[Pw_forum] use of Virtual.x

[Jérémie Teyssier]

To people doing virtual crystal approximation

 

These last weeks, I tried to figure out whether the modified version of
virtual.x was leading to reasonable pseudopotentials for VCA.

Modified means simply that I removed the error message


I have build 2 mixable pseudopotentials; same functional, same rinners, same
number nature of wavefunctions and projectors.

Then I mixed it uo using Virtual.x

Apparently it works. You get an output pseudopotential with reasonable shape
but there are many things wrong.

-          I compared the results of calculation with that potential with
super-cell calculation with the same original potentials and the result if
not completely different reveals a huge overestimate of the exchange
splitting (I am calculating magnetic structures.)  in the case of VCA.

-          More worrying, I mixed Co and Mn pseudopotentials in a 50/50
ratio and the results (again mainly the magnetisation) is completely
different (40%) between 0.5xCo+0.5xMn and 0.5xMn+0.5xCo


-          Another trouble when I mixed Co and Mn with 100% of one of the
two compounds. Then a mix of 100% Of Co (or Mn) is not equivalent to pure Co
(or Mn) pseudo.

I would say with my small experience that virtual.x is on the way to work
but if you use it, do some simple tests before even if you thing that the
resulting potential looks correct. Surprisingly, I got reasonable results
mixing Fe and Co pseudo from pwscf library.

As a conclusion and I am now trying to compare the two ways, for people
doing VCA, I suggest building a new pseudopotential directly from Vanderbild
code. Then, even if you get weird results, you have an access to all
parameter of the pseudo generation


If people have another experience on VCA with pwscf, please contact me.

 

Jérémie

 

Jeremie Teyssier

DPMC

24 quai Ernest Ensermet

CH-1211 Geneve 4

SUISSE

tel: (41) 22 379 35 68

fax: (41) 22 379 68 69

web:  <http://optics.unige.ch/jeremie/home_jeremie.html>
http://optics.unige.ch/jeremie/home_jeremie.html

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20071211/f6578d1b/attachment.html>