Question regarding Magnetic anisotropy calculation

[topojit]

I was looking into calculating the magnetic anisotropy with both scf and nscf method and I have questions regarding them.

1. In nscf method (https://www.vasp.at/wiki/index.php/Dete ... Anisotropy), we are supposed to calculate the energy along different directions (from (2,2,2) to (2,2,-6)) in the second step. Just to be clear, does that mean we only change the SAXIS value (from (2,2,2) to (2,2,-6)) keeping other parameters same? Doing this, I am getting very low energy difference (0.001meV) (see attached), is that correct or am I doing something wrong here?

2. In the scf method (https://www.vasp.at/wiki/index.php/Incl ... t_Coupling), I calculated the energy along the paths but only changing the SAXIS. My question is, I am getting very low energy difference (see attached file). The energy difference is below 0.01meV, is that correct?

3. My ultimate goal is to calculate MAE along the xz plane. Is there any way to provide the angle values in the INCAR file (0, 30, 90 etc.)? Or do we have to use fractional SAXIS value? Such as (0.5 0 0.866) for 30 deg, (0.866 0 -0.5) for 120 deg etc.

4. Among the scf and nscf method, which one is better or preferred method (efficiency, accuracy etc. wise)?

5. I saw a post yesterday that changing SAXIS along with Hubbard U parameter gives unrealistic result if we read WAVECAR form one step to another (https://www.vasp.at/forum/viewtopic.php ... opy#p30378). The cause was "SAXIS term only sets the direction for the SOC term, it does internally not influence the LDAU term." My question is does it also affect the nscf method of MAE calculation (we read CHGCAR file instead of WAVECAR here)?

Best Regards,
Topojit

[manuel_engel1]

Hi Topojit,

Welcome to the official VASP forum. I will try to answer your questions as best as I can.

1. In nscf method (wiki/index.php/Dete ... Anisotropy), we are supposed to calculate the energy along different directions (from (2,2,2) to (2,2,-6)) in the second step. Just to be clear, does that mean we only change the SAXIS value (from (2,2,2) to (2,2,-6)) keeping other parameters same? Doing this, I am getting very low energy difference (0.001meV) (see attached), is that correct or am I doing something wrong here?
2. In the scf method (wiki/index.php/Incl ... t_Coupling), I calculated the energy along the paths but only changing the SAXIS. My question is, I am getting very low energy difference (see attached file). The energy difference is below 0.01meV, is that correct?

As far as I can tell, you are following the documentation on the wiki correctly. The energy differences can indeed be very small (µeV range), so that is to be expected.

1. My ultimate goal is to calculate MAE along the xz plane. Is there any way to provide the angle values in the INCAR file (0, 30, 90 etc.)? Or do we have to use fractional SAXIS value? Such as (0.5 0 0.866) for 30 deg, (0.866 0 -0.5) for 120 deg etc.

The angle values are given implicitly by specifying SAXIS, and they cannot currently be set directly in the INCAR file.

1. Among the scf and nscf method, which one is better or preferred method (efficiency, accuracy etc. wise)?

The difference between the two methods comes down to computational cost. They should otherwise be equivalent. With the NSCF method, you get a preconverged WAVECAR and CHGCAR file from the first VASP run. Each subsequent spin-orbit calculation should therefore be slightly faster in the NSCF method.

1. I saw a post yesterday that changing SAXIS along with Hubbard U parameter gives unrealistic result if we read WAVECAR form one step to another (forum/viewtopic.php ... opy#p30378). The cause was "SAXIS term only sets the direction for the SOC term, it does internally not influence the LDAU term." My question is does it also affect the nscf method of MAE calculation (we read CHGCAR file instead of WAVECAR here)?

No, this does not affect the NSCF method. You are free to perform both SCF and NSCF MAE calculations.

Let me know if I was able to answer your questions.

Kind regards

[topojit]

Hi,

Thank you for the clarifications. I have another question I was hoping you could clarify for me.

I was trying to calculate the electronic band structure with SOC. I couldn't find any example of SOC-band on wiki, so followed examples on the web and here are the steps:

Step1: structure relaxed without any SOC.
Step2: collinear scf calculation without any SOC.
Step3: Take CHGCAR from step2 and run non-collinear SOC scf calculations. Changed the INCAR:

Code: Select all

ICHARG = 1
LSORBIT = .TRUE.
LORBMOM = .TRUE
LNONCOLLINEAR = .TRUE.
MAGMOM = 0 0 2 0 0 -2 24*0  # total 10 atoms in the POSCAR
SAXIS = 0 0 1

Step4: Take the CHGCAR and run band calculations with a line-mode KPOINTS with high symmetry points and ICHARG = 11.

My questions are:
1. Is the above process correct? Please suggest any redundant step or possible improvements.
2. In step 3, I have seen both "ICHARG =1" and "ICHARG =11" are used, which one is correct here?

Thanks,
Topojit

[manuel_engel1]

I'm happy to help.

1. Is the above process correct? Please suggest any redundant step or possible improvements.

That seems like a very sensible workflow. You don't necessarily have to perform the preparatory step with collinear spin, but we always recommend it. That way, you are more likely to reach the correct magnetic groundstate.

2. In step 3, I have seen both "ICHARG =1" and "ICHARG =11" are used, which one is correct here?

This is quite nicely explained on the VASP wiki. Here is the relevant excerpt:

non-selfconsistent calculations: Adding 10 to the value of ICHARG, e.g., ICHARG=11 or 12 (or the less convenient value 10) means that the charge density will be kept constant during the entire electronic minimization.

That means for band-structure calculations you want to use ICHARG=11 to keep the charge density constant and perform a non-self-consistent calculation.

[topojit]

Hi,

Thank you for the clarification. Maybe I wasn't clear about my confusion in the second question. In step 4, we are calculating band with line mode KPOINTS, so using "ICHARG = 11" makes sense. But in step 3, the example used same KPOINTS file as step 2 (uniform k grid mesh) so this is not a band calculation. I want to make sure I use the correct value for this step 3. My initial thought was to use value "11" in this step, but the tip on the "ICHARG" page on vaspwiki got me confused.

Tip: To improve convergence and reduce the number of electronic steps, it is recommended to set ICHARG = 1 when starting calculations repeatedly with small changes in the input parameters.

I am not clear what counts small change in the INCAR file to justify the use of "ICHARG = 1".

[manuel_engel1]

Thank you for asking again. I think I mixed up step 3 and 4 when responding, so I understand your confusion. What I have written about band-structure calculations is absolutely right, but step 3 is a self-consistent step. In this case, it can be good to read the CHGCAR from disk but then still update it, which is facilitated by ICHARG=1.

The small change refers to all input parameters, not just INCAR. Important are the positions in the POSCAR file. In your case, they stay constant between steps 2 and 3, which is fine.

Sorry again for the confusion. I hope it's much clearer now

[topojit]

Yes, it's clear now. Thank you for the clarifications.

Best,
Topojit