I have received several emails and comments reporting issues with the pw.x + open_grid.x + wannier90 approach for band structure calculations with hybrid XC functionals. I decided to quickly test the latest version of QE and found that all is well – at least for my simple test case.
The attached run file will let you calculate the bands of bulk MgO with hybrid functionals (I tested HSE, B3LYP and PBE0 but that choice was entirely arbitrary and does not reflect any meaningful selection). I also tested that both, norm-conserving (ONCV) and ultrasoft/PAW pseudopotentials work at this stage and found no problem.
Below you will see the bands obtained for the ONCV case – note that this is just a test of the code’s capabilities and not converged (wiggles in the bands indicate that the wannierization failed to localize the states sufficiently which might require adjustment of the wannierization window)
And we can also quickly check if the hybrid functionals do what we expect them to do – fix the “band-gap problem”. Indeed, we find that the band-gap is increased from the PZ case (~4.1 eV) to a maximum value of 7.2 eV (using PBE0) – with literature values usually falling between 5.7 and 7.7 this is not a bad match at all 🙂
As always, use these with caution and make sure your calculations are converged! A run script that tests above-mentioned cases as well as an example of the output for HSE using ONCV pseudopotentials are available for download below.