Pressure fluctuation control in lammps using NPT

I am running a test setup in lammps. I am using 4000 atom hcp system with meam potential. System need to be at equilibrium at 10K and for this I am giving initial velocity at 10K and running NVE for short duration of 2ps (My timestep is 1 fs) followed by NPT at 10K for 20ps. I want a configuration where system is at 10K with small pressure.
step 1: NVE for short time
velocity all create 10 4928459
fix 1 all nve
run 20000
Output
Step Temp TotEng Press
18500 4.9325929 -6191.5587 135.54744
19000 5.0476045 -6191.5587 134.10653
19500 5.0523007 -6191.5588 133.15733
20000 5.1187478 -6191.5588 132.81736
20010 5.0694554 -6191.5587 133.90782
Final output is as aspected and system equalibriated at 5K as we started with initial velocity of 10K and 133 bar of pressure system exhibits

Step2: NPT
fix 1 all npt temp 10 10 0.1 iso 0.0 0.0 1
run 200000
OUTPUT
217000 10.04384 -6186.4403 157.98486
217500 9.7292229 -6186.5088 -11.700501
218000 9.7728786 -6186.3937 -92.777014
218500 10.165017 -6186.3155 -69.765655
219000 9.875577 -6186.5188 -10.970014
219500 9.9095766 -6186.5424 74.030269
220000 10.251927 -6186.3377 182.70663
220010 10.259622 -6186.3333 76.85412
Here I used Tdamp of 100x timestep and Pdamp of 1000x timestep. I have also changed different Pdamp value also and found that pressure varies from -200 bar to +200 bar.
I can suppress this fluctuation by introducing drag term to Nose-Hoover equation.
for example :
fix 1 all npt temp 10 10 0.1 iso 0.0 0.0 1 drag 1
output
39000 10.074096 -6186.3848 0.88364453
39500 10.001291 -6186.3747 0.29228989
40000 9.9810361 -6186.3857 -0.12452421
40010 9.7588549 -6186.392 -0.91715159
But I want to suppress these fluctuation in pressure only using real ensemble(with no drag). I tried to increase further run to 100 ps but I can still see fluctuations. How much fluctuation in pressure is accepted as these fluctuation increases with temperature also. Any further insight is highly appreciable

1 Like

Pressure always fluctuates (temperature, too). How much depends on the size of the system and its compressibility (or exchange of momentum). Condensed matter is not very compressible, so pressure fluctuations are typically much larger than temperature fluctuations, especially around zero pressure. The only “natural” ways to reduce those fluctuations is to use a larger system or to average over time (provided your system is in equilibrium and thus system average and time average are equivalent).