[Pw_forum] Negative frequencies in phonon calculation

[Alberto Milani]

Dear Jose,
Thank for your help.
I know well the problems about saddle points and negative frequencies because 
I usually calculate normal modes of molecules where this problem occurs many 
times.
Here the problem is that I am studying an infinite linear chain and so I found 
strange that some kind of saddle point could be the source of the problem for 
such a system. Many other ab-initio calculation on polyyne finite chain have 
shown that these systems are indeed stable as linear chains (no negative 
frequencies are found). So I would exclude any kind of transversal 
deformartion from linearity.
From this point of view, there is no possibility for any saddle point and the 
relaxed structure that I have obtained should be considered converged in a 
well defined minimum. In fact, no negative frequencies are found for the 
longitudinal phonons.
So I think that the problem here is not the usual saddle instability. I wonder 
if the code fails in some other way in stabilizing the linear chain (e.g. 
choice of the functional, pseudopotential, ...).
Thank you again for your attention.
Yours,
A. M.