Here I present the BSE input file for yambo and I will discuss the meaning of each parameter. For a discussion on convergence issues see my post: Reasonable parameters for Yambo calculations. To generate this input type yambo -o b -k sex -y d -p p -V qp. I will not discuss the parameters relative to the dielectric constant and the cutoff for the Coulomb interaction because they are the same of the GW.
BS_CPU= "1 30 2" # [PARALLEL] CPUs for each role BS_ROLEs= "k eh t" # [PARALLEL] CPUs roles (k,eh,t) BS_nCPU_invert= 4 # [PARALLEL] CPUs for matrix inversion BS_nCPU_diago= 4 # [PARALLEL] CPUs for matrix diagonalization Parallelization on k-points, electron-holes and transitions. The k parallelization is the more efficient, than the eh and last the t. I advice you not to use a larger number of processors in BS_nCPU_invert and BS_nCPU_diago because matrix inversions and diagonalisation are not efficiently parallelized. BSEmod= "causal" # [BSE] resonant/causal/coupling Different approximations for the BSE. Causal corresponds to the standard one, namely the Tamm-Dancoff approximation, while Coupling refers to the full BSE. See Eq. 18 in Ref.. BSSmod= "d" # [BSS] (h)aydock/(d)iagonalization/(i)nversion/(t)ddft` Method to solve the BSE equation. For large systems the iterative solution (h) can be more efficient but one loose information about the excitons. Otherwise use (d), the full diagonalization. BSKmod= "SEX" # [BSE] IP/Hartree/HF/ALDA/SEX/BSfxcBSSmod= "d" # [BSS] (h)aydock/(d)iagonalization/(i)nversion/(t)ddft` Type of correlation in the BSE. The standard one is the Screened EXchange (SEX). Other possibility are Hartree-Fock (HF), Hartree, TD-LDA, independent particles (IP). BSENGexx= 20065 RL # [BSK] Exchange components Number of G-vectors in the time-dependent Hartree-term. If this number is larger than FFTGvecs the code automatically reduces it. BSENGBlk= 3343 RL # [BSK] Screened interaction block size Number of G-vectors read of the inverse dielectric constant. This number should less or equal to the number of G-vectors you used to calculate the dielectric constant. I advice not to change it. KfnQPdb= "E < SAVE/ndb.QP" # [EXTQP BSK BSS] Database This flag indicates to the code to read the quasi-particle corrections obtained from the GW calculations. Gauge= "length" # [BSE] Gauge (length|velocity) Gauge to use in the response function, leave unchanged. % BEnRange 0.00000 | 10.00000 | eV # [BSS] Energy range % % BDmRange 0.10000 | 0.10000 | eV # [BSS] Damping range % BEnSteps= 100 # [BSS] Energy steps % BLongDir 1.000000 | 0.000000 | 0.000000 | # [BSS] [cc] Electric Field % These four variables governs the final plot. BEnRange is the energy range to plot the optical spectra, BDmRange the damping, BEnSteps the number of points and finally BLongDir the direction of the optical response. (see Eq. 60 of Ref) % BSEBands 194 | 382 | # [BSK] Bands range % Number of bands that enter in the costruction of the BSE matrix. WRbsWF # [BSS] Write to disk excitonic the FWs Write on disk both eigenvalues and eigenvectors of the BSE equation. In this way you can plot and analyze the excitonic wave-function with ypp. (see Eq. 61 of Ref)
Notice that the size of the BSE matrix grows as Nv*Nc*Nk, therefore your calculation can become prohibitive very soon with the system size. If you have a large matrix I advise you to use the Haydock approach (BSSmod= “h”) to get the spectra because it is more efficient than full diagonalization.
Finally if you are studying an isolated molecule or a slab, you should consider the Coulomb cuf-off, -r flag (-c in yambo 3.4.1) in order to reduce the interaction between the periodic replica, for more details see here.
 Application of the Green’s functions method to the study of the optical properties of semiconductors, G. Strinati
Rivista del Nuovo Cimento, Vol 11, N. 12, Pg. 1 (1988)
 Effects of the Electron–Hole Interaction on the Optical Properties of Materials: the Bethe–Salpeter Equation, G. Bussi
Physica Scripta, Volume 2004, 141 (2003)