1: \documentclass[12pt]{amsart} 

2: \usepackage{amsfonts,graphics,amsmath,amsthm,amsfonts,amscd, amssymb,amsmath,latexsym,multicol}

3: \usepackage{epsfig}

4: \usepackage{flafter}

5: 

6: \input diagrams

7: 

8: \makeatletter

9: 

10: \def\jobis#1{FF\fi

11:   \def\predicate{#1}%

12:   \edef\predicate{\expandafter\strip@prefix\meaning\predicate}%

13:   \edef\job{\jobname}%

14:   \ifx\job\predicate

15: }

16: 

17: \makeatother

18: 

19: \if\jobis{proposal}%

20:  \def\try{subsection}%

21: \else

22:   \def\try{section}%

23: \fi

24: 

25: 

26: \theoremstyle{plain}

27: \newtheorem{theorem}{Theorem}[\try]

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

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

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

31: \newtheorem{note}[theorem]{Note}

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

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

34: \newtheorem{definition-lemma}[theorem]{Definition-Lemma}

35: \newtheorem{question}[theorem]{Question}

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

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

38: \newtheorem{axiom}[theorem]{Axiom}

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

40: \newtheorem{principle}[theorem]{Principle}

41: \newtheorem{algorithm}[theorem]{Algorithm}

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

43: \newtheorem{project}[theorem]{Project}

44: \def\red{_{\text {red}}}

45: %1 Variable

46: \def\ec#1{e_{\text{top}}(#1)}

47: \def\t#1{\tilde{#1}}

48: \def\ol#1{\overline{#1}}

49: \def\ul#1{\underline{#1}}

50: \def\h#1{\hat{#1}}

51: \def\ideal#1.{I_{#1}}

52: \def\ring#1.{\mathcal {O}_{#1}}

53: \def\fring#1.{\hat{\mathcal {O}}_{#1}}

54: \def\proj#1.{\mathbb {P}(#1)}

55: \def\pr #1.{\mathbb {P}^{#1}}

56: \def\dpr #1.{\hat{\mathbb {P}}^{#1}}

57: \def\af #1.{\mathbb A^{#1}}

58: \def\Hz #1.{\mathbb F_{#1}}

59: \def\Hbz #1.{\overline{\mathbb F}_{#1}}

60: \def\fb#1.{\underset #1 {\times}}

61: \def\rest#1.{\underset {\ \ring #1.} \to \otimes}

62: \def\au#1.{\operatorname {Aut}\,(#1)}

63: \def\deg#1.{\operatorname {deg } (#1)}

64: \def\pic#1.{\operatorname {Pic}\,(#1)}

65: \def\pico#1.{\operatorname{Pic}^0(#1)}

66: \def\picg#1.{\operatorname {Pic}^G(#1)}

67: \def\ner#1.{NS (#1)}

68: \def\rdown#1.{\llcorner#1\lrcorner}

69: \def\rfdown#1.{\lfloor{#1}\rfloor}

70: \def\rup#1.{\ulcorner{#1}\urcorner}

71: \def\rcup#1.{\lceil{#1}\rceil}

72: \def\cone#1.{\operatorname {NE}(#1)}

73: \def\ccone#1.{\overline{\operatorname {NE}}(#1)}

74: \def\coef#1.{\frac{(#1-1)}{#1}}

75: \def\vit#1.{D_{\langle #1 \rangle}}

76: \def\mm#1.{\overline {M}_{0,#1}}

77: \def\H1#1.{H^1(#1,{\ring #1.})}

78: \def\ac#1.{\overline {\mathbb F}_{#1}}

79: \def\dir#1{\overset\rightarrow\to{#1}}

80: \def\dirl#1{\overset\longrightarrow\to{#1}}

81: \def\hnf#1{0=#1_0\subset #1_1\subset \dots \subset #1_r=#1}

82: \def\adj#1.{\frac {#1-1}{#1}}

83: \def\spn#1.{\overline{#1}}

84: \def\pek#1.#2.{\Cal P^{#1}(#2)}

85: \def\plk#1.#2.{\Cal P^{\leq #1}(#2)}

86: \def\ev#1.{\operatorname{ev_{#1}}}

87: \def\ilist#1.{{#1}_1,{#1}_2,\dots}

88: \def\bminv#1.{(\nu_1,s_1;\nu_2,s_2;\dots ;\nu_{#1},s_{#1};\nu_{r+1})}

89: \def\zinv#1.{(\nu_1,s_1;\nu_2,s_2;\dots ;\nu_{#1},s_{#1};0)}

90: \def\iinv#1.{(\nu_1,s_1;\nu_2,s_2;\dots ;\nu_{#1},s_{#1};\infty)}

91: 

92: %2 Variables

93: 

94: \def\kts{K_{\t{S}_0}^2}

95: \def\llist#1.#2.{{#1}_1,{#1}_2,\dots,{#1}_{#2}}

96: \def\lomitlist#1.#2.{{#1}_1,{#1}_2,\dots,\hat {{#1}_i}, \dots, {#1}_{#2}}

97: \def\lomitlistz#1.#2.{{#1}_0,{#1}_1,\dots,\hat {{#1}_i}, \dots, {#1}_{#2}}

98: \def\loc#1.#2.{\Cal O_{#1,#2}}

99: \def\fderiv#1.#2.{\frac {\partial #1}{\partial #2}}

100: \def\deriv#1.#2.{\frac {d #1}{d #2}}

101: \def\map#1.#2.{#1 \longrightarrow #2}

102: \def\rmap#1.#2.{#1 \dasharrow #2}

103: \def\emb#1.#2.{#1 \hookrightarrow #2}

104: \def\non#1.#2.{\text {Spec }#1[\epsilon]/(\epsilon)^{#2}}

105: \def\Hi#1.#2.{\text {Hilb}^{#1}(#2)}

106: \def\sym#1.#2.{\operatorname {Sym}^{#1}(#2)}

107: \def\Hb#1.#2.{\text {Hilb}_{#1}(#2)}

108: \def\Hm#1.#2.{\Hom_{#1}(#2)}

109: \def\prd#1.#2.{{#1}_1\cdot {#1}_2\cdots {#1}_{#2}}

110: \def\Bl #1.#2.{\operatorname {Bl}_{#1}#2}

111: \def\pl #1.#2.{#1^{\otimes #2}}

112: \def\mgn#1.#2.{\overline {M}_{#1,#2}}

113: \def\ialist#1.#2.{{#1}_1 #2 {#1}_2, #2\dots}

114: \def\pair#1.#2.{\langle #1, #2\rangle}

115: \def\vandermonde#1.#2.{\left|

116: \begin{matrix}

117: 1 & 1 & 1 & \dots & 1\\

118: {#1}_1 & {#1}_2 & {#1}_3 & \dots & {#1}_{#2}\\

119: {#1}_1^2 & {#1}_2^2 & {#1}_3^2 & \dots & {#1}_{#2}^2\\

120: \vdots & \vdots & \vdots & \ddots & \vdots\\

121: {#1}_1^{#2-1} & {#1}_2^{#2-1} & {#1}_2^{#2-1} & \dots & {#1}_{#2}^{#2-1}\\

122: \end{matrix}

123: \right|

124: }

125: \def\vandermondet#1.#2.{\left|

126: \begin{matrix}

127: 1 & {#1}_1   & {#1}_1^2 & \dots & {#1}_1^{#2-1}\\

128: 1 & {#1}_2   & {#1}_2^2 & \dots & {#1}_2^{#2-1}\\

129: 1 & {#1}_3   & {#1}_3^2 & \dots & {#1}_3^{#2-1}\\

130: \vdots & \vdots & \vdots & \ddots & \vdots\\

131: 1 & {#1}_{#2}& {#1}_{#2}^2 & \dots & {#1}_{#2}^{#2-1}\\

132: \end{matrix}

133: \right|

134: }

135: \def\gr#1.#2.{\mathbb{G}(#1,#2)}

136: 

137: 

138: %3 Variables 

139: 

140: \def\alist#1.#2.#3.{{#1}_1 #2 {#1}_2 #2\dots #2 {#1}_{#3}}

141: \def\zlist#1.#2.#3.{#1_0 #2 #1_1 #2\dots #2 #1_{#3}}

142: \def\lomitlist30#1.#2.#3.{{#1}_0,{#1}_1 #2 \dots #2\hat {{#1}_i} #2\dots #2 {#1}_{#3}}

143: \def\lmap#1.#2.#3.{#1 \overset{#2}{\longrightarrow} #3}

144: \def\mes#1.#2.#3.{#1 \longrightarrow #2 \longrightarrow #3}

145: \def\ses#1.#2.#3.{0\longrightarrow #1 \longrightarrow #2 \longrightarrow #3 \longrightarrow 0}

146: \def\les#1.#2.#3.{0\longrightarrow #1 \longrightarrow #2 \longrightarrow #3}

