Dear LAMMPS users,

When I ran liquid-vapor equilibra cases using TIP4P water model in an NVT ensemble at 370K, the pressure fluctuated from -50 bars to 50 bars. The error bar was very large. I am curious about the reason of such larger fluctuations. Is there any way to get the pressure more accurately?

Best regards,

Han Hu

Complex Fluids and Multiphase Transport Lab

Mechanical Engineering and Mechanics Department

Drexel University

Hess Engineering Research Laboratories

hh398@…1685…, edenkira@…24…

Dear LAMMPS users,

When I ran liquid-vapor equilibra cases using TIP4P water model in an NVT

ensemble at 370K, the pressure fluctuated from -50 bars to 50 bars. The

error bar was very large. I am curious about the reason of such larger

fluctuations. Is there any way to get the pressure more accurately?

if you check through the documentation and the mailing list archives

you'll find *many* discussions of the subject. this is a very fundamental

issue of doing MD simulations and you should have long ago learned

about it (probably without realizing the consequences in practice).

in short:

this has nothing to do with "accuracy" but rather with "precision"

(i.e. it is not about correctness of the math/physics but convergence

of statistics).

liquids in general are not very compressible and you *always*

have large *local* fluctuations. the *total* pressure fluctuations

now depend on the size of your system (since the local fluctuations

are uncorrelated, the farther they are apart). the larger the system,

the smaller the *total* fluctuations. you have the same for kinetic

energy (i.e. temperature) only that the oscillations are smaller.

what we normally measure as pressure and temperature is really

only the limit for an infinitely large system. that we map it to a

smaller system is based on the assumption of equipartitioning.

for a system in equilibrium, this is not a problem, since

the average over space should be equivalent to the

average over time.

cheers,

axel.