We present a class of new explicit and stable numerical algorithms to solve the spatially discretized linear heat or diffusion equation. After discretizing the space and the time variables like conventional finite difference methods, we do not approximate the time derivatives by finite differences, but use constant neighbor and linear neighbour approximations to decouple the ordinary differential equations and solve them analytically. During this process, the timestep-size appears not in polynomial, but in exponential form with negative exponents, which guarantees stability. We compare the performance of the new methods with analytical and numerical solutions. According to our results, the methods are first and second order in time and can be much faster than the commonly used explicit or implicit methods, especially in the case of extremely large stiff systems.
Comments: 21 pages
Subjects: Numerical Analysis (math.NA); Mathematical Physics (math-ph); Computational Physics (physics.comp-ph)
Journal reference: Numerical Methods for Partial Differential Equation, Volume 37, Issue 3, May 2021, Pages 2469-2489