• Aucun résultat trouvé

On p-adic vector bundles and local systems on diamonds Annette Werner

N/A
N/A
Protected

Academic year: 2022

Partager "On p-adic vector bundles and local systems on diamonds Annette Werner"

Copied!
20
0
0

Texte intégral

(1)

On p-adic vector bundles and local systems on diamonds

Annette Werner

Report on joint works with Marvin Anas Hahn (Leipzig) and Lucas Mann (Bonn)

Tropical Geometry, Berkovich Spaces, ArithmeticD-Modules and p-adic Local Systems

December 8-10, 2020

.

Goal:Relate p-adic vector bundles top-adic local systems.

(2)

On p-adic vector bundles and local systems on diamonds

Annette Werner

Report on joint works with Marvin Anas Hahn (Leipzig) and Lucas Mann (Bonn)

Tropical Geometry, Berkovich Spaces, ArithmeticD-Modules and p-adic Local Systems

December 8-10, 2020

.

Goal:Relate p-adic vector bundles top-adic local systems.

(3)

Motivation:

Narasimhan-Seshadri correspondence (1965) on a compact Riemann surfaceX, relating irreducible unitary representations of πtop1 (X,x) to stable vector bundles onX of degree 0.

Simpson’s correspondence (1992) on smooth projective complex varietiesX, relating finite dimensional complex representations of πtop1 (X,x) to semistable Higgs bundles onX with vanishing Chern classes.

(4)

p-adic results

Deninger/W. (2005):p-adic analog of Narasimhan-Seshadri correspondence relating vector bundles on smooth proper curves overQp with nice (potentially strongly semistable) reductions to continuousp-adic representations of the ´etale fundamental group.

(5)

p-adic results

Faltings’p-adic Simpson corrrespondence (2005) on smooth proper curvesX overQp, which provides an equivalence of categories betweenp-adic Higgs bundles on XCp and generalized

representations of the ´etale fundamental group.

See also Abbes, Gros, Tsuji (Annals of Math Studies 2016) for an alternative and more general approach.

Xu (2017): Both approaches are compatible in the case of trivial Higgs field.

Liu and Zhu’sp-adic Riemann Hilbert correspondence (2016) associates on a smooth rigid varietyX over finite extension K of Qp to everyp-adic ´etale local system a Higgs bundle on XCp with nilpotent Higgs field.

(6)

p-adic results

Faltings’p-adic Simpson corrrespondence (2005) on smooth proper curvesX overQp, which provides an equivalence of categories betweenp-adic Higgs bundles on XCp and generalized

representations of the ´etale fundamental group.

See also Abbes, Gros, Tsuji (Annals of Math Studies 2016) for an alternative and more general approach.

Xu (2017): Both approaches are compatible in the case of trivial Higgs field.

Liu and Zhu’sp-adic Riemann Hilbert correspondence (2016) associates on a smooth rigid varietyX over finite extension K of Qp to every p-adic ´etale local system a Higgs bundle on XCp with nilpotent Higgs field.

(7)

Vector bundles with numerically flat reduction

Letop⊂Cp with residue fieldk 'Fp.

LetX be a smooth proper connected variety overQp.

Denote byBX the (full) category of all vector bundles E onXCp with numerically flat reduction, i.e. such that

there exists a flat, proper scheme X of finite presentation over Zp with generic fiber X, and

there exists a vector bundleE onX ⊗op with generic fiber E such that the special fiber Ek =E ⊗k is numerically flat on Xk.

(8)

Vector bundles with numerically flat reduction

We call the bundleEk onXk numerically flat if both Ek and its dualEk are numerically effective.

Langer: This is equivalent to the fact that for allk-morphisms f :C −→ Xk from a smooth projective curveC the bundle fEk is semistable of degree 0 onC (Nori).

Note that ifX is projective, then the numerically flat line bundles are precisely the ones in the torsion componentPicτ(X).

Ek numerically flat implies thatE is numerically flat. If X is projective, this means thatE is semistable with vanishing Chern classes.

(9)

Theorem 1 (Deninger / W. 2020)

LetX be a proper, smooth, connected variety overQp.

LetE be a vector bundle onXCp. Ifα:Y −→X is finite ´etale surjective withαE inBY, i.e. ifαE has numerically flat reduction on some model ofY, thenE admitsp-adic parallel transport.

The functor ofp-adic parallel transport is a continuous functor ρE : Π1(XCp)−→VecCp

from the ´etale fundamental groupoid onXCp to the category of finite dimensionalCp-vector spaces which isx 7→Ex on objects, i.e.Cp-rational points.

In particular,E gives rise to ap-adic ´etale local system onXCp.

(10)

p-adic parallel transport

The functorE 7→ρE from Theorem 1 is functorial inE, exact and well-behaved with respect to tensor products, pullbacks, internal homomorphisms and Galois-conjugation.

Any numerically flat line bundleLsatisfies condition i) of Theorem 1, i.e.αLhas numerically flat reduction on some model of a finite

´etale coverα:Y →X.

Theorem 2 (Deninger / W. 2005, 2007)

IfX is a curve, then Theorem 1 holds for arbitrary finite coverings α:Y →X (not necessarily ´etale ones).

(11)

W¨urthen (2019)

LetX be a proper, connected seminormal rigid analytic variety overQp. Then the results of Theorem 1 hold for all analytic vector bundlesE on X with numerically flat reduction on a proper flat formal model ofX.

This allows us to drop smoothness.

Moreover, this approach works with locally free sheaves on the pro-´etale site. Using adic spaces and their integral structure sheaves makes some cumbersome arguments with models of schemes superfluous.

(12)

Fundamental open question

Find a category equivalence between genuineCp-representions of the ´etale fundamental group and a suitable category of Higgs bundles.

In the case of vanishing Higgs field, we know that vector bundles with numerically flat reductions give genuinep-adic

representations.

Assume we are given a smooth projective varietyX overQp and a numerically flat vector bundleE onX. How can we produce a finite ´etale coverα:Y −→X and a numerically flat degeneration ofαE ? We may find many models ofX andE, but how do we control the behaviour of the reductions of these models?

IfX is a curve, an arbitrary finite cover α suffices (which makes our choices even larger), but how do we control the pullbacks of the vector bundle in the special fiber under Frobenius powers?

(13)

Concrete example

Letj :C ,→P2K be the Fermat curvexd+yd+zd = 0 over some finite extensionK ofQp.

Consider the stable, rank 2, degree 0 bundle E =jSyz(x2,y2,z2)(3) onC, whereSyz(x2,y2,z2) =ker(O(−2)3 (x

2,y2,z2)

−→ O) on P2K.

(14)

We have the obvious reduction, given by the Fermat curve over OK and the pullbackE of the analogous syzygy bundle onP2OK. Its special fiberEk is semistable, but not numerically flat: Brenner (2005) shows that certain Frobenius pullbacks ofEk are no longer semistable.

So we need reductions with “bad” singularities.

Idea: UseMustafin varieties, which are degenerations of projective spacePnK depending on a choice of finitely many lattices inKn+1. Theorem (Cartwright, H¨abich, Sturmfels, W. 2011)

If the configuration Γ of lattice classes is contained in one apartment of the Bruhat-Tits building associated toPGLn+1,K, then the special fiber of the associated Mustafin varietyM(Γ) is determined by the regular mixed subdivision of the scaled simplex associated to the tropical convex hull of Γ.

(15)

RecallE =jSyz(x2,y2,z2)(3) on the Fermat curve j :C ,→P2K, whereSyz(x2,y2,z2) =ker(O(−2)3 (x

2,y2,z2)

−→ O).

Theorem (Hahn, W. 2019)

For the exampleE on the projectively embedded Fermat curve C we find

a (ramified) degree 2 cover α:D →C

a configuration Γ of three lattices such that the closure Dof D in the Mustafin varietyM(Γ) satisfies: the special fiber consists of irreducible components Ci with eachCired 'P1k a vector bundle E onD with generic fiberαE such that Ek|Cred

i is trivial for all i.

This proves that the bundleE admits p-adic parallel transport.

(16)

Local systems on diamonds: joint work with Lucas Mann

LetX be a proper adic space of finite type overCp.

Scholze’s theory of diamonds provides a fascinating tool to disguise characteristic zero objects as characteristicp. Diamonds are certain sheaves on the big pro-´etale site on the category Perf of perfectoid characteristicp spaces, which are quotients of perfectoid spaces by pro-´etale equivalence relations.

There is a functorX 7→X mapping analytic adic spaces over Zp

to (locally spatial) diamonds, such that Xet 'Xet

IfK is a non-archimedean extension ofQp, the diamond functor from seminormal rigid analyticK-spaces to diamonds overSpd(K) is fully faithful

(17)

We use thev-topology on Perf: Coverings of Z are given by collections of maps toZ such that every quasi-compact open subset ofZ can be covered by finitely many quasi-compact open subsets of our collection.

Thev-site on a diamond X is then defined as the category of small v-sheaves mapping toX, with coverings given by jointly surjective maps. There is a naturalv-structure sheaf ˇOX with integral structure sheaf ˇOX+.

For every condensed ring Λ (e.g. every metric ring), there is a notion of the asscoiated constant sheaf on a diamond, and hence a notion of local systems (locally free sheaf of Λ-modules) for the v-topology.

(18)

LetX be a proper adic space of finite type overCp. LetM+ be the full subcategory of all ˇO+X-modules E onXv such that for alln the ˇO+X/pn-moduleE/pn becomes trivial on a proper cover ofX. There is a fully faithful functor ∆+ from M+ to the category of op=OCp-local systems onX such that for everyE inM+ we have

+(E)⊗op+X=E.

(19)

LetMbe the (abelian) category of all ˇOX-modules of the form E =E[p−1] for someE inM+.

Theorem 3 (Mann, W. 2020)

Inverting p, we get a functor ∆ fromM to the category ILocCp(X) of Cp-local systems with an integral model.

The functor ∆ is an equivalence of categories, compatible with direct sums, tensor products, internal homs, exterior powers and pullback.

For every E ∈ Mthere is a natural isomorphism

∆(E)⊗CpX =E

(20)

Theorem 4 (Mann, W. 2020)

Letf:Y →X be a surjective morphism of finite type proper adic spaces overCp and assume that X is normal. Let E be an

X-module. IffE ∈ MY, thenE ∈ MX. In other words, the property of having properly trivializable reduction descends alongf.

For adic spaces associated to proper smooth algebraic curves, this generalizes the previously explained result for finite pullbacks.

Références

Documents relatifs

Although Q p is an inexact field, our model of complexity makes sense because the size of the p-adic numbers we manipulate will all have roughly the same size: for ZpLF, it is

The contributions include the concept of position-restricted seeds, particu- larly suitable for short alignments with non-uniform error distribution; a model that captures

” Some of the standard and well-established techniques of local arithmetic geometry can also be seen as involving dynamical systems.. Let K/Q p be a

[Gr-Mat 2] Green, Barry; Matignon, Michel Order p automorphisms of the open disc of a p-adic field, J.. Finite order automorphisms of a p-adic

Timings with systems of singular (q)-differential equations.. Timings in seconds of lifting of univariate representations over F 16411 [[x]] for Katsura-n... Thank you for

En permutant A et B dans la question 3, on voit que e séquent n'est pas prouvable

We have proved that if we are using the model of interval arithmetic, then the subresultant pseudo-remainder sequence algorithm will output Res(A, B) at precision O(π N − N int )

The previous Section 3 shows that the good notion of p-adic regulator comes from the expression of the p-adic finite group R K associated with the class field theory interpretation