Re: Entropy Conservation

From: Volker Springel <volker_at_MPA-Garching.MPG.DE>
Date: Mon, 26 Dec 2005 18:14:34 +0100

Yves Revaz wrote:
>
> Dear Gdget list,
>
> In the file timestep.c, we can find the explicit comment (followed by
> the corresponding instructions) :
>
> /* In case of cooling, we prevent that the entropy (and
> hence temperature decreases by more than a factor 0.5 */
>
> I was asking me if it was not better to add a constraint on dt (in
> get_timestep)
> that ensure for example that the variation of the entropy due to the
> cooling is
> less than for example 10% of its value.
>

Hi,

Limiting the timestep in this way is certainly the first idea one may have
about this, but I don't think it provides a good solution.

The cooling equation has some very stiff parts, where the cooling timescale
can become much shorter than the dynamical timescale. If you then limit the
integration step of the full dynamics to the cooling time, you will get tiny
timesteps which are overkill for the dynamics of the particles, i.e. you're
  wasting cpu time without gaining significantly in accuracy. Except if you
care about detailed calculations of the emission spectrum, for example,
where you might need the exact time evolution of the temperature. In the
latter case, it would presumably be better to subcycle the cooling equation
instead of imposing the cooling timestep on the full dynamics.

Volker


> Thanks for giving me your opignon.
>
> Cheers.
>
>
Received on 2005-12-26 18:14:42

This archive was generated by hypermail 2.3.0 : 2022-09-01 14:03:41 CEST