Convolution with the kernel \(e^{s\langle x\rangle^q}\), \(q\geq 1\), \(s>0\) within ultradistribution spaces (Q2050140)

The present paper provides a self-contained introduction to the convolution for distribution spaces with the Gaussian function. Such distributions were characterised in [\textit{P.~Wagner}, Math. Ann. 276, 467--485 (1987; Zbl 0592.46035)] by a growth condition. The same property was studied in [\textit{S.~Pilipović} and \textit{B.~Prangoski}, J. Math. Pures Appl. (9) 103, No.~2, 472--503 (2015; Zbl 1307.47052)] for non quasianalytic ultradistributions. The authors of the present paper propose to study convolution with a kernel beyond a general theory which is developed by [\textit{S. Pilipović} et al., J. Math. Pures Appl. (9) 116, 174--210 (2018; Zbl 1415.46028)]. In Section~2 the Gel'fand-Shilov test function space and its dual ultradistribution space is defined, moreover convolution of Gel'fand-Shilov ultradistributions under certain conditions is given in Theorem~2.1. As for the convolution with Gaussians $\exp^{s|x-t|^2}$ as kernel, this splits the product into two parts, which is impossible if $q\neq 2$. The bounds on the growth of the derivative of $\exp^{s \langle \cdot \rangle^q}$ are estimated in Lemmas 3.2--3.3. A sufficient condition for the convolution with $\exp^{s \langle \cdot \rangle^q}$ for the ultradistribution space with prescribed conditions is obtained in Theorem~3.6. The non-quasianalytic ultradistribution space when $q = 2$ is also considered. A necessary condition for the existence of the convolution with kernel $\exp^{s \langle \cdot \rangle^q}$ is studied in Section~4. As a consequence of Theorem~3.6 with condition $q=1$ a characterization of the Roumieu type ultradistributions which are convolvable with the exponential kernel are given in Corollaries 4.5--4.7, under certain conditions. The references of the paper are valuable to the researchers interested in convolutions and ultradistribution spaces.