This paper constructs an alternative Betti realization functor for the derived category of motives using CW complexes, showing it coincides with Ayoub's and relates to classical cycle class maps.
Contribution
It provides a new construction of the Betti realization functor via CW complexes and proves its equivalence to Ayoub's functor, connecting motives with classical cycle maps.
Findings
01
The new functor coincides with Ayoub's realization functor.
02
The realization factors through Nori motives.
03
It matches the classical cycle class map on higher Chow groups.
Abstract
We give an alternative construction of the Betti realization functor on the derived category of motives of complex algebraic varieties via the category of CW complexes instead of the category of complex analytic spaces. In particular we show that the functor we define via the category of CW complexes coincide with Ayoub's one. We deduce from this construction that Ayoub's realization functor on geometric motives factors trough Nori motives and that the image of this functor on the morphisms between the motive of a point and a shift of a Tate twist of the motive with compact support of a complex algebraic variety coincide with the classical cycle class map on higher Chow groups.
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsHomotopy and Cohomology in Algebraic Topology · Algebraic Geometry and Number Theory · Algebraic structures and combinatorial models
Full text
On the realization functor of the derived category of mixed motives
Johann Bouali
Abstract
We give an alternative construction of the Betti realization functor on the derived category of
motives of complex algebraic varieties via the category of CW complexes
instead of the category of complex analytic spaces. In particular, we show that
the functor we define via the category of CW complexes coincide with Ayoub’s one.
We deduce from this construction that Ayoub’s realization functor on geometric motives
factor through Nori motives and that the image of this functor on the morphisms
between the motive of a point and a Tate twist of the motive with compact support of a complex algebraic variety
coincide with the classical cycle class map on higher chow groups.
Notations:
•
Denote by Top the category of topological spaces.
Denote by Var(k) the category of algebraic varieties over a field k, i.e. schemes of finite type over k.
Let us call PVar(k)⊂QPVar(k)⊂Var(k) the full subcategories quasi-projective varieties and projective varieties respectively.
Let us call PSmVar(k)⊂SmVar(k)⊂Var(k) the full subcategories of smooth varieties and smooth projective varieties respectively.
Denote by CW⊂Top the full subcategory of CW complexes, by CS⊂CW the full subcategory of Δ complexes,
and by TM⊂CW the full subcategory of topological manifolds
which admits a CW structure (a topological manifold admits a CW structure if it admits a differential structure).
Denote by AnSp(C) the category of analytic spaces over C.
and by AnSm(C)⊂AnSp(C) the full subcategory of smooth analytic spaces (i.e. complex analytic varieties).
•
For V∈Var(C), we denote by Van∈AnSp(C)
the complex analytic space associated to V with the usual topology induced by the usual topology of CN.
For W∈AnSp(C), we denote by Wcw∈AnSp(C) the topological space given by W which is a CW complex.
For simplicity, for V∈Var(C), we denote by Vcw:=(Van)cw∈CW.
We have then
–
the analytical functor An:Var(C)→AnSp(C), An(V)=Van,
–
the forgetful functor Cw:AnSp(C)→CW, Cw(W)=Wcw,
–
the composite of these two functors \widetilde{\mathop{\rm Cw}}\nolimits=\mathop{\rm Cw}\nolimits\circ\mathop{\rm An}\nolimits:\mathop{\rm Var}\nolimits(\mathbb{C})\to\mathop{\rm CW}\nolimits, \widetilde{\mathop{\rm Cw}}\nolimits(V)=V^{cw}.
•
We denote by □n=(P1\{1})n⊂(P1)n. For X∈Var(C) let
Zp(X,n)⊂Zp(X×□n) be the subgroup of p codimentional cycle in X×□n meeting
all faces of □n properly.
We denote by πX:X×(P1)n→X and π(P1)n:X×(P1)n→(P1)n the projections.
•
For X∈Top a topological space, we denote by
C∙sing(X,Z)=ZHomTop(Δ∙,X) the complex
of singular chains, Δp⊂Rp being the standard simplex.
We denote by In=[0,1]n and we will consider the closed embeddings of CW complexes
in:[0,1]n↪□n:=(PC1\{1})n
whose image is the product [0,∞]n⊂□n of the segments
R−=[0,∞]⊂□1 (c.f. the definition of Tz in [11]).
In particular in send [math] to [math] and 1 to ∞
and gives a morphism of complexes i:I∗↪□∗ in Z(CW).
We denote by Dˉn⊂Cn the closed ball of radius 1,
and by i1n′:In↪Dˉn the inclusion of pro complex analytic spaces and
and i1n′cw:In↪Dˉn is the corresponding inclusion of CW complexes.
•
For a (small) category S, we denote by
PSh(S,M):=Fun(Sop,M) the big category of presheaves
on S with value in M.
If M is a model category with some extra assumptions (c.f.[4]),
the projective fibration (rep. the injective cofibration) and the termwise weak equivalence of PSh(S,M)
define a projective (resp. injective) model structure PSh(S,M).
In this paper, we will consider MP(PSh(S,M)) the projective model structure on
PSh(S,M).
presheaves.
–
For X∈S, we denote by Z(X)∈PSh(S,Ab) the preshesf
given by Yoneda embedding, that is the presheaf given by for Z∈S,
Z(X)(Z)=ZHomS(Z,X) and for f:Z′→Z a morphism in S,
Z(X)(f):g∈HomS(Z,X)↦g∘f∈HomS(Z′,X)
–
For h:X→Y a morphism in S, we denote by Z(h):Z(X)→Z(Y),
the morphism in PSh(S,C(Z) given by Yoneda embedding,
that is the morphism given by for Z∈S,
Z(h)(Z):g∈HomS(Z,X)↦h∘g∈HomS(Z,Y)
•
We denote by C(Z)=C(Ab) the category of abelian complexes, C−(Z)⊂C(Z)
the full subcategory consisting of bounded above complexes, D(Z) the derived category of
C(Z) with respect to quasi-isomorphism, and D−(C)⊂D(Z) the image
of C−(Z) under the localization functor C(Z)→D(Z).
For S∈Top, we denote by C(S):=PSh(S,C(Z)) the category of complexes of presheaves on S,
C−(Z)⊂C(Z) the full subcategory consisting of bounded above complexes,
D(S) the derived category of C(S) with respect to the morphisms of complexes of presheaves
which are quasi-isomorphisms after sheaftification,
and D−(S)⊂D(S) the image of C−(S) under the localization functor C(S)→D(S).
1 Introduction
Let S∈Var(C).
In [1],[2] and [3], J.Ayoub has given and studied
a construction of the Betti realization functor on DA−(S,Z)
the derived category of mixed motives of Var(C)/S the complex algebraic varieties
together with a morphism over S.
He mentioned also the construction of the Betti realization functor on DM−(S,Z),
the derived category of mixed motives of Var(C)/S with transfers.
Let k a field.
•
The derived category of (effective) motives of Var(k)/S is
the homotopy category of the category P−(S)=PSh(Var(k)sm/S,C−(Z)) of bounded above complexes
of presheaves on the category of algebraic varieties over k together with a smooth morphism over S,
with respect to the projective (A1,et) model structure (c.f. definition 8(i)).
•
Similarly, the derived category of (effective) motives of Var(k)/S with transfers is
the homotopy category of the category PC−(S)=PSh(CorZfs(Var(k)sm/S),C−(Z))
of bounded above complexes
of presheaves on the category of finite and surjective correspondences between algebraic varieties over k
together with a smooth morphism over S with respect to the projective (A1,et) model structure
(c.f. definition 8(ii)).
•
For X∈Var(k) and l:D↪X a (locally closed) subvariety,
the motive of the pair (X,D) is M(X,D)=D(A1,et)(Ztr(X,D))∈DM−(k,Z), where
Ztr(X,D)=coker(Ztr(l)) is the cokernel of the injective morphism of presheaves
Ztr(l):Ztr(D)↪Ztr(X)
and D(A1,et):PC−→DM−(k,Z) is the localization functor.
The construction of J.Ayoub is defined on DA−(S) via AnDA−(San),
the derived category of motives of complex analytic space, which is
the homotopy category of P−(An,San)=PShZ(AnSp(C)sm/San,C−(Z))
of bounded above complexes of presheaves on the category of complex analytic spaces together with a smooth
morphism over San with respect to the projective (D1,usu) model structure (c.f. definition 34(i)).
Similarly on DM−(S,Z) it is defined via AnDM−(San,Z),
the derived category of motives of complex analytic space, which is
the homotopy category of
PC−(An,San)=PShZ(CorZfs(AnSp(C)sm/San),C−(Z))
the category of bounded above complexes of presheaves on the category of finite and surjective correspondences
between complex analytic spaces together with a smooth morphism over S
with respect to the projective (D1,usu) model structure (c.f. definition 34(ii)).
For S∈AnSp(C), we consider the commutative diagram
[TABLE]
the morphism of sites given respectively by
the inclusion functors ean(T):Ouv(S)↪AnSp(C)sm/S,
eantr(S):Ouv(S)↪CorZfs(AnSp(C)sm/S)
and Tr(S):CorZfs(AnSp(C)sm/S)↪AnSp(C)sm/S.
The definition of Betti realization functor by J.Ayoub is
The Betti realisation functor (without transfers) is the composite :
[TABLE]
(ii)
The Betti realisation functor with transfers is the composite :
[TABLE]
Since An(S)∗ derive trivially by proposition 44(ii)
and and LTr(San)∗:AnDA−(San,Z)→AnDM−(San,Z) is the inverse of Tr(San)∗,
we have \mathop{\rm Bti}\nolimits_{0}(S)^{*}=\widetilde{\mathop{\rm Bti}}\nolimits(S)^{*}\circ L\mathop{\rm Tr}\nolimits(S)^{*}.
In [2], J.Ayoub has constructed an explicit object which gives the localization functor for the (D1,usu)
model strucure. We recall this in theorem 13(i) and give a relative version (with and without transfers)
in theorem 23 :
Theorem 1**.**
Let S∈AnSp(C),
(i)
For F∙∈PSh(AnSp(C)sm/S,C−(Z)),
singDˉ∗F∙∈PSh(AnSp(C)sm,C−(Z)) is D1 local and
the inclusion morphism S(F∙):F∙→singDˉ∗F∙ is an (D1,usu) equivalence.
(ii)
For F∙∈PShZ(CorZfs(AnSp(C)sm),C−(Z)),
singDˉ∗F∙∈PShZ(CorZfs(AnSp(C)sm),C−(Z))
is D1 local and
the inclusion morphism S(F∙):F∙→singDˉ∗F∙ is an (D1,usu) equivalence.
The categories AnDM−(S) and AnDA−(S), for S∈AnSp(C) satisfy the following (see [1] and [3])
Theorem 2**.**
(i)
The adjonction
[TABLE]
is a Quillen equivalence for the (D1,usu) model structures. That is, the derived functor
Tr(S)∗:AnDM−(S)∼AnDA−(S)
is an isomorphism and LTr(S)∗ is it inverse.
(ii)
The adjonction
(ean(S)∗,ean(S)):C−(S)⇆PShZ(AnSp(C)sm/S,C−(Z))
is a Quillen equivalence for the (I1,usu) model structures. That is, the derived functor
ean∗:D−(S)∼AnDA−(S,Z)
is an isomorphism and Rean(S)∗:AnDA−(S,Z)∼D−(S) is it inverse.
(iii)
The adjonction
(eantr(S)∗,eantr(S)∗):C−(S)⇆PShZ(CorZfs(AnSp(C)sm/S),C−(Z))
is a Quillen equivalence for the (D1,usu) model structures.
That is, the derived functor
eantr(S)∗:D−(S)∼AnDM−(S,Z)
is an isomorphism and Rean∗tr:AnDM−(S,Z)∼D−(S) is it inverse.
In this paper we give a construction of the Betti realization functor via the category of CW complexes.
The reason we do this is that the cycle class map on complex analytic spaces
and the action of correspondence on homology on smooth complex analytic spaces (i.e. complex analytic manifold)
are defined in a purely topological way so that we does not the need complex structure which gives in the smooth case
the Frölicher filtration on then
(the Frölicher filtration gives the Hodge fitration on cohomoloy in the compact Kalher case.
Let X,Y,Z∈Top.
Assume Y is Hausdorf (equivalently the diagonal ΔY⊂Y×Y is a closed subset).
There is a natural composition law (c.f.33) between closed subset of X×Y which are
finite and surjective over X and closed subset of Y×Z which are finite and surjective over Y.
We denote I1:=[0,1].
Consider now the full subcategory CW⊂Top consisting of CW complexes.
By a CW subcomplex of X∈CW, we mean a topological embbeding Z↪X with Z a CW complex.
By a closed CW subcomplex of X∈CW, we mean a topological closed embbeding Z↪X with Z a CW complex,
that is the image of the embedding is a closed subset of X.
Let X,S∈CW, S connected, and h:X→S a finite and surjective morphism in CW.
We say that X/S=(X,h)∈CW/S is reducible if X=X1∪X2, with X1,X2 closed CW subcomplexes
finite and surjective over S and X1,X2=X.
A pair Y/S=(Y,h′)∈CW/S with Y∈CW and h′:Y→S a finite and surjective morphism
is called irreducible if it is not reducible. In particular Y is connected.
•
For X∈CW, Λ a commutative ring and p∈N,
we denote by Zp(X,Λ) the free Λ module generated
by the closed CW subcomplex of X of dimension p and by
Zp(X,Λ)=ZdX−p(X,Λ)
the free Λ module generated by the closed CW subcomplex of X of codimension p.
•
For X,Y∈CW, X connected and Λ a commutative ring, we define :
Zfs/X(X×Y,Λ)⊂ZdX(X×Y,Λ)
the free Λ module generated by the closed CW subcomplexes of X×Y finite and surjective over X
which are irreducible.
•
For X,Y∈CW, and Λ a commutative ring, we define :
Zfs/X(X×Y,Λ):=⊕iZfs/Xi(Xi×Y,Λ)
where X=⊔iXi, with Xi the connected components of X.
We define the category of finite surjective correspondences on CW complexes.
Definition 2**.**
•
We define CorΛfs(CW) to be the category whose objects are CW complexes and whose space of morphisms between
X,Y∈CW is the free Λ module Zfs/X(X×Y,Λ).
The composition law is the one given by (35).
•
Let S∈CW. We define CWsm/S to be the category whose objects are X/S=(X,h) with
X∈CW and h:X→S a smooth morphism. Let X/S=(X,h1),Y/S=(Y,h2)∈CWsm/S.
A morphism f:X/S→Y/S is a morphism f:X→Y such that f∘h1=h2.
•
Let S∈CW. We define CorΛfs(CWsm/S) to be the category whose objects
are those of CWsm/S and whose space of morphisms between
X,Y∈CW is the free Λ module Zfs/X(X×SY,Λ).
The composition law is the one given by (35).
We define
•
CwDA−(Z),
the derived category of motives of CW complexes to be the homotopy category of
P−(CW)=PShZ(CW,C−(Z)),
the category of bounded above complex of presheaves on the category CW of CW complexes,
with respect to the projective (I1,usu) model structure (20(i)).
•
Similarly, we define CwDM−(Z), the derived category of motives of CW complexes, as
the homotopy category of PC−(CW)=PShZ(CorZfs(CW),C−(Z)),
the category of bounded above complex of presheaves on the category CorZfs(CW),
with respect to the projective (I1,usu) model structure (20(ii)).
•
For X∈CW and l:D↪X a CW subcomplex, the motive of the pair (X,D) is
M(X,D)=D(I1,usu)(Ztr(X,D))∈CwDM−(Z), where
Ztr(X,D)=coker(Ztr(l)) is the cokernel of the injective morphism of presheaves
Ztr(l):Ztr(D)↪Ztr(X) and
D(I1,usu):PC−(CW)→CwDM−(Z) is the localization functor.
We consider the commutative diagram
[TABLE]
the morphism of sites given respectively by the inclusion functors
ecw:{pt}↪CW,
ecwtr:{pt}↪CorZfs(CW)
and Tr:CorZfs(CW)↪CW.
For S∈CW, we define
•
CwDA−(S,Z),
the derived category of motives of CW complexes, as
the homotopy category of P−(CW,S)=PShZ(CWsm/S,C−(Z))
the category of bounded above complex of presheaves on CWsm/S.
with respect to the projective (I1,usu) model structure (38(i)).
•
Similarly, for S∈CW, we define CwDM−(S,Z),
the derived category of motives of CW complexes, as
the homotopy category of PC−(CW,S)=PShZ(CorZfs(CWsm/S),C−(Z))
the category of bounded above complex of presheaves on the category CorZfs(CWsm/S)
with respect to the projective (I1,usu) model structure (38(ii)).
For S∈CW, we consider the commutative diagram
[TABLE]
the morphism of sites given respectively by
the inclusion functors ecw(S):Ouv(S)↪CWsm/S,
ecwtr(T):Ouv(S)↪CorZfs(CWsm/S)
and Tr(S):CorZfs(CWsm/S)↪CWsm/S.
We consider first the absolute case.
We give an explicit object which induces the I1 localization functor on the category of presheaves
on CW and additive presheaves on CorZ(CW).
More precisely by, considering In:=[0,1]n,
we prove in theorem 16 the following :
Theorem 3**.**
•
For F∙∈PShZ(CW,C−(Z)),
singI∗F∙∈PShZ(CorZfs(CW),C−(Z))
is I1 local and the inclusion morphism
S(F∙):F∙→singI∗F∙ is an (I1,usu) equivalence.
•
For F∙∈PShZ(CorZfs(CW),C−(Z)),
singI∗F∙∈PShZ(CorZfs(CW),C−(Z))
is I1 local and the inclusion morphism
S(F∙):F∙→singI∗F∙ is an (I1,usu) equivalence.
Then, we prove that the categories CwDM− and CwDA− satisfy the following (c.f.theorem 17) :
Theorem 4**.**
(i)
The adjonction
(Tr∗,Tr∗):PSh(CW,C−(Z))⇆PShZ(CorZfs(CW),C−(Z))
is a Quillen equivalence for the (I1,usu) model structures. That is, the derived functor
[TABLE]
is an isomorphism
and Tr∗:CwDM−(Z)∼CwDA−(Z) is it inverse.
(ii)
The adjonction
(ecw∗,ecw∗):C−(Z)⇆PShZ(CW,C−(Z))
is a Quillen equivalence for the (I1,usu) model structures. That is, the derived functor
ecw∗:D−(Z)∼CwDA−(Z) is an isomorphism
and Recw∗:CwDA−(Z)∼D−(Z) is it inverse.
(iii)
The adjonction
(ecwtr∗,ecw∗tr):C−(Z)⇆PShZ(CorZfs(CW),C−(Z))
is a Quillen equivalence for the (I1,usu) model structures. That is, the derived functor
ecwtr∗:D−(Z)∼CwDM−(Z)
is an isomorphism and Recw∗tr:CwDM−(Z)∼D−(Z) is it inverse.
For point (i), we use proposition 21 to prove that LTr∗ is this inverse of Tr∗.
In proposition 21, we prove a key result that for X∈CW
[TABLE]
is an equivalence usu local. To see this we use the fact that a CW complex is I1
homotopy equivalent to a Δ-complex, and that for a Δ-complex there exist
a countable open covering such that the intersection of a finite number of members
of this covering is either empty or a contractible topological space.
We deduce from proposition 21 and proposition 17 the point (iii) of proposition 21
which says in particular that for X∈CW, the complex singI∗Ztr(X),
where Ztr(X) is the presheaf represented by X, is quasi-isomorphic to
the complex of singular chains C∗sing(X,Z). Indeed, considering
Z(Y,E)=coker(Z(l)), the cokernel of the injective morphism of presheaves
Z(l):Z(E)↪Z(Y), we have
Proposition 1**.**
For Y∈CW and l:E↪Y a CW subcomplex, the followings embeddings are quasi-isomorphism :
[TABLE]
where C∗sing(Y,E,Z)=cokerl∗ is the relative cohomology, with
l∗:ZHomCW(Δ∗,E)↪ZHomCW(Δ∗,Y).
Let X∈TM(R) be a differential manifold, Y∈CW and E⊂Y a subcomplex.
Let T=∑iniTi∈Zfs/In×X(In×X×Y) such that
∂I∗T:=∑i=1n(−1)n(T∣Ii,0n×X×Y−T∣Ii,1n×X×Y)=0.
Denote by pX:In×X×Y→X, pY:In×X×Y→Y and
pX×Y:In×X×Y→X×Y the projections,
mi:Ti↪In×X×Y the closed CW embeddings for all i.
Denote by pXi=pX∘mi:Ti→X and pYi=pX∘mi:Ti→Y.
The action of pX×Y(T)∈ZdX+n(X×Y,Z) on homology is
[TABLE]
where, for each i :
•
pXi∗∈HomD−(Z)(C∗(X,Z),C∗(Ti,Z)[n])
is the Gynsin morphism (pXi is proper and X∈TM(R) is a topological manifold),
•
pYi∗=Z(pYi)(Δ∗):C∗(Ti,Z)→C∗(Y,Z) is the classical map on singular chain,
•
cY,E:C∗(Y,Z)→C∗(Y,E,Z) is the quotient map.
We identify in proposition 23,
for X∈TM(R) a differential manifold, Y∈CW and E⊂Y a subcomplex, the image
of a morphism
[T]∈HomPC−(CW)(Ztr(X),singI∗Ztr(Y,E)[n])
under the (I1,usu) localization functor with the action of pX×Y(T) on homology :
Proposition 2**.**
Let X∈TM(R) connected and Y∈CW. Let l:E↪Y a CW subcomplex.
(i)
Let [T]=[∑iniTi]∈HomPC−(CW)(Ztr(X),singI∗(Ztr(Y,E))[n])
Then,
[TABLE]
(ii)
If Y is compact and E⊂Y is closed then the factorization
[TABLE]
where
[⋅]:ZdX+n(X×Y,X×E,Z)→HdX+nBM(X×Y,X×E,Z)
is the fundamental class gives the classical isomorphism comming from the six functor formalism (c.f.**[7]**)
[TABLE]
By definition, we have the following commutative diagram of sites :
[TABLE]
In proposition 25(ii), we prove that Cw∗, and hence \widetilde{\mathop{\rm Cw}}\nolimits^{*},
derive trivially for the (D1,usu), resp. (A1,et),
and (I1,usu) model structure.
Our definition of the Betti realization functor is the following :
Definition 3**.**
(i)
The CW-Betti realization functor (without transfers) is the composite :
[TABLE]
(ii)
The CW-Betti realization functor with transfers is the composite :
[TABLE]
Since \widetilde{\mathop{\rm Cw}}\nolimits^{*} derive trivially by proposition 25(ii)
and LTr∗:CwDA−(Z)→CwDM−(Z) is the inverse of Tr∗ by theorem 17(i), we have
\widetilde{\mathop{\rm Bti}}\nolimits_{0}^{*}=\widetilde{\mathop{\rm Bti}}\nolimits^{*}\circ L\mathop{\rm Tr}\nolimits^{*}.
In section 3.1 we prove (c.f. theorem 18) that the absolute version of this construction coincide with Ayoub’one.
Theorem 5**.**
(i)
For Y∈Var(C), and E⊂Y a subvariety, we have \mathop{\rm Bti}\nolimits^{*}M(Y,E)=\widetilde{\mathop{\rm Bti}}\nolimits^{*}M(Y,E)
(ii)
For X,Y∈Var(C), D⊂X, E⊂Y subvarieties,
and n∈Z, n≤0, the following diagram is commutative
[TABLE]
where we denoted for simplicity X for Xan and Xcw, and similarly for D, Y and E.
To prove this theorem, we give, for X∈AnSp(C), an equivalence (D1,usu) local
B(Ztr(X)):singD∗Ztr(X)→Cw∗singI∗Ztr(Xcw) by
considering the two canonical morphism of functors (see section 3.1),
•
the morphism ψCw, which,
for G∙∈PC−(An), associate the following canonical morphism
ψCw(G∙):Cw∗(singIan∗G∙)→singI∗Cw∗G∙ in PC−(CW),
•
the morphism ψCw, which,
for F∙∈PC−, associate the following canonical morphism
ψCw(F∙):Cw∗(singIet∗F∙)→singI∗Cw∗F∙ in PC−(CW),
we define the following two morphism of functors :
(i)
the morphism W, which,
for G∙∈PC−(An), associate the composition in PC−(CW)
[TABLE]
(ii)
the morphism W, which,
for F∙∈PC−, associate the composition in PC−(CW)
[TABLE]
The morphism of functor B is defined, by associating
to G∙∈PSh(CorZfs(AnSm(C)),C−(Z)), the composite
[TABLE]
in PSh(CorZfs(AnSm(C)),C−(Z))
We deduce this theorem from the proposition 27 :
Proposition 3**.**
(i)
For Y∈AnSp(C) and E⊂Y an analytic subset,
[TABLE]
is a quasi isomorphism in C−(Z).
(ii)
For Y∈AnSp(C), and E⊂Y an analytic subset, the morphism
[TABLE]
is an equivalence (D1,usu) local in PSh(CorZfs(AnSm(C)),C−(Z)).
We deduce the point (ii) of this proposition from point (i). To prove point (i) we reduce to the smooth case
using a desingularization by [9].
For X∈AnSp(C) the morphism of complexes
[TABLE]
is given by associating to α∈Zfs/Dnˉ(Dˉn×X), the
restriction Ztr(Xcw)(in)(αcw)=α∣In×Xcw
of αcw∈Zfs/Dˉn(Dn×Xcw) by the closed embedding of CW complexes
in×IXcw:In×Xcw↪Dn×Xcw
induced by the closed embedding in:In↪Dˉn.
