ScalingStacks

3. A formalism of equivariant motivic sheaves

Mixed ℓ\ell-adic sheaves or mixed Hodge modules are important tools in geometric representations theory. They are upgrades of the categories of ℓ\ell-adic sheaves and derived category of constructible sheaves on a complex variety with analytic topology, respectively. In particular, they are naturally equipped with a notion of weight filtration and an endofunctor (1),(1), called Tate twist, shifting this filtration. Arguments involving these weights can be very powerful.

In this section we will recall the formalism of (equivariant) motivic sheaves which has similar properties but two important technical advantages. First, the aforementioned weight filtration will be replaced by a grading. Secondly, all results work rationally and are hence indepedent of ℓ.\ell.

3.1. Recollections on motivic sheaves

Let kk be a perfect field and pt=Spec⁡(k).\operatorname{pt}=\operatorname{Spec}(k). All varieties are considered to be over k.k.

For a variety XX over kk we consider the triangulated category DM⁡(X,ℚ)\operatorname{DM}(X,\mathbb{Q}) of rational motivic sheaves on X.X.11 1 There are various definitions of DM⁡(X,ℚ)\operatorname{DM}(X,\mathbb{Q}) that are equivalent, see [CD19, Section C.3]. We leave the choice of definition to the reader. In the literature, objects in DM⁡(X,ℚ)\operatorname{DM}(X,\mathbb{Q}) are often referred to as relative motives or simply motives. However, we prefer the term motivic sheaf and reserve the term motive for objects in DM⁡(k,ℚ)\operatorname{DM}(k,\mathbb{Q}). In the special case X=Spec⁡(k),X=\operatorname{Spec}(k), the category DM⁡(k,ℚ)\operatorname{DM}(k,\mathbb{Q}) agrees with Voevodsky’s triangulated category of mixed motives over k.k.

The system of categories DM⁡(X,ℚ)\operatorname{DM}(X,\mathbb{Q}) has remarkable properties, some of which we will recall now.

First, the work of Ayoub [Ayo07a, Ayo07b] and Cisinski–Déglise [CD19] shows that DM⁡(X,ℚ)\operatorname{DM}(X,\mathbb{Q}) can be equipped with a six-functor-formalism which works very similarly as in the setting of ℓ\ell-adic sheaves. Important objects are the motive M⁡(X)\operatorname{M}(X) and motive with compact support Mc⁡(X)\operatorname{M}^{c}(X) of a variety f:X→ptf:X\to\operatorname{pt} which can be expressed in terms of the six functors as

M(X)=f!f!ℚ∈DM(k,ℚ) and Mc(X)=f∗f!ℚ∈DM(k,ℚ).\operatorname{M}(X)=f_{!}f^{!}\mathbb{Q}\in\operatorname{DM}(k,\mathbb{Q})\text{ and }\operatorname{M}^{c}(X)=f_{*}f^{!}\mathbb{Q}\in\operatorname{DM}(k,\mathbb{Q}).

There is an autoequivalence (1)=−⊗ℚ(1)(1)=-\otimes\mathbb{Q}(1) called Tate twist on DM\operatorname{DM} which commutes with the six functors and can be defined by splitting the motive of the projective line as

M⁡(ℙk1)=ℚ⊕ℚ⁡(1)​[2].\operatorname{M}(\mathbb{P}^{1}_{k})=\mathbb{Q}\oplus\mathbb{Q}(1)[2].

For each prime ℓ\ell invertible in kk, there is an ℓ\ell-adic realisation functor to the category D⁡(X,ℚℓ)\operatorname{D}(X,\mathbb{Q}_{\ell}) of ℓ\ell-adic sheaves on XX

(3.1) Realℓ:DM⁡(X,ℚ)→D⁡(X,ℚℓ),\displaystyle\operatorname{Real}_{\ell}:\operatorname{DM}(X,\mathbb{Q})\to\operatorname{D}(X,\mathbb{Q}_{\ell}),

which is compatible with the six functors and the Tate twist, see [Ayo14].

Morphisms in DM⁡(X,ℚ)\operatorname{DM}(X,\mathbb{Q}) are best understood in terms of higher Chow groups, as defined by Bloch [Blo86]. For XX smooth, there is a natural isomorphism

HomDM⁡(X,ℚ)⁡(ℚX,ℚX​(m)​[n])\displaystyle\operatorname{Hom}_{\operatorname{DM}(X,\mathbb{Q})}(\mathbb{Q}_{X},\mathbb{Q}_{X}(m)[n]) ≅HomDM⁡(k,ℚ)⁡(M⁡(X),ℚ⁡(m)​[n])\displaystyle\cong\operatorname{Hom}_{\operatorname{DM}(k,\mathbb{Q})}(\operatorname{M}(X),\mathbb{Q}(m)[n])
(3.2) ≅CHm⁡(X,2​m−n)ℚ,\displaystyle\cong\operatorname{CH}^{m}(X,2m-n)_{\mathbb{Q}},

where ℚX\mathbb{Q}_{X} denotes the tensor unit and CH\operatorname{CH} a higher Chow group. In particular, for n=2​mn=2m one obtains the usual Chow group of codimension-mm cycles CHm⁡(X,0)ℚ=CHm⁡(X)ℚ.\operatorname{CH}^{m}(X,0)_{\mathbb{Q}}=\operatorname{CH}^{m}(X)_{\mathbb{Q}}. In this case the realisation functor Realℓ\operatorname{Real}_{\ell} yields the cycle class map to ℓ\ell-adic cohomology

CHm⁡(X)ℚ→Hét2​m​(X,ℚℓ​(m)).\operatorname{CH}^{m}(X)_{\mathbb{Q}}\to H^{2m}_{\text{\'{e}t}}(X,\mathbb{Q}_{\ell}(m)).

For X=ptX=\operatorname{pt} and k=𝔽qk=\mathbb{F}_{q} or k=𝔽¯pk=\overline{\mathbb{F}}_{p} these hom-groups are particularly simple:

0MXR

Lemma 3.1. Let k=𝔽qk=\mathbb{F}_{q} or k=𝔽¯pk=\overline{\mathbb{F}}_{p}, then

HomDM⁡(k,ℚ)⁡(ℚ,ℚ⁡(m)​[n])={ℚ for ​n=m=0​ and0otherwise.\displaystyle\operatorname{Hom}_{\operatorname{DM}(k,\mathbb{Q})}(\mathbb{Q},\mathbb{Q}(m)[n])=\left\{\begin{array}[]{cl}\mathbb{Q}&\text{ for }n=m=0\text{ and}\\ 0&\text{otherwise.}\\ \end{array}\right.
0MXS

Proof. First, note that the higher KK-theory of finite fields Ki​(Spec⁡(𝔽q))K_{i}(\operatorname{Spec}(\mathbb{F}_{q})) is torsion for i>0i>0 by [Qui72, Theorem 8]. The same is true for the algebraic closure, since KK-theory commutes with filtered colimits. Hence, Ki​(k)ℚ=0K_{i}(k)_{\mathbb{Q}}=0 for i>0.i>0. By the Riemann–Roch theorem for rational higher Chow groups, see [Blo86, Theorem 9.1], CHm⁡(k,2​m−n)ℚ\operatorname{CH}^{m}(k,2m-n)_{\mathbb{Q}} is a direct summand of K2​m−n​(k)ℚ.K_{2m-n}(k)_{\mathbb{Q}}. The statement follows from (3.2). ∎

In the following the purity property in (3.1) will be crucial. For this reason, we will from now on restrict to the case k=𝔽¯p.k=\overline{\mathbb{F}}_{p}.

3.2. Mixed Tate motives

We will now recall the definition of the category of pure/mixed Tate motives, see [Lev05], which will play for us the role of a graded version of the derived category of sheaves on the point.

0MXT

Definition 3.2. The category of mixed Tate motives22 2 In the literature, there are many notations for the category of mixed Tate motives. For example, TDM\operatorname{TDM} in [HK06], MTDer\operatorname{MTDer} in [SW18, SVW18], DMT\operatorname{DMT} in [Spi16] or DTM\operatorname{DTM} in [Lev05]. is the subcategory

DTM⁡(k,ℚ)=⟨ℚ⟩≅,⨭,Δ,(±1)⊂DM⁡(k,ℚ)\operatorname{DTM}(k,\mathbb{Q})=\langle\mathbb{Q}\rangle_{\cong,\inplus,\Delta,(\pm 1)}\subset\operatorname{DM}(k,\mathbb{Q})

generated by ℚ\mathbb{Q} under isomorphism, direct summands, Tate twists and triangles. The category of pure Tate motives is the category generated by objects ℚ​(n)​[2​n]\mathbb{Q}(n)[2n]

DTM(k,ℚ)w=0=⟨ℚ(n)[2n]∣n∈ℤ⟩≅,⨭,⊕⊂DTM(k,ℚ)\operatorname{DTM}(k,\mathbb{Q})^{w=0}=\langle\mathbb{Q}(n)[2n]\mid n\in\mathbb{Z}\rangle_{\cong,\inplus,\oplus}\subset\operatorname{DTM}(k,\mathbb{Q})

with respect to isomorphism, direct summands and finite direct sums.

For example, the decompostion of a projective space into affine spaces

ℙn=𝔸0⊎𝔸1⊎⋯⊎𝔸n\mathbb{P}^{n}=\mathbb{A}^{0}\uplus\mathbb{A}^{1}\uplus\dots\uplus\mathbb{A}^{n}

induces a decomposition of its motive into a direct sum of Tate motives

M⁡(ℙn)=ℚ⊕ℚ⁡(1)​[2]⊕⋯⊕ℚ⁡(n)​[2​n]\operatorname{M}(\mathbb{P}^{n})=\mathbb{Q}\oplus\mathbb{Q}(1)[2]\oplus\dots\oplus\mathbb{Q}(n)[2n]

which shows that M⁡(ℙn)\operatorname{M}(\mathbb{P}^{n}) is pure Tate.

We will use the following generalization.

0MXU

Definition 3.3. A partition of a variety XX into subvarieties X1,…,XnX_{1},\dots,X_{n} (called strata) is an affine paving, if X≤k=⋃i=1,…,kXiX_{\leq k}=\bigcup_{i=1,\dots,k}X_{i} is closed in XX for all 1≤l≤n1\leq l\leq n and each stratum XiX_{i} is isomorphic to an affine space 𝔸n.\mathbb{A}^{n}.

0MXV

Proposition 3.4. Let XX be a variety that admits an affine paving. Then the motive with compact support Mc⁡(X)∈DTM⁡(k,ℚ)w=0\operatorname{M}^{c}(X)\in\operatorname{DTM}(k,\mathbb{Q})^{w=0} is pure Tate.

0MXW

Proof. This follows from the localisation sequence and Mc⁡(𝔸n)=ℚ⁡(n)​[2​n].\operatorname{M}^{c}(\mathbb{A}^{n})=\mathbb{Q}(n)[2n]. A proof can be found in [Ebe21, Lemma 2.3(10)]. ∎

As the notation suggests, the category of pure Tate motives is the heart of a weight structure:

0MXX

Proposition 3.5. The category DTM⁡(k,ℚ)w=0\operatorname{DTM}(k,\mathbb{Q})^{w=0} is the heart of a weight structure ww on DTM⁡(k,ℚ).\operatorname{DTM}(k,\mathbb{Q}).

0MXY

Proof. By (3.2) the collection of objects in the DTM⁡(k,ℚ)w=0\operatorname{DTM}(k,\mathbb{Q})^{w=0} is negative. Moreover DTM⁡(k,ℚ)\operatorname{DTM}(k,\mathbb{Q}) is idempotent closed and generated by DTM⁡(k,ℚ)w=0\operatorname{DTM}(k,\mathbb{Q})^{w=0}. Now the statement follows from Proposition 2.4. ∎

Using that k=𝔽¯pk=\overline{\mathbb{F}}_{p} we obtain a simple decription of the category of Tate motives:

0MXZ

Proposition 3.6. The weight complex functor yields an equivalence of categories

t:DTM⁡(k,ℚ)→Kb⁡(DTM⁡(k,ℚ)w=0)≅Db⁡(ℚ​−modℤ)t:\operatorname{DTM}(k,\mathbb{Q})\to\operatorname{K}^{b}(\operatorname{DTM}(k,\mathbb{Q})^{w=0})\cong\operatorname{D}^{b}(\mathbb{Q}\operatorname{-mod}^{\mathbb{Z}})

where ℚ​(1)​[2]\mathbb{Q}(1)[2] corresponds to the one-dimensional vector space in degree one ℚ​⟨1⟩\mathbb{Q}\langle 1\rangle in the category ℚ​−modℤ\mathbb{Q}\operatorname{-mod}^{\mathbb{Z}} of graded finite-dimensional vector spaces over ℚ.\mathbb{Q}.

0MY0

Proof. Lemma 3.1 implies that DTM⁡(k,ℚ)w=0\operatorname{DTM}(k,\mathbb{Q})^{w=0} is tilting in DTM⁡(k,ℚ).\operatorname{DTM}(k,\mathbb{Q}). Further, DM\operatorname{DM} admits an enhancement as stable ∞\infty-category. The statement follows from Corollary 2.16. ∎

We note that the weight structure on mixed Tate motives considered here is just the shadow of the Chow weight structure on DMg​m⁡(k,ℚ),\operatorname{DM}^{gm}(k,\mathbb{Q}), see Example 2.2.

3.3. Equivariant motivic sheaves

Soergel–Virk–Wendt [SVW18] introduced an equivariant version of the above formalism. For a variety XX with an action of a linear algebraic group GG they define the category DMG⁡(X,ℚ)\operatorname{DM}_{G}(X,\mathbb{Q}) of GG-equivariant motivic sheaves on X.X. This system of categories still carries a six-functor-formalism33 3 For technical reasons, Soergel–Virk–Wendt construct a six-functor-formalism for the full subcategory DMG+⁡(X,ℚ)⊂DMG⁡(X,ℚ)\operatorname{DM}^{+}_{G}(X,\mathbb{Q})\subset\operatorname{DM}_{G}(X,\mathbb{Q}) of objects that are bounded below with respect to the homotopy tt-structure, see [SVW18, Section I.6]. All objects that we consider here are automatically in DMG+⁡(X,ℚ)\operatorname{DM}^{+}_{G}(X,\mathbb{Q}) and we simply ignore this technicality. and induction/restriction functors changing the group.

Similarly to the non-equivariant case, there is a realization functor to the equivariant derived category of ℓ\ell-adic sheaves DG⁡(X,ℚℓ)\operatorname{D}_{G}(X,\mathbb{Q}_{\ell}) of Bernstein–Lunts [BL94]

Realℓ:DMG⁡(X,ℚ)→DG⁡(X,ℚℓ).\operatorname{Real}_{\ell}:\operatorname{DM}_{G}(X,\mathbb{Q})\to\operatorname{D}_{G}(X,\mathbb{Q}_{\ell}).

Moreover, for XX smooth there is a natural isomorphism

(3.5) HomDMG⁡(X,ℚ)⁡(ℚX,ℚX​(m)​[n])\displaystyle\operatorname{Hom}_{\operatorname{DM}_{G}(X,\mathbb{Q})}(\mathbb{Q}_{X},\mathbb{Q}_{X}(m)[n]) ≅CHGm​(X,2​m−n)ℚ\displaystyle\cong\operatorname{CH}^{m}_{G}(X,2m-n)_{\mathbb{Q}}

to the equivariant higher Chow groups as defined by Totaro [Tot99] and Edidin–Graham [EG98].

There is a forgetful functor from equivariant to non-equivariant motivic sheaves

For:DMG⁡(X,ℚ)→DM⁡(X,ℚ)\operatorname{For}:\operatorname{DM}_{G}(X,\mathbb{Q})\to\operatorname{DM}(X,\mathbb{Q})

commuting with the six functors.

An important property of equivariant (motivic) sheaves is the induction equivalence, which allows to describe GG-equivariant motivic sheaves on a GG-orbit G/HG/H in terms of HH-equivariant motives on a point:

0MY1

Proposition 3.7. Let i:H↪Gi:H\hookrightarrow G be a closed subgroup. Denote by s:X→G×HX,x↦[e,x].s:X\to G\times_{H}X,x\mapsto[e,x]. If the anti-diagonal action of HH on G×XG\times X is free then there is an equivalence of categories

(3.6) (i,s)∗:DMG⁡(G×HX)→∼DMH⁡(X).\displaystyle(i,s)^{*}:\operatorname{DM}_{G}(G\times_{H}X)\stackrel{{\scriptstyle\sim}}{{\to}}\operatorname{DM}_{H}(X).
0MY2

Proof. See [SVW18, Proposition I.7.4]. ∎

0MY3

Remark 3.8. In [SVW18] the language of derivators is used and it is shown that DMG⁡(X,ℚ)\operatorname{DM}_{G}(X,\mathbb{Q}) admits an enhancements as stable derivator. In [RS20], [RS21a] and [RS21b] Richarz–Scholbach provide a similar construction in the language of ∞\infty-categories which shows that DMG⁡(X,ℚ)\operatorname{DM}_{G}(X,\mathbb{Q}) admits an enhancement as stable ∞\infty-category.

3.4. Equivariant mixed Tate motives

There is also an equivariant version of the category of mixed Tate motives defined in Section 3.2. For the remainder of the section let k=𝔽¯p.k=\overline{\mathbb{F}}_{p}.

0MY4

Definition 3.9. The category of equivariant mixed Tate motives on a point

DTMG⁡(k,ℚ)⊂DMG⁡(k,ℚ)\operatorname{DTM}_{G}(k,\mathbb{Q})\subset\operatorname{DM}_{G}(k,\mathbb{Q})

is the full subcategory of objects MM such that For⁡(M)∈DTM⁡(k,ℚ),\operatorname{For}(M)\in\operatorname{DTM}(k,\mathbb{Q}),

We now explain how the explicit description of this category in [SVW18, Theorem II.3.1] can be obtained using the weight complex functor.

0MY5

Proposition 3.10. There is a weight structure ww on DTMG⁡(k,ℚ)\operatorname{DTM}_{G}(k,\mathbb{Q}) such that

M∈DTMG⁡(k,ℚ)w≥0\displaystyle M\in\operatorname{DTM}_{G}(k,\mathbb{Q})^{w\geq 0} ⇔For⁡(M)∈DTM⁡(k,ℚ)w≥0​ and\displaystyle\iff\operatorname{For}(M)\in\operatorname{DTM}(k,\mathbb{Q})^{w\geq 0}\text{ and }
M∈DTMG⁡(k,ℚ)w≤0\displaystyle M\in\operatorname{DTM}_{G}(k,\mathbb{Q})^{w\leq 0} ⇔For⁡(M)∈DTM⁡(k,ℚ)w≤0.\displaystyle\iff\operatorname{For}(M)\in\operatorname{DTM}(k,\mathbb{Q})^{w\leq 0}.
0MY6

Proof. This is [SVW18, Proposition II.4.10]. ∎

Denote by G0⊂GG^{0}\subset G the connected component of the identity. Then one can consider the object IndG0G⁡(ℚ)∈DMG⁡(k)\operatorname{Ind}^{G}_{G^{0}}(\mathbb{Q})\in\operatorname{DM}_{G}(k) which plays the role of the local system on B​GBG with free action of the component group π0​(G)=G/G0.\pi_{0}(G)=G/G^{0}. The heart of ww admits the following explicit description by [SVW18, Proposition II.4.5].

0MY7

Proposition 3.11. The heart of the weight structure on DTMG⁡(k,ℚ)\operatorname{DTM}_{G}(k,\mathbb{Q}) is generated by the objects IndG0G⁡(ℚ)​(n)​[2​n]\operatorname{Ind}^{G}_{G^{0}}(\mathbb{Q})(n)[2n] with respect to isomorphism, direct summands and finite direct sums

DTMG(k,ℚ)w=0=⟨IndG0G(ℚ)(n)[2n]∣n∈ℤ⟩≅,⨭,⊕⊂DTMG(k,ℚ).\operatorname{DTM}_{G}(k,\mathbb{Q})^{w=0}=\langle\operatorname{Ind}^{G}_{G^{0}}(\mathbb{Q})(n)[2n]\mid n\in\mathbb{Z}\rangle_{\cong,\inplus,\oplus}\subset\operatorname{DTM}_{G}(k,\mathbb{Q}).

For the moment, assume that G=G0G=G_{0} is connected. By [SVW18, Proposition I.7.6] the categories DMG\operatorname{DM}_{G} just depend on the quotient G/Ru​(G)G/R_{u}(G) of GG by its unipotent radical. Hence, we can assume that GG is reductive. Denote by TT a maximal torus in GG and by W=NG​(T)/TW=N_{G}(T)/T the Weyl group. Let S=Sym⁡(X​(T)ℚ)S=\operatorname{Sym}(X(T)_{\mathbb{Q}}) be the symmetric algebra of the rationalized character lattice of T.T. The algebra SS is isomorphic to a polynomial ring in rank⁡(T)\operatorname{rank}(T) many variables and graded where we put X⁡(T)X(T) in degree one.

Since we are only considering rational coefficients, the TT- and GG-equivariant Chow rings agree with equivariant cohomology rings. In particular, the Chern class map induces isomorphisms of graded algebras

(3.7) S≅CHT∙​(k)ℚ​ and ​SW≅CHT∙​(k)ℚW≅CHG∙​(k)ℚ,\displaystyle S\cong\operatorname{CH}^{\bullet}_{T}(k)_{\mathbb{Q}}\text{ and }S^{W}\cong\operatorname{CH}^{\bullet}_{T}(k)_{\mathbb{Q}}^{W}\cong\operatorname{CH}^{\bullet}_{G}(k)_{\mathbb{Q}},

see [Tot99] or [EG98, Section 3.2]. Similarly, for higher Chow groups

(3.8) CHG∙​(k,i)ℚ≅(CHT∙​(k,i)ℚ)W≅(S⊗CH∙⁡(k,i)ℚ)W\displaystyle\operatorname{CH}^{\bullet}_{G}(k,i)_{\mathbb{Q}}\cong(\operatorname{CH}^{\bullet}_{T}(k,i)_{\mathbb{Q}})^{W}\cong(S\otimes\operatorname{CH}^{\bullet}(k,i)_{\mathbb{Q}})^{W}

using [Kri17, Theorem 1.5] and [Kri13, Theorem 5.7].

If GG is not connected, we consider the extension algebra

E\displaystyle E =⨁n∈ℤHomDMG⁡(k,ℚ)⁡(IndG0G⁡(ℚ),IndG0G⁡(ℚ)​(n)​[2​n])\displaystyle=\bigoplus_{n\in\mathbb{Z}}\operatorname{Hom}_{\operatorname{DM}_{G}(k,\mathbb{Q})}(\operatorname{Ind}^{G}_{G^{0}}(\mathbb{Q}),\operatorname{Ind}^{G}_{G^{0}}(\mathbb{Q})(n)[2n])

which can be thought of as the Chow ring of B​GBG with coefficients in the local system given by the regular representation of π0​(G).\pi_{0}(G). For an explicit description denote by SW⋉ℚ⁡[π0​(G)]S^{W}\ltimes\mathbb{Q}[\pi_{0}(G)] the twisted group algebra, see [SVW18, A.2.3], which as a graded vector space is just the tensor product SW⊗ℚ⁡[π0​(G)]S^{W}\otimes\mathbb{Q}[\pi_{0}(G)] with ℚ​[π0​(G)]\mathbb{Q}[\pi_{0}(G)] concentrated in degree zero.

0MY8

Proposition 3.12. There is a natural isomorphism E≅SW⋉ℚ⁡[π0​(G)].E\cong S^{W}\ltimes\mathbb{Q}[\pi_{0}(G)].

0MY9

Proof. If GG is connected this is (3.7). Otherwise the statements follow from a transfer argument for finite group torsors, see [SVW18, Section A.2.2]. ∎

0MYA

Proposition 3.13. The collection of objects {IndG0G⁡(ℚ)​(n)​[2​n]∣n∈ℤ}\{\operatorname{Ind}^{G}_{G^{0}}(\mathbb{Q})(n)[2n]\mid n\in\mathbb{Z}\} is tilting.

0MYB

Proof. Assume first that GG is connected. By (3.7) it suffices to show that the higher Chow groups in (3.7) vanish for i≠0.i\neq 0. This holds by Lemma  3.1 using k=𝔽¯p.k=\overline{\mathbb{F}}_{p}. Again, the statement for GG not connected follows from [SVW18, Section A.2.2]. ∎

With Corollary 2.16 we obtain the following explicit description of DTMG⁡(k,ℚ).\operatorname{DTM}_{G}(k,\mathbb{Q}).

0MYC

Theorem 3.14. The weight complex functor induces an equivalence

t:DTMG⁡(k,ℚ)→Dperfℤ⁡(SW⋉ℚ⁡[π0​(G)]).t:\operatorname{DTM}_{G}(k,\mathbb{Q})\to\operatorname{D}^{\mathbb{Z}}_{\operatorname{perf}}(S^{W}\ltimes\mathbb{Q}[\pi_{0}(G)]).

