An $L^0$-approach to stochastic evolution equations

TL;DR

This paper develops a new $L^0$-based framework for analyzing stochastic evolution equations, focusing on pathwise regularity and numerical approximation, with novel inequalities and applications to linear parabolic equations.

Contribution

It introduces two Burkholder--Davis--Gundy type inequalities for $L^0$-valued stochastic integrals, enabling new insights into pathwise regularity and convergence rates.

Findings

Established inequalities for $L^0$-valued stochastic integrals.

Derived H"older regularity results for It extsuperscript{o} integrals.

Demonstrated numerical approximation rates for stochastic evolution equations.

Abstract

We introduce a framework for studying pathwise time regularity and numerical approximation of $L^0$-valued stochastic evolution equations. At the core of our framework are two Burkholder--Davis--Gundy type inequalities accommodating It\^o integrals with respect to only stochastically integrable processes. The first of these inequalities is formulated in suitable metrics which metrize convergence in probability on the space of integrands and integrals. The second is a modified version, tailored for deriving pathwise properties of the integral. By combining it with a refined version of the Kolmogorov continuity test, we obtain a powerful method for deriving H\"older regularity of It\^o integrals in their most general form. Moreover, it provides a simple and powerful way of deriving rates of pathwise convergence of numerical approximations of stochastic evolution equations. Both applications are illustrated for a class of linear parabolic stochastic evolution equations with generalized Whittle--Mat\'ern type noise, and our findings are verified by numerical experiments from this setting.