Line data Source code
1 : !--------------------------------------------------------------------------------------------------!
2 : ! CP2K: A general program to perform molecular dynamics simulations !
3 : ! Copyright 2000-2024 CP2K developers group <https://cp2k.org> !
4 : ! !
5 : ! SPDX-License-Identifier: GPL-2.0-or-later !
6 : !--------------------------------------------------------------------------------------------------!
7 :
8 : ! **************************************************************************************************
9 : MODULE qs_gspace_mixing
10 :
11 : USE cp_control_types, ONLY: dft_control_type
12 : USE kinds, ONLY: dp
13 : USE mathlib, ONLY: invert_matrix
14 : USE message_passing, ONLY: mp_para_env_type
15 : USE pw_env_types, ONLY: pw_env_get,&
16 : pw_env_type
17 : USE pw_methods, ONLY: pw_axpy,&
18 : pw_copy,&
19 : pw_integrate_function,&
20 : pw_scale,&
21 : pw_transfer,&
22 : pw_zero
23 : USE pw_pool_types, ONLY: pw_pool_type
24 : USE pw_types, ONLY: pw_c1d_gs_type,&
25 : pw_r3d_rs_type
26 : USE qs_density_mixing_types, ONLY: broyden_mixing_nr,&
27 : gspace_mixing_nr,&
28 : mixing_storage_type,&
29 : multisecant_mixing_nr,&
30 : pulay_mixing_nr
31 : USE qs_environment_types, ONLY: get_qs_env,&
32 : qs_environment_type
33 : USE qs_rho_atom_types, ONLY: rho_atom_type
34 : USE qs_rho_types, ONLY: qs_rho_get,&
35 : qs_rho_type
36 : #include "./base/base_uses.f90"
37 :
38 : IMPLICIT NONE
39 :
40 : PRIVATE
41 :
42 : CHARACTER(len=*), PARAMETER, PRIVATE :: moduleN = 'qs_gspace_mixing'
43 :
44 : PUBLIC :: gspace_mixing
45 :
46 : CONTAINS
47 :
48 : ! **************************************************************************************************
49 : !> \brief Driver for the g-space mixing, calls the proper routine given the
50 : !>requested method
51 : !> \param qs_env ...
52 : !> \param mixing_method ...
53 : !> \param mixing_store ...
54 : !> \param rho ...
55 : !> \param para_env ...
56 : !> \param iter_count ...
57 : !> \par History
58 : !> 05.2009
59 : !> 02.2015 changed input to make g-space mixing available in linear scaling
60 : !> SCF [Patrick Seewald]
61 : !> \author MI
62 : ! **************************************************************************************************
63 1524 : SUBROUTINE gspace_mixing(qs_env, mixing_method, mixing_store, rho, para_env, iter_count)
64 : TYPE(qs_environment_type), POINTER :: qs_env
65 : INTEGER :: mixing_method
66 : TYPE(mixing_storage_type), POINTER :: mixing_store
67 : TYPE(qs_rho_type), POINTER :: rho
68 : TYPE(mp_para_env_type), POINTER :: para_env
69 : INTEGER :: iter_count
70 :
71 : CHARACTER(len=*), PARAMETER :: routineN = 'gspace_mixing'
72 :
73 : INTEGER :: handle, iatom, ig, ispin, natom, ng, &
74 : nimg, nspin
75 : LOGICAL :: gapw
76 : REAL(dp) :: alpha
77 1524 : REAL(KIND=dp), DIMENSION(:), POINTER :: tot_rho_r
78 : TYPE(dft_control_type), POINTER :: dft_control
79 : TYPE(pw_c1d_gs_type) :: rho_tmp
80 1524 : TYPE(pw_c1d_gs_type), DIMENSION(:), POINTER :: rho_g
81 : TYPE(pw_env_type), POINTER :: pw_env
82 : TYPE(pw_pool_type), POINTER :: auxbas_pw_pool
83 1524 : TYPE(pw_r3d_rs_type), DIMENSION(:), POINTER :: rho_r
84 1524 : TYPE(rho_atom_type), DIMENSION(:), POINTER :: rho_atom
85 :
86 1524 : CALL timeset(routineN, handle)
87 :
88 1524 : CALL get_qs_env(qs_env, dft_control=dft_control, pw_env=pw_env)
89 1524 : IF (.NOT. (dft_control%qs_control%dftb .OR. dft_control%qs_control%xtb .OR. &
90 : dft_control%qs_control%semi_empirical)) THEN
91 :
92 1028 : CPASSERT(ASSOCIATED(mixing_store))
93 1028 : NULLIFY (auxbas_pw_pool, rho_atom, rho_g, rho_r, tot_rho_r)
94 1028 : CALL qs_rho_get(rho, rho_g=rho_g, rho_r=rho_r, tot_rho_r=tot_rho_r)
95 :
96 1028 : nspin = SIZE(rho_g, 1)
97 1028 : nimg = dft_control%nimages
98 1028 : ng = SIZE(rho_g(1)%pw_grid%gsq)
99 1028 : CPASSERT(ng == SIZE(mixing_store%rhoin(1)%cc))
100 :
101 1028 : alpha = mixing_store%alpha
102 1028 : gapw = dft_control%qs_control%gapw
103 :
104 1028 : IF (nspin == 2) THEN
105 114 : CALL pw_env_get(pw_env=pw_env, auxbas_pw_pool=auxbas_pw_pool)
106 114 : CALL auxbas_pw_pool%create_pw(rho_tmp)
107 114 : CALL pw_zero(rho_tmp)
108 114 : CALL pw_copy(rho_g(1), rho_tmp)
109 114 : CALL pw_axpy(rho_g(2), rho_g(1), 1.0_dp)
110 114 : CALL pw_axpy(rho_tmp, rho_g(2), -1.0_dp)
111 114 : CALL pw_scale(rho_g(2), -1.0_dp)
112 : END IF
113 :
114 1028 : IF (iter_count + 1 <= mixing_store%nskip_mixing) THEN
115 : ! skip mixing
116 8 : DO ispin = 1, nspin
117 27656 : DO ig = 1, ng
118 27652 : mixing_store%rhoin(ispin)%cc(ig) = rho_g(ispin)%array(ig)
119 : END DO
120 : END DO
121 4 : IF (mixing_store%gmix_p .AND. gapw) THEN
122 0 : CALL get_qs_env(qs_env=qs_env, rho_atom_set=rho_atom)
123 0 : natom = SIZE(rho_atom)
124 0 : DO ispin = 1, nspin
125 0 : DO iatom = 1, natom
126 0 : IF (mixing_store%paw(iatom)) THEN
127 0 : mixing_store%cpc_h_in(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef
128 0 : mixing_store%cpc_s_in(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef
129 : END IF
130 : END DO
131 : END DO
132 : END IF
133 :
134 4 : mixing_store%iter_method = "NoMix"
135 4 : IF (nspin == 2) THEN
136 0 : CALL pw_axpy(rho_g(2), rho_g(1), 1.0_dp)
137 0 : CALL pw_scale(rho_g(1), 0.5_dp)
138 0 : CALL pw_axpy(rho_g(1), rho_g(2), -1.0_dp)
139 0 : CALL pw_scale(rho_g(2), -1.0_dp)
140 0 : CALL auxbas_pw_pool%give_back_pw(rho_tmp)
141 : END IF
142 4 : CALL timestop(handle)
143 4 : RETURN
144 : END IF
145 :
146 1024 : IF ((iter_count + 1 - mixing_store%nskip_mixing) <= mixing_store%n_simple_mix) THEN
147 6 : CALL gmix_potential_only(qs_env, mixing_store, rho)
148 6 : mixing_store%iter_method = "Kerker"
149 1018 : ELSEIF (mixing_method == gspace_mixing_nr) THEN
150 40 : CALL gmix_potential_only(qs_env, mixing_store, rho)
151 40 : mixing_store%iter_method = "Kerker"
152 978 : ELSEIF (mixing_method == pulay_mixing_nr) THEN
153 184 : CALL pulay_mixing(qs_env, mixing_store, rho, para_env)
154 184 : mixing_store%iter_method = "Pulay"
155 794 : ELSEIF (mixing_method == broyden_mixing_nr) THEN
156 794 : CALL broyden_mixing(qs_env, mixing_store, rho, para_env)
157 794 : mixing_store%iter_method = "Broy."
158 0 : ELSEIF (mixing_method == multisecant_mixing_nr) THEN
159 0 : CPASSERT(.NOT. gapw)
160 0 : CALL multisecant_mixing(mixing_store, rho, para_env)
161 0 : mixing_store%iter_method = "MSec."
162 : END IF
163 :
164 1024 : IF (nspin == 2) THEN
165 114 : CALL pw_axpy(rho_g(2), rho_g(1), 1.0_dp)
166 114 : CALL pw_scale(rho_g(1), 0.5_dp)
167 114 : CALL pw_axpy(rho_g(1), rho_g(2), -1.0_dp)
168 114 : CALL pw_scale(rho_g(2), -1.0_dp)
169 114 : CALL auxbas_pw_pool%give_back_pw(rho_tmp)
170 : END IF
171 :
172 2162 : DO ispin = 1, nspin
173 1138 : CALL pw_transfer(rho_g(ispin), rho_r(ispin))
174 2162 : tot_rho_r(ispin) = pw_integrate_function(rho_r(ispin), isign=-1)
175 : END DO
176 : END IF
177 :
178 1520 : CALL timestop(handle)
179 :
180 1524 : END SUBROUTINE gspace_mixing
181 :
182 : ! **************************************************************************************************
183 : !> \brief G-space mixing performed via the Kerker damping on the density on the grid
184 : !> thus affecting only the caluclation of the potential, but not the denisity matrix
185 : !> \param qs_env ...
186 : !> \param mixing_store ...
187 : !> \param rho ...
188 : !> \par History
189 : !> 02.2009
190 : !> \author MI
191 : ! **************************************************************************************************
192 :
193 92 : SUBROUTINE gmix_potential_only(qs_env, mixing_store, rho)
194 :
195 : TYPE(qs_environment_type), POINTER :: qs_env
196 : TYPE(mixing_storage_type), POINTER :: mixing_store
197 : TYPE(qs_rho_type), POINTER :: rho
198 :
199 : CHARACTER(len=*), PARAMETER :: routineN = 'gmix_potential_only'
200 :
201 46 : COMPLEX(dp), DIMENSION(:), POINTER :: cc_new
202 : INTEGER :: handle, iatom, ig, ispin, natom, ng, &
203 : nspin
204 : LOGICAL :: gapw
205 : REAL(dp) :: alpha, f_mix
206 : TYPE(dft_control_type), POINTER :: dft_control
207 46 : TYPE(pw_c1d_gs_type), DIMENSION(:), POINTER :: rho_g
208 46 : TYPE(rho_atom_type), DIMENSION(:), POINTER :: rho_atom
209 :
210 0 : CPASSERT(ASSOCIATED(mixing_store%rhoin))
211 46 : CPASSERT(ASSOCIATED(mixing_store%kerker_factor))
212 :
213 46 : CALL timeset(routineN, handle)
214 46 : NULLIFY (cc_new, dft_control, rho_atom, rho_g)
215 :
216 46 : CALL get_qs_env(qs_env, dft_control=dft_control)
217 46 : CALL qs_rho_get(rho, rho_g=rho_g)
218 :
219 46 : nspin = SIZE(rho_g, 1)
220 46 : ng = SIZE(rho_g(1)%pw_grid%gsq)
221 :
222 46 : gapw = dft_control%qs_control%gapw
223 46 : alpha = mixing_store%alpha
224 :
225 92 : DO ispin = 1, nspin
226 46 : cc_new => rho_g(ispin)%array
227 580654 : DO ig = 1, mixing_store%ig_max ! ng
228 580608 : f_mix = mixing_store%alpha*mixing_store%kerker_factor(ig)
229 580608 : cc_new(ig) = (1.0_dp - f_mix)*mixing_store%rhoin(ispin)%cc(ig) + f_mix*cc_new(ig)
230 580654 : mixing_store%rhoin(ispin)%cc(ig) = cc_new(ig)
231 : END DO
232 92 : DO ig = mixing_store%ig_max + 1, ng
233 0 : f_mix = mixing_store%alpha
234 0 : cc_new(ig) = (1.0_dp - f_mix)*mixing_store%rhoin(ispin)%cc(ig) + f_mix*cc_new(ig)
235 46 : mixing_store%rhoin(ispin)%cc(ig) = cc_new(ig)
236 : END DO
237 :
238 : END DO
239 :
240 46 : IF (mixing_store%gmix_p .AND. gapw) THEN
241 : CALL get_qs_env(qs_env=qs_env, &
242 8 : rho_atom_set=rho_atom)
243 8 : natom = SIZE(rho_atom)
244 16 : DO ispin = 1, nspin
245 80 : DO iatom = 1, natom
246 72 : IF (mixing_store%paw(iatom)) THEN
247 : rho_atom(iatom)%cpc_h(ispin)%r_coef = alpha*rho_atom(iatom)%cpc_h(ispin)%r_coef + &
248 7408 : mixing_store%cpc_h_in(iatom, ispin)%r_coef*(1._dp - alpha)
249 : rho_atom(iatom)%cpc_s(ispin)%r_coef = alpha*rho_atom(iatom)%cpc_s(ispin)%r_coef + &
250 7408 : mixing_store%cpc_s_in(iatom, ispin)%r_coef*(1._dp - alpha)
251 7408 : mixing_store%cpc_h_in(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef
252 7408 : mixing_store%cpc_s_in(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef
253 : END IF
254 : END DO
255 : END DO
256 : END IF
257 :
258 46 : CALL timestop(handle)
259 :
260 46 : END SUBROUTINE gmix_potential_only
261 :
262 : ! **************************************************************************************************
263 : !> \brief Pulay Mixing using as metrics for the residual the Kerer damping factor
264 : !> The mixing is applied directly on the density in reciprocal space,
265 : !> therefore it affects the potentials
266 : !> on the grid but not the density matrix
267 : !> \param qs_env ...
268 : !> \param mixing_store ...
269 : !> \param rho ...
270 : !> \param para_env ...
271 : !> \par History
272 : !> 03.2009
273 : !> \author MI
274 : ! **************************************************************************************************
275 :
276 184 : SUBROUTINE pulay_mixing(qs_env, mixing_store, rho, para_env)
277 :
278 : TYPE(qs_environment_type), POINTER :: qs_env
279 : TYPE(mixing_storage_type), POINTER :: mixing_store
280 : TYPE(qs_rho_type), POINTER :: rho
281 : TYPE(mp_para_env_type), POINTER :: para_env
282 :
283 : CHARACTER(len=*), PARAMETER :: routineN = 'pulay_mixing'
284 :
285 184 : COMPLEX(dp), ALLOCATABLE, DIMENSION(:) :: cc_mix
286 : INTEGER :: handle, i, iatom, ib, ibb, ig, ispin, &
287 : jb, n1, n2, natom, nb, nb1, nbuffer, &
288 : ng, nspin
289 : LOGICAL :: gapw
290 : REAL(dp) :: alpha_kerker, alpha_pulay, beta, f_mix, &
291 : inv_err, norm, norm_c_inv, res_norm, &
292 : vol
293 184 : REAL(dp), ALLOCATABLE, DIMENSION(:) :: alpha_c, ev
294 184 : REAL(dp), ALLOCATABLE, DIMENSION(:, :) :: a, b, c, c_inv, cpc_h_mix, cpc_s_mix
295 : TYPE(dft_control_type), POINTER :: dft_control
296 184 : TYPE(pw_c1d_gs_type), DIMENSION(:), POINTER :: rho_g
297 184 : TYPE(rho_atom_type), DIMENSION(:), POINTER :: rho_atom
298 :
299 0 : CPASSERT(ASSOCIATED(mixing_store%res_buffer))
300 184 : CPASSERT(ASSOCIATED(mixing_store%rhoin_buffer))
301 184 : NULLIFY (dft_control, rho_atom, rho_g)
302 184 : CALL timeset(routineN, handle)
303 :
304 184 : CALL get_qs_env(qs_env, dft_control=dft_control)
305 184 : CALL qs_rho_get(rho, rho_g=rho_g)
306 184 : nspin = SIZE(rho_g, 1)
307 184 : ng = SIZE(mixing_store%res_buffer(1, 1)%cc)
