Charged Defects Formation Energy Correction

Top Page > Browsing

Charged Defects Formation Energy Correction

Date: 2023/10/24 19:18
Name: Marc Tunica <marc.tunica-i-rosich@universite-paris-saclay.fr>
References: Correction: Dear all,

I am currently working on structural properties of charged point defects in bulk silicon. As is known, the formation energy and defect level computed in such systems need to be corrected to remove the contributions from the spurious long-range interaction between the defect and its periodic images. To remove this effect, I would like to calculate the Freysoldt-Neugebauer-Van de Walle (FNV) correction scheme (https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.102.016402). The system I am studying is bulk cubic silicon supercell containing a charged boron atom.

After reviewing discussions on this forum, it appears that the correct method for calculating the potential alignment term included in the FNV electrostatic correction implies: i) including the VNA term in the V0 (using the flag scf.ProExpn.VNA off), ii) increasing the energy cutoff, and iii) calculating the difference between the Kohn-Sham and exchange-correlation potentials (V0 - Vxc0), which represents the electrostatic potential.

After analyzing my results, I have observed a small difference in the formation energy (0.06 eV) applying this procedure for the same system using scf.ProExpn.VNA turned off and on. I would greatly appreciate any insights or explanations about this discrepancy, in particular because I optimized the lattice parameter with scf.ProExpn.VNA turned on.

Furthermore, as for the Freysoldt approach, I would like to know if considering just the Hartree contribution (without including the electrostatic ionic interactions) in the potential alignment term is accurate enough. Any comment on that will be greatly appreciated.

Thank you in advance for any assistance

Best regards,
Marc

Page: [1]

Re: Charged Defects Formation Energy Correction ( No.1 )

Date: 2023/10/27 14:19
Name: T. Ozaki

Hi,

> After analyzing my results, I have observed a small difference in the formation energy (0.06 eV)
> applying this procedure for the same system using scf.ProExpn.VNA turned off and on.
> I would greatly appreciate any insights or explanations about this discrepancy, in particular because
> I optimized the lattice parameter with scf.ProExpn.VNA turned on.

In case of scf.ProExpn.VNA=off, the accuracy is controlled by scf.energycutoff,
while for scf.ProExpn.VNA=on, it is controlled by scf.BufferL.VNA (default:6) and scf.RadialF.VNA (default: 12).
As those parameters are set to severer, the two computational results are getting closer.
The difference of 0.06 eV might be within a variation we expect.

> Furthermore, as for the Freysoldt approach, I would like to know if considering
> just the Hartree contribution (without including the electrostatic ionic interactions)
> in the potential alignment term is accurate enough.

The neutral atom potential method is a nice way to bypass the use of the Ewald method as discussed in
https://doi.org/10.1103/PhysRevB.72.045121
The long range Coulomb interaction is included only in the Hartree potential of the difference charge.
So, I think that the elimination of spurious long range interaction needs to be considered only for
the Hartree potential.

Regards,

TO

Re: Charged Defects Formation Energy Correction ( No.2 )

Date: 2023/11/02 21:54
Name: Marc <marc.tunica-i-rosich@universite-paris-saclay.fr>
References: Correction

Dear Professor Ozaki,

I greatly appreciate your response. Everything is now much clearer to me.

With gratitude,
Marc

Page: [1]