magnetic moments in antiferromagnetic material are zero with ISPIN=2, both finite with SOC

[sophie_weber]

Hello,
I have run into this odd phenomena with two or more materials (to be specific, right now I'm looking at a (010) slab of rock-salt NiO), and was curious if anyone knows might be going on. NiO is of course an antiferromagnetic material, however, when I run an scf calculation in VASP with collinear magnetism (ISPIN=2) and set the MAGMOM flag in the INCAR to finite AFM values for the Ni atoms, when I look at the final site-projected magnetization in the OUTCAR file the moments become zero to within 0.001 uB (interestingly, for the case of NiO bulk, the moments remain what they should be, ~ 2 uB). However, when I switch SOC on for the (010) NiO slab but still set the MAGMOM flag to a collinear AFM order, and keep all other settings the same, now the converged moments in the OUTCAR are the appropriate size for Ni.

Has anyone experienced this phenomenon of atomic site moments going to zero (when they definitely should not) for an ISPIN=2 calculation, but not for the same calculation with spin-orbit coupling switched on? I can provide input files if necessary.

Thank you in advance for your help.

[merzuk.kaltak]

Dear Sophie,

please provide some input (and possibly output files, i.e. OUTCAR and stdout).

[sophie_weber]

Hello,

Sure; attached are tar files with the input files + OUTCAR for the case with no SOC (zero site-projected moments in the OUTCAR) and with SOC (finite, as-expected magnetic moments in OUTCAR). This particular SOC calculation is using constrained magnetism, so I'm running an SOC calculation without constraints to verify I get the same outcome, but based on this happening with previous materials I'm quite confident it will be the same.

Thank you so much for your help, I look forward to hearing back!

[merzuk.kaltak]

Dear Sophie,

vasp-5.4.4 is symmetrizing the magnetic structure in the 010NiO_noSOC example such that you end up with a vanishing netto magnetic moment.
In vasp-6, the routines that handle symmetry have been improved and the correct magnetic moment is calculated with your input files.
I suggest to update to vasp-6 if possible.

There is a way to achieve the correct result with vasp-5.4.4 as well, but it requires to "break" the symmetry.
This means, changing the absolute value of the magnetic moments in opposite planes slightly.
For this purpose, I have re-arranged the Ni-ions in the POSCAR for easy initialization of MAGMOM as so:

Code: Select all

Ni32 O32                                
   1.00000000000000     
     0.0000000000000000    0.0000000000000000    8.4411439999999995
     8.4411439999999995    0.0000000000000000    0.0000000000000000
     0.0000000000000000   23.4411440000000013    0.0000000000000000
   Ni   O 
    32    32
Direct
  0.0000009186760082  0.0000009186760082  0.9995904741152017 !
  0.2500009186760082  0.7500009186760082  0.9995904741152017 !
  0.5000009186760153  0.5000009186760153  0.9995904741152017 !
  0.7500009186760082  0.2500009186760082  0.9995904741152017 !
  0.7499989926286190  0.9999989926286261  0.0909087713267027 !
  0.9999989926286261  0.7499989926286190  0.0909087713267027 !
  0.2499989926286261  0.4999989926286190  0.0909087713267027 !
  0.4999989926286190  0.2499989926286261  0.0909087713267027 !
  0.0000010073713739  0.5000010073713739  0.1791662286732958 !
  0.5000010073713739  0.0000010073713739  0.1791662286732958 !
  0.7500010073713810  0.7500010073713810  0.1791662286732958 !
  0.9999990813239918  0.2499990813239918  0.2704845258847968 !
  0.4999990813239918  0.7499990813239918  0.2704845258847968 !
  0.2499990813239918  0.9999990813239918  0.2704845258847968 !
  0.7499990813239918  0.4999990813239918  0.2704845258847968 !
  0.2500010073713810  0.2500010073713810  0.1791662286732958 !
  0.0000009186760082  0.5000009186760153  0.9995904741152017 
  0.2500009186760082  0.2500009186760082  0.9995904741152017
  0.5000009186760153  0.0000009186760082  0.9995904741152017
  0.7500009186760082  0.7500009186760082  0.9995904741152017
  0.7499989926286190  0.4999989926286190  0.0909087713267027
  0.2499989926286261  0.9999989926286261  0.0909087713267027
  0.9999989926286261  0.2499989926286261  0.0909087713267027
  0.4999989926286190  0.7499989926286190  0.0909087713267027
  0.0000010073713739  0.0000010073713739  0.1791662286732958
  0.7500010073713810  0.2500010073713810  0.1791662286732958
  0.2500010073713810  0.7500010073713810  0.1791662286732958
  0.5000010073713739  0.5000010073713739  0.1791662286732958 
  0.9999990813239918  0.7499990813239918  0.2704845258847968 
  0.7499990813239918  0.9999990813239918  0.2704845258847968 
  0.4999990813239918  0.2499990813239918  0.2704845258847968  
  0.2499990813239918  0.4999990813239918  0.2704845258847968
  0.7500005855785261  0.0000005855785261  0.0075619336923012
  0.0000005855785261  0.7500005855785261  0.0075619336923012
  0.2500005855785261  0.0000005855785261  0.0075619336923012
  0.7500005855785261  0.5000005855785261  0.0075619336923012
  0.2500005855785261  0.5000005855785261  0.0075619336923012
  0.5000005855785261  0.7500005855785261  0.0075619336923012
  0.0000005855785261  0.2500005855785261  0.0075619336923012
  0.5000005855785261  0.2500005855785261  0.0075619336923012
  0.2499999012680476  0.2499999012680476  0.0986568337893132
  0.4999999012680476  0.9999999012680476  0.0986568337893132
  0.9999999012680476  0.9999999012680476  0.0986568337893132
  0.4999999012680476  0.4999999012680476  0.0986568337893132
  0.7499999012680476  0.7499999012680476  0.0986568337893132
  0.2499999012680476  0.7499999012680476  0.0986568337893132
  0.7499999012680476  0.2499999012680476  0.0986568337893132
  0.9999999012680476  0.4999999012680476  0.0986568337893132
  0.0000000987319524  0.2500000987319524  0.1714181662106853
  0.2500000987319524  0.0000000987319524  0.1714181662106853
  0.2500000987319524  0.5000000987319524  0.1714181662106853
  0.5000000987319524  0.7500000987319524  0.1714181662106853
  0.7500000987319524  0.0000000987319524  0.1714181662106853
  0.5000000987319524  0.2500000987319524  0.1714181662106853
  0.7500000987319524  0.5000000987319524  0.1714181662106853
  0.0000000987319524  0.7500000987319524  0.1714181662106853
  0.2499994144214739  0.2499994144214739  0.2625130663076973
  0.9999994144214739  0.9999994144214739  0.2625130663076973
  0.7499994144214739  0.2499994144214739  0.2625130663076973
  0.9999994144214739  0.4999994144214739  0.2625130663076973
  0.2499994144214739  0.7499994144214739  0.2625130663076973
  0.4999994144214739  0.4999994144214739  0.2625130663076973
  0.7499994144214739  0.7499994144214739  0.2625130663076973
  0.4999994144214739  0.9999994144214739  0.2625130663076973

Then the positive and negative magnetic moments can be initialized like

Code: Select all

MAGMOM = 16*-2.1 16*2.0 32*0 

This was enough to find following magnetic moments in the OUTCAR (gamma-only version):

Code: Select all

 # of ion       s       p       d       tot
------------------------------------------
    1        0.003  -0.005  -1.641  -1.644
    2        0.003  -0.005  -1.641  -1.644
    3        0.003  -0.005  -1.641  -1.644
    4        0.003  -0.005  -1.641  -1.644
    5       -0.001  -0.003  -1.606  -1.609
    6       -0.001  -0.003  -1.606  -1.609
    7       -0.001  -0.003  -1.606  -1.609
    8       -0.001  -0.003  -1.606  -1.609
    9       -0.001  -0.003  -1.606  -1.609
   10       -0.001  -0.003  -1.606  -1.609
   11       -0.001  -0.003  -1.606  -1.609
   12        0.003  -0.005  -1.641  -1.644
   13        0.003  -0.005  -1.641  -1.644
   14        0.003  -0.005  -1.641  -1.644
   15        0.003  -0.005  -1.641  -1.644
   16       -0.001  -0.003  -1.606  -1.609
   17       -0.003   0.005   1.641   1.644 
   18       -0.003   0.005   1.641   1.644 
   19       -0.003   0.005   1.641   1.644 
   20       -0.003   0.005   1.641   1.644 
   21        0.001   0.003   1.606   1.610 
   22        0.001   0.003   1.606   1.610 
   23        0.001   0.003   1.606   1.609 
   24        0.001   0.003   1.606   1.609 
   25        0.001   0.003   1.606   1.609 
   26        0.001   0.003   1.606   1.609 
   27        0.001   0.003   1.606   1.609 
   28        0.001   0.003   1.606   1.609 
   29       -0.003   0.005   1.641   1.644 
   30       -0.003   0.005   1.641   1.644 
   31       -0.003   0.005   1.641   1.644 
   32       -0.003   0.005   1.641   1.644 
   33       -0.004  -0.012   0.000  -0.015
   34       -0.004  -0.012   0.000  -0.015
   35        0.004   0.012   0.000   0.015 
   36        0.004   0.012   0.000   0.015 
   37       -0.004  -0.012   0.000  -0.015
   38        0.004   0.012   0.000   0.015 
   39        0.004   0.012   0.000   0.015 
   40       -0.004  -0.012   0.000  -0.015
   41       -0.003   0.048   0.000   0.045 
   42       -0.003   0.048   0.000   0.045 
   43        0.003  -0.048   0.000  -0.045
   44        0.003  -0.048   0.000  -0.045
   45       -0.003   0.048   0.000   0.045 
   46        0.003  -0.048   0.000  -0.045
   47        0.003  -0.048   0.000  -0.045
   48       -0.003   0.048   0.000   0.045 
   49        0.003  -0.048   0.000  -0.045
   50        0.003  -0.048   0.000  -0.045
   51       -0.003   0.048   0.000   0.045 
   52        0.003  -0.048   0.000  -0.045
   53       -0.003   0.048   0.000   0.045 
   54       -0.003   0.048   0.000   0.045 
   55        0.003  -0.048   0.000  -0.045
   56       -0.003   0.048   0.000   0.045 
   57       -0.004  -0.012   0.000  -0.015
   58        0.004   0.012   0.000   0.015
   59        0.004   0.012   0.000   0.015
   60       -0.004  -0.012   0.000  -0.015
   61        0.004   0.012   0.000   0.015
   62        0.004   0.012   0.000   0.015
   63       -0.004  -0.012   0.000  -0.015
   64       -0.004  -0.012   0.000  -0.015
--------------------------------------------------
tot          0.000  -0.001  -0.000  -0.001

You probably can achieve similar results with the anti-symmetric initialization

Code: Select all

MAGMOM=16*-2 16*2 32*0 

by switching off symmetry using ISYM=-1,
however, the job will probably take a long time to finish.

[sophie_weber]

I see, thank you so much for your work and help! It's good at least to know that it was an issue with the 5.4.4 VASP, rather than a serious error in my input files. In the future I will try to use VASP 6.

Thank you again,
Sophie

[merzuk.kaltak]

You are welcome,

As a final remark for this thread:
vasp-5.4.4 does not symmetrize the local magnetic moments correctly (see here).
This is a known issue which is fixed in vasp-6 only.