I am reading Mazza's, Voevodsky's and Weibel's book Lecture Notes on Motivic Cohomology and have grown curious about the following question:

Which cohomology theories on $Sm/k$ are representable, i.e. appear as Hom-groups, in Voevodsky's construction of $DM_{\text{Nis}}^{eff,-}(k, R)$?

If we let $X$ be a scheme of finite type over $k$ and $R$ a commutative unital ring, then we have (by construction) motivic homology with coefficients in $R$ $$H_{n,i}(X, R) = \text{Hom}_{DM_{\text{Nis}}^{eff,-}}(R(i)[n], R_{\text{tr}}(X)),$$ and motivic cohomology $$H^{n,i}(X, R) = \text{Hom}_{DM_{\text{Nis}}^{eff,-}}(R_{\text{tr}}(X), R(i)[n]).$$ From this, we obtain the algebraic singular homology (and hence Suslin's singular homology) as $$H^{\text{sing}}_n(X,R) \cong H_{n,0}(X,R)$$ and, if $X$ is a smooth separated scheme of finite type over some perfect field $k$, the higher Chow groups $$CH^q(X, 2q-p) = H^{p,q}(X,\mathbb{Z}).$$

It is clear to me that such cohomology theories must satisfy certain properties:

  • Mayer-Vietories
  • $\mathbb{A}^1$-invarience
  • Künneth-theorem
  • ...

But do we have any other examples of cohomology theories that appear as Hom-groups in $DM_{\text{Nis}}^{eff,-}(k, R)$? How about

  • Betti cohomology?
  • $l$-adic cohomology?
  • Crystalline cohomology?
  • Algebraic de Rham cohomology?

Another way of framing the question is: How far off is Voevodsky's-Morel's $DM_{\text{Nis}}^{eff,-}(k, R)$ construction from being the category of pure/mixed motives in the sense of Grothendieck.

(I realize this might be a very hard question to which not much is known.)

  • 1
    $\begingroup$ Another "obvious" requirement is that the cohomology theory must have Voevodsky transfers (i.e. factor through the category $\operatorname{Corr}_k$ of Voevodsky correspondences). $\endgroup$ Feb 27 '21 at 16:10

Recall that $\mathrm{DM}(k)$ can be described as the subcategory of $\mathrm{SH}(k)$ made of modules over the motivic cohomology. This implies that cohomologies which are representable in $\mathrm{DM}(k)$ receive a cycle class map, so they must admit additive Chern classes ($c_1(L\otimes L')=c_1(L)+c_1(L')$). In fact, I think this is an if and only if, in the sense that motivic cohomology is the universal among cohomologies with additive Chern classes, but I do not know right now if this is written anywhere in the motivic setting with integral coefficients. In any case, if you put $R=\mathbb{Q}$ then you can have any type of Chern classes, for example $K$-theory. This is all written at Cisinki-Déglise's Triangulated categories..., chapter 14, which is in general a great reference for many motivic results up until 2009.

Your question is more natural for $\mathrm{SH}(k)$, as it happens already in topology, and then you know your cohomology would belong to $\mathrm{DM}(k)$ depending on if it is a module over motivic cohomology or not. An answer to your question in the setting of $\mathrm{SH}$ is in the paper of Cisinski and Déglise, Mixed Weil cohomologies, at 2.1.5. Loosely speaking, for any "sheaf cohomology" they construct an object of $\mathrm{SH}(k)$ representing such cohomology. Their hypothesis must imply having additive Chern classes, since they are willing to define a mixed variant of Weil cohomologies (which have additive Chern classes), so their cohomologies are in $\mathrm{DM}(k)$. At the end of the paper you have their main examples.

I recall that Cisinski-Déglise's method was later rewritten and developped by Holmsotrom-Scholbach in Arakelov motivic cohomology, section 3, to represent the Deligne cohomology, and by Déglise-Mazzari in The rigid syntomic spectrum at 1.4.10 where they proved that the rygid syntomic cohomology is representable.

  • 2
    $\begingroup$ $DM$ is absolutely not a full subcategory of $SH$. It is, as you correctly mention, a category of modules. $\endgroup$ Feb 27 '21 at 10:37
  • $\begingroup$ Shame on me! ^_^' Thanks a lot for pointing out my mistake $\endgroup$
    – Tintin
    Feb 27 '21 at 11:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.