Trying to minimize the pressure fluctuations in NPT, I have realized that I don’t understand the issue of tchain and pchain. In the fix npt command I modified the pchain value to be 6 on the one hand, and to 1 on the other. I expected to see smaller fluctuations in the pchain=6, but the opposite happened with f(1)<f(3)<f(6).
I am no expert but it seems like the paper by Martyna cited in the docs explains it very clearly. They write:
…Let us consider how this situation can be improved. The
distribution itself has a Gaussian dependence on the parti-
cle momenta, p, as well as thermostat momenta, p,. While
the Gaussian fluctuations of p are driven with a thermostat,
there is nothing to drive the fluctuations of p(thermostat). Although
the postions and momenta (the q’s and p’s) are of primary
interest, if the Nose-Hoover equations are ergodic, the sys-
tem covers the whole phase space, which includes the space
of the thermostat velocities. The fluctuations of the ther-
mostat variables which clearly occur in ergodic systems
may even be important in driving the system to fill phase
space (the dynamical equations are, of course, coupled).
This suggests thermostating p(thermostat), and, by analogy, the ther-
mostat of p(thermostat) plus its thermostat, etc., to form a chain.
I am no expert but it seems like the paper by Martyna cited in the docs
explains it very clearly.
for people without journal access, I think I found them here: