65. Simulations of decaying turbulence

A group of people (with Jonathan Dursi being the only one I know personally among them) has written nice review of existing software for stimulating decaying compressible (supersonic) turbulence with the goal to compare performance of the codes. In particular, they have reviewed GADGET, PHANTOM, VINE, ENZO, FLASH, TVD, and ZEUS (the first three being SPH, i.e., particle codes, while the remaining – grid codes).

Since the authors are astrophysicists, they are clearly interested in astrophysical applications, i.e., in supersonic flows. For a non-expert – one example of such regime is realized in galactic (especially, regions where stars are created) and extragalactic molecular clouds where the random motion carries enough kinetic energy to counterbalance and even overcompensate self-gravity in the clouds.

The intersting feature of supersonic turbulence is that the energy is dissipated not only due to viscocity but also in shock waves, and typically, to correctly model a shock wave, one has to implement numerical viscocity into account. Shock waves transfer energy between strongly separated scales, removing the local nature of the usual Kolmogorov’s cascade (the regime typical for subsonic turbulence or turbulence in incompressible fluid).

The question is how this numerical viscocity influences the physical outcome of simulations, and that is another reason why the authors wanted to compare these codes.

Their main conclusion is that grid codes are 2-3 times faster than SPH codes, but PHANTOM is written so well that it outperforms all other codes.