[Pw_forum] epsilon.x

[Paolo Giannozzi]

Emine Kucukbenli wrote:

> It is clearly stated there that "the non-local contribution due to the  
> pseudopotential is neglected. Actually the correction to the matrix  
> element that take into account the non-local part of the Hamiltonian  
> is not implemented."
> So to develop the code for USPP you need to find a way to calculate &  
> add these corrections, I guess. 

it is not sufficient. For norm-conserving PP, there is a nonlocal
contribution to the matrix elements <\psi_i| r | \psi_j> (i /= j)
that is not accounted for in epsilon.x . This is the meaning of
the above statement.

For USPP, additional terms have to be added, since only the smooth
term of |\psi_i> (the one expanded into plane waves) is available.

The correct calculation of matrix elements is actually performed
in the phonon code and could be in principle imported into
epsilon.f90.

P.
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy