Partial waves and projectors in AtomPAW  [SOLVED]

External optional components of Abinit (BigDFT, Wannier90...) and codes using/providing data from/to Abinit (AtomPAW, ONCVPSP, EXC, DP, Yambo...)

Moderators: ebousquet, bxu

Locked
elchatz
Posts: 15
Joined: Wed May 03, 2017 12:17 pm

Partial waves and projectors in AtomPAW

Post by elchatz » Wed Oct 25, 2017 11:06 am

Hello all,

I am using the abinit version of AtomPAW to construct pseudopotentials for two elements: Mo and S.

I am using the data files from atompaw website for now because I had some problems with those of abinit, but I don't think that is relevant for now.

In the input file I have declared 5 valence orbitals and 2 partial waves for Mo (7 in total) and 3 valence orbitals with 4 partial waves for S (again 7).

When I run the calculation, I get 11 files for each run (Mo and S). The files are numbered wfni where i=1...11.

However, in both cases I get 7 files that are named wfn00X where X=1..7.

Does anyone know which files I should be looking at for examining the partial waves and projectors and how do I know what they represent?

Thank you

elchatz
Posts: 15
Joined: Wed May 03, 2017 12:17 pm

Re: Partial waves and projectors in AtomPAW

Post by elchatz » Thu Nov 02, 2017 10:05 am

I have found the following in the atompaw source code in

https://github.com/qsnake/atompaw/blob/ ... pseudo.f90

-------------------------------------------------------------------------------------
WRITE(6,*) 'Found ', nbase,' valence basis functions '
WRITE(6,*) 'Allocating for ', mxbase, ' total basis functions'
-------------------------------------------------------------------------------------

And then in the output file:

-------------------------------------------------------------------------------------

Found 5 valence basis functions
Allocating for 15 total basis functions
basis functions:
No. n l energy occ
l,nbase,node 0 1 3
1 4 0 -4.735126E+00 2.000000E+00
l,nbase,node 0 2 4
2 5 0 -3.176611E-01 1.000000E+00
For l = 0 there are currently 2 basis functions
enter y to add additional functions or n to go to next l
l,nbase,node 1 3 2
3 4 1 -2.836301E+00 6.000000E+00
For l = 1 there are currently 1 basis functions
enter y to add additional functions or n to go to next l
enter energy for generalized function
l,nbase,node 1 4 3
4 999 1 3.000000E+00 0.000000E+00
For l = 1 there are currently 2 basis functions
enter y to add additional functions or n to go to next l
l,nbase,node 2 5 0
5 3 2 -1.626568E+01 1.000000E+01
l,nbase,node 2 6 1
6 4 2 -2.890487E-01 5.000000E+00
For l = 2 there are currently 2 basis functions
enter y to add additional functions or n to go to next l
enter energy for generalized function
l,nbase,node 2 7 2
7 999 2 6.000000E+00 0.000000E+00
For l = 2 there are currently 3 basis functions
enter y to add additional functions or n to go to next l
completed phi basis with 7 functions
-----------------------------------------------------------------------------------------------

I still don't understand which files I should be looking at.

elchatz
Posts: 15
Joined: Wed May 03, 2017 12:17 pm

Re: Partial waves and projectors in AtomPAW  [SOLVED]

Post by elchatz » Fri Nov 03, 2017 5:29 pm

This is the reply I got from Natalie Holzwarth:

-------------------------------------------------------------------------------------

There are some outputs that were put in for diagnostics that are typically not used/needed. The outputs that are useful are wfn1, wfn2..... wfn7 assuming that you have 7 basis/projector functions. Each of these contain 4 columns: r (in Bohr) \phiAE \phiPS \p (projector) for given basis channel. These functions will be used by abinit or Quantum Espresso. The program also output additional wfn files (in your case wfn8 ....wfn11). These are generated in the subroutine logderiv which is the the pseudo.F90 module. They give some information on how well the PAW representation is working at a particular energy.

The logderiv.l files are also generated by the logderiv subroutine. That program calculates the logderiv's for all l values up to lmax+1 with the thought that the local potential should control the scattering behavior of all partial waves beyond those represented by the projector and basis functions. If you use the norm-conserving construction method, this is usually done quite well.

Locked