Local potential for potential alignment in charged systems

[Udit]

Hello world,

As part of my research, I'm performing simulation of charged dopants in Silicon for calculation of the charge transition levels. One crucial step in the formation energy estimation is the determination of an alignment constant which is determined by comparing the local potential profiles of the defective and pristine supercells. https://journals.aps.org/prl/pdf/10.1103/PhysRevLett.102.016402 (Fig. 2)

My question is, for the local potential profiles is V_HARTREE_CUBE sufficient? A lot of python codes designed for such calculations do so by evaluating the LOCPOT obtained from VASP calculations. Is V_HARTREE_CUBE similar to the LOCPOT inVASP? 

Also, a lot of these publications during alignment show smoothly varying local potential profiles. However, I find that the Hartree potential profile of the defective supercell varies a lot from that of the pristine supercell and contains a lot of oscilattions mainly around the defect due to the relaxation of atomic positions, which would potentially make the alignment procedure impractical. Any comments or experiences with Charge Transition Level calculations on condensed matter systems would be greatly appreciated!

Thanks.

[Matt W]

<self promotion>

Relation between image charge and potential alignment corrections for charged defects in periodic boundary conditions

TR Durrant, ST Murphy, MB Watkins, AL Shluger

The Journal of chemical physics 149 (2), 024103 (2019)

</self promotion>

might be useful. You could contact Tom Durrant about what he has been doing. There is a load of other stuff in the literature, of course.

Matt

On Thursday, February 6, 2020 at 6:17:54 PM UTC, Travis wrote:

Hi,

To your last question - the authors report the plane-averaged potential (refer to the caption of Fig 2 in your linked paper), not the actual potential. The actual potential should be noisy as you described.

-T

[ub]

Hi Travis,

Indeed, I looked at the difference in the Hartree potentials of the defect (As +1 charged) and bulk Si 3x3x3 supercells and the result contained a lot of oscillations. I've attached an image.

Regards,

ub

On Thursday, 6 February 2020 19:17:54 UTC+1, Travis wrote:

Hi,

To your last question - the authors report the plane-averaged potential (refer to the caption of Fig 2 in your linked paper), not the actual potential. The actual potential should be noisy as you described.

-T

On Thursday, February 6, 2020 at 2:00:32 PM UTC-4, Udit wrote:

[ub]

Thanks for the article Matt. I'll go through it to see any similarities to the issues faced.

Regards,

ub

[ma455...@gmail.com]

Hi ub,

I'm wondering have you figured out the method to solve the potential alignment correction?

Regards,

Hongyang

[Nicholas Winner]

For potential alignment corrections, one should use the electrostatic potential in the V_HARTREE cube file. 

Many potential alignment correction schemes exist, but I am partial to the one by Freysoldt (https://doi.org/10.1103/PhysRevLett.102.016402) and its extension to anisotropic systems by Kumagi (https://doi.org/10.1103/PhysRevB.89.195205), which are the two I use for charged defect calculations.

Is this what you need to know, Hongyang?

-Nick

[ma455...@gmail.com]

Hi Nick,

Yes, this is very helpful. But another issue is that V_HARTREE_CUBE is only valid for QS with GPW which is mentioned in the manual... I'm wondering is there any approach to print the elctrostatic potential for ALL_ELECTRON basis sets with GAPW?

Regards,

Hongyang

[Nicholas Winner]

A developer might be able to comment more thoroughly, but I believe you have to post-process the electron density in order to do this. It is very rare to do defect calculations with all-electron calculations, so unless you have a good reason, consider switching back to GPW.

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ff485f75-3d2d-4173-a9a7-15dbf9e4d2dbn%40googlegroups.com.

[Matt W]

Would depend what you need to apply your chosen scheme. 
Pretty much by definition you can't get the potential accurately on a regular grid near all-electron cores.

[ma455...@gmail.com]

Hi Nick and Matt,

Got it! Thanks for the information!

Regards,

Hongyang

[ma455...@gmail.com]

Hi Nick,

Sorry for troubling you again. I'm currently using Freysoldt's code to perform the potential alignment correction. If I understand it correctly, we should use "sxdefectalign -q 3 --eps 12.9 --vdef v_elec_defect.cube --vref v_elec_pristine.cube --ecut 30 --qe" to perform the calculation. However, the eps parameter (dielectric constant) is needed. I'm just wondering how could we calculate the dielectric constant from DFT in cp2k?

Regards,

Hongyang

在2021年5月25日星期二 UTC+10 上午10:29:33<nwi...@berkeley.edu> 写道：