temperature monitor and thermostat regions

[Axel]

On Nov 17, 10:41 am, Teodoro Laino <teodor... at gmail.com> wrote:
> Hi Axel,

hi teo,

i'm finally getting around to testing this
and it looks pretty darn good.


> the last step (2) of the THEMOSTATS mail :
>
> On 26 Sep 2007, at 14:28, Teodoro Laino wrote:
>
> > Next thing to come in few days:
>
> > (2) Possibility to have a user-defined region for thermostats..
>
> is now completed. It is possible to define (in a similar way to
> constraints (LIST, RANGE, QM, MM, MOLNAME)) regions of the system on
> which apply thermostats. Please have a look at the several new
> regtests added to see how this feature
> works (mainly in tests/Fist/regtest-11/ and QMMM/SE/regtest).
>
> It was a real nightmare, believe me, especially due to the particle
> distribution in parallel to get an efficient
> thermostat scheme.

i understand. the codes where i had so far had the (negative)
fun to mess with thermostats, were using a replicated coordinates
strategy. infinitely simpler, but only feasable for all-qm
calculations.

> I f you find any strange behavior, please send me a small input file
> showing the problem.

not so far, but i can imaging, there is a bunch of trouble
lurking when using constraints, particularly, if they cross
thermostat region boundaries.

> At the moment the temperature that can be applied on the several
> regions is always the same..

> We could discuss about the possibility to have different regions to
> different temperatures, but I'm quite
> skeptical whether it will really work!

i guess there are some people that would be interested in this,
but i would worry about this when they actually do ask.

what would be the cherry on of the cream on _my_ pie would
be that the MASSIVE flag (i.e. one thermostat per DOF) would
not be only a global option but a per region option. i assume
this would make everything a lot more complicated and it is
probably best to wait until we have seen how the thermostat
regions work with constraints. particularly for qm/mm simulations
i found it to be an excellent means to quickly dissipate the
"shock" from the transition from all-MM to have a massive
thermostat on the qm subsystem. but already the ability to
have separate thermostats down to a single atom region is
very helpful. i'll have to see how far i can get by writing
a small script that defines a separate region per qm atom...
good thing that we have VMD.

apropos VMD: how about making SEGNAME an alias for MOLNAME,
as this is the VMD (and pdb?) nomenclature for this flag?


thanks again for your efforts,
    axel.

>
> teo
>
> p.s.: I will add soon also the possibility to dump on file the
> temperatures of the different thermostats (tomorrow should be in)..
>
> On 5 Sep 2007, at 16:42, Axel wrote:
>
>
>
> > hi all!
>
> > here's another cp2k feature question:
>
> > it would be extremely helpful to have an option to
> > monitor 'local' temperatures and apply thermostats
> > (either a single chain or one chain per DOF) to them.
> > the case of QM/MM simulations is only one example
> > where this might be useful.