Symmetry-related errors in GW corrections

[palful]

Dear developers,

I think I found a small problem in Yambo concerning symmetries.

I'm studying bilayer hexagonal boron nitride - a system with inversion symmetry. Its input geometry in the supercell can be given in two different ways when we look at the z-coordinate of layer A and layer B:

i)
B: ++++++++++ +c
------------------------- 0.
A: ++++++++++ -c


ii)
B: ++++++++++ +2c

A: ++++++++++ 0.

In case i) there are no problems. In case ii) and analogues, Quantum Espresso disregards the inversion symmetry since I'm introducing a mismatch with the FFT grid, giving the following message:

6 Sym. Ops. (no inversion) found
(note: 6 additional sym.ops. were found but ignored
their fractional translations are incommensurate with FFT grid)

In both cases the QE run is completed with the correct final results. Now, I want to obtain the QP corrections to the band structure. When I run Yambo on top of the DFT results, of course in case ii) the inversion symmetry is not taken into account by the code. What I would expect now is that the calculation would take twice the time it takes in case i), but still give the correct results. What I find instead is that, for some k-points (e.g. K in the hexagonal Brillouin zone), the GW corrections are wrong.

I attach input geometries for QE, report and output files from Yambo for both i) and ii) (labeled INV and NO_INV respectively), as well as the yambo.in, l_dbs and LOG_list files.

Cheers,
Fulvio

[Daniele Varsano]

Dear Fulvio,
thanks for reporting, we will have a close look to understand the reason of such a big discrepancy.

Best,
Daniele

[Davide Sangalli]

Dear Fulvio,
I did some checks using you input files.
The problem is related to the use of the terminator, which is indeed a new and very experimental feature of yambo.

Without the terminator everything should work fine.

Best,
D.