A comparison of basis functions for the pseudo-spectral method for a model reaction-diffusion problem (Q579895): Difference between revisions

The authors consider two schemes for approximating the Laplacian operator in reaction-diffusion systems. Experiments are based on the model equation \[ (1)\quad u_ t=\Delta u+F(u),\text{ where } u=u(x,t),\quad x\in \Omega \subset {\mathbb{R}}^ m,\quad m=1\quad or^ 2,\quad and\text{ where } u(x,t)=0\quad for\quad x\in \partial \Omega. \] The solution u is approximated in terms of a finite basis of either eigenfunctions of the Laplacian (as in the spectral method) or Chebychev polynomials; collocation is used to derive equations for the coefficients in these bases. Due to the non-periodic boundary conditions, the eigenfunction expansion cannot be expected to converge superalgebraically. Also, the effect of nonlinearities needs to be demonstrated. The authors present numerical results for various cases of (1), over line segments, rectangles, and circles. Logarithmic error curves illustrate convergence of the eigenfunction basis scheme with order \(O(N^{-2})\) or \(O(N^{-4})\), where N is the number of basis functions, while the Chebychev basis scheme seems to converge faster than any power of N. However, the eigenfunction basis gives better (and in many cases adequate) accuracy, and is easier to program, than the Chebychev basis scheme, for small values of N. The paper is well-written.