Spherical pairs over close local fields (Q1760057)

This is a beautiful paper proving striking results. Two nonarchimedean local fields \(E,F\) are said to be \(m\)-close for a positive integer \(m\), if the residue rings \(O_E/M_E^m\) and \(O_F/M_F^m\) are isomorphic. For example, \(\mathbb F_p((t))\) is \(m\)-close to \(\mathbb Q_p(p^{1/m})\); more generally, for any \(m\), any local field of positive characteristic \(p\) is \(m\)-close to a suitable extension of \(\mathbb Q_p\). In this manner, representation theory of groups over local fields of characteristic zero can be related to that of groups over local fields of positive characteristic. The folklore result is due to D. A. Kazhdan in 1986; Kazhdan showed for a reductive group \(G\) over \(\mathbb Z\), that the Hecke algebras of \(G(E)\) and \(G(F)\) with respect to the \(l\)-th congruence subgroups \(K_l(E):=\mathrm{Ker}(G(E)\to G(O_E/M_E^l))\) and \(K_l(F)\) are isomorphic if \(E\) and \(F\) are \(m\)-close for some appropriate \(m\geq l\). The authors obtain a relative version of Kazhdan's results. In particular, they prove: Theorem 1. Let \(F\) be a nonarchimedean local field, \(G\) a reductive group defined over \(F\) and \(H\) an algebraic subgroup of \(G\) also defined over \(F\). Assume that for any parabolic subgroup \(P\) of \(G\), the number of double cosets \(P(F)\backslash G(F)/H(F)\) is finite. Suppose that for any irreducible, smooth representation \(\rho\) of \(G(F)\), the number \(\dim\Hom_{H(F)}(\rho|_{H(F)},\mathbb C)\) is finite. Then, for any compact open subgroup \(K\) of \(G(F)\), the space \(\mathcal{S}(G(F)/H(F))^K\) of \(K\)-invariant, compactly supported functions on \(G(F)/H(F)\) is finitely generated (and hence finitely presented) as a module over the Hecke algebra \(\mathcal{H}_K(G(F))\). Theorem 2. Let \(H<G\) be reductive groups defined over \(\mathbb Z\) and assume that the pair \((G,H)\) is a uniformly spherical pair. Then, for any \(l\), there exists \(n\) such that for any \(n\)-close local fields \(E\) and \(F\), the module \(\mathcal{S}(G(F)/H(F))^{K_l(F)}\) over \(\mathcal{H}_{K_l}(G(F))\) and the module \(\mathcal{S}(G(F)/H(F))^{K_l(E)}\) over \(\mathcal{H}_{K_l}(G(E))\) are isomorphic, where the Hecke algebras have been identified by Kazhdan's isomorphism. Applying these results, the authors obtain the following important results (of which the characteristic 0 version was proved by the first and the third author along with other collaborators S. Rallis, E. Sayag and G. Schiffmann): For any local field \(F\), \((GL_{n+1}(F),GL_n(F))\) is a strong Gelfand pair; that is, \(\dim\Hom_{GL_n(F)}(\pi,\tau) \leq 1\) for irreducible smooth representations \(\pi\) of \(GL_{n+1}(F)\) and \(\tau\) of \(GL_n(F)\). Further, if \(F\) has characteristic different from \(2\), then \((GL_{n+k}(F),GL_n(F)\times GL_k(F))\) is a Gelfand pair. Special cases and analogues for characteristic zero for some of the above results were obtained earlier by the authors mentioned above and B. Sun and C.-B. Zhu.