There are many analogies between $\ell$-adic sheaves on varieties over finite fields and vector bundles with connections on varieties over fields of characteristic zero. I would like to know what is the equivalent of the residue. Let me explain what I mean.

Let $(E, \nabla)$ be a vector bundle with connection on some smooth variety $X$ over a field $k$ of characteristic zero. Assume that $(E, \nabla)$ has regular singularities, meaning that there is a logarithmic extension $\bar{E} \to \bar{E} \otimes \Omega_{\bar{X}}(\log D)$, where $\bar{X}$ is a smooth compactification of $X$ such that $D=\bar{X}-X$ has normal crossings. Then, for each irreducible component of $D_i$, the composition with the Poincare residue gives a map $$ \bar{E} \to \bar{E} \otimes \mathcal{O}_{D_i} $$ which in fact restricts to an endomorphism of $\bar{E} \otimes \mathcal{O}_{D_i}$.

What is the $\ell$-adic analogue of this? Assume we are over a finite field and in the same situation as before ($X \hookrightarrow \bar{X}$ good compactification, this is not anymore automatic!). A regular singular connection should be replace by a representation $\pi_1^{et}(X) \to GL_n(\bar{\mathbb{Q}}_\ell)$ tamely ramified at $D$. How do I get the residue?

One of my concerns is that, in order to define the residue, I need to choose a logarithmic extension and the residue does depend on this choice. What is the substitute of the logarithmic extension in the $\ell$-adic context?

  • 1
    $\begingroup$ An extension of the sheaf to the Kummer \'etale site of $(X, D)$ (where we allow covers tamely ramified along $D$)? $\endgroup$ Sep 30, 2014 at 22:42
  • $\begingroup$ "singular regularities" or regular singularities? $\endgroup$
    – abx
    Oct 1, 2014 at 6:34
  • $\begingroup$ Sorry, it is corrected now. Do you have some deeper knowledge to share with us? ;) $\endgroup$
    – glen90
    Oct 1, 2014 at 6:37

1 Answer 1


So in the \'{e}tale case basically the only thing you have access to is the tame ramification group action on the nearby cycles sheaf on $D$. Because the tame ramification group is cyclic, this is basically the action of a matrix.

I guess the residue of the equation $dy/d \log x = \alpha$ is $\alpha$? So that equation corresponds to the matrix $e^{ 2\pi i \alpha}$. If this is right, then clearly you want to choose a logarithm of the matrix, and take that to be your residue map.

However you face a problem because most matrices don't have logarithms $\ell$-adiccally. So you could restrict attention to matrices that are unipotent modulo the appropriate power of $\ell$, or alternately you can base change from $\mathbb Q_\ell$ to $\mathbb C$ and take a logarithm there.

  • $\begingroup$ Hi Will, thank's for your answer! I think that's a good guess. Can you say something else about how this tame ramification group action constructed or tell me some precise place where I can read about it? $\endgroup$
    – glen90
    Oct 4, 2014 at 14:47
  • $\begingroup$ Abyankhar's lemma says that the action of the fundamental group on a tamely ramified torsion sheaf factors throughout the quotient coming from the cover where you adjoin the $n$th root of the equation defining the divisor. This group is cyclic. Then for an $\ell$-adic cover you get an action of the inverse limit group, which is infinite cyclic. $\endgroup$
    – Will Sawin
    Oct 4, 2014 at 19:00

Your Answer

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

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