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

Numerical integrals d_lP(R) for periodic RI-RS fitting. More...

Functions/Subroutines

subroutine, public compute_d_lp_r (bs_env, grid_points, cells_r_d_lp, d_lp_r, qpoints, charge_lq)
 Allocate d_lP(R), prepare its integration grid, and evaluate the integrals.
subroutine, public compute_d_lp_cutoffs (bs_env, cutoffs)
 Automatic l/P cutoff from the overlap of the diffuse AO-field square and RI potential.
subroutine, public get_local_l_range (n_grid, n_ranks, i_rank, first_l, last_l)
 Distribute individual l rows evenly across MPI ranks; empty ranges are valid.
subroutine, public write_ri_rs_progress (unit_nr, phase, n_done, n_total, next_progress, phase_start)
 Print one rank-local RI-RS work milestone without synchronizing MPI ranks. Percentages count completed local units; time is the elapsed wall time since the phase began.

Detailed Description

Numerical integrals d_lP(R) for periodic RI-RS fitting.

Author
Jan Wilhelm (with help of ChatGPT, GPT-6 Sol and GPT-6 Astra)

Function/Subroutine Documentation

◆ compute_d_lp_r()

subroutine, public ri_rs_kp_integrals::compute_d_lp_r ( type(post_scf_bandstructure_type), pointer bs_env,
real(dp), dimension(:, :), intent(in) grid_points,
integer, dimension(:, :), intent(in) cells_r_d_lp,
real(dp), dimension(:, :, :), intent(out), allocatable d_lp_r,
real(dp), dimension(:, :), intent(in) qpoints,
complex(dp), dimension(:, :), intent(out), allocatable charge_lq )

Allocate d_lP(R), prepare its integration grid, and evaluate the integrals.

Parameters
bs_env...
grid_points...
cells_R_d_lPRetained integer cell vectors R, indexed (3,R).
d_lP_RNumerical integrals indexed by local point l, RI function P and cell R.
qpoints...
charge_lq...

Definition at line 103 of file ri_rs_kp_integrals.F.

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

◆ compute_d_lp_cutoffs()

subroutine, public ri_rs_kp_integrals::compute_d_lp_cutoffs ( type(post_scf_bandstructure_type), pointer bs_env,
real(dp), dimension(:), intent(out), allocatable cutoffs )

Automatic l/P cutoff from the overlap of the diffuse AO-field square and RI potential.

Outside the metric cutoff sphere, conservative Gaussian envelopes are f_l(r)^2 ~ exp[-alpha_AO |r-r_l|^2] and v_PT(r) ~ exp[-alpha_RI (|r-R_PT|-r_c)^2/2]. Maximizing their product over r gives exp[-gamma (|r_l-R_PT|-r_c)^2], with 1/gamma=1/alpha_AO+2/alpha_RI. Both methods use the same EPS_FILTER-dependent lattice domain.

Parameters
bs_env...
cutoffs...

Definition at line 148 of file ri_rs_kp_integrals.F.

Here is the caller graph for this function:

◆ get_local_l_range()

subroutine, public ri_rs_kp_integrals::get_local_l_range ( integer, intent(in) n_grid,
integer, intent(in) n_ranks,
integer, intent(in) i_rank,
integer, intent(out) first_l,
integer, intent(out) last_l )

Distribute individual l rows evenly across MPI ranks; empty ranges are valid.

Parameters
n_gridNumber of RI-RS grid points.
n_ranksNumber of MPI ranks.
i_rankMPI rank.
first_lFirst locally owned grid row.
last_lLast locally owned grid row.

Definition at line 186 of file ri_rs_kp_integrals.F.

Here is the caller graph for this function:

◆ write_ri_rs_progress()

subroutine, public ri_rs_kp_integrals::write_ri_rs_progress ( integer, intent(in) unit_nr,
character(len=*), intent(in) phase,
integer(kind=int_8), intent(in) n_done,
integer(kind=int_8), intent(in) n_total,
integer, intent(inout) next_progress,
real(kind=dp), intent(in) phase_start )

Print one rank-local RI-RS work milestone without synchronizing MPI ranks. Percentages count completed local units; time is the elapsed wall time since the phase began.

Parameters
unit_nr...
phase...
n_done...
n_total...
next_progress...
phase_start...

Definition at line 206 of file ri_rs_kp_integrals.F.

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