Better fit with pair style table

Dear all users,

   I have tried to use pair_style table for interpolating a Lennard-Jones 18-6 potential and checked the interpolation with pair_write. I need to use this potential because I want to simulate a "Prymary-Knock on Atom" dynamic. In this case, atoms reach positions very out of equilibrium and repulsion forces can be strong. Then, the potential closer to atom positions has very steep parts. The pair_table seems unable to fit well the potential. There is figure for the table and the fitted potential in attachment. Red line is the table potential and the blue dotted line is the written potential.

   As suggested by the manual, I tried to use one, ten and one hundred thousand points to get better resolution. I applied the different styles (lookup, linear and spline). I am using "RSQ" parameter to avoid additional interpolation. I have not seen much difference between the different conditions. My cuttof is one angstrom as one can note from the figure, but I imagine that I will need to use a inner cutoff. However, for a first good interpolation, I think this cutoff is acceptable.

   Does anyone have any advice for getting a better fit using pair_style table? I attached my tabulated potential data file (very long) and the input script if anyone needs to help me.

   Thanks in advance,

   Pedro Moreira


rsq.table (1.03 MB)

input.table (553 Bytes)

please try linear interpolation which is much more stable for steep
repulsive potentials or consider - in addition - overlaying pair style
zbl for the repulsion and make the tabulation flat for r -> 0