Surjectivity of Hadamard type operators on spaces of smooth functions (Q2314687)

An Hadamard operator on a space of (generalized) functions is an operator such that all monomials are eigenvectors. The present paper, which continues work by the same authors in [Trans. Am. Math. Soc. 372, No. 9, 6017--6086 (2019; Zbl 1478.46019)], is a comprehensive study of Hadamard operators on \(C^\infty(\Omega)\). The main result is Theorem 5.3 which states, for open, symmetric and multiplicatively convex \(\Omega\), that an Hadamard operator \(H_T : C^\infty(\Omega) \to C^\infty(\Omega) \) is surjective if and only if its Mellin transform is slowly decreasing and has the division property, the open set \(\Omega\) is deeply \(H_T\)-convex, and zero is not an eigenvalue. We will explain these notions in the sequel. The central tool in these investigations is the so-called multiplicative convolution operator. In consequence, the Mellin transform replaces the Fourier-Laplace transform. Mellin transforms of Hadamard operators are holomorphic in a neighborhood of infinity. The authors then adapt the notion of a slowly decreasing function, which was first employed in [\textit{L. Ehrenpreis}, Am. J. Math. 82, 522--588 (1960; Zbl 0098.08401)] to the present situation. It is shown that, even in the case \(\Omega = \mathbb R^d \), this property alone is not sufficient for surjectivity. Therefore, the division property is introduced. Roughly speaking, it means the following: If \( \mathcal H_{\mathcal M} \) is the space of all Mellin transforms, \( \mathcal M(T) \) is the Mellin transform of \(T\), and if \( g \in \mathcal H_{\mathcal M} \) is defined in some neighborhood \(U\) of infinity and can be divided by \( \mathcal M(T) \) in some arbitrily small neighborhood of \(\infty\), then it can already be divided in a fixed neighborhood that only depends on \(U\). Deep \(H_T\)-convexity should be seen as an analogue of \(P(D)\)-convexity in the case of partial differential operators. Note that \(\mathbb R^d\) is deeply \(H_T\)-convex for all \(T\). En route to this result, several other interesting results are obtained. One is in Theorem 3.7 the characterizarion of moment sequences of distributions with compact support. Another one is the theorem of supports for the multiplicative convolution. The latter is subtler than the one in the additive case due to the special role of zero.