Strong and weak convergence rates of a spatial approximation for stochastic partial differential equation with one-sided lipschitz coefficient

Jianbo Cui, Jialin Hong

Research output: Journal article publicationJournal articleAcademic researchpeer-review

17 Citations (Scopus)

Abstract

Strong and weak approximation errors of a spatial finite element method are ana- lyzed for the stochastic partial differential equations (SPDEs) with one-sided Lipschitz coefficients, including the stochastic Allen{Cahn equation, driven by additive noise. In order to give the strong convergence rate of the finite element method, we present an appropriate decomposition and some a priori estimates of the discrete stochastic convolution. To the best of our knowledge, there have been no essentially sharp weak convergence rates of spatial approximations for parabolic SPDEs with non-globally Lipschitz coefficients. To investigate the weak error, we first regularize the original equation by the splitting technique and obtain the regularity estimate the corresponding regularized Kolmogorov equation. Meanwhile, we present the refined estimates and the regularity estimate in the Malliavin sense of the finite element methods. Combining with the regularity of regularized Kolmogorov equation and Malliavin integration by parts, the weak convergence rate is shown to be twice the strong convergence rate.

Original languageEnglish
Pages (from-to)1815-1841
Number of pages27
JournalSIAM Journal on Numerical Analysis
Volume57
Issue number4
DOIs
Publication statusE-pub ahead of print - 25 Jul 2019
Externally publishedYes

Keywords

  • Finite element method
  • Kolmogorov equation
  • Malliavin calculus
  • One-sided Lipschitz coefficient
  • Stochastic Allen-Cahn equation
  • Strong and weak convergence rate

ASJC Scopus subject areas

  • Numerical Analysis
  • Computational Mathematics
  • Applied Mathematics

Fingerprint

Dive into the research topics of 'Strong and weak convergence rates of a spatial approximation for stochastic partial differential equation with one-sided lipschitz coefficient'. Together they form a unique fingerprint.

Cite this