Population1_ensemble has no imaginary freuquency, but population2_ensemble has it. Why?

[LHDphy]

Dear all,

Here, I have two questions that I would like to ask everyone. I hope those who have encountered similar situations will still be kind enough to offer their advice.

In the first generation, I generated 5000 configurations (in fact, this was already quite a lot), and then I didn't perform any convergence. This was the first strange thing or problem I encountered. Because this unit cell is very small, containing only two atoms, I performed the SSCHA calculation using a 3X3X1 scale-up of the unit cell. Is there a method to accelerate convergence? The generated "population1_minimization.pdf" and the used "minimize.py" are shown in the following picture.

Another more important issue is that the phonon spectra produced by the first generation do not have imaginary frequencies, but those produced by the second generation do have imaginary frequencies. Why is that? The pictures are still included. ps: at the temperature of 0K.

[DjordjeDangic]

Hi,

I would guess all of the imaginary phonon modes are at interpolated q points. There is nothing stopping them to be imaginary. It might indicate you have a too small supercell, but only once you have your frequencies converged.

It is always better to start with a smaller number of configurations for the first few populations, around 100, depending how symmetric your structure is. Also, very rarely one converges with just one population.