308 184 : vol = rho_g(1)%pw_grid%vol
309 :
310 184 : alpha_kerker = mixing_store%alpha
311 184 : beta = mixing_store%pulay_beta
312 184 : alpha_pulay = mixing_store%pulay_alpha
313 184 : nbuffer = mixing_store%nbuffer
314 184 : gapw = dft_control%qs_control%gapw
315 184 : IF (gapw) THEN
316 12 : CALL get_qs_env(qs_env=qs_env, rho_atom_set=rho_atom)
317 12 : natom = SIZE(rho_atom)
318 : END IF
319 :
320 552 : ALLOCATE (cc_mix(ng))
321 :
322 184 : IF (nspin == 2) THEN
323 14 : CALL pw_axpy(rho_g(1), rho_g(2), 1.0_dp, -1.0_dp)
324 96782 : DO ig = 1, ng
325 96782 : mixing_store%rhoin(2)%cc(ig) = mixing_store%rhoin(1)%cc(ig) - mixing_store%rhoin(2)%cc(ig)
326 : END DO
327 14 : IF (gapw .AND. mixing_store%gmix_p) THEN
328 0 : DO iatom = 1, natom
329 0 : IF (mixing_store%paw(iatom)) THEN
330 0 : rho_atom(iatom)%cpc_h(2)%r_coef = rho_atom(iatom)%cpc_h(1)%r_coef - rho_atom(iatom)%cpc_h(2)%r_coef
331 0 : rho_atom(iatom)%cpc_s(2)%r_coef = rho_atom(iatom)%cpc_s(1)%r_coef - rho_atom(iatom)%cpc_s(2)%r_coef
332 : mixing_store%cpc_h_in(iatom, 2)%r_coef = mixing_store%cpc_h_in(iatom, 1)%r_coef - &
333 0 : mixing_store%cpc_h_in(iatom, 2)%r_coef
334 : mixing_store%cpc_s_in(iatom, 2)%r_coef = mixing_store%cpc_s_in(iatom, 1)%r_coef - &
335 0 : mixing_store%cpc_s_in(iatom, 2)%r_coef
336 : END IF
337 : END DO
338 : END IF
339 : END IF
340 :
341 382 : DO ispin = 1, nspin
342 198 : ib = MODULO(mixing_store%ncall_p(ispin), nbuffer) + 1
343 :
344 198 : mixing_store%ncall_p(ispin) = mixing_store%ncall_p(ispin) + 1
345 198 : nb = MIN(mixing_store%ncall_p(ispin), nbuffer)
346 198 : ibb = MODULO(mixing_store%ncall_p(ispin), nbuffer) + 1
347 :
348 8641926 : mixing_store%res_buffer(ib, ispin)%cc(:) = CMPLX(0._dp, 0._dp, KIND=dp)
349 198 : norm = 0.0_dp
350 1306566 : IF (nb == 1) mixing_store%rhoin_buffer(1, ispin)%cc = mixing_store%rhoin(ispin)%cc
351 198 : res_norm = 0.0_dp
352 8641926 : DO ig = 1, ng
353 8641728 : f_mix = mixing_store%kerker_factor(ig)
354 : mixing_store%res_buffer(ib, ispin)%cc(ig) = f_mix*(rho_g(ispin)%array(ig) - &
355 8641728 : mixing_store%rhoin_buffer(ib, ispin)%cc(ig))
356 : res_norm = res_norm + &
357 : REAL(mixing_store%res_buffer(ib, ispin)%cc(ig), dp)*REAL(mixing_store%res_buffer(ib, ispin)%cc(ig), dp) + &
358 8641926 : AIMAG(mixing_store%res_buffer(ib, ispin)%cc(ig))*AIMAG(mixing_store%res_buffer(ib, ispin)%cc(ig))
359 : END DO
360 :
361 198 : CALL para_env%sum(res_norm)
362 198 : res_norm = SQRT(res_norm)
363 :
364 198 : IF (res_norm < 1.E-14_dp) THEN
365 0 : mixing_store%ncall_p(ispin) = mixing_store%ncall_p(ispin) - 1
366 : ELSE
367 198 : nb1 = nb + 1
368 792 : ALLOCATE (a(nb1, nb1))
369 5202 : a = 0.0_dp
370 594 : ALLOCATE (b(nb1, nb1))
371 5202 : b = 0.0_dp
372 792 : ALLOCATE (c(nb, nb))
373 3518 : c = 0.0_dp
374 594 : ALLOCATE (c_inv(nb, nb))
375 3518 : c_inv = 0.0_dp
376 594 : ALLOCATE (alpha_c(nb))
377 842 : alpha_c = 0.0_dp
378 198 : norm_c_inv = 0.0_dp
379 594 : ALLOCATE (ev(nb1))
380 1040 : ev = 0.0_dp
381 :
382 : END IF
383 :
384 842 : DO jb = 1, nb
385 644 : mixing_store%pulay_matrix(jb, ib, ispin) = 0.0_dp
386 33413252 : DO ig = 1, ng
387 33412608 : f_mix = mixing_store%special_metric(ig)
388 : mixing_store%pulay_matrix(jb, ib, ispin) = mixing_store%pulay_matrix(jb, ib, ispin) + &
389 : f_mix*(REAL(mixing_store%res_buffer(jb, ispin)%cc(ig), dp) &
390 : *REAL(mixing_store%res_buffer(ib, ispin)%cc(ig), dp) + &
391 : AIMAG(mixing_store%res_buffer(jb, ispin)%cc(ig))* &
392 33413252 : AIMAG(mixing_store%res_buffer(ib, ispin)%cc(ig)))
393 : END DO
394 644 : CALL para_env%sum(mixing_store%pulay_matrix(jb, ib, ispin))
395 644 : mixing_store%pulay_matrix(jb, ib, ispin) = mixing_store%pulay_matrix(jb, ib, ispin)*vol*vol
396 842 : mixing_store%pulay_matrix(ib, jb, ispin) = mixing_store%pulay_matrix(jb, ib, ispin)
397 : END DO
398 :
399 198 : IF (nb == 1 .OR. res_norm < 1.E-14_dp) THEN
400 1306406 : DO ig = 1, ng
401 1306368 : f_mix = alpha_kerker*mixing_store%kerker_factor(ig)
402 : cc_mix(ig) = rho_g(ispin)%array(ig) - &
403 1306368 : mixing_store%rhoin_buffer(ib, ispin)%cc(ig)
404 : rho_g(ispin)%array(ig) = f_mix*cc_mix(ig) + &
405 1306368 : mixing_store%rhoin_buffer(ib, ispin)%cc(ig)
406 1306406 : mixing_store%rhoin_buffer(ibb, ispin)%cc(ig) = rho_g(ispin)%array(ig)
407 : END DO
408 38 : IF (mixing_store%gmix_p) THEN
409 2 : IF (gapw) THEN
410 18 : DO iatom = 1, natom
411 18 : IF (mixing_store%paw(iatom)) THEN
412 : mixing_store%cpc_h_res_buffer(ib, iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef - &
413 1850 : mixing_store%cpc_h_in(iatom, ispin)%r_coef
414 : mixing_store%cpc_s_res_buffer(ib, iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef - &
415 1850 : mixing_store%cpc_s_in(iatom, ispin)%r_coef
416 :
417 : rho_atom(iatom)%cpc_h(ispin)%r_coef = alpha_kerker*rho_atom(iatom)%cpc_h(ispin)%r_coef + &
418 1850 : (1.0_dp - alpha_kerker)*mixing_store%cpc_h_in(iatom, ispin)%r_coef
419 : rho_atom(iatom)%cpc_s(ispin)%r_coef = alpha_kerker*rho_atom(iatom)%cpc_s(ispin)%r_coef + &
420 1850 : (1.0_dp - alpha_kerker)*mixing_store%cpc_s_in(iatom, ispin)%r_coef
421 :
422 926 : mixing_store%cpc_h_in_buffer(ib, iatom, ispin)%r_coef = mixing_store%cpc_h_in(iatom, ispin)%r_coef
423 926 : mixing_store%cpc_s_in_buffer(ib, iatom, ispin)%r_coef = mixing_store%cpc_s_in(iatom, ispin)%r_coef
424 :
425 1850 : mixing_store%cpc_h_in_buffer(ibb, iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef
426 1850 : mixing_store%cpc_s_in_buffer(ibb, iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef
427 : END IF
428 : END DO
429 : END IF
430 : END IF
431 : ELSE
432 :
433 3404 : b(1:nb, 1:nb) = mixing_store%pulay_matrix(1:nb, 1:nb, ispin)
434 3404 : c(1:nb, 1:nb) = b(1:nb, 1:nb)
435 766 : b(nb1, 1:nb) = -1.0_dp
436 766 : b(1:nb, nb1) = -1.0_dp
437 160 : b(nb1, nb1) = 0.0_dp
438 :
439 160 : CALL invert_matrix(c, c_inv, inv_err, improve=.TRUE.)
440 766 : alpha_c = 0.0_dp
441 766 : DO i = 1, nb
442 3404 : DO jb = 1, nb
443 2638 : alpha_c(i) = alpha_c(i) + c_inv(jb, i)
444 3244 : norm_c_inv = norm_c_inv + c_inv(jb, i)
445 : END DO
446 : END DO
447 766 : alpha_c(1:nb) = alpha_c(1:nb)/norm_c_inv
448 7335520 : cc_mix = CMPLX(0._dp, 0._dp, KIND=dp)
449 766 : DO jb = 1, nb
450 32107006 : DO ig = 1, ng
451 : cc_mix(ig) = cc_mix(ig) + &
452 : alpha_c(jb)*(mixing_store%rhoin_buffer(jb, ispin)%cc(ig) + &
453 32106846 : mixing_store%pulay_beta*mixing_store%res_buffer(jb, ispin)%cc(ig))
454 : END DO
455 : END DO
456 7335520 : mixing_store%rhoin_buffer(ibb, ispin)%cc = CMPLX(0._dp, 0._dp, KIND=dp)
457 160 : IF (alpha_pulay > 0.0_dp) THEN
458 233290 : DO ig = 1, ng
459 233280 : f_mix = alpha_pulay*mixing_store%kerker_factor(ig)
460 : rho_g(ispin)%array(ig) = f_mix*rho_g(ispin)%array(ig) + &
461 233280 : (1.0_dp - f_mix)*cc_mix(ig)
462 233290 : mixing_store%rhoin_buffer(ibb, ispin)%cc(ig) = rho_g(ispin)%array(ig)
463 : END DO
464 : ELSE
465 7102230 : DO ig = 1, ng
466 7102080 : rho_g(ispin)%array(ig) = cc_mix(ig)
467 7102230 : mixing_store%rhoin_buffer(ibb, ispin)%cc(ig) = rho_g(ispin)%array(ig)
468 : END DO
469 : END IF
470 :
471 160 : IF (mixing_store%gmix_p .AND. gapw) THEN
472 10 : CALL get_qs_env(qs_env=qs_env, rho_atom_set=rho_atom)
473 90 : DO iatom = 1, natom
474 90 : IF (mixing_store%paw(iatom)) THEN
475 10 : n1 = SIZE(rho_atom(iatom)%cpc_h(ispin)%r_coef, 1)
476 10 : n2 = SIZE(rho_atom(iatom)%cpc_h(ispin)%r_coef, 2)
477 40 : ALLOCATE (cpc_h_mix(n1, n2))
478 30 : ALLOCATE (cpc_s_mix(n1, n2))
479 :
480 : mixing_store%cpc_h_res_buffer(ib, iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef - &
481 9250 : mixing_store%cpc_h_in_buffer(ib, iatom, ispin)%r_coef
482 : mixing_store%cpc_s_res_buffer(ib, iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef - &
483 9250 : mixing_store%cpc_s_in_buffer(ib, iatom, ispin)%r_coef
484 4630 : cpc_h_mix = 0.0_dp
485 4630 : cpc_s_mix = 0.0_dp
486 48 : DO jb = 1, nb
487 : cpc_h_mix(:, :) = cpc_h_mix(:, :) + &
488 : alpha_c(jb)*mixing_store%cpc_h_in_buffer(jb, iatom, ispin)%r_coef(:, :) + &
489 17594 : alpha_c(jb)*beta*mixing_store%cpc_h_res_buffer(jb, iatom, ispin)%r_coef(:, :)
490 : cpc_s_mix(:, :) = cpc_s_mix(:, :) + &
491 : alpha_c(jb)*mixing_store%cpc_s_in_buffer(jb, iatom, ispin)%r_coef(:, :) + &
492 17604 : alpha_c(jb)*beta*mixing_store%cpc_s_res_buffer(jb, iatom, ispin)%r_coef(:, :)
493 : END DO
494 : rho_atom(iatom)%cpc_h(ispin)%r_coef = alpha_pulay*rho_atom(iatom)%cpc_h(ispin)%r_coef + &
495 4630 : (1.0_dp - alpha_pulay)*cpc_h_mix
496 : rho_atom(iatom)%cpc_s(ispin)%r_coef = alpha_pulay*rho_atom(iatom)%cpc_s(ispin)%r_coef + &
497 4630 : (1.0_dp - alpha_pulay)*cpc_s_mix
498 9250 : mixing_store%cpc_h_in_buffer(ibb, iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef
499 9250 : mixing_store%cpc_s_in_buffer(ibb, iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef
500 10 : DEALLOCATE (cpc_h_mix)
501 10 : DEALLOCATE (cpc_s_mix)
502 : END IF
503 : END DO
504 : END IF
505 : END IF ! nb==1
506 :
507 382 : IF (res_norm > 1.E-14_dp) THEN
508 198 : DEALLOCATE (a)
509 198 : DEALLOCATE (b)
510 198 : DEALLOCATE (ev)
511 198 : DEALLOCATE (c, c_inv, alpha_c)
512 : END IF
513 :
514 : END DO ! ispin
515 :
516 184 : DEALLOCATE (cc_mix)
517 :
518 184 : IF (nspin == 2) THEN
519 14 : CALL pw_axpy(rho_g(1), rho_g(2), 1.0_dp, -1.0_dp)
520 96782 : DO ig = 1, ng
521 96782 : mixing_store%rhoin(2)%cc(ig) = mixing_store%rhoin(1)%cc(ig) - mixing_store%rhoin(2)%cc(ig)
522 : END DO
523 14 : IF (gapw .AND. mixing_store%gmix_p) THEN
524 0 : DO iatom = 1, natom
525 0 : IF (mixing_store%paw(iatom)) THEN
526 0 : rho_atom(iatom)%cpc_h(2)%r_coef = rho_atom(iatom)%cpc_h(1)%r_coef - rho_atom(iatom)%cpc_h(2)%r_coef
527 0 : rho_atom(iatom)%cpc_s(2)%r_coef = rho_atom(iatom)%cpc_s(1)%r_coef - rho_atom(iatom)%cpc_s(2)%r_coef
528 : mixing_store%cpc_h_in(iatom, 2)%r_coef = mixing_store%cpc_h_in(iatom, 1)%r_coef - &
529 0 : mixing_store%cpc_h_in(iatom, 2)%r_coef
530 : mixing_store%cpc_s_in(iatom, 2)%r_coef = mixing_store%cpc_s_in(iatom, 1)%r_coef - &
531 0 : mixing_store%cpc_s_in(iatom, 2)%r_coef
532 : END IF
533 : END DO
534 : END IF
535 :
536 : END IF
537 :
538 184 : CALL timestop(handle)
539 :
540 552 : END SUBROUTINE pulay_mixing
541 :
542 : ! **************************************************************************************************
543 : !> \brief Broyden Mixing using as metrics for the residual the Kerer damping factor
544 : !> The mixing is applied directly on the density expansion in reciprocal space
545 : !> \param qs_env ...
546 : !> \param mixing_store ...
547 : !> \param rho ...
548 : !> \param para_env ...
549 : !> \par History
550 : !> 03.2009
551 : !> \author MI
552 : ! **************************************************************************************************
553 :
554 794 : SUBROUTINE broyden_mixing(qs_env, mixing_store, rho, para_env)
555 :
556 : TYPE(qs_environment_type), POINTER :: qs_env
557 : TYPE(mixing_storage_type), POINTER :: mixing_store
558 : TYPE(qs_rho_type), POINTER :: rho
559 : TYPE(mp_para_env_type), POINTER :: para_env
560 :
561 : CHARACTER(len=*), PARAMETER :: routineN = 'broyden_mixing'
562 :
563 : COMPLEX(dp) :: cc_mix
564 794 : COMPLEX(dp), ALLOCATABLE, DIMENSION(:) :: res_rho
565 : INTEGER :: handle, i, iatom, ib, ig, ispin, j, jb, &
566 : kb, n1, n2, natom, nb, nbuffer, ng, &
567 : nspin
568 : LOGICAL :: gapw, only_kerker
569 : REAL(dp) :: alpha, dcpc_h_res, dcpc_s_res, &
570 : delta_norm, f_mix, inv_err, res_norm, &
571 : rho_norm, valh, vals, w0
572 794 : REAL(dp), ALLOCATABLE, DIMENSION(:) :: c, g
573 794 : REAL(dp), ALLOCATABLE, DIMENSION(:, :) :: a, b
574 : TYPE(dft_control_type), POINTER :: dft_control
575 794 : TYPE(pw_c1d_gs_type), DIMENSION(:), POINTER :: rho_g
576 794 : TYPE(rho_atom_type), DIMENSION(:), POINTER :: rho_atom
577 :
578 0 : CPASSERT(ASSOCIATED(mixing_store%res_buffer))
579 794 : CPASSERT(ASSOCIATED(mixing_store%rhoin))
580 794 : CPASSERT(ASSOCIATED(mixing_store%rhoin_old))
581 794 : CPASSERT(ASSOCIATED(mixing_store%drho_buffer))
582 794 : NULLIFY (dft_control, rho_atom, rho_g)
583 794 : CALL timeset(routineN, handle)
584 :
585 794 : CALL get_qs_env(qs_env, dft_control=dft_control)
586 794 : CALL qs_rho_get(rho, rho_g=rho_g)
587 :
588 794 : nspin = SIZE(rho_g, 1)
589 794 : ng = SIZE(mixing_store%res_buffer(1, 1)%cc)
590 :
591 794 : alpha = mixing_store%alpha
592 794 : w0 = mixing_store%broy_w0
593 794 : nbuffer = mixing_store%nbuffer
594 794 : gapw = dft_control%qs_control%gapw
595 :
596 2382 : ALLOCATE (res_rho(ng))
597 :
598 794 : mixing_store%ncall = mixing_store%ncall + 1
599 794 : IF (mixing_store%ncall == 1) THEN
600 : only_kerker = .TRUE.
601 : ELSE
602 650 : only_kerker = .FALSE.
603 650 : nb = MIN(mixing_store%ncall - 1, nbuffer)
604 650 : ib = MODULO(mixing_store%ncall - 2, nbuffer) + 1
605 : END IF
606 :
607 794 : IF (gapw) THEN
608 102 : CALL get_qs_env(qs_env=qs_env, rho_atom_set=rho_atom)
609 102 : natom = SIZE(rho_atom)
610 : ELSE
611 : natom = 0
612 : END IF
613 :
614 794 : IF (nspin == 2) THEN
615 100 : CALL pw_axpy(rho_g(1), rho_g(2), 1.0_dp, -1.0_dp)
616 9659620 : DO ig = 1, ng
617 9659620 : mixing_store%rhoin(2)%cc(ig) = mixing_store%rhoin(1)%cc(ig) - mixing_store%rhoin(2)%cc(ig)
618 : END DO
619 100 : IF (gapw .AND. mixing_store%gmix_p) THEN
620 90 : DO iatom = 1, natom
621 90 : IF (mixing_store%paw(iatom)) THEN
622 60560 : rho_atom(iatom)%cpc_h(2)%r_coef = rho_atom(iatom)%cpc_h(1)%r_coef - rho_atom(iatom)%cpc_h(2)%r_coef
623 60560 : rho_atom(iatom)%cpc_s(2)%r_coef = rho_atom(iatom)%cpc_s(1)%r_coef - rho_atom(iatom)%cpc_s(2)%r_coef
624 : mixing_store%cpc_h_in(iatom, 2)%r_coef = mixing_store%cpc_h_in(iatom, 1)%r_coef - &
625 60560 : mixing_store%cpc_h_in(iatom, 2)%r_coef
626 : mixing_store%cpc_s_in(iatom, 2)%r_coef = mixing_store%cpc_s_in(iatom, 1)%r_coef - &
627 60560 : mixing_store%cpc_s_in(iatom, 2)%r_coef
628 : END IF
629 : END DO
630 : END IF
631 : END IF
632 :
633 1688 : DO ispin = 1, nspin
634 68680020 : res_rho = CMPLX(0.0_dp, 0.0_dp, KIND=dp)
635 68680020 : DO ig = 1, ng
636 68680020 : res_rho(ig) = rho_g(ispin)%array(ig) - mixing_store%rhoin(ispin)%cc(ig)
637 : END DO
638 :
639 1688 : IF (only_kerker) THEN
640 8257236 : DO ig = 1, ng
641 8257062 : mixing_store%last_res(ispin)%cc(ig) = res_rho(ig)
642 8257062 : f_mix = alpha*mixing_store%kerker_factor(ig)
643 8257062 : rho_g(ispin)%array(ig) = mixing_store%rhoin(ispin)%cc(ig) + f_mix*res_rho(ig)
644 8257062 : mixing_store%rhoin_old(ispin)%cc(ig) = mixing_store%rhoin(ispin)%cc(ig)
645 8257236 : mixing_store%rhoin(ispin)%cc(ig) = rho_g(ispin)%array(ig)
646 : END DO
647 :
648 174 : IF (mixing_store%gmix_p) THEN
649 24 : IF (gapw) THEN
650 216 : DO iatom = 1, natom
651 216 : IF (mixing_store%paw(iatom)) THEN
652 : mixing_store%cpc_h_lastres(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef - &
653 245780 : mixing_store%cpc_h_in(iatom, ispin)%r_coef
654 : mixing_store%cpc_s_lastres(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef - &
655 245780 : mixing_store%cpc_s_in(iatom, ispin)%r_coef
656 :
657 : rho_atom(iatom)%cpc_h(ispin)%r_coef = alpha*rho_atom(iatom)%cpc_h(ispin)%r_coef + &
658 245780 : mixing_store%cpc_h_in(iatom, ispin)%r_coef*(1._dp - alpha)
659 : rho_atom(iatom)%cpc_s(ispin)%r_coef = alpha*rho_atom(iatom)%cpc_s(ispin)%r_coef + &
660 245780 : mixing_store%cpc_s_in(iatom, ispin)%r_coef*(1._dp - alpha)
661 :
662 122972 : mixing_store%cpc_h_old(iatom, ispin)%r_coef = mixing_store%cpc_h_in(iatom, ispin)%r_coef
663 122972 : mixing_store%cpc_s_old(iatom, ispin)%r_coef = mixing_store%cpc_s_in(iatom, ispin)%r_coef
664 245780 : mixing_store%cpc_h_in(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef
665 245780 : mixing_store%cpc_s_in(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef
666 : END IF
667 : END DO
668 : END IF
669 : END IF
670 : ELSE
671 :
672 2880 : ALLOCATE (a(nb, nb))
673 20148 : a = 0.0_dp
674 2160 : ALLOCATE (b(nb, nb))
675 20148 : b = 0.0_dp
676 2160 : ALLOCATE (c(nb))
677 3528 : c = 0.0_dp
678 1440 : ALLOCATE (g(nb))
679 3528 : g = 0.0_dp
680 :
681 720 : delta_norm = 0.0_dp
682 720 : res_norm = 0.0_dp
683 720 : rho_norm = 0.0_dp
684 60422784 : DO ig = 1, ng
685 60422064 : mixing_store%res_buffer(ib, ispin)%cc(ig) = res_rho(ig) - mixing_store%last_res(ispin)%cc(ig)
686 60422064 : mixing_store%last_res(ispin)%cc(ig) = res_rho(ig)
687 : res_norm = res_norm + &
688 : REAL(res_rho(ig), dp)*REAL(res_rho(ig), dp) + &
689 60422064 : AIMAG(res_rho(ig))*AIMAG(res_rho(ig))
690 : delta_norm = delta_norm + &
691 : REAL(mixing_store%res_buffer(ib, ispin)%cc(ig), dp)* &
692 : REAL(mixing_store%res_buffer(ib, ispin)%cc(ig), dp) + &
693 : AIMAG(mixing_store%res_buffer(ib, ispin)%cc(ig))* &
694 60422064 : AIMAG(mixing_store%res_buffer(ib, ispin)%cc(ig))
695 : rho_norm = rho_norm + &
696 : REAL(rho_g(ispin)%array(ig), dp)*REAL(rho_g(ispin)%array(ig), dp) + &
697 60422784 : AIMAG(rho_g(ispin)%array(ig))*AIMAG(rho_g(ispin)%array(ig))
698 : END DO
699 60422784 : DO ig = 1, ng
700 : mixing_store%drho_buffer(ib, ispin)%cc(ig) = &
701 : mixing_store%rhoin(ispin)%cc(ig) - &
702 60422784 : mixing_store%rhoin_old(ispin)%cc(ig)
703 : END DO
704 720 : CALL para_env%sum(delta_norm)
705 720 : delta_norm = SQRT(delta_norm)
706 720 : CALL para_env%sum(res_norm)
707 720 : res_norm = SQRT(res_norm)
708 720 : CALL para_env%sum(rho_norm)
709 720 : rho_norm = SQRT(rho_norm)
710 :
711 720 : IF (res_norm > 1.E-14_dp) THEN
712 60422784 : mixing_store%res_buffer(ib, ispin)%cc(:) = mixing_store%res_buffer(ib, ispin)%cc(:)/delta_norm
713 60422784 : mixing_store%drho_buffer(ib, ispin)%cc(:) = mixing_store%drho_buffer(ib, ispin)%cc(:)/delta_norm
714 :
715 720 : IF (mixing_store%gmix_p .AND. gapw) THEN
716 252 : DO iatom = 1, natom
717 252 : IF (mixing_store%paw(iatom)) THEN
718 28 : n1 = SIZE(mixing_store%cpc_s_in(iatom, ispin)%r_coef, 1)
719 28 : n2 = SIZE(mixing_store%cpc_s_in(iatom, ispin)%r_coef, 2)
720 616 : DO i = 1, n2
721 12964 : DO j = 1, n1
722 : mixing_store%dcpc_h_in(ib, iatom, ispin)%r_coef(j, i) = &
723 : (mixing_store%cpc_h_in(iatom, ispin)%r_coef(j, i) - &
724 12348 : mixing_store%cpc_h_old(iatom, ispin)%r_coef(j, i))/delta_norm
725 : dcpc_h_res = ((rho_atom(iatom)%cpc_h(ispin)%r_coef(j, i) - &
726 : mixing_store%cpc_h_in(iatom, ispin)%r_coef(j, i)) - &
727 12348 : mixing_store%cpc_h_lastres(iatom, ispin)%r_coef(j, i))/delta_norm
728 : mixing_store%cpc_h_lastres(iatom, ispin)%r_coef(j, i) = rho_atom(iatom)%cpc_h(ispin)%r_coef(j, i) - &
729 12348 : mixing_store%cpc_h_in(iatom, ispin)%r_coef(j, i)
730 :
731 : mixing_store%dcpc_s_in(ib, iatom, ispin)%r_coef(j, i) = &
732 : (mixing_store%cpc_s_in(iatom, ispin)%r_coef(j, i) - &
733 12348 : mixing_store%cpc_s_old(iatom, ispin)%r_coef(j, i))/delta_norm
734 : dcpc_s_res = ((rho_atom(iatom)%cpc_s(ispin)%r_coef(j, i) - &
735 : mixing_store%cpc_s_in(iatom, ispin)%r_coef(j, i)) - &
736 12348 : mixing_store%cpc_s_lastres(iatom, ispin)%r_coef(j, i))/delta_norm
737 : mixing_store%cpc_s_lastres(iatom, ispin)%r_coef(j, i) = rho_atom(iatom)%cpc_s(ispin)%r_coef(j, i) - &
738 12348 : mixing_store%cpc_s_in(iatom, ispin)%r_coef(j, i)
739 :
740 : mixing_store%dcpc_h_in(ib, iatom, ispin)%r_coef(j, i) = &
741 : alpha*dcpc_h_res + &
742 12348 : mixing_store%dcpc_h_in(ib, iatom, ispin)%r_coef(j, i)
743 : mixing_store%dcpc_s_in(ib, iatom, ispin)%r_coef(j, i) = &
744 : alpha*dcpc_s_res + &
745 12936 : mixing_store%dcpc_s_in(ib, iatom, ispin)%r_coef(j, i)
746 : END DO
747 : END DO
748 : END IF
749 : END DO
750 : END IF
751 :
752 20148 : a(:, :) = 0.0_dp
753 60422784 : DO ig = 1, ng
754 60422064 : f_mix = alpha*mixing_store%kerker_factor(ig)
755 : mixing_store%drho_buffer(ib, ispin)%cc(ig) = &
756 : f_mix*mixing_store%res_buffer(ib, ispin)%cc(ig) + &
757 60422784 : mixing_store%drho_buffer(ib, ispin)%cc(ig)
758 : END DO
759 3528 : DO jb = 1, nb
760 13242 : DO kb = jb, nb
761 1475240898 : DO ig = 1, mixing_store%ig_max !ng
762 : a(kb, jb) = a(kb, jb) + mixing_store%p_metric(ig)*( &
763 : REAL(mixing_store%res_buffer(jb, ispin)%cc(ig), dp)* &
764 : REAL(mixing_store%res_buffer(kb, ispin)%cc(ig), dp) + &
765 : AIMAG(mixing_store%res_buffer(jb, ispin)%cc(ig))* &
766 1475240898 : AIMAG(mixing_store%res_buffer(kb, ispin)%cc(ig)))
767 : END DO
768 12522 : a(jb, kb) = a(kb, jb)
769 : END DO
770 : END DO
771 720 : CALL para_env%sum(a)
772 :
773 3528 : C = 0.0_dp
774 3528 : DO jb = 1, nb
775 2808 : a(jb, jb) = w0 + a(jb, jb)
776 326542848 : DO ig = 1, mixing_store%ig_max !ng
777 : c(jb) = c(jb) + mixing_store%p_metric(ig)*( &
778 : REAL(mixing_store%res_buffer(jb, ispin)%cc(ig), dp)*REAL(res_rho(ig), dp) + &
779 326542128 : AIMAG(mixing_store%res_buffer(jb, ispin)%cc(ig))*AIMAG(res_rho(ig)))
780 : END DO
781 : END DO
782 720 : CALL para_env%sum(c)
783 720 : CALL invert_matrix(a, b, inv_err)
784 :
785 720 : CALL dgemv('T', nb, nb, 1.0_dp, B, nb, C, 1, 0.0_dp, G, 1)
786 : ELSE
787 0 : G = 0.0_dp
788 : END IF
789 :
790 60422784 : DO ig = 1, ng
791 60422064 : cc_mix = CMPLX(0.0_dp, 0.0_dp, kind=dp)
792 386961384 : DO jb = 1, nb
793 386961384 : cc_mix = cc_mix - G(jb)*mixing_store%drho_buffer(jb, ispin)%cc(ig)
794 : END DO
795 60422064 : f_mix = alpha*mixing_store%kerker_factor(ig)
796 :
797 60422064 : IF (res_norm > 1.E-14_dp) rho_g(ispin)%array(ig) = mixing_store%rhoin(ispin)%cc(ig) + &
798 60422064 : f_mix*res_rho(ig) + cc_mix
799 60422064 : mixing_store%rhoin_old(ispin)%cc(ig) = mixing_store%rhoin(ispin)%cc(ig)
800 60422784 : mixing_store%rhoin(ispin)%cc(ig) = rho_g(ispin)%array(ig)
801 : END DO
802 :
803 720 : IF (mixing_store%gmix_p) THEN
804 28 : IF (gapw) THEN
805 252 : DO iatom = 1, natom
806 252 : IF (mixing_store%paw(iatom)) THEN
807 28 : n1 = SIZE(mixing_store%cpc_s_in(iatom, ispin)%r_coef, 1)
808 28 : n2 = SIZE(mixing_store%cpc_s_in(iatom, ispin)%r_coef, 2)
809 616 : DO i = 1, n2
810 12964 : DO j = 1, n1
811 12348 : valh = 0.0_dp
812 12348 : vals = 0.0_dp
813 61740 : DO jb = 1, nb
814 49392 : valh = valh - G(jb)*mixing_store%dcpc_h_in(jb, iatom, ispin)%r_coef(j, i)
815 61740 : vals = vals - G(jb)*mixing_store%dcpc_s_in(jb, iatom, ispin)%r_coef(j, i)
816 : END DO
817 12936 : IF (res_norm > 1.E-14_dp) THEN
818 : rho_atom(iatom)%cpc_h(ispin)%r_coef(j, i) = &
819 : alpha*rho_atom(iatom)%cpc_h(ispin)%r_coef(j, i) + &
820 12348 : mixing_store%cpc_h_in(iatom, ispin)%r_coef(j, i)*(1._dp - alpha) + valh
821 : rho_atom(iatom)%cpc_s(ispin)%r_coef(j, i) = &
822 : alpha*rho_atom(iatom)%cpc_s(ispin)%r_coef(j, i) + &
823 12348 : mixing_store%cpc_s_in(iatom, ispin)%r_coef(j, i)*(1._dp - alpha) + vals
824 : END IF
825 : END DO
826 : END DO
827 :
828 12964 : mixing_store%cpc_h_old(iatom, ispin)%r_coef = mixing_store%cpc_h_in(iatom, ispin)%r_coef
829 12964 : mixing_store%cpc_s_old(iatom, ispin)%r_coef = mixing_store%cpc_s_in(iatom, ispin)%r_coef
830 25900 : mixing_store%cpc_h_in(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_h(ispin)%r_coef
831 25900 : mixing_store%cpc_s_in(iatom, ispin)%r_coef = rho_atom(iatom)%cpc_s(ispin)%r_coef
832 : END IF
833 : END DO
834 : END IF
835 : END IF
836 :
837 720 : DEALLOCATE (a, b, c, g)
838 : END IF
839 :
840 : END DO ! ispin
841 794 : IF (nspin == 2) THEN
842 100 : CALL pw_axpy(rho_g(1), rho_g(2), 1.0_dp, -1.0_dp)
843 9659620 : DO ig = 1, ng
844 9659620 : mixing_store%rhoin(2)%cc(ig) = mixing_store%rhoin(1)%cc(ig) - mixing_store%rhoin(2)%cc(ig)
845 : END DO
846 100 : IF (gapw .AND. mixing_store%gmix_p) THEN
847 90 : DO iatom = 1, natom
848 90 : IF (mixing_store%paw(iatom)) THEN
849 60560 : rho_atom(iatom)%cpc_h(2)%r_coef = rho_atom(iatom)%cpc_h(1)%r_coef - rho_atom(iatom)%cpc_h(2)%r_coef
850 60560 : rho_atom(iatom)%cpc_s(2)%r_coef = rho_atom(iatom)%cpc_s(1)%r_coef - rho_atom(iatom)%cpc_s(2)%r_coef
851 : mixing_store%cpc_h_in(iatom, 2)%r_coef = mixing_store%cpc_h_in(iatom, 1)%r_coef - &
852 60560 : mixing_store%cpc_h_in(iatom, 2)%r_coef
853 : mixing_store%cpc_s_in(iatom, 2)%r_coef = mixing_store%cpc_s_in(iatom, 1)%r_coef - &
854 60560 : mixing_store%cpc_s_in(iatom, 2)%r_coef
855 : END IF
856 : END DO
857 : END IF
858 :
859 : END IF
860 :
861 794 : DEALLOCATE (res_rho)
862 :
863 794 : CALL timestop(handle)
864 :
865 2382 : END SUBROUTINE broyden_mixing
866 :
867 : ! **************************************************************************************************
868 : !> \brief Multisecant scheme to perform charge density Mixing
869 : !> as preconditioner we use the Kerker damping factor
870 : !> The mixing is applied directly on the density in reciprocal space,
871 : !> therefore it affects the potentials
872 : !> on the grid but not the density matrix
873 : !> \param mixing_store ...
874 : !> \param rho ...
875 : !> \param para_env ...
876 : !> \par History
877 : !> 05.2009
878 : !> \author MI
879 : ! **************************************************************************************************
880 0 : SUBROUTINE multisecant_mixing(mixing_store, rho, para_env)
881 :
882 : TYPE(mixing_storage_type), POINTER :: mixing_store
883 : TYPE(qs_rho_type), POINTER :: rho
884 : TYPE(mp_para_env_type), POINTER :: para_env
885 :
886 : CHARACTER(len=*), PARAMETER :: routineN = 'multisecant_mixing'
887 : COMPLEX(KIND=dp), PARAMETER :: cmone = (-1.0_dp, 0.0_dp), &
888 : cone = (1.0_dp, 0.0_dp), &
889 : czero = (0.0_dp, 0.0_dp)
890 :
891 : COMPLEX(dp) :: saa, yaa
892 0 : COMPLEX(dp), ALLOCATABLE, DIMENSION(:) :: gn_global, pgn, res_matrix_global, &
893 0 : tmp_vec, ugn
894 0 : COMPLEX(dp), ALLOCATABLE, DIMENSION(:, :) :: a_matrix, res_matrix, sa, step_matrix, ya
895 0 : COMPLEX(dp), DIMENSION(:), POINTER :: gn
896 : INTEGER :: handle, ib, ib_next, ib_prev, ig, &
897 : ig_global, iig, ispin, jb, kb, nb, &
898 : nbuffer, ng, ng_global, nspin
899 : LOGICAL :: use_zgemm, use_zgemm_rev
900 : REAL(dp) :: alpha, f_mix, gn_norm, gn_norm_old, inv_err, n_low, n_up, pgn_norm, prec, &
901 : r_step, reg_par, sigma_max, sigma_tilde, step_size
902 0 : REAL(dp), ALLOCATABLE, DIMENSION(:) :: norm_res, norm_res_low, norm_res_up
903 0 : REAL(dp), ALLOCATABLE, DIMENSION(:, :) :: b_matrix, binv_matrix
904 0 : REAL(dp), DIMENSION(:), POINTER :: g2
905 : REAL(dp), SAVE :: sigma_old = 1.0_dp
906 0 : TYPE(pw_c1d_gs_type), DIMENSION(:), POINTER :: rho_g
907 :
908 0 : CPASSERT(ASSOCIATED(mixing_store))
909 :
910 0 : CALL timeset(routineN, handle)
911 :
912 0 : NULLIFY (gn, rho_g)
913 :
914 0 : use_zgemm = .FALSE.
915 0 : use_zgemm_rev = .TRUE.
916 :
917 : ! prepare the parameters
918 0 : CALL qs_rho_get(rho, rho_g=rho_g)
919 :
920 0 : nspin = SIZE(rho_g, 1)
921 : ! not implemented for large grids.
922 0 : CPASSERT(rho_g(1)%pw_grid%ngpts < HUGE(ng_global))
923 0 : ng_global = INT(rho_g(1)%pw_grid%ngpts)
924 0 : ng = SIZE(mixing_store%rhoin_buffer(1, 1)%cc)
925 0 : alpha = mixing_store%alpha
926 :
927 0 : sigma_max = mixing_store%sigma_max
928 0 : reg_par = mixing_store%reg_par
929 0 : r_step = mixing_store%r_step
930 0 : nbuffer = mixing_store%nbuffer
931 :
932 : ! determine the step number, and multisecant iteration
933 0 : nb = MIN(mixing_store%ncall, nbuffer - 1)
934 0 : ib = MODULO(mixing_store%ncall, nbuffer) + 1
935 0 : IF (mixing_store%ncall > 0) THEN
936 0 : ib_prev = MODULO(mixing_store%ncall - 1, nbuffer) + 1
937 : ELSE
938 : ib_prev = 0
939 : END IF
940 0 : mixing_store%ncall = mixing_store%ncall + 1
941 0 : ib_next = MODULO(mixing_store%ncall, nbuffer) + 1
942 :
943 : ! compute the residual gn and its norm gn_norm
944 0 : DO ispin = 1, nspin
945 0 : gn => mixing_store%res_buffer(ib, ispin)%cc
946 0 : gn_norm = 0.0_dp
947 0 : DO ig = 1, ng
948 0 : gn(ig) = (rho_g(ispin)%array(ig) - mixing_store%rhoin_buffer(ib, ispin)%cc(ig))
949 : gn_norm = gn_norm + &
950 0 : REAL(gn(ig), dp)*REAL(gn(ig), dp) + AIMAG(gn(ig))*AIMAG(gn(ig))
951 : END DO
952 0 : CALL para_env%sum(gn_norm)
953 0 : gn_norm = SQRT(gn_norm)
954 0 : mixing_store%norm_res_buffer(ib, ispin) = gn_norm
955 : END DO
956 :
957 0 : IF (nb == 0) THEN
958 : !simple mixing
959 0 : DO ispin = 1, nspin
960 0 : DO ig = 1, ng
961 0 : f_mix = alpha*mixing_store%kerker_factor(ig)
962 : rho_g(ispin)%array(ig) = mixing_store%rhoin_buffer(1, ispin)%cc(ig) + &
963 0 : f_mix*mixing_store%res_buffer(1, ispin)%cc(ig)
964 0 : mixing_store%rhoin_buffer(ib_next, ispin)%cc(ig) = rho_g(ispin)%array(ig)
965 : END DO
966 : END DO
967 0 : CALL timestop(handle)
968 0 : RETURN
969 : END IF
970 :
971 : ! allocate temporary arrays
972 : ! step_matrix S ngxnb
973 0 : ALLOCATE (step_matrix(ng, nb))
974 : ! res_matrix Y ngxnb
975 0 : ALLOCATE (res_matrix(ng, nb))
976 : ! matrix A nbxnb
977 0 : ALLOCATE (a_matrix(nb, ng_global))
978 : ! PSI nb vector of norms
979 0 : ALLOCATE (norm_res(nb))
980 0 : ALLOCATE (norm_res_low(nb))
981 0 : ALLOCATE (norm_res_up(nb))
982 :
983 : ! matrix B nbxnb
984 0 : ALLOCATE (b_matrix(nb, nb))
985 : ! matrix B_inv nbxnb
986 0 : ALLOCATE (binv_matrix(nb, nb))
987 :
988 0 : ALLOCATE (gn_global(ng_global))
989 0 : ALLOCATE (res_matrix_global(ng_global))
990 : IF (use_zgemm) THEN
991 : ALLOCATE (sa(ng, ng_global))
992 : ALLOCATE (ya(ng, ng_global))
993 : END IF
994 : IF (use_zgemm_rev) THEN
995 0 : ALLOCATE (tmp_vec(nb))
996 : END IF
997 0 : ALLOCATE (pgn(ng))
998 0 : ALLOCATE (ugn(ng))
999 :
1000 0 : DO ispin = 1, nspin
1001 : ! generate the global vector with the present residual
1002 0 : gn => mixing_store%res_buffer(ib, ispin)%cc
1003 0 : gn_global = CMPLX(0.0_dp, 0.0_dp, KIND=dp)
1004 0 : DO ig = 1, ng
1005 0 : ig_global = mixing_store%ig_global_index(ig)
1006 0 : gn_global(ig_global) = gn(ig)
1007 : END DO
1008 0 : CALL para_env%sum(gn_global)
1009 :
1010 : ! compute steps (matrix S) and residual differences (matrix Y)
1011 : ! with respect to the present
1012 : ! step and the present residual (use stored rho_in and res_buffer)
1013 :
1014 : ! These quantities are pre-conditioned by means of Kerker factor multipliccation
1015 0 : g2 => rho_g(1)%pw_grid%gsq
1016 :
1017 0 : DO jb = 1, nb + 1
1018 0 : IF (jb < ib) THEN
1019 : kb = jb
1020 0 : ELSEIF (jb > ib) THEN
1021 0 : kb = jb - 1
1022 : ELSE
1023 : CYCLE
1024 : END IF
1025 0 : norm_res(kb) = 0.0_dp
1026 0 : norm_res_low(kb) = 0.0_dp
1027 0 : norm_res_up(kb) = 0.0_dp
1028 0 : DO ig = 1, ng
1029 :
1030 0 : prec = 1.0_dp
1031 :
1032 : step_matrix(ig, kb) = prec*(mixing_store%rhoin_buffer(jb, ispin)%cc(ig) - &
1033 0 : mixing_store%rhoin_buffer(ib, ispin)%cc(ig))
1034 : res_matrix(ig, kb) = (mixing_store%res_buffer(jb, ispin)%cc(ig) - &
1035 0 : mixing_store%res_buffer(ib, ispin)%cc(ig))
1036 : norm_res(kb) = norm_res(kb) + REAL(res_matrix(ig, kb), dp)*REAL(res_matrix(ig, kb), dp) + &
1037 0 : AIMAG(res_matrix(ig, kb))*AIMAG(res_matrix(ig, kb))
1038 0 : IF (g2(ig) < 4.0_dp) THEN
1039 : norm_res_low(kb) = norm_res_low(kb) + &
1040 : REAL(res_matrix(ig, kb), dp)*REAL(res_matrix(ig, kb), dp) + &
1041 0 : AIMAG(res_matrix(ig, kb))*AIMAG(res_matrix(ig, kb))
1042 : ELSE
1043 : norm_res_up(kb) = norm_res_up(kb) + &
1044 : REAL(res_matrix(ig, kb), dp)*REAL(res_matrix(ig, kb), dp) + &
1045 0 : AIMAG(res_matrix(ig, kb))*AIMAG(res_matrix(ig, kb))
1046 : END IF
1047 0 : res_matrix(ig, kb) = prec*res_matrix(ig, kb)
1048 : END DO
1049 : END DO !jb
1050 :
1051 : ! normalize each column of S and Y => Snorm Ynorm
1052 0 : CALL para_env%sum(norm_res)
1053 0 : CALL para_env%sum(norm_res_up)
1054 0 : CALL para_env%sum(norm_res_low)
1055 0 : norm_res(1:nb) = 1.0_dp/SQRT(norm_res(1:nb))
1056 : n_low = 0.0_dp
1057 : n_up = 0.0_dp
1058 0 : DO jb = 1, nb
1059 0 : n_low = n_low + norm_res_low(jb)/norm_res(jb)
1060 0 : n_up = n_up + norm_res_up(jb)/norm_res(jb)
1061 : END DO
1062 0 : DO ig = 1, ng
1063 0 : IF (g2(ig) > 4.0_dp) THEN
1064 0 : step_matrix(ig, 1:nb) = step_matrix(ig, 1:nb)*SQRT(n_low/n_up)
1065 0 : res_matrix(ig, 1:nb) = res_matrix(ig, 1:nb)*SQRT(n_low/n_up)
1066 : END IF
1067 : END DO
1068 0 : DO kb = 1, nb
1069 0 : step_matrix(1:ng, kb) = step_matrix(1:ng, kb)*norm_res(kb)
1070 0 : res_matrix(1:ng, kb) = res_matrix(1:ng, kb)*norm_res(kb)
1071 : END DO
1072 :
1073 : ! compute A as [(Ynorm^t Ynorm) + (alpha I)]^(-1) Ynorm^t
1074 : ! compute B
1075 0 : DO jb = 1, nb
1076 0 : DO kb = 1, nb
1077 0 : b_matrix(kb, jb) = 0.0_dp
1078 0 : DO ig = 1, ng
1079 : ! it is assumed that summing over all G vector gives a real, because
1080 : ! y(-G,kb) = (y(G,kb))*
1081 0 : b_matrix(kb, jb) = b_matrix(kb, jb) + REAL(res_matrix(ig, kb)*res_matrix(ig, jb), dp)
1082 : END DO
1083 : END DO
1084 : END DO
1085 :
1086 0 : CALL para_env%sum(b_matrix)
1087 0 : DO jb = 1, nb
1088 0 : b_matrix(jb, jb) = b_matrix(jb, jb) + reg_par
1089 : END DO
1090 : ! invert B
1091 0 : CALL invert_matrix(b_matrix, binv_matrix, inv_err)
1092 :
1093 : ! A = Binv Ynorm^t
1094 0 : a_matrix = CMPLX(0.0_dp, 0.0_dp, KIND=dp)
1095 0 : DO kb = 1, nb
1096 0 : res_matrix_global = CMPLX(0.0_dp, 0.0_dp, KIND=dp)
1097 0 : DO ig = 1, ng
1098 0 : ig_global = mixing_store%ig_global_index(ig)
1099 0 : res_matrix_global(ig_global) = res_matrix(ig, kb)
1100 : END DO
1101 0 : CALL para_env%sum(res_matrix_global)
1102 :
1103 0 : DO jb = 1, nb
1104 0 : DO ig = 1, ng
1105 0 : ig_global = mixing_store%ig_global_index(ig)
1106 : a_matrix(jb, ig_global) = a_matrix(jb, ig_global) + &
1107 0 : binv_matrix(jb, kb)*res_matrix_global(ig_global)
1108 : END DO
1109 : END DO
1110 : END DO
1111 0 : CALL para_env%sum(a_matrix)
1112 :
1113 : ! compute the two components of gn that will be used to update rho
1114 0 : gn => mixing_store%res_buffer(ib, ispin)%cc
1115 0 : pgn_norm = 0.0_dp
1116 :
1117 : IF (use_zgemm) THEN
1118 :
1119 : CALL zgemm("N", "N", ng, ng_global, nb, cmone, step_matrix(1, 1), ng, &
1120 : a_matrix(1, 1), nb, czero, sa(1, 1), ng)
1121 : CALL zgemm("N", "N", ng, ng_global, nb, cmone, res_matrix(1, 1), ng, &
1122 : a_matrix(1, 1), nb, czero, ya(1, 1), ng)
1123 : DO ig = 1, ng
1124 : ig_global = mixing_store%ig_global_index(ig)
1125 : ya(ig, ig_global) = ya(ig, ig_global) + CMPLX(1.0_dp, 0.0_dp, KIND=dp)
1126 : END DO
1127 :
1128 : CALL zgemv("N", ng, ng_global, cone, sa(1, 1), &
1129 : ng, gn_global(1), 1, czero, pgn(1), 1)
1130 : CALL zgemv("N", ng, ng_global, cone, ya(1, 1), &
1131 : ng, gn_global(1), 1, czero, ugn(1), 1)
1132 :
1133 : DO ig = 1, ng
1134 : pgn_norm = pgn_norm + REAL(pgn(ig), dp)*REAL(pgn(ig), dp) + &
1135 : AIMAG(pgn(ig))*AIMAG(pgn(ig))
1136 : END DO
1137 : CALL para_env%sum(pgn_norm)
1138 : ELSEIF (use_zgemm_rev) THEN
1139 :
1140 : CALL zgemv("N", nb, ng_global, cone, a_matrix(1, 1), &
1141 0 : nb, gn_global(1), 1, czero, tmp_vec(1), 1)
1142 :
1143 : CALL zgemv("N", ng, nb, cmone, step_matrix(1, 1), ng, &
1144 0 : tmp_vec(1), 1, czero, pgn(1), 1)
1145 :
1146 : CALL zgemv("N", ng, nb, cmone, res_matrix(1, 1), ng, &
1147 0 : tmp_vec(1), 1, czero, ugn(1), 1)
1148 :
1149 0 : DO ig = 1, ng
1150 : pgn_norm = pgn_norm + REAL(pgn(ig), dp)*REAL(pgn(ig), dp) + &
1151 0 : AIMAG(pgn(ig))*AIMAG(pgn(ig))
1152 0 : ugn(ig) = ugn(ig) + gn(ig)
1153 : END DO
1154 0 : CALL para_env%sum(pgn_norm)
1155 :
1156 : ELSE
1157 : DO ig = 1, ng
1158 : pgn(ig) = CMPLX(0.0_dp, 0.0_dp, KIND=dp)
1159 : ugn(ig) = CMPLX(0.0_dp, 0.0_dp, KIND=dp)
1160 : ig_global = mixing_store%ig_global_index(ig)
1161 : DO iig = 1, ng_global
1162 : saa = CMPLX(0.0_dp, 0.0_dp, KIND=dp)
1163 : yaa = CMPLX(0.0_dp, 0.0_dp, KIND=dp)
1164 :
1165 : IF (ig_global == iig) yaa = CMPLX(1.0_dp, 0.0_dp, KIND=dp)
1166 :
1167 : DO jb = 1, nb
1168 : saa = saa - step_matrix(ig, jb)*a_matrix(jb, iig)
1169 : yaa = yaa - res_matrix(ig, jb)*a_matrix(jb, iig)
1170 : END DO
1171 : pgn(ig) = pgn(ig) + saa*gn_global(iig)
1172 : ugn(ig) = ugn(ig) + yaa*gn_global(iig)
1173 : END DO
1174 : END DO
1175 : DO ig = 1, ng
1176 : pgn_norm = pgn_norm + REAL(pgn(ig), dp)*REAL(pgn(ig), dp) + &
1177 : AIMAG(pgn(ig))*AIMAG(pgn(ig))
1178 : END DO
1179 : CALL para_env%sum(pgn_norm)
1180 : END IF
1181 :
1182 0 : gn_norm = mixing_store%norm_res_buffer(ib, ispin)
1183 0 : gn_norm_old = mixing_store%norm_res_buffer(ib_prev, ispin)
1184 : IF (ib_prev /= 0) THEN
1185 0 : sigma_tilde = sigma_old*MAX(0.5_dp, MIN(2.0_dp, gn_norm_old/gn_norm))
1186 : ELSE
1187 : sigma_tilde = 0.5_dp
1188 : END IF
1189 0 : sigma_tilde = 0.1_dp
1190 : ! Step size for the unpredicted component
1191 0 : step_size = MIN(sigma_tilde, r_step*pgn_norm/gn_norm, sigma_max)
1192 0 : sigma_old = step_size
1193 :
1194 : ! update the density
1195 0 : DO ig = 1, ng
1196 0 : prec = mixing_store%kerker_factor(ig)
1197 : rho_g(ispin)%array(ig) = mixing_store%rhoin_buffer(ib, ispin)%cc(ig) &
1198 0 : - prec*step_size*ugn(ig) + prec*pgn(ig) ! - 0.1_dp * prec* gn(ig)
1199 0 : mixing_store%rhoin_buffer(ib_next, ispin)%cc(ig) = rho_g(ispin)%array(ig)
1200 : END DO
1201 :
1202 : END DO ! ispin
1203 :
1204 : ! Deallocate temporary arrays
1205 0 : DEALLOCATE (step_matrix, res_matrix)
1206 0 : DEALLOCATE (norm_res)
1207 0 : DEALLOCATE (norm_res_up)
1208 0 : DEALLOCATE (norm_res_low)
1209 0 : DEALLOCATE (a_matrix, b_matrix, binv_matrix)
1210 0 : DEALLOCATE (ugn, pgn)
1211 : IF (use_zgemm) THEN
1212 : DEALLOCATE (sa, ya)
1213 : END IF
1214 : IF (use_zgemm_rev) THEN
1215 0 : DEALLOCATE (tmp_vec)
1216 : END IF
1217 0 : DEALLOCATE (gn_global, res_matrix_global)
1218 :
1219 0 : CALL timestop(handle)
1220 :
1221 0 : END SUBROUTINE multisecant_mixing
1222 :
1223 : END MODULE qs_gspace_mixing
|
