1: \documentclass[11pt,final]{ucthesis}

2: \usepackage{amsmath}

3: \usepackage{amssymb}

4: \usepackage{amsfonts}

5: \usepackage{fancyheadings}

6: \usepackage{amscd}

7: \usepackage{latexsym}

8: \usepackage{theorem}

9: % \usepackage{doublespace}

10: \usepackage{epsfig}

11: \usepackage{psfrag}

12: \usepackage{bbold}

13: 

14: % \include{ltxtemp_th}

15: 

16: \newtheorem{theorem}{Theorem}[chapter]

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

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

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

20: \newtheorem{claim}[theorem]{Claim}

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

22: 

23: {\theorembodyfont{\rmfamily}

24: \theoremstyle{plain}

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

26: \newtheorem{convention}[theorem]{Notational Convention}

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

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

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

30: }

31: 

32: \input{xy}

33: \xyoption{all}

34: 

35: % Margin stuff from jason

36: % \oddsidemargin=0pt

37: % \evensidemargin=0pt

38: % \topmargin=0in

39: \headheight=14pt

40: % \headsep=0pt

41: % \setlength{\textheight}{9in}

42: % \setlength{\textwidth}{6.5in}

43: 

44: \newcommand{\Vol}{\operatorname{Vol}}

45: \renewcommand{\Re}{\operatorname{Re}}

46: \renewcommand{\Im}{\operatorname{Im}}

47: \newcommand{\im}{\operatorname{im}}

48: \newcommand{\id}{\operatorname{id}}

49: \newcommand{\Coker}{\operatorname{Coker}}

50: \renewcommand{\ker}{\operatorname{Ker}}

51: \renewcommand{\dim}{\operatorname{dim}}

52: \newcommand{\Ann}{\operatorname{Ann}}

53: \newcommand{\Sym}{\operatorname{Sym}}

54: \newcommand{\Hom}{\operatorname{Hom}}

55: \newcommand{\tr}{\operatorname{tr}}

56: 

57: \newcommand{\Ld}{\displaystyle\lim}

58: \newcommand{\Od}{\displaystyle\bigoplus}

59: \newcommand{\Pd}{\displaystyle\prod}

60: \newcommand{\Sd}{\displaystyle\sum}

61: 

62: \newcommand{\C}{{\mathbb{C}}}

63: \newcommand{\Z}{{\mathbb{Z}}}

64: \newcommand{\Q}{{\mathbb{Q}}}

65: \renewcommand{\H}{{\mathbb{H}}}

66: \newcommand{\R}{{\mathbb{R}}}

67: \newcommand{\Zt}{{\Z_2}}

68: 

69: \newcommand{\otn}{\{1,\ldots,n\}}

70: \newcommand{\half}{\frac{1}{2}}

71: \newcommand{\bd}{\partial}

72: 

73: \newcommand{\bigmid}{\hs\Big{|}\hs}

74: \newcommand{\subs}{\subseteq}

75: \newcommand{\sups}{\supseteq}

76: \newcommand{\hookto}{{\hookrightarrow}}

77: \newcommand{\onto}{{\twoheadrightarrow}}

78: \renewcommand{\iff}{\Leftrightarrow}

79: \newcommand{\impl}{\Rightarrow}

80: \newcommand{\becircled}{\mathaccent "7017}

81: \newcommand{\hs}{\hspace{3pt}}

82: \renewcommand{\ss}{\substack}

83: 

84: \newcommand{\D}{\Delta}

85: \renewcommand{\i}{\iota}

86: \renewcommand{\a}{\alpha}

87: \renewcommand{\b}{\beta}

88: \newcommand{\s}{\sigma}

89: \newcommand{\eps}{\varepsilon}

90: \renewcommand{\k}{\kappa}

91: \newcommand{\K}{\kappa}

92: \newcommand{\ga}{\gamma}

93: \renewcommand{\t}{\tau}

94: 

95: \renewcommand{\cot}{T^*\C^n}

96: \newcommand{\Cn}{\C^n}

97: \newcommand{\Cno}{\C^{n+1}}

98: 

99: \newcommand{\Mso}{\Ma^{\so}}

100: \newcommand{\Macs}{\Ma^{\gm}}

101: \newcommand{\Ma}{\M}

102: \newcommand{\Xa}{\X}

103: \newcommand{\Gap}{\Gamma_p}

104: \newcommand{\Gapt}{\widehat{\Gamma}_p}

105: \newcommand{\Gapto}{\Gapt\!\becircled{}}

106: 

107: \newcommand{\muc}{\mu_{\C}}

108: \newcommand{\mur}{\mu_{\R}}

109: \newcommand{\barmuc}{\bar{\mu}_{\C}}

110: \newcommand{\barmur}{\bar{\mu}_{\R}}

111: \newcommand{\mr}{\mur}

112: \newcommand{\mc}{\muc}

113: \newcommand{\mhk}{\mu_{\text{HK}}}

114: \newcommand{\mh}{\mr\oplus\mc}

115: \newcommand{\ka}{K_{\a}}

116: \newcommand{\hka}{\hk_{\a}}

117: \newcommand{\hk}{H\!K}

118: \newcommand{\htk}{H^*_{T^k}(pt)}

119: \renewcommand{\mod}{{\!/\!\!/}}

120: \newcommand{\mmod}{{\!/\!\!/\!\!/\!\!/}}

121: 

122: \newcommand{\SO}{{SO(3)}}

123: \newcommand{\SL}{{SL(2,\C)}}

124: \newcommand{\sutd}{{\mathfrak{su}(2)}^*}

125: \newcommand{\sut}{{\mathfrak{su}(2)}}

126: \newcommand{\slt}{{\mathfrak{sl}(2,\C)}}

127: \newcommand{\so}{S^1}

128: \newcommand{\gm}{\C^{\!\times}}

129: 

130: \newcommand{\GC}{G_{\C}}

131: \newcommand{\gdc}{\gd_{\C}}

132: \newcommand{\gd}{\g^*}

133: \newcommand{\g}{\mathfrak{g}}

134: \newcommand{\Tk}{T^k}

135: \newcommand{\Tn}{T^n}

136: \newcommand{\Td}{T^d}

137: \newcommand{\Tkd}{(\tk_{\Z})^*}

138: \newcommand{\Tnd}{(\tn_{\Z})^*}

139: \newcommand{\Tdd}{(\td_{\Z})^*}

140: \newcommand{\tk}{\mathfrak{t}^k}

141: \newcommand{\tn}{\mathfrak{t}^n}

142: \newcommand{\td}{\mathfrak{t}^d}

143: \newcommand{\tkd}{(\tk)^*}

144: \newcommand{\tndz}{(\tn_{\Z})^*}

145: \newcommand{\tnz}{\tn_{\Z}}

146: \newcommand{\tdz}{\td_{\Z}}

147: \newcommand{\tnd}{(\tn)^*}

148: \newcommand{\tdd}{(\td)^*}

149: \newcommand{\tdu}{\mathfrak{t}^*}

150: 

151: \newcommand{\E}{\mathcal{E}}

152: \newcommand{\EA}{\E_A}

153: \renewcommand{\L}{\mathcal{L}}

154: 

155: \newcommand{\Af}{\operatorname{Rep}(Q)}

156: \newcommand{\Diag}{\operatorname{Diag}}

157: 

158: % ht commands

159: \newcommand{\ktd}{\k_{\Td}}

160: \newcommand{\ktdso}{\k_{\Td\times\so}}

161: \newcommand{\kso}{\k_{\so}}

162: \newcommand{\otd}{\{1,\ldots,d\}}

163: \newcommand{\Tdrs}{T_{\R}^d\times\Zt}

164: \newcommand{\Tdr}{T_{\R}^d}

165: \newcommand{\Tdso}{\Td\times S^1}

166: \newcommand{\Tnso}{\Tn\times S^1}

167: \newcommand{\hma}{H^*(\Ma)}

168: \newcommand{\htm}{H^*_{\Td}(\Ma)}

169: \newcommand{\htsm}{H^*_{\Td\times S^1}(\Ma)}

170: \newcommand{\hsm}{H^*_{S^1}(\Ma)}

171: \newcommand{\hr}{H^*_{\Tdr}(\M_{\R};\Zt)}

172: \newcommand{\hrs}{H^*_{\Tdr\times\Zt}(\M_{\R};\Zt)}

173: \newcommand{\hscomp}{H^*_{\Zt}(\MA;\Zt)}

174: \newcommand{\hcomp}{H^*(\MA;\Zt)}

175: \newcommand{\bomp}{\C^d\setminus\cup_{i=1}^n H_i^{\C}}

176: \newcommand{\os}{\mathcal{OS}}

177: \newcommand{\ost}{\mathcal{OS}\otimes\Zt}

178: 

179: % os commands

180: \newcommand{\A}{\mathcal{A}}

181: \newcommand{\MA}{\mathcal{M}(\A)}

182: \newcommand{\hz}{H_{\Zt}^*(pt)}

183: 

184: % hp commands

185: \newcommand{\ctn}{{\C^{2n}}}

186: \newcommand{\ses}{st}

187: \newcommand{\Ae}{[A;e_1,\ldots,e_n]}

188: \newcommand{\pathsp}{\prod_{i=1}^n S^2_{\a_i}}

189: \newcommand{\polsa}{\X_S}

190: \newcommand{\pola}{\X}

191: \newcommand{\mcs}{\mathcal{S}}

192: \newcommand{\sa}{\mcs}

193: \newcommand{\spa}{\mcs'}

194: \newcommand{\core}{\mathfrak{L}}

195: \newcommand{\corea}{\mathfrak{L}_{\a}}

196: \newcommand{\zns}{z_{\ns}

197: \newcommand{\zms}{z_{\ms}}}

198: \newcommand{\ms}{m_S}

199: \newcommand{\ns}{n_S}

200: \newcommand{\bi}{d_i}

201: \newcommand{\bk}{d_k}

202: \newcommand{\bj}{d_j}

203: \newcommand{\ba}{d_A}

204: \newcommand{\bns}{d_{\ns}}

205: \newcommand{\ci}{c_i}

206: \newcommand{\ck}{c_k}

207: \newcommand{\cns}{c_{\ns}}

208: \newcommand{\Sb}{\bar{S}}

209: \newcommand{\Lb}{\overline{S^c}}

210: \newcommand{\Ac}{A^c}

211: \newcommand{\asb}{A\cap\Sb}

212: \newcommand{\acsb}{\Ac\cap\Sb}

213: \newcommand{\pis}{\prod_{i\in\Sb}}

214: \newcommand{\pjl}{\prod_{j\in\Lb}}

215: \newcommand{\vs}{v_S}

216: \newcommand{\vst}{\tilde{v}_S}

217: \newcommand{\twn}{\{2,\ldots,n\}}

218: \newcommand{\Zb}{\Z[d_1,\ldots,d_n]/\langle\bk^2-d_1\bk\mid k=2,\ldots,n\rangle}

219: \newcommand{\Qb}{\Q[d_1,\ldots,d_n]/\langle\bk^2-d_1\bk\mid k=2,\ldots,n\rangle}

220: \newcommand{\us}{U_S}

221: \newcommand{\mtus}{X_T\cap\us}

222: \newcommand{\ts}{T\sups S}

223: \newcommand{\sz}{|S|-1}

224: 

225: % hm commands

226: \newcommand{\Kh}{\hat\k}

227: \newcommand{\Hn}{\H^n}

228: \newcommand{\rgt}{r_T^G}

229: \newcommand{\phiw}{\phi^W}

230: \newcommand{\phat}{\hat{\phi}}

231: \newcommand{\phatw}{\hat{\phi}^W}

232: \newcommand{\IM}{\int_M}

233: \newcommand{\IN}{\int_N}

234: \newcommand{\IG}{\int_G}

235: \newcommand{\IF}{\int_F}

236: \newcommand{\intxt}{\int_{X\mmod T}}

237: \newcommand{\intxg}{\int_{X\mmod G}}

238: \newcommand{\I}{\mathcal{I}}

239: \newcommand{\J}{\mathcal{J}}

240: \newcommand{\imkt}{\left(\Im\K_T\right)}

241: \newcommand{\hso}{H_{\so}^*}

242: \newcommand{\hhso}{\widehat{H}_{\so}^*}

243: \newcommand{\hmg}{H_{\so}^*(M\mmod G)}

244: \newcommand{\hmt}{H_{\so}^*(M\mmod T)}

245: \newcommand{\hxg}{H_{\so}^*(X\mmod G)}

246: \newcommand{\hxt}{H_{\so}^*(X\mmod T)}

247: \newcommand{\hhxg}{\widehat{H}_{\so}^*(X\mmod G)}

248: \newcommand{\hhxt}{\widehat{H}_{\so}^*(X\mmod T)}

249: \newcommand{\hhst}{\widehat{H}_{\so\times T}^*(X)}

250: \newcommand{\hhsg}{\widehat{H}_{\so\times G}^*(X)}

251: \newcommand{\hp}{H_{S^1}^*(pt)}

252: \newcommand{\hhp}{{\mathbb K}}

253: \newcommand{\hm}{\hso(M)}

254: \newcommand{\hhn}{\hhso(N)}

255: \newcommand{\hhm}{\hhso(M)}

256: \newcommand{\X}{{\mathfrak{X}}}

257: \newcommand{\M}{{\mathfrak{M}}}

258: % \newcommand{\M}{M}

259: \newcommand{\MM}{{\M\times\M}}

260: \newcommand{\NN}{\mathfrak{N}}

261: \newcommand{\dso}{\Diag_*(1)}

262: \newcommand{\muh}{\mu_{G}}

263: \newcommand{\mubh}{\mu_{T}}

264: \newcommand{\bigmod}{{\Big/\!\!\!\!\Big/}}

265: \newcommand{\bigmmod}{{\Big/\!\!\!\!\Big/\!\!\!\!\Big/\!\!\!\!\Big/}}

266: 

267: % kernel commands

268: \newcommand{\ra}{\rangle}

269: \newcommand{\Hir}{H^r_i}

270: \newcommand{\PrA}{P^r_{\! A}}

271: \renewcommand{\Pr}{P^r}

272: \newcommand{\Dr}{\D^r}

273: \newcommand{\DrA}{\D^r_A}

274: \newcommand{\DrzA}{\D^{r_0}_A}

275: \newcommand{\rt}{\tilde r}

276: \newcommand{\odra}{\mathbf{1}_{\DrA}}

277: \newcommand{\odr}{\mathbf{1}_{\Dr}}

278: \newcommand{\WA}{W_A}

279: \newcommand{\WAbd}{W_A^{bd}}

280: \newcommand{\Fbd}{\mathcal{F}^{bd}}

281: \newcommand{\F}{\mathcal{F}}

282: 

283: \newcommand{\<}{\left<}

284: \renewcommand{\>}{\right>}

285: \newcommand{\la}{\langle}

286: \renewcommand{\ra}{\rangle}

287: \renewcommand{\(}{\left(}

288: \renewcommand{\)}{\right)}

289: 

290: \newcommand{\qed}{\hfill \mbox{$\Box$}\medskip\newline}

291: \newenvironment{proof}{\noindent {\bf Proof:}}{\qed \par}

292: \newenvironment{noproof}{\noindent {\bf Proof:}}{\newline}

293: \newenvironment{sketch}{\noindent {\bf Sketch of Proof:}}{\qed \par}

294: 

295: \newenvironment{proofcomb}{\noindent {\bf Proof of \ref{comb}:}}{\qed\par}

296: \newenvironment{proofint}{\noindent {\bf Proof of \ref{int}:}}{}

297: \newenvironment{proofintegration}{\noindent {\bf Proof of \ref{integration}:}}{\qed \par}

298: \newenvironment{proofordinary}{\noindent {\bf Proof of \ref{ordinary}:}}{\qed \par}

299: \newenvironment{proofhtsm}{\noindent {\bf Proof of \ref{htsm}:}}{\qed \par}

300: \newenvironment{proofp}{\noindent {\bf Proof of \ref{placement}:}}{\qed \par}

301: \newenvironment{proofc}{\noindent {\bf Proof of \ref{coorientation}:}}{\qed \par}

302: \newenvironment{proofhsm}{\noindent {\bf Proof of \ref{hsm}:}}{\qed \par}

303: \newenvironment{proofmain}{\noindent {\bf Proof of \ref{main}:}}{\qed \par}

304: \newenvironment{prooftech}{\noindent {\bf Proof of \ref{tech}:}}{\qed \par}

305: \newenvironment{componentproof}{\noindent {\bf Proof of \ref{component}:}}{\qed \par}

306: \newenvironment{usproof}{\noindent {\bf Proof of \ref{us}:}}{\qed \par}

307: \newenvironment{mainproof}{\noindent {\bf Proof of \ref{main}:}}{\qed \par}

308: \newenvironment{eqcoreproof}{\noindent {\bf Proof of \ref{eqcore}:}}{\qed \par}

309: 

310: % \newcommand{\thisisstupid}{\S}

311: % \makeatletter

312: % \renewcommand{\@seccntformat}[1]{$\thisisstupid$

313: % \csname the#1\endcsname\hspace{0.5em}}

314: % \makeatother

315: 

316: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%

317: %

318: % Double spacing, if you want it.

319: %

320: %%%%%%%%%%%%%%%%%%%%%%%%%%%%

321: % \def\dsp{\def\baselinestretch{2.0}\large\normalsize}

322: % \dsp

323: 

324: % \addtolength{\voffset}{+0.1in}

325: % %\addtolength{\hoffset}{+0.5in}

326: % \addtolength{\textwidth}{-0.2in}

327: % \addtolength{\textheight}{-0.4in}

328: 

329: \hyphenation{mar-gin-al-ia}

330: 

331: \begin{document}

332: 

333: \title{Hyperk\"ahler Analogues of K\"ahler Quotients}

334: \author{Nicholas James Proudfoot}

335: \degreesemester{Spring}

336: \degreeyear{2004}

337: \degree{Doctor of Philosophy}

338: \chair{Professor Allen Knutson}

339: \othermembers{Professor Robion Kirby \\

340:   Professor Robert Littlejohn}

341: \numberofmembers{3}

342: \prevdegrees{A.B. (Harvard University) 2000}

343: \field{Mathematics}

344: \campus{Berkeley}

345:  

346: \maketitle

347: \approvalpage

348: \copyrightpage

349: \pagestyle{fancyplain}

350: % \include{abstract}

351: 

352: \begin{abstract}

353: Let $\X$ be a K\"ahler manifold that is presented as a K\"ahler

354: quotient of $\Cn$ by the linear action of a compact group $G$.

355: We define the {\em hyperk\"ahler analogue} $\M$ of $\X$ as a

356: hyperk\"ahler quotient of the cotangent bundle $\cot$ by the induced

357: $G$-action.  Special instances of this construction include

358: hypertoric varieties \cite{BD,K1,HS,HP1} and quiver varieties \cite{N1,N2,N3}.

359: One of our aims is to provide a unified treatment of these

360: two previously studied examples.

361: 

362: The hyperk\"ahler analogue $\M$ is noncompact, but this noncompactness

363: is often ``controlled'' by an action of $\gm$ descending from

364: the scalar action on the fibers of $\cot$.  Specifically, we are

365: interested in the case where the moment map for the action

366: of the circle $\so\subs\gm$

367: is proper.  In such cases, we

368: define the {\em core} of $\M$, a reducible, compact subvariety onto which

369: $\M$ admits a circle-equivariant deformation retraction.  One of the components

370: of the core is isomorphic to the original K\"ahler manifold $\X$.

371: % When $\X$ is a toric variety, every other core component is also a toric variety.  

372: When $\X$ is a moduli space of polygons in $\R^3$, we interpret

373: each of the other core components of $\M$ as related polygonal moduli spaces.

374: 

375: Using the circle action with proper moment map, we define an integration

376: theory on the circle-equivariant cohomology of $\M$, 

377: motivated by the well-known localization theorem 

378: of \cite{AB} and \cite{BV}.  This allows us to prove a hyperk\"ahler

379: analogue of Martin's theorem \cite{Ma}, which describes the cohomology

380: ring of an arbitrary K\"ahler quotient in terms of the 

381: cohomology of the quotient by a maximal torus.

382: This theorem, along with a direct analysis of the equivariant cohomology

383: ring of a hypertoric variety, gives us a method for computing

384: the equivariant cohomology ring of many hyperk\"ahler analogues,

385: including all quiver varieties.

386: \abstractsignature

387: \end{abstract}

388: 

389: \pagestyle{plain}

390: \begin{frontmatter}

391: \pagenumbering{roman}

392: % \begin{dedication}

393: % \null\vfil

394: % {\large

395: % \begin{center}

396: % To the fond memory of a great teacher, \\

397: % Edwin L. Shay

398: % \end{center}}

399: % \vfil\null

400: % \end{dedication}

401:  

402: \tableofcontents

403: %\addcontentsline{toc}{sec_name}{Chapter 1}

404: %\listoffigures

405: %\listoftables

406: \lhead[\fancyplain{}{}]{\fancyplain{}{}}

407: \rhead[\fancyplain{}{}]{\fancyplain{}{}}

408: 

409: \begin{acknowledgements}

410: During my years at Berkeley, I have had countless invaluable conversations

411: with other students, post-docs, professors, and visitors to Berkeley and MSRI.

412: Many of these conversations

413: have been relevant to the work described here, and many more

414: have greatly influenced my general mathematical perspective.

415: I would like to particularly acknowledge my coauthors, Megumi Harada

416: and Tam\'as Hausel, and my advisor, Allen Knutson.

417: I feel extremely lucky to be part of such a warm and lively community.

418: \end{acknowledgements}

419: 

420: \end{frontmatter}

421: 

422: \pagestyle{fancy}

423: \pagenumbering{arabic}

424: \setlength{\headrulewidth}{1pt}

425: \lhead[\rm\thepage]{\sl\leftmark}

426: \rhead[\sl\leftmark]{\rm\thepage}

427: \renewcommand{\chaptermark}[1]{\markboth{\chaptername\hs\thechapter.  #1}{ }}

428: \cfoot{}

429: % \renewcommand{\chaptermark}[1]{\markboth{#1}{ }}

430: % \rhead[\fancyplain{}{\bfseries \leftmark}]{\fancyplain{}{\bfseries \thepage}}

431: 

432: \begin{chapter}{Introduction}

433: % A {\em K\"ahler} manifold is a riemannian manifold $(X,g)$ with a parallel, 

434: % orthogonal complex structure $J$.  These conditions imply that the $2$-form

435: % $\omega(v,w) := g(Jv,w)$ is closed and nondegenerate, 

436: % hence a K\"ahler manifold may be thought of

437: % as a manifold which is simultaneously riemannian, 

438: % complex, and symplectic, in a compatible way.

439: % Interesting examples of K\"ahler manifolds include 

440: % toric varieties \cite{Do}, polygon spaces \cite{HK},

441: % and the moduli space of stable vector bundles on a curve \cite{NS}.

442: We begin with a quick overview of some of the structures that we will consider

443: in this thesis, and the types of questions that we will be asking.  

444: Detailed definitions will be deferred until the next chapter.

445: 

446: Let $G$ be a compact Lie group acting linearly on $\Cn$,

447: with moment map $\mu:\Cn\to\gd$, and suppose we are given 

448: a central (i.e. $G$-fixed) regular value $\a\in\gd$ of $\mu$.  

449: From this data, we may define the K\"ahler 

450: quotient $$\X := \Cn\mod G = \mu^{-1}(\a)/G,$$ which itself inherits the structure of a 

451: K\"ahler manifold. 

452: % (If the complexification $G^{\C}$ 

453: % of $G$ is algebraic, and acts algebraically

454: % on $\Cn$, then $\X$ may be thought of as 

455: (We may also think of $\X$ as the geometric invariant theory

456: quotient of $\Cn$ by $G^{\C}$ in the sense of Mumford \cite[\S 8]{MFK};

457: see Proposition \ref{hkhs}.)

458: % Toric varieties arise from the case where $G$ is abelian, and polygon spaces

459: % arise when the vector space in question is the space of representations of a

460: % certain quiver (see Section \ref{quiver}).

461: % The moduli space of stable vector bundles on a curve arises from an analogous

462: % construction in which $\Cn$ is replaced by the infinite-dimensional affine space

463: % of holomorphic structures on a fixed smooth bundle $E$, which is acted on by

464: % the gauge group of automorphisms of $E$.

465: A {\em hyperk\"ahler} manifold is a riemannian manifold $(M,g)$ equipped

466: with three orthogonal complex structures $J_1, J_2, J_3$ and three compatible symplectic

467: forms $\omega_1, \omega_2, \omega_3$ such that $J_1J_2= -J_2J_1 = J_3$

468: for $i=1,2$, and $3$.

469: The cotangent bundle $\cot$ has a natural hyperk\"ahler structure,

470: and this structure is preserved by the induced action of $G$.

471: Furthermore, there exist maps $\mu_i:\cot\to\gd$ for $i=1,2,3$

472: such that $\mu_i$ is a moment map for the action of $G$ with respect

473: to the symplectic form $\omega_i$.

474: We define the {\em hyperk\"ahler analogue} of $\X$

475: to be the hyperk\"ahler quotient

476: $$\M := \cot\mmod G = \Big(\mu_1^{-1}(\a)\cap\mu_2^{-1}(0)

477: \cap\mu_3^{-1}(0)\Big)\Big/G.$$

478: (The set

479: $\mu_2^{-1}(0)\cap\mu_3^{-1}(0)\subs\cot$

480: is a complex subvariety with respect to $J_1$, and $\M$

481: may be thought of as the geometric invariant theory quotient

482: of this variety by $G^{\C}$.)

483: The quotient $\M$ is a complete hyperk\"ahler manifold \cite{HKLR},

484: containing $T^*\X$ as a dense open subset (see Proposition \ref{denseopen}).

485: The following is a description of some well-known classes of K\"ahler quotients,

486: along with their hyperk\"ahler analogues.

487: 

488: \vspace{.2cm}

489: \noindent{\bf Toric and hypertoric varieties.}

490: These examples comprise the case where $G$ is abelian.  

491: The geometry of toric varieties is deeply related to the combinatorics of polytopes;

492: for example, Stanley \cite{St} used the hard Lefschetz theorem for toric varieties

493: to prove certain inequalities for the $h$-numbers of a simplicial polytope.

494: Hypertoric varieties, introduced by Bielawski and Dancer \cite{BD}, interact in a similar

495: way with the combinatorics of real hyperplane arrangements.

496: Following Stanley's work, Hausel and Sturmfels \cite{HS} used the hard Lefschetz theorem

497: on a hypertoric variety to give a geometric interpretation of some previously-known

498: inequalities for the $h$-numbers of a rationally representable matroid.

499: In Chapter \ref{ht} we will explore further combinatorial properties

500: of the various equivariant cohomology rings of a hypertoric variety.

501: 

502: \vspace{.2cm}

503: \noindent{\bf Quiver varieties.}

504: For any directed graph,

505: Nakajima \cite{N1,N2,N3} defined a {\em quiver variety}

506: to be the hyperk\"ahler analogue of the moduli space of framed representations of that graph

507: (see Section \ref{quiver}).

508: Examples include the Hilbert scheme of $n$ points in $\C^2$ \cite{N4}, the moduli space

509: of instantons on an ALE space \cite{N1}, and Konno's hyperpolygon spaces \cite{K2,HP2},

510: which are the hyperk\"ahler analogues of the moduli spaces of $n$-sided polygons in $\R^3$ with

511: fixed edge lengths.

512: Quiver varieties have received much attention from representation theorists due to the actions

513: of various infinite-dimensional Lie algebras on the cohomology and

514: K-theory of a quiver variety (see, for example, \cite{N3,N5}).

515: 

516: \vspace{.2cm}

517: \noindent{\bf Moduli spaces of bundles and connections.}

518: Narasimhan and Seshadri \cite{NS} defined a notion of stability for a vector bundle

519: on a Riemann surface $\Sigma$, and proved that the moduli space of stable 

520: holomorphic bundles on $\Sigma$ may be identified with the moduli space of irreducible, 

521: flat, unitary connections.  

522: Atiyah and Bott presented this space as a K\"ahler quotient of the affine

523: space of all connections on a fixed bundle $E$ by the gauge group

524: of automorphisms of $E$.  This picture can be complexified by replacing holomorphic

525: bundles with Higgs bundles, and unitary connections with arbitrary ones \cite{Hi}.

526: The correspondence between Higgs bundles, flat connections, and representations

527: of the fundamental group is known as {\em nonabelian Hodge theory}, and has

528: been studied and generalized extensively by Simpson \cite{Si} in addition to

529: many other authors.

530: These constructions involve taking a quotient of an infinite dimensional affine

531: space by an infinite dimensional group, and therefore lie

532: it is beyond the scope of this work.

533: Many of our techniques, however, can be applied in this context.

534: See for example \cite{H1, HT1, HT2}.

535: % A Higgs bundle on a Riemann surface $\Sigma$ is a pair $(E,\theta)$, where

536: % $E$ is a holomorphic vector bundle on $\Sigma$, and $\Phi$ is an $\operatorname{End}(E)$

537: % valued $1$-form of fixed determinant.  Hitchin \cite{Hi} generalized the notion

538: % of stability of vector bundles to Higgs bundles, 

539: % and interpreted the moduli space of stable rank $2$ Higgs bundles on $\Sigma$

540: % as the moduli space of self-dual solutions to the Yang-Mills equations.

541: % In arbitrary rank, Simpson \cite{Si} showed that 

542: % the moduli space of stable Higgs bundles

543: % is diffeomorphic to the moduli space of connections on $\Sigma$ of constant central

544: % curvature.  Taking this constant to be zero, we obtain the (singular) moduli space

545: % of representations of the fundamental group of $\Sigma$.

546: % This correspondence between Higgs bundles, flat connections, and representations

547: % of the fundamental group is known as {\em nonabelian Hodge theory}, and has

548: % been studied extensively by other authors.

549: 

550: \vspace{.2cm}

551: Consider the action of the multiplicative group

552: $\gm$ on $\cot$ given by scalar multiplication of the fibers.

553: The action of the compact subgroup $\so\subs\gm$ is hamiltonian

554: with respect to the first symplectic form $\omega_1$, and descends

555: to a circle action on $\M$ with moment map $\Phi:\M\to\R$, which is a Morse-Bott function.

556: The geometry and topology 

557: associated with this action will be our main object of study.

558: In Chapter \ref{analogues} we give a detailed discussion of the construction of $\M$,

559: along with the action of $\gm$.  In the case where $\Phi$ is proper,

560: we describe a reducible subvariety $\core\subs\M$ called the {\em core} of $\M$,

561: onto which $\M$ retracts $\so$-equivariantly.

562: The core $\core$ has $\X$ 

563: as one of its components, and if $\M$ is smooth, then $\core$ is equidimensional of dimension

564: $\dim\X = \half\dim\M$.  In particular, the fundamental classes of the components

565: of $\core$ provide a natural basis for the top degree cohomology of $\M$.  This fact is

566: exploited for hypertoric varieties in \cite{HS}, and for quiver varieties in various

567: papers of Nakajima.  Building on \cite{HP1}, this thesis is the first 

568: unified treatment of hyperk\"ahler analogues and their cores,

569: encompassing both hypertoric varieties and quiver varieties.

570: 

571: The geometry of the core of $\M$ will be one of two major themes that we consider.

572: In the case where $\M$ is a hypertoric variety, each of the components of the core

573: $\core$ is itself a toric variety (Lemma \ref{htcorecomp}), as first shown in \cite{BD}.

574: In section \ref{htcore}, we give an explicit description of the action of 

575: $\gm$ and the gradient flow of $\Phi$ on each piece.  The case of hyperpolygon spaces

576: is more interesting.  The ordinary polygon space $\X$ is the moduli space of $n$-sided polygons

577: in $\R^3$ with fixed edge lengths.  In Section \ref{modcore}, we show that the other

578: core components are smooth, and may themselves be interpreted as moduli spaces

579: of polygons in $\R^3$ satisfying certain conditions (Theorem \ref{us}).

580: Thus, for the special case of hyperpolygon spaces, 

581: we have solved the following general problem:

582: 

583: \begin{problem}

584: Given any moduli space $\X$ that can be constructed as a K\"ahler reduction

585: (or GIT quotient) of complex affine space, is it possible

586: to interpret the core of the hyperk\"ahler analogue

587: $\X$ as a union of moduli spaces corresponding to other, 

588: related moduli problems?

589: \end{problem}

590: 

591: Our second major theme will be the circle-equivariant cohomology ring of $\M$.

592: In Chapter \ref{ht} we compute the circle-equivariant

593: cohomology ring of a hypertoric variety, and as an application compute the

594: $\Zt = Gal(\C/\R)$ equivariant cohomology ring of the complement of a complex

595: hyperplane arrangement defined over $\R$.  The purpose of Chapter \ref{abelianization} is 

596: to extend to the hyperk\"ahler setting a theorem of Martin \cite{Ma}, which describes

597: how to compute the cohomology ring of a K\"ahler quotient $X\mod G$ in terms

598: of the cohomology ring of the abelian quotient $X\mod T$, where $T\subs G$ is a maximal

599: torus.  The main technical difficulty arises from the fact that 

600: Martin's theorem relies heavily on computing integrals, which is not possible

601: on the noncompact hyperk\"ahler analogues that we have defined.

602: Our approach is to make use of the localization theorem of \cite{AB,BV}, which

603: allows us to define an integration theory in the circle-equivariant cohomology

604: of $\so$-manifolds with compact, oriented fixed point set.  This is perhaps the single

605: most important reason for considering the circle action on a hyperk\"ahler analogue.

606: In Section \ref{hp} we combine the results of Chapters \ref{ht} and \ref{abelianization}

607: to compute the equivariant cohomology ring of a hyperpolygon space, and of each

608: of its core components.

609: 

610: Most of Chapter \ref{ht} (with the exception of Section \ref{cogs})

611: appeared first in \cite{HP1}, and Chapter \ref{abelianization}

612: is a reproduction of \cite{HP}.  Chapter \ref{hpspaces} 

613: is taken primarily from \cite{HP2}, with the exception of Section \ref{s1},

614: which comes from \cite{HP}.

615: \end{chapter}

616: 

617: \begin{chapter}{Hyperk\"ahler analogues}\label{analogues}

618: Our plan for this chapter is to provide a unified approach to the constructions of

619: hypertoric varieties and quiver varieties, which are the two major classes of examples

620: of hyperk\"ahler analogues of familiar K\"ahler varieties that appear in the literature.

621: In Section \ref{reduction} we give the basic construction of the hyperk\"ahler analogue

622: $\M$ of a K\"ahler quotient $\X=\Cn\mod G$, and show that $\M$ may be understood

623: as a partial compactification of the cotangent bundle to $\X$ 

624: (Proposition \ref{denseopen}).

625: In Section \ref{circleaction}, we define a natural action of the group $\gm$

626: on $\M$, which is holomorphic with respect to one of the complex structures.

627: This action will be our main tool for studying the geometry of $\M$

628: in future chapters.

629: Some of this material appeared first in \cite[\S 1]{HP1}.

630: 

631: \begin{section}{Hyperk\"ahler and holomorphic 

632: symplectic reduction}\label{reduction}

633: A {\em hyperk\"ahler manifold} is a Riemannian manifold $(M,g)$ along with

634: three orthogonal, parallel complex structures, $J_1,J_2,J_3$,

635: satisfying the usual quaternionic relations. 

636: These three complex structures allow us to define three symplectic forms

637: $$\omega_1(v,w) = g(J_{1}v, w), \hs\omega_2(v,w) = g(J_{2}v,w), 

638: \hs\omega_3(v,w) = g(J_{3}v,w),$$ so that $(g, J_i, \omega_i)$ is a K\"ahler

639: structure on $M$ for $i=1,2,3$.

640: The complex-valued two-form $\omega_2 + i

641: \omega_3$ is closed, nondegenerate, and holomorphic

642: with respect to the complex structure $J_1$.

643: Any hyperk\"ahler manifold can therefore be considered as a {\em holomorphic

644: symplectic} manifold with complex structure $J_1$, real symplectic

645: form $\omega_{\R} := \omega_1$, and holomorphic symplectic form

646: $\omega_{\C} := \omega_2 + i\omega_3$.

647: This is the point of view that we will adopt.

648: 

649: We will refer to an action of $G$ on a hyperk\"ahler manifold

650: $M$ as {\em hyperhamiltonian}

651: if it is hamiltonian with respect to $\omega_{\R}$ and holomorphic

652: hamiltonian with respect to $\omega_{\C}$, with $G$-equivariant

653: moment map

654: $$\mhk:=\mr\oplus\mc:M\to\gd\oplus\g_{\C}^*.$$

655: The following theorem describes the {\em hyperk\"ahler quotient}

656: construction, a quaternionic analogue of the K\"ahler quotient.

657: 

658: \begin{theorem}\label{hklr}{\em\cite{HKLR}}

659: Let $M$ be a hyperk\"ahler manifold equipped with a hyperhamiltonian action of a

660: compact Lie group $G$, with moment maps $\mu_1, \mu_2, \mu_3$.

