[Pw_forum] wrong representation error in PH

[Samuel Poncé]

Dear PW community,

I have the following error

     Error in routine set_irr_sym_new (511):
     wrong representation

That I cannot get rid off while trying to do a phonon calculation.

My PH input
--
&inputph
  recover  = .true.,
  prefix   = 'x',
  fildyn   = 'x.dyn',
  ldisp    = .true.,
  epsil    = .true.,
  trans    = .true.,
  fildvscf = 'dvscf',
  niter_ph = 800,
  nq1=8,
  nq2=8,
  nq3=2,
  tr2_ph   =  1.0d-14,
  max_seconds = 86000,
 /

Prior to that I made a vc-relax (cell and ions) calculation with
    tprnfor         = .true.,
    tstress         = .true.,
    forc_conv_thr   = 1d-5, ! Ry/Bohr
which converged (with the correct sym). I then used these lattice paramters
for the ph calculation (through a first scf run).

The q-point that crashes is the following one:
     Calculation of q =    0.0000000  -0.5773503  -0.1312539

The q-points before that one are fine.

Things I've checked so far:

1) change the q-grid to 8x8x3. The problematic q-point is then
q =    0.0000000   0.1443376   0.0875026
In that case the problem is not the same (no error) but the calculation
never converged
      iter # 361 total cpu time : 81102.7 secs   av.it.:  16.5
      thresh= 1.419E-03 alpha_mix =  0.700 |ddv_scf|^2 =  3.809E-04

      iter # 362 total cpu time : 81178.6 secs   av.it.:  13.8
      thresh= 1.952E-03 alpha_mix =  0.700 |ddv_scf|^2 =  9.521E-05
....


2) I've increased ecutwfc from 90 to 120 in the scf run
same error

3) I've increase kgrid from 10x10x3 to 16x16x4
same error

4) I went into /PHonon/PH/set_irr_sym.f90
And changed
              if (jpert.ne.ipert .and. abs(wrk)> 1.d-6 ) &
                     call errore('set_irr_sym_new','wrong
representation',100*irr+10*jpert+ipert)
              if (jpert.eq.ipert .and. abs(wrk-1.d0)> 1.d-6 ) &
                     call errore('set_irr_sym_new','wrong
representation',100*irr+10*jpert+ipert)
into
              if (jpert.ne.ipert .and. abs(wrk)> 1.d-3 ) &
                     call errore('set_irr_sym_new','wrong
representation',100*irr+10*jpert+ipert)
              if (jpert.eq.ipert .and. abs(wrk-1.d0)> 1.d-3 ) &
                     call errore('set_irr_sym_new','wrong
representation',100*irr+10*jpert+ipert)
Same error


What else can I test?

PS: I'm using the latest trunk of QE.

Cheers,

Samuel Ponce
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20150928/1ecae736/attachment.html>