Band gap error for monolayer MoS2

[Tianshu]

Dear developers,

I'm trying to calculate the GW correction for monolayer MoS2 with yambo. Monolayer MoS2 should have a direct band gap. The calculated results show that PBE band gap is correct, which is a direct gap. However, after adding GW correction, the band gap becomes indirect. I adjusted the parameters and test the convergences, but get nothing helpful. I upload the input file and some related output, including report file, I hope you can help me for this issue. Thank you very much.

Best regards,
Tianshu Li

[Daniele Varsano]

Dear Tianshu Li,
I'm not an expert on this system, and maybe someone familiar with MoS2 monolayer can help you in detail.
Anyway, for what I can see from your files you are not cutting the interaction along the z-direction, so resulting in a layered system.
If I'm not wrong bulk MoS2 has an indirect gap. You have large space among layer, anyway, it is possible it is not enough to really isolate the monolayer.
If you look at the echo of the input file in the report and output file you will see:

Code: Select all

| CUTGeo= "none"                 # [CUT] Coulomb Cutoff geometry: box/cylinder/sphere/ws X/Y/Z/XY..
# | % CUTBox
# |  0.00     | 0.00     | 0.00     |        # [CUT] [au] Box sides
# | %
# | CUTRadius= 0.000000            # [CUT] [au] Sphere/Cylinder radius

I can see that in the input file you have included the cutoff coulomb correction but for some reason I do ignore, it has been discarded by the code.
Sometimes some hidden character, it can happen when copy paste files, but just a supposition.
The syntax seems to me to be correct, try to regenerate the input file and see if it works. In any case, I suggest you to assign a value just slighter smaller than your cell in the z-axis, something like:

CUTGeo= "box z" # [CUT] Coulomb Cutoff geometry: box/cylinder/sphere/ws X/Y/Z/XY..
% CUTBox
0.00 | 0.00 | 38.00 | # [CUT] [au] Box sides
%

and check in the report if the cutoff coulomb potential has been calculated, and the ndb.cutoff database is created.
I suggest you repeat the calculation in a separate directory or redirect the calculation where previous databases ndb.pp* ndb.HF* ndb.QP are not present otherwise yambo will read them and not recalculate.

Best,
Daniele

[Daniele Varsano]

Dear Tianshu Li,
I've shown your input to a researcher who calculated MoS2 monolayer in the past and she point out possible source of errors:

1) you have 18 valence bands: usually accurate calculations are done using 26 valence bands, i.e. all spd of Mo included in the valence, so check your pseudo and think about including semicore electrons.
2) You are including non linear core corrections, generally it is safer do not include them. Moreover, in the version of yambo you are using the cutoff energy for the <Vxc> matrix elements is fixed according to EXXRLvcs.
In some case for PBE functional this could be not enough and should be raised. You can make a test considering higher values and check in the report if the <DFT> value is converged.
I suggest you to do all the tests by calculating few QP corrections,
e.g. the gap at K.
%QPkrange # [GW] QP generalized Kpoint/Band indices
61| 61| 18| 19|
%
this will save you a lot of time.

3) You use 60 bands in the summation, usually about 200 bands are needed to get converged results.

In general, unless you are not interested in deep bands I suggest you to calculate the QP corrections for few states around the gap.

Best,
Daniele