Converse theorems assuming a partial Euler product (Q2426719)

The context for this article, its predecessor [\textit{J. B. Conrey} and \textit{D. W. Farmer}, Int. Math. Res. Not. 1995, No. 9, 445--463 (1995; Zbl 0849.11042)] and several further papers of the past four decades on modular forms and Dirichlet series with functional equations was established by \textit{A.Weil}'s often-sited work dedicated to C. L. Siegel [Math. Ann. 168, 149--156 (1967; Zbl 0158.08601)]. This context is described succinctly in the Conrey-Farmer paper, which I paraphrase and quote below. Let \(F(A)= \sum^\infty_1 a_n n^{-s}\), a Dirichlet series convergent in some right half-plane and extendable to an entire function such that \(\Gamma(s)F(s)\) remains entire and is, furthermore, bounded in vertical strips. (Call this restriction EBV.) For \(N\in\mathbb Z^+\) and real \(k\), define the completed Dirichlet series by \[ \Phi(s)= \Biggl({2\pi\over\sqrt{N}}\Biggr)^{-s}\Gamma(s) F(s). \] The interest here is in \(F(s)\) such that \(\Phi(s)\) satisfies the functional equation \[ \Phi(s)= \pm e^{\pi ik/2}\Phi(k-s).\tag{1} \] For \(f\) a function defined in the upper half-plane \(H= \{z: \text{Im}\,z> 0\}\) and \(\gamma=\left(\begin{matrix} a & b\\ c & d\end{matrix}\right)\) in \(\text{GL}_2(\mathbb R)^+\), define the ``slash operator'', \[ (f|_k\gamma)(z):= (\det\,\gamma)^{k/2}(cz+ d)^{-k} f(\gamma z). \] Also let \(\Gamma_0(N)\) denote the Hecke congruence group \[ \Gamma_0(N):= \left\{\begin{pmatrix} a & b\\ c & d\end{pmatrix}\in \text{SL}_2(\mathbb Z): N\mid c\right\}. \] To the Dirichlet series \(F(s)\) associate the exponential series \(f(z)= \sum^\infty_1 a_n e^{2\pi inz}\). (The assumption on \(F(s)\) implies that \(f(z)\) is holomorphic in \(H\).) Then the usual converse argument in the Riemann-Hecke correspondence shows that \[ \left(f|_k\begin{pmatrix} 0 & -1\\ N & 1\end{pmatrix}\right)(z)= \pm e^{\pi ik/2} f(z). \] Since \(f(z+ 1)= \left(f|_k\begin{pmatrix} 1 & 1\\ 0 & 1\end{pmatrix}\right)(z)= f(z)\), we have Hecke's converse theorem. Assume \(1\leq N\leq 4\). If \(F(s)\) is EBV and satisfies the functional equation (1), then \(f|_k\gamma= f\) for all \(\gamma\) in \(\Gamma_0(N)\). This follows from the fact that \[ \left\langle\begin{pmatrix} 1 & 1\\ 0 & 1\end{pmatrix}, \begin{pmatrix} 0 & -1\\ N & 1\end{pmatrix}\right\rangle\supset\Gamma_0(N) \] for \(1\leq N\leq 4\). Conrey and Farmer comment: ``If \(N\geq 5\), then the above argument fails \(\dots\)\ . In order to get the desired conclusion that \(f\) is invariant under \(\Gamma_0(N)\), we must put further restrictions on \(F\). Weil [loc. cit.] conceived of the important idea of requiring that the twists of \(F\) by Dirichlet characters satisfy an appropriate functional equation. Later versions of \textit{M. J. Razar} [Trans. Am. Math. Soc. 231, 489--495 (1977; Zbl 0364.10015)] and \textit{W.-C. W. Li} [Recent progress in analytic number theory, Symp. Durham 1979, Vol. 2, 119--153 (1981; Zbl 0459.10019)] reduced the number of twists to a finite number depending on \(N\).'' It should be pointed out that the twists in these latter articles are not exactly the same as those Weil introduced. Conrey and Farmer go on, ``in this paper we have partial success at replacing the assumption on twists of \(F\) by the assumption of \(F\) having an Euler product of the appropriate form.'' Indeed, their theorem (extends the Hecke converse theorem above) to the values \(5\leq N\leq 12\), \(14\leq N\leq 17\), \(N= 23\). For each of these values of \(N\) the Conrey-Farmer proof uses the Euler product at only finitely many primes. The Farmer-Wilson paper under review continues the work of Conrey-Farmer by assuming (a modified, equivalent version of) the functional equation (1) and an Euler product factor at the prime 2 only: \[ L(s,f):= F(s)= (1- a_2 2^{-s}+ 2^{k-1-2s})^{-1} \sum_{n \text{odd}} a_n n^{-s}.\tag{2} \] The authors prove an interesting result (Theorem 2.1) which is, however, conditional upon two ``Pairing Assumptions'' and a weak form of Artin's conjecture on primitive roots (stated precisely in \S4). To clarify this result, let \(T= \left(\begin{matrix} 1 & 1\\ 0 & 1\end{matrix}\right)\), \(T_2=\) the index 2 Hecke operator and \(H_N= \left(\begin{matrix} 0 & -1\\ 1 & N\end{matrix}\right)\), the Fricke involution. Then Theorem 2.1 may be stated as follows: Under the three assumptions mentioned above, if \(f\) satisfies \(f|_k T= f\), \(f|_k T_2= f\) and \(f|_k H_N= f\), then (3) \(f|_k\gamma= \chi(d) f\) for all \(\gamma= \begin{pmatrix} * & *\\ * & d\end{pmatrix}\) in \(\Gamma\), where \(\Gamma\) is a subgroup of \(\text{SL}(2, \mathbb Z)\) such that \(\Gamma_1(N)\subset\Gamma\subset\Gamma_0(N)\). Here \(\chi\) is a Dirichlet character modulo \(N\). Note that: (i) \(\Gamma\) has finite index in \(\text{SL}(2, \mathbb Z)\); (ii) the Dirichlet character arises in the definition of the Hecke operator \(T_2\). The pairing assumptions are essential to the authors' derivation of (3). They are stated explicitly in \S3.2.