[QE-users] absolute magnetization and negative rho going to very large values

wigbert polet wigbert.polet at gmail.com
Mon Jul 22 15:27:17 CEST 2019


Dear QuantumEspresso user group,

i have a problem with my calculations. I investigate tungsten carbide
(WC) layer with 14 A vacuum and adsorb atomic oxygen on the surface.

I use PWSCF v.6.3MaX

If try a relax Job with a starting magnetization for the oxygen, then
the following happens while the SCF-Cycles:

The first optimization step converge in a range of 17 iterations.

Step 1:

     negative rho (up, down):  1.163E-01 7.746E-02
     total energy              =  -18470.12492847 Ry
     Harris-Foulkes estimate   =  -18470.12491708 Ry
     estimated scf accuracy    <       0.00000263 Ry
     magnetization             =    -1.69 Bohr mag/cell
     absolute magnetization    =     2.25 Bohr mag/cell
--
     negative rho (up, down):  1.168E-01 7.723E-02
     total energy              =  -18470.12493846 Ry
     Harris-Foulkes estimate   =  -18470.12493246 Ry
     estimated scf accuracy    <       0.00000146 Ry
     total magnetization       =    -1.72 Bohr mag/cell
     absolute magnetization    =     2.30 Bohr mag/cell
--
     negative rho (up, down):  1.176E-01 7.701E-02
     total cpu time spent up to now is     3103.3 secs
     total energy              =  -18470.12494014 Ry
     Harris-Foulkes estimate   =  -18470.12494049 Ry
     estimated scf accuracy    <       0.00000116 Ry
     total magnetization       =    -1.75 Bohr mag/cell
     absolute magnetization    =     2.34 Bohr mag/cell

All seems in a region to be okay for the first step, but then some
parameter going crazy,
Step 2
     negative rho (up, down):  1.103E+02 1.177E+02
     total energy              =  -18691.45057043 Ry
     Harris-Foulkes estimate   =  -18587.46248750 Ry
     estimated scf accuracy    <   45389.31161404 Ry
     total magnetization       =    -1.39 Bohr mag/cell
     absolute magnetization    =    36.78 Bohr mag/cell
--
     negative rho (up, down):  1.211E+02 1.247E+02
     total energy              =  -26401.86319230 Ry
     Harris-Foulkes estimate   =  -20544.68384009 Ry
     estimated scf accuracy    <   91328.60181762 Ry
     total magnetization       =    -0.44 Bohr mag/cell
     absolute magnetization    =    23.60 Bohr mag/cell

I try a lot of standard things like:

testing: k-points 221 - 881 ,  ecutwfc ecutrho 60-100 600-1000,
different smearing methods and degauss values 0.005 - 0.04,

change mixing_mode, mixing_beta between 0.7-0.1, different
diagonalization themes, nbnd between 196-300, some pseudo potentials
PAW/US,
verify my input via xcrysden, using nspin=1 instead of 2


but nothing works. Can someone explain to me how the absolute
magnetization can rise up to 600 and higher? While the scf accuracy

goes to infinity?

And what can I do to get rid of this errors? I have no ideas left ...
maybe I overlook something easy.

I will print out one of my millions of input data at the end.

Many thanks and
Kind Regards,

Wigbert Polet


MY FIRST INPUT:

// n_make_cells.sh
&control
 calculation = 'relax',
 pseudo_dir = 'NODES_QE_PP',
 prefix = 'NODES_QE_PREFIX',
 outdir = 'NODES_QE_OUT',
 nstep = 200,
 etot_conv_thr = 0.000001,
 forc_conv_thr = 0.0003,
 /

&system
 ibrav = 0,
 nat = 49, ntyp = 3,
 ecutwfc = 60, ecutrho = 600, nbnd = 280,
 nspin = 2,
 starting_magnetization(1)=0.0,
 starting_magnetization(2)=0.0,
 starting_magnetization(3)=1.0,
 occupations = 'smearing',
 smearing = 'methfessel-paxton',
 degauss = 0.005,
 input_dft = 'vdw-df2-b86r'
 /

