math0512170/a1.tex
1: \documentclass[12pt]{amsart}
2: 
3: \usepackage[T2A,OT1]{fontenc}
4: \usepackage[ot2enc]{inputenc}
5: \usepackage[russian,english]{babel}
6: 
7: 
8: \usepackage{amssymb}
9: \usepackage{amscd}
10: 
11: % slovak style soft l for transcription of russian << l' >>
12: \def\softl{l\kern-0.25ex\raise0.07ex\hbox{'}\kern-0.20ex}
13: \def\Drinfeld{{Drinfe\softl d }}
14: 
15: % arrows
16: \newcommand{\isomto}{\overset{\sim}{\rightarrow}}
17: \newcommand{\isomfrom}{\overset{\sim}{\leftarrow}}
18: \newcommand{\surjto}{\twoheadrightarrow}
19: \newcommand{\injto}{\hookrightarrow}
20: 
21: \newcommand{\plim}{\mathop{\varprojlim}\limits}
22: 
23: % blackbord boald by double letters
24: \def\CC{{\mathbb{C}}} %complex numbers
25: \def\RR{{\mathbb{R}}} %reals
26: \def\QQ{{\mathbb{Q}}} %rationals
27: \def\ZZ{{\mathbb{Z}}} %integers
28: \def\NN{{\mathbb{N}}} %nonnegatives
29: \def\FF{{\mathbb{F}}} %finite field
30:  \def\Fp{{\mathbb{F}_p}} %.. of p elements
31:  \def\Fq{{\mathbb{F}_q}} %.. of q elements
32: \def\GG{{\mathbb{G}}} %algebraic group
33:  \def\Ga{{\mathbb{G}_{\rm a}}} %additive
34:  \def\Gm{{\mathbb{G}_{\rm m}}} %multiplicative
35: \def\PP{{\mathbb{P}}} %projective space
36: \def\AA{{\mathbb{A}}} %affine space / adeles
37: 
38: 
39: % caligraphic by lower c
40: \def\Dc{{\mathcal{D}}} %sheaf
41: \def\Ec{{\mathcal{E}}} %sheaf
42: \def\Oc{{\mathcal{O}}} %sheaf
43: 
44: % other caligraphics:
45: \def\Ell{{\mathcal{E}\ell\ell}} % D-elliptic sheaves
46: 
47: % bold by single letter
48: \def\C{{\mathbf{C}}}
49: 
50: % standard functions and functors
51: \def\H{{\rm H}} %homology
52: \def\SL{{\rm SL}} %algebraic groups
53: \def\GL{{\rm GL}} %algebraic groups
54: \def\M{{\rm M}} %matrix ring
55: \def\Spec{{\rm Spec}} 
56: \def\Proj{{\rm Proj}}
57: \def\Lie{{\rm Lie}} 
58: \def\Cl{{\rm Cl}}
59: \def\ord{{\rm ord}}
60: \def\rn{{\rm rn}}
61: \def\rk{{\rm rk}}
62: \def\ab{{\rm ab}}
63: \def\op{{\rm op}}
64: \def\del{{\partial}}
65: \def\tr{{\rm tr}}
66: \def\deg{{\rm deg}}
67: \def\End{{\rm End}}
68: \def\Hom{{\rm Hom}}
69: \def\Res{{\rm Res}}
70: \def\Gal{{\rm Gal}}
71: \def\Aut{{\rm Aut}}
72: \def\Sh{{\rm Sh}}
73: \def\an{{\rm an}}
74: \def\un{{\rm un}}
75: \def\sep{{\rm sep}}
76: \def\Bad{{\rm Bad}}
77: \def\id{{\rm id}}
78: \def\pr{{\rm pr}}
79: \def\i{{\imath}}
80: 
81: \swapnumbers
82: \theoremstyle{plain}
83: \newtheorem{theorem}[subsection]{Theorem}
84: \newtheorem{corollary}[subsection]{Corollary}
85: \newtheorem{lemma}[subsection]{Lemma}
86: \newtheorem{maintheorem}[subsection]{Main Theorem}
87: \newtheorem{proposition}[subsection]{Proposition}
88: \newtheorem{definition}[subsection]{Definition}
89: \theoremstyle{definition}
90: \newtheorem{example}[subsection]{Example}
91: \newtheorem{remark}[subsection]{Remark}
92: \newtheorem{question}[subsection]{Question}
93: \newtheorem{conjecture}[subsection]{Conjecture}
94: \newtheorem{hypothesis}[subsection]{Hypothesis}
95: \newtheorem{problem}[subsection]{Problem}
96: \newtheorem{todo}[subsection]{TODO}
97: \newtheorem{void}[subsection]{}
98: 
99: \begin{document}
100: 
101: \markboth{Lenny Taelman}
102:  {$\Dc$-elliptic sheaves and uniformisation}
103: 
104: 
105: \title
106: {$\Dc$-elliptic sheaves and uniformisation}
107: 
108: 
109: \author{Lenny Taelman}
110: \email{lenny@math.rug.nl}
111: 
112: \maketitle
113: 
114: %NOTATIE
115: % 
116: %
117: %
118: 
119: \thispagestyle{empty}
120: 
121: \section{Introduction}
122: 
123: This text seeks to expose a function field analogue - in the
124: sense of $\Fp[t]$ versus $\ZZ$ - to 
125: objects such as modular curves and quaternionic Shimura curves.
126: 
127: Over the field of complex numbers, these curves permit a dual
128: description as either algebraic varieties or Riemann surfaces. The
129: former is obtained as a parameter space for certain Abelian varieties,
130: the latter as a quotient space of a homogeneous space, to wit the complex upper half plane, by a discrete group action. 
131: 
132: Modelled after the classical modular curves are the \Drinfeld modular
133: curves and the higher dimensional \Drinfeld modular varieties.
134: Similarly, they can be treated either algebraically or analytically  (see \cite{Drinfeld74}). A wider class of
135: varieties, containing at once the \Drinfeld modular varieties and the
136: proper counterparts of the quaternionic Shimura curves, consists of
137: the moduli spaces of Laumon, Rapoport and Stuhler. They are
138: algebraic varieties that parametrise certain algebraic objects,
139: namely $\Dc$-elliptic sheaves (see \cite{Laumon93}). However,
140: the dual, analytic, description is absent from the literature.
141: 
142: Thus, in a slightly more precise formulation, this manuscript
143: aims to give the moduli spaces for $\Dc$-elliptic sheaves an
144: analytic description as quotients of 
145: homogeneous spaces. This proceeds roughly in three steps. First,
146: $\Dc$-elliptic sheaves are shown to form a particular subclass of the
147: category of Anderson $t$-motives (\S 2, proposition
148: \ref{equivalence}). Second, these Anderson $t$-motives
149: are proven to be \emph{uniformisable}, meaning that they can be
150: obtained as quotients of vector spaces by lattices, and the class of
151: lattices so obtained is determined (\S 3,
152: propositions \ref{uniformizability} and \ref{fullness}). Finally, these
153: lattices are parametrised by quotients of homogeneous spaces; this
154: yields the main result (\S 4, theorem \ref{thm}).
155: 
156: This uniformisation of the varieties of Laumon, Rapoport and Stuhler
157: at the ``infinite'' valuation complements recent work of Hausberger
158: (\cite{Hausberger05}). In the spirit of \Drinfeld and Cherednik's
159: $p$-adic uniformisation of Shimura curves, he treats the
160: analytic structure at the ``ramified'' places. Hausberger has confided
161: to the author that he has also obtained, but not published, a
162: uniformisation at the ``infinite'' valuation.
163: 
164: 
165: \section{$\Dc$-elliptic sheaves and Anderson $A$-motives}
166: 
167: \begin{void}
168: \marginpar{\tiny standard Drinfeld data} % standard Drinfeld data
169: The following data are fixed. A complete, smooth and
170: geometrically irreducible curve $X$ over a finite field $\Fq$ of
171: characteristic $p$ and a closed
172: point $\infty\in X$ called infinity. The degree of the residue
173: class field of $\infty$ over $\Fq$ is denoted by $\alpha$. The
174: function field of $X$ is denoted by $F=\Fq(X)$ and the subring of
175: functions which are regular
176: everywhere except possibly at infinity by $A=\H^0(X-\infty,\Oc_X)$.
177: On $A$, $\deg$ is defined as the pole order at infinity.
178: The completion of $F$ with respect to the infinite valuation is
179: $F_\infty$. Also, a complete and algebraically closed field $\C$
180: containing $F_\infty$ is fixed and the resulting embeddings denoted by
181: $\i:F_\infty\to \C$, $\i:F\to\C$ and $\i:A\to\C$. Of course, the kernel
182: of $\i:A\to\C$ is trivial, this corresponds to what is usually called
183: ``infinite characteristic'', and reflects the particular interest into
184: analytic structures taken in this exposition. Put differently:
185: reductions modulo primes of $A$ will not be touched upon. Finally,
186: denote by $\hat{A}$ the product of all completions $A_v$ over all
187: places $v\neq \infty$ and by
188: $\AA_A=\hat{A}\otimes_A F$ the ad\`{e}le ring away from
189: infinity.
190: \end{void}
191: 
192: \begin{void}
193: \marginpar{\tiny the central simple algebra} % the central simple algebra
194: Also, a central simple $F$-algebra $D$, unramified at $\infty$ is fixed.
195: Denote the dimension of $D$ over its centre by $d^2$. Let $L$ be an 
196: algebraically closed field containing $\Fq$, then 
197: $D\otimes_{\Fq}L\cong D\otimes_F L(X)$ is isomorphic to the full matrix algebra $\M(d,L(X))$, since $L(X)$ is a $C_1$ field.
198: 
199: Fix a subsheaf of $\Oc_X$-algebras $\Dc$ inside the constant sheaf
200: $D$ with the property that for all points $x\in X$, $\Dc_x$ is
201: a maximal order in $D_x$ (it is shown in \cite{Laumon93} that such sheaves exist.) Then $O_D:=\H^0(X-\infty,\Dc)$ is a maximal
202: $A$-order in $D$. 
203: \end{void}
204: 
205: \begin{void}
206: \marginpar{\tiny the sheaves} % the sheaves
207: Let $B$ be a $\C$-algebra and denote $X\times_{\Fq}\Spec(B)$ by
208: $X_B$. The curve $X_B$ on $\Spec(B)$ comes equipped with a
209: section $\i^\#:\Spec(B)\to X_B$ which derives from the
210: embedding $\i$ of the function field of $X$ into $\C$ and takes as
211: constant value the generic point of $X$. 
