Marcel,
The current version of gcmc turns the full_energy option on automatically when pair style hybrid is used. This explains why you are still seeing a time difference. You can probably turn this off easily in the source code.
Are you sure that the system is reaching equilibrium? It seems you are attempting a small number of MC moves and it only makes sense to compare equilibrium properties derived from each run, not some arbitrary quantity after some small number of moves that do not cause the system to equilibrate.
-CBL
Gcmc equilibrates at about 400 000 insertion moves so in order to provide quick results i did not try it as it would take about 50 hours. I will try and report also. But i must state that, from previous simulations i made, with kspace enabled, equilibrated results also differ, and still only one or two molecules are inserted succesfully in the system with lammps with full_energy.
Also is it still normal that the insertion successes being so different (38 vs 2 ) even if the system is not equilbirated yet?
Thanks
Marcel
Marcel,
I just posted a gcmc bug to the mailing list which is extremely relevant to your issue. Basically when inserting charged molecules, the coulomb energy is incorrectly calculated when using the full_energy off option. This in turn messes up the insertion probability. I have not run detailed tests yet, but I think the full_energy on option works properly.
So what does this mean for you? Any simulations for charged molecules run with full_energy off (this includes the old 9Dec14) are probably wrong. Additionally, it was mentioned by someone else on the list not to compare results obtained using the 9Dec14 and the latest version of lammps (with kspace on) since the old version only considered short range pair interactions.
As for the different number of successful insertions, this is probably due to the insertion bug I mentioned above. However, in general I do not think you can expect the convergence “path” of two markov chains to be exactly equivalent. All that is expected is they eventually converge to the same stationary distribution.
-CBL
Marcel,
Just a quick clarification. In my last message I mentioned ‘charged molecules’. What I meant by this is a molecule which contains charged atoms. The molecule as a whole can be neutral.
-CBL
So trying a “non-complicated” system with full_energy on and off would be meaningless right now as gcmc is bugged if full_energy is off, am i right?
Marcel
Correct, if your “non-complicated” system has molecules which contain charged atoms. If this is not the case, full_energy off is okay. You can also alter the source to fix the bug for your specific case until a patch is released, that’s what I ended up doing.
-CBL