Rankin-Selberg \(L\)-functions on the critical line (Q2487842)

The author considers subconvex estimates for Rankin-Selberg \(L\)-functions \(L(s,f\times g)\), where \(f\) and \(g\) are cusp forms with \(g\) fixed. More precisely, the cusp forms in question can be either holomorphic forms of weight \(k\), level \(N\) and character \(\chi \), or weight zero Maass forms of level \(N\), character \(\chi \) and spectral parameter \(t=\sqrt{\lambda -1/4}\), \(\lambda \) being the Laplace eigenvalue. The estimates are given in terms of the ``infinity parameter'' \(\mu \), which equals \((k-1)/2\) if \(f\) is a holomorphic form, and \(t\) if \(f\) is a Maass form. The main result is the estimate \[ L(s,f\times g) \ll _{g,s,N_f,\varepsilon} (1+| \mu | )^{(6-2\theta )/(7-4\theta )+\varepsilon }, \] where \(\Re s =1/2\) and \(\theta \) is an admissible exponent for the Ramanujan-Petersson conjecture. The known value \(\theta = 7/64\) gives the exponent \(0.880...\) in the preceding estimate, and the point is that this exponent is less than 1. \textit{P. Sarnak} [J. Funct. Anal. 184, No. 2, 419--453 (2001; Zbl 1006.11022)] was the first to give subconvex estimates of the above type. Briefly, the argument runs as follows. Instead of a single function, a mean square over a family of functions is considered, and this leads to the ``shifted convolution'' problem which is treated by the reviewer's variant of the circle method and Kuznetsov's trace formula for sums of Kloosterman sums. Reviewer's remark: In the case of the full modular group, the subconvex exponent has been reduced to \(2/3+\varepsilon \), with a relatively wide uniformity in \(s\), in a forthcoming paper by \textit{Y. Motohashi} and the reviewer, to appear in Acta Math.).