661: Suppose $\xi = \xi_{\R}\oplus \xi_{\C}$ is a

662: central regular value of $\mhk$, with $G$ acting freely

663: on $\mhk^{-1}(\xi)$. Then there is a unique

664: hyperk\"ahler structure

665: on the hyperk\"ahler quotient

666: $\M = M \mmod_{\!\!\xi}G := \mhk^{-1}(\xi)/G$,

667: with associated symplectic and holomorphic symplectic forms

668: $\omega^{\xi}_{\R}$ and $\omega^{\xi}_{\C}$, such that

669: $\omega^{\xi}_{\R}$ and $\omega^{\xi}_{\C}$ pull back to the restrictions

670: of $\omega_{\R}$ and $\omega_{\C}$ to $\mhk^{-1}(\xi)$.

671: \end{theorem}

672: 

673: For a general regular value $\xi$, the action of $G$ on

674: $\mhk^{-1}(\xi)$ will not be free, but only locally free.

675: To deal with this situation we must introduce the notion

676: of a hyperk\"ahler orbifold.

677: 

678: An {\em orbifold} is a topological space $M$ locally modeled

679: on finite quotients of euclidean space.

680: More precisely, $M$ is a Hausdorff topological space,

681: equipped with an {\em atlas} $\mathcal U$ {\em of uniformizing charts}.

682: This consists of a collection of quadruples

683: $(\tilde U, \Gamma, U, \phi)$, where $\tilde U$ is an open subset

684: of euclidean space, $\Gamma$ is a finite group acting on $\tilde U$ and

685: fixing a set of codimension at least $2$, 

686: $U$ is an open subset of $M$, and $\phi$ is a homeomorphism from

687: $\tilde U/\Gamma$ to $U$.  The sets $U$ are required to cover $M$,

688: and the quadruples must satisfy a list of compatibility

689: conditions, as in \cite{LT}.

690: 

691: Given a point $p\in M$, the {\em orbifold group at $p$} is the isotropy

692: group $\Gamma_p\subs\Gamma$ of a point 

693: $\tilde p \in \phi^{-1}(p)\subs\tilde U$

694: for any quadruple $(\tilde U, \Gamma, U, \phi)$ such that $U$ contains $p$.

695: The orbifold tangent space $T_pM = T_{\tilde p}\tilde U_{\tilde p}$ 

696: should be thought of not as a vector space,

697: but rather as a representation of $\Gamma_p$ (see Proposition \ref{coreprops}).

698: A differential form on an orbifold may be thought of as a collection

699: of $\Gamma$-invariant differential forms on the open sets $\tilde U$,

700: subject to certain compatibility conditions.  We may define riemannian metrics,

701: complex structures, vector bundles, K\"ahler structures, and hyperk\"ahler

702: structures on orbifolds in a similar manner.

703: 

704: \begin{example}

705: Let $Z$ be a smooth manifold, and let $G$ be a compact Lie group

706: acting locally freely on $Z$.  Then $Z/G$ inherits the structure 

707: of an orbifold.  The orbifold group of an orbit of a point $z\in Z$ is simply

708: the stabilizer $G_z\subs G$.

709: Any $G$-invariant tensor on $Z$ descends to an orbifold

710: tensor on $Z/G$.  Any $G$-equivariant vector bundle on $Z$ descends

711: to a vector bundle on $Z/G$.

712: {\em All of the orbifolds that we consider will be of this form}

713: (and it is not known whether any other examples exist).

714: \end{example}

715: 

716: These definitions allow for a straightforward extension of 

717: Theorem \ref{hklr} to the case where $\xi$ is an arbitrary regular value

718: of $\mhk$.  This implies, by the moment map condition, that $G$

719: acts locally freely on $\mhk^{-1}(0)$, and that the quotient

720: $\mhk^{-1}(0)/G$ inherits the structure of a hyperk\"ahler orbifold.

721: 

722: Orbifolds are in many ways just as nice as manifolds; for example,

723: it is possible to adapt Morse theory to the orbifold case, as in \cite{LT},

724: which we will use in the next section.  When we say that a certain

725: K\"ahler or hyperk\"ahler quotient is an orbifold, we wish to express

726: the opinion that it is relatively well behaved, rather than

727: the opinion that it is nasty and singular.  For this reason, we will

728: use the positively connoted adjective $\Q$-{\em smooth}

729: to refer to orbifolds.

730: 

731: We now specialize to the case where $M = \cot$, and the action

732: of $G$ on $\cot$ is induced from a linear action of $G$ on $\Cn$

733: with moment map $\mu:\Cn \to \gd$.

734: Choose an identification of $\Hn$ with $\cot$

735: such that the complex structure $J_1$ on $\Hn$

736: given by right multiplication by $i$

737: corresponds to the natural complex structure on $\cot$.

738: Then $\cot$ inherits a hyperk\"ahler structure.

739: The real symplectic form $\omega_{\R}$ is given by

740: adding the pullbacks of the standard forms on $\Cn$ and $(\Cn)^*$,

741: and the holomorphic symplectic form $\omega_{\C} = d\eta$, where $\eta$ is the canonical

742: holomorphic 1-form on $\cot$.

743:  

744: We note that $G$ acts $\H$-linearly on $\cot\cong\Hn$

745: (where $n\times n$ matrices

746: act on the left on the space of column vectors $\Hn$, and scalar multiplication by $\H$

747: is on the right).

748: This action is hyperhamiltonian

749: with moment map $\mhk=\mc\oplus\mr$,

750: where

751: $$\mr(z,w) = \mu(z)-\mu(w)\hspace{10pt}\text{and}

752: \hspace{10pt}\mc(z,w)(v) = w(\hat{v}_z)$$

753: for $w\in T^*_z\Cn,\hs v\in\g_{\C}$, and $\hat{v}_z$ the element

754: of $T_z\Cn$ induced by $v$.

755: Given a central element $\a\in\gd$,

756: we refer to the hyperk\"ahler quotient $$\M = \cot \mmod_{\!\!(\a,0)} G$$

757: as the {\em hyperk\"ahler analogue} of the corresponding K\"ahler quotient

758: $$\X=\Cn\mod_{\!\!\a} G := \mu^{-1}(\a)/G.$$

759: In future sections we will often fix a parameter $\a$ and drop it from the notation.

760: 

761: At times it will be convenient to think of K\"ahler quotients

762: in terms of geometric invariant theory, as follows.

763: Let $G_{\C}$ be an algebraic group acting on an affine variety $V$,

764: and let $\chi:G_{\C}\to\gm$ be a character of $G_{\C}$.

765: This defines a lift of the action of $G_{\C}$ to the trivial line

766: bundle $V\times \C$ by the formula

767: $$g\cdot(v,z) = (g\cdot v, \chi(g)^{-1}z).$$

768: The {\em semistable locus} $V^{s}$ with respect to 

769: $\chi$ is defined to be the set

770: of points $v\in V$ such that for $z\neq 0$, the closure

771: of the orbit $G_{\C}(v,z)\subs V\times\C$ is disjoint from the

772: zero section $V\times\{0\}$ (see \cite{MFK} or \cite{N4}).

773: The {\em geometric invariant theory} (GIT) quotient $V\mod_{\!\!\chi}G_{\C}$

774: of $V$ by $G_{\C}$ at $\chi$ is an algebraic variety with underlying

775: space $V^{ss}\!/\!\!\sim$, where $v\sim w$ 

776: if and only if the closures $\overline{G_{\C}v}$

777: and $\overline{G_{\C}w}$ intersect in $V^{ss}$.

778: The {\em stable locus} $V^{st}$ with respect to $\chi$ is the set of points

779: $v\in V^{ss}$ such that the $G_{\C}$ orbit through $v$ is closed in

780: $V^{ss}$.  Clearly the geometric quotient $V^{st}\!/G_{\C}$ is an open set

781: inside of the categorical quotient $V\mod_{\!\!\chi} G_{\C}$.

782: The following theorem is due to Kirwan in the projective case \cite{Ki};

783: our formulation of it is taken from \cite[\S 3]{N4}

784: and \cite[\S 8]{MFK}.

785: 

786: \begin{theorem}\label{hkhs}

787: Let $G$ be a compact Lie group acting linearly on a complex vector

788: space $V$ with moment map $\mu:V\to\gd$.  Let $G_{\C}$ be the complexification

789: of $G$, with its induced action on $V$.  Let $\chi$ be a character of

790: $G$, and let $d\chi$ be the associated element of 5 center of $\gd$.

791: Then $v\in V^{ss}$ if and only if $G_{\C}v\cap\mu^{-1}(d\chi)\neq\emptyset$,

792: and the inclusion $\mu^{-1}(d\chi)\subs V^{ss}$ induces a homeomorphism

793: from $V\mod_{\! d\chi} G$ to $V\mod_{\!\chi} G_{\C}$.  

794: Furthermore, $d\chi$ is a regular

795: value of $\mu$ if and only if $V^{ss}=V^{st}$.

796: \end{theorem} 

797: 

798: Given a regular value $\a\in\gd$, Theorem \ref{hkhs} tells

799: us that we may interpret $V\mod_{\!\!\a} G$ as a GIT quotient

800: only in the case where $\a$ comes from a character of $G$.

801: We note, however, that the stability and semistability conditions 

802: are unchanged when $\chi$ is replaced by a high power of itself,

803: hence we may apply Theorem \ref{hkhs} whenever some multiple

804: of $\a$ comes from a character.  Furthermore, the GIT stability condition

805: is locally constant as a function of $\chi$.  Hence for any central 

806: regular value $\a\in\gd$,

807: we may perturb $\a$ to a ``rational'' point, thereby 

808: interpret $V\mod_{\!\!\a}G$ as a GIT quotient.

809: Accordingly, we will call an element of $V$ stable with respect

810: to $\a\in\gd$ if and only if it is stable with respect to $\chi$

811: for all $\chi$ such that $d\chi$ is close to a multiple of $\a$,

812: and we will write $V\mod_{\!\!\a}G_{\C} = V^{ss}\!/\!\!\sim$.

813: 

814: We may also use this theorem to formulate the hyperk\"ahler quotient

815: construction purely in terms of algebraic geometry.  

816: Theorem \ref{hkhs} says that, for $\a$ a regular value of $\mur$,

817: $$\cot\mod_{\!\!\a}G\,\cong\,\cot\mod_{\!\!\a}G_{\C}\,\cong\,\(\cot\)^{st}\!/G_{\C}.$$

818: Since $\muc:\cot\to\gd_{\C}$ is equivariant, we may take its vanishing locus

819: on both sides of the above equation, and we obtain the identity

820: $$\cot\mmod_{\!\!(\a,0)}G\,\cong\,\muc^{-1}(0)\mod_{\!\!\a}G_{\C}

821: \,\cong\,\muc^{-1}(0)^{st}/G_{\C}.$$

822: 

823: The following proposition is proven for the case

824: where $G$ is a torus in \cite[7.1]{BD}.

825:  

826: \begin{proposition}\label{denseopen}

827: Suppose that $\a$ and $(\a,0)$ are regular values for $\mu$ and $\mhk$,

828: respectively.

829: The cotangent bundle $T^*\X$ is isomorphic to an open subset of $\M$,

830: and is dense if it is nonempty.

831: \end{proposition}

832:  

833: \begin{proof}

834: Let $Y = \{(z,w)\in\mc^{-1}(0)^{st}\mid

835: z \in (\Cn)^{st}\}$,

836: where we ask $z$ to be semistable with respect to $\a$

837: for the action of

838: $G_{\C}$ on $\Cn$, so that $\X\cong (\Cn)^{st}/G_{\C}$.

839: Let $[z]$ denote the element of $\X$ represented by $z$.

840: The tangent space $T_{[z]}\X$ is equal to the quotient of $T_{z}\Cn$

841: by the tangent space to the $G_{\C}$ orbit through $z$,

842: hence

843: $$T^*_{[z]}\X\cong\{w\in T^*_{z}\Cn\mid w(\hat{v}_z)=0\text{ for all }

844: v\in\g_{\C}\} = \{w\in(\Cn)^*\mid\mc(z,w)=0\}.$$

845: Then $$T^*\X \cong \{(z,w)\mid z\in(\Cn)^{st}\text{ and }

846: \mc(z,w)=0\}/G_{\C} = Y/G_{\C}.$$

847: By the definition of semistability, $Y$ is an open subset of $\mc^{-1}(0)$,

848: and is dense if nonempty.  This completes the proof.

849: \end{proof}

850: 

851: \vspace{-\baselineskip}

852: \begin{remark}

853: We may significantly generalize the construction of hyperk\"ahler analogues

854: as follows.  Replace $\Cn$ by a smooth complex variety $X$, equipped with an

855: action of an algebraic group $G_{\C}$, an ample line bundle $L$, and a lift

856: of the action to $L$.  Then the cotangent bundle $T^*X$ is holomorphic symplectic,

857: and carries a natural holomorphic hamiltonian action of $G_{\C}$, along with a lift

858: of this action to the pullback of $L$.  We may then define the holomorphic symplectic

859: analogue of the GIT quotient $\X = X\mod G_{\C}$ to be the GIT quotient of the

860: zero level of the holomorphic moment map in $T^*X$, where the semistable

861: sets are defined by the action of $G_{\C}$ on $L$.  Theorem \ref{hkhs} tells us

862: that this agrees with our construction if $X=\Cn$, and Proposition \ref{denseopen}

863: generalizes to say that the holomorphic symplectic analogue of $\X$ is a partial

864: compactification of its cotangent bundle.

865: 

866: The reason for relegating this definition to a remark is that when $X$

867: is not equal to $\Cn$, its cotangent bundle $T^*X$ may not be the best holomorphic

868: symplectic manifold with which to replace it.  For example, if $X$ is itself a

869: K\"ahler quotient, then the holomorphic symplectic analogue of $X$ modulo the trivial

870: group, in the sense of the previous paragraph, would simply be the cotangent bundle

871: to $X$.  But this would (usually) not agree with the hyperk\"ahler

872: analogue of $X$.

873: \end{remark}

874: 

875: \end{section}

876:  

877: \begin{section}{The $\bf{\gm}$ action and the core}\label{circleaction}

878: Consider the action of $\gm$ on $T^*\Cn$ given by

879: scalar multiplication on the fibers, that is $\tau\cdot(z,w) = (z,\tau w)$.

880: The holomorphic moment map $\muc:\cot\to\gdc$ is $\gm$-equivariant

881: with respect to the scalar action on $\gdc$, 

882: hence $\gm$ acts on $\muc^{-1}(0)$.  Linearity of the action of $G$

883: on $\Cn$ implies that the actions of $\GC$ and $\gm$ on $\cot$ commute, 

884: therefore we obtain a $J_1$-holomorphic action of $\gm$ on 

885: $\Ma = \muc^{-1}(0)\mod\GC$.

886: Note that the $\gm$ action does {\em not} preserve the holomorphic symplectic form

887: or the hyperk\"ahler structure on $\Ma$;

888: rather we have $\t^*\omega_{\C}=\t\omega_{\C}$ for $\t\in\gm$.

889: 

890: If $\Ma$ is $\Q$-smooth, then the action of the compact subgroup

891: $\so\subs\gm$ is hamiltonian with respect to the real symplectic

892: structure $\omega_{\R}$, with moment map

893: $\Phi[z,w]_{\R} = \half |w|^2$.

894: This map is an orbifold Morse-Bott function\footnote{For a detailed discussion

895: of hamiltonian circle actions and Morse theory on orbifolds,

896: see \cite{LT}.} with image

897: contained in the non-negative real numbers, and

898: $\Phi^{-1}(0)=\Xa\subs\Ma$.

899: 

900: \begin{proposition}\label{proper}

901: If the original moment map $\mu:\Cn\to\gd$ is proper,

902: then so is $\Phi:\Ma\to\R$.

903: \end{proposition}

904: 

905: \begin{proof}

906: We must show that $\Phi^{-1}[0,R]$ is compact

907: for any $R\in\R$.  We have

908: $$\Phi^{-1}[0,R] = \{(z,w)\mid \mr(z,w)=\a,\hs

909: \mc(z,w)=0,\hs\Phi(z,w)\leq R\}\big/G$$ and $G$ is compact,

910: hence it is sufficient to show that

911: $\{(z,w)\mid \mr(z,w)=\a,\hs\Phi(z,w)\leq R\}$

912: is compact.  Since $\mr(z,w)=\mu(z)-\mu(w)$, this set is a closed subset

913: of $$\mu^{-1}\left\{\a+\mu(w) \,\Bigg{|}\, \half |w|^2\leq R\right\}

914: \times\left\{w \,\Bigg{|}\, \half |w|^2 \leq R\right\},$$

915: which is compact by the properness of $\mu$.

916: \end{proof}

917: 

918: \vspace{-\baselineskip}

919: \begin{remark}\label{tc}

920: In the case where $G$ is abelian and $\Xa$ is a nonempty toric

921: variety, properness of $\mu$ (and therefore of $\Phi$) is equivalent

922: to compactness of $\Xa$.

923: \end{remark}

924: % \end{section}

925: 

926: % \begin{section}{The core}

927: Suppose that $\Ma$ is $\Q$-smooth and $\Phi$ is proper.

928: We define the {\em core} $\core\subs\Ma$ to be the union

929: of those $\gm$ orbits whose closures are compact.

930: Properness of $\Phi$ implies that $\Ld_{\t\to 0}\t\cdot p$ exists for all $p\in \M$,

931: hence we may write 

932: $$\core = \{p\in\Ma\mid\Ld_{\t\to\infty}\t\cdot p\text{  exists}\}.$$

933: For $F$ a connected component of $\Mso=\Macs$, 

934: let $U_F$ be the closure of the set

935: of points $p\in \M$ such that $\Ld_{\t\to\infty}\t\cdot p\in F$.

936: In Morse-theoretic language, 

937: $U(F)$ is the closure of the unstable orbifold of the critical set $F$.

938: We may then write $\core$ as a finite union of irreducible, compact

939: varieties as follows: $$\core = \bigcup_{F\subs\Macs}U_F.$$

940: 

941: \begin{proposition}\label{coreprops}

942: The core of $\Ma$ has the following properties:

943: \begin{enumerate}

944: \item $\core$ is an $\so$-equivariant deformation retract of $\Ma$

945: \item $U_F$ is isotropic with respect to $\omega_{\C}$

946: \item If $\Ma$ is smooth at $F$, then $\dim U_F = \half\dim\Ma$.

947: \end{enumerate}

948: \end{proposition}

949: 

950: \begin{proof}

951: Let $f:\Ma\to [0,1]$ be a smooth, $\so$-invariant function with $f^{-1}(0)=\core$.

952: For all $p\in \M$ and $t\geq 0$, let $\rho_t(p) = e^{f(p)t}\cdot p$.

953: This defines a smooth family of $\so$-equivariant maps $\rho_t:\Ma\to\Ma$, 

954: fixing $\core$, with $\rho_0 = \id$.

955: The limit $\rho_{\infty} = \Ld_{t\to\infty}\rho_t$ is a well-defined

956: smooth map from $\M$ to $\core$, hence (1) is proved.

957: 

958: Suppose that $\Ma$ is smooth at $F$ and consider a point $p\in F$.  

959: Since $p$ is a fixed point, $\so$ acts on $T_p\Ma$, 

960: and we may write $$T_p\Ma = \Od_{s\in\Z}H_s,$$

961: where $H_s$ is the $s$ weight space for the circle action.

962: Since $\t^*\omega_{\C}=\t\omega_{\C}$ and $\omega_{\C}$ is a nondegenerate

963: bilinear form on $T_p\Ma$, $\omega_{\C}$ restricts to a perfect pairing

964: $H_s\times H_{1-s}\to\C$ for all $s\in\Z$.

965: In particular, $$T_pU_F = \Od_{s\leq 0}H_s$$ is a maximal isotropic subspace

966: of $T_p\Ma$, thus proving (2) and (3).

967: 

968: Now suppose that $\Ma$ is only $\Q$-smooth at $F$, and let $\Gap$ be the

969: orbifold group at $p$.  A circle action on the orbifold tangent

970: space $T_p\M$ is an action of a group $\Gapt$,

971: where $\Gapt$ is an extension of $\so$ by $\Gap$.  

972: Let $\Gapto$ be the connected

973: component of the identity in $\Gapt$.  Then $\Gapto\,$ is itself isomorphic

974: to a circle, and maps to the original circle $\so$ with some degree $d\geq 1$.

975: We now decompose $T_p\Ma = \bigoplus H_s$ into $\Gapto$ weight spaces.

976: Again $\omega_{\C}$ is nondegenerate on $T_p\Ma$, but now

977: $\hat{\t}^*\omega_{\C} = \hat{\t}^d\omega_{\C}$ 

978: for $\hat{\t}\in\Gapto\cong\so$, hence $\omega_{\C}$ restricts to a perfect

979: pairing $H_s\times H_{d-s}\to\C$ for all $s\in\Z$.

980: It follows that $T_pU_F = \bigoplus_{s\leq 0}H_s$ is isotropic (though not necessarily

981: maximally isotropic\footnote{See Example \ref{orbi}.})

982: with respect to $\omega_{\C}$.  This completes the proof

983: of (2) in the orbifold case.

984: \end{proof}

985: 

986: \vspace{-\baselineskip}

987: \begin{remark}\label{cpt}

988: Proposition \ref{coreprops} provides a new way to understand Proposition \ref{denseopen}

989: in the case where $\M$ is $\Q$-smooth and $\Phi$ is proper.

990: The K\"ahler quotient $\Xa$ is an $\omega_{\C}$-lagrangian

991: suborbifold of $\Ma$, hence $\omega_{\C}$ identifies the normal

992: bundle to $\Xa$ in $\Ma$ with the cotangent bundle of $\Xa$.

993: The Proposition \ref{denseopen} follows from the fact that the normal bundle to

994: $\Xa$ in $\Ma$ can be identified with the dense open set of points in $\Ma$

995: that flow down to $\Xa=\Phi^{-1}(0)$ along the gradient of $\Phi$.

996: This also demonstrates that the $\gm$ action on $\M$ restricts to scalar multiplication

997: on the fibers of $T^*\Xa$.

998: \end{remark}

999: 

1000: Given a space $M$ equipped with the action of a group $G$, we say that $M$

1001: is {\em equivariantly formal} if the equivariant cohomology ring

1002: $H^*_G(M)$ is a free module over $H^*_G(pt)$.

1003: We end the section with the statement of a fundamental theorem

1004: which we will use repeatedly throughout the paper.  Theorem \ref{formality}

1005: is proven in the compact case in \cite{Ki}, and the proof goes through

1006: in the noncompact case as long as $\Phi$ is proper 

1007: (see, for example, \cite[\S 2.2]{H1} or \cite[4.2]{TW}).

1008: 

1009: \begin{proposition}\label{formality}

1010: Let $M$ be a symplectic

1011: orbifold with a hamiltonian circle action

1012: such that the moment map $\Phi:M\to\R$ is proper and bounded below,

1013: and has finitely many critical values.

1014: Then $H^*_{\so}(M)$ is a free module over $H^*_{\so}(pt)$.

1015: Moreover, if the action of $\so$ commutes with the action of another

1016: torus $T$, then $H^*_{T\times\so}(M)$ is a free module

1017: over $H^*_{\so}(pt)$.

1018: \end{proposition}

1019: \end{section}

1020: \end{chapter}

1021: 

1022: \begin{chapter}{Hypertoric varieties}\label{ht}

1023: In this chapter we give a detailed analysis of the construction described

1024: in Chapter \ref{analogues} in the special case where the group $G$ is abelian.

1025: In this case the K\"ahler quotient $\Xa$ is called a {\em toric variety}, and

1026: its hyperk\"ahler analogue $\Ma$ is called a {\em hypertoric variety}.\footnote{Also

1027: known as a {\em toric hyperk\"ahler variety}.}  

1028: These latter spaces were first studied

1029: systematically in \cite{BD}; other references include \cite{K1}, \cite{HS}, and 

1030: \cite{HP1}.  

1031: 

1032: Just as there is a strong relationship between the geometry of toric

1033: varieties and the combinatorics of polytopes (see, for example, \cite{St}), 

1034: the geometry of hypertoric

1035: varieties interacts with the combinatorics of real hyperplane arrangements.

1036: Hausel and Sturmfels \cite{HS} gave an interpretation of the cohomology

1037: ring of a hypertoric variety as the Stanley-Reisner ring of the matroid

1038: associated to the corresponding arrangement of hyperplanes (Theorem \ref{htm}).  

1039: In Section \ref{htcoh}

1040: we interpret the $\so$-equivariant cohomology ring as an invariant of the oriented

1041: matroid, a richer combinatorial structure that can be associated to a hyperplane

1042: arrangement that is defined over the real numbers (Theorem \ref{htsm}

1043: and Remark \ref{om}).

1044: This result is applied in Section \ref{ossection} to obtain a combinatorial

1045: presentation of the $\Zt$-equivariant cohomology ring of the complement

1046: of an arrangment of complex hyperplanes defined over $\R$, thus enhancing

1047: the classical result of Orlik and Solomon \cite{OS}.

1048: In Section \ref{cogs}, we use the cogenerator approach of \cite{HS}

1049: to explore an aspect of the relationship between the cohomology rings of toric and

1050: hypertoric varieties.

1051: Most of the material presented here, with the exception of 

1052: the entirety of Section \ref{cogs}, has been taken from \cite{HP1} in a modified form.

1053: 

1054: \begin{section}{Hypertoric varieties and hyperplane arrangements}\label{arrangements}

1055: Let $\tn$ and $\td$ be real vector spaces of dimensions $n$ and $d$, respectively,

1056: with integer lattices $\tnz\subs\tn$ and $\tdz\subs\td$.

1057: Let $\{x_1,\ldots,x_n\}$ be an integer basis for $\tnz$,

1058: and let $\{\bd_1,\ldots,\bd_n\}$ be the dual basis for the dual lattice

1059: $\tndz\subs\tnd$.

1060: Suppose given $n$ nonzero integer vectors

1061: $\{\a_1,\ldots,\a_n\}\subs\tdz$ that span $\td$ over the real numbers.\footnote{In 

1062: each of \cite{BD,K1,HS,HP1}, some additional assumption is placed on the vectors

1063: $\{a_i\}$.  Sometimes they are assumed to be primitive, and sometimes 

1064: they are assumed to generate the lattice $\tdz$ over the integers.  Here

1065: we make neither assumption.}

1066: Define $\pi:\tn\to\td$ by $\pi(x_i)=a_i$, and let $\tk$ be the kernel of $\pi$,

1067: so that we have an exact sequence

1068: $$0 \longrightarrow \tk \stackrel{\i}{\longrightarrow} \tn

1069: \stackrel{\pi}{\longrightarrow} \td\longrightarrow 0.$$

1070: % Because $\{a_i\}$ spans $\tdz$, this sequence 

1071: % exponentiates to an exact sequence of compact tori

1072: This sequence exponentiates to an exact sequence of abelian groups

1073: $$0 \longrightarrow \Tk \stackrel{\i}{\longrightarrow} \Tn

1074: \stackrel{\pi}{\longrightarrow} \Td\longrightarrow 0,$$

1075: where $$\Tn=\tn/\tnz,\hs\hs\Td=\td/\tdz,\hs\hs\text{and }\Tk=\ker(\pi:\Tn\to\Td).$$

1076: Thus $\Tk$ is a compact abelian group with Lie algebra $\tk$,

1077: which is connected if and only if the vectors $\{a_i\}$ span the lattice

1078: $\tdz$ over the integers.

1079: It is clear that every closed subgroup of $\Tn$ arises in this way.

1080: 

1081: The restriction to $\Tk$ of the standard action of $\Tn$ on $\cot$

1082: is hyperhamiltonian with hyperk\"ahler moment map

1083: $$\mr\oplus\mc:\cot\to\tkd\oplus(\tk_{\C})^*,$$ where

1084: $$\mur(z,w) = \i^* \left(\half \sum_{i=1}^n (|z_i|^2 - |w_i|^2) \bd_i

1085: \right) \hspace{15pt}

1086: \text{and}\hspace{15pt}\muc(z,w) = \i^*_{\C} \left(\sum_{i=1}^n (z_i w_i)\bd_i\right).$$

1087: Given an element $\a\in\tkd$ with lift $r = (r_1,\ldots,r_n)\in\tnd$,

1088: the K\"ahler quotient

1089: $$\Xa= \Cn\mod\Tk = \mu^{-1}(\a)/\Tk$$

1090: is called a {\em toric variety}, and its hyperk\"ahler

1091: analogue $$\Ma=\cot\mmod\Tk = 

1092: \Big(\mur^{-1}(\a)\cap\muc^{-1}(0)\Big)\Big/\Tk$$

1093: is called a {\em hypertoric variety}.

1094: Both of these spaces admit an effective residual action of the torus $\Td=\Tn/\Tk$

1095: which is hamiltonian in the case of $\Xa$, and hyperhamiltonian in the case

1096: of $\Ma$, with hyperk\"ahler moment map

1097: \begin{eqnarray*}

1098: \barmur[z,w]_{\R}\oplus\barmuc[z,w]_{\R}

1099: &=& \half \sum_{i=1}^n (|z_i|^2 - |w_i|^2 - r_i)\,\bd_i 

1100: \hs\oplus\hs\sum_{i=1}^n (z_i w_i)\,\bd_i\\

1101: &\in&\ker(\i^*)\oplus\ker(\i^*_{\C}) = \tdd\oplus(\td_{\C})^*.

1102: \end{eqnarray*}

1103: In fact, this property may be used to give intrinsic definitions

1104: of toric and hypertoric varieties in certain categories, 

1105: as demonstrated by the following two theorems.

1106: % \footnote{

1107: % Theorem \ref{bielawski} is proven only in the smooth case, but presumably

1108: % a similar result holds for orbifolds.}

1109: 

1110: \begin{theorem}{\em\cite{De,LT}}

1111: Any connected symplectic orbifold of real dimension $2d$ which admits

1112: an effective, hamiltonian $\Td$ action with proper moment map is 

1113: $\Td$-equivariantly symplectomorphic to a toric variety.

1114: \end{theorem}

1115: 

1116: \begin{theorem}\label{bielawski}{\em\cite{Bi}}

1117: Any complete, connected, hyperk\"ahler manifold of real dimension $4d$

1118: which admits an effective, hyperhamiltonian $\Td$ action is $\Td$-equivariantly

1119: diffeomorphic, and Taub-NUT deformation equivalent, to a hypertoric variety.

1120: \end{theorem}

1121: 

1122: The data that were used to construct $\Xa$ and $\Ma$ consist of a collection of

1123: nonzero vectors $a_i\in\tdz$ and an element $\a\in\tkd$.  It is convenient

1124: to encode in terms of an arrangement of affine hyperplanes in $\tdd$ with some additional

1125: structure.

1126: A {\em rational, weighted, cooriented, affine hyperplane} 

1127: $H\subs\tdd$ is an affine hyperplane

1128: along with a choice of nonzero normal vector $a\in\tdz$.  The word rational

1129: refers to integrality of $a$, and weighted means that $a$ is not required

1130: to be primitive. 

1131: % Let $r = (r_1,\ldots,r_n)\in\tnd$ be a lift of $\a$ along $\i^*$, and

1132: Consider the rational, weighted, cooriented hyperplane

1133: $$H_i =\{v\in\tdd \mid v\cdot a_i + r_i = 0\}$$

1134: with normal vector $a_i\in\tdz$, along with the two half-spaces

1135: \begin{equation}\label{fg}

1136: F_i = \{v\in\tdd \mid v\cdot a_i + r_i \geq 0\}\hspace{10pt}\text{and}

1137: \hspace{10pt}G_i = \{v\in\tdd \mid v\cdot a_i + r_i \leq 0\}.

1138: \end{equation}

1139: Let $$\D = \bigcap_{i=1}^n F_i=\{v \mid v\cdot a_i+r_i\geq 0

1140: \text{  for all  }i\leq n\}$$

1141: be the (possibly empty) weighted polyhedron in $\tdd$ defined by the 

1142: weighted, cooriented, 

1143: affine, hyperplane arrangement $\A = \{H_1,\ldots,H_n\}$.  

1144: % The condition that $\{a_1,\ldots,a_n\}$

1145: % spans $\tdz$ over the natural numbers implies that $\D$ is bounded

1146: % for every value of $r\in\tnd$, which in turn

1147: % implies compactness of $\Xa$.

1148: Choosing a different lift $r'$ of $\a$ corresponds combinatorially to translating

1149: $\A$ inside of $\tdd$, and geometrically to shifting the K\"ahler and hyperk\"ahler 

1150: moment maps for the residual $\Td$ actions by $r'-r \in\ker\i^*=\tdd$.

1151: Our picture-drawing convention will be to encode the coorientations of the hyperplanes

1152: by shading $\D$, as in Figure \ref{teepee}.  In every example that we consider,

1153: all hyperplanes will have weight $1$; in other words we 

1154: will choose the primitive integer

1155: normal vector inducing the indicated coorientation.

1156: 

1157: \begin{figure}[h]

1158: \centerline{\epsfig{figure=htv2.eps, height=4cm}}

1159: \caption{A cooriented arrangement representing a toric variety of complex dimension $2$,

1160: or a hypertoric variety of complex dimension $4$, 

1161: obtained from an action of $T^2$ on $\C^4$.}

1162: \label{teepee}

1163: \end{figure}

1164: 

1165: We call the arrangement $\A$ {\em simple} if every subset of $m$ hyperplanes

1166: with nonempty intersection intersects in codimension $m$.

1167: We call $\A$ {\em smooth} if every collection of $d$ linearly independent vectors

1168: $\{a_{i_1},\ldots,a_{i_d}\}$ spans $\tdd$.

1169: An element $r\in\tnd$ or $\a\in\tkd$ will be called simple if 

1170: the corresponding arrangement $\A$ is simple.

1171: 

1172: \begin{theorem}\label{simple}{\em\cite[3.2,3.3]{BD}}

1173: The hypertoric variety $\Ma$ is $\Q$-smooth if and only if $\A$ is simple,

1174: and smooth if and only if $\A$ is smooth.

1175: \end{theorem}

1176: 

1177: % indexed by subsets $A\subs\{1,\ldots,n\}$.

1178: % Let $$\bmr\oplus\bmc:\Ma\to\tdd\oplus(\td_{\C})^*$$ be the hyperk\"ahler moment map

1179: % for the action of $\Td$ on $\Ma$.

1180: % For each $A\subs\{1,\ldots,n\}$, let $$M_A = \bmr^{-1}(\Delta_A)\cap\bmc^{-1}(0).$$

1181: % The K\"ahler submanifold

1182: % $\left(M_A,\omega_{\R}|_{M_A}\right)$ of $\left(M,\omega_{\R}\right)$

1183: % is $d$-dimensional and invariant under the action of $\Td$,

1184: % and is therefore $\Td$-equivariantly isomorphic

1185: % to the K\"ahler toric variety determined by $\Delta_A$ \cite[6.5]{BD}.

1186: % We define the {\em core} $C$ and {\em extended core} $D$

1187: % of a hypertoric variety by setting $$C = \!\!\bigcup_{\Delta_A\text{ bounded}}M_A\hspace{17pt}\text{ and }

1188: % \hspace{17pt} D = \bigcup_{\text{all }A}{M_A} = \mc^{-1}(0) =

1189: % \{[z,w]\mid z_iw_i=0\text{  for all  }i\}.$$

1190: % Bielawski and Dancer \cite{BD} show that

1191: % $C$ and $D$ are each $\Td$-equivariant

1192: % deformation retracts of $M$.  See Corollary \ref{unstable} for a Morse theoretic proof.

1193: 

1194: Let us pause to point out the different ways in which 

1195: $\Xa$ and $\Ma$ depend on the

1196: arrangement $\A$.  The toric variety $\Xa$ is in fact determined by the weighted

1197: polyhedron $\D$ \cite{LT},

1198: and is therefore oblivious to any hyperplane $H_i$ such that $\D$ is contained in the interior of $F_i$.

1199: Thus the toric variety corresponding to Figure \ref{teepee} is 

1200: $\C P^2$, the toric variety associated to a triangle.

1201: This is not the case for $\Ma$; we will see, in fact, that the

1202: hypertoric variety of Figure \ref{teepee} is topologically distinct

1203: from the one that we would obtain by deleting the third hyperplane.

1204: For this reason, it is slightly

1205: misleading to call $\Ma$ the hyperk\"ahler analogue of $\Xa$;

1206: more precisely, it is the hyperk\"ahler analogue of

1207: {\em a given presentation} of $\Xa$ as a K\"ahler quotient of a 

1208: complex vector space.

1209: 

1210: Just as the toric variety $\Xa$ fails to retain all of the data of

1211: the arrangement $\A$, there is some data that goes unnoticed by the

1212: hypertoric variety $\Ma$, as evidenced by the two following results.

1213: 

1214: \begin{lemma}\label{placement}

1215: % If $\a, \a'\in\tdd$ are both simple, then

1216: % $M_{\a}$ and $M_{\a'}$ are $\Td$-equivariantly diffeomorphic, and their

1217: % cohomology rings can be naturally identified.

