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 : !> \brief Main routines for GW + Bethe-Salpeter for computing electronic excitations
10 : !> \par History
11 : !> 04.2024 created [Maximilian Graml]
12 : ! **************************************************************************************************
13 :
14 : MODULE bse_main
15 :
16 : USE bse_full_diag, ONLY: create_A,&
17 : create_B,&
18 : create_hermitian_form_of_ABBA,&
19 : diagonalize_A,&
20 : diagonalize_C
21 : USE bse_iterative, ONLY: do_subspace_iterations,&
22 : fill_local_3c_arrays
23 : USE bse_util, ONLY: deallocate_matrices_bse,&
24 : estimate_BSE_resources,&
25 : mult_B_with_W,&
26 : print_BSE_start_flag,&
27 : truncate_BSE_matrices
28 : USE cp_fm_types, ONLY: cp_fm_release,&
29 : cp_fm_type
30 : USE group_dist_types, ONLY: group_dist_d1_type
31 : USE input_constants, ONLY: bse_abba,&
32 : bse_both,&
33 : bse_fulldiag,&
34 : bse_iterdiag,&
35 : bse_tda
36 : USE kinds, ONLY: dp
37 : USE message_passing, ONLY: mp_para_env_type
38 : USE mp2_types, ONLY: mp2_type
39 : #include "./base/base_uses.f90"
40 :
41 : IMPLICIT NONE
42 :
43 : PRIVATE
44 :
45 : CHARACTER(len=*), PARAMETER, PRIVATE :: moduleN = 'bse_main'
46 :
47 : PUBLIC :: start_bse_calculation
48 :
49 : CONTAINS
50 :
51 : ! **************************************************************************************************
52 : !> \brief Main subroutine managing BSE calculations
53 : !> \param fm_mat_S_ia_bse ...
54 : !> \param fm_mat_S_ij_bse ...
55 : !> \param fm_mat_S_ab_bse ...
56 : !> \param fm_mat_Q_static_bse ...
57 : !> \param fm_mat_Q_static_bse_gemm ...
58 : !> \param Eigenval ...
59 : !> \param homo ...
60 : !> \param virtual ...
61 : !> \param dimen_RI ...
62 : !> \param dimen_RI_red ...
63 : !> \param gd_array ...
64 : !> \param color_sub ...
65 : !> \param mp2_env ...
66 : !> \param unit_nr ...
67 : ! **************************************************************************************************
68 4 : SUBROUTINE start_bse_calculation(fm_mat_S_ia_bse, fm_mat_S_ij_bse, fm_mat_S_ab_bse, &
69 : fm_mat_Q_static_bse, fm_mat_Q_static_bse_gemm, &
70 4 : Eigenval, homo, virtual, dimen_RI, dimen_RI_red, &
71 : gd_array, color_sub, mp2_env, unit_nr)
72 :
73 : TYPE(cp_fm_type), INTENT(IN) :: fm_mat_S_ia_bse, fm_mat_S_ij_bse, &
74 : fm_mat_S_ab_bse
75 : TYPE(cp_fm_type), INTENT(INOUT) :: fm_mat_Q_static_bse, &
76 : fm_mat_Q_static_bse_gemm
77 : REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :), &
78 : INTENT(IN) :: Eigenval
79 : INTEGER, DIMENSION(:), INTENT(IN) :: homo, virtual
80 : INTEGER, INTENT(IN) :: dimen_RI, dimen_RI_red
81 : TYPE(group_dist_d1_type), INTENT(IN) :: gd_array
82 : INTEGER, INTENT(IN) :: color_sub
83 : TYPE(mp2_type) :: mp2_env
84 : INTEGER, INTENT(IN) :: unit_nr
85 :
86 : CHARACTER(LEN=*), PARAMETER :: routineN = 'start_bse_calculation'
87 :
88 : INTEGER :: handle, homo_red, virtual_red
89 : LOGICAL :: my_do_abba, my_do_fulldiag, &
90 : my_do_iterat_diag, my_do_tda
91 : REAL(KIND=dp) :: diag_runtime_est
92 4 : REAL(KIND=dp), ALLOCATABLE, DIMENSION(:) :: Eigenval_reduced
93 4 : REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :) :: B_abQ_bse_local, B_bar_iaQ_bse_local, &
94 4 : B_bar_ijQ_bse_local, B_iaQ_bse_local
95 : TYPE(cp_fm_type) :: fm_A_BSE, fm_B_BSE, fm_C_BSE, fm_inv_sqrt_A_minus_B, fm_mat_S_ab_trunc, &
96 : fm_mat_S_bar_ia_bse, fm_mat_S_bar_ij_bse, fm_mat_S_ia_trunc, fm_mat_S_ij_trunc, &
97 : fm_sqrt_A_minus_B
98 : TYPE(mp_para_env_type), POINTER :: para_env
99 :
100 4 : CALL timeset(routineN, handle)
101 :
102 4 : para_env => fm_mat_S_ia_bse%matrix_struct%para_env
103 :
104 4 : my_do_fulldiag = .FALSE.
105 4 : my_do_iterat_diag = .FALSE.
106 4 : my_do_tda = .FALSE.
107 4 : my_do_abba = .FALSE.
108 : !Method: Iterative or full diagonalization
109 4 : SELECT CASE (mp2_env%ri_g0w0%bse_diag_method)
110 : CASE (bse_iterdiag)
111 0 : my_do_iterat_diag = .TRUE.
112 : !MG: Basics of the Davidson solver are implemented, but not rigorously checked.
113 0 : CPABORT("Iterative BSE not yet implemented")
114 : CASE (bse_fulldiag)
115 4 : my_do_fulldiag = .TRUE.
116 : END SELECT
117 : !Approximation: TDA and/or full ABBA matrix
118 6 : SELECT CASE (mp2_env%ri_g0w0%bse_approx)
119 : CASE (bse_tda)
120 2 : my_do_tda = .TRUE.
121 : CASE (bse_abba)
122 2 : my_do_abba = .TRUE.
123 : CASE (bse_both)
124 0 : my_do_tda = .TRUE.
125 4 : my_do_abba = .TRUE.
126 : END SELECT
127 :
128 4 : CALL print_BSE_start_flag(my_do_tda, my_do_abba, unit_nr)
129 :
130 4 : CALL fm_mat_S_ia_bse%matrix_struct%para_env%sync()
131 :
132 : ! Reduce matrices in case of energy cutoff for occupied and unoccupied in A/B-BSE-matrices
133 : CALL truncate_BSE_matrices(fm_mat_S_ia_bse, fm_mat_S_ij_bse, fm_mat_S_ab_bse, &
134 : fm_mat_S_ia_trunc, fm_mat_S_ij_trunc, fm_mat_S_ab_trunc, &
135 : Eigenval(:, 1, 1), Eigenval_reduced, &
136 : homo(1), virtual(1), dimen_RI, unit_nr, &
137 : homo_red, virtual_red, &
138 4 : mp2_env)
139 : ! \bar{B}^P_rs = \sum_R W_PR B^R_rs where B^R_rs = \sum_T [1/sqrt(v)]_RT (T|rs)
140 : ! r,s: MO-index, P,R,T: RI-index
141 : ! B: fm_mat_S_..., W: fm_mat_Q_...
142 : CALL mult_B_with_W(fm_mat_S_ij_trunc, fm_mat_S_ia_trunc, fm_mat_S_bar_ia_bse, &
143 : fm_mat_S_bar_ij_bse, fm_mat_Q_static_bse_gemm, &
144 4 : dimen_RI_red, homo_red, virtual_red)
145 :
146 4 : IF (my_do_iterat_diag) THEN
147 : CALL fill_local_3c_arrays(fm_mat_S_ab_trunc, fm_mat_S_ia_trunc, &
148 : fm_mat_S_bar_ia_bse, fm_mat_S_bar_ij_bse, &
149 : B_bar_ijQ_bse_local, B_abQ_bse_local, B_bar_iaQ_bse_local, &
150 : B_iaQ_bse_local, dimen_RI_red, homo_red, virtual_red, &
151 : gd_array, color_sub, para_env)
152 : END IF
153 :
154 4 : IF (my_do_fulldiag) THEN
155 : ! Quick estimate of memory consumption and runtime of diagonalizations
156 : CALL estimate_BSE_resources(homo_red, virtual_red, unit_nr, my_do_abba, &
157 4 : para_env, diag_runtime_est)
158 : ! Matrix A constructed from GW energies and 3c-B-matrices (cf. subroutine mult_B_with_W)
159 : ! A_ia,jb = (ε_a-ε_i) δ_ij δ_ab + α * v_ia,jb - W_ij,ab
160 : ! ε_a, ε_i are GW singleparticle energies from Eigenval_reduced
161 : ! α is a spin-dependent factor
162 : ! v_ia,jb = \sum_P B^P_ia B^P_jb (unscreened Coulomb interaction)
163 : ! W_ij,ab = \sum_P \bar{B}^P_ij B^P_ab (screened Coulomb interaction)
164 : CALL create_A(fm_mat_S_ia_trunc, fm_mat_S_bar_ij_bse, fm_mat_S_ab_trunc, &
165 : fm_A_BSE, Eigenval_reduced, unit_nr, &
166 : homo_red, virtual_red, dimen_RI, mp2_env, &
167 4 : para_env)
168 4 : IF (my_do_abba) THEN
169 : ! Matrix B constructed from 3c-B-matrices (cf. subroutine mult_B_with_W)
170 : ! B_ia,jb = α * v_ia,jb - W_ib,aj
171 : ! α is a spin-dependent factor
172 : ! v_ia,jb = \sum_P B^P_ia B^P_jb (unscreened Coulomb interaction)
173 : ! W_ib,aj = \sum_P \bar{B}^P_ib B^P_aj (screened Coulomb interaction)
174 : CALL create_B(fm_mat_S_ia_trunc, fm_mat_S_bar_ia_bse, fm_B_BSE, &
175 2 : homo_red, virtual_red, dimen_RI, unit_nr, mp2_env)
176 : ! Construct Matrix C=(A-B)^0.5 (A+B) (A-B)^0.5 to solve full BSE matrix as a hermitian problem
177 : ! (cf. Eq. (A7) in F. Furche J. Chem. Phys., Vol. 114, No. 14, (2001)).
178 : ! We keep fm_sqrt_A_minus_B and fm_inv_sqrt_A_minus_B for print of singleparticle transitions
179 : ! of ABBA as described in Eq. (A10) in F. Furche J. Chem. Phys., Vol. 114, No. 14, (2001).
180 : CALL create_hermitian_form_of_ABBA(fm_A_BSE, fm_B_BSE, fm_C_BSE, &
181 : fm_sqrt_A_minus_B, fm_inv_sqrt_A_minus_B, &
182 2 : homo_red, virtual_red, unit_nr, mp2_env, diag_runtime_est)
183 : END IF
184 4 : CALL cp_fm_release(fm_B_BSE)
185 4 : IF (my_do_tda) THEN
186 : ! Solving the hermitian eigenvalue equation A X^n = Ω^n X^n
187 : CALL diagonalize_A(fm_A_BSE, homo_red, virtual_red, homo(1), &
188 2 : unit_nr, diag_runtime_est, mp2_env)
189 : END IF
190 : ! Release to avoid faulty use of changed A matrix
191 4 : CALL cp_fm_release(fm_A_BSE)
192 4 : IF (my_do_abba) THEN
193 : ! Solving eigenvalue equation C Z^n = (Ω^n)^2 Z^n .
194 : ! Here, the eigenvectors Z^n relate to X^n via
195 : ! Eq. (A10) in F. Furche J. Chem. Phys., Vol. 114, No. 14, (2001).
196 : CALL diagonalize_C(fm_C_BSE, homo_red, virtual_red, homo(1), &
197 : fm_sqrt_A_minus_B, fm_inv_sqrt_A_minus_B, &
198 2 : unit_nr, diag_runtime_est, mp2_env)
199 : END IF
200 : ! Release to avoid faulty use of changed C matrix
201 4 : CALL cp_fm_release(fm_C_BSE)
202 : END IF
203 :
204 : CALL deallocate_matrices_bse(fm_mat_S_bar_ia_bse, fm_mat_S_bar_ij_bse, &
205 : fm_mat_S_ia_trunc, fm_mat_S_ij_trunc, fm_mat_S_ab_trunc, &
206 4 : fm_mat_Q_static_bse, fm_mat_Q_static_bse_gemm)
207 4 : DEALLOCATE (Eigenval_reduced)
208 4 : IF (my_do_iterat_diag) THEN
209 : ! Contains untested Block-Davidson algorithm
210 : CALL do_subspace_iterations(B_bar_ijQ_bse_local, B_abQ_bse_local, B_bar_iaQ_bse_local, &
211 : B_iaQ_bse_local, homo(1), virtual(1), mp2_env%ri_g0w0%bse_spin_config, unit_nr, &
212 0 : Eigenval(:, 1, 1), para_env, mp2_env)
213 : ! Deallocate local 3c-B-matrices
214 0 : DEALLOCATE (B_bar_ijQ_bse_local, B_abQ_bse_local, B_bar_iaQ_bse_local, B_iaQ_bse_local)
215 : END IF
216 :
217 4 : IF (unit_nr > 0) THEN
218 2 : WRITE (unit_nr, '(T2,A4,T7,A53)') 'BSE|', 'The BSE was successfully calculated. Have a nice day!'
219 : END IF
220 :
221 4 : CALL timestop(handle)
222 :
223 8 : END SUBROUTINE start_bse_calculation
224 :
225 : END MODULE bse_main
|
