Multiplicity one theorem for \((\mathrm{GL}_{n+1}, \mathrm{GL}_n)\) over a local field of positive characteristic (Q2663085): Difference between revisions

Let \(\mathbb{F}\) be a non-Archimedean field of characteristic different from \(2\), and embed \(\mathrm{GL}_{n}(\mathbb{F})\) into \(\mathrm{GL}_{n+1}(\mathbb{F})\) via the standard upper-left embedding. The main result of this paper states that each distribution on \(\mathrm{GL}_{n+1}(\mathbb{F})\) which is invariant under the conjugation action of \(\mathrm{GL}_n(\mathbb{F})\) is also invariant under transposition. Several corollaries follow in short order: Each irreducible smooth representation of \(\mathrm{GL}_{n}(\mathbb{F})\) occurs with multiplicity at most \(1\) in the restriction of each irreducible smooth representation of \(\mathrm{GL}_{n+1}(\mathbb{F})\); and if \(P_n\) is the parabolic subgroup consisting of matrices whose final column is \((0,\ldots,0,1)\), then every \(P_n\)-invariant distribution on \(\mathrm{GL}_{n}(\mathbb{F})\) is \(\mathrm{GL}_{n}(\mathbb{F})\)-invariant. These corollaries are previously known, but are nevertheless nice illustrations of the utility of the main result. The proof of the main theorem adapts a proof given in [\textit{A.\ Aizenbud} et al., Ann.\ Math. (2) 172, No.\ 2, 1407--1434 (2010; Zbl 1202.22012)] of the same result for the case where \(\mathbb{F}\) has characteristic zero, following essentially the same logical argument as given in [\textit{A.\ Aizenbud} and \textit{D.\ Gourevitch}, Sel.\ Math., New Ser.\ 15, No.\ 2, 271--294 (2009; Zbl 1185.22006)]. Thus, the proof proceeds by letting \(\tilde{G}\) be the semidirect product of \(\mathrm{GL}_n\) with a group of order \(2\) acting by transposition, and considering \((\tilde{G},\chi)\)-equivariant distributions on \(\mathrm{gl}(V) \times V \times V^{\ast}\), where \(\chi\) is the pull-back of the sign character. The main theorem is reformulated to state that all such equivariant distributions are zero, and then restrictions are derived on the possible support of such a distribution. The main difference in logical argument between this paper and these references is that Harish-Chandra descent is avoided here, which is necessary to avoid restrictions on the characteristic.