1218: The hypertoric variety $\Ma$ is independent, up to $\Td$-equivariant

1219: diffeomorphism,\footnote{Bielawski and Dancer \cite{BD} 

1220: prove a weaker version of this statement,

1221: involving the (nonequivariant) homeomorphism type of $\Ma$.} 

1222: of the choice of a simple element $\a\in\tkd$.

1223: \end{lemma}

1224: 

1225: \begin{lemma}\label{coorientation}

1226: The hypertoric variety $\Ma$ is independent, up to $\Td$-equivariant isometry,

1227: of the coorientations of the hyperplanes $\{H_i\}$.

1228: \end{lemma}

1229: 

1230: \begin{proofp}

1231: The set of nonregular values for $\mur\oplus\muc$

1232: has codimension 3 inside of 

1233: $\tkd \oplus (\tk_{\C})^*$ \cite{BD}, hence we may

1234: choose a path connecting the two regular values $(\alpha,0)$

1235: and $(\alpha',0)$ for any simple $\a,\a'\in\tkd$, 

1236: and this path is unique up to homotopy.

1237: Since the moment map $\mur\oplus\muc$

1238: is not proper, we must take some care in showing that

1239: two fibers are diffeomorphic. To this end, we note that the

1240: norm-square function $\psi(z,w) = \|z\|^2 + \|w\|^2$ is $T^n$-invariant

1241: and proper on $\cot$.

1242: Let $(\cot)_{reg}$ denote the open submanifold of $\cot$ consisting of the

1243: preimages of the regular values of $\mur\oplus\muc$.

1244: By a direct computation, it is easy

1245: to see that the kernels of $d\psi$ and $d\mur\oplus d\muc$ intersect

1246: transversely at any point $p \in (\cot)_{reg}$.  Using the

1247: $T^n$-invariant hyperk\"ahler metric on $\cot$, 

1248: we define an Ehresmann connection

1249: on $(\cot)_{reg}$ with respect to $\mur\oplus\muc$

1250: such that the horizontal subspaces are contained in the kernel of $d\psi$.

1251: %Note that this connection can be extended to an Ehresmann

1252: %connection with respect to $\iota^{*} \bar{\mu}_{\R} \oplus

1253: %\iota_{\C}^{*} \bar{\mu}_{\C} \oplus \psi.$ 

1254: 

1255: This connection allows us to lift a path connecting the two

1256: regular values to a horizontal vector field on its preimage in

1257: $(\cot)_{reg}$. Since the horizontal subspaces are tangent to

1258: the kernel of $d\psi$, the flow preserves level sets of $\psi$. Note that the

1259: function $$\mur\oplus\muc\oplus\psi: \cot \to \tkd \oplus (\tk_{\C})^{*} \oplus\R$$

1260: {\em is} proper. By a theorem of Ehresmann \cite[8.12]{BJ}, the

1261: properness of this map implies that the flow of this vector field exists for all time, and

1262: identifies the inverse image of $(\alpha,0)$ with that of

1263: $(\alpha',0)$. Since the metric, $\psi$, and $\mur\oplus\muc$ are

1264: all $T^n$-invariant, the Ehresmann connection is also $T^n$-invariant,

1265: therefore the diffeomorphism identifying the fibers

1266: is $T^n$-equivariant, and

1267: the reduced spaces are $T^d$-equivariantly diffeomorphic.

1268: \end{proofp}

1269: 

1270: \vspace{-\baselineskip}

1271: \begin{proofc}

1272: It suffices to consider the case when we change the orientation of a

1273: single hyperplane within the arrangement.  Changing the coorientation

1274: of a hyperplane $H_m$ is equivalent to defining a new map

1275: $\pi':\tn\to\td$, with $\pi'(x_i)=a_i$ for $i\neq m$, 

1276: and $\pi'(x_m) = -a_m$.

1277: This map exponentiates to a map $\pi':\Tn\to\Td$, with $\ker(\pi')$

1278: conjugate to $\ker(\pi)$ inside of $\operatorname{GL}_n(\H)$ 

1279: (the group of quaternion-linear

1280: automorphisms of $\cot\cong\Hn$) by the element

1281: $(1,\ldots,1,j,1,\ldots,1)\in \operatorname{GL}_1(\H)^n\subs \operatorname{GL}_n(\H)$,

1282: where the $j$ appears in the $m^{\text{th}}$ slot.

1283: Hence the hyperk\"ahler quotient by $\ker(\pi')$ is isomorphic

1284: to the hyperk\"ahler quotient by $\ker(\pi)$.

1285: \end{proofc}

1286: 

1287: \vspace{-\baselineskip}

1288: \begin{example}\label{same}

1289: The three cooriented arrangements of Figure 2 all specify

1290: the same hyperk\"ahler variety $\M$ up to equivariant

1291: diffeomorphism.  The first has $\X\cong\widetilde{\C P^2}$ 

1292: (the blow-up of $\C P^2$ at a point),

1293: and the second and the third have $\X\cong\C P^2$.

1294: Note that if we reversed the coorientation of $H_3$ in Figure 2(a)

1295: or 2(c), then we would get a noncompact

1296: $\X\cong\widetilde{\C^2}$.  If we reversed the coorientation of $H_3$ in

1297: Figure 2(b), then $\X$ would be empty, 

1298: but the topology of $\Ma$ would not change.

1299: 

1300: \begin{figure}[h]

1301: \centerline{\epsfig{figure=htv1.eps}}

1302: \caption{Three arrangements

1303: related by reversing coorientations and translating hyperplanes.}

1304: \label{triplet}

1305: \end{figure}

1306: \end{example}

1307: 

1308: Our purpose is to study not just the geometry

1309: of $\Ma$, but the geometry of $\Ma$ along with the hamiltonian circle action

1310: defined in Section \ref{circleaction}.

1311: In order to define this action, we used the fact that we were reducing at a regular value

1312: of the form $(\a,0)\in\tdd\oplus(\td_{\C})^*$.

1313: Although the set of regular values of $\mur\oplus\muc$ is simply connected,

1314: the set of regular values of the form $(\a,0)$ is disconnected, therefore the

1315: diffeomorphism of Lemma \ref{placement} is not circle-equivariant.

1316: Furthermore, left multiplication by the diagonal matrix 

1317: $(1,\ldots,1,j,1,\ldots,1) \in 

1318: \operatorname{GL}_n(\H)$ is not an $\so$-equivariant 

1319: automorphism of $\cot\cong\Hn$, 

1320: therefore the topological type of $\Ma$ along with a $\so$ action 

1321: may depend nontrivially

1322: on both the locations and the coorientations of the hyperplanes in $\A$.

1323: Indeed it must, because we can recover $\Xa$ from $\Ma$ by taking the minimum

1324: $\Phi^{-1}(0)$ of the moment map $\Phi:\Ma\to\R$, and we know that $\Xa$ depends in an essential way

1325: on the combinatorial type of the polyhedron $\D$.

1326: In this sense, the structure of a hypertoric variety $\Ma$

1327: along with a hamiltonian circle action may be regarded as 

1328: the universal geometric object associated to $\A$ from which

1329: both $\Ma$ and $\Xa$ can be recovered.

1330: \end{section}

1331: 

1332: \begin{section}{Geometry of the core}\label{htcore}

1333: In this section we give a combinatorial description of the fixed point set 

1334: $\M^{\gm} = \Ma^{\so}$

1335: and the core $\core$ of a $\Q$-smooth hypertoric variety $\Ma$.  

1336: We will assume that $\Phi$ is proper.  (If $\D$ is nonempty, this is equivalent to asking

1337: that $\D$ be bounded, or that $\Xa$ be compact.)

1338: % The parameter $\a$ will be fixed throughout

1339: % that section, hence we will omit the subscript.

1340: First, we note that the holomorphic moment map $\barmuc:\Ma\to(\td_{\C})^*$

1341: is $\gm$-equivariant with respect to the scalar action on $(\td_{\C})^*$,

1342: hence both $\Ma^{\gm}$ and $\core$ will be contained in

1343: $$\E=\barmuc^{-1}(0)=\Big\{[z,w]\in 

1344: \Ma\bigmid z_iw_i=0\text{ for all } i\Big\},$$ 

1345: which we call the {\em extended core} of $\Ma$.

1346: It is clear from the defining equations that the restriction of

1347: $\barmur$ from $\E$ to $\tdd$ is surjective.

1348: The extended core naturally breaks into components

1349: $$\EA = \Big\{[z,w]\in \Ma\bigmid w_i=0

1350: \text{ for all $i\in A$ and $z_i=0$ for all $i\in A^c$}\Big\},$$

1351: indexed by subsets $A\subs\otn$.

1352: When $A=\emptyset$, $\EA = \X \subs \Ma$.  More generally,

1353: the variety $\EA\subs \Ma$ is a $d$-dimensional K\"ahler subvariety 

1354: of $\Ma$ with an effective hamiltonian $\Td$-action, and is therefore itself a toric variety.

1355: (It is the K\"ahler quotient by $\Tk$ of an $n$-dimensional 

1356: coordinate subspace of $\cot$,

1357: contained in $\muc^{-1}(0)$.)

1358: The hyperplanes $\{H_i\}$ divide $\tdd$ into a union of closed,

1359: (possibly empty) convex polyhedra $$\Delta_A = 

1360: \bigcap_{i\in A}F_i\,\,\cap\,\,\bigcap_{i\in A^c}G_i.$$

1361: 

1362: \begin{lemma}\label{sides}

1363: If $w_i=0$, then $\barmur[z,w]_{\R}\in F_i$.

1364: If $z_i=0$, then $\barmur[z,w]_{\R}\in G_i$.

1365: \end{lemma}

1366: 

1367: \begin{proof}

1368: We have

1369: $$\barmur[z,w]_{\R}\cdot a_i + r_i = \mur(z,w)\cdot x_i

1370: =\half\Big(|z_i|^2 - |w_i|^2\Big),$$

1371: hence the statement follows from Equation~\eqref{fg}.

1372: \end{proof}

1373: 

1374: \vspace{-\baselineskip}

1375: \begin{lemma}\label{htcorecomp}{\em\cite{BD}}

1376: The core component $\EA$ is isomorphic to the toric variety 

1377: corresponding to the weighted polytope $\D_A$.

1378: \end{lemma}

1379: 

1380: \begin{proof}

1381: Lemma \ref{sides} tells us that $\barmur(\EA)\subs\Delta_A$,

1382: and surjectivity of $\barmur|_{\E}$ says that this inclusion is an equality.

1383: The lemma then follows from the classification theorems of \cite{De,LT}.

1384: \end{proof}

1385: 

1386: Although $\gm$ does not act on $\Ma$ as a subtorus of $\Td_{\C}$, we show

1387: below that when restricted to any single

1388: component $\EA$ of the extended core, $\gm$

1389: {\em does} act as a subtorus of $\Td_{\C}$, with the subtorus depending

1390: combinatorially on $A$.  This will allow us to give a combinatorial

1391: analysis of the gradient flow of $\Phi$ on the extended core.

1392: 

1393: Suppose that $[z,w]\in\EA$.  Then for $\tau\in\gm$,

1394: $$\tau[z,w]=[z,\tau w] = 

1395: [\tau_1 z_1,\ldots \tau_n z_n, \tau_1^{-1}w_1,\ldots \tau_n^{-1}w_n], \text{ where }

1396: \tau_i =

1397: \begin{cases}

1398: \tau^{-1} & \text{if }i\in A,\\

1399: 1 & \text{if }i\notin A.

1400: \end{cases}

1401: $$

1402: In other words, the $\gm$ action on $\EA$

1403: is given by the one dimensional subtorus

1404: $(\tau_1, \ldots, \tau_n)$

1405: of the original torus $T^n_{\gm}$,

1406: hence the moment map $\Phi|_{\EA}$ for the action of $\so\subs\gm$ is given by

1407: $$\Phi[z,w] = \left<\mr[z,w],\,\,\sum_{i\in A}a_i\right>.$$

1408: This formula allows us to compute the fixed points of the circle action.

1409: % Since $S^1$ acts freely on $(\td_{\C})^{*} \backslash \{0\}$

1410: % and $\mc: \M \to (\td_{\C})^{*}$ is $S^1$-equivariant,

1411: % we must have $M^{S^1} \subseteq \mc^{-1}(0) = D$.

1412: For any subset $B\subs\{1,\ldots,n\}$, let $\EA^B$ be the toric subvariety

1413: of $\EA$ defined by the conditions $z_i=w_i=0$ for all $i\in B$.

1414: Geometrically, $\EA^B$ is defined by the (possibly empty) intersection of the hyperplanes

1415: $\{H_i\mid i\in B\}$ with $\Delta_A$.

1416: % $\displaystyle\bigcap_{i\in B}H_i \cap \Delta_A$ of the polyhedron $\Delta_A$.

1417: 

1418: \begin{proposition}

1419: The fixed point set of the action of $S^1$ on $\EA$

1420: is the union of those toric subvarieties $\EA^B$ such that

1421: $\sum_{i\in A}a_i\in \td_B:=\operatorname{Span}_{j\in B}a_j$.

1422: \end{proposition}

1423: 

1424: \begin{proof}

1425: The moment map $\Phi|_{\EA^B}$ will be constant if and only if

1426: $\sum_{i\in A}a_i$ is perpendicular to

1427: the face $\Phi(\EA^B)$,

1428: i.e. if $\sum_{i\in A}a_i$ lies in the kernel of the projection

1429: $\td\onto\td/\td_B$.

1430: \end{proof}

1431: 

1432: \vspace{-\baselineskip}

1433: \begin{corollary}

1434: Every vertex $v\in\tdd$ of the polyhedral complex $|\A|$ given by our arrangement

1435: is the image of an $S^1$-fixed point in $\M$.

1436: Every component of $\M^{S^1}$ maps to a face of $|\A|$.

1437: % has dimension less than or equal to $d$,

1438: % and the only component of dimension $d$ is $\E_{\emptyset}=X=\Phi^{-1}(0)$.

1439: \end{corollary}

1440: 

1441: \begin{proposition}

1442: The core $\core$ of $\Ma$ is equal to the union of those

1443: subvarieties $\EA$ such that $\Delta_A$ is bounded.

1444: \end{proposition}

1445: 

1446: \begin{proof}

1447: Because $\gm$ acts on $\EA$ as a subtorus of the complex torus $\Td_{\C}$,

1448: the set $$\{p\in\EA\mid\Ld_{\t\to\infty}\t\cdot p\text{ exists}\}$$

1449: is a (possibly reducible) toric subvariety of $\EA$, i.e. a union of subvarieties

1450: of the form $\EA^B$.  Fix a subset $B\subs\otn$.

1451: The variety $\EA^B$ is stable under the $\gm$ action, hence

1452: if $\EA^B$ is compact, then $\EA^B\subs\core$.  On the other hand if $\EA^B$ is noncompact,

1453: then properness of $\Phi$ precludes it from being part of the core,

1454: hence $$\core = \{p\in\E\mid\Phi(p)\text{ lies on a bounded face of }|\A|\}.$$

1455: By \cite[6.7]{HS}, the bounded complex of $|\A|$ has pure dimension $d$,

1456: and is therefore equal to the union of those polyhedra $\Delta_A$ that are bounded.

1457: \end{proof}

1458: 

1459: % For any point $p\in \M^{S^1}$, the {\em stable orbifold}

1460: % $S(p)$ at $p$ is defined to be the set of $x\in \M$ such

1461: % that $x$ approaches $p$ when flowing along the vector field

1462: % $-\operatorname{grad}(\Phi)$,

1463: % and the {\em unstable orbifold} $U(p)$ at $p$ is defined to be the stable orbifold with respect

1464: % to the function $-\Phi$.

1465: % For any suborbifold $Y\subs \M^{S^1}$, the unstable orbifold $U(Y)$ at $Y$ is defined

1466: % to be the union of $U(y)$ for all $y\in Y$.  In general, for $y\in Y$, we have the identity

1467: % $\dim_{\R} U(Y) + \dim_{\R} S(y) = 4d$.

1468: % 

1469: % Let $Y\subs \M^{S^1}$ be a component of the fixed point set of $M$.

1470: % Let $v\in\tdd$ be a vertex in the polytope $\mr(Y)$, and let $y$

1471: % be the unique preimage of $v$ in $Y$.

1472: % 

1473: % \begin{proposition}

1474: % The unstable orbifold $U(Y)$ is a complex suborbifold of complex

1475: % dimension at most $d$, contained in the core $C\subs \M$.

1476: % If $\A$ is smooth at $y$, then $\dim_{\C}U(Y)=d$, and the closure of $U(Y)$

1477: % is an irreducible component of $C$.

1478: % \end{proposition}

1479: % 

1480: % \begin{proof}

1481: % For simplicity, we will assume that $v=\cap_{j=1}^dH_j$.

1482: % For all $l\in\otd$, let $b_l\in\td_{\Z}$ be the smallest integer vector

1483: % such that $\left<a_j,b_l\>=0$ for $j\neq l$ and $\left<a_l,b_l\> >0$.

1484: % Geometrically, $b_l$ is the primitive integer vector on the line

1485: % $\cap_{j\neq l}H_j$ pointing in the direction of $\Delta$.

1486: % Note that $M$ is smooth at the $T^d$-fixed point above $v$

1487: % if and only if $\left<a_l,b_l\>=1$ for all $l\in\otd$.

1488: % Let $R_l\subs\tdd$ be the ray emanating from $v$ in the direction of $b_l$,

1489: % and ending before it hits another vertex.  Let $Q_l$ be the analogous ray in the opposite

1490: % direction.

1491: % 

1492: % Let $\Delta_A$ be a region (not necessarily bounded) of the polyhedral

1493: % complex defined by $\A$ adjacent to $R_l$.

1494: % The preimage $\mr^{-1}(R_l)\cap D$ of $R_l$ in $D$ is a complex line,

1495: % and it is contained in the unstable

1496: % orbifold at $U(Y)$ if and only if $\left<b_l,\sum_{i\in A}a_i\right>\geq 0$.

1497: % If $\left<b_l,\sum_{i\in A}a_i\right> <0$, it is contained in the stable orbifold $S(y)$.

1498: % The preimage $\mr^{-1}(Q_l)\cap D$ of $Q_l$ in $D$ is also a complex line,

1499: % contained in the unstable

1500: % orbifold $U(Y)$ if and only if $\<-b_l,a_l+\sum_{i\in A}a_i\>\geq 0$,

1501: % and otherwise in $S(y)$.

1502: % Since $\<b_l,\sum_{i\in A}a_i\>+\<-b_l,a_l+\sum_{i\in A}a_i\>

1503: % =-\left<a_l,b_l\> <0$, at most one of these two directions can be unstable.

1504: % In the smooth case, $\left<a_l,b_l\>=1$ for all $l$, and exactly one of

1505: % the two directions is unstable.

1506: % 

1507: % Consider the polytope $\Delta_v$ incident to $v$ and characterized by the property that its

1508: % edges at the vertex $v$ are exactly the unstable directions.

1509: % The toric variety $X_{\Delta_v}\subs D$ is contained

1510: % in the closure of $U(Y)$, and a dimension count tells us that this containment is an equality.

1511: % In the smooth case, $\Delta_v$ is $d$-dimensional, and $X_{\Delta_v}$ is a component of the core.

1512: % \end{proof}

1513: % 

1514: % Note that, even in the smooth case, it is not necessarily the case that the

1515: % $R_l$ direction is stable and the $Q_l$ direction is unstable.  See, for example,

1516: % the vertex $v = H_1\cap H_2$ in Figure 2(c).

1517: 

1518: \vspace{-\baselineskip}

1519: \begin{corollary}\label{bij}

1520: There is a natural injection from the set of bounded regions

1521: $\{\Delta_A\mid A\in I\}$

1522: to the set of connected components of $\M^{\gm}$.

1523: If $\A$ is smooth, this map is a bijection.

1524: \end{corollary}

1525: 

1526: \begin{proof}

1527: To each $A\in I$, we associate the fixed subvariety $\EA^B$ corresponding

1528: to the face of $\Delta_A$ on which the linear functional $\sum_{i\in A}a_i$

1529: is minimized, so that $\EA=U(\EA^B)$.

1530: Proposition \ref{coreprops}~(3) tells us that if $\A$ is smooth and 

1531: $F$ is a component

1532: of $\Ma^{\gm}$, then we will have $U(F)=\EA$ for some $A\subs\otn$.

1533: \end{proof}

1534: 

1535: \vspace{-\baselineskip}

1536: \begin{example}

1537: In Figure 3, representing a reduction of $T^*\C^5$ by $T^3$,

1538: we choose a metric on $(\mathfrak{t}^2)^*$ in order

1539: to draw the linear functional $\sum_{i\in A}a_i$ as a vector

1540: in each region $\Delta_A$.  We see that $\M^{S^1}$ has three components,

1541: one of them $X\cong\widetilde{\C P}^2$, one of them a projective line with another 

1542: $\widetilde{\C P}^2$ as its associated core component, and one of them a point with 

1543: core component $\C P^2$.

1544: 

1545: \begin{figure}[h]

1546: \centerline{\epsfig{figure=flow.eps}}

1547: \caption{The gradient flow of $\Phi:\Ma\to\R$.}

1548: \end{figure}

1549: \end{example}

1550: 

1551: \begin{example}\label{orbi}

1552: The hypertoric variety represented by Figure 4 has a fixed point set with four

1553: connected components (three points and a $\C P^2$), but only three components in its core.

1554: This phenomenon can be blamed on the orbifold point $p$ represented by the

1555: intersection of $H_3$ and $H_4$, 

1556: which has only a one-dimensional unstable orbifold (to its northwest).

1557: In other words, this example illustrates the necessity of the smoothness

1558: assumption to obtain a bijection in Corollary \ref{bij}.

1559: 

1560: \begin{figure}[h]

1561: \centerline{\epsfig{figure=orbi.eps}}

1562: \caption{A singular example.}

1563: \end{figure}

1564: \end{example}

1565: \end{section}

1566: 

1567: \begin{section}{Cohomology rings}\label{htcoh}

1568: In this section we compute the $\so$ and $\Tdso$-equivariant cohomology

1569: rings of of a $\Q$-smooth hypertoric variety $\Ma$, 

1570: extending the computations of the ordinary and $\Td$-equivariant

1571: rings given in \cite{K1} and \cite{HS}.

1572: For the sake of simplicity, and with an eye toward the applications in Chapter

1573: \ref{abelianization}, we will restrict our attention

1574: to the case where $\Phi$ is proper (see Remark \ref{tc}).

1575: This assumption will be necessary for the application of

1576: Proposition \ref{formality} and the proof of Theorem \ref{tara}.

1577: 

1578: \begin{remark}\label{qorz}

1579: Because we wish to treat the smooth and $\Q$-smooth cases simultaneously,

1580: we will work with cohomology over the rational numbers.  We note, however,

1581: that Konno proves Theorem \ref{htm} over the integers in the smooth case,

1582: and therefore our Theorem \ref{htsm} holds over the integers when $\A$ is smooth.

1583: This fact will be significant in Section \ref{ossection}, when we will need to reduce our

1584: coefficients modulo 2.

1585: \end{remark}

1586: 

1587: Consider the hyperk\"ahler Kirwan maps

1588: $$\ktd:H^*_{\Tn}(\cot)\to\htm\hspace{15pt}\text{and}

1589: \hspace{15pt}\k:H^*_{\Tk}(\cot)\to H^*(\Ma)$$

1590: induced by the $\Tn$-equivariant inclusion of $\mur^{-1}(\a)\cap\muc^{-1}(0)$

1591: into $\cot$.  Because the vector space $\cot$ is 

1592: $\Tn$-equivariantly contractible, we have

1593: $$H^*_{\Tn}(\cot)=\Sym\tnd\cong\Q[\bd_1,\ldots,\bd_n]$$ and

1594: $$H^*_{\Tk}(\cot)=\Sym\tkd\cong\Q[\bd_1,\ldots,\bd_n]/\ker(\i^*).$$

1595: 

1596: \begin{theorem}\label{htm}\label{hm}{\em\cite{K1,HS}}

1597: The Kirwan maps $\ktd$ and $\k$ are surjective, and the kernels of both are

1598: generated by the elements

1599: $$\prod_{i\in S}\bd_i\hspace{15pt}\text{  for all $S\subs\otn$ such that  }

1600: \bigcap_{i\in S} H_i = \emptyset.$$

1601: \end{theorem}

1602: 

1603: \begin{remark}

1604: The kernel of $\ktd$ is precisely

1605: the Stanley-Reisner ideal of

1606: the matroid of linear dependencies among the vectors $\{a_i\}$ \cite{HS}

1607: (see Remark \ref{om}).

1608: \end{remark}

1609: 

1610: The inclusion of $\mur^{-1}(\a)\cap\muc^{-1}(0)$

1611: into $\cot$ is also $\so$-equivariant, hence we may consider the 

1612: analogous circle-equivariant Kirwan maps

1613: $$\ktdso:H^*_{\Tn\times\so}(\cot)\to\htsm\hspace{15pt}\text{and}

1614: \hspace{15pt}\k:H^*_{\Tk\times\so}(\cot)\to H^*_{\so}(\Ma),$$

1615: where $$H^*_{\Tn\times\so}(\cot)\cong\Q[\bd_1,\ldots,\bd_n,x]$$ and

1616: $$H^*_{\Tk}(\cot)\cong\Q[\bd_1,\ldots,\bd_n,x]/\ker(\i^*).$$

1617: The remainder of this section will be devoted to proving the following theorem.

1618: 

1619: \begin{theorem}\label{htsm}\label{hsm}

1620: The circle-equivariant Kirwan maps $\ktdso$ and $\kso$ are surjective,

1621: and the kernels of both are generated by the elements

1622: \begin{eqnarray*}

1623: \prod_{i\in S_1}\bd_i\times\prod_{j\in S_2}(x-\bd_j)\hspace{15pt}

1624: &\text{for}&\!\!\text{all disjoint pairs $S_1,S_2\subs\otn$}\\

1625: &&\text{such that }

1626: \bigcap_{i\in S_1} G_i\,\,\cap\,\,\bigcap_{j\in S_2} F_j = \emptyset.

1627: \end{eqnarray*}

1628: \end{theorem}

1629: 

1630: \begin{remark}\label{om}

1631: For all $i\in\otn$, let $b_i = a_i\oplus 0\in\td\oplus\R$, 

1632: and let $b_0=0\oplus 1$.  The

1633: {\em pointed matroid} associated to $\A$ is a combinatorial object that 

1634: tells us which subsets of $\{b_0,\ldots,b_n\}$ are linearly dependent.

1635: By simplicity of $\A$, this is equivalent to knowing which subsets

1636: $S\subs\otn$ have the property that $$\bigcap_{i\in S} H_i=\emptyset,$$ 

1637: which is in turn

1638: equivalent to knowing the dependence relations among the vectors $\{a_i\}$.

1639: In particular, it does not depend on the relative positions of the hyperplanes,

1640: encoded by the parameter $\a\in\tkd$.

1641: 

1642: The {\em pointed oriented matroid} associated to $\A$ encodes the data

1643: of which subsets of $\{\pm b_0,\ldots,\pm b_n\}$ are linearly dependent

1644: over the {\em positive} real numbers.  This is equivalent to knowing

1645: which pairs of subsets $S_1,S_2\subs\otn$ have the property that

1646: $$\bigcap_{i\in S_1} G_i\,\,\cap\,\,\bigcap_{j\in S_2} F_j = \emptyset,$$

1647: which does indeed depend on $\a$.

1648: Hence Theorem \ref{htm} shows that $\htm$ is an invariant

1649: of the pointed matroid of $\A$, and Theorem \ref{htsm} demonstrates that $\htsm$

1650: is an invariant of the pointed oriented matroid of $\A$.  

1651: For more on this perspective, see \cite{H3} and \cite{Pr}.

1652: \end{remark}

1653: 

1654: % In order to make use of Theorem \ref{htm}, we must apply

1655: % the principle of {\em equivariant formality},

1656: % proven for compact manifolds in \cite{Ki}, which we adapt to our situation

1657: % as follows.

1658: % 

1659: % \begin{proposition}\label{formality}

1660: % Let $M$ be a symplectic

1661: % orbifold, possibly noncompact but of finite topological type.

1662: % Suppose that $M$ admits a hamiltonian action of a torus $T\times S^1$,

1663: % such that the fixed point set $M^{\so}$ has an even-dimensional,

1664: % $T$-equivariant cell decomposition,

1665: % and the $S^1$-component $\Phi:M\to\R$ of the moment map

1666: % is proper and bounded below.  Then

1667: % $H^*_{T\times S^1}(M)$ is a free module over $H^*_{T\times\so}(pt)$.

1668: % \end{proposition}

1669: % 

1670: % \begin{proof}

1671: % Because $\Phi$ is a moment map, it is a Morse-Bott function

1672: % such that all of the critical suborbifolds and their normal bundles

1673: % carry almost complex structures.  

1674: % Thus we get a Morse-Bott stratification

1675: % of $M$ into even-dimensional $T\times\so$-invariant suborbifolds.

1676: % This tells us, as in \cite[5.8]{Ki},

1677: % that the spectral sequence associated to the fibration

1678: % $M\hookto EG\times_G \M\to BG$ collapses, where $G=T\times\so$,

1679: % and we get the desired result.

1680: % \end{proof}

1681: Consider the following commuting square of maps, where

1682: $\phi$ and $\psi$ are each given by setting the image of 

1683: $x\in\ H^*_{\so}(pt)\cong\Q[x]$ to zero.

1684: 

1685: $$\begin{CD}

1686: H^*_{\Tn\times S^1}(\cot) @>\ktdso >> \htsm\\

1687: @V\phi VV @ VV\psi V\\

1688: H^*_{\Tn}(\cot) @>\ktd >> \htm\\

1689: \end{CD}$$

1690: 

1691: \begin{lemma}\label{surjective}

1692: The equivariant Kirwan map $\ktdso$ is surjective.

1693: \end{lemma}

1694: 

1695: \begin{proof}

1696: Suppose that $\gamma\in\htsm$ is a homogeneous class of minimal degree

1697: that is {\em not} in the image of $\ktdso$.  By Theorem \ref{htm} $\ktd$ is surjective, 

1698: hence we may choose a class $\eta\in\phi^{-1}\ktd^{-1}\psi(\gamma)$.

1699: Theorem \ref{formality} tells us that the kernel of $\psi$ is generated by $x$,

1700: hence $\ktdso(\eta)-\gamma = x\delta$ for some $\delta\in\htsm$.

1701: Then $\delta$ is a class of lower degree that is not in the image of $\ktdso$.

1702: \end{proof}

1703: 

1704: \vspace{-\baselineskip}

1705: \begin{lemma}\label{enough}

1706: If $\mathcal{I}\subs\ker\ktdso$ and $\phi(\mathcal{I}) = \ker\ktd$,

1707: then $\mathcal{I}=\ker\ktdso$.

1708: \end{lemma}

1709: 

1710: \begin{proof}

1711: Suppose that $a\in\ker\ktdso\smallsetminus\mathcal{I}$ 

1712: is a homogeneous class of minimal degree,

1713: and choose $b\in\mathcal{I}$ such that $\phi(a-b)=0$.  Then $a-b = cx$ for some

1714: $c\in H^*_{\Tn\times S^1}(\cot)$.  By Proposition \ref{formality},

1715: $cx\in\ker\ktdso\impl c\in\ker\ktdso$, hence $c\in\ker\ktdso\smallsetminus\mathcal{I}$

1716: is a class of lower degree than $a$.

1717: \end{proof}

1718: 

1719: \vspace{-\baselineskip}

1720: \begin{proofhtsm}

1721: For any element $$h\in H^2_{T^n\times S^1}(\cot)\cong\Q\{\bd_1,\ldots,\bd_n,x\},$$

1722: let $\tilde{L}_h = \cot\times\C_{h}$

1723: be the $\Tn\times S^1$-equivariant line bundle on $\cot$

1724: with equivariant Euler class $h$.

1725: Let $$L_h = \tilde{L}_h\Big{|}_{\mur^{-1}(\a)\cap\muc^{-1}(0)}\Big/\Tk$$

1726: be the quotient $\Td\times S^1$-equivariant line bundle on $\Ma$.

1727: We will write $\tilde L_i = \tilde L_{\bd_i}$ and

1728: $\tilde K = \tilde L_x$,

1729: with quotients $L_i$ and $K$.

1730: Since the $\Tdso$-equivariant Euler class $e(L_i)$

1731: is the image of $\bd_i$ under the hyperk\"ahler Kirwan map

1732: $H^*_{T^n\times S^1}(\cot)\to \htsm$, we will abuse notation and

1733: denote it by $\bd_i$.  Similarly, we will denote $e(K)$ by $x$.

1734: Lemma \ref{surjective} tells us that

1735: $\htsm$ is generated by $\bd_1,\ldots,\bd_n,x$.

1736: 

1737: Consider the $\Tn\times S^1$-equivariant section $\tilde s_i$ of $\tilde{L}_i$

1738: given by the function $\tilde s_i(z,w)=z_i$.

1739: This descends to a $\Tdso$-equivariant section $s_i$ of $L_i$

1740: with zero-set $$Z_i

1741: := \{[z,w]\in \Ma\mid z_i=0\}.$$

1742: Similarly, the function $\tilde t_i(z,w)=w_i$ defines a $\Tdso$-equivariant

1743: section of $L_i^*\otimes K$ with zero set $$W_i := \{[z,w]\in \Ma\mid w_i=0\}.$$

1744: Thus the divisor $Z_i$ represents the cohomology class $\bd_i$,

1745: and $W_i$ represents $x-\bd_i$.

1746: Note, that by Lemma \ref{sides},

1747: we have $\mr(Z_i) \subs G_i$ and $\mr(W_i) \subs F_i$ for all $i\in\otn$.

1748: 

1749: Let $S_1$ and $S_2$ be a pair of subsets of $\{1,\ldots n\}$ such that

1750: $\big(\cap_{i\in S_1}G_i\big)\cap

1751: \big(\cap_{j\in S_2}F_j\big) = \emptyset,$ and hence

1752: $$\big(\cap_{i\in S_1}Z_i\big)\cap

1753: \big(\cap_{j\in S_2}W_j\big)\subs\mr^{-1}\bigg(

1754: \big(\cap_{i\in S_1}G_i\big)\cap

1755: \big(\cap_{j\in S_2}F_j\big)\bigg) = \emptyset.$$

1756: Now consider the vector bundle

1757: $E=\Od_{i\in S_1}L_i\,\,\oplus\,\,\Od_{j\in S_2}L_j^*\otimes K$

1758: with equivariant Euler class 

1759: $$e(E)=\prod_{i\in S_1}\bd_i\times\prod_{j\in S_2}(x-\bd_j).$$

1760: The section 

1761: $\left(\oplus_{i\in S_1}s_i\right)\oplus\left(\oplus_{i\in S_2}t_i\right)$

1762: is a nonvanishing equivariant

1763: global section of $E$,

1764: hence $e(E)$ is trivial in $\htsm$.

1765: Theorem \ref{htm} and Lemma \ref{enough} 

1766: tell us that we have found all of the relations.

1767: The proofs of the analogous statements for $\hsm$ are identical.

1768: \end{proofhtsm}

1769: 

1770: How sensitive are the invariants $\htsm$ and $\hsm$?

1771: We can recover $\htm$ and $\hma$

1772: by setting $x$ to zero, hence they are at least as fine

1773: as the ordinary or $\Td$-equivariant cohomology rings.

1774: The ring $\htsm$ does {\em not} depend on coorientations,

1775: for if $\M'$ is related to $\M$ by flipping the coorientation

1776: of the $l^{\text{th}}$ hyperplane $H_k$, then the map taking 

1777: % $\bd_i$ to $\bd_i$ for $i\neq l$ and 

1778: $\bd_l$ to $x-\bd_l$ is an isomorphism

1779: between $\htsm$ and $H^*_{\Td\times S^1}(M')$.\footnote{The 

1780: oriented matroid of a collection of nonzero vectors

1781: in a real vector space does not change when one of the vectors is negated, hence

1782: the independence of $\htsm$ on coorientations can be deduced from Remark \ref{om}.}

1783: The ring {\em does}, however, depend on $\a$, as we see in Example \ref{ts}.

1784: 

1785: \begin{example}\label{ts}

1786: We compute the equivariant cohomology ring $\htsm$

1787: for the hypertoric varieties $\M_a$, $\M_b$, and $\M_c$

1788: defined by the arrangements in Figure~\ref{triplet}(a), (b), and (c), 

1789: respectively.

1790: Note that each of these arrangements is smooth, hence Theorem \ref{htsm}

1791: holds over the integers, as in Remark \ref{qorz}.

1792: $$H^*_{\Td\times S^1}(\M_a)=\Z[\bd_1,\ldots,\bd_4,x]\big/

1793: \< \bd_2\bd_3, \bd_1(x-\bd_2)\bd_4, \bd_1\bd_3\bd_4 \>,$$

1794: $$H^*_{\Td\times S^1}(\M_b)=\Z[\bd_1,\ldots,\bd_4,x]\big/

1795: \< (x-\bd_2)\bd_3, \bd_1\bd_2\bd_4, \bd_1\bd_3\bd_4 \>,$$

1796: $$H^*_{\Td\times S^1}(\M_c)=\Z[\bd_1,\ldots,\bd_4,x]\big/

1797: \< \bd_2\bd_3, (x-\bd_1)\bd_2(x-\bd_4), \bd_1\bd_3\bd_4 \>.$$

1798: As we have already observed, the rings 

1799: $H^*_{\Td\times S^1}(\M_a)$ and $H^*_{\Td\times S^1}(\M_b)$

1800: are isomorphic by interchanging $\bd_2$ with $x-\bd_2$.

1801: % We now show that $H^*_{\Td\times S^1}(\M_a)$ is not isomorphic

1802: % to $H^*_{\Td\times S^1}(\M_c)$.  

1803: One can check that the annihilator of $\bd_2$ in $H^*_{\Td\times S^1}(\M_a)$

1804: is the principal ideal generated by $\bd_3$, while the ring $H^*_{\Td\times S^1}(\M_c)$

1805: has no degree $2$ element whose annihilator is generated by a single element of degree $2$.

1806: Hence $H^*_{\Td\times S^1}(\M_c)$ is not isomorphic to the other two rings.

1807: % Since $(x-\bd_2)(x-\bd_3)$ is the unique

1808: % degree $4$ relation in both rings, any isomorphism 

1809: % between the two rings

1810: % would have to have to take $x-\bd_2$ and $x-\bd_3$ 

1811: % to scalar multiples of $x-\bd_2$ and $x-\bd_3$, possibly in the reverse order.

1812: % One can check that the only degree $4$ classes

1813: % in $H^*_{\Td\times S^1}(\M_a)$ that annihilate $x-\bd_2$ are multiples of $x-\bd_3$,

1814: % whereas in $H^*_{\Td\times S^1}(\M_c)$, $x-\bd_2$ is killed by $\bd_1\bd_4$

1815: % and $x-\bd_3$ is killed by $(x-\bd_1)(x-\bd_4)$.

1816: % Hence $x-\bd_2$ can be taken neither to a multiple of $x-\bd_2$ nor of $x-\bd_3$,

1817: % therefore the rings cannot be isomorphic.

1818: \end{example}

1819: 

1820: The ring $\hsm$, on the other hand, is sensitive to coorientations

1821: as well as the parameter $\a$, as we see in Example \ref{s}.

1822: 

1823: \begin{example}\label{s}

1824: We now compute the ring $\hsm$ for

1825: $\M_a$, $\M_b$, and $\M_c$ of Figure 2.

1826: Theorem \ref{hsm} tells us that we need only to quotient the ring

1827: $\htsm$ by $\ker(\iota^*)$.

1828: For $\M_a$, the kernel of $\iota_a^*$ is generated by

1829: $\bd_1+\bd_2-\bd_3$ and $\bd_1-\bd_4$, hence we have

1830: \begin{eqnarray*}

1831: H^*_{S^1}(\M_a)&=&\Z[\bd_2,\bd_3,x]\big/

1832: \< \bd_2\bd_3, (\bd_3-\bd_2)^2(x-\bd_2), (\bd_3-\bd_2)^2\bd_3 \>\\

1833: &\cong & \Z[\bd_2,\bd_3,x]\big/\< \bd_2\bd_3, (\bd_3-\bd_2)^2(x-\bd_2), \bd_3^3 \>.\\

1834: \end{eqnarray*}

1835: Since the hyperplanes of 2(c) have the same coorientations as those

1836: of 2(a), we have $\ker\iota_b^*=\ker\iota_a^*$, hence

1837: \begin{eqnarray*}

1838: H^*_{S^1}(\M_c) &=& \Z[\bd_2,\bd_3,x]\big/

1839: \< \bd_2\bd_3, (x-\bd_3+\bd_2)^2\bd_2, (\bd_3-\bd_2)^2\bd_3 \>\\

1840: &\cong & \Z[\bd_2,\bd_3,x]\big/\< \bd_2\bd_3, (x-\bd_3+\bd_2)^2\bd_2, \bd_3^3 \>.\\

1841: \end{eqnarray*}

1842: Finally, since Figure 2(b) is obtained from 2(a) by flipping the coorientation

1843: of $H_2$, we find that $\ker(\iota_b^*)$ is generated by $\bd_1-\bd_2-\bd_3$

1844: and $\bd_1-\bd_4$, therefore

1845: $$H^*_{S^1}(\M_b)=\Z[\bd_2,\bd_3,x]\big/

1846: \< (x-\bd_2)\bd_3, (\bd_2+\bd_3)^2\bd_2, (\bd_2+\bd_3)^2\bd_3 \>.$$

1847: As in Example \ref{ts}, $H^*_{S^1}(\M_a)$ and $H^*_{S^1}(\M_c)$

1848: can be distinguished by the fact that the annihilator

1849: of $\bd_2\in H^*_{S^1}(\M_a)$ is generated by a single element of degree $2$,

1850: and no element of $H^*_{S^1}(\M_c)$ has this property.

1851: On the other hand, $H^*_{S^1}(\M_b)$ is distinguished from

1852: $H^*_{S^1}(\M_a)$ and $H^*_{S^1}(\M_c)$ by the fact that neither $x-\bd_2$ nor $\bd_3$ cubes to zero.

1853: % Each of three rings have two linear elements, unique up to scale, that multiply to zero.

1854: % In $H^*_{S^1}(\M_c)$, both of them cube to zero, while

1855: % in $H^*_{S^1}(\M_a)$ and $H^*_{S^1}(\M_b)$, only $(x-\bd_3)$ cubes to zero.

1856: % By changing coordinates, we can write

1857: % $$H^*_{S^1}(\M_a) = \Z[x,y,z]\big/\<yz, (z-y+x)^2(x-y), z^3\>$$ and

1858: % $$H^*_{S^1}(\M_b) = \Z[x,y,z]\big/\<yz, (z-y)^2(x-y), z^3\>.$$

1859: % Any isomorphism from $H^*_{S^1}(\M_a)$ to $H^*_{S^1}(\M_b)$ would have to take

1860: % $y$ and $z$ to multiples of themselves, and $x$ to an expression of the form

1861: % $\a x+\b y+ \gamma z$, with $\a\neq 0$.  Then a contradiction arises from the

1862: % fact that the first degree six relation in $H^*_{S^1}(\M_a)$ is cubic in $x$,

1863: % while $H^*_{S^1}(\M_b)$ has no such nontrivial relations.

1864: % Hence all three rings are nonisomorphic.

1865: \end{example}

1866: 

1867: % \begin{remark}\label{lawrence}

1868: % Theorems \ref{htsm} and \ref{hsm} can be interpreted

1869: % in light of the recent work of Hausel and Sturmfels \cite{HS}

1870: % on Lawrence toric varieties.

1871: % The Lawrence toric variety $N$ associated to the arrangement $\A$

1872: % is the K\"ahler reduction $\cot\mod\Tk$,

1873: % so that $M$ sits inside of $N$ as the complete

1874: % intersection cut out by the equation $\bar\mc(z,w)=0$.

1875: % The residual torus acting on $N$ has dimension $d+n$,

1876: % and includes the $(d+1)$-dimensional torus $\Td\times S^1$

1877: % acting on $M$, and

1878: % the inclusion of $M$ into

1879: % $N$ induces an isomorphism on $\Td\times S^1$-equivariant cohomology.

1880: % One can use geometric arguments similar to those that were applied to prove

1881: % Theorem \ref{htsm},

1882: % or the purely combinatorial approach of \cite{HS},

1883: % to show that

1884: % $$H_{T^{d+n}}^*(N)=\Q[\bd_1,\ldots,\bd_n,v_1,\ldots,v_n]\bigg/

1885: % \< \prod_{i\in S_1}\bd_i\times

1886: % \prod_{j\in S_2}v_j\hs\bigg{\vert}\hs\bigcap_{i\in S} H_i = \emptyset\>.$$

1887: % From here we can recover $\htsm = H_{\Td\times S^1}^*(N)$

1888: % by setting $\bd_i+v_i = \bd_j+v_j$ for all $i,j\leq n$.

1889: % Note that Hausel and Sturmfels' work applies to the general orbifold case.

1890: % \end{remark}

1891: \end{section}

1892: 

1893: \begin{section}{The equivariant Orlik-Solomon algebra}\label{ossection}

1894: In this section we restrict our attention to smooth arrangements.

1895: When $\A$ is smooth, all of the computations of Section \ref{htcoh}

1896: hold over the integers (see Remark \ref{qorz}).  Since the rings in question

1897: are torsion-free, the presentations are also valid when the coefficients

1898: are taken in the field field $\Zt$.

1899: 

1900: Let $\M_{\R} \subs \M$ be the {\em real locus}

1901: $\{[z,w]\in \M\mid z,w \text{ real}\}$ of $\M$ 

1902: with respect to the complex structure $J_1$.

1903: The full group $\Td\times S^1$ does not act on $\M_{\R} $,

1904: but the subgroup $\Tdr\times\Zt$ does act,

1905: where $\Tdr :=\Z_2^d\subs \Td$ is the fixed point set

1906: of the involution of $\Td$ given by complex

1907: conjugation.\footnote{It is interesting to note that

1908: the real locus with respect to the complex structure $J_1$

1909: is in fact a complex submanifold with respect to $J_3$,

1910: on which $\Tdr$ acts holomorphically and $\Zt$ acts anti-holomorphically.}

1911: % one of the other complex structures on $\M$.

1912: % The action of $\Tdr$ is holomorphic because $\Tdr$ is a subgroup of $\Td$,

1913: % which preserves all of the complex structures on $\M$.

1914: % The action of $\Zt$, on the other hand, is anti-holomorphic,

1915: % i.e. it can be thought of as complex conjugation.}

1916: In this section we will study the geometry of the real locus,

1917: focusing in particular on the properties of the $\Zt$ action.

1918: The following theorem is an extension of the results of \cite{BGH} and \cite{Sc}

1919: to the noncompact case of hypertoric varieties.

1920: 

1921: \begin{theorem}\label{tara}{\em\cite{HH}}

1922: Let $G = \Tdso$ or $\Td$,

1923: and $G_{\R} = \Tdr\times\Zt$ or $\Tdr$, respectively.

1924: Then we have $H^{*}_{G}(\M;\Zt)\cong

1925: H^{*}_{G_{\R}}(\M_{\R} ;\Zt)$

1926: by an isomorphism that halves the grading.\footnote{In particular,

1927: $H^{*}_{G}(\M;\Zt)$ is concentrated in even degree.}

1928: Furthermore, this isomorphism takes the cohomology classes represented

1929: by the $G$-stable submanifolds $Z_i$ and $W_i$ to those represented by 

1930: the $G_{\R}$-stable submanifolds $Z_i\cap \M_{\R}$ and $W_i\cap \M_{\R}$.

1931: \end{theorem}

1932: 

1933: % \begin{sketch}

1934: % Consider the injection $H^*_G(\M;\Zt)\hookto H^*_G(\M^G;\Zt)$

1935: % given by the inclusion of the fixed point set into $\M$.

1936: % % Let \(\M^{(1)}\) be the {\em one-skeleton} of $\M$, i.e. the set of 

1937: % % points in $\M$ whose $G$-orbits are 0- or 1-dimensional. 

1938: % % Let \(\M^{(0)}\) denote the fixed points. Let 

1939: % % \(j: \M^{(1)} \hookrightarrow \M\) and \(i: \M^{(0)} \hookrightarrow \M\) 

1940: % % denote the natural inclusions. 

1941: % The essential idea is to show that

1942: % % the image \(j^{*}(H^*_{G}(\M^{(1)})

1943: % % \subset H^{*}_{G}(\M^{(0)})\) is the same as the image 

1944: % a class in $H^*_G(\M^G;\Zt)$ extends over $\M$ if and only

1945: % if it extends to the set of points on which $G$ acts with

1946: % a stabilizer of codimension at most $1$,

1947: % and then to show that a similar

1948: % statement in $G_{\R}$-equivariant cohomology

1949: % also holds for the real locus $\M_{\R} $ with its $G_{\R}$

1950: % action. One then uses a canonical isomorphism $H^2_G(pt,\Z_2) \cong 

1951: % H^1_{G_{\R}}(pt,\Z_2)$ to give the result.

1952: % 

1953: % The key to the proof is a noncompact $G_{\R}$ version

1954: % of the proposition in

1955: % \cite{TW} stating that the $G_{\R}$-equivariant Euler class of

1956: % the negative normal bundle of a critical point $p$

1957: % is not a zero divisor, which can be shown explicitly

1958: % using a local normal form for the actions of $G$ and $G_{\R}$.

1959: % % Using a local normal form theorem for 

1960: % % Hamiltonian $G$-manifolds equipped with a compatible 

1961: % % antisymplectic involution $\sigma$ and properties of 

1962: % % the $G$ weights and $G_{\R}$ weights at the fixed points, 

1963: % % one can compute

1964: % % explicitly the $G_{\R}$ equivariant Euler class and verify 

1965: % %that it is not a zero divisor.  

1966: % The proposition then follows

1967: % from standard $G_{\R}$ versions of the Thom isomorphism theorem

1968: % with coefficients in $\Z_2$.

1969: % Since a component of the

1970: % moment map is proper, bounded below, and has finitely many

1971: % fixed points, one can then check that the inductive argument, given

1972: % in Section 3 of \cite{TW}

1973: % to complete the proof of \cite[Thm 1]{TW} also holds in this case.

1974: % \end{sketch}

1975: 

1976: Consider the restriction $\Psi$ of the hyperk\"ahler moment map

1977: $\mr\oplus\mc$ to $\M_{\R} $.  Since $z$ and $w$ are real

1978: for every $[z,w]\in \M_{\R} $, the map $\mc$

1979: takes values in $\td_{\R}\subs\td_{\C}$, which we will identify

1980: with $i\R^d$, so that $\Psi$ takes values

1981: in $\R^d\oplus i\R^d\cong\C^d$.

1982: Note that $\Psi$ is $\Zt$-equivariant, with $\Zt$ acting on

1983: $\Cn$ by complex conjugation.

1984: 

1985: \begin{lemma}

1986: The map $\Psi:{\M_{\R}}\to\C^d$ is surjective, 

1987: and the fibers are the orbits of $\Tdr$.

1988: The stabilizer of a point $p\in {\M_{\R}}$ has order

1989: $2^r$, where $r$ is the number of hyperplanes in the complexified

1990: arrangement $\A_{\C}$ containing the point $\Psi(p)$.

1991: \end{lemma}

1992: 

1993: \begin{proof}

1994: For any point $a+bi\in\C^d$, choose a point $[z,w]\in \M$ such that

1995: $\mr[z,w]=a$ and $\mc[z,w]=b$.  After moving $[z,w]$ by an element of $\Td$

1996: we may assume that $z$ and $w$ are real, hence we may assume

1997: that $[z,w]\in {\M_{\R}}$.

1998: Then $$\Psi^{-1}(a+bi)=\mur^{-1}(a)\cap\muc^{-1}(b)\cap {\M_{\R}}

1999: = \Td[z,w]\cap {\M_{\R}} = \Tdr[z,w].$$

2000: The second statement follows easily from \cite[3.1]{BD}.

2001: \end{proof}

2002: 

2003: Let $Y\subs {\M_{\R}}$ be the locus of points on which $\Tdr$

2004: acts freely, i.e. the preimage under $\Psi$ of the space

2005: $\MA:=\bomp$.

2006: The inclusion map $Y\hookto {\M_{\R}}$

2007: induces maps backward on cohomology, which we will denote

2008: $$\phi:\hr\to H^*_{\Tdr}(Y;\Zt)\hspace{15pt}\text{and}\hspace{15pt}

2009: \phi_2:\hrs\to H^*_{\Tdr\times\Zt}(Y;\Zt).$$

2010: By Theorem \ref{tara}, we have

2011: $$\hr\cong H^{*}_{\Td}(\M;\Zt)\hspace{15pt}\text{and}\hspace{15pt}

2012: \hrs\cong H^{*}_{\Tdso}(\M;\Zt).$$

2013: Furthermore, since $\Tdr$ acts freely on $Y$ with quotient $\MA$, we have

2014: $$H^*_{\Tdr}(Y;\Zt)\cong H^*(\MA;\Zt)\hspace{15pt}\text{and}\hspace{15pt}

2015: H^*_{\Tdr\times\Zt}(Y;\Zt)\cong H_{\Zt}^*(\MA;\Zt),$$ hence we may write

2016: $$\phi:H^{*}_{\Td}(\M;\Zt)\to H^*(\MA;\Zt)\hspace{15pt}\text{and}\hspace{15pt}

2017: \phi_2:H^{*}_{\Tdso}(\M;\Zt)\to H_{\Zt}^*(\MA;\Zt).$$

2018: The ring $H^*(\MA;\Zt)$ is a classical invariant of the arrangement $\A$ known as

2019: the {\em Orlik-Solomon algebra} (with coefficients in $\Zt$),

2020: and is denoted by $A(\A;\Zt)$ \cite{OT}.

2021: The ring $H^*_{\Zt}(\MA;\Zt)$ was introduced in \cite{HP1} and further studied

2022: in \cite{Pr}.  We call it the {\em equivariant Orlik-Solomon algebra}

2023: and denote it by $A_2(\MA;\Zt)$.

2024: 

2025: \begin{proposition}\label{ztf}{\em\cite[2.4]{Pr}}

2026: The space $\MA$ is $\Zt$-equivariantly formal, i.e. $A_2(\A;\Zt)$ is a free module

2027: over $\Zt[x]=\hz$.

2028: \end{proposition}

2029: 

2030: \begin{theorem}\label{surj}

2031: Both $\phi$ and $\phi_2$ are surjective, with kernels

2032: $$\ker\phi = \Big\la\bd_i^2\,\bigmid\, i\in\otn\Big\ra\hspace{10pt}\text{and}

2033: \hspace{10pt}\ker\phi_2 = \Big\la\bd_i\,(x-\bd_i)\,\bigmid\, i\in\otn\Big\ra.$$

2034: \end{theorem}

2035: 

2036: \begin{proof}

2037: Theorem \ref{tara} tells us that $\phi_2(\bd_i)$ is represented in 

2038: $H^*_{\Tdr\times\Zt}(Y;\Zt)$ by the submanifold $Z_i\cap Y$, and likewise

2039: $\phi_2(x-\bd_i)$ by the submanifold $W_i\cap Y$.  Since $\mur(Z_i\cap W_i)\subs H_i$,

2040: we have $Z_i\cap W_i\cap Y=\emptyset$, hence $\bd_i(x-\bd_i)$ lies in the kernel of

2041: $\phi_2$ (and therefore $\bd_i^2$ lies in the kernel of $\phi$).

2042: 

2043: By Proposition \ref{ztf} and a pair of formal arguments

2044: identical to those of Lemmas \ref{surjective} and \ref{enough}, it is sufficient

2045: to prove Theorem \ref{surj} only for $\phi$.  

2046: Quotienting $Z_i\cap Y$ by $\Tdr$, we find that $\phi(\bd_i)$ is represented in $A(\A;\Zt)$

2047: by the submanifold

2048: $$\{v\in\MA\mid v\cdot a_i + r_i \in\R^-\}.$$

2049: The standard presentation of $A(\A;\Zt)$ (see, for example, \cite{OT}) says that

2050: these classes generate the ring, and that all relations between them are generated

2051: by the monomials of Theorem \ref{hm} and $\bd_i^2$ for all $i$.

2052: \end{proof}

2053: 

2054: \vspace{-\baselineskip}

2055: \begin{remark}

2056: Theorems \ref{htsm} and \ref{surj} combine to give a presentation of the equivariant

2057: Orlik-Solomon algebra $A_2(\A;\Zt)$ in the case where $\A$ is rational, simple, and smooth.

2058: This presentation first appeared in \cite{HP1}.  In \cite{Pr}, we generalize this

2059: presentation to arbitrary real hyperplane arrangements, and in fact to arbitrary

2060: pointed oriented matroids.

2061: \end{remark}

2062: 

2063: \begin{remark}

2064: The ring $A_2(\A;\Zt)$ is a deformation over the affine line $\operatorname{Spec}\Zt[x]$

2065: from the ordinary Orlik-Solomon algebra $A(\A;\Zt)$ to the Varchenko-Gel$'$fand ring

2066: $VG(\A;\Zt)$ of locally constant $\Zt$-valued functions on the real points of $\MA$ \cite{Pr}.

2067: While the rings $A(\A;\Zt)$ and $VG(\A;\Zt)$ depend only on the matroid associated

2068: to $\A$, the deformation $A_2(\A;\Zt)$ depends on the richer structure

2069: of the pointed oriented matroid (see Remark \ref{om}).  

2070: \end{remark}

2071: 

2072: \begin{example}\label{first}

2073: Consider the arrangements $\A_a$ and $\A_c$ in Figure 2(a) and 2(c).

2074: By Theorem \ref{surj} and Example \ref{ts} we have

2075: $$H^*_{\Zt}(\mathcal{M}(\A_a);\Zt)\cong

2076: \Zt [\bd_1,\ldots,\bd_4,x]\bigg/

2077: \< \begin{array}{c}

2078: \bd_1(x-\bd_1), \bd_2(x-\bd_2), \bd_3(x-\bd_3), \bd_4(x-\bd_4),\\

2079: \bd_2\bd_3, \bd_1(x-\bd_2)\bd_4, \bd_1\bd_3\bd_4 

2080: \end{array}\>$$

2081: and

2082: $$H^*_{\Zt}(\mathcal{M}(\A_c);\Zt)\cong

2083: \Zt [\bd_1,\ldots,\bd_4,x]\bigg/

2084: \< \begin{array}{c}

2085: \bd_1(x-\bd_1), \bd_2(x-\bd_2), \bd_3(x-\bd_3), \bd_4(x-\bd_4),\\

2086: \bd_2\bd_3, (x-\bd_1)\bd_2(x-\bd_4), \bd_1\bd_3\bd_4

2087: \end{array}\>.$$

2088: The map $f:H^*_{\Zt}(\mathcal{M}(\A_a);\Zt)\to H^*_{\Zt}(\mathcal{M}(\A_b);\Zt)$

2089: given by

2090: $$f(\bd_1) = \bd_1+\bd_2, f(\bd_2) = \bd_2+\bd_3+x, f(\bd_3) = \bd_3,

2091: f(\bd_4) = \bd_2+\bd_4,\text{ and }f(x)=x$$

2092: is an isomorphism of graded $\Zt[x]$-algebras, despite the fact that the

2093: oriented matroids of the two arrangements differ.\footnote{We thank

2094: Graham Denham for finding this isomorphism.}

2095: \end{example}

2096: 

2097: \begin{example}\label{second}

2098: Now consider the arrangements $\A_a'$ and $\A_c'$ obtained from $\A_a$ and $\A_c$

2099: by adding a vertical line on the far left, as shown in Figure \ref{vertical}.

2100: \begin{figure}[h]

2101: \begin{center}

2102: \psfrag{Ha'}{$\A_a'$}

2103: \psfrag{Hc'}{$\A_c'$}

2104: \includegraphics{htv3.eps}

2105: \caption{Adding a vertical line to $\A_a$ and $\A_c$.}\label{vertical}

2106: \end{center}

2107: \end{figure}

2108: Again by Theorem \ref{surj}, we have

2109: $$H^*_{\Zt}(\mathcal{M}(\A_a');\Zt)\cong

2110: \Zt [\bd_1,\ldots,\bd_4,x]\bigg/

2111: \< \begin{array}{c}

2112: \bd_1(x-\bd_1), \bd_2(x-\bd_2), \bd_3(x-\bd_3), \bd_4(x-\bd_4),\\ \bd_5(x-\bd_5), 

2113: \bd_2\bd_3, (x-\bd_1)\bd_5, \bd_1(x-\bd_2)\bd_4,\\ \bd_1\bd_3\bd_4, (x-\bd_2)\bd_4\bd_5, \bd_3\bd_4\bd_5 

2114: \end{array}\>$$

2115: and

2116: $$H^*_{\Zt}(\mathcal{M}(\A_c');\Zt)\cong

2117: \Zt [\bd_1,\ldots,\bd_4,x]\bigg/

2118: \< \begin{array}{c}

2119: \bd_1(x-\bd_1), \bd_2(x-\bd_2), \bd_3(x-\bd_3), \bd_4(x-\bd_4),\\ \bd_5(x-\bd_5),

2120: \bd_2\bd_3, (x-\bd_1)\bd_5, (x-\bd_1)\bd_2(x-\bd_4),\\ \bd_1\bd_3\bd_4, (x-\bd_2)\bd_4\bd_5, \bd_3\bd_4\bd_5 

2121: \end{array}\>.$$

2122: We have used Macaulay 2 \cite{M2} to check that the annihilator of the element

2123: $\bd_2\in H^*_{\Zt}(\mathcal{M}(\A_a');\Zt)$ is generated by two linear elements

2124: (namely $\bd_3$ and $x-\bd_2$) and nothing else, while there is no element

2125: of $H^*_{\Zt}(\mathcal{M}(\A_c');\Zt)$ with this property.

2126: Hence the two rings are not isomorphic.

2127: \end{example}

2128: \end{section}

2129: 

2130: \begin{section}{Cogenerators}\label{cogs}

2131: Consider the K\"ahler Kirwan map

2132: $$\ka:\Sym\tkd\cong H^*_{\Tk}(\C^n)\to H^*(\Xa_{\a})$$ 

2133: induced by the $\Tk$-equivariant

2134: inclusion of $\mu^{-1}(\a)$ into $\Cn$.

2135: In this section we would like to consider simultaneously the Kirwan

2136: maps $\ka$ for many different values of $\a$, so almost all of the

2137: notation that we use will have a subscript or superscript indicating

2138: the parameter $\a\in\tkd$ or a lift $r\in\tnd$.

2139: An exception to this rule will be the 

2140: hyperk\"ahler Kirwan map $$\k:\Sym\tkd\to H^*(\Ma_{\a}),$$

2141: which, by Lemma \ref{placement} or Theorem \ref{hm}, 

2142: is independent of our choice of simple $\a\in\tkd$.

2143: The main result of this section is the following.

2144: 

2145: \begin{theorem}\label{int}

2146: The kernel of the hyperk\"ahler Kirwan map $\k$ is equal to the intersection over all simple

2147: $\a$ of the kernels of the K\"ahler Kirwan maps $\ka$.

2148: \end{theorem}

2149: 

2150: \begin{remark}

2151: Konno \cite[7.6]{K2} proves an analogous theorem about the kernels

2152: of the Kirwan maps to the cohomology rings of polygon and

2153: hyperpolygon spaces.  We may therefore conjecture a generalization

2154: of Theorem  \ref{int} in which $\Tk$ is replaced by an arbitrary compact

2155: group $G$.  Note that our proof of Theorem \ref{int}

2156: depends strongly on the combinatorics associated to hypertoric varieties.

2157: \end{remark}

2158: 

2159: We approach Theorem \ref{int} by describing the kernels of $\k$ and $\ka$ not

2160: in terms of generators, but rather in terms of cogenerators.

2161: Given an ideal $\mathcal{I}\subs\Sym\tkd$, a set of {\em cogenerators} for $\mathcal{I}$

2162: is a collection of polynomials $\{f_i\}\subs\Sym\tk$ such that

2163: $$\mathcal{I}=\{\bd\in\Sym\tkd\mid\bd\cdot f_i = 0\text{ for all }i\}.$$

2164: 

2165: The volume function $\Vol\Dr$ is locally polynomial in $r$.

2166: More precisely, for every simple $r\in\tnd$, there exists a degree $d$ polynomial

2167: $\Pr\in\Sym^d\tn$ such that for every simple $s\in\tnd$ lying in the

2168: same connected component of the set of simple elements as $r$,

2169: we have $$\Vol\D^{s}=\Pr\!(s).$$  We will refer to $\Pr$ as the

2170: {\em volume polynomial} of $\Dr$.  The fact that the volume of a polytope

2171: is translation invariant tells us that $\Pr$ lies in the image of the inclusion

2172: $\i:\Sym^d\tk\hookto\Sym^d\tn$.

2173: 

2174: \begin{theorem}\label{toric}{\em\cite{GS,KP}}

2175: Let $r\in\tnd$ be a simple element with $\i^*(r)=\a$.  Then

2176: % $$\ker (\ka\circ\i^*) = \Ann\{\Pr\} = \big\{\bd\in\Sym\tnd \mid \bd\cdot\Pr = 0\big\}$$ and 

2177: $$\ker\ka = \Ann\{\i^{-1}\Pr\} = 

2178: \big\{\bd\in\Sym\tkd \mid \bd\cdot(\i^{-1}\Pr) = 0\big\}.$$

2179: \end{theorem}

2180: 

2181: A similar description of the cohomology ring of a hypertoric variety

2182: is given in \cite{HS}.  For any subset

2183: $A\subs\otn$, consider the polyhedron $\DrA$ introduced in Section \ref{htcore}.

2184: If $\DrA$ is nonempty, then it is bounded if and only if the vectors

2185: $\{\eps_1(A)a_1,\ldots,\eps_n(A)a_n\}$ span $\td$ over the non-negative

2186: real numbers,

2187: where $\eps_i(A) = (-1)^{|A\cap\{i\}|}$.

2188: We call such an $A$ {\em admissible}.  For all admissible $A$,

2189: there exists a degree $d$ polynomial

2190: $\PrA\in\Sym^d\tn$ such that for every simple $s\in\tnd$

2191: lying in the same connected component of the set of simple elements as $r$,

2192: we have $$\Vol\D^{s}_A=\PrA(s).$$

2193: Once again, the translation invariance of volume implies that

2194: $\PrA$ lies in the image of the inclusion

2195: $\i:\Sym^d\tk\hookto\Sym^d\tn$.

2196: Consider the linear span

2197: $$U^r = \Q\big\{\PrA \mid A\text{ admissible}\big\}.$$

2198: 

2199: \begin{theorem}\label{hypertoric}{\em\cite{HS}}

2200: Let $r\in\tnd$ be a simple element with $\i^*(r)=\a$.  Then

2201: % $$\ker (\hk\circ\i^*) = \Ann U^r = \big\{\bd\in\Sym\tnd \mid 

2202: % \bd\cdot P = 0\text{ for all }P\in U^r\big\}$$ and 

2203: $$\ker\k = \Ann \,\i^{-1}U^r = \big\{\bd\in\Sym\tkd \mid 

2204: \bd\cdot \i^{-1}P = 0\text{ for all }P\in U^r\big\}.$$

2205: \end{theorem}

2206: 

2207: \begin{remark}\label{wow}

2208: It is clear from the statement of Theorem \ref{hypertoric}

2209: that $H^*(\Ma)$ does not depend on the coorientations of the hyperplanes

2210: $\{H_1^r,\ldots,H_n^r\}$, as has been observed in Lemma \ref{coorientation}

2211: and throughout Section \ref{htcoh}.  Indeed,

2212: the polynomials $\PrA$ for $A$ admissible

2213: are simply the volume polynomials of the maximal regions of $|\A|$.

2214: What is not clear from this presentation is the independence of $H^*(\Ma)$

2215: on $\a$.  In other words, it is a nontrivial

2216: fact that the vector space $U^r$ is independent of the parameter $r\in\tnd$.

2217: \end{remark}

2218: 

2219: \begin{proofint}

2220: The statement of Theorem \ref{int} equates the kernel of $\k$, 

2221: which is cogenerated by the vector

2222: space $\i^{-1}U^r$, with the intersection of the kernels of the maps $\ka$,

2223: each of which is cogenerated by the element $\i^{-1}P^r$.  Intersection of ideals

2224: corresponds to linear span on the level of cogenerators, hence we have

2225: $$\bigcap_{\a\text{ simple}}\ker\ka = \Ann \,\i^{-1}V,\hspace{15pt}\text{where}

2226: \hspace{15pt}V = \Q\big\{\Pr\mid r\text{ simple}\}.$$

2227: Our plan is to show that $V = U^r$ for any simple $r$.

2228: Recall that the assignment of $\Pr$ to $r$ is locally constant on the set

2229: of simple elements of $\tnd$, hence $V$ is finite-dimensional.

2230: Since $\Pr\in U^r$ and $U^r$ does not depend on $r$ (see Remark \ref{wow}),

2231: it is clear that $V\subs U^r$.  Thus to prove Theorem \ref{int}, it will suffice

2232: to prove the opposite inclusion, as stated below.

2233: 

2234: \begin{proposition}\label{comb}

2235: We have $\PrA\in V$ for every admissible $A\subs\otn$.

2236: \end{proposition}

2237: 

2238: Let $\F$ be the infinite dimensional vector space consisting of all real-valued

2239: functions on $\tdd$, and let $\Fbd$ be the subspace consisting of functions

2240: with bounded support.  For all subsets $A\subs\otn$, let

2241: $$\WA = \Q\big\{\odra\mid r\text{ simple}\big\}$$

2242: be the subspace of $\F$ consisting of finite linear combinations

2243: of characteristic functions of polyhedra $\DrA$, and let

2244: $$\WAbd = \WA\cap\Fbd.$$

2245: Note that $\WAbd=\WA$ if and only if $A$ is admissible.

2246: 

2247: \begin{lemma}\label{key}

2248: For all $A,A'\subs\otn$, $\WAbd = W_{A'}^{bd}$.

2249: \end{lemma}

2250: 

2251: \begin{proof}

2252: We may immediately reduce to the case where $A' = A\cup\{j\}$.

2253: Fix a simple $r\in\tnd$.

2254: Let $\rt\in\tnd$ be another simple element obtained from $r$ by putting

2255: $\rt_i=r_i$ for all $i\neq j$, and $\rt_j = N$ for some $N\gg 0$.

2256: Then $\D_A^r\subs\D_A^{\rt}$, and 

2257: \begin{eqnarray*}

2258: \D_{A}^{\rt} \smallsetminus \D_{A}^{r} &=& 

2259: \big\{v\in\tdd\mid \eps_i(A')(v\cdot a_i+r_i)\geq 0\text{  for all  }i\leq n

2260: \text{ and } v\cdot a_j + N\geq 0\}\\

2261: &=& \D_{A'}^{r} \cap G_j^{N}.

2262: \end{eqnarray*}

2263: Suppose that $f\in\Fbd$ can be written as a linear combination

2264: of functions of the form $\mathbf{1}_{\D_{A'}^{r}}$.  Choosing $N$ large enough

2265: that the support of $f$ is contained in $G_j^{N}$,

2266: the above computation shows that $f$ can be written

2267: as a linear combination of functions of the form $\odra$,

2268: hence $W_{A'}^{bd}\subs W_{A}^{bd}$.  The reverse inclusion is obtained

2269: by an identical argument.

2270: \end{proof}

2271: 

2272: \vspace{-\baselineskip}

2273: \begin{example}\label{pictures}

2274: Suppose that we want to write the characteristic function for the

2275: upper triangle $\Delta_{\{1,4\}}$ 

2276: in Figure~\ref{triplet}(c) as linear combination

2277: of characteristic functions of the shaded regions obtained by translating

2278: the hyperplanes in any possible way.

2279: Since $\{1,4\}$ has two elements, the procedure described

2280: in Lemma \ref{key} must be iterated twice, and the result will have a total

2281: of $2^2=4$ terms, as illustrated in Figure \ref{char}.

2282: The first iteration exhibits $\mathbf{1}_{\Delta_{\{1,4\}}}$ as an element

2283: of $W_{\{4\}}^{bd}$ by expressing it as the difference of the characteristic

2284: functions of two (unbounded) regions.

2285: With the second iteration, we attempt to express each of these two characteristic functions

2286: as elements of $W_{\{1,4\}}^{bd} = W_{\{1,4\}}$.

2287: This attempt must fail, because each of the two functions that we try

2288: to express has unbounded support.  But the failures cancel out, and we succeed

2289: in expressing the {\em difference} as an element of $W_{\{1,4\}}$.

2290: \begin{figure}[h]

2291: \begin{center}

2292: \includegraphics[height=140mm]{newchar.eps}

2293: \caption{An equation of characteristic functions.

2294: We write two numbers next to each hyperplane:  the first is the 

2295: index $i\in\{1,\ldots,4\}$, and the second is the parameter $r_i$

2296: specifying the distance from the origin (denoted by a black dot) to $H_i$.

2297: The two iterations of Lemma \ref{key} have produced two undetermined

2298: large numbers, which we call $N$ and $N'$.}\label{char}

2299: \end{center}\end{figure}

2300: \end{example}

2301: 

2302: \begin{proofcomb}

2303: By Lemma \ref{key}, we may write

2304: $$\odra = \sum_{j=1}^m\eta_j \mathbf{1}_{\D^{r(j)}}$$

2305: for any simple $r$ and admissible $A$, where $\eta_j\in\Z$ and

2306: $r(j)$ is a simple element of $\tnd$ for all $j\leq m$.

2307: Taking volumes of both sides of the equation, we have

2308: \begin{equation}\label{polys}

2309: \PrA(r) = \sum_{j=1}^m\eta_j P^{r(j)}\big(r(j)\big).

2310: \end{equation}

2311: Furthermore, we observe from the proof of Lemma \ref{key}

2312: that for all $j\leq m$ and all $i\leq n$, the $i^{\text{th}}$

2313: coordinate $r_i(j)$ of $r(j)$ is either equal to $r_i$, or to

2314: some large number number $N\gg 0$.  The Equation~\eqref{polys}

2315: still holds if we wiggle these large numbers a little bit,

2316: hence the polynomial $P^{r(j)}$ must be independent of the variable

2317: $r_i(j)$ whenever $r_i(j)\neq r_i$.  Thus we may substitute

2318: $r$ for each $r(j)$ on the right-hand side, and we obtain the equation

2319: \begin{equation*}

2320: \PrA(r) = \sum_{j=1}^m\eta_j P^{r(j)}(r).

2321: \end{equation*}

2322: This equation clearly holds in a neighborhood of $r$,

2323: hence we obtain an equation of polynomials

2324: $$\PrA = \sum_{j=1}^m\eta_j P^{r(j)}.$$

2325: This completes the proof of

2326: Proposition \ref{comb}, and therefore also of Theorem \ref{int}.

2327: \end{proofcomb}

2328: \end{proofint}

2329: 

2330: \vspace{-\baselineskip}

2331: \begin{example}

2332: Let's see what happens when we take volume polynomials in the equation

2333: of Figure \ref{char}.  The two polytopes on the top line have different

2334: volumes, but the same volume polynomial, hence these two terms cancel.

2335: We are left with the equation

2336: $$P_{\{1,4\}}^{(0,1,1,0)} = P^{(0,1,1,0)} - P^{(N,1,1,0)},$$

2337: which translates as

2338: $$\half\(-x_1+x_2-x_4\)^2 = \half\(x_1+x_3+x_4\)^2 - 

2339: \(x_2+x_3\)\(x_1+x_4+\half x_3 - \half x_2\).$$

2340: \end{example}

2341: \end{section}

2342: \end{chapter}

2343: 

2344: \begin{chapter}{Abelianization}\label{abelianization}

2345: Let $X$ be a symplectic manifold equipped with a hamiltonian

2346: action of a compact Lie group $G$.  Let $T\subs G$

2347: be a maximal torus, let $\Delta\subset\tdu$ 

2348: be the set of roots\footnote{Not to be confused with the polyhedron $\Delta$

2349: of Chapter \ref{ht}.} of $G$,

2350: and let $W=N(T)/T$ be the Weyl group of $G$.

2351: If $\mu:X\to\gd$ is a moment map for the action of $G$, then $pr\circ\mu:X\to\tdu$

2352: is a moment map for the action of $T$, where $pr:\gd\to\tdu$ is the standard projection.

2353: Suppose that $0\in\gd$ and $0\in\tdu$ are regular values for the two moment maps.

2354: If the symplectic quotients

2355: $$X\mod G = \mu^{-1}(0)/G\hspace{15pt}\text{and}

2356: \hspace{15pt}X\mod T=(pr\circ\mu)^{-1}(0)/T$$ 

2357: are both compact, then Martin's theorem \cite[Theorem A]{Ma}

2358: relates the cohomology of $X\mod G$ to the cohomology of $X\mod T$.

2359: Specifically, it says that $$H^*(X\mod G)\cong\frac{H^*(X\mod T)^W}{ann(e_0)},$$

2360: where $$e_0 = \prod_{\a\in\Delta}\a\in\left(\operatorname{Sym}\tdu\right)^W

2361: \cong H^*_T(pt)^W,$$ which acts naturally on

2362: $H^*(X\mod T)^W \cong H^*_T(\mu_T^{-1}(0))^W$.

2363: In the case where $X$ is a complex vector space and $G$ acts linearly

2364: on $X$, a similar result was obtained by Ellingsrud and Str\o mme \cite{ES}

2365: using different techniques.

2366: 

2367: Our goal is to state and prove an analogue of this theorem for

2368: hyperk\"ahler quotients.  There are two main obstacles to this goal.

2369: First, hyperk\"ahler quotients are rarely compact.

2370: The assumption of compactness in Martin's theorem is crucial

2371: because his proof involves integration.

2372: Our answer to this problem is to work

2373: with equivariant cohomology

2374: of {\em circle compact} manifolds, by which we mean oriented manifolds

2375: with an action of $\so$ such that the fixed point set is oriented and compact.

2376: Using the localization theorem of Atiyah-Bott \cite{AB} and Berline-Vergne \cite{BV},

2377: as motivation, we show that integration in rationalized $\so$-equivariant cohomology

2378: of circle compact manifolds can be defined in terms of integration

2379: on their fixed point sets.  Section~\ref{thepush} is devoted to

2380: making this statement precise by defining a well-behaved

2381: push forward in the rationalized $\so$-equivariant cohomology

2382: of circle compact manifolds.

2383: 

2384: The second obstacle is that Martin's result uses surjectivity

2385: of the K\"ahler Kirwan map from $H^*_G(X)$ to $H^*(X\mod G)$ \cite{Ki}.  

2386: The analogous map for circle compact

2387: hyperk\"ahler quotients is conjecturally surjective, but

2388: only a few special cases are known (see Theorems \ref{hm} and \ref{hpsurj}, and 

2389: Remarks \ref{generation} and \ref{horn}).

2390: Our approach is to assume that the rationalized Kirwan map is

2391: surjective, which is equivalent to saying that the cokernel

2392: of the non-rationalized Kirwan map

2393: $$\K_G:H^*_{\so\times G}(M)\to\hmg$$

2394: is torsion as a module over $\hp$.  This is a weaker assumption

2395: than surjectivity of $\K_G$; in particular, we show in Section

2396: \ref{quiver} that this assumption holds for quiver varieties,

2397: as a consequence of the work of Nakajima.

2398: 

2399: Under this assumption,

2400: Theorem~\ref{main} computes the rationalized equivariant cohomology

2401: of $M\mmod G$ in terms of that of $M\mmod T$.

2402: We show that, at regular values of the hyperk\"ahler moment maps,

2403: $$\hhso(M\mmod G)\cong\frac{\hhso(M\mmod T)^W}{ann(e)},$$

2404: where $\hhso$ denotes rationalized equivariant cohomology

2405: (see Definition \ref{rat}), and

2406: $$e = \Pd_{\a\in\Delta}\a(x-\a) \in (\operatorname{Sym}\tdu)^W\otimes\Q[x]

2407: \cong H^*_{\so\times T}(pt)^W\subs\widehat{H}^*_{\so\times T}(pt)^W.$$

2408: Theorem~\ref{ordinary}

2409: describes the image of the non-rationalized Kirwan map

2410: in a similar way:

2411: $$\hso(M\mmod G)\supseteq\Im(\K_G)\cong\frac{\imkt^W}{ann(e)},$$

2412: where $\K_T:H^*_{\so\times T}(M)\to \hso(M\mmod T)$ is the Kirwan

2413: map for the abelian quotient.

2414: In many situations, such as when $M=\cot$,

2415: $\K_T$ is known to be surjective (Theorem \ref{hm}).

2416: 

2417: This Chapter is a reproduction of \cite[\S 1-3]{HP}.

2418: 

2419: \begin{section}{Integration}\label{thepush}

2420: The localization theorem of Atiyah-Bott \cite{AB} and Berline-Vergne \cite{BV} says that

2421: given a manifold $M$ with a circle action, the restriction map from

2422: the circle-equivariant cohomology of $M$ to the circle-equivariant

2423: cohomology of the fixed point set $F$ is an isomorphism modulo torsion.

2424: In particular, integrals on $M$ can be computed in terms of integrals on $F$.

2425: If $F$ is compact, it is possible to use the Atiyah-Bott-Berline-Vergne formula

2426: to {\em define} integrals on $M$.

2427: 

2428: We will work in the category of {\em circle compact}

2429: manifolds, by which we mean oriented $\so$-manifolds with compact

2430: and oriented fixed point sets.

2431: Maps between circle compact

2432: manifolds are required to be equivariant.

2433: % Our goal is to define a cohomological pushforward in the category

2434: % of circle compact manifolds, and derive some of its basic properties.

2435: 

2436: \begin{definition}\label{rat}

2437: Let $\hhp = \Q(x)$, the rational function field of $\hp\cong \Q[x]$.

2438: For a circle compact manifold $M$, let $\hhm = \hm\otimes\hhp$,

2439: where the tensor product is taken over the ring $\hp$. We call $\hhm$

2440: the {\em rationalized} $\so$-equivariant cohomology of $M$.

2441: Note that because $\operatorname{deg}(x)=2$, $\hhm$ is supergraded,

2442: and supercommutative with respect to this supergrading.

2443: \end{definition}

2444: 

2445: An immediate consequence of \cite{AB} is that restriction gives an isomorphism

2446: \begin{equation}\label{isomorphic}\hhm\cong \hhso(F)\cong H^*(F)\otimes_{\Q}\hhp,\end{equation}

2447: where $F=M^{S^1}$ denotes

2448: the compact fixed point set of $M$.  In particular

2449: $\hhm$ is a finite dimensional vector space over $\hhp$,

2450: and trivial if and only if $F$ is empty.

2451: 

2452: Let $i:N\hookto M$ be a closed embedding.

2453: There is a standard notion of proper pushforward

2454: $$i_*:\hso(N)\to\hm$$ given by the formula $i_*=r\circ\Phi$,

2455: where $r:\hso(M,M\setminus N)\to\hso(M)$ is the restriction map,

2456: and $\Phi:\hso(N)\to\hso(M,M\setminus N)$ is the Thom isomorphism.

2457: We will also denote the induced map $\hhso(N)\to\hhm$ by $i_*$.

2458: Geometrically, $i_*$ can be understood as the inclusion

2459: of cycles in Borel-Moore homology.

2460: 

2461: This map satisfies two important formal properties

2462: \cite{AB}:

2463: \begin{equation}\label{functor}

2464: \text{Functoriality:   } (i\circ j)_* = i_*\circ j_*

2465: \end{equation}

2466: \begin{equation}\label{module}

2467: \text{Module homomorphism:   } i_*(\ga\cdot i^*\a)

2468: =i_\ga\cdot\a

2469: \text{   for all   }\a\in\hhm, \ga\in\hhn.

2470: \end{equation}

2471: We will denote the Euler

2472: class $i^*i_*(1)\in\hhso(N)$ by $e(N)$.

2473: If a class $\ga\in\hhn$ is in the image of $i^*$,

2474: then property (\ref{module}) tells us that

2475: $i^*i_*\ga = e(N)\ga$.  Since the pushforward construction is

2476: local in a neighborhood

2477: of $N$ in $M$, we may assume that $i^*$ is surjective, hence this identity

2478: holds for all $\ga\in\hhn$.

2479: 

2480: Let $F= M^{\so}$ be the fixed point set of $M$.

2481: Since $M$ and $F$ are each oriented, so is the normal bundle

2482: to $F$ inside of $M$.  The following result is

2483: standard, see e.g. \cite{Ki}.

2484: 

2485: \begin{lemma}

2486: The Euler class $e(F)\in\hhso(F)$ of the normal bundle to $F$ in $M$

2487: is invertible.

2488: \end{lemma}

2489: 

2490: \begin{proof}

2491: Let $\{F_1,\ldots,F_d\}$ be the connected components of $F$.

2492: Since $\hhso(F)\cong\bigoplus\hhso(F_i)$ and $e(F)=\oplus e(F_i)$,

2493: our statement is equivalent to showing that $e(F_i)$ is invertible for all $i$.

2494: Since $\so$ acts trivially on $F_i$, $\hhso(F_i)\cong H^*(F_i)\otimes_{\Q}\hhp$.

2495: We have $e(F_i) = 1\otimes ax^k + nil$, where

2496: $k= \operatorname{codim}(F_i)$, $a$ is the product of the weights of the $S^1$

2497: action on any fiber of the normal bundle, and $nil$ consists of terms

2498: of positive degree in $H^*(F_i)$.  Since $F_i$ is a component of the fixed

2499: point set, $\so$ acts freely on the complement of the zero section

2500: of the normal bundle, therefore $a\neq 0$.  Since $ax^k$ is invertible and

2501: $nil$ is nilpotent, we are done.

2502: \end{proof}

2503: 

2504: \vspace{-\baselineskip}

2505: \begin{definition}

2506: For $\a\in\hhm$, let $$\IM\a=\IF\frac{\a|_F}{e(F)}\in\hhp.$$

2507: \end{definition}

2508: 

2509: Note that this definition does not depend on our choice

2510: of orientation of $F$.  Indeed, reversing the orientation of $F$

2511: has the effect of negating $e(F)$, {\em and} introducing a second factor

2512: of $-1$ coming from the change in fundamental class.  These two effects

2513: cancel.

2514: 

2515: For this definition to be satisfactory, we must be able to prove

2516: the following lemma, which is standard in the setting of

2517: ordinary cohomology of compact manifolds.

2518: 

2519: \begin{lemma}\label{Npush}

2520: Let $i:N\hookto M$ be a closed immersion.

2521: Then for any $\a\in\hhm,\ga\in\hhso(N)$, we have

2522: $\IM\a\cdot i_*\ga = \IN i^*\a\cdot\ga$.

2523: \end{lemma}

2524: 

2525: \begin{proof}

2526: Let $G=N^{\so}$, let $j:G\to F$ denote the restriction of $i$ to $G$,

2527: and let $\phi:F\to M$ and $\psi:G\to N$ denote the inclusions of $F$ and $G$

2528: into $M$ and $N$, respectively.

2529: $$\begin{CD}

2530: N        @>i>> M\\

2531: @A\psi AA @AA\phi A\\

2532: G        @>j>> F

2533: \end{CD}$$

2534: Then $$\IM\a\cdot i_*\ga = \IF\frac{\phi^*\a\cdot\phi^*i_*\ga}{e(F)},$$

2535: and $$\IN i^*\a\cdot\ga = \IG\frac{\psi^*i^*\a\cdot\psi^*\ga}{e(G)}

2536: =\IG\frac{j^*\phi^*\a\cdot\psi^*\ga}{e(G)}

2537: =\IF\phi^*\a\cdot j_*\left(\frac{\psi^*\ga}{e(G)}\right),$$

2538: where the last equality is simply the integration formula applied

2539: to the map $j:G\to F$ of compact manifolds \cite{AB}.

2540: Hence it will be sufficient to prove that

2541: $$\phi^*i_*\ga=e(F)\cdot j_*\left(\frac{\psi^*\ga}{e(G)}\right)\in\hhso(F).$$

2542: To do this, we will

2543: show that the difference of the two classes lies in

2544: the kernel of $\phi_*$, which we know is trivial because the composition

2545: $\phi^*\phi_*$ is given by multiplication by the invertible

2546: class $e(F)\in\hhso(F)$.

2547: On the left hand side we get

2548: $$\phi_*\phi^*i_*\ga = \phi_*(1)\cdot i_*\ga\hspace{15pt}

2549: \text{by    }(\ref{module}),

2550: $$

2551: and on the right hand side we get

2552: \begin{eqnarray*}

2553: \phi_*\left(e(F)\cdot j_*\left(\frac{\psi^*\ga}{e(G)}\right)\right)

2554: &=& \phi_*\left(\phi^*\phi_*(1)

2555: \cdot j_*\left(\frac{\psi^*\ga}{e(G)}\right)\right)\\

2556: &=& \phi_*(1)\cdot \phi_*j_*\left(\frac{\psi^*\ga}{e(G)}\right)

2557: \hspace{1in}\text{    by    }(\ref{module})\\

2558: &=& \phi_*(1)\cdot i_*\psi_*\left(\frac{\psi^*\ga}{e(G)}\right)

2559: \hspace{1in}\text{    by    }(\ref{functor}).

2560: \end{eqnarray*}

2561: It thus remains only to show that

2562: $\ga = \psi_*\left(\frac{\psi^*\ga}{e(G)}\right)$.

2563: This is seen by applying $\psi^*$ to both sides,

2564: which is an isomorphism (working over the field $\hhp$) by \cite{AB}.

2565: % But this is simply the localization theorem of \cite{AB}, proven by applying

2566: % $\psi^*$ to both sides.  (Atiyah and Bott prove that

2567: % $\psi^*:\hso(N)\to\hso(G)$ is injective

2568: % modulo torsion; by tensoring with $\hhp$, we are working modulo torsion.)

2569: \end{proof}

2570: 

2571: For $\a_1,\a_2\in\hhm$, consider the

2572: symmetric, bilinear, $\hhp$-valued pairing

2573: $$\la\a_1,\a_2\ra_M=\IM\a_1\a_2.$$

2574: 

2575: \begin{lemma}[Poincar\'e Duality]\label{perfect}

2576: This pairing is nondegenerate.

2577: \end{lemma}

2578:  

2579: \begin{proof}

2580: Suppose that $\a\in\hhm$ is nonzero, and therefore $\phi^*\a\neq 0$.

2581: Since $F$ is

2582: compact, there must exist a class $\ga\in\hhso(F)$ such

2583: that $0\neq\IF\phi^*\a\cdot\ga=\IM\a\cdot \phi_*\ga=\la\a,\phi_*\ga\ra_M$.

2584: \end{proof}

2585: 

2586: \vspace{-\baselineskip}

2587: \begin{definition}\label{push}

2588: For an arbitrary equivariant map $f:N\to M$, we may now define

2589: the pushforward $$f_*:\hhn\to\hhm$$ to

2590: be the adjoint of $f^*$ with respect to the pairings

2591: $\la\cdot,\cdot\ra_N$ and $\la\cdot,\cdot\ra_M$.

2592: This is well defined because, according to (\ref{isomorphic}),

2593: $\hhm$ and $\hhn$ are finite dimensional

2594: vector spaces over the field $\hhp$. Lemma~\ref{Npush} tells us that this definition generalizes

2595: the definition for closed immersions.  Furthermore,

2596: properties (\ref{functor}) and (\ref{module}) for pushforwards along

2597: arbitrary maps

2598: are immediate corollaries of the definition.

2599: If $f$ is a projection, then $f_*$ will be given by

2600: integration along the fibers.

2601: Using the fact that

2602: every map factors through its graph as a closed immersion

2603: and a projection, we always have a geometric interpretation of the pushforward.

2604: \end{definition}

2605: 

2606: As an application, let us consider the manifold $M\times M$,

2607: along with the two projections $\pi_1$ and $\pi_2$,

2608: and the diagonal map $\Diag:M\to M\times M$.

2609: Suppose that we can write $$\Diag_*(1)=\sum\pi_1^*a_i\cdot\pi_2^*b_i$$

2610: for a finite collection of classes $a_i,b_i\in\hhm$.

2611: The following Proposition will be used in Section~\ref{quiver}.

2612: 

2613: \begin{proposition}\label{basis}

2614: The set $\{b_i\}$ is an additive spanning set for $\hhm$.

2615: \end{proposition}

2616: 

2617: \begin{proof}

2618: For any $\a\in\hhm$, we have

2619: \begin{eqnarray*}

2620: \a &=& \id_*\id^*\a\\

2621: &=& (\pi_2\circ\Diag)_*(\pi_1\circ\Diag)^*\a\\

2622: &=& \pi_{2*}\big(\Diag_*\left(1\cdot\Diag^*\pi_1^*\a\right)\big)\\

2623: &=& \pi_{2*}\big(\pi_1^*\a\cdot\Diag_*(1)\big)\\

2624: &=& \pi_{2*}\left(\sum\pi_1^*(a_i\a)\cdot\pi_2^*b_i\right)\\

2625: &=& \sum\pi_{2*}\pi_1^*(a_i\a) \cdot b_i\\

2626: &=& \sum\la a_i,\a\ra\cdot b_i,

2627: \end{eqnarray*}

2628: hence $\a$ is in the span of $\{b_i\}$.

2629: \end{proof}

2630: \end{section}

2631: 

2632: \begin{section}{Hyperk\"ahler abelianization}

2633: Let $M$ be a hyperk\"ahler manifold with a circle action, and

2634: suppose that a compact Lie group $G$ acts hyperhamiltonianly on $M$

2635: with hyperk\"ahler moment map

2636: $$\muh=\mur\oplus\muc:M\to\gd\oplus\gd_{\C},$$

2637: where $\muc$ is holomorphic with respect to the distinguished

2638: complex structure $I$.

2639: We require that the action of $G$ commute with the action of $\so$,

2640: that $\mur$ is $\so$-invariant, and that $\muc$ is $\so$-equivariant

2641: with respect to the action of $\so$ on $\gd_{\C}$ by complex multiplication.

2642: 

2643: Let $T\subs G$ be a maximal torus, and let $pr:\gd\to\tdu$

2644: be the natural projection.  Then $T$ acts on $M$ with hyperk\"ahler

2645: moment map

2646: $$\mubh=pr\!\circ\!\mur\,\oplus\, pr_{\C}\!\circ\!\muc:

2647: M\to\tdu\oplus\tdu_{\C}.$$

2648: Let $\xi\in\gd$ be a central element such that $(\xi,0)$ is a regular

2649: value of $\muh$ and $(pr(\xi),0)$ is a regular value of $\mubh$.

2650: Assume further that $G$ acts freely on $\muh^{-1}(\xi,0)$,

2651: and $T$ acts freely on $\mubh^{-1}(pr(\xi),0)$.\footnote{We make this

2652: simplifying assumption in order to talk about manifolds, rather than

2653: orbifolds, which makes the integration formulae easier to state.

2654: In fact, Theorems \ref{main} and \ref{ordinary}

2655: generalize easily to the orbifold case, as in \cite[\S 6]{Ma}.}

2656: Let $$M\mmod G = \muh^{-1}(\xi,0)/G\hs\hs\hs\text{   and   }\hs\hs\hs

2657: M\mmod T = \mubh^{-1}(pr(\xi),0)/T$$

2658: be the hyperk\"ahler quotients of $M$ by $G$ and $T$, respectively.

2659: Because $\muh$ and $\mubh$ are circle-equivariant,

2660: the action of $S^1$ on $M$ descends to actions on the hyperk\"ahler quotients.

2661: Note that $M\mmod T$ also inherits an action of the Weyl group $W$ of $G$.

2662: 

2663: \begin{example}\label{standard}

2664: The main example to keep in mind is $M=\cot$, where $\so$

2665: acts on $M$ by scalar multiplication on the fibers and the action

2666: of $G$ on $M$ is induced by a linear action of $G$ on $\Cn$,

2667: as in Chapter \ref{analogues}.

2668: \end{example}

2669: 

2670: Consider the Kirwan maps

2671: $$\k_G:H_{\so\times G}^*(M)\to \hso(M\mmod G)\hs\hs\hs\text{   and   }\hs\hs\hs

2672: \k_T:H_{\so\times T}^*(M)\to \hso(M\mmod T),$$

2673: induced by the inclusions of $\muh^{-1}(\xi,0)$ and $\mubh^{-1}(pr(\xi),0)$ into $M$,

2674: along with their rationalizations

2675: $$\hat\k_G:\widehat H_{\so\times G}^*(M)\to \hhso(M\mmod G)\hs\hs\hs\text{   and   }\hs\hs\hs

2676: \hat\k_T:\widehat H_{\so\times T}^*(M)\to \hhso(M\mmod T).$$

2677: Let $$\rgt:\widehat{H}^*_{\so\times G}(M)\to\widehat{H}^*_{\so\times T}(M)^W$$

2678: be the standard isomorphism.

2679: 

2680: % All of these maps are compatible with the module structures

2681: % of each of the rings over the ground ring

2682: % $H_{\so\times G}(pt) \cong (\operatorname{Sym}\tdu)^W\otimes\Q[x]$,

2683: % which includes into the rings $\widehat{H}_{\so\times G}(pt)$,

2684: % $H_{\so\times T}(pt)$, and $\widehat{H}_{\so\times T}(pt)$.

2685: Let $\Delta=\Delta^+\sqcup\Delta^-\subset\tdu$ be the set of roots of $G$.

2686: Let $$e = \Pd_{\a\in\Delta}\a(x-\a) \in (\operatorname{Sym}\tdu)^W\otimes\Q[x]

2687: \cong H_{\so\times G}(pt)\subs\widehat{H}_{\so\times G}(pt),$$

2688: and $$e' = \Pd_{\a\in\Delta^-}\a\cdot\Pd_{\a\in\Delta}(x-\a)

2689: \in\operatorname{Sym}\tdu\otimes\Q[x]\cong H_{\so\times T}(pt)

2690: \subs\widehat{H}_{\so\times T}(pt).$$

2691: The following two theorems are analogues of

2692: Theorems B and A of \cite{Ma}, adapted to circle compact

2693: hyperk\"ahler quotients.  Our proofs follow closely those of Martin.

2694: 

2695: \begin{theorem}\label{integration}

2696: Suppose that $M\mmod G$ and $M\mmod T$ are both circle compact.

2697: If $\ga\in\hhsg$, then

2698: $$\intxg \Kh_G(\ga) = \frac{1}{|W|}\intxt \Kh_T\circ\rgt(\ga)\cdot e.$$

2699: \end{theorem}

2700: 

2701: \begin{theorem}\label{main}

2702: Suppose that $M\mmod G$ and $M\mmod T$ are both circle compact,

2703: and that the rationalized Kirwan map $\Kh_G$

2704: is surjective.  Then

2705: $$\hhso(M\mmod G)\cong\frac{\hhso(M\mmod T)^W}{ann(e)}

2706: \cong\left(\frac{\hhso(M\mmod T)}{ann(e')}\right)^W.$$

2707: % and $$\hso(M\mmod G)\supseteq\Im(\K_G)\cong\frac{\hso(M\mmod T)^W}{ann(e)}

2708: % \cong\left(\frac{\hso(\M\mmod T)}{ann(e')}\right)^W,$$

2709: \end{theorem}

2710: 

2711: \begin{proofintegration}

2712: Consider the following pair of maps:

2713: $$\begin{CD}

2714: \muh^{-1}(\xi,0)/T  @>i>> \mubh^{-1}(pr(\xi),0)/T\cong M\mmod T\\

2715: @V\pi VV \\

2716: \muh^{-1}(\xi,0)/G\cong M\mmod G.

2717: \end{CD}$$

2718: Each of these spaces is a complex $S^1$-manifold with a compact, complex

2719: fixed point set, and therefore satisfies the hypotheses of Section

2720: \ref{thepush}.

2721: Let $$b = \Pd_{\a\in\Delta^+}\a\in H^*_{\so\times T}(pt)$$ be the product of the

2722: positive roots of $G$, which we will think of as an

2723: element of $\hhxt$.  Martin shows that $\pi_*i^*b=|W|$,

2724: and that $i^*\circ\Kh_T\circ\rgt = \pi^*\Kh_G$ \cite{Ma}, hence

2725: we have

2726: \begin{eqnarray*}

2727: \intxg\Kh_G(\ga) &=& \frac{1}{|W|}\intxg\Kh_G(\ga)\cdot\pi_*i^*b\\

2728: &=& \frac{1}{|W|}\int_{\muh^{-1}(\xi,0)/T}\pi^*\Kh_G(\ga)\cdot i^*b\hspace{20pt}

2729: \text{by Definition~\ref{push}}\\

2730: &=& \frac{1}{|W|}\int_{\muh^{-1}(\xi,0)/T}i^*\circ\Kh_T\circ\rgt(\ga)\cdot i^*b\\

2731: &=& \frac{1}{|W|}\intxt \Kh_T\circ\rgt(\ga)\cdot b \cdot i_*(1)\hspace{20pt}

2732: \text{by Lemma~\ref{Npush}}.

2733: \end{eqnarray*}

2734: It remains to compute $i_*(1)\in\hhxt$.

2735: For $\a\in\Delta$, let $$L_{\a} = \mubh^{-1}((pr(\xi),0)\times_T\C_{\a}$$

2736: be the line bundle on $M\mmod T$ with $S^1$-equivariant

2737: Euler class $\a$.

2738: Similarly, let $L_x$ be the (topologically trivial)

2739: line bundle with $S^1$-equivariant

2740: Euler class $x$.

2741: Following the idea of \cite[1.2.1]{Ma}, we observe

2742: that the restriction of $\muh - (\xi,0)$ to

2743: $\mubh^{-1}(pr(\xi),0)$ defines

2744: an $\so\times T$-equivariant map $$s:\mubh^{-1}(pr(\xi),0)

2745: \to V\oplus V_{\C},$$ where $V = pr^{-1}(0)$

2746: and $V_{\C}=pr_{\C}^{-1}(0)$.

2747: This descends to an $S^1$-equivariant section of the

2748: bundle $$E = \mubh^{-1}(pr(\xi),0)\times_T \left(V\oplus V_{\C}\right)$$

2749: with zero locus $\muh^{-1}(\xi,0)/T$.  The fact that $(\xi,0)$

2750: is a regular value implies that this section is generic,

2751: hence the equivariant Euler class $e(E)\in\hhxt$ is equal to $i_*(1)$.

2752: 

2753: The vector space $V$ is isomorphic as a $T$-representation to

2754: $\bigoplus_{\a\in\Delta^-}\C_{\a}$,

2755: with $S^1$ acting trivially.

2756: Similarly, $V_{\C}$ is isomorphic to $V\otimes\C\cong V\oplus V^*$,

2757: with $S^1$ acting diagonally by scalars.

2758: Hence

2759: \begin{eqnarray*}

2760: E &\cong& \bigoplus_{\a\in\Delta^-}L_{\a}

2761: \oplus\bigoplus_{\a\in\Delta^-}\left(L_x\otimes L_{\a}\right)

2762: \oplus \left(L_x\otimes L_{-\a}\right)\\

2763: &\cong& \bigoplus_{\a\in\Delta^-}L_{\a}\oplus

2764: \bigoplus_{\a\in\Delta}L_x\otimes L_{-\a},

2765: \end{eqnarray*}

2766: and therefore $$i_*(1)=e(E) = 

2767: \prod_{\a\in\Delta^-}\a\cdot\prod_{\a\in\Delta}(x-\a)=e'.$$

2768: Multiplying by $b$ we obtain $e$, and the theorem is proved.

2769: \end{proofintegration}

2770: 

2771: % THIS IS STUPID; ANY RESTRICTION TO THE FIXED POINT SET

2772: % PROVIDES SUCH AN EXAMPLE.

2773: % Before proceeding with the proof of Theorem \ref{main} we present the

2774: % following example of Michael Thaddeus \cite{Th}, 

2775: % which shows that it is possible

2776: % for an inclusion of circle compact manifolds to induce a surjection

2777: % in rationalized $\so$-equivariant cohomology, 

2778: % while not inducing one in non-rationalized $\so$-equivariant cohomology.

2779: % 

2780: % \begin{example}\label{thaddeus}

2781: % Let $\so$ act on  $\C P^1\times\C P^1$ by

2782: % $$((x:y),(u:v))\mapsto ((\lambda x :y),(u:v))$$ and on $\C P^3$ by

2783: % $$(z_1:z_2:z_3:z_4)\mapsto (\lambda z_1: \lambda z_2: z_3:z_4).$$

2784: % Then the Segr\'e embedding $i:\C P^1\times\C P^1\to \C P^3$ given by

2785: % $$i\big((x:y),(u:v)\big)=(xu:xv:yu:yv)$$

2786: % is $\so$-equivariant, and clearly induces an isomorphism

2787: % on the fixed point sets of the

2788: % $\so$ action. Therefore $i^*: \hhso(\C P^3)\to \hhso(\C P^1\times \C P^1)$ 

2789: % is surjective, and in fact an isomorphism, however

2790: % $i^*:\hso(\C P^3)\to \hso(\C P^1\times \C P^1)$ is only an 

2791: % injection and therefore cannot be surjective.

2792: % \end{example}

2793: 

2794: \begin{proofmain}

2795: Observe that the restriction

2796: of $\pi^*$ to the Weyl-invariant part

2797: $\hhso\left(\muh^{-1}(\xi,0)/T\right)^W$ is given by the composition of

2798: isomorphisms

2799: $$\hhso\left(\muh^{-1}(\xi,0)/T\right)^W

2800: \cong\widehat{H}^*_{\so\times T}\left(\muh^{-1}(\xi,0)\right)^W

2801: \cong\widehat{H}^*_{\so\times G}\left(\muh^{-1}(\xi,0)\right)

2802: \cong\hhxg,$$

2803: hence we may define

2804: $$i^*_W:= (\pi^*)^{-1}\circ i^*:

2805: \hhxt^W\to\hhso\left(\muh^{-1}(\xi,0)/T\right)^W.$$

2806: Furthermore, we have

2807: $\Kh_G = i^*_W\circ\Kh_T\circ\rgt$, hence

2808: $i^*_W$ is surjective.

2809: As in \cite[\S 3]{Ma},

2810: \begin{eqnarray*}

2811: i^*_W(a)=0 &\iff& \forall c\in\hhxt^W, \intxg i^*_W(c)\cdot i^*_W(a)=0\hspace{15pt}

2812: \text{by \ref{perfect} and surjectivity of $i^*_W$}\\

2813: &\iff& \forall c\in\hhxt^W, \intxt c\cdot a\cdot e = 0\hspace{15pt}

2814: \text{by Theorem~\ref{integration}}\\

2815: &\iff& \forall d\in\hhxt, \intxt d\cdot a\cdot e=0\hspace{15pt}

2816: \text{by using $W$ to average $d$}\\

2817: &\iff& a\cdot e=0\hspace{15pt}\text{by Lemma~\ref{perfect}},

2818: \end{eqnarray*}

2819: hence $\ker i^*_W = ann(e)$.

2820: By surjectivity of $i^*_W$,

2821: \begin{eqnarray*}

2822: \hhso(M\mmod G) &\cong&

2823: \frac{\hhxt^W}{\ker i^*_W}\cong\frac{\hhxt^W}{ann(e)}.

2824: % &\cong& \frac{\hhxt^W}{ann(e)}\cong\left(\frac{\hhxt}{ann(e')}\right)^W.

2825: \end{eqnarray*}

2826: By a second application of Lemma~\ref{perfect}, for any $a\in\hhxt$, we have

2827: \begin{eqnarray*}

2828: i^*(a) = 0 &\impl& \forall f\in\hhso(\muh^{-1}(\xi,0)/T),

2829: \int_{\muh^{-1}(\xi,0)/T}f\cdot i^*(a) = 0\\

2830: &\impl& \forall c\in\hhxt,

2831: \int_{\muh^{-1}(\xi,0)/T} i^*(c)\cdot i^*(a)=0\\

2832: &\impl& \forall c\in\hhxt,

2833: \intxt c\cdot a\cdot i_*(1)=0\hspace{15pt}\text{by Lemma~\ref{Npush}}\\

2834: &\impl& a\cdot e'= a\cdot i_*(1) = 0\hspace{15pt}\text{by Lemma~\ref{perfect}},

2835: \end{eqnarray*}

2836: hence $\ker i^* \subs ann(e')$.

2837: This gives us a natural surjection

2838: $$\frac{\hhxt^W}{ann(e)}=\frac{\hhxt^W}{\ker i^*_W}

2839: \cong \left(\frac{\hhxt}{\ker i^*}\right)^W

2840: \to \left(\frac{\hhxt}{ann(e')}\right)^W,$$

2841: which is also injective because $e'$ divides $e$.

2842: This completes the proof of Theorem \ref{main}.

2843: \end{proofmain}

2844: 

2845: For the non-rationalized version of Theorem~\ref{main}, we make the additional

2846: assumption that $M\mmod G$ and $M\mmod T$ are equivariantly formal

2847: $\so$-manifolds, i.e. that $\hxg$ and $\hxt$ are free modules

2848: over $\hp$.  Proposition \ref{formality} tells us that this 

2849: is the case whenever the circle action

2850: is hamiltonian and its moment map is proper and bounded below.

2851: 

2852: \begin{theorem}\label{ordinary}

2853: Suppose that $M\mmod G$ and $M\mmod T$ are equivariantly formal, circle compact, and

2854: that the rationalized Kirwan map $\Kh_G$ is surjective.

2855: % and that at least one of the ordinary Kirwan maps $\K_G$ or $\K_T$

2856: % is surjective as well.

2857: Then

2858: $$\hso(M\mmod G)\supseteq\Im(\K_G)\cong\frac{(\Im\K_T)^W}{ann(e)}

2859: \cong\left(\frac{\Im\K_T}{ann(e')}\right)^W.$$

2860: \end{theorem}

2861: 

2862: \begin{remark}\label{notsobad}

2863: In the context of Example~\ref{standard} with $pr\circ\mu$ proper,

2864: $M\mmod G$ and $M\mmod T$ are both circle compact and equivariantly

2865: formal (Proposition \ref{formality}) and $\K_T$ is always surjective 

2866: (Theorem \ref{hm}).

2867: \end{remark}

2868: 

2869: \begin{proofordinary}

2870: Consider the following exact commutative diagram

2871: $$

2872: \xymatrix{

2873: 0 \ar[r] & A \ar[r]\ar[d] & \hxt^W \ar[r]^{i_W^*} \ar[d]

2874: & \hxg \ar[d]\\

2875: 0 \ar[r] & \widehat{A} \ar[r] & \hhxt^W \ar[r]^{i_W^*} & \hhxg.}

2876: $$

2877: Equivariant formality implies that the downward maps in the above

2878: diagram are inclusions, hence the map on top labeled $i^*_W$

2879: is simply the restriction of the map on the bottom

2880: to the subring $\hxt\subs\hhxt$.

2881: We therefore have $$A = \widehat{A}\cap\hxt^W = ann(e).$$

2882: Just as in the rationalized case, we have $\K_G = i_W^*\circ\K_T\circ\rgt$,

2883: hence

2884: $$\Im(\K_G)\cong i_W^*\left(\Im\K_T\circ\rgt\right)

2885: \cong \frac{(\Im\K_T)^W}{ann(e)}.$$

2886: 

2887: Now consider the analogous diagram

2888: $$

2889: \xymatrix{

2890: 0 \ar[r] & B \ar[r]\ar[d] & \hxt \ar[r]^{\!\!\!\!\!\!\!\!\! i^*} \ar[d]

2891: & \hso(\muh^{-1}(\xi,0)/T) \ar[d]\\

2892: 0 \ar[r] & \widehat{B} \ar[r] & \hhxt \ar[r]^{\!\!\!\!\!\!\!\!\! i^*} 

2893: & \hso(\muh^{-1}(\xi,0)/T).}

2894: $$

2895: Since we have not assumed that $\muh^{-1}(\xi,0)/T$ is equivariantly formal,

2896: we only know that the first two downward arrows are inclusions, and hence

2897: can only conclude that $B$ is contained in the annihilator of $e'$.

2898: Since $e'$ divides $e$, we have a series of natural surjections

2899: $$\frac{\imkt^W}{ann(e)}\cong

2900: \frac{\imkt^W}{A}\cong

2901: \left(\frac{\Im\K_T}{B}\right)^W\to

2902: \left(\frac{\Im\K_T}{ann(e')}\right)^W\to

2903: \left(\frac{\Im\K_T}{ann(e)}\right)^W.$$

2904: The composition of these maps is an isomorphism, hence so is each one.

2905: \end{proofordinary}

2906: \end{section}

2907: \end{chapter}

2908: 

2909: \begin{chapter}{Hyperpolygon spaces}\label{hpspaces}

2910: A hyperpolygon space is the hyperk\"ahler analogue of a polygon space,

2911: which parameterizes $n$-sided polygons in $\R^3$ with fixed edge lengths.

2912: It is also an example of a quiver variety, introduced by Nakajima \cite{N1,N2},

2913: and since studied by many authors.  In Section \ref{quiver} we give the basic

2914: constructions of quiver varieties and hyperpolygon spaces, and show that they

2915: satisfy all of the hypotheses of Chapter \ref{abelianization}.

2916: Section \ref{modcore} is devoted to understanding the components of the core

2917: of a hyperpolygon space; in particular, we show that they are all smooth

2918: (Theorem \ref{component}), and interpret them as moduli spaces of spatial polygons

2919: with certain properties (Theorem \ref{us}).

2920: Sections \ref{s1} and \ref{last} contain computations of the $\so$-equivariant

2921: cohomology rings of the hyperpolygon space as well as its core components,

2922: making use of the abelianization technique of Chapter \ref{abelianization}.

2923: 

2924: This chapter is taken from \cite{HP2} and \cite{HP}.  The reader is warned

2925: that our notation differs significantly from that of \cite{HP2}; most glaring

2926: is the fact that our spaces $\X$ and $\M$ correspond to the spaces 

2927: $M$ and $X$ (respectively) in \cite{HP2}.  This abrupt switch is necessary to

2928: conform with the conventions of Chapters \ref{analogues}-\ref{abelianization}.

2929: 

2930: \begin{section}{Quiver varieties}\label{quiver}

2931: Let $Q$ be a quiver with vertex set $I$ and edge set $E\subs I\times I$,

2932: where $(i,j)\in E$ means that $Q$ has an arrow pointing from $i$ to $j$.

2933: We assume that

2934: $Q$ is connected and has no oriented cycles.

2935: Suppose given two collections of vector spaces $\{V_i\}$ and $\{W_i\}$,

2936: each indexed by $I$, and consider the affine space

2937: $$\Af = \bigoplus_{(i,j)\in E}\Hom(V_i,V_j)

2938: \,\oplus\,\bigoplus_{i\in I}\Hom(V_i,W_i).$$

2939: The group $U(V) = \prod_{i\in I}U(V_i)$ acts on $\Af$ by conjugation, and

2940: this action is hamiltonian.  Given an element

2941: $$(B,J) = \bigoplus_{(i,j)\in E}B_{ij}\oplus\bigoplus_{i\in I}J_i$$

2942: of $\Af$,

2943: the $\mathfrak{u}(V_i)^*$ component of the moment map

2944: is $$\mu_i(B,J) = J_i^{\dagger}J_i+\sum_{(i,j)\in E}B_{ij}^{\dagger}B_{ij},$$

2945: where $\dagger$ denotes adjoint, and $\mathfrak{u}(V_i)^*$

2946: is identified with with the set of hermitian matrices via

2947: the trace form.

2948: Given a generic central element $\xi\in\mathfrak{u}(V)^*$,

2949: the K\"ahler quotient $\Af\,\mod_{\!\xi}U(V)$ parameterizes 

2950: isomorphism classes of $\xi$-stable,

2951: framed representations

2952: of $Q$ of fixed dimension \cite{N2}.  If $W_i=0$ for all $i$, then

2953: the diagonal circle $U(1)$ in the center of $U(V)$ acts trivially,

2954: and we instead quotient by

2955: $PU(V) = U(V)/U(1)$.

2956: 

2957: Consider the hyperk\"ahler quotient $$\M = T^*\Af\,\mmod_{\!(\xi,0)}U(V),$$

2958: or, if $W_i=0$ for all $i$,

2959: $$\M = T^*\Af\,\mmod_{\!(\xi,0)}PU(V).$$

2960: As in Section~\ref{circleaction}, $\M$ has a natural action 

2961: of $\gm$ induced from scalar

2962: multiplication on the fibers of $T^*\Af$.

2963: We now show that $M=T^*\Af$ satisfies

2964: the hypotheses of Theorems~\ref{main} and \ref{ordinary}.

2965: 

2966: \begin{proposition}

2967: Let $T(V)\subs U(V)$ be a maximal torus, and let $pr:\mathfrak{u}(V)^*\to\mathfrak{t}(V)^*$

2968: be the natural projection.

2969: The moment maps $\mu=\Od_{i\in I}\mu_i:\Af\to\mathfrak{u}(V)^*$

2970: and $pr\circ\mu:\Af\to\mathfrak{t}(V)^*$ are each proper.

2971: \end{proposition}

2972: 

2973: \begin{proof}

2974: To show that $\mu$ and $pr\circ\mu$ is proper, it suffices to find an element

2975: $t\in T(V)\subs U(V)$ such that the weights of the

2976: action of $t$ on $\Af$ are all strictly positive.

2977: Let $\lambda=\{\lambda_i\mid i\in I\}$ be a collection of integers,

2978: and let $t\in T(V)$ be the central element of $U(V)$ that acts

2979: on $V_i$ with weight $\lambda_i$ for all $i$.

2980: Then $t$ acts on $\Hom(V_i,V_j)$ with weight $\lambda_j-\lambda_i$,

2981: and on $\Hom(V_i,W_i)$ with weight $-\lambda_i$.

2982: Hence we have reduced the problem to showing that it is possible

2983: to choose $\lambda$ such that $\lambda_i<0$ for all $i\in I$ and

2984: $\lambda_i<\lambda_j$ for all $(i,j)\in E$.

2985: 

2986: We proceed by induction on the order of $I$.

2987: Since $Q$ has no oriented cycles, there must exist a source $i\in I$;

2988: a vertex such that for all $j\in I$, $(j,i)\notin E$.

2989: Deleting $i$ gives a smaller (possibly disconnected) quiver with

2990: no oriented cycles, and therefore we may choose

2991: $\left\{\lambda_j\mid j\in I\smallsetminus\{i\}

2992: \right\}$ such that $\lambda_j<0$ for all $j\in I\smallsetminus\{i\}$

2993: and $\lambda_j<\lambda_k$ for all $(j,k)\in E$.

2994: We then choose $\lambda_i<\min\left\{\lambda_j\mid j\in I\smallsetminus\{i\}\right\}$,

2995: and we are done.

2996: \end{proof}

2997: 

2998: \vspace{-\baselineskip}

2999: \begin{proposition}

3000: The rationalized Kirwan map

3001: $\Kh_{U(V)}:\widehat{H}^*_{\so\times U(V)}\Big(T^*\Af\Big)\to\hhso(\M)$

3002: is surjective.

3003: \end{proposition}

3004: 

3005: \begin{proof}

3006: Nakajima \cite[\S 7.3]{N2} shows that there exist cohomology classes

3007: $a_i, b_i$ in the image of $\Kh_{U(V)}$

3008: such that $\dso = \sum\pi_1^*a_i\cdot\pi_2^*b_i$.

3009: (Nakajima uses a slightly different circle action, but

3010: his proof is easily adapted to the circle action that we have defined.)

3011: It follows from Proposition~\ref{basis} that

3012: the classes $\{b_i\}$ generate $\hhso(\M)$.

3013: \end{proof}

3014: 

3015: \vspace{-\baselineskip}

3016: \begin{remark}\label{generation}

3017: This Proposition shows that the assumptions of 

3018: Theorems~\ref{integration},~\ref{main},

3019: and~\ref{ordinary} are satisfied for Nakajima's quiver varieties.

3020: Thus integration in equivariant cohomology yields a

3021: description of the rationalized $\so$-equivariant cohomology, and also of

3022: the image of the non-rationalized Kirwan map $\K_G$. Therefore

3023: if we know that $\K_G$ is surjective for a particular quiver variety,

3024: then we have a concrete description of the ($\so$-equivariant)

3025: cohomology ring of that quiver variety.

3026: It is known that $\K_G$ is surjective

3027: for Hilbert schemes of $n$ points on an ALE space,

3028: so our theory applies and gives a description

3029: of the cohomology ring of these quiver varieties.

3030: More examples of quiver varieties with surjective Kirwan map,

3031: including hyperpolygon spaces,

3032: are discussed in Remark \ref{horn}.

3033: \end{remark}

3034: 

3035: \begin{remark}

3036: Another interesting application of Proposition~\ref{basis} is

3037: to the moduli space ${\mathcal M}$ of stable rank $n$ and degree $1$

3038: Higgs bundles  on a genus $g>1$ smooth projective algebraic curve $C$ (see \cite{H2}).

3039: It is an easy exercise to write down the cohomology class 

3040: of the diagonal in ${\mathcal M}\times {\mathcal M}$

3041: as an expression in a certain set of tautological classes.

3042: Proposition~\ref{basis} implies that

3043: the rationalized $\so$-equivariant cohomology ring $\hhso({\mathcal M})$ 

3044: is generated by these classes.

3045: In fact the same result follows from the argument of \cite{HT1}. 

3046: There ${\mathcal M}$ was embedded into a circle compact manifold

3047: ${\mathcal M}_\infty$, whose cohomology is the free algebra on the 

3048: tautological classes. The argument in \cite{HT1} then goes by

3049: showing that the embedding of the $\so$-fixed point set of ${\mathcal M}$ 

3050: in that of ${\mathcal M}_\infty$ induces a surjection

3051: on cohomology. This already implies that $\hhso({\mathcal M}_\infty)$ 

3052: surjects onto $\hhso(\mathcal M)$. In \cite{HT1} it is shown

3053: that in the rank $2$ case this embedding also implies the surjection on 

3054: non-rationalized cohomology, and then

3055: a companion paper \cite{HT2} describes the cohomology ring of ${\mathcal M}$ explicitly. 

3056: However for higher rank

3057: this part of the argument of \cite{HT1} breaks down. 

3058: Later Markman \cite{Mk} used similar diagonal arguments on

3059: certain compactifications of $\mathcal M$ and

3060: Hironaka's celebrated theorem on desingularization of algebraic 

3061: varieties to deduce that the cohomology ring of ${\mathcal M}$

3062: is generated by tautological classes for all $n$.

3063: \end{remark}

3064: 

3065: A {\em hyperpolygon space}, introduced in \cite{K2},

3066: is a quiver variety associated to the

3067: following quiver (Figure \ref{fig:quiver}),

3068: with $V_0 = \C^2$, $V_i=\C^1$ for $i\in\{1,\ldots,n\}$, and $W_i=0$ for all $i$.

3069: % It is so named because, for

3070: % $$\xi = \left(-\half\sum_{i=1}^n\xi_i;\hspace{2pt}\xi_1,\ldots,\xi_n\right)

3071: % \in\mathfrak{pu}(V)^*\subs\mathfrak{u}(2)^*\oplus\mathfrak{u}(1)^n,$$

3072: % the K\"ahler quotient $\Af\mod_{\!\xi}PU(V)\cong(\C^2)^n\mod_{\!\xi}PU(V)$

3073: % parameterizes $n$-sided polygons in $\R^3$

3074: % with edge lengths $\xi_1,\ldots,\xi_n$, up to rotation \cite{HK}.

3075: \begin{figure}[h]

3076: \centerline{\epsfig{figure=polyquiv.eps,height=4cm}}

3077: \caption{The quiver for a hyperpolygon space.}\label{fig:quiver}

3078: \end{figure}

3079: 

3080: Let $$G := PU(V) = \Big(SU(2)\times U(1)^n\Big)\Big/\Z_2,$$

3081: and $$\GC := PGL(V) = \Big(\SL\times (\gm)^n\Big)\Big/\Z_2,$$

3082: where $\Zt$ acts by multiplying each factor by $-1$.

3083: We represent an element of $\Af\cong\ctn$ by an $n$-tuple

3084: of column vectors $$q = (q_1,\ldots,q_n).$$

3085: Following the conventions in \cite{K2}, we consider the {\em right}

3086: action of $G\subs\GC$ on $\Af$ given explicitly by

3087: $$q[\Theta;e_1,\ldots,e_n] = (\Theta^{-1}q_1e_1,\ldots,\Theta^{-1}q_ne_n).$$

3088: The compact group $G$ acts with moment map $\mu:\ctn\to\sutd\oplus\tnd$

3089: given by the equation $$\mu(q) = \sum_{i=1}^n (q_iq_i^*)_0 \oplus 

3090: \left(\half|q_1|^2,\ldots,\half|q_n|^2\right),$$

3091: where $q_i^*$ denotes the conjugate transpose of $q_i$, $(q_iq_i^*)_0$ 

3092: denotes the

3093: traceless part of $q_iq_i^*$, and $\sutd$ is identified with 

3094: $i\cdot\sut$ via the trace form.

3095: Given an $n$-tuple of real numbers $(\a_1,\ldots,\a_n)$, 

3096: we define the {\em polygon space} $$\X(\a) := \ctn\mod_{\!\!\a}G,$$

3097: where $\a = 0\oplus (\a_1,\ldots,\a_n)\in\sutd\oplus\tnd$.

3098: If we break the reduction into two steps, reducing first by $U(1)^n$

3099: and then by $SU(2)$, we find that

3100: \begin{equation}\label{polygon}

3101: \X(\a)\cong\left\{(v_1,\ldots,v_n)\in (\R^3)^n\,\bigg|\, 

3102: \|v_i\|=\a_i\text{ and }\sum v_i=0\right\}\bigg/SO(3)

3103: \end{equation}

3104: (see Remark \ref{interp} and the proof of Theorem \ref{us}).

3105: Here $\sutd$ is being identified with $\R^3$, 

3106: and the coadjoint action of $SU(2)$

3107: on $\sutd$ is being replaced by the standard action of $SO(3)$ on $\R^3$ \cite{HK}.

3108: This space, therefore, may be thought of as the moduli space of $n$-sided polygons in $\R^3$,

3109: with fixed edge lengths $(\a_1,\ldots,\a_n)$, up to rotation.  

3110: In particular, $\X(\a)$ is empty unless $\a_i\geq 0$ for all $i$.

3111: 

3112: We call $\a$ {\em generic} if there does not exist a

3113: % map $\eps:\{1,\ldots,n\}\to\{\pm 1\}$

3114: subset $S\subs\{1,\ldots,n\}$

3115: such that $\sum_{i\in S}\a_i=\sum_{j\in S^c}\a_j$.

3116: Geometrically, this means that there is no element of $\X(\a)$ represented by a polygon that is

3117: contained in a single line in $\R^3$.  If $\a$ is generic, then $\X(\a)$ is smooth \cite{HK}.

3118: Throughout this paper we will assume that $\a$ is generic, and that $\a_i> 0$ for all $i$.

3119: 

3120: To define the hyperk\"ahler analogue of $\X(\a)$, we consider the induced action of $G$

3121: on $\cot$.

3122: % \cong\H^{2n}$, a $(2n)$-dimensional quaternionic "vector space".

3123: Explicitly, we write an element of $T^*\Af$ as $(p,q)$,

3124: where $q = (q_1,\ldots,q_n)$ is an $n$-tuple

3125: of column vectors

3126: and $p = (p_1,\ldots,p_n)$ an $n$-tuple

3127: of row vectors, and we put

3128: $$(p,q)[\Theta;e_1,\ldots,e_n] = 

3129: \big((e_1^{-1}p_1\Theta,\ldots,e_n^{-1}p_n\Theta),(\Theta^{-1}q_1e_1,\ldots,\Theta^{-1}q_ne_n)\big).$$

3130: The action of $G$ on $T^*\Af$ is hyperhamiltonian

3131: with hyperk\"ahler moment map

3132: $$\mur\oplus\muc:\cot\to\Big(\sutd\oplus\tnd\Big)\oplus\Big(\slt^*\oplus

3133: (\mathfrak{u}(1)^n_{\C})^*\Big)$$

3134: given by the equations

3135: \begin{equation*}\label{mur}

3136: \mu_{\R}(p,q)  =  \frac{\sqrt{-1}}{2}\sum_{i=1}^n \left(q_i q_i^* - p_i^* p_i\right)_0

3137: \oplus\left(\half\left(|q_1|^2 - |p_1|^2\right), \ldots, \half\left(|q_n|^2 - |p_n|^2\right)\right) 

3138: \end{equation*}

3139: and

3140: \begin{equation*}\label{muc}

3141: \mu_{\C}\left(p,q\right)  =  - \sum_{i=1}^n \left(q_i p_i\right)_0\oplus \left(\sqrt{-1} p_1 q_1,

3142: \ldots, \sqrt{-1} p_n q_n\right).

3143: \end{equation*}

3144: We then define the hyperpolygon space to be the hyperk\"ahler quotient

3145: $$\M(\a) := \cot\mmod_{(\a,0)}G = 

3146: \Big(\mur^{-1}(\a)\cap\muc^{-1}(0)\Big)\Big/G,$$

3147: a smooth, noncompact hyperk\"ahler manifold of complex dimension $2(n-3)$.

3148: Recall that we also have

3149: $$\X(\a)\cong\left(\ctn\right)^{\ses}\!\big/G_{\C}

3150: \hspace{.7cm}\text{ and }\hspace{.7cm}

3151: \M(\a)\cong\muc^{-1}(0)^{\ses}\big/G_{\C},$$

3152: where $\ses$ means stable with respect to the weight $\a$ in the sense of

3153: geometric invariant theory (see Section \ref{reduction}).\footnote{Recall

3154: from Theorem \ref{hkhs} that the notions of stability and semistability

3155: agree for generic $\a$.}

3156: Nakajima gives a stability criterion for general quiver varieties \cite{N1,N2},

3157: which Konno interprets in the special case of hyperpolygon spaces.

3158: Call a subset $S\subs\{1,\ldots,n\}$ {\em short} if $\sum_{i\in S}\a_i < \sum_{j\in S^c}\a_j$,

3159: and call it {\em long} if its complement is short.  (Assuming that $\a$ is generic is equivalent to assuming

3160: that every subset is either short or long.)

3161: Given a point $(p,q)\in\cot$ and a subset $S\subs\{1,\ldots,n\}$,

3162: we will say that $S$ is {\em straight} 

3163: in $(p,q)$ if $q_i$ is proportional to $q_j$ for every $i,j\in S$.

3164: The terminology comes from K\"ahler polygon spaces, in which this condition

3165: is equivalent to asking that the vectors $v_i$ and $v_j$ be proportional

3166: over $\R_+$, or that the edges of lengths $\a_i$ and $\a_j$ (if they happen to be adjacent)

3167: line up to make a single edge of length $\a_i+\a_j$,

3168: as in Figure \ref{straight}.

3169: \begin{figure}[h]

3170: \begin{center}

3171: \psfrag{a1}{$\a_1$}

3172: \psfrag{a2}{$\a_2$}

3173: \psfrag{a3}{$\a_3$}

3174: \psfrag{a4}{$\a_4$}

3175: \psfrag{a5}{$\a_5$}

3176: \psfrag{a6}{$\a_6$}

3177: \psfrag{a7}{$\a_7$}

3178: \psfrag{a8}{$\a_8$}

3179: \includegraphics[height=40mm]{straight.eps}

3180: \caption{The subset $\{1,2,3\}$ is straight.}\label{straight}

3181: \end{center}\end{figure}

3182: 

3183: \begin{theorem}\label{stability}{\em\cite[4.2]{K2}}

3184: Suppose that $\a$ is generic, and $\a_i> 0$ for all $i$.

3185: Then a point $(p,q)\in\cot$ is stable with respect to $\a$

3186: if and only if the following two conditions are satisfied:

3187: \begin{eqnarray*}

3188: &1)& q_i\neq 0 \text{ for all }i,\text{ and}\\

3189: &2)& \text{if $S$ is straight and $p_j=0$ for all $j\in S^c$, then $S$ is short.}

3190: \end{eqnarray*}

3191: \end{theorem}

3192: As in Chapter \ref{analogues}, we will use the notation $[p,q]$ to denote

3193: the $\GC$-equivalence class of a point $(p,q)\in\muc^{-1}(0)^{\ses}$,

3194: and $[p,q]_{\R}$ to denote

3195: the $G$-equivalence class of a point $(p,q)\in\mur^{-1}(\a)\cap\muc^{-1}(0)$.

3196: Recall that $\X$ sits inside of $\M$ 

3197: as the locus of points $[p,q]$ with $p=0$.

3198: This observation, along with Theorem \ref{stability}, allows us to recover

3199: the stability condition for the action of $G$ on $\ctn$.

3200: A point $q\in\ctn$ is stable if and only if $q_i\neq 0$ for all $i$,

3201: and no long subset is straight, as first shown in \cite{Kl}.

3202: The polygonally-minded reader is warned that in the hyperpolygon space $\M(\a)$,

3203: long subsets {\em can} be straight.

3204: \end{section}

3205: 

3206: \begin{section}{Moduli theoretic interpretation of the core}\label{modcore}

3207: For the rest of the section we fix a generic 

3208: $\a = 0\oplus (\a_1,\ldots,\a_n)\in\sutd\oplus\tnd$,

3209: with $\a_i>0$ for all $i$, and write $\M=\M(\a)$, $\X=\X(\a)$.

3210: % Consider the action of $\C^*$ on $\M$ given in the complex description

3211: % by $\la\cdot[p,q]=[\la p, q]$.  

3212: % The circle $S^1\subs\C^*$ preserves

3213: % $\omega_{\R}$, and acts with moment map $\Phi:\M\to\R$ given in the symplectic

3214: % description by $\Phi\left([p,q]_{\R}\right) = \half\sum |p_i|^2$.

3215: Following Konno, we define

3216: $$\mathcal{S} = \big\{S\subs\{1,\ldots,n\}\hs\hs\big{|}\hs\hs S\text{ is short}\big\}$$

3217: and $$\spa = \big\{S\in\mathcal{S}\hs\hs\big{|}\hs\hs |S|\geq 2\big\}.$$

3218: 

3219: \begin{theorem}\label{fixed}{\em\cite{K2}}

3220: The fixed point set 

3221: $\displaystyle{\M^{\gm} = \M^{S^1} = \pola\cup\bigcup_{S\in\spa}\M_S,}$

3222: where

3223: $$\M_S = \big\{[p,q]\hs\hs\big{|}\hs\hs S\text{ and }S^c\text{ are each straight, and }

3224: p_j = 0\text{ for all }j\in S^c\big\}.$$  Furthermore, $\M_S$

3225: is diffeomorphic to $\C P^{|S|-2}$.

3226: \end{theorem}

3227: 

3228: For all $S\in\spa$, let

3229: $U_S = U_{\polsa}$ be the piece of the core $\core\subs\M$

3230: defined in Section \ref{circleaction}.

3231: {\em A priori} we know only that $U_S$ is an irreducible, isotropic

3232: subvariety of dimension at most $n-3$ (Proposition \ref{coreprops}).

3233: 

3234: \begin{theorem}\label{component}

3235: The core component $U_S$ is smooth of complex

3236: dimension $n-3$, and we have

3237: $$U_S = \big\{[p,q]\bigmid S\text{ is straight, and }

3238: p_j = 0\text{ for all }j\in S^c\big\}.$$

3239: \end{theorem}

3240: 

3241: Before proving Theorem \ref{component}, 

3242: we describe the way in which the various components

3243: of the core fit together.

3244: For all $S\in\spa$, let $$\polsa = U_S\cap \pola =

3245: \big\{[0,q]\hs\hs\big{|}\hs\hs S\text{ is straight}\big\}.$$

3246: We call this space the {\em polygon subspace} of $\pola$ 

3247: corresponding to the short subset $S$.

3248: Note that $\polsa$ is itself a polygon space with $n-|S|+1$ edges, of lengths

3249: $\{\a_j\mid j\in S^c\}\cup\{\sum_S\a_i\}$.  In particular, it is smooth.

3250: Now suppose given two short subsets $S,T\in\spa$, 

3251: and consider the intersection $U_S\cap U_T$.

3252: \begin{itemize}

3253: \item If $S\cap T=\emptyset$, then $U_S\cap U_T = \polsa\cap X_T$, a polygon subspace

3254: both of $\polsa$ and of $X_T$.

3255: \item If $S\cap T\neq\emptyset$

3256: and $S\cup T$ is long, then $U_S\cap U_T=\emptyset$.

3257: \item If $S\cap T\neq\emptyset$ and 

3258: $S\cup T$ is short,

3259: then $$U_S\cap U_T = \big\{[p,q]\bigmid S\cup T\text{ is straight, and }p_j=0\text{ for all }

3260: j\in (S\cap T)^c\big\}.$$

3261: This is a subvariety of $U_{S\cup T}$ given by taking the closure inside of $U_{S\cup T}$

3262: of a certain subbundle of the conormal bundle to $X_{S\cup T}\subs\pola$,

3263: defined by setting $p_j=0$ for all $j\in (S\cap T)^c\supseteq (S\cup T)^c$.

3264: \end{itemize}

3265: Each of these descriptions generalizes to higher intersections

3266: without modification.

3267: 

3268: Finally, we compute the fixed point set $U_S^{\gm}$.

3269: If $[p,q]\in U_S^{\gm}$, then either $p=0$ and $[p,q]\in\polsa$,

3270: or $[p,q]\in \X_T$ for some $T\in\spa$.  If $[p,q]\in \X_T$ then Theorem \ref{fixed}

3271: tells us that $T$ and $T^c$ are each straight, hence $S\subs T$ or $S\subs T^c$.

3272: Since $p\neq 0$, we must have $S\subs T$.  Indeed, $U_S\cap \X_T$ is the linear subspace

3273: of $\X_T\cong \C P^{|T|-2}$ given by the condition $p_j=0$ for all $j\in T\smallsetminus S$.

3274: In particular, $U_S\cap \X_T$ is isomorphic to $\C P^{|S|-2}$ for any $T\supseteq S$.

3275: 

3276: \begin{example}\label{ooh}

3277: Let $n=5$, $\a_1=\a_2=1$, and $\a_3=\a_4=\a_5=3$, and consider the short subset

3278: $S=\{1,2\}$.  The fixed point set of $U_S$ consists of $\polsa\cong\C P^1$,

3279: and four points $\X_S$, $U_S\cap \X_{T_3}$, 

3280: $U_S\cap \X_{T_4}$, and $U_S\cap \X_{T_5}$,

3281: where $T_j = \{1,2,j\}$ for $j=3,4,5.$  

3282: For each $j$, $U_S\cap U_{T_j}$ is isomorphic to $\C P^1$,

3283: and touches $\polsa$ at the point $X_{T_j}$.

3284: In the following picture, an ellipse represents a copy of $\C P^1$ flowing between two

3285: fixed points, where the numbers or pairs of numbers

3286: indicate subsets that are straight on this $\C P^1$.  (For example, 

3287: $12,45$ means that $1$ and $2$ are straight,

3288: as are $4$ and $5$.)

3289: We will revisit this example at the end of Section \ref{last}.

3290: \begin{figure}[h]

3291: \begin{center}

3292: \psfrag{Xs}{$\M_S$}

3293: \psfrag{Xt3}{$\M_{T_3}$}

3294: \psfrag{Xt4}{$\M_{T_4}$}

3295: \psfrag{Xt5}{$\M_{T_5}$}

3296: \psfrag{Ms(a)}{$\X_{S}$}

3297: \psfrag{Mt3(a)}{$\X_{T_3}$}

3298: \psfrag{Mt4(a)}{$\X_{T_4}$}

3299: \psfrag{Mt5(a)}{$\X_{T_5}$}

3300: \includegraphics[height=40mm]{comp.eps}

3301: \caption{$U_S$, with $S=\{1,2\}$}

3302: \label{surface}

3303: \end{center}

3304: \end{figure}

3305: \end{example}

3306: 

3307: \begin{componentproof}

3308: % The fact that $\dim U_S = \half\dim \M = n-3$ is a general property of

3309: % core components of quiver varieties \cite{N1}.  

3310: % Thus, by a dimension count, it is enough to show that the set

3311: % $\big\{[p,q]\bigmid S\text{ is straight, and }

3312: % p_j = 0\text{ for all }j\in S^c\big\}$ is contained in $U_S$.

3313: % 

3314: Consider a point $[p,q]\in \M$ with $S$ straight, and $p_j=0$ for all $j\in S^c$.

3315: By applying an element of $G$, we may assume that $q_i = \binom{1}{0}$ for all $i\in S$.

3316: Suppose further that there exists an $i\in S$ with $p_i\neq 0$, and that

3317: no strict superset of $S$ is straight.  In other words, if 

3318: $q_j = \binom{a_j}{b_j}$ for $j\in S^c$,

3319: suppose that $b_j\neq 0$.

3320: For $t\in\gm$, let $\Theta(t) = 

3321: \footnotesize{\Big(\begin{array}{cc}

3322: t&0\\

3323: 0&t^{-1}

3324: \end{array}

3325: \Big)}$, 

3326: let $e_i(t)=t$ for all $i\in S$,

3327: and let $e_j(t)=t^{-1}$ for all $j\in S^c$.

3328: Then for $i\in S$, we have $e_i(t)^{-1}p_i \Theta(t) = t^{-2} p_i$ and 

3329: $\Theta(t)^{-1}q_i e_i = q_i$.

3330: For $j\in S^c$, we have $\Theta(t)^{-1}q_j e_j = \binom{t^{-2}a_j}{b_j}$.

3331: Hence

3332: \begin{eqnarray*}

3333: \lim_{t\to\infty}t\cdot[p,q] &=& \lim_{t\to\infty}t^2\cdot[p,q]\\

3334: &=& \lim_{t\to\infty}[t^2 p,q]\\

3335: &=& \lim_{t\to\infty}[t^2 e(t)^{-1}p \Theta(t), \Theta(t)^{-1}q e(t)]\\

3336: &=& [p,q'],

3337: \end{eqnarray*}

3338: where $q'_i = q_i$ for $i\in S$, and $q'_j = \binom{0}{b_j}$ for $j\in S^c$.

3339: Since we have assumed that $b_j\neq 0$ for all $j\in S^c$ and that $p_i\neq 0$

3340: for some $i\in S$, $(p,q')$ is stable, and hence defines an element of $\M_S$.

3341: Since $U_S$ is defined to be the closure of the set of elements that flow up to $\M_S$,

3342: it includes all $[p,q]$ with $S$ straight and $p_j=0$ for all $j\in S^c$.

3343: By dimension count, this containment is an equality, and

3344: we have $\dim U_S = n-3$.

3345: 

3346: To see that $U_S$ is smooth, it is sufficient to show that $U_S$ is smooth

3347: at $[p,q]$ for all $[p,q]\in \M^{\gm}$.

3348: First suppose that $[p,q]\in \M_T$ for some $T\in\spa$ containing $S$.

3349: Suppose, without loss of generality, that

3350: $T=\{1,\ldots,l\}$ and $S=\{1,\ldots,m\}$ for some $l\leq m$.

3351: Konno computes an explicit local complex chart for $\M$ at the point $[p,q]$,

3352: with coordinates $\{z_i, w_i\mid 3\leq i\leq n-1\}$ \cite{K2}.

3353: With respect to these coordinates, a point $[p',q']$ has the property that

3354: $S$ is straight and $p'_j=0$ for all $j\in S^c$ if and only if 

3355: $w_i=0$ for all $3\leq i\leq l$

3356: and $z_j=0$ for all $l+1 \leq j\leq n-1$.

3357: Hence $U_S$ is smooth at $[p,q]$.

3358: 

3359: It remains only to show that $U_S$ is smooth at $\polsa = U_S\cap\pola$.

3360: Let $$E = \{(p,q)\mid S\text{ is straight, }

3361: p_j = 0\text{ for all }j\in S^c,\text{ and }\mu_{\C}(p,q)=0\},$$ and let 

3362: $N = \{(p,q)\in E\mid p=0\}$.  The natural projection from $E$ to $N$

3363: exhibits $E$ as a vector bundle over $N$, because the equation $\mu_{\C}(p,q)=0$

3364: is linear in $p$.  We have $U_S = E\mod G = E^{\ses}/G$, and 

3365: $\polsa = N\mod G = N^{\ses}/G$.  

3366: The set $E\vert_{N^{\ses}}/G\subs E^{\ses}/G$ is

3367: an open neighborhood of $\polsa$ inside of $U_S$, and is isomorphic to a vector

3368: bundle over $\polsa$.

3369: Since $\polsa$ is a polygon space it is smooth, hence $U_S$ is smooth 

3370: in a neighborhood of $\polsa$.

3371: \end{componentproof}

3372: 

3373: \vspace{-\baselineskip}

3374: \begin{corollary}\label{compactification}

3375: $U_S$ is a compactification of the conormal

3376: bundle to $\polsa$ in $\pola$.

3377: \end{corollary}

3378: 

3379: \begin{proof}

3380: Choose a point $[q,0]\in\polsa$, and a decomposition

3381: $$T_{[q,0]}\M = \nu_1\oplus \nu_2\oplus T_{[q,0]}\polsa\oplus E,$$

3382: where $\nu_1$ is the normal space to $\polsa$ inside of $U_S$,

3383: and $\nu_2$ is the normal space to $\polsa$ inside of $\pola$.

3384: Proposition \ref{coreprops} tells us that $U_S$ and $\pola$

3385: are both $\omega_{\C}$-lagrangian submanifolds of $\M$,

3386: hence $\omega_{\C}$ gives a perfect pairing between $T_{[q,0]}\polsa$

3387: and $E$.  It follows that $\omega_{\C}$ also gives a perfect pairing between

3388: $\nu_1$ and $\nu_2$, and therefore that the normal bundle to

3389: $\polsa$ inside of $U_S$ is dual to the normal bundle of $\polsa$

3390: inside of $\pola$.

3391: % We must show that the normal bundle to $\polsa$ in $U_S$

3392: % is dual to the normal bundle to $\polsa$ in $\pola$.

3393: % We use only general facts about quiver varieties from \cite{N1},

3394: % and the additional information that $U_S$ is smooth,

3395: % from Theorem \ref{component}.

3396: % Consider a point $[0,q]\in\polsa$, and let $H_0$ and $H_1$ be the

3397: % $0$ and $1$ weight spaces of the $\C^*$ action on $T_{[0,q]}\M$.  

3398: % The holomorphic symplectic

3399: % form $\omega_{\C}$ is being rotated by $\C^*$ with weight $1$ \cite[5.1]{N1},

3400: % hence it defines a perfect pairing between $H_0$ and $H_1$.

3401: % The fiber at $[0,q]$ of the normal bundle to $\polsa$ in $\pola$

3402: % is $H_0/T_{[0,q]}\polsa$, which is dual by $\omega_{\C}$ to the annihilator

3403: % of $T_{[0,q]}\polsa$ inside of $H_1$.

3404: % Since $U_S$ is $\C^*$-invariant, we may write 

3405: % $$T_{[0,q]}U_S = T_{[0,q]}U_S\cap H_0\oplus T_{[0,q]}U_S\cap H_1

3406: % = T_{[0,q]}\polsa\oplus T_{[0,q]}U_S\cap H_1.$$

3407: % To prove Corollary \ref{compactification}, we must show that

3408: % $T_{[0,q]}U_S\cap H_1$ is equal to the annihilator of $T_{[0,q]}\polsa$.

3409: % The fact that $T_{[0,q]}U_S\cap H_1$

3410: % is contained in the annihilator of $T_{[0,q]}\polsa$ follows from the

3411: % fact that $U_S$ is lagrangian with respect to $\omega_{\C}$ \cite{N1} 

3412: % (here we use

3413: % smoothness of $U_S$ at $[0,q]$).

3414: % Equality is then obtained by dimension count.

3415: \end{proof}

3416: 

3417: \vspace{-\baselineskip}

3418: \begin{remark}

3419: This argument generalizes to the smooth intersection of any two

3420: lagrangian submanifolds of a symplectic manifold.

3421: \end{remark}

3422: 

3423: We next describe $U_S$ in polygon-theoretic terms, as a certain moduli

3424: space of pairs of polygons in $\R^3$.

3425: % satisfying certain conditions,

3426: % up to rotation.

3427: 

3428: \begin{theorem}\label{us}

3429: The core component $U_S$ is 

3430: homeomorphic to the moduli space

3431: of $n+1$ vectors $$\{u_i, v_j, w\in\R^3\mid i\in S, j\in S^c\},$$

3432: taken up to rotation, satisfying the following conditions:

3433: \begin{eqnarray*}

3434: &1)&\hs\hs w + \sum_{j\in S^c}v_j = 0\\

3435: &2)&\hs\hs \sum_{i\in S}u_i = 0\\

3436: &3)&\hs\hs u_i\cdot w = 0\hs\hs\text{ for all }i\in S\\

3437: &4)&\hs\hs \|v_j\| = \a_j\hs\hs\text{ for all }j\in S^c\\

3438: &5)&\hs\hs \|w\| = \sum_{i\in S}\sqrt{\a_i^2 + \|u_i\|^2}.

3439: \end{eqnarray*}

3440: \end{theorem}

3441: 

3442: \begin{remark}\label{interp}

3443: In more descriptive terms, a point in $U_S$ specifies two polygons in $\R^3$,

3444: as in Figure~\ref{fig:Uspoly}.

3445: \begin{figure}[h]

3446: \begin{center}

3447: \psfrag{w}{$w$}

3448: \psfrag{v1}{$v_1$}

3449: \psfrag{v2}{$v_2$}

3450: \psfrag{v3}{$v_3$}

3451: \psfrag{vnS}{$v_{n-|S|}$}

3452: \psfrag{u1}{$u_1$}

3453: \psfrag{u2}{$u_2$}

3454: \psfrag{uS}{$u_{|S|}$}

3455: \includegraphics[height=60mm]{Uspoly.eps}

3456: \caption{

3457: % The two polygons specified by a point in $U_S$. On the K\"ahler

3458: % polygon space at the minimum of $\Phi$ on $U_S$, the vectors $u_i$ are all

3459: % zero, and the vector $w$ corresponds to the $q_i$ for $i\in S$, which

3460: % are straight in $X_S$. At the maximum of $\Phi$ in $U_S$, the $v_i$

3461: % all lie on the same line as $w$, perpendicular to the $u_i$, making a

3462: % degenerate (linear) polygon. Then the only data left is given by the

3463: % second polygon specified by the $u_i$.

3464: An element of $U_S$, represented by a spatial polygon with a distinguished

3465: edge, and a planar polygon perpendicular to that edge.}

3466: \label{fig:Uspoly}

3467: \end{center}

3468: \end{figure}

3469: The first is the $n - |S| + 1$ sided polygon consisting of the vectors

3470: $\{v_j\mid j\in S^c\}$ and $w$.  Each vector $v_j$ has length $\a_j$, and

3471: $w$ has a variable length, always greater than or equal to $\sum_{i\in S}\a_i$.

3472: This variable length is determined by the lengths of the edges in the second polygon, 

3473: which consists of $|S|$ vectors $\{u_i\mid i\in S\}$, 

3474: all contained in the plane perpendicular to $w$.

3475: Note that this description also applies to the K\"ahler polygon space $\X$ 

3476: by taking $S=\emptyset$.

3477: 

3478: By setting $u_i = 0$ for all $i$ we get $\polsa$,

3479: the minimum of the Morse-Bott function $\Phi$ on $U_S$.

3480: On the other hand, consider the submanifold of $U_S$ obtained by imposing

3481: the extra condition that $\|w\| = \sum_{j\in S^c}\|v_j\|$.

3482: Then the first of the two polygons is forced to be linear,

3483: and we are left with $|S|$ vectors $\{u_i\}$ in the perpendicular plane

3484: satisfying a certain norm condition

3485: and adding to zero. 

3486: Identifying this plane with $\C$ and dividing by the circle action rotating this plane, 

3487: we obtain $\C P^{|S|-2}\cong \M_S$,

3488: the maximum of $\Phi$ on $U_S$.

3489: Other critical points of $\Phi$ occur whenever the first polygon is linear,

3490: which is possible for finitely many values of $\|w\|$.

3491: \end{remark}

3492: 

3493: \begin{usproof}

3494: Suppose given a point $[p,q]_{\R}\in U_S$,

3495: and let

3496: $$u_i = q_ip_i + p_i^*q_i^*\hs\hs\text{ for all }i\in S,$$

3497: $$v_j = (q_j q_j^*)_0\hs\hs\text{ for all }j\in S^c,$$

3498: $$w = \sum_{i\in S}(q_iq_i^*)_0-(p_i^*p_i)_0.$$

3499: These vectors live in $i\cdot\sut\cong\sut^*\cong\R^3$, which is endowed with the

3500: metric

3501: $u\cdot v = \frac{1}{2}\operatorname{tr}uv$,

3502: invariant under the coadjoint action.

3503: With respect to this metric, we have the equalities

3504: $\|(qq^*)_0\| = \half |q|^2$ and $\|(p^*p)_0\| = \half |p|^2$,

3505: hence conditions (1), (2), and (4) are immediate consequences 

3506: of the moment map equations.

3507: 

3508: To verify condition (3), note that the vectors $\{q_i\mid i\in S\}$

3509: are all proportional over $\C$, which implies that the vectors $(q_iq_i^*)_0$

3510: are positive scalar multiples of each other.

3511: Furthermore, the moment map equation $p_iq_i=0$ implies that $(p_i^*p_i)_0$

3512: is a non-positive scalar multiple of $(q_iq_i^*)_0$, therefore

3513: $w = \sum(q_iq_i^*)_0-(p_i^*p_i)_0$ is proportional over $\R_+$ to

3514: $(q_iq_i^*)_0$ for any $i\in S$.

3515: Then $u_i\cdot w = \half\tr u_i w$ is a multiple of

3516: $$\tr u_i(q_iq_i^*)_0 = \tr u_iq_iq_i^* = \tr p_i^*q_i^*q_iq_i^*

3517: = |q_i|^2\tr p_i^*q_i^* = 0,$$

3518: where the first equality comes from the fact that

3519: $q_iq_i^*-(q_iq_i^*)_0$ is a scalar multiple of the identity,

3520: and $\tr u_i = 0$.

3521: 

3522: To check condition (5), we first compute the norm of $u_i$:

3523: \begin{eqnarray*}

3524: \|u_i\|^2 &=& \half\tr u_i^2\\

3525: &=& \half\tr(q_ip_ip_i^*q_i^* + p_i^*q_i^*q_ip_i)\\

3526: &=& |q_i|^2|p_i|^2\\

3527: &=& |q_i|^2 (|q_i|^2 - 2\a_i).

3528: \end{eqnarray*}

3529: Since all of the vectors

3530: $\{(q_iq_i^*)_0 -(p_i^*p_i)_0\mid i\in S\}$ point in the same direction,

3531: we have $$\|w\| = \sum_{i\in S}\|(q_iq_i^*)_0\| + \|(p_i^*p_i)_0\|

3532: = \sum_{i\in S} \half|q_i|^2 + \half|p_i|^2

3533: = \sum_{i\in S} |q_i|^2 - \a_i

3534: = \sum_{i\in S} \sqrt{\a_i^2 + \|u_i\|^2}.$$

3535: 

3536: We have defined a continuous map from $U_S$ to the moduli space of

3537: vectors $\{u_i,v_j,w\}$ satisfying conditions (1)-(5),

3538: and we claim that this map is a homeomorphism.

3539: Since the source of this map is compact and the target is Hausdorff,

3540: it is sufficient to show that the map is bijective.

3541: 

3542: Suppose given a collection of vectors $\{u_i,v_j,w\}\subs\sut$ 

3543: satisfying conditions (1)-(5).  Using the adjoint action of $SU(2)$,

3544: we may assume that $w$ is a positive scalar multiple of

3545: $\footnotesize{\Big(\begin{array}{cc}

3546: 1&0\\

3547: 0& -1

3548: \end{array}

3549: \Big)}$.

3550: By condition (3), this implies that for all $i\in S$, there exists $t_i\in\C$

3551: with $u_i =

3552: \footnotesize{\Big(\begin{array}{cc}

3553: 0& t_i\\

3554: \bar{t_i}& 0

3555: \end{array}

3556: \Big)}$. 

3557: For $j\in S^c$, we choose $q_j\in\C^2$ with $(q_jq_j^*)_0=v_j$,

3558: and observe that $q_j$ is unique up to the action of $U(1)^n$.

3559: We know that for all $i\in S$, $(q_iq_i^*)_0$ must be a positive multiple of $w$,

3560: hence there exist $a_i,b_i\in\C$ such that

3561: $$q_i=\binom{a_i}{0}\hs\text{ and }\hs p_i = (0\hs b_i)$$

3562: for all $i\in S$.

3563: In order to have $u_i = q_ip_i + p_i^*q_i^*$ and $\half|q_i|^2-\half|p_i|^2=\a_i$, 

3564: we must have $$a_i b_i = t_i\hs\text{ and }\hs 

3565: \half|a_i|^2-\half|b_i|^2=\a_i.$$

3566: These equations uniquely define $a_i$ and $b_i$ up to the action of $U(1)^n$.

3567: It follows from conditions (1)-(5) that 

3568: $(p,q)\in\mur^{-1}(\a)\cap\muc^{-1}(0)$

3569: and that $w=\sum_{i\in S}(q_iq_i^*)_0-(p_i^*p_i)_0$. 

3570: This shows that our map is bijective, and thus completes the proof of Theorem \ref{us}.

3571: \end{usproof}

3572: 

3573: \vspace{-\baselineskip}

3574: \begin{remark}

3575: Suppose that $S$ has only two elements; without loss of generality

3576: we will assume that $S=\{1,2\}$.  Then forgetting $u_1$ and $u_2$

3577: gives a diffeomorphism from $U_S$ to the ``vertical polygon space''

3578: $V\!P(\a_3,\ldots,\a_n,\a_1+\a_2)$

3579: defined in \cite{HK}, shown to be diffeomorphic to a toric variety.

3580: More generally with $S=\{1,\ldots,k\}$, given any two-element subset $T\subs S$, the subvariety

3581: of $U_S$ given by the equations $u_i=0$ for all $i\in S\smallsetminus T$

3582: is diffeomorphic to $V\!P(\a_{k+1},\ldots,\a_n,\sum_T\a_i)$.

3583: \end{remark}

3584: \end{section}

3585: 

3586: \begin{section}{Cohomology rings}\label{s1}

3587: In this section we use Theorem \ref{ordinary} to compute the circle-equivariant

3588: cohomology of a hyperpolygon space $\M$, thus reproducing (by different means)

3589: the results of \cite[\S 3]{HP2}.

3590: Recall that we have $$\M = T^*\ctn\mmod G,$$ where $G$ is a quotient

3591: of $U(1)^n\times SU(2)$ by $\Zt$.

3592: We will simplify our computations by dividing first by

3593: the torus $U(1)^n$.

3594: We have

3595: \begin{eqnarray*}

3596: \M &=& \left(T^*\C^{2n}\right)\bigmmod G\\

3597: &\cong&\left(\left(T^*\C^2\right)^n\bigmmod U(1)^n\right)\bigmmod SU(2)\\

3598: &\cong& \prod_{i=1}^nT^*\C P^1\bigmmod SU(2),

3599: \end{eqnarray*}

3600: where the action of $SU(2)$ on each copy of $T^*\C P^1$

3601: is induced by the rotation action on $\C P^1\cong S^2$.

3602: 

3603: \begin{proposition}\label{hpsurj}

3604: The non-rationalized Kirwan map

3605: $\K_{U(V)}:H^*_{\so\times U(V)}(T^*\C^{2n})\to\hso(\M)$ is surjective.

3606: \end{proposition}

3607: 

3608: \begin{proof}

3609: The map $\K_{U(V)}$ factors as a composition

3610: $$H^*_{\so\times U(V)}(T^*\C^{2n})

3611: \to H^*_{\so\times SU(2)}\left(\prod_{i=1}^nT^*\C P^1\right)

3612: \overset{\K_{SU(2)}}\Longrightarrow \hso(\M),$$

3613: where the first map is the Kirwan map for a toric hyperk\"ahler variety,

3614: and therefore surjective by \cite{HP1}.  Hence it suffices to show that

3615: $\K_{SU(2)}$ is surjective.

3616: 

3617: The level set $\muc^{-1}(0)$ for the action

3618: of $SU(2)$ on $\prod_{i=1}^nT^*\C P^1$ is a subbundle of the cotangent bundle,

3619: given by requiring the $n$ cotangent vectors to add to zero after being restricted

3620: to the diagonal $\C P^1$.  In particular this set is smooth,

3621: and its $\so\times SU(2)$-equivariant cohomology ring is canonically

3622: isomorphic to that of $\prod_{i=1}^nT^*\C P^1$.

3623: Hence $\K_{SU(2)}$ factors as

3624: $$H^*_{\so\times SU(2)}\left(\prod_{i=1}^nT^*\C P^1\right)

3625: \cong H^*_{\so\times SU(2)}\Big(\muc^{-1}(0)\Big)\to

3626: \hso\Big(\muc^{-1}(0)\bigmod SU(2)\Big)

3627: \cong \hso(\M),$$ where the map in the middle is the K\"ahler

3628: Kirwan map.

3629: Surjectivity of this map follows from the following more general lemma,

3630: applied to the manifold $\muc^{-1}(0)$.

3631: 

3632: \begin{lemma}\label{cut}

3633: Suppose given a hamiltonian action of $\so\times G$ on a symplectic

3634: manifold $M$, such that the $\so$ component of the moment map is proper

3635: and bounded below with finitely many critical values.

3636: Then the K\"ahler Kirwan map

3637: $\K:H^*_{\so\times G}(M)\to\hso(M\mod G)$ is surjective.

3638: \end{lemma}

3639: 

3640: \begin{proof}

3641: Extend the action of $S^1$ to an action on $M\times\C$ by letting

3642: $S^1$ act only on the left-hand factor.  On the other hand,

3643: consider a second copy of the circle, which we will call $\mathbb T$ to avoid confusion,

3644: acting diagonally on $M\times\C$.

3645: Choose $r\in \operatorname{Lie}(\mathbb T)^*\cong\R$ greater than

3646: the largest critical value of the $\mathbb T$-moment map, and consider

3647: the space $$Cut(M\mod G):=\left(M\times\C\right)\mod_{\! r} \mathbb T\times G

3648: \cong\big((M\mod G)\times\C\big)\mod_{\! r} \mathbb T.$$

3649: This space, which is called the {\em symplectic cut} of $M\mod G$ \cite{Le},

3650: is an $S^1$-equivariant (orbifold) compactification of $M\mod G$.

3651: We then have a commutative diagram

3652: $$\begin{CD}

3653: H^*_{\so\times G\times \mathbb T}(M\times\C)        @>>> H^*_{\so\times G}(M)\\

3654: @VVV @VV\K V\\

3655: \hso(Cut(M\mod G))        @>>> \hso(M\mod G).

3656: \end{CD}$$

3657: The vertical map on the left is surjective because the $G\times \mathbb T$ moment map is proper,

3658: and the map on the bottom is surjective because the long

3659: exact sequence in cohomology for $M\mod G\subs Cut(M\mod G)$

3660: splits naturally, hence $\K$ is surjective as well.

3661: \end{proof}

3662: 

3663: \noindent By applying Lemma~\ref{cut} to $M = \muc^{-1}(0)$,

3664: this completes the proof of Proposition~\ref{hpsurj}.

3665: \end{proof}

3666: 

3667: \vspace{-\baselineskip}

3668: \begin{remark}\label{horn}

3669: The argument in Proposition \ref{hpsurj} generalizes immediately

3670: to show that the hyperk\"ahler Kirwan map for the quotient

3671: $$\left(\prod_{i=1}^nT^*Flag(\C^k)\right)\bigmmod SU(k)$$

3672: is surjective.  This is itself a quiver variety, and like the

3673: hyperpolygon space, it has a moduli theoretic interpretation.

3674: The K\"ahler quotient

3675: $$\left(\prod_{i=1}^nFlag(\C^k)\right)\bigmod SU(k)$$

3676: is isomorphic to the space of $n$-tuples of $k\times k$

3677: hermitian matrices with fixed eigenvalues adding to zero, modulo

3678: conjugation.

3679: This space has been studied by many authors.  The classical

3680: problem, due to Horn, of determining the values of

3681: the moment map for which it is

3682: nonempty, has only recently been solved \cite{KT}.  For a survey, see \cite{Fu}.

3683: \end{remark}

3684: 

3685: To compute the kernel of the hyperk\"ahler Kirwan map for the

3686: hyperpolygon space, we first need to study the abelian quotient

3687: $$\NN := \prod_{i=1}^nT^*\C P^1\bigmmod T,$$ where $T\cong U(1)\subs SU(2)$ is a

3688: maximal torus.\footnote{This is the hyperk\"ahler analogue of the {\em abelian

3689: polygon space} from \cite{HK}.}

3690: The space $\prod_{i=1}^nT^*\C P^1$ is a hypertoric variety

3691: given by an arrangement of $2n$ hyperplanes in $\R^n$,

3692: where the $(2i-1)^{\text{st}}$ and $(2i)^{\text{th}}$ hyperplanes

3693: are given by the equations $x_i = \pm\xi_i$.

3694: Taking the hyperk\"ahler quotient by $T$ corresponds on the level of arrangements

3695: to restricting this arrangement to the hyperplane

3696: $\{x\in\R^n\mid\sum x_i = 0\}$.

3697: By Theorem \ref{hsm}, we have

3698: $$\hso\left(\NN\right)\cong

3699: \Q[a_1,b_1,\ldots,a_n,b_n,\delta,x]\Big/

3700: \Big\la a_i-b_i-\delta, \hs a_ib_i\hs\Big{|}\hs i\leq n\Big\ra+

3701: \Big\la A_S, B_S\hs\Big{|}\hs S\text{   short}\Big\ra,

3702: $$

3703: % where $$\mathcal{I} = \left\la (a_i-b_i)-(a_j-b_j)\hs\hs\Big{|}\hs\hs i,j\leq n\right\ra,$$

3704: % $$\mathcal{M} = \left\la \prod_{i\in S}(x-a_i)\prod_{j\in S^c}b_j,\hs\hs

3705: % \prod_{i\in S}(x-b_i)\prod_{j\in S^c}a_j\hs\hs\Big{|}\hs\hs S\text{  short}\right\ra,$$

3706: % and $$\mathcal{K2}=\left\la c_id_i\mid i\leq n\right\la.$$

3707: where

3708: \begin{equation*}\label{asbs}

3709: A_S= \prod_{i\in S}(x-a_i)\prod_{j\in S^c}b_j\hs\hs

3710: \text{   and   }\hs\hs B_S = \prod_{i\in S}(x-b_i)\prod_{j\in S^c}a_j.

3711: \end{equation*}

3712: Here $\delta$ is the image in $\hso\left(\NN\right)$

3713: of the unique positive root of $SU(2)$.

3714: The Weyl group $W$ of $SU(2)$, isomorphic to $\Z/2\Z$, acts on this ring by fixing $x$

3715: and switching $a_i$ and $b_i$ for all $i$.

3716: Let $c_i = a_i+b_i$, and let $C_S=A_S+B_S$.

3717: Let $$P=\Q[c_1,\ldots,c_n,\delta,x]\Big/

3718: \Big\la c_i^2-\delta^2\hs\Big{|}\hs i\leq n\Big\ra$$ and

3719: $$Q=P^W=\Q[c_1,\ldots,c_n,\delta^2,x]\Big/

3720: \Big\la c_i^2-\delta^2\hs\Big{|}\hs i\leq n\Big\ra.$$

3721: Let $$\I=\Big\la A_S, B_S\hs\Big{|}\hs S\text{   short}\Big\ra\subs P

3722: \hspace{15pt}\text{and}\hspace{15pt}

3723: \J=\Big\la C_S\hs\Big{|}\hs S\text{   short}\Big\ra\subs Q,$$

3724: so that $$\hso(\NN)\cong P/\I

3725: \hspace{15pt}\text{and}\hspace{15pt}

3726: \hso(\NN)^W\cong Q/\J.$$

3727: Note that all odd powers of $\delta$ in the expression for $C_S=A_S+B_S$ cancel out.

3728: Then by Theorem~\ref{ordinary} and Remark~\ref{notsobad},

3729: \begin{eqnarray*}

3730: \hso(\M)&\cong& \frac{\hso\left(\NN\right)^W}{ann(e)}

3731: \cong \frac{Q}{(e:\J)},

3732: \end{eqnarray*}

3733: where $e=\delta^2(x^2-\delta^2)$, and $(e:\J)$ is the ideal of elements

3734: of $Q$ whose product with $e$ lies in $\J$.

3735: 

3736: If $S$ is a nonempty short subset, let

3737: $m_S$ be the smallest element of $S$, $n_S$ the smallest

3738: element of $S^c$, and $$D_S = \prod_{m_S\neq i\in S}(c_i-x)

3739: \,\,\cdot\!\prod_{n_S\neq j\in S^c}(c_{n_S}+c_j)\in Q.$$

3740: 

3741: \begin{theorem}\label{hp}

3742: The circle-equivariant cohomology ring of the hyperpolygon space

3743: $\M$ is isomorphic to\footnote{The class

3744: denoted by $c_i$ in \cite{HP2} differs from our $c_i$ by a sign, hence

3745: to recover the presentation of \cite{HP2} we must replace $c_i-x$

3746: with $c_i+x$ in the expression for $D_S$.}

3747: $$Q\big/\big\la

3748: D_S\mid \emptyset\neq S\text{   short}\big\ra.$$

3749: \end{theorem}

3750: 

3751: \begin{proof}

3752: We begin by proving that $e\cdot D_S\in\mathcal{J}$ for all nonempty

3753: short subsets $S\subs\{1,\ldots,n\}$.  We will in fact prove the slightly

3754: stronger statement

3755: $$e\cdot D_S\in\Big\la C_T\hs\Big{|}\hs T\subs S\text{   short}\Big\ra

3756: \subs\mathcal{J},$$

3757: proceeding by induction on $|S|$.

3758: We will assume, without loss of generality, that $n\in S$.

3759: The base case occurs when $S=\{n\}$, and in this case

3760: we observe that

3761: $$e\cdot D_S = 2^{n-3}\cdot(x+c_n)\cdot\Big((2x-c_n)\cdot C_{\emptyset} - c_n\cdot C_S\Big).$$

3762: We now proceed to the inductive step, assuming that the proposition

3763: is proved for all short subsets of size less than $|S|$, and all values of $n$.

3764: For all $T\subs S\smallsetminus\{n\}$, we have

3765: $$\half\Big(C_T-C_{T\cup\{n\}}\Big)=(c_n-x)\cdot C'_T,$$

3766: where $C'_T$ is the polynomial in the variables $\{c_1,\ldots,c_{n-1},\delta^2\}$

3767: corresponding to the short subset $T\subs\{1,\ldots,n-1\}$.

3768: Since $S\smallsetminus\{n\}$ is a short subset of $\{1,\ldots,n-1\}$

3769: of size strictly smaller than $S$,

3770: our inductive hypothesis tells us that $e\cdot D_S/(c_n-x)$

3771: can be written as a linear combination of polynomials $C'_T$, where the coefficients

3772: are quadratic polynomials in $\{c_1,\ldots,c_{n-1},\delta^2\}$.

3773: Replacing $C'_T$ with $\half\left(C_T-C_{T\cup\{n\}}\right)=(c_n-x)\cdot C'_T$,

3774: we have expressed $e\cdot D_S$ in terms of the appropriate polynomials.

3775: This completes the induction.

3776: 

3777: Suppose that $F\in Q$ is an element of degree less than $n-2$

3778: such that $e\cdot F\in\J$.  By the second isomorphism

3779: of Theorem~\ref{ordinary}, this implies that $e'\cdot F\in\I\subs P$,

3780: where $e'=\delta(x^2-\delta^2)$.

3781: Consider the quotient ring $R$ of $P$ obtained by setting

3782: $a_i^2=b_i^2=x=0$ for all $i$.  (Recall that $a_i=\half(c_i+\delta)$

3783: and $b_i=\half(c_i-\delta)$.)  Then element $e'$ maps to zero in $R$,

3784: while the generators $\{A_S, B_S\}$ of $\I$ descend to a basis

3785: for the $n^{\text{th}}$ degree part of $R$.  This means that we

3786: must have $e'\cdot F=0\in P$.

3787: Using the additive basis for $P$ consisting of monomials

3788: that are squarefree in the variables $c_1,\ldots,c_n$,

3789: it is easy to check that $e'$ is not a zero divisor in $P$,

3790: and therefore that $F=0$.

3791: 

3792: Finally, we must show that $\{D_S\mid \emptyset\neq S\text{  short}\}$

3793: generates all elements of $(e:\J)$ of degree at least $n-2$.

3794: We obtain this fact from the following technical lemma, the proof of which we defer

3795: until the end of the section.

3796: 

3797: \begin{lemma}\label{tech}

3798: The set $\{D_S\mid \emptyset\neq S\text{  short}\}$ 

3799: descends to a basis for the degree $n-2$

3800: part of the quotient ring $Q/\la x\ra$.

3801: \end{lemma}

3802: 

3803: Let $F$  be an element of minimal degree $k\geq n-2$ that is in

3804: $(e:\J)$ but not $\la D_S\mid \emptyset\neq S\text{  short}\ra$.

3805: By Lemma \ref{tech}, $F$

3806: differs from an element of

3807: $\la D_S\mid \emptyset\neq S\text{  short}\ra$ by $x\cdot F'$ for some $F'$

3808: of degree $k-1$.  By equivariant formality of $\hso(\M)$,

3809: $$x\cdot F' = F\in (e:\J)\impl F'\in (e:\J),$$

3810: which contradicts the minimality of $k=\operatorname{deg}F$.

3811: Hence $\la D_S\mid \emptyset\neq S\text{  short}\ra=(e:\J)$, and the proposition

3812: is proved.

3813: \end{proof}

3814: 

3815: \vspace{-\baselineskip}

3816: \begin{corollary}\label{konno}

3817: The ordinary cohomology ring $H^*(\M)$ is isomorphic to

3818: $$\Q[c_1,\ldots,c_n]\Big/\big\la c_i^2-c_j^2\mid i,j\leq n\big\ra

3819: +\la\text{all monomials of degree $n-2$}\ra.$$

3820: \end{corollary}

3821: 

3822: \begin{proof}

3823: This follows from the fact that $H^*(\M)\cong \hso(\M)/\la x\ra$

3824: for any equivariantly formal space $M$, and the observation in \cite{HP2}

3825: that $\{D_S\mid \emptyset\neq S\text{  short}\}$ descends to a basis for the degree $n-2$

3826: part of $Q/\la x\ra$.

3827: \end{proof}

3828: 

3829: \begin{prooftech}

3830: Let $\bk = \half(c_1+\ck)$ for all $k$, so that $c_k = 2\bk-d_1$.

3831: The relations $c_k^2 = c_1^2$ translate to $d_k^2 = d_1d_k$ for all $k$,

3832: and we have

3833: $$Q/\la x\ra = \Q[d_1,\ldots,d_n]\Big/\big< d_k^2-d_1d_k\hs\big{|}\hs k\in\twn\big>.$$

3834: For all short subsets $S$, put

3835: $$\Sb = S\smallsetminus\{m_S\}\hspace{15pt}\text{and}\hspace{15pt}

3836: \Lb = S^c\smallsetminus\{n_S\},$$

3837: and let

3838: $$v_S = (-1)^n\pjl (\bj+\bns-d_1)\times\pis(2\bi-d_1).$$

3839: For all $A\subs\twn$, let

3840: $$\ba = (-1)^{|A|} d_1^{n-2-|A|} \prod_{k\in A} d_k$$ for all $A \subsetneq \twn$.

3841: Then $\{d_A\}$

3842: is a basis for

3843: the $(n-2)^{\text{nd}}$ graded piece of $Q/\la x\ra$,

3844: and $v_S$ is equal to $(-1)^n\cdot 2^{-|\Lb|}$ times the image of $D_S$ in $Q/\la x\ra$.

3845: Hence our the statement of Lemma \ref{tech} is that for all $A$, $\ba$ may be 

3846: expressed as a linear combination of the elements $\{v_S\mid S\in\sa\}$.

3847: 

3848: \begin{convention}

3849: The notation $S^c$ refers to the complement of $S$ inside of the set $\otn$,

3850: while the notation $A^c$ refers to the complement of $A$ inside of the set $\twn$.

3851: \end{convention}

3852: 

3853: \begin{claim}\label{vs}

3854: We have the following expression for $v_S$ in terms of the basis $\{d_A\}$:

3855: $$\vs =

3856: \begin{cases}

3857: \displaystyle{\sum_{\ss{\Lb\hs\subseteq A \\ m_S\notin A}}} \hs 2^{|\asb|} \hs\ba

3858: & \text{if}\hspace{10pt}1 \in S^c;\vspace{.5cm}\\

3859: \displaystyle{\sum_{S^c \nsubseteq A}} \hs 2^{|\asb|} 

3860: \hs\ba & \text{if}\hspace{10pt}1 \in S.\\

3861: \end{cases}

3862: $$

3863: \end{claim}

3864: 

3865: \begin{proof}

3866: Any degree $n-2$ monomial

3867: in $d_1,\ldots,d_n$ is equal to $(-1)^{|A|}d_A$, where $A$ is the set of $k>1$ such that

3868: $d_k$ appears in the monomial.

3869: Expanding $\vs$, we need to count (with sign) the occurrence of $d_A$ for each $A$.

3870: In most cases we find that there is no cancellation, and the claim is straightforward.

3871: The most difficult case occurs when

3872: $1\in S$ (therefore $n_S=1$) and $\ms\in A$; 

3873: in this case the number of times (with multiplicity) that $d_A$

3874: occurs in $\vs$ is 

3875: \begin{eqnarray*}

3876: && (-1)^n(-1)^{|A|}(-1)^{|\acsb|}\hs 2^{|\asb|}\sum_{E\subsetneq\Ac\cap\Lb}(-1)^{|E|}

3877: \vspace{.3cm}\\

3878: \vspace{.3cm}&=& (-1)^n(-1)^{|A|}(-1)^{|\acsb|}\hs 2^{|\asb|}

3879: \left((1-1)^{|\Ac\cap\Lb|}-(-1)^{|\Ac\cap\Lb|}\right)\vspace{.3cm}\\

3880: % &=& (-1)^n(-1)^{|A|}(-1)^{|\acsb|}\hs 2^{|\asb|}(-1)^{|\Ac\cap\Lb|+1}\\

3881: \vspace{.3cm}&=& (-1)^{n+|A|+|\acsb|+|\Ac\cap\Lb|+1}\hs 2^{|\asb|}\\

3882: % &=& (-1)^{n+|A|+|\Ac|+1}\hs 2^{|\asb|}\\

3883: \vspace{.3cm}&=& (-1)^{2n}\hs 2^{|\asb|}\\

3884: \vspace{.3cm}&=& \hs 2^{|\asb|}.

3885: \end{eqnarray*}

3886: We leave the remaining cases to be checked by the reader.

3887: \end{proof}

3888: 

3889: \vspace{-\baselineskip}

3890: \begin{claim}\label{ws}

3891: Suppose that $1\in S$.

3892: Let $S_0 = S$, and for $1\leq k\leq |S|$, let $S_k = S_{k-1}\setminus\{m_{S_{k-1}}\}$.

3893: (In other words, let $S_k$ consist of the $|S|-k$ largest elements of $S$).

3894: Then $$v_S + \displaystyle{\sum_{k=1}^{|S|-1}} \hs 2^{k-1}\hs v_{S_k} 

3895: = \displaystyle{\sum_A}\hs 2^{|\asb|}\hs d_A.$$

3896: \end{claim}

3897: 

3898: \begin{proof}

3899: We proceed by induction to show that

3900: $$v_S + \displaystyle{\sum_{k=1}^l} \hs 2^{k-1}\hs v_{S_k} =

3901: \displaystyle{\sum_A} \hs 2^{|\asb|}\hs d_A - 

3902: \hs 2^l\cdot\!\displaystyle{\sum_{\overline{S^c_{l+1}}\subs A}}

3903: 2^{|\asb_l|}\hs d_A.$$

3904: The case $l=|S|-1$ is the statement of the claim.

3905: The base case $l=0$ follows from Claim \ref{vs},

3906: together with the observation that $\overline{S^c_{1}}=S^c$.

3907: More generally, for all $l\geq 1$, we have 

3908: $$\overline{S^c_{l+1}} = S^c\cup\{m_{S_1},\ldots,m_{S_l}\}.$$

3909: Then 

3910: \begin{eqnarray*}

3911: v_S + \displaystyle{\sum_{k=1}^{l+1}} \hs 2^{k-1}\hs v_{S_k} &=&

3912: v_S + \displaystyle{\sum_{k=1}^l}\hs 2^{k-1}\hs v_{S_k} + 2^l\hs v_{S_{l+1}}\\

3913: % &=& \displaystyle{\sum_A} \hs 2^{|\asb|}\hs d_A 

3914: % - 2^l\hs\displaystyle{\sum_{\overline{S^c_{l+1}}\subs A}}

3915: % 2^{|\asb_l|}\hs d_A + 2^l\hs v_{S_{l+1}}\hs\hs\hs\hs\text{inductive hypothesis}\\

3916: &=& \displaystyle{\sum_A} \hs 2^{|\asb|}\hs d_A - 2^l\cdot\!\displaystyle{\sum_{\overline{S^c_{l+1}}\subs A}}

3917: 2^{|\asb_l|}\hs d_A + 2^l\cdot\!\sum_{\ss{\overline{S^c_{l+1}}\subs A \\ m_{S_{l+1}}\notin A}}

3918: 2^{|\asb_{l+1}|}\hs d_A

3919: \end{eqnarray*}

3920: by the inductive hypothesis and Claim \ref{vs}.

3921: Using the fact that

3922: $\asb_{l+1}=\asb_l$ when $m_{S_{l+1}}\notin A$,

3923: this is equal to

3924: $$\sum_A\hs 2^{|\asb|}\hs d_A - 2^l\cdot\!\sum_{\overline{S^c_{l+1}}\cup\{m_{S_{l+1}}\}\subs A}

3925: 2^{|\asb_l|}.$$

3926: Finally, since $|\asb_{l+1}|=|\asb_l|-1$ when $m_{S_{l+1}}\in A$,

3927: this reduces to

3928: $$\sum_A \hs 2^{|\asb|}\hs d_A - 2^{l+1}\hs\cdot\!\displaystyle{\sum_{\overline{S^c_{l+2}}

3929: \subs A}}

3930: 2^{|\asb_{l+1}|}\hs d_A,$$ thus proving our claim.

3931: \end{proof}

3932: 

3933: For all short subsets $T$ containing $1$,

3934: let $w_T=\displaystyle{\sum_A}2^{|A\cap \bar{T}|}\hs d_A$, 

3935: which by Claim \ref{ws} is expressible as a linear combination of elements

3936: of the set $\{v_S\mid \emptyset\neq S\in\mathcal{S}\}$.

3937: Let $$x_S = 

3938: \begin{cases}

3939: \displaystyle{\sum_{1\in T\subs S}}(-1)^{|S|+|T|}w_T & \text{if $1\in S$,}\\

3940: v_S & \text{if $1\in S^c$.}

3941: \end{cases}$$

3942: Our last task will be to prove that the transition matrix $\Upsilon$ 

3943: taking the basis $\{d_A\}$ to 

3944: the set $\{x_S\}$

3945: is lower triangular with ones on the diagonal, and therefore invertible.

3946: In order to make sense of ``the diagonal," 

3947: we must first give an explicit bijection between the set of proper subsets of $\twn$

3948: and the set of nonempty short subsets of $\otn$.

