Non-linear twists of \(L\)-functions: a survey (Q627055): Difference between revisions

The author gives a very interesting survey, about non-linear twists, in the by-now classical environment of ``Selberg Class'', abbrev. SC, i.e., a class of \(L\)-functions axiomatized, in the 1989 paper [Proceedings of the Amalfi conference on analytic number theory, Maiori, Amalfi, Italy, 1989. Salerno: Universitá di Salerno, 367--385 (1992; Zbl 0787.11037)], by \textit{A. Selberg}. We will not quote these axioms, only say that SC contains the most important \(L\)-functions, like Riemann zeta, Dedekind zeta functions \(L_K(s)\) (associated to algebraic number fields \(K\)), the Dirichlet \(L(s,\chi)\) and, resp., the Hecke \(L\)-functions \(L_K(s,\chi)\) with primitive Dirichlet, resp., Hecke, characters \(\chi\) and the normalized \(L\)-functions \(L(s,f)\) associated with newforms \(f(z)\) for the Hecke groups. Also, under very famous (still) unproved hypotheses, like the Artin Conjecture and, resp., the Ramanujan Conjecture, we have that the Artin \(L\)-functions, resp., the cuspidal \(\mathrm{GL}(n)\) automorphic \(L\)-functions, belong to SC. Also, we have a more general setting (missing \(2\) axioms out of \(5\)), the ``Extended Selberg Class'', abbrev. ESC. (Just to give an idea, a typical function in ESC is a real coefficients linear combination of elements in SC, satisfying the same Functional Equation, FE; FE is one of the common ESC and SC axioms.) One important ``invariant'', in ESC, is the degree of \(F\), say \(d=d_F\geq 0\), which is related to the Functional Equation and, at least for the time being, only conjecturally an integer. After a short parade of SC and ESC known results, given \[ F(s)=\sum_{n=1}^{\infty}{{a(n)}\over {n^s}}, \quad F_d(s,\alpha):=\sum_{n=1}^{\infty}{{a(n)}\over {n^s}}e(-\alpha n^{1/d}), \; \alpha>0, \] where \(d=d_F\) is the degree of \(F\) in ESC, is defined as the ``standard (non-linear) twist'', say, of our \(F\). Here, as usual, \(e(\beta):=e^{2\pi i\beta}\) is the additive character; terminology ``non-linear'', here, is to distinguish the linear twist with \(e(-n\alpha)\) (that is linked, via Gauss sums, to primitive Dirichlet characters). Then, Theorem 1 (from a joint work with Kaczorowski) gives analytic informations on \(F\) of degree \(d>0\) in ESC, starting from its standard twist (saying also it is always meromorphic over \(\mathbb C\)): based on the nature of \(\alpha>0\), we may distinguish: \(F\) is entire OR has (at most simple) poles, located at (rational) points depending on \(d\), with residues depending on \(d\), the conductor \(q=q_F\) (another important invariant) and \(\alpha\). The (difficult and technical) proof, whose sketch is given here, goes through the known analytic properties of ``incomplete hypergeometric functions''. The list of consequences of this Theorem looks impressive and goes in the direction of the very important problem of classifying both SC and ESC (and for ESC is a much harder task: compare the paper's Remark 3): 1) prove the \textit{degree Conjecture}: \(d\) is always an integer (in SC or, even more, in ESC); 2) then, prove the \textit{general converse problem}: SC with \(d\) fixed is the set of degree \(d\) automorphic \(F(s)\). (As the author points out, 2) is a more general instance of the so-called ``Langlands program'', here!) First of all, they (P. and Kaczorowski) give (as a consequence of Theorem 1) the following (strong) version of Hamburger's Theorem (characterizing Riemann's \(\zeta\)): if \(F\) is in ESC with \(d_F=q_F=1\), then \(F=c \zeta\); i.e., it's a constant (complex) multiple of Riemann's zeta. Then, they prove in a very short way (even with one-line proofs!) a long list of known results (and proved in many other ways!); just to give an idea, I will quote the following results from the survey: {\parindent=6mm {a)} Theorem 2, a characterization of degree \(d_F=0\) functions \(F\) in SC/ESC (first, by Conrey and Ghosh); {b)} Theorem 3, saying that \(0<d<1\) is never possible in ESC (again Conrey and Ghosh, after 50s works); {c)} Theorem 4, a complete characterization of both SC and ESC for the degree one; {d)} Theorem 5, giving an Omega-estimate for the remainder of \(a(n)\)-partial sums (up to \(x\to \infty\)) as the \(s=1\) residue of \(x^sF(s)/s\), which is ``due essentially to K. Chandrasekharan and R. Narasimhan''. } In search of degrees \(d>1\) classification for the SC (and ESC), the standard non-linear twist has been generalized by the author and Kaczorowski to the (theory of) multidimensional non-linear twists, see paper's \(\S3\). Last but not least, we mention, as an application of these multidimensional non-linear twists, the beautiful result in the direction of proving the degree Conjecture: ESC is empty, as \(1<d<2\), as proved by \textit{J. Kaczorowski} and \textit{A. Perelli}, see [Ann. Math. (2) 173, No. 3, 1397--1441 (2011; Zbl 1235.11085)]!