Stochastic integration with respect to cylindrical Lévy processes (Q682265)

A process \((L_t)_{t\geq 0}\) on a Hilbert space \(H\) is a ``cylindrical Lévy process'' when its finite-dimensional marginal projections \((L_th_1,\ldots, L_th_n)_{t\geq 0, n\in\mathbb{N}^*, h_1,\ldots h_n\in H}\) are \(\mathbb{R}^n\) valued Lévy processes. This definition is justified by the non-existence of a proper \(H\)-valued Brownian motion. The authors consider càdlàg adapted stochastic integrands \(\Psi\) taking values in the space \(\mathcal{L}(H,K)\) of Hilbert-Schmidt operators (for another Hilbert space \(K\)). They show that a stochastic integral \(\int_0^t \Psi_s\,dL_s\,\) makes sense in \(K\), by precisely establishing that the limit of sequences \(\int_0^t \Psi_s^n\,dL_s\,\) (where the \(\Psi^n\) are elementary integrands converging to \(\Psi\) in probability in the Skorokhod metric), exists in probability, does not depend on the sequence \(\Psi^n\), and moreover defines an adapted càdlàg semimartingale. Neither the integrator nor the integrands are required to satisfy any moment or boundedness condition.