is defined everywhere;

2751: similarly, $f_i\,\theta^{0}_i$ is everywhere defined thanks to the fact

2752: that $f_i$ has support in $U_i$, $i=1,2$.

2753: 

2754: Consider rational sections $s$ and $s'$ of $L$ and $L'$ such that

2755: $\mathit{div}(s)=D$ and $\mathit{div}(s')=D'$ as above. With respect to

2756: the two-element cover $\cover{U}_X=\{U_1,U_2\}$, the section $s$

2757: corresponds to the pair $\{s_1,s_2\}$, and similarly for

2758: $s'=\{s'_1,s'_2\}$. Since $\mathrm{supp}(D)$ is contained in $U_1$ but

2759: not in $U_2$, and the other way around for $D'$, it follows that $s_2$

2760: and $s'_1$ are actually invertible functions over their respective

2761: domains. Following \cite{lang:arakelov}, we can assume that $s$ and $s'$

2762: are in fact the rational section $\mathbf{1}$, so that $s_2=1$ and

2763: $s'_1=1$, and therefore:

2764: \begin{gather*}

2765:   \abs{s}^2_\rho =

2766:   \begin{cases}

2767:     \log \rho_1 + \log \abs{s_1}^2 & \text{on } U_1\,,\\

2768:     \log \rho_2 & \text{on } U_2\,,

2769:   \end{cases}\\ \intertext{and}

2770:   \abs{s'}^2_{\rho'} =

2771:   \begin{cases}

2772:     \log\, \rho'_1 & \text{on } U_1\,,\\

2773:     \log \rho'_2 + \log\, \abs{s'_2}^2 & \text{on } U_2\,.

2774:   \end{cases}\\

2775: \end{gather*}

2776: Furthermore, 

2777: \begin{math}

2778:   \log \, \abs{g'_{21}} = \log \, \abs{s'_2}

2779: \end{math}

2780: on $U_1\cap U_2$. 

2781: Let us denote by $\norm{\cdot}=\abs{\cdot}_\rho= \abs{\cdot}_{\rho'}$

2782: for simplicity. Using the relevant entries from~\eqref{eq:74} we have:

2783: \begin{equation*}

2784:   \begin{gathered}

2785:     \theta^{0}_1 = c_1(\rho)\,\log \norm{s'}\,,\\

2786:     \theta^{0}_2 = \delb\del \log \rho_2 \; \onehalf \log \rho'_2

2787:     = dd^c \,\log\,\norm{s} \; \onehalf \log\rho'_2\,.    

2788:   \end{gathered}

2789: \end{equation*}

2790: An elementary integration by parts leads to:

2791: \begin{equation*}

2792:   \begin{split}

2793:     \int_X f_2 \theta^{0}_2

2794:     &= \int_X f_2 \log\,\norm{s}\, c_1(\rho') \\

2795:     &+ \int_X \log\,\norm{s} \, df_2 \wedge \onehalf d^c \log\rho'_2

2796:     - \int_X \onehalf \log\rho'_2  \, df_2 \wedge d^c \log\,\norm{s}\,.