212: 
213: The projections $X_B \to X$ and $X_B\to \Spec(B)$ are denoted
214: by $\pr_X$ and $\pr_S$ respectively. The
215: pullback of the geometric $q$-Frobenius on $\Spec(B)$ to
216: $X_B$ by $\pr_X$ by $\sigma$. Thus $\sigma$ acts trivially on
217: $X$ and as the $q$-th power endomorphism on
218: the ring of functions $B$.
219: \end{void}
220: 
221: \begin{definition} \label{ellsh}
222:  A \emph{$\Dc$-elliptic sheaf} on $X_B$ is a commutative
223: diagram:
224: \[ \begin{CD}
225: \ldots @>j>> \Ec_{-1} @>j>> \Ec_0 @>j>> \Ec_1 @>j>> \Ec_2 @>j>> \ldots \\
226:   @.    @AtAA @AtAA @AtAA @AtAA \\
227: \ldots @>j>> \sigma^\ast\Ec_{-2} @>j>> \sigma^\ast\Ec_{-1} @>j>>
228: \sigma^\ast\Ec_0 @>j>> \sigma^\ast\Ec_1 @>j>>
229: \ldots \end{CD} \]
230: where the $\Ec_i$ are locally free right $\pr_X^\ast\Dc$-modules of rank
231: $1$ and  the $t$ and the $j$ are $\pr_X^\ast\Dc$-linear
232: injections. These data are constraint to the following conditions:
233: \begin{enumerate}
234:  \item \emph{Periodicity}: the composition of $d\alpha$
235:   consecutive $j$'s defines an identification
236:   $\Ec_{i+d\alpha}=\Ec_i\otimes_{\Oc_{X_B}} \
237:   (\Oc_X(\infty)\boxtimes \Oc_{\Spec(B)})$
238:   through the natural embedding $\Oc_X\injto \Oc_X(\infty)$;
239:  \item \emph{Pole}: the direct image of $\Ec_i/j(\Ec_{i-1})$ by
240:   $\pr_B$ is a locally free $\Oc_{\Spec(B)}$-module of rank
241:    $d$;
242:  \item \emph{Zero}: $\Ec_i/t(\sigma^\ast\Ec_{i-1})$ is 
243:  the direct image by $\i^\#:\Spec(B)\to X_B$ of a locally free
244:  $\Oc_{\Spec(B)}$-module.
245: \end{enumerate}
246:  A shorthand notation for such an object is $(\Ec_i,j,t)$. A
247:  \emph{morphism of $\Dc$-elliptic sheaves} on $X_B$ is a
248:  collection of $\pr_X^\ast\Dc$-linear morphisms $\Ec_i\to\Ec'_i$ 
249:  that commute with the $j$'s and the $t$'s.
250: \end{definition}
251: 
252: \begin{void}\label{decalage}
253: \marginpar{\tiny d\'{e}calage} % decalage
254:   It will be necessary to consider elliptic
255:  $\Dc$-sheaves up to \emph{d\'{e}calage}. Formally this can be done by
256:  allowing more morphisms. Namely, a morphism of  $\Dc$-elliptic sheaves up to
257:  d\'{e}calage consists of a fixed integer $n$ and a collection
258:  of $\pr_X^\ast\Dc$-linear morphisms
259:  $\Ec_i\to\Ec'_{i+n}$ commuting with the $j$'s and $t$'s.
260:  \end{void}
261: 
262: \begin{definition} 
263: \marginpar{\tiny the modules} % the modules
264:  An Anderson $A$-motive with multiplication by $O_D$ over $\C$
265:  is a left $O_D^\op\otimes_\Fq \C[\sigma]$-module $M$ satisfying
266:  \begin{enumerate}
267:   \item $M$ is projective of rank $1$ over $O_D^\op\otimes \C$
268:   \item $M$ is projective of rank $d$ over $\C[\sigma]$
269:   \item there exists a positive integer $n$ such that
270:     $(a-\i(a))^n(M/\sigma M)=0$ for all $a\in A$.
271:  \end{enumerate}
272: \end{definition}
273: 
274: \begin{void}\label{altdef}
275:  Denote by $\sigma$ the endomorphism of $O_D^\op\otimes_\Fq \C$
276:  that acts trivially on $O_D^\op$ and as the $q$-th power map
277:  on $\C$.
278:  Then an Anderson $A$-motive with multiplication by $O_D$ can
279:  also be described as a
280:  $O_D^\op\otimes \C$-module $M$ on which the action of $\sigma$ is
281:  described by either a linear morphism
282:  $\tau:\sigma^\ast M \to M$ or a semi-linear map from $M$ to
283:  itself, also (abusively) denoted by $\tau:M\to M$.
284: \end{void}
285: 
286:  As is already suggested in \cite[\S 3]{Laumon93}, there is a close
287:  relationship between $\Dc$-elliptic sheaves and Anderson
288:  $t$-motives (in this manuscript called $A$-motives). This
289:  kinship is best understood through Drinfe\softl d's
290:  theory of vector bundles on the non-commutative projective line.
291:  For more details on these objects, one can consult \S 3
292:  of \emph{loc. cit.}, which establishes most of what is needed
293:  in order to properly relate Anderson $A$-motives with
294:  multiplication by $O_D$ to $\Dc$-elliptic sheaves. In the following
295:  paragraphs the remaining steps are taken.
296: 
297: \begin{void}\label{begineq}
298: \marginpar{\tiny from sheaves to modules} % from sheaves to modules
299:  Since by the pole axiom, the subsequent quotients
300:  $\Ec_{i+1}/\Ec_i$ in
301:  a $\Dc$-elliptic sheaf $(\Ec_i,j,t)$ on $X_\C$ are
302:  supported at $\infty$,
303:  the $O_D^\op\otimes_k \C$-module $M:=\H^0(X_\C-\infty,\Ec_i)$ is
304:  independent of $i$. Therefore the
305:  $t:\sigma^\ast \Ec_i \to \Ec_{i+1}$ induce a morphism 
306:  $\tau:\sigma^\ast M\to M$ which turns the
307:  $O_D^\op\otimes_k \C$-module
308:  $M$ into a $O_D^\op\otimes_k \C[\sigma]$-module
309:  (as in \ref{altdef}). By construction,
310:  $M$ depends only on the given $\Dc$-elliptic sheaf up to
311:  d\'{e}calage. It follows from \cite[3.17]{Laumon93} that $M$
312:  is an Anderson $A$-motive with multiplication by $O_D$.
313: \end{void}
314: 
315: \begin{void}
316: \marginpar{\tiny from modules to sheaves} % from modules to sheaves
317:  The above construction is functorial and it is possible to
318:  define an inverse functor. It suffices to extend the given
319:  $\C[\sigma]$-module $M$ to a vector bundle on the non-commutative
320:  projective line with coordinate $\sigma$ and then apply the
321:  equivalence 3.17 of \emph{loc. cit.} This is tantamount to
322:  a module-theoretic construction, that will be described next,
323:  first in the rank one \Drinfeld module case, that is, for $D=F$,
324:  and then for a general $D$.
325: \end{void}
326: 
327: \begin{void}
328:  Let $D=F$. The non-commutative ring of power series in
329:  $\sigma^{-1}$, denoted by $\C[[\sigma^{-1}]]$, has the skew field
330:  of Laurent series $\C((\sigma^{-1}))$ as quotient field. Given
331:  an $M$, a Anderson $A$-motive with multiplication by
332:  $O_D=A$, under the aforementioned equivalence, the definition
333:  of a $\Dc$-elliptic sheaf $(\Ec_i,j,t)$ such
334:  that $M=\H^0(X_\C-\infty,\Ec_i)$ boils down to the definition of a
335:  free rank one $\C[[\sigma^{-1}]]$-submodule
336:   $$ W \subset \C((\sigma^{-1})) \otimes_{\C[\sigma]} M .$$
337:  The right-hand side of this inclusion is isomorphic to
338:  $\C((\sigma^{-1}))$, and the corresponding submodules are precisely
339:  those generated by a $\sigma^m$ for an integer $m$. It can be
340:  checked that every such submodule defines a vector bundle on
341:  the non-commutative projective line with the correct
342:  properties, hence also an elliptic sheaf. Different
343:  choices of $m$ define elliptic sheaves that are
344:  equivalent under d\'{e}calage.
345: \end{void}
346: 
347: \begin{void}\label{morita}
348: \marginpar{\tiny Morita equivalence} % morita equivalence
349:  Extending the above reasoning to the general case is a matter
350:  of applying Morita equivalence. As this equivalence
351:  is also central to several of the proofs yet to come, it is
352:  worthwhile to recall a precise formulation here.
353:  Let $R$ be a ring with unity. Then to every $R$-module $N$ it is
354:  possible to associate the $\M(n,R)$-module $N\oplus N\oplus \cdots
355:  \oplus N$, the direct sum of $n$ copies of $N$.
356:  \emph{Morita equivalence} states that this
357:  defines an equivalence from the category of  $R$-modules to
358:  the category of $\M(n,R)$-modules, and that this equivalences
359:  preserves such properties as projectivity, injectivity, and
360:  finite generation. Also, the construction is functorial in $R$.
361:  A detailed exposition can be found in \cite[\S 7]{Lam99}.
362: \end{void}
363:  
364: 
365: \begin{void}\label{endeq}
366: \marginpar{\tiny from modules to sheaves (continued)}
367:                            % from modules to sheaves (continued)
368:  The restriction on $D$ is now dismissed. Now
369:  $\C((\sigma^{-1})) \otimes_{\C[\sigma]} M$ is a module over
370:  $F_\infty \otimes_A O_D^\op$ and the submodule
371:   $$ W \subset \C((\sigma^{-1})) \otimes_{\C[\sigma]} M $$
372:  is required to be isomorphic to $\C[[\sigma^{-1}]]^d$ and to
373:  be equipped with a continuous action by
374:  $\Dc_\infty\subset F_\infty\otimes_A O_D^\op$. Since $\Dc_\infty$
375:  is a matrix algebra inside the matrix algebra
376:  $F_\infty\otimes_A O_D^\op$, the machinery of Morita applies.
377:  It turns the set of such $W$ into the set of
378:  $\C[[\sigma^{-1}]]$-submodules of $\C((\sigma^{-1}))$. As above,
379:  these are generated by some $\sigma^m$ and different $m$ define
380:  different representatives of the same class of 
381:  $\Dc$-elliptic sheaves up to d\'{e}calage.
382: \end{void}
383: 
384: The above discussion (\ref{begineq}-\ref{endeq}) sums up to:
385: 
386: \begin{proposition}\label{equivalence}
387:  The category of $\Dc$-elliptic sheaves on $X_\C$ up to
388:  d\'{e}calage is equivalent to the category of
389:  Anderson $A$-motives with multiplication by $O_D$.
390: \end{proposition}
391: 
392: 
393: \begin{example}
394: \marginpar{\tiny \Drinfeld modules} %  drinfeld modules
395:  If $O_D$ is the algebra of $d\times d$ matrices over $A$ then Morita
396: equivalence implies the equivalence of Anderson $A$-motives with
397: multiplication by $O_D$ and Anderson $A$-motives of dimension $1$ and
398: rank $d$. Thus the theory of $\Dc$-elliptic sheaves includes the
399: \Drinfeld modules as a special case.
400: \end{example}
401: 
402:  
403: \begin{void} 
404: \marginpar{\tiny level structures} % level structures 
405: Given an effective divisor $I$ on $X-\infty$ it is possible to define
406: level $I$-structures on $A$-motives with multiplication by $O_D$
407: as well
408: as on $\Dc$-elliptic sheaves such that the equivalence between the two
409: categories extends to an equivalence of the categories ``with level
410: structure''. For the sake of clarity, the results in this text
411: are first stated and proved ignoring level structures, and
412: the corresponding propositions with level structure are simply
413: stated as corollaries. The omitted proofs are never hard.
414: \end{void}
415: 
416: The $I$ in the succeeding definitions refers in \emph{abus de
417: notation} to both the divisor $I$ on $X$ and the corresponding
418: ideal $I\subset A$. 
419: 
420: \begin{void} 
421: \marginpar{\tiny level structures on sheaves} % level structures on sheaves
422: Since $\infty$ and $I$  are disjoint, the
423: restriction $\Ec_I$ of the $X_B$-sheaf $\Ec_i$ to $I_B$ is
424: independent of $i$ and for the same reason the injection $t$ restricts
425: to an isomorphism $t:\sigma^\ast\Ec_I\to \Ec_I$.  
426: A \emph{level $I$-structure on a $\Dc$-elliptic sheaf $(\Ec_i,j,t)$}
427: is an isomorphism of $\Oc_{I_B}$-sheaves:
428: $\Dc_I\boxtimes\Oc_B \to \Ec_I$ such
429: that the Frobenius $\sigma^\ast \Oc_B\to \Oc_B$ corresponds to 
430: $t:\sigma^\ast\Ec_I\to \Ec_I$.
431: \end{void}
432: 
433: \begin{void} 
434: \marginpar{\tiny level structures on modules} % level structures on modules
435:  A \emph{level $I$ structure on an Anderson $A$-motive
436:  with multiplication by $O_D$} is the choice of an isomorphism
437:  $(O_D^\op/IO_D^\op)\otimes_{\Fq} \C\to M/IM$
438:  such that the Frobenius on $\C$ matches the action of $\sigma$
439:  on $M/IM$.
440: \end{void} 
441: 
442: 
443: 
444: \section{Uniformisation of the Anderson $A$-motives}
445: 
446: \marginpar{\tiny introduction} % introduction
447: Whereas all Abelian varieties over the field of the complex numbers
448: can be obtained as the quotient of a complex vector space by a
449: lattice, the parallel statement for Anderson $A$-motives is false
450: in general. Those Anderson $A$-motives that \emph{can} be obtained
451: in such a way are called  \emph{uniformisable}. It 
452: will be shown in this section that the $A$-motives considered above,
453: \emph{videlicet} the $A$-motives with multiplication by $O_D$, are
454: in fact uniformisable.
455: 
456: \begin{void}
457: \marginpar{\tiny on uniformisability} % on uniformisability
458:  A very brief overview of the results of \cite{Anderson86}
459:  is given. There is an anti-equivalence of categories that
460:  assigns to an Anderson $A$-motive $M$ of dimension $d$,
461:  an action of $A$ on an algebraic group $E(M)\cong\GG_a^d$,
462:  satisfying certain properties. This $E(M)$ occurs naturally
463:  in an exact sequence of \emph{analytic} $A$-modules:
464: \[ \begin{CD}
465:   \Lambda @>>> \Lie(E(M)) @>{\exp_{E(M)}}>> E(M)
466:   \end{CD} \]
467:  Here $\Lie(E(M))$ is a $d$-dimensional vector space over $\C$
468:  on which $A$ acts by the embedding $\i:A\injto \C$, and $\Lambda$
469:  is a lattice, \emph{i.e.} a discrete and projective $A$-submodule.
470:  The Anderson $A$-motive $M$ can be recovered from this lattice
471:  if $\exp$ is surjective. There exist, however, $M$ for which
472:  this is not the case. $M$ is called \emph{uniformisable}
473:  when $\exp_{E(M)}$ \emph{is} surjective and by
474:  \cite[\S 2]{Anderson86} this
475:  is the case precisely when the rank of $\Lambda$ equals the rank
476:  of $M$ (in general $\rk_A(\Lambda)\leq \rk_{A\otimes\C}(M)$). \end{void}
477: 
478: \begin{void}
479: \marginpar{\tiny a criterion for uniformisability}
480:                          % a criterion for uniformisability
481:  Fix a subring $\Fp[a]\subset A$ such that $a$ has a pole of order
482:  prime to $p$ at $\infty$ and such that the resulting fraction
483:  field extension $F/\Fp(a)$ is separable. If the rank of $A$
484:  over $\Fp[a]$ is
485:  $r$, then an Anderson $A$-motive with multiplication by $O_D$ is an
486:  Anderson $\Fp[a]$-motive of rank $rd^2$ and dimension $d$. Given an
487:  Anderson $A$-motive over $\C$ it is now possible to extend its
488:  scalars from $\C[a]$ to $\C\{\{a\}\}$ - the ring of power series with
489:  convergence radius at least $1$ - which yields a module
490:  $\tilde{M}=M\otimes_{\C[a]}\C\{\{a\}\}$. The semilinear action
491:  $\tau:M\to M$ extends to a semilinear
492:  $\tilde{\tau}:\tilde{M}\to\tilde{M}$ and $M$ is said to be
493:  \emph{analytically trivial} if $\tilde{M}$ has a $\C\{\{a\}\}$-basis
494:  consisting of $\tilde{\tau}$-invariant vectors. This is all
495:  that is needed to
496:  state Anderson's criterion for uniformisability
497:  \cite[\S 2]{Anderson86}. Namely, it states that $M$ is
498:  uniformisable if and only if it is analytically trivial. 
499: \end{void}
500: 
501: 
502: \begin{proposition}\label{uniformizability}
503: \marginpar{\tiny uniformisability of the modules}
504:                             % uniformisability of the modules
505:  Over $\C$, all $d$-dimensional $A$-motives with multiplication
506:  by $O_D$ are uniformisable.
507: \end{proposition}
508: 
509: \begin{proof}
510:  Fix an isomorphism
511:  $O_D^\op\otimes_{\Fq}\C\cong \M(d,A\otimes_\Fq\C)$.
512: 
513:  Let $M$ be  an $A$-motive as in the proposition. Then $M$ is a
514:  projective $O_D^\op\otimes \C$-module of rank one together with a 
515:  semilinear map $\tau:M\to M$ commuting with the action
516:  of $O_D^\op$
517:  and having a determinant equal to $(a-\i(a))^d$ up to a unit.
518:  Morita equivalence associates to these data a module $N$ of rank
519:  $n/d$ over $A\otimes \C$ equipped
520:  with a semilinear $\tau_N:  N\to N$ of determinant a
521:  unit times $(a-\i(a))$ such that $M$ can be recovered from $N$
522:  as a $d$-fold direct sum: $M\cong N\oplus N \oplus \cdots \oplus N$. 
523:  The $A$-motive $N$ has dimension $1$, thus determines a
524:  \Drinfeld module. \Drinfeld modules are uniformisable
525:  (\cite{Drinfeld74}), and it
526:  follows from Anderson's analytic triviality criterion that $N$ is
527:  analytically trivial. This shows that
528:  $\tilde{M}=\tilde{N}\oplus\tilde{N}\oplus\cdots\oplus\tilde{N}$
529:  has a $\tilde{\tau}$-invariant $\C\{\{a\}\}$-basis, hence $M$
530:  is uniformisable.
531: \end{proof}
532: 
533: \begin{void}
534: \marginpar{\tiny lattices and representations} % lattices and representations
535: Thus, over $\C$, to every $d$-dimensional $A$-motive $M$ with
536: multiplication by $O_D$, or to every $\Dc$-elliptic sheaf on
537: $X_\C$, a discrete lattice $\Lambda$ of $A$-rank $r^2$ in
538: $V:=\Lie(E(M))$ is assigned. This assignment is a faithful functor
539: on the category of uniformisable $A$-motives
540: \cite[2.12.2]{Anderson86}, and this immediately implies that
541: $\Lambda\subset V$ is more than just an $A$-lattice in a
542: $\C$-vector space: it naturally carries the structure of an $O_D$-lattice
543: in a faithful $\C$-linear representation of $D$. The following proposition 
544: confirms that every pair $\Lambda\subset V$ of a rank one
545: $O_D$-lattice in a faithful
546: $d$-dimensional representation of $D$ occurs in such a way.
547: \end{void}
548: 
549: \begin{proposition}\label{fullness}
550: \marginpar{\tiny surjectivity of the construction} % surjectivity of the construction
551:  Let $\Lambda\subset V$ be a discrete projective
552:  and rank one $O_D$-submodule of a $d$-dimensional faithful linear
553:  representation of $D$. There exists a uniformisable Anderson $
554:  A$-motive with multiplication by $O_D$ and
555:  an isomorphism $\Lie(E(M))\to V$ that surjects $\ker(\exp_{E(M)})$
556:  onto $\Lambda$.
557: \end{proposition}
558: 
559: \begin{proof}
560:  By \cite[corollary 3.5.1]{Anderson86}, it is sufficient to find
561:  an $A$-lattice
562:  $\Lambda'\subset V$ whose $F_\infty$-span equals the $F_\infty$-span of
563: $\Lambda$ and which is the kernel of the exponential function of some
564: uniformisable abelian $A$-module. Since $D$ is unramified at the infinite place,
565: $D\otimes F_\infty$ is isomorphic to $\M(d,F_\infty)$, hence the
566: $F_\infty$-span of $\Lambda$ splits under the action of the diagonal
567: idempotents as a direct sum of $d$ terms each lying within a one-dimensional linear
568: subspace of $V$. Each such term is the span of a suitable
569: lattice in a one-dimensional linear space and the direct sum of these
570: lattices is the $\Lambda'$ which is being sought for. It is the kernel
571: of the exponential function of an abelian $A$-module which is
572: uniformisable because it is the
573: direct sum of $d$ suitable \Drinfeld modules.
574: \end{proof}
575: 
576: \begin{corollary}\label{eq2}
577:  The category of Anderson $A$-motives with multiplication by $O_D$
578:  is anti-equivalent with the category of pairs $(V,\Lambda)$, where
579:  $V$ is a faithful $d$-dimensional representation of $D$ and
580:  $\Lambda\subset V$ an $O_D$-lattice.
581: \end{corollary}
582: 
583: 
584: 
585: 
586: \section{Analytic moduli spaces of $\Dc$-elliptic sheaves}
587: 
588: In this section, all $\Dc$-elliptic sheaves are considered up to
589: d\'{e}calage (see \ref{decalage}).
590: 
591: The requisite preparations have now been made in order to
592: give the varieties of Laumon, Rapoport and Stuhler an analytic
593: description. The main ingredient is the classification of the 
594: pairs $(V,\Lambda)$ of corollary \ref{eq2}. These are classified
595: by quotients of the \Drinfeld symmetric space
596: $\Omega^d(\C)=\PP^{d-1}(\C)-\PP^{d-1}(F_\infty)$, in very much
597: the same way that some Shimura varieties are constructed as
598: quotients of the complex upper half plane. In what follows
599: $D^\ast(R)$ denotes $(D\otimes_F R)^\ast$, for an $F$-algebra $R$.
600: 
601: \begin{proposition}\label{classlat}
602: \marginpar{\tiny classification of lattices}
603:                             % classification of lattices
604:  There is a natural bijection between the set of isomorphism
605:  classes of pairs $(V,\Lambda)$ of $d$-dimensional $\C$-linear
606:  representations of $D$ with $O_D$-lattice $\Lambda\subset V$ and
607:  the double co-sets in
608:   \[ D^\ast \backslash
609:    \left[\Omega^{d}(\C) \times D^\ast(\AA_A)  /
610:     U \right] \]
611:   where $D^\ast$ acts diagonally on the product and in particular on
612:   $\Omega^{d}(\C)$ by the choice of an isomorphism
613:   $D\otimes_F F_\infty\cong \M(d,F_\infty)$ and where $U$ is
614:   the compact open subgroup
615:   $(O_D\otimes_A\hat{A})^\ast\subset D^\ast(\AA_A)$.
616: 
617: \end{proposition}
618: 
619: \begin{proof}
620:  Start of with a pair $(V,\Lambda)$. Since all
621:  representations are conjugate, it is possible to assume without
622:  loss of generality that
623:  $V=F_\infty^d\otimes \C=\C^d$ on which $D$ acts by a fixed
624:  isomorphism $D\otimes_A F_\infty \cong \M(d,F_\infty)$.
625:  Consider the $F$-span
626:  $F\Lambda$ of the lattice. This is a free module of rank $1$ over
627:  $D$ lying inside $V$ and the choice of a generator
628:  marks a point on $\PP^{d-1}(\C)$ and identifies $F\Lambda$ with
629:  $D$. The marked point lies in
630:  $\Omega^d(\C)\subset\PP^{d-1}(\C)$ by the discreteness of $\Lambda$.
631:  The embedding $\Lambda\subset D$ can be tensored to an
632:  embedding $\hat{A}\Lambda \subset D\otimes \AA_A$
633:  and the former can be recovered from the latter as
634:  $\Lambda=\hat{A}\Lambda \cap D\otimes 1$. But all projective
635:  $O_D\otimes\hat{A}$-modules are free, consequently the projective
636:  $O_D$-submodules $\Lambda\subset D$ of rank one are in
637:  bijection with the free rank
638:  one $O_D\otimes\hat{A}$-submodules of $D\otimes\AA_A$ and the latter
639:  are in bijection with
640:  $(D\otimes\AA_A)^\ast/(O_D\otimes\hat{A})^\ast$. It remains to
641:  mod out by the choice of the generator of $F\Lambda$, \emph{i.e.}
642:  by $D^\ast$, to establish the desired one-to-one
643:  correspondence. 
644: \end{proof}
645: 
646: \begin{void}
647: \marginpar{\tiny analysis of the proposition}
648:                             % analysis of the proposition
649: Chaining the last proposition with the equivalences \ref{eq2} and
650: \ref{equivalence} furnishes a bijection between the set
651: of isomorphism classes of $\Dc$-elliptic sheaves on $X_\C$
652: and the above double co-sets. This double co-set space inherits the
653: structure of a rigid analytic space from the rigid analytic
654: space $\Omega^d$ (see \cite{Drinfeld74}). So, at least point-wise,
655: $\Dc$-elliptic sheaves are classified by a rigid analytic space, but
656: this set-theoretic result is not enough to give an analytic structure
657: to the algebraic moduli spaces of Laumon, Rapoport and Stuhler.
658: \end{void}
659: 
660: \begin{void}
661: \marginpar{\tiny analytic moduli} % analytic moduli
662: A close examination of the proof of proposition \ref{classlat}
663: shows that it establishes a slightly stronger result. In fact,
664: it demonstrates that every analytic family of pairs $(V,\Lambda)$
665: over a rigid analytic $\C$-space $Y$ - the definition of such a 
666: family should be clear - results in an analytic map from
667: $Y$ to the double co-set space. In fact, also this stronger statement
668: can be easily pulled through the chain of equivalences \ref{eq2} and
669: \ref{equivalence}, using the results of \cite{Boeckle05}. The
670: outcome is a procedure that assigns to every $\Dc$-elliptic sheaf
671: on $X_B$ for some $\C$-algebra $B$, a morphism of rigid
672: analytic $\C$-spaces from $\Spec(B)^\an$ to the double co-set space
673: of the proposition. 
674: \end{void}
675: 
676: \begin{void}
677: \marginpar{\tiny algebraic moduli} % algebraic moduli
678:  In \cite{Laumon93} it is shown that when $I\neq A$, 
679:  the functor that ``maps'' a $\C$-algebra $B$ to the set of
680:  isomorphism classes of $\Dc$-elliptic sheaves on $X_B$ is
681:  representable by a quasi-projective $\C$-scheme of
682:  dimension $d-1$, denoted by $\Ell_{X,\Dc,I}$. Taking the
683:  quotient by a finite group yields a coarse moduli
684:  scheme $\Ell_{X,\Dc,A}$ classifying 
685:  $\Dc$-elliptic sheaves without level structure. All these
686:  varieties are shown to be smooth and under the condition that
687:  $D$ be a division algebra they are shown to be complete
688:  in \emph{loc.cit.}
689: \end{void}
690: 
691: \begin{void}
692: \marginpar{\tiny relating algebraic and analytic moduli}
693:                         % relating algebraic and analytic moduli
694:   Now the above considerations define a natural map of rigid
695:   analytic spaces
696:    \[ \Ell_{X,\Dc,A}(\C)^\an \to D^\ast \backslash
697:    \left[\Omega^{d}(\C) \times D^\ast(\AA_A)  /
698:     U\right], \]
699:   which is a bijection on the sets of $\C$-valued points. Since both
700:   spaces are reduced, this has to be an isomorphism. This proves the
701:   fundamental case of the main theorem.
702: \end{void}
703: 
704: \begin{theorem}
705: \marginpar{\tiny principal case of the theorem}
706:                            % principal case of the theorem
707:  There is a natural isomorphism of rigid analytic spaces
708:    \[ \Ell_{X,\Dc,A}(\C)^\an \cong D^\ast \backslash
709:    \left[\Omega^{d}(\C) \times D^\ast(\AA_A)  /
710:     U\right]. \]
711: \end{theorem}
712: 
713: \begin{void}
714: \marginpar{\tiny free vs. projective} % free vs. projective
715:  A triple $(\Ec_i,j,t)$ is said to be \emph{free} if and only
716:  if the restrictions of the $\Ec_i$ to $X\times s$ are free for
717:  all geometric points $s$ of the base. For example, when $X$ is the
718:  projective line and $\infty$ is rational then all elliptic
719:  $\Dc$-sheaves are free. This notion
720:  readily translates to Anderson $A$-motives. $M$ being free then
721:  signifies $M$ being free as $A\otimes \C$-module.
722:  Free $\Dc$-elliptic sheaves are
723:  classified by closed and open subspaces
724:  $\Ell^0_{X,\Dc,I}\subset \Ell_{X,\Dc,I}$. 
725:   The above theorem
726:  quite easily generalises to the main result of this exposition.
727: \end{void}
728: 
729: 
730: \begin{theorem}\label{thm}
731: \marginpar{\tiny main theorem} % main theorem
732:  The analytification of the moduli spaces $\Ell^0$ and $\Ell$ is
733:  $$ \Ell^0_{X,\Dc,I}(\C)^\an \cong G(I) \backslash \Omega^{d}(\C) $$
734: and
735:  $$ \Ell_{X,\Dc,I}(\C)^\an \cong
736:    D^\ast \backslash \left[\Omega^{d}(\C)
737:    \times D^\ast(\AA_A) / U(I)\right]
738: $$
739: where $G(I)\subset O_D^\ast$ and $U(I)\subset U$ are the subgroups
740: of elements
741: that reduce to $1$ modulo $I$ and where $D^\ast$ acts on
742: $\Omega^{d}(\C)$ by the choice of an isomorphism
743: $D\otimes_F F_\infty\cong \M(d,F_\infty)$.
744: \end{theorem}
745: 
746: If $D=\M(d,F)$, then the above result boils down to the
747: uniformisation of the \Drinfeld modular varieties $M^d_I$
748: as established in \cite{Drinfeld74}. 
749: 
750: \subsection{Example}
751: \marginpar{\tiny quaternionic Shimura curves}
752:                           % quaternionic shimura curves
753: When $O_D$ is a maximal order in a quaternion algebra over $F$ that is
754: unramified at $\infty$, then for every level $I$, the moduli space described
755: above is a complete smooth curve over $\C$. By the above theorem, these spaces
756: are quotients of $\Omega^2\subset \PP^1(\C)$ by some subgroup $G$ of
757: $O_D^\ast\subset D\otimes_F \C\cong \M(2,\C)$. It follows from
758: the theory of Mumford curves (see \emph{e.g.}
759: \cite{Mumford72}, \cite{Gerritzen80}) that the genus of this curve is
760: the rank of $G$, \emph{i.e.} the rank of the Abelianisation of $G$. It
761: is in general however not straightforward to calculate the rank of a
762: given $G$. Usually it is easier to find a larger group $H \supset G$
763: such that $H\backslash\Omega^2$ has genus zero, and to calculate the
764: genus of $G\backslash\Omega^2$ from the ramification of
765: $G\backslash\Omega^2\to H\backslash\Omega^2\cong\PP^1$.
766: 
767: 
768: 
769: 
770: \bibliographystyle{plain}
771: \bibliography{master}
772: 
773: 
774: 
775: 
776: 
777: 
778: \end{document}
779: 
780: 
781: