Exponential sums and rigid cohomology (Q2135567)

\(L\)-functions defined by exponential sums may be studied using cohomological methods. The \(\ell\)-adic cohomology theory seems to be ill-suited for the investigation of slopes (the only tool that I know of is Deligne's integrality theorem). Among \(p\)-adic cohomology theories associated with exponential sums, Adolphson-Sperber utilized a complex of Banach spaces, while Berthelot introduced the rigid cohomology with coefficients in a ``Dwork crystal''. There are also variants of these theories. In the paper under review, the author proves a comparison theorem, which equates an overconvergent variant of the Adolphson-Sperber cohomology with rigid cohomology. To prove the comparison theorem, he begins by explaining how to translate differential operators as well as Frobenii of Adolphson-Sperber (which were defined by the Artin-Hasse exponential) to those used in rigid cohomology (which were defined by Dwork exponentials, a truncation of the Artin-Hasse exponential). Then he proves an overconvergent vanishing theorem which allows him to get rid of higher direct images in the argument. The proof concludes with a Mayer-Vietoris argument.