Hello,
I am just trying basic MD simulation with CO2 molecules and EPM2 force field.
I am running NPT simulation and measuring density.
When i set same density and temperature to NVT simulation, i get different pressure.
I am using LAMMPS simulation tool. There are 500 molecules of CO2 in my simulation. And i am averaging density (in NPT) and pressure (in NVT) every step over 1ns (1 million steps with 1fs step).
Does anyone know, where is problem?
Thank you!
EDIT:
Here is my result. Experimental data was taken from NIST database. NPT data was taken by setting pressure and measuring density. NVT data was taken by setting density of system and measuring pressure. All done by LAMMPS tool.
for T = 365 K
I would suggest that you first also show the statistical errors: in NVT, pressure is not constant and it does fluctuate, and in NPT volume (and so the density) fluctuates. In NPT, in addition, depending on the barostat you may use, the pressure can also fluctuate.
The second comment I have is related to the number of particles, 500 molecules may not be such big number to expect a good equivalence, I think. So, you may try for a bigger system.
The third observation is related to the curves, it seems a (almost) parallel shift between the experimental and the simulation results. I would also check pressure calculations: it should two terms, the density dependence and temperature term, plus the virial term. In addition, I would also check for long range corrections in pressure (see Daan Frenkel and Berend Smit book). Furthermore, is the pressure calculated based on atoms or molecules? I would check this as well, it may affect the fluctuations of pressure.
I would suggest that you first also show the statistical errors: in NVT, pressure is not constant and it does fluctuate, and in NPT volume (and so the density) fluctuates. In NPT, in addition, depending on the barostat you may use, the pressure can also fluctuate.
The second comment I have is related to the number of particles, 500 molecules may not be such big number to expect a good equivalence, I think. So, you may try for a bigger system.
The third observation is related to the curves, it seems a (almost) parallel shift between the experimental and the simulation results. I would also check pressure calculations: it should two terms, the density dependence and temperature term, plus the virial term. In addition, I would also check for long range corrections in pressure (see Daan Frenkel and Berend Smit book). Furthermore, is the pressure calculated based on atoms or molecules? I would check this as well, it may affect the fluctuations of pressure.
Another possible source of differences is that the system get trapped in some metastable state. For instance, although you should be supercritical, it is possible that with a small number of molecules and very short runs, some clusters (maybe due to initial condition) are present. To avoid this problem you can eventually run a long simulation at higher temperature (to destroy all possible structures) and then cool the system to the desired value of T.
Anyway, as suggested by Hiqmet Kamberaj, the first check should be to see what heppens increasing the number of particles (at least one order of magnitude) and add error bars.
You have at least three choices:
1. berendsen barostat/thermostat (but can not require time intefgration & change .fix_press+berendsen.cpp for triclinic boxes -->> very dangerous!)
2. NVE+NVT+(fix box/relax + pressure tool).
3. NVE+NVT+fix press/berendsen (dangerous because of the non-time-integrational press/berendsen tool).
May be second is the best choice.
Cheers
- Badges
- Science topic