Hello All,
I’m sure a similar question has been discussed on multiple occasions, sorry for asking.
I’m simulating a system (15000 atoms) with the charmm22 force field. I’m getting pressure fluctuations (fix npt) ranging from -1100 to 1200 atm, with the average sitting at -19 atm (wanted 1 atm) after 6 ns. I’m wondering if these are normal fluctuations and average? I know that pressure fluctuations will become smaller as system size grows, but how big does one have to go to minimize fluctuations? I ask because you guys are the PhDs and I’m the grad student trying to learn 
Thanks for your input, and have a good holiday.
Hello All,
I'm sure a similar question has been discussed on multiple occasions, sorry
for asking.
I'm simulating a system (15000 atoms) with the charmm22 force field. I'm
getting pressure fluctuations (fix npt) ranging from -1100 to 1200 atm, with
the average sitting at -19 atm (wanted 1 atm) after 6 ns. I'm wondering if
these are normal fluctuations and average? I know that pressure
fluctuations will become smaller as system size grows, but how big does one
have to go to minimize fluctuations? I ask because you guys are the PhDs
and I'm the grad student trying to learn
...and - as usual - you'll get the "it depends" answer. 
seriously, there is no absolute statement that can be made.
most importantly, the magnitude of the fluctuations has to be
inversely proportional to the compressibility of the material.
but there are all kinds of additional things related to the
other settings in the input and the type of potential and so on.
cheers,
axel.
Axel, thank you for your reply!
Do you know of any good references for the inverse relationship between compressibility and fluctuations?
Are these fluctuations studied or the methods to control them based more on experience?
What I’ve noticed is that using Berendsen with the modulus set at 20000 (as specified in the charmm papers), I get much smaller fluctuations than with NPT.
These simulations are supposed to be in the solid state with box dimensions around 7 to 10 nm, so I was expecting smaller pressure fluctuations.
Thank you again for your input.
Axel, thank you for your reply!
Do you know of any good references for the inverse relationship between
compressibility and fluctuations?
no. probably any text book on condensed matter physics.
but it also just comes out of simple consideration of the
properties of matter.
Are these fluctuations studied or the methods to control them based more on
experience?
yes, this is part of statistical mechanics.
What I've noticed is that using Berendsen with the modulus set at 20000 (as
specified in the charmm papers), I get much smaller fluctuations than with
NPT.
that is no surprise. the berendsen method is based on an exponential
decay, i.e. damping and has no rigorous statistical mechanical derivation.
the NPT integrator, however, is coupling the motions of the cell dimensions
to a single or a chain of harmonic oscillators, which can be shown to be a
rigorous representation of the NPT ensemble. for large enough systems,
small enough damping (i.e. large enough time pdamp time constant) those
should converge to the same limit.
These simulations are supposed to be in the solid state with box dimensions
around 7 to 10 nm, so I was expecting smaller pressure fluctuations.
solid state systems are problematic. particularly with PBC, you always
have to deal with not only the "natural" fluctuations, but sound waves
travelling through your system that won't go away quickly. thus it is
advisable - as explained many times - to equilibrate the system with a
dissipative thermalization method like langevin with a short time constant,
increase that time constant and then switch to a desired setup. similarly,
any operation that may result in a "kick" or "shock" to the system has
to be avoided.
length of trajectory has little impact at the time scales
that we're talking about here.
axel