Wong-Zakai type approximation of SPDEs of Lévy noise (Q2385537)

The author considers a Lévy-process driven stochastic partial differential equation of the form \[ dX(t)= (AX(t-)+f(X(t-))dt + \int_Z g(X(t-);z)\eta(dz;dt), \qquad X(0)=x_0.\tag{1} \] Here \(A\) is an operator generating a compact and analytic semi-group of contractions on a suitable Banach space and \(\eta\) is a Poisson random measure on a complete probability space. The aim of the article is to set up a Wong-Zakai type approximation to the solution of (1). Thus, firstly, the underlying path \(L(t)= \int_0^t \int_Z z \eta(dz;ds)\) of the driving noise is approximated by a linear interpolation type process \(L^{(n)}(t)= \sum_{i=1}^{[2^nt]} \zeta_i^n\), if \([2^nt] = [2^nT]/n\), and \(= \sum_{i=1}^{[2^nt]} \zeta_i^n + (t-[2^nt]/n)\zeta_{[tn]+1}^n\), if \([2^nt] \neq [2^nT]/n\), where \(\zeta_i^n\) is a sequence of identically distributed random variables given by \(\zeta_i^n= \int_Zz \eta(dz; ((i-1)2^{-n},i 2^{-n}])\) for \(i=1,\ldots, [2^nT]\). Then the Wong-Zakai approximation of (1) is defined as \[ dX^{(n)}(t)= (AX^{(n)}(t)+f(X^{(n)}(t))dt + g(X^{(n)}(t);\zeta_{[2^nt]+1}^n)\frac{dt}{2^{-n}}. \] For this equation the author proves that for all natural numbers \(n\) there exists a unique solution which is uniformly integrable in \(q\)-th mean and the laws of the family \(\{X^{(n)}\}\) form a tight sequence in an appropriate space. In the proof of the latter assertion the main ingredients are Aldous' condition and the compact containment condition.