[Pw_forum] negative values in the electron-phonon calculation.

[Paolo Giannozzi]

On Thursday 09 September 2004 16:26, Stefano Baroni wrote:

> As a general remark, negative frequencies (better, "imaginary"
> frequencies: i.e. negative squared frequencies), signal the instability
> of a crystal structure towards a deformation along the phonon
> distortion pattern. In principles, there is nothing wrong in this.

actually the original reference to "negative frequencies" :

> > This is one of the output file with negative frequences:

was misleading: it is the phonon linewidth "gamma" (as defined 
and calculated in PH/elphon.f90) that seems to be negative for
some values of the parameters (the electron-phonon coupling
coefficient "lambda" = gamma/(pi N(Ef) omega^2  is negative 
as a consequence if gamma is negative).

I think that it might be simply a problem of bad convergence
with respect to k-points and gaussian broadenings. Electron 
phonon coupling coefficients - in particular, the double sum
over delta functions - converge very slowly with respect to 
these parameters. This is why that particular sum is done on 
a very dense grid of k-points, using several values of the 
gaussian broadenings: one has to find a region of parameters
in which the results do not change appreciably. It seems to me
that the results are still far from convergence with the parameters
used.

I am a little bit puzzled because I was expecting that the calculated
gamma is a positive definite quantity, even far from convergence,
but maybe this is not true. 

Paolo

-- 
Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
Piazza dei Cavalieri 7      I-56126 Pisa, Italy