Stochastic partial differential equations with unbounded and degenerate coefficients (Q627677)

This paper studies the problem of existence and uniqueness of solutions of a stochastic partial differential equation (SPDE) of parabolic type \[ du_{t}=({\mathcal L}_{t}u_{t}+f_{t})\, dt + ({\mathcal M}_{t}^{\ell}u_{t} + g_{t}^{\ell})\, dB_{t}^{\ell}, \] in \({\mathbb R}^{d}\), where \({\mathcal L}_{t}\) is the random parabolic operator \[ {\mathcal L}_{t}(\omega)u:=\partial_{i}(a_{t}^{ij}(\omega,t)\partial_{j}u)+ \partial_{i}(b_{t}^{i}(\omega,x)u)+c_{t}(\omega, x)u \] with unbounded and degenerate coefficients; \(B_{t}^{\ell}\), \(\ell \in {\mathbb N}\) is a collection of independent Wiener processes; and \[ {\mathcal M}_{t}^{\ell}(\omega)u:=\sigma_{t}^{i\ell}(\omega, x)\partial_{i}u+ h_{t}^{\ell}(\omega, x)u, \] where again the coefficients are allowed to be unbounded and degenerate. The forcing terms \(f\) and \(g\) are also assumed to be random and possibly unbounded. This class of SDPEs and its extensions arise in a number of important applications which include filtering theory. The author obtains existence and uniqueness results for this problem under a variety of assumptions on the coefficients and the data which are collected in Theorems 2.3, 2.4, 2.5, 2.6 of the paper, whose assumptions (among others) allow ``quadratic'' growth in \(x\) for \(a^{ij}\) and ``linear'' growth in \(x\) for \(b^{i}\). The approach to the problem is via an extension of the DiPerna-Lions theory of generalized or distribution solutions for SPDEs in conjuction with mollification techniques for the coefficients of the operator. After guaranteeing existence, under general classes of assumptions on the coefficients, the author proves a maximal principle for SPDEs, of interest in its own right, which is then applied to obtain uniqueness results for the generalized solutions. The problem of \(L^{1}\) integrability of generalized solutions as well as weak continuity, is treated under two alternative sets of conditions on the coefficients. An example from nonlinear filtering theory is treated in detail and extensions to nonlinear problems are discussed in the final section. The paper contributes important results to a problem of considerable interest in the field which will undoubtedly lead to interesting followup results to other types of SPDEs.