[Pw_forum] q2r and electron-phonon interaction

[Paolo Giannozzi]

On Nov 29, 2010, at 22:59 , sama at ns.dcci.unipi.it wrote:

> In line 358 of code q2r.f90 is read the variable *nstar*, which for
> count_q=2 takes the value nstar=3. In the next subroutine read_gamma
> the code expects then a series of information dimensioned 3, while
> actually in file a2Fq2r.51 the series is dimensioned 6.

this is actually a bug, that happens whenever -q is not in the star  
of q.
In this case, there is still a symmetry operation (time reversal)  
sending
q into -q, but it is not a point group symmetry. Or at least, this is  
what
I vaguely remember. The code writes the number nq of q's in the star,
then it writes the results for all q's, plus those for all -q's.  
Workaround:
in file PH/elphon.f90, locate line
      write(iuelph,*) nq
replace with
      IF ( imq == 0 ) THEN
         write(iuelph,*) 2*nq
      ELSE
         write(iuelph,*) nq
      ENDIF
No warranty! Thank you for the detailed report.

By the way:
1) in principle you should use pseuopotentials generated for the same
exchange-correlation potential
2) If you use norm-conserving pseudopopotentials, you should not set a
value of ecutrho (it will be set to 4*ecutwfc)
3) "la2F=.true." should be set only in the first scf run, the one with
very dense k-point grid

Paolo
---
Paolo Giannozzi, Dept of Chemistry&Physics, Univ. Udine
via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222