RPA absorption spectrum analysis

[nthiliniek]

Hello,

Since I wanted to analyze the optical spectrum of RPA, I switched to e-h space, But the two spectra I got from both G-space and e-h space are comparatively different. I used the same value for BSENGexx and NGsBlkXd and I used coupling also. The only difference is BSE bands. I tried to include all the bands (345), but the calculation did not work. Therefore I reduced the bands up to 20 (occupied 10, unoccupied 10).
Is there a way to include all the bands? Is it possible to get a spectrum close to the G-space spectrum by reducing the number of bands?
Here are the details of the bands in my system:

[X] Filled Bands : 288
[X] Empty Bands : 289 345

I am not sure, how should I get a proper ratio for BSE bands parameter. Should I include more empty bands than filled bands in the ratio?
Here I am attaching my report files for both calculations.

And If it is a memory problem, I am not sure how I distribute the memory for the calculation.

Thanks in advance.

TD-rpa.rar

[Daniele Varsano]

Dear Niranji,

calculations in G space or transition space are totally equivalent, and they provide the same results if the same parameters are used.
When working in transition space, adding bands increase the dimension of the matrix you want to diagonalize so probably you cannot include all of them.

The matrix of your calculation (279-298) anyway is small, and you can increase it.
I suggest you to enlarge the bands windows e.g. (277-300), (275-302), etc. until you reach convergence of the low energy peaks that I presume they are the one you want to analyze.

Best,
Daniele