[Pw_forum] Problem with scf convergence

Tue Mar 20 18:36:40 CET 2012

Dear Jonathan,

I am not sure if you tried, but the local-density-dependent TF screening is very important when the cell is large, like in your case. I did not find the PPs you are using in the quantum espresso webpage, so I downloaded the usual USPP and repeated the calculation. Here it is the input and the output that I obtained. It took a significant number of scf cycles to converge (more than 100, I had to restart the calculation), so probably a better optimization of the smearing/charge cutoff/etc... parameters can still be done. But considering it is Co-MgO we are talking about, I would say that it's ok. I hope it helps!

All best!
Fabio Negreiros Ribeiro,
Post-Doc, IPCF - Consiglio Nazionalle delle Ricerche, Pisa, Italy

-----Original Message-----
From: Jonathan Trinastic <jptrinastic at phys.ufl.edu>
To: pw_forum <pw_forum at pwscf.org>
Sent: Tue, Mar 20, 2012 4:15 pm
Subject: [Pw_forum] Problem with scf convergence

Hi,

I'm having an issue getting a Co-MgO interface to converge properly.  
Basically, the setup includes 8 layers of bcc Co  interfaced with 8 
layers of MgO, periodic in x and y and a vacuum between interfaces in 
the z direction.  I'm trying to find the Co-O distance that gives the 
energy minimum, so I'm trying to relax structures with varying distances 
- I'm using a 2.17 A distance to start based on experimental values.  
I've been able to relax the Co and MgO parts separately without a 
problem, but when I combine them they won't converge.

I've included my pw.x input below, however I have also tried adjust many 
of the parameters based on feedback in other posts and the QE User 
guide, such as: increasing degauss or adding bands to account for 
metallic nature of Co, decreasing mixing beta, increasing the energy 
cutoff value, however all these end up having the same convergence 
problems where the estimated scf accuracy hovers around 1-10.  Does 
anyone see potential issues with the input I'm using that I'm missing?  
I've chosen my energy and charge density cutoffs based on the 
recommended choices for the oxygen PAW potential, which is higher than 
for Co or Mg.

Thanks,
Jonathan

&control
     calculation='scf',
     restart_mode='from_scratch',
     pseudo_dir = '/scratch/hpc/jptrinas/espresso-4.3.2/pseudo/',
     outdir='./tmp/'
     prefix='bcc.co.mgo'
     tstress = .true.
     tprnfor = .true.
  /
&system
     ibrav = 6, celldm(1) =5.32, celldm(3) = 19.52, nat= 24, ntyp= 3
     starting_magnetization(1)=0.7, nspin=2
     ecutwfc = 38.0, ecutrho = 151.0
     occupations='smearing', smearing='gaussian', degauss=0.02
  /
&electrons
     diagonalization='cg'
     conv_thr = 1.0e-6
     mixing_beta = 0.3
  /
ATOMIC_SPECIES
  Co 58.93 Co.pbe-p-paw.UPF
  Mg 24.31 Mg.pbe-nsp-bpaw.UPF
  O  15.99 O.pbe-kjpaw.UPF
ATOMIC_POSITIONS (alat)
  Co 0.0 0.0 5.000
  Co 0.5 0.5 5.500
  Co 0.0 0.0 6.000
  Co 0.5 0.5 6.500
  Co 0.0 0.0 7.000
  Co 0.5 0.5 7.500
  Co 0.0 0.0 8.000
  Co 0.5 0.5 8.500
  O  0.5 0.5 9.270
  Mg 0.0 0.0 9.270
  O  0.0 0.0 10.02
  Mg 0.5 0.5 10.02
  O  0.5 0.5 10.77
  Mg 0.0 0.0 10.77
  O  0.0 0.0 11.52
  Mg 0.5 0.5 11.52
  O  0.5 0.5 12.27
  Mg 0.0 0.0 12.27
  O  0.0 0.0 13.02
  Mg 0.5 0.5 13.02
  O  0.5 0.5 13.77
  Mg 0.0 0.0 13.77
  O  0.0 0.0 14.52
  Mg 0.5 0.5 14.52
K_POINTS (automatic)
  6 6 1 0 0 0

Typical scf accuracy estimates by end of run:
  estimated scf accuracy <      16.36480484 Ry
      estimated scf accuracy <       8.36686889 Ry
      estimated scf accuracy <       2.33384573 Ry
      estimated scf accuracy <       6.92471882 Ry
      estimated scf accuracy <      13.18681083 Ry
      estimated scf accuracy <      11.95347196 Ry
      estimated scf accuracy <      11.49117957 Ry
      estimated scf accuracy <      12.65621865 Ry
      estimated scf accuracy <       8.53355241 Ry
      estimated scf accuracy <       6.20738100 Ry
      estimated scf accuracy <       7.91117866 Ry
      estimated scf accuracy <      17.89939870 Ry
      estimated scf accuracy <       4.29389382 Ry
      estimated scf accuracy <       8.11074063 Ry
      estimated scf accuracy <       3.91904239 Ry
      estimated scf accuracy <       2.99042194 Ry
      estimated scf accuracy <       2.47072016 Ry
      estimated scf accuracy <       5.16323880 Ry
      estimated scf accuracy <       4.17983416 Ry
      estimated scf accuracy <      16.77910831 Ry
      estimated scf accuracy <       8.52666365 Ry
      estimated scf accuracy <       4.02303973 Ry
      estimated scf accuracy <      11.00391780 Ry
      estimated scf accuracy <       4.89255713 Ry

-- 
Jonathan Trinastic, M.S.
Graduate Student
Department of Physics
University of Florida

_______________________________________________
Pw_forum mailing list
Pw_forum at pwscf.org
http://www.democritos.it/mailman/listinfo/pw_forum

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20120320/7503c053/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: input.in
Type: application/octet-stream
Size: 1242 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20120320/7503c053/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: input.out
Type: application/octet-stream
Size: 98325 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20120320/7503c053/attachment-0001.obj>