High-Wavenumber Finite Differences and Turbulence Simulations

Abstract

We introduce a fluid dynamics algorithm that performs with nearly spectral accuracy, but uses finite-differences instead of FFTs to compute gradients and thus executes 10 times faster. The finite differencing is not based on a high-order polynomial fit. The polynomial scheme has supurb accuracy for low-wavenumber gradients but fails at high wavenumbers. We instead use a scheme tuned to enhance high-wavenumber accuracy at the expense of low wavenumbers, although the loss of low-wavenumber accuracy is negligibly slight. A tuned gradient is capable of capturing all wavenumbers up to 80 percent of the Nyquist limit with an error of no worse than 1 percent. The fact that gradients are based on finite differences enables diverse geometries to be considered and eliminates the parallel communications bottleneck.