(git:27c966b)
Loading...
Searching...
No Matches
bse_iterative Module Reference

Iterative solution of the Bethe-Salpeter equation: the block Davidson solvers on top of the matrix-free application of A and B, and the post-processing of their eigenvectors. The solvers live in bse_davidson, apart from this driver, so that a later Lanczos-Haydock solver for the spectrum can share the matrix-free A and B of bse_matvec and this driver. More...

Functions/Subroutines

subroutine, public solve_bse_iteratively (eigenval_reduced, homo_red_arr, virt_red_arr, homo, dimen_ri, do_tda, do_abba, bse_env, mo_coeff, unit_nr)
 Lowest excitations of a closed-shell reference from the block Davidson solvers on top of the matrix-free application of A and B, post-processed as in the full diagonalization. TDA (bse_davidson_tda): sum_jb A_ia,jb X_jb^n = Ω^n X_ia^n. Full problem, by ABBA_SOLVER: MK_DAVIDSON, sum_jb [(A+B)(A-B)]_ia,jb x_jb^n = (Ω^n)^2 x_ia^n with x^n = X^n - Y^n; OLSEN_STRATMANN, sum_jb (A+B)_ia,jb y_jb^n = Ω^n x_ia^n and sum_jb (A-B)_ia,jb x_jb^n = Ω^n y_ia^n with y^n = X^n + Y^n. Neither A nor B is formed: bse_matvec_apply contracts the RI slabs, (A Z)_ia = (ε_a-ε_i) Z_ia + α sum_P B^P_ia sum_jb B^P_jb Z_jb - w sum_Pjb \bar{B}^P_ij B^P_ab Z_jb (B Z)_ia = α sum_P B^P_ia sum_jb B^P_jb Z_jb - w sum_Pjb \bar{B}^P_ib B^P_ja Z_jb with α, w and \bar{B} set by create_matvec_env from SPIN_CONFIG and SCREENING_IN_W. The five slabs are released once the solver holds its RI-sliced copies.

Detailed Description

Iterative solution of the Bethe-Salpeter equation: the block Davidson solvers on top of the matrix-free application of A and B, and the post-processing of their eigenvectors. The solvers live in bse_davidson, apart from this driver, so that a later Lanczos-Haydock solver for the spectrum can share the matrix-free A and B of bse_matvec and this driver.

History
09.2026 created [Maximilian Graml]

Function/Subroutine Documentation

◆ solve_bse_iteratively()

subroutine, public bse_iterative::solve_bse_iteratively ( real(kind=dp), dimension(:), intent(in) eigenval_reduced,
integer, dimension(:), intent(in) homo_red_arr,
integer, dimension(:), intent(in) virt_red_arr,
integer, dimension(:), intent(in) homo,
integer, intent(in) dimen_ri,
logical, intent(in) do_tda,
logical, intent(in) do_abba,
type(bse_env_type), intent(inout) bse_env,
type(cp_fm_type), dimension(:), intent(in) mo_coeff,
integer, intent(in) unit_nr )

Lowest excitations of a closed-shell reference from the block Davidson solvers on top of the matrix-free application of A and B, post-processed as in the full diagonalization. TDA (bse_davidson_tda): sum_jb A_ia,jb X_jb^n = Ω^n X_ia^n. Full problem, by ABBA_SOLVER: MK_DAVIDSON, sum_jb [(A+B)(A-B)]_ia,jb x_jb^n = (Ω^n)^2 x_ia^n with x^n = X^n - Y^n; OLSEN_STRATMANN, sum_jb (A+B)_ia,jb y_jb^n = Ω^n x_ia^n and sum_jb (A-B)_ia,jb x_jb^n = Ω^n y_ia^n with y^n = X^n + Y^n. Neither A nor B is formed: bse_matvec_apply contracts the RI slabs, (A Z)_ia = (ε_a-ε_i) Z_ia + α sum_P B^P_ia sum_jb B^P_jb Z_jb - w sum_Pjb \bar{B}^P_ij B^P_ab Z_jb (B Z)_ia = α sum_P B^P_ia sum_jb B^P_jb Z_jb - w sum_Pjb \bar{B}^P_ib B^P_ja Z_jb with α, w and \bar{B} set by create_matvec_env from SPIN_CONFIG and SCREENING_IN_W. The five slabs are released once the solver holds its RI-sliced copies.

Parameters
Eigenval_reducedenergies of the homo_red + virt_red active levels in Hartree, QP or DFT per USE_KS_ENERGIES
homo_red_arroccupied levels in the active window, per spin; entry 1 is used
virt_red_arrvirtual levels in the active window, per spin; entry 1 is used
homooccupied levels of the full reference, per spin; postprocess_bse indexes the MOs with it
dimen_RInumber of RI functions; only the debug construction of the explicit A and B uses it
do_tdasolve the TDA problem
do_abbasolve the full problem; after the TDA in a TDA+ABBA run, seeded with X^TDA
bse_envthe BSE environment: the BSE_ITERAT settings, BSE_DEBUG_PRINT and the TDDFPT control flags; the five slabs of spin 1, released here: fm_S_ia = B^P_ia (N_RI x homo_red*virt_red), fm_S_ij = B^P_ij, fm_S_ab = B^P_ab, fm_S_bar_ia = \bar{B}^P_ia = sum_Q [1+Q(0)]^-1_PQ B^Q_ia (the slab of W_ib,aj) and fm_S_bar_ij = \bar{B}^P_ij (the slab of W_ij,ab)
mo_coeffMO coefficients of the reference, handed to postprocess_bse
unit_nroutput unit, positive on the writing rank only

Definition at line 88 of file bse_iterative.F.

Here is the call graph for this function:
Here is the caller graph for this function: