I try to perform grand canonical Monte Carlo simulation for binary mixture by LAMMPS.
I want a pure GCMC simulator.
Firstly, I test the program by the simulation of one component hard sphere fluid. The cutoff distance of LJ 12-6 potential is set equal to sigma, it is not a perfect hard sphere system, but the density of the system from the simulation is very close to the theoretical one.
However, there is a problem when I do the simulation for a binary hard-sphere mixture. The densities predicted by the simulator are always far away from the confirmed theoretical ones.
The GCMC exchange is performed by these two commands in my script:
fix 1 O gcmc 1 1 1 1 29494 $T 3.982254 0.1 group O
fix 2 C gcmc 1 1 1 2 29494 $T 1.065682 0.1 group C
In addition, the density of the atoms handled by the second fix command is always close to zero. Here, the density of group C is almost zero, but if I change the order of those two fix commands, then the density of group O is close to zero.
The LAMMPS Version is (6 Jul 2017).
Please find the full .in script and the .data file from the attachment.
There are a lot of complicating factors in your example. Also, it is gigantic. I can’t check every possible thing you might be doing wrong. However, I did run my own tests on a modified version of in.gcmc.lj, in which I ran two gcmc fixes with two LAMMPS types with identical interaction coefficients. The total density and pressure of the system came out slightly lower than for the 1-component case, while the excess chemical potential came out slightly higher. I think is due to the fact the ideal gas contribution to chemical potential is different by shifted by kTln(2) for 2-component case. In any event, the numbers of types 1 and 2 are the same, within statistical noise.
There are a lot of complicating factors in your example. Also, it is gigantic. I can’t check every possible thing you might be doing wrong. However, I did run my own tests on a modified version of in.gcmc.lj, in which I ran two gcmc fixes with two LAMMPS types with identical interaction coefficients. The total density and pressure of the system came out slightly lower than for the 1-component case, while the excess chemical potential came out slightly higher. I think is due to the fact the ideal gas contribution to chemical potential is different by shifted by kTln(2) for 2-component case. In any event, the numbers of types 1 and 2 are the same, within statistical noise.