mirror of
https://github.com/LCPQ/quantum_package
synced 2024-11-08 23:23:57 +01:00
365 lines
13 KiB
Fortran
365 lines
13 KiB
Fortran
BEGIN_PROVIDER [ double precision, Fock_matrix_mo, (mo_tot_num_align,mo_tot_num) ]
|
|
&BEGIN_PROVIDER [ double precision, Fock_matrix_diag_mo, (mo_tot_num)]
|
|
implicit none
|
|
BEGIN_DOC
|
|
! Fock matrix on the MO basis.
|
|
! For open shells, the ROHF Fock Matrix is
|
|
!
|
|
! | F-K | F + K/2 | F |
|
|
! |---------------------------------|
|
|
! | F + K/2 | F | F - K/2 |
|
|
! |---------------------------------|
|
|
! | F | F - K/2 | F + K |
|
|
!
|
|
! F = 1/2 (Fa + Fb)
|
|
!
|
|
! K = Fb - Fa
|
|
!
|
|
END_DOC
|
|
integer :: i,j,n
|
|
if (elec_alpha_num == elec_beta_num) then
|
|
Fock_matrix_mo = Fock_matrix_alpha_mo
|
|
else
|
|
|
|
do j=1,elec_beta_num
|
|
! F-K
|
|
do i=1,elec_beta_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j))&
|
|
- (Fock_matrix_beta_mo(i,j) - Fock_matrix_alpha_mo(i,j))
|
|
enddo
|
|
! F+K/2
|
|
do i=elec_beta_num+1,elec_alpha_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j))&
|
|
+ 0.5d0*(Fock_matrix_beta_mo(i,j) - Fock_matrix_alpha_mo(i,j))
|
|
enddo
|
|
! F
|
|
do i=elec_alpha_num+1, mo_tot_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j))
|
|
enddo
|
|
enddo
|
|
|
|
do j=elec_beta_num+1,elec_alpha_num
|
|
! F+K/2
|
|
do i=1,elec_beta_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j))&
|
|
+ 0.5d0*(Fock_matrix_beta_mo(i,j) - Fock_matrix_alpha_mo(i,j))
|
|
enddo
|
|
! F
|
|
do i=elec_beta_num+1,elec_alpha_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j))
|
|
enddo
|
|
! F-K/2
|
|
do i=elec_alpha_num+1, mo_tot_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j))&
|
|
- 0.5d0*(Fock_matrix_beta_mo(i,j) - Fock_matrix_alpha_mo(i,j))
|
|
enddo
|
|
enddo
|
|
|
|
do j=elec_alpha_num+1, mo_tot_num
|
|
! F
|
|
do i=1,elec_beta_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j))
|
|
enddo
|
|
! F-K/2
|
|
do i=elec_beta_num+1,elec_alpha_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j))&
|
|
- 0.5d0*(Fock_matrix_beta_mo(i,j) - Fock_matrix_alpha_mo(i,j))
|
|
enddo
|
|
! F+K
|
|
do i=elec_alpha_num+1,mo_tot_num
|
|
Fock_matrix_mo(i,j) = 0.5d0*(Fock_matrix_alpha_mo(i,j)+Fock_matrix_beta_mo(i,j)) &
|
|
+ (Fock_matrix_beta_mo(i,j) - Fock_matrix_alpha_mo(i,j))
|
|
enddo
|
|
enddo
|
|
|
|
endif
|
|
do i = 1, mo_tot_num
|
|
Fock_matrix_diag_mo(i) = Fock_matrix_mo(i,i)
|
|
enddo
|
|
END_PROVIDER
|
|
|
|
|
|
|
|
BEGIN_PROVIDER [ double precision, Fock_matrix_alpha_ao, (ao_num_align, ao_num) ]
|
|
&BEGIN_PROVIDER [ double precision, Fock_matrix_beta_ao, (ao_num_align, ao_num) ]
|
|
implicit none
|
|
BEGIN_DOC
|
|
! Alpha Fock matrix in AO basis set
|
|
END_DOC
|
|
|
|
integer :: i,j
|
|
do j=1,ao_num
|
|
!DIR$ VECTOR ALIGNED
|
|
do i=1,ao_num
|
|
Fock_matrix_alpha_ao(i,j) = ao_mono_elec_integral(i,j) + ao_bi_elec_integral_alpha(i,j)
|
|
Fock_matrix_beta_ao (i,j) = ao_mono_elec_integral(i,j) + ao_bi_elec_integral_beta (i,j)
|
|
enddo
|
|
enddo
|
|
|
|
END_PROVIDER
|
|
|
|
|
|
BEGIN_PROVIDER [ double precision, ao_bi_elec_integral_alpha, (ao_num_align, ao_num) ]
|
|
&BEGIN_PROVIDER [ double precision, ao_bi_elec_integral_beta , (ao_num_align, ao_num) ]
|
|
use map_module
|
|
implicit none
|
|
BEGIN_DOC
|
|
! Alpha Fock matrix in AO basis set
|
|
END_DOC
|
|
|
|
integer :: i,j,k,l,k1,r,s
|
|
integer*8 :: p,q
|
|
double precision :: integral
|
|
double precision :: ao_bielec_integral
|
|
if (do_direct_integrals) then
|
|
|
|
ao_bi_elec_integral_alpha = 0.d0
|
|
ao_bi_elec_integral_beta = 0.d0
|
|
!$OMP PARALLEL DEFAULT(NONE) &
|
|
!$OMP PRIVATE(i,j,l,k1,k,integral,ii,jj,kk,ll,i8,keys,values,p,q,r,s)&
|
|
!$OMP SHARED(ao_num,HF_density_matrix_ao_alpha,HF_density_matrix_ao_beta,&
|
|
!$OMP ao_integrals_map,ao_integrals_threshold, ao_bielec_integral_schwartz, &
|
|
!$OMP ao_overlap_abs) &
|
|
!$OMP REDUCTION(+:ao_bi_elec_integral_alpha,ao_bi_elec_integral_beta)
|
|
|
|
allocate(keys(1), values(1))
|
|
|
|
q = ao_num*ao_num*ao_num*ao_num
|
|
!$OMP DO SCHEDULE(dynamic)
|
|
do p=1_8,q
|
|
call bielec_integrals_index_reverse(kk,ii,ll,jj,p)
|
|
if ( (kk(1)>ao_num).or. &
|
|
(ii(1)>ao_num).or. &
|
|
(jj(1)>ao_num).or. &
|
|
(ll(1)>ao_num) ) then
|
|
cycle
|
|
endif
|
|
k = kk(1)
|
|
i = ii(1)
|
|
l = ll(1)
|
|
j = jj(1)
|
|
|
|
if (ao_overlap_abs(k,l)*ao_overlap_abs(i,j) &
|
|
< ao_integrals_threshold) then
|
|
cycle
|
|
endif
|
|
if (ao_bielec_integral_schwartz(k,l)*ao_bielec_integral_schwartz(i,j) &
|
|
< ao_integrals_threshold) then
|
|
cycle
|
|
endif
|
|
values(1) = ao_bielec_integral(k,l,i,j)
|
|
if (abs(values(1)) < ao_integrals_threshold) then
|
|
cycle
|
|
endif
|
|
do k2=1,8
|
|
if (kk(k2)==0) then
|
|
cycle
|
|
endif
|
|
i = ii(k2)
|
|
j = jj(k2)
|
|
k = kk(k2)
|
|
l = ll(k2)
|
|
integral = (HF_density_matrix_ao_alpha(k,l)+HF_density_matrix_ao_beta(k,l)) * values(1)
|
|
ao_bi_elec_integral_alpha(i,j) += integral
|
|
ao_bi_elec_integral_beta (i,j) += integral
|
|
integral = values(1)
|
|
ao_bi_elec_integral_alpha(l,j) -= HF_density_matrix_ao_alpha(k,i) * integral
|
|
ao_bi_elec_integral_beta (l,j) -= HF_density_matrix_ao_beta (k,i) * integral
|
|
enddo
|
|
enddo
|
|
!$OMP END DO
|
|
deallocate(keys,values)
|
|
!$OMP END PARALLEL
|
|
else
|
|
PROVIDE ao_bielec_integrals_in_map
|
|
|
|
integer(omp_lock_kind) :: lck(ao_num)
|
|
integer*8 :: i8
|
|
integer :: ii(8), jj(8), kk(8), ll(8), k2
|
|
integer(cache_map_size_kind) :: n_elements_max, n_elements
|
|
integer(key_kind), allocatable :: keys(:)
|
|
double precision, allocatable :: values(:)
|
|
|
|
ao_bi_elec_integral_alpha = 0.d0
|
|
ao_bi_elec_integral_beta = 0.d0
|
|
!$OMP PARALLEL DEFAULT(NONE) &
|
|
!$OMP PRIVATE(i,j,l,k1,k,integral,ii,jj,kk,ll,i8,keys,values,n_elements_max,n_elements)&
|
|
!$OMP SHARED(ao_num,HF_density_matrix_ao_alpha,HF_density_matrix_ao_beta,&
|
|
!$OMP ao_integrals_map) &
|
|
!$OMP REDUCTION(+:ao_bi_elec_integral_alpha,ao_bi_elec_integral_beta)
|
|
|
|
call get_cache_map_n_elements_max(ao_integrals_map,n_elements_max)
|
|
allocate(keys(n_elements_max), values(n_elements_max))
|
|
|
|
!$OMP DO SCHEDULE(dynamic)
|
|
do i8=0_8,ao_integrals_map%map_size
|
|
n_elements = n_elements_max
|
|
call get_cache_map(ao_integrals_map,i8,keys,values,n_elements)
|
|
do k1=1,n_elements
|
|
call bielec_integrals_index_reverse(kk,ii,ll,jj,keys(k1))
|
|
|
|
do k2=1,8
|
|
if (kk(k2)==0) then
|
|
cycle
|
|
endif
|
|
i = ii(k2)
|
|
j = jj(k2)
|
|
k = kk(k2)
|
|
l = ll(k2)
|
|
integral = (HF_density_matrix_ao_alpha(k,l)+HF_density_matrix_ao_beta(k,l)) * values(k1)
|
|
ao_bi_elec_integral_alpha(i,j) += integral
|
|
ao_bi_elec_integral_beta (i,j) += integral
|
|
integral = values(k1)
|
|
ao_bi_elec_integral_alpha(l,j) -= HF_density_matrix_ao_alpha(k,i) * integral
|
|
ao_bi_elec_integral_beta (l,j) -= HF_density_matrix_ao_beta (k,i) * integral
|
|
enddo
|
|
enddo
|
|
enddo
|
|
!$OMP END DO
|
|
deallocate(keys,values)
|
|
!$OMP END PARALLEL
|
|
|
|
endif
|
|
|
|
END_PROVIDER
|
|
|
|
|
|
|
|
|
|
|
|
|
|
BEGIN_PROVIDER [ double precision, Fock_matrix_alpha_mo, (mo_tot_num_align,mo_tot_num) ]
|
|
implicit none
|
|
BEGIN_DOC
|
|
! Fock matrix on the MO basis
|
|
END_DOC
|
|
double precision, allocatable :: T(:,:)
|
|
allocate ( T(ao_num_align,mo_tot_num) )
|
|
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: T
|
|
call dgemm('N','N', ao_num, mo_tot_num, ao_num, &
|
|
1.d0, Fock_matrix_alpha_ao,size(Fock_matrix_alpha_ao,1), &
|
|
mo_coef, size(mo_coef,1), &
|
|
0.d0, T, ao_num_align)
|
|
call dgemm('T','N', mo_tot_num, mo_tot_num, ao_num, &
|
|
1.d0, mo_coef,size(mo_coef,1), &
|
|
T, size(T,1), &
|
|
0.d0, Fock_matrix_alpha_mo, mo_tot_num_align)
|
|
deallocate(T)
|
|
END_PROVIDER
|
|
|
|
|
|
BEGIN_PROVIDER [ double precision, Fock_matrix_beta_mo, (mo_tot_num_align,mo_tot_num) ]
|
|
implicit none
|
|
BEGIN_DOC
|
|
! Fock matrix on the MO basis
|
|
END_DOC
|
|
double precision, allocatable :: T(:,:)
|
|
allocate ( T(ao_num_align,mo_tot_num) )
|
|
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: T
|
|
call dgemm('N','N', ao_num, mo_tot_num, ao_num, &
|
|
1.d0, Fock_matrix_beta_ao,size(Fock_matrix_beta_ao,1), &
|
|
mo_coef, size(mo_coef,1), &
|
|
0.d0, T, ao_num_align)
|
|
call dgemm('T','N', mo_tot_num, mo_tot_num, ao_num, &
|
|
1.d0, mo_coef,size(mo_coef,1), &
|
|
T, size(T,1), &
|
|
0.d0, Fock_matrix_beta_mo, mo_tot_num_align)
|
|
deallocate(T)
|
|
END_PROVIDER
|
|
|
|
BEGIN_PROVIDER [ double precision, HF_energy ]
|
|
implicit none
|
|
BEGIN_DOC
|
|
! Hartree-Fock energy
|
|
END_DOC
|
|
HF_energy = nuclear_repulsion
|
|
|
|
integer :: i,j
|
|
do j=1,ao_num
|
|
do i=1,ao_num
|
|
HF_energy += 0.5d0 * ( &
|
|
(ao_mono_elec_integral(i,j) + Fock_matrix_alpha_ao(i,j) ) * HF_density_matrix_ao_alpha(i,j) +&
|
|
(ao_mono_elec_integral(i,j) + Fock_matrix_beta_ao (i,j) ) * HF_density_matrix_ao_beta (i,j) )
|
|
enddo
|
|
enddo
|
|
|
|
END_PROVIDER
|
|
|
|
|
|
BEGIN_PROVIDER [ double precision, Fock_matrix_ao, (ao_num_align, ao_num) ]
|
|
implicit none
|
|
BEGIN_DOC
|
|
! Fock matrix in AO basis set
|
|
END_DOC
|
|
|
|
if (elec_alpha_num == elec_beta_num) then
|
|
integer :: i,j
|
|
do j=1,ao_num
|
|
!DIR$ VECTOR ALIGNED
|
|
do i=1,ao_num_align
|
|
Fock_matrix_ao(i,j) = Fock_matrix_alpha_ao(i,j)
|
|
enddo
|
|
enddo
|
|
else
|
|
double precision, allocatable :: T(:,:), M(:,:)
|
|
! F_ao = S C F_mo C^t S
|
|
allocate (T(ao_num_align,ao_num),M(ao_num_align,ao_num))
|
|
call dgemm('N','N', ao_num,ao_num,ao_num, 1.d0, &
|
|
ao_overlap, size(ao_overlap,1), &
|
|
mo_coef, size(mo_coef,1), &
|
|
0.d0, &
|
|
M, size(M,1))
|
|
call dgemm('N','N', ao_num,mo_tot_num,mo_tot_num, 1.d0, &
|
|
M, size(M,1), &
|
|
Fock_matrix_mo, size(Fock_matrix_mo,1), &
|
|
0.d0, &
|
|
T, size(T,1))
|
|
call dgemm('N','T', mo_tot_num,ao_num,mo_tot_num, 1.d0, &
|
|
T, size(T,1), &
|
|
mo_coef, size(mo_coef,1), &
|
|
0.d0, &
|
|
M, size(M,1))
|
|
call dgemm('N','N', ao_num,ao_num,ao_num, 1.d0, &
|
|
M, size(M,1), &
|
|
ao_overlap, size(ao_overlap,1), &
|
|
0.d0, &
|
|
Fock_matrix_ao, size(Fock_matrix_ao,1))
|
|
|
|
deallocate(T)
|
|
endif
|
|
END_PROVIDER
|
|
|
|
subroutine Fock_mo_to_ao(FMO,LDFMO,FAO,LDFAO)
|
|
implicit none
|
|
integer, intent(in) :: LDFMO ! size(FMO,1)
|
|
integer, intent(in) :: LDFAO ! size(FAO,1)
|
|
double precision, intent(in) :: FMO(LDFMO,*)
|
|
double precision, intent(out) :: FAO(LDFAO,*)
|
|
|
|
double precision, allocatable :: T(:,:), M(:,:)
|
|
! F_ao = S C F_mo C^t S
|
|
allocate (T(ao_num_align,ao_num),M(ao_num_align,ao_num))
|
|
call dgemm('N','N', ao_num,ao_num,ao_num, 1.d0, &
|
|
ao_overlap, size(ao_overlap,1), &
|
|
mo_coef, size(mo_coef,1), &
|
|
0.d0, &
|
|
M, size(M,1))
|
|
call dgemm('N','N', ao_num,mo_tot_num,mo_tot_num, 1.d0, &
|
|
M, size(M,1), &
|
|
FMO, size(FMO,1), &
|
|
0.d0, &
|
|
T, size(T,1))
|
|
call dgemm('N','T', mo_tot_num,ao_num,mo_tot_num, 1.d0, &
|
|
T, size(T,1), &
|
|
mo_coef, size(mo_coef,1), &
|
|
0.d0, &
|
|
M, size(M,1))
|
|
call dgemm('N','N', ao_num,ao_num,ao_num, 1.d0, &
|
|
M, size(M,1), &
|
|
ao_overlap, size(ao_overlap,1), &
|
|
0.d0, &
|
|
FAO, size(FAO,1))
|
|
deallocate(T,M)
|
|
end
|
|
|