Input Parameters for Spin Polarised Calculations

[satvik_kashyap]

Hello,

I want to relax AlV with spin polarisation turned on.

The system has 16 atoms of Al and V each. I am using only the Gamma Point.
I relaxed the structure without Spin-Polarisation and then used the CONTCAR, WAVECAR and CHGCAR to relax with Spin-Polarisation (as suggested on VASP Wiki).

Upon using ISPIN = 2, with MAGMOM = 16*1.5 16*4.5, the result yields a structure with a magnetization of around 4. However, AlV being non-magnetic, the expected result is zero.
There isn't a significant difference in the result even after reducing the values of magnetic moments in MAGMOM (including omitting the tag).
Upon setting the individual magnetic moment to zero, the net magnetization is close to zero. However, the obtained structure has greater energy than structures with finite magnetization.

Upon testing different Smearing Configurations, the following was the result obtained (picture attached)-

Screenshot from 2023-06-15 14-34-51.png

The issue repeats for other intermetallics like Al5Co2 and AlCo

Any suggestions would be really helpful regarding the choice of parameters to obtain accurate results.

[marie-therese.huebsch]

Hi,

Maybe I don't get your point, but -210.15eV < -204.17eV.

By the way, omitting MAGMOM means its default is used, and that is 1 mu_B on each site.

Marie-Therese

[satvik_kashyap]

Hi,

The comment I made about energy was when I tested configurations with different magnetic moments in MAGMON. The magnetic structure seemed to have lower energy than the non-magnetic structure, which didn't make sense considering the system is supposed to be non-magnetic.

When I tested the system by keeping MAGMOM constant and varying the ISMEAR and SIGMA, the results obtained were as mentioned in the attached picture.
Yes, -210.15eV < -204.17eV, but the Entropy corresponding to the system having E = -210.12 eV is much higher than the other systems (in absolute value), which is supposed to be kept low (as per VASP Wiki, less than 1 meV/atom). Do we consider the negative sign while checking if the entropy is lesser than the desired value?

Also, since this is a metallic system, shouldn't the configuration with ISMEAR = 1 yield better results?

[marie-therese.huebsch]

Hi,

Generally, energies cannot be compared across different settings for ISMEAR and SIGMA. You can only interpret the energy difference between different magnetic configurations or sizes of the moments if all other settings, e.g., smearing, k-point sampling, PAW basis, etc., are the same.

ISMEAR > 0 is a good choice for metals. But your result, i.e., which magnetic configuration is most stable, should not critically depend on the choice of the smearing. That means the energy difference btw two calculations that defer only in the magnetic configuration should be nearly independent of the smearing method. If that were not the case, I would worry about convergence or consider if the compared systems are degenerate (at least within spin-density functional theory, if not in reality).

If there is a specific question about a specific calculation, could you please upload the input and main output files as stated in the forum guidelines? Thanks.

Marie-Therese

[satvik_kashyap]

Hi,

To clarify-

There are two sets of experiments.

In the first set of experiments, parameters like ISMEAR, SIGMA, ENCUT etc., were kept constant, and the value of MAGMOM was varied.
For all values of MAGMOM except the setting where the magnetic moment on the individual atoms was set to zero, the results yielded a magnetic system. Only when the MAGMOM was set as zero, keeping the magnetic moment on each atom as zero, was the system found to be non-magnetic.
However, the non-magnetic system had a higher energy than the magnetic system. This was counterintuitive as the system is supposed to be non-magnetic.

In the second set of experiments, MAGMOM was kept constant, and the parameters related to smearing- ISMEAR and SIGMA were varied.
The attached picture of the results (in the first post on the topic) corresponds to this experiment. The main objective here was to compare the Entropy and Magnetization values.
The absolute value of entropy when the configuration yielded a non-magnetic solution, was much higher than the entropy obtained from the configuration that yielded a magnetic system. I am given to understand that the entropy needs to be kept minimum. But in doing so, the magnetization in the system is unphysical.

The results of these two experiments have put me in a dilemma as to what parameters do I use