I have a problem with low density while running NPT in polymers
I am running a small simulation of bulk polymer using 3D periodic cell. The polymer has 3 chains of 50 monomers for a total of 2256 atoms. Force field is Dreiding with harmonic bonds, angles, torsions and impropers; and a buck/coul/cut VdW potential.
Samples are created in C2 and equilibriated with a series of NVT compressions-->minimizations-->dynamics (i.e., I believe I have fairly stable samples). I stop when I get the sample to have the density of bulk PMMA +15% (hence high internal pressure); then switch to Lammps for the production NPT runs. At the bottom you will find the script with the simulation set-up, the first pressure cycle (which is a decompression to atmospheric pressure) and the first temperature cycle. This is followed by 2 pressure cycles and a final T cycle.
My question is: Density from the simulation is ~25% lower than expected across all the pressure range. I should be getting 1.30g/cc at 10,000 atm, and I am getting that value at 25,000 atm instead. Density at atmospheric pressure should be 1.15g/cc, I get 0.8.
I played with the Pdamp factor (from 10.0 to 5000.0) and with the drag factor (from 0.2 to 14!) and though I got some improvements, the density is still 25% off. I recently tried reseting the velocities in every run and it helped: Initially, density was going to 0.1g/cc during the T cycles due to the uncontrolled expansion generated while decompressing; now it only goes down to 0.7g/cc.
I double checked my FF parameters, thrice, and they are right. I went back to 2006 in the Lammps mail-list archives and found some tips, but none in particular about low density or problems with the buck potential. Now I am running the same simulation in C2 in order to compare the energies and decide whether the problem is in the model or in the algorithm.
Funny thing, I created a coarse grain FF based on the same Dreiding parameters, replaced the atoms in my sample by beads and ran the same simulation using the coarse grained model/FF. The densities are much closer to the experimental values.
Any help would be greatly appreciated.