Pathwise uniform convergence of numerical approximations for a two-dimensional stochastic Navier-Stokes equation with no-slip boundary conditions

Binjie Li, Xiaoping Xie, Qin Zhou · 2024

Read Pathwise uniform convergence of numerical approximations for a two-dimensional stochastic Navier-Stokes equation with no-slip boundary conditions by Binjie Li, Xiaoping Xie, Qin Zhou (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-04231.txt.

Introduction

Binjie Li, Xiaoping Xie, and Qin Zhou prove how fast computer approximations settle down for a fluid model buffeted by random noise. Their equation is the two-dimensional stochastic Navier-Stokes system, the workhorse model of moving fluids, with the fluid pinned to zero speed at the walls. Earlier work mostly measured average errors. This paper asks for something stronger: that whole simulated paths converge, uniformly, with high probability.

For a fully discrete finite element scheme, they show the error shrinks at nearly order three halves as the mesh refines in space. It shrinks at nearly order one half as the time step shrinks. Those are strong rates for a noisy nonlinear system with boundary constraints. The proof technique may interest other researchers wrestling with random partial differential equations.

Catalog record

CREATORBinjie Li, Xiaoping Xie, Qin Zhou
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE116 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-04231

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Pathwise uniform convergence of numerical approximations for a two-dimensional stochastic Navier-Stokes equation with no-slip boundary conditions Thanks: This work was partially supported by the National Natural Science Foundation of China under grants 12301525 and 12171340, and by the Natural Science Foundation of Sichuan Province under grant 2023NSFSC1324. Binjie Li, Xiaoping Xie, Qin Zhou Abstract This paper investigates the pathwise uniform convergence in probability of fully discrete finite-element approximations for the two-dimensional stochastic Navier-Stokes equations with multiplicative noise, subject to no-slip boundary conditions. We demonstrate that the full discretization achieves nearly 3 / 2 3/2 -order convergence in space and nearly half-order convergence in time. Keywords: stochastic Navier-Stokes equation, no-slip boundary conditions, finite element method, pathwise uniform convergence 1 Introduction Let 0<T<\infty , and consider \mathcal{O}\subset\mathbb{R}^{2} as a bounded, connected domain with a sufficiently smooth boundary \partial\mathcal{O} . We study the following incompressible stochastic Navier-Stokes equation (SNSE): \begin{cases}\begin{aligned} \mathrm{d}y(t,x)&=\Bigl[\Delta y(t,x)-(y(t,x)\cdot\nabla)y(t,x)-\nabla\varphi(t,x)\Bigr]\,\mathrm{d}t\\ &\hskip 9.24994pt+\sum_{n\in\mathbb{N}}\bigl[f_{n}(x,y(t,x))-\nabla\widetilde{\varphi}_{n}(t,x)\bigr]\,\mathrm{d}\beta_{n}(t),&&t\in[0,T],\,x\in\mathcal{O},\\ \nabla\cdot y(t,x)&=0,&&t\in[0,T],\,x\in\mathcal{O},\\ y(t,x)&=0,&&t\in[0,T],\,x\in\partial\mathcal{O},\\ y(0,x)&=y_{0}(x),&&x\in\mathcal{O}.\end{aligned}\end{cases} (1.1) In this system, y represents the velocity field of interest, \varphi denotes the pressure field, and (\widetilde{\varphi}_{n})_{n\in\mathbb{N}} is a sequence of unknown auxiliary scalar processes that ensures the divergence-free property of the diffusion term. The initial velocity field is given by y_{0} , (\beta_{n})_{n\in\mathbb{N}} is a sequence of independent Brownian motions, and the diffusion coefficients (f_{n})_{n\in\mathbb{N}} will be detailed later. Due to their important applications to the theory of turbulent flows and its significance from a pure mathematics perspective, the SNSEs have garnered significant interest. These equations have been extensively studied in the literature; see [ 3 , 14 , 17 , 24 , 40 ] and the reference therein. Consequently, over the past decades, the numerical analysis of SNSEs, as well as their simplified variants, the stochastic Stokes equations, has attracted considerable interest from the numerical community. For the cases of space-periodic boundary conditions, significant advancements have been observed in recent years. We briefly summarize some recent contributions as follows. Dörsek [ 21 ] derived rates of convergence for full discretizations that use spectral Galerkin method in space and semigroup splitting or cubature methods in time, for a two-dimensional SNSE with additive noise. In the context of finite-element-based full discretizations, Carelli and Prohl [ 18 ] made notable progress by establishing linear convergence in space and almost \frac{1}{4} -order convergence in time with respect to convergence in probability. Their findings were later improved upon by Breit and Dodgson [ 9 ] , who demonstrated enhanced theoretical results with linear convergence in space and nearly \frac{1}{2} -order convergence in time. Bessaih et al. [ 4 ] derived almost \frac{1}{2} -order temporal convergence in probability for a splitting up method. Bessaih and Millet [ 5 ] established strong L^{2} convergence for both the Euler scheme and a splitting scheme, and they subsequently extended their analysis to a full discretization that combines the Euler scheme for temporal approximation with the finite element method for spatial approximation, proving strong L^{2} convergence once more [ 6 ] . Hausenblas and Randrianasolo [ 34 ] studied a time-discretization with penalty-projection method, and derived a convergence rate of \frac{1}{4} in probability. A recent contribution by Breit et al. [ 10 ] derived a convergence rate of \frac{1}{2} in the mean square error for a time-discretization of a two-dimensional SNSE with transport noise. We also direct the reader to [ 22 , 23 ] and the references therein for the numerical analysis of stochastic Stokes equations. Of particular note is the recent work by Li et al. [ 38 ] , which established a strong convergence rate of O(h^{2}+\tau^{1/2}) in the L^{\infty}(0,T;L^{2}(\Omega;L^{2})) norm for finite-element-based approximations of two- and three-dimensional stochastic Stokes equations under the stress boundary conditions. Despite significant progress in this field, the numerical analysis of SNSEs subject to no-slip boundary conditions remains a relatively underdeveloped area of research. A fundamental difficulty arises from the fact that the Helmholtz projection \mathcal{P} fails to preserve the zero-trace property. This limitation implies that the range of \mathcal{P} is not necessarily contained within the domain of A_{2}^{1/4} (let alone A_{2}^{1/2} ), where A_{2} represents the realization of the negative Stokes operator in L^{2}(\mathcal{O};\mathbb{R}^{2}) with homogeneous Dirichlet boundary conditions. Moreover, the important identity \int_{\mathcal{O}}\big((u\cdot\nabla)u\big)\cdot A_{2}u\,\mathrm{d}x=0 , which holds on the domain of the Stokes operator under space-periodic boundary conditions, fails to hold under no-slip boundary conditions. These mathematical obstacles, combined with the temporal irregularities induced by the Brownian motions, substantially diminishes the solution’s regularity when compared to scenarios with space-periodic boundary conditions. Specifically, in the model problem Eq. 1.1 , the trajectories of the process y generally do not possess L^{2}(0,T;\dot{H}^{3/2,2}) -regularity, where \dot{H}^{3/2,2} denotes the domain of A_{2}^{3/4} , unless the critical structural assumptions—that the diffusion coefficients are divergence-free and vanish on the boundary \partial\mathcal{O} —are satisfied. Here is a brief summary of some recent works on the numerical analysis of SNSEs with no-slip boundary conditions: • Breckner [ 8 ] demonstrated the convergence in mean square for a spectral-Galerkin spatial semidiscretization. • Brzeźniak and Prohl [ 15 ] addressed both two- and three-dimensional cases, showing the convergence of finite element approximations to a unique strong solution in two dimensions using compactness arguments. • Doghman [ 20 ] extended the work [ 15 ] by proving the convergence of finite element approximations for the two-dimensional SNSEs through the artificial compressibility method. • Ondreját et al. [ 43 ] achieved convergence for a class of finite element approximations to a weak martingale solution in three dimensions. • To the best of our knowledge, Breit and Prohl [ 11 ] were the first to report strong convergence rates in probability for a fully discrete scheme applied to two-dimensional SNSEs with multiplicative noise, yielding the following convergence result: \lim_{\begin{subarray}{c}h\to 0\\ \tau\to 0\end{subarray}}\mathbb{P}\left[\frac{\max_{m}\|\bm{u}(t_{m})-\bm{u}_{h,m}\|_{L_{x}^{2}}^{2}+\sum_{m}\tau\|\nabla(\bm{u}(t_{m})-\bm{u}_{h,m})\|_{L_{x}^{2}}^{2}}{h^{\alpha}+\tau^{\beta}}>\xi\right]=0, for any \alpha\in(1,2) , \beta\in(0,1) , and \xi>0 . However, the numerical analysis presented in [ 11 ] is generally not applicable to the model problem Eq. 1.1 , as it relies on the aforementioned structural assumption that the diffusion coefficients are divergence-free and vanish on the boundary \partial\mathcal{O} . The model problem Eq. 1.1 and its more complex variants, where the diffusion coefficients depend on the gradient of y , have been extensively studied; see, for instance, [ 16 , 41 , 42 , 54 ] . Despite considerable research efforts, to the best of our knowledge, rigorous numerical analyses of finite element approximations directly applicable to the model problem Eq. 1.1 are limited to those presented in the aforementioned works [ 15 , 20 ] . However, these studies have not provided explicit convergence rates in probability. This significant gap in the existing literature motivates our current research. The remainder of this manuscript is organized as follows. 1. Section 2 introduces the fundamental notations and presents the concept of global mild solutions. 2. In Section 3 , we conduct a rigorous investigation into the regularity properties of the model problem Eq. 1.1 , which forms the cornerstone of our subsequent numerical analysis. Building upon the theoretical foundations established in [ 53 , 54 ] , we prove that the global mild solution of Eq. 1.1 satisfies the probabilistic estimate \mathbb{P}\left(\lVert{y}\rVert_{C([0,T];\dot{H}^{\varrho,2})}\leqslant R\right)\geqslant 1-\frac{c}{\ln(1+R)},\quad\forall R>1, where 1<\varrho<\frac{3}{2} , \dot{H}^{\varrho,2} denotes the domain of A_{2}^{\varrho/2} , and c>0 represents a constant independent of R

Read the complete text