An asymptotic parallel-in-time method for highly oscillatory PDEs (Q2875002)

A new time-stepping algorithm for nonlinear partial differential equations (PDEs) is presented that exhibits scale separation in time of a highly oscillatory nature. The algorithm combines the parareal method -- a parallel-in-time scheme introduced by \textit{J.-L. Lions} et al. [C. R. Acad. Sci., Paris, Sér. I, Math. 332, No. 7, 661--668 (2001; Zbl 0984.65085)] -- with techniques from the heterogeneous multiscale method (cf. [\textit{W. E} and \textit{B. Engquist}, Notices Am. Math. Soc. 50, No. 9, 1062--1070 (2003; Zbl 1032.65013)]), which make use of the slow asymptotic structure of the equations (cf. [\textit{A. J. Majda} and \textit{P. Embid}, Theor. Comput. Fluid Dyn. 11, No. 3--4, 155--169 (1998; Zbl 0923.76339)]). Error bounds are presented, based on the analysis of \textit{M. J. Gander} and \textit{E. Hairer} [Lect. Notes Comput. Sci. Eng. 60, 45--56 (2008; Zbl 1140.65336)] and \textit{G. Bal} [Lect. Notes Comput. Sci. Eng. 40, 425--432 (2005; Zbl 1066.65091)], that demonstrate convergence of the method. A complexity analysis also demonstrates that the parallel speedup increases arbitrarily with greater scale separation. Finally, the accuracy and efficiency of the method is demonstrated on the (one-dimensional) rotating shallow water equations, which is a standard test problem for new algorithms in geophysical fluid problems. Compared to exponential integrators such as ETDRK4 and Strang splitting -- which solve the stiff oscillatory part exactly -- it is found that coarse time steps can be used that are orders of magnitude larger (for a comparable accuracy), yielding an estimated parallel speedup of approximately 100 for physically realistic parameter values. For the (one-dimensional) shallow water equations, it is also shown that the estimated parallel speedup of this ``asymptotic parareal method'' is more than a factor of 10 greater than the speedup obtained from the standard parareal method.