On completeness of local intertwining periods
Hengfei Lu, Nadir Matringe
TL;DR
The paper develops a framework to describe invariant linear forms on induced representations in terms of intertwining periods for unimodular tempered symmetric pairs. It formulates and motivates a conjecture that, when the inducing data is square-integrable, the entire space $\mathrm{Hom}_H(I_P^G(\sigma),\mathbb{C})$ is generated by regularized (and in favorable cases normalized) intertwining periods attached to admissible parabolic orbits, providing a canonical description up to normalization. The authors verify the conjecture in several key settings, including the group case, Galois pairs $({\mathrm{GL}}_n(E),{\mathrm{U}}_n(E/F))$, inner forms of $\mathrm{GL}_n$, and rank-one cases with $G={\mathrm{SL}}_2$, and develop a geometric lemma to study the support of these periods. Their results connect the local relative Langlands program to explicit period calculations, offering practical tools for understanding distinction and multiplicities in induced representations. The work lays a path toward a general normalization framework and shows how, in important instances, distinction can be analyzed through intertwining operators and open double coset geometry.
Abstract
In this paper we study the problem of explicitly describing the space of invariant linear forms on induced distinguished representations in terms of invariant linear forms on the inducing representation. More precisely, for certain tempered reductive symmetric pairs (G,H) over a local field of characteristic zero, which we call unimodular in this paper, we study under which condition on the inducing representation, the space of H-invariant linear forms on a parabolically induced representation of G is generated by regularized intertwining periods attached to admissible parabolic orbits in G{H, as defined in the work of Matringe--Offen--Yang. We conjecture that it is the case when the inducing representation is square-integrable. Under this assumption we actually conjecture that one can replace regularized by normalized intertwining periods. We then verify the conjecture on known examples, and prove it for various pairs where G has semi-simple split rank one.
