1: \documentclass{ert-l}

2: \usepackage{amsmath, amssymb, amscd}

3: 

4: \newtheorem{thm}{Theorem}

5: \newtheorem{prop}[thm]{Proposition}

6: \newtheorem{lem}[thm]{Lemma}

7: \newtheorem{claim}[thm]{Claim}

8: \newtheorem{cor}[thm]{Corollary}

9: 

10: %\theoremstyle{remark}

11: \newtheorem{rem}{Remark}

12: 

13: %\theoremstyle{definition}

14: \newtheorem{df}{Definition}

15: \newtheorem{ex}[df]{Example}

16: 

17: \renewcommand{\phi}{\varphi}

18: \renewcommand{\epsilon}{\varepsilon}

19: 

20: \newcommand{\cal}{\mathcal}

21: \newcommand{\BB}{\mathbb}

22: \newcommand{\g}{\mathfrak}

23: \newcommand{\separate}{\vskip5pt}

24: \newcommand{\VF}{\operatorname{VF}}

25: \newcommand{\den}{\operatorname{Denominator}}

26: \newcommand{\talpha}{\widetilde{\alpha}}

27: \newcommand{\supp}{\operatorname{supp}}

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

29: \newcommand{\re}{\operatorname{Re}}

30: \newcommand{\tr}{\operatorname{Tr}}

31: \newcommand{\End}{\operatorname{End}}

32: \newcommand{\rhom}{\operatorname{RHom}}

33: \newcommand{\modg}{\operatorname{Mod}_{quasi-G_{\BB C}-eq}}

34: \newcommand{\modge}{\operatorname{Mod}_{G_{\BB C}-eq}}

35: \newcommand{\thom}{\operatorname{Thom}}

36: \newcommand{\parag}{${\mathcal x}$}

37: 

38: 

39: \begin{document}

40: 

41: \title[Riemann-Roch-Hirzebruch integral formula]

42: {Riemann-Roch-Hirzebruch integral formula for characters of

43: reductive Lie groups}

44: 

45: \author{Matvei Libine}

46: 

47: \address{Department of Mathematics and Statistics, University of Massachusetts,

48: Lederle Graduate Research Tower, 710 North Pleasant Street, Amherst, MA 01003}

49: 

50: \email{matvei@math.umass.edu}

51: 

52: \keywords{

53: equivariant sheaves and ${\cal D}$-modules,

54: characteristic cycles of sheaves and ${\cal D}$-modules,

55: integral character formula, fixed point integral localization formula,

56: fixed point character formula, representations of reductive Lie groups,

57: equivariant forms,}

58: 

59: \subjclass[2000]{Primary 22E45; Secondary 32C38, 19L10, 55N91}

60: 

61: \begin{abstract}

62: Let $G_{\BB R}$ be a real reductive Lie group acting on a manifold $M$.

63: M.~Kashiwara and W.~Schmid in \cite{KaSchm} constructed representations

64: of $G_{\BB R}$ using sheaves and quasi-$G_{\BB R}$-equivariant

65: ${\cal D}$-modules on $M$.

66: In this article we prove an integral  character formula for these

67: representations (Theorem \ref{main}).

68: Our main tools will be the integral localization formula recently proved

69: in \cite{L3} and the integral character formula proved by W.~Schmid and

70: K.~Vilonen in \cite{SchV2} (originally established by W.~Rossmann in \cite{Ro})

71: in the important special case when the manifold $M$ is the flag variety of

72: $\BB C \otimes_{\BB R} \g g_{\BB R}$ -- the complexified Lie algebra of

73: $G_{\BB R}$.

74: In the special case when $G_{\BB R}$ is commutative and the

75: ${\cal D}$-module is the sheaf of sections of a

76: $G_{\BB R}$-equivariant line bundle over $M$ this integral character

77: formula will reduce to the classical Riemann-Roch-Hirzebruch formula.

78: As an illustration we give a concrete example on the enhanced flag variety.

79: \end{abstract}

80: 

81: \maketitle

82: 

83: \tableofcontents

84: 

85: \begin{section}

86: {Introduction}

87: \end{section}

88: 

89: We consider pairs of Lie groups: a real linear reductive group $G_{\mathbb R}$

90: sitting inside its complexification $G_{\mathbb C}$.

91: For example:

92: $$

93: \begin{matrix}

94: GL(n, \mathbb R) & \subset & GL(n, \mathbb C) \\

95: GL^+(n, \mathbb R) & \subset & GL(n, \mathbb C) \\

96: U(n) & \subset & GL(n, \mathbb C)

97: \end{matrix}

98: \qquad

99: \begin{matrix}

100: SL(n, \mathbb R) & \subset & SL(n, \mathbb C) \\

101: SO(n, \mathbb C) & \subset & SL(n, \mathbb C)  \\

102: SU(n) & \subset & SL(n, \mathbb C)

103: \end{matrix}

104: \qquad

105: Sp(n, \mathbb R) \subset Sp(n, \mathbb C).

106: $$

107: Suppose that $G_{\mathbb C}$ acts algebraically on a smooth complex

108: projective variety $M$.

109: Fix a $G_{\mathbb C}$-invariant open algebraic subset $U \subset M$,

110: and take a $G_{\mathbb C}$-equivariant algebraic line bundle

111: $({\bf E}, \nabla_{\bf E})$ over $U$ with a $G_{\mathbb C}$-invariant

112: algebraic flat connection $\nabla_{\bf E}$.

113: %The flat connection allows us to regard the sheaf of sections

114: %${\cal O}({\bf E})$ as a ${\cal D}$-module over $U$ and also over $M$.

115: Let $S_{\mathbb R} \subset M$ be an open $G_{\mathbb R}$-invariant subset

116: (which may or may not be $G_{\mathbb C}$-invariant) and

117: consider the cohomology spaces

118: \begin{equation}  \label{hspaces}

119: H^p(S_{\mathbb R}, {\cal O}({\bf E})), \qquad p \in \mathbb Z.

120: \end{equation}

121: 

122: The classical Riemann-Roch-Hirzebruch formula computes the index of

123: ${\bf E}$, i.e. the alternating sum

124: $$

125: \sum_p (-1)^p \dim H^p(S_{\mathbb R}, {\cal O}({\bf E}))

126: $$

127: with $S_{\mathbb R} = U = M$.

128: For general $S_{\mathbb R}$ and $U$, however, these dimensions can be infinite.

129: To work around this problem we regard the vector spaces (\ref{hspaces})

130: as representations of $G_{\mathbb R}$, and, as a substitute for the index,

131: we ask for the character of the virtual representation

132: $$

133: \sum_p (-1)^p H^p(S_{\mathbb R}, {\cal O}({\bf E})).

134: $$

135: (Recall that for finite-dimensional representations the value of the character

136: at the identity element $e \in G$ equals the dimension of the representation.)

137: 

138: In this article we establish an integral formula for characters of

139: virtual representations of this kind in a more general setting

140: of ${\cal D}$-modules.

141: Existence of such integral character formula was conjectured by W.~Schmid.

142: My work toward this formula has led to the integral localization formula

143: proved in \cite{L3}.

144: 

145: We use the following convention: whenever $A$ is a subset of $B$,

146: we denote the inclusion map $A \hookrightarrow B$ by $j_{A \hookrightarrow B}$.

147: 

148: \separate

149: 

150: \begin{section}

151: {Preliminary results}

152: \end{section}

153: 

154: Let $G_{\BB C}$ be a connected complex algebraic reductive group

155: which is defined over $\BB R$.

156: We will be primarily interested in representations of a subgroup

157: $G_{\BB R}$ of $G_{\BB C}$ lying between the group of real points

158: $G_{\BB C}(\BB R)$ and the identity component $G_{\BB C}(\BB R)^0$.

159: We regard $G_{\BB R}$ as a real reductive Lie group.

160: Let $\g g_{\BB C}$ and $\g g_{\BB R}$ be the respective Lie algebras

161: of $G_{\BB C}$ and $G_{\BB R}$.

162: Let ${\cal B}$ denote the flag variety of $\g g_{\mathbb C}$.

163: It is a smooth complex projective variety consisting of all Borel

164: subalgebras of $\g g_{\BB C}$. The group $G_{\mathbb C}$ acts on ${\cal B}$

165: transitively.

166: 

167: If the group $G_{\mathbb R}$ is compact, then all irreducible representations

168: of $G_{\mathbb R}$ can be enumerated by their highest weights $\lambda$

169: lying in the weight lattice $\Lambda \subset i \g g_{\mathbb R}^*$.

170: The Borel-Weil-Bott theorem can be regarded as an explicit

171: construction of a holomorphic $G_{\mathbb R}$-equivariant line bundle

172: ${\cal L}_{\lambda} \to {\cal B}$ such that the resulting

173: representation of $G_{\mathbb R}$ in the cohomology groups is:

174: \begin{eqnarray*}

175: && H^p({\cal B}, {\cal O}({\cal L}_{\lambda}))=0 \text{\quad if $p \ne 0$}, \\

176: && H^0({\cal B}, {\cal O}({\cal L}_{\lambda})) \simeq \pi_{\lambda},

177: \end{eqnarray*}

178: where ${\cal O}({\cal L}_{\lambda})$ is the sheaf of sections of

179: ${\cal L}_{\lambda}$ and

180: $\pi_{\lambda}$ denotes the irreducible representation of

181: $G_{\mathbb R}$ of highest weight $\lambda$.

182: Then N.~Berline and M.~Vergne \cite{BV}, \cite{BGV} observed that the

183: character of $\pi_{\lambda}$, as a function on $\g g_{\mathbb R}$,

184: can be expressed as an integral over ${\cal B}$ of a certain naturally

185: defined equivariantly closed form.

186: They proved it by applying their famous integral localization

187: formula and matching contributions from fixed points with terms

188: of the Weyl character formula.

189: (This is a restatement of Kirillov's character formula.)

190: 

191: \separate

192: 

193: M.~Kashiwara and W.~Schmid \cite{KaSchm}

194: generalize the Borel-Weil-Bott construction.

195: Instead of line bundles over the flag variety ${\cal B}$ they consider

196: $G_{\mathbb R}$-equivariant sheaves\footnote{Strictly speaking,

197: ${\cal F}$ is not a sheaf on ${\cal B}$

198: but rather an element of the ``$G_{\mathbb R}$ equivariant derived category

199: on ${\cal B}$ with twist $(-\lambda-\rho)$'' denoted by

200: $\operatorname{D}_{G_{\BB R}} ({\cal B})_{-\lambda}$

201: (see \cite{SchV2} and Remark \ref{twistings}).}

202: ${\cal F}$ and, for each integer

203: $p \in \BB Z$, they define representations of $G_{\mathbb R}$ in

204: $\operatorname{Ext}^p({\cal F},{\cal O}_{{\cal B}^{an}})$, where

205: ${\cal O}_{{\cal B}^{an}}$ denotes the sheaf of analytic functions on

206: ${\cal B}$.

207: In other words, M.~Kashiwara and W.~Schmid prove that these vector spaces

208: possess a natural Fr\'echet topology and the resulting representations of

209: $G_{\mathbb R}$ are admissible of finite length.

210: Hence these representations have characters in the sense of Harish-Chandra.

211: Let $\tilde \theta$ be the character of the virtual representation of

212: $G_{\mathbb R}$

213: $$

214: \sum_p (-1)^p \operatorname{Ext}^p

215: \bigl( \BB D_{\cal B} {\cal F}, {\cal O}_{{\cal B}^{an}}(\lambda) \bigr),

216: $$

217: where $\BB D_{\cal B} {\cal F}$ denotes the Verdier dual of ${\cal F}$,

218: ${\cal O}_{{\cal B}^{an}}$ 

219: and $\lambda$ is some twisting parameter lying in $\g h_{\BB C}^*$ --

220: the dual space of the universal Cartan algebra of $\g g_{\BB C}$.

221: Here the character $\tilde \theta$ is a distribution on $\g g_{\mathbb R}$.

222: Then W.~Schmid and K.~Vilonen \cite{SchV2} prove

223: two character formulas for $\tilde \theta$.

224: 

225: The integral character formula expresses $\tilde \theta$ as an integral of a

226: certain differential form (independent of ${\cal F}$) over the characteristic

227: cycle $Ch({\cal F})$ of ${\cal F}$.

228: Characteristic cycles were introduced by M.~Kashiwara and their

229: definition can be found in \cite{KaScha}. On the other hand,

230: W.~Schmid and K.~Vilonen give a geometric way to understand

231: characteristic cycles in \cite{SchV1}.

232: A comprehensive treatment of characteristic cycles can be found in \cite{Schu}.

233: The cycle $Ch({\cal F})$ is a conic Lagrangian Borel-Moore homology cycle

234: lying inside the cotangent space $T^*{\cal B}$.

235: If the sheaf ${\cal F}$ happens to be perverse,

236: the characteristic cycle of ${\cal F}$ equals the characteristic cycle of the

237: holonomic ${\cal D}$-module corresponding to ${\cal F}$ via the

238: Riemann-Hilbert correspondence.

239: Originally existence of such

240: formula over an unspecified cycle was established by W.~Rossmann in \cite{Ro}.

241: In the special case when the group $G_{\mathbb R}$ is compact, the integral

242: character formula reduces to Kirillov's character formula.

243: 

244: On the other hand, Harish-Chandra showed that the distribution $\tilde\theta$

245: is given by integration against a certain function $F_{\tilde\theta}$

246: on $\g g_{\mathbb R}$.

247: This function $F_{\tilde\theta}$ is an $Ad(G_{\mathbb R})$-invariant locally

248: $L^1$-function, its restriction to the set of regular semisimple

249: elements $\g g_{\mathbb R}^{rs}$ of $\g g_{\mathbb R}$ can be represented

250: by an analytic function.

251: According to Harish-Chandra, $F_{\tilde\theta}$ can be expressed as follows.

252: For a regular semisimple element $X \in \g g_{\mathbb R}^{rs}$ we denote by

253: $\g t_{\mathbb C}(X) \subset \g g_{\mathbb C}$ the unique (complex) Cartan

254: algebra containing $X$ and by $\Psi(X) \subset \g t_{\mathbb C}(X)^*$ we

255: denote the root system of $\g g_{\mathbb C}$ with respect to

256: $\g t_{\mathbb C}(X)$.

257: Since $X$ is regular semisimple, the number of Borel subalgebras

258: $\{\g b_1,\dots,\g b_{|W|}\}$ containing $X$ is exactly the order of the

259: Weyl group $W$.

260: Each Borel subalgebra $\g b_k \in {\cal B}$ containing $X$

261: (and hence containing $\g t_{\mathbb C}(X)$) determines a positive root system 

262: $\Psi_{\g b_k}^+(X) \subset \Psi(X)$ consisting of all those roots whose root

263: spaces are {\em not} contained in $\g b_k$;

264: that is $\g b_k$ contains all negative root spaces relative to

265: $\Psi_{\g b_k}^+(X)$. Then

266: \begin{equation}  \label{fpcharformula}

267: F_{\tilde\theta} (X) =

268: \sum_{k=1}^{|W|}

269: m_{\g b_k}(X) \cdot \frac {e^{\langle X,\lambda_{\g b_k} \rangle}}

270: {\prod_{\alpha \in \Psi_{\g b_k}^+(X)} \alpha(X)},

271: \end{equation}

272: where $m_{\g b_k}(X)$'s are some integer multiplicities.

273: W.~Schmid and K.~Vilonen give a formula for $m_{\g b_k}(X)$'s in terms of

274: local cohomology of ${\cal F}$ (formula (5.25b) in \cite{SchV2}).

275: Pick another positive root system $\Psi^{\le}(X) \subset \Psi(X)$ such that

276: $$

277: \re \alpha (X) \le 0 \qquad \text{for all $\alpha \in \Psi^{\le}(X)$.}

278: $$

279: (If there are roots in $\Psi$ which take purely imaginary values on $X$,

280: there will be different possible choices for $\Psi^{\le}(X)$.)

281: Let $B(X) \subset G_{\BB C}$ be the Borel subgroup whose Lie algebra

282: contains $\g t_{\BB C}(X)$ and the positive root spaces corresponding

283: to $\Psi^{\le}(X)$.

284: The action of $B(X)$ on ${\cal B}$ has precisely $|W|$ orbits

285: $O_1, \dots, O_{|W|}$. Each of these orbits $O_k$ is a locally closed

286: subset of ${\cal B}$ and contains exactly one of the Borel subalgebras

287: $\{\g b_1,\dots,\g b_{|W|}\}$ containing $X$.

288: We order the Borel subalgebras containing $X$ and the orbits of $B(X)$ so that

289: $\g b_k$ is contained in $O_k$, $k=1,\dots,|W|$.

290: Then

291: \begin{equation}  \label{m}

292: m_{\g b_k}(X) = \chi

293: \bigl( {\cal H}^*_{O_k} (\mathbb D_{\cal B} {\cal F})_{\g b_k} \bigr)

294: = \chi \bigl( (j_{\{\g b_k\} \hookrightarrow O_k})^* \circ

295: (j_{O_k \hookrightarrow {\cal B}})^! \mathbb D_{\cal B} {\cal F}) \bigr)

296: \end{equation}

297: (recall that $\BB D_{\cal B} {\cal F}$ denotes the Verdier dual of ${\cal F}$).

298: These multiplicities are exactly the local contributions of points

299: $\g b_k \in {\cal B}$ to the Lefschetz fixed point formula, as generalized

300: to sheaf cohomology by M.~Goresky and R.~MacPherson \cite{GM}.

301: W.~Schmid and K.~Vilonen call (\ref{fpcharformula}) combined with (\ref{m})

302: the fixed point character formula because the set of all Borel subalgebras

303: $\{\g b_1,\dots,\g b_{|W|}\}$ containing $X$ can be expressed as the set of

304: zeroes of the vector field generated by the infinitesimal action of

305: $X$ on ${\cal B}$.

306: In the special case when the group $G_{\mathbb R}$ is compact,

307: and  $\lambda+\rho$ is an integral weight, all the multiplicities

308: $m_{\g b_k}(X)$ are equal to each other, say,

309: $$

310: m_{\g b_1}(X) = \dots = m_{\g b_{|W|}}(X) = \kappa \in \mathbb Z,

311: $$

312: and (\ref{fpcharformula}) reduces to $\kappa$ times the Weyl character formula.

313: The fixed point formula was conjectured by M.~Kashiwara

314: \cite{Ka}, and its proof uses the above-mentioned generalization of the

315: Lefschetz fixed point formula to sheaf cohomology \cite{GM}.

316: 

317: 

318: 

319: There is a striking relationship between these two character formulas.

320: In the compact group case N.~Berline and M.~Vergne \cite{BV}, \cite{BGV}

321: gave a simple proof of Kirillov's character formula using their

322: integral localization formula; they matched contributions from zeroes of

323: vector fields with terms of the Weyl character formula.

324: However, in the non-compact group case their argument breaks down because

325: their localization formula works for compact groups only.

326: Originally W.~Schmid and K.~Vilonen \cite{SchV2} proved these

327: character formulas independently of each other using representation

328: theory methods. Thus, besides an important representation-theoretical result,

329: they formally established existence of an integral localization formula

330: in a very special case.

331: In the announcement \cite{Sch} W.~Schmid posed a question:

332: ``Can this equivalence of character formulas be seen directly without

333: a detour to representation theory, just as in the compact case.''

334: In \cite{L1}, \cite{L2} I provide such a geometric link, then in \cite{L3}

335: I establish a general integral localization formula for non-compact group

336: actions.

337: In turn, this article uses the new localization formula for

338: non-compact group actions to give a generalization of the integral character

339: formula to representations associated to sheaves on manifolds other than the

340: flag manifold of $\g g_{\mathbb C}$.

341: 

342: \separate

343: 

344: We describe the integral character formula for $\tilde \theta$ in more detail;

345: its ingredients will be used in our character formula (\ref{mainintegral}).

346: The character $\tilde\theta$ is a distribution on

347: $\Omega_c^{top}(\g g_{\BB R})$ -- the space of smooth compactly

348: supported differential forms on $\g g_{\BB R}$ of top degree.

349: If $\phi \in \Omega_c^{top}(\g g_{\BB R})$, then we define its Fourier

350: transform $\hat \phi \in {\cal C}^{\infty}(\g g_{\BB C}^*)$

351: as in \cite{L1}, \cite{L2}, \cite{L3}, \cite{Ro} and \cite{SchV2}:

352: \begin{equation}  \label{FT}

353: \hat \phi (\xi) = \int_{\g g_{\BB R}}

354: e^{\langle X, \xi \rangle} \phi(X),

355: \qquad X \in \g g_{\BB R}, \: \xi \in \g g_{\BB C}^*,

356: \end{equation}

357: without the customary factor of $i = \sqrt{-1}$ in the exponent.

358: 

359: For a smooth complex algebraic manifold $M$ we denote by $\sigma_M$ the

360: canonical complex symplectic form on the holomorphic cotangent space $T^*M$.

361: In general, when the group $G_{\mathbb C}$ acts on a complex manifold $M$,

362: for $X \in \g g_{\BB C}$, we denote by $X_M$ the vector field on

363: $M$ given by (notice the minus sign)

364: \begin{equation}  \label{vfield}

365: (X_M \cdot \phi) (x) =

366: \frac d{d\epsilon} \phi \bigl( \exp (-\epsilon X)x \bigr) \Bigr|_{\epsilon=0},

367: \qquad \phi \in {\cal C}^{\infty}(M).

368: \end{equation}

369: Then the {\em moment map} on the holomorphic cotangent space

370: $\mu_M : T^*M \to \g g_{\BB C}^*$ is defined by

371: \begin{equation}  \label{mu}

372: \mu_M (\zeta): X \mapsto - \langle \zeta, X_M \rangle,

373: \qquad X \in \g g_{\BB C},\: \zeta \in T^*M.

374: \end{equation}

375: When the manifold $M$ is the flag variety ${\cal B}$ we get the complex

376: algebraic symplectic form $\sigma_{\cal B}$ on $T^*{\cal B}$, the moment map

377: $\mu_{\cal B}: T^*{\cal B} \to \g g_{\BB C}^*$ and the vector field

378: $X_{\cal B}$ for each element $X \in \g g_{\mathbb C}$.

379: 

380: We fix a compact real form (i.e. a maximal compact subgroup)

381: $U_{\BB R} \subset G_{\BB C}$ with Lie algebra $\g u_{\mathbb R}$.

382: We will use Rossmann's map

383: ${\cal B} \ni x \mapsto \lambda_x \in \g g_{\BB C}^*$

384: defined in \cite{Ro} and Section 8 of \cite{SchV1}

385: (here we use the notations of \cite{SchV1} and \cite{SchV2}).

386: Recall that the twisting parameter $\lambda$ is an element of the dual of

387: the universal Cartan algebra $\g h_{\mathbb C}$ of $\g g_{\mathbb C}$; 

388: $\g h_{\mathbb C}$ is not a subalgebra of $\g g_{\mathbb C}$, but is

389: canonically isomorphic to any Cartan subalgebra

390: $\g t_{\mathbb C} \subset \g g_{\mathbb C}$ equipped with a specified choice

391: of positive root system $\Phi^+$.

392: For $x \in {\cal B}$, we denote by $\g b_x \subset \g g_{\mathbb C}$ the Borel

393: subalgebra corresponding to $x$. Then $\g h_{\mathbb C}$ is canonically

394: isomorphic to the quotient $\g b_x / [\g b_x, \g b_x]$, so that $\g b_x$

395: contains all the negative root spaces.

396: Thus we have an exact sequence of vector spaces

397: \begin{equation}  \label{split}

398: 0 \to [\g b_x, \g b_x] \to \g b_x \to \g h_{\mathbb C} \to 0.

399: \end{equation}

400: In general, this sequence does not have a canonical splitting.

401: But once a choice of a compact real form $U_{\mathbb R} \subset G_{\BB C}$

402: is made, we can split this sequence as follows.

403: Let $T_{\mathbb R}(x)$ be the stabilizer of $x$ in $U_{\mathbb R}$, it is

404: a maximal torus in $U_{\mathbb R}$, and set

405: $$

406: \g t_{\mathbb R}(x) = Lie (T_{\mathbb R}(x)) \subset \g u_{\mathbb R} \subset

407: \g g_{\mathbb C},

408: \qquad

409: \g t_{\mathbb C}(x) = \g t_{\mathbb R}(x) \otimes_{\mathbb R} \mathbb C

410: \subset \g g_{\mathbb C}.

411: $$

412: Then $\g t_{\mathbb C}(x)$ is a Cartan subalgebra of $\g g_{\mathbb C}$

413: which lies in $\g b_x$ and comes equipped with a system of positive roots

414: $\Psi^+_{\g b_x}$ so that $\g b_x$ contains all the negative root spaces.

415: Hence we get a composition of maps

416: $$

417: \g h_{\mathbb C} \tilde \rightarrow \g t_{\mathbb C}(x) \hookrightarrow 

418: \g b_x

419: $$

420: which splits (\ref{split}):

421: $$

422: \g b_x \simeq \g t_{\mathbb C}(x) \oplus [\g b_x, \g b_x]

423: \simeq \g h_{\mathbb C} \oplus [\g b_x, \g b_x].

424: $$

425: We also get a splitting of $\g g_{\mathbb C}$:

426: $$

427: \g g_{\mathbb C} \simeq \g t_{\mathbb C}(x) \oplus

428: \Bigl(\bigoplus \text{root spaces of $\g t_{\mathbb C}(x)$} \Bigr)

429: \simeq \g h_{\mathbb C} \oplus

430: \Bigl(\bigoplus \text{root spaces of $\g t_{\mathbb C}(x)$} \Bigr).

431: $$

432: Taking duals we get a splitting of $\g g_{\mathbb C}^*$:

433: $$

434: \g g_{\mathbb C}^* \simeq \g t_{\mathbb C}(x)^* \oplus

435: \Bigl(\bigoplus \text{root spaces of $\g t_{\mathbb C}(x)$} \Bigr)^*

436: \simeq \g h_{\mathbb C}^* \oplus

437: \Bigl(\bigoplus \text{root spaces of $\g t_{\mathbb C}(x)$} \Bigr)^*.

438: $$

439: Hence $\lambda \in \g h_{\mathbb C}^*$ gets identified via this splitting

440: with an element $\lambda_x \in \g g_{\mathbb C}^*$.

441: The map $\lambda_x: {\cal B} \to \g g_{\BB C}^*$ is smooth, real algebraic

442: and $U_{\mathbb R}$-equivariant, but in general not $G_{\mathbb C}$- or

443: $G_{\mathbb R}$-equivariant, nor complex algebraic.

444: When $\lambda$ is regular, the {\em twisted moment map}

445: $$

446: \mu_{\lambda} =_{def} \mu_{\cal B} + \lambda_x :

447: T^*{\cal B} \to \g g_{\mathbb C}^*

448: $$

449: is a real algebraic $U_{\mathbb R}$-equivariant diffeomorphism of $T^*{\cal B}$

450: onto the coadjoint orbit of $G_{\mathbb C}$ of any point

451: $\lambda_x \in \g g_{\mathbb C}^*$.

452: 

453: We will also use a $U_{\BB R}$-invariant 2-form $\tau_{\lambda}$ on

454: ${\cal B}$ defined by the formula

455: $$

456: \tau_{\lambda}(X_x,Y_x) = \lambda_x([X,Y]),

457: $$

458: where $X_x, Y_x \in T_x{\cal B}$ are the tangent vectors at $x \in {\cal B}$

459: induced by $X, Y \in \g u_{\BB R}$ via differentiation of the

460: $U_{\BB R}$-action.

461: It is an important property of $\lambda_x$ and $\tau_{\lambda}$

462: that together they make a (non-homogeneous) differential form on ${\cal B}$

463: $$

464: \langle X, \lambda_x \rangle - \tau_{\lambda}, \qquad X \in \g g_{\BB C},

465: $$

466: which depends holomorphically on $X$ (in fact linearly) and which is

467: {\em equivariantly closed} with respect to $U_{\BB R}$

468: (see \cite{BGV}, \cite{GS} or \cite{L3} for the definition of equivariantly

469: closed forms). It follows that the forms

470: $$

471: e^{\langle X, \lambda_x \rangle - \tau_{\lambda}}

472: \qquad \text{and} \qquad

473: e^{-\langle X, \lambda_x \rangle + \tau_{\lambda}}

474: $$

475: are $U_{\BB R}$-equivariantly closed forms too.

476: For a differential form $\omega$ which is possibly non-homogeneous,

477: we denote by $\omega_{[k]}$ its homogeneous component of degree $k$.

478: Then the integral character formula (Theorem 3.8 in \cite{SchV2})

479: says that the value of the character

480: $\tilde\theta$ at $\phi \in \Omega_c^{top}(\g g_{\BB R})$ is given by

481: \begin{multline}  \label{intcharformula}

482: \theta (\phi) =

483: \frac 1{(2\pi i)^dd!} \int_{Ch({\cal F})}

484: (\hat\phi \circ \mu_{\lambda}) \cdot (-\sigma_{\cal B}+\tau_{\lambda})^d  \\

485: = \frac 1{(-2\pi i)^d} \int_{Ch({\cal F})} \Bigl( \int_{\g g_{\BB R}}

486: e^{\langle X, \mu_{\cal B}(\zeta) + \lambda_x \rangle

487: + \sigma_{\cal B} - \tau_{\lambda}}

488: \wedge \phi(X) \Bigr)_{[2d]},  \\

489: \qquad X \in \g g_{\BB R},\: \zeta \in T^*{\cal B},

490: \end{multline}

491: where $\mu_{\lambda} = \mu_{\cal B} + \lambda_x$ and

492: $d = \dim_{\mathbb C} {\cal B}$.

493: This integral converges because the expression $\hat\phi \circ \mu_{\lambda}$

494: involves the Fourier transform of $\phi$

495: and decays rapidly along the support of $Ch({\cal F})$.

496: 

497: 

498: 

499: \separate

500: 

501: \begin{section}

502: {Setup}

503: \end{section}

504: 

505: Recall that $G_{\BB C}$ is a connected complex algebraic reductive group

506: which is defined over $\BB R$, $G_{\BB C}$ acts algebraically on a

507: smooth complex projective variety $M$.

508: And we are primarily interested in representations of a real reductive

509: group $G_{\BB R} \subset G_{\BB C}$ lying between the group of real points

510: $G_{\BB C}(\BB R)$ and the identity component $G_{\BB C}(\BB R)^0$.

511: 

512: We will be using the concepts of ${\cal D}$-modules and derived categories;

513: \cite{Bo} and \cite{KaScha} are good introductions to these subjects.

514: Let ${\cal O}_M$ denote the structure sheaf on $M$ and

515: let ${\cal D}_M$ denote the sheaf of linear differential operators on $M$

516: with algebraic coefficients; ${\cal D}_M$ acts on ${\cal O}_M$.

517: The definition of a quasi-equivariant ${\cal D}$-module can be found

518: in many different sources; for convenience, we copy the definition given in

519: \cite{KaSchm}.

520: 

521: Let $G_{\BB C}$ be a connected complex algebraic reductive group

522: acting algebraically on a smooth complex algebraic variety $M$,

523: $\g g_{\mathbb C} = Lie(G_{\BB C})$.

524: We write $\mu$ for the action morphism and $\pi$ for the projection map

525: $G_{\BB C} \times M \to M$:

526: $$

527: \mu, \pi: G_{\BB C} \times M \to M,

528: \quad \mu(g, x) = gx, \quad \pi(g, x) = x

529: \qquad g \in G_{\BB C},\: x \in M.

530: $$

531: We also consider three maps

532: $$

533: q_j: G_{\BB C} \times G_{\BB C} \times M \to G_{\BB C} \times M,

534: \qquad j=1,2,3,

535: $$

536: $$

537: q_1(g_1,g_2,x)=(g_1,g_2x), \qquad

538: q_2(g_1,g_2,x)=(g_1g_2,x), \qquad

539: q_3(g_1,g_2,x)=(g_2,x).

540: $$

541: Then we have the following identities:

542: \begin{align*}

543: \mu \circ q_1 = \mu \circ q_2 &: \quad (g_1,g_2,x) \mapsto g_1g_2x,  \\

544: \pi \circ q_2 = \pi \circ q_3 &: \quad (g_1,g_2,x) \mapsto x, \\

545: \mu \circ q_3 = \pi \circ q_1 &: \quad (g_1,g_2,x) \mapsto g_2x.

546: \end{align*}

547: 

548: \begin{df}

549: We denote by ${\cal O}_{G_{\mathbb C}} \boxtimes {\cal D}_M$ the subalgebra

550: ${\cal O}_{G_{\mathbb C} \times M} \otimes_{\pi^{-1} {\cal O}_M}

551: \pi^{-1}{\cal D}_M$ of ${\cal D}_{G_{\mathbb C} \times M}$.

552: A {\em quasi-$G_{\BB C}$-equivariant ${\cal D}_M$-module} is a

553: ${\cal D}_M$-module $\g M$ equipped with an

554: ${\cal O}_{G_{\mathbb C}} \boxtimes {\cal D}_M$-linear

555: isomorphism $\beta: \mu^* \g M \,\tilde\longrightarrow\, \pi^* \g M$, such that

556: the composition of isomorphisms

557: $$

558: \begin{CD}

559: q_2^* \mu^* \g M  \simeq q_1^* \mu^* \g M @>{q_1^*\beta}>>

560: q_1^* \pi^* \g M  \simeq  q_3^* \mu^* \g M  @>{q_3^*\beta}>> q_3^* \pi^* \g M

561: \simeq q_2^* \pi^* \g M

562: \end{CD}

563: $$

564: coincides with

565: $$

566: \begin{CD}

567: q_2^* \mu^* \g M   @>{q_2^*\beta}>> q_2^* \pi^* \g M.

568: \end{CD}

569: $$

570: If $\beta$ is linear even over ${\cal D}_{G_{\mathbb C} \times M}$,

571: this reduces to the usual definition of a {\em $G_{\mathbb C}$-equivariant

572: ${\cal D}_M$-module}.

573: \end{df}

574: 

575: Informally speaking, this definition can be interpreted as follows.

576: For each $g \in G_{\mathbb C}$, denote by $\mu_g :M \to M$ the translation

577: by $g$, i.e. $\mu_g: x \mapsto gx$.

578: Then $\beta$ consists of a family of isomorphisms of ${\cal D}_M$-modules

579: $\beta_g: \mu_g^* \g M \,\tilde\longrightarrow\, \g M$,

580: depending algebraically on $g$ and multiplicative in the variable $g$.

581: 

582: 

583: \begin{ex}  \label{line_bundle}

584: Let $({\bf E}, \nabla_{\bf E})$ be a $G_{\BB C}$-equivariant algebraic line

585: bundle over a $G_{\BB C}$-invariant open algebraic subset

586: $U \subset M$ with a $G_{\BB C}$-invariant algebraic flat connection

587: $\nabla_{\bf E}$.

588: Take $\g M$ to be the direct image of the sheaf of sections

589: ${\cal O}({\bf E})$, under the inclusion map $U \hookrightarrow M$.

590: Then ${\cal D}_M$ acts on ${\cal O}({\bf E})$ via the flat

591: connection making ${\cal O}({\bf E})$ a quasi-$G_{\BB C}$-equivariant

592: ${\cal D}_M$-module.

593: 

594: Each element $X \in \g g_{\mathbb C}$ acts on $\g M$ in two ways.

595: One way is by inducing the vector field $-X_M$ given by (\ref{vfield})

596: which in turn operates on $\g M$ via the connection.

597: The other way is by infinitesimal translation of the sections of the

598: $G_{\BB C}$-equivariant line bundle ${\bf E}$.

599: When these two actions of $\g g_{\mathbb C}$ coincide, the ${\cal D}_M$-module

600: $\g M$ is $G_{\mathbb C}$-equivariant.

601: \end{ex}

602: 

603: 

604: Let $\g M$ be a coherent, quasi-$G_{\BB C}$-equivariant ${\cal D}_M$-module.

605: Recall that, for $X \in \g g_{\BB C}$, $X_M$ denotes the vector field on $M$

606: given by (\ref{vfield}).

607: Following \cite{KaSchm} we get two different actions of $\g g_{\BB C}$ on

608: $\g M$. The first action is via

609: $\g g_{\BB C} \ni X \mapsto -X_M \in \Gamma({\cal D}_M)$

610: -- the global sections of ${\cal D}_M$ -- followed by the ${\cal D}_M$-module

611: structure; we denote this action by

612: $\alpha_{\cal D}$.

613: And the second action is through differentiation of the $G_{\BB C}$-action

614: when we regard $\g M$ as a $G_{\BB C}$-equivariant ${\cal O}_M$-module;

615: this action is denoted by $\alpha_t$.

616: We set $\gamma = \alpha_t - \alpha_{\cal D}$, then $\gamma$ is a Lie algebra

617: homomorphism

618: \begin{equation}  \label{gamma}

619: \gamma: \g g_{\BB C} \longrightarrow \End_{{\cal D}_M} (\g M).

620: \end{equation}

621: This way $\g M$ becomes a $({\cal D}_M, {\cal U}(\g g_{\BB C}))$-module,

622: where ${\cal U}(\g g_{\BB C})$ denotes the universal enveloping algebra

623: of $\g g_{\BB C}$.

624: The quasi-$G_{\BB C}$-equivariant ${\cal D}_M$-module $\g M$ is

625: $G_{\BB C}$-equivariant precisely when $\gamma \equiv 0$.

626: 

627: We say that the quasi-$G_{\BB C}$-equivariant ${\cal D}_M$-module $\g M$ is

628: {\em ${\cal Z}(\g g_{\BB C})$-finite} if some ideal of finite codimension

629: ${\cal I} \subset {\cal Z}(\g g_{\BB C})$ (the center of 

630: ${\cal U}(\g g_{\BB C})$) annihilates $\g M$ via the $\gamma$-action.

631: 

632: We denote by $Ch(\g M)$ the characteristic cycle of

633: $\g M$ which lies in $T^*M$.

634: Pick a Borel subalgebra $\g b_{\BB C} \subset \g g_{\BB C}$

635: and define a subset of $\g g_{\BB C}^*$

636: $$

637: \g b_{\BB C}^{\perp} =

638: \{ \xi \in \g g_{\BB C}^* ;\: \xi|_{\g b_{\BB C}} \equiv 0 \}.

639: $$

640: Then the ${\cal D}_M$-module $\g M$ is called {\em admissible} if

641: $$

642: Ch(\g M) \cap \mu_M^{-1} (\g b_{\BB C}^{\perp}) \quad \subset \quad T^*M

643: $$

644: is a Lagrangian variety.

645: When $\g M$ is ${\cal Z}(\g g_{\BB C})$-finite,

646: the variety $Ch(\g M) \cap \mu_M^{-1} (\g b_{\BB C}^{\perp})$

647: is known to be involutive (\cite{KaMF} or \cite{Gi}),

648: hence the above condition is equivalent to

649: $$

650: \dim_{\BB C} \bigl( Ch(\g M) \cap \mu_M^{-1} (\g b_{\BB C}^{\perp}) \bigr)

651: = \dim_{\BB C} M.

652: $$

653: Because of $G_{\BB C}$-invariance of $Ch(\g M)$, if this condition is

654: satisfied for one particular Borel subalgebra

655: $\g b_{\BB C} \subset \g g_{\BB C}$,

656: then it is satisfied for all Borel subalgebras of $ \g g_{\BB C}$,

657: and this definition is independent of the choice of the Borel subalgebra

658: $\g b_{\BB C}$.

659: 

660: 

661: \separate

662: 

663: Let ${\cal S}$ be a $G_{\BB R}$-equivariant constructible sheaf on $M$.

664: We denote by ${\cal O}_{M^{an}}$ the sheaf of holomorphic functions on $M$.

665: Then M.~Kashiwara and W.~Schmid \cite{KaSchm} equip the vector spaces

666: \begin{equation} \label{ext}

667: \rhom^p_{{\cal D}_M} ( \g M \otimes {\cal S}, {\cal O}_{M^{an}}),

668: \qquad p \in \BB Z,

669: \end{equation}

670: with a natural Fr\'echet topology and prove that the resulting virtual

671: representation of $G_{\BB R}$

672: \begin{equation}  \label{vrep}

673: \sum_p (-1)^p

674: \rhom^p_{{\cal D}_M} ( \g M \otimes {\cal S}, {\cal O}_{M^{an}})

675: \end{equation}

676: is admissible of finite length whenever the ${\cal D}_M$-module $\g M$

677: is admissible and ${\cal Z}(\g g_{\BB C})$-finite.

678: In particular, the representation (\ref{vrep}) has a character $\theta$

679: in the sense of Harish-Chandra. As before, $\theta$ is a distribution on

680: $\g g_{\mathbb R}$.

681: 

682: In this paragraph we outline M.~Kashiwara and W.~Schmid's construction of

683: topology on the spaces (\ref{ext}). Oversimplifying and ignoring the

684: $G_{\mathbb C}$- and $G_{\mathbb R}$-actions, suppose first that the sheaf

685: ${\cal S}$ is $(j_{U \hookrightarrow M})_! \mathbb C_U$,

686: where $U \subset M$ is an open semi-algebraic $G_{\mathbb R}$-invariant subset

687: and $\mathbb C_U$ is the constant sheaf on $U$.

688: Furthermore, suppose that $\g M$ is a {\em locally free}

689: quasi-$G_{\mathbb C}$-equivariant ${\cal D}_M$-module, i.e.

690: $\g M = {\cal D}_M \otimes_{{\cal O}_M} \g F$ for some coherent,

691: locally free, $G_{\mathbb C}$-equivariant ${\cal O_M}$-module $\g F$ on $M$.

692: They replace the sheaf ${\cal O}_{M^{an}}$ with the ${\cal C}^{\infty}$

693: Dolbeault complex $\Omega_M^{(0,\cdot)}$, to which it is quasi-isomorphic,

694: and write out isomorphisms of complexes of vector spaces without any topology:

695: \begin{multline}  \label{topology}

696: \rhom_{{\cal D}_M} ( \g M \otimes {\cal S}, {\cal O}_{M^{an}})

697: \simeq \rhom_{{\cal D}_M} ( {\cal D}_M \otimes_{{\cal O}_M} \g F

698: \otimes (j_{U \hookrightarrow M})_! \mathbb C_U, \Omega_M^{(0,\cdot)})  \\

699: \simeq \rhom_{{\cal O}_M} ( \g F \otimes

700: (j_{U \hookrightarrow M})_! \mathbb C_U, \Omega_M^{(0,\cdot)})

701: \simeq \operatorname{R}\Gamma ( U;

702: (\g F^*)^{an} \otimes_{{\cal O}_{M^{an}}} \Omega_M^{(0,\cdot)})  \\

703: \simeq \Gamma ( U;

704: (\g F^*)^{an} \otimes_{{\cal O}_{M^{an}}} \Omega_M^{(0,\cdot)}),

705: \end{multline}

706: where $(\g F^*)^{an}$ denotes the sheaf of analytic sections of the dual

707: of $\g F$. The complex on the right has a natural Fr\'echet topology -- the

708: ${\cal C}^{\infty}$ topology for differential forms -- and continuous

709: $G_{\mathbb R}$ action.

710: General $\g M$ and ${\cal S}$, have resolutions by locally free

711: quasi-$G_{\mathbb C}$-equivariant ${\cal D}_M$-modules

712: (Lemma 4.7 in \cite{KaSchm}) and by sheaves of the type

713: $(j_{U \hookrightarrow M})_! \mathbb C_U$ respectively.

714: This, combined with the acyclicity of the right hand side of (\ref{topology}),

715: makes it possible to equip the vector spaces (\ref{ext}) with a

716: Fr\'echet topology. Then M.~Kashiwara and W.~Schmid work hard to show that

717: this topology does not depend on the choices involved. They do it by

718: proving that the topology on the spaces (\ref{ext}) is that of the

719: maximal globalization of their underlying Harish-Chandra modules.

720: 

721: \separate

722: 

723: Recall that ${\cal B}$ denotes the flag variety of $\g g_{\BB C}$.

724: We will establish a formula for this character as an integral over a

725: cycle in $T^*({\cal B} \times M)$ under an additional assumption that

726: $\g M$ has an infinitesimal character, i.e. ${\cal Z}(\g g_{\BB C})$

727: acts on $\g M$ by a character.

728: Even if this condition is not satisfied, since $\g M$ is assumed to be 

729: ${\cal Z}(\g g_{\BB C})$-finite, there is a finite filtration of $\g M$

730: by quasi-$G_{\BB C}$-equivariant ${\cal D}_M$-submodules such that the

731: successive quotients have infinitesimal characters, and we can apply this

732: integral character formula to each of these quotients separately.

733: 

734: In the special case when $G_{\BB C} = \BB C^{\times}$, the

735: ${\cal D}_M$-module $\g M$ is the sheaf of sections of a

736: $G_{\BB C}$-equivariant line bundle $({\bf E}, \nabla_{\bf E})$ over $M$

737: as in Example \ref{line_bundle} with $U=M$,

738: and ${\cal S}$ is the constant sheaf $\BB C_M$,

739: the flag variety ${\cal B}$ consists of just one  point and the integral

740: character formula will reduce to the classical Riemann-Roch-Hirzebruch formula.

741: 

742: \begin{rem}

743: We do not need $M$ to be projective to establish that (\ref{vrep}) is

744: admissible of finite length. In fact it is sufficient to assume that $M$ is

745: a smooth {\em quasi-projective} variety.

746: The compactness of $M$ will be needed for the integral character formula.

747: 

748: On the other hand, the result of Sumihiro \cite{Su} restated as

749: Proposition 4.6 in \cite{KaSchm} together with Theorem 5.12 there

750: show that we can always embed $M$ into a smooth projective variety

751: and there is no loss of generality in assuming that $M$ is projective.

752: We will illustrate this in Section \ref{example}.

753: \end{rem}

754: 

755: Our derivation of the integral character formula for $\theta$ follows the

756: following scheme:

757: \begin{itemize}

758: \item

759: Replace the pair $(\g M, {\cal S})$ which lives on $M$ with a new pair

760: $(\widetilde{\g M}, \widetilde{\cal S})$ which lives on the flag variety

761: ${\cal B}$ such that the virtual representation (\ref{vrep})

762: stays unchanged (\ref{newpair});

763: \item

764: Write out the fixed point character formula for $\theta$ in terms of

765: $(\widetilde{\g M}, \widetilde{\cal S})$ and zeroes on ${\cal B}$;

766: \item

767: We want to prove that the integral (\ref{intcharformula}) represents $\theta$;

768: first we apply the integral localization formula \cite{L3} to

769: (\ref{intcharformula});

770: \item

771: Combine similar terms in the result of the previous step and match them with

772: the terms in the fixed point character formula for $\theta$ obtained in the

773: earlier step.

774: \end{itemize}

775: 

776: 

777: 

778: \separate

779: 

780: \begin{section}

781: {Statement of the main result}

782: \end{section}

783: 

784: In the previous section we assumed that $\g M$ was a 

785: quasi-$G_{\BB C}$-equivariant coherent ${\cal D}_M$-module on $M$

786: which was admissible and had an infinitesimal character.

787: We will follow \cite{KaSchm} and index characters of ${\cal Z}(\g g_{\BB C})$

788: by linear functionals $\lambda$ on the universal Cartan without the

789: customary shift by $\rho$ (half sum of the positive roots);

790: in other words, $\chi_{\lambda}: {\cal Z}(\g g_{\BB C}) \to \BB C$

791: denotes the character by which ${\cal Z}(\g g_{\BB C})$ acts on the Verma

792: module with highest weight $\lambda$. Then $\chi_{\lambda} = \chi_{\mu}$

793: if and only if $\lambda + \rho$ is conjugate to $\mu + \rho$ under the

794: action of the Weyl group $W$.

795: 

796: \begin{rem}  \label{twistings}

797: In this article we use results from \cite{KaSchm} and \cite{SchV2}.

798: Unfortunately these two sources use different conventions for labeling

799: characters and twists. We follow the conventions of \cite{KaSchm} explained

800: above. On the other hand \cite{SchV2} shift these notations by $\rho$ and

801: they define, for instance, the ``$G_{\BB R}$-equivariant derived category on

802: ${\cal B}$ with twist $(\lambda-\rho)$'' denoted by

803: $\operatorname{D}_{G_{\BB R}} ({\cal B})_{\lambda}$

804: which becomes the bounded $G_{\BB R}$-equivariant derived category in the

805: usual ``untwisted'' sense precisely when $\lambda = \rho$.

806: Similarly, ${\cal O}_{\cal B}(\lambda)$ in \cite{SchV2} denotes

807: the twisted sheaf of holomorphic functions with twist $(\lambda-\rho)$

808: and which becomes the sheaf of ordinary holomorphic functions precisely

809: when $\lambda = \rho$.

810: \end{rem}

811: 

812: 

813: 

814: We assume that $\g M$ is an object in $\modg^{coh,\lambda-\rho}({\cal D}_M)$

815: -- the category of quasi-$G_{\BB C}$-equivariant coherent

816: ${\cal D}_M$-modules on $M$ with infinitesimal character $\chi_{\lambda-\rho}$.

817: 

818: The Kashiwara-Schmid construction of Fr\'echet topology on the spaces

819: (\ref{ext}) was carried out on the level of derived categories.

820: This means that instead of a sheaf ${\cal S}$ we have an element

821: ${\cal S} \in D^b_{G_{\BB R}, \BB R-c} (\BB C_M)$ -- the bounded

822: derived category of $G_{\BB R}$-equivariant $\BB R$-constructible

823: sheaves, and we have a pairing

824: $$

825: \modg^{coh,\lambda-\rho}({\cal D}_M) \times D^b_{G_{\BB R}, \BB R-c} (\BB C_M)

826: \quad \longrightarrow \quad D^b({\cal F}_{G_{\BB R}}),

827: $$

828: $$

829: (\g M, {\cal S}) \quad \mapsto \quad

830: \rhom^{top}_{{\cal D}_M} ( \g M \otimes {\cal S}, {\cal O}_{M^{an}}),

831: $$

832: where $D^b({\cal F}_{G_{\BB R}})$ denotes the derived category

833: of $G_{\BB R}$-representations defined in \parag 3 of \cite{KaSchm}.

834: The category $D^b({\cal F}_{G_{\BB R}})$ is built on complexes of topological

835: vector spaces with continuous $G_{\mathbb R}$-actions $(C,d_C)$,

836: the differential maps $d_C^n : C^n \to C^{n+1}$ are required to be continuous

837: and $G_{\mathbb R}$-equivariant. A complex $(C,d_C)$ is {\em exact} if,

838: for all $n$,

839: $$

840: d_C^n : C^n \to \ker d_C^{n+1}

841: $$

842: is onto and is an open map relative to the subspace topology on

843: $\ker d_C^{n+1}$.

844: A complex $(C,d_C)$ becomes zero in $D^b({\cal F}_{G_{\BB R}})$ precisely

845: when it is exact.

846: 

847: Because we will apply results of A.~Beilinson and J.~Bernstein on equivalences

848: of categories \cite{BB}, we will also assume that $\lambda$ is

849: {\em integrally dominant}, i.e.

850: $$

851: \langle \check \alpha, \lambda \rangle \notin \BB Z_{<0},

852: \qquad \text{for every positive coroot $\check \alpha$.}

853: $$

854: This can always be achieved by replacing $\lambda$ with an appropriate

855: $W$-translate.

856: 

857: \separate

858: 

859: We denote by $\modge({\cal D}_{{\cal B},\,\lambda-\rho})$ the category of

860: modules over the sheaf of twisted differential operators

861: ${\cal D}_{{\cal B},\,\lambda-\rho}$ on the flag variety ${\cal B}$.

862: The sheaf ${\cal D}_{{\cal B},\,\lambda-\rho}$ is defined in \cite{BB},

863: but we follow the twisting conventions of \cite{KaSchm}

864: as explained in Remark \ref{twistings}, so that

865: ${\cal D}_{{\cal B},\,0} = {\cal D}_{{\cal B}}$ -- the sheaf of differential

866: operators on ${\cal B}$ without any twist.

867: We form a product space ${\cal B} \times M$ with diagonal $G_{\BB C}$-action

868: and consider the sheaf of twisted differential operators

869: ${\cal D}_{{\cal B} \times M,\,\lambda-\rho} =_{def}

870: {\cal D}_{{\cal B},\,\lambda-\rho} \boxtimes {\cal D}_M$,

871: the twisting is confined to the factor ${\cal B}$.

872: We also denote by $\modge({\cal D}_{{\cal B} \times M,\,\lambda-\rho})$

873: the category of $G_{\BB C}$-equivariant

874: ${\cal D}_{{\cal B} \times M,\,\lambda-\rho}$-modules on ${\cal B} \times M$.

875: Let $p$ and $q$ be the projection maps

876: $$

877: \begin{CD}

878: {\cal B} @<p<< {\cal B} \times M \\ & & @VVqV \\ & & M,

879: \end{CD}

880: $$

881: and let $\tilde p$ and $\tilde q$ be the induced projections on the

882: cotangent bundles:

883: $$

884: \begin{CD}

885: T^*{\cal B} @<\tilde p<< T^*({\cal B} \times M) \\ & & @VV\tilde qV \\

886: & & T^*M.

887: \end{CD}

888: $$

889: 

890: M.~Kashiwara and W.~Schmid \cite{KaSchm} use results of

891: A.~Beilinson and J.~Bernstein on equivalence of categories \cite{BB}

892: to prove that the pair $(\g M, {\cal S})$ on $M$ can be replaced with a

893: ${\cal D}_{{\cal B},\,\rho-\lambda}$-module and a complex of sheaves on

894: ${\cal B}$.

895: We will make this statement precise.

896: First of all, they show that there exists a coherent holonomic module

897: $\g L \in \modge({\cal D}_{{\cal B} \times M,\,\lambda-\rho})$ such that

898: $$

899: q_* \g L = \g M  \qquad \text{and} \qquad

900: R^k q_* \g L = 0 \text{ if $k \ne 0$,}

901: $$

902: namely one can take

903: $$

904: \g L = {\cal D}_{{\cal B},\,\lambda-\rho} \boxtimes \g M /

905: \gamma_{{\cal B} \times M}

906: (\g g_{\BB C})({\cal D}_{{\cal B},\,\lambda-\rho} \boxtimes \g M),

907: $$

908: where

909: $\gamma_{{\cal B} \times M}

910: (\g g_{\BB C})({\cal D}_{{\cal B},\,\lambda-\rho} \boxtimes \g M)$

911: denotes the image in ${\cal D}_{{\cal B},\,\lambda-\rho} \boxtimes \g M$ of

912: the module

913: $\g g_{\BB C} \otimes ({\cal D}_{{\cal B},\,\lambda-\rho} \boxtimes \g M)$

914: under the map $\gamma_{{\cal B} \times M}$ given by the equation (\ref{gamma})

915: with the ambient manifold ${\cal B} \times M$.

916: 

917: Then we apply the twisted deRham functor to obtain a complex of sheaves

918: $$

919: {\cal L} = DR_{{\cal B} \times M} (\g L) =_{def}

920: R{\cal H}om_{{\cal D}_{{\cal B} \times M, \, \lambda-\rho}}

921: ({\cal O}_{\cal B}(\lambda-\rho) \boxtimes {\cal O}_M, \g L),

922: $$

923: which is an element of the bounded $G_{\BB C}$-equivariant,

924: $\BB C$-constructible derived category with twist $(\rho-\lambda)$ along the

925: ${\cal B}$-factor denoted

926: $D^b_{G_{\BB C},\, \rho-\lambda, \BB C-c} (\BB C_{{\cal B} \times M})$.

927: Here ${\cal O}_{\cal B}(\lambda-\rho)$ is a twisted sheaf of holomorphic

928: functions on ${\cal B}$, with twist $(\lambda-\rho)$

929: (so that ${\cal O}_{\cal B}(0)$ is just the sheaf of functions on ${\cal B}$

930: with no twist at all).

931: 

932: Let $d = \dim_{\BB C} {\cal B}$, and let

933: $(\Omega_{\cal B}^d)^{-1} \simeq {\cal O}_{\cal B} (2\rho)$ denote the

934: reciprocal of the canonical sheaf.

935: Combining the equations (6.7), (7.14) and (7.15) from \cite{KaSchm}

936: we obtain

937: \begin{multline}  \label{maineqn}

938: \rhom^{top}_{{\cal D}_M} ( \g M \otimes {\cal S}, {\cal O}_{M^{an}})  \\

939: \simeq

940: \rhom^{top}_{{\cal D}_{\cal B},\, \rho-\lambda}

941: \bigl( {\cal D}_{{\cal B},\, \rho-\lambda} \otimes_{{\cal O}_{\cal B}}

942: (\Omega_{\cal B}^d)^{-1} \otimes Rp_*({\cal L} \otimes q^{-1}{\cal S}),

943: {\cal O}_{{\cal B}^{an}}(\rho-\lambda) \bigr) [d - 2\dim_{\BB C} M]  \\

944: \simeq

945: \rhom^{top}_{{\cal D}_{\cal B}, -\lambda -\rho}

946: \bigl( {\cal D}_{{\cal B}, -\lambda -\rho}

947: \otimes Rp_*({\cal L} \otimes q^{-1}{\cal S}),

948: {\cal O}_{{\cal B}^{an}}(-\lambda -\rho) \bigr) [d - 2\dim_{\BB C} M]  \\

949: \simeq

950: \rhom \bigl( Rp_*({\cal L} \otimes q^{-1}{\cal S}), 

951: {\cal O}_{{\cal B}^{an}}(-\lambda -\rho) \bigr) [d - 2\dim_{\BB C} M]

952: \end{multline}

953: as elements of the derived category of $G_{\BB R}$-representations

954: $D^b({\cal F}_{G_{\BB R}})$.

955: Here we view $Rp_*({\cal L} \otimes q^{-1}{\cal S})$ as an object in

956: $D^b_{G_{\BB R},-\lambda-\rho, \BB R-c} (\BB C_{\cal B})$,

957: which makes sense because $2\rho$ is an integral weight, and this implies

958: the existence of a canonical isomorphism (6.8) in \cite{KaSchm}

959: $$

960: D^b_{G_{\BB R},\, \rho-\lambda, \BB R-c} (\BB C_{\cal B})

961: \quad\simeq\quad

962: D^b_{G_{\BB R},-\lambda-\rho, \BB R-c} (\BB C_{\cal B}).

963: $$

964: That is, the pair $(\g M, {\cal S})$ on the variety $M$ is replaced by a pair

965: \begin{equation}  \label{newpair}

966: \bigl( {\cal D}_{{\cal B}, -\lambda -\rho},

967: Rp_*({\cal L} \otimes q^{-1}{\cal S}) \bigr)

968: \end{equation}

969: on the flag variety ${\cal B}$, with an additional twist by $(-\lambda -\rho)$.

970: 

971: \separate

972: 

973: Our starting point is the integral character formula (\ref{intcharformula})

974: proved by W.~Schmid and K.~Vilonen in \cite{SchV2}. We apply it to the

975: right hand side of (\ref{maineqn}).

976: We fix a compact real form $U_{\BB R} \subset G_{\BB C}$.

977: Recall that $\mu_{\cal B}: T^*{\cal B} \to \g g_{\BB C}^*$ is the moment

978: map defined by (\ref{mu}), $\sigma_{\cal B}$ is the canonical complex

979: algebraic holomorphic symplectic form on $T^*{\cal B}$,

980: $\lambda_x : {\cal B} \to \g g_{\BB C}^*$ is the $U_{\mathbb R}$-equivariant

981: Rossmann's map, and $\tau_{\lambda}$ is a certain $U_{\BB R}$-invariant

982: 2-form on ${\cal B}$.

983: The character of the virtual representation (\ref{vrep}) is a distribution

984: on $\Omega_c^{top}(\g g_{\BB R})$ -- the space of smooth compactly

985: supported differential forms on $\g g_{\BB R}$ of top degree.

986: For an element $\phi \in \Omega_c^{top}(\g g_{\BB R})$, its Fourier transform 

987: $\hat \phi \in {\cal C}^{\infty}(\g g_{\BB C}^*)$ is defined by (\ref{FT})

988: without the customary factor of $i = \sqrt{-1}$ in the exponent.

989: Then the integral character formula says that the character

990: $\theta$ of the virtual representation (\ref{maineqn}) of $G_{\BB R}$,

991: as a distribution on $\Omega_c^{top}(\g g_{\BB R})$, is

992: \begin{multline}  \label{intcharformula1}

993: \theta(\phi) = \frac 1{(2\pi i)^d d!}

994: \int_{Ch(Rp_*({\cal L} \otimes q^{-1}{\cal S}))^a}

995: (\hat\phi \circ \mu_{-\lambda}) \cdot (\sigma_{\cal B} + \tau_{\lambda})^d  \\

996: = \frac 1{(2\pi i)^d}

997: \int_{Ch(Rp_*({\cal L} \otimes q^{-1}{\cal S}))^a}

998: \Bigl( \int_{\g g_{\BB R}}

999: e^{\langle X, \mu_{\cal B}(\zeta) - \lambda_x \rangle

1000: + \sigma_{\cal B} + \tau_{\lambda}}

1001: \wedge \phi(X) \Bigr)_{[2d]},  \\

1002: \qquad X \in \g g_{\BB R},\: \zeta \in T^*{\cal B}.

1003: \end{multline}

1004: Here $\mu_{-\lambda} = \mu_{\cal B} - \lambda_x$,

1005: $d= \dim_{\BB C} {\cal B}$, $a: T^*{\cal B} \to T^*{\cal B}$

1006: is the antipodal map $\zeta \mapsto -\zeta$, and

1007: $Ch(Rp_*({\cal L} \otimes q^{-1}{\cal S}))^a$ denotes

1008: the image under this antipodal map of the characteristic cycle of 

1009: $Rp_*({\cal L} \otimes q^{-1}{\cal S})$ (which is a cycle in $T^*{\cal B}$).

1010: 

1011: \begin{rem}

1012: If $Z$ is a complex manifold and

1013: $Z^{\BB R}$ is the underlying real analytic manifold,

1014: then the holomorphic symplectic form $\sigma_Z$ is defined on the

1015: holomorphic cotangent bundle $T^*Z$, while the characteristic cycles

1016: of constructible sheaves on $Z$ lie in the real cotangent bundle

1017: $T^*(Z^{\BB R})$. Hence we need to identify $T^*Z$ with $T^*(Z^{\BB R})$.

1018: There are at least two different but equally natural ways of doing this,

1019: we use the convention (11.1.2) of \cite{KaScha}, Chapter XI;

1020: the same convention is used in \cite{L1}, \cite{L2}, \cite{L3} and

1021: \cite{SchV2}. Under this convention, if

1022: $\sigma_{Z,\BB R}$ is the canonical real symplectic form on $T^*Z^{\BB R}$

1023: and $\sigma_Z$ is the canonical complex holomorphic symplectic form on $T^*Z$,

1024: then $\sigma_{Z,\BB R}$ gets identified with $2 \re \sigma_Z$.

1025: % dz <--> 2dx,  -idz <--> 2dy

1026: \end{rem}

1027: 

1028: Another important ingredient is a generalization of the Hopf index theorem

1029: stated as Corollary 9.5.2 in \cite{KaScha}.

1030: Let ${\cal T}$ be a constructible sheaf on $M$ or

1031: an element of the derived category $D^b_{\BB R-c} (\BB C_M)$,

1032: and let $\chi(M, {\cal T})$ denote the Euler characteristic of $M$ with

1033: respect to ${\cal T}$.

1034: Then

1035: \begin{equation}  \label{hopf}

1036: \chi(M, {\cal T}) = \# \bigl( [M] \cap Ch({\cal T}) \bigr),

1037: \end{equation}

1038: where $[M]$ denotes the fundamental cycle of $M$.

1039: Alternatively one can apply the equation (5.30) from \cite{Schu}.

1040: Let $\thom_{T^*M}$ denote the Thom form of the cotangent bundle

1041: $T^*M \twoheadrightarrow M$.

1042: That is $\thom_{T^*M}$ is a closed differential form on $T^*M$ of degree

1043: $2\dim_{\BB C}M$ which decays rapidly along the fiber (or even compactly

1044: supported along the fiber) and such that

1045: $$

1046: \int_{T_x^*M} \thom_{T^*M} = (2\pi)^{\dim_{\BB C}M}, \qquad \forall x \in M.

1047: $$

1048: We regard $M$ as a submanifold of $T^*M$ via the zero section inclusion.

1049: Then the restriction of the Thom form to $M$ is the Euler form of $M$.

1050: Since the form $(2\pi)^{-\dim_{\BB C}M} \thom_{T^*M}$ is Poincar\'e dual to

1051: the homology class of $[M]$ in $T^*M$, the Hopf index theorem (\ref{hopf})

1052: can be rewritten as

1053: \begin{equation}  \label{bonnet}

1054: \chi(M, {\cal T}) = \# \bigl( [M] \cap Ch({\cal T}) \bigr)

1055: = (2\pi)^{-\dim_{\BB C}M} \int_{Ch({\cal T})} \thom_{T^*M},

1056: \end{equation}

1057: which is a generalization of the Gauss-Bonnet theorem.

1058: 

1059: Recall that $U_{\BB R}$ is a compact real form of $G_{\BB C}$.

1060: The form $\thom_{T^*M}$ may be chosen to be $U_{\BB R}$-invariant.

1061: If, in addition, the cotangent bundle $T^*M$ has a spin structure,

1062: then V.~Mathai and D.~Quillen showed in \cite{MQ}

1063: (see also Section 7.7 in \cite{BGV})

1064: that $\thom_{T^*M}$ can be realized as the top degree part of a

1065: $U_{\BB R}$-equivariantly closed form on $T^*M$ in a canonical way.

1066: 

1067: \separate

1068: 

1069: The last essential ingredient is the deformation argument for integrals

1070: of equivariantly closed forms from \cite{L3}.

1071: This argument requires that any maximal complex torus

1072: $T_{\BB C} \subset G_{\BB C}$ acts on $M$ with isolated fixed points.

1073: (Since $M$ is compact, there will be only finitely many of those.)

1074: If this condition is satisfied for one particular torus, then it is

1075: satisfied for all tori because all maximal complex tori are conjugate

1076: by elements of $G_{\BB C}$.

1077: (This condition is satisfied in all application we have in mind.)

1078: 

1079: We will combine the integral character formula (\ref{intcharformula1})

1080: and the Gauss-Bonnet formula (\ref{bonnet}) to get a formula for the

1081: character of the virtual representation (\ref{maineqn}) as an integral over

1082: a cycle in $T^*({\cal B} \times M)$.

1083: Set

1084: $$

1085: \widetilde{\cal S} = {\cal L} \otimes q^{-1}{\cal S}

1086: \qquad \text{and} \qquad

1087: \Lambda = Ch(\widetilde{\cal S})^a

1088: = Ch(\BB D_{{\cal B} \times M} (\widetilde{\cal S}))

1089: \subset T^*({\cal B} \times M).

1090: $$

1091: Here $a: T^*({\cal B} \times M) \to T^*({\cal B} \times M)$

1092: is the antipodal map $\zeta \mapsto -\zeta$ and

1093: $Ch(\widetilde{\cal S})^a$ denotes the image under this antipodal map of

1094: the characteristic cycle of $\widetilde{\cal S}$; the operator

1095: $\BB D_{{\cal B} \times M}$ is the Verdier duality operator

1096: (see (\ref{D_XxM}) below for its properties).

1097: 

1098: 

1099: \begin{thm}  \label{main}

1100: Suppose that any maximal complex torus $T_{\BB C} \subset G_{\BB C}$ acts on

1101: $M$ with isolated fixed points.

1102: Then the value of the character $\theta$ of the virtual representation

1103: (\ref{vrep}) on $\phi \in \Omega_c^{top}(\g g_{\BB R})$ is

1104: \begin{multline}  \label{mainintegral}

1105: \theta(\phi) =

1106: \frac {i^n}{(2\pi i)^{d+n} d!}

1107: \int_{\Lambda}

1108: \tilde p^* \bigl( (\hat\phi \circ \mu_{-\lambda}) \cdot

1109: (\sigma_{\cal B} + \tau_{\lambda})^d \bigr)

1110: \wedge \tilde q^*\thom_{T^*M}  \\

1111: = \frac {i^n}{(2\pi i)^{d+n}}

1112: \int_{\Lambda} \Bigl( \int_{\g g_{\BB R}} 

1113: \tilde p^* e^{\langle X, \mu_{\cal B}(\zeta) - \lambda_x \rangle +

1114: \sigma_{\cal B} + \tau_{\lambda}}

1115: \wedge \tilde q^*\thom_{T^*M} \wedge \phi(X) \Bigr)_{[2(d+n)]},

1116: \end{multline}

1117: where 

1118: $\mu_{-\lambda} = \mu_{\cal B} - \lambda_x$,

1119: $d =\dim_{\BB C} {\cal B}$ and $n=\dim_{\BB C} M$.

1120: \end{thm}

1121: 

1122: \begin{rem}

1123: Suppose that the group $G_{\BB C} = \BB C^{\times}$, the

1124: ${\cal D}_M$-module $\g M$ is the sheaf of sections of a

1125: $G_{\BB C}$-equivariant line bundle $({\bf E}, \nabla_{\bf E})$

1126: over $M$ as in Example \ref{line_bundle},

1127: and ${\cal S}$ is the constant sheaf $\BB C_M$.

1128: Then the flag variety ${\cal B}$ is just one  point,

1129: the cycle $\Lambda =[M]$, and because of the Riemann-Roch relationship

1130: (8.4) in \cite{MQ} (or Theorem 7.44 in \cite{BGV})

1131: the above integral character formula will reduce to the classical

1132: Riemann-Roch-Hirzebruch formula.

1133: There is no curvature of the line bundle $({\bf E}, \nabla_{\bf E})$

1134: present in the character formula (\ref{mainintegral})

1135: because we assume that the connection $\nabla_{\bf E}$ is flat so that

1136: ${\cal O}({\bf E})$ is a quasi-$G_{\BB C}$-equivariant ${\cal D}_M$-module.

1137: \end{rem}

1138: 

1139: 

1140: 

1141: \separate

1142: 

1143: \begin{section}

1144: {Proof of Theorem \ref{main}}

1145: \end{section}

1146: 

1147: We start our proof by applying the fixed point character formula

1148: (\ref{fpcharformula}) combined with (\ref{m})

1149: due to W.~Schmid and K.~Vilonen \cite{SchV2} to the right hand side

1150: of (\ref{maineqn}).

1151: Thus the character $\theta$ of the virtual representation (\ref{maineqn})

1152: is given by integration against a function $F_{\theta}$ on $\g g_{\BB R}$:

1153: \begin{equation}  \label{F}

1154: \theta(\phi) = \int_{\g g_{\BB R}} F_{\theta} \phi,

1155: \qquad \phi \in \Omega_c^{top}(\g g_{\BB R}).

1156: \end{equation}

1157: This function $F_{\theta}$ is an $Ad(G_{\BB R})$-invariant, locally $L^1$

1158: function on $\g g_{\BB R}$ whose restriction to the set of regular semisimple

1159: elements $\g g_{\BB R}^{rs}$ can be represented by a real analytic function.

1160: The value of this analytic function at $X \in \g g_{\BB R}^{rs}$ is determined

1161: by zeroes of the vector field $X_{\cal B}$ on the flag variety ${\cal B}$

1162: as follows.

1163: 

1164: Recall that $\g t_{\BB C}(X) \subset \g g_{\BB C}$ is the unique Cartan

1165: subalgebra containing $X \in \g g_{\BB R}^{rs}$.

1166: Let $\Psi(X) \subset \g t^*_{\BB C}(X)$ be the root system of $\g g_{\BB C}$

1167: with respect to $\g t_{\BB C}(X)$ and pick a positive root system

1168: $\Psi^{\le}(X) \subset \Psi$ such that

1169: $$

1170: \re \alpha (X) \le 0 \qquad \text{for all $\alpha \in \Psi^{\le}(X)$.}

1171: $$

1172: Let $B(X) \subset G_{\BB C}$ be the Borel subgroup whose Lie algebra

1173: contains $\g t_{\BB C}(X)$ and the positive root spaces corresponding

1174: to $\Psi^{\le}(X)$.

1175: The action of $B(X)$ on ${\cal B}$ has exactly the same number of orbits

1176: $O_1, \dots, O_{|W|}$ as the order of the Weyl group $W$ of $G_{\BB C}$.

1177: Each of these orbits $O_k$ contains exactly one zero of the vector field

1178: $X_{\cal B}$, and we order the zeroes $\{\g b_1,\dots,\g b_{|W|}\}$ of

1179: $X_{\cal B}$ so that $\g b_k$ is contained in $O_k$, $k=1,\dots,|W|$.

1180: The set of zeroes $\{\g b_1,\dots,\g b_{|W|}\}$ is precisely the set of Borel

1181: subalgebras containing $\g t_{\BB C}(X)$.

1182: Let $\Psi^+_{\g b_k}(X) \subset \Psi(X)$ be the

1183: positive root system such that $\g b_k$ contains all the negative root spaces,

1184: $k=1,\dots,|W|$.

1185: Then the fixed point character formula \cite{SchV2} says that

1186: the function $F_{\theta}$ which appeared in the equation (\ref{F}) is

1187: \begin{equation}  \label{fpcharf}

1188: F_{\theta} (X) = (-1)^d \sum_{k=1}^{|W|} m_{\g b_k}(X)

1189: \frac {e^{-\langle X , \lambda_{\g b_k} \rangle}}

1190: {\prod_{\alpha \in \Psi^+_{\g b_k}(X)} \alpha(X)},

1191: \end{equation}

1192: where $m_{\g b_k}(X)$'s are integer multiplicities given by the formula

1193: \begin{equation}  \label{m1}

1194: m_{\g b_k}(X) = \chi

1195: \bigl( {\cal H}^*_{O_k} (Rp_*(\widetilde{\cal S}))_{\g b_k} \bigr)

1196: = \chi \bigl( (j_{\{\g b_k\} \hookrightarrow O_k})^* \circ

1197: (j_{O_k \hookrightarrow {\cal B}})^! \bigl(

1198: Rp_*(\widetilde{\cal S}) \bigr) \bigr).

1199: \end{equation}

1200: 

1201: Let $\BB D_{\cal B}$ and $\BB D_{{\cal B} \times M}$ denote the

1202: Verdier duality operators on ${\cal B}$ and ${\cal B} \times M$ respectively:

1203: $$

1204: \BB D_{\cal B} : \quad

1205: D^b_{G_{\BB R},-\lambda-\rho,\BB R-c} (\BB C_{\cal B})

1206: \quad \tilde\longrightarrow \quad

1207: D^b_{G_{\BB R},\lambda+\rho,\BB R-c} (\BB C_{\cal B});

1208: $$

1209: \begin{equation} \label{D_XxM}

1210: \BB D_{{\cal B} \times M} : \quad

1211: D^b_{G_{\BB R},-\lambda-\rho,\BB R-c} (\BB C_{{\cal B} \times M})

1212: \quad \tilde\longrightarrow \quad

1213: D^b_{G_{\BB R},\lambda+\rho,\BB R-c} (\BB C_{{\cal B} \times M}).

1214: \end{equation}

1215: The effect of the Verdier duality operator $\BB D_Z$ on the

1216: characteristic cycle of an $\BB R$-constructible sheaf ${\cal T}$

1217: (or an element of $D^b_{\BB R-c} (\BB C_Z)$) on any

1218: smooth quasi-projective variety $Z$ is described by

1219: $$

1220: Ch(\BB D_Z ({\cal T})) = Ch({\cal T})^a,

1221: $$

1222: where $a: T^*Z \to T^*Z$ is the antipodal map $\zeta \mapsto -\zeta$ and

1223: $Ch({\cal T})^a$ denotes the image under this antipodal map of

1224: the characteristic cycle of ${\cal T}$.

1225: 

1226: 

1227: \separate

1228: 

1229: 

1230: For a regular semisimple element $X \in \g g_{\BB C}^{rs}$ we denote by

1231: $T_{\BB C}(X) = \exp (\g t_{\BB C}(X)) \subset G_{\BB C}$

1232: the maximal complex torus corresponding to the unique Cartan

1233: subalgebra containing $X$.

1234: If $x \in M$ is a point fixed by $T_{\BB C}(X)$; then $T_{\BB C}(X)$ acts

1235: linearly on the tangent space $T_xM$.

1236: We denote by $\Delta (X) \subset \g t_{\BB C}(X)^*$ the set of weights

1237: of $\g t_{\BB C}(X)$ which are either roots of $\g g_{\BB C}$ or occur in

1238: the tangent space $T_xM$ of some point $x \in M$ fixed by $T_{\BB C}(X)$.

1239: 

1240: Let $\g g_{\BB R}'$ denote the set of {\em strongly regular semisimple}

1241: elements $X \in \g g_{\BB R}^{rs}$ which satisfy the following additional

1242: properties. If $\g t_{\BB R}(X) \subset \g g_{\BB R}$ and

1243: $\g t_{\BB C}(X) \subset \g g_{\BB C}$ are the unique Cartan

1244: subalgebras in $\g g_{\BB R}$ and $\g g_{\BB C}$ respectively

1245: containing $X$, then:

1246: \begin{enumerate}

1247: \item

1248: The set of zeroes of the vector field $X_M$ is exactly the set of points in

1249: $M$ fixed by the complex torus

1250: $T_{\BB C}(X) = \exp (\g t_{\BB C}(X)) \subset G_{\BB C}$;

1251: \item

1252: $\beta(X) \ne 0$ for all $\beta \in \Delta(X) \subset \g t_{\BB C}(X)^*$;

1253: \item

1254: For each $\beta \in \Delta(X)$, we have either

1255: \begin{equation}  \label{g'}

1256: \re(\beta)|_{\g t_{\BB R}(X)} \equiv 0

1257: \qquad \text{or} \qquad \re(\beta(X)) \ne 0.

1258: \end{equation}

1259: \end{enumerate}

1260: Clearly, $\g g_{\BB R}'$ is an open subset of $\g g_{\BB R}$;

1261: since $M$ is compact and $\Delta(X)$ is finite, the complement of

1262: $\g g_{\BB R}'$ in $\g g_{\BB R}$ has measure zero.

1263: 

1264: From now on we will assume that the element $X \in \g g_{\BB R}$ is

1265: not only regular semisimple, but also lies in $\g g_{\BB R}'$.

1266: Then applying the integral localization formula of \cite{L3} to the

1267: integral character formula (\ref{intcharformula1}), with the global formula

1268: for the multiplicities, we can rewrite the formula (\ref{m1}) as

1269: \begin{equation}  \label{m2}

1270: m_{\g b_k}(X) = \chi \bigl( {\cal B}, \bigl( \BB D_{\cal B}

1271: (Rp_*(\widetilde{\cal S})) \bigr)_{O_k} \bigr)

1272: = \chi \bigl( {\cal B},

1273: (j_{O_k \hookrightarrow {\cal B}})_! \circ

1274: (j_{O_k \hookrightarrow {\cal B}})^* \bigl( \BB D_{\cal B}

1275: (Rp_*(\widetilde{\cal S})) \bigr) \bigr).

1276: \end{equation}

1277: Applying Proposition 2.5.11 from \cite{KaScha} to the Cartesian square

1278: $$

1279: \begin{matrix}

1280: O_k \times M & \hookrightarrow & {\cal B} \times M \\

1281: \downarrow & & \downarrow \\

1282: O_k & \hookrightarrow & {\cal B}

1283: \end{matrix}

1284: $$

1285: and using that the projection map $p$ is proper we obtain:

1286: \begin{multline}  \label{m3}

1287: m_{\g b_k}(X) =

1288: \chi \bigl( {\cal B},

1289: (j_{O_k \hookrightarrow {\cal B}})_! \circ

1290: (j_{O_k \hookrightarrow {\cal B}})^* \bigl( Rp_!

1291: (\BB D_{{\cal B} \times M} (\widetilde{\cal S})) \bigr) \bigr)  \\

1292: = \chi \bigl( {\cal B},

1293: (j_{O_k \hookrightarrow {\cal B}})_! \circ Rp_! \circ

1294: (j_{O_k \times M \hookrightarrow {\cal B} \times M})^*

1295: (\BB D_{{\cal B} \times M} (\widetilde{\cal S})) \bigr)  \\

1296: = \chi \bigl( {\cal B}, Rp_! \circ

1297: (j_{O_k \times M \hookrightarrow {\cal B} \times M})_! \circ

1298: (j_{O_k \times M \hookrightarrow {\cal B} \times M})^*

1299: (\BB D_{{\cal B} \times M} (\widetilde{\cal S})) \bigr)  \\

1300: = \chi \bigl( {\cal B} \times M,

1301: (j_{O_k \times M \hookrightarrow {\cal B} \times M})_! \circ

1302: (j_{O_k \times M \hookrightarrow {\cal B} \times M})^*

1303: (\BB D_{{\cal B} \times M} (\widetilde{\cal S})) \bigr)  \\

1304: = \chi \bigl( {\cal B} \times M,

1305: (\BB D_{{\cal B} \times M} (\widetilde{\cal S}))_{O_k \times M} \bigr).

1306: \end{multline}

1307: 

1308: \separate

1309: 

1310: Next we will compare this result with the result of application of

1311: the integral localization formula of \cite{L3} to the integral

1312: (\ref{mainintegral}). Let

1313: $$

1314: M = \coprod_{\{x \in M;\: X_M(x) =0 \}} \tilde O_x

1315: $$

1316: be the Bialynicki-Birula decomposition \cite{Bi} of M into attracting sets

1317: (relative to $X$), as described in \cite{L3}.

1318: To obtain this decomposition we pick any

1319: $X' \in \g t_{\mathbb R}(X) \cap \g g_{\mathbb R}'$ in the same connected

1320: component of $\g t_{\mathbb R}(X) \cap \g g_{\mathbb R}'$ and such that

1321: $$

1322: \re \beta(X) >0 \: \Longleftrightarrow \: \re \beta(X') >0

1323: \qquad \text{and} \qquad

1324: \re \beta(X) <0 \: \Longleftrightarrow \: \re \beta(X') <0

1325: $$

1326: for all $\beta \in \Delta(X)$, and the complex 1-dimensional subspace

1327: $\{tX';\: t \in \BB C \} \subset \g g_{\BB C}$ is the Lie algebra of

1328: a closed algebraic subgroup $\BB C^{\times}(X') \subset G_{\BB C}$

1329: isomorphic to $\BB C^{\times}$.

1330: We fix an embedding of $\BB C^{\times}(X') \simeq \BB C^{\times}$ into

1331: $\mathbb C$ so that the tangent map sends

1332: $X' \in T_e (\BB C^{\times}(X'))$ into an element with

1333: nonnegative real part. This embedding allows us to take limits as

1334: $z \in \BB C^{\times}(X')$ approaches to zero,

1335: and for each $x \in M$ with $X_M(x) =0$ we set

1336: $$

1337: \tilde O_x = \{ y \in M;\: \lim_{z \to 0} z^{-1} \cdot y =x \}.

1338: $$

1339: The sets $\tilde O_x$ are smooth locally closed algebraic subvarieties of $M$.

1340: 

1341: Then

1342: \begin{multline*}

1343: \{x \in {\cal B} \times M;\: X_{{\cal B} \times M} (x) =0 \} \\

1344: = \quad

1345: \{x \in {\cal B};\: X_{\cal B}(x) =0 \} \times \{x \in M;\: X_M(x) =0 \}  \\

1346: = \quad \{\g b_1,\dots,\g b_{|W|}\} \times \{x \in M;\: X_M(x) =0 \}

1347: \end{multline*}

1348: is the set of zeroes of the vector field $X_{{\cal B} \times M}$, and

1349: $$

1350: {\cal B} \times M =

1351: \coprod_{\begin{matrix} k=1,\dots,|W| \\ \{x \in M;\:X_M(x) =0 \} \end{matrix}}

1352: O_k \times \tilde O_x

1353: $$

1354: is the Bialynicki-Birula decomposition of ${\cal B} \times M$

1355: into attracting sets (relative to $X$).

1356: Applying the integral localization formula to the integral (\ref{mainintegral})

1357: yields

1358: $$

1359: \frac {i^n}{(2\pi i)^{d+n}}

1360: \int_{\Lambda} \Bigl( \int_{\g g_{\BB R}} 

1361: \tilde p^* e^{\langle X, \mu_{\cal B}(\zeta) - \lambda_x \rangle +

1362: \sigma_{\cal B} + \tau_{\lambda}}

1363: \wedge \tilde q^* \thom_{T^*M} \wedge \phi(X) \Bigr)_{[2(d+n)]}

1364: = \int_{\g g_{\BB R}} \tilde F_{\theta} \phi

1365: $$

1366: for any $\phi \in \Omega_c^{top}(\g g_{\BB R})$

1367: with $\supp \phi \subset \g g_{\BB R}'$,

1368: where the function $\tilde F_{\theta}$ is

1369: \begin{equation}  \label{Ftilde}

1370: \tilde F_{\theta}(X) =

1371: (-1)^d \sum_{k=1}^{|W|} \sum_{\{x \in M;\:X_M(x) =0 \}} \tilde m_{k,x}(X)

1372: \frac {e^{-\langle X , \lambda_{\g b_k} \rangle}}

1373: {\prod_{\alpha \in \Psi^+_{\g b_k}(X)} \alpha(X)},

1374: \end{equation}

1375: and

1376: \begin{multline*}

1377: \tilde m_{k,x}(X) =

1378: \chi \bigl( {\cal B} \times M, (\BB D_{{\cal B} \times M}

1379: (\widetilde{\cal S}))_{O_k \times \tilde O_x} \bigr)  \\

1380: = \chi({\cal B} \times M, 

1381: (j_{O_k \times \tilde O_x \hookrightarrow {\cal B} \times M})_! \circ

1382: (j_{O_k \times \tilde O_x \hookrightarrow {\cal B} \times M})^*

1383: (\BB D_{{\cal B} \times M} (\widetilde{\cal S})) \bigr).

1384: \end{multline*}

1385: We want to show that $\tilde F_{\theta}(X) = F_{\theta}(X)$. We have:

1386: \begin{multline*}

1387: \sum_{\{x \in M;\:X_M(x) =0 \}} \tilde m_{k,x}(X) =

1388: \sum_{\{x \in M;\:X_M(x) =0 \}}

1389: \chi \bigl( {\cal B} \times M, (\BB D_{{\cal B} \times M}

1390: (\widetilde{\cal S}))_{O_k \times \tilde O_x} \bigr)  \\

1391: = \chi \bigl( {\cal B} \times M,

1392: (\BB D_{{\cal B} \times M} (\widetilde{\cal S}))_{O_k \times M} \bigr)

1393: = m_{\g b_k}(X).

1394: \end{multline*}

1395: Combining this with the equations (\ref{fpcharf}) and (\ref{Ftilde}) we obtain

1396: $$

1397: \tilde F_{\theta}(X) =

1398: (-1)^d \sum_{k=1}^{|W|} m_{x_k}

1399: \frac {e^{-\langle X , \lambda_{\g b_k} \rangle}}

1400: {\prod_{\alpha \in \Psi^+_{\g b_k}(X)} \alpha(X)} = F_{\theta}(X),

1401: $$

1402: which proves (\ref{mainintegral}) at least when

1403: $\supp(\phi) \subset \g g_{\BB R}'$.

1404: Since the function $F_{\theta}$ is known to be a character of some

1405: representation (namely (\ref{vrep})),

1406: it is locally $L^1$ and the main result of \cite{L3} now proves

1407: (\ref{mainintegral}) for all $\phi \in \Omega_c^{top}(\g g_{\BB R})$.

1408: 

1409: 

1410: \separate

1411: 

1412: \begin{section}

1413: {An example on the enhanced flag variety}  \label{example}

1414: \end{section}

1415: 

1416: The purpose of this section is to show how Theorem \ref{main} can be applied

1417: to smooth quasi-projective varieties which are not necessarily projective.

1418: We let our group $G_{\BB R}$ be

1419: $GL(N, \BB R) \subset GL(N, \BB C) = G_{\BB C}$, for some

1420: $N \in \BB N$; we denote by $B_{\BB C}$ the group of all invertible

1421: lower-triangular matrices and by $N_{\BB C} \subset B_{\BB C}$ the group

1422: of lower-triangular unipotent matrices, we set

1423: $B_{\BB R} = B_{\BB C} \cap G_{\BB R}$ and

1424: $N_{\BB R} = N_{\BB C} \cap G_{\BB R}$.

1425: We will be interested in the complex homogeneous space

1426: $Z_{\BB C} = G_{\BB C} / N_{\BB C}$ and its real submanifold

1427: $Z_{\BB R} = G_{\BB R} / N_{\BB R} \subset Z_{\BB C}$.

1428: The space $Z_{\BB C}$ can be regarded as a fiber bundle over the flag variety

1429: ${\cal B} = G_{\BB C} / B_{\BB C}$ with fibers isomorphic to

1430: $H_{\BB C} =_{def} B_{\BB C} /N_{\BB C} \simeq (\BB C^{\times})^N$.

1431: It is easy to see that these fibers can be identified with each other in a

1432: canonical way. This allows us to think of $Z_{\BB C}$ as a trivial principal

1433: $H_{\BB C}$-bundle which is also a $G_{\BB C}$-equivariant fiber

1434: bundle, and the actions of $H_{\BB C}$ and $G_{\BB C}$ commute.

1435: W.~Schmid and K.~Vilonen call $Z_{\mathbb C}$ the {\em enhanced flag variety}

1436: of $\g g_{\BB C}$ \cite{SchV2}.

1437: 

1438: Let ${\cal S}' = (j_{Z_{\BB R} \hookrightarrow Z_{\BB C}})_*

1439: (\BB C_{Z_{\BB R}})$ denote the sheaf on $Z_{\BB C}$ which is the direct

1440: image of the constant sheaf on $Z_{\BB R}$.

1441: Fix a $\lambda \in \g h_{\mathbb C}^*$ and

1442: denote by $\ker \chi_{\lambda-\rho} \subset {\cal Z}(\g g_{\mathbb C})$

1443: the kernel of the character

1444: $\chi_{\lambda-\rho} : {\cal Z}(\g g_{\mathbb C}) \to \mathbb C$, and take a

1445: ${\cal D}_{Z_{\BB C}}$-module

1446: $$

1447: \g M' = {\cal D}_{Z_{\BB C}} /

1448: \gamma_{Z_{\BB C}} (\ker \chi_{\lambda-\rho}) {\cal D}_{Z_{\BB C}}.

1449: $$

1450: Notice that by construction ${\cal Z}(\g g_{\mathbb C})$ acts on $\g M'$

1451: by character $\chi_{\lambda-\rho}$.

1452: We form a virtual representation of $G_{\BB R}$

1453: \begin{equation}  \label{Whittaker}

1454: \rhom^{top}_{{\cal D}_{Z_{\BB C}}} ( \g M' \otimes {\cal S}',

1455: {\cal O}_{Z_{\BB C}^{an}}).

1456: \end{equation}

1457: The composition factors of this representation are known to be

1458: the principal series representations of $G_{\BB R}$ which have the same

1459: character on $\g g_{\BB R}$.

1460: We will obtain an integral formula for the character of this representation.

1461: Of course, the homogeneous space $Z_{\BB C}$ is not compact,

1462: and in order to apply Theorem \ref{main} we need to compactify

1463: $Z_{\BB C}$ first.

1464: 

1465: Recall that $Z_{\BB C}$ is a principal ($H_{\BB C}, G_{\BB C}$)-equivariant

1466: fiber bundle with commuting $H_{\BB C}$- and $G_{\BB C}$-actions.

1467: Hence we can embed

1468: $H_{\BB C} =_{def} B_{\BB C} /N_{\BB C} \simeq (\BB C^{\times})^N$

1469: into the vector space $\BB C^N$ and

1470: form a $G_{\BB C}$-equivariant vector bundle

1471: $\tilde Z_{\BB C} \simeq \BB C^N \times {\cal B}$

1472: with an embedding $Z_{\BB C} \hookrightarrow \tilde Z_{\BB C}$

1473: of $G_{\BB C}$-equivariant fiber bundles over ${\cal B}$.

1474: Next we form a line bundle $L_0 = \BB C \times {\cal B}$ over ${\cal B}$

1475: with trivial $G_{\BB C}$-action on the first factor and form a

1476: projectivization $M =_{def} \BB P(\tilde Z_{\BB C} \oplus L_0)$.

1477: Then $M$ is a smooth complex projective variety, $G_{\BB C}$ acts on $M$

1478: algebraically, and $M$ contains $Z_{\BB C}$ as a dense open subset.

1479: Define

1480: $$

1481: \g M = {\cal D}_M / \gamma_M (\ker \chi_{\lambda-\rho}) {\cal D}_M

1482: \qquad \text{and} \qquad

1483: {\cal S} = (j_{Z_{\BB C} \hookrightarrow M})_! {\cal S}'

1484: = (j_{Z_{\BB R} \hookrightarrow M})_! (\BB C_{Z_{\BB R}}).

1485: $$

1486: 

1487: For convenience we restate Theorem 5.12 of \cite{KaSchm}:

1488: \begin{thm}

1489: Let $f: X \to Y$ be a $G_{\BB C}$-equivariant morphism between complex

1490: algebraic, quasi-projective $G_{\BB C}$-manifolds $X$, $Y$.

1491: If $f$ is projective, there exists an isomorphism

1492: $$

1493: \rhom^{top}_{{\cal D}_X} ( \g M \otimes f_{an}^{-1} {\cal T},

1494: {\cal O}_{X^{an}}) \quad \simeq \quad

1495: \rhom^{top}_{{\cal D}_Y} \Bigl ( \int_f \g M \otimes {\cal T},

1496: {\cal O}_{Y^{an}} \Bigr ) [-\dim X/Y],

1497: $$

1498: %functorially in $\g M \in D^b_{G_{\BB C}, coh}({\cal D}_X)$

1499: %(the bounded derived category of $\modg^{coh}({\cal D}_X)$)

1500: %and ${\cal T} \in D^b_{G_{\BB R}, \BB R-c}(\BB C_Y)$.

1501: functorially in ${\cal T} \in D^b_{G_{\BB R}, \BB R-c}(\BB C_Y)$ and

1502: $\g M \in D^b_{G_{\BB C}, coh}({\cal D}_X)$

1503: (the bounded derived category of $\modg^{coh}({\cal D}_X)$).

1504: If $f$ is smooth,

1505: 

1506: 

1507: $$

1508: \rhom^{top}_{{\cal D}_X} (Lf^* \g M \otimes {\cal T}, {\cal O}_{X^{an}})

1509: \quad \simeq \quad

1510: \rhom^{top}_{{\cal D}_Y} (\g M \otimes R(f_{an})_! {\cal T},

1511: {\cal O}_{Y^{an}}) [-2\dim X/Y],

1512: $$

1513: functorially in $\g M \in D^b_{G_{\BB C}, coh}({\cal D}_Y)$

1514: and ${\cal T} \in D^b_{G_{\BB R}, \BB R-c}(\BB C_X)$.

1515: \end{thm}

1516: 

1517: Applying the second part of this theorem to the open inclusion

1518: $Z_{\BB C} \hookrightarrow M$

1519: we can rewrite our virtual representation (\ref{Whittaker}) as

1520: $$

1521: \rhom^{top}_{{\cal D}_{Z_{\BB C}}} ( \g M' \otimes {\cal S}',

1522: {\cal O}_{Z_{\BB C}^{an}}) \quad \simeq \quad

1523: \rhom^{top}_{{\cal D}_M} ( \g M \otimes {\cal S}, {\cal O}_{M^{an}}).

1524: $$

1525: The ${\cal D}_M$-module $\g M$ is ${\cal Z}(\g g_{\mathbb C})$-finite and

1526: lies in $\modg^{coh,\lambda-\rho}({\cal D}_M)$ essentially by construction.

1527: It follows from the Bruhat decomposition of $G_{\mathbb C}$ that

1528: $B_{\BB C}$ acts on $M$ with finitely many orbits; this implies that

1529: $\g M$ is admissible.

1530: It is easy to see that each maximal torus $T_{\BB C}$ acts on $M$ with

1531: finitely many fixed points.

1532: Therefore Theorem \ref{main} applies here and we obtain an integral

1533: formula for the character of the virtual representation (\ref{Whittaker}).

1534: This is probably the most complicated formula for the principal series

1535: character there is.

1536: 

1537: 

1538: \separate

1539: 

1540: \begin{thebibliography}{[KaSchm]}

1541: %\bibitem[A]{A} M. Atiyah, {\em Characters of Semisimple Lie Groups},

1542: %Lecture notes (1976), reprinted in Collected Works of Michael Atiyah,

1543: %Vol. 4, Springer, 1988, pp. 491-557.

1544: \bibitem[BB]{BB} A.~Beilinson and J.~Bernstein,

1545: {\em Localisation de $\g g$-modules}, C. R. Acad. Sci. Paris {\bf 292}

1546: (1981), 15-18.

1547: \bibitem[BGV]{BGV} N. Berline, E. Getzler, M. Vergne,

1548: {\em Heat Kernels and Dirac Operators}, Springer-Verlag, 1992.

1549: \bibitem[BV]{BV} N.~Berline and M.~Vergne,

1550: {\em Classes caract\'eristiques \'equivariantes.

1551: Formules de localisation en cohomologie \'equivariante},

1552: C. R. Acad. Sci. Paris {\bf 295} (1982), 539-541.

1553: %\bibitem[BV]{BV} N. Berline, M. Vergne,

1554: %{\em Fourier Transform of Orbits of the Coadjoint Representation},

1555: %Representation Theory of Reductive Groups,

1556: %Birkh\"auser, Basel, 1983, 53-57.

1557: \bibitem[Bi]{Bi} A.~Bialynicki-Birula,

1558: {\em Some theorems on actions of algebraic groups}, Ann. of Math.,

1559: {\bf 98} (1973), 480-497. 

1560: \bibitem[Bo]{Bo} A. Borel et al., {\em Algebraic ${\cal D}$-modules},

1561: Perspectives in Mathematics, Academic Press, 1987.

1562: %\bibitem[ChG]{ChG} N.~Chriss and V.~Ginzburg,

1563: %{\em Representation Theory and Complex Geometry},

1564: %Birkh\"auser, 1997.%\bibitem[DHV]{DHV} M. Duflo, G. Heckman, M. Vergne,

1565: %{\em Projection d'orbites, formule de Kirillov et formule de Blattner},

1566: %Mem. Soc. Math. France {\bf 15} (1984), 65-128.

1567: %\bibitem[Fo]{Fo} G. B. Folland,

1568: %{\em Real Analysis: Modern Techniques and Their Applications, Second edition},

1569: %Pure and Applied Mathematics, John Wiley \& Sons, Inc., New York, 1999.

1570: %\bibitem[DM]{DM} L. van den Dries and C. Miller,

1571: %{\em Geometric categories and o-minimal structures},

1572: %Duke Math. Jour. {\bf 84} (1996), 497-540.

1573: \bibitem[Gi]{Gi} V.~Ginzburg, {\em $\g g$-modules, Springer's representations

1574: and bivariant Chern classes}, Advances in Math. {\bf 61} (1986), 385-438.

1575: \bibitem[GM]{GM} M.~Goresky and R.~MacPherson,

1576: {\em Local contribution to the Lefschetz fixed point formula},

1577: Inventiones Math. {\bf 111} (1993), 1-33.

1578: %\bibitem[GM]{GM} M. Goresky and R. MacPherson,

1579: %{\em Stratified Morse Theory}, Ergebnisse der Mathematik,

1580: %vol 14, Springer-Verlag, Heidelberg, 1989.

1581: %\bibitem[Ka]{Ka} M. Kashiwara, {\em Index theorem for constructible sheaves},

1582: %Ast\'erisque {\bf 130} (1985), 193-209.

1583: \bibitem[GS]{GS} V. Guillemin and S. Sternberg,

1584: {\em Supersymmetry and Equivariant de Rham Theory},

1585: Springer-Verlag, 1999.

1586: %\bibitem[HC]{HC} Harish-Chandra, {\em Invariant eigendistributions on a

1587: %semisimple Lie algebra},

1588: %Inst. Hautes \'Etudes Sci. Publ. Math. {\bf 27} (1965), 5-54.

1589: \bibitem[Ka]{Ka} M.~Kashiwara, {\em Character, character cycle,

1590: fixed point theorem, and group representations},

1591: Advanced Studies in Pure Mathematics, vol. 14, Kinokuniya, Tokyo,

1592: 1988, 369-378.

1593: \bibitem[KaMF]{KaMF} M.~Kashiwara and T.~Monteiro-Fernandes,

1594: {\em Involutivit\'e des vari\'et\'es microcaract\'eristiques},

1595: Bull. Soc. Math. France {\bf 114} (1986), 393-402.

1596: \bibitem[KaScha]{KaScha} M.~Kashiwara and P.~Schapira,

1597: {\em Sheaves on Manifolds}, Springer, 1990.

1598: \bibitem[KaSchm]{KaSchm} M.~Kashiwara and W.~Schmid,

1599: {\em Quasi-equivariant ${\cal D}$-modules, equivariant derived category, and

1600: representations of reductive Lie groups},

1601: Lie Theory and Geometry, in Honor of Bertram Kostant,

1602: Progress in Mathematics, vol. 123, Birkh\"auser, Boston, 1994, pp. 457-488.

1603: %\bibitem[Kn]{Kn} A.~Knapp, {\em Lie Groups Beyond an Introduction},

1604: %Progress in Mathematics, vol. 140, Birkh\"auser, 2002.

1605: \bibitem[L1]{L1} M.~Libine,

1606: {\em A localization argument for characters of reductive Lie groups},

1607: Jour. Func. Anal. {\bf 203} (2003), 197-236, also math.RT/0206019.

1608: \bibitem[L2]{L2} M.~Libine,

1609: {\em A localization argument for characters of reductive Lie groups:

1610: an introduction and examples} in P.~Delorme, M.~Vergne (Eds.),

1611: ``Noncommutative Harmonic Analysis: In Honor of Jacques Carmona'',

1612: Progress in Mathematics, vol. 220, Birkh\"auser, 2004, pp. 375-394;

1613: also math.RT/0208024.

1614: \bibitem[L3]{L3} M.~Libine,

1615: {\em Integrals of equivariant forms and

1616: a Gauss-Bonnet theorem for constructible sheaves},

1617: math.DG/0306152, 2003.

1618: %\bibitem[LSch]{LSch} M.~Libine and W.~Schmid,

1619: %{\em Geometric methods in representation theory}, 2003,

1620: %to appear in the Proceedings of the International Euroschool and

1621: %Euroconference PQR2003.

1622: \bibitem[MQ]{MQ} V.~Mathai and D.~Quillen,

1623: {\em Superconnections, Thom classes and equivariant differential forms},

1624: Topology {\bf 25} (1986), 85-110.

1625: %\bibitem[Par]{Par} P.-E.~Paradan, {\em The Fourier Transform of

1626: %Semi-Simple Coadjoint Orbits}, Jour. Func. Anal. {\bf 163} (1999), 152-179.

1627: \bibitem[Ro]{Ro} W. Rossmann, {\em Invariant Eigendistributions on a

1628: Semisimple Lie Algebra and Homology Classes on the Conormal Variety I, II},

1629: Jour. Func. Anal. {\bf 96} (1991), 130-193.

1630: %\bibitem[R2]{R2} W.~Rossmann, {\em Kirillov's Character Formula for Reductive

1631: %Lie Groups}, Invent. Math. {\bf 48} (1978), 207-220.

1632: %\bibitem[R3]{R3} W.~Rossmann, {\em Nilpotent orbital integrals in a real

1633: %semisimple Lie algebra and representations of Weyl groups},

1634: %Operator algebras, unitary representations, enveloping algebras,

1635: %and invariant theory (Paris, 1989),

1636: %Progress in Mathematics, vol. 92, Birkh\"auser, Boston, 1990, 263-287.

1637: %\bibitem[Ru]{Ru} W. Rudin, {\em Functional Analysis},

1638: %International Series in Pure and Applied Mathematics, McGraw-Hill, 1991.

1639: \bibitem[Sch]{Sch} W.~Schmid,

1640: {\em Character formulas and localization of integrals},

1641: Deformation Theory and Symplectic Geometry, Mathematical Physics Studies,

1642: {\bf 20} (1997), Kluwer Academic Publishers, 259-270.

1643: \bibitem[SchV1]{SchV1} W. Schmid and K. Vilonen,

1644: {\em Characteristic cycles of constructible sheaves},

1645: Inventiones Math. {\bf 124} (1996), 451-502.

1646: \bibitem[SchV2]{SchV2} W. Schmid and K. Vilonen,

1647: {\em Two geometric character formulas for reductive Lie groups},

1648: Jour. AMS {\bf 11} (1998), 799-876.

1649: %\bibitem[SchV3]{SchV3} W. Schmid and K. Vilonen,

1650: %{\em Characteristic cycles and wave front cycles of representations of

1651: %{reductive groups}, Annals of Math. (2), {\bf 151} (2000), 1071-1118.

1652: \bibitem[{Sch\"u}]{Schu} J.~Sch\"urmann,

1653: {\em Topology of Singular Spaces and Constructible Sheaves},

1654: Monografie Matematyczne, vol. 63, Birkh\"auser, 2003.

1655: \bibitem[Su]{Su} H.~Sumihiro, {\em Equivariant Completion},

1656: J.~Math.~Kyoto~Univ. {\bf 14} (1974), 1-28.

1657: %\bibitem[V]{V} M. Vergne,

1658: %{\em On Rossmann's Character Formula for Discrete Series},

1659: %Invent. Math. {\bf 54} (1979), 11-14.

1660: \end{thebibliography}

1661: 

1662: \end{document}

1663: 

1664: