7
$\begingroup$

I begin by clarifying that the "higher-dimensional" in my question refers to analogues of Artin L-functions over higher dimensional base schemes than $\mathrm{Spec}(\mathbb{Z})$.

Now for the set-up. This will be very similar to the set-up of Chapter 9 of Serre's book "Lectures on N_X(p)". Let $R$ be a finite type flat $\mathbb{Z}$-algebra of dimension $n$ with fraction field $K$ and let $X = \mathrm{Spec}(R)$. Let $L/K$ be a finite Galois extension with Galois group $G$ and let $Y$ be the normalisation of $X$ in $L$. At any point I am quite happy to shrink $X$. For example, I may assume that $X$ is regular and that $Y/X$ is finite étale.

Now for the analogue of Artin L-functions. For a complex representation $\rho$ of $G$ we define $$L(s,\rho) = \prod_{\substack{\text{closed points} \\ x \in X}}\mathrm{det}\left(I - \frac{\rho(\mathrm{Frob}_x)}{N(x)^s}\right)^{-1}, \quad \text{re}(s) > n.$$ See the above book of Serre for the definition of Frobenius elements in this setting. That the above Euler product converges in the range $\text{re}(s) > n$ follows from the Weil conjectures.

My first question is the following.

  1. Have such L-functions been studied before?

It seems that Deligne's Weil II paper should be relevant here, but alas I don't understand it well enough.

But what I would really like to know is the following.

  1. Assume that $\rho$ is irreducible and not the trivial representation. Then does $L(s,\rho)$ admit a holomorphic continuation to the line $\text{re}(s) = n$ without zeros?

This latter fact is proved in the classical case where $n = 1$ by using Brauer's induction theorem to reduce to the case where $G$ is abelian. Here one essentially obtains a dirichlet L-function, where the result is well-known (being part of the proof of Dirichlet's theorem on primes in arithmetic progressions). I imagine a similar reduction to the abelian case can be made in my setting, but of course the point is that I don't know how to deal with the abelian case itself.

$\endgroup$
  • $\begingroup$ There is an article of Serre on zeta functions associated to schemes $X$ of finite type over $\mathbb{Z}$. If I recall correctly, if $G$ is a finite group acting on $X$ by automorphisms, and $\chi$ is a complex character of $G$, then you can define a zeta function $\zeta(X,\chi,s)$. Does this give the same $L$-function you suggest to look at? $\endgroup$ – François Brunault Oct 15 '15 at 9:13
  • $\begingroup$ Quite possibly. Which article are you referring to? Serre has many... $\endgroup$ – Daniel Loughran Oct 15 '15 at 9:17
  • $\begingroup$ I don't have Serre's collected papers at hand, but I think the article in question is MR0194396 (33 #2606) Serre, Jean-Pierre . Zeta and L functions. 1965 Arithmetical Algebraic Geometry (Proc. Conf. Purdue Univ., 1963) pp. 82--92 Harper & Row, New York. $\endgroup$ – François Brunault Oct 15 '15 at 9:20
  • $\begingroup$ Thanks for the reference. I have only managed to find the paper so far in Russian.... I will keep on looking. $\endgroup$ – Daniel Loughran Oct 15 '15 at 14:12
3
$\begingroup$

According to Serre in "Zeta and L-functions" (MR0194396), these L-functions have been defined by Artin himself. Moreover, it seems that in the formula $$ L(s,\rho) = \prod_{\substack{\text{closed points} \\ x \in X}}\mathrm{det}\left(I - \frac{\rho(\mathrm{Frob}_x)}{N(x)^s}\right)^{-1} $$ one should define $\rho(\mathrm{Frob}_x)$ as the average of the $\rho(g)$ for $g \in G$ mapping to $\mathrm{Frob}_x$.

Serre further says that $L(s,\rho)$ can be continued to a meromorphic function in the half-plane $\mathrm{Re}(s)>n-\frac12$, and that the singularities « can be determined, or rather reduced to the classical case $n=1$ » (there is an interesting lemma in the case $Y \to Y'$ is a fibration of curves, enabling to do induction on the dimension of $Y$). I don't know whether this answers your Question 2.

Artin's paper is Zur Theorie der L-Reihen mit allgemeinen Gruppencharakteren. Abh. Math. Sem. Univ. Hamburg 8 (1931), no. 1, 292--306 (MR3069563).

$\endgroup$
  • $\begingroup$ Thanks very much for the reference (I have now found a copy of it in English). In fact Serre mentions a result in his paper which is almost exactly the answer to Q2. Namely the final Corollary in the paper states that $L(s,\rho)$ is holomorphic and non-zero at $s=n$. I wonder if it is somehow implicit in this work that $L(s,\rho)$ is also holomorphic and non-zero on the line $\text{re}(s) = n$? $\endgroup$ – Daniel Loughran Oct 15 '15 at 16:20
  • $\begingroup$ It is a shame that this paper contains essentially no proofs, especially of the interesting result you mention. Any idea where I can find proofs? $\endgroup$ – Daniel Loughran Oct 15 '15 at 16:21
  • $\begingroup$ It seems that Serre's inductive argument on the dimension to show $L(\rho,n) \neq 0$ could also be used on the line $\mathrm{Re}(s)=n$, but I'm really no expert here. Unfortunately I know of no other reference than this article. It would be indeed very nice to have a reference with more detailed proofs. $\endgroup$ – François Brunault Oct 15 '15 at 17:34
  • $\begingroup$ It would be also very nice to define a $G$-equivariant L-function in this setting. I have not seen a reference for this. This would not really be a new object, but this would tie things together nicely I think. $\endgroup$ – François Brunault Oct 15 '15 at 17:39
  • $\begingroup$ After chasing references, I have managed to find a proof of the result I want. It can be found in Section 2.1. of Faltings - Complements to Mordell. Thank you for your help on this matter. $\endgroup$ – Daniel Loughran Oct 20 '15 at 10:52

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.