2797:   \end{split}

2798: \end{equation*}

2799: On the other hand, we have

2800: \begin{equation*}

2801:   \int_X df_2 \wedge \theta^{1}_{12} =

2802:   \int_X \log\,\norm{s}\; df_2\wedge d^c \log\,\abs{s'_2}

2803:   -\int_X \log\,\abs{s'_2} \; df_2\wedge d^c \log\,\norm{s}\,,

2804: \end{equation*}

2805: and putting all terms together we obtain

2806: \begin{equation}

2807:   \label{eq:79}

2808:   \begin{split}

2809:     \int_X f_1\theta^{0}_1 +f_2\theta^{1}_2

2810:     + df_2 \wedge \theta^{1}_{21}

2811:     &=\int_X f_1\,c_1(\rho)\,\log \norm{s'}

2812:     +\int_X f_2 \log\,\norm{s}\, c_1(\rho') \\

2813:     &+ \int_X \log\,\norm{s}\; df_2\wedge d^c \log\,\norm{s'}

2814:   +\int_X \log\,\norm{s'} \; df_1\wedge d^c \log\,\norm{s}

2815:   \end{split}

2816: \end{equation}

2817: which, if expressed in terms of the squares of the norms, is (up

2818: to a factor) the logarithm of $\norm{\dual{s}{s'}}$, as it is

2819: found in \cite[formula 6.5.1]{MR89b:32038}. This version is due

2820: to O. Gabber. Via the Poincar\'e-Lelong lemma (see,

2821: e.g.\ \cite{gh:alg_geom})

2822: \begin{equation*}

2823:   \delb\del \log\,\norm{s} = c_1(\rho) + \tate\,[D]\,,

2824: \end{equation*}

2825: where $[D]$ is the delta-current supported at the divisor of $s$,

2826: and similarly for $s'$, formula~\eqref{eq:79} can be recast into:

2827: \begin{equation}

2828:   \label{eq:80}

2829:   \begin{split}

2830:     \int_X f_1\theta^{0}_1 +f_2\theta^{1}_2

2831:     + df_2 \wedge \theta^{1}_{21} &=

2832:     \int_X dd^c \log\,\norm{s}\; \log\,\norm{s'}

2833:     -\tate\,\log\,\norm{s} [D']

2834:     -\tate\,\log\,\norm{s'} [D] \\

2835:     &= -\tate\; \log\,\norm{\dual{s}{s'}}

2836:   \end{split}

2837: \end{equation}

2838: which is what we wanted to show.

2839: \end{proof}

2840: This allows us to recast the Liouville functional for conformal metrics

2841: on $X$ in the following form.

2842: \begin{corollary}

2843:   \label{corollary:2}

2844:   The exponential of the Liouville functional defines an hermitian

2845:   metric on the complex line $\dual{T_X}{T_X}$, namely for a conformal

2846:   metric $\rho\in \cm{X}$ we have:

2847:   \begin{equation*}

2848:     \exp S[\log\rho] = \norm{\dual{T_X}{T_X}}\, \exp

2849:     \tfrac{1}{2\pi} A_X(\rho)\,,

2850:   \end{equation*}

2851:   where $A_X(\rho)$ is the area of $X$ with respect to $\rho$.

2852: \end{corollary}

2853: \begin{remark}

2854:   The above corollary justifies the choice made in

2855:   Definition~\ref{def:2} for the various factors $\tate$.

2856: \end{remark}

2857: A similar result has been obtained in ref.\ \cite{zograf1990} by

2858: considering the Liouville action functional defined on the Schottky

2859: space, and in fact the results in loc.\ cit. are formulated in terms of

2860: a Schottky \emph{family.}

2861: 

2862: Indeed, the statement in Corollary~\ref{corollary:2} can be immediately

2863: reformulated for a family $\pi \colon X\rightarrow S$ with base

2864: parameter space $S$ by considering the relative holomorphic tangent line

2865: bundle $T_{X/S}$ with an hermitian fiber metric $\rho$. (Thus $\rho_s\in

2866: \cm{X_s}$ for every fiber $X_s\,, s\in S$.) Notice that the fiber metric

2867: $\rho$ needs not be critical (i.e.\ satisfying the fiberwise constant

2868: negative curvature condition).

2869: 

2870: 

2871: \section{Conclusions and outlook}

2872: \label{sec:conclusions-outlook}

2873: 

2874: In this paper we have provided a geometric description of the

2875: construction of the Liouville action functional for constant

2876: negative curvature metrics on compact Riemann surfaces of genus

2877: $g\geq 2$.

2878: 

2879: Our approach was to construct a Hermitian holomorphic Deligne

2880: cohomology class as the cup square of the metrized holomorphic

2881: tangent bundle $T_X$, and then show at the level of cocycles that

2882: (modulo an area term) this construction agrees with those

2883: in~\cite{math.CV/0204318} and earlier works.

