Hi, Tyler,
Berendsen is very strong it just damps any departure of temperature
from the target value exponentially, but it does not give the
canonical ensemble, while
Nose-Hoover is more delicate, in LAMMPS it has
this Tdamp parameter
which is related to the extended degrees of freedom "mass" and it
therefore affects the T fluctuations
(Tdamp of a few hundreds time steps is the best, the documentation says)
and Tchain (=3 by default) which tells how many of those extended
degrees you do actually have. The more of them, the better ergodicity
is.
Playing with these two might help in your case?
P.S. I'm trying to learn the correct way to reply to the messages in
this mailing list, hope I'm doing it right and this message reaches
the right destination.
Best regards, Maxim