Reflectivity calculations

[jess_white]

Hi,

I am currently running calculations to find the reflectivity of monazite which I know has to follow four stages from a tutorial I have found. Unfortunately though for the last step even after 48 hours the results are not completed as ran out of time. I have to use the high memory queue due to large memory resources. For reflectivity calculations one cannot just copy CONTCAR to POSCAR to continue the calculations can they? It gets up to WFULL0030 and then says due to time limit the job was ended. Any help would be greatly appreciated

This is the INCAR:

PREC = Normal
ENCUT = 520
ALGO = GW0
ISMEAR = 0
SIGMA = 0.05
NBANDS = 216
NOMEGA = 72
LSPECTRAL= .TRUE.
LOPTICS= .TRUE.
OMEGAMAX= 10
NBANDSO = 16
NBANDSV = 16
NEDOS=3000

[alexey.tal]

Dear jess_white,

Thank you for your question.

Could you post the link for the tutorial you are following? I see that you are running a GW calculation. However, you set tags like NBANDSO and NBANDSV which are only relevant for BSE or TDDFT. Also, you set LOPTICS, which you shouldn't need in the eigenvalue GW calculations.

Also, could you please provide all the input and output files from your calculation (see guidelines)?

Best wishes,
Alexey

[jess_white]

Hi Alexey,

Thank you for your response. So the example calculation I got from vaspkit examples and that is the file called Si_BSE_optical. It seemed like they had four INCARs so I assumed that they did one after another but I may have been poorly mistaken. I have not ventured into spec/reflectivity in VASP but helping out a colleague so I apologise with my probably very wrong way of approaching the system.

The La_2 is my attempt - I did have to delete the WAVECAR, WAVEDAR and the WFULL files as too large.

Any help is greatly appreciated!

Thanks,
Jess

[alexey.tal]

Thank you for providing the files.

We have made our own quite comprehensive tutorials on GW and BSE calculations, which should be a good starting point to learn about these methods.
Unfortunately, it is not currently possible to interrupt and restart a GW calculation. However, your system is quite large, so you should try the low-scaling GW algorithm which is substantially faster for large systems than the standard GW algorithm.

[jess_white]

Thank you Alexey! What is the difference between GW and BSE? I am a bit confused which one to use

[alexey.tal]

If you are interested in the reflectivity you need to calculate the dielectric function first. In principle, both GW and BSE allow one to calculate the dielectric function. However, the electron-hole interaction is taken into account in BSE, but not in GW. If the electron-hole interaction is strong in your system, the BSE is the right approach. Furthermore, the GW approach calculates the dielectric function in the RPA approximation, which usually improves the dielectric function very little compared to the independent particle approximation, so one might as well just use the independent particle approach (see LOPTICS). Nevertheless, the GW approximation is typically used for the approximation to the self-energy, which allows to improved the electronic structure description and the band gap compared to DFT with local exchange-correlation potentials.

You can also have a look at our lecture on GW.

Could you provide more detail about the system you are interested in studying and what properties you would like to calculate?