[lammps-users] SW potential mixing rules

I was wondering if anyone knows the details for the mixing rules used in
the Stillinger-Weber model in LAMMPS. In particular, I am concerned
about the sigmas in the 3-body term. The SW was originally developed
for a single atom type, but when mixing of two or more atom types, it
seems the sigmas used in the 3-body part should correspond to the
distance between the two atoms used in that portion of the calculation
of the 3-body term. This is how it was done by Ethier and Lewis (J.
Mat. Res., vol 7, p2817, 1992). In other words for the 3-body part as
described in the LAMMPS manual
(http://lammps.sandia.gov/doc/pair_sw.html), the part where r (r_ij) is
used, sigma_ij would be used and the part where s (r_ik) is used,
sigma_ik is used. But the way the sw pair coefficient files are
defined, it is not clear to me whether this approach is being used
opposed to just using a single sigma. For example consider the
coefficients for Si Ge Si and Si Si Ge. In both of these cases it
appears that only one sigma is being used regardless of whether the
portion of the 3-body part is between Si-Si or Si-Ge, is this a correct
assumption about the LAMMPS SW model? If so, how are folks defining the
sigmas when calculating the potential between one atom (Si) and two
other atoms, one which is the same as the central atom, such as Si Si

Thank you,