[Pw_forum] Problems with phonon calculation in one-dimensional chains

[Jin Zhang]

Dear all,

I have encountered several issues related to Gamma point phonon calculation
in simple 1d chains(including aluminum, gold and platinum). The symptom is
*sometimes* there are one or more eigen-*optical*-modes having
negative(usually large) frequencies. The system I considered are all
2-atom/cell 1d chains(to produce optical branches), with large vacuum region
around them(usually 16-20Bohrs). For example, the aluminum input are listed
below,
---------------------------------------------------------
pw.x----------------------------------------------------------
Infinite Al Chain NCPP
   &CONTROL
      calculation = 'scf',
      prefix='al',
      nstep=100,
      tstress = .TRUE.,
      tprnfor = .TRUE.,
      outdir='./tmp/',
      pseudo_dir = '/home/i/opt/espresso- 3.2/pseudo',
      wf_collect=.true.,
   /
   &SYSTEM
      ibrav=8,
      celldm(1)=20.0,
      celldm(2)=1.0,
      celldm(3)=0.9,
      nat=2,
      ntyp=1,
      ecutwfc=30.0,
      occupations='smearing',
      smearing='cold',
      degauss=0.05,
      !nbnd=40,
      la2F=.true.,
   /
   &ELECTRONS
      electron_maxstep = 100,
      mixing_beta = 0.5,
      conv_thr =  1.0d-8,
   /
   &IONS
   /
ATOMIC_SPECIES
Al  26.98154  Al.pz-vbc.UPF
ATOMIC_POSITIONS {crystal}
Al    0.000    0.0   0.0
Al    0.000    0.0   0.5
K_POINTS (automatic)
  1 1 40  0 0 1
------------------------------------------------------------
ph.x---------------------------------------------------------------
 &inputph
  tr2_ph=1.0d-14,
  recover = .false.
  prefix='al',
  fildvscf='dvscf',
  alpha_mix=0.5,
  amass(1)=26.98154,
  outdir='./tmp',
  fildyn='dyn',
  elph=.false.,
  trans=.true.,
  ldisp=.true.,
  nq1 =1, nq2 =1, nq3 = 1,
/
-------------------------------------------------------------------------------------------------------------------------------

The obtained frequency is

-------------------------------------------------------------------------------------------------------------------------------


     omega( 1) =      -3.982548 [THz] =    -132.844370 [cm-1]
 (  0.000000  0.000000  0.000000  0.000000  0.711951  0.000000 )
 (  0.000000  0.000000  0.000000  0.000000 -0.702229  0.000000 )
     omega( 2) =       0.377569 [THz] =      12.594436 [cm-1]
 ( -0.762581  0.000000  0.000000  0.000000  0.000000  0.000000 )
 ( -0.646893  0.000000  0.000000  0.000000  0.000000  0.000000 )
     omega( 3) =       0.377569 [THz] =      12.594436 [cm-1]
 (  0.000000  0.000000 -0.762581  0.000000  0.000000  0.000000 )
 (  0.000000  0.000000 -0.646893  0.000000  0.000000  0.000000 )
     omega( 4) =       0.525357 [THz] =      17.524138 [cm-1]
 (  0.000000  0.000000  0.000000  0.000000 -0.702229  0.000000 )
 (  0.000000  0.000000  0.000000  0.000000 -0.711951  0.000000 )
     omega( 5) =       1.396045 [THz] =      46.567374 [cm-1]
 (  0.000000  0.000000 -0.646893  0.000000  0.000000  0.000000 )
 (  0.000000  0.000000  0.762581  0.000000  0.000000  0.000000 )
     omega( 6) =       1.396045 [THz] =      46.567374 [cm-1]
 ( -0.646893  0.000000  0.000000  0.000000   0.000000  0.000000 )
 (  0.762581  0.000000  0.000000  0.000000  0.000000  0.000000 )
-----------------------------------------------------------------------------------------------------------------------------

As you can see, the LO mode is largely negative while other modes(including
acoustic ones) seem fine. I have made convergence test to k-points,  ecut
and degauss. In other cases like the gold and platinum chains, the negative
modes are usually TO ones. So my question is: Why does this happen? Is there
any physics I missed in the above calculation, or just some naive settings
make the result weird?

Looking forward to your help. Thanks in advance!

Best wishes,
Jin Zhang
-- 
MPhil
Dept. of Physics, Peking University
Beijing, 100871, P.R.China
Phone: 86-10-62768590
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20070527/e216d2a3/attachment.html>