[Pw_forum] Dear all, where can I get the detailed instructions to get the right pseudopotentials?

[Lorenzo Paulatto]

In data 12 giugno 2009 alle ore 00:12:09, 文沈 <wshen02 at gmail.com> ha  
scritto:
> I find the pseudopotential I get never provide the correct result.
> anybody who have the experience to get the pseudopotential please tell me
> where I can get the detailed instruction on making pseudopotential?

Dear Wen Shen,
*after* reading the notes from Paolo Giannozzi you have to study the  
problem in a systematic way.

The starting point is the periodic table of elements, it is not called  
"periodic" by chance. Elements tend to have similar properties to the ones  
above and under them, from a pseudopotential generation point of view also  
elements close to the right or to the left are "similar." You can pick the  
example file from atomic_doc/pseudo_library and adapt them to similar  
elements.

Notice how the cutoff radiuses are related to the physical radiuses of the  
orbitals: <r>, <r^2>, r_max, which are all printed during the all-electron  
calculation.

Be careful about the syntax: ld1.x is very picky, try to not use more  
variables than are actually necessary.

If you are trying to do an ultrasoft pseudopotential start with a signle  
projector per channel, than look at the logarithmic derivative (good  
range: -10:+10Ry) and add the second projectors (one at a time) trying to  
move the inevitable ghost-state as high in energy as possible.

Do systematic generations, e.g. write a small script that generates a  
dozen of different pseudopotentials with all parameters equal but one, and  
plot their logarithmic derivatives. Look at all the derivatives: which  
value will be better for the parameter.

Finally, keep in mind that light elements are easier, and elements with d  
orbitals are harder, elements from the first 3 or 4 columns will likely  
require one or two semi-core states to be include in valence. Lanthanides  
and actinides can take one week of work, especially if you have little  
experience.

That's about it, if you have a specific input don't be afraid to ask for  
help.

regards


-- 
Lorenzo Paulatto
SISSA  &  DEMOCRITOS (Trieste)
phone: +39 040 3787 511
skype: paulatz
www:   http://people.sissa.it/~paulatto/

     *** save italian brains ***
  http://saveitalianbrains.wordpress.com/