Line data Source code
1 : !--------------------------------------------------------------------------------------------------!
2 : ! CP2K: A general program to perform molecular dynamics simulations !
3 : ! Copyright 2000-2026 CP2K developers group <https://cp2k.org> !
4 : ! !
5 : ! SPDX-License-Identifier: GPL-2.0-or-later !
6 : !--------------------------------------------------------------------------------------------------!
7 :
8 : ! **************************************************************************************************
9 : !> \brief Routines needed for cubic-scaling RPA and SOS-Laplace-MP2 forces
10 : !> \author Augustin Bussy
11 : ! **************************************************************************************************
12 : MODULE rpa_im_time_force_methods
13 : USE admm_methods, ONLY: admm_projection_derivative
14 : USE admm_types, ONLY: admm_type,&
15 : get_admm_env
16 : USE ao_util, ONLY: exp_radius_very_extended
17 : USE atomic_kind_types, ONLY: atomic_kind_type,&
18 : get_atomic_kind_set
19 : USE basis_set_types, ONLY: gto_basis_set_p_type,&
20 : gto_basis_set_type
21 : USE bibliography, ONLY: Bussy2023,&
22 : cite_reference
23 : USE cell_types, ONLY: cell_type,&
24 : pbc
25 : USE cp_blacs_env, ONLY: cp_blacs_env_type
26 : USE cp_control_types, ONLY: dft_control_type
27 : USE cp_dbcsr_api, ONLY: &
28 : dbcsr_add, dbcsr_clear, dbcsr_complete_redistribute, dbcsr_copy, dbcsr_create, &
29 : dbcsr_distribution_new, dbcsr_distribution_release, dbcsr_distribution_type, &
30 : dbcsr_get_block_p, dbcsr_iterator_blocks_left, dbcsr_iterator_next_block, &
31 : dbcsr_iterator_start, dbcsr_iterator_stop, dbcsr_iterator_type, dbcsr_multiply, &
32 : dbcsr_p_type, dbcsr_release, dbcsr_scale, dbcsr_set, dbcsr_type, dbcsr_type_antisymmetric, &
33 : dbcsr_type_no_symmetry, dbcsr_type_symmetric
34 : USE cp_dbcsr_cholesky, ONLY: cp_dbcsr_cholesky_decompose,&
35 : cp_dbcsr_cholesky_invert
36 : USE cp_dbcsr_contrib, ONLY: dbcsr_add_on_diag,&
37 : dbcsr_frobenius_norm
38 : USE cp_dbcsr_diag, ONLY: cp_dbcsr_power
39 : USE cp_dbcsr_operations, ONLY: copy_dbcsr_to_fm,&
40 : copy_fm_to_dbcsr,&
41 : cp_dbcsr_dist2d_to_dist,&
42 : cp_dbcsr_sm_fm_multiply,&
43 : dbcsr_allocate_matrix_set,&
44 : dbcsr_deallocate_matrix_set
45 : USE cp_eri_mme_interface, ONLY: cp_eri_mme_update_local_counts
46 : USE cp_fm_struct, ONLY: cp_fm_struct_create,&
47 : cp_fm_struct_release,&
48 : cp_fm_struct_type
49 : USE cp_fm_types, ONLY: cp_fm_create,&
50 : cp_fm_release,&
51 : cp_fm_set_all,&
52 : cp_fm_to_fm,&
53 : cp_fm_type
54 : USE dbt_api, ONLY: &
55 : dbt_batched_contract_finalize, dbt_batched_contract_init, dbt_clear, dbt_contract, &
56 : dbt_copy, dbt_copy_matrix_to_tensor, dbt_copy_tensor_to_matrix, dbt_create, dbt_destroy, &
57 : dbt_filter, dbt_get_info, dbt_mp_environ_pgrid, dbt_pgrid_create, dbt_pgrid_destroy, &
58 : dbt_pgrid_type, dbt_scale, dbt_type
59 : USE distribution_2d_types, ONLY: distribution_2d_type
60 : USE gaussian_gridlevels, ONLY: gaussian_gridlevel
61 : USE hfx_admm_utils, ONLY: tddft_hfx_matrix
62 : USE hfx_derivatives, ONLY: derivatives_four_center
63 : USE hfx_exx, ONLY: add_exx_to_rhs
64 : USE hfx_ri, ONLY: get_2c_der_force,&
65 : get_force_from_3c_trace,&
66 : get_idx_to_atom,&
67 : hfx_ri_update_forces
68 : USE hfx_types, ONLY: alloc_containers,&
69 : block_ind_type,&
70 : dealloc_containers,&
71 : hfx_compression_type,&
72 : hfx_type
73 : USE input_constants, ONLY: do_admm_aux_exch_func_none,&
74 : do_eri_gpw,&
75 : do_eri_mme,&
76 : do_potential_id,&
77 : ri_rpa_method_gpw
78 : USE input_section_types, ONLY: section_vals_get,&
79 : section_vals_get_subs_vals,&
80 : section_vals_type,&
81 : section_vals_val_get
82 : USE iterate_matrix, ONLY: matrix_exponential
83 : USE kinds, ONLY: dp,&
84 : int_8
85 : USE libint_2c_3c, ONLY: libint_potential_type
86 : USE machine, ONLY: m_flush,&
87 : m_walltime
88 : USE mathconstants, ONLY: fourpi
89 : USE message_passing, ONLY: mp_cart_type,&
90 : mp_para_env_release,&
91 : mp_para_env_type
92 : USE mp2_eri, ONLY: integrate_set_2c
93 : USE mp2_eri_gpw, ONLY: calc_potential_gpw,&
94 : cleanup_gpw,&
95 : prepare_gpw,&
96 : virial_gpw_potential
97 : USE mp2_types, ONLY: mp2_type
98 : USE orbital_pointers, ONLY: ncoset
99 : USE parallel_gemm_api, ONLY: parallel_gemm
100 : USE particle_methods, ONLY: get_particle_set
101 : USE particle_types, ONLY: particle_type
102 : USE pw_env_types, ONLY: pw_env_get,&
103 : pw_env_type
104 : USE pw_methods, ONLY: pw_axpy,&
105 : pw_copy,&
106 : pw_integral_ab,&
107 : pw_scale,&
108 : pw_transfer,&
109 : pw_zero
110 : USE pw_poisson_methods, ONLY: pw_poisson_solve
111 : USE pw_poisson_types, ONLY: pw_poisson_type
112 : USE pw_pool_types, ONLY: pw_pool_type
113 : USE pw_types, ONLY: pw_c1d_gs_type,&
114 : pw_r3d_rs_type
115 : USE qs_collocate_density, ONLY: calculate_rho_elec,&
116 : collocate_function
117 : USE qs_core_matrices, ONLY: core_matrices,&
118 : kinetic_energy_matrix
119 : USE qs_density_matrices, ONLY: calculate_whz_matrix
120 : USE qs_environment_types, ONLY: get_qs_env,&
121 : qs_environment_type,&
122 : set_qs_env
123 : USE qs_force_types, ONLY: qs_force_type
124 : USE qs_fxc, ONLY: qs_fxc_create
125 : USE qs_integral_utils, ONLY: basis_set_list_setup
126 : USE qs_integrate_potential, ONLY: integrate_pgf_product,&
127 : integrate_v_core_rspace,&
128 : integrate_v_rspace
129 : USE qs_interactions, ONLY: init_interaction_radii_orb_basis
130 : USE qs_kind_types, ONLY: qs_kind_type
131 : USE qs_ks_methods, ONLY: calc_rho_tot_gspace
132 : USE qs_ks_reference, ONLY: ks_ref_potential
133 : USE qs_ks_types, ONLY: set_ks_env
134 : USE qs_linres_types, ONLY: linres_control_type
135 : USE qs_matrix_w, ONLY: compute_matrix_w
136 : USE qs_mo_types, ONLY: get_mo_set,&
137 : mo_set_type
138 : USE qs_neighbor_list_types, ONLY: neighbor_list_set_p_type,&
139 : release_neighbor_list_sets
140 : USE qs_overlap, ONLY: build_overlap_matrix
141 : USE qs_p_env_methods, ONLY: p_env_create,&
142 : p_env_psi0_changed
143 : USE qs_p_env_types, ONLY: p_env_release,&
144 : qs_p_env_type
145 : USE qs_rho_atom_types, ONLY: rho_atom_type
146 : USE qs_rho_types, ONLY: qs_rho_create,&
147 : qs_rho_get,&
148 : qs_rho_set,&
149 : qs_rho_type
150 : USE qs_tensors, ONLY: &
151 : build_2c_derivatives, build_2c_integrals, build_2c_neighbor_lists, build_3c_derivatives, &
152 : build_3c_neighbor_lists, calc_2c_virial, calc_3c_virial, compress_tensor, &
153 : decompress_tensor, get_tensor_occupancy, neighbor_list_3c_destroy
154 : USE qs_tensors_types, ONLY: create_2c_tensor,&
155 : create_3c_tensor,&
156 : create_tensor_batches,&
157 : distribution_3d_create,&
158 : distribution_3d_type,&
159 : neighbor_list_3c_type
160 : USE realspace_grid_types, ONLY: map_gaussian_here,&
161 : realspace_grid_type
162 : USE response_solver, ONLY: response_equation_new
163 : USE rpa_im_time, ONLY: compute_mat_dm_global,&
164 : propagator_sector_occupied,&
165 : propagator_sector_virtual
166 : USE rpa_im_time_force_types, ONLY: im_time_force_type
167 : USE rs_pw_interface, ONLY: potential_pw2rs
168 : USE task_list_types, ONLY: task_list_type
169 : USE time_frequency_grids, ONLY: time_frequency_grid_type
170 : USE virial_types, ONLY: virial_type
171 : #include "./base/base_uses.f90"
172 :
173 : IMPLICIT NONE
174 :
175 : PRIVATE
176 :
177 : CHARACTER(len=*), PARAMETER, PRIVATE :: moduleN = 'rpa_im_time_force_methods'
178 :
179 : PUBLIC :: init_im_time_forces, calc_laplace_loop_forces, calc_post_loop_forces, &
180 : keep_initial_quad, calc_rpa_loop_forces
181 :
182 : CONTAINS
183 :
184 : ! **************************************************************************************************
185 : !> \brief Initializes and pre-calculates all needed tensors for the forces
186 : !> \param force_data ...
187 : !> \param fm_matrix_PQ ...
188 : !> \param t_3c_M the 3-center M tensor to be used as a template
189 : !> \param unit_nr ...
190 : !> \param mp2_env ...
191 : !> \param qs_env ...
192 : ! **************************************************************************************************
193 50 : SUBROUTINE init_im_time_forces(force_data, fm_matrix_PQ, t_3c_M, unit_nr, mp2_env, qs_env)
194 :
195 : TYPE(im_time_force_type), INTENT(INOUT) :: force_data
196 : TYPE(cp_fm_type), INTENT(IN) :: fm_matrix_PQ
197 : TYPE(dbt_type), INTENT(INOUT) :: t_3c_M
198 : INTEGER, INTENT(IN) :: unit_nr
199 : TYPE(mp2_type) :: mp2_env
200 : TYPE(qs_environment_type), POINTER :: qs_env
201 :
202 : CHARACTER(LEN=*), PARAMETER :: routineN = 'init_im_time_forces'
203 :
204 : INTEGER :: handle, i_mem, i_xyz, ibasis, ispin, &
205 : n_dependent, n_mem, n_rep, natom, &
206 : nkind, nspins
207 : INTEGER(int_8) :: nze, nze_tot
208 50 : INTEGER, ALLOCATABLE, DIMENSION(:) :: dist1, dist2, dist_AO_1, dist_AO_2, &
209 50 : dist_RI, dummy_end, dummy_start, &
210 100 : end_blocks, sizes_AO, sizes_RI, &
211 50 : start_blocks
212 : INTEGER, DIMENSION(2) :: pdims_t2c
213 : INTEGER, DIMENSION(3) :: nblks_total, pcoord, pdims, pdims_t3c
214 100 : INTEGER, DIMENSION(:), POINTER :: col_bsize, row_bsize
215 : LOGICAL :: do_periodic, use_virial
216 : REAL(dp) :: compression_factor, eps_pgf_orb, &
217 : eps_pgf_orb_old, memory, occ
218 : TYPE(cell_type), POINTER :: cell
219 : TYPE(cp_blacs_env_type), POINTER :: blacs_env
220 : TYPE(dbcsr_distribution_type) :: dbcsr_dist
221 100 : TYPE(dbcsr_p_type), DIMENSION(:), POINTER :: matrix_s, rho_ao
222 : TYPE(dbcsr_type) :: dbcsr_work, dbcsr_work2, dbcsr_work3
223 100 : TYPE(dbcsr_type), DIMENSION(1) :: t_2c_int_tmp
224 350 : TYPE(dbcsr_type), DIMENSION(1, 3) :: t_2c_der_tmp
225 250 : TYPE(dbt_pgrid_type) :: pgrid_t2c, pgrid_t3c
226 1000 : TYPE(dbt_type) :: t_2c_template, t_2c_tmp, t_3c_template
227 50 : TYPE(dbt_type), ALLOCATABLE, DIMENSION(:, :, :) :: t_3c_der_AO_prv, t_3c_der_RI_prv
228 : TYPE(dft_control_type), POINTER :: dft_control
229 : TYPE(distribution_2d_type), POINTER :: dist_2d
230 : TYPE(distribution_3d_type) :: dist_3d, dist_vir
231 : TYPE(gto_basis_set_p_type), ALLOCATABLE, &
232 50 : DIMENSION(:), TARGET :: basis_set_ao, basis_set_ri_aux
233 : TYPE(gto_basis_set_type), POINTER :: orb_basis, ri_basis
234 : TYPE(libint_potential_type) :: identity_pot
235 50 : TYPE(mp_cart_type) :: mp_comm_t3c, mp_comm_vir
236 : TYPE(mp_para_env_type), POINTER :: para_env
237 : TYPE(neighbor_list_3c_type) :: nl_3c
238 : TYPE(neighbor_list_set_p_type), DIMENSION(:), &
239 50 : POINTER :: nl_2c
240 50 : TYPE(particle_type), DIMENSION(:), POINTER :: particle_set
241 50 : TYPE(qs_kind_type), DIMENSION(:), POINTER :: qs_kind_set
242 : TYPE(qs_rho_type), POINTER :: rho
243 : TYPE(section_vals_type), POINTER :: qs_section
244 : TYPE(virial_type), POINTER :: virial
245 :
246 50 : NULLIFY (dft_control, para_env, particle_set, qs_kind_set, dist_2d, nl_2c, blacs_env, matrix_s, &
247 50 : rho, rho_ao, cell, qs_section, orb_basis, ri_basis, virial)
248 :
249 50 : CALL cite_reference(Bussy2023)
250 :
251 50 : CALL timeset(routineN, handle)
252 :
253 : CALL get_qs_env(qs_env, natom=natom, nkind=nkind, dft_control=dft_control, para_env=para_env, &
254 50 : particle_set=particle_set, qs_kind_set=qs_kind_set, cell=cell, virial=virial)
255 50 : IF (dft_control%qs_control%gapw) THEN
256 0 : CPABORT("Low-scaling RPA/SOS-MP2 forces only available with GPW")
257 : END IF
258 :
259 50 : use_virial = virial%pv_availability .AND. (.NOT. virial%pv_numer)
260 :
261 50 : do_periodic = .FALSE.
262 128 : IF (ANY(cell%perd == 1)) do_periodic = .TRUE.
263 50 : force_data%do_periodic = do_periodic
264 :
265 : !Dealing with the 3-center derivatives
266 50 : pdims_t3c = 0
267 50 : CALL dbt_pgrid_create(para_env, pdims_t3c, pgrid_t3c)
268 :
269 : !Make sure we use the proper QS EPS_PGF_ORB values
270 50 : qs_section => section_vals_get_subs_vals(qs_env%input, "DFT%QS")
271 50 : CALL section_vals_val_get(qs_section, "EPS_PGF_ORB", n_rep_val=n_rep)
272 50 : IF (n_rep /= 0) THEN
273 0 : CALL section_vals_val_get(qs_section, "EPS_PGF_ORB", r_val=eps_pgf_orb)
274 : ELSE
275 50 : CALL section_vals_val_get(qs_section, "EPS_DEFAULT", r_val=eps_pgf_orb)
276 50 : eps_pgf_orb = SQRT(eps_pgf_orb)
277 : END IF
278 50 : eps_pgf_orb_old = dft_control%qs_control%eps_pgf_orb
279 :
280 200 : ALLOCATE (sizes_RI(natom), sizes_AO(natom))
281 400 : ALLOCATE (basis_set_ri_aux(nkind), basis_set_ao(nkind))
282 50 : CALL basis_set_list_setup(basis_set_ri_aux, "RI_AUX", qs_kind_set)
283 50 : CALL get_particle_set(particle_set, qs_kind_set, nsgf=sizes_RI, basis=basis_set_ri_aux)
284 50 : CALL basis_set_list_setup(basis_set_ao, "ORB", qs_kind_set)
285 50 : CALL get_particle_set(particle_set, qs_kind_set, nsgf=sizes_AO, basis=basis_set_ao)
286 :
287 150 : DO ibasis = 1, SIZE(basis_set_ao)
288 100 : orb_basis => basis_set_ao(ibasis)%gto_basis_set
289 100 : CALL init_interaction_radii_orb_basis(orb_basis, eps_pgf_orb)
290 100 : ri_basis => basis_set_ri_aux(ibasis)%gto_basis_set
291 150 : CALL init_interaction_radii_orb_basis(ri_basis, eps_pgf_orb)
292 : END DO
293 :
294 : CALL create_3c_tensor(t_3c_template, dist_RI, dist_AO_1, dist_AO_2, pgrid_t3c, &
295 50 : sizes_RI, sizes_AO, sizes_AO, map1=[1], map2=[2, 3], name="der (RI AO | AO)")
296 :
297 1550 : ALLOCATE (t_3c_der_RI_prv(1, 1, 3), t_3c_der_AO_prv(1, 1, 3))
298 200 : DO i_xyz = 1, 3
299 150 : CALL dbt_create(t_3c_template, t_3c_der_RI_prv(1, 1, i_xyz))
300 200 : CALL dbt_create(t_3c_template, t_3c_der_AO_prv(1, 1, i_xyz))
301 : END DO
302 :
303 50 : IF (use_virial) THEN
304 52 : ALLOCATE (force_data%t_3c_virial, force_data%t_3c_virial_split)
305 4 : CALL dbt_create(t_3c_template, force_data%t_3c_virial)
306 4 : CALL dbt_create(t_3c_M, force_data%t_3c_virial_split)
307 : END IF
308 50 : CALL dbt_destroy(t_3c_template)
309 :
310 50 : CALL dbt_mp_environ_pgrid(pgrid_t3c, pdims, pcoord)
311 50 : CALL mp_comm_t3c%create(pgrid_t3c%mp_comm_2d, 3, pdims)
312 : CALL distribution_3d_create(dist_3d, dist_RI, dist_AO_1, dist_AO_2, &
313 50 : nkind, particle_set, mp_comm_t3c, own_comm=.TRUE.)
314 :
315 : !In case of virial, we need to store the 3c_nl
316 50 : IF (use_virial) THEN
317 4 : ALLOCATE (force_data%nl_3c)
318 4 : CALL mp_comm_vir%create(pgrid_t3c%mp_comm_2d, 3, pdims)
319 : CALL distribution_3d_create(dist_vir, dist_RI, dist_AO_1, dist_AO_2, &
320 4 : nkind, particle_set, mp_comm_vir, own_comm=.TRUE.)
321 : CALL build_3c_neighbor_lists(force_data%nl_3c, basis_set_ri_aux, basis_set_ao, basis_set_ao, &
322 : dist_vir, mp2_env%ri_metric, "RPA_3c_nl", qs_env, op_pos=1, &
323 4 : sym_jk=.FALSE., own_dist=.TRUE.)
324 : END IF
325 :
326 : CALL build_3c_neighbor_lists(nl_3c, basis_set_ri_aux, basis_set_ao, basis_set_ao, dist_3d, &
327 : mp2_env%ri_metric, "RPA_3c_nl", qs_env, op_pos=1, sym_jk=.TRUE., &
328 50 : own_dist=.TRUE.)
329 50 : DEALLOCATE (dist_RI, dist_AO_1, dist_AO_2)
330 :
331 : !Prepare the resulting 3c tensors in the format of t_3c_M for compatible traces: (RI|AO AO), split blocks
332 50 : CALL dbt_get_info(t_3c_M, nblks_total=nblks_total)
333 250 : ALLOCATE (force_data%bsizes_RI_split(nblks_total(1)), force_data%bsizes_AO_split(nblks_total(2)))
334 50 : CALL dbt_get_info(t_3c_M, blk_size_1=force_data%bsizes_RI_split, blk_size_2=force_data%bsizes_AO_split)
335 200 : DO i_xyz = 1, 3
336 150 : CALL dbt_create(t_3c_M, force_data%t_3c_der_RI(i_xyz))
337 200 : CALL dbt_create(t_3c_M, force_data%t_3c_der_AO(i_xyz))
338 : END DO
339 :
340 : !Keep track of atom index corresponding to split blocks
341 100 : ALLOCATE (force_data%idx_to_at_RI(nblks_total(1)))
342 50 : CALL get_idx_to_atom(force_data%idx_to_at_RI, force_data%bsizes_RI_split, sizes_RI)
343 :
344 100 : ALLOCATE (force_data%idx_to_at_AO(nblks_total(2)))
345 50 : CALL get_idx_to_atom(force_data%idx_to_at_AO, force_data%bsizes_AO_split, sizes_AO)
346 :
347 50 : n_mem = mp2_env%ri_rpa_im_time%cut_memory
348 50 : CALL create_tensor_batches(sizes_RI, n_mem, dummy_start, dummy_end, start_blocks, end_blocks)
349 50 : DEALLOCATE (dummy_start, dummy_end)
350 :
351 212300 : ALLOCATE (force_data%t_3c_der_AO_comp(n_mem, 3), force_data%t_3c_der_RI_comp(n_mem, 3))
352 1100 : ALLOCATE (force_data%t_3c_der_AO_ind(n_mem, 3), force_data%t_3c_der_RI_ind(n_mem, 3))
353 :
354 50 : memory = 0.0_dp
355 50 : nze_tot = 0
356 150 : DO i_mem = 1, n_mem
357 : CALL build_3c_derivatives(t_3c_der_RI_prv, t_3c_der_AO_prv, mp2_env%ri_rpa_im_time%eps_filter, &
358 : qs_env, nl_3c, basis_set_ri_aux, basis_set_ao, basis_set_ao, &
359 : mp2_env%ri_metric, der_eps=mp2_env%ri_rpa_im_time%eps_filter, op_pos=1, &
360 300 : bounds_i=[start_blocks(i_mem), end_blocks(i_mem)])
361 :
362 450 : DO i_xyz = 1, 3
363 300 : CALL dbt_copy(t_3c_der_RI_prv(1, 1, i_xyz), force_data%t_3c_der_RI(i_xyz), move_data=.TRUE.)
364 300 : CALL dbt_filter(force_data%t_3c_der_RI(i_xyz), mp2_env%ri_rpa_im_time%eps_filter)
365 300 : CALL get_tensor_occupancy(force_data%t_3c_der_RI(i_xyz), nze, occ)
366 300 : nze_tot = nze_tot + nze
367 :
368 300 : CALL alloc_containers(force_data%t_3c_der_RI_comp(i_mem, i_xyz), 1)
369 : CALL compress_tensor(force_data%t_3c_der_RI(i_xyz), force_data%t_3c_der_RI_ind(i_mem, i_xyz)%ind, &
370 300 : force_data%t_3c_der_RI_comp(i_mem, i_xyz), mp2_env%ri_rpa_im_time%eps_compress, memory)
371 300 : CALL dbt_clear(force_data%t_3c_der_RI(i_xyz))
372 :
373 300 : CALL dbt_copy(t_3c_der_AO_prv(1, 1, i_xyz), force_data%t_3c_der_AO(i_xyz), move_data=.TRUE.)
374 300 : CALL dbt_filter(force_data%t_3c_der_AO(i_xyz), mp2_env%ri_rpa_im_time%eps_filter)
375 300 : CALL get_tensor_occupancy(force_data%t_3c_der_AO(i_xyz), nze, occ)
376 300 : nze_tot = nze_tot + nze
377 :
378 300 : CALL alloc_containers(force_data%t_3c_der_AO_comp(i_mem, i_xyz), 1)
379 : CALL compress_tensor(force_data%t_3c_der_AO(i_xyz), force_data%t_3c_der_AO_ind(i_mem, i_xyz)%ind, &
380 300 : force_data%t_3c_der_AO_comp(i_mem, i_xyz), mp2_env%ri_rpa_im_time%eps_compress, memory)
381 1000 : CALL dbt_clear(force_data%t_3c_der_AO(i_xyz))
382 : END DO
383 : END DO
384 50 : CALL neighbor_list_3c_destroy(nl_3c)
385 200 : DO i_xyz = 1, 3
386 150 : CALL dbt_destroy(t_3c_der_RI_prv(1, 1, i_xyz))
387 200 : CALL dbt_destroy(t_3c_der_AO_prv(1, 1, i_xyz))
388 : END DO
389 :
390 50 : CALL para_env%sum(memory)
391 50 : compression_factor = REAL(nze_tot, dp)*1.0E-06_dp*8.0_dp/memory
392 50 : IF (unit_nr > 0) THEN
393 : WRITE (UNIT=unit_nr, FMT="((T3,A,T66,F11.2,A4))") &
394 25 : "MEMORY_INFO| Memory for 3-center derivatives (compressed):", memory, ' MiB'
395 :
396 : WRITE (UNIT=unit_nr, FMT="((T3,A,T60,F21.2))") &
397 25 : "MEMORY_INFO| Compression factor: ", compression_factor
398 : END IF
399 :
400 : !Dealing with the 2-center derivatives
401 50 : CALL get_qs_env(qs_env, distribution_2d=dist_2d, blacs_env=blacs_env, matrix_s=matrix_s)
402 50 : CALL cp_dbcsr_dist2d_to_dist(dist_2d, dbcsr_dist)
403 150 : ALLOCATE (row_bsize(SIZE(sizes_RI)))
404 100 : ALLOCATE (col_bsize(SIZE(sizes_RI)))
405 212 : row_bsize(:) = sizes_RI(:)
406 212 : col_bsize(:) = sizes_RI(:)
407 :
408 50 : pdims_t2c = 0
409 50 : CALL dbt_pgrid_create(para_env, pdims_t2c, pgrid_t2c)
410 : CALL create_2c_tensor(t_2c_template, dist1, dist2, pgrid_t2c, force_data%bsizes_RI_split, &
411 50 : force_data%bsizes_RI_split, name='(RI| RI)')
412 50 : DEALLOCATE (dist1, dist2)
413 :
414 50 : CALL dbcsr_create(t_2c_int_tmp(1), "(P|Q) RPA", dbcsr_dist, dbcsr_type_symmetric, row_bsize, col_bsize)
415 200 : DO i_xyz = 1, 3
416 : CALL dbcsr_create(t_2c_der_tmp(1, i_xyz), "(P|Q) RPA der", dbcsr_dist, &
417 200 : dbcsr_type_antisymmetric, row_bsize, col_bsize)
418 : END DO
419 :
420 50 : IF (use_virial) THEN
421 4 : ALLOCATE (force_data%RI_virial_pot, force_data%RI_virial_met)
422 : CALL dbcsr_create(force_data%RI_virial_pot, "RI_virial", dbcsr_dist, &
423 4 : dbcsr_type_no_symmetry, row_bsize, col_bsize)
424 : CALL dbcsr_create(force_data%RI_virial_met, "RI_virial", dbcsr_dist, &
425 4 : dbcsr_type_no_symmetry, row_bsize, col_bsize)
426 : END IF
427 :
428 : ! Main (P|Q) integrals and derivatives
429 : ! Integrals are passed as a full matrix => convert to DBCSR
430 50 : CALL dbcsr_create(dbcsr_work, template=t_2c_int_tmp(1))
431 50 : CALL copy_fm_to_dbcsr(fm_matrix_PQ, dbcsr_work)
432 :
433 : ! We need the +/- square root of (P|Q)
434 50 : CALL dbcsr_create(dbcsr_work2, template=t_2c_int_tmp(1))
435 50 : CALL dbcsr_create(dbcsr_work3, template=t_2c_int_tmp(1))
436 50 : CALL dbcsr_copy(dbcsr_work2, dbcsr_work)
437 50 : CALL cp_dbcsr_power(dbcsr_work, -0.5_dp, 1.0E-7_dp, n_dependent, para_env, blacs_env) !1.0E-7 ev qunenching thresh
438 :
439 : ! Transfer to tensor format with split blocks
440 50 : CALL dbt_create(dbcsr_work, t_2c_tmp)
441 50 : CALL dbt_copy_matrix_to_tensor(dbcsr_work, t_2c_tmp)
442 50 : CALL dbt_create(t_2c_template, force_data%t_2c_pot_msqrt)
443 50 : CALL dbt_copy(t_2c_tmp, force_data%t_2c_pot_msqrt, move_data=.TRUE.)
444 50 : CALL dbt_filter(force_data%t_2c_pot_msqrt, mp2_env%ri_rpa_im_time%eps_filter)
445 :
446 50 : CALL dbcsr_multiply('N', 'N', 1.0_dp, dbcsr_work2, dbcsr_work, 0.0_dp, dbcsr_work3)
447 50 : CALL dbt_copy_matrix_to_tensor(dbcsr_work3, t_2c_tmp)
448 50 : CALL dbt_create(t_2c_template, force_data%t_2c_pot_psqrt)
449 50 : CALL dbt_copy(t_2c_tmp, force_data%t_2c_pot_psqrt, move_data=.TRUE.)
450 50 : CALL dbt_filter(force_data%t_2c_pot_psqrt, mp2_env%ri_rpa_im_time%eps_filter)
451 50 : CALL dbt_destroy(t_2c_tmp)
452 50 : CALL dbcsr_release(dbcsr_work2)
453 50 : CALL dbcsr_release(dbcsr_work3)
454 50 : CALL dbcsr_clear(dbcsr_work)
455 :
456 : ! Deal with the 2c potential derivatives. Only precompute if not in PBCs
457 50 : IF (.NOT. do_periodic) THEN
458 : CALL build_2c_neighbor_lists(nl_2c, basis_set_ri_aux, basis_set_ri_aux, mp2_env%potential_parameter, &
459 26 : "RPA_2c_nl_pot", qs_env, sym_ij=.TRUE., dist_2d=dist_2d)
460 : CALL build_2c_derivatives(t_2c_der_tmp, mp2_env%ri_rpa_im_time%eps_filter, qs_env, nl_2c, &
461 26 : basis_set_ri_aux, basis_set_ri_aux, mp2_env%potential_parameter)
462 26 : CALL release_neighbor_list_sets(nl_2c)
463 :
464 104 : DO i_xyz = 1, 3
465 78 : CALL dbt_create(t_2c_der_tmp(1, i_xyz), t_2c_tmp)
466 78 : CALL dbt_copy_matrix_to_tensor(t_2c_der_tmp(1, i_xyz), t_2c_tmp)
467 78 : CALL dbt_create(t_2c_template, force_data%t_2c_der_pot(i_xyz))
468 78 : CALL dbt_copy(t_2c_tmp, force_data%t_2c_der_pot(i_xyz), move_data=.TRUE.)
469 78 : CALL dbt_filter(force_data%t_2c_der_pot(i_xyz), mp2_env%ri_rpa_im_time%eps_filter)
470 78 : CALL dbt_destroy(t_2c_tmp)
471 104 : CALL dbcsr_clear(t_2c_der_tmp(1, i_xyz))
472 : END DO
473 :
474 26 : IF (use_virial) THEN
475 : CALL build_2c_neighbor_lists(force_data%nl_2c_pot, basis_set_ri_aux, basis_set_ri_aux, &
476 : mp2_env%potential_parameter, "RPA_2c_nl_pot", qs_env, &
477 0 : sym_ij=.FALSE., dist_2d=dist_2d)
478 : END IF
479 : END IF
480 : ! Create a G_PQ matrix to collect the terms for the force trace in the periodic case
481 50 : CALL dbcsr_create(force_data%G_PQ, "G_PQ", dbcsr_dist, dbcsr_type_no_symmetry, row_bsize, col_bsize)
482 :
483 : ! we need the RI metric derivatives and the inverse of the integrals
484 : CALL build_2c_neighbor_lists(nl_2c, basis_set_ri_aux, basis_set_ri_aux, mp2_env%ri_metric, &
485 50 : "RPA_2c_nl_metric", qs_env, sym_ij=.TRUE., dist_2d=dist_2d)
486 : CALL build_2c_integrals(t_2c_int_tmp, mp2_env%ri_rpa_im_time%eps_filter, qs_env, nl_2c, &
487 50 : basis_set_ri_aux, basis_set_ri_aux, mp2_env%ri_metric)
488 : CALL build_2c_derivatives(t_2c_der_tmp, mp2_env%ri_rpa_im_time%eps_filter, qs_env, nl_2c, &
489 50 : basis_set_ri_aux, basis_set_ri_aux, mp2_env%ri_metric)
490 50 : CALL release_neighbor_list_sets(nl_2c)
491 :
492 50 : IF (use_virial) THEN
493 : CALL build_2c_neighbor_lists(force_data%nl_2c_met, basis_set_ri_aux, basis_set_ri_aux, &
494 : mp2_env%ri_metric, "RPA_2c_nl_metric", qs_env, sym_ij=.FALSE., &
495 4 : dist_2d=dist_2d)
496 : END IF
497 :
498 50 : CALL dbcsr_copy(dbcsr_work, t_2c_int_tmp(1))
499 50 : CALL cp_dbcsr_cholesky_decompose(dbcsr_work, para_env=para_env, blacs_env=blacs_env)
500 50 : CALL cp_dbcsr_cholesky_invert(dbcsr_work, para_env=para_env, blacs_env=blacs_env, uplo_to_full=.TRUE.)
