1: %

2: \documentclass[12pt]{article}

3: \usepackage{amsmath, amsthm} 

4: \usepackage{multirow} 

5: \usepackage{amsfonts}

6: \usepackage{amssymb,graphics,psfrag}

7: \usepackage{array,epsfig,multirow,stmaryrd,graphicx}

8: \usepackage{comment}

9: %\usepackage[notref,notcite]{showkeys}

10: %\usepackage{cite}

11: %\usepackage{showkeys}

12: %

13: %

14: \def\hybrid{\topmargin -20pt    \oddsidemargin 0pt

15:         \headheight 0pt \headsep 0pt

16: %       \textwidth 6.5in        % US paper

17: %       \textheight 9in         % US paper

18:         \textwidth 6.25in       % A4 paper

19:         \textheight 9.5in       % A4 paper

20:         \marginparwidth .875in

21:         \parskip 5pt plus 1pt   \jot = 1.5ex}

22: %       The default is set to be hybrid

23: \hybrid

24: \numberwithin{equation}{section}

25: \numberwithin{table}{section}\setlength{\multlinegap}{25pt}   

26: 

27: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

28: % abbreviate  environments

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

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

31: \newcommand{\be}{\begin{equation}}

32: \newcommand{\ee}{\end{equation}}

33: %\newcommand{\bi}{\begin{itemize}}

34: %\newcommand{\ei}{\end{itemize}}

35: \newcommand{\bea}{\begin{eqnarray}}

36: \newcommand{\eea}{\end{eqnarray}}   

37: \newcommand{\ben}{\begin{eqnarray*}}

38: \newcommand{\een}{\end{eqnarray*}}                  

39: \newcommand{\ba}{\begin{aligned}}

40: \newcommand{\ea}{\end{aligned}}

41: %\newcommand{\ba}{\begin{array}}

42: %\newcommand{\ea}{\end{array}}

43: \newcommand{\bt}{\begin{tabular}}

44: \newcommand{\et}{\end{tabular}}

45: \newcommand{\bc}{\begin{center}}

46: \newcommand{\ec}{\end{center}}

47: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

48: % abbreviate Greek

49: \newcommand{\ax}{\alpha}

50: \newcommand{\bx}{\beta}

51: \newcommand{\cx}{\gamma}

52: \newcommand{\dx}{\delta}

53: \newcommand{\ox}{\omega}

54: \newcommand{\lx}{\lambda}

55: \newcommand{\ab}{\bar\alpha}

56: \newcommand{\bb}{\bar\beta}

57: \newcommand{\cb}{\bar\gamma}

58: \newcommand{\db}{\bar\delta}

59: \newcommand{\Sx}{\Sigma}

60: \newcommand{\Lx}{\Lambda}

61: \newcommand{\Ox}{\Omega}

62: \newcommand{\Dx}{\Delta}

63: \newcommand{\Gx}{\Gamma}

64: \newcommand{\Oxb}{\bar{\Omega}}

65: %

66: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

67: % Cal

68: \newcommand{\cO}{\mathcal{O}}

69: \newcommand{\cT}{\mathcal{T}}

70: \newcommand{\cE}{\mathcal{E}}

71: \newcommand{\cP}{\mathcal{P}}

72: \newcommand{\cC}{\mathcal{C}}

73: \newcommand{\cD}{\mathcal{D}}

74: \newcommand{\cL}{\mathcal{L}}

75: \newcommand{\cS}{\mathcal{S}}

76: \newcommand{\cK}{\mathcal{K}}

77: \newcommand{\cN}{\mathcal{N}}

78: \newcommand{\cW}{\mathcal{W}}

79: \newcommand{\cG}{\mathcal{G}}

80: \newcommand{\cA}{\mathcal{A}}

81: \newcommand{\cH}{\mathcal{H}}

82: \newcommand{\cB}{\mathcal{B}}

83: \newcommand{\cF}{\mathcal{F}}

84: \newcommand{\cI}{\mathcal{I}}

85: \newcommand{\cJ}{\mathcal{J}}

86: \newcommand{\cR}{\mathcal{R}}

87: \newcommand{\Ac}{\mathcal{A}}

88: \newcommand{\cV}{\mathcal{V}}

89: \newcommand{\Bc}{\mathcal{B}} 

90: \newcommand{\KK}{\mathcal{K}}

91: \newcommand{\MM}{\mathcal{M}}

92: \newcommand{\cM}{\mathcal M}

93: \newcommand{\cQ}{\mathcal Q}

94: \newcommand{\OO}{\mathcal{O}}

95: \newcommand{\Vw}{{\mathcal K}_w\vphantom{{\mathcal V}_w}}

96: \newcommand{\Gw}{{\mathcal G}_w\vphantom{{\mathcal G}_w}}

97: %\newcommand{\Vw}{\mathcal K}

98: %\newcommand{\Gw}{\mathcal G}

99: 

100: \newcommand{\IF}{\text{Im}\, \mathcal{F}}

101: \newcommand{\IM}{\text{Im}\, \mathcal{M}}

102: \newcommand{\RF}{\text{Re}\, \mathcal{F}}

103: \newcommand{\RM}{\text{Re}\, \mathcal{M}}

104: \newcommand{\I}{\text{Im}}

105: \newcommand{\R}{\text{Re}}

106: 

107: \newcommand{\Kcs}{K^{\text{cs}}}

108: \newcommand{\Kks}{K^{\text{ks}}}

109: 

110: \newcommand{\volume}{\text{{\small} vol}\, }

111: \newcommand{\ds}{\displaystyle}

112: 

113: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

114: % indices

115: \newcommand{\bi}{{\bar \imath}}

116: \newcommand{\ib}{{\bar\imath }}

117: \newcommand{\jb}{{\bar\jmath }}

118: \newcommand{\bj}{{\bar\jmath}}

119: \newcommand{\bk}{\bar{k}}

120: \newcommand{\bl}{\bar{l}}

121: \newcommand{\Kh}{{\hat{K}}}

122: \newcommand{\Lh}{{\hat{L}}}

123: \newcommand{\Ah}{{\hat{A}}}

124: \newcommand{\Bh}{{\hat{B}}}

125: \newcommand{\Ch}{{\hat{C}}}

126: \newcommand{\Dh}{{\hat{D}}}

127: \newcommand{\Mh}{{\hat{M}}}

128: \newcommand{\Nh}{{\hat{N}}}

129: \newcommand{\kh}{{\hat{k}}}

130: \newcommand{\ah}{{\hat{a}}}

131: \newcommand{\bh}{{\hat{b}}}

132: \newcommand{\ch}{{\hat{c}}}

133: \newcommand{\lh}{{\hat{l}}}

134: \newcommand{\mh}{{\hat{m}}}

135: \newcommand{\nh}{{\hat{n}}}

136: 

137: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

138: % Dan's macros

139: \DeclareMathOperator{\SU}{\mathit{SU}}

140: \DeclareMathOperator{\SO}{\mathit{SO}}

141: \DeclareMathOperator{\Symp}{\mathit{Sp}}

142: \DeclareMathOperator{\Spin}{\mathit{Spin}}

143: \DeclareMathOperator{\so}{\mathit{so}}

144: \DeclareMathOperator{\su}{\mathit{su}}

145: \DeclareMathOperator{\symp}{\mathit{sp}}

146: \DeclareMathOperator{\spin}{\mathit{spin}}

147: \DeclareMathOperator{\GL}{\mathit{GL}}

148: \DeclareMathOperator{\SL}{\mathit{SL}}

149: 

150: \DeclareMathOperator{\vol}{vol}

151: 

152: \newcommand{\rep}[1]{\mathbf{#1}}

153: 

154: \newcommand{\dd}{d}

155: \newcommand{\ii}{\mathrm{i}}

156: 

157: \newcommand{\bbZ}{\mathbb{Z}}

158: \newcommand{\bbR}{\mathbb{R}}

159: \newcommand{\bbC}{\mathbb{C}}

160: \newcommand{\bbP}{\mathbb{P}}

161: \newcommand{\bbF}{\mathbb{F}}

162: 

163: \newcommand{\bfO}{\mathbf{\Omega}}

164: \newcommand{\bfJ}{\mathbf{J}}

165: \newcommand{\bfC}{\mathbf{C}}

166: 

167: 

168: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

169: % misc

170: \newcommand{\CY}{Calabi--Yau}

171: \newcommand{\half}{\frac12}

172: \newcommand{\quart}{\frac14}

173: \newcommand{\nn}{\nonumber}

174: \def\Tnote#1{{\bf[TG: #1]}}

175: \def\Mnote#1{{\bf[MW: #1]}}

176: \newcommand{\RE}{\textrm{Re} \,}

177: \newcommand{\?}{{\bf [??]}}

178: \newcommand{\addref}{{\bf [add ref]}}

179: \newcommand{\cref}{{\bf [check ref]}}

180: \newcommand{\chec}{{\bf [check]}}

181: \newcommand{\park}{{\bf [the following text/formulas are just being parked here]}}

182: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

183: \newcommand{\M}{M}

184: \newcommand{\N}{\Theta}

185: \newcommand{\Weff}{W^{\rm (eff)}}

186: \newcommand{\Weffb}{\bar W^{\rm (eff)}}

187: \newcommand{\Y}{Y}

188: \newcommand{\G}{\mathcal{I}}

189: \newcommand{\CHI}{\mathcal{I}}

190: \newcommand{\f}{}

191: \newcommand{\Jc}{J_{\rm c}}

192: \newcommand{\Omegac}{\Omega_{\rm c}}

193: \newcommand{\cc}{c}

194: \newcommand{\CC}{C}

195: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

196: \newcommand{\hW}{\hat{W}}

197: \newcommand{\hK}{\hat{K}}

198: \newcommand{\hWb}{\hat{\bar W}}

199: \newcommand{\hphi}{{\phi}}

200: \newcommand{\Gt}{G^{(3)}\vphantom{G}} 

201: \newcommand{\Gtb}{\bar{G}^{(3)}\vphantom{\bar G}} 

202: \newcommand{\ha}{\hat{a}}

203: \newcommand{\hb}{\hat{b}}

204: \newcommand{\hab}{\hat{ a}}

205: \newcommand{\hbb}{\hat{ b}}

206: \newcommand{\cha}{\chi_{\hat{a}}}

207: \newcommand{\chab}{\bar{\chi}_{\hat{a}}}

208: \newcommand{\chb}{\chi_{\hat{b}}}

209: \newcommand{\chbb}{\bar{\chi}_{\hat{b}}}

210: \newcommand{\Dth}{\rm (D3)}

211: \newcommand{\eff}{\rm (eff)}

212: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

213: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

214: \newcommand{\rprop}{\oblong}

215: \newcommand{\tr}{\mathrm{Tr}\:}

216: \newcommand{\id}{\mathbf{1}}

217: \newcommand{\com}[2]{\big[ {#1},{#2} \big]}

218: \newcommand{\lie}[2]{\left[ {#1},{#2} \right]}

219: \newcommand{\ins}[1]{\mathrm{i}_{#1}}

220: \newcommand{\D}{\mathrm{D}}        

221: \newcommand{\Kw}{\mathcal{K}_w}    % Warped CY volume

222: \newcommand{\Em}{\varphi}          % Embedding map of the world-volume

223: \newcommand{\WV}{\mathcal{W}}      % Worldvolume

224: \newcommand{\FD}{F}

225: \newcommand{\FA}{F_\mathrm{A}}%\vphantom{F_\mathrm{A}}}  % Field strength - Brane

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

227: \newcommand{\Riem}[4]{R_{#1\hphantom{#2}#3#4}^{\hphantom{#1}#2}}

228: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

229: \newcommand{\simga}{\sigma}

230: 

231: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

232: %Bold-Math

233: \newcommand{\rhob}{\mbox{\boldmath$\rho$}}

234: 

235: 

236: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

237: %%%   caligraphic definitions        %%%%

238: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

239: \newcommand{\CA}{{\cal A}}

240: \newcommand{\CB}{{\cal B}}

241: %\newcommand{\CC}{{\cal C}}

242: \newcommand{\CE}{{\cal E}}

243: \newcommand{\CF}{{\cal F}}

244: \newcommand{\CG}{{\cal G}}

245: \newcommand{\CH}{{\cal H}}

246: \newcommand{\CI}{{\cal I}}

247: \newcommand{\CJ}{{\cal J}}

248: \newcommand{\CK}{{\cal K}}

249: \newcommand{\CL}{{\cal L}}

250: \newcommand{\CM}{{\cal M}}

251: \newcommand{\CN}{{\cal N}}

252: \newcommand{\CO}{{\cal O}}

253: \newcommand{\CP}{{\cal P}}

254: \newcommand{\CQ}{{\cal Q}}

255: \newcommand{\CR}{{\cal R}}

256: \newcommand{\CS}{{\cal S}}

257: \newcommand{\CT}{{\cal T}}

258: \newcommand{\CU}{{\cal U}}

259: \newcommand{\CV}{{\cal V}}

260: \newcommand{\CW}{{\cal W}}

261: 

262: 

263: 

264: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

265: %%% math symbols for Z,C,R... %%%

266: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

267: \def\IZ{{\mathbb Z}}

268: \def\IR{{\mathbb R}}

269: \def\IC{{\mathbb C}}

270: \def\IP{{\mathbb P}}

271: \def\IT{{\mathbb T}}

272: \def\IS{{\mathbb S}}

273: %%%%%%%%%%%%%%%%%%%%%%%%%%

274: %  others     %%%%%%%%%%%%%

275: %%%%%%%%%%%%%%%%%%%%%%%%%%%%

276: %\newcommand{\tr}{{\rm Tr}}

277: \newcommand{\re}{{\rm e}}

278: \newcommand{\ri}{{\rm i}}

279: \newcommand{\rd}{{\rm d}}

280: \newcommand{\Li}{{\rm Li}}

281: \psfrag{n1}{$\nu_1$}

282: \psfrag{n2}{$\nu_2$}

283: \psfrag{n1'}{$\nu_1'$}

284: \psfrag{n2'}{$\nu_2'$}

285: \psfrag{n9}{$\nu_9$}

286: \psfrag{n10'}{$\nu_{10}'$}

287: \psfrag{t1}{$\tilde{\nu}_1$}

288: \psfrag{t2}{$\tilde{\nu}_2$}

289: \psfrag{t9}{$\tilde{\nu}_9$}

290: \psfrag{t1'}{$\tilde{\nu}_1'$}

291: \psfrag{t2'}{$\tilde{\nu}_2'$}

292: \psfrag{t10'}{$\tilde{\nu}_{10}'$}

293: \newtheorem{lem14.1}{Lemma}

294: \newtheorem{cor14.2}[lem14.1]{Corollary}

295: \newtheorem{cor3}[lem14.1]{Corollary}

296: \newtheorem{cor4}[lem14.1]{Corollary}

297: \newtheorem{cor5}[lem14.1]{Corollary}

298: 

299: \newcommand{\tD}{t_D}

300: \newcommand{\mf}{\Psi}

301: \newcommand{\coeff}{c}

302: 

303: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

304: %%                              TABLEAUX.TEX

305: %%      This  macro file is for producing a ``Young Tableau'' which is

306: %%      an array of little squares sometimes used in mathematical physics.

307: %%      For instance, the command $\tableau{6 3 2}$ will produce a tableau

308: %%      with 6 squares in the top row, 3 in the next, and 2 in the last.

309: %%                                  OOOOOO

310: %%      This tableau will look like OOO    but made of squares instead of O's.

311: %%                                  OO

312: %%      Any number of rows may be present, each having a nonzero number of

313: %%      squares.

314: %%

315: %%      A tableau is math mode material, so use $ or $$ to enclose it.

316: %%

317: %%      The size and line-thickness of the little boxes are controlled by the

318: %%      dimension parameters --

319: %%              \tableauside=1.0ex              %(size)

320: %%              \tableaurule=0.4pt              %(line-thickness)

321: %%      Change them if you want.

322: %%

323: %%                                                      -- Doug Eardley 9/19/8%%

324: %%

325: \newdimen\tableauside\tableauside=1.0ex

326: \newdimen\tableaurule\tableaurule=0.4pt

327: \newdimen\tableaustep

328: \def\phantomhrule#1{\hbox{\vbox to0pt{\hrule height\tableaurule width#1\vss}}}

329: \def\phantomvrule#1{\vbox{\hbox to0pt{\vrule width\tableaurule height#1\hss}}}

330: \def\sqr{\vbox{%

331:   \phantomhrule\tableaustep

332:   \hbox{\phantomvrule\tableaustep\kern\tableaustep\phantomvrule\tableaustep}%

333:   \hbox{\vbox{\phantomhrule\tableauside}\kern-\tableaurule}}}

334: \def\squares#1{\hbox{\count0=#1\noindent\loop\sqr

335:   \advance\count0 by-1 \ifnum\count0>0\repeat}}

336: \def\tableau#1{\vcenter{\offinterlineskip

337:   \tableaustep=\tableauside\advance\tableaustep by-\tableaurule

338:   \kern\normallineskip\hbox

339:     {\kern\normallineskip\vbox

340:       {\gettableau#1 0 }%

341:      \kern\normallineskip\kern\tableaurule}%

342:   \kern\normallineskip\kern\tableaurule}}

343: \def\gettableau#1{\ifnum#1=0\let\next=\null\else

344: \squares{#1}\let\next=\gettableau\fi\next}

345: 

346: \tableauside=1.0ex

347: \tableaurule=0.4pt

348: \def\IE{\mathbb{E}}

349: 

350: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

351: \newcommand{\onefigure}[2]{\begin{figure}[htbp]

352:          \caption{\small #2\label{#1}(#1)}

353:          \end{figure}}

354: \newcommand{\onefigurenocap}[1]{\begin{figure}[h]

355:          \begin{center}\leavevmode\epsfbox{#1.eps}\end{center}

356:          \end{figure}}

357: \renewcommand{\onefigure}[2]{\begin{figure}[htbp]

358:          \begin{center}\leavevmode\epsfbox{#1.eps}\end{center}

359:          \caption{\small #2\label{#1}}

360:          \end{figure}}

361: \newcommand{\figref}[1]{Fig.~\protect\ref{#1}}

362: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

363: 

364: \def\blfootnote{\xdef\@thefnmark{}\@footnotetext} 

365: \long\def\symbolfootnote[#1]#2{\begingroup%

366: \def\thefootnote{\fnsymbol{footnote}}\footnote[#1]{#2}\endgroup}

367: 

368: \begin{document}

369: 

370: 

371: \begin{titlepage}

372: 

373: \hfill\vbox{

374: \hbox{CERN-PH-TH/2007-039}

375: \hbox{MAD-TH-07-04}

376: }

377: 

378: \vspace*{ 2cm}

379: 

380: \centerline{\Large \bf  Direct Integration of the Topological String} 

381: 

382: \medskip

383: 

384: \vspace*{4.0ex}

385: 

386: 

387: 

388: \centerline{\large \rm

389: Thomas W.~Grimm$^a$,\ Albrecht Klemm$^a$,\ Marcos Mari\~no$^b$ and Marlene Weiss$^{b,c}$\symbolfootnote[0]{\tt \begin{tabular}{ll}$^a$ &grimm@physics.wisc.edu,\ aklemm@physics.wisc.edu\\[.1cm] 

390:                      $^b$ &marcos@mail.cern.ch,\  marlene.weiss@cern.ch\end{tabular}}} 

391: 

392: 

393: \vspace*{4.0ex}

394: \begin{center}

395: {\em $^a$\ \ Department of Physics, University of Wisconsin, \\[.1cm]

396:         Madison, WI 53706, USA}

397:  

398: 

399: \vspace*{1.8ex}

400: 

401: 

402: {\em $^b$ Department of Physics, CERN\\[.1cm]

403: Geneva 23, CH-1211 Switzerland}

404: 

405: \vspace*{1.8ex}

406: 

407: {\em $^c$ Institut f\"ur Theoretische Physik, ETH H\"onggerberg\\[.1cm]

408: CH-8093 Z\"urich, Switzerland}

409: 

410: 

411: 

412: 

413: \vskip 0.5cm

414: \end{center}

415: 

416: \centerline{\bf Abstract}

417: \medskip

418: We present a new method to solve the holomorphic anomaly equations governing the free energies of type B topological 

419: strings. The method is based on direct integration with respect to the non--holomorphic dependence of the amplitudes, 

420: and relies on the interplay between non--holomorphicity and modularity properties 

421: of the topological string amplitudes. We develop a formalism valid for any Calabi--Yau manifold and we study 

422: in detail two examples, providing closed expressions for the amplitudes at low genus, as well as a discussion of 

423: the boundary conditions that fix the holomorphic ambiguity. The first example is the non-compact Calabi--Yau underlying 

424: Seiberg--Witten theory and its gravitational corrections. The second example is the Enriques Calabi--Yau, which we 

425: solve in full generality up to genus six. We discuss various aspects of this model: we obtain a new method to generate 

426: holomorphic automorphic forms on the Enriques moduli space, we write down a new product formula for the fiber amplitudes 

427: at all genus, and we analyze in detail the field theory limit. This allows us to uncover the modularity properties of 

428: $SU(2)$, $\CN=2$ super Yang--Mills theory with four massless hypermultiplets. 

429: 

430:                  

431: 

432: \vskip 1cm

433: 

434: 

435: 

436: 

437: 

438: \noindent February 2007

439: \end{titlepage}

440: 

441: 

442: 

443: \tableofcontents

444: 

445: 

446: \section{Introduction}

447: Topological string theory has played an important role in the quest for a 

448: better understanding of both physical and mathematical aspects of string theory. 

449: There are two different topological string theories related to each other by mirror symmetry, 

450: and known as the A and B-model. 

451: They are obtained from an $\CN=2$ superconformal field theory, twisted in two distinct 

452: ways to become type A or type B topological sigma models that are then coupled to gravity.  

453: The physical relevance of these theories lies in their intimate connection to type II superstring theory. 

454: In particular, the topological string on a given Calabi-Yau manifold computes higher derivative F-terms 

455: in the 4d effective action of the corresponding type II theory. 

456: {}From a mathematical point of view, the topological string partition function provides a generating functional for Gromov-Witten 

457: invariants in enumerative geometry. 

458: 

459: It is therefore desirable to solve the topological string on a given Calabi-Yau 

460: manifold, that is to say, to compute all the topological amplitudes $F^{(g)}$ in the genus expansion of the partition 

461: function. While this problem is completely solved for the case of non-compact toric Calabi-Yau manifolds thanks 

462: to the techniques of localization and the topological vertex, it remains a challenge for the compact case. 

463: One of the main tools in solving topological string theory, which also applies to compact Calabi-Yau manifolds, 

464: is the holomorphic anomaly equations for the B-model found in \cite{bcov}. In this work we present a new approach

465: to solving these equations. We make use of the fact that for each Calabi-Yau manifold there exists a target space 

466: symmetry group which provides a symmetry of the topological partition function \cite{abk} and thereby drastically

467: reduces the space of candidate solutions. The topological string amplitudes 

468: $F^{(g)}$ turn out to be polynomials in a finite set of generators which transform 

469: in a particularly simple way under the space-time symmetry group. 

470: Moreover, it can be shown that all non-holomorphic dependence in these amplitudes arises 

471: through a very special set of generators that are suitable generalizations of the 

472: non-holomorphic Eisenstein function $E_2(\tau, \bar \tau)$. The remaining generators are holomorphic. 

473: Keeping track of these non-holomorphic contributions we will be able to directly integrate the 

474: holomorphic anomaly equations. This method turns out to be very efficient and gives 

475: us  rich new information about the remaining holomorphic generators. A similar approach to the 

476: holomorphic anomaly equations was sketched in \cite{bcov}, in the analysis of toroidal orbifolds. 

477: For the quintic Calabi-Yau manifold a more complicated method was outlined in~\cite{Yamaguchi:2004bt}.

478: Other related approaches have been used before in \cite{hosono, hosonorev} to analyze rational elliptic 

479: surfaces, and in \cite{mnw,mnvw} to study noncritical strings and $\CN=4$ super Yang--Mills theory. 

480: 

481: 

482: The direct integration of the holomorphic anomaly equations can be performed for 

483: a generic Calabi-Yau manifold, as we will show in the final section of this 

484: work. However, in order to fully exploit the interplay of the holomorphic 

485: anomaly with the space-time symmetry, we will intensively discuss specific examples. 

486: To illustrate the general ideas we first study the local Calabi-Yau manifold 

487: associated to the Seiberg-Witten curve. Here the target-space symmetry group is 

488: a subgroup of $Sl(2,\bbZ)$ and the generating modular functions are well-known. 

489: 

490: Applying these methods to a compact Calabi-Yau manifold is far more 

491: involved. In the main part of the paper we will focus on the specific example of the Enriques Calabi-Yau \cite{fhsv}, 

492: arguably the simplest 

493: Calabi-Yau compactification with nontrivial topological string amplitudes \cite{km,mp}. 

494: This manifold can be obtained as the free quotient $({\rm K3}\times \mathbb{T}^2)/\bbZ_2$, where $\bbZ_2$ acts as 

495: the Enriques involution on the K3 fibers. The  

496: target space duality group of the Enriques Calabi-Yau is shown to be the discrete group $Sl(2,\bbZ)\times O(10,2,\bbZ)$,

497: with the factors corresponding to the $\mathbb{T}^2$ base and Enriques fiber, respectively. The generating modular forms 

498: for $Sl(2,\bbZ)$ are well-known, therefore 

499: we will be particularly concerned with the contributions from the Enriques fiber

500: and specially their mixing with the $\mathbb{T}^2$ base. 

501: 

502: After integrating the holomorphic anomaly equations the only problem remaining is to fix  

503: the holomorphic ambiguities, i.e.~the boundary conditions in the integration of the 

504: equations. These ambiguities are constrained by information coming from boundaries of the moduli space where the $F^{(g)}$ are known 

505: explicitly. In the Enriques case one can use the fiber limit, where all amplitudes can be determined by heterotic-type II duality \cite{km}, 

506: and a field theory limit where the manifold degenerates to give rise to $SU(2)$, $N_f=4$ Seiberg-Witten theory. By making use

507: of these boundary conditions we determine the full topological string amplitudes up to genus $6$, improving in this 

508: way previous results in \cite{km}. 

509: As a bonus of our analysis, we clarify the modularity properties of the 

510: conformal $N_f=4$ theory and its gravitational corrections described in \cite{Nek}. At present the available boundary conditions 

511: are not enough to completely solve topological string theory on the Enriques Calabi--Yau, but 

512: we provide efficient tools to 

513: analyze the amplitudes at all genus with the method of direct integration.

514: 

515: 

516: The organization of this paper is as follows. In section \ref{anomaly} we review the derivation 

517: of the holomorphic anomaly equations. Section \ref{Seiberg-Witten} gives a first simple example 

518: of the method of direct integration and the fixing of holomorphic ambiguities by application to 

519: Seiberg-Witten theory. Section \ref{sec:Enriques} reviews what will be our main focus, the Enriques Calabi-Yau. 

520: We introduce modular and automorphic forms which will be relevant later and discuss the topological amplitudes

521: on the Enriques fiber. Also an all-genus product formula for the fiber partition function will 

522: be introduced. 

523: Section \ref{sec:diE} constitutes the core of this work. We show explicitly how one can solve for $F^{(g)}$

524:  up to genus six and present the general recursive formalism. Furthermore, boundary conditions and a 

525:  reduced Enriques model where part of the moduli space is blown down are investigated. 

526:  In section \ref{sec:ftlim} we analyze the field theory limit corresponding to $N_f=4$ SYM and we relate it in 

527: detail to the Enriques 

528:  Calabi--Yau. In section \ref{generic} 

529:  we present a formalism for direct integration on generic Calabi-Yau manifolds. Section \ref{conclusion} contains 

530:  conclusions and an outlook on further directions of investigation. Appendix \ref{N=2sp} reviews some special geometry. 

531:  Appendix \ref{theta} collects some useful formulae for theta functions and modular forms. 

532:  Appendix \ref{heteroticFg} reviews the heterotic computation of the amplitudes 

533: in \cite{mm,km} and presents improved formulae for their antiholomorphic dependence. Finally, appendix \ref{Cal_big_Fg} presents the holomorphic anomaly equations on the so-called big moduli space. 

534: 

535: 

536: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

537: %%%%%%%%

538: %%%%%%%%     Section 2

539: %%%%%%%%

540: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

541: 

542: \section{The holomorphic anomaly equations \label{anomaly}}

543: 

544: In this section we will briefly recall some basics about topological 

545: string theory to set the stage for the following sections 

546: and to fix our conventions. This will force us to introduce some world-sheet notations and techniques.

547: However, for the rest of this work we will mostly need only the explicit form of the 

548: holomorphic anomaly equations. For a more detailed 

549: introduction to topological string theory the reader might want to consult references \cite{mirror,vonk,nv,mbook, klemm}.

550: 

551: Type II string theory on Calabi-Yau threefold $Y$ yields a  

552: superconformal field theory  with left and right moving $(2,2)$

553: supersymmetry on the world-sheet. This structure admits two

554: topological string theories: the A--and the B--model. 

555: The key quantity in these topological theories is their all genus partition function 

556: %

557: \beq \label{part-function}

558:     Z=\text{exp}\ \sum^\infty_{g=0} g_s^{2g -2} F^{(g)} \ .

559: \eeq

560: %

561: This formal expansion in the string coupling $g_s$  contains the 

562: topological string amplitudes $F^{(g)}$ for maps from genus $g$ Riemann surfaces 

563: into a target Calabi-Yau manifold. 

564: The topological string amplitudes of the A-- and B--model are identified by mirror symmetry, which 

565: maps one theory on $Y$ to its dual on the corresponding

566: mirror Calabi-Yau. 

567: 

568: We will now briefly recall the B--model definitions of the free energies 

569: $F^{(g)}$. The B--model describes constant maps from a world-sheet 

570: Riemann surface $\Sigma_g$ to points in the Calabi-Yau space $Y$. Therefore,

571: the B--model definition of the $F^{(g)}$ involves only the integration over the

572: moduli space ${\overline{\cM}_g}$ of the world-sheet  

573: and not over the moduli space of maps.

574: More precisely, let us denote by  $({\underline m},{\overline {\underline m}})$ 

575: coordinates on ${\overline{\cM}_g}$ and abbreviate the correlators of the world-sheet CFT by 

576: $\big<\cdot\big>_g$. The free energies $F^{(g)}$ are then defined by

577: %

578: \begin{equation}

579:           F^{(g)}=\langle 1\rangle_g=\int_{\overline{\cM}_g}\ \langle \prod_{k=1}^{3g-3} \beta^k {\bar \beta}^k \rangle_g \ \,

580:           [\dd {\underline m} \wedge \dd {\overline {\underline m}}]\ .

581: \end{equation} 

582: %

583: Here we inserted the operators $\beta^k=\int_{\Sigma_g} G^- \mu^k$ and their complex conjugates to 

584: obtain the correct measure on the moduli space. $\beta^k$ and ${\bar \beta}^k$ contain the 

585: the world-sheet Beltrami differentials

586: $\mu^k \in H^1(T\Sigma_g)$ and the world-sheet supersymmetry generators $G^-,\bar G^-$.

587: The contraction of $[\dd {\underline m} \wedge \dd {\overline {\underline m}}]$ with 

588: the $(\beta^k,\bar \beta^k)$ factor is

589: antisymmetric due to the presence of $G^-,\bar G^-$ and  

590: yields a top form on the complex $3g-3$ dimensional moduli space ${\overline{\cM}_g}$. 

591: The fact that one has to integrate only over the moduli space of the world-sheet

592: makes the B-model far simpler to solve than the A-model. Therefore, it is often easier to use the B-model 

593: and the mirror map to determine A-model quantities.

594:  

595: 

596: {}From the point of view of the four-dimensional effective action, one is interested in the dependence 

597: of the $F^{(g)}$ on the

598: complex moduli  $t^i,\bar t^i$ in the vector multiplets. These parametrize marginal 

599: deformations, which in the B-model correspond to complex structure deformation of the Calabi-Yau manifold.

600: Infinitesimally  the world-sheet action is  

601: perturbed by the $t^i,\bar t^i$ as follows 

602: \begin{equation}   

603: S=S_0+ 

604:  t^i\ \int_{\Sigma_g} \cO_i^{(2)}+

605:  {\bar t}^{i}\ \int_{\Sigma_g} 

606: \bar \cO_{i}^{(2)}\ ,

607: \end{equation}     

608: where the sums run over $i=1,\ldots,h^1(Y,TY)=h^{(2,1)}(Y)$.

609: Here the marginal two-form operators are obtained using the descent 

610: equations as

611: % 

612: \begin{equation} 

613: \label{descent}

614: {\cal O}_i^{(2)}= \{G_0^-,[\bar G_0^-,{\cal O}_i^{(0)}]\} \rd z \rd {\bar z}\ , \qquad  

615: \overline {{\cal O}^{(2)}_{\bar \imath}}=\{G_0^+,[\bar G_0^+,\bar {\cal O}^{(0)}_{\bar \imath}]\}

616: \rd z \rd {\bar z}\ ,

617: \end{equation}

618: %

619: where $G_0^+,G_0^-$ are the zero modes of the twisted world-sheet supersymmetries $G^+,G^-$.

620: In these equations we denoted by ${\cal O}_i^{(0)}$ the zero-form cohomological operators, which are in one-to-one 

621: correspondence with the $H^1(Y,TY)$ cohomology of the target space. 

622: 

623: {}From the point of view of the target space Calabi-Yau

624: the complex fields $t^i,\bar t^i$ provide a set of local coordinates on the moduli 

625: space of complex structure deformations $\cM$. This space is shown to 

626: be a special K\"ahler manifold with K\"ahler potential 

627: \beq

628:    K(t,\bar t)=-\log i\int_Y \Omega(t) \wedge \bar \Omega(\bar t)\ ,

629: \label{kaehlerpotential}

630: \eeq

631: where $\Omega(t)$ is the holomorphic three-form on $Y$ varying holomorphically with a 

632: change of the complex structure. $\Omega(t)$ is only unique up to rescalings by a 

633: holomorphic function and hence should be viewed as a section of the line bundle $\cL$ over 

634: the moduli space $\cM$. 

635: In appendix \ref{N=2sp} we review how the special geometry of $\cM$ can 

636: be entirely encoded by a single holomorphic section of ${\cal L}^{2}$, 

637: the prepotential $F^{(0)}=\cF(t)$. {}From a world-sheet point of view one 

638: does not obtain $F^{(0)}$ directly, but rather finds the three-point function

639: \beq

640:    C^{(0)}_{ijk} =\langle \cO_{i}^{(0)}\cO_{j}^{(0)}\cO_{k}^{(0)} \rangle_g=- \int_Y \Omega(t) \wedge \partial_i \partial_j \partial_k \Omega(t)\ ,

641: \eeq

642: where $\partial_i$ are derivatives with respect to $t^i$.

643: 

644:  

645: At higher genus a more involved world-sheet analysis can be applied to investigate the 

646: properties of the higher $F^{(g)}$. It turns out that the  higher genus topological 

647: string amplitudes $F^{(g)}$ are not holomorphic, but rather fulfill  specific 

648: holomorphic anomaly equations. These equations are recursive in the genus  and 

649: determine the anti-holomorphic derivative of $F^{(g)}$.  Therefore, even if the genus zero 

650: data are given they determine $F^{(g)}$ only up to a holomorphic ambiguity. 

651: We will now briefly state the essential features and results of the work of

652: Bershadsky, Cecotti, Ooguri and Vafa \cite{bcov}, who have shown that 

653: 

654: \noindent

655: {\sl i}.) The $F^{(g)}$ transform as section of ${\cal L}^{2-2g}$ with the connection (\ref{cov_D}).

656: 

657: \noindent

658: {\sl ii}.) The topological B-model correlation functions 

659: \beq 

660: \label{C_prop}

661:   C^{(g)}_{i_1 \ldots i_n}=

662: \left\{ \begin{array}{ll}

663: \langle \int_{\Sigma_g}  \cO_{i_1}^{(2)} \cdots  \int_{\Sigma_g}  \cO_{i_n}^{(2)}\rangle_g= 

664: D_{i_1}\ldots D_{i_n} F^{(g)} &\ \text{for}\ \ g\ge1\\ [3 mm]

665: \langle \cO_{i_1}^{(0)}\cO_{i_2}^{(0)}\cO_{i_3}^{(0)}  \int_{\Sigma_g} 

666: \cO_{i_{4}}^{(2)} \cdots  \int_{\Sigma_g}  \cO_{i_n}^{(2)}\rangle_g=

667: D_{i_{4}}\ldots D_{i_n} C^{(0)}_{i_1 i_2 i_3} &\ \text{for}\ \ g=0 \end{array}\right.

668: \eeq

669: can be obtained using the covariant derivatives \eqref{cov_D} and obey

670: \beq \label{C_prop_cond}

671:    C^{(g)}_{i_1 \ldots i_n} = 0 \quad \text{for} \ \ 2g-2 + n \le 0 \ .

672: \eeq

673: 

674: \noindent

675: {\sl iii}.) The anti-holomorphic derivative  $\partial_{\bar \imath}=

676: \frac{\partial}{\partial {\bar t^{i}}}$ of the $F^{(g)}$,

677: \begin{equation} 

678: \bar \partial_{\bar \imath} F^{(g)}=

679: \int_{\overline{\cM}_g} \bar \partial_{\bar \imath} \mu_g=

680: \int_{\overline{\cM}_g} \partial_m \bar \partial_{\bar m} \lambda_{\bar \imath , g}=

681: \int_{\partial \overline{\cM}_g} \lambda_{\bar \imath, g},

682: \end{equation}  

683: receives only contributions from the complex codimension one locus in the moduli space of Riemann surfaces corresponding

684:  to world-sheets which are degenerate with lower genus components.  These boundary 

685: contributions can be worked out and yield recursive equations for the $F^{(g)}$. For $g>1$ one gets

686: \beq \label{rec_Fg}

687:    \bar \partial_{\bar \imath} F^{(g)} = \tfrac{1}{2} \bar C^{(0)jk}_{\bar \imath} 

688:    \Big(D_j D_k F^{(g-1)} + \sum_{r=1}^{g-1}D_j F^{(r)} D_k F^{(g-r)} \Big)  

689: \eeq 

690: and for $g=1$ a generalisation of the Quillen anomaly

691: \beq \label{anomaly_F1}

692:    \partial_{i} \bar \partial_{\bar \jmath} F^{(1)} = \tfrac{1}{2}     C^{(0)}_{i k l } \bar C^{(0)kl}_{\bar \jmath}  - 

693: \Big(\frac{\chi}{24} -1 \Big)G_{i\bar \jmath}\ .

694: \eeq

695: Here we defined 

696: \be

697: \bar C^{(0)kl}_{\bar \jmath}= e^{2K} G^{k\bar k} 

698: G^{l \bar l} \bar C^{(0)}_{\bar \jmath \bar k \bar l}\ ,

699: \eeq

700: where $G_{k \bar k}=\partial_k \bar \partial_{\bar k} K$ is the Weil-Petersson metric of the K\"ahler potential (\ref{kaehlerpotential}).   

701: 

702: These are the recursive holomorphic anomaly equations, which we want to integrate directly  

703: in this paper. Note that there is no holomorphic anomaly at genus zero. 

704: $C^{(0)}_{ijk}$ has no world-sheet moduli dependence, 

705: hence no boundaries, and is therefore holomorphic. 

706: The genus zero data thus have to be provided from the outset. 

707: They can be determined from the 

708: period integrals of the manifold  $Y$.    

709: 

710: It is further shown in ref.~\cite{bcov} that (\ref{rec_Fg}) can be integrated recursively. 

711: With an iterative procedure of complexity growing exponentially with the genus, 

712: one rewrites (\ref{rec_Fg}) as 

713: \begin{equation}

714: \label{feynman1}

715: { \partial}_{\bar k} F^{(g)}(t,\bar t)=\bar \partial_{\bar k} 

716: \Gamma^{(g)}( \hat \Delta^{ij}, \hat \Delta^{i}, \hat \Delta, C^{(r<g)}_{i_1 \ldots i_n}) \ ,

717: \end{equation}

718: and integrates it to 

719: \begin{equation}

720: \label{Fgwithf}

721: F^{(g)}(t, \bar t)= \Gamma^{(g)}( \hat \Delta^{ij}, \hat \Delta^{i}, \hat \Delta, C^{(r<g)}_{i_1 \ldots i_n})+f^{(g)}(t)\ . 

722: \end{equation}

723: Here $\Gamma^{(g)}$ is a functional of some propagators $\hat \Delta^{ij}, \hat \Delta^{i}, \hat \Delta$ 

724: and the lower genus vertices $C^{(r)}_{i_1\ldots i_n}$ with $r<g$.

725: The holomorphic ambiguity $f^{(g)}(t)$  arises as an integration constant.

726: To prove that the functional $\Gamma^{(g)}$ exists at every genus,~\cite{bcov} show that it is 

727: the disconnected Feynman graph expansion of an auxiliary action with the above vertices and propagators, 

728: whose partition function fulfills a master equation equivalent to (\ref{rec_Fg}) and (\ref{anomaly_F1}).

729: The propagators can be defined using the genus zero data as follows. Since

730: \begin{equation} 

731: \bar D^{\phantom{0}}_{\bar \imath} \bar C^{(0)}_{\bar \jmath \bar k \bar l}=\bar D^{\phantom{0}}_{\bar \jmath} \bar C^{(0)}_{\bar 

732: \imath \bar k \bar l}

733: \end{equation}

734: one can integrate 

735: \begin{equation}

736: \bar C^{(0)}_{\bar \jmath \bar k \bar l}=-\tfrac12 e^{-2 K} \bar D_{\bar \imath}\bar D_{\bar \jmath}

737: \bar \partial_{\bar k} \hat \Delta

738: \end{equation}

739: as

740: \begin{equation}\label{def-small-Delta}

741: G_{\bi j} \hat \Delta^{j} = \tfrac12 \bar \partial_\bi \hat \Delta\ ,\qquad \quad G_{\bi k}\hat  \Delta^{kj}= 

742: \bar \partial_{\bi} \hat \Delta^j\ ,\qquad \quad \bar C^{(0) jk}_{\bar \imath}=  \bar \partial_{\bi} \hat \Delta^{jk}\ .

743: \end{equation} 

744: Note that the propagators are defined by these equations only up to holomorphic 

745: ambiguities arising in the integration steps. Fixing these ambiguities directly affects 

746: the definition of the holomorphic functions $f^{(g)}(t)$ in \eqref{Fgwithf}. 

747: It turns out that a preferred choice for this ambiguity 

748: is provided by relating the propagators in a canonical way to $F^{(1)}(t,\bar t)$ \cite{abk}.

749: 

750: The combinatorics of the Feynman graph expansion are useful to establish 

751: some general properties of the $F^{(g)}$, but its complexity 

752: grows exponentially with the genus.  However, the $F^{(g)}$ are 

753: invariant under space-time modular transformations which are 

754: a symmetry of the full string compactification. As we will discuss 

755: later, they generically admit a split into a universal factor times a modular 

756: form. Here the weights of the modular  

757: forms grow linearly with the genus. Since the ring of modular forms 

758: is finitely generated, the complexity of modular invariant expressions

759: grows only polynomially with the genus. The method of direct 

760: integration that we develop in this paper uses this connection 

761: with modular forms such that its complexity also 

762: grows only polynomially with the genus. It has the advantage that 

763: the modular properties of the amplitudes are manifest 

764: in all steps of the derivation.   

765:   

766: 

767: 

768: 

769: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

770: %%%%%%%%

771: %%%%%%%%     Section 3

772: %%%%%%%%

773: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

774: 

775: 

776: \section{Solving Seiberg-Witten theory by direct integration \label{Seiberg-Witten}}   

777:        

778: 

779: Local Calabi-Yau geometries provide simple and instructive examples for the interplay

780: between holomorphicity and modular invariance in topological string theory. 

781: In this section we will explain the key features using the simplest example, namely 

782: the local Calabi-Yau corresponding to $SU(2)$ Seiberg-Witten theory with no matter \cite{sw1}. 

783: In section \ref{SWgeometry} we first recall the geometry of Seiberg-Witten theory.

784: We show that all genus zero data can be expressed in terms of a finite set of holomorphic 

785: modular forms. All higher amplitudes $F^{(g)}$ are invariant under the modular 

786: group. In section \ref{SWintegration}

787: we directly integrate the holomorphic anomaly equations, determining 

788: all $F^{(g)}$ up to a holomorphic modular ambiguity. 

789: Modularity restricts this ambiguity so much 

790: that simple boundary conditions set by the effective action near 

791: special points in the moduli space allow one to reconstruct all $F^{(g)}$. We review 

792: such a convenient  set of boundary conditions in section \ref{SWboundary}.

793: The general philosophy presented in this section will be later applied to

794: the more complicated case of compact Calabi-Yau manifolds.

795: 

796: \subsection{The Seiberg-Witten geometry \label{SWgeometry}} 

797: 

798: 

799: Seiberg-Witten theory  with no matter \cite{sw1} can be obtained in the A--model as a limit of the local 

800: Calabi-Yau geometry 

801: ${\cal O}(-2,-2)\rightarrow \mathbb{P}^1\times \mathbb{P}^1$ \cite{kkv}. The mirror 

802: B--model geometry of this limit is the Seiberg-Witten elliptic curve ${\cal E}$

803: %

804: \begin{equation} 

805: y^2=(x-u)(x-\Lambda^2)(x+\Lambda^2)\ ,

806: \end{equation}

807: % 

808: whose modular group is $\Gamma(2)$. This subgroup of $ Sl(2,\bbZ)$ 

809: acts on the period integrals

810: \beq \label{def-ttD}

811:   t=\int_{a} \lambda\ ,\qquad \qquad t_D=\int_{b} \lambda\ ,  

812: \eeq

813: where $\lambda=\frac{\sqrt{2}}{2 \pi }\frac{y}{x^2-1} {\rm d}x$

814: is the Seiberg-Witten meromorphic differential. In the limit described above, $\lambda$ is 

815: obtained as a reduction of the holomorphic $(3,0)$ form of the Calabi-Yau manifold. 

816: Rigid special geometry guarantees 

817: the existence of a prepotential $F^{(0)}={\cal F}(t)$ with the properties 

818: \begin{equation} 

819: t_D=\frac{\partial {\cal F}}{\partial t}\ ,\qquad \qquad 

820: \tau = -\frac{1}{4 \pi} \frac{\partial^2 {\cal F}}{\partial^2 t} \ .

821: \end{equation}  

822: These conditions are obtained as the rigid limit of the

823: special geometry relations presented in Appendix \ref{N=2sp}. 

824: Note that $\tau$ is precisely the complex structure parameter 

825: of the torus and hence parametrizes the 

826: upper half-plane. In particular, $\I \tau >0$ 

827: is guaranteed by the Riemann inequality consistent with the fact that 

828: $\I \tau$ is the gauge kinetic coupling function of Seiberg-Witten theory.

829: Moreover, a modular transformation acts on $\tau$ as

830: %

831: \beq \label{modulartrans}

832:    \tau \ \mapsto \ \frac{a\tau + b}{c\tau + d}\ .

833: \eeq

834: The genus zero data are functions of $\tau$ and transform 

835: in a particularly simple way under \eqref{modulartrans}. They 

836: can be expressed in terms of a finite set of modular generators,

837: which we will specify in the following.  

838: 

839: A modular function $f(\tau)$ of weight $m$ is defined 

840: to transform as $f(\tau) \, \mapsto \, (c \tau+d)^m f(\tau)$ under 

841: \eqref{modulartrans}. Focusing on the modular group of the Seiberg-Witten 

842: curve, we note that 

843: the ring of modular functions of $\Gamma(2)$ can be expressed as 

844: powers of the Jacobi $\theta$-functions. Relevant properties of  the 

845: Jacobian $\theta$-functions are summarized in Appendix B. We introduce 

846: two generators 

847: \begin{equation} 

848: K_2=\vartheta_3^4+\vartheta_4^4, \qquad  \qquad  K_4=\vartheta_2^8\ , % \Delta = \theta_3^4 \theta_4^4 \ , 

849: \label{gamma2generators} 

850: \end{equation} 

851: which are of modular weight two and four respectively. 

852: The modular transformation properties follow

853: from (\ref{thetatransformation}).

854: $K_2, K_4$ generate the graded ring of  holomorphic modular forms 

855: ${\cal M}_*(\Gamma(2))$ of $\Gamma(2)$, which we will also denote by 

856: $\mathbb{C}[K_2, K_4]$. 

857: It turns out to be useful to also introduce

858: \beq \label{ring_hE4}

859:   h=K_2\ ,\qquad \qquad  E_4=\tfrac{1}{4} (K_2^2 + 3 K_4)\ .

860: \eeq

861: As we will see when we develop the method of direct integration, it is natural to take 

862: $h$, $E_4$ as the generators of the ring ${\cal M}_*(\Gamma(2))$.

863: 

864: Let us now express the genus zero data in terms of modular forms.

865: The connection with the geometry of the Seiberg-Witten curve is given by the following relation 

866: \begin{equation}

867: u(\tau)=\frac{K_2}{\sqrt{K_4}}\ . 

868: \label{ut}

869: \end{equation}

870: The combination $z(\tau)=1/u^2(\tau)$ is modular invariant and can be viewed as the analog of the 

871: mirror map for this non-compact Calabi-Yau manifold.

872: The analog of the holomorphic triple coupling is 

873: \begin{equation}

874: \label{xisw}

875:  C \equiv C^{(0)}_{ttt}=

876: \frac{\partial \tau}{\partial t}=\frac{32 K_4^{1/ 4}}{K_2^2-K_4}\ 

877: \end{equation}

878: Note that $C^2$ is a form of weight $-6$ 

879: under the modular transformations in  $\Gamma(2)$.

880: The modular group $\Gamma(2)$ also

881: determines the periods $t,t_D$ as weight $1$ objects \footnote{They 

882: can be calculated likewise using the Picard-Fuchs equation.}

883: \begin{equation}

884: t(\tau)={E_2(\tau)+ K_2(\tau)\over 3 K_4^{1/4}(\tau)}, \qquad \quad 

885: t_D(\tau_D)=-i{ 2 E_2(\tau_D)- K_2 (\tau_D)- 3K_4^{1/2}(\tau_D) \over 3 \big(2 K_2(\tau_D)- 2 K_4^{1/2}(\tau_D) \big)^{1/2}}\ ,

886: \label{t-td-period}

887: \end{equation}

888: where $\tau_D=-\frac{1}{\tau}$ and $E_2$ is the second Eisenstein series 

889: defined in \eqref{geneis}. It is natural to give the periods in the above parameters. 

890: In the electric phase of Seiberg-Witten theory the $q=e^{2 \pi i \tau}$ series converges and

891: $t$ is the physical expansion parameter, while in the magnetic phase the $q_D=e^{2 \pi i \tau_D}$ 

892: series converges and $t_D$ is the physical expansion parameter. Of course $t_D(\tau)$ and 

893: $t(\tau_D)$ can be obtained by performing an $S$-duality transformation on $E_2$ and the Jacobi theta 

894: functions.               

895: 

896: 

897: 

898: 

899: \subsection{Direct integration \label{SWintegration}}

900: 

901: Having discussed the genus zero  geometry, let us now turn to the 

902: higher genus  free energies $F^{(g)}$ and their holomorphic anomaly. Starting with 

903: $F^{(1)}$, we note that the holomorphic anomaly equation (\ref{anomaly_F1}) specializes to 

904: %

905: \begin{equation} \label{SWF1anomaly}

906: \partial_t \partial_{\bar t} F^{(1)}=\tfrac{1}{2}  C^{(0)\, tt}_{\bar t} C^{(0)}_{ttt} \ .

907: \end{equation}

908: %

909: where the indices are raised with the Weil-Petersson metric $G_{t\bar t}=2 \I \tau$.

910: This equation integrates immediately to

911: %

912: \begin{equation} \label{SWF1withtau}

913: F^{(1)}= - \tfrac12 \log \,\I \tau - 

914:      \log |\Phi(\tau)|\ ,

915: \end{equation}

916: % 

917: where $\partial \tau/\partial t$ is evaluated using \eqref{xisw}.

918: The holomorphic object $\Phi(\tau)$ is the ambiguity at genus one. It is 

919: determined from modular constraints and the physical requirement that 

920: $F^{(1)}$ should only be singular at the discriminant of ${\cal E}$. 

921: Note that under a modular transformation \eqref{modulartrans} one 

922: finds that $\I \tau\ \mapsto\ {|c \tau+d|^{-2}}\I \tau$.

923: Together with the invariance of $F^{(1)}$ this implies that $\Phi(\tau)$ must be a 

924: modular form of weight $1$.  The only modular 

925: form of weight $1$ which has only poles at the discriminant of ${\cal E}$

926: is the square of the $\eta$ function given in \eqref{dede}. This fixes the ambiguity at genus one as $\Phi(\tau) = \eta^2 (\tau)$.

927: 

928: At genus one the non-holomorphic dependence was induced through the appearance of

929: $\I \tau$. As dictated by the holomorphic anomaly equations, 

930: all higher $F^{(g)}$ also depend on $\bar t$.

931: We now show that this dependence arises through the propagator $\hat \Delta^{tt}$ only.

932: $\hat \Delta^{tt}$ is obtained in the local limit of \eqref{def-small-Delta} and thus obeys

933: \beq \label{proploc}

934:     {\partial}_{\bar t} \hat \Delta^{tt}= C^{(0)\, tt}_{\bar t} \ .

935: \eeq

936: All other propagators vanish in this limit.

937: To integrate this condition, we first multiply both sides in \eqref{proploc} by 

938: $C^{(0)}_{ttt}$. The result is easily compared to the holomorphic 

939: anomaly equation \eqref{SWF1anomaly} of $F^{(1)}$. 

940: Changing derivatives 

941: by inserting ${\partial \tau}/{\partial t}=C^{(0)}_{ttt}$

942: one evaluates with the help of 

943:  (\ref{Eta-der})

944: \begin{equation}

945: \hat \Delta^{tt}=2 \partial_\tau F^{(1)}(\tau,\bar \tau)=-\tfrac{1}{12} \widehat E_2(\tau,\bar \tau)\ ,\qquad \qquad \partial_\tau = (2\pi i)^{-1} \tfrac{ \partial }{\partial \tau}

946: \label{swpropagator}

947: \end{equation} 

948: The occurrence of the non-holomorphic extension 

949: of the second Eisenstein series $E_2(\tau)$      

950: \begin{equation}

951: \widehat E_2(\tau,\bar \tau)= E_2(\tau)-{3 \over \pi  \I \tau} \ .

952: \label{hatE2}

953: \end{equation}

954: is forced by modular invariance. Since $F^{(1)}(\tau,\bar \tau)$ 

955: is a modular function of weight zero, its derivative must be a 

956: modular form of weight $2$ which is not holomorphic. 

957: The only form with these properties is the almost 

958: holomorphic form $ \widehat E_2(\tau,\bar \tau)$. This form is 

959: the canonical, almost holomorphic extension of the second Eisenstein series $E_2$, where $E_2$ is the unique holomorphic quasimodular form of weight 2  

960: transforming as

961: \begin{equation}

962: E_2(\tau) \quad \mapsto\quad (c\tau + d)^2 E_2(\tau) - \tfrac{6}{ \pi} i c (c \tau+d)\ 

963: \label{E2transformation} 

964: \end{equation} 

965: under a modular transformation \eqref{modulartrans}.

966: The shift in the transformation of the anholomorphic piece in (\ref{hatE2}) cancels 

967: precisely the shift in (\ref{E2transformation}). More generally the ring 

968: ${\hat {\cal M}}_*$ of almost holomorphic forms of $\Gamma(2)$ is generated 

969: as $\mathbb{C}[\widehat E_2,h,\Delta]$. 

970: 

971: 

972: 

973: Using  the propagator and general properties of the Feynman graph expansion 

974: one can extract the fact that the higher genus $F^{(g)}$ are weight 0 forms 

975: with the structure

976: %

977: \begin{equation}

978: F^{(g)}(\tau, \bar \tau)=C^{2g-2} \sum_{k=0}^{3g-3}

979: \widehat E_2^{k}(\tau,\bar \tau) c^{(g)}_k(\tau)\ ,\qquad g>1\ ,

980: \label{eq:generallocalform}

981: \end{equation}

982: where we defined $C=C_{ttt}^{(0)}$.

983: Modular invariance implies then that the holomorphic forms $c^{(g)}_k(\tau)$ are modular 

984:  of weight $6(g-1)-2k$ in $\mathbb{C}[h,\Delta]$. We will show 

985: next that all forms $c^{(g)}_k(\tau)$ with $k>0$ are very easily determined by 

986: direct integration of the holomorphic anomaly equation. The 

987: form $c^{(g)}_0(\tau)$ is not determined in this way and  corresponds to

988: a holomorphic modular ambiguity.        

989: 

990: In order to analyze the holomorphic anomaly equations in the local case, it turns out to be 

991: very useful to discuss some general properties related to modular transformations. Let us 

992: first discuss how derivatives transform under the modular transformation \eqref{modulartrans}.

993: Denoting by $f_k$ a modular form of weight $k$ it is elementary 

994: to check that its derivative transforms under \eqref{modulartrans} as

995: \begin{equation} \label{deriv_shift}

996: \partial_{\tau} f_k \quad \mapsto \quad (c\tau+d)^{k+2}\partial_\tau f_k +

997: \frac{k}{2 \pi i} c(c \tau +d)^{k+1} f_k \ .

998: \end{equation}

999: Similarly, we can evaluate $\partial_{t} f_k = C^{-1} \partial_{\tau} f_k$, where as above 

1000: $C=C^{(0)}_{ttt}$. In order to cancel the shift in \eqref{deriv_shift} 

1001: we will now introduce covariant derivatives. There are two possible 

1002: ways to achieve this\footnote{We thank Don Zagier for explaining us several manipulations involved in the following.}. 

1003: Firstly, one can cancel the shift against the shift of $(\I \tau)^{-1}$ and set

1004: \begin{equation}  \label{der_nonhol}

1005:   D_t f_k=\Big(\partial_t-{ k C \over 4 \pi \I \tau}\Big)f_k \ ,\qquad \quad D_\tau f_k=\Big(\partial_\tau-{ k\over 4 \pi \I \tau}\Big)f_k\ .

1006: \end{equation}

1007: Here $D_t$ is the covariant derivative to the Weil-Petersson metric $G_{t\bar t}$ and $D_\tau$ is the so-called Mass derivative.

1008: $D_t$ maps almost holomorphic forms of $\Gamma(2)$ of weight $k$ 

1009: into almost  holomorphic forms of weight $k-1$, while $D_\tau$ increases the weight

1010: from $k$ to $k+2$. Note that both derivatives in \eqref{der_nonhol} are non-holomorphic 

1011: due to the appearance of $\I \tau$.

1012: There is however  a second possibility to cancel the shift \eqref{deriv_shift} which is manifestly

1013: holomorphic. More precisely, one can cancel the shift against the shift \eqref{E2transformation} of $E_2(\tau)$ and

1014: define

1015: \beq \label{der_hol}

1016:   \hat D_t f_k = \big(\partial_t-\tfrac{1}{12}  k C E_2 \big) f_k\ ,\qquad \quad  \hat D_\tau f_k = \big(\partial_\tau-\tfrac{1}{12}  k E_2 \big) f_k\ .

1017: \eeq  

1018: In this case $\hat D_\tau$ is known as the Serre derivative.

1019: Both $\hat D_t$ and $\hat D_\tau$ are holomorphic. They map holomorphic 

1020: modular forms of weight $k$ to holomorphic modular forms of weight $k-1$ and $k+2$ respectively.

1021: It is easy to check that the following identity holds

1022: \beq \label{splitSW}

1023:     D_t f_k = \hat D_t f_k + \tfrac{1}{12}k C \widehat E_2\, f_k\ ,\qquad D_\tau f_k = \hat D_\tau f_k +\tfrac{1}{12}k  

1024: \widehat E_2\, f_k\ .

1025: \eeq

1026: These equations also imply that whenever $f_k$ is holomorphic all the non-holomorphic 

1027: dependence of $D_t f_k$ and $D_\tau f_k$ lies in a term involving the propagator.

1028: In other words, once again all anti-holomorphic dependence arises through the propagator $\widehat E_2$ only.

1029: The generalizations of the modular derivatives \eqref{der_nonhol} and \eqref{der_hol} will 

1030: reappear in later sections of this work. For the Enriques Calabi-Yau they are given in \eqref{cov_D},\eqref{def-KY}

1031: and \eqref{der_holE}, while in the general discussion of compact Calabi-Yau manifolds they appear in 

1032: \eqref{cov_Dgen1},\eqref{cov_Dgen2} and \eqref{Dhol}.

1033: 

1034: Here we will us the covariant derivatives  \eqref{der_nonhol} and \eqref{der_hol} 

1035: to rewrite the holomorphic anomaly equations  (\ref{rec_Fg}). Firstly,

1036: we will apply modularity and the fact that all non-holomorphic 

1037: dependence arises through the propagator $\widehat E_2(\tau ,\tau)$ to convert anti-holomorphic 

1038: derivatives into derivatives with respect to $\widehat E_2$. Using \eqref{splitSW} we will be able to 

1039: carefully keep track of the $\widehat E_2$ 

1040: dependence in the holomorphic anomaly equations. Eventually, a solution will be simply obtained by

1041:  direct integration of a polynomial in $\widehat E_2$. 

1042:  

1043: To begin with, note that the holomorphic anomaly equations specialize in the local limit to

1044: \beq

1045:   \partial_{\bar t}   F^{(g)} = \tfrac{1}{2} C^{(0)tt}_{\bar t} \Big(D_t \partial_t F^{(g-1)}+

1046: \sum_{r=1}^{g-1} \partial_t F^{(r)} \partial_t F^{(g-r)}\Big)\ .

1047: \eeq

1048: Using the fact that all non-holomorphic dependence arises only through the 

1049: propagator $\widehat E_2(\tau,\bar \tau)$, this equation can be rewritten as 

1050: \beq \label{SWholan2}

1051: \frac{\partial   F^{(g)}}{\partial \widehat E_2}= \tfrac{1}{48} \Big(D_t \partial_t F^{(g-1)}+

1052: \sum_{r=1}^{g-1} \partial_t F^{(r)} \partial_t F^{(g-r)}\Big)\ .

1053: \eeq

1054: Here we used \eqref{proploc} to substitute $C^{(0)tt}_{\bar t} $ with the derivative $\partial_{\bar t} \widehat E_2$, 

1055: which then cancels with the same factor arising on the left-hand side of this equation. 

1056: Let us now manipulate the right-hand side of \eqref{SWholan-final} and split off the derivative 

1057: of $F^{(1)}$ in the second term

1058: \beq \label{SWholan3}

1059: \frac{\partial   F^{(g)}}{\partial \widehat E_2}= \left\{\begin{array}{ll} \tfrac{1}{48}\Big(D_t \partial_t F^{(1)}+  (\partial_t F^{(1)})^2\Big)  &g=2\ ,\\ 

1060: \ds{\tfrac{1}{48} \Big((D_t + 2 \partial_t F^{(1)}) \partial_t F^{(g-1)}+

1061: \sum_{r=2}^{g-2} \partial_t F^{(r)} \partial_t F^{(g-r)}\Big)}\qquad \qquad& g>2 \ ,\end{array}\right.

1062: \eeq

1063: where the sum now runs from $r=2$ to $r=g-2$. One then notes that $\partial_t F^{(1)}$ can be replaced

1064: by $-\frac{1}{24} C\widehat E_2$ by using \eqref{swpropagator}. Furthermore, we replace

1065: the non-holomorphic derivative $D_t$ with its holomorphic 

1066: counterpart $\hat D_t$ via \eqref{splitSW}. Altogether, one evaluates

1067: \beq \label{F2D}

1068:    \frac{\partial   F^{(2)}}{\partial \widehat E_2}= -\tfrac{1}{48 \cdot 24}\Big(\hat D_t (C \widehat E_2) -  \tfrac18 ( C \widehat E_2)^2\Big) 

1069: \eeq

1070: for genus two and for $g>2$

1071: \beq \label{SWholan-final}

1072: \frac{\partial   F^{(g)}}{\partial \widehat E_2}=\tfrac{1}{48} \Big((\hat D_t -\tfrac16 C \widehat E_2) \partial_t F^{(g-1)}+

1073: \sum_{r=2}^{g-2} \partial_t F^{(r)} \partial_t F^{(g-r)}\Big)\ .

1074: \eeq

1075: 

1076: We are now in the position to make the dependence on $\widehat E_2$ explicit. This can be done

1077: by rewriting the right-hand side of \eqref{SWholan-final} using \eqref{der_hol}. We also define

1078: $\hat d_t$ and $\hat d_\tau$ as covariant derivatives $D_t,\hat{D}_\tau$ not acting on the propagators $\widehat E_2$, such 

1079: that e.g. $

1080:    \hat d_\tau( \hat E^k_2\, c^{(r)}_k)=   \hat E^k_2\,  \hat D_\tau c^{(r)}_k $.

1081: Applying the chain rule we find 

1082: \beq \label{first_D}

1083:    \partial_t F^{(r)}  = \big[\hat d_t + (\hat D_t \widehat E_2) {\partial}_{\widehat E_2} \big] F^{(r)}

1084:                                 = C \big[\hat d_\tau - \tfrac{1}{12} (E_4+ \widehat E_2^2) {\partial}_{\widehat E_2} \big] F^{(r)}\ ,

1085: \eeq

1086: where \eqref{hatE2}, \eqref{der_hol}  and  \eqref{E-der} are applied to evaluate the derivative of $E_2$. 

1087: The Eisenstein series $E_4$ arises naturally in rewriting the derivatives. 

1088: We will therefore work with the ring $\bbC[\widehat E_2, h,E_4]$ introduced in \eqref{ring_hE4}.

1089: 

1090: Similarly, we rewrite the second derivative 

1091: \bea \label{second_D}

1092:    \hat D_t \partial_t F^{(g-1)}

1093:    &=&\tfrac{1}{12^2}C^2\Big(12^2 \hat d_\tau^2+6^2 h \hat d_\tau + 2   E_4(\widehat E_2{\partial}_{\widehat E_2}  +   \widehat E_2^2 {\partial}_{\widehat E_2}^2)\nn \\

1094:      &&     -( 3 h  + 12 \hat d_\tau )\widehat E_2^2 {\partial}_{\widehat E_2}+  2 \widehat E_2^3{\partial}_{\widehat E_2}+  \widehat E_2^4  {\partial}_{\widehat E_2}^2 \\

1095:     && +  (  -9 E_4 h + 2 h^3 - 12  E_4 \hat d_\tau   ) {\partial}_{\widehat E_2} +E_4^2 {\partial}_{\widehat E_2}^2 \Big)F^{(g-1)}\ ,\nn 

1096: \eea

1097: where we have used that the derivative 

1098: of $C$ is given by $ \hat D_\tau C =\tfrac14  h\, C$.

1099: This is how the holomorphic modular form $h$  defined in \eqref{ring_hE4} arises in the direct integration. 

1100: 

1101: We can now actually perform the direct integration. This is done by inserting the expressions 

1102: \eqref{first_D} and \eqref{second_D} for $\partial_t F^{(r)}$ and $\hat D_t \partial_t F^{(g-1)}$ into the holomorphic 

1103: anomaly equation \eqref{SWholan-final}. Replacing all $F^{(r)}$ for $1<r<g$ with their propagator expansion 

1104: \eqref{eq:generallocalform}, it is then straightforward to keep track of the number 

1105: of propagators $\widehat E_2$ in each term of the right-hand side of \eqref{SWholan-final}. Finally, $F^{(g)}$ is  

1106: determined up to a $\widehat E_2-$independent ambiguity by integrating the resulting polynomial in $\widehat E_2$.

1107: Without much effort this procedure can be repeated iteratively up 

1108: to the desired genus.

1109: 

1110: Note that the equation \eqref{F2D} for $F^{(2)}$ is particularly simple to integrate. 

1111: Using  \eqref{der_hol}  and  \eqref{E-der}  one evaluates  

1112: \beq

1113:  \hat D_t (C \widehat E_2) -  \tfrac18 ( C \widehat E_2)^2=\tfrac{1}{24}C^2 \big(-5 \widehat E_2^2+ 6 \widehat E_2 h-2 E_4 \big)\ .

1114: \eeq

1115: Inserted into \eqref{SWholan-final} it is straightforward to integrate this quadratic polynomial in $\widehat E_2$ 

1116: to derive $F^{(2)}$ as

1117: \begin{equation}

1118: {F}^{(2)}(\tau,\bar \tau)= \tfrac{1}{2\cdot 24^3} C^2\left(

1119: \tfrac{5}{3}\, \widehat E_2^3 - 3\,h \widehat E_2^2 + 2\,E_4 \widehat E_2\right) +C^2 c^{(2)}_0,

1120: \label{swF2}

1121: \end{equation}

1122: where $c^{(2)}_0(h,E_4)$ is the holomorphic ambiguity which can be fixed

1123: by additional boundary conditions as we discuss in the next section.

1124: For genus up to $7$ the expressions for $F^{(g)}$ were 

1125: calculated in \cite{hk} using the Feynman graph expansion. 

1126: The direct integration using (\ref{SWholan-final}) provides a far more effective 

1127: method to solve Seiberg-Witten theory and confirms the results of \cite{hk}.

1128: Furthermore, the modular properties of the expressions are 

1129: manifest at each step. As we will discuss in the later sections, similar 

1130: constructions will  provide us with a 

1131: powerful tool to determine the set of candidate modular generators

1132: for more complicated Calabi-Yau manifolds. In particular, holomorphic 

1133: modular forms are needed to parametrize the holomorphic ambiguity. 

1134: In case we know the ring of holomorphic modular forms, fixing the ambiguity

1135: reduces to a determination of a finite set of numerical factors 

1136: at each genus. For Seiberg-Witten theory this can be done systematically, 

1137: as we will discuss in the next section.

1138: 

1139: 

1140:    

1141: \subsection{Boundary conditions \label{SWboundary}}

1142: 

1143: To systematically fix the  $c^{(g)}_0$ we have to understand the boundary behavior

1144: of the $F^{(g)}$. As it is well known, there are three distinguished regions in the moduli space of pure $SU(2)$

1145: $\cN=2$ SYM which correspond to the geometrical 

1146: singularities of ${\cal E}$. We will parametrize the moduli space by the vacuum expectation 

1147: value  $u=\langle {\rm Tr} \Phi^2\rangle$ of the scalar $\Phi$ in the $\cN=2$ vector

1148: multiplet. The first  region occurs at $u\sim \frac{1}{2} t^2 \rightarrow 

1149: \infty$, and it corresponds physically to the semiclassical regime. 

1150: The monopole region occurs near $u\rightarrow \Lambda^2$, where a magnetic monopole of charge $(e,m)=(0,1)$ becomes 

1151: massless and the electric $SU(2)$ theory with gauge coupling ${\rm Im} \tau$ is 

1152: strongly coupled. At the point $u\rightarrow -\Lambda^2$  a dyon of charge $(e,m)=(-1,1)$

1153: becomes massless. However, this point is identified with the monopole point by a $\mathbb{Z}_2$ exact 

1154: quantum symmetry. For this reason there are no independent boundary 

1155: conditions at $u\rightarrow -\Lambda^2$ and we focus on  

1156: $u\rightarrow \Lambda^2$ and $u\sim \infty$. In both cases the 

1157: elliptic curve acquires a node, i.e. a local singularity of 

1158: the form $\xi^2+\eta^2=(u\pm \Lambda^2)$, where a cycle of $\IS^1$ topology shrinks. 

1159: In string theory, a point in the moduli space where a node in the target 

1160: geometry develops is called a conifold point.     

1161: 

1162: 

1163: The natural physical parameter in  the magnetic monopole region $u\rightarrow \Lambda^2$

1164: is $t_D$.  We get first a convergent expansion for the  $F^{(g)}$ in the 

1165: variable $q_D=\exp(2 \pi i \tau_D)$ for \mbox{$\tau_D=-{1\over \tau}\rightarrow i \infty$}, which corresponds to 

1166: $t_D\rightarrow 0$.  This is obtained by an $S$- transformation of the modular 

1167: expressions for the $F^{(g)}(\tau,\bar \tau)$ such as (\ref{swF2}), which converge 

1168: in the semiclassical region. The holomorphic magnetic 

1169: expansions ${\cal F}^{(g)}_D(\tau_D)$ can be obtained by formally taking the limit 

1170: $\bar \tau_D\rightarrow \infty$, while keeping $\tau_D$ fixed. Finally we obtain 

1171: the expansion in $t_D$ by inverting (\ref{t-td-period}). In these magnetic expansions, a gap structure was observed near the monopole (or conifold) point \cite{hk}. 

1172: One finds that the leading behavior of ${\cal F}^{(g)}_D(\tau_D)$ is of the form

1173: %  

1174: \begin{equation} 

1175: {\cal F}_D^{(g)}={B_{2g}\over 2 g ( 2g -2) \tilde t_D^{2g-2}}+ k^{(g)}_1 \tilde t_D+

1176: {\cal O}(\tilde t_D^2)\ ,

1177: \label{gap}

1178: \end{equation}

1179: %

1180: where the $B_{n}$ are the Bernoulli numbers and we used a rescaled variable 

1181: $\tilde t_D=i{t_D\over 2}$. The knowledge of the leading coefficients and the absence 

1182: of the remaining $2g-3$ sub-leading negative powers in the $\tilde t_D$ expansion imposes

1183: $2g-2$ conditions. Since ${\dim }M_{6g-3}(\Gamma(2))=\left[\frac{3 g-1}{2}\right]$ 

1184: this overdetermines the $c^{(g)}_0$, e.g.~for $g=2$ we 

1185: find $c^{(2)}_0=- \frac{1}{2\cdot 24^3}\big(\frac{1}{2}E_4\,h + \frac{1}{30}h^3\big)$. 

1186: It is very easy to integrate (\ref{SWholan-final}) using (\ref{first_D}), (\ref{second_D}) and 

1187: the gap condition, which fixes the ambiguity to arbitrary genus. This solves the 

1188: theory completely. One finds moreover a pattern in the first subleading term in the 

1189: magnetic expansion

1190: %

1191: \begin{equation}

1192: k^{(g)}_1={((2g-3)!!)^3\over g! 2^{7 g-2}}\ .

1193: \end{equation}

1194:  

1195:  

1196: The gap can be explained by using the embedding of Seiberg-Witten theory into type IIA string 

1197: theory compactified on a suitable Calabi--Yau manifold. The most generic singularity of a $d$ complex dimensional 

1198: manifold is a node where an $\IS^d$ shrinks. The 

1199: codimension one locus in the moduli space 

1200: where this happens is called the conifold. It was argued 

1201: in~\cite{strominger,Vafa:1995ta} that at the conifold a 

1202: RR-hypermultiplet becomes massless. This hypermultiplet 

1203: is charged and couples to the $U(1)$ vector multiplets. 

1204: Its one loop effect on the kinetic terms 

1205: of the vector multiplets in the effective action 

1206: is captured by the local expansion of $F^{(0)}$~\cite{strominger}. 

1207: A gravitational one-loop effect yields the moduli dependence 

1208: of the $R_+^2$ term in the effective action and is given by  local 

1209: expansion $F^{(1)}$~\cite{Vafa:1995ta}. Using further one-loop arguments

1210: it was shown that the $F^{(g)}$, which capture the moduli dependence 

1211: of the coupling of the self-dual part of the curvature to the 

1212: self-dual part of the graviphoton $R^2_+ \, F_+^{2g-2}$, have the following 

1213: gap structure

1214: %

1215: \begin{eqnarray} \label{thegap}

1216: F^{(g)}_{\textrm{conifold}}=\frac{(-1)^{g-1}B_{2g}}{2g(2g-2)t_D^{2g-2}}

1217: +\mathcal{O}(t_D^0),

1218: \end{eqnarray}

1219: %

1220: where $t_D$ is a suitable coordinate transverse to the conifold 

1221: divisor \cite{hkq}. The Seiberg-Witten gauge theory embedded in type IIA string theory inherits this structure, and the massless 

1222: hypermultiplet at the conifold is identified as a monopole becoming massless at the monopole 

1223: point. In this way, (\ref{thegap}) explains the field theory result (\ref{gap}) and extends it to the full 

1224: supergravity action. 

1225: 

1226: Once the Seiberg-Witten amplitudes $F^{(g)}$ have been determined in terms of modular 

1227: functions, these can be expanded around every point in the moduli space.

1228: For example, in the semiclassical regime

1229: $\tau\rightarrow i \infty$, $u\rightarrow \infty$ one finds the 

1230: holomorphic amplitudes

1231: %

1232: \begin{equation} 

1233: {\cal F}^{(g)}={(-1)^gB_{2g}\over  g ( 2g -2) (2 t)^{2g-2}}+ 

1234: {l^{(g)}_{2g+6} \over t^{2g+6}}+{\cal O}(t^{2g+10})\ .

1235: \end{equation}

1236: %

1237: The higher order terms in this expansion correspond to gauge theory 

1238: instantons and have been computed in \cite{Nek}.

1239: 

1240: 

1241: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1242: %%%%%%%%

1243: %%%%%%%%     Section 4

1244: %%%%%%%%

1245: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1246: 

1247: 

1248: \section{A first look at the Enriques Calabi-Yau}\label{sec:Enriques}

1249: 

1250: In this section we review some basic properties of topological string theory on

1251: the Enriques Calabi-Yau. We begin by reviewing the $N=2$ special 

1252: geometry of the classical moduli space of K\"ahler and complex structure deformations in 

1253: section \ref{specialE}. 

1254: The first world-sheet instanton corrections arise from genus one Riemann surfaces 

1255: as shown in refs.~\cite{fhsv,hmfhsv,km}. The holomorphic  higher genus free energies, restricted to the K3 fiber,

1256: can be also derived by using heterotic-type II duality \cite{km}. We briefly summarize these results

1257: in section \ref{genus1}. In understanding and deriving the expression for the 

1258: full $F^{(g)}$ an important hint is given by their transformation properties 

1259: under the symmetry group of the full topological string theory on the Enriques Calabi-Yau. 

1260: More precisely, 

1261: generalizing the results of the previous section, one expects 

1262: that all $F^{(g)}$ are built out of functions transforming in 

1263: a particularly simple way under the group $Sl(2,\bbZ) \times O(10,2,\bbZ)$.

1264: In section \ref{aut_prop} we review some essentials about these modular 

1265: and automorphic functions and forms.  

1266: 

1267: 

1268: 

1269: \subsection{Special geometry of the classical moduli space \label{specialE}}

1270: 

1271: The Enriques Calabi-Yau can be viewed as the first non-trivial generalization of 

1272: the product space $\IT^2 \times {\rm K3}$. It is defined as the orbifold 

1273: $(\IT^2\times {\rm K3})/\bbZ_2$, where $\bbZ_2$ acts as a free involution \cite{fhsv}. 

1274: This involution inverts the coordinates of the torus and acts as the Enriques 

1275: involution on the K3 surface. 

1276: The cohomology lattice of $\IT^2\times {\rm K3}$ takes the form \cite{Aspinwall:1996mn}

1277: \begin{equation} 

1278: \Gamma^{6,22}=\Gamma^{2,2}\oplus [\Gamma^{1,1}\oplus E_8(-1)]_1

1279: \oplus[\Gamma^{1,1}\oplus E_8(-1)]_2\oplus \Gamma^{1,1}_g \oplus \Gamma^{1,1}_s\ ,

1280: \label{coveringlattice}

1281: \end{equation}

1282: where the inner products on the sublattices $E_8(-1)$ and $\Gamma^{1,1}$ are given by   

1283: \beq

1284:   (C^{\alpha \beta})=-C_{E_8}\ ,\qquad \qquad (C^{ij})= \left(\begin{array}{cc}0 & 1\\ 1 & 0\end{array} \right)\ . 

1285: \eeq

1286: with $\alpha,\beta=1,\ldots,8$ and $i,j=1,2$. 

1287: Here $C_{E_8}$ is the Cartan matrix of the exceptional group $E_8$.

1288: The lattice \eqref{coveringlattice} splits into $H^1(\mathbb{T}^2)\oplus H_1(\mathbb{T}^2)=\Gamma^{2,2}$ 

1289: and $H^*(K3)=\Gamma^{4,20}$. Under heterotic-type II duality it can be 

1290: identified with the Narain lattice of the heterotic compactification on $\mathbb{T}^6$. 

1291: The $\bbZ_2$ involution on the Enriques Calabi-Yau 

1292: acts on the five terms of the lattice \eqref{coveringlattice} as~\cite{fhsv} \footnote{The  effect of the phase factor 

1293: on the type II side was interpreted as turning on a Wilson line~\cite{fhsv}.}

1294: \begin{equation}

1295: |p_1,p_2,p_3,p_4,p_5\rangle\rightarrow  e^{\pi i \delta\cdot p_5} |-p_1,p_3,p_2,-p_4,p_5\rangle\ ,

1296: \end{equation} 

1297: where $p_i$ is an element of the $i$-th term in \eqref{coveringlattice} and we denoted

1298: $\delta=(1,-1)\in \Gamma^{1,1}_s$.                

1299: 

1300: 

1301: The Enriques Calabi-Yau has holonomy group $SU(2)\times \bbZ_2$. 

1302: This implies that type II string theory compactified on the Enriques Calabi-Yau

1303: will lead to a four-dimensional theory with $\cN=2$ supersymmetry. 

1304: Nevertheless, due to the fact that it does not have the full $SU(3)$ holonomy

1305: of generic Calabi-Yau threefolds, various special properties related to $\cN=4$ 

1306: compactification on $\IT^2 \times K3$ are inherited. 

1307: 

1308: As an example of the close relation of the Enriques Calabi-Yau to 

1309: its $\cN=4$ counterpart $\IT^2 \times K3$ one notes that the moduli 

1310: space of K\"ahler and complex structure deformations are simply 

1311: cosets. The complex dimensions of these moduli spaces are given by

1312: the dimensions $h^{(1,1)}$ and $h^{(2,1)}$ of the cohomologies 

1313: $H^{(1,1)}$ and $H^{(2,1)}$. They can be determined constructing 

1314: a basis of $H^{(p,q)}$ of forms of K3 and $\IT^2$ invariant 

1315: under the free involution. One obtains \cite{fhsv}

1316: \beq

1317:     h^{(2,1)}=h^{(1,1)}=11\ ,

1318: \eeq

1319: while $H^{(0,0)},\ H^{(3,3)}$ as well as $H^{(3,0)}$ are one-dimensional. 

1320: Moreover, one can show that the Enriques Calabi-Yau is self-mirror and

1321: that both the K\"ahler and complex structure moduli spaces are given by the coset

1322: %

1323: \be

1324: \label{coset}

1325:  \cM=\frac{Sl(2,\bbR)}{SO(2)} \times \CN_8\ ,

1326: \ee

1327: %

1328: where 

1329: %

1330: \be

1331: \label{nscoset}

1332: \CN_s= 

1333: {O(s+2,2) \over O(s+2) \times O(2)}.

1334: \ee

1335: %

1336: The actual moduli space is obtained after dividing $\CM$ by the discrete groups $Sl(2,\bbZ)\times O(10,2;\IZ)$. 

1337: $\cM$ is a simple example 

1338: of a special K\"ahler manifold. We will discuss its properties in the following. 

1339: 

1340: It is a well-known fact that the geometric moduli space of a Calabi-Yau manifold 

1341: consists of two special K\"ahler manifolds corresponding to K\"ahler and 

1342: complex structure deformations. A summary of some of the 

1343: basic definitions and identities of special geometry can be found in 

1344: appendix \ref{N=2sp}. Essentially all information is encoded in one 

1345: holomorphic function, the prepotential $\cF$. Let us for 

1346: concreteness consider the moduli space of K\"ahler structure 

1347: deformations of the Enriques Calabi-Yau which is of the form \eqref{coset}.

1348: Denoting by $\hat \omega$ the harmonic $(1,1)$-form in the $\IT^2$-base

1349: and by $\omega_a$ the $(1,1)$ forms in the Enriques fiber, we 

1350: obtain complex coordinates $S,t^a$ by expanding

1351:   the combination 

1352:   %

1353:   \beq

1354:   \label{kahl}

1355:  J+i B_2=\ S\, \hat \omega +t^a\, \omega_a \ ,\qquad \quad a=1,\ldots,10\ ,

1356: \eeq

1357: %

1358: where $J$ is the K\"ahler form on the Enriques Calabi-Yau and $B_2$ is the NS-NS two-form.

1359: Note that in our conventions $\R\, S >0$ and $\R\, t^a > 0$ such that the world-sheet  instantons

1360: arise as series in $q_S= e^{-S}$ and $q_{t^a}=e^{-t^a}$ in the large radius expansion. We note that 

1361: these complexified K\"ahler parameters $t^a$ can be regarded as a parametrization of the coset $\CN_8$. The parametrization 

1362: we are using here is the one suitable for the conventional large radius limit and corresponds to what was called in \cite{km} the 

1363: geometric reduction. In terms of (\ref{kahl}), the prepotential takes the form

1364: %

1365: \beq \label{def-Eprepot}

1366:   \cF = - \tfrac{i}2 C_{ab} t^a t^b S \ .

1367: \eeq

1368: %

1369: For the Enriques Calabi-Yau the cubic expression for the genus zero free energy $F^{(0)}=\cF$ is exact and

1370: world-sheet instanton corrections will only arise at higher genus. This is precisely the reason 

1371: for the simple form \eqref{coset} of the moduli space.

1372: The symmetric matrix $C_{ab}$ in \eqref{def-Eprepot} 

1373: encodes the intersections in the Enriques fiber $E$ such that 

1374: \be

1375: C_{ab}=\int_E \omega_a \wedge \omega_b\ . 

1376: \ee

1377: The inverse matrix $C^{ab}\equiv C^{-1\, ab}$ can be calculated explicitly and coincide in an appropriate basis 

1378: with the intersection matrix of the $\bbZ_2$ invariant lattice of the second and the third factor in 

1379: (\ref{coveringlattice}), i.e.   

1380: %

1381: \be \label{def-Gamma_E}

1382: \Gamma_E=\Gamma^{1,1}\oplus E_8 (-1)\ , \qquad \quad (C^{ab})=\left(\begin{array}{cc}0 & 1\\ 1 & 0\end{array} \right) \times (-C_{E_8}).

1383: \ee

1384: %

1385: Here $C_{E_8}$ is the  Cartan matrix of the exeptional group $E_8$. The

1386: lattice $\Gamma_E$ is identified with the second cohomology group of the Enriques surface. 

1387: 

1388: The prepotential for the Enriques Calabi-Yau encodes the classical geometry 

1389: of the moduli space \eqref{coset}. The K\"ahler potential is derived using equation

1390: \eqref{KpotII} to be of the form\beq \label{def-KY}

1391: K=-\log \big[  Y (S+\bar S)\big]\ ,\qquad Y=\tfrac12 C_{a b} (t^a+\bar t^a)  (t^b+\bar t^b) \ .

1392: \eeq

1393: Note that $K$ as given in \eqref{KpotII} contains a term 

1394: $-\log |X^0|^2$, with $X^0$ being the fundamental period. Such a term can be removed 

1395: by a K\"ahler transformation $K\rightarrow K-f-\bar f$, where 

1396: $f$ is a holomorphic function, such that our expression \eqref{def-KY} corresponds to 

1397: a certain K\"ahler gauge. In general, all objects we will consider below are sections of 

1398: a line bundle $\cL$ which parametrizes such holomorphic rescalings $V\rightarrow e^{f}V$. 

1399: As an example $e^{-K}$ is a section of $\cL\otimes \bar \cL$.

1400: Such K\"ahler transformations do not change the K\"ahler metric which is 

1401: obtained by evaluating the holomorphic and anti-holomorphic derivative of $K$.

1402: The K\"ahler metric splits into two pieces 

1403: \beq

1404:   G_{S\bar S} = \frac{1}{(S+\bar S)^2}\ ,\qquad G_{a \bar b} = -\frac{C_{a b}}{Y} + \frac{ C_{ac}(t+\bar t)^c C_{bd} (t+\bar t)^c}{Y^2} \ ,

1405: \eeq

1406: with all other components vanishing.

1407: The Christoffel symbols for this metric are easily evaluated to be

1408: %

1409: \be \label{EnriquesChris}

1410: \Gamma^S_{SS}=2 K_S\ , \qquad

1411: \Gamma^c_{ab}=K_e C^{ed}  \hat \Gamma^c_{ab|d} \ ,

1412: \ee

1413: where $K_S$ and $K_a$ are the first derivatives of the K\"ahler potential \eqref{def-KY} 

1414: and we have defined  

1415: \beq \label{def-hatGamma}

1416:   \hat \Gamma^b_{ac|d} = \big(\delta^b_c C_{ad} +\delta^b_a C_{cd} - \delta^b_d C_{ac} \big)\ .

1417: \eeq

1418: It is also easy to derive the holomorphic Yukawa couplings $ C^{(0)}_{ijk}$ defined in \eqref{def-CC}. 

1419: In coordinates $S,t^a$ one uses the prepotential \eqref{def-Eprepot} to 

1420: show 

1421: \beq

1422:     C^{(0)}_{S ab}= C_{ab}\ .

1423: \eeq

1424: 

1425: In general $C^{(0)}_{S ab}$ is a section of $\CL^{2}\otimes {\rm Sym}^3(T^*{\CM})$. In the case of the Enriques 

1426: Calabi--Yau it is 

1427: constant in the K\"ahler gauge and coordinates chosen above, and covariantly constant in a general gauge. 

1428: The covariant derivative, acting on a section of 

1429: $\cL^{m}\otimes \bar \cL^{n}$, is \eqref{cov_D} 

1430: %

1431: \be

1432: D_a=\partial_a+m K_a,\qquad D_{\bar{a}}=\partial_{\bar a}+n K_{\bar a},

1433: \ee

1434: %

1435: and includes the Christoffel symbols when acting on tensors. 

1436: Applied to $C^{(0)}_{S ab}$ one shows

1437: \be

1438: D_c C^{(0)}_{abS} = -\Gamma_{ca}^d C_{db} -\Gamma_{cb}^d C_{ad} + 2 \partial_c K C_{ab}=0\ , 

1439: \ee

1440: %

1441: which vanishes by means of the equation \eqref{EnriquesChris} for the Christoffel symbols. 

1442: A similar equation holds for the covariant derivative $D_S C^{(0)}_{abS}$, showing that $C^{(0)}_{abS}$

1443: is indeed covariantly constant. Once again, this special property of the Yukawa couplings is immediately 

1444: traced back to the fact that the prepotential $\cF$ receives no instanton corrections.

1445: 

1446: The space $\CM$ has two different types of singular loci in complex codimension one on the moduli 

1447: space~\cite{fhsv,aspinwall} which lead to conformal field theories in four dimensions. 

1448: The first degeneration comes from the shrinking of a smooth rational

1449: curve $e\in \Gamma_E$ with $e^2=-2$. 

1450: The shrinking $\IP^1$ leads to an $SU(2)$ gauge symmetry enhancement 

1451: together with a massless hypermultiplet, also in

1452: the adjoint representation of the gauge group. We then obtain for

1453: this point the massless spectrum of ${\cal N}=4$ supersymmetric 

1454: gauge theory. In terms of the complexified K\"ahler parameters introduced 

1455: in (\ref{kahl}) this singular locus occurs along  

1456: %

1457: \be

1458: \label{slocusone}

1459: t^1 =t^2. 

1460: \ee

1461: %

1462: In order to understand the second singular locus, we first point out that the coset $\CN_8$ can be parametrized in many 

1463: different ways. In \cite{km} it was noticed that there is a parametrization of this coset in terms of some coordinates $\tD^a$, $a=1, \cdots, 10$  

1464: which are related to what was called there the BHM reduction. By using the formulae in \cite{km} it is easy to see that the coordinates $t^a$ and 

1465: $\tD^a$ are related by the following simple projective transformation, 

1466: %

1467: \be

1468: \label{projtrans}

1469: \ba

1470: t^1&=\tD^1 -{1\over 4 \tD^2}\sum_{i=3}^{10} (\tD^i)^2,\\

1471: t^2&={2 \pi^2 \over \tD^2},\\

1472: t^i &=-\pi \ri { \tD^i \over \tD^2},\quad i=3, \cdots, 10.

1473: \ea

1474: \ee

1475: %

1476: The second singular locus occurs when 

1477: %

1478: \be

1479: \label{slocustwo}

1480: \tD^1=\tD^2. 

1481: \ee

1482: %

1483: On this locus one gets as well an $SU(2)$ gauge symmetry enhancement. In addition

1484: one gets four hypermultiplets in the fundamental representation

1485: of $SU(2)$, and the resulting gauge theory

1486: is ${\cal N}=2$, $SU(2)$ Yang-Mills theory with four massless hypermultiplets. In \figref{modulispace} we represent schematically the two singular 

1487: loci in moduli space, related by the projective transformation  (\ref{projtrans}). In sections \ref{sec:diE} and \ref{sec:ftlim} of this paper we will explore in some detail the 

1488: field theory limit of the topological string amplitudes and we will verify this picture of the moduli space.  

1489: 

1490: \begin{figure}[!ht]

1491: \leavevmode

1492: \begin{center}

1493: %\epsfysize=5.5cm

1494: \includegraphics[height=5cm]{moduli.eps} 

1495: \end{center}

1496: \caption{The singular loci in the moduli space $\CN_8$, leading to two different gauge theories in the field theory limit. }

1497: \label{modulispace}

1498: \end{figure}

1499: %

1500: 

1501: \subsection{Genus one and the free energies on the Enriques fiber\label{genus1}}

1502: 

1503: So far we have discussed the classical moduli space of the Enriques Calabi-Yau $Y$. 

1504: We introduced the prepotential $\cF$ which is cubic in the K\"ahler structure deformations and 

1505: receives no worldsheet instanton corrections. One expects that such a simple structure will no longer persist at 

1506: higher genus. This is already true at genus one as was shown in \cite{hmfhsv,km}. Heterotic--type II 

1507: duality can also be used to determine all higher genus free energies on the 

1508: K3 fibers of the Enriques Calabi-Yau \cite{km}. In this section we will summarize some results of \cite{km}

1509: and present a closed expression for the fiber free energies also including the anti-holomorphic dependence. 

1510: 

1511: Let us begin with a brief discussion of the free energies for the Enriques fiber. The fiber limit of the topological 

1512: string amplitudes corresponds to blowing up the volume of the base space by taking 

1513: %

1514: \beq

1515: S \ \rightarrow \ \infty\ , \qquad \qquad  q_S \equiv e^{-S} \ \rightarrow \ 0\ .

1516: \eeq

1517: %

1518: In what follows we will need to distinguish the full topological string amplitudes 

1519: $F^{(g)}$ from their fiber limits as well as from their holomorphic limits. We will denote,

1520: %

1521: \be

1522: F^{(g)}_E(t,\bar{t})=\lim_{S\rightarrow \infty}F^{(g)}(t,\bar{t})

1523: \ee

1524: %

1525: and

1526: %

1527: \be

1528: \label{holofiber}

1529: \CF^{(g)}_E(t)=\lim_{\bar{t}\rightarrow \infty}F^{(g)}_E(t,\bar{t}).

1530: \ee

1531: %

1532: The fiber limit $F^{(g)}_E(t,\bar t)$ can be calculated using heterotic-type II duality \cite{agnt,mm,km}. In the heterotic string 

1533: they are given by a one--loop computation of the form

1534: %

1535:  \be

1536:  \label{hetint}

1537: F^{(g)}_E (t, \bar t)=\int \rd\tau \, {\overline \Theta}^g_{\Gamma} (\tau, v^+) f_g(\tau, \bar \tau) /Y^{g-1}

1538: \ee

1539: %

1540: where $Y$ is defined in \eqref{def-KY}, and $\Theta^g_{\Gamma} (\tau, v^+) $ is a theta function with an insertion of $2g-2$ powers of the right--moving 

1541: heterotic momentum. We will not need the precise definitions of $\Theta^g_{\Gamma}$ and $f_{g}$ here. However, 

1542: it is important to note that

1543: these amplitudes can be evaluated in closed form by using standard techniques for one--loop integrals. 

1544: The holomorphic limit (\ref{holofiber}) was determined in \cite{km} and 

1545: it is given by

1546: %

1547: \beq

1548: \label{gr}

1549:   \cF^{(g)}_E (t) = \sum_{r>0} c_g(r^2) \Big[2^{3-2g} \text{Li}_{3-2g}(e^{-r\cdot t}) -  \text{Li}_{3-2g}(e^{-2r\cdot t}) \Big] \ ,

1550: \eeq

1551: %

1552: where $\text{Li}_{n}$ is the polylogarithm of index $n$ defined as

1553: \beq

1554: {\rm Li}_n (x) =\sum_{d=1}^{\infty} {x^d\over d^n}\ .

1555: \eeq

1556: In formula \eqref{gr} we have also set $r^2 =C^{ab}r_a r_b$ and $r\cdot t = r_a t^a$. We will sometimes write

1557: %

1558: \be

1559: r=(n,m, \vec q).

1560: \ee

1561: %

1562: The restriction $r>0$ means $n>0$, 

1563:  or $n=0, m>0$, or $n=m=0$, $\vec q>0$.

1564: Finally, we need to define the coefficients $c_g(n)$. They can be 

1565: identified as the expansion coefficients of a particular quasi-modular form

1566: %

1567: \be

1568: \label{geomrmod}

1569: \sum_n c_g(n) q^n =-2 \frac{{\cal P}_{g}(q)}{\eta^{12}(2\tau)},

1570: \ee

1571: %

1572: with $\CP_g(q)$ given by

1573: %

1574: \be

1575: \label{defpg}

1576: \biggl( { 2\pi  \eta^3 \lambda \over \vartheta_1(\lambda|\tau)}\biggr)^2=

1577: \sum_{g=0}^{\infty} (2 \pi \lambda)^{2g} {\cal P}_{g}(q).

1578: \ee

1579: %

1580: The definition of $\eta(\tau)$ and the theta-function $\vartheta_1(\lambda|\tau)$ can be found 

1581: in Appendix \ref{theta}. {}From the definition \eqref{defpg} and the identities summarized in 

1582: Appendix  \ref{theta} one also infers that the $\cP_g$ are quasimodular forms of weight $2g$ 

1583: and can be written as polynomials in the Eisenstein series $E_2, E_4, E_6$. We have for example

1584: % 

1585: \be

1586: \label{casesps}

1587: \CP_1(q)=\tfrac{1}{12} E_2(q)\ , \,\,\,\,\,\, \quad \CP_2(q)=\tfrac{1}{1440} (5 E_2^2 + E_4)\ .

1588: \ee

1589: %

1590: 

1591: In general, as we will see in section \ref{sec:diE}, it is very hard to include the $\IT^2$-base in

1592: order to obtain the expressions $\cF^{(g)}$ for the full Enriques Calabi-Yau. It turns out that only $\cF^{(1)}$

1593: factorizes nicely, namely we can write the A--model free energy $\cF^{(1)}$ as \cite{hmfhsv,km}

1594: %

1595: \beq

1596:    \cF^{(1)}(S,t) = \cF^{(1)}_{\rm base} +\cF^{(1)}_E  \ ,

1597: \eeq  

1598: %

1599: where $\cF^{(1)}_{\rm base}$ and $\cF^{(1)}_E $ are the contributions from the $\IT^2$ base 

1600: and the K3 fiber. $\cF^{(1)}_{\rm base}$ is the torus free energy given by \cite{bcov1}

1601: \beq

1602:   \cF^{(1)}_{\rm base} = - 12 \log \eta(S)\ ,

1603: \eeq

1604: %

1605: where $\eta(S)$ is defined in \eqref{dede}, while 

1606: %

1607: \be

1608: \label{gonehol}

1609: \cF^{(1)}_E=- \tfrac{1}{2} \log \Phi(t), 

1610: \ee

1611: %

1612: where $\Phi(t)$ is the infinite product

1613: %

1614: \be \label{def-Phi}

1615:   \Phi(t) =  \prod_{r>0}  \left(\frac{1  -e^{-r\cdot t}}{1 + e^{-r\cdot t}}\right)^{2c_1(r^2)} \ .

1616: \eeq

1617: %

1618: This infinite product first appeared in the work of Borcherds \cite{borcherds}. As we will discuss in more detail later on, $\Phi(t)$ is the key example of a  

1619: holomorphic automorphic form for the Enriques Calabi-Yau. It is also convenient to introduce, 

1620: %

1621: \be

1622: \Phi(S,t) = \eta^{24}(S) \Phi(t), 

1623: \ee

1624: %

1625: so that we can write

1626: %

1627: \be

1628:   \cF^{(1)}(S,t)=-\tfrac{1}{2} \log \, \Phi(S,t).

1629:   \ee

1630: 

1631: We presented above formulae for the holomorphic limit of $F^{(g)}_E(t,\bar t)$, but heterotic-type II duality can be 

1632: used as well to obtain the antiholomorphic dependence on $\bar t$. At genus one, one finds \cite{hm,mm} 

1633: %

1634: \be

1635: \label{feoneanti}

1636: F^{(1)}_E (t, \bar t) =-2\log Y -\log \big|\Phi(t) \big|.

1637: \ee

1638: %

1639:  The antiholomorphic dependence on $\bar S$ is the usual one for the torus \cite{bcov1} and one has 

1640:  %

1641: \beq \label{F1Enriques}

1642:  F^{(1)}(S, \bar S,t,\bar t) =  F^{(1)}_E(t,\bar t) - 6 \log \Bigl( (S+ \bar S) |\eta^2(S)|^2\Bigr).

1643: \eeq

1644: %

1645: Equivalently, we can write

1646: %

1647: \be \label{F1Enriques2}

1648:  F^{(1)}(S, \bar S,t,\bar t)=  - 2\log \big[ (S+\bar S)^3 Y \big] - \log \big|\Phi(S,t)\big|.

1649: \ee

1650: %

1651: As a consistency check one shows that this anti-holomorphic dependence can also be 

1652: inferred from the holomorphic anomaly equation \eqref{anomaly_F1} for $F^{(1)}$.

1653: 

1654: The antiholomorphic dependence in the heterotic calculation at higher genus is much more complicated, 

1655: but was written down for the STU model in \cite{mm}. As we show in Appendix \ref{heteroticFg}, this 

1656: computation can be considerably simplified and adapted to the Enriques case. 

1657: We find that the non-holomorphic free energy $F^{(g)}_E(t,\bar t)$ can be cast into the 

1658: form

1659: %

1660: \bea\label{Fgantihol}

1661: F^{(g)}_E (t, \bar t)

1662: &=&\sum_{l=0}^{g-1}\sum_{C=0}^{\begin{subarray}{c}{\rm min}\\(l,2g-3-l)\end{subarray}}  

1663: \left(\begin{array}{c}2g-3-l \\ C \end{array}\right)

1664:   {(t+\bar{t})^{a_1}\ldots(t+\bar{t})^{a_{l-C}}\partial_{a_1}\ldots \partial_{a_{l-C}}\CF^{(g-l)}_E(t)\over(l-C)! 2^l\ Y^l}\nn\\

1665:   &&-{1\over 2^{g-2}(g-1)Y^{g-1}}\ ,

1666: \eea

1667: %

1668: where $\cF^{(r)}_E(t)$ is the holomorphic fiber expression given in \eqref{gr}. 

1669: It is easy to check that the $F^{(g)}_E(t, \bar t)$ fulfill the holomorphic anomaly equation on the fiber.

1670: 

1671: So far we have discussed the heterotic results for the fiber limit by using the K\"ahler parameters (\ref{kahl}) 

1672: appropriate for the large radius limit. As shown in \cite{km}, one can also compute them in the 

1673: coordinates $\tD^a$ introduced in (\ref{projtrans}). This was called the BHM reduction in \cite{km}, and leads to the 

1674: holomorphic couplings,  

1675: %

1676: \be

1677: \label{finalfgbor}

1678: \CF^{(g)}_E (\tD)=\sum_{r>0} d_g(r^2/2)(-1)^{n+m} {\rm Li}_{3-2g}({\rm e}^{-r \cdot \tD })

1679: \ee

1680: %

1681: where the coefficients $d_g(n)$ are defined by

1682: %

1683: \be

1684: \sum_n d_g (n) q^n =\frac{2^{2+g} {\cal P}_{g}(q^4)-2^{2-g} {\cal P}_{g}(q)}{\eta^{12}(2\tau)},

1685: \label{finalmodcoef}

1686: \ee

1687: %

1688: and in (\ref{finalfgbor}) we regard $r$ as a vector in $\Gamma^{1,1}\oplus E_8(-2)$. Note that

1689: in comparison to \eqref{def-Gamma_E} we now need to include the lattice 

1690: $E_8(-2)$ with inner product given by $-2$ times the Cartan matrix of $E_8$, such that $r^2=2nm-2 \vec q^{~2}$. One has, in particular, 

1691: %

1692: \be

1693: \CF^{(1)}_E (\tD)=-\tfrac{1}{2} \log \Phi_B(\tD)\ , 

1694: \ee

1695: %

1696: where 

1697: %

1698: \be

1699: \Phi_B(\tD)=\prod_{r>0} \Bigl(1- \re^{-r \cdot \tD}\Bigr)^{(-1)^{n+m} c_{B}(r^2/2)}

1700: \ee

1701: %

1702: with coefficients

1703: %

1704: \be

1705: \sum_n c_{B} (n) q^n ={\eta(2 \tau)^8 \over \eta(\tau)^8 \eta(4 \tau)^8}.

1706: \ee

1707: %

1708: This is the modular form introduced by Borcherds in \cite{borcherdsone}, and the above

1709: expression for $F_1$ agrees with that found by Harvey and Moore in \cite{hmfhsv} (up to a factor of $1/2$ due to

1710: different choice of normalizations). 

1711: 

1712: \subsection{An all--genus product formula on the fiber}

1713: 

1714: As we have already mentioned, the infinite product (\ref{def-Phi}) was first considered by Borcherds in \cite{borcherds}. Borcherds also noticed 

1715: that (\ref{def-Phi}) is the denominator formula for a generalized Kac--Moody (or Borcherds) superalgebra (see \cite{gebert, hm} for a review of 

1716: Borcherds algebras). The root lattice of this superalgebra 

1717: is $\Gamma^{1,1}\oplus E_8(-1)$ (i.e. the cohomology lattice of the Enriques surface), and 

1718: the simple roots  are the positive, norm $0$ vectors. Each simple root appears also as a superroot, both with multiplicity $8$, and 

1719: this is why the product of (\ref{def-Phi}) has a ``supersymmetric" structure: the numerator is a trace over fermionic degrees of freedom, while the 

1720: denominator traces over bosonic degrees of freedom. Both have the same multiplicity $2c_1 (r^2)$. In addition, the fact that $c_1(-1)=0$ is 

1721: equivalent to the absence of tachyons in the spectrum. 

1722: 

1723: %Mathematically, this leads to the absence of real roots in the superalgebra (all roots are imaginary). 

1724: 

1725: We will now write down a formula for the total partition function of topological string theory, restricted to the fiber, and we will show that it preserves the 

1726: structure found by Borcherds for (\ref{def-Phi}). As a first step, we define a generating functional $\xi(q,g_s)$ closely related to (\ref{defpg}),

1727: %

1728: \be

1729:   \xi(q,g_s)= \prod_{n=1}^{\infty} 

1730:  {(1-q^n)^2 \over 1-2 q^n \cos \, g_s + q^{2n}}.

1731:  \ee

1732:  %

1733: We have the identity

1734: %

1735: \be

1736:  \sum_{g=0}^{\infty} \CP_g(q) g_s^{2g-2} =\biggl( 2 \sin \, {g_s \over 2} \biggr)^{-2} \xi^2(q, g_s), 

1737:  \ee

1738: %

1739: Let us now define the Enriques degeneracies $\Omega_E(r, \ell)$ as 

1740: %

1741: \be

1742: \label{omegae}

1743: \sum_{r,\ell} 8 \Omega_E(r,\ell)q^{r^2} q_s^{ \ell} ={2\over  (q_s^{1\over 4} - q_s^{-{1\over 4}})^2} {1\over \eta^{12}(2\tau)} (\xi^2(q, g_s/2)-\xi^2(-q, g_s/2)), 

1744: \ee

1745: %

1746: where

1747: %

1748: \be

1749: q_s=\re^{ \ri g_s}

1750: \ee

1751: %

1752: The r.h.s. of (\ref{omegae}) only involves {\it integer} powers of $q_s^{\pm 1}$. We can collect the Enriques degeneracies in the generating polynomials

1753: %

1754: \be

1755: \Omega_n(z) =\sum_{r^2=2n,\ell\ge 0} \Omega_E(r,\ell) z^{\ell},

1756: \ee

1757: %

1758: which are of degree $n$ in $z$. We have for the first few:

1759: %

1760: \be

1761: \ba

1762: \Omega_0(z)&=1, \\

1763: \Omega_1(z) &=12 + 2z,\\

1764: \Omega_2(z)&= 90 + 24 z + 3 z^2,\\

1765: \Omega_3(z)&=520 + 180 z + 36 z^2 + 4 z^3,\\

1766: \Omega_4(z)&=2538 + 1040 z+ 270 z^2 + 48 z^3+ 5 z^4,\\

1767: \Omega_5(z)&=10944 + 5070 z + 1560 z^2 + 360 z^3 + 60 z^4 + 6 z^5.

1768: \ea

1769: \ee

1770: %

1771: Notice that the constant terms of $\Omega_n(z)$ are closely related to the Euler characteristics of the Hilbert schemes of the Enriques surface, but there are 

1772: ``deviations" which become more and more important as the degree increases.  Finally, notice that 

1773: %

1774: \be

1775: \sum_{\ell} \Omega_E(r,\ell)q^{r^2} q_s^{  \ell } =\Omega_n(q_s) + \Omega_n(q_s^{-1}) -\Omega_n(0).

1776: \ee

1777: %

1778: We now define 

1779: %

1780: \be

1781: F_E=\sum_{g=1}^{\infty} g_s^{2g-2}\CF^{(g)}_E(t), \qquad Z_E =\re^{-2 F_E}.

1782: \ee

1783: %

1784: Notice that, as $g_s\rightarrow 0$, $Z_E$ is precisely the Borcherds product $\Phi(t)$. It is now an easy exercise to 

1785: evaluate it for finite $g_s$ from (\ref{gr}), and we find

1786: %

1787: \be

1788: \label{allgproduct}

1789: Z_E(g_s,t)=\prod_{r,\ell} \biggl( { 1- q_s^{\ell} \re^{ -r\cdot t} \over 1+q_s^{\ell} \re^{  -r\cdot t} }\biggr)^{8 \Omega_E(r,\ell)}.

1790: \ee

1791: %

1792: As in the $g=1$ case, (\ref{allgproduct}) has a supersymmetric structure, with the same degeneracies for 

1793: fermionic and bosonic states. This formula in fact suggests the existence of a superalgebra structure for the 

1794: all--genus result as well. By including $g_s$ we have extended the lattice to

1795: %

1796: \be

1797: \Gamma^{1,1}\oplus E_8(-1) \rightarrow \Gamma^{1,1}\oplus E_8(-1)\oplus \IZ

1798: \ee

1799: %

1800: which is reminiscent of the growth of an eleven--dimensional direction associated to the string coupling constant. The fact that 

1801: the all--genus heterotic results seem to lead to an extra direction in the heterotic lattice has been pointed out in \cite{d,kawai}. 

1802:  It would be very interesting to see if there is indeed a superalgebra associated to the all--genus result (\ref{allgproduct}). If this was the case,  the 

1803:  quantities $8 \Omega_E(r,\ell)$ would correspond to root multiplicities. 

1804:  

1805: Finally, we mention that according to the conjecture in \cite{mnop} and the results of \cite{mp}, (\ref{allgproduct}) is 

1806: essentially the generating functional of an infinite family of Donaldson--Thomas invariants on the 

1807: Enriques surface (written already in the right variables). Such product formulas for $Z$ 

1808: exist generically if the latter is expressed in terms of of Gopakumar-Vafa invariants~\cite{Klemm:2004km}.

1809: Our comments above indicate that  the Donaldson--Thomas theory on this manifold has a highly nontrivial algebraic 

1810: structure (see section 3.2.6 in \cite{mp} for a  related observation). 

1811: 

1812: 

1813: 

1814: 

1815: \subsection{Automorphic forms \label{aut_prop}}

1816: 

1817: The free energies $F^{(g)}_E (t, \bar t)$ on the fiber turn out to be automorphic forms on the coset space $\CN_8$. 

1818: Here we will study in some detail automorphic forms on the space $\CN_s$. We will say that

1819: a function on the moduli space $\CN_s$ is {\it automorphic} 

1820: if it has well--defined transformation properties under the discrete subgroup $ O(s+2,2;\IZ)$.

1821: 

1822: The transformation properties are easier to understand if we consider explicit generators of 

1823: the symmetry group. We consider the explicit parametrization of the coset space (\ref{coset}) induced by a reduction 

1824: %

1825: \be

1826: \Gamma^{s+2,2} =\Gamma^{s+1,1} \oplus \Gamma^{1,1}, 

1827: \ee

1828: %

1829: and let $t \in \IC^{s+1,1}$ be the vector of complex coordinates parametrizing the coset. Our conventions are such that 

1830: $t$ has {\it positive} real part.

1831: For an element $t^a \in \IC^{s+1,1}$ we define the inner product 

1832: %

1833: \be

1834: t^2=\tfrac{1}{2} C_{ab} t^a t^b, 

1835: \ee

1836: %

1837: where $C_{ab}$ is the intersection matrix. 

1838: 

1839: The generators of the symmetry group are taken to be \cite{hm}:

1840: %

1841: \begin{itemize}

1842: \item $t \mapsto t+2\pi i \lambda$, $\lambda \in \Gamma^{s+1,1}$.

1843: 

1844: \item $ t \mapsto w(t)$, $w\in O(s+1,1;\IZ)$.

1845: 

1846: \item The automorphic analog of an S--duality transformation

1847: %

1848: \be

1849: \label{saut}

1850: t^a \quad  \mapsto\quad \tilde t^a={ t^a \over t^2}.

1851: \ee

1852: %

1853: \end{itemize}

1854: 

1855: We say that a function $\mf(t)$ is an {\it automorphic function of weight $k$} if it is invariant under the first two transformations above, and 

1856: if under (\ref{saut}), 

1857: it behaves as follows:

1858: %

1859: \be

1860: \mf_k(\tilde t)= t^{2 k} \mf_k(t).

1861: \ee

1862: %

1863: We can also have automorphic forms of weight $(k, \bar k)$ which transform as

1864: %

1865: \be \label{def-Phi_kk}

1866: \mf_{k,\bar k} (\tilde t)= t^{2k}\ \bar t^{\, 2 \bar k} \mf_{k,\bar k} (t).

1867: \ee

1868: %

1869: Although we have not indicated it explicitly, these functions might have a non-holomorphic dependence on $\bar t$. 

1870: Automorphic forms are in general non-holomorphic. Some automorphic forms are meromorphic (they have 

1871: poles at divisors). If they do not have poles, they are 

1872: called {\it holomorphic}. 

1873: 

1874: Notice that (\ref{saut}) transforms the metric $Y=(t+\bar t)^2$ on the ``upper half plane" as follows:

1875: \beq \label{trans-Y}

1876: Y \ \mapsto \  t^{-2}\ \bar t^{-2} Y\ .

1877: \ee

1878: Following the definition \eqref{def-Phi_kk} this identifies $Y$ as an automorphic form of weight $(-1, -1)$.

1879: Recalling the form of the K\"ahler potential for the classical moduli space \eqref{def-KY} this is nothing but a K\"ahler transformation \cite{cardoso}

1880: \beq \label{transK}

1881:    K\ \mapsto\ K+\log t^2 +\log \bar t^2\ .

1882: \eeq

1883: %

1884: in special coordinates where $X^0=1$.  

1885: Note that, if we keep $X^0$, this shift can be absorbed by the transformation of $X^0$

1886: %

1887: \beq \label{trans-X0}

1888:    X^0\ \mapsto\ t^2 X^0\ .

1889: \eeq

1890: %

1891: This can be traced back to the fact that $K$ as given in \eqref{KpotII} is a scalar under the full 

1892: symplectic group.

1893: 

1894:  In order to understand how the automorphic properties mix with taking derivatives, it is useful to derive the Jacobian $J_a^b$ of the change of 

1895:  coordinates (\ref{saut}). We immediately find, 

1896:  %

1897:  \be \label{def-J}

1898: {\partial \tilde t^a \over \partial t^b} \equiv  (J^{-1})^a_{b} = {1\over t^4} \Bigl( \delta^a_{~b} t^2 - t^a C_{be}t^e \Bigr)\ , \qquad \qquad 

1899: {\partial t^a \over \partial \tilde t^b} =J^a_{b} =\delta^a_{~b} t^2 - t^a C_{be}t^e\ .

1900:  \ee

1901: Notice that $J_a^b$ obeys the following useful identities  

1902: %

1903: \be

1904: \label{jcont}

1905:  J_a^b = t^4  (J^{-1})^b_{a}\ , \quad  \qquad C_{ab}=t^{-4} C_{cd} J^c_a J^d_b \ ,\quad   \qquad  C^{ab}J_a^c J_b^d=t^4 C^{cd}\ .

1906: \ee

1907: %

1908: Let us now assume that $\mf$ is an automorphic form of weight $(k,0)$. We want to determine the transformation 

1909: behavior of $D_a\mf$ and $D_a D_b \mf$ under the dualities \eqref{saut}. $D_a$ are here the derivatives covariant both with respect to Christoffel connection and the canonical connection on the vacuum bundle $\CL$, as introduced in section \ref{specialE}. 

1910: Therefore,

1911: \be

1912: D_a \mf =(\partial_a - k K_a)\mf\ . 

1913: \ee

1914: %

1915: Notice that, since $K$ transforms as given in \eqref{transK}, its first derivative $K_a$ shifts as

1916: \be

1917:     K_a\ \mapsto\ J_a^b \big(K_b + t^{-2} C_{bc} t^c \big)\ .

1918: \ee

1919: Combining this with the transformation of the automorphic form $\mf$ itself 

1920: we conclude 

1921: \beq \label{trans_D_aPhi}

1922:    D_a \mf \ \mapsto \  t^{2k} J^b_a D_b \mf\ .

1923: \eeq

1924: Similarly, we show that the second derivative of $\mf$ transforms as 

1925: \beq \label{trans_D_aD_bPhi}

1926:    D_b D_a \mf\ \mapsto\ t^{2k} J_d^b J_a^c D_b D_c \mf\ ,

1927: \eeq

1928: where we have used that the Christoffel symbols in the second connection 

1929: transform as 

1930: \beq

1931:      J^d_b \partial_d J^c_a - \widetilde \Gamma^d_{ba} J_a^c = \Gamma^d_{ba} J_a^c\ . 

1932: \eeq

1933: Hence, we have shown that the covariant derivatives $D_a$ of $\mf$ 

1934: transform with a factor $t^{2k}$ but are also rotated by the Jacobian $J^a_b$ 

1935: containing another factor of $t^2$. 

1936: Note however, that we can easily obtain automorphic forms 

1937: containing the derivatives $D_a \mf$. More precisely, if $\mf$ and $\mf'$ 

1938: are automorphic forms of weight $(k,0)$ and $(k',0)$ we find by 

1939: using \eqref{jcont} that 

1940: % 

1941: \beq \label{aut-forms}

1942:    C^{ab} D_a D_b \mf\ , \qquad \quad C^{ab} D_a \mf D_b \mf'

1943: \eeq 

1944: %

1945: are automorphic forms of weight $k+2$ and $k+k'+2$ respectively. Such 

1946: automorphic combinations arise in the derivation of all $F^{(g)}(S, \bar S, t, \bar t)$, $g>1$.

1947: More precisely, we will argue in the next sections that as function of $t, \bar t$, $F^{(g)}(S, \bar S, t, \bar t)$ itself is 

1948: an automorphic form of weight $(2g-2,0)$ such that

1949: %

1950: \beq \label{trans-Fg}

1951:    F^{(g)}\ \mapsto\ t^{4g-4} F^{(g)}\qquad \text{for} \quad g>1 \ .

1952: \eeq

1953: %

1954: An important example of an automorphic form is the heterotic integral (\ref{hetint}). It is 

1955: easy to show from the properties of the Narain--Siegel theta function that it has 

1956: weight $(2g-2,0)$. Since this integral gives the fiber limit $F^{(g)}_E$, we obtain a check of the general 

1957: property (\ref{trans-Fg}) from heterotic/type II duality. Note that it is straightforward to define

1958: amplitudes $F^{(g)}$ invariant under automorphic transformations by  

1959: \beq \label{inv_comb}

1960:    (X^{0})^{2-2g}\, F^{(g)}\ .

1961: \eeq

1962: The invariance of this combination is readily checked by using \eqref{trans-X0} and \eqref{trans-Fg}. 

1963: The expressions \eqref{inv_comb} are shown to be invariant under the full target space symmetry group

1964: $Sl(2,\bbZ)\times O(10,2)$. They are the direct analogs of the invariant free energies encountered in the Seiberg-Witten example in 

1965: section \ref{Seiberg-Witten}.

1966: 

1967: 

1968: 

1969: A particularly important and simple example occurs at $g=1$. Since $F^{(1)}_E$ is invariant, one 

1970: deduces from (\ref{feoneanti}) and \eqref{trans-Y} that 

1971: $\Phi(t)$ is an automorphic form of weight $(4,0)$ i.e.

1972: %

1973: \beq \label{trans-Phi}

1974:   \Phi(\tilde t) = t^{8}\ \Phi(t)\ , \qquad \qquad \tilde t^a = \frac{t^a}{ t^2}\ .

1975: \eeq

1976: %

1977: One can also show that $\Phi(t)$ is holomorphic. This is proved in \cite{borcherds}, and it 

1978: is in fact a consequence of the regularity of $\CF^{(g)}_E (t)$ at the singular 

1979: locus (\ref{slocusone}), which will be discussed in more detail in section \ref{boundaries}. In addition, $\Phi(t)$ is what is called a 

1980: singular automorphic form (see \cite{binfinite}, section 3, for a definition). Singular automorphic forms are known to 

1981: satisfy a wave equation 

1982: %

1983: \be

1984: C^{ab} {\partial^2 \over \partial t^a \partial t^b} \Phi(t)=0. 

1985: \ee

1986: %

1987: Equivalently, they have Fourier expansions involving only vectors of zero norm. It follows that $\CF^{(1)}_E(t)$ satisfies

1988: %

1989: \be

1990: \label{wavefone}

1991: C^{ab}\partial_a \partial_{b} \CF^{(1)}_E= 2C^{ab} \partial_a \CF^{(1)}_E\, \partial_b \CF^{(1)}_E\ .

1992: \ee

1993: %

1994: This is equivalent to the recursive relation found in \cite{mp} for genus one invariants on the fiber, and proves that the expression for $ \CF^{(1)}_E(t)$ 

1995: obtained in \cite{km} agrees with the Gromov--Witten calculation of \cite{mp}. 

1996: 

1997: 

1998: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1999: %%%%%%%%

2000: %%%%%%%%     Section 5

2001: %%%%%%%%

2002: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2003: 

2004: 

2005: \section{Direct Integration on the Enriques Calabi-Yau}\label{sec:diE}

2006: 

2007: In this section we illustrate the power of the method of direct integration by 

2008: studying the topological string amplitudes $F^{(g)}$ on the Enriques 

2009: Calabi-Yau. Our approach will follow and generalize the strategy 

2010: developed for the Seiberg-Witten example in section \ref{Seiberg-Witten}.

2011: To begin with, we perform a direct integration along the

2012: $\mathbb{T}^2-$base in section \ref{simple}. Using the fiber results 

2013: obtained in the previous section as additional input, the first 

2014: six free energies $F^{(g)}$ can be determined in a closed form. 

2015: We then present a more general formalism combining direct integration in base and fiber directions. In section \ref{propagators_auto}, we introduce 

2016: the relevant holomorphic and non-holomorphic $O(10,2,\bbZ)$ forms.

2017: A closed recursive expression for $F^{(g)}$ will be derived in section \ref{direct_fiber_base}. 

2018: It determines the $F^{(g)}$ up to a holomorphic ambiguity and we will briefly discuss 

2019: possible boundary conditions in section \ref{boundaries}. Finally, in section 

2020: \ref{red-model} we consider a reduced Enriques  

2021: model with three parameters only, which was already studied in \cite{km}. This model has the advantage that the mirror map 

2022: can be determined explicitly. 

2023: We also study in more 

2024: detail the boundary conditions (such as the gap condition), which 

2025: lead to valuable conclusions also applying to the full model.

2026: 

2027: 

2028: 

2029: \subsection{A simple direct integration and $F^{(g)}$ to genus six \label{simple}}

2030: 

2031: Let us now perform the direct integration along the $\mathbb{T}^2$ base and derive the first few amplitudes $F^{(g)}$.

2032: In order to do that we carefully keep track of their dependence of on the base direction $S,\bar S$.  

2033: As in the case of Seiberg--Witten theory studied in section 3, it is easy to see from the 

2034: structure of the holomorphic anomaly equations that the only antiholomorphic dependence of 

2035: $F^{(g)}$ on $\bar S$ appears through $\widehat E_2(S, \bar S)$. By taking derivatives with respect to $S$ we will also generate 

2036: in the holomorphic anomaly equations the modular forms $E_4(S), E_6(S)$, and by keeping track of the modular weight one immediately finds 

2037: that $F^{(g)}$ is an element of weight $2g-2$ in the ring 

2038: generated by

2039: %

2040: \be

2041: \label{genS}

2042:  \widehat E_2(S, \bar S), \quad E_4 (S), \quad E_6(S)\ .

2043: \ee

2044: %

2045: Our only assumption here is that the holomorphic ambiguity for $F^{(g)}$ is also a modular form of weight $2g-2$ in this ring. This assumption

2046: (as well as the details of the direct integration) 

2047: can be checked in a highly nontrivial way by comparing the resulting expressions to the field theory limit in the $N_f=4$ locus of \figref{modulispace}. This 

2048: check will be performed in section \ref{sec:ftlim}. 

2049: 

2050: To perform the direct integration 

2051: let us first rewrite the holomorphic anomaly equation for the base direction $\bar S$. 

2052: The general expression (\ref{rec_Fg}) reduces to

2053: %

2054: \be

2055: \partial_{\bar S}F^{(g)}=- \frac{1}{2}{C^{ab} \over (S+\bar S)^2} \Big(D_a D_b F^{(g-1)}+ \sum_{r=1}^{g-1}D_a F^{(r)}  D_b F^{(g-r)} \Big)\ .

2056: \ee

2057: %

2058: We now convert the derivative $\partial_{\bar S}$ into a derivative with respect to $\widehat E_2$.

2059: The definition of $\widehat E_2$ was already given in \eqref{hatE2}. Since we now consider an expansion in $q_S=e^{-S}$ 

2060: it takes the form

2061: %

2062: \be \label{def-widhatE2}

2063: \widehat E_2 (S, \bar S)=-\frac{12}{S+\bar S}+E_2(S)\ .

2064: \ee

2065: %

2066: Using the above assumption that the dependence of $F^{(g)}$ on $\bar S$ is only through this quantity, we can rewrite the 

2067: anomaly equation as

2068: %

2069: \be

2070: \label{partialhol}

2071: {\partial F^{(g)} \over \partial \widehat E_2}=- \tfrac{1}{24} C^{ab} \Big(D_a D_b F^{(g-1)}+ \sum_{r=1}^{g-1}D_a F^{(r)}  D_b F^{(g-r)} \Big)\ .

2072: \ee

2073: %

2074: Here the covariant derivatives $D_a$ are only taken with respect to the fiber directions and do not depend on 

2075: the base due to the simple special geometry of the Enriques Calabi-Yau. This implies that all dependence  

2076: on $\widehat E_2$ arises directly through the $F^{(r)}$.

2077: We thus expand $F^{(g)}$ in powers of $\widehat E_2$ by writing 

2078: %

2079: \be

2080:     \label{fgexp}

2081: F^{(g)}=\sum_{k=0}^{g-1} \widehat E_2^k (S,\bar S)\ \coeff^{(g)}_{k}\ , \qquad \quad g> 1\ .

2082: \ee

2083: %

2084: We see that (\ref{partialhol}) determines all the coefficients $c^{(g)}_{ k}$ for $k=1, \ldots, g-1$ in terms of quantities at lower genera. Explicitly, we have 

2085: the solution

2086: %

2087: \be

2088: \label{cksol}

2089: c^{(g)}_{ k}=- \tfrac{1}{24 k} C^{ab}  \Big(D_a D_b c^{(g-1)}_{ k-1} + \sum_{r=1}^{g-1}\sum_{l+m=k-1} D_a c^{(r)}_{l} D_b c^{(g-r)}_{ m}\Big)\ ,

2090: \ee

2091: %

2092: where we have set

2093: %

2094: \be \label{def-F1coeff}

2095:    \coeff^{(1)}_{0}=F^{(1)}\ , \qquad  \qquad \coeff^{(1)}_{i}=0\ ,\qquad i \neq 0\ .

2096:  \ee

2097:  %

2098: The $\widehat E_2$-independent term $c^{(g)}_{0}$ arises as an integration constant and hence 

2099: cannot be determined by the holomorphic anomaly equation. 

2100: However, given our assumptions, we can fix it up to genus 6 as follows. Let us 

2101: denote the coefficients in the fiber limit by

2102: %

2103: \be

2104:    c^{(g)}_{E|\, k}=\lim_{S, \bar S \rightarrow \infty} c^{(g)}_{k}.

2105: \ee

2106: %

2107: By also taking the fiber limit of (\ref{fgexp}) we find

2108: %

2109: \be

2110: \label{sumfiber}

2111: \sum_{k=0}^{g-1}  c^{(g)}_{E|\, k}=F^{(g)}_E(t,\bar t).

2112: \ee

2113: %

2114: The free energies $F^{(g)}_E(t,\bar t)$ are known from the  heterotic computation and given in \eqref{Fgantihol}. Together with the fact that 

2115: all $c^{(g)}_{E|\, k}$ for $k\ge 1$ are uniquely determined  by the direct integration we can use \eqref{sumfiber} to derive $c^{(g)}_{E|\, 0}$ i.e.~the fiber limit of the integration 

2116: constant. But the condition that $c^{(g)}_{0}$ is a modular form in the ring generated by (\ref{genS}) and does not involve $\widehat E_2$ fixes it uniquely in terms 

2117: of $c^{(g)}_{E|\, 0}$ as 

2118: %

2119: \be

2120: \ba \label{fix_c}

2121:  c^{(2)}_{0}&= 0\ ,   \qquad \quad &

2122:  c^{(3)}_{0}&=c^{(3)}_{E|\, 0}\  E_4\ ,  \qquad \quad &

2123:  c^{(4)}_{0}&= c^{(4)}_{E|\, 0}\  E_6\ , \\

2124:  c^{(5)}_{0}&= c^{(5)}_{E|\, 0}\ E^2_4\ ,  \qquad \quad & 

2125:  c^{(6)}_{0}&= c^{(6)}_{E|\, 0} \  E_4 \,E_6\ ,

2126:  \ea

2127:  \ee

2128:  %

2129:  where $E_4(S)$ and $E_6(S)$ are the two holomorphic generators in \eqref{genS}. This can be checked 

2130:  by noting that the definition \eqref{geneis} of the Eisenstein series implies that

2131:  \beq \label{Elimit}

2132:     E_2\ ,\ \ E_4\ , \ \ E_6\quad \rightarrow\quad 1\ , 

2133:  \eeq

2134:  in the fiber limit $S,\bar S\rightarrow \infty$.

2135:  For $g\ge 7$, the number of possible modular forms is greater than one and $c^{(g)}_{E|\, 0}$ is no longer uniquely determined 

2136:  in terms of its fiber limit. For example, at genus seven $c^{(7)}_0$ can contain terms proportional to 

2137:  $E_4^3$ as well as $E_6^2$.

2138:  

2139:  Let us now write down some explicit formula for lower genera. For $g=2$ we find, 

2140:  %

2141: \be

2142: F^{(2)}(S, \bar S,t,\bar t) =\widehat E_2(S, \bar S)\ c^{(2)}_{1}\ , 

2143: \ee

2144: %

2145: where we use \eqref{fix_c} and apply (\ref{cksol}) to derive

2146: %

2147: \be

2148: \label{f2sol}

2149:  c^{(2)}_{1}= -\tfrac{1}{24} C^{ab}  \left(D_a D_b F_E^{(1)}+ D_a F_E^{(1)} D_b F_E^{(1)} \right). 

2150: \ee

2151:  %

2152: Consistency of the fiber limit requires that

2153: %

2154: %\be

2155: $c^{(2)}_{1}=F_E^{(2)}(t, \bar t)$. 

2156: %\ee

2157: %

2158: This can be checked by using the heterotic expression (\ref{Fgantihol}) for $F_E^{(2)}(t, \bar t)$, the property (\ref{wavefone}), and the identity \cite{km}

2159: %

2160: \be

2161:  \label{identity}

2162:  \CF^{(2)}_E=-\tfrac{1}{16} C^{ab}\partial_a \partial_b \CF^{(1)}_E\ ,

2163:   \ee

2164:  %

2165: which follows directly from (\ref{hetint}). In the holomorphic limit we find, 

2166: %

2167: \be 

2168: \CF^{(2)}(S,t)=E_2(S) \CF^{(2)}_E(t)\ , 

2169: \ee

2170: %

2171: in agreement with the results of \cite{km,mp}. In the following sections we will also need a slightly different 

2172: form of $F^{(2)}$. Namely, it is straightforward to apply \eqref{wavefone} to write

2173: \beq \label{result_F2}

2174:    F^{(2)} = -\tfrac{1}{8} C^{ab} \partial_a F^{(1)} \partial_b F^{(1)}\ .

2175: \eeq

2176: 

2177: Let us now consider the $g=3$ case. The amplitude $F^{(3)}$ can be expanded by using \eqref{fgexp} and

2178: \eqref{fix_c} as 

2179: %

2180: \be

2181: F^{(3)}=\widehat E_2^2(S, \bar S)\, c^{(3)}_{2} + E_4(S)\, c^{(3)}_{E|\, 0}\ .

2182: \ee

2183: %

2184: Using the result of the direct integration (\ref{cksol}) we obtain

2185: %

2186: \be

2187:   c^{(3)}_{2}=- \tfrac{1}{48} C^{ab} \Big(   D_aD_b F^{(2)}_E + 2 D_a F^{(2)}_E D_b F^{(1)}_E\Big)\ .

2188: \ee

2189: %

2190: To determine $c^{(3)}_{E|\, 0}$ we use (\ref{sumfiber}), which gives 

2191: %

2192: \be

2193:  c^{(3)}_{2} +  c^{(3)}_{E|\,0}=F^{(3)}_E(t, \bar t)\ . 

2194: \ee

2195: %

2196: On the other hand,  one finds that

2197: %

2198: \be

2199: \label{antiholfthree}

2200: F^{(3)}_E(t, \bar t)=-\tfrac{1}{24} C^{ab} D_a D_b F^{(2)}_E.

2201: \ee

2202: %

2203: This can be derived in the holomorphic limit by using (\ref{hetint}), and it is similar to (\ref{identity}). The antiholomorphic part can be 

2204: checked with (\ref{Fgantihol}). Using all this, we finally obtain the following simple expression for 

2205: $F^{(3)}(S, \bar S, t, \bar t)$,

2206: %

2207: \be

2208: F^{(3)}=-\tfrac{1}{24} E_4\, C^{ab}  D_a D_b F^{(2)}_E  

2209:  - \tfrac{1}{48} (\widehat E_2^2-E_4) 

2210: C^{ab} \big(  D_a D_b F^{(2)}_E + 2 D_a F^{(2)}_E D_b F^{(1)}_E \big)\ ,

2211: \ee

2212: %

2213: with the holomorphic limit

2214: %

2215: \be

2216: \label{finalfthree}

2217: \CF^{(3)}(S,t)= -\tfrac{1}{24}E_4\, C^{ab}  \partial_a \partial_b \CF^{(2)}_E- \tfrac{1}{48} (E_2^2-E_4)

2218: C^{ab} \big(  \partial_a \partial_b \CF^{(2)}_E + 2 \partial_a \CF^{(2)}_E \partial_b \CF^{(1)}_E\big)\ .

2219: \ee

2220: %

2221: Note that the second term in these expressions vanishes identically in the fiber limit where $E_2,E_4\rightarrow 1$. 

2222: As we will discuss in more detail in section \ref{boundaries}, this is the first $F^{(g)}$ where the inclusion of the base yields

2223: a behavior near the singular loci that differs significantly from the fiber limit.\\

2224: {}Explicit calculations at genus $4$ proceed in the same way. Modular invariance with respect to $S$ gives

2225: %

2226: \be

2227: F^{(4)}(S, \bar S, t, \bar t)=\widehat E_2^3 \ c^{(4)}_{E|\, 3} + \widehat E_2 E_4\ c^{(4)}_{E|\, 1} + E_6\ c^{(4)}_{E|\, 0}.

2228: \ee

2229: %

2230: Once again, the general equation (\ref{gencksol}) allows us to determine the coefficients as

2231: %

2232: \be \label{c4}

2233: \ba

2234:  c^{(4)}_{E|\, 3}&=- \tfrac{1}{72} C^{ab} \big( D_aD_b c^{(3)}_{E|\,2} + 2 D_a F^{(1)}_E D_b c^{(3)}_{E|\, 2} + D_a F^{(2)}_ED_bF^{(2)}_E \big)\ , \\ 

2235:  c^{(4)}_{E|\, 1}&=-\tfrac{1}{24} C^{ab} \big( D_aD_b  c^{(3)}_{E|\, 0}  + 2 D_a  F^{(1)}_E D_b  c^{(3)}_{E|\, 0} \big)\ .

2236: \ea

2237: \ee

2238: % 

2239: The ambiguity $c^{(4)}_{E|\, 0}$ is again determined by the heterotic computation in the fiber limit. 

2240: More precisely, one specializes \eqref{sumfiber} to 

2241: %

2242: \be

2243:   c^{(4)}_{E|\, 0}+ c^{(4)}_{E|\, 1}+ c^{(4)}_{E|\,3} = F^{(4)}_E(t, \bar t)\ ,

2244: \ee

2245: %

2246: and solves for $c^{(4)}_{E|\, 0}$ by inserting the fiber result \eqref{Fgantihol}. This determines 

2247: the free energy $F^{(4)}$. A similar analysis also applies to $g=5,6$. As already discussed 

2248: above, the main obstacle that has to be overcome in order to proceed to higher genus is the difficulty to fix the ambiguities $c^{(g)}_0$. 

2249: We will discuss possible additional boundary conditions in sections \ref{boundaries}, \ref{red-model}

2250: and \ref{sec:ftlim}.  

2251:  

2252: 

2253: 

2254: \subsection{Propagators and homolomorphic automorphic forms \label{propagators_auto}}

2255: 

2256: In the previous section we calculated the first free energies $F^{(g)}$ 

2257: by a direct integration along the base direction. The results were expressed 

2258: in terms of the holomorphic fiber energies $\cF^{(g)}_E$, which are known 

2259: from heterotic-type II duality. Even though the results were rather compact and 

2260: transparent, the information we have extracted is somewhat partial, since 

2261: we have not used the holomorphic anomaly equations for the fiber moduli. In order to exploit the information they contain, we 

2262: will construct building blocks for the automorphic forms in the fiber which enable us to 

2263: perform the direct integration of the remaining holomorphic anomaly equations. Recall that we argued in the previous sections that 

2264: the almost holomorphic modular form 

2265: \beq \label{def-E2}

2266:    \widehat E_2(S,\bar S) = -\frac{12}{S+\bar S}+E_2(S)\ ,\qquad  \qquad E_2(S) = \partial_S \log \Phi\ ,

2267: \eeq

2268: contains all non-holomorphic dependence of $F^{(g)}$ along the base direction $S$. It will be 

2269: the task of this section to introduce the analog of $\widehat E_2$ for the fiber directions $t^a$. 

2270: Furthermore we will define the fiber analogs of the holomorphic modular forms 

2271: $E_4(S)$ and $E_6(S)$. This will lead us to the definition of a new class of holomorphic 

2272: automorphic forms of $O(10,2,\bbZ)$. Eventually, in section \ref{direct_fiber_base} 

2273: we will argue that a direct integration along the fiber direction allows us to 

2274: express all $F^{(g)}$ in terms of these almost holomorphic and holomorphic 

2275: forms of $O(10,2,\bbZ)$.

2276: 

2277: Let us now introduce the fiber analog of the almost holomorphic modular 

2278: form $\widehat E_2(S,\bar S)$. This can be done by recalling that 

2279: the genus one free energy $F^{(1)}$ is an invariant of the full 

2280: symmetry group $Sl(2,\bbZ)\times O(10,2,\bbZ)$ and hence its

2281: first derivatives transform in a particularly simple way. For the derivative with respect

2282: to $S$ one finds $\partial_S F^{(1)} = \frac12 \widehat E_2$. The derivative with 

2283: respect to $t^a$ we denote by $ \Delta^a = -\tfrac{1}{2} C^{a b} \partial_b F^{(1)}$ and evaluate

2284: \beq \label{def-Deltaa}

2285:      \Delta^a =\frac{t^a+\bar t^a}{Y}  + \epsilon^a(t)=\epsilon^a(t)-K_b(t) C^{ba}\ , \qquad \quad \epsilon^a(t) = \tfrac14  C^{ab} \partial_{t^b} \log \Phi \ ,

2286: \eeq

2287: where $Y=\frac12 C_{ab}(t+\bar t)^b (t+\bar t)^b$ and $\Phi$ is given in \eqref{def-Phi}. The function $\epsilon^a(t)$ is holomorphic in the coordinates 

2288: $t^a$ and is the fiber analog of $E_2(S)$, while $\Delta^a$ plays the role of $\widehat{E}_2$. To see this note that $\epsilon^a$ transforms with a 

2289: shift under the duality $t^a \mapsto t^a /t^2$:

2290: \beq \label{trans-epsilon}

2291:   \epsilon^a \quad \mapsto\quad  t^4 (J^{-1})^a_b (\epsilon^b + t^{-2} t^b)\ .

2292: \eeq

2293: This shift is precisely canceled by the shift of the non-holomorphic term in \eqref{def-Deltaa}

2294: such that $\Delta^a$ simply transforms as

2295: \beq \label{trans-Delta}

2296:    \Delta^a \quad \mapsto\quad  t^{4} (J^{-1})_b^a \Delta^b(t)\ .

2297: \eeq

2298: Note that $\widehat E_2$ and $\Delta^a$ are sufficient to parametrize all propagators $\hat \Delta^{ij},\hat \Delta^{i},\hat \Delta$

2299: introduced in \eqref{def-small-Delta}. Indeed, one has 

2300: \begin{align} \label{Enriques-props}

2301:   \hat \Delta^{a b} &=  -\tfrac{1}{12} C^{a b} \widehat E_2 \ ,\qquad

2302:   &&\hat \Delta^{a S} =  \Delta^a\ ,\\

2303:   \hat \Delta^S& =- \tfrac{1}{2} C_{a b} \Delta^{a} \Delta^{b}\ ,\qquad

2304:   &&\hat \Delta^a = \tfrac{1}{12} \widehat E_2 \Delta^a \ ,\qquad \quad

2305:  \hat \Delta = -\tfrac{1}{12}  \widehat E_2 C_{a b} \Delta^{a} \Delta^{b}\ .\nn

2306: \end{align}

2307: Using the explicit form of $\widehat E_2$ and $\Delta^a$ it is straightforward to check 

2308: that these propagators fulfill the defining conditions \eqref{def-small-Delta}.

2309: The fact that all $\hat \Delta-$propagators can be expressed as polynomials 

2310: in $\widehat E_2$ and $\Delta^a$ will be used in the next section to 

2311: argue that all non-holomorphic dependence of $F^{(g)}$ only arises through $\widehat E_2, \Delta^a$.

2312: However, we also have to extract the non-holomorphic dependence in the 

2313: covariant derivatives $D_a$ defined in \eqref{cov_D}. Following the logic 

2314: of section \ref{Seiberg-Witten} we will show that each derivative can be split into 

2315: a holomorphic covariant  derivative $\hat D_a$ plus holomorphic terms times the 

2316: propagators $\Delta^a$. As an important byproduct, the definition of $\hat D_a$ will

2317: also allow us to find an interesting construction of holomorphic automorphic forms.

2318: 

2319: Let us now construct a holomorphic covariant derivative $\hat D_a$, which 

2320: has the same properties as $D_a$ under automorphic 

2321: transformations \eqref{saut}. More precisely, given 

2322: an automorphic form $\mf$ of weight $k$

2323: we define its first derivative as 

2324: \beq\label{der_holE}

2325:     \hat D_a \mf \equiv \big(\partial_a - k C_{ab} \epsilon^b\big) \mf\ , 

2326: \eeq 

2327: where $\epsilon^a$ is defined in \eqref{def-Deltaa}, and note that $\hat D_a =D_a-kC_{ab}\Delta^b$. $\hat D_a$ can be viewed 

2328: as the analog of the Serre derivative \eqref{der_hol} for modular forms of subgroups of $Sl(2,\bbZ)$.

2329: It is not hard to check that it transforms under \eqref{saut} exactly as $D_a$. This 

2330: transformation property was given in \eqref{trans_D_aPhi}.

2331: Note however, that $\hat D_a$ maps holomorphic 

2332: forms into holomorphic forms, while $D_a$ contains an anti-holomorphic contribution.

2333: Moreover, by definition of $\epsilon^a$ one has

2334: \beq

2335:    \hat D_a \Phi(t) = 0\ ,

2336: \eeq 

2337: for the automorphic form $\Phi(t)$ given in \eqref{def-Phi}. In order to 

2338: evaluate second derivatives we need to introduce the holomorphic analog of 

2339: the Christoffel symbol in the definition \eqref{cov_D} of $D_k$.

2340: To do that, let us consider a section $ \mf_a$ which transforms as 

2341: $ \mf_a\, \mapsto\, t^{2k} J^b_a \mf_b$ 

2342: under the action \eqref{saut}.

2343: The covariant derivative is then defined to act as

2344: \beq \label{def-hatD}

2345:   \hat D_a  \mf_b = \big(\partial_a - k C_{ac} \epsilon^c\big) \mf_b - \hat \Gamma^{c}_{ab} \mf_c\ .

2346: \eeq

2347: Here we have included the holomorphic Christoffel symbol 

2348: \beq 

2349:    \hat \Gamma_{ab}^c= \hat \Gamma^{c}_{ab|d} \epsilon^d= \tfrac{1}{2} \hat C^{cd}\big(\partial_b \hat C_{da}+\partial_a \hat C_{db}-\partial_d \hat C_{ab} \big)\ ,   

2350: \eeq

2351: where $\hat \Gamma^b_{cd|a}$ is defined in \eqref{def-hatGamma} and related to $\Gamma^b_{cd}$ by $\Gamma^b_{cd}=\hat \Gamma^b_{cd|a}C^{ae}K_e$. We also have introduced the 

2352: holomorphic `metric' $\hat C_{ab}$. Explicitly, $\hat C_{ab}$ is defined as

2353: \beq

2354:   \hat C_{ab} = \Phi^{1/2} C_{ab}\ ,\qquad \hat C_{ab}\ \mapsto\ J^c_a J^d_b \hat C_{cd}\ ,

2355: \eeq

2356: where $\Phi$ is given in \eqref{def-Phi} and  we have also displayed the transformation behavior of $\hat C_{ab}$ under \eqref{saut}

2357: as inferred from \eqref{trans-Phi} and \eqref{jcont}. 

2358: Once again we evaluate the transformation behavior of $\hat D_a \mf_b$ under \eqref{saut}

2359: and finds the holomorphic analog of \eqref{trans_D_aD_bPhi}.

2360: It is now easy to show that every non-holomorphic 

2361: derivative $D_a$ can be split as 

2362: \beq \label{split_Dhol}

2363:    D_a \mf_b = \hat D_a \mf_b +k C_{ac} \Delta^c\, \mf_b+ \hat \Gamma^{c}_{ab|d} \Delta^d\, \mf_c\ .

2364: \eeq

2365: In other words, whenever $\mf_b$ is holomorphic the non-holomorphic 

2366: dependence in $D_a \mf_b$ arises through the propagators $\Delta^a$ 

2367: only.

2368: 

2369: 

2370:  Let us now discuss a second interesting application of the holomorphic 

2371:  covariant derivative $\hat D_a$. Namely, we will now show how it can be used to 

2372:  construct new holomorphic automorphic forms. To start with 

2373:  let us note that $\epsilon_a=C_{ab} \epsilon^b$ transforms in \eqref{trans-epsilon} similarly to 

2374:  a vector field. We can use this analogy and define a field strength

2375:  \beq \label{def-epsilon4}

2376:       \epsilon^{4}_{ab} = \partial_{a} \epsilon_b -\tfrac12 \hat \Gamma^{c}_{ab} \epsilon_c=\partial_{a} \epsilon_b -\epsilon_a\epsilon_b+C_{ab}\epsilon^2\ ,\qquad \qquad \epsilon_a=C_{ab} \epsilon^b\ ,

2377:  \eeq

2378: which transforms covariantly, $\epsilon_{ab}^4\, \mapsto\, J^c_a J^d_b \epsilon^4_{cd}$,

2379: under automorphic transformations \eqref{saut}. 

2380: Note that by using the wave-equation \eqref{wavefone} one shows that $\partial_a \epsilon^a = - 4 C_{ab} \epsilon^a \epsilon^b$

2381:  such that 

2382: \beq \label{Cepsilon4=0}

2383:   C^{ab}\epsilon_{ab}^4 = 0\ .

2384: \eeq

2385: Nevertheless, we can use $\epsilon^4_{ab}$ to construct holomorphic automorphic 

2386: forms. To do that, we define

2387: \beq\label{epsilon2k}

2388:   \epsilon^{2k}_{a_1\ldots a_{k}} = \hat D^{\phantom{4}}_{a_k} \ldots \hat D^{\phantom{4}}_{a_3} \epsilon^4_{a_2 a_1}\ ,

2389: \eeq

2390: which is shown to be totally symmetric in the indices.

2391: Holomorphic automorphic forms are now constructed by contraction with 

2392: $C^{ab}$. For example, forms of weight $4$ and $6$ are given by

2393: \bea \label{holfroms}

2394:   \text{weight}\ 4: & \qquad&  C^{ab} C^{cd}  \epsilon^4_{ac} \epsilon^4_{bd}\ ,\\

2395:    \text{weight}\ 6: & \qquad&  C^{ac} C^{be} C^{df}  \epsilon^4_{ab} \epsilon^4_{cd} \epsilon^4_{ef} \ ,\qquad 

2396:                 C^{ac} C^{be} C^{df}  \epsilon^{6}_{abd} \epsilon^6_{cef} \ .\nn

2397: \eea 

2398: It is tempting to conjecture that holomorphic automorphic forms of this type are sufficient to parametrize 

2399: the holomorphic ambiguity of $F^{(g)}$. The fact that there is no holomorphic weight $2$

2400: automorphic form of this type due to \eqref{Cepsilon4=0} matches nicely the fact that there is 

2401: no holomorphic ambiguity for $F^{(2)}$. Also the forms in \eqref{holfroms} can be shown to be sufficient 

2402: to parametrize the ambiguities of $F^{(3)}$ and $F^{(4)}$. This will be analyzed in further work.

2403: 

2404: 

2405: 

2406: 

2407: \subsection{Direct integration of the holomorphic anomaly \label{direct_fiber_base}}

2408: 

2409: We will now use the material developed in the previous section to perform the direct integration in both fiber 

2410: and base directions. This will allow us to give closed expressions which determine 

2411: the $F^{(g)}$ up to a holomorphic ambiguity. 

2412: To begin with, we show that each 

2413: $F^{(g)}$ can be written as 

2414: \beq \label{Eexp}

2415:   F^{(g)} = \sum_{k=0}^{g-1}\, \sum_{n=0}^{2g-2}\ \widehat E_2^k \Delta^{a_1} \ldots \Delta^{a_n} \coeff^{(g)}_{k\, |\, a_1 \ldots a_n} \ ,\qquad g>1

2416: \eeq

2417: where $\coeff^{(g)}_{k\, |\, a_1 \ldots a_n} $ are holomorphic functions of $S,\, t^a$ and all anti-holomorphic 

2418: dependence arises through the propagators $ \Delta^a$ and $\widehat E_2$ introduced in \eqref{def-E2} and \eqref{def-Deltaa}.

2419: Note that by using the transformation properties of $F^{(g)}$ and $\Delta^a$ given in \eqref{trans-Fg} and \eqref{trans-Delta}

2420: one infers that

2421: \beq

2422:    \coeff^{(g)}_{k\, |\, a_1 \ldots a_n} \ \mapsto\  t^{4g-4-4n} J^{b_1}_{a_1} \ldots J^{b_n}_{a_n}\, \coeff^{(g)}_{k\,|\,b_1 \ldots b_n}

2423: \eeq

2424: under automorphic transformations \eqref{saut}.

2425: 

2426: Let us now show that each $F^{(g)}$ for $g>1$ can indeed be written as \eqref{Eexp} 

2427: by using induction. 

2428: We first note that $F^{(2)}$ is of the form \eqref{Eexp}, 

2429: \beq \label{trans-k}

2430:   F^{(2)} = -\tfrac{1}2 \widehat E_2 C_{ab} \Delta^a \Delta^b\ ,

2431: \eeq

2432: as is immediately inferred from \eqref{result_F2} and \eqref{def-Deltaa}.

2433: So let us assume that \eqref{Eexp} is true for all $r<g$ and 

2434: show that this implies that \eqref{Eexp} is true for $g$. In order to do that 

2435: we use the Feynman graph expansion \eqref{Fgwithf} of $F^{(g)}$ \cite{bcov}, which states

2436: that each $F^{(g)}$ can be written as an expansion with propagators $\hat \Delta^{ij},\hat \Delta^i,\hat \Delta$

2437: and vertices $C^{(r)}_{i_1\ldots i_n}$ with $r<g$. We have already shown that the $\hat \Delta$-propagators 

2438: are polynomials in $\widehat E_2$ and $\Delta^a$ in \eqref{Enriques-props}. Hence, it remains to 

2439: show that also the vertices $C^{(r)}_{i_1\ldots i_n}$ are polynomials in $\widehat E_2$ and $\Delta^a$.

2440: By definition \eqref{C_prop} and our assertion, the vertices are defined as the covariant derivatives of 

2441: amplitudes $F^{(r)}$ of the form \eqref{Eexp}. Using \eqref{split_Dhol} each of these covariant derivatives 

2442: $D_a$ can be split into a holomorphic covariant derivative $\hat D_a$ and an expansion in $\Delta^a$. 

2443: So we only have to show that $\hat D_a \Delta^b$ admits again an expansion into $\Delta$'s.

2444: A straightforward computation shows that 

2445: \beq \label{der_Delta}

2446:   \hat D_a \Delta^b = C^{bd} \epsilon^4_{da} - \tfrac12 \hat \Gamma^b_{cd|a} \Delta^c \Delta^d\ ,

2447: \eeq

2448: where $\epsilon^4_{ab}$ and $ \hat \Gamma^b_{cd|a}$ are defined in \eqref{def-epsilon4} and \eqref{def-hatGamma}.

2449: Altogether one infers that all vertices and $\hat \Delta$-propagators are polynomial in $\Delta^a$ and 

2450: hence that $F^{(g)}$ is of the form \eqref{Eexp}.

2451: 

2452: 

2453: Having shown that every $F^{(g)}$ is of the form \eqref{Eexp} we will now derive a 

2454: closed expression for $F^{(g)}$ by direct integration of the holomorphic 

2455: anomaly equation \eqref{rec_Fg}.

2456: Applying the definition \eqref{def-small-Delta} of the propagators we can write 

2457: the holomorphic anomaly equation as

2458: \beq \label{rec_Fg1}

2459:    \partial_{\bar \imath} F^{(g)} = \tfrac12 \partial_{\bar \imath} \hat \Delta^{ik}

2460:    \Big(D_j D_k F^{(g-1)} + \sum_{r=1}^{g-1}D_j F^{(r)} D_k F^{(g-r)} \Big)  \ .

2461: \eeq

2462: This equation captures the anti-holomorphic derivatives $\partial_{\bar S} F^{(g)}$

2463: along the base as well as the derivative $\partial_{\bar a} F^{(g)}$ along the fiber 

2464: of the Enriques Calabi-Yau. Recall that the only non-vanishing propagators are 

2465: $\hat \Delta^{ab} =-\frac{1}{12} C^{ab}\widehat E_2$ and $\Delta^a =\hat \Delta^{aS}$.

2466: As we have shown, they contain all anti-holomorphic dependence

2467: such that we can rewrite \eqref{rec_Fg1} as

2468: \bea \label{directE}

2469:    \frac{\partial F^{(g)} }{\partial \widehat E_2}&=&  -\tfrac{1}{24}C^{ab}

2470:    \Big(D_a D_b F^{(g-1)} + \sum_{r=1}^{g-1}D_a F^{(r)} D_b F^{(g-r)} \Big)\ ,\\

2471:    \label{directDelta} \frac{\partial F^{(g)} }{\partial \Delta^a}&=& 

2472:    D_a D_S F^{(g-1)} + \sum_{r=1}^{g-1} D_a F^{(r)} D_S F^{(g-r)} \ . 

2473: \eea

2474: %

2475: As we have seen above, the first equation is already very powerful and can be integrated easily. We can write the 

2476: solution (\ref{cksol}) as 

2477: %

2478: \beq

2479: \label{gencksol}

2480:   F^{(g)}=  - \tfrac{1}{24} \sum_{k=1}^\infty \tfrac{1}{k} \widehat E_2^k C^{ab}

2481:     \Big(D_a D_b \coeff^{(g-1)}_{k-1}   

2482:    +\sum_{r=1}^{g-1} \sum_{l+m=k-1} D_a \coeff^{(r)}_{l} D_b \coeff^{(g-r)}_{m} \Big)+\coeff^{(g)}_{0}\ ,

2483: \eeq

2484: %

2485: where $\coeff^{(1)}_m$ is defined in \eqref{def-F1coeff}. Note that $\coeff^{(g)}_{0}(\Delta,S,t)$ 

2486: arises an integration constant of the $\widehat E_2$ integration and hence  

2487: can be a function of the propagators $\Delta^a$ but not 

2488: $\widehat E_2$. 

2489: 

2490: Let us now determine a second closed expression for $F^{(g)}$ by

2491: integrating the second anomaly equation \eqref{directDelta}. Since $F^{(1)}$

2492: is not of the form \eqref{Eexp} we first split off terms involving $F^{(1)}$.

2493: Inserting the definitions of the propagators $\Delta^a$ and $\widehat E_2$ we find for $g>2$ that

2494: \beq \label{delta-ano1}

2495:   \frac{\partial F^{(g)} }{\partial \Delta^a} = 

2496:   (D_S+ \tfrac{1}{2} \widehat E_2 )D_a  F^{(g-1)} -2C_{ac}\Delta^c D_S F^{(g-1)}

2497:   +\sum_{r=2}^{g-2} D_a F^{(r)} D_S F^{(g-r)}.

2498: \eeq

2499: To make the dependence on the propagators $\Delta^a$ explicit we 

2500: expand the covariant derivative $D_aF^{(g)}$. The covariant 

2501: derivative $D_a$ can be split into a holomorphic derivative $\hat D_a$ defined in \eqref{def-hatD} plus 

2502: a propagator expansion using \eqref{split_Dhol}. 

2503: Moreover, using the chain rule one rewrites 

2504: \beq \label{chain_dec}

2505:     \hat D_a = \hat d_a + (\hat D_a \Delta^b) \partial_{\Delta^b}\ ,

2506: \eeq

2507: where $\hat d_a$ is the covariant holomorphic derivative not acting on the 

2508: propagators, i.e.~we set 

2509: \beq

2510:   \hat d_a \big(\Delta^{a_1}\ldots \Delta^{a_n} \coeff_{a_1\ldots a_n}\big) =  \Delta^{a_1}\ldots \Delta^{a_n} \hat D_a \coeff_{a_1\ldots a_n}\ .

2511: \eeq

2512: Combining \eqref{split_Dhol}, \eqref{chain_dec} and \eqref{der_Delta} we immediately derive

2513: \bea \label{dFa}

2514:   D_a F^{(g)} &=& \Big[\hat d_a + \epsilon^4_{ac}  C^{cb} \partial_{\Delta^b} 

2515:              +(2g-2) C_{ad}\Delta^d - \tfrac{1}{2}  \hat \Gamma^b_{cd|a} \Delta^c \Delta^d \partial_{\Delta^b}\Big]F^{(g)}\ .

2516: \eea

2517: This expansion makes the dependence of $D_a$ on the propagators $\Delta^a$ explicit. 

2518: We note that the $\hat d_a$ term on the right-hand side of this expansion does not 

2519: change the number of propagators. The second term lowers the number of propagators by one, 

2520: while the two last terms raise the number of propagators by one.

2521: Inspecting the holomorphic anomaly equation we note that only 

2522: the first derivative along the fiber direction appears on the right-hand side of  \eqref{delta-ano1}. 

2523: Hence, at least for the integration of  \eqref{delta-ano1} it will not be necessary to evaluate 

2524: the second derivative $D_a D_b F^{(g)}$ as a propagator expansion.

2525: 

2526: 

2527: To integrate  expressions such as \eqref{dFa} for $D_a F^{(g)}$ we also need to keep track of the number 

2528: of propagators in the expansion of $F^{(g)}$. Therefore, we introduce the following 

2529: short-hand notation

2530: \beq \label{Fgincg}

2531:   F^{(g)} = \sum_n \coeff^{(g)}_{\ (n)}\ , \qquad \qquad \coeff ^{(g)}_{\ (n)} = \sum_{k=0}^{g-1} \widehat E_2^k \Delta^{a_1} \ldots \Delta^{a_n} \coeff ^{(g)}_{k\,|\,a_1 \ldots a_n}\ ,

2532: \eeq

2533: where each $c^{(g)}_{\ (n)}$ contains $n$ propagators $\Delta^a$.

2534: By counting the number of propagators one finds

2535: \beq

2536:   \int D_a F^{(g)} d\Delta^a =  \sum_n \Big\{ \tfrac{1}{n+1}  \Delta^a \hat d_a 

2537:             + \tfrac{1}{n}\Delta^a \epsilon^4_{ac} C^{cb} \partial_{\Delta^b}     

2538:             +  \tfrac{4g-4-n}{n+2} \Delta^2 \Big\}\, \coeff^{(g)}_{\ (n)}\ ,

2539: \eeq

2540: where as defined above $\Delta^2=\frac12 C_{ab}\Delta^a \Delta^b$.

2541: This integral together  with similar ones for the remaining terms in \eqref{delta-ano1} yields 

2542: a closed expression for $F^{(g)}$ of the form 

2543: \bea \label{general-direct-delta}

2544: F^{(g)} &=&  \big(D_S + \tfrac{1}{2}\widehat E_2) \sum_n \Big\{ \tfrac{1}{n+1} \Delta^a \hat d_a  + \tfrac{1}{n}\Delta^a \epsilon^4_{ac} C^{cb} \partial_{\Delta^b} 

2545:            +  \tfrac{4g-8-n}{n+2} \Delta^2 \Big\}\coeff^{(g-1)}_{\ (n)}\nn \\

2546:    && -  \sum_n \tfrac{4}{n+2} \Delta^2  D_S c^{(g-1)}_{\ (n)} 

2547:          + \sum_{r=2}^{g-2} \sum_n \sum_{k+l=n} D_S \coeff^{(g-r)}_{\ (l)} \Big\{ \tfrac{1}{n+1} \Delta^a \hat d_a\nn \\

2548: &&  + \tfrac{1}{n}\Delta^a \epsilon^4_{ac} C^{cb} \partial_{\Delta^b} 

2549:              +  \tfrac{4r-4-n}{n+2} \Delta^2 

2550:                 \Big\} \coeff^{(r)}_{\ (k)} 

2551:                +\coeff^{(g)}_{\ (0)}\ .

2552: \eea

2553: Here $\coeff^{(g)}_{\ (0)}(\widehat E_2,S,t)$ is the integration constant of the $\Delta^a$ integration and hence 

2554: can depend on $\widehat E_2$ but not on $\Delta^a$.

2555: 

2556: Before turning to the discussion of an explicit example, let us consider the 

2557: fiber limit of \eqref{general-direct-delta}. We therefore 

2558: apply \eqref{E-der} and \eqref{Elimit} to show that

2559: \beq  \label{DSF=0}

2560:  \lim_{S,\bar S\rightarrow \infty}  D_S F^{(g)} =0\ .

2561: \eeq

2562: We also denote by $c^{(g)}_{E\,(k)}$ the fiber limit of the coefficients $c^{(g)}_{\ (k)}$ in \eqref{Fgincg}.

2563: Inserting \eqref{DSF=0} into the formula \eqref{general-direct-delta} for direct integration along the fiber direction

2564: one finds

2565: \beq \label{general-direct-deltaE}

2566: F^{(g)}_E = \tfrac{1}{2} \sum_n \Big( \tfrac{1}{n+1} \Delta^a \hat d_a  + \tfrac{1}{n}\Delta^a \epsilon^4_{ac} C^{cb} \partial_{\Delta^b} 

2567:            +  \tfrac{4g-8-n}{n+2} \Delta^2 \Big)\coeff^{(g-1)}_{E\, (n)} + \coeff^{(g)}_{E\, (0)}\ ,

2568: \eeq

2569: %

2570: where $c^{(g)}_{E\, (0)}(t)$ is a holomorphic ambiguity in the fiber. Recall that the full expression \eqref{Fgantihol} for

2571: $F^{(g)}_E(t,\bar t)$ is known from heterotic-type II duality. Therefore, verifying that this closed expression 

2572: fulfills the differential equation \eqref{general-direct-deltaE} provides a non-trivial 

2573: check of our derivations.

2574: 

2575: Let us end this section by presenting the first non-trivial solution to the closed expressions \eqref{gencksol} and \eqref{general-direct-delta}

2576: for $F^{(g)}$. More precisely, one derives that the free energy $F^{(3)}$ admits the 

2577: following propagator expansion

2578: \bea

2579:   F^{(3)} &=& -\tfrac{1}{48}\hat E^2_2 \big( 14 \Delta^4 + 10 \epsilon^4_{ab} \Delta^a \Delta^b

2580:     - \epsilon^4_{ac} \epsilon^4_{bd} C^{ab} C^{cd}   \big) \nn \\

2581:                   &&     -\tfrac{1}{48} E_4  \big( - 2 \Delta^4+2 \epsilon^4_{ab} \Delta^a \Delta^b

2582:                  -  \epsilon^4_{ac} \epsilon^4_{bd} C^{ab} C^{cd}  \big)\ ,

2583: \eea

2584: where $\epsilon^4_{ab}$ is defined in \eqref{def-epsilon4}.

2585: Note that the last term in the first line has to be determined by the direct integration with respect to 

2586: $\widehat E_2$ by using \eqref{gencksol}. Moreover, the purely holomorphic term 

2587: \beq \label{def-f3full}

2588:   f^{(3)}(S,t)= \tfrac{1}{48} E_4 \epsilon^4_{ac} \epsilon^4_{bd} C^{ab} C^{cd}

2589: \eeq

2590: is the holomorphic ambiguity at genus $3$, determined by the fiber limit. 

2591: In other words, applying \eqref{Elimit} one easily derives 

2592: \beq

2593:   F^{(3)}_E = -\tfrac{1}{4} \epsilon^4_{ab} \Delta^a \Delta^b - \tfrac14 \Delta^4

2594:     + \tfrac{1}{24} \epsilon^4_{ac} \epsilon^4_{bd} C^{ab} C^{cd}\ ,

2595: \eeq

2596: which is readily compared with the general expression \eqref{Fgantihol} for 

2597: the fiber free energies. It is straightforward to derive all  $F^{(g)}$ for $g<7$ by

2598: evaluating \eqref{gencksol} and \eqref{general-direct-delta} and fixing the ambiguity by comparison with the 

2599: fiber result \eqref{Fgantihol}. Clearly,  at genus greater than $6$ we will encounter the same difficulties 

2600: as in section \ref{simple}. Only additional boundary conditions can help to 

2601: fix the ambiguities in these cases. In the next section we will summarize possible additional conditions.

2602: 

2603: \subsection{Boundary conditions}\label{boundaries}

2604: 

2605: One important feature of the formalism of direct integration is that modular  

2606: and holomorphic properties of the $F^{(g)}$ are manifest. In particular the 

2607: ambiguity is holomorphic, modular invariant and for given genus expressible 

2608: in terms of a modular form of finite weight. This implies that a finite number of 

2609: data will fix it. The latter must be provided from additional 

2610: information at the boundaries of the moduli space of the Calabi-Yau manifold. Let 

2611: us give a short overview over the the nature of these boundary conditions.  

2612: 

2613: In the large radius limit the holomorphic limit  of the 

2614: $F^{(g)}$ has an expansion in terms of Gromov-Witten invariants $N^{(g)}_\beta$. 

2615: Since the an-holomorphic part is fixed, the $F^{(g)}$ can be completely 

2616: determined by calculating a finite number of Gromov-Witten invariants. 

2617: The reorganisation of the expansion  in terms of Gopakumar-Vafa  

2618:  invariants $n^{(g)}_\beta$ is useful here, because the latter 

2619: vanish if the degree is higher then the maximal degree for which 

2620: a smooth curve exists in a given class.                  

2621: 

2622: For K3-fibered Calabi-Yau threefolds, the limit of large 

2623: base volume corresponds generically to a perturbative heterotic 

2624: string theory on K3$\times \IT^2$. If the heterotic theory is known one

2625: can calculate the dependence of the $F^{(g)}$ on the fiber moduli  

2626: by calculating a BPS saturated one loop amplitude in the heterotic 

2627: string~\cite{mm,km}. In the Enriques CY case this yields most 

2628: of the information and is the reason that one can tackle an

2629: $11$ parameter model at all. Even if the heterotic dual is not known, 

2630: one may get all the holomorphic $F^{(g)}$ in the fiber from the 

2631: modular properties of the B-model on the K3 and the formula 

2632: for the cohomology of the Hilbert scheme of points on the 

2633: fiber~\cite{Klemm:2004km}.

2634: 

2635: If the Calabi-Yau admits controllable local limits, e.g. 

2636: to toric Fano varieties with anti-canonical 

2637: bundle, then the $F^{(g)}$ can be 

2638: unambiguously calculated using the topological 

2639: vertex~\cite{vertex}.   

2640: 

2641: One can also find boundary conditions by looking at the behavior of the 

2642: topological string amplitudes near the conifold point, as we discussed in section \ref{SWboundary}. When there is 

2643: only one hypermultiplet becoming massless at the conifold point, the amplitudes behave like (\ref{thegap}), where 

2644: $t_D$ is a suitable coordinate transverse to the conifold 

2645: divisor. This yields $2g-2$ independent conditions 

2646: on the holomorphic ambiguity.    

2647: 

2648: 

2649: In contrast to generic $\CN=2$ compactifications, the four dimensional massless 

2650: spectrum at singularities of the Enriques Calabi-Yau is conformal, which 

2651: requires hyper- and vector multiplets to become simultaneously massless. 

2652: The leading behavior of the corresponding effective action is less 

2653: characteristic. We will find a partial gap in the reduced model 

2654: considered in section \ref{red-model}, which is similar to the partial

2655: gap structures that were found in~\cite{hkq} at a point where likewise 

2656: several RR states become massless. 

2657: The determination of the subleading  behavior is possible in the field theory limit and 

2658: yields conditions on the anomaly. We will consider here only the complex codimension 

2659: singularities that we discussed in section \ref{sec:Enriques}. The nontrivial information about the 

2660: $F^{(g)}$ comes from the $N_f=4$ locus: as we will show in section \ref{sec:ftlim}, 

2661: the residue of the leading singularity near (\ref{slocusone}) can be computed using 

2662: instanton counting in field theory. 

2663: 

2664: Let us now analyze the leading singularity of $\CF^{(g)}$ near the singular loci in the fiber limit. This 

2665: can be done with the heterotic computations of \cite{km} reviewed in section \ref{sec:Enriques}. These 

2666: computations give us expansions around two special regions in moduli space, the large radius limit 

2667: (where $t^a$ are large) and the region appropriate to the BHM reduction (where $t_D^a$ are large). 

2668: As in \cite{agnt,mm}, we can use the computation at large radius to obtain the leading behavior of 

2669: the fiber amplitudes near (\ref{slocusone}), and the computation in the BHM reduction to obtain the 

2670: behavior near (\ref{slocustwo}). 

2671: 

2672: Let us first look at the behavior near (\ref{slocusone}). A possible singular behavior there must come from the vector 

2673: $r=(1, -1)$ in (\ref{gr}), since this leads to a polylogarithm which, when expanded at the singular locus (\ref{slocusone}),

2674: %

2675: \be

2676: \label{polyexp}

2677: {\rm Li}_{3-2g}(\re^{-z})={(2g-3)! \over z^{2g-2}} +\CO(z^0), \quad g\ge 2,

2678: \ee

2679: %

2680: exhibits a pole. Here, $z=t^1-t^2$. However, since $c_g(-2)=0$, the coefficient of this polylogarithm vanishes and 

2681: we conclude that the amplitudes are regular at (\ref{slocusone}). This is indeed 

2682: consistent with the fact that the field theory limit of this model at (\ref{slocusone}) is massless 

2683: $SU(2)$, $\CN=4$ super Yang--Mills theory, which has $\CF^{(g)}=0$ for all $g\ge 2$ \cite{Nek,no,bfmt}. 

2684: 

2685: 

2686: Let us now look at the behavior near (\ref{slocustwo}). To understand this, we look at the heterotic 

2687: result for the holomorphic couplings in the BHM reduction (\ref{finalfgbor}). Again, 

2688: the singular behavior comes from the vector $r=(1,-1)$. Since the coefficients are defined 

2689: now by (\ref{finalmodcoef}), we find

2690: %

2691: \be

2692: d_g(-1)={4^g-1 \over 2^{g-2}} {|B_{2g} | \over 2g (2g-2)!}.

2693: \ee

2694: %

2695: If we set 

2696: %

2697: \be

2698: \mu=t^1_D -t^2_D,

2699: \ee

2700: %

2701: and we take into account the behavior of the polylogarithm (\ref{polyexp}), 

2702: we find that the singular behavior of $\CF^{(g)}_E(t_D)$ near (\ref{slocustwo}) is given by

2703: %

2704: \be

2705: \label{fgsingen}

2706: \CF^{(g)}_{E} (t_D) \rightarrow {4^g-1 \over 2^{g-2}}   {|B_{2g}| \over 2g (2g-2)}{1\over  \mu^{2g-2}} + 

2707: \CO(\mu^0) 

2708: \ee

2709: %

2710: for $g\ge 2$, while for $g=1$ we have a logarithm singularity %

2711: \be

2712: -{1\over 2} \log \, \mu.

2713: \ee

2714: %

2715: Since the full $\CF^{(g)}(S,t_D)$ can be written for $g\le 6$ in terms of (\ref{finalfgbor}), as we showed in 

2716: section \ref{simple}, we can compute its leading singular behavior at 

2717: the locus (\ref{slocustwo}). This will be useful in section \ref{sec:ftlim} to compare to the field theory limit. 

2718: The above computation shows that along the fiber direction the topological 

2719: string amplitudes $\CF_E^{(g)}$ show the gap behavior discovered in \cite{hk, hkq}. 

2720: In order to see if the gap also holds in the mixed directions, it is clear from the formulae above 

2721: that we need a precise knowledge of the regular terms in $\mu$ in the expansion of $\CF_E^{(g)}$. 

2722: Unfortunately, this is something we cannot extract from the heterotic expressions. We will however

2723: be able to do this in the reduced model introduced in \cite{km} and studied in more detail below. 

2724: We will see that indeed the strong gap condition obtained for the fiber direction in (\ref{fgsingen}) 

2725: does not hold for the mixed directions.  

2726: 

2727: 

2728: 

2729: 

2730: 

2731: \subsection{The reduced Enriques model \label{red-model}}

2732: 

2733: In this section we discuss a reduced model for the Enriques Calabi-Yau introduced in \cite{km}.

2734: The main advantage of this model is that the target symmetry group becomes much simpler, and 

2735: one can easily parametrize the holomorphic functions which appear in the 

2736: expansion of $F^{(g)}$ in the propagators $\Delta^a(t,\bar t)$ 

2737: and $\widehat E_2(S,\bar S)$. In particular, the holomorphic ambiguity can be parametrized in terms of a finite number 

2738: of coefficients at each genus. Also the mirror map is known explicitly and can be used to translate the 

2739: $F^{(g)}$ into a simple polynomial form.

2740: In these aspects, the reduced model is very closely related

2741: to the Seiberg--Witten theory studied in section \ref{Seiberg-Witten}. 

2742: 

2743: 

2744: \subsubsection{Special geometry and the mirror map}

2745: 

2746: We begin with a brief discussion of the reduced special geometry and recall the mirror map 

2747: derived in \cite{km}. Out of the eleven special coordinates $S,t^a$ the reduced model 

2748: is only parametrized by three parameters. More precisely, it is

2749: obtained by shrinking $8$ out of the $10$ cycles in the Enriques fiber as

2750: \beq \label{reductionlimit}

2751:    (S,t^a) = (S,t^i ,t^\alpha) \quad \rightarrow \quad (S,t^i,0)\ , \qquad i=1,2\ , \quad \alpha=3,\ldots, 10\ .

2752: \eeq

2753: We denote the reduced moduli space spanned by the remaining coordinates $S,t^1,t^2$ by $\cM_{\rm r}$. 

2754: Explicitly, the full coset \eqref{coset} reduces in this limit to 

2755: \beq \label{cosetr}

2756:    \cM_{\rm r} = \frac{Sl(2,\bbR)}{SO(2)}  \times \left( \frac{Sl(2,\bbR)}{SO(2)} \right)^2\ ,

2757: \eeq 

2758: inducing a split of the full target space symmetry group as

2759: \beq \label{group-reduction}

2760:   Sl(2,\bbZ) \times O(10,2,\bbZ) \quad \rightarrow\quad Sl(2,\bbZ) \times \Gamma(2) \times \Gamma(2)\ .

2761: \eeq

2762: The generators of $Sl(2,\bbZ)$ are precisely the Eisenstein series $\widehat E_2(S,\bar S),\ E_4(S),\ E_6(S)$

2763: as already introduced for the full model in \eqref{genS}. The generators for $\Gamma(2)$ have been introduced 

2764: in the Seiberg-Witten section \ref{Seiberg-Witten}. More precisely, we will generate the ring of almost holomorphic 

2765: modular forms of $\Gamma(2)$ by $\widehat E_2(t,\bar t)$, $K_2(t)$ and $K_4(t)$ explicitly defined in \eqref{def-widhatE2} and \eqref{gamma2generators}.

2766: In the following we will simplify expressions by abbreviating

2767: \begin{align} \label{def-tildeK}

2768:   E_2 &= E_2(t^1)\ , \qquad & K_2&= K_2(t^1)\ , \qquad &  K_4&= K_4(t^1)\ ,\nn \\

2769:   \tilde E_2 & = E_2(t^2)\ , \qquad & \tilde K_2&= K_2(t^2)\ , \qquad& \tilde K_4&= K_4(t^2)\ .

2770: \end{align}

2771: Whenever not stated otherwise, we will keep the $S$-dependence explicit.

2772: Let us also note that the matrix $C_{ab}$ splits as

2773: \beq

2774:    C_{ab} = \left(\begin{array}{cc} C_{ij}& 0\\ 0 & C_{\alpha \beta} \end{array} \right)\ , \qquad C_{ij} =\left(\begin{array}{cc} 0& 1\\ 1 & 0 \end{array} \right)\ ,

2775: \eeq

2776: as already given in \eqref{def-Gamma_E}.

2777: Hence, the holomorphic prepotential \eqref{def-Eprepot} and the fiber K\"ahler potential $Y=(t+\bar t)^2$ reduce to

2778: \beq \label{red-prepot}

2779:   \cF_{\rm r}= iS t^1 t^2\ ,\qquad \qquad Y_{\rm r} = (t^1+\bar t^1)(t^2+\bar t^2)\ .

2780: \eeq

2781: As we have already noted in section \ref{specialE} 

2782: this prepotential and fiber potential are exact and receive no 

2783: instanton corrections.

