Query regarding HSE functional

[Jyotirmoy Deb]

I want to calculate band structure of a 2D material using HSE functional. I have optimized the structure initially using GGA-PBE method and starting from this geomerty I have started the calculation for optimization using HSE functional (Two compute node used each having 24 processor). The unit cell consists of 26 atoms. But the job is running since last ten days. Actually I am very much confused whether the calculation is running in right direction or not.  Kindly suggest me. Also give me an idea how to get band gap using HSE functional within a limited time period.   

[Petr Khomyakov]

HSE functional is computationally much more demanding than GGA. So, that is expected that the computational time has significantly increased. It might be that you could optimize your settings to make it somewhat faster. Could you post your python script and related log-file to see if everything is fine with the current calculation, in principle?

Note again that the computational cost for HSE calculations is very much dependent on the number of k-points used for sampling the Brillouin zone, as well as on the wave function cutoff and the number of unoccupied bands included in the calculation. 

[Jyotirmoy Deb]

Here I am attaching the python script. Kindly suggest me how to change the input so that I will get the optimized geometry in a limited time period. My actual interest is to calculate the band gap using HSE functional.

[Petr Khomyakov]

Your calculation looks healthy. I see no issue: stress and forces are gradually decreasing.

I do not think you need 2 k-points in the out-of-plane direction, i.e., nc can probably be set to 1. I would also reduce the number of k-points in the other directions, as your in-plane unit cell looks relatively large. I would first test basic convergence with respect to k-point sampling using GGA functional that is much cheaper, and do it for different k-point grids to see how much that affects the results. You may also do HSE optimization with less k-points and then increase the grid density when computing the band structure/band gap.

Is there any band gap in this structure when using GGA? You may also check whether your GGA-optimized structure and current HSE-optimized structure have similar GGA (or HSE) gaps or band structure in general. That would allow you checking if using HSE is actually important for relaxation or GGA works good enough.