[Pw_forum] Trouble with the Fe PBE PAW functional.

[Joshua Davis]

Dear QE users

*Summary:* I am attempting a wave-function energy convergence on my Iron
containing compound, but I am have confusing errors that include lack of
electronic convergence, and the statement "g-vectors missing."  The
suggested density cut-off is really high for Iron.  I think this may have
some thing to do with the errors.

*General:*
I am using QE 5.1 on Ubuntu 12.04 LTS The pseudopotentials I am using are:

Li.pbe-s-kjpaw_psl.0.2.1.UPF
Fe.pbe-spn-kjpaw_psl.0.2.1.UPF
O.pbe-n-kjpaw_psl.0.1.UPF

The  general script I use to make my input files and run them is:


#!/bin/bash
##################################################
# This is a e-cut convergence  for a wave function
# of a cell of LiFeO2. $ECUT is in Ry.
##################################################

for ECUT in 50 70 80 90 100 110 120 130 140 150 160
do
rm -rf ./scratch/*
cat > LiFeO2_ec$ECUT.in << EOF
 &CONTROL
   title = 'LiFeO2ECUTtest',
   calculation = 'vc-relax',
   pseudo_dir = '../pot',
   outdir = './scratch',
   prefix = 'lifeo2ecut',
   etot_conv_thr = 1.0D-5,
   forc_conv_thr = 1.0D-4
 /

 &SYSTEM
   ibrav = 8,
   a = 5.51600,
   b = 6.41390,
   c = 5.07890,
   nat = 16,
   ntyp = 3,
   starting_magnetization(1) = 0,
   starting_magnetization(2) = 6,
   starting_magnetization(3) = 0,
   ecutwfc = $ECUT,
   nspin = 2,
   occupations = 'smearing',
   smearing = 'gaussian',
   degauss = 2.0D-2
 /

 &ELECTRONS
   mixing_beta = 0.7,
   electron_maxstep = 200,
   conv_thr = 1.D-8
 /
 &IONS
   ion_dynamics = 'bfgs'
 /

 &CELL
   cell_dynamics = 'bfgs',
   cell_dofree = 'xyz'
 /

ATOMIC_SPECIES
 Li 6.94 Li.pbe-s-kjpaw_psl.0.2.1.UPF
 Fe 55.845 Fe.pbe-spn-kjpaw_psl.0.2.1.UPF
 O 15.999 O.pbe-n-kjpaw_psl.0.1.UPF

ATOMIC_POSITIONS (crystal)
...
...
..


K_POINTS (gamma)

EOF

echo $ECUT running...

mpirun -np 4 pw.x -nt 4 -nd 4 <LiFeO2_ec$ECUT.in> LiFeO2_ec$ECUT.out

echo $ECUT finished


I am essentially just running my LiFeO2  variable cell relaxtion at 50, 70,
80, 90, 100, 110, 120, 130, 140, 150,  and 160 Ry.

My problems are arising from some of the individual calculations not
completing properly.

Wave-function cut-offs of 50 Ry and 70 Ry do not converge electronically
(the calculation reaches the max electron step w/o converging) which makes
sense because they are small basis sets comparatively.
The calculation for a cut-off 80 Ry does converge. However, for a cut-off
90, the relaxtion converges, but  for the final single point calculation
done at the end with the new cell parameters, the calculation stops with
the error:
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine ggen (2):
     g-vectors missing !
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


the cut-offs of 100, 110, 120, and 130 all work fine, but the the
calculation with a wave function cut-off of 140 Ry again gives me the same
"g-vectors missing" error.  Then for a cut-off of 150 Ry, the calculation
will again not converge electronically, but the calculation will converge
with a wave-function cut-off of 160 Ry.


The suggested wave-function and density cut-off for Lithium  are 78 and 355
Ry respectively, and for Oxygen it is 47 and 187 Ry.  While for Iron it is
64 and 782 Ry(???)
So it makes sense as why a wave-function cut-off of 50 and 70 Ry would not
converge, but I am not entirely sure why there are problems at 90, 140, and
150 Ry. I have not changed the default density cut-off yet, so the density
cut-off for eat calculation is  just 4x the wave-function cut-off at this
point.

I haven't done much with the density cut-off, but I think that might be
where some of the issues are coming from.  The suggested cut-offs for
Lithium and Oxygen seem normal, but the density cut-off for Iron seems
really really high.  It is about 13x higher than the wave function
cut-off.  Is it supposed to be this high, or is it a typo.  I have tried
doing a calculation with this high of a density cut-off, and It either
takes too long or gives me a "too many g-vectors" error.  There just seems
like there is something off about the Iron PAW potential that I am not
getting.

Any help would be appreciated.

I have atompaw compiled so I'm not gonna shy away from making a new
potential, but I am not well versed in making psuedopotentials and I will
probably be more inclined to make one from the data sets already on the
atompaw website.

*Summary Again:* I am attempting a wave-function energy convergence on my
Iron containing compound, but I am have confusing errors that include lack
of electronic convergence, and the statement "g-vectors missing."  The
suggested density cut-off is really high for Iron.  I think this may have
some thing to do with the errors.

Again any help would be much appreciated

----------------------------------------------------------------------------------------------------------------
Joshua D. Davis

Graduate Assistant
Department of Chemistry
Michigan State University

578 S. Shaw Lane, room 432
East Lansing, MI 48824
-----------------------------------------------------------------------------------------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20150327/9a45076b/attachment.html>