please always copy the mailing list on your replies, not just
individual people. thanks.
Alex,
Thanks for your quick and useful response!1: I'm starting with a solid alpha-crystalline system of MEAM Sn, generating
a velocity distribution and then thermostatting just below its melting
temperature.
good sampling near phase transitions is always difficult to get. you
also have to worry about finite size effects.
2: I was starting with NPT right away, unaware that the relaxation time was
so long. If I run under NVT for long enough to see many thermostat
oscillations, it appears I get nearly identical results under NPT.
neither are good thermostats to initialize a solid system with. better
would be fix langevin in combination with fix nve and even better
should be the CSVR thermostat (temp/csvr) that was added in today's
patch. those provide better equipartitioning than what you get with
fix nvt/npt. in the latter case, you usually need to couple each DOF
to a single NH-chain to enforce equipartitioning during the initial
step of bringing your system to equilibrium. in (ideal crystalline)
solids, the thermostat oscillation (and even more so the
barostat/pressure oscillations) can be quite large. fix nvt/npt have
the drag option for compensation, but that will kill the ensemble.
Perhaps I am further from equilibrium than I thought.
might be, or your system is too small or both.
axel.