[gmx-users] Highest possible temperature in NPT?

[chris.neale at utoronto.ca]

It is not immediately clear how one would implement REMD with NPT. In the
simplest instance, your monte carlo test would be based on unscaled volumes
(that would later scale) and including data from some time period X after a move
would lead to a non-maxwellian distribution that was overly spread to other
temperature umbrellas (worse, your lowest temperature would be unbalanced by the
presence of only higher-temperature skewing data.)

Probably you need to scale volume at the same time that you scale temperature in
your REMD system. This should solve your other problems.