A nonnegativity-preserving finite element method for a class of parabolic SPDEs with multiplicative noise

Ana Djurdjevac, Claude Le Bris, Endre Süli · 2025

Read A nonnegativity-preserving finite element method for a class of parabolic SPDEs with multiplicative noise by Ana Djurdjevac, Claude Le Bris, Endre Süli (2025) 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-2502-16854.txt.

Introduction

Ana Djurdjevac, Claude Le Bris, and Endre Süli design a simulation method that keeps solutions from going negative when the true answer cannot, in a 2025 paper. Their setting is a heat-style equation driven by random noise that multiplies the state. Such equations model spread under chance effects, and their true solutions stay nonnegative when they start that way. Many standard numerical schemes break that rule, which can wreck their meaning.

The team's finite element scheme provably keeps the numerical solution nonnegative at every step, with no condition on the mesh size, and it converges to the true solution. A fully step-by-step version for the linear case keeps the same guard. Tests against older schemes that lack the guard show the gain in practice.

Catalog record

CREATORAna Djurdjevac, Claude Le Bris, Endre Süli
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE78 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-16854

Browse the full library · About the Tower of Babel

Full text

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

A nonnegativity-preserving finite element method for a class of parabolic SPDEs with multiplicative noise Ana Djurdjevac, Claude Le Bris, Endre Süli Abstract We consider a prototypical parabolic SPDE with finite-dimensional multiplicative noise, which, subject to a nonnegative initial datum, has a unique nonnegative solution. Inspired by well-established techniques in the deterministic case, we introduce a finite element discretization of this SPDE that is convergent and which, subject to a nonnegative initial datum and unconditionally with respect to the spatial discretization parameter, preserves nonnegativity of the numerical solution throughout the course of evolution. We perform a mathematical analysis of this method. In addition, in the associated linear setting, we develop a fully discrete scheme that also preserves nonnegativity, and we present numerical experiments that illustrate the advantages of the proposed method over alternative finite element and finite difference methods that were previously considered in the literature, which do not necessarily guarantee nonnegativity of the numerical solution. 1 Introduction We consider the following prototypical stochastic partial differential equation (henceforth abbreviated as SPDE) with multiplicative noise: \,\mathrm{d}u-\Delta u\,\mathrm{d}t=f(u)\,\mathrm{d}W (1) (with the usual notation; see ( 6 ) below for the precise mathematical meaning) that preserves nonnegativity at the continuous level: the nonnegativity of an initial datum u_{0} for ( 1 ) is propagated into the associated solution u(t) for all t\geq 0 . The purpose of this article is to construct and analyze a finite element method (FEM) for this SPDE that is accurate and that preserves nonnegativity unconditionally in the discretization parameters (the time step \Delta t and the spatial mesh size h ). Our motivation for this study stems from our interest in the Dean–Kawasaki equation and related equations, and more generally a large class of SPDEs that have nonnegative solutions because their solution models the density of a system of possibly infinitely many particles (see [ 22 , 16 , 43 , 41 ] for further details on this class of problems and for some illustrative and recent mathematical contributions to the topic). The present paper may be seen as a very preliminary step towards the design of suitable numerical methods for such equations. In any event, the equation we study and approximate numerically is sufficiently ubiquitous to be relevant in many practical settings. We therefore hope that the considerations herein will serve as a guideline for the construction of numerical methods for more sophisticated equations. The preservation of nonnegativity for parabolic SPDEs with multiplicative noise at the continuous level has also been explored in other contexts in several publications; see, for example, [ 6 , 10 , 17 , 25 ] . Numerical methods for SPDEs have been extensively studied in the literature. For general results, we refer to [ 35 , 39 , 51 ] and references therein. As was mentioned above, the specific feature of the SPDE we shall be focussed on here is nonnegativity-preservation. Our concern is therefore at the intersections of two different classes of questions, namely: (i) nonnegativity-preservation for finite element discretizations of deterministic partial differential equations; and (ii) nonnegativity-preservation for time-stepping schemes for stochastic differential equations. Even in the absence of noise, that is, for deterministic PDEs, preservation of nonnegativity, as stated in item (i) above, is a challenging requirement since it is intimately related to the construction of numerical methods that satisfy a discrete maximum principle; see, for example, the recent monograph [ 4 ] . Since we cannot expect any simplifications in this respect in the presence of noise in an otherwise deterministic equation, our numerical method must comply with nonnegativity-preservation for PDEs. We will make these requirements precise later on (in Section 3 ). They concern the geometry of the mesh and the use of a specific version of the mass matrix, the so-called lumped mass matrix, in the finite element discretization, first introduced in the parabolic setting in [ 26 ] (see also [ 50 , 9 ] ). Briefly, our choice of the spatial discretization is dictated by our wish to preserve nonnegativity. However, this comes at a price. A certain number of mathematical difficulties follow from the technical choices we are bound to make at the finite element level. Put differently, our analysis would have been much simpler and more standard to conduct had we opted for a discretization that is not required to systematically guarantee nonnegativity-preservation. To start with, in the absence of mass-lumping, we could have considered a formulation of the SPDE based on semi-group theory (a formulation that is routinely considered for generic SPDEs; see, for example, [ 28 , 27 , 49 , 34 , 29 , 52 ] ), and then our analysis of the discretization error would also have become much simpler and would have followed through the use of standard techniques. In addition, ignoring for the moment the dependence of the solution on the spatial variable and concentrating on the random variable, nonnegativity-preservation in the context of time-stepping schemes for SDEs is a well-established topic that has been of particular relevance in, for example, financial mathematics. We will benefit from the knowledge accumulated in the field concerning this topic to formulate our time discretization; see [ 47 , 48 , 42 , 37 ] for some specific contributions in this direction. In particular, we have been greatly inspired by the publications of A. Alfonsi initiated in [ 1 ] and pursued, for example, in [ 2 , 3 ] . In the context of SPDEs, publications that rigorously address nonnegativity-preservation at the fully-discrete level are rare. We are aware of only a handful of papers on this topic. They all address finite difference type methods (as opposed to finite element methods) and we shall only mention a couple: the recent work [ 53 ] introduces a nonnegativity-preserving finite difference scheme for the linear stochastic heat equation with finite-dimensional noise (as will be the case in the present study). In the same spirit, the work [ 7 ] develops a nonnegativity-preserving Lie–Trotter splitting scheme using a finite difference method in space for a nonlinear heat equation driven by multiplicative space-time white noise, and therefore improves on the former work in terms of generality of the noise but is restricted to the one-dimensional setting. Positivity preservation for SPDEs has also been investigated in the recent work [ 18 ] , where energy-regularized logarithmic SPDE models and structure-preserving numerical approximations are studied. It is important to note that the issue of nonnegativity-preservation cannot be overlooked. In the particular case of the Dean–Kawasaki equation, which we have mentioned as our motivation for the present work, it was shown in [ 21 ] that a classical spatial finite-volume approximation subject to a nonnegative initial condition may produce numerical solutions with negative values at subsequent times. Ad hoc techniques to overcome this shortcoming have therefore been developed and then adapted to the finite element context in [ 43 ] . This suffices to indicate the need for the development of a more systematic approach, which is our objective here. It is also important to emphasize outright that our wish to consider finite element methods (as opposed to finite difference schemes), in addition to the fact that the former are of variational nature (and therefore more amenable

Read the complete text