arXiv reader

Duality Theorem for Motives

Ivan Panin and Serge Yagunov

arXiv
0810.2160v1 · PDF · 13 October 2008
Categories
math.AG, math.KT
Licence
CC BY 4.0 The site shows the text.

Abstract

Using Dold–Puppe category approach to the duality in topology, we prove general duality theorem for the category of motives. As one of the applications of this general result we obtain, in particular, a generalization of Friedlander–Voevodsky’s duality to the case of arbitrary base field characteristic.

Items 10

Example

Example 1

A0KC

One sets \mathfrak M=\mathbf{DM^-(k)} and the functor \bf M to be the corresponding motive functor from [SV, Vo2]. Then \bf M happens to be a functor with transfers for projective morphisms and our list of axioms is fulfilled. The Cancelation axiom was checked in this context in [Vo1]. Construction of the transfer and verification of other axioms belong to the base properties of Voevodsky’s motives (see, for example [SV, Section 4]).

Example

Example 2

PKMU

Changing, in our settings, \bf Sm/k with a category of CW-complexes and taking the derived homotopy category of singular complexes with \mathbb Z_2-coefficients as \mathfrak M, we obtain, as one can easily see, the standard definitions of singular (co-)homology. To make the indexes consistent, it suffices to put M(i)[j]:=M[j-i], where on the right-hand-side the usual triangulated category shift is employed. Alteratively, one can just rewrite all the constructions with a single index. The role of projective morphisms is played in this case by proper (in the topological sense) maps. The transfer construction can be found in almost any algebraic topology textbook. Direct verification shows that all the axioms hold in the considered case. As a result we obtain a proof of the classical Poincaré duality theorem in the spirit of Dold–Puppe’s paper [DP].

Theorem

Theorem 3

HYFH

For every orientable theory of motives \bf M and arbitrary varieties X,Y\in\bf Sm/k with projective equidimensional X there is the canonical isomorphism of abelian groups:

\operatorname{Hom}(M(Y)(i)[j],M(X))\simeq \operatorname{Hom}(M(Y)\otimes M(X),\mathbb Z(d-i)[2d-j]),

contravariant with respect to Y. Here d=\dim X.

Used by Proof, Proof
Corollary

Corollary 4

1FMZ

For an arbitrary smooth projective equidimensional variety X of dimension d there exists the canonical isomorphism:

\mathrm{H\!M}_{*,*}(X)\simeq \mathrm{H\!M}^{2d-*,d-*}(X).
Used by Proof
Proof of Corollary 4

Set Y=\operatorname{pt} in the theorem statement.

Corollary

Corollary 5

VNPP

Let \bf M be the theory of motives considered above in Example 1. Then for every smooth projective equidimensional variety X the constructed duality morphism

\psi\colon M(X)\to\text{\it \underline{Hom}}(M(X),\mathbb Z(d)[2d])

is an isomorphism.

Proof of Corollary 5

Complete the morphism \psi up to a distinguished triangle in the category \mathfrak M and denote the third vertex of the triangle by \text{\rm Cone}(\psi). As it follows from Theorem 3, the relation \operatorname{Hom}(M(Y)(p)[q],\text{\rm Cone}(\psi))=0 holds for every smooth irreducible variety Y and arbitrary integers p,q. Because the category \mathfrak M is weakly generated by motives of such a kind, one can see that the object \text{\rm Cone}(\psi) is isomorphic to the zero-object. Triangulated category axioms easily imply that the morphism with zero cone is an isomorphism.

Proof

Proof (Theorem )

BHML

In order to construct isomorphisms mentioned in the Theorem, one needs the following ingredients. First of all, denote by 1\in\mathrm{H\!M}^{0,0}(\operatorname{pt})=\mathrm{H\!M}_{0,0}(\operatorname{pt}) the element of (co-)homology group of the point, corresponding to the identity morphism \operatorname{id}\colon\operatorname{pt}\to\operatorname{pt}. Let us note that for a morphism of varieties f\colon X\to Y the morphism M(f) induces the natural maps f_*\colon\mathrm{H\!M}_{*,*}(X)\to\mathrm{H\!M}_{*,*}(Y) in motivic homology and f^*\colon\mathrm{H\!M}^{*,*}(Y)\to\mathrm{H\!M}^{*,*}(X) respectively, in cohomology. In case then f is projective of codimension d, motivic transfer map f^! induces corresponding transfers

f_!\colon\mathrm{H\!M}^{*,*}(X)\to\mathrm{H\!M}^{*+2d,*+d}(Y)\hskip 0.3cm\text{and}\hskip 0.3cm f^!\colon\mathrm{H\!M}_{*,*}(Y)\to\mathrm{H\!M}_{*-2d,*-d}(X)

in (co-)homology. For a variety X of dimension d consider the diagonal and projection morphisms:

X\times X\overset\Delta\leftarrow X\overset p\rightarrow\operatorname{pt}

and call the elements

[X]_*=\Delta_*p^!(1)\in\mathrm{H\!M}_{2d,d}(X\times X)

and

[X]^*=\Delta_!p^*(1)\in\mathrm{H\!M}^{2d,d}(X\times X)

fundamental and cofundamental classes of X, correspondingly. Let us also define slant-products

/\colon \mathrm{H\!M}^{i,j}(X\times Y)\otimes\mathrm{H\!M}_{m,n}(Y)\to \mathrm{H\!M}^{i-m,j-n}(X)

and

\backslash\colon \mathrm{H\!M}_{m,n}(X\times Y)\otimes\mathrm{H\!M}^{i,j}(Y)\to \mathrm{H\!M}_{m-i,n-j}(X)

in the following way. For elements \alpha\in \mathrm{H\!M}^{i,j}(X\times Y) and a\in \mathrm{H\!M}_{m,n}(Y) set:

\alpha/a: M(X)(m)=M(X)\otimes\mathbb Z(m)\overset{1\otimes a}\to M(X)\otimes M(Y)[-n]\overset\alpha\to \mathbb Z(i)[j-n]

for the first product, and symmetrically:

\beta\backslash b:\mathbb Z(m)[n-j]\overset b\to M(X)\otimes M(Y)[-j]\overset{1\otimes \beta}\to M(X)\otimes\mathbb Z(i)=M(X)(i)

for the second one, provided that b\in \mathrm{H\!M}_{m,n}(X\times Y) and \beta\in \mathrm{H\!M}^{i,j}(Y). (Here and below we implicitly use Cancelation axiom.) Set Poincaré Duality homomorphisms for the case of Corollary 4, letting:

\mathcal D_\bullet(-)=[X]^*/-\colon \mathrm{H\!M}_{*,*}(X)\to \mathrm{H\!M}^{2d-*,d-*}(X)

and

\mathcal D^\bullet(-)=-\backslash [X]_*\colon \mathrm{H\!M}^{*,*}(X)\to \mathrm{H\!M}_{2d-*,d-*}(X),

verbatim as it was done in [PY].

Then, consider a natural extension of these homomorphisms to the general case of the theorem. Denote, for brevity, the motive M(Y)(i)[j] by \mathcal Y, and the index shift and twist (d)[2d] by \{d\}. Construct a map

\mathcal D_\bullet\colon \operatorname{Hom}(\mathcal Y,-)\to\operatorname{Hom}(\mathcal Y\otimes -,\mathbb Z\{d\})

as follows. For a morphism a\in \operatorname{Hom}(\mathcal Y,M(X)) give D_\bullet(a) by the formula:

\mathcal Y\otimes M(X)\overset{a\otimes 1}\to M(X)\otimes M(X)\overset{\Delta^!}\to M(X)\{d\}\overset{p_*}\to \mathbb Z\{d\}.

The inverse map

\mathcal D^\bullet\colon\operatorname{Hom}(\mathcal Y\otimes -,\mathbb Z\{d\})\to \operatorname{Hom}(\mathcal Y,-)

is given for a morphism \alpha\in \operatorname{Hom}(\mathcal Y\otimes M(X),\mathbb Z\{d\}) as:

\mathcal Y\overset{1\otimes p^!}\to \mathcal Y\otimes M(X)\{-d\}\overset{1\otimes \Delta_*}\to \mathcal Y\otimes M(X) \otimes M(X)\{-d\}\overset{\alpha\otimes 1}\to \mathbb Z\otimes M(X).

The constructed maps are obviously contravariant with respect to the variable \mathcal Y. Theorem 3 now results from the Cancelation axiom and commutativity of two diagrams below(In the consequent diagrams M(X)^{\otimes n} denotes n-fold tensor product and \Delta^i is a diagonal morphism applied to the i-th factor.): $$ YZM(X)_1p^!1[d]^id@=[dr]
[dr]|

Proposition

Proposition 6

Z4GU

For projective equidimensional varieties X,Y\in\bf Sm/k and a morphism f\colon X\to Y, one has:

f_!=\mathcal D_\bullet^Y f_* \mathcal D^\bullet_X\hskip 1cm\text{and}\hskip 1cm f^!=\mathcal D^\bullet_X f^* \mathcal D_\bullet^Y.

Here \mathcal D_\bullet^X and \mathcal D_\bullet^Y denote Poincaré duality isomorphisms applied to varieties X and Y, respectively.

Proof of Proposition 6

The first equality can be easily derived from the relation

f_*(\alpha\backslash [X]_*)=f_!(\alpha)\backslash [Y]_*,

which, in its row, yields from the commutativity of the diagram:

\xymatrix{&&M(X)\otimes M(X)\ar^{1\otimes f_*}[d]\ar^(0.55){\alpha\otimes 1}[r]&\mathbb Z\otimes M(X)\ar^{1\otimes f_*}[d]\\ \mathbb Z\ar^{p_X^!}[r]\ar_-{p_Y^!}[dr]&M(X)\ar@{}[dr]|{\blacksquare}\ar[r]_-{\Gamma^f_*}\ar^-{\Delta^X_*}[ur]&M(X)\otimes M(Y)\ar[r]^(0.55){\alpha\otimes 1}&\mathbb Z\otimes M(Y)\\ &M(Y)\ar_-{\Delta^Y_*}[r]\ar^{f^!}[u]&M(Y)\otimes M(Y).\ar_{f^!\otimes 1}[u]}

(Here all the shifts and twists are omitted for simplicity.) The square marked by \blacksquare is induced by the transversal graph diagram of the morphism f:

\xymatrix{X\ar^-{\Gamma^f}[r]\ar_f[d]&X\times Y\ar^{f\times\operatorname{id}}[d]\\ Y\ar_-\Delta[r]&Y\times Y}

Hence, it is commutative. Commutativity in all other diagram parts is obvious. The second equality can be proven in a similar way.

Source and licence

The text on this page is from arXiv:0810.2160v1 by Ivan Panin and Serge Yagunov, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/0810/0810.2160.jsonl.