501 :
502 50 : CALL dbt_create(dbcsr_work, t_2c_tmp)
503 50 : CALL dbt_copy_matrix_to_tensor(dbcsr_work, t_2c_tmp)
504 50 : CALL dbt_create(t_2c_template, force_data%t_2c_inv_metric)
505 50 : CALL dbt_copy(t_2c_tmp, force_data%t_2c_inv_metric, move_data=.TRUE.)
506 50 : CALL dbt_filter(force_data%t_2c_inv_metric, mp2_env%ri_rpa_im_time%eps_filter)
507 50 : CALL dbt_destroy(t_2c_tmp)
508 50 : CALL dbcsr_clear(dbcsr_work)
509 50 : CALL dbcsr_clear(t_2c_int_tmp(1))
510 :
511 200 : DO i_xyz = 1, 3
512 150 : CALL dbt_create(t_2c_der_tmp(1, i_xyz), t_2c_tmp)
513 150 : CALL dbt_copy_matrix_to_tensor(t_2c_der_tmp(1, i_xyz), t_2c_tmp)
514 150 : CALL dbt_create(t_2c_template, force_data%t_2c_der_metric(i_xyz))
515 150 : CALL dbt_copy(t_2c_tmp, force_data%t_2c_der_metric(i_xyz), move_data=.TRUE.)
516 150 : CALL dbt_filter(force_data%t_2c_der_metric(i_xyz), mp2_env%ri_rpa_im_time%eps_filter)
517 150 : CALL dbt_destroy(t_2c_tmp)
518 200 : CALL dbcsr_clear(t_2c_der_tmp(1, i_xyz))
519 : END DO
520 :
521 : !Pre-calculate matrix K = metric^-1 * V^0.5
522 50 : CALL dbt_create(t_2c_template, force_data%t_2c_K)
523 : CALL dbt_contract(1.0_dp, force_data%t_2c_inv_metric, force_data%t_2c_pot_psqrt, &
524 : 0.0_dp, force_data%t_2c_K, &
525 : contract_1=[2], notcontract_1=[1], &
526 : contract_2=[1], notcontract_2=[2], &
527 50 : map_1=[1], map_2=[2], filter_eps=mp2_env%ri_rpa_im_time%eps_filter)
528 :
529 : ! Finally, we need the overlap matrix derivative and the inverse of the integrals
530 50 : CALL dbt_destroy(t_2c_template)
531 50 : CALL dbcsr_release(dbcsr_work)
532 50 : CALL dbcsr_release(t_2c_int_tmp(1))
533 200 : DO i_xyz = 1, 3
534 200 : CALL dbcsr_release(t_2c_der_tmp(1, i_xyz))
535 : END DO
536 :
537 50 : DEALLOCATE (row_bsize, col_bsize)
538 150 : ALLOCATE (row_bsize(SIZE(sizes_AO)))
539 100 : ALLOCATE (col_bsize(SIZE(sizes_AO)))
540 212 : row_bsize(:) = sizes_AO(:)
541 212 : col_bsize(:) = sizes_AO(:)
542 :
543 : CALL create_2c_tensor(t_2c_template, dist1, dist2, pgrid_t2c, force_data%bsizes_AO_split, &
544 50 : force_data%bsizes_AO_split, name='(AO| AO)')
545 50 : DEALLOCATE (dist1, dist2)
546 :
547 200 : DO i_xyz = 1, 3
548 : CALL dbcsr_create(t_2c_der_tmp(1, i_xyz), "(P|Q) RPA der", dbcsr_dist, &
549 200 : dbcsr_type_antisymmetric, row_bsize, col_bsize)
550 : END DO
551 :
552 50 : identity_pot%potential_type = do_potential_id
553 : CALL build_2c_neighbor_lists(nl_2c, basis_set_ao, basis_set_ao, identity_pot, &
554 50 : "RPA_2c_nl_metric", qs_env, sym_ij=.TRUE., dist_2d=dist_2d)
555 : CALL build_2c_derivatives(t_2c_der_tmp, mp2_env%ri_rpa_im_time%eps_filter, qs_env, nl_2c, &
556 50 : basis_set_ao, basis_set_ao, identity_pot)
557 50 : CALL release_neighbor_list_sets(nl_2c)
558 :
559 50 : IF (use_virial) THEN
560 : CALL build_2c_neighbor_lists(force_data%nl_2c_ovlp, basis_set_ao, basis_set_ao, identity_pot, &
561 4 : "RPA_2c_nl_metric", qs_env, sym_ij=.FALSE., dist_2d=dist_2d)
562 : END IF
563 :
564 50 : CALL dbcsr_create(force_data%inv_ovlp, template=matrix_s(1)%matrix)
565 50 : CALL dbcsr_copy(force_data%inv_ovlp, matrix_s(1)%matrix)
566 50 : CALL cp_dbcsr_cholesky_decompose(force_data%inv_ovlp, para_env=para_env, blacs_env=blacs_env)
567 50 : CALL cp_dbcsr_cholesky_invert(force_data%inv_ovlp, para_env=para_env, blacs_env=blacs_env, uplo_to_full=.TRUE.)
568 :
569 200 : DO i_xyz = 1, 3
570 150 : CALL dbt_create(t_2c_der_tmp(1, i_xyz), t_2c_tmp)
571 150 : CALL dbt_copy_matrix_to_tensor(t_2c_der_tmp(1, i_xyz), t_2c_tmp)
572 150 : CALL dbt_create(t_2c_template, force_data%t_2c_der_ovlp(i_xyz))
573 150 : CALL dbt_copy(t_2c_tmp, force_data%t_2c_der_ovlp(i_xyz), move_data=.TRUE.)
574 150 : CALL dbt_filter(force_data%t_2c_der_ovlp(i_xyz), mp2_env%ri_rpa_im_time%eps_filter)
575 150 : CALL dbt_destroy(t_2c_tmp)
576 200 : CALL dbcsr_clear(t_2c_der_tmp(1, i_xyz))
577 : END DO
578 :
579 : !Create the rest of the 2-center AO tensors
580 50 : nspins = dft_control%nspins
581 324 : ALLOCATE (force_data%P_virt(nspins), force_data%P_occ(nspins))
582 274 : ALLOCATE (force_data%sum_YP_tau(nspins), force_data%sum_O_tau(nspins))
583 112 : DO ispin = 1, nspins
584 62 : ALLOCATE (force_data%P_virt(ispin)%matrix, force_data%P_occ(ispin)%matrix)
585 62 : ALLOCATE (force_data%sum_YP_tau(ispin)%matrix, force_data%sum_O_tau(ispin)%matrix)
586 62 : CALL dbcsr_create(force_data%P_virt(ispin)%matrix, template=matrix_s(1)%matrix)
587 62 : CALL dbcsr_create(force_data%P_occ(ispin)%matrix, template=matrix_s(1)%matrix)
588 62 : CALL dbcsr_create(force_data%sum_O_tau(ispin)%matrix, template=matrix_s(1)%matrix)
589 62 : CALL dbcsr_create(force_data%sum_YP_tau(ispin)%matrix, template=matrix_s(1)%matrix)
590 :
591 62 : CALL dbcsr_copy(force_data%sum_O_tau(ispin)%matrix, matrix_s(1)%matrix)
592 62 : CALL dbcsr_copy(force_data%sum_YP_tau(ispin)%matrix, matrix_s(1)%matrix)
593 :
594 62 : CALL dbcsr_set(force_data%sum_O_tau(ispin)%matrix, 0.0_dp)
595 112 : CALL dbcsr_set(force_data%sum_YP_tau(ispin)%matrix, 0.0_dp)
596 : END DO
597 :
598 : !Populate the density matrices: 1 = P_virt*S +P_occ*S ==> P_virt = S^-1 - P_occ
599 50 : CALL get_qs_env(qs_env, rho=rho)
600 50 : CALL qs_rho_get(rho, rho_ao=rho_ao)
601 50 : CALL dbcsr_copy(force_data%P_occ(1)%matrix, rho_ao(1)%matrix)
602 50 : IF (nspins == 1) THEN
603 38 : CALL dbcsr_scale(force_data%P_occ(1)%matrix, 0.5_dp) !because double occupency
604 : ELSE
605 12 : CALL dbcsr_copy(force_data%P_occ(2)%matrix, rho_ao(2)%matrix)
606 : END IF
607 112 : DO ispin = 1, nspins
608 62 : CALL dbcsr_copy(force_data%P_virt(ispin)%matrix, force_data%inv_ovlp)
609 112 : CALL dbcsr_add(force_data%P_virt(ispin)%matrix, force_data%P_occ(ispin)%matrix, 1.0_dp, -1.0_dp)
610 : END DO
611 :
612 150 : DO ibasis = 1, SIZE(basis_set_ao)
613 100 : orb_basis => basis_set_ao(ibasis)%gto_basis_set
614 100 : CALL init_interaction_radii_orb_basis(orb_basis, eps_pgf_orb_old)
615 100 : ri_basis => basis_set_ri_aux(ibasis)%gto_basis_set
616 150 : CALL init_interaction_radii_orb_basis(ri_basis, eps_pgf_orb_old)
617 : END DO
618 :
619 50 : CALL dbt_destroy(t_2c_template)
620 50 : CALL dbcsr_release(dbcsr_work)
621 200 : DO i_xyz = 1, 3
622 200 : CALL dbcsr_release(t_2c_der_tmp(1, i_xyz))
623 : END DO
624 50 : DEALLOCATE (row_bsize, col_bsize)
625 50 : CALL dbt_pgrid_destroy(pgrid_t3c)
626 50 : CALL dbt_pgrid_destroy(pgrid_t2c)
627 50 : CALL dbcsr_distribution_release(dbcsr_dist)
628 50 : CALL timestop(handle)
629 :
630 600 : END SUBROUTINE init_im_time_forces
631 :
632 : ! **************************************************************************************************
633 : !> \brief Updates the cubic-scaling SOS-Laplace-MP2 contribution to the forces at each quadrature point
634 : !> \param force_data ...
635 : !> \param mat_P_omega ...
636 : !> \param t_3c_M ...
637 : !> \param t_3c_O ...
638 : !> \param t_3c_O_compressed ...
639 : !> \param t_3c_O_ind ...
640 : !> \param fm_mo_coeff real Gamma-point MO coefficients ...
641 : !> \param homo ...
642 : !> \param starts_array_mc ...
643 : !> \param ends_array_mc ...
644 : !> \param starts_array_mc_block ...
645 : !> \param ends_array_mc_block ...
646 : !> \param nmo ...
647 : !> \param Eigenval ...
648 : !> \param grid ...
649 : !> \param cut_memory ...
650 : !> \param Pspin ...
651 : !> \param Qspin ...
652 : !> \param open_shell ...
653 : !> \param unit_nr ...
654 : !> \param dbcsr_time ...
655 : !> \param dbcsr_nflop ...
656 : !> \param mp2_env ...
657 : !> \param qs_env ...
658 : !> \note In open-shell, we need to take Q from one spin, and everything from the other
659 : ! **************************************************************************************************
660 130 : SUBROUTINE calc_laplace_loop_forces(force_data, mat_P_omega, t_3c_M, t_3c_O, t_3c_O_compressed, &
661 52 : t_3c_O_ind, fm_mo_coeff, homo, starts_array_mc, ends_array_mc, &
662 26 : starts_array_mc_block, ends_array_mc_block, &
663 26 : nmo, Eigenval, grid, cut_memory, Pspin, Qspin, &
664 : open_shell, unit_nr, dbcsr_time, dbcsr_nflop, mp2_env, qs_env)
665 :
666 : TYPE(im_time_force_type), INTENT(INOUT) :: force_data
667 : TYPE(dbcsr_p_type), DIMENSION(:, :), INTENT(INOUT) :: mat_P_omega
668 : TYPE(dbt_type), INTENT(INOUT) :: t_3c_M, t_3c_O
669 : TYPE(hfx_compression_type), DIMENSION(:) :: t_3c_O_compressed
670 : TYPE(block_ind_type), DIMENSION(:), INTENT(INOUT) :: t_3c_O_ind
671 : TYPE(cp_fm_type), DIMENSION(:), INTENT(IN) :: fm_mo_coeff
672 : INTEGER, DIMENSION(:), INTENT(IN) :: homo, starts_array_mc, ends_array_mc, &
673 : starts_array_mc_block, &
674 : ends_array_mc_block
675 : INTEGER, INTENT(IN) :: nmo
676 : REAL(KIND=dp), DIMENSION(:, :), INTENT(IN) :: Eigenval
677 : TYPE(time_frequency_grid_type), INTENT(IN) :: grid
678 : INTEGER, INTENT(IN) :: cut_memory, Pspin, Qspin
679 : LOGICAL, INTENT(IN) :: open_shell
680 : INTEGER, INTENT(IN) :: unit_nr
681 : REAL(dp), INTENT(INOUT) :: dbcsr_time
682 : INTEGER(int_8), INTENT(INOUT) :: dbcsr_nflop
683 : TYPE(mp2_type) :: mp2_env
684 : TYPE(qs_environment_type), POINTER :: qs_env
685 :
686 : CHARACTER(LEN=*), PARAMETER :: routineN = 'calc_laplace_loop_forces'
687 :
688 : INTEGER :: dummy_int, handle, handle2, i_mem, i_xyz, ibasis, ispin, j_xyz, jquad, k_xyz, &
689 : n_mem_RI, n_rep, natom, nkind, nspins, num_integ_points, unit_nr_dbcsr
690 : INTEGER(int_8) :: flop, nze, nze_ddint, nze_der_AO, &
691 : nze_der_RI, nze_KQK
692 26 : INTEGER, ALLOCATABLE, DIMENSION(:) :: atom_of_kind, batch_blk_end, &
693 26 : batch_blk_start, batch_end_RI, &
694 26 : batch_start_RI, kind_of, mc_ranges, &
695 26 : mc_ranges_RI
696 26 : INTEGER, DIMENSION(:, :), POINTER :: dummy_ptr
697 : LOGICAL :: memory_info, use_virial
698 : REAL(dp) :: eps_filter, eps_pgf_orb, &
699 : eps_pgf_orb_old, fac, occ, occ_ddint, &
700 : occ_der_AO, occ_der_RI, occ_KQK, &
701 : omega, pref, t1, t2, tau
702 : REAL(dp), DIMENSION(3, 3) :: work_virial, work_virial_ovlp
703 26 : TYPE(atomic_kind_type), DIMENSION(:), POINTER :: atomic_kind_set
704 : TYPE(cell_type), POINTER :: cell
705 26 : TYPE(dbcsr_p_type), DIMENSION(:), POINTER :: matrix_ks, matrix_s
706 26 : TYPE(dbcsr_p_type), DIMENSION(:, :, :), POINTER :: propagator
707 : TYPE(dbcsr_type) :: dbcsr_work1, dbcsr_work2, dbcsr_work3, &
708 : exp_occ, exp_virt, R_occ, R_virt, &
709 : virial_ovlp, Y_1, Y_2
710 1274 : TYPE(dbt_type) :: t_2c_AO, t_2c_RI, t_2c_RI_2, t_2c_tmp, t_3c_0, t_3c_1, t_3c_3, t_3c_4, &
711 1274 : t_3c_5, t_3c_6, t_3c_7, t_3c_8, t_3c_help_1, t_3c_help_2, t_3c_ints, t_3c_sparse, &
712 1456 : t_3c_work, t_dm_occ, t_dm_virt, t_KQKT, t_M_occ, t_M_virt, t_Q, t_R_occ, t_R_virt
713 26 : TYPE(dbt_type), ALLOCATABLE, DIMENSION(:) :: t_P
714 : TYPE(dft_control_type), POINTER :: dft_control
715 : TYPE(gto_basis_set_p_type), ALLOCATABLE, &
716 26 : DIMENSION(:), TARGET :: basis_set_ao, basis_set_ri_aux
717 : TYPE(gto_basis_set_type), POINTER :: orb_basis, ri_basis
718 : TYPE(libint_potential_type) :: identity_pot
719 : TYPE(mp_para_env_type), POINTER :: para_env
720 26 : TYPE(particle_type), DIMENSION(:), POINTER :: particle_set
721 26 : TYPE(qs_force_type), DIMENSION(:), POINTER :: force
722 26 : TYPE(qs_kind_type), DIMENSION(:), POINTER :: qs_kind_set
723 : TYPE(section_vals_type), POINTER :: qs_section
724 : TYPE(virial_type), POINTER :: virial
725 :
726 26 : NULLIFY (matrix_s, dummy_ptr, atomic_kind_set, force, matrix_s, matrix_ks)
727 26 : NULLIFY (dft_control, virial, particle_set, cell, para_env, orb_basis, ri_basis, qs_section)
728 26 : NULLIFY (qs_kind_set)
729 :
730 26 : CALL timeset(routineN, handle)
731 :
732 26 : num_integ_points = SIZE(grid%imaginary_time)
733 :
734 26 : NULLIFY (propagator)
735 :
736 : CALL get_qs_env(qs_env, matrix_s=matrix_s, natom=natom, atomic_kind_set=atomic_kind_set, &
737 : force=force, matrix_ks=matrix_ks, dft_control=dft_control, virial=virial, &
738 : particle_set=particle_set, cell=cell, para_env=para_env, nkind=nkind, &
739 26 : qs_kind_set=qs_kind_set)
740 26 : eps_filter = mp2_env%ri_rpa_im_time%eps_filter
741 26 : nspins = dft_control%nspins
742 :
743 26 : memory_info = mp2_env%ri_rpa_im_time%memory_info
744 26 : IF (memory_info) THEN
745 0 : unit_nr_dbcsr = unit_nr
746 : ELSE
747 26 : unit_nr_dbcsr = 0
748 : END IF
749 :
750 26 : use_virial = virial%pv_availability .AND. (.NOT. virial%pv_numer)
751 :
752 26 : IF (use_virial) virial%pv_calculate = .TRUE.
753 :
754 26 : IF (use_virial) THEN
755 2 : qs_section => section_vals_get_subs_vals(qs_env%input, "DFT%QS")
756 2 : CALL section_vals_val_get(qs_section, "EPS_PGF_ORB", n_rep_val=n_rep)
757 2 : IF (n_rep /= 0) THEN
758 0 : CALL section_vals_val_get(qs_section, "EPS_PGF_ORB", r_val=eps_pgf_orb)
759 : ELSE
760 2 : CALL section_vals_val_get(qs_section, "EPS_DEFAULT", r_val=eps_pgf_orb)
761 2 : eps_pgf_orb = SQRT(eps_pgf_orb)
762 : END IF
763 2 : eps_pgf_orb_old = dft_control%qs_control%eps_pgf_orb
764 :
765 16 : ALLOCATE (basis_set_ri_aux(nkind), basis_set_ao(nkind))
766 2 : CALL basis_set_list_setup(basis_set_ri_aux, "RI_AUX", qs_kind_set)
767 2 : CALL basis_set_list_setup(basis_set_ao, "ORB", qs_kind_set)
768 :
769 8 : DO ibasis = 1, SIZE(basis_set_ao)
770 4 : orb_basis => basis_set_ao(ibasis)%gto_basis_set
771 4 : CALL init_interaction_radii_orb_basis(orb_basis, eps_pgf_orb)
772 4 : ri_basis => basis_set_ri_aux(ibasis)%gto_basis_set
773 6 : CALL init_interaction_radii_orb_basis(ri_basis, eps_pgf_orb)
774 : END DO
775 : END IF
776 :
777 : !We follow the general logic of the compute_mat_P_omega routine
778 268 : ALLOCATE (t_P(nspins))
779 26 : CALL dbt_create(force_data%t_2c_K, t_2c_RI)
780 26 : CALL dbt_create(force_data%t_2c_K, t_2c_RI_2)
781 26 : CALL dbt_create(force_data%t_2c_der_ovlp(1), t_2c_AO)
782 :
783 26 : CALL get_atomic_kind_set(atomic_kind_set, kind_of=kind_of, atom_of_kind=atom_of_kind)
784 :
785 : ! Always do the batching of the MO on mu and sigma, such that it is consistent between
786 : ! the occupied and the virtual quantities
787 78 : ALLOCATE (mc_ranges(cut_memory + 1))
788 78 : mc_ranges(:cut_memory) = starts_array_mc_block(:)
789 26 : mc_ranges(cut_memory + 1) = ends_array_mc_block(cut_memory) + 1
790 :
791 : ! Also need some batching on the RI, because it loses sparsity at some point
792 26 : n_mem_RI = cut_memory
793 : CALL create_tensor_batches(force_data%bsizes_RI_split, n_mem_RI, batch_start_RI, batch_end_RI, &
794 26 : batch_blk_start, batch_blk_end)
795 78 : ALLOCATE (mc_ranges_RI(n_mem_RI + 1))
796 78 : mc_ranges_RI(1:n_mem_RI) = batch_blk_start(1:n_mem_RI)
797 26 : mc_ranges_RI(n_mem_RI + 1) = batch_blk_end(n_mem_RI) + 1
798 26 : DEALLOCATE (batch_blk_start, batch_blk_end)
799 :
800 : !Pre-allocate all required tensors and matrices
801 60 : DO ispin = 1, nspins
802 60 : CALL dbt_create(t_2c_RI, t_P(ispin))
803 : END DO
804 26 : CALL dbt_create(t_2c_RI, t_Q)
805 26 : CALL dbt_create(t_2c_RI, t_KQKT)
806 26 : CALL dbt_create(t_2c_AO, t_dm_occ)
807 26 : CALL dbt_create(t_2c_AO, t_dm_virt)
808 :
809 : !note: t_3c_O and t_3c_M have different mappings (map_1d, map_2d)
810 26 : CALL dbt_create(t_3c_O, t_M_occ)
811 26 : CALL dbt_create(t_3c_O, t_M_virt)
812 26 : CALL dbt_create(t_3c_O, t_3c_0)
813 :
814 26 : CALL dbt_create(t_3c_O, t_3c_1)
815 26 : CALL dbt_create(t_3c_O, t_3c_3)
816 26 : CALL dbt_create(t_3c_O, t_3c_4)
817 26 : CALL dbt_create(t_3c_O, t_3c_5)
818 26 : CALL dbt_create(t_3c_M, t_3c_6)
819 26 : CALL dbt_create(t_3c_M, t_3c_7)
820 26 : CALL dbt_create(t_3c_M, t_3c_8)
821 26 : CALL dbt_create(t_3c_M, t_3c_sparse)
822 26 : CALL dbt_create(t_3c_O, t_3c_help_1)
823 26 : CALL dbt_create(t_3c_O, t_3c_help_2)
824 26 : CALL dbt_create(t_2c_AO, t_R_occ)
825 26 : CALL dbt_create(t_2c_AO, t_R_virt)
826 26 : CALL dbt_create(t_3c_M, t_3c_ints)
827 26 : CALL dbt_create(t_3c_M, t_3c_work)
828 :
829 : !Pre-define the sparsity of t_3c_4 as a function of the derivatives
830 26 : occ_der_AO = 0; nze_der_AO = 0
831 26 : occ_der_RI = 0; nze_der_RI = 0
832 104 : DO i_xyz = 1, 3
833 260 : DO i_mem = 1, cut_memory
834 : CALL decompress_tensor(force_data%t_3c_der_RI(i_xyz), force_data%t_3c_der_RI_ind(i_mem, i_xyz)%ind, &
835 156 : force_data%t_3c_der_RI_comp(i_mem, i_xyz), mp2_env%ri_rpa_im_time%eps_compress)
836 156 : CALL get_tensor_occupancy(force_data%t_3c_der_RI(i_xyz), nze, occ)
837 156 : occ_der_RI = occ_der_RI + occ
838 156 : nze_der_RI = nze_der_RI + nze
839 156 : CALL dbt_copy(force_data%t_3c_der_RI(i_xyz), t_3c_sparse, summation=.TRUE., move_data=.TRUE.)
840 :
841 : CALL decompress_tensor(force_data%t_3c_der_AO(i_xyz), force_data%t_3c_der_AO_ind(i_mem, i_xyz)%ind, &
842 156 : force_data%t_3c_der_AO_comp(i_mem, i_xyz), mp2_env%ri_rpa_im_time%eps_compress)
843 156 : CALL get_tensor_occupancy(force_data%t_3c_der_AO(i_xyz), nze, occ)
844 156 : occ_der_AO = occ_der_AO + occ
845 156 : nze_der_AO = nze_der_AO + nze
846 156 : CALL dbt_copy(force_data%t_3c_der_AO(i_xyz), t_3c_sparse, order=[1, 3, 2], summation=.TRUE.)
847 546 : CALL dbt_copy(force_data%t_3c_der_AO(i_xyz), t_3c_sparse, summation=.TRUE., move_data=.TRUE.)
848 : END DO
849 : END DO
850 26 : occ_der_RI = occ_der_RI/3.0_dp
851 26 : occ_der_AO = occ_der_AO/3.0_dp
852 26 : nze_der_RI = nze_der_RI/3
853 26 : nze_der_AO = nze_der_AO/3
854 :
855 26 : CALL dbcsr_create(R_occ, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
856 26 : CALL dbcsr_create(R_virt, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
857 26 : CALL dbcsr_create(dbcsr_work1, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
858 26 : CALL dbcsr_create(dbcsr_work2, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
859 26 : CALL dbcsr_create(dbcsr_work3, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
860 26 : CALL dbcsr_create(exp_occ, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
861 26 : CALL dbcsr_create(exp_virt, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
862 26 : IF (use_virial) CALL dbcsr_create(virial_ovlp, template=dbcsr_work1)
863 :
864 26 : CALL dbt_batched_contract_init(t_3c_0, batch_range_2=mc_ranges, batch_range_3=mc_ranges)
865 26 : CALL dbt_batched_contract_init(t_3c_1, batch_range_2=mc_ranges, batch_range_3=mc_ranges)
866 26 : CALL dbt_batched_contract_init(t_3c_3, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
867 26 : CALL dbt_batched_contract_init(t_M_occ, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
868 26 : CALL dbt_batched_contract_init(t_M_virt, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
869 :
870 26 : CALL dbt_batched_contract_init(t_3c_ints, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
871 26 : CALL dbt_batched_contract_init(t_3c_work, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
872 :
873 : CALL dbt_batched_contract_init(t_3c_4, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
874 26 : batch_range_3=mc_ranges)
875 : CALL dbt_batched_contract_init(t_3c_5, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
876 26 : batch_range_3=mc_ranges)
877 : CALL dbt_batched_contract_init(t_3c_6, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
878 26 : batch_range_3=mc_ranges)
879 : CALL dbt_batched_contract_init(t_3c_7, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
880 26 : batch_range_3=mc_ranges)
881 : CALL dbt_batched_contract_init(t_3c_8, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
882 26 : batch_range_3=mc_ranges)
883 : CALL dbt_batched_contract_init(t_3c_sparse, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
884 26 : batch_range_3=mc_ranges)
885 :
886 26 : work_virial = 0.0_dp
887 26 : work_virial_ovlp = 0.0_dp
888 104 : DO jquad = 1, num_integ_points
889 78 : tau = grid%imaginary_time(jquad)
890 78 : omega = grid%time_weights_at_zero_frequency(jquad)
891 78 : fac = -2.0_dp*omega*mp2_env%scale_S
892 78 : IF (open_shell) fac = 0.5_dp*fac
893 78 : occ_ddint = 0; nze_ddint = 0
894 :
895 78 : CALL para_env%sync()
896 78 : t1 = m_walltime()
897 :
898 : !Deal with the force contributions where there is no explicit 3-center quantities, i.e. the
899 : !forces due to the metric and potential derivatives
900 180 : DO ispin = 1, nspins
901 102 : CALL dbt_create(mat_P_omega(jquad, ispin)%matrix, t_2c_tmp)
902 102 : CALL dbt_copy_matrix_to_tensor(mat_P_omega(jquad, ispin)%matrix, t_2c_tmp)
903 102 : CALL dbt_copy(t_2c_tmp, t_P(ispin), move_data=.TRUE.)
904 102 : CALL dbt_filter(t_P(ispin), eps_filter)
905 180 : CALL dbt_destroy(t_2c_tmp)
906 : END DO
907 :
908 : !Q = K^T*P*K, open-shell: Q is from one spin, everything else from the other
909 : CALL dbt_contract(1.0_dp, t_P(Qspin), force_data%t_2c_K, 0.0_dp, t_2c_RI, &
910 : contract_1=[2], notcontract_1=[1], &
911 : contract_2=[1], notcontract_2=[2], &
912 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
913 78 : flop=flop, unit_nr=unit_nr_dbcsr)
914 78 : dbcsr_nflop = dbcsr_nflop + flop
915 : CALL dbt_contract(1.0_dp, force_data%t_2c_K, t_2c_RI, 0.0_dp, t_Q, &
916 : contract_1=[1], notcontract_1=[2], &
917 : contract_2=[1], notcontract_2=[2], &
918 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
919 78 : flop=flop, unit_nr=unit_nr_dbcsr)
920 78 : dbcsr_nflop = dbcsr_nflop + flop
921 78 : CALL dbt_clear(t_2c_RI)
922 :
923 : CALL perform_2c_ops(force, t_KQKT, force_data, fac, t_Q, t_P(Pspin), t_2c_RI, t_2c_RI_2, &
924 78 : use_virial, atom_of_kind, kind_of, eps_filter, dbcsr_nflop, unit_nr_dbcsr)
925 78 : CALL get_tensor_occupancy(t_KQKT, nze_KQK, occ_KQK)
926 :
927 : !Calculate the pseudo-density matrix in tensor form. There are a few useless arguments for SOS-MP2
928 : CALL compute_mat_dm_global(grid, nmo, fm_mo_coeff(Pspin), homo(Pspin), propagator, &
929 : matrix_s, Pspin, Eigenval(:, Pspin), 0.0_dp, eps_filter, &
930 : mp2_env%ri_rpa_im_time%memory_info, unit_nr, &
931 78 : jquad, .FALSE., .FALSE., qs_env, dummy_int, dummy_ptr, para_env)
932 :
933 78 : CALL dbt_create(propagator(propagator_sector_occupied, jquad, 1)%matrix, t_2c_tmp)
934 78 : CALL dbt_copy_matrix_to_tensor(propagator(propagator_sector_occupied, jquad, 1)%matrix, t_2c_tmp)
935 78 : CALL dbt_copy(t_2c_tmp, t_dm_occ, move_data=.TRUE.)
936 78 : CALL dbt_filter(t_dm_occ, eps_filter)
937 78 : CALL dbt_destroy(t_2c_tmp)
938 :
939 78 : CALL dbt_create(propagator(propagator_sector_virtual, jquad, 1)%matrix, t_2c_tmp)
940 78 : CALL dbt_copy_matrix_to_tensor(propagator(propagator_sector_virtual, jquad, 1)%matrix, t_2c_tmp)
941 78 : CALL dbt_copy(t_2c_tmp, t_dm_virt, move_data=.TRUE.)
942 78 : CALL dbt_filter(t_dm_virt, eps_filter)
943 78 : CALL dbt_destroy(t_2c_tmp)
944 :
945 : !Deal with the 3-center quantities.
946 : CALL perform_3c_ops(force, t_R_occ, t_R_virt, force_data, fac, cut_memory, n_mem_RI, &
947 : t_KQKT, t_dm_occ, t_dm_virt, t_3c_O, t_3c_M, t_M_occ, t_M_virt, t_3c_0, t_3c_1, &
948 : t_3c_3, t_3c_4, t_3c_5, t_3c_6, t_3c_7, t_3c_8, t_3c_sparse, t_3c_help_1, t_3c_help_2, &
949 : t_3c_ints, t_3c_work, starts_array_mc, ends_array_mc, batch_start_RI, &
950 : batch_end_RI, t_3c_O_compressed, t_3c_O_ind, use_virial, &
951 : atom_of_kind, kind_of, eps_filter, occ_ddint, nze_ddint, dbcsr_nflop, &
952 78 : unit_nr_dbcsr, mp2_env)
953 :
954 78 : CALL timeset(routineN//"_dbcsr", handle2)
955 : !We go back to DBCSR matrices from now on
956 : !Note: R matrices are in fact symmetric, but use a normal type for convenience
957 78 : CALL dbt_create(matrix_s(1)%matrix, t_2c_tmp)
958 78 : CALL dbt_copy(t_R_occ, t_2c_tmp, move_data=.TRUE.)
959 78 : CALL dbt_copy_tensor_to_matrix(t_2c_tmp, R_occ)
960 :
961 78 : CALL dbt_copy(t_R_virt, t_2c_tmp, move_data=.TRUE.)
962 78 : CALL dbt_copy_tensor_to_matrix(t_2c_tmp, R_virt)
963 :
964 : !Iteratively calculate the Y1 and Y2 matrices
965 : CALL dbcsr_multiply('N', 'N', tau, force_data%P_occ(Pspin)%matrix, &
966 78 : matrix_ks(Pspin)%matrix, 0.0_dp, dbcsr_work1)
967 78 : CALL build_Y_matrix(Y_1, dbcsr_work1, force_data%P_occ(Pspin)%matrix, R_virt, eps_filter)
968 78 : CALL matrix_exponential(exp_occ, dbcsr_work1, 1.0_dp, 1.0_dp, eps_filter)
969 :
970 : CALL dbcsr_multiply('N', 'N', -tau, force_data%P_virt(Pspin)%matrix, &
971 78 : matrix_ks(Pspin)%matrix, 0.0_dp, dbcsr_work1)
972 78 : CALL build_Y_matrix(Y_2, dbcsr_work1, force_data%P_virt(Pspin)%matrix, R_occ, eps_filter)
973 78 : CALL matrix_exponential(exp_virt, dbcsr_work1, 1.0_dp, 1.0_dp, eps_filter)
974 :
975 : !The force contribution coming from [-S^-1*(e^-tau*P_virt*F)^T*R_occ*S^-1
976 : ! +tau*S^-1*Y_2^T*F*S^-1] * der_S
977 78 : CALL dbcsr_multiply('N', 'N', 1.0_dp, R_occ, force_data%inv_ovlp, 0.0_dp, dbcsr_work1)
978 78 : CALL dbcsr_multiply('T', 'N', 1.0_dp, exp_virt, dbcsr_work1, 0.0_dp, dbcsr_work3)
979 78 : CALL dbcsr_multiply('N', 'N', 1.0_dp, force_data%inv_ovlp, dbcsr_work3, 0.0_dp, dbcsr_work2)
980 :
981 78 : CALL dbcsr_multiply('N', 'T', tau, force_data%inv_ovlp, Y_2, 0.0_dp, dbcsr_work3)
982 78 : CALL dbcsr_multiply('N', 'N', 1.0_dp, dbcsr_work3, matrix_ks(Pspin)%matrix, 0.0_dp, dbcsr_work1)
983 78 : CALL dbcsr_multiply('N', 'N', 1.0_dp, dbcsr_work1, force_data%inv_ovlp, 0.0_dp, dbcsr_work3)
984 :
985 78 : CALL dbcsr_add(dbcsr_work2, dbcsr_work3, 1.0_dp, -1.0_dp)
986 :
987 78 : CALL dbt_copy_matrix_to_tensor(dbcsr_work2, t_2c_tmp)
988 78 : CALL dbt_copy(t_2c_tmp, t_2c_AO, move_data=.TRUE.)
989 :
990 78 : pref = -1.0_dp*fac
991 : CALL get_2c_der_force(force, t_2c_AO, force_data%t_2c_der_ovlp, atom_of_kind, &
992 78 : kind_of, force_data%idx_to_at_AO, pref, do_ovlp=.TRUE.)
993 :
994 78 : IF (use_virial) CALL dbcsr_add(virial_ovlp, dbcsr_work2, 1.0_dp, pref)
995 :
996 : !The final contribution from Tr[(tau*Y_1*P_occ - tau*Y_2*P_virt) * der_F]
997 : CALL dbcsr_multiply('N', 'N', tau*fac, Y_1, force_data%P_occ(Pspin)%matrix, 1.0_dp, &
998 78 : force_data%sum_YP_tau(Pspin)%matrix, retain_sparsity=.TRUE.)
999 : CALL dbcsr_multiply('N', 'N', -tau*fac, Y_2, force_data%P_virt(Pspin)%matrix, 1.0_dp, &
1000 78 : force_data%sum_YP_tau(Pspin)%matrix, retain_sparsity=.TRUE.)
1001 :
1002 : !Build-up the RHS of the response equation.
1003 78 : pref = -omega*mp2_env%scale_S
1004 : CALL dbcsr_multiply('N', 'N', pref, R_virt, exp_occ, 1.0_dp, &
1005 78 : force_data%sum_O_tau(Pspin)%matrix, retain_sparsity=.TRUE.)
1006 : CALL dbcsr_multiply('N', 'N', -pref, R_occ, exp_virt, 1.0_dp, &
1007 78 : force_data%sum_O_tau(Pspin)%matrix, retain_sparsity=.TRUE.)
1008 : CALL dbcsr_multiply('N', 'N', pref*tau, matrix_ks(Pspin)%matrix, Y_1, 1.0_dp, &
1009 78 : force_data%sum_O_tau(Pspin)%matrix, retain_sparsity=.TRUE.)
1010 : CALL dbcsr_multiply('N', 'N', pref*tau, matrix_ks(Pspin)%matrix, Y_2, 1.0_dp, &
1011 78 : force_data%sum_O_tau(Pspin)%matrix, retain_sparsity=.TRUE.)
1012 :
1013 78 : CALL timestop(handle2)
1014 :
1015 : !Print some info
1016 78 : CALL para_env%sync()
1017 78 : t2 = m_walltime()
1018 78 : dbcsr_time = dbcsr_time + t2 - t1
1019 :
1020 78 : IF (unit_nr > 0) THEN
1021 : WRITE (unit_nr, '(/T3,A,1X,I3,A)') &
1022 39 : 'RPA_LOW_SCALING_INFO| Info for time point', jquad, ' (gradients)'
1023 : WRITE (unit_nr, '(T6,A,T56,F25.6)') &
1024 39 : 'Execution time (s):', t2 - t1
1025 : WRITE (unit_nr, '(T6,A,T63,ES7.1,1X,A1,1X,F7.3,A1)') &
1026 39 : 'Occupancy of 3c AO derivs:', REAL(nze_der_AO, dp), '/', occ_der_AO*100, '%'
1027 : WRITE (unit_nr, '(T6,A,T63,ES7.1,1X,A1,1X,F7.3,A1)') &
1028 39 : 'Occupancy of 3c RI derivs:', REAL(nze_der_RI, dp), '/', occ_der_RI*100, '%'
1029 : WRITE (unit_nr, '(T6,A,T63,ES7.1,1X,A1,1X,F7.3,A1)') &
1030 39 : 'Occupancy of the Docc * Dvirt * 3c-int tensor', REAL(nze_ddint, dp), '/', occ_ddint*100, '%'
1031 : WRITE (unit_nr, '(T6,A,T63,ES7.1,1X,A1,1X,F7.3,A1)') &
1032 39 : 'Occupancy of KQK^T 2c-tensor:', REAL(nze_KQK, dp), '/', occ_KQK*100, '%'
1033 39 : CALL m_flush(unit_nr)
1034 : END IF
1035 :
1036 : !intermediate clean-up
1037 78 : CALL dbcsr_release(Y_1)
1038 78 : CALL dbcsr_release(Y_2)
1039 416 : CALL dbt_destroy(t_2c_tmp)
1040 : END DO !jquad
1041 :
1042 26 : CALL dbt_batched_contract_finalize(t_3c_0)
1043 26 : CALL dbt_batched_contract_finalize(t_3c_1)
1044 26 : CALL dbt_batched_contract_finalize(t_3c_3)
1045 26 : CALL dbt_batched_contract_finalize(t_M_occ)
1046 26 : CALL dbt_batched_contract_finalize(t_M_virt)
1047 :
1048 26 : CALL dbt_batched_contract_finalize(t_3c_ints)
1049 26 : CALL dbt_batched_contract_finalize(t_3c_work)
1050 :
1051 26 : CALL dbt_batched_contract_finalize(t_3c_4)
1052 26 : CALL dbt_batched_contract_finalize(t_3c_5)
1053 26 : CALL dbt_batched_contract_finalize(t_3c_6)
1054 26 : CALL dbt_batched_contract_finalize(t_3c_7)
1055 26 : CALL dbt_batched_contract_finalize(t_3c_8)
1056 26 : CALL dbt_batched_contract_finalize(t_3c_sparse)
1057 :
1058 : !Calculate the 2c and 3c contributions to the virial
1059 26 : IF (use_virial) THEN
1060 2 : CALL dbt_copy(force_data%t_3c_virial_split, force_data%t_3c_virial, move_data=.TRUE.)
1061 : CALL calc_3c_virial(work_virial, force_data%t_3c_virial, 1.0_dp, qs_env, force_data%nl_3c, &
1062 : basis_set_ri_aux, basis_set_ao, basis_set_ao, mp2_env%ri_metric, &
1063 2 : der_eps=mp2_env%ri_rpa_im_time%eps_filter, op_pos=1)
1064 :
1065 : CALL calc_2c_virial(work_virial, force_data%RI_virial_met, 1.0_dp, qs_env, force_data%nl_2c_met, &
1066 2 : basis_set_ri_aux, basis_set_ri_aux, mp2_env%ri_metric)
1067 2 : CALL dbcsr_clear(force_data%RI_virial_met)
1068 :
1069 2 : IF (.NOT. force_data%do_periodic) THEN
1070 : CALL calc_2c_virial(work_virial, force_data%RI_virial_pot, 1.0_dp, qs_env, force_data%nl_2c_pot, &
1071 0 : basis_set_ri_aux, basis_set_ri_aux, mp2_env%potential_parameter)
1072 0 : CALL dbcsr_clear(force_data%RI_virial_pot)
1073 : END IF
1074 :
1075 2 : identity_pot%potential_type = do_potential_id
1076 : CALL calc_2c_virial(work_virial_ovlp, virial_ovlp, 1.0_dp, qs_env, force_data%nl_2c_ovlp, &
1077 2 : basis_set_ao, basis_set_ao, identity_pot)
1078 2 : CALL dbcsr_release(virial_ovlp)
1079 :
1080 8 : DO k_xyz = 1, 3
1081 26 : DO j_xyz = 1, 3
1082 78 : DO i_xyz = 1, 3
1083 : virial%pv_mp2(i_xyz, j_xyz) = virial%pv_mp2(i_xyz, j_xyz) &
1084 54 : - work_virial(i_xyz, k_xyz)*cell%hmat(j_xyz, k_xyz)
1085 : virial%pv_overlap(i_xyz, j_xyz) = virial%pv_overlap(i_xyz, j_xyz) &
1086 54 : - work_virial_ovlp(i_xyz, k_xyz)*cell%hmat(j_xyz, k_xyz)
1087 : virial%pv_virial(i_xyz, j_xyz) = virial%pv_virial(i_xyz, j_xyz) &
1088 : - work_virial(i_xyz, k_xyz)*cell%hmat(j_xyz, k_xyz) &
1089 72 : - work_virial_ovlp(i_xyz, k_xyz)*cell%hmat(j_xyz, k_xyz)
1090 : END DO
1091 : END DO
1092 : END DO
1093 : END IF
1094 :
1095 : !Calculate the periodic contributions of (P|Q) to the force and the virial
1096 26 : work_virial = 0.0_dp
1097 26 : IF (force_data%do_periodic) THEN
1098 10 : IF (mp2_env%eri_method == do_eri_gpw) THEN
1099 6 : CALL get_2c_gpw_forces(force_data%G_PQ, force, work_virial, use_virial, mp2_env, qs_env)
1100 4 : ELSE IF (mp2_env%eri_method == do_eri_mme) THEN
1101 4 : CALL get_2c_mme_forces(force_data%G_PQ, force, mp2_env, qs_env)
1102 4 : IF (use_virial) CPABORT("Stress tensor not available with MME intrgrals")
1103 : ELSE
1104 0 : CPABORT("Periodic case not possible with OS integrals")
1105 : END IF
1106 10 : CALL dbcsr_clear(force_data%G_PQ)
1107 : END IF
1108 :
1109 26 : IF (use_virial) THEN
1110 26 : virial%pv_mp2 = virial%pv_mp2 + work_virial
1111 26 : virial%pv_virial = virial%pv_virial + work_virial
1112 2 : virial%pv_calculate = .FALSE.
1113 :
1114 6 : DO ibasis = 1, SIZE(basis_set_ao)
1115 4 : orb_basis => basis_set_ao(ibasis)%gto_basis_set
1116 4 : CALL init_interaction_radii_orb_basis(orb_basis, eps_pgf_orb_old)
1117 4 : ri_basis => basis_set_ri_aux(ibasis)%gto_basis_set
1118 6 : CALL init_interaction_radii_orb_basis(ri_basis, eps_pgf_orb_old)
1119 : END DO
1120 : END IF
1121 :
1122 : !clean-up
1123 26 : IF (ASSOCIATED(dummy_ptr)) DEALLOCATE (dummy_ptr)
1124 60 : DO ispin = 1, nspins
1125 60 : CALL dbt_destroy(t_P(ispin))
1126 : END DO
1127 26 : CALL dbt_destroy(t_3c_0)
1128 26 : CALL dbt_destroy(t_3c_1)
1129 26 : CALL dbt_destroy(t_3c_3)
1130 26 : CALL dbt_destroy(t_3c_4)
1131 26 : CALL dbt_destroy(t_3c_5)
1132 26 : CALL dbt_destroy(t_3c_6)
1133 26 : CALL dbt_destroy(t_3c_7)
1134 26 : CALL dbt_destroy(t_3c_8)
1135 26 : CALL dbt_destroy(t_3c_sparse)
1136 26 : CALL dbt_destroy(t_3c_help_1)
1137 26 : CALL dbt_destroy(t_3c_help_2)
1138 26 : CALL dbt_destroy(t_3c_ints)
1139 26 : CALL dbt_destroy(t_3c_work)
1140 26 : CALL dbt_destroy(t_R_occ)
1141 26 : CALL dbt_destroy(t_R_virt)
1142 26 : CALL dbt_destroy(t_dm_occ)
1143 26 : CALL dbt_destroy(t_dm_virt)
1144 26 : CALL dbt_destroy(t_Q)
1145 26 : CALL dbt_destroy(t_KQKT)
1146 26 : CALL dbt_destroy(t_M_occ)
1147 26 : CALL dbt_destroy(t_M_virt)
1148 26 : CALL dbcsr_release(R_occ)
1149 26 : CALL dbcsr_release(R_virt)
1150 26 : CALL dbcsr_release(dbcsr_work1)
1151 26 : CALL dbcsr_release(dbcsr_work2)
1152 26 : CALL dbcsr_release(dbcsr_work3)
1153 26 : CALL dbcsr_release(exp_occ)
1154 26 : CALL dbcsr_release(exp_virt)
1155 :
1156 26 : CALL dbt_destroy(t_2c_RI)
1157 26 : CALL dbt_destroy(t_2c_RI_2)
1158 26 : CALL dbt_destroy(t_2c_AO)
1159 26 : CALL dbcsr_deallocate_matrix_set(propagator)
1160 :
1161 26 : CALL timestop(handle)
1162 :
1163 112 : END SUBROUTINE calc_laplace_loop_forces
1164 :
1165 : ! **************************************************************************************************
1166 : !> \brief Updates the cubic-scaling RPA contribution to the forces at each quadrature point. This
1167 : !> routine is adapted from the corresponding Laplace SOS-MP2 loop force one.
1168 : !> \param force_data ...
1169 : !> \param mat_P_omega ...
1170 : !> \param t_3c_M ...
1171 : !> \param t_3c_O ...
1172 : !> \param t_3c_O_compressed ...
1173 : !> \param t_3c_O_ind ...
1174 : !> \param fm_mo_coeff real Gamma-point MO coefficients ...
1175 : !> \param homo ...
1176 : !> \param starts_array_mc ...
1177 : !> \param ends_array_mc ...
1178 : !> \param starts_array_mc_block ...
1179 : !> \param ends_array_mc_block ...
1180 : !> \param nmo ...
1181 : !> \param Eigenval ...
1182 : !> \param e_fermi ...
1183 : !> \param grid ...
1184 : !> \param cut_memory ...
1185 : !> \param ispin ...
1186 : !> \param open_shell ...
1187 : !> \param unit_nr ...
1188 : !> \param dbcsr_time ...
1189 : !> \param dbcsr_nflop ...
1190 : !> \param mp2_env ...
1191 : !> \param qs_env ...
1192 : ! **************************************************************************************************
1193 180 : SUBROUTINE calc_rpa_loop_forces(force_data, mat_P_omega, t_3c_M, t_3c_O, t_3c_O_compressed, &
1194 72 : t_3c_O_ind, fm_mo_coeff, homo, starts_array_mc, ends_array_mc, &
1195 36 : starts_array_mc_block, ends_array_mc_block, &
1196 36 : nmo, Eigenval, e_fermi, grid, cut_memory, ispin, open_shell, unit_nr, dbcsr_time, &
1197 : dbcsr_nflop, mp2_env, qs_env)
1198 :
1199 : TYPE(im_time_force_type), INTENT(INOUT) :: force_data
1200 : TYPE(dbcsr_p_type), DIMENSION(:, :), INTENT(INOUT) :: mat_P_omega
1201 : TYPE(dbt_type), INTENT(INOUT) :: t_3c_M, t_3c_O
1202 : TYPE(hfx_compression_type), DIMENSION(:) :: t_3c_O_compressed
1203 : TYPE(block_ind_type), DIMENSION(:), INTENT(INOUT) :: t_3c_O_ind
1204 : TYPE(cp_fm_type), DIMENSION(:), INTENT(IN) :: fm_mo_coeff
1205 : INTEGER, DIMENSION(:), INTENT(IN) :: homo, starts_array_mc, ends_array_mc, &
1206 : starts_array_mc_block, &
1207 : ends_array_mc_block
1208 : INTEGER, INTENT(IN) :: nmo
1209 : REAL(KIND=dp), DIMENSION(:, :), INTENT(IN) :: Eigenval
1210 : REAL(KIND=dp), INTENT(IN) :: e_fermi
1211 : TYPE(time_frequency_grid_type), INTENT(IN) :: grid
1212 : INTEGER, INTENT(IN) :: cut_memory, ispin
1213 : LOGICAL, INTENT(IN) :: open_shell
1214 : INTEGER, INTENT(IN) :: unit_nr
1215 : REAL(dp), INTENT(INOUT) :: dbcsr_time
1216 : INTEGER(int_8), INTENT(INOUT) :: dbcsr_nflop
1217 : TYPE(mp2_type) :: mp2_env
1218 : TYPE(qs_environment_type), POINTER :: qs_env
1219 :
1220 : CHARACTER(LEN=*), PARAMETER :: routineN = 'calc_rpa_loop_forces'
1221 :
1222 : INTEGER :: dummy_int, handle, handle2, i_mem, i_xyz, ibasis, iquad, j_xyz, jquad, k_xyz, &
1223 : n_mem_RI, n_rep, natom, nkind, nspins, num_integ_points, unit_nr_dbcsr
1224 : INTEGER(int_8) :: flop, nze, nze_ddint, nze_der_AO, &
1225 : nze_der_RI, nze_KBK
1226 36 : INTEGER, ALLOCATABLE, DIMENSION(:) :: atom_of_kind, batch_blk_end, &
1227 36 : batch_blk_start, batch_end_RI, &
1228 36 : batch_start_RI, kind_of, mc_ranges, &
1229 36 : mc_ranges_RI
1230 36 : INTEGER, DIMENSION(:, :), POINTER :: dummy_ptr
1231 : LOGICAL :: memory_info, use_virial
1232 : REAL(dp) :: eps_filter, eps_pgf_orb, eps_pgf_orb_old, fac, occ, occ_ddint, occ_der_AO, &
1233 : occ_der_RI, occ_KBK, pref, spin_fac, t1, t2, tau, weight
1234 : REAL(dp), DIMENSION(3, 3) :: work_virial, work_virial_ovlp
1235 36 : TYPE(atomic_kind_type), DIMENSION(:), POINTER :: atomic_kind_set
1236 : TYPE(cell_type), POINTER :: cell
1237 : TYPE(cp_blacs_env_type), POINTER :: blacs_env
1238 36 : TYPE(dbcsr_p_type), DIMENSION(:), POINTER :: mat_P_tau, matrix_ks, matrix_s
1239 36 : TYPE(dbcsr_p_type), DIMENSION(:, :, :), POINTER :: propagator
1240 : TYPE(dbcsr_type) :: dbcsr_work1, dbcsr_work2, dbcsr_work3, &
1241 : dbcsr_work_symm, exp_occ, exp_virt, &
1242 : R_occ, R_virt, virial_ovlp, Y_1, Y_2
1243 1764 : TYPE(dbt_type) :: t_2c_AO, t_2c_RI, t_2c_RI_2, t_2c_tmp, t_3c_0, t_3c_1, t_3c_3, t_3c_4, &
1244 1764 : t_3c_5, t_3c_6, t_3c_7, t_3c_8, t_3c_help_1, t_3c_help_2, t_3c_ints, t_3c_sparse, &
1245 2016 : t_3c_work, t_dm_occ, t_dm_virt, t_KBKT, t_M_occ, t_M_virt, t_P, t_R_occ, t_R_virt
1246 36 : TYPE(dbt_type), ALLOCATABLE, DIMENSION(:) :: t_B
1247 : TYPE(dft_control_type), POINTER :: dft_control
1248 : TYPE(gto_basis_set_p_type), ALLOCATABLE, &
1249 36 : DIMENSION(:), TARGET :: basis_set_ao, basis_set_ri_aux
1250 : TYPE(gto_basis_set_type), POINTER :: orb_basis, ri_basis
1251 : TYPE(libint_potential_type) :: identity_pot
1252 : TYPE(mp_para_env_type), POINTER :: para_env
1253 36 : TYPE(particle_type), DIMENSION(:), POINTER :: particle_set
1254 36 : TYPE(qs_force_type), DIMENSION(:), POINTER :: force
1255 36 : TYPE(qs_kind_type), DIMENSION(:), POINTER :: qs_kind_set
1256 : TYPE(section_vals_type), POINTER :: qs_section
1257 : TYPE(virial_type), POINTER :: virial
1258 :
1259 36 : NULLIFY (matrix_s, dummy_ptr, atomic_kind_set, force, matrix_s, matrix_ks)
1260 36 : NULLIFY (dft_control, virial, particle_set, cell, blacs_env, para_env, orb_basis, ri_basis)
1261 36 : NULLIFY (qs_kind_set)
1262 :
1263 36 : CALL timeset(routineN, handle)
1264 :
1265 36 : num_integ_points = SIZE(grid%imaginary_time)
1266 :
1267 36 : NULLIFY (propagator)
1268 :
1269 : CALL get_qs_env(qs_env, matrix_s=matrix_s, natom=natom, atomic_kind_set=atomic_kind_set, &
1270 : force=force, matrix_ks=matrix_ks, dft_control=dft_control, virial=virial, &
1271 : particle_set=particle_set, cell=cell, blacs_env=blacs_env, para_env=para_env, &
1272 36 : qs_kind_set=qs_kind_set, nkind=nkind)
1273 36 : eps_filter = mp2_env%ri_rpa_im_time%eps_filter
1274 36 : nspins = dft_control%nspins
1275 :
1276 36 : memory_info = mp2_env%ri_rpa_im_time%memory_info
1277 36 : IF (memory_info) THEN
1278 0 : unit_nr_dbcsr = unit_nr
1279 : ELSE
1280 36 : unit_nr_dbcsr = 0
1281 : END IF
1282 :
1283 36 : use_virial = virial%pv_availability .AND. (.NOT. virial%pv_numer)
1284 :
1285 36 : IF (use_virial) virial%pv_calculate = .TRUE.
1286 :
1287 36 : IF (use_virial) THEN
1288 2 : qs_section => section_vals_get_subs_vals(qs_env%input, "DFT%QS")
1289 2 : CALL section_vals_val_get(qs_section, "EPS_PGF_ORB", n_rep_val=n_rep)
1290 2 : IF (n_rep /= 0) THEN
1291 0 : CALL section_vals_val_get(qs_section, "EPS_PGF_ORB", r_val=eps_pgf_orb)
1292 : ELSE
1293 2 : CALL section_vals_val_get(qs_section, "EPS_DEFAULT", r_val=eps_pgf_orb)
1294 2 : eps_pgf_orb = SQRT(eps_pgf_orb)
1295 : END IF
1296 2 : eps_pgf_orb_old = dft_control%qs_control%eps_pgf_orb
1297 :
1298 16 : ALLOCATE (basis_set_ri_aux(nkind), basis_set_ao(nkind))
1299 2 : CALL basis_set_list_setup(basis_set_ri_aux, "RI_AUX", qs_kind_set)
1300 2 : CALL basis_set_list_setup(basis_set_ao, "ORB", qs_kind_set)
1301 :
1302 8 : DO ibasis = 1, SIZE(basis_set_ao)
1303 4 : orb_basis => basis_set_ao(ibasis)%gto_basis_set
1304 4 : CALL init_interaction_radii_orb_basis(orb_basis, eps_pgf_orb)
1305 4 : ri_basis => basis_set_ri_aux(ibasis)%gto_basis_set
1306 6 : CALL init_interaction_radii_orb_basis(ri_basis, eps_pgf_orb)
1307 : END DO
1308 : END IF
1309 :
1310 : !We follow the general logic of the compute_mat_P_omega routine
1311 36 : CALL dbt_create(force_data%t_2c_K, t_2c_RI)
1312 36 : CALL dbt_create(force_data%t_2c_K, t_2c_RI_2)
1313 36 : CALL dbt_create(force_data%t_2c_der_ovlp(1), t_2c_AO)
1314 :
1315 36 : CALL get_atomic_kind_set(atomic_kind_set, kind_of=kind_of, atom_of_kind=atom_of_kind)
1316 :
1317 : ! Always do the batching of the MO on mu and sigma, such that it is consistent between
1318 : ! the occupied and the virtual quantities
1319 108 : ALLOCATE (mc_ranges(cut_memory + 1))
1320 108 : mc_ranges(:cut_memory) = starts_array_mc_block(:)
1321 36 : mc_ranges(cut_memory + 1) = ends_array_mc_block(cut_memory) + 1
1322 :
1323 : ! Also need some batching on the RI, because it loses sparsity at some point
1324 36 : n_mem_RI = cut_memory
1325 : CALL create_tensor_batches(force_data%bsizes_RI_split, n_mem_RI, batch_start_RI, batch_end_RI, &
1326 36 : batch_blk_start, batch_blk_end)
1327 108 : ALLOCATE (mc_ranges_RI(n_mem_RI + 1))
1328 108 : mc_ranges_RI(1:n_mem_RI) = batch_blk_start(1:n_mem_RI)
1329 36 : mc_ranges_RI(n_mem_RI + 1) = batch_blk_end(n_mem_RI) + 1
1330 36 : DEALLOCATE (batch_blk_start, batch_blk_end)
1331 :
1332 : !Pre-allocate all required tensors and matrices
1333 36 : CALL dbt_create(t_2c_RI, t_P)
1334 36 : CALL dbt_create(t_2c_RI, t_KBKT)
1335 36 : CALL dbt_create(t_2c_AO, t_dm_occ)
1336 36 : CALL dbt_create(t_2c_AO, t_dm_virt)
1337 :
1338 : !note: t_3c_O and t_3c_M have different mappings (map_1d, map_2d)
1339 36 : CALL dbt_create(t_3c_O, t_M_occ)
1340 36 : CALL dbt_create(t_3c_O, t_M_virt)
1341 36 : CALL dbt_create(t_3c_O, t_3c_0)
1342 :
1343 36 : CALL dbt_create(t_3c_O, t_3c_1)
1344 36 : CALL dbt_create(t_3c_O, t_3c_3)
1345 36 : CALL dbt_create(t_3c_O, t_3c_4)
1346 36 : CALL dbt_create(t_3c_O, t_3c_5)
1347 36 : CALL dbt_create(t_3c_M, t_3c_6)
1348 36 : CALL dbt_create(t_3c_M, t_3c_7)
1349 36 : CALL dbt_create(t_3c_M, t_3c_8)
1350 36 : CALL dbt_create(t_3c_M, t_3c_sparse)
1351 36 : CALL dbt_create(t_3c_O, t_3c_help_1)
1352 36 : CALL dbt_create(t_3c_O, t_3c_help_2)
1353 36 : CALL dbt_create(t_2c_AO, t_R_occ)
1354 36 : CALL dbt_create(t_2c_AO, t_R_virt)
1355 36 : CALL dbt_create(t_3c_M, t_3c_ints)
1356 36 : CALL dbt_create(t_3c_M, t_3c_work)
1357 :
1358 : !Before entring the loop, need to compute the 2c tensors B = (1 + Q(w))^-1 - 1, for each
1359 : !frequency grid point, before doing the transformation to the time grid
1360 416 : ALLOCATE (t_B(num_integ_points))
1361 128 : DO jquad = 1, num_integ_points
1362 128 : CALL dbt_create(t_2c_RI, t_B(jquad))
1363 : END DO
1364 :
1365 200 : ALLOCATE (mat_P_tau(num_integ_points))
1366 128 : DO jquad = 1, num_integ_points
1367 92 : ALLOCATE (mat_P_tau(jquad)%matrix)
1368 128 : CALL dbcsr_create(mat_P_tau(jquad)%matrix, template=mat_P_omega(jquad, ispin)%matrix)
1369 : END DO
1370 :
1371 36 : CALL dbcsr_create(dbcsr_work_symm, template=force_data%G_PQ, matrix_type=dbcsr_type_symmetric)
1372 36 : CALL dbt_create(dbcsr_work_symm, t_2c_tmp)
1373 :
1374 : !loop over freqeuncies
1375 128 : DO iquad = 1, num_integ_points
1376 :
1377 : !calculate (1 + Q(w))^-1 - 1 for the given freq.
1378 : !Always take spin alpha (get 2*alpha in closed shell, and alpha+beta in open-shell)
1379 92 : CALL dbcsr_copy(dbcsr_work_symm, mat_P_omega(iquad, 1)%matrix)
1380 92 : CALL dbt_copy_matrix_to_tensor(dbcsr_work_symm, t_2c_tmp)
1381 92 : CALL dbt_copy(t_2c_tmp, t_2c_RI, move_data=.TRUE.)
1382 :
1383 : CALL dbt_contract(1.0_dp, t_2c_RI, force_data%t_2c_K, 0.0_dp, t_2c_RI_2, &
1384 : contract_1=[2], notcontract_1=[1], &
1385 : contract_2=[1], notcontract_2=[2], &
1386 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
1387 92 : flop=flop, unit_nr=unit_nr_dbcsr)
1388 92 : dbcsr_nflop = dbcsr_nflop + flop
1389 : CALL dbt_contract(1.0_dp, force_data%t_2c_K, t_2c_RI_2, 0.0_dp, t_2c_RI, &
1390 : contract_1=[1], notcontract_1=[2], &
1391 : contract_2=[1], notcontract_2=[2], &
1392 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
1393 92 : flop=flop, unit_nr=unit_nr_dbcsr)
1394 92 : CALL dbt_copy(t_2c_RI, t_2c_tmp, move_data=.TRUE.)
1395 92 : CALL dbt_copy_tensor_to_matrix(t_2c_tmp, dbcsr_work_symm)
1396 92 : CALL dbcsr_add_on_diag(dbcsr_work_symm, 1.0_dp)
1397 :
1398 92 : CALL cp_dbcsr_cholesky_decompose(dbcsr_work_symm, para_env=para_env, blacs_env=blacs_env)
1399 92 : CALL cp_dbcsr_cholesky_invert(dbcsr_work_symm, para_env=para_env, blacs_env=blacs_env, uplo_to_full=.TRUE.)
1400 :
1401 92 : CALL dbcsr_add_on_diag(dbcsr_work_symm, -1.0_dp)
1402 :
1403 372 : DO jquad = 1, num_integ_points
1404 :
1405 : !the P matrix to time.
1406 244 : weight = grid%cosine_frequency_to_time(jquad, iquad)
1407 244 : IF (open_shell) THEN
1408 64 : IF (ispin == 1) THEN
1409 : !mat_P_omega contains the sum of alpha and beta spin => we only want alpha
1410 32 : CALL dbcsr_add(mat_P_tau(jquad)%matrix, mat_P_omega(iquad, 1)%matrix, 1.0_dp, weight)
1411 32 : CALL dbcsr_add(mat_P_tau(jquad)%matrix, mat_P_omega(iquad, 2)%matrix, 1.0_dp, -weight)
1412 : ELSE
1413 32 : CALL dbcsr_add(mat_P_tau(jquad)%matrix, mat_P_omega(iquad, 2)%matrix, 1.0_dp, weight)
1414 : END IF
1415 : ELSE
1416 : !factor 0.5 because originam matrix Q is scaled by 2 in RPA (spin)
1417 180 : weight = 0.5_dp*weight
1418 180 : CALL dbcsr_add(mat_P_tau(jquad)%matrix, mat_P_omega(iquad, 1)%matrix, 1.0_dp, weight)
1419 : END IF
1420 :
1421 : ! The energy derivative uses the weighted adjoint of the forward transform.
1422 : weight = grid%cosine_time_to_frequency(iquad, jquad)* &
1423 244 : grid%frequency_weights(iquad)
1424 244 : CALL dbt_copy_matrix_to_tensor(dbcsr_work_symm, t_2c_tmp)
1425 244 : CALL dbt_scale(t_2c_tmp, weight)
1426 336 : CALL dbt_copy(t_2c_tmp, t_B(jquad), summation=.TRUE., move_data=.TRUE.)
1427 : END DO
1428 : END DO
1429 36 : CALL dbt_destroy(t_2c_tmp)
1430 36 : CALL dbcsr_release(dbcsr_work_symm)
1431 36 : CALL dbt_clear(t_2c_RI)
1432 36 : CALL dbt_clear(t_2c_RI_2)
1433 :
1434 : !Pre-define the sparsity of t_3c_4 as a function of the derivatives
1435 36 : occ_der_AO = 0; nze_der_AO = 0
1436 36 : occ_der_RI = 0; nze_der_RI = 0
1437 144 : DO i_xyz = 1, 3
1438 360 : DO i_mem = 1, cut_memory
1439 : CALL decompress_tensor(force_data%t_3c_der_RI(i_xyz), force_data%t_3c_der_RI_ind(i_mem, i_xyz)%ind, &
1440 216 : force_data%t_3c_der_RI_comp(i_mem, i_xyz), mp2_env%ri_rpa_im_time%eps_compress)
1441 216 : CALL get_tensor_occupancy(force_data%t_3c_der_RI(i_xyz), nze, occ)
1442 216 : occ_der_RI = occ_der_RI + occ
1443 216 : nze_der_RI = nze_der_RI + nze
1444 216 : CALL dbt_copy(force_data%t_3c_der_RI(i_xyz), t_3c_sparse, summation=.TRUE., move_data=.TRUE.)
