Second or fourth-order finite difference operators, which one is most effective?
IJSM
Second or fourth-order finite difference operators,
which one is most effective?
Kolade M. Owolabi
Department of Mathematical Sciences, Federal University of Technology, Akure, P.O. Box 704, Ondo State, Nigeria.
Email: mkowolax@yahoo.com
This paper presents higher-order finite difference (FD) formulas for the spatial
approximation of the time-dependent reaction-diffusion problems with a clear justification
through examples, “why fourth-order FD formula is preferred to its second-order
counterpart” that has been widely used in literature. As a consequence, methods for the
solution of initial and boundary value PDEs, such as the method of lines (MOL), is of broad
interest in science and engineering. This procedure begins with discretizing the spatial
derivatives in the PDE with algebraic approximations. The key idea of MOL is to replace the
spatial derivatives in the PDE with the algebraic approximations. Once this procedure is
done, the spatial derivatives are no longer stated explicitly in terms of the spatial
independent variables. In other words, only one independent variable is remaining, the
resulting semi-discrete problem has now become a system of coupled ordinary differential
equations (ODEs) in time. Thus, we can apply any integration algorithm for the initial value
ODEs to compute an approximate numerical solution to the PDE. Analysis of the basic
properties of these schemes such as the order of accuracy, convergence, consistency,
stability and symmetry are well examined.
2010 Mathematics Subject Classification: {92B05, 92C15, 35K55, 65M20, 76R50}
Keywords: Consistency, Finite difference, PDEs, Reaction-diffusion, Stability, Symmetry.
INTRODUCTION
Reaction-diffusion equations are classified as a special class of parabolic time-dependent partial differential
equations. The major way of solving the class of these equations is through discretization. A well-known approach to
solve time-dependent partial differential equation, with solutions varying in both time and space, is the method of
lines (MOL), see Ascher et al. (1995), Schisser and Griffits (2009), Holden and Karlsen (2012) and Strikwerda
(2004) for details. Application of this method requires to first constructing a semi-discrete approximation to the
problem by setting up a regular grid in space, this is achieved by discretising the spatial independent variables with
boundary constraints. Hence, a couple systems of ordinary differential equations are generated in time, which is
associated with the initial value. Once that is done, we numerically approximate the solutions to the original time-
dependent partial differential equation by marching forward in time on this grid. Conveniently, we can now apply any
existing, and generally well established, time-stepping numerical methods such as the implicit-explicit (IMEX)
schemes, Runge-Kutta methods or exponential time differencing (ETD) schemes among many others.
In this paper, for the spatial discretisation, we are primarily concerned with the use of higher-order finite difference
method. The discrete approximation to the derivatives will be converted into Toeplitz matrices. The discretisation in
time uses majorly the exponential time differencing schemes, other time-stepping methods include but not limited to
the fourth-order Runge-Kutta (RK4) method and implicit-explicit schemes. A brief of each of these methods will be
International Journal of Statistics and Mathematics
Vol. 1(3), pp. 044-054, August, 2014. © www.premierpublishers.org, ISSN: 2375-0499x
Review