Abstract:We address an original approach for the convergence analysis of a finite-volume scheme for the approximation of a stochastic diffusion-convection equation with multiplicative noise in a bounded domain of $\mathbb{R}^d$ (with $d=2$ or $3$) and with homogeneous Neumann boundary conditions. The idea behind our approach is to avoid using the stochastic compactness method. We study a numerical scheme that is semi-implicit in time and in which the convection and the diffusion terms are respectively approximated by means of an upwind scheme and the so called two-point flux approximation scheme (TPFA). By adapting well-known methods for the time discretization of stochastic PDEs and combining them with deterministic techniques applied to spatial discretization, we show strong convergence of our scheme towards the unique variational solution of the continuous problem in $L^p(0,T;L^2(\Omega;L^2(\Lambda)))$, for any finite $p\geq 1$.
What problem does this paper attempt to address?
### What problem does this paper attempt to solve?
This paper aims to solve the problem of finite - volume approximation of a diffusion - convection equation with multiplicative random forces. Specifically, the authors focus on the numerical solution of the stochastic diffusion - convection equation with homogeneous Neumann boundary conditions in the bounded region \(\Lambda\subset\mathbb{R}^d\) (where \(d = 2\) or \(d=3\)). The equation has the form:
\[
du-\Delta u\,dt+\text{div}_x(vu)\,dt = g(u)\,dW(t)+\beta(u)\,dt\quad\text{in }\Omega\times(0,T)\times\Lambda
\]
where:
- \(u(0,\cdot)=u_0\) is the initial condition,
- \(\nabla u\cdot n = 0\) is the homogeneous Neumann boundary condition.
The main challenges include:
1. **Handling nonlinear terms**: The equation contains nonlinear diffusion and convection terms.
2. **Multiplicative noise**: The equation is affected by the multiplicative random noise \(g(u)\,dW(t)\).
3. **Selection of numerical methods**: Selecting numerical methods suitable for dealing with stochastic partial differential equations (SPDEs).
### Main contributions of the paper
1. **New convergence analysis method**: The authors propose a new method to prove the convergence of the finite - volume scheme, avoiding the use of traditional stochastic compactness methods. This makes the analysis more concise and does not require complex stochastic analysis tools.
2. **Considering convection terms**: Compared with previous work, this paper introduces the convection term \(\text{div}_x(vu)\), which is more common in practical applications, such as convection - diffusion problems in fluid mechanics.
3. **Relaxing assumptions**: Relative to previous work, this paper relaxes the requirements for the initial condition, only requiring \(u_0\in L^2(\Omega; L^2(\Lambda))\) instead of the stronger \(u_0\in L^2(\Omega; H^1(\Lambda))\).
4. **Three - dimensional space extension**: The convergence analysis is extended from two - dimensional space to three - dimensional space, increasing the universality of the results.
### Numerical methods
The authors adopt a method that is semi - implicit in time and uses the upwind scheme and the two - point flux approximation scheme (TPFA) to discretize the convection and diffusion terms in space respectively. By combining deterministic techniques (for spatial discretization) and known time - discretization methods (for stochastic PDEs), they prove that the numerical scheme strongly converges to the unique variational solution of the continuous problem in \(L^p(0,T; L^2(\Omega; L^2(\Lambda)))\).
### Conclusion
This research fills the gaps in the existing literature, proposes a new finite - volume method to approximate the diffusion - convection equation with multiplicative random noise, and successfully proves its convergence. This result is not only of great theoretical significance but also provides a powerful numerical tool for practical applications.