Hecke's converse theorem (Q1187138)

This article treats the Riemann-Hecke-Weil correspondence \(F(z)\leftrightarrow\varphi(s)\) where \(F(z)\) is a modular form and \(\varphi(s)\) is a Dirichlet series with a single gamma-function factor, satisfying a Riemann functional equation. (\(F\) and \(\varphi\) are related by the Mellin transform.) The setting, broader than the usual one, is informed by Eichler cohomology: ``modular integrals'' with a cocycle of ``period functions'' replace the modular forms. This work thus brings to mind \textit{S. Bochner}'s treatment of modular relations with ``residual functions''. [See Ann. Math. (2) 53, 332--363 (1951; Zbl 0042.32101) and related later articles of Bochner.] Relevant as well are works of the reviewer [Duke Math. J. 45, 47--62 (1978; Zbl 0374.10014); Glasg. Math. J. 22, 185--197 (1981; Zbl 0459.10017) and others] which study the Hecke correspondence within the context of modular integrals. The author identifies several important special examples of spaces \(V\) suitable to contain the cocycle of period functions arising from a modular integral \(F(z)\) associated, as above, to a Dirichlet series \(\varphi(s)\). (In the classical Hecke correspondence, \(V=\{0\}.)\) Such \(V\) are naturally required to be invariant under the stroke operator by elements of \(\mathrm{GL}(2,\mathbb{Q})\) (considered as linear fractional transformations). Spaces \(V\) with the additional property that they contain no nonconstant periodic functions play a special role within the theory. The principal result is the ``Converse Theorem'' (Theorem 8.1), which generalizes the Hecke converse theorem -- or, more properly, Weil's extension of Hecke's converse theorem -- to the congruence subgroups \(\Gamma_0(N)\). In contrast to Weil's theorem, here \(\varphi(s)\) may have any (finite) number of poles in the plane, but it is restricted by the assumption that it has a ``partial Euler product'' (i.e., \(\varphi(s)\) is expressible as a single Euler factor multiplied by a related Dirichlet series). Then, in the presence of the usual functional equations for the ``twists'' of \(\varphi(s)\) by Dirichlet characters, the conclusion is virtually the same as in Weil: \(F(z)\) is a modular form on \(\Gamma_0(N)\). The weight \(k\) of \(F\) is in \(\mathbb{Z}^+\). For \(k=2\) the conclusion must be weakened slightly: \(F(z)+bE_2(z)\) is a modular form of weight 2 on \(\Gamma_0(N)\), where \(b\) is a constant and \(E_2(z)\) is the usual Eisenstein series of weight 2 on the full modular group. \((E_2(z)\) is a modular integral rather than a modular form.) In the author's ``Converse Theorem'' the assumption that a partial Euler product exists cannot be removed.