LDA+U, Spin-Orbit Coupling and Orbital Moment

[boris]

Dear all,

I'm using vasp 5.2 and I want to calculate the orbital moment carried by uranium atoms in uranium dioxide UO2.

I'm using the LDA+U formalism and I included the spin-orbit coupling in the calculations. I'm printing the orbital moment by setting LORBMOM=.TRUE.

The result is, however, inconsistent with the value given in the literature. I'm supposed to find something like -3.5 MuB for the orbital moment and vasp yields 0.0 MuB.

Is there any known issue with the spin-orbit coupling implementation in vasp together with the LDA+U formalism?

Thank you for your answers.

Regards

[Danny]

Maybe you need to initialise the orbital momenta yourself (similar to magmom) on a value close to what it should be?

I don't seem to be able to find any reference to LORBMOM in the online manual...

cheers
Danny

[boris]

There is indeed no reference to LORBMOM in the manual. I managed to find this keyword thanks to a question asked on this forum.

Unfortunately, there was nothing more about it.

[admin]

just a silly question: have you given all other parameters to do SOC as well (LSORBIT, LNONCOLLINEAR, SAXIS, ISYM=-1)?

[boris]

Yes, I have set LSORBIT = .TRUE., LNONCOLLINEAR is automatically set to TRUE as well, SAXIS = 0 0 1 but I set ISYM to 0, not to -1. Does this make any difference?

Thank you for your answer

[boris]

I have performed another calculation using ISYM=-1. There is no difference compared to the ISYM=0 case.

Here is the result:

orbital moment (z)

 # of ion     p       d       f       tot
 ------------------------------------------------
  1        -0.002  -0.002  -0.368  -0.372
  2        -0.002  -0.002  -0.368  -0.372
  3         0.002   0.002   0.368   0.372
  4         0.002   0.002   0.368   0.372
  5         0.000   0.000   0.000   0.000
  6         0.000   0.000   0.000   0.000
  7         0.000   0.000   0.000   0.000
  8         0.000   0.000   0.000   0.000
  9         0.000   0.000   0.000   0.000
 10         0.000   0.000   0.000   0.000
 11         0.000   0.000   0.000   0.000
 12         0.000   0.000   0.000   0.000
 ------------------------------------------------
            0.000   0.000   0.000   0.000

There is no orbital moment in the other directions. The spin moment is given by :

magnetization (z)

# of ion     s       p       d       f       tot
------------------------------------------------
  1        0.013  -0.001   0.030   1.967   2.010
  2        0.013  -0.001   0.030   1.967   2.010
  3       -0.013   0.001  -0.030  -1.967  -2.010
  4       -0.013   0.001  -0.030  -1.967  -2.010
  5        0.000   0.000   0.000   0.000   0.000
  6        0.000   0.000   0.000   0.000   0.000
  7        0.000   0.000   0.000   0.000   0.000
  8        0.000   0.000   0.000   0.000   0.000
  9        0.000   0.000   0.000   0.000   0.000
 10        0.000   0.000   0.000   0.000   0.000
 11        0.000   0.000   0.000   0.000   0.000
 12        0.000   0.000   0.000   0.000   0.000
------------------------------------------------
tot        0.000   0.000   0.000   0.000   0.000

Therefore, it makes a total magnetic moment of 1.64 MuB, which I think is in very good agreement with the experimental value (1.74 MuB).

However, other people have calculated it using first principles DFT+U calculations and found the orbital moment to be -3.5.

I can't figure out what is wrong.

Thank you




<span class='smallblacktext'>[ Edited Mon Apr 19 2010, 03:08PM ]</span>

[boris]

Ok, I found what was going wrong.

It is written in the vasp manual that one should first calculate a collinear WAVECAR and CHGCAR file, and then add the spin-orbit coupling while restarting with the two files.

Actually,it appears that starting from a charge density without any spin-orbit coupling terms is not a good method.

I started from scratch directly with the spin-orbit coupling and it gave an orbital moment of -3 MuB.