How to use a smaller simulation box with data files designed for larger box

Dear All,

I have a simulation box of a particular size with all proteins and water molecules in it, but it takes very long time to run. I know I can change simulation box size to a smaller one and recreate the data files for all the molecules to be inside the box, but that is very time consuming as well, and I was wondering how to tell LAMMPS to just simulate a smaller region without changing all data files.

Thank you so much,
Regards,
Delaram

What you request cannot be done and it doesn’t make sense. Even if possible, it would lead to tainted results. If you reduce the amount of water in a simulation of solvated proteins for example, the proteins would have unphysical behavior due to interacting with themselves or being in an extremely high and also unphysical concentration. Even if the individual simulation would run faster, you are still wasting time. It does not pay to get bad results faster.

If you have not enough compute resources, you have to look into finding additional ones. Most universities have HPC clusters and most countries have regional and national supercomputing centers where you or your adviser/supervisor can apply for time.

1 Like

I think I did not explain my question correctly. I do not want to change the density of proteins or water or CNT, My CNT is long and Imy question is if there is any way for LAMMPS to look at part of it interacting with proteins and water without deleting any molecule or changing density. My last option would be for sure using additional computational sources.

This doesn’t really change my assessment. This will also impact the validity of your simulation, since you are effectively embedding your “production” system into a block of ice at 0K. Furthermore, the only interactions you could avoid would be the bonded and non-bonded interactions of the area that you freeze. You cannot avoid the interactions between the mobile part and the immobile part and you won’t save any time on the long-range Coulomb solver, since that must be applied to the entire volume. By effectively freezing the outer atoms you are also suppressing fluctuations of your system and thus are creating additional unphysical behavior.

In summary, the savings are likely much less than you are expecting and you still will have tainted results. Similar savings may be possible using the r-RESPA integrator at the expense of some loss in energy conservation, e.g. by evaluating kspace only every second timestep.

Utimately, saving CPU time is no excuse for tainted simulation results and thus you should focus your energy on obtaining access to more compute resources, if needed through collaboration.

1 Like

Thank you so much for detailed explanation.