Dear LAMMPS users,

Sorry to ask again this question. I didn’t get a clear answer on it. In tersoff SiC.Tersoff file for instance, for [ C Si Si ] the parameter R come from the mixing rule of pure C and Si. But, for [ Si Si C ] it also come from the mixing rule. While here the bonding is Si-Si and C is only affecting it. Shouldn’t it come from [ Si Si Si ] or I’m doing something wrong ?

Thanks,

Ali

Maybe the joke about Axel’s bad-cop routine was unclear to you, but it definitely was not meant as an answer to your question.

The answer to your question by Ray Shan was actually quite clear, even for non-experts in Tersoff potentials. For more details, you were advised to read the relevant papers: did you?

Hi Giacomo,

Yes I did it many before. The point is I can obtain SiC elastic constant using Tersoff. But, for SiN it is so odd. I used Brito motta parameters. But, Initially I have a very high stress component. therefore, when I use fix/relax to have a reasonable structure, although the high stress alleviates, the elastic constant that I obtain, especially C33, is completely false. So, it means there is something wrong with using those parameters I hypothesize. So, my main focus by now is to understand how to solve it. And, actually I got used to Axel’s joke and I knew that his answer was not a real answer :)) I know that in the manual it says that lam2 B R D lam 1 and A can be set to zero. But, I asked about the general case. May be the incorrect result that I obtain is because of it. Meanwhile, any comment would be helpful and appreciable.

Regards,

Ali

You are not getting the main point in my previous replies. In your Si-N structure do you even have Si-Si-N bonds – meaning can a Si bond to a Si and a N at the same time within such a short cutoff (not common in oxides and nitrides)? If not, why does it matter what values of interaction parameters you set? In general the AAB pair parameters are the same as those for AAA, but special adjustments can be made on a case-by-case basis as necessary so there is no hard set rule. I think I have replied this to you previously.

Have you verified and demonstrated the “brito motta” parameters reproduce a good C33 compared to experiments or DFT calculations? Have you verified the convergence of elastic constants as function of finite deformation sizes? Have you verified your minimizations during each strain in the elastic constants calculations actually converged?

Hi Ray,

I will try my best to give you a short summary of my problem in this Email. In my perfect structure I don’t have Si-Si bond. According to your Email, I should get the same answer, I mean same pressure component, in case of making Si-Si-N rows as zeros. But, look:

Si-Si-N with nonzero parameters:

-88946.335 -90527.826 1047121.7 1097.6496 1.4639344e-010 -9.7595629e-011

Si-Si-N with zeros parameters:

101790.1 -102774.09 431222.21 747.34748 8.8812022e-010 -1.9519126e-011

Brito motta, in their paper obtained a good C33 almost similar to the experiment.

Yes, I checked the convergence of my answers. As I change the deformation size, I still get the same value for C33.

Yes, I checked the convergence during minimization. For all of the components, it converges below 50 steps of minimization.

Ray, the main problem is with the initial pressure value. For example, in z direction it is almost around 100GPa initially. Obviously, as I use box/relax it goes. But, in this scenario the lattice constant in z direction changes from 2.9 to 2.99 and with the second system I’m getting incorrect C33.

Thanks,

Ali

This is still not conclusive as nobody can know if you really don’t have Si-Si bonds within your cutoff (which was not given). In any case, helping to construct a custom potential parameter set is not the purpose of the LAMMPS mailing list.