I'm still having some problems with my barostat, but I've been able to
isolate the problem to the constraint virial **.
Can anyone with experience programming a calculation of the constraint
virial (preferably in the context of Velocity-Verlet integration) please
reply to my private email so we can discuss the finer points of the
algorithm and hopefully resolve my dilemma?
Dzeinkuje,
Dave
>>>>>>>>>>>>>>>>>>>>>>>>>>
** Proof of point:
- When I simulate an ideal gas, my barostat takes the system to the
correct volume given the applied external pressure (this means that the
kinetic energy term in the pressure calculation is correct).
- When I simulate a Lennard-Jones gas with a sizeable epsilon parameter,
my computed energies are in close agreement (1%) with AMBER and the
final system volume is within 0.3% of that obtained by AMBER. I think
there must still be some long-range correction running around that I can't
quite get rid of, but the essential result is that my code and AMBER
produce very similar final volumes for this gas system, and those final
volumes are a significant (12-fold decrease) departure from the volume of
a similar amount of an ideal gas. So, the non-bonded virial (at least in
the case of monatomic particles) is correct.
- When I simulate a system of TIP3P water molecules without using SETTLE
but rather a crude system of bonds between O-H1, O-H2, and H1-H2, the
system also reaches a reasonable and stable final volume, close to a
density of 1.0 g/cm3 (it is reasonable at this point to assume that any
deviation is a result of not making the molecules rigid). So, the
non-bonded virial, again, is correct, and I can simulate tri-atomic
molecules under NPT conditions without trouble.
- When I try to simulate TIP3P water (or any other three-site water) with
SETTLE for rigid molecules, the system blows up (slowly but exponentially,
it expands until a numerical catastrophe hits). My constraint virial is
therefore obviously the problem.
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Feb 14 2007 - 06:07:37 PST
