Velocity discrepancy - Kinetic theory and MD using LAMMPS

LAMMPS LAMMPS Mailing List Mirror
1

Hi Axel,

Thank you so much for the prompt reply. Really sorry for not getting back to you earlier, I was down with fever and cough for the past week. I will try and provide more information about my simulation here.

** I am attaching an input script and logfile from a simulation: 2000 nitrogen molecules in a box at 10 bar and 300K.

** And yes, I am using NVE
** For this case, the simulation domain has an area of cross-section of 5 nm * 5 nm and a length of 340 nm.
** I do run simulations for a long time. Each 0.1 ns corresponds to 50,000 steps run. That way I run till 20 ns and observe no change in the plots. So that’s a total of 50,000 * 20 steps.

Problem / Doubt:
** My problem still is that when I plot the V-rms from MD I get around 750 m/s whereas kinetic theory assuming 3 Degree of Freedom (DOF) gives just 550 m/s. This is at 300K.

** I don’t get this discrepancy when I deal with Helium with is mono-atomic and perfectly spherical. I get really good agreement of RMS with kinetic theory. Now helium can only have 3 DOF.

----- Here are my questions:
** when I dump vx, vy and vz from lammps, are they only translational components or they have a contribution from rotational components as well.

** 300-350K is very low for vibrational modes to be activated, so no need to worry on that front
** Or can the difference for nitrogen : (MD giving rms = 750m/s and kinetic theory: 550m/s) be justified saying kinetic theory assumes single hard-sphere models and also no attraction between particles. Whereas in MD, N2 is definitely not one sphere and LJ potentials have attraction and repulsion sitting inside of them.

Will be looking forward to your reply. Have a nice day. Do let me know if you need any more information from my end.

Regards,

Suman

log.lammps.20200222_010920 (9.53 KB)

in-script (1.63 KB)

2

Hi Axel,

Thank you so much for the prompt reply. Really sorry for not getting back to you earlier, I was down with fever and cough for the past week. I will try and provide more information about my simulation here.

** I am attaching an input script and logfile from a simulation: 2000 nitrogen molecules in a box at 10 bar and 300K.

** And yes, I am using NVE
** For this case, the simulation domain has an area of cross-section of 5 nm * 5 nm and a length of 340 nm.
** I do run simulations for a long time. Each 0.1 ns corresponds to 50,000 steps run. That way I run till 20 ns and observe no change in the plots. So that’s a total of 50,000 * 20 steps.

Problem / Doubt:
** My problem still is that when I plot the V-rms from MD I get around 750 m/s whereas kinetic theory assuming 3 Degree of Freedom (DOF) gives just 550 m/s. This is at 300K.

** I don’t get this discrepancy when I deal with Helium with is mono-atomic and perfectly spherical. I get really good agreement of RMS with kinetic theory. Now helium can only have 3 DOF.

----- Here are my questions:
** when I dump vx, vy and vz from lammps, are they only translational components or they have a contribution from rotational components as well.

vx vy vz are the coordinates of the individual atoms. being point particles, they have no rotation, but they have a total of 3 DOFs,. which means each n2 molecule is considered to have 6 DOFs since you don’t use fix shake to form a rigid bond, that would remove one DOF. for the global temperature, also 3 DOFs from the total DOFs are removed as the system should be invariant against translation. apropos, using fix recenter is not really enforcing that, it will merely hide any effect of a center of mass drift from the positions of the trajectory, the velocities are untouched. you would need fix momentum, if you want to cancel a center of mass drift.

** 300-350K is very low for vibrational modes to be activated, so no need to worry on that front
** Or can the difference for nitrogen : (MD giving rms = 750m/s and kinetic theory: 550m/s) be justified saying kinetic theory assumes single hard-sphere models and also no attraction between particles. Whereas in MD, N2 is definitely not one sphere and LJ potentials have attraction and repulsion sitting inside of them.

in your case, you have a particularly large cutoff for the lennard-jones potential. did you have a same magnitude cutoff for your helium tests?
i have not looked into a text book about these matters for a long time, so i cannot really comment on how applicable this is in your specific case. i would expect that it will depend a lot on the density and typical MD simulation systems are of rather high density to assume ideal gas or slightly perturbed ideal gas behavior.

Will be looking forward to your reply. Have a nice day. Do let me know if you need any more information from my end.

this is really more of a question about the theory and less a question of using LAMMPS and thus i would recommend to discuss with your adviser or experienced colleagues. i cannot provide more help than what i already did.

axel.