14
$\begingroup$

As far as I understand, for a smooth variety $X$ its motivic cohomology could be described as the corresponding piece of the $\gamma$-filtration of (Quillen's) $K^*(X)$; this is completely true for $\mathbb{Q}$-coefficients, and true up to bounded denominators for $\mathbb{Z}$-coefficients.

My question is: is there a similar result for singular varieties? Here for motivic cohomology I would like to take $Hom_{DM}(M(X),\mathbb{Z}(p)[q]$; $DM$ is the category of Voevodsky's motives, and $M(X)$ is the motif of $X$ (I don't want to take the motif with compact support instead). These cohomology theories satisfy cdh-descent, but have no easy descriptions in terms of complexes of algebraic cycles. Unfortunately, I don't know much about the $\gamma$-filtration of $K$-theory.

Actually, I would like to prove the following fact: if a morphism $X\to Y$ of varieties induces an isomorphism for $K^*$, then the exponents of the kernels and cokernels of the corresponding morphisms for motivic cohomology are bounded (by a constant that depends only only on the dimensions of $X$ and $Y$). Any hints and/or references would be very welcome!

$\endgroup$

1 Answer 1

10
$\begingroup$

The precise relationship between K-theory and motivic cohomology for smooth schemes is the (analog of the) Atiya-Hirzebruch spectral sequence. This generalizes to non-smooth schemes if one uses K'-theory and higher Chow groups.

It is easy to see that motivic cohomology cannot be used to recover K-theory, because it does not detect nilpotents. For example, $k[t]/t^2$ has K-theory different from $k$, but motivic cohomology is the same. cdh-descent than gives examples for reduced schemes (look at a cusp).

There are some approaches (Bloch-Esnault) to beef up cycles to get results in the above examples.

As for your original question, recover motivic cohomology from K-theory, this might be possible, but I have no result to offer.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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