[QE-developers] rgd_blk subroutine

[Lorenzo Paulatto]

>       nr1x = int ( sqrt (geg) / &
>                    sqrt (bg (1, 1) **2 + bg (2, 1) **2 + bg (3, 1) **2) 
Hello Lorenzo,
I think that initial calculation is only used to give a very rough 
estimate of the number of points required. Later on geg is actually 
recomputed accurately.

If you want to re-implement this routine with another unit of measure, 
you can safely assume that it was written thinking of silicon, with 
alat=10.2 and 2pi/alat = 0.62¹.

It does not really matter, as long as you are going far enough the Ewald 
sum will be converged.



[1] This happens to be close to the Å to bohr conversion, and a frequent 
cause of headaches when learning Wannier functions with the silicon example.

cheers


-- 
Lorenzo Paulatto - Paris