[QE-users] Error in piezoelectric coefficients calculation (Berry phase)

[Paolo Giannozzi]

On 04/08/2023 16:13, Zimmi Singh wrote:

> 1) In the nscf calculations, I added the 'nbnd' tag in the input file 
> (nscf1.in). However, when examining the output file 
> (nscf1.out), the polarization value was displayed as NaN

It happens only if you use many processors and many empty bands. The 
reason is quite stupid: a matrix that should be 1 on the diagonal for 
empty bands is instead equal to the number of processors. This does not 
affect the final result but may produce floating-point overflow when 
computing the determinant. It will be fixed in the next version. For the 
time being, use less empty bands (they are useless anyway for Berry's 
phase calculation)

Paolo
-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 206, 33100 Udine Italy, +39-0432-558216