2884: 

2885: Furthermore, our construction leads to the identification of the

2886: class corresponding to the Liouville action with the determinant

2887: of cohomology construction. Hence it could serve as an

2888: alternative construction of the latter in terms of different

2889: choices of cocycles.

2890: 

2891: One of the most important properties of the Liouville functional

2892: from the works~\cite{zogtak1987-1,zogtak1987-2,zograf1990} is the

2893: link with the Weil-Petersson geometry of the Teichm\"uller space.

2894: From this point of view, an analysis of the behavior of the

2895: Liouville action for families of Riemann surfaces is crucial. In

2896: particular, a delicate analytic computation of the variation of

2897: the Action with respect to the moduli was carried out in

2898: the afore mentioned works to establish the link with the

2899: Weil-Petersson metric.

2900: 

2901: We have not pursued these matters in the present work limiting

2902: ourselves to establish the existence of a class for a metrized

2903: bundle on the base of a family $\pi\colon X\to S$. A more precise

2904: analysis would require, not only a more explicit description of

2905: the map $\int_\pi$ (also advocated in~\cite{brymcl:deg4_II}), but

2906: a full computation of the Leray sequence associated to the family

2907: $\pi$. This is instrumental in defining \emph{relative} Hermitian

2908: holomorphic classes, and in analyzing their variation with

2909: respect to base parameters.

2910: 

2911: We hope to pursue this and other problems related to the

2912: extension of the present work to singular metrics in a different

2913: publication.

2914: 

2915: 

2916: \appendix

2917: 

2918: \section{Cones}

2919: \label{sec:Cones}

2920: 

2921: In the main body of the paper we have used iterated cones to

2922: define the hermitian holomorphic Deligne complexes. One technical

2923: problem one has to face concerns the homotopy (graded)

2924: commutativity of the modified \bei\ product defined in

2925: eq.~\eqref{eq:5}. A problem arises because the factors in the

2926: cones are cones themselves and therefore they have multiplication

2927: structures which are graded commutative up to homotopy to begin

2928: with. We want to show that even in this situation the final

2929: resulting product on cones is again homotopy graded commutative.

2930: This ensures that on cohomology the product will be genuinely

2931: graded commutative, so that in particular hermitian holomorphic

2932: Deligne cohomology as defined in

2933: section~\ref{sec:herm-holom-deligne} has the correct product

2934: structure.

2935: 

2936: 

2937: \subsection{Cones and homotopies}

2938: \label{sec:cones-homotopies}

2939: 

2940: We consider the following situation. For $i=1,2,3$ we have maps of

2941: complexes: 

2942: \begin{math}

2943:   f_i \colon A^\bullet_i \rightarrow B^\bullet_i\,,

2944: \end{math}

2945: and for $i<j$ maps

2946: \begin{math}

2947:   a_{ji} \colon A^\bullet_{i} \rightarrow A^\bullet_j

2948: \end{math}

2949: and

2950: \begin{math}

2951:   b_{ji} \colon B^\bullet_{i} \rightarrow B^\bullet_j\,.

2952: \end{math}

2953: Also, let

2954: \begin{math}

2955:   C^\bullet(f_i) =\cone (f_i:A^\bullet_i\rightarrow B^\bullet_i) \,,

2956: \end{math}

2957: for $i=1,2,3$. 

2958: 

2959: First, consider the homotopy commutative diagram:

2960: \begin{equation}

2961:   \label{eq:81}\vcenter{%

2962:   \xymatrix{%

2963:     A^\bullet_j \ar[r]^{f_j} \ar[d]_{a_{ij}} & B^\bullet_j

2964:     \ar[d]^{b_{ij}} \ar@{=>}[dl]^{s_{ij}} \\

2965:     A^\bullet_i \ar[r]_{f_i} & B^\bullet_i}}

2966: \end{equation}

2967: where

2968: \begin{math}

2969:   s_{ij} \colon A^\bullet_j \rightarrow B^{\bullet -1}_i

2970: \end{math}

2971: is the homotopy map of complexes:

2972: \begin{equation*}

2973:   f_ia_{ij} -b_{ij}f_j  = d\,s_{ij} + s_{ij}d\,.

