[Pw_forum] Na phonon issues

[Nicola Marzari]

Miguel Martínez Canales wrote:

> What would one do in case long range effects are not negligible? For 
> example, fcc Li at 30 GPa has a soft phonon in the \Gamma-K branch. 
> Reproducing it via the IFC matrix seems to require a not so coarse q 
> point grid. What would one do? Calculate the elastic constants from 
> classic lattice distortions? Introduce phonons from the soft region to 
> include in the q2r.x list? Forget about getting accurate or even 
> physical results with matdyn?


Not exactly sure - but a P=0 calculation as discussed before
would give you the right phonon dispersions (with the acoustic
sum rules) say on the Gamma-K branch, and a phonon calculation
explicitly done on the Gamma-K line would give you a sense of the
numerical error at Gamma and outside Gamma.

So, you might discover that the error at Gamma for the lowest
frequency is 10 cm^-1, 10% outside gamma is 3 cm^-1, and 20%
outside gamma is 1 cm^-1.

This gives you a ballpark for what to do at P=30 GPa - i.e. what are
the q where the error is really small. Note that in this I'm making the 
implicit suggestion that the "errors" are taking place mostly at gamma 
due to the long-wavelength nature of the modes, and would not be there
at a finite q, even if the phonon frequencies were getting close to zero.

Also, you could use extrapolations from calculations at higher
pressures (where the phonon frequency would have gone strongly
imaginary, at a given q), to make sure it's all consistent.

Still, it would be great to understand what is the origin
ofthis discrepancy, and why it is not cured by increasing the
cutoffs and k-point sampling - rather converging smoothly to a non-zero
value.

			nicola



---------------------------------------------------------------------
Prof Nicola Marzari   Department of Materials Science and Engineering
13-5066   MIT   77 Massachusetts Avenue   Cambridge MA 02139-4307 USA
tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu