Re: [Wannier] Recreating radial wavefunctions from pw2wannier90 or wannier90 outputs

2020-11-03 Thread Stepan Tsirkin 

Dear  Erica, dear Lorenzo,

just to add to what Lorenzo said:

 * as far as I know mostly the PAW coded write the pseudo-wavefunction
   into the UNK files, therefore the number of nodes cannot be
   reproduced in Wannier functions. At least it is true for  VASP and
   QuantumEspresso. The reasons are twofold: (i) noone bothered to code
   plotting all-electron wavefunction, (ii) the all-electron
   wavefunction strongly oscillates near the core, therefore a very
   dense real-space grid is needed to represent it.
 * The Wannier function may have additional nodes far from the the
   center (usually near the neighboring atoms, or even further) but
   those are not related to the nodes of atomic wavefunctions.
 * According to the page http://www.wannier.org/features/ , Wannier90
   is interfaced with Fleur and Wien2k, but I am not familiar with
   those, I do not know if/how they write the UNK files.
 * I fully agree with Lorenzo's conclusion:

Finally, if all you need is a pretty picture, I don't think it is
worth it. If you need the correct all-electron wavefunction close
to the core, I'm not sure wannierization is the best method.


Regards,

Stepan


On 03.11.20 10:41, Lorenzo Paulatto wrote:



On 11/3/20 7:55 AM, Erica Kotta wrote:
The plots in xcrysden I've come across only show the lowest-level 
orbitals for each l-value (for example only 2p orbital character, 
even when you know that the bands of interest are 3p or 4p). I 
thought that by getting the underlying xsf numerical data or 
recreating the plots from the Umn/UNK files, I would be able to 
resolve the "missing" nodes.

Hello Erica,
this is not a limitation of the wannierization method, but of the 
pseudo-potential approximation. The Wannier90 only works with what it 
gets from the total energy code.


I don't know if Wannier90 can interface with any all-electron code

In principle if you use PAW pseudopotential, or with a GIPAW-like 
method, it is possible to reconstructs the all-electron wavefunctions 
from a pseudopotential calculation, but I do not think it has ever 
been done in conjunction with wannierization.


Finally, if all you need is a pretty picture, I don't think it is 
worth it. If you need the correct all-electron wavefunction close to 
the core, I'm not sure wannierization is the best method.


cheers


___
Wannier mailing list
Wannier@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/wannier

Re: [Wannier] Recreating radial wavefunctions from pw2wannier90 or wannier90 outputs

2020-11-03 Thread Lorenzo Paulatto 




On 11/3/20 7:55 AM, Erica Kotta wrote:
The plots in xcrysden I've 
come across only show the lowest-level orbitals for each l-value (for 
example only 2p orbital character, even when you know that the bands of 
interest are 3p or 4p). I thought that by getting the underlying xsf 
numerical data or recreating the plots from the Umn/UNK files, I would 
be able to resolve the "missing" nodes.

Hello Erica,
this is not a limitation of the wannierization method, but of the 
pseudo-potential approximation. The Wannier90 only works with what it 
gets from the total energy code.


I don't know if Wannier90 can interface with any all-electron code

In principle if you use PAW pseudopotential, or with a GIPAW-like 
method, it is possible to reconstructs the all-electron wavefunctions 
from a pseudopotential calculation, but I do not think it has ever been 
done in conjunction with wannierization.


Finally, if all you need is a pretty picture, I don't think it is worth 
it. If you need the correct all-electron wavefunction close to the core, 
I'm not sure wannierization is the best method.


cheers


--
Lorenzo Paulatto - Paris
___
Wannier mailing list
Wannier@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/wannier

Re: [Wannier] Recreating radial wavefunctions from pw2wannier90 or wannier90 outputs

2020-10-29 Thread Giovanni Pizzi 
Dear Erica,
I am not sure if I'm misunderstanding your question, anyway here are some 
answers based on what I think you might need:

- the initial projections are generated by the interface code (e.g. 
pw2wannier90) so you should try to have them printed there. Anyway, their 
analytical form is tabulated in the user guide so you should be able to create 
their numerical representation on a grid yourself.

- Those, anyway, are just the starting point of the Wannierisation. The code 
computes the WFs and then uses the minimised U matrices (unitary transforms), 
and the UNK files from the ab-initio code (periodic part of the Kohn-Sham 
wavefunctions) to plot the final WFs in real space (there is a flag in the 
input to have W90 print the xsf files, or also in cube format). The flags, and 
the detailed specification of the format of the various files, is explained in 
detail in the Wannier90 user guide.

- if what you want is just the numerical data in the XSF (so, the Wannier 
functions in real space), you can just open and parse the XSF file, the numbers 
are in there (the XSF format is well documented, just google for 'xsf 
specification'). There might even be codes around that already parse it, 
probably.

Best,
Giovanni

P.S.: remember to sign your emails with your affiliation

--
Giovanni Pizzi
Theory and Simulation of Materials and MARVEL, EPFL
http://people.epfl.ch/giovanni.pizzi
http://nccr-marvel.ch/en/people/profile/giovanni-pizzi

On 28 Oct 2020, at 19:40, Erica Kotta mailto:eck...@nyu.edu>> 
wrote:

Hello,

I have a Wannier90 output where I have projected a group of isolated bands onto 
a set of p- and d-orbitals.

Is it possible to see the radial wave functions used by the Wannier90 software? 
I see in the User Guide tables 3.1-3.3 they list the angular functions and then 
a possible choice of radial functions, but is there a way to see the actual 
form that was used, that represents the actual basis? Does it entail outputting 
the Umn matrices, and using the UNK...NC to recreate them as in Equations 1&2 
in Wu's paper 
(https://www.labxing.com/files/lab_data/496-1580695056-bV5W1jk8.pdf)?
(And if this is the case, how do you open/read these UNK...NC files?)

Basically, I was wondering if I could get some set of numerical data 
representing what we see when we plot the xsf files in xcrysden.

Thank you very much in advance,
Erica
___
Wannier mailing list
Wannier@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/wannier

___
Wannier mailing list
Wannier@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/wannier