Dear,
I'd like to calculate the optical properties of a crystal structure using the Quantum Espresso package. Most of the tutorials I've seen used the epsilon.x function. However, I read a comment that epsilon.x only works with NC pseudopotentials.
At the beginning of this work, I was using the ONCV NC PPs from the SG15 database and PseudoDojo (PBE), but both PPs gave me lattice parameters shorter than the crystal values, after optimization. These values were
Dear Jose,
1) You can in principle use the experimental lattice parameters and optimise only the atomic positions (not the lattice constants). This might provide a more accurate description, and a few people would recommend this approach rather than a full unit cell optimisation.
2) Quantum Espresso used to have a quite good page for pseudopotentials. But the Pseudodojo is a viable alternative and offer a quite efficient implementation.
3) You can try with Yambo. Run the QE calculation, a QE nscf calculation, and afterwards doing an Independent Particle calculation with Yambo should provide the same results as the epsilon.x with the NC pseudopotentials. The calculation is quite fast but obviously it requires you have Yambo compiled on your machine a possibly difficult task. However, it gives you some control over Yambo and thus the possibility of improving your calculation for example by studying excitons.
I hope this helps.
Best regards,
Roberto
Dear Roberto D'Agosta
Thank you for your answer.
1) I'll try it.
2) Yes, on the QE page there are a number of PPs. However, in the part where the NC PPs are shown, there is a warning (not recommended, kept for reference). So, I went to PseudoDojo.
3) I'll read about Yambo. Does it require a new PP file?
Is it possible that I'm doing something wrong during the vc-relax calculation with PseudoDojo PPs? Despite the small difference, I've never seen the lattice parameters smaller than experimental after GGA optimization.
A colleague performed the same calculation using CASTEP (with normal NC PPs) and the parameters were larger than the experimental. So, I tried to change all of the threshold values (ecutwfc, ecutrho, conv, forc), but I only observed a better value after increasing the difference between ecutwfc-ecutrho, but it was not enough to show larger values after vc-relax.
- Badges
- Science topic