The symmetric cube (Q1923250): Difference between revisions

Let \(\pi\) be a cuspidal automorphic representation of \(GL(2,{\mathbb{A}})\), where \({\mathbb{A}}\) is the adele ring of a global field \(F\). Suppose further that \(F\) contains the cube roots of unity and that the central character \(\chi=\prod_v \chi_v\) of \(\pi\) is unitary. In this paper, the authors construct and analyze a Rankin-Selberg integral for the symmetric cube \(L\)-function \(L(s,\pi,\vee^3)\). The meromorphic continuation of this \(L\)-function was first obtained by Langlands and Shahidi [\textit{F. Shahidi}, Compos. Math. 70, 243-273 (1989; Zbl 0684.10026)], by the study of Eisenstein series on the exceptional group \(G_2\). In particular Shahidi, making use of his theory of local coefficients, was able to establish a functional equation with \(L\)-factors at every place, and to prove that this \(L\)-function has no poles for Re\((s)\geq1\), except when \(\pi\) is a monomial representation attached to a Hecke character of a quadratic extension and \(s=1\). The Rankin-Selberg integral exhibited here allows the authors to improve this: Suppose that for every archimedean place \(v\) of \(F\), \(\chi_v\) is trivial, and let \(S\) be a finite set of places of \(F\) containing all archimedean places and all places where \(\pi_v\) is not a spherical principal series representation. Then the authors show that the partial \(L\)-function \(L_S(s,\pi,\vee^3)\) has no pole with Re\((s)\geq 3/4\) except possibly at \(s=1\). This improvement has an important application: the standard \(L\)-function \(L(s,\pi)\) does not have any Siegel zeroes. This follows from the result cited together with the work of Hoffstein-Ramakrishnan, which in turn builds on previous work of Gelbart-Jacquet, Bump-Ginzburg, Hoffstein-Lockhart, and Goldfeld-Hoffstein-Lieman. The Rankin-Selberg integral exhibited is one case of a general, still mostly undeveloped, principle originating in the work of \textit{D. Bump} and \textit{J. Hoffstein} [Duke Math. J. 53, 1047-1072 (1986; Zbl 0613.10028)]. To describe this, suppose for the moment that the base field contains the \(n\)-th roots of unity, and let \(\theta_n\) be the theta function on the \(n\)-fold cover of \(GL(n)\). This object is the residue of an Eisenstein series on the \(n\)-fold cover of \(GL(n)\), does have a Whittaker model, and may be described as a representation of small Gelfand-Kirillov dimension [\textit{D. A. Kazhdan} and \textit{S. J. Patterson}, Publ. Math., Inst. Haut. Etud. Sci. 59, 35-142 (1984; Zbl 0559.10026)]. Then the Bump-Hoffstein principle is that convolving (in the Rankin-Selberg sense) an automorphic object with \(\theta_n\) is always Eulerian! For \(n=2\), this is already a theme in the work of Shimura; the new idea here is the case \(n>2\). This generalization is surprising, since \(\theta_2\) may be constructed using the Weil representation, while for \(n>2\) no such construction is known. Though the theory of Rankin-Selberg convolutions for automorphic forms on various algebraic groups is somewhat developed, most of these Rankin-Selberg integrals do not carry over to the metaplectic context. Thus carrying out the convolution envisioned by Bump-Hoffstein, even in low-rank cases, typically requires new ideas. Returning to the paper under review, the authors achieve the convolution of \(\pi\) with \(\theta_3\). To do so, they introduce and study a new Rankin-Selberg integral (not the standard one for \(GL(2)\times GL(3)\)). Namely, they induce \(\theta_3\) up to a metaplectic Eisenstein series \(E_s\) on the cubic cover of \(GSp(6)\), via the Siegel parabolic. They then integrate over a subgroup of the form \(GL(2)U\), where \(GL(2)\) is embedded diagonally and \(U\) is a unipotent subgroup. The integrand is \(E_s\) times a function in the space of \(\pi\) and a character of \(U\). The authors analyze this convolution by unfolding, and carry out the unramified computation. In this, they make use of the Whittaker values of \(\theta_3\), computed by Bump-Hoffstein and by Proskurin. The local factor at a good place is shown to be the quotient of \(L(s,\pi_v,\vee^3)\) by the product of two \(GL(1)\) \(L\)-functions. They next analyze the local integrals at the bad places, showing for the nonarchimedean places that they have analytic continuation and that the local data may be chosen so as to make them nonzero at a given point. At the archimedean places they prove absolute convergence in a certain right-half-plane. Combining this with a list of the possible poles of the Eisenstein series \(E_s\), the theorem follows. Further study of the archimedean integral could lead to further improvements in the study of the poles. The authors note that, ``Except for the previously mentioned examples of Bump and Hoffstein, the present paper is the first known example of an Eulerian Rankin-Selberg integral employing an \(n\)-fold metaplectic cover with \(n>2\).'' This was the case when the core calculations described here were carried out, but other such examples are now known. \textit{T. Goezte} [Euler products associated to metaplectic automorphic forms on the 3-fold cover of \(GSp(4)\), Trans. Am. Math. Soc. (in press)] gives such an integral by convolving a generic automorphic form on the cubic cover of \(GSp(4)\) with \(\theta_3\). \textit{T. Suzuki} [Metaplectic Eisenstein series and the Bump-Hoffstein conjecture (preprint)] has convolved automorphic forms on the \(n\)-fold cover of \(GL(r)\) with \(\theta_n\) for any \(n\) to obtain Euler products. And the reviewer and \textit{D. Bump} [Metapletic generating functions and Shimura integrals, Contemp. Math. (to appear)] have analyzed such integrals by the study of non-unique models in certain cases.