[Pw_forum] Berry phase electric field calculations with empty bands

[Alexander Urban]

Dear Quantum Espresso users,

is it possible to perform Berry phase electric field calculations for a 
periodic insulator including empty bands (1) in principle and (2) using 
PWSCF?

I know how to do Berry phase electric field calculations with PWSCF 
(using the `lelfield' keyword) and I also verified the polarizations of 
a number of isolated systems with the results of the sawtooth potential 
(using the `tefield' keyword).  But when I include unoccupied bands, by 
setting `nbnd' to a value larger than nelec/2, the polarization in the 
output does not look reasonable anymore and does not converge.  However, 
the SCF loop converges and also the eigenvalues (i.e. the band 
structures) look fine.

In the PWSCF source code (QE version 4.3.2), namely in 
`c_phase_field.f90', I found the following lines of code:

!  --- Check that we are working with an insulator with no empty bands ---
    IF ((degauss > 0.01d0) .OR. (nbnd /= nelec/2)) &
         WRITE (stdout,*) 'PAY ATTENTION: EL FIELD AND OCCUPATIONS'
    !  CALL errore('c_phase', &
    !        'Polarization only for insulators and no empty bands',1)

The call to the `errore' subroutine is commented out, how do I have to 
interpret this?  Does not the electric polarization only depend on the 
occupied bands, anyway?  Can I trust the SCF wavefunctions?

Sorry for the long e-mail,

Alex


----------------------

Alexander Urban (PhD student)
Interdisciplinary Center for Molecular Materials (ICMM)
Computer Chemistry Center (CCC)
University of Erlangen-Nuremberg
Germany