Refined global Gross-Prasad conjecture on special Bessel periods and Böcherer's conjecture (Q2031669)

Given an odd orthogonal group \(G\) over a number field \(F\), its Bessel subgroup has the form \(R=D\ltimes S\) where \(S\) is a unipotent subgroup and \(D\) is the reductive part isomorphic to \(E^*/F^*\); \(E\) is a quadratic extension of \(F\). For an automorphic form \(\phi\) on \(G\), its Bessel period is \(\int_{R(F)\backslash R(A_F)}\phi(r)\psi^{-1}(r)\ dr\) where \(\psi\) is a character of \(R\) determined by an additive character on \(A_F/F\). The general Gan-Gross-Prasad conjecture applies to this period, and predicts a relation of the period with \(L\)-values. This paper established such a relation under mild restriction: that \(\phi\) is in the space of an irreducible cuspidal representation \(\pi\) whose Archimedean components are discrete series and whose Bessel periods do not always vanish. The main result is an explicit identity between the Bessel period and the \(L\)-values, similar to the identity for Gross-Prasad period conjectured by Ichino-Ikeda. One important application of the result is the proof of the long-standing Bocherer's conjecture. That conjecture is stated for Siegel modular forms, which can be considered as cusp forms on \(\mathrm{PGSp}_2\cong \mathrm{SO}(5)\). The main result readily applies as the Archimedean components are discrete series in this case. Bocherer's original conjecture was stated as an equality up to a constant, the paper explicates that constant. In this paper only the level one modular forms are considered, but it is remarked that the result can be extended to odd square free levels. The proof of the main result uses a period transition argument. Through theta correspondence, the Bessel period on \(G\) can be transited to a Whittaker period on a metaplectic group. The Ichino-Ikeda type formula for Whittaker period on a metaplectic group was established by \textit{E. Lapid} and \textit{Z. Mao} [Algebra Number Theory 11, No. 3, 713--765 (2017; Zbl 1418.11076)] under the same assumption on Archimedean components of the representation. To complete the argument, one also needs the transition between the inner products on \(G\) and on the metaplectic group, which is supplied by Rallis inner product formula.