Weak order for the discretization of the stochastic heat equation driven by impulsive noise (Q1935447)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Weak order for the discretization of the stochastic heat equation driven by impulsive noise |
scientific article |
Statements
Weak order for the discretization of the stochastic heat equation driven by impulsive noise (English)
0 references
15 February 2013
0 references
The authors study the weak convergence of numerical approximations to the solution of a linear parabolic stochastic partial differential equation on a Hilbert space \(H\). In abstract form the considered equation reads \[ dX_t+AX_tdt=Q^{1/2}dZ_t, \quad X_0=x\in H,\quad t\in[0,T], \] which is additively driven by a cylindrical impulsive stochastic process \(Z_t\) on \(H\), i.e., a process defined via a space-time Poisson random measure. The authors discuss thoroughly this class of processes and stochastic integration with respect to it. Essentially, the assumptions on the operators \(A\) and \(Q\) in the considered SPDE are standard regularity assumptions in the semigroup approach to SPDEs, i.e., \(A\) is a strictly positive, symmetric, unbounded operator on \(H\) the domain of which is dense and compactly embedded in \(H\) and the covariance operator \(Q\) is nonnegative, bounded and symmetric. Furthermore, it is assumed that there exists \(\alpha> 0\), \(\beta\in(1-\alpha,\alpha]\) such that \(A^{-\alpha}\) is a finite trace operator and \(A^\beta Q\) is bounded. Finally, key to establish the convergence result are suitable regularity assumptions on the intensity measure of the space-time Poisson random measure. For the numerical approximation the authors present a discretisation using the \(\theta\)-method in time, a finite element approximation in space and the noise process is discretised as a projection on the finite elements subspace of the noise increment. The main result is that under the above conditions for all \(\gamma<1-\alpha+\beta\leq 1\) and all \(\varphi\in C_b^2(H)\) the weak error estimate \[ |\operatorname{E}\varphi(X^N_h)-\operatorname{E}\varphi(X_T)|\leq C(h^{2\gamma}+(\Delta t)^\gamma) \] holds, where \(\Delta t=T/N\leq 1\) and the constant \(C\) is independent of \(N\) and \(h\). \(X_T\) is the value of the mild solution to the SPDE at time \(T\) and \(X_h^N\) is the numerical approximation corresponding to the time discretisation corresponding to an equidistant step-size \(T/N\) and finite element discretisation space with parameter \(h\). The paper follows (and extends) the study [\textit{A. Debussche} and \textit{J. Printems}, Math. Comput. 78, No. 266, 845--863 (2009; Zbl 1215.60043)] wherein an analogous convergence result is proven for an SPDE of the above type driven by a cylindrical Wiener process.
0 references
stochastic heat equation
0 references
additive noise
0 references
impulsive cylindrical noise
0 references
weak approximation
0 references
finite elements
0 references
\(\theta\)-method
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references