2974: \end{equation*}

2975: An immediate verification yields

2976: \begin{lemma}

2977:   \label{lemma:cone-map}

2978:   The diagram~\eqref{eq:81} can be extended to

2979:   \begin{equation*}

2980:     \xymatrix{%

2981:       A^\bullet_j \ar[r]^{f_j} \ar[d]_{a_{ij}} &

2982:       B^\bullet_j \ar[r] \ar[d]^{b_{ij}} \ar@{=>}[dl]^{s_{ij}} &

2983:       C^\bullet (f_j) \ar[d]^{c_{ij}} \ar[r]^{[1]} &

2984:       A^\bullet_j \ar[d]^{a_{ij}}\\

2985:       A^\bullet_i \ar[r]_{f_i} & B^\bullet_i \ar[r] &

2986:       C^\bullet (f_i) \ar[r]_{[1]} & A^\bullet_i

2987:     }

2988:   \end{equation*}

2989:   where the map $c_{ij}$ is given by

2990:   \begin{math}

2991:     \bigl(

2992:     \begin{smallmatrix}

2993:       a_{ij} & 0 \\ -s_{ij} & b_{ij}

2994:     \end{smallmatrix}

2995:     \bigr)

2996:   \end{math}

2997:   and the squares containing the cones are in fact strictly commutative.

2998: \end{lemma}

2999: \begin{remark}

3000:   This lemma is nothing other than the statement that any

3001:   homotopy commutative diagram of the form~\eqref{eq:81} in the

3002:   category of complexes in an abelian category can be extended to

3003:   a (homotopy) commutative diagram of distinguished triangles,

3004:   that is, one of the axioms defining a triangulated category,

3005:   see, e.g. \cite{weibel_hom_alg}.

3006: \end{remark}

3007: For $k<j<i$ consider the homotopy commutative triangle

3008: \begin{equation*}

3009:   \xymatrix{%

3010:     & A^\bullet_j  \ar[dr]^{a_{ij}} \ar@{=>}[d]^(.6){\alpha_{ijk}}& \\

3011:     A^\bullet_k \ar[0,2]_{a_{ik}} \ar[ur]^{a_{jk}} & &

3012:     A^\bullet_i}

3013: \end{equation*}

3014: where

3015: \begin{math}

3016:   a_{ik} -a_{ij}a_{jk} = d\alpha_{ijk} + \alpha_{ijk} d\,,

3017: \end{math}

3018: and similarly for the complexes $B^\bullet_i$, with a

3019: corresponding homotopy $\beta_{ijk}$. Thus

3020: \begin{math}

3021:   \alpha_{ijk} \colon

3022:   A^\bullet_k \rightarrow A^{\bullet -1}_i

3023: \end{math}

3024: and

3025: \begin{math}

3026:   \beta_{ijk} \colon

3027:   B^\bullet_k \rightarrow B^{\bullet -1}_i\,.

3028: \end{math}

3029: Now consider the diagram:

3030: \begin{equation}

3031:   \label{eq:82}\vcenter{%

3032:   \xymatrix@!{%

3033:     A^\bullet_k \ar[0,2]^{a_{jk}} \ar[dr]_(.7){a_{ik}} \ar[2,0]_{f_k} & &

3034:     A^\bullet_j \ar[dl]^{a_{ij}} \ar[2,0]^{f_j}

3035:     \ar@{=>}[];[dl]+/ul 5ex/\\

3036:     & A^\bullet_i \ar[2,0]^(.4){f_i} & \\

3037:     B^\bullet_k \ar'[r]^(.7){b_{jk}}[rr] \ar[dr]_{b_{jk}}

3038:     \ar@{=>}[ur] \ar@2{-->}[-1,0];[-2,1] & &

3039:     B^\bullet_j \ar[dl]^{b_{ij}} \ar@{=>}[ul]

