Distinguished regular supercuspidal representations (Q2308338)

Let \(F\) be a nonarchimedean local field, \(G\) a reductive \(F\)-group, \(\theta\colon G\to G\) an involution, and \(G^\theta\) the fixed points of~\(\theta\). A (smooth, complex) representation \(\pi\) of \(G(F)\) is said to be \textit{\((\theta,\chi)\)-distinguished}, where \(\chi\) is a character of~\(G^\theta(F)\), if \[ \mathrm{Hom}_{G^\theta(F)}(\pi,\chi) \neq 0. \] In particular, we say \(\pi\) is \textit{\(\theta\)-distinguished} if it is \((\theta,\text{triv})\)-distinguished. Identifying which representations are distinguished gives information about automorphic periods, such as their nonvanishing. Distinction also features in the conjectural framework of the relative Langlands program [\textit{Y. Sakellaridis} and \textit{A. Venkatesh}, Periods and harmonic analysis on spherical varieties. Paris: Société Mathématique de France (SMF) (2017; Zbl 1479.22016)]. The article under review proves a simple criterion for distinction which applies to most supercuspidal representations. It also verifies in important special cases that distinction satisfies an expected compatibility with respect to the local Langlands correspondence. The representations treated in the article are the regular supercuspidal representations of \textit{T. Kaletha} [J. Am. Math. Soc. 32, No. 4, 1071--1170 (2019; Zbl 1473.22012)]. Such representations are in canonical bijection with \(G(F)\)-conjugacy classes of pairs \((S,\mu)\), where \(S\) is an elliptic maximal torus of~\(G\) and \(\mu\) is a regular character of~\(S(F)\). Write \(\pi=\pi_{(S,\mu)}\) for the representation corresponding to~\((S,\mu)\). The tool for testing distinction is a pair of characters of~\(S(F)\). The first, \(\eta_S\), stems from earlier and related work of \textit{J. Hakim} and \textit{F. Murnaghan} [IMRP, Int. Math. Res. Pap. 2008, Article ID rpn005, 166 p. (2008; Zbl 1160.22008)]. The second, \(\varepsilon_S\), stems from work of Lusztig on distinction problems over finite fields [\textit{G. Lusztig}, Prog. Math. None, 57--81 (1990; Zbl 0773.20012)]. Zhang shows that \(\pi_{(S,\mu)}\) is \(\theta\)-distinguished if and only if (1) \(S\) is \(\theta\)-stable (after possibly conjugating by an element of~\(G(F)\)) and (2) \(\mu\) is \((\theta, \varepsilon_S\cdot\eta_S)\)-distinguished. As for the Langlands correspondence, Lapid has conjectured that if \(\pi\) is \(\theta\)-distinguished then its \(L\)-packet is preserved by the involution \(\tau\mapsto\tau^\vee\circ\theta\). For Kaletha's regular \(L\)-packets, Zhang computes the effect of the involutions \(\tau\mapsto\tau\circ\theta\) and \(\tau\mapsto\tau^\vee\) on \(L\)-parameters. They correspond to composition with the dual of \(\theta\) and the Chevalley involution, respectively. If the character \(\eta_S\) is symmetric, meaning that \(\eta_S\circ\theta = \eta_S^{-1}\), the results are stronger: any \(\theta\)-distinguished \(\pi\) is isomorphic to \(\pi^\vee\circ\theta\) and thus satisfies Lapid's conjecture, in fact, a strengthening of it. Zhang suspects that \(\eta_S\) is symmetric in general, and can prove it when \(\pi\) has depth zero or is epipelagic.