[Pw_forum] Negative frequencies in phonon calculation

[Jose C. Conesa]

Hello,
  Assuming that there are no problems of numerical inaccuracy or inadequate convergence 
thresholds, negative frequencies at gamma point in a geometry-relaxed system usually mean that 
(within the system and the calculation method used) there is some distortion (decreasing the 
symmetry of the system) which leads to lower energy, i.e. you are at a saddle point in the energy 
hypersurface. You must move some atom out of the initial symmetry and optimize geometry 
again starting from that modified situation. If you can obtain the vector coordinates for the normal 
mode of vibration corresponding to the negative frequency, a displacement of the involved atoms, 
proportional to that normal mode, by e.g. a maximum of 0.2 angstrom will give to you a starting 
displaced geometry closer to the true energy minimum. Sometimes you need to do this several 
times if not all of the negative frequencies disappear after the first attempt.
If, after obtaining a new geometry with no negative frequencies at the gamma point, you still have 
negative frequencies at other BZ points, this means that your system is unstable against some 
atomic displacements that multiply some of the primitive cell vectors by e.g. 2 or 3, leading to a 
supercell. Curing this situation may be more tedious.
Good luck,
Jose C. Conesa
Instituto de Catalisis y Petroleoquimica, CSIC
Campus de Cantoblanco
28049 Madrid  -  Spain
Phone Nr. 34-91-5854766
Fax Nr. 34-91-5854760