Rankin-Selberg periods for spherical principal series (Q2139841)

This paper is motivated by the subconvexity problem for the Rankin-Selberg \(L\)-function \(L(s,f\times \bar g)\) attached to a pair of automorphic forms \(f\) (on \(\mathrm{GL}(n)\)) and \(g\) (on \(\mathrm{GL}(n')\)). The Rankin-Selberg \(L\)-function \(L(s,f\times \bar g)\) is a Dirichlet series which can be seen as, loosely speaking, a generating function for the Fourier-Whittaker coefficients of \(f\) and \(g\). A fundamental problem in the theory of \(L\)-functions is obtaining bounds for \(L(s,f\times \bar g)\) along the critical line \(s = 1/2 + i\mathbb{R}\). There is a standard way of obtaining the so-called convexity bound using the Phragmén-Lindelöf principle. A bound that improves upon this is called a \textit{subconvexity bound}. One way of producing subconvexity bounds is by using period integrals. The period integral \(\Lambda(s,f\times \bar g)\) is the integral of \(f\) and \(g\) (suitably modified, in some cases) over a certain locally symmetric space; for example, when \(n' = n-1\) we have \[ \Lambda(s,f\times \bar g) = \int_{\text{SL}(n',\mathbb{Z})\backslash \text{GL}(n',\mathbb{R})} f(z) \overline{g(z)}|\det(z)|^{s-\frac 12}\ dz. \] Using standard methods, one shows that the period integral is -- up to certain factors that are easy to control -- equal to the product a local Archimedean \(L\)-factor \(G(s)\) and the original \(L\)-function \(L(s,f\times \bar g)\). Thus, bounds for the Rankin-Selberg \(L\)-function can be obtained by bounding the period integral. In the paper under review, the authors propose a way of obtaining bounds for the period integral based on a method developed by Bernstein and Reznikov, as follows. The period integral \(\Lambda(s,f\times \bar g)\) gives rise to an invariant functional on the tensor product of automorphic representations corresponding to \(f\) and \(g\). The proposed method consists of two steps: \begin{itemize} \item[1.] Directly construct explicit invariant functionals on the tensor product of representations coming from \(f\) and \(g\). \end{itemize} By the multiplicity one property (proved in this setting by \textit{B. Sun} and \textit{C.-B. Zhu} [Ann. Math. (2) 175, No. 1, 23--44 (2012; Zbl 1239.22014)] and \textit{F. Chen} and \textit{B. Sun} [Int. Math. Res. Not. 2015, No. 14, 5849--5873 (2015; Zbl 1333.22013)], such a functional must be equal to the functional coming from \(\Lambda(s,f\times \bar g)\), up to a scalar factor. Thus, the second step is \begin{itemize} \item[2.] Obtain bounds for the proportionality scalar. \end{itemize} In this paper the authors focus on the first step. To be precise, they construct explicit invariant functionals on tensor products of spherical principal series representations. The authors also conjecture that the special value of these invariant functionals at the spherical vectors is equal in absolute value to the Archimedean \(L\)-factor \(G(s)\) up to a constant. Finally, they prove this conjecture for \((n,n') = (2,1), (2,2), (3,1), (3,2)\).