ADP potential file format

Hi Everyone,

I got the February 2012 package the other day to start working on fitting an ADP style potential. From the “pair_style adp” manual page, I read…

"Following the Nelements sections, Nr values for each pair potential phi® array are listed for all i,j element pairs in the same format as other arrays. Since these interactions are symmetric (i,j = j,i) only phi arrays with i >= j are listed, in the following order: i,j = (1,1), (2,1), (2,2), (3,1), (3,2), (3,3), (4,1), …, (Nelements, Nelements).

After the phi® arrays, each of the u® arrays are listed in the same order with the same assumptions of symmetry. Directly following the u®, the w® arrays are listed."

… and so I wrote a script to generate an input file with tabulated F(rho), rho®, phi®, u®, w®, etc. It soon became clear that something was wrong, and I recalled that in pair_style eam/alloy, for example, r*phi® is tabulated instead of phi®. I made the change and things worked closer to my expectations. I also compared some of the ADP potentials that came with the LAMMPS distribution and plots of phi®, u®, etc. in the corresponding journal papers and they matched.

So it seems to me that the ADP file format uses F(rho), rho®, r*phi®, u®, w®, etc. In other words, phi® is the only function scaled by ®. Can someone else verify this for me? Also, if this is correct, might I suggest adding the line…

“The tabulated values for each phi function are listed in setfl files directly as r*phi (in units of eV-Angstroms), since they are for atom pairs.”

… from the “pair_style eam/alloy” manual page to the “pair_style adp” manual page? This would perhaps save someone else some confusion.

Thanks for your consideration and to all the LAMMPS developers for their hard work.

Jason Gruber

I think you are correct about r*phi(). I will send
an email to the ADP authors to verify, then edit
the WWW page as you suggest.

Thanks for ferreting this out rather than just
posting an email like "help, ADP isn't giving
me the results I expect".


Jason - you were correct. It is r*phi(r). I changed
the doc page.