GW PPA termination with no error

[Daniele Varsano]

Dear Shobhit,

as you can see from the report file, yambo does not recognize the grid as uniform and this is why the hf input file cannot be generated:

Code: Select all

X grid is uniform                                : no

When using the NoDiagSC keyword, the algorithm is not able to resolve the grid.

Most probably this happens because you are using a shifted k point grid.
Using a gamma-centred grid in nscf input file should solve the problem.

Code: Select all

K_POINTS {automatic}
    18 18 1 0 0 0 

Best,
Daniele

[shobhit21287]

Dear Daniele,
We reran the nscf files with diagonalisation parameters, and it worked somehow. However, the GW correction and after running the GW-PPA file is wrong. The band in our structure changes from a direct to an indirect gap. Attaching the gw_input, nscf.in and ypp input files for your reference. Also, we ran DFT Convergences for good measure and to speed up the calculations.

Note that the band structure produced by ypp without the flag
GfnQPdb= "E < ./allk_5nbd/ndb.QP"
is correct however, after the flag is added, the structure changes resulting in a new indirect bandgap.

gw_final.txt

ypp_bands.txt

nscf.txt

Best,
Shobhit

[Daniele Varsano]

Dear Shobhit,

I'm not an expert on MoSe2, but I suspect that it's happening something similar to MoS2. The minimum of the two valleys in conduction are nearly degenerates. Small details (e.g. pseudopotentials) in the GW calculation can switch the character from direct to indirect.
In the case of MoS2, for instance Giustino et al. reports that it is needed to include the effect of the substrate to obtain the direct nature in GW approximation: (PHYSICAL REVIEW B 103, 125401 (2021)).

Anyway, I suggest you to check carefully your converge parameter and explore the role of pseudopotentials.

Best,

Daniele

[shobhit21287]

Dear Daniele,
We did what you said and changed the pseudopotentials from GGA to LDA. However, we now face a new issue in the GW calculations. Upon running YPP for band interpolation, the band gap changes from direct to indirect. I am attaching the GW input and YPP input for your reference. We have faced these issues in 2 novel materials now. I would also like to point out that converging the DFT parameters for MoS2 with GGA pseudopotential fixed this indirect gap but not with MoSe2.

ypp_input_MoSSe_newrelaxed.txt

gw_input_MoSSe_newrelaxed.txt

Best,
Shobhit

[Daniele Varsano]

Dear Shobhit,

the interpolation procedure can be inaccurate. In any case, I suggest you to use the BOLTZ procedure instead of NN

Code: Select all

INTERP_mode= "BOLTZ" 

About direct/indirect character issue, this is a known issue, the direct and indirect gaps are nearly degenerate (see e.g. https://cmrdb.fysik.dtu.dk/c2db/row/MoSe2-f61b14d398c7) and the absolute value is very sensitive to small variation of the lattice parameter and calculation details.

Best,
Daniele