Since SW⋉ℚ⁡[π0​(G)]S^{W}\ltimes\mathbb{Q}[\pi_{0}(G)] has finite cohomological dimension, Dperfℤ⁡(SW⋉ℚ⁡[π0​(G)])\operatorname{D}^{\mathbb{Z}}_{\operatorname{perf}}(S^{W}\ltimes\mathbb{Q}[\pi_{0}(G)]) is just the bounded derived category of finitely generated graded modules. A similar result was shown in [SVW18, Theorem II.3.1] using slightly different arguments.

3.5. Gradings

Let ℓ≠p\ell\neq p be a prime. The categories of equivariant mixed Tate motives DTMG⁡(k)\operatorname{DTM}_{G}(k) can be regarded as graded versions of the categories of equivariant sheaves DG⁡(pt,ℚℓ)\operatorname{D}_{G}(\operatorname{pt},\mathbb{Q}_{\ell}) defined by Bernstein–Lunts [BL94]. Under the realisation functor Realℓ,\operatorname{Real}_{\ell}, see (3.1), the Tate motive ℚ⁡(1)\mathbb{Q}(1) gets mapped to the Tate module

Realℓ⁡(ℚ⁡(1))=ℚℓ​(1)=lim←⁡μℓn⊗ℤℓℚℓ\operatorname{Real}_{\ell}(\mathbb{Q}(1))=\mathbb{Q}_{\ell}(1)=\varprojlim\mu_{\ell^{n}}\otimes_{\mathbb{Z}_{\ell}}\mathbb{Q}_{\ell}

which can be identified with ℚℓ\mathbb{Q}_{\ell} by choosing a compatible system of ℓn\ell^{n}-th roots of unity in k=𝔽¯p.k=\overline{\mathbb{F}}_{p}. This induces a natural equivalence of functors

(3.9) Realℓ∘(1)→Realℓ.\displaystyle\operatorname{Real}_{\ell}\circ(1)\to\operatorname{Real}_{\ell}.

Hence, intuitively, the Tate twist (1)(1) can be regarded as a shift of grading and Realℓ\operatorname{Real}_{\ell} as a functor forgetting the grading. Restricted to mixed Tate motives, the functor Realℓ\operatorname{Real}_{\ell} becomes a degrading functor in the sense of [BGS96, Section 4.3].

0MYD

Proposition 3.15. The equivalence in (3.9) induces an isomorphism

⨁n∈ℤHomDMG⁡(k,ℚℓ)⁡(M,N⁡(n))→HomDG⁡(k,ℚℓ)⁡(Realℓ⁡(M),Realℓ⁡(N))\bigoplus_{n\in\mathbb{Z}}\operatorname{Hom}_{\operatorname{DM}_{G}(k,\mathbb{Q}_{\ell})}(M,N(n))\to\operatorname{Hom}_{\operatorname{D}_{G}(k,\mathbb{Q}_{\ell})}(\operatorname{Real}_{\ell}(M),\operatorname{Real}_{\ell}(N))

for M,N∈DTMG⁡(k).M,N\in\operatorname{DTM}_{G}(k). If M,N∈DTMG⁡(k)w=0M,N\in\operatorname{DTM}_{G}(k)^{w=0} then all summands for n≠0n\neq 0 vanish and the functor Realℓ\operatorname{Real}_{\ell} gives an isomorphism

HomDMG⁡(k,ℚℓ)⁡(M,N)→HomDG⁡(k,ℚℓ)⁡(Realℓ⁡(M),Realℓ⁡(N)).\operatorname{Hom}_{\operatorname{DM}_{G}(k,\mathbb{Q}_{\ell})}(M,N)\to\operatorname{Hom}_{\operatorname{D}_{G}(k,\mathbb{Q}_{\ell})}(\operatorname{Real}_{\ell}(M),\operatorname{Real}_{\ell}(N)).
0MYE

Proof. First, by Proposition 3.11 and induction it suffices to show the statement for objects of the form IndG0G​ℚ​(n)​[m].\operatorname{Ind}_{G_{0}}^{G}\mathbb{Q}(n)[m]. If GG is connected the homomorphims of objects of the form IndG0G⁡ℚ⁡(n)​[m]=ℚ⁡(n)​[m]\operatorname{Ind}_{G_{0}}^{G}\mathbb{Q}(n)[m]=\mathbb{Q}(n)[m] are described in terms of SWS^{W} in both DMG⁡(k,ℚ),\operatorname{DM}_{G}(k,\mathbb{Q}), see Proposition 3.12, and in DG⁡(pt,ℚℓ),\operatorname{D}_{G}(\operatorname{pt},\mathbb{Q}_{\ell}), see [BL94, Section 13.10], and the statement is easily seen to be true. The case that GG is not connected can be handled as described in [SVW18, Theorem A.2.8]. ∎

3.6. Pointwise Tate

We need a notion of local purity for motivic sheaves on a variety X.X. That is, we want to consider motivic sheaves whose restriction to every point is pure or mixed Tate:

0MYF

Definition 3.16. Let ?∈{∗,!}.?\in\{*,!\}. An object M∈DMG⁡(X)M\in\operatorname{DM}_{G}(X) is called ??-pointwise mixed Tate or ??-pointwise pure Tate, respectively, if for each point ix:pt→Xi_{x}:\operatorname{pt}\to X

ix?​For​M∈DTM⁡(k,ℚ)​ or ​ix?​For​M∈DTM⁡(k,ℚ)w=0​, respectively.\displaystyle i_{x}^{?}\operatorname{For}M\in\operatorname{DTM}(k,\mathbb{Q})\text{ or }i_{x}^{?}\operatorname{For}M\in\operatorname{DTM}(k,\mathbb{Q})^{w=0}\text{, respectively.}

Here the functor ix?​Fori_{x}^{?}\operatorname{For} is the composition

DMG⁡(X,ℚ)→ForDM⁡(X,ℚ)→ix?DM⁡(k,ℚ).\operatorname{DM}_{G}(X,\mathbb{Q})\stackrel{{\scriptstyle\operatorname{For}}}{{\to}}\operatorname{DM}(X,\mathbb{Q})\stackrel{{\scriptstyle i_{x}^{?}}}{{\to}}\operatorname{DM}(k,\mathbb{Q}).

The object MM is called pointwise mixed (pure) Tate if it is ∗*- and !!-pointwise mixed (pure) Tate.

It is sometimes convenient to work with the following equivalent orbitwise definition.

0MYG

Proposition 3.17. Let ?∈{∗,!}?\in\{*,!\} and M∈DMG⁡(X).M\in\operatorname{DM}_{G}(X). Then MM is ??-pointwise mixed or pure Tate if and only if for each orbit 𝒪↪X\mathcal{O}\hookrightarrow X

(i,s)∗​j?​M∈DTMH⁡(k,ℚ)​ or ​(i,s)∗​j?​M∈DTMH⁡(k,ℚ)w=0​, respectively.\displaystyle(i,s)^{*}j^{?}M\in\operatorname{DTM}_{H}(k,\mathbb{Q})\text{ or }(i,s)^{*}j^{?}M\in\operatorname{DTM}_{H}(k,\mathbb{Q})^{w=0}\text{, respectively.}

Here j:G/H≅𝒪↪Xj:G/H\cong\mathcal{O}\hookrightarrow X and the functor (i,s)∗​j?(i,s)^{*}j^{?} is the composition

DMG⁡(X,ℚ)→j?DMG⁡(G/H,ℚ)→(i,s)∗DMH⁡(pt,ℚ)\operatorname{DM}_{G}(X,\mathbb{Q})\stackrel{{\scriptstyle j^{?}}}{{\to}}\operatorname{DM}_{G}(G/H,\mathbb{Q})\stackrel{{\scriptstyle(i,s)^{*}}}{{\to}}\operatorname{DM}_{H}(\operatorname{pt},\mathbb{Q})

of pullback to the orbit and the induction equivalence (3.6).

0MYH

Remark 3.18. In [SVW18] this equivalent orbitwise definition is used.

Objects that are pointwise pure Tate have remarkable properties. They behave very similarly to pure Tate objects on a point, particulary if there are only finitely many GG-orbits. They satisfy the following extension vanishing:

0MYI

Proposition 3.19. Assume that the GG-action on XX has finitely many orbits. Let M,N∈DMG⁡(X)M,N\in\operatorname{DM}_{G}(X) be ∗*- and !!-pointwise pure Tate. Then HomDMG⁡(X)⁡(M,N⁡[n])=0\operatorname{Hom}_{\operatorname{DM}_{G}(X)}(M,N[n])=0 for all n≠0.n\neq 0.

0MYJ

Proof. The statement can be shown by an induction on the number of orbits. Denote by j:G/H≅𝒪↪Xj:G/H\cong\mathcal{O}\hookrightarrow X and i:Z=X\𝒪→Xi:Z=X\backslash\mathcal{O}\to X the inclusion of an open orbit 𝒪\mathcal{O} and its closed complement Z.Z. Then the localisation triangle induces an exact sequence

HomDMG⁡(Z)(i∗M,i!N[n]){\lx@inpgf@ignorespaces\operatorname{Hom}_{\operatorname{DM}_{G}(Z)}(i^{*}M,i^{!}N[n])}HomDMG⁡(X)⁡(M,N⁡[n]){\lx@inpgf@ignorespaces\operatorname{Hom}_{\operatorname{DM}_{G}(X)}(M,N[n])}HomDMG⁡(𝒪)(j∗M,j!N[n]).{\lx@inpgf@ignorespaces\operatorname{Hom}_{\operatorname{DM}_{G}(\mathcal{O})}(j^{*}M,j^{!}N[n]).}

Since i∗​Mi^{*}M and i!Ni^{!}N are ∗*- and !!-pointwise pure Tate, respectively, the first term of the sequence vanishes by induction. The last term vanishes since by assumption j∗​Mj^{*}M and j!Nj^{!}N correspond to objects in DTMH⁡(k,ℚ)w=0\operatorname{DTM}_{H}(k,\mathbb{Q})^{w=0} via the induction equivalence (3.6) and thus have no non-trivial extension by Proposition 3.12. See [SVW18, Corollary II.4.19] for a similar proof. ∎

Restricted to pointwise mixed Tate objects Realℓ\operatorname{Real}_{\ell} is a degrading functor after passing to ℚℓ\mathbb{Q}_{\ell}-coefficients.

0MYK

Proposition 3.20. Assume that the GG-action on XX has finitely many orbits. Let ℓ≠p\ell\neq p be a prime and M,N∈DMG⁡(X).M,N\in\operatorname{DM}_{G}(X). Then the natural isomorphisms Realℓ∘(1)→Realℓ,\operatorname{Real}_{\ell}\circ(1)\to\operatorname{Real}_{\ell}, see (3.9), induces isomorphisms

⨁n∈ℤHomDMG⁡(k,ℚℓ)⁡(M,N⁡(n))→∼HomDG⁡(k,ℚℓ)⁡(Realℓ⁡(M),Realℓ⁡(N))\bigoplus_{n\in\mathbb{Z}}\operatorname{Hom}_{\operatorname{DM}_{G}(k,\mathbb{Q}_{\ell})}(M,N(n))\stackrel{{\scriptstyle\sim}}{{\to}}\operatorname{Hom}_{\operatorname{D}_{G}(k,\mathbb{Q}_{\ell})}(\operatorname{Real}_{\ell}(M),\operatorname{Real}_{\ell}(N))

if M,NM,N are ∗*- and !!-pointwise mixed Tate, respectively, and

HomDMG⁡(X,ℚℓ)⁡(M,N)→∼HomDG⁡(X,ℚℓ)⁡(Realℓ⁡(M),Realℓ⁡(N))\operatorname{Hom}_{\operatorname{DM}_{G}(X,\mathbb{Q}_{\ell})}(M,N)\stackrel{{\scriptstyle\sim}}{{\to}}\operatorname{Hom}_{\operatorname{D}_{G}(X,\mathbb{Q}_{\ell})}(\operatorname{Real}_{\ell}(M),\operatorname{Real}_{\ell}(N))

if M,NM,N are ∗*- and !!-pointwise pure Tate, respectively.

0MYL

Proof. As in the proof of Proposition 3.19 the statement can be reduced to the case of a point where it is the same as Proposition 3.15. ∎

Pointwise pure Tate objects can be obtained from pushforwards along proper maps whose fibers have pure Tate motives.

0MYM

Proposition 3.21. Let μ:M→N\mu:M\to N be a GG-equivariant proper map. Assume that MM is smooth and that the motives of the fibers of μ\mu are pure Tate,

M⁡(μ−1​({x})∈DTM⁡(k,ℚ)w=0CLOSE.\operatorname{M}(\mu^{-1}(\{x\})\in\operatorname{DTM}(k,\mathbb{Q})^{w=0}.

Then the object μ!(ℚM)∈DMG(N,ℚ)\mu_{!}(\mathbb{Q}_{M})\in\operatorname{DM}_{G}(N,\mathbb{Q}) is pointwise pure Tate.

0MYN

Proof. We first show that μ!(ℚM)\mu_{!}(\mathbb{Q}_{M}) is ∗*-pointwise pure Tate. Let ix:pt→Ni_{x}:\operatorname{pt}\to N be the inclusion of a point x∈X.x\in X. We have to show that

ix∗Forμ!(ℚM)∈DTM(k,ℚ)w=0.i_{x}^{*}\operatorname{For}\mu_{!}(\mathbb{Q}_{M})\in\operatorname{DTM}(k,\mathbb{Q})^{w=0}.

By applying base change with respect to the Cartesian diagram

μ−1​(x){\lx@inpgf@ignorespaces\mu^{-1}(x)}M{\lx@inpgf@ignorespaces M}{x}{\lx@inpgf@ignorespaces\{x\}}N{\lx@inpgf@ignorespaces N}l\scriptstyle{\lx@inpgf@ignorespaces l}μ′\scriptstyle{\lx@inpgf@ignorespaces\mu^{\prime}}μ\scriptstyle{\lx@inpgf@ignorespaces\mu}ix\scriptstyle{\lx@inpgf@ignorespaces i_{x}}

and the fact that For\operatorname{For} commutes with the six operations, we have

ix∗Forμ!(ℚM)=μ!′l∗ℚM=μ!′ℚμ−1​(x)∈DM(k,ℚ).i_{x}^{*}\operatorname{For}\mu_{!}(\mathbb{Q}_{M})=\mu^{\prime}_{!}l^{*}\mathbb{Q}_{M}=\mu^{\prime}_{!}\mathbb{Q}_{\mu^{-1}(x)}\in\operatorname{DM}(k,\mathbb{Q}).

Now μ′!ℚμ−1​(x)\mu^{\prime}_{!}\mathbb{Q}_{\mu^{-1}(x)} is pure Tate since it is Verdier dual to the motive Mc⁡(μ−1​({x})=M⁡(μ−1​({x})CLOSECLOSE\operatorname{M}^{c}(\mu^{-1}(\{x\})=\operatorname{M}(\mu^{-1}(\{x\}) and Verdier duality preserves pure Tate motives.

That μ!(ℚM)=μ∗(ℚM)\mu_{!}(\mathbb{Q}_{M})=\mu_{*}(\mathbb{Q}_{M}) is !!-pointwise pure Tate follows by using Verdier dual arguments. ∎

Original mathematics by the credited authors. Source collection and HTML conversion remain in progress.

Jens Niklas Eberhardt, Catharina Stroppel

Original source: arXiv:2109.00305v2