[Pw_forum] XSpectra and US or PAW pseudos?

[Adrian Hühn]

Hello Lorenzo,

this is an extremely useful feature indeed. I am also a bit confused 
that PAW is not compatible, because within the PAW formalism we should 
get back to a AE wavefunction. I read the basic papers of both the PAW 
and the US formalism and see, that they are quiet similar (i.e. both 
sorts of PPs in the PSlibrary share the same generation parameters). I 
also read somewhere that both ansatzes are in principle the same while 
the algorithms differ.
Besides Matteo said XSpectra is not compatible with PAW so. And he was 
involved in the implementation of the Lanczos algorithm with USPPs (PRB 
80, 075102 (2009)). I will try to do some small test calculations, for 
comparison.

Concerning the reference state: for instance this is the input for a Mg 
speudopotential (PP) from the PSlibrary (Mg.pbe-n-kjpaw_psl.0.3.0.upf, 
the one I used for my tests)

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

  &input
    title='Mg c',
    zed=12.0,
    rel=1,
    config='[Ne] 3s2.0 3p0 3d-2',
    iswitch=3,
    dft='PBE'
  /
  &inputp
    lpaw=.true.,
    pseudotype=3,
    file_pseudopw='Mg.pbe-n-kjpaw_psl.0.3.0.UPF',
    author='ADC',
    lloc=2,
    which_augfun='PSQ',
    rmatch_augfun=1.9,
    nlcc=.true.,
    new_core_ps=.true.,
    rcore=2.1,
    tm=.true.
    lgipaw_reconstruction=.true.
    use_paw_as_gipaw=.true.
  /
5
3S  1  0  2.00  0.00  2.30  2.50  0.0
3S  2  0  0.00  6.50  1.40  2.50  0.0
3P  2  1  0.00  0.00  2.40  2.60  0.0
3P  3  1  0.00  6.30  1.40  2.60  0.0
3D  3  2 -2.00  0.30  2.30  2.30  0.0

-----------------------------
Its reference state is the atomic ground state. GIPAW projectors are the 
same as PAW (?).To create the core-hole PP one should change the config 
tag to

---
    config='1s1 2s2 2p6 3s2.0 3p0 3d-2',
---

and I am done.

What about USPP? The associated Mg USPP looks like the same:
-----------------------------
  &input
    title='Mg c',
    zed=12.0,
    rel=1,
    config='[Ne] 3s2.0 3p0 3d-2',
    iswitch=3,
    dft='PBE'
  /
  &inputp
    pseudotype=3,
    file_pseudopw='Mg.pbe-n-rrkjus_psl.0.3.0.UPF',
    author='ADC',
    lloc=2,
    which_augfun='PSQ',
    rmatch_augfun=1.9,
    nlcc=.true.,
    new_core_ps=.true.,
    rcore=2.1,
    tm=.true.
  /
5
3S  1  0  2.00  0.00  2.30  2.50  0.0
3S  2  0  0.00  6.50  1.40  2.50  0.0
3P  2  1  0.00  0.00  2.40  2.60  0.0
3P  3  1  0.00  6.30  1.40  2.60  0.0
3D  3  2 -2.00  0.30  2.30  2.30  0.0
-----------------------------
Creating the core-hole should trivially apply here as well. But how do I 
consistently add GIPAW information? Do I take simply the same radii? How 
does the syntax look like?

For the normconserving case it is nicely described in the tutorial so no 
questions here :-).

A further question: In the tutorials it was suggested to ionize the 
reference state a bit. Is this crucial?

Regards
Adrian

Am 31.08.2015 um 12:53 schrieb Lorenzo Paulatto:
> On Monday, August 31, 2015 11:49:28 AM Adrian Hühn wrote:
>> How do I correctly include a reference state to a pseudopotential from
>> the PSlibrary?
>>
> At the beginning of the .UPF file you normally find a copy of the input used
> to generate it, you can then regenerate the pseudo adding reference
> information for GIPAW (ask further if you need help about this).
>
> About PAW dataset not being compatible, I think they can be if you generate
> them with the use_paw_as_gipaw=.true. keyword, but I leave the final word to
> experts or to testing .
>
> regards
>