[QE-developers] rgd_blk subroutine

[Lorenzo Monacelli]

Dear QE developers.

I'm trying to implement the effective charge interpolation of the phonon 
dynamical matrix by using the rgd_blk subroutine (rigid.f90 file in the 
PH source)

However I noticed something strange.

This subroutine computes automatically the minimal mesh of G vectors in 
the reciprocal space to converge the Ewald sum only using the reciprocal 
space expression.

However, this estimation is based on the following equation:


      nr1x = int ( sqrt (geg) / &
                   sqrt (bg (1, 1) **2 + bg (2, 1) **2 + bg (3, 1) **2) 
) + 1


And geg is determined as:

   gmax= 14.d0
   alph= 1.0d0
   geg = gmax*alph*4.0d0

Unfortunately, while geg is independent on the unit of measure, bg is in 
2pi / alat. This means that if I calculate the same structure by 
choosing a different alat units (that is arbitrary), I will get a 
different mesh. I believe there is a mistake here (a bug?) or maybe I am 
missing something?

Bests,

Lorenzo