Is it the right procedure for magnetic materials ?

[salah-eddine_bouzarmine]

Hello everyone,

I would like to know if i am doing things right.

I do calculations in the following order:

1.1. I start with structure relaxation.
1.2. SCF calculation directly using DFT+U+SOC.
in this step, i used to work with the default algo in VASP, which gives unrealistic results (underestimate or overestimate the orbital magnetic moment contribution). So, with some extended research i found out that using ALGO = All solve this issue and i'm using it and works very well.
I do this calculation for each magnetic configuration to determine the ground state one to with which i can continue the calculation. However, for AFM structures i do not get zero total magnetic moment in the cell and to solve this i use I_constrained_M = 4 , LAMBDA, RWIGS and M_CONSTR to set the direction of spins.

1.3. For DOS i do post nscf calculations by setting icharg = 11 & istart = 2 and NEDOS = 2000 to get a smooth DOS curves.
1.4. For Bandstructure i only change kpoints to the k-path and all is good.

2. For magnetocrystalline anisotropy, i used the procedure recommended in the wiki with running scf calculation without spin-orbit coupling then adding spin-orbit coupling with each direction to compute the MAE.
What i observed in the OUTCAR files is the following:
When using scf without spin-orbit we only get the spin contribution which is normal. But after adding spin-orbit we do get the orbital contribution to the magnetic moment but its value is very underestimated compared to the procedure i already described and this orbital contribution is the only one that follows the SAXIS we set at each direction and the energy difference in the range of meV which seems right.

So, is my procedure right ?

[alexey.tal]

Dear salah-eddine_bouzarmine,

Thank you for your question. And sorry for the late reply.

The first procedure seems to be fine for finding the ground state magnetic configuration and calculating DOS or band structure of that state. However, to be certain one should provide all the input and output files for every step. Furthermore, when you use constrained magnetic moments you should keep in mind that the total energy should not be compares as I_CONSTRAINED_M adds a penalty contribution to the total energy.

For calculating Magnetocrystalline Anisotropy Energy with SOC you should read the following guide on our wiki.

Could you please provide the OUTCARs from the calculations where you compare the orbital contribution to the magnetic moment? I'm not sure I understood how these are different.

Please see the forum guidelines.

Best wishes,
Alexey

[salah-eddine_bouzarmine]

Dear alexey.tal,

Thank you so much for your reply, and sorry for not replying so fast.

I have included the files for your reference.

I used the procedure described in the wiki, i started a calculation of my system without using SOC with std version of vasp. When done i added SOC and changed icharg = 11 with different saxis diections calculated with ncl version.

I included a regular scf oucar for comparison, and the outcar file for the calculation without SOC then the outcar with added SOC with saxis 001.

Best regards,
Salah-Eddine

[alexey.tal]

Thank you for providing the files.

2. For magnetocrystalline anisotropy, i used the procedure recommended in the wiki with running scf calculation without spin-orbit coupling then adding spin-orbit coupling with each direction to compute the MAE.
What i observed in the OUTCAR files is the following:
When using scf without spin-orbit we only get the spin contribution which is normal. But after adding spin-orbit we do get the orbital contribution to the magnetic moment but its value is very underestimated compared to the procedure i already described and this orbital contribution is the only one that follows the SAXIS we set at each direction and the energy difference in the range of meV which seems right.

I see that the orbital moment along z-direction is quite substantial, i.e., ~ 0.3. Why do you believe that this value is underestimated?
Did you not find a reasonable MAE following this procedure?

[salah-eddine_bouzarmine]

Thank you for the reply.

In the SCF calculation when SOC is included from the beginning of the calculation, the spin component is around 2.9 muB while the orbital one is about -1.4 muB. So when we do the calculation for magnetocrystalline anisotropy the procedure reproduces the same spin value but it only gives -0.3 for the orbital contribution! That's why i questioned my self if this difference would affect the MAE and determination of easy and hard directions.
In the mean time, the values i get are quite reasonable (-197.88833 eV for easy axis, -197.88820 eV hard axis giving MAE = 0.071 MJ/m^3).

[alexey.tal]

As I wrote above, constraining the moments can be used to converged the system to the desired solution, however, one should not constrain the moments in the MAE calculation as the energies are affected by the penalty term.
And the reason the orbital moments are so different in your "scf" calculation is probably due to the Wigner-Seitz radius RWIGS.