The A1–homotopy type of Atiyah–Hitchin schemes I: the geometry of complex points

Description

Niveau: Supérieur, Doctorat, Bac+8
The A1–homotopy type of Atiyah–Hitchin schemes I: the geometry of complex points CHRISTOPHE CAZANAVE Given a smooth algebraic variety Y , we construct a family of new algebraic varieties RnY indexed by a positive integer n , which we baptize the Atiyah– Hitchin schemes of Y . This paper is the first of a series devoted to the study of the A1 –homotopy type (in the sense of Morel and Voevodsky) of these schemes. The interest of the Atiyah–Hitchin schemes is that we conjecture that, as n tends to infinity, the sequence of spaces RnY converges, in a precise sense, to ?P 1?P1 Y , the free P1 –loop space generated by Y . This first paper focuses on the geometry of the schemes RnY : the slogan is that RnY is a scheme-theoretic “completion” of the unordered configuration space of n distinct points in A1 with labels in Y . This makes RnY analogous—although in general different—to the May–Milgram model. 55P35, 55R80, 14F42; 57N80 1 Introduction This is the first of a series of papers devoted to the introduction and the study of some in- teresting families of algebraic varieties which we baptize the Atiyah–Hitchin schemes. There is a family of Atiyah–Hitchin schemes attached to any given algebraic variety Y ; they are indexed by a positive integer n and we denote them RnY . The fundamental example at the source of the definition, corresponding to the case where Y is the affine line minus the origin, is given by the schemes of pointed degree n rational functions (Fn)n>0 (see example

• functions has


• has multiplicity


• called whitney


• main result


• rny


• topological space


• plex rational


• bundles over


• functions

Sujets

• Lannes
• Topological space
• Function

Informations

• Publié par: profil-zyak-2012
• Nombre de lectures: 16
• Langue: English

Extrait

1

The A 1 –homotopy type of Atiyah–Hitchin schemes I: the geometry of complex points C HRISTOPHE C AZANAVE

Given a smooth algebraic varietyY, we construct a family of new algebraic varietiesR n Yindexed by a positive integern, which we baptize theAtiyah– Hitchin schemesofY . Thispaper is the rst of a series devoted to the study of the A 1 –homotopy type (in the sense of Morel and Voevodsky) of theseschemes. The interest of the Atiyah–Hitchin schemes is that we conjecture that, asntends to innity, the sequence of spacesR n Yconverges, in a precise sense, toΩ P 1 Σ P 1 Y, the freeP 1 –loop space generated byY. This rst paper focuses on the geometry of the schemesR n Y: the slogan is that R n Yheftorunredeondcrugoitaapsnfoecisascheme-theoreit“cocpmelitnoo” ndistinct points inA 1 with labels inY. This makesR n Yanahugholt—ausgolo in general different—to the May–Milgram model.

55P35, 55R80, 14F42; 57N80

Introduction

This is the rst of a series of papers devoted to the introduction and the study of some in- teresting families of algebraic varieties which we baptize theAtiyah–Hitchin scheme.s There is a family of Atiyah–Hitchin schemes attached to any given algebraic varietyY; they are indexed by a positive integernand we denote themR n Y. The fundamental example at the source of the denition, corresponding to the case whereYis the afne line minus the origin, is given by the schemes of pointed degreenrational functions (F n ) n>0 (seeexample 2.4–(3) homo- Thefor the precise denition of these schemes). topy type of the topological spaceF n (C) of complex pointed degreenrational functions has been studied by Graeme Segal in his seminal article [31 proved that “the]. Segal sequence of topological spacesF n (C) converges, asntends to innity, to the double loop spaceΩ 2 S 3 ”. (The meaning of this last statement is that the natural inclusion map ofF n (C) into the degreencomponent of the space of continuous pointed maps map ∗ (P 1 (C),P 1 (C))—a space homotopy equivalent toΩ 2 S 3 —induces isomorphisms on thenresult has been a large source of inspiration for This rst homotopy groups.)

2

Christophe Cazanave

many authors, among which C Boyer, F Cohen, R Cohen, M. Guest, J Hurtubise, S. Kallel, Y. Kamiyama, F Kirwan, A. Kozlowzki, B. Mann, J. Milgram, M Murayama, J. Mostovoy, D. Shimamoto, K. Yamaguchi [7,8,9,16,17,18,19,20,21,22,28,34]. In particular, one can nd in the literature several constructions inspired bySegal's paper [31nite-dimensional manifolds approximating certain mapping spaces.] of successive

The objective of this series of articles is twofold. First, we would like a unifying framework for the various topological results mentioned above. As a partial answer, we construct a general “machine” which, when fed with a smooth connected manifold Y a “nice” topological space(resp. withY), returns a familyR n Yof smooth nite- dimensional manifolds (resp. topological spaces) which are successive approximations ofΩ 2 Σ 2 Y, the double loop space freely generated byY construction of. (TheR n Y is given in this article; that it does approximateΩ 2 Σ 2 Ywill be proved in [3].) Our second aim, which is more speculative, explores a new direction of generalization: that ofA 1 –homotopy theory, the homotopy theory of schemes developed by Morel and Voevodsky [27 our construction is algebraic:]. Indeed, fed with a smooth when algebraic varietyY, the “machine” returns a family of smooth algebraic varieties, the Atiyah–Hitchin schemesR n Y this series of articles, we will give evidence that. In when the algebraic varietyYisA 1 –connected then “the sequence of schemesR n Y converges (in the homotopy category) toΩ P 1 Σ P 1 Y, theP 1 –loop space freely generated byY”. (As above, one can give a precise meaning to this statement.)

This rst paper is devoted to the analysis of the geometry of the Atiyah–Hitchin schemes. The slogan is that for every integern>1,R n Yhas to be thought of as a scheme-theoretic “completion” of the unordered conguration space ofndistinct points inA 1 with labels inY. (Here by “completion” one should understand that a generic point of (R n Y)(C) belongs to the above conguration space but that there are also moredegenerate illustrate this by describing the geometry ofcongurations.) We the complex manifold (R n Y)(C) associated to a complex algebraic varietyY. The geometry of (R n Y)(C) is reminiscent of an other well-known approximation of the double loop spaceΩ 2 Σ 2 Y(C): the so-called May–Milgram modelC 2 Y(C) (see May's book [26 The, construction 2.4]). space (R n Y)(C) is closely related to then-th term F n (C 2 Y(C)) in the canonical ltration ofC 2 Y(C we will prove in [). Indeed,3] (see also [6, chapitre 5]) that there is astablehomotopy equivalence Σ ∞ (R n Y)(C)≈Σ ∞ F n (C 2 Y(C)) which is compatible with the Snaith splitting ofΩ 2 Σ 2 Y(C the special case of com-). (In plex rational functions, this stable homotopy equivalence was rst proved by F Cohen,

TheA 1 –homotopy type of Atiyah–Hitchin schemes I

3

R Cohen, B Mann and J Milgram [7,8].) However,unstablythe two spaces (R n Y)(C) and F n C 2 Y(C) may differ: for an integerd>1 and forn>1, (R n (A d − {0}))(C) and F n C 2 (C d − {0}) are homotopy equivalent if and only ifd>1 (see R Cohen and D Shimamoto [9] and Totaro [32]).

We leave to the next articles [3,4] the study of the (unstable and stable) homotopy type of the Atiyah–Hitchin schemes and its relation toΩ P 1 Σ P 1 Y. There is also an interesting connection with our previous work in [5] on the algebraic connected components of the schemes of pointed rational functionsF n which suggests the existence of a version of the little disks operad and of the group completion theorem inA 1 –homotopy theory.

Let us give a avor of the results and techniques contained in the paper. LetYbe a xed smooth complex algebraic variety and letn Abe a positive integer. point in the space (R n Y)(C) is a pair (A,B) where: •A=X n +a n−1 X n−1 +  +a 0 is a monic degreenpolynomial with complex coefcients •Bcorresponds to the datum, for each rootαof the polynomialA, of a point in the total space of a certain vector bundleJ α overY(C vector bundles). These J α are “jet-like” bundles overY(C) of order the multiplicity of the rootα. For example, whenαis a simple rootJ α is the zero-dimensional vector bundle over Y(C); whenαhas multiplicity two,J α is the tangent bundle ofY(C) and so on. (See the introduction ofSection 3for a more precise account about this.)

This description leads to a decomposition of the complex manifold (R n Y)(C)as a set as a disjoint union of complex submanifolds, each one individually well understood. For example, the open stratum (corresponding to the locus where the rstcoordinateA has all its roots simple) is homeomorphic to the space of unordered congurations ofn distinct points inCwith labels inY(C). Notethis is exactly the same space which that appears in the denition of F n (C 2 Y(C)), then-th term of the canonical ltration of the May–Milgram modelC 2 Y(C) forΩ 2 Σ 2 Y(C). In general, each piece of (R n Y)(C) isup to homotopya space of congurations of a certain number of points inCwith labels inY(Chere whether the congurations are ordered or not,see) (we leave vague denition 3.1.1). A full understanding of the geometry of the space (R n Y)(C) requires also information about how the different pieces are glued together. We can provide this information by using the fact that the decomposition of (R n Y)(C) satises a strong regularity condition: it is a so-calledWhitney stratication. For such stratications, the work of J

4

Christophe Cazanave

Mather in the 1970's (based on the former contribution of R Thom) provides a suitable notion of tubular neighbourhood of one stratum into another adjacent stratum, which we can be described. Although our description requires a combinatorial formalism which is a bit intricate, the two main underlying ideas are simple.

• The geometry of the strata of (R n Y)(C in a if,) has the following property: conguration of points inCwith labels in 1 Y(C) (which represents a point in some stratum), two pointsα 1 6=α 2 ∈Chave the same label, sayy 1 = y 2 =y∈Y(C), then, as the two pointsα 1 andα 2 tend to a common value, say α∈C, then the conguration [(α 1 ,y),(α 2 ,y),(α i ,y i )] tends to the conguration [(α,y),(α i ,y i  turns out that It(which represents a point in a lower stratum).)] this information sufces to capture the homotopy type of (R n Y)(C). • The stratication of the space of monic complex polynomials associated to the multiplicity of the roots is closely related to the stratication of (R n Y)(C). (In fact, it corresponds to the special caseY= strata are genuinept.) Its conguration spaces of points inC the attaching maps between the strata. And can be described by some versions of the structure maps in the little disks operad on the level of conguration spaces of points a little disk into: embedding another one is analogous to blowing up a multiple root into other roots of lower multiplicities.

A rough summary of our main result (theorem 3.4.3 is a natural There) is the following. homotopy equivalence between (R n Y)(C) and a space denotedΨ n (Y(C)) obtained by attaching conguration spaces of little discs inCwith labels inY(C) togethervia structure maps in the little discs operad and diagonals on the labels inY(C). In particular, our description implies that the homotopy type of the topological space (R n Y)(Conly depends on the homotopy type of) Y(C), which is not obvious from the denition ofR n Y(and which is convenient to use in [3 when the algebraic]). Moreover, varietyYis dened over the eld of real numbersR, complex conjugation induces an involution on both (R n Y)(C) andΨ n (Y(C analysis of the stratication is)). Our compatible with these involutions: we have a homotopy equivalenceΨ n (Y(C))≈ (R n Y)(C) as Z  2 –spaces. This suggests that theA 1 –homotopy type of the spaceR n Y should only depend on theA 1 –homotopy type ofY, a result which we could not prove (seequestion 3.6.2).

Overview of the article

1 Here,Y(C) is identied with the zero sections of the “jet-like” bundelsJ α .

TheA 1 –homotopy type of Atiyah–Hitchin schemes I

•

•

•

5

Section 2 Theiris devoted to the denition of the Atiyah–Hitchin schemes. functor of points are easily dened and shown to be representable. Somebasic properties are established. The algebraic geometry here is elementary and aimed attopologists(inparticular,weavoidusingthetheoryofHilbertschemes).We also discuss examples we nd illuminating.

Section 3describes completely the stratication of the space of complex points (R n Y)(C) whenYis dened overC more details:. In – §3.1describes the topology of the strata of (R n Y)(C are the). These conguration spaces of points inCwith labels in the “jet-like” bundles overY(C) alluded to in the introduction. – §3.2 weis a warm-up: analyse in details the stratications of (R 2 Y)(C) and (R 3 Y)(C study of (). TheR 3 Y)(C) requires already all the technical difculty contained in the use of the Thom–Mather theory of controlled tubular neighbourhoods in Whitney stratications. It is written in order to motivate and illustrate the general method. – To treat the general case, we rst introduce in §3.3the required formalism to handle the combinatorics. This allows us to dene in §3.4the functor Ψ n :Top−→Topfor anynand to state our main result (theorem 3.4.3): there is a natural homotopy equivalence between (R n Y)(C) andΨ n (Y(C)). The proof is then given in §3.5. – In §3.6, we study the case when the varietyYis dened over the eld of real numbersR this case, complex conjugation induces an involution. In on both spacesΨ n (Y(C)) and (R n Y)(C) and the homotopy equivalence (R n Y)(C)≈Ψ n (Y(Cis shown to be compatible with this action.)) Appendix Ais a recollection of the necessary material on stratications. In particular, we briey present the Thom–Mather theory of controlled tubulra neighbourhoods for Whitney stratications, which is used as an important in- gredient in our description of the homotopy type of (R n Y)(C).

Acknowledgements

I present here one part of my Ph.D. thesis [6]. I am very much indebted to Jean Lannes for his precious and generous advises. It's a pleasure to warmly thank h im here. This research was partially supported by the project ANR blanc BLAN08-2 338236, HGRT.