Godement-Jacquet L-functions and full theta lifts (Q1650196)

The authors prove a very useful criterion for the big theta lift in theta correspondence for dual reductive pairs of type II. We explain in more detail: let \(D\) be a central division algebra of finite dimension \(d^2\) over \(F\), where \(F\) is a local field. Let \(G=\mathrm{GL}_n(D)\). Let \(S\) be the space of Schwartz (\(F\) Archimedean) or Bruhat-Schwartz (\(F\) non-Archimedean) functions on \(M_n(D)\). This space, with respect to the usual right and left translations, carries, up to a twist by a character, the theta representation for a dual reductive pair \((G,G)\) of type II. It is of interest, for an irreducible admissible (complex) representation \(\sigma\) of \(G\), to understand the reducibility of the full theta lift \(\Theta (\sigma)\). It is already known for a long time that \(\Theta (\sigma)\) has a unique irreducible quotient which is isomorphic to \(\widetilde{\sigma}\) ( \({}\widetilde{}\) denotes the contragredient), and that for a supercuspidal \(\sigma\), \(\Theta (\sigma)\) is irreducible. But, for the applications in the theory of automorphic forms it is important to know when \(\Theta (\sigma)\) is irreducible. The authors obtain the following criterion. Let \(L(s,\sigma)\) be the Godement-Jacquet L-function. Then, if \(L(s,\sigma)\) has no pole for \(s=\frac{1}{2}\) or \(L(s,\widetilde{\sigma})\) has no pole for \(s=\frac{1}{2}\), the lift is irreducible (thus, isomorphic to \(\widetilde{\sigma}\)). Then, the authors go on to prove that the requirements of this criterion are fulfilled for all the generic representations of \(\mathrm{GL}_n(F)\). The idea of the proof of this criterion is to use the local Godement-Jacquet zeta integral, where \(L(s,\sigma)\) arises as the denominator and the numerator is a certain holomorphic function. Then, the results of Godement and Jacquet relate a possible pole of \(L(s,\sigma)\) for \(s=\frac{1}{2}\) with the behavior of that zeta integral restricted to the subspace of Schwartz functions on \(\mathrm{GL}_n(D)\). Besides the usability of this result, a nice feature of the proof is that it treats both the Archimedean and non-Archimedean cases in a parallel way, breaking the proof down into steps which work in both cases.