435
ISSN 1064-5624, Doklady Mathematics, 2018, Vol. 98, No. 2, pp. 435–438. © Pleiades Publishing, Ltd., 2018.
Original Russian Text © K.K. Sabelfeld, 2018, published in Doklady Akademii Nauk, 2018, Vol. 482, No. 2.
A Mesh Free Stochastic Algorithm for Solving
Diffusion–Convection–Reaction Equations on Complicated Domains
1
K. K. Sabelfeld
Presented by Academician of the RAS A.N. Konovalov March 13, 2018
Received March 19, 2018
Abstract—A mesh free stochastic algorithm for solving transient diffusion–convection–reaction problems on
domains with complicated structure is suggested. For the solutions of this kind of equations exact represen-
tations of the survival probabilities, the probability densities of the first passage time and position on a sphere
are obtained. Based on these representations we construct a stochastic algorithm which is simple in imple-
mentaion for solving one- and three-dimensional diffusion–convection–reaction equations. The method is
continuous both in space and time, and its advantages are particularly well manifested in solving problems on
complicated domains, calculating fluxes to parts of the boundary, and other integral functionals, for instance,
the total concentration of the particles which have been reacted to a time instant .
DOI: 10.1134/S1064562418060108
t
Stochastic methods for solving boundary value
problems for parabolic and elliptic differential equa-
tions are based on probabilistic representations in the
form of mathematical expectations in the space of dif-
fusion processes [5]. Implementation of stochastic
algorithms is then carried out by numerical solution of
stochastic differential equations governing the trajec-
tories of the diffusion processes [6]. However this
approach needs constructions of temporal and spatial
meshes which complicates the implementation on
complicated domains and significantly reduces the
efficiency of the method. For equations with constant
coefficients a Random Walk on Spheres (RWS) algo-
rithm is extremely efficient, it was first suggested in [7]
for the Laplace equation, and in [4] for the heat equa-
tion. This class of methods is mesh free, and is effi-
ciently implemened on complicated domains. The
RWS method was then extended on other equations
mainly related to the Laplace equation and isotropic
diffusion [2, 8, 9]. In all these studies the isotropy
property of the Wiener process is used. In practice,
when dealing with convection, one uses an approxi-
mation by adding a small drift on a small time step.
This however again brings the method back to the class
1
The article was translated by the author.
of methods with meshes, and introduces an additional
error related to an artificial separation of the diffusion
and convection transfer. In this paper we suggest a
mesh free RWS method for solving high-dimensional
diffusion-convection-reaction problems which is
based on a simple idea: we derive an exact distribution
(in time and position) of the process with constant
convection on an arbitrary sphere. This enables to
carry out exact simulations of the particle transport in
space and time governed by the diffusion–convec-
tion–reaction equations. In the case when the diffu-
sion coefficient, the convection and reaction terms are
varying in space the random walking spheres are cho-
sen small enough, and the RWS method remains mesh
free.
1. DIFFUSION–CONVECTION–REACTION
EQUATION
Our goal is to construct a mesh free algorithm for
solving a transient diffusion–convection–reaction
equation based on simulation of particle trajectories
governed by this equation. We call the random particle
trajectory also as a diffusion–convection–reaction
process. Physically, these processes describe the
motion of a particle with a diffusion mobility charac-
terized by the diffusion coefficient in a given
velocity field , and under the reaction (say, chem-
ical reactions) rate . We first consider for sim-
plicity a one-dimensional case: we consider on a
direct equation for the Green function
() D x
() vx
λ
2
() x
, [0 ] L
MATHEMATICS
Institute of Computational Mathematics and Mathematical
Geophysics, Russian Academy of Sciences, Novosibirsk,
630090 Russia
e-mail: karl@osmf.sscc.ru