3040:     \ar@{=>}[];[dl]+/ul 5ex/ \\

3041:     & B^\bullet_i &

3042:   }}

3043: \end{equation}

3044: The faces in~\eqref{eq:82} are homotopy commutative, however we assume

3045: that composing the \emph{faces} is strictly commutative, namely the two

3046: possible homotopies

3047: \begin{math}

3048:   b_{ij}\,b_{jk}\,f_k \Longrightarrow f_i\,a_{ik}

3049: \end{math}

3050: must be equal. Concretely, this corresponds to the relation

3051: \begin{equation}

3052:   \label{eq:83}

3053:   s_{ik} + \beta_{ijk}\,f_{k} = f_i\,\alpha_{ijk}

3054:   +s_{ij}\,a_{jk} +b_{ij}\,s_{jk}\,.

3055: \end{equation}

3056: We have:

3057: \begin{lemma}

3058:   \label{lemma:cone-homotopy}

3059:   The map 

3060:   \begin{equation*}

3061:     \bigl(

3062:     \begin{smallmatrix}

3063:       -\alpha_{ijk} & 0 \\ 0 & \beta_{ijk}

3064:     \end{smallmatrix}

3065:     \bigr) \colon C^\bullet (f_k) \longrightarrow

3066:     C^{\bullet -1}(f_i)

3067:   \end{equation*}

3068:   realizes the homotopy

3069:   \begin{equation*}

3070:     \xymatrix{%

3071:       & C^\bullet (f_j) \ar[dr]^{c_{ij}} \ar@{=>}[d]& \\

3072:       C^\bullet (f_k) \ar[ur]^{c_{jk}} \ar[rr]_{c_{ik}} & &

3073:       C^\bullet (f_i)}

3074:   \end{equation*}

3075: \end{lemma}

3076: \begin{proof}

3077:   It is an elementary calculation based on writing 

3078:   \begin{math}

3079:     c_{ik} - c_{ik}\,c_{jk}

3080:   \end{math}

3081:   explicitly via the matrix representation given in

3082:   Lemma~\ref{lemma:cone-map} and using eq.~\eqref{eq:83}.

3083: \end{proof}

3084: 

3085: 

3086: \subsection{Applications}

3087: \label{sec:applications}

3088: 

3089: Consider the same setup as in section~\ref{sec:cup-products-cones}, with

3090: the same complexes $X^\bullet_i$, etc., and diagrams:

3091: \begin{equation*}

3092:   \mathcal{D}_i \eqdef

3093:   X^\bullet_i \overset{f_i}{\longrightarrow} Z^\bullet_i

3094:   \overset{g_i}{\longleftarrow} Y^\bullet_i

3095: \end{equation*}

3096: from which we construct the cones

3097: \begin{equation*}

3098:   C(\mathcal{D}_i) =

3099:   \cone (X^\bullet_i\oplus Y^\bullet_i

3100:   \xrightarrow{f_i-g_i} Z^\bullet_i)[-1]\,,

3101:   \quad i=1,2,3\,.

3102: \end{equation*}

3103: Moreover, following ref.\ \cite{bei:hodge_coho}, define

3104: $\mathcal{D}_i\otimes \mathcal{D}_j$ by taking the tensor product

3105: component-wise. Thus

3106: \begin{equation*}

3107:   \mathcal{D}_1\otimes \mathcal{D}_2 = 

3108:   X^\bullet_1 \otimes X^\bullet_2

3109:   \xrightarrow{f_1\otimes f_2}

3110:   Z^\bullet_1\otimes Z^\bullet_2

3111:   \xleftarrow{g_1\otimes g_2}

3112:   Y^\bullet_1\otimes Y^\bullet_2\,.

