I’m sorry about yet another (long) mail about the Tersoff potential parameters, as there have been so much already (most recently last week…). I write this mail because recently I found an error in my own potential file for BN, which was based on the work of Albe and Möller[1,2], which I used a couple of years ago. This made me look into the standard Tersoff potential files again to create a correct one, after which I was left confused. This comes mainly from subtle differences between ordering in SiC_Erhart-Albe.tersoff and BNC.tersoff . There are some other issues with the latter file I found, which I will mention below.
For the SiC_Erhart-Albe potential different coefficients are given for C-C, Si-Si and C-Si interactions (from ). For the two body interactions everything is fine and done according to the manual. For the three-body interactions the coefficients are taken like i,k,j. Example:
C Si C --> C-C coefficients
C C Si --> C-Si coefficients etc.
Meaning that k, the second element in the potential file is not really relevant for the 3-body interactions.
For BNC this is done differently, ignoring the C interactions one keeps interactions for B-N, B-B and N-N. Here the three-body terms are taken like i,j,k, which seems more logical. For example:
B N B --> B-N coefficients
B B N --> B-B coefficients etc.
To me it seems, maybe naively, that one of them should be wrong. Looking through the mail list and documentation I find some conflicting information. For example, the i,k,j scheme is mentioned in:
while here it seems that it should be i,j,k:
The documentation seems to be ambiguous about this fact, as it mentions that SiCSi and SiSiC will (in general) not be the same. This statement is true for both i,j,k ordering and i,k,j. For most cases, eg. perfect solids, it goes well either way. When one has defects like two interchanged atoms, these differences become important.
If somebody could clarify this it would be greatly appreciated.