3949: We do this as follows:  given $A\subsetneq\twn$, let 

3950: $$S(A) = \begin{cases}

3951: \Ac &\text{ if $\Ac$ is short,}\\

3952: \otn\setminus\Ac = A\cup\{1\} &\text{ if $\Ac$ is long.}

3953: \end{cases}$$

3954: The rows of $\Upsilon$ will be indexed by $A$, 

3955: and the sets will appear in lexicographic order within cardinality

3956: class.  For example, when $n=4$, the order of the rows will be 

3957: $\emptyset$, $\{2\}$, $\{3\}$, $\{4\}$,

3958: $\{2,3\}$, $\{2,4\}$, $\{3,4\}$.  

3959: The columns will be indexed by $S$ according to the bijection described above.

3960: 

3961: \begin{claim}\label{tri}

3962: The matrix $\Upsilon$ is lower triangular with ones on the diagonal.

3963: \end{claim}

3964: 

3965: \begin{proof}

3966: First consider a column corresponding to a short subset $S$ that does {\it not} contain 1.

3967: The entries in this column correspond to the coefficient of $d_A$ in $x_S = v_S$.

3968: From Claim \ref{vs}, we see that $d_A$ appears in $v_S$ only if $\Lb\subs A\subs \Lb\cup\Sb$,

3969: and if so it appears with a coefficient of $2^{|\asb|}$.

3970: Since $1\notin S$, we have $\Lb = S^c\setminus\{1\} = \twn\setminus S$.

3971: The diagonal entry corresponds to the set $A = \twn\setminus S = \Lb$, therefore

3972: in this row we get the number $2^{|\asb|} = 2^{|\Lb\cap\Sb|} = 1$.

3973: Since the set $A$ corresponding to a given row can never contain the set $B$ corresponding

3974: to a 

3975: lower row, the rows above the diagonal fail to satisfy the condition $\Lb\subs A$,

3976: and we get all zeros.

3977: 

3978: Now consider a column corresponding to a short subset $S$ that {\it does} contain $1$.

3979: In this case, the coefficient of $d_A$ in $x_S$ is

3980: $$(-1)^{|S|}\displaystyle{\sum_{1\in T\subs S}}(-1)^{|T|}2^{|A\cap\bar{T}|}.$$

3981: The diagonal entry corresponds to the set $A=\Sb$, and we get

3982: \begin{eqnarray*}

3983: (-1)^{|S|}\displaystyle{\sum_{1\in T\subs S}}(-1)^{|T|}2^{|\bar{T}|}

3984: &=& (-1)^{|\Sb|}\displaystyle{\sum_{1\in T\subs S}}(-2)^{|\bar{T}|}\\

3985: &=& (-1)^{|\Sb|}(1-2)^{|\Sb|} = 1.

3986: \end{eqnarray*}

3987: Any row above the diagonal corresponds to a set $A$ which does not contain $\Sb$.

3988: Choose an element $l\in\Sb\setminus A$.  Then

3989: \begin{eqnarray*}

3990: (-1)^{|S|}\displaystyle{\sum_{1\in T\subs S}}(-1)^{|T|}2^{|A\cap\bar{T}|}

3991: &=& (-1)^{|S|}\displaystyle{\sum_{l\in T}}(-1)^{|T|}2^{|A\cap\bar{T}|}

3992: + (-1)^{|S|}\displaystyle{\sum_{l\notin T}}(-1)^{|T|}2^{|A\cap\bar{T}|}\\

3993: &=& (-1)^{|S|}\displaystyle{\sum_{l\notin T}}\left[(-1)^{|T|}2^{|A\cap\bar{T}|}

3994: + (-1)^{|T\cup\{l\}|}2^{|A\cap\bar{T}|}\right]\\

3995: &=& 0.

3996: \end{eqnarray*}

3997: Hence $\Upsilon$ is lower triangular.

3998: \end{proof}

3999: 

4000: Claim \ref{tri} tells us that each $d_A$ can be expressed as a linear

4001: combination of elements of the form $x_S$, and therefore of elements

4002: of the form $v_S$.  This completes the proof of Lemma \ref{tech}

4003: \end{prooftech}

4004: \end{section}

4005: 

4006: \begin{section}{Cohomology of the core components}\label{last}

4007: In this section we compute the $S^1$-equivariant 

4008: and ordinary cohomology rings of the core component

4009: $U_S$ corresponding to a short subset $S\subs\otn$.  Since $U_S$ is the closure of a cell

4010: in an even-dimensional equivariant cellular decomposition of $\M$, the restriction

4011: map $\hso(\M)\to\hso(U_S)$ is surjective.  In particular, $\hso(U_S)$ 

4012: is generated by restrictions

4013: of the Kirwan classes $c_1,\ldots,c_n,x$.  

4014: For our presentation, it will be convenient

4015: to assume that $1\in S$, and to work with the classes 

4016: $d_k = \half(c_1+c_k)$ introduced in the proof of Lemma \ref{tech}.  

4017: With respect to these generators, we obtain the following result.

4018: 

4019: \begin{theorem}\label{eqcore}

4020: The equivariant cohomology ring $\hso(U_S)$ is isomorphic to

4021: $\Q[d_1,\ldots,d_n,x]/\mathcal{J}_S,$

4022: where $\mathcal{J}_S$ is generated by the following four families:

4023: \begin{eqnarray*}&1)&\hs\hs d_1 - d_i \hs\hs\text{ for all }\hs\hs i\in S\\

4024: &2)&\hs\hs

4025: d_j(d_1 - d_j)

4026: \hs\hs\text{ for all }\hs\hs j\in S^c\\

4027: &3)&\hs\hs \prod_{j\in R}d_j\hs\hs\text{ for all }\hs\hs R\subs S^c\text{ such that }R\cup S

4028: \text{ is long}\\

4029: &4)&\hs\hs (d_1 + x)^{|S|-1}\cdot\frac{1}{d_1}\left(\prod_{j\in L}(d_j-d_1)-\prod_{j\in L}d_j\right)

4030: \hs\hs\text{ for all long subsets }L\subs S^c.

4031: \end{eqnarray*}

4032: \end{theorem}

4033: 

4034: \begin{corollary}\label{ordcore}

4035: The ordinary cohomology ring $H^*(U_S)$ is isomorphic to

4036: $\Q[d_1,\ldots,d_n]/\mathcal{I}_S,$

4037: where $\mathcal{I}_S$ is generated by the following four families:

4038: \begin{eqnarray*}&1)&\hs\hs d_1 - d_i \hs\hs\text{ for all }\hs\hs i\in S\\

4039: &2)&\hs\hs

4040: d_j(d_1 - d_j)

4041: \hs\hs\text{ for all }\hs\hs j\in S^c\\

4042: &3)&\hs\hs \prod_{j\in R}d_j\hs\hs\text{ for all }\hs\hs R\subs S^c\text{ such that }R\cup S

4043: \text{ is long}\\

4044: &4)&\hs\hs d_1^{|S|-2}\prod_{j\in L}(d_j-d_1)

4045: \hs\hs\text{ for all long subsets }L\subs S^c.

4046: \end{eqnarray*}

4047: \end{corollary}

4048: 

4049: \begin{remark}\label{integer}

4050: Each of these relations has a geometric interpretation.

4051: For $i\in\{1,\ldots,n\}$, it is possible to construct a line bundle on $\M$ with equivariant

4052: Euler class $d_i - d_1$

4053: which has a section supported on the locus

4054: where $q_1q_1^*$ and $q_iq_i^*\in\R^3$ point in opposite directions.  

4055: Since this locus is disjoint

4056: from $U_S$ when $i\in S$, we have $$1)\hs\hs d_i = d_1 \in \hso(\us)\hs\text{ for all }\hs i\in S.$$

4057: Similarly,

4058: $-d_j = -\half(c_1+c_j)$ is represented by the divisor $Z_{1j}\subs\M$ of points

4059: on which $q_1q_1^*$ and $q_iq_i^*\in\R^3$ point in the same direction

4060: \cite[\S 3]{HP2}.

4061: Then by the previous reasoning, we obtain 

4062: $$2)\hs\hs d_j(d_1 - d_j) = 0\in\hso(\us)\hs\text{ for all }\hs j\in S^c.$$

4063: For any $R\subs S^c$, we may intersect the divisors $Z_{ij}\subs\M$ 

4064: (defined in the analogous way) for all $j\in R$

4065: to find that

4066: the cohomology class $(-1)^{|R|}\prod_{j\in R}d_j$ is represented

4067: by the subvariety $Z_R\subs \M$ of points with $q_j$ proportional to $q_1$

4068: for all $j\in R$.

4069: When restricted to $U_S$, this becomes $U_S\cap U_{R\cup S}$,

4070: the closure of the unstable manifold for the critical locus $\M_{R\cup S}\cap U_S$

4071: of the Morse-Bott function $\Phi|_{U_S}$.

4072: In particular, we have $$3)\hs\hs \prod_{j\in R}d_j=0\in\hso(\us)\text{ if }R\cup S\hs\text{ is long}.$$

4073: To understand the fourth family of relations, we note that the class

4074: $$d_1+x = 2d_i-d_1+x = c_i+x \in \hso(U_S)$$ 

4075: is represented by the divisor $W_i$ of points

4076: with $p_i = 0$ for any $i\in S$ \cite[\S 3]{HP2}.  In particular,

4077: $(d_1+x)^{|S|-1}$ 

4078: is represented by the subvariety of points in $U_S$ on which $p_i = 0$

4079: for all $i\in\Sb$, which is equal to $\X_S$ by the complex moment map condition.

4080: Hence the fourth family of generators of $\mathcal{J}_S$ (or of $\mathcal{I}_S$)

4081: can be interpreted geometrically as $(d_1+x)^{|S|-1}$ 

4082: (respectively $d_1^{|S|-1}$

4083: in the nonequivariant case) times

4084: classes that vanish in $\hso(\X_S)$ (see Lemma \ref{pols}).

4085: % 

4086: % The critical submanifold $J_{R\cup S}\cap U_S$ is a projective

4087: % space with cohomology ring generated over $\Z[x]$ by the restriction

4088: % of $d_1+x$.  It is shown in $\cite{HK}$ that the cohomology ring

4089: % of $\X_S$ is also generated over the integers by the restrictions

4090: % of $d_1,\ldots,d_n$.  Hence Morse-Bott theory tells us that

4091: % $d_1,\ldots,d_n,x$ is actually an {\em integer} basis for $\hso(U_S)$.

4092: \end{remark}

4093: 

4094: \begin{eqcoreproof}

4095: Let $\phi:\Q[d_1,\ldots,d_n,x]\to\hso(\us)$ denote the composition of the Kirwan

4096: map with restriction to $\us$.  Our claim is that $\ker\phi = \mathcal{J}_S$.

4097: For every short subset $T$ containing $S$,

4098: let $$\phi_T:\Q[d_1,\ldots,d_n,x]\to\hso(\mtus)$$ denote

4099: the composition of the Kirwan map with restriction to $\mtus$,

4100: and let $$J_T = \ker\phi_T.$$

4101: Similarly, let $$\phi_{\emptyset}:\Q[d_1,\ldots,d_n,x]\to\hso(\X_S)$$

4102: be the natural map, and let

4103: $$J_{\emptyset} = \ker\phi_{\emptyset}.$$

4104: The kernel of the restriction map $\hso(U_S)\to\hso(U_S^{S^1})$

4105: to the fixed point set of $U_S$ is a torsion module over $\hso(pt)$ \cite[3.5]{AB}, 

4106: and Proposition \ref{formality}

4107: tells us that $\hso(U_S)$ is a free $\hso(pt)$-module. 

4108: Hence the restriction map is injective,

4109: and we have $$\ker\phi = \ker\phi_{\emptyset}\cap\bigcap_{\ts}\ker\phi_T.$$

4110: We know that $\mtus\cong\C P^{|S|-2}$ for all short $\ts$,

4111: therefore $$\hso(\mtus)\cong\Q[h,x]/h^{\sz}.$$

4112: Furthermore, we know that for all $i\in T$, the restriction of $d_i + x$ 

4113: to $\hso(U_T)$ is represented by the divisor $W_i\cap U_T$ (see Remark \ref{integer}),

4114: and therefore restricts to the class of a hyperplane on $\mtus$.

4115: Hence $\phi_T(d_i+x)=h$ for all $i\in T$.

4116: On the other hand, for $j\in T^c$, the class $d_j$ 

4117: is represented by the divisor $Z_{1j}$ on $\M$, which is disjoint from $\mtus$,

4118: hence $\phi(d_j) = 0$ for all $j\in T^c$.

4119: Thus we conclude that

4120: $$\ker\phi_T = \<d_1-d_i, \hs d_j, (d_1+x)^{\sz}\mid i\in T, j\in T^c\>.$$

4121: 

4122: \begin{lemma}\label{jt}

4123: The intersection 

4124: $\displaystyle{

4125: \bigcap_{\ts}}\ker\phi_T$

4126: is equal to

4127: $$\<d_1-d_i, \hs d_j(d_1 - d_j), 

4128: \displaystyle{\prod_{j\in R}d_j},

4129: (d_1+x)^{\sz}

4130: % where $i$ ranges over $S$, $j$ ranges over $S^c$,

4131: % and $R\subs S^c$ is any subset such that $R\cup S$ is long.

4132: \hs\hs\bigg |\hs\hs i\in S, j\in S^c, R\cup S\text{ long}\>.$$

4133: \end{lemma}

4134: 

4135: \begin{proof}

4136: First, since the variable $x$ appears only in the generator $(d_1+x)^{\sz}$,

4137: which is contained in every ideal on both sides of the equation, we may reduce

4138: the problem to showing that

4139: \begin{equation}\label{reduce}

4140: \bigcap_{\ts}\big\langle d_1-d_i, \hs d_j\bigmid i\in T, j\in T^c\big\rangle = 

4141: \<d_1-d_i, \hs d_j(d_1-d_j), \prod_{j\in R}d_j

4142: \hs\hs\bigg |\hs\hs i\in S, j\in S^c, R\cup S\text{ long}\>

4143: \end{equation}

4144: in the ring $\Q[d_1,\ldots,d_n]$.

4145: Both ideals cut out the (reducible) variety 

4146: $$\bigcup_{\ts}Y_T\subs\operatorname{Spec}\Q[d_1,\ldots,d_n],$$

4147: where $$Y_T = \big\{(z_1,\ldots,z_n\bigmid

4148: z_i=z_1\hs\forall i\in S, z_j=0\hs\forall j\in S^c\big\}.$$

4149: The left hand side of Equation \eqref{reduce} is an intersection of prime ideals,

4150: and is therefore radical.

4151: Thus by Hilbert's Nullstellensatz, it is sufficient to prove that

4152: the right hand side of Equation \eqref{reduce} is radical.

4153: This involves showing that the ideal is saturated, with

4154: Hilbert polynomial equal to the constant $\#\{\text{short }\ts\}$.

4155: 

4156: The degree $k$ piece of the quotient

4157: $$\Q[d_1,\ldots,d_n]/\<d_1-d_i, \hs d_j(d_1-d_j)

4158: \mid i\in S, j\in S^c\>$$

4159: has a basis of elements of the form

4160: $$d_1^{e_1}\prod_{j\in S^c}d_j^{e_j},$$

4161: where $e_j \in\{0,1\}$ for all $j>0$, and $e_1+\sum_{j\in S^c}e_j = k$.

4162: The subset of these elements with the property that

4163: $S\cup\{j\mid e_j = 1\}$ is short descends to a basis for the

4164: degree $k$ part of the ring

4165: $$\Q[d_1,\ldots,d_n]\bigg/\<d_1-d_i, \hs d_j(d_1-d_j), \prod_{j\in R}d_j

4166: \hs\hs\bigg |\hs\hs i\in S, j\in S^c, R\cup S\text{ long}\>,$$

4167: hence our ideal has the desired Hilbert polynomial.

4168: It is also clear from this description that if an element

4169: $a$ of the quotient ring is nonzero, so is $d_1^d\cdot a$ for any $d\geq 0$,

4170: hence our ideal is saturated.

4171: \end{proof}

4172: 

4173: It now remains to show that

4174: $$\mathcal{J}_S = \<d_1-d_i, \hs d_j(d_1 - d_j), \prod_{j\in R}d_j,

4175: \hs (d_1+x)^{\sz} \hs\hs\bigg |\hs\hs i\in S, j\in S^c, R\cup S\text{ long}\> \cap \ker\phi_{\emptyset}.$$

4176: The fact that $\mathcal{J}_S$ is contained in the intersection is clear.

4177: To show the opposite containment, consider an element

4178: $$a + \eta\cdot (d_1+x)^{\sz}\in \<d_1-d_i, \hs d_j(d_1 - d_j), \prod_{j\in R}d_j,

4179: \hs (d_1+x)^{\sz} \hs\hs\bigg |\hs\hs i\in S, j\in S^c, R\cup S\text{ long}\>,$$

4180: with $$a\in \<d_1-d_i, \hs d_j(d_1 - d_j), \prod_{j\in R}d_j\hs\hs\bigg |\hs\hs

4181: i\in S, j\in S^c, R\cup S\text{ long}\>,$$

4182: and suppose that we also have

4183: $$a + \eta\cdot (d_1+x)^{\sz}\in\ker\phi_{\emptyset}.$$

4184: 

4185: \begin{lemma}\label{pols}{\em \cite{HK}}

4186: The kernel of $\phi_{\emptyset}$

4187: is equal to 

4188: $$\<d_1-d_i, \hs d_j(d_1-d_j), \prod_{j\in R}d_j, 

4189: \hs (d_1+x)^{\sz}d_1^{-1}\left(\prod_{j\in L}(d_j-d_1)-\prod_{j\in L}d_j\right)\>,$$

4190: where $i\in S, j\in S^c$, and $R,L\subs S^c$, with $R\cup S$ and $L$ both long.

4191: % \mid i\in S, j\in S^c, R\cup S\text{ long}, L\subs S^c\text{ long}\>.$$

4192: \end{lemma}

4193: 

4194: Lemma \ref{pols} tells us that $a\in \ker\phi_{\emptyset}$,

4195: therefore $$\eta\cdot (d_1+x)^{\sz}\in\ker\phi_{\emptyset}.$$

4196: But $(d_1+x)^{\sz}$ is represented in $\hso(U_S)$ by the subvariety

4197: $\X_S$ (see Remark \ref{integer}), hence 

4198: $$0 = \phi_{\emptyset}(\eta\cdot (d_1+x)^{\sz})

4199: = \phi_{\emptyset}(\eta)\cdot e(\X_S),$$

4200: where $e(\X_S)$ is the equivariant Euler class of the normal bundle

4201: to $\X_S$ inside of $U_S$.  Since the equivariant Euler class

4202: of the normal bundle to a component of the fixed point set is never a zero-divisor,

4203: we have $\eta\in\ker\phi_{\emptyset}$.

4204: Then by Equation \ref{pols}, $$a+\eta\cdot (d_1+x)^{\sz}\in\mathcal{J}_S.$$

4205: This completes the proof of Theorem \ref{eqcore}.

4206: \end{eqcoreproof}

4207: 

4208: \vspace{-\baselineskip}

4209: \begin{example}\label{projective}

4210: For arbitrary $n$ and $\a$, suppose that $S$ is a maximal short subset.

4211: Then Corollary \ref{ordcore} tells us that

4212: $H^*(U_S)\cong\Q[d_1]/\langle d_1^{n-2}\rangle$.

4213: We conjecture that in this case we in fact have $U_S\cong \C P^{n-3}$.

4214: \end{example}

4215: 

4216: \begin{example}\label{again}

4217: Consider the core component pictured in Example \ref{ooh}.

4218: By Theorem \ref{ordcore} and Remark \ref{integer},

4219: $$H^*(U_S) \cong \Q[d_1,d_3,d_4,d_5]

4220: \Bigg/

4221: \left<\begin{array}{c}

4222: d_3(d_1-d_3),\hs d_4(d_1-d_4),\hs d_5(d_1-d_5),\hs d_3d_4,\hs d_3d_5,\hs d_4d_5,\\

4223: d_1(d_1-d_3-d_4),\hs  d_1(d_1-d_3-d_5),\hs d_1(d_1-d_4-d_5)

4224: \end{array}\right>,$$

4225: where $d_1$ is the fundamental class of $\X_S$,

4226: and $d_3$, $d_4$, and $d_5$ are the negatives of the fundamental classes of the curves

4227: labeled $123$, $124$, and $125$, respectively.

4228: Because the transverse intersection of two complex varieties is positive,

4229: we know that $-d_1d_3[U_S] = 1$.

4230: With respect to the basis $$\{d_1-d_3-d_4-d_5,\, d_3,\, d_4,\, d_5\},$$

4231: the intersection form on $H^2(U_S)$ is represented by the matrix

4232: $$\left(\begin{array}{cccc}

4233: 1&&&\\

4234: &-1&&\\

4235: &&-1&\\

4236: &&&-1\end{array}\right).$$

4237: It is likely that $U_S$ is isomorphic 

4238: to the blow-up of $\C P^2$ at three points.

4239: \end{example}

4240: 

4241: \begin{example}

4242: Using the same $\a= (1,1,3,3,3)$,

4243: consider the short subset $S=\{1,3\}$.

4244: In this case, Theorem \ref{ordcore} tells us that

4245: $$H^*(U_S)\cong\Q[d_1,d_2]/\left<d_1^2, d_2(d_1-d_2)\right>.$$

4246: With respect to the basis $\{d_1-d_2,\, d_2\}$,

4247: the intersection form on $H^2(U_S)$ is represented by the matrix

4248: $\footnotesize{\Big(\begin{array}{cc}

4249: -1&0\\

4250: 0& 1

4251: \end{array}

4252: \Big)}$,

4253: hence $U_S$ is homeomorphic to the blow-up of $\C P^2$ at a single point.

4254: \end{example}

4255: \end{section}

4256: \end{chapter}

4257: 

4258: \footnotesize{

4259: \begin{thebibliography}{10}

4260: 

4261: \bibitem[AB]{AB}

4262: M.~Atiyah and R.~Bott.

4263: \newblock The moment map and equivariant cohomology.

4264: \newblock {\em Topology} 23 (1984), 1--28.

4265: 

4266: \bibitem[BV]{BV} N.~Berline and M.~Vergne.

4267: \newblock Z\'eros d'un champ de vecteurs et classes caract\'eristiques  

4268: \'equivariantes.  

4269: \newblock {\em Duke Math. J.}  50 (1983) no. 2, 539--549.

4270: 

4271: \bibitem[Bi]{Bi}

4272: R.~Bielawski.

4273: \newblock Complete hyperkaehler $4n$-manifolds with a local tri-Hamiltonian $R^n$-action.

4274: \newblock {\em Math. Ann.} 314 (1999) no. 3, 505--528.

4275: 

4276: \bibitem[BD]{BD}

4277: R.~Bielawski and A.~Dancer.

4278: \newblock The geometry and topology of toric hyperk\"ahler manifolds.

4279: \newblock {\em Comm. Anal. Geom.} 8 (2000), 727--760.

4280: 

4281: \bibitem[BGH]{BGH}

4282: D.~Biss, T.~Holm, and V.~Guillemin.

4283: \newblock The mod 2 cohomology of fixed point sets

4284: of anti-symplectic involutions.

4285: \newblock ArXiv:  math.SG/0107151.

4286: 

4287: \bibitem[BJ]{BJ}

4288: T.~Brocker and K.~Janich.

4289: \newblock {\em Introduction to Differential Topology}.

4290: \newblock Cambridge University Press, 1982.

4291: 

4292: \bibitem[De]{De}

4293: T.~Delzant.

4294: \newblock Hamiltoniens p\'eriodiques et images convexes de l'application moment.

4295: \newblock {\em Bull. Soc. Math. France} 116  (1988),  no. 3, 315--339.

4296: 

4297: \bibitem[Do]{Do}

4298: I.~Dolgachev.

4299: \newblock {\em Lectures on invariant theory}.

4300: \newblock London Mathematical Society Lecture Note Series, 296.

4301: Cambridge University Press, 2003.

4302: 

4303: \bibitem[ES]{ES}

4304: G.~Ellingsrud and S.A.~Str\o mme.

4305: \newblock On the Chow ring of a geometric quotient.

4306: \newblock {\em Ann. of Math.} (2) 130 (1989), no. 1, 159--187.

4307: 

4308: \bibitem[Fu]{Fu}

4309: W.~Fulton.

4310: \newblock Eigenvalues, invariant factors, highest weights, 

4311: and Schubert calculus.

4312: \newblock {\em Bull. Amer. Math. Soc.} 37 (2000), no. 3, 209--249 .

4313: 

4314: \bibitem[M2]{M2}

4315: D.~Grayson and M.~Stillman.

4316: \newblock Macaulay 2: a computer algebra system for algebraic geometry.

4317: \newblock Version 0.8.56, www.math.uiuc.edu/macaulay2, 1999.

4318: 

4319: \bibitem[GS]{GS}

4320: V.~Guillemin and S.~Sternberg.

4321: \newblock The coefficients of the Duistermaat-Heckman polynomial 

4322: and the cohomology ring of reduced spaces. 

4323: \newblock {\em Geometry, topology, \& physics},  202--213,

4324: Conf. Proc. Lecture Notes Geom. Topology, IV,

4325: Internat. Press, Cambridge, MA, 1995. 

4326: 

4327: \bibitem[HH]{HH}

4328: M.~Harada and T.~Holm.

4329: \newblock The equivariant cohomology 

4330: of hypertoric varieties and their real loci.

4331: \newblock In preparation.

4332: 

4333: \bibitem[HP1]{HP1}

4334: M.~Harada and N.~Proudfoot.

4335: \newblock Properties of the residual circle action on a hypertoric

4336: variety.

4337: \newblock To appear in {\em Pacific J. Math.}

4338: \newblock ArXiv:  math.DG/0207012.

4339: 

4340: \bibitem[HP2]{HP2}

4341: M.~Harada and N.~Proudfoot.

4342: \newblock Hyperpolygon spaces and their cores.

4343: \newblock To appear in {\em Trans. A.M.S.}

4344: \newblock ArXiv:  math.AG/0308218.

4345: 

4346: \bibitem[H1]{H1}

4347: T.~Hausel.

4348: \newblock Geometry of the moduli space of Higgs bundles.

4349: \newblock Thesis for Ph.D. in Pure Mathematics, 

4350: DPMMS, Cambridge University, August 1998.

4351: \newblock ArXiv:  math.AG/0107040.

4352: 

4353: \bibitem[H2]{H2}

4354: T.~Hausel.

4355: \newblock Compactification of moduli of Higgs bundles.

4356: \newblock {\em J. Reine Angew. Math.} 503 (1998), 169--192.

4357: \newblock ArXiv:  math.AG/9804083.

4358: 

4359: \bibitem[H3]{H3}

4360: T.~Hausel.

4361: \newblock Quaternionic geometry of matroids.

4362: \newblock ArXiv:  math.AG/0308146.

4363: 

4364: \bibitem[HP]{HP}

4365: T.~Hausel and N.~Proudfoot.

4366: \newblock Abelianization for hyperk\"ahler quotients.

4367: \newblock ArXiv:  math.SG/0310141.

4368: 

4369: \bibitem[HS]{HS}

4370: T.~Hausel and B.~Sturmfels.

4371: \newblock Toric hyperk\"ahler varieties.

4372: \newblock {\em Doc. Math.}  7  (2002), 495--534 (electronic).

4373: \newblock ArXiv:  math.AG/0203096.

4374: 

4375: \bibitem[HT1]{HT1} T.~Hausel and M.~Thaddeus.

4376: \newblock  Generators for the cohomology ring of the

4377: moduli space of rank 2 Higgs bundles.

4378: \newblock To appear in {\em Proc. London Mat. Soc.}

4379: \newblock ArXiv:  math.AG/0003093.

4380: 

4381: \bibitem[HT2]{HT2} T.~Hausel and M.~Thaddeus.

4382: \newblock   Relations in the cohomology ring of the 

4383: moduli space of rank 2 Higgs bundles,

4384: \newblock {\em J. Amer. Math. Soc.}, 16 (2003), 303--329

4385: \newblock ArXiv: math.AG/0003094.

4386: 

4387: \bibitem[HK]{HK}

4388: J-C.~Hausmann, A.~Knutson.

4389: \newblock The cohomology ring of polygon spaces.

4390: \newblock {\em Ann. Inst. Fourier, Grenoble} 48 (1998) no. 1, 281--321.

4391: 

4392: % \bibitem[HL]{HL} 

4393: % P.~Heinzner and F.~Loose.

4394: % \newblock Reduction of complex Hamiltonian $G$-spaces.

4395: % \newblock {\em Geom. Funct. Anal.} 4 (1994), no. 3, 288--297.

4396: % 

4397: \bibitem[Hi]{Hi}

4398: N.~Hitchin.

4399: \newblock The self-duality equations on a Riemann surface.

4400: \newblock {\em Proc. London Math. Soc.} (3) 55 (1987) no. 1, 59--126.

4401: 

4402: \bibitem[HKLR]{HKLR}

4403: N.~Hitchin, A.~Karlhede, U.~Lindstr\"om, and M.~Ro\v{c}ek.

4404: \newblock Hyper-K\"ahler metrics and supersymmetry.

4405: \newblock {\em Comm. Math. Phys.} 108 (1987)

4406: no. 4, 535--589.

4407: 

4408: \bibitem[KP]{KP}

4409: A.~Khovanskii and A.~Pukhlikov.

4410: \newblock The Riemann-Roch theorem for integrals and sums of quasipolynomials

4411: on virtual polytopes.

4412: \newblock {\em St. Petersburg Math. J.} 4 (1993), 789--812.

4413: 

4414: \bibitem[Ki]{Ki}

4415: F.~Kirwan.

4416: \newblock {\em Cohomology of quotients in symplectic and algebraic geometry}.

4417: \newblock Mathematical Notes, 31.

4418: \newblock Princeton University Press, 1984.

4419: 

4420: \bibitem[Kl]{Kl}

4421: A.~Klyachko.

4422: \newblock Spatial polygons and stable configurations of points in the projective line.

4423: \newblock Algebraic geometry and its applications (Yaroslavl, 1992),

4424: {\em Aspects Math.}, Vieweg, Braunschweig (1994) 67--84.

4425: 

4426: \bibitem[KT]{KT}

4427: A.~Knutson and T.~Tao.

4428: \newblock

4429: The honeycomb model of $GL(n)$ tensor products I:

4430: proof of the saturation conjecture.

4431: \newblock {\em Journal of the A.M.S.} {\bf 12} (1999) no. 4, 1055--1090.

4432: 

4433: \bibitem[K1]{K1}

4434: H.~Konno.

4435: \newblock Cohomology rings of toric hyperk\"ahler manifolds.

4436: \newblock {\em Int. J. of Math.} 11 (2000)

4437: no. 8, 1001--1026.

4438: 

4439: \bibitem[K2]{K2}

4440: H.~Konno.

4441: \newblock On the cohomology ring of the HyperK\"ahler analogue of the

4442: Polygon Spaces.

4443: \newblock Integrable systems, topology, and physics (Tokyo, 2000), 129--149,

4444: {\em Contemp. Math.}, 309,

4445: Amer. Math. Soc., Providence, RI, 2002.

4446: 

4447: \bibitem[Le]{Le}

4448: E.~Lerman.

4449: \newblock Symplectic cuts.

4450: \newblock {\em Math. Res. Lett.} 2 (1995) no. 3, 247--258.

4451: 

4452: \bibitem[LT]{LT}

4453: E.~Lerman and S.~Tolman.

4454: \newblock Hamiltonian torus actions on symplectic orbifolds and toric varieties.

4455: \newblock {\em Trans. Amer. Math. Soc.} 349 (1997) no. 10, 4201--4230.

4456: 

4457: \bibitem[Mk]{Mk}

4458: E.~Markman.

4459: \newblock Generators of the cohomology ring of moduli spaces 

4460: of sheaves on symplectic surfaces. 

4461: \newblock {\emph J. Reine Angew. Math.}  544  (2002), 61--82.

4462: 

4463: \bibitem[Ma]{Ma}

4464: S.~Martin.

4465: \newblock Symplectic quotients by a nonabelian group and

4466: by its maximal torus.

4467: \newblock To appear in the {\em Ann. of Math.}

4468: \newblock ArXiv: math.SG/0001002.

4469: 

4470: \bibitem[MNS]{MNS}

4471: G.~Moore, N.~Nekrasov, and S.~Shatashvili.

4472: \newblock \emph{Integrating over Higgs branches}.

4473: \newblock {\em Comm. Math. Phys.} 209 (2000) no.~1, 97--121.

4474: 

4475: \bibitem[MFK]{MFK}

4476: D.~Mumford, J.~Fogarty, and F.~Kirwan.

4477: \newblock {\em Geometric Invariant Theory}.

4478: \newblock Third edition. Ergebnisse der Mathematik und ihrer

4479: Grenzgebiete (2) [Results in Mathematics and Related Areas (2)], 34.

4480: \newblock Springer-Verlag, 1994.

4481: 

4482: \bibitem[N1]{N1}

4483: H.~Nakajima.

4484: \newblock Instantons on ALE spaces, quiver varieties, and Kac-Moody algebras.

4485: \newblock {\em Duke Math. J.} 76 (1994) no. 2, 365--416.

4486: 

4487: \bibitem[N2]{N2}

4488: H.~Nakajima.

4489: \newblock Varieties associated with quivers.

4490: \newblock {\em Canadian Math. Soc. Conf. Proc.} 19 (1996), 139--157.

4491: 

4492: \bibitem[N3]{N3}

4493: H.~Nakajima.

4494: \newblock Quiver varieties and Kac-Moody algebras.

4495: \newblock {\em Duke Math. J.} 91 (1998) no. 3, 515--560.

4496: 

4497: \bibitem[N4]{N4}

4498: H.~Nakajima.

4499: \newblock {\em Lectures on Hilbert schemes of points on surfaces}. 

4500: \newblock University Lecture Series, 18.

4501: \newblock American Mathematical Society, 1999.

4502: 

4503: \bibitem[N5]{N5}

4504: H.~Nakajima.

4505: \newblock Quiver varieties and finite-dimensional representations of quantum affine algebras. 

4506: \newblock {\em J. Amer. Math. Soc.} 14 (2001) no. 1, 145--238.

4507: 

4508: \bibitem[NS]{NS}

4509: M.~Narasimhan and C.~Seshadri.

4510: \newblock Stable bundles and unitary bundles on a compact Riemann surface.

4511: \newblock {\em Proc. Nat. Acad. Sci. U.S.A.} 52 (1964) 207--211.

4512: 

4513: \bibitem[OS]{OS}

4514: P.~Orlik and L.~Solomon.

4515: \newblock Combinatorics and topology of complements of hyperplanes.

4516: \newblock {\em Invent. Math.} 56 (1980), 167--189.

4517: 

4518: \bibitem[OT]{OT}

4519: P.~Orlik and H.~Terao.

4520: \newblock {\em Arrangements of Hyperplanes}.

4521: \newblock Grundlehren der mathematischen Wissenschaften, 300.

4522: \newblock Springer-Verlag, 1992.

4523: 

4524: % \bibitem[Pa]{Pa} P.~Paradan. 

4525: % \newblock The moment map and equivariant cohomology with generalized coefficients.

4526: % \newblock\emph{Topology} {39} (2000), no. 2, 401--444.

4527: % 

4528: \bibitem[Pr]{Pr}

4529: N.~Proudfoot.

4530: \newblock {\em The equivariant Orlik-Solomon algebra}.

4531: \newblock ArXiv:  math.CO/0306013.

4532: 

4533: \bibitem[Sc]{Sc}

4534: C.~Schmid.

4535: \newblock {\em Cohomologie \'equivariante des certaines vari\'et\'es

4536: hamiltoniennes et de leur partie r\'eelle.}

4537: \newblock Th\`ese \`a Universit\'e de Gen\`eve.

4538: \newblock Available at http://www.unige.ch/math/biblio/these/theses.html.

4539: 

4540: \bibitem[Si]{Si}

4541: C.~Simpson.

4542: \newblock Higgs bundles and local systems.  

4543: \newblock {\em Inst. Hautes Études Sci. Publ. Math.} 75 (1992), 5--95.

4544:  

4545: % \bibitem[Sj]{Sj}

4546: % R.~Sjamaar.

4547: % \newblock Holomorphic slices, symplecitic reduction, and multiplicities of representations.

4548: % \newblock {\em Ann. Math.} 141 (1995), 87--129.

4549: % 

4550: \bibitem[St]{St}

4551: R.~Stanley.

4552: \newblock The number of faces of a simplicial convex polytope.

4553: \newblock {\em Adv. in Math.} 35 (1980), no. 3, 236--238.

4554: 

4555: \bibitem[Th]{Th} M.~Thaddeus. 

4556: \newblock Private communication.

4557: 

4558: \bibitem[TW]{TW}

4559: S.~Tolman and J.~Weitsman.

4560: \newblock Cohomology rings of symplectic quotients.

4561: \newblock {\em Comm. An. Geom.} 11 (2003) no. 4, 751--773.

4562: \end{thebibliography}

4563: }

4564: % \end{spacing}

4565: \end{document}

4566: