[Wien] Too large orbital polarization with orbital polarization potential Vop And Total energy of easy axis and hard axis is reversed !!

[Gerhard H Fecher]

Our measurements (XMCD) revealed for thin fcc and L1_0 FePt films
0.2 and 0.24 muB for the orbital moments at the Fe atoms, at room temperature 
(300K) and an applied induction field of about 2T. 
Values for bulk material should probably be smaller, what may be compensated 
at 0K partially.
There may also be some form anisotropy, that is your magnetization curve may 
depend on the shape of the sample and its alignment with respect to the 
applied magnetic field.

However, there are several other things unknown if analyzing XMCD data, 
starting with the background being usually assumed to be more or less 
step-like, some rather rough estimate.
Further, that are the unknown Tz and the unknown number of d-holes. Both 
cannot be measured easily. Sum rule analysis does neither account for the 
variation of the matrix elements nor of the relativistic effects like the 
j-dependency of the wavefunctions for both large and small components.

Eperimentally, XMCD integrates over the excited shell, that is the extend of 
the core holes: here the Fe 2p1/2 and 2p3/2. The Fe 3d is without doubt 
delocalized, so I call in question that one is able to sparate the atom 
resolved magnetic moments at all.

In most cases I found that the experimental and calculated orbital to spin 
moment ratios aggreed much better than the pure orbital moments.

If you are interested, from full relativistic KKR calculations (ASA) I found 
the following values:

fcc:
pure VWN
Fe  ms = 2.945  ml = 0.072
Pt  ms = 0.237 ml = 0.037

VWN + OP
Fe  ms=2.942, ml=0.115
Pt  ms = 0.232, ml=0.042


L1_0
VWN + OP
Fe  ms=2.859, ml=0.117
Pt  ms = 0.318, ml=0.044

The OP algorithm is usefull, as you will find a lot of examples were pure LDA 
or GGA calculations fail to give the order of orbital moments correctly.
Roughly spoken: The problem is that spin polarized calculations assume only a 
spin dependent coupling of states neglecting orbital contributions that are 
only indirectly present due to spin-orbit coupling. This just does not 
describe the problem fully and OP is just one approximation to correct for 
that.
However, you should have a clear mind what you compare and how.
Note that under certain conditions, you will not even find the easy axis of 
the elemental ferromagnets correctly ( there are pretty lot examples 
published), from pure LDA calculations neither.

About the MAEs: Did you check the charge convergence or only the energy 
criterium ? In case the first was still bad, you may restart the so far 
convereged calculation and check for charge consistence.

Anyway, just to say it in flapsy words: To calculate the MAE from the total 
energies means to do the following experiment: Magnetize youre sample along 
the hard axis, wait that it switches to the easy direction. The difference in 
the temperature times Boltzmann constant gives you the MAE. No one will even 
think about doing that experiment. 

Ciao
Gerhard

PS.: In my opinion one should play Wien2k and its options (as for other band 
structure programs) more like a music instrument, or does someone scratch a 
violine with a diamond file and expects nice tune ?



Am Freitag, 2. September 2005 10:12 schrieb yasuharu_shiraishi at fujifilm.co.jp:
> 
> Dear Gerhard and Steffan,
> 
>   Thank you!!
>   I will calculate orbital moment with different Muffin-tin radius from now
>   on!
> 
>   But I think that orbital moment of Fe is too large as compared with
>   experimental value.(Physical Review B, 63, 144409)
>   My calculation and experimental value of orbital moment is as follows.
> 
>   my calculation
>   Fe  0.298
>   Pt  0.054
> 
>   experimental value
>   Fe  0.07
>   Pt  0.10
> 
>   And sorry, I have more one question about MAE.
>   If I calculate total energy with orbital polarization potential Vop,
>   Total energy of easy axis and hard axis is reversed !!
>   Outcome is as follows !!
> 
>   My calculation of Total energy with FePt on spin-polarization
>   with spin-orbit interaction is as follows.
> 
>   FePt[001](easy axis)     -39414.569518(Ry)
>   FePt[110](hard axis)      -39414.569367(Ry)
> 
>   I get MAE as follows
>     MAE = Total energy[1 1 0] (hard axis) - Total energy[0 0 1] (easy axis)
> 
>   So I get MAE as 0.000151(Ry).
> 
>   But  I calculate MAE on spin polarization with spin-orbit interaction and
>   orbital polarization potential.
>   Outcome is as follows.
> 
>   FePt[001](easy axis)       -39414.570737(Ry)
>   FePt[110](hard axis)        -39414.572893(Ry)
> 
>   So  I get MAE as -0.002156 Ry, total energy of easy axis and hard axis is
>   reversed!!
> 
>   Experimental  value  is  0.000090Ry,  so  if I calculate MAE with orbital
>   polarization,
>   I  get  strange  value  and  total  energy  of easy axis and hard axis is
>   reversed!!
>   I think if I want to calculate MAE, I should not use orbital polarization
>   potential
>   Vop in WIEN2k. Is it Ok?
>   And is algorithm of orbital polarization in Wien2k useful?
>   Best regards.
> 
> 
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>