Stationary solutions of SPDEs and infinite horizon BDSDEs with non-Lipschitz coefficients (Q2269623): Difference between revisions

It is shown that Stochastic Partial Differential Equations (SPDEs) of the type \[ dv(t,x) = [\mathcal{L}v(t,x) + f(x, v(t,x), \sigma^*(x) Dv(t,x))]dt + g(x,v(t,x), \sigma^*(x) Dv(t,x))dB_t \] admit stationary weak solutions under certain assumptions on the coefficients. Here \(\mathcal{L}\) is a second order differential operator, more precisely it corresponds to the infinitesimal generator of a diffusion process. The stationary solution can be represented as the unique solution of the associated infinite horizon (Forward) Backward Doubly Stochastic Differential Equation (BDSDE): \[ e^{-Ks} Y_s^{t,x} = \int_s^\infty e^{-Kr}f(X_r^{t,x}, Y_r^{t,x}, Z_r^{t,x}) dr + \int_s^\infty K e^{-Kr}Y_r^{t,x} dr \] \[ - \sum_{j=1}^\infty \int_s^\infty e^{-Kr} g_j(X_r^{t,x}, Y_r^{t,x}, Z_r^{t,x}) d\hat\beta_j - \int_s^\infty e^{-Kr}\langle Z_r^{t,x}, dW_r \rangle \] The stationarity is pathwise: \(v_t(\omega) = v_0 (\theta_t \omega)\) for a measure-preserving transformation \(\theta_t\) on \(\Omega\). It is remarkable that with this approach one can treat equations with multiplicative noise \(g\) that depends on the derivative of \(v\). The coefficients are not assumed to be Lipschitz continuous. Instead they have to satisfy some boundedness conditions as well as a monotonicity condition.