1445 :
1446 : CALL decompress_tensor(force_data%t_3c_der_AO(i_xyz), force_data%t_3c_der_AO_ind(i_mem, i_xyz)%ind, &
1447 216 : force_data%t_3c_der_AO_comp(i_mem, i_xyz), mp2_env%ri_rpa_im_time%eps_compress)
1448 216 : CALL get_tensor_occupancy(force_data%t_3c_der_AO(i_xyz), nze, occ)
1449 216 : occ_der_AO = occ_der_AO + occ
1450 216 : nze_der_AO = nze_der_AO + nze
1451 216 : CALL dbt_copy(force_data%t_3c_der_AO(i_xyz), t_3c_sparse, order=[1, 3, 2], summation=.TRUE.)
1452 756 : CALL dbt_copy(force_data%t_3c_der_AO(i_xyz), t_3c_sparse, summation=.TRUE., move_data=.TRUE.)
1453 : END DO
1454 : END DO
1455 36 : occ_der_RI = occ_der_RI/3.0_dp
1456 36 : occ_der_AO = occ_der_AO/3.0_dp
1457 36 : nze_der_RI = nze_der_RI/3
1458 36 : nze_der_AO = nze_der_AO/3
1459 :
1460 36 : CALL dbcsr_create(R_occ, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
1461 36 : CALL dbcsr_create(R_virt, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
1462 36 : CALL dbcsr_create(dbcsr_work_symm, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_symmetric)
1463 36 : CALL dbcsr_create(dbcsr_work1, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
1464 36 : CALL dbcsr_create(dbcsr_work2, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
1465 36 : CALL dbcsr_create(dbcsr_work3, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
1466 36 : CALL dbcsr_create(exp_occ, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
1467 36 : CALL dbcsr_create(exp_virt, template=matrix_s(1)%matrix, matrix_type=dbcsr_type_no_symmetry)
1468 36 : IF (use_virial) CALL dbcsr_create(virial_ovlp, template=dbcsr_work1)
1469 :
1470 36 : CALL dbt_batched_contract_init(t_3c_0, batch_range_2=mc_ranges, batch_range_3=mc_ranges)
1471 36 : CALL dbt_batched_contract_init(t_3c_1, batch_range_2=mc_ranges, batch_range_3=mc_ranges)
1472 36 : CALL dbt_batched_contract_init(t_3c_3, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
1473 36 : CALL dbt_batched_contract_init(t_M_occ, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
1474 36 : CALL dbt_batched_contract_init(t_M_virt, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
1475 :
1476 36 : CALL dbt_batched_contract_init(t_3c_ints, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
1477 36 : CALL dbt_batched_contract_init(t_3c_work, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges)
1478 :
1479 : CALL dbt_batched_contract_init(t_3c_4, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
1480 36 : batch_range_3=mc_ranges)
1481 : CALL dbt_batched_contract_init(t_3c_5, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
1482 36 : batch_range_3=mc_ranges)
1483 : CALL dbt_batched_contract_init(t_3c_6, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
1484 36 : batch_range_3=mc_ranges)
1485 : CALL dbt_batched_contract_init(t_3c_7, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
1486 36 : batch_range_3=mc_ranges)
1487 : CALL dbt_batched_contract_init(t_3c_8, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
1488 36 : batch_range_3=mc_ranges)
1489 : CALL dbt_batched_contract_init(t_3c_sparse, batch_range_1=mc_ranges_RI, batch_range_2=mc_ranges, &
1490 36 : batch_range_3=mc_ranges)
1491 :
1492 36 : fac = 1.0_dp/fourpi*mp2_env%ri_rpa%scale_rpa
1493 36 : IF (open_shell) fac = 0.5_dp*fac
1494 :
1495 36 : work_virial = 0.0_dp
1496 36 : work_virial_ovlp = 0.0_dp
1497 128 : DO jquad = 1, num_integ_points
1498 92 : tau = grid%imaginary_time(jquad)
1499 92 : occ_ddint = 0; nze_ddint = 0
1500 :
1501 92 : CALL para_env%sync()
1502 92 : t1 = m_walltime()
1503 :
1504 : !Deal with the force contributions where there is no explicit 3-center quantities, i.e. the
1505 : !forces due to the metric and potential derivatives
1506 92 : CALL dbt_create(mat_P_tau(jquad)%matrix, t_2c_tmp)
1507 92 : CALL dbt_copy_matrix_to_tensor(mat_P_tau(jquad)%matrix, t_2c_tmp)
1508 92 : CALL dbt_copy(t_2c_tmp, t_P, move_data=.TRUE.)
1509 92 : CALL dbt_filter(t_P, eps_filter)
1510 92 : CALL dbt_destroy(t_2c_tmp)
1511 :
1512 : CALL perform_2c_ops(force, t_KBKT, force_data, fac, t_B(jquad), t_P, t_2c_RI, t_2c_RI_2, &
1513 92 : use_virial, atom_of_kind, kind_of, eps_filter, dbcsr_nflop, unit_nr_dbcsr)
1514 92 : CALL get_tensor_occupancy(t_KBKT, nze_KBK, occ_KBK)
1515 :
1516 : !Calculate the pseudo-density matrix in tensor form. There are a few useless arguments for SOS-MP2
1517 : CALL compute_mat_dm_global(grid, nmo, fm_mo_coeff(ispin), homo(ispin), propagator, &
1518 : matrix_s, ispin, Eigenval(:, ispin), e_fermi, eps_filter, &
1519 : mp2_env%ri_rpa_im_time%memory_info, unit_nr, &
1520 92 : jquad, .FALSE., .FALSE., qs_env, dummy_int, dummy_ptr, para_env)
1521 :
1522 92 : CALL dbt_create(propagator(propagator_sector_occupied, jquad, 1)%matrix, t_2c_tmp)
1523 92 : CALL dbt_copy_matrix_to_tensor(propagator(propagator_sector_occupied, jquad, 1)%matrix, t_2c_tmp)
1524 92 : CALL dbt_copy(t_2c_tmp, t_dm_occ, move_data=.TRUE.)
1525 92 : CALL dbt_filter(t_dm_occ, eps_filter)
1526 92 : CALL dbt_destroy(t_2c_tmp)
1527 :
1528 92 : CALL dbt_create(propagator(propagator_sector_virtual, jquad, 1)%matrix, t_2c_tmp)
1529 92 : CALL dbt_copy_matrix_to_tensor(propagator(propagator_sector_virtual, jquad, 1)%matrix, t_2c_tmp)
1530 92 : CALL dbt_copy(t_2c_tmp, t_dm_virt, move_data=.TRUE.)
1531 92 : CALL dbt_filter(t_dm_virt, eps_filter)
1532 92 : CALL dbt_destroy(t_2c_tmp)
1533 :
1534 : !Deal with the 3-center quantities.
1535 : CALL perform_3c_ops(force, t_R_occ, t_R_virt, force_data, fac, cut_memory, n_mem_RI, &
1536 : t_KBKT, t_dm_occ, t_dm_virt, t_3c_O, t_3c_M, t_M_occ, t_M_virt, t_3c_0, t_3c_1, &
1537 : t_3c_3, t_3c_4, t_3c_5, t_3c_6, t_3c_7, t_3c_8, t_3c_sparse, t_3c_help_1, t_3c_help_2, &
1538 : t_3c_ints, t_3c_work, starts_array_mc, ends_array_mc, batch_start_RI, &
1539 : batch_end_RI, t_3c_O_compressed, t_3c_O_ind, use_virial, &
1540 : atom_of_kind, kind_of, eps_filter, occ_ddint, nze_ddint, dbcsr_nflop, &
1541 92 : unit_nr_dbcsr, mp2_env)
1542 :
1543 92 : CALL timeset(routineN//"_dbcsr", handle2)
1544 : !We go back to DBCSR matrices from now on
1545 : !Note: R matrices are in fact symmetric, but use a normal type for convenience
1546 92 : CALL dbt_create(matrix_s(1)%matrix, t_2c_tmp)
1547 92 : CALL dbt_copy(t_R_occ, t_2c_tmp, move_data=.TRUE.)
1548 92 : CALL dbt_copy_tensor_to_matrix(t_2c_tmp, R_occ)
1549 :
1550 92 : CALL dbt_copy(t_R_virt, t_2c_tmp, move_data=.TRUE.)
1551 92 : CALL dbt_copy_tensor_to_matrix(t_2c_tmp, R_virt)
1552 :
1553 : !Iteratively calculate the Y1 and Y2 matrices
1554 92 : CALL dbcsr_copy(dbcsr_work_symm, matrix_ks(ispin)%matrix)
1555 92 : CALL dbcsr_add(dbcsr_work_symm, matrix_s(1)%matrix, 1.0_dp, -e_fermi)
1556 : CALL dbcsr_multiply('N', 'N', tau, force_data%P_occ(ispin)%matrix, &
1557 92 : dbcsr_work_symm, 0.0_dp, dbcsr_work1)
1558 92 : CALL build_Y_matrix(Y_1, dbcsr_work1, force_data%P_occ(ispin)%matrix, R_virt, eps_filter)
1559 92 : CALL matrix_exponential(exp_occ, dbcsr_work1, 1.0_dp, 1.0_dp, eps_filter)
1560 :
1561 : CALL dbcsr_multiply('N', 'N', -tau, force_data%P_virt(ispin)%matrix, &
1562 92 : dbcsr_work_symm, 0.0_dp, dbcsr_work1)
1563 92 : CALL build_Y_matrix(Y_2, dbcsr_work1, force_data%P_virt(ispin)%matrix, R_occ, eps_filter)
1564 92 : CALL matrix_exponential(exp_virt, dbcsr_work1, 1.0_dp, 1.0_dp, eps_filter)
1565 :
1566 : !The force contribution coming from [-S^-1*(e^-tau*P_virt*F)^T*R_occ*S^-1
1567 : ! +tau*S^-1*Y_2^T*F*S^-1] * der_S
1568 : !as well as -tau*e_fermi*Y_1*P^occ + tau*e_fermi*Y_2*P^virt
1569 92 : CALL dbcsr_multiply('N', 'N', 1.0_dp, R_occ, force_data%inv_ovlp, 0.0_dp, dbcsr_work1)
1570 92 : CALL dbcsr_multiply('T', 'N', 1.0_dp, exp_virt, dbcsr_work1, 0.0_dp, dbcsr_work3)
1571 92 : CALL dbcsr_multiply('N', 'N', 1.0_dp, force_data%inv_ovlp, dbcsr_work3, 0.0_dp, dbcsr_work2)
1572 :
1573 92 : CALL dbcsr_multiply('N', 'T', tau, force_data%inv_ovlp, Y_2, 0.0_dp, dbcsr_work3)
1574 92 : CALL dbcsr_multiply('N', 'N', 1.0_dp, dbcsr_work3, dbcsr_work_symm, 0.0_dp, dbcsr_work1)
1575 92 : CALL dbcsr_multiply('N', 'N', -1.0_dp, dbcsr_work1, force_data%inv_ovlp, 1.0_dp, dbcsr_work2)
1576 :
1577 92 : CALL dbcsr_multiply('N', 'T', tau*e_fermi, force_data%P_occ(ispin)%matrix, Y_1, 1.0_dp, dbcsr_work2)
1578 92 : CALL dbcsr_multiply('N', 'T', -tau*e_fermi, force_data%P_virt(ispin)%matrix, Y_2, 1.0_dp, dbcsr_work2)
1579 :
1580 92 : CALL dbt_copy_matrix_to_tensor(dbcsr_work2, t_2c_tmp)
1581 92 : CALL dbt_copy(t_2c_tmp, t_2c_AO, move_data=.TRUE.)
1582 :
1583 92 : pref = -1.0_dp*fac
1584 : CALL get_2c_der_force(force, t_2c_AO, force_data%t_2c_der_ovlp, atom_of_kind, &
1585 92 : kind_of, force_data%idx_to_at_AO, pref, do_ovlp=.TRUE.)
1586 :
1587 92 : IF (use_virial) CALL dbcsr_add(virial_ovlp, dbcsr_work2, 1.0_dp, pref)
1588 :
1589 : !The final contribution from Tr[(tau*Y_1*P_occ - tau*Y_2*P_virt) * der_F]
1590 : CALL dbcsr_multiply('N', 'N', fac*tau, Y_1, force_data%P_occ(ispin)%matrix, 1.0_dp, &
1591 92 : force_data%sum_YP_tau(ispin)%matrix, retain_sparsity=.TRUE.)
1592 : CALL dbcsr_multiply('N', 'N', -fac*tau, Y_2, force_data%P_virt(ispin)%matrix, 1.0_dp, &
1593 92 : force_data%sum_YP_tau(ispin)%matrix, retain_sparsity=.TRUE.)
1594 :
1595 92 : spin_fac = 0.5_dp*fac
1596 92 : IF (open_shell) spin_fac = 2.0_dp*spin_fac
1597 : !Build-up the RHS of the response equation.
1598 : CALL dbcsr_multiply('N', 'N', 1.0_dp*spin_fac, R_virt, exp_occ, 1.0_dp, &
1599 92 : force_data%sum_O_tau(ispin)%matrix, retain_sparsity=.TRUE.)
1600 : CALL dbcsr_multiply('N', 'N', -1.0_dp*spin_fac, R_occ, exp_virt, 1.0_dp, &
1601 92 : force_data%sum_O_tau(ispin)%matrix, retain_sparsity=.TRUE.)
1602 : CALL dbcsr_multiply('N', 'N', tau*spin_fac, dbcsr_work_symm, Y_1, 1.0_dp, &
1603 92 : force_data%sum_O_tau(ispin)%matrix, retain_sparsity=.TRUE.)
1604 : CALL dbcsr_multiply('N', 'N', tau*spin_fac, dbcsr_work_symm, Y_2, 1.0_dp, &
1605 92 : force_data%sum_O_tau(ispin)%matrix, retain_sparsity=.TRUE.)
1606 :
1607 92 : CALL timestop(handle2)
1608 :
1609 : !Print some info
1610 92 : CALL para_env%sync()
1611 92 : t2 = m_walltime()
1612 92 : dbcsr_time = dbcsr_time + t2 - t1
1613 :
1614 92 : IF (unit_nr > 0) THEN
1615 : WRITE (unit_nr, '(/T3,A,1X,I3,A)') &
1616 46 : 'RPA_LOW_SCALING_INFO| Info for time point', jquad, ' (gradients)'
1617 : WRITE (unit_nr, '(T6,A,T56,F25.6)') &
1618 46 : 'Time:', t2 - t1
1619 : WRITE (unit_nr, '(T6,A,T63,ES7.1,1X,A1,1X,F7.3,A1)') &
1620 46 : 'Occupancy of 3c AO derivs:', REAL(nze_der_AO, dp), '/', occ_der_AO*100, '%'
1621 : WRITE (unit_nr, '(T6,A,T63,ES7.1,1X,A1,1X,F7.3,A1)') &
1622 46 : 'Occupancy of 3c RI derivs:', REAL(nze_der_RI, dp), '/', occ_der_RI*100, '%'
1623 : WRITE (unit_nr, '(T6,A,T63,ES7.1,1X,A1,1X,F7.3,A1)') &
1624 46 : 'Occupancy of the Docc * Dvirt * 3c-int tensor', REAL(nze_ddint, dp), '/', occ_ddint*100, '%'
1625 : WRITE (unit_nr, '(T6,A,T63,ES7.1,1X,A1,1X,F7.3,A1)') &
1626 46 : 'Occupancy of KBK^T 2c-tensor:', REAL(nze_KBK, dp), '/', occ_KBK*100, '%'
1627 46 : CALL m_flush(unit_nr)
1628 : END IF
1629 :
1630 : !intermediate clean-up
1631 92 : CALL dbcsr_release(Y_1)
1632 92 : CALL dbcsr_release(Y_2)
1633 496 : CALL dbt_destroy(t_2c_tmp)
1634 :
1635 : END DO !jquad
1636 :
1637 36 : CALL dbt_batched_contract_finalize(t_3c_0)
1638 36 : CALL dbt_batched_contract_finalize(t_3c_1)
1639 36 : CALL dbt_batched_contract_finalize(t_3c_3)
1640 36 : CALL dbt_batched_contract_finalize(t_M_occ)
1641 36 : CALL dbt_batched_contract_finalize(t_M_virt)
1642 :
1643 36 : CALL dbt_batched_contract_finalize(t_3c_ints)
1644 36 : CALL dbt_batched_contract_finalize(t_3c_work)
1645 :
1646 36 : CALL dbt_batched_contract_finalize(t_3c_4)
1647 36 : CALL dbt_batched_contract_finalize(t_3c_5)
1648 36 : CALL dbt_batched_contract_finalize(t_3c_6)
1649 36 : CALL dbt_batched_contract_finalize(t_3c_7)
1650 36 : CALL dbt_batched_contract_finalize(t_3c_8)
1651 36 : CALL dbt_batched_contract_finalize(t_3c_sparse)
1652 :
1653 : !Calculate the 2c and 3c contributions to the virial
1654 36 : IF (use_virial) THEN
1655 2 : CALL dbt_copy(force_data%t_3c_virial_split, force_data%t_3c_virial, move_data=.TRUE.)
1656 : CALL calc_3c_virial(work_virial, force_data%t_3c_virial, 1.0_dp, qs_env, force_data%nl_3c, &
1657 : basis_set_ri_aux, basis_set_ao, basis_set_ao, mp2_env%ri_metric, &
1658 2 : der_eps=mp2_env%ri_rpa_im_time%eps_filter, op_pos=1)
1659 :
1660 : CALL calc_2c_virial(work_virial, force_data%RI_virial_met, 1.0_dp, qs_env, force_data%nl_2c_met, &
1661 2 : basis_set_ri_aux, basis_set_ri_aux, mp2_env%ri_metric)
1662 2 : CALL dbcsr_clear(force_data%RI_virial_met)
1663 :
1664 2 : IF (.NOT. force_data%do_periodic) THEN
1665 : CALL calc_2c_virial(work_virial, force_data%RI_virial_pot, 1.0_dp, qs_env, force_data%nl_2c_pot, &
1666 0 : basis_set_ri_aux, basis_set_ri_aux, mp2_env%potential_parameter)
1667 0 : CALL dbcsr_clear(force_data%RI_virial_pot)
1668 : END IF
1669 :
1670 2 : identity_pot%potential_type = do_potential_id
1671 : CALL calc_2c_virial(work_virial_ovlp, virial_ovlp, 1.0_dp, qs_env, force_data%nl_2c_ovlp, &
1672 2 : basis_set_ao, basis_set_ao, identity_pot)
1673 2 : CALL dbcsr_release(virial_ovlp)
1674 :
1675 8 : DO k_xyz = 1, 3
1676 26 : DO j_xyz = 1, 3
1677 78 : DO i_xyz = 1, 3
1678 : virial%pv_mp2(i_xyz, j_xyz) = virial%pv_mp2(i_xyz, j_xyz) &
1679 54 : - work_virial(i_xyz, k_xyz)*cell%hmat(j_xyz, k_xyz)
1680 : virial%pv_overlap(i_xyz, j_xyz) = virial%pv_overlap(i_xyz, j_xyz) &
1681 54 : - work_virial_ovlp(i_xyz, k_xyz)*cell%hmat(j_xyz, k_xyz)
1682 : virial%pv_virial(i_xyz, j_xyz) = virial%pv_virial(i_xyz, j_xyz) &
1683 : - work_virial(i_xyz, k_xyz)*cell%hmat(j_xyz, k_xyz) &
1684 72 : - work_virial_ovlp(i_xyz, k_xyz)*cell%hmat(j_xyz, k_xyz)
1685 : END DO
1686 : END DO
1687 : END DO
1688 : END IF
1689 :
1690 : !Calculate the periodic contributions of (P|Q) to the force and the virial
1691 36 : work_virial = 0.0_dp
1692 36 : IF (force_data%do_periodic) THEN
1693 18 : IF (mp2_env%eri_method == do_eri_gpw) THEN
1694 6 : CALL get_2c_gpw_forces(force_data%G_PQ, force, work_virial, use_virial, mp2_env, qs_env)
1695 12 : ELSE IF (mp2_env%eri_method == do_eri_mme) THEN
1696 12 : CALL get_2c_mme_forces(force_data%G_PQ, force, mp2_env, qs_env)
1697 12 : IF (use_virial) CPABORT("Stress tensor not available with MME intrgrals")
1698 : ELSE
1699 0 : CPABORT("Periodic case not possible with OS integrals")
1700 : END IF
1701 18 : CALL dbcsr_clear(force_data%G_PQ)
1702 : END IF
1703 :
1704 36 : IF (use_virial) THEN
1705 26 : virial%pv_mp2 = virial%pv_mp2 + work_virial
1706 26 : virial%pv_virial = virial%pv_virial + work_virial
1707 2 : virial%pv_calculate = .FALSE.
1708 :
1709 6 : DO ibasis = 1, SIZE(basis_set_ao)
1710 4 : orb_basis => basis_set_ao(ibasis)%gto_basis_set
1711 4 : CALL init_interaction_radii_orb_basis(orb_basis, eps_pgf_orb_old)
1712 4 : ri_basis => basis_set_ri_aux(ibasis)%gto_basis_set
1713 6 : CALL init_interaction_radii_orb_basis(ri_basis, eps_pgf_orb_old)
1714 : END DO
1715 : END IF
1716 :
1717 : !clean-up
1718 36 : IF (ASSOCIATED(dummy_ptr)) DEALLOCATE (dummy_ptr)
1719 128 : DO jquad = 1, num_integ_points
1720 128 : CALL dbt_destroy(t_B(jquad))
1721 : END DO
1722 36 : CALL dbt_destroy(t_P)
1723 36 : CALL dbt_destroy(t_3c_0)
1724 36 : CALL dbt_destroy(t_3c_1)
1725 36 : CALL dbt_destroy(t_3c_3)
1726 36 : CALL dbt_destroy(t_3c_4)
1727 36 : CALL dbt_destroy(t_3c_5)
1728 36 : CALL dbt_destroy(t_3c_6)
1729 36 : CALL dbt_destroy(t_3c_7)
1730 36 : CALL dbt_destroy(t_3c_8)
1731 36 : CALL dbt_destroy(t_3c_sparse)
1732 36 : CALL dbt_destroy(t_3c_help_1)
1733 36 : CALL dbt_destroy(t_3c_help_2)
1734 36 : CALL dbt_destroy(t_3c_ints)
1735 36 : CALL dbt_destroy(t_3c_work)
1736 36 : CALL dbt_destroy(t_R_occ)
1737 36 : CALL dbt_destroy(t_R_virt)
1738 36 : CALL dbt_destroy(t_dm_occ)
1739 36 : CALL dbt_destroy(t_dm_virt)
1740 36 : CALL dbt_destroy(t_KBKT)
1741 36 : CALL dbt_destroy(t_M_occ)
1742 36 : CALL dbt_destroy(t_M_virt)
1743 36 : CALL dbcsr_release(R_occ)
1744 36 : CALL dbcsr_release(R_virt)
1745 36 : CALL dbcsr_release(dbcsr_work_symm)
1746 36 : CALL dbcsr_release(dbcsr_work1)
1747 36 : CALL dbcsr_release(dbcsr_work2)
1748 36 : CALL dbcsr_release(dbcsr_work3)
1749 36 : CALL dbcsr_release(exp_occ)
1750 36 : CALL dbcsr_release(exp_virt)
1751 :
1752 36 : CALL dbt_destroy(t_2c_RI)
1753 36 : CALL dbt_destroy(t_2c_RI_2)
1754 36 : CALL dbt_destroy(t_2c_AO)
1755 36 : CALL dbcsr_deallocate_matrix_set(propagator)
1756 36 : CALL dbcsr_deallocate_matrix_set(mat_P_tau)
1757 :
1758 36 : CALL timestop(handle)
1759 :
1760 200 : END SUBROUTINE calc_rpa_loop_forces
1761 :
1762 : ! **************************************************************************************************
1763 : !> \brief This subroutines performs the 2c tensor operations that are common accros low-scaling RPA
1764 : !> and SOS-MP2, including forces and virial
1765 : !> \param force ...
1766 : !> \param t_KBKT returns the 2c tensor product of K*B*K^T
1767 : !> \param force_data ...
1768 : !> \param fac ...
1769 : !> \param t_B depending on RPA or SOS-MP2, t_B contains (1 + Q)^-1 - 1 or simply Q, respectively
1770 : !> \param t_P ...
1771 : !> \param t_2c_RI ...
1772 : !> \param t_2c_RI_2 ...
1773 : !> \param use_virial ...
1774 : !> \param atom_of_kind ...
1775 : !> \param kind_of ...
1776 : !> \param eps_filter ...
1777 : !> \param dbcsr_nflop ...
1778 : !> \param unit_nr_dbcsr ...
1779 : ! **************************************************************************************************
1780 170 : SUBROUTINE perform_2c_ops(force, t_KBKT, force_data, fac, t_B, t_P, t_2c_RI, t_2c_RI_2, use_virial, &
1781 170 : atom_of_kind, kind_of, eps_filter, dbcsr_nflop, unit_nr_dbcsr)
1782 :
1783 : TYPE(qs_force_type), DIMENSION(:), POINTER :: force
1784 : TYPE(dbt_type), INTENT(INOUT) :: t_KBKT
1785 : TYPE(im_time_force_type), INTENT(INOUT) :: force_data
1786 : REAL(dp), INTENT(IN) :: fac
1787 : TYPE(dbt_type), INTENT(INOUT) :: t_B, t_P, t_2c_RI, t_2c_RI_2
1788 : LOGICAL, INTENT(IN) :: use_virial
1789 : INTEGER, DIMENSION(:), INTENT(IN) :: atom_of_kind, kind_of
1790 : REAL(dp), INTENT(IN) :: eps_filter
1791 : INTEGER(int_8), INTENT(INOUT) :: dbcsr_nflop
1792 : INTEGER, INTENT(IN) :: unit_nr_dbcsr
1793 :
1794 : CHARACTER(LEN=*), PARAMETER :: routineN = 'perform_2c_ops'
1795 :
1796 : INTEGER :: handle
1797 : INTEGER(int_8) :: flop
1798 : REAL(dp) :: pref
1799 2890 : TYPE(dbt_type) :: t_2c_tmp, t_2c_virial
1800 :
1801 170 : CALL timeset(routineN, handle)
1802 :
1803 170 : IF (use_virial) CALL dbt_create(force_data%RI_virial_pot, t_2c_virial)
1804 :
1805 : !P^T*K*B + P*K*B^T (note we calculate and save K*B*K^T for later, and P=P^T)
1806 : CALL dbt_contract(1.0_dp, force_data%t_2c_K, t_B, 0.0_dp, t_2c_RI, &
1807 : contract_1=[2], notcontract_1=[1], &
1808 : contract_2=[1], notcontract_2=[2], &
1809 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
1810 170 : flop=flop, unit_nr=unit_nr_dbcsr)
1811 170 : dbcsr_nflop = dbcsr_nflop + flop
1812 :
1813 : CALL dbt_contract(1.0_dp, t_2c_RI, force_data%t_2c_K, 0.0_dp, t_KBKT, &
1814 : contract_1=[2], notcontract_1=[1], &
1815 : contract_2=[2], notcontract_2=[1], &
1816 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
1817 170 : flop=flop, unit_nr=unit_nr_dbcsr)
1818 170 : dbcsr_nflop = dbcsr_nflop + flop
1819 :
1820 : CALL dbt_contract(2.0_dp, t_P, t_2c_RI, 0.0_dp, t_2c_RI_2, & !t_2c_RI_2 holds P^T*K*B
1821 : contract_1=[2], notcontract_1=[1], &
1822 : contract_2=[1], notcontract_2=[2], &
1823 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
1824 170 : flop=flop, unit_nr=unit_nr_dbcsr)
1825 170 : dbcsr_nflop = dbcsr_nflop + flop
1826 170 : CALL dbt_clear(t_2c_RI)
1827 : !t_2c_RI_2 currently holds 2*P^T*K*B = P^T*K*B + P*K*B^T (because of symmetry)
1828 :
1829 : !For the metric contribution, we need S^-1*(P^T*K*B + P*K*B^T)*K^T
1830 : CALL dbt_contract(1.0_dp, force_data%t_2c_inv_metric, t_2c_RI_2, 0.0_dp, t_2c_RI, &
1831 : contract_1=[2], notcontract_1=[1], &
1832 : contract_2=[1], notcontract_2=[2], &
1833 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
1834 170 : flop=flop, unit_nr=unit_nr_dbcsr)
1835 170 : dbcsr_nflop = dbcsr_nflop + flop
1836 :
1837 : CALL dbt_contract(1.0_dp, t_2c_RI, force_data%t_2c_K, 0.0_dp, t_2c_RI_2, &
1838 : contract_1=[2], notcontract_1=[1], &
1839 : contract_2=[2], notcontract_2=[1], &
1840 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
1841 170 : flop=flop, unit_nr=unit_nr_dbcsr)
1842 170 : dbcsr_nflop = dbcsr_nflop + flop
1843 :
1844 : !Here we do the trace for the force
1845 170 : pref = -1.0_dp*fac
1846 : CALL get_2c_der_force(force, t_2c_RI_2, force_data%t_2c_der_metric, atom_of_kind, &
1847 170 : kind_of, force_data%idx_to_at_RI, pref, do_mp2=.TRUE.)
1848 170 : IF (use_virial) THEN
1849 12 : CALL dbt_copy(t_2c_RI_2, t_2c_virial)
1850 12 : CALL dbt_scale(t_2c_virial, pref)
1851 12 : CALL dbt_copy_tensor_to_matrix(t_2c_virial, force_data%RI_virial_met, summation=.TRUE.)
1852 12 : CALL dbt_clear(t_2c_virial)
1853 : END IF
1854 :
1855 : !For the potential contribution, we need S^-1*(P^T*K*B + P*K*B^T)*V^-0.5
1856 : !some of it is still in t_2c_RI: ( S^-1*(P^T*K*B + P*K*B^T) )
1857 : CALL dbt_contract(1.0_dp, t_2c_RI, force_data%t_2c_pot_msqrt, 0.0_dp, t_2c_RI_2, &
1858 : contract_1=[2], notcontract_1=[1], &
1859 : contract_2=[1], notcontract_2=[2], &
1860 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
1861 170 : flop=flop, unit_nr=unit_nr_dbcsr)
1862 170 : dbcsr_nflop = dbcsr_nflop + flop
1863 :
1864 : !Here we do the trace for the force. In the periodic case, we store the matrix in G_PQ for later
1865 170 : pref = 0.5_dp*fac
1866 170 : IF (force_data%do_periodic) THEN
1867 76 : CALL dbt_scale(t_2c_RI_2, pref)
1868 76 : CALL dbt_create(force_data%G_PQ, t_2c_tmp)
1869 76 : CALL dbt_copy(t_2c_RI_2, t_2c_tmp, move_data=.TRUE.)
1870 76 : CALL dbt_copy_tensor_to_matrix(t_2c_tmp, force_data%G_PQ, summation=.TRUE.)
1871 76 : CALL dbt_destroy(t_2c_tmp)
1872 : ELSE
1873 : CALL get_2c_der_force(force, t_2c_RI_2, force_data%t_2c_der_pot, atom_of_kind, &
1874 94 : kind_of, force_data%idx_to_at_RI, pref, do_mp2=.TRUE.)
1875 :
1876 94 : IF (use_virial) THEN
1877 0 : CALL dbt_copy(t_2c_RI_2, t_2c_virial)
1878 0 : CALL dbt_scale(t_2c_virial, pref)
1879 0 : CALL dbt_copy_tensor_to_matrix(t_2c_virial, force_data%RI_virial_pot, summation=.TRUE.)
1880 0 : CALL dbt_clear(t_2c_virial)
1881 : END IF
1882 : END IF
1883 :
1884 170 : CALL dbt_clear(t_2c_RI)
1885 170 : CALL dbt_clear(t_2c_RI_2)
1886 :
1887 170 : IF (use_virial) CALL dbt_destroy(t_2c_virial)
1888 :
1889 170 : CALL timestop(handle)
1890 :
1891 170 : END SUBROUTINE perform_2c_ops
1892 :
1893 : ! **************************************************************************************************
1894 : !> \brief This subroutines performs the 3c tensor operations that are common accros low-scaling RPA
1895 : !> and SOS-MP2, including forces and virial
1896 : !> \param force ...
1897 : !> \param t_R_occ ...
1898 : !> \param t_R_virt ...
1899 : !> \param force_data ...
1900 : !> \param fac ...
1901 : !> \param cut_memory ...
1902 : !> \param n_mem_RI ...
1903 : !> \param t_KBKT ...
1904 : !> \param t_dm_occ ...
1905 : !> \param t_dm_virt ...
1906 : !> \param t_3c_O ...
1907 : !> \param t_3c_M ...
1908 : !> \param t_M_occ ...
1909 : !> \param t_M_virt ...
1910 : !> \param t_3c_0 ...
1911 : !> \param t_3c_1 ...
1912 : !> \param t_3c_3 ...
1913 : !> \param t_3c_4 ...
1914 : !> \param t_3c_5 ...
1915 : !> \param t_3c_6 ...
1916 : !> \param t_3c_7 ...
1917 : !> \param t_3c_8 ...
1918 : !> \param t_3c_sparse ...
1919 : !> \param t_3c_help_1 ...
1920 : !> \param t_3c_help_2 ...
1921 : !> \param t_3c_ints ...
1922 : !> \param t_3c_work ...
1923 : !> \param starts_array_mc ...
1924 : !> \param ends_array_mc ...
1925 : !> \param batch_start_RI ...
1926 : !> \param batch_end_RI ...
1927 : !> \param t_3c_O_compressed ...
1928 : !> \param t_3c_O_ind ...
1929 : !> \param use_virial ...
1930 : !> \param atom_of_kind ...
1931 : !> \param kind_of ...
1932 : !> \param eps_filter ...
1933 : !> \param occ_ddint ...
1934 : !> \param nze_ddint ...
1935 : !> \param dbcsr_nflop ...
1936 : !> \param unit_nr_dbcsr ...
1937 : !> \param mp2_env ...
1938 : ! **************************************************************************************************
1939 170 : SUBROUTINE perform_3c_ops(force, t_R_occ, t_R_virt, force_data, fac, cut_memory, n_mem_RI, &
1940 : t_KBKT, t_dm_occ, t_dm_virt, t_3c_O, t_3c_M, t_M_occ, t_M_virt, t_3c_0, t_3c_1, &
1941 : t_3c_3, t_3c_4, t_3c_5, t_3c_6, t_3c_7, t_3c_8, t_3c_sparse, t_3c_help_1, t_3c_help_2, &
1942 170 : t_3c_ints, t_3c_work, starts_array_mc, ends_array_mc, batch_start_RI, &
1943 170 : batch_end_RI, t_3c_O_compressed, t_3c_O_ind, use_virial, &
1944 170 : atom_of_kind, kind_of, eps_filter, occ_ddint, nze_ddint, dbcsr_nflop, &
1945 : unit_nr_dbcsr, mp2_env)
1946 :
1947 : TYPE(qs_force_type), DIMENSION(:), POINTER :: force
1948 : TYPE(dbt_type), INTENT(INOUT) :: t_R_occ, t_R_virt
1949 : TYPE(im_time_force_type), INTENT(INOUT) :: force_data
1950 : REAL(dp), INTENT(IN) :: fac
1951 : INTEGER, INTENT(IN) :: cut_memory, n_mem_RI
1952 : TYPE(dbt_type), INTENT(INOUT) :: t_KBKT, t_dm_occ, t_dm_virt, t_3c_O, t_3c_M, t_M_occ, &
1953 : t_M_virt, t_3c_0, t_3c_1, t_3c_3, t_3c_4, t_3c_5, t_3c_6, t_3c_7, t_3c_8, t_3c_sparse, &
1954 : t_3c_help_1, t_3c_help_2, t_3c_ints, t_3c_work
1955 : INTEGER, DIMENSION(:), INTENT(IN) :: starts_array_mc, ends_array_mc, &
1956 : batch_start_RI, batch_end_RI
1957 : TYPE(hfx_compression_type), DIMENSION(:) :: t_3c_O_compressed
1958 : TYPE(block_ind_type), DIMENSION(:), INTENT(INOUT) :: t_3c_O_ind
1959 : LOGICAL, INTENT(IN) :: use_virial
1960 : INTEGER, DIMENSION(:), INTENT(IN) :: atom_of_kind, kind_of
1961 : REAL(dp), INTENT(IN) :: eps_filter
1962 : REAL(dp), INTENT(INOUT) :: occ_ddint
1963 : INTEGER(int_8), INTENT(INOUT) :: nze_ddint, dbcsr_nflop
1964 : INTEGER, INTENT(IN) :: unit_nr_dbcsr
1965 : TYPE(mp2_type) :: mp2_env
1966 :
1967 : CHARACTER(LEN=*), PARAMETER :: routineN = 'perform_3c_ops'
1968 :
1969 : INTEGER :: dummy_int, handle, handle2, i_mem, &
1970 : i_xyz, j_mem, k_mem
1971 : INTEGER(int_8) :: flop, nze
1972 : INTEGER, DIMENSION(2, 1) :: ibounds, jbounds, kbounds
1973 : INTEGER, DIMENSION(2, 2) :: bounds_2c
1974 : INTEGER, DIMENSION(2, 3) :: bounds_cpy
1975 : INTEGER, DIMENSION(3) :: bounds_3c
1976 : REAL(dp) :: memory, occ, pref
1977 170 : TYPE(block_ind_type), ALLOCATABLE, DIMENSION(:, :) :: blk_indices
1978 : TYPE(hfx_compression_type), ALLOCATABLE, &
1979 170 : DIMENSION(:, :) :: store_3c
1980 :
1981 170 : CALL timeset(routineN, handle)
1982 :
1983 170 : CALL dbt_get_info(t_3c_M, nfull_total=bounds_3c)
1984 :
1985 : !Pre-compute and compress KBK^T * (pq|R)
1986 360910 : ALLOCATE (store_3c(n_mem_RI, cut_memory))
1987 1700 : ALLOCATE (blk_indices(n_mem_RI, cut_memory))
1988 170 : memory = 0.0_dp
1989 170 : CALL timeset(routineN//"_pre_3c", handle2)
1990 : !temporarily build the full int 3c tensor
1991 170 : CALL dbt_copy(t_3c_O, t_3c_0)
1992 510 : DO i_mem = 1, cut_memory
1993 : CALL decompress_tensor(t_3c_O, t_3c_O_ind(i_mem)%ind, t_3c_O_compressed(i_mem), &
1994 340 : mp2_env%ri_rpa_im_time%eps_compress)
1995 340 : CALL dbt_copy(t_3c_O, t_3c_ints)
1996 340 : CALL dbt_copy(t_3c_O, t_3c_0, move_data=.TRUE., summation=.TRUE.)
1997 :
1998 1190 : DO k_mem = 1, n_mem_RI
1999 2040 : kbounds(:, 1) = [batch_start_RI(k_mem), batch_end_RI(k_mem)]
2000 :
2001 680 : CALL alloc_containers(store_3c(k_mem, i_mem), 1)
2002 :
2003 : !contract with KBK^T over the RI index and store
2004 680 : CALL dbt_batched_contract_init(t_KBKT)
2005 : CALL dbt_contract(1.0_dp, t_KBKT, t_3c_ints, 0.0_dp, t_3c_work, &
2006 : contract_1=[2], notcontract_1=[1], &
2007 : contract_2=[1], notcontract_2=[2, 3], &
2008 : map_1=[1], map_2=[2, 3], filter_eps=eps_filter, &
2009 680 : bounds_2=kbounds, flop=flop, unit_nr=unit_nr_dbcsr)
2010 680 : CALL dbt_batched_contract_finalize(t_KBKT)
2011 680 : dbcsr_nflop = dbcsr_nflop + flop
2012 :
2013 680 : CALL dbt_copy(t_3c_work, t_3c_M, move_data=.TRUE.)
2014 : CALL compress_tensor(t_3c_M, blk_indices(k_mem, i_mem)%ind, store_3c(k_mem, i_mem), &
2015 1020 : mp2_env%ri_rpa_im_time%eps_compress, memory)
2016 : END DO
2017 : END DO !i_mem
2018 170 : CALL dbt_clear(t_3c_M)
2019 170 : CALL dbt_copy(t_3c_M, t_3c_ints)
2020 170 : CALL timestop(handle2)
2021 :
2022 170 : CALL dbt_batched_contract_init(t_R_occ)
2023 170 : CALL dbt_batched_contract_init(t_R_virt)
2024 510 : DO i_mem = 1, cut_memory
2025 1020 : ibounds(:, 1) = [starts_array_mc(i_mem), ends_array_mc(i_mem)]
2026 :
2027 : !Compute the matrices M (integrals in t_3c_0)
2028 340 : CALL timeset(routineN//"_3c_M", handle2)
2029 340 : CALL dbt_batched_contract_init(t_dm_occ)
2030 : CALL dbt_contract(1.0_dp, t_3c_0, t_dm_occ, 0.0_dp, t_3c_1, &
2031 : contract_1=[3], notcontract_1=[1, 2], &
2032 : contract_2=[1], notcontract_2=[2], &
2033 : map_1=[1, 2], map_2=[3], filter_eps=eps_filter, &
2034 340 : bounds_3=ibounds, flop=flop, unit_nr=unit_nr_dbcsr)
2035 340 : dbcsr_nflop = dbcsr_nflop + flop
2036 340 : CALL dbt_batched_contract_finalize(t_dm_occ)
2037 340 : CALL dbt_copy(t_3c_1, t_M_occ, order=[1, 3, 2], move_data=.TRUE.)
2038 :
2039 340 : CALL dbt_batched_contract_init(t_dm_virt)
2040 : CALL dbt_contract(1.0_dp, t_3c_0, t_dm_virt, 0.0_dp, t_3c_1, &
2041 : contract_1=[3], notcontract_1=[1, 2], &
2042 : contract_2=[1], notcontract_2=[2], &
2043 : map_1=[1, 2], map_2=[3], filter_eps=eps_filter, &
2044 340 : bounds_3=ibounds, flop=flop, unit_nr=unit_nr_dbcsr)
2045 340 : dbcsr_nflop = dbcsr_nflop + flop
2046 340 : CALL dbt_batched_contract_finalize(t_dm_virt)
2047 340 : CALL dbt_copy(t_3c_1, t_M_virt, order=[1, 3, 2], move_data=.TRUE.)
2048 340 : CALL timestop(handle2)
2049 :
2050 : !Compute the R matrices
2051 340 : CALL timeset(routineN//"_3c_R", handle2)
2052 1020 : DO k_mem = 1, n_mem_RI
2053 : CALL decompress_tensor(t_3c_M, blk_indices(k_mem, i_mem)%ind, store_3c(k_mem, i_mem), &
2054 680 : mp2_env%ri_rpa_im_time%eps_compress)
2055 680 : CALL dbt_copy(t_3c_M, t_3c_3, move_data=.TRUE.)
2056 :
2057 : CALL dbt_contract(1.0_dp, t_M_occ, t_3c_3, 1.0_dp, t_R_occ, &
2058 : contract_1=[1, 2], notcontract_1=[3], &
2059 : contract_2=[1, 2], notcontract_2=[3], &
2060 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
2061 680 : flop=flop, unit_nr=unit_nr_dbcsr)
2062 680 : dbcsr_nflop = dbcsr_nflop + flop
2063 :
2064 : CALL dbt_contract(1.0_dp, t_M_virt, t_3c_3, 1.0_dp, t_R_virt, &
2065 : contract_1=[1, 2], notcontract_1=[3], &
2066 : contract_2=[1, 2], notcontract_2=[3], &
2067 : map_1=[1], map_2=[2], filter_eps=eps_filter, &
2068 680 : flop=flop, unit_nr=unit_nr_dbcsr)
2069 1020 : dbcsr_nflop = dbcsr_nflop + flop
2070 : END DO
2071 340 : CALL dbt_copy(t_3c_M, t_3c_3)
2072 340 : CALL dbt_copy(t_3c_M, t_M_virt)
2073 340 : CALL timestop(handle2)
2074 :
2075 340 : CALL dbt_copy(t_M_occ, t_3c_4, move_data=.TRUE.)
2076 :
2077 340 : IF (cut_memory > 0) CALL dbt_batched_contract_init(t_KBKT)
2078 1020 : DO j_mem = 1, cut_memory
2079 2040 : jbounds(:, 1) = [starts_array_mc(j_mem), ends_array_mc(j_mem)]
2080 :
2081 2040 : bounds_cpy(:, 1) = [1, bounds_3c(1)]
2082 2040 : bounds_cpy(:, 2) = [starts_array_mc(i_mem), ends_array_mc(i_mem)]
2083 2040 : bounds_cpy(:, 3) = [starts_array_mc(j_mem), ends_array_mc(j_mem)]
2084 680 : CALL dbt_copy(t_3c_sparse, t_3c_7, bounds=bounds_cpy)
2085 :
2086 680 : CALL dbt_batched_contract_init(t_dm_virt)
2087 2040 : DO k_mem = 1, n_mem_RI
2088 4080 : bounds_2c(:, 1) = [batch_start_RI(k_mem), batch_end_RI(k_mem)]
2089 4080 : bounds_2c(:, 2) = [starts_array_mc(i_mem), ends_array_mc(i_mem)]
2090 :
2091 1360 : CALL timeset(routineN//"_3c_dm", handle2)
2092 :
2093 : !Calculate (mu nu| P) * D_occ * D_virt
2094 : !Note: technically need M_occ*D_virt + M_virt*D_occ, but it is equivalent to 2*M_occ*D_virt
2095 : CALL dbt_contract(2.0_dp, t_3c_4, t_dm_virt, 0.0_dp, t_3c_5, &
2096 : contract_1=[3], notcontract_1=[1, 2], &
2097 : contract_2=[1], notcontract_2=[2], &
2098 : map_1=[1, 2], map_2=[3], filter_eps=eps_filter, &
2099 1360 : bounds_2=bounds_2c, bounds_3=jbounds, flop=flop, unit_nr=unit_nr_dbcsr)
2100 1360 : dbcsr_nflop = dbcsr_nflop + flop
2101 :
2102 1360 : CALL get_tensor_occupancy(t_3c_5, nze, occ)
2103 1360 : nze_ddint = nze_ddint + nze
2104 1360 : occ_ddint = occ_ddint + occ
2105 :
2106 : ! Skip the expensive KBK^T contraction when the intermediate block is empty
2107 1360 : IF (nze == 0) THEN
2108 0 : CALL dbt_clear(t_3c_5)
2109 0 : CYCLE
2110 : END IF
2111 :
2112 1360 : CALL dbt_copy(t_3c_5, t_3c_6, move_data=.TRUE.)
2113 1360 : CALL timestop(handle2)
2114 :
2115 : !Calculate the contraction of the above with K*B*K^T
2116 1360 : CALL timeset(routineN//"_3c_KBK", handle2)
2117 : CALL dbt_contract(1.0_dp, t_KBKT, t_3c_6, 0.0_dp, t_3c_7, &
2118 : contract_1=[2], notcontract_1=[1], &
2119 : contract_2=[1], notcontract_2=[2, 3], &
2120 : map_1=[1], map_2=[2, 3], &
2121 1360 : retain_sparsity=.TRUE., flop=flop, unit_nr=unit_nr_dbcsr)
2122 1360 : dbcsr_nflop = dbcsr_nflop + flop
2123 1360 : CALL timestop(handle2)
2124 6120 : CALL dbt_copy(t_3c_7, t_3c_8, summation=.TRUE.)
2125 :
2126 : END DO !k_mem
2127 1020 : CALL dbt_batched_contract_finalize(t_dm_virt)
2128 : END DO !j_mem
2129 340 : IF (cut_memory > 0) CALL dbt_batched_contract_finalize(t_KBKT)
2130 :
2131 340 : CALL dbt_copy(t_3c_8, t_3c_help_1, move_data=.TRUE.)
2132 :
2133 340 : pref = 1.0_dp*fac
2134 1020 : DO k_mem = 1, cut_memory
2135 2720 : DO i_xyz = 1, 3
2136 2040 : CALL dbt_clear(force_data%t_3c_der_RI(i_xyz))
2137 : CALL decompress_tensor(force_data%t_3c_der_RI(i_xyz), force_data%t_3c_der_RI_ind(k_mem, i_xyz)%ind, &
2138 2720 : force_data%t_3c_der_RI_comp(k_mem, i_xyz), mp2_env%ri_rpa_im_time%eps_compress)
2139 : END DO
2140 : CALL get_force_from_3c_trace(force, t_3c_help_1, force_data%t_3c_der_RI, atom_of_kind, kind_of, &
2141 1020 : force_data%idx_to_at_RI, pref, do_mp2=.TRUE., deriv_dim=1)
2142 : END DO
2143 :
2144 340 : IF (use_virial) THEN
2145 24 : CALL dbt_copy(t_3c_help_1, t_3c_help_2)
2146 24 : CALL dbt_scale(t_3c_help_2, pref)
2147 24 : CALL dbt_copy(t_3c_help_2, force_data%t_3c_virial_split, summation=.TRUE., move_data=.TRUE.)
2148 : END IF
2149 :
2150 340 : CALL dbt_copy(t_3c_help_1, t_3c_help_2)
2151 340 : CALL dbt_copy(t_3c_help_1, t_3c_help_2, order=[1, 3, 2], move_data=.TRUE., summation=.TRUE.)
2152 1020 : DO k_mem = 1, cut_memory
2153 2720 : DO i_xyz = 1, 3
2154 2040 : CALL dbt_clear(force_data%t_3c_der_AO(i_xyz))
2155 : CALL decompress_tensor(force_data%t_3c_der_AO(i_xyz), force_data%t_3c_der_AO_ind(k_mem, i_xyz)%ind, &
2156 2720 : force_data%t_3c_der_AO_comp(k_mem, i_xyz), mp2_env%ri_rpa_im_time%eps_compress)
2157 : END DO
2158 : CALL get_force_from_3c_trace(force, t_3c_help_2, force_data%t_3c_der_AO, atom_of_kind, kind_of, &
2159 1020 : force_data%idx_to_at_AO, pref, do_mp2=.TRUE., deriv_dim=3)
2160 : END DO
2161 :
2162 1190 : CALL dbt_clear(t_3c_help_2)
2163 : END DO !i_mem
2164 170 : CALL dbt_batched_contract_finalize(t_R_occ)
2165 170 : CALL dbt_batched_contract_finalize(t_R_virt)
2166 :
2167 510 : DO k_mem = 1, n_mem_RI
2168 1190 : DO i_mem = 1, cut_memory
2169 1020 : CALL dealloc_containers(store_3c(k_mem, i_mem), dummy_int)
2170 : END DO
2171 : END DO
2172 850 : DEALLOCATE (store_3c, blk_indices)
2173 :
2174 170 : CALL timestop(handle)
2175 :
2176 340 : END SUBROUTINE perform_3c_ops
2177 :
2178 : ! **************************************************************************************************
2179 : !> \brief All the forces that can be calculated after the loop on the Laplace quaradture, using
2180 : !> terms collected during the said loop. This inludes the z-vector equation and its reponse
2181 : !> forces, as well as the force coming from the trace with the derivative of the KS matrix
2182 : !> \param force_data ...
2183 : !> \param unit_nr ...
2184 : !> \param qs_env ...
2185 : ! **************************************************************************************************
2186 50 : SUBROUTINE calc_post_loop_forces(force_data, unit_nr, qs_env)
2187 :
2188 : TYPE(im_time_force_type), INTENT(INOUT) :: force_data
2189 : INTEGER, INTENT(IN) :: unit_nr
2190 : TYPE(qs_environment_type), POINTER :: qs_env
2191 :
2192 : CHARACTER(len=*), PARAMETER :: routineN = 'calc_post_loop_forces'
2193 :
2194 : INTEGER :: handle, ispin, nao, nao_aux, nocc, nspins
2195 : LOGICAL :: do_exx
2196 : REAL(dp) :: focc
2197 : TYPE(admm_type), POINTER :: admm_env
2198 : TYPE(cp_fm_struct_type), POINTER :: fm_struct
2199 50 : TYPE(cp_fm_type), ALLOCATABLE, DIMENSION(:) :: cpmos, mo_occ
2200 : TYPE(cp_fm_type), POINTER :: mo_coeff
2201 50 : TYPE(dbcsr_p_type), DIMENSION(:), POINTER :: dbcsr_p_work, matrix_p_mp2, &
2202 50 : matrix_p_mp2_admm, matrix_s, &
2203 50 : matrix_s_aux, work_admm, YP_admm
2204 : TYPE(dft_control_type), POINTER :: dft_control
2205 : TYPE(linres_control_type), POINTER :: linres_control
2206 50 : TYPE(mo_set_type), DIMENSION(:), POINTER :: mos
2207 : TYPE(qs_p_env_type), POINTER :: p_env
2208 : TYPE(section_vals_type), POINTER :: hfx_section, lr_section
2209 :
2210 50 : NULLIFY (linres_control, p_env, dft_control, matrix_s, mos, mo_coeff, fm_struct, lr_section, &
2211 50 : dbcsr_p_work, YP_admm, matrix_p_mp2, admm_env, work_admm, matrix_s_aux, matrix_p_mp2_admm)
2212 :
2213 50 : CALL timeset(routineN, handle)
2214 :
2215 50 : CALL get_qs_env(qs_env, dft_control=dft_control, matrix_s=matrix_s, mos=mos)
2216 50 : nspins = dft_control%nspins
2217 :
2218 : ! Setting up for the z-vector equation
2219 :
2220 : ! Initialize linres_control
2221 50 : lr_section => section_vals_get_subs_vals(qs_env%input, "DFT%XC%WF_CORRELATION%LOW_SCALING%CPHF")
2222 :
2223 50 : ALLOCATE (linres_control)
2224 50 : CALL section_vals_val_get(lr_section, "MAX_ITER", i_val=linres_control%max_iter)
2225 50 : CALL section_vals_val_get(lr_section, "EPS_CONV", r_val=linres_control%eps)
2226 50 : CALL section_vals_val_get(lr_section, "PRECONDITIONER", i_val=linres_control%preconditioner_type)
2227 50 : CALL section_vals_val_get(lr_section, "ENERGY_GAP", r_val=linres_control%energy_gap)
2228 :
2229 50 : linres_control%do_kernel = .TRUE.
2230 50 : linres_control%lr_triplet = .FALSE.
2231 50 : linres_control%converged = .FALSE.
2232 50 : linres_control%eps_filter = qs_env%mp2_env%ri_rpa_im_time%eps_filter
2233 :
2234 50 : CALL set_qs_env(qs_env, linres_control=linres_control)
2235 :
2236 50 : IF (unit_nr > 0) THEN
2237 25 : WRITE (unit_nr, *)
2238 25 : WRITE (unit_nr, '(T3,A)') 'MP2_CPHF| Iterative solution of Z-Vector equations'
2239 25 : WRITE (unit_nr, '(T3,A,T45,ES8.1)') 'MP2_CPHF| Convergence threshold:', linres_control%eps
2240 25 : WRITE (unit_nr, '(T3,A,T45,I8)') 'MP2_CPHF| Maximum number of iterations: ', linres_control%max_iter
2241 : END IF
2242 :
2243 350 : ALLOCATE (p_env)
2244 50 : CALL p_env_create(p_env, qs_env, orthogonal_orbitals=.TRUE., linres_control=linres_control)
2245 50 : CALL p_env_psi0_changed(p_env, qs_env)
2246 :
2247 : ! Matrix allocation
2248 50 : CALL dbcsr_allocate_matrix_set(p_env%p1, nspins)
2249 50 : CALL dbcsr_allocate_matrix_set(p_env%w1, nspins)
2250 50 : CALL dbcsr_allocate_matrix_set(dbcsr_p_work, nspins)
2251 112 : DO ispin = 1, nspins
2252 62 : ALLOCATE (p_env%p1(ispin)%matrix, p_env%w1(ispin)%matrix, dbcsr_p_work(ispin)%matrix)
2253 62 : CALL dbcsr_create(matrix=p_env%p1(ispin)%matrix, template=matrix_s(1)%matrix)
2254 62 : CALL dbcsr_create(matrix=p_env%w1(ispin)%matrix, template=matrix_s(1)%matrix)
2255 62 : CALL dbcsr_create(matrix=dbcsr_p_work(ispin)%matrix, template=matrix_s(1)%matrix)
2256 62 : CALL dbcsr_copy(p_env%p1(ispin)%matrix, matrix_s(1)%matrix)
2257 62 : CALL dbcsr_copy(p_env%w1(ispin)%matrix, matrix_s(1)%matrix)
2258 62 : CALL dbcsr_copy(dbcsr_p_work(ispin)%matrix, matrix_s(1)%matrix)
2259 62 : CALL dbcsr_set(p_env%p1(ispin)%matrix, 0.0_dp)
2260 62 : CALL dbcsr_set(p_env%w1(ispin)%matrix, 0.0_dp)
2261 112 : CALL dbcsr_set(dbcsr_p_work(ispin)%matrix, 0.0_dp)
2262 : END DO
2263 :
2264 50 : IF (dft_control%do_admm) THEN
2265 16 : CALL get_admm_env(qs_env%admm_env, matrix_s_aux_fit=matrix_s_aux)
2266 16 : CALL dbcsr_allocate_matrix_set(p_env%p1_admm, nspins)
2267 16 : CALL dbcsr_allocate_matrix_set(work_admm, nspins)
2268 36 : DO ispin = 1, nspins
2269 20 : ALLOCATE (p_env%p1_admm(ispin)%matrix, work_admm(ispin)%matrix)
2270 20 : CALL dbcsr_create(p_env%p1_admm(ispin)%matrix, template=matrix_s_aux(1)%matrix)
2271 20 : CALL dbcsr_copy(p_env%p1_admm(ispin)%matrix, matrix_s_aux(1)%matrix)
2272 20 : CALL dbcsr_set(p_env%p1_admm(ispin)%matrix, 0.0_dp)
2273 20 : CALL dbcsr_create(work_admm(ispin)%matrix, template=matrix_s_aux(1)%matrix)
2274 20 : CALL dbcsr_copy(work_admm(ispin)%matrix, matrix_s_aux(1)%matrix)
2275 36 : CALL dbcsr_set(work_admm(ispin)%matrix, 0.0_dp)
2276 : END DO
2277 : END IF
2278 :
2279 : ! Preparing the RHS of the z-vector equation
2280 50 : CALL prepare_for_response(force_data, qs_env)
2281 324 : ALLOCATE (cpmos(nspins), mo_occ(nspins))
2282 112 : DO ispin = 1, nspins
2283 62 : CALL get_mo_set(mo_set=mos(ispin), mo_coeff=mo_coeff, nao=nao, homo=nocc)
2284 62 : NULLIFY (fm_struct)
2285 : CALL cp_fm_struct_create(fm_struct, ncol_global=nocc, &
2286 62 : template_fmstruct=mo_coeff%matrix_struct)
2287 62 : CALL cp_fm_create(cpmos(ispin), fm_struct)
2288 62 : CALL cp_fm_set_all(cpmos(ispin), 0.0_dp)
2289 62 : CALL cp_fm_create(mo_occ(ispin), fm_struct)
2290 62 : CALL cp_fm_to_fm(mo_coeff, mo_occ(ispin), nocc)
2291 174 : CALL cp_fm_struct_release(fm_struct)
2292 : END DO
2293 :
2294 : ! in case of EXX, need to add the HF Hamiltonian to the RHS of the Z-vector equation
2295 : ! Strategy: we take the ks_matrix, remove the current xc contribution, and then add the RPA HF one
2296 50 : do_exx = .FALSE.
2297 50 : IF (qs_env%mp2_env%method == ri_rpa_method_gpw) THEN
2298 28 : hfx_section => section_vals_get_subs_vals(qs_env%input, "DFT%XC%WF_CORRELATION%RI_RPA%HF")
2299 28 : CALL section_vals_get(hfx_section, explicit=do_exx)
2300 : END IF
2301 :
2302 50 : IF (do_exx) THEN
2303 : CALL add_exx_to_rhs(rhs=force_data%sum_O_tau, &
2304 : qs_env=qs_env, &
2305 : ext_hfx_section=hfx_section, &
2306 : x_data=qs_env%mp2_env%ri_rpa%x_data, &
2307 : recalc_integrals=.FALSE., &
2308 : do_admm=qs_env%mp2_env%ri_rpa%do_admm, &
2309 : do_exx=do_exx, &
2310 18 : reuse_hfx=qs_env%mp2_env%ri_rpa%reuse_hfx)
2311 : END IF
2312 :
2313 50 : focc = 2.0_dp
2314 50 : IF (nspins == 1) focc = 4.0_dp
2315 112 : DO ispin = 1, nspins
2316 62 : CALL get_mo_set(mo_set=mos(ispin), mo_coeff=mo_coeff, homo=nocc)
2317 : CALL cp_dbcsr_sm_fm_multiply(force_data%sum_O_tau(ispin)%matrix, mo_occ(ispin), &
2318 : cpmos(ispin), nocc, &
2319 112 : alpha=focc, beta=0.0_dp)
2320 : END DO
2321 :
2322 : ! The z-vector equation and associated forces
2323 50 : CALL response_equation_new(qs_env, p_env, cpmos, unit_nr)
2324 :
2325 : ! Save the mp2 density matrix
2326 50 : CALL get_qs_env(qs_env, matrix_p_mp2=matrix_p_mp2)
2327 50 : IF (ASSOCIATED(matrix_p_mp2)) CALL dbcsr_deallocate_matrix_set(matrix_p_mp2)
2328 112 : DO ispin = 1, nspins
2329 62 : CALL dbcsr_copy(dbcsr_p_work(ispin)%matrix, p_env%p1(ispin)%matrix)
2330 112 : CALL dbcsr_add(dbcsr_p_work(ispin)%matrix, force_data%sum_YP_tau(ispin)%matrix, 1.0_dp, 1.0_dp)
2331 : END DO
2332 50 : CALL set_ks_env(qs_env%ks_env, matrix_p_mp2=dbcsr_p_work)
2333 :
2334 50 : IF (dft_control%do_admm) THEN
2335 16 : CALL dbcsr_allocate_matrix_set(YP_admm, nspins)
2336 16 : CALL get_qs_env(qs_env, matrix_p_mp2_admm=matrix_p_mp2_admm, admm_env=admm_env)
2337 16 : nao = admm_env%nao_orb
2338 16 : nao_aux = admm_env%nao_aux_fit
2339 16 : IF (ASSOCIATED(matrix_p_mp2_admm)) CALL dbcsr_deallocate_matrix_set(matrix_p_mp2_admm)
2340 36 : DO ispin = 1, nspins
2341 :
2342 : !sum_YP_tau in the auxiliary basis
2343 20 : CALL copy_dbcsr_to_fm(force_data%sum_YP_tau(ispin)%matrix, admm_env%work_orb_orb)
2344 : CALL parallel_gemm('N', 'N', nao_aux, nao, nao, 1.0_dp, admm_env%A, admm_env%work_orb_orb, &
2345 20 : 0.0_dp, admm_env%work_aux_orb)
2346 : CALL parallel_gemm('N', 'T', nao_aux, nao_aux, nao, 1.0_dp, admm_env%work_aux_orb, admm_env%A, &
2347 20 : 0.0_dp, admm_env%work_aux_aux)
2348 20 : CALL copy_fm_to_dbcsr(admm_env%work_aux_aux, work_admm(ispin)%matrix, keep_sparsity=.TRUE.)
2349 :
2350 : !save the admm representation od sum_YP_tau
2351 20 : ALLOCATE (YP_admm(ispin)%matrix)
2352 20 : CALL dbcsr_create(YP_admm(ispin)%matrix, template=work_admm(ispin)%matrix)
2353 20 : CALL dbcsr_copy(YP_admm(ispin)%matrix, work_admm(ispin)%matrix)
2354 :
2355 36 : CALL dbcsr_add(work_admm(ispin)%matrix, p_env%p1_admm(ispin)%matrix, 1.0_dp, 1.0_dp)
2356 :
2357 : END DO
2358 16 : CALL set_ks_env(qs_env%ks_env, matrix_p_mp2_admm=work_admm)
2359 : END IF
2360 :
2361 : !Calculate the response force and the force from the trace with F
2362 50 : CALL update_im_time_forces(p_env, force_data%sum_O_tau, force_data%sum_YP_tau, YP_admm, qs_env)
2363 :
2364 : !clean-up
2365 50 : IF (dft_control%do_admm) CALL dbcsr_deallocate_matrix_set(YP_admm)
2366 :
2367 50 : CALL cp_fm_release(cpmos)
2368 50 : CALL cp_fm_release(mo_occ)
2369 50 : CALL p_env_release(p_env)
2370 50 : DEALLOCATE (p_env)
2371 :
2372 50 : CALL timestop(handle)
2373 :
2374 100 : END SUBROUTINE calc_post_loop_forces
2375 :
2376 : ! **************************************************************************************************
2377 : !> \brief Prepares the RHS of the z-vector equation. Apply the xc and HFX kernel on the previously
2378 : !> stored sum_YP_tau density, and add it to the final force_data%sum_O_tau quantity
2379 : !> \param force_data ...
2380 : !> \param qs_env ...
2381 : ! **************************************************************************************************
2382 50 : SUBROUTINE prepare_for_response(force_data, qs_env)
2383 :
2384 : TYPE(im_time_force_type), INTENT(INOUT) :: force_data
2385 : TYPE(qs_environment_type), POINTER :: qs_env
2386 :
2387 : CHARACTER(LEN=*), PARAMETER :: routineN = 'prepare_for_response'
2388 :
2389 : INTEGER :: handle, ispin, nao, nao_aux, nspins
2390 : LOGICAL :: do_hfx, do_tau, do_tau_admm
2391 : REAL(dp) :: ehartree
2392 : TYPE(admm_type), POINTER :: admm_env
2393 50 : TYPE(dbcsr_p_type), DIMENSION(:), POINTER :: dbcsr_p_work, ker_tau_admm, matrix_s, &
2394 50 : matrix_s_aux, work_admm
2395 : TYPE(dbcsr_type) :: dbcsr_work
2396 : TYPE(dft_control_type), POINTER :: dft_control
2397 : TYPE(pw_c1d_gs_type) :: rhoz_tot_gspace, zv_hartree_gspace
2398 50 : TYPE(pw_c1d_gs_type), DIMENSION(:), POINTER :: rhoz_g
2399 : TYPE(pw_env_type), POINTER :: pw_env
2400 : TYPE(pw_poisson_type), POINTER :: poisson_env
2401 : TYPE(pw_pool_type), POINTER :: auxbas_pw_pool
2402 : TYPE(pw_r3d_rs_type) :: zv_hartree_rspace
2403 50 : TYPE(pw_r3d_rs_type), DIMENSION(:), POINTER :: rhoz_r, tauz_r, v_xc, v_xc_tau
2404 : TYPE(qs_rho_type), POINTER :: rho, rho_aux_fit, rhoz
2405 50 : TYPE(rho_atom_type), DIMENSION(:), POINTER :: rho0_atom_set, rho1_atom_set
2406 : TYPE(section_vals_type), POINTER :: hfx_section, xc_section
2407 : TYPE(task_list_type), POINTER :: task_list_aux_fit
2408 :
2409 50 : NULLIFY (pw_env, rhoz_r, rhoz_g, tauz_r, v_xc, v_xc_tau, &
2410 50 : poisson_env, auxbas_pw_pool, dft_control, admm_env, xc_section, rho, rho_aux_fit, &
2411 50 : task_list_aux_fit, ker_tau_admm, work_admm, dbcsr_p_work, matrix_s, hfx_section)
2412 50 : NULLIFY (rho0_atom_set, rho1_atom_set)
2413 :
2414 50 : CALL timeset(routineN, handle)
2415 :
2416 50 : CALL get_qs_env(qs_env, dft_control=dft_control, pw_env=pw_env, rho=rho, matrix_s=matrix_s)
2417 50 : CALL pw_env_get(pw_env, auxbas_pw_pool=auxbas_pw_pool, poisson_env=poisson_env)
2418 50 : nspins = dft_control%nspins
2419 :
2420 50 : CALL dbcsr_allocate_matrix_set(dbcsr_p_work, nspins)
2421 112 : DO ispin = 1, nspins
2422 62 : ALLOCATE (dbcsr_p_work(ispin)%matrix)
2423 62 : CALL dbcsr_create(matrix=dbcsr_p_work(ispin)%matrix, template=matrix_s(1)%matrix)
2424 62 : CALL dbcsr_copy(dbcsr_p_work(ispin)%matrix, matrix_s(1)%matrix)
2425 112 : CALL dbcsr_set(dbcsr_p_work(ispin)%matrix, 0.0_dp)
2426 : END DO
2427 :
2428 : !Apply the kernel on the density saved in force_data%sum_YP_tau
2429 374 : ALLOCATE (rhoz_r(nspins), rhoz_g(nspins))
2430 112 : DO ispin = 1, nspins
2431 62 : CALL auxbas_pw_pool%create_pw(rhoz_r(ispin))
2432 112 : CALL auxbas_pw_pool%create_pw(rhoz_g(ispin))
2433 : END DO
2434 50 : CALL auxbas_pw_pool%create_pw(rhoz_tot_gspace)
2435 50 : CALL auxbas_pw_pool%create_pw(zv_hartree_rspace)
2436 50 : CALL auxbas_pw_pool%create_pw(zv_hartree_gspace)
2437 :
2438 50 : CALL pw_zero(rhoz_tot_gspace)
2439 112 : DO ispin = 1, nspins
2440 : CALL calculate_rho_elec(ks_env=qs_env%ks_env, matrix_p=force_data%sum_YP_tau(ispin)%matrix, &
2441 62 : rho=rhoz_r(ispin), rho_gspace=rhoz_g(ispin))
2442 112 : CALL pw_axpy(rhoz_g(ispin), rhoz_tot_gspace)
2443 : END DO
2444 :
2445 : CALL pw_poisson_solve(poisson_env, rhoz_tot_gspace, ehartree, &
2446 50 : zv_hartree_gspace)
2447 :
2448 50 : CALL pw_transfer(zv_hartree_gspace, zv_hartree_rspace)
2449 50 : CALL pw_scale(zv_hartree_rspace, zv_hartree_rspace%pw_grid%dvol)
2450 :
2451 50 : CALL qs_rho_get(rho, tau_r_valid=do_tau)
2452 50 : IF (do_tau) THEN
2453 : BLOCK
2454 : TYPE(pw_c1d_gs_type) :: tauz_g
2455 24 : ALLOCATE (tauz_r(nspins))
2456 8 : CALL auxbas_pw_pool%create_pw(tauz_g)
2457 16 : DO ispin = 1, nspins
2458 8 : CALL auxbas_pw_pool%create_pw(tauz_r(ispin))
2459 :
2460 : CALL calculate_rho_elec(ks_env=qs_env%ks_env, matrix_p=force_data%sum_YP_tau(ispin)%matrix, &
2461 16 : rho=tauz_r(ispin), rho_gspace=tauz_g, compute_tau=.TRUE.)
2462 : END DO
2463 16 : CALL auxbas_pw_pool%give_back_pw(tauz_g)
2464 : END BLOCK
2465 : END IF
2466 :
2467 50 : IF (dft_control%do_admm) THEN
2468 16 : CALL get_qs_env(qs_env, admm_env=admm_env)
2469 16 : xc_section => admm_env%xc_section_primary
2470 : ELSE
2471 34 : xc_section => section_vals_get_subs_vals(qs_env%input, "DFT%XC")
2472 : END IF
2473 :
2474 : !Primary XC kernel
2475 50 : ALLOCATE (rhoz)
2476 50 : CALL qs_rho_create(rhoz)
2477 50 : IF (ASSOCIATED(rhoz_r)) THEN
2478 50 : CALL qs_rho_set(rhoz, rho_r=rhoz_r, rho_r_valid=.TRUE.)
2479 : END IF
2480 50 : IF (ASSOCIATED(rhoz_g)) THEN
2481 50 : CALL qs_rho_set(rhoz, rho_g=rhoz_g, rho_g_valid=.TRUE.)
2482 : END IF
2483 50 : IF (ASSOCIATED(tauz_r)) THEN
2484 8 : CALL qs_rho_set(rhoz, tau_r=tauz_r, tau_r_valid=.TRUE.)
2485 : END IF
2486 : !
2487 : CALL qs_fxc_create(qs_env, rho, rhoz, rho0_atom_set, xc_section, .FALSE., &
2488 : v_xc, v_xc_tau, rho1_atom_set, &
2489 50 : dispersion_env=qs_env%dispersion_env)
2490 : !
2491 50 : DEALLOCATE (rhoz)
2492 :
2493 112 : DO ispin = 1, nspins
2494 62 : CALL pw_scale(v_xc(ispin), v_xc(ispin)%pw_grid%dvol)
2495 62 : CALL pw_axpy(zv_hartree_rspace, v_xc(ispin))
2496 : CALL integrate_v_rspace(qs_env=qs_env, &
2497 : v_rspace=v_xc(ispin), &
2498 : hmat=dbcsr_p_work(ispin), &
2499 62 : calculate_forces=.FALSE.)
2500 :
2501 112 : CALL auxbas_pw_pool%give_back_pw(v_xc(ispin))
2502 : END DO
2503 50 : CALL auxbas_pw_pool%give_back_pw(rhoz_tot_gspace)
2504 50 : CALL auxbas_pw_pool%give_back_pw(zv_hartree_rspace)
2505 50 : CALL auxbas_pw_pool%give_back_pw(zv_hartree_gspace)
2506 50 : DEALLOCATE (v_xc)
2507 :
2508 50 : IF (do_tau) THEN
2509 16 : DO ispin = 1, nspins
2510 8 : CALL pw_scale(v_xc_tau(ispin), v_xc_tau(ispin)%pw_grid%dvol)
2511 : CALL integrate_v_rspace(qs_env=qs_env, &
2512 : v_rspace=v_xc_tau(ispin), &
2513 : hmat=dbcsr_p_work(ispin), &
2514 : compute_tau=.TRUE., &
2515 8 : calculate_forces=.FALSE.)
2516 16 : CALL auxbas_pw_pool%give_back_pw(v_xc_tau(ispin))
2517 : END DO
2518 8 : DEALLOCATE (v_xc_tau)
2519 : END IF
2520 :
2521 : !Auxiliary xc kernel (admm)
2522 50 : IF (dft_control%do_admm) THEN
2523 16 : CALL get_qs_env(qs_env, admm_env=admm_env)
2524 : CALL get_admm_env(admm_env, matrix_s_aux_fit=matrix_s_aux, &
2525 16 : task_list_aux_fit=task_list_aux_fit, rho_aux_fit=rho_aux_fit)
2526 :
2527 16 : CALL qs_rho_get(rho_aux_fit, tau_r_valid=do_tau_admm)
2528 :
2529 16 : CALL dbcsr_allocate_matrix_set(work_admm, nspins)
2530 16 : CALL dbcsr_allocate_matrix_set(ker_tau_admm, nspins)
2531 36 : DO ispin = 1, nspins
2532 20 : ALLOCATE (work_admm(ispin)%matrix, ker_tau_admm(ispin)%matrix)
2533 20 : CALL dbcsr_create(work_admm(ispin)%matrix, template=matrix_s_aux(1)%matrix)
2534 20 : CALL dbcsr_copy(work_admm(ispin)%matrix, matrix_s_aux(1)%matrix)
2535 20 : CALL dbcsr_set(work_admm(ispin)%matrix, 0.0_dp)
2536 20 : CALL dbcsr_create(ker_tau_admm(ispin)%matrix, template=matrix_s_aux(1)%matrix)
2537 20 : CALL dbcsr_copy(ker_tau_admm(ispin)%matrix, matrix_s_aux(1)%matrix)
2538 36 : CALL dbcsr_set(ker_tau_admm(ispin)%matrix, 0.0_dp)
2539 : END DO
2540 :
2541 : !get the density in the auxuliary density
2542 16 : CPASSERT(ASSOCIATED(admm_env%work_orb_orb))
2543 16 : CPASSERT(ASSOCIATED(admm_env%work_aux_orb))
2544 16 : CPASSERT(ASSOCIATED(admm_env%work_aux_aux))
2545 16 : nao = admm_env%nao_orb
2546 16 : nao_aux = admm_env%nao_aux_fit
2547 36 : DO ispin = 1, nspins
2548 20 : CALL copy_dbcsr_to_fm(force_data%sum_YP_tau(ispin)%matrix, admm_env%work_orb_orb)
2549 : CALL parallel_gemm('N', 'N', nao_aux, nao, nao, 1.0_dp, admm_env%A, admm_env%work_orb_orb, &
2550 20 : 0.0_dp, admm_env%work_aux_orb)
2551 : CALL parallel_gemm('N', 'T', nao_aux, nao_aux, nao, 1.0_dp, admm_env%work_aux_orb, admm_env%A, &
2552 20 : 0.0_dp, admm_env%work_aux_aux)
2553 36 : CALL copy_fm_to_dbcsr(admm_env%work_aux_aux, ker_tau_admm(ispin)%matrix, keep_sparsity=.TRUE.)
2554 : END DO
2555 :
2556 16 : IF (.NOT. qs_env%admm_env%aux_exch_func == do_admm_aux_exch_func_none) THEN
2557 36 : DO ispin = 1, nspins
2558 20 : CALL pw_zero(rhoz_r(ispin))
2559 20 : CALL pw_zero(rhoz_g(ispin))
2560 : CALL calculate_rho_elec(ks_env=qs_env%ks_env, matrix_p=ker_tau_admm(ispin)%matrix, &
2561 : rho=rhoz_r(ispin), rho_gspace=rhoz_g(ispin), &
2562 36 : basis_type="AUX_FIT", task_list_external=task_list_aux_fit)
2563 : END DO
2564 :
2565 16 : IF (do_tau_admm) THEN
2566 : BLOCK
2567 : TYPE(pw_c1d_gs_type) :: tauz_g
2568 0 : CALL auxbas_pw_pool%create_pw(tauz_g)
2569 0 : DO ispin = 1, nspins
2570 0 : CALL pw_zero(tauz_r(ispin))
2571 : CALL calculate_rho_elec(ks_env=qs_env%ks_env, matrix_p=ker_tau_admm(ispin)%matrix, &
2572 : rho=tauz_r(ispin), rho_gspace=tauz_g, &
2573 : basis_type="AUX_FIT", task_list_external=task_list_aux_fit, &
2574 0 : compute_tau=.TRUE.)
2575 : END DO
2576 0 : CALL auxbas_pw_pool%give_back_pw(tauz_g)
2577 : END BLOCK
2578 : END IF
2579 :
2580 16 : xc_section => admm_env%xc_section_aux
2581 16 : ALLOCATE (rhoz)
2582 16 : CALL qs_rho_create(rhoz)
2583 16 : IF (ASSOCIATED(rhoz_r)) THEN
2584 16 : CALL qs_rho_set(rhoz, rho_r=rhoz_r, rho_r_valid=.TRUE.)
2585 : END IF
2586 16 : IF (ASSOCIATED(rhoz_g)) THEN
2587 16 : CALL qs_rho_set(rhoz, rho_g=rhoz_g, rho_g_valid=.TRUE.)
2588 : END IF
2589 16 : IF (ASSOCIATED(tauz_r)) THEN
2590 0 : CALL qs_rho_set(rhoz, tau_r=tauz_r, tau_r_valid=.TRUE.)
2591 : END IF
2592 : !
2593 : CALL qs_fxc_create(qs_env, rho_aux_fit, rhoz, rho0_atom_set, xc_section, .FALSE., &
2594 16 : v_xc, v_xc_tau, rho1_atom_set)
2595 : !
2596 16 : DEALLOCATE (rhoz)
2597 :
2598 36 : DO ispin = 1, nspins
2599 20 : CALL pw_scale(v_xc(ispin), v_xc(ispin)%pw_grid%dvol)
2600 : CALL integrate_v_rspace(qs_env=qs_env, &
2601 : v_rspace=v_xc(ispin), &
2602 : hmat=work_admm(ispin), &
2603 : calculate_forces=.FALSE., &
2604 : basis_type="AUX_FIT", &
2605 20 : task_list_external=task_list_aux_fit)
2606 36 : CALL auxbas_pw_pool%give_back_pw(v_xc(ispin))
2607 : END DO
2608 16 : DEALLOCATE (v_xc)
2609 :
2610 16 : IF (do_tau_admm) THEN
2611 0 : DO ispin = 1, nspins
2612 0 : CALL pw_scale(v_xc_tau(ispin), v_xc_tau(ispin)%pw_grid%dvol)
2613 : CALL integrate_v_rspace(qs_env=qs_env, &
2614 : v_rspace=v_xc_tau(ispin), &
2615 : hmat=work_admm(ispin), &
2616 : calculate_forces=.FALSE., &
2617 : basis_type="AUX_FIT", &
2618 : task_list_external=task_list_aux_fit, &
2619 0 : compute_tau=.TRUE.)
2620 0 : CALL auxbas_pw_pool%give_back_pw(v_xc_tau(ispin))
2621 : END DO
2622 0 : DEALLOCATE (v_xc_tau)
2623 : END IF
2624 : END IF !admm
2625 : END IF
2626 :
2627 112 : DO ispin = 1, nspins
2628 62 : CALL auxbas_pw_pool%give_back_pw(rhoz_r(ispin))
2629 112 : CALL auxbas_pw_pool%give_back_pw(rhoz_g(ispin))
2630 : END DO
2631 50 : DEALLOCATE (rhoz_r, rhoz_g)
2632 :
2633 50 : IF (do_tau) THEN
2634 16 : DO ispin = 1, nspins
2635 16 : CALL auxbas_pw_pool%give_back_pw(tauz_r(ispin))
2636 : END DO
2637 8 : DEALLOCATE (tauz_r)
2638 : END IF
2639 :
2640 : !HFX kernel
2641 50 : hfx_section => section_vals_get_subs_vals(qs_env%input, "DFT%XC%HF")
2642 50 : CALL section_vals_get(hfx_section, explicit=do_hfx)
2643 50 : IF (do_hfx) THEN
2644 32 : IF (dft_control%do_admm) THEN
2645 16 : CALL tddft_hfx_matrix(work_admm, ker_tau_admm, qs_env, .FALSE., .FALSE.)
2646 :
2647 : !Going back to primary basis
2648 16 : CALL dbcsr_create(dbcsr_work, template=dbcsr_p_work(1)%matrix)
2649 16 : CALL dbcsr_copy(dbcsr_work, dbcsr_p_work(1)%matrix)
2650 16 : CALL dbcsr_set(dbcsr_work, 0.0_dp)
2651 36 : DO ispin = 1, nspins
2652 20 : CALL copy_dbcsr_to_fm(work_admm(ispin)%matrix, admm_env%work_aux_aux)
2653 : CALL parallel_gemm('N', 'N', nao_aux, nao, nao_aux, 1.0_dp, admm_env%work_aux_aux, admm_env%A, &
2654 20 : 0.0_dp, admm_env%work_aux_orb)
2655 : CALL parallel_gemm('T', 'N', nao, nao, nao_aux, 1.0_dp, admm_env%A, admm_env%work_aux_orb, &
2656 20 : 0.0_dp, admm_env%work_orb_orb)
2657 20 : CALL copy_fm_to_dbcsr(admm_env%work_orb_orb, dbcsr_work, keep_sparsity=.TRUE.)
2658 36 : CALL dbcsr_add(dbcsr_p_work(ispin)%matrix, dbcsr_work, 1.0_dp, 1.0_dp)
2659 : END DO
2660 16 : CALL dbcsr_release(dbcsr_work)
2661 16 : CALL dbcsr_deallocate_matrix_set(ker_tau_admm)
2662 : ELSE
2663 16 : CALL tddft_hfx_matrix(dbcsr_p_work, force_data%sum_YP_tau, qs_env, .FALSE., .FALSE.)
2664 : END IF
2665 : END IF
2666 :
2667 112 : DO ispin = 1, nspins
2668 112 : CALL dbcsr_add(force_data%sum_O_tau(ispin)%matrix, dbcsr_p_work(ispin)%matrix, 1.0_dp, 1.0_dp)
2669 : END DO
2670 :
2671 50 : CALL dbcsr_deallocate_matrix_set(dbcsr_p_work)
2672 50 : CALL dbcsr_deallocate_matrix_set(work_admm)
2673 :
2674 50 : CALL timestop(handle)
2675 :
2676 250 : END SUBROUTINE prepare_for_response
2677 :
2678 : ! **************************************************************************************************
2679 : !> \brief Calculate the force and virial due to the (P|Q) GPW integral derivatives
2680 : !> \param G_PQ ...
2681 : !> \param force ...
2682 : !> \param h_stress ...
2683 : !> \param use_virial ...
2684 : !> \param mp2_env ...
2685 : !> \param qs_env ...
2686 : ! **************************************************************************************************
2687 12 : SUBROUTINE get_2c_gpw_forces(G_PQ, force, h_stress, use_virial, mp2_env, qs_env)
2688 :
2689 : TYPE(dbcsr_type), INTENT(INOUT) :: G_PQ
2690 : TYPE(qs_force_type), DIMENSION(:), POINTER :: force
2691 : REAL(dp), DIMENSION(3, 3), INTENT(INOUT) :: h_stress
2692 : LOGICAL, INTENT(IN) :: use_virial
2693 : TYPE(mp2_type), INTENT(INOUT) :: mp2_env
2694 : TYPE(qs_environment_type), POINTER :: qs_env
2695 :
2696 : CHARACTER(len=*), PARAMETER :: routineN = 'get_2c_gpw_forces'
2697 :
2698 : INTEGER :: atom_a, color, handle, i, i_RI, i_xyz, iatom, igrid_level, ikind, ipgf, iset, j, &
2699 : j_RI, jatom, lb_RI, n_RI, natom, ncoa, ncoms, nkind, nproc, nseta, o1, offset, pdims(2), &
2700 : sgfa, ub_RI
2701 24 : INTEGER, ALLOCATABLE, DIMENSION(:) :: atom_of_kind, iproc_map, kind_of, &
2702 12 : sizes_RI
2703 24 : INTEGER, DIMENSION(:), POINTER :: col_dist, la_max, la_min, npgfa, nsgfa, &
2704 12 : row_dist
2705 12 : INTEGER, DIMENSION(:, :), POINTER :: first_sgfa, pgrid
2706 : LOGICAL :: found, one_proc_group
2707 : REAL(dp) :: cutoff_old, radius, relative_cutoff_old
2708 12 : REAL(dp), ALLOCATABLE, DIMENSION(:) :: e_cutoff_old, wf_vector
2709 : REAL(dp), DIMENSION(3) :: force_a, force_b, ra
2710 : REAL(dp), DIMENSION(3, 3) :: my_virial_a, my_virial_b
2711 12 : REAL(KIND=dp), DIMENSION(:, :), POINTER :: h_tmp, I_ab, pab, pblock, sphi_a, zeta
2712 12 : TYPE(atomic_kind_type), DIMENSION(:), POINTER :: atomic_kind_set
2713 : TYPE(cell_type), POINTER :: cell
2714 : TYPE(dbcsr_distribution_type) :: dbcsr_dist
2715 : TYPE(dbcsr_type) :: tmp_G_PQ
2716 : TYPE(dft_control_type), POINTER :: dft_control
2717 : TYPE(gto_basis_set_p_type), ALLOCATABLE, &
2718 12 : DIMENSION(:), TARGET :: basis_set_ri_aux
2719 : TYPE(gto_basis_set_type), POINTER :: basis_set_a
2720 12 : TYPE(mo_set_type), DIMENSION(:), POINTER :: mos
2721 : TYPE(mp_para_env_type), POINTER :: para_env, para_env_ext
2722 : TYPE(neighbor_list_set_p_type), DIMENSION(:), &
2723 12 : POINTER :: sab_orb
2724 12 : TYPE(particle_type), DIMENSION(:), POINTER :: particle_set
2725 48 : TYPE(pw_c1d_gs_type) :: dvg(3), pot_g, rho_g, rho_g_copy
2726 : TYPE(pw_env_type), POINTER :: pw_env_ext
2727 : TYPE(pw_poisson_type), POINTER :: poisson_env
2728 : TYPE(pw_pool_type), POINTER :: auxbas_pw_pool
2729 : TYPE(pw_r3d_rs_type) :: psi_L, rho_r
2730 12 : TYPE(qs_kind_type), DIMENSION(:), POINTER :: qs_kind_set
2731 12 : TYPE(realspace_grid_type), DIMENSION(:), POINTER :: rs_v
2732 : TYPE(task_list_type), POINTER :: task_list_ext
2733 :
2734 12 : NULLIFY (sab_orb, task_list_ext, particle_set, qs_kind_set, dft_control, pw_env_ext, auxbas_pw_pool, &
2735 12 : poisson_env, atomic_kind_set, para_env, cell, rs_v, mos, basis_set_a)
2736 :
2737 12 : CALL timeset(routineN, handle)
2738 :
2739 : CALL get_qs_env(qs_env, dft_control=dft_control, para_env=para_env, sab_orb=sab_orb, &
2740 : natom=natom, nkind=nkind, qs_kind_set=qs_kind_set, particle_set=particle_set, &
2741 12 : mos=mos, cell=cell, atomic_kind_set=atomic_kind_set)
2742 :
2743 : !The idea is to use GPW to compute the integrals and derivatives. Because the potential needs
2744 : !to be calculated for each phi_j (column) of all AO pairs, and because that is expensive, we want
2745 : !to minimize the amount of time we do that. Therefore, we work with a special distribution, where
2746 : !each column of the resulting DBCSR matrix is mapped to a sub-communicator.
2747 :
2748 : !Try to get the optimal pdims (we want a grid that is flat: many cols, few rows)
2749 12 : IF (para_env%num_pe <= natom) THEN
2750 : pdims(1) = 1
2751 : pdims(2) = para_env%num_pe
2752 : ELSE
2753 0 : DO i = natom, 1, -1
2754 0 : IF (MODULO(para_env%num_pe, i) == 0) THEN
2755 0 : pdims(1) = para_env%num_pe/i
2756 0 : pdims(2) = i
2757 0 : EXIT
2758 : END IF
2759 : END DO
2760 : END IF
2761 :
2762 48 : ALLOCATE (row_dist(natom), col_dist(natom))
2763 48 : DO iatom = 1, natom
2764 48 : row_dist(iatom) = MODULO(iatom, pdims(1))
2765 : END DO
2766 48 : DO jatom = 1, natom
2767 48 : col_dist(jatom) = MODULO(jatom, pdims(2))
2768 : END DO
2769 :
2770 48 : ALLOCATE (pgrid(0:pdims(1) - 1, 0:pdims(2) - 1))
2771 12 : nproc = 0
2772 24 : DO i = 0, pdims(1) - 1
2773 48 : DO j = 0, pdims(2) - 1
2774 24 : pgrid(i, j) = nproc
2775 36 : nproc = nproc + 1
2776 : END DO
2777 : END DO
2778 :
2779 12 : CALL dbcsr_distribution_new(dbcsr_dist, group=para_env%get_handle(), pgrid=pgrid, row_dist=row_dist, col_dist=col_dist)
2780 :
2781 : !The temporary DBCSR integrals and derivatives matrices in this flat distribution
2782 12 : CALL dbcsr_create(tmp_G_PQ, template=G_PQ, matrix_type=dbcsr_type_no_symmetry, dist=dbcsr_dist)
2783 12 : CALL dbcsr_complete_redistribute(G_PQ, tmp_G_PQ)
2784 :
2785 84 : ALLOCATE (basis_set_ri_aux(nkind), sizes_RI(natom))
2786 12 : CALL basis_set_list_setup(basis_set_ri_aux, "RI_AUX", qs_kind_set)
2787 12 : CALL get_particle_set(particle_set, qs_kind_set, nsgf=sizes_RI, basis=basis_set_ri_aux)
2788 48 : n_RI = SUM(sizes_RI)
2789 :
2790 12 : one_proc_group = mp2_env%mp2_num_proc == 1
2791 12 : ALLOCATE (para_env_ext)
2792 12 : IF (one_proc_group) THEN
2793 : !one subgroup per proc
2794 4 : CALL para_env_ext%from_split(para_env, para_env%mepos)
2795 : ELSE
2796 : !Split the communicator accross the columns of the matrix
2797 8 : ncoms = MIN(pdims(2), para_env%num_pe/mp2_env%mp2_num_proc)
2798 16 : DO i = 0, pdims(1) - 1
2799 32 : DO j = 0, pdims(2) - 1
2800 24 : IF (pgrid(i, j) == para_env%mepos) color = MODULO(j + 1, ncoms)
2801 : END DO
2802 : END DO
2803 8 : CALL para_env_ext%from_split(para_env, color)
2804 : END IF
2805 :
2806 : !sab_orb and task_list_ext are essentially dummies
2807 : CALL prepare_gpw(qs_env, dft_control, e_cutoff_old, cutoff_old, relative_cutoff_old, para_env_ext, pw_env_ext, &
2808 12 : auxbas_pw_pool, poisson_env, task_list_ext, rho_r, rho_g, pot_g, psi_L, sab_orb)
2809 :
2810 12 : IF (use_virial) THEN
2811 4 : CALL auxbas_pw_pool%create_pw(rho_g_copy)
2812 16 : DO i_xyz = 1, 3
2813 16 : CALL auxbas_pw_pool%create_pw(dvg(i_xyz))
2814 : END DO
2815 : END IF
2816 :
2817 36 : ALLOCATE (wf_vector(n_RI))
2818 :
2819 12 : CALL get_atomic_kind_set(atomic_kind_set, kind_of=kind_of, atom_of_kind=atom_of_kind)
2820 :
2821 36 : ALLOCATE (iproc_map(natom))
2822 :
2823 : !Loop over the atomic blocks
2824 48 : DO jatom = 1, natom
2825 :
2826 : !Only calculate if on the correct sub-communicator/proc
2827 36 : IF (one_proc_group) THEN
2828 12 : iproc_map = 0
2829 48 : DO iatom = 1, natom
2830 48 : IF (pgrid(row_dist(iatom), col_dist(jatom)) == para_env%mepos) iproc_map(iatom) = 1
2831 : END DO
2832 30 : IF (.NOT. ANY(iproc_map == 1)) CYCLE
2833 : ELSE
2834 24 : IF (.NOT. MODULO(col_dist(jatom) + 1, ncoms) == color) CYCLE
2835 : END IF
2836 :
2837 60 : lb_RI = SUM(sizes_RI(1:jatom - 1))
2838 30 : ub_RI = lb_RI + sizes_RI(jatom)
2839 872 : DO j_RI = lb_RI + 1, ub_RI
2840 :
2841 830 : wf_vector = 0.0_dp
2842 830 : wf_vector(j_RI) = 1.0_dp
2843 :
2844 : CALL collocate_function(wf_vector, psi_L, rho_g, atomic_kind_set, qs_kind_set, cell, &
2845 : particle_set, pw_env_ext, dft_control%qs_control%eps_rho_rspace, &
2846 830 : basis_type="RI_AUX")
2847 :
2848 830 : IF (use_virial) THEN
2849 166 : CALL calc_potential_gpw(rho_r, rho_g, poisson_env, pot_g, mp2_env%potential_parameter, dvg)
2850 :
2851 166 : wf_vector = 0.0_dp
2852 664 : DO iatom = 1, natom
2853 : !only compute if i,j atom pair on correct proc
2854 498 : IF (one_proc_group) THEN
2855 498 : IF (.NOT. iproc_map(iatom) == 1) CYCLE
2856 : END IF
2857 :
2858 498 : CALL dbcsr_get_block_p(tmp_G_PQ, iatom, jatom, pblock, found)
2859 498 : IF (.NOT. found) CYCLE
2860 :
2861 996 : i_RI = SUM(sizes_RI(1:iatom - 1))
2862 14940 : wf_vector(i_RI + 1:i_RI + sizes_RI(iatom)) = pblock(:, j_RI - lb_RI)
2863 : END DO
2864 :
2865 166 : CALL pw_copy(rho_g, rho_g_copy)
2866 : CALL collocate_function(wf_vector, psi_L, rho_g, atomic_kind_set, qs_kind_set, cell, &
2867 : particle_set, pw_env_ext, dft_control%qs_control%eps_rho_rspace, &
2868 166 : basis_type="RI_AUX")
2869 :
2870 : CALL calc_potential_gpw(psi_L, rho_g, poisson_env, pot_g, mp2_env%potential_parameter, &
2871 166 : no_transfer=.TRUE.)
2872 : CALL virial_gpw_potential(rho_g_copy, pot_g, rho_g, dvg, h_stress, &
2873 166 : mp2_env%potential_parameter, para_env_ext)
2874 : ELSE
2875 664 : CALL calc_potential_gpw(rho_r, rho_g, poisson_env, pot_g, mp2_env%potential_parameter)
2876 : END IF
2877 :
2878 830 : NULLIFY (rs_v)
2879 830 : CALL pw_env_get(pw_env_ext, rs_grids=rs_v)
2880 830 : CALL potential_pw2rs(rs_v, rho_r, pw_env_ext)
2881 :
2882 3356 : DO iatom = 1, natom
2883 :
2884 : !only compute if i,j atom pair on correct proc
2885 2490 : IF (one_proc_group) THEN
2886 498 : IF (.NOT. iproc_map(iatom) == 1) CYCLE
2887 : END IF
2888 :
2889 2490 : force_a(:) = 0.0_dp
2890 2490 : force_b(:) = 0.0_dp
2891 2490 : IF (use_virial) THEN
2892 498 : my_virial_a = 0.0_dp
2893 498 : my_virial_b = 0.0_dp
2894 : END IF
2895 :
2896 2490 : ikind = kind_of(iatom)
2897 2490 : atom_a = atom_of_kind(iatom)
2898 :
2899 2490 : basis_set_a => basis_set_ri_aux(ikind)%gto_basis_set
2900 2490 : first_sgfa => basis_set_a%first_sgf
2901 2490 : la_max => basis_set_a%lmax
2902 2490 : la_min => basis_set_a%lmin
2903 2490 : nseta = basis_set_a%nset
2904 2490 : nsgfa => basis_set_a%nsgf_set
2905 2490 : sphi_a => basis_set_a%sphi
2906 2490 : zeta => basis_set_a%zet
2907 2490 : npgfa => basis_set_a%npgf
2908 :
2909 2490 : ra(:) = pbc(particle_set(iatom)%r, cell)
2910 :
2911 2490 : CALL dbcsr_get_block_p(tmp_G_PQ, iatom, jatom, pblock, found)
2912 2490 : IF (.NOT. found) CYCLE
2913 :
2914 : offset = 0
2915 15936 : DO iset = 1, nseta
2916 14442 : ncoa = npgfa(iset)*ncoset(la_max(iset))
2917 14442 : sgfa = first_sgfa(1, iset)
2918 :
2919 131472 : ALLOCATE (h_tmp(ncoa, 1)); h_tmp = 0.0_dp
2920 99102 : ALLOCATE (I_ab(nsgfa(iset), 1)); I_ab = 0.0_dp
2921 117030 : ALLOCATE (pab(ncoa, 1)); pab = 0.0_dp
2922 :
2923 97110 : I_ab(1:nsgfa(iset), 1) = 2.0_dp*pblock(offset + 1:offset + nsgfa(iset), j_RI - lb_RI)
2924 : CALL dgemm("N", "N", ncoa, 1, nsgfa(iset), 1.0_dp, sphi_a(1, sgfa), SIZE(sphi_a, 1), &
2925 14442 : I_ab(1, 1), nsgfa(iset), 0.0_dp, pab(1, 1), ncoa)
2926 :
2927 28884 : igrid_level = gaussian_gridlevel(pw_env_ext%gridlevel_info, MINVAL(zeta(:, iset)))
2928 :
2929 : ! The last three parameters are used to check whether a given function is within the own range.
2930 : ! Here, it is always the case, so let's enforce it because mod(0, 1)==0
2931 14442 : IF (map_gaussian_here(rs_v(igrid_level), cell%h_inv, ra, 0, 1, 0)) THEN
2932 28884 : DO ipgf = 1, npgfa(iset)
2933 14442 : o1 = (ipgf - 1)*ncoset(la_max(iset))
2934 14442 : igrid_level = gaussian_gridlevel(pw_env_ext%gridlevel_info, zeta(ipgf, iset))
2935 :
2936 : radius = exp_radius_very_extended(la_min=la_min(iset), la_max=la_max(iset), &
2937 : lb_min=0, lb_max=0, ra=ra, rb=ra, rp=ra, &
2938 : zetp=zeta(ipgf, iset), &
2939 : eps=dft_control%qs_control%eps_gvg_rspace, &
2940 14442 : prefactor=1.0_dp, cutoff=1.0_dp)
2941 :
2942 : CALL integrate_pgf_product( &
2943 : la_max=la_max(iset), zeta=zeta(ipgf, iset), la_min=la_min(iset), &
2944 : lb_max=0, zetb=0.0_dp, lb_min=0, &
2945 : ra=ra, rab=[0.0_dp, 0.0_dp, 0.0_dp], &
2946 : rsgrid=rs_v(igrid_level), &
2947 : hab=h_tmp, pab=pab, &
2948 : o1=o1, &
2949 : o2=0, &
2950 : radius=radius, &
2951 : calculate_forces=.TRUE., &
2952 : force_a=force_a, force_b=force_b, &
2953 28884 : use_virial=use_virial, my_virial_a=my_virial_a, my_virial_b=my_virial_b)
2954 :
2955 : END DO
2956 :
2957 : END IF
2958 :
2959 14442 : offset = offset + nsgfa(iset)
2960 15936 : DEALLOCATE (pab, h_tmp, I_ab)
2961 : END DO !iset
2962 :
2963 5976 : force(ikind)%mp2_non_sep(:, atom_a) = force(ikind)%mp2_non_sep(:, atom_a) + force_a + force_b
2964 10790 : IF (use_virial) h_stress = h_stress + my_virial_a + my_virial_b
2965 :
2966 : END DO !iatom
2967 : END DO !j_RI
2968 : END DO !jatom
2969 :
2970 12 : IF (use_virial) THEN
2971 4 : CALL auxbas_pw_pool%give_back_pw(rho_g_copy)
2972 16 : DO i_xyz = 1, 3
2973 16 : CALL auxbas_pw_pool%give_back_pw(dvg(i_xyz))
2974 : END DO
2975 : END IF
2976 :
2977 : CALL cleanup_gpw(qs_env, e_cutoff_old, cutoff_old, relative_cutoff_old, para_env_ext, pw_env_ext, &
2978 12 : task_list_ext, auxbas_pw_pool, rho_r, rho_g, pot_g, psi_L)
2979 :
2980 12 : CALL dbcsr_release(tmp_G_PQ)
2981 12 : CALL dbcsr_distribution_release(dbcsr_dist)
2982 12 : DEALLOCATE (col_dist, row_dist, pgrid)
2983 :
2984 12 : CALL mp_para_env_release(para_env_ext)
2985 :
2986 12 : CALL timestop(handle)
2987 :
2988 36 : END SUBROUTINE get_2c_gpw_forces
2989 :
2990 : ! **************************************************************************************************
2991 : !> \brief Calculate the forces due to the (P|Q) MME integral derivatives
2992 : !> \param G_PQ ...
2993 : !> \param force ...
2994 : !> \param mp2_env ...
2995 : !> \param qs_env ...
2996 : ! **************************************************************************************************
2997 16 : SUBROUTINE get_2c_mme_forces(G_PQ, force, mp2_env, qs_env)
2998 :
2999 : TYPE(dbcsr_type), INTENT(INOUT) :: G_PQ
3000 : TYPE(qs_force_type), DIMENSION(:), POINTER :: force
3001 : TYPE(mp2_type), INTENT(INOUT) :: mp2_env
3002 : TYPE(qs_environment_type), POINTER :: qs_env
3003 :
3004 : CHARACTER(len=*), PARAMETER :: routineN = 'get_2c_mme_forces'
3005 :
3006 : INTEGER :: atom_a, atom_b, G_count, handle, i_xyz, iatom, ikind, iset, jatom, jkind, jset, &
3007 : natom, nkind, nseta, nsetb, offset_hab_a, offset_hab_b, R_count, sgfa, sgfb
3008 16 : INTEGER, ALLOCATABLE, DIMENSION(:) :: atom_of_kind, kind_of
3009 16 : INTEGER, DIMENSION(:), POINTER :: la_max, la_min, lb_max, lb_min, npgfa, &
3010 16 : npgfb, nsgfa, nsgfb
3011 16 : INTEGER, DIMENSION(:, :), POINTER :: first_sgfa, first_sgfb
3012 : LOGICAL :: found
3013 : REAL(dp) :: new_force, pref
3014 16 : REAL(dp), ALLOCATABLE, DIMENSION(:, :) :: hab
3015 16 : REAL(dp), ALLOCATABLE, DIMENSION(:, :, :) :: hdab
3016 16 : REAL(dp), DIMENSION(:, :), POINTER :: pblock
3017 : REAL(KIND=dp), DIMENSION(3) :: ra, rb
3018 16 : REAL(KIND=dp), DIMENSION(:, :), POINTER :: sphi_a, sphi_b, zeta, zetb
3019 16 : TYPE(atomic_kind_type), DIMENSION(:), POINTER :: atomic_kind_set
3020 : TYPE(cell_type), POINTER :: cell
3021 : TYPE(dbcsr_iterator_type) :: iter
3022 : TYPE(gto_basis_set_p_type), ALLOCATABLE, &
3023 16 : DIMENSION(:), TARGET :: basis_set_ri_aux
3024 : TYPE(gto_basis_set_type), POINTER :: basis_set_a, basis_set_b
3025 : TYPE(mp_para_env_type), POINTER :: para_env
3026 16 : TYPE(particle_type), DIMENSION(:), POINTER :: particle_set
3027 16 : TYPE(qs_kind_type), DIMENSION(:), POINTER :: qs_kind_set
3028 :
3029 16 : NULLIFY (qs_kind_set, basis_set_a, basis_set_b, pblock, particle_set, &
3030 16 : cell, la_max, la_min, lb_min, npgfa, lb_max, npgfb, nsgfa, &
3031 16 : nsgfb, first_sgfa, first_sgfb, sphi_a, sphi_b, zeta, zetb, &
3032 16 : atomic_kind_set, para_env)
3033 :
3034 16 : CALL timeset(routineN, handle)
3035 :
3036 : CALL get_qs_env(qs_env, qs_kind_set=qs_kind_set, nkind=nkind, particle_set=particle_set, &
3037 16 : cell=cell, atomic_kind_set=atomic_kind_set, natom=natom, para_env=para_env)
3038 :
3039 16 : CALL get_atomic_kind_set(atomic_kind_set, kind_of=kind_of, atom_of_kind=atom_of_kind)
3040 :
3041 80 : ALLOCATE (basis_set_ri_aux(nkind))
3042 16 : CALL basis_set_list_setup(basis_set_ri_aux, "RI_AUX", qs_kind_set)
3043 :
3044 16 : G_count = 0; R_count = 0
3045 :
3046 16 : CALL dbcsr_iterator_start(iter, G_PQ)
3047 116 : DO WHILE (dbcsr_iterator_blocks_left(iter))
3048 :
3049 100 : CALL dbcsr_iterator_next_block(iter, row=iatom, column=jatom)
3050 100 : CALL dbcsr_get_block_p(G_PQ, iatom, jatom, pblock, found)
3051 100 : IF (.NOT. found) CYCLE
3052 100 : IF (iatom > jatom) CYCLE
3053 64 : pref = 2.0_dp
3054 64 : IF (iatom == jatom) pref = 1.0_dp
3055 :
3056 64 : ikind = kind_of(iatom)
3057 64 : jkind = kind_of(jatom)
3058 :
3059 64 : atom_a = atom_of_kind(iatom)
3060 64 : atom_b = atom_of_kind(jatom)
3061 :
3062 64 : basis_set_a => basis_set_ri_aux(ikind)%gto_basis_set
3063 64 : first_sgfa => basis_set_a%first_sgf
3064 64 : la_max => basis_set_a%lmax
3065 64 : la_min => basis_set_a%lmin
3066 64 : nseta = basis_set_a%nset
3067 64 : nsgfa => basis_set_a%nsgf_set
3068 64 : sphi_a => basis_set_a%sphi
3069 64 : zeta => basis_set_a%zet
3070 64 : npgfa => basis_set_a%npgf
3071 :
3072 64 : basis_set_b => basis_set_ri_aux(jkind)%gto_basis_set
3073 64 : first_sgfb => basis_set_b%first_sgf
3074 64 : lb_max => basis_set_b%lmax
3075 64 : lb_min => basis_set_b%lmin
3076 64 : nsetb = basis_set_b%nset
3077 64 : nsgfb => basis_set_b%nsgf_set
3078 64 : sphi_b => basis_set_b%sphi
3079 64 : zetb => basis_set_b%zet
3080 64 : npgfb => basis_set_b%npgf
3081 :
3082 64 : ra(:) = pbc(particle_set(iatom)%r, cell)
3083 64 : rb(:) = pbc(particle_set(jatom)%r, cell)
3084 :
3085 256 : ALLOCATE (hab(basis_set_a%nsgf, basis_set_b%nsgf))
3086 256 : ALLOCATE (hdab(3, basis_set_a%nsgf, basis_set_b%nsgf))
3087 64 : hab(:, :) = 0.0_dp
3088 64 : hdab(:, :, :) = 0.0_dp
3089 :
3090 64 : offset_hab_a = 0
3091 756 : DO iset = 1, nseta
3092 692 : sgfa = first_sgfa(1, iset)
3093 :
3094 692 : offset_hab_b = 0
3095 6340 : DO jset = 1, nsetb
3096 5648 : sgfb = first_sgfb(1, jset)
3097 :
3098 : CALL integrate_set_2c(mp2_env%eri_mme_param%par, mp2_env%potential_parameter, la_min(iset), &
3099 : la_max(iset), lb_min(jset), lb_max(jset), npgfa(iset), npgfb(jset), &
3100 : zeta(:, iset), zetb(:, jset), ra, rb, hab, nsgfa(iset), nsgfb(jset), &
3101 : offset_hab_a, offset_hab_b, 0, 0, sphi_a, sphi_b, sgfa, sgfb, &
3102 : nsgfa(iset), nsgfb(jset), do_eri_mme, hdab=hdab, &
3103 5648 : G_count=G_count, R_count=R_count)
3104 :
3105 6340 : offset_hab_b = offset_hab_b + nsgfb(jset)
3106 : END DO
3107 756 : offset_hab_a = offset_hab_a + nsgfa(iset)
3108 : END DO
3109 :
3110 256 : DO i_xyz = 1, 3
3111 143832 : new_force = pref*SUM(pblock(:, :)*hdab(i_xyz, :, :))
3112 192 : force(ikind)%mp2_non_sep(i_xyz, atom_a) = force(ikind)%mp2_non_sep(i_xyz, atom_a) + new_force
3113 256 : force(jkind)%mp2_non_sep(i_xyz, atom_b) = force(jkind)%mp2_non_sep(i_xyz, atom_b) - new_force
3114 : END DO
3115 :
3116 216 : DEALLOCATE (hab, hdab)
3117 : END DO
3118 16 : CALL dbcsr_iterator_stop(iter)
3119 :
3120 16 : CALL cp_eri_mme_update_local_counts(mp2_env%eri_mme_param, para_env, G_count_2c=G_count, R_count_2c=R_count)
3121 :
3122 16 : CALL timestop(handle)
3123 :
3124 48 : END SUBROUTINE get_2c_mme_forces
3125 :
3126 : ! **************************************************************************************************
3127 : !> \brief This routines gather all the force updates due to the response density and the trace with F
3128 : !> Also update the forces due to the SCF density for XC and exact exchange
3129 : !> \param p_env the p_env coming from the response calculation
3130 : !> \param matrix_hz the matrix going into the RHS of the response equation
3131 : !> \param matrix_p_F the density matrix with which we evaluate Trace[P*F]
3132 : !> \param matrix_p_F_admm ...
3133 : !> \param qs_env ...
3134 : !> \note very much inspired from the response_force routine in response_solver.F, especially for virial
3135 : ! **************************************************************************************************
3136 50 : SUBROUTINE update_im_time_forces(p_env, matrix_hz, matrix_p_F, matrix_p_F_admm, qs_env)
3137 :
3138 : TYPE(qs_p_env_type), POINTER :: p_env
3139 : TYPE(dbcsr_p_type), DIMENSION(:), POINTER :: matrix_hz, matrix_p_F, matrix_p_F_admm
3140 : TYPE(qs_environment_type), POINTER :: qs_env
3141 :
3142 : CHARACTER(len=*), PARAMETER :: routineN = 'update_im_time_forces'
3143 :
3144 : INTEGER :: handle, i, idens, ispin, n_rep_hf, nao, &
3145 : nao_aux, nder, nimages, nocc, nspins
3146 : LOGICAL :: do_exx, do_hfx, do_tau, do_tau_admm, &
3147 : use_virial
3148 : REAL(dp) :: dummy_real1, dummy_real2, ehartree, exc, &
3149 : focc
3150 : REAL(dp), DIMENSION(3, 3) :: h_stress, pv_loc
3151 : TYPE(admm_type), POINTER :: admm_env
3152 50 : TYPE(atomic_kind_type), DIMENSION(:), POINTER :: atomic_kind_set
3153 50 : TYPE(dbcsr_p_type), DIMENSION(:), POINTER :: current_density, current_density_admm, &
3154 50 : current_mat_h, matrix_p_mp2, matrix_p_mp2_admm, matrix_s, matrix_s_aux_fit, matrix_w, &
3155 50 : rho_ao, rho_ao_aux, scrm, scrm_admm
3156 50 : TYPE(dbcsr_p_type), DIMENSION(:, :), POINTER :: dbcsr_work_h, dbcsr_work_p, mpa2
3157 : TYPE(dbcsr_type) :: dbcsr_work
3158 : TYPE(dft_control_type), POINTER :: dft_control
3159 50 : TYPE(hfx_type), DIMENSION(:, :), POINTER :: x_data
3160 50 : TYPE(mo_set_type), DIMENSION(:), POINTER :: mos
3161 : TYPE(mp_para_env_type), POINTER :: para_env
3162 : TYPE(neighbor_list_set_p_type), DIMENSION(:), &
3163 50 : POINTER :: sab_orb, sac_ae, sac_ppl, sap_ppnl
3164 50 : TYPE(particle_type), DIMENSION(:), POINTER :: particle_set
3165 : TYPE(pw_c1d_gs_type) :: rho_tot_gspace, rhoz_tot_gspace, &
3166 : zv_hartree_gspace
3167 50 : TYPE(pw_c1d_gs_type), DIMENSION(:), POINTER :: rhoz_g
3168 : TYPE(pw_env_type), POINTER :: pw_env
3169 : TYPE(pw_poisson_type), POINTER :: poisson_env
3170 : TYPE(pw_pool_type), POINTER :: auxbas_pw_pool
3171 : TYPE(pw_r3d_rs_type) :: vh_rspace, vhxc_rspace, zv_hartree_rspace
3172 50 : TYPE(pw_r3d_rs_type), DIMENSION(:), POINTER :: rhoz_r, tauz_r, v_xc, v_xc_tau, &
3173 50 : vadmm_rspace, vtau_rspace, vxc_rspace
3174 50 : TYPE(qs_force_type), DIMENSION(:), POINTER :: force
3175 50 : TYPE(qs_kind_type), DIMENSION(:), POINTER :: qs_kind_set
3176 : TYPE(qs_rho_type), POINTER :: rho, rho_aux_fit, rhoz
3177 50 : TYPE(rho_atom_type), DIMENSION(:), POINTER :: rho0_atom_set, rho1_atom_set
3178 : TYPE(section_vals_type), POINTER :: hfx_section, xc_section
3179 : TYPE(task_list_type), POINTER :: task_list_aux_fit
3180 : TYPE(virial_type), POINTER :: virial
3181 :
3182 50 : NULLIFY (scrm, rho, dft_control, matrix_p_mp2, matrix_s, &
3183 50 : matrix_p_mp2_admm, admm_env, sab_orb, dbcsr_work_p, &
3184 50 : dbcsr_work_h, sac_ae, sac_ppl, sap_ppnl, force, virial, &
3185 50 : qs_kind_set, atomic_kind_set, particle_set, pw_env, poisson_env, &
3186 50 : auxbas_pw_pool, task_list_aux_fit, matrix_s_aux_fit, scrm_admm, &
3187 50 : rho_aux_fit, rho_ao_aux, x_data, hfx_section, xc_section, &
3188 50 : para_env, rhoz_g, rhoz_r, tauz_r, v_xc, v_xc_tau, vxc_rspace, &
3189 50 : vtau_rspace, vadmm_rspace, rho_ao, matrix_w)
3190 50 : NULLIFY (rho0_atom_set, rho1_atom_set)
3191 :
3192 50 : CALL timeset(routineN, handle)
3193 :
3194 : CALL get_qs_env(qs_env, rho=rho, dft_control=dft_control, matrix_s=matrix_s, admm_env=admm_env, &
3195 : sab_orb=sab_orb, sac_ae=sac_ae, sac_ppl=sac_ppl, sap_ppnl=sap_ppnl, force=force, &
3196 : virial=virial, particle_set=particle_set, qs_kind_set=qs_kind_set, &
3197 50 : atomic_kind_set=atomic_kind_set, x_data=x_data, para_env=para_env)
3198 50 : nspins = dft_control%nspins
3199 :
3200 50 : use_virial = virial%pv_availability .AND. (.NOT. virial%pv_numer)
3201 50 : IF (use_virial) virial%pv_calculate = .TRUE.
3202 :
3203 : !Whether we replace the force/energy of SCF XC with HF in RPA
3204 50 : do_exx = .FALSE.
3205 50 : IF (qs_env%mp2_env%method == ri_rpa_method_gpw) THEN
3206 28 : hfx_section => section_vals_get_subs_vals(qs_env%input, "DFT%XC%WF_CORRELATION%RI_RPA%HF")
3207 28 : CALL section_vals_get(hfx_section, explicit=do_exx)
3208 : END IF
3209 :
3210 : !Get the mp2 density matrix which is p_env%p1 + matrix_p_F
3211 50 : CALL get_qs_env(qs_env, matrix_p_mp2=matrix_p_mp2, matrix_p_mp2_admm=matrix_p_mp2_admm)
3212 :
3213 : !The kinetic term (only response density)
3214 50 : NULLIFY (scrm)
3215 50 : mpa2(1:nspins, 1:1) => matrix_p_mp2(1:nspins)
3216 : CALL kinetic_energy_matrix(qs_env, matrix_t=scrm, matrix_p=mpa2, &
3217 : matrix_name="KINETIC ENERGY MATRIX", &
3218 : basis_type="ORB", &
3219 50 : sab_orb=sab_orb, calculate_forces=.TRUE.)
3220 50 : CALL dbcsr_deallocate_matrix_set(scrm)
3221 :
3222 : !The pseudo-potential terms (only reponse density)
3223 50 : CALL dbcsr_allocate_matrix_set(scrm, nspins)
3224 112 : DO ispin = 1, nspins
3225 62 : ALLOCATE (scrm(ispin)%matrix)
3226 62 : CALL dbcsr_create(scrm(ispin)%matrix, template=matrix_s(1)%matrix)
3227 62 : CALL dbcsr_copy(scrm(ispin)%matrix, matrix_s(1)%matrix)
3228 112 : CALL dbcsr_set(scrm(ispin)%matrix, 0.0_dp)
3229 : END DO
3230 :
3231 50 : nder = 1
3232 50 : nimages = 1
3233 424 : ALLOCATE (dbcsr_work_p(nspins, 1), dbcsr_work_h(nspins, 1))
3234 112 : DO ispin = 1, nspins
3235 62 : dbcsr_work_p(ispin, 1)%matrix => matrix_p_mp2(ispin)%matrix
3236 112 : dbcsr_work_h(ispin, 1)%matrix => scrm(ispin)%matrix
3237 : END DO
3238 :
3239 50 : CALL core_matrices(qs_env, dbcsr_work_h, dbcsr_work_p, .TRUE., nder)
3240 :
3241 50 : DEALLOCATE (dbcsr_work_p, dbcsr_work_h)
3242 :
3243 50 : IF (use_virial) THEN
3244 4 : h_stress = 0.0_dp
3245 52 : virial%pv_xc = 0.0_dp
3246 4 : NULLIFY (vxc_rspace, vtau_rspace, vadmm_rspace)
3247 : CALL ks_ref_potential(qs_env, vh_rspace, vxc_rspace, vtau_rspace, vadmm_rspace, &
3248 4 : dummy_real1, dummy_real2, h_stress)
3249 52 : virial%pv_ehartree = virial%pv_ehartree + h_stress/REAL(para_env%num_pe, dp)
3250 52 : virial%pv_virial = virial%pv_virial + h_stress/REAL(para_env%num_pe, dp)
3251 4 : IF (.NOT. do_exx) THEN
3252 : !if RPA EXX, then do not consider XC virial (replaced by RPA%HF virial)
3253 52 : virial%pv_exc = virial%pv_exc - virial%pv_xc
3254 52 : virial%pv_virial = virial%pv_virial - virial%pv_xc
3255 : END IF
3256 : ELSE
3257 46 : CALL ks_ref_potential(qs_env, vh_rspace, vxc_rspace, vtau_rspace, vadmm_rspace, dummy_real1, dummy_real2)
3258 : END IF
3259 50 : do_tau = ASSOCIATED(vtau_rspace)
3260 :
3261 : !Core forces from the SCF
3262 50 : CALL integrate_v_core_rspace(vh_rspace, qs_env)
3263 :
3264 : !The Hartree-xc potential term, P*dVHxc (mp2 + SCF density x deriv of the SCF potential)
3265 : !Get the total density
3266 50 : CALL qs_rho_get(rho, rho_ao=rho_ao)
3267 112 : DO ispin = 1, nspins
3268 112 : CALL dbcsr_add(rho_ao(ispin)%matrix, matrix_p_mp2(ispin)%matrix, 1.0_dp, 1.0_dp)
3269 : END DO
3270 :
3271 50 : CALL get_qs_env(qs_env, pw_env=pw_env)
3272 : CALL pw_env_get(pw_env, auxbas_pw_pool=auxbas_pw_pool, &
3273 50 : poisson_env=poisson_env)
3274 50 : CALL auxbas_pw_pool%create_pw(vhxc_rspace)
3275 :
3276 98 : IF (use_virial) pv_loc = virial%pv_virial
3277 :
3278 50 : IF (do_exx) THEN
3279 : !Only want response XC contribution, but SCF+response Hartree contribution
3280 44 : DO ispin = 1, nspins
3281 : !Hartree
3282 26 : CALL pw_transfer(vh_rspace, vhxc_rspace)
3283 : CALL integrate_v_rspace(v_rspace=vhxc_rspace, &
3284 : hmat=scrm(ispin), pmat=rho_ao(ispin), &
3285 26 : qs_env=qs_env, calculate_forces=.TRUE.)
3286 : !XC
3287 26 : CALL pw_transfer(vxc_rspace(ispin), vhxc_rspace)
3288 : CALL integrate_v_rspace(v_rspace=vhxc_rspace, &
3289 : hmat=scrm(ispin), pmat=matrix_p_mp2(ispin), &
3290 26 : qs_env=qs_env, calculate_forces=.TRUE.)
3291 44 : IF (do_tau) THEN
3292 : CALL integrate_v_rspace(v_rspace=vtau_rspace(ispin), &
3293 : hmat=scrm(ispin), pmat=matrix_p_mp2(ispin), &
3294 0 : qs_env=qs_env, calculate_forces=.TRUE., compute_tau=.TRUE.)
3295 : END IF
3296 : END DO
3297 : ELSE
3298 68 : DO ispin = 1, nspins
3299 36 : CALL pw_transfer(vh_rspace, vhxc_rspace)
3300 36 : CALL pw_axpy(vxc_rspace(ispin), vhxc_rspace)
3301 : CALL integrate_v_rspace(v_rspace=vhxc_rspace, &
3302 : hmat=scrm(ispin), pmat=rho_ao(ispin), &
3303 36 : qs_env=qs_env, calculate_forces=.TRUE.)
3304 68 : IF (do_tau) THEN
3305 : CALL integrate_v_rspace(v_rspace=vtau_rspace(ispin), &
3306 : hmat=scrm(ispin), pmat=rho_ao(ispin), &
3307 8 : qs_env=qs_env, calculate_forces=.TRUE., compute_tau=.TRUE.)
3308 : END IF
3309 : END DO
3310 : END IF
3311 50 : CALL auxbas_pw_pool%give_back_pw(vhxc_rspace)
3312 :
3313 98 : IF (use_virial) virial%pv_ehartree = virial%pv_ehartree + (virial%pv_virial - pv_loc)
3314 :
3315 : !The admm projection contribution (mp2 + SCF densities). If EXX, then only mp2 density
3316 50 : IF (dft_control%do_admm) THEN
3317 : CALL get_admm_env(admm_env, task_list_aux_fit=task_list_aux_fit, rho_aux_fit=rho_aux_fit, &
3318 16 : matrix_s_aux_fit=matrix_s_aux_fit)
3319 16 : CALL qs_rho_get(rho_aux_fit, rho_ao=rho_ao_aux)
3320 16 : CALL dbcsr_allocate_matrix_set(scrm_admm, nspins)
3321 36 : DO ispin = 1, nspins
3322 20 : ALLOCATE (scrm_admm(ispin)%matrix)
3323 20 : CALL dbcsr_create(scrm_admm(ispin)%matrix, template=matrix_s_aux_fit(1)%matrix)
3324 20 : CALL dbcsr_copy(scrm_admm(ispin)%matrix, matrix_s_aux_fit(1)%matrix)
3325 36 : CALL dbcsr_set(scrm_admm(ispin)%matrix, 0.0_dp)
3326 : END DO
3327 :
3328 64 : IF (use_virial) pv_loc = virial%pv_virial
3329 16 : IF (.NOT. qs_env%admm_env%aux_exch_func == do_admm_aux_exch_func_none) THEN
3330 36 : DO ispin = 1, nspins
3331 36 : IF (do_exx) THEN
3332 : CALL integrate_v_rspace(v_rspace=vadmm_rspace(ispin), &
3333 : hmat=scrm_admm(ispin), pmat=matrix_p_mp2_admm(ispin), &
3334 : qs_env=qs_env, calculate_forces=.TRUE., &
3335 8 : basis_type="AUX_FIT", task_list_external=task_list_aux_fit)
3336 : ELSE
3337 12 : CALL dbcsr_add(rho_ao_aux(ispin)%matrix, matrix_p_mp2_admm(ispin)%matrix, 1.0_dp, 1.0_dp)
3338 : CALL integrate_v_rspace(v_rspace=vadmm_rspace(ispin), &
3339 : hmat=scrm_admm(ispin), pmat=rho_ao_aux(ispin), &
3340 : qs_env=qs_env, calculate_forces=.TRUE., &
3341 12 : basis_type="AUX_FIT", task_list_external=task_list_aux_fit)
3342 12 : CALL dbcsr_add(rho_ao_aux(ispin)%matrix, matrix_p_mp2_admm(ispin)%matrix, 1.0_dp, -1.0_dp)
3343 : END IF
3344 : END DO
3345 : END IF
3346 64 : IF (use_virial) virial%pv_ehartree = virial%pv_ehartree + (virial%pv_virial - pv_loc)
3347 :
3348 16 : CALL tddft_hfx_matrix(scrm_admm, rho_ao_aux, qs_env, .FALSE., .FALSE.)
3349 :
3350 16 : IF (do_exx) THEN
3351 4 : CALL admm_projection_derivative(qs_env, scrm_admm, matrix_p_mp2)
3352 : ELSE
3353 12 : CALL admm_projection_derivative(qs_env, scrm_admm, rho_ao)
3354 : END IF
3355 : END IF
3356 :
3357 : !The exact-exchange term (mp2 + SCF densities)
3358 50 : xc_section => section_vals_get_subs_vals(qs_env%input, "DFT%XC")
3359 50 : hfx_section => section_vals_get_subs_vals(xc_section, "HF")
3360 50 : CALL section_vals_get(hfx_section, explicit=do_hfx)
3361 :
3362 50 : IF (do_hfx) THEN
3363 32 : CALL section_vals_get(hfx_section, n_repetition=n_rep_hf)
3364 32 : CPASSERT(n_rep_hf == 1)
3365 80 : IF (use_virial) virial%pv_fock_4c = 0.0_dp
3366 :
3367 : !In case of EXX, only want to response HFX forces, as the SCF will change according to RI_RPA%HF
3368 32 : IF (do_exx) THEN
3369 8 : IF (dft_control%do_admm) THEN
3370 4 : CALL get_admm_env(admm_env, rho_aux_fit=rho_aux_fit)
3371 4 : CALL qs_rho_get(rho_aux_fit, rho_ao=rho_ao_aux, rho_ao_kp=dbcsr_work_p)
3372 4 : IF (x_data(1, 1)%do_hfx_ri) THEN
3373 :
3374 : CALL hfx_ri_update_forces(qs_env, x_data(1, 1)%ri_data, nspins, &
3375 : x_data(1, 1)%general_parameter%fraction, &
3376 : rho_ao=dbcsr_work_p, rho_ao_resp=matrix_p_mp2_admm, &
3377 0 : use_virial=use_virial, resp_only=.TRUE.)
3378 : ELSE
3379 : CALL derivatives_four_center(qs_env, dbcsr_work_p, matrix_p_mp2_admm, hfx_section, para_env, &
3380 4 : 1, use_virial, resp_only=.TRUE.)
3381 : END IF
3382 : ELSE
3383 8 : DO ispin = 1, nspins
3384 8 : CALL dbcsr_add(rho_ao(ispin)%matrix, matrix_p_mp2(ispin)%matrix, 1.0_dp, -1.0_dp)
3385 : END DO
3386 4 : CALL qs_rho_get(rho, rho_ao_kp=dbcsr_work_p)
3387 4 : IF (x_data(1, 1)%do_hfx_ri) THEN
3388 :
3389 : CALL hfx_ri_update_forces(qs_env, x_data(1, 1)%ri_data, nspins, &
3390 : x_data(1, 1)%general_parameter%fraction, &
3391 : rho_ao=dbcsr_work_p, rho_ao_resp=matrix_p_mp2, &
3392 0 : use_virial=use_virial, resp_only=.TRUE.)
3393 : ELSE
3394 : CALL derivatives_four_center(qs_env, dbcsr_work_p, matrix_p_mp2, hfx_section, para_env, &
3395 4 : 1, use_virial, resp_only=.TRUE.)
3396 : END IF
3397 8 : DO ispin = 1, nspins
3398 8 : CALL dbcsr_add(rho_ao(ispin)%matrix, matrix_p_mp2(ispin)%matrix, 1.0_dp, 1.0_dp)
3399 : END DO
3400 : END IF !admm
3401 :
3402 : ELSE !No Exx
3403 24 : IF (dft_control%do_admm) THEN
3404 12 : CALL get_admm_env(admm_env, rho_aux_fit=rho_aux_fit)
3405 12 : CALL qs_rho_get(rho_aux_fit, rho_ao=rho_ao_aux, rho_ao_kp=dbcsr_work_p)
3406 24 : DO ispin = 1, nspins
3407 24 : CALL dbcsr_add(rho_ao_aux(ispin)%matrix, matrix_p_mp2_admm(ispin)%matrix, 1.0_dp, 1.0_dp)
3408 : END DO
3409 12 : IF (x_data(1, 1)%do_hfx_ri) THEN
3410 :
3411 : CALL hfx_ri_update_forces(qs_env, x_data(1, 1)%ri_data, nspins, &
3412 : x_data(1, 1)%general_parameter%fraction, &
3413 : rho_ao=dbcsr_work_p, rho_ao_resp=matrix_p_mp2_admm, &
3414 0 : use_virial=use_virial, resp_only=.FALSE.)
3415 : ELSE
3416 : CALL derivatives_four_center(qs_env, dbcsr_work_p, matrix_p_mp2_admm, hfx_section, para_env, &
3417 12 : 1, use_virial, resp_only=.FALSE.)
3418 : END IF
3419 24 : DO ispin = 1, nspins
3420 24 : CALL dbcsr_add(rho_ao_aux(ispin)%matrix, matrix_p_mp2_admm(ispin)%matrix, 1.0_dp, -1.0_dp)
3421 : END DO
3422 : ELSE
3423 12 : CALL qs_rho_get(rho, rho_ao_kp=dbcsr_work_p)
3424 12 : IF (x_data(1, 1)%do_hfx_ri) THEN
3425 :
3426 : CALL hfx_ri_update_forces(qs_env, x_data(1, 1)%ri_data, nspins, &
3427 : x_data(1, 1)%general_parameter%fraction, &
3428 : rho_ao=dbcsr_work_p, rho_ao_resp=matrix_p_mp2, &
3429 0 : use_virial=use_virial, resp_only=.FALSE.)
3430 : ELSE
3431 : CALL derivatives_four_center(qs_env, dbcsr_work_p, matrix_p_mp2, hfx_section, para_env, &
3432 12 : 1, use_virial, resp_only=.FALSE.)
3433 : END IF
3434 : END IF
3435 : END IF !do_exx
3436 :
3437 32 : IF (use_virial) THEN
3438 52 : virial%pv_exx = virial%pv_exx - virial%pv_fock_4c
3439 52 : virial%pv_virial = virial%pv_virial - virial%pv_fock_4c
3440 : END IF
3441 : END IF
3442 :
3443 : !retrieve the SCF density
3444 50 : CALL qs_rho_get(rho, rho_ao=rho_ao)
3445 112 : DO ispin = 1, nspins
3446 112 : CALL dbcsr_add(rho_ao(ispin)%matrix, matrix_p_mp2(ispin)%matrix, 1.0_dp, -1.0_dp)
3447 : END DO
3448 :
3449 : !From here, we need to do everything twice. Once for the response density, and once for the
3450 : !density that is used for the trace Tr[P*F]. The reason is that the former is needed for the
3451 : !eventual overlap contribution from matrix_wz
3452 : !Only with the mp2 density
3453 :
3454 436 : ALLOCATE (current_density(nspins), current_mat_h(nspins), current_density_admm(nspins))
3455 150 : DO idens = 1, 2
3456 224 : DO ispin = 1, nspins
3457 224 : IF (idens == 1) THEN
3458 62 : current_density(ispin)%matrix => matrix_p_F(ispin)%matrix
3459 62 : current_mat_h(ispin)%matrix => scrm(ispin)%matrix
3460 62 : IF (dft_control%do_admm) current_density_admm(ispin)%matrix => matrix_p_F_admm(ispin)%matrix
3461 : ELSE
3462 62 : current_density(ispin)%matrix => p_env%p1(ispin)%matrix
3463 62 : current_mat_h(ispin)%matrix => matrix_hz(ispin)%matrix
3464 62 : IF (dft_control%do_admm) current_density_admm(ispin)%matrix => p_env%p1_admm(ispin)%matrix
3465 : END IF
3466 : END DO
3467 :
3468 : !The core-denstiy derivative
3469 748 : ALLOCATE (rhoz_r(nspins), rhoz_g(nspins))
3470 224 : DO ispin = 1, nspins
3471 124 : CALL auxbas_pw_pool%create_pw(rhoz_r(ispin))
3472 224 : CALL auxbas_pw_pool%create_pw(rhoz_g(ispin))
3473 : END DO
3474 100 : CALL auxbas_pw_pool%create_pw(rhoz_tot_gspace)
3475 100 : CALL auxbas_pw_pool%create_pw(zv_hartree_rspace)
3476 100 : CALL auxbas_pw_pool%create_pw(zv_hartree_gspace)
3477 :
3478 100 : CALL pw_zero(rhoz_tot_gspace)
3479 224 : DO ispin = 1, nspins
3480 : CALL calculate_rho_elec(ks_env=qs_env%ks_env, matrix_p=current_density(ispin)%matrix, &
3481 124 : rho=rhoz_r(ispin), rho_gspace=rhoz_g(ispin))
3482 224 : CALL pw_axpy(rhoz_g(ispin), rhoz_tot_gspace)
3483 : END DO
3484 :
3485 100 : IF (use_virial) THEN
3486 :
3487 8 : CALL get_qs_env(qs_env, rho=rho)
3488 8 : CALL auxbas_pw_pool%create_pw(rho_tot_gspace)
3489 :
3490 8 : CALL calc_rho_tot_gspace(rho_tot_gspace, qs_env, rho)
3491 :
3492 8 : h_stress(:, :) = 0.0_dp
3493 : CALL pw_poisson_solve(poisson_env, &
3494 : density=rhoz_tot_gspace, &
3495 : ehartree=ehartree, &
3496 : vhartree=zv_hartree_gspace, &
3497 : h_stress=h_stress, &
3498 8 : aux_density=rho_tot_gspace)
3499 :
3500 8 : CALL auxbas_pw_pool%give_back_pw(rho_tot_gspace)
3501 :
3502 : !Green contribution
3503 104 : virial%pv_ehartree = virial%pv_ehartree + 2.0_dp*h_stress/REAL(para_env%num_pe, dp)
3504 104 : virial%pv_virial = virial%pv_virial + 2.0_dp*h_stress/REAL(para_env%num_pe, dp)
3505 :
3506 : ELSE
3507 : CALL pw_poisson_solve(poisson_env, rhoz_tot_gspace, ehartree, &
3508 92 : zv_hartree_gspace)
3509 : END IF
3510 :
3511 100 : CALL pw_transfer(zv_hartree_gspace, zv_hartree_rspace)
3512 100 : CALL pw_scale(zv_hartree_rspace, zv_hartree_rspace%pw_grid%dvol)
3513 100 : CALL integrate_v_core_rspace(zv_hartree_rspace, qs_env)
3514 :
3515 100 : IF (do_tau) THEN
3516 : BLOCK
3517 : TYPE(pw_c1d_gs_type) :: tauz_g
3518 16 : CALL auxbas_pw_pool%create_pw(tauz_g)
3519 48 : ALLOCATE (tauz_r(nspins))
3520 32 : DO ispin = 1, nspins
3521 16 : CALL auxbas_pw_pool%create_pw(tauz_r(ispin))
3522 :
3523 : CALL calculate_rho_elec(ks_env=qs_env%ks_env, matrix_p=current_density(ispin)%matrix, &
3524 32 : rho=tauz_r(ispin), rho_gspace=tauz_g, compute_tau=.TRUE.)
3525 : END DO
3526 16 : CALL auxbas_pw_pool%give_back_pw(tauz_g)
3527 : END BLOCK
3528 : END IF
3529 :
3530 : !Volume contribution to the virial
3531 100 : IF (use_virial) THEN
3532 : !Volume contribution
3533 : exc = 0.0_dp
3534 16 : DO ispin = 1, nspins
3535 : exc = exc + pw_integral_ab(rhoz_r(ispin), vxc_rspace(ispin))/ &
3536 16 : vxc_rspace(ispin)%pw_grid%dvol
3537 : END DO
3538 8 : IF (ASSOCIATED(vtau_rspace)) THEN
3539 0 : DO ispin = 1, nspins
3540 : exc = exc + pw_integral_ab(tauz_r(ispin), vtau_rspace(ispin))/ &
3541 0 : vtau_rspace(ispin)%pw_grid%dvol
3542 : END DO
3543 : END IF
3544 32 : DO i = 1, 3
3545 24 : virial%pv_ehartree(i, i) = virial%pv_ehartree(i, i) - 4.0_dp*ehartree/REAL(para_env%num_pe, dp)
3546 24 : virial%pv_exc(i, i) = virial%pv_exc(i, i) - exc/REAL(para_env%num_pe, dp)
3547 : virial%pv_virial(i, i) = virial%pv_virial(i, i) - 4.0_dp*ehartree/REAL(para_env%num_pe, dp) &
3548 32 : - exc/REAL(para_env%num_pe, dp)
3549 : END DO
3550 : END IF
3551 :
3552 : !The xc-kernel term.
3553 100 : IF (dft_control%do_admm) THEN
3554 32 : CALL get_qs_env(qs_env, admm_env=admm_env)
3555 32 : xc_section => admm_env%xc_section_primary
3556 : ELSE
3557 68 : xc_section => section_vals_get_subs_vals(qs_env%input, "DFT%XC")
3558 : END IF
3559 :
3560 196 : IF (use_virial) virial%pv_xc = 0.0_dp
3561 :
3562 100 : ALLOCATE (rhoz)
3563 100 : CALL qs_rho_create(rhoz)
3564 100 : IF (ASSOCIATED(rhoz_r)) THEN
3565 100 : CALL qs_rho_set(rhoz, rho_r=rhoz_r, rho_r_valid=.TRUE.)
3566 : END IF
3567 100 : IF (ASSOCIATED(rhoz_g)) THEN
3568 100 : CALL qs_rho_set(rhoz, rho_g=rhoz_g, rho_g_valid=.TRUE.)
3569 : END IF
3570 100 : IF (ASSOCIATED(tauz_r)) THEN
3571 16 : CALL qs_rho_set(rhoz, tau_r=tauz_r, tau_r_valid=.TRUE.)
3572 : END IF
3573 : !
3574 : CALL qs_fxc_create(qs_env, rho, rhoz, rho0_atom_set, xc_section, .FALSE., &
3575 : v_xc, v_xc_tau, rho1_atom_set, &
3576 : dispersion_env=qs_env%dispersion_env, &
3577 100 : compute_virial=use_virial, virial_xc=virial%pv_xc)
3578 : !
3579 100 : DEALLOCATE (rhoz)
3580 :
3581 100 : IF (use_virial) THEN
3582 104 : virial%pv_exc = virial%pv_exc + virial%pv_xc
3583 104 : virial%pv_virial = virial%pv_virial + virial%pv_xc
3584 :
3585 104 : pv_loc = virial%pv_virial
3586 : END IF
3587 :
3588 100 : CALL qs_rho_get(rho, rho_ao_kp=dbcsr_work_p)
3589 224 : DO ispin = 1, nspins
3590 124 : CALL pw_scale(v_xc(ispin), v_xc(ispin)%pw_grid%dvol)
3591 124 : CALL pw_axpy(zv_hartree_rspace, v_xc(ispin))
3592 : CALL integrate_v_rspace(qs_env=qs_env, &
3593 : v_rspace=v_xc(ispin), &
3594 : hmat=current_mat_h(ispin), &
3595 : pmat=dbcsr_work_p(ispin, 1), &
3596 124 : calculate_forces=.TRUE.)
3597 224 : CALL auxbas_pw_pool%give_back_pw(v_xc(ispin))
3598 : END DO
3599 100 : CALL auxbas_pw_pool%give_back_pw(rhoz_tot_gspace)
3600 100 : CALL auxbas_pw_pool%give_back_pw(zv_hartree_rspace)
3601 100 : CALL auxbas_pw_pool%give_back_pw(zv_hartree_gspace)
3602 100 : DEALLOCATE (v_xc)
3603 :
3604 100 : IF (do_tau) THEN
3605 32 : DO ispin = 1, nspins
3606 16 : CALL pw_scale(v_xc_tau(ispin), v_xc_tau(ispin)%pw_grid%dvol)
3607 : CALL integrate_v_rspace(qs_env=qs_env, &
3608 : v_rspace=v_xc_tau(ispin), &
3609 : hmat=current_mat_h(ispin), &
3610 : pmat=dbcsr_work_p(ispin, 1), &
3611 : compute_tau=.TRUE., &
3612 16 : calculate_forces=.TRUE.)
3613 32 : CALL auxbas_pw_pool%give_back_pw(v_xc_tau(ispin))
3614 : END DO
3615 16 : DEALLOCATE (v_xc_tau)
3616 : END IF
3617 :
3618 196 : IF (use_virial) virial%pv_ehartree = virial%pv_ehartree + (virial%pv_virial - pv_loc)
3619 :
3620 100 : IF (do_hfx) THEN
3621 64 : IF (dft_control%do_admm) THEN
3622 72 : DO ispin = 1, nspins
3623 72 : CALL dbcsr_set(scrm_admm(ispin)%matrix, 0.0_dp)
3624 : END DO
3625 32 : CALL qs_rho_get(rho_aux_fit, tau_r_valid=do_tau_admm)
3626 :
3627 32 : IF (.NOT. admm_env%aux_exch_func == do_admm_aux_exch_func_none) THEN
3628 32 : CALL get_admm_env(admm_env, rho_aux_fit=rho_aux_fit)
3629 72 : DO ispin = 1, nspins
3630 40 : CALL pw_zero(rhoz_r(ispin))
3631 40 : CALL pw_zero(rhoz_g(ispin))
3632 : CALL calculate_rho_elec(ks_env=qs_env%ks_env, matrix_p=current_density_admm(ispin)%matrix, &
3633 : rho=rhoz_r(ispin), rho_gspace=rhoz_g(ispin), &
3634 72 : basis_type="AUX_FIT", task_list_external=task_list_aux_fit)
3635 : END DO
3636 :
3637 32 : IF (do_tau_admm) THEN
3638 : BLOCK
3639 : TYPE(pw_c1d_gs_type) :: tauz_g
3640 0 : CALL auxbas_pw_pool%create_pw(tauz_g)
3641 0 : DO ispin = 1, nspins
3642 0 : CALL pw_zero(tauz_r(ispin))
3643 : CALL calculate_rho_elec(ks_env=qs_env%ks_env, matrix_p=current_density(ispin)%matrix, &
3644 : rho=tauz_r(ispin), rho_gspace=tauz_g, &
3645 : basis_type="AUX_FIT", task_list_external=task_list_aux_fit, &
3646 0 : compute_tau=.TRUE.)
3647 : END DO
3648 0 : CALL auxbas_pw_pool%give_back_pw(tauz_g)
3649 : END BLOCK
3650 : END IF
3651 :
3652 : !Volume contribution to the virial
3653 32 : IF (use_virial) THEN
3654 : exc = 0.0_dp
3655 16 : DO ispin = 1, nspins
3656 : exc = exc + pw_integral_ab(rhoz_r(ispin), vadmm_rspace(ispin))/ &
3657 16 : vadmm_rspace(ispin)%pw_grid%dvol
3658 : END DO
3659 32 : DO i = 1, 3
3660 24 : virial%pv_exc(i, i) = virial%pv_exc(i, i) - exc/REAL(para_env%num_pe, dp)
3661 32 : virial%pv_virial(i, i) = virial%pv_virial(i, i) - exc/REAL(para_env%num_pe, dp)
3662 : END DO
3663 :
3664 104 : virial%pv_xc = 0.0_dp
3665 : END IF
3666 :
3667 32 : xc_section => admm_env%xc_section_aux
3668 32 : ALLOCATE (rhoz)
3669 32 : CALL qs_rho_create(rhoz)
3670 32 : IF (ASSOCIATED(rhoz_r)) THEN
3671 32 : CALL qs_rho_set(rhoz, rho_r=rhoz_r, rho_r_valid=.TRUE.)
3672 : END IF
3673 32 : IF (ASSOCIATED(rhoz_g)) THEN
3674 32 : CALL qs_rho_set(rhoz, rho_g=rhoz_g, rho_g_valid=.TRUE.)
3675 : END IF
3676 32 : IF (ASSOCIATED(tauz_r)) THEN
3677 0 : CALL qs_rho_set(rhoz, tau_r=tauz_r, tau_r_valid=.TRUE.)
3678 : END IF
3679 : !
3680 : CALL qs_fxc_create(qs_env, rho_aux_fit, rhoz, rho0_atom_set, xc_section, .FALSE., &
3681 : v_xc, v_xc_tau, rho1_atom_set, &
3682 32 : compute_virial=use_virial, virial_xc=virial%pv_xc)
3683 : !
3684 32 : DEALLOCATE (rhoz)
3685 :
3686 32 : IF (use_virial) THEN
3687 104 : virial%pv_exc = virial%pv_exc + virial%pv_xc
3688 104 : virial%pv_virial = virial%pv_virial + virial%pv_xc
3689 :
3690 104 : pv_loc = virial%pv_virial
3691 : END IF
3692 :
3693 32 : CALL qs_rho_get(rho_aux_fit, rho_ao_kp=dbcsr_work_p)
3694 72 : DO ispin = 1, nspins
3695 40 : CALL pw_scale(v_xc(ispin), v_xc(ispin)%pw_grid%dvol)
3696 : CALL integrate_v_rspace(qs_env=qs_env, &
3697 : v_rspace=v_xc(ispin), &
3698 : hmat=scrm_admm(ispin), &
3699 : pmat=dbcsr_work_p(ispin, 1), &
3700 : calculate_forces=.TRUE., &
3701 : basis_type="AUX_FIT", &
3702 40 : task_list_external=task_list_aux_fit)
3703 72 : CALL auxbas_pw_pool%give_back_pw(v_xc(ispin))
3704 : END DO
3705 32 : DEALLOCATE (v_xc)
3706 :
3707 32 : IF (do_tau_admm) THEN
3708 0 : DO ispin = 1, nspins
3709 0 : CALL pw_scale(v_xc_tau(ispin), v_xc_tau(ispin)%pw_grid%dvol)
3710 : CALL integrate_v_rspace(qs_env=qs_env, &
3711 : v_rspace=v_xc_tau(ispin), &
3712 : hmat=scrm_admm(ispin), &
3713 : pmat=dbcsr_work_p(ispin, 1), &
3714 : calculate_forces=.TRUE., &
3715 : basis_type="AUX_FIT", &
3716 : task_list_external=task_list_aux_fit, &
3717 0 : compute_tau=.TRUE.)
3718 0 : CALL auxbas_pw_pool%give_back_pw(v_xc_tau(ispin))
3719 : END DO
3720 0 : DEALLOCATE (v_xc_tau)
3721 : END IF
3722 :
3723 128 : IF (use_virial) virial%pv_ehartree = virial%pv_ehartree + (virial%pv_virial - pv_loc)
3724 : END IF
3725 :
3726 32 : CALL tddft_hfx_matrix(scrm_admm, current_density_admm, qs_env, .FALSE., .FALSE.)
3727 :
3728 32 : CALL qs_rho_get(rho, rho_ao_kp=dbcsr_work_p)
3729 32 : CALL admm_projection_derivative(qs_env, scrm_admm, dbcsr_work_p(:, 1))
3730 :
3731 : !If response density, need to get matrix_hz contribution
3732 32 : CALL dbcsr_create(dbcsr_work, template=matrix_s(1)%matrix)
3733 32 : IF (idens == 2) THEN
3734 16 : nao = admm_env%nao_orb
3735 16 : nao_aux = admm_env%nao_aux_fit
3736 36 : DO ispin = 1, nspins
3737 20 : CALL dbcsr_copy(dbcsr_work, matrix_hz(ispin)%matrix)
3738 20 : CALL dbcsr_set(dbcsr_work, 0.0_dp)
3739 :
3740 : CALL cp_dbcsr_sm_fm_multiply(scrm_admm(ispin)%matrix, admm_env%A, &
3741 20 : admm_env%work_aux_orb, nao)
3742 : CALL parallel_gemm('T', 'N', nao, nao, nao_aux, &
3743 : 1.0_dp, admm_env%A, admm_env%work_aux_orb, 0.0_dp, &
3744 20 : admm_env%work_orb_orb)
3745 20 : CALL copy_fm_to_dbcsr(admm_env%work_orb_orb, dbcsr_work, keep_sparsity=.TRUE.)
3746 36 : CALL dbcsr_add(matrix_hz(ispin)%matrix, dbcsr_work, 1.0_dp, 1.0_dp)
3747 : END DO
3748 : END IF
3749 :
3750 32 : CALL dbcsr_release(dbcsr_work)
3751 : ELSE !no admm
3752 :
3753 : !Need the contribution to matrix_hz as well
3754 32 : IF (idens == 2) THEN
3755 16 : CALL tddft_hfx_matrix(matrix_hz, current_density, qs_env, .FALSE., .FALSE.)
3756 : END IF
3757 : END IF !admm
3758 : END IF !do_hfx
3759 :
3760 224 : DO ispin = 1, nspins
3761 124 : CALL auxbas_pw_pool%give_back_pw(rhoz_r(ispin))
3762 224 : CALL auxbas_pw_pool%give_back_pw(rhoz_g(ispin))
3763 : END DO
3764 100 : DEALLOCATE (rhoz_r, rhoz_g)
3765 :
3766 250 : IF (do_tau) THEN
3767 32 : DO ispin = 1, nspins
3768 32 : CALL auxbas_pw_pool%give_back_pw(tauz_r(ispin))
3769 : END DO
3770 16 : DEALLOCATE (tauz_r)
3771 : END IF
3772 : END DO !idens
3773 50 : CALL dbcsr_deallocate_matrix_set(scrm_admm)
3774 :
3775 50 : DEALLOCATE (current_density, current_mat_h, current_density_admm)
3776 50 : CALL dbcsr_deallocate_matrix_set(scrm)
3777 :
3778 : !The energy weighted and overlap term. ONLY with the response density
3779 50 : focc = 2.0_dp
3780 50 : IF (nspins == 2) focc = 1.0_dp
3781 50 : CALL get_qs_env(qs_env, mos=mos)
3782 112 : DO ispin = 1, nspins
3783 62 : CALL get_mo_set(mo_set=mos(ispin), homo=nocc)
3784 : CALL calculate_whz_matrix(mos(ispin)%mo_coeff, matrix_hz(ispin)%matrix, &
3785 112 : p_env%w1(ispin)%matrix, focc, nocc)
3786 : END DO
3787 50 : IF (nspins == 2) CALL dbcsr_add(p_env%w1(1)%matrix, p_env%w1(2)%matrix, 1.0_dp, 1.0_dp)
3788 :
3789 : !Add to it the SCF W matrix, except if EXX (because taken care of by HF response)
3790 50 : IF (.NOT. do_exx) THEN
3791 32 : CALL compute_matrix_w(qs_env, calc_forces=.TRUE.)
3792 32 : CALL get_qs_env(qs_env, matrix_w=matrix_w)
3793 32 : CALL dbcsr_add(p_env%w1(1)%matrix, matrix_w(1)%matrix, 1.0_dp, 1.0_dp)
3794 32 : IF (nspins == 2) CALL dbcsr_add(p_env%w1(1)%matrix, matrix_w(2)%matrix, 1.0_dp, 1.0_dp)
3795 : END IF
3796 :
3797 50 : NULLIFY (scrm)
3798 : CALL build_overlap_matrix(qs_env%ks_env, matrix_s=scrm, &
3799 : matrix_name="OVERLAP MATRIX", &
3800 : basis_type_a="ORB", basis_type_b="ORB", &
3801 : sab_nl=sab_orb, calculate_forces=.TRUE., &
3802 50 : matrix_p=p_env%w1(1)%matrix)
3803 :
3804 50 : IF (.NOT. do_exx) THEN
3805 32 : CALL dbcsr_add(p_env%w1(1)%matrix, matrix_w(1)%matrix, 1.0_dp, -1.0_dp)
3806 32 : IF (nspins == 2) CALL dbcsr_add(p_env%w1(1)%matrix, matrix_w(2)%matrix, 1.0_dp, -1.0_dp)
3807 68 : DO ispin = 1, nspins
3808 68 : CALL dbcsr_set(matrix_w(ispin)%matrix, 0.0_dp)
3809 : END DO
3810 : END IF
3811 :
3812 50 : IF (nspins == 2) CALL dbcsr_add(p_env%w1(1)%matrix, p_env%w1(2)%matrix, 1.0_dp, -1.0_dp)
3813 50 : CALL dbcsr_deallocate_matrix_set(scrm)
3814 :
3815 50 : IF (use_virial) virial%pv_calculate = .FALSE.
3816 :
3817 : !clean-up
3818 50 : CALL auxbas_pw_pool%give_back_pw(vh_rspace)
3819 :
3820 112 : DO ispin = 1, nspins
3821 62 : CALL auxbas_pw_pool%give_back_pw(vxc_rspace(ispin))
3822 62 : IF (ASSOCIATED(vtau_rspace)) THEN
3823 8 : CALL auxbas_pw_pool%give_back_pw(vtau_rspace(ispin))
3824 : END IF
3825 112 : IF (ASSOCIATED(vadmm_rspace)) THEN
3826 20 : CALL auxbas_pw_pool%give_back_pw(vadmm_rspace(ispin))
3827 : END IF
3828 : END DO
3829 50 : DEALLOCATE (vxc_rspace)
3830 50 : IF (ASSOCIATED(vtau_rspace)) DEALLOCATE (vtau_rspace)
3831 50 : IF (ASSOCIATED(vadmm_rspace)) DEALLOCATE (vadmm_rspace)
3832 :
3833 50 : CALL timestop(handle)
3834 :
3835 100 : END SUBROUTINE update_im_time_forces
3836 :
3837 : ! **************************************************************************************************
3838 : !> \brief Iteratively builds the matrix Y = sum_k Y_k until convergence, where
3839 : !> Y_k = 1/k*2^n (A/2^n) Y_k-1 + 1/k!*2^n * PR(n) * (A/2^n)^(k-1)
3840 : !> n is chosen such that the norm of A is < 1 (and e^A converges fast)
3841 : !> PR(n) = e^(A/2^n)*PR(n-1) + PR(n-1)*e^(A/2^n), PR(0) = P*R^T
3842 : !> \param Y ...
3843 : !> \param A ...
3844 : !> \param P ...
3845 : !> \param R ...
3846 : !> \param filter_eps ...
3847 : ! **************************************************************************************************
3848 340 : SUBROUTINE build_Y_matrix(Y, A, P, R, filter_eps)
3849 :
3850 : TYPE(dbcsr_type), INTENT(OUT) :: Y
3851 : TYPE(dbcsr_type), INTENT(INOUT) :: A, P, R
3852 : REAL(dp), INTENT(IN) :: filter_eps
3853 :
3854 : CHARACTER(LEN=*), PARAMETER :: routineN = 'build_Y_matrix'
3855 :
3856 : INTEGER :: handle, k, n
3857 : REAL(dp) :: norm_scalar, threshold
3858 : TYPE(dbcsr_type) :: A2n, exp_A2n, PRn, work, work2, Yk
3859 :
3860 340 : CALL timeset(routineN, handle)
3861 :
3862 340 : threshold = 1.0E-16_dp
3863 :
3864 : !Find n such that norm(A) < 1 and we insure convergence of the exponential
3865 340 : norm_scalar = dbcsr_frobenius_norm(A)
3866 :
3867 : !checked: result invariant with value of n
3868 340 : n = 1
3869 466 : DO
3870 806 : IF ((norm_scalar/2.0_dp**n) < 1.0_dp) EXIT
3871 466 : n = n + 1
3872 : END DO
3873 :
3874 : !Calculate PR(n) recursively
3875 340 : CALL dbcsr_create(PRn, template=A, matrix_type=dbcsr_type_no_symmetry)
3876 340 : CALL dbcsr_create(work, template=A, matrix_type=dbcsr_type_no_symmetry)
3877 340 : CALL dbcsr_multiply('N', 'N', 1.0_dp, P, R, 0.0_dp, work, filter_eps=filter_eps)
3878 340 : CALL dbcsr_create(exp_A2n, template=A, matrix_type=dbcsr_type_no_symmetry)
3879 :
3880 1146 : DO k = 1, n
3881 806 : CALL matrix_exponential(exp_A2n, A, 1.0_dp, 0.5_dp**k, threshold)
3882 806 : CALL dbcsr_multiply('N', 'N', 1.0_dp, exp_A2n, work, 0.0_dp, PRn, filter_eps=filter_eps)
3883 806 : CALL dbcsr_multiply('N', 'N', 1.0_dp, work, exp_A2n, 1.0_dp, PRn, filter_eps=filter_eps)
3884 1146 : CALL dbcsr_copy(work, PRn)
3885 : END DO
3886 340 : CALL dbcsr_release(exp_A2n)
3887 :
3888 : !Calculate Y iteratively, until convergence
3889 340 : CALL dbcsr_create(A2n, template=A, matrix_type=dbcsr_type_no_symmetry)
3890 340 : CALL dbcsr_copy(A2n, A)
3891 340 : CALL dbcsr_scale(A2n, 0.5_dp**n)
3892 340 : CALL dbcsr_create(Y, template=A, matrix_type=dbcsr_type_no_symmetry)
3893 340 : CALL dbcsr_create(Yk, template=A, matrix_type=dbcsr_type_no_symmetry)
3894 340 : CALL dbcsr_create(work2, template=A, matrix_type=dbcsr_type_no_symmetry)
3895 :
3896 : !k=1
3897 340 : CALL dbcsr_scale(PRn, 0.5_dp**n)
3898 340 : CALL dbcsr_copy(work, PRn)
3899 340 : CALL dbcsr_copy(work2, PRn)
3900 340 : CALL dbcsr_add(Y, PRn, 1.0_dp, 1.0_dp)
3901 :
3902 340 : k = 1
3903 1908 : DO
3904 2248 : k = k + 1
3905 2248 : CALL dbcsr_multiply('N', 'N', 1.0_dp/REAL(k, dp), A2n, work, 0.0_dp, Yk, filter_eps=filter_eps)
3906 2248 : CALL dbcsr_multiply('N', 'N', 1.0_dp/REAL(k, dp), work2, A2n, 0.0_dp, PRn, filter_eps=filter_eps)
3907 :
3908 2248 : CALL dbcsr_add(Yk, PRn, 1.0_dp, 1.0_dp)
3909 2248 : CALL dbcsr_add(Y, Yk, 1.0_dp, 1.0_dp)
3910 :
3911 2248 : IF (dbcsr_frobenius_norm(Yk) < threshold) EXIT
3912 1908 : CALL dbcsr_copy(work, Yk)
3913 1908 : CALL dbcsr_copy(work2, PRn)
3914 : END DO
3915 :
3916 340 : CALL dbcsr_release(work)
3917 340 : CALL dbcsr_release(work2)
3918 340 : CALL dbcsr_release(PRn)
3919 340 : CALL dbcsr_release(A2n)
3920 340 : CALL dbcsr_release(Yk)
3921 :
3922 340 : CALL timestop(handle)
3923 :
3924 340 : END SUBROUTINE build_Y_matrix
3925 :
3926 : ! **************************************************************************************************
3927 : !> \brief Overwrites the "optimal" Laplace quadrature with that of the first step
3928 : !> \param grid ...
3929 : !> \param do_laplace ...
3930 : !> \param do_im_time ...
3931 : !> \param unit_nr ...
3932 : !> \param qs_env ...
3933 : ! **************************************************************************************************
3934 214 : SUBROUTINE keep_initial_quad(grid, do_laplace, do_im_time, unit_nr, qs_env)
3935 :
3936 : TYPE(time_frequency_grid_type), INTENT(INOUT) :: grid
3937 : LOGICAL, INTENT(IN) :: do_laplace, do_im_time
3938 : INTEGER, INTENT(IN) :: unit_nr
3939 : TYPE(qs_environment_type), POINTER :: qs_env
3940 :
3941 : INTEGER :: jquad
3942 :
3943 214 : IF (do_laplace .OR. do_im_time) THEN
3944 172 : IF (.NOT. ALLOCATED(qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%imaginary_time)) THEN
3945 1410 : ALLOCATE (qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%imaginary_time, SOURCE=grid%imaginary_time)
3946 0 : ALLOCATE (qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%time_weights_at_zero_frequency, &
3947 1410 : SOURCE=grid%time_weights_at_zero_frequency)
3948 : ELSE
3949 : !If weights already stored, we overwrite the new ones
3950 152 : grid%imaginary_time(:) = qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%imaginary_time(:)
3951 : grid%time_weights_at_zero_frequency(:) = &
3952 194 : qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%time_weights_at_zero_frequency(:)
3953 : END IF
3954 : END IF
3955 214 : IF (.NOT. do_laplace) THEN
3956 150 : IF (.NOT. ALLOCATED(qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%frequency)) THEN
3957 1348 : ALLOCATE (qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%frequency, SOURCE=grid%frequency)
3958 0 : ALLOCATE (qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%frequency_weights, &
3959 1348 : SOURCE=grid%frequency_weights)
3960 122 : IF (do_im_time) THEN
3961 0 : ALLOCATE (qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%cosine_time_to_frequency, &
3962 14260 : SOURCE=grid%cosine_time_to_frequency)
3963 0 : ALLOCATE (qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%cosine_frequency_to_time, &
3964 14260 : SOURCE=grid%cosine_frequency_to_time)
3965 : END IF
3966 : ELSE
3967 120 : grid%frequency(:) = qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%frequency(:)
3968 120 : grid%frequency_weights(:) = qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%frequency_weights(:)
3969 28 : IF (do_im_time) THEN
3970 : grid%cosine_time_to_frequency(:, :) = &
3971 184 : qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%cosine_time_to_frequency(:, :)
3972 : grid%cosine_frequency_to_time(:, :) = &
3973 184 : qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%cosine_frequency_to_time(:, :)
3974 : END IF
3975 : END IF
3976 : END IF
3977 214 : IF (unit_nr > 0) THEN
3978 : !Printing order same as in mp2_grids.F for consistency
3979 107 : IF (ALLOCATED(qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%frequency) .AND. (.NOT. do_laplace)) THEN
3980 : WRITE (UNIT=unit_nr, FMT="(T3,A,T75,i6)") &
3981 75 : "MINIMAX_INFO| Number of integration points:", SIZE(grid%frequency)
3982 : WRITE (UNIT=unit_nr, FMT="(T3,A,T54,A,T72,A)") &
3983 75 : "MINIMAX_INFO| Minimax params (freq grid, scaled):", "Weights", "Abscissas"
3984 612 : DO jquad = 1, SIZE(grid%frequency)
3985 : WRITE (UNIT=unit_nr, FMT="(T41,F20.10,F20.10)") &
3986 612 : grid%frequency_weights(jquad), grid%frequency(jquad)
3987 : END DO
3988 75 : CALL m_flush(unit_nr)
3989 : END IF
3990 107 : IF (ALLOCATED(qs_env%mp2_env%ri_rpa_im_time%time_frequency_grid%imaginary_time)) THEN
3991 : WRITE (UNIT=unit_nr, FMT="(T3,A,T75,i6)") &
3992 86 : "MINIMAX_INFO| Number of integration points:", SIZE(grid%imaginary_time)
3993 : WRITE (UNIT=unit_nr, FMT="(T3,A,T54,A,T72,A)") &
3994 86 : "MINIMAX_INFO| Minimax params (time grid, scaled):", "Weights", "Abscissas"
3995 645 : DO jquad = 1, SIZE(grid%imaginary_time)
3996 : WRITE (UNIT=unit_nr, FMT="(T41,F20.10,F20.10)") &
3997 645 : grid%time_weights_at_zero_frequency(jquad), grid%imaginary_time(jquad)
3998 : END DO
3999 86 : CALL m_flush(unit_nr)
4000 : END IF
4001 : END IF
4002 :
4003 214 : END SUBROUTINE keep_initial_quad
4004 :
4005 : END MODULE rpa_im_time_force_methods
|
