Density convergence of a fully discrete finite difference method for stochastic Cahn--Hilliard equation

Jialin Hong,Diancong Jin,Derui Sheng
2023-08-17
Abstract:This paper focuses on investigating the density convergence of a fully discrete finite difference method when applied to numerically solve the stochastic Cahn--Hilliard equation driven by multiplicative space-time white noises. The main difficulty lies in the control of the drift coefficient that is neither globally Lipschitz nor one-sided Lipschitz. To handle this difficulty, we propose a novel localization argument and derive the strong convergence rate of the numerical solution to estimate the total variation distance between the exact and numerical solutions. This along with the existence of the density of the numerical solution finally yields the convergence of density in $L^1(\mathbb{R})$ of the numerical solution. Our results partially answer positively to the open problem emerged in [J. Cui and J. Hong, J. Differential Equations (2020)] on computing the density of the exact solution numerically.
Numerical Analysis
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the convergence problem of the probability density function (PDF) of the numerical solution of the stochastic Cahn–Hilliard equation. Specifically, the author is concerned with whether the probability density function of the numerical solution obtained by the fully - discrete finite - difference method can effectively approximate the probability density function of the exact solution when solving the stochastic Cahn–Hilliard equation driven by multiplicative white - in - time noise. ### Main problems and challenges of the paper 1. **Non - global Lipschitz condition**: The drift coefficient in the stochastic Cahn–Hilliard equation is neither globally Lipschitz continuous nor one - sided Lipschitz continuous. This brings difficulties to the analysis of the strong convergence of the numerical solution and the convergence of the probability density function. 2. **Choice of numerical method**: To deal with the above problems, the author chooses a fully - discrete finite - difference method (FDM) and uses the implicit Euler method for discretization in time. 3. **Convergence of probability density function**: The author needs to prove that the probability density function of the numerical solution converges to the probability density function of the exact solution in the $L^1(\mathbb{R})$ norm. This is a relatively less - studied problem, especially for stochastic partial differential equations (SPDEs) with polynomial nonlinearity and multiplicative noise. ### Solutions To solve these problems, the author takes the following measures: - **Localization argument**: A new localization argument method is introduced to deal with the non - globally monotonic coefficient $\Delta f$. Through this method, the author transforms the estimation problem of the total variation distance into the estimation problem of its localized version. - **Strong convergence rate**: By using the interpolation method and the auxiliary process $\tilde{U}(t)$, the author obtains the strong convergence rate of the numerical solution and further deduces the convergence of the probability density function. - **Estimation of Malliavin derivative**: To control the total variation distance, the author applies the negative moment estimation of the Malliavin derivative and the convergence results in the Malliavin - Sobolev space. ### Main contributions 1. **Optimal strong convergence rate**: The optimal strong convergence rate of the fully - discrete finite - difference method for the stochastic Cahn–Hilliard equation with polynomial nonlinearity and multiplicative noise is given. 2. **Density convergence**: For the first time, the convergence of the probability density function of the numerical solution of SPDEs with polynomial nonlinearity is proved, which partially solves the open problem of the numerical method for calculating the probability density function of the exact solution. 3. **Localization criterion**: A criterion for transforming the total variation distance of a random variable into its localized version is proposed and successfully applied to derive the convergence of the probability density function of the fully - discrete numerical method. This method is also considered applicable to other SPDEs with non - global Lipschitz coefficients, such as the stochastic Allen - Cahn equation. In conclusion, this paper solves the complex problem of the convergence of the probability density function of the numerical solution of the stochastic Cahn–Hilliard equation by introducing innovative methods and techniques, providing new ideas and tools for research in related fields.