Numerical Approximation of Stochastic Time-Fractional Diffusion

Bangti Jin,Yubin Yan,Zhi Zhou

DOI: https://doi.org/10.48550/arXiv.1810.01822

2018-10-04

Abstract:We develop and analyze a numerical method for stochastic time-fractional diffusion driven by additive fractionally integrated Gaussian noise. The model involves two nonlocal terms in time, i.e., a Caputo fractional derivative of order $\alpha\in(0,1)$, and fractionally integrated Gaussian noise (with a Riemann-Liouville fractional integral of order $\gamma \in[0,1]$ in the front). The numerical scheme approximates the model in space by the Galerkin method with continuous piecewise linear finite elements and in time by the classical Grünwald-Letnikov method, and the noise by the $L^2$-projection. Sharp strong and weak convergence rates are established, using suitable nonsmooth data error estimates for the deterministic counterpart. Numerical results are presented to support the theoretical findings.

Numerical Analysis

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to develop and analyze numerical methods for solving stochastic time - fractional diffusion equations driven by additive fractional - order integral Gaussian noise. Specifically, the author focuses on the time - fractional diffusion equation in the following form:

\[
\partial_t^\alpha u(t) + A u(t) = {}_0I_t^\gamma \dot{W}(t), \quad 0 < t \leq T,
\]

where:
- \(\partial_t^\alpha\) represents the Caputo fractional - order derivative of order \(\alpha\in(0, 1)\);
- \(A\) is the negative Laplace operator \(-\Delta\) with zero Dirichlet boundary conditions;
- \({}_0I_t^\gamma\) represents the Riemann - Liouville fractional - order integral of order \(\gamma\in[0, 1]\);
- \(W(t)\) is a Wiener process.

### Main problems

1. **Development of numerical methods**: The author proposes a numerical scheme based on the Galerkin finite element method (spatial discretization) and the Grünwald - Letnikov method (temporal discretization), and uses the \(L^2\) projection to approximate the noise term.

2. **Convergence analysis**: The author establishes the optimal convergence rates of this numerical method in the sense of strong convergence and weak convergence. In particular, they prove the almost optimal strong convergence rate and weak convergence rate of the fully discrete approximation.

3. **Theoretical contributions**: By introducing non - smooth data error estimates, the author addresses the challenges different from those of classical diffusion equations due to the existence of fractional - order integrals. In addition, they also use Malliavin calculus tools to analyze weak convergence.

### Key challenges

- **Treatment of non - local terms**: The model contains two non - local terms (Caputo fractional - order derivative and fractional - order integral noise), which makes traditional numerical analysis methods no longer applicable.

- **Influence of noise**: Fractional - order integral noise has a significant impact on the convergence rate, especially in terms of strong convergence and weak convergence. The author analyzes in detail the influence of different parameters \(\alpha\) and \(\gamma\) on the convergence rate.

### Conclusion

This research fills the gap in the field of numerical analysis of stochastic time - fractional partial differential equations, especially when the noise is fractional - order integral Gaussian noise. The author not only provides effective numerical methods but also verifies the effectiveness and accuracy of these methods through rigorous mathematical analysis.