3113: \end{equation*}

3114: Assuming as in section~\ref{sec:cup-products-cones} that the product

3115: maps are compatible with the $f_i$, etc., the diagram

3116: \begin{equation*}

3117:   \mathcal{D}_1\otimes \mathcal{D}_2 \rightarrow \mathcal{D}_3

3118: \end{equation*}

3119: is of the same type as~\eqref{eq:81}, and therefore

3120: lemma~\ref{lemma:cone-map} implies lemma~\ref{lemma:mod-cup-prod}.

3121: 

3122: Now, let the multiplication maps

3123: \begin{math}

3124:   X^\bullet_1 \otimes X^\bullet_2 \rightarrow X^\bullet_3

3125: \end{math}

3126: be graded commutative up to homotopy and similarly for the

3127: $Y^\bullet_i$ and the $Z^\bullet_i$. We are interested in the

3128: commutativity properties of multiplication map given by the \bei\

3129: product~\eqref{eq:5}. 

3130: \begin{proposition}

3131:   \label{prop:4}

3132:   The multiplication map

3133:   \begin{math}

3134:     \cup_\alpha\colon C(\mathcal{D}_1)\otimes C(\mathcal{D}_2)

3135:     \longrightarrow C(\mathcal{D}_3)

3136:   \end{math}

3137:   given by~\eqref{eq:5} is homotopy graded commutative.

3138: \end{proposition}

3139: \begin{proof}

3140:   The permutation operation on tensor products induces the diagram

3141:   \begin{equation*}

3142:     \xymatrix{%

3143:       {\mathcal{D}_1}\otimes {\mathcal{D}_2}  \ar[dr] \ar[rr] & &

3144:       {\mathcal{D}_2}\otimes {\mathcal{D}_1} \ar[dl] \\

3145:       & {\mathcal{D}}_3 & }

3146:   \end{equation*}

3147:   which is of type~\eqref{eq:82} and we can apply

3148:   lemma~\ref{lemma:cone-homotopy}.

3149: \end{proof}

3150: It follows from the proposition that the cohomology inherits a well

3151: defined graded commutative product. This in particular applies to the

3152: definition of hermitian holomorphic Deligne cohomology that uses the

3153: cone~\eqref{eq:22}. Therefore we conclude that the cup

3154: product~\eqref{eq:27} is graded commutative, as wanted.

3155: 

3156: \section{Conventions on Kleinian groups and fractional linear

3157:   transformations} 

3158: \label{sec:kleinian-groups}

3159: 

3160: As a reference the reader can consult, among many others, the

3161: book \cite{MR99g:30055}. Let $\Gamma$ be a finitely generated

3162: purely loxodromic non-elementary Kleinian group of the second

3163: kind, so there is a nonempty discontinuity region

3164: $U_\Gamma\subset \pione$. The limit set is $L_\Gamma = \pione

3165: \setminus U_\Gamma$. According to Ahlfors' finiteness theorem the

3166: quotient $X=U_\Gamma/\Gamma$ is a finite union of analytically

3167: finite Riemann surfaces. Thus:

3168: \begin{equation*}

3169:   U_\Gamma/\Gamma = U_1/\Gamma_1 \sqcup \dots \sqcup U_n/\Gamma_n\,,

3170: \end{equation*}

3171: where $U_1,\dots,U_n$ are the inequivalent components of

3172: $U_\Gamma$ and $\Gamma_1, \dots, \Gamma_n$ their stabilizers. By

3173: way of example, a Schottky group has just one component, whereas

3174: a Fuchsian or quasi-Fuchsian group will have exactly two

3175: components.

3176: 

3177: We consider the map $U_\Gamma \to X$ as an \'etale cover, and

3178: \cech\ cohomology with respect to it translates into group

3179: cohomology for the group $\Gamma$, where the coefficient modules

3180: are sections over $U_\Gamma$ of the relevant sheaves. The group

3181: action is by pull-back.  According to the conventions of ref.\ 

3182: \cite{aldtak2000} we will write the coboundary operation as:

3183: \begin{equation}

3184:   \label{eq:58}

3185:     (\deltacheck c)_{\gamma_1,\dots,\gamma_n} =

3186:     c_{\gamma_2,\dots,\gamma_n} 

3187:     +\sum_{i=1}^{n-1}(-1)^i

3188:     c_{\gamma_1,\dots,\gamma_i\gamma_{i+1},\dots,\gamma_n}

3189:     +(-1)^n (c_{\gamma_1,\dots,\gamma_{n-1}})\cdot {\gamma_n}\,,

3190: \end{equation}

3191: where $c$ is an $n$-cochain with values in some \emph{right}

3192: $\Gamma$-module $A$. The expression~\eqref{eq:58} is the \cech\ 

3193: coboundary applied to the nerve of the cover $U_\Gamma \to X$.

3194: 

3195: We assume that $\Gamma$ is normalized, namely the point $\infty$

3196: belongs to the limit set $L_\Gamma$. If $\gamma\in \Gamma$

3197: corresponds to the fractional linear transformation:

3198: \begin{equation*}

3199:   \pione \ni z \longmapsto \gamma (z) = \frac{az +b}{cz +d}\,,

3200: \end{equation*}

3201: we have

3202: \begin{equation}

3203:   \label{eq:59}

3204:   \gamma' (z) = \frac{\det \gamma}{c^2(z -z_\gamma)^2}\,,

3205: \end{equation}

3206: where $z_\gamma = -\frac{d}{c} \equiv \gamma^{-1}(\infty)$. Set

3207: \begin{equation*}

3208:   c(\gamma) \equiv c_\gamma \eqdef \frac{\det \gamma}{c^2}\,.

3209: \end{equation*}

3210: The following properties are easily verified. If $z_0$ and $z_\infty$

3211: are the attracting and repelling fixed points, respectively, then

3212: \begin{equation*}

3213:   c_\gamma =

3214:   \frac{(z_0 -z_\infty)^2\lambda_\gamma}{(1-\lambda_\gamma)^2} \,,

3215: \end{equation*}

3216: where $\lambda_\gamma$ is the dilating factor. For $\gamma_i$ and

3217: $\gamma_j$ two elements of $\Gamma$, denote:

3218: \begin{equation*}

3219:   z_i = \gamma^{-1}(\infty)\,, \quad

3220:   z_{ij} = (\gamma_i\gamma_j)^{-1}(\infty) =

3221:   \gamma^{-1}_j (z_i)\,, \quad c_i = c_{\gamma_i}\,, \quad

3222:   c_{ij} = c_{\gamma_i\gamma_j}\,.

3223: \end{equation*}

3224: We have the following relation:

3225: \begin{equation}

3226:   \label{eq:60}

3227:   c_{\gamma_1\gamma_2}

3228:   = \frac{c_{\gamma_1}}{c_{\gamma_2}}\, (z_{12} -z_2)^2\,.

3229: \end{equation}

3230: Finally, given four points $z_1,z_2, z_3, z_4\in \pione$, we define

3231: their cross-ratio by:

3232: \begin{equation}

3233:   \label{eq:61}

3234:   \CR{z_1}{z_2}{z_3}{z_4}

3235:   = \frac{(z_1-z_2)(z_3-z_4)}{(z_1-z_4)(z_3-z_2)}\,.

3236: \end{equation}

3237: 

3238: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3239: \bibliography{general}

3240: \bibliographystyle{hamsplain}

3241: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3242: 

3243: \end{document}

3244: 

3245: 

3246: % LocalWords:  holomorphic uniformization Liouville Ettore Aldrovandi Deligne

3247: % LocalWords:  Brylinski Quillen Kleinian Wigner Poincar Polyakov's variational

3248: % LocalWords:  Teichm uller homological ref Takhtajan cocycle etale Fuchsian

3249: % LocalWords:  loc Krasnov diffeomorphisms refs cohomological Friedan Shenker

3250: % LocalWords:  dilogarithms regularized submanifold imbedding

3251: