GCMC with full-energy

Hi Lammps users

I have previously stated that the new fix gcmc module was much slower with the new “full-energy” option on (about 30 times slower) and it has been said that it was normal to be so. I will again keep you busy with a question about the new algorithm

While it is acceptable, i have a new problem with the new module, which is big difference between the sorption results and with difference i mean either one or none molecules sorped when using the new lammps with gcmc module. With i have about 50 insertion successes of 45000 insertion moves with older lammps, i have either one or none successes with the new lammps. Is this much difference also normal? I have also must state that the results of older lammps are matching well with experimental results we made in our lab.

With old lammps i mean 9Dec14 version and with new one i mean 15may15 version, and all simulations are carried out with 12 intel xeon cores. I both use pair style hybrid and kspace solver for long range electrostatics. The sample input and sample output files are available in attachment.

co2.txt (483 Bytes)

lognew.lammps (16.2 KB)

logold.lammps (14.2 KB)

inputgcmc.in (11.1 KB)

dengeli6BpB.data (2.15 MB)

If you are comparing the 9Dec14 version with the 15May15 version then you must be using kspace in the new one but not in the old one, right? That could make a huge difference.

In other words, try turning off kspace and pair hybrid with the 15May15 version and make your problem exactly the same as the old case using the 9Dec14 version. Then do you get the same answer with the 9Dec14 version vs the 15May15 version?

Stan

I am trying that since your answer, yet despite I decreased gcmc moves 10 times, the one with 15May15 lammps still did not produce result despite 9Dec version resulted in 15 seconds. I will report the results as soon as i get them.

Thank you,
Marcel

I attached the results of both 9Dec14 and 15May15 lammps results for same gcmc simulation which has a pair_style hybrid but no kspace calculations. To sum up, the first gcmc ( i have two in the input) insterts, 38 molecules succesfully with 9Dec14 version, while the number is only 2 for 15May15 version and it take an hour to make the move with 15 May version while 9Dec version does it at only 140 seconds.

co2.txt (483 Bytes)

dengeli6BpB.data (2.15 MB)

inputgcmc.in (10.4 KB)

logoldnokspace.lammps (13.5 KB)

Hi lammps community

I have been working on to simulate same structure, with or without full_energy. Also i removed charges from the inserted molecule as it was told that there was a bugged line of code for charged molecules when full_energy was off.

There is again a huge difference between the results of simulations and the simulation with full_energy option on, failed to sorp any molecule. You can find the files attached.

co2.txt (429 Bytes)

logfull.lammps (3.49 KB)

lognofull.lammps (3.49 KB)

inputgcmc.in (1.42 KB)

6BpB.data (2.04 MB)