(git:27c966b)
Loading...
Searching...
No Matches
gw_ri_rs_small_cell_full_kp_gamma.F File Reference

Go to the source code of this file.

Modules

module  gw_ri_rs_small_cell_full_kp_gamma
 Gamma-point (q = 0) correction of W and Σ for periodic RI-RS GW.

Functions/Subroutines

subroutine, public gw_ri_rs_small_cell_full_kp_gamma::prepare_ri_rs_gamma_correction (qs_env, bs_env)
 Preserve distributed AO position, Hamiltonian and overlap images before QS release. Only the sparse operators and their immutable image traversal survive; SCF grids and neighbor lists can be released by the common GW initialization path.
subroutine, public gw_ri_rs_small_cell_full_kp_gamma::gamma_point_correction (bs_env, qs_env, gwkp_env)
 Gamma-correction head and wing coefficients for the near-Gamma cells. Construct the q->0 dielectric limit along face-cubature and mesh-sample directions from SCF dipole transitions and the directly fitted Gamma Gaussian body.
subroutine, public gw_ri_rs_small_cell_full_kp_gamma::singular_head_cell_factors (bs_env, cells, factors)
 Integrate 1/q**2 (3D) or 1/|q| (2D slab) over the distinct shifted-mesh cells around Gamma. The regular mesh sum already contains the near-Gamma model samples, which are subtracted here. In 3D the returned coefficient multiplies the factored head density 4*pi*Z_l0(0)*conjg(Z_l'0(0))/M_0(0)**2; in 2D it has units of inverse in-plane momentum. The face-to-polar Jacobian resolves the integrable singularity without sampling q=0.
subroutine, public gw_ri_rs_small_cell_full_kp_gamma::add_gamma_corr_cell_correction (bs_env, cells, model, i_t, z_head, first_pair, w_grid_r)
 Add the Gamma-correction head and wings of one time point to the local grid pairs. The wing of cell R is w_l(R,τ) = Σ_ω cos(τ,ω) Σ_c wing_cartesian_lc(ω) c_c(R,ω); one GEMM per cell forms it together with its conjugate-side partner.
subroutine, public gw_ri_rs_small_cell_full_kp_gamma::add_gamma_corr_pair_cell (bs_env, cells, icell, head_scalar, head_vector, first_pair, values_r, wing_left, wing_right)
 Scatter one low-rank Gamma correction into the centred grid-pair images. For s=nearest_integer[H^{-1}(r_l-r_l')], C_ll'(R)=C0_ll'(R-s). R-s is wrapped into the finite FFT cell with its transfer-mesh Bloch twist. Half-cell ties average adjacent integer images, retaining Hermiticity and cell-wrapping covariance. The mirrored conjugate-transpose contribution enforces reciprocity also on Nyquist faces. No grid-pair angular tensor is formed. values_R holds flattened column-major pairs beginning at first_pair.