5
$\begingroup$

Suppose that $X/\mathbb{Q}_p$ is a smooth, projective variety, and choose a prime $\ell\neq p$. Then the weight-monodromy conjecture says that the graded pieces $\mathrm{Gr}_k^M$ of the monodromy filtration on the $i$th cohomology groups $H^i_\mathrm{et}(X_{\overline{\mathbb{Q}}_p},\mathbb{Q}_\ell)$ are pure of weight $i+k$.

I was wondering if there is a similar conjecture for open varieties, coming from looking at a good compactification $X\hookrightarrow \overline{X}$. If $X$ is a curve, then I can get my hands on what's happening explicitly, so suppose that $X$ is a curve with compactification $\overline{X}$ and $D$ is the set of missing points.

Then the only interesting group is $H^1_\mathrm{et}(X_{\overline{\mathbb{Q}}_p},\mathbb{Q}_\ell)$ which by excision sits in an exact sequence $$ 0 \rightarrow H^1_\mathrm{et}(\overline{X}_{\overline{\mathbb{Q}}_p},\mathbb{Q}_\ell) \rightarrow H^1_\mathrm{et}(X_{\overline{\mathbb{Q}}_p},\mathbb{Q}_\ell) \rightarrow V \rightarrow 0 $$ where $V$ is a subspace of $H^0_\mathrm{et}(D_{\overline{\mathbb{Q}}_p},\mathbb{Q}_\ell)(-1)$, it is therefore pure of weight $2$. So (pretty tautologically) $V$ turns up in the 2nd graded piece $\mathrm{Gr}_2^W$ of the weight filtration, but with some basic linear algebra I think I've convinced myself that $V$ actually turns up in the 0th graded piece $\mathrm{Gr}_0^M$ of the monodromy filtration on $H^1_\mathrm{et}(X_{\overline{\mathbb{Q}}_p},\mathbb{Q}_\ell)$, and apart from this 'extra' $V$, the graded pieces of the monodromy filtration for $X$ are the same as those for $\overline{X}$.

In other words, $\mathrm{Gr}^M_{-1}$ and $\mathrm{Gr}_1^M$ are pure of weights 0 and 2 respectively, but $\mathrm{Gr}^M_1$ is mixed with weights 1,2. So the naive weight monodromy conjecture fails, but you can still say something - the graded pieces $\mathrm{Gr}_k^M$ are mixed of weights $\geq i+k$.

So I guess this is my question: is there a conjectural 'weight monodromy' for open varieties which says something like the above, e.g. the $k$th graded pieces is mixed with weights in $[i+k,2i+k]$? Maybe this would easily follow from the 'usual' weight monodromy by choosing a good compactification and then just using some linear algebra?

| cite | improve this question | | | | |
$\endgroup$
7
$\begingroup$

Tony Scholl gave a talk at a conference in Warwick in 2013 on exactly this topic (his talk was called "Remarks on monodromy and weights"). He explained how to formulate a precise version of weight-monodromy for arbitrary varieties over p-adic fields (not necessarily proper or smooth).

The idea is that for any field $K$, and any finite-type $K$-scheme $X$, if we set $H^i(X) = H^i(X_{\overline{K}}, Q_\ell)$ for some $\ell \ne char(K)$, there is a canonical increasing filtration $W_n H^i(X)$ (the "geometric weight filtration") of $H^i(X)$ such that $gr_n^W H^i(X)$ is a subquotient of $H^n(Y)$ for some smooth proper $Y$. (This is due to Deligne and de Jong.)

So it's reasonable to conjecture that if $K$ is a $p$-adic field and $\ell \ne p$, the pieces $gr_n^W H^i$ should be "quasi-pure of weight n" as representations of $G_K$, which means that all eigenvalues of Frobenius are Weil numbers of some weight, and for all $r$ the monodromy operator $N^r$ sends the weight $n + r$ generalized eigenspace isomorphically to the weight $n - r$ one. (Scholl just calls this "pure of weight $n$", but this conflicts with the more restrictive usage of "pure" in your question.)

In Scholl's talk he announced a proof that the usual weight-monodromy conjecture (i.e. for $X$ proper and smooth over $K$) actually implies this more general conjecture. This implication is not obvious, because being quasi-pure is not preserved under passing to subquotients (and Tony didn't say how he gets around this, as far as I can recall).

(It's not totally clear if this answers the question, because one would have to understand how the geometric weight filtration relates to the filtration on $H^i(X)$ given by the Grothendieck abstract-monodromy theorem -- they are definitely not the same, because the latter encodes information about the singularities of the special fibre.)

| cite | improve this answer | | | | |
$\endgroup$

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.