[QE-users] Eigenvalues for vc relax not converging

[Ari P Seitsonen]

Dear Priya,

   Some quick comments:

  - Are you sure that the ecutwfc = 20 Ry is sufficient? The 40 Ry that you 
mention sounds much more realistic, if not even higher value would be 
needed; the calculation of the stress tensor ("vc-relax") requires a 
stricter convergence with respect to the cut-off energy/basis set than for 
example total energy as the former is a (second) derivative of the latter, 
and the basis set would in principle be a function of the lattice 
constant, so better be well converged in the basis set

  - Since you have an FCC lattice you could easily calculate the total 
energy of the ("relax"'ed) system as a function of the lattice parametre 
and check if you converge toward the same value as the 'vc-relax'

  - Do I understand you correctly that you can reach convergence in the 
self-consistent loop, but in the subsequent loops the electronic structure 
does not convergence (as the D3 does not affect the electronic structure 
explicitly, only via the modified lattice constant)

  - Your value for smearing is quite small; you could at least temporarily 
try to increase it

  - What is the lowest value of convergence that you reach in, say 100 
iterations?

     Good Luck, Greetings from Helsinki,

        apsi

-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
   Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/
     Ecole Normale Supérieure (ENS), Département de Chimie, Paris
     Mobile (F) : +33 789 37 24 25    (CH) : +41 79 71 90 935


On Tue, 8 May 2018, Priya Shrivastava wrote:

> 
> Dear All,
> 
> I am trying to reproduce published data but wihtout Vdw i got wrong results and then after including DFT-D3 , my
> calculations are not converging for vc-relax.Please see my i/p  file. I have just started using QE so i am not
> sure whats wrong however i have read threads containing similar problems. I tried reducing beta to 0.2 and change
> nbnd and diagonalization but to no avail . I have also used the suggested ecutwfc=40 in the paper  but my
> calculations didnt converge even after 500 iterations. Please help me.
> &CONTROL
>  calculation='vc-relax',
>  restart_mode='from_scratch',
>  prefix='Cs',
>  pseudo_dir="/home/external/iitb/priyas/PSEUDOPOTENTIALS_PBE",
>  tstress = .true.
>  tprnfor = .true.
>  verbosity= 'high',
>  forc_conv_thr=1.0d-3,
>  etot_conv_thr=1.0d-5,
> /
> &SYSTEM
>  a           = 10.6212,
>  ibrav       = 2,
>  nat         = 3,
>  ntyp        = 3,
>  ecutwfc= 20.0,
>  ecutrho=400.0,
>  vdw_corr='DFT-D3',
>  nbnd=35,
>  occupations='smearing',
>  degauss=0.001, 
> /
> &ELECTRONS
>  mixing_beta=0.2,
>  diagonalization='cg',
>  conv_thr=1d-8,
>  electron_maxstep=500,
> /
> &IONS
>  ion_dynamics='bfgs',
> /
> &CELL
>  cell_dynamics='bfgs'
>  press=0.0,
>  press_conv_thr=0.5,
> /
> ATOMIC_SPECIES
> Cs 132.905 Cs.pbe-spn-kjpaw_psl.1.0.0.UPF
> Ti 47.867  Tl.pbe-dn-kjpaw_psl.1.0.0.UPF
> Br 79.904  Br.pbe-n-kjpaw_psl.1.0.0.UPF
> ATOMIC_POSITIONS crystal
> Cs      0.250000   0.250000   0.250000
> Ti      0.000000   0.000000   0.000000
> Br      0.234000   0.000000   0.000000
> K_POINTS automatic
> 4 4 4 1 1 1
> ,
> Regards,
> Priya,IITB
> 
> 
>