1: % Izzet:

2: %  

3: %  definability and degree

4: %  tau-invertible sheaves and degrees.  how many up to isom?

5: %

6: %%  EDIT;  FIX

7: %%  Fix \PO

8: %%  After ++++++ is scraps.

9: %%  See %% ??

10: 

11: %\documentclass[11pt]{article}

12: \documentclass{amsart}

13:           

14: \raggedbottom

15: 

16: \usepackage{fullpage,eucal,amsmath,amsthm,amsfonts,verbatim,amsopn,amssymb}

17: 

18: \usepackage[all]{xy}

19: 

20: \newcommand\dmo{\DeclareMathOperator}

21: 

22: \renewcommand\sec{\section}

23: \newcommand\subs{\subsection*}

24: %\newcommand\ssubs{\subsubsection*}

25: \newcommand\para{\paragraph*}

26: \newcommand\bea{\begin{array}{r@{,\ \ \ \ \ \ \ }l}}

27: \newcommand\ena{\end{array}}

28: \newcommand\barr{\begin{array}}

29: \newcommand\earr{\end{array}}

30: \newcommand\ben{\begin{enumerate}}

31: \newcommand\een{\end{enumerate}}

32: \newcommand\beq{\begin{equation}}

33: \newcommand\eeq{\end{equation}}

34: \newcommand\bqa{\begin{eqnarray*}}

35: \newcommand\eqa{\end{eqnarray*}}

36: \newcommand\bqan{\begin{eqnarray}}

37: \newcommand\eqan{\end{eqnarray}}

38: \newcommand\bit{\begin{itemize}}

39: \newcommand\eit{\end{itemize}}

40: \newcommand\bfi{\begin{figure}[htbp]}

41: \newcommand\efi{\end{figure}}

42: \newcommand\bce{\begin{center}}

43: \newcommand\ece{\end{center}}

44: \newcommand\note{\marginpar}

45: \newcommand\bpr{\begin{proof}}

46: \newcommand\epr{\end{proof}}

47: 

48: \newcommand{\ignore}[1]{}

49: 

50: %% theorems

51: 

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

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

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

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

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

57: 

58: \theoremstyle{definition}

59: 

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

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

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

63: \newtheorem{remarks}[theorem]{Remarks}

64: \newtheorem{examples}[theorem]{Examples}

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

66: 

67: \newcommand\bthm{\begin{theorem}}

68: \newcommand\ethm{\end{theorem}}

69: \newcommand\bcn{\begin{conjecture}}

70: \newcommand\ecn{\end{conjecture}}

71: \newcommand\bla{\begin{lemma}}

72: \newcommand\ela{\end{lemma}}

73: \newcommand\bco{\begin{corollary}}

74: \newcommand\eco{\end{corollary}}

75: \newcommand\bpro{\begin{proposition}}

76: \newcommand\epro{\end{proposition}}

77: \newcommand\bdf{\begin{definition}}

78: \newcommand\edf{\end{definition}}

79: \newcommand\bex{\begin{example}}

80: \newcommand\eex{\end{example}}

81: \newcommand\brm{\begin{remark}}

82: \newcommand\erm{\end{remark}}

83: \newcommand\brms{\begin{remarks}}

84: \newcommand\erms{\end{remarks}}

85: \newcommand\bqu{\begin{question}}

86: \newcommand\equ{\end{question}}

87: 

88: %% symbol abbrevatiations

89: 

90: \newcommand\ind{\indent}

91: \newcommand\no{\noindent}

92: \newcommand\bsn{\bigskip \noindent}

93: \newcommand\bsk{\bigskip}

94: \newcommand\msn{\medskip \noindent}

95: \newcommand\msk{\medskip}

96: \newcommand\ssn{\smallskip \noindent}

97: \newcommand\ssk{\smallskip}

98:   

99: \newcommand\tu{\textup}

100: 

101: %% math and logical symbols

102: 

103: \newcommand\Prf{\textsc{Proof}}

104: \newcommand\prf{\textsc{Proof. \quad}}

105: \newcommand\prfprop{\textsc{Proof of Proposition. \quad}}

106: \newcommand\sprf{\textsc{Sketch of Proof. \quad}}

107: 

108: \newcommand\sr{\stackrel}

109: \newcommand\<{\langle}

110: \renewcommand\>{\rangle}

111: \newcommand\ot{\otimes}

112: \newcommand\op{\oplus}

113: 

114: \newcommand\sub{\subseteq}

115: \renewcommand\sup{\supseteq}

116: \newcommand\subn{\subset}

117: \newcommand\supn{\supset}

118: \newcommand\all{\forall}  

119: \newcommand\ex{\exists}

120: \newcommand\surj{\twoheadrightarrow}

121: \newcommand\inj{\hookrightarrow}

122: \newcommand\im{\Rightarrow}

123: \newcommand\rim{\Leftarrow}

124: \newcommand\ra{\rightarrow}

125: \newcommand\la{\leftarrow}

126: \newcommand\lra{\longrightarrow}

127: \newcommand\lla{\longleftarrow}

128: \newcommand\hra{\hookrightarrow}

129: \newcommand\map{\mapsto}

130: \newcommand\lmap{\longmapsto}

131: \newcommand\seq{\sim}

132: \newcommand\teq{\approx}

133: \newcommand\teqq{\parallel}

134: \newcommand\ch{\choose}

135: \newcommand\eq{\equiv}

136: \newcommand\dv{\uplus}

137: \newcommand\rst{|}

138: 

139: %% letters and fonts

140: 

141: \newcommand\al{{\alpha}}

142: \newcommand\be{{\beta}}

143: \newcommand\ga{{\gamma}}

144: \newcommand\dl{\delta}

145: \newcommand\dltl{\tilde{\delta}}

146: \newcommand\lm{\lambda}

147: \newcommand\Lm{\Lambda}

148: \newcommand\cm{\otimes}

149: \newcommand\om{\omega}

150: \newcommand\omt{\omega^\tau}

151: \newcommand\Om{\Omega}

152: \newcommand\Omt{\Omega^\tau}

153: \newcommand\Omtp{\Omega^{\tau,p}}

154: \newcommand\Omti{\Omega^{\tau,i}}

155: \newcommand\Omtcx{\Omega^{\tau\bullet}}

156: 

157: \newcommand\pd{\partial}

158: \newcommand\xtl{\tilde{X}}

159: 

160: \newcommand\xbar{\overline{x}}

161: \newcommand\ybar{\overline{y}}

162: \newcommand\zbar{\overline{z}}

163: \newcommand\abar{\overline{a}}

164: \newcommand\bbar{\overline{b}}

165: \newcommand\cbar{\overline{c}}

166: \newcommand\albar{{\overline{\alpha}}}

167: \newcommand\bebar{{\overline{\beta}}}

168: \newcommand\gabar{{\overline{\gamma}}}

169: 

170: \newcommand\A{{\mathbb{A}}}

171: \newcommand\M{{\mathbb{M}}}

172: \newcommand\N{{\mathbb{N}}}

173: \newcommand\Z{{\mathbb{Z}}}

174: \newcommand\K{{\mathbb{K}}}

175: \renewcommand\L{{\mathbb{L}}}

176: \newcommand\F{{\mathbb{F}}}

177: \newcommand\Q{{\mathbb{Q}}}

178: \newcommand\R{{\mathbb{R}}}

179: \newcommand\C{{\mathbb{C}}}

180: \renewcommand\H{{\mathbb{H}}}

181: \renewcommand\P{{\mathbb{P}}}

182: \renewcommand\S{{\mathbb{S}}}

183: \newcommand\T{{\mathbb{T}}}

184: 

185: \renewcommand\a{{\mathfrak{a}}}

186: \newcommand\g{{\mathfrak{g}}}

187: \newcommand\h{{\mathfrak{h}}}

188: \renewcommand\i{{\mathfrak{i}}}

189: \renewcommand\j{{\mathfrak{j}}}

190: \renewcommand\k{{\mathfrak{k}}}

191: \renewcommand\l{{\mathfrak{l}}}

192: \newcommand\m{{\mathfrak{m}}}

193: \newcommand\n{{\mathfrak{n}}}

194: \newcommand\p{{\mathfrak{p}}}

195: \newcommand\s{{\mathfrak{s}}}

196: \renewcommand\u{{\mathfrak{u}}}

197: \renewcommand\v{{\mathfrak{v}}}

198: \newcommand\x{{\mathfrak{x}}}

199: \newcommand\z{{\mathfrak{z}}}

200: \newcommand\gl{{\mathfrak{gl}}}

201: \newcommand\so{{\mathfrak{so}}}

202: \newcommand\U{{\mathfrak{U}}}

203: \newcommand\V{{\mathfrak{V}}}

204: \newcommand\X{{\mathfrak{X}}}

205: \newcommand\Y{{\mathfrak{Y}}}

206: 

207: \newcommand\DD{{\mathfrak{D}}}

208: \newcommand\calp{{\mathfrak{p}}}

209: \newcommand\calO{{\mathcal{O}}}

210: \renewcommand\O{{\mathcal{O}}}

211: 

212: \newcommand\sym{{\mathbb{S}}}

213: 

214: \newcommand\Dun{{\underline{D}}}

215: \newcommand\dun{{\underline{d}}}

216: 

217: %% specific abbreviations

218: 

219: \newcommand\aff{\textup{Aff}}

220: \newcommand\Hom{\textup{Hom}}

221: \newcommand\Alg{\textup{Alg}}

222: \newcommand\Ext{\textup{Ext}}

223: \renewcommand\dim{\textup{dim}}

224: \newcommand\HH{\textup{H}}

225: \newcommand\hh{h}

226: \newcommand\Spec{\textup{Spec}}

227: \newcommand\BSpec{\textup{\textbf{Spec}}}

228: \newcommand\DSpec{\textup{DSpec}}

229: \newcommand\Der{\textup{Der}}

230: \newcommand\Ker{\textup{Ker}}

231: \newcommand\Mor{\textup{Mor}}

232: \newcommand\HS{\textup{HS}}

233: \newcommand\acl{\textup{acl}}

234: \newcommand\cl{\textup{cl}}

235: \dmo{\spec}{Spec}

236: 

237: 

238: \newcommand\sm{strongly minimal\,}

239: \newcommand\dring{$\DD$-ring}

240: \newcommand\drings{$\DD$-rings}

241: \newcommand\dfield{$\DD$-field}

242: \newcommand\dfields{$\DD$-fields}

243: \newcommand\dsch{$\DD$-\textup{scheme}}

244: \newcommand\prl{^{(1)}}

245: 

246: \newcommand{\GL}{\textup{GL}}

247: \newcommand{\Diff}{\textup{Diff}}

248: \newcommand{\Aff}{\textup{Aff}}

249: \newcommand{\Tr}{\textup{Tr}}

250: \newcommand\MA{{\mathfrak{A}}}

251: \newcommand\MO{{\mathcal{O}}}

252: \newcommand\ME{{\mathcal{E}}}

253: \newcommand\MF{{\mathcal{F}}}

254: \newcommand\MG{{\mathcal{G}}}

255: \newcommand\MH{{\mathcal{H}}}

256: \newcommand\MFT{{\mathcal{F}^\tau}}

257: \newcommand\MGT{{\mathcal{G}^\tau}}

258: \newcommand\tis{$\tau$-invertible sheaf}

259: \newcommand\tiss{$\tau$-invertible sheaves}

260: \newcommand\tauf{$\tau$-form}

261: \newcommand\taufs{$\tau$-forms}

262: \newcommand\PO{PO}

263: 

264: %% end abbreviations

265: 

266: \begin{document}

267: 

268: \title{Order 1 strongly minimal sets in differentially closed fields}

269: \author{Eric Rosen}

270: \address{Department of Mathematics\\

271: Massachusetts Institute of Technology\\

272: Cambridge, MA 02139}

273: \email{rosen@math.mit.edu}

274: \urladdr{http://math.mit.edu/~rosen/}

275: \thanks{This paper constitutes a revised version of part of my dissertation, at the 

276: University of Illinois at Chicago (2005).  I am grateful to Lawrence Ein,

277: Henri Gillet, David Marker, and Thomas Scanlon for helpful discussions on this material.}

278: 

279: \date{\today}

280: 

281: \begin{abstract}

282: We give a classification of non-orthogonality classes of trivial order 

283: 1 strongly minimal sets in differentially closed fields.  A central idea 

284: is the introduction of $\tau$-forms, functions on the prolongation of 

285: a variety which are analogous to 1-forms.  Order 1 strongly minimal sets then 

286: correspond to smooth projective curves with $\tau$-forms.  We also

287: formulate our results scheme-theoretically, in terms of $\tau$-differentials

288: and $\tau$-invertible sheaves on curves, thereby obtaining additional 

289: information about the strongly minimal sets.

290: This work partially generalizes and extends results of Hrushovski and Itai. 

291: \end{abstract}

292: 

293: \maketitle

294: 

295: \sec*{Introduction}

296: 

297: This paper addresses the problem of classifying the strongly minimal sets

298: definable in differentially closed fields.  A nice description of 

299: the background to the question and known results can be found in the 

300: introduction of \cite{HI},

301: so we only give a brief introduction here.  For more on the model theory of 

302: differential fields, see also \cite{MMP}, \cite{Pil02}, 

303: \cite{HS}, and \cite{HICM}.

304: The precise problem is to classify strongly minimal\ sets up to non-orthogonality, 

305: a natural logical equivalence relation.  Associated to each strongly minimal\ set

306: is a combinatorial geometry, which is either non-locally modular, locally

307: modular non-trivial, or trivial, thus dividing strongly minimal\ sets into three

308: classes.  Hrushovski and Sokolovi\'c \cite{HS} classified

309: both the non-locally modular and the locally modular non-trivial strongly minimal\ sets.

310: The former are non-orthogonal to the field of constants, which is strongly minimal,

311: and the latter correspond to (isogeny classes of) simple abelian varieties 

312: that do not descend to the constants.  What remains is to classify the 

313: trivial ones.

314: 

315: Let $K$ be a universal domain for DCF, the theory of differentially closed

316: fields, and let $k$ denote the field of constants.

317: A strongly minimal\ set, or formula, determines a unique strongly minimal\ type, and sets/formulas 

318: determining the same type are non-orthogonal, so one can equally well consider 

319: strongly minimal\ types.  Recall that if $q$ is a strongly minimal\ type over a differential field $L$, 

320: then the {\em order} of $q$ is $\textup{td}((\abar, \abar', \ldots)/L)$,

321: for $\abar$ realizing $q$, which is always finite.  

322: It seems to be unknown whether there are (trivial)

323: strongly minimal\ types of arbitrarily large order, though there are examples of order 2.

324: One consequence of \cite{HI} is that there are many types of order 1.

325: 

326: Hrushovski \cite{HJ} proved that every order 1 strongly minimal\ type that is orthogonal to $k$

327: is $\om$-{\em categorical} and thus trivial, since the classification of 

328: the non-trivial locally modular strongly minimal\ types by Manin kernels implies that

329: they are never $\om$-categorical.  

330: (Alternatively, one can derive the triviality of 

331: order 1 locally modular strongly minimal sets 

332: as a consequence of the fact that any non-trivial locally modular 

333: strongly minimal set is non-orthogonal to the Manin kernel of a simple abelian

334: variety~\cite{HS} and Buium's proof that any such Manin kernel has transcendence 

335: degree at least 2).  With Itai, Hrushovski then gives a quite precise

336: description of the trivial strongly minimal\ types over $k$, which we summarize below.

337: In this paper, we consider all order 1 strongly minimal\ types,

338: partially generalizing and extending some results of \cite{HI}.

339: 

340: \subs{Background}

341: It is well-known, and easy to see, that (complete, stationary) $n$-types in

342: pure algebraically closed fields correspond to irreducible Zariski closed 

343: subsets of $\A^n$.  Call such a set an {\em embedded affine variety},

344: as it is an affine variety together with a closed embedding into affine space.

345: We recall a similar  geometric description of complete strongly minimal\ types in DCF, 

346: which also holds more generally for types of finite rank.  

347: (See \cite{HI} or \cite{Pil02} for more details.)  

348: 

349: Say that two types $p$ and $q$ are {\em interdefinable}

350: if, perhaps passing to non-forking extensions over a common base field,

351: there is a definable bijection between their sets of realizations.  Then

352: any strongly minimal\ type $p$ in DCF is interdefinable with a type $q$ such that for

353: any realization $\abar$ of $q$, $\dl(\abar) = s(\abar)$, where $s(\xbar)$ is a

354: tuple of polynomials.  Since interdefinable types are non-orthogonal, we

355: may assume that all (strongly minimal) types are of this form.  

356: 

357: Given an affine variety $V \sub \A^n$, the map $\dl$ above, acting 

358: componentwise, is a section of the first prolongation $\tau V \sub

359: \A^{2n}$, a (possibly reducible) affine variety defined below,

360: also called the {\em shifted tangent bundle}.

361: By the above reduction, a strongly minimal\ type $p$ corresponds to a pair

362: $(V,s)$, where $V \sub \A^n$ is an embedded affine variety,

363: and $s: V \ra \tau V\sub \A^{2n}$ 

364: is a section of the {\em first prolongation}.  In this case,

365: one says that $p$ {\em lives on} the variety $V$.  It is easy to check

366: that the order of $p$ equals the dimension of $V$, so 

367: the order 1 types are exactly those that live on curves.

368: 

369: In general, a pair $(V,s)$ as above determines a Kolchin closed set

370: of finite Morley rank, $\Xi(V,s):= \{v \in V | \dl(v)=s(v) \}$.  When $V$

371: is a curve, this set is necessarily strongly minimal, so there's actually a bijection

372: between (Kolchin closed) order 1 strongly minimal sets and pairs $(C,s)$, $C$ an embedded

373: affine curve and $s$ a section of the prolongation, $s:C\ra \tau C$.

374: (Moreover, this bijection preserves fields of definition.)

375: This gives a precise, purely geometric characterization of order 1

376: strongly minimal sets.

377: What remains, though, is to determine whether or not such a 

378: strongly minimal\ set $\Xi(C,s)$ is trivial, and to understand non-orthogonality.

379: 

380: If $X$ is a strongly minimal\ set living on a curve $C$, and $C$ is birational to

381: some $C'$, then $X$ is interdefinable with some $X'$ living on $C'$.

382: Since every curve is birational to a unique smooth projective curve,

383: it suffices to consider such sets living on (embedded) smooth projective

384: curves.  Strictly speaking, then, we want to classify strongly minimal

385: sets of the form $\Xi(C,s)$, $C$ an embedded smooth projective

386: curve, $s: C \ra \tau C$.  But we will consider $C$ as an

387: abstract curve, since the particular embedding in $\P^n$ is unimportant,

388: as different embeddings yield interdefinable strongly minimal sets.

389: 

390: Given a curve $C$ (or any variety) defined over $k$, the prolongation 

391: $\tau C$ equals the tangent variety $TC$.  In this case, a section

392: of the prolongation $s: C \ra \tau C$ is just a vector field. For curves, 

393: there is also a natural bijection between vector fields and 1-forms.

394: Given a curve $C$ and a vector field $s$, let $\om$ be the 1-form such

395: that $\om (s) = 1$ almost everywhere.  Thus, order 1 strongly minimal\ sets defined over

396: the constants are also represented by pairs $(C,\om)$, defined over

397: the constants.  

398: 

399: Using this representation, Hrushovski and Itai obtain a rather complete 

400: description of  non-orthogonality classes of trivial order 1 strongly minimal\ sets 

401: defined over the constants.  To any such class they associate a unique 

402: (smooth projective) curve, the main idea being to pick a canonical 

403: {\em strictly minimal} set in any such class.  Second, they prove that for 

404: any curve of genus $\geq 2$, there are many classes associated to that curve, 

405: thereby proving that there are indeed `very many' trivial order 1 strongly minimal\ sets 

406: defined over the constants.

407: 

408: \subs{Results}

409: 

410: The original motivation for this work was to generalize results from

411: \cite{HI} to all order 1 strongly minimal\ sets.  One of the main new ideas is the 

412: introduction of $\tau$-forms, which are functions on the prolongation of

413: a curve, analogous to 1-forms, which are functions on the tangent variety.

414: This enables one to develop a `geometric approach' to order one strongly minimal\ sets

415: as in \cite{HI}, with a pair $(C,\omt)$, $C$ a curve, $\omt$ a $\tau$-form, 

416: representing a strongly minimal\ set.  We can then characterize non-orthogonality

417: classes of trivial strongly minimal\ sets in terms of `essential $\tau$-forms', 

418: appropriately defined, generalizing a central result of \cite{HI}.

419: 

420: In a separate paper~\cite{Ros}, we introduce and develop the theory of $\tau$-differentials,

421: the algebraic analog of $\tau$-forms. We use this material here to give an alternative

422: characterization of order one strongly minimal sets in terms of curves with

423: $\tau$-invertible sheaves.  

424: This makes it possible to determine the dimension of the space of global 

425: $\tau$-forms on a curve, in terms of its Kodaira-Spencer class.  

426: Further, we show that the set of global $\tau$-forms on a curve is (uniformly) definable,

427: as is the dimension of this set.

428: It also leads to the introduction of the prolongation cone of a curve, a vector bundle

429: into which both the tangent variety and prolongation naturally embed.

430: 

431: \subs{Open questions}

432: 

433: The main question left open by this work is whether for every curve $C$

434: of genus $>1$ there are $\sim$-essential \taufs.  For curves defined over

435: $k$, a positive answer follows from \cite{HI}.  A positive

436: answer for all curves would show that there are order 1 trivial strongly minimal\ sets 

437: that are orthogonal to any such set defined over $k$.  

438: %It also remains to 

439: %look more carefully at strongly minimal\ sets on curves of genus 0 or 1.  When everything

440: %is defined over $k$, then the situation is well-understood, but the general

441: %case may be more complex, especially for elliptic curves.

442: 

443: Two main problems about trivial strongly minimal\ sets in DCFs, asked by

444: Hrushovski, remain completely open.  Are they all $\om$-categorical?

445: And can they be classified?  Extending the classification beyond order 1

446: certainly seems difficult.  For example, presumably one would need to 

447: be able to determine whether a finite rank type given as a pair $(V,s)$ 

448: is strongly minimal.  But perhaps there are particular cases of order 2 types,

449: thus living on surfaces, that are more accessible.  

450: There are also related questions, which might be approached

451: by existing methods.  Hrushovski suggests that one could try to show

452: that the solutions to a generic order 2 equation are strongly minimal, 

453: to answer a question of Poizat.  

454: Another well-known question is whether every U-Rank 1 type is strongly minimal.  

455: This is related to Hrushovski and Scanlon's result~\cite{HrSc} 

456: that U-Rank and Morley rank differ in DCF.

457: 

458: \sec{Prolongations}

459: \label{prolong}

460: We recall the construction of the prolongation of a variety over a differentially

461: closed field.  Buium's original definition uses the language of schemes, 

462: though we give a description in local coordinates used more often by

463: model-theorists.  For a variety $V$ defined over the constants, its prolongation

464: is just the tangent variety $TV$.  In general, though, it is a $TV$-{\em torsor}.

465: Thus it is also called the {\em shifted tangent bundle}.  

466: In this case, the fibers of the canonical projection to $V$ are no longer

467: vector spaces, but affine spaces, defined below.

468: 

469: \para{Affine Spaces}  

470: An affine space is essentially a vector space without a 

471: distinguished point as origin.  Affine spaces arise in model theory as combinatorial

472: geometries that are locally modular but not modular.   Just as a locally

473: modular geometry becomes modular when one fixes any point, choosing a point in 

474: an affine space naturally gives a vector space.  The presentation here

475: differs from than that in, for example, Hodges' textbook~\cite{Hod}, 

476: but is basically 

477: equivalent.  (The word ``affine'' will also be used in a 

478: completely different sense, in connection with affine varieties.)

479: 

480: \bdf

481: \label{aff}

482: Let $K$ be a field.  A {\em K-affine space} is a triple $(A, V, \alpha)$,

483: where $A$ is a set, $V$ is a $K$-vector space, and $\alpha$ is a regular

484: action of $V$ on $A$. The {\em dimension} of $A$, $\dim(A)$, is just $\dim(V)$.

485: \edf

486: 

487: We will write $\alpha(v,a)$ as $v \cdot a$, and often omit the action $\alpha$ 

488: when it is understood.

489: 

490: \brm

491: \label{affbasics}

492: \ben

493: \item  For any $K$-vector space $V$, $(V, V)$ is an affine space in a natural way.

494: \item  Given $(A, V)$ and $a \in A$, there is a natural bijection $i_a:V\lra A$

495: given by $i_a(v) = v\cdot a$.

496: \een

497: \erm

498: 

499: \bdf

500: An {\em affine map} between $K$-affine spaces $(A, V, \alpha)$ and $(B, W, \beta)$

501: is a map $f:A \lra B$ such that there is a linear map $\lambda f: V \lra W$ so that 

502: for all $a\in A, v \in V$, $f(v\cdot a) = (\lambda f (v))\cdot f(a)$.

503: Let $\Aff(A,B)$ denote the set of all affine maps from $A$ to $B$.

504: (By putting coordinates on $A$ and $B$, this set can be endowed with a vector

505: space structure.)

506: 

507: \edf

508:    

509: \brm

510: \label{affrems}

511: \ben

512: \item  Given an affine map $f$, the linear map $\lambda f$ is uniquely determined.

513: In fact, $\lambda$ is a functor from the category of $K$-affine spaces to the 

514: category of $K$-vector spaces.

515: \item Given affine maps $f, g$, from $(A, V)$ to $(B, W)$, $\lambda f = \lambda g$ 

516: if and only if there is a $w \in W$ such that for all $a \in A$, $f(a) = w \cdot g(a)$.

517: \item There is a short exact sequence 

518: %\note{[?!]}

519: $$

520: 0 \lra B \stackrel{\mu}{\lra}\aff(A, B)\stackrel{\lambda}{\lra}\Hom(V, W) \lra 0

521: $$

522: where for all $b_0 \in B, \mu(b_0)$ is the constant affine map 

523: $\mu(b_0): A \lra B$ mapping each $a \in A$ to $b_0$.  Thus $\dim(\aff(A, B)) = 

524: \dim(B) + \dim(A)\dim(B)$.

525: \een

526: \erm

527: 

528: We are particularly interested in the case, $W = K$, i.e., when $B$ is 

529: 1-dimensional.

530: Then $\aff(A, B)$ is something like a dual space to $(A, V)$, though 

531: $\dim(\aff(A,B))=\dim(A) + 1$.    

532: 

533: \brm

534: Let $(A,V)$ be an affine space.  Recall that the {\em affine group} of $V$,

535: denoted $\Aff(V)$, is the group generated by the translation group $\Tr(V)$

536: and $\GL(V)$.  In fact, $\Aff(V)=\Tr(V)\rtimes \GL(V)$.

537: %%change \bowtie to \rtimes

538: The automorphism group of $(A,V)$ is naturally isomorphic to $\Aff(V)$.

539: \erm

540: 

541: \para{Varieties, tangent spaces, and prolongations}

542: A prevariety is a topological space with an open cover,

543: $W = W_1 \cup \ldots \cup W_m$ and compatible coordinate charts, $f_i:W_i \lra V_i$,

544: $V_i$ an affine variety.  For all $i, j  \leq m$, let $U_{i, j} = f_i(W_i \cap 

545: W_j) \sub V_i$ and $f_{i, j}: U_{i, j} \lra U_{j, i}$ be $f_j\circ f_i^{-1}$.

546: For our purposes, a variety will be an irreducible, smooth separated prevariety.

547: 

548: Recall that, given a polynomial $p(X)$, 

549: $p^\delta(X)$ denotes the polynomial obtained by applying $\delta$ to

550: each of the coefficients.  For more information, see 

551: Marker~\cite{Mar03} or Pillay~\cite{Pil02}.  The following easy observation 

552: will be useful.

553: 

554: \bla

555: For any $n$, let $\epsilon: K[x_1, \ldots , x_n] \lra K[x_1, \ldots , x_n]$

556: be the map that takes any polynomial $f$ to $f^\dl$, which is obtained 

557: from $f$ by taking the derivative of each coefficient.  Then $\epsilon$

558: is a derivation on $K[x_1, \ldots , x_n]$.  Further, $\epsilon$ commutes

559: with each derivation $\frac{d}{dx_i}$. 

560: \ela

561: 

562: \bpr

563: Straightforward.

564: \epr

565: 

566: \bdf

567: Let $V \subseteq K^n$ be an irreducible smooth affine variety, 

568: $I(V) = \<p_1,\ldots , p_m\>$.  The tangent space of $V$ is

569: $$

570: TV = \{(a, u) \in K^{2n} : \sum_{i=1}^n\frac{\partial p_j}{\partial x_i}(a) 

571: \cdot

572: u_i = 0, j = 1, \ldots , m\}

573: $$

574: The first prolongation of $V$ is 

575: $$

576: \tau V = \{(a, u) \in K^{2n} : \sum_{i=1}^n\frac{\partial p_j}{\partial x_i}(a) 

577: \cdot u_i  + p_j^\delta(a)= 0, j = 1, \ldots , m\}

578: $$

579: 

580: Since everything is functorial, one can also define $TV$ and $\tau V$

581: for general varieties, in a coordinate free manner.

582: \edf

583: 

584: \brm

585: There are natural projection maps $\pi_T: TV \lra V$ and $\pi_\tau: \tau V \lra

586: V$.  For smooth $V$, and $v \in V$, $\pi_T^{-1}(v)=TV_a$ is a $\dim(V)$-dimensional

587: vector space, and $\pi_\tau^{-1}(v)=\tau (V)_a$ is a $\dim(V)$-dimensional 

588: affine space naturally acted on by $\pi_T^{-1}(v)$.

589: 

590: Further, let $TV \times_V\tau V = \{(a, u, w) \in K^{3n} : (a, u) \in TV, (a, w) \in 

591: \tau V\}$.  The map $p: TV \times_V\tau V \lra \tau V$ given by

592: $p(a, u, w) = p(a, u+w)$ makes $\tau V$ a {\em torsor} under $TV$.  

593: \erm

594: 

595: We now introduce a new construction, the prolongation cone of a variety,

596: into which both the tangent variety and prolongation naturally embed.  For more

597: details, see~\cite{Ros}.

598: 

599: \bdf

600: Let $V \subseteq K^n$ be an irreducible smooth affine variety, 

601: $I(V) = \<p_1,\ldots , p_m\>$.  The {\em prolongation cone} of $V$ is

602: $$

603: \tilde{V} = \{(a, u, u') \in K^{2n+1} : \sum_{i=1}^n\frac{\partial p_j}{\partial x_i}(a) 

604: \cdot

605: u_i + p_j^\delta(a)\cdot u'= 0, j = 1, \ldots , m\}

606: $$

607: 

608: As in the case of the tangent variety, there is a  natural projection map

609: $\tilde{\pi}:\tilde{V} \lra V$, whose fibres are vector spaces, now of dimension

610: $\dim (V) + 1$.  More generally, the prolongation cone is a $\dim(V) + 1$ vector

611: bundle over $V$.  Again, the construction also globalizes to general varieties.

612: 

613: Observe that the intersection of the prolongation cone with the hyperplane $u' = 0$

614: is the tangent variety.  Likewise, the intersection with the hyperplane $u' = 1$

615: is the prolongation.

616: \edf

617: 

618: \brm  

619: The tangent space $TV$ is a variety with `additional linear structure' on the fibers 

620: $TV_v$.  We recall

621: how to make this notion precise and indicate how to make rigorous the 

622: notion that the fibers $\tau V_v$ of the prolongation are affine spaces.

623: 

624: One can define a vector bundle as follows (see \cite{GHL}, p.\ 15).

625: Let $E$ and $V$ be smooth varieties over a field $K$, $\pi:E\lra V$ a regular map.

626: We say that $(\pi , E, V)$ is a {\em vector bundle of rank n} if the following

627: conditions hold.

628: \ben

629: \item  $\pi$ is surjective.

630: \item  There exists a finite open cover $(U_i)_{i\in I}$ of $V$, and

631:   isomorphisms $h_i:\pi^{-1}(U_i)\lra U_i\times K^n$ such that for any

632:   $x\in U_i$, $h_i(\pi^{-1}(x))=\{x\}\times K^n$.

633: \item  For any $i,j \in I$, there is a regular map $g_{ij}:U_i\cap U_j \lra

634: \GL_n(K)$ such that the map

635: $$

636: h_i \circ h_j^{-1}:(U_i\cap U_j) \times K^n \lra (U_i\cap U_j) \times K^n 

637: $$

638: is of the form $h_i\circ h_j^{-1}(x,v)=(x,g_{ij}(x)\cdot v)$.

639: \een

640:   

641: $V\times K^n$ is the {\em trivial vector bundle} over $V$, and the definition 

642: of a vector bundle says that it is {\em locally trivial}.  Likewise, the 

643: prolongation $\tau V$ is a locally trivial affine bundle, in a similar sense.  

644: That is, let $V$

645: be a variety and $(K^n,K^n)$ $n$-dimensional affine space.  Then 

646: $(\pi , E, V)$ as above is a {\em trivial affine bundle of rank n} if it is 

647: isomorphic to $V\times K^n$, where we consider $K^n$ as an affine, rather than

648: vector, space.  More generally, $(\pi , E, V)$ as above is

649: an {\em affine bundle of rank n} if we have the following.

650: \ben

651: \item  $\pi$ is surjective.

652: \item  There exists a finite open cover $(U_i)_{i\in I}$ of $V$, and

653:   isomorphisms $h_i:\pi^{-1}(U_i)\lra U_i\times K^n$ such that for any

654:   $x\in U_i$, $h_i(\pi^{-1}(x))=\{x\}\times K^n$.

655: \item  For any $i,j \in I$, there is a map $g_{ij}:U_i\cap U_j \lra

656: \Aff(K^n)$ such that the map

657: $$

658: h_i \circ h_j^{-1}:(U_i\cap U_j) \times K^n \lra (U_i\cap U_j) \times K^n 

659: $$

660: is of the form $h_i\circ h_j^{-1}(x,v)=(x,g_{ij}(x)\cdot v)$.

661: \een

662: 

663: \erm

664: 

665: \brm 

666: Another way to formalize the notion of an affine bundle is suggested by

667: the notion of a fiber bundle associated to a principal bundle

668: (See \cite{KN96}, p.\ 50--55.  Compare also section III.3 on 

669: Affine Connections, p.\ 125.)  

670: Roughly, given a manifold $M$ and a Lie group $G$,

671: a principal $G$-bundle $P(G,M)$ is a fiber bundle 

672: $\pi:E\lra M$, with a free $G$-action

673: on each fiber $\pi^{-1}(x), x\in M$.  If $F$ is some other manifold, with 

674: a $G$-action, then one can construct a fiber bundle $E(G,M,F,P)$ which is

675: the fiber bundle over $M$, with standard fiber $F$, and structure group

676: $G$, associated to the principal fiber bundle $P$.  In our setting, $M$ is our 

677: variety $V$, $G$ is $\Aff(K^n)$, and $F$ is our affine space.

678: \erm

679: 

680: The following lemma is due to Buium \cite{Bui93}.

681: \bla

682: Let $V$ be a smooth variety of dimension $n$.  Then $\tau V$ is an affine bundle

683: of rank $n$.

684: \ela

685: 

686: \para{Tangent and lifting maps}

687: We now consider maps between varieties.  Let $V,W$ be varieties,

688: and $\phi:V \ra W$ a regular map.  The map $\phi$ determines a map

689: from $TV$ to $TW$, written $T\phi$,  the {\em tangent map} of $\phi$,

690: which restricts, for each $a \in V$,

691: to a linear map on fibers, $T\phi_a:TV_a \ra TW_{f(a)}$.

692: 

693: An important special case occurs when $f$ is a regular function on $V$,

694: viewed as a map from $V$ to $\A^1$.  We consider the {\em differential},

695: $df$, which is the composition $df = \pi\circ Tf$, where 

696: $\pi$ is the projection from $T\A^1 \ra \A^1$ onto the {\em tangent vector

697: component}.  Thus, $df$ is a regular function on $TV$, and we have a

698: $K$-linear derivation $d:K[V]\ra K[TV]$.  Alternatively, $d$ is a

699: derivation from $K[V]$ to $\Om[V]$, the regular differential forms

700: on $V$, which is a $K[V]$-module with a natural embedding in $K[TV]$.

701: 

702: 

703: For prolongations, there is also a 

704: {\em lifting map} from $\tau V$ to $\tau W$, 

705: which we write $\phi^{(1)}$, which restricts, for each $a\in V$, to an 

706: affine map on fibers, $\phi^{(1)}_a:\tau V_a \ra \tau W_{f(a)}$.

707: For affine varieties, $V \sub K^n, W \sub K^m$, the map $\phi^{(1)}$

708: is given by $\phi^{(1)}_a((a,u)) = (\phi(a),d\phi_a(u)+\phi^\dl(a))$.

709: 

710: When $f$ is a regular function on $V$, composing $f^{(1)}$ with 

711: $\pi$, as above, one gets a map $\tau f = \pi\circ f^{(1)}$,

712: which we call a {\em $\tau$-differential}.

713: Below, Lemma~\ref{twder},

714: we will see that $\tau:K[V]\ra K[\tau V]$ is a derivation (in fact,

715: a $\tau$-derivation, as defined below.

716: 

717: \brms

718: \ben

719: \item  When the derivation $\dl$ on $K$ is trivial, then for any variety

720: $V$, $\tau V = TV$.  Likewise, given any regular map between varieties,

721: $f:V \ra W$, $f\prl= Tf$.  More generally, this holds true over an

722: arbitrary differential field if everything is defined over the field of 

723: constants.

724: 

725: \item  In \cite{Ros}, Proposition 3.15, it is shown that that the map 

726: $\tau$ coincides with something introduced by Buium, 

727: in a different context. 

728: \een

729: \erms

730: 

731: The next lemma follows from the description, above, of the 

732: lifting map in local coordinates.

733: 

734: \bla

735: \label{taudiff}

736: Let $V$ be a variety, $f$ a regular function on $V$.  For each 

737: $a\in V$, $\lm(\tau f_a)=df_a$.

738: \ela

739: 

740: %One can define the tangent space and prolongation of a variety by patching.

741: %Also, what about other charts? \ldots

742: 

743: \sec{$\tau$-forms}

744: \label{tauforms}  

745: Our main idea is to look at an analog of 1-forms on $\tau V$, which will be functions

746: $f:\tau V \lra K$ such that on each fiber of $\pi_\tau : \tau V \lra V$, $f$ is 

747: an affine map, as defined above.

748: We recall the definition of 1-forms, following the presentation of 

749: Shafarevich~\cite{Sha}.  We then use this formalism

750: to introduce $\tau$-forms.   

751: For an alternative, more general treatment of this material, 

752: in the language of commutative algebra and scheme theory, see \cite{Ros}.

753: 

754: Let $V$ be a variety, and let $\Phi [V]$ be the set of all functions 

755: $\phi$ mapping each point $v \in V$ to a linear map 

756: $\phi (v): TV_v\lra K$.

757: Note that $\Phi [V]$ naturally forms a (very large) $K[V]$-module. 

758: Given $f \in K[V]$, the differential $df$ is a function in $\Phi [V]$.

759: One could look at the submodule of $\Phi[V]$ generated by

760: $\{df: f \in K[V]\}$, but this is somewhat too small.

761: Instead, we say that an element $\phi \in \Phi [V]$ is a regular differential

762: 1-form on $V$ if every $v\in V$ has a neighborhood $U$ such that the 

763: restriction of $\phi$ to $U$ belongs to the $K[U]$-submodule of 

764: $\Phi[U]$ generated by the elements $df, f \in K[U]$.

765: 

766: The regular differential 1-forms on $V$ form a $K[V]$-module, 

767: denoted $\Om [V]$.

768: 

769: \bla

770: The map $d: K[V]\lra\Om[V]$ satisfies 

771: \ben

772: \item  $dc=0$ for all $c \in K$.

773: \item  $d(f + g) = df+dg$

774: \item  $d(fg)= fdg + gdf$.

775: \een

776: \ela

777: 

778: \bpro

779: Every $v\in V$ has an affine neighborhood $U$ such that 

780: $\Om [U]$ is a free $K[U]$-module of rank $\dim (V)$.

781: \epro

782: 

783: We now introduce $\tau$-forms, imitating this construction.

784: Let $\Psi[V]$ be the set of all functions $\psi$ mapping each 

785: $v \in V$ to an affine map $\psi(v): \tau V_v\lra K$.

786: As above, $\Psi [V]$ forms a $K[V]$-module and,

787: given $f \in K[V]$, the $\tau$-differential $\tau f$ is in $\Psi [V]$.  

788:   

789: \bdf

790: We say that $\psi\in \Psi[V]$ is a {\em regular $\tau$-form} if every $v\in V$ 

791: has a neighborhood $U$ such that the restriction of $\psi$ to $U$ belongs 

792: to the $K[U]$-submodule of $\Psi[U]$ generated by the elements

793: $\tau f,f \in K[U]$.  

794: 

795: The regular $\tau$-forms on $V$ form a $K[V]$-module, 

796: denoted $\Omt [V]$.

797: \edf

798: 

799: Note that for any variety $V$, \tauf\ $\omt \in \Omt [V]$,

800: and $v \in V$, there is an open neighborhood $U$ of $v$ such 

801: that locally, on $U$,

802: $$

803: \omt = (\sum_{i=1}^ng_i\tau f_i)

804: $$

805: for $g_i,f_i\in K[U]$.

806: 

807: \brm

808: Given a variety $V$, note that a differential form on $V$ is a 

809: regular function on $TV$.  Thus the module $\Om[V]$ embeds

810: naturally in $K[TV]$, and the differential map is a $K$-linear derivation,

811: $d: K[V]\ra K[TV]$.  Likewise, a $\tau$-form on $V$ is a regular

812: function on $\tau V$.  By Lemma~\ref{twder}, below, $\tau$ is actually

813: a derivation.

814: \erm

815: 

816: \brm

817: \label{iotaiota}

818: From the definition, there is a map $\tau: K[V]\lra\Omt[V]$, much

819: like the derivation map $d: K[V]\lra\Om[V]$ to 1-forms.  

820: But for \taufs\ there is also a natural embedding of $K[V]$ into

821: $\Omt[V]$, which we now describe.

822: Because $K$ is differentially closed, there is a $c\in K$ 

823: such that $\dl(c)=1$.  So $c\in K[V]$ and $\tau c\in \Omt[V]$ is

824: the constant function on $\tau V$ with value 1.  By the definition

825: of $\Omt[V]$, for each $f\in K[V], f\tau c \in \Omt[V]$, where

826: $f\tau c$ is constant on each fiber $\tau V_v$, with value $f(v)$.

827: One sees immediately that the map $\iota:K[V]\lra \Omt [V]$, given by

828: $\iota(f)=f\tau c$, is an embedding (and is independent of the choice

829: of $c$).  Occasionally, given $f \in K[V]$, we will also write $f \in \Omt[V]$, 

830: where to be more precise we mean that $\iota f\in  \Omt[V]$.

831: 

832: Call a \tauf\ $\omt$ {\em trivial} if it equals $\iota f$, for some $f\in K[V]$.

833: %because it is a constant map on each fiber $\tau V_v, v \in V$.

834: (Clearly, a \tauf\ is trivial if and only if it is trivial

835: on some non-empty open subset.)  Further, given $e\in K \sub K[V]$,

836: call $\iota e$ a {\em constant trivial} $\tau$-form.

837: \erm

838: 

839: \bpro

840: \label{freemod}

841: Every $v\in V$ has an (affine) neighborhood $U$ such that 

842: $\Omt [U]$ is a free $K[U]$-module of rank $\dim (V) + 1$.

843: \epro

844: 

845: \prf 

846: %\textsf{(There must also be a more direct argument.  One should be

847: %able to imitate the proof that $\Om[V]$ is a locally free module,

848: %rather than using this fact in the proof.)}

849: %

850: We know that $v\in V$ has a neighborhood $U_1$ such that 

851: $\Om[U_1]$ is a free $K[U_1]$-module of rank $\dim (V)$.

852: Clearly, there is a neighborhood $U_2$ of $v$ such that there is a 

853: regular section $E_2$ of $\tau U_2$.  In other words, there is 

854: a map $\phi_2: U_2 \lra \tau U_2$ that is an isomorphism from 

855: $U_2$ to $E_2$, whose inverse is the projection map 

856: $\pi_\tau|_{U_2}$.  Let $U=U_1\cap U_2$, $E=E_2\cap\tau U$,

857: and $\phi =\phi_2|_U$, so that $E$ is a section of $U$.

858: 

859: $\tau U$ is a $TU$-torsor, and the section $E$ gives us an

860: isomorphism $\phi'$ from $TU$ to $\tau U$, which maps the 

861: 0-section of $TU$ to $E$.  Given $u\in TU_a$,

862: $$

863: \phi'(u)=u\cdot\phi(\pi_T(u))

864: $$

865: where $u\cdot\phi(\pi_T(u))$ denotes the action of $TU_a$

866: on $\tau U_a$.

867: (On each affine fiber, $(\tau V_a, TV_a)$, this is the 

868: bijection from $TV_a$ to $\tau V_a$ that one gets by

869: fixing the point $\phi(a) \in \tau V$, as in Remark~\ref{affbasics}.2.)

870: 

871: The isomorphism $\phi'$ determines a bijection $\Phi'$ between 

872: $\Om[U]$ and the set $\Psi\sub\Omt[U]$ of \taufs\ on $U$ 

873: that take the value 0 on all of $E$.  Given $\om\in\Om[U]$, let

874: $\Phi'(\om)$ be the \tauf\ on $U$ such that for $u \in \tau U$,

875: $$

876: \Phi'(\om)(u) = \om(\phi'^{-1}(u)).

877: $$

878: In fact, this bijection is an isomorphism of $K[U]$-modules.

879: 

880: We want to show that $\Omt[U]=\Psi \oplus \iota K[U]$.

881: Let $\omt\in\Omt[U]$, and let $g\in K[U]$ be the function

882: $g(a)=\omt(\phi(a))$.  Then $\omt_0=\omt -\iota g$ is in 

883: $\Psi$, which implies that $\Omt[U]=\Psi + \iota K[U]$.

884: 

885: Finally, observe that $\Psi + \iota K[U]$ is a free direct

886: sum $\Psi \oplus \iota K[U]$, making $\Omt[U]$ into a free $K[U]$-module of dimension

887: $\dim(V)+1$.  For suppose that there is an $\omt \in \Psi,

888: \omt=\Phi'(\om)$, $\om\in\Om[U]$, and a $g\in K[U]$ such that 

889: $\omt=\iota g$.  Then $\omt$ is constant on each affine $\tau$-fiber,

890: so $\om$ must be the trivial 1-form, and $\omt$ must be identically

891: zero.  Thus, $g$ is also identically zero, as desired.

892: \qed

893: 

894: \msk

895: 

896: The following lemma is suggestive.

897: 

898: \bla

899: \label{twder}

900: The map $\tau: K[V]\lra\Omt[V]$ satisfies 

901: \ben

902: \item  $\tau c=\dl c$ for all $c \in K$.

903: \item  $\tau(f + g) = \tau f + \tau g$

904: \item  $\tau(fg)= f\tau g + g\tau f$.

905: \een

906: \ela

907: 

908: \prf  Conditions 1.\ and 2.\ are immediate from the definitions.

909: To prove 3., we show first that 

910: $(fg)^\delta = f^\delta g + g^\delta f$.  To simplify notation,

911: we assume that $f,g$ are polynomials in one variable.  Let

912: $f(x)=\sum_{i=0}^ma_ix^i$ and  $g(x) = \sum_{j=0}^nb_jx^j$.  Then

913: $$

914: \begin{array}{lll}

915: (fg)^\delta&=&\sum_{k=0}^{m+n}\sum_{i=0}^k(a_i^\delta b_{k-i} + 

916: a_ib^\delta_{k-i})x^k\\\\

917: &=&\sum_{k=0}^{m+n}\sum_{i=0}^ka_i^\delta b_{k-i}x^k +

918: \sum_{k=0}^{m+n}\sum_{i=0}^ka_ib^\delta_{k-i}x^k\\\\

919: &=&f^\delta g + fg^\delta

920: \end{array}

921: $$

922: as desired.

923: 

924: Then for $(a, u) \in \tau (V)$, 

925: $$

926: \begin{array}{lll}

927: \tau(fg)(a,u) &=& d(fg)_a\cdot u + (fg)^\delta(a)\\\\

928: &=& f(a)dg_a\cdot u + g(a)df_a\cdot u + f(a)g^\delta(a)+g(a)f^\delta(a)

929: \\\\

930: &=&f(a)\tau g(a,u) + g(a)\tau f(a,u)

931: \end{array}

932: $$  

933: which completes the proof.

934: \qed

935: 

936: \brm

937: This lemma says that the map $\tau: K[V]\lra\Omt[V]$ is a {\em $\tau$-derivation},

938: that is, a derivation such that for all $a,b \in K$, $\dl(a)\tau(b) = \dl(b)\tau(a)$.

939: In fact, this map is the fundamental example of such a derivation.  

940: For more information, see~\cite{Ros}.

941: \erm

942: 

943: \bco

944: Let $V$ be a variety.  For any polynomial $F\in K[T_1,\ldots,T_m]$,

945: and functions $f_1,\ldots,f_m\in K[V]$,

946: $$

947: \begin{array}{lll}

948: \tau(F(f_1,\ldots,f_m))&=&\sum_{i=1}^m\frac{\partial F}{\partial T_i}

949: (f_1,\ldots,f_m)\tau f_i + F^\delta(f_1,\ldots,f_m)

950: \end{array}

951: $$

952: \eco

953: \prf

954: It suffices to consider $F(T_1,\ldots,T_m)=

955: cT_1^{a_1}\cdots T_m^{a_m}$ a monomial.  Then

956: $$

957: \begin{array}{lll}

958: \tau (cf_1^{a_1}\cdots f_m^{a_m}) &=&

959: \sum_{i=1}^m (a_i f_i^{a_i-1}\tau f_i(\prod_{j\neq i}f_j^{a_j}))+

960: (\tau c)(f_1^{a_1}\cdots f_m^{a_m})

961: \\\\

962: &=&\sum_{i=1}^m\frac{\partial F}{\partial T_i}

963: (f_1,\ldots,f_m) \tau f_i+ F^\delta(f_1,\ldots,f_m)

964: \end{array}

965: $$

966: as desired.

967: \qed

968: 

969: \brm      

970: We now define rational 1-forms and $\tau$-forms on $V$.

971: Notice first that the above definition actually gives a sheaf

972: of modules of 1-forms and $\tau$-forms.  Given any 

973: open set $U \sub V$, one can define $\Om [U]$ and $\Omt [U]$

974: as above, and for open subsets $U \sub W \sub V$,

975: there are natural restriction maps $\rho_{W,U}:\Om [W] \lra \Om [U]$

976: and $\rho_{W,U}^\tau:\Omt [W] \lra \Omt [U]$.

977: Define an equivalence relation on 1-forms,

978: where $\om_1 \in \Om [U_1]$ and $\om_2 \in \Om [U_2]$, $U_i$ non-empty,

979: are equivalent if they agree on $U_1 \cap U_2$ (or on any open set).

980: A rational 1-form is an equivalence class under this relation,

981: and $\Om (V)$ denotes the set of rational 1-forms.

982: One can then easily define the domain of a rational 1-form.

983: Recall that $\Om (V)$ is a $\dim(V)$-dimensional vector space

984: over $K(V)$, the field of rational function on $V$.

985:   

986: The rational $\tau$-forms $\Omt (V)$ are defined in exactly the same way

987: and also form a $K(V)$-vector space.  

988: As with regular \taufs, there is a natural embedding

989: $\iota: K(V)\lra\Omt(V)$.  

990: \erm

991: 

992: The next result follows immediately from Proposition~\ref{freemod}.

993: 

994: \bpro    

995: \label{taurat}

996: Given an $n$-dimensional variety $V$, $\Omt(V)$ is an 

997: $(n+1)$-dimensional $K(V)$-vector space.

998: \epro

999: 

1000: From now on, by 1-form or $\tau$-form we will mean rational

1001: 1-form or $\tau$-form.  Following Hrushovski-Itai~\cite{HI},

1002: regular forms will be referred to as {\em global forms}.

1003: 

1004: \brm

1005: In the language of schemes, we can express Proposition~\ref{freemod} by

1006: saying that, given a smooth variety $V$, $\Omt_V$ is a locally free

1007: sheaf of dimension $\dim(V) + 1$.  In fact, elements of $\Omt_V$ are

1008: exactly the rational functions on the prolongation cone $\tilde{V}$ that

1009: are linear on each fiber.  Equivalently, they are sections of the dual

1010: bundle of $\tilde{V}$.  This situation is analogous to the fact that 

1011: 1-forms are functions on the tangent bundle and sections of the cotangent

1012: bundle.

1013: \erm

1014: 

1015: \subs{The $\Lm$ map}

1016: We describe a functor $\Lambda$ from $\tau$-forms to 1-forms, that

1017: is a precise analog of the functor $\lambda$ from affine spaces to 

1018: vector spaces, defined above.

1019: 

1020: \bla

1021: \label{lambda}

1022: Let $V$ be an irreducible smooth variety.  There is a natural map

1023: $\Lambda_V: \Omega^\tau (V) \lra \Omega (V)$, such that for each 

1024: $\omega^\tau \in \Omega^\tau(V)$, and $v\in V$,

1025: 

1026: $$

1027: \Lambda_V(\omega^\tau)_v = \lambda(\omega^\tau_v)

1028: $$

1029: where $\lambda(\omega^\tau_v): TV_v\lra K$ is the linear map

1030: associated to the affine map $\omega^\tau_v: \tau V_v \lra K$.

1031: 

1032: Given $f \in K(V)$ and $\tau f\in\Omt(V)$,

1033: then $\Lm_V(\tau )=df\in\Om(V)$.

1034: \ela

1035: 

1036: \prf

1037: Fix $\omt\in\Omt(V)$, and let $\om =\Lm_V(\omt)$. By the definition of $\Lm_V$, 

1038: for each $v\in V$, $\om_v$ is a linear map on $TV_v$, so one must show

1039: that these linear maps vary smoothly on $V$.  It suffices to check

1040: this locally.  

1041: 

1042: Let $v\in V$, and choose an open neighborhood $U\sub V$ of $v$, such

1043: that on $U$, $\omt$ is given by 

1044: $\sum_{i=1}^ng_idf_i + \iota h$, $f_i,g_i,h \in K[U]$.

1045: Then $\om|_U=\sum_{i=1}^ng_idf_i$, as desired.

1046: 

1047: The last assertion follows from Lemma~\ref{taudiff}.

1048: \qed

1049:   

1050: \msk

1051: 

1052: Notice that $\Lambda_V: \Omega^\tau(V) \lra \Omega(V)$ is clearly surjective.

1053: The map $\Lambda_V$ restricts to a map from global $\tau$-forms to global

1054: 1-forms, but this map is in general {\em not} surjective.

1055: Below, the remarks preceding Proposition~\ref{cohom} provide more information

1056: about this restricted map.

1057: 

1058: Observe also that the composition of $\iota: K(V) \lra \Omt(V)$ with

1059: $\Lambda_V: \Omt(V)$ is trivial, i.e., $\Lambda_V \circ \iota$ is the zero

1060: map.  As $\iota$ is injective, and $\dim(\Omt(V)) = \dim(\Om(V)) + 

1061: \dim (K(V)) = n + 1$, as $K(V)$-vector spaces, one has the following

1062: proposition.

1063: 

1064: \bpro

1065: For any variety $V$, there is a short exact sequence

1066: $$

1067: 0 \lra K(V) \lra \Omt(V) \lra \Om(V) \lra 0.

1068: $$

1069: \epro

1070: 

1071: A description of this sequence, 

1072: reformulated in terms of locally free sheaves on $V$

1073: is given in~\cite{Ros}, where it is shown that, as an extension of

1074: $\Om(V)$ by $K(V)$, it corresponds to the Kodaira-Spencer class of $V$.

1075: 

1076: We now describe how to pullback \taufs.  Recall that, given a 

1077: rational map $\phi:V\lra W$ between varieties, there is a pullback map

1078: $\phi^*:\Om(W)\lra\Om(V)$ that can be defined as follows.

1079: Let $\om\in\Om(W)$, $v\in V$ and $y=\phi(v)$.  Then for all

1080: $a\in TV_v$, $\phi^*\om_v(a)=\om_y(T\phi(a))$.

1081: 

1082: \bdf

1083: Let $\phi:V\lra W$ be a rational map between varieties.  For each

1084: $\omt\in\Omt(W)$, we define the {\em pullback} of $\omt$ by $\phi$, written

1085: $\phi^{\tau*}\omt$, to be the \tauf\ on $V$, given as follows.

1086: For $v\in V, y=\phi(v)$, and $a\in\tau V_v$, then

1087: $\phi^{\tau*}\omt(a)= \omt_y(\phi\prl(a))$.

1088: \edf

1089: 

1090: \bla

1091: \label{comm}

1092: Let $V, W$ be varieties, $\phi: V \lra W$ a morphism.  The following diagram is 

1093: commutative.

1094: 

1095: $$

1096: \xymatrix{

1097: \Omt (V) \ar[d]^{\Lm_V} &\Omt (W) \ar[l]_{\phi^{\tau *}} \ar[d]^{\Lm_W}\\

1098: \Om (V) &\Om (W)\ar[l]_{\phi^*}

1099: }

1100: $$

1101: \ela

1102: 

1103: \prf It suffices to check this on fibers $\tau V_a$ and $TV_a$, $a \in V$,

1104: $\tau W_{\phi(a)}$ and $TW_{\phi(a)}$, $\phi(a) \in W$.

1105: We have $\phi\prl_a: \tau(V)_a \lra \tau W_{\phi(a)}$ and $T\phi_a:

1106: TV_a \lra TW_{\phi (a)}$; notice that $\lambda(\tau\phi_{a})=d\phi_{(a)}$.

1107: Then $(\phi^{\tau *}\omega^\tau)_a = \omega^\tau\circ\phi\prl_a$ and 

1108: $\lambda(\phi^{\tau *}\omega^\tau)_a)=\lambda(\omega^\tau)\circ\lambda(\tau\phi_a)$.

1109: Going around the square in the other direction, 

1110: $\phi^*(\lambda(\omega^\tau_{\phi (a)})) = 

1111: \lambda(\omega^\tau)\circ\lambda(\tau\phi_a)$,

1112: as desired.

1113: \qed

1114:   

1115: \brm

1116: By the previous lemma, one can make $\Lm$ into a functor from the 

1117: category of \taufs\ to the category of 1-forms.  Given $\phi:V\lra W$, 

1118: we have $\Lm(\phi^{\tau*})=\phi^*$.

1119: (Alternatively, one can define $\Lm$ explicitly and fiberwise, as in 

1120: Lemma~\ref{lambda}.)

1121: \erm

1122: 

1123: \sec{Algebraic curves}

1124: \label{algcurves}

1125: 

1126: %+++

1127: 

1128: \subs{Global $\tau$-forms}

1129:   

1130: In this section, by curve we mean a smooth projective curve, unless noted otherwise.

1131: Recall that $\Omega [V]$ denotes the set of global 1-forms on a variety $V$.

1132: 

1133: \bdf 

1134: Given a variety $V$, let $\Omega^\tau [V]$ denote the set of global $\tau$-forms 

1135: on $V$, a $K$-subspace of $\Omega^\tau (V)$.

1136: \edf

1137: 

1138: Note that for any variety $V$, the map $\Lambda_V:\Omega^\tau(V) \lra \Omega(V)$

1139: restricts to a map from $\Omega^\tau[V]$ to $\Omega[V]$.  The next lemma is immediate.

1140: 

1141: \bla

1142: \label{kern}

1143: Let $C$ be a curve.

1144: The kernel of $\Lambda_C:\Omt [C] \lra \Om [C]$, $\textup{Ker}(\Lm_C)$, is the set of

1145: $\omt \in \Omt[C]$ such that for all $c \in C$, $\omt_c$ is a constant affine

1146: map on $\tau C_c$.

1147: Thus each $\omt \in \textup{Ker}(\Lm_C)$ determines in a natural way a regular

1148: function on $C$.  Since the only regular functions on $C$ are the 

1149: constant functions, there is a natural isomorphism $\textup{Ker}(\Lm_C) \cong K$.

1150: 

1151: More generally, the same holds for any smooth projective variety.

1152: \ela

1153: 

1154: We can now describe the fibers of the map $\Lm_C : \Omt [C] \lra \Om [C]$.

1155: For all $\omt_1, \omt_2 \in \Omt [C]$, $\Lm_C (\omt_1) = \Lm_C (\omt_2)$ if and only if

1156: $\Lm_C (\omt_1 - \omt_2) = 0 \in \Om [C]$ if and only if there is an $e \in K$ such that

1157: for all $a \in \tau C$, $\omt_1(a) = \omt_2(a) + e$.  In this 

1158: case, we say that $\omt_1$ is a {\em translate} of $\omt_2$, or that they

1159: are {\em parallel}.  Clearly, parallelism classes are 1-dimensional $K$-affine

1160: spaces and are the fibers of the map $\Lambda_C$.

1161: 

1162: Recall that on a curve $C$ of genus $g$, $\Omega[C]$ is (by definition) a 

1163: $g$-dimensional $K$-vector space. 

1164: Here, we note that the computation of dimension of $\Omt[C]$ is somewhat

1165: more complicated, depending on the {\em Kodaira-Spencer class} of $C$.

1166: 

1167: \bpro

1168: \label{globdim}

1169: Let $C$ be a curve of genus $g$.  Then 

1170: $1 \leq \dim_K \Omt[C] \leq g+1$.

1171: Further, $\dim_K \Omt[C] = g + 1$ if and only if

1172: $C$ is isomorphic to a curve defined over the field $k$ of constants.

1173: \epro

1174: 

1175: \prf

1176: The first statement follows immediately from the above remarks as, given the 

1177: map $\Lm_C : \Omt [C] \lra \Om [C]$, it follows that $\dim_K \Omt [C] 

1178: \leq \dim_K \Om[C]  + \dim_K \Ker (\Lm_C) = g +1$.  The other inequality 

1179: follows from the natural embedding $\iota$ of $K[C] \cong K$ into $\Omt[C]$.

1180: The proof of the second statement,which is equivalent to Proposition~\ref{cohom}, 

1181: uses scheme-theoretic machinery developed below.

1182: \qed

1183: 

1184: \msk

1185: 

1186: For curves of genus 1, this yields a complete description of the global $\tau$-forms.

1187: 

1188: \bco

1189: Let $C$ be a curve of genus 1.  If $C$ is isomorphic to a curve defined over $k$,

1190: then $\dim_K \Omt[C] =2$.  Otherwise, $\dim_K\Omt[C]=1$.  Further, in both cases,

1191: we can give an explicit description of the global $\tau$-forms. 

1192: \eco

1193: 

1194: \prf

1195: Suppose first that $C$ is defined over $k$.  Then, $\tau C \cong TC$, the tangent 

1196: variety.  In this case, it is clear that any \tauf\ can be written (uniquely) as a sum of

1197: a 1-form and of a \tauf\ of the form $\iota f, f \in K(C)$ (as defined in 

1198: Remark~\ref{iotaiota}), and likewise for global forms.  

1199: Thus $\dim_K\Omt[C] = \dim_K\Om[C] + \dim_K K[C] = 1+1 = 2$.

1200: Otherwise, if $C$ is not defined over $k$, then $\dim_K \Omt[C] = 1$, by the above

1201: proposition.  Here, $\Omt[C] = \{\iota e | e\in K, \iota e$ a constant trivial 

1202: $\tau$-form$\}$.  

1203: \qed

1204: 

1205: \msk

1206: 

1207: Recall that Hrushovski and Itai call a 1-form on a curve {\em essential} if it is

1208: not the pullback of any other 1-form, and show that there are many essential 

1209: global 1-forms on any curve of genus $\geq 2$.  

1210: 

1211: \bpro 

1212: \label{hrit}

1213: \textup{(Hrushovski-Itai)}  Let $C$ be a curve of genus $\geq 2$,

1214: defined over $k$.  There exists a finite or countable union  

1215: ${\mathcal{E}} = \cup_lS_l$ of proper subspaces of $\Om [C]$, such that any 1-form in 

1216: $\Om [C] \backslash {\mathcal{E}}$ is essential.  

1217: There exists an essential global 1-form, defined over $k$.

1218: \epro

1219: 

1220: Here, we are unable to determine whether an analgous statement holds for global $\tau$-forms.

1221: Another natural question is whether the global \taufs\ on a curve of positive genus

1222: are exactly the pullbacks of invariant global \taufs\ on its Jacobian, as is true

1223: of 1-forms.

1224: 

1225: \subs{Rational $\tau$-forms}

1226: 

1227: In this section, we introduce an equivalence relation on 

1228: $\tau$-forms that will be useful for the study of strongly minimal 

1229: sets in the next section.  Here, there is no need to assume that

1230: curves are projective.

1231:   

1232: \bdf    

1233: Let $C$ be a curve, $\pi_\tau: \tau C \ra C$ the canonical projection.

1234: Let $\omt$ be a \tauf\, on a curve $C$.  For each $c\in C$, $\omt_c$

1235: is either 

1236: (i) everywhere defined on $\pi_\tau^{-1}(c)$, and a bijective map to $K$,

1237: (ii) everywhere defined on $\pi_\tau^{-1}(c)$, but a constant map to $K$,

1238: (iii) everywhere undefined on the affine fiber  $\pi_\tau^{-1}(c)$.

1239: In case (ii), say that $c$ is a {\em zero} of $\omt$ (where zero here means that

1240: given the affine map $\omt_c:\pi_\tau^{-1}(c)\lra K$, the corresponding

1241: linear map $\lm(\omt_c): K \lra K$ is the zero map).

1242: In case (iii), say that $c$ is a {\em pole} of $\omt$.

1243: 

1244: Let $Z_{\omt}$ denote the set of zeros of $\omt$, and

1245: $P_{\omt}$ its set of poles.  

1246: \edf

1247: 

1248: Observe that for any $\omt \in \Omt (C)$, either $Z_{\omt}$ is finite

1249: or $\omt$ is trivial, in which case $Z_{\omt}$ is an open subset of $C$.

1250: 

1251: \bdf

1252: Let $C$ be a curve, $\omt \in \Omt (C)$ non-trivial.  

1253: The {\em null set} of $\omt$ is

1254: $$

1255: N_{\omt} = \{a \in \tau C | \omt(a) = 0\}.

1256: $$

1257: This is a rational section of the algebraic variety $\tau C$,

1258: and thus birational to $C$.

1259: 

1260: 

1261: Say that $\omt_1, \omt_2 \in \Omt (C)$ are {\em $\seq$-equivalent},

1262: written $\omt_1\seq\omt_2$, if

1263: $N_{\omt_1}$ and $N_{\omt_2}$ are almost equal.

1264: 

1265: Given a \tauf\ $\omt$, let $\omt / \sim$ denote its $\seq$-class.  

1266: \edf

1267: 

1268: \bla  

1269: \label{normform}

1270: Let $C$ be a curve, and $\omt \in\Omt (C)$ a non-trivial $\tau$-form.

1271: For every $\omt_1 \in \Omt(C)$, there are unique $f,g\in K(C)$ such

1272: that $\omt_1=f\omt+\iota g$.

1273: \ela

1274: 

1275: \prf

1276: (The idea is similar to the proof of Proposition~\ref{freemod}.)

1277: If $\omt_1$ is trivial, i.e., there is a $g \in K(C)$ such that

1278: $\omt_1 = \iota g$, then $\omt_1 = 0\omt+\iota g$.  So we may

1279: suppose that $\omt_1$ is non-trivial.

1280: 

1281: Assume first that $\omt$ and $\omt_1$ are $\seq$-equivalent.

1282: Choose some small enough quasi-affine neighborhood of $U \sub C$

1283: such that

1284: \ben

1285: \item  $U$ is disjoint from all the zeros and poles of $\omt$ and $\omt_1$;

1286: \item  $N_{\omt}\cap U = N_{\omt_1 }\cap U$;

1287: \item  on $U$, $\omt$ and $\omt_1$ are of the form

1288: $$

1289: \omt = \sum_{i=1}^mg_i\tau f_i

1290: $$

1291: %

1292: $$

1293: \omt_1 = \sum_{j=1}^n\gamma_j\tau \phi_j

1294: $$

1295: with $g_i,f_i,\gamma_j,\phi_j, \in K[U]$.

1296: \een

1297: In particular, $\omt$ and $\omt_1$ are regular functions on the quasi-affine 

1298: variety $\tau U$.  Furthermore, on each fiber $\tau U_u$,

1299: there is a single point $x \in N_{\omt} = N_{\omt_1}$ such that 

1300: $\omt(x)=\omt_1(x)=0$.  Since $\omt_u$ and $\omt_{1,u}$ are non-constant

1301: affine maps on $\tau U_u$ with the same zero, one must be a 

1302: constant multiple of the other.  That is, there is an $e_u\in K$,

1303: such that $e_u\omt_u=\omt_{1,u}$.  From the definition of $\omt$ and

1304: $\omt_1$ it is clear that the function $e$ on $U$, $e(u) = e_u$, 

1305: is a regular function on $U$. So $\omt_1=e\omt$, as desired.

1306: 

1307: We now consider the general case.  There is a finite set $A \sub N_{\omt}$, such that

1308: $N_{\omt} - A$ is a regular section of $\tau U$, so there is an isomorphism 

1309: $\psi:U\lra N_{\omt} - A$.  Let $g\in K[U]$ be the regular function

1310: $g=\omt_1\circ \psi$, and define $\omt_2 = \omt_1 - \iota g$.

1311: Then $\omt$ and $\omt_2$ are $\seq$-equivalent, so there is an

1312: $f \in K[U]$ such that $\omt_2=f\omt$.  Finally,

1313: $\omt_1 = f\omt + \iota g$, as desired.

1314: \qed

1315:     

1316: \brm

1317: Note that this lemma gives another proof that 

1318: $\Omt(C)$ is a 2-dimensional $K(C)$-vector space,

1319: which also follows from Proposition~\ref{taurat}.

1320: \erm

1321:  

1322: \bco

1323: \label{seqcl}

1324: Let $C$ be a curve, $\omt_1, \omt_2 \in \Omt(C)$, both non-trivial.  

1325: Then $\omt_1$ and $\omt_2$ are $\seq$-equivalent if and only if there is a rational 

1326: function $f\in K(C)$ such that $\omt_1=f\omt_2$.

1327: 

1328: Given a \tauf\ $\omt$, the class $\omt / \sim$ is a 1-dimensional 

1329: subspace of $\Omt(C)$.

1330: \eco  

1331: 

1332: \prf

1333: One direction of the first claim follows immediately from the above proof.  

1334: In the other, suppose that $\omt_1=f\omt_2$.  Then on any open set $U\sub C$ 

1335: disjoint from the zeros and poles of $f, \omt_1$, and $\omt_2$, $N_{\omt_1}|_U

1336: =N_{\omt_2}|_U$.

1337: 

1338: The second claim follows from the the first and the observation that 

1339: $\omt / \sim$ is closed under addition.  

1340: \qed

1341:       

1342: \bla

1343: \label{sigmanull}

1344: Let $C$ be a curve, $\omt_1, \omt_2 \in \Omt(C)$, both non-trivial.  

1345: Suppose that $N_{\omt_1} \cap N_{\omt_2}$ is infinite.  Then 

1346: $\omt_1$ and $\omt_2$ are $\seq$-equivalent.  In other words,

1347: if  $N_{\omt_1} \cap N_{\omt_2}$ is infinite, then 

1348: $N_{\omt_1}$ and $N_{\omt_2}$ are almost equal.

1349: \ela

1350:   

1351: \prf

1352: $N_{\omt_1}$ and $N_{\omt_2}$ are each rational

1353: sections of $\tau C$, so each is a curve on the surface $\tau C$.

1354: If these curves have infinite intersection, they must be equal, up 

1355: to a finite set.

1356: \qed

1357: 

1358: \msk

1359: 

1360: Since we will be concerned with $\sim$-equivalence classes of 

1361: \taufs, the following relativized versions of global and essential \taufs\ 

1362: are important for the main results about strictly minimal sets below.  

1363: Equivalent, perhaps more natural, definitions of 

1364: these notions, in the language of schemes, are given in Section~\ref{tauinv}.

1365: 

1366: \bdf  Let $C$ be a smooth projective curve, $\omt$ a rational \tauf.

1367: Say that $\omt$ is $\sim$-{\em global} if it is $\sim$-equivalent to 

1368: a global \tauf, in which case we also say that the class $\omt / \sim$ is 

1369: {\em global}.

1370: Say that $\omt$ is $\sim$-{\em essential} if every

1371: $\sim$-equivalent \tauf\ is essential.  

1372: \edf  

1373:  

1374: \brm

1375: Let $C, C'$ be curves, and $f:C \lra C'$ a morphism between them.

1376: It is easy to see that for any $\sim$-equivalent \taufs\ $\omt_1$

1377: and $\omt_2$ on $C'$, then  $f^*\omt_1$ and $f^*\omt_2$ 

1378: are $\sim$-equivalent on $C$.

1379: \erm

1380: 

1381: \brm

1382: Let $\omt$ be a non-trivial global \tauf\ on a curve $C$ of genus $g \geq 1$.  

1383: Then for any $c \in K$, $\omt_c : = c + \omt$ is global and not 

1384: $\sim$-equivalent to $\omt$.  In  particular, there is a family

1385: of global $\sim$-classes of $\tau$-forms indexed by $K$.  

1386: \erm

1387: 

1388: Recall that we defined two global \taufs\ on a curve $C$ to be parallel 

1389: if they are both in the same fiber of the map $\Lm_C$.  This notion also

1390: makes sense for rational \taufs.

1391: 

1392: \bdf

1393: Let $C$ be a curve.  We say that two \taufs\ $\omt_1$ and $\omt_2$ are 

1394: {\em parallel}, written $\omt_1\teqq\omt_2$, if there is a $g\in K(C)$ such that 

1395: $\omt_2=\omt_1+g$.  Equivalently, $\Lm_C(\omt_1) = \Lm_C(\omt_2)$.

1396: \edf

1397: 

1398: Parallelism is clearly an equivalence relation.

1399: 

1400: \bla  Fix a curve $C$.

1401: \ben

1402: \item  The trivial \taufs\ on $C$ are a parallelism class.

1403: \item  On the non-trivial \taufs, $\seq$-equivalence and parallelism 

1404:   are cross-cutting equivalence relations.  More precisely, for 

1405:   any non-trivial $\omt_1, \omt_2$,

1406:   $|\omt_1/ \seq \cap \omt_2/ \teqq|=1$.

1407: \een

1408: \ela

1409: \prf

1410: Immediate.

1411: \qed

1412: 

1413: \para{The $\Lm$ map on curves}  

1414: Fix a curve $C$, and a non-trivial \tauf\ $\omt_0\in\Omt(C)$.  

1415: By Lemma ~\ref{normform}, every $\omt\in\Omt(C)$ 

1416: is equal to $f\omt_0+g, f,g\in K(C)$, and, by Corollary~\ref{seqcl}

1417: the $\seq$-equivalence class of $\omt_0$ is 

1418: $\omt_0/ \seq=\{f\omt_0:f\in K(C)-\{0\}\}$.

1419: 

1420: The following lemma can be easily verified.

1421: 

1422: \bla

1423: Let $C$ be a curve, $f\in K(C)$, and $\omt$ a non-trivial $\tau$-form on $C$.  

1424: Then $\Lm_C(f\omt)=f\Lm_C(\omt)$.

1425: %Also, $\Lm_C(f(\omt-1)+1)=f\Lm_C(\omt)$

1426: %

1427: Thus the map $\Lm_C$ induces a bijection between the 

1428: equivalence class $\omt/ \seq$ and the non-trivial 1-forms.

1429: %In particular, $\Lm_C$ is surjective.

1430: %Also, $\Lm_C$ induces a bijection between the 

1431: %equivalence class $\omt/_\teq$ and the non-trivial 1-forms.

1432: \ignore{

1433: \item  For any $\omt_1,\omt_2\in\Omt(C)$, $\Lm_C(\omt_1)=\Lm_C(\omt_2)$ if 

1434: and only if $\omt_1\teqq\omt_2$.  Equivalently, for any $\om\in\Om(C)$, the 

1435: fiber $\Lm^{-1}_C(\om)$ is a parallelism class.  

1436: }

1437: %end{ignore}

1438: \ela

1439: 

1440: \sec{Order one strongly minimal sets}

1441: \label{sm}  

1442: 

1443: We aim to classify order one strongly minimal sets, up to

1444: non-orthogonality.  

1445: Hrushovski proved that every such set is either

1446: trivial or non-orthogonal to the constant field $k$, so we will

1447: be investigating trivial sets.

1448: We first briefly recall the analysis from Hrushovski-Itai of order 

1449: one \sm sets over the constants.

1450:   

1451: Say that two strongly minimal sets $X$ and $Y$ are {\em birationally 

1452: isomorphic}, or {\em birational}, if there is an almost everywhere defined 

1453: bijective map from $X$ to $Y$.  Any order one strongly minimal set $Y$ is 

1454: birational to one of the form

1455: $$

1456: X = \Xi(C, s) = \{ x \in C : \dl x = s(x)\}

1457: $$

1458: where $s: C \lra \tau (C)$ is a rational section of the prolongation.

1459: In this case, one says that $X$ lives on $C$, and the differential

1460: function field of the Kolchin closed set $X$ equals the function 

1461: field of $C$.

1462: 

1463: Suppose now that $C$ is a curve defined over $k$, so $\tau C = TC$,

1464: the tangent space of $C$.  Then the \sm sets $X$ defined over $k$

1465: and living on $C$ are given as $X = \Xi (C, s)$, $s$ a 

1466: rational vector field also defined over $k$.  There is a natural

1467: 1-1 correspondence between such vector fields $s$  and 1-forms

1468: $\om$, given by the relation $\om(x)s(x) = 1$ for almost all $x$.  

1469: So order one \sm sets over the constants also correspond to sets

1470: $$

1471: X = \Xi(C, \om) = \{a \in C : \om(a)\dl a=1\}

1472: $$

1473: where this is taken to include a pole $p$ of $\om$ if $\dl p = 0$.

1474: (This last condition ensures that $\Xi(C, \om)$ is Kolchin closed.)

1475: 

1476: In this setting, Hrushovski and Itai analyze \sm sets in terms of 

1477: pairs $(C, \om)$, $C$ a smooth projective curve and $\om \in \Om (C)$,

1478: both defined over $k$.  We now show how to extend their results

1479: to all curves, using $\tau$-forms.

1480: 

1481: \bdf

1482: Let $C$ be a smooth projective curve and  $\omt \in \Omt (C)$ a

1483: non-zero $\tau$-form.  Let $P \sub C$ be the set of poles $p$ of $\omt$

1484: such that $\dl p=0$, and define

1485: $$

1486: \Xi (C, \omt) = \{a \in C : \omt(a)\dl a = 0\} \cup P

1487: $$

1488: \edf

1489: As above, one can check that $\Xi (C, \omt)$ is always Kolchin closed.

1490: 

1491: Observe that $X = (C, \omt) = \{a \in C : \omt(a)\dl a = 0\} $ is a strongly minimal 

1492: set.  Indeed, let $s$ be a section of $\tau C$ so that $\omt s = 1$ almost 

1493: everywhere.  Then $\Xi(C, s) = \Xi(C, \omt)$, up to a finite set.

1494: On the other hand, 

1495: this is no longer a 1-1 correspondence, as many $\tau$-forms correspond to the 

1496: same section and therefore determine the same \sm set.  

1497: 

1498: \bla

1499: Let $C$ be a curve, $\omt_1, \omt_2 \in \Omt(C)$, both non-trivial.

1500: Then $\omt_1$ and $\omt_2$ are $\seq$-equivalent if and only if 

1501: $\Xi(C, \omt_1)$ and $\Xi(C, \omt_2)$ are almost equal.

1502: \ela

1503: 

1504: \prf

1505: From the definition, $\Xi(C,\omt)$ only depends on $N_{\omt}$,

1506: so if $\omt_1$ and $\omt_2$ are $\seq$-equivalent, then 

1507: $\Xi(C, \omt_1)$ and $\Xi(C, \omt_2)$ are almost equal.

1508: 

1509: In the other direction, if

1510: $\Xi(C, \omt_1)$ and $\Xi(C, \omt_2)$ are almost equal,

1511: then $N_{\omt_1}\cap N_{\omt_2}$ is infinite, so 

1512: $\omt_1$ and $\omt_2$ are $\seq$-equivalent, by 

1513: Lemma~\ref{seqcl}.

1514: \qed

1515: 

1516: \bco

1517: \label{formz}

1518: There is a 1-1 correspondence between $\seq$-equivalence

1519: classes of \taufs\, in $\Omt (C)$ and strongly minimal sets living on $C$.

1520: \eco

1521: Notice that we do not have to treat the case $C(k) \sub C$, corresponding

1522: to $s(x)=0$ separately.  

1523:    

1524: The following series of lemmas provides versions of 

1525: Lemmas 2.4 and 2.6 -- 2.9 of \cite{HI}, for \taufs.  

1526: (Lemma 2.5 becomes false.)  The proofs are the same.

1527: 

1528: For the proof of the next lemma, we recall the following easy fact.

1529: Given $\phi:V\lra W$, and $v\in V$, $(\tau \phi)(\dl v)=\dl (\phi(v))$.

1530: 

1531: \bla  

1532: \label{taupullback}

1533: Let $g: C_1 \lra C_2$ be a dominant  regular map between nonsingular curves, and 

1534: $\omt_2$ a \tauf\, on $C_2$.  Let $\omt_1 = g^{\tau *}\omt_2$ be the pullback of

1535: $\omt_2$ by $g$ to $C_1$.  Then 

1536: $g^{-1}\Xi(C_2, \omt_2) = \Xi(C_1, \omt_1)$.  

1537: \ela

1538: 

1539: \prf  

1540: Let $c_1\in C_1, c_2=g(c_1)$.  Then

1541: $$

1542: \omt_1(c_1)\dl c_1=\omt_2(c_2)(g\prl)(\dl c_1)=\omt_2(c_2)\dl 

1543: (g(c_1))=\omt_2(c_2)\dl c_2.

1544: $$

1545: Thus $\omt_1(c_1)\dl c_1=0$ if and only if $\omt_2(c_2)\dl c_2=0$.

1546: \qed

1547: 

1548: \bla

1549: \label{infint}

1550: If $\Xi(C, \omt_1)$ and $\Xi(C, \omt_2)$ have infinite intersection,

1551: $\omt_1 $ and $\omt_2$ are $\seq$-equivalent.

1552: \ela

1553: 

1554: \prf

1555: For infinitely many points $c$ in the intersection $\Xi(C, \omt_1)

1556: \cap \Xi(C, \omt_2)$, $\omt_1(c)\dl c=0=\omt_2(c)\dl c$.  Thus,

1557: $N_{\omt_1}$ and $N_{\omt_2}$ have infinite intersection, and are

1558: therefore $\seq$-equivalent, by Lemma~\ref{sigmanull}.

1559: \qed

1560: 

1561: \bla

1562: \label{compat}

1563: Let $g:C_1 \lra C_2$ be a dominant rational map between non-singular curves.

1564: Let $\omt_i$ be a \tauf\, on $C_i, i = 1,2$.  Assume

1565: $g\Xi(C_1, \omt_1) \sub \Xi(C_2, \omt_2)$ (perhaps up to a finite set).

1566: Then $\omt_1$ and $g^{\tau *}\omt_2$ are $\seq$-equivalent.

1567: \ela  

1568:    

1569: \prf

1570: By Lemma~\ref{taupullback}, $g^{-1}\Xi(C_2, \omt_2) = \Xi(C_1, g^{\tau *}\omt_2)$

1571: (a.e.).  Thus the intersection of $\Xi(C_1, \omt_1)$ with $\Xi(C_1, g^{\tau *}\omt_2)$

1572: is infinite.  By the previous lemma, $\omt_1$ and $g^{\tau*}\omt_2$

1573: are $\seq$-equivalent.  

1574: \qed

1575: 

1576: \bla

1577: Let $C_1$ be a complete nonsingular curve, $\omt_1$ a global \tauf\, on $C_1$.

1578: Let $g:C_1 \lra C_2$ be a rational function into a complete nonsingular curve,

1579: and $\omt_2$ a \tauf\, on $C_2$ such that $g\Xi(C_1, \omt_1) = 

1580: \Xi(C_2, \omt_2)$.  Then $\omt_1$ is a $\sim$-global \tauf\ on $C_2$.

1581: \ela

1582:   

1583: \prf

1584: As pointed out in \cite{HI}, $g: C_1 \lra C_2$ is a surjective morphism.

1585: It then follows from definitions that $\omt_2$ is a global \tauf\, on $C_2$

1586: if and only if its pullback, $g^{\tau *}\omt_2$ is a global \tauf\, on $C_1$.

1587: By Lemma~\ref{taupullback}, $g^{-1}\Xi(C_2, \omt_2)=\Xi(C_1, g^{\tau *}\omt_1)$,

1588: so $\Xi(C_1, g^{\tau *}\omt_2) \cap \Xi(C_1, \omt_1)$ is infinite.

1589: By Lemma~\ref{infint}, $g^{\tau *}\omt_2 \sim \omt_1$, so by

1590: the above remark, $\omt_1$ is a $\sim$-global form.

1591: \qed  

1592: 

1593: \bla

1594: \label{noeqrel}

1595: Let $\omt$ be a $\sim$-essential \tauf\ on a curve $C$.  Let $E$ be a 

1596: definable equivalence relation on $\Xi(C,\omt)$, with finite classes.

1597: Then almost every class of $E$ has one element.

1598: \ela

1599: 

1600: \prf

1601: Suppose not.  By \cite{HI}, Lemma 1.2, there is a curve $C_0$ and a map

1602: $g: C \ra C_0$ with deg$(g) > 1$ such that $g(\Xi(C,\omt))$ lives on $C_0$.

1603: Thus there is a \tauf\ $\omt_0$ on $C_0$ such that $g(\Xi(C,\omt))=

1604: \Xi(C_0,\omt_0)$, up to a finite set.  But by Lemma~\ref{compat},

1605: $\omt$ and $g^{\tau*}(\omt_0)$ are $\sim$-equivalent, contradicting the

1606: assumption that $\omt$ is $\sim$-essential.

1607: \qed

1608: 

1609: \bla

1610: \label{nonorthcon}

1611: Let $C$ be a curve of genus $\geq 1$ and $\omt$ a \tauf\ on $C$.  

1612: If $\Xi(C,\omt)$ is non-orthogonal to the constants, then there is a 

1613: regular map $g:C \ra \P^1$ with $g(\Xi(C,\omt))=k$, up to a finite set.  

1614: In particular, there is a \tauf\ $\omt_0$ on $\P^1$ such that 

1615: $\Xi(\P^1,\omt_0) = k$ and $g^{\tau*}\omt_0$ is $\sim$-equivalent 

1616: to $\omt$ on $C$.  Thus, $\omt$ is not $\sim$-essential.

1617: \ela

1618: 

1619: \prf

1620: Let $X = \Xi(C,\omt)$ be non-orthogonal to the constants.  Then there is 

1621: a definable differential rational function $f$ with $f(X) \sub k$, up to a

1622: finite set (e.g., see \cite{HI}, proof of Lemma 2.10).  Since for all 

1623: $x \in X, \dl(x)=s(x)$, for some polynomial $s$, we can assume that

1624: $f$ is a rational function.  Thus $f$ extends to a regular, dominant map 

1625: $f:C \ra \P^1$, proving the first claim.  The rest of the lemma follows

1626: from Lemma~\ref{compat}. 

1627: \qed

1628: 

1629: \msk

1630: 

1631: We are now able to generalize a central result of \cite{HI}.  Recall

1632: that a strongly minimal set $X$ defined over a set $A$ is {\em strictly minimal}

1633: over $A$ if $X \cap \acl(A) = \emptyset$ and, for all $a \in X, \acl(aA) \cap X

1634: = \{a\}$.  We say that $X$ is strictly minimal if it is strictly minimal over

1635: some set over which it is defined.  A strongly minimal set $X$ 

1636: is {\em $\omega$-categorical} if for any finite set $B \sub X$, $\acl_X(B)$

1637: is finite.  Hrushovski's theorem that all order one strongly minimal\ sets are 

1638: $\omega$-categorical~\cite{HJ} plays an important role in the proof of the 

1639: following result of \cite{HI}.

1640: 

1641: \bpro

1642: \label{strictmin}

1643: Let $C$ be a complete nonsingular curve over $k$ of genus $>1$.  Let $\om$ 

1644: be an essential 1-form on $C$, defined over $k$.  Then, after perhaps removing

1645: a finite set, $\Xi(C,\om)$ is strictly minimal, with trivial induced 

1646: structure.  Two such sets $\Xi(C_1,\om_1)$

1647: and $\Xi(C_2,\om_2)$ are non-orthogonal if and only if there exists an isomorphism 

1648: $g:C_1 \ra C_2$ with $\om_1 = g^*\om_2$.

1649: \epro

1650: (In \cite{HI}, they state this result for global 1-forms, which implies

1651: additionally that $\Xi(C,\om)$ has no points in the algebraic closure of

1652: the parameters.)

1653: 

1654: We now state one of our main results.

1655: 

1656: \bthm

1657: \label{genprop}

1658: Let $C$ be a smooth projective curve of genus $g \geq 1$.  Let $\omt$ be

1659: a $\sim$-essential \tauf.  Then there is a finite set $A \sub \Xi(C,\omt)$

1660: such that $\Xi(C,\omt) - A$ is trivial strictly minimal. Further, for any

1661: definable equivalence relation with finite classes on $\Xi(C,\omt)$, 

1662: over any set of parameters, almost every class has one element.

1663: 

1664: Two such sets, $\Xi(C_1,\omt_1)$ and $\Xi(C_2,\omt_2)$ are 

1665: not orthogonal if and only if there is an isomorphism $f:C_1 \ra C_2$ such that

1666: $f^{\tau *}\omt_2$ is $\sim$-equivalent to $\omt_1$.

1667: \ethm

1668: 

1669: \prf

1670: By Lemma~\ref{nonorthcon}, $X = \Xi(C,\omt)$ is a trivial strongly minimal set and

1671: by Lemma~\ref{noeqrel}, for any definable equivalence relation with finite classes 

1672: on $X$, over any set of parameters, almost every class has one element.

1673: Since $X$ is $\omega$-categorical, the set $A$ of elements of $X$ which are 

1674: algebraic over the parameters defining $C$ and $\omt$ is finite.  Thus 

1675: $X - A$ is strictly minimal.

1676: 

1677: Suppose $\Xi(C_1,\omt_1)$ and $\Xi(C_2,\omt_2)$ are non-orthogonal.  

1678: After removing a finite set of points from each, one is left with two

1679: strictly minimal sets with trivial geometry, so there is a definable 

1680: bijection between them.  Thus there is a birational map $g: C_1 \ra C_2$,

1681: with $\omt_1 \sim g^{\tau *}\omt_2$, which is an isomorphism.

1682: \qed

1683: 

1684: \brm

1685: In a sense, this completely classifies the trivial strictly minimal sets,

1686: in terms of $\sim$-essential forms, but it is not very explicit.  

1687: In particular, it leaves open the question whether there are any such forms.

1688: For curves over $k$, one gets an affirmative answer from \cite{HI}, but the general

1689: case remains open.

1690: 

1691: A key part of Hrushovski and Itai's proof that there are many trivial strictly 

1692: minimal sets living on every curve defined over $k$ of genus $>1$,

1693: which are all orthogonal, is the fact that on such curves generic global 

1694: 1-forms are essential.  Proving an analogous statement for \taufs\ seems difficult,

1695: since the dimension of global \taufs\ does not depend only on the genus.

1696: Further, in the more general case, one needs to consider $\sim$-essential \taufs,

1697: which are essentially equivalence classes.  For these reasons, it is not

1698: clear how one might be able to adapt ideas from \cite{HI} to answer this question.

1699: \erm

1700: 

1701: Let us call an equivalence relation on a set $X$ {\em trivial} if it is definable

1702: in the language of equality.  Equivalently, there is either a cofinite equivalence class

1703: or all but finitely many elements are in classes of size 1.

1704: 

1705: In the statement of Proposition~\ref{strictmin}, there are no non-trivial equivalence 

1706: relations on $\Xi(C,\omega)$ defined over $k$.   One might ask whether it is possible

1707: to define such a relation using parameters from $K$.  If this were the case, then

1708: one would have pairs $(C,\omega)$ and $(C_0,\omt_0)$, $C \not\cong C_0$,

1709: $C$ a curve defined over $k$, $\omega$ an essential

1710: 1-form defined over $k$, $\Xi(C,\omega)$ trivial strictly minimal over $k$, and

1711: $\Xi(C_0,\omt_0)$ strictly minimal over $K$, with  a finite-to-one map from 

1712: $\Xi(C,\omega)$ to $\Xi(C_0,\omt_0)$.    Below, we argue that this can not in fact

1713: happen.  The key point is the following lemma.

1714:   

1715: \bla

1716: \label{eqprm}

1717: Let $K$ be an $\om$-stable structure, and let $X$ be a trivial strictly minimal set in $K$

1718: defined over a set $A$.  Then no non-trivial equivalence relation on $X$ is definable

1719: with parameters from $K$.

1720: \ela

1721: 

1722: \prf

1723: As usual, we may assume that the set $A$ over which $X$ is defined is empty.

1724: Suppose for contradiction that there is a tuple $\abar$ and a formula $\phi(\xbar,y,z)$

1725: such that $\phi(\abar,y,z)$ defines a non-trivial equivalence relation $E$ on $X$.

1726: By strong minimality of $X$, it is clear that there is a finite $n$ such that every

1727: class in $E$ is of size $\leq n$ and all but finitely many classes are of the same

1728: size $m > 1$.  Let $\theta(\xbar)$ be a formula true of $\abar$ such that for all

1729: $\bbar$, if $K \models \theta(\bbar)$, then $\phi(\bbar,y,z)$ defines an equivalence relation

1730: on $X$ such that all but finitely many elements of $X$ belong to a class of size $m$.

1731: 

1732: Let $T$ be the set of equivalence relations $T = \{\phi(\bbar,y,z) | K \models \theta(\bbar)\}$.

1733: Note that, even without elimination of imaginaries in $K$, one can treat $T$ as a definable

1734: set that can be quantified over in a natural way.  Observe also that $T$ is infinite

1735: as follows.  For if not, choose an $E \in T$ and a pair $(p,q)$ in $E$.  

1736: Then $q \in \acl (p)$, contradicting the fact that $X$ is trivial strictly minimal.

1737: 

1738: Let $\Sigma$ be the set of finite sequences of 0's and 1's, partially ordered such that,

1739: for $\rho, \sigma \in \Sigma$,

1740: $\rho \leq \sigma$ if and only if $\rho$ is an initial segment of $\sigma$.

1741: For each $\sigma \in \Sigma$, we define a pair $\pi_\sigma = (p_\sigma,q_\sigma)$

1742: of elements in $X$ and a `definable' set $T_\sigma \sub T$ with the following properties.  

1743: \ben

1744: \item $T_\sigma$ is infinite and definable over parameters 

1745: $\cup_{\rho < \sigma}\{p_\rho,q_\rho\}$.

1746: 

1747: \item $T_\sigma$ is the disjoint union of $T_{\sigma,0}$ and $T_{\sigma,1}$.

1748: 

1749: \item $T_{\sigma,0}:=\{E \in T_\sigma | E(p_\sigma,q_\sigma)\}$ and 

1750:       $T_{\sigma,1}:=\{E \in T_\sigma | \neg E(p_\sigma,q_\sigma)\}$ are both non-empty.

1751: \een

1752: Since it is clear that such a set of $T_\sigma$'s contradicts $\omega$-stability,

1753: this will yield the desired contradiction.

1754: 

1755: We proceed by induction.  

1756: Let $T_\emptyset = T$.  Choose a pair $\pi_\emptyset=(p_\emptyset,q_\emptyset)$ such that 

1757: there are $E,E' \in T_\emptyset$ 

1758: with $E(p_\emptyset,q_\emptyset)$ and $\neg E'(p_\emptyset,q_\emptyset)$.

1759: (If this were not possible, then $T$ would only contain one element.)  

1760: 

1761: Suppose now that for all $\rho \leq \sigma$, $\pi_\rho$ and $T_\rho$ have been defined.

1762: Let $T_{\sigma,0} = \{E \in T_\sigma| E(p_\sigma,q_\sigma)\}$ and 

1763: $T_{\sigma,1} = \{E \in T_\sigma| \neg E(p_\sigma,q_\sigma)\}$.

1764: By construction, both $T_{\sigma,0}$ and $T_{\sigma,1}$ are non-empty.

1765: In fact, we claim that both sets must be infinite.  For if not, it is easy to see

1766: that $q_\sigma \in \acl(\{p_\rho,q_\rho | \rho < \sigma\}\cup\{p_\sigma\})$,

1767: contradicting the fact that $X$ is trivial strictly minimal.

1768: Now choose $\pi_{\sigma,i}=(p_{\sigma,i},q_{\sigma,i})$, $i = 0,1$,  disjoint from all

1769: $\pi_\rho$, $\rho \leq \sigma$, such that 

1770: there are $E_i, F_i \in T_{\sigma,i}$ such that $E_i(p_{\sigma,i},q_{\sigma,i})$

1771: and $\neg F_i(p_{\sigma,i},q_{\sigma,i})$.

1772: 

1773: It is clear that this construction has the required properties.

1774: \qed

1775: 

1776: \bco

1777: Let $C$ be a complete nonsingular curve over $k$ of genus $>1$.  Let $\om$ 

1778: be an essential 1-form on $C$, defined over $k$, so that after perhaps removing 

1779: a finite set, $\Xi(C,\om)$ is strictly minimal with trivial induced 

1780: structure.  Then there is no non-trivial equivalence relation definable on $\Xi(C,\om)$,

1781: even with arbitrary parameters.

1782: \eco

1783:   

1784: \prf

1785: By Proposition~\ref{strictmin} and Lemma~\ref{eqprm}.

1786: \qed

1787: 

1788: \section{Another description of $\tau$-forms}

1789: \label{tauinv}

1790: 

1791: %%% Redo intro

1792: 

1793: In this section, we give a new algebraic characterization of the 

1794: strongly minimal sets living on a curve $C$, using the theory of $\tau$-differentials

1795: developed in \cite{Ros}.   Of course, it is equivalent

1796: to the earlier formulation, but has the advantage that it does not involve

1797: considering equivalence classes, so it is easier to work with.

1798: Below, we use the language of schemes, so a variety is an integral separated

1799: scheme of finite type over an algebraically closed field.

1800: We continue to assume that curves are smooth and projective.

1801: 

1802: Recall that there is a natural bijection between rank $n$ locally free

1803: sheaves on a variety and rank $n$ vector bundles over it.

1804: For example, the sheaf of differential forms on a variety $X$,

1805: $\Om_X$, corresponds to the cotangent bundle $T^*X$, 

1806: and can be viewed as the sections of this bundle.  Given a curve $C$,  

1807: the locally free rank 2 sheaf of $\tau$-differentials on $C$, as defined in \cite{Ros},

1808: and also denoted $\Omt_C$, 

1809: corresponds exactly to the set of rational $\tau$-forms, as defined above,

1810: with a $\tau$-form $\omt$ corresponding to a section $s \in \Omt(U)$,

1811: for open $U \sub C$.

1812: 

1813: Given a rank 2 bundle, one can consider all the rank 1 subbundles.

1814: (To be precise, a rank 1 subbundle is a rank 1 bundle together

1815: with a bundle embedding.)  Passing to sheaves, rank

1816: 1 subbundles correspond to invertible (i.e., rank 1) subsheaves.

1817: Note that it is possible for one rank 1 sheaf to be properly embedded in

1818: another (in which case the quotient will be a torsion sheaf), essentially

1819: because this can also happen with free rank 1 modules.  Nevertheless,

1820: in our situation, any invertible subsheaf of a rank 2 locally free sheaf 

1821: embeds in a unique maximal invertible subsheaf, as described below.

1822: 

1823: We have seen that \taufs\ correspond to sections of a rank 2 locally

1824: free sheaf.  Since the $\seq$-equivalence class of such a form $\omt$

1825: consists of those forms $g\omt, g \in K(C)$, such a class should correspond

1826: to a (unique) maximal subsheaf containing $\omt$ as a section on an open set.

1827: The following easy lemma makes this precise.

1828: 

1829: \bla

1830: Let $C$ be a smooth curve, $\MH$ a rank 2 locally free sheaf on $C$,

1831: and $s \in \MH(U)$ a section.  Then there is a unique maximal 

1832: invertible subsheaf $\MF \sub \MH$ such that $s \in \MF(U)$.  

1833: Further, the quotient $\MH / \MF$ is an invertible sheaf.

1834: \ela

1835: 

1836: \prf

1837: By possibly passing to a smaller open set, we may assume that $U = \Spec\ R$

1838: is affine, $\MH(U) \cong \tilde{M}$, $M = \<m,n \>$ a free rank 2 

1839: $R$-module, and $s = am + bn\in M, a,b \in R$.  Since $R$ is a UFD 

1840: (e.g.\ \cite{Mat}, Thm.\ 20.1), $a$ and $b$ have a greatest common divisor

1841: $c$.  Letting $a' = a/c$, $b'=b/c$, and $s'=a'm + b'n$, the module 

1842: $N = Rs' \sub M$ is the unique largest rank 1 free submodule of $M$

1843: containing $s$.  Let $d,e \in R$ be such that $da'+eb'=1$, and let

1844: $L = R(em-dn)$.  Then it is easy to check that $M  = N \op L$, and hence that 

1845: there is a short exact sequence $0 \lra N \lra M \lra L \lra 0$.  

1846: 

1847: We want to define $\MF$ such that $\MF(U) = \tilde{N}$.  Since a sheaf is 

1848: determined by its stalk at each point $P$, it suffices to show how, for each 

1849: $P \in C - U$, $\MF(U)$ determines a unique maximal stalk $\MF_P$

1850: that will make $\MF$ locally free.  

1851: 

1852: For each such $P$, choose an affine scheme $\ W \sub U \cup \{P\}$

1853: containing $P$, $W \cong \Spec\ S$, such that $\MH(W)$ is free on $W$.  

1854: Note that $ W':= W - P =

1855: \Spec\ S_f$, for some $f \in S$, and $P = V(f)$.  Let $\MH(W') = \tilde{J}$

1856: and choose basis elements $\<i,j\> \in J$ such that $\MF(W') = \tilde{I}$,

1857: where $I := Ri \sub J$.  We then define $\MF_P$ to be the submodule of

1858: $\tilde{J}_P$ generated by $\<i\>$.  

1859: Clearly, this determines an extension of $\MF$ to $U \cup P$ and thus,

1860: continuing in this way to all of $C$.  (As an aside, note that one could 

1861: also define $\MF_P$ to be generated by $f^mi$, for any positive $m$.

1862: Choosing $m > 1$ would yield a non-maximal invertible subsheaf.)

1863: \qed

1864: 

1865: \msk

1866: 

1867: It is easy to see that two sections determine the same invertible subsheaf

1868: if and only if, on some non-empty open set on which they are both defined,

1869: each is a multiple of the other.  In particular, $\sim$ equivalence classes

1870: of \taufs\ correspond to maximal invertible subsheaves, as desired.

1871: 

1872: \bdf

1873: Let $C$ be a curve.  A {\em $\tau$-invertible sheaf} on $C$, denoted $\MFT$,

1874: is a maximal invertible subsheaf of $\Omt_C$, i.e, an invertible sheaf $\MF$, 

1875: together with an embedding $f:\MF\ra\Omt_C$.

1876: \edf

1877: 

1878: \brm

1879: For any curve $C$, there is a {\em trivial} \tis, corresponding to

1880: the natural embedding of $\O_C$ into $\Omt_C$.  Clearly, this \tis\

1881: corresponds to the $\sim$-equivalence class of trivial \taufs.  

1882: Below, by \tis\ we will always mean non-trivial \tis.

1883: \erm

1884: 

1885: 

1886: We can reformulate Corollary~\ref{formz} as follows.

1887:   

1888: \bpro

1889: Let $C$ be a curve.  There is a natural bijection between strongly 

1890: minimal sets living on $C$ and $\tau$-invertible sheaves on $C$.

1891: \epro

1892: 

1893: As before, given a \tis\ $\MFT$ on $C$, let $\Xi(C,\MFT)$ denote the

1894: corresponding strongly minimal\ set.

1895: 

1896: %The following lemma is clear.

1897: %\bla

1898: %Let $C,C'$ be curves, $f:C\ra C'$ a morphism, and $\MFT$ a $\tau$-invertible

1899: %sheaf on $C'$.  Then the pullback $f^*\MFT$ is a $\tau$-invertible sheaf

1900: %on $C$.

1901: %\ela

1902: 

1903: We now introduce the analog of a global \tauf.  

1904: Since we are dealing with equivalence classes of \taufs,

1905: the correct notion is that of being a \tauf\ that is $\seq$-equivalent

1906: to a global \tauf.  Passing to $\tau$-invertible sheaves, these are

1907: the sheaves $\MFT$ that contain a global section.

1908: 

1909: \bdf

1910: Let $C$ be a curve, $\MFT$ a \tis. Say that $\MFT$ is a {\em global \tis}

1911: if $H^0(C,\MFT) \neq 0$, that is, if $\MFT$ has a global section.

1912: \edf

1913: 

1914: Next, we define the appropriate notion of a $\tau$-invertible

1915: sheaf being essential, which is more subtle than for 1-forms.

1916: First, we establish the following results.

1917: 

1918: \bpro

1919: \label{injpull}

1920: Let $f: X \ra Y$ be a finite morphism of curves.  

1921: There is an exact sequence of sheaves on $X$,

1922: $$

1923: 0 \lra f^*\Omt_Y \sr{\al}{\lra} \Omt_X \lra \Om_{X/Y} \lra 0

1924: $$

1925: \epro

1926: 

1927: \prf

1928: To check that $\alpha$ is injective, it suffice to check this at the 

1929: stalk of the generic point.  Here, injectivity is obvious.

1930: \qed

1931: 

1932: The following proposition follows immediately.

1933: 

1934: \bpro

1935: \label{tauinvpull}

1936: Let $f:X \ra Y$ be a finite morphism of curves, and $\MFT$ a (non-trivial)

1937: \tis\ on $Y$.

1938: Then there is a unique  \tis\ $\MGT$ on $Y$ such that $\al(f^*\MFT) \sub \MGT$.

1939: %the map $f^*\Omt_Y \lra \Omt_X$

1940: %determines an injective homomorphism of $f^*\MFT$ into $\MGT$.  

1941: Further, there is the following commutative diagram, with exact rows,

1942: $$

1943: \xymatrix{

1944: 0 \ar[r] & f^*\MFT \ar[r]\ar[d] & \MGT \ar[r]\ar[d] & \Om_{X/Y} \ar[r]\ar[d] & 0 \\

1945: 0 \ar[r] & f^*\Omt_Y \ar[r]^\al & \Omt_X \ar[r] & \Om_{X/Y} \ar[r] & 0,\\

1946: }

1947: $$

1948: such that each vertical arrow is an injection, and $\MH$ is a torsion sheaf.

1949: \epro

1950: 

1951: \bpr

1952: The only point that needs to be checked is that the cokernel of the map

1953: $f^*\MFT \lra \MGT$ is $\Om_{X/Y}$.  To verify this, it suffices to compute

1954: the map locally around each ramification point of $X$.

1955: \epr

1956: 

1957: \brm

1958: This proposition says that, given a morphism of curves,

1959: $f:X\ra Y$, the pullback of a \tis\ is not necessarily

1960: a \tis, even though the pullback of a $\tau$-form is a $\tau$-form.

1961: By analogy, the pullback of a 1-form on $Y$ is a 1-form on $X$

1962: though, in general,  $f^*\Om_Y \neq \Om_X$.  But there is an 

1963: injective homomorphism $f^*\Om_Y \ra \Om_X$, just as there is

1964: an injective homomorphism $f^*\MFT \ra \MGT$.

1965: 

1966: Note that given a morphism between curves (or more generally varieties), 

1967: $f: C_1 \lra C_2$, the pullback of the trivial \tis\ on $C_2$ is equal to 

1968: the trivial \tis\ on $C_1$, precisely because $f^*\O_{C_2} = \O_{C_1}$.

1969: \erm

1970: 

1971: \bdf

1972: In the notation of the previous proposition, we say that $\MGT$ is the 

1973: {\em weak pullback} of $\MFT$ (along $f$).  

1974: In case the natural map $f^*\MFT \ra \MGT$ is an isomorphism, we say 

1975: that $\MGT$ is the {\em strong pullback} of $\MFT$.

1976: \edf

1977: 

1978: The following lemma follows immediately from the definitions.

1979: 

1980: \bla

1981: Let $f: X \ra Y$ be a morphism of curves, $\omt$ a $\tau$-form on $Y$,

1982: and $\MFT$ the \tis\ corresponding to $\omt / \sim$.  Then $\MGT$, the weak

1983: pullback of $\MFT$, is the \tis\ corresponding to $f^*\omt$.

1984: \ela

1985: 

1986: We can now reformulate Lemma~\ref{taupullback} in terms of \tiss.

1987: \bla

1988: Let $f: X \ra Y$ be a morphism of curves, $\MFT$ a \tis\ on $Y$,

1989: and $\MGT$ the weak pullback of $\MFT$ on $X$.  Then

1990: $f^{-1}\Xi(Y,\MFT)=\Xi(X,\MGT)$.

1991: \ela

1992: 

1993: \bdf

1994: Let $X$ be a curve.  A \tis\ $\MGT$ on $X$ is {\em essential}

1995: if there does not exist a curve $Y$ and a morphism $f:X\ra Y$ 

1996: such that $\MGT$ is the weak pullback of some \tis\ on $Y$.

1997: \edf

1998: 

1999: \brm

2000: Let $X$ be a curve, and $f:X \ra Y$ a regular map to another curve $Y$.

2001: By Proposition~\ref{tauinvpull}, there is a natural embedding of $\O_X$-sheaves,

2002: $\alpha: f^*\Omt_Y \lra \Omt_X$.  Notice that for any \tis\ 

2003: $\MGT\sub \Omt_X$, there is maximal invertible subsheaf of 

2004: $\alpha(f^*\Omt_Y)$ that embeds in $\MGT$, 

2005: namely $\MGT \cap \alpha(f^*\Omt_Y)$.

2006: Let us call this subsheaf $\MH$.  If $\MGT$ is essential, then

2007: $\MH$ cannot be of the form $f^*\MFT$, for $\MFT$ a \tis\ on $Y$.

2008: 

2009: In particular, although every section of $f^*\Omt_Y$ is naturally a

2010: function on $\tau X$, not every such function is the pullback along

2011: $f$ of a $\tau$-form on $Y$.  Functions arising in this way

2012: naturally correspond instead to sections of $f^{-1}\Omt_Y$, but this 

2013: is not an $\O_X$-sheaf.  Recall that by definition, 

2014: $f^*\Omt_Y = f^{-1}\Omt_Y \ot_{f^{-1}\O_Y}\O_X$.

2015: 

2016: As sections of a \tis\ $\MFT$ on $Y$ correspond to a $\sim$-equivalence class of functions 

2017: on $\tau Y$, sections of $f^{-1}\MFT$ correspond to the pullback of these

2018: functions to $\tau X$ along the lifting map $f\prl$, as defined in Section~\ref{prolong}.

2019: As $f^*\MFT = f^{-1}\MFT\ot_{f^{-1}\O_Y}\O_X$, sections of $f^*\MFT$ correspond to 

2020: products of functions from $f^{-1}\MFT$ with rational functions on $X$.

2021: \erm

2022: 

2023: \brm

2024: Observe that a $\tau$-form $\omt$ is $\sim$-essential, as defined in

2025: Section~\ref{algcurves}, 

2026: if and only if the \tis\ corresponding to $\omt / \sim$ is essential.

2027: \erm

2028: 

2029: We can now restate Theorem~\ref{genprop}.

2030: 

2031: \bthm

2032: Let $C$ be a smooth projective curve of genus $g \geq 1$.  Let $\MFT$ be

2033: an essential \tis.  Then there is a finite set $A \sub \Xi(C,\MFT)$

2034: such that $\Xi(C,\MFT) - A$ is strictly minimal. Further, for any

2035: definable equivalence relation with finite classes on $\Xi(C,\MFT)$, 

2036: over any set of parameters, almost every class has one element.

2037: Two such sets, $\Xi(C_1,\mathcal{F}^\tau_1)$ and 

2038: $\Xi(C_2,\mathcal{F}^\tau_2)$ are 

2039: non-orthogonal if and only if there is an isomorphism $f:C_1 \ra C_2$ such that

2040: $\mathcal{F}^\tau_1$ is the weak pullback of $f^{*}\mathcal{F}^\tau_2$.

2041: \ethm

2042: 

2043: We now complete the proof of Proposition~\ref{globdim}.  Given a (locally free) 

2044: sheaf $\MF$ on a projective variety $V$, let $\hh^0(V,\MF)= \dim_K H^0(V,\MF)$,  

2045: the dimension of the space of global sections of $\MF$.

2046: 

2047: Given a curve $C$, we will call $\hh^0(C,\Omt_C)$ the {\em $\tau$-rank} of $C$, 

2048: denoted $g^\tau_C$ or simply $g^\tau$.  The following proposition shows that $g^\tau$

2049: can be easily determined by the Kodaira-Spencer rank of $C$, which we now explain.

2050: 

2051: By \cite{Ros}, we have the following short exact sequence 

2052: $$

2053: \xi:\qquad \ 0 \lra \O_C \lra \Omt_C \lra \Om_{C/K} \lra 0.

2054: $$

2055: Recall that $\xi$ is called an extension of $\Om_{C/K}$ by $\O_C$,

2056: and that there is a natural bijection between the set of such extensions and 

2057: the cohomology class $\Ext^1(\Om_{C/K}, \O_C)$, such that 

2058: the split extension corresponds to $0\in \Ext^1(\Om_{C/K}, \O_C)$ 

2059: (e.g., \cite{CW}, Theorem 3.4.3).  In our case, this mapping is given as follows.

2060: Given $f:C \lra \Spec\, K$, consider the sequence 

2061: $$

2062: \eta: \qquad 0 \lra f^*\Om_K \lra \Om_C \lra \Om_{C/K} \lra 0

2063: $$

2064: (with $\Om_C$ projective).  Applying $\Hom(-,\O_C)$ yields the exact sequence

2065: $$

2066: \Hom(\Om_C,\O_C) \lra \Hom(f^*\Om_K, \O_C) \sr{\partial}{\lra} 

2067: \Ext^1(\Om_{C/K},\O_C) \lra 0

2068: $$

2069: Given $x\in \Ext^1(\Om_{C/K},\O_C)$, choose $\beta \in \Hom(f^*\Om_K, \O_C)$

2070: such that $\pd(\beta)=x$.  We then let $\zeta$, the pushout of $\eta$ along $\beta$,

2071: be the extension corresponding to $x$.

2072: $$

2073: \xymatrix{

2074: \eta: & 0 \ar[r] & f^*\Om_K \ar[r]\ar[d]^\beta 

2075: & \Om_C \ar[r]\ar[d] & \Om_{C/K} \ar[r]\ar[d] & 0 \\

2076: %

2077: \zeta: & 0 \ar[r] & \O_C \ar[r] & \MG \ar[r] & \Om_{C/K} \ar[r] & 0

2078: }

2079: $$

2080: Thus, by \cite{Ros}, the extension $\xi$ corresponds to the element

2081: $\pd(\dltl) \in \Ext^1(\Om_{C/K},\O_C)$, where $\dltl$ is determined by the derivation 

2082: on $K$.

2083: 

2084: There are natural isomorphisms $\Ext^1(\Om_{C/K},\O_C) \cong 

2085: \Ext^1(\O_C, \Theta_{C/K}) \cong H^1(C,\Theta_{C/K})$, 

2086: with $\Theta_{C/K} = \Om_{C/K}^\vee$, the tangent sheaf

2087: of $C$ (the dual of $\Om_{C/K}$) (e.g., \cite{RH}, pages 234-5).

2088: Thus, there is a natural map, which we also denote $\pd$,

2089: $\pd:\Hom(f^*\Om_K,C) \lra H^1(C,\Theta_{C/K})$, 

2090: which Buium~(\cite{Bui93}, p. 1395) calls the Kodaira-Spencer map of $C$.  

2091: The element $\pd(\dltl) \in \Ext^1(\Om_{C/K},\O_C)$, or $H^1(C,\Theta_{C/K})$,

2092: is the Kodaira-Spencer class of $C$.

2093: 

2094: From the sequence $\eta$, corresponding to the Kodaira-Spencer class of $C$,

2095: one obtains a long exact sequence of cohomology groups.

2096: $$

2097: 0 \lra \HH^0(C,\O_C) \lra \HH^0(C,\Omt_C) \lra \HH^0(C,\Om_{C/K})

2098: \stackrel{\partial}{\lra}

2099: \HH^1(C,\O_C) \lra \ldots

2100: $$

2101: Let us call $\dim_K \partial (\HH^0(C,\Om_{C/K}))$ the {\em Kodaira-Spencer} rank of

2102: $C$, denoted $KS_C$.  Clearly, if the Kodaira-Spencer class of $C$ is trivial,

2103: that is, if $\eta$ splits, then $KS_C = 0$, though the converse does not hold in 

2104: general.  

2105: For curves of genus 1, though, $\eta$ splits if and only if $KS_C = 0$.  

2106: This follows, for example, 

2107: from Atiyah's classification of rank 2 bundles over elliptic curves \cite{At}.

2108: This yields a complete understanding of global $\tau$-forms on elliptic curves.

2109: 

2110: The following proposition is equivalent to Proposition~\ref{globdim} above.

2111: 

2112: \bpro

2113: \label{cohom}

2114: Let $C$ be a smooth projective curve.  

2115: Then $1 \leq h^0(C,\Omt_C)\leq g + 1$.

2116: Precisely, $h^0(C,\Omt_C) = g+1 - KS_C$.

2117: In particular, if $C$ is defined over $k$, the field of constants, then 

2118: $h^0(C,\Omt_C) = g+1 $.

2119: \epro

2120: 

2121: \prf

2122: The first two statements follow from remarks above.  The third follows from 

2123: the fact that a curve defined over $k$ has trivial Kodaira-Spencer class.

2124: (Since $H^1(C,\Theta_{C/K})$ classifies $TC$-torsors (e.g., 

2125: see~\cite{JM}), this is equivalent to the fact that for such curves, 

2126: $\tau C \cong_C TC$.)

2127: \qed

2128: 

2129: 

2130: 

2131: \ignore{

2132: To prove the third,

2133: recall that the class $H^1(C,\Theta_{C/K})$ classifies $TC$-torsors (e.g., 

2134: see~\cite{JM}), that and Buium~(\cite{Bui93}, p. 1396) proves that the 

2135: {\em Kodaira-Spencer class}\

2136: $\pd(\dltl)$ corresponds with the $TC$-torsor $\tau C$.  The torsor is  

2137: trivial, that is, isomorphic to $TC$, if and only if $\pd(\dltl) =0$, which

2138: holds exactly when the original sequence $\xi$ splits.  Putting everything together,

2139: one gets that $\xi$ splits if and only if $\tau C \cong_C TC$, which happens exactly

2140: when $C$ is defined over $k$, as desired.  This completes the proof.

2141: \qed

2142: }

2143: %%end{ignore}

2144: 

2145: \subsection{Definability results}

2146: 

2147: Recall that the {\em degree} of an invertible sheaf on a variety is defined to 

2148: be the degree of the corresponding divisor.  We now observe that there is 

2149: a (uniform) bound on the degree of a $\tau$-invertible sheaf on a curve.  

2150: 

2151: \bla

2152: Let $V$ be a projective curve, $\MF$ a coherent sheaf on $V$.  Then for any 

2153: invertible subsheaf $\MG$ of $V$, $\deg \MG \leq \hh^0(\MF) + g(C) - 1$.

2154: \ela

2155: 

2156: \bpr

2157: By the Riemann-Roch theorem, 

2158: $\deg \MG = \hh^0(\MG) - \hh^0(\om \otimes \MG^{-1}) + g - 1 \leq 

2159: \hh^0(\MG) + g - 1 $.  Since $\hh^0(\MG) \leq \hh^0(\MF)$, this yields the 

2160: desired bound.

2161: \epr

2162: 

2163: \bco

2164: For any curve $C$, the degree of any \tis\ is $\leq 2g$.

2165: \eco

2166: 

2167: The next result then follows by standard arguments.

2168: 

2169: \bco

2170: For any curve $C$, the set of global $\tau$-forms on $C$ is definable,

2171: as is the equivalence relation $\sim$ on these forms.  Further, this remains

2172: true for families of curves (of fixed genus).  Finally, the dimension of 

2173: the space of global $\tau$-forms, as a $K$-vector space, is uniformly definable.

2174: \eco

2175: 

2176: Observe that, by Proposition~\ref{cohom}, this last claim is equivalent to the known fact

2177: that the Kodaira-Spencer rank of a curve is definable.

2178: 

2179: %% ??

2180:  

2181: \ignore{

2182: \newpage

2183: 

2184: We can now state one of our main results.

2185: All curves are smooth projective curves over an uncountable algebraically 

2186: closed field $K$.  Given a curve $C$ and a sheaf $\MF$, recall that 

2187: $h^i(C,\MF) = \dim_K(H^i(C,\MF))$.  (I'll specify the notion of

2188: generic curve later.)

2189: 

2190: \bthm

2191: Let $C$ be a generic curve of genus $g \geq 2$.  Then every 'non-trivial'

2192: global $\tau$-invertible sheaf is essential.

2193: \ethm

2194: 

2195: The proof uses the following proposition, which was provided by L. Ein.

2196: 

2197: \bpro

2198: Let $C$ be a generic curve of genus $g \geq 2$.  For any curve $C'$,

2199: morphism $f:C \ra C'$, and invertible sheaf $\MF$ on $C'$,

2200: if $h^0(C',\MF) = 0$, then $h^0(C,f^*\MF) = 0$.

2201: \epro

2202: 

2203: \prf

2204: The following fact is well-known.  (The proof involves a counting

2205: argument.  The moduli space of curves of genus $g \geq 2$ is 

2206: $(3g-3)$-dimensional and, by the Hurwitz formula, if $f: C_1 \ra C_2$

2207: is a morphism between curves, then $g(C_2) \leq g(C_1)$.)

2208: 

2209: \bla

2210: Let $C$ be a generic curve of genus $g \geq 2$, $C$ any curve, and

2211: $f:C\ra C'$ a morphism of degree $\geq 2$.  Then $C' \cong \P^1$.

2212: \ela

2213: 

2214: We now assume that $C'\cong\P^1$, and let $\MF$ be an invertible sheaf

2215: on $\P^1$, $\MF=\O_{\P^1}(n), n\in\Z$.  Then 

2216: $

2217: h^0(C,f^*\MF)=h^0(\P^1,f_*f^*\MF)=h^0(C',\MF\otimes f_*\O_C).

2218: $

2219: The first equality is obvious, and the second follows immediately from 

2220: the  Projection Formula (\cite{RH}, p.\ 124).  

2221: 

2222: We claim that $f_*\O_C$ is a locally free sheaf of the form

2223: $\O_{\P^1}\oplus(\oplus_i\O_{\P^1}(a_i)),$

2224: with $a_i < 0$, for all $i$.  We have a short exact sequence,

2225: $$

2226: 0\lra \O_{\P^1}\stackrel{\alpha}\lra f_*\O_C\lra \ME \lra 0.

2227: $$

2228: Let $\beta: f_*\O_C \ra \O_{\P^1}$ be the map $\beta = \frac{1}{d}t$,

2229: where $t$ is the standard trace map.  Then $\beta\circ\alpha$ is the 

2230: identity on $\O_{\P^1}$, so the above sequence splits,

2231: i.e., $f_*\O_C=\O_{\P^1}\oplus\ME$.  

2232: 

2233: Next, observe that $\ME$ is locally free.  To show this, it is sufficient

2234: to show that $\ME_x$ is locally free, for all $x\in \PO$ 

2235: (\cite{RH}, p.\ 124).  For the generic point $x_0\in \PO$,

2236: we get an exact sequence of vector spaces over the field $K(y)$.

2237: $$

2238: 0\lra \O_{\P^1,x_0}\lra f_*\O_{C,x_0}\lra \ME_{x_0} \lra 0

2239: $$

2240: as desired.  For non-generic $x\in\PO$, we get a corresponding short

2241: exact sequence, with $f_*\O_{C,x}$ a finitely generated module over

2242: $\O_{\PO,x}$, which is a discrete valuation ring, and hence a 

2243: principal ideal domain.  Clearly, $f_*\O_{C,x}$ is torsion free, and hence free

2244: ([L], p.\ 147).  Likewise, one can easily show that $\ME_x$ is torsion free,

2245: and hence free.

2246: 

2247: Every locally free sheaf on $\P^1$ is of the form $\oplus_i\O_{P_1}(a_i)$,

2248: $a_i\in\Z$ (\cite{RH}, p.\ 384).  

2249: Observe that $\ME$ has no global sections, so it must be of

2250: the form $\oplus_i\O_{P_1}(a_i), a_i < 0$.  Indeed, one has the following exact

2251: sequence.

2252: $$

2253: 0\lra H^0(\PO,\O_\PO)\lra H^0(\PO,f_*\O_C)\lra H^0(\PO,\ME)

2254: \lra H^1(\PO,\O_\PO)

2255: $$

2256: Since $h^0(\PO,f_*\O_C)=1$ and  $h^1(\PO,\O_\PO)=\textup{genus}(\PO)=0$,

2257: we get $h^0(\PO,\ME)=0$, as desired.  This completes the proof of the claim.

2258: 

2259: Finally, let $\MF$ be an invertible sheaf on $\PO$ with $h^0(\PO,\MF)=0$,

2260: so $\MF=\O_\PO(b), b < 0$.  By above, to show that $h^0(C,f^*\MF)=0$,

2261: it suffices to show $h^0(C,\MF\otimes f_*\O_C)=0$.  But

2262: $$

2263: \MF\otimes f_*\O_C=\O_\PO(b)\otimes(\O_{\P^1}\oplus(\oplus_i\O_{P_1}(a_i)))=

2264: \O_\PO(b) \oplus (\oplus_i\O_{P_1}(a_i+b))

2265: $$

2266: which obviously has no global sections, 

2267: completing the proof of the proposition.

2268: \qed

2269: 

2270: We introduce a second, similar equivalence relation on \taufs\

2271: that looks somewhat less natural but will be more useful for

2272: later applications.  

2273: 

2274: \bdf

2275: Let $C$ be a curve, $\omt \in \Omt(C)$.  The {\em unary set} of $\omt$ is 

2276: $$

2277: M_{\omt} = \{(a, u): a \in C-Z_{\omt}, u\in \tau C_a \text{ and }\omt_a(u) = 1\}.

2278: $$

2279: This is a rational section of the algebraic variety $\tau C$,

2280: and thus birational to $C$.

2281: 

2282: 

2283: Say that $\omt_1, \omt_2 \in \Omt (C)$ are {\em $\seq$-equivalent},

2284: written $\omt_1\teq\omt_2$, if

2285: $M_{\omt_1}$ and $M_{\omt_2}$ are almost equal.

2286: \edf

2287: 

2288: \bla

2289: \label{teqcl}

2290: Let $C$ be a curve, $\omt_1, \omt_2 \in \Omt(C)$, both non-trivial.

2291: Then $\omt_1$ and $\omt_2$ are $\seq$-equivalent if and only if there is a rational 

2292: function $f\in K(C)$ such that $\omt_1=f(\omt_2-1) + 1$.

2293: \ela

2294:  

2295: \prf

2296: If $\omt_1=f(\omt_2-1) + 1$, then it is clear $M_{\omt_1}$ and

2297: $M_{\omt_2}$ are almost equal.  The other direction follows as

2298: with $\seq$-equivalence.

2299: \qed

2300:   

2301: \bla

2302: \label{taunull}

2303: Let $C$ be a curve, $\omt_1, \omt_2 \in \Omt(C)$, both non-trivial.  

2304: Suppose that $M_{\omt_1} \cap M_{\omt_2}$ is infinite.  Then 

2305: $\omt_1$ and $\omt_2$ are $\seq$-equivalent.  In other words,

2306: if  $M_{\omt_1} \cap M_{\omt_2}$ is infinite, then 

2307: $M_{\omt_1}$ and $M_{\omt_2}$ are almost equal.

2308: \ela

2309: 

2310: \prf

2311: Like the proof of Lemma~\ref{sigmanull}.

2312: \qed

2313: 

2314: %\bqu

2315: %Are the global $\tau$-forms on any abelian variety exactly the invariant 

2316: %$\tau$-forms, 

2317: %since any abelian variety is a quotient of a Jacobian variety [I think]?

2318: %\equ

2319: 

2320: To prove the second statement, we first determine $\ker(\dltl)$ and $\ker(\be)$.

2321: Setting $R = K$ in the above diagram, one obtains

2322: $$

2323: \xymatrix{

2324: 0 \sr[r] &\Om_K \ar[r]^\alpha \ar[d]^{\tilde{\dl}} & \Om_K \ar[r] 

2325: \ar[d]^\beta & \Om_{K/K}\ar[r]\ar[d]^{=} & 0 \\

2326: 0 \ar[r] &K \ar[r]^{\iota} & \Omt_K \ar[r]^{\lm} & \Om_{K/K} \ar[r] & 0

2327: }

2328: $$

2329: with $\Om_{K/K}=0$, $\dltl$ and $\beta$ surjective. 

2330: By definition, $\ker(\be)=\< \dl(a)db-\dl(b)da | a,b\in K\>\sub \Om_K$,

2331: which we call $M$.  By the Snake Lemma, $\ker(\dltl)= M$ also,

2332: so one gets a short exact sequence,

2333: $0\lra M \lra \Om_K \lra K \lra 0$.

2334: 

2335: Now let $R$ be any $K$-algebra.  Tensoring the previous sequence with $R$,

2336: one gets, $0\lra R\ot_K M \lra R \ot_K\Om_K \lra R \lra 0$.  By the definition

2337: of $\Omt_R$, one also has $0 \lra R\ot_KM\lra\Om_R\lra\Omt_R\lra 0$.

2338: By Lemma~\ref{tdfeq}, these sequences fit together to form the following

2339: diagram.

2340: $$

2341: \xymatrix{

2342: 0 \ar[r] & R \ot_K M \ar[r] \ar[d]^= & R\ot_K \Om_K \ar[r]^\dltl \ar[d]^\al

2343: & R \ar[r] \ar[d]^\iota & 0 \\

2344: 0\ar[r] & R \ot_K M \ar[r] & \Om_R \ar[r]^\be & \Omt_R \ar[r] & 0  

2345: }

2346: $$

2347: By the Snake Lemma again, $\Ker(\al) \cong \Ker (\iota)$, which completes the 

2348: proof.

2349: 

2350: There is also hidden my original construction of $\tau$-differentials.

2351: 

2352: 

2353: \brm 

2354: There is a more algebraic way to construct the module of 1-forms

2355: on a variety, via the notion of K\"{a}hler differentials.

2356: Let $R = K[V]$ be a $K$-algebra, e.g.\ the ring of regular functions on an

2357: affine  $K$-variety.

2358: Then one defines the $R$-module $\Omega_{R/K}$ as a certain universal object.

2359: For more information, see Eisenbud~(\cite{Eis}, Chapter 16)

2360: or Hartshorne~(\cite{RH}, Chapter~II.8).

2361: %

2362: We introduce K\"{a}hler $\tau$-differentials, which should

2363: provide a purely algebraic construction of \taufs.  I will develop

2364: %\marginpar{Expand.}

2365: and investigate these ideas later.

2366: \erm

2367: %   

2368: Let $A$ be a ring, $B$ an $A$-algebra, and $M$ a $B$-module. Recall that a 

2369: derivation is an abelian group homomorphism $d:B\lra M$ satisfying the Leibniz 

2370: rule, $d(bc)=bdc+cdb$.  The derivation is $A$-linear if it is a homomorphism of 

2371: $A$-modules.

2372: %  

2373: \bdf  

2374: Let $(A,\dl)$ be a differential ring and $B$ an $A$-algebra.

2375: \ben

2376: \item  

2377: A $B$-$\tau$-module is a pair $M^\tau=(M,\iota_M)$ such that $M$ is a

2378: $B$-module and $\iota_M$ is an injective $B$-module map from $B$ itself

2379: to $M$, $\iota_M : B\lra M$.  (Thus, a $B$-$\tau$-module is a 

2380: $B$-module together with a distinguished free rank 1 submodule of $M$.)

2381: % 

2382: Given two $B$-$\tau$-modules $M_0^\tau$ and $M_1^\tau$, a 

2383: homomorphism $f$

2384: from $M_0^\tau$ to $M_1^\tau$ is a $B$-module homomomorphism from $M_0$

2385: to $M_1$ such that for all $b\in B$, $f\circ \iota_{M_0} (b)=\iota_{M_1}(b)$.

2386: %

2387: \item 

2388: A derivation $\tau$ from $B$ to a $B$-$\tau$-module $M^\tau$ is

2389: $A$-quasilinear if for all $a\in A$, $\tau a = \iota (\dl (a))$.

2390: %

2391: \een

2392: \edf

2393: %    

2394: The following lemma explains the choice of the term `quasilinear'.

2395: (The formula for $\tau(ab)$ bears a certain resemblance to Pillay \&

2396: Ziegler's definition of a $\partial$-module in their paper,

2397: beginning of Section 3.  They have an additive endomorphism $D:V\lra V$

2398: such that for $c\in K$, $v \in V$, $D(cv)= cDv+\dl(c)v$.)

2399: %

2400: \bla  Let $(A,\dl)$ be a differential ring, $B$ an $A$-algebra, $M^\tau$ a 

2401: $B$-$\tau$-module, and $\tau :B\lra M^\tau$ an $A$-quasilinear derivation.

2402: Then for $a\in A$, $b\in B$, $\tau(ab)=a\tau b + \dl (a) \iota b$.

2403: %

2404: In particular, if $\dl$ is the trivial derivation on $A$, then

2405: $\tau$ is an $A$-linear derivation.

2406: \ela

2407: %  

2408: \prf

2409: Let $a\in A, b \in B$.  Then $\tau(ab)=$

2410: $$

2411: a\tau(b) + b\tau(a)=

2412: a\tau b + b\iota(\dl(a))=

2413: a\tau b + \iota(b\dl(a))=

2414: a\tau b + \dl(a)\iota(b).

2415: $$

2416: %

2417: The second statement follows immediately.

2418: %

2419: \qed

2420: %

2421: \bdf

2422: The module of K\"{a}hler $\tau$-differentials of $B$ over $A$,

2423: written $\Omt_{A/B}$, is the $B$-$\tau$-module generated by 

2424: $\{\tau b : b \in B\} \cup \{\iota b : b \in B\}$, with the following relations

2425: holding for all $a\in A$, $b,b' \in B$.

2426: $$

2427: \tau(bb')=b\tau b' + b'\tau b

2428: $$

2429: %  

2430: $$

2431: \tau(a)=\iota(\dl(a))

2432: $$  

2433: The map $\tau : B \lra\Omt_{A/B}$, sending $b$ to $\tau b$ is an 

2434: $A$-quasilinear derivation, called the universal

2435: $A$-quasilinear derivation.

2436: \edf

2437: %

2438: The next lemma follows immediately from the previous definition.

2439: %  

2440: \bla

2441: $\Omt_{A/B}$ has the expected universal property.  That is,

2442: given a $B$-$\tau$-module $M^\tau$ and a $\tau$-derivation

2443: $\sigma: B \lra M$, there is a unique homomorphism of $B$-$\tau$-modules 

2444: $\phi: \Omt_{A/B}\lra M$ such that $\sigma=\phi\circ\tau$.

2445: %

2446: Equivalently, there is a natural bijection,

2447: $$

2448: \text{Der}^\tau_B(B,M)\cong\text{Hom}^\tau(\Omt_{B/A},M).

2449: $$

2450: \ela  

2451: %    

2452: The significance of the following lemma will become clearer after 

2453: the introduction below of the $\Lm$-map from \taufs\ to 1-forms

2454: in Lemma~\ref{lambda}.  The proof is immediate.

2455: %  

2456: \bla 

2457: Let $(A,\dl)$ be a differential ring, $B$ an $A$-algebra.  There 

2458: is a canonical homomorphism $\widehat{\Lambda}$ from $\Omt_{B/A}$ to $\Om_{B/A}$, 

2459: that sends each $\tau b \in \Omt_{B/A}$ to $db \in \Om_{B/A}$ and has kernel 

2460: $\iota (B)$.

2461: %

2462: Thus, there is a $B$-module isomorphism, $\Omt_{B/A}\cong\Om_{B/A}\oplus B.$

2463: \ela

2464: }

2465: %% end ignore

2466: 

2467: \bibliography{jsl-smsets.bib}

2468: \bibliographystyle{alpha}

2469: 

2470: \end{document}

2471:  

2472: