[Pw_forum] Time of phonon calculation

Ruslan S. Yagufarov yagufarov at phys.tsu.ru
Fri May 15 13:24:33 CEST 2009


Hello, dear Pwscf users,

I have a question about phonon dispersion calculation in Ni2MnGa.
The problem is that it takes too much time and the iteration procedure does
not converge at all. I am tried to use various parameters, but without any
success. After doing 100 iterations,
Maybe I'm using wrong threshholds??? (my .scf and .ph inputs are attached
below). As it seems to me? according to the previous post on this forum the
everything is right.

Actually, the questions are:

1)How much time does it usually take to calculate correct phonon dispersion
for a system with only 4 atoms (like Ni2MnGa) in the case of non-parallel
execution (assuming that you are using quite good computer with 2 processors
and all parameters are good). Is it possible to have more than 50 hours
calculation???

2)Can anybody tell me what is wrong with this:

SCF calculation

 &control
calculation = 'scf',
restart_mode='from_scratch',
prefix='Ni2MnGa',
pseudo_dir = '/home2/ruslan/q/pseudo',
outdir= '/home2/ruslan/out'
tprnfor=.true.
tstress=.true.
/
&system
ibrav=  2, celldm(1) =11.011  , nat =  4, ntyp = 3,
ecutwfc = 27.0,  ecutrho =350 , degauss=0.01
smearing = 'mp', occupations = 'smearing',  nspin = 2,
starting_magnetization(2) = 1.0
starting_magnetization(1) = 0.3
starting_magnetization(3) = 0
/
&electrons
/
ATOMIC_SPECIES
Ni     58.69      Ni.pbe-nd-rrkjus.UPF
Mn     54.93        Mn.pbe-sp-van.UPF
Ga     69.72        Ga.pbe-nsp-van.UPF
ATOMIC_POSITIONS alat
Ni       0.25 0.25 0.25
Ni      -0.25 -0.25 -0.25
Mn      -0.50 -0.50 -0.50
Ga       0.00 0.00 0.00
K_POINTS automatic
8 8 8 0 0 0

PH calculation

&inputph
 prefix='Ni2MnGa',
 ldisp=.true.,
 alpha_mix(1)=0.1,
 tr2_ph=1.0d-12,
 nq1=4,
 nq2=4,
 nq3=4,
 amass(1)=58.6943,
 amass(2)=54.938049,
 amass(3)=69.72,
 outdir='/home2/ruslan/out',
 fildyn='/home2/ruslan/out/ni2mnga.dynX',
/

Thank you!
Best regards, Ruslan
Tomsk State University, Institute of Material Science SB RAS 


More information about the Pw_forum mailing list