Unipotent monodromy and arithmetic \({\mathcal {D}}\)-modules (Q1744809): Difference between revisions

A cohomology theory in algebraic geometry is a functor which associates to algebraic varieties over a field \(k\) vector spaces over another field \(K\) (subject to some further conditions). A number of cohomology theories are in use in algebraic geometry, and with each of these comes a ``category of coefficients'' whose objects encode variations in cohomology and which are stable under Grothendieck's six functors. For example, when \(k\) is of characteristic 0 and \(K = k\), we can consider de Rham cohomology. Systems of differential equations (more precisely, bounded derived catgories of holonomic D-modules) on varieties over \(k\) provide a category of coefficients for this cohomology theory. The paper under review is an exploration of a category of coefficients in the setting of rigid cohomology, where \(k\) is a perfect field of characteristic \(p > 0\) and \(K\) is a complete discrete valuation field of characteristic 0 with residue field \(k\). Rigid cohomology roughly entails lifting varieties over \(k\) to characteristic 0 and considering the de Rham cohomology thereof. Correspondingly, a category of coefficients for rigid cohomology is built roughly out of \(p\)-adic differential equations, ie, D-modules on characteristic 0 lifts. This is the content of the highly technical theory of ``arithmetic D-modules.'' Much work has gone into constructing a category of coefficients for rigid cohomology out of arithmetic D-modules, and it is now known that overholonomic arithmetic D-modules with Frobenius structures form a suitable category of coefficients for rigid cohomology. The presence of these Frobenius structures is an important technical point in this theory, since Frobenius structures guarantee finite dimensionality of natural cohomological constructions. It is of some interest, particularly to those interested in \(p\)-adic differential equations, to consider what happens without insisting on the presence of Frobenius structures. In this more general setting, finite dimensionality of the same natural cohomological constructions mentioned above can be guaranteed by imposing ``\(p\)-adic non-Liouville hypotheses.'' The paper under review defines a notion of arithmetic D-modules having ``potentially unipotent monodromy'' in order to impose these \(p\)-adic non-Liouville hypotheses. The author then goes on to prove that this condition can be used to provide a category of coefficients for rigid cohomology, ie, that this condition is suitably preserved by Grothendieck's six functors.