The subconvexity problem for Rankin-Selberg \(L\)-functions and equidistribution of Heegner points. II (Q818640)

One central problem of \(L\)-functions is to study their sizes on the critical line and the ultimate goal is to establish the (Grand) Lindelöf hypothesis. There is a generic method, though being unable to prove the Lindelöf hypothesis, to give upper estimates by means of the convexity principle (or Phragmén-Lindelöf principle) in complex analysis. Subconvexity bounds are referred to results better than these upper estimates. In the vast family of \(L\)-functions, subconvexity bounds (in various aspects) are known for a few members. This article is a continuation of the important work [Ann. Math. (2) 160, 185--236 (2004; Zbl 1068.11033) and Duke Math. J. 114, 123--191 (2002; Zbl 1035.11018)] to solve the subconvexity problem for Rankin-Selberg \(L\)-functions in the level aspect. The authors prove the following theorem for the Rankin-Selberg \(L\)-function \(L(f\otimes g, s)\). Theorem: Let \(f\) and \(g\) be two primitive (either Maass or holomorphic) cusp forms of level \(q\), \(D\) and nebentypus \(\chi_f, \chi_g\) respectively. Suppose \(\chi_f\chi_g\) is not trivial. Then \[ L(f\otimes g,1/2+it)\ll q^{1/2-\delta}\qquad(t\in\mathbb R) \] where \(\delta= \frac1{2648}\) and the implied constant depends polynomially on \(|t|\), \(D\) and the weights or the Laplacian eigenvalues of \(f\) and \(g\). This new result extends to general \(g\) (both holomorphic and Maass cusp forms) and assures that the implied constant has at most polynomial growth in any other parameters. Interestingly some subconvexity bounds for twisted \(L\)-functions are established in order to achieve the goal. (A further investigation on the bounds for twisted \(L\)-functions is carried out in [Forum Math. 19, 61--105 (2007; Zbl 1168.11014)].) The paper under review is indeed very substantial, providing a discussion on recent developments of the subconvexity problem, a review of automorphic forms and an application to Heegner points on Shimura curves, in addition to the delicate proof. The proof makes use of various important tools and techniques, such as the functional equation of the Rankin-Selberg \(L\)-function, Voronoi summation formulae, spectral summation formulae, the amplification method and Jutila's method of overlapping intervals (a variant of the \(\delta\)-symbol method). A vital part is to tackle an independently interesting and important problem of (a sum of) shifted convolution sums, whose novel treatment starts with Jutila method.