[Pw_forum] Difference in implementation in matdyn.x between versions 4.2.1 and 5.0.0?

[Matthijs Mentink]

Dear all,

I am trying to calculate the phonon DOS of Nb3Sn, using matdyn.x. For this
I do a self-consistent calculation with 666 automatic k-points and then a
phonon calculation with 666 automatic q-points. After doing q2r (see
below), I calculate the phonon dos with matdyn.x (also see below).

-- q2r.in
&input
  fildyn='Nb3Sn.dyn', flfrc='Nb3Sn.fc',
 /

-- matdyn.x
 &input
  asr='simple', amass(1)=92.90638, amass(2)=118.710,
  flfrc='Nb3Sn.fc', dos=.true.,
  fldos='Nb3Sn_300K_pz_free.phdos',
  flfrq='Nb3Sn_300K_pz_free.freq'
  nk1=24, nk2=24, nk3=24
 /

Basically, I extrapolate from 666 automatic q-points to 24-24-24 automatic
q-points and then calculate the phonon dos.

This seems to work just fine in espresso-4.2.1, with a phonon dos very
close to what I expect, but the result is dramatically different in
espresso 5.0.0. Furthermore, when I use espresso 5.0.0 to calculate
everything and then use matdyn.x from version 4.2.1, I get a very spiky
unfiltered version with a result close to what I expect (and want), whereas
in version 5.0.0 of matdyn.x I get a phonon dos with very broad peaks, with
considerable softening at low frequencies, which is something that I do not
expect to see. It should also be noted that I am using 1.90e-3 Ry of
broadening.

Basically, it seems that the implementation of matdyn.x has changed from
version 4.2.1 (which incidentally gave the same result as 4.3.2) to version
5.0.0. Anybody have any idea what I am doing wrong here?

thanks in advance,

Matthijs
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20120806/be051c98/attachment.html>