2784: 

2785: 

2786: 

2787: Let us now turn to a discussion of the mirror map for the reduced Enriques model.

2788: In order to determine this duality map we first note that the reduced Enriques has 

2789: an algebraic realization. Applying standard techniques, one can thus 

2790: derive the three Picard-Fuchs equations for the holomorphic three-form

2791: $\Omega(z)$ as

2792: \beq \label{PF-eqn}

2793:   \cL_1 \Omega(z) =0\ ,\qquad \quad \cL_2 \Omega(z)=0 \ , \qquad \quad \cL_3 \Omega(z)= 0 \ ,

2794: \eeq

2795: where $z^i(t),z^3(S)$ with $i=1,2$ are the mirror coordinates of $t^i,S$ respectively.

2796: The Picard-Fuchs operators are found to be 

2797: \bea

2798:   \cL_i &=& \theta^2_i - 4(4\theta_i +4 \theta_j - 3)(4\theta_i + 4\theta_j-1)z_i\ ,\qquad \quad i,j=1,2\ , \quad i\neq j\ , \\

2799:   \cL_3 &=& 36 (z^3-1)^2  (z^3-2) \theta_3^2 + 36 z^3 (z^3-1) \theta_3 + z^3 \big(8z^3-4(z^3)^2-31\big) \ ,

2800: \eea

2801: where $\theta_i=z^i \frac{\partial }{\partial z^i}$.

2802: The Picard-Fuchs equations \eqref{PF-eqn} can be solved to determine the mirror maps $z^i(t),z^3(S)$. 

2803: This was done in \cite{km} and we will only quote the result here. We first abbreviate 

2804: \beq

2805:    \label{def-zonly}

2806:    z(q_i)= \frac{K_4(t^i)}{ K_2^2(t^i)}\ .

2807: \eeq

2808: Using this shorthand notation the fiber mirror map reads 

2809: \begin{equation} \label{def-mirrorz}

2810:  z^1(t)=z(q_1) \big(1- z(q_2)  \big)\ ,\qquad \qquad 

2811:   z^2(t)= z(q_2)  \big(1- z(q_1)  \big)\ . 

2812: \end{equation}

2813: These coordinates are related by a factor of $64$ to $z_1,z_2$ used in ref.~\cite{km}.

2814: In the base one evaluates

2815: \beq

2816:   z^3(S)= 1-E_4^{-3/2} E_6\ . 

2817: \eeq

2818: Compared to \cite{km} we have rescaled $z^3$ by a factor $864$. Using these

2819: explicit expressions for $z^1,z^2$ and $z^3$, one immediately verifies 

2820: their invariance under the target space symmetry group $Sl(2,\bbZ) \times \Gamma(2) \times \Gamma(2)$.

2821: Also the fundamental period $X^0$ can be obtained from the Picard-Fuchs system 

2822: \eqref{PF-eqn} and reads 

2823: \beq \label{def-X0}

2824:   X^0  = x^0 \hat X^0\ ,\qquad \qquad (\hat X^0)^2 = \tfrac{1}{4} K_2 \tilde K_2\ ,\qquad \qquad (x^0)^4 = E_4\ .

2825: \eeq

2826: We immediately verify that $X^0$ is not invariant under  

2827: the symmetry group $Sl(2,\bbZ) \times \Gamma(2) \times \Gamma(2)$. 

2828: The S-duality transformation \eqref{saut} reads for the reduced model 

2829: $t^1  \mapsto 1/t^2$ and $t^2  \mapsto 1/t^1$. 

2830: Applied to $X^0$ this yields precisely the transformation behavior given in \eqref{trans-X0}.

2831: Before turning to the higher genus amplitudes in the next section let us also

2832: note that the discriminant of the reduced model 

2833: is given by

2834: \beq

2835:    \Delta(z^1,z^2)\ D(z^3)\ , 

2836: \eeq 

2837: where $\Delta(z^1,z^2)$ is the discriminant along the fiber and 

2838: $D(z^3)$ is the discriminant along the base. Explicitly, we find in the 

2839: coordinates \eqref{def-zonly} and \eqref{def-mirrorz} that 

2840: \bea \label{def-disc1}

2841:     \Delta(z^1,z^2) &=& \big(1-z(q_1)-z(q_2)\big)^2  \\ 

2842:                 &=&1- 2(z^1 + z^2 + z^1 z^2) + (z^1)^2 + (z^2)^2\ .

2843: \eea 

2844: The second discriminant $D(z^3)$ is given by

2845: \beq \label{def-disc2}

2846:  D(z^3)=\tfrac{1}{2^6 3^3} \big( (z^3)^2-z^3\big) \ .

2847: \eeq

2848: In the next section we will use the mirror coordinates $z^1,z^2$ to express the 

2849: reduced free energies $F^{(g)}_r$. Since along the base direction all equations are 

2850: expressed in terms of simple Eisenstein series $E_{2n}(S)$ we choose to keep this 

2851: $S$-parametrization also in the following discussions.

2852: 

2853: 

2854: \subsubsection{Reduced free energies and direct integration}

2855: 

2856: Let us now discuss the free energies $F^{(g)}_r$  and their holomorphic limits $\cF^{(g)}_r$ 

2857: for the reduced model. In the limit \eqref{reductionlimit} they are simply defined as

2858: %

2859: \be

2860: F^{(g)}_r(S,t^1, t^2) = F^{(g)}(S,t^1, t^2, t^{\alpha}=0)\ .

2861: \ee

2862: %

2863: The reduced form of $F^{(1)}$ can be derived by direct computation as was already discussed in \cite{km}. 

2864: Explicitly one finds 

2865: %

2866: \beq

2867:    F^{(1)}_{\rm r} = - 2\log\big[(S+\bar S)^3 (t^1+\bar t^1)(t^2 +\bar t^2)\big] - \log |\Phi_{\rm r}(S,t)| \ ,

2868: \eeq

2869: where

2870: \beq \label{def-Phir}

2871:     \Phi_{\rm r}(S,t^1,t^2) = \eta^{24}(S) \prod_{m,n} \Big(\frac{1-q^n q^m}{ 1+q^n q^m} \Big)^{c^{\rm r}_1(2mn)}\ .

2872: \eeq

2873: The coefficients $c_1^r(n)$ are given through the modular form 

2874: \beq \label{reduced_c}

2875:   \sum_n c^{\rm r}_1(n) q^n = - \frac{64}{3 \eta^6(q) \vartheta^6_2(q)} E_2(q) E_4(q^2)\ .

2876: \eeq

2877: Note that in comparison with the expression \eqref{geomrmod} for the full Enriques model the 

2878: Eisenstein series $E_4(q^2)$ appears in \eqref{reduced_c}. This extra factor arises due to the summation 

2879: over the $E_8$ vectors in \eqref{def-Phi} and precisely counts their degeneracy.

2880: It was further shown in \cite{km} that the following denominator formula holds

2881: \beq

2882:      \Phi_{\rm r}(S,t^1,t^2) = \tfrac{1}{16} \eta^{24}(S)\ \delta =  \eta^{24}(S) (\hat X^0)^{4} \Delta^{1/2}

2883: \eeq

2884: where 

2885: \beq

2886:    \delta(t^1,t^2) = K^2_2 \tilde K^2_2 - K_4 \tilde K_2^2 - K_2^2 \tilde K_4\ .

2887: \eeq

2888: Here the $\Gamma(2)$ generators $K_2,\tilde K_2 $ as well as $K_4,\tilde K_4$ are defined in \eqref{def-tildeK}, while 

2889: the fundamental period $\hat X^0$ and the discriminant $\Delta$ were given in \eqref{def-X0} and \eqref{def-disc1}.

2890: 

2891: The holomorphic reduced amplitudes restricted to the Enriques fiber can also be 

2892: computed directly by reducing the heterotic expressions (\ref{gr}) and (\ref{finalfgbor}).

2893: The result reads \cite{km}

2894: %

2895: \be

2896: \label{red}

2897: \ba

2898:   \cF^{(g)}_{{\rm r},E} (t) &= \sum_{r>0} c^{\rm r}_g(r^2) \Big[2^{3-2g} \text{Li}_{3-2g}(e^{-r\cdot t}) -  \text{Li}_{3-2g}(e^{-2r\cdot t}) \Big]\ , \\

2899:   \CF^{(g)}_{{\rm r},E} (t_D)&=\sum_{r>0} d^{\rm r}_g(r^2/2)(-1)^{n+m} {\rm Li}_{3-2g}({\rm e}^{-r \cdot t_D })\ ,

2900:   \ea

2901: \ee

2902: %

2903: where the coefficients $c^{\rm r}(n)$, $d^{\rm r}_g(n)$ are defined by

2904: %

2905: \be

2906: \label{redco}

2907: \ba

2908: \sum_n c^{\rm r}_g(n) q^n& =-2 E_4(q^2) \frac{{\cal P}_{g}(q)}{\eta^{12}(2\tau)}\ ,\\

2909: \sum_n d^{\rm r}_g (n) q^n &=E_4(q^2) \frac{2^{2+g} {\cal P}_{g}(q^4)-2^{2-g} {\cal P}_{g}(q)}{\eta^{12}(2\tau)}\ .

2910: \ea

2911: \ee

2912: %

2913: Once again we recognize the additional factor $E_4(q^2)$ counting the degeneracies of the $E_8$ 

2914: lattice. Clearly, also the expressions $\cF^{(g)}_{{\rm r},E}(t)$ and $\CF^{(g)}_{{\rm r},E} (t_D)$

2915: can be expressed in terms of the holomorphic generators \eqref{def-tildeK} depending on $t^i$ and $t^i_D$ 

2916: respectively. 

2917: 

2918: Let us now turn to the discussion of the complete reduced amplitudes including the base and 

2919: the non-holomorphic dependence. In order to do that we describe the direct integration 

2920: for the reduced model focusing on the essential differences to the considerations 

2921: presented in section \ref{direct_fiber_base}.

2922: To begin with, note that the 

2923: propagators of the full model reduce as

2924: \beq \label{def-rprop}

2925:    \Delta^i\quad \rightarrow \quad \rprop^i  \ ,\qquad \qquad%= -C^{ij}\big( \hat K_j -\tfrac18 \partial_j \log \Delta \big)\ ,\\

2926:    \Delta^\alpha\quad \rightarrow \quad 0\ ,

2927: \eeq

2928: where $\rprop^i$ is obtained from \eqref{def-Deltaa} by setting $t^\alpha=0$ and using \eqref{def-Phir}.

2929: That $ \Delta^\alpha$ reduces to zero arises from the fact that in

2930: summation over the $E_8$ lattice the vectors cancel pairwise.

2931: In order to perform the direct integration we first have to 

2932: find recursive relations which are valid for the reduced free energies 

2933: $F^{(g)}_{\rm r}$.

2934: Recall that in the full Enriques model we found two sorts of 

2935: recursive relations \eqref{directE} and \eqref{directDelta} capturing the properties $F^{(g)}$

2936: in the base and in the fiber of the Enriques. It turns out that only the second anomaly equation  

2937:  \eqref{directDelta}  admits a simple reduction. More precisely, it can be rewritten for the 

2938:  reduced model as

2939:  \beq \label{recrel_reduced}

2940:    \frac{\partial F^{(g)}_{\rm r} }{\partial \rprop^i}= 

2941:   D_S  D_i F^{(g-1)}_{\rm r} + \sum_{r=1}^{g-1} D_i F^{(r)}_{\rm r} D_S F^{(g-r)}_{\rm r} \ ,

2942: \eeq

2943: since performing the reduction $t^\alpha=0$ interchanges with the differentiation with respect 

2944: to $t^1,t^2$. Note that this is no longer true for derivatives with respect to $t^\alpha$. In particular,

2945: the first equation \eqref{directE} involves a summation over the $\alpha$ indices and one shows 

2946: that the resulting terms do not vanish under the reduction $t^\alpha=0$.

2947: Nevertheless, one can directly integrate \eqref{recrel_reduced} for the reduced free energies

2948: \beq \label{small_propexp}

2949:     F^{(g)}_{\rm r} = \sum_{n=1}^{2g-2} \rprop^{i_1} \ldots \rprop^{i_n} \hat c^{(g)}_{i_1 \ldots i_n} + \hat c^{(g)}\ ,\qquad \quad g>1\ .

2950: \eeq

2951: The function $ \hat c^{(g)}$ is holomorphic in $t^i$ and generally depends on $\widehat E_2(S,\bar S),E_4(S),E_6(S)$. 

2952: Note that due to \eqref{def-rprop} the coefficients of the full and reduced model 

2953: are related by $\hat c^{(g)}_{i_1 \ldots i_n}=c^{(g)}_{i_1 \ldots i_n}(t^\alpha=0)$.

2954: The direct integration is performed in analogy to the integration in the full model 

2955: and results in a closed expression similar to \eqref{general-direct-delta}.

2956: The important difference is that the $\epsilon^4_{\text{r}\, ij}$ as well as the covariant  

2957: derivatives $\hat D^{\rm r}_a$ are not obtained from the full $\epsilon^4_{ab}$ and $\hat D_a$ by simply restricting to 

2958: the $i,j$ indices and setting $t^\alpha=0$. Both $\epsilon^4_{\text{r}\, ij}$

2959: as well as $\hat D^{\rm r}_a$ have to be defined with respect to a new holomorphic 

2960: metric $\hat C_{ij}^{\rm r}=\Phi^{1/2}_{\rm r} C_{ij}$ but otherwise analog to \eqref{def-epsilon4} and \eqref{def-hatD}.

2961: If one had been using the old connection, an additional summation over the $\alpha$ indices would arise and 

2962: yield extra contributions. Applied to the specific free energy $F^{(3)}$ one finds the reduction of 

2963: the holomorphic ambiguity \eqref{def-f3full}

2964: \beq

2965:      f^{(3)}_{\rm r}(S,t)= \tfrac{1}{48} E_4 \big( \epsilon^4_{\text{r}\, ik} \epsilon^4_{\text{r}\, jl} C^{ij} C^{kl} +\tfrac18 (\epsilon^4_{\text{r}\, ij} C^{ij})^2 \big)

2966: \eeq

2967: After these considerations it is not surprising that the contraction of the new $\epsilon^4_{\text{r}\, ij} $ with 

2968: $C^{ij}$ does not vanish as it is the case in the full model \eqref{Cepsilon4=0}.

2969: 

2970: 

2971: 

2972: \subsubsection{The free energies $F^{(g)}$ on the mirror}

2973: 

2974: So far the reduced free energies $F^{(g)}_{\rm r}$ were expressed as

2975: functions of the variables $t^i,S$ or $t_D^i,S$.

2976: In the reduced model we know the mirror map explicitly and thus will 

2977: be able to translate the expansion  \eqref{small_propexp} 

2978: of $F^{(g)}_{\rm r} $ into a function of the complex coordinates $z^i$. 

2979: We will show that the holomorphic coefficients become polynomials in 

2980: $z^i$ divided by an appropriate power of the discriminant.

2981: Since the dependence of $F^{(g)}_{\rm r}$ is rather transparent we 

2982: chose to keep this variable and do not replace it by its mirror 

2983: counterpart $z^3$. 

2984: 

2985: The $F^{(g)}$ transform non-trivially 

2986: under the reduced automorphic transformations. We already discussed 

2987: the actually invariant combination in \eqref{inv_comb}. In the coordinates 

2988: $z^i,S$ we thus set 

2989: \beq

2990:     F^{(g)} (z,\bar z,S,\bar S) = (\hat X^0)^{2-2g}\, F^{(g)}(t,\bar t,S,\bar S) \ .

2991: \eeq

2992: This definition is consistent with the fact that the $z^i(t)$ are invariant 

2993: under the target space group \eqref{group-reduction}, while $(\hat X^0)^{2g-2}$ transforms exactly as 

2994: $F^{(g)}(t,S)$. To rewrite the expansion  \eqref{small_propexp} we first note that 

2995: the propagator $\rprop^i$ can be written in the $z^i$ coordinates as

2996: \beq \label{trans-rprop}

2997:     \rprop^i = (\hat X^0)^2 \frac{\partial t^i}{\partial z^j}\rprop^{z^j}\ , \qquad \qquad \rprop^{z^i} = -C^{z^i z^j}\big( \hat K_{z^j} -\tfrac18 \partial_{z^j} \log \Delta \big)\ ,

2998: \eeq

2999: where $\rprop^{z^i}$ is a function of $z^i,\bar z^i$ and we have used 

3000: \beq \label{trans-CC}

3001:   C_{ij}= (\hat X^0)^{-2} C_{z^k z^l} \frac{\partial z^k}{\partial t^i} \frac{\partial z^l}{\partial t^j}\  ,\qquad \quad

3002:   \hat K(z,\bar z) \equiv - \log\big[|\hat X^0|^2 Y_{\rm r}(z,\bar z) \big]  . 

3003: \eeq

3004: It is not hard to use the expressions \eqref{def-mirrorz} for $z^1$ and $z^2$ to evaluate $C_{z^i z^j}$ explicitly as

3005: \beq

3006:   C_{z^1 z^2} = \frac{1}{z^1 z^2 \Delta}(1-z^1-z^2) \ , \qquad C_{z^1 z^1} =  \frac{1}{z^1 z^2 \Delta} 2 z^2\ , \qquad C_{z^2 z^2} = \frac{1}{z^1 z^2\Delta} 2z^1\ . 

3007: \eeq

3008: Once again \eqref{trans-rprop} and \eqref{trans-CC} are in accordance with the transformation behavior of the 

3009: $\rprop^i$ and $C_{ij}$ given in \eqref{trans-Delta} and \eqref{jcont}.

3010: Similarly, we transform the coefficients $\hat c^{(g)}_{i_1\ldots i_n}$ in \eqref{small_propexp} and set 

3011: \beq \label{coeffinz}

3012:     \hat c^{(g)}_{i_1 \ldots i_n } = (\hat X^0)^{2g-2 - 2n} \

3013:     \frac{\partial z^{j_1}}{\partial t^{i_1}}\ldots  \frac{\partial z^{j_n}}{\partial t^{i_n}}\ \hat c^{(g)}_{z^{j_1} \ldots z^{j_n}}(z)\ , 

3014: \eeq

3015: which is consistent with \eqref{trans-k}. It is also straightforward to rewrite the 

3016: direct integration expression for $F^{(g)}_{\rm r}$

3017: by using the $z^i$ coordinates. Let us once again only discuss the appearing

3018: building blocks. We begin by noting that the holomorphic covariant derivative transforms as

3019: \beq

3020:   \hat D_i V_j = (\hat X^0)^{2k} \frac{\partial z^l}{\partial t^i} \frac{\partial z^m}{\partial t^j}  \hat D_{z^l} V_{z^m}\ ,

3021: \eeq 

3022: where the covariant derivative $\hat D_{z^i}$ is given by 

3023: \beq

3024:    \hat D_{z^i} V_{z^j} = \partial_{z^i} V_{z^j} - \tfrac{k}8 (\partial_{z^i}\log \Delta) V_{z^j}+ \hat \Gamma^{z^l}_{z^i z^j} V_{z^l}\ .

3025: \eeq

3026: The holomorphic Christoffel symbol in this expression is defined by 

3027: \beq

3028:    \hat  \Gamma^{z^l}_{z^i z^j} =\tfrac12 \hat C^{z^l z^m} \big(\partial_{z^i} \hat C_{z^m z^j}+\partial_{z^j} \hat C_{z^i z^m}- \partial_{z^m} \hat C_{z^i z^j} \big)\ , \qquad \hat C_{z^i z^j} = \Delta^{1/4} C_{z^i z^j}\ .

3029: \eeq

3030: The second important object in the general equation for the direct integration 

3031: is the automorphic form $\epsilon^4_{\text{r}\, ij}$. One shows that it can be decomposed as 

3032: \beq \label{def-eps4z}

3033:   \epsilon^4_{\text{r}\, ij} =   \frac{1}{z^1 z^2 \Delta^2}\frac{\partial z^l}{\partial t^i} \frac{\partial z^m}{\partial t^j}\ \epsilon^4_{z^l z^m} \ .

3034: \eeq

3035: where for $i=1,2$, $i\neq j$ one finds that

3036: \bea

3037:   \epsilon^4_{z^i z^i}&=&- \tfrac1{16} z^j\,\big( (z^i)^2\,\left( 1 + 3\, z^j \right)  + 

3038:         {\left( -1 +z^j \right) }^2\,\left( 1 + 3\, z^j \right)  - 

3039:         2\, z^i\,\left( 1 - 5\, z^j + 3\, (z^j)^2 \right) \big)\ ,  \nn\\[.2cm]

3040:   \epsilon^4_{z^i z^j}&=&    \tfrac{3}{16}\,z^i\,z^j\,\left( -2 + z^i + (z^i)^2 + z^j + (z^j)^2- 

3041:       2\,z^i\,z^j \right) \ .

3042: \eea

3043: Note that $\epsilon^4_{z^i z^i}$ is polynomial due to the fact that we extracted the denominator $z^1 z^2 \Delta^2$

3044: in \eqref{def-eps4z}. This turns out to be possible for all the coefficients $ \hat c^{(g)}_{z^{i_1} \ldots z^{i_n}}$

3045: appearing in \eqref{coeffinz}. We thus define

3046: \beq

3047:    P^{(g)}_{i_1 \ldots i_n}(z,\widehat E_2,E_4,E_6) = (z^1 z^2 \Delta)^{g-1}\  \hat c^{(g)}_{z^{i_1} \ldots z^{i_n}}\ ,

3048: \eeq

3049: where $P^{(g)}$ are polynomials in $z^i$ as well as $\widehat E_2,E_4,E_6$.

3050: The reduced free energies are thus of the form

3051: \beq

3052:   F^{(g)}_{\rm r}(z,\bar z,S,\bar S) =\frac{1}{ (z^1 z^2 \Delta)^{g-1} } \sum_{n} \rprop^{z^{i_1}} \ldots \rprop^{z^{i_n}}  P^{(g)}_{i_1 \ldots i_n} \ ,\qquad g>1\ .  

3053: \eeq

3054: In particular, this implies that at each genus the holomorphic ambiguity is

3055: parametrized by a polynomial $P^{(g)}(z,E_4,E_6)$ holomorphic 

3056: in $z^i$ and $S$. As it was the case before the coefficients in $P^{(g)}$ 

3057: have to be determined by boundary conditions. For the lower genera this 

3058: can be done explicitly by using the fiber limes. At higher genus 

3059: additional information are needed and we will discuss in the next section 

3060: the possible input from a small gap condition. We believe that essentially all

3061: results on the mirror rewriting can be generalized to the full model in case one 

3062: is able to determine the full mirror map. For the ten parameters along the fiber 

3063: this is however a technically challenging task.  

3064:  

3065: 

3066: \subsubsection{Boundary conditions and the small gap}

3067: As we have seen in (\ref{group-reduction}), the automorphism group acting on 

3068: the fiber variables is simply 

3069: %

3070: \be

3071: \Gamma(2) \times \Gamma(2)\ ,

3072: \ee

3073: %

3074: where these groups act on $t^{1,2}$, respectively,

3075: plus the exchange $t^1 \leftrightarrow t^2$. Moreover, we see from (\ref{projtrans}) that the $\{t^i=0: i=3, \cdots, 10\}$ locus 

3076: maps to the 

3077: $\{t^i_D=0: i=3, \cdots, 10\}$ locus. If we now define 

3078: %

3079: \be

3080: 2\pi \ri \tau^{1,2} =-t^{1,2}, \quad 2\pi \ri \tau_D^{1,2}= -t_D^{1,2} .

3081: \ee

3082: %

3083: we see that the transformation (\ref{projtrans}) relating the geometric and the BHM expressions 

3084: reduces to 

3085: %

3086: \be

3087: \label{strans}

3088: \tau^1_D=\tau^1, \quad  \tau_D^2 = -{1\over 2} {1\over \tau^2}.

3089: \ee

3090: %

3091: By using the explicit expressions for $ \cF^{(g)}_{{\rm r},E} (t)$ in terms of modular forms 

3092: (which can be obtained for example by direct integration), one finds that under (\ref{strans}) 

3093: %

3094: \be

3095:  \cF^{(g)}_{{\rm r},E} (t) \rightarrow 2^{1-g}  \cF^{(g)}_{{\rm r},E} (t_D),

3096:  \ee

3097:  %

3098: where the factor of $2$ is inherited from the factor of $2$ in (\ref{strans}) and $\cF^{(g)}_{{\rm r},E} (t_D)$ are also given in (\ref{red}). 

3099: Therefore, one can obtain expressions for the amplitudes in the BHM reduction 

3100: in terms of modular forms by simply applying the transformation (\ref{strans}) to the results of the direct integration in the reduced model (which are 

3101: valid for the geometric reduction). 

3102: 

3103: These expressions for the BHM amplitudes can also be used to study in detail the behavior near the singularity (\ref{slocustwo}), and in particular to 

3104: calculate the subleading terms. One can verify that the discriminant (\ref{def-disc1}) transforms under (\ref{strans}) 

3105: as 

3106: %

3107: \be

3108: \Delta(t^1, t^2)\quad \mapsto \quad \Delta(t^1_D, t^2_D)= (z(q_D^1)-z(q_D^2))^2,

3109: \ee

3110: %

3111: which vanishes at the locus (\ref{slocustwo}). This leads to the singular behavior of $\cF^{(g)}_{{\rm r}} (t_D)$, and 

3112: one can now verify the behavior (\ref{fgsingen}) by expanding the expressions in terms of modular forms. One finds, 

3113:  %

3114: \be

3115: \ba

3116: \CF_{{\rm r}, E}^{(1)}(t_D)&=-{1\over 2}\log(\mu)-{1\over 2}\log\Big[\frac{1}{128} K_2K_4(K_2^2-K_4)(q_D^2) \Big]+\CO(\mu), \\

3117: \CF_{{\rm r}, E}^{(2)}(t_D)&={1\over 16\mu^2}-{80E_2^2K_2^2-16K_2^4+3K_2^2K_4+9K_4^2+16E_2(K_2^2+3K_2K_4)\over 9216 K_2^2}(q_D^2)+\CO(\mu),\\

3118: \CF_{{\rm r}, E}^{(3)}(t_D)&={1\over 32\mu^4}+{1\over 53084160 K_2^4}\left(-800E_2^4K_2^4+214K_2^8-726K_2^6K_4+1431K_2^4K_4^2\right.\\

3119: &+405K_4^4-320E_2^3(K_2^5+3K_2^3K_4)+120E_2^2(10K_2^6-15K_2^4K_4+9K_2^2K_4^2)\\

3120: &\left.-540K_2^2K_4-40E_2(14K_2^7-54K_2^5K_4+27K_2^3K_4^2-27K-2K_4^3)\right)(q_D^2)+O(\mu).\\[.2cm]

3121: \ea

3122: \ee

3123: %

3124: However, if one includes the base directions, the gap is ``partially filled'' starting at genus three (for $\CF^{(2)}(S, t_D)$, the gap property away from the fiber limit is trivially satisfied). 

3125: Indeed, one finds that the term $C^{ab}\partial_a\CF^{(1)}_E(t_D) \partial_b \CF^{(2)}_E (t_D)$, leads, in the reduced model, to the expansion

3126: %

3127: \be

3128: \ba

3129: &\partial_1\CF^{(1)}_{{\rm r},E} (t_D) \partial_2 \CF^{(2)}_{{\rm r},E} (t_D) + \partial_2\CF^{(1)}_{{\rm r},E} (t_D) \partial_1 \CF^{(2)}_{{\rm r},E} (t_D)=\\

3130: & -{1\over 8\mu^4}-{20E_2^2K_2+17K_2^3+3K_2K_4+4E_2(K_2^2+3K_4)\over 9216 K_2}(q_D^2){1\over \mu^2}+\cdots \ea

3131: \ee

3132: %

3133: Although there are some nontrivial cancellations (for example, there is no term in $\mu^{-3}$), 

3134: generically one finds, for finite $S$, singular terms in $\mu$ beyond the leading one. 

3135: 

3136: 

3137: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3138: 

3139: \section{The field theory limit }\label{sec:ftlim}

3140: 

3141: As we reviewed in section \ref{sec:Enriques}, there is a line of enhanced symmetry in the moduli space of the Enriques Calabi--Yau 

3142: which leads in the field theory limit to $SU(2)$, $\CN=2$ QCD with four massless hypermultiplets. This occurs at the locus (\ref{slocustwo}). 

3143: Similarly to what happens for other K3 fibrations \cite{kklmv}, we expect that near this locus the leading 

3144: singularities of the topological string partition functions become field theory amplitudes of the $N_f=4$ theory. 

3145: At genus zero one should recover the 

3146: prepotential, and at higher genus the gravitational amplitudes introduced by Nekrasov in \cite{Nek} by using instanton counting techniques. In this section 

3147: we will explain this in some detail, and as spin-off we will obtain some new results on the modularity properties of the $N_f=4$ theory and its gravitational 

3148: corrections. 

3149: 

3150: We first note that the behavior of the amplitudes near (\ref{slocusone}), in the fiber limit, has been already determined with heterotic techniques in (\ref{fgsingen}). 

3151: The results of section \ref{sec:diE} including the base were obtained in principle in the large radius limit, in terms of the ``electric" coordinates $t$. 

3152: However, the calculations of $F^{(g)}$ performed there are also valid in the $t_D$ coordinates, due to general covariance. 

3153: In particular, the holomorphic limit $\CF^{(g)}(S,t_D)$ can be expanded in polynomials in $E_2(S), E_4(S), E_6(S)$ as explained before (\ref{genS}), and we 

3154: can write

3155: %

3156: \be

3157: \CF^{(g)}(S,t_D)=\sum_k p^g_k (S) f^g_k(t_D). 

3158: \ee

3159: %

3160: Near the locus (\ref{slocustwo}) the $f^g_k$ should show display a singular behavior of the form 

3161: %

3162: \be

3163: f^g_k (t_D) = {b_k^g \over \mu^{2g-2}} + \cdots,

3164: \ee

3165: %

3166: as we checked in the fiber limit in (\ref{fgsingen}). How does this compare to the field theory?

3167: 

3168: The prepotential and gravitational corrections of the 

3169: massless $N_f=4$, $SU(2)$ $\CN=2$ Yang--Mills theory depend on the vector multiplet variable $a$ and on the 

3170: microscopic coupling $\tau_0$. They can be put together into a generating functional 

3171: %

3172: \be

3173: \label{genfield}

3174:  \CF^{\rm YM}(a, \tau_0, \hbar) =\sum_{g=0}^{\infty} \hbar^{2g} \CF_g^{\rm YM}(a, \tau_0),

3175: \ee

3176:  %

3177:  where $\CF_0^{\rm YM}(a, \tau_0)$ is the $\CN=2$ prepotential and the higher $g$ amplitudes are the gravitational corrections. 

3178: The statement that the type II theory on the Enriques Calabi--Yau has this gauge theory as its field theory limit near the locus (\ref{slocustwo}) implies that 

3179: the leading singularity of the topological string amplitudes is given by

3180: %

3181: \be

3182: \label{fscomp}

3183: \CF^{(g)}(S,t_D) \rightarrow {1\over \mu^{2g-2}} \sum_k b^g_k p^g_k(S) =\CF_g^{\rm YM}(a, \tau), 

3184: \ee

3185: %

3186: where $S$ is related to the coupling constant of the theory $\tau_0$, and 

3187: $\mu$ is related to the $a$ variable of Seiberg and Witten in a way that we will make precise in a moment.

3188: Let us first look at the prepotential. 

3189: While it has been originally assumed \cite{sw} that the prepotential of the self-dual theories with $\mathcal{N}=2$, gauge group $SU(N)$ and $2N$ flavors 

3190: is classically exact, it was found in \cite{dkm1} that it does get instanton corrections.

3191: Those can however be absorbed in the following redefinition of the coupling \cite{dkm2}, 

3192: %

3193: \be

3194: \tau_0\rightarrow\tau ={1\over 2} {\partial^2\over\partial a^2}\CF^{\rm YM}_0(a, \tau_0)=\tau_0+\sum_{k}c_k q_0^k,

3195: \ee

3196: where

3197: %

3198: \be

3199: q_0=\exp(2 \pi i \tau_0).

3200: \ee

3201: %

3202: We then have for the instanton-corrected prepotential 

3203: %

3204: \be

3205: \CF^{\rm YM}_0(a, \tau_0)={1\over 2} \tau a^2,

3206: \ee

3207: % 

3208: in terms of the renormalized coupling $\tau$.

3209: This is needed in order to match the type II prepotential (\ref{def-Eprepot}), which does not exhibit instanton corrections. 

3210: We will then express the $\CF^{\rm YM}_g$ obtained by instanton computations 

3211: not as functions of $q_0$, but of $q=\re^{2 \pi i \tau}$. 

3212: 

3213: The computation of the field theory amplitudes proceeds as follows. The functional (\ref{genfield}) has the structure

3214: %

3215: \be\label{gf}

3216:  \CF^{\rm YM}(a, \tau_0,  \hbar)=\CF^{\rm YM}_{\rm pert}(a, \hbar) -\hbar ^2 \log Z(a,\tau_0, \hbar), 

3217: \ee

3218: %

3219: where 

3220: %

3221: \be

3222: \label{pertfield}

3223: \CF^{\rm YM}_{{\rm pert},g}(a, \hbar)= -{2B_{2g}\over 4^{(g-1)}2g(2g-2)}(1-4^g){1\over a^{2g-2}}

3224: \ee

3225: %

3226: is the perturbative piece computed in \cite{no}, and 

3227: %

3228: \be

3229: Z(a,\tau_0, \hbar)=\sum_k Z_k(a,\hbar) q_0^k

3230: \ee

3231: %

3232: is an instanton sum. Nekrasov's formula for the $k$-instanton contribution to the partition sum $Z_k(a,\hbar)$ can be written as \cite{bfmt}

3233: %

3234: \be\label{ZNek}

3235: Z_k(a,\hbar)=\sum_{\{Y_\lambda\}}\prod_{\lambda}^{N}\prod_{s\in Y_\lambda}{\varphi_\lambda(s)^4\over \prod_{\tilde{\lambda}} E(s)^2}.

3236: \ee

3237: %

3238: The sum runs over sets $\{Y_\lambda\}$ of Young diagrams labeled in the $SU(2)$ case by $\lambda=1,2$.

3239: For massless flavors,

3240: %

3241: \be

3242: \varphi_\lambda(s)=a_{\lambda}-(s_j-s_i)\hbar,

3243: \ee

3244: %

3245: where $s_i,s_j$ are the coordinates of the cell s inside the Young diagram $Y_\lambda$. We also have

3246: %

3247: \be

3248: E(s)=a_{\lambda\tilde{\lambda}}-\hbar(h(s)-v(s)-1),\qquad h(s)=\nu_{s_i}-s_j,\qquad v(s)=\tilde{\nu}'_{s_j}-s_i,

3249: \ee

3250: %

3251: where $\nu_{s_i}$ is the length of row $s_i$ in $Y_\lambda$, $\tilde{\nu}'_{s_j}$ the length of column $s_j$ in $Y_{\tilde{\lambda}}$ and $h(s),v(s)$ are the number of boxes to the right of s inside $Y_{\lambda}$ respectively above s inside $Y_{\tilde{\lambda}}$, see \figref{Young}. The constants $a_\lambda=(a_1,a_2)$ are set to $(-a,a)$.

3252: 

3253: %

3254: 

3255: \begin{figure}[!ht]

3256: \leavevmode

3257: \begin{center}

3258: %\epsfysize=7.5cm

3259: \includegraphics[height=6.5cm]{young.eps}

3260: \end{center}

3261: \caption{A sample pair of Young diagrams $Y_\lambda,Y_{\tilde{\lambda}}$ contributing to \eqref{ZNek}.}

3262: \label{Young}

3263: \end{figure}

3264: %

3265: 

3266: The relative normalizations between the results in \cite{Nek} and 

3267: the Calabi--Yau case can be obtained from the limit $q \rightarrow 0$, which is the limit $S\rightarrow \infty$. The 

3268: only remaining singularity on the Enriques is then (\ref{fgsingen}), while in the Yang--Mills case we are left with the perturbative piece 

3269: (\ref{pertfield}). Comparing this to (\ref{fgsingen}) and taking into account the relative 

3270: sign in (\ref{fscomp}) we find

3271: %

3272: \be

3273: (-2)^{g-1}{a^{2g-2}\over \mu^{2g-2}}=1,

3274: \ee 

3275: %

3276: and one can immediately read off the normalization of a with respect to $\mu=t_D^1-t_D^2$:

3277: %

3278: \be

3279: a={\mu\over i\sqrt{2}}.

3280: \ee

3281: %

3282: We notice the following factorization, 

3283: %

3284: \be

3285: \CF^{\rm YM}_g (q_0, a)={1\over a^{2g-2}} \Xi_g(q_0), 

3286: \ee

3287: %

3288: where $\Xi_g(q_0)$ is a power series in $q_0$. The relation between $q_0$ and $q$ is defined by

3289: %

3290: \be

3291: q =q_0 \exp[\Xi_0(q_0)], 

3292: \ee

3293: %

3294: which can be inverted to obtain the relation between $q_0 $ and $q$. The explicit power series one finds is

3295: %

3296:  \be

3297:  \label{mmap}

3298: q_0=q-\frac{q^2}{2}+\frac{11 q^3}{64}-\frac{3 q^4}{64}+\frac{359 q^5}{32768}-\frac{75

3299:    q^6}{32768}+\frac{919 q^7}{2097152}-\frac{41 q^8}{524288}+\CO(q^{9}).

3300: \ee

3301: %

3302: If we now plug this series into $\CF^{\rm YM}_g (a, q_0)$ we find that all gravitational couplings are functions of $q^2$, that is to say, there are no 

3303: odd instanton contributions, as it should be since those are forbidden  by a $\mathbb{Z}_2$--symmetry of the theory \cite{sw}. 

3304: The power series (\ref{mmap}) should be given by a mirror map, corresponding to some algebraic realization of an elliptic curve. Indeed, 

3305: when expressed in terms of 

3306: %

3307: \be

3308: q = 2^4 q^{1\over 2}_S, \quad q_S=\re^{-S},

3309: \ee

3310: %

3311: we find

3312: %

3313: \be

3314: q_0=16 \, q_S^{1\over 2}  - 128 \, q_S + 704 \, q_S^{3\over 2} -3072 \, q_S^2 +\cdots= {\vartheta_2^4(q_S) \over \vartheta_3^4 (q_S)},

3315: \ee

3316: %

3317: which is (up to an overall factor $16$) the Hauptmodul of $\Gamma_0(4)$. This equality between $q_0$ and the Hauptmodul has only be checked for the 

3318: first few terms of the instanton expansion, and we don't have a general proof. 

3319: 

3320: We can now express the couplings $\CF^{\rm YM}_g (a, q_0)$, computed from (\ref{ZNek}), in terms of $q_S, \mu$. Due to the connection to the 

3321: Enriques results and the 

3322: field theory limit (\ref{fscomp}), we expect  them to be (up to an overall factor $\mu^{2-2g}$) quasi--modular forms in $q_S$ of weight $2g-2$, and belonging 

3323: to the ring generated by 

3324: $E_2(S)$, $E_4(S)$ and $E_6(S)$. The results obtained with the instanton expansion are in perfect agreement with this. We find at $g=2$

3325: %

3326: \be

3327: \ba

3328: \mu^2 \CF_2^{\rm YM} &=\frac{1}{16}-\frac{3 q_S}{2}-\frac{9 q_S^2}{2}-6 q_S^3-\frac{21 q_S^4}{2}-9 q_S^5+18 q_S^6+\CO\left(q_S^7\right)\\

3329: &=\frac{1}{2^4} E_2(q_S)\ .

3330: \ea

3331: \ee

3332: %

3333: Proceeding in the same way we find, 

3334: %

3335: \be

3336: \ba

3337: \mu^4\CF_3^{\rm YM} &=\frac{1}{2^5}\Big( \frac{2}{3} E_2^2+\frac{1}{3} E_4 \Big)\ ,\\[.2cm]

3338: \mu^6\CF_4^{\rm YM} &=\frac{1}{2^6}\Big(\frac{11}{12} E_2^3+\frac{4}{3} E_2 E_4 +\frac{7}{12} E_6\Big)\ ,\\[.2cm]

3339: \mu^8\CF_5^{\rm YM} &=\frac{1}{2^7}\Big(\frac{17}{9}E_2^4+\frac{97}{18}E_2^2 E_4+\frac{32}{9}E_4^2+\frac{14}{3}E_2 E_6 \Big)\ ,\\[.2cm]

3340: \mu^{10}\CF_6^{\rm YM} &=\frac{1}{2^8}\Big(\frac{619}{120}E_2^5+\frac{218}{9}E_2^3 E_4+\frac{427}{9}E_2 E_4^2+\frac{4501}{144}E_2^2 E_6

3341:     +\frac{4337}{144}E_4 E_6 \Big)\ ,\\[.2cm]

3342: \mu^{12}\CF_7^{\rm YM} &=\frac{1}{2^9}\Big(\frac{1418}{81}E_2^6+\frac{52837}{432}E_2^4 E_4+\frac{12848}{27}E_2^2 E_4^2 +\frac{22631}{108}E_2^3 E_6

3343: +\frac{5423}{9}E_2 E_4 E_6 \\

3344:    &\quad \ \ \ \ \ +\frac{6529}{54}E_6^2 +\frac{352069}{1296}E_4^3 \Big)\ ,

3345: \ea

3346: \ee

3347: %

3348: We point out that we have not proved these equalities, but rather verified them by 

3349: using the instanton expansion up to high order. It is however highly non--trivial that 

3350: this expansion can be matched to a quasimodular form of the required weight. In addition, one can 

3351: verify that the coefficients of the above combinations agree with the Enriques results. For example, 

3352: if we look at the singular behavior of 

3353: (\ref{finalfthree}) by using (\ref{fgsingen}), one finds, 

3354: %

3355: \be

3356: \CF^{(3)}(S,t_D) \rightarrow {1\over 32\mu^4} E_4(S) + {1\over 48 \mu^4}(E_2^2(S)-E_4(S)) = {1\over 96}(2 E_2^2(S) + E_4(S)), 

3357: \ee

3358: %

3359: in agreement with the result above. We have checked that the above polynomials are in accordance with the 

3360: field theory limit of the Enriques model also for $g=4,5,6$. For higher genus the instanton results for the $N_f=4$ theory provide a boundary condition for the 

3361: holomorphic anomaly equation, since they determine the coefficient of the leading singularity near (\ref{slocustwo}) as a function of $S$, and generalize the 

3362: heterotic result (\ref{fgsingen}) away from the fiber. 

3363: 

3364: In summary, we have verified with the instanton computations of \cite{Nek} our general results about the structure of the topological string amplitudes in the 

3365: Enriques Calabi--Yau (in particular our assumption after (\ref{genS}) about the modular properties of the holomorphic ambiguity). Conversely, the 

3366: results on the Enriques side have been instrumental in clarifying the modularity structure of the massless $N_f=4$ theory.

3367: 

3368: 

3369: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3370: %%%%

3371: %%%%      Section 7

3372: %%%%

3373: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3374: 

3375: \section{Direct integration on generic Calabi-Yau manifolds}\label{generic}

3376: 

3377: In this section we present a general formalism which allows for direct integration of the 

3378: holomorphic anomaly equation \eqref{rec_Fg} for a generic Calabi-Yau manifold.

3379: In order to do that we will first have to rewrite these equations by using new

3380: coordinates and introduce the so-called big moduli space $\widehat \cM$ in 

3381: section \ref{sec:bigrecanomaly}. The holomorphic anomaly equations on the big

3382: moduli space have been also discussed in~\cite{Dijkgraaf:2002ac,Verlinde:2004ck}.

3383: The target space symmetry group acts naturally on the coordinates 

3384: of this extended moduli space and we will briefly discuss modular forms 

3385: on $\widehat \cM$ in section \ref{symp_group}. This has been also studied in~\cite{abk}, 

3386: see also~\cite{Gunaydin:2006bz}. Alternatively to the direct integration, the higher 

3387: genus amplitudes can be derived using a Feynman graph expansion in generalization 

3388: of \cite{bcov,Verlinde:2004ck,abk}. We introduce the appropriate propagators and vertices in section \ref{sec:vertprop}.

3389: Finally, in section \ref{direct-int} we derive a closed expression for the $F^{(g)}$ 

3390: using direct integration. This can be viewed as the generalization of the discussion 

3391: of the Seiberg-Witten example in  section \ref{Seiberg-Witten} for compact Calabi-Yau manifolds with 

3392: an arbitrary number of moduli.

3393: 

3394: 

3395: \subsection{The recursive anomaly for $F^{(g)}$ \label{sec:bigrecanomaly}}

3396: 

3397: 

3398: 

3399: In this section  we rewrite the holomorphic anomaly equations \eqref{rec_Fg}  for an enlarged moduli space in which the $2h^{(2,1)}$

3400: coordinates $t^i,\bar t^i$ on $\cM$ are promoted to $2h^{(2,1)}+2$ coordinates $Y^K,\bar Y^K$. 

3401: {}From a geometric point of view, this amounts to working on the moduli space of 

3402: complex normalized $(3,0)$-forms $\Omega$ on the Calabi-Yau manifold under consideration.

3403: We denote this moduli space by $\widehat \cM$ and call it the big moduli space. 