If X is smooth connected, we see using a covering by geodesically convex open subset,
we see that there exist a countable open covering by open subsets isomorphic to open balls in CdX
such that the intersection of a finite number of members are either empty or isomorphic to an open ball in CdX.
In the section 3.2, we use our construction of the Betti realization functor via CW complexes
to give, for X,Y∈Var(C) and E⊂Y a closed subvariety,
an explicit image of a morphism α∈HomPC−(Ztr(X),Ztr(Y,E)[n])
by the Betti realization functor.
Definition 4**.**
(i)
For V∈Var(C) and p∈N, we consider
the Bloch cycle complex Zp(V,∗), and
Zp(Vcw,∗) the cubical complex such that for n∈N,
Zp(Vcw,n)⊂Zp(In×Xcw) is the abelian subgroup
consisting of cycle meeting the face of In properly
and whose differential is given by, for γ∈Zp(Vcw,n),
∂Iγ=∑i=1n(−1)i(γ∣Ii,0n−1×Vcw−γ∣Ii,1n−1×Vcw)
where Ii,0,Ii,1⊂In are the faces.
We then denote
[TABLE]
the map complexes given by restriction with respect to the closed embedding of CW complexes
in×IV:In×Vcw↪□n×Vcw whose image
is [0,∞]n×Vcw⊂□n×Vcw (c.f. notations).
(ii)
Let V∈Var(C) quasi-projective.
Let Y∈PVar(C) be a compactification of V (e.g. the projectivisation of V)
and E=Y\V.
The higher cycle class map (**[11]**) is the morphism of complexes of abelian groups :
[TABLE]
where αˉ∈Zp(Y×□∗) is the closure of α and
pY:Ycw×I∗→Ycw is the projection.
Let X∈SmVar(C) and Y∈Var(C).
Let E⊂Y be a closed subset and V=Y\E.
Denote by j:V↪Y the open embeddings.
We have the open embedding IX×j:X×V↪X×Y.
The map of complexes
[TABLE]
induces maps denoted by the same way on the subcomplexes indicated in the following diagram
in C−(Z):
Recall that Dtr(A1,et):PC−→DM−(C,Z)) and
Dtr(I1,usu):PC−(CW)→CwDM−(Z) are the canonical localization functors.
We prove in proposition 28 the following
Proposition 4**.**
Let X∈SmVar(C) and Y∈Var(C).
Let E⊂Y be a closed subset and V=Y\E.
For n∈Z, n≤0,
the following diagram is commutative
[TABLE]
On the other side, we identified in proposition 23 in section 2.3 the image
of a morphism
T∈HomPC−(CW)(Ztr(Xcw),singI∗Ztr(Ycw,Ecw)[n])
under the (I1,usu) localization functor with the action on homology :
Using proposition 28 and proposition 23 (i),
we immediately deduce from theorem 18 the following (c.f.3):
Corollary 1**.**
Let X∈SmVar(C), Y∈Var(C), E⊂Y a closed subvariety
and V=Y\E the open complementary.
Let n∈Z, n≤0. Then,
(i)
the following diagram (90) is commutative
[TABLE]
where we denoted for simplicity X for Xan and Xcw, and similarly for Y and E.
(ii)
for α∈HomPC−(Ztr(X),C∗Z(Y,E)[n]),
we have
[TABLE]
where pX×Y:In×Xcw×Ycw→Xcw×Ycw is the projection.
We deduce from this corollary 3(ii),
proposition 29, end lemma 11, the following main result
which say that Ayoub’s Betti realization functor factor through Nori motives.
Consider the functor N:CbCorZ(SmVar(C))→Db(N)
from the category of bounded complexes of correspondences of algebraic varieties
to the derived category of Nori motives (c.f.[12]).
This functor factor through the localization functor :
[TABLE]
Denote by oN:Db(N)→Db(Z) the forgetful functor.
In theorem 19 we prove :
Theorem 6**.**
(i)
For X∈SmVar(C),
Bti∗∘D(A1,et)(Z(X))=oN∘N(X)
(ii)
For X,Y∈SmVar(C), the following diagram commutes
[TABLE]
(iii)
The Betti realisation functor factor through Nori motives. That is
Bti∗=oN∘Nˉ
Let V∈Var(C) quasi-projective.
Let Y∈PVar(C) a compactification of V and E=Y\V.
Denote by j:V↪Y the open embedding.
•
For p≤dV, consider the following composition of isomorphisms of abelian groups
[TABLE]
•
For p≥dV and E′=(Y×Pp−dV)\(V×Ap−dV),
consider the following composition of isomorphisms of abelian groups
[TABLE]
.
We prove that under these identifications, the image of the Betti realization
functor on morphism coincide with the Bloch cycle class map (c.f. theorem 20) :
Theorem 7**.**
Let V∈Var(C). Let Y∈PVar(C) be a compactification of V and E=Y\V.
Then,
(i)
for p≤dV, the following diagram commutes :
[TABLE]
(ii)
for p≥dV, the following diagram commutes :
[TABLE]
In the last section we give a relative version of theorem 18.
We first give in theorem 25 a relative version of theorem 16.
Theorem 8**.**
Let S∈CW.
(i)
For F∙∈PShZ(CorZfs(CWsm/S),C−(Z)),
singI∗F∙∈PShZ(CorZfs(CWsm/S),C−(Z))
is I1 local and the inclusion morphism
S(F∙):F∙→singI∗F∙ is an (I1,usu) equivalence.
(ii)
For F∙∈PShZ(CorZfs(CWsm/S),C−(Z)),
singI∗F∙∈PShZ(CorZfs(CWsm/S),C−(Z))
is I1 local and the inclusion morphism
S(F∙):F∙→singI∗F∙ is an (I1,usu) equivalence.
We then prove in theorem 26, a relative version of point (ii) and (iii) of 17 :
Theorem 9**.**
Let S∈CW.
(i)
The adjonction
(ecw(S)∗,ecw(S)∗):C−(S)⇆PShZ(CWsm/S,C−(Z))
is a Quillen equivalence for the (I1,usu) model structures. That is, the derived functor
ecw(S)∗:D−(S)∼CwDA−(S,Z) is an isomorphism
and Recw(S)∗:CwDA−(S,Z)∼D−(S) is it inverse.
(ii)
The adjonction
(ecwtr(S)∗,ecwtr(S)∗):C−(S)⇆PShZ(CorZfs(CWsm/S),C−(Z))
is a Quillen equivalence for the (I1,usu) model structures. That is, the derived functor
ecwtr(S)∗:D−(S)∼CwDM−(S,Z) is an isomorphism
and Recwtr(S)∗:CwDM−(S,Z)∼D−(S) is it inverse.
By definition, for each S∈Var(C), we have (c.f.114)
the following commutative diagram of sites DCat(S)
[TABLE]
For T,S∈Var(C) and f:T→S a morphism, the morphism of sites
•
P(f):Var(C)/T→Var(C)/S, P(fan):AnSp(C)/Tan→AnSp(C)/San,
and P(fcw):CW/Tcw→CW/Scw
given by the pullback functor,
•
P(f):CorZfs(Var(C)sm/T)→CorZfs(Var(C)sm/S),
P(fan):CorZfs(AnSp(C)/Tan)→CorZfs(AnSp(C)/San), and
P(fcw):CorZfs(CW/Tcw)→CorZfs(CWsm/Scw)
given by the pullback functor,
gives a morphism of diagram of sites
[TABLE]
Our definition of the Betti realization functor in the relative setting is :
Definition 5**.**
Let S∈Var(C).
(i)
The CW-Betti realization functor (without transfers) is the composite :
[TABLE]
(ii)
The CW-Betti realisation functor with transfers is the composite :
[TABLE]
Similarly, in the relative case, since \widetilde{\mathop{\rm Cw}}\nolimits(S)^{*} derive trivially by proposition 44(ii)
and and LTr(Scw)∗:CwDA−(Scw,Z)→CwDM−(Scw,Z) is the inverse of Tr(Scw)∗
(c.f.remark 1), we have \widetilde{\mathop{\rm Bti}}\nolimits_{0}(S)^{*}=\widetilde{\mathop{\rm Bti}}\nolimits(S)^{*}\circ L\mathop{\rm Tr}\nolimits(S)^{*}.
As in Ayoub’s definition, it defines morphisms of homotopic 2-functors (c.f. theorem 28 :
We finally prove in theorem 29 a relative version of the theorem 18,
that is the construction of the relative Betti realization functor via CW complexes coincide
with Ayoub’s one via analytic spaces :
Theorem 10**.**
Let S∈Var(C).
Let M∈DM−(S,Z) is a constructible motive.
(i)
We have \mathop{\rm Bti}\nolimits^{*}M=\widetilde{\mathop{\rm Bti}}\nolimits^{*}M
(ii)
Let M1,M2∈DM−(S,Z) contructible motives.
Let F1∙,F2∙∈PC−(S)
such that Mi=D(A1,et)(S)(Fi∙)
for i=1,2. The following diagram is commutative
[TABLE]
As in the absolute case, we define for S∈AnSp(C), the morphism of functor B(S), by associating
to G∙∈PSh(CorZfs(AnSpsm(C)/S),C−(Z)),
the morphism B(S)(G∙) which is the composite
[TABLE]
in PSh(CorZfs(AnSp(C)sm/S),C−(Z)),
and deduce this theorem from the point (ii) of following proposition :
Proposition 5**.**
Let S∈Var(C) and F∙∈PC−(S)
such that D(A1,et)(S)(F∙)∈DM−(S,Z) is a constructible motive. Then,
(i)
e^{tr}_{an}(S)_{*}B(S)(F^{\bullet}):\mathop{\rm sing}\nolimits_{\bar{\mathbb{D}}^{*}}\mathop{\rm An}\nolimits(S)^{*}F^{\bullet}\to\mathop{\rm sing}\nolimits_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}*
is an equivalence usu local in C−(San).*
(ii)
B(S)(F^{\bullet}):\underline{\mathop{\rm sing}\nolimits}_{\bar{\mathbb{D}}^{*}}\mathop{\rm An}\nolimits(S)^{*}F^{\bullet}\to\mathop{\rm Cw}\nolimits(S^{an})_{*}\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}*
is an equivalence (D1,usu) local.*
As in the absolute case, we deduce the point (ii) of this proposition from point (i).
We prove the point (i) of this proposition using the absolute case
(point (i) of the proposition 27).
2 Derived categories of motives of algebraic varieties, analytic spaces, and CW complexes
2.1 The derived category of mixed motives of algebraic varieties
For X∈Var(k), and Λ a commutative ring and p∈N,
we denote by Zp(X,Λ) the free Λ module
generated by the irreducible closed subspaces of X of dimension p,
by Zp(X,Λ)=ZdX−p(X,Λ)
the free Λ module generated by the irreducible closed subspaces of X of codimension p.
An algebraic variety X∈Var(k) is aid to be proper (or complete) if the terminal map aX:X→Speck
is universally closed, that is for all Y∈Var(k),
the projection pY:X×kY→Y is closed (note that the topology on X×kY is finer than the
product topology on the underlying topological spaces).
For X∈Var(k), a compactification of X is a complete variety Xˉ∈Var(k) such that X⊂Xˉ
is an open subset. For X∈Var(k) quasi-projective,
the projectivisation Xˉ∈PVar(k) is a compactification of X.
For X∈Var(k), we have the Bloch cycle complex Z(X,∗), with for n∈N,
Zp(X,n)⊂Zp(X×□n) is the subgroup of the p codimentional irreducible closed
subspace of X×□n meeting all faces of □n properly.
For X,Y∈Var(k) and Λ a commutative ring, we will use the following notations :
•
if X is smooth connected Zfs/X(X×kY,Λ)⊂ZdX(X×kY,Λ)
is the free Λ submodule generated by the irreducible closed subspaces
of X×kY which are finite and surjective over X,
•
if X is smooth, Zfs/X(X×Y,Λ):=⊕iZfs/Xi(Xi×Y,Λ)
where X=⊔iXi, with Xi the connected components of X,
•
Zed(r)/X(X×kY,Λ)⊂ZdX+r(X×kY,Λ)
the free Λ module generated by the irreducible closed subspaces of X×kY
dominant over X, and whose fibers over X are either empty or equidimensional of relative dimension r.
By definition, for X smooth,
Zfs/X(X×kY,Λ)⊂Zed(0)/X(X×kY,Λ)
and Zed(p−dX)/□n(□n×kX,Λ)⊂Zp(X,n)⊗ZΛ.
For Y∈Var(k) irreducible and j:V↪Y an open embedding,
we denote by
[TABLE]
Definition 6**.**
[15]**
We define CorΛfs(SmVar(k)) to be the category whose objects are smooth algebraic varieties over k
and whose space of morphisms between
X,Y∈SmVar(k) is the free Λ module Zfs/X(X×kY,Λ).
The composition of morphisms is defined in [15]
We have
•
the additive embedding of categories Tr:Z(SmVar(k))↪CorZfs(SmVar(k))
which gives the corresponding morphism of sites Tr:CorZfs(SmVar(k))→Z(SmVar(k)).
•
the inclusion functor
evar:{pt}↪SmVar(C),
which gives the corresponding morphism of sites
evar:SmVar(C)→{pt},
•
the inclusion functor evartr:=Tr∘evar:{pt}↪CorZfs(SmVar(C))
which gives the corresponding morphism of sites
evartr:=Tr∘evar:CorZfs(SmVar(C))→{pt}.
We consider the following two big categories :
•
PSh(SmVar(k),C−(Z))=PShZ(Z(SmVar(k)),C−(Z)),
the category of bounded above complexes of presheaves on SmVar(k), or equivalently additive presheaves on Z(SmVar(k)),
sometimes, we will write for short P−=PSh(SmVar(C),C−(Z)),
•
PShZ(CorZfs(SmVar(k)),C−(Z)),
the category of bounded above complexes of additive presheaves on CorZfs(SmVar(k))
sometimes, we will write for short PC−=PSh(CorZfs(SmVar(C)),C−(Z)),
given by Tr:CorZfs(SmVar(k))→Z(SmVar(k)), evar:SmVar(k)→{pt}
and evartr:CorZfs(SmVar(k))→{pt} respectively.
We denote by aet:PShZ(SmVar(k),Ab)→ShZ,et(SmVar(k),Ab)
the etale sheaftification functor.
For X∈SmVar(k), we denote by
[TABLE]
the presheaves represented by X. They are etale sheaves.
which are etale sheaves.
Of course, if X∈PVar(k), then Zeq(X,0)=Ztr(X).
We consider
•
the usual monoidal strutcure on PSh(SmVar(k)),C−(Z)) and the associated internal Hom
given by, for F∙,G∙∈PSh(SmVar(k)),C−(Z)) and Y∈SmVar(k),
[TABLE]
•
the unique monoidal strutcure on PShZ(CorZfs(SmVar(k)),C−(Z))
such that,for X,Y∈SmVar(k), Ztr(X)⊗Ztr(Y):=Ztr(X×kY)
and wich commute with colimites.
It has an internal Hom which is given, for X,Y∈SmVar(k) and
F∙∈PShZ(CorZfs(SmVar(k)),C−(Z)),
Hom(Ztr(Y),F∙):X↦F∙(X×kY)
Together with these monoidal structure, the functor
We say that a morphism ϕ:G1∙→G2∙ in PSh(SmVar(k),C−(Z)) is an
etale local equivalence if ϕ∗:aetHk(G1∙)→aetHk(G2∙) is an isomorphism for all k∈Z.
The projective etale topology model structure on PSh(SmVar(k),C−(Z))
is the left Bousfield localization of the projective model structure MP(PShZ(SmVar(k),C−(Z)))
with respect to the etale local equivalence.
(ii)
We say that a morphism
ϕ:G1∙→G2∙ in PShZ(CorZfs(SmVar(k)),C−(Z)) is an
etale local equivalence if and only if its restriction to SmVar(C)Tr∗ϕ:Tr∗G1∙→Tr∗G2∙ is an etale local equivalence.
The projective etale topology model structure on PShZ(CorZfs(SmVar(k)),C−(Z))
is the left Bousfield localization of the projective model structure
MP(PShZ(CorZfs(SmVar(k)),C−(Z)))
with respect to the etale local equivalence.
Definition 8**.**
(i)
The projective (Ak1,et) model structure on PSh(SmVar(k),C−(Z)) is the left Bousfield localization
of the projective etale topology model structure (c.f. definition 7(i))
with respect to the class of maps
{Z(X×Ak1)[n]→Z(X)[n],X∈SmVar(k),n∈Z}.
(ii)
The projective (Ak1,et) model structure on PShZ(CorZfs(SmVar(k)),C−(Z))
is the left Bousfield localization of the projective etale topology model structure (c.f. definition 7(ii))
with respect to the class of maps
{Ztr(X×Ak1)[n]→Z(X)[n],X∈SmVar(k),n∈Z}.
Definition 9**.**
(i)
We define
DM−(k,Z)et:=HoAk1,et(PShZ(CorZfs(SmVar(k)),C−(Z))),
to be the derived category of (effective) motives, it is
the homotopy category of the category PSh(CorZfs(SmVar(k)),C−(Z))
with respect to the projective (A1,et) model structure (cf. definition 8(ii)).
We denote by
[TABLE]
the canonical localisation functor.
(ii)
By the same way, we denote
DA−(k,Z)et:=HoAk1,et(PSh(SmVar(k),C−(Z)))(cf.8(i)) and
[TABLE]
the canonical localisation functor.
For X∈Var(k),
•
the (derived) motive of X is M(X)=D(A1,et)(Ztr(X))∈DM−(k).
•
the (derived) motive with compact support of X is Mc(X)=D(A1,et)(Zeq(X,0))∈DM−(k,Z).
Of course, if X∈PVar(k), then Zeq(X,0)=Ztr(X), so that Mc(X)=M(X).
For F∙∈PSh(SmVar(k),Ab) and X∈SmVar(k),
we have the complex F(X×□k∗) associated to the cubical object
F(X×□k∗) in the category of abelian groups.
•
If F∙∈PSh(SmVar(k),C−(Z)) ,
[TABLE]
is the total complex of presheaves associated to the bicomplex of presheaves X↦F∙(□k∗×kX),
and C∗F∙:=evar∗C∗F∙=F∙(□k∗)∈C−(Z).
We denote by S(F∙):F∙→C∗F∙,
[TABLE]
the inclusion morphism of PSh(SmVar(k),C−(Z)) :
For f:F1∙→F2∙ a morphism PSh(SmVar(k),C−(Z)), we denote by
S(f):C∗F1∙→C∗F2∙,
the morphism of PSh(SmVar(k),C−(Z)) given by for X∈SmVar(k),
[TABLE]
•
If F∙∈PShZ(CorZfs(SmVar(k)),C−(Z)),
[TABLE]
is the complex of presheaves
associated to the bicomplex of presheaves X↦F∙(□k∗×kX),
and C∗F∙:=evar∗trC∗F∙=F∙(□k∗)∈C−(Z).
We have the inclusion morphism (12)
[TABLE]
which is a morphism in PSh(CorZfs(SmVar(k)),C−(Z))
denoted the same way S(F∙):F∙→C∗F∙.
For f:F1∙→F2∙ a morphism PC−, we
have the morphism (13)
[TABLE]
which is a morphism in PC− denoted the same way
S(f):C∗F1∙→C∗F2∙.
For F∙∈PSh(CorZfs(SmVar(k)),C−(Z)), we have by definition
Tr∗C∗F∙=C∗Tr∗F∙ and Tr∗S(F∙)=S(Tr∗F∙).
We now make the following definition
Definition 10**.**
Let X∈Var(k) and D⊂X a subvariety. Denote by l:D↪X the locally closed embbeding.
We define
•
Z(X,D)=coker(Z(l))∈PShZ(SmVar(k),C−(Z))*
to be the cokernel of the injective morphism Z(l):Z(D)↪Z(X).
By definition, we have the following exact sequence*
[TABLE]
in PShZ(SmVar(k),C−(Z)).
•
Ztr(X,D)=coker(Ztr(l))∈PShZ(CorZfs(SmVar(k)),C−(Z))*
to be the cokernel of the injective morphism Ztr(l):Ztr(D)↪Ztr(X).
By definition, we have the following exact sequence*
[TABLE]
in PShZ(CorZfs(SmVar(k)),C−(Z)).
By the exact sequences, we have Tr∗Z(X,D)=Ztr(X,D)
In particular, we get the following exact sequence in PShZ(CorZfs(SmVar(k)),C−(Z))
[TABLE]
We define
[TABLE]
to be the relative motive of the pair (Y,E).
We now look at the behavior of the functors mentionned above with respect to the (A1,et) model structure
(Tr∗,Tr∗):PSh(SmVar(k),C−(Z))⇆PShZ(CorZfs(SmVar(k)),C−(Z))*
is a Quillen adjonction for the etale topology model structures (c.f. definition 7 (i) and (ii) respectively)
and a Quillen adjonction for the (A1,et) model structures (c.f. definition 8 (i) and (ii) respectively).*
(ii)
(evar∗,evar∗):PSh(SmVar(C),C−(Z))⇆C−(Z)*
is a Quillen adjonction for the etale topology model structure (c.f. definition 7 (i))
and a Quillen adjonction for the (A1,et) model structure (c.f. definition 8 (i)).*
(iii)
(evartr∗,evar∗tr):PSh(CorZfs(SmVar(k)),C−(Z))⇆C−(Z)*
is a Quillen adjonction for the etale topology model structure (c.f. definition 7 (ii))
and a Quillen adjonction for the (A1,et) model structure (c.f. definition 8 (ii)).*
For F∙∈PSh(SmVar(k),C−(Z)),
C∗F∙∈PSh(SmVar(k),C−(Z)) is Ak1 local and
the inclusion morphism S(F∙):F∙→C∗F∙ is an (Ak1,et) equivalence.
(ii)
For F∙∈PShZ(CorZfs(SmVar(k)),C−(Z)),
C∗F∙∈PShZ(CorZfs(SmVar(k)),C−(Z))
is Ak1 local and the inclusion morphism
S(F∙):F∙→C∗F∙ is an (Ak1,et) equivalence.
Theorem 12**.**
[2]**
The adjonction
(Tr∗,Tr∗):PSh(SmVar(k),C−(Z))⇆PShZ(CorZfs(SmVar(k)),C−(Z))
is a Quillen equivalence for the (A1,et) model structures.
That is, the derived functor
[TABLE]
is an isomorphism and
Tr∗:DM−(k,Z)et∼DA−(k,Z)et is it inverse.
The following proposition identify the relative motive of a closed pair (Y,E) of projective varieties to the
motive of compact support of V=Y\E
Proposition 8**.**
[15]**
Let V∈Var(k) quasi projective.
Let Y∈PVar(k) a compactification of V with E=Y\V.
Denote by j:V↪Y the open embedding and i:E↪Y.
Then we have the following exact sequence in PShZ(CorZfs(SmVar(k)),C−(Z))
[TABLE]
with j∗(X):=(IX×j)∗:Zfs/X(X×Y,Z)→Zed(0)/X(X×V,Z).
This say (c.f. definition 10) that j∗
induces a quasi-isomorphism in PShZ(CorZfs(SmVar(k)),C−(Z))
[TABLE]
We finish this subsection by the following result of Suslin and Voevodsky :
Proposition 9**.**
Let X∈SmVar(k), Y∈PVar(k) and n∈Z, n<0.
The morphism of abelian group
[TABLE]
of the functor
D(A1,et):PShZ(CorZfs(SmVar(k)),C−(Z))→DM−(k,Z)
is an isomorphism.
Proof.
For X∈SmVar(k) and Y∈Var(k), we have the commutative diagram
[TABLE]
The equality on the right follow from the fact that
S(Ztr(X)):Ztr(X)→C∗Ztr(X) is an equivalence (A1,et) local.
Moreover, we have (c.f.[15])
Hetn(X,C∗Ztr(Y))=HZarn(X,C∗Ztr(Y))
Now if Y∈PVar(k) is projective, C∗Ztr(Y)=C∗Zeq(Y,0)
satisfy Zariski descent (c.f.[15]), hence h(X) is an isomorphism.
∎
2.2 The derived category of mixed motives of analytic spaces
For X∈AnSp(C), and Λ a commutative ring and p∈N,
we denote by Zp(X,Λ) the free Λ module generated
by the irreducible closed analytic subspaces of X of dimension p.
For X∈AnSp(C) irreducible, and Λ a commutative ring and p∈N,
we denote by Zp(X,Λ)=ZdX−p(X,Λ)
the free Λ module generated by the irreducible closed analytic subspaces of X of codimension p.
For X,Y∈AnSp(C) and Λ a commutative ring, we denote by
•
if X is smooth connected, Zfs/X(X×Y,Λ)⊂ZdX(X×Y,Λ)
the free Λ submodule generated by the irreducible closed analytic subspaces of X×Y which are finite and surjective over X,
•
if X is smooth, Zfs/X(X×Y,Λ):=⊕iZfs/Xi(Xi×Y,Λ)
where X=⊔iXi, with Xi the connected components of X.
Definition 11**.**
[3]**
Let CorΛfs(AnSm(C)) be the category whose objects are complex analytic varieties over C and whose space of morphisms between
X,Y∈AnSm(C) is the free Λ module Zfs/X(X×Y,Λ).
The composition of morphisms is defined in [3].
We have
•
the additive embedding of categories Tr:Z(AnSm(C))↪CorZfs(AnSm(C))
which gives the corresponding morphism of sites Tr:CorZfs(AnSm(C))→Z(AnSm(C)),
•
the inclusion functor
ean:{pt}↪AnSm(C),
which gives the corresponding morphism of sites
ean:AnSm(C)→{pt},
•
the inclusion functor eantr:=Tr∘ean:{pt}↪CorZfs(AnSm(C))
which gives the corresponding morphism of sites
eantr:=Tr∘ean:CorZfs(AnSm(C))→{pt}.
We consider the following two big categories :
•
PSh(AnSm(C),C−(Z))=PShZ(Z(AnSm(C),C−(Z))),
the category of bounded above complexes of presheaves on AnSm(C),
or equivalently additive presheaves on Z(AnSm(C)),
sometimes, we will write for short P−(An)=PSh(AnSm(C),C−(Z)),
•
PSh(CorZfs(AnSm(C)),C−(Z)),
the category of bounded above complexes of additive presheaves on CorZfs(AnSm(C))
sometimes, we will write for short PC−(An)=PSh(CorZfs(AnSm(C)),C−(Z)),
given by Tr:CorΛfs(AnSm(C))→Λ(AnSm(C)),
eanAnSm(C)→{pt} and eantr:CorZfs(AnSm(C))→{pt}.
We denote by
ausu:PShZ(AnSm(C),Ab)→ShZ,usu(AnSm(C),Ab)
the sheaftification functor for the usual topology.
For X∈AnSm(C), we denote by
[TABLE]
the presheaves represented by X. They are usu sheaves.
For X∈AnSp(C) a complex analytic space non smooth,
[TABLE]
is also an usu sheaf,
We consider
•
the usual monoidal strutcure on PSh(AnSm(C)),C−(Z)) and the associated internal Hom
given by, for F∙,G∙∈PSh(AnSm(C)),C−(Z)) and Y∈AnSm(C),
[TABLE]
•
the unique monoidal strutcure on PShZ(CorZfs(AnSm(C)),C−(Z))
such that,for X,Y∈AnSm(C), Ztr(X)⊗Ztr(Y):=Ztr(X×Y)
and wich commute with colimites.
It has an internal Hom which is given, for X,Y∈AnSm(C) and
F∙∈PShZ(CorZfs(AnSm(C)),C−(Z)),
Hom(Ztr(Y),F∙):X↦F∙(X×Y)
Together with these monoidal structure, the functor
We say that a morphism ϕ:G1∙→G2∙ in PSh(AnSm(C),C−(Z)) is an
usu local equivalence if ϕ∗:ausuHk(G1∙)→ausuHk(G2∙) is an isomorphism for all k∈Z.
The projective usual topology model structure on PSh(AnSm(C),C−(Z))
is the left Bousfield localization of the projective model structure MP(PShZ(AnSm(C),C−(Z)))
with respect to the usu local equivalence.
(ii)
We say that a morphism ϕ:G1∙→G2∙
in PShZ(CorZfs(AnSm(C)),C−(Z)) is an
usu local equivalence if and only if its restriction to AnSm(C)Tr∗ϕ:Tr∗G1∙→Tr∗G2∙ is an usu local equivalence.
The projective usual topology model structure on PShZ(CorZfs(AnSm(C)),C−(Z))
is the left Bousfield localization of the projective model structure
MP(PShZ(CorZfs(AnSm(C)),C−(Z)))
with respect to the usu local equivalence.
Definition 13**.**
(i)
The projective (D1,usu) model structure on PSh(AnSm(C),C−(Z)) is the left Bousfield localization
of the projective usual topology model structure (c.f. definition 12(i))
with respect to the class of maps
{Z(X×D1)[n]→Z(X)[n],X∈AnSm(C),n∈Z}.
(ii)
The projective (D1,usu) model structure on PShZ(CorZfs(AnSm(C)),C−(Z)) is the left Bousfield localization of the projective usual topology model structure (c.f. definition 12(ii))
with respect to the class of maps
{Ztr(X×D1)[n]→Z(X)[n],X∈AnSm(C),n∈Z}.
Definition 14**.**
(i)
We define
AnDM−(Z):=HoD1,usu(PShZ(CorZfs(AnSm(C)),C−(Z))),
to be the derived category of motives of complex analytic space, it is
the homotopy category of the category PShZ(CorZfs(AnSm(C)),C−(Z))
with respect to the projective (D1,usu) model structure (c.f. definition 13(ii)).
We denote by
[TABLE]
the canonical localization functor.
(ii)
We denote by the same way
AnDA−(Z):=HoD1,usu(PSh(AnSm(C),C−(Z))) (c.f. definition 13(i)) and
[TABLE]
the canonical localization functor.
For X∈AnSp(C), the (derived) motive of X is
M(X)=D(D1,usu)(Ztr(X))∈AnDM−(Z).
Recall Δ denote the simplicial category
Let
•
pΔ:AnSm(C)→Δ×AnSm(C)
be the morphism of site given by the projection functor
(X,i)∈Δ×AnSm(C)↦pΔ((X,i))=X∈AnSm(C),
•
pΔ:CorZ(AnSm(C))→Δ×CorZ(AnSm(C))
be the morphism of site given by the projection functor
(X,i)∈Δ×AnSm(C)↦pΔ((X,i))=X∈AnSm(C)
In following lemma point (ii) is a generalization of point (i) to hypercoverings.
We will only use point (i) in this paper.
Lemma 1**.**
(i)
Let X∈AnSm(C) and X=∪i∈JUi a covering by open subsets Ui⊂X, J being a countable set.
We denote by ji:Ui↪X the open embedding.
For I⊂J a finite subset, let UI=∩i∈IUi.
Then,
–
[⋯→⊕cardI=rZ(UI)→⋯→⊕i∈JZ(Ui)]⊕i∈JZ(ji)Z(X)*
is an usu local equivalence in P−(An),*
–
[⋯→⊕cardI=rZtr(UI)→⋯→⊕i∈JZtr(Ui)]⊕i∈JZtr(ji)Ztr(X)*
is an usu local equivalence in PC−(An)*
(ii)
Let X∈AnSm(C) and j∙:U∙→X an hypercovering of X by open subset
(i.e. for all n∈Δ, Un is an open subset and jn:Un↪X is the open embedding),
and (U∙,∙):⋯→(Un,n)→⋯ the associated complex in Δ×AnSm(C)
Then,
–
j:LpΔ∗(Z(U∙,∙))→Z(X)* is an isomorphism
in Housu(PSh(AnSm(C),C−(Z)))*
–
j:LpΔ∗(Ztr(U∙,∙))→Ztr(X)* is an isomorphism
in Housu(PShZ(CorZ(AnSm(C)),C−(Z)))*
Proof.
(i): It follows from the following two facts :
•
the sequence
[⋯→⊕cardI=rZ(UI)→⋯→⊕i∈JZ(Ui)]
is clearly exact in P−(An),
•
the exactness of
[⊕i∈JZ(Ui)⊕i∈JZ(ji)Z(X)→0]
in Housu(PSh(AnSm(C),C−(Z))),
follows from the fact that for Y∈AnSm(C), y∈Y and f:Y→X a morphism,
there exists an open subset V(y)⊂Y containing y such that f(V(y))⊂Ui, with
Ui containing f(y).
Similary, the second point follows from the following two facts :
•
the sequence
[⋯→⊕cardI=rZtr(UI)→⋯→⊕i∈JZtr(Ui)]
is clearly exact in PC−(An),
•
the exactness of
[⊕i∈JZtr(Ui)⊕i∈JZtr(ji)Ztr(X)→0]
in Housu(PShZ(CorZfs(AnSm(C)),C−(Z))),
follows from the fact that for Y∈AnSm(C), y∈Y and α∈Ztr(X)(Y) irreducible,
there exists an open subset V(y)⊂Y containing y such that α∣V(y)∈Ztr(Ui)(V(y)), with
Ui containing α∗y.
(ii): see [1] Proposition 1.4 Etape 1 with our notation
pΔ∗ for pΔ.
∎
The following proposition is to use point (i) instead of point (ii)
in the proof of the smooth case of point (i) of the proposition 27.
Proposition 10**.**
Let X∈AnSm(C) connected. Then there exist a countable open covering X=∪i∈JDi such that
for all finite subset I⊂J, DI:=∩i∈IDi=∅ or DI≃DdX is
bihomolomorphic to an open ball DdX⊂CdX.
Proof.
Let x∈X. As X is smooth, there exist an open neighborhood Dx⊂X of x in X such that
Dx is geodesically convex and such that Dx↪CdX.
As X a countable union of compact subset, we can extract a countable covering X=∪i∈JDJ of
the open covering X=∪x∈XDx. As DI is geodesically convex and DI admits an open embedding
DI↪CdX in CdX, DI≃DdX.
∎
We denote Dˉn=Dˉ(0,1)n⊂Cn. We see it as a pro-objet of AnSm(C) ([3]).
For F∙∈PSh(AnSm(C),Ab) and X∈AnSm(C)),
we have the complex F(X×Dˉ∗) associated to the cubical object
F(X×Dˉ∗) in the category of abelian groups.
•
If F∙∈PSh(AnSm(C),C−(Z)),
[TABLE]
is the total complex of presheaves associated to the bicomplex of presheaves X↦F∙(Dˉ∗×X),
and
singDˉF∙:=ean∗singDˉ∗F∙=F∙(Dˉ∗)∈C−(Z).
We denote by
S(F∙):F∙→singDˉ∗F∙,
[TABLE]
the inclusion morphism in PSh(AnSm(C),C−(Z)).
For f:F1∙→F2∙ a morphism PSh(AnSm(C),C−(Z)), we denote by
S(f):singDˉ∗F1∙→singDˉ∗F2∙
the morphism of P−(An) given by, for X∈AnSm(C),
[TABLE]
•
If F∙∈PShZ(CorZfs(AnSm(C)),C−(Z)),
[TABLE]
is the total complex of presheaves associated to the bicomplex of presheaves X↦F∙(Dˉ∗×X),
and
singDˉF∙:=ean∗trsingDˉ∗F∙=F∙(Dˉ∗)∈C−(Z).
We have the inclusion morphism (22)
[TABLE]
which is a morphism in PSh(CorZfs(AnSm(C)),C−(Z))
denoted the same way S(F∙):F∙→singDˉ∗F∙.
For f:F1∙→F2∙ a morphism PC−(An), we
have the morphism (23)
[TABLE]
which is a morphism in PC−(An)
denoted the same way
S(f):singDˉ∗F1∙→singDˉ∗F2∙.
For F∙∈PSh(CorZfs(AnSm(C)),C−(Z)), we have by definition
Tr∗singDˉ∗F∙=singDˉ∗Tr∗F∙
and Tr∗S(F∙)=S(Tr∗F∙).
We now make the following definition
Definition 15**.**
Let X∈AnSp(C) and D⊂X an analytic subspace.
Denote by l:D↪X the locally closed embbeding.
We define
(i)
Z(X,D)=coker(Z(l))∈PSh(AnSm(C),C−(Z))*
to be the cokernel of the injective morphism Z(l):Ztr(D)↪Z(X).
By definition, we have the following exact sequence*
[TABLE]
in PSh(AnSm(C),C−(Z)).
(ii)
Ztr(X,D)=coker(Ztr(l))∈PShZ(CorZfs(AnSm(C)),C−(Z))*
to be the cokernel of the injective morphism Ztr(l):Ztr(D)↪Ztr(X).
By definition, we have the following exact sequence*
[TABLE]
in PShZ(CorZfs(SmVar(k)),C−(Z)).
By the exact sequences, we have Tr∗Z(X,D)=Ztr(X,D)
In particular, we get the following exact sequence in PShZ(CorZfs(AnSm(C)),C−(Z))
[TABLE]
We define
[TABLE]
to be the relative motive of the pair (Y,E).
We now look at the behavior of the functors mentionned above with respect to the (D1,et) model structure
(Tr∗,Tr∗):PSh(AnSm(k),C−(Z))⇆PShZ(CorZfs(AnSm(C)),C−(Z))*
is a Quillen adjonction for the usual topology model structures (c.f. definition 12 (i) and (ii) respectively)
and a Quillen adjonction for the (D1,usu) model structures (c.f. definition 13 (i) and (ii) respectively).*
(ii)
(ean∗,ean∗):PSh(AnSm(C),C−(Z))⇆C−(Z)*
is a Quillen adjonction for the usual topology model structure (c.f. definition 12 (i))
and a Quillen adjonction for the (D1,usu) model structure (c.f. definition 13 (i)).*
(iii)
(eantr∗,ean∗tr):PSh(CorZfs(AnSm(C)),C−(Z))⇆C−(Z)*
is a Quillen adjonction for the usual topology model structure (c.f. definition 12 (ii))
and a Quillen adjonction for the (D1,usu) model structure (c.f. definition 13 (ii)).*
Lemma 2**.**
(i)
A complex of presheaves
F∙∈PSh(CorZfs(AnSm(C)),C−(Z)) is
D1 local if and only if
Tr∗F∙∈PSh(AnSm(C),C−(Z)) is
D1 local.
(ii)
A morphism
ϕ:F∙→G∙ in PSh(CorZfs(AnSm(C)),C−(Z))
is an (D1,usu) local equivalence if and only if
Tr∗ϕ:Tr∗F∙→Tr∗G∙ is
is an (D1,usu) local equivalence.
Proof.
(i): Let g:F∙→L∙
be an usu local equivalence in PSh(CorZfs(AnSm(C)),C−(Z))
with L∙ usu fibrant. Then,
•
F∙ is D1 local if and only if L∙ is D1 local.
•
By definition Tr∗ preserve usu local equivalence, hence
Tr∗g:Tr∗F∙→Tr∗L∙ is an usu local equivalence in
PSh(AnSm(C),C−(Z)).
Thus, Tr∗F∙ is D1 local if and only if Tr∗L∙ is D1 local.
•
As indicated in [3], Tr∗L∙ is also usu fibrant.
By Yoneda lemma, for X∈AnSm(C), we have
[TABLE]
Hence L∙ is D1 local if and only if Tr∗L∙ is D1 local.
Let ϕ:F∙→G∙ in PSh(CorZfs(AnSm(C)),C−(Z)) such that
Tr∗ϕ:Tr∗F∙→Tr∗G∙ is an equivalence (D1,usu) local
in PSh(AnSm(C),C−(Z)). Since by definition Tr∗ detecte and preserve usu local equivalence,
we can, up to replace F∙ and G∙ by usu equivalent presheaves,
assume that F∙ and G∙ are usu fibrant.
Let K∙∈PSh(CorZfs(AnSm(C)),C−(Z)) be an D1 local object.
By (i), Tr∗K∙ is D1 local.
Hence, we have,
[TABLE]
This proves the if part.
∎
Proposition 12**.**
(i)
The functor
Tr∗:PShZ(CorZfs(AnSm(C)),C−(Z))→PSh(AnSm(C)),C−(Z))
derive trivially.
(ii)
For K∙∈C−(Z), ean∗K∙ is D1 local.
(iii)
For K∙∈C−(Z), eantr∗K∙ is D1 local.
Proof.
(i): By lemma 2 (ii), Tr∗ preserve (D1,usu) local equivalence.
(iii): We have Tr∗eantr∗K∙=ean∗K∙.
By (ii), ean∗K∙ is D1 local. By lemma 2 (i), Tr∗ detect D1 local object.
This proves (iii).
∎
Theorem 13**.**
(i)
For F∙∈PSh(AnSm(C),C−(Z)),
singDˉ∗F∙∈PSh(AnSm(C),C−(Z)) is D1 local and
the inclusion morphism S(F∙):F∙→singDˉ∗F∙ is an (D1,usu) equivalence.
(ii)
For F∙∈PShZ(CorZfs(AnSm(C)),C−(Z)),
singDˉ∗F∙∈PShZ(CorZfs(AnSm(C)),C−(Z))
is D1 local and
the inclusion morphism S(F∙):F∙→singDˉ∗F∙ is an (D1,usu) equivalence.
(ii):
By (i), Tr∗singDˉ∗F∙=singDˉ∗Tr∗F∙
is D1 local.
By lemma 2 (i), Tr∗ detect D1 local object.
Thus,
singDˉ∗F∙=singDˉ∗F∙
is D1 local.
This proves the first part of the assertion.
It follows from (i) that
[TABLE]
is an (D1,usu) equivalence.
Since, by lemma 2 (ii), Tr∗ detect (D1,usu) equivalence,
S(F∙):F∙→singDˉ∗F∙
is an (D1,usu) equivalence.
The adjonction
(Tr∗,Tr∗):PSh(AnSm(C),C−(Z))⇆PShZ(CorZfs(AnSm(C)),C−(Z))
is a Quillen equivalence for the (D1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Tr∗:AnDM−(Z)∼AnDA−(Z) is it inverse.
(ii)
The adjonction
(ean∗,ean∗):C−(Z)⇆PShZ(AnSm(C),C−(Z))
is a Quillen equivalence for the (I1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Rean∗:AnDA−(Z)∼D−(Z) is it inverse.
(iii)
The adjonction
(eantr∗,ean∗tr):C−(Z)⇆PShZ(CorZfs(AnSm(C)),C−(Z))
is a Quillen equivalence for the (D1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Rean∗tr:AnDM−(Z)∼D−(Z) is it inverse.
Definition 16**.**
(i)
We say that a morphism ϵ:X′→X with X,X′∈AnSp(C) is
a proper modification (or abstract blow up) if it is proper and if there exist
Z⊂X a closed analytic subspace (the discriminant) such that ϵ
induces an isomorphism ϵX\Z:X′\E∼X\Z
with E=ϵ−1(Z).
(ii)
The cdh topology on AnSp(C) is the minimal Grothendieck topology generated by
open covering for the usual topology and covers c=ϵ⊔l:X′⊔Z→X corresponding
to proper modifications ϵ:X′→X, with l:Z↪X the closed embedding.
(iii)
We say that a morphism ϵ:X′→X in AnSp(C) is
a resolution of X if it is a proper modification and if X′ is smooth.
We recall a version of Hironaka’s desingulariation theorem
Theorem 15**.**
[9]**
For all X∈AnSp(C), there exist a proper modification ϵ:X′→X
with discriminant Z⊂X such that X′ is smooth and ϵ−1(Z)⊂X′ is a normal crossing divisor.
We denote by ιan:AnSm(C)↪AnSp(C) the full embedding functor,
which gives the morphism of sites
ιan:AnSp(C)↪AnSm(C).
If F∈PSh(AnSm(C),Ab) is an usu sheaf, ιan∗F is a cdh sheaf.
If F∈PSh(AnSp(C),Ab) is an usu sheaf, we have ιan∗ιan∗F=acdhF.
Corollary 2**.**
Let F∈PSh(AnSm(C),Ab) be an usu sheaf, then ιan∗F=0
if and only for all X∈AnSm(C) and all α∈F(X), there exist
a composition of blowups along smooth centers e:Xr→⋯→X such that
F(e)(α)=0∈F(Xr).
Let F∈PSh(AnSm(C),Ab).
Let where F→I∙ is an usu local equivalence in P−(An)
with I∙ a complex of injective objects (that is I∙ is projectively fibrant).
(i)
We say that F is homotopy invariant
if F(pX):F(X)→F(X×D1) is an isomorphism,
where pX:X×D1→X is the projection.
(ii)
We say that F is strictly homotopy invariant
if the presheaf Hn(I∙):X∈AnSm(C)↦Hn(X,F),
is homotopy invariant for all n∈N.
Proposition 13**.**
Let F∈PSh(AnSm(C),Ab).
Then F is homotopy invariant if and only if it is strictly homotopy invariant.
Lemma 3**.**
(i)
Let e:X′→X be the blow up of a smooth X∈AnSm(C) along a smooth center
Z⊂X. Denote by l:Z↪X the closed embedding.
Let C (resp. Q) the cokernel of Ztr(e):Ztr(X′)→Ztr(X)
(resp. the cokernel of
Ztr(l)⊕Ztr(e):Ztr(Z)⊕Ztr(X′)→Ztr(X)).
Then for any homotopy invariant sheaf F∈PSh(Cor(AnSm(C)),Ab),
Extn(C,F)=Extn(Q,F)=0 for all n∈N.
(ii)
Let F∈PSh(Cor(AnSm(C)),Ab) be an usu sheaf such that ιan∗F=0.
Then ausuHnsingD∗F=0 for all n∈Z, n≤0.
Proof.
(i): As in the proof of [15, proposition 13.19],
it follows from the fact that locally for the usual topology, since X and Z are smooth, X′ is the blow up of
Z×AdX−dS along Z×0.
(ii): As in the proof of [15, theorem 13.25],
(ii) follows from (i), corollary 2, and the fact that HnsingD∗ is homotopy
invariant, hence strictly homotopy invariant by proposition 13.
∎
Proposition 14**.**
Let ϵ:X′→X be a proper modification, X,X′∈AnSp(C).
Then, denoting
ϵZ∗+l∗′=S(Ztr(ϵZ)⊕Ztr(l′)) and
ϵ∗+l∗=S(Ztr(ϵ)⊕Ztr(l)),
the morphism in PC−(An)
[TABLE]
is an usu local equivalence.
Proof.
Denote by K∈PSh(CorZ(AnSm(C)),Ab), the cokernel of
Ztr(l)⊕Ztr(ϵ):Ztr(Z)⊕Ztr(X′)→Ztr(X).
The sequence
[TABLE]
is clearly exact in PC−(An),
In particular, the sequence
[TABLE]
is exact in PC−(An).
Hence we have to show that
ausuHnsingDˉ∗K=HnsingDˉ∗ausuK=0
for all n∈Z, n≤0, ausu being an exact functor..
By lemma 3(ii), it suffice to show that ιan∗ausuK=0.
But for all U∈AnSm(C), α∈Ztr(X)(U) and
α′∈ZdX′(U×X′) the proper transform of α by IU×ϵ,
there exist, by platification and
resolution of singularities, a proper modification e:U′′→U, with U′′∈AnSm(C) such that
the proper transform α′′∈ZdX′(U′′×X′) of α′ by e×IX′ is finite
(and surjective) over U′′, so that
Ztr(X)(e)(α):=α∘e=Ztrϵ(X′)(α′′)∈Ztr(X)(U′′)
with α′′∈Ztr(X′)(U′′). It then follows from corollary 2.
∎
We will use in the next section the following :
Proposition 15**.**
(i)
Let G1∙,G2∙∈PShZ(CorZfs(AnSm(C)),C−(Z))
and f:G1∙→G2∙ a morphism.
If
[TABLE]
is a quasi-isomorphism in C−(Z),
then f:G1∙→G2∙ is an (D1,usu) local equivalence.
(ii)
Let G1∙,G2∙∈PShZ(CorZfs(AnSm(C)),C−(Z))
and f:G1∙→G2∙ a morphism.
If
–
G1∙* and G2∙ are D1 local and*
–
ean∗trf:ean∗trG1∙→ean∗trG2∙*
is a quasi-isomorphism in C−(Z),*
then f:G1∙→G2∙ is an (D1,usu) local equivalence.
is an (D1,usu) local equivalence (even a quasi-isomorphism),
since ean∗trS(f):singD∗ˉG1∙→singD∗ˉG2∙
is a quasi-isomorphism in C−(Z) by hypothesis.
Hence, by the second commutative diagram,
[TABLE]
is an (D1,usu) local equivalence.
Since S(G1∙) and S(G2∙) are (D1,usu) local equivalence by theorem 13 (ii),
and S(f) is a (D1,usu) local equivalence, f:G1∙→G2∙
is a (D1,usu) local equivalence by the first commutative diagram.
(ii): Consider again the commutative diagram
[TABLE]
Since S(G1∙) and is a (D1,usu) local equivalence and G1∙ is D1 local,
S(G1∙) is an usu local equivalence.
Hence ean∗trS(G1∙) is a quasi isomorphism.
Since S(G2∙) and is a (D1,usu) local equivalence and G2∙ is D1 local,
S(G2∙) is an usu local equivalence.
Hence ean∗trS(G2∙) is a quasi isomorphism.
On the other side, by hypothesis, ean∗trf is a quasi isomorphism.
Hence, by the above diagram,
ean∗trS(f):singD∗ˉG1∙→singD∗ˉG2∙
is a quasi-isomorphism. It follows then by (i) that f:G1→G2 is a (D1,usu) local equivalence.
∎
2.3 Presheaves and transfers on the category of CW complexes
Let X,Y∈Top. A continous map f:X→Y is said to be proper if it is universally closed, that is for all T∈Top,
f×IT:X×T→Y×T is closed.
A continous map f:X→Y is proper if and only if it is closed and,
for all y∈Y, Xy=f−1(y) is quasi-compact (c.f.[5]).
A continous map f:X→Y is said to be finite if it is proper and,
for all y∈Y, Xy=f−1(y) is a finite set.
A continous map f:X→Y is said to be dominant if f(X)⊂Y has non empty interior.
For X∈Top, there exist a one point compactification Xˉ∈Top of X, that is
Xˉ is quasi-compact and there is an open embedding X↪Xˉ.
Moreover Xˉ is compact (i.e. quasi-compact and Hausdorf) if and only if
X is locally compact and Hausdorf.
Let X,Y,Z∈Top and denote pX:X×Y→X and pY:Y×Z→Z,
pXZ:X×Y×Z→X×Z the projections.
Assume that Y is Hausdorf (equivalently the diagonal ΔY⊂Y×Y is a closed subset).
Let Γ1⊂X×Y, Γ2⊂Y×Z, closed subsets.
Hence, since Y is Hausdorf, Γ1×YΓ2⊂X×Y×Z is a closed subset.
Now assume that pX∣Γ1:Γ1→X and pY∣Γ2:Γ2→Y are finite and surjective.
Consider the commutative diagram
[TABLE]
The projection pΓ1:Γ1×YΓ2→Γ1 is finite and surjective
(since pY∣Γ2:Γ2→Y is finite and surjective by hypothesis and Y is Hausdorf).
The projection pX∣Γ1:Γ1→X is finite and surjective by hypothesis.
Hence, by the diagram (32), the composition
[TABLE]
is finite and surjective. Thus,
pX∣pXZ(Γ1×YΓ2):pXZ(Γ1×YΓ2)→X
is finite and surjective, that is
[TABLE]
is a closed subset of X×Z which is finite and surjective on X.
We consider now the full subcategory CW⊂Top consisting of CW complexes.
Recall that CW complexes are Hausdorf, locally contractible and locally compact.
By a CW subcomplex of X∈CW, we mean a topological embbeding Z↪X with Z a CW complex.
By a closed CW subcomplex of X∈CW, we mean a topological closed embbeding Z↪X with Z a CW complex,
that is the image of the embedding is a closed subset of X.
Let X,S∈CW, S connected, and h:X→S a finite and surjective morphism in CW.
We say that X/S=(X,h)∈CW/S is reducible if X=X1∪X2, with X1,X2 closed CW subcomplexes
finite and surjective over S and X1,X2=X.
A pair Y/S=(Y,h′)∈CW/S with Y∈CW and h′:Y→S a finite and surjective morphism
is called irreducible if it is not reducible. In particular Y is connected.
•
For X∈CW, Λ a commutative ring and p∈N,
we denote by Zp(X,Λ) the free Λ module generated
by the closed CW subcomplex of X of dimension p and by
Zp(X,Λ)=ZdX−p(X,Λ)
the free Λ module generated by the closed CW subcomplex of X of codimension p.
•
For X,Y∈CW, X connected, and Λ a commutative ring, we define :
Zfs/X(X×Y,Λ)⊂ZdX(X×Y,Λ)
the free Λ module generated by the closed CW subcomplexes of X×Y finite and surjective over X
which are irreducible.
Note that if Z⊂X×Y is a closed CW subcomplex finite and surjective over X which is irreducible,
then if Z=Z1∪Z2 with Z1,Z2⊂Z closed CW subcomplex finite and surjective over X,
then Z1=Z2=Z.
•
For X,Y∈CW, and Λ a commutative ring, we define :
Zfs/X(X×Y,Λ):=⊕iZfs/Xi(Xi×Y,Λ)
where X=⊔iXi, with Xi the connected components of X.
For Y∈CW and j:V↪Y an open embedding,
we denote by
[TABLE]
Let X,Y,Z∈CW and denote pX:X×Y→X and pY:Y×Z→Z,
pXZ:X×Y×Z→X×Z the projections.
Let Γ1⊂X×Y be a closed CW subcomplex finite and surjective over X which is irreducible.
Let Γ2⊂Y×Z be a closed CW subcomplex finite and surjective over Y which is irreducible.
Hence, since Y is Hausdorf,
Γ1×YΓ2⊂X×Y×Z is a closed CW subcomplex finite and surjective over X,
in particular all its irreducible components are finite and surjective over X.
Thus, by diagram 32
[TABLE]
is a closed CW subcomplex of X×Z finite and surjective over X,
in particular all its irreducible components are finite and surjective over X and
so define an element of Zfs/X(X×Z) .
Definition 18**.**
We define CorΛfs(CW) to be the category whose objects are CW complexes and whose space of morphisms between
X,Y∈CW is the free Λ module Zfs/X(X×Y,Λ).
The composition is the one given by (35).
We have
•
the additive embedding of categories Tr:Z(CW)↪CorZfs(CW)
which gives the corresponding morphism of sites Tr:CorZfs(CW)→Z(CW),
•
the inclusion functor
ecw:{pt}↪CW,
which gives the corresponding morphism of sites
ecw:CW→{pt},
•
the inclusion functor ecwtr:=Tr∘ecw:{pt}↪CorZfs(CW)
which gives the corresponding morphism of sites
ecwtr:=Tr∘ecw:CorZfs(CW)→{pt}.
We consider the following two big categories :
•
PSh(CW,C−(Z))=PShZ(Z(CW),C−(Z))),
the category of bounded above complexes of presheaves on CW,
or equivalently additive presheaves on Z(CW), sometimes, we will write for short
P−(CW)=PSh(CW,C−(Z)),
•
PSh(CorZfs(CW),C−(Z)),
the category of bounded above complexes of additive presheaves on CorZfs(CW),
sometimes, we will write for short
PC−(CW)=PSh(CorZfs(CW),C−(Z)),
and the adjonctions :
•
(Tr∗,Tr∗):PSh(CW,C−(Z))⇆PSh(CorZfs(CW),C−(Z)),
•
(ecw∗,ecw∗):PSh(CW,C−(Z))⇆C−(Z),
•
(ecwtr∗,ecw∗tr):PSh(CorZfs(CW),C−(Z))⇆C−(Z),
given by Tr:CorΛfs(CW)→Λ(CW),
ecwCW→{pt} and ecwtr:CorZfs(CW)→{pt}.
For X∈CW, we denote by
[TABLE]
the presheaves represented by X. They are usu sheaves.
We denote by
ausu:PShZ(CW,Ab)→ShZ,usu(CW,Ab)
the usu sheaftification functor.
We consider
•
the usual monoidal strutcure on PSh(CW,C−(Z)) and the associated internal Hom
given by, for F∙,G∙∈PSh(CW,C−(Z)) and Y∈CW,
[TABLE]
•
the unique monoidal strutcure on PShZ(CorZfs(CW),C−(Z))
such that,for X,Y∈CW, Ztr(X)⊗Ztr(Y):=Ztr(X×Y)
and wich commute with colimites.
It has an internal Hom which is given, for X,Y∈CW and
F∙∈PShZ(CorZfs(CW),C−(Z)),
Hom(Ztr(Y),F∙:X↦F∙(X×Y)
Together with these monoidal structure, the functor
[TABLE]
is monoidal.
Definition 19**.**
(i)
We say that a morphism ϕ:G1∙→G2∙ in PSh(CW,C−(Z)) is an
usu local equivalence if ϕ∗:ausuHk(G1∙)→ausuHk(G2∙) is an isomorphism for all k∈Z.
The projective usual topology model structure on PSh(CW,C−(Z))
is the left Bousfield localization of the projective model structure MP(PShZ(CW,C−(Z)))
with respect to the usu local equivalence.
(ii)
We say that a morphism ϕ:G1∙→G2∙ in PShZ(CorZfs(CW),C−(Z)) is an
usu local equivalence if and only if its restriction to CWTr∗ϕ:Tr∗G1∙→Tr∗G2∙ is an usu local equivalence.
The projective usual topology model structure on PShZ(CorZfs(CW),C−(Z))
is the left Bousfield localization of the projective model structure
MP(PShZ(CorZfs(CW),C−(Z)))
with respect to the usu local equivalence.
Definition 20**.**
(i)
The projective (I1,usu) model structure on PSh(CW,C−(Z)) is the left Bousfield localization
of the projective usual topology model structure (c.f.19(i)) with respect to the class of maps
{Z(X×I1)[n]→Z(X)[n],X∈CW,n∈Z}.
(ii)
The projective (I1,usu) model structure on PShZ(CorZfs(CW),C−(Z)) is the left Bousfield localization of the projective usual topology model structure (c.f.19(ii)) with respect to the class of maps
{Ztr(X×I1)→Z(X),X∈CW,n∈Z}.
Definition 21**.**
(i)
We define CwDM−(Z):=HoI1,usu(PShZ(CorZfs(AnSm(C)),C−(Z))),
the derived category of motives of CW complexes, it is
the homotopy category of PShZ(CorZfs(CW),C−(Z))
with respect to the projective (I1,usu) model structure (20(ii)).
We denote by
[TABLE]
the canonical localisation functor.
(ii)
We denote by the same way
CwDA−(Z):=HoI1,usu(PSh(CW,C−(Z))) (20(i)) and
[TABLE]
the canonical localisation functor.
We recall for convenience to the reader the definition of I1 homotopy :
•
Let X,Y∈Top. We say that two maps f0:X→Y, f1:X→Y are I1 homotopic,
if there exist h:X×I1→Y such that f0=h∘(IX×i0) and f1=h∘(IX×i1), with
(IX×i0):X×{0}↪X×I1
and (IX×i1):X×{1}↪X×I1
the inclusions.
•
Let X,Y∈Top. We say that X is I1 homotopy equivalent to Y if there
exist two maps f:X→Y, g:Y→X such that
g∘f is I1 homotopic to IX and f∘g is I1 homotopic to IY.
•
Let F∙,G∙∈PSh(CW,C−(Z)).
We say that two maps ϕ0:F∙→G∙ and ϕ1:F∙→G∙
are I1 homotopic if there exist
ϕ~:F∙→Hom(Z(I1),G∙) such that
ϕ0=G∙(i0)∘ϕ~ and ϕ1=G∙(i1)∘ϕ~, where,
–
G∙(i0):Hom(Z(I1),G∙)→G∙ is
induced by i0:{pt}→I1, that is, for X∈CW,
G∙(i0)(X)=G∙(IX×i0):G∙(X×I1)→G∙(X),
–
G∙(i1):Hom(Z(I1),G∙)→G∙ is
induced by i1:{pt}→I1, that is, for X∈CW,
G∙(i1)(X)=G∙(IX×i1):G∙(X×I1)→G∙(X).
•
Let F∙,G∙∈PSh(CW,C−(Z)).
We say that F∙ is I1 homotopy equivalent to G∙ if there
exist two maps ϕ:F∙→G∙ and ψ:F∙→G∙ such that
ψ∘ϕ is I1 homotopic to IF∙
and ϕ∘ψ is I1 homotopic to IG∙.
We have the following easy lemma
Lemma 4**.**
Let X,Y∈CW and f0:X→Y, f1:X→Y two morphisms.
If f0 and f1 are I1 homotopic, then
•
Z(f0):Z(X)→Z(Y)* and Z(f1):Z(X)→Z(Y)
are I1 homotopic in PSh(CW,C−(Z)),*
•
Tr∗Ztr(f0):Tr∗Ztr(X)→Tr∗Ztr(Y)*
and Tr∗Ztr(f1):Tr∗Ztr(X)→Tr∗Ztr(Y)
are I1 homotopic in PSh(CW,C−(Z)).*
Proof.
Let h:X×I1→Y such that f0=h∘(IX×i0) and f1=h∘(IX×i1), with
(IX×i0):X×{0}↪X×I1
and (IX×i1):X×{1}↪X×I1
the inclusions.
Then
•
ϕ(h):Z(X)→Hom(Z(I1),Z(Y)), given by for Z∈CW
[TABLE]
satisfy
Z(Y)(i0)∘ϕ(h)=Z(f0) and Z(Y)(i1)∘ϕ(h)=Z(f1).
•
ϕ(h):Ztr(X)→Hom(Z(I1),Ztr(Y)), given by for Z∈CW
[TABLE]
satisfy
Ztr(Y)(i0)∘ϕ(h)=Ztr(f0) and
Ztr(Y)(i1)∘ϕ(h)=Ztr(f1).
∎
Recall Δ denote the simplicial category
Let
•
pΔ:CW→Δ×CW
be the morphism of site given by the projection functor
(X,i)∈Δ×CW↦pΔ((X,i))=X∈CW,
•
pΔ:CorZ(CW)→Δ×CorZ(CW)
be the morphism of site given by the projection functor
(X,i)∈Δ×CW↦pΔ((X,i))=X∈CW
In the following lemma point (i) is a generalization of point (i) to usu hypercover.
We will only use point (i) in this paper in the proof of proposition 21(ii).
Lemma 5**.**
(i)
: Let X∈CW and X=∪i∈JUi a covering by open subsets Ui⊂X, J being a countable set.
We denote by ji:Ui↪X the open embedding.
For I⊂J a finite subset, let UI=∩i∈IUi.
Then,
–
[⋯→⊕cardI=rZ(UI)→⋯→⊕i∈JZ(Ui)]⊕i∈JZ(ji)Z(X)*
is an usu local equivalence in P−(CW).*
–
[⋯→⊕cardI=rZtr(UI)→⋯→⊕i∈JZtr(Ui)]⊕i∈JZtr(ji)Ztr(X)*
is an usu local equivalence in PC−(CW).*
(ii)
Let X∈CW and j∙:U∙→X an hypercovering of X by open subsets
(i.e. for all n∈Δ, Un is an open subset and jn:Un↪X is the open embedding),
and (U∙,∙):⋯→(Un,n)→⋯ the associated complex in Δ×CW
Then,
–
j:LpΔ∗(Z(U∙,∙))→Z(X)* is an isomorphism
in Housu(PSh(CW,C−(Z)))*
–
j:LpΔ∗(Ztr(U∙,∙))→Ztr(X)* is an isomorphism
in Housu(PShZ(CorZ(CW),C−(Z)))*
Proof.
(i): The first point follows from the following two facts :
•
[⋯→⊕cardI=rZ(UI)→⋯→⊕i∈JZ(Ui)]
is clearly exact in P−(CW).
•
the exactness of
[⊕i∈JZ(Ui)⊕i∈JZ(ji)Z(X)→0]
in Housu(PSh(CW,C−(Z))),
follows from the fact that for Y∈CW, y∈Y and f:Y→X a morphism,
there exists an open subset V(y)⊂Y containing y such that f(V(y))⊂Ui, with
Ui containing f(y).
Similary, the second point follows from the following two facts :
•
[⋯→⊕cardI=rZtr(UI)→⋯→⊕i∈JZtr(Ui)]
is clearly exact in PShZ(CorZfs(CW),C−(Z))).
•
the exactness of
[⊕i∈JZtr(Ui)⊕i∈JZtr(ji)Ztr(X)→0]
in Housu(PShZ(CorZfs(CW),C−(Z))),
follows from the fact that for Y∈CW, y∈Y and α∈Ztr(X)(Y) irreducible,
there exists an open subset V(y)⊂Y containing y such that α∣V(y)∈Ztr(Ui)(V(y)), with
Ui containing α∗y.
The following proposition is to use point (i) instead of point (ii) in the proof
of proposition 21(ii). Recall CS⊂CW is the full subcategory of Δ complexes.
Proposition 16**.**
(i)
Let X∈CW. There exist X′∈CS homotopy equivalent to X,
that is there exist g:X′→X and h:X→X′ such that h∘g is homotopic to IX′
and g∘h is homotopic to IX.
(ii)
Let X∈CS. There exist a countable open covering X=∪i∈JUi such that
for all finite subset I⊂J, UI:=∩i∈IUi=∅ or UI is contractible.
(ii): Take an open star of each vertices of X. As X is a countable union of compact set,
we can extract a countable subcovering of this open covering.
∎
We denote, for all n∈N, In:=[0,1]n∈CW.
It gives, together with the face maps
∂ϵ,in:In−1↪In, ϵ=0,1, i=1,⋯n
a cubical object of CW, denoted I∗.
In particular, for n=1, we have the canonical maps
i0:{0}↪I1i1:{1}↪I1,
and the terminal map a=aI1:I1→{pt}.
For F∙∈PSh(CW,Ab) and X∈CW,
we have the complex F(X×I∗) associated to the cubical object
F(X×I∗) in the category of abelian groups, whose differential maps are
[TABLE]
There is ([14]) a canonical morphism L:I∗→Δ∗ of complexes of Z(CW).
This gives for F∈PSh(CW,Ab) and X∈CW, the morphism of complexes of abelian groups
[TABLE]
with L×IX:I∗×X→Δ∗×X
the corresponding morphism of complexes of Z(CW).
We have then following :
Proposition 17**.**
[14]**
For X∈CW, Z(X)(L):ZHom(Δ∗,X)→singI∗Z(X)
is a quasi-isomorphism of complexes of abelian groups.
We now introduce an explicit localization functor for the (I1,usu) model structure.
•
If F∙∈PSh(CW,C−(Z)) ,
[TABLE]
is the total complex of presheaves associated to the bicomplex of presheaves X↦F∙(I∗×X),
and singI∗F∙:=ecw∗singI∗F∙=F∙(I∗)∈C−(Z).
We denote by
S(F∙):F∙→singI∗F∙
[TABLE]
the inclusion morphism in PSh(CW),C−(Z)).
For f:F1∙→F2∙ a morphism of PSh(CW,C−(Z)), we denote by
S(f):singI∗F1∙→singI∗F2∙
the morphism of P−(CW) given by, for X∈CW,
[TABLE]
•
If F∙∈PShZ(CorZfs(CW),C−(Z)),
[TABLE]
is the total complex of presheaves associated to the bicomplex of presheaves X↦F∙(I∗×X),
and singI∗F∙:=ecw∗trsingI∗F∙=F∙(I∗)∈C−(Z).
We have the inclusion morphism (40)
[TABLE]
which is a morphism in PSh(CorZfs(CW),C−(Z))
denoted the same way S(F∙):F∙→singI∗F∙.
For f:F1∙→F2∙ a morphism PC−(CW), we
have the morphism (41)
[TABLE]
which is a morphism in PC−(CW) denoted the same way
S(f):singI∗F1∙→singI∗F2∙.
For F∙∈PSh(CorZfs(CW),C−(Z)), we have by definition
Tr∗singI∗F∙=singI∗Tr∗F∙
and Tr∗S(F∙)=S(Tr∗F∙).
We now make the following definition
Definition 22**.**
Let Y∈CW and E⊂Y a CW subcomplex. Denote by l:E↪X the topological embedding.
We define
•
Z(Y,E)=coker(Z(l))∈PShZ(CW,C−(Z))*
to be the cokernel of the injective morphism Z(l):Z(E)↪Z(X).*
•
Ztr(Y,E)=coker(Ztr(l))∈PShZ(CorZfs(CW),C−(Z))*
to be the cokernel of the injective morphism Ztr(l):Ztr(E)↪Ztr(Y),
by definition, we have the following exact sequence in PC−(CW)*
[TABLE]
•
Z(Y,E)=coker(ad(l!,l!)(ZX))∈Sh(Y)*
the cokernel of the injective morphism
ad(l!,l!)(ZY):l!l!ZY↪ZY in Sh(Y)
By definition, there is a distingushed triangle in Db(Y)*
[TABLE]
The relative Borel Moore homology of the pair (Y,E)
is HpBM(Y,E,Z)=Hom(Z,aX!aX!Z(Y,E)).
In particular, we get from the second point the following exact sequence in
PShZ(CorZfs(CW),C−(Z))
[TABLE]
We define
[TABLE]
to be the relative motive of the pair (Y,E).
We now look at the behavior of the functors mentionned above with respect to the (I,usu) model structures.
We start we a lemma concerning the properties of I1 homotopy maps.
Lemma 6**.**
Let F∙∈PSh(CW,C−(Z)).
(i)
Let X,Y∈CW and f0:X→Y, f1:X→Y be two maps.
If f0 and f1 are I1 homotopic, then the maps of complexes
–
singI∗F∙(f0):TotF∙(Y×I∗)→TotF∙(X×I∗)* and*
–
singI∗F∙(f1):TotF∙(Y×I∗)→TotF∙(X×I∗)**
induces the same map on homology.
(ii)
Let X,Y∈CW, if f:X→Y is a I1 homotopy equivalence then
[TABLE]
is a quasi-isomorphism of complexes of abelian groups.
(iii)
Let F∙,G∙∈PSh(CW,C−(Z)) and
ϕ0:F∙→G∙, ϕ1:F∙→G∙ be two maps.
If ϕ0 and ϕ1 are I1 homotopic, then ϕ0=ϕ1∈CwDA−.
(iv)
Let F∙,G∙∈PSh(CW,C−(Z)), if
ϕ:F∙→G∙ is a I1 homotopy equivalence then ϕ is a (I1,usu) local equivalence.
Proof.
(i): Let h:X×I1→Y an homotopy from f0=h∘(IX×i0) to f1=h∘(IX×i1).
Let α∈Fk(Y×In). Then, we have
[TABLE]
This proves (i).
(ii): Follow imediately from (i).
(iii): Let ϕ~:F∙→Hom(Z(I1),G∙) an homotopy
from ϕ0=G∙(i0)∘ϕ~ to ϕ1=G∙(i1)∘ϕ~.
Let
[TABLE]
be the map induced by aI1:I1→{pt}, that is, for X∈CW,
G∙(p)(X)=G∙(pX):G∙(X)→G∙(X×I1),
with pX=IX×aI1:X×I1→X is the projection.
Then, we have
[TABLE]
Hence, it suffice to show that
G∙(p):G∙→Hom(Z(I1),G∙)
is an (I1,usu) local equivalence, since
then G∙(i0)=G∙(i1)∘G∙(p)−1∈CwDA−.
So, consider ψ:G∙→L∙ a morphism in PSh(CW,C−(Z)) with L∙I1 local.
Consider the commutative diagramm in HousuPSh(CW,C−(Z))
[TABLE]
Since we
now consider morphism in the homotopy category with respect to the usu model structure,
we can assume, up to replace L∙ by an usu local equivalent object that L∙ is usu fibrant.
Then, L∙L∙(p)Hom(Z(I1),L∙)
is an usu local equivalence and
[TABLE]
is the only map making the diagram 46 commute. This prove that
G∙(p) is a (I1,usu) local equivalence.
(iv): Follows imediately from (iii).
∎
Lemma 7**.**
(i)
A complex of presheaves F∙∈PSh(CorZfs(CW),C−(Z)) is
I1 local if and only if
Tr∗F∙∈PSh(CW,C−(Z)) is
I1 local.
(ii)
A morphism ϕ:F∙→G∙ in PSh(CorZfs(CW),C−(Z)) is
an (I1,usu) local equivalence if and only if
Tr∗ϕ:Tr∗F∙→Tr∗G∙ is an (I1,usu) local equivalence.
Proof.
(i): Let h:F∙→L∙
be an usu local equivalence in PSh(CorZfs(CW),C−(Z))
with L∙ usu fibrant.
Then,
•
F∙ is I1 local if and only if L∙ is I1 local.
•
By definition Tr∗ preserve usu local equivalence, hence
Tr∗h:Tr∗F∙→Tr∗L∙ is an usu local equivalence in PSh(CW,C−(Z)).
Thus, Tr∗F∙ is I1 local if and only if Tr∗L∙ is I1 local.
•
As indiquated in [3],since , Tr∗L∙ is also usu fibrant.
By Yoneda lemma, for X∈CW, we have
[TABLE]
Hence L∙ is I1 local if and only if Tr∗L∙ is I1 local.
This proves (i).
(ii): Let us prove (ii)
•
The only if part is similar to [2], lemma 2.111, we check that, for Y∈CW,
[TABLE]
is an equivalence (I1,usu) local,
where pY:Y×I1→Y is the projection.
By lemma 6(iv), it suffice to show that Tr∗Ztr(pY)
is an I1 homotopy equivalence in P−(CW).
By lemma 4, it suffice to show that pY is an I1 homotopy equivalence in CW.
But we have the map
[TABLE]
which is an homotopy from IY×I1=θ12(Y)∘(IY×I1×i0) to
IY×0=θ12(Y)∘(IY×I1×i1).
On the other hand, pY∘(IY×0)=IY.
This proves the only if part.
•
Let ϕ:F∙→G∙ in PSh(CorZfs(CW),C−(Z)) such that
Tr∗ϕ:Tr∗F∙→Tr∗G∙ is an equivalence (I1,usu) local
in PSh(CW,C−(Z)). Since by definition Tr∗ preserve and detect usu local equivalence,
we can assume, up to replace F∙ and G∙ by usu equivalent presheaves,
that F∙ and G∙ are usu fibrant.
Let K∙∈PSh(CorZfs(CW),C−(Z)) an I1 local object.
By (i), Tr∗K∙ is I1 local.
Hence, we have,
[TABLE]
This proves the if part.
∎
Proposition 18**.**
(i)
(Tr∗,Tr∗):PSh(CW,C−(Z))⇆PShZ(CorZfs(CW),C−(Z))*
is a Quillen adjonction for the usu topology model structures (c.f. definition 19 (i) and (ii) respectively)
and a Quillen adjonction for the (I1,usu) model structures (c.f. definition 20 (i) and (ii) respectively).*
(ii)
(ecw∗,ecw∗):PSh(CW,C−(Z))⇆C−(Z).
is a Quillen adjonction for the usual topology model structure (c.f. definition 19 (i))
and a Quillen adjonction for the (I1,usu) model structure (c.f. definition 20 (i)).
(iii)
(ecwtr∗,ecw∗tr):PSh(CorZfs(CW),C−(Z))⇆C−(Z).
is a Quillen adjonction for the usual topology model structure (c.f. definition 19 (ii))
and a Quillen adjonction for the (I1,usu) model structure (c.f. definition 20 (ii)).
Proof.
(i): By lemma 7 (i), Tr∗ preserve I1 local objects, hence preserve the fibrations of
the (I1,usu) model structures.
By lemma 7 (ii), Tr∗ preserve (I1,usu) equivalence.
Thus, Tr∗ preserve the trivial fibrations of the (I1,usu) model structures.
(ii): It is clear that ecw∗ is a left Quillen functor, that is preserve cofibrations and trivial cofibrations.
(iii): It is clear that ecwtr∗ is a left Quillen functor, that is preserve cofibrations and trivial cofibrations.
∎
Proposition 19**.**
(i)
For F∙∈PSh(CW,C−(Z)), the adjonction morphism
[TABLE]
is an equivalence usu local.
(ii)
For F∙∈PShZ(CorZfs(CW),C−(Z)), the adjonction morphism
[TABLE]
is an equivalence usu local.
(iii)
For K∙∈C−(Z), the adjonction morphisms
[TABLE]
are isomorphisms.
Proof.
(i): The question is local. It then follows from the fact that the CW complexes are locally contractible.
Indeed, let X∈CW. Since the question is local and CW complexes are locally contractile,
we can assume after shrinking X that X is contractible.
Let x∈X. Denoting aX:X→{pt} the terminal map
and ex:{pt}→X the point map, there exist an homotopy between ex∘aX and the identity IX of X,
that is pX is a I1 homotopy equivalence.
Hence, by lemma 6 (ii),
[TABLE]
is an homotopy equivalence of complexes of abelian groups.
(ii): It is a particular case of point (i) since, by definition (19), Tr∗ detect and preserve usu local equivalence.
Indeed, by (i)
[TABLE]
is an equivalence usu local. Since Tr∗ preserve usu local equivalence, this prove (ii).
(iii): Trivial.
∎
Proposition 20**.**
(i)
The functor
Tr∗:PShZ(CorZfs(CW),C−(Z))→PSh(CW,C−(Z))
derive trivially.
(ii)
For K∙∈C−(Z), ecw∗K∙ is I1 local.
(iii)
For K∙∈C−(Z), ecwtr∗K∙ is I1 local.
Proof.
(i): By lemma 7, Tr∗ preserve (I1,usu) local equivalences.
(ii): Let ecw∗K∙→L∙ an usu local equivalence, with L∙ usu fibrant.
Since ecw∗K∙ is usu equivalent to L∙ it suffices to prove that L∙ is I1 local.
Since L∙ is usu fibrant, we have to prove that
[TABLE]
is an equivalence usu local
The proof is now similar to [1] proposition 1.6 etape B.
(iii): We have Tr∗ecwtr∗K∙=ecw∗K∙.
By (ii), ecw∗K∙ is I1 local. By lemma 7 (i), Tr∗ detect I1 local object.
This proves (iii).
∎
Theorem 16**.**
(i)
For F∙∈PSh(CW,C−(Z)),
singI∗F∙∈PSh(CW,C−(Z)) is I1 local and the inclusion morphism
S(F∙):F∙→singI∗F∙ is an (I1,usu) equivalence.
(ii)
For F∙∈PShZ(CorZfs(CW),C−(Z)),
singI∗F∙∈PShZ(CorZfs(CW),C−(Z))
is I1 local and the inclusion morphism
S(F∙):F∙→singI∗F∙ is an (I1,usu) equivalence.
Proof.
(i): Let us prove (i)
•
By proposition 19 (i) and proposition 20 (ii),
singI∗F∙∈PSh(CW,C−(Z)) is I1 local.
This proves the first part of the assertion.
•
Consider the commutative diagram
[TABLE]
By the diagram (57), to prove that
S(F∙):F∙→singI∗F∙
is an (I1,usu) equivalence, it suffices to show, as in [2], theorem 2.23 etape 2,
that for all n∈Z, the morphism
[TABLE]
is an equivalence (I1,usu) local.
For X∈CW, consider the map
[TABLE]
We have,
[TABLE]
that is θ1,n(X)
define an I1 homotopy from IIn×X to 0×IX ;
on the other side
pX∘(0×IX)=IX, with pX:In×X→X the projection.
Thus,
[TABLE]
that is
F∙(θ1,n)
define an I1 homotopy from F∙(IIn) to
F∙(0) ; on the other side,
F∙(0)∘F∙(pn)=I,
with
–
F∙(0):Hom(Z(In),F∙)→Hom(Z(In),F∙), given by
X∈CW↦F∙(0)(X)=F∙(0×IX):F∙(In×X)→F∙(In×X)
–
F∙(IX):Hom(Z(In),F∙)→Hom(Z(In),F∙), given by
X∈CW↦F∙(IX)(X)=F∙(IIn×X)
–
F∙(θ1,n):Hom(Z(I1),Hom(Z(In),F∙))→Hom(Z(In),F∙), given by
X∈CW↦F∙(θ1,n)(X)=F∙(θ1,n(X)):F∙(In×X)→F∙(In×I1×X).
Hence,
F∙(pIn):F∙→Hom(Z(In),F∙)
is an homotopy equivalence, hence an equivalence (I1,usu) local by lemma 6 (iv).
(ii): Let us prove (ii)
•
By (i),
Tr∗singI∗F∙=singI∗Tr∗F∙
is I1 local.
By lemma 7 (i), Tr∗ detect I1 local object.
Thus,
singI∗F∙=singI∗F∙
is I1 local.
This proves the first part of the assertion.
•
It follows from (i) that
[TABLE]
is an (I1,usu) equivalence.
Since, by lemma 7 (ii), Tr∗ detect (I1,usu) equivalence,
S(F∙):F∙→singI∗F∙
is an (I1,usu) equivalence.
∎
Proposition 21**.**
(i)
For F∙∈PSh(CorZ(CW),C−(Z)),
[TABLE]
is an isomorphism in PSh(CorZ(CW),C−(Z)).
(ii)
For X∈CW, the embedding
[TABLE]
in PSh(CW,C−(Z)) is an equivalence usu local
Proof.
(i): Obvious
(ii):
By proposition 16(i), there exist an homotopy equivalence g:X′→X, with X′∈CS a Δ complex.
Then,
•
S(Z(g)):singI∗Z(X′)→singI∗Z(X)
is an I1 homotopy equivalence by lemma 4,
thus an usu local equivalence by lemma 6(iv) and theorem 16(i),
•
Tr∗S(Ztr(g)):Tr∗singI∗Ztr(X′)→Tr∗singI∗Z(X)
is an I1 homotopy equivalence by lemma 4,
thus an usu local equivalence by lemma 6(iv) and theorem 16(ii)
and the fact that Tr∗ preserve I1 local object by lemma 7.
Consider the following commutative diagram in PSh(CW,C−(Z))
[TABLE]
By diagram (59) and the fact that S(Z(g)) and Tr∗S(Ztr(g)) are usu local
equivalence, it suffice to show that
[TABLE]
is an usu local equivalence.
By proposition 16(ii), there exist a countable open covering X′=∪i∈JUi such that for all
finite subset I⊂J, UI:=∩i∈IUi=∅ or UI is contractible.
Let (U∙,∙) be the complex in Δ×CW associated to this covering.
Let
•
h:Z(U∙,∙)→K∙ be an equivalence usu local
in PSh(Δ×CW,C−(Z)) with K∙ usu fibrant
•
l:Ztr(U∙,∙)→L∙ be an equivalence usu local
in PSh(Δ×CorZ(CW),C−(Z)) with L∙ usu fibrant
We have then a unique morphism m:K∙→Tr∗L∙
such that the following diagram in Housu(PSh(Δ×CW,C−(Z))) commutes
[TABLE]
We have then the following commutative diagram in Housu(PSh(CW,C−(Z))) :
[TABLE]
By lemma 5 (i), and the fact that Tr∗ preserve usu local equivalence,
are isomorphisms.
On the other hand, U[n] is contractible for all [n]∈Δ, this means that the canonical morphism
aU[n]:U[n]→{pt} in CW is an homotopy equivalence.
Hence,
•
Z(aU[n]):Z(U[n])→Z({pt}) is an
homotopy equivalence in PSh(CW,C−(Z)), thus an (I1,usu) local equivalence
by lemma 6(iv) ;
•
Tr∗Ztr(aU[n]):Tr∗Ztr(U[n])→Tr∗Ztr({pt}) is an
homotopy equivalence in PSh(CW,C−(Z)), thus an (I1,usu) local equivalence
by lemma 6(iv).
Thus,
[TABLE]
is an (I1,usu) local equivalence in PSh(Δ×CW,C−(Z)).
The diagram (60) then shows that m:K∙→Tr∗L∙
is an isomorphism in HoI1,usu(PSh(Δ×CW,C−(Z)).
Since pΔ∗ derive trivially by definition, this implies that
[TABLE]
is an isomorphism in CwDM(Z)−
The diagram (61), then shows that
ad(Tr∗,Tr∗)(Z(X′)):Z(X)→Tr∗Ztr(X′)
is an isomorphism in CwDM(Z)−.
∎
Theorem 17**.**
(i)
The adjonction
(Tr∗,Tr∗):PSh(CW,C−(Z))⇆PShZ(CorZfs(CW),C−(Z))
is a Quillen equivalence for the (I1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Tr∗:CwDM−(Z)∼CwDA−(Z) is it inverse.
(ii)
The adjonction
(ecw∗,ecw∗):C−(Z)⇆PShZ(CW,C−(Z))
is a Quillen equivalence for the (I1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Recw∗:CwDA−(Z)∼D−(Z) is it inverse.
(iii)
The adjonction
(ecwtr∗,ecw∗tr):C−(Z)⇆PShZ(CorZfs(CW),C−(Z))
is a Quillen equivalence for the (I1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Recw∗tr:CwDM−(Z)∼D−(Z) is it inverse.
Proof.
(i): It follows from proposition 21 and theorem 16.
(ii): It follows from proposition 19(i) and theorem 16(i).
(iii): It follows from proposition 19(ii) and theorem 16(ii).
It also follows from (i) and (ii).
∎
We deduce from proposition 21 (ii) and proposition 17 the following :
Proposition 22**.**
For Y∈CW and l:E↪Y a CW subcomplex, the followings embeddings are quasi-isomorphism :
[TABLE]
where C∗sing(Y,E,Z)=cokerl∗ is the relative cohomology, with
l∗:ZHomCW(Δ∗,E)↪ZHomCW(Δ∗,Y).
Proof.
Follows from proposition 21 (ii) and proposition 17.
∎
Let X,Y∈CW and E⊂Y a subcomplex, we have
•
the morphism
[TABLE]
given by the composition
Sn(X,(Y,E))(H)=H∘S(Ztr(X)) with the inclusion morphism S(Ztr(X)),
•
the morphism
[TABLE]
given by,
for [T]∈HomPC−(CW)(Ztr(X),singI∗Ztr(Y,E)[n]), and Z∈CW,
[TABLE]
with IIn×γ∈Zfs/In+p×Z((In×Ip×Z)×(In×X)).
Lemma 8**.**
Let X,Y∈CW and E⊂Y a subcomplex, we have
(i)
Sn(X,(Y,E))∘Rn(X,(Y,E))=I,
(ii)
for [T]∈HomPC−(CW)(Ztr(X),singI∗Ztr(Y,E)[n]),
the equality of morphism of CwDM−(Z)
[TABLE]
Proof.
(i): Obvious
(ii): Follows from (i).
∎
Recall TM(R)⊂CW is the full subcategory of topological manifolds.
Let X∈TM(R) connected topological manifold, Y∈CW a CW complex,
and l:E↪Y a CW subcomplex.
Denote by pX:In×X×Y→X, pY:In×X×Y→Y and
pX×Y:In×X×Y→X×Y the projections.
Let
[TABLE]
such that ∂I∗T=0,
where mi:Ti↪In×X×Y is a closed CW subcomplex for all i.
Denote pXi=pX∘mi:Ti→X and pYi=pX∘mi:Ti→Y.
Consider the class of T :
[TABLE]
We have then
•
its image
[TABLE]
by the composite functor
Recw∗tr∘D(I1,usu):PSh(CorZ(CW),C−(Z))→D−(Z),
where the last equality follows from the fact that by theorem 16(ii)
–
S(Ztr(X))):Ztr(X)→singI∗Ztr(X)
is an equivalence (I1,usu) local and
–
singI∗Ztr(X) and singI∗Ztr(Y)
are I1 local objects,
•
the action of pX×Y(T)∈ZdX+n(X×Y,Z)
[TABLE]
on homology, where, for each i :
–
pXi∗∈HomD−(Z)(C∗(X,Z),C∗(Ti,Z)[n])
is the Gynsin morphism (pXi is proper and X∈TM(R) is a topological manifold),
–
pYi∗=Z(pYi)(Δ∗):C∗(Ti,Z)→C∗(Y,Z) is the classical map on singular chain,
–
cY,E:C∗(Y,Z)→C∗(Y,E,Z) is the quotient map.
Here, since X is a connected topological manifold, we have HdXBM(X,Z)=[X]=Z and
since Ti⊂In×X×Y is finite surjective and irreducible over In×X,
we have HdX+nBM(Ti\∂(Ti),Z)=[Ti]=Z.
We have the following :
Lemma 9**.**
Let X∈TM(R) connected, Y∈CW and E⊂Y a subcomplex.
Let T=∑iniTi∈singInZtr(Y,E)(X) such that ∂I∗T=0
and let γ=∑jmjγj∈Cp(X,Z) such that ∂γ=0.
Then
[TABLE]
Proof.
Denote by Ti,γj:=pY(Ti∘(IIn×γj))⊂Y
where pY:Ip+n×Y→Y is the projection.
We have then dimTi,γj≤dim(Ti∘(IIn×γj))=p+n
and the factorization
Ti∘(IIn×γj)=iγj∘Ti,γjo
where,
•
Ti,γjo⊂Ip+n×Ti,γj is finite and surjective over Ip+n,
so that dimTi,γjo=p+n≥dimTi,γj,
•
iγj:Ti,γj↪Y is the closed embedding.
We can assume without loss of generality that dimTi,γj=p+n. We have then,
[TABLE]
where [⋅] denote the fundamental class in Borel Moore homology and ∂ the boundary.
∎
We deduce from lemma 8(ii) and lemma 9 the following :
Proposition 23**.**
Let X∈TM(R) connected and Y∈CW. Let l:E↪Y a CW subcomplex.
(i)
Let [T]=[∑iniTi]∈HomPC−(CW)(Ztr(X),singI∗(Ztr(Y,E))[n])
Then,
[TABLE]
(ii)
If Y is compact and E⊂Y is closed then the factorization
[TABLE]
where
[⋅]:ZdX+n(X×Y,X×E,Z)→HdX+nBM(X×Y,X×E,Z)
is the fundamental class gives the classical isomorphism
Hence Recw∗tr∘D(I1,usu)
is an isomorphism if X∈TM, Y is compact and E is closed
since Kn(X,(Y,E)) is an isomorphism by the six functor formalism as indiquated in [7]
if X∈TM, Y is compact and E is closed.
On the other hand, by theorem 17 (iii)
[TABLE]
is an isomorphism.
This proves the proposition.
∎
3 The Betti realization functor on the derived category of motives of complex algebraic varieties
We will consider,
•
the analytical functor An:Var(C)→AnSp(C)
given by An(V)=Van, An(g)=gan,
and the analytical functor on transfers
An:CorΛfs(SmVar(C))→CorZfs(AnSm(C))
given by An(V)=Van,An(Γ)=Γan,
•
the forgetful functor Cw:AnSp(C)→CW
given by Cw(W)=Wcw, Cw(g)=gcw,
and the forgetful functor on transfers
Cw:CorZfs(AnSm(C))→CorZfs(CW) given by Cw(W)=Wcw,
Cw(Γ)=Γcw,
•
the composites \widetilde{\mathop{\rm Cw}}\nolimits=\mathop{\rm Cw}\nolimits\circ\mathop{\rm An}\nolimits:\mathop{\rm Var}\nolimits(\mathbb{C})\to\mathop{\rm CW}\nolimits,
given by \widetilde{\mathop{\rm Cw}}\nolimits(V)=V^{cw}, \widetilde{\mathop{\rm Cw}}\nolimits(g)=g^{cw},
and \widetilde{\mathop{\rm Cw}}\nolimits=\mathop{\rm Cw}\nolimits\circ\mathop{\rm An}\nolimits:\mathop{\rm Cor}\nolimits^{fs}_{\mathbb{Z}}(\mathop{\rm SmVar}\nolimits(\mathbb{C}))\to\mathop{\rm Cor}\nolimits^{fs}_{\mathbb{Z}}(\mathop{\rm CW}\nolimits),
given by \widetilde{\mathop{\rm Cw}}\nolimits(V)=V^{cw}, \widetilde{\mathop{\rm Cw}}\nolimits(\Gamma)=\Gamma^{cw}.
•
the embeddings of categories
ιan:AnSm(C)→AnSp(C) and
ιvar:SmVar(C)→Var(C).
By definition, we have the following commutative diagram of sites
Let K∙∈PSh(CW,C−(Z)) be an I1 local object.
Let h:K∙→L∙ an equivalence usu local with L∙ usu fibrant.
Then,
–
L∙ is I1 local,
S(L∙):L∙→singI∗L∙ is an equivalence usu local,
–
singI∗L∙ is usu fibrant.
Since Cw∗ preserve usu local equivalence and usu fibrant object
(for X∈AnSm(C), the restriction of the functor Cw to the small site of open subset of X is fully faithfull),
–
Cw∗(S(L∙)∘h)=(Cw∗S(L∙))∘(Cw∗h):Cw∗K∙→Cw∗singI∗L∙ is an equivalence usu local
–
Cw∗singI∗L∙ is usu fibrant.
Now, for X∈AnSm(C),
[TABLE]
Moreover, the map
pXcw=IXcw×aD1:Xcw×D1→Xcw is an homotopy equivalence.
Hence, by lemma 6(ii),
[TABLE]
is a quasi-isomorphism.
In particular Hn(singI∗L∙(pXcw)) is an isomorphism.
This proves that Cw∗singI∗L∙ is D1 local.
Now, since
–
Cw∗(S(L∙)∘h) is an equivalence usu local
–
Cw∗singI∗L∙ is D1 local,
Cw∗K∙ is D1 local.
•
Let K∙∈PSh(CorZfs(CW),C−(Z)) be an I1 local object.
By lemma 7, Tr∗K is I1 local
Hence by above, Tr∗Cw∗K∙=Cw∗Tr∗K∙ is D1 local.
Thus, by lemma 7Tr∗K∙ is D1 local.
•
Let f:G1→G2 an equivalence (D1,usu) local in PSh(AnSm(C),C−(Z)).
Let K∈PSh(CW,C−(Z)) a I1 local object. Up to replace K by an usu equivalent presheaf,
we may assume that K is usu fibrant. Then Cw∗K is also usu fibrant.
Consider the following commutative diagram :
[TABLE]
By above Cw∗K is D1 local, and f is an equivalence (D1,usu) local.
Hence, Z(Cw∗K)(f) is an isomorphism since Cw∗K is usu fibrant.
The diagram shows then that Z(K)(Cw∗f) is an isomorphism.
This proves that Cw∗f:Cw∗G1→Cw∗G2 is an equivalence (I1,usu) local since K is usu fibrant.
•
Let f:G1→G2 an equivalence (D1,usu) local in PSh(CorZfs(AnSm(C)),C−(Z)).
Let K∈PSh(CorZfs(CW),C−(Z)) a I1 local object.
Up to replace K by an usu equivalent presheaf, we may assume that K is usu fibrant. Then Cw∗K is also usu fibrant.
Consider the following commutative diagram :
[TABLE]
By above Cw∗K is D1 local, and f is an equivalence (D1,usu) local.
Hence, Z(Cw∗K)(f) is an isomorphism since Cw∗K is usu fibrant.
The diagram shows then that Z(K)(Cw∗f) is an isomorphism.
This proves that Cw∗f:Cw∗G1→Cw∗G2 is an equivalence (I1,usu) local
since K is usu fibrant.
(iii): Follows from (i) and (ii) since by definition \widetilde{\mathop{\rm Cw}}\nolimits=\mathop{\rm Cw}\nolimits\circ\mathop{\rm An}\nolimits.
∎
3.1 Ayoub’s Betti realization functor and the Betti realisation functor via CW commplexes
We recall the definition of Ayoub’s realization functor :
The Betti realisation functor (without transfers) is the composite :
[TABLE]
(ii)
The Betti realisation functor with transfers is the composite :
[TABLE]
Since An∗ derive trivially by proposition 25(i)
and LTr∗:AnDA−(Z)→AnDM−(Z) is the inverse of Tr∗ by theorem 14(i), we have
Bti0∗=Bti∗∘LTr∗.
We now define a Betti realization functor via CW complexes.
The main result of this subsection will be that this functor coincide with Ayoub’s one.
Definition 24**.**
(i)
The CW-Betti realization functor (without transfers) is the composite :
[TABLE]
(ii)
The CW-Betti realisation functor with transfers is the composite :
[TABLE]
Since \widetilde{\mathop{\rm Cw}}\nolimits^{*} derive trivially by proposition 25(ii)
and LTr∗:CwDA−(Z)→CwDM−(Z) is the inverse of Tr∗ by theorem 17(i), we have
\widetilde{\mathop{\rm Bti}}\nolimits_{0}^{*}=\widetilde{\mathop{\rm Bti}}\nolimits^{*}\circ L\mathop{\rm Tr}\nolimits^{*}.
Denote by i∗:Iet∗↪□C∗=□∗ the embeddings
of pro-objects in SmVar(C)
[TABLE]
indexed by the filtrant category of etale neighborhood of [0,∞]∗ in □C∗,
The morphism i∗ is given by the etale morphisms i∗(l):Ul→□C∗ associated to
l∈Vet([0,∞]∗,□C∗).
Denote by i∗′:Ian∗i1∗′Dˉ∗i2∗′AC∗
the embeddings of the pro-objects in AnSm(C)
•
Ian∗:=(Uj,ajk)j,k∈Van(I∗,AC∗),j≤k indexed
by the filtrant category of etale analytic neighborhood of I∗ in AC∗, and
•
Dˉ∗:=(Ul,alm)l,m∈Van(Dˉ∗,AC∗),l≤m indexed
by the filtrant category of etale analytic neighborhood of I∗ in AC∗.
The morphism i1∗′ is given by the identities i1∗′(l):Uτ(l)=UlIUlUl, where
τ:Van(Dˉ∗,AC∗)→Van(I∗,AC∗)
is the natural embedding of categories.
The embedding i∗:Iet∗↪□∗ of pro-objets in SmVar(C) gives for,
F∙∈PSh(CorZfs(SmVar(C)),C−(Z)),
the following morphism in PSh(CorZfs(SmVar(C)),C−(Z)),
[TABLE]
given by for X∈SmVar(C), the morphism of complexes
[TABLE]
The morphism i1∗′:Ian∗→Dˉ∗ of pro-objet in AnSm(C) gives for,
G∙∈PSh(CorZfs(AnSm(C)),C−(Z)),
the following morphism in PSh(CorZfs(AnSm(C)),C−(Z)),
[TABLE]
given by for X∈SmVar(C), the morphism of complexes of abelian groups
[TABLE]
We have two canonical morphism of functors :
•
the morphism ψCw, which,
for G∙∈PSh(CorZfs(AnSm(C)),C−(Z)), associate the morphism
[TABLE]
in PSh(CorZfs(CW),C−(Z)) ;
the morphism ψCw(G∙) is given by, for Z∈CW,
[TABLE]
given by
(f:Xcw→Z)↦(f×II∗:(X×Ian∗)cw→Z×I∗)
and the identity of G∙(X×Ian∗) ;
•
the morphism ψCw, which,
for F∙∈PSh(CorZfs(SmVar(C)),C−(Z)), associate the morphism
[TABLE]
in PSh(CorZfs(CW),C−(Z)) ;
the morphism ψCw(F∙) is given by, for Z∈CW,
[TABLE]
given by
(f:Xcw→Z)↦(f×II∗:(X×Iet∗)cw→Z×I∗)
and the identity of F∙(X×Iet∗).
Definition 25**.**
We define the following two morphism of functors :
(i)
the morphism W, which,
for G∙∈PSh(CorZfs(AnSm(C)),C−(Z)), associate the composition
[TABLE]
in PSh(CorZfs(CW),C−(Z)),
(ii)
the morphism W, which,
for F∙∈PSh(CorZfs(SmVar(C)),C−(Z)), associate the composition
[TABLE]
in PSh(CorZfs(CW),C−(Z)).
Proposition 26**.**
(i)
For G∙∈PSh(CorZfs(AnSm(C)),C−(Z)),
W(G∙):Cw∗(singDˉ∗G∙)→singI∗Cw∗G∙
is an equivalence (I1,usu) local in PSh(CorZfs(CW),C−(Z)),
(ii)
For F∙∈PSh(CorZfs(SmVar(C)),C−(Z)),
\widetilde{W}(F^{\bullet}):\widetilde{\mathop{\rm Cw}}\nolimits^{*}(\underline{C}_{*}F^{\bullet})\to\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits^{*}F^{\bullet}
is an (I1,usu) local equivalence in PSh(CorZfs(CW),C−(Z)).
Proof.
(i):Consider the following commutative diagram
[TABLE]
•
By theorem 13(ii) S(G∙) is a (D1,et) equivalence.
Hence by proposition 25 (ii) Cw∗S(G∙) is a (I1,usu) equivalence.
•
By theorem 16 (ii) S(Cw∗G∙) is a (I1,usu) equivalence.
Now, since Cw∗S(G∙) and S(Cw∗G∙) are (I1,usu) equivalence,
the diagram (79) shows that W(G∙) is a (I1,usu) equivalence.
(ii): Consider the following commutative diagram
[TABLE]
•
By theorem 11(ii) S(F∙) is a (A1,et) equivalence.
Hence, by proposition 25 (iii) \widetilde{\mathop{\rm Cw}}\nolimits^{*}S(F^{\bullet}) is a (I1,usu) equivalence.
•
By theorem 16 (ii) S(\widetilde{\mathop{\rm Cw}}\nolimits^{*}F^{\bullet}) is a (I1,usu) equivalence.
Now, since \widetilde{\mathop{\rm Cw}}\nolimits^{*}S(F^{\bullet}) and S(\widetilde{\mathop{\rm Cw}}\nolimits^{*}F^{\bullet}) are (I1,usu) equivalence,
the diagram (80) shows that W(F∙) is a (I1,usu) equivalence.
∎
Definition 26**.**
We define the morphism of functor B, by associating
to G∙∈PSh(CorZfs(AnSm(C)),C−(Z)), the composite
[TABLE]
in PSh(CorZfs(AnSm(C)),C−(Z))
We have now the following key proposition.
Proposition 27**.**
(i)
For Y∈AnSp(C) and E⊂Y an analytic subset,
[TABLE]
is a quasi isomorphism in C−(Z).
(ii)
For Y∈AnSp(C), and E⊂Y an analytic subset, the morphism
[TABLE]
is an equivalence (D1,usu) local in PSh(CorZfs(AnSm(C)),C−(Z)).
Proof.
(i):
•
Consider first Y∈AnSm(C).
By proposition 10, there exist a covering Y=∪i∈JDi by a countable family of open balls
Di≃DdY such that DI=∅ or DI≃DdY,
for all I={i1,⋯il}⊂J. Denote by jI:DI↪Y the open embedding.
We have then the following commutative diagram in PSh(CorZfs(AnSm(C)),C−(Z)) :
[TABLE]
This gives, after applying the functor ean∗tr to (81), the commutative diagram in C−(Z):
[TABLE]
Since DI≃DdY,
–
Tot(H0(singI∗Ztr(DIcw))):Tot∙,∗(⊕cardI=∙singI∗Ztr(DIcw))→⊕cardI=∙H0(singI∗Ztr(DIcw))
is a quasi-isomorphism in C−(Z) by proposition 22 ;
–
Tot(H0(singDˉ∗Ztr(DI))):Tot∙,∗(⊕cardI=∙singDˉ∗Ztr(DI))→⊕cardI=∙H0(singDˉ∗Ztr(DI))
is a quasi-isomorphism in C−(Z)
since singDˉ∗Ztr(DI)→Ztr(pt) is an equivalence
usu local, since, by theorem 13(ii), it is an equivalence (D1,usu) local between
D1 local objects ;
–
H0(ean∗trB(Ztr(DI))):⊕cardI=∙H0(singDˉ∗Ztr(DI))→⊕cardI=∙H0(singI∗Ztr(DIcw))
is a quasi-isomorphism in C−(Z) by proposition by the two preceding points.
Hence,
[TABLE]
is a quasi-isomorphism in C−(Z).
On the other hand,
–
Tot(ean∗trS(Z(jI))):Tot∙,∗(⊕cardI=∙singDˉ∗Ztr(DI))→singDˉ∗Ztr(Y) is a quasi-isomorphism in C−(Z)
since
[TABLE]
is a (D1,usu) local equivalence in PSh(CorZfs(AnSm(C)),C−(Z))
by lemma 1 and theorem 13(ii),
–
Tot(ecw∗trS(Z(jIcw))):Tot∙,∗(⊕cardI=∙singI∗Ztr(DIcw))→singI∗Ztr(Ycw) is a quasi-isomorphism in C−(Z)
since
[TABLE]
is an equivalence (I1,usu) local in PSh(CorZfs(CW),C−(Z))
by lemma 5 and theorem 16(ii)
Consider now Y∈AnSp(C).
We prove that ean∗trB(Ztr(Y)) is a quasi-isomorphism by induction on dY=dimY.
If dY=0 there is nothing to prove.
By theorem 15, there exist a proper modification
ϵ:Y′→Y such that Y′∈AnSm(C) and E=ϵ−1(Z)⊂Y′ is a normal crossing divisor.
Denote by l:Z↪Y and l′:E↪Y′ the closed embeddings and
ϵZ:E→Z the morphism such that l∘ϵZ=ϵ∘l′.
We have then, with
–
b=ϵZ∗+l∗′=S(Ztr(ϵZ)⊕Ztr(l′)) and
a=ϵ∗+l∗=S(Ztr(ϵ)⊕Ztr(l)),
–
d=ϵZ∗cw+l∗′cw=S(Ztr(ϵZcw)⊕Ztr(l′cw)) and
c=ϵ∗cw+l∗cw=S(Ztr(ϵcw)⊕Ztr(lcw)),
the following commutative diagram in PSh(CorZfs(AnSm(C)),C−(Z)) :
[TABLE]
But,
–
[0→singDˉ∗Ztr(E)bsingDˉ∗Ztr(Z)⊕singDˉ∗Ztr(Y′)]asingDˉ∗Ztr(Y)
is a quasi-isomorphism by proposition 14,
–
[0→singI∗Ztr(Ecw)dsingI∗Ztr(Zcw)⊕singI∗Ztr(Y′cw)]csingI∗Ztr(Ycw)
is a quasi-isomorphism by proposition 17 and the fact that the 2-cubical
diagram associated to a proper modification is of cohomological descent (see e.g. [13]),
–
as Y′∈SmVar(C) is smooth and dimE=dY−1<dY, dimZ<dY,
we have, by the smooth case we proved above and by the induction hypothesis,
that ean∗trB(Ztr(Z)⊕Ztr(Y′)) and ean∗trB(Ztr(E))
are quasi-isomorphisms,
thus, the diagram (83) shows that ean∗trB(Ztr(Y)) is a quasi-isomorphism.
•
Consider now Y∈AnSp(C) as before and E⊂Y an analytic subspace.
Denote by l:E↪Y the locally closed embedding.
We have then the following commutative diagram in PC−(An) :
[TABLE]
where the first row is the exact sequence (27), and the second row is an exact sequence
since the sequence (43) is exact and Cw∗ is an exact functor.
Since we just showed that ean∗trB(Ztr(Y)) and ean∗trB(Ztr(E))
are quasi-isomorphism, the diagram (84) implies that ean∗trB(Ztr(Y,E))
is a quasi isomorphism.
(ii): Follows by (i).Let us explain.
•
On the one hand,
–
By theorem 13 (ii), singDˉ∗Ztr(Y,E) is D1 local.
–
By theorem 16(ii), singI∗Ztr(Ycw,Ecw) is I1 local.
Hence, Cw∗singI∗Ztr(Ycw,Ecw) is D1 local,
by proposition 25 (ii).
•
On the other hand by (i)
ean∗tr(B(Ztr(Y,E))):singD∗ˉZtr(Y,E)→singI∗Ztr(Ycw,Ecw)
is a quasi isomorphism in C−(Z).
The main result of this subsection is the following :
Theorem 18**.**
(i)
For Y∈Var(C), and E⊂Y a subvariety, we have \mathop{\rm Bti}\nolimits^{*}M(Y,E)=\widetilde{\mathop{\rm Bti}}\nolimits^{*}M(Y,E)
(ii)
For X,Y∈Var(C), D⊂X, E⊂Y subvarieties,
and n∈Z, n≤0, the following diagram is commutative
[TABLE]
where we denoted for simplicity X for Xan and Xcw, and similarly for D, Y and E.
Proof.
(i): By definition, Bti∗M(Y,E)=Rean∗tr(Ztr(Yan,Ean)).
Since, by theorem 13(ii),
•
S(Ztr(Yan,Ean)):Ztr(Yan,Ean)→singD∗Ztr(Yan,Ean)
is an equivalence (D1,usu) local in PC−(An) and
•
singD∗Ztr(Yan,Ean)
is a D1 local object,
we have
Bti∗M(Y,E)=ean∗tr(singD∗Ztr(Yan,Ean))=singD∗Ztr(Yan,Ean).
Since
•
B(Ztr(Yan,Ean)):singD∗Ztr(Yan,Ean)→Cw∗singI∗Ztr(Ycw,Ecw)
is an equivalence (D1,usu) local in PC−(An)
by proposition 27 (ii), and
•
Cw∗singI∗Ztr(Ycw,Ecw)
is a D1 local object by theorem 16(ii) and proposition 25(ii),
we have
[TABLE]
By definition, \widetilde{\mathop{\rm Bti}}\nolimits^{*}M(X)=Re^{tr}_{cw*}(\mathbb{Z}_{tr}(Y^{cw},E^{cw})).
Since, by theorem 16(ii),
•
S(Ztr(Ycw,Ecw)):Ztr(Ycw,Ecw)→singI∗Ztr(Ycw,Ecw)
is an equivalence (I1,usu) local in PC−(CW) and
•
singI∗Ztr(Ycw,Ecw) is an I1 local object,
we have
[TABLE]
This proves (i).
(ii):
Let α∈HomDM−(C,Z)(M(X,D),M(Y,E)[n]).
Consider the commutative diagram in AnDM(Z)
[TABLE]
Since
singD∗Ztr(Xan,Dan) and singD∗Ztr(Yan,Ean)
are D1 local objects by theorem 13(ii),
[TABLE]
Thus,
[TABLE]
Since
singI∗Ztr(Xcw,Dcw) and singI∗Ztr(Ycw,Ecw)
are I1 local objects by theorem 16(ii),
[TABLE]
Thus
[TABLE]
Since
Cw∗singI∗Ztr(Xcw,Dcw) and Cw∗singI∗Ztr(Ycw,Ecw)
are D1 local objects by theorem 16(ii) and proposition 25 (ii),
[TABLE]
Thus
[TABLE]
∎
3.2 The image of algebraic correspondences after localization by Ayoub’s Betti realization functor
Definition 27**.**
Let V∈Var(C) quasi projective and p,q∈N.
Recall Zp(V,∗)⊂Zp(□∗×V,Z) is the Bloch cycle complex
consisting of closed subset meeting the face of □∗ properly.
(i)
Let Zq(Vcw,∗)⊂Zq(I∗×Vcw,Z)
be the abelian subgroup consisting of closed subset meeting the face of In properly.
We have the restriction map induced by the closed embedding of CW complexes
in×IV:In×Vcw↪□n×Vcw,
which gives the morphism of complexes of abelian groups
[TABLE]
(ii)
Let Y∈PVar(C) be a compactification of V and E=Y\V.
The higher cycle class map (**[11]**) is the morphism of complexes of abelian groups :
[TABLE]
αˉ∈Zp(Y×□n)* is the closure of α and
pY:Ycw×I∗→Ycw is the projection*
Let X∈SmVar(C) and Y∈Var(C).
Let E⊂Y be a closed subset and V=Y\E.
Denote by j:V↪Y the open embeddings.
We have the open embedding IX×j:X×V↪X×Y.
The map of complexes T^X×YdY induces a map
denoted by the same way on the subcomplexes indicated in the following diagram
in C−(Z):
[TABLE]
Lemma 10**.**
Let Y∈Var(C), X∈SmVar(C), E⊂Y a closed subset and V=Y\E
the open complementary. Let n∈Z, n≤0. Then, for
[TABLE]
we have the following equality of morphisms of PSh(CorZ(CW),C−(Z)).
[TABLE]
where HnT^X×V is the map induced in cohomology of the map of complex T^X×V.
Proof.
Let αn∈CnZtr(Y,E)(X) such that
α=[αn]∈HnC∗Ztr(Y,E)(X).
We have
[TABLE]
∎
We deduce from this lemma 10 and proposition 26(ii) the following
Proposition 28**.**
Let X∈SmVar(C), Y∈Var(C), E⊂Y a closed subset
and V=Y\E the open complementary.
Let n∈Z, n≤0. Then, the following diagram is commutative
[TABLE]
Proof.
Let α∈HomPC−(Ztr(X),C∗Ztr(Y,E)[n]).
We have
[TABLE]
∎
Using proposition 28 and proposition 23 (i),
we immediately deduce from theorem 18 the following :
Corollary 3**.**
Let X∈SmVar(C), Y∈Var(C), E⊂Y a closed subvariety
and V=Y\E the open complementary.
Let n∈Z, n≤0. Then,
(i)
the following diagram is commutative
[TABLE]
where we denoted for simplicity X for Xan and Xcw, and similarly for Y and E.
(ii)
for α∈HomPC−(Ztr(X),C∗Z(Y,E)[n]),
we have
[TABLE]
where pX×Y:Xcw×Ycw×In→Xcw×Ycw is the projection.
Proof.
(i): By theorem 18 (ii) and proposition 28,
the following diagram is commutative
[TABLE]
where we denote for simplicity X for Xan and Xcw, and similary for Y and E.
This proves (i).
(ii) : We have,
[TABLE]
This proves (ii).
∎
3.3 Ayoub’s Betti realization functor and Nori motives
We denote by Cb(CorZ(SmVar(C))) the category of bounded complexes of the
category CorZ(SmVar(C)). We have the Yoneda embedding
[TABLE]
A (small) diagram is a 1-simplicial set, that is a functor from Δ1 to Set
We denote by Var2(C) the diagram of pairs
whose set of vertices are the triplet (X,D,i), X,D∈Var(C),
D⊂X a closed subvariety and i∈N, and whose set of edges
are the morphism of pairs (X,D,i)→(Y,E,i) and if (Z,K)⊂(X,D),
there is an edge (X,D,i)→(Z,K,i−1).
We have a morphism of diagram
[TABLE]
We denote by GVar2(C)⊂Var2(C) the subdiagram of good pairs
that is (X,D,i) is a good pair if X,D are affine, dimX=i and if Hp(X,D,Z)=0 is p=i.
We recall the definition of Nori motives.
Let X∈Var(C),
we have the following isomorphism in Db(Z) :
[TABLE]
(ii)
Let X,Y∈SmVar(C), then the following diagram commutes
[TABLE]
We will use the following lemma
Lemma 11**.**
Let C,S two categories, and D:C→C′ a localization functor.
Let F1,F2:C′→S
two functors. If
[TABLE]
then F1=F2.
Proof.
Let f′:C1→C2 a morphism in C′.
Without loss of generality, we can assume that f′=f∘s−1 with
f:C0→C2 and s:C0→C1 morphisms in C, such that
s belongs to the class of morphisms of C we localize.
We have then,
[TABLE]
∎
We deduce from theorem 18, corollary 3(ii),
proposition 29, end lemma 11, the following main result :
Theorem 19**.**
(i)
For X∈SmVar(C),
Bti∗∘D(A1,et)(Z(X))=oN∘N(X)
(ii)
For X,Y∈SmVar(C), the following diagram commutes
[TABLE]
(iii)
The Betti realisation functor factor through Nori motives. That is
Bti∗=oN∘Nˉ
Proof.
(i): We have
[TABLE]
(ii): Let α∈HomPC−(Z(X),Z(Y)),
we have
[TABLE]
(iii): By (i) and (ii), we have the equality
[TABLE]
Lemma 11 applied to this equality (92) say that Bti∗=oN∘Nˉ.
This proves (iii).
∎
3.4 The image of Ayoub’s Betti realization functor on morphism and Bloch cycle class map
Let V∈Var(C) quasi-projective.
Let Y∈PVar(C) a compactification of V and E=Y\V
Denote by j:V↪Y the open embedding.
For r∈N, we denote by
iY:Y↪Ar×Y and
iV=iY∣V:V↪Ar×V the inclusions,
by pY:Ar×Y→Y and pV=pY∣V:Ar×V→V the projections,
by a:Ar↪Pr the open embedding,
and by E′=Y×Pr\(V×Ar).
We consider
•
for p≤dV, the inclusion of complexes
iV∗:Zp(V,∗)↪ZdV(AdV−p×V,∗),
which is a quasi isomorphism.
•
for p≥dV the inclusion of complexes
pV∗:Zp(Ap−dV,∗×V)↪Zp(V,∗),
which is a quasi-isomorphism.
Since Y is projective, and Ar is smooth,
•
D(A1,et):HomPC−(Ztr(Ar),C∗Z(Y,E)[n])→HomDM−(C,Z)(M(Ar),M(Y,E)[n])
is an isomorphism by proposition 9
•
D(I1,et):HomPC−(CW)(Ztr(Ar,cw),singI∗Ztr(Ycw,Ecw)[n])→HomCwDM−(Z)(M(Ar,cw),M(Ycw,Ecw)[n])
is an isomorphism by proposition 24.
On the other side, the inclusion of complexes of abelian groups
•
for p≤dV,
(IA×j)∗:C∗Ztr(Y,E)(AdV−p)↪ZdV(AdV−p×V,∗)
•
for p≥dV,
(a×j)∗:C∗Ztr(Pp−dV×Y,E′)(Spec(C))↪Zp(Ap−dV×V,∗)
is a quasi-isomorphisms.
Definition 28**.**
Let V∈Var(C) quasi-projective.
Let Y∈PVar(C) a compactification of V and E=Y\V.
Recall that j:V↪Y is the open embedding.
(i)
For p≤dV, we consider the following composition of isomorphisms of abelian groups
[TABLE]
(ii)
For p≥dV, we consider the following composition of isomorphisms of abelian groups
[TABLE]
We now prove that under these identifications, the image of the Betti realization
functor on morphism coincide with the Bloch cycle class map.
Theorem 20**.**
Let V∈Var(C). Let Y∈PVar(C) be a compactification of V and E=Y\V.
Then,
(i)
for p≤dV, the following diagram commutes :
[TABLE]
(ii)
for p≥dV, the following diagram commutes :
[TABLE]
Proof.
(i):
By corollary 3(ii), the second square of following diagram commutes
[TABLE]
where p:AdV−p×V×In→AdV−p×V is the projection in CW.
Moreover Kn(AdV−p,(Y,E))∘[⋅]=Kn(AdV−p,(Y,E))
by definition.
On the other hand, we have the following commutative diagram.
[TABLE]
This proves (i).
(ii):
By corollary 3(ii), the second square of following diagram commutes
[TABLE]
where p:Pp−dV×Y×In→Pp−dV×Y is the projection in CW.
Moreover
Kn(pt,Pp−dV×Y,E′))∘[⋅]=Kn(pt,Pp−dV×Y,E′))
by definition.
On the other hand, we have the following commutative diagram.
[TABLE]
This proves (ii).
∎
4 The relative case
4.1 The derived category of relative motives of algebraic varieties
Let S∈Var(k). The category Var(k)/S is the category whose objects are
X/S=(X,h) with X∈Var(k) and h:X→S is a morphism,
and whose space of morphisms between
X/S=(X,h1) and Y/S=(Y,h2)∈Var(k)/S are the morphism g:X→Y such that h2∘g=h1.
We denote by Var(k)sm/S⊂Var(k)/S the full subcategory consisting of the objects
X/S=(X,h) with X∈Var(k), such that h:X→S is a smooth morphism.
For X/S=(X,h)∈Var(k)/S, and n∈N, we denote by
•
(X×A1/S):=(X×kA1,h∘pX)=(X×S(A1×S)/S),
where pX:X×kA1→X is the projection.
•
(X×□n/S):=(X×k□n,h∘pX)=(X×S(□n×S)/S),
where pX:X×k□n→X is the projection.
Definition 29**.**
[6]**
Let S∈Var(k).
We define CorΛfs(Var(k)sm/S) to be the category whose objects are the one of Var(k)sm/S
and whose space of morphisms between
X/S and Y/S∈Var(k)sm/S is the free Λ module Zfs/X(X×SY,Λ).
The composition of morphisms is defined similary then in the absolute case (see [6])
We have
•
the additive embedding of categories
Tr(S):Z(Var(k)sm/S)↪CorZfs(Var(k)sm/S)
which gives the corresponding morphism of sites
Tr(S):CorZfs(Var(k)sm/S)→Z(Var(k)sm/S).
•
the inclusion functor
evar(S):Ouv(S)↪Var(k)sm/S,
which gives the corresponding morphism of sites
evar(S):Var(k)sm/S→Ouv(S),
•
the inclusion functor
evartr(S):=Tr∘evarOuv(S)↪CorZfs(Var(k)sm/S)
which gives the corresponding morphism of sites
evartr(S):=Tr∘evar:CorZfs(Var(k)sm/S)→Ouv(S).
For each morphism f:T→S in Var(C), we have
•
the pullback functor
P(f):Var(C)/S→Var(C)/T, P(f)(X/S)=(X×ST/T), P(f)(h)=hT,
which gives the morphism of sites P(f):Var(C)/T→Var(C)/S
•
the pullback functor
P(f):CorZfs(Var(C)sm/S)→CorZfs(Var(C)sm/T),
P(f)(X/S)=(X×ST/T), P(f)(h)=hT,
which gives the morphism of sites
P(f):CorZfs(Var(C)sm/T)→CorZfs(Var(C)sm/S).
For S∈Var(k), we consider the following two big categories :
•
PSh(Var(k)sm/S,C−(Z))=PShZ(Z(Var(k)sm/S),C−(Z)),
the category of bounded above complexes of presheaves on Var(k)sm/S,
or equivalently additive presheaves on Z(Var(k)sm/S),
sometimes, we will write for short P−(S)=PSh(Var(k)sm/S,C−(Z)),
•
PShZ(CorZfs(Var(k)sm/S),C−(Z)),
the category of bounded above complexes of additive presheaves on CorZfs(Var(k)sm/S)
sometimes, we will write for short PC−(S)=PSh(CorZfs(Var(k)sm/S),C−(Z)),
given by Tr(S):CorZfs(Var(k)sm/S)→Z(Var(k)sm/S),
evar(S):Var(k)sm/S→Ouv(S) and
evartr(S):CorZfs(Var(k)sm/S)→Ouv(S) respectively.
We denote by aet:PShZ(Var(k)sm/S,Ab)→ShZ,et(Var(k)sm/S,Ab)
the etale sheaftification functor.
For X/S∈Var(k)sm/S, we denote by
[TABLE]
the presheaves represented by X. They are etale sheaves.
For X/S=(X,h)∈Var(k)/S with h:X→S non smooth,
[TABLE]
is also an etale sheaf ; of course if h:X→S is not dominant then Ztr(X/S)=0
For a morphism f:T→S in Var(k) we have the adjonctions
given by P(f):Var(k)sm/T→Var(k)sm/S and
P(f):CorZfs(Var(k)sm/T)→CorZfs(Var(k)sm/S),
respectively.
Definition 30**.**
(i)
The projective etale toplogy model structure on PSh(Var(k)sm/S,C−(Z))
is defined in the similar way of the absolute case (c.f. definition 7(i)).
(ii)
The projective (Ak1,et) model structure on the category
PSh(Var(k)sm/S,C−(Z)) is the left Bousfield localization
of the projective etale topology model structure with respect to the class of maps
{Z(X×Ak1/S)[n]→Z(X/S)[n],X/S∈Var(k)sm/S,n∈Z}.
(iii)
The projective etale toplogy model structure on PSh(Corfs(Var(k)sm/S),C−(Z))
is defined in the similar way of the absolute case (c.f. definition 7(ii)).
(iv)
The projective (Ak1,et) model structure on the category
PShZ(CorZfs(Var(k)sm/S),C−(Z))
is the left Bousfield localization of the projective etale topology model structure with respect to the class of maps
{Ztr(X×Ak1/S)[n]→Z(X/S)[n],X/S∈Var(k)sm/S,n∈Z}.
Definition 31**.**
Let S∈Var(k)
(i)
We define
DM−(S,Z)et:=HoAk1,et(PShZ(CorZfs(Var(k)sm/S),C−(Z))),
to be the derived category of (effective) motives, it is
the homotopy category of PSh(CorZfs(Var(k)sm/S),C−(Z))
with respect to the projective (A1,et) model structure (c.f. definition 30(ii)).
We denote by
[TABLE]
the canonical localization functor.
(ii)
By the same way, we denote
DA−(S,Z)et:=HoAk1,et(PSh(Var(k)sm/S,C−(Z))) (c.f.30(i)) and
[TABLE]
the canonical localization functor.
Proposition 30**.**
Let S∈Var(k)
(i)
(Tr(S)∗,Tr(S)∗):PSh(Var(k)sm/S,C−(Z))⇆PShZ(CorZfs(Var(k)sm/S),C−(Z))*
is a Quillen adjonction for the etale topology model structures
and a Quillen adjonction for the (A1,et) model structures
(c.f. definition 30 (i) and (ii) respectively).*
(ii)
(evar(S)∗,evar(S)∗):PSh(Var(k)sm/S,C−(Z))⇆C−(Z)*
is a Quillen adjonction for the etale topology model structures
and a Quillen adjonction for the (A1,et) model structures
(c.f. definition 30 (i)).*
(iii)
(evar(S)tr∗,evartr(S)∗):PSh(CorZfs(Var(k)sm/S),C−(Z))⇆C−(Z)*
is a Quillen adjonction for the etale topology model structures
and a Quillen adjonction for the (A1,et) model structures
(c.f. definition 30 (ii)).*
Proof.
(i): Follows from the fact that Tr(S)∗ derive trivially hence is a right Quillen functor.
(ii): Follows from the fact that evar(S)∗ derive trivially hence is a left Quillen functor.
(iii): Follows from the fact that evartr(S)∗ derive trivially hence is a left Quillen functor.
Point (i) follows imediately from definition of A1 local objects.
Point (ii) follows from point (i).
∎
We immediately deduce the following
Proposition 32**.**
Let f:T→S a morphism in Var(k).
(i)
The adjonction
(f∗,f∗):PSh(Var(k)sm/S,C−(Z))⇆PSh(Var(k)sm/T,C−(Z))
is a Quillen adjonction with respect to the etale model structures and the (A1,et) model structures
(ii)
The adjonction
(f∗,f∗):PSh(CorZfs(Var(k)sm/S),C−(Z))⇆PSh(CorZfs(Var(k)sm/T),C−(Z)),
is a Quillen adjonction with respect to the etale model structures and the (A1,et) model structures
Proof.
By proposition 31(ii), f∗ is a left Quillen functor.
∎
Theorem 21**.**
[6]**
Let S∈Var(k). The adjonction
(Tr(S)∗,Tr(S)∗):PSh(Var(k)sm/S,C−(Z))⇆PShZ(CorZfs(Var(k)sm/S),C−(Z))
is a Quillen equivalence, that is the derived functor
[TABLE]
is an isomorphism and
Tr(S)∗:DM−(S,Z)∼DA−(S,Z)PShZ(CorZfs(Var(k)sm/S),C−(Z))
is its inverse
We have all the property of the six functor formalism on Var(k) :
Theorem 22**.**
[6]**
We have the six functor formalism on DM−:Var(k)→TriCat, S∈Var(k)↦DM−(S,Z).
Indeed, the 2-functor DM−:Var(k)→TriCat is an homotopic 2-functor is the sense of [4].
In particular, we have well a defined pair of adjoint functors
(f!,f!):DM−(T,Z)⇆DM−(S,Z), such that
•
f!=Rf∗* if f:T→S is proper,*
•
j!* is the left adjoint of j∗ if j:So↪S is an open embbeding
(j∗ admits a left adjoint since it is a smooth morphism).*
Proof.
See [6] or [4].
We just recall the definition of f!.
Let f:T→S a morphism in Var(k). Let Tˉ∈PVar(k) a compactification of T.
Take T^⊂Tˉ×S the closure of the graph of f.
Then, f=f^∘j where j:T↪T^ is the open embedding and
f^:=pS∣Tˉ is a proper morphism, with pS:Tˉ×S→S the projection.
Then for F∙∈PC−(T) a A1 local and etale fibrant object
f!F∙:=D(A1,et)(S)(f^∗jlF∙)
does not depends of the compactification of f by the support property.
∎
Definition 32**.**
Let S∈Var(C).
•
Let M∈DM−(S,Z) and p∈Z.
The Tate twist of M is M⊗D(A1,et)(ZS(p)) with
ZS(p):=(Ztr(P1×S,∞))⊗p∈PC−(S).
For F∙∈PC−(S), we denote F∙(p):=F∙⊗ZS(p)∈PC−(S).
•
A motive M∈DM−(S,Z) is called constructible if it belongs to the thick
subcategory generated by motives of the form D(A1,et)(S)(Ztr(X/S)(p)),
with X/S=(X,h), where h:X→S is a smooth morphism, and p∈Z.
Definition-Proposition 2**.**
Let S,X∈Var(k) and h:X→S a morphism.
(i)
The motive of X/S=(X,h) is M(X/S):=h!h!Ztr(S/S)∈DM−(S,Z).
It is a constructible object.
(ii)
The Borel-Moore motive of X/S=(X,h)
is MBM(X/S):=h!h∗Ztr(S/S)=h!Ztr(X/X)∈DM−(S,Z).
It is a constructible object.
(iii)
The cohomological motive of X/S=(X,h) is
M(X/S):=Rh∗h∗Ztr(S/S)=Rh∗Ztr(X/X)∈DM−(S,Z).
It is a constructible object.
(iv)
The motive with compact support of X/S=(X,h) is
M(X/S):=Rh∗h!Ztr(S/S)∈DM−(S,Z). It is a constructible object.
Proof.
The fact that these four object associated to X/S=(X,h)
are constructible follows from [6] section 4.
∎
4.2 The derived category of relative motives of analytic spaces
Let S∈AnSp(C). The category AnSp(C)/S is the category whose objects are
X/S=(X,h) with X∈AnSp(C) and h:X→S is a morphism,
and whose space of morphisms between
X/S=(X,h1) and Y/S=(Y,h2)∈SmVar(S) are the morphism g:X→Y such that h2∘g=h1.
We denote by AnSp(C)sm/S⊂AnSp(C)/S the full subcategory constisting of the objects
X/S=(X,h) with X∈AnSp(C), such that h:X→S is a smooth morphism.
For X/S=(X,h)∈AnSp(C)/S, and n∈N, we denote by
•
(X×D1/S):=(X×D1,h∘pX)=(X×S(D1×S)/S),
where pX:X×D1→X is the projection.
•
(X×Dˉn/S):=(X×Dˉn,h∘pX)=(X×S(Dˉn×S)/S),
where pX:X×Dˉn→X is the projection.
Definition 33**.**
Let S∈AnSp(C).
We define CorΛfs(AnSp(C)sm/S)
to be the category whose objects are the one of AnSp(C)sm/S
and whose space of morphisms between
X/S and Y/S∈AnSp(C)sm/S is the free Λ module Zfs/X(X×SY,Λ).
The composition of morphisms is defined similary then in the absolute case.
We have
•
the additive embedding of categories
Tr(S):Z(AnSp(C)sm/S)↪CorZfs(AnSp(C)sm/S)
which gives the corresponding morphism of sites
Tr(S):CorZfs(AnSp(C)sm/S)→Z(AnSp(C)sm/S).
•
the inclusion functor
ean(S):Ouv(S)↪AnSp(C)sm/S,
which gives the corresponding morphism of sites
ean(S):AnSp(C)sm/S→Ouv(S),
•
the inclusion functor
eantr(S):=Tr∘eanOuv(S)↪CorZfs(AnSp(C)sm/S)
which gives the corresponding morphism of sites
eantr(S):=Tr∘ean:CorZfs(AnSp(C)sm/S)→Ouv(S).
For each morphism f:T→S in AnSp(C), we have
•
the pullback functor
P(f):AnSp(C)/S→AnSp(C)/T, P(f)(X/S)=(X×ST/T), P(f)(h)=hT,
which gives the morphism of sites P(f):Var(C)/T→Var(C)/S
•
the pullback functor
P(f):CorZfs(AnSp(C)sm/S)→CorZfs(AnSp(C)sm/T),
P(f)(X/S)=(X×ST/T), P(f)(h)=hT,
which gives the morphism of sites
P(f):CorZfs(AnSp(C)sm/T)→CorZfs(AnSp(C)sm/S).
For S∈AnSp(C), we consider the following two big categories :
•
PSh(AnSp(C)sm/S,C−(Z))=PShZ(Z(AnSp(C)sm/S),C−(Z)),
the category of bounded above complexes of presheaves on AnSp(C)sm/S,
or equivalently additive presheaves on Z(AnSp(C)sm/S),
sometimes, we will write for short P−(S)=PSh(AnSp(C)sm/S,C−(Z)),
•
PShZ(CorZfs(AnSp(C)sm/S),C−(Z)),
the category of bounded above complexes of additive presheaves on CorZfs(AnSp(C)sm/S)
sometimes, we will write for short PC−(S)=PSh(CorZfs(AnSp(C)sm/S),C−(Z)),
given by Tr(S):CorZfs(AnSp(C)sm/S)→Z(AnSp(C)sm/S),
ean(S):AnSp(C)sm/S→Ouv(S) and
eantr(S):CorZfs(AnSp(C)sm/S)→Ouv(S) respectively.
We denote by aet:PShZ(AnSp(C)sm/S,Ab)→ShZ,et(AnSp(C)sm/S,Ab)
the etale sheaftification functor.
For X/S∈AnSp(C)sm/S, we denote by
[TABLE]
the presheaves represented by X. They are usu sheaves.
For a morphism f:T→S in AnSp(C) we have the adjonction
given by P(f):AnSp(C)sm/T→AnSp(C)sm/S and
P(f):CorZfs(AnSp(C)sm/T)→CorZfs(AnSp(C)sm/S),
respectively.
Definition 34**.**
Let S∈AnSp(C).
(i)
The projective usual toplogy model structure on PSh(AnSp(C)sm/S,C−(Z))
is defined in the similar way of the absolute case (c.f. definition 12(i)).
(ii)
The projective (D1,et) model structure on the category
PSh(AnSp(C)sm/S,C−(Z)) is the left Bousfield localization
of the projective usual topology model structure with respect to the class of maps
{Z(X×D1/S)[n]→Z(X/S)[n],X/S∈AnSp(C)sm/S,n∈Z}.
(iii)
The projective usual toplogy model structure on PSh(Corfs(AnSp(C)sm/S),C−(Z))
is defined in the similar way of the absolute case (c.f. definition 12(ii)).
(iv)
The projective (D1,et) model structure on the category
PShZ(CorZfs(AnSp(C)sm/S),C−(Z))
is the left Bousfield localization of the projective usual topology model structure with respect to the class of maps
{Ztr(X×D1/S)[n]→Z(X/S)[n],X/S∈AnSp(C)sm/S,n∈Z}.
Definition 35**.**
Let S∈AnSp(C).
(i)
We define
AnDM−(S,Z):=HoD1,et(PShZ(CorZfs(AnSp(C)sm/S),C−(Z))),
to be the derived category of (effective) motives, it is
the homotopy category of PSh(CorZfs(AnSp(C)sm/S),C−(Z))
with respect to the projective (D1,usu) model structure (c.f. definition 34(ii)).
We denote by
[TABLE]
the canonical localization functor.
(ii)
By the same way, we denote
AnDA−(S,Z):=HoD1,usu(PSh(AnSp(C)sm/S,C−(Z))) (c.f.34(i)) and
[TABLE]
the canonical localization functor.
We now look at an explicit localization functor.
For F∙∈PSh(AnSp(C)sm/S,Ab) and X/S∈AnSp(C)sm/S,
we have the complex F(X×Dˉ∗/S) associated to the cubical object
F(X×Dˉ∗/S) in the category of abelian groups.
•
If F∙∈PSh(AnSp(C)sm/S,C−(Z)) ,
[TABLE]
is the total complex of presheaves associated to the bicomplex of presheaves X/S↦F∙(Dˉ∗×X/S),
and singDˉ∗F∙:=ean(S)∗singDˉ∗F∙∈C−(Z).
We denote by S(F∙):F∙→singDˉ∗F∙,
[TABLE]
the inclusion morphism of PSh(AnSp(C)sm/S,C−(Z)) :
For f:F1∙→F2∙ a morphism in P−(An,S), we denote by
S(f):singDˉ∗F1∙→singDˉ∗F2∙,
the morphism of PSh(AnSp(C)sm/S,C−(Z)) given by for X/S∈AnSp(C)sm/S,
[TABLE]
•
If F∙∈PShZ(CorZfs(AnSp(C)sm/S),C−(Z)),
[TABLE]
is the complex of presheaves
associated to the bicomplex of presheaves X/S↦F∙(Dˉ∗×X/S),
and singD∗ˉF∙:=eantr(S)∗singDˉ∗F∙∈C−(Z).
We have the inclusion morphism (97)
[TABLE]
which is a morphism in PSh(CorZfs(AnSp(C)sm/S),C−(Z))
denoted the same way S(F∙):F∙→singDˉ∗F∙.
For f:F1∙→F2∙ a morphism PC−(S), we
have the morphism (98)
[TABLE]
which is a morphism in PC−(An,S) denoted the same way
S(f):singDˉ∗F1∙→singDˉ∗F2∙.
For F∙∈PSh(CorZfs(AnSp(C)sm/S),C−(Z)), we have by definition
Tr(S)∗singDˉ∗F∙=singDˉ∗Tr(S)∗F∙ and
Tr(S)∗S(F∙)=S(Tr(S)∗F∙).
Proposition 33**.**
Let S∈AnSp(C)
(i)
(Tr(S)∗,Tr(S)∗):PSh(AnSp(C)sm/S,C−(Z))⇆PShZ(CorZfs(AnSp(C)sm/S),C−(Z))*
is a Quillen adjonction for the usual topology model structures
and a Quillen adjonction for the (D1,et) model structures
(c.f. definition 30 (i) and (ii) respectively).*
(ii)
(ean(S)∗,evar(S)∗):PSh(AnSp(C)sm/S,C−(Z))⇆C−(Z)*
is a Quillen adjonction for the usual topology model structures
and a Quillen adjonction for the (D1,et) model structures
(c.f. definition 30 (i)).*
(iii)
(ean(S)tr∗,evartr(S)∗):PSh(CorZfs(AnSp(C)sm/S),C−(Z))⇆C−(Z)*
is a Quillen adjonction for the usual topology model structures
and a Quillen adjonction for the (D1,et) model structures
(c.f. definition 30 (ii)).*
Proof.
(i): Follows from the fact that Tr(S)∗ derive trivially hence is a right Quillen functor.
(ii): Follows from the fact that evar(S)∗ derive trivially hence is a left Quillen functor.
(iii): Follows from the fact that evartr(S)∗ derive trivially hence is a left Quillen functor.
Point (i) follows immediately from definition of D1 local objects.
Point (ii) follows from point (i).
∎
We immediately deduce the following
Proposition 35**.**
Let f:T→S a morphism in AnSp(C).
(i)
The adjonction
(f∗,f∗):PSh(AnSp(C)sm/S,C−(Z))⇆PSh(AnSp(C)sm/T,C−(Z))
is a Quillen adjonction with respect to the usual topology model structures and the (D1,et) model structures
(ii)
The adjonction
(f∗,f∗):PSh(CorZfs(AnSp(C)sm/S),C−(Z))⇆PSh(CorZfs(AnSp(C)sm/T),C−(Z)),
is a Quillen adjonction with respect to the usual topology model structures and the (D1,et) model structures
Proof.
(i): By [4], f∗ derive trivially for the usual topology model structures.
Now, by proposition 34(ii), f∗ derive trivially for the (D1,usu) model structures.
In particular f∗ is a left Quillen functor.
(ii): Similar to point (i).
∎
Theorem 23**.**
Let S∈AnSp(C)
(i)
For F∙∈PSh(AnSp(C)sm/S,C−(Z)),
singDˉ∗F∙∈PSh(AnSp(C)sm/S,C−(Z))
is D1 local and
the inclusion morphism S(F∙):F∙→singDˉ∗F∙
is an (D1,usu) equivalence.
(ii)
For F∙∈PShZ(CorZfs(AnSp(C)sm/S),C−(Z)),
singDˉ∗F∙∈PShZ(CorZfs(AnSp(C)sm),C−(Z))
is D1 local and
the inclusion morphism S(F∙):F∙→singDˉ∗F∙
is an (D1,usu) equivalence.
Proof.
Similar to the proof of the absolute case.
∎
Theorem 24**.**
Let S∈AnSp(C).
(i)
The adjonction
(Tr(S)∗,Tr(S)∗):PSh(AnSp(C)sm/S,C−(Z))⇆PShZ(CorZfs(AnSp(C)sm/S),C−(Z))
is a Quillen equivalence for the (D1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Tr∗:AnDM−(S,Z)∼AnDA−(S,Z) is it inverse.
(ii)
The adjonction
(ean(S)∗,ean(S)∗):C−(Z)⇆PShZ(AnSm(C),C−(Z))
is a Quillen equivalence for the (I1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Rean(S)∗:AnDA−(S,Z)∼D−(S) is it inverse.
(iii)
The adjonction
(eantr(S)∗,eantr(S)∗):C−(S)⇆PShZ(CorZfs(AnSpsm(C)/S),C−(Z))
is a Quillen equivalence for the (D1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Reantr(S)∗:AnDM−(S,Z)∼D−(S) is it inverse.
Let f:T→S a morphism in AnSp(C).
There is a canonical morphism of functor ϕ(f∗,S) which associate to
F∙∈PShZ(CorZfs(AnSp(C)sm/S),C−(Z))
the morphism
[TABLE]
in PShZ(CorZfs(AnSp(C)sm/T),C−(Z))
given by for Y/T∈AnSp(C)sm/T, the morphism
[TABLE]
given by (h:Y/T→XT/T)↦(h∘pY:Y×Dˉ∗/T→XT/T)
and F∙(pX):(X×Dˉ∗/S)→F∙(X/S).
Let f:T→S a morphism in AnSp(C).
There is also canonical morphism of functor ϕ(f,ean) which associate to
F∙∈PShZ(CorZfs(AnSp(C)sm/S),C−(Z))
the following morphism in C−(T)
[TABLE]
given by the adjonction morphisms and
denoting for simplicity eS=eantr(S) and eT=eantr(T).
Let f:T→S a morphism in AnSp(C).
We denote by ϕ(f∗,ean,S) the morphism of functor, which for
F∙∈PShZ(CorZfs(AnSp(C)sm/S),C−(Z)),
associate the following composition in C−(T)
[TABLE]
By definition, we have:
Proposition 36**.**
Let f:T→S a morphism in AnSp(C). For F∙∈PC−(An,S), the morphism
ϕ(f∗,e,S)(F∙):f∗singDˉ∗F∙→singDˉ∗f∗F∙
in C−(T) is an isomorphism.
Proof.
By definition, the morphism ϕ(f∗,e,S)(F∙)
is given by for To⊂T,
[TABLE]
given by the isomorphism (h:To/T→XT/T)↦(h∘pT:To×Dˉ∗/T→XT/T)
and F∙(pSo):F∙(So×Dˉ∗/S)→F∙(So/S).
∎
We will use in the last subsection the following :
Proposition 37**.**
(i)
Let G1∙,G2∙∈PShZ(CorZfs(AnSp(C)sm/S),C−(Z))
and f:G1∙→G2∙ a morphism. If
[TABLE]
is an equivalence usu local in C−(S),
then f:G1∙→G2∙ is an (D1,usu) local equivalence.
(ii)
Let G1∙,G2∙∈PShZ(CorZfs(AnSp(C)sm/S),C−(Z))
and f:G1∙→G2∙ a morphism. If
–
G1∙* and G2∙ are D1 local,*
–
eantr(S)∗f:ean∗trG1∙→ean∗trG2∙* is an equivalence usu local in C−(S),*
then f:G1∙→G2∙ is an (D1,usu) local equivalence.
Proof.
Similar to the proof of proposition 15. Point (i) follows from theorem 24(iii),
and point (ii) follows from point (i).
∎
4.3 Presheaves and transfers on relative CW complexes
Let S∈Top. The category Top/S is the category whose objects are
X/S=(X,h) with X∈Top and h:X→S is a morphism,
and whose space of morphisms between
X/S=(X,h1) and Y/S=(Y,h2)∈Top/S are the morphism g:X→Y such that h2∘g=h1.
We now restrict to the full subcategory CW⊂Top of CW complexes.
Definition 36**.**
A morphism f:X→Y, X,Y∈CW with X connected is said to be smooth if
for all x∈X, there exist a neighborhood Ux⊂X of x in X and
an open embedding k:Ux↪Ux×RdX such that
f∣Ux=pUx∘k.
Let S∈CW. The category CW/S is the category whose objects are
X/S=(X,h) with X∈CW and h:X→S is a morphism,
and whose space of morphisms between
X/S=(X,h1) and Y/S=(Y,h2)∈CW/S are the morphism g:X→Y such that h2∘g=h1.
We denote by CWsm/S⊂CW/S the full subcategory consisting of the objects
X/S=(X,h) with X∈CW, such that h:X→S is a smooth morphism.
For X/S=(X,h)∈CW/S, and n∈N, we denote by
•
(X×I1/S):=(X×I1,h∘pX)=(X×S(I1×S)/S),
where pX:X×kI1→X is the projection.
•
(X×In/S):=(X×In,h∘pX)=(X×S(In×S)/S),
where pX:X×In→X is the projection.
Definition 37**.**
Let S∈CW.
We define CorΛfs(CWsm/S) to be the category whose objects are the one of CWsm/S
and whose space of morphisms between
X/S and Y/S∈CWsm/S is the free Λ module Zfs/X(X×SY,Λ).
The composition of morphisms is defined similarly then in the absolute case.
We have
•
the additive embedding of categories
Tr(S):Z(CWsm/S)↪CorZfs(CWsm/S)
which gives the corresponding morphism of sites
Tr(S):CorZfs(CWsm/S)→Z(CWsm/S).
•
the inclusion functor
ecw(S):Ouv(S)↪CWsm/S,
which gives the corresponding morphism of sites
ecw(S):CWsm/S→Ouv(S),
•
the inclusion functor
ecwtr(S):=Tr∘ecwOuv(S)↪CorZfs(CWsm/S)
which gives the corresponding morphism of sites
ecwtr(S):=Tr∘ecw:CorZfs(CWsm/S)→Ouv(S).
For each morphism f:T→S in CW, we have
•
the pullback functor
P(f):CW/S→CW/T, P(f)(X/S)=(X×ST/T), P(f)(h)=hT,
which gives the morphism of sites P(f):CW/T→CW/S
•
the pullback functor
P(f):CorZfs(CWsm/S)→CorZfs(CWsm/T),
P(f)(X/S)=(X×ST/T), P(f)(h)=hT,
which gives the morphism of sites
P(f):CorZfs(CWsm/T)→CorZfs(CWsm/S).
For S∈CW, we consider the following two big categories :
•
PSh(CWsm/S,C−(Z))=PShZ(Z(CWsm/S),C−(Z)),
the category of bounded above complexes of presheaves on CWsm/S,
or equivalently additive presheaves on Z(CWsm/S),
sometimes, we will write for short P−(CW,S)=PSh(CWsm/S,C−(Z)),
•
PShZ(CorZfs(CWsm/S),C−(Z)),
the category of bounded above complexes of additive presheaves on CorZfs(CWsm/S)
sometimes, we will write for short PC−(CW,S)=PSh(CorZfs(CWsm/S),C−(Z)),
given by Tr(S):CorZfs(CWsm/S)→Z(CWsm/S),
ecw(S):CWsm/S→Ouv(S) and
ecwtr(S):CorZfs(CWsm/S)→Ouv(S) respectively.
We denote by aet:PShZ(CWsm/S,Ab)→ShZ,et(CWsm/S,Ab)
the etale sheaftification functor.
For X/S∈CWsm/S, we denote by
[TABLE]
the presheaves represented by X. They are usu sheaves.
For X/S=(X,h)∈CW/S with h:X→S non smooth,
[TABLE]
is also an usu sheaf ; of course if h:X→S is not dominant then Ztr(X/S)=0
For a morphism f:T→S in CW we have the adjonctions
given by P(f):CWsm/T→CWsm/S and
P(f):CorZfs(CWsm/T)→CorZfs(CWsm/S),
respectively.
Definition 38**.**
(i)
The projective usual topology model structure on PSh(CWsm/S,C−(Z))
is defined in the similar way of the absolute case (c.f. definition 19(i)).
(ii)
The projective (I1,usu) model structure on the category
PSh(CWsm/S,C−(Z)) is the left Bousfield localization
of the projective usual topology model structure with respect to the class of maps
{Z(X×I1/S)[n]→Z(X/S)[n],X/S∈CWsm/S,n∈Z}.
(iii)
The projective usual topology model structure on PSh(Corfs(CWsm/S),C−(Z))
is defined in the similar way of the absolute case (c.f. definition 19(ii)).
(iv)
The projective (I1,usu) model structure on the category
PShZ(CorZfs(CWsm/S),C−(Z))
is the left Bousfield localization of the projective usual topology model structure with respect to the class of maps
{Ztr(X×I1/S)[n]→Z(X/S)[n],X/S∈CWsm/S,n∈Z}.
Definition 39**.**
Let S∈CW.
(i)
We define
CwDM−(S,Z):=HoI1,usu(PShZ(CorZfs(CWsm/S),C−(Z))),
to be the derived category of (effective) motives, it is
the homotopy category of PSh(CorZfs(CWsm/S),C−(Z))
with respect to the projective (I1,usu) model structure (c.f. definition 38(ii)).
We denote by
[TABLE]
the canonical localization functor.
(ii)
By the same way, we denote
CwDA−(S,Z):=HoI1,usu(PSh(CWsm/S,C−(Z))) (c.f.38(i)) and
[TABLE]
the canonical localization functor.
We now look at an explicit localization functor.
For F∙∈PSh(CWsm/S,Ab) and X/S∈CWsm/S,
we have the complex F(X×I∗/S) associated to the cubical object
F(X×I∗/S) in the category of abelian groups.
•
If F∙∈PSh(CWsm/S,C−(Z)) ,
[TABLE]
is the total complex of presheaves associated to the bicomplex of presheaves X/S↦F∙(I∗×X/S),
and singI∗F∙:=ecw(S)∗singI∗F∙∈C−(Z).
We denote by S(F∙):F∙→singI∗F∙,
[TABLE]
the inclusion morphism of PSh(CWsm/S,C−(Z)) :
For f:F1∙→F2∙ a morphism PSh(CWsm/S,C−(Z)), we denote by
S(f):singI∗F1∙→singI∗F2∙,
the morphism of PSh(CWsm/S,C−(Z)) given by for X/S∈CWsm/S,
[TABLE]
•
If F∙∈PShZ(CorZfs(CWsm/S),C−(Z)),
[TABLE]
is the complex of presheaves
associated to the bicomplex of presheaves X/S↦F∙(I∗×X/S),
and singI∗F∙:=ecwtr(S)∗singI∗F∙∈C−(Z).
We have the inclusion morphism (106)
[TABLE]
which is a morphism in PSh(CorZfs(CWsm/S),C−(Z))
denoted the same way S(F∙):F∙→singI∗F∙.
For f:F1∙→F2∙ a morphism PC−(S), we
have the morphism (107)
[TABLE]
which is a morphism in PC−(S) denoted the same way
S(f):singI∗F1∙→singI∗F2∙.
For F∙∈PSh(CorZfs(CWsm/S),C−(Z)), we have by definition
Tr(S)∗singI∗F∙=singI∗Tr(S)∗F∙ and
Tr(S)∗S(F∙)=S(Tr(S)∗F∙).
We recall the definition of homotopy in the relative setting :
•
Let S,X,Y∈Top and hX:X→S, hY:Y→S two maps.
Let f0:X→Y, f1:X→Y be two maps such that hY∘f0=hX and hY∘f0=hX,
that is such that they define morphisms from X/S=(X,hX) to Y/S=(Y,hY) in Top/S.
We say that f0, f1 are I1 homotopic,
if there exist h:X×I1→Y such that
–
hY∘h=hX∘pX that is h:X×I1/S→Y/S is a morphism in Top/S,
–
f0=h∘(IX×i0) and f1=h∘(IX×i1),
with (IX×i0):X×{0}↪X×I1
and (IX×i1):X×{1}↪X×I1
the inclusions and pX:X×I1→X the projection.
•
Let S∈CW. Let F∙,G∙∈PSh(CW/S,C−(Z)).
We say that two maps ϕ0:F∙→G∙ and ϕ1:F∙→G∙
are I1 homotopic if there exist
ϕ~:F∙→Hom(Z(I1×S),G∙) such that
ϕ0=G∙(i0)∘ϕ~ and ϕ1=G∙(i1)∘ϕ~, where,
–
G∙(i0):Hom(Z(I1),G∙)→G∙ is
induced by i0:{pt}→I1, that is, for X/S∈CW/S,
G∙(i0)(X/S)=G∙(IX×i0):G∙(X×I1/S)→G∙(X/S),
–
G∙(i1):Hom(Z(I1),G∙)→G∙ is
induced by i1:{pt}→I1, that is, for X/S∈CW/S,
G∙(i1)(X/S)=G∙(IX×i1):G∙(X×I1)→G∙(X/S).
Similary to the absolute case, we have the following lemmas :
Lemma 12**.**
Let X/S,Y/S∈CW/S and f0:X/S→Y/S, f1:X/S→Y/S two morphisms.
If f0 and f1 are I1 homotopic, then
•
Z(f0):Z(X/S)→Z(Y/S)* and Z(f1):Z(X/S)→Z(Y)
are I1 homotopic in PSh(CW/S,C−(Z)),*
•
Tr∗Ztr(f0):Tr∗Ztr(X/S)→Tr∗Ztr(Y/S)*
and Tr∗Ztr(f1):Tr∗Ztr(X/S)→Tr∗Ztr(Y/S)
are I1 homotopic in PSh(CW/S,C−(Z)).*
Proof.
Similar to the absolute case.
∎
Lemma 13**.**
Let S∈CW. Let F∙∈PSh(CW/S,C−(Z)).
(i)
Let X/S,Y/S∈CW/S and f0:X/S→Y/S, f1:X/S→Y/S be two morphism.
If f0 and f1 are I1 homotopic, then the maps of complexes
–
singI∗F∙(f0):TotF∙(Y×I∗/S)→TotF∙(X×I∗/S)* and*
–
singI∗F∙(f1):TotF∙(Y×I∗/S)→TotF∙(X×I∗/S)**
induces the same map on homology.
(ii)
Let X/S,Y/S∈CW/S, if f:X→Y is a I1 homotopy equivalence then
[TABLE]
is a quasi-isomorphism of complexes of abelian groups.
(iii)
Let F∙,G∙∈PSh(CW/S,C−(Z)) and
ϕ0:F∙→G∙, ϕ1:F∙→G∙ be two maps.
If ϕ0 and ϕ1 are I1 homotopic, then ϕ0=ϕ1∈CwDA−(S).
(iv)
Let F∙,G∙∈PSh(CW/S,C−(Z)), if
ϕ:F∙→G∙ is a I1 homotopy equivalence then ϕ is a (I1,usu) local equivalence.
Proof.
Similar to the absolute case.
∎
Lemma 14**.**
(i)
A complex of presheaves F∙∈PSh(CorZfs(CWsm/S),C−(Z)) is
I1 local if and only if
Tr(S)∗F∙∈PSh(CWsm/S,C−(Z)) is
I1 local.
(ii)
A morphism ϕ:F∙→G∙ in PSh(CorZfs(CWsm/S),C−(Z)) is
an (I1,usu) local equivalence if and only if
Tr(S)∗ϕ:Tr(S)∗F∙→Tr∗G∙ is an (I1,usu) local equivalence.
Proof.
(i): Similar to the absolute case.
(ii): As in the absolute case the only if part follows from lemma 12(ii) and the if part follows from (i).
∎
Proposition 38**.**
Let S∈CW
(i)
(Tr(S)∗,Tr(S)∗):PSh(CWsm/S,C−(Z))⇆PShZ(CorZfs(CWsm/S),C−(Z))*
is a Quillen adjonction for the etale topology model structures
and a Quillen adjonction for the (I1,et) model structures
(c.f. definition 38 (i) and (ii) respectively).*
(ii)
(ecw(S)∗,ecw(S)∗):PSh(CWsm/S,C−(Z))⇆C−(Z)*
is a Quillen adjonction for the etale topology model structures
and a Quillen adjonction for the (I1,et) model structures
(c.f. definition 38 (i)).*
(iii)
(ecw(S)tr∗,ecwtr(S)∗):PSh(CorZfs(CWsm/S),C−(Z))⇆C−(Z)*
is a Quillen adjonction for the etale topology model structures
and a Quillen adjonction for the (I1,et) model structures
(c.f. definition 38 (ii)).*
Proof.
(i): Follows from the fact that Tr(S)∗ by lemme 14 derive trivially hence is a right Quillen functor.
(ii): Follows from the fact that ecw(S)∗ derive trivially hence is a left Quillen functor.
(iii): Follows from the fact that ecwtr(S)∗ derive trivially hence is a left Quillen functor.
Point (i) follows immediately from definition of I1 local objects.
Point (ii) follows from point (i).
∎
We immediately deduce the following
Proposition 40**.**
Let f:T→S a morphism in CW.
(i)
The adjonction
(f∗,f∗):PSh(CWsm/S,C−(Z))⇆PSh(CWsm/T,C−(Z))
is a Quillen adjonction with respect to the usu model structures and the (I1,usu) model structures
(ii)
The adjonction
(f∗,f∗):PSh(CorZfs(CWsm/S),C−(Z))⇆PSh(CorZfs(CWsm/T),C−(Z)),
is a Quillen adjonction with respect to the usu model structures and the (I1,usu) model structures
Proof.
(i): By [4], f∗ derive trivially for the usual topology model structure.
Now, by proposition 39(ii), f∗ derive trivially for the (I1,usu) model structure.
In particular, f∗ is a left Quillen functor.
(i): Similar to point (i).
∎
We deduce from lemma 13(ii), the point (i) of the following proposition :
Proposition 41**.**
Let S∈CW
(i)
For F∙∈PSh(CWsm/S,C−(Z)), the adjonction morphism
[TABLE]
is an equivalence usu local.
(ii)
For F∙∈PShZ(CorZfs(CWsm/S),C−(Z)), the adjonction morphism
(i): Let X/S∈CWsm/S. Since the question is local,
we can assume, after shrinking X/S,
that X/S=(V×So,pS), with So⊂S an open subset
and V⊂RdX a contractible open subset and
pS:RdX→S the projection.
It then follows from lemma 13(ii).
(ii): It is a particular point of (i), since Tr(S)∗ preserve usu local equivalences.
(iii): Obvious
∎
Proposition 42**.**
Let S∈CW
(i)
The functor
Tr(S)∗:PShZ(CorZfs(CW),C−(Z))→PSh(CW,C−(Z))
derive trivially.
(ii): Let ecw(S)∗K∙→L∙ an usu local equivalence, with L∙ usu fibrant.
Since ecw(S)∗K∙ is usu equivalent to L∙ it suffices to prove that L∙ is I1 local.
Since L∙ is usu fibrant, we have to prove that
[TABLE]
is an equivalence usu local
The proof is now similar to [1] proposition 1.6 etape B.
For F∙∈PSh(CWsm/S,C−(Z)),
singI∗F∙∈CWsm/S,C−(Z))
is I1 local and
the inclusion morphism S(F∙):F∙→singI∗F∙
is an (I1,usu) equivalence.
(ii)
For F∙∈PShZ(CorZfs(CWsm/S),C−(Z)),
singIˉ∗F∙∈PShZ(CorZfs(CWsm/S),C−(Z))
is I1 local and
the inclusion morphism S(F∙):F∙→singI∗F∙
is an (I1,usu) equivalence.
Proof.
(i): As in the absolute case, the fact that singI∗F∙
is a I1 local object follows from proposition 41(i) and proposition 42(ii).
We now prove that S(F∙) is an (I1,usu) local equivalence.
As in the absolute case, it suffice to show that
that for all n∈Z, the morphism
[TABLE]
is an equivalence (I1,usu) local.
For X/S∈CW/S, the morphism
[TABLE]
define an I1 homotopy from IIn×X to 0×IX ;
on the other side
pX∘(0×IX)=IX, with pX:In×X→X the projection.
Thus, F∙(θ1,n)
define an I1 homotopy from F∙(IIn) to
F∙(0) ; on the other side,
F∙(0)∘F∙(pn)=I.
This proves (i).
(ii): As in the absolute case, follows from (i).
∎
Theorem 26**.**
Let S∈CW. Then,
(i)
The adjonction
(ecw(S)∗,ecw(S)∗):C−(Z)⇆PShZ(CWsm/S,C−(Z))
is a Quillen equivalence for the (I1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Recw(S)∗:CwDA−(Z)∼D−(S) is it inverse.
(ii)
The adjonction
(ecwtr(S)∗,ecwtr(S)∗):C−(Z)⇆PShZ(CorZfs(CWsm/S),C−(Z))
is a Quillen equivalence for the (I1,usu) model structures.
That is, the derived functor
[TABLE]
is an isomorphism
and Recwtr(S)∗:CwDM−(Z)∼D−(S) is it inverse.
(iii)
The functor
Tr(S)∗:PSh(CWsm/S,C−(Z))→PShZ(CorZfs(CWsm/S),C−(Z))
induces an isomorphism
and Tr(S)∗:CwDM−(S,Z)∼CwDA−(S,Z).
Proof.
(i): It follows from proposition 41(i) and theorem 25(i).
(ii): It follows from proposition 41(ii) and theorem 25(ii).
It also follows from (i) and (ii).
(iii): It follows from (i) and (ii).
∎
Remark 1**.**
As in the absolute case, for F∙∈PSh(CorZ(CWsm/S),C−(Z)),
ad(Tr∗,Tr∗)(F∙):F∙→Tr∗Tr∗F∙
is an isomorphism in PSh(CorZ(CWsm),C−(Z)) and we can prove that
for X/S∈CWsm/S, the embedding
[TABLE]
in PSh(CWsm/S,C−(Z)) is an equivalence usu local.
We would deduce from this that LTr∗ is the inverse of Tr∗.
We will not do it since we don’t use it.
Let f:T→S a morphism in CW.
There is a canonical morphism of functor ϕ(f∗,S) which associate to
F∙∈PShZ(CorZfs(CWsm/S),C−(Z))
the morphism
[TABLE]
in PShZ(CorZfs(CWsm/T),C−(Z))
given by for Y/T∈CWsm/T, the morphism
[TABLE]
given by (h:Y/T→XT)↦(h∘pY:Y×I∗/T→XT)
and F∙(pX):(X×I∗/S)→F∙(X/S).
Let f:T→S a morphism in CW.
There is also canonical morphism of functor ϕ(f,ecw) which associate to
F∙∈PShZ(CorZfs(CWsm/S),C−(Z))
the morphism in C−(T)
[TABLE]
given by the adjonction morphisms and denoting for simplicity
eS=ecwtr(S), eT=ecwtr(T).
Let f:T→S a morphism in CW.
We denote by ϕ(f∗,ecw,S) the morphism of functor, which for
F∙∈PShZ(CorZfs(CWsm/S),C−(Z)),
associate the following composition in C−(T)
[TABLE]
By definition, we have :
Proposition 43**.**
Let f:T→S a morphism in CW. For F∙∈PC−(CW,S), the morphism in C−(T)ϕ(f∗,e,S)(F∙):f∗singI∗F∙→singI∗f∗F∙
is an isomorphism in C−(T).
Proof.
By definition, the morphism ϕ(f∗,e,S)(F∙)
is given by for To⊂T,
[TABLE]
given by the isomorphism (h:To/T→XT/T)↦(h∘pT:To×I∗/T→XT/T)
and F∙(pSo):F∙(So×I∗/S)→F∙(So/S).
∎
4.4 The relative Betti realisation functor
•
For each S∈Var(C) we have
the analytical functor An(S):Var(C))/S→AnSp(C)/San
given by (V/S)↦Van/San on objects and g↦gan on morphisms,
the analytical functor on transfers
An(S):CorΛfs(Var(C)sm/S)→CorZfs(AnSp(C)sm/San))
given by V/S↦Van/San on objects and Γ↦Γan on morphisms,
•
For each S∈AnSp(C), we have
the forgetful functor Cw(S):AnSp(C)sm/S→CW/Scw
given by W/S↦Wcw/Scw on objects and g↦gcw on morphisms,
and the forgetful functor on transfers
Cw(S):CorZfs(AnSm(C)sm/S)→CorZfs(CWsm/Scw)
given by W/S↦Wcw/Scw on objects and Γ↦Γcw on morphisms,
•
For each S∈Var(C), we have
the composites \widetilde{\mathop{\rm Cw}}\nolimits(S)=\mathop{\rm Cw}\nolimits(S^{an})\circ\mathop{\rm An}\nolimits(S):\mathop{\rm Var}\nolimits(\mathbb{C})/S\to\mathop{\rm CW}\nolimits/S,
given by V/S↦Vcw/Scw on objects and g↦gcw on morphisms, and
\widetilde{\mathop{\rm Cw}}\nolimits(S)=\mathop{\rm Cw}\nolimits(S^{an})\circ\mathop{\rm An}\nolimits(S):\mathop{\rm Cor}\nolimits^{fs}_{\mathbb{Z}}(\mathop{\rm Var}\nolimits(\mathbb{C})^{sm}/S)\to\mathop{\rm Cor}\nolimits^{fs}_{\mathbb{Z}}(\mathop{\rm CW}\nolimits/S),
given by V/S↦Vcw/Scw and Γ↦Γcw.
•
For each S∈Var(C) and S′∈AnSp(C), we have
the embeddings of categories ιan(S′):AnSp(C)sm/S′→AnSp(C)/S′ and
ιvar(S):Var(C)sm/S→Var(C)/S.
By definition, for each S∈Var(C), we have the following commutative diagram of sites DCat(S)
[TABLE]
For T,S∈Var(C) and f:T→S a morphism, the morphism of sites
•
P(f):Var(C)/T→Var(C)/S, P(fan):AnSp(C)/Tan→AnSp(C)/San,
and P(fcw):CW/Tcw→CW/Scw
given by the pullback functor,
•
P(f):CorZfs(Var(C)sm/T)→CorZfs(Var(C)sm/S),
P(fan):CorZfs(AnSp(C)/Tan)→CorZfs(AnSp(C)/San), and
P(fcw):CorZfs(CW/Tcw)→CorZfs(CWsm/Scw)
given by the pullback functor,
derive trivially for the (A1,et) and (I1,usu) model structures.
Proof.
The proof is completely similar to the proof of proposition 25 using
lemma 13 (ii), since for X/S∈AnSp(C)/S,
pX:Xcw×D1→Xcw is a homotopy equivalence in CW/Scw.
∎
Now we have :
•
The 2-functor DM−:Var(k)→TriCat is an homotopic 2-functor is the sense of [4]
(theorem22).
•
The 2-functor D−:Var(C)→TriCat, S∈Var(C)↦D−(Scw).
is an homotopic 2-functor (see [10])
The Betti realisation functor (without transfers) is the composite :
[TABLE]
(ii)
The Betti realization functor with transfers is the composite :
[TABLE]
Since An(S)∗ derive trivially by proposition 44(i)
and and LTr(San)∗:AnDA−(San,Z)→CwDM−(San,Z) is the inverse of Tr(San)∗
(c.f.theorem 24(i)), we have \widetilde{\mathop{\rm Bti}}\nolimits_{0}(S)^{*}=\widetilde{\mathop{\rm Bti}}\nolimits(S)^{*}\circ L\mathop{\rm Tr}\nolimits(S)^{*}.
We have the following :
Theorem 27**.**
[1]**
The Betti realization functor define morphisms of homotopic 2-functors :
•
S∈Var(C)↦(Bti0(S):DA−(S,Z)→D−(S))**
•
S∈Var(C)↦(Bti(S):DM−(S,Z)→D−(S)).
As in the absolute case, we define :
Definition 41**.**
Let S∈Var(C).
(i)
The CW-Betti realization functor (without transfers) is the composite :
[TABLE]
(ii)
The CW-Betti realisation functor with transfers is the composite :
[TABLE]
Similarly, in the relative case, since \widetilde{\mathop{\rm Cw}}\nolimits(S)^{*} derive trivially by proposition 44(ii)
and LTr(Scw)∗:CwDA−(Scw,Z)→CwDM−(Scw,Z) is the inverse of Tr(Scw∗
(c.f.remark 1), we have \widetilde{\mathop{\rm Bti}}\nolimits_{0}(S)^{*}=\widetilde{\mathop{\rm Bti}}\nolimits(S)^{*}\circ L\mathop{\rm Tr}\nolimits(S)^{*}.
We also have the following
Theorem 28**.**
The Betti realization functor define morphisms of homotopic 2-functors :
Let S∈AnSp(C).
For G∙∈PC−(An,S),
W(G∙)(S):Cw(S)∗(singDˉ∗G∙)→singI∗Cw(S)∗G∙
is an equivalence (I1,usu) local in PC−(CW,Scw).
(ii)
Let S∈Var(C).
For F∙∈PC−(S),
\widetilde{W}(S)(F^{\bullet}):\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}(\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}F^{\bullet})\to\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}
is an (I1,usu) local equivalence in PC−(CW,Scw).
For S∈AnSp(C), we define the morphism of functor B(S), by associating
to G∙∈PSh(CorZfs(AnSpsm(C)/S),C−(Z)),
the morphism B(S)(G∙) which is the composite
[TABLE]
in PSh(CorZfs(AnSp(C)sm/S),C−(Z))
We have the following key proposition.
Proposition 46**.**
Let S∈Var(C) and F∙∈PC−(S)
such that D(A1,et)(S)(F∙)∈DM−(S,Z) is a constructible motive. Then,
(i)
e^{tr}_{an}(S)_{*}B(S)(F^{\bullet}):\mathop{\rm sing}\nolimits_{\bar{\mathbb{D}}^{*}}\mathop{\rm An}\nolimits(S)^{*}F^{\bullet}\to\mathop{\rm sing}\nolimits_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}*
is an equivalence usu local in C−(San).*
(ii)
B(S)(F^{\bullet}):\underline{\mathop{\rm sing}\nolimits}_{\bar{\mathbb{D}}^{*}}\mathop{\rm An}\nolimits(S)^{*}F^{\bullet}\to\mathop{\rm Cw}\nolimits(S^{an})_{*}\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}*
is an equivalence (D1,usu) local.*
Proof.
(i): As M=D(A1,et)(S)(F∙) is a constructible motive,
we may assume using induction that there exist X/S,Y/S∈Var(C)sm/S,
X/S=(X,h1),Y/S=(X,h2) such that
[TABLE]
with α∈HomPC−(S)(F(X/S),F(Y/S,p,n)), where we have chosen
k1:Ztr(X/S)→F(X/S) and
k2:Ztr(Y/S)(p)[n]→F(Y/S,p,n)
equivalences (A1,et) local with
F(X/S) and F(X/S,p,n) are A1 local and etale fibrant objects.
Consider the following commutative diagrams in PC−(An,San) :
are equivalences (I1,usu) local by proposition 44 (iii) and theorem 25(ii).
On the other side,
•
ean(S)∗B(Ztr(Xan/San)):singD∗Ztr(Xan/San)→singI∗Ztr(Xcw/Scw)
is an equivalence usu local by proposition 27(i) applied to
[TABLE]
for each s∈S : since h1:X→S is smooth,
is∗Ztr(Xan/San)=Ztr(Xsan)
and is∗Ztr(Xcw/Scw)=Ztr(Xscw),
where is:{s}↪S is the closed embedding.
•
ean(S)∗B(Ztr(Y/S)(p)[n]):singD∗Ztr(Yan/Yan)(p)[n]→singI∗Ztr(Ycw/Scw)[n]
is an equivalence usu local by proposition 27(i) applied to
[TABLE]
for each s∈S : since h2:Y→S is smooth
is∗Ztr(Yan/San)=Ztr(Ysan)
and is∗Ztr(Ycw/Scw)=Ztr(Yscw),
where is:{s}↪S is the closed embedding.
The diagram (1) then shows that
[TABLE]
is an equivalence usu local in C−(San),
and the diagram (2) that
[TABLE]
is equivalence usu local in C−(San).
(ii): Follows from (i)
Let us explain.
•
On the one hand,
–
By theorem 23 (ii),
singDˉ∗An(S)∗F∙
is D1 local.
–
By theorem 16(ii),
\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}
is I1 local.
Hence, \mathop{\rm Cw}\nolimits(S^{an})_{*}\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}
is D1 local, by proposition 25 (ii).
•
On the other hand by (i)
e^{tr}_{an}(S)_{*}(B(S)(F^{\bullet})):\mathop{\rm sing}\nolimits_{\bar{\mathbb{D}}^{*}}\mathop{\rm An}\nolimits(S)^{*}F^{\bullet}\to\mathop{\rm sing}\nolimits_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}
is a quasi isomorphism in C−(S).
The proposition 46 gives the following relative version of theorem 18.
Theorem 29**.**
Let S∈Var(C).
Let M∈DM−(S,Z) is a constructible motive.
(i)
We have \mathop{\rm Bti}\nolimits^{*}M=\widetilde{\mathop{\rm Bti}}\nolimits^{*}M
(ii)
Let M1,M2∈DM−(S,Z) contructible motives.
Let F1∙,F2∙∈PC−(S)
such that Mi=D(A1,et)(S)(Fi∙)∈DM−(S,Z)
for i=1,2. The following diagram is commutative
[TABLE]
Proof.
(i): By definition, Bti(S)∗M=Reantr(S)∗An(S)∗M.
Since, by theorem 23(ii),
•
S(F∙):An(S)∗F∙→singD∗An(S)∗F∙
is an equivalence (D1,usu) local in PC−(An,San) and
•
singD∗An(S)∗F∙ is a D1 local object,
we have, since An(S)∗ derive trivially by proposition 44(i),
Bti∗M=eantr(S)∗(singD∗An(S)∗F∙)=singD∗An(S)∗F∙.
Since
•
B(\mathop{\rm An}\nolimits(S)^{*}F^{\bullet}):\underline{\mathop{\rm sing}\nolimits}_{\mathbb{D}^{*}}\mathop{\rm An}\nolimits(S)^{*}F^{\bullet}\to\mathop{\rm Cw}\nolimits_{*}\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}
is an equivalence (D1,usu) local in PC−(An,San)
by proposition 46 (ii), and
•
\mathop{\rm Cw}\nolimits(S)_{*}\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}
is a D1 local object by theorem 25(ii) and proposition 44(ii),
we have, since \widetilde{\mathop{\rm Cw}}\nolimits(S)^{*} derive trivially by proposition 44(iii),
[TABLE]
By definition, \widetilde{\mathop{\rm Bti}}\nolimits^{*}M=Re^{tr}_{cw*}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}M.
Since, by theorem 25(ii),
•
S(\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}):\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}\to\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet}
is an equivalence (I1,usu) local in PC−(CW,Scw) and
•
\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F^{\bullet} is an I1 local object,
we have
[TABLE]
This proves (i).
(ii):
Let α∈HomDM−(C,Z)(M1,M2).
Consider the commutative diagram in AnDM(S,Z)
[TABLE]
Since
singD∗An∗F1∙ and singD∗An∗F2∙
are D1 local objects by theorem 23(ii)
and since An(S) derive trivially by proposition 44(i),
[TABLE]
Thus,
[TABLE]
Since
\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F_{1}^{\bullet} and
\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F_{2}^{\bullet}
are I1 local objects by theorem 25(ii),
[TABLE]
Thus, since \widetilde{\mathop{\rm Cw}}\nolimits(S)^{*} derive trivially by proposition 44(iii),
[TABLE]
Since
\mathop{\rm Cw}\nolimits(S)_{*}\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F_{1}^{\bullet} and
\mathop{\rm Cw}\nolimits(S)_{*}\underline{\mathop{\rm sing}\nolimits}_{\mathbb{I}^{*}}\widetilde{\mathop{\rm Cw}}\nolimits(S)^{*}F_{2}^{\bullet}
are D1 local objects by theorem 25(ii) and proposition 44 (ii),
[TABLE]
Thus
[TABLE]
∎
Bibliography15
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] J.Ayoub, Note sur les opérations de Grothendieck et la realisation de Betti , Journal of the Institute of Mathematics of Jussieu, Volume 9, Issue 02, April 2010, pp.225-263.
2[2] J.Ayoub, L’algebre de Hopf et le groupe de Galois motiviques d’un corps de caractéristique nulle I , Journal die reine (Crelles Journal), Volume 2014, Issue 693, pp.1-149.
3[3] J.Ayoub, L’algebre de Hopf et le groupe de Galois motiviques d’un corps de caractéristique nulle II , Journal die reine (Crelles Journal), Volume 2014, Issue 693, pp.151-126.
4[4] J.Ayoub, Les six opérations de Grothendieck et le formalisme des cycles évanescents dans le monde motivique I et II , Société de Mathématiques de France, Astérisque, Volume 314 and 315, 2006.
5[5] Eléments de Mathématiques, Topologie générale , Chapitre 1 a 4, 2007.