Given a spherical variety X for a group G over a nonarchimedean local field k, the Plancherel decomposition for L 2 (X) should be related to "distinguished" Arthur parameters into a dual group closely related to that defined by Gaitsgory and Nadler. Motivated by this, we develop, under some assumptions on the spherical variety, a Plancherel formula for L 2 (X) up to discrete (modulo center) spectra of its "boundary degenerations", certain G-varieties with more symmetries which model X at infinity. Along the way, we discuss the asymptotic theory of subrepresentations of C ∞ (X) and establish conjectures of Ichino-Ikeda and Lapid-Mao. We finally discuss global analogues of our local conjectures, concerning the period integrals of automorphic forms over spherical subgroups.
Contents8 This also holds for any quasi-affine spherical variety, if H denotes the stabilizer of a point on the open G-orbit -cf. the remark after Lemma 6.6 in loc.cit.. 9 In cases like the Whittaker model, where X = H\G and we have a morphism Λ : H → Ga whose composition with a complex character of k gives rise to the line bundle 16 Here is how to see the equivalence: If they are simple and orthogonal, the only root system they can generate is A1 × A1. Vice versa, if there are no more roots in their linear span, we can find real functionals ℓ1 and ℓ2, such that ℓ1 is positive on α, β and ℓ2 is zero on ±α, ±β and non-zero on all other roots. Then, for s ≫ 0, the functional ℓ1 + sℓ2 distinguishes a set of positive roots which must have α and β as its simple elements, because it takes larger values on every other positive root. We thank Vladimir Drinfeld for pointing out this equivalence.