ABINIT Discussion Forums

Dear there,

I tried to generate a PAW datasets with the atompaw code for potassium with the configuration of only the 4s electron (no 3p 3s electrons).

In norm-conserving psp, it is possible to add the nonlinear core-correction, is there any way to do this in PAW. Because I found my 4s-only K-PAW gives a much smaller lattice parameter for BCC K. I tried to avoid using the 3p 3s of K for PAW, since the computational cost increases rapidly with the number of electrons in my case.

thanks a lot

best
Chen

Hi Chen,
There is no need for a "nonlinear core-correction" in the PAW formalism because it is an all-electron like method, i.e. the wave functions are the all-electron ones. The only approximation you make in paw (provided that you pseudo are correct) is the freezing of the core electrons. In your case, how did you test your pseudo-potentials ? Did you check the log derivatives ? Also, do you know if it is at all possible to freeze the 3p and 3s electrons of potassium without drastically changing the chemical behaviour ?

Regards,

Emmanuel

Hello Chen,

Plot the 3s and 3p radial wavefunctions and you'll see that you will have a very hard time not considering these orbitals as valence for any reasonable rpaw value.

Kind regards,

Alain