[Pw_forum] regression in phonon with SOC between v5.0.1 and v5.0.2

[Alaska Subedi]

Dear all,

There seems to be a regression in phonon calculations with SOC between
versions 5.0.1 and 5.0.2.

I tested Pb (input files are below). While I seem to get results
similar to dal Corso's J. Phys.: Condens. Matter 20, 445202 (2008)
with version 5.0.1, the results from version 5.0.2 significantly
differ. In particular, I get unstable modes away from Gamma.

>From version 5.0.1:

     q = (   -0.250000000   0.250000000  -0.250000000 )

 **************************************************************************
     omega( 1) =       0.686784 [THz] =      22.908647 [cm-1]
     omega( 2) =       0.686784 [THz] =      22.908647 [cm-1]
     omega( 3) =       1.839691 [THz] =      61.365485 [cm-1]
 **************************************************************************

>From version 5.0.2:

     q = (   -0.250000000   0.250000000  -0.250000000 )

 **************************************************************************
     omega( 1) =      -0.312682 [THz] =     -10.429943 [cm-1]
     omega( 2) =      -0.312682 [THz] =     -10.429943 [cm-1]
     omega( 3) =       1.434735 [THz] =      47.857599 [cm-1]
 **************************************************************************

The input files are same for both versions.

Pb.scf.in:
&control
    calculation='scf'
    restart_mode='from_scratch',
    tprnfor = .true.
    tstress = .true.
    prefix='Pb-ph0',
    pseudo_dir = './pseudo/',
    outdir='./tmp/'
/

&system
    ibrav = 2
    celldm(1) =  9.23
    nat = 1
    ntyp = 1
    ecutwfc = 40
    ecutrho = 400
    occupations = 'smearing'
    degauss = 0.010
    lspinorb = .true.
    noncolin = .true.
/

&electrons
    conv_thr = 1.0d-12
    diagonalization = 'cg'
!    mixing_mode = 'local-TF'
    mixing_beta = 0.3d0
/

ATOMIC_SPECIES
 Pb  0.0     Pb.rel-pz-dn-rrkjus.UPF
ATOMIC_POSITIONS crystal
Pb       0.000   0.000  0.000
K_POINTS  automatic
  8  8  8  0  0  0

Pb.ph.in:
phonons of Pb
&inputph
  tr2_ph = 1.0d-18
  prefix = 'Pb-ph0'
  alpha_mix(1) = 0.2
  nmix_ph = 5,
  amass(1) = 207.21
  outdir = './tmp/'
  fildyn = 'Pb-ph0.dyn'
  ldisp = .true.
  nq1 = 8, nq2 = 8, nq3 = 8
/

Thanks,
Alaska
CPHT, Ecole Polytechnique, Paris