[CP2K:809] applicability of TEMP_TOL to thermostat regions...

[Teodoro Laino]

On 13 Mar 2008, at 02:45, Axel wrote:

>
> hi everybody,
>
> since we can apply thermostats to individual regions, i was just
> wondering, if TEMP_TOL should be applied to each region individually
> or only globally. i guess, it could make some sense to have this per
> region if those are "large enough" (but what would be large enough?).
>
> ... or would using a CSVR type thermostat with a short time-constant
> have a similar (if not better) effect (removal of excessive kinetic
> energy
> in certain areas of my system at the beginning of equilibration).
Exactly.. a short timecon has the same effect..
Giovanni can comment on that better than I can..

When we (me and Giovanni) were implementing his (Giovanni) thermostat  
in CP2K my
idea was to get rid of the TEMP_TOL keyword (since TEMP_TOL (i.e. the  
rescaling of the velocity)) is much more
clean and powerful than a simple TEMP_TOL. There was a kind of  
rebellion for reasons that I won't comment here.

I can survive with having a nice thermostat (CSVR) and an ugly  
rescaling option (TEMP_TOL) in two different places (quite well
distinct),
BUT  you will never see the TEMP_TOL keyword inside a thermostat  
section (it's absolutely dirty and would
degrade all the efforts done in cleaning that part of the code).

So.. I have no problem to convert TEMP_TOL into a kind of special  
thermostat, but TEMP_TOL (as a keyword itself)
would be deleted. Since there's no overall agreement (unless things  
have been changed in the last weeks) I would suggest
you to use CSVR with a small TIMECON.

cheers,
Teo

>
> comments? suggestions?
>
> cheers,
>     axel.
> >