Hi Axel,

Thanks for your reply. In my experience, after minimization the pressure is very small, i.e. 10^(-8) or even smaller. Compared with this, I just said the pressure was high in my buck/coul calculation.

I just want to make sure that the structure after minimization is “correct”. So in this case if I don’t misunderstand, what you mean is that I have found the optimized structure since the minimization is converged. Would you please give some comments on that? Many thanks again!

Best,

Siming

Hi Axel,

Thanks for your reply. In my experience, after minimization the pressure is very small, i.e. 10^(-8) or even smaller. Compared with this, I just said the pressure was high in my buck/coul calculation.

this can be due to chance or due to having a system that is simple to minimize (like an FCC metal lattice) and has a smooth potential. for complex force fields and structures, it is not always as straightforward for the minimizer and if the potential hypersurface is noisy/rugged, then the minimizer may stop when it is close to the minimum, but not at the theoretically exact minimum.

I just want to make sure that the structure after minimization is “correct”. So in this case if I don’t misunderstand, what you mean is that I have found the optimized structure since the minimization is converged. Would you please give some comments on that? Many thanks again!

the minimizer will tell you what critera were met when it stops. look at the output. whether a final structure is the actual global minimum is impossible to state, especially from remote. you are looking at a high-dimensional optimization problem, where you are making some assumptions how you can reach the minimum quickly. but it is easily possible, that you get “trapped” in a local minimum. there are different possible reasons for that. also, there are different methods to test whether a minimum is a local minimum vs. a global minimum. it helps to start with a geometry that is close to the minimum (i.e. not from completely random positions). one possible test of being in a significant minimum would be to do some (small) random displacements of atoms and minimize again. when the resulting geometry is (numerically) sufficiently similar, then you are likely in a converged minimum. there are other possible tests like computing phonon spectra. but explaining all this here is going too far. there is plenty of textbook literature and review papers on minimization of geometries for molecular or atomic systems. that is where you should look for further guidance.

axel.