Lévy processes and Lévy white noise as tempered distributions (Q682269)

It is well known that a Gaussian white noise in \(\mathbb{R}^d\) is a generalized random field that can be seen as a random element of \(\mathcal{S}'(\mathbb{R}^d)\). An analogous result in the context of Lévy processes has been shown in [\textit{J. Fageot} et al., J. Fourier Anal. Appl. 20, No. 6, 1179--1211 (2014; Zbl 1306.60036)], provided the associated Lévy measure has finite absolute moment of some order. In the paper under review, the authors show that this condition is also necessary. More precisely, in the single-parameter case, they show that a Lévy white noise \(\dot{X}\) (corresponding to the Lévy process \(X\) and defined as \(\langle \dot{X}(\omega),\varphi \rangle:=-\int_{\mathbb{R}_+}X_t(\omega)\varphi'(t) dt\), \(\omega\in\Omega\), \(\varphi\in\mathcal{D}(\mathbb{R})\)) satisfies \(\dot{X}\in\mathcal{S}'(\mathbb{R})\) \(\mathbb{P}\)-a.s. if, and only if, the associated Lévy measure \(\nu\) satisfies \(\int_{\{|x|\geq1\}}|x|^\eta\nu(dx)<\infty\) for some \(\eta>0\). Analogously, in the multi-parameter case, they show that a Lévy white noise \(\dot{X}\) (corresponding to the \(d\)-parameter Lévy field \(X\) and defined as \(\langle \dot{X}(\omega),\varphi \rangle:=(-1)^d\int_{\mathbb{R}^d_+}X_t(\omega)\frac{\partial^d \varphi}{\partial t_1\cdots\partial t_d}(t) dt\), \(\omega\in\Omega\), \(\varphi\in\mathcal{D}(\mathbb{R}^d)\)) satisfies \(\dot{X}\in\mathcal{S}'(\mathbb{R}^d)\) \(\mathbb{P}\)-a.s. if, and only if, the associated Lévy measure \(\nu\) again satisfies \(\int_{\{|x|\geq1\}}|x|^\eta\nu(dx)<\infty\) for some \(\eta>0\).