I've had to put aside the Monte-Carlo barostat implementation in mdgx for
the moment, but I think the Langevin superposition barostat is ready for
some test runs on simple cases. It works with any of the thermostats,
fortunately, so we can do rigid waters with the barostat.
More recently, I've been able to exploit repeating pseudo-random number
sequences in the barostat, allowing me to get phase transition properties
of (TIP3P) water to remarkable precision with only a few tens of
nanoseconds of MD. Boiling and freezing TIP3P water of course doesn't
produce a very accurate phase diagram, but if I properly calibrate the
random sequence length to match the pressure coupling time constant, I'm
able to converge multiple trajectories starting from independent starting
configurations. So the technique definitely "works," it's merely a matter
of feeding it a water model that can realistically venture as far as the
sampling model will take it.
Dave
> Hi everyone:
>
> I want to give you a heads-up about a new thermostat we are working on.
> It's
> based on the Langevin idea of adding a random-force componenent, but
> borrows
> an idea from quantum computing: the extra forces can represent several
> different temperatures at once. This has some important implications.
> For
> example, if your dynamics undergoes spontaneous conformational transitions
> (e.g trp-cage folding/unfolding) then you can extract the entire
> temperature
> dependence from a single run. Or (for more complex sampling problems)
> temperature replica exchange simulations can be carried out with a single
> replica.
>
> This is not yet ready for general use, but the basic ideas have been
> implemented in sample code. Right now, it doesn't work with SHAKE
> constraints, and the some instabilites set in if you have more than a few
> dozen temperatures in a single run. But we can get (SPC-fw) water to go
> from
> supercooled to just under boiling: this is not a classical cycling of
> temperature from low to high values, but a true quantum-like superposition
> of
> temperatures.
>
> (Dave Cerutti thinks he can make a barostat do the same thing.)
>
> ...details later....dac
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Apr 01 2011 - 13:00:02 PDT
