Re: Memory issue with TDDFT calculation
1. I will use gw POTCAR during whole procedure.
Super!
2. Could you please comment if I want to perform BSE@GW or TDHF with parameters from GW to account for electron hole interaction at all for such large systems is it not feasible with our HPC configurations? Because similar memory issue i faced during GW band structure calculations.
If you would like to use the GW algorithm for calculating the quasiparticles and the screening you should use the low-scaling GW algorithm
as it is much faster for large systems. You can find the description of this method together with the memory estimates on our wiki page.
If you would like to include the electron-hole interaction the LADDER=.TRUE. have to be included in your INCAR. Furthermore, you can try to use a model dielectric function as discussed in Tal et al., Phys. Rev. Research, 032019 (2020). In this case you can neglect the local exchange-correlation kernel, i.e., set LFXC=.TRUE. and avoid the allocation of the large array for \(f_{xc}(G,G')\).
3. Could you please comment on how to select NBANDO/V ?
The number of bands should be selected based on the energy range for the dielectric function that is relevant in the calculation, i.e., all the bands that contribute to the dielectric function at the relevant energy range need to be included. In principle, one might need to include all the occupied bands and twice the number of empty bands. But these can be reduced if only a small part of the spectrum is required.