[Pw_forum] Bizarre (and rather unexpected) absolute magnetization?

[Van den Berg, Jan (JA)]

Dear PWscf users,

I am in the process of doing some LDA+U calculations on binary systems, and discovered a peculiar behaviour with Ru-based binary systems. I have been using the PAW pseudopotential Ru.pbe-spn-kjpaw_psl.0.3.0.UPF - it is not in the standard set, but available from THEOS under the PSEUDOPOTENTIALS_TOT/  directory. Aside from the fact that CG style diagonalization suffers (as seen from many iterations where eigenvalues are not converged), the bizarre result is seen in the absolute magnetization.

TYPICAL INPUT FILE:
&CONTROL
  title = "single-point calculation"
  calculation  = 'scf'
  prefix       = "fcc-RuAl"
  pseudo_dir   = "~/mypseudo/"
  wf_collect=.true.
/
&SYSTEM
  ibrav     = 0
  nat       = 4
  ntyp      = 3
  ecutwfc   = 80
  occupations = 'smearing'
  smearing = 'methfessel-paxton'
  degauss = 0.01
  nspin = 2
  starting_magnetization(1) = 0.1,0.1,0.01
  lda_plus_U=.true.
  Hubbard_U(1)=1D-40,Hubbard_alpha(1)=1D-40
  Hubbard_U(2)=1D-40,Hubbard_alpha(2)=1D-40
  Hubbard_U(3)=1D-40,Hubbard_alpha(3)=1D-40
  nosym_evc=.true.
/
&ELECTRONS
  startingwfc='atomic+random'
  startingpot='atomic'
  conv_thr    = 1.D-6
  mixing_beta = 0.5D0
  electron_maxstep = 4000
  diagonalization = 'cg'
/
ATOMIC_SPECIES
Ru1  101.07  Ru.pbe-spn-kjpaw_psl.0.3.0.UPF
Ru2  101.07  Ru.pbe-spn-kjpaw_psl.0.3.0.UPF
Al1  26.982  Al.pbe-n-kjpaw_psl.0.1.UPF
ATOMIC_POSITIONS (crystal)
Al1  0.0  0.0  0.0
Ru1  0.0  0.5  0.5
Al1  0.5  0.0  0.5
Ru2  0.5  0.5  0.0
K_POINTS (automatic)
  9 9 9 0 0 0
CELL_PARAMETERS (angstrom)
   3.6 0.0 0.0
   0.0 3.6 0.0
   0.0 0.0 3.6

TYPICAL OUTPUT:
....

Iteration #3:
     Magnetic moment per site:
     atom:    1    charge:    1.5011    magn:    0.0183    constr:    0.0000
     atom:    2    charge:   14.3856    magn:    0.4645    constr:    0.0000
     atom:    3    charge:    1.4893    magn:    0.0330    constr:    0.0000
     atom:    4    charge:   14.4619    magn:    0.4294    constr:    0.0000

     total cpu time spent up to now is      618.8 secs

     total energy              =   -5726.20306650 Ry
     Harris-Foulkes estimate   =   -4814.95640865 Ry
     estimated scf accuracy    <   25264.02550193 Ry

     total magnetization       =     0.11 Bohr mag/cell
     absolute magnetization    =  2115.04 Bohr mag/cell

Iteration #4:
     Magnetic moment per site:
     atom:    1    charge:    1.5539    magn:   -0.0043    constr:    0.0000
     atom:    2    charge:   14.7995    magn:    0.3472    constr:    0.0000
     atom:    3    charge:    1.5631    magn:    0.0130    constr:    0.0000
     atom:    4    charge:   14.1133    magn:    0.2910    constr:    0.0000

     total cpu time spent up to now is      847.5 secs

     total energy              =   -6800.82522725 Ry
     Harris-Foulkes estimate   =   -5911.29292705 Ry
     estimated scf accuracy    <   12707.95018218 Ry

     total magnetization       =    -0.07 Bohr mag/cell
     absolute magnetization    =  1617.61 Bohr mag/cell


....

     iteration #113     ecut=    80.00 Ry     beta=0.50
     CG style diagonalization
     c_bands:  1 eigenvalues not converged
     c_bands:  1 eigenvalues not converged
     ethr =  1.00E-02,  avg # of iterations =  3.1

     negative rho (up, down):  8.618E+02 8.613E+02
     Magnetic moment per site:
     atom:    1    charge:    1.2423    magn:   -0.0288    constr:    0.0000
     atom:    2    charge:   15.6339    magn:   -0.0550    constr:    0.0000
    atom:    3    charge:    1.3867    magn:   -0.0177    constr:    0.0000
     atom:    4    charge:   15.5383    magn:   -0.0419    constr:    0.0000

     total cpu time spent up to now is    10453.0 secs

     total energy              =   -8411.78286402 Ry
     Harris-Foulkes estimate   =   -8410.36053542 Ry
     estimated scf accuracy    <      28.63071702 Ry

     total magnetization       =    -0.20 Bohr mag/cell
     absolute magnetization    =   144.41 Bohr mag/cell

     iteration #114     ecut=    80.00 Ry     beta=0.50
     CG style diagonalization
     c_bands:  1 eigenvalues not converged
     ethr =  1.00E-02,  avg # of iterations =  3.0

     negative rho (up, down):  8.618E+02 8.614E+02
     Magnetic moment per site:
     atom:    1    charge:    1.2416    magn:   -0.0288    constr:    0.0000
     atom:    2    charge:   15.6333    magn:   -0.0558    constr:    0.0000
     atom:    3    charge:    1.3864    magn:   -0.0179    constr:    0.0000
     atom:    4    charge:   15.5402    magn:   -0.0446    constr:    0.0000

     total cpu time spent up to now is    10529.2 secs

     total energy              =   -8411.31255477 Ry
     Harris-Foulkes estimate   =   -8411.86059006 Ry
     estimated scf accuracy    <      24.75641734 Ry

     total magnetization       =    -0.20 Bohr mag/cell
     absolute magnetization    =   142.71 Bohr mag/cell



At this stage I terminated the program.  I would expect that both total magnetization and absolute magnetization go to zero, but instead large absolute magnetizations are maintained.  Although I could do the calculation without spin, I want the code to find the nonmagnetic state - so far without much luck.

Any advice would be highly appreciated!

Regards,
Jan-Albert

NOTICE: Please note that this eMail, and the contents thereof, is subject to the standard Sasol eMail legal notice which may be found at: http://www.sasol.com/legalnotices
If you cannot access the legal notice through the URL attached and you wish to receive a copy thereof please send an eMail to legalnotice at sasol.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20150715/869556b0/attachment.html>