Influence of the regularity of the test functions for weak convergence in numerical discretization of SPDEs (Q2283124)

The paper studies the order of weak convergence of the numerical method for stochastic partial differential equations in dependence on the regularity of test functions. The authors focus on linear parabolic stochastic equation with additive space-time white noise, which is equipped with homogeneous Dirichlet boundary conditions and where the domain for space variable is one dimensional. This equation is solved using the spectral Galerkin method for spatial discretization and the backward Euler scheme for temporal discretization. First, the class of bounded continuous functions is considered, and it is proved that in this case, the supremum of the weak error over this class of functions does not converge to zero. Then, the class of bounded Lipschitz continuous functions is studied, and it is shown that the order of convergence for the weak error is equal to the order of convergence for the strong error. These results are different from the previous results for at least twice continuously differentiable functions, where the order of convergence for the weak error is twice as large as the order of convergence for the strong error.