&ELECTRONS
 conv_thr = 1.D-6,
 mixing_mode='local-TF',
 mixing_beta = 0.5,
 diagonalization = 'david',
 electron_maxstep = 250
 /

&IONS
 ion_dynamics='bfgs'
 /

ATOMIC_SPECIES
C   12.00000000    C.pbe-n-rrkjus_psl.1.0.0.UPF
W  183.9509312     W.pbe-spn-rrkjus_psl.1.0.1.UPF (This doesn't work
see https://dalcorso.github.io/pslibrary/PP_list.html) We used later
other
O   15.99491461956 O.pbe-n-rrkjus_psl.1.0.0.UPF

CELL_PARAMETERS (angstrom)
8.502074  0.000000   0.000000
0.000000 10.084398   0.000005
0.000000 -0.000010 -19.823287

ATOMIC_POSITIONS (angstrom)
 C 7.085057 8.407257 0.000011 0 0 0
 C 4.251032 8.407257 0.000011 0 0 0
 C 1.417007 8.407257 0.000011 0 0 0
 C 7.085057 3.365058 0.000008 0 0 0
 C 4.251032 3.365058 0.000008 0 0 0
 C 1.417007 3.365058 0.000008 0 0 0
 W 5.668050 5.038705 0.000002 0 0 0
 W 2.834025 5.038705 0.000002 0 0 0
 W 0.000000 5.038705 0.000002 0 0 0
 W 5.668050 -0.003494 -0.000001 0 0 0
 W 2.834025 -0.003494 -0.000001 0 0 0
 W 0.000000 -0.003494 -0.000001 0 0 0
 W 5.668050 7.559787 -1.455798 1 1 1
 W 2.834025 7.559787 -1.455798 1 1 1
 W 0.000000 7.559787 -1.455798 1 1 1
 W 5.668050 2.517588 -1.455800 1 1 1
 W 2.834025 2.517588 -1.455800 1 1 1
 W 0.000000 2.517588 -1.455800 1 1 1
 C 7.085057 5.886154 -1.455824 1 1 1
 C 4.251032 5.886154 -1.455824 1 1 1
 C 1.417007 5.886154 -1.455824 1 1 1
 C 7.085057 0.843955 -1.455827 1 1 1
 C 4.251032 0.843955 -1.455827 1 1 1
 C 1.417007 0.843955 -1.455827 1 1 1
 C 7.085057 8.407256 -2.911633 1 1 1
 C 4.251032 8.407256 -2.911633 1 1 1
 C 1.417007 8.407256 -2.911633 1 1 1
 C 7.085057 3.365057 -2.911635 1 1 1
 C 4.251032 3.365057 -2.911635 1 1 1
 C 1.417007 3.365057 -2.911635 1 1 1
 W 5.668050 5.038704 -2.911642 1 1 1
 W 2.834025 5.038704 -2.911642 1 1 1
 W 0.000000 5.038704 -2.911642 1 1 1
 W 5.668050 -0.003495 -2.911644 1 1 1
 W 2.834025 -0.003495 -2.911644 1 1 1
 W 0.000000 -0.003495 -2.911644 1 1 1
 W 5.668050 7.559786 -4.367441 1 1 1
 W 2.834025 7.559786 -4.367441 1 1 1
 W 0.000000 7.559786 -4.367441 1 1 1
 W 5.668050 2.517587 -4.367444 1 1 1
 W 2.834025 2.517587 -4.367444 1 1 1
 W 0.000000 2.517587 -4.367444 1 1 1
 C 7.085057 5.886152 -4.367467 1 1 1
 C 4.251032 5.886152 -4.367467 1 1 1
 C 1.417007 5.886152 -4.367467 1 1 1
 C 7.085057 0.843953 -4.367470 1 1 1
 C 4.251032 0.843953 -4.367470 1 1 1
 C 1.417007 0.843953 -4.367470 1 1 1
 O 2.834025 2.517587 -6.367444 1 1 1

K_POINTS {automatic}
2 2 1 0 0 0
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20190722/e0449056/attachment.html>


More information about the users mailing list