Convergence Advice on MoS2

[jmullen]

Greetings,

I am experimenting again with YAMBO and sharpening my skills using MoS2. However, I have run into a general problem that I am hoping people here who have worked with these systems can help me with. First, system calculation details

Initial calculations with Quantum Espresso
Cells optimized, atomic positions relaxed
scf and bands calculations are run with LDA and GGA pseudopotentials
calculations performed with non relativistic (and at least one relativisitic)
calculations performed on 18x18, 27x27, 36x36 grid (nscf) with 300 bands

Band structures conform to expected values for all of the calculations with Quantum Espresso. For each of the systems, I then follow the steps

p2y
yambo -i -V RL
yambo ( I accept the default max G vecs before this run)
yambo -x -k hartree -p p -g n (I accept the defaults for all variables but for the ones listed below)

Then I start studying the variables for convergence, and this is where I get stuck. The basic problem I am having is that K is not the lowest conduction band value in the complete list of QP's. I am working with the following variables for convergence

FFTGvecs
EXXRLVcs
BndsRnXp
NGsBlkXp
GbndRnge

For the most part, I have found the bands variables should be kept at the maximum number of bands which were run in the NSCF calculation and the NGsBlkXp seems to converge between 10 and 20. As for the FFTGVecs and the EXXRLVcs variables, the energy gap at K converges to published values very quickly (less than 1000 for both variables), but this problem of conduction band energies lower than the K point conduction band energy does not go away even for variable values of > 5000.

What I am looking for is guidance about the space I am probing for convergence before I consume large amount of computational time by moving into 10000+ G vecs. I have followed the tutorials on the subject and tried to follow advice on similar matters I have found in this forum, but nothing so far has worked.

Thanks
Jeff Mullen
NCSU

[Palummo Maurizia]

Hi Jeff

to my experience on MoS2 reasonable convregence parameters (using a NC pseudo with ecut = 60 Ry at DFT level) are

FFTGvecs =50 Ry
EXXRLVcs = 50 Ry
BndsRnXp = > 200-300 ( if you are interested only to gaps corrections and not in absolute corrections to each state, if this is the case I think you are obliged to use much more bands and also size of dielectric matrix!
NGsBlkXp = 10 Ry
GbndRnge = generally I used the same as BndsRnXp but again this is ok if you wish to have an estimate of only the QP corrections to the gaps / transitions and not to the energy of each state.

Regarding the direct/indirect gap at K of MoS2 I know this is very critical even at DFT level! How is yours at DFT level?
In any case what it is important is to use Mo pseudos with sp in the valence if not the exchange part of the self-energy comes out wrong and the dispersion of the QP bands could be wrong.

Best

Maurizia

[jmullen]

Thanks for the advice.

I have indeed converged the gaps. I had not anticipated that converging the gaps would be done and the actual conduction bands would be so far off. I am still not sure of why/how that is so, but I will study the issue further.

Cheers
Jeff

[Codest]

Thanks for the advice.

I have indeed made the best out of PhenQ and converged the gaps. I had not anticipated that converging the gaps would be done and the actual conduction bands would be so far off. I am still not sure of why/how that is so, but I will study the issue further.

Cheers
Jeff

Why is it that whenever I converge the gaps they revert back to before later on. My saving isn't working.

[Daniele Varsano]

Dear codest
Please fill your signature with your name and affiliation. It is a rule of the forum.
Next can you explain more explicitely what problem are you encountering?
Best
Daniele