GW+BSE calculation stucking at Transition Groups Buld-up

[burkzdemir]

Dear developers,

I am running yambo-4.4.0 and trying to perform GW+BSE calculation and my calculation is stuck at Transition Groups Build-up. Attached is my input file (I could not attach my report file since it was too large).

--------------------------------------------------------------------------------------
<22m-43s> P0001: X@q[580] |########################################| [100%] --(E) --(X)
<22m-43s> P0001: Writing pp_fragment_580
<22m-43s> P0001: [05] Response Functions in Transition space
<22m-43s> P0001: Response_T_space parallel ENVIRONMENT is incomplete. Switching to defaults
<22m-43s> P0001: [PARALLEL Response_T_space for K(ibz) on 3 CPU] Loaded/Total (Percentual):194/580(33%)
<22m-43s> P0001: [PARALLEL Response_T_space for (e/h) Groups on 10 CPU] Loaded/Total (Percentual):194/5800(3%)
<22m-43s> P0001: [PARALLEL Response_T_space for (e/h)->(e/h)' Transitions (ordered) on 1 CPU] Loaded/Total (Percentual):562794/16822900(3%)
<22m-43s> P0001: [PARALLEL Response_T_space for CON bands on 10 CPU] Loaded/Total (Percentual):3/21(14%)
<22m-43s> P0001: [PARALLEL Response_T_space for VAL bands on 1 CPU] Loaded/Total (Percentual):22/22(100%)
<22m-43s> P0001: [05.01] Transition Groups build-up

----------------------------------------------------------------------------------------

[Daniele Varsano]

Dear Burak,
Without looking at your report and complete setup I can only suppose what is going on.
Most probably it is a memory issue:
You are using many bands in the BSE calculation (I do not know which is your last occupied bands so I cannot estimate the size of the BSE matrix),
anyway, I do suspect that you have a very large matrix (Nc x Nv x Nk), note Nk is the total k point grid in the whole Bz (reported in the report).
Usually, in BSE calculations spectra are converged by using few bands across the gap. My suggestion is to reduce your number of BSEbands and increase step by step until convergence. Next, please also note that you did not specify any parallelization strategy, in this case, Yambo switch to the default:

Code: Select all

Response_T_space parallel ENVIRONMENT is incomplete. Switching to defaults

and it is not guaranteed that it is the best strategy for memory saving.
Moreover, even if unrelated with your problem, note that in your input you are considering the electron-hole attraction by using only 1 reciprocal lattice vector:

Code: Select all

BSENGBlk= 1 RL
NGsBlkXp= 1 RL  

and this will provide for sure unconverged results.

Maybe you want to have a look at the tutorial dealing with BSE calculations:
Tutorial1
Tutorial2

Best,
Daniele