BndsRnXp and NGsBlkXp

[kxzhou]

Dear all,
The calculated GW indirect bandgap shows opposite trends with respect to the parameters BndsRnXp and NGsBlkXp, which is somewhat confusing to me, as I expected them to exhibit similar trends. Should I increase both parameters simultaneously until the relative variation of the final indirect bandgap with respect to each parameter is less than 10 meV?
Best,
Star

[Daniele Varsano]

Dear Kexin,

yes, the two parameters are not independent, and both need to be converged as you are doing.

Best,

Daniele

[kxzhou]

Dear Daniele，
When performing GW calculations with the PPA (Plasmon-Pole Approximation), can I set other parameters (such as Bnds, NGsB, Gbnd) to smaller values while testing the convergence of the k-point grid, or must these parameters be fixed at their converged values?

Best，

Star

[Daniele Varsano]

Dear Kexin Zhou,

yes, k point sampling convergence is independent of the other parameters.

Best,

Daniele

[kxzhou]

Dear Daniele,

yes, k point sampling convergence is independent of the other parameters.

I tried fixing BndsRnXp, NGsBlkXp, and GbndRnge at smaller values (rather than the already converged values) and tested convergence by varying the DFT k-point grid density. As a result, when the k-point grid increased, the positive band gap even turned into a negative band gap (the system should have a positive band gap). Can reasonable convergence parameters still be obtained under these circumstances?

Another question, for systems with strong spin-orbit coupling, can the parameters obtained from convergence tests without spin-orbit coupling be used for cases with spin-orbit coupling?

Thanks!
Best,
Kexin

[Daniele Varsano]

Dear Kexin,

you can converge k point and other parameters independently, but final values should be obtained with all the parameters at convergence.
Usually yes, convergence parameter obtained without soc can be used also when adding soc.

Best,
Daniele