New COMB parameters + performance problems of the COMB potential in a hybrid

Dear LAMMPS community,

We are interested in using the COMB potential in a system with Ti-Si Ti-O
and Si-O interactions and on top of that using it as a part of a hybrid. I
have two questions regarding this (I'm not sure if it's here that I should
post the first one, please feel free to just ignore it and direct me to the
right person for this) :

1) We have very nearly succeeded in generating the new lines to be added to
the parameter file ffield.comb but the few parameters whose usage we
haven't been able to deduce from the 4 or 5 published papers on the subject
are : A123, Aconf and R_omega_{a,b,c,d}. On

A123 is defined as the equilibrium O-M-O or M-O-M bond angle. Does this
mean that for a triplet like, say, Si-Si-Ti, it is not defined or that it's
just a typo and I should find out the eqm angle for ANY triplet using DFT
or such? The same goes also for Aconf.

As far as R_omega_{a,b,c,d} are concerned, it would be sufficient for me to
make sure that they (like all parameters unless otherwise stated) refer to
element 1.

Once we've successfully made this addition and tested it, I'll be happy to

2) I believe that as of the 16/Aug built, there seems to be a slight
problem with the usage of the COMB potential as a part of a hybrid. To
illustrate this, when I run the HfO2 example provided in the examples/
folder, it runs fine. When I add a fictitious third species (a single
proton) that is supposed to interact with everyone else through coul/cut, I
get the following error :

ERROR: Pair coeff for hybrid has invalid style (pair_hybrid.cpp:330)

I am attaching the input, data file and the log file. We have tried several
variations of this and COMB doesn't seem to like being in a hybrid
situation. A cursory look at the suggested code snippet

  int none = 0;
   if (m == nstyles) {
     if (strcmp(arg[2],"none") == 0) none = 1;
     else error->all(FLERR,"Pair coeff for hybrid has invalid style");

and around it seems to suggest (I'm only guessing) that maybe comb is not a
defined keyword for the hybrid style? More likely of course, it's our
mistake and we have looked through the documentation to single out any
possible clashes but probably missed something. Has anyone been able to use
comb as a part of a hybrid? If yes, I would love to take a look at their
input or be grateful if they pointed out the mistake.

Many thanks in advance and apologies for the lengthy message.
Best wishes,
Hande Toffoli

in.comb.HfO2-hybrid (1.25 KB)

data.m-HfO2-hybrid (103 KB)

log.lammps (767 Bytes)

Ray can likely answer your first Q.

Re: hybrid, this line in your script
pair_coeff * * ffield.comb Hf O none
needs to be
pair_coeff * * comb ffield.comb Hf O none


Ray can likely answer your first Q.

Great thanks!

Re: hybrid, this line in your script
pair_coeff * * ffield.comb Hf O none
needs to be
pair_coeff * * comb ffield.comb Hf O none

That's an embarassing mistake! Works like a charm now. Thank you very much.


Hi Hande,

A123 and Aconf are equilibrium angles and strength coefficients for a
bond-order independent 3-body term. Equilibrium bond angles are
obtained from DFT or expt, while Aconf via fitting procedures. It was
originally designed for OMO and MOM, but you can apply it to any trio
such as Si-Si-Ti. Set both to 0.0 for most cases.

R_omega is more complicated. Mixing rules for mixing A, B, lambda and
alpha coefficients for pair interactions were later found to be
inadequate for most cases, so we added R_omega_a, b, c, d to scale the
mixed A, B, lambda and alpha coefficients. These scalers are also
obtained via fitting to DFT data. Setting these scalers to 1.0 for
Si-Ti interactions indicates using purely the mixing rules. Frankly,
you need some DFT data to fit these scalers for Si-Ti to. We did not
do this, and this is why we do not have Si-Ti entries in the
ffield.comb file.

Please let me know if you have more questions, thanks.