On stochastic partial differential equations with spatially correlated noise: smoothness of the law. (Q1888759): Difference between revisions

A stochastic partial differential equation \[ Lu(t,x) = a(u(t,x))\dot F(t,x) + b(u(t,x)), \quad t\geq 0, \;x\in \mathbb R^ {d}, \tag{1} \] with a zero initial condition is considered. It is assumed that \(F\) is a Gaussian generalized random field with mean zero and the covariance functional \[ J(\varphi ,\psi ) = \int _ {\mathbb R_ +}\int _ {\mathbb R^ {d}} \bigl (\varphi (s,\cdot )*\widetilde \psi (s,\cdot )\bigr )(x)\,d\Gamma (x) \,ds, \quad \varphi ,\psi \in {\mathcal D}(\mathbb R^ {d+1}), \] where \(\widetilde \psi (s,x) = \psi (s,-x)\) and \(\Gamma \) is a nonnegative and nonnegatively definite tempered measure. Let \(a,b\in C^ \infty (\mathbb R)\) have bounded derivatives of all orders, and suppose that \(\inf _ {v\in \mathbb R}| a(v)| > 0\). \(L\) is a second-order differential operator such that the fundamental solution \(S\) to \(Lu=0\) satisfies hypotheses introduced by \textit{R. C. Dalang} [Electron. J. Probab. 4, Paper No. 6 (1999; Zbl 0922.60056)] which are sufficient for a mild solution of (1) to exist. Let \(u\) be a solution to (1); under additional assumptions on \(S\) it is proven that for all \(t>0\) and \(x\in \mathbb R^ {d}\) the law of \(u(t,x)\) has a density \(p_ {t,x}\) with respect to Lebesgue measure on \(\mathbb R\), and \(p_ {t,x}\in C^ \infty (\mathbb R)\). The general theory is then applied to heat equations on \(\mathbb R^ {d}\), \(d\geq 1\), and to wave equations on \(\mathbb R\) and \(\mathbb R^ 2\).