On de Jong's conjecture (Q877470)

A few years ago, \textit{A. J. de Jong} proposed a significant conjecture on Galois representations of arithmetic fundamental groups of normal varieties over finite fields [Isr. J. Math. 121, 62--84 (2001; Zbl 1054.11032)]. A particular geometric version of de Jong's conjecture can be formulated as follows: Let \(X\) be a smooth projective curve over a finite field \(\mathbb F_q\), and let \(\rho:\pi_1(X)\to \text{GL}_n({\mathbf F})\) be a continuous representation of its arithmetic fundamental group \(\pi_1(X)\) over \({\mathbf F}=\mathbb F_\ell(t))\), with \(\mathbb F_\ell\) being another finite field of order \(\ell\) coprime to \(q\). Assume that the induced representation of the geometric fundamental group \(\pi_1(\overline X)\subset\pi_1(X)\) is absolutely irreducible. Then, under these conditions, the group \(\rho (\pi_1(\overline X))\) is finite. As A. J. de Jong already pointed out, a proof of this conjecture would follow from a version of the geometric Langlands conjecture, the relevant formulation of which is given by the following statement: If \(\rho:\pi_1(X) \to \text{GL}_n({\mathbf F})\) is a representation as in de Jong's conjecture, then there exists a non-zero cuspidal Hecke eigenform \(f_\rho\) with eigenvalues corresponding to this representation \(\rho\). In the paper under review, the author sketches a possible proof of this version of Langland's conjecture, and therefore also a proof of the above-mentioned version of de Jong's conjecture. However, as the author points out, his approach relies on two mathematical theories that have not yet completely worked out, and which actually do not exist in the published literature. One of them is the theory of étale sheaves with coefficients in \({\mathbf F}\), which is expected to be parallel to the corresponding theory of \({\mathbf F}'\)-adic sheaves for a local field \({\mathbf F}'\) of characteristic zero. Taking the existence of such a theory for granted, including the existence of a suitable abelian category of perverse sheaves, together with, the necessary sheaf-function correspondence, the author reformulates the above-mentioned geometric version of Langland's conjecture in this sheaf-theoretic context, and proves it then in this form. Actually, to a representation \(\rho\) as above, he associates an object \(S_\rho\) in the derived category of the moduli stack of rank \(n\) bundles on the curve \(X\) such that \(S_\rho\) is cuspidal and satisfies the Hecke eigencondition with respect to the representation \(\rho\). Applying the assumed sheaf-function correspondence in this sheaf-theoretic framework, he then obtains a function \(f_\rho\) on \(\text{Bun}_n(\mathbb F_q)\), which indeed appears as a cuspidal Hecke eigenform with eigenvalues corresponding to the original representation \(\rho\). This proof is carried out under the assumption that \(\ell>2n\), but it is also indicated that it works for any \(\ell\neq 2\). To summarize, the author gives a proof of the special versions of the conjectures formulated as above (Langlands, de Jong) in the case \(\ell\neq 2\), modulo the still incomplete theory of étale \({\mathbf F}\)-sheaves, which quite conceivably should exist in the form used in the present paper.