My Community

Hello there,

I would like to make a few tests regarding the inclusion/removal of the 3d states of Zn, and how that affects the electronic band structure of some Zn-based semiconductors. I have been looking for a Zn PAW-PBE pot car file with a 4s2 valence, but all I could find is a 3d10 4s2 species (or even harder). I
understand why the simpler 4s2 version is not shipped in the official bundle, but I was wondering if you could provide it to me anyways at my own
risk?

Many thanks in advance, with kind regards,

José

Dear José,

As you can imagine, we get many requests for different types of pseudo-potentials from different users, but not all of them can be quickly and easily generated or very useful for most people.
It looks that such a pseudo-potential would not yield correct ground-state properties, since the 4s and 3d states are close in energy and will likely interact and hybridize.
But I am not the expert on this area. I will ask and get back to you ASAP.

Kind regards

Dear Henrique,

Thanks for your reply.

Agree. 4s and 3d are rather close in energy. However, they are also well displaced in space. The expectation values for their respective radii are 0.9 and 2.9 Bohr for 3d and 4s. One example where I am expecting a Zn(4s2) pseudo to work reasonably well is for a substitutional Zn impurity in crystalline Si which acts as a double acceptor (grabs two electrons from the Si host to form a total of four bonds). The 3d electrons form a closed shell and should not participate on the valence (bonding with neighbouring Si atoms).

With kind regards,
/J

While discussing this question with some of my colleagues, we were wondering what is the reason you would like to have this pseudopotential.

Is it because of having less electrons to describe in the valence and having the calculations run faster?
In this case, I would say that generating and minimally testing such pseudopotential would likely be much more work intensive than just having the 3d states in the valence.

If the reason is that the 3d states don't have the correct energy because of the inherent limitations of DFT then perhaps you could remedy this by including an U term.

Or is there another argument why such a pseudopotential might be useful in general?

Dear Henrique,

Thanks for you reply. Indeed I are planning to investigate a few Zn-based materials and wonder how likely are we to avoid having the 3d shell in the valence. Several people have been showing that for instance Ga(4s24p1) is able to describe well defects in GaN (see for instance APL-109-162107(2016), doi:10.1063/1.4964831). The overlap between the Ga(3d) with the outer shells should be comparable to that of Zn(3d). I understand that developing a new pseudo may be difficult and that is fine. Honestly I though that you could have one "on the shelf". Thanks anyway for considering it. We will find some other way to sort it out.

Thanks a lot,

/J

I see.
I think the reason you can treat the 3d states of Ga in the core is because they are well separated from the 4s (about 10 eV separation).
So for Ga you find two versions of the POTCAR, one with the 3d states in the valence (Ga_d) and one without (Ga).
In Zn, they are much closer in energy (about 4 eV separation). We only provide a POTCAR Zn with 3d states in the valence.
I have never seen a pseudo for Zn with only 4s in the valence (that is not to say that it does not exist or is impossible to create).

Perhaps when you have a Zn defect in crystalline Si the 3d bonds are well separated from the rest (I don't know) in which case such a pseudopotential could work, but then it might not be transferable to other cases.

I will ask if it's possible and practical to generate such a pseudopotential and get back to you.

I have e-mailed you directly about this request.