BSE with RIM cutoff

[sdwang]

Dear all,
I am testing RIM_W algorithm of version 5.1 for 2D hBN. And the BSE excitation energies with and without RIM_W are different as below:
without RIM_W (cut off is 'slab_z'):
# E [ev] Strength Index
#
5.28477526 0.288948715 1.00000000
5.28508425 1.00000000 2.00000000
6.12765455 0.500189364E-1 3.00000000
6.12791538 0.405531764 4.00000000
6.26788950 0.477613789E-7 5.00000000
6.31209612 0.318419770E-2 6.00000000
6.31224060 0.116705261E-1 7.00000000
6.34610367 0.126361414E-6 8.00000000
6.60957623 0.207954161E-1 9.00000000
...

with RIM_W (cut off is 'slab_z'):
# E [ev] Strength Index
#
5.72337484 0.281303704 1.00000000
5.72371817 1.00000000 2.00000000
6.54615641 0.512173101E-1 3.00000000
6.54645443 0.407920659 4.00000000
6.69065762 0.452553444E-7 5.00000000
6.72918653 0.554423500E-2 6.00000000
6.72934961 0.157248992E-1 7.00000000
6.77091455 0.152845715E-6 8.00000000
7.02254725 0.263438858E-1 9.00000000
We can find a 0.44 eV of the excitation energy difference. My question is if the RIM_W algorithm is helpful to the BSE convergence.

Best,

Shudong

[Daniele Varsano]

Dear Shudong,

the RIM_W has been conceived for GW calculations. It can be useful also for BSE if excitons are formed by transitions in a localized part of the BZ.
We observed the algorithm is working for some systems we are dealing with, anyway it has not been extensively tested, so I suggest you to use it with care, you can perform a k point convergence and realize if you are obtaining meaningful results.

Best,
Daniele