Remote Access Mathematics of Computation
Green Open Access

Mathematics of Computation

ISSN 1088-6842(online) ISSN 0025-5718(print)



Convergence of a random particle method to solutions of the Kolmogorov equation $ u\sb t=\nu u\sb {xx}+u(1-u)$

Author: Elbridge Gerry Puckett
Journal: Math. Comp. 52 (1989), 615-645
MSC: Primary 65C05; Secondary 65M99
MathSciNet review: 964006
Full-text PDF

Abstract | References | Similar Articles | Additional Information

Abstract: We study a random particle method for solving the reaction-diffusion equation $ {u_t} = \nu {u_{xx}} + f(u)$ which is a one-dimensional analogue of the random vortex method. It is a fractional step method in which $ {u_t} = \nu {u_{xx}}$ is solved by random walking the particles while $ {u_t} = f(u)$ is solved with a numerical ordinary differential equation solver such as Euler's method. We prove that the method converges when $ f(u) = u(1 - u)$, i.e. the Kolmogorov equation, and that when the time step $ \Delta t$ is $ O({\sqrt[4]{N}^{ - 1}})$ the rate of convergence is like $ \ln N \cdot \,{\sqrt[4]{N}^{ - 1}}$ where N denotes the number of particles. Furthermore, we show that this rate of convergence is uniform as the diffusion coefficient $ \nu$ tends to 0. Thus, travelling waves with arbitrarily steep wavefronts may be modeled without an increase in the computational cost. We also present the results of numerical experiments including the use of second-order time discretization and second-order operator splitting and use these results to estimate the expected value and standard deviation of the error.

References [Enhancements On Off] (What's this?)

Similar Articles

Retrieve articles in Mathematics of Computation with MSC: 65C05, 65M99

Retrieve articles in all journals with MSC: 65C05, 65M99

Additional Information

Keywords: Kolmogorov equation, particle method, random vortex method, random walk
Article copyright: © Copyright 1989 American Mathematical Society

American Mathematical Society