Rate of convergence of a semi-implicit time Euler scheme for a 2D Bénard-Boussinesq model

Hakima Bessaih, Annie Millet · 2024

Read Rate of convergence of a semi-implicit time Euler scheme for a 2D Bénard-Boussinesq model by Hakima Bessaih, Annie Millet (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-2411-02590.txt.

Description

Abstract . We prove that a semi-implicit time Euler scheme for the two-dimensional Bénard-Boussinesq model on the torus D D converges. The rate of convergence in probability is almost 1/2 for a multiplicative noise; this relies on moment estimates in various norms for the processes and the scheme.

Catalog record

CREATORHakima Bessaih, Annie Millet
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE132 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2411-02590

Browse the full library · About the Tower of Babel

Full text

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

Rate of convergence of a semi-implicit time Euler scheme for a 2D Bénard-Boussinesq model Hakima Bessaih, Annie Millet Abstract . We prove that a semi-implicit time Euler scheme for the two-dimensional Bénard-Boussinesq model on the torus D D converges. The rate of convergence in probability is almost 1/2 for a multiplicative noise; this relies on moment estimates in various norms for the processes and the scheme. In case of an additive noise, due to the coupling of the equations, provided that the difference on temperature between the top and bottom parts of the torus is not too big compared to the viscosity and thermal diffusivity, a strong polynomial rate of convergence (almost 1/2) is proven in ( L 2 ​ ( D ) ) 2 (L^{2}(D))^{2} for the velocity and in L 2 ​ ( D ) L^{2}(D) for the temperature. It depends on exponential moments of the scheme; due to linear terms involving the other quantity in both evolution equations, the proof has to be done simultaneaously for both the velocity and the temperature. These rates in both cases are similar to that obtained for the Navier-Stokes equation. 2000 Mathematics Subject Classification Primary 60H15, 60H35, 65M12; Secondary 76D03, 76M35. 1. Introduction The Boussinesq equations have been used as a model in many geophysical applications. They have been widely studied in a both the deterministic and stochastic settings. We take random forcings into account and formulate the Bénard convection problem as a system of stochastic partial differential equations (SPDEs). The need to take stochastic effects into account for modeling complex systems has now become widely recognized. Stochastic partial differential equations (SPDEs) arise naturally as mathematical models for nonlinear macroscopic dynamics under random influences. It is a coupled system of the stochastic Navier–Stokes equations and the stochastic transport equation for temperature. For us this system will be subject to an additive random perturbation which will be defined later. Here, u describes the fluid velocity field, while {\mathcal{T}} describes the temperature of the buoyancy driven fluid, and \pi is the fluid’s pressure. We study the Bénard-Boussinesq equations on the torus D=[0,L]^{2} subject to a multiplicative (an additive) stochastic perturbation. Unlike [ 5 ] we impose different values of the temperature {\mathcal{T}} on the top and bottom parts of the torus, that is {\mathcal{T}}(t,.)={\mathcal{T}}_{0} on \{x_{2}=0\} and {\mathcal{T}}(t,.)={\mathcal{T}}_{L} on \{x_{2}=L\} for t\in[0,T] . The fluid’s velocity u , temperature {\mathcal{T}} and pressure \pi satisfy the following Bénard problem (see e.g. [ 22 ] and [ 15 ] ): \displaystyle\partial_{t}u-\nu\Delta u+(u\cdot\nabla)u+\nabla{\pi} \displaystyle=({\mathcal{T}}-T_{L})v_{2}+G(u)\,dW\quad\mbox{\rm in }\quad(0,T)\times D, (1.1) \displaystyle\partial_{t}{\mathcal{T}}-\kappa\Delta{\mathcal{T}}+(u\cdot\nabla{\mathcal{T}}) \displaystyle={\mathcal{G}}({\mathcal{T}})\,d\tilde{W}\quad\mbox{\rm in }\quad(0,T)\times D, (1.2) \displaystyle\mbox{\rm div }u \displaystyle=0\quad\mbox{\rm in }\quad[0,T]\times D, where T>0 . The processes u:\Omega\times(0,T)\times D\to{\mathbb{R}}^{2} , and {\mathcal{T}}:\Omega\times(0,T)\times D\to{\mathbb{R}} have initial conditions u_{0} and {\mathcal{T}}_{0} in D respectively, and \pi:\Omega\times(0,T)\times D\to{\mathbb{R}} . The parameter \nu>0 denotes the kinematic viscosity of the fluid, \kappa>0 its thermal diffusivity, and (v_{1},v_{2}) is the canonical basis of {\mathbb{R}}^{2} . On [0,T]\times\partial D these fields satisfy the following boundary conditions \displaystyle u(t,x+Lv_{i})=u(t,x), \displaystyle\;i=1,2,\quad{\mathcal{T}}(t,x+Lv_{1})={\mathcal{T}}(t,x),\quad\partial_{1}{\mathcal{T}}(t,x+Lv_{1})={\mathcal{T}}(t,x), \displaystyle\pi(t,x+Lv_{1})=\pi(t,x), \displaystyle\quad\partial_{1}\pi(t,x+Lv_{1})=\pi(t,x) (1.3) where v_{i} , i=1,2 denotes the canonical basis of {\mathbb{R}}^{2} , More information on the stochastic perturbation will be given in the next section. We then introduce a classical change of processes to describe the temperature and pressure (see e.g. [ 22 ] and [ 15 ] ): \theta(t,x)={\mathcal{T}}(t,x)-{\mathcal{T}}_{0}-x_{2}\frac{{\mathcal{T}}_{L}-{\mathcal{T}}_{0}}{L},\quad\hat{\pi}(t,x)=\pi(t,x)+({\mathcal{T}}_{L}-{\mathcal{T}}_{0})x_{2}+\frac{({\mathcal{T}}_{L}-{\mathcal{T}}_{0})x_{2}^{2}}{2L}. (1.4) We deduce that \theta and \hat{\pi} have the same periodic properties as {\mathcal{T}} and \pi , that \theta(t,.)=0 on \{x_{2}=0\}\cup\{x_{2}=L\} and that the pair (u,\theta) satisfies the more symmetric Boussinesq equations \displaystyle\partial_{t}u-\nu\Delta u+(u\cdot\nabla)u+\nabla\hat{\pi} \displaystyle=\theta v_{2}+G(u)dW\quad\mbox{\rm in }\quad(0,T)\times D, (1.5) \displaystyle\partial_{t}\theta-\kappa\Delta\theta+(u\cdot\nabla\theta)+C_{L}u_{2} \displaystyle=\tilde{G}(\theta)d\tilde{W}\quad\mbox{\rm in }\quad(0,T)\times D, (1.6) \displaystyle\mbox{\rm div }u \displaystyle=0\quad\mbox{\rm in }\quad[0,T]\times D, where C_{L}:=\frac{{\mathcal{T}}_{L}-{\mathcal{T}}_{0}}{L}\in{\mathbb{R}} and {\tilde{G}} is defined by {\tilde{G}}(\theta)={\mathcal{G}}\big({\mathcal{T}}-{\mathcal{T}}_{0}-x_{2}\frac{{\mathcal{T}}_{L}-{\mathcal{T}}_{0}}{L}\big) . In two dimensions, the deterministic version of this system (without stochastic perturbations) has been extensively studied, yielding a comprehensive understanding of its well-posedness and long-time behavior; see [ 22 ] . Further investigations have addressed the cases where either \nu=0 or \kappa=0 , albeit with only partial results. In [ 17 ] , the authors establish the existence and uniqueness of solutions when \nu>0 and \kappa=0 for initial conditions u_{0}\in(H^{m})^{2}({\mathbb{R}}^{2}) and \theta_{0}\in H^{m-1}({\mathbb{R}}^{2}) for m\geq 3 . In [ 19 ] , the case where viscosity acts only in the horizontal direction (i.e., replacing \nu\Delta u by \nu\partial_{1}^{2}u ) and \kappa=0 is analyzed. The authors prove the existence of a global solution for initial data \theta_{0}\in L^{2}(D) and a divergence-free velocity field u_{0}\in\big(W^{1,2}(D)\big)^{2} . They reformulate the system in terms of vorticity and establish uniqueness under stronger regularity assumptions, leveraging techniques inspired by Yudovich. To the best of our knowledge, the case \nu=\kappa=0 remains an open problem. In the stochastic setting, well-posedness and the large deviation principle have been established for the Bénard system with \nu>0 and \kappa>0 ; see [ 10 ] . Random attractors are investigated in [ 14 ] for a more general ocean-atmospheric model, while invariant measures are studied in [ 15 ] using dissipativity arguments. Let V^{0}=\{u\in(L^{2})^{2}:{\rm div}\,u=0\} , \Pi:(L^{2})^{2}\to V^{0} denote the Leray projection on divergence free fields, and A=-\Pi\Delta denote he Stokes operator. Projecting on divergence-free fields, we consider the system \displaystyle\partial_{t}u+\nu Au+(u\cdot\nabla)u \displaystyle=\Pi(\theta v_{2})+G(u)dW\quad\mbox{\rm in }\quad(0,T)\times D, (1.7) \displaystyle\partial_{t}\theta-\kappa\Delta\theta+(u\cdot\nabla\theta)+C_{L}u_{2} \displaystyle={\tilde{G}}(\theta)d\tilde{W}\quad\mbox{\rm in }\quad(0,T)\times D, (1.8) If u_{0} and \theta_{0} are square integrable, it is known that the random system ( 1.5 )–( 1.6 ) is well-posed, and that there exists a unique solution (u\times\theta) in L^{2}([0,T];(L^{2})^{2}\times L^{2})\cap L^{2}(\Omega;(H^{1})^{2}\times H^{1}) ; see e.g. [ 10 ] . Numerical schemes and algorithms have been introduced to best approximate the solution to non-linear PDEs. The time approximation is either an implicit Euler or a time splitting scheme coupled with Galerkin approximation or finite elements to approximate the space variable. The literature on numerical analysis for SPDEs is now very extensive. In many papers the models are either linear, have global Lipschitz properties, or more generally some monotonicity property. In this case the convergence is proven to be in mean square. When nonlinearities are involved that are not of Lipschitz or monotone type, then a rate of convergence in mean square is more difficult to obtain. Indeed, because of the stochastic perturbation, one may not use the Gronwall lemma after taking the expectation of the error bound since it involves a nonlinear term which is often quadratic; such a nonlinearity requires some localization. In a random setting, the discretization of the Navier-Stokes equations has been intensively investigated. Various space-time numerical schemes have been studied for the stochastic Navier-Stokes equations with a multiplicative or an additive noise, that is where in the right hand side of ( 1.5 ) (with no \theta ) we have either G(u)\,dW or dW . We refer to [ 8 , 13 , 6 , 9 , 7 ] , where convergence in probability is stated with various rates of convergence in a multiplicative setting. As stated previously, the main tool to get the convergence in probability is the localization of the nonlinear term over a space of large probability. We studied the strong (that is L^{2}(\Omega) ) rate of convergence of the time implicit Euler scheme (resp. space-time implicit Euler scheme coupled with finite element space discretization) in our previous papers [ 2 ] (resp. [ 3 ] ) for an H^{1} -valued

Read the complete text