Regarding two Tersoff potentials

Hello all,

I am using hybrid pair style under which i am using two tersoff potentials. With first tersoff i am taking the interaction between Be, Be and C and the second tersoff potential is used for C. I assume the C-C interactions with first tersoff are replaced by second tersoff potential. I want to know is it right way to assign C-C interaction by second tersoff.(The program is running without any error)

# There are three types of atoms Cu, Be and C. *Cu:*1-3, Be-4, C:5-7.
pair_style hybrid tersoff tersoff morse 2.5 eam

pair_coeff * * tersoff 1 BeC.abop.tersoff NULL NULL NULL Be C C C
pair_coeff * * tersoff 2 C.abop.tersoff NULL NULL NULL NULL C C C
pair_coeff 13 57 morse 0.087 5.14 2.05
pair_coeff 13 13 eam Cu_u3.eam
pair_coeff 1*3 4 morse 2.14 5 2.065

Hello all,
I am using hybrid pair style under which i am using two tersoff
potentials. With first tersoff i am taking the interaction between Be, Be
and C and the second tersoff potential is used for C. I assume the C-C
interactions with first tersoff are replaced by second tersoff potential. I
want to know is it right way to assign C-C interaction by second
tersoff.(The program is running without any error)

as has been explained *many* times on this mailing list, ‚Äčtersoff is not a
pairwise additive po‚Äčtential, so you do not only need to consider
interactions between pairs of elements (i.e. Be-Be, Be-C, C-C) but also you
have terms for triples. with your setup, you will have an inconsistent
representation: how would terms of the kind C-C-Be or Be-C-C or C-Be-C be
represented? the neighbor list for the first tersoff potential will not
contain the C-C pairs, while the second tersoff potential will not know
about pairs containing Be atoms. so the only way to implement this kind of
mix of parameters is to merge the various parameter terms from the two
potential files into a single potential file, i.e. replace terms in
BeC.abop.tersoff with terms from C.abop.tersoff.

however, it is considered a bad idea to mix terms for the same potential
function from different parameterizations. each parameterization can follow
a different strategy to balance the components contributing to the
interactions. after all you are reducing something where everything
interacts with everything (e.g. as represented in a quantum mechanical
hamiltonian) with a simplified model via some kind of empirical mean-field
approach. this does not only apply to tersoff, but to all potentials
implemented in LAMMPS. if you mix contributions from different
parameterizations, the result has to be considered a new parameterization
and thus first needs to be characterized in the same way with sufficient
detail.

axel.