BKS parametrization of buck/coul/long pair_style for silica


actually simulating systems like that was my original reason to start working on USER-CUDA (though I investigated ion conducting glasses).

I have attached a modified input script from the one you referenced where I copied in my pair parameter definitions. I also attached a plot of the pair potential as calculated from LAMMPS compared with the tabulated data (the commands to generate the new table list are included in the script). You see they are the same except for distances smaller 0.15A. This is because the buck/coul potential has a singularity at r=0 which you can avoid in a tabled potential by just turning off interaction below a certain cutoff.

For optimal performance you should probably use npt, nvt, and nve integrators (not nph and langevin) since those are directly supported right now. In order to run the simulation on the GPU you need to provide -sf cuda as commandline parameter to lammps.


-------- Original-Nachricht --------

in.SiO2-b (1013 Bytes)


Dear Christian,

Thank you kindly for your very informative answer.

I have plotted the output of your script against the tables provided
by Dr. Crozier and now understand the vertical asymptote at r=0 to be
the root of the discrepancies I had encountered when using the BKS
parametrization of the buck/coul/long pair_style.


Benjamin N. Doblack
Graduate Student
School of Engineering
University of California at Merced
[email protected]...