Benchmarking the propagator method for nonlinear systems: a Burgers-Korteweg-deVries equation (Q915414)

The analytical basis of the new method presented here is the construction of propagators [due to \textit{R. B. Perez} and \textit{V. Protonopeccu}, Math. Phys. 29, 353-361 (1988; M.R. 89b:58087)], which ``propagate'' the initial distribution of the unknown solution from the phase-space location (0,x) to the phase-space location (t,x) by an explicit integral expression. Thus, the role played by them is entirely similar to that played by the customary Green's function encountered in linear problems. However, the propagators depend (nonlinearly) on the unknown solution, thus the original nonlinear partial differential equation of third order (including both the Burgers equation and the Korteweg-de Vries equation) recasts into an implicit nonlinear integral equation. But the use of a special propagator (exactly expressed via Fourier and Laplace transforms) reduces the problem to an explicit nonlinear integral equation in a fixed-point form. To solve it numerically the authors implement a spectral method (a truncated Fourier series) for the spatial variable with a class of implicit Runge-Kutta methods for the temporal integral. To obtain a comparative evaluation of this method the authors use as benchmarks a finite difference scheme and a local method based on inverse scattering transforms [\textit{T. R. Taha}, \textit{M. J. Ablowitz}, J. Comput. Physics 55, 231-253 (1984; Zbl 0541.65083)]. All this indicates that the new method is efficient, accurate, and robust.