Fix normalization of wavefunctions

[antoine-levitt]

@azadoks Do you agree? The wavefunctions are not in units of energy, so there should not be a factor of 2 here.

[mfherbst]

@azadoks ping 😄

[azadoks]

Ah yes, my bad the factor of two should probably not be there. However, in checking this, I found something else to consider:

• ABINIT does rescale these functions so that their integrals are correct.
• QE also has code for rescaling and checking for zero-valued pseudo-atomic wavefunctions.

[antoine-levitt]

I don't understand your comment, is the current version (with fix in) correct?

[azadoks]

I think the factor of two needs to go, but we should also probably be normalizing the wavefunctions by their norm. Abinit does chi ./= sqrt.(integrate(r, chi.^2)) where chi is the data from the file (i.e. with the factor of r included). QE does the same thing, but it also sets the occupations for any wavefunctions which have a norm < 1e-8 to a small negative number so that they are not used in the initial guess wavefunction.

[antoine-levitt]

OK, I'll open an issue