A spectral method in space and time to solve the advection-diffusion and wave equations in a bounded domain (1401.1373v1)

Abstract: The advection-diffusion and wave equations are the fundamental equations governing any physical law and therefore arise in many areas of physics and astrophysics. For complex problems and geometries, only numerical simulations can give insight into quantitative and accurate behavior of the seeked solutions. The standard numerical algorithm to solve partial differential equations is to split the space and time discretisation separately into different uncorrelated methods. Time is usually advanced by explicit schemes, or, for too restrictive time steps, by implicit or semi-implicit algorithms. This separate time and space slicing is artificial and sometimes unpractical. Indeed, treating space and time directions symmetrically and simultaneously without splitting is highly recommended in some problems like diffusion. It is the purpose of this work to present a simple numerical algorithm to solve the standard linear scalar advection-diffusion and wave equations using a fully spectral method in a two-dimensional Cartesian $(x,t)$ bounded space-time domain. Generalization in three-dimensions $(x,y,t)$ is shown for the pure diffusion problem. The basic idea is to expand the unknown function in Chebyshev polynomials for the spatial variables $(x,y)$ as well as for the time variable $t$. We show typical examples and demonstrate the spectral accuracy of the method. The great advantage of fully spectral methods resides in their high-accuracy for a relatively small number of grid points (for sufficiently smooth solutions) compared to standard time-stepping techniques.