1: 

2: 

3: \documentclass[amssymb,twoside,10pt,amscd,reqno]{amsart}

4: \usepackage{amsmath}

5: \usepackage{amscd}

6: \usepackage{latexsym}

7: \usepackage{amsthm}

8: \usepackage{amssymb}

9: %%\usepackage[active]{srcltx} % SRC Specials: DVI [Inverse] Search

10: \textwidth = 330 pt \textheight = 530 pt

11: 

12: \theoremstyle{plain}

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

14: \newtheorem{thm}[theorem]{Theorem}

15: \newtheorem{cor}[theorem]{Corollary}

16: \newtheorem{lem}[theorem]{Lemma}

17: \newtheorem{prop}[theorem]{Proposition}

18: \newtheorem{defn}[theorem]{Definition}

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

20: 

21: \theoremstyle{definition}

22: %\newtheorem{defn}{Definition}

23: %\newtheorem{defndonstr}[defn]{Definition and Construction}

24: \newtheorem{ex}{Example}

25: \newtheorem{corex}[ex]{Corollary and Example}

26: 

27: \theoremstyle{remark}

28: \newtheorem{remark}{Remark}

29: 

30: %\newcommand{\binom}[2]{\left( \begin{array}{cc} #1 \\ #2 \end{array} \right)}

31: \newcommand{\Hdg}[2]{\mathcal{H}^{#1,#2}}

32: \newcommand{\pr}{\text{pr}}

33: \newcommand{\mc}{\mathcal}

34: \newcommand{\NN}{\mathbb{N}}

35: \newcommand{\QQ}{\mathbb{Q}}

36: \newcommand{\ZZ}{\mathbb{Z}}

37: \newcommand{\CC}{\mathbb{C}}

38: \newcommand{\RR}{\mathbb{R}}

39: \newcommand{\AAA}{\mathbb{A}}

40: \newcommand{\PP}{\mathbb{P}}

41: \newcommand{\PPP}{\PP_{\CC}}

42: \newcommand{\SP}{\text{Spec }}

43: \newcommand{\HB}[3]{\text{Hilb}_{#1}({#2}/{#3})}

44: \newcommand{\HI}[2]{\text{Hilb}_{#1}(#2)}

45: \newcommand{\Hi}[1]{\text{Hilb}_{#1}}

46: \newcommand{\GR}[3]{\text{Grass}_{#1}(#2,#3)}

47: \newcommand{\kgn}[4]{\text{M}_{#1,#2}(#3,#4)}

48: 

49: \def\ul{\underline}

50: \def\la{\longrightarrow}

51: \def\ni{\noindent}

52: \def\cl{\centerline}

53: 

54: \def\Sym{{\rm Sym}}

55: \def\pic{{\rm Pic}}

56: 

57: \def\A{{\mathbb A}}

58: \def\C{{\mathbb C}}

59: \def\F{{\mathbb F}}

60: \def\G{{\mathbb G}}

61: \def\P{{\mathbb P}}

62: \def\Q{{\mathbb Q}}

63: \def\R{{\mathbb R}}

64: \def\Z{{\mathbb Z}}

65: 

66: \def\cC{{\mc C}}

67: \def\cG{{\mc G}}

68: \def\cX{{\mc X}}

69: \def\cE{{\mc E}}

70: \def\cF{{\mc F}}

71: \def\cH{{\mc H}}

72: \def\cL{{\mc L}}

73: \def\cK{{\mc K}}

74: 

75: \def\ps{\vspace{4pt}}

76: 

77: \newcommand{\lt}{\left}

78: \newcommand{\rt}{\right}

79: 

80: \newcommand{\g}{{\frak g}}

81: \newcommand{\gs}{{\frak g}^{\ast}}

82: \newcommand{\HH}{\mathcal H}

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

84: \newcommand{\Ci}{C^{\infty}}

85: \newcommand{\B}{{\cal B}}

86: \newcommand{\Cs}{C^{\ast}}

87: \newcommand{\Ws}{W^{\ast}}

88: \newcommand{\mapdown}[1]{\downarrow

89: \rlap{$\vcenter{\hbox{$\scriptstyle#1$}}$}}

90: 

91: \setcounter{tocdepth}{1}

92: 

93: 

94: \begin{document}

95: 

96: \title{Quot functors for Deligne-Mumford stacks}

97: \author[Olsson]{Martin Olsson}

98: 

99: \author[Starr]{Jason Starr}

100: 

101: 

102: 

103: 

104: 

105: \maketitle

106: 

107: \noindent \small

108: \address{Department of Mathematics,  Massachusetts Institute of

109: Technology,  Cambridge, MA 02139,} \email{molsson@math.mit.edu}

110: 

111: \

112: 

113: \noindent

114: \address{Department of Mathematics, Massachusetts Institute of

115: Technology, Cambridge, MA 02139,} \email{jstarr@math.mit.edu}

116: 

117: \

118: 

119: \normalsize  \centerline {\emph{Dedicated to S. Kleiman on the

120: occasion of his 60th birthday}}

121: 

122: \begin{abstract}

123:   Given a separated and locally finitely-presented Deligne-Mumford

124:   stack $\cX$ over an algebraic space $S$, and a locally

125:   finitely-presented $\OO_{\cX}$-module $\cF$, we prove that the Quot

126:   functor $\text{Quot}(\cF/\cX/S)$ is represented by a separated and

127:   locally finitely-presented algebraic space over $S$.  Under

128:   additional hypotheses, we prove that the connected components of

129:   $\text{Quot}(\cF/\cX/S)$ are quasi-projective over $S$.

130: \end{abstract}

131: 

132: %\maketitle

133: 

134: \tableofcontents

135: 

136: \section{Statement of results}

137: 

138: Let $p:\cX\rightarrow S$ be a separated, locally finitely-presented

139: $1$-morphism from a Deligne-Mumford stack $\cX$ to an algebraic space

140: $S$.  Let $\cF$ be a quasi-coherent $\OO_\cX$-module (on the \'etale

141: site of $\cX$) such that $\cF$ is locally finitely-presented.  Define

142: a contravariant functor

143: \begin{equation}

144: Q=Q(\cF/\cX/S): S-\text{schemes} \rightarrow \text{Sets}

145: \end{equation}

146: as follows.  For each $S$-scheme $f:Z\rightarrow S$, define $\cX_Z$ to

147: be $\cX\times_S Z$, and define $\cF_Z$ to be the pullback of $\cF$ to

148: $\cX_Z$.  We define $Q(Z)$ to be the set of $\OO_{\cX_Z}$-module

149: quotients $\cF_Z\rightarrow G$ which satisfy

150: \begin{enumerate}

151: \item $G$ is a quasi-coherent $\OO_{\cX_Z}$-module which is locally

152:   finitely-presented,

153: \item $G$ is flat over $Z$,

154: \item the support of $G$ is proper over $Z$.

155: \end{enumerate}

156: All of these properties are preserved by base-change on $Z$, and

157: therefore pullback makes $Q$ into a contravariant functor.

158: 

159: \begin{thm}~\label{thm-thm1}  $Q$ is represented by an algebraic space

160:   which is separated and locally finitely-presented over $S$.  If

161:   $\cF$ has proper support over $S$, then $Q\rightarrow S$ satisfies

162:   the valuative criterion for properness.

163: \end{thm}

164: 

165: \textbf{Remark:} Under the hypothesis that $\cF$ has proper support

166: over $S$, we are not claiming that $Q\rightarrow S$ is proper, because

167: we do not show that $Q\rightarrow S$ is quasi-compact (in general it

168: need not be).

169: 

170: \

171: 

172: We give a better description of $Q$ under additional hypotheses on

173: $\cX$.  Our first hypothesis is that $\cX$ is a global quotient.

174: 

175: \begin{defn}~\label{defn-gq}  Let $S$ be an algebraic space.  A

176:   \emph{global quotient stack} over $S$ is an (Artin) algebraic stack

177:   $\cX$ over $S$ which is isomorphic to a stack of the form $[Z/G],$

178:   where $Z$ is an algebraic space which is finitely-presented over

179:   $S$, and $G$ is a flat, finitely-presented group scheme over $S$

180:   which is a subgroup scheme of the general linear group scheme

181:   $\text{GL}_{n,S}$ for some $n$.

182: \end{defn}

183: 

184: This is essentially ~\cite[definition 2.9]{EHKV} (with the Noetherian

185: hypotheses replaced by a finite-presentation hypothesis). We remind

186: the reader of the following characterization in~\cite[remark

187: 2.11]{EHKV}.

188: 

189: \begin{lem}~\label{rmk-gq}

190:   Suppose that $\cX$ is a global quotient stack over $S$ which is

191:   isomorphic to $[Z/G]$ as above.  Then the diagonal action of $G$ on

192:   $Z\times_S GL_{n,S}$ is free and the quotient $Z'=[Z\times_S

193:   GL_{n,S}/G]$ is an algebraic space with a right action of

194:   $GL_{n,S}$.  The quotient stack $[Z'/GL_{n,S}]$ is isomorphic to the

195:   original stack $[Z/G]$.

196: \end{lem}

197: 

198: So every quotient stack is isomorphic to a stack of the form

199: $[Z'/GL_{n,S}]$.

200: 

201: \

202: 

203: Our second hypothesis is that $\cX$ is a \emph{tame} Deligne-Mumford

204: stack.

205: 

206: \begin{defn}~\label{defn-tame}

207:   A Deligne-Mumford stack $\cX$ is \emph{tame} if for any

208:   algebraically closed field $k$ and any $1$-morphism $\zeta:\SP

209:   k\rightarrow \cX$, the stabilizer group $\text{Aut}(\zeta)(\SP k)$

210:   has order prime to char$(k)$, where $\text{Aut}(\zeta)$ is the

211:   finite $k$-group scheme defined to be the Cartesian product of the

212:   diagram

213: \begin{equation}

214: \begin{CD}

215:   @. \SP k\\

216:   @. @VV(\zeta, \zeta )V \\

217:   \cX @>\Delta >> \cX \times \cX.

218: \end{CD}

219: \end{equation}

220: Here $\Delta:\cX\rightarrow \cX \times \cX$ is the diagonal morphism.

221: \end{defn}

222: 

223: \begin{thm}~\label{thm-thm2}

224:   Suppose that $S$ is an affine scheme, and let $f:\cX\rightarrow S$

225:   be a separated $1$-morphism from a tame Deligne-Mumford stack to $S$

226:   such that $\cX$ is a global quotient over $S$ and such that the

227:   coarse moduli space $X$ of $\cX$ is a quasi-projective $S$-scheme

228:   (resp. projective $S$-scheme).  Then the connected components of $Q$

229:   are quasi-projective $S$-schemes (resp.  projective $S$-schemes).

230: \end{thm}

231: 

232: \textbf{Remark:} The existence of the coarse moduli space for $\cX$

233: follows from ~\cite{K-M}.

234: 

235: \

236: 

237: \textbf{Remark:} The condition that $S$ be an affine scheme is

238: required because the property of being quasi-projective is not Zariski

239: local on the base.

240: 

241: \

242: 

243: \textbf{Acknowledgments:}  The authors benefitted from useful

244: conversations with Johan de Jong.  The authors also

245: wish to thank the referee,

246: particularly for his suggestion to simplify section~\ref{gensheavsec}

247: by including theorem~\ref{thm-genpt}.

248: 

249: This work was largely done

250: during the semester program on Algebraic Stacks, Intersection Theory

251: and Non-Abelian Hodge Theory at MSRI during the spring of 2002

252: where both authors were MSRI

253: Postdoctoral Fellows.  The first author was partially supported by an NSF

254: Postdoctoral Fellowship.

255: The second author was partially supported by NSF grant DMS-0201423.

256: 

257: \section{Representability by an algebraic space}~\label{sec-M}

258: 

259: 

260: %Let $S$ be an algebraic space,  $f:\cX\rightarrow S$  a morphism of

261: %finite presentation from a Deligne-Mumford stack to $S$, and let

262: %$\mc F$ be a coherent sheaf on $X$.  Define $Q$ to be the functor

263: %\[

264: %Q:(\hbox{$S$-schemes})\longrightarrow Set

265: %\]

266: %which to any $T/S$ associates the isomorphism classes of quotients

267: %$\mc F_T\rightarrow \mc G$, where $\mc G$ is a coherent sheaf on

268: %$X_T$, flat over $T$,  with proper support. Here $X_T := X\times

269: %_ST$ and $\mc F_T$ denotes the pullback of $\mc F$ to $X_T$.

270: 

271: %\begin{thm}\label{1:mainthm} $Q$ is representable by an algebraic space.  If $f$ is

272: %separated (resp. proper), then $Q$ is separated (resp. satisfies

273: %the valuative criterion for properness).

274: %\end{thm}

275: 

276: %The proof occupies the remainder of the section.

277: 

278: In this section we prove theorem \ref{thm-thm1}.

279: 

280: Note first that $Q$ is a sheaf for the fppf-topology by descent

281: theory, and is limit preserving. In addition, for each open

282: substack $\mc U\subset \cX$ there is a natural open immersion

283: $Q(\mc F|_{\mc U}/\mc U/S)\subset Q$.  Moreover, $Q$ is the union

284: over finitely presented open substacks $\mc U\subset \mc X$ of the

285: $Q(\mc F|_{\mc U}/\mc U/S)$.  We may therefore assume that $\mc X$

286: is of finite presentation over $S$. Since the question of

287: representability of $Q$ is \'etale local on $S$ we may assume that

288: $S$ is an affine scheme, and by a standard limit argument we may

289: assume that $S$ is of finite type over $\text{Spec}(\mathbb{Z})$.

290: 

291: Under these assumptions we

292:  prove the theorem by verifying the conditions of theorem 5.3 of

293: \cite{Artin}.

294: 

295: \

296: 

297: \noindent \emph{Commutation with inverse limits.}

298: We need the Grothendieck existence theorem for Deligne-Mumford stacks:

299: \begin{prop}\label{stackgrothex}

300: Let $A$ be a complete noetherian local ring, $\cX /A$  a

301: Deligne-Mumford stack of finite type, and let $A_n =

302: A/\mathfrak{m}_A^{n+1}$, $\cX_n = \cX\otimes _AA_n$. Then the

303: natural functor

304: \[

305: \begin{CD}

306: (\hbox{coherent sheaves on $\cX$ with support proper over $A$})\\

307: @VVV \\

308: (\hbox{compatible families of coherent sheaves on the $\cX_n$ with

309: proper support})

310: \end{CD}

311: \]

312: is an equivalence of categories.

313: \end{prop}

314: \begin{proof}

315: This proposition is perhaps best viewed in a context of formal

316: algebraic stacks, but since we do not want to develop such a

317: theory here we take a more ad-hoc approach.

318: 

319: Let $\widehat {\cX}$ be the ringed topos $(\cX_{0,et}, \mc O_{\widehat

320: {\cX}})$, where $\mc O_{\widehat {\cX}}$ is the sheaf of rings which

321: to any \'etale $\cX_0$-scheme $U_0\rightarrow \cX_0$ associates

322: \[

323: \varprojlim \Gamma (U_n, \mc O_{U_n}).

324: \]

325: Here $U_n$ denotes the unique lifting of $U_0$ to an \'etale

326: $\cX_n$-scheme.  There is a natural morphism of ringed topoi

327: \[

328: j:\widehat \cX\rightarrow \cX_{et}.

329: \]

330: If $\mc F$ is a sheaf of $\mc O_\cX$-modules, then we denote by

331: $\widehat {\mc F}$ the sheaf $j^*\mc F$.  Note that the functor $\mc F\mapsto

332: \widehat {\mc F}$ is an exact functor.

333: 

334: \begin{lem}\label{cohlem}

335: If $\mc F$ and $\mc G$ are coherent sheaves on $\cX$ with proper

336: support over $A$, then for every integer $n$ the natural map

337: \[

338: \hbox{Ext}^n_{\mc O_\cX}(\mc F, \mc G)\longrightarrow {Ext}^n_{\mc

339: O_{\widehat {\cX}}}(\widehat {\mc F}, \widehat {\mc G})

340: \]

341: is an isomorphism.

342: \end{lem}

343: \begin{proof}

344: Observe first that the natural map

345: \[

346: \mc Ext^n_{\mc O_\cX }(\mc F, \mc G)^{\wedge }\rightarrow \mc

347: Ext^n_{\mc O_{\widehat \cX }}(\widehat {\mc F},\widehat  {\mc G})

348: \]

349: is an isomorphism.

350: Indeed this can be verified locally and so follows from the theory of formal

351: schemes.

352: From this and the local-to-global spectral sequence for $\hbox{Ext}$ it

353: follows that it suffices to show that for any coherent sheaf $\mc

354: F$ with proper support, the natural map

355: \begin{equation}\label{cohmap}

356: H^n(\cX, \mc F)\longrightarrow H^n(\widehat {\cX}, \widehat {\mc F})

357: \end{equation}

358: is an isomorphism for every $n$.  We prove this by induction on

359: the dimension of $\cX$.

360: 

361: The key observation is that if $a:U\rightarrow \cX$ is a finite

362: morphism from a scheme to $\cX$, then the map \ref{cohmap} is known

363: to be an isomorphism for all $n$ in the case when $\mc F$ is equal

364: to $a_*\mc F'$ for some coherent sheaf $\mc F'$ on $U$ (since $a$

365: is finite and \cite{EGA}. III.5).

366: 

367: If the dimension of $\cX$ is zero, then we can by (\cite{LM-B},

368: 16.6) find a finite \'etale cover $a:U\rightarrow \cX$ of $\cX$ by a

369: scheme $U$. If $b:U\times _\cX U\rightarrow \cX$ denotes the canonical

370: map, then the sequences

371: \[

372: \mc F\rightarrow a_*a^*\mc F\rightrightarrows b_*b^*\mc F

373: \]

374: \[

375: \widehat {\mc F}\rightarrow a_*a^*\widehat {\mc

376: F}\rightrightarrows b_*b^*\widehat {\mc F}

377: \]

378: are exact so the result is true for $n = 0$.

379: 

380: To prove the result for general $n$, assume the result is true for

381: $n-1$ and let $\mc G = a_*a^*\mc F/\mc F$ so that we have an exact

382: sequence of coherent sheaves

383: \[

384: 0\rightarrow \mc F \rightarrow a_*a^*\mc F\rightarrow \mc

385: G\rightarrow 0.

386: \]

387: Then the commutative diagram

388: \[

389: \begin{matrix}

390: H^{n-1}(\cX, a_*a^*\mc F)& \rightarrow & H^{n-1}(\cX, \mc G) &

391: \rightarrow

392: & H^n(\cX, \mc F) & \rightarrow &H^n(\cX, a_*a^*\mc F)\\

393: \downarrow && \downarrow && \downarrow && \downarrow \\

394: H^{n-1}(\widehat \cX, a_*a^*\widehat {\mc F})& \rightarrow &

395: H^{n-1}(\widehat \cX, \widehat {\mc G}) & \rightarrow & H^n(\widehat

396: \cX, \widehat {\mc F}) & \rightarrow &H^n(\widehat \cX, a_*a^*\widehat

397: {\mc F})

398: \end{matrix}

399: \]

400: shows that the map $H^n(\cX, \mc F)\rightarrow H^n(\widehat \cX,

401: \widehat {\mc F})$ is injective.  But then the map $H^n(\cX, \mc

402: G)\rightarrow H^n(\widehat {\cX}, \widehat {\mc G})$ is also

403: injective, and an analysis of the diagram

404: \[

405: \begin{matrix}

406: H^{n-1}(\cX, \mc G)& \rightarrow & H^{n}(\cX, \mc F) & \rightarrow

407: & H^n(\cX, a_*a^*\mc F) & \rightarrow &H^n(\cX, \mc G)\\

408: \downarrow && \downarrow && \downarrow && \downarrow \\

409: H^{n-1}(\widehat \cX, \widehat {\mc G})& \rightarrow &

410: H^{n}(\widehat \cX, \widehat {\mc F})& \rightarrow & H^{n}(\widehat

411: \cX, a_*a^*\widehat {\mc F}) & \rightarrow & H^n(\widehat \cX,

412: \widehat {\mc G})

413: \end{matrix}

414: \]

415: reveals that the map $H^n(\cX, \mc F)\rightarrow H^n(\widehat \cX,

416: \widehat {\mc F})$ is an isomorphism.  This completes the proof of

417: the case when $dim(\cX) = 0$.

418: 

419: To prove the result for general $\cX$ we assume the result is true

420: for $dim(\cX)-1$ and  proceed by induction on $n$. By (\cite{LM-B},

421: 16.6) there exists a finite surjective morphism $a:U\rightarrow \cX$

422: which is generically \'etale.  Let $b:U\times _{\cX }U\rightarrow \cX$ be

423: the natural map and let $K$ be the kernel of $a_*a^*\mc

424: F\rightarrow b_*b^*\mc F$. Note that the map \ref{cohmap} for $K$

425: is an isomorphism when $n=0$. There is a natural map $\mc

426: F\rightarrow K$ which is generically an isomorphism.  Let $\mc F'$

427: be the kernel of this map and let $\mc F^{\prime \prime }$ be the

428: cokernel. Then $\mc F'$ and $\mc F^{\prime \prime }$ have

429: lower-dimensional support and so the map \ref{cohmap} is an

430: isomorphism for these sheaves.  If $\mc G$ denotes the image of

431: $\mc F\rightarrow K$, then the map $H^0(\cX, \mc G)\rightarrow

432: H^0(\widehat {\cX}, \widehat {\mc G})$ is an isomorphism by the

433: corresponding result for $K$ and $\mc F^{\prime \prime }$ and the

434: exact sequence

435: \[

436: 0\rightarrow \mc G\rightarrow K\rightarrow \mc F^{\prime \prime

437: }\rightarrow 0.

438: \]

439: Then from the exact sequence

440: \[

441: 0\rightarrow \mc F'\rightarrow \mc F\rightarrow \mc G\rightarrow 0

442: \]

443: we deduce that the map $H^0(\cX, \mc F)\rightarrow H^0(\widehat \cX,

444: \widehat {\mc F})$ is an isomorphism.

445: 

446: To prove the result for $n$ assuming the result for $n-1$, note

447: that a similar argument to the one above shows that if  the map

448: $H^n(\cX, K)\rightarrow H^n(\widehat {\cX}, \widehat K)$ is injective

449: (resp. an isomorphism) then the map \ref{cohmap} is injective

450: (resp. an isomorphism).  Therefore the proof is completed by using

451: the argument of the case $dim(\cX) = 0$ with $K$ replacing $\mc F$.

452: \end{proof}

453: 

454: %The lemma implies that the functor in \ref{stackgrothex} is fully

455: %faithful.  To prove essential surjectivity, note that we can view

456: Now observe that the category of compatible families of coherent sheaves on the

457: $\cX_n$ with proper support is naturally viewed

458: as a full subcategory of the category of

459: sheaves of $\mc O_{\widehat \cX}$-modules on $\widehat {\cX}$ using

460: the functor which sends a family $\{\mc F_n\}$ to the sheaf

461: $\widehat {\mc F}$ associated to the presheaf

462: \[

463: U_0\mapsto \varprojlim \Gamma (U_0, \mc F_n).

464: \]

465: The functor $\{\mc F_n\}\mapsto \widehat {\mc F}$ is fully

466: faithful and identifies the category of compatible systems of coherent

467: sheaves on the $\cX_n$ with proper support with a subcategory of the

468: category of sheaves of $\mc O_{\widehat {\cX}}$-modules which is

469: closed under the formation of kernels, cokernels, and extensions.

470: Indeed, these assertions can be verified locally on $\cX$ and hence follow from

471: the corresponding statements for formal schemes.  From this it and the

472: lemma it follows that the functor in \ref{stackgrothex} is fully faithful.

473: 

474: Now the

475: functor from coherent sheaves on $\cX$ with proper support to

476: sheaves of $\mc O_{\widehat \cX}$-modules on $\widehat \cX$ identifies

477: the category of such sheaves with a full subcategory of the

478: category of sheaves of $\mc O_{\widehat \cX}$-modules which is

479: stable under the formation of kernels, cokernels, and extensions

480: (by lemma \ref{cohlem}).  Thus the following two lemmas prove

481: proposition \ref{stackgrothex}.

482: 

483: \begin{lem}

484: Let $\mc A$ be an abelian category and let $\mc A'\subset \mc A$

485: be a full subcategory which is stable under the formation of

486: kernels, cokernels, and extensions.  Then any object of $\mc A$

487: which  admits a morphism to an object of $\mc A'$ such that the

488: kernel and cokernel are in $\mc A'$ is in $\mc A'$.

489: \end{lem}

490: \begin{proof}

491: By assumption such an object $\mc F\in \mc A$ sits in an exact

492: sequence

493: \[

494: 0\rightarrow K\rightarrow \mc F\rightarrow \mc F'\rightarrow

495: Q\rightarrow 0

496: \]

497: where $K, \mc F', Q\in \mc A'$.  Let $K' = Ker(\mc F'\rightarrow

498: Q)$.  Then $K'\in \mc A'$, and we have an exact sequence

499: \[

500: 0\rightarrow K\rightarrow \mc F\rightarrow K'\rightarrow 0.

501: \]

502: Therefore $\mc F\in \mc A'$.

503: \end{proof}

504: 

505: 

506: \begin{lem}

507: For every compatible family of coherent sheaves $\{\mc F_n\}$ with

508: proper support and associated sheaf $\widehat {\mc F}$ on

509: $\widehat {\cX}$, there exists a morphism $\widehat {\mc

510: F}\rightarrow j^*\mc G$ for some coherent sheaf $\mc G$ on $\cX$

511: with proper support such that the kernel and cokernel are

512: isomorphic to the pullbacks of coherent sheaves on $\cX$ with proper

513: support.

514: \end{lem}

515: \begin{proof}

516: We proceed by induction on the dimension of the support of

517: $\widehat {\mc F}$.

518: 

519: If the support of $\widehat {\mc F}$ has dimension zero, then for

520: any \'etale cover $a:U\rightarrow \cX$ the inverse image of the

521: support of $\widehat {\mc F}$ in $\widehat {U}$ is a closed

522: subscheme which is of dimension zero, hence is proper.  Therefore,

523: there exists  a unique sheaf on $U$ inducing the restriction of

524: $\widehat {\mc F}$ to $\widehat {U}$ (by \cite{EGA} III.5).

525: Moreover, by the uniqueness this sheaf comes with descent datum

526: relative to $a$. Hence $\widehat {\mc F}$ is induced from a

527: coherent sheaf on $\cX$ with proper support.

528: 

529: As for the general case, choose a morphism $a:U\rightarrow \cX$

530: which is finite and generically \'etale (such a morphism exists by

531: \cite{LM-B}, 16.6), and let $b:U\times _{\cX}U\rightarrow \cX$ be the

532: natural map.  Then $a_*a^*\widehat {\mc F}$ and $b_*b^*\widehat

533: {\mc F}$ are obtained from coherent sheaves on $\cX$, and hence so

534: is

535: \[

536: K := Ker(a_*a^*\widehat {\mc F}\rightarrow b_*b^*\widehat {\mc

537: F}).

538: \]

539: Moreover, there is a natural map $\widehat {\mc F}\rightarrow K$

540: which is generically an isomorphism.  Hence the kernels and

541: cokernels have lower dimensional support and by induction are

542: obtained from coherent sheaves on $\cX$.

543: \end{proof}

544: 

545: \end{proof}

546: 

547: \noindent \emph{Separation Conditions.}  These follow by the same

548: reasoning as in (\cite{Artin}, page 64).

549: 

550: \

551: 

552: \noindent \emph{Deformation theory.}

553: 

554: Suppose given a deformation situation (in the sense of

555: \cite{Artin})

556: \[

557: A'\rightarrow A\rightarrow A_0

558: \]

559: and a quotient $\mc F_A\rightarrow \mc G_A$ over $\cX_A =

560: \cX\times _S\text{Spec}(A)$ giving an element of $Q(A)$.  Suppose

561: further that a map $B\rightarrow A$ is given.  Then it is

562: well-known (see for example \cite{Sch}, 3.4) that the map

563: \begin{equation}\label{schmap}

564: Q(A'\times _BA)\longrightarrow Q(A')\times _{Q(A)}Q(B)

565: \end{equation}

566: is a bijection.

567: 

568: 

569: If $M = Ker(A'\rightarrow A)$, then it follows from the

570: bijectivity of \ref{schmap} that a deformation theory in the sense

571: of (\cite{Artin}) is provided by the module $Q_{\mc G_0}(A_0[M])$

572: (see for example loc. cit. page 47).  Here $A_0[M]$ denotes the

573: ring with underlying module $A_0\oplus M$ and multiplication

574: \[

575: (a, m)\cdot (a', m') = (aa', am'+a'm),

576: \]

577: and $Q_{\mc G_0}(A_0[M])$ denotes the set of elements in

578: $Q(A_0[M])$ whose image in $Q(A_0)$ is the reduction $\mc G_0$ of

579: $\mc G$.  The conditions on the deformation theory of

580: (\cite{Artin}, theorem 5.3) are therefore satisfied by the

581: following lemma and standard properties of cohomology.

582: \begin{lem}

583: Let $\mc H_0 = Ker(\mc F_0\rightarrow \mc G_0)$ (where $\mc F_0$

584: denotes the reduction of $\mc F$ to $\cX_{A_0}$). Then there is a

585: natural $A_0$-module isomorphism

586: \[

587: Q_{\mc G_0}(A_0[M]) \simeq \hbox{Ext}^0(\mc H_0, \mc G_0\otimes

588: M).

589: \]

590: \end{lem}

591: \begin{proof}

592: Let

593: \[

594: \begin{CD}

595: 0@>>> \mc G_0\otimes M@>>> \mc G_0\otimes M\oplus \mc F_0@>>> \mc F_0@>>> 0

596: \end{CD}

597: \]

598: be the sequence obtained by pushing out the sequence

599: \[

600: \begin{CD}

601: 0@>>> \mc F_0\otimes M@>>> \mc F_0\oplus \mc F_0\otimes M@>>> \mc G_0@>>> 0

602: \end{CD}

603: \]

604: via the given map $\mc F_0\rightarrow \mc G_0$.  Then to give a lifting of

605: $\mc G_0$ is equivalent to giving a sub-$\mc O_{\cX _{A_0}}$-module

606: $\mc H\subset \mc F_0\oplus \mc G_0\otimes M$ such that the induced map to $\mc F_0$

607: induces an isomorphism $\mc H\simeq \mc H_0$.  Note that such a sub-module

608: is automatically a sub-$\mc O_{\cX _{A_0[M]}}$-module.

609: In other words, the set of liftings

610: of $\mc G_0$ is in natural bijection with the set of maps $\mc H_0\rightarrow \mc F_0

611: \oplus \mc G_0\otimes M$ lifting the inclusion $\mc H_0\subset \mc F_0$.  But to give

612: such a map is precisely equivalent to giving a morphism $\mc H_0\rightarrow \mc G_0

613: \otimes M$.  The verification that this bijection is a module homomorphism

614: is left to the reader.

615: 

616: 

617: %We define a map

618: %\begin{equation}\label{Qcanmap}

619: %Q_{\mc G_0}(A_0[M]) \longrightarrow \hbox{Ext}^0(\mc H_0, \mc

620: %G_0\otimes M)

621: %\end{equation}

622: %as follows.  If $\mc F_0\oplus \mc F_0\otimes M\rightarrow \mc G$

623: %is a surjection defining a lifting of $\mc G_0$ to $A_0[M]$, then

624: %the composite

625: %\[

626: %\begin{CD}

627: %\mc H_0@>h\mapsto (h,0)>> \mc F_0\oplus \mc F_0\otimes M@>>> \mc

628: %G@>>> \mc G_0

629: %\end{CD}

630: %\]

631: %is zero, and hence $\mc G$ defines a map $\mc H_0\rightarrow \mc

632: %G_0\otimes M$.

633: 

634: %To check that \ref{Qcanmap} is an isomorphism, we may work locally

635: %on $\cX_{A_0}$ (ignoring the issue of proper support) since sheaves

636: %and morphisms of sheaves can be constructed locally.  Thus it

637: %suffices to consider the case when $\cX_{A_0}$ is an affine scheme.

638: %In this case, any extension of $\mc G_0$ by $\mc G_0\otimes M$ is

639: %split, so all extensions are obtained from a map $\gamma :\mc

640: %F_0\oplus \mc F_0\otimes M\rightarrow \mc G_0\oplus \mc G_0\otimes

641: %M$ filling in the diagram

642: %\[

643: %\begin{matrix}

644: %0&\rightarrow &\mc F_0\otimes M&\rightarrow &\mc F_0\oplus \mc

645: %F_0\otimes M&\rightarrow &\mc F_0&\rightarrow &0\\

646: %&&\downarrow &&  && \downarrow &&\\

647: %0&\rightarrow &\mc G_0\otimes M&\rightarrow &\mc G_0\oplus \mc

648: %G_0\otimes M&\rightarrow &\mc G_0&\rightarrow &0.

649: %\end{matrix}

650: %\]

651: %Now to give such a map $\gamma $ is equivalent to giving a map

652: %$\alpha :\mc F_0\rightarrow \mc G_0\otimes M$, and changing the

653: %choice of the splitting of

654: %\[

655: %0\rightarrow \mc G_0\otimes M\rightarrow \mc G_0\oplus \mc

656: %G_0\otimes M\rightarrow \mc G_0\rightarrow 0

657: %\]

658: %changes the map $\alpha $ by translation by a map $\beta :\mc

659: %G_0\rightarrow \mc G_0\otimes M$.  Thus $Q_{\mc G_0}(A_0[M])$ is

660: %canonically in bijection with

661: %\[

662: %Cok(\hbox{Hom}(\mc G_0, \mc G_0\otimes M)\rightarrow

663: %\hbox{Hom}(\mc F_0, \mc G_0\otimes M)) = \hbox{Hom}(\mc H_0, \mc

664: %G_0\otimes M).

665: %\]

666: %>From this it follows that the map is an isomorphism.

667: \end{proof}

668: 

669: 

670: 

671: 

672: \noindent \emph{Conditions on the obstructions.}

673: 

674: Our understanding of the obstruction theory of $Q$ will be in a

675: 2-step approach (correcting a mistake in \cite{Artin}).  Let

676: $A'\rightarrow A\rightarrow A_0$ be a deformation situation as

677: above and $\mc F_A\rightarrow \mc G_A$ an object in $Q(A)$.

678: For any quotient $\epsilon :M\rightarrow M_\epsilon $, let $A_\epsilon $ be the quotient

679: of $A'$ by the kernel of $\epsilon $.

680: For such an $\epsilon $,

681: the

682: first obstruction to lifting $\mc G_A$ to $A_\epsilon $ is that the map

683: \[

684: M\otimes \mc F_{A_0}\longrightarrow M_\epsilon \otimes \mc G_{A_0}

685: \]

686: factors through $M\mc F_{A'}$.  If we let $\mc T$ be the kernel

687: of $M\otimes \mc F_{A_0}\rightarrow \mc F_{A'}$, then we want that the map

688: \begin{equation}\label{tormap}

689: \mc T\longrightarrow M_\epsilon \otimes \mc G_{A_0}

690: \end{equation}

691: is zero.  If this is the case, then there is a canonical map $M\mc

692: F_{A'}\rightarrow M_\epsilon \otimes \mc G_{A_0}$ and the condition that

693: there exists an element in $Q(A_\epsilon )$ inducing $\mc G$ is equivalent

694: to the statement that the resulting extension

695: \[

696: 0\rightarrow M_\epsilon \otimes \mc G_{A_0}\rightarrow E\rightarrow \mc

697: F_A\rightarrow 0

698: \]

699: is obtained from an extension of $\mc G_{A}$ by $M_\epsilon \otimes \mc

700: G_{A_0}$. In other words, if the map \ref{tormap} is zero, then if

701: we let $\mc H =Ker(\mc F_A\rightarrow \mc G_A)$ there exists a

702: canonical obstruction in

703: \[

704: \hbox{Ext}^1_{\cX_{A'}}(\mc H, M_\epsilon \otimes \mc G_{A_0})

705: \]

706: whose vanishing is necessary and sufficient for the existence of a

707: lifting.

708: 

709: Now the only condition on the obstructions in (\cite{Artin}, 5.3)

710: which does not follow immediately from the bijectivity of

711: \ref{schmap}, is condition (5.3, [5${}'$].c).

712: 

713: Thus suppose given a deformation situation as before, $\xi \in

714: Q(A)$, and suppose further that $M$ is free of rank $n$.  Let $K$

715: be the field of fractions of $A_0$, and denote by a subscript $K$

716: the localizations at the generic point of $\text{Spec}(A_0)$.  We

717: suppose that for every one-dimensional quotient $M_K\rightarrow

718: M_K^*$ there is a non-trivial obstruction to lifting $\xi _K$ to

719: $Q(A_K^*)$, where $A_K^*$ denotes the extension defined by

720: $M_K^*$.  Then we have to show that there exists a non-empty open

721: subset $U\subset \text{Spec}(A_0)$ such that for every quotient

722: $\epsilon $ of $M$ of length one with support in $U$, $\xi $ does

723: not lift to $A_\epsilon $ (the extension obtained from $\epsilon

724: $).

725: 

726: Let

727: \[

728: \phi \in \hbox{Ext}^0(\mc T, M\otimes \mc

729: G_{A_0})\simeq \hbox{Ext}^0(\mc T, \mc

730: G_{A_0})\otimes M

731: \]

732: be the class defined by \ref{tormap} in the case when $\epsilon $ is the identity.

733:  We reduce to the case when

734: $\phi  = 0$.  Once this reduction is made the argument of

735: (\cite{Artin}, page 66) will finish the proof.

736: 

737: 

738: To make the reduction, we can by shrinking on $\text{Spec}(A_0)$

739: assume that

740: $$ \hbox{Ext}^0(\mc T, \mc G_{A_0})$$ is a free module; say of

741: rank $r$.  In addition, by the argument of (\cite{Artin}, page

742: 66), we can after shrinking $\text{Spec}(A_0)$ assume that for

743: each point $s\in \text{Spec}(A_0)$, the natural map

744: \begin{equation}\label{pointeq}

745: \hbox{Ext}^0(\mc T, \mc G_{A_0})\otimes

746: k(s) \longrightarrow  \hbox{Ext}^0(\mc T,

747: \mc G_{A_0}\otimes k(s))

748: \end{equation}

749: is an isomorphism.

750: 

751: Choosing a basis for $ \hbox{Ext}^0(\mc T,

752: \mc G_{A_0})$ we can think of $\phi $ as an element

753: \[

754: \phi = (\phi _1, \dots, \phi _r)\in M^r.

755: \]

756: Let $N\subset M$ be the submodule of $M$ generated by the $\phi

757: _i$, and let $M' = M/N$.  After further shrinking

758: $\text{Spec}(A_0)$ we can assume that $M'$ is a free module.  Now

759: note that any length-one quotient $M\rightarrow M_{\epsilon }$ for

760: which the obstruction to lifting $\xi $ goes to zero factors

761: through $M'$ by \ref{pointeq}.

762:  Moreover, any such quotient which does not factor through $M'$ is

763: obstructed. Therefore, we may replace $M$ by $M'$ and hence are

764: reduced to the case when $\phi = 0$.

765: 

766: \

767: 

768: \noindent \emph{Valuative criteria for properness when $\mc F$ has

769: proper support}

770: 

771: Let $R$ be a discrete valuation ring with field of fractions $K$,

772: and let $i:\cX_K\hookrightarrow \cX$ be the inclusion of the generic

773: fiber.  We suppose that we have a flat quotient $\mc

774: F_K\rightarrow \mc G_K$ over the generic fiber which we wish to

775: extend to $\cX$.  For this we take $\mc G$ to be the image of the

776: map $\mc F\rightarrow i_*\mc G_K$.  The image is evidently a

777: coherent sheaf, and has proper support since $\mc F$ has  proper support.  It

778: is flat because it is a subsheaf of a torsion free sheaf, and by

779: definition $\mc G$ induces $\mc G_K$ on the generic fiber.

780: 

781: This completes the proof of theorem \ref{thm-thm1}. \qed

782: 

783: 

784: 

785: 

786: 

787: 

788: 

789: 

790: 

791: 

792: 

793: 

794: 

795: 

796: 

797: \section{Flattening stratifications}~\label{sec-FS}

798: 

799: As a corollary of the representability result in the last section,

800: when the locally finitely-presented sheaf $\mc F$ has proper support

801: over $S$, we can construct the \emph{flattening stratification} of

802: $\mc F$ as an algebraic space.  But in fact this algebraic space is

803: representable and quasi-affine over $S$.  This is a crucial step in

804: the proof of theorem~\ref{thm-thm2}.  Therefore we include a proof of

805: the following fact:

806: 

807: \begin{prop}~\label{prop-qaff}  Suppose $f:Y\rightarrow X$ is a

808:   finite-type, separated, quasi-finite morphism of algebraic spaces.

809:   Then f is quasi-affine, in particular $f$ is representable by

810:   schemes.

811: \end{prop}

812: 

813: \begin{proof}

814:   We need to prove that the natural morphism of $X$-schemes,

815: \begin{equation}

816: \iota:Y\rightarrow \SP_X f_*\OO_Y

817: \end{equation}

818: is an open immersion.  By ~\cite[proposition II.4.18]{Kn}, $f_*\OO_Y$

819: commutes with flat base change on $X$, so the formation of $\iota$

820: commutes with flat base change on $X$.  Moreover one may check that a

821: morphism of $X$-schemes is an open immersion after fpqc base change.

822: Thus, without loss of generality, we may suppose that $X$ is an affine

823: scheme.

824: 

825: \

826: 

827: To prove $\iota$ is an open immersion, it suffices to check that for

828: each point $p\in Y$, the following two conditions hold:

829: \begin{enumerate}

830: \item $\iota$ is \'etale at $p$, and

831: \item there is an open set $p\in U\subset Y$ which is disjoint from

832: the image of $Y\times_{\iota,\iota} Y - \Delta(Y)$ under the

833: projection $\text{pr}_1:Y\times_{\iota,\iota} Y \rightarrow Y$.

834: \end{enumerate}

835: Here $Y\times_{\iota,\iota} Y$ is the fiber product of $Y$ with itself

836: over $\SP_X f_*\OO_Y$ and $\Delta:Y\rightarrow Y\times_{\iota,\iota}

837: Y$ is the diagonal morphism.  It is clear that if both $(1)$ and $(2)$

838: hold for each point $p\in Y$, then $\iota$ is an \'etale monomorphism,

839: and therefore $\iota$ is an open immersion.

840: 

841: \

842: 

843: The claim is that for given $p\in Y$, we may check $(1)$ and $(2)$

844: after passing to an \'etale neighborhood of $q=f(p)\in X$, i.e. if

845: $X'\rightarrow X$ is an \'etale morphism, $q'\in X'$ is a point lying

846: over $q$, and if $p'\in Y':= X'\times_X Y$ is the point lying over

847: $q'$ and $p$, then it suffices to check $(1)$ and $(2)$ for $p'$.  We

848: have mentioned that the natural morphism of $X'$-schemes

849: \begin{equation}

850: \iota':Y'\rightarrow \SP_{X'} (f')_*\OO_{Y'}

851: \end{equation}

852: is the base-change of $\iota$.  The property of being \'etale at a

853: point can be checked after \'etale (and even flat) base-change, so

854: if $(1)$ holds for $p'$ then $(1)$ holds for $p$.  Suppose $(2)$

855: holds for $p'$ and let $U'$ be an open set $p'\in U'\subset Y'$ as

856: in $(2)$. Let $p\in U\subset Y$ be the open image of $U'$ under

857: $g:Y'\rightarrow Y$.  We have the equality

858: \begin{eqnarray}

859: g(U') \cap \text{pr}_1\lt(Y\times_{\iota,\iota} Y - \Delta(Y)\rt) =

860: g\lt( U' \cap g^{-1} \text{pr}_1 \lt( Y\times_{\iota,\iota} Y -

861: \Delta(Y)\rt)\rt) = \\

862: g\lt( U' \cap \text{pr}'_1 \lt(

863: Y'\times_{\iota',\iota'} Y' - \Delta'(Y')\rt) \rt).

864: \end{eqnarray}

865: But $U'\cap \text{pr}'_1 \lt( Y'\times_{\iota',\iota'} Y' -

866: \Delta'(Y') \rt) = \emptyset$ by assumption.  Thus $U=g(U')$ satisfies

867: condition $(2)$ for $p\in Y$.  So it suffices to check $(1)$ and $(2)$

868: for $p'\in Y'$.

869: 

870: \

871: 

872: Now let $Z\rightarrow Y$ be an \'etale morphism of a scheme $Z$ to $Y$

873: and $r\in Z$ a point mapping to $p$.  Then $Z\rightarrow X$ is

874: finite-type, separated and quasi-finite.  By ~\cite[proposition

875: 2.3.8(a)]{Ner}, we may find an \'etale morphism $X'\rightarrow X$ with

876: $X'$ an affine scheme, and a point $q'\in X'$ mapping to $q=f(p)$ such

877: that if $r'\in Z'$ is the point lying over $q'$ and $r$, and if $Z'_0$

878: is the connected component of $Z'$ containing $r'$, then

879: $Z'_0\rightarrow X'$ is a finite morphism of affine schemes.  Let

880: $Y'_0\subset Y'$ be the connected component of $Y'$ containing the

881: image of $Z'_0$.  Then $Z'_0\rightarrow Y'_0$ is \'etale with dense

882: image, but it is also proper.  Therefore $Z'_0\rightarrow Y'_0$ is

883: finite,\'etale and surjective.  So, by Knudsen's version of

884: Chevalley's theorem~\cite[theorem III.4.1]{Kn}, $Y'_0$ is an affine

885: scheme.

886: 

887: \

888: 

889: Let $Y'_r$ denote the union of  connected components $Y'-Y'_0$.

890: Then we have a decomposition of $\text{Spec}_{X'}(f')_* \OO_{Y'}$

891: into  components, $\text{Spec}_{X'}(f'_0)_* \OO_{Y'_0}$ union

892: $\text{Spec}_{X'}(f'_r)_* \OO_{Y'_r}$ and a decomposition of

893: $\iota'$ as the ``disjoint union'' of the two morphisms:

894: \begin{eqnarray}

895: \iota'_0:Y'_0 \rightarrow \SP_{X'}(f'_0)_* \OO_{Y'_0}, \\

896: \iota'_r:Y'_r \rightarrow \SP_{X'}(f'_r)_* \OO_{Y'_r}.

897: \end{eqnarray}

898: And $\iota'$ is an isomorphism.  Thus $\iota'$ is \'etale at $p'$.

899: And defining $U'=Y'_0$, we see that $U'$ satisfies the condition $(2)$

900: for $p'\in Y'$.  So $(1)$ and $(2)$ hold for $p'$ and thus $(1)$ and

901: $(2)$ hold for $p\in Y$.  Since this holds for every $p\in Y$, we

902: conclude that $\iota$ is an open immersion, i.e. $f:Y\rightarrow X$ is

903: quasi-affine.

904: \end{proof}

905: 

906: \textbf{Remark:} We could not find precisely this statement in

907: ~\cite{Kn}, although it follows easily from results proved there.  If

908: one further assumes that $f$ is finitely-presented, then this result

909: also follows easily from ~\cite[th\'eor\`eme 16.5]{LM-B}.

910: 

911: \

912: 

913: Now suppose $S$ is an algebraic space, $f:\cX\rightarrow S$ is a

914: 1-morphism from a Deligne-Mumford stack to $S$ which is separated and

915: locally finitely-presented.  Let $\cF$ be a locally finitely-presented

916: $\OO_\cX$-module.  Consider the functor

917: \begin{equation}

918: \Sigma:\lt(S-\text{schemes}\rt)^{\text opp}\rightarrow \text{Sets},

919: \end{equation}

920: which to any morphism $g:T\rightarrow S$ associates $\lt\{*\rt\}$ if

921: the pullback of $\cF$ to $T\times_S \cX$ is flat over $T$, and which

922: associates $\emptyset$ otherwise.  Given a morphism of $S$-schemes,

923: $h:T_1\rightarrow T_2$, the morphism $\Sigma(h)$ is defined to be the

924: unique map $\Sigma(T_2)\rightarrow \Sigma(T_1)$.  For this to make

925: sense, we must check that if $\Sigma(T_2)$ is nonempty, then so is

926: $\Sigma(T_1)$.  But this is clear, if the pullback of $\cF$ to

927: $T_2\times_S \cX$ is flat over $T_2$, then the pullback of this sheaf

928: over $T_1\times_{T_2} \lt(T_2 \times_S \cX\rt)$ is flat over $T_1$.

929: And this pullback is isomorphic to the pullback of $\cF$ to

930: $T_1\times_S \cX$.  So $\Sigma(T_1)$ is nonempty.

931: 

932: \

933: 

934: \begin{thm}~\label{thm-fs}

935:   Let $f:\cX\rightarrow S$, $\cF$, and $\Sigma$ be as above.

936: \begin{enumerate}

937: \item $\Sigma$ is an fpqc sheaf which is limit preserving and

938:   $\Sigma\rightarrow S$ is a monomorphism.

939: \item If $\cF$ has proper support over $S$, then $\Sigma$ is an

940:   algebraic space and $\Sigma\rightarrow S$ is a surjective,

941:   finitely-presented monomorphism.  In particular, $\Sigma\rightarrow

942:   S$ is quasi-affine.

943: \end{enumerate}

944: \end{thm}

945: 

946: \begin{proof}

947:   It is immediate that $\Sigma\rightarrow S$ is a monomorphism.  Since

948:   one may check that a quasi-coherent sheaf on $T\times_S \cX$ is flat

949:   over $T$ after performing an fpqc base change of $T$, it follows

950:   that $\Sigma$ is an fpqc sheaf.  The fact that $\Sigma$ is

951:   limit-preserving follows from ~\cite[th\'eor\`eme IV.11.2.6]{EGA}.

952:   So (1) is proved.

953: 

954: \

955: 

956: To prove (2), first notice that we may use (1) to reduce to the case

957: that $S=\SP A$ is a Noetherian affine scheme, and $\cX$ is the support

958: of $\cF$ which is a proper, finitely-presented Deligne-Mumford stack

959: over $S$.  By theorem~\ref{thm-thm1} we know the Quot functor of $\cF$

960: is represented by an algebraic space $Q\rightarrow S$ which is

961: separated and locally finitely-presented.  Denote by

962: \begin{equation}

963: \begin{CD}

964:   0 @>>> {\mc K} @>>> \text{pr}_2^* \cF @>>> {\mc G} @>>> 0

965: \end{CD}

966: \end{equation}

967: the universal short exact sequence on $Q\times_S \cX$.  Since

968: $\text{pr}_2^* \cF$ has proper support over $Q$,  ${\mc K}$ also

969: has proper support over $Q$.  Define $U$ to be the complement in $Q$

970: of the image of the support of ${\mc K}$.  The restriction of the

971: universal short exact sequence over $U$ is simply the identity

972: morphism of $\text{pr}_2^* \cF$ to itself, i.e. the pullback of

973: $\cF$ to $U$ is flat over $U$.  So we have an induced morphism

974: $U\rightarrow \Sigma$.  Conversely there is an obvious morphism

975: $\Sigma\rightarrow Q$ which factors through $U$, and we see that

976: $U\rightarrow \Sigma$ is a natural isomorphism of algebraic spaces.

977: So $\Sigma$ is an algebraic space and $\Sigma\rightarrow S$ is a locally

978: finitely-presented monomorphism.

979: 

980: \

981: 

982: Any module over a field is flat over that field, therefore for each

983: field $K$ and each morphism $\SP K \rightarrow S$, this morphism

984: factors through $\Sigma\rightarrow S$.  So $\Sigma\rightarrow S$ is

985: surjective.  The claim is that for any

986: surjective, locally finitely-presented monomorphism $h:Y\rightarrow S$

987: of an algebraic space to a Noetherian affine scheme, $Y$ is

988: quasi-compact.

989: 

990: \

991: 

992: We will prove this claim by induction on the dimension of $S$.

993: If $S=\SP K$ for some field $K$, it is obvious.  If

994: $Y^{\text red}$ is quasi-compact, the same is true of $Y$, so we may

995: reduce to the case that $S$ is reduced.  A finite union of

996: quasi-compact sets is quasi-compact, so we may reduce to the case that

997: $S$ is integral.  Now suppose the result is proved for all schemes $S$

998: of dimension at most $n$ and suppose $S$ is an integral scheme of

999: dimension $n+1$.  Let $\SP K$ be the generic point of $S$ and let

1000: $U\subset Y$ be the open set where $U\rightarrow S$ is \'etale.  Then

1001: $\SP K$ factors through $Y$, and in fact it is contained in $U$.  Thus

1002: $U\rightarrow S$ is an \'etale monomorphism, i.e. an open immersion

1003: which has dense image.  Let the complement of $U$

1004: in $S$ be $C$ (with reduced scheme structure)

1005: and let the preimage of $C$ be $Z$.  Then $Z\rightarrow

1006: C$ is again a surjective, locally finitely-presented monomorphism, and

1007: $C$ has dimension at most $n$.  So by the induction assumption, $Z$ is

1008: quasi-compact.  Since $U$ is an open subset of a Noetherian scheme, it

1009: is quasi-compact.  Thus $Y=U\cup Z$ is quasi-compact and the claim is

1010: proved by induction.

1011: 

1012: \

1013: 

1014: By the last paragraph, we conclude that $\Sigma\rightarrow S$ is a

1015: surjective, finitely-presented monomorphism of algebraic spaces.  So

1016: by proposition~\ref{prop-qaff}, we conclude that $\Sigma \rightarrow

1017: S$ is quasi-affine.

1018: \end{proof}

1019: 

1020: \textbf{Remarks:} (1)  If the support of $\cF$ is a scheme and the

1021: morphism to $S$ is projective, then it follows from ~\cite[theorem,

1022: p.55]{Mum} that $\Sigma\rightarrow S$ is a disjoint union of locally

1023: closed immersions.  While one can find examples of surjective,

1024: finitely-presented monomorphisms $Y\rightarrow S$ not a disjoint union

1025: of locally closed immersions, we conjecture that $\Sigma\rightarrow S$

1026: is a disjoint union of locally closed immersions whenever $\cF$

1027: has proper support over $S$.

1028: 

1029: \

1030: 

1031: (2)  Again in the case that the support of $\cF$ is projective

1032:      over $S$, the methods in ~\cite[section 8]{Mum} provide a

1033:      \emph{global construction} of the flattening stratification

1034:      $\Sigma\rightarrow S$ along with a partition labelled by the

1035:      Hilbert polynomial of the fibers of $\cF$.  In the case that

1036:      $\cX$ is a tame, global quotient with projective coarse

1037:      moduli space, we believe that there is again a \emph{global

1038:      construction} of the flattening stratification $\Sigma\rightarrow

1039:      S$ along with a partition labelled by the Hilbert polynomial (as

1040:      defined in section~\ref{sec-HP}).  However we don't know what

1041:      this global construction is, and the proof of the existence of

1042:      $\Sigma\rightarrow S$ given above is the one truly

1043:      non-constructive step in the proof of theorm~\ref{thm-thm2}.

1044: 

1045: \section{Hilbert polynomials}~\label{sec-HP}

1046: 

1047: 

1048: Recall that a Deligne-Mumford stack $\cX$ is \emph{tame} if for

1049: each algebraically-closed field $k$ and each 1-morphism $\zeta:\SP

1050: k\rightarrow \cX$, the $k$-valued points of

1051: \begin{equation}

1052: G_{\zeta} := \SP k \times _{(\zeta,\zeta),\mc X\times \mc

1053: X,\Delta} {\mc

1054:   X}

1055: \end{equation}

1056: form a group of order  prime to char$(k)$.   We remind the reader

1057: of some facts about tame Deligne-Mumford stacks.

1058: 

1059: \

1060: 

1061: \begin{lem}~\label{lem-tame}

1062: Let $\cX$ be a tame Deligne-Mumford stack, $\pi:{\mc

1063:   X}\rightarrow X$ its coarse moduli space.

1064: \begin{enumerate}

1065: \item The additive functor $\pi_*$ from the category of

1066:   $\OO_{\cX}$-modules

1067:   to the category of $\OO_X$-modules

1068:   maps quasi-coherent sheaves to

1069:   quasi-coherent sheaves and maps locally finitely-presented sheaves

1070:   to locally finitely-presented sheaves.

1071: \item The additive functor $\pi_*$ is exact, in particular

1072:   $R^i\pi_*\cF=0$ for $i>0$ and $\cF$ any quasi-coherent $\OO_{\cX}$-module.

1073: \item Suppose $g:X\rightarrow S$ is a morphism of algebraic spaces and

1074:   suppose $\cF$ is a quasi-coherent sheaf on $\cX$ which is

1075:   flat over $S$.  Then also $\pi_* \cF$ is flat over $S$.

1076: \end{enumerate}

1077: \end{lem}

1078: 

1079: \begin{proof}

1080:   (1) and (2) form ~\cite[lemma 2.3.4]{AV}.  And (3) follows by the

1081:   same local analysis in the proof of ~\cite[lemma 2.3.4]{AV} since

1082:   the invariant submodule $M^\Gamma$ (in the notation of ~\cite{AV})

1083:   of an $S$-flat module $M$ is a direct summand, and so it is also

1084:   $S$-flat.

1085: \end{proof}

1086: 

1087: 

1088: Suppose $k$ is a field and $\cX$ is a separated, locally

1089: finitely-presented tame Deligne-Mumford stack over $\SP{k}$, and

1090: let $\pi :\cX\rightarrow X$ be its coarse moduli space. Suppose

1091: that $\cF$ is a coherent $\OO_\cX$-module with proper support.

1092: Then by lemma \ref{lem-tame}, $H^i_{\text{\'et}}(\cX, \cF) =

1093: H^i_{\text{\'et}}(X, \pi _*\cF)$ for all $i$, and hence these

1094: groups are zero for $i$ bigger than the dimension of $\cX$.

1095: Therefore the sum

1096: \begin{equation}

1097: \chi(\cX,\cF) = \sum_{i=0}^\infty (-1)^i \text{dim}_k

1098: H^i_{\text{\'et}}(\cX,\cF)

1099: \end{equation}

1100: is finite.  For each locally free sheaf $E$ of finite rank on

1101: $\cX$, we also have that $E\otimes_{\OO_\cX} \cF$ is again

1102: coherent with proper support. Since $E\mapsto \chi({\mc

1103: X},E\otimes_{\OO_\cX} \cF)$ is additive in short exact sequences,

1104: we have a well-defined group homomorphism

1105: \begin{equation}

1106: P_\cF:K^0(\cX)\rightarrow \ZZ, [E]\mapsto \chi({\mc

1107:   X},E\otimes_{\OO_\cX} \cF).

1108: \end{equation}

1109: 

1110: \

1111: 

1112: \begin{defn}~\label{defn-HP}

1113:   Given a homomorphism of Abelian groups $i:A\rightarrow

1114:   K^0(\cX)$, define the \emph{A-Hilbert polynomial} of $\cF$,

1115:   $P_{A,\cF}$, to be the function $P_{A,\cF}=P_\cF\circ i$.

1116: \end{defn}

1117: 

1118: \textbf{Remarks:} (1) If ${\mc L}$ is an invertible sheaf on $\cX$,

1119: $A=\ZZ[x,x^{-1}]$ and $i:A\rightarrow K^0(\cX)$ is the group homomorphism

1120: such that $i(x^d)=[{\mc L}^d]$, then the $A$-Hilbert polynomial of

1121: $\cF$,

1122: $P_{A,\cF}$ is the usual

1123: Hilbert polynomial of $\cF$ with respect to ${\mc L}$.  We will

1124: need to consider cases where $i:A\rightarrow K^0({\mc F})$ cannot be

1125: reduced to this form, which is why our notion of Hilbert polynomial is

1126: so general.

1127: 

1128: \

1129: 

1130: (2) The most instructive example, from our point of view, is when

1131:     $\cX = BG$ for some finite, \'etale $k$-group scheme $G$.

1132:     Then $K^0(BG)$ is naturally isomorphic to the Grothendieck group of the

1133:     category of finite $k[G]$-modules, i.e. the representation ring of

1134:     $k[G]$.  And the Hilbert polynomial $P_\cF$ determines the

1135:     image $[\cF]$ of $\cF$ in $K^0(BG)$.

1136: 

1137: \

1138: 

1139: 

1140: Now suppose that $f:\cX\rightarrow S$ is a 1-morphism from a

1141: tame Deligne-Mumford stack to a connected algebraic space such that $f$ is

1142: separated and locally finitely-presented,  Define $A=K^0(\cX)$ and

1143: for each field $k$ and each morphism $g:\SP k\rightarrow S$, define

1144: $i_g:A\rightarrow K^0(\SP k\times_S \cX)$ to be the pullback map

1145: $K^0(\text{pr}_2)$.

1146: 

1147: \

1148: 

1149: \begin{lem}~\label{lem-HPconst}

1150: Suppose that $\cF$ is a locally finitely-presented quasi-coherent

1151: sheaf on $\cX$ which is flat over $S$ and which has proper support

1152: over $S$.  Then there exists a function $P:A\rightarrow \ZZ$ such that

1153: for all $g:\SP k\rightarrow S$, $P_{A,g^*\cF} = P$.

1154: \end{lem}

1155: 

1156: \begin{proof}

1157: We need to show that for each locally free sheaf $\cE$ on ${\mc

1158:   X}$, the function

1159: \begin{equation}

1160: (g:\SP k\rightarrow S) \mapsto \chi\lt(g^*\lt(\cE\otimes_{\OO_{\mc

1161:     X}}\cF\rt)\rt)

1162: \end{equation}

1163: is constant.  Since $R^i\pi_*$ vanishes on all quasi-coherent modules

1164: for $i>0$, we have

1165: \begin{equation}

1166: \chi\lt(g^*\lt(\cE\otimes_{\OO_\cX}\cF\rt)\rt) =

1167: \chi\lt(g^*\pi_*\lt( \cE\otimes_{\OO_\cX}\cF\rt)\rt).

1168: \end{equation}

1169: And by lemma~\ref{lem-tame} (3), we know that $\pi_*({\mc

1170:   E}\otimes_{\OO_\cX}\cF)$ is an $S$-flat, locally

1171:   finitely-presented sheaf with proper support over $S$.  So by

1172:   ~\cite[corollary, p. 50]{Mum}, we conclude that there is some

1173:   $P([\cE])\in \ZZ$ such that for all $g:\SP k\rightarrow S$, we

1174:   have

1175: \begin{equation}

1176: \chi\lt(g^*\pi_*\lt( \cE\otimes_{\OO_\cX}\cF\rt)\rt) =

1177:   P([\cE]).

1178: \end{equation}

1179: \end{proof}

1180: 

1181: Fix an additive homomorphism $P:K^0(\cX)\rightarrow \ZZ$ and

1182: define $Q^P = Q^P(\cF/\cX/S)$ to be the subfunctor of $Q({\mc

1183:   F}/\cX/S)$ such that for each $S$-scheme, $g:Z\rightarrow S$ we

1184: have $Q^P(g:Z\rightarrow S)$ is the set of quotients $[g^*{\mc

1185:   F}\rightarrow {\mc G}]\in Q(g:Z\rightarrow S]$ such that for each

1186: field $k$ and each morphism $h:\SP k \rightarrow Z$, we have

1187: $P_{A,h^*{\mc G}}=P$.  By lemma~\ref{lem-HPconst}, we see that $Q^P$ is

1188: an open and closed subfunctor (possibly empty) of $Q$ and that $Q$ is

1189: the disjoint union of $Q^P$ as $P$ ranges over all $P$.

1190: 

1191: \

1192: 

1193: Observe that theorem~\ref{thm-thm2} is implied by the following refinement.

1194: 

1195: \

1196: 

1197: \begin{thm}~\label{thm-thm3}

1198:   Suppose that $S$ is an affine scheme.  Suppose that $\cX$

1199:   is a tame Deligne-Mumford stack which is separated and

1200:   finitely-presented over $S$, whose coarse moduli space is a

1201:   quasi-projective $S$-scheme (resp. projective $S$-scheme), and which

1202:   is a global quotient over $S$.

1203:   Then for each locally

1204:   finitely-presented quasi-coherent sheaf $\cF$ on $\cX$ and

1205:   for each homomorphism $P:K^0(\cX)\rightarrow \ZZ$, the functor

1206:   $Q^P(\cF/\cX/S)$ is represented by an algebraic space $Q^P$ which

1207:   admits a factorization $Q^P\rightarrow Q'\rightarrow S$ where $Q'$

1208:   is projective over $S$ and $Q^P\rightarrow Q'$ is a

1209:   finitely-presented, quasi-finite monomorphism.

1210:   If $\cF$ has proper support over $S$, then

1211:   $Q^P\rightarrow Q'$ is a finitely-presented closed immersion.

1212: \end{thm}

1213: 

1214: \textbf{Remark:}  In particular, if $S$ is affine, then

1215: $Q^P(\cF/\cX/S)$ is represented by a quasi-projective $S$-scheme

1216: (which is projective if the support of $\cF$ is proper over $S$).

1217: 

1218: \

1219: 

1220: We will prove theorem~\ref{thm-thm3} in sections \ref{gensheavsec} and

1221: \ref{nattransec}.

1222: 

1223: \section{Generating sheaves}\label{gensheavsec}

1224: Let $\cX$ be a tame Deligne-Mumford stack with coarse

1225: moduli space

1226: $\pi:\cX\rightarrow X$.  For each locally free sheaf $\cE$ on

1227: $\cX$, define additive functors

1228: \begin{eqnarray}

1229: F_\cE:\text{Quasi-coherent}_\cX \rightarrow

1230: \text{Quasi-coherent}_X, \\

1231: G_\cE:\text{Quasi-coherent}_\cX \rightarrow

1232: \text{Quasi-coherent}_\cX

1233: \end{eqnarray}

1234: by the formulas

1235: \begin{equation}

1236: F_\cE(\cF) = \pi_*\textit{Hom}_{\OO_{\cX}}(\cE,\cF), \ \ \ G_\cE =

1237: \pi^*\lt(F_\cE(\cF)\rt)\otimes_{\OO_{\cX}}\cE

1238: \end{equation}

1239: where $\mc F$ is a quasi-coherent sheaf on $\mc X$.

1240: 

1241: 

1242: \

1243: 

1244: By lemma~\ref{lem-tame}, $F_\cE$ is an exact functor which

1245: preserves flatness and preserves the property of being locally

1246: finitely-presented.  And $G_\cE$ is a right-exact functor which

1247: preserves the property of being locally finitely-presented.  Moreover

1248: there is a natural transformation $\theta_\cE:G_\cE\Rightarrow

1249: \text{Id}$ where $\text{Id}$ is the identity functor on the category

1250: of quasi-coherent sheaves on $\cX$ and for a quasi-coherent sheaf

1251: $\cF$ on $\cX$, the morphism

1252: \begin{equation}

1253: \theta_{\cE}(\cF):\pi^* \lt( \pi_* \textit{Hom}_{\OO_{\cX}} (\cE,\cF)

1254: \rt) \otimes_{\OO_{\cX}} \cE \rightarrow \cF

1255: \end{equation}

1256: is the left adjoint to the natural morphism

1257: \begin{equation}

1258: \pi^*\pi_*\textit{Hom}_{\OO_{\cX}}(\cE,\cF)\rightarrow

1259: \textit{Hom}_{\OO_{\cX}}(\cE,\cF),

1260: \end{equation}

1261: which is itself the left adjoint of the identity morphism

1262: \begin{equation}

1263: \pi_* \textit{Hom}_{\OO_{\cX}} (\cE,\cF) \rightarrow \pi_*

1264: \textit{Hom}_{\OO_{\cX}} (\cE,\cF).

1265: \end{equation}

1266: 

1267: \textbf{Remark:} Since $\cX$ is a tame stack, $\pi_*$ from

1268: quasi-coherent sheaves on $\cX$ to quasi-coherent sheaves on $X$

1269: is compatible with \emph{arbitrary} base changes $T\rightarrow X$

1270: (see for example the proof of ~\cite[2.3.3]{AV}).  It follows that

1271: also the functors $F$, $G$ and the natural transformation $\theta$

1272: are compatible with \emph{arbitrary} base changes $T\rightarrow

1273: X$.

1274: 

1275: 

1276: \begin{defn}~\label{defn-gen}  With notation as above, $\cE$ is a

1277:   \emph{generator } for $\cF$ if $\theta_{\cE}(\cF)$ is surjective.

1278: \end{defn}

1279: 

1280: \textbf{Example:}  Suppose that $G$ is a finite group and $\cX =

1281: BG\times_{\SP \ZZ} \SP k$.  Then the quasi-coherent

1282: $\OO_{\cX}$-modules correspond to (left-)modules over $k[G]$.  Let $\cE$ be

1283: the locally free sheaf corresponding to the left regular

1284: representation of $G$ on $k[G]$.  Then for every quasi-coherent sheaf

1285: $\cF$, $\cE$ is a generator for $\cF$.  For that matter, if $M$ is any

1286: $k[G]$-module which contains every irreducible representation of $G$

1287: as a submodule, then the locally free sheaf corresponding to $M$ is a

1288: generator for every $\cF$.

1289: 

1290: \

1291: 

1292: The goal of this section is to prove that the previous example is

1293: typical for separated, tame, Deligne-Mumford stacks which are global

1294: quotients.  As suggested by the referee, we first prove the following

1295: \emph{pointwise} condition for a locally free sheaf $\cE$ to be a generator.

1296: 

1297: \begin{thm}~\label{thm-genpt}

1298: Suppose that $\cX$ is a tame Deligne-Mumford stack as above and

1299: $\cE$ is a locally free sheaf on $\cX$. For each

1300: algebraically-closed field $k$ and each $1$-morphism $\zeta:\SP k

1301: \rightarrow \cX$ with stabilizer group $G_k$ (c.f.

1302: section~\ref{sec-HP}), let $\widetilde{\zeta}:BG_\zeta \rightarrow

1303: \cX$ be the natural map.

1304: 

1305: Suppose that for some $1$-morphism $\zeta:\SP k \rightarrow \cX$,

1306: the locally free sheaf $\cE_\zeta :=

1307: \lt(\widetilde{\zeta}\rt)^*\cE$ is a generator for the left

1308: regular representation of $G_\zeta$, i.e. considered as a

1309: finite-dimensional $k[G_\zeta]$-module, $\cE_\zeta$ contains every

1310: irreducible representation of $G_\zeta$ as a submodule. Then there

1311: exists a Zariski open subset $U\subset X$ containing the image of

1312: $\zeta(\SP k)$ such that the restriction of $\cE$ to $\cX \times_X

1313: U$ is a generator for every quasi-coherent sheaf on $\cX\times_X

1314: U$. In particular, if $\lt(\widetilde{\zeta} \rt)^*\cE$ is a

1315: generator for the left regular representation of $G_\zeta$ for

1316: every geometric point $\zeta:\SP k \rightarrow \cX$, then $\cE$ is

1317: a generator for every quasi-coherent sheaf on $\cX$.

1318: \end{thm}

1319: 

1320: \textbf{Remarks:} (1) The converse to theorem~\ref{thm-genpt} is

1321: obvious.  Suppose that $\cE$ is a locally free sheaf on

1322: $\cX$ and suppose that $\widetilde{\zeta}:BG_\zeta

1323: \rightarrow \cX$ is as above.  Let $\cF$ be the push-forward by

1324: $\widetilde{\zeta}$ of the sheaf on $BG_\zeta$ corresponding to the

1325: left regular representation of $G_\zeta$.  Then $\cE$ is a generator

1326: for $\cF$ iff $\lt(\widetilde{\zeta}\rt)^*\cE$ is a generator for the

1327: left regular representation of $G_\zeta$.

1328: 

1329: \

1330: 

1331: (2) Suppose $\cX$ is a quotient stack of the form $\cX = [Y/GL_n]$.

1332:     Then for each geometric point $\zeta:\SP k \rightarrow \cX$ the

1333:     stabilizer group $G_\zeta$ is a closed subgroup scheme of

1334:     $GL_{n,k}$ compatibly with the composite map

1335: \begin{equation}

1336: BG_\zeta \xrightarrow{\widetilde{\zeta}} \cX \rightarrow BGL_n.

1337: \end{equation}

1338: Of course every representation of $GL_n$ over $k$ induces a

1339: representation of $G_\zeta$, i.e. every quasi-coherent sheaf on

1340: $BGL_{n,k}$ pulls back to a quasi-coherent sheaf on $BG_\zeta$.

1341: Moreover, it is well known that there exists a

1342: $GL_{n,\ZZ}$-submodule

1343: \begin{equation} W\subset

1344: H^0(GL_{n,\ZZ},\OO_{GL_{n,\ZZ}})\end{equation} which is a

1345: finitely-generated free $\ZZ$-module such that the composite map

1346: \begin{equation}

1347: W \otimes_\ZZ k \rightarrow H^0(GL_{n,k},\OO_{GL_{n,k}}) \rightarrow

1348: H^0(G_\zeta, \OO_{G_\zeta})

1349: \end{equation}

1350: is surjective.  So for each point $\zeta:\SP k \rightarrow \cX$, there

1351: is a $GL_{n,\ZZ}$ representation $W$ which is a finitely-generated

1352: free $\ZZ$-module such that the pullback $\cE$ under $\cX

1353: \rightarrow BGL_{n,\ZZ}$ of the locally free sheaf associated to $W$

1354: satisfies the condition in theorem~\ref{thm-genpt}.  In order to apply

1355: theorem~\ref{thm-genpt}, we need to find a single representation $W$

1356: of $GL_{n,\ZZ}$ which works for every geometric point $\zeta:\SP k

1357: \rightarrow \cX$.

1358: 

1359: \

1360: 

1361: (3) The stack $BG_{\zeta }$ is naturally isomorphic to the fiber

1362: product

1363: \begin{equation}

1364: \cX \times _{X, \pi \circ \zeta }\SP k

1365: \end{equation}

1366: and the map $\widetilde {\zeta }$ is simply the projection onto

1367: the first factor.

1368: 

1369: \

1370: 

1371: We will prove theorem~\ref{thm-genpt} after proving a proposition

1372: about quotient stacks by a finite group.

1373: Suppose that $\cE$ is a locally free sheaf on

1374: $\cX$ which satisfies the hypotheses of theorem~\ref{thm-genpt} and

1375: suppose that $\cF$ is a quasi-coherent sheaf on $\cX$.  To show that

1376: $\cE$ is a generator for $\cF$, it suffices to prove this after

1377: performing a faithfully-flat base change of the coarse moduli space

1378: $X$.  By ~\cite[lemma 2.2.3]{AV}, there is an \'etale cover $(X_\alpha

1379: \rightarrow X)$ such that for each $\alpha$ the fiber product $\cX

1380: \times_X X_\alpha$ is of the form $[Y/G]$ where $Y$ is a finite scheme

1381: over $X_\alpha$ and $G$ is a finite group which acts on $Y$ via

1382: $X_\alpha$-morphisms.  Moreover, since $\cX$ is tame, we may assume

1383: that the order of $G$ is prime to the characteristic of every residue

1384: field of $X_\alpha$.  So the first step in proving

1385: theorem~\ref{thm-genpt} is the case that $\cX = [Y/G]$ as above.

1386: 

1387: Now there is a map $f: [Y/G] \rightarrow BG$.  Let $E$ denote the

1388: locally free sheaf on $BG$ corresponding to the left regular

1389: representation of $G$.  If $\cE$ is a locally free sheaf on $\cX$

1390: which is a generator for every quasi-coherent sheaf, then in

1391: particular it is a generator for $f^*E$.  The following proposition

1392: shows that this is a sufficient condition for $\cE$ to be a generator

1393: for every quasi-coherent sheaf.

1394: 

1395: \begin{prop}~\label{prop-genBG}

1396: Suppose that $\cX$ is a tame, separated Deligne-Mumford stack of the form

1397: $\cX=[Y/G],$ as above.  Let $f:\cX \rightarrow BG$ and $E$ be as above.

1398: Then $f^*E$ is a

1399: generator for every quasi-coherent $\OO_{\cX}$-module $\cF$.

1400: \end{prop}

1401: 

1402: \begin{proof}

1403: Let $g:Y\rightarrow \cX$ denote the quotient $1$-morphism, and let

1404: $p:\cX\rightarrow X$ denote the map to the coarse moduli space of

1405: $\cX$.

1406: Observe

1407: that $f^* E$ is simply $g_*\OO_Y$.

1408: The

1409: composition $p\circ g:Y\rightarrow X$ is a finite, surjective morphism

1410: of algebraic spaces.  In particular, it is affine and for each

1411: quasi-coherent $\OO_Y$-module $\cG$, the induced morphism

1412: \begin{equation}

1413: \alpha:(p\circ g)^*(p\circ g)_* \cG \rightarrow \cG,

1414: \end{equation}

1415: is surjective.  By adjointness of $p_*$ and $p^*$, there is an induced

1416: morphism

1417: \begin{equation}

1418: p^* (p\circ g)_* \cG \rightarrow g_* \cG,

1419: \end{equation}

1420: in fact this is precisely $\theta_{\OO_{\cX}}(g_*\cG)$.

1421: Since $g_*\cG$ is a module over $g_*\OO_Y$, we get an induced

1422: morphism of $\OO_{\cX}$-modules

1423: \begin{equation}

1424: \phi:g_*\OO_Y \otimes_{\OO_{\cX}} p^*(p\circ g)_* \cG \rightarrow g_*\cG.

1425: \end{equation}

1426: The claim is that $\phi$ is surjective; let us assume this for a

1427: moment.  The canonical injection $g^\#:\OO_{\cX}\rightarrow

1428: g_*\OO_Y$ induces a morphism of $\OO_{\cX}$-modules

1429: \begin{equation}

1430: \psi:g_*\OO_Y \otimes_{\OO_{\cX}}p^*

1431: p_*\textit{Hom}_{\OO_{\cX}}(g_*\OO_Y,g_*\cG) \rightarrow

1432: g_*\OO_Y \otimes_{\OO_{\cX}}p^*

1433: p_*\textit{Hom}_{\OO_{\cX}}(\OO_{\cX},g_*\cG).

1434: \end{equation}

1435: Since $g$ is representable, finite and flat, there is a surjective

1436: trace map

1437: \begin{equation}

1438: t:g_*\OO_Y\rightarrow \OO_{\cX},

1439: \end{equation}

1440: which splits the injection $g^\#$.

1441: Therefore $\psi$ is surjective.  And

1442: $\theta_{\cE}(g_*\cG)$ is the composite $\phi\circ\psi$.  Since $\psi$

1443: and $\phi$ are both surjective, we conclude that

1444: $\theta_{\cE}(g_*\cG)$ is surjective.  So to prove that

1445: $\theta_{\cE}(g_*\cG)$ is surjective, it suffice to prove the claim

1446: that $\phi$ is surjective.

1447: 

1448: \

1449: 

1450: To see that $\phi$ is surjective, we apply $g^*$; since $g$ is

1451: faithfully flat, we may check surjectivity after base-change by $g$.

1452: Since $g$ is affine, the canonical morphism

1453: \begin{equation}

1454:  (g^* g_* \OO_Y)\otimes_{\OO_Y}\cG\rightarrow g^*g_*\cG

1455: \end{equation}

1456: is an isomorphism.

1457: Similarly, the canonical morphism

1458: \begin{equation}

1459: (g^* g_* \OO_Y)\otimes_{\OO_Y} (p\circ g)^*(p\circ g)_*\cG\rightarrow

1460: g^*\lt(g_*\OO_Y \otimes_{\OO_{\cX}} p^*(p\circ g)_* \cG\rt)

1461: \end{equation}

1462: is also an isomorphism.

1463: And we have a commutative diagram

1464: \begin{equation}

1465: \begin{CD}

1466: (g^* g_* \OO_Y)\otimes_{\OO_Y} (p\circ g)^*(p\circ g)_*\cG @>

1467: 1\otimes\alpha >> (g^* g_* \OO_Y)\otimes_{\OO_Y}\cG \\

1468: @VVV  @VVV \\

1469: g^*\lt(g_*\OO_Y \otimes_{\OO_{\cX}} p^*(p\circ g)_* \cG\rt) @>g^*\phi

1470: >> g^*g_*\cG

1471: \end{CD}

1472: \end{equation}

1473: Since $\alpha$ is surjective, so is $1\otimes \alpha$.  Therefore

1474: $g^*\phi$ is surjective, and it follows that $\phi$ is surjective.

1475: This proves the claim, and we conclude that $f^* E$ is a generator for

1476: all sheaves of the form $g_*\cG$ with $\cG$ a quasi-coherent

1477: $\OO_Y$-module.

1478: 

1479: \

1480: 

1481: Let $\cF$ be a quasi-coherent $\OO_{\cX}$-module.

1482: Since $g$ is affine, the

1483: canonical morphism of $\OO_{\cX}$-modules

1484: \begin{equation}

1485: g_*\OO_Y \otimes_{\OO_{\cX}} \cF \rightarrow g_* g^* \cF,

1486: \end{equation}

1487: is an isomorphism.  Therefore

1488: $f^* E$ is a generator for $g_*\OO_Y \otimes_{\OO_{\cX}} \cF$.

1489: And we have a surjective morphism of $\OO_{\cX}$-modules,

1490: \begin{equation}

1491: t\otimes 1: g_*\OO_Y \otimes_{\OO_{\cX}} \cF \rightarrow \cF.

1492: \end{equation}

1493: Since $G_{f^*E}$ is a right-exact functor, we conclude that $f^* E$ is

1494: also a generator for $\cF$.

1495: \end{proof}

1496: 

1497: \begin{cor}~\label{cor-genBG}

1498: Suppose that $\cX$, $f:\cX \rightarrow BG$, and $E$ are as in

1499: proposition~\ref{prop-genBG}.  If $\cE$

1500: is a locally free sheaf on $\cX$ which generates $f^* E$, then $\cE$

1501: generates $\cF$ for every quasi-coherent $\OO_{\cX}$-module $\cF$.

1502: \end{cor}

1503: 

1504: \begin{proof}

1505: Since $p_*$ is exact, for any quasi-coherent $\OO_X$-module $\cG$ and

1506: any quasi-coherent $\OO_{\cX}$-module $\cF$, we have that

1507: \begin{equation}

1508: p_*(p^*\cG\otimes_{\OO_{\cX}} \cF) = \cG \otimes_{\OO_{X}} p_*\cF.

1509: \end{equation}

1510: To see this, it suffices to work locally over affine opens in $X$, so

1511: we may suppose that $\cG$ is a colimit of finitely-presented

1512: $\OO_X$-modules.  Since $p_*$ commutes with colimits, we are reduced

1513: to the case that $\cG$ is finitely-presented.  Since $p^*$ is

1514: right-exact and since $p_*$ is exact, we are reduced to the case that

1515: $\cG=\OO_X$, which is trivial.

1516: 

1517: \

1518: 

1519: Similarly, we conclude that $\theta_{\cE}(p^*\cG \otimes_{\OO_{\cX}}

1520: \cF)$ is just $1_{p^* \cG}\otimes \theta_{\cE}(\cF)$.  In particular,

1521: if $\cE$ generates $\cF$, then $\cE$ generates

1522: $p^*\cG\otimes_{\OO_{\cX}} \cF$.  Therefore,  $\cE$ generates every

1523: sheaf of the

1524: form $p^*\cG\otimes_{\OO_{\cX}} f^* E$.  But by

1525: proposition~\ref{prop-genBG}, we conclude that every quasi-coherent

1526: sheaf $\cF$ is a surjective image of a sheaf of the form

1527: $p^*\cG\otimes_{\OO_{\cX}} f^*E$.  Since $G_{\cE}$ is right-exact, we

1528: conclude that $\cE$ generates $\cF$.

1529: \end{proof}

1530: 

1531: Now we come to the proof of theorem~\ref{thm-genpt}:

1532: 

1533: \begin{proof}

1534: Suppose $\cE$ satisfies the hypotheses of theorem~\ref{thm-genpt}.

1535: By ~\cite[lemma 2.2.3]{AV}, there is an \'etale morphism $g:X'

1536: \rightarrow X$ whose image contains the image of $\zeta(\SP k)$ and

1537: such that $\cX' = \cX \times_X X'$ is of the form $[Y/G]$ for some

1538: finite $X'$-scheme $Y$, for some finite group $G$ and an action of $G$

1539: on $Y$ via $X'$-morphisms.  Let $f:[Y/G] \rightarrow BG$ and $E$ be as

1540: in proposition~\ref{prop-genBG}.  Let $\cE'$ denote the pullback of

1541: $\cE$ to $\cX'$.  Consider the morphism $\theta_{\cE'}(f^* E)$.

1542: Denote the cokernel of this morphism by $\cG$.  Then $\cG$ is a

1543: finitely-generated quasi-coherent sheaf, and therefore has closed

1544: support $Z\subset \cX$.

1545: 

1546: Since $k$ is algebraically-closed and since

1547: $X'\rightarrow X$ is finite, we can find a factorization $\xi:\SP k

1548: \rightarrow X'$ of $\SP k \rightarrow X$.

1549: Now the functors $F$ and $G$

1550: and the natural transformation $\theta$ are compatible with arbitrary

1551: base change on $X$.  Therefore the pullback of $\theta_{\cE'}(f^* E)$

1552: equals $\theta_{\xi^*\cE'}(\xi^* f^* E)$.  Of course we have

1553: \begin{equation}

1554: \cX' \times_{X'} \SP k = \cX \times_X \SP k = BG_\zeta.

1555: \end{equation}

1556: By hypothesis $\xi^* \cE' = \widetilde{\zeta}(\cE)$ generates the

1557: regular representation of $G_\zeta$.  So by

1558: corollary~\ref{cor-genBG}, we conclude that

1559: $\theta_{\xi^*\cE'}(\xi^* f^* E)$ is surjective.  So the image of

1560: $\xi(\SP k)$ is not in the support of $\cG$.  Denoting by

1561: $V\subset X'$ the complement of the image of the support of $\cG$,

1562: we have that $V$ contains $\xi(\SP k)$ and the restriction of

1563: $\cE'$ over $V$ generates $f^* E$.

1564: 

1565: Let $U$ be the image of $V\rightarrow X$ so that the canonical

1566: morphism $h:\cX \times_X V

1567: \rightarrow \cX \times_X U$ is faithfully flat and $U$ contains the

1568: image of $\zeta(\SP k)$.

1569: Let $\cF$ be any quasi-coherent sheaf on $\cX \times_X U$.  Then

1570: $h^*\theta_{\cE}(\cF)$ is $\theta_{h^*\cE}(h^*\cF)$.  By

1571: corollary~\ref{cor-genBG}, we know that $\theta_{h^*\cE}(h^*\cF)$ is

1572: surjective.  Since $h$ is faithfully flat, we conclude that also

1573: $\theta_{\cE}(\cF)$ is surjective.  So the restriction of $\cE$ to

1574: $\cX \times_X U$ is a generator for every quasi-coherent sheaf $\cF$

1575: on $\cX \times_X U$.  This completes the proof of the theorem.

1576: \end{proof}

1577: 

1578: Now we come to the main theorem of this section:

1579: 

1580: \begin{thm}~\label{thm-genthm}

1581: Suppose that $\cX$ is a quasi-compact, tame Deligne-Mumford stack of

1582: the form $[Y/GL_{n,\ZZ}]$ for some scheme $Y$ and an action of

1583: $GL_{n,\ZZ}$.  Let $f:\cX \rightarrow BGL_{n,\ZZ}$ be the obvious

1584: $1$-morphism.

1585: There exists a finitely-generated free

1586: $\ZZ$-module, $W$ which is a representation of $GL_{n,\ZZ}$ such that

1587: if $E$ on $BGL_{n,\ZZ}$ is the induced locally free sheaf then $f^* E$

1588: has the following property:  For every morphism of algebraic spaces

1589: $g:X'\rightarrow X$ with corresponding map $h:\cX \times_X X'

1590: \rightarrow \cX$ and for every quasi-coherent sheaf $\cF$ on $\cX

1591: \times_X X'$, we have that $h^*f^* E$ is a generator for $\cF$.

1592: \end{thm}

1593: 

1594: \begin{proof}

1595: First of all suppose we prove the result when $g:X'\rightarrow X$ is

1596: the identity map.  Then by remark $(1)$ following the statement of

1597: theorem~\ref{thm-genpt}, it follows that for every

1598: algebraically-closed field $k$ and every $1$-morphism $\zeta:\SP k

1599: \rightarrow \cX$

1600: we have that $\widetilde{\zeta}^*f^* E$ generates the left regular

1601: representation of $G_\zeta$.

1602: Now suppose that $g:X'\rightarrow X$ is an arbitrary morphism, $k$ is

1603: an algebraically-closed field and $\xi:\SP k \rightarrow \cX \times_X

1604: X'$ is a $1$-morphism.  Let $\zeta:\SP k \rightarrow \cX$ be $h\circ

1605: \xi$.  Then $G_\xi = G_\zeta$, and $\xi^*h^*f^*E = \zeta^* f^* E$.  So

1606: we see that $h^* f^*E$ on $\cX \times_X X'$ satisfies the hypotheses

1607: of theorem~\ref{thm-genpt}.  So $h^* f^* E$ generates every

1608: quasi-coherent sheaf.  Therefore to prove our theorem, we need only

1609: consider the case that $g:X'\rightarrow X$ is the identity map.

1610: 

1611: By remark $(2)$ following the statement of theorem~\ref{thm-genpt},

1612: for each geometric point $\zeta:\SP k \rightarrow \cX$, there is a

1613: finitely-generated free $\ZZ$-module $W_\zeta$ which is a

1614: representation of $GL_{n,\ZZ}$ and such that if $E_\zeta$ is the

1615: corresponding locally free sheaf on $BGL_{n,\ZZ}$, then

1616: $\widetilde{\zeta}^* f^* W_\zeta$ generates the left regular

1617: representation.  Then by theorem~\ref{thm-genpt}, there is a Zariski

1618: open neighborhood $U_\zeta \subset X$ of the image of $\zeta(\SP k)$

1619: such that the restriction of $f^* E_\zeta$ to $\cX \times_X U_\zeta$

1620: generates all quasi-coherent sheaves.  Allowing $k$ and $\zeta$ to

1621: vary over all algebraically-closed fields and $1$-morphisms, we obtain

1622: a Zariski open cover $\{ U_\zeta\}$ of $X$.  Since $\cX$ is

1623: quasi-compact, some finite subset covers $X$, say

1624: $\zeta_1,\dots,\zeta_N$.  Let $W$ be the $GL_{n,\ZZ}$-representation

1625: $W_{\zeta_1} \oplus \dots \oplus W_{\zeta_N}$.  Then clearly $W$

1626: satisfies the conditions in the theorem.

1627: \end{proof}

1628: 

1629: 

1630: \section{Natural transformation of Quot functors}\label{nattransec}

1631: Suppose that $S$ is a quasi-compact algebraic space, $f:\cX\rightarrow

1632: S$ is a separated, finitely-presented $1$-morphism of a tame

1633: Deligne-Mumford stack to $S$ such that $\cX$ is a global quotient over

1634: $S$, say $\cX=[Y/GL_{n,X}]$.  By theorem~\ref{thm-genthm}, there is a

1635: $GL_{n,\ZZ}$ representation $V$ which is a finite free $\ZZ$-module

1636: and such that for every morphism of algebraic spaces $T\rightarrow S$

1637: and every quasi-coherent sheaf $\cG$ on $T\times_S \cX$, $\cE_V$

1638: generates $\cG$.

1639: 

1640: \

1641: 

1642: Suppose that $\cF$ is a locally finitely-presented $\OO_{\cX}$-module.

1643: Suppose that $P:K^0(\cX)\rightarrow \ZZ$ is a given Hilbert

1644: polynomial.

1645: Let $p:\cX\rightarrow X$ denote the coarse moduli scheme of $\cX$ and

1646: let $P_V:K^0(X)\rightarrow \ZZ$ denote the map

1647: \begin{equation}

1648: P_V([E]) = P([\textit{Hom}_{\OO_{\cX}}(\cE_V,p^*E)]).

1649: \end{equation}

1650: Suppose $T\rightarrow S$ is a morphism of a scheme to $S$ and suppose

1651: that $\cF_T\rightarrow \cG$ is an element of $Q^P(\cF/\cX/S)(T)$.

1652: Consider $F_{\cE_V}(\cF_T)\rightarrow F_{\cE_V}(\cG)$.  Since

1653: $F_{\cE_V}$ is exact, this is still surjective.  By

1654: lemma~\ref{lem-tame}, $F_{\cE_V}(\cG)$ is locally finitely-presented,

1655: has proper support over $T$ and is flat over $T$.  Moreover every

1656: geometric fiber has Hilbert polynomial $P_V$,

1657: i.e. $F_{\cE_V}(\cF_T)\rightarrow F_{\cE_V}(\cG)$ is an element of

1658: $Q^{P_V}(F_{\cE_V}/X/S)(T)$.  This defines a natural transformation

1659: \begin{equation}

1660: F_{\cE_V}:Q^P(\cF/\cX/S) \Rightarrow Q^{P_V}(F_{\cE_V}(\cF)/X/S).

1661: \end{equation}

1662: 

1663: \begin{lem}~\label{lem-mono}

1664: The natural transformation $F_{\cE_V}$ is a monomorphism, i.e. for

1665: each $T\rightarrow S$ the morphism of sets

1666: \begin{equation}

1667: F_{\cE_V}(T):Q^P(\cF/\cX/S)(T)\rightarrow

1668: Q^{P_V}(F_{\cE_V}(\cF)/X/S)(T),

1669: \end{equation}

1670: is an injection of sets.

1671: \end{lem}

1672: 

1673: \begin{proof}

1674: Given any morphism $T\rightarrow S$ of a scheme to $S$ and given any

1675: element $\alpha:F_{\cE_V}(\cF)_T\rightarrow G$ in

1676: $Q^{P_V}(F_{\cE_V}(\cF)/X/S)(T)$, define $\beta:K\rightarrow

1677: F_{\cE_V}(\cF)_T$ to be the kernel of

1678: $\alpha$ and define $\eta_T(\alpha):\cF\rightarrow \cG$ to be the

1679: cokernel of the composition:

1680: \begin{equation}

1681: \begin{CD}

1682: p^*K\otimes_{\OO_{\cX}} \cE_V @> p^*\beta\otimes 1 >> G_{\cE_V}(\cF_T)

1683: @> \theta_{\cE_V}(\cF_T) >> \cF_T.

1684: \end{CD}

1685: \end{equation}

1686: 

1687: \

1688: 

1689: Now suppose given $\nu:\cF_T\rightarrow \cG$ in $Q^P(\cF/\cX/S)(T)$

1690: and let $\mu:\cK\rightarrow \cF_T$ denote the kernel of $\nu$.  Then

1691: we have a short exact sequence:

1692: \begin{equation}

1693: \begin{CD}

1694: 0 @>>> F_{\cE_V}(\cK) @> F_{\cE_V}(\mu) >> F_{\cE_V}(\cF)_T @>

1695: F_{\cE_V}(\nu) >> F_{\cE_V}(\cG) @>>> 0.

1696: \end{CD}

1697: \end{equation}

1698: Notice that if $\alpha=F_{\cE_V}(\nu)$, then $\beta=F_{\cE_F}(\mu)$.

1699: Since $p^*$ is right-exact, we have a commutative diagram of exact

1700: sequences:

1701: \begin{equation}

1702: \begin{CD}

1703: & & G_{\cE_V}(\cK) @> p^*\beta\otimes 1 >>

1704:     G_{\cE_V}(\cF_T) @> p^*\alpha\otimes 1 >>

1705:       G_{\cE_V}(\cG) @>>> 0 \\

1706: & & @V \theta_{\cE_V}(\cK) VV     @V \theta_{\cE_V}(\cF_T) VV

1707:      @V \theta_{\cE_V}(\cG) VV \\

1708: 0 @>>> \cK @> \mu >> \cE_T @> \nu >> \cK @>>> 0

1709: \end{CD}

1710: \end{equation}

1711: It follows by the snake lemma that we have the formula:

1712: \begin{equation}~\label{eqn-main}

1713: \eta_T(F_{\cE_V}(\mu)) = \mu.

1714: \end{equation}

1715: Therefore $F_{\cE_V}$ is injective.

1716: \end{proof}

1717: 

1718: \textbf{Remark:}  Notice that the association $T\mapsto \eta_T$ is

1719: functorial for arbitrary $S$-morphisms $T_1\rightarrow T_2$.  This

1720: might seem odd since the formation of the kernel $K$ is not compatible

1721: arbitrary base-change.  But one could define $\eta_T$ equivalently to

1722: be the coequalizer of the diagram:

1723: \begin{equation}

1724: \begin{CD}

1725: G_{\cE_V}(\cF_T) @> p^*\alpha \otimes 1 >> p^* G \otimes_{\OO_{\cX}}

1726: \cE_V \\

1727: @V \theta_{\cE_V}(\cF_T) VV \\

1728: \cF_T

1729: \end{CD}

1730: \end{equation}

1731: Since coequalizers are compatible with arbitrary base-change, we see

1732: that $\eta$ is compatible with arbitrary $S$-morphisms $T_1\rightarrow

1733: T_2$.

1734: 

1735: \

1736: 

1737: \begin{prop}~\label{prop-relrep}

1738: The monomorphism $F_{\cE_V}$ of functors is relatively

1739: representable by schemes.  In fact, $F_{\cE_V}$ is a

1740: finitely-presented, quasi-finite, monomorphism.  If $\cF$ has

1741: proper support over $S$, then $F_{\cE_V}$ is a finitely-presented,

1742: finite, monomorphism, i.e. a finitely-presented closed immersion.

1743: \end{prop}

1744: 

1745: \begin{proof}

1746: Suppose $T\rightarrow S$ is a morphism of a scheme to $S$ and suppose

1747: that $\alpha:F_{\cE_V}(\cF_T)\rightarrow G$ is an element of

1748: $Q^{P_V}(F_{\cE_V}(\cF)/X/S)$.  Now form

1749: $\eta_T(\alpha):\cF_T\rightarrow \cG$.  By theorem~\ref{thm-fs}, there

1750: is a flattening stratification $g:\Sigma\rightarrow T$ for $\cG$ and

1751: $g$ is a surjective, finitely-presented, quasi-finite monomorphism.

1752: By lemma~\ref{lem-HPconst}, for each connected component $\Sigma_i$ of

1753: $\Sigma$, the restriction of $\cG$ to $\Sigma_i$ has constant Hilbert

1754: polynomial $P_i$.  In particular, there is a connected component

1755: (possibly empty) $\Sigma_i$ of $\Sigma$ on which the restriction of

1756: $\cG$ has Hilbert polynomial $P$.  Of course we have an induced

1757: natural transformation of functors

1758: \begin{equation}

1759: \Sigma_i\rightarrow T\times_{\alpha,Q^{P_V}(F_{\cE_V}(\cF)/X/S)}

1760: Q^P(\cF/\cX/S).

1761: \end{equation}

1762: By equation~\ref{eqn-main}, we have an inverse natural

1763: transformation.  Thus we conclude that the fiber functor

1764: $T\times_{\alpha,Q^{P_V}(F_{\cE_V}(\cF)/X/S)} Q^P(\cF/\cX/S)$ is

1765: represented by the morphism $g_i:\Sigma_i\rightarrow T$.

1766: Notice that this is a finitely-presented, quasi-finite monomorphism of

1767: schemes.

1768: 

1769: \

1770: 

1771: In case $\cF$ has proper support over $S$, we know from

1772: theorem~\ref{thm-thm1} that $Q^P\rightarrow S$ satisfies the valuative

1773: criterion of properness.  Therefore $g_i:\Sigma_i\rightarrow T$

1774: satisfies the valuative criterion of properness, i.e. $g_i$ is

1775: finite.  But a finite monomorphism is precisely a closed immersion,

1776: therefore $g_i$ is a finitely-presented closed immersion.

1777: \end{proof}

1778: 

1779: 

1780: Now we are in a position to prove theorem~\ref{thm-thm3}.  By

1781: proposition~\ref{prop-relrep}, we know that

1782: \begin{equation}

1783: Q^P(\cF/\cX/S) \rightarrow Q^{P_V}(F_{\cE_V}(\cF)/X/S)

1784: \end{equation}

1785: is relatively representable by a

1786: finitely-presented, quasi-finite monomorphism

1787: (resp. finitely-presented closed immersion).

1788: 

1789: \

1790: 

1791: Now the proof that $Q^{P_V}(F_{\cE_V}(\cF)/X/S)\rightarrow S$ is

1792: represented by a scheme which is quasi-projective over $S$

1793: is essetially ~\cite[th\'eor\`eme 3.2, part IV]{FGA}.  In his proof,

1794: Grothendieck makes some Noetherian hypotheses which are eliminated in

1795: a standard way, cf.~\cite[section IV.8.9]{EGA}.  For the sake of

1796: completeness, we include the proof here.

1797: 

1798: \

1799: 

1800: We have a locally closed immersion $X\hookrightarrow \PP^N_S$ for some

1801: $N$.

1802: By ~\cite[proposition IV.8.9.1]{EGA}, we can find a finite-type affine

1803: $\ZZ$-scheme, $S'$, a quasi-projective (resp. projective) $S'$-scheme,

1804: $X'$, and a coherent sheaf $F'$ on $X'$ along with a morphism

1805: $S\rightarrow S'$ such that $X$ is isomorphic to $S\times_{S'} X'$,

1806: and under this isomorphism $F_{\cE_V}(\cF)$ is isomorphic to the

1807: pullback of $F'$.  Let $p:\ZZ[t,t^{-1}]\rightarrow \ZZ$ be the

1808: polynomial $p(t^n) = P_V([\OO_X(n)])$ where $\OO_X(n)$ is the pullback

1809: to $X$ of the invertible sheaf $\OO_{\PP^N_S}(n)$ on $\PP^N_S$.

1810: polynomial, and consider the Quot functor $Q^p(F'/X'/S')$

1811: which is the connected component of $Q(F'/X'/S')$ parametrizing

1812: families of quotients $F'_T\rightarrow G$ such that for each closed

1813: point $t\in T$, we have $\chi(G_t\otimes \OO(n)) = p(t^n)$.  By

1814: lemma~\ref{lem-HPconst}, it follows

1815: that $Q^{P_V}(F_{\cE_V}(\cF)/X/S)$ is isomorphic to a connected

1816: component of the fiber product $S\times_{S'} Q^p(F'/X'/S')$.  So to

1817: prove that $Q^{P_V}(F_{\cE_V}(\cF)/X/S)\rightarrow S$ is

1818: quasi-projective, it suffices

1819: to show that $Q^p(F'/X'/S')\rightarrow S'$ is quasi-projective.

1820: 

1821: \

1822: 

1823: By ~\cite[th\'eor\`eme 3.2, part IV]{FGA}, we know that the

1824: restriction of the functor $Q^p(F'/X'/S')$ to the category of

1825: locally Noetherian $S'$-schemes is represented by a

1826: quasi-projective $S'$-scheme.  Using ~\cite[proposition

1827: IV.8.9.1]{EGA} and ~\cite[th\'eor\`eme IV.11.2.6]{EGA}, it follows

1828: that in fact this quasi-projective $S'$-scheme represents

1829: $Q^p(F'/X'/S')$ on the category of all $S'$-schemes.  Thus

1830: $Q^{P_V}(F_{\cE_V}(\cF)/X/S)\rightarrow S$ is represented by a

1831: quasi-projective scheme.  This completes the proof of

1832: theorem~\ref{thm-thm3}.

1833: 

1834: \bibliography{my}

1835: \bibliographystyle{abbrv}

1836: 

1837: \end{document}

1838: