On implicit and explicit discretization schemes for parabolic SPDEs in any dimension (Q2485474): Difference between revisions

Consider approximation methods for solutions \(u=u(t,x)\) of stochastic partial differential equations \[ \frac{\partial u}{\partial t} = \Delta u + b(t,x,u) + \sigma (t,x,u) \frac{\partial^2 F}{\partial t \partial x} \] with homogeneous boundary condition \(u| _{\partial D} = 0\) and initial condition \(u(0,x)=u_0(x)\) on \(D = [0,1]^d\) with dimension \(d \geq 1\), driven by \(L^2(P)\)-valued centered, space-correlated Gaussian process \(F\). The authors present certain equidistant explicit and implicit discretization schemes for \(u(t,x)\). The rate of \(L^p(P)\)-convergence of those discretizations is studied for \(p \geq 2\), and how the correlation of \(F\) has influence on it. The proofs are based on the product form of related Green function and its approximation. Some numerical results are presented too.