Question About Continuing SSCHA Minimization After MAX_ITERATIONS

[masoumehmahmodi-debug]

Hello,

I have a question about continuing an SSCHA calculation after reaching the maximum number of iterations.

Suppose I run SSCHA with:
N_CONFIGS = 50
MAX_ITERATIONS = 20

After iteration 20 finishes, the calculation stops because the maximum number of iterations is reached, but from the free-energy vs iteration plots it is clear that the energy has not fully converged yet.

In this situation, is it possible (and theoretically correct) to:

Take the last structure / dynamical matrix produced at iteration 20,

Use it as the new starting point (input structure and force constants), and

Rerun SSCHA to effectively “continue” the minimization from where it stopped,

instead of starting again from the original harmonic calculation?

Does your answer change between the purely harmonic starting point and the anharmonic (already partially optimized) case?

Thank you very much for your help and for developing SSCHA.

[mesonepigreco]

Of course, the correct thing to do is to restart from the last optimized dynamical matrix. Indeed, the result should not depend on the original point unless there is a phase transition involved, for which the starting point should be the one of the phase you want to study.