GCMC

Matt McGrath obfis... at gmail.com
Tue Aug 23 08:21:41 UTC 2011


Hmm.  I was thinking about that, and I didn't think it would work
before I responded to your original question.  But now, I'm not so
sure.  If you set all the MM interactions in the vapor box to have
epsilon=0 and no charges (so you don't need Ewald sums), then you have
an ideal gas, so...maybe.

A quick way to check would be to try it with MM in both boxes, but set
all the LJ parameters and charges equal to zero in the vapor box.
Choose it to be large enough that the density doesn't really change
over the course of the simulation, don't do any molecule displacements
in the vapor box, and compare it against GCMC with a different code.
One concern I have is that, on swap moves, the code is going to check
for overlaps in the vapor box, which it should do for true GCMC.  If
you try this test and get an answer that is close, and you feel
comfortable changing the source code, I can let you know which lines
to change to remove those checks so we can see if it works.  I know of
codes that do GCMC as basically GEMC with a box that has no
interactions in it (and is big...500 molecules or more), so this
doesn't seem so different.

                                              Cheers, Matt


On Aug 22, 10:57 pm, Isaak Daniels <isaakdan... at gmail.com>
wrote:
> Can one do it so that the box not of interest (representing the
> environment) is MM while the other is done with QM, so one can avoid
> the pitfalls of enlarging the box?
>
> Thank you
>
> On Aug 21, 7:42 pm, Matt McGrath <obfis... at gmail.com> wrote:
>
>
>
>
>
>
>
> > Hi Isaak.  Unfortunately, this isn't really possible at the moment,
> > unless you do Gibbs ensemble Monte Carlo and make the second box huge,
> > i.e. essentially a reservoir.  Of course, if you're using any grid-
> > based method to compute the energy (Quickstep, Ewald) this adds a
> > whole lot of expense to the simulation as well, and probably isn't
> > worth it.
>
> > GCMC has been in the pipeline for a while (there have been issues with
> > the choice of reference state for QS calculations with GCMC), and is
> > the next big project on my list.  I'm about 30% done with the current
> > project (making the MC routines more streamlined), so it might get
> > implemented in the next year...but that's just a guess.  Could be six
> > months, could be a couple years, so if you're working on a major
> > project, it's probably best to use a different code.  Sorry!
>
> >                                 Cheers, Matt
>
> > On Aug 21, 2:51 am, Isaak Daniels <isaakdan... at gmail.com> wrote:
>
> > > Is there a way to run a sort of GCMC on Cp2k?
>
> > > Thank you


More information about the CP2K-user mailing list