Comparsion between hiphive and phonopy

hiPhive
1

Dear Developers,

I am comparing results from hiphive and phonopy. For the most part the dispersion agrees well however acoustical phonon along a certain direction does not agree.

I tried the following:

  1. increasing number of supercells within hiphive
  2. tried sveral different cutoffs distance upto 5th order
  3. using both mc rattle and phonon rattle
  4. the supercell used is quite large (a=b=c~ 18 A)

however the agreement does not improve. Any tips on this will be helpful.

phonons
phonons2067×1610 363 KB

Many thanks,
Alex

2

Without more info here is what you can try/think about from the top of my head.

  1. Rmse test and train vs cutoff plot is always helpful.
    2.Did you try to include some short range third order interactions too?
  2. Is the material 2D or related then try rotational sum rules.
  3. What system is it? How many parameters did you get? What are the cutoffs?
  4. My guess is that in this direction the cutoffs are not sufficiently converged.
  5. Whate cell size did you use for phonopy?