MANYBODY hybrid potential

Dear LAMMPS Users,

I am trying to mix different potentials in order to describe the interaction of a silicon tip with an alumina surface.

I chose to adopt potentials belonging to MANYBODY group, since they are suitable for this kind of application and it looks like they can be mixed without losing the physical meaning of the simulation. Please correct me if I am wrong.

The interaction between Al and Si is described by the MEAM potential Jelinek_2012_meam_alsimgcufe, while the interactions between Si O and Al O are described, respectively, by the SiO.1997.vashishta file and an equivalent file containing the parameters for Al and O, which have been found in a paper by the same author.

this is conceptually incorrect. you are trying to use manybody
potentials as if they were pairwise additive, but they are not. so
they not only consider pairs of atom types, but also triples and some
even groups of 4 atoms. also a potential for SiO2 is not applicable
for pure Si.

I am using the command hybrid/overlay in order to add the contributions coming from every potential to the final definition of the forces on the atoms.

using hybrid/overlay is the wrong choice here, as this will not
prevent you from double counting interactions. your input *must* work
correctly with just hybrid or else it won't be correct.

I noticed that the parameters describing the forces between O-O are different in the Si-O and the Al-O vashishta potentials and I would like to be sure to avoid the addition of meaningless contribution to the hybrid potential.

yes, they *must* be different, because empirical potentials are
parameterized for specific compounds, not elements. you are not doing
quantum chemistry here.

In my opinion, I should keep the O-O interaction parameters coming from alumina sourcefile and set to zero the contribution from SiO.1997.vashishta file, since it should be applied to oxygen atoms in silica, which is not present.

this in conceptually incorrect, as explained above.

Following the same principle and considering that meam potential embeds the Al-Al and Si-Si interactions, I should not consider the contributions for the same couples coming from the vashishta potential, avoiding the application of a double interaction between the same couples of atoms.

you are disregarding, that also triples of atoms need to be considered.

I will be grateful if someone more expert than me comments my statements and corrects me if I am wrong.

you are wrong. what you propose is bogus. you *can* use a suitable
manybody potential each (e.g. tersoff) for Si alone, and for alumina
alone, but you cannot "connect" the two systems with a manybody
potential, particularly not a different kind. these interactions must
be a pairwise additive potential (e.g. lj/cut or morse). for an
interaction of a tip with a surface, e.g. to do nanomachining, that is
quite ok, since you are really only requiring something the models
hard repulsion between the two materials well, and you won't have
subtle/soft interactions.
otherwise, you have to use a single parameterset that can handle
*both* materials at the same time. given the difference in properties
chemical nature, this is probably only possible with reaxff, provided
somebody has created such a parameter set (note, that those are not
very transferable either, but also have to be parameterized
specifically for the materials modeled and the conditions).