Conflicting OSZICAR and OUTCAR Moment Results

[lizd]

I am attempting to perform a variety of total-moment-constrained calculations of a Heusler alloy. I am using NUPDOWN = [0.2 to 9.2 in steps of 0.2] to achieve this effect, and utilizing a fixed MAGMOM (MAGMOM = 3 -3 0 0 0 0 0 0 for *all* calcs) with a WAVECAR and CHGCAR from a pre-converged non-magnetic (ISPIN = 1) calculation, using ICHARG = 1 for the magnetic calculation.

I am using LORBIT = 11 to get the site-projected moments (and DOSs), and seeing a confusing/inconsistent result. For an extreme example, take NUPDOWN = 5.0:

Code: Select all

OSZICAR:
       N       E                     dE             d eps       ncg     rms          rms(c)
DAV:   1    -0.621885723238E+02   -0.62189E+02   -0.98219E+00166672   0.241E+01    0.103E+01
DAV:   2    -0.645414846099E+02   -0.23529E+01   -0.75025E+00142916   0.193E+01    0.169E+01
[...]
DAV:   9    -0.627650008903E+02   -0.25584E-04   -0.15702E-04165106   0.771E-02    0.626E-02
DAV:  10    -0.627650152948E+02   -0.14404E-04   -0.27906E-05132982   0.393E-02    0.401E-02
DAV:  11    -0.627650157311E+02   -0.43632E-06   -0.42667E-06 92242   0.197E-02
   1 F= -.62765016E+02 E0= -.62765016E+02  d E =0.000000E+00  mag=     0.0000

Code: Select all

OUTCAR:
[...]
 magnetization (x)
 
# of ion     s       p       d       f       tot
------------------------------------------------
  1        0.041   0.054   3.420   0.000   3.514
  2       -0.038  -0.025  -2.293   0.000  -2.357
  3        0.008   0.005   0.790   0.004   0.807
  4        0.019   0.007   0.771   0.001   0.799
  5        0.019   0.007   0.771   0.001   0.799
  6        0.008   0.005   0.790   0.004   0.807
  7        0.003   0.029   0.021   0.000   0.053
  8        0.003   0.029   0.021   0.000   0.053
------------------------------------------------
tot        0.063   0.111   4.291   0.010   4.475
[...]

While I understand that the projections done in LORBIT=11 may leave some "vacuum" charge/moment, this mismatch (0 vs 4.4) is a little hard to justify.

At each step, in the OUTCAR, the magnetization is reported as 0, e.g.:

Code: Select all

 ----------------------------------------- Iteration    1(  20)  ---------------------------------------

[...]
 
  eigenvalue-minimisations  : 59222
  total energy-change (2. order) : 0.1606531E-07  (-0.7278164E-07)
  number of electron     122.0000002 magnetization       0.0000000
  augmentation part       28.0579698 magnetization       0.0000000

If I plot the site-projected DOS I get a symmetric DOS, consistent with the mu_b = 0 configuration. But the energy is different from the NUPDOWN = 0 case!

What should I believe here, and what is possibly going on?

[admin]

by setting NUPDOWN the total magnetic moment of the unit cell is fixed, this is NOT equivalent to the
sum of local moments. The mismatch is related to the 'interstitial' region in the cell (or double-counting of
volume, if neighboring spheres overlap). There is no general rule to decide whether a mismatch between NUPDOWN and the sum of
local moments is reasonable .
in OSZICAR, the final moment in the cell is the same value of NUPDOWN. as set in INCAR.
Also, the magnetization as written in OUTCAR is the value set in INCAR. If it is not,
there must have been an error in reading the tag from INCAR, or some of the files got mixed up.
to give a more detailled answer, one should have a look at the details of both calculations, please send a complete error report to
the vaspmaster