2D polarization calculation

[pooja_rani]

Hi,

I am trying to perform berry phase polarization calculation for a 2D material. I used the tag LCALCPOL, but every time i am getting some warning in the output, which i attached in the form of a screenshot here. I tried to use a dense K-grid and also used the other method by using LBERRY tags, but every time I get the same warning. I also use the DIPOL tag with LCALCPOL (like DIPOL = 0.5 0.5 0.144) and tried to vary these values, but I am still getting the same warning. Can you please help me with this problem? The interesting thing in that is it's printing the dipole values in OUTCAR, but that is incorrect. So, I am performing a calculation for a trilayer with centrosymmetric stacking, and the polarization is actually zero, but in OUTCAR I am getting some values, which I think is incorrect. This I am doing to compare with two bilayer polarizations. I hope you understand what i am trying to do.

Also, this is a nonmagnetic material. I do not understand why this is writing "spin channel" in the output. I started with the calculation for the example that is given with the LCALCPOL tag. The calculation was completed, but it did not print any dipole moment in the INCAR file. Can you please correct me if I missed something in this calculation?

Besides that, I have a query, do we need to perform SCF calculation before the Berry phase calculation? I just want to know the proper procedure to get the polarization of a 2D material.

Thank you and kind regards,
Pooja

[zahedzx]

Hi Pooja,

The warning

Code: Select all

PEAD_POLARIZATION_CALC: WARNING: contribution from string ... not well clustered

indicates that the Berry phases from different k-point strings are not sufficiently consistent. This is usually related to k-point sampling, insufficient wavefunction convergence, or a very small band gap (near-degenerate occupied states), rather than the DIPOL tag itself.

The "spin channel 1" message is normal even for a non-spin-polarized calculation (ISPIN=1) and is not an indication that the calculation is magnetic.

Regarding the non-zero polarization: in the modern Berry-phase formalism, the absolute polarization is only defined modulo a polarization quantum. Therefore, a centrosymmetric structure does not necessarily give a printed value of exactly zero. The physically meaningful quantity is usually the change in polarization between two structures.

For polarization calculations, it is recommended to first perform a fully converged SCF calculation and then run a static calculation with LCALCPOL = .TRUE. using the converged charge density/wavefunctions.

Best,
Zahed

[pooja_rani]

Dear Zahed,

Thank you so much for the reply!
I get it now. But I still have a few questions:
1) I already was trying this calculation with denser KPOINTS, I tried different sets, but each time it was showing me this warning. Can you please guide me if there is a way to get exact KPOINTS to do this calculation? For other calculation i am using 11 11 1.
2) For 2D material, how to choose the values for the DIPOL tag. FOr now I am taking the average of the atomic positions, is that right?

Kind Regards,
Pooja Rani

[zahedzx]

Dear Pooja,

There is no universal choice of k-points for Berry-phase calculations. The mesh should be converged for your specific system, so it has to be determined by a convergence test. An 11×11×1 mesh may be sufficient for some 2D systems but not for others.

Regarding DIPOL, this tag does not affect the Berry-phase polarization calculated with LCALCPOL. It is used for the dipole correction (LDIPOL) and defines the reference point for that correction.
If you use LDIPOL, choosing a point near the center of the slab (i.e. center of mass of the atoms in the cell) is reasonable. Otherwise, the DIPOL tag is not relevant for the Berry-phase polarization.

Best,
Zahed