3404: The coordinates $Y^K$ on $\widehat \cM$ are defined as functions 

3405: of $t^k$ by using the homogeneous coordinates $X^K(t)$ arising in the expansion \eqref{exp_Omega} of the holomorphic three-form $\Omega$.

3406: Explicitly, we define 

3407: \beq

3408:    Y^I = \lambda^{-1} X^I(t)\ ,\qquad I = 0,\ldots,h^{(2,1)}\ ,

3409: \eeq

3410: where $\lambda$ is the complex string coupling.

3411: The big moduli space $\widehat \cM$ 

3412: is shown to be a rigid special K\"ahler manifold with K\"ahler potential $\widehat K $ and 

3413: K\"ahler metric $\widehat K_{I\bar J}$ given by

3414: \beq \label{def-Khat}

3415:   \widehat K =\tfrac{i}{2}\big( Y^K \bar \cF_K(\bar Y) - \bar Y^K \cF_K(Y)\big)\ , \qquad 

3416:   \widehat K_{I\bar J} = \partial_{I}\partial_{\bar J}\widehat K = \I\tau_{IJ} \ ,

3417: \eeq

3418: where $\partial_I \equiv \partial_{Y^I}$ and $\partial_{\bar I} \equiv \partial_{\bar Y^I}$

3419: are the derivatives with respect to the coordinates on $\widehat \cM$ and $\tau_{IJ}=\partial_I \partial_J \cF$

3420: is the second derivative of the prepotential.

3421: Note that the K\"ahler metric $\I \tau_{IJ}$ is not positive definite, but rather has complex signature $(h^{2,1},1)$, 

3422: i.e.~has one complex negative direction.   

3423: The metric connection is shown to be 

3424: \beq \label{biggamma}

3425:    \Gamma^I_{JK} = \widehat K^{I \bar M} \partial_J \widehat K_{K \bar M} = -\tfrac{i}{2} C_{JK}^{\ \ I}\ ,

3426: \eeq

3427: where $C_{IJK}(Y)=\partial_I \partial_J \partial_K \cF$ is the third derivative of the prepotential $\cF$.

3428: This implies that the covariant derivative of a tensor $V_K$ on $\widehat M$ is given by

3429: \beq\label{cov_Dgen1}

3430:    D_I V_K \equiv \partial_{I} V_K -\Gamma^J_{IK}V_J =  \partial_{I} V_K + \tfrac{i}{2} C_{IK}^{\ \ J} V_J \ .

3431: \eeq

3432: Here and in the following, we will raise and lower indices using the metric $\widehat K_{I\bar J} = \I\tau_{IJ} $.

3433: For a more exhaustive discussion of rigid 

3434: special geometry we refer to the existing literature \cite{N=2rev,Craps:1997gp,Freed}.

3435: 

3436: Let us now lift the holomorphic anomaly equations  \eqref{rec_Fg} for the free energies $F^{(g)}$ 

3437: to the big moduli space $\widehat \cM$. In order to do that, we evaluate $F^{(g)}(t,\bar t)$ as functions of the 

3438: homogeneous coordinates $X^K$. As reviewed in section \ref{anomaly}, they transform as sections of $\cL^{2-2g}$ such that 

3439: \beq \label{Fg_Y}

3440:    F^{(g)}(Y,\bar Y) = \lambda^{2g-2} F^{(g)}(X,\bar X)\ ,\qquad \qquad Y^K \partial_K F^{(g)} = (2-2g) F^{(g)}\ . 

3441: \eeq

3442: Rewriting the holomorphic anomaly equations \eqref{rec_Fg} using the $Y^K$ coordinates and the functions $F^{(g)}(Y,\bar Y)$

3443: we find  

3444: \beq \label{big_Fg}

3445:        \partial_{\bar I} F^{(g)} = -\tfrac{i}{8} \bar C_{I}^{\ JK} \Big( 

3446:        D_J\partial_{K} F^{(g-1)} +\sum_{r=1}^{g-1} \partial_{J} F^{(r)} \partial_{K} F^{(g-r)} \Big)\ .

3447: \eeq

3448: A detailed derivation of \eqref{big_Fg} can be found in appendix \ref{Cal_big_Fg1}.

3449:  We can also lift the equation \eqref{anomaly_F1} for $F^{(1)}$ to the 

3450: big moduli space $\widehat \cM$ in a way similar to the lift of the holomorphic anomaly equations for $g>1$. 

3451: First recall that $F^{(1)}$ is a section of $\cL^0$ and hence 

3452: as a function of the homogeneous coordinates $X^K(t)$ homogeneous of degree 

3453: $0$ as seen in \eqref{Fg_Y}. This implies that 

3454: \beq \label{F1_hom}

3455:   Y^K \partial_{Y^K} \partial_{\bar Y^M}F^{(1)} = \bar Y^K \partial_{\bar Y^K} \partial_{Y^M}F^{(1)} = 0\ .

3456: \eeq

3457: Using this property and the special geometry identities summarized in appendix \ref{N=2sp}, one derives 

3458: the holomorphic anomaly for $F^{(1)}(Y,Y)$ on the big moduli space 

3459: (see appendix \ref{Cal_big_F1})

3460: \beq \label{big_F1}

3461:   \partial_{I} \partial_{\bar J} F^{(1)} =  \tfrac18 C_{ILM} \bar C_{J}^{\ LM} 

3462:    - \Big(\frac{\chi}{24} -1 \Big) K_{I \bar J}(Y,\bar Y) \ ,

3463: \eeq

3464: where the second derivative of the K\"ahler potential \eqref{kaehlerpotential} is shown to be

3465: \beq \label{deriv_KY}

3466:   K_{I \bar J}\equiv \partial_{Y^I} \partial_{\bar Y^J} K(Y,\bar Y) = 2e^{K} \widehat K_{IJ}+ 4 e^{2K} \bar Y_{I} Y_J\ ,

3467: \eeq

3468: and indices were lowered by contraction with the metric \eqref{def-Khat}.

3469: Note that the last term in the expression for $ K_{I \bar J}$

3470: ensures that the holomorphic anomaly  \eqref{big_F1}

3471: also implies \eqref{F1_hom}.

3472: In this big moduli space formulation, it is straightforward to integrate the holomorphic 

3473: anomaly equation \eqref{big_F1} for $F^{(1)}$. One thus shows that the genus one 

3474: free energy is locally of the form 

3475: \beq \label{F_1solution}

3476:   F^{(1)}(Y,\bar Y) = - \tfrac{1}{2}\log \det(\I\tau_{IJ}) - \Big(\frac{\chi}{24}-1\Big)K(Y,\bar Y) - \ln |\Phi|+f^{(1)}+\bar f^{(1)}\ ,

3477: \eeq

3478: where $\Phi(Y)$ and $f^{(1)}(Y)$ are holomorphic functions arising as integration constants. 

3479: For reasons which will become clear later, we introduced the seemingly artificial split of the 

3480: holomorphic ambiguity into $\Phi$ and $f^{(1)}$.  

3481: The expression \eqref{F_1solution} provides the direct generalization for $F^{(1)}$ in Seiberg-Witten theory \eqref{SWF1withtau} and 

3482: also reduces to the Enriques result \eqref{F1Enriques2}. Clearly, the holomorphic anomaly does not determine $\Phi,\ f^{(1)}$

3483: which were derived in the Seiberg-Witten and Enriques example by using additional information due to modularity

3484: and string dualities. In the next section we will briefly discuss modularity from the point of 

3485: view of the big moduli space $\widehat \cM$.

3486: 

3487: 

3488: \subsection{Monodromy, symplectic group and modular forms \label{symp_group}}

3489: 

3490: In this section we discuss the action of the target space symmetry group on 

3491: the coordinates of the big moduli space $\widehat \cM$ and introduce some 

3492: basic modular forms and modular derivatives.

3493: To begin with, let us note that there is a natural symplectic

3494: action on the periods $(\cF_J,Y^I)$ of the holomorphic three-form given by

3495: \beq \label{period_rot}

3496:    \left(\begin{array}{cc}a & b \\ c & d \end{array}\right) \left(\begin{array}{c}\cF \\ Y \end{array}\right)=\left(\begin{array}{c}\cF' \\ Y' \end{array}\right)\ ,

3497: \eeq

3498: where $a,b,c$ and $d$ are real integer-valued matrices obeying 

3499: \beq \label{def-M}

3500:   a^T c=c^T a\ ,\qquad \quad b^T d = d^T b\ , \qquad \quad a^T d - c^T b = 1\ .

3501: \eeq

3502: These transformations change the basis of the third cohomology of the Calabi-Yau manifold 

3503: and form the symplectic  group $Sp(H^3,\mathbb{Z})$. Note that in general only 

3504: a subgroup $\Gamma_M$ of $Sp(H^3,\mathbb{Z})$ provides a true 

3505: symmetry of the topological string theory. $\Gamma_M$ is the monodromy 

3506: group. We encountered specific examples for $\Gamma_M$ in the sections on Seiberg-Witten 

3507: theory and the Enriques Calabi-Yau: $\Gamma_M(\text{SW})=\Gamma(2)$ and $\Gamma_M(\text{E})=Sl(2,\bbZ)\times O(10,2,\bbZ)$.

3508: The monodromy group $\Gamma_M$ is a symmetry of all higher genus amplitudes 

3509: $F^{(g)}(Y,\bar Y)$.

3510: 

3511: Given the action of $Sp(H^3,\mathbb{Z})$ on the periods, we can also investigate 

3512: its induced action on the geometrical objects on $\widehat \cM$. 

3513: First note that both K\"ahler potentials $K$ and $\widehat K$ are 

3514: invariant under \eqref{period_rot} since they contain the symplectic 

3515: scalars $Y^K \bar \cF_K - \bar Y^K \cF_K$. The second derivative 

3516: $\tau_{IJ}=\partial_{Y^I} \cF_J$ of the prepotential transforms as

3517: \beq

3518:   \tau \quad \mapsto \quad (a\tau +b)(c\tau +d)^{-1}\ .

3519: \label{tau_{IJ}transformation}

3520: \eeq

3521: This implies that $\tau_{IJ}$ transforms as a modular 

3522: parameter and is the higher-dimensional analog of \eqref{modulartrans}.

3523: Once again one easily shows that the inverse of $\I \tau_{IJ}$ transforms 

3524: with a shift

3525: \beq

3526:    \I \tau^{IJ}\quad \mapsto \quad (c\tau + d)_{\ K}^I (c\tau +d)_{\ L}^J \I \tau^{KL} -2i c^{IK} (c\tau +d)^J_K\ .

3527: \label{imtau^{IJ}transformation}

3528: \eeq

3529: On the other hand, the third derivative $C_{IJK}$ of the prepotential $\cF$ 

3530: transforms without such a shift

3531: \beq

3532:   C_{IJK}\quad \mapsto \quad (c\tau + d)_I^{-1\, M}(c\tau +d)_J^{-1\, N}(c\tau + d)^{-1\, P}_K C_{MNP} 

3533: \eeq

3534: This is precisely the transformation property of a modular form of weight $-3$.  

3535: In general, we say that a modular form is of weight $-n$ if it transforms as

3536: \beq

3537:    M_{I_1 \ldots I_n}\quad \mapsto \quad (c\tau + d)_{I_1}^{-1\, J_1} \ldots (c\tau + d)_{I_n}^{-1\, J_n} M_{J_1 \ldots J_n}\ .

3538: \eeq

3539: The holomorphic form $\Phi$ appearing in \eqref{F_1solution} has no indices, but nevertheless transforms 

3540: under the modular group $\Gamma_M$. It is chosen such that $F^{(1)}$ as well as $f^{(1)}$ are invariant.

3541: This implies that it has to transform as 

3542: \beq

3543:    \Phi\quad \mapsto \quad \det (c \tau +d)\ \Phi

3544: \eeq

3545: to compensate the transformation  of $\det(\I \tau_{IJ})$ in \eqref{F_1solution}. A major challenge is to find the appropriate 

3546: $\Phi$ for a given Calabi-Yau manifold and show that it can be expressed as a function of $\tau_{IJ}$ 

3547: only. In order to do that one can change $f^{(1)}$ by holomorphic modular invariant combinations.

3548: $\Phi(\tau_{IJ})$ can be explicitly derived for the Enriques Calabi-Yau. It is desirable 

3549: to explore further examples such as the quintic Calabi-Yau.

3550: 

3551: As we have seen before, the derivative $\partial_{I_0} M_{I_1 \ldots I_n}$

3552: of a modular form $M$ is no longer a modular form, since the derivative also acts on the matrices $(C\tau + D)_{I_i}^{-1\, J_i}$. 

3553: However, this action can be compensated by using covariant derivatives on $\widehat \cM$. One easily shows 

3554: that the Christoffel symbols \eqref{biggamma} shift under \eqref{period_rot} such that the covariant 

3555: derivative $D_{I_0} M_{I_1 \ldots I_n}$ of a modular form is again a modular form of weight reduced by one.

3556: If we express $M_{K}$ as a function of $\tau_{IJ}$, we can also take derivatives 

3557: \beq\label{cov_Dgen2}

3558:    D^{IJ}M_K\equiv  \partial_{\tau_{IJ}}M_K -  \tfrac{i}{2}\delta_K^{\{J} \I \tau^{I \}L} M_L\ ,\qquad \quad D_I = C_{IJK} D^{JK}\ ,

3559: \eeq

3560: where $\{IJ\}$ indicates symmetrization in the indices $I$ and $J$ with symmetry factor $\frac12$. In order to relate 

3561: $D_I$ and $D^{JK}$ we have used that $\partial_I \tau_{KL}=C_{IKL}$. Since $C_{IKL}$ has weight $-3$

3562: this also implies that $D^{IJ}$ raises the weight of the modular form by $2$.

3563:  Let us note that $D_I$ and 

3564: $D^{IJ}$ are the higher-dimensional analogs of the derivatives $D_t$ and $D_\tau$ displayed in \eqref{der_nonhol}.

3565: 

3566: 

3567: 

3568: 

3569: 

3570: \subsection{Feynman rules for $F^{(g)}$: vertices and the propagators \label{sec:vertprop}}

3571: 

3572: Let us now come back to the discussion of the holomorphic anomaly equations \eqref{big_Fg}.

3573:  As argued in \cite{bcov} and briefly recalled in section \ref{anomaly}, the traditional way of finding a solution to equations 

3574: of the form \eqref{big_Fg} is via a Feynman 

3575: graph expansion.

3576: In this section we will derive the vertices and propagators to describe 

3577: such an expansion in the large moduli space. This can be done by first directly solving 

3578: \eqref{big_Fg} for the smallest possible genus $g=2$. The resulting section $F^{(2)}$

3579: can be identified as a sum over Feynman graphs counting degeneracies of Riemann surfaces.

3580: This example allows us to identify the vertices and propagators, which can be used to 

3581: systematically construct every solution $F^{(g)}$. The generating functional encoding 

3582: these Feynman rules is then derived and can be shown to be equivalent to the generating 

3583: functional of Bershadsky, Cecotti, Ooguri and Vafa \cite{bcov}.

3584:   

3585: In order to extract the solutions for the free energies $F^{(g)}$ we first define  

3586: complex tensors  

3587: \beq \label{def-C^g}

3588:   C^{(g)}_{I_1 \ldots I_n}(Y,\bar Y) \equiv  \left\{ \begin{array}{ll}D_{I_1}\ldots D_{I_n} F^{(g)}(Y,\bar Y) &\ \text{for}\ \ g\ge1\\

3589:                                                                           i D_{I_1}\ldots D_{I_n} C_{I_{n-2} I_{n-1} I_{n}} &\ \text{for}\ \ g=0 \end{array}\right.\ .

3590: \eeq

3591: and demand

3592: \beq\label{def-C^g2}

3593:    C^{(g)}_{I_1 \ldots I_n} = 0 \quad \text{for} \ \ 2g-2 + n \le 0 \ .

3594: \eeq

3595: These two equations are the big moduli space equivalents of \eqref{C_prop} and \eqref{C_prop_cond}. 

3596: They imply that $C^{(g)}_{I_1 \ldots I_n}$ is a section of  $\cL^{2-2g-n}$ such that

3597: we can infer the homogeneity relation

3598: \beq \label{proj_2}

3599:   Y^K C^{(g)}_{K I_1 \ldots I_n} = (2-2g-n) C^{(g)}_{I_1 \ldots I_n}\ .

3600: \eeq

3601:  

3602: In equation \eqref{F_1solution} we already displayed the general local form of

3603: solutions for the free energy $F^{(1)}$. The next function to determine is 

3604: $F^{(2)}(Y,\bar Y)$. Evaluating \eqref{big_Fg} for $g=2$ one obtains 

3605: \beq \label{F_2big}

3606:     \partial_{\bar Y^I} F^{(2)} = -\tfrac{i}{8} \bar C_{I}^{\ JK} \Big( 

3607:     D_J \partial_{K} F^{(1)} + \partial_{J} F^{(1)} \partial_{K} F^{(1)} \Big) \ ,

3608: \eeq

3609: As we discuss in appendix \ref{Cal_big_Fg1} such an equation can be solved by 

3610: an integration by parts method. This amounts 

3611: to writing the right-hand side of \eqref{F_2big} as an anti-holomorphic 

3612: derivative of some expression $\Gamma^{(2)}(Y,\bar Y)$. The solution 

3613: to \eqref{F_2big} is then given by $F^{(2)} = \Gamma^{(2)}(Y,\bar Y) + f^{(2)}(Y)$,

3614: where  $f^{(2)}$ is the holomorphic ambiguity at genus two.

3615: This method of solving \eqref{F_2big} is equivalent to the one used in ref.~\cite{bcov} 

3616: to solve the holomorphic anomaly equations \eqref{rec_Fg} for $F^{(g)}(t,\bar t)$.

3617: However, in contrast to \cite{bcov} it will be sufficient to introduce one type of propagator 

3618: denoted by $\Delta^{IJ}$. The propagator $\Delta^{IJ}$ has to be chosen such that

3619: \beq \label{d_prop}

3620:    \partial_{\bar K} \Delta^{IJ} = \tfrac{i}4 \bar C_K^{\ IJ}\ .

3621: \eeq

3622: Clearly, this fixes the form of $\Delta^{IJ}$ only up to an holomorphic function. 

3623: As for the examples discussed in the previous sections this ambiguity can be fixed by modular invariance and

3624: compatibility with the solution $F^{(1)}$. 

3625: 

3626: In order to derive an explicit expression for the propagator $\Delta^{IJ}$ we note that a solution to \eqref{d_prop} is

3627: always of the form 

3628: \beq

3629: \label{DIJanholomorphic}

3630:   \Delta^{IJ} =-\tfrac12 \I \tau^{IJ} + {\cal E}^{IJ}(Y)\ ,

3631: \eeq

3632: where ${\cal E}^{IJ}(Y)$ is a holomorphic function, which compensates 

3633: the shift transformation \eqref{imtau^{IJ}transformation} of $\I \tau^{IJ}$.

3634: As in the Seiberg-Witten and Enriques example we want to express $\cE^{IJ}$

3635: as a derivative of the holomorphic part of $F^{(1)}$ given in \eqref{F_1solution}.

3636: In order to do that, let us assume that 

3637: we can express $\Phi(Y)$ as a function of $\tau_{IJ}$ itself. To achieve this,

3638: it might be necessary to appropriately split the holomorphic ambiguity of 

3639: $F^{(1)}$ into $\Phi(\tau)$ and an additional function $f^{(1)}(Y)$. 

3640: $f^{(1)}$ is a modular invariant function which might not be 

3641: expressible as a function of $\tau_{IJ}$.

3642: We then identify the 

3643: holomorphic part in \eqref{DIJanholomorphic} to be

3644: \beq

3645:   \cE^{IJ}=  -\frac{i}{\Phi} \frac{\partial \Phi(\tau)}{\partial \tau_{IJ}}\ .

3646: \eeq

3647: {}From this definition one can immediately conclude that $\Delta^{IJ}$ is a modular form of weight 

3648: $2$ under the target space symmetry group $\Gamma_M$.

3649: To see this, note that since $F^{(1)}$ and $K$ are invariant under $\Gamma_M$   

3650: also the section

3651: \beq \label{def-F1tilde}

3652:    \tilde F^{(1)} =- \tfrac{1}{2}\log \det(\I\tau_{IJ}) - \ln |\Phi(\tau)|+f^{(1)}+\bar f^{(1)}\ ,

3653: \eeq

3654: is trivially invariant under $\Gamma_M$. But evaluating the first derivative on the weight 

3655: zero forms $\tilde F^{(1)}$ and $f^{(1)}$ one finds

3656: \beq \label{Deltaasder}

3657:   \partial_I \tilde F^{(1)}=-\tfrac12 C^{(0)}_{IJK} \Delta^{JK}+\partial_I f^{(1)}\ ,

3658: \eeq

3659: and infers from the discussion of section \ref{symp_group} that $\Delta^{IJ}$ is of weight $2$ and does not shift under

3660: $\Gamma_M$.

3661: 

3662: 

3663: Now that we have discussed the propagator $\Delta^{IJ}$, let us turn to the definition of the 

3664: vertices. We do that by continuing the evaluation of the $F^{(2)}$ example.

3665: In appendix \ref{Cal_big_Fg1} we determine by the partial integration method of \cite{bcov} that 

3666: $F^{(2)}(Y,\bar Y)$ to be

3667: \bea \label{F_2_big}

3668:    F^{(2)}(Y,\bar Y)&=& f^{(2)} - \Delta^{JK} \Big( 

3669:    \tfrac12 \tilde C^{(1)}_{JK} + \tfrac12 \tilde C^{(1)}_J \tilde C^{(1)}_K \Big)

3670:    -\Delta^{JK}\Delta^{LM} \Big(\tfrac18 C^{(0)}_{KLMJ} + \tfrac12 C^{(0)}_{JLM}   \tilde C_K^{(1)} \Big)  \nn \\

3671:          &&- \Delta^{JK} \Delta^{LM} \Delta^{QP}\Big(\tfrac{1}{12} C^{(0)}_{KMQ} C^{(0)}_{PLJ} + \tfrac18 C^{(0)}_{KJQ} C^{(0)}_{PML} \Big)  \ ,

3672: \eea

3673: where  $f^{(2)}(Y)$ is the holomorphic ambiguity.

3674: Note that in this expansion we introduced the shifted $F^{(1)}$ vertices 

3675: \beq

3676:     \tilde C^{(1)}_{JK} = C^{(1)}_{JK} + \big(\tfrac{\chi}{24} -1 \big) K_{J} K_K\ ,\qquad \qquad  \tilde C^{(1)}_K = C^{(1)}_K + \big(\tfrac{\chi}{24} -1 \big)  K_{K} \ .

3677: \eeq

3678: It is not hard to interpret the resulting $F^{(2)}$ as being obtained from a Feynman graph expansion. Each term 

3679: in \eqref{F_2_big} corresponds to one Feynman diagram representing a degeneration of a genus $2$ Riemann surface.

3680: The vertices are $C^{(0)}_{IJK},\ C^{(0)}_{IJKL}$ and $\tilde C^{(1)}_{I},\ \tilde C^{(1)}_{IJ}$ which are connected 

3681: by propagators $\Delta^{IJ}$. The whole Feynman sum is shown in Figure \ref{F_2pic}.

3682: 

3683: \begin{figure}[!ht]

3684: \leavevmode

3685: \begin{center}

3686: %\epsfysize=7.5cm

3687: \includegraphics[height=2.4cm]{two.eps}

3688: \end{center}

3689: \caption{The Feynman graph expansion for $F^{(2)}$.}

3690: \label{F_2pic}

3691: \end{figure}

3692: %

3693: 

3694: 

3695: 

3696: 

3697: {}From this example we can also infer the general Feynman rules which generate the solutions $F^{(g)}(Y,\bar Y)$

3698: to the holomorphic anomaly equation \eqref{big_Fg}. The propagator is defined in \eqref{DIJanholomorphic} and \eqref{Deltaasder} as 

3699: the derivative of $\tilde F^{(1)}$. The vertices take the form 

3700: \bea \label{def-tildeC}

3701:   \tilde C^{(g)}_{I_1 \ldots I_n} &=& C^{(g)}_{I_1 \ldots I_n} \qquad \text{for}\quad g \neq 1 \ ,\qquad \qquad  \tilde C^{(1)} = 0\ .\\

3702:   \tilde C^{(1)}_{I_1 \ldots I_n} &=& C^{(1)}_{I_1 \ldots I_n} + (n-1)! \big(\tfrac{\chi}{24} -1 \big) K_{I_1}\ldots K_{I_n}\qquad  \text{for}\quad n >0 \ ,

3703: \eea

3704: where $K_I = \partial_{Y^I}K(Y,\bar Y)$ are the first derivatives of the K\"ahler potential \eqref{kaehlerpotential}.

3705: It is straightforward to check that by using $K_I = - \widehat K_I /\widehat K$ and $D_I \widehat K_J=0$ 

3706: one finds 

3707: \beq

3708:  \tilde C^{(1)}_{I_1 \ldots I_n}= D_{I_1} \ldots D_{I_n}  \tilde F^{(1)} \ ,

3709: \eeq

3710: where $\tilde  F^{(1)} $ is defined in \eqref{def-F1tilde}.

3711:  This Feynman graph expansion

3712: can be obtained as a saddle point expansion of the formal integral

3713: \beq \label{Z[Y]}

3714:     \hat Z\big[Y \big] = \int dZ\ \sqrt{\det \Delta}\ \text{exp}\big(-\tfrac12 g_s^{-2}\ \Delta^{-1}_{IJ}\, Z^I Z^J  +  W[Z;\, Y,\bar Y] \big)\ ,

3715: \eeq

3716: where $g_s$ is the expansion constant playing the role of $\hbar$.

3717: Here $W[Z;\, Y,\bar Y]$ contains the vertices \eqref{def-tildeC} and reads

3718: \bea \label{big_W}

3719:   W[Z;\, Y,\bar Y]&=& \sum_{g=0}^{\infty} \sum_{n=0}^{\infty} \tfrac{1}{n!}\, g_s^{2g-2}\, \tilde C^{(g)}_{I_1\ldots I_n} Z^{I_1} \ldots  Z^{I_n}\\

3720:     &=&  \sum_{g=0}^{\infty} \sum_{n=0}^{\infty} \tfrac{1}{n!}\, g_s^{2g-2}\, C^{(g)}_{I_1\ldots I_n} Z^{I_1} \ldots  Z^{I_n} - \big(\tfrac{\chi}{24} -1 \big) \ln\big(1-Z^I K_I \big)\ .\nn

3721: \eea 

3722: Note that the holomorphic anomaly equations on the big moduli space together with (\ref{def-C^g}) 

3723: and (\ref{def-C^g2}) imply that the integrand of (\ref{Z[Y]}) transforms as a wavefunction 

3724: \cite{Witten:1993ed,Dijkgraaf:2002ac,Verlinde:2004ck,abk}. Moreover, following~\cite{bcov,Verlinde:2004ck} 

3725: one shows that $\hat Z[Y]$ is actually holomorphic 

3726: in $Y^I$. One thus concludes that each coefficient of $g_s^{2g-2}$ in the saddle point 

3727: expansion of $\log \hat Z$ is a holomorphic ambiguity $f^{(g)}(Y)$. On the other hand, 

3728: each coefficient is of the form $F^{(g)}(Y,\bar Y)- \Gamma^{(g)}(Y,\bar Y)$, 

3729: where $\Gamma^{(g)}$ are the Feynman graphs described above.  

3730: We thus solve for $F^{(g)}= \Gamma^{(g)}(Y,\bar Y)+f^{(g)}(Y)$ and find the desired result. 

3731: In the remainder of this section, we will argue that 

3732: the big moduli space formulation is indeed completely 

3733: equivalent to the results obtained in \cite{bcov}.

3734: 

3735: 

3736: 

3737: 

3738: 

3739: Let us now turn to the comparison of the big moduli space formalism with the

3740: standard results of \cite{bcov} reviewed in section \ref{anomaly}.

3741: Firstly, note that we only needed one type of propagator $\Delta^{IJ}$. 

3742: This propagator is related to the propagators $ \hat  \Delta^{ij}, \hat\Delta^i $ and $\hat \Delta $

3743: introduced in \eqref{def-small-Delta} by 

3744: \bea

3745: \label{DIJ}

3746:   \Delta^{IJ}  &=&  \chi^I_i \hat  \Delta^{ij} \chi^J_j -  \chi^I_i \hat\Delta^i X^J -  X^I \hat \Delta^i \chi^J_i + X^I \hat\Delta X^J\ ,\nn \\

3747:     &=&  \left(\begin{array}{cc}X^I  &\chi_i^I  \end{array} \right)

3748:           \left(\begin{array}{cc} \hat \Delta & -\hat \Delta^{j}\\ -\hat \Delta^{i} &\hat \Delta^{ij} \end{array} \right)

3749:           \left(\begin{array}{c} X^J \\  \chi_j^J\end{array} \right)\ ,

3750: \label{DIJM}

3751: \eea

3752: where $\chi_i^I$ is defined in \eqref{exp_Omega}.

3753: To check this identity, we can evaluate the $\bar t^i$-derivative of $\Delta^{IJ}$. Clearly,

3754: from the form \eqref{DIJanholomorphic} we find 

3755: \beq

3756:    \partial_{\bar t^i} \Delta^{IJ}= \tfrac{i}{4} \lambda^{-1} \bar \chi^K_{\bi} \bar C_{K}^{\ \ IJ} \ .

3757: \eeq

3758: Precisely the same equation is obtained by using the identification \eqref{DIJ} the special geometry identities 

3759: \eqref{special_geom} and the derivatives \eqref{def-small-Delta} of the small propagators $ \hat  \Delta^{ij}, \hat\Delta^i $ and $\hat \Delta $.

3760: In other words, we found a non-holomorphic lift of the small propagators $\hat \Delta$ to 

3761: $\widehat \cM$ such that $\Delta^{IJ}$ takes the simple form \eqref{DIJanholomorphic}.

3762: Even though we did not completely specify the holomorphic dependence of $\Delta^{IJ}$

3763: we already notice that all non-holomorphic dependence entirely arises through the inverse 

3764: of $\I \tau_{IJ}$. This already hints to the fact that in the formulation on $\widehat \cM$

3765: we have much better control over the $\bar Y^I$ dependence of each $F^{(g)}(Y,\bar Y)$. In section 

3766: \ref{direct-int} we will show that this fact can be used to directly integrate the holomorphic anomaly 

3767: equations, which provides an efficient and direct method to find $F^{(g)}$. In order to show

3768: that expressions such as \eqref{F_2_big} are completely equivalent to the ones of \cite{bcov}, 

3769: we also need the projection of the vertices $\tilde C^{(g)}_{I_1 \ldots I_n}$.

3770: These vertices are related to the correlation functions 

3771: $C^{(g)}_{i_1 \ldots i_n}$ defined in \eqref{C_prop}  by

3772: \beq \label{proj_1}

3773:   C^{(g)}_{i_1 \ldots i_n}(t,\bar t) = \lambda^{2-2g-n} \chi^{I_1}_{i_1} \ldots \chi^{I_n}_{i_n} \tilde C^{(g)}_{I_1 \ldots I_n}(Y,\bar Y) \ .

3774: \eeq

3775: In order to derive this equation we have used the special geometry relations 

3776: \eqref{special_geom} as well as the scaling behavior of $C^{(g)}_{I_1 \ldots I_n}$ when inserting 

3777: $Y^K ={ \lambda}^{-1} X^{K}(t)$. This equation also holds for $\tilde C^{(1)}_{I_1\ldots I_n}$

3778: since due to \eqref{K_Iproj} the additional terms are zero under the contraction with $\chi^I_{i}$. They 

3779: are however of importance once one contracts $\tilde C^{(g)}_{I_1\ldots I_n}$ by $Y^K$ yielding

3780: \beq

3781:    Y^K \tilde C^{(g)}_{KI_1\ldots I_n} = %(2-2g-n) C^{(g)}_{I_1\ldots I_n} - n! \delta^1_{g} K_{I_1}\ldots K_{I_n}=

3782:                                                             (2-2g -n)\, \tilde C^{(g)}_{I_1 \ldots I_n}\ .

3783: \eeq

3784: By using these identities and the identification \eqref{DIJ} of the propagators 

3785: the expansion \eqref{F_2_big} on $\widehat \cM$ gets transformed into the 

3786: known result of \cite{bcov}. Moreover, also the generating function \eqref{big_W}

3787: reduces to the one found in \cite{bcov} if we identify 

3788: \beq

3789:    Z^I = -\varphi Y^I +  x^i \chi^I_i(Y,\bar Y) \ .

3790: \eeq

3791: This proves that the solutions for $F^{(g)}$ are actually 

3792: identical in the formulation of section \ref{anomaly} and the big

3793: moduli space formalism presented here. As already mentioned above,

3794: the advantage of this new formulation is that all non-holomorphic 

3795: dependence arises entirely through $\I \tau^{IJ}$ in the unified 

3796: propagator $\Delta^{IJ}$ and the covariant derivatives $D_I$.

3797: We will use this fact in the next section to perform a direct integration 

3798: and to derive a closed expression for $F^{(g)}$.

3799: 

3800: 

3801: 

3802: 

3803: \subsection{Direct integration of the holomorphic anomaly \label{direct-int}}

3804: 

3805: 

3806: In this section we make use of the special properties of the big moduli space formulation 

3807: to directly integrate the holomorphic anomaly equations \eqref{big_Fg}.

3808: To begin with, we will argue that every $F^{(g)}$ for $g>1$ can be expressed as 

3809: a finite power series in the propagators $\Delta^{IJ}$ as

3810: \beq \label{Fgexp}

3811:     F^{(g)}(Y,\bar Y) = \sum^{3g-3}_{k=0}  \Delta^{I_1 J_1} \ldots \Delta^{I_k J_k}\ c^{(g)}_{I_1 J_1 \ldots I_k J_k}\ ,

3812: \eeq

3813: where $c^{(g)}$ without indices is the holomorphic ambiguity at genus $g$.

3814: Due to modular invariance of $F^{(g)}$ the coefficients $c^{(g)}_{I_1 J_1 \ldots I_k J_k}(Y)$ are shown to be 

3815: holomorphic modular forms of weight $-2k$ on the big moduli space $\widehat \cM$. All non-holomorphic dependence of 

3816: $F^{(g)}$ arises entirely through $\I \tau^{IJ}$ appearing in the propagators $\Delta^{IJ}$ defined in \eqref{DIJanholomorphic},

3817: It is this fact which will allow us to directly integrate the holomorphic anomaly equations \eqref{big_Fg}.

3818: 

3819: First of all, we have to show that indeed each $F^{(g)}$ for $g>1$ can be written as a power series in 

3820: the propagators $\Delta^{IJ}$ with holomorphic coefficients. We check this for $F^{(2)}$ first. 

3821: $F^{(2)}$ was expressed in \eqref{F_2_big} as a power series in $\Delta^{IJ}$ with coefficients 

3822: containing $\tilde C^{(0)}_{IJKL},\ \tilde C^{(1)}_I$ and  $\tilde C^{(1)}_{IJ}$. {}From their definitions 

3823: \eqref{def-tildeC}, it is clear that these three quantities are not holomorphic. Hence, in order to establish that \eqref{Fgexp} is true for $F^{(2)}$,

3824: they have to be written as power series in $\Delta^{IJ}$. For $\tilde C^{(0)}_{IJKL} \equiv D_I C^{(0)}_{JKL}$ this requires 

3825: that we have to expand the connection $D_I$. Note that $D_I$ contains the Christoffel symbol

3826: $\Gamma^{K}_{IJ}=-\frac{i}{2} C_{IJL}\I \tau^{LK}$ and is only non-holomorphic due to the appearance of $\I \tau^{IJ}$.

3827: However, by using \eqref{DIJanholomorphic} one 

3828: can replace $\I \tau^{LK}$ and split the connection as

3829: \beq \label{Dsplit}

3830:    D_I V_J = \check D_I V_J - C^{(0)}_{IJK} \Delta^{KL} V_L\ ,\qquad \qquad \Gamma^{K}_{IJ}=\check \Gamma^K_{IJ} + C^{(0)}_{IJM} \Delta^{MK}\ ,

3831: \eeq

3832: where we introduced the holomorphic connection

3833: \beq\label{Dhol}

3834:   \check D_I V_J = \partial_I V_J - \check \Gamma^{M}_{IJ} V_M= \partial_I V_J + i C_{IJK} \cE^{KM} V_M\ .

3835: \eeq

3836: The holomorphic connection $\check D_I$ maps holomorphic sections 

3837: $V_K(Y)$ into holomorphic sections $\check D_K V_L(Y)$. Moreover, it maps modular

3838: forms into modular forms, decreasing the weight of the modular form by one. 

3839: $\check D_I$ are the generalizations of the holomorphic covariant 

3840: derivatives \eqref{der_hol} and \eqref{def-hatD} for the Seiberg-Witten example and the Enriques Calabi-Yau.

3841: We can now split $\tilde C^{(0)}_{IJKL}$ into a holomorphic part and a term linear in the propagator

3842: \beq \label{tildeC04}

3843:    \tilde C^{(0)}_{IJKL} = \check D_I  C^{(0)}_{JKL} - \Delta^{MN} \big(C^{(0)}_{IJM} C^{(0)}_{NKL}+  C^{(0)}_{IKM}  C^{(0)}_{NJL}+ C^{(0)}_{ILM} C^{(0)}_{NJK}\big) \ .

3844: \eeq

3845: Clearly, due to the holomorphicity of $C^{(0)}_{IJK}$ both $ \check D_I  C^{(0)}_{JKL}$ and the coefficient of $\Delta^{MN}$ are holomorphic 

3846: functions.

3847: 

3848: Let us now evaluate $\tilde C^{(1)}_I$ and $\tilde C^{(1)}_{IJ}$.

3849: The first derivative $\tilde C^{(1)}_I =\partial_I \tilde F^{(1)}$ was already given in \eqref{Deltaasder}

3850: and shown to have an expansion in the propagator $\Delta^{IJ}$ with 

3851: holomorphic coefficients.

3852: In order to also evaluate the remaining vertices, we will need to take derivatives of $\cE^{IJ}$. 

3853: As in the examples, 

3854: note that the first derivative $\partial_K \cE^{IJ}$ is not a modular form, but rather transforms with a 

3855: shift. These shift transformations can be compensated by adding another 

3856: term quadratic in $\cE^{IJ}$. Indeed, we find that the linear 

3857: combination 

3858: \beq \label{E(Phi)}

3859:   \cE^{KL}_{\ \ \ I} \equiv \partial_I \cE^{KL} - \cE^{KM} \cE^{LN} C^{(0)}_{IMN}\ ,

3860: \eeq 

3861: transforms as a modular form without an additional shift. Not surprisingly, $\cE^{KL}_{\ \ \ I}$ is not the same as 

3862: $\check D_I \cE^{KL}$ but rather the field strength of $\cE^{IJ}$. However, there is another important 

3863: representation of $\cE^{KL}_{\ \ \ I}$ in terms of derivatives of $\Phi(\tau)$. Using \eqref{E(Phi)}

3864: one finds 

3865: \beq

3866:    \cE^{KL}_{\ \ \ I} = \cE^{KLMN}_4 C^{(0)}_{MNI}\ ,

3867: \eeq

3868: where 

3869: \beq

3870:   \cE^{I_1 J_1 \ldots I_k J_k}_{2k} = (-i)^{k}\frac{1}{\Phi} \frac{\partial \Phi(\tau)}{\partial \tau_{I_1 J_1} \ldots \partial \tau_{I_k J_k}}\ .

3871: \eeq

3872: These holomorphic modular forms of weight $2k$ are the direct generalizations of the forms $\epsilon^{2k}_{a_1\ldots a_k}$ introduced 

3873: in \eqref{epsilon2k}. A direct calculation shows that we can also express the holomorphic 

3874: modular derivative of $\Delta^{IJ}$ as a propagator expansion,

3875: \beq \label{dDelta}

3876:   %\partial_I \Delta^{KL} &=& -\Delta^{KM} \Delta^{LN} C^{(0)}_{IMN} - \Delta^{KM} \check \Gamma^L_{IM} - \Delta^{LM} \check \Gamma^K_{IM}

3877:      %                                      + \cE^{KLMN}_4 C^{(0)}_{MNI} \ ,\nn \\

3878:  \check D_I \Delta^{KL} =- \Delta^{KM} \Delta^{LN} C^{(0)}_{MNI} 

3879:                                            + \cE^{KLMN}_4 C^{(0)}_{MNI}\ .

3880: \eeq

3881: 

3882: We are now in the position to evaluate the vertex $\tilde C^{(1)}_{IJ} \equiv D_J \partial_I \tilde \cF^{(1)}$.

3883: Using the derivatives \eqref{dDelta} of the propagators together 

3884: with \eqref{Dsplit}, one easily derives 

3885: \bea \label{tildeC^1_IJ}

3886:     \tilde C^{(1)}_{IJ}  &=&-\tfrac12 \cE^{KLMN}_4 C^{(0)}_{MNJ} C^{(0)}_{IKL} + \check D_J\partial_I f^{(1)}

3887:     -\tfrac12 \Delta^{KL}\big( \check D_J C^{(0)}_{IKL} + 2 C^{(0)}_{IJK} \partial_L f^{(1)}\big)

3888:       \nn \\

3889:      &&\qquad 

3890:      + \tfrac{1}{2}\Delta^{KL} \Delta^{MN} \big(C^{(0)}_{JI M} C^{(0)}_{NKL} +  C^{(0)}_{J K M} C^{(0)}_{NIL}\big)\ .

3891: \eea

3892: Inserting \eqref{Deltaasder},  \eqref{tildeC04} and \eqref{tildeC^1_IJ} into the expansion \eqref{F_2_big} for 

3893: $F^{(2)}$ one finds 

3894: \bea \label{F2exp}

3895:    F^{(2)}& =&  \Delta^{I_1 J_1} \Delta^{I_2 J_2} \Delta^{I_3 J_3}\big( \tfrac{1}{24} C^{(0)}_{I_1 I_2 I_3} C^{(0)}_{J_1 J_2 J_3} +  \tfrac{1}{16} C^{(0)}_{I_1 J_1 I_2} C^{(0)}_{J_2 I_3 J_3}  \big) \\

3896:     && - \tfrac18 \Delta^{I_1 J_1} \Delta^{I_2 J_2}\big(\check D_{I_1} C^{(0)}_{J_1 I_2 J_2}

3897:    + 4 C^{(0)}_{I_1 J_1 I_2} \partial_{J_2} f^{(1)} \big)\nn \\

3898:      && - \tfrac14 \Delta^{I_1 J_1} \cE_4^{KLMN} C^{(0)}_{I_1 MN} C^{(0)}_{J_1KL}  

3899:       + \tfrac12 \check D_{I_1} \partial_{J_1} f^{(1)} + \tfrac{1}2 \partial_{I_1} f^{(1)} \partial_{J_1} f^{(1)}+c^{(2)}\ .\nn

3900: \eea

3901: This shows that the calculation of $F^{(2)}$ using the partial integration and the expansion of the non-holomorphic

3902: coefficients yields the desired expansion \eqref{Fgexp} of $F^{(2)}$. 

3903: We will now use an inductive argument to show that every $F^{(g)}$ can be 

3904: written in the form \eqref{Fgexp} and derive a recursive expression 

3905: by direct integration.

3906: 

3907: Let us now go one step further and show that if all $F^{(r)}$  for $1<r<g$ 

3908: can be written in the form \eqref{Fgexp} also $F^{(g)}$ itself 

3909: admits this expansion. To do that, we use the Feynman graph expansion 

3910: introduced in section \ref{sec:vertprop}. It was shown there that each 

3911: $F^{(g)}(Y,\bar Y)$ can be obtained from vertices $\tilde C^{(r)}_{I_1\ldots I_k}$, $r<g$

3912: connected with propagators $\Delta^{IJ}$. But it is not hard to see that $\tilde C^{(r)}_{I_1\ldots I_k}$

3913: is actually an expansion in $\Delta^{IJ}$ with holomorphic coefficients. More precisely, note that 

3914: the vertices are obtained 

3915: by taking covariant derivatives $D_I$ of $F^{(r)}$ and 

3916: we can apply \eqref{Dsplit} to rewrite these into holomorphic 

3917: covariant derivatives $\check D_I$ and a propagator contribution. But since by our

3918: induction assumption $F^{(r)}$ is of the form \eqref{Fgexp} for $r<g$ we can apply \eqref{dDelta} 

3919: to show that this is equally true for $F^{(g)}$ itself. This proves that \eqref{Fgexp} is true for all $g>1$.

3920: It is also straightforward to 

3921: count  the number of propagators arising in the expansion \eqref{Fgexp}. 

3922: One simply notes that the term in the Feynman graph expansion 

3923: with coefficients $C^{(0)}_{I JK}$ only is already an

3924: expansion in $\Delta^{IJ}$ with holomorphic coefficients. It has the 

3925: maximal number of propagators, namely $3g-3$.

3926: 

3927: Having shown that $F^{(g)}$ can be always brought to the form

3928: \eqref{Fgexp}, let us now determine a closed expression by direct integration.