147: \def\res#1.#2.#3.{#1 \longrightarrow #2 \longrightarrow #3\longrightarrow 0}

148: \def\Hi#1.#2.#3.{\text {Hilb}^{#1}_{#2}(#3)}

149: \def\ten#1.#2.#3.{#1\underset {#2}{\otimes} #3}

150: \def\lomitlist30#1.#2.#3.{{#1}_0 #2 {#1}_1 #2 \dots #2 \hat {{#1}_i} #2 \dots #2 {#1}_{#3}}

151: 

152: 

153: 

154: %Operatornames 

155: 

156: \def\discrep{\operatorname{discrep}}

157: \def\Diff{\operatorname{Diff}}

158: \def\SHom{\operatorname{\mathcal{H}om}}

159: \def\CHom{\operatorname{\mathbf{Hom}}}

160: \def\Hom{\operatorname{Hom}}

161: \def\ext{\operatorname{Ext}}

162: \def\tor{\operatorname{Tor}}

163: \def\SExt{\operatorname{\Cal Ext}}

164: \def\Specan{\operatorname{\Cal Specan}}

165: \def\sp{\operatorname{Spec}}

166: \def\Proj{\operatorname{Proj}}

167: \def\Supp{\operatorname{Supp}}

168: \def\dif{\diff}

169: \def\dim{\operatorname{dim}}

170: \def\cod{\operatorname{codim}}

171: \def\deg{\operatorname{deg}}

172: \def\prj{\operatorname{Proj}}

173: \def\sg{\operatorname{Sing}}

174: \def\lg{\operatorname{length}}

175: \def\Pic{\operatorname{Pic}}

176: \def\Aut{\operatorname{Aut}}

177: \def\sln{\operatorname{SL}}

178: \def\det{\operatorname{det}}

179: \def\trdeg{\operatorname{trdeg}}

180: \def\coh{\operatorname{H}}

181: \def\vlm{\operatorname{Vol}}

182: \def\ker{\operatorname{Ker}}

183: \def\rk{\operatorname{rk}}

184: \def\im{\operatorname{Im}}

185: \def\re{\operatorname{Re}}

186: \def\nm{\operatorname{Nm}}

187: \def\SL{\operatorname{SL}}

188: \def\pgl{\operatorname{PGL}}

189: \def\gl{\operatorname{GL}}

190: \def\lcc{\operatorname{LLC}}

191: \def\lcs{\operatorname{LCS}}

192: \def\lcp{\operatorname{LCS^+}}

193: \def\cs{\operatorname{CC}}

194: \def\ann{\operatorname{Ann}}

195: \def\top{\operatorname{\mathfrak{Top}}}

196: \def\var{\operatorname{\Cal Var}}

197: \def\Hilb{\operatorname{Hilb}}

198: \def\Div{\operatorname{Div}}

199: \def\Chow{\operatorname{Chow}}

200: \def\mult{\operatorname{mult}}

201: \def\mov{\operatorname{Mov}}

202: \def\ord{\operatorname{ord}}

203: \def\resi{\operatorname{Res}}

204: \def\rest{\operatorname{res}}

205: \def\Sym{\operatorname{Sym}}

206: \def\gal{\operatorname{Gal}}

207: \def\inv{\operatorname{inv}}

208: \def\vol{\operatorname{vol}}

209: %\def\res{\operatorname{res}}

210: \def\obj{\operatorname{Obj}}

211: \def\op{\operatorname{op}}

212: \def\sign{\operatorname{sign}}

213: \def\Bs{\operatorname{Bs}}

214: \def\ch{\operatorname{CH}}

215: 

216: %Acronyms

217: 

218: \def\rar{\rightarrow} 

219: \def\tensor{\otimes}

220: \def\sd{\Sigma a_i D_i}

221: \def\C{\mathbb C}

222: \def\etl{\'etale }

223: \def\kd{K_X + \Delta}

224: \def\kb{K_X + B}

225: \def\kd{K + \Delta}

226: \def\pon{(P^1)^{\times n}}

227: \def\kpd{K_{X/Y} + \Sigma a_i D_i}

228: \def\cl{{\Cal L}}

229: \def\cq{{\Cal Q}}

230: %\def\red{_{\text {red}}}

231: \def\demop{\demo{Proof}}

232: \def\p{\mathbb P}

233: \def\mi{\mu _{\text {min}}}

234: \def\ma{\mu _{\text {max}}}

235: \def\qf{{\mathbb Q}-\text{factorial}}

236: \def\qc{{\mathbb Q}-\text{Cartier}}

237: \def\gb{\goth B}

238: \def\ts{\tilde{S}}

239: \def\ta{\tilde{A}}

240: \def\tb{\tilde{B}}

241: \def\hs{\hat S}

242: \def\hb{\hat B}

243: \def\dd{\Cal D}

244: \def\e{\Cal E}

245: \def\ZZ{\mathbb Z}

246: \def\gH{\goth H}

247: \def\alg{\pi_1^{\text {alg}}}

248: \def\orb{\pi_1^{\text {alg}}}

249: \def\orbalg{\pi_1^{\text {orb-alg}}}

250: \def\aos{{\mathbb A}^1_*}

251: \def\os{\ol{S_0}}

252: \def\ko{K_{\os}}

253: \def\kt{K_{\t{S}_0}}

254: \def\sb{\ol{\Sigma}}

255: \def\ef{E_{\infty}}

256: \def\eo{E_0}

257: \def\e1{E_1}

258: \def\e2{E_2}

259: \def\del{\bar {\partial}}

260: \def\OO{\Cal O}

261: \def\cc{\Cal E}

262: \def\xderiv{x\frac d{dx}}

263: \def\ds{\displaystyle}

264: \def\mn{\overline {M}_{0,n}}

265: \def\idear{\trianglelefteq}

266: \def\qle{\sim_{\mathbb Q}}

267: \def\bd{_{\bullet}}

268: \def\st{_{\text st}}

269: \def\Bd{\mathbb}

270: \def\ilim{\mathop{\varprojlim}\limits}  

271: \def\Top{\mathcal{Top}}

272: \def\wbM{{}^\sharp \mathbf{M}}

273: \def\wbE{{}^\sharp \mathbf{E}}

274: \def\wM{{}^\sharp M}

275: \def\wX{{}^\sharp X}

276: \def\wY{{}^\sharp Y}

277: \def\wW{{}^\sharp W}

278: \def\wZ{{}^\sharp Z}

279: \def\wF{{}^\sharp F}

280: \def\wR{{}^\sharp R}

281: \def\wE{{}^\sharp E}

282: \def\wG{{}^\sharp G}

283: \def\wB{{}^\sharp B}

284: \def\wbD{{}^\sharp \mathbf{D}}

285: \def\wD{{}^\sharp D}

286: \def\wU{{}^\sharp U}

287: \def\wK{{}^\sharp K}

288: \def\wH{{}^\sharp H}

289: %\def{\wX}{\widetilde{X}}

290: %\def{\wY}{\widetilde{Y}}

291: %\def{\wZ}{\widetilde{Z}}

292: %\def{\wF}{\widetilde{F}}

293: %\def{\wR}{\widetilde{R}}

294: %\def{\wE}{\widetilde{E}}

295: %\def{\wD}{\widetilde{\mathbb{D}}}

296: \def\wf{{}^\sharp f}

297: \def\wg{{}^\sharp g}

298: 

299: %Commutative diagram 

300: 

301: \def\normalbaselines{\baselineskip20pt\lineskip3pt\lineskiplimit3pt}

302: \def\llongrightarrow{\relbar\joinrel\longrightarrow}

303: \def\beginCD{\matrix\def\normalbaselines{

304: \baselineskip20pt\lineskip3pt\lineskiplimit3pt}}

305: \def\doneCD{\endmatrix}

306: \def\mapright#1{\smash{\mathop{\llongrightarrow}\limits^{#1}}}

307: \def\mapdown#1{\big\downarrow\rlap{$\vcenter

308: {\hbox{$\scriptstyle#1$}}$}}

309: \def\lmapdown#1{\big\downarrow\llap{$\vcenter

310: {\hbox{$\scriptstyle{#1}\,\,\,\,$}}$}}

311: \def\Mapdown#1{\mapdown{#1}}

312: \def\Mapup#1{\big\uparrow\rlap{$\vcenter{\hbox{$\scriptstyle#1$}}$}}

313: \def\mapse#1{

314: {\vcenter{\hbox{$\mathop{\smash{\raise1pt\hbox{$\diagdown$}\!\lower7pt

315: \hbox{$\searrow$}}\vphantom{p}}\limits_{#1}\vphantom{\mapdown{}}$}}}}

316: \def\Llongrightarrow{\relbar\joinrel\relbar\joinrel\longrightarrow}

317: \def\Mapright#1{\smash{\mathop{\Llongrightarrow}\limits^{#1}}}

318: \def\nsim{\,\,\hbox{{$\backslash$}\llap{$\sim$}}\,\,}

319: 

