Correspondance de Langlands‐Kazhdan explicite dans le cas non ramifié (Q4024449)

Given a local non-archimedean field of residual characteristic \(p\), and an unramified extension \(E\) of \(F\) of degree \(n\), there are (at least) two different procedures for attaching to each regular character \(\theta\) of \(E^ times\) a class of irreducible admissible supercuspidal representations of \(\text{GL}_ n(F)\). One works by explicit induction off a compact-modulo center subgroup [cf. \textit{P. Gerardin}, Proc. Symp. Pure Math. 33, Part I, 157-170 (1979; Zbl 0416.22019)]; denote this correspondence by \(\theta\to \pi'(\theta)\). A second correspondence, denoted \(\theta\to\pi(\theta)\), was constructed by \textit{D. Kazhdan} using global methods and the trace formula [Lect. Notes Math. 1041, 209-249 (1984; Zbl 0538.20014)]. The purpose of the paper under review is to prove the following relationship between these two correspondences: \(\pi'(\theta)=\pi(\theta)\) if \(n\) is odd, whereas for \(n\) even, \(\pi'(\theta)=\pi(\theta\omega)\), with \(\omega\) the (unique) unramified character of \(E^ times\) of order 2. The significance of this result for the ``local Langlands conjecture'' for \(\text{GL}(n)\) is as follows. According to Langlands, there should be a canonical bijection \(\sigma\leftrightarrow \pi(\sigma)\) between the set of classes of \(n\)- dimensional irreducible complex representations of the Weil group of \(F\) and the set of classes of irreducible supercuspidal representations of \(\text{GL}(n,F)\). On the other hand, and independently of this conjecture, we know how to directly construct all the irreducible supercuspidals of \(\text{GL}(n,F)\) via induction from compact-modulo center subgroups [cf. \textit{C. J. Bushnell} and \textit{P. C. Kutzko}, The admissible dual of \(\text{GL}(N)\) via compact open subgroups, Ann. Math. Stud. 129, Princeton University Press, 1993]. So a natural problem is to describe Langlands' \(\pi(\sigma)\) explicitly in terms of the constructions just alluded to above; in particular, for \(\sigma\) equal to a monomial representation \(\sigma(\theta)\), induced from a character of \(E^ times\) regarded as a subgroup of the Weil group. This is precisely what this paper does (since Kazhdan's \(\pi(\theta)\) is really Langlands' \(\pi(\sigma)\) with \(\sigma=\sigma(\theta))\).