1: 

2: 

3: \documentclass[amssymb,twoside,12 pt,amscd,reqno]{amsart}

4: \usepackage{amsmath}

5: \usepackage{amscd}

6: \usepackage{latexsym}

7: \addtolength{\textwidth}{+4cm}

8: \addtolength{\textheight}{+2cm}

9: \addtolength{\baselineskip}{+0.7cm}

10: \hoffset-2cm

11: \voffset-1cm

12: 

13: \theoremstyle{plain}

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

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

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

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

18: \newtheorem{Claim}[thm]{Claim}

19: 

20: \theoremstyle{definition}

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

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

23: \newtheorem{Remark}[thm]{Remark}

24: 

25: 

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

27: \newcommand{\pr}{\text{pr}}

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

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

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

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

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

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

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

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: 

43: \def\ul{\underline}

44: \def\la{\longrightarrow}

45: \def\ni{\noindent}

46: \def\cl{\centerline}

47: 

48: \def\Sym{{\rm Sym}}

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

50: 

51: \def\A{{\mathbb A}}

52: \def\C{{\mathbb C}}

53: \def\F{{\mathbb F}}

54: \def\G{{\mathbb G}}

55: \def\P{{\mathbb P}}

56: \def\Q{{\mathbb Q}}

57: \def\R{{\mathbb R}}

58: \def\Z{{\mathbb Z}}

59: 

60: \def\cC{{\cal C}}

61: \def\cX{{\cal X}}

62: \def\cH{{\cal H}}

63: 

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

65: 

66: \newcommand{\lt}{\left}

67: \newcommand{\rt}{\right}

68: 

69: \newcommand{\g}{{\frak g}}

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

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

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

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

74: \newcommand{\B}{{\cal B}}

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

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

77: 

78: \setcounter{tocdepth}{1}

79: 

80: 

81: \begin{document}

82: 

83: \title{Abel-Jacobi Maps Associated to Smooth Cubic Threefolds}

84: \author[Harris]{Joe Harris} 

85: \address{Department of Mathematics \\ Harvard University \\ Cambridge MA 02138}

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

87: \author[Roth]{Mike Roth} 

88: \address{Department of Mathematics \\ University of Michigan \\ Ann

89: Arbor, MI 48109}  

90: \email{mikeroth@math.lsa.umich.edu}

91: \author[Starr]{Jason Starr}\thanks{The third author was partially

92: supported by an NSF Graduate Research Fellowship and a Sloan

93: Dissertation Fellowship} 

94: \address{Department of Mathematics \\ Massachusetts Institute of Technology \\ Cambridge MA 02139}

95: \email{jstarr@math.mit.edu}

96: \date{\today}

97: 

98: \maketitle

99: 

100: \tableofcontents

101: 

102: \begin{abstract}

103: In this article we consider the spaces $\Hdg{d}{g}(X)$ parametrizing

104: curves of degree $d$ and genus $g$ on a smooth cubic threefold $X

105: \subset \P^4$, with regard in particular to the Abel-Jacobi map $u_d :

106: \Hdg{d}{g}(X) \to J^3(X)$ to the intermediate Jacobian $J^3(X)$ of

107: $X$.  Our principle result is that for all $d \leq 5$ the map $u_d$

108: coincides with the maximal rationally connected fibration of

109: $\Hdg{d}{g}(X)$.

110: \end{abstract}

111: 

112: \section{Introduction} \label{sec-intro}

113: 

114: In this paper we will study spaces parametrizing curves on a smooth

115: cubic threefold $X \subset \P^4$. There has been a great deal of work

116: in recent years on the geometry of spaces parametrizing rational

117: curves on a variety $X$.  Most of it has focussed on the enumerative

118: geometry of these spaces: the description of their Chow rings and the

119: evaluation of certain products in their Chow rings. Here we will be

120: concerned with a very different sort of question: we will be concerned

121: with the birational geometry of the spaces $M$.

122: 

123: \ps

124: 

125: We should start by explaining, at least in part, our motivation. The

126: central object in curve theory---the one that links together every

127: aspect of the theory, and whose study yields the majority of theorems

128: in the subject---is the Abel-Jacobi map. This is the map from the

129: symmetric product $C_d$ of a curve $C$, parametrizing $0$-cycles of a

130: given degree $d$ on $C$, to the Jacobian $\pic^d(C) \cong J(C)$,

131: defined variously as the space of cycles of degree $d$ mod linear

132: equivalence (that is, the space of line bundles of degree $d$ on $C$)

133: or, over the complex numbers, as the complex torus

134: $H^1(C,K)^*/H_1(C,\Z)$.

135: 

136: \ps

137: 

138: Under the circumstances, it's natural to ask what sort of analogue of

139: this we might be able to define and study in higher dimensions. There

140: have been numerous constructions proposed to this end; they have been

141: too varied to categorize easily, but the majority adopt one of two

142: approaches, corresponding to the definitions of $\pic^d(C) \cong J(C)$

143: for a curve. In the first, we look again at the space of cycles on a

144: variety $X$ mod some equivalence relation (typically rational

145: equivalence); in the second, we try to form a geometric object out of

146: the Hodge structure of $X$.

147: 

148: \ps

149: 

150: Both approaches suffer from some seemingly unavoidable

151: difficulties. In the first, the quotients of the spaces on cycles on a

152: variety $X$ by rational equivalence tends to be either too big or too

153: small: if we mod out by algebraic equivalence we tend to lose too much

154: information, while if we mod out by rational equivalence the quotient

155: is too large (and in particular too hard to calculate even in simple

156: concrete cases).  As for the second, we simply don't know how to make

157: an algebrao-geometric object out of a Hodge structure except in very

158: special cases---specifically, the case of Hodge structures of odd

159: weight with all but two of the Hodge groups vanishing, where we can

160: form the intermediate Jacobian and it will be an Abelian variety. In

161: these cases the theory has had spectacular successes (such as the

162: proof of the irrationality of cubic threefolds), but it's undeniably

163: frustrating that it has no wider application.

164: 

165: \ps

166: 

167: In this paper we'd like to suggest another approach to the problem,

168: based on a construction of Koll\'ar, Miyaoka and Mori: the {\it

169: maximal rationally connected fibration} associated to a variety

170: $X$. Briefly, we say that a smooth projective variety $X$ is {\it

171: rationally connected} if two general points of $X$ can be connected by

172: a chain of rational curves. This condition is equivalent to

173: rationality and unirationality for curves and surfaces; in higher

174: dimensions it is weaker than either.  More generally, the maximal

175: rationally connected fibration associates to a variety $X$ a

176: (birational isomorphism class of) variety $Z$ and a rational map $\phi

177: : X \to Z$ with the properties that

178: \begin{itemize}

179: \item the fibers $X_z$ of $\phi$ are rationally connected; and conversely 

180: \item almost all the rational curves in $X$ lie in fibers of $\phi$:

181: for a very general point $z \in Z$ any rational curve in $X$ meeting

182: $X_z$ lies in $X_z$.

183: \end{itemize}

184: \ni The variety $Z$ and morphism $\phi$ are unique up to birational 

185: isomorphism, and are called the

186: \textit{mrc quotient} and \textit{mrc fibration} of $X$, respectively.

187: They measure the failure of $X$ to be rationally connected: if $X$ is

188: rationally connected, $Z$ is a point, while if $X$ is not uniruled we

189: have $Z=X$.

190: 

191: \ps

192: 

193: The point of this is, if we start with the symmetric product $C_d$ of

194: a curve $C$ and apply this construction, we find (for $d$ large, at

195: any rate) that the mrc fibration of $C_d$ is the Abel-Jacobi map $\phi

196: : C_d

197: \to J(C)$. In other words, we don't have to define the Jacobian 

198: $J(C)$ either in terms of cycles mod linear

199: equivalence or via Hodge theory; we can realize it simply the mrc

200: quotient of the space $C_d$ parametrizing $0$-cycles of reasonably

201: large degree $d$ on $C$. The obvious question is then: what happens

202: when we take the mrc quotient of the variety parametrizing

203: cycles\footnote{Note that we have a choice of parameter spaces for the

204: cycles on $X$: the Chow variety, the Hilbert scheme and, in the case

205: of 1-dimensional cycles, the Kontsevich space. But since we are

206: concerned with the birational geometry of these spaces it really

207: doesn't matter which we choose to work with, except for technical

208: convenience.} on a higher-dimensional variety?

209: 

210: \ps

211: 

212: To begin with, taking the mrc quotient of the varieties parametrizing

213: $0$-cycles on $X$ doesn't seem to yield much: if $X$ is rationally

214: connected, so will be its symmetric powers; and if $X$ possesses any

215: holomorphic forms one expects the dimensions of the mrc quotients of

216: $X_d$ tend  

217: to $\infty$ with $d$ (c.f. ~\cite{RE}).  We

218: turn our attention next to curves on $X$ and ask: what can we say

219: about the mrc quotients of the varieties $\Hdg{d}{g}(X)$ parametrizing

220: curves of degree $d$ and genus $g$ on $X$? For various reasons it

221: makes sense to look primarily at the spaces parametrizing rational

222: curves---for one thing, we don't want to get involved in the geometry

223: of the moduli space $M_g$ of curves of genus $g$ when we are only

224: interested in invariants of $X$---but in the course of studying

225: rational curves we will also discover facts about the geometry of

226: $\Hdg{d}{g}(X)$ for $g > 0$.

227: 

228: \ps

229: 

230: For $X = \P^n$ and for $X \subset \P^{n+1}$ a quadric hypersurface,

231: the answer is known, at least in the case of rational curves, and is

232: trivial: the variety $\Hdg{d}{0}(X)$ parametrizing rational curves of

233: degree $d$ on $X$ is itself rationally connected in both cases. The

234: first real test is thus cubic hypersurfaces, and here we will focus

235: specifically on the case of a smooth cubic threefold $X

236: \subset

237: \P^4$.

238: 

239: \ps

240: 

241: We start by asking the most naive possible question: is the variety

242: $\Hdg{d}{g}(X)$ rationally connected? The answer is that it can never

243: be: we have the Abel Jacobi map $$ u : \Hdg{d}{g}(X) \to J^3(X) $$

244: from the Kontsevich space to the intermediate Jacobian $$ J_3(X) =

245: H^{2,1}(X)^*/H_3(X,\Z).  $$ Since any map of a rational curve to a

246: complex torus/Abelian variety is constant, any rational curve on

247: $\Hdg{d}{g}(X)$ must lie in a fiber of $u$. The Abel-Jacobi map thus

248: factors through the mrc fibration $\phi : \Hdg{d}{g}(X) \to Z$ of

249: $\Hdg{d}{g}(X)$.

250: 

251: \ps

252: 

253: Given this, the second most naive question would be: is the

254: intermediate Jacobian $J^3(X)$ the mrc quotient of $\Hdg{d}{g}(X)$?

255: This may seem equally naive, but it turns out to hold for small values

256: of $d$. The main result of this paper is the

257: 

258: \ps

259: 

260: \begin{thm}\label{mainthm}  If $X \subset \P^4$ is any smooth 

261: cubic threefold, for any $d \leq 5$ the Abel-Jacobi

262: map $$ u : \Hdg{d}{g}(X) \to J^3(X) $$ is the maximal rationally

263: connected fibration of $\Hdg{d}{g}(X)$.

264: \end{thm}

265: 

266: 

267: As we indicated, we view this result as simply the first test of a

268: general hypothesis. It would be very interesting to see if the

269: analogous statement holds for other rationally connected threefolds.

270: Work is being done in this direction: Ana-Maria Castravet has proved

271: in ~\cite{C} that for $X \subset \P^5$ the intersection of two quadrics, the

272: analogous statement---that the mrc quotient of the space

273: $\Hdg{d}{0}(X)$ is the intermediate Jacobian---holds for all $d$. And

274: work of Harris and Starr~\cite{HS} shows that the same is true at least

275: for components of the variety $\Hdg{d}{0}(X)$ corresponding to an open

276: cone of curve classes $\beta \in H_2(Z,\Z)$ in a blow-up $X$ of

277: $\P^3$.

278: 

279: \ps

280: 

281: Ultimately, though, we want to extend this research to

282: higher-dimensional varieties $X$, where the Hodge theory doesn't seem

283: to provide us with a candidate for the mrc quotients of the spaces of

284: rational curves on $X$. To relate this to a specific issue: for smooth

285: cubic threefolds $X$, the fact that the Hodge structure on $H^3(X,\Z)$

286: is not isomorphic to a product of Hodge structures of curves implies

287: that $X$ is not rational. A close examination of the Hodge structure

288: of a very general cubic fourfold $X \subset \P^5$ suggests a similar

289: finding: the Hodge structure on $H^4(X,\Z)$ does not appear to be a

290: product of factors of Hodge structures of surfaces (see [Has]), and if

291: this is indeed the case it would imply that $X$ is irrational. But the

292: fact that the Hodge structure is not conveyed in the form of a

293: geometric object---there is no construction analogous to that of the

294: intermediate Jacobian of a cubic threefold---has frustrated our

295: attempts to make this into a proof. But now we can ask: if the

296: intermediate Jacobian of a cubic threefold $X$ may be realized as the

297: mrc quotient of the space of rational curves on $X$, what happens when

298: we take the mrc quotient of the space of rational curves on a cubic

299: fourfold?

300: 

301: \ps

302: 

303: In this paper, though, we will be concerned exclusively with the

304: geometry of cubic threefolds. We start some preliminary sections on

305: the Abel-Jacobi map for curves on threefolds in general and on the

306: geometry of curves of low degree on cubic threefolds in particular. We

307: then launch into the analysis of the Abel-Jacobi map for curves of

308: degree 3, 4 and 5 on a smooth cubic threefold $X$. The basic technique

309: here is {\it residuation}: we associate to a curve $C \subset X$ on

310: $X$ and a surface $S \subset \P^4$ containing it, a residual curve

311: $C'$ with (roughly) $C +C' = S

312: \cap X$. 

313: %In this way we relate the parameter space for curves $C$ to

314: %that for curves $C'$; we use this first to prove 

315: %the irreducibility of the spaces $\Hdg{d}{g}(X)$ and then to prove

316: %that the fibers of the Abel-Jacobi map $\Hdg{d}{g}(X) \to J^3(X)$ are

317: %unirational for $d \leq 5$.

318: In this way we relate the parameter space for curves $C$ to that for

319: curves $C'$.

320: In ~\cite{HRS1} it is proved

321: that for $d\leq 5$, each of the spaces $\Hdg{d}{g}(X)$ is

322: irreducible of dimension $2d$.  Combining this with the residuation

323: technique, we prove that for $d\leq 5$ the fibers of the Abel-Jacobi

324: map $\Hdg{d}{g}(X)\to J^3(X)$ are unirational.

325: 

326: \ps

327: 

328: Many of the results in this paper have also been proved in ~\cite{MT}

329: and ~\cite{AM} by considering moduli of vector bundles on $X$.  Our

330: theorem~\ref{thm-st30} corresponds to \cite[proposition 3.2]{MT}.  Our

331: corollary~\ref{cor-51fib} corresponds to ~\cite[theorem 5.6(iii)]{MT}

332: and ~\cite[theorem 3.2]{AM}.  Our proof is original and does not use

333: vector bundles on $X$, but rather an analysis of the divisors in the

334: intermediate Jacobian.  In addition, we give an analysis of quartic

335: rational curves (which are also considered in ~\cite{MT}), of quintic

336: rational curves, and of quintic curves of genus $2$.  Furthermore,

337: we prove that for every degree $d>3$, the

338: Abel-Jacobi map $\alpha_{d,0}:\Hdg{d}{0}(X)\rightarrow

339: J(X)$ is dominant and the general fiber is irreducible.

340: 

341: \subsection{Notation} \label{sec-not}

342: 

343: All schemes in this paper will be schemes over $\CC$.  All absolute

344: products will be understood to be fiber products over

345: $\text{Spec}(\CC)$.

346: 

347: \ps

348: 

349: For a projective variety $X$ and a numerical polynomial $P(t)$,

350: $\HI{P(t)}{X}$ denotes the corresponding Hilbert scheme.  For integers

351: $d,g$, $\Hdg{d}{g}(X)\subset \HI{dt+1-g}{X}$ denotes the open

352: subscheme parametrizing smooth, connected curves of degree $d$ and

353: genus $g$, and $\overline{\Hdg{d}{g}}(X)$ denotes the closure of

354: $\Hdg{d}{g}(X)$ in $\HI{dt+1-g}{X}$.

355:  

356: 

357: \section{Review of the Abel-Jacobi Map} \label{sec-rev}

358: 

359: Our object of study are the \emph{Abel-Jacobi maps} associated 

360: to families of 1-cycles on a smooth cubic hypersurface $X\subset

361: \PP^4$.  The reader is referred to ~\cite{CG} and ~\cite{VHS}

362: for full definitions.  Here we recall only a few facts

363: about Abel-Jacobi maps.

364: 

365: \ps

366: 

367: Associated to a smooth, projective threefold $X$ there is a complex

368: torus  

369: \begin{equation}

370: J^2(X) = H^3_\ZZ(X)\backslash H^3(X,\CC) / \lt(H^{3,0}(X)\oplus

371: H^{2,1}(X)\rt).

372: \end{equation}

373: In case $X$ is a cubic hypersurface in $\PP^4$ (in fact for any

374: \emph{rationally connected} threefold) then $J^2(X)$ is a principally

375: polarized

376: abelian variety with theta divisor $\Theta$.  Given an algebraic

377: 1-cycle $\gamma\in A_1(X)$ which 

378: is \emph{homologically equivalent to zero}~\cite[13]{VHS}, one can

379: associate a point $u_2(\alpha)$.  The construction is analogous to

380: the Abel-Jacobi map for a smooth, projective algebraic curve $C$ which 

381: associates to each 0-cycle $\gamma\in A_0(C)$ which is homologically

382: equivalent to zero a point $u_1(\alpha)\in J^1(C)$, the Jacobian

383: variety of $C$.  In particular $u_2:A_1(X)^{hom}\rightarrow J^2(X)$ is 

384: a group homomorphism.

385: 

386: \ps

387: 

388: Suppose that $B$ is a normal, connected variety of dimension $n$ and

389: $\Gamma\in A_{n+1}(B\times X)$ is an $(n+1)$-cycle such that for each

390: closed point $b\in B$ the corresponding cycle $\Gamma_b\in

391: A_1(X)$~\cite[\S 10.1]{F} is homologically equivalent to zero.  Then

392: in this case the set map $b\mapsto u_1(\Gamma_b)\in J^2(X)$ comes from 

393: a (unique) algebraic morphism $u=u_\Gamma:B\rightarrow J^2(X)$.  We

394: call this morphism the \emph{Abel-Jacobi map} determined by $\Gamma$.

395: 

396: \ps

397: 

398: More generally, suppose $B$ as above, $\Gamma\in A_{n+1}(B\times

399: X)$ is any $(n+1)$-cycle, and suppose $b_0\in B$ is some base-point.

400: Then we can form a new cycle $\Gamma'=\Gamma - \pi_2^*\Gamma_{b_0}$,

401: and for all $b\in B$ we have $\Gamma'_b = \Gamma_b - \Gamma_{b_0}$ is

402: homologically equivalent to zero.  Thus we have an algebraic morphism

403: $u=u_{\Gamma'}:B\rightarrow J^2(X)$.  Of course this morphism depends

404: on the choice of a base-point, but changing the base-point only

405: changes the morphism by a constant translation.  Thus we shall speak

406: of any of the morphisms $u_{\Gamma'}$ determined by $\Gamma$ and the

407: choice of a base-point as an \emph{Abel-Jacobi map} determined by

408: $\Gamma$.  

409: 

410: \ps

411: 

412: Suppose that $\Gamma_1,\Gamma_2\in A_{n+1}(B\times X)$ are two

413: $(n+1)$-cycles.  Then $u_{\Gamma_1+\Gamma_2}$ is the pointwise sum

414: $u_{\Gamma_1}+u_{\Gamma_2}$.  This trivial observation is frequently

415: useful.  

416: 

417: \ps 

418: 

419: \textbf{The Residuation Trick:}

420: Another useful observation is that any Abel-Jacobi morphism

421: $\alpha_\Gamma$ contracts all rational curves on $B$, since an Abelian 

422: variety contains no rational curves.  Combined with the observation in

423: the last paragraph, this leads to the residuation trick:  Suppose that

424: $B$ is a normal, rationally-connected variety and $\Gamma\in A_{n+1}(B\times 

425: X)$ is an $(n+1)$-cycle.  Then $u_\Gamma:B\rightarrow J(X)$ is a

426: constant map.  Now suppose that $B'\subset B$ is a normal closed

427: subvariety and that $\Gamma|_{B'\times X}$ decomposes as a sum of

428: cycles $\Gamma_1+\Gamma_2$.  Since $u_{\Gamma_1}+u_{\Gamma_2}$ equals

429: a constant map, we conclude that $u_{\Gamma_1}$ is the pointwise

430: inverse of $u_{\Gamma_2}$, up to a fixed additive constant.  

431: 

432: 

433: \section{Lines, Conics and Plane Cubics}~\label{sec-123}

434: 

435: The study of the Abel-Jacobi map associated to the space

436: $\Hdg{1}{0}(X)$ of lines on $X$ was carried out in  ~\cite{CG}.  

437: In this section we will summarize their results, which will also be

438: useful to us for studying Abel-Jacobi maps of higher degree curves.

439: In this section we also consider the Abel-Jacobi maps associated to

440: the spaces $\Hdg{2}{0}(X)$ and $\Hdg{3}{1}(X)$ of plane conics and

441: plane cubics in $X$.  In each case the Abel-Jacobi is trivial to

442: describe.  

443: 

444: \subsection{Lines}\label{subsec-lines}

445: 

446: For brevity we refer to the \emph{Fano scheme of lines},

447: $\Hdg{1}{0}(X)$, simply as $F$.  

448: Two general lines $L_1,L_2\subset \PP^4$ determine a hyperplane by

449: $\text{span}(L_1,L_2)$.  We generalize this as follows:

450: Let $(F\times F-\Delta)\xrightarrow{\Phi} 

451: \PP^{4\vee}$ denote the following set map:

452: \begin{equation}\label{eqn-16}

453:  \Phi \left( [L_1,L_2] \right) =

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

455:     \lt[ \text{span} \left( L_1,L_2 \right)\rt] &\text{if }L_1\cap L_2

456:   = \emptyset,  \\

457:   \lt[T_p X\rt]  & \text{if }p\in L_1\cap L_2

458:  \end{array} \right.

459: \end{equation} 

460: By~\cite[lemma 12.16]{CG}, $\Phi$ is algebraic.  

461: Let $X^\vee\subset \PP^{4\vee}$ denote the dual variety of $X$,

462: i.e. the variety 

463: parametrizing tangent hyperplanes to $X$.  Let $X^\vee_s\subset

464: X^\vee$ denote 

465: the subvariety parametrizing hyperplanes $H$ which are tangent to $X$ and

466: such that the singular locus of $H\cap X$ is not simply a single ordinary 

467: double point.  Let $U_s\subset U\subset F\times F$ denote the open

468: sets $\Phi^{-1}\lt(\PP^{4\vee} - X^\vee\rt)\subset \Phi^{-1}\lt(\PP^{4\vee} - 

469: X^\vee_s\rt)$.  And let $I\subset F\times F$ denote the divisor

470: parametrizing 

471: incident lines, i.e. $I$ is the closure of the set $\{\lt([L_1],[L_2]\rt) 

472: : L_1\neq L_2, L_1\cap L_2\neq\emptyset\}$.

473: In~\cite{CG}, Clemens and Griffiths completely describe both

474: the total 

475: Abel-Jacobi map $F\times F\xrightarrow{\psi} J(X)$ and the

476: Abel-Jacobi map $F\xrightarrow{i} J(X)$.  Here is a summary of their results

477: 

478: \ps

479: 

480: \begin{thm} \label{thm-cg}

481: \begin{enumerate} 

482: 

483: \item The Fano variety $F$ is a smooth surface and the 

484: Abel-Jacobi

485: map $F\xrightarrow{u} J(X)$ is a closed immersion~\cite[theorem 7.8,

486: theorem 12.37]{CG}. 

487: 

488: \item The induced map $Alb(F)=J^2(F)\rightarrow J(X)$ is an isomorphism of 

489: principally polarized Abelian varieties~\cite[theorem 11.19]{CG}.

490: \item The class of $u(F)$ in $J(X)$

491: is $\frac{[\Theta]^3}{3!}$~\cite[proposition 13.1]{CG}.

492: 

493: \item The difference of 

494: Abel-Jacobi maps 

495: \begin{equation}

496: \psi:F\times F \rightarrow J(X), \ \ \psi([L],[L']) = u([L]) - u([L'])

497: \end{equation}

498: maps $F\times F$ generically $6$-to-$1$ to the theta divisor

499: $\Theta~\subset ~J(X)$~\cite[section 13]{CG}.

500: 

501: \item Let $(\Theta-\{0\}) \xrightarrow{\mathcal{G}} 

502: \PP H^{1,2}(X)^\vee$ denote the Gauss map.  If we identify

503: $\PP H^{1,2}(X)$ with $\PP^4$ via the

504: Griffiths residue calculus~\cite{GR}, then the composite map

505: \begin{equation}

506: (F\times_C F-\Delta)\xrightarrow{\psi}(\Theta -\{0\})\xrightarrow{\mathcal{G}}

507: \PP^{4\vee}

508: \end{equation}

509: is just the map $\Phi$ defined above~\cite[formula 13.6]{CG}.

510: 

511: \item The fibers of the Abel-Jacobi map form a \emph{ Schl\"afli

512: double-six}, i.e. the general fiber of $\psi:F\times F\rightarrow J$

513: is of the form $\{(E_1,G_1),\dots,(E_6,G_6)\}$ where the lines

514: $E_i,G_j$ lie in a smooth hyperplane section of $X$, the $E_i$ are

515: pairwise skew, the $G_j$ are pairwise skew, and $E_i$ and $G_j$ are

516: skew iff $i=j$.  

517: 

518: There is a more precise result than above.  Let 

519: \begin{equation}R' \subset \lt(U\times_{\PP^{4\vee}} U\rt) \times F \times

520: \GR{}{3}{V}\times \GR{}{3}{V}\end{equation} 

521: be the closed subscheme parametrizing data

522: $\lt(\lt([L_1],[L_2]\rt),\lt([L_3],[L_4]\rt),[l], [H_1], [H_3]\rt)$

523: such that for each $i=1,\dots,4$, $l\cap L_i\neq\emptyset$ and such

524: that $H_1\cap X = l\cup L_1\cup L_4$, $H_2\cap X = l\cup L_2\cup L_3$.

525: Let $R\subset U\times_{\PP^{4\vee}} U$ be the image of $R'$ under the

526: projection map.  Let $\Delta\subset U\times U$ be the diagonal.  Then

527: the fiber product $U\times_\Theta U\subset U\times U$ is just the

528: union $R\cup\Delta$~\cite[p. 347-348]{CG}

529: 

530: \item The branch locus of $\Theta\xrightarrow{\mathcal{G}} \PP^{4\vee}$ equals

531: the branch locus of $F\times F \xrightarrow{\Phi} \PP^{4\vee}$ equals   

532: the dual variety of $X$, i.e. the variety 

533: parametrizing the tangent hyperplanes to $X$.  The 

534: ramification locus of $U \xrightarrow{\Phi} \PP^{4\vee}$ equals the

535: ramification 

536: locus of $U\xrightarrow{\psi} \Theta$ equals

537: the divisor $I$.  Each

538: such pair is a simple ramification point of both $\psi$ and

539: $\Phi$~\cite[lemma 13,8]{CG}.  

540: \end{enumerate}

541: \end{thm}

542: 

543: In~\cite{Tj}, Tjurin also analyzed cubic threefolds and the associated 

544: intermediate Jacobians.  We summarize his results:

545: 

546: \ps

547: 

548: \begin{thm}~\cite{Tj} \label{thm-tj}  

549: Let $\widetilde{J(X)}$ be the variety obtained by blowing up $0\in J(X)$ and

550: let $\widetilde{\Theta}$ be the proper transform of $\Theta$.  Let 

551: $\widetilde{F\times F}$

552: be the variety obtained by blowing up the diagonal in $F\times F$.  The

553: exceptional divisor $E\subset\widetilde{J(X)}$ is isomorphic to $\PP^4$ and

554: this isomorphism identifies the intersection $\widetilde{\Theta}\cap E$ with

555: our original cubic threefold $X$.  The exceptional divisor $E'\subset

556: \widetilde{F\times F}$ is the projective bundle $\PP T_F$.  

557: In fact there is

558: an isomorphism of sheaves of the tautological rank 2 sheaf $S=S(2,V)$

559: and the tangent bundle $T_F$ so that $E'\cong \PP_{F}S$.  The induced

560: morphism $\PP_{F}S\xrightarrow{\tilde{\psi}} \PP^4$ is just the usual

561: morphism induced by the map of sheaves $S\hookrightarrow V\otimes_\CC \OO_F$.

562: In particular, this morphism is generically $6$-to-$1$ onto $X$ with 

563: ramification locus $\Sigma_1\subset F$ consisting of the lines of "second

564: type", i.e. lines $L\subset X$ such that $N_{L/X}\cong \OO_L(-1)\oplus\OO_L(1)$.

565: Let $D\subset F\times F$ denote the divisor of intersecting lines and

566: let $\widetilde{D}$ be the proper transform of $D$.  Then $\widetilde{D}\cap E'

567: =\Sigma_1$ so that we finally have the result: The morphism

568: $\widetilde{F\times F}\rightarrow{\widetilde{\Theta}}$ is finite of degree

569: $6$ with ramification locus $\widetilde{D}$.

570: \end{thm}

571: 

572: \subsection{Conics}\label{subsec-cons}

573: 

574: Next we consider the space $\Hdg{2}{0}(X)$ parametrizing smooth plane

575: conics on $X$.  Given a plane conic $C\subset X$, consider the plane

576: cubic curve which is the intersection $\text{span}(C)\cap X$.  This

577: curve contains $C$ is an irreducible component, and the residual

578: component is a line $L\subset X$.  In this way we have a morphism

579: $\Hdg{2}{0}(X) \rightarrow F$ by $[C]\mapsto [L]$.  Conversely, given

580: a line $L\subset X$ and a $2$-plane $P\subset \PP^4$ which contains

581: $L$, then the residual to $L$ in $P\cap X$ is a plane conic.  In this

582: way one sees that $\Hdg{2}{0}(X)$ is isomorphic to an open subset of

583: the $\PP^2$-bundle

584: \begin{equation}

585: \PP Q = \{([L],[P])\in F\times \mathbf{G}(2,4) | L\subset P\}.

586: \end{equation}

587: In ~\cite[prop 3.4]{HRS1} we prove the stronger result:

588: 

589: \begin{prop}~\label{prop-cons1}  

590: The morphism

591: $\Hdg{2}{0}(X)\rightarrow F$ is isomorphic to an open subset of a

592: $\PP^2$-bundle $\PP Q\rightarrow F$.  In particular, $\Hdg{2}{0}(X)$

593: is smooth and connected of dimension $4$.  Moreover $\PP Q$ is the

594: normalization of $\HI{2t+1}{X}$.  

595: \end{prop}

596: 

597: Next we describe the Abel-Jacobi map for conics.  

598: By the residuation trick, the Abel-Jacobi map $u_{D_2}:\PP

599: Q\rightarrow J(X)$ is, up to a 

600: fixed translation, the pointwise inverse of the composite $\PP

601: Q\xrightarrow{\pi} F\xrightarrow{u} J(X)$.  Thus the fibers of

602: $u_{D_2}:\PP Q\rightarrow J(X)$ are just the fibers of $\pi$,

603: i.e. $\PP^2$'s.  

604: 

605: 

606: \subsection{Plane Cubics}\label{sec-pcubs}

607: 

608: Every curve $C\subset \PP^4$ with Hilbert polynomial $3t$ is a plane

609: cubic, and the 2-plane $P=\text{span}(C)$ is unique; we have

610: that $C=X\cap P$.  Therefore the Hilbert scheme $\HI{3t}{X}$ is just

611: the Grassmannian $\mathbb{G}(2,\PP^4)$ of 2-planes in $\PP^4$ and

612: $\Hdg{3}{1}(X)$ is just an open subset of $\mathbb{G}(2,\PP^4)$.    

613: Since

614: $\mathbb{G}(2,\PP^4)$ is rational, it follows that both the total Abel-Jacobi

615: map $\psi:\HI{3t}{X}\times \HI{3t}{X}\rightarrow J(X)$ and the

616: Abel-Jacobi map $i:\HI{3t}{X}\rightarrow J(X)$ are constant maps.  

617: 

618: \subsection{Two Useful Results}\label{sec-res}

619: 

620: Recall that $\psi:F\times F\rightarrow J(X)$ is defined as the

621: \emph{difference map} $\psi = u\circ \pi_1-u\circ \pi_2$.  Now we

622: consider the sum map $\psi'=u\circ\pi_1+u\circ\pi_2$.  

623: Alternatively one may define $\psi'$ to be the (Weil extension of the)

624: morphism $(F\times F)-\Delta \rightarrow J(X)$ which is the

625: Abel-Jacobi map corresponding to the flat family $Z\xrightarrow{\pi}

626: (F\times F-\Delta)$ defined as follows.  For $L_1$ and $L_2$ skew

627: lines, we define $Z_{\lt([L_1],[L_2]\rt)}$ to be the disjoint union of

628: $L_1$ and $L_2$.  If $p\in L_1\cap L_2$, we define

629: $Z_{\lt([L_1],[L_2]\rt)}$ to be the subscheme whose reduced scheme is

630: just the union $L_1\cup L_2$ and which has an embedded point at $p$

631: corresponding to the normal direction of $\text{span}(L_1,L_2)\subset

632: T_p X$.

633: 

634: \ps

635: 

636: Clearly $\psi'$ commutes with the involution 

637: \begin{equation}

638: \iota:F\times F \rightarrow F\times F, ([L_1],[L_2])\mapsto

639: ([L_2],[L_1]). 

640: \end{equation}

641: So $\psi'$ factors through the

642: quotient $F\times F\rightarrow \text{Sym}^2 F$.  We will denote by

643: $\text{Sym}^2 F

644: \xrightarrow{\psi''} J(X)$ the induced morphism.

645: Let us define $\Theta'$ to be the scheme-theoretic image of $\psi'$.

646: 

647: \ps

648: 

649: First we prove an easy lemma about

650: cohomology of complex tori.  Given a $g$-dimensional complex torus $A$

651: and a sequence of nonzero integers $(n_1,\dots,n_r)$, consider the

652: holomorphic map $f=f_{(n_1,\dots,n_r)}:A^r\rightarrow A$ defined by

653: $f(a_1,\dots,a_r) = n_1a_1 + \dots + n_r a_r$.  There is an induced

654: Gysin image map on cohomology $f_*:H^{p+2(r-1)g}(A^r)\rightarrow

655: H^p(A)$.  Now by the K\"unneth formula, the cohomology $H^q(A^r)$ is a

656: direct sum of K\"unneth components $H^{q_1}(A)\otimes \dots \otimes

657: H^{q_r}(A)$ with $q=q_1+\dots+q_r$.

658: 

659: \begin{lem}\label{lem-cohom}  For each integer $r\geq 1$, for each

660: integer $0\leq p\leq 2g$, and for each decomposition $(q_1,\dots,q_r)$

661: with $q_1+\dots+q_r = p + 2(r-1)g$, there is a homomorphism

662: \begin{equation}

663: g=g_{(q_1,\dots,q_r)}:H^{q_1}(A)\otimes\dots\otimes

664: H^{q_r}(A)\rightarrow H^p(A)

665: \end{equation}

666: such that for each sequence of nonzero integers $(n_1,\dots,n_r)$,

667: with $f$ defined as above, the restriction of the Gysin image map

668: \begin{equation}

669: f_*:H^{q_1}(A)\otimes \dots\otimes H^{q_r}(A)\rightarrow H^p(A)

670: \end{equation}

671: satisfies $f_* = \lt(n_1^{2g-q_1}\cdot \dots \cdot n_r^{2g-q_r}\rt) g$.  

672: \end{lem}

673: 

674: \begin{proof}  This is just a computation.  By Poincar\'e duality, to

675: give the homomorphism $g$, it is equivalent to give a bilinear pairing

676: \begin{equation}

677: \lt(H^{q_1}(A)\otimes \dots \otimes H^{q_r}(A)\rt) \times H^{2g-p}(A)

678: \rightarrow \ZZ. 

679: \end{equation}

680: Moreover, since $H^{2g-p}(A)= \bigwedge^{2g-p}H^1(A)$, it suffices to

681: define the pairing for pure wedge powers

682: $\alpha = \alpha_1\wedge\dots\wedge\alpha_{2g-p}\in H^{g-p}(A)$.  

683: Define $S$ to be the set of functions

684: $\sigma:\{1,2,\dots,2g-p\}\rightarrow {1,2,\dots, r}$ such that for each 

685: $i=1,\dots, r$, we have $q_i+\#\sigma^{-1}(i) = 2g$.  

686: We define the pairing by taking the wedge product in $H^*(A^r)$ and

687: then taking the degree as follows:

688: \begin{equation}

689: \langle \beta = \beta_1\otimes \dots \otimes \beta_r, \alpha_1\wedge\dots

690: \wedge \alpha_{2g-p}\rangle = \text{deg}\sum_{\sigma\in S}

691: \pi_1^*\beta_1 \wedge \dots \wedge \pi_r^*\beta_r \wedge

692: \pi_{\sigma(1)}^*\alpha_1 \wedge \dots \wedge

693: \pi_{\sigma(2g-p)}^*\alpha_{2g-p}.

694: \end{equation}

695: 

696: \

697: 

698: The fact that $f_* = \lt(n_1^{2g-q_1}\cdot \dots \cdot

699: n_r^{2g-q_r}\rt) g$ follows from the projection formula

700: \begin{equation}

701: f_*(\beta)\wedge \alpha = f_*(\beta\wedge f^*\alpha)

702: \end{equation}

703: together with the formula

704: \begin{eqnarray}

705: f^*(\alpha_1\wedge\dots\wedge\alpha_r) = f^*(\alpha_1)\wedge \dots

706: \wedge f^*(\alpha_r) =  \\

707: \lt( \sum_{\sigma(1)=1}^r

708: n_{\sigma(1)}\pi_{\sigma(1)}^*(\alpha_1)\rt) \wedge \dots \wedge \lt(

709: \sum_{\sigma(2g-p)=1}^r

710: n_{\sigma(2g-p)}\pi_{\sigma(2g-p)}^*(\alpha_{2g-p})\rt). \nonumber

711: \end{eqnarray}

712: 

713: \end{proof}

714: 

715: \ps

716: 

717: \begin{thm}\label{sec-thmu}

718: The image $\Theta'$ is a divisor in $J(X)$ which is algebraically 

719: equivalent to $3\Theta$ and $\psi':\text{Sym}^2(F)\rightarrow \Theta'$

720: is birational.  The singular locus of $\Theta'$ has codimension

721: $2$.  For general $a\in J(X)$, we have that $\Theta'\cap(a+\Theta)$

722: is irreducible.

723: \end{thm}

724: 

725: \begin{proof}

726: For divisors on $J(X)$, algebraic equivalence is equivalent to

727: homological equivalence, so we shall establish that $\Theta'$ is

728: homologically equivalent to $3\Theta$.  

729: 

730: \

731: 

732: Using the notation of the previous lemma,

733: both $\Theta$ and $\Theta'$ are images of the subvariety $u(F)\times

734: u(F) \subset J \times J$ under the morphism $f_{(1,-1)}$ and

735: $f_{(1,1)}$ respectively.  By lemma~\ref{lem-cohom} we know there is a

736: linear map $g:H^4(J)\otimes H^4(J)\rightarrow H^2(J)$ such that 

737: \begin{eqnarray}

738: f_{(1,-1)*}(P.D.[u(F)\times u(F)]) = (1)^6(-1)^6 g(P.D.[u(F)\times

739: u(F)]), \\

740: f_{(1,1)*}(P.D.[u(F)\times u(F)]) = (1)^6(1)^6 g(P.D.[u(F)\times

741: u(f)]) \nonumber 

742: \end{eqnarray}

743: where $P.D.[u(F)\times u(F)]$ is the Poincar\'e dual of the homology

744: class of $u(F)\times u(F)$.  We know the mapping $\psi: F\times F

745: \rightarrow \Theta$ has degree $6$.  So if $\psi': F\times

746: F\rightarrow \Theta'$ has degree $d$, then it follows that

747: \begin{equation}

748: d P.D.[\Theta'] = f_{(1,1)*}(P.D.[u(F)\times u(F)]) =

749: f_{(1,-1)*}(P.D.[u(F)\times u(F)]) = 6 P.D.[\Theta].

750: \end{equation}

751: In other words, the cohomology class of $\Theta'$ is equal to

752: $\frac{6}{d}$ times the cohomology class of $\Theta$.  Of course $d$

753: is divisible by $2$ because $\psi'$ factors through $F\times F

754: \rightarrow \text{Sym}^2(F)$.  It remains to prove that $d$ is

755: precisely $2$.

756: 

757: \ps

758: 

759: Now let $\mathcal{G}':(\Theta')^{ns}\rightarrow \PP^{4\vee}$ be the

760: Gauss map defined on the nonsingular locus of $\Theta'$.  Of course

761: the differentials $d\psi$ and $d\psi'$ are simply given by $d u\circ

762: d\pi_1 - d u\circ d\pi_2$ and by $d u\circ d\pi_1 + d u\circ d\pi_2$

763: respectively.  In particular for each point $(L,M)\in F\times F$, the

764: subspace $\text{image}(d\psi)\subset T_0 J$ and

765: $\text{image}(d\psi)\subset T_0 J$ are equal, i.e. the composites

766: \begin{eqnarray}

767: (F\times F-\Delta)\xrightarrow{\psi} \Theta \xrightarrow{\mathcal{G}}

768: \PP^{4\vee} \\

769: F\times F\xrightarrow{\psi'} \Theta'

770: \xrightarrow{{\mathcal{G}}'} \PP^{4\vee}

771: \end{eqnarray}

772: are equal as rational maps; so both are equal to the rational map

773: $\Phi$.  Therefore, for each point $\lt([L_1],[L_2]\rt)$ whose image

774: under $\psi'$ lies in the nonsingular locus of $\Theta'$, we see that

775: the fiber of $\psi'$ containing $\lt([L_1],[L_2]\rt)$ is contained in

776: the fiber of $\Phi$ containing $\lt([L_1],[L_2]\rt)$.  So, we are

777: reduced to showing that, for generic $\lt([L_1],[L_2]\rt)$, if

778: $\psi'\lt([L_1],[L_2]\rt)=\psi'\lt([L_3],[L_4]\rt)$ with $L_1,L_2,L_3$

779: and $L_4$ contained in a smooth hyperplane section of $X$, then either

780: $\lt([L_1],[L_2]\rt)=\lt([L_3],[L_4]\rt)$ or

781: $\lt([L_2],[L_1]\rt)=\lt([L_4],[L_3]\rt)$.

782: 

783: \ps

784: 

785: A bit more generally, suppose that $\lt([L_1],[L_2]\rt)\in U$ is a pair

786: of skew lines.  We will show

787: that the fiber $\lt(\psi'\rt)^{-1}\lt(\psi'\lt([L_1],[L_2]\rt)\rt) = \lt\{

788: \lt([L_1],[L_2]\rt),\lt([L_2],[L_1]\rt)\rt\}$.  Indeed, suppose that 

789: $\psi'\lt([L_1],[L_2]\rt) =\psi'\lt([L_3],[L_4]\rt)$ and suppose

790: that $\lt([L_3],[L_4]\rt)\neq \lt([L_2],[L_1]\rt)$.  Then

791: $\psi\lt( [L_1],[L_4]\rt)=\psi\lt([L_3],[L_2]\rt)$.  Therefore

792: $\Phi\lt([L_1],[L_4]\rt)=\Phi\lt([L_3],[L_2]\rt)$.  Let us call this

793: common hyperplane $H$.  Then $L_1\subset H$ and $L_2\subset H$.  Therefore

794: $H=\Phi\lt([L_1],[L_2]\rt)$.  Since we have $L_4\neq L_1$ and $L_3\neq L_2$,

795: we conclude that $ \lt([L_1],[L_4]\rt), \lt([L_3],[L_2]\rt)\in U$.

796: Therefore, by theorem~\ref{thm-cg} (7), we have either $\lt([L_1],[L_4]\rt)=

797:  \lt([L_3],[L_2]\rt)$, or else there exists a line $l\in H$ such that

798: $l\cup L_1\cup L_2$ is the intersection of $X$ with a $\PP^2$.  But $L_1$

799: and $L_2$ cannot lie in a common $\PP^2$ since they are skew.  Therefore

800: we conclude that $\lt([L_1],[L_4]\rt)=

801:  \lt([L_3],[L_2]\rt)$, i.e. $\lt([L_3],[L_4]\rt)=\lt([L_1],[L_2]\rt)$.

802: 

803: \ps

804: 

805: So we deduce that $d=2$ and thus $\Theta'$ is algebraically equivalent 

806: to $3\Theta$.  But we deduce even more.  The image of 

807: $U-I\cap U$ in $\text{Sym}^2 F$ is smooth, let's call it $U'$.  Since

808: the map $\psi':U'\rightarrow\psi'(U')$ is bijective and since

809: $\text{rank}(d\psi') = \text{rank}(d\psi) = 4$ on $U - I\cap U$, 

810: it follows from Zariski's main

811: theorem~\cite[p. 288-289]{M} that 

812: $\psi'(U')$ is 

813: smooth and $U'\xrightarrow{\psi'}\psi'{U'}$ is an isomorphism.  Now the

814: complement of $U$ in $F\times F$ has codimension $2$.  Therefore

815: $\psi'(F\times F-U)$ has codimension at least $2$ inside of $\Theta'$.

816: But in fact $\psi'(I)$ also has codimension $2$.  Consider the rational

817: map $I\xrightarrow{\rho} F$ defined by sending a pair of incident lines

818: $\lt([L_1],[L_2]\rt)$ to the residual line $l$ such that

819: $\text{span}(L_1,L_2)\cap X = l\cup L_1\cup L_2$.  By the residuation trick

820: we have that the restriction of $\psi'$ to $I$ equals the composition 

821: of $\rho$ with the pointwise negative Abel-Jacobi map $-i$.  In

822: particular $\psi'(I)$ is just $-u(F)$ up to translation.  So $\psi'(I)$ has

823: codimension $2$ inside of $\Theta'$.  Therefore $\psi'(U')\subset\Theta'$

824: has complement of codimension $2$.  So the singular locus

825: of  $\Theta'$, has codimension $2$.

826: 

827: \ps

828: 

829: 

830: Since

831: $\Theta'\cap(a+\Theta)$ is a positive dimensional intersection of

832: ample divisors,  

833: $\Theta'\cap(a+\Theta)$ is connected.

834: Since $\Theta'\cap(a+\Theta)$ is a

835: complete intersection of Cartier divisors, it follows from

836: Hartshorne's connectedness theorem \cite[thereom 18.12]{E} that

837: $\Theta'\cap(a+\Theta)$ is connected in codimension $2$.  Thus to

838: prove that $\Theta'\cap (a+\Theta)$ is irreducible, it suffices to

839: prove that the singular locus of $\Theta'\cap (a+\Theta)$ has

840: codimension at least $2$.

841: By the Bertini-Kleiman theorem~\cite[theorem III.10.8]{H}, we

842: have that for general $a\in J(X)$, the intersection $\Theta'\cap(a+\Theta)$

843: is smooth away from the intersection of each divisor with the singular 

844: locus of the other divisor.  But by the last paragraph and by

845: theorem~\ref{thm-tj}, we see that the singular loci of $\Theta$ and

846: $\Theta'$ both have codimension at least $2$ (in $\Theta$ or $\Theta'$ 

847: respectively).  So it follows that for $a$ general, the singular locus 

848: of $\Theta'\cap(a+\Theta)$ has codimension at least $2$.  Thus we

849: conclude that for general $a\in J(X)$, the intersection

850: $\Theta'\cap(a+\Theta)$ is irreducible.

851: 

852: \end{proof}

853: 

854: We also use the following enumerative lemma, which is proved

855: in~\cite[lemma 4.2]{HRS1}.  

856: 

857: \ps

858: 

859: \begin{lem}\label{lem-enum}  Suppose that $C\subset X$ is a smooth

860: curve of genus $g$ 

861: and degree $d$.  Let $B_C\subset F$ denote the scheme parametrizing

862: lines in $X$ which intersect $C$ in a scheme of degree $2$ or more.

863: Define $b(C)=\frac{5d(d-3)}{2}+6-6g$.  If

864: $B_C$ is not positive dimensional and if $b(C)\geq 0$, then the degree of

865: $B_C$ is $b(C)$.  

866: \end{lem}

867: 

868: 

869: \section{Twisted cubics}~\label{sec-30}

870: 

871: We  now begin in earnest the analysis of the geometry of cubic,

872: quartic and quintic curves on our cubic threefold 

873: $X \subset \P^4$. In each case our goal is the proof of the Main

874: Theorem \ref{mainthm} for the variety 

875: $\Hdg{d}{g}(X)$ parametrizing curves of this degree and genus. We

876: start with the variety $\Hdg{3}{0}(X)$ 

877: parametrizing rational curves of degree 3---that is \emph{twisted cubics}. 

878: 

879: \ps

880: 

881: In ~\cite[theorem 4.4]{HRS1} we prove the following result

882: 

883: \begin{thm}\label{thm-irr30} 

884: The space $\Hdg{3}{0}(X)$ is a smooth,

885: irreducible $6$-dimensional variety.

886: \end{thm}

887: 

888: We have a morphism 

889: \begin{equation}

890: \Hdg{3}{0}(\PP^4)\xrightarrow{\sigma^{3,0}} \PP^{4\vee}

891: \end{equation}

892: defined by sending $[C]$ to $\text{span}(C)$.

893: This morphism makes $\Hdg{3}{0}(\PP^4)$ 

894: into a locally trivial bundle over $\PP^{4\vee}$ with fiber

895: $\Hdg{3}{0}(\PP^3)$ .  Recall from section~\ref{sec-123} that we defined

896: $X^\vee

897: \subset \PP^{4\vee}$ to be the dual variety of $X$ which parametrizes

898: tangent hyperplanes to $X$ and we defined $U$ to

899: be the complement of $X^\vee$ in $\PP^{4\vee}$.  Then we define 

900: $\Hdg{3}{0}_U(X)$ to be the

901: open subscheme of $\Hdg{3}{0}(X)$ which parametrizes twisted cubics,

902: $C$, in $X$ 

903: such that $\sigma^{3,0}([C])\in U$.  By the graph construction we may

904: consider $\Hdg{3}{0}_U(X)$ as a locally closed subvariety of

905: $\HI{3t+1}{X}\times U$.  

906: Let $\overline{\mathcal{H}}\subset \HI{3t+1}{X}\times U$ denote the

907: closure of $\Hdg{3}{0}_U(X)$ with the reduced induced scheme structure.

908: Denote by $\overline{\mathcal{H}} \xrightarrow{f} U(X)$ the projection map. 

909: 

910: \begin{thm}\label{thm-st30}  

911: Let

912: $\overline{\mathcal{H}}\xrightarrow{f''} U'\xrightarrow{f'} 

913: U(X)$ 

914: be the Stein factorization of $\overline{\mathcal{H}}\xrightarrow{f} U(X)$.

915: Then 

916: $\overline{\mathcal{H}}\xrightarrow{f''} U'$ is isomorphic to a $\PP^2$-bundle

917: $\PP_{U'}(E)\xrightarrow{\pi} U'$ with $E$ a locally free sheaf of rank $3$.

918: And $U'\xrightarrow{f'} U$ is an unramified finite morphism of degree

919: $72$.  Moreover, the Abel-Jacobi map $\overline{\mathcal{H}}\xrightarrow{i}

920: J(X)$ 

921: factors as $\overline{\mathcal{H}}\xrightarrow{f''} U'\xrightarrow{i'}

922: J(X)$ where 

923: $U'\xrightarrow{i'} J(X)$ is a birational morphism of $U'$ to a translate

924: of $\Theta$.

925: \end{thm} 

926: 

927: \begin{proof}

928: This theorem is  ~\cite[theorem 4.5]{HRS1}, and is proved there.  For

929: brevity's sake, we recall just the sketch of the proof.

930: 

931: \ps

932: 

933: One can form the subvariety $\mathcal{X}\subset U\times X$ as the

934: universal smooth hyperplane section of $X$.  Each fiber of

935: $\mathcal{X}\rightarrow U$ is a smooth cubic surface.  We associate to

936: this family of cubic surfaces the finite \'etale morphism

937: \begin{equation}

938: \rho: \text{Pic}^{3,0}(\mathcal{X}/U) \rightarrow U

939: \end{equation}

940: whose fiber over a point $[H]\in U$ is simply the set of divisor

941: classes $D$ on the smooth cubic surface $H\cap U$ such that $D.D =1$

942: and $D.h=3$

943: with $h$ the hyperplane class.  Now the Weyl group $\mathcal{W}(E_6)$

944: acts transitively on the set of such divisor classes $D$. It follows

945: that with respect to the model of a cubic surface as 

946: $\PP^2$ blown up at 6 points, each divisor class $D$ above corresponds

947: to the class of a line on $\PP^2$.  From this it follows that the

948: induced morphism 

949: \begin{equation}

950: g:\Hdg{3}{0}(X)\rightarrow \text{Pic}^{3,0}(\mathcal{X}/U))

951: \end{equation}

952: is surjective and isomorphic to an open subset of a $\PP^2$-bundle on

953: $\text{Pic}^{3,0}(\mathcal{X}/U)$.  In particular, since

954: $u:\Hdg{3}{0}(X)\rightarrow J(X)$ contracts all rational curves, there

955: is an induced morphism $u':\text{Pic}^{3,0}(\mathcal{X}/U)\rightarrow

956: J(X)$ such that $u=u'\circ g$.  

957: 

958: \ps

959: 

960: It only remains to show that $u'$ maps

961: $\text{Pic}^{3,0}(\mathcal{X}/U)$ birationally to its image.  This is

962: proved by analyzing ``Z''s of lines, i.e. configurations of lines on a

963: cubic surface $H\cap X$ whose dual graph is just the connected graph with 3

964: vertices and no loops.  Using ~\ref{thm-cg}, part 5, we conclude

965: that two ``Z''s have the same image in $J(X)$ iff they are in the same

966: linear equivalence class on $H\cap X$.  

967: \end{proof}

968: 

969: \begin{cor}\label{cor-aj30}  The Abel-Jacobi map

970: $u_{3,0}:\Hdg{3}{0}(X)\rightarrow J(X)$ dominates a translate of

971: $\Theta$ and is birational to a $\PP^2$-bundle over its image.

972: \end{cor}

973: 

974: \begin{proof}  Since $\Hdg{3}{0}(X)$ is irreducible, $\Hdg{3}{0}_U(X)$ 

975: is dense in $\Hdg{3}{0}(X)$.

976: \end{proof}

977: 

978: 

979: \subsection{Quartic Elliptic Curves}

980: 

981: Recall that the normalization of $\HI{2t+1}{X}$ is isomorphic to the

982: $\PP^2$-bundle $\PP 

983: Q\rightarrow F$ which parametrizes pairs $(L,P)$ which $L\subset X$

984: a line and $P\subset\PP^4$ a $2$-plane containing $L$.    

985: Let $A\xrightarrow{g} \PP Q$ denote the $\PP^1$-bundle

986: which parametrizes triples $(L,P,H)$ with $H$ a hyperplane containing

987: $P$.  Let $I_{4,1}\xrightarrow{h} A$ denote the $\PP^4$-bundle parametrizing

988: 4-tuples $(L,P,H,Q)$ where $Q\subset H$ is a quadric surface

989: containing the conic $C\subset X\cap P$. 

990: Notice that $I_{4,1}$ is smooth

991: and connected of dimension $4+1+4=9$.

992: 

993: \ps

994: 

995: Let $D\subset I_{4,1}\times X$

996: denote the intersection of the universal quadric surface over $I_{4,1}$ with 

997: $I_{4,1}\times X\subset I_{4,1}\times \PP^4$.  Then $D$ is a local complete

998: intersection scheme.  By the Lefschetz hyperplane theorem, $X$

999: contains no quadric surfaces; therefore $D\rightarrow I_{4,1}$ has constant

1000: fiber dimension $1$ and so is flat.  Let $D_1\subset I_{4,1}\times X$ denote

1001: the pullback from $\PP Q\times X =\HI{2t+1}{X}\times X$ of the

1002: universal family of conics.  Since $I_{4,1}\times X\rightarrow \PP Q\times

1003: X$ is smooth and the universal family of conics is a local complete

1004: intersection which is flat over $\PP Q$, we conclude that also $D_1$ is a

1005: local complete intersection which is flat over $I_{4,1}$.  Clearly

1006: $D_1\subset D$.  Thus by corollary~\cite[corollary 2.7]{HRS1}, we see that the

1007: residual $D_2$ of $D_1\subset D$ is Cohen-Macaulay and flat over $I_{4,1}$.

1008: 

1009: \ps

1010: 

1011: By the base-change property in corollary~\cite[corollary 2.7]{HRS1},

1012: we see that 

1013: the fiber of $D_1\rightarrow I_{4,1}$ over a point $(L,P,H,Q)$ is simply the

1014: residual of $C\subset Q\cap X$.  If we choose $Q$ to be a smooth

1015: quadric, i.e. $Q\cong \PP^1\times \PP^1$, then $C\subset Q$ is a

1016: divisor of type $(1,1)$ and $X\cap Q\subset Q$ is a divisor of type

1017: $(3,3)$.  Thus the residual curve $E$ is a divisor of type $(2,2)$,

1018: i.e. a quartic curve of arithmetic genus 1.  Thus $D_2\subset I_{4,1}\times

1019: X$ is a family of connected, closed subschemes of $X$ with Hilbert

1020: polynomial $4t$.  So we have an induced map $f:I_{4,1}\rightarrow

1021: \HI{4t}{X}$.  

1022: 

1023: \ps

1024: 

1025: \begin{prop}\label{lem-qell}  The image of the morphism above

1026: $f:I_{4,1}\rightarrow\HI{4t}{X}$ is the closure $\overline{\Hdg{4}{1}}(X)$ 

1027: of $\Hdg{4}{1}(X)$.  Moreover the open set $f^{-1}\Hdg{4}{1}(X)\subset 

1028: I_{4,1}$ is a $\PP^1$-bundle over $\Hdg{4}{1}(X)$.  Thus $\Hdg{4}{1}(X)$ is

1029: smooth and connected of dimension 8.

1030: \end{prop}

1031: 

1032: \begin{proof}

1033: This is ~\cite[proposition 5.1]{HRS1}.

1034: \end{proof}

1035: 

1036: If $(L,P,H,Q)$ is a point in the fiber over $[E]$, then

1037: $H=\text{span}(E)$.  And since there is no rational curve in $F$, we

1038: also have that $L$ is constant in the fiber.

1039: So we have a well-defined morphism $m:\Hdg{4}{1}(X)\rightarrow \PP

1040: Q^\vee$, where $\PP Q^\vee$ is the $\PP^2$-bundle over $F$ parametrizing pairs

1041: $([L],[H]), L\subset H$.  For a general $H$, the intersection $Y=H\cap X$ is

1042: a smooth cubic surface.  And the fiber $m^{-1}([L],[H])$ is an open

1043: subset of the complete linear series $|\OO_Y{L+h}|$, where $h$ is the

1044: hyperplane class on $Y$.  Thus $m:\Hdg{4}{1}(X)\rightarrow \PP Q^\vee$

1045: is a morphism of smooth connected varieties which is birational to a

1046: $\PP^4$-bundle.  Composing $m$ with the projection $\PP Q^\vee$ yields

1047: a morphism $n:\Hdg{4}{1}(X)\rightarrow F$ which is birational to a

1048: $\PP^4$-bundle over a $\PP^2$-bundle.  

1049: 

1050: \begin{cor}~\label{cor-qell}

1051: By the residuation trick, we conclude 

1052: the Abel-Jacobi map

1053: $u_{4,1}:\Hdg{4}{1}(X)\rightarrow J(X)$ is equal, up to a

1054: fixed translation, to the composite:

1055: \begin{equation}\begin{CD}

1056: \Hdg{4}{1}(X) @> n >> F @> u_{1,0} >> J(X).

1057: \end{CD}\end{equation}

1058: Thus the general fiber of $u_{4,1}$ equals the general fiber

1059: of $n$, and so is isomorphic to an open subset of a $\PP^4$-bundle

1060: over $\PP^2$.  

1061: \end{cor}

1062: 

1063: \section{Cubic scrolls and applications}~\label{sec-cuscr}

1064: 

1065: \subsection{Preliminaries on cubic scrolls}~\label{subsec-prelim}

1066: 

1067: In the next few sections we will use residuation in a cubic surface

1068: scroll. We start by collecting 

1069: some basic facts about these surfaces.

1070: 

1071: \ps

1072: 

1073: There are several equivalent descriptions of cubic scrolls.  

1074: \begin{enumerate}

1075: \item A cubic scroll $\Sigma \subset \PP^4$ is a connected, smooth

1076: surface with Hilbert polynomial $P(t) = \frac{3}{2}t^2 + \frac{5}{2}t

1077: + 1$. 

1078: \item A cubic scroll $\Sigma\subset

1079: \PP^4$ is the determinantal variety defined by the $2\times 2$ minors of 

1080: a matrix of linear forms:

1081: \begin{equation}\lt[ \begin{array} {ccc}

1082:  L_1 & L_2 & L_3 \\

1083:  M_1 & M_2 & M_3 

1084: \end{array} \rt]\end{equation}

1085: such that for each row or column, the linear forms in

1086: that row or column are linearly independent 

1087: \item A cubic scroll $\Sigma\subset \PP^4$ is the \emph{join} of an

1088: isomorphism $\phi:L\rightarrow C$.  Here $L\subset \PP^4$ is a line and

1089: $C\subset \PP^4$ such that $L\cap

1090: \text{span}(C) =\emptyset$.  The join of $\phi$ is defined as the

1091: union over all $p\in L$ of the line $\text{span}(p,\phi(p))$.  

1092: \item A cubic scroll $\Sigma\subset \PP^4$ is the image of a morphism $f:\PP

1093: E\rightarrow \PP^4$ where $E$ is the rank 2 vector bundle on $\PP^1$, 

1094: $E=\OO_{\PP^1}(-1)\oplus \OO_{\PP^1}(-2)$, and the morphism $f:\PP

1095: E\rightarrow \PP^4$ is such that $f^*\OO_{\PP^4}(1) = \OO_{\PP E}(1)$

1096: and the pullback map $H^0(\PP^4,\OO_{\PP^4}(1))\rightarrow H^0(\PP

1097: E,\OO_{\PP E}(1)$ is an isomorphism.

1098: \item A cubic scroll $\Sigma\subset \PP^4$ is as

1099: a minimal variety, i.e. $\Sigma\subset \PP^4$ is any smooth connected 

1100: surface with $\text{span}(\Sigma)=\PP^4$ which has the minimal

1101: possible degree for such a surface, namely $\text{deg}(\Sigma)=3$. 

1102: \item A cubic scroll $\Sigma\subset \PP^4$ is a smooth surface residual to a

1103: 2-plane $\Pi$ in the base locus of a pencil of quadric hypersurfaces

1104: which contain $\Pi$.  

1105: \end{enumerate}

1106: 

1107: \ps

1108: 

1109: From the fourth description $\Sigma = f(\PP E)$ we see that

1110: $\text{Pic}(\Sigma) = \text{Pic}(\PP E) \cong \ZZ^2$.  Let $\pi:\PP

1111: E\rightarrow \PP^1$ denote the projection morphism and let

1112: $\sigma:\PP^1\rightarrow \PP E$ denote the unique section whose image

1113: $D=\sigma(\PP^1)$ has self-intersection $D.D=-1$.  Then $f(D)$ is a

1114: line on $\Sigma$ called the \emph{directrix}.  And for each $t\in

1115: \PP^1$, $f(\pi^{-1}(t))$ is a line called a \emph{line of the ruling} of

1116: $\Sigma$.  Denote by $F$ the divisor class of any $\pi^{-1}(t)$.  Then

1117: $\text{Pic}(\Sigma) = \ZZ\{D,F\}$ and the intersection pairing on

1118: $\Sigma$ is determined by $D.D = -1, D.F =1, F.F =0$.  The hyperplane

1119: class is $H=D+2F$ and the canonical class is $K = -2D-3F$.  

1120: 

1121: \ps

1122: 

1123: Using the fourth description of a cubic scroll, we see that any two

1124: cubic scrolls differ only by the choice of the isomorphism

1125: $H^0(\PP^4,\OO_{\PP^4}(1)) \rightarrow H^0(\PP E,\OO_{\PP E}(1))$.

1126: Therefore any two cubic scrolls are

1127: conjugate under the action of $\text{PGL}(5)$.  So the open set

1128: $U\subset \text{Hilb}_{P(t)}(\PP^4)$ parametrizing cubic scrolls is a

1129: homogeneous space for 

1130: $\text{PGL}(5)$, in particular it is smooth, connected and rational.

1131: So the Abel-Jacobi map $U\rightarrow J(X)$ associated to the family of

1132: intersections $\Sigma\cap X\subset X$ is a constant map.  

1133: 

1134: \subsection{Cubic Scrolls and Quartic Rational

1135: Curves}~\label{subsec-scr40}

1136: 

1137: Recall that $\text{Pic}(\Sigma) = \ZZ\{D,F\}$ where $D$ is the

1138: directrix and $F$ is the class of a line of ruling.  The intersection

1139: product is given by $D^2 = -1, D.F = 1, F^2 =0$.  The canonical

1140: class is given by $K_\Sigma = -2D-3F$ and the hyperplane class is

1141: given by $H=D+2F$.  The linear system $|F|$ is nef because it is the

1142: pullback of $\OO_{\PP^1}(1)$ under the projection

1143: $\pi:\Sigma\rightarrow \PP^1$.  Similarly, $|D+F|$ is nef because it

1144: contains all the conics obtained as the residuals to lines of the

1145: ruling in $|H|$.  Thus for any effective curve class $aD+bF$ we have

1146: the two inequalities $a=(aD+bF).F \geq 0$, $b=(aD+bF).(D+F) \geq 0$.  

1147: 

1148: \ps

1149: 

1150: Suppose that $C\subset \Sigma$ is an effective

1151: divisor of degree $4$ and arithmetic genus $0$.  By the adjunction

1152: formula 

1153: \begin{equation}

1154: K_\Sigma.[C] + [C].[C] = 2p_a - 2 = -2.

1155: \end{equation}

1156: So if $[C]=aD+bF$, then we have the conditions

1157: \begin{equation}

1158: a\geq 0, b\geq 0, a+b = 4, a^2 -2ab +a+2b = 2.

1159: \end{equation}

1160: It is easy to check that there are precisely two solutions $[C]=2D+2F,

1161: [C]=D+3F$.  We will see that both possibilities occur and describe

1162: some constructions related to each possibility.

1163: 

1164: \ps

1165: 

1166: \begin{lem}\label{lem-g1240}  Let $C\subset \PP^4$ be a smooth quartic 

1167: rational curve and let $V\subset |\OO_C(2)|$ be a pencil of degree

1168: $2$-divisors on $C$ without basepoints.  There exists a unique map of a

1169: Hirzebruch surface $\mathbf{F}_1$, $f:\Sigma\rightarrow \PP^4$ such

1170: that $(f^* H)^2=3$  

1171: and a factorization $i:C\rightarrow \Sigma$ of $C\rightarrow \PP^4$

1172: such that $i(C)\sim 2D+2F$ and the pencil of degree $2$ divisors

1173: $F\cap C$ is the pencil $V$.

1174: \end{lem}

1175: 

1176: \begin{proof}

1177: This is ~\cite[lemma 6.7]{HRS1}.  

1178: \end{proof}

1179: 

1180: \textbf{Remark}

1181: While we are at it, let's mention a specialization of the construction

1182: above, namely what happens if $V$ is not basepoint free.  Then

1183: $V=p+|\OO_C(1)|$ where $p\in C$ is some basepoint.  Consider

1184: the projection morphism $f:\PP^4\rightarrow \PP^3$ obtained by

1185: projection from $p$ (this is a rational map undefined at $p$).  The

1186: image of $C$ is a rational cubic curve $B$ (possibly a singular plane

1187: cubic).  Consider the cone $\Sigma'$ in $\PP^4$ over $B$ with vertex

1188: $p$.  This 

1189: surface contains $C$.  If we blowup $\PP^4$ at $p$, then the proper

1190: transform of $\Sigma'$ in $\widetilde{\PP^4}$ is a surface whose

1191: normalization $\Sigma$ is a 

1192: Hirzebruch surface $\mathbf{F}_3$ (normalization is only necessary if

1193: $B$ is a plane curve).  The directrix $D$ of $\Sigma$ is the pullback of

1194: the exceptional divisor of $\widetilde{\PP^4}$.  The inclusion

1195: $C\subset \Sigma'$ induces a factorization $i:C\rightarrow \Sigma$ of

1196: $C\rightarrow \PP^4$, and $[i(C)] = D + 4F$.  The intersection of $D$

1197: and $i(C)$ is precisely the point $p$.  And the linear system $i^*|F|$

1198: is exactly $|\OO_C(1)|$.  

1199: 

1200: \ps

1201: 

1202: Next we consider the case of a rational curve $C\subset \Sigma$ such

1203: that $[C]=D+3F$.  

1204: 

1205: \begin{lem}~\label{lem-scr40}  

1206: Let $C\subset \PP^4$ be a smooth quartic rational curve 

1207: and let $L\subset \PP^4$ be a line such that $L\cap C=Z$ is a degree 2

1208: divisor.  

1209: Let $\phi:C\rightarrow L$ be an isomorphism such that $\phi(Z)=Z$ and

1210: $\phi|_Z$ is the identity map.  Then there exists a unique triple

1211: $(h,i,j)$ where $h:\Sigma\rightarrow \PP^4$ is a finite map of a 

1212: Hirzebruch surface $\Sigma~\equiv~\mathbf{F}_1$, and 

1213: $i:C\rightarrow\PP^4$, $j:L\rightarrow\PP^4$ are factorizations of

1214: $C\rightarrow \PP^4,L\rightarrow \PP^4$

1215: such that $j(L)=D$ is the directrix, such that $[i(C)]=D+3F$ and such

1216: that the composition of $i:C\rightarrow \Sigma$ with the projection

1217: $\pi:\Sigma\rightarrow D$ equals $j\circ \phi$.  

1218: \end{lem}

1219: 

1220: \begin{proof}

1221: This is ~\cite[lemma 6.8]{HRS1}.

1222: \end{proof}

1223: 

1224: \subsection{Cubic Scrolls and Quintic Elliptics}~\label{subsec-scr51}

1225: 

1226: Recall that our fourth description of a cubic scroll was the image

1227: of a morphism $f:\Sigma \rightarrow \PP^4$ where $\Sigma$ is the

1228: Hirzebruch surface $\mathbb{F}_1$, $f^*\OO(1)\sim \OO_{\Sigma}(1) =

1229: \OO_{\PP E}(D+2F)$, and $f:\Sigma \rightarrow \PP^4$ is given by the

1230: complete linear series of $\OO_{\Sigma}(D+2F)$.  In the next 

1231: sections it

1232: will be useful to weaken this last condition.

1233: 

1234: \ps

1235: 

1236: \begin{defn} A \emph{cubic scroll} in $\PP^n$ is a finite morphism

1237: $f:\Sigma\rightarrow \PP^n$ where $\Sigma$ is isomorphic to the

1238: Hirzebruch surface $\mathbb{F}_1$ and such that $f^*\OO_{\PP^n}(1)$ is

1239: isomorphic to $\OO_{\Sigma}(D+2F)$. 

1240: \end{defn}

1241: 

1242: Let $H = D+2F$ denote the pullback of the hyperplane class.

1243: Now suppose that $E\subset \Sigma$ is an effective Cartier divisor

1244: with $p_a(E)=1$ and $E.H = 5$.  Since $F$ and $D+F$ are effective and

1245: move, we have $E.F, E.(D+F) \geq 0$.  Writing $E=aD + bF$ we see

1246: $(a,b)$ satisfies the relations $a,b \geq 0, a+b = 5$ and

1247: $a(b-3)+b(a-2) -a(a-2) = 0$.  These relations give the unique solution

1248: $E=2D+3F = -K$.  In particular, if $E$ is smooth then

1249: $\pi:E\rightarrow \PP^1$ is a finite morphism of degree 2, i.e. a

1250: $g^1_2$ on $E$.  Thus a pair $(f:\Sigma\rightarrow \PP^n, E\subset

1251: \Sigma)$ of a cubic scroll and a quintic elliptic determines a pair

1252: $(g:E\rightarrow \PP^n, \pi:E\rightarrow \PP^1)$ where $g:E\rightarrow

1253: \PP^n$ is a quintic elliptic and $\pi:E\rightarrow \PP^1$ is a degree

1254: 2 morphism.  

1255: 

1256: \ps

1257: 

1258: Suppose we start with a pair $(g:E\rightarrow \PP^n, \pi:E\rightarrow

1259: \PP^1)$ where $g:E\rightarrow \PP^n$ is an embedding of a quintic

1260: elliptic curve and $\pi:E\rightarrow \PP^1$ is a degree 2 morphism.

1261: Consider the rank 2 vector bundle $\pi_* g^* \OO_{\PP^n}(1)$.  

1262: 

1263: \ps

1264: 

1265: \begin{lem} Suppose $E$ is an elliptic curve and $\pi:E\rightarrow

1266: \PP^1$ is a degree 2 morphism.  Suppose $L$ is an invertible sheaf on

1267: $E$ of degree $d$.  Then we have

1268: 

1269: \begin{equation}

1270: \pi_* L \cong \lt\{ \begin{array}{ll}

1271:  \OO_{\PP^1}(e) \oplus \OO_{\PP^1}(e-1) & d = 2e+1, \\

1272:  \OO_{\PP^1}(e) \oplus \OO_{\PP^1}(e-2) & d=2e, L\cong

1273:         \pi^*\OO_{\PP^1}(e), \\

1274:  \OO_{\PP^1}(e-1) \oplus \OO_{\PP^1}(e-1) & d=2e, L\not\cong \pi^*

1275:         \OO_{\PP^1}(e) 

1276: \end{array} \rt.

1277: \end{equation}

1278: \end{lem}

1279: 

1280: \begin{proof}

1281: This is ~\cite[lemma 6.10]{HRS1}.

1282: \end{proof}

1283: 

1284: By the lemma we see that the vector bundle $G:=\pi_* g^*

1285: \OO_{\PP^n}(1)$ is isomorphic to $\OO_{\PP^1}(1)~\oplus~

1286: \OO_{\PP^1}(2)$.  Associated to the linear series

1287: $\OO_E^{n+1}\rightarrow g^*\OO_{\PP^n}(1)$ defining the embedding $g$,

1288: we have the push-forward linear series $\OO_{\PP^1}^{n+1}\rightarrow

1289: G$.  Since $g$ is an embedding, for each pair of points $\{p,q\}

1290: \subset E$ (possibly infinitely near), we have that

1291: $\OO_E^{n+1}\rightarrow g^*\OO_{\PP^n}(1)|_{\{p,q\}}$ is surjective.

1292: In particular taking $\{p,q\}=\pi^{-1}(t)$ for $t\in \PP^1$, we

1293: conclude that $\OO_{\PP^1}^{n+1}\rightarrow F|_t$ is surjective.  Thus

1294: we have an induced morphism $\PP G^\vee \rightarrow \PP^n$ which pulls

1295: back $\OO_{\PP^n}(1)$ to $\OO_{\PP G^\vee }(1)$.  Let us

1296: denote $\Sigma := \PP G^\vee $ and let us denote the morphism by

1297: $f:\Sigma\rightarrow \PP^n$.  Abstractly $\Sigma$ is isomorphic to

1298: $\mathbb{F}_1$ and $f:\Sigma\rightarrow \PP^n$ is a cubic scroll.

1299: 

1300: \ps

1301: 

1302: The tautological map $\pi^*\pi_* g^*\OO_{\PP^n}(1)\rightarrow

1303: g^*\OO_{\PP^n}(1)$ is clearly surjective.  Thus there is an induced

1304: morphism $h:E\rightarrow \Sigma$.  Chasing definitions, we see that

1305: $g=f\circ h$.  So we conclude that given a pair $(g:E\rightarrow

1306: \PP^n, \pi:E\rightarrow \PP^1)$ as above, we obtain a pair

1307: $(f:\Sigma\rightarrow \PP^n, h:E\rightarrow \Sigma)$.  Thus we

1308: have prove the following:

1309: 

1310: \begin{lem}\label{lem-equiv1}

1311: There is an equivalence between the collection of pairs

1312: $(f:\Sigma\rightarrow \PP^n,h:E\rightarrow \Sigma)$ with

1313: $f:\Sigma\rightarrow \PP^n$ a cubic scroll and $f\circ h:E\rightarrow

1314: \PP^n$ an embedded quintic elliptic curve and the collection of pairs

1315: $(g:E\rightarrow \PP^n, \pi:E\rightarrow \PP^1)$ where $g:E\rightarrow

1316: \PP^n$ is an embedded quintic elliptic curve and $\pi:E\rightarrow

1317: \PP^1$ is a degree 2 morphism.  

1318: \end{lem}

1319: 

1320: Stated more precisely, this gives an

1321: isomorphism of the parameter schemes of such pairs, but we don't need

1322: such a precise result.

1323: 

1324: 

1325: \subsection{Cubic Scrolls and Quintic Rational

1326: Curves}~\label{subsec-scr50}

1327: 

1328: If one carries out the analogous computations as at the beginning of

1329: subsection~\ref{subsec-scr40} one sees that the only effective divisor

1330: classes $aD+bF$ on a cubic scroll $\Sigma$ with degree $5$ and

1331: arithmetic genus $0$ are $D+4F$ and $3D+2F$.  But the divisor class

1332: $3D+2F$ cannot be the divisor of an irreducible curve because

1333: $(3D+2F).D = -1$.  Thus if $C\subset \Sigma$ is an irreducible curve

1334: of degree $5$ and arithmetic genus $0$, then $[C]=D+4F$.  

1335: 

1336: \begin{lem}~\label{lem-scr50}

1337: Let $C\subset \PP^4$ be a smooth quintic rational curve

1338: and let $L\subset \PP^4$ be a line such that $L\cap C$ is a degree 3

1339: divisor $Z$.  Let $\phi:C\rightarrow L$ be an isomorphism such that

1340: $\phi(Z)=Z$ and $\phi|_Z$ is the identity map.  Then there exists a

1341: unique triple $(h,i,j)$ such that $h:\Sigma\rightarrow\PP^4$ is a

1342: finite map of a Hirzebruch surface $\Sigma~\equiv~\PP^4$, and

1343: $i:C\rightarrow\PP^4, j:L\rightarrow\PP^4$ are factorizations of

1344: $C\rightarrow\PP^4, L\rightarrow\PP^4$ such that $j(L)=D$ is the

1345: directrix, such that $[i(C)]=D+4F$ and such that the composition of

1346: $i$ with the projection $\pi:\Sigma \rightarrow D$ equals $j\circ

1347: \phi$.  

1348: \end{lem}

1349: 

1350: \begin{proof}

1351: This is ~\cite[lemma 6.12]{HRS1}.

1352: \end{proof}

1353: 

1354: \section{Quartic Rational Curves}

1355: 

1356: In this section we will prove that the Abel-Jacobi map

1357: $u_{4,0}:\Hdg{4}{0}(X)\rightarrow J(X)$ is dominant and the general

1358: fiber is irreducible of dimension $3$.  In a later section we will

1359: prove that the general fiber is unirational.

1360: 

1361: \ps

1362: 

1363: The construction we use to understand quartic rational curves is as

1364: follows.  For any quartic rational curve

1365: $C\subset X$, define $A_C\subset F$ to be the scheme parametrizing

1366: $2$-secant lines to $C$.  By lemma~\ref{lem-enum}, $A_C$ is either

1367: positive-dimensional or else has length $16$.  Define $I\subset

1368: \Hdg{4}{0}(X)\times \HI{2}{F}$ to be the space of pairs $([C],[Z])$

1369: where $Z\subset A_C$ is a $0$-dimensional length 2 subscheme of $A_C$.  

1370: Denote by $\Lambda\subset Z\times \PP^4$ the flat family of lines

1371: determined by $Z\subset F$.  

1372: 

1373: \begin{defn} We say $Z$ is \emph{planar} if there exists a $2$-plane

1374: $P\subset \PP^4$ such that $\Lambda\subset Z\times P$.  We say $Z$ is

1375: \emph{nonplanar} if $Z$ is not planar.

1376: \end{defn}

1377: 

1378: By lemma 7.1, lemma 7.2 and theorem 7.3 of ~\cite{HRS1}, we have

1379: the following result

1380: 

1381: \begin{thm}~\label{thm-irr40} 

1382: The morphism $I\rightarrow \Hdg{4}{0}(X)$ is generically

1383: finite, $I$ is irreducible of dimension $8$ and therefore

1384: $\Hdg{4}{0}(X)$ is irreducible of dimension $8$.  For a general pair

1385: $([C],[Z])\in I$, $C$ is nondegenerate and $Z$ is reduced and nonplanar.

1386: \end{thm}

1387: 

1388: 

1389: Let $([C],[Z])\in I$ be a general pair.

1390: Now we may also consider $Z$ as a subscheme of

1391: $\text{Sym}^2(C)\cong\PP^2$.  Since $Z$ is non-planar, in particular

1392: the span of $Z\subset \PP^2$ yields a pencil of degree $2$ divisors on

1393: $C$ without basepoints.  By lemma~\ref{lem-g1240}, there is a cubic

1394: scroll $f:\Sigma\rightarrow\PP^4$ along with a factorization

1395: $i:C\rightarrow\PP^4$ such that the pencil of degree $2$ divisors on

1396: $C$ is just the pencil of intersections of $C$ with the lines of

1397: ruling of $\Sigma$.  The residual of $C\cup \Lambda$ in $\Sigma\cap X$

1398: is a curve in $\Sigma$ of degree $3$ and arithmetic genus $0$.  As a

1399: corollary of the proof of ~\cite[theorem 7.3]{HRS1}, for $([C],[Z])$

1400: a general pair, the residual curve $D$ is a twisted cubic curve.

1401: As in theorem~\ref{thm-st30}, let $U'\subset J(X)$ denote the

1402: Abel-Jacobi image of the locus of twisted cubics $D\subset X$ such

1403: that $\text{span}(D)\cap X$ is a smooth cubic surface.  And recall

1404: from theorem~\ref{sec-thmu} that the Abel-Jacobi map

1405: $\HI{2}{F}\rightarrow \Theta'$ is birational.  Define $R=\Theta'\times

1406: U'$ and define a rational transformation 

1407: \begin{equation}

1408: h:I\rightarrow R, \lt([C],[Z]\rt)\mapsto (u[Z],u([D])).

1409: \end{equation}

1410: 

1411: \begin{lem}  The rational transformation $h:I\rightarrow R$ is

1412: birational.

1413: \end{lem}

1414: 

1415: \begin{proof}  By Zariski's main theorem, it suffices to prove that

1416: for the general element $(x,y)\in R$, there is a unique pair

1417: $([C],[Z])$ such that $h([C],[Z])=(x,y)$.  Now $x=\{L_1,L_2\}$ is a

1418: general pair of disjoint lines in $X$.  And $y$ is a linear

1419: equivalence class $|D|$ of twisted cubics on a general hyperplane

1420: section $H\cap X$.  Then $L_i\cap H=\{p_i\}$ is a general point on $H\cap X$

1421: (since every point of $X$ lies on a line, we may assume this point is

1422: general on $H\cap X$).  And there is a unique twisted cubic $D\subset

1423: H\cap X$ in the linear equivalence class $|D|$ and which contains the

1424: two points $p_1$ and $p_2$.  If $h([C],[Z])=(x,y)$, then the residual

1425: to $C\cup L_1\cup L_2$ in $\Sigma$ can only be $D$.  

1426: 

1427: \ps

1428: 

1429: Now consider the hyperplane $H'=\text{span}(L_1,L_2)$.  The

1430: intersection $H'\cap D$ consists of three points $p_1,p_2$ and a third

1431: point $q$.  Moreover the directrix $M$ of $\Sigma$ is contained in

1432: $H$.  And by a divisor class calculation, $M\cap D$ consists of a

1433: point other than $p_1, p_2$.  The only possibility is that $M\cap D =

1434: \{q\}$.  Notice that given two skew lines $L_1, L_2$ in a $3$-plane

1435: $H'$ and given a point $q\in H'$ not lying on $L_1\cup L_2$, there is

1436: a unique line $M\subset H'$ which contains $q$ and intersects each of

1437: $L_1,L_2$.  Indeed, if we project from $q$ then $L_1,L_2$ project to

1438: distinct lines in $H/q\cong \PP^2$ which intersect in a unique point.

1439: And $M$ is simply the cone over this point.  So we conclude that the

1440: directrix line $M$ is uniquely determined by $(x,y)$.  

1441: 

1442: \

1443: 

1444: Finally, projection $\Sigma\rightarrow M$ to the directrix determines

1445: an isomorphism 

1446: $\phi:D\rightarrow M$ such that for each $r\in D,

1447: \text{span}(r,\phi(r))$ is a line of the ruling of $\Sigma$, in

1448: particular $\phi(q)=q$.  Conversely, given a line $M$ which intersects

1449: $C$ in one point $q$ and given an isomorphism $\phi:C\rightarrow M$

1450: such that $\phi(q)=q$, then the union of the lines

1451: $\text{span}(r,\phi(r))$ is a cubic scroll $\Sigma$.  Thus to uniquely

1452: specify the cubic scroll $\Sigma$, we have only to determine $\phi$.

1453: But notice also that we already know $\phi(p_1),\phi(p_2)$ are the

1454: unique points of intersection of $M$ with $L_1$ and $L_2$

1455: respectively.  Thus $\phi$ is also uniquely determined by $(x,y)$.

1456: Altogether we conclude that $\Sigma$ is uniquely determined by

1457: $(x,y)$.  But then we can recover/construct $C$ as the residual to

1458: $D\cup L_1\cup L_2$ in $\Sigma\cap X$.  This proves that $h$ is

1459: birational.

1460: \end{proof}

1461: 

1462: \ps

1463: 

1464: \begin{thm}\label{thm-irrfib} 

1465: The composite of $I\rightarrow

1466: \Hdg{4}{0}(X)$ with the Abel-Jacobi map

1467: $u_{4,0}:\Hdg{4}{0}(X)\rightarrow J(X)$ is dominant and the

1468: general fiber is irreducible.  Therefore $u_{4,0}$ is dominant and the 

1469: general fiber is irreducible.

1470: \end{thm}

1471: 

1472: \begin{proof}

1473: By the residuation trick, the composite equals (as a rational map) the

1474: pointwise inverse of the composition

1475: \begin{equation}

1476: I\xrightarrow{h} R = \Theta'\times U' \rightarrow J(X)

1477: \end{equation}

1478: where the second map is the restriction to $\Theta'\times U'\subset

1479: J(X)\times J(X)$ of the addition map.  Clearly $\Theta'+U'= J(X)$ since

1480: $\Theta'$ is a divisor. $U'\subset \Theta$ is a Zariski-dense open

1481: set, and $\Theta$ is not contained in any translate of

1482: $\Theta'$.  Thus $I\rightarrow J(X)$ is dominant.  

1483: 

1484: \ps

1485: 

1486: Using the fact that $\Theta$ is a symmetric divisor, we see that the

1487: general fiber of the map $\Theta'\times\Theta \rightarrow J(X)$ is an

1488: intersection $\Theta\cap (a+\Theta')$.  By theorem~\ref{sec-thmu}, for 

1489: general $a$ this intersection is irreducible.  Thus we conclude that

1490: the general fiber of $I\rightarrow J(X)$ is irreducible.

1491: This proves the theorem.

1492: \end{proof}

1493: 

1494: \section{Quintic Elliptics}~\label{subsec-irr51}

1495: 

1496: In this section we will prove that the Abel-Jacobi map

1497: $u_{5,1}:\Hdg{5}{1}(X)\rightarrow J(X)$ is dominant and the general

1498: fiber is an irreducible $5$-fold.  In the next section we will see

1499: that the fibers are unirational.

1500: 

1501: \ps

1502: 

1503: The construction we use to understand quintic elliptics is as

1504: follows.  Define $g:\widetilde{H}\rightarrow \Hdg{5}{1}(X)$ to be the

1505: relative $\text{Pic}^2$ of the universal family of elliptic curves

1506: $\mathcal{C}\rightarrow \Hdg{5}{1}(X)$.  By lemma~\ref{lem-equiv1}

1507: $\widetilde{H}$ is also the parameter space for pairs $(f,h)$ where

1508: $f:\Sigma\rightarrow \PP^4$ is a generalized cubic scroll and

1509: $h:C\rightarrow \Sigma$ is a curve such that $f(h(C))\subset X$ is a

1510: smooth quintic elliptic.  The residual to $C$ in $\Sigma\cap X$ is a

1511: curve $C'$ of degree $4$ and arithemtic genus $0$.  

1512: 

1513: \begin{thm}~\label{thm-irr51}  

1514: The scheme $\widetilde{H}$ is irreducible of dimension

1515: $11$.  For a general pair $(f,h)\in \widetilde{H}$, the residual curve

1516: $D$ is a smooth, nondegenerate, quartic rational curve.  

1517: \end{thm}

1518: 

1519: \begin{proof}

1520: This follows from the proof of ~\cite[theorem 8.1]{HRS1}.

1521: \end{proof}

1522: 

1523: We have an induced rational transformation $g':\widetilde{H}\rightarrow

1524: \Hdg{4}{0}(X)$ which sends a pair $(f,h)$ to the residual curve $C'$.

1525: Also by lemma~\ref{lem-scr40}, we see that $g':\widetilde{H}\rightarrow

1526: \Hdg{4}{0}(X)$ is (birationally) the parameter space for pairs

1527: $(f',h')$ where $f':\Sigma\rightarrow \PP^4$ is a cubic scroll and

1528: $h':C'\rightarrow \Sigma$ is a curve which intersects the directrix in

1529: a degree $2$ divisor and lines of the ruling in a degree $1$ divisor,

1530: and such that $f'(h'(C'))\subset X$ is a quartic rational curve.  Given

1531: such a pair, the residual to $C'$ in $\Sigma\cap X$ is the quintic

1532: elliptic $C$ we started with, so both descriptions of $\widetilde{H}$

1533: are equivalent.

1534: 

1535: \ps

1536: 

1537: Notice that by lemma~\ref{lem-scr40}, the fiber of $g'$ over a general

1538: point $[C']$ is the irreducible (rational) variety parametrizing pairs

1539: $(D,\phi)$ 

1540: where $D$ is a $2$-secant line to $C'$ and where $\phi:C'\rightarrow

1541: D$ is an isomorphism such that $\phi$ is the identity on $C'\cap D$.  

1542: 

1543: \begin{thm}~\label{thm-irrfib2} The composite of

1544: $g:\widetilde{H}\rightarrow \Hdg{5}{1}(X)$ and the 

1545: Abel-Jacobi map

1546: $u_{5,1}:\Hdg{5}{1}(X)\rightarrow J(X)$ is dominant and the general

1547: fiber is an irreducible $6$-fold.  Therefore the Abel-Jacobi map

1548: $u_{5,1}:\Hdg{5}{1}(X)\rightarrow J(X)$ is dominant and the general

1549: fiber is an irreducible $5$-fold.

1550: \end{thm}

1551: 

1552: \begin{proof}

1553: The space of all embedded cubic scrolls $f:\Sigma \rightarrow

1554: \PP^4$ is clearly unirational, 

1555: in fact it is a homogeneous space for $\text{PGL}(5)$ and so it is even

1556: rational.  So the Abel-Jacobi map is constant on the family of

1557: complete intersections $f^*X$.  

1558: By the residuation trick, the following two birational transformations

1559: are pointwise (additive) inverses:

1560: \begin{eqnarray}

1561: \widetilde{H}\xrightarrow{g} \Hdg{5}{1}(X) \xrightarrow{u_{5,1}} J(X) \\

1562: \widetilde{H}\xrightarrow{g'} \Hdg{4}{0}(X) \xrightarrow{u_{4,0}}

1563: J(X). \nonumber

1564: \end{eqnarray}

1565: By theorem~\ref{thm-irrfib}, $u_{4,0}$ is dominant and the general

1566: fiber is irreducible.  We have seen that $\widetilde{H}\rightarrow

1567: \Hdg{4}{0}(X)$ is dominant and the general fiber is irreducible.  Thus

1568: one, and hence both 

1569: of the morphisms $\widetilde{H}_1\rightarrow J(X)$ are dominant and the

1570: general fiber is irreducible.  Since $\widetilde{H}\rightarrow

1571: \Hdg{5}{1}(X)$ is dominant, we conclude that $u_{5,1}$ is dominant and

1572: the general fiber is irreducible.  

1573: \end{proof}

1574: 

1575: \section{Double Residuation and Unirationality of the Fibers}

1576: 

1577: In the last section we introduced the space $\widetilde{H}$ which

1578: parametrizes pairs $(f:\Sigma\rightarrow \PP^4,

1579: h:E\rightarrow \Sigma)$ where $f:\Sigma\rightarrow \PP^4$ is a cubic

1580: scroll and $E\subset \Sigma$ is an elliptic curve such that $f\circ

1581: h:E\rightarrow \PP^4$ is an embedding of $E$ as a quintic elliptic

1582: curve in $X$.  Equivalently $\widetilde{H}$ parametrizes pairs

1583: $(f:\Sigma\rightarrow \PP^4, k:C\rightarrow \Sigma)$ where $C$ is the

1584: rational curve residual to $E$ in $f^*X$ and $f\circ k:C\rightarrow X$

1585: is a quartic rational curve.  In this section we will use

1586: $\widetilde{H}$ to prove that the fibers of $u_{4,0}$ and $u_{5,1}$

1587: are unirational.

1588: 

1589: \ps

1590: 

1591: We need a partial compactification of $\widetilde{H}$.  Let

1592: $\overline{H}^{5,1} \subset \text{Hilb}_{5t}(X)$ denote the open

1593: subset of the closure of

1594: $\Hdg{5}{1}(X)$ which parametrizes Cohen-Macaulay curves (i.e. curves with

1595: no embedded points).  Let $P(t)$ be the numerical polynomial $P(t) = 

1596: \frac{3}{2}t^2 + \frac{5}{2}t + 1$.  Let $M~\subset~

1597: \text{Hilb}_{P(t)}(\PP^4)$ denote the open subscheme parametrizing

1598: connected, reduced, local complete intersection subschemes of $\PP^4$

1599: with Hilbert polynomial $P(t)$.  The space $M$ contains an open subset

1600: parametrizing embedded cubic scrolls, but $M$ also parametrizes mild

1601: degenerations of embedded cubic scrolls.  We will refer to the schemes

1602: parametrized by $M$ as \emph{generalized cubic scrolls}.  Let

1603: $N\subset M\times \overline{H}^{5,1}$ denote the locally

1604: closed subscheme parametrizing pairs $(\Sigma,E)$ such that $X\cap

1605: \Sigma$ is a reduced Weil divisor and $E\subset X\cap\Sigma$.   

1606: Notice that the Lefschetz hyperplane

1607: theorem shows that $X$ contains no generalized cubic scrolls.  Thus

1608: $X\cap \Sigma$ is a Cartier divisor on $\Sigma$ which contains $E$.

1609: By ~\cite[corollary 2.7]{HRS1} the family of

1610: residual curves to $E\subset X\cap \Sigma$ is a flat family of

1611: Cohen-Macaulay curves.

1612: The residual divisor $C\subset \Sigma$ is a connected, arithmetic

1613: genus 0 curve of degree 4 contained in $X$, i.e. a point of

1614: $\overline{H}^{4,0}$.  So we have 2 projection morphisms

1615: $p_1:N\rightarrow \overline{H}^{5,1}$ and

1616: $p_2:N\rightarrow \overline{H}^{4,0}$.  

1617: 

1618: \ps

1619: 

1620: One thing to notice is that if $E$ is a quintic elliptic curve whose

1621: span is all of $\PP^4$, then $E$ is not contained in any hyperplanes,

1622: any quadric surfaces, or in any cone over a twisted cubic.  Thus for

1623: every point $(\Sigma,E)\in N$, we have that $\Sigma$ is a

1624: smooth cubic scroll.  

1625: 

1626: \ps

1627: 

1628: Now define $N^2$ to be the fiber product

1629: $N\times_{\overline{H}^{5,1}}N$.  We define the two maps $i=1,2$, 

1630: $p_{2,i}:N^2 \rightarrow \overline{H}^{4,0}$ to be the

1631: compositions of the projection $p_i:N^2\rightarrow N$ with

1632: $p_2:N\rightarrow \overline{H}^{4,0}$.  

1633: 

1634: \ps

1635: 

1636: \begin{lem}~\label{lem-urfib}  For a general quartic rational curve

1637: $[C]\in \Hdg{4}{0}(X)$, 

1638: the fiber of $p_{2,i}:N^2\rightarrow \overline{H}^{4,0}$ is

1639: unirational of dimension 4.

1640: \end{lem}

1641: 

1642: \begin{proof}

1643: Let $C\subset X$ be a general quartic rational curve.  Let $T\subset

1644: \PP^4$ be the threefold swept out by all 2-secant lines and tangent

1645: lines to $C$.  Then $T\cap X$ is a surface.  Let $U\subset X$ denote

1646: the complement of $T$.  We construct a rational transformation $\rho:C

1647: \times U \rightarrow N^2$ as follows.  For each point $p\in C$ and

1648: $q\in U$, there is a line $L=\text{span}(p,q)$.  The quotient

1649: projective space $\PP^4/L$ is isomorphic to $\PP^2$.  Since $L\cap C =

1650: \{p\}$, we see that the image of $C$ under the rational projection

1651: $\PP^4 \rightarrow \PP^2$ is a singular cubic curve.  Let $V\subset

1652: C\times U$ denote the open set such that this singular cubic curve is

1653: a nodal cubic.  Then the node corresponds to the unique 2-secant line

1654: $D\subset \PP^4$ to $C$ which intersects $L$.  And there is a unique

1655: isomorphism $\pi:C\rightarrow D$ such that $\pi(r)=r$ for each $r\in

1656: C\cap D$ and such that $\pi(p) = s$ where $D\cap L = \{s\}$.  By

1657: subsection~\ref{subsec-scr40} the 4-tuple $(C,D,\pi,C\cap D)$

1658: determines a cubic scroll 

1659: $f:\Sigma\rightarrow \PP^4$ which contains $C$ and $D$.  Let $E$

1660: denote the residual curve to $C$ in $f^*X$.  Then $(\Sigma,E)$ is a

1661: point of $\widetilde{H}$.  Also we know that $q$ is on $E$.  The

1662: linear system $|2q|$ on $E$ is a $g^1_2$, i.e. a degree 2 morphism

1663: $\pi:E\rightarrow \PP^1$.  By lemma\ref{lem-equiv1}

1664: $\pi:E~\rightarrow \PP^1$ determines

1665: a second cubic scroll 

1666: $\Sigma'$ containing $E$ such that $|2q|$ is just the linear system of 

1667: intersections of $E$ with the lines of ruling of $\Sigma'$.  For

1668: $(p,q)\in V$ we define $\rho(p,q) = ( 

1669: (\Sigma,E), (\Sigma',E))$.

1670: 

1671: \ps

1672: 

1673: Since every $g^1_2$ on $E$ can be expressed as $|2q|$ for some $q\in

1674: E$, it is clear that we can obtain every cubic scroll $\Sigma'$

1675: containing $E$ simply by varying the point $q$.  Thus we conclude that

1676: the map $\rho:V\rightarrow N^2$ dominates the fiber of

1677: $p_{2,i}:N^2 \rightarrow \overline{H}^{4,0}$ over $[C]$.  Since

1678: $V$ is an open subset of the product of unirational varieties $C\times

1679: X$, we conclude that $V$ is unirational.  Since the image of a

1680: unirational variety is unirational, we conclude that the fiber of

1681: $p_{2,i}:N^2 \rightarrow \overline{H}^{4,0}$ over $[C]$ is unirational.

1682: \end{proof}

1683: 

1684: Let $P\subset N^2$ denote the irreducible component whose general

1685: member is a point $(E,\Sigma_1,\Sigma_2)$ where $E$ is a smooth quintic

1686: elliptic and $\Sigma_1,\Sigma_2$ are cubic scrolls.

1687: Now consider the morphism 

1688: \begin{equation}

1689: r:P\rightarrow

1690: \overline{H}^{4,0}\times \overline{H}^{4,0},

1691: r(E,\Sigma_1,\Sigma_2) = (p_2(E,\Sigma_1),p_2(E,\Sigma_2))

1692: \end{equation}

1693: i.e. the pair of residual quartic curves to $E$ in $\Sigma_1\cap X$

1694: and $\Sigma_2 \cap X$.

1695: 

1696: \begin{thm}~\label{thm-1dfib}

1697:  For a general pair $(C_1,C_2)$ of smooth quartic rational

1698: curves in the image of $r$, the fiber is 1-dimensional.  And the

1699: general fiber of the Abel-Jacobi map 

1700: \begin{equation} 

1701: u_{4,0}:\Hdg{4}{0}(X)\rightarrow J(X)

1702: \end{equation}

1703: it one of the irreducible, unirational 3-folds

1704: $p_{2,2}(p_{2,1}^{-1}([C]))$ for some $[C]\in\Hdg{4}{0}(X)$.

1705: \end{thm}

1706: 

1707: \begin{proof}

1708: 

1709: Let $\Pi\subset \PP^4$ be a hyperplane such that $Y=\Pi\cap X$ is a

1710: smooth cubic surface.  For a general point $p\in Y$ we can

1711: find a quartic elliptic curve $B\subset Y$ with $p\in B$: in the

1712: model of $Y$ as the blow-up of $\PP^2$ at 6 points, $B$ is

1713: the proper transform of a plane cubic passing through 5 of those

1714: points and the additional point $p$.  Also for a general point

1715: $p\in Y$, we can find a line of type I, $L\subset X$, such that

1716: $L\cap \Pi =\{p\}$.  Define $E=B\cup L$, so $E$ is a connected, nodal

1717: curve of arithmetic genus 1 and degree 5.  

1718: 

1719: \begin{Claim}\label{claim-a1}

1720: The curve $E$ satisfies the conditions in

1721: ~\cite[lemma 2.3]{HRS1},

1722: i.e. 

1723: \begin{equation}

1724: H^1(L,N_{L/X}(-p))=H^1(B,N_{B/X})=0.

1725: \end{equation}  

1726: Therefore $\HI{5t}{X}$ is

1727: smooth at $[E]$ and deformations of $E$ smooth the node at $p$.

1728: \end{Claim}

1729: 

1730: We have the following short exact sequences of coherent sheaves:

1731: \begin{equation}

1732: \begin{CD}

1733: 0 @>>> N_{B/X} @>>> N_{E/X}|_{B} @>>> \OO_p @>>> 0 \\

1734: 0 @>>> N_{L/X} @>>> N_{E/X}|_L @>>> \OO_p @>>> 0 \end{CD}

1735: \end{equation}

1736: We also have a short exact sequence:

1737: \begin{equation}

1738: \begin{CD}

1739: 0 @>>> N_{B/Y} @>>> N_{B/X} @>>> N_{Y/X}|_{B} @>>> 0 \end{CD}

1740: \end{equation}

1741: The self-intersection of a quartic rational curve on a smooth cubic

1742: scroll is $4$.  And $N_{Y/X}|_{B} = \OO_{\Pi}(1)|_E$ is also degree

1743: 4.  So by Riemann-Roch we conclude that

1744: $H^1(B,N_{B/Y})=H^1(B,N_{Y/X}|_{B}) = 0$.  Applying the long exact

1745: sequence in cohomology to our last short exact sequence, we conclude

1746: that $H^1(B,N_{B/X}) = 0$.  Applying the long exact sequence in

1747: cohomology to our first short exact sequence above, we conclude that

1748: $H^1(B,N_{E/X}|_{B}) = 0$.  

1749: 

1750: \ps

1751: 

1752: Twisting the second exact sequence above by $\OO_L(-p)$ yields an

1753: exact sequence:

1754: \begin{equation}

1755: \begin{CD}

1756: 0 @>>> N_{L/X}(-p) @>>> N_{E/X}|_L(-p) @>>> \OO_p @>>> 0 \end{CD}

1757: \end{equation}

1758: By assumption $N_{L/X} \cong \OO_L \oplus \OO_L$, thus $N_{L/X}(-p)

1759: \cong \OO_L(-1)\oplus \OO_L(-1)$.  In particular, $H^1(L,N_{L/X}(-p))

1760: = 0$.  Applying the long exact sequence in cohomology to this short

1761: exact sequence, we conclude that $H^1(L,N_{E/X}(-p)) = 0$.  Finally,

1762: we have the short exact sequence:

1763: \begin{equation}

1764: \begin{CD}

1765: 0 @>>> N_{E/X}|_L(-p) @>>> N_{E/X} @>>> N_{E/X}|_{B} @>>> 0 \end{CD}

1766: \end{equation}

1767: Applying the long exact sequence in cohomology to this short exact

1768: sequence, we conclude that $H^1(E,N_{E/X}) = 0$.  Thus $[E]\in

1769: \HI{5t}{X}$ is unobstructed.  This finishes the proof of

1770: claim~\ref{claim-a1}. 

1771: 

1772: \ps

1773: 

1774: For a general line $M\subset \Pi$ containing $p$, the residual to $L$

1775: in $\text{span}(L,M)\cap X$ is a smooth conic.  Let $M_1,M_2$ be two

1776: such lines.

1777: Without loss of

1778: generality, we may also suppose that $M_1, M_2$ are 2-secant lines to

1779: $B$: since we are free to choose $B$ a general quartic elliptic, we may

1780: first choose $M_1, M_2$, and then choose $B$ to pass through $p$ and one of

1781: the other two points of $M_i\cap Y$ for each of $i=1,2$.  Let $M_i\cap B

1782: = \{p, q_i\}$ and let 

1783: $r_i$ denote the third point of $M_i\cap Y$.  

1784: 

1785: \ps

1786: 

1787: Define $S_i' =\text{span}(L,M_i)$ and let $S_1'', S_2''\subset \Pi$ be

1788: smooth quadric surfaces containing $B\cup M_1$ and $B\cup M_2$

1789: respectively.  Define 

1790: $S_1=S_1'\cup S_1'', S_2=S_2'\cup S_2''$, thus $S_1, S_2$ are each a union

1791: of a 2-plane and a smooth quadric surface.  

1792: By ~\cite[lemma 6.2]{HRS1}

1793: such a surface is a specialization of a cubic

1794: scroll, thus $[S_1],[S_2]\in \HI{P(t)}{\PP^4}$.  

1795: 

1796: \begin{Claim}~\label{claim-a2} For

1797: $i=1,2$, we have $H^1(S_i,I_{E/S_i}N_{S_i/\PP^4})=

1798: H^2(S_i,I_{E/S_i}N_{S_i/\PP^4})=0$ where $I_{E/S_i}$ is the ideal

1799: sheaf of $E\subset S_i$.  

1800: \end{Claim}

1801: 

1802: By the deformation theory argument in the proof of ~\cite[lemma

1803: 6.2]{HRS1}, 

1804: this vanishing result implies that the

1805: morphism $N^2\rightarrow \overline{H}^{5,1}$ is smooth at

1806: $(E,S_1,S_2)$. 

1807: 

1808: \ps

1809: 

1810: We have a short exact sequence:

1811: \begin{equation}

1812: \begin{CD}

1813: 0 @>>> N_{S_i/\PP^4}|_{S_i'}(-M_i-L) @>>> I_{E/S_i}N_{S_i/\PP^4} @>>>

1814: N_{S_i/\PP^4}|_{S_i''}(-B) @>>> 0 \end{CD}

1815: \end{equation}

1816: And we have the two short exact sequences:

1817: \begin{eqnarray}

1818: \begin{CD}

1819: 0 @>>> N_{S_i'/\PP^4}(-M_i-L) @>>> N_{S_i/\PP^4}|_{S'}(-M_i-L) @>>>

1820: N_{M_i/S_i''}(-p) @>>> 0 \end{CD} \\

1821: \begin{CD}

1822: 0 @>>> N_{S_i''/\PP^4}(-B) @>>> N_{S_i/\PP^4}|_{S_i''}(-B) @>>>

1823: N_{M_i/S'}\otimes N_{M_i/S_i''}(-p-q) @>>> 0 \end{CD}

1824: \end{eqnarray}

1825: From the proof of lemma~\cite[lemma 6.3]{HRS1}, we know that 

1826: \begin{equation}

1827: N_{S_i'/\PP^4}

1828: \cong \OO_{\PP^2}(1) \oplus \OO_{\PP^2}(1), N_{S_i''/\PP^4} \cong

1829: \OO_{\PP^1\times \PP^1}(1,1) \oplus \OO_{\PP^1\times\PP^1}(2,2).

1830: \end{equation}

1831: Thus we have $N_{S'/\PP^4}(-M_i-L)\cong \OO_{\PP^2}(-1)\oplus

1832: \OO_{\PP^2}(-1)$ and $N_{S_i''/\PP^4}(-B)\cong

1833: \OO_{\PP^1\times\PP^1}(-1,-1) \oplus \OO_{\PP^1\times\PP^1}$.

1834: Moreover $N_{M_i/S'}\cong \OO_{\PP^1}(1)$ and $N_{M_i/S_i''}\cong

1835: \OO_{\PP^1}$.  Thus our two exact sequences are:

1836: \begin{equation}

1837: \begin{CD}

1838: 0 @>>> \OO_{\PP^2}(-1) \oplus \OO_{\PP^2}(-1) @>>>

1839: N_{S_i/\PP^4}|_{S_i'}(-M_i-L) @>>> \OO_{\PP^1}(-1) @>>> 0 \\

1840: 0 @>>> \OO_{\PP^1\times\PP^1}(-1,-1) \oplus \OO_{\PP^1\times\PP^1}

1841: @>>> N_{S_i/\PP^4}|_{S_i''}(-B) @>>> \OO_{\PP^1}(-1) @>>> 0 \end{CD}

1842: \end{equation}

1843: It quickly follows from the long exact sequence in cohomology that

1844: \begin{equation}

1845: H^{j>0}(S',N_{S_i/\PP^4}|_{S'}(-M_i-L)) =

1846: H^{j>0}(S_i'',N_{S_i/\PP^4}|_{S_i''}(-B)) = 0

1847: \end{equation}

1848: Applying the long exact sequence in cohomology to our first short

1849: exact sequence, we conclude that $H^{j>0}(S_i,I_{E/S_i}N_{S_i/\PP^4})

1850: = 0$.  This proves claim~\ref{claim-a2}

1851: 

1852: \ps

1853: 

1854: We conclude that $N^2\rightarrow \overline{H}^{5,1}$ is

1855: smooth at $(E,S_1,S_2)$.  Since $[E]$ is a smooth point of

1856: $\text{Hilb}_{5t}(X)$ which lies in $\overline{H}^{5,1}$, we conclude

1857: that $N^2$ is smooth at $(E,S_1,S_2)$ and the irreducible

1858: component of $N^2$ which contains $(E,S_1,S_2)$ dominates

1859: $\overline{H}^{5,1}$, i.e. the irreducible component is $P$.  

1860: 

1861: \ps

1862: 

1863: \begin{Claim}~\label{claim-a3}

1864: The fiber of $r:P\rightarrow

1865: \overline{H}^{4,0}\times \overline{H}^{4,0}$ containing $(E,S_1,S_2)$

1866: is one-dimensional at $(E,S_1,S_2)$.

1867: \end{Claim}

1868:  

1869: Of course the residual to $L$ in

1870: $S_i'\cap X$ is a smooth conic $D_i'$ which intersects $M_i$ in

1871: $\{q_i,r_i\}$.  And 

1872: the residual to $B$ in $S_i''\cap X$ is a smooth conic $D_i''$ which

1873: contains $r_i$.  Let us define $D_i=D_i'\cup D_i''$.  Then

1874: $r(E,S_1,S_2) = 

1875: (D_1,D_2)$.  

1876: 

1877: \ps

1878: 

1879: Now we want to determine the dimension of every irreducible component

1880: of the fiber $\Phi=r^{-1}(D_1,D_2)$ through $(E,S_1,S_2)$.  By

1881: lemma~\cite[lemma 6.3]{HRS1}, we may restrict our attention to the open

1882: subset of $\Phi$ parametrizing pairs $(C,R_1,R_2)$ such that each of

1883: $R_1,R_2$ is in $T\cup U$, i.e. $R_i$ is either a cubic scroll or

1884: the union of a 2-plane and a quadric surface along a line.  

1885: But for any conic $G$ in a cubic scroll $\Sigma$ with directrix

1886: $D$ and fiber $F$, we know that $G\sim D+F$.  Given a union of 2 such

1887: conics, $G',G''$, we see that $3H-G'-G'' \sim D+4F$.  Thus the

1888: residual to such a curve in $\Sigma\cap X$ would be a quintic curve of

1889: arithmetic genus 0, not arithmetic genus 1.  We conclude that for

1890: $(C,R_1,R_2)$ in $\Phi$, we must have $R_i$ is in $T$.  

1891: 

1892: \ps

1893: 

1894: For

1895: an open neighborhood of $(E,S_1,S_2)$ in $\Phi$, we must have that

1896: $D_i'$ lies in the irreducible component of $R_i$ which is a 2-plane.

1897: Now suppose given such a $(C,R_1,R_2)$.  There is a unique 2-plane

1898: $S_i'$ which contains $D_i'$.  Thus $R_i = S_i'\cup R_i''$ for some

1899: smooth quadric 

1900: surface $R_i''$.  And $C=L\cup A$ for some quartic elliptic $A$.  Now

1901: $\text{span}(A) $ contains both $\text{span}(D_1'')$ and

1902: $\text{span}(D_2'')$.  Thus $\text{span}(A)=\Pi$.  Finally, since

1903: $S_i'\cap \Pi = M_i$, we can only have $S_i'\cap R_i'' = M_i$.  In

1904: particular, we conclude that $M_i$ intersects $A$ in the points

1905: $p,q_i$.  Thus $A\subset Y=\Pi\cap X$ is a quartic elliptic curve

1906: passing through $p,q_1$ and $q_2$.  These points impose independent

1907: conditions on the 4-dimensional linear system of quartic elliptic

1908: curves residual to $D_i''$.  Thus there is a 1-dimensional linear

1909: system of $A$'s.  

1910: 

1911: \ps

1912: 

1913: For each $A$ in this pencil of quartic elliptic curves, we have $A\cup

1914: D_i''\subset Y$ lies in $|\OO_Y(2)|$.  Since $Y$ is linearly normal

1915: and lies in no quadric surfaces, we have $|\OO_Y(2)|=|\OO_{\Pi}(2)|$.

1916: Thus there is a unique quadric surface $R_i''$ containing $A\cup

1917: D_i''$.  Since $\{p,q_i,r_i\}\subset M_i\cap R_i''$, we conclude that

1918: $M_i\subset R_i''$.  Thus $R_i=S_i'\cup R_i''$ is a surface in $T$,

1919: and $(A,R_1,R_2)$ is a point in the fiber $\Phi$.  So the fiber of

1920: $\Phi$ is one-dimensional at $(E,S_1,S_2)$, which finishes the proof

1921: of claim~\ref{claim-a3}.  

1922: 

1923: \ps

1924: 

1925: By the above, $\Phi$ is a 1-dimensional irreducible variety in a

1926: neighborhood of $(E,S_1,S_2)$.  It follows by upper semicontinuity of

1927: the fiber dimension that the general fiber of $r:\PP

1928: \rightarrow \overline{H}^{4,0}(X)\times \overline{H}^{4,0}(X)$ is at most

1929: 1-dimensional.  

1930: 

1931: \ps

1932: 

1933: For a general rational quartic $[C]\in\Hdg{4}{0}(X)$, it follows by

1934: lemma~\ref{lem-urfib} that $p_{2,1}^{-1}([C])\subset N^2$ is a

1935: unirational 4-fold.  So the subvariety

1936: $Z_C:=p_{2,2}(p_{2,1}^{-1}([C])$ is an irreducible, unirational

1937: variety.  Also observe that the subvarieties $Z_C$ sweep out

1938: $\Hdg{4}{0}(X)$: given a curve $[D]\in\Hdg{4}{0}(X)$, if we choose any 

1939: curve $[C]\in Z_D$, then also $[D]$ is contained in $Z_C$.  

1940: 

1941: \ps

1942: 

1943: Since the fiber dimension of $p_{2,i}:N^2\rightarrow \Hdg{4,0}(X)$ is

1944: at most 1, we conclude the dimension of $Z_C$ is at least $4-1=3$.

1945: Since $J(X)$ 

1946: contains no unirational subvarieties, $Z_C$ is

1947: contained in a fiber of the Abel-Jacobi map

1948: $u_{4,0}:\Hdg{4,0}(X)\rightarrow J(X)$.  By theorem~\ref{thm-irrfib},

1949: the general fiber of $u_{4,0}$ is irreducible of dimension $3$.

1950: Combining this with the fact that a general point of

1951: $\Hdg{4}{0}(X)$ is contained in an irreducible 3-fold $Z_C$, we

1952: conclude that the general fiber of $u_{4,0}$ is one of the general

1953: subvarieties $Z_C$, and vice versa.  As a consequence, observe that

1954: the general fiber dimension of $u_{4,0}$ is 1 (and not 0).  This

1955: completes the proof of the theorem.

1956: \end{proof}

1957: 

1958: \ps

1959: 

1960: \begin{cor}~\label{cor-51fib}  The general fiber of the Abel-Jacobi

1961: map

1962: $u_{5,1}:\Hdg{5}{1}(X)\rightarrow J(X)$ is an irreducible, unirational

1963: 5-fold.

1964: \end{cor}

1965: 

1966: \begin{proof} Recall from the proof of lemma~\ref{lem-urfib} that not

1967: only did we show that $p_{2,i}^{-1}([C])$ is unirational, but we showed

1968: that there is a dominant, generically-finite morphism $B\rightarrow

1969: p_{2,i}^{-1}([C])$ such that $B$ is unirational and on $B$ we can 

1970:  produce a

1971: section $\sigma$ of the family of elliptic curves $E$ -- in fact we used

1972: $|2\sigma|$ as the $g^1_2$'s on $E$ to produce the surface

1973: $\Sigma'$.  Thus we have a distinguished line on each $\Sigma'$

1974: corresponding to $|2\sigma|$.  And if $C'$ is the residual to

1975: $E\subset \Sigma'\cap X$, this line intersects $C'$ in a distinguished

1976: point.  What this shows is that there is a unirational variety $B$

1977: dominating a general fiber $Z=u_{4,0}^{-1}(p)$ such that after we

1978: base-change to $B$, we have a section $\tau$ of the family of quartic

1979: rational curves $C'$.  Thus the family of quartic rational curves over

1980: $B$ is a conic bundle with a section; therefore it is a

1981: $\PP^1$-bundle.  

1982: 

1983: \ps

1984: 

1985: It is easy to see that given a quartic rational curve $C'$ along with

1986: a point $p$, the locus of cubic scrolls containing $C$ is canonically

1987: birational to $\text{Sym}^2(C)\times \CC^*$, which is canonically

1988: birational to $\AAA^3$.  Thus the fiber product

1989: $\widetilde{H} \times_{\Hdg{4}{0}(X)} B\rightarrow B$ is canonically

1990: birational to $\AAA^3\times B\rightarrow B$.  In particular, since $B$

1991: is itself unirational, we conclude that

1992: $\widetilde{H}\times_{\Hdg{4}{0}(X)} B$ is also unirational.

1993: 

1994: \ps

1995: 

1996: Consider the

1997: composite morphism:

1998: \begin{equation}\begin{CD}

1999: \widetilde{H}\times_{\Hdg{4}{0}(X)} B @>>> \widetilde{H} @>>>

2000: \Hdg{5}{1}(X).

2001: \end{CD}\end{equation}

2002: Define $Y$ to be the image.  Because $\widetilde{H}\rightarrow

2003: \Hdg{5}{1}(X)$ has fiber dimension, we conclude that 

2004: \begin{equation}\text{dim}(Y)\geq \text{dim}

2005: \widetilde{H}\times_{\Hdg{4}{0}(X)} B - 1 = \text{dim}(B) + 3 -1 = 3+3-1

2006: = 5.\end{equation}

2007: So $Y$ is a unirational variety of dimension 5.  Since $J(X)$ contains

2008: no unirational varieties, we conclude that $Y$ is contained in a fiber

2009: of $u_{5,1}$.  Since $\widetilde{H}\rightarrow \Hdg{5}{1}(X)$ is

2010: dominant, we 

2011: conclude that a general point of $\Hdg{5}{1}(X)$ is contained in one

2012: of the varieties $Y$. 

2013: Finally, by theorem~\ref{thm-irrfib2}, we see that the general fiber of

2014: $u_{5,1}$ is an irreducible 5-fold.  Thus we have that $Y$ equals a

2015: fiber of $u_{5,1}$, so the general fiber of $u_{5,1}$ is an

2016: irreducible, unirational 5-fold.

2017: \end{proof}

2018: 

2019: \section{Quintic Rational Curves}

2020: 

2021: In this section we will prove that the Abel-Jacobi map

2022: $u_{5,0}:\Hdg{5}{0}(X)\rightarrow J(X)$ is dominant and the general

2023: fiber is an irreducible, unirational $5$-fold.  

2024: 

2025: \

2026: 

2027: The construction we use to understand quintic rational curves is as

2028: follows.  Define $I\subset \Hdg{5}{0}(X)\times \mathbb{G}(1,4)$ to be

2029: the locally closed subvariety parametrizing pairs $([C],[L])$ where

2030: $L$ is a $3$-secant line to $C$ which is not a $4$-secant line.  There

2031: is a unique isomorphism $\phi:C\rightarrow L$ which is the identity on

2032: $C\cap L$.  By lemma~\ref{lem-scr50}, there is a cubic scroll

2033: $h:\Sigma\rightarrow\PP^4$ such that $L$ is the directrix of $\Sigma$,

2034: $C\subset \Sigma$, and $\phi$ is the restriction of the projection

2035: $\pi:\Sigma\rightarrow L$ to $C$.  The residual to $C$ in $\Sigma\cap

2036: X$ is a quartic rational curve $C'$ (which does not intersect $L$ and which

2037: intersects lines of the ruling in a degree $2$ divisor).  

2038: 

2039: \begin{lem}~\label{lem-irr50}

2040: The scheme $\Hdg{5}{0}(X)$ is irreducible of dimension $10$.  The

2041: morphism $I\rightarrow \Hdg{5}{0}(X)$ is birational.  For a general

2042: pair $([C],[L])$ the residual quartic curve $C'$ is a smooth,

2043: nondegenerate quartic rational curve.

2044: \end{lem}

2045: 

2046: \begin{proof}

2047: This follows from ~\cite[corollary 9.3, theorem 9.4]{HRS1}.  

2048: \end{proof}  

2049: 

2050: What are the fibers of $I\rightarrow \Hdg{4}{0}(X)$?  By

2051: lemma~\ref{lem-g1240}, the fiber over $[C']$ for $C'$ a general

2052: quartic rational curve is simply an open subset of the $\PP^2$ which

2053: parametrizes the 

2054: collection of $g^1_2$'s on $C'$.  

2055: 

2056: \begin{thm}~\label{thm-urfib3}  The general fiber of

2057: $u_{5,0}:\Hdg{5,0}(X)\rightarrow J(X)$ is an irreducible, unirational

2058: $5$-fold.

2059: \end{thm}

2060: 

2061: \begin{proof}

2062: Since $I\rightarrow \Hdg{5}{0}(X)$ is birational, we will show that the

2063: general fiber of $u_{5,0}:I\rightarrow J(X)$ is an irreducible,

2064: unirational $5$-fold.  

2065: 

2066: \ps

2067: 

2068: Now $I$ surjects to an open subscheme of $\Hdg{4,0}(X)$ with

2069: irreducible fibers isomorphic to open subsets of $\PP^2$.  By the

2070: residuation trick, we know that $u_{5,0}$ is the pointwise inverse (up

2071: to constant translation) of the composite map:

2072: \begin{equation}\begin{CD}

2073: I @>>> \Hdg{4}{0}(X) @> u_{4,0} >> J(X)

2074: \end{CD}\end{equation}

2075: We know the general fiber of $u_{4,0}$ is an irreducible $3$-fold.

2076: Thus we conclude that the general fiber of $I_1\rightarrow J(X)$ is an

2077: irreducible $5$-fold.

2078: 

2079: \ps

2080: 

2081: To see that the general fiber is unirational, we once again use the

2082: fact that for general $p\in J(X)$ there is a morphism $B\rightarrow

2083: u^{-1}_{4,0}(p)$ such that 

2084: $B$ is unirational and the base-change to $B$ of the universal curve

2085: over $\Hdg{4}{0}(X)$ admits a section.  Thus the base-change of the

2086: universal curve is birational to $B\times \AAA^1$.  So the ``relative

2087: symmetric product'' of the universal curve is birational to

2088: $B\times\AAA^2$.  By lemma~\ref{lem-irr50}

2089: $I\rightarrow \Hdg{4}{0}(X)$ is 

2090: isomorphic to an open subset of the relative second symmetric product

2091: of the universal curve.  Thus the fiber product

2092: $B\times_{\Hdg{4}{0}(X)} I$ is birational to $B\times \AAA^2$.  Since

2093: $B$ is unirational, so is $B\times \AAA^2$.  And

2094: $B\times_{\Hdg{4}{0}(X)} I$ dominates the fiber over $p\in J(X)$.

2095: Thus we conclude that the general fiber of $I\rightarrow J(X)$ is

2096: unirational.  

2097: \end{proof}

2098: 

2099: \section{Quintic Curves of Genus $2$}

2100: 

2101: By B\'ezout's theorem, $X$ cannot contain a plane curve of degree

2102: $d>3$.  Thus the next case after quintic elliptic curves is quintic

2103: curves of genus $2$. 

2104: 

2105: \ps

2106: 

2107: In this section we will show that the Abel-Jacobi map

2108: $u_{5,2}:\Hdg{5}{2}(X)\rightarrow J(X)$ has image

2109: $\text{image}(u_{5,2}) = u_{1,0}(F)$.  Moreover the general fiber is

2110: irreducible and rational of dimension $8$.  The construction we

2111: will use to prove this is as follows.  Suppose that $C\subset X$ is a

2112: smooth quintic curve of genus $2$.  By Riemann-Roch, 

2113: $h^0(C,\OO_{\PP^4}(1)|_C) = 4$, so $C$ is contained in a hyperplane

2114: section $H\cap X$.  Similarly, $h^0(C,\OO_{\PP^4}(2)|_C)=9<10$, so $C$

2115: is contained in a quadric surface $Q\subset H$.  The residual to $C$

2116: in $Q\cap X$ is a line $L\subset X$.  Thus there is a morphism

2117: $f:\widetilde{\Hdg{5}{1}}(X)\rightarrow F$ where

2118: $\widetilde{\Hdg{5}{1}}(X)$ is the normalization of $\Hdg{5}{1}(X)$.

2119: 

2120: \begin{lem}~\label{lem-irr52}

2121: The scheme $\Hdg{5}{2}(X)$ is irreducible of dimension $10$ and the

2122: morphism $f:\widetilde{\Hdg{5}{2}}(X)\rightarrow F$ is dominant.

2123: \end{lem}

2124: 

2125: \begin{proof}

2126: This follows from the proof of ~\cite[theorem 10.1]{HRS1}.

2127: \end{proof}

2128: 

2129: What is the fiber of $f$?  Given a line $L\subset X$, to specify the

2130: curve $C$, it suffices to specify the hyperplane $H$ containing $L$

2131: and the quadric surface $Q\subset H$ containing $L$.  The collection

2132: of pairs 

2133: \begin{equation}

2134: \PP Q^\vee = \{([L],[H])\in F\times \PP^{4\vee}: L\subset H\}

2135: \end{equation}

2136: is a $\PP^2$-bundle over $F$.  And the collection of triples

2137: \begin{equation}

2138: I=\{([L],[H],[Q]): L\subset Q\subset H\}

2139: \end{equation}

2140: is a $\PP^6$-bundle over $\PP Q^\vee$.  By a dimension count we

2141: conclude that the morphism $g:\widetilde{\Hdg{5}{2}}(X)\rightarrow I$

2142: maps birationally to an open subset of $I$.  Thus the fibers of $f$

2143: are irreducible, rational varieties of dimension $8$.

2144: 

2145: \begin{cor}~\label{cor-urfib52}

2146: The Abel-Jacobi map $u_{5,2}$ has image a (translate of) an open

2147: subset of 

2148: $-u_{1,0}(F)$, and the general fiber of $u_{5,2}$

2149: is an irreducible, rational $8$-fold.

2150: \end{cor}

2151: 

2152: \begin{proof}

2153: By the residuation trick we know $u_{5,2}$ is the pointwise inverse of

2154: $u_{1,0}\circ g$.  And $u_{1,0}$ is an embedding.  Since $g$ is

2155: dominant, we conclude that $\text{image}(u_{5,2})$ is an open subset

2156: of $-u_{1,0}(X)$.  Since the fibers of $u_{5,2}$ equal the fibers of

2157: $g$, we conclude the general fiber of $u_{5,2}$ is an irreducible,

2158: rational variety of dimension $8$.

2159: \end{proof}

2160: 

2161: 

2162: \subsection{Irreducibility for $\Hdg{d}{0}(X)$}

2163: 

2164: We have avoided using the following result in the previous section,

2165: but we present it here to mention one important corollary.

2166: 

2167: \ps

2168: 

2169: \begin{thm}~\label{thm-irrd0}  For each $d$ the space $\Hdg{d}{0}(X)$

2170: is an irreducible,  reduced, local complete intersection scheme of

2171: dimension $2d$.  Moreover the general point of $\Hdg{d}{0}(X)$ is an

2172: unobstructed curve.   

2173: \end{thm}

2174: 

2175: This is theorem 1 of~\cite{HRS}.  We will not discuss the proof here,

2176: but we will prove a corollary that follows from

2177: theorem~\ref{thm-irrd0} and our analysis of $\Hdg{4}{0}(X)$ and

2178: $\Hdg{5}{0}(X)$.  Let $\widetilde{\Hdg{d}{0}}(X)$ denote the normalization

2179: of $\Hdg{d}{0}(X)$.

2180: 

2181: \ps

2182: 

2183: \begin{thm}~\label{thm-irrd02}  For each $d\geq 4$, the Abel-Jacobi

2184: morphism $\alpha_{d,0}:\widetilde{\Hdg{d}{0}}(X)\rightarrow J(X)$ is

2185: dominant and 

2186: the general fiber is irreducible.

2187: \end{thm}

2188: 

2189: \begin{proof}  

2190: Let $\mathcal{H}^d\rightarrow \overline{\Hdg{d}{0}(X)}$ denote the

2191: normalization of the closure of $\Hdg{d}{0}(X)$.  We will actually

2192: show that $\alpha_{d,0}:\mathcal{H}^d\rightarrow J(X)$ is dominant and the

2193: general fiber is irreducible. 

2194: Let us denote the Stein factorization of

2195: $\alpha_{d,0}$ as follows:

2196: \begin{equation}\begin{CD}

2197: \mathcal{H}^d @> \beta_d >> Z_d @> \gamma_d >> J(X)

2198: \end{CD}\end{equation}

2199: We need to prove that $\gamma_d$ is an open immersion.  By

2200: theorem~\ref{thm-irrd0}, we know that $Z_d$ is irreducible.  We will

2201: prove by induction that there is a rational section

2202: $\epsilon_d:J(X)\rightarrow Z_d$.  It then follows that $\gamma_d$ is

2203: an open immersion.  We have already established this result in case

2204: $d=4$ or $5$.  Therefore suppose that $d\geq 6$ and suppose that the

2205: theorem has been proved for all integers less than $d$ (and greater

2206: than $3$). 

2207: 

2208: \ps

2209: 

2210: Let $\Hdg{d-2}{0}(X)_{u}\subset \Hdg{d-2}{0}(X)$ denote the

2211: open subscheme of $\Hdg{d-2}{0}(X)$ such that the corresponding curve

2212: is unobstructed. 

2213: Let $\pi:\mathcal{C}^{d-2}\rightarrow \Hdg{d-2}{0}(X)_{u}$ denote the

2214: universal curve.

2215: The points of $\mathcal{C}^{d-2}$ parametrize pairs $([C],p)$ where

2216: $[C]\in\overline{\Hdg{d-2}{0}}(X)$ and $p\in C$.  

2217:   Let $L\subset X$ be a general line.  Let

2218: $U\subset \mathcal{C}^{d-2}$ denote the open subscheme parametrizing

2219: pairs $([C],p)$ such that $p\in X\setminus L$.  We have a family of

2220: $2$-planes $P\subset U\times \PP^4$ whose fiber over $([C],p)$ is

2221: $\text{span}(L,p)$.  Let $D\subset U\times X$ denote the intersection

2222: of $U\times X$ with $P$ inside $U\times \PP^4$.  Let $D_1\subset D$

2223: denote the constant family $U\times L\subset U\times \PP^4$.  Let

2224: $D_2$ denote the residual family of conics.  By

2225: ~\cite[corollary 2.7]{HRS1}, $D_2$ is flat over $U$.  Let $V\subset U$

2226: denote the open subscheme parametrizing pairs $([C],p)$ such that the

2227: corresponding conic $C'$ is smooth and such that $C$

2228: intersects $C'$ transversally at $p$ and in no other

2229: points.  Then over $V$ the union $\mathcal{C}^{d-2}\cup D_2\subset

2230: V\times X$ is a flat family of curves of degree $d$ and arithmetic

2231: genus $0$.  So there is an induced morphism $f:V\rightarrow

2232: \HI{dt+1}{X}$.  

2233: 

2234: \ps

2235: 

2236: We will prove that $f(V)\subset \overline{\Hdg{d}{0}}(X)$.  In fact we

2237: will prove that every point in $f(V)$ satisfies the conditions in

2238: ~\cite[lemma 2.3]{HRS1}, namely $H^1(C',N_{C'/X}(-p))=H^1(C,N_{C/X})=0$.

2239: It then

2240: follows that such a point is a smooth point of $\HI{dt+1}{X}$ and

2241: also that the node smooths so that the point is in

2242: $\overline{\Hdg{d}{0}}(X)$. 

2243: Let $([C],p)$

2244: be a point of $V$, and let $C'$ be the corresponding conic.  Let

2245: $B=C\cup C'$ denote the union.  By construction $B$ is a connected

2246: nodal curve.  We need to prove that $H^1(B,N_{B/X})=0$.  

2247: 

2248: \ps

2249: 

2250: We have the

2251: exact sequence:

2252: \begin{equation}\begin{CD}

2253: 0 @>>> N_{B/X}|_{C'}(-p) @>>> N_{B/X} @>>> N_{B/X}|_C @>>> 0.

2254: \end{CD}\end{equation}

2255: And we have two exact sequences:

2256: \begin{equation}\begin{CD}

2257: 0 @>>> N_{C'/X}(-p) @>>> N_{B/X}|_{C'}(-p) @>>> \OO_p @>>> 0 \\

2258: 0 @>>> N_{C/X} @>>> N_{B/X}|_{C} @>>> \OO_p @>>> 0

2259: \end{CD}.\end{equation}

2260: We have seen that $N_{C'/X}$ is either $\OO_{C'}(1)\oplus \OO_{C'}(1)$

2261: or $\OO_{C'}(2)\oplus \OO_{C'}$.  Thus $H^1(C',N_{C'/X}(-p))=0$.  By

2262: the long exact sequence in cohomology we conclude that

2263: $H^1(C',N_{B/X}|_{C'}(-p))=0$.  By assumption $H^1(C,N_{C/X})=0$,

2264: therefore also $H^1(C,N_{B/X}|_C)=0$.  By the long exact sequence in

2265: cohomology we conclude that $H^1(B,N_{B/X})=0$, i.e. $B$ is

2266: unobstructed.  

2267: 

2268: \ps

2269: 

2270: We conclude that $f(V)$ is contained in the smooth locus of

2271: $\overline{\Hdg{d}{0}}(X)$.  Thus we can factor $f$ as $g:V\rightarrow

2272: \mathcal{H}^d$.  By additivity $\theta(B)=\theta(C)+\theta(C')$.  And by

2273: residuation we have $\theta(C')=-\theta(L)$ (up to a fixed constant).

2274: Thus we conclude that the composite map:

2275: \begin{equation}\begin{CD}

2276: V @> g >> \mathcal{H}^d @> \alpha_{d,0} >> J(X)

2277: \end{CD}\end{equation}

2278: equals the pointwise inverse (up to a constant translation) of the

2279: composite:

2280: \begin{equation}\begin{CD}

2281: V @>>> \mathcal{H}^{d-2} @> \alpha_{d-2,0} >> J(X).

2282: \end{CD}\end{equation}

2283: Since $V\rightarrow \mathcal{H}^{d-2}$ has irreducible fibers, the Stein

2284: factorization of this composite is just the Stein factorization of

2285: $\alpha_{d-2,0}$ (or more precisely an open subscheme).  By the

2286: induction assumption $\gamma_{d-2}:Z_{d-2}\rightarrow J(X)$ is an open

2287: immersion.  By the universal property of the Stein factorization of

2288: $V\rightarrow J(X)$, there is an induced morphism

2289: $\epsilon_d:Z_{d-2}\rightarrow Z_d$.  This is a rational section of

2290: $\gamma_d:Z_d\rightarrow J(X)$, which shows that $\gamma_d$ is an open

2291: immersion.  Therefore the morphism

2292: $\alpha_{d,0}:\mathcal{H}^d\rightarrow J(X)$ is dominant and the

2293: general fiber is irreducible, as was to be shown.

2294: \end{proof}

2295: 

2296: \bibliography{my}

2297: \bibliographystyle{abbrv}

2298:  

2299: \end{document}  

2300: