Invariant measures for stochastic heat equations with unbounded coefficients. (Q2574540)

Existence of invariant measures for a semilinear stochastic heat equation \[ du = \{\lambda \Delta u + f(u)\}\,dt + \sigma (u)\, dW \quad \text{in \(\mathbb R_ {+}\times \mathbb R^ {d}\)} \tag{1} \] is studied. Suppose that \(\lambda >0\), \(\sigma :\mathbb R\to \mathbb R\) is globally Lipschitz continuous, \(f:\mathbb R\to \mathbb R\) is locally Lipschitz and satisfies \(yf(y)\leq \kappa (1+| y| ^ 2)\), \(| f(y)-f(x)| \leq c| y-x| (1+| y| ^ {\nu -1}+| x| ^ {\nu -1})\) for some constants \(c>0\), \(\kappa >0\) , \(\nu \geq 1\) and all \(x,y\in \mathbb R\). The driving process \(W\) is a Wiener process in \(L^ 2(\mathbb R^ {d})\) with a nuclear covariance operator \(\sum ^ \infty _ {j=1} a_ {j}e_ {j}\otimes e_ {j}\), where \(a_ {j}\), \(j\geq 1\), are nonnegative numbers satisfying \(\sum ^ \infty _ {j=1} a_ {j}< \infty \) and \(\{e_ {j}\}_ {j\geq 1}\) is an orthonormal basis in \(L^ 2(\mathbb R^ {d})\) such that \(\sup _ {j\geq 1}\| e_ {j}\| _ {\infty } <\infty \). If \(d=1\), a standard cylindrical Wiener process on \(L^ 2(\mathbb R^ {d})\) may be considered as well. For \(p\geq 2\), \(\rho >d\) let us denote by \(L^ {p}_ \rho \) the weighted space \(L^ {p}(\mathbb R^ {d};(1+| x| ^ 2)^ {- \rho /2}\text dx)\). If \(p=\max (2,\nu )\), then the equation (1) defines a Markov process \(\mathbf u\) on \(L^ {p}_ \rho \). It is shown that if \(\rho \) and \(\overline \rho \) are such that \(\rho >\overline \rho + d\), \(\overline \rho >d\), then the process \(\mathbf u\) on \(L^ {p}_ \rho \) has an invariant measure, provided that there exists a solution of (1) which is bounded in probability in the space \(L^ {p\nu }_ {\overline \rho }\). Moreover, by using a comparison theorem it is proved that (1) admits a solution bounded in probability if \(uf(u)\leq C-\mu | u| ^ 2\), \(u\in \mathbb R\), for a sufficiently large constant \(\mu \).