ABINIT Discussion Forums

I have heard, but am not sure, that when doing a GAMMA phonon calculation, sometimes seeing small negative phonons w/ frequencies no less than -20 cm-1 are really not "there"; that they really are an artifact of the calculation. Is this true? And if true, what is a good cut off frequency to use as being "true" or being "false"? thanks, jb

Processing with anaddb and application of the acoustic sum rule will take care of these artifacts, and also give you an indication of the extent to which the sum rule was violated.

Thanks, Professor.
I reviewed the ABinit input vars and, using the same script as before, included rfasr 2
and it worked like a charm eliminating the negative phonons and a small positive one as well.

I've met this problem before,

first of all, the structural must be tested for convergence, i.e. the structural must be fully relaxed.

The Phonon is low-energy excitation, should have higher energy than ground state. However, if the structure is not relaxed fully, with residual stress, then the "Ground State"
is not a ground state any more, and atomic displacement will lead to a lower energy and more "relaxed" structure, which give negative phonon energy.

Only after that, the sum rule will be meaningful.