3929: Since all non-holomorphic dependence arises through $\Delta^{IJ}$, the 

3930: holomorphic anomaly equation can be rewritten as 

3931: \beq \label{reg_Fgtointegrate}

3932:   \frac{\partial F^{(g)}}{\partial \Delta^{IJ}}=\tfrac12 D_I \partial_J F^{(g-1)} + \tfrac12 \sum^{g-1}_{r=1} \partial_I F^{(r)} \partial_J F^{(g-r)}\ . 

3933: \eeq

3934: To integrate this expression we introduce the following shorthand notation

3935: \beq \label{m_kex}

3936:      F^{(g)}(Y,\bar Y) = \sum^{3g-3}_{k=0}  c^{(g)}_{\ (k)} \ , \qquad \quad  

3937:      c^{(r)}_{\ (k)} = \Delta^{I_1 J_1} \ldots \Delta^{I_k J_k}\,  c^{(r)}_{I_1 J_1 \ldots I_k J_k}\ ,

3938: \eeq

3939: where $c^{(g)}_{\ (k)}$ is the term containing $k$ propagators $\Delta^{IJ}$.

3940: We also rewrite the right-hand side of the holomorphic anomaly equation 

3941: as 

3942: \beq

3943:   D_I \partial_J F^{(g-1)} + \partial_I \tilde F^{(1)} \partial_J F^{(g-1)} + \partial_I F^{(g-1)} \partial_J \tilde F^{(1)}

3944:                        + \sum_{r=2}^{g-2} \partial_I F^{(r)} \partial_J F^{(g-r)}\ .

3945: \eeq 

3946: Here the first three terms can be rewritten as 

3947: \bea \label{DDexpression}

3948:   && D_I \partial_J F^{(g-1)} + \partial_I \tilde F^{(1)} \partial_J F^{(g-1)} + \partial_I F^{(g-1)} \partial_J \tilde F^{(1)} \\

3949:   &&\ = \check D_I \partial_J F^{(g-1)} 

3950:          -\Delta^{KL} C^{(0)}_{IJK} \partial_{L} F^{(g-1)} -\Delta^{KL} C^{(0)}_{KL\{I} \partial_{J\} } F^{(g-1)} + 2 \partial_{\{ I} f^{(1)} \partial_{J\} } F^{(g-1)} \ , \nn

3951: \eea

3952: where we have applied \eqref{Dsplit} and inserted \eqref{Deltaasder}.

3953:  We also introduce the derivative $\check d$,

3954: which acts on the coefficients of the $\Delta$-expansion as the holomorphic covariant derivative 

3955: $\check D_I$ but leaving $\Delta^{IJ}$ invariant. For $c^{(g)}_{\ (k)}$ given in \eqref{m_kex} we thus set

3956: \beq

3957:   \check d_{I} c^{(g)}_{\ (k)}=\Delta^{I_1 J_1} \ldots \Delta^{I_k J_k}\,  \check D_I c^{(g)}_{I_1 J_1 \ldots I_k J_k}(Y) 

3958: \eeq

3959: Using this definition, we calculate

3960: \bea \label{firstbig_der}

3961:   \partial_I c^{(g)}_{\ (k)}&=&  \check d_I c^{(g)}_{\ (k)}  + \big(\check D_I \Delta^{MN} \big) \frac{\partial }{\partial \Delta^{MN}} c^{(g)}_{\ (k)}   \\

3962:                    &=& \check d_I c^{(g)}_{\ (k)}

3963:                    + C^{(0)}_{IPQ}\big(\cE^{PQMN}_4  - \Delta^{PM} \Delta^{QN} \big)    \frac{\partial }{\partial \Delta^{MN}} c^{(g)}_{\ (k)}\ . 

3964: \eea

3965: Note that the first term is homogeneous of degree $k$ in $\Delta$, the second is homogeneous of degree $k-1$, while the last is 

3966: homogeneous of degree $k+1$. 

3967: We also evaluate the second derivative

3968: \bea \label{secondbig_der}

3969:  \check D_I \partial_J c^{(g-1)}_{\ (k)} &=& 

3970:  \Big[\cE_4^{MNQP} \cE_4^{TURS} C^{(0)}_{IMN} C^{(0)}_{JTU} \Big] \frac{\partial^2}{\partial \Delta^{QP} \partial \Delta^{RS}} c^{(g-1)}_{\ (k)}  \\

3971: &+&\Big[\cE^{CDKLMN}_6 C^{(0)}_{KLI} C^{(0)}_{MNJ}  + \cE^{CDFG}_4 \check  C^{(0)}_{JFGI}+ 2  \cE^{CDFG}_4 C^{(0)}_{FG\{I}\check d^{\phantom{(}}_{J\}} \Big] \frac{\partial}{\partial \Delta^{CD}} c^{(g-1)}_{\ (k)}\nn \\

3972: &+&\Big[\check d_I  \check d_J - 2 \cE^{QCFG}_4  C^{(0)}_{FG\{ I} C^{(0)}_{J \}QB} \Delta^{BD} \frac{\partial}{\partial \Delta^{CD}} \nn \\

3973:   &&    - 2 \cE^{CDEF}_4 C^{(0)}_{MN\{I} C^{(0)}_{J\}EF} \Delta^{MQ} \Delta^{NP}  \frac{\partial^2}{\partial \Delta^{QP} \partial \Delta^{CD}} \Big] c^{(g-1)}_{\ (k)} \nn \\

3974:   &-& \Big[\check C^{(0)}_{IJAB}+2 C^{(0)}_{AB \{I} \check d^{\phantom{(}}_{J\} } \Big] \Delta^{AC} \Delta^{BD}  \frac{\partial}{\partial \Delta^{CD}} c^{(g-1)}_{\ (k)}\nn\\

3975:   &+& \Big[ 2 C^{(0)}_{IMN} C^{(0)}_{JQB} \Delta^{QM} \Delta^{NC} \Delta^{BD}  \frac{\partial}{\partial \Delta^{CD}}  \nn \\

3976:   &&       +   C^{(0)}_{IMN} C^{(0)}_{JAB}  \Delta^{QM} \Delta^{NP} \Delta^{AC} \Delta^{BD}   \frac{\partial^2}{\partial \Delta^{QP} \partial \Delta^{CD}}  \Big] c^{(g-1)}_{\ (k)}\nn\ ,

3977: \eea

3978: where $\{IJ\}$ indicates the symmetrization of the indices and we abbreviated 

3979: \beq

3980:    \check C^{(0)}_{IJKL}= \check D_I C^{(0)}_{JKL}\ .

3981: \eeq

3982: Once again, we can specify the $\Delta$-homogeneity of the terms: first line $k-2$, second line $k-1$, third and fourth line $k$,

3983: fifth line $k+1$, sixth and seventh line $k+2$. In performing the direct integration of the holomorphic anomaly equation

3984: we keep track of the number of propagators on the right-hand side of \eqref{reg_Fgtointegrate}. We can do this explicitly 

3985: by inserting \eqref{DDexpression} together with \eqref{firstbig_der} and \eqref{secondbig_der} 

3986: into \eqref{reg_Fgtointegrate}. Due to the vast number of indices 

3987: the result looks rather complicated and will be presented in the following. 

3988: 

3989: 

3990: Performing the direct integration one finds

3991: \bea

3992:   F^{(g)} &=& \tfrac12 \sum_{k=0}^{3g-6}\Big[ \tfrac{1}{k-1}

3993:   \cE_4^{MNQP} \cE_4^{TURS} C^{(0)}_{IMN} C^{(0)}_{JTU} \Delta^{IJ} \frac{\partial^2}{\partial \Delta^{QP} \partial \Delta^{RS}}  \nn \\

3994: &+&\tfrac{1}{k}\Big(\cE^{CDKLMN}_6 C^{(0)}_{KLI} C^{(0)}_{MNJ}  + \cE^{CDFG}_4 \check C^{(0)}_{JFGI}

3995:        + 2  \cE^{CDFG}_4 C^{(0)}_{FGI}\big(\check d^{\phantom{(}}_{J} +  \partial_{J} f^{(1)}\big) \Big) 

3996:          \Delta^{IJ} \frac{\partial}{\partial \Delta^{CD}} \nn \\

3997: &+&\tfrac{1}{k+1}\Big( \big(\check d_I + \partial_{I} f^{(1)} \big) \big(\check d_J + \partial_{J} f^{(1)} \big) \Delta^{IJ} - \partial_{I} f^{(1)}   \partial_{J} f^{(1)} \Delta^{IJ} \nn \\

3998: &&- 2 \big( \cE^{QCFG}_4  C^{(0)}_{FGI} C^{(0)}_{J QB}  \Delta^{IJ} \Delta^{BD}+\cE^{PQCD}_4  C^{(0)}_{IJK}  C^{(0)}_{LPQ} \Delta^{IJ} \Delta^{KL} \big) \frac{\partial}{\partial \Delta^{CD}} \nn \\

3999:   && \qquad\   - 2 \cE^{CDEF}_4 C^{(0)}_{MNI} C^{(0)}_{JEF}  \Delta^{IJ}\Delta^{MQ} \Delta^{NP}  \frac{\partial^2}{\partial \Delta^{QP} \partial \Delta^{CD}} \Big)  \nn \\

4000:   &-& \tfrac{1}{k+2}\Big(2 C^{(0)}_{IJK}      \check d_L  \Delta^{IJ} \Delta^{KL}  + \big(\check C^{(0)}_{IJAB}+2 C^{(0)}_{ABI}( \check d^{\phantom{(}}_{J} + \partial_J f^{(1)})\big) \Delta^{IJ} \Delta^{AC} \Delta^{BD}  \frac{\partial}{\partial \Delta^{CD}} \Big) \nn\\

4001:   &+&\tfrac{1}{k+3} \Big( 2 \big( C^{(0)}_{IJK}     C^{(0)}_{LPQ}\Delta^{IJ}  \Delta^{KL} \Delta^{PC} \Delta^{QD}  + C^{(0)}_{IMN} C^{(0)}_{JQB} \Delta^{IJ} \Delta^{QM} \Delta^{NC} \Delta^{BD}\big)\frac{\partial }{\partial \Delta^{CD}} \nn \\

4002:   &&    \qquad\     +   C^{(0)}_{IMN} C^{(0)}_{JAB} \Delta^{IJ} \Delta^{QM} \Delta^{NP} \Delta^{AC} \Delta^{BD}   \frac{\partial^2}{\partial \Delta^{QP} \partial \Delta^{CD}}  \Big) \Big] c^{(g-1)}_{\ (k)}\nn \\

4003:   &+&\tfrac12 \sum_{r=2}^{g-2} \sum_{k=0}^{3g-6} \sum_{m+n=k} \Big[  \tfrac{1}{k+1} \Delta^{IJ} \big(\check d_I   c^{(g-r)}_{\ (m)} \big) \big(\check d_J c^{(r)}_{\ (n)}\big)\nn \\

4004:   &+&\tfrac{1}{k-1}\cE^{PQMN}_4 \cE^{RSTU}_4  C^{(0)}_{IPQ}  C^{(0)}_{JRS} \Delta^{IJ} \Big(    \frac{\partial }{\partial \Delta^{MN}}   c^{(g-r)}_{\ (m)} \Big) 

4005:    \Big(    \frac{\partial }{\partial \Delta^{TU}}   c^{(r)}_{\ (n)} \Big)\nn\\

4006:   &+&  \tfrac{1}{k+3}  C^{(0)}_{IPQ}  C^{(0)}_{JRS}  \Delta^{IJ} \Delta^{PM} \Delta^{QN}\Delta^{PT}\Delta^{SU} \Big(    \frac{\partial }{\partial \Delta^{MN}}   c^{(g-r)}_{\ (m)} \Big) 

4007:    \Big(    \frac{\partial }{\partial \Delta^{TU}}   c^{(r)}_{\ (n)} \Big) \nn\\

4008:    &+&\tfrac{1}{k} \cE^{RSTU}_4 C^{(0)}_{JRS} \Delta^{IJ}   \Big\{ \big(\check d_I   c^{(g-r)}_{\ (m)} \big)    \Big( \frac{\partial }{\partial \Delta^{TU}}   c^{(r)}_{\ (n)} \Big)  +

4009:                                \big(  \frac{\partial }{\partial \Delta^{TU}}   c^{(g-r)}_{\ (m)} \big)    \Big(\check d_I  c^{(r)}_{\ (n)} \Big)    \Big\} \nn \\

4010:    &-&\tfrac{1}{k+2} C^{(0)}_{JRS} \Delta^{IJ} \Delta^{PT} \Delta^{SU}  \Big\{ \big(\check d_I   c^{(g-r)}_{\ (m)} \big)    \Big( \frac{\partial }{\partial \Delta^{TU}}   c^{(r)}_{\ (n)} \Big)  +

4011:                                \big(  \frac{\partial }{\partial \Delta^{TU}}   c^{(g-r)}_{\ (m)} \big)    \Big(\check d_I  c^{(r)}_{\ (n)} \Big)    \Big\}                                \Big]\nn \\

4012:    &-&\tfrac{1}{k+1}  \cE^{PQMN}_4 C^{(0)}_{IPQ} C^{(0)}_{JRS} \Delta^{IJ} \Delta^{PT} \Delta^{SU} \Big\{

4013:           \Big(    \frac{\partial }{\partial \Delta^{MN}}   c^{(g-r)}_{\ (m)} \Big)  \Big(    \frac{\partial }{\partial \Delta^{TU}}   c^{(r)}_{\ (n)} \Big) \nn\\

4014:           &&\qquad +

4015:           \Big(     \frac{\partial }{\partial \Delta^{TU}}   c^{(g-r)}_{\ (m)} \Big)  \Big(   \frac{\partial }{\partial \Delta^{MN}}   c^{(r)}_{\ (n)} \Big)     \Big\} \Big] + c^{(g)}_{\ (0)}\ .

4016: \eea

4017: Let us end with some brief remarks about the properties of the direct integration in the big moduli space.

4018: Firstly, we note that the building blocks of $F^{(g)}$ are the propagators $\Delta^{IJ}$ as well as the holomorphic modular forms 

4019: \beq \label{hol_mod_forms}

4020:    \cE^{I_1\ldots J_k}_{2k}\ ,\qquad \check D_{I_1}\ldots \check D_{I_k} f^{(1)}\ ,\qquad \check D_{I_1}\ldots \check D_{I_k} C^{(0)}_{KLM}\ ,

4021: \eeq

4022: induced by $F^{(1)}$ and $F^{(0)}$. It seems likely that also the holomorphic 

4023: ambiguity can be parametrized by \eqref{hol_mod_forms}. To determine these forms it is essential to 

4024: find $\Phi(\tau)$, which will be harder for examples other than the Enriques Calabi-Yau. 

4025: Moreover, in order to efficiently derive all $F^{(g)}$ one also needs to show that the 

4026: forms \eqref{hol_mod_forms} are generated by a finite number of holomorphic modular forms of $\Gamma_M$.

4027: Clearly, the most challenging task is then to fix the ambiguity by appropriate boundary conditions.

4028: To explore these issues for other interesting examples will be left for further work.

4029: 

4030: \section{Conclusion and Outlook}\label{conclusion}

4031: 

4032: In this paper we have developed a new approach to solving the holomorphic anomaly equations of \cite{bcov}, 

4033: based on the interplay between modularity and non-holomorphicity, which makes possible to perform 

4034: a direct integration of the equations at each genus. This approach is more efficient than the diagram 

4035: expansion of \cite{bcov} and leads to closed expressions for the topological string amplitudes, 

4036: once the ambiguities are fixed by appropriate boundary conditions. The amplitudes obtained with this procedure  

4037: can be written as polynomials in a finite set of generators that transform in a particularly simple way under 

4038: the space-time symmetry group, making the modularity properties manifest. 

4039: There are many open questions and possible avenues for future work. We list here some of them. 

4040: 

4041:  Although we have been able to improve the results of \cite{km} for the Enriques Calabi--Yau manifold, 

4042: it would be interesting to push further the formalism developed in this paper. In section \ref{sec:diE} we have introduced 

4043: a set of holomorphic automorphic forms on the Enriques moduli space which might be enough to parametrize the holomorphic 

4044: ambiguity. Using these forms, the boundary conditions obtained from the field theory and the fiber limits, and some extra 

4045: information coming for example from Gromov--Witten theory, one might be able to obtain the topological string amplitudes 

4046: at higher genus. 

4047: 

4048: As explained in \cite{kkvbh}, Gopakumar--Vafa invariants should provide a microscopic counting of degrees of freedom for 

4049: 5d spinning black holes, although in order to make contact with the macroscopic Bekenstein--Hawking entropy one typically 

4050: needs a knowledge of these invariants (therefore of the topological string amplitudes $F^{(g)}$) at arbitrary high genus. 

4051: Some of the results of this paper might be useful in studies of these black holes. For example, the all--genus fiber result 

4052: for the Enriques Calabi--Yau manifold should give a detailed microscopic counting for small 5d black holes obtained by 

4053: wrapping M2 branes in the Enriques fiber.

4054: 

4055: Vast progress has been made in the understanding of compactifications which allow to stabilize many or

4056: all moduli in $\cN=1$ supersymmetric vacua \cite{Douglas:2006es}. 

4057: These vacua often rely on the inclusion of background fluxes 

4058: and D brane instanton effects. Orientifolds of the Enriques Calabi-Yau might serve as 

4059: very controllable examples in which certain corrections to the $\cN=1$ 

4060: low energy effective theory can be derived. In particular, it is an interesting 

4061: task to identify and compute corrections to the four-dimensional super- and K\"ahler potentials 

4062: encoded by the higher genus amplitudes.

4063:   

4064: 

4065: As shown in  \cite{hk,emo} the free energies of matrix models satisfy the holomorphic 

4066: anomaly conditions. Hence, the techniques of this paper could lead to a useful method to analyze matrix models. Using 

4067: matrix model technology one can also write down holomorphic anomaly equations for open string amplitudes in local Calabi--Yau 

4068: manifolds \cite{emo}, and it would be interesting to study them using the methods of this paper. In view of the results of 

4069: \cite{mmopen}, this could lead to a powerful  approach to compute open string amplitudes on toric Calabi--Yau manifolds. 

4070: 

4071: 

4072: 

4073: 

4074: 

4075: \section*{Acknowledgments}

4076: We would like to thank Mina Aganagic, Richard Borcherds, Vincent Bouchard, Robbert Dijkgraaf, 

4077: Ian Ellwood, Eberhard Freitag, Min-xin Huang, Peter Mayr, Nikita Nekrasov and Don Zagier, 

4078: for very  valuable discussions and correspondences. 

4079: M.M.~would like to thank Greg Moore for conversations on the field theory limit of the 

4080: FHSV model in 1998. A.K.~and T.G.~are supported by DOE Grant 

4081: DE-FG-02-95ER40896. M.W.~is supported by the Marie Curie EST program. This work has been initiated during a stay of A.K., M.M.~and T.G.~at 

4082: the MSRI in Berkeley. 

4083: 

4084: \bigskip \bigskip

4085: 

4086: \appendix

4087: 

4088: \noindent {\bf \Large Appendices}

4089: 

4090: \section{$\cN=2$ special geometry \label{N=2sp}}

4091: 

4092: In this appendix we summarize some basics about

4093: $\cN=2$ special geometry~\cite{CandelasdellaOssa,N=2rev,Craps:1997gp,Freed}.

4094: Let $Y$ be a Calabi-Yau threefold, i.e.~a complex three-dimensional K\"ahler 

4095: manifold with $SU(3)$ or $SU(2)\times \bbZ_2$, but no smaller, holonomy 

4096: group.  In particular $Y$ has a 

4097: no-where vanishing holomorphic three-form $\Omega$, which is unique 

4098:  up to complex rescaling. $\Omega$ depends on the complex structure 

4099: of $Y$ and hence varies over the space of complex structure deformations $\cM$.

4100: Local  coordinates on $\cM$ are denoted by $t^i,\bar t^i$. $\Omega(t)$ can be used to define a 

4101: K\"ahler potential 

4102: \beq \label{Kpot}

4103: K(t,\bar t)= - \log\big[i \int_Y \Omega \wedge \bar \Omega \big]

4104: %=:-\log \langle \bar\Omega, \Omega\rangle \ . 

4105: \eeq

4106: $K$ induces the following K\"ahler metric structures on $\cM$ 

4107: \begin{equation}

4108: \begin{array}{rl}

4109: G_{i\bar \jmath}&=\partial_i\partial_{\bar \jmath} K, 

4110: \quad \ 

4111: \Gamma^{k}_{ij}=G^{k\bar l}\partial_i G_{j\bar l},

4112: \quad  \ 

4113: \Gamma^{\bar k}_{\bar \imath \bar \jmath}=G^{l\bar k}{\bar \partial}_{\bar \imath} G_{l\bar \jmath}\\[0.2cm]

4114: R_{i \bar \jmath k \bar l}&=-\partial_i \bar \partial_{\bar \jmath} G_{k\bar l}+

4115: G^{m\bar n} (\partial_i G_{k\bar n})(\bar \partial_{\bar \jmath} G_{m\bar l}), 

4116: \quad \ \ 

4117: R_{i\bar j k}^{\phantom{i \bar \jmath k} l}=-\bar \partial_{\bar \jmath} \Gamma^l_{ik}

4118: \\[0.2cm]

4119: R_{i\bar \jmath}&\equiv G^{k\bar l}R_{i\bar \jmath k\bar l}=-\partial_i 

4120: \bar \partial_{\bar \jmath} \log \det(G_{i\bar \jmath})\ .

4121: \end{array}

4122: \label{kaehlersimplifications}

4123: \end{equation}

4124: $\Omega$ and $\bar \Omega$ are sections of holomorphic and anti-holomorphic 

4125: lines bundles ${\cal L}$ and ${\overline  {\cal L}}$ over ${\cal M}$ 

4126: respectively and holomorphic gauge transformations 

4127: $\Omega\rightarrow e^f\Omega$  in ${\cal L}$ correspond to 

4128: K\"ahler transformations, i.e. $e^{-K}\in {\cL}\otimes {\overline {\cL}}$.  

4129: The derivatives $\partial_i$ are with respect to coordinates $t_i$ of ${\cal M}$, 

4130:  and sections like $V_{j\bar \jmath}$ in 

4131: $T\cM^*_{(1,0)}\otimes T\cM^*_{(0,1)} 

4132: \otimes {\cal L}^m  \otimes {\overline  {\cal L}}^{n}$ have a natural 

4133: connection with respect to the Weil-Petersson metric $G_{i\bar \jmath}$ and the 

4134: line bundle $K_i=\partial_i K$, $K_{\bar \imath} =\partial_{\bar \imath } K$

4135: \begin{equation}

4136: \label{cov_D}

4137: D_i V_{j\bar\jmath}=\partial_i- \Gamma_{ij}^l V_{l\bar \jmath}+ m K_i V_{j\bar \jmath},\quad

4138: D_{\bar \imath} V_{j\bar\jmath}=\partial_{\bar \imath}- 

4139: \Gamma_{\bar \imath \bar \jmath}^{\bar l} V_{j \bar l}+ n K_{\bar \imath} V_{j\bar \jmath}\ .

4140: \end{equation}

4141: For a given complex structure $\Omega$ defines a Hodge decomposition 

4142: \beq

4143:   H^{3}(Y,\bbC) = H^{(3,0)} \oplus H^{(2,1)} \oplus H^{(1,2)} \oplus   H^{(0,3)}\ .

4144: \label{hodgedecomposition}

4145: \eeq

4146: The forms $\Omega$, $\chi_i\equiv D_i\Omega$, ${\overline\chi}_{\bar \imath} 

4147: \equiv D_{\bar \imath} \bar \Omega$ and $\bar \Omega$ provide a basis which 

4148: spans the above cohomology groups over $\bbC$. Since it depends on the complex

4149: structure we call it the moving basis.  

4150: By Kodaira theory, infinitesimal deformations of the complex structure  

4151: are elements of  $H^1(Y,TY)$. $\Omega$ induces an isomorphism 

4152: $H^1(Y,TY)\sim H^{(2,1)}(Y)$.  Hence the harmonic  

4153: $(2,1)$-forms $\chi_i$, $i=1,\ldots,h^{21}$ can be identified as (co)tangent 

4154: vectors to ${\cal M}$ and these deformations are unobstructed on a CY 

4155: manifold \cite{TianTodorov}.

4156: 

4157: We also introduce a fixed integer symplectic basis $(\alpha_K,\beta^L)$ of $H^{3}(Y,\bbZ)$ with 

4158: \begin{equation} \label{integralsymplecticbasis}

4159: \int_Y \alpha_K \wedge \beta^L =- \int_Y \beta^L \wedge \alpha_K = \delta_K^L\ , 

4160: \quad \qquad

4161: \int_Y \alpha_K \wedge \alpha_L = \int_Y \beta^K \wedge \beta^L = 0\ ,

4162: \end{equation} 

4163: which is independent of the complex structure. We can expand the moving 

4164: basis in terms of the fixed basis 

4165: \beq \label{exp_Omega}

4166:   \Omega = X^I \alpha_I - \cF_I \beta^I \ , \qquad \chi_i = 

4167: \chi_i^I \alpha_I - \chi_{Ii} \beta^I\ , \qquad \text{etc\ .}

4168: \eeq

4169: The  expansion coefficients are the period integrals 

4170: \begin{equation}

4171: \label{periods}

4172: X^I=\int_{A^I} \Omega, \quad \quad F_I=\int_{B_I}\Omega,\quad \quad \chi^I_i=\int_{A^I}\chi_i, 

4173: \quad \quad \chi_{I\, i}=\int_{B_I}\chi_i\ ,

4174: \end{equation}  

4175: where $(A^K,B_I)$ is a basis  of $H_{3}(Y,\bbZ)$ dual to $(\alpha_K,\beta^L)$.

4176: Using (\ref{integralsymplecticbasis}) and (\ref{periods}) we can express (\ref{Kpot}) in terms of the periods

4177: \beq 

4178: \label{KpotII}

4179:   K=- \log i \big[ \bar X^K \cF_K  -  X^K \bar \cF_K\big] \ .

4180: \eeq

4181: Note that $X^I\in \cL$, ${\overline \cF}_I\in {\overline \cL}$, 

4182: $\chi^I_i \in T^*_{(1,0)}\cM \otimes \cL$ etc. 

4183: Obviously the periods carry the information about 

4184: the complex structure deformations. The  $X^I$, $I=0,\ldots h^{21}$ 

4185: can serve locally as homogeneous coordinates on ${\cal M}$. 

4186: Local special coordinates on ${\cal M}$ are 

4187: defined by $t^i=X^i/X^0$, $i=1,\ldots h^{(2,1)}$. 

4188: The $\cF_I$ on the other hand are not independent. It follows rather 

4189: from 

4190: %

4191: \be

4192: \int_Y\Omega \wedge {\partial \over \partial X^I}\Omega=0

4193: \ee

4194: % 

4195: that there is a holomorphic section $\cF$ of ${\cL}^{2}$   

4196: called prepotential obeying

4197: %

4198: \be

4199: \cF=\tfrac{1}{2}  X^I F_I\ ,\qquad \quad \cF_I = \partial_{X^I} \cF\ .

4200: \ee

4201: %

4202: This also implies that $\cF(X)$ is homogeneous of degree two in $X^I$.

4203: In special coordinates $t^i$ one also writes $\cF(t)=

4204: (X^0)^{-2}{\cF(X)}$.

4205:  It turns out to be useful to introduce the 

4206: second and third derivative of the prepotential as

4207: \beq

4208:    \tau_{IJ} = \partial_I \partial_J \cF\ ,\qquad \quad C_{IJK} = \partial_I \partial_J \partial_K \cF\ ,

4209: \eeq

4210: which are homogeneous of degree zero and minus one respectively.

4211: 

4212: 

4213: 

4214: Special K\"ahler geometry describes the relation between the metric structure 

4215: and the Yukawa coupling 

4216: \beq

4217: \label{def-CC}

4218:   C^{(0)}_{ijk} \equiv iC_{ijk} \equiv - \int_Y \Omega \wedge 

4219: \partial_i \partial_j \partial_k \Omega= - \int_Y \Omega \wedge D_i D_j D_k \Omega\ ,

4220: %i\langle \Omega, \partial_i \partial_j \partial_k \Omega\rangle=

4221: \eeq

4222: a section of $C_{ijk}\in {\rm Sym}^3 (T^*_{(1,0)})\otimes \cL^2$.

4223: Using $\langle \chi_i ,\bar \chi_{\bar \imath} \rangle =G_{i\bar \jmath} e^{-K}$ 

4224: and transversality of $\langle,\rangle$ under the the decomposition 

4225: (\ref{hodgedecomposition}), i.e. $\langle \gamma_{(k,l)},\gamma_{(m,n)} \rangle=0$ unless

4226: $k+m=l+n=3$  one gets the special geometry identities \cite{CandelasdellaOssa} 

4227: \beq \label{special_geom}

4228:   D_i X^I \equiv \chi^I_i \ ,\qquad 

4229:   D_i \chi_j^I  = i C_{ijk} G^{k\bar k} \bar \chi^I_{\bar k} e^K\ , \qquad 

4230:   D_i \bar \chi_{\bar j}^I = G_{i\bar j} \bar X^I\ .

4231: \eeq

4232: {}From (\ref{cov_D}) and (\ref{kaehlersimplifications}) follows 

4233: $[D_i,D_{\bar \jmath}]\chi_k=-G_{i\bar \jmath}+R_{i\bar \jmath k}^{\ \ l} \chi_l$ 

4234: and using (\ref{special_geom}) one gets 

4235: \begin{equation}

4236: [D_i,D_{\bar k}]^{\ \, l}_j=R^{\ \ \ l}_{i\bar k j}=G_{i\bar k}\delta^l_j+G_{j\bar k}\delta^l_i-C_{ijm} {\bar C}_{\bar k}^{ml}\ ,

4237: \label{eq:special}

4238: \end{equation}

4239: where we abbreviated  

4240: \begin{equation} \label{def-Cud}

4241: {\bar C}_{\bar k}^{(0) ml} = 

4242: e^{2K}\bar C^{(0)}_{\bar k\bar \imath\bar \jmath} 

4243: G^{m\bar\imath} G^{l \bar \jmath}\ ,\qquad \qquad {\bar C}_{\bar k}^{ml}=i{\bar C}_{\bar k}^{(0) ml}\ .

4244: \end{equation}

4245: 

4246: Let us also summarize some relations obeyed by $\tau_{IJ}$ and $C_{IJK}$. 

4247: One first notes that by homogeneity and \eqref{def-CC} and \eqref{special_geom} one has

4248: \beq

4249: \label{vielbein}

4250:  C_{IJK}X^K=0\ ,\qquad \qquad  C_{ijk} = C_{IJK}  \chi^I_i  \chi^J_j  \chi^K_k\ .

4251: \eeq

4252: Using the above definitions and the degree two 

4253: homogeneity of $\cF$ one also shows that

4254: \beq

4255: \label{transversality}

4256: 2 e^{K} X^I\I \tau_{IJ} \bar X^J=1\ ,\qquad \bar X^I \I \tau_{IJ} \chi_i^J=0\ ,

4257: \qquad   2 e^K \chi^I_i \I \tau_{IJ} \bar \chi^J_{\bar j}=G_{i \bar j}\ .  

4258: \eeq

4259: Denoting by $\I \tau^{IJ}$ the inverse of $\I \tau_{IJ}$ it follows from these conditions that

4260: \beq \label{G-tau}

4261: \chi^I_i G^{i \bar j} \bar \chi^J_{\bar j} e^K  = \tfrac{1}{2}  \I \tau^{IJ} + X^I \bar X^J e^K\ .

4262: \eeq

4263: 

4264: 

4265: 

4266: 

4267: 

4268: \section{Theta functions and modular forms}\label{theta}

4269: 

4270: 

4271: 

4272: Our conventions for the Jacobi theta functions are:

4273: %

4274: \be

4275: \begin{aligned}

4276: \vartheta_1(\nu|\tau)&=\vartheta [^1_1](\nu|\tau)=

4277: i \sum_{n \in {\bf Z}} (-1)^n q^{{1\over 2}(n+1/2)^2} e^{i \pi (2n +1) \nu},\\

4278: \vartheta_2(\nu|\tau)&= \vartheta [^1_0](\nu|\tau)=

4279: \sum_{n \in {\bf Z}} q^{{1\over 2}(n+1/2)^2} e^{i \pi (2n +1) \nu},\\

4280: \vartheta_3(\nu|\tau)&= \vartheta [^0_0](\nu|\tau)=

4281: \sum_{n \in {\bf Z}} q^{{1\over 2} n^2} e^{i \pi 2n  \nu},\\

4282: \vartheta_4(\nu|\tau)&= \vartheta [^0_1](\nu|\tau) =

4283: \sum_{n \in {\bf Z}} (-1)^n q^{{1\over 2}n^2} e^{i \pi 2n  \nu},

4284: \end{aligned}

4285: \ee

4286: %

4287: where $q=e^{2\pi i \tau}$. When $\nu=0$ we will simply denote $\vartheta_2(\tau)=

4288: \vartheta_2(0|\tau)$ (notice that $\vartheta_1(0|\tau)=0$).

4289: The theta functions $\vartheta_2(\tau)$, $\vartheta_3(\tau)$ and $\vartheta_4(\tau)$ have the

4290: following product representation:

4291: %

4292: \be

4293: \begin{aligned}

4294: \vartheta_2(\tau)&  =2 q^{1/8}\prod_{n=1}^{\infty} (1-q^n)(1+q^n)^2,\\

4295: \vartheta_3(\tau)&  = \prod_{n=1}^{\infty} (1-q^n)(1+q^{n-\half} )^2,\\

4296: \vartheta_4(\tau)& = \prod_{n=1}^{\infty} (1-q^n)(1-q^{n-\half} )^2

4297: \end{aligned}

4298: \ee

4299: %

4300: and under modular transformations they behave as:

4301: %

4302: \be

4303: \begin{aligned}

4304: \vartheta_2 (-1/\tau)= &{\sqrt { \tau \over i}} \vartheta_4 (\tau),\\

4305: \vartheta_3 (-1/\tau)= &{\sqrt { \tau \over i}} \vartheta_3 (\tau),\\

4306: \vartheta_4 (-1/\tau)= &{\sqrt { \tau \over i}} \vartheta_2 (\tau),

4307: \end{aligned}

4308: \quad\quad

4309: \begin{aligned}

4310: \vartheta_2 (\tau +1)= &{\rm e}^{i\pi/4} \vartheta_2 (\tau),\\

4311: \vartheta_3 (\tau +1)= &\vartheta_4 (\tau),\\

4312: \vartheta_4 (\tau +1)= & \vartheta_3 (\tau).

4313: \end{aligned}

4314: \label{thetatransformation} 

4315: \ee

4316: %

4317: The theta function $\vartheta_1(\nu|\tau)$ has the product representation

4318: %

4319: \be

4320: \label{prodone}

4321: \vartheta_1(\nu|\tau)=-2 q^{1\over 8} \sin (\pi \nu) \prod_{n=1}^{\infty} (1-q^n) (1-2 \cos (2 \pi \nu) q^n + q^{2n}).

4322: \ee

4323: %

4324: We also have the following useful identities:

4325: %

4326: \be

4327: \label{sumthet}

4328: \vartheta_3^4 (\tau) = \vartheta_2^4 (\tau) + \vartheta_4^4 (\tau),

4329: \ee

4330: %

4331: and

4332: %

4333: \be

4334: \label{prodtheta}

4335: \vartheta_2 (\tau)\vartheta_3 (\tau)\vartheta_4 (\tau) = 2\, \eta^{3}(\tau),

4336: \ee

4337: %

4338: where

4339: %

4340: \be

4341: \label{dede}

4342: \eta(\tau)= q^{1/24} \prod_{n=1}^{\infty} (1- q^n)

4343: \ee

4344: %

4345: is the Dedekind eta function. One has the following doubling formulae,

4346: %

4347: \be

4348: \label{deta}

4349: \begin{aligned}

4350: \eta(2\tau)=&{\sqrt {\eta(\tau)\vartheta_2(\tau) \over 2}}, \qquad

4351: \vartheta_2(2 \tau) ={\sqrt { {\vartheta_3^2(\tau) -\vartheta_4^2(\tau) \over 2}}}, \\

4352: \vartheta_3(2 \tau) =&{\sqrt { {\vartheta_3^2(\tau) +\vartheta_4^2(\tau) \over 2}}},\qquad

4353: \vartheta_4(2 \tau) ={\sqrt { \vartheta_3(\tau) \vartheta_4(\tau)}},\\

4354: \eta(\tau/2)=&{\sqrt {\eta(\tau) \vartheta_4(\tau)}}.

4355: \end{aligned}

4356: \ee

4357: %

4358: The Eisenstein series are defined by

4359: %

4360: \be

4361: \label{geneis}

4362: E_{2n}(q)=1-{4n \over B_{2n}}\sum_{k=1}^{\infty}{k^{2n-1} q^{k}\over 1-q^{k}},

4363: \ee

4364: %

4365: where $B_{m}$ are the Bernoulli numbers. The covariant version of $E_2$ is

4366: %

4367: \be

4368: \widehat E_2(\tau, \bar \tau)=E_2(\tau) -{3\over \pi {\rm Im}\, \tau}=E_2(\tau)-{6\ri \over \pi (\tau-\bar \tau)}.

4369: \ee

4370: %

4371: 

4372: The formulae for the derivatives of the theta functions are also useful:

4373: %

4374: \be

4375: \begin{aligned}

4376: q {d \over dq}\log\, \vartheta_4=&{1\over 24}\biggl( E_2 -\vartheta_2^4 -\vartheta_3^4\biggr),\\

4377: q {d \over dq}\log\, \vartheta_3=&{1\over 24}\biggl( E_2 +\vartheta_2^4 -\vartheta_3^4\biggr),\\

4378: q {d \over dq}\log\, \vartheta_2=&{1\over 24}\biggl( E_2 +\vartheta_3^4 +\vartheta_4^4\biggr),

4379: \end{aligned}

4380: \ee

4381: %

4382: and from these one finds 

4383: %

4384: \be

4385: q {d \over dq}\log\, \eta ={1\over 24}E_2(\tau)

4386: \label{Eta-der}

4387: \ee

4388: %

4389: and the Ramanujan identities 

4390: %

4391: \be\label{E-der}

4392: \ba

4393: q {d \over dq}\ E_2(q) =& {1\over 12}(E_2^2(q)-E_4(q)),\\

4394: q {d \over dq}\ E_4(q) =& {1\over 3}(E_2(q)E_4(q)-E_6(q)),\\

4395: q {d \over dq}\ E_6(q) =& {1\over 2}(E_2(q)E_6(q)-E_4^2(q)).

4396: \ea

4397: \ee

4398: %

4399: These can be used to compute the $q$-derivatives of the generators $K_2$, $K_4$ introduced in (\ref{gamma2generators}):

4400: %

4401: \be

4402: \ba

4403: q \partial_{q} K_2 &= {1\over 6} E_2(q) K_2(q) +{1\over 4} K_4(q) -{1\over 12} K_2^2(q),\\

4404: q \partial_q K_4 &= {1\over 3} K_4(q) (E_2(q) + K_2(q)).

4405: \ea

4406: \ee

4407: %

4408: The doubling formulae for $E_2(\tau), E_4(\tau)$ are

4409: %

4410: \be

4411: \label{deis}

4412: \begin{aligned}

4413: E_2(2 \tau)= &{1\over 2} E_2(\tau) + {1\over 4}(\vartheta_3^4(\tau) +\vartheta_4^4(\tau)),\\

4414: E_4(2 \tau)= & {1\over 16} E_4(2\tau) + {15\over 16}\vartheta_3^4 (\tau)\vartheta_4^4 (\tau).

4415: \end{aligned}

4416: \ee

4417: %

4418: 

4419: %\subsection{Borcherds automorphic forms}

4420: 

4421: 

4422: 

4423: \section{The antiholomorphic dependence of the heterotic $F^{(g)}$}\label{heteroticFg}

4424: In this Appendix we find the antiholomorphic dependence of $F^{(g)}(t,\bar t)$ in the heterotic theory. In section \ref{stu}, we show how the complicated 

4425: result of the heterotic computation of the $F^{(g)}$ in the STU-model given in \cite{mm} can be simplified, along the lines of \cite{borcherds}. In section \ref{enr} we 

4426: write down the result for $F^{(g)}_E$ in the Enriques Calabi-Yau and derive \eqref{Fgantihol}.

4427: 

4428: \subsection{A simple form for $F^{(g)}$ in the STU-model}\label{stu}

4429: 

4430: In \cite{mm}, an explicit expression for the holomorphic and antiholomorphic dependence of the topological amplitudes in the fiber limit of the STU-model was found. 

4431: This expression is obtained from a one--loop computation in the dual heterotic theory, given by the integral \eqref{hetint}, which is then 

4432: performed by using the technique of lattice reduction \cite{borcherds}. One finds that $F^{(g)}=F^{(g)}_{\rm deg}+F^{(g)}_{\rm ndeg}$, where \cite{mm}

4433: %

4434: \be\label{fgdeg}

4435: F^{(g)}_{\rm deg}= 4 \pi^2 U_1 \delta_{g,1} +{2^{2g-1} \pi^{4g-3} \over

4436: T_1^{2g-3}} \sum_{l=0}^{g} c_g (0,l) {l!\over \pi^{l+3} }

4437: \biggl( {T_1 \over

4438: U_1} \biggr)^l \zeta (2(2+l-g)),

4439: \ee

4440: \be\label{fgnondeg}

4441: \ba

4442: &F^{(g>1)}_{\rm ndeg} = 4\pi^{2g-2}(-1)^{g-1}\sum_{r\not=

4443: 0 }

4444:  \sum_{l=0}^{g}

4445:  \sum_{h=0}^{2g-2} \sum_{j=0}^{[g-1-h/2]}

4446: \sum_{a=0}^s c_g ( r^2/2, l)

4447: {(2\pi)^l(2g-2)! \over  j! h! (2g-h-2j-2)!}\\ &  \times  {(-1)^{j+h} \over  2^{j+a} }{(s+a)! \over  a! (s-a)!}({\rm

4448: sgn} \left({\rm Re}(r\cdot y)\right)^h

4449: {1\over (T_1U_1)^l} \left({\rm Re}(r \hat{\cdot} y)\right)^{l-j-a}

4450: \Li _{3+a+j+l-2g}(\re^{-r \hat{\cdot} y} )

4451: \\

4452: & +{2\pi^{3g-3}c_g(0,g-1)\over (T_1 U_1)^{g-1}}

4453: \sum_{s=0}^{g-1}  (-1)^s { (2g-2)!  \over s! (g-1-s)!} \psi ({1\over 2}+s)  \\

4454: & + \sum_{\begin{subarray}{c}l=0\\l\neq g-1\end{subarray}}^{g} 4^{l+g}\pi^{2g+l-5/2}c_g(0,l){\zeta (3+2(l-g)) \over

4455: (T_1U_1)^l} \\

4456: & \times \sum_{s=0}^{g-1}(-1)^s 2^{2(s-2g)+5} { (2g-2)! \over (2s)!

4457: (g-1-s)!}   \Gamma \Bigl({3\over 2}+  s+l -g\Bigr).

4458: \ea

4459: \ee 

4460: %

4461: We refer to $F^{(g)}_{\rm deg}$, $F^{(g)}_{\rm ndeg}$ as the degenerate and nondegenerate contributions, respectively. 

4462: Also, $s:=|2g-2-h-j-l-1/2|-{1/2};\quad y=(T,U),\quad$ the complex norm is defined as $r^2=2r_1r_2$, and

4463: %

4464: \be

4465: r\hat{\cdot}y\equiv |{\rm Re}(r\cdot y)|+{\rm i}{\rm Im}(r\cdot y)\nonumber.

4466: \ee

4467: % 

4468: The coefficients $c_g (m,l) $ can be obtained from the expansion

4469: %

4470: \be\label{expex}

4471: {E_4 E_6 \over \eta^{24}} \widehat{\CP}_{g}= \sum_{m\in \mathbb{Q}}

4472: \sum_{l\ge 0} c_g (m,l) q^m \tau_2^{-l},

4473: \ee

4474: %

4475: where $\widehat{\CP}_{g}$ are defined by

4476: %

4477: \be

4478: \biggl( { 2\pi  \eta^3 \lambda \over \vartheta_1(\lambda|\tau)}\biggr)^2\re^{-{\pi\lambda^2\over \tau_2}}=

4479: \sum_{g=0}^{\infty} (2 \pi \lambda)^{2g} {\widehat{\cal P}}_{g}(\tau, \bar \tau).

4480: \ee

4481: %

4482: Note that these $\widehat{\CP}_{g}(\tau,\bar{\tau})$ are the modular, almost holomorphic extensions of the $\CP_{g}(\tau)$ defined in \eqref{defpg}, that is,  $\widehat{\CP}_{g}$ is obtained from $\CP_{g}$ by replacing $E_2\rightarrow \widehat{E}_2$.

4483: The only antiholomorphic dependence in $\widehat{\CP}_{g}$ thus lies in the $\widehat{E}_2(\tau,\bar{\tau})$. Using the explicit expressions for $\widehat{\CP}_{g}$ given in \cite{km}, one can show that independently of the specific model,

4484: %

4485: \be\label{rec}

4486: c_g(m,l)={(-1)^l\over l!(4\pi)^l}c_{g-l}(m),

4487: \ee

4488: where $c_g(m)$ are defined analogously to \eqref{geomrmod}, that is

4489: %

4490: \be

4491: \sum_n c_g(n) q^n ={\cal P}_{g}(q){E_4E_6\over \eta^{24}}.

4492: \ee

4493: %

4494: In what follows, we will systematically express everything in terms of the coefficients $c_g(m)$.

4495: 

4496: It turns out that \eqref{fgnondeg} can be dramatically simplified. We will need the identity:

4497: %

4498: \be

4499: \label{blemma}

4500: \sum_j(-1)^j\binom{C}{j}\binom{A-2j+C-B-1}{A-2j}=\sum_j(-1)^j\binom{C}{A-j}\binom{B}{j}

4501: \ee

4502: %

4503: This is valid for any $A,B,C \in\mathbb{Z}$, see \cite{borcherds} for a proof. A special case of the above formula is the following. 

4504: Let $C$, $l$ and $m^+-h^+$ be integers such that $0\leq l<C<m^+-h^+-l$. Then,

4505: %

4506: \be

4507: \label{cor2}

4508: \sum_{\begin{subarray}{c}j\end{subarray}}{(-1)^j(m^+-h^++C-2j-1-l)!\over j!(m^+-h^+-2j)!(C-j)!}=0.\nonumber

4509: \ee

4510: %

4511: The proof of this statement is easy. Set $B=l, A=m^+-h^+$ in (\ref{blemma}) to obtain

4512: %

4513: \be

4514: \sum_{\begin{subarray}{c}j\end{subarray}}(-1)^j{(m^+-h^++C-2j-1-l)!C!\over j!(m^+-h^+-2j)!(C-j)!(C-1-l)!}=\sum_j\binom{C}{m^+-h^+-j}\binom{l}{j}.

4515: \ee

4516: %

4517: Since $C > l\geq 0$, any non-vanishing term on the right-hand side must fulfill $m^+-h^+-C\leq j\leq l$, in contradiction with the assumption $C<m^+-h^+-l$. 

4518: 

4519: We also have the following three additional nontrivial identities. First of all, let $s:=|2g-2-h-j-l-1/2|-{1/2}$. Then, 

4520: \ben

4521: &&\sum_{h=0}^{2g-2}\sum_{j=0}^{C}{(2g-2)!\over 2^{2g-2}}{(s+C-j)!(-1)^{C-j}\over l!h!j!(2g-2-h-2j)!(s-C+j)!(C-j)!}\\

4522: &&\hspace{4cm}=\left\{\begin{array}{ll}\binom{2g-3-l}{C}{1\over (l-C)!} & C\leq{\rm min}(l,2g-3-l)\\

4523: 0&{\rm otherwise.}\end{array}\right.

4524: \label{cor3}

4525: \een

4526: %

4527: This is valid for any pair of positive integers $g,l$. The second identity reads, 

4528: %

4529: \be

4530: \label{cor4}

4531: \sum_{s=0}^{g-1}(-1)^s{(2g-2)!\over s!(g-1-s)!}\psi\Bigl(s+{1\over 2}\Bigr)=-2^{(2g-2)}(g-2)!.

4532: \ee

4533: %

4534: The final identity we will need is

4535: %

4536: \be

4537: \label{cor5}

4538: \sum_{s=0}^{g-1}(-1)^{l+s}2^{2(s-2g)+5}{(2g-2)!\over (2s)!(g-1-s)!}\Gamma\Bigl({3\over 2}+s+l-g \Bigr)={(-1)^{g-1}(2g-3-l)!\over (2g-3-2l)!}\sqrt{\pi}4^{-l}.

4539: \ee

4540: %

4541: %

4542: which is valid for any $l\in \mathbb{N}, l<g-1$. Making use of (\ref{cor2}) and (\ref{cor3}), we can convert the sums over $h,j,a$ in \eqref{fgnondeg} into a single one over $C=j+a=\{0,\cdots,l\}$. Then, 

4543: (\ref{cor4}) and (\ref{cor5}) can be used to simplify the second respectively third term in \eqref{fgnondeg}. The sum over $r$ can be restricted for all $g\geq 3$ to a sum over $r$ for which ${\rm Re}(r\cdot y)<0$, or equivalently to a sum over positive $r$ and a finite number of boundary cases. At genus 2, however, there is a contribution from ${\rm Re}(r\cdot y)>0$, it reads \cite{mm}

4544: %

4545: \be\label{negr}

4546: {c_0(r^2/2)\over 16 T_1U_1}\Li_3(\re^{-r\cdot y}).

4547: \ee

4548: % 

4549: We can then write down a simplified expression for the nondegenerate part of $F^{(g)}$ in the STU model:

4550: %

4551: \be\label{Fgndfin1}

4552: \ba

4553: &F^{(g>1)}_{\rm nd,STU}\\

4554: &=

4555: \sum_{l=0}^{g-1}\sum_{C=0}^{{\rm min}(l,2g-3-l)}\sum_{r>0}{\binom{2g-l-3}{C}\over(l-C)! 2^C}{(-{\rm Re}(r\cdot y))^{l-C}\over (2T_1U_1)^l}c_{g-l}({r^2\over 2})\Li_{3-2g+l+C}(\re^{-r\cdot y})\\

4556: &+{22\over 2^g(g-1)}{1\over (2T_1U_1)^{g-1}}+\sum_{l=0}^{g-2}{c_{g-l}(0)\over l!(4T_1U_1)^l}\zeta(3+2(l-g)){(2g-3-l)!\over (2g-3-2l)!},

4557: \ea

4558: \ee

4559: %

4560: where we also have used the fact that in the STU model, \mbox{$c_1(0)=-22$}, and we have removed an overall prefactor of $4(2\pi \ri)^{2g-2}$ to agree with the normalization of the 

4561: topological string amplitudes. 

4562:  

4563: %%%%%%%%%%%

4564: \subsection{Application to the Enriques Calabi-Yau}\label{enr}

4565: The above expressions have to be adapted slightly for the Enriques Calabi-Yau. We only consider here the geometric reduction suited to the large radius limit. As shown in \cite{km}, 

4566: the polylogarithm is replaced by $\Li_m(x)\rightarrow 2^m\Li_m(x^{1\over 2})-\Li_m(x)$, and the norm of the reduced lattice is doubled. We also replace the quantity $2T_1U_1$ appearing in the STU-model by $Y=\re^{-K}$ as in \eqref{def-KY}, and the coefficients $c_g(m)$ are now defined by \eqref{geomrmod}. There is a new important simplification: $c_{0}(r^2)$ and $c_{g>1}(0)$ vanish, 

4567: and thus there is no contribution from negative $r$ at any genus $g>1$, since \eqref{negr} becomes

4568: %

4569: \be

4570: {c_0(r^2)\over 8 Y}\left(8\Li_3(\re^{-r\cdot y})-\Li_3(\re^{-2r\cdot y})\right)=0.

4571: \ee

4572: %. 

4573: Furthermore, the degenerate contribution \eqref{fgdeg} and the last term in \eqref{fgnondeg} vanish for all $g>1$,  while $c_1(0)=4$, and the full $F^{(g)}_E(t,\bar{t})$ for the Enriques reads

4574: %

4575: \be

4576: \ba

4577: &\hspace{-1cm}F^{(g>1)}_E(t,\bar{t})=\sum_{l=0}^{g-1}\sum_{C=0}^{{\rm min}(l,2g-3-l)}\sum_{r>0}{\binom{2g-l-3}{C}\over(l-C)! 2^C}{(-2{\rm Re}(r\cdot t))^{l-C}\over Y^l}c_{g-l}(r^2)\\

4578: &\cdot\left(2^{3-2g+l+C}\Li_{3-2g+l+C}(\re^{-r\cdot t})-\Li_{3-2g+l+C}(\re^{-2r\cdot t})\right)-{1\over 2^{g-2}(g-1)}{1\over Y^{g-1}}.

4579: \ea

4580: \ee

4581: %

4582: Using

4583: %

4584: \be

4585: {\rm Re}(t^a)\partial_{t^a}\Li_n(\re^{-r\cdot t})=-{\rm Re}(r\cdot t)\Li_{n-1}(\re^{-r\cdot t}),

4586: \ee

4587: %

4588: this can be cast into the following recursive form: 

4589: %

4590: \be

4591: \ba

4592: &F^{(g)}_E(t,\bar{t})\\

4593: =&\sum_{l=0}^{g-1}\sum_{C=0}^{\begin{subarray}{c}{\rm min}\\(l,2g-3-l)\end{subarray}}{(2g-3-l)!\over (2g-3-l-C)!(l-C)!C!2^l}{(t^{a_1}+\bar{t}^{a_1})\cdots(t^{a_{l-C}}+\bar{t}^{a_{l-C}})\partial_{a_1}\cdots\partial_{a_{l-C}}\CF^{(g-l)}(t)\over Y^l}\\

4594: &\hspace{2cm}-{1\over 2^{g-2}(g-1)Y^{g-1}}.\\

4595: \ea

4596: \ee

4597: %

4598: Notice that this exhibits the structure of the antiholomorphic amplitudes written down in \cite{abk}. 

4599: 

4600: \section{Anomaly equations for $F^{(g)}$ on the big moduli space \label{Cal_big_Fg}}

4601: 

4602: \subsection{Anomaly equation for $F^{(g)}\ (g>1)$ \label{Cal_big_Fg1}}

4603: Here we provide some details on the calculation of the recursive anomaly equations on the 

4604: big moduli space. 

4605: We like to rewrite the equation \eqref{rec_Fg} in terms of the variables $Y^K = \lambda^{-1} X^K(t)$ and

4606: $\bar Y^K$. First note that 

4607: \beq

4608:    \frac{\partial}{\partial t^i} - K_i \lambda \frac{\partial}{\partial \lambda} = \lambda^{-1} \chi^I_i \frac{\partial}{\partial Y^K}\ ,

4609: \eeq

4610: where $\chi_i^I$ is defined in \eqref{exp_Omega}.

4611: This implies that the first derivative of $F^{(g)}$ can be written as

4612: \beq \label{D_iF}

4613:   D_i F^{(g)} = \lambda^{-2g+1} \chi^I_i \, \partial_{ Y^I} F^{(g)}(Y)\ .

4614: \eeq

4615: where we have used the fact that $\lambda \partial_\lambda F^{(g)}(Y) = (2g-2) F^{(g)}(Y)$ 

4616: due to \eqref{Fg_Y}. Moreover, one derives that the second derivative reads

4617: \bea

4618:    D_i D_j F^{(g-1)} &=&   \lambda^{-2g+3} (D_i \chi^I_j) \, \partial_{ Y^I} F^{(g-1)}(Y) +  \lambda^{-2g+3}  \chi^I_j D_i \, \partial_{Y^I} F^{(g-1)}(Y) \nn \\

4619:   &=& i \lambda^{-2g+3} C_{ijk} G^{k\bar k} \bar \chi^I_{\bar k}\, \partial_{Y^I} F^{(g-1)}(Y) + \lambda^{-2g+2}  \chi^I_i \chi^J_j\,  \partial_{Y^I} \partial_{Y^J} F^{(g-1)}(Y) \nn \\

4620:   &=& \lambda^{-2g+2}  \chi^I_i \chi^J_j  \Big[ \tfrac{i}2 C_{IJ}^{(Y)\, K}  \partial_{Y^K} F^{(g-1)}(Y)  +   \partial_{Y^I} \partial_{Y^J}  F^{(g-1)}(Y) \Big]\ .

4621: \eea

4622: In order to evaluate the second identity we have used the special geometry relation \eqref{special_geom} and \eqref{D_iF} 

4623: while for the third identity we have used

4624: \eqref{G-tau}.

4625: Also notice that from \eqref{D_iF} one infers that

4626: \beq

4627:   \sum_{r=1}^{g-1} D_i F^{(r)} D_j F^{(g-r)} =  \lambda^{-2g+2} \chi^I_i \chi^J_j  \sum_{r=1}^{g-1} \partial_{Y^I} F^{(r)}(Y)\, \partial_{Y^J} F^{(g-r)}(Y)

4628: \eeq

4629: Finally, we need the identity 

4630: \beq

4631:   \tfrac{i}2 e^{2K} \bar C_{\bar i \bar j \bar k} G^{\bar j j} G^{\bar k k} \chi^I_j \chi^J_k =  \tfrac{i}8 \lambda^{-1}\bar \chi^K_{\bar i} \bar C_{\ K}^{(Y)\, IJ}\ .

4632: \eeq

4633: Hence, we conclude that 

4634: \beq

4635:   \partial_{\bar Y^I} F^{(g)} =  \tfrac{i}8 \bar C_{K}^{\ IJ}  \Big[ \partial_{Y^I} \partial_{Y^J}  F^{(g-1)} + \sum_{r=1}^{g-1} \partial_{Y^I} F^{(r)}\, \partial_{Y^J} F^{(g-r)} \Big]

4636:                                                        -  \tfrac{1}{16} \bar C_{K}^{\ IJ}  C_{IJ}^{\ \ K}  \partial_{Y^K} F^{(g-1)}\ ,

4637: \eeq

4638: where $C_{IJK}$ and $F^{(r)}$ are functions of $Y^K,\bar Y^K$. This equation is precisely the recursive anomaly equation 

4639: given in \eqref{big_Fg}.

4640: 

4641: Let us also present the derivation of the simplest solution to \eqref{big_Fg}. In other words, we 

4642: calculate $F^{(2)}$ by using  the integration by parts method of \cite{bcov}.

4643: To do that we use the definition \eqref{d_prop} of the propagator to

4644: replace $\bar C_K^{\ IJ}$ in \eqref{F_2big}. We pull the derivative $\partial_{\bar I}$ in front 

4645: of all the terms and evaluate

4646: \bea

4647:   &&\partial_{\bar I} \Big[ F^{(2)} + \tfrac{1}{2} \Delta^{JK} \big( 

4648:    D_J \partial_{K} F^{(1)} + \partial_{J} F^{(1)} \partial_{K} F^{(1)} \big)\Big]  

4649:   =-\big(\tfrac{\chi}{24} -1 \big) \partial_{\bar I} \big[\Delta^{JK} K_{J}\big] \partial_K F^{(1)} \nn \\

4650:   &&-

4651:   \tfrac{1}{8} \partial_{\bar I} \big[\Delta^{JK} \Delta^{LM}\big] \big(C^{(0)}_{KLMJ} + 4 C^{(0)}_{JLM} \partial_{K} F^{(1)} \big)

4652:   -\tfrac12 \big(\tfrac{\chi}{24} -1 \big) \partial_{\bar I} \big[\Delta^{JK} K_{J} K_K\big]

4653:       \ ,

4654: \eea

4655: where $C^{(0)}_{IJK}=iC_{IJK}$ as defined in \eqref{def-C^g}.

4656: In performing the derivative we used the equation \eqref{F_2big} to eliminate the 

4657: terms arising when $\partial_{\bar I}$ hits the propagator. Furthermore 

4658: we commuted $\partial_{\bar I}$ with the covariant derivative $D_J$ by using the identity 

4659: \beq

4660:     \big[ \partial_{\bar I},D_J\big] V_K= \tfrac{1}{4} C_{JK}^{\ \ \, P} \bar C_{IP}^{\ \ \, M} V_M\ .

4661: \eeq

4662: One can then eliminate the second derivative $\partial_{\bar I} \partial_K F^{(1)}$ by inserting the 

4663: equation \eqref{big_F1} and applying the useful identities

4664: \beq

4665:    D_I \widehat K_{I \bar J}=D_I  \widehat K_J = 0 \ , \qquad \qquad \Delta^{IJ} D_I K_{J \bar K} = 2\Delta^{IJ} K_{I\bar K} K_J\ ,

4666:    \qquad K_J \partial_{\bar I}  \Delta^{JK} =0\ .  

4667: \eeq

4668: In the next step we once again pull the derivative $\partial_{\bar I}$ in front of all terms

4669: and evaluate 

4670: \bea

4671:   &&\partial_{\bar I} \Big[ F^{(2)} + \tfrac{1}{2} \Delta^{JK} \big( 

4672:    D_J \partial_{K} F^{(1)} + \partial_{J} F^{(1)} \partial_{K} F^{(1)} \big)

4673:    +\tfrac12 \big(\tfrac{\chi}{24} -1 \big) \Delta^{JK} K_{J} K_K \\

4674:   && +\tfrac{1}{8} \Delta^{JK}\Delta^{LM} \big(C^{(0)}_{KLMJ} + 4 C^{(0)}_{JLM} \partial_{K} F^{(1)} \big)

4675:      + \big(\tfrac{\chi}{24} -1 \big) \Delta^{JK} K_{J} \partial_K F^{(1)}\Big] \nn \\

4676:   &&=  -\partial_{\bar I} \Big[ \tfrac{1}2 \big(\tfrac{\chi}{24} -1 \big)  C^{(0)}_{JLM} \Delta^{JK} \Delta^{LM} K_J 

4677:          + \tfrac{1}2 \big(\tfrac{\chi}{24} -1 \big)^2 \Delta^{JK} K_J K_K \nn\\

4678:          &&\ \ + \Delta^{JK} \Delta^{LM} \Delta^{QP}\big(\tfrac1{12} C^{(0)}_{KMQ} C^{(0)}_{PLJ} +\tfrac{1}{8}  C^{(0)}_{KJQ} C^{(0)}_{PML} \big) \Big]\ . \nn

4679: \eea

4680: We are now in the position to read off $F^{(2)}(Y,\bar Y)$ up to a holomorphic ambiguity $f^{(2)}(Y)$.

4681: The corresponding solution can be found in \eqref{F_2_big}.

4682: 

4683: \subsection{Anomaly equation for $F^{(1)}$ on big phase space \label{Cal_big_F1}}

4684: 

4685: In this appendix we discuss the lift of the holomorphic anomaly equation \ref{anomaly_F1} for $F^{(1)}$ 

4686: to the big moduli space $\widehat \cM$. To begin with let us first note that 

4687: \beq

4688:    |\lambda|^{-2} \chi^I_i \bar \chi^J_{\bar j} \partial_{Y^I} \partial_{\bar Y^J} K= G_{i\bar \jmath}\ , \qquad  

4689:    Y^I \partial_{Y^I} \partial_{\bar Y^J} K= 0\ ,\qquad 

4690:    \bar Y^J \partial_{Y^I} \partial_{\bar Y^J} K= 0\ .

4691: \eeq

4692: where $K$ is the K\"ahler potential \eqref{KpotII} and $G_{i\bar \jmath}$ is the Weil-Petersson metric.

4693: We also evaluate the first derivative $K_I$ of $K$ and show that it satisfies

4694: \beq \label{K_Iproj}

4695:    K_I \chi^I_i = 0\ , \qquad \qquad K_I Y^I = -1 \ .

4696: \eeq

4697: With these identities at hand we now lift the holomorphic anomaly equation \eqref{anomaly_F1}.

4698: Using the homogeneity condition \eqref{F1_hom} one derives 

4699: \beq \label{F1rewrite1}

4700:    \partial_{i} \partial_{\bar j} F^{(1)} = |\lambda|^{-2} \chi^I_i \bar \chi^J_{\bar j} \partial_{Y^I} \partial_{\bar Y^J} F^{(1)}(Y)  

4701: \eeq

4702: Moreover, one shows that 

4703: \beq

4704:   \tfrac{1}{2} e^{2K} G^{k\bar k} G^{l \bar l} C_{i k l } \bar C_{\bar j \bar k \bar l} = |\lambda|^{-2} \chi^I_i \bar \chi^J_{\bar j} \tfrac18 C_{ILM} \bar C_{J}^{\ LM}\ ,

4705: \eeq

4706: as well as

4707: \beq \label{F1rewrite3}

4708:   \Big(\frac{\chi}{24} -1 \Big)G_{i\bar j} = |\lambda|^{-2}\chi^I_i \bar \chi^J_{\bar j} \Big(\frac{\chi}{12} -2 \Big) e^{K(Y,\bar Y)} \I \tau_{IJ}\ .

4709: \eeq

4710: Inserting \eqref{F1rewrite1}-\eqref{F1rewrite3} into the anomaly equation \eqref{anomaly_F1}

4711: we verify its big moduli space counterpart \eqref{big_F1}. It is straightforward to 

4712: integrate \eqref{big_F1} to find the local solution \eqref{F_1solution} for $F^{(1)}$ by 

4713: applying the identity

4714: \beq

4715:   R_{IJ} =  \partial_{Y^I}  \partial_{\bar Y^J} \log \det \I \tau = - \tfrac14 C_{I KL} \bar C_J^{\ KL}\ .

4716: \eeq

4717: It is however instructive to also recall a second alternative approach which integrates 

4718: \eqref{anomaly_F1} rather then  \eqref{big_F1}.

4719: 

4720: Let us end this appendix by recalling the direct integration of \eqref{anomaly_F1}.

4721: First note that the Riemann tensor on a special K\"ahler manifold is given by 

4722: \beq

4723:    R_{i\bar j l \bar m} = G_{i \bar j} G_{l \bar m} + G_{i \bar m} G_{l \bar j} - e^{2K} C_{ilp} \bar C_{\bar j \bar m \bar p} G^{p\bar p}\ .   

4724: \eeq

4725: The Ricci tensor takes the form

4726: \beq

4727:   R_{i\bar j} = \partial_i \partial_{\bar j} \log \det G = G_{i \bar j} (h^{2,1}+1) - e^{2K} C_{ilp} \bar C_{\bar j \bar m \bar p} G^{l\bar m} G^{p\bar p}\ .\ , 

4728: \eeq

4729: such that 

4730: \beq

4731:    \tfrac{1}{2} e^{2K} G^{k\bar k} G^{l \bar l} C_{i k l } \bar C_{\bar j \bar k \bar l} = G_{i \bar j} (h^{2,1}+1) -  

4732:    \partial_i \partial_{\bar j} \log \det G\ . 

4733: \eeq

4734: Using this equation we solve \eqref{anomaly_F1} as \footnote{The derivative of the determinate of the matrix $A$ is given by

4735:  $\partial_x \det A = \det A \cdot A^{-1\, IJ} \partial_x A_{IJ}$ }

4736: \beq

4737:   F^{(1)} = -\tfrac{1}{2} \log \det G + \big(\tfrac{1}{2} (h^{2,1} + 1) -\frac{\chi}{24} + 1 \big)K + h(t) + \bar h(\bar t)

4738: \eeq

4739: where $h(t)$ is a holomorphic function arising as integration constant.

4740: Now note that it follows from \eqref{G-tau} that \cite{N=2rev} 

4741: \beq

4742:   \det(2\I\tau_{IJ}) = - \det( G_{i\bar j}) e^{-(h^{(2,1)}+1)K} |\det(\chi^I_i, X^I)|^{-2}\ .

4743: \eeq

4744: This equation can be used to rewrite $F^{(1)}$ as

4745: \beq \label{F1_somlutionapp}

4746:   F^{(1)} = - \tfrac{1}{2}\log \det(2\I\tau_{IJ})+ \Big(1-\frac{\chi}{24}\Big)K  +\tfrac{1}{2} \log\Big( -|\det(\chi^I_i, X^I)|^{-2}\Big)  + h(t) + \bar h(\bar t)

4747: \eeq

4748: One can evaluate the determinate of the coordinate change and shows \cite{N=2rev}

4749: \beq

4750:  |\det(\chi^I_i, X^I)|^{-2} =|X^0|^{-2(h^{2,1} + 1)}|\det e^i_j|^{-2}\ .

4751: \eeq

4752: %\beq

4753: %   \log \det ( 2\I \tau )=\log (-\det G) - (h^{2,1} + 1) K  - (h^{2,1} + 1)\log (X^0 \bar X^0)-2 \log(|\det e^i_j|) \ .

4754: %\eeq

4755: where $e^i_j = \partial_{t^i} (X^i / X^0)$. But since $X^0$ and $e^i_j$ are holomorphic in the coordinates $t^i$

4756: they can be absorbed into $h$ such that \eqref{F1_somlutionapp} becomes \eqref{F_1solution}.

4757: 

4758: 

4759: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4760: \begin{thebibliography}{99}

4761: %%%%%%%%%%%%%%%%%%%%%%%%%%%%

4762: 

4763: 

4764: 

4765: 

4766: 

4767: \bibitem{abk}

4768: M. Aganagic, V. Bouchard and A. Klemm, ``Topological strings and (almost) modular forms," 

4769: hep-th/0607100.

4770:  %%CITATION = HEP-TH 0607100;%%

4771: 

4772: \bibitem{vertex}

4773:   M.~Aganagic, A.~Klemm, M.~Mari\~no and C.~Vafa,

4774:   ``The topological vertex,''

4775:   Commun.\ Math.\ Phys.\  {\bf 254}, 425 (2005)

4776:   [arXiv:hep-th/0305132].

4777:   %%CITATION = CMPHA,254,425;%%

4778: 

4779: \bibitem{agnt}

4780: I.~Antoniadis, E.~Gava, K.~S.~Narain and T.~R.~Taylor,

4781: ``N=2 type II heterotic duality and higher derivative F terms,''

4782:   Nucl.\ Phys.\ B {\bf 455}, 109 (1995)

4783:   [arXiv:hep-th/9507115].

4784:   %%CITATION = HEP-TH 9507115;%%

4785: 

4786: \bibitem{aspinwall}

4787: P.~S.~Aspinwall, ``An N=2 Dual Pair and a Phase Transition,''

4788:   Nucl.\ Phys.\ B {\bf 460}, 57 (1996)

4789:   [arXiv:hep-th/9510142].

4790:   %%CITATION = HEP-TH 9510142;%%

4791: 

4792: \bibitem{Aspinwall:1996mn}

4793:   P.~S.~Aspinwall,

4794:   ``K3 surfaces and string duality,''

4795:   arXiv:hep-th/9611137.

4796:   %%CITATION = HEP-TH/9611137;%%

4797:   

4798: \bibitem{bcov1}

4799:   M.~Bershadsky, S.~Cecotti, H.~Ooguri and C.~Vafa,

4800:   ``Holomorphic anomalies in topological field theories,''

4801:   Nucl.\ Phys.\ B {\bf 405}, 279 (1993)

4802:   [arXiv:hep-th/9302103].

4803:   %%CITATION = HEP-TH 9302103;%%

4804: 

4805: 

4806: \bibitem{bcov}

4807: M.~Bershadsky, S.~Cecotti, H.~Ooguri and C.~Vafa, ``Kodaira-Spencer theory of gravity and exact results for quantum string

4808: amplitudes,''

4809:   Commun.\ Math.\ Phys.\  {\bf 165}, 311 (1994)

4810:   [arXiv:hep-th/9309140].

4811:   %%CITATION = HEP-TH 9309140;%%

4812: 

4813: 

4814: 

4815: \bibitem{binfinite}

4816: R.E. Borcherds, ``Automorphic forms in $O_{s+2,2}(\IR)$ and infinite products," Inv. Math. {\bf 120} (1995) 161.

4817: 

4818: \bibitem{borcherdsone}

4819: R. E. Borcherds, ``The moduli space of Enriques surfaces and the fake monster Lie superalgebra,''

4820: Topology {\bf 35}, 699 (1996).

4821: 

4822: \bibitem{borcherds}

4823: R. E. Borcherds, ``Automorphic forms with singularities on Grassmannians,''

4824: Invent. Math. {\bf 132}, 491 (1998) [arXiv:alg-geom/9609022].

4825: 

4826: \bibitem{borcea}

4827: C.~Borcea,

4828: ``K3-surfaces with involutions and Mirror pairs of Calabi-Yau manifolds'', in Mirror symetry II, Ed. B.~Greene and S.S.~Yau (1997), AMS/International Press

4829: 

4830: \bibitem{bfmt}

4831:   U.~Bruzzo, F.~Fucito, J.~F.~Morales and A.~Tanzini,

4832:   ``Multi-instanton calculus and equivariant cohomology,''

4833:   JHEP {\bf 0305} (2003) 054

4834:   [arXiv:hep-th/0211108].

4835:   %%CITATION = HEP-TH 0211108;%%

4836:   

4837:   \bibitem{CandelasdellaOssa}

4838: P.~Candelas and X.~de la Ossa, ``Moduli space of Calabi-Yau manifolds,''

4839:   Nucl.\ Phys.\  B {\bf 355}, 455 (1991).

4840:   %%CITATION = NUPHA,B355,455;%%  

4841: 

4842: 

4843: 

4844: \bibitem{N=2rev}

4845:   A.~Ceresole, R.~D'Auria and S.~Ferrara,

4846:   ``The Symplectic Structure of N=2 Supergravity and its Central Extension,''

4847:   Nucl.\ Phys.\ Proc.\ Suppl.\  {\bf 46}, 67 (1996)

4848:   [arXiv:hep-th/9509160];\\

4849:   %%CITATION = HEP-TH 9509160;%%

4850:   L.~Andrianopoli, M.~Bertolini, A.~Ceresole, R.~D'Auria, S.~Ferrara, P.~Fre and T.~Magri,

4851:    ``N = 2 supergravity and N = 2 super Yang-Mills theory on general scalar

4852:   manifolds: Symplectic covariance, gaugings and the momentum map,''

4853:   J.\ Geom.\ Phys.\  {\bf 23}, 111 (1997)

4854:   [arXiv:hep-th/9605032].

4855:   %%CITATION = HEP-TH 9605032;%%

4856: 

4857: \bibitem{Craps:1997gp}

4858:   B.~Craps, F.~Roose, W.~Troost and A.~Van Proeyen,

4859:   ``What is special Kaehler geometry?,''

4860:   Nucl.\ Phys.\ B {\bf 503}, 565 (1997)

4861:   [arXiv:hep-th/9703082].

4862: 

4863: 

4864: 

4865: 

4866: \bibitem{DijkgraafZagier} R.~Dijkgraaf, ``Mirror Symmetry and Elliptic Curves,''

4867: and M.~Kaneko and D.~Zagier, ``A generalized Jacobi Theta function and Quasimodular Forms,''

4868: in {\em The Moduli Space of Curves}, Prog. Math. {\bf 129} (Birkh\"auser, 1995),

4869: 149 and 165.

4870: 

4871: \bibitem{d}

4872: R. Dijkgraaf, personal communication (1998). 

4873: 

4874: \bibitem{Dijkgraaf:2002ac}

4875:   R.~Dijkgraaf, E.~P.~Verlinde and M.~Vonk,

4876:   ``On the partition sum of the NS five-brane,''

4877:   arXiv:hep-th/0205281.

4878:   %%CITATION = HEP-TH/0205281;%%

4879: 

4880: \bibitem{Douglas:2006es}

4881:   M.~R.~Douglas and S.~Kachru,

4882:   ``Flux compactification,''

4883:   arXiv:hep-th/0610102.

4884:   %%CITATION = HEP-TH/0610102;%%

4885: 

4886: \bibitem{dkm1}

4887:   N.~Dorey, V.~V.~Khoze and M.~P.~Mattis,

4888:    ``Multi-instanton calculus in N = 2 supersymmetric gauge theory.  II:

4889:   Coupling to matter,''

4890:   Phys.\ Rev.\ D {\bf 54} (1996) 7832

4891:   [arXiv:hep-th/9607202].

4892:   %%CITATION = HEP-TH 9607202;%%

4893: 

4894: \bibitem{dkm2}

4895:   N.~Dorey, V.~V.~Khoze and M.~P.~Mattis,

4896:   ``On N = 2 supersymmetric {QCD} with 4 flavors,''

4897:   Nucl.\ Phys.\ B {\bf 492} (1997) 607

4898:   [arXiv:hep-th/9611016].

4899:   %%CITATION = HEP-TH 9611016;%%

4900: 

4901: \bibitem{emo}

4902: B.~Eynard, M.~Mari\~no and N.~Orantin, ``Holomorphic anomaly and matrix models,''

4903:   arXiv:hep-th/0702110.

4904:   %%CITATION = HEP-TH/0702110;%%

4905: 

4906: \bibitem{fhsv}

4907: S.~Ferrara, J.~A.~Harvey, A.~Strominger and C.~Vafa, ``Second quantized mirror symmetry,''

4908:   Phys.\ Lett.\ B {\bf 361}, 59 (1995) [arXiv:hep-th/9505162].

4909:   %%CITATION = HEP-TH 9505162;%%

4910: 

4911: \bibitem{Freed}

4912:   D.~S.~Freed,

4913:   ``Special Kaehler manifolds,''

4914:   Commun.\ Math.\ Phys.\  {\bf 203}, 31 (1999)

4915:   [arXiv:hep-th/9712042].

4916:   %%CITATION = HEP-TH 9712042;%%

4917: 

4918: \bibitem{gebert}

4919: R.~W.~Gebert, ``Introduction To Vertex Algebras, Borcherds Algebras, And The Monster Lie Algebra,''

4920:   Int.\ J.\ Mod.\ Phys.\  A {\bf 8}, 5441 (1993)

4921:   [arXiv:hep-th/9308151].

4922:   %%CITATION = IMPAE,A8,5441;%%

4923: 

4924: \bibitem{gv} R.~Gopakumar and C.~Vafa, ``M-theory and topological strings. I\& II,"

4925: [arXiv:hep-th/9809187] and [arXiv:hep-th/9812127].

4926:  %%CITATION = HEP-TH 9809187;%%

4927: %%CITATION = HEP-TH 9812127;%%

4928: 

4929: 

4930: \bibitem{Gunaydin:2006bz}

4931:   M.~Gunaydin, A.~Neitzke and B.~Pioline,

4932:   ``Topological wave functions and heat equations,''

4933:   JHEP {\bf 0612}, 070 (2006)

4934:   [arXiv:hep-th/0607200].

4935:   %%CITATION = JHEPA,0612,070;%%

4936: 

4937: 

4938: \bibitem{hm}

4939: J.~A.~Harvey and G.~W.~Moore,

4940:   ``Algebras, BPS States, and Strings,''

4941:   Nucl.\ Phys.\ B {\bf 463}, 315 (1996)

4942:   [arXiv:hep-th/9510182].

4943:   %%CITATION = HEP-TH 9510182;%%

4944: 

4945: 

4946: \bibitem{hmfhsv}

4947: J.~A.~Harvey and G.~W.~Moore,

4948:   ``Exact gravitational threshold correction in the FHSV model,''

4949:   Phys.\ Rev.\ D {\bf 57}, 2329 (1998)

4950:   [arXiv:hep-th/9611176].

4951:   %%CITATION = HEP-TH 9611176;%%

4952: 

4953: \bibitem{mirror}

4954: K. Hori {\it et al.}, {\it Mirror symmetry}, AMS, Providence 2003. 

4955: 

4956: \bibitem{hosono}

4957: S.~Hosono, M.~H.~Saito and A.~Takahashi,

4958:   ``Holomorphic anomaly equation and BPS state counting of rational  elliptic surface,''

4959:   Adv.\ Theor.\ Math.\ Phys.\  {\bf 3}, 177 (1999)

4960:   [arXiv:hep-th/9901151].

4961:   %%CITATION = HEP-TH 9901151;%%

4962: 

4963: \bibitem{hosonorev}

4964: S.~Hosono, ``Counting BPS states via holomorphic anomaly equations,''

4965:   arXiv:hep-th/0206206.

4966:   %%CITATION = HEP-TH/0206206;%%

4967:   

4968: \bibitem{hk}

4969: M.~x.~Huang and A.~Klemm,

4970:   ``Holomorphic anomaly in gauge theories and matrix models,''

4971:   arXiv:hep-th/0605195.

4972:   %%CITATION = HEP-TH 0605195;%%

4973: 

4974: \bibitem{hkq}

4975:   M.~x.~Huang, A.~Klemm and S.~Quackenbush, ``Topological string theory on compact Calabi-Yau: Modularity and boundary conditions,''

4976:   arXiv:hep-th/0612125.

4977:   %%CITATION = HEP-TH 0612125;%%

4978: 

4979: \bibitem{kklmv}

4980: S.~Kachru, A.~Klemm, W.~Lerche, P.~Mayr and C.~Vafa, ``Nonperturbative results on the point particle limit of N=2 heterotic string compactifications,''

4981:   Nucl.\ Phys.\  B {\bf 459}, 537 (1996)

4982:   [arXiv:hep-th/9508155].

4983:   %%CITATION = NUPHA,B459,537;%%

4984:   

4985:   \bibitem{kkv}

4986: S.~H.~Katz, A.~Klemm and C.~Vafa, ``Geometric engineering of quantum field theories,''

4987:   Nucl.\ Phys.\  B {\bf 497}, 173 (1997)

4988:   [arXiv:hep-th/9609239].

4989:   %%CITATION = NUPHA,B497,173;%%  

4990: 

4991: \bibitem{kkvbh}

4992: S.~H.~Katz, A.~Klemm and C.~Vafa, ``M-theory, topological strings and spinning black holes,''

4993:   Adv.\ Theor.\ Math.\ Phys.\  {\bf 3}, 1445 (1999)

4994:   [arXiv:hep-th/9910181].

4995:   %%CITATION = 00203,3,1445;%%

4996: 

4997: \bibitem{kawai}

4998: T.~Kawai and K.~Yoshioka, ``String partition functions and infinite products,''

4999:   Adv.\ Theor.\ Math.\ Phys.\  {\bf 4}, 397 (2000)

5000:   [arXiv:hep-th/0002169].

5001:   %%CITATION = 00203,4,397;%%  

5002:   

5003: \bibitem{klemm}

5004: A.~Klemm, ``Topological string theory and integrable structures,''

5005:   Fortsch.\ Phys.\  {\bf 53}, 720 (2005).

5006:   %%CITATION = FPYKA,53,720;%%

5007: 

5008: \bibitem{Klemm:2004km}

5009:   A.~Klemm, M.~Kreuzer, E.~Riegler and E.~Scheidegger,

5010:   ``Topological string amplitudes, complete intersection Calabi-Yau spaces  and

5011:   threshold corrections,''

5012:   JHEP {\bf 0505}, 023 (2005)

5013:   [arXiv:hep-th/0410018].

5014:   %%CITATION = JHEPA,0505,023;%%.          

5015:   

5016: \bibitem{km}

5017:   A.~Klemm and M.~Mari\~no,

5018:   ``Counting BPS states on the Enriques Calabi-Yau,''

5019:   arXiv:hep-th/0512227.

5020:   %%CITATION = HEP-TH 0512227;%%

5021: 

5022: 

5023: \bibitem{Klingen} H.~Klingen, {\it Introductory lectures on Siegel modular forms}, Cambridge University Press, 1990.    

5024: 

5025: \bibitem{cardoso}

5026: G.~Lopes Cardoso, D.~Lust and T.~Mohaupt, ``Threshold corrections and symmetry enhancement in string

5027: compactifications,''

5028:   Nucl.\ Phys.\ B {\bf 450}, 115 (1995)

5029:   [arXiv:hep-th/9412209].

5030:   %%CITATION = HEP-TH 9412209;%%

5031: 

5032:     

5033: \bibitem{mbook}

5034: M. Mari\~no, {\it Chern--Simons theory, matrix models, and topological strings}, Oxford University Press, 2005.

5035: 

5036:  \bibitem{mmopen}

5037:  M.~Mari\~no, ``Open string amplitudes and large order behavior in topological string theory,''

5038:   arXiv:hep-th/0612127.

5039:   %%CITATION = HEP-TH/0612127;%%

5040:   

5041: \bibitem{mm}

5042: M.~Mari\~no and G.~W.~Moore, ``Counting higher genus curves in a Calabi-Yau manifold,''

5043:   Nucl.\ Phys.\ B {\bf 543}, 592 (1999)

5044:   [arXiv:hep-th/9808131].

5045:   %%CITATION = HEP-TH 9808131;%%

5046: 

5047: \bibitem{mnop}

5048: D. Maulik, N. Nekrasov, A. Okounkov, and R. Pandharipande, ``Gromov-Witten theory and Donaldson-Thomas theory, I and II," 

5049: math.AG/0312059 and math.AG/0406092.

5050: 

5051: \bibitem{mp}

5052: D. Maulik and R. Pandharipande, ``New calculations in Gromov--Witten theory," math.AG/0601395.

5053: 

5054: \bibitem{mnw}

5055: J.~A.~Minahan, D.~Nemeschansky and N.~P.~Warner, ``Partition functions for BPS states of the non-critical E(8) string,''

5056:   Adv.\ Theor.\ Math.\ Phys.\  {\bf 1}, 167 (1998)

5057:   [arXiv:hep-th/9707149]; ``Instanton expansions for mass deformed N = 4 super Yang-Mills theories,''

5058:   Nucl.\ Phys.\  B {\bf 528}, 109 (1998) [arXiv:hep-th/9710146].

5059:   %%CITATION = NUPHA,B528,109;%%

5060:   %%CITATION = 00203,1,167;%%

5061:   

5062: \bibitem{mnvw}

5063: J.~A.~Minahan, D.~Nemeschansky, C.~Vafa and N.~P.~Warner, ``E-strings and N = 4 topological Yang-Mills theories,''

5064:   Nucl.\ Phys.\  B {\bf 527}, 581 (1998)

5065:   [arXiv:hep-th/9802168].

5066:   %%CITATION = NUPHA,B527,581;%%

5067:   

5068: \bibitem{nv}

5069:  A.~Neitzke and C.~Vafa, ``Topological strings and their physical applications,''

5070:   arXiv:hep-th/0410178.

5071:   %%CITATION = HEP-TH/0410178;%%

5072: 

5073: 

5074:   

5075: \bibitem{Nek}

5076:  N.~A.~Nekrasov,

5077:   ``Seiberg-Witten prepotential from instanton counting,''

5078:   Adv.\ Theor.\ Math.\ Phys.\  {\bf 7} (2004) 831

5079:   [arXiv:hep-th/0206161].

5080:   %%CITATION = HEP-TH 0206161;%%

5081: 

5082: \bibitem{no}

5083:   N.~Nekrasov and A.~Okounkov,

5084:   ``Seiberg-Witten theory and random partitions,''

5085:   arXiv:hep-th/0306238.

5086:   %%CITATION = HEP-TH 0306238;%%

5087: 

5088: 

5089: \bibitem{sw1}

5090:   N.~Seiberg and E.~Witten,

5091:    ``Electric - magnetic duality, monopole condensation, and confinement in N=2

5092:   supersymmetric Yang-Mills theory,''

5093:   Nucl.\ Phys.\ B {\bf 426} (1994) 19

5094:   [Erratum-ibid.\ B {\bf 430} (1994) 485]

5095:   [arXiv:hep-th/9407087].

5096:   %%CITATION = HEP-TH 9407087;%%

5097: 

5098: \bibitem{sw}

5099:   N.~Seiberg and E.~Witten,

5100:    ``Monopoles, duality and chiral symmetry breaking in N=2 supersymmetric

5101:   QCD,''

5102:   Nucl.\ Phys.\ B {\bf 431} (1994) 484

5103:   [arXiv:hep-th/9408099].

5104:   %%CITATION = HEP-TH 9408099;%%

5105: 

5106: \bibitem{strominger}

5107: A.~Strominger, ``Massless black holes and conifolds in string theory,''

5108:   Nucl.\ Phys.\  B {\bf 451}, 96 (1995)

5109:   [arXiv:hep-th/9504090].

5110:   %%CITATION = NUPHA,B451,96;%%

5111:   

5112:   \bibitem{TianTodorov}

5113:   G. Tian, ``Smoothness of the universal deformation space of compact Calabi--Yau manifolds and its Petersson--Weil metric," 

5114:   in S.--T. Yau (ed.), {\it Mathematical aspects of string theory}, World Scientific, Singapore, 1987, p. 629. A. N. Todorov, ``The Weil-Petersson geometry 

5115:   of the moduli space of $SU(n \ge3)$ (Calabi--Yau) manifolds I," Commun. Math. Phys. {\bf 126} (1989) 325. 

5116:   

5117: \bibitem{Vafa:1995ta}

5118:   C.~Vafa,

5119:   ``A Stringy test of the fate of the conifold,''

5120:   Nucl.\ Phys.\ B {\bf 447} (1995) 252

5121:   [arXiv:hep-th/9505023].

5122:   %%CITATION = HEP-TH 9505023;%%

5123: 

5124: \bibitem{Verlinde:2004ck}

5125:   E.~P.~Verlinde,

5126:   ``Attractors and the holomorphic anomaly,''

5127:   arXiv:hep-th/0412139.

5128:   %%CITATION = HEP-TH/0412139;%%

5129: 

5130: \bibitem{vonk}

5131: M.~Vonk, ``A mini-course on topological strings,''

5132:   arXiv:hep-th/0504147.

5133:   %%CITATION = HEP-TH/0504147;%%

5134: 

5135: \bibitem{Witten:1993ed}

5136:   E.~Witten,

5137:   ``Quantum background independence in string theory,''

5138:   arXiv:hep-th/9306122.

5139:   %%CITATION = HEP-TH/9306122;%%

5140: 

5141: \bibitem{Yamaguchi:2004bt}

5142:   S.~Yamaguchi and S.~T.~Yau,

5143:   ``Topological string partition functions as polynomials,''

5144:   JHEP {\bf 0407}, 047 (2004)

5145:   [arXiv:hep-th/0406078].

5146:   %%CITATION = JHEPA,0407,047;%%

5147:  

5148: 

5149: 

5150: 

5151: \end{thebibliography}

5152: \end{document}

5153: 

5154: 

5155: 

5156: 

5157: 

5158: 

5159: 

5160: 

5161: 

5162: 

5163: 

5164: 

5165: 

5166: 

5167: 

5168: 

5169: