1: \documentclass[twoside,reqno]{amsart}

2: \usepackage{amsmath}

3: \usepackage{amscd}

4: \usepackage{latexsym}

5: \usepackage{amsthm}

6: \usepackage{amssymb}

7: 

8: \theoremstyle{plain}

9: \newtheorem{thm}{Theorem}[section]

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

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

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

13: \newtheorem{obs}[thm]{Observation}

14: 

15: \theoremstyle{definition}

16: \newtheorem{defn}[thm]{Definition}

17: \newtheorem{ex}[thm]{Example}

18: \newtheorem{notat}[thm]{Notation}

19: \newtheorem{hyp}[thm]{Hypothesis}

20: 

21: \theoremstyle{remark}

22: \newtheorem{rmk}[thm]{Remark}

23: 

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

25: \newcommand{\pr}{\text{pr}}

26: \newcommand{\mc}{\mathcal}

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

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

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

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

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

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

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

34: \newcommand{\GG}{\mathbb{G}}

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

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

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

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

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

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

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

42: \newcommand{\Kgnb}[2]{\overline{\mc M}_{#1}({#2})}

43: \newcommand{\Kbm}{\overline{\mc M}}

44: \newcommand{\Ext}{\mathbb{E}\text{xt}}

45: \newcommand{\EExt}[1]{\mathbb{E}\text{xt}_{\OO_{#1}}}

46: 

47: \def\ul{\underline}

48: \def\la{\longrightarrow}

49: \def\ni{\noindent}

50: \def\cl{\centerline}

51: 

52: \def\Sym{{\rm Sym}}

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

54: 

55: \def\A{{\mathbb A}}

56: \def\C{{\mathbb C}}

57: \def\F{{\mathbb F}}

58: \def\G{{\mathbb G}}

59: \def\P{{\mathbb P}}

60: \def\Q{{\mathbb Q}}

61: \def\R{{\mathbb R}}

62: \def\Z{{\mathbb Z}}

63: 

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

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

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

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

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

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

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

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

72: 

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

74: 

75: \newcommand{\lt}{\left}

76: \newcommand{\rt}{\right}

77: 

78: \newcommand{\g}{{\frak g}}

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

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

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

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

83: \newcommand{\B}{{\cal B}}

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

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

86: 

87: \setcounter{tocdepth}{1}

88: 

89: 

90: \begin{document}

91: 

92: \title[Rational curves on hypersurfaces, II]{Rational curves on

93: hypersurfaces of low degree, II}  

94: 

95: \author[J. Harris]{Joe Harris} 

96:  \address{Department of Mathematics \\ 

97:    Harvard University \\ 

98:    Cambridge MA 02138}

99:  \email{harris@math.harvard.edu}

100: 

101: \author[J. Starr]{Jason Starr} 

102:  \address{Department of Mathematics \\ 

103:   Massachusetts Institute of Technology \\ 

104:   Cambridge MA 02139}

105:  \email{jstarr@math.mit.edu} \date{\today}

106: 

107: \begin{abstract}

108:   This is a continuation of ~\cite{HRS2} in which we proved

109:   irreducibility of spaces of rational curves on a general

110:   hypersurface $X_d\subset \PP^n$ of degree $d<\frac{n+1}{2}$.  In

111:   this paper, we prove that if $d^2 + d + 2 \leq n$ and if $d\geq 3$,

112:   then the spaces of rational curves are themselves rationally

113:   connected.

114: \end{abstract}

115: 

116: \maketitle

117: 

118: \tableofcontents

119: 

120: \section{Statement of results}~\label{sec-results}

121: 

122: In ~\cite{HRS2}, it is proved that if $X_d \subset \PP^n$ is a general

123: hypersurface of degree $d < \frac{n+1}{2}$, then each space

124: $\text{RatCurves}^e(X)$ parametrizing smooth rational curves of degree

125: $e$ on $X$, is itself an integral, local complete intersection scheme

126: of the expected dimension $(n+1-d)e+(n-4)$.  More precisely, it is

127: proved that for every stable $A$-graph $\tau$ and every flag $f\in

128: \text{Flag}(\tau)$, the Behrend-Manin stack $\Kbm(X,\tau)$ is an

129: integral, local complete intersection stack of the expected dimension

130: $\text{dim}(X,\tau)$, and the evaluation morphism

131: $\text{ev}_f:\Kbm(X,\tau) \rightarrow X$ is flat of the expected fiber

132: dimension $\text{dim}(X,\tau) - \text{dim}(X)$.  Since

133: $\text{RatCurves}^e(X)$ is a Zariski open set in the stack

134: $\Kgnb{0,0}{X,e}$, the result on $\text{RatCurves}^e(X)$ follows.

135: 

136: \

137: 

138: After establishing irreducibility and the dimension of the spaces

139: $\text{RatCurves}^e(X)$, the next question is to determine the Kodaira

140: dimension of $\text{RatCurves}^e(X)$.  For a general Fano hypersurface

141: $X_d \subset \PP^n$ with $d\leq n$, determining the Kodaira dimensions

142: of $\text{RatCurves}^e(X)$ is subtle.  For instance for smooth cubic

143: threefolds $X_3 \subset \PP^4$, $X$ has a nontrivial intermediate

144: Jacobian $J(X)$, and the Abel-Jacobi maps $\text{RatCurves}^e(X)

145: \rightarrow J(X)$ is dominant for $e \geq 4$.  So the Kodaira

146: dimension is at least $0$; conjecturally the fibers of the Abel-Jacobi

147: map are rationally connected so that the Kodaira dimension is exactly

148: $0$.

149: 

150: \

151: 

152: On the other hand for $d =1,2$, it is a theorem of Kim and

153: Pandharipande ~\cite[theorem 3]{KP}, that each of the spaces

154: $\text{RatCurves}^e(X)$ is itself a rational variety, and thus has

155: Kodaira dimension $-\infty$.  In this paper we present the following

156: generalization of ~\cite[theorem 3]{KP}:

157: 

158: \begin{thm} \label{thm-thm1}

159:   Given positive integers $(d,n)$ with $d^2 + d + 2 \leq n$ and $d\geq

160:   3$, for $X_d \subset \PP^n$ a general hypersurface of degree $d$,

161:   each of the spaces $\Kgnb{0,0}{X,e}$ is rationally connected.  Thus

162:   $\text{RatCurves}^e(X)$ has a rationally connected compactification.

163: \end{thm}

164: 

165: To remind the reader, a variety $V$ is rationally connected if given

166: two general points $p,q\in V$, there is a map $f:\PP^1 \rightarrow V$

167: whose image contains $p$ and $q$.  This property is strictly weaker

168: than rationality, and it is unknown whether this property is the same

169: as unirationality.  It is a priori a much simpler property to check.

170: And any rationally connected variety has Kodaira dimension $-\infty$,

171: therefore each of the schemes $\text{RatCurves}^e(X)$ has Kodaira

172: dimension $-\infty$.

173:    

174: \

175: 

176: The proofs rely on ~\cite[theorem IV.3.7]{K}: given a smooth variety

177: $V$ and a morphism $f:\PP^1 \rightarrow V$ such that $f^* T_V$ is an

178: ample vector bundle, then $V$ is rationally connected.  This criterion

179: also works for smooth Deligne-Mumford stacks $V$.  For readers not

180: versed in stacks, this is a moot point -- every morphism of $\PP^1$

181: into a stack constructed in this paper can be deformed to a map

182: contained in the fine moduli locus of the stack.  For a Behrend-Manin

183: stack, $\Kbm(X,\tau)$, a morphism $f:\PP^1 \rightarrow \Kbm(X,\tau)$

184: is equivalent to a fibered surface $\pi:\Sigma \rightarrow \PP^1$,

185: with some collection of sections $\sigma_1,\dots,\sigma_r:\PP^1

186: \rightarrow \Sigma$, and a map $f:\Sigma \rightarrow X$.  Assuming

187: that $f:\PP^1 \rightarrow \Kbm(X,\tau)$ maps into the

188: \emph{unobstructed locus}, the bundle $f^*T_{\Kbm(X,\tau)}$ can be

189: computed from a universal construction applied to the datum

190: $(\pi:\Sigma \rightarrow \PP^1,\sigma_1,\dots,\sigma_r,f)$.  Thus, to

191: prove $\Kbm(X,\tau)$ is rationally connected, we are reduced to

192: finding a datum $(\pi:\Sigma \rightarrow

193: \PP^1,\sigma_1,\dots,\sigma_r,\PP^1)$ satisfying certain properties.

194: 

195: \

196: 

197: In the proof of the induction step, we will use the following

198: technical hypotheses:

199: 

200: \begin{hyp}\label{hyp-1}

201:   For each contraction of genus $0$ stable $A$-graphs,

202:   $\phi:\sigma\rightarrow \tau$, the codimension of the image of the

203:   corresponding morphism of Behrend-Manin stacks $\Kbm(X,\sigma)

204:   \rightarrow \Kbm(X,\tau)$ equals $\text{dim}(X,\tau) -

205:   \text{dim}(X,\sigma)$.

206: \end{hyp}

207: 

208: In particular, by \cite[proposition 7.4]{HRS2}, for a general

209: $X_d\subset \PP^n$ with $d<\frac{n+1}{2}$, each stack $\Kbm(X,\sigma)$

210: has the expected dimension, and hypothesis~\ref{hyp-1} holds for $X$.

211: 

212: \begin{hyp}\label{hyp-1.5}

213:   The general fiber of the evaluation map $\text{ev}:\Kgnb{0,1}{X,1}

214:   \rightarrow X$ is irreducible.

215: \end{hyp}

216: 

217: In particular, for a general pair $([X],[p])$ consisting of a

218: hypersurface $X\subset \PP^n$ of degree $d$ and a point $p\in X$, the

219: associated fiber of $\text{ev}$ is a subvariety $Z\subset \PP^{n-1}$

220: which is a complete intersection of a general sequence of

221: hypersurfaces $Y_1,\dots, Y_d$ in $\PP^n$ with $\text{deg}(Y_i)=i$.

222: By the Bertini-Kleiman theorem, a general such complete intersection

223: is smooth and connected if $d < n-1$.

224: 

225: \begin{hyp}\label{hyp-1.75}

226:   For each integer $e\geq 0$, the locus in $\Kgnb{0,1}{X,e}$

227:   parametrizing stable maps with nontrivial automorphism group has

228:   codimension at least $2$.

229: \end{hyp}

230: 

231: Of course any stable map with nontrivial automorphism group has an

232: irreducible component which is a multiple cover of its image.  In

233: light of \cite[proposition 7.4]{HRS2}, a simple parameter count shows

234: that for a general hypersurface $X_d \subset \PP^n$ with $d\leq

235: \frac{n+1}{2}$, hypothesis~\ref{hyp-1.75} is satisfied.

236: 

237: \subsection{Acknowledgments}

238: 

239: We are very grateful to Johan de Jong and Steven Kleiman for many

240: useful conversations.

241: 

242: \section{Deformation ample}~\label{sec-DA}

243: Unless stated otherwise, all schemes will be finite type, separated

244: schemes over $\SP \CC$.  All absolute fiber products will be fiber

245: products over $\SP \CC$.

246: 

247: \

248: 

249: The following lemma is completely trivial.

250: 

251: \begin{lem}~\label{lem-gend}

252:   Let $B$ be a connected, proper, prestable curve of arithmetic genus

253:   $0$ and let $E$ be a locally free sheaf on $E$ which is generated by

254:   global sections.  Then $H^1(B,E)$ is zero.  Moreover if $p\in B$ is

255:   any point and ${\mc I}_p\subset \OO_B$ is the corresponding ideal

256:   sheaf, then $H^1(B,{\mc I}_p\cdot E)$ is zero.

257: \end{lem}

258: 

259: \begin{proof}

260:   Since $E$ is generated by global sections, there is a short exact

261:   sequence of the form:

262: \begin{equation}

263: \begin{CD}

264: 0 @>>> K @>>> \OO_B^{\oplus N} @>>> E @>>> 0

265: \end{CD}

266: \end{equation}

267: Since $B$ is a curve, $H^2(B,K)=0$.  Therefore we have a surjection

268: $H^1(B,\OO_B)^{\oplus N} \rightarrow H^1(B,E)$.  Since $B$ is

269: connected of arithmetic genus $0$, $H^1(B,\OO_B)$ is zero.  Hence

270: $H^1(B,E)$ is zero.

271: 

272: \

273: 

274: Now for any point $p\in B$ we have a short exact sequence:

275: \begin{equation}

276: \begin{CD}

277: 0 @>>> {\mc I}_p\cdot E @>>> E @>>> E|_p @>>> 0

278: \end{CD}

279: \end{equation}

280: This gives rise to a long exact sequence in cohomology:

281: \begin{equation}

282: \begin{CD}

283: H^0(B,E) @>>> E|_p @>>> H^1(B,{\mc I}_p\cdot E) @>>> H^1(B,E)

284: \end{CD}

285: \end{equation}

286: By the last paragraph, $H^1(B,E)$ is zero.  And since $E$ is generated

287: by global sections, $H^0(B,E)\rightarrow E|_p$ is surjective.

288: Therefore $H^1(B,{\mc I}_p \cdot E)$ is zero.

289: \end{proof}

290: 

291: Suppose given a connected, proper, prestable curve $B$ of arithmetic

292: genus $0$ and a locally free sheaf $E$ of positive rank on $B$.  Any

293: pair $({\mc B},{\mc E})$ consisting of a flat family of connected,

294: proper prestable curves over a DVR, say $\pi:{\mc B} \rightarrow \SP

295: R$ along with a locally free sheaf ${\mc E}$ on ${\mc B}$ such that

296: the closed fiber of $\pi$ is isomorphic to $B$, such that the generic

297: fiber $B_\eta$ of $\pi$ is smooth, and such that the restriction of

298: ${\mc E}$ to $B\subset {\mc B}$ is isomorphic to $E$ will be called a

299: \emph{smoothing} of the pair $(B,E)$.  We want to know when $(B,E)$

300: satisfies the condition that for every smoothing $({\mc B},{\mc E})$,

301: the restriction ${\mc E}_\eta$ of ${\mc E}$ to the generic fiber is an

302: ample locally free sheaf.  Certainly if $E$ is ample, this is true,

303: but $E$ need not be ample for this condition to hold: e.g. if $E$ is

304: an invertible sheaf such that the total degree of $E$ is positive,

305: then every smoothing of $(B,E)$ will have ample generic fiber.

306: Although it is not the most general criterion, we find the following

307: notion to be useful and it is the one we use in the remainder of the

308: paper.

309: 

310: \

311: 

312: \begin{defn}~\label{def-DA}

313:   Let $B$ be a connected, proper, prestable curve of arithmetic genus

314:   $0$.  A locally free sheaf $E$ on $B$ with positive rank is

315:   \emph{deformation ample} if

316:   \begin{enumerate}

317:     \item $E$ is generated by global sections, and

318:     \item $H^1(B,E(K_B))$ is zero, where $\OO_B(K_B)$ is the dualizing

319:       sheaf of $B$.

320:   \end{enumerate}

321: \end{defn}

322: 

323: \begin{rmk}~\label{rmk-DA}

324:   \begin{enumerate}

325:     \item Conditions (1) and (2) above are independent.

326:     \item If $E$ is invertible, then $E$ is deformation ample iff the

327:       restriction of $E$ to every irreducible component has

328:       nonnegative degree and the restriction to at least one

329:       irreducible component has positive degree.

330:     \item For a general $E$, one can determine whether $E$ is

331:       deformation ample in terms of the splitting type of the

332:       restriction of $E$ to each irreducible component together with

333:       the patching isomorphisms at the nodes of $B$.

334:   \end{enumerate}

335: \end{rmk}

336: 

337: \begin{defn}~\label{defn-fDA}

338:   Let $T$ be a scheme and let $\pi:B\rightarrow T$ be a family of

339:   prestable curves of arithmetic genus $0$.  A locally free sheaf $E$

340:   on $B$ with positive rank is \emph{$\pi$-relatively deformation

341:     ample} (or simply deformation ample if $\pi$ is understood) if

342:   \begin{enumerate}

343:     \item the canonical map $\pi^*\pi_*E\rightarrow E$ is surjective, and

344:     \item $R^1\pi_*(E(K_\pi))$ is zero, where $\OO_B(K_\pi)$ is the

345:       relative dualizing sheaf of $\pi$.

346:   \end{enumerate}

347: \end{defn}

348: 

349: \begin{lem}~\label{lem-bcDA}

350:   With notation as in definition~\ref{defn-fDA}, suppose

351:   $f:T'\rightarrow T$ is a morphism of schemes and let

352:   $\pi':B'\rightarrow T'$ be the base-change of $\pi$ by $f$.  Let

353:   $E'$ be the pullback of $E$ by the projection $g:B'\rightarrow B$.

354:   If $E$ is $\pi$-relatively deformation ample, then $E'$ is

355:   $\pi'$-relatively deformation ample.  If $f'$ is surjective, the

356:   converse also holds.

357: \end{lem}

358: 

359: \begin{proof}

360:   For the main direction, by ~\cite[section 8.5.2, proposition

361:   8.9.1]{EGA4}, it suffices to consider the case when $T$ and $T'$ are

362:   Noetherian affine schemes.

363: 

364: \

365: 

366: There is a canonical map of $\OO_{T'}$-modules $\alpha:f^*\pi_*

367: E\rightarrow (\pi')_*g^* E$.  There is a commutative diagram:

368: \begin{equation}

369: \begin{CD}

370: (\pi')^*f^*\pi_* E @> = >> g^* \pi^* \pi_* E \\

371: @V (\pi')^*\alpha VV @VVV \\

372: (\pi')^*(\pi')_* E' @>>> E'

373: \end{CD}

374: \end{equation}

375: Since $\pi^*\pi_*E \rightarrow E$ is surjective, we conclude that

376: $g^*\pi^*\pi_* E\rightarrow E'$ is surjective.  Therefore also

377: $(\pi')^*(\pi')_*E' \rightarrow E'$ is surjective.

378:  

379: 

380: \

381: 

382: Now $R^2\pi_* E(K)$ is identically zero.  So by \cite[theorem

383: III.12.11(b)]{H}, we conclude that for every closed point $t\in T$, we

384: have $H^1(B_t,E(K)|_{B_t}) = 0$.  By ~\cite[prop. III.9.3]{H}, we

385: conclude that for every closed point $t'\in T'$, we have

386: $H^1(B'_{t'},E'(K')|_{B'_{t'}}) = 0$.  So by \cite[theorem

387: III.12.11(a)]{H} and Nakayama's lemma, we conclude that

388: $R^1\pi'_*(E'(K'))$ is identically zero.  This shows that $E'$ is

389: $\pi'$-relatively deformation ample.

390: 

391: \

392: 

393: For the converse in case $T'\rightarrow T$ is surjective, we may

394: reduce to the case that $T$ is a Noetherian affine scheme.  By the

395: argument above, we conclude that for each closed point $t'\in T'$, we

396: have $H^1(B'_{t'},E'(K')|_{B'_{t'}}) = 0$.  Since $T'\rightarrow T$ is

397: surjective, we conclude by ~\cite[prop. III.9.3]{H} that for each

398: closed point $t\in T$ we have $H^1(B_t,E(K)|_{B_t}) = 0$.  So by

399: ~\cite[theorem III.12.11(a)]{H} and Nakayama's lemma, we conclude that

400: $R^1\pi_*(E(K))$ is identically zero.

401: 

402: \

403: 

404: It remains to show that $E$ is $\pi$-relatively generated by global

405: sections.  For each point $t\in T$, there is some point $t'\in T'$

406: mapping to $t$.  Since $E'|_{B'_{t'}}$ is generated by global

407: sections, it follows that $E|_{B_t}$ is generated by global sections.

408: So, by lemma~\ref{lem-gend}, we conclude that $H^1\lt( B_t, E|_{B_t}

409: \rt)$ is zero.  So by ~\cite[theorem III.12.11(a)]{H} and Nakayama's

410: lemma, we conclude that $R^1\pi_*(E)$ is identically zero.  Since $E$

411: is flat over $T$, it follows by a standard argument that for any

412: coherent $\OO_T$-module ${\mc F}$, we have that $R^1\pi_*( \pi^*{\mc

413:   F}\otimes E)$ is also zero.  In particular, applying the long exact

414: sequence of higher direct images to the short exact sequence:

415: \begin{equation}

416: \begin{CD}

417: 0 @>>> \pi^*{\mc I}_t\otimes E @>>> E @>>> E|_{B_t} @>>> 0

418: \end{CD}

419: \end{equation}

420: we conclude that $\pi_*(E) \rightarrow H^0\lt( B_t, E|_{B_t} \rt)$ is

421: surjective.  Since $E|_{B_t}$ is generated by global sections for each

422: $t\in T$, we conclude that $E$ is $\pi$-relatively generated by global

423: sections.

424: \end{proof}

425: 

426: \begin{lem}\label{lem-smDA}

427:   With notation as in definition~\ref{defn-fDA}, if $\pi$ is smooth,

428:   then $E$ is $\pi$-relatively deformation ample iff $E$ is

429:   $\pi$-relatively ample.

430: \end{lem}

431: 

432: \begin{proof}

433:   Both properties are local on $T$ and can be checked after \'{e}tale,

434:   surjective base-change of $T$.  Thus we may assume that

435:   $\pi:B\rightarrow T$ is isomorphic to $\pi_1:T\times \PP^1

436:   \rightarrow T$.  Define $F=(\pi_1)_*(E\otimes

437:   \pi_2^*\OO_{\PP^1}(-1))$.  There is a natural map $\alpha:

438:   (\pi_1)^*F \otimes\pi_2^*\OO_{\PP^1}(1)\rightarrow E$.  Suppose that

439:   $E$ is deformation ample.  Then the claim is that $\alpha$ is

440:   surjective.  To prove this, it suffices to prove the following:

441: \begin{enumerate}

442: \item For each geometric point $t$ of $T$ with residue field $\kappa$,

443:   we have $H^1(\PP^1_\kappa, E|_{B_t}\otimes \OO_{\PP^1}(-1))$ is

444:   trivial,

445: \item $\pi^* F|_{B_t} = H^0(\PP^1_\kappa,E|_{B_t}\otimes

446:   \OO_{\PP^1}(-1))$, and

447: \item the map $H^0(\PP^1_\kappa,E|_{B_t}\otimes

448:   \OO_{\PP^1}(-1))\otimes \OO_{\PP^1}(1) \rightarrow E|_{B_t}$ is

449:   surjective. 

450: \end{enumerate}

451: 

452: \

453: 

454: Now by Grothendieck's lemma ~\cite[exercise V.2.6]{H}, $E|_{B_t}$

455: splits as a direct sum $\OO_{\PP^1}(a_1)\oplus \dots\oplus

456: \OO_{\PP^1}(a_r)$ for some integers $a_1\leq \dots \leq a_r$.  By

457: lemma~\ref{lem-bcDA}, we know that $E|_{B_t}$ is deformation ample.

458: It follows that $a_1\geq 1$.  Thus $H^1(B_t, E|_{B_t}\otimes

459: \OO_{\PP^1}(-1))=0$ and $(1)$ is established.  Combined

460: with~\cite[theorem III.12.11(b)]{H}, also $(2)$ follows.  Finally, for

461: $a_i \geq 1$, we clearly have $H^0(\PP^1_\kappa,

462: \OO_{\PP^1}(a_i-1))\otimes \OO_{\PP^1}(1)\rightarrow \OO_{\PP^1}(a_i)$

463: is surjective.  Thus $(3)$ holds and the claim is proved.  Now

464: $(\pi_1)^*F\otimes (\pi_2)^*\OO_{\PP^1}(1)$ is $\pi_1$-relatively

465: ample.  Since $E$ is a quotient of $(\pi_1)^*F\otimes

466: (\pi_2)^*\OO_{\PP^1}(1)$, we conclude that $E$ is also

467: $\pi_1$-relatively ample.

468: 

469: \

470: 

471: The converse result follows in the same way.

472: \end{proof}

473: 

474: \begin{lem}~\label{lem-opDA}

475:   With notation as in lemma~\ref{defn-fDA}, there exists an open

476:   subscheme $i:U\rightarrow T$ with the following property: for every

477:   morphism $f:T'\rightarrow T$, $f(T')$ is contained in $U$ iff $E'$

478:   is $\pi'$-relatively deformation ample.

479: \end{lem}

480: 

481: \begin{proof}

482:   By \cite[section 8.5.2,proposition 8.9.1]{EGA4}, we may reduce to

483:   the case that $T$ and $T'$ are Noetherian affine schemes.

484: 

485: \

486: 

487: Let $Z_1\subset T$ be the closed subset with is the image under $f$ of

488: the support of $\text{coker}(\pi^*\pi_* E\rightarrow E)$.  Let

489: $Z_2\subset T$ be the closed subset which is the support of

490: $R^1\pi_*(E(K_\pi))$.  Let $i:U\rightarrow T$ be the open complement

491: of $Z_1\cup Z_2$.

492: 

493: \

494: 

495: Suppose $f:T'\rightarrow T$ is a morphism of schemes.  By

496: ~\cite[theorem III.12.11, prop. III.9.3]{H} and Nakayama's lemma,

497: $R^1\pi'_*(E'(K'))$ is identically zero iff for each $t'\in T'$ with

498: $t=f(t')$, we have $H^1(B_t,E(K)|_{B_t})$ is zero, i.e. if $t$ is

499: contained in the complement of $Z_2$.  So $R^1\pi'_*(E'(K'))$ is

500: identically zero iff $f(T)$ is contained in the complement of $Z_2$.

501: If $R^1\pi'_*(E'(K'))$ is zero, then also $R^1\pi'_*(E')$ is zero (by

502: the same argument as in the proof of lemma~\ref{lem-bcDA}).  So using

503: ~\cite[theorem III.12.11, prop. III.9.3]{H} again, $E'$ is

504: $\pi'$-relatively generated by global sections iff for each $t'\in T'$

505: with $t=f(t')$, we have $E_t$ is generated by global sections, i.e.

506: $t$ is in the complement of $Z_1$.  So we conclude that $E'$ is

507: $\pi'$-relatively deformation ample iff $f(T)$ is contained in the

508: complement of $Z_1\cup Z_2$.

509: \end{proof}

510: 

511: \begin{lem}~\label{lem-secDA} 

512:   With notation as in definition~\ref{defn-fDA}:

513: \begin{enumerate}

514: \item If $E_1\rightarrow E_2$ is a morphism of locally free

515:   sheaves on $B$ whose cokernel is torsion in every fiber

516:   (in particular, if the

517:   morphism is surjective), if $E_2$ is nonzero, 

518:   and if $E_1$ is $\pi$-relatively deformation ample,

519:   then also $E_2$ is $\pi$-relatively deformation ample.

520: \item Given a short exact sequence of nonzero locally free sheaves on

521:   $B$,

522: \begin{equation}

523: \begin{CD}

524: 0 @>>> E_1 @>>> E_2 @>>> E_3 @>>> 0 

525: \end{CD}

526: \end{equation}

527: if $E_1$ and $E_3$ are $\pi$-relatively deformation ample, then $E_2$

528: is also $\pi$-relatively deformation ample.

529: \item If $E$ is $\pi$-relatively deformation ample, then for each integer

530:   $n\geq 1$, also $E^{\otimes n}$ is $\pi$-relatively deformation

531:   ample.

532: \end{enumerate}

533: \end{lem}

534: 

535: \begin{proof}

536:   First we prove $(1)$.  Let $Q$ denote the cokernel, which is assumed

537:   to be torsion.  Now $R^1\pi_*$ is right exact on coherent sheaves,

538:   $R^1\pi_*(Q(K))$ is zero (because $Q(K)$ is torsion in fibers), and

539:   $R^1\pi_*(E_1(K))=0$ by assumption.  So we conclude that also

540:   $R^1\pi_*(E_2(K))=0$.  Let $I\subset E_2$ denote the image sheaf of

541:   $E_1\rightarrow E_2$.  The surjective composition map

542: \begin{equation}

543: \begin{CD}

544: \pi^*\pi_*E_1 @>>> E_1 @>>> I

545: \end{CD}

546: \end{equation}

547: factors through $\pi^*\pi_* I\rightarrow I$.  Therefore $I$ is

548: $\pi$-relatively generated by global sections.  In particular,

549: $R^1\pi_* I$ is zero (because $R^1\pi_* \OO_B$ is zero).  Since $Q$ is

550: fiberwise torsion, it is $\pi$-relatively generated by global

551: sections.  Now we have a short exact sequence of coherent sheaves:

552: \begin{equation}

553: \begin{CD}

554: 0 @>>> I @>>> E_2 @>>> Q @>>> 0.

555: \end{CD}

556: \end{equation}

557: Applying the long exact sequence of higher direct images, and using

558: that $R^1\pi_*I$ is zero, locally on $B$ all global sections of $Q$

559: lift to global sections of $E_2$.  Since also $I$ is $\pi$-relatively

560: generated by global sections, we conclude that $E_2$ is

561: $\pi$-relatively generated by global sections.  Thus $E_2$ is

562: $\pi$-relatively deformation ample.

563: 

564: \

565: 

566: Next we prove $(2)$.  The long exact sequence of higher direct images

567: and the vanishings $R^1\pi_*(E_1(K))=R^1\pi_*(E_3(K))=0$ imply that

568: $R^1\pi_*(E_2(K))$ is also trivial.  Now $E_1$ and $E_3$ are

569: $\pi$-relatively generated by global sections an so (locally on $T$)

570: they are quotients of a trivial sheaf $\OO_B^{\oplus N}$.  Since

571: $R^1\pi_*\OO_B$ is zero, and since $R^1\pi_*$ is right exact for short

572: exact sequences of coherent sheaves, we conclude that $R^1\pi_*E_1 =

573: R^1\pi_* E_3 = 0$.  Applying the long exact sequence of higher direct

574: images to the short exact sequence above, we conclude that

575: $\pi_*(E_2)\rightarrow \pi_*(E_3)$ is surjective.  Since $E_3$ is

576: $\pi$-relatively generated by global sections, we conclude that

577: $\pi^*\pi_*E_2\rightarrow E_2 \rightarrow E_3$ is surjective.

578: Therefore the map from $E_1$ to the cokernel of

579: $\pi^*\pi_*E_2\rightarrow E_2$ is surjective.  But since $E_1$ is

580: $\pi$-relatively generated by global sections, this map is zero and we

581: conclude $\pi^*\pi_*E_2\rightarrow E_2$ is surjective.  So $E_2$ is

582: $\pi$-relatively deformation ample.

583: 

584: \

585: 

586: Finally we prove $(3)$ by induction on $n$.  It suffices to consider

587: the case when $T$ is affine.  For $n=1$ it is trivial.  Suppose $n>1$

588: and suppose the result is known for all smaller values of $n$.  In

589: particular, $E^{\otimes(n-1)}$ is $\pi$-relatively generated by global

590: sections, and we have a natural surjection

591: \begin{equation}

592: \pi_*\pi^*(E^{\otimes(n-1)})\otimes_{\OO_B} E \rightarrow E^{\otimes n}.

593: \end{equation}

594: Now we can find a surjective map $\OO_T^{\oplus r}\rightarrow

595: \pi_*(E^{\otimes(n-1)})$.  Thus we have a surjection $E^{\oplus

596:   r}\rightarrow E^{\otimes n}$.  By $(2)$ above and induction, we have

597: that $E^{\oplus r}$ is $\pi$-relatively deformation ample.  By $(1)$

598: above, we conclude that $E^{\otimes n}$ is $\pi$-relatively

599: deformation ample, and we have proved $(3)$ by induction on $n$.

600: \end{proof}

601: 

602: \begin{lem}~\label{lem-DAcrit}

603:   Suppose now that $T=\SP k$ for some algebraically closed field $k$.

604:   If $E$ satisfies the hypotheses

605: \begin{enumerate}

606: \item For every irreducible component $B_i\subset B$ we have

607:   $E|_{B_i}$ is generated by global sections, and

608: \item for some nonempty closed subcurve $B'\subset B$, $E|_{B'}$ is

609:   deformation ample,

610: \end{enumerate}

611: then $E$ is deformation ample.

612: \end{lem}

613: 

614: \begin{proof}

615:   Let $\delta$ be the number of irreducible components of $B$ which

616:   aren't contained in $B'$.  We prove the result by induction on

617:   $\delta$.  If $\delta=0$, then $B=B'$ and there is nothing to prove.

618:   Thus suppose that $\delta >0$ and suppose the result has been proved

619:   for all smaller values of $\delta$.

620: 

621: \

622: 

623: Let $B_1\subset B$ be an irreducible component of $B$ which is not in

624: $B'$, let $B_2\subset B$ denote the union of all irreducible

625: components other than $B_1$, and suppose that $B_1$ intersects $B_2$

626: in precisely one node $p$ of $B$.  By the induction assumption

627: $E|_{B_2}$ is deformation ample.

628: 

629: \

630: 

631: First we prove that $E$ is generated by global sections.  Let

632: $F\subset E$ be the image of $H^0(B,E)\otimes_k \OO_B \rightarrow E$.

633: We have a short exact sequence of coherent sheaves:

634: \begin{equation}

635: \begin{CD}

636: 0 @>>> E|_{B_1}(-p) @>>> E @>>> E|_{B_2} @>>> 0.

637: \end{CD}

638: \end{equation}

639: Since $E|_{B_1}$ is a locally free sheaf on $\PP^1$ generated by

640: global sections, Grothendieck's lemma and cohomology of line bundles

641: on $\PP^1$ imply that $H^1(B,E|_{B_1}(-p))=0$.  Thus all the global

642: sections of $E|_{B_2}$ lift to global sections of $E$, i.e.

643: $F\rightarrow E|_{B_2}$ is surjective.  So $E/F$ is supported on $B_1$

644: and thus is a quotient of $E|_{B_1}$.  Since $E|_{B_1}$ is generated

645: by global sections, also $E/F$ is generated by global sections.  We

646: have a short exact sequence

647: \begin{equation}

648: \begin{CD}

649: 0 @>>> F @>>> E @>>> E/F @>>> 0.

650: \end{CD}

651: \end{equation}

652: Since $H^1(B,\OO_B)=0$, also $H^1(B,H^0(B,E)\otimes_k \OO_B) = 0$.

653: Since $H^1(B,*)$ is right exact, $H^1(B,F)=0$.  Thus all the global

654: sections of $E/F$ lift to global sections of $E$.  This can only hold

655: if $E/F=0$, i.e. if $E$ is generated by global sections.

656: 

657: \

658: 

659: Next we prove that $H^1(B,E(K_B))=0$.  We have a short exact sequence

660: of sheaves:

661: \begin{equation}

662: \begin{CD}

663: 0 @>>> E(K_B)|_{B_2}(-p) @>>> E(K_B) @>>> E(K_B)|_{B_1} @>>> 0.

664: \end{CD}

665: \end{equation}

666: This gives an exact sequence of vector spaces:

667: \begin{equation}

668: \begin{CD}

669: H^1(B,E(K_B)|_{B_2}(-p)) @>>> H^1(B,E(K_B)) @>>> H^1(B,E(K_B)|_{B_1})

670: @>>> 0.

671: \end{CD}

672: \end{equation}

673: By standard results on dualizing sheaves and finite morphisms, we have

674: \begin{equation}

675: K_{B_2}=\textit{Hom}_{\OO_B}(\OO_{B_2},K_B) = K_B|_{B_2}(-p).

676: \end{equation}

677: Therefore $H^1(B,E(K_B)|_{B_2}(-p)) = H^1(B_2,E|_{B_2}(K_{B_2}))$,

678: which is zero by the induction assumption.  Similarly, $E(K_B)|_{B_1}$

679: equals $E|_{B_1}(-1)$ (identifying $B_1$ with $\PP^1$).  Since

680: $E|_{B_1}$ is generated by global sections, it follows by

681: Grothendieck's lemma and cohomology of line bundles on $\PP^1$ that

682: $H^1(B_1,E|_{B_1}(-1))$ is trivial.  Therefore we conclude that

683: $H^1(B,E(K_B))$ is trivial.  So $E$ is deformation ample, and the

684: result is proved by induction on $\delta$.

685: \end{proof}

686: 

687: \begin{rmk} \label{rmk-DAcrit}

688:   A particular case of lemma~\ref{lem-DAcrit} is when $B'$ is one

689:   irreducible component of $B$.  Then the lemma says that a locally

690:   free sheaf on $B$ which is \emph{generically ample} in the sense of

691:   Lazarsfeld ~\cite{F} is deformation ample.

692: \end{rmk}

693: 

694: \section{Some deformation theory} \label{sec-def}

695: In the next section we will need some deformation theory of stable

696: maps.

697: 

698: \begin{notat} \label{not-cpx}

699:   Suppose $T$ is a scheme and suppose

700: \begin{equation}

701: \zeta = ((\pi:B\rightarrow T,\sigma_1,\dots, \sigma_r), g:B

702: \rightarrow X)

703: \end{equation}

704: is a family of marked prestable maps to a smooth scheme $X$.  Denote

705: by $L_\zeta$ the complex of coherent sheaves on $B$

706: \begin{equation}

707: \begin{CD}

708: -1 & & 0 \\

709: f^*\Omega_X @> (df)^\dagger >> \Omega_\pi(\sigma_1 + \dots + \sigma_r)

710: \end{CD}

711: \end{equation}

712: Denote by $L_\zeta^\vee$ the object

713: \begin{equation}

714: L_\zeta^\vee :=

715: \mathbb{R}\textit{Hom}_{\OO_B}(L_\zeta,\OO_B)

716: \end{equation} 

717: in the derived category of $B$.

718: \end{notat}

719: 

720: The relevance of the complex $L_\zeta^\vee$ is the following.

721: 

722: \begin{lem} \label{lem-cpx}

723:   Suppose $X$ is a smooth projective scheme.  Let $\pi:{\mc

724:     B}\rightarrow \Kgnb{g,r}{X,\beta}$ denote the universal curve, let

725:   $\sigma_i: \Kgnb{g,r}{X,\beta} \rightarrow {\mc B}$ denote the

726:   universal sections, and let $f:{\mc B} \rightarrow X$ denote the

727:   universal map, i.e.

728: \begin{equation}

729: \zeta = \lt( \lt( \pi:{\mc B} \rightarrow \Kgnb{g,r}{X,\beta},

730: \sigma_1, \dots, \sigma_r \rt), f:{\mc B} \rightarrow X \rt)

731: \end{equation}

732: is the universal family of stable maps.  There is an \emph{obstruction

733:   theory} for $\Kgnb{g,r}{X,\beta}$ in the sense of \cite[definition

734: 4.4]{BF} of the form

735: \begin{equation}

736: \phi:\lt( \mathbb{R}\pi_*(L_\zeta^\vee) [1] \rt)^\vee \rightarrow

737: L_{\Kgnb{g,r}{X,\beta}}.

738: \end{equation}

739: \end{lem}

740: 

741: \begin{proof} Essentially this follows from ~\cite{BF} and

742:   ~\cite{B}.

743: \end{proof}

744: 

745: \begin{rmk} \label{rmk-cpx}

746:   Explicitly, if $\zeta = ((B,p_1,\dots,p_r),f:B\rightarrow X)$ is a

747:   stable map (i.e $T = \SP C$), then the space of first order

748:   deformations of $\zeta$ is given by $\EExt{B}^1(L_\zeta,\OO_B)$ and

749:   the obstruction group is a subgroup of $\EExt{B}^2(L_\zeta,\OO_B)$.

750:   In particular, if $\EExt{B}^2(L_\zeta,\OO_B)$ vanishes, then

751:   $\Kgnb{g,r}{X,\beta}$ is smooth at the point $[\zeta]$.

752: \end{rmk}

753: 

754: \subsection{Contracting unstable components}\label{subsec-unstable}

755: 

756: In this subsection we wish to investigate the relationship between

757: $\EExt{B}(L_\zeta, \OO_B)$ and $\EExt{B'}(L_\zeta',\OO_B')$ where

758: \begin{equation}

759: \zeta = ((B,p_1,\dots,p_r),f:B \rightarrow X)

760: \end{equation}

761: is a prestable map, where

762: \begin{equation}

763: (B',p'_1,\dots, p'_r,q'_1,\dots,q'_s)

764: \end{equation}

765: is a prestable curve, where

766: \begin{equation}

767: h:(B',p'_1,\dots, p'_r) \rightarrow (B,p_1,\dots, p_r)

768: \end{equation} 

769: is a map which contracts some of the unstable components of

770: $(B',p_1,\dots,p_r)$, where $f' = f\circ h$ and where $\zeta'$ is the

771: prestable map

772: \begin{equation}

773: \zeta' = ((B',p'_1,\dots, p'_r,q'_1, \dots, q'_s), f':B' \rightarrow

774: X).

775: \end{equation}

776: 

777: \

778: 

779: Any morphism $h:B'\rightarrow B$ as above can be factored as a

780: sequence of \emph{elementary} maps.  We begin by analyzing the case

781: $\zeta = (B,f:B\rightarrow X)$ is a prestable map without marked

782: points, $h:B'\rightarrow B$ contracts a single unstable component, and

783: $\zeta' = (B', f' = f\circ h:B' \rightarrow X)$ (again without marked

784: points).  We call such maps either \emph{type I} or \emph{type II},

785: depending on whether the image under $h$ of the contracted component

786: is a smooth point of $B$ or a node of $B$.  After this, we analyze the

787: case where $\zeta = (B,(p_1,\dots,p_n),f:B\rightarrow X)$ is a marked

788: prestable map, where $B'$ is the same prestable map, but with one

789: extra marked point, and where $h:B'\rightarrow B$ is the identity map.

790: We call such a map \emph{type III}.

791: 

792: \

793: 

794: To simplify calculations, we replace $L_\zeta$ by a quasi-isomorphic

795: complex as follows.  Choose a regular embedding $i:B\rightarrow S$ of

796: $B$ into a smooth surface $S$.  Then the morphism $(f,i):B \rightarrow

797: X \times S$ is a regular embedding.  Let $N_{(f,i)}$ denote the normal

798: bundle of the embedding.  There are induced morphisms $\alpha_f:

799: N_{(f,i)}^\vee \rightarrow f^* \Omega_X$ and $\beta_i: N_{(f,i)}^\vee

800: \rightarrow i^* \Omega_S$ which are the components of the canonical

801: morphism $N_{(f,i)}^\vee \rightarrow (f,i)^* \Omega_{X \times S}$.

802: Define the complex $L_{(f,i)}$ to be

803: \begin{equation}

804: \begin{CD}

805: -1 & & 0 \\

806: N_{(f,i)}^\vee @> \beta_i >> i^* \Omega_S

807: \end{CD}

808: \end{equation}

809: There is a map of complexes $\gamma_{(f,i)}:L_{(f,i)} \rightarrow

810: L_\zeta$ defined by the commutative diagram

811: \begin{equation}

812: \begin{CD}

813: N_{(f,i)}^\vee @> \beta_i >> i^* \Omega_S \\

814: @V \alpha_f VV  @VV (di)^\dagger V \\

815: f^* \Omega_X @> (df)^\dagger >> \Omega_B

816: \end{CD}

817: \end{equation}

818: 

819: \begin{lem} \label{lem-qism}

820:   The morphism $\gamma_{(f,i)}: L_{(f,i)} \rightarrow L_\zeta$ is a

821:   quasi-isomorphism of complexes.

822: \end{lem}

823:  

824: \begin{proof} 

825:   This is an easy local argument which is left as an exercise for the

826:   reader.

827: \end{proof}

828: 

829: As a corollary of the lemma, we see that $L_\zeta^\vee$ is represented

830: in $D(\OO_{B})$ by the complex $L_{(f,i)}^\vee$ defined to be

831: \begin{equation}

832: \begin{CD}

833: 0 & & 1 \\

834: i^* T_S @> (\beta_i)^\vee >> N_{(f,i)}

835: \end{CD}

836: \end{equation}

837: 

838: \

839: 

840: Now suppose that $p\in B \subset S$ is a point, define

841: $g:S'\rightarrow S$ to be the blow up of $S$ at $p$ with exceptional

842: divisor $E$, and define $i':B' \rightarrow S'$ to be the reduced total

843: transform of $B$.

844: 

845: \

846: 

847: We break up our analysis according to the type of behavior of $p\in

848: B$.  If $p\in B$ is a smooth point, we call this \emph{type (I)}.  If

849: $p\in B$ is a node, we call this \emph{type (II)}.  We further

850: decompose each type as follows.  If $p\in B$ is a smooth point which

851: lies on a stable component, we say this is \emph{type (Ia)}.  If $p\in

852: B$ is a smooth point which lies on an unstable component, we say this

853: is \emph{type (Ib)}.  If $p\in B$ is a node, and the first order

854: deformations of $\zeta$ smooth the node, we say this is \emph{type

855:   (IIa)}.  If the first order deformations of $\zeta$ don't smooth the

856: node, we say this is \emph{type (IIb)}.

857: 

858: \

859: 

860: For type (I), we have $g^* B$ equals $B'$ as Cartier divisors.  For

861: type (II), we have $g^* B$ equals $B' + E$ as Cartier divisors.  For

862: both types, we define $h:B' \rightarrow B$ to be the restriction of

863: $g$ and we define $f':B' \rightarrow B$ to be $f' = f\circ h$.

864: 

865: \

866: 

867: Of course we have $g_*\OO_{S'} = \OO_S$ and $R^{k>0}g_* \OO_{S'}$ is

868: zero.  For type (I), we have $\OO_{S'}(-B') = g^* \OO_S(-B)$.  So by

869: the projection formula we have $g_*\OO_{S'}(-B') = \OO_S(-B)$ and

870: $R^{k>0}g_*\OO_{S'}(-B')$ is zero.  Also we have that $g_*\OO_{S'}(E)

871: = \OO_S$ and $R^{k>0}g_* \OO_{S'}(E)$ is zero.  So also for type (II)

872: we have $g_*\OO_{S'}(-B') = \OO_S(-B)$ and $R^{k>0}g_* \OO_{S'}(-B')$

873: is zero.

874: 

875: \

876: 

877: Using the resolution of $\OO_{B'}$

878: \begin{equation}

879: \begin{CD}

880: 0 @>>> \OO_{S'}(-B') @>>> \OO_{S'} @>>> \OO_{B'} @>>> 0

881: \end{CD}

882: \end{equation}

883: we conclude that $h_* \OO_{B'} = \OO_B$ and $R^{k>0}h_* \OO_{B'}$ is

884: zero.  In other words, the canonical morphism $\OO_B \rightarrow

885: \mathbb{R}h_* \OO_{B'}$ is a quasi-isomorphism.  From this it follows

886: by the projection formula that the canonical morphism

887: \begin{equation}

888: L_{(f,i)}^\vee \rightarrow 

889: \mathbb{R}h_* \mathbb{L}h^* (L_{(f,i)}^\vee)

890: \end{equation}

891: is a quasi-isomorphism.  Therefore the pullback morphisms

892: \begin{equation}

893: \mathbb{H}^k(B,L_\zeta^\vee) \rightarrow

894: \mathbb{H}^k(B',g^*L_\zeta^\vee)

895: \end{equation}

896: are isomorphisms.

897: 

898: \

899: 

900: Now there is a canonical morphism $\OO_{S'}(B') \rightarrow g^*

901: \OO_S(B)$.  For type (I), this morphism is an isomorphism.  For type

902: (II), this morphism is injective and the cokernel is $g^*\OO_S(B)|_E$,

903: i.e. $\OO_E \otimes_\CC M$ where $M= \OO_S(B)|_p$ is a one-dimensional

904: vector space.  For type (I), we conclude that the canonical morphism

905: $N_{(f',i')} \rightarrow h^* N_{(f,i)}$ is an isomorphism.  For type

906: (II), we conclude that there is an exact sequence:

907: \begin{equation}

908: 0 \rightarrow \textit{Tor}_1^{\OO_{S'}}(\OO_{B'},\OO_E)\otimes_\CC M

909: \rightarrow N_{(f',i')} \rightarrow h^* N_{(f,i)} \rightarrow \OO_E

910: \otimes_\CC M \rightarrow 0

911: \end{equation}

912: For both types, denote by $\Gamma \subset B'$ the subcurve which is

913: the union of all components other than $E$, and let $D = E\cap

914: \Gamma$.  From the resolution of $\OO_E$

915: \begin{equation}

916: \begin{CD}

917: 0 @>>> \OO_{S'}(-E) @>>> \OO_S @>>> \OO_E @>>> 0

918: \end{CD}

919: \end{equation}

920: we have the relation

921: \begin{equation}

922: \textit{Tor}_1^{\OO_{S'}}(\OO_{B'},\OO_E) = {\mc

923:   I}_{\Gamma}\otimes_{\OO_{S'}}\OO_{S'}(-E) = \OO_E(-E)(-D).

924: \end{equation}

925: where ${\mc I}_{\Gamma}$ is the ideal sheaf of $\OO_{B'}$ defining

926: $\Gamma \subset B'$, i.e. $\OO_E(-D)$.  So in case $(2)$, we have an

927: exact sequence:

928: \begin{equation}

929: 0 \rightarrow \OO_E(-E)(-D)\otimes_\CC M \rightarrow N_{(f',i')}

930: \rightarrow h^*N_{(f,i)} \rightarrow \OO_E \otimes_\CC M \rightarrow

931: 0.

932: \end{equation}

933: 

934: \

935: 

936: For both types, we have a short exact sequence of $\OO_{S'}$-modules

937: \begin{equation}

938: \begin{CD}

939: 0 @>>> g^* \Omega_S @> (dg)^\dagger >> \Omega_{S'} @>>> \Omega_E @>>>

940: 0.

941: \end{CD}

942: \end{equation}

943: Applying $\mathbb{R}\textit{Hom}_{\OO_{S'}}(*,\OO_{S'})$, we have an

944: exact sequence

945: \begin{equation}

946: \begin{CD}

947: 0 @>>> T_{S'} @>>> g^* T_S @>>>

948: \textit{Ext}^1_{\OO_{S'}}(\OO_E,\OO_{S'}) @>>> 0 

949: \end{CD}

950: \end{equation}

951: Using the resolution of $\OO_E$ in the last paragraph, we have that

952: \begin{equation}

953: \textit{Ext}^1_{\OO_{S'}}(\OO_E,\OO_{S'}) = \OO_E(E).

954: \end{equation}

955: So we have an exact sequence

956: \begin{equation}

957: \begin{CD}

958: 0 @>>> T_{S'} @>>> g^* T_S @>>> T_E(E) @>>> 0

959: \end{CD}

960: \end{equation}

961: Using our \emph{Tor} result from the last paragraph, we have a short

962: exact sequence

963: \begin{equation}

964: 0 \rightarrow T_E(-D) \rightarrow (i')^*T_{S'} \xrightarrow{dg} h^*

965: i^* T_S \rightarrow T_E(E) \rightarrow 0.

966: \end{equation}

967: Notice this holds in both cases.

968: 

969: \

970: 

971: The maps $N_{(f',i')} \rightarrow h^* N_{(f,i)}$ and $(i')^*T_{S'}

972: \rightarrow h^* i^* T_S$ considered in the last two paragraphs are

973: compatible with $\alpha_{i'}$ and $h^*\alpha_i$.  So we have an

974: induced map of complexes $dg: L_{\zeta'}^\vee \rightarrow h^*

975: L_{\zeta}^\vee$.  Define $I\hookrightarrow h^* L_\zeta^\vee$ to be the

976: image complex of $dg:L_{\zeta'}^\vee \rightarrow h^* L_{\zeta}^\vee$.

977: For type (I), we define two complexes of coherent sheaves on $B'$,

978: $K_I$ and $Q_I$, by $K_I = T_E(-D)[0]$ and $Q_I = T_E(E)[0]$.  For

979: type (II), we define complexes of coherent sheaves on $B'$, $K_{II}$

980: and $Q_{II}$ by

981: \begin{eqnarray}

982: K_{II} = T_E(-D)[0] \oplus\lt( \OO_E(-E)(-D)\otimes_\CC M\rt)[-1] \\

983: Q_{II} = T_E(E)[0] \oplus \lt(\OO_E \otimes_\CC M\rt)[-1]

984: \end{eqnarray}

985: Then we have exact sequences of complexes

986: \begin{equation}

987: \begin{CD}

988: 0 @>>> K @>>> L_{\zeta'}^\vee @> dg >> I @>>> 0 \\

989: 0 @>>> I @>>> h^* L_{\zeta}^\vee @>>> Q @>>> 0

990: \end{CD}

991: \end{equation}

992: 

993: \

994: 

995: For type (I), we have that $\mathbb{H}^0(B',K_I) = H^0(E,T_E(-D))$ is

996: $2$-dimensional, because $T_E(-D) \cong \OO_E(1)$.  And $\mathbb{H}^{k

997:   > 0}(B',K_1)$ is zero.  Similarly $\mathbb{H}^0(B',Q_I) =

998: H^0(E,T_E(E))$ is $2$-dimensional and $\mathbb{H}^{k>0}(B',Q_I)$ is

999: zero.  Therefore we have a long exact sequence of hypercohomology

1000: groups:

1001: \begin{eqnarray*}

1002: 0 \rightarrow H^0(E,T_E(-D)) \rightarrow

1003: \mathbb{H}^0(B',L_{(f',i')}^\vee) \rightarrow 

1004: \mathbb{H}^0(B, L_{(f,i)}^\vee) \rightarrow \dots \\ 

1005: \dots  \rightarrow  H^0(E, T_E(E)) \rightarrow

1006: \mathbb{H}^1(B',L_{(f',i')}^\vee) \rightarrow

1007:  \mathbb{H}^1(B, L_{(f,i)}^\vee ) \rightarrow 0 \\

1008:   0 \rightarrow \mathbb{H}^2(B',L_{(f',i')}^\vee) \rightarrow \mathbb{H}^2(B,

1009:  L_{(f,i)}^\vee ) \rightarrow 0

1010: \end{eqnarray*}

1011: 

1012: \

1013: 

1014: For type (Ia), the map

1015: \begin{equation}

1016: \mathbb{H}^0(B,L_{(f,i)}^\vee) \rightarrow H^0(E,T_E(E))

1017: \end{equation}

1018: is the zero map.  So we have proved the following lemma.

1019: 

1020: \begin{lem} \label{lem-def1a}

1021:   Suppose that $h:B'\rightarrow B$ is type (Ia).  Then we have exact

1022:   sequences:

1023: \begin{eqnarray}

1024: 0 \rightarrow H^0(E,T_E(-D)) \rightarrow

1025: \mathbb{H}^0(B',L_{(f',i')}^\vee) \rightarrow 

1026: \mathbb{H}^0(B, L_{(f,i)}^\vee) \rightarrow 0 \\ 

1027: 0 \rightarrow  H^0(E, T_E(E)) \rightarrow

1028: \mathbb{H}^1(B',L_{(f',i')}^\vee) \rightarrow

1029:  \mathbb{H}^1(B, L_{(f,i)}^\vee ) \rightarrow 0 \\

1030:   0 \rightarrow \mathbb{H}^2(B',L_{(f',i')}^\vee) \rightarrow \mathbb{H}^2(B,

1031:  L_{(f,i)}^\vee ) \rightarrow 0

1032: \end{eqnarray}

1033: So the canonical map from the Lie algebra of infinitesimal

1034: automorphisms of $\zeta'$ to the Lie algebra of infinitesimal

1035: automorphisms of $\zeta$ is surjective with $2$-dimensional kernel,

1036: the canonical map from the space of first order deformations of

1037: $\zeta'$ to the space of first order deformations of $\zeta$ is

1038: surjective with $2$-dimensional kernel, and the obstruction space of

1039: $\zeta'$ equals the obstruction space of $\zeta$.

1040: \end{lem}

1041: 

1042: \

1043: 

1044: For type (Ib), the map

1045: \begin{equation}

1046: \mathbb{H}^0(B,L_{(f,i)}^\vee) \rightarrow H^0(E,T_E(E))

1047: \end{equation}

1048: has a $1$-dimensional image; we will call it $N$.  We have proved the

1049: following lemma.

1050: 

1051: \begin{lem} \label{lem-def1b}

1052:   Suppose that $h:B'\rightarrow B$ is type (Ib).  Then there is a

1053:   $1$-dimensional subspace $N\subset H^0(E,T_E(E))$ such that we have

1054:   exact sequences:

1055: \begin{eqnarray}

1056: 0 \rightarrow H^0(E,T_E(-D)) \rightarrow

1057: \mathbb{H}^0(B',L_{(f',i')}^\vee) \rightarrow 

1058: \mathbb{H}^0(B, L_{(f,i)}^\vee) \rightarrow N \rightarrow 0 \\ 

1059: 0 \rightarrow  H^0(E, T_E(E))/N \rightarrow

1060: \mathbb{H}^1(B',L_{(f',i')}^\vee) \rightarrow

1061:  \mathbb{H}^1(B, L_{(f,i)}^\vee ) \rightarrow 0 \\

1062:   0 \rightarrow \mathbb{H}^2(B',L_{(f',i')}^\vee) \rightarrow 

1063: \mathbb{H}^2(B, L_{(f,i)}^\vee ) \rightarrow 0

1064: \end{eqnarray}

1065: So the canonical map from the Lie algebra of infinitesimal

1066: automorphisms of $\zeta'$ to the Lie algebra of infinitesimal

1067: automorphisms of $\zeta$ has both $1$-dimensional kernel and cokernel,

1068: the canonical map from the space of first order deformations of

1069: $\zeta'$ to the space of first order deformations of $\zeta$ is

1070: surjective with $1$-dimensional kernel, and the obstruction space of

1071: $\zeta'$ equals the obstruction space of $\zeta$.

1072: \end{lem}

1073: 

1074: \

1075: 

1076: Next we consider type (II).  Then $\OO_E(-E)(-D)$ is isomorphic to

1077: $\OO_E(-1)$.  Since $H^0(E,\OO_E(-1)) = H^1(E,\OO_E(-1)) = 0$, the

1078: terms $\OO_E(-E)(-D)\otimes_\CC M[-1]$ do not contribute to the

1079: hypercohomology of $K_{II}$.  And $T_E(-D)$ is isomorphic to $\OO_E$.

1080: So we have $\mathbb{H}^0(B',K_{II}) = H^0(E,T_E(-D))$ is

1081: $1$-dimensional and $\mathbb{H}^{k>0}(B',K_{II})$ is zero.

1082: 

1083: \

1084: 

1085: For $Q_{II}$ both terms contribute to the cohomology.  We have

1086: $T_E(E)$ is isomorphic to $\OO_E(1)$ so that $\mathbb{H}^0(B',Q_{II})

1087: = H^0(E,T_E(E))$ is $2$-dimensional, and $\mathbb{H}^1(B',Q_{II}) = M

1088: = \OO_S(B)|_p$ is $1$-dimensional.  Therefore we have a long exact

1089: sequence in hypercohomology:

1090: \begin{eqnarray*}

1091: 0 \rightarrow H^0(E,T_E(-D)) \rightarrow

1092: \mathbb{H}^0(B',L_{(f',i')}^\vee) \rightarrow

1093: \mathbb{H}^0(B,L_{(f,i)}^\vee) \rightarrow \dots \\

1094: \dots \rightarrow H^0(E,T_E(E)) \rightarrow

1095: \mathbb{H}^1(B',L_{(f',i')}^\vee) \rightarrow

1096: \mathbb{H}^1(B,L_{(f,i)}^\vee) \rightarrow \dots \\

1097: \dots \rightarrow M \rightarrow \mathbb{H}^2(B',L_{(f',i')}^\vee)

1098: \rightarrow \mathbb{H}^2(B,L_{(f,i)}^\vee) \rightarrow 0

1099: \end{eqnarray*}

1100: 

1101: \

1102: 

1103: Geometrically, every infinitesimal automorphism of $\zeta$ lifts to an

1104: infinitesimal automorphism of $\zeta'$, so that

1105: $\mathbb{H}^0(B,L_{(f,i)}^\vee)\rightarrow H^0(E,T_E(E))$ is the zero

1106: map.  Similarly, the map $\mathbb{H}^1(B,L_{(f,i)}^\vee) \rightarrow

1107: \OO_S(B)|_p$ is nonzero iff there are deformations of $\zeta$ which

1108: smooth the node $p$ to first order.  So we have the following lemma:

1109: 

1110: \begin{lem} \label{lem-def2a}

1111:   Suppose that $h:B'\rightarrow B$ is type (IIa).  Then we have exact

1112:   sequences:

1113: \begin{eqnarray}

1114: 0 \rightarrow H^0(E,T_E(-D)) \rightarrow

1115: \mathbb{H}^0(B',L_{(f',i')}^\vee) \rightarrow

1116: \mathbb{H}^0(B,L_{(f,i)}^\vee) \rightarrow 0 \\

1117: 0 \rightarrow H^0(E,T_E(E)) \rightarrow

1118: \mathbb{H}^1(B',L_{(f',i')}^\vee) \rightarrow

1119: \mathbb{H}^1(B,L_{(f,i)}^\vee) \rightarrow \OO_S(B)|_p \rightarrow 0 \\

1120: 0 \rightarrow

1121: \mathbb{H}^2(B',L_{(f',i')}^\vee) \rightarrow \mathbb{H}^2(B,

1122: L_{(f,i)}^\vee) \rightarrow 0 

1123: \end{eqnarray}

1124: So the canonical map from the space of first order deformations of

1125: $\zeta'$ to the space of first order deformations of $\zeta$ has both

1126: $1$-dimensional kernel and cokernel, and the obstruction space to

1127: $\zeta'$ equals the obstruction space to $\zeta$.

1128: \end{lem}

1129: 

1130: \begin{lem} \label{lem-def2b}

1131:   Suppose that $h:B'\rightarrow B$ is type (IIb).  Then we have exact

1132:   sequences:

1133: \begin{eqnarray}

1134: 0 \rightarrow H^0(E,T_E(-D)) \rightarrow

1135: \mathbb{H}^0(B',L_{(f',i')}^\vee) \rightarrow

1136: \mathbb{H}^0(B,L_{(f,i)}^\vee) \rightarrow 0 \\

1137: 0 \rightarrow H^0(E,T_E(E)) \rightarrow

1138: \mathbb{H}^1(B',L_{(f',i')}^\vee) \rightarrow

1139: \mathbb{H}^1(B,L_{(f,i)}^\vee) \rightarrow 0 \\

1140: 0 \rightarrow \OO_S(B)|_p \rightarrow

1141: \mathbb{H}^2(B',L_{(f',i')}^\vee) \rightarrow \mathbb{H}^2(B,

1142: L_{(f,i)}^\vee) \rightarrow 0 

1143: \end{eqnarray}

1144: So the canonical map from the space of first order deformations of

1145: $\zeta'$ to the space of first order deformations of $\zeta$ is

1146: surjective with $1$-dimensional kernel, and the map from the

1147: obstruction space of $\zeta'$ to the obstruction space of $\zeta$ is

1148: surjective and has a $1$-dimensional kernel.

1149: \end{lem}

1150: 

1151: \

1152: 

1153: Finally, we consider the case when $h:B'\rightarrow B$ is the identity

1154: map, but there is one extra marked point $q\in B'$ which is not in

1155: $B$; we call this \emph{type (III)}.  We further break this up as

1156: follows.  If $q\in B$ lies on an unstable component, we call this

1157: \emph{type (IIIa)}.  If $q\in B$ lies on a stable component, we call

1158: this \emph{type (IIIb)}.

1159: 

1160: \

1161:   

1162: For type (III), there is a canonical short exact sequence of

1163: complexes:

1164: \begin{equation}

1165: \begin{CD}

1166: 0 @>>> L_\zeta @>>> L_\zeta' @>>> \Omega_B(q)|_q[0] @>>> 0

1167: \end{CD}

1168: \end{equation}

1169: Of course we have:

1170: \begin{eqnarray}

1171: \EExt{B}^0(\Omega_B(q)|_q[0],\OO_B) = 0, \\

1172: \EExt{B}^1(\Omega_B(q)|_q[0],\OO_B) = T_B|_q, \\

1173: \EExt{B}^2(\Omega_B(q)|_q[0],\OO_B) = 0.

1174: \end{eqnarray}

1175: The induced map $\mathbb{E}\text{xt}^0_{\OO_B}(L_\zeta,\OO_B)

1176: \rightarrow T_B|_q$ is nonzero iff $q$ lies on an unstable component

1177: of $\zeta$.  Thus we have the following lemma:

1178: 

1179: \begin{lem} \label{lem-def3a}

1180:   Suppose that $h:B'\rightarrow B$ is type (IIIa).  Then we have exact

1181:   sequences:

1182: \begin{eqnarray}

1183: 0 \rightarrow \EExt{B'}^0(L_{\zeta'},\OO_{B'}) \rightarrow

1184: \EExt{B}^0(L_\zeta,\OO_B) \rightarrow T_B|_q \rightarrow 0 \\

1185: 0 \rightarrow \EExt{B'}^1(L_{\zeta'},\OO_{B'}) \rightarrow

1186: \EExt{B}^1(L_\zeta,\OO_B) \rightarrow 0 \\

1187: 0 \rightarrow \EExt{B'}^2(L_{\zeta'},\OO_{B'}) \rightarrow

1188: \EExt{B}^2(L_\zeta,\OO_B) \rightarrow 0 

1189: \end{eqnarray}

1190: So the Lie algebra of infinitesimal automorphisms of $\zeta'$ has

1191: codimension $1$ in the Lie algebra of infinitesimal automorphisms of

1192: $\zeta$, the space of first order deformations of $\zeta'$ equals the

1193: space of first order deformations of $\zeta$, and the obstruction

1194: space of $\zeta'$ equals the obstruction space of $\zeta$.

1195: \end{lem}

1196: 

1197: \begin{lem} \label{lem-def3b}

1198:   Suppose that $h:B'\rightarrow B$ is type (IIIb).  Then we have exact

1199:   sequences:

1200: \begin{eqnarray}

1201: 0 \rightarrow \EExt{B'}^0(L_{\zeta'},\OO_{B'}) \rightarrow

1202: \EExt{B}^0(L_\zeta,\OO_B) \rightarrow 0 \\

1203: 0 \rightarrow T_B|_q \rightarrow \EExt{B'}^1(L_{\zeta'},\OO_{B'}) 

1204: \rightarrow \EExt{B}^1(L_\zeta,\OO_B) \rightarrow 0 \\

1205: 0 \rightarrow \EExt{B'}^2(L_{\zeta'},\OO_{B'}) \rightarrow

1206: \EExt{B}^2(L_\zeta,\OO_B) \rightarrow 0 

1207: \end{eqnarray}

1208: So the Lie algebra of infinitesimal automorphisms of $\zeta'$ equals

1209: the Lie algebra of infinitesimal automorphisms of $\zeta$, the

1210: canonical map from the space of first order deformations of $\zeta'$

1211: to the space of first order deformations of $\zeta$ is surjective with

1212: $1$-dimensional kernel, and the obstruction space of $\zeta'$ equals

1213: the obstruction space of $\zeta'$.

1214: \end{lem}

1215: 

1216: Combining lemma~\ref{lem-def1a} through lemma~\ref{lem-def3b}, one can

1217: analyze the associated maps of vector spaces

1218: $\EExt{B'}^k(L_{\zeta'},\OO_{B'}) \rightarrow

1219: \EExt{B}^k(L_\zeta,\OO_B)$ for any morphism $h:B'\rightarrow B$ which

1220: removes some subset of marked points from $B'$ and then contracts some

1221: subset of the unstable components.

1222: 

1223: \subsection{Gluing stable curves}\label{subsec-glue}

1224: 

1225: Just as one has an obstruction theory for $\Kgnb{g,r}{X,\beta}$ of the

1226: form $\lt( {\mathbb R}\pi_*\lt(L_\zeta^\vee\rt)[1]\rt)^\vee$, also for

1227: any stable $A$-graph $\tau$, one has an analogous obstruction theory

1228: for each of the Behrend-Manin stacks $\Kbm(X,\tau)$ (c.f.~\cite{BM}

1229: for the definition of $\Kbm(X,\tau)$).  We will not describe this

1230: obstruction theory here.  In \cite{B}, a \emph{relative obstruction

1231:   theory} for the morphism $\Kbm(X,\tau) \rightarrow {\mathfrak

1232:   M}(\tau)$ is given from which an absolute obstruction theory for

1233: $\Kbm(X,\tau)$ can be deduced.

1234: 

1235: \

1236: 

1237: Suppose that $\tau$ is a stable $A$-graph, and suppose that

1238: $\{f_1,f_2\}$ is a disconnecting edge of $\tau$.  Let $\tau_1\subset

1239: \tau$ be the maximal connected subgraph which contains $f_1$ and not

1240: $f_2$, and let $\tau_2\subset \tau$ be the maximal connected subgraph

1241: which contains $f_2$ and not $f_1$.  So we have forgetful

1242: $1$-morphisms $\Kbm(X,\tau) \rightarrow \Kbm(X,\tau_i)$ for $i=1,2$.

1243: 

1244: \begin{lem}\label{lem-glue1}

1245:   Suppose that $\zeta:T \rightarrow \Kbm(X,\tau)$ is a $1$-morphism

1246:   and let $\zeta_i:T \rightarrow \Kbm(X,\tau_i)$, $i=1,2$ be the

1247:   composition of $\zeta$ with the forgetful $1$-morphism above.

1248:   Suppose that for each point $t\in T$, the stable maps $\zeta_1(t)\in

1249:   \Kbm(X,\tau_1)$ and $\zeta_2(t) \in \Kbm(X,\tau_2)$ are unobstructed

1250:   (in the sense that the obstruction groups described above are zero)

1251:   and the evaluation morphism $\text{ev}_{f_1}:\Kbm(X,\tau_1)

1252:   \rightarrow X$ is smooth at $\zeta_1(t)$.  Let $T_{\text{ev}_{f_1}}$

1253:   denote the dual of the sheaf of relative differentials of

1254:   $\text{ev}_{f_1}$.  Then also $\zeta(t)\in \Kbm(X,\tau)$ is

1255:   unobstructed, and there is a short exact sequence:

1256: \begin{equation}

1257: \begin{CD}

1258: 0 @>>> \zeta_1^*T_{\text{ev}_{f_1}} @>>> \zeta^* T_{\Kbm(X,\tau)} @>>>

1259: \zeta_2^* T_{\Kbm(X,\tau_2)} @>>> 0

1260: \end{CD}

1261: \end{equation}

1262: \end{lem}

1263: 

1264: \begin{proof}

1265:   The proof essentially follows from the fact that $\Kbm(X,\tau)$ is

1266:   an open substack of the $2$-fiber product:

1267: \begin{equation}

1268: \Kbm(X,\tau_1) \times_{\text{ev}_{f_1},X,\text{ev}_{f_2}}

1269: \Kbm(X,\tau_2).

1270: \end{equation}

1271: The details are left to the reader.

1272: \end{proof}

1273: 

1274: Now suppose that $\phi:\tau \rightarrow \sigma$ is the contraction of

1275: stable $A$-graphs which contracts the edge $\{f_1,f_2\}$.  The induced

1276: $1$-morphism $\Kbm(X,\tau) \rightarrow \Kbm(X,\sigma)$ is unramified

1277: of codimension at most $1$.  In some circumstances, it is the

1278: normalization of a Cartier divisor.

1279: 

1280: \begin{lem}\label{lem-glue2}

1281:   With the same notation as in lemma~\ref{lem-glue1}, suppose that

1282:   $\tau$ is a genus $0$ tree.  For $i=1,2$, let the domain of

1283:   $\zeta_i$ be given by $\pi_i:C_i \rightarrow T$, let $g_i:C_i

1284:   \rightarrow X$ be the map of $\zeta_i$, and let $s_i:T \rightarrow

1285:   C_i$ be the section corresponding to the flag $f_i$ of $\tau_i$.

1286:   Denote by $T_{\pi_i}$ the dual of the sheaf of relative

1287:   differentials of $\pi_i$.

1288:   

1289:   Suppose that for every point $t\in T$, $\Kbm(X,\tau_2)$ is

1290:   unobstructed at $\zeta_2(t)$, and suppose that $g_1^*T_X$ is

1291:   $\pi_1$-relatively generated by global sections.  Then for each

1292:   point $t\in T$, $\Kbm(X,\sigma)$ is unobstructed at $\zeta(t)$, the

1293:   morphism $\Kbm(X,\tau) \rightarrow \Kbm(X,\sigma)$ is a regular

1294:   embedding of codimension $1$ at $\zeta(t)$, and we have a short

1295:   exact sequence:

1296: \begin{equation}

1297: \begin{CD}

1298: 0 @>>> \zeta^*T_{\Kbm(X,\tau)} @>>> \zeta^* T_{\Kbm(X,\sigma)} @>>>

1299: s_1^*T_{\pi_1} \otimes s_2^* T_{\pi_2} @>>> 0

1300: \end{CD}

1301: \end{equation}

1302: \end{lem}

1303: 

1304: \begin{proof}

1305:   Let $\pi:C\rightarrow T$ be the family of curves obtained by

1306:   identifying the section $s_1$ of $\pi_1$ and the section $s_2$ of

1307:   $\pi_2$.  Let $s:T \rightarrow C$ be the section corresponding to

1308:   $s_1$ and $s_2$.  Let $g:C \rightarrow X$ be the map obtained by

1309:   gluing $g_1$ and $g_2$.

1310: 

1311: \

1312: 

1313: First we prove that for any point $t\in T$ we have that

1314: $\Kbm(X,\sigma)$ is unobstructed at $\zeta(t)$ and the morphism

1315: $\Kbm(X,\tau) \rightarrow \Kbm(X,\sigma)$ is a regular embedding of

1316: codimension $1$.  The two statements together are equivalent to the

1317: statement that there are first order deformations of the map

1318: $\zeta(t)\in \Kbm(X,\sigma)$ which smooth the node $s(t)\in C_t$.

1319: 

1320: \

1321: 

1322: Now there is an exact sequence for $\zeta(t)\in \Kbm(X,\sigma)$:

1323: \begin{equation}

1324: T_{\Kbm(X,\sigma)}|_{\zeta(t)} \rightarrow \text{Ext}^1\lt(

1325: \Omega_{C_t},\OO_{C_t} \rt) \rightarrow H^1\lt( C_t,g^*T_X \rt) \rightarrow

1326: \text{Obs}(\zeta(t))  

1327: \rightarrow 0

1328: \end{equation}

1329: Here $\text{Obs}(\zeta)$ is the obstruction group to $\Kbm(X,\sigma)$

1330: at $\zeta$.  Similarly, we have an exact sequence for $\zeta_2(t)$:

1331: \begin{equation}

1332: \begin{CD}

1333: T_{\Kbm(X,\tau_2)}|_{\zeta_2(t)} @>>> \text{Ext}^1\lt(

1334: \Omega_{(C_2)_t} \rt) @>>> H^1\lt( (C_2)_t, g_2^*T_X \rt) @>>> 0

1335: \end{CD}

1336: \end{equation}

1337: 

1338: \

1339: 

1340: We have a short exact sequence of sheaves on $C_t$:

1341: \begin{equation}

1342: \begin{CD}

1343:   0 @>>> g_1^*T_X\lt(-s(t)\rt) @>>> g^*T_X @>>> g_2^*T_X @>>> 0

1344: \end{CD}

1345: \end{equation}

1346: By assumption, $g_1^*T_X$ is generated by global sections.  Thus by

1347: lemma~\ref{lem-gend}, we conclude that $H^1\lt(C_t, g_1^*T_X\lt(-s(t)

1348: \rt) \rt)$ is zero.  Thus we have an identification of $H^1\lt(

1349: C_t,g^*T_X \rt)$ and $H^1\lt( (C_2)_t, g_2^*T_X \rt)$.  Also,

1350: $\text{Ext}^1(\Omega_{C_t},\OO_{C_t})$ is canonically isomorphic to

1351: the product over all nodes $q\in C_t$ of $T'_q\otimes T''_q$ where

1352: $T'_q$ and $T''_q$ are the tangent spaces to the two branches of $C_t$

1353: at $q$.  We have the analogous result for $(C_2)_t$.  Via these

1354: identifications, we have a commutative diagram:

1355: \begin{equation}

1356: \begin{CD}

1357: \text{Ext}^1\lt( \Omega_{C_t},\OO_{C_t} \rt) @>>> H^1\lt( C_t, g^* T_X

1358: \rt) \\

1359: @V p VV @VV = V \\

1360: \text{Ext}^1\lt( \Omega_{(C_2)_t}, \OO_{(C_2)_t} \rt) @>>> H^1\lt(

1361: (C_2)_t, g_2^* T_X \rt)

1362: \end{CD}

1363: \end{equation}

1364: where $p$ is the canonical projection.  Choose any section

1365: \begin{equation}

1366: \phi:T'_{s(t)}\otimes T''_{s(t)} \rightarrow \text{Ext}^1\lt(

1367: \Omega_{C_t}, \OO_{C_t} \rt)

1368: \end{equation}

1369: and any section

1370: \begin{equation}

1371: \psi:\text{Ext}^1\lt( \Omega_{(C_2)_t}, \OO_{(C_2)_t} \rt) \rightarrow

1372: \text{Ext}^1\lt( \Omega_{C_t},\OO_{C_t} \rt)

1373: \end{equation}

1374: of the canonical projections.  Choose any element $u\in

1375: T'_{s(t)}\otimes T''_{s(t)}$ and consider the image

1376: $\overline{\phi(u)}$ of $\phi(u)$ in $H^1\lt( C_t, g^* T_X \rt)$.

1377: Since the map

1378: \begin{equation}

1379: \begin{CD}

1380: \text{Ext}^1\lt( \Omega_{(C_2)_t}, \OO_{(C_2)_t} \rt) @>>> H^1\lt(

1381: (C_2)_t, g_2^* T_X \rt)

1382: \end{CD}

1383: \end{equation}

1384: is surjective, we can find some element $v\in \text{Ext}^1\lt(

1385: \Omega_{(C_2)_t}, \OO_{(C_2)_t} \rt)$ such that $\overline{\psi(v)}$

1386: equals $\overline{\phi(u)}$.  Consider $\phi(u)-\psi(v)$.  This has

1387: image $0$ in $H^1\lt( C_t, g^*T_X \rt)$.  Therefore we conclude that

1388: it is in the image of $T_{\Kbm(X,\sigma)}|_{\zeta(t)}$.  So we

1389: conclude that $u$ is in the image of the projection map

1390: $T_{\Kbm(X,\sigma)} \rightarrow T'_{s(t)}\otimes T''_{s(t)}$, i.e.

1391: this projection map is surjective.  So the deformations of $\zeta(t)$

1392: smooth the node $s(t)$.  Therefore $\Kbm(X,\sigma)$ is smooth at

1393: $\zeta(t)$ and the morphism $\Kbm(X,\tau) \rightarrow \Kbm(X,\sigma)$

1394: is a regular embedding of codimension $1$ at $\zeta(t)$.

1395: 

1396: \

1397: 

1398: Finally, the short exact sequence above is just the globalized version

1399: of the projection map $T_{\Kbm(X,\sigma} \rightarrow T'_{s(t)}\otimes

1400: T''_{s(t)}$ appearing in the last paragraph.

1401: \end{proof}

1402: 

1403: \section{Properties of families of stable maps} \label{sec-props}

1404: 

1405: In this section we introduce some definitions and lemmas regarding

1406: properties of families of stable maps.  Recall, to prove that

1407: $\Kgnb{0,0}{X,e}$ is rationally connected, we have to find a

1408: \emph{very free} $1$-morphism $\zeta:\PP^1 \rightarrow

1409: \Kgnb{0,0}{X,e}$, i.e. a $1$-morphism whose image is contained in the

1410: smooth locus and such that $\zeta^*T_{\Kgnb{0,0}{X,e}}$ is an ample

1411: vector bundle.  Our proof that such a $1$-morphism exists is by

1412: induction, where the induction step consists of attaching a

1413: $1$-parameter family of lines to our $1$-parameter family of degree

1414: $e$ stable maps.  To make the induction argument work, we need a bit

1415: more than a very free $1$-morphism $\zeta:\PP^1 \rightarrow

1416: \Kgnb{0,0}{X,e}$.  The property we need is what we call a \emph{very

1417:   positive} $1$-morphism.  Additionally, we need that our

1418: $1$-parameter family of lines has a property which we call \emph{very

1419:   twisting}.  Finally, because of an operation we perform on

1420: $1$-morphisms which we call \emph{modification}, and which we

1421: introduce in the next section, we need to consider the case of

1422: $1$-morphisms with reducible domain, i.e. $\zeta:B \rightarrow

1423: \Kgnb{0,1}{X,e}$ where $B$ is a connected, prestable curve of

1424: arithmetic genus $0$.  This is the level of generality in which we

1425: make all our definitions.

1426: 

1427: \begin{defn}  Given a genus $0$ stable map

1428:   $\zeta = ((B,p_1,\dots,p_n),f:B\rightarrow X)$, we say $\zeta$ is

1429:   \emph{very stable} if the unmarked prestable map $(B,f:B\rightarrow

1430:   X)$ is stable.

1431: \end{defn}

1432: 

1433: \begin{notat} \label{not-1mor}

1434:   Given a closed subscheme $X\subset \PP^n$ and a scheme $B$, a

1435:   $1$-morphism $\zeta:B\rightarrow \Kgnb{g,r}{X,e}$ is equivalent to a

1436:   datum:

1437: \begin{equation}

1438: \zeta = \lt( \lt(p_\zeta:\Sigma_\zeta \rightarrow B,

1439: \sigma_{\zeta,1},\dots, \sigma_{\zeta,r} \rt), g_\zeta \rt).

1440: \end{equation}

1441: Here $p_\zeta:\Sigma_\zeta\rightarrow B$ is a family of prestable

1442: curves, $\sigma_{\zeta,i}:B\rightarrow \Sigma_\zeta$ is a collection

1443: of sections, $g_\zeta:\Sigma_\zeta\rightarrow X$ is a morphism of

1444: schemes and we denote $h_{\zeta,i}=g_\zeta\circ \sigma_{\zeta,i}$.

1445: When there is no risk of confusion, we will suppress the $\zeta$

1446: subscripts.

1447: \end{notat}

1448: 

1449: \begin{defn}~\label{defn-twisting}

1450:   Suppose $\pi:B\rightarrow T$ is a family of prestable, geometrically

1451:   connected curves of arithmetic genus $0$.  Suppose given a

1452:   $1$-morphism $\zeta:B \rightarrow \Kgnb{0,1}{X,1}$, i.e. a datum

1453: \begin{equation}

1454: \zeta = \lt( p:\Sigma \rightarrow B, \sigma: B \rightarrow \Sigma,

1455: g:\Sigma \rightarrow X \rt)

1456: \end{equation}

1457: such that $X$ is smooth along $g(\Sigma)$.  The $1$-morphism

1458: $\zeta:B\rightarrow \Kgnb{0,1}{X,1}$ is \emph{twisting} if

1459: \begin{enumerate}

1460: \item The data $(\pi:B\rightarrow T, h:B \rightarrow X)$ is a

1461:   family of stable maps to $X$, i.e. a $1$-morphism $\xi:T\rightarrow

1462:   \Kgnb{0,0}{X,e}$ for some $e\geq 0$.

1463: \item The image of $\xi:T\rightarrow \Kgnb{0,0}{X,e}$ is contained in

1464:   the \emph{very unobstructed} locus of $\Kgnb{0,0}{X}$.

1465: \item The image of $\zeta:T \rightarrow \Kgnb{0,1}{X,1}$ is contained

1466:   in the \emph{very unobstructed} locus of the evaluation morphism

1467:   $\text{ev}:\Kgnb{0,1}{X,1} \rightarrow X$.

1468: \item Denoting by $T_{\text{ev}}$ the dual of the sheaf of relative

1469:   differentials $\Omega_{\text{ev}}$, the pullback bundle

1470:   $\zeta^*T_{\text{ev}}$ is

1471:   $\pi$-relatively generated by global sections.

1472: \item Denoting by $\text{pr}:\Kgnb{0,1}{X,1}\rightarrow \Kgnb{0,0}{X,1}$

1473:   the projection map, and by $T_{\text{pr}}$ the dual of the sheaf of

1474:   relative differentials $\Omega_{\text{pr}}$, the pullback bundle

1475:   $\zeta^*T_{\text{pr}}$, i.e. the line bundle

1476:   $\sigma^*\OO_\Sigma(\sigma)$, is $\pi$-relatively generated by

1477:   global sections.

1478: \end{enumerate}

1479: \end{defn}

1480: 

1481: \

1482: 

1483: \begin{defn}~\label{defn-verytwisting}

1484:   With notation as in definition~\ref{defn-twisting}, a morphism

1485:   $\zeta:T\rightarrow \Kgnb{0,1}{X,1}$ is \emph{very twisting} if it

1486:   is twisting and if $\zeta^*T_{\text{ev}}$ is $\pi$-relatively

1487:   deformation ample.

1488: \end{defn}

1489: 

1490: \

1491: 

1492: \begin{rmk}~\label{rmk-twisting} 

1493:   Regarding the definitions above:

1494: \begin{enumerate}

1495:   

1496: \item In $(2)$ and $(3)$ of defintion~\ref{defn-twisting}, \emph{very

1497:     unobstructed} means that the naive obstruction group vanishes.

1498:   For $(2)$ this means that for each $t\in T$ and the corresponding

1499:   stable map $(h_t:B_t\rightarrow X)$, the following group vanishes:

1500: \begin{equation}

1501: \begin{CD}

1502:   & -1 &  & 0 \\

1503: \mathbb{E}\text{xt}^1_{B_t}( & h_t^*\Omega_X @>>> \Omega_{B_t},\ & \OO_{B_t})

1504: \end{CD}

1505: \end{equation}

1506: 

1507: \item It is easy to see that $\zeta^* T_{\text{pr}}$ is just

1508:   $\sigma^*\OO_{\Sigma}(\sigma)$.  

1509:   

1510: \item Observe the product morphism $(p,g):\Sigma\rightarrow B \times

1511:   X$ is a regular embedding.  Denote by ${\mc N}$ the normal bundle of

1512:   this regular embedding.  Then $(3)$ is equivalent to the condition

1513:   that $R^1p_*\lt({\mc N}(-\sigma)\rt)$ is trivial.  In this case

1514:   $\zeta^*T_{\text{ev}}$ is the locally free sheaf $p_*\lt({\mc

1515:     N}(-\sigma)\rt)$.

1516: 

1517: \item Since the prestable family of maps $(\pi:B \rightarrow T, \xi:B

1518:   \rightarrow X)$  is stable, clearly also $(\pi:B \rightarrow T,

1519:   \zeta: B \rightarrow \Kgnb{0,1}{X,1})$ is stable.

1520:   

1521: \item There are some degree conditions implicit in these definitions.

1522:   The total degree of $\sigma^*\OO_{\Sigma}(\sigma)$ is simply

1523:   $s=2e-e'$ where $e$ is the degree of $h:B \rightarrow \PP^N$ and

1524:   $e'$ is the degree of $g:\Sigma \rightarrow \PP^N$ (both degrees

1525:   with respect to $\OO_{\PP^N}(1)$).  So if $\zeta$ is twisting, we

1526:   have that $2e \geq e'$ and if $\zeta$ is very twisting, we have that

1527:   $2e > e'$.

1528:   

1529: \item Additionally, given a twisting family $\zeta$ a point $b\in B$,

1530:   and a deformation of the line $g_b:\Sigma_b \rightarrow X$ which

1531:   continues to contain $h(b)$, there must be a deformation of the

1532:   whole family $\zeta$ giving rise to the deformation of $\Sigma_b$

1533:   and which does not deform $h:B \rightarrow X$.  In particular, if

1534:   $h:B \rightarrow X$ is also an embedded line, then the map of the

1535:   surface $g:\Sigma \rightarrow X$ must have degree $1$ or $2$ and

1536:   must deform along with a line which intersects $h(B)$ in a fixed

1537:   point.  If $X \subset \PP^n$ is a hypersurface of low degree $d > 1$

1538:   such that to a general line $h:B \rightarrow X$ there is a

1539:   corresponding twisting family $\zeta$ (what we refer to as a

1540:   \emph{twistable line} below), then we must have that $h:B

1541:   \rightarrow X$ is an embedding of a smooth quadric surface.  The

1542:   condition on such $X$ that a general line is \emph{twistable} is

1543:   essentially that, given two general intersecting lines $B$ and $L$

1544:   in $X$, there is a smooth quadric surface $\Sigma$ in $X$ which

1545:   contains both $B$ and $L$.  In a later section we will see that this

1546:   condition does hold for a general hypersurface $X\subset \PP^n$ of

1547:   degree $d$ when $d^2 \leq n+1$.

1548: 

1549: \end{enumerate}

1550: \end{rmk}

1551: 

1552: \begin{lem} \label{lem-twisttogether}

1553:   Suppose $B = B_1 \cup B_2$ is a prestable, geometrically connected

1554:   curve of arithmetic genus $0$ where $B_1$ and $B_2$ are connected

1555:   subcurves such that $B_1 \cap B_2$ is a single node of $B$.  Suppose

1556:   given $\zeta:B \rightarrow \Kgnb{0,1}{X,1}$ such that $\zeta|_{B_i}

1557:   : B_i \rightarrow \Kgnb{0,1}{X,1}$ is twisting for $i=1,2$.  Then

1558:   $\zeta$ is twisting.  If, in addition, at least one of $\zeta_i$ is

1559:   very twisting, then $\zeta$ is very twisting.

1560: \end{lem}

1561: 

1562: \begin{proof}

1563:   This is an easy consequence of lemma~\ref{lem-DAcrit}.

1564: \end{proof}

1565: 

1566: 

1567: \begin{lem} \label{lem-twistopen}

1568:   Let $\pi:B\rightarrow T$ be a family of prestable, geometrically

1569:   connected curves of arithmetic genus $0$ and let $\zeta:B

1570:   \rightarrow \Kgnb{0,1}{X,1}$ be a morphism.  There is an open

1571:   subscheme $U_{\text{twist}}\subset T$ (resp. $U_{vtwist}\subset T$)

1572:   with the following property: for any morphism of schemes

1573:   $f:T'\rightarrow T$, the pullback family $f^*\pi:f^*B \rightarrow

1574:   T'$ and $f^*\zeta: f^*B \rightarrow \Kgnb{0,1}{X,1}$ is twisting

1575:   (resp. very twisting) iff $f(T')\subset U_{\text{twist}}$ (resp.

1576:   $f(T')\subset U_{\text{vtwist}}$).

1577: \end{lem}

1578: 

1579: \begin{proof}

1580:   By \cite[lemma 1]{B} there is an universal open subscheme

1581:   $U_1\subset T$ over which $(\pi:B\rightarrow T, h:B\rightarrow X)$

1582:   is a family of stable maps.  It is clear that $U$, if it exists,

1583:   must also be contained in the complement of the support of

1584: \begin{equation}

1585: \mathbb{R}^1\lt(\pi_*\textit{Hom}_{\OO_B}\rt)\lt(h^*\Omega_X

1586: \rightarrow \Omega_\pi, \OO_B\rt),

1587: \end{equation}

1588: and in the complement of the image under $\pi$ of the supports of the

1589: sheaves:

1590: \begin{eqnarray}

1591: R^1p_*\lt({\mc N}(-\sigma)\rt), \\

1592: \text{coker}\lt( \pi^*\pi_* \sigma^*\OO_{\Sigma}(\sigma) \rightarrow

1593: \sigma^*\OO_{\Sigma}(\sigma) \rt).

1594: \end{eqnarray}

1595: Let $U_2$ denote the complement of these sets in $U_1$.  On $U_2$ all

1596: of the conditions to be twisting (resp. very twisting) are satisfied

1597: except the condition that $\zeta^*T_{\text{ev}}$ is $\pi$-relatively

1598: generated by global sections (resp. $\pi$-relatively deformation

1599: ample).  So we define $U_{\text{twist}}$ to be the complement in $U_2$

1600: of the image under $\pi$ of the cokernel of the morphism

1601: \begin{equation}

1602: \pi^*\pi_* \zeta^*T_{\text{ev}}\rightarrow \zeta^*T_{\text{ev}}.

1603: \end{equation}

1604: And, using lemma~\ref{lem-opDA}, we define $U_{\text{vtwist}}$ to be

1605: the universal open subscheme of $U_2$ over which

1606: $\zeta^*T_{\text{ev}}$ is $\pi$-relatively deformation ample.  It

1607: follows immediately from the construction that $U_{\text{twist}}$ and

1608: $U_{\text{vtwist}}$ have the desired universal properties.

1609: \end{proof}

1610: 

1611: \begin{defn} \label{defn-twistable}

1612:   Suppose $(\pi:B\rightarrow T, h:B\rightarrow X)$ is a family of

1613:   genus $0$ stable maps, i.e. a $1$-morphism $\xi:T\rightarrow

1614:   \Kgnb{0,0}{X,e}$ for some $e \geq 0$.  We say $\xi:T \rightarrow

1615:   \Kgnb{0,0}{X,e}$ is \emph{twistable} (resp. \emph{very twistable})

1616:   if there exists a surjective \'etale morphism $u:T'\rightarrow T$

1617:   and a morphism $\zeta:u^*B \rightarrow \Kgnb{0,1}{X,1}$ with

1618:   $h_\zeta = u^* h$ such that $\zeta$ is twisting (resp. very

1619:   twisting).

1620: \end{defn}

1621: 

1622: \begin{prop} \label{prop-twistopen}

1623:   Let $\xi = (\pi:B\rightarrow T, h:B\rightarrow X)$ be a $1$-morphism

1624:   $\xi:T \rightarrow \Kgnb{0,0}{X,e}$.  There is an open subscheme

1625:   $U_{t-able}\subset T$ (resp $U_{vt-able}\subset T$) such that for

1626:   each morphism of schemes $f:T'\rightarrow T$, the pullback $(f^*\pi:

1627:   f^*B \rightarrow T', f^*h: f^*B \rightarrow X)$ is twistable (resp.

1628:   very twistable) iff $f(T') \subset U_{t-able}$ (resp. $f(T')\subset

1629:   U_{vt-able}$).

1630: \end{prop}

1631: 

1632: \begin{proof}

1633:   It suffices to check that if $t_0\in T$ is a geometric point such

1634:   that $h_{t_0}:B_{t_0} \rightarrow X$ is twistable (resp. very

1635:   twistable), then there is an \'etale neighborhood of $t_0\in T$ over

1636:   which $\xi$ is twistable (resp. very twistable).  Denote by

1637:   $\zeta_0:B_{t_0} \rightarrow \Kgnb{0,1}{X,1}$ the twisting morphism.

1638:   We consider $\Kgnb{0,1}{X,1}$ as a projective scheme via the

1639:   Pl\"ucker and Segr\'e embeddings of $\GG(1,n) \times \PP^n

1640:   \hookrightarrow \PP^{\frac{n(n+1)^2}{2}-1}$.  Let $\beta$ denote the

1641:   degree of the stable map $\zeta_0$.

1642: 

1643: \

1644: 

1645: Define ${\mc M} = T \times \Kgnb{0,0}{\Kgnb{0,1}{X,1}, \beta}$, i.e.

1646: ${\mc M}$ parametrizes pairs $(t,\zeta)$ where $t\in T$ is a point and

1647: where $\zeta:B \rightarrow \Kgnb{0,1}{X,1}$ is a genus $0$ stable map

1648: of degree $\beta$.  Denote the universal stable map by

1649: \begin{eqnarray}

1650: \rho:{\mc B} \rightarrow \Kgnb{0,0}{\Kgnb{0,1}{X,1},\beta}, \\

1651: \zeta: {\mc B} \rightarrow \Kgnb{0,1}{X,1}.

1652: \end{eqnarray}

1653: As in notation~\ref{not-1mor}, let $p:\Sigma \rightarrow {\mc B}$ be

1654: the pullback by $\zeta$ of the universal curve over $\Kgnb{0,1}{X,1}$,

1655: let $\sigma:{\mc B} \rightarrow \Sigma$ be the pullback of the

1656: universal section, let $g:\Sigma \rightarrow X$ be the pullback of the

1657: universal map, and let $h = g \circ \sigma$.  So we have a family of

1658: prestable maps 

1659: \begin{equation}

1660: \widetilde{\xi} = \lt( \rho:{\mc B} \rightarrow \Kgnb{0,0}{\Kgnb{0,1}{X,1},

1661:   \beta}, h: {\mc B} \rightarrow X \rt).

1662: \end{equation}

1663: By \cite[lemma 1]{B} there is a maximal open substack ${\mc U}_e \subset

1664: \Kgnb{0,0}{\Kgnb{0,1}{X,1}, \beta}$ over which $\widetilde{\xi}$ is

1665: stable of degree $e$.  By assumption, $(t_0,\zeta_0)$ is in $T \times

1666: {\mc U}_e$.  

1667: 

1668: \

1669: 

1670: In the last paragraph we constructed a $1$-morphism

1671: \begin{equation}

1672: (1_T,\widetilde{\xi}): T \times {\mc U}_e \rightarrow T \times

1673: \Kgnb{0,0}{X,e}.

1674: \end{equation}

1675: We also saw that $(t_0,\zeta_0)$ is in the domain of this

1676: $1$-morphism.  The claim is that $(1_T,\widetilde{\xi})$ is smooth on

1677: a neighborhood of $(t_0,\zeta_0)$, i.e. that $\widetilde{\xi}: {\mc

1678:   U}_e \rightarrow \Kgnb{0,0}{X,e}$ is smooth at $\zeta_0$.  First we

1679: will show that ${\mc U}_e$ is smooth at $\zeta_0$.  The space of first

1680: order deformations and the obstruction space of ${\mc U}_e$ at

1681: $\zeta_0$ are given by

1682: \begin{equation}

1683: \mathbb{E}\text{xt}^i_{B_{t_0}}(L^\cdot_{\zeta_0},\OO_{B_{t_0}}),

1684: \end{equation}

1685: for $i=1,2$ respectively, where $L^\cdot_{\zeta_0}$ is the complex

1686: \begin{equation}

1687: \begin{CD}

1688: -1 & & 0 \\

1689: \zeta_0^* \Omega_{\Kgnb{0,1}{X,1}} @> d(\zeta_0)^\dagger >> \Omega_{B_{t_0}}.

1690: \end{CD}

1691: \end{equation}

1692: Now the induced morphism $\xi_0:B_{t_0} \rightarrow X$ by $\xi_0 =

1693: \text{ev} \circ \zeta_0$ also has an associated complex

1694: $L^\cdot_{\xi_0}$:

1695: \begin{equation}

1696: \begin{CD}

1697: -1 & & 0 \\

1698: \xi_0^* \Omega_X @> d(\xi_0)^\dagger >> \Omega_{B_{t_0}}.

1699: \end{CD}

1700: \end{equation}

1701: There is a morphism of complexes:

1702: \begin{eqnarray}

1703: \gamma: L^\cdot_{\xi_0} \rightarrow L^\cdot_{\zeta_0} \\

1704: \gamma^0 = \text{id} : \Omega_B \rightarrow \Omega_B \\

1705: \gamma^{-1} = \zeta_0^*\lt( d(\text{ev})^\dagger \rt): \zeta_0^*

1706: \text{ev}^* \Omega_X \rightarrow \zeta_0^* \Omega_{\Kgnb{0,1}{X,1}}.

1707: \end{eqnarray}

1708: There is also a morphism of complexes:

1709: \begin{equation}

1710: \delta: L^\cdot_{\zeta_0} \rightarrow \zeta_0^* \Omega_{\text{ev}}[1],

1711: \end{equation}

1712: where $\delta^{-1}:\zeta_0^* \Omega_{\Kgnb{0,1}{X,1}} \rightarrow

1713: \zeta_0^*\Omega_{\text{ev}}$ is the pullback of the canonical

1714: surjection.  And the triple:

1715: \begin{equation}

1716: \begin{CD}

1717: L^\cdot_{\xi_0} @> \gamma >> L^\cdot_{\zeta_0} @> \delta >> \zeta_0^*

1718: \Omega_{\text{ev}}[1] 

1719: \end{CD}

1720: \end{equation}

1721: is an exact triangle.  Thus there is a corresponding long exact

1722: sequence of $\mathbb{E}\text{xt}$'s.  Condition $(2)$ of

1723: definition~\ref{defn-twisting} says that

1724: $\mathbb{E}\text{xt}^1_{B_{t_0}}(L^\cdot_{\xi_0},\OO_{B_{t_0}})$ is

1725: zero.  By condition $(4)$ of the definition, $\zeta_0^* T_{\text{ev}}$

1726: is generated by global sections.  Since $B_{t_0}$ is connected of

1727: arithmetic genus $0$, we have that $H^1(B_{t_0},\OO_{B_{t_0}})$ is

1728: zero.  So for any trivial bundle, $H^1$ is zero.  Since $H^2$ vanishes

1729: on all coherent sheaves, we conclude that for any sheaf generated by

1730: global sections, $H^1$ is zero.  Thus we have that the group

1731: \begin{equation}

1732: \mathbb{E}\text{xt}^2(\zeta_0^*\Omega_{\text{ev}}, \OO_{B_{t_0}}) =

1733: H^1(B_{t_0}, T_{\text{ev}}),

1734: \end{equation}

1735: is also zero.  By the long exact sequence, we conclude that

1736: $\mathbb{E}\text{xt}^1_{B_{t_0}}( L^\cdot_{\zeta_0}, \OO_{B_{t_0}} )$

1737: is also zero.  So the obstruction group vanishes and ${\mc U}_e$ is

1738: smooth at $\zeta_0$.

1739: 

1740: \

1741: 

1742: By condition $(2)$ the image point $\xi_0\in \Kgnb{0,0}{X,e}$ is a

1743: smooth point of $\Kgnb{0,0}{X,e}$.  Thus to prove that

1744: $\widetilde{\xi}: {\mc U}_e \rightarrow \Kgnb{0,0}{X,e}$ is smooth, it

1745: suffices to prove that derivative map $d(\widetilde{\xi})$ is

1746: surjective on the space of first order deformations.  The map

1747: \begin{equation}

1748: d(\widetilde{\xi}):

1749: \mathbb{E}\text{xt}^1_{B_{t_0}}(L^{\cdot}_{\zeta_0}, \OO_{B_{t_0}})

1750: \rightarrow \mathbb{E}\text{xt}^1_{B_{t_0}}(L^\cdot_{\xi_0},

1751: \OO_{B_{t_0}}),

1752: \end{equation}

1753: is precisely the map occurring in the long exact sequence of

1754: $\mathbb{E}\text{xt}$'s from the paragraph above.  By the long exact

1755: sequence, the cokernel of this map is a subgroup of

1756: $H^1(B_{t_0},T_{\text{ev}})$, and this is zero as we have seen.

1757: Therefore $\widetilde{\xi}$ is smooth at $\zeta_0\in {\mc U}_e$.  

1758: 

1759: \

1760: 

1761: Consider the morphism $(1_T,\xi): T\rightarrow T \times

1762: \Kgnb{0,0}{X,e}$.  We can form the fiber product ${\mc M}$ of

1763: $(1_T,\xi)$ with the morphism $(1_T,\widetilde{\xi}): T\times {\mc

1764:   U}_e \rightarrow T \times \Kgnb{0,0}{X,e}$.  The fiber product ${\mc

1765:   M}$ exactly parametrizes triples $(t,\zeta,\theta)$ where $t\in T$

1766: is a point, $\zeta:B \rightarrow \Kgnb{0,1}{X,1}$ is a point in ${\mc

1767:   U}_e$, and $\theta:\xi_t \rightarrow \widetilde{\zeta}$ is an

1768: equivalence of objects in the groupoid $\Kgnb{0,0}{X,e}(\SP

1769: \kappa(t))$.  The projection map $\text{pr}_1:{\mc M} \rightarrow T$

1770: is smooth at $(t_0,\zeta_0)$ by the last paragraph.  So we can find an

1771: \'etale morphism $f:M \rightarrow {\mc M}$ of a scheme to ${\mc M}$

1772: whose image contains $(t_0,\zeta_0)$, and such that $M\rightarrow T$

1773: is smooth.  Thus there is an \'etale morphism $u:T' \rightarrow T$ and

1774: a section $z:T' \rightarrow M$.  Define $\zeta:T'\rightarrow {\mc

1775:   U}_e$ to be the composition $\text{pr}_2\circ g \circ z$.

1776: 

1777: \

1778: 

1779: We also denote by $\zeta:B' \rightarrow \Kgnb{0,1}{X,1}$ the pullback

1780: by $\zeta:T'\rightarrow {\mc U}$ of the universal stable map.  As

1781: $\widetilde{\xi}(\zeta): B' \rightarrow \Kgnb{0,0}{X,e}$ is equivalent

1782: to $u^* \xi: u^*B \rightarrow \Kgnb{0,0}{X,e}$, after replacing $T'$

1783: by an \'etale, cover, we may suppose that $B'=u^*B$ as $T'$-schemes,

1784: and $\widetilde{\xi}(\zeta) = u^*\xi$.  Now the fiber of

1785: $\zeta:u^*B\rightarrow \Kgnb{0,1}{X,1}$ over any preimage of

1786: $(t_0,\zeta_0)$ is twisting.  So by lemma~\ref{lem-twistopen}, up to

1787: replacing $T'$ by a Zariski open subscheme, we may suppose that

1788: $\zeta:u^*B \rightarrow \Kgnb{0,1}{X,1}$ is twisting.  Similarly, if

1789: $(t_0,\zeta_0)$ is very twisting, we may suppose that $\zeta$ is very

1790: twisting.  So we conclude that on the Zariski open subscheme of $T$

1791: which is the image of $u:T'\rightarrow T$, the family $\xi:B

1792: \rightarrow \Kgnb{0,0}{X,e}$ is twistable (resp. very twistable).

1793: Since this holds for every point $t_0\in T$ where $\xi_0$ is

1794: twistable, the lemma is proved.

1795: \end{proof}

1796: 

1797: \begin{lem} \label{lem-twisttogether2}

1798:   Suppose given two families

1799: \begin{equation}

1800: \xi_i = \lt( \lt( \pi_i:B_i \rightarrow T, \sigma_i:T \rightarrow B_i \rt),

1801: h_i:B_i \rightarrow X \rt), i=1,2.

1802: \end{equation}

1803: such that for each $(\pi_i:B_i \rightarrow T, h_i: B_i \rightarrow X)$

1804: is twistable, and such that $h_1\circ \sigma_1 = h_2 \circ \sigma_2$.

1805: For each $t\in T$, assume that the locus of free lines in $X$ passing

1806: through $h_1\circ\sigma_1(t) = h_2\circ \sigma_2(t)$ is irreducible.

1807: 

1808: \

1809: 

1810: Let us denote by

1811: \begin{equation}

1812: \xi =  \lt( \pi:B \rightarrow T, h:B \rightarrow X \rt)

1813: \end{equation}

1814: the family obtained by taking $B$ to be the connected sum of $B_1$ and

1815: $B_2$ where the section $\sigma_1$ is identified with the section

1816: $\sigma_2$.  Then $\xi$ is a twistable family.  Moreover, if at least

1817: one of $\xi_1, \xi_2$ is very twistable, then $\xi$ is very twistable.

1818: \end{lem}

1819: 

1820: \begin{proof}

1821:   This follows essentially by lemma~\ref{lem-twisttogether}.  First of

1822:   all, using proposition~\ref{prop-twistopen}, it suffices to prove

1823:   the result when $T=\SP k$ for some algebraically closed field $k$.

1824:   We suppose that we are in this case.

1825: 

1826: \

1827: 

1828: For each of $i=1,2$, let ${\mc M}_1$ denote the fiber product

1829: constructed in the proof of proposition~\ref{prop-twistopen}, i.e.

1830: ${\mc M}_1$ parametrizes pairs $(\zeta_i,\theta_i)$ where $\zeta_i:B_i

1831: \rightarrow \Kgnb{0,1}{X,1}$ is a twisting family (resp. very twisting

1832: family) such that the induced map

1833: \begin{equation}

1834: \widetilde{\zeta}_i = \lt( (B_i,\sigma_i),g_i\circ \rho_i:B_i

1835: \rightarrow X \rt)

1836: \end{equation}

1837: is stable, and where $\theta_i:\xi_i \rightarrow

1838: \widetilde{\zeta}_i$ is an equivalence of objects.

1839: Since each of $\xi_i$ is twistable, we see that each of ${\mc M}_i$ is

1840: nonempty.  

1841: 

1842: \

1843: 

1844: By the proof of proposition~\ref{prop-twistopen}, each of ${\mc M}_i$

1845: is smooth.  By the definition of twisting families, for each $i=1,2$

1846: the morphism

1847: \begin{equation}

1848: e_i: {\mc M}_i \rightarrow \Kgnb{0,1}{X,1},\ \zeta_i\mapsto

1849: \zeta_i(\sigma_i) 

1850: \end{equation}

1851: has image

1852: contained in the unobstructed locus of $\text{ev}:

1853: \Kgnb{0,1}{X,1}\rightarrow X$.  Let $P\subset \Kgnb{0,1}{X,1}$ be the

1854: preimage under $\text{ev}$ of the point

1855: $p=h_1(\sigma_1)=h_2(\sigma_2)$.  The image of $e_i$ is contained in

1856: the smooth locus of $P$.  The claim is that $e_i:{\mc M}_i \rightarrow

1857: P$ is smooth.  By \cite[proposition I.2.14.2]{K}, the obstruction space at a

1858: point $\zeta_i$ is contained in the cohomology group $H^1\lt( B_i,

1859: \zeta_i^* T_{\text{ev}}(-\sigma_i) \rt)$.  By the definition of a

1860: twisting family, $\zeta_i^* T_{\text{ev}}$ is generated by global

1861: sections.  Thus, by lemma~\ref{lem-gend}, the cohomology group above

1862: is zero.  Since the obstruction space vanishes, we conclude that $e_i$

1863: is smooth.  

1864: 

1865: \

1866: 

1867: Since both $e_1:{\mc M}_1 \rightarrow P$ and $e_2:{\mc M}_2

1868: \rightarrow P$ are smooth, both have nonempty, open image.  And $P$ is

1869: irreducible by assumption.  Therefore the image of $e_1$ and the image

1870: of $e_2$ intersect.  If we choose a family $\zeta_1\in {\mc M}_1$ and

1871: $\zeta_2\in {\mc M}_2$ such that $e_1(\zeta_1)=e_2(\zeta_2)$, then we

1872: can glue $\zeta_1$ and $\zeta_2$ to obtain a morphism $\zeta:B

1873: \rightarrow \Kgnb{0,1}{X,1}$ such that $\zeta|_{B_1}=\zeta_1$ and

1874: $\zeta|_{B_2} = \zeta_2$.  By lemma~\ref{lem-twisttogether}, we

1875: conclude that $\zeta$ is twisting.  Moreover, if at least one of

1876: $\zeta_i, i=1,2$ is very twisting, then $\zeta$ is very twisting.  And

1877: $\widetilde{\zeta} = \xi$.  This shows that $\xi$ is twistable, and it

1878: is very twistable if at least one of $\xi_i, i=1,2$ is very twistable.

1879: \end{proof}

1880: 

1881: \begin{hyp} \label{hyp-2}

1882:   Let $U \subset \Kgnb{0,1}{X,1}$ denote the preimage of

1883:   $U_{\text{t-able}} \subset \Kgnb{0,0}{X,1}$ under $\text{pr}$.  The

1884:   evaluation morphism $\text{ev}:U \rightarrow X$ has Zariski dense

1885:   image.

1886: \end{hyp}

1887: 

1888: \begin{defn} \label{defn-pos}

1889:   Suppose $\pi:B \rightarrow T$ is a family of prestable,

1890:   geometrically connected curves of arithmetic genus $0$.  A

1891:   $1$-morphism $\zeta:B \rightarrow \Kgnb{0,1}{X,e}$ is

1892:   \emph{positive} (resp. \emph{very positive}) if:

1893: \begin{enumerate}

1894: \item The data $(\pi:B \rightarrow T, h:B \rightarrow X)$ is a family

1895:   of stable maps to $X$, i.e. a $1$-morphism $\xi:T \rightarrow

1896:   \Kgnb{0,0}{X,\epsilon}$ for some $\epsilon\geq 0$.

1897: \item The image of $\xi:T \rightarrow \Kgnb{0,0}{X,\epsilon}$ is contained in

1898:   the \emph{very unobstructed} locus of $\Kgnb{0,0}{X,\epsilon}$.

1899: \item The image of $\text{pr}\circ \zeta: T\rightarrow \Kgnb{0,0}{X,e}$

1900:   is contained in the \emph{very unobstructed} locus of

1901:   $\Kgnb{0,0}{X,e}$.

1902: \item The pullback bundle $(\text{pr}\circ \zeta)^*

1903:   T_{\Kgnb{0,0}{X,e}}$ is $\pi$-relatively deformation ample.

1904: \item The pullback line bundle $\sigma^*\OO_{\Sigma}(\sigma)$ is

1905:   $\pi$-relatively generated by global sections

1906:   (resp. $\pi$-relatively ample).  

1907: \end{enumerate}

1908: \end{defn}

1909: 

1910: \begin{rmk} \label{rmk-pos}

1911:   Regarding the definition above:

1912: \begin{enumerate}

1913:   

1914: \item This definition is very similar to

1915:   definition~\ref{defn-twisting}.  It differs in that $e$ need not

1916:   equal $1$ and that we only require $\text{pr}\circ \zeta$ to have

1917:   image in the very unobstructed locus, instead of requiring $\zeta$

1918:   to have image in the very unobstructed locus of $\text{ev}$.

1919:   

1920: \item Consider the case when $T = \SP \kappa$ for some field $\kappa$,

1921:   and suppose that $B$ is smooth, i.e. $B \cong \PP^1_\kappa$.  If

1922:   $\zeta:B \rightarrow \Kgnb{0,1}{X,e}$ is positive, then the morphism

1923:   $\text{pr} \circ \zeta: B \rightarrow \Kgnb{0,0}{X,e}$ is \emph{very

1924:     free} in the sense of Debarre \cite[p. 86]{De}.

1925: 

1926: \end{enumerate}

1927: \end{rmk}

1928: 

1929: \begin{lem}\label{lem-posopen}

1930:   Let $\pi:B \rightarrow T$ be a family of prestable, geometrically

1931:   connected curves of arithmetic genus $0$ and let $\zeta:B

1932:   \rightarrow \Kgnb{0,1}{X,1}$ be a $1$-morphism.  There is an open

1933:   subscheme $U_{\text{pos}} \subset T$ (resp. $U_{\text{v-pos}}\subset

1934:   T$) with the following property: for any morphism of schemes

1935:   $f:T'\rightarrow T$, the pullback family $f^*\pi: f^*B \rightarrow

1936:   T'$ and $f^*\zeta: f^*B \rightarrow \Kgnb{0,1}{X,1}$ is positive

1937:   (resp. very positive) iff $f(T')\subset U_{\text{free}}$ (resp.

1938:   $f(T')\subset U_{\text{v-free}}$).

1939: \end{lem}

1940: 

1941: \begin{proof}

1942:   The proof is almost identical to the proof of

1943:   lemma~\ref{lem-twistopen}.

1944: \end{proof}

1945: 

1946: \begin{lem} \label{lem-posdef}

1947:   Suppose that $T=\SP k$ is a point and $\zeta:B\rightarrow

1948:   \Kgnb{0,1}{X,e}$ is a positive $1$-morphism whose image is contained

1949:   in the locus of very stable maps.

1950: \begin{enumerate}

1951: 

1952: \item If $B$ is smooth, then $\zeta:B \rightarrow \Kgnb{0,1}{X,e}$ is

1953:   \emph{free} in the sense of Koll\'ar ~\cite[definition II.3.11]{K}.

1954:   If $\zeta$ is very positive, then $\zeta$ is \emph{very free}.

1955: 

1956: \item In any case, the $1$-morphism is \emph{unobstructed} in the

1957:   sense of Koll\'ar \cite[definition I.2.6]{K}; in particular it is

1958:   the specialization of a positive $1$-morphism $\zeta_\eta:B_\eta

1959:   \rightarrow \Kgnb{0,1}{X,e}$ with $B_\eta$ geometrically connected

1960:   and smooth, and whose image is contained in the locus of very stable

1961:   maps.

1962: \end{enumerate}

1963: \end{lem}

1964: 

1965: \begin{proof}

1966:   Suppose that the image of $\zeta$ lies in the locus of very stable

1967:   maps.  By the relative version of lemma~\ref{lem-def3a}, we have

1968:   that the image of $\zeta$ is in the smooth locus of

1969:   $\Kgnb{0,1}{X,e}$.  And we have a short exact sequence:

1970: \begin{equation}

1971: \begin{CD}

1972: 0 @>>> \sigma^* \OO_{\Sigma}(\sigma) @>>> \zeta^* T_{\Kgnb{0,1}{X,e}}

1973: @>>> (\text{pr}\circ \zeta)^* T_{\Kgnb{0,0}{X,e}} @>>> 0

1974: \end{CD}

1975: \end{equation}

1976: By condition $(4)$ of definition ~\ref{defn-pos}, $(\text{pr}\circ

1977: \zeta)^* T_{\Kgnb{0,0}{X,e}}$ is deformation ample.  And by condition

1978: $(5)$ of definition ~\ref{defn-pos}, $\sigma^*\OO_{\Sigma}(\sigma)$ is

1979: generated by global sections (resp. deformation ample).  Therefore

1980: $\zeta^* T_{\Kgnb{0,1}{X,e}}$ is generated by global sections.  And if

1981: $\zeta$ is very positive, then $\zeta^* T_{\Kgnb{0,1}{X,e}}$ is

1982: deformation ample by $(2)$ of lemma ~\ref{lem-secDA}.  So if $B$ is

1983: smooth, then $\zeta$ is free, and it is very free if $\zeta$ is very

1984: positive.  This proves $(1)$.

1985: 

1986: \

1987: 

1988: Since $\zeta^* T_{\Kgnb{0,1}{X,e}}$ is generated by global sections,

1989: by lemma ~\ref{lem-gend}, $H^1(B,\zeta^* T_{\Kgnb{0,1}{X,e}})$ is

1990: zero.  Therefore $\zeta$ is unobstructed in the sense of Koll\'ar.

1991: Now let $\pi: {\mc B} \rightarrow \SP R$ be a smoothing of $B$, i.e. a

1992: flat family of proper, geometrically connected, prestable curves of

1993: arithmetic genus $0$ over a DVR such that the special fiber is

1994: isomorphic to $B$ and such that the general fiber $B_\eta$ is smooth.

1995: By ~\cite[theorem I.2.10]{K}, the projection of the relative

1996: Hom-scheme,

1997: \begin{equation}

1998: \textit{Hom}_{\SP R}({\mc B}, \SP R \times \Kgnb{0,1}{X,e})

1999: \rightarrow \SP R,

2000: \end{equation}

2001: is smooth at $[\zeta]$.  Therefore, after making some finite, flat

2002: base change $\SP R' \rightarrow \SP R$ we may suppose that $\zeta$ is

2003: the specialization of a $1$-morphism $\zeta_R:{\mc B} \rightarrow

2004: \Kgnb{0,1}{X,e}$.  By lemma ~\ref{lem-posopen}, we have that $\zeta_R$

2005: is positive, in particular $\zeta_\eta:B_\eta \rightarrow

2006: \Kgnb{0,1}{X,e}$ is positive.  Since the locus of very stable maps in

2007: $\Kgnb{0,1}{X,e}$ is open, we conclude that the image of $\zeta_R$ is

2008: contained in this locus.

2009: \end{proof}

2010: 

2011: Now we come to the main notion of this section.

2012: 

2013: \begin{defn} \label{defn-inducts}

2014:   Suppose $\pi:B \rightarrow T$ is a family of prestable,

2015:   geometrically connected curves of arithmetic genus $0$.  An

2016:   \emph{inducting pair} of degree $e$ is a pair

2017: \begin{equation}

2018: \lt( \zeta_1:B\rightarrow \Kgnb{0,1}{X,1}, \overline{\zeta}_e:B

2019: \rightarrow \Kgnb{0,1}{X,e} \rt),

2020: \end{equation}

2021: such that:

2022: \begin{enumerate}

2023: 

2024: \item $\zeta_1$ is very twisting,

2025: 

2026: \item $\overline{\zeta}_e$ is very positive and the image of

2027:   $\overline{\zeta}_e$ is contained in the locus of very stable maps,

2028:   and

2029: 

2030: \item the two morphisms $h_{\zeta_1}:B \rightarrow X$ and

2031:   $h_{\overline{\zeta}_e}: B \rightarrow X$ are equal.

2032: \end{enumerate}

2033: \end{defn}

2034: 

2035: \begin{lem} \label{lem-inducting}

2036:   Let $\pi:B \rightarrow T$ be a family of prestable, geometrically

2037:   connected curves of arithmetic genus $0$, and let

2038: \begin{equation}

2039: \lt( \zeta_1:B

2040: \rightarrow \Kgnb{0,1}{X,1}, \overline{\zeta}_e:B \rightarrow

2041: \Kgnb{0,1}{X,e} \rt),

2042: \end{equation}

2043: be a pair of $1$-morphisms such that $h_{\zeta_1} =

2044: h_{\overline{\zeta}_e}$.  Then there is an open subscheme

2045: $U_{\text{induct}}\subset T$ with the following property: for any

2046: morphism of schemes $f:T'\rightarrow T$, the pullback of

2047: $(\zeta_1,\overline{\zeta}_e)$ is inducting iff $f(T')\subset U$.

2048: \end{lem}

2049: 

2050: \begin{proof}

2051:   We just define $U_{\text{induct}}$ to be the intersection of the

2052:   open subset $U_{\text{vtwist}}\subset T$ as in

2053:   lemma~\ref{lem-twistopen} for $\zeta_1$ and the open subset

2054:   $U_{\text{v-pos}}\subset T$ as in lemma~\ref{lem-posopen} for

2055:   $\overline{\zeta}_e$.

2056: \end{proof}

2057: 

2058: We finally come to our last definition.

2059: 

2060: \begin{defn} \label{defn-inductable}

2061:   Suppose $(\pi:B\rightarrow T, \overline{\zeta}_e:B \rightarrow

2062:   \Kgnb{0,1}{X,e})$ is a very positive family whose image is contained

2063:   in the locus of very stable maps.  We say $\overline{\zeta}_e$ is

2064:   \emph{inductable} if there is a surjective \'etale morphism $u:T'

2065:   \rightarrow T$ and a morphism $\zeta_1:u^*B \rightarrow

2066:   \Kgnb{0,1}{X,1}$ with $h_{\zeta_1} = u^* h_{\overline{\zeta}_e}$

2067:   such that $(\zeta_1,\overline{\zeta}_e)$ is an inducting pair.

2068: \end{defn}

2069: 

2070: \begin{lem} \label{lem-inductable}

2071:   Let $(\pi:B \rightarrow T, \overline{\zeta}_e:B \rightarrow

2072:   \Kgnb{0,1}{X,e})$ be a very free family.  There is an open subscheme

2073:   $U_{i-able} \subset T$ such that for each morphism of schemes

2074:   $f:T'\rightarrow T$, the pullback $(f^*\pi: f^*B \rightarrow T',

2075:   f^*\overline{\zeta}_e: f^* B \rightarrow \Kgnb{0,1}{X,e})$ is

2076:   inductable iff $f(T')\subset U_{i-able}$.

2077: \end{lem}

2078: 

2079: \begin{proof}

2080:   We simply apply proposition~\ref{prop-twistopen} to

2081: \begin{equation}

2082: \xi := (\pi:B

2083: \rightarrow T, h_{\overline{\zeta}_e}: B \rightarrow X).

2084: \end{equation}

2085: \end{proof}

2086: 

2087: \section{The induction argument} \label{sec-induct}

2088: 

2089: In this section we will show that given an inductable $1$-morphism

2090: $\overline{\zeta}_e: B \rightarrow \Kgnb{0,1}{X,e}$, this gives rise

2091: to an inductable $1$-morphism $\overline{\zeta}_{e+1}: B \rightarrow

2092: \Kgnb{0,1}{X,e+1}$.  The basic idea is, given an inducting pair

2093: $(\zeta_1,\overline{\zeta}_{e+1})$ to form the family of connected

2094: sums.  This isn't quite an inductable $1$-morphism, but after

2095: deforming and then performing a simple operation which we call a

2096: \emph{modification}, we do obtain an inductable $1$-morphism

2097: $\overline{\zeta}_{e+1}$.

2098: 

2099: \begin{notat} \label{not-divs}

2100:   We will follow \cite{QDiv} in our notation of the tautological

2101:   divisors on $\Kgnb{0,1}{\PP^N,e}$. Specifically, in

2102:   $A^1(\Kgnb{0,1}{\PP^N,e})_\QQ$ we denote by $\Delta_{(e_1,e_2)}$ the

2103:   $\QQ$-divisor whose general point parametrizes a reducible embedded

2104:   curve with one irreducible component of degree $e_1$, one

2105:   irreducible component of degree $e_2$ and where the marked point is

2106:   on the first irreducible component.  We denote by ${\mc L}$ the

2107:   divisor class $\text{ev}^*\OO_{\PP^N}(1)$.  And we denote by ${\mc

2108:     H}$ the divisor which parametrizes stable maps whose image in

2109:   $\PP^r$ intersects a given codimension $2$ linear space.  Given a

2110:   closed subscheme $X\subset \PP^N$, we also denote by

2111:   $\Delta_{(e_1,e_2)}$, ${\mc L}$ and ${\mc H}$ the pullbacks of the

2112:   divisors given above by the induced $1$-morphism $\Kgnb{0,1}{X,e}

2113:   \rightarrow \Kgnb{0,1}{\PP^N,e}$.

2114: \end{notat}

2115: 

2116: Before proceeding to the main result of this section, we describe an

2117: operation which we will perform repeatedly in the proof.  Suppose that

2118: $\zeta:B \rightarrow \Kgnb{g,r}{X,e}$ is a family of stable maps as in

2119: notation ~\ref{not-1mor}.  Suppose $b\in B$ is a point whose image

2120: $\zeta(b)$ is a stable map

2121: \begin{equation}

2122: ((\Sigma_b, p_1, \dots, p_r), g_b:\Sigma_b

2123: \rightarrow X).

2124: \end{equation}  

2125: Suppose that $L\subset \Sigma_b$ is an irreducible component which is

2126: not contracted by $g_b$ and $p_i\in L$ is one of the marked points.

2127: For simplicity assume that $L$ contains no nodes, in particular this

2128: is the case when $L$ has genus $0$.  Let $M\subset \Sigma_b$ denote

2129: all the irreducible components of $\Sigma_b$ other than $L$.  Let

2130: $R=(r_1,\dots, r_l)$ denote the set of intersection points of $L$ and

2131: $M$ with some ordering.  Let $D = (p_{j_1},\dots, p_{j_m})$ denote the

2132: set of marked points which lie on $L$ other than $p_i$ and let

2133: $E=(p_{k_1},\dots, p_{k_n})$ denote the set of marked points which lie

2134: on $M$.

2135: 

2136: \

2137: 

2138: Form the product surface $L\times L$ with diagonal $\Delta:L

2139: \rightarrow L\times L$ and let $u:\Lambda \rightarrow L$ denote the

2140: blowing up of $L\times L$ along the set of points $\Delta(R\cup D)$.

2141: For each point $p \in R\cup D$, let $F_{p} \subset \Lambda$ be the

2142: proper transform of $L \times \{ p \}$.  Let $F_{p_i}$ denote the

2143: proper transform of the diagonal $\Delta(L) \subset L \times L$.

2144: Consider $\text{pr}_1\circ u: \Lambda \rightarrow L$ as a family of

2145: prestable curves parametrized by $L$.  Then the data

2146: \begin{equation}

2147: \widetilde{\zeta}_{\Lambda} = 

2148: (\text{pr}_1 \circ u: \Lambda

2149: \rightarrow L, (\sigma_{p_i}, \sigma_p | p\in R\cup D) ),

2150: \end{equation}

2151: where $\sigma_p:L \rightarrow F_p$ is the unique isomorphism such that

2152: $\text{pr}_1\circ \sigma_p$ is the identity, gives a family of

2153: prestable marked curves parametrizing by $L$, it is essentially the

2154: constant family $L\times (L,\{p_i\}\cup R \cup D) \rightarrow L$

2155: except that we are allowing $p_i$ to vary among all points in $L$ and

2156: then blowing up to obtain a stable family.

2157: 

2158: \

2159: 

2160: Next we form the constant family of prestable marked curves

2161: parametrized by $L$:

2162: \begin{equation}

2163: \widetilde{\zeta}_{L\times M} =

2164: (\text{pr}_1: L \times M 

2165: \rightarrow L, (s_p | p \in R \cup E)

2166: \end{equation}

2167: where $s_p: L \rightarrow L\times M$ is simply $s_p(t)=(t,p)$.  We can

2168: glue $\widetilde{\zeta}_{\Lambda}$ and $\widetilde{\zeta}_{L\times M}$

2169: as follows.  For each $p\in R$, we identify the section $\sigma_p$ of

2170: $\widetilde{\zeta}_{\Lambda}$ with the section $s_p$ of

2171: $\widetilde{\zeta}_{L\times M}$.  Here the identification is the

2172: unique one compatible with projection to $L$.  Let us denote the new

2173: family of prestable marked curves by

2174: \begin{equation}

2175: (\rho:\Pi \rightarrow L, (\phi_j: L

2176: \rightarrow \Pi | j=1,\dots, r)

2177: \end{equation}

2178: where $\Pi$ is the surface obtained by gluing $\Lambda$ and $L\times

2179: M$ as above, and where

2180: \begin{equation}

2181: \phi_j = \lt\{ \begin{array}{ll}

2182:                 \sigma_{p_j} & ,p_j\in L \\

2183:                 s_{p_j}      & ,p_j\in M

2184:                \end{array} \rt.

2185: \end{equation}

2186: Notice that there is a unique morphism $\text{pr}_2:\Pi \rightarrow

2187: \Sigma_b$ whose restriction to $\Lambda$ is $\text{pr}_2\circ u:

2188: \Lambda \rightarrow L \subset \Sigma_b$ and whose restriction to $L

2189: \times M$ is $\text{pr}_2: L \times M \rightarrow M\subset \Sigma_b$.

2190: We form a family of stable maps parametrized by $L$,

2191: \begin{equation}

2192: \widetilde{\zeta}_\Pi = \lt( \lt( \rho: \Pi \rightarrow L, \phi_1,\dots,

2193: \phi_r \rt), g_b 

2194: \circ \text{pr}_2: \Pi \rightarrow X \rt).

2195: \end{equation}

2196: Notice that the family $\widetilde{\zeta}_\Pi$ is stable, and if we

2197: remove the section $\phi_i$ and stabilize, we just get the constant

2198: family parametrized by $L$ whose image is the stabilization of

2199: $\zeta(b)$ upon removing $p_i$.  Also, the image

2200: $\widetilde{\zeta}_\Pi(p_i)$ is precisely $\zeta(b)$.  Let $\tilde{B}$

2201: be the connected sum of $B$ and $L$ where $b\in B$ is identified with

2202: $p_i\in L$.  Since $\widetilde{\zeta}_\Pi(p_i) = \zeta(b)$, we may

2203: form a $1$-morphism $\widetilde{\zeta}: \widetilde{B} \rightarrow

2204: \Kgnb{g,r}{X,e}$ such that $\widetilde{\zeta}$ restricted to $B$ is

2205: $\zeta$, and $\widetilde{\zeta}$ restricted to $L$ is

2206: $\widetilde{\zeta}_\Pi$.

2207: 

2208: \begin{notat} \label{not-modif}

2209:   Given a $1$-morphism $\zeta:B \rightarrow \Kgnb{g,r}{X,e}$, a point

2210:   $b\in B$, an irreducible component $L\subset \Sigma_b$, and a marked

2211:   point $p_i\in L$ as above, we call the $1$-morphism

2212:   $\widetilde{\zeta}:\widetilde{B} \rightarrow \Kgnb{g,r}{X,e}$

2213:   constructed in the last paragraph the \emph{modification} of $\zeta$

2214:   determined by $b\in B$, by $L$ and by $p_i$.

2215: \end{notat}

2216: 

2217: \begin{lem} \label{lem-modif}

2218:   Suppose given a $1$-morphism $\zeta:B \rightarrow \Kgnb{0,r}{X,e}$,

2219:   a point $b\in B$, an irreducible component $L\subset \Sigma_b$ which

2220:   is not contracted, and a marked point $p_i\in L$ such that when we

2221:   remove $p_i$, the resulting stable map is a smooth point of

2222:   $\Kgnb{0,r-1}{X,e}$ (this condition is equivalent to the condition

2223:   that the image of $\widetilde{\zeta}_\Pi$ is contained in the smooth

2224:   locus of $\Kgnb{0,r}{X,e}$).  Then we have the vanishing

2225: \begin{equation}

2226: H^1 \lt( L,\widetilde{\zeta}_\Pi^*T_{\Kgnb{0,r}{X,e}} \rt) = 0.

2227: \end{equation}

2228: In particular, if $\zeta:B \rightarrow \Kgnb{0,r}{X,e}$ is a free

2229: morphism of a rational curve into the smooth locus, then there are

2230: deformations of $\widetilde{\zeta}:\widetilde{B} \rightarrow

2231: \Kgnb{0,r}{X,e}$ which smooth the node of $\widetilde{B}$.

2232: \end{lem}

2233: 

2234: \begin{proof}

2235:   This is an application of the deformation theory of section

2236:   ~\ref{sec-def}.  Let $\tau$ denote the dual graph of $\zeta(b)$ and

2237:   let $\psi:\tau \rightarrow \tau'$ be the combinatorial morphism of

2238:   graphs which removes the tail associated to $p_i$.  The morphism

2239:   $\Kbm(X,\psi):\Kbm(X,\tau) \rightarrow \Kbm(X,\tau')$ is smooth

2240:   along the image of $\widetilde{\zeta}_\Pi$ and the pullback of the

2241:   vertical tangent bundle is simply $T_L$.  So

2242:   $\widetilde{\zeta}_\Pi^* T_{\Kbm(X,\tau)}$ is generated by global

2243:   sections.  The pullback of the normal sheaf ${\mc N}$ of

2244:   $\Kbm(X,\tau) \rightarrow \Kgnb{0,r}{X,e}$ is the direct sum over

2245:   all $r_j\in R$ of $N_{F_j/\Lambda} \otimes_{\CC} T_{r_j} M$.  As the

2246:   normal bundle of $N_{F_j/\Lambda}$ is just $\OO_L(-1)$, we conclude

2247:   that $H^1(L,{\mc N}) = 0$.  We have a short exact sequence:

2248: \begin{equation}

2249: \begin{CD}

2250: 0 @>>> \widetilde{\zeta}_\Pi^*T_{\Kbm(X,\tau)} @>>> 

2251: \widetilde{\zeta}_\Pi^*T_{\Kgnb{0,r}{X,e}}  @>>> {\mc N} @>>> 0

2252: \end{CD}

2253: \end{equation}

2254: In the corresponding long exact sequence of cohomology, $H^1$ of the

2255: first and third terms vanishes.  Therefore we have the vanishing

2256: result.

2257: 

2258: \

2259: 

2260: Suppose that $\zeta:B \rightarrow \Kgnb{0,r}{X,e}$ is a free

2261: $1$-morphism of a rational curve into the smooth locus of

2262: $\Kgnb{0,r}{X,e}$.  Then $\zeta^*T_{\Kgnb{0,r}{X,e}}$ is generated by

2263: global sections, so $H^1(B,\zeta^* T_{\Kgnb{0,r}{X,e}}(-b))$ is zero

2264: by lemma~\ref{lem-gend} We have a short exact sequence:

2265: \begin{equation}

2266: \begin{CD}

2267: 0 @>>> \zeta^* T_{\Kgnb{0,r}{X,e}}(-b) @>>> \widetilde{\zeta}^*

2268: T_{\Kgnb{0,r}{X,e}} @>>> \widetilde{\zeta}_\Pi^* T_{\Kgnb{0,r}{X,e}}

2269: @>>> 0

2270: \end{CD}

2271: \end{equation}

2272: In the corresponding long exact sequence of cohomology, $H^1$ of the

2273: first and third terms vanishes.  Therefore

2274: $H^1(\widetilde{B},\widetilde{\zeta}^* T_{\Kgnb{0,r}{X,e}} )$

2275: vanishes.  This cohomology group is the obstruction to smoothing the

2276: node, therefore there are deformations of

2277: $\widetilde{\zeta}:\widetilde{B}\rightarrow \Kgnb{0,r}{X,e}$ which

2278: smooth the node of $\widetilde{B}$.

2279: \end{proof}

2280: 

2281: \begin{rmk} \label{rmk-modif}

2282:   In case the line bundle $\sigma^*\OO_{\Sigma}(\sigma)$ is generated

2283:   by global sections and $L$ contains only one node of $\Sigma_b$, we

2284:   have a simpler proof of the deformation result.  We have a short

2285:   exact sequence of sheaves on $\Sigma$:

2286: \begin{equation}

2287: \begin{CD}

2288: 0 @>>> \OO_{\Sigma}(\sigma) @>>> \OO_{\Sigma}(\sigma + L) @>>>

2289: N_{L/\Sigma}(p_i) @>>> 0 

2290: \end{CD}

2291: \end{equation}

2292: Since $L$ contains only one node, $N_{L/\Sigma}\cong \OO_L(-1)$, so

2293: the last term is isomorphic to $\OO_L$.  So $\OO_{\Sigma}(\sigma+L)$

2294: is generated by global sections.  A small deformation $\sigma'$ of

2295: $\sigma + L$ in the linear series $|\sigma + L|$ will be a section of

2296: $\pi:\Sigma \rightarrow B$, and the stabilization of the $1$-morphism

2297: $B \rightarrow \Kgnb{0,r}{X,e}$ which removes the section $\sigma$

2298: from $\zeta$ and replaces it by $\sigma'$ will be a small deformation

2299: of $\widetilde{\zeta}$ which smooths the node of $\widetilde{B}$.

2300: \end{rmk}

2301: 

2302: \

2303: 

2304: Now we come to the main theorem of this section, which we use for the

2305: induction step in the proof of theorem~\ref{thm-thm1}.

2306: 

2307: \begin{thm} \label{thm-induction}

2308:   Suppose that $X$ satisfies hypothesis~\ref{hyp-1},

2309:   hypothesis~\ref{hyp-1.5}, hypothesis~\ref{hyp-1.75}, and

2310:   hypothesis~\ref{hyp-2}.  For each integer $e\geq 1$, if there exists

2311:   an inductable map $\overline{\zeta}_e: B_e \rightarrow

2312:   \Kgnb{0,1}{X,e}$, then there exists an inductable map

2313:   $\overline{\xi}_{e+1}: B_{e+1} \rightarrow \Kgnb{0,1}{X,e+1}$.

2314:   

2315:   More precisely, suppose that $(\zeta_1,\overline{\zeta}_e)$ is an

2316:   inducting pair.  Let us denote:

2317: \begin{eqnarray}

2318: s=\text{deg}(\zeta_1)^*(2{\mc

2319: L} - {\mc H}), \\

2320: \overline{s}=\text{deg}(\overline{\zeta}_e)^*(2{\mc

2321: L} - {\mc H})

2322: \end{eqnarray}  

2323: Then for each $k=1,\dots,\overline{s}$, there is an inducting pair

2324: $(\xi^k_1,\overline{\xi}^k_{e+1})$ satisfying the following.

2325: \begin{enumerate}

2326:   

2327: \item We have

2328: \begin{equation}

2329: \text{deg}\lt((\xi^k_1)^*{\mc H}\rt) = \text{deg}\lt(\zeta_1^*{\mc H}

2330: \rt) + 2k. 

2331: \end{equation}

2332: 

2333: \item We have

2334: \begin{equation}

2335: \text{deg}\lt( (\xi^k_1)^*{\mc L} \rt) = \text{deg}\lt(

2336: (\overline{\xi}^k_{e+1})^*{\mc L} \rt) = \text{deg}\lt( \zeta_1^*{\mc L}

2337: \rt) + k.

2338: \end{equation}

2339: 

2340: \item We have

2341: \begin{equation}

2342: \text{deg}\lt( (\overline{\xi}^k_{e+1})^*{\mc H} \rt) = \text{deg} \lt(

2343: \zeta_1^*{\mc H} \rt) + \text{deg} \lt( \overline{\zeta}_e^*{\mc H}

2344: \rt).

2345: \end{equation}

2346: 

2347: \item For each $e_1 + e_2 = e$ with both $e_1, e_2 \geq 2$, we have

2348: \begin{equation}

2349: \text{deg}\lt( (\overline{\xi}^k_{e+1})^* \Delta_{(e_1+1,e_2)} \rt) = 

2350: \text{deg}\lt( \overline{\zeta}_{e}^* \Delta_{(e_1,e_2)} \rt).

2351: \end{equation}

2352: 

2353: \item If $e > 1$, we have

2354: \begin{equation}

2355: \text{deg}\lt( (\overline{\xi}^k_{e+1})^* \Delta_{(e,1)} \rt)

2356: =\text{deg}\lt( \overline{\zeta}_{e}^* \Delta_{(e-1,1)} \rt)+ s+ k,

2357: \end{equation}

2358: and we have

2359: \begin{equation}

2360: \text{deg}\lt( (\overline{\xi}^k_{e+1})^* \Delta_{(1,e)} \rt) = \overline{s}-k.

2361: \end{equation}

2362: 

2363: \item If $e=1$, we have

2364: \begin{equation}

2365: \text{deg}\lt( (\overline{\xi}^k_{2})^* \Delta_{(1,1)} \rt) = s+\overline{s}.

2366: \end{equation}

2367: \end{enumerate}

2368: 

2369: \end{thm}

2370: 

2371: \begin{proof}

2372:   By lemma ~\ref{lem-posdef} and lemma ~\ref{lem-inductable}, we may

2373:   suppose that $B_e$ is smooth.  Moreover, in this case

2374:   $\overline{\zeta}_e:B_e \rightarrow \Kgnb{0,1}{X,e}$ is free (in

2375:   fact very free).  Therefore, we may suppose that

2376:   $\overline{\zeta}_e(B_e)$ is in general position: for any finite

2377:   collection of codimension $2$ subvarieties $(Z_\alpha |

2378:   \alpha=1,\dots,M)$ and any finite collection of divisors $(D_\beta |

2379:   \beta=1,\dots, N)$, we may suppose that $\overline{\zeta}_e(B_e)$ is

2380:   disjoint from each $Z_\alpha$ and has $0$-dimensional intersection

2381:   with each $D_\beta$.

2382: 

2383: \

2384:  

2385: Let us denote the family of stable maps $\overline{\zeta}_e$ by:

2386: \begin{equation}

2387: \lt( \overline{p}:\overline{\Sigma} \rightarrow B,

2388: \overline{\sigma}:B \rightarrow \overline{\Sigma}, \overline{g}:

2389: \overline{\Sigma} \rightarrow X \rt).

2390: \end{equation}

2391: And let us denote the family of stable maps $\zeta_1$ by

2392: \begin{equation}

2393: \lt( p:\Sigma \rightarrow B, \sigma:B \rightarrow \Sigma, g: \Sigma

2394: \rightarrow X \rt).

2395: \end{equation}

2396: The basic idea is to form the connected sum of the surfaces $\Sigma$

2397: and $\overline{\Sigma}$ glued along the sections $\sigma$ and

2398: $\overline{\sigma}$.  The actual family $\overline{\zeta}_{e+1}$ is a

2399: bit more complicated.

2400: 

2401: \

2402: 

2403: Define $\pi':\Sigma'\rightarrow B$ to be the family of curves obtained

2404: by taking the connected sum of $\Sigma$ and $\overline{\Sigma}$ glued

2405: along the sections $\sigma$ and $\overline{\sigma}$.  Here $\pi'$ is

2406: the unique morphism such that $\pi'|_{\Sigma} = \pi$ and

2407: $\pi'|_{\overline{\Sigma}} = \overline{\pi}$.  Define $g':\Sigma'

2408: \rightarrow X$ to be the unique morphism such that $g'|_{\Sigma} = g$

2409: and $g'|_{\overline{\Sigma}} = \overline{g}$.  Then $\zeta' =

2410: (\pi':\Sigma'\rightarrow B, g':\Sigma' \rightarrow X)$ is a family of

2411: stable maps in the boundary divisor $\Delta_{e,1}$ of

2412: $\Kgnb{0,0}{X,e+1}$.  Moreover, $\zeta'$ clearly factors through the

2413: Behrend-Manin stack $\Kbm(X,\tau) \rightarrow X$ where $\tau$ is the

2414: genus $0$ stable $A$-graph with two vertices $v_1, v_2$ with

2415: $\beta(v_1)=1$ and $\beta(v_2)=e$.  By lemma~\ref{lem-glue1}, we have

2416: a short exact sequence:

2417: \begin{equation}

2418: \begin{CD}

2419: 0 @>>> \zeta_1^* T_{\text{ev}} @>>> (\zeta')^* T_{\Kbm(X,\tau)} @>>>

2420: \overline{\zeta}_e^* T_{\Kgnb{0,1}{X,e}} @>>> 0

2421: \end{CD}

2422: \end{equation}

2423: Since $\zeta_1$ is very twisting, by definition

2424: ~\ref{defn-verytwisting} we have that $\zeta_1^* T_{\text{ev}}$ is

2425: ample.  Since $\overline{\zeta}_e$ is very positive, by lemma

2426: ~\ref{lem-posdef}, we have that $\overline{\zeta}_e^*

2427: T_{\Kgnb{0,1}{X,e}}$ is ample.  Therefore $(\zeta')^*

2428: T_{\Kbm(X,\tau)}$ is ample.  By lemma~\ref{lem-glue2}, we have a short

2429: exact sequence:

2430: \begin{equation}

2431: 0 \rightarrow (\zeta')^* T_{\Kbm(X,\tau)} \rightarrow (\zeta')^*

2432: T_{\Kgnb{0,0}{X,e+1}} \rightarrow \sigma^* \OO_{\Sigma}(\sigma) \otimes

2433: \overline{\sigma}^* \OO_{\overline{\Sigma}}(\overline{\sigma}) \rightarrow 0

2434: \end{equation}

2435: By definition ~\ref{defn-verytwisting} and definition ~\ref{defn-pos},

2436: we have that both $\sigma^*\OO_{\Sigma}(\sigma)$ and

2437: $\overline{\sigma}^* \OO_{\overline{\Sigma}}(\overline{\sigma})$ are

2438: ample.  Therefore their tensor product is ample, and we conclude that

2439: $(\zeta')^* T_{\Kgnb{0,0}{X,e+1}}$ is ample.

2440: 

2441: 

2442: \

2443: 

2444: Denote by $\overline{s}$ the self-intersection of

2445: $\overline{\sigma}\subset \overline{\Sigma}$, i.e. the degree of the

2446: invertible sheaf

2447: $\overline{\sigma}^*\OO_{\overline{\Sigma}}(\overline{\sigma})$.

2448: Notice that we have

2449: \begin{equation}

2450: \overline{s} = \text{deg}\lt( 2\overline{\zeta}_e^*{\mc L} -

2451: \overline{\zeta}_e^*{\mc H} \rt).

2452: \end{equation}

2453: Let $\sigma':B \rightarrow \overline{\Sigma}$ be a general member of

2454: the linear series of $|\overline{\sigma}|$.  Since

2455: $\overline{\sigma}^*\OO_{\overline{\Sigma}}(\overline{\sigma})$ is

2456: generated by global sections, we can find such a $\sigma'$ which has

2457: transverse intersections with $\overline{\sigma}$ at points

2458: $p_1,\dots,p_{\overline{s}} \in \overline{\Sigma}$.  Define

2459: $b:\widetilde{\overline{\Sigma}}\rightarrow \overline{\Sigma}$ to be

2460: the blowing up of $\overline{\Sigma}$ at the points

2461: $p_1,\dots,p_{\overline{s}}$.  Let

2462: $\widetilde{\overline{\pi}}:\widetilde{\overline{\Sigma}} \rightarrow

2463: \overline{\Sigma}$ denote the projection $\overline{\pi}\circ b$.  Let

2464: $\widetilde{\overline{g}}$ denote $\overline{g}\circ b$.  Let

2465: $\widetilde{\overline{\sigma}}:B \rightarrow

2466: \widetilde{\overline{\Sigma}}$ and $\widetilde{\sigma}:B \rightarrow

2467: \widetilde{\overline{\Sigma}}$ denote the proper transforms of

2468: $\overline{\sigma}$ and $\sigma'$ respectively.  Notice that

2469: $\widetilde{\overline{\sigma}}$ and $\widetilde{\sigma}$ are disjoint

2470: sections.  So the data

2471: \begin{equation}

2472: \lt( \lt(\widetilde{\overline{p}}:\widetilde{\overline{\Sigma}}

2473: \rightarrow B, \widetilde{\overline{\sigma}}, \widetilde{\sigma} \rt),

2474: \widetilde{\overline{f}}: 

2475: \widetilde{\overline{\Sigma}} \rightarrow X \rt)

2476: \end{equation}

2477: is a family of stable pointed maps, i.e. a $1$-morphism

2478: $\widetilde{\overline{\zeta}}_e: B \rightarrow \Kgnb{0,2}{X,e}$. By

2479: the deformation theory in subsection~\ref{subsec-unstable}, we have a

2480: short exact sequence:

2481: \begin{equation}

2482: \begin{CD}

2483: 0 @>>> (\widetilde{\sigma})^*

2484: \OO_{\widetilde{\overline{\Sigma}}}(\widetilde{\sigma}) @>>> \lt( 

2485: \widetilde{\overline{\zeta}}_e \rt)^* T_{\Kgnb{0,2}{X,e}} @>>> \lt(

2486: \overline{\zeta} \rt)_e^* T_{\Kgnb{0,1}{X,e}} @>>> 0

2487: \end{CD}

2488: \end{equation}

2489: Of course we have that $(\widetilde{\sigma})^*

2490: \OO_{\widetilde{\overline{\Sigma}}}(\widetilde{\sigma})$ is the

2491: trivial invertible sheaf $\OO_B$.  In particular, we have that $\lt(

2492: \widetilde{\overline{\zeta}}_e \rt)^* T_{\Kgnb{0,2}{X,e}}$ is

2493: generated by global sections.

2494: 

2495: 

2496: \

2497: 

2498: Define $\widetilde{\pi}:\widetilde{\Sigma} \rightarrow B$ to be the

2499: family of curves obtained by taking the connected sum of $\Sigma$ and

2500: $\widetilde{\overline{\Sigma}}$ glued along the sections $\sigma$ and

2501: $\tau$ respectively.  Here $\widetilde{\pi}$ is the unique morphism

2502: such that $\widetilde{\pi}|_{\widetilde{\overline{\Sigma}}} =

2503: \widetilde{\overline{\pi}}$ and $\widetilde{\pi}|_\Sigma = \pi$.

2504: Define $\widetilde{g}:\widetilde{\Sigma}\rightarrow X$ to be the

2505: unique morphism such that

2506: $\widetilde{g}|_{\widetilde{\overline{\Sigma}}} =

2507: \widetilde{\overline{g}}$ and such that $\widetilde{g}|_{\Sigma} = g$.

2508: Define $\widetilde{\sigma}:B \rightarrow \widetilde{\Sigma}$ to be the

2509: section from the last paragraph.  This gives a family of stable maps:

2510: \begin{equation}

2511: \widetilde{\zeta} = \lt( \lt( \widetilde{\pi}:\widetilde{\Sigma}

2512: \rightarrow B, \widetilde{\sigma}:B \rightarrow \widetilde{\Sigma}

2513: \rt), \widetilde{g}: \widetilde{\Sigma} \rightarrow X \rt).

2514: \end{equation}

2515: Define $\tau$ to be the unique genus $0$ stable $A$-graph with two

2516: vertices $v_1,v_2$, with one edge joining $v_1$ and $v_2$, with one

2517: flag attached to $v_1$, with $\beta(v_1)=e$, and with $\beta(v_2)=1$.

2518: In other words, $\tau$ is the dual graph of a stable map with one

2519: marked point, with reducible domain consisting of two irreducible

2520: components, where the component with the marked point has degree $e$

2521: and where the other component has degree $1$.  Then

2522: $\widetilde{\zeta}:B \rightarrow \Kgnb{0,1}{X,e+1}$ is a $1$-morphism

2523: which factors through the Behrend-Manin stack $\Kbm(X,\tau)

2524: \rightarrow \Kgnb{0,1}{X,e+1}$.  By lemma~\ref{lem-glue1}, we have a

2525: short exact sequence:

2526: \begin{equation}

2527: \begin{CD}

2528: 0 @>>> \zeta_1^* T_{\text{ev}} @>>> \lt( \widetilde{\zeta} \rt)^*

2529: T_{\Kbm(X,\tau)} @>>>

2530: \lt( \widetilde{\overline{\zeta}}_e \rt)^* T_{\Kgnb{0,2}{X,e}} @>>> 0

2531: \end{CD}

2532: \end{equation}

2533: As the first and third term in this exact sequence are generated by

2534: global sections, also $\lt( \widetilde{\zeta} \rt)^*

2535: T_{\Kbm(X,\gamma)}$ is generated by global sections.  Finally, by

2536: lemma~\ref{lem-glue2} we have a short exact sequence:

2537: \begin{equation}

2538: 0 \rightarrow \lt( \widetilde{\zeta} \rt)^* T_{\Kbm(X,\gamma)} \rightarrow \lt(

2539: \widetilde{\zeta} \rt)^* T_{\Kgnb{0,1}{X,e+1}} \rightarrow \sigma^*

2540: \OO_{\Sigma}(\sigma) \otimes_{\OO_B} \lt(

2541: \widetilde{\overline{\sigma}} \rt)^*

2542: \OO_{\widetilde{\overline{\Sigma}}}(\widetilde{\overline{\sigma}})

2543: \rightarrow 0

2544: \end{equation}

2545: Of course $\lt( \widetilde{\overline{\sigma}} \rt)^*

2546: \OO_{\widetilde{\overline{\Sigma}}}(\widetilde{\overline{\sigma}})$ is

2547: the trivial invertible sheaf.  And $\sigma^* \OO_{\Sigma}(\sigma)$ is

2548: just $\OO_B(s)$.  Thus the last term in the sequence is an ample

2549: invertible sheaf.  In particular, $\lt( \widetilde{\zeta} \rt)^*

2550: T_{\Kgnb{0,1}{X,e+1}}$ is generated by global sections.  So the

2551: $1$-morphism $\widetilde{\zeta}: B \rightarrow \Kgnb{0,1}{X,e+1}$ is

2552: free.  Also, the pullback by $\widetilde{\zeta}$ of the normal sheaf

2553: of the unramified $1$-morphism, $\Kbm(X,\tau) \rightarrow

2554: \Kgnb{0,1}{X,e+1}$ is the last term in the sequence, and so has

2555: positive degree.

2556: 

2557: \

2558:   

2559: One final remark, when $e>1$, the image of $\widetilde{\zeta}$

2560: intersects the divisor $\Delta_{1,e}$ transversely precisely at the

2561: images of the points $p_1,\dots, p_{\overline{s}} \in B$, in

2562: particular the degree of the $\QQ$-Cartier divisor class

2563: $\widetilde{\sigma}^*\OO_{\Kgnb{0,1}{X,e+1}}(\Delta_{1,e})$ is

2564: positive.  In the special case $e=1$, we have that

2565: $\Delta_{1,e}=\Delta_{e,1} =\Delta_{1,1}$.  In this case

2566: $\Kbm(X,\tau)$ is the normalization of $\Delta_{1,1}$ (at least in a

2567: neighborhood of the image of $\widetilde{\sigma}$).  So the degree of

2568: $\widetilde{\sigma}^*\OO_{\Kgnb{-,1}{X,e+1}}(\Delta_{1,1})$ is the sum

2569: of the degree of the pullback of the normal sheaf of

2570: $\Kbm(X,\tau)\rightarrow \Kgnb{0,1}{X,e+1}$ and the divisor $p_1 +

2571: \dots + p_{\overline{s}}$, i.e. $s+ \overline{s}$.  So in this case,

2572: the degree is again positive.

2573: 

2574: \

2575: 

2576: Since $\widetilde{\zeta}:B \rightarrow \Kgnb{0,1}{X,e+1}$ is free we

2577: can find deformations of $\widetilde{\zeta}$ which are in general

2578: position.  Now by hypothesis ~\ref{hyp-1}, the locus parametrizing

2579: stable maps with at least three irreducible components in their domain

2580: has codimension $2$.  By hypothesis~\ref{hyp-1.75} the locus

2581: parametrizing stable maps with automorphisms has codimension at least

2582: $2$.  Therefore we can find a deformation $\overline{\xi}_{e+1}^0:B

2583: \rightarrow \Kgnb{0,1}{X,e+1}$ of $\widetilde{\zeta}$ such that the

2584: image of $\overline{\xi}_{e+1}^0:B \rightarrow \Kgnb{0,1}{X,e+1}$

2585: misses the locus parametrizing stable maps with at least three

2586: irreducible components in their domain and misses the locus

2587: parametrizing stable maps with automorphisms.  As well, we can assume

2588: that the pullback of $T_{\Kgnb{0,0}{X,e+1}}$ by $\text{pr}\circ

2589: \overline{\xi}_{e+1}^0: B \rightarrow \Kgnb{0,0}{X,e+1}$ is an ample

2590: vector bundle since the pullback of $T_{\Kgnb{0,0}{X,e+1}}$ by

2591: $\text{pr}\circ \widetilde{\zeta}$, i.e. by $\zeta':B \rightarrow

2592: \Kgnb{0,0}{X,e+1}$, is an ample bundle.  Let us denote

2593: $\overline{\xi}_{e+1}^0$ by

2594: \begin{equation}

2595: \overline{\xi}_{e+1}^0 = \lt( \lt( \varpi^0:\Xi^0 \rightarrow B, \lambda^0:B

2596: \rightarrow \Xi^0 \rt), g^0:\Xi^0 \rightarrow B \rt).

2597: \end{equation}

2598: Define $h^0:B \rightarrow X$ to be $g^0\circ \lambda^0$.  By

2599: assumption, $\overline{h}:B \rightarrow X$ is very twistable, and

2600: $h^0:B \rightarrow X$ is a small deformation of $\overline{h}$.  So by

2601: proposition ~\ref{prop-twistopen}, $h^0$ is very twistable.

2602: 

2603: \

2604: 

2605: Since

2606: $\widetilde{\sigma}^*\OO_{\widetilde{\Sigma}}(\widetilde{\sigma})$ is

2607: trivial, we may also assume that $\lt( \lambda^0 \rt)^*

2608: \OO_{\Xi^0}(\lambda^0)$ is trivial.  Finally, we may assume that all

2609: intersections of the image of $\overline{\xi}_{e+1}^0$ and the divisor

2610: $\Delta_{1,e}$ are transverse and occur at general points of

2611: $\Delta_{1,e}$.  In particular, if $b\in B$ is such a point, we may

2612: assume that the corresponding stable map $\overline{\xi}_{e+1}^0(b)$

2613: is of the form $((\Xi^0_b,\lambda^0_b),h^0_b:\Xi^0_b \rightarrow X)$

2614: where $\Xi^0_b$ is a reducible curve $L \cup M$, with $L\cap M$

2615: consisting of one node of $\Xi^0_b$ which is a general point of $X$,

2616: with $\lambda^0_q \in L$, and such that $L \rightarrow X$ is an

2617: embedding of a line which is twistable.

2618: 

2619: \

2620: 

2621: Now we define $\widetilde{\overline{\xi}}^0_{e+1}$ to be the

2622: modification of $\overline{\xi}^0_{e+1}$ associated to the point $b\in

2623: B$, to $L\subset \Xi^0_b$ and $\lambda^0_q \in L$ (c.f.

2624: notation~\ref{not-modif}).  We saw above that $h^0:B \rightarrow X$ is

2625: very twistable.  And the evaluation map $\widetilde{h}_L:L \rightarrow

2626: X$ associated to the restriction $\widetilde{\overline{\xi}}^0_{e+1}:

2627: L \rightarrow \Kgnb{0,1}{X,e+1}$ is just the embedding $L\subset X$,

2628: which is twistable.  By hypothesis~\ref{hyp-1.5} and the assumption

2629: that $L\cap M$ is a general point of $X$, we see that the hypotheses

2630: of lemma~\ref{lem-twisttogether2} are satisfied.  So by

2631: lemma~\ref{lem-twisttogether2} we have that the evaluation map

2632: $\widetilde{h}:\widetilde{B}\rightarrow X$ associated to

2633: $\widetilde{\overline{\xi}}^0_{e+1}$ is also very twistable.  Also,

2634: notice that the image of $\widetilde{\overline{\xi}}^0_{e+1}$ is

2635: contained in the locus of very stable maps.

2636: 

2637: \

2638: 

2639: The $1$-morphism $\widetilde{\overline{\xi}}^0_{e+1}$ satifies the

2640: criterion of lemma~\ref{lem-modif}, in fact the criterion of

2641: remark~\ref{rmk-modif}, thus we can smooth the node of

2642: $\widetilde{B}$.  Let $\overline{\xi}^1_{e+1}:B \rightarrow $ denote a

2643: small deformation which smooths the node of $\widetilde{B}$.  We will

2644: denote this by

2645: \begin{equation}

2646: \overline{\xi}^1_{e+1} = \lt( \lt( \pi^1:\Xi^1\rightarrow B, \sigma^1:

2647: B \rightarrow \Xi^1 \rt), g^1: \Xi^1 \rightarrow X \rt).

2648: \end{equation}

2649: The claim is that $\overline{\xi}^1_{e+1}$ is inductable.  Denote $h^1

2650: = g^1\circ \sigma^1$.  Since $\widetilde{h}$ is very twistable and

2651: since $h^1$ is a small deformation of $\widetilde{h}$, it follows by

2652: proposition~\ref{prop-twistopen} that $h^1$ is very twistable.  Now

2653: the image of $\widetilde{\overline{\xi}}^0_{e+1}$ is not contained in

2654: the locus of very stable maps, precisely because of the point $r\in L$

2655: where $L\cap M = \{r\}$: the stable map

2656: $\widetilde{\overline{\xi}}^0_{e+1}(r)$ is not very stable.  But from

2657: the description of the smoothing of

2658: $\widetilde{\overline{\xi}}^0_{e+1}$ given in remark~\ref{rmk-modif},

2659: we can find small deformations whose image is contained in the very

2660: stable locus.  So we may assume the image of $\overline{\xi}^1_{e+1}$

2661: is contained in the locus of very stable maps.  It remains only to

2662: show that $\overline{\xi}^1_{e+1}$ is positive.

2663: 

2664: \

2665: 

2666: That $h^1:B \rightarrow T$ is a stable map follows from the fact that

2667: $B$ is smooth and $h^1$ is non-constant: in fact the degree of

2668: $h^1(B)$ equals the degree of $\widetilde{h}(\widetilde{B})$, which is

2669: just $\text{deg}(h(B)) +1$.  This proves ($1$) of

2670: definition~\ref{defn-pos}.  To show that $h^1:B\rightarrow X$ is

2671: unobstructed, it suffices to prove that

2672: $\widetilde{h}:\widetilde{B}\rightarrow X$ is unobstructed.  By

2673: assumption, $h^0:B \rightarrow X$ is unobstructed (being a small

2674: deformation of $h:B \rightarrow X$).  And the restriction to $L$ of

2675: $T_X$ is generated by global sections (since $L$ is general).  Thus

2676: $\widetilde{h}:\widetilde{B} \rightarrow X$ is unobstructed.  This

2677: proves ($2$) of definition~\ref{defn-pos}. The map

2678: \begin{equation}

2679: \text{pr}\circ \widetilde{\overline{\xi}}^0_{e+1}: \widetilde{B}

2680: \rightarrow \Kgnb{0,0}{X,e}

2681: \end{equation}

2682: is constant on $L\subset \widetilde{B}$, and on $B\subset

2683: \widetilde{B}$, it is just $\zeta'$.  Thus the image is contained in

2684: the unobstructed locus.  So we have ($3$) of definition

2685: ~\ref{defn-pos}.  The pullback of $T_{\Kgnb{0,0}{X,e}}$ to

2686: $\widetilde{B}$ is trivial on $L$ and equals

2687: $(\zeta')^*T_{\Kgnb{0,0}{X,e}}$ on $B$.  Since

2688: $(\zeta')^*T_{\Kgnb{0,0}{X,e}}$ is ample, we conclude from

2689: lemma~\ref{lem-DAcrit} that the pullback to $\widetilde{B}$ of

2690: $T_{\Kgnb{0,0}{X,e}}$ is deformation ample.  Since

2691: $\overline{\xi}^1_{e+1}$ is a small deformation of

2692: $\widetilde{\overline{\xi}}^0_{e+1}$, the pullback of

2693: $T_{\Kgnb{0,0}{X,e}}$ via $\overline{\xi}^1_{e+1}$ is deformation

2694: ample by lemma~\ref{lem-opDA}.  This proves ($4$) of definiton

2695: ~\ref{defn-pos}.  Now for $\widetilde{\overline{\zeta}}^0_{e+1}$, the

2696: pullback of $\widetilde{\sigma}^*\OO_{\widetilde{\Sigma}}(\sigma)$ is

2697: trivial when restricted to $B\subset \widetilde{B}$, but on $L\subset

2698: \widetilde{B}$ it is $\OO_L(1)$.  So by lemma~\ref{lem-DAcrit}, it is

2699: deformation ample.  Since $\overline{\zeta}^1_{e+1}$ is a small

2700: deformation of $\widetilde{\overline{\xi}}^0_{e+1}$, by

2701: lemma~\ref{lem-opDA} we have that $(\sigma^1)^*\OO_{\Xi^1}(\sigma^1)$

2702: is deformation ample.  So $\overline{\xi}^1_{e+1}$ is positive.

2703: Therefore it is inductable.

2704: 

2705: \

2706: 

2707: To define the $1$-morphisms $\overline{\xi}^k_{e+1}:B \rightarrow

2708: \Kgnb{0,1}{X,e+1}$ we repeat the procedure above.  Given

2709: $\overline{\xi}^k_{e+1}$ and a point $b\in B$ whose image is a general

2710: point of $\Delta_{(1,e)}$, say $(\Xi^k_b,\lambda^k_b,g^k_b:\Xi^k_b

2711: \rightarrow X)$ with $\Xi^k_b = L \cup M$, $\lambda^k_b \in L$ and

2712: $g^k_b:L \rightarrow X$ an embedding of a twistable line, we define

2713: $\widetilde{B}^k$ to be the connected sum of $B$ and $L$ with $b\in B$

2714: identified with $\lambda^k_b \in L$.  Then we define

2715: \begin{equation}

2716: \widetilde{\overline{\xi}}^k_{e+1}:\widetilde{B}^k \rightarrow

2717: \Kgnb{0,1}{X,e+1}

2718: \end{equation}

2719: to be the modification of $\overline{\xi}^k_{e+1}$ associated to $b\in

2720: B$, $L\subset \Xi^k_b$ and $\lambda^k_b \in L$.  By the same argument

2721: as above, deformations of $\widetilde{\overline{\xi}}^k_{e+1}$ smooth

2722: the node of $\widetilde{B}^k$, and a small deformation

2723: $\overline{\xi}^{k+1}_{e+1}$ of $\widetilde{\overline{\xi}}^k_{e+1}$

2724: is inductable.

2725: 

2726: \

2727: 

2728: It is quite simple to work out the degrees of the pullbacks by

2729: $\widetilde{\overline{\xi}}^0_{e+1}$ of the tautological divisors of

2730: $\Kgnb{0,1}{\PP^N,e+1}$ in terms of the degrees of the pullbacks by

2731: $\zeta_1$ and $\overline{\zeta}_e$ of the tautological divisors.  This

2732: is left to the interested reader.

2733: \end{proof}

2734: 

2735: \section{Twistable lines} \label{sec-twist}

2736: 

2737: In this section, we will prove that if $n+1\geq d^2$, then for a

2738: general hypersurface $X_d \subset \PP^n$ of degree $d$ and a general

2739: line $L\subset X$, we have that $L$ is a twistable line on $X$.  To

2740: prove this we introduce some incidence correspondences.  Let $N_d =

2741: \binom{n+d}{n}-1$ and let $\PP^{N_d}$ denote the projective space

2742: parametrizing hypersurfaces $X_d\subset \PP^n$ of degree $d$.  Let

2743: ${\mc X} \subset \PP^{N_d} \times \PP^n$ denote the universal family

2744: of degree $d$ hypersurfaces in $\PP^n$.  Let $\GG(1,n)$ denote the

2745: Grassmannian variety of lines in $\PP^n$.  Let $F({\mc X}) \subset

2746: \PP^{N_d} \times \GG(1,n)$ denote the parameter space of pairs

2747: $([X],[L])$ where $X\subset \PP^n$ is a hypersurface of degree $d$,

2748: $L\subset \PP^n$ is a line and $L\subset X$.  Observe that the

2749: projection $F({\mc X}) \rightarrow \GG(1,d)$ is a projective bundle of

2750: relative dimension $N_d - (d+1)$.

2751: 

2752: \

2753: 

2754: Let $P(t) = (t+1)^2$ denote the Hilbert polynomial of a quadric

2755: surface in $\PP^3$, and let $U \subset \textit{Hilb}^{P(t)}({\PP^n})$

2756: denote the open subscheme parametrizing subschemes $\Sigma \subset

2757: \PP^n$ which are projectively equivalent to a smooth, quadric surface

2758: in $\PP^3 \subset \PP^n$.  Let $V \subset U \times \GG(1,n)$ denote

2759: the parameter space of pairs $([\Sigma],[L])$ where $\Sigma \subset

2760: \PP^n$ is a smooth quadric surface, where $L\subset \PP^n$ is a line,

2761: and where $L\subset \Sigma$.  The projection map $V\rightarrow U$ has

2762: a Stein factorization $V \rightarrow \tilde{U} \rightarrow U$ where

2763: $\tilde{U} \rightarrow U$ is a finite, \'etale double cover, and where

2764: $V\rightarrow \tilde{U}$ is a $\PP^1$-bundle.  Let $W\subset \PP^{N_d}

2765: \times U \times \GG(1,n)$ denote the parameter space of triples

2766: $([X],[\Sigma],[L])$ where $X\subset \PP^n$ is a hypersurface of

2767: degree $d$, $([\Sigma],[L])$ is a point of $V$ and where $\Sigma

2768: \subset X$.  The projection map $W \rightarrow V$ is a projective

2769: bundle of relative dimension $N_d - (d+1)^2$.

2770: 

2771: \

2772: 

2773: Now for a triple $([X],[\Sigma],[L])\in W$, there is a map

2774: (well-defined up to nonzero scalar) $d_X:\CC^{n+1} \rightarrow

2775: H^0(\PP^n,\OO_{\PP^n}(d-1))$ which evaluates the partial derivatives

2776: of a defining equation of $X$.  We may compose this map with the

2777: restriction map $H^0(\PP^n,\OO_{\PP^n}(d)) \rightarrow

2778: H^0(\Sigma,\OO_{\Sigma}(d-1))$. Let this map be denoted by

2779: $d_{X,\Sigma}:\CC^{n+1} \rightarrow H^0(\Sigma,\OO_{\Sigma}(d-1))$.

2780: More precisely, let $E$ be the trivial vector bundle on $W$ of rank

2781: $n+1$, let $G$ be the vector bundle on $U$ whose fiber at a point

2782: $\Sigma$ is just $H^0(\Sigma,\OO_{\Sigma}(d-1))$, and let $F$ be the

2783: vector bundle on $W$ which is

2784: $\text{pr}_1^*(\OO_{\PP^{N_d}}(1))\otimes \text{pr}_2^*G$.  Then there

2785: is a map of vector bundles $d:E \rightarrow F$ whose fiber over

2786: $([X],[\Sigma],[L])$ is the map $d_{X,\Sigma}$ constructed above.  Let

2787: $W^o\subset W$ denote the open subscheme (possibly empty) over which

2788: $d$ is surjective (i.e. the complement of the support of the cokernel

2789: of $d$).

2790: 

2791: \begin{lem} \label{lem-quadric1}

2792:   For any point $([X],[\Sigma],[L])\in W^o$, we have

2793: \begin{enumerate}

2794: 

2795: \item $X$ is smooth along $\Sigma$,

2796: 

2797: \item $H^i\lt( \Sigma, N_{\Sigma/X} \rt)$ is zero for $i>0$,

2798: 

2799: \item $H^i\lt( \Sigma,N_{\Sigma/X}(-L) \rt)$ is zero for $i>0$,

2800:   

2801: \item $H^i\lt( \Sigma,N_{\Sigma/X}\otimes \OO_{\Sigma}(-1) \rt)$ is

2802:   zero for $i>0$,

2803: 

2804: \item $H^1\lt( L, N_{L/X} \rt)$ is zero,

2805: 

2806: \item $H^1\lt( L, N_{L/X}(-1) \rt)$ is zero,

2807:   

2808: \item the projection morphism $W \rightarrow \PP^{N_d}$ given by

2809:   $([X],[\Sigma],[L]) \mapsto [X]$ is smooth at $([X],[\Sigma],[L])$,

2810:   

2811: \item the projection morphism $F({\mc X}) \rightarrow \PP^{N_d}$ given

2812:   by $([X],[L]) \mapsto [X]$ is smooth at $([X],[L])$, and

2813: 

2814: \item the projection morphism $\pi:W\rightarrow F({\mc X})$ given by

2815:   $([X],[\Sigma],[L]) \mapsto ([X],[L])$ is smooth at

2816:   $([X],[\Sigma],[L])$.  

2817: \end{enumerate}

2818: \end{lem}

2819: 

2820: \begin{proof}

2821:   Since the partial derivatives of a defining equation of $X$ generate

2822:   $H^0(\Sigma,\OO_{\Sigma}(d-1))$, in particular the locus where they

2823:   all vanish is disjoint from $\Sigma$.  By the Jacobian criterion, we

2824:   conclude that $X$ is smooth along $\Sigma$.

2825: 

2826: \

2827: 

2828: For a smooth quadric surface $\Sigma \subset \PP^3 \subset \PP^n$, we

2829: have a short exact sequence:

2830: \begin{equation}

2831: \begin{CD}

2832: 0 @>>> N_{\Sigma/\PP^3} @>>> N_{\Sigma/\PP^n} @>>>

2833: N_{\PP^3/\PP^n}|_{\Sigma} @>>> 0

2834: \end{CD}

2835: \end{equation}

2836: Since $N_{\Sigma/\PP^3} \cong \OO_{\Sigma}(2)$ and since

2837: $N_{\PP^3/\PP^n}|_{\Sigma} \cong \OO_{\Sigma}(1)^{\oplus (n-3)}$, we

2838: have a short exact sequence:

2839: \begin{equation}

2840: \begin{CD}

2841: 0 @>>> \OO_{\Sigma}(2) @>>> N_{\Sigma/\PP^n} @>>>

2842: \OO_{\Sigma}(1)^{\oplus (n-3)} @>>> 0

2843: \end{CD}

2844: \end{equation}

2845: From this it is easy to compute that $H^i\lt( \Sigma,N_{\Sigma/\PP^n}

2846: \rt)$, $H^i\lt( \Sigma,N_{\Sigma/\PP^n}\otimes \OO_{\Sigma}(-1) \rt)$

2847: and $H^i\lt( \Sigma,N_{\Sigma/\PP^n}\otimes \OO_{\Sigma}(-L) \rt)$ are

2848: all zero for $i>0$.

2849: 

2850: \

2851: 

2852: There is a short exact sequence:

2853: \begin{equation}

2854: \begin{CD}

2855: 0 @>>> N_{\Sigma/X} @>>> N_{\Sigma/\PP^n} @>>> N_{X/\PP^n}|_\Sigma

2856: @>>> 0

2857: \end{CD}

2858: \end{equation}

2859: Of course $N_{X/\PP^n}|_{\Sigma} \cong \OO_{\Sigma}(d)$, and for ${\mc

2860:   L} = \OO_{\Sigma}$, for ${\mc L} = \OO_{\Sigma}(-L)$, and for ${\mc

2861:   L} = \OO_{\Sigma}(-1)$, we compute that $H^i\lt( \Sigma,

2862: \OO_{\Sigma}(d) \otimes {\mc L} \rt)$ is zero for $i>0$.  The bundle

2863: $N_{\Sigma/\PP^n}$ was computed in the last paragraph.  For the line

2864: bundles ${\mc L} = \OO_\Sigma$, ${\mc L} = \OO_{\Sigma}(-L)$ and ${\mc

2865:   L} = \OO_{\Sigma}(-1)$, we computed that $H^i(\Sigma,

2866: N_{\Sigma/\PP^n}\otimes {\mc L})$ is zero for $i>0$.  It immediately

2867: follows from the long exact sequence in cohomology that $H^2\lt(

2868: \Sigma, N_{\Sigma/X} \otimes {\mc L} \rt)$ is zero.  We also conclude

2869: that $H^1(\Sigma, N_{\Sigma/X}\otimes {\mc L})$ is zero iff the

2870: corresponding map

2871: \begin{equation}

2872: H^0(\Sigma, N_{\Sigma/\PP^n}\otimes {\mc L}) \rightarrow H^0(\Sigma,

2873: N_{X/\PP^n}|_\Sigma) 

2874: \end{equation}

2875: is surjective.  

2876: 

2877: \

2878: 

2879: In the case that ${\mc L} = \OO_\Sigma(-1)$, the map from the last

2880: paragraph factors the map

2881: \begin{equation}

2882: H^0(\PP^n,T_{\PP^n}(-1)) \rightarrow H^0(\Sigma,\OO_{\Sigma}(d-1)).

2883: \end{equation}

2884: But this map is precisely the map $d_{X,\Sigma}:\CC^{n+1} \rightarrow

2885: H^0(\Sigma,\OO_{\Sigma}(d-1))$.  Since $d_{X,\Sigma}$ is surjective,

2886: we conclude that $H^1(\Sigma, N_{\Sigma/X}\otimes \OO_{\Sigma}(-1))$

2887: is zero.

2888: 

2889: \

2890: 

2891: To see that $H^1(\Sigma,N_{\Sigma/X}\otimes \OO_{\Sigma}(-L))$ is

2892: zero, observe that we have $\OO_{\Sigma}(1) \cong \OO_{\Sigma}(L+L')$

2893: where $L'\subset \Sigma$ is any line in the ruling opposite to the

2894: ruling of $L$.  Thus we have a commutative diagram:

2895: \begin{equation}

2896: \begin{CD}

2897: H^0( \Sigma , N_{\Sigma/\PP^n}(-1) ) \otimes_{\CC}

2898: H^0( \Sigma, \OO_{\Sigma}(L') ) @>>> H^0( \Sigma, N_{\Sigma/\PP^n}(-L) \\

2899: @VVV @VVV \\

2900: H^0( \Sigma , N_{X/\PP^n}|_{\Sigma}(-1) )

2901: \otimes_{\CC} H^0( \Sigma , \OO_{\Sigma}(L') ) @>>>

2902: H^0( \Sigma, N_{X/\PP^n}|_{\Sigma}(-L) ) \\

2903: \end{CD}

2904: \end{equation}

2905: The top vertical horizontal arrow is surjective by the last paragraph.

2906: Moreover, the right vertical arrow is

2907: \begin{equation}

2908: \begin{array}{cc}

2909: H^0\lt( \Sigma,

2910: \OO_{\Sigma} \lt( (d-1)L + (d-1)L' \rt) \rt) \otimes H^0\lt( \Sigma,

2911: \OO_{\Sigma}(L') \rt) \ \ \ \ \ \ \ \ \ \ \ \ \ \\

2912: \ \ \ \ \ \ \ \ \ \ \ \ \rightarrow 

2913: H^0\lt( \Sigma, \OO_{\Sigma}\lt( (d-1)L + dL' \rt) \rt),

2914: \end{array}

2915: \end{equation}

2916: which is clearly surjective.  Therefore we conclude that the bottom

2917: horizontal arrow is also surjective, i.e. $H^1(\Sigma,

2918: N_{\Sigma/X}(-L) )$ is zero.  The proof that $H^1(\Sigma,

2919: N_{\Sigma/X})$ is zero is almost identical to the proof that

2920: $H^1(\Sigma, N_{\Sigma/X}(-L)$ is zero and is left to the reader.

2921: 

2922: \

2923: 

2924: To see that $H^1\lt( L, N_{L/X}(-1) \rt)$ is zero, first observe we

2925: have a short exact sequence:

2926: \begin{equation}

2927: \begin{CD}

2928: 0 @>>> N_{\Sigma/X}(-1) @>>> N_{\Sigma/X}(-L') @>>> N_{\Sigma/X}|_L(-1) @>>> 0

2929: \end{CD}

2930: \end{equation}

2931: By our computations and the long exact sequence in cohomology, we

2932: conclude that $H^1\lt( L, N_{\Sigma/X}|_L(-1) \rt)$ is zero.  Next

2933: observe that we have a short exact sequence:

2934: \begin{equation}

2935: \begin{CD}

2936: 0 @>>> N_{L/\Sigma}(-1) @>>> N_{L/X}(-1) @>>> N_{\Sigma/X}|_L(-1) @>>> 0

2937: \end{CD}

2938: \end{equation}

2939: Of course $N_{L/\Sigma} \cong \OO_L(1)$, so $H^1\lt(

2940: L,N_{L/\Sigma}(-1) \rt)$ is zero.  And we have seen that $H^1\lt(

2941: L,N_{\Sigma/X}|_L(-1) \rt)$ is zero.  Therefore by the long exact

2942: sequence in cohomology, we conclude that $H^1\lt(L, N_{L/X}(-1) \rt)$

2943: is zero. By an almost identical argument, we also conclude that

2944: $H^1\lt( L, N_{L/X} \rt)$ is zero.

2945: 

2946: \

2947: 

2948: Now by \cite[proposition I.2.14.2]{K}, the obstruction space for the

2949: relative Hilbert scheme $\textit{Hilb}^{P(t)}({\mc X}/\PP^{N_d})$ at a

2950: point $([X],[\Sigma])$ is contained in $H^1(\Sigma,N_{\Sigma/X})$.

2951: Since the obstruction space vanishes, it follows by \cite[theorem

2952: 2.10]{K} that $\textit{Hilb}^{P(t)}({\mc X}/\PP^{N_d}) \rightarrow

2953: \PP^{N_d}$ is smooth at $([X],[\Sigma])$.  As we have seen $W

2954: \rightarrow \textit{Hilb}^{P(t)}({\mc X}/\PP^{N_d})$ is smooth.

2955: Therefore the composition $W \rightarrow \PP^{N_d}$ is smooth along

2956: $W^o$.

2957: 

2958: \

2959: 

2960: For basically the same reason as above, we conclude that the

2961: projection map $F({\mc X}) \rightarrow \PP^{N_d}$ is smooth along the

2962: image of $\pi:W^o \rightarrow F({\mc X})$.  Since $W^o \rightarrow

2963: \PP^{N_d}$ is smooth, and since $F({\mc X}) \rightarrow \PP^{N_d}$ is

2964: smooth along the image of $W^o$, to prove that $\pi$ is smooth along

2965: $W^o$, it suffices to check that the derivative map $d\pi:

2966: T_{W^o/\PP^{N_d}} \rightarrow \pi^* T_{F({\mc X})/\PP^{N_d}}$ is

2967: surjective at every point.  This exactly reduces to the statement that

2968: $H^0 \lt( \Sigma, N_{\Sigma/X} \rt) \rightarrow H^0\lt( L,

2969: N_{\Sigma/X}|_L \rt)$ is surjective.  Since the cokernel is contained

2970: in $H^1\lt( \Sigma, N_{\Sigma/X}(-L) \rt)$, which is zero, we conclude

2971: the map is surjective.  Therefore $\pi:W^o \rightarrow F({\mc X})$ is

2972: smooth.

2973: \end{proof}

2974: 

2975: Now suppose that $([X],[\Sigma],[L])$ is a point in $W^o$.  We

2976: associate to this triple a morphism $\zeta: L \rightarrow

2977: \Kgnb{0,1}{X,1}$ as follows.  Let $\sigma:L \rightarrow \Sigma$ be the

2978: inclusion and let $\text{pr}_L: \Sigma \rightarrow L$ be the unique

2979: projection such that $\sigma$ is a section of $\text{pr}_L$.  Let

2980: $g:\Sigma \rightarrow X$ be the inclusion.  Then we have a morphism

2981: $\zeta:L \rightarrow \Kgnb{0,1}{X,1}$ given by the data:

2982: \begin{equation}

2983: \zeta = \lt(\text{pr}_L:\Sigma \rightarrow L, \sigma:L \rightarrow

2984: \Sigma, g:\Sigma 

2985: \rightarrow X \rt).

2986: \end{equation}

2987: 

2988: \begin{lem} \label{lem-quadric2}

2989:   If $([X],[\Sigma],[L])$ is a triple in $W^o$ and if $X$ is smooth,

2990:   then the corresponding morphism $\zeta:L \rightarrow

2991:   \Kgnb{0,1}{X,1}$ is flexible.

2992: \end{lem}

2993: 

2994: \begin{proof}

2995:   We need to check the axioms of definition~\ref{defn-twisting}.

2996:   Since $g\circ \sigma:L \rightarrow X$ is an embedding, in particular

2997:   this map is stable, i.e. axiom $(1)$ is satisfied.  By part $(5)$ of

2998:   lemma~\ref{lem-quadric1}, we conclude that $\Kgnb{0,0}{X,1}$ is

2999:   unobstructed at $[g\circ \sigma: L \rightarrow X]$, i.e. axiom $(2)$

3000:   is satisfied.

3001: 

3002: \

3003: 

3004: To check axiom $(3)$, consider the normal bundle ${\mc N}$ of the

3005: regular embedding $(\text{pr}_L, g): \Sigma \rightarrow L \times X$.

3006: This fits into a short exact sequence:

3007: \begin{equation}

3008: \begin{CD}

3009: 0 @>>> \text{pr}_L^* T_L @>>> {\mc N} @>>> N_{\Sigma/X} @>>> 0

3010: \end{CD}

3011: \end{equation}

3012: By part $(3)$ of remark~\ref{rmk-twisting}, we need to check that $R^1

3013: \lt(\text{pr}_L\rt)_* {\mc N}(-\sigma)$ is zero.  

3014: It is clear that for each fiber $L'$ of $\text{pr}_L:\Sigma

3015: \rightarrow L$, we have that ${\mc N}(-\sigma)|_{L'}$ is just

3016: $N_{L'/X}(-1)$.  By part $(6)$ of lemma~\ref{lem-quadric1}, we

3017: conclude that $H^1\lt(L',N_{L'/X}(-1)\rt)$ is zero.  Therefore we

3018: conclude that $R^1 \lt( \text{pr}_L \rt)_* {\mc N}(-\sigma)$ is zero.

3019: 

3020: \

3021: 

3022: By part $(3)$ of remark~\ref{rmk-twisting}, we have that $\zeta^*

3023: T_{\text{ev}}$ is equivalent to $\lt( \text{pr}_L \rt)_* {\mc

3024:   N}(-\sigma)$.  Twisting the short exact sequence above by

3025: $\OO_{\Sigma}(-L)$ and applying the long exact sequence of higher

3026: direct images, we see that $\lt( \text{pr}_L \rt)_* {\mc N}(-\sigma)$

3027: fits between $\lt( \text{pr}_L \rt)_* \text{pr}_L^* T_L(-L)$ and $\lt(

3028: \text{pr}_L \rt)_* N_{\Sigma/X}(-L)$ with cokernel $R^1 \lt(

3029: \text{pr}_L \rt)_* \text{pr}_L^* T_L(-L)$.  For any fiber $L'$ of

3030: $\text{pr}_L$, we have $T_L(-L)|_{L'}$ is isomorphic to

3031: $\OO_{L'}(-1)$.  Therefore $\lt(\text{pr}_L\rt)_* \text{pr}_L^*

3032: T_L(-1)$ and $R^1 \lt(\text{pr}_L \rt)_* \text{pr}_L^* T_L(-L)$ are

3033: both zero, i.e. $\lt( \text{pr}_L \rt)_* {\mc N}(-L)$ is isomorphic to

3034: $\lt( \text{pr}_L \rt)_* N_{\Sigma/X}(-L)$.

3035: 

3036: \

3037: 

3038: To show that axiom $(4)$ holds, we want to prove that for any point

3039: $p'\in L$ with corresponding fiber $L' =

3040: \text{pr}_L^{-1}\lt\{p'\rt\}$, we have that $\lt( \text{pr}_L \rt)_*

3041: N_{\Sigma/X}(-L)\otimes \OO_L(-p')$ has no $H^1$.  Observe that since

3042: $R^1\lt( \text{pr}_L \rt)_* \text{pr}_L^* T_L(-L-L')$ and $R^1 \lt(

3043: \text{pr}_L \rt)_* {\mc N}(-L -L')$ are both zero, it follows from the

3044: long exact sequence of higher direct images that also $R^1 \lt(

3045: \text{pr}_L \rt)_* N_{\Sigma/X}(-L-L')$ is zero.  Therefore by the

3046: Leray spectral sequence, we conclude that $H^1\lt( \Sigma,

3047: N_{\Sigma/X}(-L-L') \rt)$ equals $H^1 \lt( L, \lt( \text{pr}_L

3048: \rt)_*\lt( N_{\Sigma/X}(-L) \rt)(-p') \rt)$.  But by part $(4)$ of

3049: lemma~\ref{lem-quadric1}, this is zero.  Thus axiom $(4)$ is

3050: satisfied.

3051: 

3052: \

3053: 

3054: Finally, observe that $\sigma^* \OO_{\Sigma}(\sigma)$ is the trivial

3055: line bundle, so axiom $(5)$ is satisfied.

3056: \end{proof}

3057: 

3058: By lemma~\ref{lem-quadric2}, for any pair $([X],[L])\in F({\mc X})$,

3059: if we can find a corresponding triple $([X],[\Sigma],[L])\in W^o$,

3060: then it follows that $L$ is a twistable line on $X$.  Now we come to

3061: the main result of this section.

3062: 

3063: \begin{prop} \label{prop-quadric}

3064:   If $n+1 \geq d^2$ and $d\geq 2$, then $W^o\rightarrow F({\mc X})$ is

3065:   dominant.  Thus for a general pair $([X],[L])\in F({\mc X})$, we

3066:   have that $L$ is a twistable line on $X$.

3067: \end{prop}

3068: 

3069: \begin{proof}

3070:   By part $(9)$ of lemma~\ref{lem-quadric1}, it suffices to prove that

3071:   $W^o$ is nonempty.  Let $I_d$ be the set of pairs of integers

3072:   $I_d=\{(i,j) :0\leq i,j \leq d-1, i+j \geq 2\}$.  Choose coordinates

3073:   on $\PP^n$ of the form $(Y_0,Y_1,Y_2,Y_3) \cup \lt( X_{i,j}

3074:   \rt)_{(i,j)\in I_d} \cup (Z_m: m = 1,\dots, n+1-d^2)$.  Let $\Sigma

3075:   \subset \PP^n$ be the smooth quadric surface with ideal

3076: \begin{equation}

3077: I_{\Sigma} = \langle Y_0 Y_3 - Y_1 Y_2 \rangle + \langle  X_{i,j} |

3078: (i,j) \in I_d \rangle + \langle Z_m | m= 1, \dots, n+1-d^2 \rangle.

3079: \end{equation}

3080: This is the image of the embedding $f:\PP^1 \times \PP^1 \rightarrow

3081: \PP^n$ given by sending a point $\lt( [U_0:U_1], [V_0:V_1] \rt)\in

3082: \PP^1 \times \PP^1$ to

3083: \begin{equation}

3084: \lt( [U_0:U_1],[V_0:V_1] \rt) \mapsto [U_0V_0: U_0V_1: U_1V_0: U_1V_1:

3085: 0 : \dots :0 ].

3086: \end{equation}

3087: 

3088: \

3089: 

3090: We make the following convention.  Given a pair $(i,j)$ of integers,

3091: we set $k=\min(i,j)$, we set $i'=i-k$ and we set $j'=j-k$.  Consider

3092: the hypersurface $X\subset \PP^n$ with defining equation

3093: \begin{equation}

3094: F = \lt( Y_0 Y_3 - Y_1 Y_2 \rt) Y_3^{d-2} + \sum_{(i,j) \in I_d} Y_0^k

3095: Y_1^{i'} Y_2^{j'} Y_3 X_{i,j}.

3096: \end{equation}

3097: It is clear that $\Sigma \subset X$.  We claim that the derivative map

3098: $dF:\CC^{n+1} \rightarrow H^0\lt(\Sigma,\OO_{\Sigma}(d-1) \rt)$ is

3099: surjective.  Observe first that we have

3100: \begin{eqnarray}

3101: \frac{\partial F}{\partial Y_0} \mapsto U_1^{d-1}V_1^{d-1},

3102: \frac{\partial F}{\partial Y_1} \mapsto U_1^{d-1}V_0V_1^{d-2}, \\

3103: \frac{\partial F}{\partial Y_2} \mapsto U_0U_1^{d-2}V_1^{d-1},

3104: \frac{\partial F}{\partial Y_3} \mapsto U_0U_1^{d-2}V_0V_1^{d-2}.

3105: \end{eqnarray}

3106: And observe that for $(i,j)\in I_d$, we have that

3107: \begin{equation}

3108: \frac{\partial F}{\partial X_{i,j}} \mapsto

3109: U_0^iU_1^{d-1-i}V_0^jV_1^{d-1-j}.

3110: \end{equation}

3111: Since the partial derivatives of the form $\frac{\partial F}{\partial

3112:   Y_i}$ give the terms $U_0^iU_1^{d-1-i}V_0^jV_1^{d-1-j}$ with $(i,j)

3113: = (0,0), (0,1), (1,0),$ and $(1,1)$, and since these are precisely the

3114: pairs $(i,j)$ not contained in $I_d$, we conclude that $dF$ is

3115: surjective.  Thus, for any line $L\subset \Sigma$, we have that

3116: $([X],[\Sigma],[L])$ is in $W^o$.

3117: \end{proof}

3118: 

3119: \section{A very positive, very twisting family of lines}

3120: 

3121: In the last section, we proved that if $n+1 \geq d^2$, and $d\geq 2$,

3122: then for a general hypersurface $X_d\subset \PP^n$ of degree $d$, a

3123: general line $L\subset X$ is twistable, in other words

3124: hypothesis~\ref{hyp-2} holds.  In this section, we will prove that if

3125: $n\geq d^2 + d+2$, and $d\geq 3$ then there exists a morphism

3126: $\overline{\zeta}_1:\PP^1 \rightarrow \Kgnb{0,1}{X,1}$ which is both

3127: very twistable and very positive.  This provides the base case for the

3128: induction argument of section~\ref{sec-induct}.

3129: 

3130: \

3131: 

3132: The arguments in this section are very similar to those of the last

3133: section.  In that section, the key result was that for $([X],[L])$

3134: general, there is a quadric surface $\Sigma$ with $L\subset \Sigma

3135: \subset X$ such that $H^i\lt(\Sigma,N_{\Sigma/X}(-1)\rt)$ is zero for

3136: $i>0$.  This result in turn reduced to finding a single degree $d$

3137: polynomial $F$ on $\PP^n$, vanishing on some quadric surface $\Sigma$,

3138: and such that

3139: \begin{equation}

3140: d_{F,\Sigma}:\CC^{n+1} \rightarrow H^0\lt(\Sigma,

3141: \OO_{\Sigma}(d-1)\rt)

3142: \end{equation}

3143: is surjective.  

3144: 

3145: \

3146: 

3147: In this section, the role of $L\subset X$ will be replaced by a

3148: rational normal curve $C_0\subset X$ of some degree $k\leq n$ (in the

3149: end we will only need the case $k=2d-3$).  The role of the quadric

3150: surface will be replace by a rational normal scroll $\Sigma$ of degree

3151: $2k-1$ such that $C_0 \subset \Sigma \subset X$.  The cohomology

3152: vanishing result of the last section will be replaced by the vanishing

3153: of $H^i\lt(\Sigma, N_{\Sigma/X}(-C_0 - 2L) \rt)$ for $i>0$, where $L$

3154: is any line of ruling of $\Sigma$.  The computation in this section

3155: will be to find a single degree $d$ polynomial $F$ on $\PP^n$,

3156: vanishing on $\Sigma$, and such that the image of the map,

3157: \begin{equation}

3158: d_{F,\Sigma}: \CC^{n+1} \rightarrow H^0\lt( \Sigma,\OO_{\Sigma}(d-1)

3159: \rt),

3160: \end{equation}

3161: let's call it $W\subset H^0\lt( \Sigma, \OO_{\Sigma}(d-1) \rt)$, has

3162: the property that the induced map

3163: \begin{equation}

3164: W \otimes H^0 \lt(\Sigma,\OO_{\Sigma}\lt( (k-3)L \rt) \rt) \rightarrow

3165: H^0 \lt( \Sigma, \OO_{\Sigma}(d-1) \otimes \OO_{\Sigma}\lt( (k-3) L

3166: \rt) \rt)

3167: \end{equation}

3168: is surjective.  A similar polynomial $F$ to that of the last section

3169: satisfies this condition.

3170: 

3171: \subsection{Generating Linear Systems on ${\mathbb F}_1$}

3172: In the last section, the relevant surface was the Hirzebruch surface

3173: ${\mathbb F}_0 = \PP^1 \times \PP^1$ embedded as a quadric surface.

3174: In this section, the relevant surface is the Hirzebruch surface

3175: ${\mathbb F}_1$ embedded as a rational normal scroll of degree $2k-1$.

3176: We will perform our computations using the projective model of

3177: ${\mathbb F}_1$:

3178: \begin{equation}

3179: {\mathbb F}_1 = \lt\{ ([T_0:T_1],[T_0U:T_1U:V]) \in \PP^1 \times \PP^2

3180: | T_0 (T_1U) = T_1 (T_0U) \rt\}.

3181: \end{equation}

3182: In the equation above, $T_0U$ and $T_1U$ are simply variables on

3183: $\PP^2$.  We denote the projection maps by $\text{pr}_1:{\mathbb F}_1

3184: \rightarrow \PP^1$ and $\text{pr}_2: {\mathbb F}_1 \rightarrow \PP^2$.

3185: We denote by $\OO_{{\mathbb F}_1}(F)$ the line bundle $\text{pr}_1^*

3186: \OO_{\PP^1}$ and by $\OO_{{\mathbb F}_1}(E+F)$ the line bundle

3187: $\text{pr}_2^* \OO_{\PP^2}$.  Here $\OO_{{\mathbb F}_1}(E)$ is the

3188: divisor class of the directrix $E\subset {\mathbb F}_1$.  This

3189: explains our terminology $T_0U$ and $T_1U$: $U$ is a nonzero element

3190: of $H^0\lt({\mathbb F}_1,\OO_{{\mathbb F}_1}(E) \rt)$, and $T_0U$ and

3191: $T_1U$ are the products of $U$ with the two global sections $T_0$ and

3192: $T_1$ of $H^0\lt({\mathbb F}_1, \OO_{{\mathbb F}_1}(F) \rt)$.

3193: 

3194: \

3195: 

3196: We note that the line bundles $\OO_{{\mathbb F}_1}(E+F)$ and

3197: $\OO_{{\mathbb F}_1}(F)$ generate the Picard group of ${\mathbb F}_1$.

3198: Thus we adopt the terminology for line bundles on ${\mathbb F}_1$:

3199: \begin{equation}

3200: \OO(a,b) := \OO_{{\mathbb F}_1}\lt( a(E+F) + bF \rt).

3201: \end{equation}

3202: Note that $E+F$ and $F$ are each NEF, but not ample.  We conclude that

3203: these two line bundles generate the NEF cone.  Thus every NEF line

3204: bundle on ${\mathbb F}_1$ is of the form $\OO(a,b)$ for some

3205: nonnegative integers $a,b$.

3206: 

3207: \

3208: 

3209: Now suppose that $\OO(a,b)$ is some NEF line bundle and $W\subset

3210: H^0\lt( {\mathbb F}_1, \OO(a,b) \rt)$ is a linear series.

3211: 

3212: \begin{defn} \label{defn-cgen}

3213:   For an integer $c\geq 0$, we say that $W$ is a $c$-\emph{generating

3214:     linear system}, if the associated map

3215: \begin{equation}

3216: \mu_{W,c}: W \otimes H^0\lt({\mathbb F}_1, \OO(0,c)\rt)

3217: \rightarrow H^0\lt( {\mathbb F}_1, \OO(a,b+c) \rt)

3218: \end{equation}

3219: is surjective.

3220: \end{defn}

3221: 

3222: The question we want to answer is, when is $W$ a $c$-generating linear

3223: system.  In particular, what is the minimal necessary dimension of a

3224: $c$-generating linear system?  To simplify the answer, we write $b-1 =

3225: \beta_d(c+1) + \beta_r$ where $\beta_d,\beta_r$ are integers with $0

3226: \leq \beta_r < c+1$, and we write $a+b-1 = \alpha_d(c+1) + \alpha_r$

3227: where $\alpha_d, \alpha_r$ are integers with $0\leq \alpha_r < c+1$.

3228: 

3229: \

3230: 

3231: \begin{lem} \label{lem-comput}

3232:   Define the functions

3233: \begin{eqnarray}

3234: M(a,b,c) = a^2 + \lt(2b+3(c+1) \rt)a + 2b +4(c+1) + 2, \\

3235: E(a,b,c) = \lt(\beta_r^2 - (c+1)\beta_r\rt) - 

3236: \lt(\alpha_r^2 - (c-1)\alpha_r\rt) 

3237: \end{eqnarray}

3238: The minimal necessary dimension for a $c$-generating linear system

3239: \begin{equation}

3240: W\subset H^0\lt( {\mathbb F}_1, \OO(a,b)

3241: \rt)

3242: \end{equation}

3243: is $\text{dim}(W) = \frac{1}{2(c+1)}\lt( M(a,b,c) + E(a,b,c) \rt)$.

3244: \end{lem}

3245: 

3246: \begin{proof}

3247:   This is just a computation.  For any nonnegative integers $a',b'$

3248:   there is a decreasing filtration on $H^0\lt( {\mathbb F}_1,

3249:   \OO(a',b') \rt)$ given by

3250: \begin{equation}

3251: F^iH^0\lt( {\mathbb F}_1, \OO(a',b')\rt) = H^0\lt( {\mathbb F}_1,

3252:  \OO(a',b')(-iE)  \rt). 

3253: \end{equation}

3254: For any linear system $W\subset H^0\lt( {\mathbb F}_1, \OO_(a,b)

3255: \rt)$, there is an induced filtration $F^iW = F^i\cap W$.  And the

3256: multiplication map $\mu_{W,c}$ respects the filtrations on $W$ and on

3257: $H^0\lt( {\mathbb F}_1, \OO(a,b+c) \rt)$.  If $\mu_{W,c}$ is

3258: surjective, then the associated graded pieces

3259: \begin{equation}

3260: \text{gr}^i\mu_{W,c}: \text{gr}^iW \otimes H^0\lt( {\mathbb F}_1,

3261: \OO(0,c) \rt) \rightarrow \text{gr}^i H^0\lt( {\mathbb

3262: F}_1, \OO(a,b+c) \rt)

3263: \end{equation}

3264: are all surjective.  As the dimension of $W$ is the sum of the

3265: dimensions of the pieces $\text{gr}^i W$, we should compute the

3266: minimum possible dimension of a vector subspace $W^i \subset

3267: \text{gr}^i H^0\lt( {\mathbb F}_1, \OO(a,b) \rt)$ such that

3268: $\text{gr}^i\mu_{W^i,c}$ is surjective (where $\text{gr}^i\mu_{W^i,c}$

3269: is the obvious map).

3270: 

3271: \

3272: 

3273: Of course the associated graded parts for $\OO(a',b')$ are just

3274: \begin{equation}

3275: \text{gr}^i H^0\lt( {\mathbb F}_1, \OO(a',b') \rt)

3276:        = \lt\{ \begin{array}{ll}

3277:                   H^0\lt( E, \OO_E(b'+i) \rt), & 0\leq i \leq a' \\

3278:                   \{0\}, & i > a' 

3279:                 \end{array} \rt.

3280: \end{equation}

3281: So we are looking for a subset $W^i \subset H^0\lt( E, \OO_E(b+i)

3282: \rt)$ such that the multiplication map

3283: \begin{equation}

3284: \text{gr}^i \mu_{W^i,c}: W^i \otimes H^0\lt( E, \OO_E(c) \rt)

3285: \rightarrow H^0\lt( E, \OO_E(b+c+i) \rt)

3286: \end{equation}

3287: is surjective.  Counting the dimensions of the spaces on the left and

3288: the right, we have

3289: \begin{equation}

3290: \text{dim}(W^i)\lt(c+1\rt) \geq \lt(b+c+i+1\rt),

3291: \end{equation}

3292: in other words,

3293: \begin{equation}

3294: \text{dim}(W^i) \geq \lt\lfloor\frac{b+i-1}{c+1} \rt\rfloor +2.

3295: \end{equation}

3296: On the other hand, we can acheive this bound: simply take $W^i$ to be

3297: generated by the set of monomials:

3298: \begin{equation}

3299: \{ U^i V^{a-i} T_0^{(b+i)-j(c+1)}T_1^{j(c+1)}| j=1,\dots,r\}\cup  

3300: \{ U^i V^{a-i} T_1^{b+i}\}

3301: \end{equation}

3302: where $r = \lt\lfloor\frac{b+i-1}{c+1} \rt\rfloor$.  Thus we have that

3303: the minimum necessary dimension for a $c$-generating linear series is

3304: \begin{equation}

3305: \text{dim}(W) = 2a+2 + \sum_{i=0}^a \lt\lfloor \frac{b+i-1}{c+1}

3306: \rt\rfloor.

3307: \end{equation}

3308: If we write $b-1 = \beta_d(c+1) + \beta_r$ with $0\leq \beta_r < c+1$

3309: and if we write $a+b-1 = \alpha_d(c+1) + \alpha_r$ with $0 \leq

3310: \alpha_r < c+1$, then the sum above is just

3311: \begin{equation}

3312: 2a+2 + \sum_{i=0}^a \lt\lfloor \frac{b+i-1}{c+1} \rt\rfloor =

3313: \frac{1}{2(c+1)}\lt( 

3314: M(a,b,c) + E(a,b,c) \rt)

3315: \end{equation}

3316: where $M(a,b,c)$ and $E(a,b,c)$ are as above.

3317: \end{proof}

3318: 

3319: For the next result, we introduce the Cox homogeneous coordinate ring:

3320: \begin{equation}

3321: S = S({\mathbb F}_1) := \CC[T_0,T_1,U,V] = \oplus_{(a,b) \in \ZZ^2}

3322: H^0\lt( {\mathbb F}_1, \OO(a,b) \rt). 

3323: \end{equation}

3324: This is a $\ZZ^2$-graded ring, graded by $\text{deg}(T_0) =

3325: \text{deg}(T_1) = (0,1)$, $\text{deg}(V) = (1,0)$ and $\text{deg}(U) =

3326: (1,-1)$.  For any multi-degree $(a,b)\in \ZZ^2$, we have that

3327: $S_{(a,b)} = H^0\lt( {\mathbb F}_1, \OO(a,b) \rt)$.  We put a graded

3328: lexicographical monomial order on $S$ where the grading is by total

3329: degree ($\text{deg(a,b)} = a+b$), and where $U>V>T_0>T_1$.

3330: 

3331: \

3332: 

3333: In the proof of the lemma above, a special role was played by the

3334: linear system

3335: \begin{eqnarray}

3336: W_0(a,b,c) = \text{span}\lt\{ U^i V^{a-i}

3337: T_0^{(b+i)-j(c+1)}T_1^{j(c+1)}|i=0,\dots, a,

3338: j=1,\dots,r(i) \rt\}  \\

3339: + \text{span}\lt\{U^i V^{a-i}

3340: T_1^{b+i}\rt|i=0,\dots,a\}.

3341: \end{eqnarray} 

3342: Here $r(i) = \lt\lfloor\frac{b+i-1}{c+1} \rt\rfloor$.

3343: 

3344: \begin{lem}\label{lem-comput2}

3345:   Suppose $W\subset H^0\lt( {\mathbb F}_1, \OO(a,b) \rt)$ is a linear

3346:   system.  If the vector space of initial terms $\text{IN}(W)$

3347:   contains $W_0(a,b,c)$, then $W$ is a $c$-generating linear system.

3348: \end{lem}

3349: 

3350: \begin{proof}

3351:   Clearly we have that the vector space of initial terms of

3352:   $\text{image}(\mu_{W,c})$ satisfies

3353: \begin{equation}

3354: \text{IN}(W)\cdot S_{(0,c)} \subset \text{IN}\lt(\text{image}(\mu_{W,c})

3355: \rt).

3356: \end{equation}

3357: So if $\text{IN}(W)$ contains $W_0(a,b,c)$, then we have that

3358: \begin{equation}

3359: W_0(a,b,c)\cdot S_{(0,c)} \subset \text{IN}\lt( \text{image}(\mu_{W,c})

3360: \rt).

3361: \end{equation}

3362: By construction, $W_0(a,b,c)\cdot S_{(0,c)} = S_{(a,b+c)}$.  So we

3363: have $\text{IN}\lt( \text{image}(\mu_{W,c}) \rt) = S_{(a,b+c)}$, and

3364: therefore $\text{image}(\mu_{W,c}) = S_{(a,b+c)}$.  Therefore $W$ is a

3365: $c$-generating linear system.

3366: \end{proof}

3367: 

3368: An important special case for us is when $a=d-1, b=(d-1)(k-1)$ and

3369: $c=k-3$ for some positive integers $d$ and $k$ (here $d$ will be the

3370: degree of the hypersurface $X\subset \PP^n$, and $k$ will be the

3371: degree of the curve $C_0\subset X$).  In particular, if $k\geq 2d-3$,

3372: then we have $b-1 = (d-1)(k-2) + d-2$ and $a+b-1 = (d-1)(k-2) + 2d-3$.

3373: So the formulas above reduce to

3374: \begin{equation}

3375: \begin{array}{rcl}

3376: M(d-1,(d-1)(k-1),k-3) & =  & 

3377: 2(k-2)(d-1)^2 + \\

3378: 5(k-2)(d-1) + 

3379: 4(k-2) & + &

3380: \lt(3(d-1)^2 + 2(d-1) -2 \rt), \\

3381: E(d-1,(d-1)(k-2),k-3) & = & 

3382: (k-2)(d-1) - \\

3383: \lt( 3(d-1)^2 + 2(d-1) - 2 \rt)

3384: \end{array}

3385: \end{equation}

3386: So, if $k\geq 2d-3$, we have

3387: \begin{equation}

3388: \frac{1}{2(k-2)}(M+E) = (d-1)^2 + 3(d-1) + 2 = d^2 + d.

3389: \end{equation}

3390: 

3391: \subsection{Cohomology Results}

3392: We introduce some incidence correspondences, analogous to those

3393: introduced in section~\ref{sec-twist}.  Let $N_d = \binom{n+d}{n} -1$

3394: and let $\PP^{N_d}$ denote the projective space parametrizing

3395: hypersurfaces $X_d \subset \PP^n$ of degree $d$.  Let ${\mc X} \subset

3396: \PP^{N_d} \times \PP^n$ denote the universal family of degree $d$

3397: hypersurfaces in $\PP^n$.  Let $k$ be any integer with $1\leq k \leq

3398: \frac{n}{2}$ (later we will only need the case that $k=2d-3$).  Let

3399: ${\mc R}^k(\PP^n) \subset \textit{Hilb}^{kt+1}(\PP^n)$ denote the open

3400: subscheme parametrizing curves $C_0\subset \PP^n$ which are

3401: projectively equivalent to a degree $k$ rational normal curve $C_0

3402: \subset \PP^k \subset \PP^n$.  Observe that ${\mc R}^k\lt(\PP^n\rt)$

3403: is a homogeneous space of $\text{PGL}_{n+1}$, and therefore is smooth

3404: and connected.  Let ${\mc R}^k({\mc X}) \subset \PP^{N_d} \times {\mc

3405:   R}^k(\PP^n)$ denote the parameter space for pairs $([X],[C_0])$

3406: where $C_0\subset X$.  Observe that the projection ${\mc R}^k({\mc X})

3407: \rightarrow {\mc R}^k(\PP^n)$ is a projective bundle of relative

3408: dimension $N_d - (kd+1)$.

3409: 

3410: \

3411: 

3412: Let $Q(t) = \frac{1}{2}(t+1)((2k-1)t+2)$ denote the Hilbert polynomial

3413: of a rational normal scroll of degree $2k-1$ in $\PP^{2k}$.  Let ${\mc

3414:   U}\subset \textit{Hilb}^{Q(t)}({\PP^n})$ denote the open subscheme

3415: parametrizing subschemes $\Sigma \subset \PP^n$ which are projectively

3416: equivalent to a rational normal scroll of degree $2k-1$ in

3417: $\PP^{2k}\subset \PP^n$ which is abstractly isomorphic to ${\mathbb

3418:   F}_1$.  Let ${\mc V}\subset {\mc U} \times {\mc R}^k(\PP^n)$ denote

3419: the parameter space of pairs $([\Sigma],[C_0])$ where $C_0\subset

3420: \Sigma$ and such that, using the isomorphism of $\Sigma$ and ${\mathbb

3421:   F}_1$, the line bundle of $C_0$ is $\OO(1,0)$.  The projection map

3422: ${\mc V} \rightarrow {\mc U}$ factors as an open subset (with nonempty

3423: fibers) of a projective bundle over ${\mc U}$ of relative dimension

3424: $2$ (actually each fiber is isomorphic to the $\AAA^2$ of irreducible

3425: curves in the linear system $|\OO(1,0)|$).

3426: 

3427: \

3428: 

3429: Let ${\mc W} \subset \PP^{N_d}\times {\mc U} \times {\mc R}^k(\PP^n)$

3430: denote the parameter space for triples $([X],[\Sigma],[C_0])$ where

3431: $([\Sigma],[C_0])$ is in ${\mc V}$ and where $\Sigma \subset X$.  The

3432: projection map ${\mc W} \rightarrow {\mc V}$ is a projective bundle of

3433: relative dimension $N_d - Q(d)$.

3434: 

3435: \

3436: 

3437: Now for a triple $([X],[\Sigma],[C_0]) \in {\mc W}$, we define

3438: $d_{X,\Sigma}: \CC^{n+1} \rightarrow H^0\lt( \Sigma, \OO_{\Sigma}(d-1)

3439: \rt)$ as in section~\ref{sec-twist}.  More precisely, let ${\mc E}$ be

3440: the trivial vector bundle on ${\mc W}$ of rank $n+1$, let ${\mc G}$ be

3441: the vector on ${\mc U}$ whose fiber at a point $\Sigma$ is just

3442: $H^0\lt(\Sigma, \OO_{\Sigma}(d-1)\rt)$, and let ${\mc F}$ be the

3443: vector bundle on ${\mc W}$ which is

3444: $\text{pr}_1^*(\OO_{\PP^{N_d}}(1))\otimes \text{pr}_2^* {\mc G}$.

3445: Then there is a map of vector bundles $d:{\mc E} \rightarrow {\mc F}$

3446: whose fiber over $([X],[\Sigma],[C_0])$ is the map $d_{X,\Sigma}$

3447: constructed above.  Let ${\mc W}^o\subset {\mc W}$ be the open

3448: subscheme (possibly empty) parametrizing pairs $([X],[\Sigma],[C_0])$

3449: such that the image of $d_{X,\Sigma}$ is a $(k-3)$-generating linear

3450: series in $H^0\lt( \Sigma, \OO_{\PP^n}(d-3)|_\Sigma \rt)$.

3451: 

3452: \begin{lem} \label{lem-scroll1}

3453:   Let $f:{\mathbb F}_1 \rightarrow \Sigma$ be an isomorphism to a

3454:   degree $2k-1$ rational normal scroll $\Sigma \subset \PP^{2k}\subset

3455:   \PP^n$.  For each pair of integers $a,b \geq 0$, consider the bundle

3456: \begin{equation}

3457: N(a,b) = f^*\lt( N_{\Sigma/\PP^n(-1}) \rt) \otimes 

3458:   \OO(a,b) 

3459: \end{equation}

3460: and the subbundle

3461: \begin{equation}

3462: N'(a,b) = f^*\lt( N_{\Sigma/\PP^{2k} } \rt) \otimes

3463:   \OO(a,b).

3464: \end{equation}

3465: Then we have the following:

3466: \begin{enumerate}

3467: 

3468: \item $N'(0,0)$ is generated by global sections and satisfies $H^i\lt(

3469:   {\mathbb F}_1, N'(0,0) \rt)$ is zero for $i>0$,

3470: 

3471: \item $N(0,0)$ is generated by global sections and satisfies $H^i\lt(

3472:   {\mathbb F}_1, N(0,0) \rt)$ is zero for $i>0$,

3473: 

3474: \item if ${\mathcal F}$ is any coherent sheaf on ${\mathbb F}_1$ which

3475:   is generated by global sections and satisfies $H^i\lt( {\mathbb

3476:   F}_1, {\mathcal F} \rt)$ is zero for $i>0$, then for every pair of

3477:   nonnegative integers $(a,b)$ we have that ${\mathcal F}(a,b) := {\mathcal

3478:   F}\otimes \OO(a,b)$ is generated by global

3479:   sections and satisfes $H^i\lt( {\mathbb F}_1, {\mathcal F}(a,b)

3480:   \rt)$ is zero for $i>0$.

3481: 

3482: \end{enumerate}

3483: In particular, we conclude that for any pair of nonnegative integers

3484: $(a,b)$, we have that $N(a,b)$ (resp. $N'(a,b)$) is generated by

3485: global sections and satisfies $H^i\lt({\mathbb F}_1, N(a,b)\rt)$ is

3486: zero for $i>0$ (resp. $H^i\lt( {\mathbb F}_1, N'(a,b) \rt)$ is zero

3487: for $i>0$).

3488: \end{lem}

3489: 

3490: \begin{proof}

3491:   Recall that $\text{pr}_1:{\mathbb F}_1 \rightarrow \PP^1$ is the

3492:   projection morphism such that $\text{pr}_1^*\OO_{\PP^1}(1) =

3493:   \OO_{{\mathbb F}_1}(f)$.  Via $\text{pr}_1$, ${\mathbb F}_1$ is

3494:   isomorphic as a $\PP^1$-scheme to the total space of the projective

3495:   bundle:

3496: \begin{equation}

3497: {\mathbb F}_1 \cong \PP\lt( \OO_{\PP^1}(-k) \oplus \OO_{\PP^1}(-(k-1))

3498: \rt).

3499: \end{equation}

3500: Under this isomorphism $\OO(1,k-1)$ corresponds to the tautological

3501: quotient bundle $\OO(1)$ on $\PP\lt( \OO_{\PP^1}(-k) \oplus

3502: \OO_{\PP^1}(-(k-1)) \rt)$.  In other words, up to projective

3503: equivalence, the map $f:{\mathbb F}_1 \rightarrow \PP^{2k}$

3504: corresponds to the complete linear system of $\OO(1)$, i.e.

3505: $f^*\OO_{\PP^{2k}}(1) \cong \OO(1)$.  Using this identification, it is

3506: easy to see that we have a short exact sequence of vector bundles on

3507: ${\mathbb F}_1$:

3508: \begin{equation}

3509: 0 \rightarrow \text{pr}_1^* T_{\PP^1} \rightarrow \text{pr}_1^*\lt(

3510: \OO_{\PP^1}(1)^{\oplus (2k-1)} \rt)\otimes f^*\OO_{\PP^{2k}}(1) \rightarrow

3511: f^* N_{\Sigma/\PP^{2k}} \rightarrow 0

3512: \end{equation}

3513: Twisting by $f^*\OO_{\PP^{2k}}(-1)$, we observe that $N'(0,0)$ is a

3514: quotient of $\text{pr}_1^*\lt(\OO_{\PP^1}(1)^{\oplus (2k-1)} \rt)$,

3515: and so is generated by global sections.  Observe that we have the

3516: vanishing

3517: \begin{equation}

3518: \lt(\text{pr}_1\rt)_*\lt(\text{pr}_1^*T_{\PP^1}\otimes

3519: f^*\OO_{\PP^{2k}}(-1) \rt) =

3520: R^1\lt(\text{pr}_1\rt)_*\lt(\text{pr}_1^*T_{\PP^1}\otimes 

3521: f^*\OO_{\PP^{2k}}(-1) \rt) = \{0\}.

3522: \end{equation}

3523: Applying the long exact sequence of higher direct images to our short

3524: exact sequence (after twisting by $f^*\OO_{\PP^{2k}}(-1)$), we have

3525: that $R^1\lt( \text{pr}_1 \rt)_* \lt(f^*N_{\Sigma/\PP^{2k}}(-1) \rt)$

3526: is zero, and $\lt( \text{pr}_1 \rt)_* \lt(f^*N_{\Sigma/\PP^{2k}}(-1)

3527: \rt)$ is $\OO_{\PP^1}(1)^{\oplus (2k-1)}$.  From this and the Leray

3528: spectral sequence associated to $\text{pr}_1:{\mathbb F}_1 \rightarrow

3529: \PP^1$, we conclude that $H^i\lt({\mathbb F}_1, N'(0,0) \rt)$ is zero

3530: for $i>0$.  This proves item $(1)$ of the lemma.

3531: 

3532: \

3533: 

3534: Also observe that we have a short exact sequence:

3535: \begin{equation}

3536: \begin{CD}

3537: 0 @>>> N'(0,0) @>>> N(0,0) @>>> \OO(0,0)^{\oplus (n-2k)}

3538: @>>> 0

3539: \end{CD}

3540: \end{equation}

3541: Since $H^i\lt( {\mathbb F}_1, \OO_{{\mathbb F}_1} \rt)$ is zero for

3542: $i>0$, we conclude that $N(0,0)$ is generated by global sections and

3543: satisfies $H^i\lt({\mathbb F}_1, N(0,0) \rt)$ is zero for $i>0$.

3544: This proves item $(2)$ of the lemma.

3545: 

3546: \

3547: 

3548: Now suppose that ${\mathcal F}$ is a coherent sheaf on ${\mathbb F}_1$

3549: such that ${\mathcal F}$ is generated by global sections and such that

3550: $H^i\lt( {\mathbb F}_1, {\mathcal F} \rt)$ is zero for $i>0$.  We will

3551: prove by double induction on $(a,b)$ that the same is true for

3552: ${\mathcal F}(a,b) := {\mathcal F}\otimes \OO(a,b)$.

3553: 

3554: \

3555: 

3556: First we prove the result when $b=0$.  We proceed by induction on $a$.

3557: If $a=0$, the result is tautological.  Thus suppose that $a>0$ and

3558: suppose the result is proved for $a-1$.  Let $D\subset {\mathbb F}_1$

3559: be a general member of the linear system $|\OO(1,0)|$.  Then $D$ is a

3560: smooth curve isomorphic to $\PP^1$.  Since $D$ is general, we have a

3561: short exact sequence:

3562: \begin{equation}

3563: \begin{CD}

3564: 0 @>>> {\mathcal F}(a-1,0) @>>> {\mathcal F}(a,0) @>>> {\mathcal F}(a,0)|_D

3565: @>>> 0

3566: \end{CD}

3567: \end{equation}

3568: Now $\mathcal{F}|_D$ is generated by global sections.  And

3569: $\OO_{{\mathbb F}_1}(a(e+f))|_D$ is isomorphic to $\OO_{\PP^1}(a)$.

3570: Thus we conclude that also ${\mathcal F}(a,0)|_D$ is generated by

3571: global sections.  By the induction assumption, $H^1\lt(({\mathbb

3572:   F}_1,{\mathcal F}(a-1,0) \rt)$ is zero, we conclude by the long

3573: exact sequence of cohomology associated to the short exact sequence

3574: above, that all the global sections of ${\mathcal F}(a,0)|_D$ lift to

3575: global sections of ${\mathcal F}(a,0)$.  Therefore ${\mathcal F}(a,0)$

3576: is generated by global sections.  Also, a coherent sheaf on $\PP^1$

3577: which is generated by global sections has no higher cohomology.

3578: Combining this with the induction assumption and using the long exact

3579: sequence in cohomology associated to the short exact sequence above,

3580: we conclude that $H^i\lt( {\mathbb F}_1, {\mathcal F}(a,0) \rt)$ is

3581: zero for $i>0$.  Therefore we conclude by induction that for all

3582: $a>0$, ${\mathcal F}(a,0)$ is generated by global sections and

3583: $H^i\lt( {\mathbb F}_1, {\mathcal F}(a,0) \rt)$ is zero for $i>0$.

3584: 

3585: \

3586: 

3587: Now we prove the result with $b$ arbitrary.  We proceed by induction

3588: on $b$.  If $b=0$, the result was proved in the last paragraph.  Thus

3589: suppose that $b>0$ and suppose the result is proved for $b-1$.  Let $L

3590: \subset {\mathbb F}_1$ be a general fiber of $\text{pr}_1$.  Then $L$

3591: is smooth and isomorphic to $\PP^1$.  Since $L$ is general, we have a

3592: short exact sequence:

3593: \begin{equation}

3594: \begin{CD}

3595: 0 @>>> {\mathcal F}(a,b-1) @>>> {\mathcal F}(a,b) @>>> {\mathcal

3596:   F}(a,b)|_L @>>> 0

3597: \end{CD}

3598: \end{equation}

3599: Now $\OO_{{\mathbb F}_1}(a(e+f)+bf)|_L$ is isomorphic to

3600: $\OO_{\PP^1}(a)$.  By a similar analysis to that in the last

3601: paragraph, we conclude that ${\mathcal F}(a,b)$ is generated by global

3602: sections and that $H^i\lt( {\mathbb F}_1, {\mathcal F}(a,b) \rt)$ is

3603: zero for $i>0$.  So item $(3)$ is proved by induction on $b$.

3604: \end{proof}

3605: 

3606: \begin{lem} \label{lem-scroll2}

3607:   Let $([X],[\Sigma],[C_0]) \in {\mc W}^o$ be any triple, and let

3608:   $f:{\mathbb F}_1 \rightarrow \Sigma$ be some fixed isomorphism.  Let

3609:   $N'(a,b)$ and $N(a,b)$ be as in lemma~\ref{lem-scroll1}.  Also let

3610:   us denote

3611: \begin{equation}

3612: N_X(a,b) f^*\lt( N_{\Sigma/X}(-1) \rt)\otimes \OO(a,b)

3613: \end{equation}  

3614: Then we have the following:

3615: \begin{enumerate}

3616: 

3617: \item $X$ is smooth along $\Sigma$

3618:   

3619: \item for each pair of nonnegative integers $(a,b)$, we have $H^i\lt(

3620:   {\mathbb F}_1, N_X(a,b+k-3) \rt)$ is zero for $i>0$,

3621:   

3622: \item for any line of ruling $L\subset \Sigma$, $H^1\lt( L,

3623:   N_{L/X}(a-1) \rt)$ is zero for $a\geq 0$ any integer,

3624:   

3625: \item $H^1\lt( C_0, N_{C_0/X}(a-2) \rt)$ is zero for $a\geq 0$ any

3626:   integer,

3627:   

3628: \item the projection morphism ${\mc W} \rightarrow \PP^{N_d}$ given by

3629:   $([X],[\Sigma],[C_0]) \mapsto [X]$ is smooth at

3630:   $([X],[\Sigma],[C_0])$,

3631:   

3632: \item for any line $L\subset \Sigma$, the projection morphism $F({\mc

3633:     X}) \rightarrow \PP^{N_d}$ given by $([X],[L]) \mapsto [X]$ is

3634:   smooth at $([X],[L])$,

3635:   

3636: \item the projection morphism ${\mc R}^k({\mc X}) \rightarrow

3637:   \PP^{N_d}$ given by $([X],[C_0]) \mapsto [X]$ is smooth at

3638:   $([X],[C_0])$, and

3639:   

3640: \item the projection morphism $\pi: {\mc W} \rightarrow {\mc R}^k({\mc

3641:     X})$ given by $([X],[\Sigma],[C_0]) \mapsto ([X],[C_0])$ is smooth

3642:   at $([X],[\Sigma],[C_0])$.

3643: \end{enumerate}

3644: 

3645: \end{lem} 

3646: 

3647: \begin{proof}

3648:   Since the partial derivatives of a defining equation of $X$ give a

3649:   $c$-generating linear series, in particular they generate the sheaf

3650:   $\OO_{\Sigma}(d-1)$.  Thus, there is no point of $\Sigma$ at which

3651:   all the partial derivatives vanish.  By the Jacobian criterion, we

3652:   conclude that $X$ is smooth along $\Sigma$.  This proves item $(1)$.

3653: 

3654: \

3655: 

3656: For item $2$, we observe that we have a short exact sequence:

3657: \begin{equation}

3658: \begin{CD}

3659: 0 @>>> N_{\Sigma/X} @>>> N_{\Sigma/\PP^n} @>>> N_{X/\PP^n}|_\Sigma

3660: @>>> 0

3661: \end{CD}

3662: \end{equation}

3663: For ease of notation, define $\alpha = a + (d-1)$ and $\beta =

3664: b+(d-1)(k-1)$.  We have a short exact sequence:

3665: \begin{equation}

3666: \begin{CD}

3667: 0 @>>> N_X(a,b) @>>> N(a,b) @>>> \OO(\alpha,\beta) @>>> 0 

3668: \end{CD}

3669: \end{equation}

3670: When $a,b \geq 0$, it follows by lemma~\ref{lem-scroll1} that $H^i\lt(

3671: {\mathbb F}_1, N(a,b) \rt)$ is zero for $i\geq 0$.  By a simple

3672: calculation, we also see that

3673: \begin{equation}

3674: H^i\lt( {\mathbb F}_1,\OO(\alpha,\beta) \rt) = \{0\}, i\geq 0.

3675: \end{equation}

3676: So we conclude that $H^2\lt( {\mathbb F}_1,N_X(a,b) \rt)$ is zero, and

3677: $H^1\lt( {\mathbb F}_1, N_X(a,b) \rt)$ is zero iff the following map

3678: is surjective:

3679: \begin{equation}

3680: H^0\lt( {\mathbb F}_1, N(a,b) \rt) \rightarrow H^0\lt( {\mathbb F}_1,

3681: \OO(\alpha,\beta) \rt).

3682: \end{equation}

3683: We have a commutative diagram:

3684: \begin{equation}

3685: \begin{CD}

3686: H^0\lt( {\mathbb F}_1, N(a,b) \rt) \otimes H^0\lt( {\mathbb F}_1,

3687: \OO(a',b') \rt) @>>> H^0\lt( {\mathbb F}_1, N(a+a',b+b') \rt) \\

3688: @VVV @VVV \\

3689: H^0\lt( {\mathbb F}_1, \OO\lt(\alpha,\beta

3690: \rt) \rt) \otimes H^0\lt( {\mathbb F}_1, \OO(a',b') \rt) @>>>

3691: H^0\lt( {\mathbb F}_1,

3692: \OO\lt( \alpha+a', \beta+b' \rt) \rt)

3693: \end{CD}

3694: \end{equation}

3695: We conclude that if $H^1\lt( {\mathbb F}_1, N_X(a,b) \rt)$ is zero and

3696: if $a',b' \geq 0$, then we also have that $H^1\lt( {\mathbb F}_1,

3697: N_X(a+a',b+b') \rt)$.  So to prove item $(2)$, we are reduced to the

3698: case $a=0, b=k-3$.  But then the commutative diagram above factors as

3699: \begin{equation}

3700: \begin{CD}

3701: H^0\lt( {\mathbb F}_1, T_{\PP^n} \rt) \otimes H^0\lt( {\mathbb F}_1,

3702: \OO(0,k-3) \rt) @>>> H^0\lt( {\mathbb F}_1, N(0,k-3) \rt) \\

3703: @VVV @VVV \\

3704: H^0\lt( {\mathbb F}_1, \OO\lt(d-1,(d-1)(k-1)

3705: \rt) \rt) \otimes H^0\lt( {\mathbb F}_1, \OO(0,k-3) \rt) @>>>  

3706: H^0\lt( {\mathbb F}_1, \OO\lt(\alpha,\beta \rt) \rt)

3707: \end{CD}

3708: \end{equation}

3709: By definition, the composition

3710: \begin{equation}

3711: H^0\lt( {\mathbb F}_1 T_{\PP^n} \rt) \otimes

3712: H^0\lt( {\mathbb F}_1, \OO(0,k-3) \rt) \rightarrow H^0\lt( {\mathbb

3713:   F}_1, \OO\lt(\alpha, \beta \rt) \rt)

3714: \end{equation}

3715: is surjective iff the triple $([X],[\Sigma],[C_0])$ is in ${\mathcal

3716:   W}^o$.

3717: So if $([X],[\Sigma],[C_0])$ is in ${\mathcal W}^o$, then $H^1\lt(

3718:   {\mathbb F}_1, N_X(0,k-3) \rt)$ is zero.  Thus item $(2)$ holds.

3719: 

3720: \

3721: 

3722: For item $(3)$, observe that we have a short exact sequence:

3723: \begin{equation}

3724: \begin{CD}

3725: 0 @>>> N_{L/\Sigma}(a-1) @>>> N_{L/X}(a-1) @>>> N_{\Sigma/X}|_L(a-1) @>>> 0

3726: \end{CD}

3727: \end{equation}

3728: Of course $N_{L/\Sigma} \cong \OO_L$, thus $H^1\lt( L,

3729: N_{L/\Sigma}(a-1) \rt)$ is zero for $a\geq 0$.  So to prove item

3730: $(3)$, it suffices to prove that $H^1\lt(L, N_{\Sigma/X}|_L(a-1) \rt)$

3731: is zero.  Since $\OO(a-1,b)|_L \cong \OO_L(a-1)$, we have a short

3732: exact sequence:

3733: \begin{equation}

3734: \begin{CD}

3735: 0 @>>> N_X(a,k-3) @>>> N_X(a,k-2) @>>> N_{\Sigma/X}|_L(a-1) @>>> 0

3736: \end{CD}

3737: \end{equation}

3738: By item $(2)$, for $a\geq 0$ the higher cohomology of the first two

3739: terms vanishes.  Thus by the long exact sequence in cohomology

3740: associated to this short exact sequence, we have that $H^1\lt( L,

3741: N_{\Sigma/X}|_L(a-1) \rt)$ is zero for $a\geq 0$.  This proves item

3742: $(3)$.

3743: 

3744: \

3745: 

3746: Item $(4)$ is almost identical to item $(3)$ and is left as an

3747: exercise to the reader.

3748: 

3749: \

3750: 

3751: By \cite[proposition 2.14.2]{K}, the obstruction space for the

3752: relative Hilbert scheme $\textit{Hilb}^{Q(t)}({\mc X}/\PP^{N_d})$ 

3753: at a point $([X],[\Sigma])$ is contained in $H^1\lt(\Sigma,

3754: N_{\Sigma/X} \rt)$.  For a triple $([X],[\Sigma],[C_0])$ in ${\mc

3755: W}^o$, it follows from item $(2)$ that 

3756: \begin{equation}

3757: H^1\lt( \Sigma, N_{\Sigma/X} \rt) = H^1\lt( {\mathbb F}_1, N_X(1,k-1)

3758: \rt)

3759: \end{equation} 

3760: vanishes.  It follows by \cite[theorem 2.10]{K} that

3761: $\textit{Hilb}^{Q(t)}({\mc X}/\PP^{N_d}) \rightarrow \PP^{N_d}$ is

3762: smooth at $([X],[\Sigma])$.  Of course the projection ${\mc W}^o

3763: \rightarrow \textit{Hilb}^{Q(t)}({\mc X}/\PP^{N_d})$ is an open subset

3764: of a projective bundle, and so is smooth.  Thus we conclude that the

3765: composite morphism ${\mc W}^o \rightarrow \PP^{N_d}$ is smooth.  This

3766: proves item $(5)$.

3767: 

3768: \

3769: 

3770: Item $(6)$ is similar to item $(5)$ and follows from item $(3)$ which

3771: shows that $H^1\lt( L, N_{L/X} \rt)$ is zero.

3772: 

3773: \

3774: 

3775: Item $(7)$ is similar to item $(5)$ and follows from item $(4)$ which

3776: shows that $H^1\lt( C_0, N_{C_0/X} \rt)$ is zero.

3777: 

3778: \

3779: 

3780: Since ${\mc W}^o \rightarrow \PP^{N_d}$ is smooth at

3781: $([X],[\Sigma],[C_0])$ and since ${\mc R}^k \rightarrow \PP^{N_d}$ is

3782: smooth at $([X],[C_0])$, to prove that $\pi:{\mc W}^0 \rightarrow {\mc

3783:   R}^k({\mc X})$ is smooth at $([X],[\Sigma],[C_0])$, it suffices to

3784: check that the derivative map $d\pi: T_{{\mc W}^0/\PP^{N_d}}

3785: \rightarrow \pi^* T_{{\mc R}^k({\mc X})/\PP^{N_d}}$ is surjective at

3786: $([X],[\Sigma],[C_0])$.  This exactly reduces to the statement that

3787: $H^0 \lt( \Sigma, N_{\Sigma/X} \rt) \rightarrow H^0 \lt( C_0,

3788: N_{\Sigma/X}|_{C_0} \rt)$ is surjective.  Since the cokernel is

3789: contained in $H^1\lt( {\mathbb F}_1, N_X(0,k-1) \rt)$, which is zero

3790: by item $(3)$, we conclude the derivative $d\pi$ is surjective.

3791: Therefore $\pi:{\mc W}^o \rightarrow {\mc R}^k({\mc X})$ is smooth.

3792: This proves item $(8)$.

3793: \end{proof}

3794: 

3795: Now suppose that $([X],[\Sigma],[C_0])$ is a point in ${\mc W}^o$.

3796: Let $\sigma:C_0 \rightarrow \Sigma$ be the inclusion and let

3797: $\text{pr}_{C_0}: \Sigma \rightarrow C_0$ be the unique projection

3798: such that $\sigma$ is a section of $\text{pr}_L$ (in the model of

3799: $\Sigma$ as ${\mathbb F}_1$, $\text{pr}_{C_0}$ is simply

3800: $\text{pr}_1$).  Let $g:\Sigma \rightarrow X$ be the inclusion.  Then

3801: we have an induced morphism $\zeta:C_0 \rightarrow \Kgnb{0,1}{X,1}$

3802: given by the diagram:

3803: \begin{equation}

3804: \begin{CD}

3805: \Sigma @> g >> X \\

3806: @VV \text{pr}_{C_0} V \\

3807: C_0

3808: \end{CD}

3809: \end{equation}

3810: 

3811: \begin{lem} \label{lem-scroll3}

3812:   If $([X],[\Sigma],[C_0])$ is a triple in ${\mc W}^o$, then the

3813:   corresponding morphism $\zeta:C_0 \rightarrow \Kgnb{0,1}{X,1}$ is

3814:   very twisting and very positive.

3815: \end{lem}

3816: 

3817: \begin{proof}

3818:   First we check the axioms of definition~\ref{defn-verytwisting}.

3819:   Since $g\circ \sigma:C_0 \rightarrow X$ is an embedding, in

3820:   particular this map is stable, i.e. axiom $(1)$ is satisfied.  By

3821:   item $(7)$ of lemma~\ref{lem-scroll2}, we conclude that

3822:   $\Kgnb{0,0}{X,k}$ is unobstructed at $[g\circ \sigma: C_0

3823:   \rightarrow X]$, i.e. axiom $(2)$ is satisfied.

3824: 

3825: \

3826: 

3827: The argument that axiom $(3)$ holds is exactly the same as the

3828: argument that axiom $(3)$ holds in the proof of

3829: lemma~\ref{lem-quadric2}, where item $(6)$ of lemma~\ref{lem-quadric1}

3830: is replaced by item $(3)$ of lemma~\ref{lem-scroll2}.

3831: 

3832: \

3833: 

3834: As in the proof of lemma~\ref{lem-quadric2}, we have that

3835: $\zeta^*T_{\text{ev}}$ is isomorphic to $\lt( \text{pr}_{C_0} \rt)_*

3836: N_X(0,k-1)$.  Of course $\zeta^*T_{\text{ev}}$ is ample iff $H^1\lt(

3837: C_0, \zeta^*T_{\text{ev}}(-2) \rt)$ is zero.  By the isomorphism above

3838: and a Leray spectral sequence argument analogous to the one in the

3839: proof of lemma~\ref{lem-quadric2}, this cohomology group equals

3840: $H^1\lt( {\mathbb F}_1, N_X(0,k-3) \rt)$.  By item $(2)$ of

3841: lemma~\ref{lem-scroll2}, this group is zero.  Therefore

3842: $\zeta^*T_{\text{ev}}$ is an ample bundle.  So axiom $(4)$ is

3843: satisfied.

3844: 

3845: \

3846: 

3847: Finally, observe that $\sigma^* \OO_{\Sigma}(\sigma)$ is

3848: $\OO_{C_0}(1)$ and so is ample.  So axiom $(5)$ is satisfied.  Thus

3849: $\zeta$ is a very twisting family.

3850: 

3851: \

3852: 

3853: Next we consider the axioms in definition~\ref{defn-pos}.  Axioms

3854: $(1)$, $(2)$ and $(3)$ follow immediately from axioms $(1)$, $(2)$ and

3855: $(3)$ of definition~\ref{defn-twisting} proved above.  To see that

3856: axiom $(4)$ holds, observe that we have a short exact sequence:

3857: \begin{equation}

3858: \begin{CD}

3859: 0 @>>> \zeta^* T_{\text{ev}} @>>> \zeta^* T_{\Kgnb{0,1}{X,1}} @>>>

3860: \lt( g\circ \sigma\rt)^* T_X @>>> 0

3861: \end{CD}

3862: \end{equation}

3863: We have seen above that $\zeta^* T_{\text{ev}}$ is ample.  Moreover,

3864: it follows by item $(4)$ of lemma~\ref{lem-scroll2} that $N_{C_0/X}$

3865: is an ample vector bundle.  Since also $T_{C_0}$ is an ample line

3866: bundle, we conclude that $T_X|_{C_0}$ is an ample vector bundle.

3867: Since the first and third terms in the short exact sequence above are

3868: ample, we conclude that $\zeta^* T_{\Kgnb{0,1}{X,1}}$ is an ample

3869: vector bundle.  Since $\zeta^* \text{pr}^* T_{\Kgnb{0,0}{X,1}}$ is a

3870: quotient bundle of $\zeta^* T_{\Kgnb{0,1}{X,1}}$, we also have that

3871: $\zeta^* \text{pr}^* T_{\Kgnb{0,0}{X,1}}$ is an ample vector bundle.

3872: So axiom $(4)$ holds.

3873: 

3874: \

3875: 

3876: Finally, we have seen above that $\sigma^*\OO_{\Sigma}(\sigma)$ equals

3877: $\OO_{C_0}(1)$, which is ample.  Thus axiom $(5)$ holds.  So $\zeta$

3878: is a very positive family.

3879: \end{proof} 

3880: 

3881: \begin{prop} \label{prop-scroll}

3882: If $n \geq d^2 + d + 2$ and if $d\geq 3$, for $k=2d-3$, we have that

3883: ${\mc W}^o \rightarrow {\mc R}^k({\mc X})$ is dominant, and ${\mc

3884: R}^k({\mc X}) \rightarrow \PP^{N_d}$ is dominant.  So for $[X]\in

3885: \PP^{N_d}$ general, there exists a very twisting, very positive family

3886: $\zeta: C_0 \rightarrow \Kgnb{0,1}{X,1}$.

3887: \end{prop}

3888: 

3889: \begin{proof}

3890:   By item $(8)$ of lemma~\ref{lem-scroll2}, it suffices to prove that

3891:   ${\mc W}^o$ is nonempty.  We have to find a pair $([X],[\Sigma])$

3892:   such that for $a=d-1$, $b=(d-1)(k-1)$ and for $c=k-3$, we have that

3893:   the image of the derivative map

3894: \begin{equation}

3895: d_{X,\Sigma}:H^0\lt( (\PP^n,T_{\PP^n}(-1) \rt) \rightarrow H^0\lt(

3896: {\mathbb F}_1, \OO(a,b) \rt)

3897: \end{equation}

3898: is a $c$-generating linear system.

3899: 

3900: \

3901: 

3902: Recall that $S = \CC[T_0,T_1,U,V]$ is the $\ZZ^2$-graded Cox

3903: homogeneous coordinate ring of ${\mathbb F}_1$.  Let $A_d$ denote the

3904: set of monomials in the vector subspace $S_{(d-1,(d-1)(k-1))}$ which

3905: occur in the linear system $W_0(a,b,c)$, i.e.

3906: \begin{equation}

3907: \begin{array}{c}

3908: A_d = \\

3909: \lt\{ U^i V^{d-1-i}

3910: T_0^{((d-1)(k-1)+i)-j(k-2)}T_1^{j(k-2)}|i=0,\dots, d-1,

3911: j=1,\dots,r(i) \rt\}  \\

3912: \cup \lt\{U^i V^{d-1-i}

3913: T_1^{(d-1)(k-1)+i}\rt|i=0,\dots,d-1\}

3914: \end{array}

3915: \end{equation}

3916: where $r(i) = d-1+\lt\lfloor\frac{d-2+i}{k-2} \rt\rfloor$.

3917: Let $B_d$ denote the set of

3918: monomials of the form 

3919: \begin{equation}

3920: \begin{array}{c}

3921: B_d = \\

3922: \lt\{U^{d-1} T_0^{(d-1)k - i(k-2)} T_1^{i(k-2)} | i=0,\dots, d-2

3923: \rt\} \cup \\

3924: \lt\{ U^{d-2} V T_0^{(d-1)k-1-i(k-2)} T_1^{i(k-2)} | i=0,\dots, d-2 \rt\}

3925: \cup \\

3926: \lt\{ U^{d-3} V^2 T_0^{(d-1)k-2-(j+2)(k-2)} T_1^{(j+2)(k-2)} |

3927: j=0,\dots, d-4 \rt\} \cup \\

3928: \lt\{ U^{d-4} V^3 T_0^{(d-1)k-3-(j+2)(k-2)} T_1^{(j+2)(k-2)} | j=0,\dots,

3929: d-4 \rt\}.

3930: \end{array}

3931: \end{equation}

3932: Let $C_d$ denote the set of monomials $C_d = A_d - B_d$.  Now $A_d$

3933: contains $d^2 + d$ monomials, and $B_d$ contains $4d-8$ monomials.

3934: Choose homogeneous coordinates on $\PP^n$ of the form

3935: \begin{equation}

3936: \lt\{Y_0,\dots,Y_{2k} \rt\} \cup \lt\{ X_M | M \in C_d \rt\} \cup

3937: \lt\{Z_l | l=1,\dots, n-(d^2+d+2) \rt\}.

3938: \end{equation}

3939: 

3940: \

3941: 

3942: Let $f:{\mathbb F}_1 \rightarrow \PP^{2k} \subset \PP^n$ be the map

3943: defined by sending a point $([T_0:T_1],[T_0U:T_1U:V])$ to the point in

3944: $\PP^n$ with coordinates $X_m = 0, m\in C_d$, with $Z_l=0, l=1,\dots,

3945: n-(d^2+d)$, and with

3946: \begin{eqnarray}

3947: Y_0 = T_0^kU,\dots, Y_i = T_0^{k-i}T_1^iU, \dots, Y_k = T_1^kU, \\

3948: Y_{k+1} = T_0^{k-1}V,\dots, Y_{k+1+j} = T_0^{k-1-j}T_1^j V,\dots,

3949: Y_{2k} = T_1^{k-1}V.

3950: \end{eqnarray}

3951: This is an embedding whose image $\Sigma = f({\mathbb F}_1)$ is a

3952: rational normal scroll of degree $2k-1$.  

3953: 

3954: \

3955: 

3956: Now the pullback map $H^0\lt( (\PP^{2k},\OO_{\PP^{2k}}(1) \rt)

3957: \rightarrow H^0\lt( {\mathbb F}_1, \OO(1,k-1) \rt)$ is surjective by

3958: construction.  And the natural map

3959: \begin{equation}

3960: \text{Sym}^{d-1}H^0 \lt( {\mathbb F}_1, \OO(1,k-1) \rt) \rightarrow

3961: H^0 \lt( {\mathbb F}_1, \OO\lt(d-1,(d-1)(k-1)\rt) \rt)

3962: \end{equation} 

3963: is surjective.  Therefore the pullback map 

3964: \begin{equation}

3965: H^0\lt( \PP^{2k},\OO_{\PP^{2k}}(d-1) \rt) \rightarrow H^0\lt( {\mathbb

3966: F}_1, \OO\lt(d-1,(d-1)(k-1) \rt) \rt)

3967: \end{equation}

3968: is surjective.  For each monomial $M\in C_d$, choose a polynomial

3969: $G_M(Y_0,\dots,Y_{2k})$ such that $f^*G_M = M$.  

3970: 

3971: \

3972: 

3973: Consider the hypersurface $X\subset \PP^n$ with defining equation

3974: \begin{eqnarray}

3975: F = \sum_{i=0}^{d-2} \lt(Y_iY_{k+2+i} - Y_{i+1}Y_{k+1+i}\rt)

3976: Y_0^{d-2-i}Y_{k-3}^i + \\

3977: \sum_{j=0}^{d-4}\lt(Y_{d-1+j}Y_{k+d+1+j} - Y_{d+j}Y_{k+d+j}

3978: \rt)Y_0^{d-4-j}Y_{k-3}^jY_{2k-3}Y_{2k+1-d} + \\ 

3979: \sum_{M\in C_d}G_M(Y_0,\dots,Y_{2k})X_M.

3980: \end{eqnarray}

3981: The corresponding derivative map $d_{X,\Sigma}$ acts on the partial

3982: derivatives $\frac{\partial F}{\partial Y_i}$ by

3983: \begin{equation}

3984: \begin{array}{lcl}

3985: \frac{\partial F}{\partial Y_0} & \mapsto & U^{d-2}VT_0^{(d-1)k-2}; \\

3986: \frac{\partial F}{\partial Y_{i+1}} & \mapsto &

3987: -U^{d-2} V T_0^{(d-1)k-1-i(k-2)} T_1^{i(k-2)} + \\

3988: & & U^{d-2} V T_0^{(d-1)k-2-(i+1)(k-2)} T_1^{(i+1)(k-2)}, \\

3989: & & i=0,\dots,d-3,

3990: \\

3991: \frac{\partial F}{\partial Y_{d-1}} & \mapsto &

3992: -U^{d-2} V T_0^{(d-1)k-1-(d-2)(k-2)} T_1^{(d-2)(k-2)} + \\

3993: & & U^{d-4} V^3 T_0^{(d-1)k-4-2(k-2)} T_1^{2(k-2)+1}, \\

3994: \frac{\partial F}{\partial Y_{d+j}} & \mapsto &

3995: -U^{d-4} V^3 T_0^{(d-1)k-3-(j+2)(k-2)} T_1^{(j+2)(k-2)} + \\

3996: & & U^{d-4} V^3 T_0^{(d-1)k-4-(j+3)(k-2)} T_1^{(j+3)(k-2)},\\ 

3997: & & j=0,\dots,

3998: d-5, \\

3999: \frac{\partial F}{\partial Y_{k-1}} & \mapsto & 

4000: -U^{d-4} V^3

4001: T_0^{(d-1)k-3-(d-2)(k-2)} T_1^{(d-2)(k-2)}, \\

4002: \frac{\partial F}{\partial Y_k} & \mapsto & 0, \\

4003: \frac{\partial F}{\partial Y_{k+1}} & \mapsto & -U^{d-1} T_0^{(d-1)k-1}T_1, \\

4004: \frac{\partial F}{\partial Y_{k+2+i}} & \mapsto & U^{d-1} T_0^{(d-1)k -

4005:   i(k-2)} T_1^{i(k-2)} - \\

4006: & & U^{d-1} T_0^{(d-1)k-1-(i+1)(k-2)} T_1^{(i+1)(k-2)+1},\\

4007: & & i=0,\dots, d-3,

4008: \\

4009: \frac{\partial F}{\partial Y_{k+d}} & \mapsto & U^{d-1}

4010: T_0^{(d-1)k-(d-2)(k-2)} T_1^{(d-2)(k-2)} - \\

4011: & & U^{d-3} V^2 T_0^{(d-1)k - 3 -2(k-2)} T_1^{2(k+2) + 1}, \\

4012: \frac{\partial F}{\partial Y_{k+d+1+j}} & \mapsto & U^{d-3} V^2

4013: T_0^{(d-1)k-2-(j+2)(k-2)} T_1^{(j+2)(k-2)} - \\

4014: & & U^{d-3} V^2 T_0^{(d-1)k-3-(j+3)(k-2)} T_1^{(j+3)(k-2)+1},\\

4015: & & j=0,\dots,

4016: d-5, \\

4017: \frac{\partial F}{\partial Y_{2k}} & \mapsto & U^{d-3} V^2

4018: T_0^{(d-1)k-2-(d-2)(k-2)} T_1^{(d-2)(k-2)}.

4019: \end{array}

4020: \end{equation}

4021: 

4022: \

4023: 

4024: In the list above, every partial derivative is a sum of at most two

4025: monomials, i.e. each partial derivative is a binomial, and the first

4026: term listed is the initial term with respect to our monomial order.

4027: From this list, it is clear that every monomial in $B_d$ occurs as the

4028: initial term of some partial derivative.  Also, for each $M\in C_d$,

4029: the partial derivative $\frac{\partial F}{\partial X_M}$ simply maps

4030: to $M$.  Thus every monomial in $C_d$ occurs as the initial term of

4031: some partial derivative.  Thus every monomial in $A_d$ occurs as the

4032: initial term of some partial derivative.  So the vector space of

4033: initial terms of $\text{image}(d_{X,\Sigma})$ contains $W_0(a,b,c)$.

4034: Therefore, by lemma~\ref{lem-comput2}, we conclude that

4035: $\text{image}(d_{X,\Sigma})$ is a $c$-generating linear system.  So,

4036: for any irreducible curve $C_0$ in the linear system of $|\OO(1,0)|$,

4037: we have that $([X],[\Sigma],[C_0])$ is in ${\mc W}^o$.

4038: \end{proof}

4039: 

4040: \section{Proof of the main theorem}

4041: 

4042: In this section we prove theorem~\ref{thm-thm1}.  As explained at the

4043: end of section~\ref{sec-results}, if $d< \frac{n+1}{2}$, then for a

4044: general hypersurface $X_d\subset \PP^n$, hypothesis~\ref{hyp-1},

4045: hypothesis~\ref{hyp-1.5}, and hypothesis~\ref{hyp-1.75} are satisfied.

4046: By proposition~\ref{prop-quadric}, if $d\geq 2$ and $d^2 \leq n+1$,

4047: then for $X_d\subset \PP^n$ general we have that

4048: hypothesis~\ref{hyp-2} is satisfied.  Finally, if $d\geq 3$ and if

4049: $d^2+d+2 \leq n$, then for $X_d\subset \PP^n$ general there exists a

4050: very twisting, very positive family $\zeta:C_0 \rightarrow

4051: \Kgnb{0,1}{X,1}$.  Thus $(\zeta,\zeta)$ is an inducting pair.  Now by

4052: theorem~\ref{thm-induction}, we conclude that for every $e \geq 1$

4053: there exists an inducting pair $(\zeta_1,\overline{\zeta}_e)$.  In

4054: particular, there exists a very positive $1$-morphism

4055: $\overline{\zeta}_e:C \rightarrow \Kgnb{0,1}{X,e}$.  As seen in the

4056: proof of theorem~\ref{thm-induction}, we may assume that $C$ is

4057: smooth, i.e. $C$ is equal to $\PP^1$, and that the image of

4058: $\text{pr}\circ \overline{\zeta}_e:\PP^1 \rightarrow \Kgnb{0,0}{X,e}$

4059: is contained in the smooth part of the fine moduli locus.  So, by item

4060: $(2)$ of remark~\ref{rmk-pos}, we conclude that $\overline{\zeta}_e$

4061: is a \emph{very free} morphism.  And by \cite[proposition 7.4]{HRS2},

4062: we have that $\Kgnb{0,0}{X,e}$ is an irreducible variety.  Therefore

4063: by \cite[theorem IV.3.7]{K}, we conclude that $\Kgnb{0,0}{X,e}$ is

4064: rationally connected.

4065: 

4066: \bibliography{my}

4067: \bibliographystyle{abbrv}

4068: 

4069: \end{document}  

4070: 

4071: 

4072: 

4073: 

4074: 

4075: 

4076: 

4077: 

4078: 

4079: 

4080: 

4081: