Are PAW projectors used in .spn file?

[stepan-tsirkin]

Hi @Chengcheng-Xiao

Just one question to make sure my understanding is correct. In your patch, to calculate the spn file is the full PAW wavefunction is used? Or only the plane-wave part - the one given by plane-wave coeficients in the WAVECAR file?

In turn, my utility vaspspn uses only the plane-wave part (written in the WAVECAR) and normalizes it. This is a rough approximation but it used to work qualitatively.

Here @ma455173220 reported wrong results with vaspspn, but good results with your patch. So I want to find out if it is a bug that somehow crawled into vaspspn over time, or is it the issue of using only plane-wave part.

Best,
Stepan.

[Chengcheng-Xiao]

Hello again @stepan-tsirkin!

To answer your question, my patch calculates the .spn file using the full PAW formalism, including the augmented part (the projector stuff you mentioned).

my utility vaspspn uses only the plane-wave part (written in the WAVECAR) and normalizes it.

I am actually not sure if normalizing the pseudo-wavefunction directly is a good idea. In VASP, the correct moments are recovered by adding the "compensation charges" using the multipole expansion method (see https://journals.aps.org/prb/pdf/10.1103/PhysRevB.59.1758). Directly normalizing the wavefunctions might be problematic. Regrettably, off the top of my head, I don't really have a good idea of calculating the .spn file with only the pseudo-wavefunctions and no information from the PAW dataset.

Best,
Chengcheng

[stepan-tsirkin]

Thanks @Chengcheng-Xiao ,

I am actually not sure if normalizing the pseudo-wavefunction directly is a good idea.

It is a bad idea in general. It only makes sense in the limit where a spin state is a pure spin state. Say, for some wavefunciton $\psi$ only the "spin-up" component of the spinor is non-zero., then we will get $\langle \psi| \sigma_z | \psi>==1$ both in case we use afull PAW wavefunction or a normalized pseudo-wavefunction.

But for off-diagonal matrix elements, I guess, it is strongly wrong.