[Pw_forum] big differences in relativistic phonon calculations between v5.0 and v5.3

[Alaska Subedi]

Dear all,

When I calculate the phonon dispersions of PbTe using the relativistic
pseudopotentials from pslibrary, I find that v5.0 gives stable phonon
dispersions while v5.3 (and v5.2 as well) gives unstable phonons along
q = (0, x, 0). There is also a published work
http://prb.aps.org/abstract/PRB/v85/i18/e184303 with stable PbTe
phonons using relativistic pseudopotentials of pslibrary. However, I
cannot get stable phonon frequencies anymore with the new version of
the code with the sets of relativistic pseudopotentials that I tried:
lda and gga ultrasoft and normconserving.

Could anyone please check what has gone wrong in newer versions?

The input files are below.

PbTe.scf.in:

&control
    calculation = 'scf'
    restart_mode = 'from_scratch',
    tprnfor = .true.
    tstress = .true.
    prefix = 'PT-ph2',
    pseudo_dir = './pseudo/'
    outdir = './tmp/'
/

&system
    ibrav = 2
    celldm(1) = 12.056453 ! 6.38 angstroms
    nat = 2
    ntyp = 2
    ecutwfc = 50
    ecutrho = 500
    occupations = 'fixed'
    lspinorb = .true.
    noncolin = .true.
/

&electrons
    conv_thr = 1.0d-12
!    diagonalization = 'cg'
!    mixing_mode = 'local-TF' ! 'plain'
    mixing_beta = 0.2d0
/

ATOMIC_SPECIES
 Pb  207.21      Pb.rel-pz-dn-rrkjus_psl.0.2.2.UPF
 Te  127.60      Te.rel-pz-dn-rrkjus_psl.0.2.2.UPF
ATOMIC_POSITIONS crystal
Pb       0.500   0.500  0.500
Te       0.000   0.000  0.000
K_POINTS automatic
  8  8  8  0  0  0


PbTe.ph.in:

&inputph
  tr2_ph=1.0d-18,
  prefix='PT-ph2',
  alpha_mix(1)=0.2,
  nmix_ph=5,
  amass(1) = 207.21
  amass(2) = 127.60
  outdir='./tmp/',
  fildyn='PT-ph0.dyn2'
/
   0.000000000000000E+00   0.5000000000000E+00   0.000000000000000E+00