Dear QE users,
I want to ask you about the refractive indices calculation of material on Quantum Espresso. I performed the SCF calculation. Then I extracted the real and imaginary parts of the dielectric function. But I obtained for some regions of energy, different values of dielectric tensor (i.e. epsilon(11) and epsilon (22) are different) that are supposed to be equal since I am investigating a uniaxial material. Is that normal on Quantum Espresso? If it so, should I use the average in each energy point to calculate the refractive index using its formula?
Any help will be appreciated.
Best regards.
L. Ait lamine.
Dear Lahcen,
I had the same issue and once I've run a nscf calculation with a very high k-mesh and band number (in my case 2 * number of valence bands) before calculating the dielectric functin using epsilon.x, all values for epsilon (eps_x ,eps_y, eps_z) where the same except numerical noise.
Best,
Simon
Dear Simon,
thank you for these information. I appreciate it, However, is it mandatory to perform NSCF calculation to extract the optical properties? or just SCF calculation that should be performed?
Since optical properties require a much denser k-mesh in the BZ and epsilon.x can't use symmetry operations to reduce the BZ further, you need to increase the k-mesh really much to sample the while BZ. So you can try to do the following:
- first doing a scf calculation with reasonble small k-mesh to obtain the ground state
- secondly do a nscf with a very dense k-mesh and more bands to sample the BZ (set nosym = .true., noinv = .true. to sample the whole BZ and not use symmetry operations)
- finally run epsilon.x with a large nw value
I am no expert on QE but that worked for me, even this can take quite some computing time so maybe there is a more convienent way to do it.
PS: I think epsilon.x can be run after the scf calculation as well if you provide the pw.x scf input with a large k-mesh throughout the whole BZ without using symmetry, but that would take even more computing time I belive.
Dear Simon,
Thank you very much for your explanation, I do appreciate it.
Sometimes we get first three negative frequencies which can be attributed to three funny modes arising because of some numerical noise in our calculations.
Applying the Acoustic sum rule (ASR) corrections in phonon calculations are helpful in such cases.
- Badges
- Science topic
- Similar topics
- Chemistry
- Nanotechnology
- Nanostructures
- Thin Films