320: % Boxes for Exams

321: 

322: 

323: \def\shadow#1{\rightline

324: {\vbox{\hrule \hbox{\vrule\vbox{\hbox{\vbox to .5in{\hsize = 1.5in #1}}}

325: }\hrule height2pt}

326: }

327: }

328: 

329: \def\oldshadow#1{\rightline

330: {\vbox{\hrule \hbox{\vrule\vbox{\hbox{\vbox to .5in{\hsize = 1.5in #1}}}

331: \vrule width2pt}\hrule height2pt}

332: }

333: }

334: 

335: 

336: \def\pagebox{\rlap

337: {{ \vbox{\hrule \hbox{\vrule\vbox{\hbox{\vbox to

338: .3in{\hsize = .3in\ }}}\vrule}\hrule}}

339: }

340: }

341: 

342: \def\VR#1.{height#1pt&\omit&&\omit&&\omit&&\omit&&\omit&\cr}

343: \def\CR{\VR 8.}

344: \def\SR{\VR 2.}

345: \def\LINE{\CR\tablerule\CR}

346: 

347: 

348: \def\VRT#1.{height#1pt&\omit&&\omit&\cr}

349: \def\CRT{\VRT 5.}

350: \def\SRT{\VRT 6.}

351: \def\LINET{\CRT\tablerule\CRT}

352: 

353: % Dynkin Diagrams

354: 

355: \def\dnl{=\hspace{-.2cm}>\hspace{-.2cm}=}

356: \def\dnr{=\hspace{-.2cm}<\hspace{-.2cm}=}

357: \def\sn{-\hspace{-.17cm}-}

358: 

359: 

360: 

361: \begin{document}

362: \title{Shokurov's Rational Connectedness Conjecture}

363: \author{Christopher D. Hacon} 

364: \address{Department of Mathematics \\  

365: University of Utah\\  

366: 155 South 1400 E\\

367: JWB 233\\

368: Salt Lake City, UT 84112, USA}

369: \email{hacon@math.utah.edu}

370: \author{James M\textsuperscript{c}Kernan} 

371: \address{Department of Mathematics\\ 

372: University of California at Santa Barbara\\ 

373: Santa Barbara, CA 93106, USA} 

374: \email{mckernan@math.ucsb.edu}

375: 

376: 

377: \begin{abstract} We prove a conjecture of V. V. Shokurov which in particular implies 

378: that the fibers of a resolution of a variety with divisorial log terminal singularities

379: are rationally chain connected.

380: \end{abstract} 

381: 

382: \thanks{The first author was partially supported by NSA research grant no:

383:   MDA904-03-1-0101 and by a grant from the Sloan Foundation.  Part of this work was

384:   completed whilst the second author was visiting the Pontificia Universidad Cat\'olica

385:   del Per\'u, Lima and the Tokyo Mathematics Department.}

386: 

387: \maketitle

388: \pagestyle{plain}

389: \section{Introduction}

390: \label{s_introduction} 

391: 

392: In recent years it has become increasingly clear that the geometry of higher dimensional

393: varieties is closely related to the geometry of rational curves on these varieties. From

394: the point of view of the minimal model program, one expects that rational curves on

395: varieties with mild singularities (e.g. log terminal singularities) share many of the

396: basic properties of rational curves on smooth varieties.  Surprisingly, very little seems

397: to be known in this direction.  The purpose of this paper is to give an affirmative answer

398: to several natural questions that arise in this context.  For example we show that:

399: 

400: \textit{If $(X, \Delta)$ is a divisorially log terminal pair and

401:   $f:\map Y.X.$ is a birational morphism, then for any $x\in X$,

402:   $f^{-1}(x)$ is rationally chain connected and in particular covered

403:   by rational curves.}

404: 

405: Two immediate consequences of this are:

406: \begin{enumerate} 

407: \item \textit{If $(X, \Delta )$ is a divisorially log terminal pair, and $g: \rmap X.Z.$

408:   is a rational map to a proper variety which is not everywhere defined, then $Z$ contains

409:   a rational curve.}

410: \item \textit{If $(X, \Delta )$ is a divisorially log terminal pair,

411:   then $X$ is rationally chain connected if and only if it is

412:   rationally connected.}

413: \end{enumerate}  

414: 

415: We now turn to a more detailed discussion of the results of this paper.

416: 

417: \begin{definition}\label{d_mod} Let $X$ be a reduced, separated scheme of finite type 

418: over an algebraically closed field (so that every irreducible component of $X$ is a

419: variety) and let $V$ be any subset.  We will say that a curve $C$ is a \textbf{chain

420:   modulo $V$} if $C$ union a subset of $V$ is connected.  We will say that \textbf{$X$ is

421:   rationally chain connected modulo $V$}, if any two points $x$ and $y$, which belong to

422: the same connected component of $X$, belong to a chain of rational curves modulo $V$.

423: \end{definition}

424: 

425: Note that if $V$ is empty and $X$ is irreducible, then $X$ is rationally chain connected

426: in the usual sense.  Note also that the disjoint union of two copies of $\pr 1.$ is

427: rationally chain connected but not connected.  Here is the main result of this paper:

428: 

429: \begin{theorem}\label{t_main} Let $(X,\Delta)$ be a log pair and let $f\colon\map X.S.$ be 

430: a projective morphism such that $-K_X$ is relatively big and $\ring X.(-m(K_X+\Delta))$ is

431: relatively generated, for some $m>0$.  Let $g\colon\map Y.X.$ be any birational morphism

432: and let $\pi\colon\map Y.S.$ be the composite morphism.

433: 

434: Then every fibre of $\pi$ is rationally chain connected modulo the inverse image of the

435: locus of log canonical singularities.

436: \end{theorem}

437: 

438: We note that we can weaken the hypothesis if we assume that $K_X+\Delta$ is kawamata log

439: terminal, see \eqref{l_semi}.  This result has some interesting consequences.  We start

440: with a very general result about the fundamental group of a log pair:

441: 

442: \begin{corollary}\label{c_main} Let $(X,\Delta)$ be a projective log pair such that 

443: $-(K_X+\Delta)$ is semiample and $-(K_X+\Delta)$ is big.

444: 

445: Then the fundamental group of $X$ is a quotient of the fundamental group of the locus of log

446: canonical singularities of the pair $(X,\Delta)$.  

447: \end{corollary}

448: 

449: Some other interesting consequences of \eqref{t_main} arise when we eliminate the locus of

450: log canonical singularities from the statement:

451: 

452: \begin{corollary}\label{c_rcck} Let $(X,\Delta)$ be a kawamata log terminal pair. 

453: Let $f\colon\map X.S.$ be a projective morphism such that $-(K_X+\Delta)$ is relatively nef 

454: and $-K_X$ is relatively big.  

455: 

456:  Then every fibre of $f$ is rationally chain connected. 

457: \end{corollary}

458: 

459:  With a little more work, we can also prove:

460: 

461: \begin{corollary}\label{c_rccl} Let $(X,\Delta)$ be a log canonical pair.  Let 

462: $f\colon\map X.S.$ be a projective morphism such that $-(K_X+\Delta)$ is relatively ample.

463: 

464:  Then every fibre of $f$ is rationally chain connected. 

465: \end{corollary}

466: 

467: \eqref{c_rccl} was conjectured by Shokurov in \cite{Shokurov00a}, and was proved there,

468: assuming the MMP.  In the same paper, he also conjectured the following, which he proved under

469: the same assumptions:

470: 

471: \begin{corollary}\label{c_dlt} Let $(X,\Delta)$ be a divisorially log terminal pair. 

472: 

473: If $g\colon\map Y.X.$ is any birational morphism then the fibres of $g$ are rationally chain

474: connected.

475: \end{corollary}

476: 

477: It was pointed out in \cite{Shokurov00a} that one then gets the following:

478: 

479: \begin{corollary}\label{c_covered} Let $f:\rmap X.Y.$ be a rational morphism of normal proper 

480: varieties such that $(X,\Delta)$ is a divisorially log terminal pair for some effective

481: divisor $\Delta$.  Then, for each closed point $x\in X$, the indeterminacy locus of $x$ is

482: covered by rational curves.

483: \end{corollary}

484: 

485: Recall that if $W\subset X\times Y$ is the closure of the graph of $f$ and $p$ and $q$ are

486: the projections from $W$ to $X$, and $Y$, then the indeterminacy locus of $x$ is defined

487: as $q(p^{-1}(x))$.

488: 

489: Another very interesting consequence of \eqref{c_dlt} is:

490: 

491: \begin{corollary}\label{c_equivalent} Let $(X,\Delta)$ be a divisorially log terminal pair.  

492: Then $X$ is rationally chain connected iff it is rationally connected.  

493: \end{corollary}

494: 

495:  The same methods which are used to prove \eqref{t_main}, in conjunction with 

496: \eqref{c_equivalent}, yield the following:

497: 

498: \begin{corollary}\label{c_over} Let $(X,\Delta)$ be a kawamata log terminal pair.  

499: Let $f\colon\map X.S.$ be a projective morphism with connected fibres such that

500: $-(K_X+\Delta)$ is relatively nef and $-K_X$ is relatively big.  Let $g\colon\map Y.X.$ be

501: any birational morphism, and let $\pi\colon\map Y.S.$ be the composition.  Let $T$ be any

502: irreducible subset of $S$ and let $W$ be the inverse image of $T$ inside $Y$.

503: 

504: Then there is an irreducible closed subset $E$ of $W$, which dominates $T$ and which has

505: connected and rationally connected fibres.

506: \end{corollary}

507: 

508: One interesting feature of rationally connected varieties is that a family of rationally

509: connected varieties over a curve always admits a section.  As a consequence of

510: \eqref{c_over}, we are able to show that a similar result holds for families of Fano

511: varieties, a result which was also conjectured by Shokurov:

512: 

513: \begin{corollary}\label{c_section} Let $(X,\Delta)$ be a kawamata log terminal pair.  

514: Let $f\colon\map X.S.$ be a projective morphism with connected fibres such that

515: $-(K_X+\Delta)$ is relatively nef and $-K_X$ is relatively big.  Let $g\colon\map Y.X.$ be

516: any birational morphism, and let $\pi\colon\map Y.S.$ be the composition.

517: 

518:  Then $\pi$ has a section over any curve.  

519: \end{corollary}

520: 

521: Another reason why rationally connected and rationally chain connected varieties are

522: interesting, is because their topology is particularly simple.  In particular this means

523: that the intersection theory of a divisorially log terminal pair and a resolution are closely

524: related:

525: 

526: \begin{corollary}\label{c_chow} Let $(X,\Delta)$ be a kawamata log terminal pair.  

527: Let $f\colon\map X.S.$ be a projective morphism with connected fibres such that

528: $-(K_X+\Delta)$ is relatively nef and $-K_X$ is relatively big.  Let $g\colon\map Y.X.$ be

529: any birational morphism, and let $\pi\colon\map Y.S.$ be the composition.

530: 

531: Then the natural map 

532: $$

533: \pi_*\colon\map \ch^0(Y).\ch^0(S).,

534: $$

535: is an isomorphism.  

536: \end{corollary}

537: 

538:    The following, although not a direct consequence of \eqref{t_main}, is proved using very

539: similar methods:

540: 

541: \begin{corollary}\label{c_iitaka} Let $(X,\Delta)$ be a kawamata log terminal projective pair

542: and let $\rmap X.Z.$ be the Iitaka fibration associated to $-(K_X+\Delta)$.

543: 

544:  Then $Z$ is rationally connected.  

545: \end{corollary}

546: 

547:  As a consequence we obtain:

548: \begin{corollary}\label{c_fano} Let $(X,\Delta)$ be a kawamata log terminal projective pair

549: and suppose that that $-(K_X+\Delta)$ is big and nef.  

550: 

551:  Then $X$ is rationally connected.  

552: \end{corollary}

553: 

554: We remark that \eqref{c_fano} has also been recently proved by Zhang, using a similar

555: argument, see \cite{Zhang04}.  We were working on this paper, when his result appeared on

556: the archive.  As Zhang points out in \cite{Zhang04}, \eqref{c_fano} implies the following:

557: 

558: \begin{corollary}\label{c_simply} Let $(X,\Delta)$ be a kawamata log terminal projective pair

559: and suppose that that $-(K_X+\Delta)$ is big and nef. 

560: 

561:  Then $X$ is simply connected.  

562: \end{corollary}

563: 

564: It should be pointed out though that it is more natural to prove the stronger result that

565: the smooth locus of $X$ has finite fundamental group.  The only known case of this much

566: stronger result, is for surfaces, see \cite{KM99}.  It is proved in \cite{McKernan02} that

567: at least the algebraic fundamental group of the smooth locus is finite.

568: 

569: \begin{remark}\label{r_replace} Note that \eqref{c_rcck} and \eqref{c_simply}, which 

570: are stated for kawamata log terminal pairs extend to the case when the locus of log

571: canonical singularities is a point (equivalently, by connectedness, when $V$ has dimension

572: zero).

573: \end{remark}

574: 

575: We now give a quick sketch of the proof of \eqref{t_main}.  Let us suppose that we want to

576: prove that $F$ is rationally connected.  By the main result of \cite{GHS03}, it suffices

577: to prove that whenever we have a test rational map $t\colon\map F.Z.$, then either $Z$ is

578: uniruled or it is a point, see \eqref{l_sufficient}.  Suppose $Z$ is not uniruled.  By the

579: main result of \cite{BDPP04}, $K_Z$ is pseudo-effective.

580: 

581: Suppose for a moment, that we can produce a divisor $\Theta$ with three key properties:

582: \begin{itemize} 

583: \item[(a)] $K_F+\Theta$ has Kodaira dimension zero, 

584: \item[(b)] there is an ample $\mathbb{Q}$-divisor $H$ on $Z$ such that $t^*H\leq \Theta$, and 

585: \item[(c)] $K_F+\Theta$ is log terminal on the general fibre.  

586: \end{itemize} 

587: 

588: By log additivity, it follows that the Kodaira dimension of $K_F+\Theta$ is at least the

589: dimension of $Z$, so that $Z$ is a point by (a), see \eqref{p_uni}.  It remains to

590: indicate how to produce the divisor $\Theta$.  In our case, $F$ is a smooth divisor and a

591: component of the fibre of a resolution of a Fano fibration $f\colon\map X.S.$.  The

592: existence of a divisor $\Delta$ on the total space of the Fano fibration satisfying (a),

593: (b) and (c) is quite straightforward, and it is equally straightforward to lift this to a

594: resolution $g\colon\map Y.X.$ of the total space.  Conditions (b) and (c) then descend to

595: $F$.  The tricky part is ensuring that condition (a) continues to hold.

596: 

597: The first step is to realise $F$ as a component of the locus of log canonical

598: singularities, with respect to some divisor.  At this point we apply the main technical

599: result of \cite{HM05b}, which says, roughly speaking, that we can lift appropriate sections from $F$

600: to $Y$.

601: 

602: \section{Notation and conventions}

603: \label{s_notation}

604: 

605: We work over the field of complex numbers $\mathbb{C}$.  A $\mathbb{Q}$-Cartier divisor

606: $D$ on a normal variety $X$ is \textit{nef} if $D\cdot C\geq 0$ for any curve $C\subset

607: X$.  We say that two $\mathbb{Q}$-divisors $D_1$, $D_2$ are $\mathbb{Q}$-linearly

608: equivalent ($D_1\sim _{\mathbb{Q}} D_2$) if there exists an integer $m>0$ such that $mD_i$

609: are linearly equivalent.  Given a morphism of normal varieties $g\colon\map Y.X.$, we say

610: that two $\mathbb{Q}$-divisors $D_1$ and $D_2$ are $\mathbb{Q}$-$g$-linearly equivalent

611: ($D_1\sim _{g,\mathbb{Q}} D_2$ if there is a positive integer $m$ and a Cartier divisor

612: $B$ on $X$ such that $mD_1\sim mD_2+f^*B$.  We say that a $\mathbb{Q}$-Weil divisor $D$ is

613: big if we may find an ample divisor $A$ and an effective divisor $B$, such that $D \sim

614: _{\mathbb{Q}} A+B$.  A \textit{log pair} $(X,\Delta)$ is a normal variety $X$ and an

615: effective $\mathbb{Q}$-Weil divisor $\Delta$ such that $K_X+\Delta$ is

616: $\mathbb{Q}$-Cartier.  A projective morphism $g \colon\map Y.X.$ is a \textit{log

617:   resolution} of the pair $(X,\Delta )$ if $Y$ is smooth and $g^{-1}(\Delta )\cup

618: \{\,\text {exceptional set of $g$}\,\}$ is a divisor with normal crossings support.  We

619: write $g^*(K_X +\Delta )=K_Y +\Gamma$ and $\Gamma =\sum a_i\Gamma _i$ where $\Gamma _i$

620: are distinct reduced irreducible divisors.  The log discrepancy of $\Gamma_i$ is $1-a_i$.

621: The \textit{locus of log canonical singularities of the pair $(X,\Delta)$}, denoted

622: $\lcs(X,\Delta)$, is equal to the image of those components of $\Gamma$ of coefficient at

623: least one (equivalently log discrepancy at most zero).  The pair $(X,\Delta )$ is

624: \textit{kawamata log terminal} if for every (equivalently for one) log resolution

625: $g\colon\map Y.X.$ as above, the coefficients of $\Gamma$ are strictly less than one, that

626: is $a_i<1$ for all $i$.  Equivalently, the pair $(X,\Delta)$ is kawamata log terminal if

627: the locus of log canonical singularities is empty.  We say that the pair $(X,\Delta)$ is

628: \textit{divisorially log terminal} if the coefficients of $\Delta$ lie between zero and

629: one, and there is a log resolution such that the coefficients of the $g$-exceptional

630: divisors are all less than one.

631: 

632: Given a pair $(X,\Delta)$ and a birational morphism, we will often consider decompositions

633: of the form

634: $$

635: K_Y+\Gamma=\pi^*(K_X+\Delta)+E,

636: $$

637: where $\Gamma$ and $E$ are effective, often with no common components, and where $E$ is

638: exceptional.  Here, even though the equals sign denotes $\mathbb{Q}$-linearly equivalence,

639: in fact it is often implicitly assumed that $\Gamma$ is a combination of exceptional

640: divisors and the strict transform of $\Delta$.  In fact, under the further assumption that

641: $\Gamma$ and $E$ have no common components, such a decomposition is unique.  However, we

642: will often make no such assumption concerning the support of $\Gamma$, so that such a

643: decomposition is far from unique.

644: 

645: Let $\phi\colon\rmap X.Y.$ be a rational map.  We say that a subset $V$ of $X$

646: \textit{dominates} $Y$, if the inverse image of $V$, in the graph of $\phi$, dominates

647: $Y$.

648: 

649: Let $f\colon\map X.S.$ be a morphism.  We say that a divisor $E$ is exceptional for $f$,

650: if $f(E)$ has codimension at least two in $S$.

651: 

652: \section{Some Examples}

653: \label{s_examples}

654: 

655: In this section we collect together some examples, whose purpose is to show that the

656: results stated in \S \ref{s_introduction} are in some sense best possible, and to motivate

657: their proofs.

658: 

659:  

660: 

661: We remark that in theorem \eqref{c_fano}, one can not remove the hypothesis that

662: $-(K_X+\Delta)$ is nef or that $(X,\Delta )$ is kawamata log terminal as shown by the

663: following well known example.  Let $f\colon\map S.C.$ be any $\pr 1.$-bundle over an

664: elliptic curve and let $E$ be a section of minimal self-intersection.  Then $f$ is the

665: maximal rationally connected fibration of $S$.  Suppose first that $E^2<0$.  Then

666: $$

667: -(K_S+tE) \qquad \text{is} \qquad  

668: \begin{cases} \text{big}   & \text{for any $t<2$} \\

669:               \text{ample} & \text{for any $1<t<2$} \\

670:               \text{nef}   & \text{for any $1\leq t\leq 2$} \\

671:               \text{lc}    & \text{for any $t\leq 1$} \\

672:               \text{klt}   & \text{for any $t<1$} \\

673: \end{cases}

674: $$

675: Clearly $S$ is not rationally chain connected.  If $t=1$ then $K_S+E$ is log canonical

676: and $-(K_S+E)$ is nef and big.  If $t<1$ then $K_S+tE$ is kawamata log terminal and

677: $-(K_S+tE)$ is big but not nef.  If $1<t<2$ then $-(K_S+tE)$ is ample but not log canonical.

678: Notice that by contracting the negative section $E$, we get a rationally chain connected

679: surface $T$.  Therefore, rational chain connectedness is not a birational property of $S$;

680: of course if $f\colon\map X'.X.$ is a birational morphism of varieties and $X'$ is

681: rationally chain connected, then $X$ is also rationally chain connected.

682: 

683:   One might well ask the following:

684: \begin{question}\label{q_ask} Let $(X,\Delta)$ be a klt pair, $\Delta$ effective and

685: $-(K_X+\Delta)$ nef.  If $X\dashrightarrow W$ is the MRCC fibration then does it follow

686: that 

687: $$

688: \dim X\geq \dim W +\kappa (-(K_X+\Delta))\ ?

689: $$ 

690: \end{question}

691: 

692: Now suppose that we let $E_2$ be the unique non-split extension of $\ring C.$ by $\ring

693: C.$ and set $S$ to be the projectivisation of $E_2$.  Then $E$ has self-intersection zero.

694: $-K_S=2E$, so $-K_S$ is nef of numerical dimension one.  But in fact no multiple of $E$

695: moves.  Indeed if it did, then there would be an \'etale cover $\pi\colon\map C'.C.$

696: such that $\pi^*E_2$ splits.  It is easy to see that in characteristic zero this never

697: happens.  Thus the Kodaira dimension of $-K_S$ is zero and the sum of the Kodaira

698: dimension plus the dimension of $C$ is less than the dimension of $S$ so that the answer

699: to \eqref{q_ask} is no, in general.  

700: 

701: Finally, we give an easy example, to illustrate the fact that one cannot expect every

702: component of the exceptional locus of a resolution of a divisorially log terminal

703: singularity to be rationally connected.  In fact it is rarely the case that each

704: individual component is rationally chain connected.

705: 

706: Let $X$ be a smooth threefold.  Pick a point $x\in X$ and let $\map Y_1.X.$ blow up this

707: point.  Now let $\map Y.Y_1.$ blow up a cubic curve in the exceptional divisor, which is a

708: copy of $\pr 2.$.  Consider the birational morphism $\pi\colon\map Y.X.$.  Denote the

709: first exceptional divisor, a copy of $\pr 2.$, by $E$ and the other by $F$.  It is easy to

710: check that $-(K_Y+4/5E+1/5F)$ is relatively ample.  Note that $E\cup F$ is rationally

711: chain connected, but that $F$ is not even rationally chain connected.  Indeed $F$ is once

712: again a $\pr 1.$-bundle over an elliptic curve.

713: 

714: \section{Uniruled, rationally connected and rationally chain connected varieties}

715: \label{s_uni}

716: 

717: In this section we give sufficient conditions to ensure that a variety is either uniruled,

718: or rationally connected, or rationally chain connected.  First a criterion 

719: to ensure that a variety is uniruled:

720: 

721: \begin{proposition}\label{p_uni} Let $(X,\Delta)$ be a projective log pair and let $h\colon\map X.F.$ 

722: and $t\colon\rmap F.Z.$ be a morphism and a rational map, where $F$ and $Z$ are

723: projective, with the following properties:

724: \begin{enumerate} 

725: \item the locus of log canonical singularities of  $K_X+\Delta$ does not dominate $Z$, where

726: $\rmap X.Z.$ is the composition of $t$ and $h$,

727: \item $K_X+\Delta$ has Kodaira dimension at least zero on the general fibre of $\rmap

728: X.Z.$, (that is, if $g\colon\map Y.X.$ resolves the indeterminacy of $\rmap X.Z.$ then

729: $g^*(K_X+\Delta)$ has Kodaira dimension at least zero on the general fibre of the induced

730: morphism $\map Y.Z.$),

731: \item $K_X+\Delta$ has Kodaira dimension at most zero, and

732: \item there is an ample divisor $A$ on $F$ such that $h^*A\leq\Delta$.  

733: \end{enumerate} 

734:  

735:  Then either $Z$ is a point or it is uniruled.  

736: \end{proposition}

737: \begin{proof} Suppose that $Z$ is not uniruled.  Blowing up $Z$, we may assume that $Z$ is

738: smooth.  Let $g\colon\map Y.X.$ be a log resolution of $X$, such that the induced rational

739: map $\rmap Y.Z.$ is in fact a morphism $\psi\colon\map Y.Z.$ .  We may write

740: $$

741: K_Y+\Theta=g^*(K_X+\Delta)+E,

742: $$

743: where $\Theta$ and $E$ are effective, with no common components and $E$ is exceptional.  Now

744: set $\Gamma=\Theta+\epsilon E'$, where $\epsilon$ is a sufficiently small rational number,

745: and $E'$ is the support of the exceptional locus.  With this choice of $\Gamma$, the

746: Kodaira dimension of $K_Y+\Gamma$ is at least zero on the general fibre of $\psi$, and the

747: support of $\Gamma$ contains the full exceptional locus.

748: 

749: As $Z$ is not uniruled, it follows by the main result of \cite{BDPP04} that $K_Z$ is

750: pseudo-effective.  By (4), and our choice of $\Gamma$, $\Gamma$ contains the pullback of

751: an ample divisor from $F$.  Possibly replacing $\Gamma$ by a linearly equivalent divisor,

752: we may find an ample divisor $G$ on $Z$ such that $\psi^*G\leq\Gamma$.  Since $K_Y+\Gamma$

753: is log terminal on the general fibre of $\psi$, it follows by log additivity of the

754: Kodaira dimension, see Corollary 2.11 of \cite{HM05b}, that the Kodaira dimension of

755: $K_Y+\Gamma$ is at least the dimension of $Z$.  As the Kodaira dimension of $K_Y+\Gamma$

756: is at most the Kodaira dimension of $K_X+\Delta$, it follows that $Z$ is a point.

757: \end{proof}

758: 

759:  It is easy to use \eqref{p_uni} to prove that a variety is rationally chain connected:

760: 

761: \begin{lemma}\label{l_sufficient} Let $F$ be a normal variety.  

762: \begin{enumerate} 

763: \item $F$ is rationally connected iff for every non-constant dominant rational map $t\colon\rmap

764: F.Z.$, $Z$ is uniruled.

765: \item $F$ is rationally chain connected modulo $V$ iff for every non-constant dominant

766: rational map $t\colon\rmap F.Z.$, either $Z$ is uniruled or $V$ dominates $Z$.

767: \end{enumerate} 

768: \end{lemma}

769: \begin{proof} It is clear that the image of every rationally connected variety is rationally 

770: connected.  In particular the image of every rationally connected variety is either

771: uniruled or a point.  If $V$ does not dominate $Z$, then pick two general points $x$ and

772: $y$ of $F$.  Then they are connected by a chain of rational curves modulo $V$, and the

773: image of one of these curves must be a rational curve through the general point of $Z$.

774: 

775: 

776: To prove the reverse direction, first observe that $F$ is uniruled, by applying the basic

777: criterion to the identity map $\map F.F.$.  Let $F'$ be a smooth model of $F$, and let

778: $\rmap F'.Z.$ be the maximal rationally connected fibration of $F'$.  Blowing up, we may

779: assume that this map is a morphism.  Since $F$ is birational to $F'$, this induces a

780: rational map $t\colon\rmap F.Z.$.  By assumption either $Z$ is a point, or $Z$ is

781: uniruled, or $V$ dominates $Z$.

782: 

783: As $F$ is uniruled, $\map F'.Z.$ is a non-trivial fibration.  This morphism has the

784: defining property that if a rational curve $C$ meets a very general fibre $G$, then $C$ is

785: contained in $G$.  By the main result of \cite{GHS03}, we can lift any rational curve

786: which passes through the general point of $Z$ to $F$.  Thus $Z$ is not uniruled.  If $Z$

787: is a point, then $F$ is rationally connected.  The only other possibility is that $V$

788: dominates $Z$.  In this case $F$ is certainly rationally chain connected modulo $V$.

789: \end{proof}

790: 

791: \begin{corollary}\label{c_uni} Let $(X,\Delta)$ be a log pair and let $h\colon\map X.F.$

792: be a morphism.  Suppose that for every rational map $t\colon\rmap

793: F.Z.$, either the locus of log canonical singularities of $K_X+\Delta$

794: dominates $Z$, or (2-4) of \eqref{p_uni} hold.

795:  

796: Then $F$ is rationally chain connected modulo the image $R$ of the locus where $K_X+\Delta$

797: is not kawamata log terminal. 

798: \end{corollary}

799: \begin{proof} We are going to apply the criterion of \eqref{l_sufficient}.  Suppose we are

800: given a rational map $t\colon\rmap F.Z.$.  If $R$ dominates $Z$ there is nothing to prove.

801: Otherwise we just need to prove that $Z$ is either uniruled or a point and for this we

802: just need to observe that conditions (1-4) of \eqref{p_uni} hold.  \end{proof}

803: 

804: \section{A fine analysis of the fibres of a Fano fibration}

805: \label{s_fine}

806: 

807: In this section we state a detailed theorem about the fibres of a Fano fibration, which

808: despite being technical in nature, we expect will be of independent interest.  We fix some

809: notation which will hold throughout the section.  Let $(X,\Delta)$ be a log pair and let

810: $f\colon\map X.S.$ be a morphism such that $K_X+\Delta \sim _{\mathbb{Q},f} 0$ and

811: $\Delta$ is $f$-big.

812: 

813: Let $s\in S$ be any closed point and let $K$ be any effective $\mathbb{Q}$-Cartier divisor

814: on $S$.  Let $g\colon\map Y.X.$ be any birational morphism, such that the fibre of the

815: composite morphism $\pi\colon\map Y.S.$ over $s$ union the exceptional locus of $g$ union

816: the strict transform of $f^*K+\Delta$ is a divisor with normal crossings.  Let $\llist

817: F.k.$ be the components of the fibre $\pi^{-1}(s)$, which have log discrepancy greater

818: than zero, with respect to $K_X+\Delta$, and let $F$ be their union.  

819: 

820: Given $t\in [0,1]$, set $\Delta_t=\Delta+tf^*K$, $\Gamma_t=\Gamma+t\pi^*K$ (where $\Gamma$

821: is defined below) and let $V_t$ be the closure of

822: $$

823: \lcs(Y,\Gamma_t)-\lcs(Y,\Gamma).

824: $$

825: We let $\Gamma'_t$ denote the fractional part of  $\Gamma_t$.  

826: 

827: \begin{theorem}\label{t_fibres} With the notation above, we may pick $K$ and divisors 

828: $\Gamma$ and $E$ on $Y$, with the following properties:

829: \begin{enumerate} 

830: \item The equation

831: $$

832: K_Y+\Gamma=g^*(K_X+\Delta)+E,

833: $$

834: holds, where $\Gamma$ and $E$ are effective, with no common components and $E$ is

835: $g$-exceptional.  Moreover, we may write $\Gamma=A+B$, where $A$ is $\pi$-ample and $B$ is

836: effective.

837: \item $F=V_1$.  

838: \item Possibly relabelling and rescaling, we may assume that there are rational numbers,

839: $0=\zlist t.,.k.=1$, such that $V_i=\alist F.\cup.i.$, where, here and elsewhere, we adopt

840: the shorthand subscript $i$ in lieu of $t_i$.  For $\epsilon$ sufficiently small,

841: $V_{t_i-\epsilon}=V_{t_{i-1}}$.  Denote by $\Theta_i$ the restriction of $\Gamma'_i$ to

842: $F_i$.

843: \item Let $H$ be any $\pi$-ample divisor.  Then there is a constant $M$ such that 

844: $$

845: h^0(X,\ring X.(m(K_{F_i}+\Theta_i)+H))\leq M

846: $$

847: for all sufficiently divisible positive integers $m$, and any $1\leq i\leq k$. 

848: \item $F_i$ is rationally chain connected modulo $W_i=F_i\cap \lcs(Y,\Gamma_{i-1})$.  

849: \item If $K_Y+\Gamma$ is kawamata log terminal then $F_1$ is rationally connected.  

850: \item If $K_X+\Delta$ is kawamata log terminal then we may pick

851: $\Gamma$ so that $K_Y+\Gamma$ is kawamata log terminal.

852: \end{enumerate} 

853: \end{theorem}

854: 

855: We prove each part of \eqref{t_fibres} in a series of lemmas: 

856: 

857: \begin{lemma}\label{l_one} (1) of \eqref{t_fibres} holds.  

858: \end{lemma}

859: \begin{proof} We may write

860: $$

861: K_Y+\Gamma_1=g^*(K_X+\Delta)+E_1,

862: $$

863: where $\Gamma_1$ and $E_1$ are effective, with no common components and $E_1$ is

864: $g$-exceptional.  Now let $E'$ be the sum of all the exceptional divisors, taken with

865: coefficient one.  Set $\Gamma_2=\Gamma_1+\delta E'$, $E_2=E_1+\delta E'$, for some

866: positive rational number $\delta$.  If we choose $\delta$ small enough, then we do not

867: change the locus of log canonical singularities.  As $\Gamma_2$ and the strict transform

868: of $\Delta$ union the exceptional locus have the same support, it follows that $\Gamma_2$

869: is $\pi$-big.  It follows that we may write $\Gamma_2\sim_{\mathbb{Q},\pi} A+B$, where $A$ is

870: $\pi$-ample and $B$ is effective.  Let

871: $$

872: K_Y+\Gamma=g^*(K_X+\Delta)+E,

873: $$

874: be the decomposition obtained by cancelling like terms on both sides of

875: $$

876: K_Y+((1-\epsilon)\Gamma_2+\epsilon B)+\epsilon A=g^*(K_X+\Delta)+E_2.

877: $$

878: Thus $\Gamma$ and $E$ are effective with no common components, $E$ is exceptional and of

879: course $\epsilon A\leq\Gamma$ is relatively ample.

880: \end{proof}

881: 

882: \begin{lemma}\label{l_two} (2) of \eqref{t_fibres} holds.    

883: \end{lemma}

884: \begin{proof} If we pick $K$ sufficiently singular at $s$, then we may assume that

885: $F\subset V_1$.  Possibly replacing $K$ by a linearly equivalent $\mathbb{Q}$-divisor, we

886: may then assume that $F=V_1$.  \end{proof}

887: 

888: \begin{lemma}\label{l_three} (3) of \eqref{t_fibres} holds.    

889: \end{lemma}

890: \begin{proof} By (2), for every $0\leq t\leq 1$, $V_t$ is a union of components of $F$.  Let

891: $t_i$ be the smallest value of $t$, such that $F_i$ is a component of $\lcs(Y,\Gamma_t)$.

892: Possibly perturbing $K$, we may assume that $t_i\neq t_j$, if $i\neq j$ and so possibly

893: re-ordering and rescaling, we may assume that $0<\alist t.<.k.=1$. \end{proof}

894: 

895: \begin{lemma}\label{l_four} (4) of \eqref{t_fibres} holds.    

896: \end{lemma}

897: \begin{proof} This is the most technical part of \eqref{t_fibres}. 

898: 

899: We are certainly free to replace $H$ by a multiple.  Pick a divisor $G$ on $X$ such that

900: $g^*G\geq (\dim X+2)H$.  Then 

901: \begin{align*} 

902: g_*\ring Y.(m(K_Y+\Gamma_i)+(\dim X+2)H)  &\subset g_*\ring Y.(g^*G+mE)\\

903:                                           &=\ring X.(G),\\

904: \end{align*} 

905: for $m$ sufficiently divisible.  Therefore $\pi _*\ring Y.(m(K_Y+\Gamma_i)+(\dim

906: X+2)H)\subset f_*\ring X.(G)$.  On the other hand, by (3.17) of \cite{HM05b}, since

907: $\Theta_i$ does not contain $F_i$, we may lift any section of

908: $$

909: \pi_*(\ring X.(m(K_{F_i}+\Theta_i)+H))=H^0(X,\ring X.(m(K_{F_i}+\Theta_i)+H)),

910: $$

911: to $\pi_*(\ring Y.(m(K_Y+\Gamma_i)+(\dim X+2)H))$.

912: \end{proof}

913: 

914: 

915: \begin{lemma}\label{l_five} (5) of \eqref{t_fibres} holds.    

916: \end{lemma}

917: \begin{proof} We are going to apply \eqref{c_uni}.  Let $\psi\colon\rmap F_i.Z.$ be any

918: rational map, such that the locus where $K_{F_i}+(\Gamma_i-F_i)|_{F_i}$ is not kawamata

919: log terminal does not dominate $Z$.  We just need to check that conditions (2-4) of

920: \eqref{p_uni} hold.  Now

921: $$

922: K_Y+\Gamma_i=E_i,

923: $$

924: where $E_i$ is effective and exceptional, and $F_i$ is not contained in the support of

925: $E_i$.  It follows that the Kodaira dimension of $K_Y+\Gamma_i$ restricted to the general

926: fibre of $\psi$ is at least zero.  We may assume that $W_i$ does not dominate $Z$, so that

927: $\Gamma_i=\Gamma_i'$, on the general fibre of $\psi$.  It follows that $K_{F_i}+\Theta_i$

928: has Kodaira dimension at least zero on the general fibre of $\psi$.  Thus (2) holds.  (3)

929: is an immediate consequence of (4) of \eqref{t_fibres}.  (4) holds by (1) of

930: \eqref{t_fibres}.

931: \end{proof}

932: 

933: \begin{lemma}\label{l_six} (6) of \eqref{t_fibres} holds.    

934: \end{lemma}

935: \begin{proof} Since $W_0=\lcs(Y,\Gamma)$ is by assumption empty, it follows that 

936: $F_1$ is rationally chain connected.  As $F_1$ is smooth, it is in fact rationally

937: connected.

938: \end{proof}

939: 

940: \begin{lemma}\label{l_seven} (7) of \eqref{t_fibres} holds.    

941: \end{lemma}

942: \begin{proof} Clear from the construction of $\Gamma$ given in \eqref{l_one}.  

943: \end{proof}

944: 

945: \section{Proof of \eqref{t_main}}

946: \label{s_proof}

947: 

948: \begin{lemma}\label{l_alter} If $-(K_X+\Delta)$ is semiample and $-K_X$ is big, 

949: then we may find a divisor $\Delta'\geq\Delta$, such that some multiple of $K_X+\Delta'$

950: is linearly equivalent to zero, where $\Delta'$ contains an ample divisor, and where the locus

951: of log canonical singularities of $K_X+\Delta'$ is contained in the locus of log canonical

952: singularities of $K_X+\Delta$.  

953: \end{lemma}

954: \begin{proof} If we pick a general element $D\in |-m(K_X+\Delta)|$, where $m$ is sufficiently 

955: divisible, then $K_X+\Delta+D/m$ has the same locus of log canonical singularities as

956: $K_X+\Delta$.  Replacing $\Delta$ by $\Delta+D/m$, we may assume therefore that $\Delta$

957: is big and that some multiple of $K_X+\Delta$ is linearly equivalent to zero.  As $\Delta$

958: is big, we may write $\Delta\sim_{\mathbb{Q}}A+B$, where $A$ is ample and $B$ is

959: effective.  Let $\Delta'=(1-\epsilon)\Delta+\epsilon A+\epsilon B$.  Then for $\epsilon$

960: sufficiently small, the locus of log canonical singularities of $K_X+\Delta'$ is contained

961: in the locus of log canonical singularities of $K_X+\Delta$.

962: \end{proof}

963: 

964: \begin{proof}[Proof of \eqref{t_main}] Let $s\in S$.  This result is local over

965: $s\in S$.  Passing to an open neighbourhood of $s\in S$, we may as well assume that

966: $-(K_X+\Delta)$ is semiample.  By \eqref{l_alter} we may assume therefore that

967: $\Delta=A+B$, where $A$ is ample and $B$ is effective, and that some multiple of

968: $K_X+\Delta$ is linearly equivalent to zero.

969: 

970: Note that as $g$ has connected fibres, it follows that if $\map Y'.Y.$ is any birational

971: map, then we are free to replace $Y$ by $Y'$.  Thus we may assume that the hypothesis and

972: notation of \S \ref{s_fine} hold.

973: 

974: By induction on $i$, it suffices to prove that $F_i$ is rationally chain connected modulo

975: $W_i$, which is (5) of \eqref{t_fibres}.

976: \end{proof}

977: 

978: \section{Proof of Corollaries}

979: \label{s_corollaries}

980: 

981: \begin{proof}[Proof of \eqref{c_main}] Since $X$ is covered by rational curves which 

982: intersect an irreducible component of $V$, the main theorem of \cite{Campana91} implies

983: that the image of the fundamental group of $V$ is of finite index in the fundamental group

984: of $X$.

985: 

986:  Let $\pi\colon\map Y.X.$ be any \'etale morphism of degree $r$, where $Y$ is connected,

987: let $W$ be the inverse image of $V$, and set $\Gamma=\pi^*\Delta$.  Then

988: $K_Y+\Gamma=\pi^*(K_X+\Delta)$, so that $W$ is the locus of log canonical singularities of

989: $K_Y+\Gamma$ and $-(K_Y+\Gamma)$ is big and nef.  Note that $W$ is connected, by the

990: connectedness theorem.

991: 

992: The proof now divides into two cases.  If $V$ is non-empty, then the result follows

993: immediately from the fact that $\pi$ is arbitrary and $W$ is always connected.  

994: 

995: Now suppose that $V$ is empty.  In this case, we adopt the convention that the fundamental

996: group of the empty set is the trivial group.  By what we have already proved the

997: fundamental group is finite and so it suffices to prove that the algebraic fundamental

998: group is trivial.  Now

999: $$

1000: \chi(\ring Y.)=r\chi(\ring X.).

1001: $$

1002: On the other hand, 

1003: $$

1004: h^i(Y,\ring Y.)=h^i(X,\ring X.)=0 \qquad \text{for $i>0$,} \qquad 

1005: $$

1006: by Kawamata-Viehweg vanishing, so that 

1007: $$

1008: \chi(\ring Y.)=\chi(\ring X.)=1,

1009: $$

1010: whence $r=1$.  But then the fundamental group of $X$ is trivial and the fundamental

1011: group of $V$ certainly surjects onto the fundamental group of $X$.  

1012: \end{proof}

1013: 

1014: \begin{lemma}\label{l_semi} Let $(X,\Delta)$ be a kawamata log terminal pair and

1015: let $f\colon\map X.S.$ be a projective morphism.  Suppose that $-(K_X+\Delta)$ is

1016: relatively nef and $-K_X$ is relatively big.

1017: 

1018:  Then $-(K_X+\Delta)$ is relatively semiample. 

1019: \end{lemma}

1020: \begin{proof} It suffices to exhibit a divisor $\Theta$, such that $K_X+\Theta$ is 

1021: kawamata log terminal and $-(K_X+\Theta)$ is ample, since then, by the

1022: base point free theorem, every nef divisor is relatively semiample.

1023: 

1024:   By assumption we may write 

1025: $$

1026: -K_X\sim_{\mathbb{Q},f} A+B,

1027: $$

1028: where $A$ is relatively ample and $B$ is effective.  Let 

1029: $$

1030: \Theta=(1-\epsilon)\Delta+\epsilon B.

1031: $$

1032: 

1033:  Then 

1034: $$

1035: -(K_X+\Theta)\sim_{\mathbb{Q},f} -(1-\epsilon)(K_X+\Delta)+\epsilon A

1036: $$

1037: is relatively ample, as it is the sum of a relatively nef divisor and a relatively

1038: ample divisor.  In particular $K_X+\Theta$ is $\mathbb{Q}$-Cartier.  But then

1039: $$

1040: K_X+\Theta=K_X+(1-\epsilon)\Delta+\epsilon B,

1041: $$

1042: is certainly kawamata log terminal, for $\epsilon$ small enough.  

1043: \end{proof}

1044: 

1045: 

1046: \begin{proof}[Proof of \eqref{c_rcck}] By \eqref{l_semi} $\ring X.(-m(K_X+\Delta))$ is 

1047: relatively generated, for some $m>0$, and so the result follows from \eqref{t_main}, since

1048: the locus of log canonical singularities is empty.

1049: \end{proof}

1050: 

1051: \begin{proof}[Proof of \eqref{c_rccl}] By \eqref{t_main}, we already know that the fibres

1052: of $f$ are rationally chain connected modulo the locus $V$ of log canonical singularities.

1053: It remains to prove that the fibres of $V$ over $S$ are rationally chain connected.  By

1054: induction on the dimension, it suffices to prove that if $W$ is a log canonical centre,

1055: then $W$ is rationally chain connected modulo the union $R$ of the log canonical centres

1056: properly contained in $W$.  As in the proof of \eqref{t_main}, by \eqref{l_alter}, we may

1057: assume that $\Delta$ contains an ample divisor and that some multiple of $K_X+\Delta$

1058: is linearly equivalent to zero.  Let $g\colon\map Y.X.$ be a log resolution of the pair

1059: $(X,\Delta)$.  We may write

1060: $$

1061: K_Y+\Gamma=g^*(K_X+\Delta)+E,

1062: $$

1063: where $\Gamma$ and $E$ are effective, with no common components and $E$ is

1064: $g$-exceptional.  Let $F$ a log canonical centre of $K_Y+\Gamma$ which dominates $W$,

1065: minimal with this property.  Then $F$ is the intersection of divisors of log discrepancy

1066: zero, with centre $W$.  In particular we may find a divisor $\Theta$ such that

1067: $$

1068: (K_Y+\Gamma)|_F=K_F+\Theta.

1069: $$

1070: We are going to apply \eqref{c_uni} to $h\colon\map F.W.$.  Let $t\colon\rmap W.Z.$ be

1071: a rational map.  Blowing up further, we may assume that the induced map $\psi\colon\map

1072: F.Z.$ is a morphism.  Note that the image in $W$ of the locus of log canonical

1073: singularities of $K_F+\Theta$ is equal to $R$, by minimality of $F$.  So we may assume

1074: that that the locus of log canonical singularities of $K_F+\Theta$ does not dominate $Z$.

1075: We need to check that conditions (2-4) hold.  (2) holds, exactly as in the proof of

1076: \eqref{t_main}.  (3) holds by (4) of \eqref{t_fibres} and (4) holds by assumption.

1077: \end{proof}

1078: 

1079: \begin{proof}[Proof of \eqref{c_dlt}] By (2.43) of \cite{KM98}, we may assume that 

1080: the pair $(X,\Delta)$ is kawamata log terminal, and in this case we may apply

1081: \eqref{t_main} with $f$ the identity.  \end{proof}

1082: 

1083: \begin{proof}[Proof of \eqref{c_covered}] Immediate from \eqref{c_dlt}.  

1084: \end{proof}

1085: 

1086: \begin{proof}[Proof of \eqref{c_equivalent}] If $X$ is rationally connected then it is

1087: certainly rationally chain connected, so suppose that $X$ is rationally chain connected.

1088: Let $g\colon\map Y.X.$ be a resolution of singularities.  By \eqref{c_dlt} $g$ has

1089: rationally chain connected fibres.  It follows that $Y$ is rationally chain connected. 

1090: But then $Y$ is rationally connected, as it is smooth, and so $X$ is rationally connected 

1091: as well.  

1092: \end{proof}

1093: 

1094: \begin{proof}[Proof of \eqref{c_over}] We work locally in a neighbourhood of the generic

1095: point of $T$.  \eqref{l_semi} implies that $-(K_X+\Delta)$ is relatively semiample.  As in

1096: the proof of \eqref{t_main} and \eqref{l_alter}, we may therefore assume $-(K_X+\Delta)$

1097: is ample and that $\Delta$ contains an ample divisor $A$,

1098: 

1099: Suppose that $T\neq S$.  Pick any ample divisor $H$ in $S$ which contains $T$.  Let $t$ be

1100: the largest rational number such that $K_X+\Delta+tf^*H$ is log canonical.  Possibly

1101: perturbing $\Delta$ by an ample divisor in $X$, we may assume that there is a unique log

1102: canonical centre $V$ of $K_X+\Delta+tf^*H$.  By connectedness $V$ has connected fibres

1103: over $T$.  Since $V$ is the unique log canonical centre, it is certainly a minimal log

1104: canonical centre.  Thus $V$ is normal.  It follows that $V$ has irreducible fibres.  By

1105: Kawamata's subadjunction formula, see Theorem 1 of \cite{Kawamata97}, there is a boundary

1106: $\Theta$ on $V$, such that

1107: $$

1108: (K_X+\Delta)|_V=K_V+\Theta,

1109: $$

1110: where $K_V+\Theta$ is kawamata log terminal.  Possibly replacing $A$ by a linearly 

1111: equivalent divisor, we may assume that $\Theta$ contains an ample divisor.  Thus we 

1112: are free to restrict to $V$ and replace $X$ by $V$ and $S$ by the image of $V$.  

1113: 

1114: Thus by induction on the dimension, we may assume that $T=S$.  In this case the general

1115: fibre over $T$ is kawamata log terminal and so by \eqref{t_main} and \eqref{c_equivalent}

1116: the general fibre is rationally connected.  \end{proof}

1117: 

1118: \begin{proof}[Proof of \eqref{c_section}] Pick a resolution $g\colon\map Y.X.$ and let 

1119: $\pi\colon\map Y.S.$ be the composition.  Pick a curve $C\subset S$, and let 

1120: $E$ be a subvariety of $Y$ with rationally connected fibres over $C$, whose existence

1121: is guaranteed by \eqref{c_over}.   Now apply the main result of \cite{GHS03}.  \end{proof}

1122: 

1123: \begin{proof}[Proof of \eqref{c_chow}] Even though this would seem to be a fairly standard

1124: result, we could not find a reference, and so we include a proof for completeness.

1125: 

1126: First note that if $\pi$ is any surjective morphism, then $\pi_*$ is always surjective, as

1127: $\ch^0(S)$ is generated by the points of $S$.  Now suppose that $\alpha$ is a cycle in

1128: $\ch^0(Y)$, whose image $\beta$ in $\ch^0(S)$ is zero.  It remains to prove that $\alpha$

1129: is equivalent to zero.  First pick a curve $\Sigma$ in $Y$ which is the intersection of

1130: general hyperplanes, containing the support of $\alpha$.  Then we may replace $\alpha$ by

1131: an equivalent cycle in $\Sigma$, in such a way that the support of $\alpha$ belongs to the

1132: general fibre.

1133: 

1134: As $\beta$ is equivalent to zero, by definition of $\ch^0(S)$, it follows that there is a

1135: curve $C$ in $S$ such that the support of $\beta$ belongs to $C$, and such that $\beta$ is

1136: linearly equivalent to zero in $C$.  By \eqref{c_section} we may find a curve $D\subset Y$

1137: which maps birationally down to $C$.  Let $\alpha'$ be the pullback of $\beta$ to $D$.  As

1138: $\alpha$ is equivalent to zero in $C$, $\alpha'$ is certainly equivalent to zero in $D$.  

1139: 

1140: On the the other hand, by our choice of $\alpha$, the fibres of $\pi$ over the support of

1141: $\beta$ are rationally connected.  It follows that $\alpha$ and $\alpha'$ are equivalent,

1142: so that $\alpha$ is indeed zero in $\ch^0(Y)$.  \end{proof}

1143: 

1144: \begin{proof}[Proof of \eqref{c_iitaka}] If $K_X+\Delta$ is kawamata log terminal, then 

1145: the locus of log canonical singularities is empty and the result is therefore immediate

1146: from \eqref{c_uni}.  \end{proof}

1147: 

1148: \begin{proof}[Proof of \eqref{c_fano}] By assumption the Iitaka fibration of $-(K_X+\Delta)$ 

1149: is birational, and as rational connectedness is a birational invariant, the result follows

1150: from \eqref{c_iitaka}.

1151: \end{proof}

1152: 

1153: \begin{proof}[Proof of \eqref{c_simply}] The pair $(X,\Delta)$ is kawamata log terminal iff 

1154: the locus of log canonical singularities is empty, and so this result is an immediate

1155: consequence of \eqref{c_main}.  \end{proof}

1156: 

1157: \bibliographystyle{hamsplain} 

1158: 

1159: \bibliography{/home/mckernan/Jewel/Tex/math}

1160: 

1161: 

1162: \end{document}

1163: 

1164: 

1165: