9
1
mirror of https://github.com/QuantumPackage/qp2.git synced 2024-12-21 11:03:29 +01:00

added ao_many_one_e_ints/ bi_ortho_mos/

This commit is contained in:
eginer 2023-02-06 18:17:56 +01:00
parent ddf2035d2b
commit 17d8197a67
28 changed files with 6602 additions and 1 deletions

@ -1 +1 @@
Subproject commit 242151e03d1d6bf042387226431d82d35845686a
Subproject commit f40bde0925808bbec0424b57bfcef1b26473a1c8

View File

@ -0,0 +1,5 @@
ao_one_e_ints
ao_two_e_ints
becke_numerical_grid
mo_one_e_ints
dft_utils_in_r

View File

@ -0,0 +1,25 @@
==================
ao_many_one_e_ints
==================
This module contains A LOT of one-electron integrals of the type
A_ij( r ) = \int dr' phi_i(r') w(r,r') phi_j(r')
where r is a point in real space.
+) ao_gaus_gauss.irp.f: w(r,r') is a exp(-(r-r')^2) , and can be multiplied by x/y/z
+) ao_erf_gauss.irp.f : w(r,r') is a exp(-(r-r')^2) erf(mu * |r-r'|)/|r-r'| , and can be multiplied by x/y/z
+) ao_erf_gauss_grad.irp.f: w(r,r') is a exp(-(r-r')^2) erf(mu * |r-r'|)/|r-r'| , and can be multiplied by x/y/z, but evaluated with also one gradient of an AO function.
Fit of a Slater function and corresponding integrals
----------------------------------------------------
The file fit_slat_gauss.irp.f contains many useful providers/routines to fit a Slater function with 20 gaussian.
+) coef_fit_slat_gauss : coefficients of the gaussians to fit e^(-x)
+) expo_fit_slat_gauss : exponents of the gaussians to fit e^(-x)
Integrals involving Slater functions : stg_gauss_int.irp.f
Taylor expansion of full correlation factor
-------------------------------------------
In taylor_exp.irp.f you might find interesting integrals of the type
\int dr' exp( e^{-alpha |r-r|' - beta |r-r'|^2}) phi_i(r') phi_j(r')
evaluated as a Taylor expansion of the exponential.

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,150 @@
subroutine phi_j_erf_mu_r_dxyz_phi(i,j,mu_in, C_center, dxyz_ints)
implicit none
BEGIN_DOC
! dxyz_ints(1/2/3) = int dr phi_i(r) [erf(mu |r - C|)/|r-C|] d/d(x/y/z) phi_i(r)
END_DOC
integer, intent(in) :: i,j
double precision, intent(in) :: mu_in, C_center(3)
double precision, intent(out):: dxyz_ints(3)
integer :: num_A,power_A(3), num_b, power_B(3),power_B_tmp(3)
double precision :: alpha, beta, A_center(3), B_center(3),contrib,NAI_pol_mult_erf,coef,thr
integer :: n_pt_in,l,m,mm
thr = 1.d-12
dxyz_ints = 0.d0
if(ao_overlap_abs(j,i).lt.thr)then
return
endif
n_pt_in = n_pt_max_integrals
! j
num_A = ao_nucl(j)
power_A(1:3)= ao_power(j,1:3)
A_center(1:3) = nucl_coord(num_A,1:3)
! i
num_B = ao_nucl(i)
power_B(1:3)= ao_power(i,1:3)
B_center(1:3) = nucl_coord(num_B,1:3)
do l=1,ao_prim_num(j)
alpha = ao_expo_ordered_transp(l,j)
do m=1,ao_prim_num(i)
beta = ao_expo_ordered_transp(m,i)
coef = ao_coef_normalized_ordered_transp(l,j) * ao_coef_normalized_ordered_transp(m,i)
if(dabs(coef).lt.thr)cycle
do mm = 1, 3
! (d/dx phi_i ) * phi_j
! d/dx * (x - B_x)^b_x exp(-beta * (x -B_x)^2)= [b_x * (x - B_x)^(b_x - 1) - 2 beta * (x - B_x)^(b_x + 1)] exp(-beta * (x -B_x)^2)
!
! first contribution :: b_x (x - B_x)^(b_x-1) :: integral with b_x=>b_x-1 multiplied by b_x
power_B_tmp = power_B
power_B_tmp(mm) += -1
contrib = NAI_pol_mult_erf(A_center,B_center,power_A,power_B_tmp,alpha,beta,C_center,n_pt_in,mu_in)
dxyz_ints(mm) += contrib * dble(power_B(mm)) * coef
! second contribution :: - 2 beta * (x - B_x)^(b_x + 1) :: integral with b_x=> b_x+1 multiplied by -2 * beta
power_B_tmp = power_B
power_B_tmp(mm) += 1
contrib = NAI_pol_mult_erf(A_center,B_center,power_A,power_B_tmp,alpha,beta,C_center,n_pt_in,mu_in)
dxyz_ints(mm) += contrib * (-2.d0 * beta ) * coef
enddo
enddo
enddo
end
subroutine phi_j_erf_mu_r_dxyz_phi_bis(i,j,mu_in, C_center, dxyz_ints)
implicit none
BEGIN_DOC
! dxyz_ints(1/2/3) = int dr phi_j(r) [erf(mu |r - C|)/|r-C|] d/d(x/y/z) phi_i(r)
END_DOC
integer, intent(in) :: i,j
double precision, intent(in) :: mu_in, C_center(3)
double precision, intent(out):: dxyz_ints(3)
integer :: num_A,power_A(3), num_b, power_B(3),power_B_tmp(3)
double precision :: alpha, beta, A_center(3), B_center(3),contrib,NAI_pol_mult_erf
double precision :: thr, coef
integer :: n_pt_in,l,m,mm,kk
thr = 1.d-12
dxyz_ints = 0.d0
if(ao_overlap_abs(j,i).lt.thr)then
return
endif
n_pt_in = n_pt_max_integrals
! j == A
num_A = ao_nucl(j)
power_A(1:3)= ao_power(j,1:3)
A_center(1:3) = nucl_coord(num_A,1:3)
! i == B
num_B = ao_nucl(i)
power_B(1:3)= ao_power(i,1:3)
B_center(1:3) = nucl_coord(num_B,1:3)
dxyz_ints = 0.d0
do l=1,ao_prim_num(j)
alpha = ao_expo_ordered_transp(l,j)
do m=1,ao_prim_num(i)
beta = ao_expo_ordered_transp(m,i)
do kk = 1, 2 ! loop over the extra terms induced by the d/dx/y/z * AO(i)
do mm = 1, 3
power_B_tmp = power_B
power_B_tmp(mm) = power_ord_grad_transp(kk,mm,i)
coef = ao_coef_normalized_ordered_transp(l,j) * ao_coef_ord_grad_transp(kk,mm,m,i)
if(dabs(coef).lt.thr)cycle
contrib = NAI_pol_mult_erf(A_center,B_center,power_A,power_B_tmp,alpha,beta,C_center,n_pt_in,mu_in)
dxyz_ints(mm) += contrib * coef
enddo
enddo
enddo
enddo
end
subroutine phi_j_erf_mu_r_xyz_dxyz_phi(i,j,mu_in, C_center, dxyz_ints)
implicit none
BEGIN_DOC
! dxyz_ints(1/2/3) = int dr phi_j(r) x/y/z [erf(mu |r - C|)/|r-C|] d/d(x/y/z) phi_i(r)
END_DOC
integer, intent(in) :: i,j
double precision, intent(in) :: mu_in, C_center(3)
double precision, intent(out):: dxyz_ints(3)
integer :: num_A,power_A(3), num_b, power_B(3),power_B_tmp(3)
double precision :: alpha, beta, A_center(3), B_center(3),contrib,NAI_pol_mult_erf
double precision :: thr, coef
integer :: n_pt_in,l,m,mm,kk
thr = 1.d-12
dxyz_ints = 0.d0
if(ao_overlap_abs(j,i).lt.thr)then
return
endif
n_pt_in = n_pt_max_integrals
! j == A
num_A = ao_nucl(j)
power_A(1:3)= ao_power(j,1:3)
A_center(1:3) = nucl_coord(num_A,1:3)
! i == B
num_B = ao_nucl(i)
power_B(1:3)= ao_power(i,1:3)
B_center(1:3) = nucl_coord(num_B,1:3)
dxyz_ints = 0.d0
do l=1,ao_prim_num(j)
alpha = ao_expo_ordered_transp(l,j)
do m=1,ao_prim_num(i)
beta = ao_expo_ordered_transp(m,i)
do kk = 1, 4 ! loop over the extra terms induced by the x/y/z * d dx/y/z AO(i)
do mm = 1, 3
power_B_tmp = power_B
power_B_tmp(mm) = power_ord_xyz_grad_transp(kk,mm,i)
coef = ao_coef_normalized_ordered_transp(l,j) * ao_coef_ord_xyz_grad_transp(kk,mm,m,i)
if(dabs(coef).lt.thr)cycle
contrib = NAI_pol_mult_erf(A_center,B_center,power_A,power_B_tmp,alpha,beta,C_center,n_pt_in,mu_in)
dxyz_ints(mm) += contrib * coef
enddo
enddo
enddo
enddo
end

View File

@ -0,0 +1,426 @@
! ---
subroutine overlap_gauss_xyz_r12_ao(D_center,delta,i,j,gauss_ints)
implicit none
BEGIN_DOC
! gauss_ints(m) = \int dr AO_i(r) AO_j(r) x/y/z e^{-delta |r-D_center|^2}
!
! with m == 1 ==> x, m == 2 ==> y, m == 3 ==> z
END_DOC
integer, intent(in) :: i,j
double precision, intent(in) :: D_center(3), delta
double precision, intent(out) :: gauss_ints(3)
integer :: num_a,num_b,power_A(3), power_B(3),l,k,m
double precision :: A_center(3), B_center(3),overlap_gauss_r12,alpha,beta,gauss_ints_tmp(3)
gauss_ints = 0.d0
if(ao_overlap_abs(j,i).lt.1.d-12)then
return
endif
num_A = ao_nucl(i)
power_A(1:3)= ao_power(i,1:3)
A_center(1:3) = nucl_coord(num_A,1:3)
num_B = ao_nucl(j)
power_B(1:3)= ao_power(j,1:3)
B_center(1:3) = nucl_coord(num_B,1:3)
do l=1,ao_prim_num(i)
alpha = ao_expo_ordered_transp(l,i)
do k=1,ao_prim_num(j)
beta = ao_expo_ordered_transp(k,j)
call overlap_gauss_xyz_r12(D_center,delta,A_center,B_center,power_A,power_B,alpha,beta,gauss_ints_tmp)
do m = 1, 3
gauss_ints(m) += gauss_ints_tmp(m) * ao_coef_normalized_ordered_transp(l,i) &
* ao_coef_normalized_ordered_transp(k,j)
enddo
enddo
enddo
end
double precision function overlap_gauss_xyz_r12_ao_specific(D_center,delta,i,j,mx)
implicit none
BEGIN_DOC
! \int dr AO_i(r) AO_j(r) x/y/z e^{-delta |r-D_center|^2}
!
! with mx == 1 ==> x, mx == 2 ==> y, mx == 3 ==> z
END_DOC
integer, intent(in) :: i,j,mx
double precision, intent(in) :: D_center(3), delta
integer :: num_a,num_b,power_A(3), power_B(3),l,k
double precision :: gauss_int
double precision :: A_center(3), B_center(3),overlap_gauss_r12,alpha,beta
double precision :: overlap_gauss_xyz_r12_specific
overlap_gauss_xyz_r12_ao_specific = 0.d0
if(ao_overlap_abs(j,i).lt.1.d-12)then
return
endif
num_A = ao_nucl(i)
power_A(1:3)= ao_power(i,1:3)
A_center(1:3) = nucl_coord(num_A,1:3)
num_B = ao_nucl(j)
power_B(1:3)= ao_power(j,1:3)
B_center(1:3) = nucl_coord(num_B,1:3)
do l=1,ao_prim_num(i)
alpha = ao_expo_ordered_transp(l,i)
do k=1,ao_prim_num(j)
beta = ao_expo_ordered_transp(k,j)
gauss_int = overlap_gauss_xyz_r12_specific(D_center,delta,A_center,B_center,power_A,power_B,alpha,beta,mx)
overlap_gauss_xyz_r12_ao_specific = gauss_int * ao_coef_normalized_ordered_transp(l,i) &
* ao_coef_normalized_ordered_transp(k,j)
enddo
enddo
end
subroutine overlap_gauss_r12_all_ao(D_center,delta,aos_ints)
implicit none
double precision, intent(in) :: D_center(3), delta
double precision, intent(out):: aos_ints(ao_num,ao_num)
integer :: num_a,num_b,power_A(3), power_B(3),l,k,i,j
double precision :: A_center(3), B_center(3),overlap_gauss_r12,alpha,beta,analytical_j
aos_ints = 0.d0
do i = 1, ao_num
do j = 1, ao_num
if(ao_overlap_abs(j,i).lt.1.d-12)cycle
num_A = ao_nucl(i)
power_A(1:3)= ao_power(i,1:3)
A_center(1:3) = nucl_coord(num_A,1:3)
num_B = ao_nucl(j)
power_B(1:3)= ao_power(j,1:3)
B_center(1:3) = nucl_coord(num_B,1:3)
do l=1,ao_prim_num(i)
alpha = ao_expo_ordered_transp(l,i)
do k=1,ao_prim_num(j)
beta = ao_expo_ordered_transp(k,j)
analytical_j = overlap_gauss_r12(D_center,delta,A_center,B_center,power_A,power_B,alpha,beta)
aos_ints(j,i) += analytical_j * ao_coef_normalized_ordered_transp(l,i) &
* ao_coef_normalized_ordered_transp(k,j)
enddo
enddo
enddo
enddo
end
! ---
! TODO :: PUT CYCLES IN LOOPS
double precision function overlap_gauss_r12_ao(D_center, delta, i, j)
BEGIN_DOC
! \int dr AO_i(r) AO_j(r) e^{-delta |r-D_center|^2}
END_DOC
implicit none
integer, intent(in) :: i, j
double precision, intent(in) :: D_center(3), delta
integer :: power_A(3), power_B(3), l, k
double precision :: A_center(3), B_center(3), alpha, beta, coef, coef1, analytical_j
double precision, external :: overlap_gauss_r12
overlap_gauss_r12_ao = 0.d0
if(ao_overlap_abs(j,i).lt.1.d-12) then
return
endif
power_A(1:3) = ao_power(i,1:3)
power_B(1:3) = ao_power(j,1:3)
A_center(1:3) = nucl_coord(ao_nucl(i),1:3)
B_center(1:3) = nucl_coord(ao_nucl(j),1:3)
do l = 1, ao_prim_num(i)
alpha = ao_expo_ordered_transp (l,i)
coef1 = ao_coef_normalized_ordered_transp(l,i)
do k = 1, ao_prim_num(j)
beta = ao_expo_ordered_transp(k,j)
coef = coef1 * ao_coef_normalized_ordered_transp(k,j)
if(dabs(coef) .lt. 1d-12) cycle
analytical_j = overlap_gauss_r12(D_center, delta, A_center, B_center, power_A, power_B, alpha, beta)
overlap_gauss_r12_ao += coef * analytical_j
enddo
enddo
end function overlap_gauss_r12_ao
! --
double precision function overlap_abs_gauss_r12_ao(D_center, delta, i, j)
BEGIN_DOC
! \int dr AO_i(r) AO_j(r) e^{-delta |r-D_center|^2}
END_DOC
implicit none
integer, intent(in) :: i, j
double precision, intent(in) :: D_center(3), delta
integer :: power_A(3), power_B(3), l, k
double precision :: A_center(3), B_center(3), alpha, beta, coef, coef1, analytical_j
double precision, external :: overlap_abs_gauss_r12
overlap_abs_gauss_r12_ao = 0.d0
if(ao_overlap_abs(j,i).lt.1.d-12) then
return
endif
power_A(1:3) = ao_power(i,1:3)
power_B(1:3) = ao_power(j,1:3)
A_center(1:3) = nucl_coord(ao_nucl(i),1:3)
B_center(1:3) = nucl_coord(ao_nucl(j),1:3)
do l = 1, ao_prim_num(i)
alpha = ao_expo_ordered_transp (l,i)
coef1 = ao_coef_normalized_ordered_transp(l,i)
do k = 1, ao_prim_num(j)
beta = ao_expo_ordered_transp(k,j)
coef = coef1 * ao_coef_normalized_ordered_transp(k,j)
if(dabs(coef) .lt. 1d-12) cycle
analytical_j = overlap_abs_gauss_r12(D_center, delta, A_center, B_center, power_A, power_B, alpha, beta)
overlap_abs_gauss_r12_ao += dabs(coef * analytical_j)
enddo
enddo
end function overlap_gauss_r12_ao
! --
subroutine overlap_gauss_r12_ao_v(D_center, LD_D, delta, i, j, resv, LD_resv, n_points)
BEGIN_DOC
!
! \int dr AO_i(r) AO_j(r) e^{-delta |r-D_center|^2}
!
! n_points: nb of integrals <= min(LD_D, LD_resv)
!
END_DOC
implicit none
integer, intent(in) :: i, j, LD_D, LD_resv, n_points
double precision, intent(in) :: D_center(LD_D,3), delta
double precision, intent(out) :: resv(LD_resv)
integer :: ipoint
integer :: power_A(3), power_B(3), l, k
double precision :: A_center(3), B_center(3), alpha, beta, coef, coef1
double precision, allocatable :: analytical_j(:)
resv(:) = 0.d0
if(ao_overlap_abs(j,i) .lt. 1.d-12) then
return
endif
power_A(1:3) = ao_power(i,1:3)
power_B(1:3) = ao_power(j,1:3)
A_center(1:3) = nucl_coord(ao_nucl(i),1:3)
B_center(1:3) = nucl_coord(ao_nucl(j),1:3)
allocate(analytical_j(n_points))
do l = 1, ao_prim_num(i)
alpha = ao_expo_ordered_transp (l,i)
coef1 = ao_coef_normalized_ordered_transp(l,i)
do k = 1, ao_prim_num(j)
beta = ao_expo_ordered_transp(k,j)
coef = coef1 * ao_coef_normalized_ordered_transp(k,j)
if(dabs(coef) .lt. 1d-12) cycle
call overlap_gauss_r12_v(D_center, LD_D, delta, A_center, B_center, power_A, power_B, alpha, beta, analytical_j, n_points, n_points)
do ipoint = 1, n_points
resv(ipoint) = resv(ipoint) + coef * analytical_j(ipoint)
enddo
enddo
enddo
deallocate(analytical_j)
end subroutine overlap_gauss_r12_ao_v
! ---
double precision function overlap_gauss_r12_ao_with1s(B_center, beta, D_center, delta, i, j)
BEGIN_DOC
!
! \int dr AO_i(r) AO_j(r) e^{-beta |r-B_center^2|} e^{-delta |r-D_center|^2}
!
END_DOC
implicit none
integer, intent(in) :: i, j
double precision, intent(in) :: B_center(3), beta, D_center(3), delta
integer :: power_A1(3), power_A2(3), l, k
double precision :: A1_center(3), A2_center(3), alpha1, alpha2, coef1, coef12, analytical_j
double precision :: G_center(3), gama, fact_g, gama_inv
double precision, external :: overlap_gauss_r12, overlap_gauss_r12_ao
if(beta .lt. 1d-10) then
overlap_gauss_r12_ao_with1s = overlap_gauss_r12_ao(D_center, delta, i, j)
return
endif
overlap_gauss_r12_ao_with1s = 0.d0
if(ao_overlap_abs(j,i) .lt. 1.d-12) then
return
endif
! e^{-beta |r-B_center^2|} e^{-delta |r-D_center|^2} = fact_g e^{-gama |r - G|^2}
gama = beta + delta
gama_inv = 1.d0 / gama
G_center(1) = (beta * B_center(1) + delta * D_center(1)) * gama_inv
G_center(2) = (beta * B_center(2) + delta * D_center(2)) * gama_inv
G_center(3) = (beta * B_center(3) + delta * D_center(3)) * gama_inv
fact_g = beta * delta * gama_inv * ( (B_center(1) - D_center(1)) * (B_center(1) - D_center(1)) &
+ (B_center(2) - D_center(2)) * (B_center(2) - D_center(2)) &
+ (B_center(3) - D_center(3)) * (B_center(3) - D_center(3)) )
if(fact_g .gt. 10d0) return
fact_g = dexp(-fact_g)
! ---
power_A1(1:3) = ao_power(i,1:3)
power_A2(1:3) = ao_power(j,1:3)
A1_center(1:3) = nucl_coord(ao_nucl(i),1:3)
A2_center(1:3) = nucl_coord(ao_nucl(j),1:3)
do l = 1, ao_prim_num(i)
alpha1 = ao_expo_ordered_transp (l,i)
coef1 = fact_g * ao_coef_normalized_ordered_transp(l,i)
if(dabs(coef1) .lt. 1d-12) cycle
do k = 1, ao_prim_num(j)
alpha2 = ao_expo_ordered_transp (k,j)
coef12 = coef1 * ao_coef_normalized_ordered_transp(k,j)
if(dabs(coef12) .lt. 1d-12) cycle
analytical_j = overlap_gauss_r12(G_center, gama, A1_center, A2_center, power_A1, power_A2, alpha1, alpha2)
overlap_gauss_r12_ao_with1s += coef12 * analytical_j
enddo
enddo
end function overlap_gauss_r12_ao_with1s
! ---
subroutine overlap_gauss_r12_ao_with1s_v(B_center, beta, D_center, LD_D, delta, i, j, resv, LD_resv, n_points)
BEGIN_DOC
!
! \int dr AO_i(r) AO_j(r) e^{-beta |r-B_center^2|} e^{-delta |r-D_center|^2}
! using an array of D_centers.
!
END_DOC
implicit none
integer, intent(in) :: i, j, n_points, LD_D, LD_resv
double precision, intent(in) :: B_center(3), beta, D_center(LD_D,3), delta
double precision, intent(out) :: resv(LD_resv)
integer :: ipoint
integer :: power_A1(3), power_A2(3), l, k
double precision :: A1_center(3), A2_center(3), alpha1, alpha2, coef1
double precision :: coef12, coef12f
double precision :: gama, gama_inv
double precision :: bg, dg, bdg
double precision, allocatable :: fact_g(:), G_center(:,:), analytical_j(:)
if(ao_overlap_abs(j,i) .lt. 1.d-12) then
return
endif
ASSERT(beta .gt. 0.d0)
if(beta .lt. 1d-10) then
call overlap_gauss_r12_ao_v(D_center, LD_D, delta, i, j, resv, LD_resv, n_points)
return
endif
resv(:) = 0.d0
! e^{-beta |r-B_center^2|} e^{-delta |r-D_center|^2} = fact_g e^{-gama |r - G|^2}
gama = beta + delta
gama_inv = 1.d0 / gama
power_A1(1:3) = ao_power(i,1:3)
power_A2(1:3) = ao_power(j,1:3)
A1_center(1:3) = nucl_coord(ao_nucl(i),1:3)
A2_center(1:3) = nucl_coord(ao_nucl(j),1:3)
allocate(fact_g(n_points), G_center(n_points,3), analytical_j(n_points))
bg = beta * gama_inv
dg = delta * gama_inv
bdg = bg * delta
do ipoint = 1, n_points
G_center(ipoint,1) = bg * B_center(1) + dg * D_center(ipoint,1)
G_center(ipoint,2) = bg * B_center(2) + dg * D_center(ipoint,2)
G_center(ipoint,3) = bg * B_center(3) + dg * D_center(ipoint,3)
fact_g(ipoint) = bdg * ( (B_center(1) - D_center(ipoint,1)) * (B_center(1) - D_center(ipoint,1)) &
+ (B_center(2) - D_center(ipoint,2)) * (B_center(2) - D_center(ipoint,2)) &
+ (B_center(3) - D_center(ipoint,3)) * (B_center(3) - D_center(ipoint,3)) )
if(fact_g(ipoint) < 10d0) then
fact_g(ipoint) = dexp(-fact_g(ipoint))
else
fact_g(ipoint) = 0.d0
endif
enddo
do l = 1, ao_prim_num(i)
alpha1 = ao_expo_ordered_transp (l,i)
coef1 = ao_coef_normalized_ordered_transp(l,i)
do k = 1, ao_prim_num(j)
alpha2 = ao_expo_ordered_transp (k,j)
coef12 = coef1 * ao_coef_normalized_ordered_transp(k,j)
if(dabs(coef12) .lt. 1d-12) cycle
call overlap_gauss_r12_v(G_center, n_points, gama, A1_center, A2_center, power_A1, power_A2, alpha1, alpha2, analytical_j, n_points, n_points)
do ipoint = 1, n_points
coef12f = coef12 * fact_g(ipoint)
resv(ipoint) += coef12f * analytical_j(ipoint)
enddo
enddo
enddo
deallocate(fact_g, G_center, analytical_j)
end subroutine overlap_gauss_r12_ao_with1s_v
! ---

View File

@ -0,0 +1,94 @@
BEGIN_PROVIDER [integer, n_max_fit_slat]
implicit none
BEGIN_DOC
! number of gaussian to fit exp(-x)
!
! I took 20 gaussians from the program bassto.f
END_DOC
n_max_fit_slat = 20
END_PROVIDER
BEGIN_PROVIDER [double precision, coef_fit_slat_gauss, (n_max_fit_slat)]
&BEGIN_PROVIDER [double precision, expo_fit_slat_gauss, (n_max_fit_slat)]
implicit none
include 'constants.include.F'
BEGIN_DOC
! fit the exp(-x) as
!
! \sum_{i = 1, n_max_fit_slat} coef_fit_slat_gauss(i) * exp(-expo_fit_slat_gauss(i) * x**2)
!
! The coefficient are taken from the program bassto.f
END_DOC
expo_fit_slat_gauss(01)=30573.77073000000
coef_fit_slat_gauss(01)=0.00338925525
expo_fit_slat_gauss(02)=5608.45238100000
coef_fit_slat_gauss(02)=0.00536433869
expo_fit_slat_gauss(03)=1570.95673400000
coef_fit_slat_gauss(03)=0.00818702846
expo_fit_slat_gauss(04)=541.39785110000
coef_fit_slat_gauss(04)=0.01202047655
expo_fit_slat_gauss(05)=212.43469630000
coef_fit_slat_gauss(05)=0.01711289568
expo_fit_slat_gauss(06)=91.31444574000
coef_fit_slat_gauss(06)=0.02376001022
expo_fit_slat_gauss(07)=42.04087246000
coef_fit_slat_gauss(07)=0.03229121736
expo_fit_slat_gauss(08)=20.43200443000
coef_fit_slat_gauss(08)=0.04303646818
expo_fit_slat_gauss(09)=10.37775161000
coef_fit_slat_gauss(09)=0.05624657578
expo_fit_slat_gauss(10)=5.46880754500
coef_fit_slat_gauss(10)=0.07192311571
expo_fit_slat_gauss(11)=2.97373529200
coef_fit_slat_gauss(11)=0.08949389001
expo_fit_slat_gauss(12)=1.66144190200
coef_fit_slat_gauss(12)=0.10727599240
expo_fit_slat_gauss(13)=0.95052560820
coef_fit_slat_gauss(13)=0.12178961750
expo_fit_slat_gauss(14)=0.55528683970
coef_fit_slat_gauss(14)=0.12740141870
expo_fit_slat_gauss(15)=0.33043360020
coef_fit_slat_gauss(15)=0.11759168160
expo_fit_slat_gauss(16)=0.19982303230
coef_fit_slat_gauss(16)=0.08953504394
expo_fit_slat_gauss(17)=0.12246840760
coef_fit_slat_gauss(17)=0.05066721317
expo_fit_slat_gauss(18)=0.07575825322
coef_fit_slat_gauss(18)=0.01806363869
expo_fit_slat_gauss(19)=0.04690146243
coef_fit_slat_gauss(19)=0.00305632563
expo_fit_slat_gauss(20)=0.02834749861
coef_fit_slat_gauss(20)=0.00013317513
END_PROVIDER
double precision function slater_fit_gam(x,gam)
implicit none
double precision, intent(in) :: x,gam
BEGIN_DOC
! fit of the function exp(-gam * x) with gaussian functions
END_DOC
integer :: i
slater_fit_gam = 0.d0
do i = 1, n_max_fit_slat
slater_fit_gam += coef_fit_slat_gauss(i) * dexp(-expo_fit_slat_gauss(i) * gam * gam * x * x)
enddo
end
subroutine expo_fit_slater_gam(gam,expos)
implicit none
BEGIN_DOC
! returns the array of the exponents of the gaussians to fit exp(-gam*x)
END_DOC
double precision, intent(in) :: gam
double precision, intent(out) :: expos(n_max_fit_slat)
integer :: i
do i = 1, n_max_fit_slat
expos(i) = expo_fit_slat_gauss(i) * gam * gam
enddo
end

View File

@ -0,0 +1,517 @@
! ---
BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2_test, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! -\frac{1}{4} x int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 [1 - erf(mu r12)]^2
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), expo_fit, coef_fit
double precision :: coef, beta, B_center(3)
double precision :: tmp
double precision :: wall0, wall1
double precision :: int_gauss, dsqpi_3_2, int_j1b
double precision :: factor_ij_1s, beta_ij, center_ij_1s(3), sq_pi_3_2
double precision, allocatable :: int_fit_v(:)
double precision, external :: overlap_gauss_r12_ao_with1s
print*, ' providing int2_grad1u2_grad2u2_j1b2_test ...'
sq_pi_3_2 = (dacos(-1.d0))**(1.5d0)
provide mu_erf final_grid_points_transp j1b_pen List_comb_thr_b3_coef
call wall_time(wall0)
int2_grad1u2_grad2u2_j1b2_test(:,:,:) = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit_v, tmp,int_gauss,int_j1b,factor_ij_1s,beta_ij,center_ij_1s) &
!$OMP SHARED (n_points_final_grid, ao_num, final_grid_points,List_comb_thr_b3_size, &
!$OMP final_grid_points_transp, ng_fit_jast, &
!$OMP expo_gauss_1_erf_x_2, coef_gauss_1_erf_x_2, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo, &
!$OMP List_comb_thr_b3_cent, int2_grad1u2_grad2u2_j1b2_test, ao_abs_comb_b3_j1b, &
!$OMP ao_overlap_abs,sq_pi_3_2)
!$OMP DO SCHEDULE(dynamic)
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
if(ao_overlap_abs(j,i) .lt. 1.d-12) then
cycle
endif
do i_1s = 1, List_comb_thr_b3_size(j,i)
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_1_erf_x_2(i_fit)
!DIR$ FORCEINLINE
call gaussian_product(expo_fit,r,beta,B_center,factor_ij_1s,beta_ij,center_ij_1s)
coef_fit = -0.25d0 * coef_gauss_1_erf_x_2(i_fit) * coef
! if(dabs(coef_fit*factor_ij_1s*int_j1b).lt.1.d-10)cycle ! old version
if(dabs(coef_fit*factor_ij_1s*int_j1b*sq_pi_3_2*(beta_ij)**(-1.5d0)).lt.1.d-10)cycle
! call overlap_gauss_r12_ao_with1s_v(B_center, beta, final_grid_points_transp, &
! expo_fit, i, j, int_fit_v, n_points_final_grid)
int_gauss = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
int2_grad1u2_grad2u2_j1b2_test(j,i,ipoint) += coef_fit * int_gauss
enddo
enddo
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 1, ao_num
do j = 1, i-1
int2_grad1u2_grad2u2_j1b2_test(j,i,ipoint) = int2_grad1u2_grad2u2_j1b2_test(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_grad1u2_grad2u2_j1b2_test', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2_test_v, (ao_num, ao_num, n_points_final_grid)]
!
! BEGIN_DOC
! !
! ! -\frac{1}{4} x int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 [1 - erf(mu r12)]^2
! !
! END_DOC
!
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), expo_fit, coef_fit
double precision :: coef, beta, B_center(3)
double precision :: tmp
double precision :: wall0, wall1
double precision, allocatable :: int_fit_v(:),big_array(:,:,:)
double precision, external :: overlap_gauss_r12_ao_with1s
print*, ' providing int2_grad1u2_grad2u2_j1b2_test_v ...'
provide mu_erf final_grid_points_transp j1b_pen
call wall_time(wall0)
double precision :: int_j1b
big_array(:,:,:) = 0.d0
allocate(big_array(n_points_final_grid,ao_num, ao_num))
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center,&
!$OMP coef_fit, expo_fit, int_fit_v, tmp,int_j1b) &
!$OMP SHARED (n_points_final_grid, ao_num, List_comb_thr_b3_size,&
!$OMP final_grid_points_transp, ng_fit_jast, &
!$OMP expo_gauss_1_erf_x_2, coef_gauss_1_erf_x_2, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo, &
!$OMP List_comb_thr_b3_cent, big_array,&
!$OMP ao_abs_comb_b3_j1b,ao_overlap_abs)
!
allocate(int_fit_v(n_points_final_grid))
!$OMP DO SCHEDULE(dynamic)
do i = 1, ao_num
do j = i, ao_num
if(ao_overlap_abs(j,i) .lt. 1.d-12) then
cycle
endif
do i_1s = 1, List_comb_thr_b3_size(j,i)
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
! if(dabs(coef)*dabs(int_j1b).lt.1.d-15)cycle
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_1_erf_x_2(i_fit)
coef_fit = -0.25d0 * coef_gauss_1_erf_x_2(i_fit) * coef
call overlap_gauss_r12_ao_with1s_v(B_center, beta, final_grid_points_transp, size(final_grid_points_transp,1),&
expo_fit, i, j, int_fit_v, size(int_fit_v,1),n_points_final_grid)
do ipoint = 1, n_points_final_grid
big_array(ipoint,j,i) += coef_fit * int_fit_v(ipoint)
enddo
enddo
enddo
enddo
enddo
!$OMP END DO
deallocate(int_fit_v)
!$OMP END PARALLEL
do i = 1, ao_num
do j = i, ao_num
do ipoint = 1, n_points_final_grid
int2_grad1u2_grad2u2_j1b2_test_v(j,i,ipoint) = big_array(ipoint,j,i)
enddo
enddo
enddo
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
int2_grad1u2_grad2u2_j1b2_test_v(j,i,ipoint) = big_array(ipoint,i,j)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_grad1u2_grad2u2_j1b2_test_v', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, int2_u2_j1b2_test, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 [u_12^mu]^2
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit, expo_fit, coef_fit
double precision :: coef, beta, B_center(3), tmp
double precision :: wall0, wall1,int_j1b
double precision, external :: overlap_gauss_r12_ao
double precision, external :: overlap_gauss_r12_ao_with1s
double precision :: factor_ij_1s,beta_ij,center_ij_1s(3),sq_pi_3_2
print*, ' providing int2_u2_j1b2_test ...'
sq_pi_3_2 = (dacos(-1.d0))**(1.5d0)
provide mu_erf final_grid_points j1b_pen
call wall_time(wall0)
int2_u2_j1b2_test = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit, tmp, int_j1b,factor_ij_1s,beta_ij,center_ij_1s) &
!$OMP SHARED (n_points_final_grid, ao_num, List_comb_thr_b3_size, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_x_2, coef_gauss_j_mu_x_2, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo,sq_pi_3_2, &
!$OMP List_comb_thr_b3_cent, int2_u2_j1b2_test,ao_abs_comb_b3_j1b)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
tmp = 0.d0
do i_1s = 1, List_comb_thr_b3_size(j,i)
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_x_2(i_fit)
coef_fit = coef_gauss_j_mu_x_2(i_fit)
!DIR$ FORCEINLINE
call gaussian_product(expo_fit,r,beta,B_center,factor_ij_1s,beta_ij,center_ij_1s)
! if(dabs(coef_fit*coef*factor_ij_1s*int_j1b).lt.1.d-10)cycle ! old version
if(dabs(coef_fit*coef*factor_ij_1s*int_j1b*sq_pi_3_2*(beta_ij)**(-1.5d0)).lt.1.d-10)cycle
! ---
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += coef * coef_fit * int_fit
enddo
! ---
enddo
int2_u2_j1b2_test(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
int2_u2_j1b2_test(j,i,ipoint) = int2_u2_j1b2_test(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_u2_j1b2_test', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, int2_u_grad1u_x_j1b2_test, (ao_num, ao_num, n_points_final_grid, 3)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 u_12^mu [\grad_1 u_12^mu] r2
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit(3), expo_fit, coef_fit
double precision :: coef, beta, B_center(3), dist
double precision :: alpha_1s, alpha_1s_inv, centr_1s(3), expo_coef_1s, coef_tmp
double precision :: tmp_x, tmp_y, tmp_z, int_j1b
double precision :: wall0, wall1, sq_pi_3_2,sq_alpha
print*, ' providing int2_u_grad1u_x_j1b2_test ...'
sq_pi_3_2 = dacos(-1.D0)**(1.d0)
provide mu_erf final_grid_points j1b_pen
call wall_time(wall0)
int2_u_grad1u_x_j1b2_test = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit, alpha_1s, dist, &
!$OMP alpha_1s_inv, centr_1s, expo_coef_1s, coef_tmp, &
!$OMP tmp_x, tmp_y, tmp_z,int_j1b,sq_alpha) &
!$OMP SHARED (n_points_final_grid, ao_num, List_comb_thr_b3_size, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_1_erf, coef_gauss_j_mu_1_erf, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo, &
!$OMP List_comb_thr_b3_cent, int2_u_grad1u_x_j1b2_test,ao_abs_comb_b3_j1b,sq_pi_3_2)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
tmp_x = 0.d0
tmp_y = 0.d0
tmp_z = 0.d0
do i_1s = 1, List_comb_thr_b3_size(j,i)
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_1_erf(i_fit)
coef_fit = coef_gauss_j_mu_1_erf(i_fit)
dist = (B_center(1) - r(1)) * (B_center(1) - r(1)) &
+ (B_center(2) - r(2)) * (B_center(2) - r(2)) &
+ (B_center(3) - r(3)) * (B_center(3) - r(3))
alpha_1s = beta + expo_fit
alpha_1s_inv = 1.d0 / alpha_1s
centr_1s(1) = alpha_1s_inv * (beta * B_center(1) + expo_fit * r(1))
centr_1s(2) = alpha_1s_inv * (beta * B_center(2) + expo_fit * r(2))
centr_1s(3) = alpha_1s_inv * (beta * B_center(3) + expo_fit * r(3))
expo_coef_1s = beta * expo_fit * alpha_1s_inv * dist
coef_tmp = coef * coef_fit * dexp(-expo_coef_1s)
sq_alpha = alpha_1s_inv * dsqrt(alpha_1s_inv)
! if(dabs(coef_tmp*int_j1b) .lt. 1d-10) cycle ! old version
if(dabs(coef_tmp*int_j1b*sq_pi_3_2*sq_alpha) .lt. 1d-10) cycle
call NAI_pol_x_mult_erf_ao_with1s(i, j, alpha_1s, centr_1s, 1.d+9, r, int_fit)
tmp_x += coef_tmp * int_fit(1)
tmp_y += coef_tmp * int_fit(2)
tmp_z += coef_tmp * int_fit(3)
enddo
! ---
enddo
int2_u_grad1u_x_j1b2_test(j,i,ipoint,1) = tmp_x
int2_u_grad1u_x_j1b2_test(j,i,ipoint,2) = tmp_y
int2_u_grad1u_x_j1b2_test(j,i,ipoint,3) = tmp_z
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
int2_u_grad1u_x_j1b2_test(j,i,ipoint,1) = int2_u_grad1u_x_j1b2_test(i,j,ipoint,1)
int2_u_grad1u_x_j1b2_test(j,i,ipoint,2) = int2_u_grad1u_x_j1b2_test(i,j,ipoint,2)
int2_u_grad1u_x_j1b2_test(j,i,ipoint,3) = int2_u_grad1u_x_j1b2_test(i,j,ipoint,3)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_u_grad1u_x_j1b2_test', wall1 - wall0
END_PROVIDER
BEGIN_PROVIDER [ double precision, int2_u_grad1u_j1b2_test, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 u_12^mu [\grad_1 u_12^mu]
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit, expo_fit, coef_fit, coef_tmp
double precision :: coef, beta, B_center(3), dist
double precision :: alpha_1s, alpha_1s_inv, centr_1s(3), expo_coef_1s, tmp
double precision :: wall0, wall1
double precision, external :: NAI_pol_mult_erf_ao_with1s
double precision :: j12_mu_r12,int_j1b
double precision :: sigma_ij,dist_ij_ipoint,dsqpi_3_2
double precision :: beta_ij,center_ij_1s(3),factor_ij_1s
print*, ' providing int2_u_grad1u_j1b2_test ...'
dsqpi_3_2 = (dacos(-1.d0))**(1.5d0)
provide mu_erf final_grid_points j1b_pen ao_overlap_abs List_comb_thr_b3_cent
call wall_time(wall0)
int2_u_grad1u_j1b2_test = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit, tmp, alpha_1s, dist, &
!$OMP beta_ij,center_ij_1s,factor_ij_1s, &
!$OMP int_j1b,alpha_1s_inv, centr_1s, expo_coef_1s, coef_tmp) &
!$OMP SHARED (n_points_final_grid, ao_num, List_comb_thr_b3_size, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_1_erf, coef_gauss_j_mu_1_erf, &
!$OMP ao_prod_dist_grid, ao_prod_sigma, ao_overlap_abs_grid,ao_prod_center,dsqpi_3_2, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo, ao_abs_comb_b3_j1b, &
!$OMP List_comb_thr_b3_cent, int2_u_grad1u_j1b2_test)
!$OMP DO
do ipoint = 1, n_points_final_grid
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-10)cycle
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
tmp = 0.d0
do i_1s = 1, List_comb_thr_b3_size(j,i)
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
dist = (B_center(1) - r(1)) * (B_center(1) - r(1)) &
+ (B_center(2) - r(2)) * (B_center(2) - r(2)) &
+ (B_center(3) - r(3)) * (B_center(3) - r(3))
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_1_erf(i_fit)
call gaussian_product(expo_fit,r,beta,B_center,factor_ij_1s,beta_ij,center_ij_1s)
if(factor_ij_1s*dabs(coef*int_j1b)*dsqpi_3_2*beta_ij**(-1.5d0).lt.1.d-15)cycle
coef_fit = coef_gauss_j_mu_1_erf(i_fit)
alpha_1s = beta + expo_fit
alpha_1s_inv = 1.d0 / alpha_1s
centr_1s(1) = alpha_1s_inv * (beta * B_center(1) + expo_fit * r(1))
centr_1s(2) = alpha_1s_inv * (beta * B_center(2) + expo_fit * r(2))
centr_1s(3) = alpha_1s_inv * (beta * B_center(3) + expo_fit * r(3))
expo_coef_1s = beta * expo_fit * alpha_1s_inv * dist
if(expo_coef_1s .gt. 20.d0) cycle
coef_tmp = coef * coef_fit * dexp(-expo_coef_1s)
if(dabs(coef_tmp) .lt. 1d-08) cycle
int_fit = NAI_pol_mult_erf_ao_with1s(i, j, alpha_1s, centr_1s, 1.d+9, r)
tmp += coef_tmp * int_fit
enddo
enddo
int2_u_grad1u_j1b2_test(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
int2_u_grad1u_j1b2_test(j,i,ipoint) = int2_u_grad1u_j1b2_test(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_u_grad1u_j1b2_test', wall1 - wall0
END_PROVIDER
! ---

View File

@ -0,0 +1,420 @@
! ---
BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! -\frac{1}{4} x int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 [1 - erf(mu r12)]^2
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit, expo_fit, coef_fit
double precision :: coef, beta, B_center(3)
double precision :: tmp
double precision :: wall0, wall1
double precision, external :: overlap_gauss_r12_ao
double precision, external :: overlap_gauss_r12_ao_with1s
print*, ' providing int2_grad1u2_grad2u2_j1b2 ...'
call wall_time(wall0)
provide mu_erf final_grid_points j1b_pen
int2_grad1u2_grad2u2_j1b2 = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit, tmp) &
!$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b3_size, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_1_erf_x_2, coef_gauss_1_erf_x_2, &
!$OMP List_all_comb_b3_coef, List_all_comb_b3_expo, &
!$OMP List_all_comb_b3_cent, int2_grad1u2_grad2u2_j1b2)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
tmp = 0.d0
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_1_erf_x_2(i_fit)
coef_fit = coef_gauss_1_erf_x_2(i_fit)
! ---
int_fit = overlap_gauss_r12_ao(r, expo_fit, i, j)
tmp += -0.25d0 * coef_fit * int_fit
! if(dabs(coef_fit*int_fit) .lt. 1d-12) cycle
! ---
do i_1s = 2, List_all_comb_b3_size
coef = List_all_comb_b3_coef (i_1s)
beta = List_all_comb_b3_expo (i_1s)
B_center(1) = List_all_comb_b3_cent(1,i_1s)
B_center(2) = List_all_comb_b3_cent(2,i_1s)
B_center(3) = List_all_comb_b3_cent(3,i_1s)
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += -0.25d0 * coef * coef_fit * int_fit
enddo
! ---
enddo
int2_grad1u2_grad2u2_j1b2(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
int2_grad1u2_grad2u2_j1b2(j,i,ipoint) = int2_grad1u2_grad2u2_j1b2(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_grad1u2_grad2u2_j1b2 =', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, int2_u2_j1b2, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 [u_12^mu]^2
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit, expo_fit, coef_fit
double precision :: coef, beta, B_center(3), tmp
double precision :: wall0, wall1
double precision, external :: overlap_gauss_r12_ao
double precision, external :: overlap_gauss_r12_ao_with1s
print*, ' providing int2_u2_j1b2 ...'
call wall_time(wall0)
provide mu_erf final_grid_points j1b_pen
int2_u2_j1b2 = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit, tmp) &
!$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b3_size, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_x_2, coef_gauss_j_mu_x_2, &
!$OMP List_all_comb_b3_coef, List_all_comb_b3_expo, &
!$OMP List_all_comb_b3_cent, int2_u2_j1b2)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
tmp = 0.d0
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_x_2(i_fit)
coef_fit = coef_gauss_j_mu_x_2(i_fit)
! ---
int_fit = overlap_gauss_r12_ao(r, expo_fit, i, j)
tmp += coef_fit * int_fit
! if(dabs(coef_fit*int_fit) .lt. 1d-12) cycle
! ---
do i_1s = 2, List_all_comb_b3_size
coef = List_all_comb_b3_coef (i_1s)
beta = List_all_comb_b3_expo (i_1s)
B_center(1) = List_all_comb_b3_cent(1,i_1s)
B_center(2) = List_all_comb_b3_cent(2,i_1s)
B_center(3) = List_all_comb_b3_cent(3,i_1s)
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += coef * coef_fit * int_fit
enddo
! ---
enddo
int2_u2_j1b2(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
int2_u2_j1b2(j,i,ipoint) = int2_u2_j1b2(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_u2_j1b2', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, int2_u_grad1u_x_j1b2, (ao_num, ao_num, n_points_final_grid, 3)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 u_12^mu [\grad_1 u_12^mu] r2
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit(3), expo_fit, coef_fit
double precision :: coef, beta, B_center(3), dist
double precision :: alpha_1s, alpha_1s_inv, centr_1s(3), expo_coef_1s, coef_tmp
double precision :: tmp_x, tmp_y, tmp_z
double precision :: wall0, wall1
print*, ' providing int2_u_grad1u_x_j1b2 ...'
call wall_time(wall0)
provide mu_erf final_grid_points j1b_pen
int2_u_grad1u_x_j1b2 = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit, alpha_1s, dist, &
!$OMP alpha_1s_inv, centr_1s, expo_coef_1s, coef_tmp, &
!$OMP tmp_x, tmp_y, tmp_z) &
!$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b3_size, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_1_erf, coef_gauss_j_mu_1_erf, &
!$OMP List_all_comb_b3_coef, List_all_comb_b3_expo, &
!$OMP List_all_comb_b3_cent, int2_u_grad1u_x_j1b2)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
tmp_x = 0.d0
tmp_y = 0.d0
tmp_z = 0.d0
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_1_erf(i_fit)
coef_fit = coef_gauss_j_mu_1_erf(i_fit)
! ---
call NAI_pol_x_mult_erf_ao_with1s(i, j, expo_fit, r, 1.d+9, r, int_fit)
tmp_x += coef_fit * int_fit(1)
tmp_y += coef_fit * int_fit(2)
tmp_z += coef_fit * int_fit(3)
! if( dabs(coef_fit)*(dabs(int_fit(1)) + dabs(int_fit(2)) + dabs(int_fit(3))) .lt. 3d-10 ) cycle
! ---
do i_1s = 2, List_all_comb_b3_size
coef = List_all_comb_b3_coef (i_1s)
beta = List_all_comb_b3_expo (i_1s)
B_center(1) = List_all_comb_b3_cent(1,i_1s)
B_center(2) = List_all_comb_b3_cent(2,i_1s)
B_center(3) = List_all_comb_b3_cent(3,i_1s)
dist = (B_center(1) - r(1)) * (B_center(1) - r(1)) &
+ (B_center(2) - r(2)) * (B_center(2) - r(2)) &
+ (B_center(3) - r(3)) * (B_center(3) - r(3))
alpha_1s = beta + expo_fit
alpha_1s_inv = 1.d0 / alpha_1s
centr_1s(1) = alpha_1s_inv * (beta * B_center(1) + expo_fit * r(1))
centr_1s(2) = alpha_1s_inv * (beta * B_center(2) + expo_fit * r(2))
centr_1s(3) = alpha_1s_inv * (beta * B_center(3) + expo_fit * r(3))
expo_coef_1s = beta * expo_fit * alpha_1s_inv * dist
coef_tmp = coef * coef_fit * dexp(-expo_coef_1s)
! if(dabs(coef_tmp) .lt. 1d-12) cycle
call NAI_pol_x_mult_erf_ao_with1s(i, j, alpha_1s, centr_1s, 1.d+9, r, int_fit)
tmp_x += coef_tmp * int_fit(1)
tmp_y += coef_tmp * int_fit(2)
tmp_z += coef_tmp * int_fit(3)
enddo
! ---
enddo
int2_u_grad1u_x_j1b2(j,i,ipoint,1) = tmp_x
int2_u_grad1u_x_j1b2(j,i,ipoint,2) = tmp_y
int2_u_grad1u_x_j1b2(j,i,ipoint,3) = tmp_z
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
int2_u_grad1u_x_j1b2(j,i,ipoint,1) = int2_u_grad1u_x_j1b2(i,j,ipoint,1)
int2_u_grad1u_x_j1b2(j,i,ipoint,2) = int2_u_grad1u_x_j1b2(i,j,ipoint,2)
int2_u_grad1u_x_j1b2(j,i,ipoint,3) = int2_u_grad1u_x_j1b2(i,j,ipoint,3)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_u_grad1u_x_j1b2 = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, int2_u_grad1u_j1b2, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 u_12^mu [\grad_1 u_12^mu]
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit, expo_fit, coef_fit, coef_tmp
double precision :: coef, beta, B_center(3), dist
double precision :: alpha_1s, alpha_1s_inv, centr_1s(3), expo_coef_1s, tmp
double precision :: wall0, wall1
double precision, external :: NAI_pol_mult_erf_ao_with1s
print*, ' providing int2_u_grad1u_j1b2 ...'
call wall_time(wall0)
provide mu_erf final_grid_points j1b_pen
int2_u_grad1u_j1b2 = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit, tmp, alpha_1s, dist, &
!$OMP alpha_1s_inv, centr_1s, expo_coef_1s, coef_tmp) &
!$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b3_size, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_1_erf, coef_gauss_j_mu_1_erf, &
!$OMP List_all_comb_b3_coef, List_all_comb_b3_expo, &
!$OMP List_all_comb_b3_cent, int2_u_grad1u_j1b2)
!$OMP DO
do ipoint = 1, n_points_final_grid
do i = 1, ao_num
do j = i, ao_num
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
tmp = 0.d0
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_1_erf(i_fit)
coef_fit = coef_gauss_j_mu_1_erf(i_fit)
! ---
int_fit = NAI_pol_mult_erf_ao_with1s(i, j, expo_fit, r, 1.d+9, r)
! if(dabs(coef_fit)*dabs(int_fit) .lt. 1d-12) cycle
tmp += coef_fit * int_fit
! ---
do i_1s = 2, List_all_comb_b3_size
coef = List_all_comb_b3_coef (i_1s)
beta = List_all_comb_b3_expo (i_1s)
B_center(1) = List_all_comb_b3_cent(1,i_1s)
B_center(2) = List_all_comb_b3_cent(2,i_1s)
B_center(3) = List_all_comb_b3_cent(3,i_1s)
dist = (B_center(1) - r(1)) * (B_center(1) - r(1)) &
+ (B_center(2) - r(2)) * (B_center(2) - r(2)) &
+ (B_center(3) - r(3)) * (B_center(3) - r(3))
alpha_1s = beta + expo_fit
alpha_1s_inv = 1.d0 / alpha_1s
centr_1s(1) = alpha_1s_inv * (beta * B_center(1) + expo_fit * r(1))
centr_1s(2) = alpha_1s_inv * (beta * B_center(2) + expo_fit * r(2))
centr_1s(3) = alpha_1s_inv * (beta * B_center(3) + expo_fit * r(3))
expo_coef_1s = beta * expo_fit * alpha_1s_inv * dist
if(expo_coef_1s .gt. 80.d0) cycle
coef_tmp = coef * coef_fit * dexp(-expo_coef_1s)
if(dabs(coef_tmp) .lt. 1d-12) cycle
int_fit = NAI_pol_mult_erf_ao_with1s(i, j, alpha_1s, centr_1s, 1.d+9, r)
tmp += coef_tmp * int_fit
enddo
! ---
enddo
int2_u_grad1u_j1b2(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
int2_u_grad1u_j1b2(j,i,ipoint) = int2_u_grad1u_j1b2(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for int2_u_grad1u_j1b2', wall1 - wall0
END_PROVIDER
! ---

View File

@ -0,0 +1,453 @@
!
!! ---
!
!BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2, (ao_num, ao_num, n_points_final_grid)]
!
! BEGIN_DOC
! !
! ! -\frac{1}{4} int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 [1 - erf(mu r12)]^2
! !
! END_DOC
!
! implicit none
! integer :: i, j, ipoint, i_1s, i_fit
! integer :: i_mask_grid
! double precision :: r(3), expo_fit, coef_fit
! double precision :: coef, beta, B_center(3)
! double precision :: wall0, wall1
!
! integer, allocatable :: n_mask_grid(:)
! double precision, allocatable :: r_mask_grid(:,:)
! double precision, allocatable :: int_fit_v(:)
!
! print*, ' providing int2_grad1u2_grad2u2_j1b2'
!
! provide mu_erf final_grid_points_transp j1b_pen
! call wall_time(wall0)
!
! int2_grad1u2_grad2u2_j1b2(:,:,:) = 0.d0
!
! !$OMP PARALLEL DEFAULT (NONE) &
! !$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center,&
! !$OMP coef_fit, expo_fit, int_fit_v, n_mask_grid, &
! !$OMP i_mask_grid, r_mask_grid) &
! !$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b3_size,&
! !$OMP final_grid_points_transp, n_max_fit_slat, &
! !$OMP expo_gauss_1_erf_x_2, coef_gauss_1_erf_x_2, &
! !$OMP List_all_comb_b3_coef, List_all_comb_b3_expo, &
! !$OMP List_all_comb_b3_cent, int2_grad1u2_grad2u2_j1b2, &
! !$OMP ao_overlap_abs)
!
! allocate(int_fit_v(n_points_final_grid))
! allocate(n_mask_grid(n_points_final_grid))
! allocate(r_mask_grid(n_points_final_grid,3))
!
! !$OMP DO SCHEDULE(dynamic)
! do i = 1, ao_num
! do j = i, ao_num
!
! if(ao_overlap_abs(j,i) .lt. 1.d-12) then
! cycle
! endif
!
! do i_fit = 1, n_max_fit_slat
!
! expo_fit = expo_gauss_1_erf_x_2(i_fit)
! coef_fit = coef_gauss_1_erf_x_2(i_fit) * (-0.25d0)
!
! ! ---
!
! call overlap_gauss_r12_ao_v(final_grid_points_transp, n_points_final_grid, expo_fit, i, j, int_fit_v, n_points_final_grid, n_points_final_grid)
!
! i_mask_grid = 0 ! dim
! n_mask_grid = 0 ! ind
! r_mask_grid = 0.d0 ! val
! do ipoint = 1, n_points_final_grid
!
! int2_grad1u2_grad2u2_j1b2(j,i,ipoint) += coef_fit * int_fit_v(ipoint)
!
! if(dabs(int_fit_v(ipoint)) .gt. 1d-10) then
! i_mask_grid += 1
! n_mask_grid(i_mask_grid ) = ipoint
! r_mask_grid(i_mask_grid,1) = final_grid_points_transp(ipoint,1)
! r_mask_grid(i_mask_grid,2) = final_grid_points_transp(ipoint,2)
! r_mask_grid(i_mask_grid,3) = final_grid_points_transp(ipoint,3)
! endif
!
! enddo
!
! if(i_mask_grid .eq. 0) cycle
!
! ! ---
!
! do i_1s = 2, List_all_comb_b3_size
!
! coef = List_all_comb_b3_coef (i_1s) * coef_fit
! beta = List_all_comb_b3_expo (i_1s)
! B_center(1) = List_all_comb_b3_cent(1,i_1s)
! B_center(2) = List_all_comb_b3_cent(2,i_1s)
! B_center(3) = List_all_comb_b3_cent(3,i_1s)
!
! call overlap_gauss_r12_ao_with1s_v(B_center, beta, r_mask_grid, n_points_final_grid, expo_fit, i, j, int_fit_v, n_points_final_grid, i_mask_grid)
!
! do ipoint = 1, i_mask_grid
! int2_grad1u2_grad2u2_j1b2(j,i,n_mask_grid(ipoint)) += coef * int_fit_v(ipoint)
! enddo
!
! enddo
!
! ! ---
!
! enddo
! enddo
! enddo
! !$OMP END DO
!
! deallocate(n_mask_grid)
! deallocate(r_mask_grid)
! deallocate(int_fit_v)
!
! !$OMP END PARALLEL
!
! do ipoint = 1, n_points_final_grid
! do i = 2, ao_num
! do j = 1, i-1
! int2_grad1u2_grad2u2_j1b2(j,i,ipoint) = int2_grad1u2_grad2u2_j1b2(i,j,ipoint)
! enddo
! enddo
! enddo
!
! call wall_time(wall1)
! print*, ' wall time for int2_grad1u2_grad2u2_j1b2', wall1 - wall0
!
!END_PROVIDER
!
!! ---
!
!BEGIN_PROVIDER [ double precision, int2_u2_j1b2, (ao_num, ao_num, n_points_final_grid)]
!
! BEGIN_DOC
! !
! ! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 [u_12^mu]^2
! !
! END_DOC
!
! implicit none
! integer :: i, j, ipoint, i_1s, i_fit
! integer :: i_mask_grid
! double precision :: r(3), expo_fit, coef_fit
! double precision :: coef, beta, B_center(3), tmp
! double precision :: wall0, wall1
!
! integer, allocatable :: n_mask_grid(:)
! double precision, allocatable :: r_mask_grid(:,:)
! double precision, allocatable :: int_fit_v(:)
!
! print*, ' providing int2_u2_j1b2'
!
! provide mu_erf final_grid_points_transp j1b_pen
! call wall_time(wall0)
!
! int2_u2_j1b2(:,:,:) = 0.d0
!
! !$OMP PARALLEL DEFAULT (NONE) &
! !$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
! !$OMP coef_fit, expo_fit, int_fit_v, &
! !$OMP i_mask_grid, n_mask_grid, r_mask_grid ) &
! !$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b3_size, &
! !$OMP final_grid_points_transp, n_max_fit_slat, &
! !$OMP expo_gauss_j_mu_x_2, coef_gauss_j_mu_x_2, &
! !$OMP List_all_comb_b3_coef, List_all_comb_b3_expo, &
! !$OMP List_all_comb_b3_cent, int2_u2_j1b2)
!
! allocate(n_mask_grid(n_points_final_grid))
! allocate(r_mask_grid(n_points_final_grid,3))
! allocate(int_fit_v(n_points_final_grid))
!
! !$OMP DO SCHEDULE(dynamic)
! do i = 1, ao_num
! do j = i, ao_num
!
! do i_fit = 1, n_max_fit_slat
!
! expo_fit = expo_gauss_j_mu_x_2(i_fit)
! coef_fit = coef_gauss_j_mu_x_2(i_fit)
!
! ! ---
!
! call overlap_gauss_r12_ao_v(final_grid_points_transp, n_points_final_grid, expo_fit, i, j, int_fit_v, n_points_final_grid, n_points_final_grid)
!
! i_mask_grid = 0 ! dim
! n_mask_grid = 0 ! ind
! r_mask_grid = 0.d0 ! val
!
! do ipoint = 1, n_points_final_grid
! int2_u2_j1b2(j,i,ipoint) += coef_fit * int_fit_v(ipoint)
!
! if(dabs(int_fit_v(ipoint)) .gt. 1d-10) then
! i_mask_grid += 1
! n_mask_grid(i_mask_grid ) = ipoint
! r_mask_grid(i_mask_grid,1) = final_grid_points_transp(ipoint,1)
! r_mask_grid(i_mask_grid,2) = final_grid_points_transp(ipoint,2)
! r_mask_grid(i_mask_grid,3) = final_grid_points_transp(ipoint,3)
! endif
! enddo
!
! if(i_mask_grid .eq. 0) cycle
!
! ! ---
!
! do i_1s = 2, List_all_comb_b3_size
!
! coef = List_all_comb_b3_coef (i_1s) * coef_fit
! beta = List_all_comb_b3_expo (i_1s)
! B_center(1) = List_all_comb_b3_cent(1,i_1s)
! B_center(2) = List_all_comb_b3_cent(2,i_1s)
! B_center(3) = List_all_comb_b3_cent(3,i_1s)
!
! call overlap_gauss_r12_ao_with1s_v(B_center, beta, r_mask_grid, n_points_final_grid, expo_fit, i, j, int_fit_v, n_points_final_grid, i_mask_grid)
!
! do ipoint = 1, i_mask_grid
! int2_u2_j1b2(j,i,n_mask_grid(ipoint)) += coef * int_fit_v(ipoint)
! enddo
!
! enddo
!
! ! ---
!
! enddo
! enddo
! enddo
! !$OMP END DO
!
! deallocate(n_mask_grid)
! deallocate(r_mask_grid)
! deallocate(int_fit_v)
!
! !$OMP END PARALLEL
!
! do ipoint = 1, n_points_final_grid
! do i = 2, ao_num
! do j = 1, i-1
! int2_u2_j1b2(j,i,ipoint) = int2_u2_j1b2(i,j,ipoint)
! enddo
! enddo
! enddo
!
! call wall_time(wall1)
! print*, ' wall time for int2_u2_j1b2', wall1 - wall0
!
!END_PROVIDER
!
!! ---
!
!BEGIN_PROVIDER [ double precision, int2_u_grad1u_x_j1b2, (ao_num, ao_num, n_points_final_grid, 3)]
!
! BEGIN_DOC
! !
! ! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2)^2 u_12^mu [\grad_1 u_12^mu] r2
! !
! END_DOC
!
! implicit none
!
! integer :: i, j, ipoint, i_1s, i_fit
! integer :: i_mask_grid1, i_mask_grid2, i_mask_grid3, i_mask_grid(3)
! double precision :: x, y, z, expo_fit, coef_fit
! double precision :: coef, beta, B_center(3)
! double precision :: alpha_1s, alpha_1s_inv, expo_coef_1s
! double precision :: wall0, wall1
!
! integer, allocatable :: n_mask_grid(:,:)
! double precision, allocatable :: r_mask_grid(:,:,:)
! double precision, allocatable :: int_fit_v(:,:), dist(:,:), centr_1s(:,:,:)
!
! print*, ' providing int2_u_grad1u_x_j1b2'
!
! provide mu_erf final_grid_points_transp j1b_pen
! call wall_time(wall0)
!
! int2_u_grad1u_x_j1b2(:,:,:,:) = 0.d0
!
! !$OMP PARALLEL DEFAULT (NONE) &
! !$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, x, y, z, coef, beta, &
! !$OMP coef_fit, expo_fit, int_fit_v, alpha_1s, dist, B_center,&
! !$OMP alpha_1s_inv, centr_1s, expo_coef_1s, &
! !$OMP i_mask_grid1, i_mask_grid2, i_mask_grid3, i_mask_grid, &
! !$OMP n_mask_grid, r_mask_grid) &
! !$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b3_size, &
! !$OMP final_grid_points_transp, n_max_fit_slat, &
! !$OMP expo_gauss_j_mu_1_erf, coef_gauss_j_mu_1_erf, &
! !$OMP List_all_comb_b3_coef, List_all_comb_b3_expo, &
! !$OMP List_all_comb_b3_cent, int2_u_grad1u_x_j1b2)
!
! allocate(dist(n_points_final_grid,3))
! allocate(centr_1s(n_points_final_grid,3,3))
! allocate(n_mask_grid(n_points_final_grid,3))
! allocate(r_mask_grid(n_points_final_grid,3,3))
! allocate(int_fit_v(n_points_final_grid,3))
!
! !$OMP DO SCHEDULE(dynamic)
! do i = 1, ao_num
! do j = i, ao_num
! do i_fit = 1, n_max_fit_slat
!
! expo_fit = expo_gauss_j_mu_1_erf(i_fit)
! coef_fit = coef_gauss_j_mu_1_erf(i_fit)
!
! ! ---
!
! call NAI_pol_x_mult_erf_ao_with1s_v0(i, j, expo_fit, final_grid_points_transp, n_points_final_grid, 1.d+9, final_grid_points_transp, n_points_final_grid, int_fit_v, n_points_final_grid, n_points_final_grid)
!
! i_mask_grid1 = 0 ! dim
! i_mask_grid2 = 0 ! dim
! i_mask_grid3 = 0 ! dim
! n_mask_grid = 0 ! ind
! r_mask_grid = 0.d0 ! val
! do ipoint = 1, n_points_final_grid
!
! ! ---
!
! int2_u_grad1u_x_j1b2(j,i,ipoint,1) += coef_fit * int_fit_v(ipoint,1)
!
! if(dabs(int_fit_v(ipoint,1)) .gt. 1d-10) then
! i_mask_grid1 += 1
! n_mask_grid(i_mask_grid1, 1) = ipoint
! r_mask_grid(i_mask_grid1,1,1) = final_grid_points_transp(ipoint,1)
! r_mask_grid(i_mask_grid1,2,1) = final_grid_points_transp(ipoint,2)
! r_mask_grid(i_mask_grid1,3,1) = final_grid_points_transp(ipoint,3)
! endif
!
! ! ---
!
! int2_u_grad1u_x_j1b2(j,i,ipoint,2) += coef_fit * int_fit_v(ipoint,2)
!
! if(dabs(int_fit_v(ipoint,2)) .gt. 1d-10) then
! i_mask_grid2 += 1
! n_mask_grid(i_mask_grid2, 2) = ipoint
! r_mask_grid(i_mask_grid2,1,2) = final_grid_points_transp(ipoint,1)
! r_mask_grid(i_mask_grid2,2,2) = final_grid_points_transp(ipoint,2)
! r_mask_grid(i_mask_grid2,3,2) = final_grid_points_transp(ipoint,3)
! endif
!
! ! ---
!
! int2_u_grad1u_x_j1b2(j,i,ipoint,3) += coef_fit * int_fit_v(ipoint,3)
!
! if(dabs(int_fit_v(ipoint,3)) .gt. 1d-10) then
! i_mask_grid3 += 1
! n_mask_grid(i_mask_grid3, 3) = ipoint
! r_mask_grid(i_mask_grid3,1,3) = final_grid_points_transp(ipoint,1)
! r_mask_grid(i_mask_grid3,2,3) = final_grid_points_transp(ipoint,2)
! r_mask_grid(i_mask_grid3,3,3) = final_grid_points_transp(ipoint,3)
! endif
!
! ! ---
!
! enddo
!
! if((i_mask_grid1+i_mask_grid2+i_mask_grid3) .eq. 0) cycle
!
! i_mask_grid(1) = i_mask_grid1
! i_mask_grid(2) = i_mask_grid2
! i_mask_grid(3) = i_mask_grid3
!
! ! ---
!
! do i_1s = 2, List_all_comb_b3_size
!
! coef = List_all_comb_b3_coef (i_1s) * coef_fit
! beta = List_all_comb_b3_expo (i_1s)
! B_center(1) = List_all_comb_b3_cent(1,i_1s)
! B_center(2) = List_all_comb_b3_cent(2,i_1s)
! B_center(3) = List_all_comb_b3_cent(3,i_1s)
!
! alpha_1s = beta + expo_fit
! alpha_1s_inv = 1.d0 / alpha_1s
! expo_coef_1s = beta * expo_fit * alpha_1s_inv
!
! do ipoint = 1, i_mask_grid1
!
! x = r_mask_grid(ipoint,1,1)
! y = r_mask_grid(ipoint,2,1)
! z = r_mask_grid(ipoint,3,1)
!
! centr_1s(ipoint,1,1) = alpha_1s_inv * (beta * B_center(1) + expo_fit * x)
! centr_1s(ipoint,2,1) = alpha_1s_inv * (beta * B_center(2) + expo_fit * y)
! centr_1s(ipoint,3,1) = alpha_1s_inv * (beta * B_center(3) + expo_fit * z)
!
! dist(ipoint,1) = (B_center(1) - x) * (B_center(1) - x) + (B_center(2) - y) * (B_center(2) - y) + (B_center(3) - z) * (B_center(3) - z)
! enddo
!
! do ipoint = 1, i_mask_grid2
!
! x = r_mask_grid(ipoint,1,2)
! y = r_mask_grid(ipoint,2,2)
! z = r_mask_grid(ipoint,3,2)
!
! centr_1s(ipoint,1,2) = alpha_1s_inv * (beta * B_center(1) + expo_fit * x)
! centr_1s(ipoint,2,2) = alpha_1s_inv * (beta * B_center(2) + expo_fit * y)
! centr_1s(ipoint,3,2) = alpha_1s_inv * (beta * B_center(3) + expo_fit * z)
!
! dist(ipoint,2) = (B_center(1) - x) * (B_center(1) - x) + (B_center(2) - y) * (B_center(2) - y) + (B_center(3) - z) * (B_center(3) - z)
! enddo
!
! do ipoint = 1, i_mask_grid3
!
! x = r_mask_grid(ipoint,1,3)
! y = r_mask_grid(ipoint,2,3)
! z = r_mask_grid(ipoint,3,3)
!
! centr_1s(ipoint,1,3) = alpha_1s_inv * (beta * B_center(1) + expo_fit * x)
! centr_1s(ipoint,2,3) = alpha_1s_inv * (beta * B_center(2) + expo_fit * y)
! centr_1s(ipoint,3,3) = alpha_1s_inv * (beta * B_center(3) + expo_fit * z)
!
! dist(ipoint,3) = (B_center(1) - x) * (B_center(1) - x) + (B_center(2) - y) * (B_center(2) - y) + (B_center(3) - z) * (B_center(3) - z)
! enddo
!
! call NAI_pol_x_mult_erf_ao_with1s_v(i, j, alpha_1s, centr_1s, n_points_final_grid, 1.d+9, r_mask_grid, n_points_final_grid, int_fit_v, n_points_final_grid, i_mask_grid)
!
! do ipoint = 1, i_mask_grid1
! int2_u_grad1u_x_j1b2(j,i,n_mask_grid(ipoint,1),1) += coef * dexp(-expo_coef_1s * dist(ipoint,1)) * int_fit_v(ipoint,1)
! enddo
!
! do ipoint = 1, i_mask_grid2
! int2_u_grad1u_x_j1b2(j,i,n_mask_grid(ipoint,2),2) += coef * dexp(-expo_coef_1s * dist(ipoint,2)) * int_fit_v(ipoint,2)
! enddo
!
! do ipoint = 1, i_mask_grid3
! int2_u_grad1u_x_j1b2(j,i,n_mask_grid(ipoint,3),3) += coef * dexp(-expo_coef_1s * dist(ipoint,3)) * int_fit_v(ipoint,3)
! enddo
!
! enddo
!
! ! ---
!
! enddo
! enddo
! enddo
! !$OMP END DO
!
! deallocate(dist)
! deallocate(centr_1s)
! deallocate(n_mask_grid)
! deallocate(r_mask_grid)
! deallocate(int_fit_v)
!
! !$OMP END PARALLEL
!
! do ipoint = 1, n_points_final_grid
! do i = 2, ao_num
! do j = 1, i-1
! int2_u_grad1u_x_j1b2(j,i,ipoint,1) = int2_u_grad1u_x_j1b2(i,j,ipoint,1)
! int2_u_grad1u_x_j1b2(j,i,ipoint,2) = int2_u_grad1u_x_j1b2(i,j,ipoint,2)
! int2_u_grad1u_x_j1b2(j,i,ipoint,3) = int2_u_grad1u_x_j1b2(i,j,ipoint,3)
! enddo
! enddo
! enddo
!
! call wall_time(wall1)
! print*, ' wall time for int2_u_grad1u_x_j1b2 =', wall1 - wall0
!
!END_PROVIDER
!

View File

@ -0,0 +1,369 @@
! ---
BEGIN_PROVIDER [ double precision, v_ij_erf_rk_cst_mu_j1b_test, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr phi_i(r) phi_j(r) 1s_j1b(r) (erf(mu(R) |r - R| - 1) / |r - R|
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s
double precision :: r(3), int_mu, int_coulomb
double precision :: coef, beta, B_center(3)
double precision :: tmp,int_j1b
double precision :: wall0, wall1
double precision, external :: NAI_pol_mult_erf_ao_with1s
double precision :: sigma_ij,dist_ij_ipoint,dsqpi_3_2
print*, ' providing v_ij_erf_rk_cst_mu_j1b_test ...'
dsqpi_3_2 = (dacos(-1.d0))**(1.5d0)
provide mu_erf final_grid_points j1b_pen
call wall_time(wall0)
v_ij_erf_rk_cst_mu_j1b_test = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, r, coef, beta, B_center, int_mu, int_coulomb, tmp, int_j1b)&
!$OMP SHARED (n_points_final_grid, ao_num, List_comb_thr_b2_size, final_grid_points, &
!$OMP List_comb_thr_b2_coef, List_comb_thr_b2_expo, List_comb_thr_b2_cent,ao_abs_comb_b2_j1b, &
!$OMP v_ij_erf_rk_cst_mu_j1b_test, mu_erf, &
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2)
!$OMP DO
!do ipoint = 1, 10
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-20)cycle
tmp = 0.d0
do i_1s = 1, List_comb_thr_b2_size(j,i)
coef = List_comb_thr_b2_coef (i_1s,j,i)
beta = List_comb_thr_b2_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b2_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
B_center(1) = List_comb_thr_b2_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b2_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b2_cent(3,i_1s,j,i)
! TODO :: cycle on the 1 - erf(mur12)
int_mu = NAI_pol_mult_erf_ao_with1s(i, j, beta, B_center, mu_erf, r)
int_coulomb = NAI_pol_mult_erf_ao_with1s(i, j, beta, B_center, 1.d+9, r)
tmp += coef * (int_mu - int_coulomb)
enddo
v_ij_erf_rk_cst_mu_j1b_test(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
v_ij_erf_rk_cst_mu_j1b_test(j,i,ipoint) = v_ij_erf_rk_cst_mu_j1b_test(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for v_ij_erf_rk_cst_mu_j1b_test', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu_j1b_test, (ao_num, ao_num, n_points_final_grid, 3)]
BEGIN_DOC
! int dr x phi_i(r) phi_j(r) 1s_j1b(r) (erf(mu(R) |r - R|) - 1)/|r - R|
END_DOC
implicit none
integer :: i, j, ipoint, i_1s
double precision :: coef, beta, B_center(3), r(3), ints(3), ints_coulomb(3)
double precision :: tmp_x, tmp_y, tmp_z
double precision :: wall0, wall1
double precision :: sigma_ij,dist_ij_ipoint,dsqpi_3_2,int_j1b,factor_ij_1s,beta_ij,center_ij_1s
print*, ' providing x_v_ij_erf_rk_cst_mu_j1b_test ...'
dsqpi_3_2 = (dacos(-1.d0))**(1.5d0)
provide expo_erfc_mu_gauss ao_prod_sigma ao_prod_center
call wall_time(wall0)
x_v_ij_erf_rk_cst_mu_j1b_test = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, r, coef, beta, B_center, ints, ints_coulomb, &
!$OMP int_j1b, tmp_x, tmp_y, tmp_z,factor_ij_1s,beta_ij,center_ij_1s) &
!$OMP SHARED (n_points_final_grid, ao_num, List_comb_thr_b2_size, final_grid_points,&
!$OMP List_comb_thr_b2_coef, List_comb_thr_b2_expo, List_comb_thr_b2_cent, &
!$OMP x_v_ij_erf_rk_cst_mu_j1b_test, mu_erf,ao_abs_comb_b2_j1b, &
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma)
! !$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2,expo_erfc_mu_gauss)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-10)cycle
tmp_x = 0.d0
tmp_y = 0.d0
tmp_z = 0.d0
do i_1s = 1, List_comb_thr_b2_size(j,i)
coef = List_comb_thr_b2_coef (i_1s,j,i)
beta = List_comb_thr_b2_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b2_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
B_center(1) = List_comb_thr_b2_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b2_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b2_cent(3,i_1s,j,i)
! if(ao_prod_center(1,j,i).ne.10000.d0)then
! ! approximate 1 - erf(mu r12) by a gaussian * 10
! !DIR$ FORCEINLINE
! call gaussian_product(expo_erfc_mu_gauss,r, &
! ao_prod_sigma(j,i),ao_prod_center(1,j,i), &
! factor_ij_1s,beta_ij,center_ij_1s)
! if(dabs(coef * factor_ij_1s*int_j1b*10.d0 * dsqpi_3_2 * beta_ij**(-1.5d0)).lt.1.d-10)cycle
! endif
call NAI_pol_x_mult_erf_ao_with1s(i, j, beta, B_center, mu_erf, r, ints )
call NAI_pol_x_mult_erf_ao_with1s(i, j, beta, B_center, 1.d+9, r, ints_coulomb)
tmp_x += coef * (ints(1) - ints_coulomb(1))
tmp_y += coef * (ints(2) - ints_coulomb(2))
tmp_z += coef * (ints(3) - ints_coulomb(3))
enddo
x_v_ij_erf_rk_cst_mu_j1b_test(j,i,ipoint,1) = tmp_x
x_v_ij_erf_rk_cst_mu_j1b_test(j,i,ipoint,2) = tmp_y
x_v_ij_erf_rk_cst_mu_j1b_test(j,i,ipoint,3) = tmp_z
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
x_v_ij_erf_rk_cst_mu_j1b_test(j,i,ipoint,1) = x_v_ij_erf_rk_cst_mu_j1b_test(i,j,ipoint,1)
x_v_ij_erf_rk_cst_mu_j1b_test(j,i,ipoint,2) = x_v_ij_erf_rk_cst_mu_j1b_test(i,j,ipoint,2)
x_v_ij_erf_rk_cst_mu_j1b_test(j,i,ipoint,3) = x_v_ij_erf_rk_cst_mu_j1b_test(i,j,ipoint,3)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for x_v_ij_erf_rk_cst_mu_j1b_test', wall1 - wall0
END_PROVIDER
! ---
! TODO analytically
BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_test, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2) u(mu, r12)
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit, expo_fit, coef_fit
double precision :: coef, beta, B_center(3)
double precision :: tmp
double precision :: wall0, wall1
double precision, external :: overlap_gauss_r12_ao_with1s
double precision :: sigma_ij,dist_ij_ipoint,dsqpi_3_2,int_j1b
print*, ' providing v_ij_u_cst_mu_j1b_test ...'
dsqpi_3_2 = (dacos(-1.d0))**(1.5d0)
provide mu_erf final_grid_points j1b_pen
call wall_time(wall0)
v_ij_u_cst_mu_j1b_test = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP beta_ij_u, factor_ij_1s_u, center_ij_1s_u, &
!$OMP coef_fit, expo_fit, int_fit, tmp,coeftot,int_j1b) &
!$OMP SHARED (n_points_final_grid, ao_num, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_x, coef_gauss_j_mu_x, &
!$OMP List_comb_thr_b2_coef, List_comb_thr_b2_expo,List_comb_thr_b2_size, &
!$OMP List_comb_thr_b2_cent, v_ij_u_cst_mu_j1b_test,ao_abs_comb_b2_j1b, &
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2)
!$OMP DO
!do ipoint = 1, 10
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-20)cycle
tmp = 0.d0
do i_1s = 1, List_comb_thr_b2_size(j,i)
coef = List_comb_thr_b2_coef (i_1s,j,i)
beta = List_comb_thr_b2_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b2_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
B_center(1) = List_comb_thr_b2_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b2_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b2_cent(3,i_1s,j,i)
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_x(i_fit)
coef_fit = coef_gauss_j_mu_x(i_fit)
coeftot = coef * coef_fit
if(dabs(coeftot).lt.1.d-15)cycle
double precision :: beta_ij_u, factor_ij_1s_u, center_ij_1s_u(3),coeftot
call gaussian_product(beta,B_center,expo_fit,r,factor_ij_1s_u,beta_ij_u,center_ij_1s_u)
if(factor_ij_1s_u*ao_overlap_abs_grid(j,i).lt.1.d-15)cycle
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += coef * coef_fit * int_fit
enddo
enddo
v_ij_u_cst_mu_j1b_test(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
v_ij_u_cst_mu_j1b_test(j,i,ipoint) = v_ij_u_cst_mu_j1b_test(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for v_ij_u_cst_mu_j1b_test', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_ng_1_test, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2) u(mu, r12) with u(mu,r12) \approx 1/2 mu e^{-2.5 * mu (r12)^2}
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s
double precision :: r(3), int_fit, expo_fit, coef_fit
double precision :: coef, beta, B_center(3)
double precision :: tmp
double precision :: wall0, wall1
double precision, external :: overlap_gauss_r12_ao_with1s
double precision :: sigma_ij,dist_ij_ipoint,dsqpi_3_2,int_j1b
dsqpi_3_2 = (dacos(-1.d0))**(1.5d0)
provide mu_erf final_grid_points j1b_pen
call wall_time(wall0)
v_ij_u_cst_mu_j1b_ng_1_test = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, r, coef, beta, B_center, &
!$OMP beta_ij_u, factor_ij_1s_u, center_ij_1s_u, &
!$OMP coef_fit, expo_fit, int_fit, tmp,coeftot,int_j1b) &
!$OMP SHARED (n_points_final_grid, ao_num, &
!$OMP final_grid_points, expo_good_j_mu_1gauss,coef_good_j_mu_1gauss, &
!$OMP expo_gauss_j_mu_x, coef_gauss_j_mu_x, &
!$OMP List_comb_thr_b2_coef, List_comb_thr_b2_expo,List_comb_thr_b2_size, &
!$OMP List_comb_thr_b2_cent, v_ij_u_cst_mu_j1b_ng_1_test,ao_abs_comb_b2_j1b, &
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2)
!$OMP DO
!do ipoint = 1, 10
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-20)cycle
tmp = 0.d0
do i_1s = 1, List_comb_thr_b2_size(j,i)
coef = List_comb_thr_b2_coef (i_1s,j,i)
beta = List_comb_thr_b2_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b2_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
B_center(1) = List_comb_thr_b2_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b2_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b2_cent(3,i_1s,j,i)
! do i_fit = 1, ng_fit_jast
expo_fit = expo_good_j_mu_1gauss
coef_fit = 1.d0
coeftot = coef * coef_fit
if(dabs(coeftot).lt.1.d-15)cycle
double precision :: beta_ij_u, factor_ij_1s_u, center_ij_1s_u(3),coeftot
call gaussian_product(beta,B_center,expo_fit,r,factor_ij_1s_u,beta_ij_u,center_ij_1s_u)
if(factor_ij_1s_u*ao_overlap_abs_grid(j,i).lt.1.d-15)cycle
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += coef * coef_fit * int_fit
! enddo
enddo
v_ij_u_cst_mu_j1b_ng_1_test(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
v_ij_u_cst_mu_j1b_ng_1_test(j,i,ipoint) = v_ij_u_cst_mu_j1b_ng_1_test(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for v_ij_u_cst_mu_j1b_ng_1_test', wall1 - wall0
END_PROVIDER
! ---

View File

@ -0,0 +1,300 @@
! ---
BEGIN_PROVIDER [ double precision, v_ij_erf_rk_cst_mu_j1b, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr phi_i(r) phi_j(r) 1s_j1b(r) (erf(mu(R) |r - R| - 1) / |r - R|
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s
double precision :: r(3), int_mu, int_coulomb
double precision :: coef, beta, B_center(3)
double precision :: tmp
double precision :: wall0, wall1
double precision, external :: NAI_pol_mult_erf_ao_with1s
print *, ' providing v_ij_erf_rk_cst_mu_j1b ...'
call wall_time(wall0)
provide mu_erf final_grid_points j1b_pen
v_ij_erf_rk_cst_mu_j1b = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, r, coef, beta, B_center, int_mu, int_coulomb, tmp) &
!$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b2_size, final_grid_points, &
!$OMP List_all_comb_b2_coef, List_all_comb_b2_expo, List_all_comb_b2_cent, &
!$OMP v_ij_erf_rk_cst_mu_j1b, mu_erf)
!$OMP DO
!do ipoint = 1, 10
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
tmp = 0.d0
! ---
coef = List_all_comb_b2_coef (1)
beta = List_all_comb_b2_expo (1)
B_center(1) = List_all_comb_b2_cent(1,1)
B_center(2) = List_all_comb_b2_cent(2,1)
B_center(3) = List_all_comb_b2_cent(3,1)
int_mu = NAI_pol_mult_erf_ao_with1s(i, j, beta, B_center, mu_erf, r)
int_coulomb = NAI_pol_mult_erf_ao_with1s(i, j, beta, B_center, 1.d+9, r)
! if(dabs(coef)*dabs(int_mu - int_coulomb) .lt. 1d-12) cycle
tmp += coef * (int_mu - int_coulomb)
! ---
do i_1s = 2, List_all_comb_b2_size
coef = List_all_comb_b2_coef (i_1s)
beta = List_all_comb_b2_expo (i_1s)
B_center(1) = List_all_comb_b2_cent(1,i_1s)
B_center(2) = List_all_comb_b2_cent(2,i_1s)
B_center(3) = List_all_comb_b2_cent(3,i_1s)
int_mu = NAI_pol_mult_erf_ao_with1s(i, j, beta, B_center, mu_erf, r)
int_coulomb = NAI_pol_mult_erf_ao_with1s(i, j, beta, B_center, 1.d+9, r)
tmp += coef * (int_mu - int_coulomb)
enddo
! ---
v_ij_erf_rk_cst_mu_j1b(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
v_ij_erf_rk_cst_mu_j1b(j,i,ipoint) = v_ij_erf_rk_cst_mu_j1b(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for v_ij_erf_rk_cst_mu_j1b', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu_j1b, (ao_num, ao_num, n_points_final_grid, 3)]
BEGIN_DOC
! int dr x phi_i(r) phi_j(r) 1s_j1b(r) (erf(mu(R) |r - R|) - 1)/|r - R|
END_DOC
implicit none
integer :: i, j, ipoint, i_1s
double precision :: coef, beta, B_center(3), r(3), ints(3), ints_coulomb(3)
double precision :: tmp_x, tmp_y, tmp_z
double precision :: wall0, wall1
print*, ' providing x_v_ij_erf_rk_cst_mu_j1b ...'
call wall_time(wall0)
x_v_ij_erf_rk_cst_mu_j1b = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, r, coef, beta, B_center, ints, ints_coulomb, &
!$OMP tmp_x, tmp_y, tmp_z) &
!$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b2_size, final_grid_points,&
!$OMP List_all_comb_b2_coef, List_all_comb_b2_expo, List_all_comb_b2_cent, &
!$OMP x_v_ij_erf_rk_cst_mu_j1b, mu_erf)
!$OMP DO
!do ipoint = 1, 10
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
tmp_x = 0.d0
tmp_y = 0.d0
tmp_z = 0.d0
! ---
coef = List_all_comb_b2_coef (1)
beta = List_all_comb_b2_expo (1)
B_center(1) = List_all_comb_b2_cent(1,1)
B_center(2) = List_all_comb_b2_cent(2,1)
B_center(3) = List_all_comb_b2_cent(3,1)
call NAI_pol_x_mult_erf_ao_with1s(i, j, beta, B_center, mu_erf, r, ints )
call NAI_pol_x_mult_erf_ao_with1s(i, j, beta, B_center, 1.d+9, r, ints_coulomb)
! if( dabs(coef)*(dabs(ints(1)-ints_coulomb(1)) + dabs(ints(2)-ints_coulomb(2)) + dabs(ints(3)-ints_coulomb(3))) .lt. 3d-10) cycle
tmp_x += coef * (ints(1) - ints_coulomb(1))
tmp_y += coef * (ints(2) - ints_coulomb(2))
tmp_z += coef * (ints(3) - ints_coulomb(3))
! ---
do i_1s = 2, List_all_comb_b2_size
coef = List_all_comb_b2_coef (i_1s)
beta = List_all_comb_b2_expo (i_1s)
B_center(1) = List_all_comb_b2_cent(1,i_1s)
B_center(2) = List_all_comb_b2_cent(2,i_1s)
B_center(3) = List_all_comb_b2_cent(3,i_1s)
call NAI_pol_x_mult_erf_ao_with1s(i, j, beta, B_center, mu_erf, r, ints )
call NAI_pol_x_mult_erf_ao_with1s(i, j, beta, B_center, 1.d+9, r, ints_coulomb)
tmp_x += coef * (ints(1) - ints_coulomb(1))
tmp_y += coef * (ints(2) - ints_coulomb(2))
tmp_z += coef * (ints(3) - ints_coulomb(3))
enddo
! ---
x_v_ij_erf_rk_cst_mu_j1b(j,i,ipoint,1) = tmp_x
x_v_ij_erf_rk_cst_mu_j1b(j,i,ipoint,2) = tmp_y
x_v_ij_erf_rk_cst_mu_j1b(j,i,ipoint,3) = tmp_z
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
x_v_ij_erf_rk_cst_mu_j1b(j,i,ipoint,1) = x_v_ij_erf_rk_cst_mu_j1b(i,j,ipoint,1)
x_v_ij_erf_rk_cst_mu_j1b(j,i,ipoint,2) = x_v_ij_erf_rk_cst_mu_j1b(i,j,ipoint,2)
x_v_ij_erf_rk_cst_mu_j1b(j,i,ipoint,3) = x_v_ij_erf_rk_cst_mu_j1b(i,j,ipoint,3)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for x_v_ij_erf_rk_cst_mu_j1b =', wall1 - wall0
END_PROVIDER
! ---
! TODO analytically
BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr2 phi_i(r2) phi_j(r2) 1s_j1b(r2) u(mu, r12)
!
END_DOC
implicit none
integer :: i, j, ipoint, i_1s, i_fit
double precision :: r(3), int_fit, expo_fit, coef_fit
double precision :: coef, beta, B_center(3)
double precision :: tmp
double precision :: wall0, wall1
double precision, external :: overlap_gauss_r12_ao_with1s
print*, ' providing v_ij_u_cst_mu_j1b ...'
call wall_time(wall0)
provide mu_erf final_grid_points j1b_pen
v_ij_u_cst_mu_j1b = 0.d0
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, i_1s, i_fit, r, coef, beta, B_center, &
!$OMP coef_fit, expo_fit, int_fit, tmp) &
!$OMP SHARED (n_points_final_grid, ao_num, List_all_comb_b2_size, &
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_x, coef_gauss_j_mu_x, &
!$OMP List_all_comb_b2_coef, List_all_comb_b2_expo, &
!$OMP List_all_comb_b2_cent, v_ij_u_cst_mu_j1b)
!$OMP DO
!do ipoint = 1, 10
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
tmp = 0.d0
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_x(i_fit)
coef_fit = coef_gauss_j_mu_x(i_fit)
! ---
coef = List_all_comb_b2_coef (1)
beta = List_all_comb_b2_expo (1)
B_center(1) = List_all_comb_b2_cent(1,1)
B_center(2) = List_all_comb_b2_cent(2,1)
B_center(3) = List_all_comb_b2_cent(3,1)
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
! if(dabs(int_fit*coef) .lt. 1d-12) cycle
tmp += coef * coef_fit * int_fit
! ---
do i_1s = 2, List_all_comb_b2_size
coef = List_all_comb_b2_coef (i_1s)
beta = List_all_comb_b2_expo (i_1s)
B_center(1) = List_all_comb_b2_cent(1,i_1s)
B_center(2) = List_all_comb_b2_cent(2,i_1s)
B_center(3) = List_all_comb_b2_cent(3,i_1s)
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += coef * coef_fit * int_fit
enddo
! ---
enddo
v_ij_u_cst_mu_j1b(j,i,ipoint) = tmp
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
v_ij_u_cst_mu_j1b(j,i,ipoint) = v_ij_u_cst_mu_j1b(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for v_ij_u_cst_mu_j1b', wall1 - wall0
END_PROVIDER
! ---

View File

@ -0,0 +1,437 @@
! ---
BEGIN_PROVIDER [ double precision, v_ij_erf_rk_cst_mu, (ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
!
! int dr phi_i(r) phi_j(r) (erf(mu(R) |r - R| - 1) / |r - R|
!
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: r(3)
double precision :: int_mu, int_coulomb
double precision :: wall0, wall1
double precision :: NAI_pol_mult_erf_ao
print*, ' providing v_ij_erf_rk_cst_mu ...'
provide mu_erf final_grid_points
call wall_time(wall0)
v_ij_erf_rk_cst_mu = 0.d0
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i, j, ipoint, r, int_mu, int_coulomb) &
!$OMP SHARED (ao_num, n_points_final_grid, v_ij_erf_rk_cst_mu, final_grid_points, mu_erf)
!$OMP DO SCHEDULE (dynamic)
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
int_mu = NAI_pol_mult_erf_ao(i, j, mu_erf, r)
int_coulomb = NAI_pol_mult_erf_ao(i, j, 1.d+9, r)
v_ij_erf_rk_cst_mu(j,i,ipoint) = int_mu - int_coulomb
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
v_ij_erf_rk_cst_mu(j,i,ipoint) = v_ij_erf_rk_cst_mu(i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print*, ' wall time for v_ij_erf_rk_cst_mu = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, v_ij_erf_rk_cst_mu_transp, (n_points_final_grid, ao_num, ao_num)]
BEGIN_DOC
! int dr phi_i(r) phi_j(r) (erf(mu(R) |r - R| - 1)/|r - R|
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: r(3)
double precision :: int_mu, int_coulomb
double precision :: wall0, wall1
double precision :: NAI_pol_mult_erf_ao
print *, ' providing v_ij_erf_rk_cst_mu_transp ...'
provide mu_erf final_grid_points
call wall_time(wall0)
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,ipoint,r,int_mu,int_coulomb) &
!$OMP SHARED (ao_num,n_points_final_grid,v_ij_erf_rk_cst_mu_transp,final_grid_points,mu_erf)
!$OMP DO SCHEDULE (dynamic)
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
int_mu = NAI_pol_mult_erf_ao(i, j, mu_erf, r)
int_coulomb = NAI_pol_mult_erf_ao(i, j, 1.d+9, r)
v_ij_erf_rk_cst_mu_transp(ipoint,j,i) = int_mu - int_coulomb
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do i = 2, ao_num
do j = 1, i-1
do ipoint = 1, n_points_final_grid
v_ij_erf_rk_cst_mu_transp(ipoint,j,i) = v_ij_erf_rk_cst_mu_transp(ipoint,i,j)
enddo
enddo
enddo
call wall_time(wall1)
print *, ' wall time for v_ij_erf_rk_cst_mu_transp = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu_tmp, (3, ao_num, ao_num, n_points_final_grid)]
BEGIN_DOC
! int dr x * phi_i(r) phi_j(r) (erf(mu(R) |r - R|) - 1)/|r - R|
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: r(3), ints(3), ints_coulomb(3)
double precision :: wall0, wall1
print*, ' providing x_v_ij_erf_rk_cst_mu_tmp ...'
call wall_time(wall0)
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,ipoint,r,ints,ints_coulomb) &
!$OMP SHARED (ao_num,n_points_final_grid,x_v_ij_erf_rk_cst_mu_tmp,final_grid_points,mu_erf)
!$OMP DO SCHEDULE (dynamic)
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
call NAI_pol_x_mult_erf_ao(i, j, mu_erf, r, ints )
call NAI_pol_x_mult_erf_ao(i, j, 1.d+9 , r, ints_coulomb)
x_v_ij_erf_rk_cst_mu_tmp(1,j,i,ipoint) = ints(1) - ints_coulomb(1)
x_v_ij_erf_rk_cst_mu_tmp(2,j,i,ipoint) = ints(2) - ints_coulomb(2)
x_v_ij_erf_rk_cst_mu_tmp(3,j,i,ipoint) = ints(3) - ints_coulomb(3)
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
do ipoint = 1, n_points_final_grid
do i = 2, ao_num
do j = 1, i-1
x_v_ij_erf_rk_cst_mu_tmp(1,j,i,ipoint) = x_v_ij_erf_rk_cst_mu_tmp(1,i,j,ipoint)
x_v_ij_erf_rk_cst_mu_tmp(2,j,i,ipoint) = x_v_ij_erf_rk_cst_mu_tmp(2,i,j,ipoint)
x_v_ij_erf_rk_cst_mu_tmp(3,j,i,ipoint) = x_v_ij_erf_rk_cst_mu_tmp(3,i,j,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print *, ' wall time for x_v_ij_erf_rk_cst_mu_tmp = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu, (ao_num, ao_num, n_points_final_grid, 3)]
BEGIN_DOC
! int dr x * phi_i(r) phi_j(r) (erf(mu(R) |r - R|) - 1)/|r - R|
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: wall0, wall1
print *, ' providing x_v_ij_erf_rk_cst_mu ...'
call wall_time(wall0)
do ipoint = 1, n_points_final_grid
do i = 1, ao_num
do j = 1, ao_num
x_v_ij_erf_rk_cst_mu(j,i,ipoint,1) = x_v_ij_erf_rk_cst_mu_tmp(1,j,i,ipoint)
x_v_ij_erf_rk_cst_mu(j,i,ipoint,2) = x_v_ij_erf_rk_cst_mu_tmp(2,j,i,ipoint)
x_v_ij_erf_rk_cst_mu(j,i,ipoint,3) = x_v_ij_erf_rk_cst_mu_tmp(3,j,i,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print *, ' wall time for x_v_ij_erf_rk_cst_mu = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu_transp, (ao_num, ao_num,3,n_points_final_grid)]
BEGIN_DOC
! int dr x * phi_i(r) phi_j(r) (erf(mu(R) |r - R|) - 1)/|r - R|
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: wall0, wall1
print *, ' providing x_v_ij_erf_rk_cst_mu_transp ...'
call wall_time(wall0)
do ipoint = 1, n_points_final_grid
do i = 1, ao_num
do j = 1, ao_num
x_v_ij_erf_rk_cst_mu_transp(j,i,1,ipoint) = x_v_ij_erf_rk_cst_mu_tmp(1,j,i,ipoint)
x_v_ij_erf_rk_cst_mu_transp(j,i,2,ipoint) = x_v_ij_erf_rk_cst_mu_tmp(2,j,i,ipoint)
x_v_ij_erf_rk_cst_mu_transp(j,i,3,ipoint) = x_v_ij_erf_rk_cst_mu_tmp(3,j,i,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print *, ' wall time for x_v_ij_erf_rk_cst_mu_transp = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu_transp_bis, (n_points_final_grid, ao_num, ao_num, 3)]
BEGIN_DOC
! int dr x * phi_i(r) phi_j(r) (erf(mu(R) |r - R|) - 1)/|r - R|
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: wall0, wall1
print *, ' providing x_v_ij_erf_rk_cst_mu_transp_bis ...'
call wall_time(wall0)
do i = 1, ao_num
do j = 1, ao_num
do ipoint = 1, n_points_final_grid
x_v_ij_erf_rk_cst_mu_transp_bis(ipoint,j,i,1) = x_v_ij_erf_rk_cst_mu_tmp(1,j,i,ipoint)
x_v_ij_erf_rk_cst_mu_transp_bis(ipoint,j,i,2) = x_v_ij_erf_rk_cst_mu_tmp(2,j,i,ipoint)
x_v_ij_erf_rk_cst_mu_transp_bis(ipoint,j,i,3) = x_v_ij_erf_rk_cst_mu_tmp(3,j,i,ipoint)
enddo
enddo
enddo
call wall_time(wall1)
print *, ' wall time for x_v_ij_erf_rk_cst_mu_transp_bis = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, d_dx_v_ij_erf_rk_cst_mu_tmp, (3, n_points_final_grid, ao_num, ao_num)]
BEGIN_DOC
! d_dx_v_ij_erf_rk_cst_mu_tmp(m,R,j,i) = int dr phi_j(r)) (erf(mu(R) |r - R|) - 1)/|r - R| d/dx (phi_i(r)
!
! with m == 1 -> d/dx , m == 2 -> d/dy , m == 3 -> d/dz
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: r(3), ints(3), ints_coulomb(3)
double precision :: wall0, wall1
print *, ' providing d_dx_v_ij_erf_rk_cst_mu_tmp ...'
call wall_time(wall0)
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,ipoint,r,ints,ints_coulomb) &
!$OMP SHARED (ao_num,n_points_final_grid,d_dx_v_ij_erf_rk_cst_mu_tmp,final_grid_points,mu_erf)
!$OMP DO SCHEDULE (dynamic)
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = 1, ao_num
call phi_j_erf_mu_r_dxyz_phi(j, i, mu_erf, r, ints)
call phi_j_erf_mu_r_dxyz_phi(j, i, 1.d+9, r, ints_coulomb)
d_dx_v_ij_erf_rk_cst_mu_tmp(1,ipoint,j,i) = ints(1) - ints_coulomb(1)
d_dx_v_ij_erf_rk_cst_mu_tmp(2,ipoint,j,i) = ints(2) - ints_coulomb(2)
d_dx_v_ij_erf_rk_cst_mu_tmp(3,ipoint,j,i) = ints(3) - ints_coulomb(3)
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
call wall_time(wall1)
print *, ' wall time for d_dx_v_ij_erf_rk_cst_mu_tmp = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, d_dx_v_ij_erf_rk_cst_mu, (n_points_final_grid, ao_num, ao_num, 3)]
BEGIN_DOC
!
! d_dx_v_ij_erf_rk_cst_mu_tmp(j,i,R,m) = int dr phi_j(r)) (erf(mu(R) |r - R|) - 1)/|r - R| d/dx (phi_i(r)
!
! with m == 1 -> d/dx , m == 2 -> d/dy , m == 3 -> d/dz
!
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: wall0, wall1
print *, ' providing d_dx_v_ij_erf_rk_cst_mu ...'
call wall_time(wall0)
do i = 1, ao_num
do j = 1, ao_num
do ipoint = 1, n_points_final_grid
d_dx_v_ij_erf_rk_cst_mu(ipoint,j,i,1) = d_dx_v_ij_erf_rk_cst_mu_tmp(1,ipoint,j,i)
d_dx_v_ij_erf_rk_cst_mu(ipoint,j,i,2) = d_dx_v_ij_erf_rk_cst_mu_tmp(2,ipoint,j,i)
d_dx_v_ij_erf_rk_cst_mu(ipoint,j,i,3) = d_dx_v_ij_erf_rk_cst_mu_tmp(3,ipoint,j,i)
enddo
enddo
enddo
call wall_time(wall1)
print *, ' wall time for d_dx_v_ij_erf_rk_cst_mu = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, x_d_dx_v_ij_erf_rk_cst_mu_tmp, (3, n_points_final_grid, ao_num, ao_num)]
BEGIN_DOC
!
! x_d_dx_v_ij_erf_rk_cst_mu_tmp(m,j,i,R) = int dr x phi_j(r)) (erf(mu(R) |r - R|) - 1)/|r - R| d/dx (phi_i(r)
!
! with m == 1 -> d/dx , m == 2 -> d/dy , m == 3 -> d/dz
!
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: r(3), ints(3), ints_coulomb(3)
double precision :: wall0, wall1
print *, ' providing x_d_dx_v_ij_erf_rk_cst_mu_tmp ...'
call wall_time(wall0)
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,ipoint,r,ints,ints_coulomb) &
!$OMP SHARED (ao_num,n_points_final_grid,x_d_dx_v_ij_erf_rk_cst_mu_tmp,final_grid_points,mu_erf)
!$OMP DO SCHEDULE (dynamic)
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = 1, ao_num
call phi_j_erf_mu_r_xyz_dxyz_phi(j, i, mu_erf, r, ints)
call phi_j_erf_mu_r_xyz_dxyz_phi(j, i, 1.d+9, r, ints_coulomb)
x_d_dx_v_ij_erf_rk_cst_mu_tmp(1,ipoint,j,i) = ints(1) - ints_coulomb(1)
x_d_dx_v_ij_erf_rk_cst_mu_tmp(2,ipoint,j,i) = ints(2) - ints_coulomb(2)
x_d_dx_v_ij_erf_rk_cst_mu_tmp(3,ipoint,j,i) = ints(3) - ints_coulomb(3)
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
call wall_time(wall1)
print *, ' wall time for x_d_dx_v_ij_erf_rk_cst_mu_tmp = ', wall1 - wall0
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, x_d_dx_v_ij_erf_rk_cst_mu, (n_points_final_grid,ao_num, ao_num,3)]
BEGIN_DOC
!
! x_d_dx_v_ij_erf_rk_cst_mu_tmp(j,i,R,m) = int dr x phi_j(r)) (erf(mu(R) |r - R|) - 1)/|r - R| d/dx (phi_i(r)
!
! with m == 1 -> d/dx , m == 2 -> d/dy , m == 3 -> d/dz
!
END_DOC
implicit none
integer :: i, j, ipoint
double precision :: wall0, wall1
print *, ' providing x_d_dx_v_ij_erf_rk_cst_mu ...'
call wall_time(wall0)
do i = 1, ao_num
do j = 1, ao_num
do ipoint = 1, n_points_final_grid
x_d_dx_v_ij_erf_rk_cst_mu(ipoint,j,i,1) = x_d_dx_v_ij_erf_rk_cst_mu_tmp(1,ipoint,j,i)
x_d_dx_v_ij_erf_rk_cst_mu(ipoint,j,i,2) = x_d_dx_v_ij_erf_rk_cst_mu_tmp(2,ipoint,j,i)
x_d_dx_v_ij_erf_rk_cst_mu(ipoint,j,i,3) = x_d_dx_v_ij_erf_rk_cst_mu_tmp(3,ipoint,j,i)
enddo
enddo
enddo
call wall_time(wall1)
print *, ' wall time for x_d_dx_v_ij_erf_rk_cst_mu = ', wall1 - wall0
END_PROVIDER
! ---

View File

@ -0,0 +1,59 @@
BEGIN_PROVIDER [ integer, n_pts_grid_ao_prod, (ao_num, ao_num)]
&BEGIN_PROVIDER [ integer, max_n_pts_grid_ao_prod]
implicit none
integer :: i,j,ipoint
double precision :: overlap, r(3),thr, overlap_abs_gauss_r12_ao,overlap_gauss_r12_ao
double precision :: sigma,dist,center_ij(3),fact_gauss, alpha, center(3)
n_pts_grid_ao_prod = 0
thr = 1.d-11
print*,' expo_good_j_mu_1gauss = ',expo_good_j_mu_1gauss
!$OMP PARALLEL DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, r, overlap, fact_gauss, alpha, center,dist,sigma,center_ij) &
!$OMP SHARED (n_points_final_grid, ao_num, thr, ao_overlap_abs_grid,n_pts_grid_ao_prod,expo_good_j_mu_1gauss,&
!$OMP final_grid_points,ao_prod_center,ao_prod_sigma,ao_nucl)
!$OMP DO
do i = 1, ao_num
! do i = 3,3
do j = 1, ao_num
! do i = 22,22
! do j = 9,9
center_ij(1:3) = ao_prod_center(1:3,j,i)
sigma = ao_prod_sigma(j,i)
sigma *= sigma
sigma = 0.5d0 /sigma
! if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-10)cycle
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
r(3) = final_grid_points(3,ipoint)
dist = (center_ij(1) - r(1))*(center_ij(1) - r(1))
dist += (center_ij(2) - r(2))*(center_ij(2) - r(2))
dist += (center_ij(3) - r(3))*(center_ij(3) - r(3))
dist = dsqrt(dist)
call gaussian_product(sigma, center_ij, expo_good_j_mu_1gauss, r, fact_gauss, alpha, center)
! print*,''
! print*,j,i,ao_overlap_abs_grid(j,i),ao_overlap_abs(j,i)
! print*,r
! print*,dist,sigma
! print*,fact_gauss
if( fact_gauss*ao_overlap_abs_grid(j,i).lt.1.d-11)cycle
if(ao_nucl(i) == ao_nucl(j))then
overlap = overlap_abs_gauss_r12_ao(r, expo_good_j_mu_1gauss, i, j)
else
overlap = overlap_gauss_r12_ao(r, expo_good_j_mu_1gauss, i, j)
endif
! print*,overlap
if(dabs(overlap).lt.thr)cycle
n_pts_grid_ao_prod(j,i) += 1
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
integer :: list(ao_num)
do i = 1, ao_num
list(i) = maxval(n_pts_grid_ao_prod(:,i))
enddo
max_n_pts_grid_ao_prod = maxval(list)
END_PROVIDER

View File

@ -0,0 +1,237 @@
! ---
BEGIN_PROVIDER [ integer, List_all_comb_b2_size]
implicit none
List_all_comb_b2_size = 2**nucl_num
END_PROVIDER
! ---
BEGIN_PROVIDER [ integer, List_all_comb_b2, (nucl_num, List_all_comb_b2_size)]
implicit none
integer :: i, j
if(nucl_num .gt. 32) then
print *, ' nucl_num = ', nucl_num, '> 32'
stop
endif
List_all_comb_b2 = 0
do i = 0, List_all_comb_b2_size-1
do j = 0, nucl_num-1
if (btest(i,j)) then
List_all_comb_b2(j+1,i+1) = 1
endif
enddo
enddo
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, List_all_comb_b2_coef, ( List_all_comb_b2_size)]
&BEGIN_PROVIDER [ double precision, List_all_comb_b2_expo, ( List_all_comb_b2_size)]
&BEGIN_PROVIDER [ double precision, List_all_comb_b2_cent, (3, List_all_comb_b2_size)]
implicit none
integer :: i, j, k, phase
double precision :: tmp_alphaj, tmp_alphak
double precision :: tmp_cent_x, tmp_cent_y, tmp_cent_z
provide j1b_pen
List_all_comb_b2_coef = 0.d0
List_all_comb_b2_expo = 0.d0
List_all_comb_b2_cent = 0.d0
do i = 1, List_all_comb_b2_size
tmp_cent_x = 0.d0
tmp_cent_y = 0.d0
tmp_cent_z = 0.d0
do j = 1, nucl_num
tmp_alphaj = dble(List_all_comb_b2(j,i)) * j1b_pen(j)
List_all_comb_b2_expo(i) += tmp_alphaj
tmp_cent_x += tmp_alphaj * nucl_coord(j,1)
tmp_cent_y += tmp_alphaj * nucl_coord(j,2)
tmp_cent_z += tmp_alphaj * nucl_coord(j,3)
enddo
if(List_all_comb_b2_expo(i) .lt. 1d-10) cycle
List_all_comb_b2_cent(1,i) = tmp_cent_x / List_all_comb_b2_expo(i)
List_all_comb_b2_cent(2,i) = tmp_cent_y / List_all_comb_b2_expo(i)
List_all_comb_b2_cent(3,i) = tmp_cent_z / List_all_comb_b2_expo(i)
enddo
! ---
do i = 1, List_all_comb_b2_size
do j = 2, nucl_num, 1
tmp_alphaj = dble(List_all_comb_b2(j,i)) * j1b_pen(j)
do k = 1, j-1, 1
tmp_alphak = dble(List_all_comb_b2(k,i)) * j1b_pen(k)
List_all_comb_b2_coef(i) += tmp_alphaj * tmp_alphak * ( (nucl_coord(j,1) - nucl_coord(k,1)) * (nucl_coord(j,1) - nucl_coord(k,1)) &
+ (nucl_coord(j,2) - nucl_coord(k,2)) * (nucl_coord(j,2) - nucl_coord(k,2)) &
+ (nucl_coord(j,3) - nucl_coord(k,3)) * (nucl_coord(j,3) - nucl_coord(k,3)) )
enddo
enddo
if(List_all_comb_b2_expo(i) .lt. 1d-10) cycle
List_all_comb_b2_coef(i) = List_all_comb_b2_coef(i) / List_all_comb_b2_expo(i)
enddo
! ---
do i = 1, List_all_comb_b2_size
phase = 0
do j = 1, nucl_num
phase += List_all_comb_b2(j,i)
enddo
List_all_comb_b2_coef(i) = (-1.d0)**dble(phase) * dexp(-List_all_comb_b2_coef(i))
enddo
print *, ' coeff, expo & cent of list b2'
do i = 1, List_all_comb_b2_size
print*, i, List_all_comb_b2_coef(i), List_all_comb_b2_expo(i)
print*, List_all_comb_b2_cent(1,i), List_all_comb_b2_cent(2,i), List_all_comb_b2_cent(3,i)
enddo
END_PROVIDER
! ---
BEGIN_PROVIDER [ integer, List_all_comb_b3_size]
implicit none
List_all_comb_b3_size = 3**nucl_num
END_PROVIDER
! ---
BEGIN_PROVIDER [ integer, List_all_comb_b3, (nucl_num, List_all_comb_b3_size)]
implicit none
integer :: i, j, ii, jj
integer, allocatable :: M(:,:), p(:)
if(nucl_num .gt. 32) then
print *, ' nucl_num = ', nucl_num, '> 32'
stop
endif
List_all_comb_b3(:,:) = 0
List_all_comb_b3(:,List_all_comb_b3_size) = 2
allocate(p(nucl_num))
p = 0
do i = 2, List_all_comb_b3_size-1
do j = 1, nucl_num
ii = 0
do jj = 1, j-1, 1
ii = ii + p(jj) * 3**(jj-1)
enddo
p(j) = modulo(i-1-ii, 3**j) / 3**(j-1)
List_all_comb_b3(j,i) = p(j)
enddo
enddo
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, List_all_comb_b3_coef, ( List_all_comb_b3_size)]
&BEGIN_PROVIDER [ double precision, List_all_comb_b3_expo, ( List_all_comb_b3_size)]
&BEGIN_PROVIDER [ double precision, List_all_comb_b3_cent, (3, List_all_comb_b3_size)]
implicit none
integer :: i, j, k, phase
double precision :: tmp_alphaj, tmp_alphak, facto
provide j1b_pen
List_all_comb_b3_coef = 0.d0
List_all_comb_b3_expo = 0.d0
List_all_comb_b3_cent = 0.d0
do i = 1, List_all_comb_b3_size
do j = 1, nucl_num
tmp_alphaj = dble(List_all_comb_b3(j,i)) * j1b_pen(j)
List_all_comb_b3_expo(i) += tmp_alphaj
List_all_comb_b3_cent(1,i) += tmp_alphaj * nucl_coord(j,1)
List_all_comb_b3_cent(2,i) += tmp_alphaj * nucl_coord(j,2)
List_all_comb_b3_cent(3,i) += tmp_alphaj * nucl_coord(j,3)
enddo
if(List_all_comb_b3_expo(i) .lt. 1d-10) cycle
ASSERT(List_all_comb_b3_expo(i) .gt. 0d0)
List_all_comb_b3_cent(1,i) = List_all_comb_b3_cent(1,i) / List_all_comb_b3_expo(i)
List_all_comb_b3_cent(2,i) = List_all_comb_b3_cent(2,i) / List_all_comb_b3_expo(i)
List_all_comb_b3_cent(3,i) = List_all_comb_b3_cent(3,i) / List_all_comb_b3_expo(i)
enddo
! ---
do i = 1, List_all_comb_b3_size
do j = 2, nucl_num, 1
tmp_alphaj = dble(List_all_comb_b3(j,i)) * j1b_pen(j)
do k = 1, j-1, 1
tmp_alphak = dble(List_all_comb_b3(k,i)) * j1b_pen(k)
List_all_comb_b3_coef(i) += tmp_alphaj * tmp_alphak * ( (nucl_coord(j,1) - nucl_coord(k,1)) * (nucl_coord(j,1) - nucl_coord(k,1)) &
+ (nucl_coord(j,2) - nucl_coord(k,2)) * (nucl_coord(j,2) - nucl_coord(k,2)) &
+ (nucl_coord(j,3) - nucl_coord(k,3)) * (nucl_coord(j,3) - nucl_coord(k,3)) )
enddo
enddo
if(List_all_comb_b3_expo(i) .lt. 1d-10) cycle
List_all_comb_b3_coef(i) = List_all_comb_b3_coef(i) / List_all_comb_b3_expo(i)
enddo
! ---
do i = 1, List_all_comb_b3_size
facto = 1.d0
phase = 0
do j = 1, nucl_num
tmp_alphaj = dble(List_all_comb_b3(j,i))
facto *= 2.d0 / (gamma(tmp_alphaj+1.d0) * gamma(3.d0-tmp_alphaj))
phase += List_all_comb_b3(j,i)
enddo
List_all_comb_b3_coef(i) = (-1.d0)**dble(phase) * facto * dexp(-List_all_comb_b3_coef(i))
enddo
print *, ' coeff, expo & cent of list b3'
do i = 1, List_all_comb_b3_size
print*, i, List_all_comb_b3_coef(i), List_all_comb_b3_expo(i)
print*, List_all_comb_b3_cent(1,i), List_all_comb_b3_cent(2,i), List_all_comb_b3_cent(3,i)
enddo
END_PROVIDER
! ---

View File

@ -0,0 +1,191 @@
BEGIN_PROVIDER [ integer, List_comb_thr_b2_size, (ao_num, ao_num)]
&BEGIN_PROVIDER [ integer, max_List_comb_thr_b2_size]
implicit none
integer :: i_1s,i,j,ipoint
double precision :: coef,beta,center(3),int_j1b,thr
double precision :: r(3),weight,dist
thr = 1.d-15
List_comb_thr_b2_size = 0
do i = 1, ao_num
do j = i, ao_num
do i_1s = 1, List_all_comb_b2_size
coef = List_all_comb_b2_coef (i_1s)
if(dabs(coef).lt.1.d-15)cycle
beta = List_all_comb_b2_expo (i_1s)
beta = max(beta,1.d-12)
center(1:3) = List_all_comb_b2_cent(1:3,i_1s)
int_j1b = 0.d0
do ipoint = 1, n_points_extra_final_grid
r(1:3) = final_grid_points_extra(1:3,ipoint)
weight = final_weight_at_r_vector_extra(ipoint)
dist = ( center(1) - r(1) )*( center(1) - r(1) )
dist += ( center(2) - r(2) )*( center(2) - r(2) )
dist += ( center(3) - r(3) )*( center(3) - r(3) )
int_j1b += dabs(aos_in_r_array_extra_transp(ipoint,i) * aos_in_r_array_extra_transp(ipoint,j))*dexp(-beta*dist) * weight
enddo
if(dabs(coef)*dabs(int_j1b).gt.thr)then
List_comb_thr_b2_size(j,i) += 1
endif
enddo
enddo
enddo
do i = 1, ao_num
do j = 1, i-1
List_comb_thr_b2_size(j,i) = List_comb_thr_b2_size(i,j)
enddo
enddo
integer :: list(ao_num)
do i = 1, ao_num
list(i) = maxval(List_comb_thr_b2_size(:,i))
enddo
max_List_comb_thr_b2_size = maxval(list)
END_PROVIDER
BEGIN_PROVIDER [ double precision, List_comb_thr_b2_coef, ( max_List_comb_thr_b2_size,ao_num, ao_num )]
&BEGIN_PROVIDER [ double precision, List_comb_thr_b2_expo, ( max_List_comb_thr_b2_size,ao_num, ao_num )]
&BEGIN_PROVIDER [ double precision, List_comb_thr_b2_cent, (3, max_List_comb_thr_b2_size,ao_num, ao_num )]
&BEGIN_PROVIDER [ double precision, ao_abs_comb_b2_j1b, ( max_List_comb_thr_b2_size ,ao_num, ao_num)]
implicit none
integer :: i_1s,i,j,ipoint,icount
double precision :: coef,beta,center(3),int_j1b,thr
double precision :: r(3),weight,dist
thr = 1.d-15
ao_abs_comb_b2_j1b = 10000000.d0
do i = 1, ao_num
do j = i, ao_num
icount = 0
do i_1s = 1, List_all_comb_b2_size
coef = List_all_comb_b2_coef (i_1s)
if(dabs(coef).lt.1.d-12)cycle
beta = List_all_comb_b2_expo (i_1s)
center(1:3) = List_all_comb_b2_cent(1:3,i_1s)
int_j1b = 0.d0
do ipoint = 1, n_points_extra_final_grid
r(1:3) = final_grid_points_extra(1:3,ipoint)
weight = final_weight_at_r_vector_extra(ipoint)
dist = ( center(1) - r(1) )*( center(1) - r(1) )
dist += ( center(2) - r(2) )*( center(2) - r(2) )
dist += ( center(3) - r(3) )*( center(3) - r(3) )
int_j1b += dabs(aos_in_r_array_extra_transp(ipoint,i) * aos_in_r_array_extra_transp(ipoint,j))*dexp(-beta*dist) * weight
enddo
if(dabs(coef)*dabs(int_j1b).gt.thr)then
icount += 1
List_comb_thr_b2_coef(icount,j,i) = coef
List_comb_thr_b2_expo(icount,j,i) = beta
List_comb_thr_b2_cent(1:3,icount,j,i) = center(1:3)
ao_abs_comb_b2_j1b(icount,j,i) = int_j1b
endif
enddo
enddo
enddo
do i = 1, ao_num
do j = 1, i-1
do icount = 1, List_comb_thr_b2_size(j,i)
List_comb_thr_b2_coef(icount,j,i) = List_comb_thr_b2_coef(icount,i,j)
List_comb_thr_b2_expo(icount,j,i) = List_comb_thr_b2_expo(icount,i,j)
List_comb_thr_b2_cent(1:3,icount,j,i) = List_comb_thr_b2_cent(1:3,icount,i,j)
enddo
enddo
enddo
END_PROVIDER
BEGIN_PROVIDER [ integer, List_comb_thr_b3_size, (ao_num, ao_num)]
&BEGIN_PROVIDER [ integer, max_List_comb_thr_b3_size]
implicit none
integer :: i_1s,i,j,ipoint
double precision :: coef,beta,center(3),int_j1b,thr
double precision :: r(3),weight,dist
thr = 1.d-15
List_comb_thr_b3_size = 0
do i = 1, ao_num
do j = 1, ao_num
do i_1s = 1, List_all_comb_b3_size
coef = List_all_comb_b3_coef (i_1s)
beta = List_all_comb_b3_expo (i_1s)
center(1:3) = List_all_comb_b3_cent(1:3,i_1s)
if(dabs(coef).lt.thr)cycle
int_j1b = 0.d0
do ipoint = 1, n_points_extra_final_grid
r(1:3) = final_grid_points_extra(1:3,ipoint)
weight = final_weight_at_r_vector_extra(ipoint)
dist = ( center(1) - r(1) )*( center(1) - r(1) )
dist += ( center(2) - r(2) )*( center(2) - r(2) )
dist += ( center(3) - r(3) )*( center(3) - r(3) )
int_j1b += dabs(aos_in_r_array_extra_transp(ipoint,i) * aos_in_r_array_extra_transp(ipoint,j))*dexp(-beta*dist) * weight
enddo
if(dabs(coef)*dabs(int_j1b).gt.thr)then
List_comb_thr_b3_size(j,i) += 1
endif
enddo
enddo
enddo
! do i = 1, ao_num
! do j = 1, i-1
! List_comb_thr_b3_size(j,i) = List_comb_thr_b3_size(i,j)
! enddo
! enddo
integer :: list(ao_num)
do i = 1, ao_num
list(i) = maxval(List_comb_thr_b3_size(:,i))
enddo
max_List_comb_thr_b3_size = maxval(list)
print*,'max_List_comb_thr_b3_size = ',max_List_comb_thr_b3_size
END_PROVIDER
BEGIN_PROVIDER [ double precision, List_comb_thr_b3_coef, ( max_List_comb_thr_b3_size,ao_num, ao_num )]
&BEGIN_PROVIDER [ double precision, List_comb_thr_b3_expo, ( max_List_comb_thr_b3_size,ao_num, ao_num )]
&BEGIN_PROVIDER [ double precision, List_comb_thr_b3_cent, (3, max_List_comb_thr_b3_size,ao_num, ao_num )]
&BEGIN_PROVIDER [ double precision, ao_abs_comb_b3_j1b, ( max_List_comb_thr_b3_size ,ao_num, ao_num)]
implicit none
integer :: i_1s,i,j,ipoint,icount
double precision :: coef,beta,center(3),int_j1b,thr
double precision :: r(3),weight,dist
thr = 1.d-15
ao_abs_comb_b3_j1b = 10000000.d0
do i = 1, ao_num
do j = 1, ao_num
icount = 0
do i_1s = 1, List_all_comb_b3_size
coef = List_all_comb_b3_coef (i_1s)
beta = List_all_comb_b3_expo (i_1s)
beta = max(beta,1.d-12)
center(1:3) = List_all_comb_b3_cent(1:3,i_1s)
if(dabs(coef).lt.thr)cycle
int_j1b = 0.d0
do ipoint = 1, n_points_extra_final_grid
r(1:3) = final_grid_points_extra(1:3,ipoint)
weight = final_weight_at_r_vector_extra(ipoint)
dist = ( center(1) - r(1) )*( center(1) - r(1) )
dist += ( center(2) - r(2) )*( center(2) - r(2) )
dist += ( center(3) - r(3) )*( center(3) - r(3) )
int_j1b += dabs(aos_in_r_array_extra_transp(ipoint,i) * aos_in_r_array_extra_transp(ipoint,j))*dexp(-beta*dist) * weight
enddo
if(dabs(coef)*dabs(int_j1b).gt.thr)then
icount += 1
List_comb_thr_b3_coef(icount,j,i) = coef
List_comb_thr_b3_expo(icount,j,i) = beta
List_comb_thr_b3_cent(1:3,icount,j,i) = center(1:3)
ao_abs_comb_b3_j1b(icount,j,i) = int_j1b
endif
enddo
enddo
enddo
! do i = 1, ao_num
! do j = 1, i-1
! do icount = 1, List_comb_thr_b3_size(j,i)
! List_comb_thr_b3_coef(icount,j,i) = List_comb_thr_b3_coef(icount,i,j)
! List_comb_thr_b3_expo(icount,j,i) = List_comb_thr_b3_expo(icount,i,j)
! List_comb_thr_b3_cent(1:3,icount,j,i) = List_comb_thr_b3_cent(1:3,icount,i,j)
! enddo
! enddo
! enddo
END_PROVIDER

View File

@ -0,0 +1,195 @@
double precision function NAI_pol_mult_erf_gauss_r12(D_center,delta,A_center,B_center,power_A,power_B,alpha,beta,C_center,mu)
BEGIN_DOC
! Computes the following integral R^3 :
!
! .. math::
!
! \int dr (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
! \frac{\erf(\mu | r - R_C | )}{ | r - R_C | }$ exp(-delta (r - D)^2 ).
!
END_DOC
implicit none
include 'constants.include.F'
double precision, intent(in) :: D_center(3), delta ! pure gaussian "D"
double precision, intent(in) :: C_center(3),mu ! coulomb center "C" and "mu" in the erf(mu*x)/x function
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
double precision :: NAI_pol_mult_erf
! First you multiply the usual gaussian "A" with the gaussian exp(-delta (r - D)^2 )
double precision :: A_new(0:max_dim,3)! new polynom
double precision :: A_center_new(3) ! new center
integer :: iorder_a_new(3) ! i_order(i) = order of the new polynom ==> should be equal to power_A
double precision :: alpha_new ! new exponent
double precision :: fact_a_new ! constant factor
double precision :: accu,coefx,coefy,coefz,coefxy,coefxyz,thr
integer :: d(3),i,lx,ly,lz,iorder_tmp(3)
thr = 1.d-10
d = 0 ! order of the polynom for the gaussian exp(-delta (r - D)^2 ) == 0
! New gaussian/polynom defined by :: new pol new center new expo cst fact new order
call give_explicit_poly_and_gaussian(A_new , A_center_new , alpha_new, fact_a_new , iorder_a_new , &
delta,alpha,d,power_A,D_center,A_center,n_pt_max_integrals)
! The new gaussian exp(-delta (r - D)^2 ) (x-A_x)^a \exp(-\alpha (x-A_x)^2
accu = 0.d0
do lx = 0, iorder_a_new(1)
coefx = A_new(lx,1)
if(dabs(coefx).lt.thr)cycle
iorder_tmp(1) = lx
do ly = 0, iorder_a_new(2)
coefy = A_new(ly,2)
coefxy = coefx * coefy
if(dabs(coefxy).lt.thr)cycle
iorder_tmp(2) = ly
do lz = 0, iorder_a_new(3)
coefz = A_new(lz,3)
coefxyz = coefxy * coefz
if(dabs(coefxyz).lt.thr)cycle
iorder_tmp(3) = lz
accu += coefxyz * NAI_pol_mult_erf(A_center_new,B_center,iorder_tmp,power_B,alpha_new,beta,C_center,n_pt_max_integrals,mu)
enddo
enddo
enddo
NAI_pol_mult_erf_gauss_r12 = fact_a_new * accu
end
subroutine erfc_mu_gauss_xyz(D_center,delta,mu,A_center,B_center,power_A,power_B,alpha,beta,n_pt_in,xyz_ints)
BEGIN_DOC
! Computes the following integral :
!
! .. math::
!
! \int dr exp(-delta (r - D)^2 ) x/y/z * (1 - erf(mu |r-r'|))/ |r-r'| * (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
!
! xyz_ints(1) = x , xyz_ints(2) = y, xyz_ints(3) = z, xyz_ints(4) = x^0
END_DOC
implicit none
include 'constants.include.F'
double precision, intent(in) :: D_center(3), delta,mu ! pure gaussian "D" and mu parameter
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3),n_pt_in
double precision, intent(out) :: xyz_ints(4)
double precision :: NAI_pol_mult_erf
! First you multiply the usual gaussian "A" with the gaussian exp(-delta (r - D)^2 )
double precision :: A_new(0:max_dim,3)! new polynom
double precision :: A_center_new(3) ! new center
integer :: iorder_a_new(3) ! i_order(i) = order of the new polynom ==> should be equal to power_A
double precision :: alpha_new ! new exponent
double precision :: fact_a_new ! constant factor
double precision :: accu,coefx,coefy,coefz,coefxy,coefxyz,thr,contrib,contrib_inf,mu_inf
integer :: d(3),i,lx,ly,lz,iorder_tmp(3),dim1,mm
integer :: power_B_tmp(3)
dim1=100
mu_inf = 1.d+10
thr = 1.d-10
d = 0 ! order of the polynom for the gaussian exp(-delta (r - D)^2 ) == 0
! New gaussian/polynom defined by :: new pol new center new expo cst fact new order
call give_explicit_poly_and_gaussian(A_new , A_center_new , alpha_new, fact_a_new , iorder_a_new , &
delta,alpha,d,power_A,D_center,A_center,n_pt_max_integrals)
! The new gaussian exp(-delta (r - D)^2 ) (x-A_x)^a \exp(-\alpha (x-A_x)^2
xyz_ints = 0.d0
do lx = 0, iorder_a_new(1)
coefx = A_new(lx,1)
if(dabs(coefx).lt.thr)cycle
iorder_tmp(1) = lx
do ly = 0, iorder_a_new(2)
coefy = A_new(ly,2)
coefxy = coefx * coefy
if(dabs(coefxy).lt.thr)cycle
iorder_tmp(2) = ly
do lz = 0, iorder_a_new(3)
coefz = A_new(lz,3)
coefxyz = coefxy * coefz
if(dabs(coefxyz).lt.thr)cycle
iorder_tmp(3) = lz
power_B_tmp = power_B
contrib = NAI_pol_mult_erf(A_center_new,B_center,iorder_tmp,power_B_tmp,alpha_new,beta,D_center,n_pt_in,mu)
contrib_inf = NAI_pol_mult_erf(A_center_new,B_center,iorder_tmp,power_B_tmp,alpha_new,beta,D_center,n_pt_in,mu_inf)
xyz_ints(4) += (contrib_inf - contrib) * coefxyz ! usual term with no x/y/z
do mm = 1, 3
! (x phi_i ) * phi_j
! x * (x - B_x)^b_x = B_x (x - B_x)^b_x + 1 * (x - B_x)^{b_x+1}
!
! first contribution :: B_x (x - B_x)^b_x :: usual integral multiplied by B_x
power_B_tmp = power_B
contrib_inf = NAI_pol_mult_erf(A_center_new,B_center,iorder_tmp,power_B_tmp,alpha_new,beta,D_center,n_pt_in,mu_inf)
contrib = NAI_pol_mult_erf(A_center_new,B_center,iorder_tmp,power_B_tmp,alpha_new,beta,D_center,n_pt_in,mu)
xyz_ints(mm) += (contrib_inf - contrib) * B_center(mm) * coefxyz
!
! second contribution :: (x - B_x)^(b_x+1) :: integral with b_x=>b_x+1
power_B_tmp(mm) += 1
contrib = NAI_pol_mult_erf(A_center_new,B_center,iorder_tmp,power_B_tmp,alpha_new,beta,D_center,n_pt_in,mu)
contrib_inf = NAI_pol_mult_erf(A_center_new,B_center,iorder_tmp,power_B_tmp,alpha_new,beta,D_center,n_pt_in,mu_inf)
xyz_ints(mm) += (contrib_inf - contrib) * coefxyz
enddo
enddo
enddo
enddo
xyz_ints *= fact_a_new
end
double precision function erf_mu_gauss(D_center,delta,mu,A_center,B_center,power_A,power_B,alpha,beta,n_pt_in)
BEGIN_DOC
! Computes the following integral :
!
! .. math::
!
! \int dr exp(-delta (r - D)^2 ) erf(mu*|r-r'|)/ |r-r'| * (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
!
END_DOC
implicit none
include 'constants.include.F'
double precision, intent(in) :: D_center(3), delta,mu ! pure gaussian "D" and mu parameter
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3),n_pt_in
double precision :: NAI_pol_mult_erf
! First you multiply the usual gaussian "A" with the gaussian exp(-delta (r - D)^2 )
double precision :: A_new(0:max_dim,3)! new polynom
double precision :: A_center_new(3) ! new center
integer :: iorder_a_new(3) ! i_order(i) = order of the new polynom ==> should be equal to power_A
double precision :: alpha_new ! new exponent
double precision :: fact_a_new ! constant factor
double precision :: accu,coefx,coefy,coefz,coefxy,coefxyz,thr,contrib,contrib_inf,mu_inf
integer :: d(3),i,lx,ly,lz,iorder_tmp(3),dim1,mm
dim1=100
mu_inf = 1.d+10
thr = 1.d-10
d = 0 ! order of the polynom for the gaussian exp(-delta (r - D)^2 ) == 0
! New gaussian/polynom defined by :: new pol new center new expo cst fact new order
call give_explicit_poly_and_gaussian(A_new , A_center_new , alpha_new, fact_a_new , iorder_a_new , &
delta,alpha,d,power_A,D_center,A_center,n_pt_max_integrals)
! The new gaussian exp(-delta (r - D)^2 ) (x-A_x)^a \exp(-\alpha (x-A_x)^2
erf_mu_gauss = 0.d0
do lx = 0, iorder_a_new(1)
coefx = A_new(lx,1)
if(dabs(coefx).lt.thr)cycle
iorder_tmp(1) = lx
do ly = 0, iorder_a_new(2)
coefy = A_new(ly,2)
coefxy = coefx * coefy
if(dabs(coefxy).lt.thr)cycle
iorder_tmp(2) = ly
do lz = 0, iorder_a_new(3)
coefz = A_new(lz,3)
coefxyz = coefxy * coefz
if(dabs(coefxyz).lt.thr)cycle
iorder_tmp(3) = lz
contrib = NAI_pol_mult_erf(A_center_new,B_center,iorder_tmp,power_B,alpha_new,beta,D_center,n_pt_in,mu)
erf_mu_gauss += contrib * coefxyz
enddo
enddo
enddo
erf_mu_gauss *= fact_a_new
end

View File

@ -0,0 +1,340 @@
! ---
double precision function overlap_gauss_r12(D_center, delta, A_center, B_center, power_A, power_B, alpha, beta)
BEGIN_DOC
!
! Computes the following integral :
!
! .. math ::
!
! \int dr exp(-delta (r - D)^2 ) (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
!
END_DOC
include 'constants.include.F'
implicit none
double precision, intent(in) :: D_center(3), delta ! pure gaussian "D"
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
double precision :: overlap_x,overlap_y,overlap_z,overlap
! First you multiply the usual gaussian "A" with the gaussian exp(-delta (r - D)^2 )
double precision :: A_new(0:max_dim,3)! new polynom
double precision :: A_center_new(3) ! new center
integer :: iorder_a_new(3) ! i_order(i) = order of the new polynom ==> should be equal to power_A
double precision :: alpha_new ! new exponent
double precision :: fact_a_new ! constant factor
double precision :: accu, coefx, coefy, coefz, coefxy, coefxyz, thr
integer :: d(3), i, lx, ly, lz, iorder_tmp(3), dim1
dim1 = 100
thr = 1.d-10
d(:) = 0 ! order of the polynom for the gaussian exp(-delta (r - D)^2 ) == 0
overlap_gauss_r12 = 0.d0
! New gaussian/polynom defined by :: new pol new center new expo cst fact new order
call give_explicit_poly_and_gaussian(A_new , A_center_new , alpha_new, fact_a_new , iorder_a_new ,&
delta,alpha,d,power_A,D_center,A_center,n_pt_max_integrals)
if(fact_a_new.lt.thr)return
! The new gaussian exp(-delta (r - D)^2 ) (x-A_x)^a \exp(-\alpha (x-A_x)^2
accu = 0.d0
do lx = 0, iorder_a_new(1)
coefx = A_new(lx,1)*fact_a_new
if(dabs(coefx).lt.thr)cycle
iorder_tmp(1) = lx
do ly = 0, iorder_a_new(2)
coefy = A_new(ly,2)
coefxy = coefx * coefy
if(dabs(coefxy) .lt. thr) cycle
iorder_tmp(2) = ly
do lz = 0, iorder_a_new(3)
coefz = A_new(lz,3)
coefxyz = coefxy * coefz
if(dabs(coefxyz) .lt. thr) cycle
iorder_tmp(3) = lz
call overlap_gaussian_xyz( A_center_new, B_center, alpha_new, beta, iorder_tmp, power_B &
, overlap_x, overlap_y, overlap_z, overlap, dim1)
accu += coefxyz * overlap
enddo
enddo
enddo
overlap_gauss_r12 = accu
end
!---
double precision function overlap_abs_gauss_r12(D_center,delta,A_center,B_center,power_A,power_B,alpha,beta)
BEGIN_DOC
! Computes the following integral :
!
! .. math ::
!
! \int dr exp(-delta (r - D)^2 ) |(x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )|
!
END_DOC
implicit none
include 'constants.include.F'
double precision, intent(in) :: D_center(3), delta ! pure gaussian "D"
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
double precision :: overlap_x,overlap_y,overlap_z,overlap
! First you multiply the usual gaussian "A" with the gaussian exp(-delta (r - D)^2 )
double precision :: A_new(0:max_dim,3)! new polynom
double precision :: A_center_new(3) ! new center
integer :: iorder_a_new(3) ! i_order(i) = order of the new polynom ==> should be equal to power_A
double precision :: alpha_new ! new exponent
double precision :: fact_a_new ! constant factor
double precision :: accu,coefx,coefy,coefz,coefxy,coefxyz,thr,dx,lower_exp_val
integer :: d(3),i,lx,ly,lz,iorder_tmp(3),dim1
dim1=50
lower_exp_val = 40.d0
thr = 1.d-12
d(:) = 0 ! order of the polynom for the gaussian exp(-delta (r - D)^2 ) == 0
overlap_abs_gauss_r12 = 0.d0
! New gaussian/polynom defined by :: new pol new center new expo cst fact new order
call give_explicit_poly_and_gaussian(A_new , A_center_new , alpha_new, fact_a_new , iorder_a_new ,&
delta,alpha,d,power_A,D_center,A_center,n_pt_max_integrals)
if(fact_a_new.lt.thr)return
! The new gaussian exp(-delta (r - D)^2 ) (x-A_x)^a \exp(-\alpha (x-A_x)^2
accu = 0.d0
do lx = 0, iorder_a_new(1)
coefx = A_new(lx,1)*fact_a_new
! if(dabs(coefx).lt.thr)cycle
iorder_tmp(1) = lx
do ly = 0, iorder_a_new(2)
coefy = A_new(ly,2)
coefxy = coefx * coefy
if(dabs(coefxy).lt.thr)cycle
iorder_tmp(2) = ly
do lz = 0, iorder_a_new(3)
coefz = A_new(lz,3)
coefxyz = coefxy * coefz
if(dabs(coefxyz).lt.thr)cycle
iorder_tmp(3) = lz
call overlap_x_abs(A_center_new(1),B_center(1),alpha_new,beta,iorder_tmp(1),power_B(1),overlap_x,lower_exp_val,dx,dim1)
call overlap_x_abs(A_center_new(2),B_center(2),alpha_new,beta,iorder_tmp(2),power_B(2),overlap_y,lower_exp_val,dx,dim1)
call overlap_x_abs(A_center_new(3),B_center(3),alpha_new,beta,iorder_tmp(3),power_B(3),overlap_z,lower_exp_val,dx,dim1)
accu += dabs(coefxyz * overlap_x * overlap_y * overlap_z)
enddo
enddo
enddo
overlap_abs_gauss_r12= accu
end
!---
! TODO apply Gaussian product three times first
subroutine overlap_gauss_r12_v(D_center, LD_D, delta, A_center, B_center, power_A, power_B, alpha, beta, rvec, LD_rvec, n_points)
BEGIN_DOC
!
! Computes the following integral :
!
! \int dr exp(-delta (r - D)^2) (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2)
! using an array of D_centers
!
! n_points: nb of integrals
!
END_DOC
implicit none
include 'constants.include.F'
integer, intent(in) :: LD_D, LD_rvec, n_points
integer, intent(in) :: power_A(3), power_B(3)
double precision, intent(in) :: D_center(LD_D,3), delta
double precision, intent(in) :: A_center(3), B_center(3), alpha, beta
double precision, intent(out) :: rvec(LD_rvec)
integer :: maxab
integer :: d(3), i, lx, ly, lz, iorder_tmp(3), ipoint
double precision :: overlap_x, overlap_y, overlap_z
double precision :: alpha_new
double precision :: accu, thr, coefxy
integer, allocatable :: iorder_a_new(:)
double precision, allocatable :: overlap(:)
double precision, allocatable :: A_new(:,:,:), A_center_new(:,:)
double precision, allocatable :: fact_a_new(:)
thr = 1.d-10
d(:) = 0
maxab = maxval(power_A(1:3))
allocate(A_new(n_points,0:maxab,3), A_center_new(n_points,3), fact_a_new(n_points), iorder_a_new(3), overlap(n_points))
call give_explicit_poly_and_gaussian_v(A_new, maxab, A_center_new, alpha_new, fact_a_new, iorder_a_new, delta, alpha, d, power_A, D_center, LD_D, A_center, n_points)
rvec(:) = 0.d0
do lx = 0, iorder_a_new(1)
iorder_tmp(1) = lx
do ly = 0, iorder_a_new(2)
iorder_tmp(2) = ly
do lz = 0, iorder_a_new(3)
iorder_tmp(3) = lz
call overlap_gaussian_xyz_v(A_center_new, B_center, alpha_new, beta, iorder_tmp, power_B, overlap, n_points)
do ipoint = 1, n_points
rvec(ipoint) = rvec(ipoint) + A_new(ipoint,lx,1) * A_new(ipoint,ly,2) * A_new(ipoint,lz,3) * overlap(ipoint)
enddo
enddo
enddo
enddo
do ipoint = 1, n_points
rvec(ipoint) = rvec(ipoint) * fact_a_new(ipoint)
enddo
deallocate(A_new, A_center_new, fact_a_new, iorder_a_new, overlap)
end subroutine overlap_gauss_r12_v
!---
subroutine overlap_gauss_xyz_r12(D_center, delta, A_center, B_center, power_A, power_B, alpha, beta, gauss_ints)
BEGIN_DOC
! Computes the following integral :
!
! .. math::
!
! gauss_ints(m) = \int dr exp(-delta (r - D)^2 ) * x/y/z (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
!
! with m == 1 ==> x, m == 2 ==> y, m == 3 ==> z
END_DOC
implicit none
include 'constants.include.F'
double precision, intent(in) :: D_center(3), delta ! pure gaussian "D"
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
double precision, intent(out) :: gauss_ints(3)
double precision :: overlap_x,overlap_y,overlap_z,overlap
! First you multiply the usual gaussian "A" with the gaussian exp(-delta (r - D)^2 )
double precision :: A_new(0:max_dim,3)! new polynom
double precision :: A_center_new(3) ! new center
integer :: iorder_a_new(3) ! i_order(i) = order of the new polynom ==> should be equal to power_A
integer :: power_B_new(3)
double precision :: alpha_new ! new exponent
double precision :: fact_a_new ! constant factor
double precision :: coefx,coefy,coefz,coefxy,coefxyz,thr
integer :: d(3),i,lx,ly,lz,iorder_tmp(3),dim1,m
dim1=100
thr = 1.d-10
d = 0 ! order of the polynom for the gaussian exp(-delta (r - D)^2 ) == 0
! New gaussian/polynom defined by :: new pol new center new expo cst fact new order
call give_explicit_poly_and_gaussian(A_new , A_center_new , alpha_new, fact_a_new , iorder_a_new , &
delta,alpha,d,power_A,D_center,A_center,n_pt_max_integrals)
! The new gaussian exp(-delta (r - D)^2 ) (x-A_x)^a \exp(-\alpha (x-A_x)^2
gauss_ints = 0.d0
do lx = 0, iorder_a_new(1)
coefx = A_new(lx,1)
if(dabs(coefx).lt.thr)cycle
iorder_tmp(1) = lx
do ly = 0, iorder_a_new(2)
coefy = A_new(ly,2)
coefxy = coefx * coefy
if(dabs(coefxy).lt.thr)cycle
iorder_tmp(2) = ly
do lz = 0, iorder_a_new(3)
coefz = A_new(lz,3)
coefxyz = coefxy * coefz
if(dabs(coefxyz).lt.thr)cycle
iorder_tmp(3) = lz
do m = 1, 3
! change (x-Bx)^bx --> (x-Bx)^(bx+1) + Bx(x-Bx)^bx
power_B_new = power_B
power_B_new(m) += 1 ! (x-Bx)^(bx+1)
call overlap_gaussian_xyz(A_center_new,B_center,alpha_new,beta,iorder_tmp,power_B_new,overlap_x,overlap_y,overlap_z,overlap,dim1)
gauss_ints(m) += coefxyz * overlap
power_B_new = power_B
call overlap_gaussian_xyz(A_center_new,B_center,alpha_new,beta,iorder_tmp,power_B_new,overlap_x,overlap_y,overlap_z,overlap,dim1)
gauss_ints(m) += coefxyz * overlap * B_center(m) ! Bx (x-Bx)^(bx)
enddo
enddo
enddo
enddo
gauss_ints *= fact_a_new
end
double precision function overlap_gauss_xyz_r12_specific(D_center,delta,A_center,B_center,power_A,power_B,alpha,beta,mx)
BEGIN_DOC
! Computes the following integral :
!
! .. math::
!
! \int dr exp(-delta (r - D)^2 ) * x/y/z (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
!
! with mx == 1 ==> x, mx == 2 ==> y, mx == 3 ==> z
END_DOC
implicit none
include 'constants.include.F'
double precision, intent(in) :: D_center(3), delta ! pure gaussian "D"
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3),mx
double precision :: overlap_x,overlap_y,overlap_z,overlap
! First you multiply the usual gaussian "A" with the gaussian exp(-delta (r - D)^2 )
double precision :: A_new(0:max_dim,3)! new polynom
double precision :: A_center_new(3) ! new center
integer :: iorder_a_new(3) ! i_order(i) = order of the new polynom ==> should be equal to power_A
integer :: power_B_new(3)
double precision :: alpha_new ! new exponent
double precision :: fact_a_new ! constant factor
double precision :: coefx,coefy,coefz,coefxy,coefxyz,thr
integer :: d(3),i,lx,ly,lz,iorder_tmp(3),dim1,m
dim1=100
thr = 1.d-10
d = 0 ! order of the polynom for the gaussian exp(-delta (r - D)^2 ) == 0
! New gaussian/polynom defined by :: new pol new center new expo cst fact new order
call give_explicit_poly_and_gaussian(A_new , A_center_new , alpha_new, fact_a_new , iorder_a_new , &
delta,alpha,d,power_A,D_center,A_center,n_pt_max_integrals)
! The new gaussian exp(-delta (r - D)^2 ) (x-A_x)^a \exp(-\alpha (x-A_x)^2
overlap_gauss_xyz_r12_specific = 0.d0
do lx = 0, iorder_a_new(1)
coefx = A_new(lx,1)
if(dabs(coefx).lt.thr)cycle
iorder_tmp(1) = lx
do ly = 0, iorder_a_new(2)
coefy = A_new(ly,2)
coefxy = coefx * coefy
if(dabs(coefxy).lt.thr)cycle
iorder_tmp(2) = ly
do lz = 0, iorder_a_new(3)
coefz = A_new(lz,3)
coefxyz = coefxy * coefz
if(dabs(coefxyz).lt.thr)cycle
iorder_tmp(3) = lz
m = mx
! change (x-Bx)^bx --> (x-Bx)^(bx+1) + Bx(x-Bx)^bx
power_B_new = power_B
power_B_new(m) += 1 ! (x-Bx)^(bx+1)
call overlap_gaussian_xyz(A_center_new,B_center,alpha_new,beta,iorder_tmp,power_B_new,overlap_x,overlap_y,overlap_z,overlap,dim1)
overlap_gauss_xyz_r12_specific += coefxyz * overlap
power_B_new = power_B
call overlap_gaussian_xyz(A_center_new,B_center,alpha_new,beta,iorder_tmp,power_B_new,overlap_x,overlap_y,overlap_z,overlap,dim1)
overlap_gauss_xyz_r12_specific += coefxyz * overlap * B_center(m) ! Bx (x-Bx)^(bx)
enddo
enddo
enddo
overlap_gauss_xyz_r12_specific *= fact_a_new
end

View File

@ -0,0 +1,121 @@
double precision function ovlp_stg_gauss_int_phi_ij(D_center,gam,delta,A_center,B_center,power_A,power_B,alpha,beta)
BEGIN_DOC
! Computes the following integral :
!
! .. math::
!
! \int dr exp(-gam (r - D)) exp(-delta * (r -D)^2) (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
!
END_DOC
implicit none
double precision, intent(in) :: D_center(3), gam ! pure Slater "D" in r-r_D
double precision, intent(in) :: delta ! gaussian in r-r_D
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
integer :: i
double precision :: integral,gama_gauss
double precision, allocatable :: expos_slat(:)
allocate(expos_slat(n_max_fit_slat))
double precision :: overlap_gauss_r12
ovlp_stg_gauss_int_phi_ij = 0.d0
call expo_fit_slater_gam(gam,expos_slat)
do i = 1, n_max_fit_slat
gama_gauss = expos_slat(i)+delta
integral = overlap_gauss_r12(D_center,gama_gauss,A_center,B_center,power_A,power_B,alpha,beta)
ovlp_stg_gauss_int_phi_ij += coef_fit_slat_gauss(i) * integral
enddo
end
double precision function erf_mu_stg_gauss_int_phi_ij(D_center,gam,delta,A_center,B_center,power_A,power_B,alpha,beta,C_center,mu)
BEGIN_DOC
! Computes the following integral :
!
! .. math::
!
! \int dr exp(-gam(r - D)-delta(r - D)^2) (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
! \frac{\erf(\mu | r - R_C | )}{ | r - R_C | }$.
!
END_DOC
implicit none
include 'constants.include.F'
double precision, intent(in) :: D_center(3), gam ! pure Slater "D" in r-r_D
double precision, intent(in) :: delta ! gaussian in r-r_D
double precision, intent(in) :: C_center(3),mu ! coulomb center "C" and "mu" in the erf(mu*x)/x function
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
integer :: i
double precision :: NAI_pol_mult_erf_gauss_r12
double precision :: integral,gama_gauss
double precision, allocatable :: expos_slat(:)
allocate(expos_slat(n_max_fit_slat))
erf_mu_stg_gauss_int_phi_ij = 0.d0
call expo_fit_slater_gam(gam,expos_slat)
do i = 1, n_max_fit_slat
gama_gauss = expos_slat(i) + delta
integral = NAI_pol_mult_erf_gauss_r12(D_center,gama_gauss,A_center,B_center,power_A,power_B,alpha,beta,C_center,mu)
erf_mu_stg_gauss_int_phi_ij += coef_fit_slat_gauss(i) * integral
enddo
end
double precision function overlap_stg_gauss(D_center,gam,A_center,B_center,power_A,power_B,alpha,beta)
BEGIN_DOC
! Computes the following integral :
!
! .. math::
!
! \int dr exp(-gam (r - D)) (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
!
END_DOC
implicit none
double precision, intent(in) :: D_center(3), gam ! pure Slater "D"
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
integer :: i
double precision :: expos_slat(n_max_fit_slat),integral,delta
double precision :: overlap_gauss_r12
overlap_stg_gauss = 0.d0
call expo_fit_slater_gam(gam,expos_slat)
do i = 1, n_max_fit_slat
delta = expos_slat(i)
integral = overlap_gauss_r12(D_center,delta,A_center,B_center,power_A,power_B,alpha,beta)
overlap_stg_gauss += coef_fit_slat_gauss(i) * integral
enddo
end
double precision function erf_mu_stg_gauss(D_center,gam,A_center,B_center,power_A,power_B,alpha,beta,C_center,mu)
BEGIN_DOC
! Computes the following integral :
!
! .. math::
!
! \int dr exp(-gam(r - D)) (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
! \frac{\erf(\mu | r - R_C | )}{ | r - R_C | }$.
!
END_DOC
implicit none
include 'constants.include.F'
double precision, intent(in) :: D_center(3), gam ! pure Slater "D"
double precision, intent(in) :: C_center(3),mu ! coulomb center "C" and "mu" in the erf(mu*x)/x function
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
integer :: i
double precision :: expos_slat(n_max_fit_slat),integral,delta
double precision :: NAI_pol_mult_erf_gauss_r12
erf_mu_stg_gauss = 0.d0
call expo_fit_slater_gam(gam,expos_slat)
do i = 1, n_max_fit_slat
delta = expos_slat(i)
integral = NAI_pol_mult_erf_gauss_r12(D_center,delta,A_center,B_center,power_A,power_B,alpha,beta,C_center,mu)
erf_mu_stg_gauss += coef_fit_slat_gauss(i) * integral
enddo
end

View File

@ -0,0 +1,101 @@
double precision function exp_dl(x,n)
implicit none
double precision, intent(in) :: x
integer , intent(in) :: n
integer :: i
exp_dl = 1.d0
do i = 1, n
exp_dl += fact_inv(i) * x**dble(i)
enddo
end
subroutine exp_dl_rout(x,n, array)
implicit none
double precision, intent(in) :: x
integer , intent(in) :: n
double precision, intent(out):: array(0:n)
integer :: i
double precision :: accu
accu = 1.d0
array(0) = 1.d0
do i = 1, n
accu += fact_inv(i) * x**dble(i)
array(i) = accu
enddo
end
subroutine exp_dl_ovlp_stg_phi_ij(zeta,D_center,gam,delta,A_center,B_center,power_A,power_B,alpha,beta,n_taylor,array_ints,integral_taylor,exponent_exp)
BEGIN_DOC
! Computes the following integrals :
!
! .. math::
!
! array(i) = \int dr EXP{exponent_exp * [exp(-gam*i (r - D)) exp(-delta*i * (r -D)^2)] (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
!
!
! and gives back the Taylor expansion of the exponential in integral_taylor
END_DOC
implicit none
double precision, intent(in) :: zeta ! prefactor of the argument of the exp(-zeta*x)
integer, intent(in) :: n_taylor ! order of the Taylor expansion of the exponential
double precision, intent(in) :: D_center(3), gam ! pure Slater "D" in r-r_D
double precision, intent(in) :: delta ! gaussian in r-r_D
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
double precision, intent(in) :: exponent_exp
integer, intent(in) :: power_A(3),power_B(3)
double precision, intent(out) :: array_ints(0:n_taylor),integral_taylor
integer :: i,dim1
double precision :: delta_exp,gam_exp,ovlp_stg_gauss_int_phi_ij
double precision :: overlap_x,overlap_y,overlap_z,overlap
dim1=100
call overlap_gaussian_xyz(A_center,B_center,alpha,beta,power_A,power_B,overlap_x,overlap_y,overlap_z,overlap,dim1)
array_ints(0) = overlap
integral_taylor = array_ints(0)
do i = 1, n_taylor
delta_exp = dble(i) * delta
gam_exp = dble(i) * gam
array_ints(i) = ovlp_stg_gauss_int_phi_ij(D_center,gam_exp,delta_exp,A_center,B_center,power_A,power_B,alpha,beta)
integral_taylor += (-zeta*exponent_exp)**dble(i) * fact_inv(i) * array_ints(i)
enddo
end
subroutine exp_dl_erf_stg_phi_ij(zeta,D_center,gam,delta,A_center,B_center,power_A,power_B,alpha,beta,C_center,mu,n_taylor,array_ints,integral_taylor)
BEGIN_DOC
! Computes the following integrals :
!
! .. math::
!
! array(i) = \int dr exp(-gam*i (r - D)) exp(-delta*i * (r -D)^2) (x-A_x)^a (x-B_x)^b \exp(-\alpha (x-A_x)^2 - \beta (x-B_x)^2 )
! \frac{\erf(\mu | r - R_C | )}{ | r - R_C | }$.
!
!
! and gives back the Taylor expansion of the exponential in integral_taylor
END_DOC
implicit none
integer, intent(in) :: n_taylor ! order of the Taylor expansion of the exponential
double precision, intent(in) :: zeta ! prefactor of the argument of the exp(-zeta*x)
double precision, intent(in) :: D_center(3), gam ! pure Slater "D" in r-r_D
double precision, intent(in) :: delta ! gaussian in r-r_D
double precision, intent(in) :: C_center(3),mu ! coulomb center "C" and "mu" in the erf(mu*x)/x function
double precision, intent(in) :: A_center(3),B_center(3),alpha,beta ! gaussian/polynoms "A" and "B"
integer, intent(in) :: power_A(3),power_B(3)
double precision, intent(out) :: array_ints(0:n_taylor),integral_taylor
integer :: i,dim1
double precision :: delta_exp,gam_exp,NAI_pol_mult_erf,erf_mu_stg_gauss_int_phi_ij
dim1=100
array_ints(0) = NAI_pol_mult_erf(A_center,B_center,power_A,power_B,alpha,beta,C_center,n_pt_max_integrals,mu)
integral_taylor = array_ints(0)
do i = 1, n_taylor
delta_exp = dble(i) * delta
gam_exp = dble(i) * gam
array_ints(i) = erf_mu_stg_gauss_int_phi_ij(D_center,gam_exp,delta_exp,A_center,B_center,power_A,power_B,alpha,beta,C_center,mu)
integral_taylor += (-zeta)**dble(i) * fact_inv(i) * array_ints(i)
enddo
end

View File

@ -0,0 +1,343 @@
BEGIN_PROVIDER [double precision, coef_xyz_ao, (2,3,ao_num)]
&BEGIN_PROVIDER [integer, power_xyz_ao, (2,3,ao_num)]
implicit none
BEGIN_DOC
! coefficient for the basis function :: (x * phi_i(r), y * phi_i(r), * z_phi(r))
!
! x * (x - A_x)^a_x = A_x (x - A_x)^a_x + 1 * (x - A_x)^{a_x+1}
END_DOC
integer :: i,j,k,num_ao,power_ao(1:3)
double precision :: center_ao(1:3)
do i = 1, ao_num
power_ao(1:3)= ao_power(i,1:3)
num_ao = ao_nucl(i)
center_ao(1:3) = nucl_coord(num_ao,1:3)
do j = 1, 3
coef_xyz_ao(1,j,i) = center_ao(j) ! A_x (x - A_x)^a_x
power_xyz_ao(1,j,i)= power_ao(j)
coef_xyz_ao(2,j,i) = 1.d0 ! 1 * (x - A_x)^a_{x+1}
power_xyz_ao(2,j,i)= power_ao(j) + 1
enddo
enddo
END_PROVIDER
BEGIN_PROVIDER [ double precision, ao_coef_ord_grad_transp, (2,3,ao_prim_num_max,ao_num) ]
&BEGIN_PROVIDER [ integer, power_ord_grad_transp, (2,3,ao_num) ]
implicit none
BEGIN_DOC
! grad AO in terms of polynoms and coefficients
!
! WARNING !!!! SOME polynoms might be negative !!!!!
!
! WHEN IT IS THE CASE, coefficients are ZERO
END_DOC
integer :: i,j,power_ao(3), m,kk
do j=1, ao_num
power_ao(1:3)= ao_power(j,1:3)
do m = 1, 3
power_ord_grad_transp(1,m,j) = power_ao(m) - 1
power_ord_grad_transp(2,m,j) = power_ao(m) + 1
enddo
do i=1, ao_prim_num_max
do m = 1, 3
ao_coef_ord_grad_transp(1,m,i,j) = ao_coef_normalized_ordered(j,i) * dble(power_ao(m)) ! a_x * c_i
ao_coef_ord_grad_transp(2,m,i,j) = -2.d0 * ao_coef_normalized_ordered(j,i) * ao_expo_ordered_transp(i,j) ! -2 * c_i * alpha_i
do kk = 1, 2
if(power_ord_grad_transp(kk,m,j).lt.0)then
ao_coef_ord_grad_transp(kk,m,i,j) = 0.d0
endif
enddo
enddo
enddo
enddo
END_PROVIDER
BEGIN_PROVIDER [ double precision, ao_coef_ord_xyz_grad_transp, (4,3,ao_prim_num_max,ao_num) ]
&BEGIN_PROVIDER [ integer, power_ord_xyz_grad_transp, (4,3,ao_num) ]
implicit none
BEGIN_DOC
! x * d/dx of an AO in terms of polynoms and coefficients
!
! WARNING !!!! SOME polynoms might be negative !!!!!
!
! WHEN IT IS THE CASE, coefficients are ZERO
END_DOC
integer :: i,j,power_ao(3), m,num_ao,kk
double precision :: center_ao(1:3)
do j=1, ao_num
power_ao(1:3)= ao_power(j,1:3)
num_ao = ao_nucl(j)
center_ao(1:3) = nucl_coord(num_ao,1:3)
do m = 1, 3
power_ord_xyz_grad_transp(1,m,j) = power_ao(m) - 1
power_ord_xyz_grad_transp(2,m,j) = power_ao(m)
power_ord_xyz_grad_transp(3,m,j) = power_ao(m) + 1
power_ord_xyz_grad_transp(4,m,j) = power_ao(m) + 2
do kk = 1, 4
if(power_ord_xyz_grad_transp(kk,m,j).lt.0)then
power_ord_xyz_grad_transp(kk,m,j) = -1
endif
enddo
enddo
do i=1, ao_prim_num_max
do m = 1, 3
ao_coef_ord_xyz_grad_transp(1,m,i,j) = dble(power_ao(m)) * ao_coef_normalized_ordered(j,i) * center_ao(m)
ao_coef_ord_xyz_grad_transp(2,m,i,j) = dble(power_ao(m)) * ao_coef_normalized_ordered(j,i)
ao_coef_ord_xyz_grad_transp(3,m,i,j) = -2.d0 * ao_coef_normalized_ordered(j,i) * ao_expo_ordered_transp(i,j) * center_ao(m)
ao_coef_ord_xyz_grad_transp(4,m,i,j) = -2.d0 * ao_coef_normalized_ordered(j,i) * ao_expo_ordered_transp(i,j)
do kk = 1, 4
if(power_ord_xyz_grad_transp(kk,m,j).lt.0)then
ao_coef_ord_xyz_grad_transp(kk,m,i,j) = 0.d0
endif
enddo
enddo
enddo
enddo
END_PROVIDER
subroutine xyz_grad_phi_ao(r,i_ao,xyz_grad_phi)
implicit none
integer, intent(in) :: i_ao
double precision, intent(in) :: r(3)
double precision, intent(out):: xyz_grad_phi(3) ! x * d/dx phi i, y * d/dy phi_i, z * d/dz phi_
double precision :: center_ao(3),beta
double precision :: accu(3,4),dr(3),r2,pol_usual(3)
integer :: m,power_ao(3),num_ao,j_prim
power_ao(1:3)= ao_power(i_ao,1:3)
num_ao = ao_nucl(i_ao)
center_ao(1:3) = nucl_coord(num_ao,1:3)
dr(1) = (r(1) - center_ao(1))
dr(2) = (r(2) - center_ao(2))
dr(3) = (r(3) - center_ao(3))
r2 = 0.d0
do m = 1, 3
r2 += dr(m)*dr(m)
enddo
! computes the gaussian part
accu = 0.d0
do j_prim =1,ao_prim_num(i_ao)
beta = ao_expo_ordered_transp(j_prim,i_ao)
if(dabs(beta*r2).gt.50.d0)cycle
do m = 1, 3
accu(m,1) += ao_coef_ord_xyz_grad_transp(1,m,j_prim,i_ao) * dexp(-beta*r2)
accu(m,2) += ao_coef_ord_xyz_grad_transp(2,m,j_prim,i_ao) * dexp(-beta*r2)
accu(m,3) += ao_coef_ord_xyz_grad_transp(3,m,j_prim,i_ao) * dexp(-beta*r2)
accu(m,4) += ao_coef_ord_xyz_grad_transp(4,m,j_prim,i_ao) * dexp(-beta*r2)
enddo
enddo
! computes the polynom part
pol_usual = 0.d0
pol_usual(1) = dr(2)**dble(power_ao(2)) * dr(3)**dble(power_ao(3))
pol_usual(2) = dr(1)**dble(power_ao(1)) * dr(3)**dble(power_ao(3))
pol_usual(3) = dr(1)**dble(power_ao(1)) * dr(2)**dble(power_ao(2))
xyz_grad_phi = 0.d0
do m = 1, 3
xyz_grad_phi(m) += accu(m,2) * pol_usual(m) * dr(m)**dble(power_ord_xyz_grad_transp(2,m,i_ao))
xyz_grad_phi(m) += accu(m,3) * pol_usual(m) * dr(m)**dble(power_ord_xyz_grad_transp(3,m,i_ao))
xyz_grad_phi(m) += accu(m,4) * pol_usual(m) * dr(m)**dble(power_ord_xyz_grad_transp(4,m,i_ao))
if(power_ord_xyz_grad_transp(1,m,i_ao).lt.0)cycle
xyz_grad_phi(m) += accu(m,1) * pol_usual(m) * dr(m)**dble(power_ord_xyz_grad_transp(1,m,i_ao))
enddo
end
subroutine grad_phi_ao(r,i_ao,grad_xyz_phi)
implicit none
integer, intent(in) :: i_ao
double precision, intent(in) :: r(3)
double precision, intent(out):: grad_xyz_phi(3) ! x * phi i, y * phi_i, z * phi_
double precision :: center_ao(3),beta
double precision :: accu(3,2),dr(3),r2,pol_usual(3)
integer :: m,power_ao(3),num_ao,j_prim
power_ao(1:3)= ao_power(i_ao,1:3)
num_ao = ao_nucl(i_ao)
center_ao(1:3) = nucl_coord(num_ao,1:3)
dr(1) = (r(1) - center_ao(1))
dr(2) = (r(2) - center_ao(2))
dr(3) = (r(3) - center_ao(3))
r2 = 0.d0
do m = 1, 3
r2 += dr(m)*dr(m)
enddo
! computes the gaussian part
accu = 0.d0
do j_prim =1,ao_prim_num(i_ao)
beta = ao_expo_ordered_transp(j_prim,i_ao)
if(dabs(beta*r2).gt.50.d0)cycle
do m = 1, 3
accu(m,1) += ao_coef_ord_grad_transp(1,m,j_prim,i_ao) * dexp(-beta*r2)
accu(m,2) += ao_coef_ord_grad_transp(2,m,j_prim,i_ao) * dexp(-beta*r2)
enddo
enddo
! computes the polynom part
pol_usual = 0.d0
pol_usual(1) = dr(2)**dble(power_ao(2)) * dr(3)**dble(power_ao(3))
pol_usual(2) = dr(1)**dble(power_ao(1)) * dr(3)**dble(power_ao(3))
pol_usual(3) = dr(1)**dble(power_ao(1)) * dr(2)**dble(power_ao(2))
do m = 1, 3
grad_xyz_phi(m) = accu(m,2) * pol_usual(m) * dr(m)**dble(power_ord_grad_transp(2,m,i_ao))
if(power_ao(m)==0)cycle
grad_xyz_phi(m) += accu(m,1) * pol_usual(m) * dr(m)**dble(power_ord_grad_transp(1,m,i_ao))
enddo
end
subroutine xyz_phi_ao(r,i_ao,xyz_phi)
implicit none
integer, intent(in) :: i_ao
double precision, intent(in) :: r(3)
double precision, intent(out):: xyz_phi(3) ! x * phi i, y * phi_i, z * phi_i
double precision :: center_ao(3),beta
double precision :: accu,dr(3),r2,pol_usual(3)
integer :: m,power_ao(3),num_ao
power_ao(1:3)= ao_power(i_ao,1:3)
num_ao = ao_nucl(i_ao)
center_ao(1:3) = nucl_coord(num_ao,1:3)
dr(1) = (r(1) - center_ao(1))
dr(2) = (r(2) - center_ao(2))
dr(3) = (r(3) - center_ao(3))
r2 = 0.d0
do m = 1, 3
r2 += dr(m)*dr(m)
enddo
! computes the gaussian part
accu = 0.d0
do m=1,ao_prim_num(i_ao)
beta = ao_expo_ordered_transp(m,i_ao)
if(dabs(beta*r2).gt.50.d0)cycle
accu += ao_coef_normalized_ordered_transp(m,i_ao) * dexp(-beta*r2)
enddo
! computes the polynom part
pol_usual = 0.d0
pol_usual(1) = dr(2)**dble(power_ao(2)) * dr(3)**dble(power_ao(3))
pol_usual(2) = dr(1)**dble(power_ao(1)) * dr(3)**dble(power_ao(3))
pol_usual(3) = dr(1)**dble(power_ao(1)) * dr(2)**dble(power_ao(2))
do m = 1, 3
xyz_phi(m) = accu * pol_usual(m) * dr(m)**(dble(power_ao(m))) * ( coef_xyz_ao(1,m,i_ao) + coef_xyz_ao(2,m,i_ao) * dr(m) )
enddo
end
subroutine test_pol_xyz
implicit none
integer :: ipoint,i,j,m,jpoint
double precision :: r1(3),derf_mu_x
double precision :: weight1,r12,xyz_phi(3),grad_phi(3),xyz_grad_phi(3)
double precision, allocatable :: aos_array(:),aos_grad_array(:,:)
double precision :: num_xyz_phi(3),num_grad_phi(3),num_xyz_grad_phi(3)
double precision :: accu_xyz_phi(3),accu_grad_phi(3),accu_xyz_grad_phi(3)
double precision :: meta_accu_xyz_phi(3),meta_accu_grad_phi(3),meta_accu_xyz_grad_phi(3)
allocate(aos_array(ao_num),aos_grad_array(3,ao_num))
meta_accu_xyz_phi = 0.d0
meta_accu_grad_phi = 0.d0
meta_accu_xyz_grad_phi= 0.d0
do i = 1, ao_num
accu_xyz_phi = 0.d0
accu_grad_phi = 0.d0
accu_xyz_grad_phi= 0.d0
do ipoint = 1, n_points_final_grid
r1(:) = final_grid_points(:,ipoint)
weight1 = final_weight_at_r_vector(ipoint)
call give_all_aos_and_grad_at_r(r1,aos_array,aos_grad_array)
do m = 1, 3
num_xyz_phi(m) = r1(m) * aos_array(i)
num_grad_phi(m) = aos_grad_array(m,i)
num_xyz_grad_phi(m) = r1(m) * aos_grad_array(m,i)
enddo
call xyz_phi_ao(r1,i,xyz_phi)
call grad_phi_ao(r1,i,grad_phi)
call xyz_grad_phi_ao(r1,i,xyz_grad_phi)
do m = 1, 3
accu_xyz_phi(m) += weight1 * dabs(num_xyz_phi(m) - xyz_phi(m) )
accu_grad_phi(m) += weight1 * dabs(num_grad_phi(m) - grad_phi(m) )
accu_xyz_grad_phi(m) += weight1 * dabs(num_xyz_grad_phi(m) - xyz_grad_phi(m))
enddo
enddo
print*,''
print*,''
print*,'i,',i
print*,''
do m = 1, 3
! print*, 'm, accu_xyz_phi(m) ' ,m, accu_xyz_phi(m)
! print*, 'm, accu_grad_phi(m) ' ,m, accu_grad_phi(m)
print*, 'm, accu_xyz_grad_phi' ,m, accu_xyz_grad_phi(m)
enddo
do m = 1, 3
meta_accu_xyz_phi(m) += dabs(accu_xyz_phi(m))
meta_accu_grad_phi(m) += dabs(accu_grad_phi(m))
meta_accu_xyz_grad_phi(m) += dabs(accu_xyz_grad_phi(m))
enddo
enddo
do m = 1, 3
! print*, 'm, meta_accu_xyz_phi(m) ' ,m, meta_accu_xyz_phi(m)
! print*, 'm, meta_accu_grad_phi(m) ' ,m, meta_accu_grad_phi(m)
print*, 'm, meta_accu_xyz_grad_phi' ,m, meta_accu_xyz_grad_phi(m)
enddo
end
subroutine test_ints_semi_bis
implicit none
integer :: ipoint,i,j,m
double precision :: r1(3), aos_grad_array_r1(3, ao_num), aos_array_r1(ao_num)
double precision :: C_center(3), weight1,mu_in,r12,derf_mu_x,dxyz_ints(3),NAI_pol_mult_erf_ao
double precision :: ao_mat(ao_num,ao_num),ao_xmat(3,ao_num,ao_num),accu1, accu2(3)
mu_in = 0.5d0
C_center = 0.d0
C_center(1) = 0.25d0
C_center(3) = 1.12d0
C_center(2) = -1.d0
ao_mat = 0.d0
ao_xmat = 0.d0
do ipoint = 1, n_points_final_grid
r1(1) = final_grid_points(1,ipoint)
r1(2) = final_grid_points(2,ipoint)
r1(3) = final_grid_points(3,ipoint)
call give_all_aos_and_grad_at_r(r1,aos_array_r1,aos_grad_array_r1)
weight1 = final_weight_at_r_vector(ipoint)
r12 = (r1(1) - C_center(1))**2.d0 + (r1(2) - C_center(2))**2.d0 + (r1(3) - C_center(3))**2.d0
r12 = dsqrt(r12)
do i = 1, ao_num
do j = 1, ao_num
ao_mat(j,i) += aos_array_r1(i) * aos_array_r1(j) * weight1 * derf_mu_x(mu_in,r12)
do m = 1, 3
ao_xmat(m,j,i) += r1(m) * aos_array_r1(j) * aos_grad_array_r1(m,i) * weight1 * derf_mu_x(mu_in,r12)
enddo
enddo
enddo
enddo
accu1 = 0.d0
accu2 = 0.d0
accu1relat = 0.d0
accu2relat = 0.d0
double precision :: accu1relat, accu2relat(3)
double precision :: contrib(3)
do i = 1, ao_num
do j = 1, ao_num
call phi_j_erf_mu_r_xyz_dxyz_phi(i,j,mu_in, C_center, dxyz_ints)
print*,''
print*,'i,j',i,j
print*,dxyz_ints(:)
print*,ao_xmat(:,j,i)
do m = 1, 3
contrib(m) = dabs(ao_xmat(m,j,i) - dxyz_ints(m))
accu2(m) += contrib(m)
if(dabs(ao_xmat(m,j,i)).gt.1.d-10)then
accu2relat(m) += dabs(ao_xmat(m,j,i) - dxyz_ints(m))/dabs(ao_xmat(m,j,i))
endif
enddo
print*,contrib
enddo
print*,''
enddo
print*,'accu2relat = '
print*, accu2relat /dble(ao_num * ao_num)
end

View File

@ -0,0 +1,11 @@
[mo_r_coef]
type: double precision
doc: right-coefficient of the i-th |AO| on the j-th |MO|
interface: ezfio
size: (ao_basis.ao_num,mo_basis.mo_num)
[mo_l_coef]
type: double precision
doc: right-coefficient of the i-th |AO| on the j-th |MO|
interface: ezfio
size: (ao_basis.ao_num,mo_basis.mo_num)

3
src/bi_ortho_mos/NEED Normal file
View File

@ -0,0 +1,3 @@
mo_basis
becke_numerical_grid
dft_utils_in_r

View File

@ -0,0 +1,70 @@
! ---
BEGIN_PROVIDER [double precision, TCSCF_bi_ort_dm_ao_alpha, (ao_num, ao_num) ]
BEGIN_DOC
! TCSCF_bi_ort_dm_ao_alpha(i,j) = <Chi_0| a^dagger_i,alpha a_j,alpha |Phi_0> where i,j are AO basis.
!
! This is the equivalent of the alpha density of the HF Slater determinant, but with a couple of bi-orthonormal Slater determinant |Chi_0> and |Phi_0>
END_DOC
implicit none
PROVIDE mo_l_coef mo_r_coef
call dgemm( 'N', 'T', ao_num, ao_num, elec_alpha_num, 1.d0 &
, mo_l_coef, size(mo_l_coef, 1), mo_r_coef, size(mo_r_coef, 1) &
!, mo_r_coef, size(mo_r_coef, 1), mo_l_coef, size(mo_l_coef, 1) &
, 0.d0, TCSCF_bi_ort_dm_ao_alpha, size(TCSCF_bi_ort_dm_ao_alpha, 1) )
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, TCSCF_bi_ort_dm_ao_beta, (ao_num, ao_num) ]
BEGIN_DOC
! TCSCF_bi_ort_dm_ao_beta(i,j) = <Chi_0| a^dagger_i,beta a_j,beta |Phi_0> where i,j are AO basis.
!
! This is the equivalent of the beta density of the HF Slater determinant, but with a couple of bi-orthonormal Slater determinant |Chi_0> and |Phi_0>
END_DOC
implicit none
PROVIDE mo_l_coef mo_r_coef
call dgemm( 'N', 'T', ao_num, ao_num, elec_beta_num, 1.d0 &
, mo_l_coef, size(mo_l_coef, 1), mo_r_coef, size(mo_r_coef, 1) &
!, mo_r_coef, size(mo_r_coef, 1), mo_l_coef, size(mo_l_coef, 1) &
, 0.d0, TCSCF_bi_ort_dm_ao_beta, size(TCSCF_bi_ort_dm_ao_beta, 1) )
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, TCSCF_bi_ort_dm_ao, (ao_num, ao_num) ]
BEGIN_DOC
! TCSCF_bi_ort_dm_ao(i,j) = <Chi_0| a^dagger_i,beta+alpha a_j,beta+alpha |Phi_0> where i,j are AO basis.
!
! This is the equivalent of the total electronic density of the HF Slater determinant, but with a couple of bi-orthonormal Slater determinant |Chi_0> and |Phi_0>
END_DOC
implicit none
PROVIDE mo_l_coef mo_r_coef
ASSERT(size(TCSCF_bi_ort_dm_ao, 1) == size(TCSCF_bi_ort_dm_ao_alpha, 1))
if(elec_alpha_num==elec_beta_num) then
TCSCF_bi_ort_dm_ao = TCSCF_bi_ort_dm_ao_alpha + TCSCF_bi_ort_dm_ao_alpha
else
ASSERT(size(TCSCF_bi_ort_dm_ao, 1) == size(TCSCF_bi_ort_dm_ao_beta, 1))
TCSCF_bi_ort_dm_ao = TCSCF_bi_ort_dm_ao_alpha + TCSCF_bi_ort_dm_ao_beta
endif
END_PROVIDER
! ---

View File

@ -0,0 +1,137 @@
! TODO: left & right MO without duplicate AO calculation
! ---
BEGIN_PROVIDER[double precision, mos_r_in_r_array, (mo_num, n_points_final_grid)]
BEGIN_DOC
! mos_in_r_array(i,j) = value of the ith RIGHT mo on the jth grid point
END_DOC
implicit none
integer :: i, j
double precision :: mos_array(mo_num), r(3)
!$OMP PARALLEL DO &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i, j, r, mos_array) &
!$OMP SHARED (mos_r_in_r_array, n_points_final_grid, mo_num, final_grid_points)
do i = 1, n_points_final_grid
r(1) = final_grid_points(1,i)
r(2) = final_grid_points(2,i)
r(3) = final_grid_points(3,i)
call give_all_mos_r_at_r(r, mos_array)
do j = 1, mo_num
mos_r_in_r_array(j,i) = mos_array(j)
enddo
enddo
!$OMP END PARALLEL DO
END_PROVIDER
! ---
BEGIN_PROVIDER[double precision, mos_r_in_r_array_transp, (n_points_final_grid, mo_num)]
BEGIN_DOC
! mos_r_in_r_array_transp(i,j) = value of the jth mo on the ith grid point
END_DOC
implicit none
integer :: i,j
do i = 1, n_points_final_grid
do j = 1, mo_num
mos_r_in_r_array_transp(i,j) = mos_r_in_r_array(j,i)
enddo
enddo
END_PROVIDER
! ---
subroutine give_all_mos_r_at_r(r, mos_r_array)
BEGIN_DOC
! mos_r_array(i) = ith RIGHT MO function evaluated at "r"
END_DOC
implicit none
double precision, intent(in) :: r(3)
double precision, intent(out) :: mos_r_array(mo_num)
double precision :: aos_array(ao_num)
call give_all_aos_at_r(r, aos_array)
call dgemv('N', mo_num, ao_num, 1.d0, mo_r_coef_transp, mo_num, aos_array, 1, 0.d0, mos_r_array, 1)
end subroutine give_all_mos_r_at_r
! ---
BEGIN_PROVIDER[double precision, mos_l_in_r_array, (mo_num, n_points_final_grid)]
BEGIN_DOC
! mos_in_r_array(i,j) = value of the ith LEFT mo on the jth grid point
END_DOC
implicit none
integer :: i, j
double precision :: mos_array(mo_num), r(3)
!$OMP PARALLEL DO &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,r,mos_array,j) &
!$OMP SHARED(mos_l_in_r_array,n_points_final_grid,mo_num,final_grid_points)
do i = 1, n_points_final_grid
r(1) = final_grid_points(1,i)
r(2) = final_grid_points(2,i)
r(3) = final_grid_points(3,i)
call give_all_mos_l_at_r(r, mos_array)
do j = 1, mo_num
mos_l_in_r_array(j,i) = mos_array(j)
enddo
enddo
!$OMP END PARALLEL DO
END_PROVIDER
! ---
subroutine give_all_mos_l_at_r(r, mos_l_array)
BEGIN_DOC
! mos_l_array(i) = ith LEFT MO function evaluated at "r"
END_DOC
implicit none
double precision, intent(in) :: r(3)
double precision, intent(out) :: mos_l_array(mo_num)
double precision :: aos_array(ao_num)
call give_all_aos_at_r(r, aos_array)
call dgemv('N', mo_num, ao_num, 1.d0, mo_l_coef_transp, mo_num, aos_array, 1, 0.d0, mos_l_array, 1)
end subroutine give_all_mos_l_at_r
! ---
BEGIN_PROVIDER[double precision, mos_l_in_r_array_transp,(n_points_final_grid,mo_num)]
BEGIN_DOC
! mos_l_in_r_array_transp(i,j) = value of the jth mo on the ith grid point
END_DOC
implicit none
integer :: i, j
do i = 1, n_points_final_grid
do j = 1, mo_num
mos_l_in_r_array_transp(i,j) = mos_l_in_r_array(j,i)
enddo
enddo
END_PROVIDER
! ---

View File

@ -0,0 +1,100 @@
BEGIN_PROVIDER[double precision, mos_r_grad_in_r_array,(mo_num,n_points_final_grid,3)]
implicit none
BEGIN_DOC
! mos_r_grad_in_r_array(i,j,k) = value of the kth component of the gradient of ith RIGHT mo on the jth grid point
!
! k = 1 : x, k= 2, y, k 3, z
END_DOC
integer :: m
mos_r_grad_in_r_array = 0.d0
do m=1,3
call dgemm('N','N',mo_num,n_points_final_grid,ao_num,1.d0,mo_r_coef_transp,mo_num,aos_grad_in_r_array(1,1,m),ao_num,0.d0,mos_r_grad_in_r_array(1,1,m),mo_num)
enddo
END_PROVIDER
BEGIN_PROVIDER[double precision, mos_r_grad_in_r_array_transp,(3,mo_num,n_points_final_grid)]
implicit none
BEGIN_DOC
! mos_r_grad_in_r_array_transp(i,j,k) = value of the kth component of the gradient of jth RIGHT mo on the ith grid point
!
! k = 1 : x, k= 2, y, k 3, z
END_DOC
integer :: m
integer :: i,j
mos_r_grad_in_r_array_transp = 0.d0
do i = 1, n_points_final_grid
do j = 1, mo_num
do m = 1, 3
mos_r_grad_in_r_array_transp(m,j,i) = mos_r_grad_in_r_array(j,i,m)
enddo
enddo
enddo
END_PROVIDER
BEGIN_PROVIDER[double precision, mos_r_grad_in_r_array_transp_bis,(3,n_points_final_grid,mo_num)]
implicit none
BEGIN_DOC
! mos_r_grad_in_r_array_transp(i,j,k) = value of the ith component of the gradient on the jth grid point of jth RIGHT MO
END_DOC
integer :: m
integer :: i,j
mos_r_grad_in_r_array_transp_bis = 0.d0
do j = 1, mo_num
do i = 1, n_points_final_grid
do m = 1, 3
mos_r_grad_in_r_array_transp_bis(m,i,j) = mos_r_grad_in_r_array(j,i,m)
enddo
enddo
enddo
END_PROVIDER
BEGIN_PROVIDER[double precision, mos_l_grad_in_r_array,(mo_num,n_points_final_grid,3)]
implicit none
BEGIN_DOC
! mos_l_grad_in_r_array(i,j,k) = value of the kth component of the gradient of ith RIGHT mo on the jth grid point
!
! k = 1 : x, k= 2, y, k 3, z
END_DOC
integer :: m
mos_l_grad_in_r_array = 0.d0
do m=1,3
call dgemm('N','N',mo_num,n_points_final_grid,ao_num,1.d0,mo_r_coef_transp,mo_num,aos_grad_in_r_array(1,1,m),ao_num,0.d0,mos_l_grad_in_r_array(1,1,m),mo_num)
enddo
END_PROVIDER
BEGIN_PROVIDER[double precision, mos_l_grad_in_r_array_transp,(3,mo_num,n_points_final_grid)]
implicit none
BEGIN_DOC
! mos_l_grad_in_r_array_transp(i,j,k) = value of the kth component of the gradient of jth RIGHT mo on the ith grid point
!
! k = 1 : x, k= 2, y, k 3, z
END_DOC
integer :: m
integer :: i,j
mos_l_grad_in_r_array_transp = 0.d0
do i = 1, n_points_final_grid
do j = 1, mo_num
do m = 1, 3
mos_l_grad_in_r_array_transp(m,j,i) = mos_l_grad_in_r_array(j,i,m)
enddo
enddo
enddo
END_PROVIDER
BEGIN_PROVIDER[double precision, mos_l_grad_in_r_array_transp_bis,(3,n_points_final_grid,mo_num)]
implicit none
BEGIN_DOC
! mos_l_grad_in_r_array_transp(i,j,k) = value of the ith component of the gradient on the jth grid point of jth RIGHT MO
END_DOC
integer :: m
integer :: i,j
mos_l_grad_in_r_array_transp_bis = 0.d0
do j = 1, mo_num
do i = 1, n_points_final_grid
do m = 1, 3
mos_l_grad_in_r_array_transp_bis(m,i,j) = mos_l_grad_in_r_array(j,i,m)
enddo
enddo
enddo
END_PROVIDER

View File

@ -0,0 +1,224 @@
! ---
subroutine ao_to_mo_bi_ortho(A_ao, LDA_ao, A_mo, LDA_mo)
BEGIN_DOC
!
! Transform A from the |AO| basis to the BI ORTHONORMAL MOS
!
! $C_L^\dagger.A_{ao}.C_R$ where C_L and C_R are the LEFT and RIGHT MO coefs
!
END_DOC
implicit none
integer, intent(in) :: LDA_ao, LDA_mo
double precision, intent(in) :: A_ao(LDA_ao,ao_num)
double precision, intent(out) :: A_mo(LDA_mo,mo_num)
double precision, allocatable :: T(:,:)
allocate ( T(ao_num,mo_num) )
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: T
! T = A_ao x mo_r_coef
call dgemm( 'N', 'N', ao_num, mo_num, ao_num, 1.d0 &
, A_ao, LDA_ao, mo_r_coef, size(mo_r_coef, 1) &
, 0.d0, T, size(T, 1) )
! A_mo = mo_l_coef.T x T
call dgemm( 'T', 'N', mo_num, mo_num, ao_num, 1.d0 &
, mo_l_coef, size(mo_l_coef, 1), T, size(T, 1) &
, 0.d0, A_mo, LDA_mo )
! call restore_symmetry(mo_num,mo_num,A_mo,size(A_mo,1),1.d-12)
deallocate(T)
end subroutine ao_to_mo_bi_ortho
! ---
subroutine mo_to_ao_bi_ortho(A_mo, LDA_mo, A_ao, LDA_ao)
BEGIN_DOC
!
! mo_l_coef.T x A_ao x mo_r_coef = A_mo
! mo_l_coef.T x ao_overlap x mo_r_coef = I
!
! ==> A_ao = (ao_overlap x mo_r_coef) x A_mo x (ao_overlap x mo_l_coef).T
!
END_DOC
implicit none
integer, intent(in) :: LDA_ao, LDA_mo
double precision, intent(in) :: A_mo(LDA_mo,mo_num)
double precision, intent(out) :: A_ao(LDA_ao,ao_num)
double precision, allocatable :: tmp_1(:,:), tmp_2(:,:)
! ao_overlap x mo_r_coef
allocate( tmp_1(ao_num,mo_num) )
call dgemm( 'N', 'N', ao_num, mo_num, ao_num, 1.d0 &
, ao_overlap, size(ao_overlap, 1), mo_r_coef, size(mo_r_coef, 1) &
, 0.d0, tmp_1, size(tmp_1, 1) )
! (ao_overlap x mo_r_coef) x A_mo
allocate( tmp_2(ao_num,mo_num) )
call dgemm( 'N', 'N', ao_num, mo_num, mo_num, 1.d0 &
, tmp_1, size(tmp_1, 1), A_mo, LDA_mo &
, 0.d0, tmp_2, size(tmp_2, 1) )
! ao_overlap x mo_l_coef
tmp_1 = 0.d0
call dgemm( 'N', 'N', ao_num, mo_num, ao_num, 1.d0 &
, ao_overlap, size(ao_overlap, 1), mo_l_coef, size(mo_l_coef, 1) &
, 0.d0, tmp_1, size(tmp_1, 1) )
! (ao_overlap x mo_r_coef) x A_mo x (ao_overlap x mo_l_coef).T
call dgemm( 'N', 'T', ao_num, ao_num, mo_num, 1.d0 &
, tmp_2, size(tmp_2, 1), tmp_1, size(tmp_1, 1) &
, 0.d0, A_ao, LDA_ao )
deallocate(tmp_1, tmp_2)
end subroutine mo_to_ao_bi_ortho
! ---
BEGIN_PROVIDER [ double precision, mo_r_coef, (ao_num, mo_num) ]
BEGIN_DOC
!
! Molecular right-orbital coefficients on |AO| basis set
!
END_DOC
implicit none
integer :: i, j
logical :: exists
PROVIDE ezfio_filename
if (mpi_master) then
call ezfio_has_bi_ortho_mos_mo_r_coef(exists)
endif
IRP_IF MPI_DEBUG
print *, irp_here, mpi_rank
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
IRP_ENDIF
IRP_IF MPI
include 'mpif.h'
integer :: ierr
call MPI_BCAST(exists, 1, MPI_LOGICAL, 0, MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
stop 'Unable to read mo_r_coef with MPI'
endif
IRP_ENDIF
if (exists) then
if (mpi_master) then
call ezfio_get_bi_ortho_mos_mo_r_coef(mo_r_coef)
write(*,*) 'Read mo_r_coef'
endif
IRP_IF MPI
call MPI_BCAST(mo_r_coef, mo_num*ao_num, MPI_DOUBLE_PRECISION, 0, MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
stop 'Unable to read mo_r_coef with MPI'
endif
IRP_ENDIF
else
print*, 'mo_r_coef are mo_coef'
do i = 1, mo_num
do j = 1, ao_num
mo_r_coef(j,i) = mo_coef(j,i)
enddo
enddo
endif
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, mo_l_coef, (ao_num, mo_num) ]
BEGIN_DOC
!
! Molecular left-orbital coefficients on |AO| basis set
!
END_DOC
implicit none
integer :: i, j
logical :: exists
PROVIDE ezfio_filename
if (mpi_master) then
call ezfio_has_bi_ortho_mos_mo_l_coef(exists)
endif
IRP_IF MPI_DEBUG
print *, irp_here, mpi_rank
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
IRP_ENDIF
IRP_IF MPI
include 'mpif.h'
integer :: ierr
call MPI_BCAST(exists, 1, MPI_LOGICAL, 0, MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
stop 'Unable to read mo_l_coef with MPI'
endif
IRP_ENDIF
if (exists) then
if (mpi_master) then
call ezfio_get_bi_ortho_mos_mo_l_coef(mo_l_coef)
write(*,*) 'Read mo_l_coef'
endif
IRP_IF MPI
call MPI_BCAST(mo_l_coef, mo_num*ao_num, MPI_DOUBLE_PRECISION, 0, MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
stop 'Unable to read mo_l_coef with MPI'
endif
IRP_ENDIF
else
print*, 'mo_l_coef are mo_coef'
do i = 1, mo_num
do j = 1, ao_num
mo_l_coef(j,i) = mo_coef(j,i)
enddo
enddo
endif
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, mo_r_coef_transp, (mo_num, ao_num)]
implicit none
integer :: j, m
do j = 1, mo_num
do m = 1, ao_num
mo_r_coef_transp(j,m) = mo_r_coef(m,j)
enddo
enddo
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, mo_l_coef_transp, (mo_num, ao_num)]
implicit none
integer :: j, m
do j = 1, mo_num
do m = 1, ao_num
mo_l_coef_transp(j,m) = mo_l_coef(m,j)
enddo
enddo
END_PROVIDER
! ---

View File

@ -0,0 +1,160 @@
BEGIN_PROVIDER [ double precision, overlap_bi_ortho, (mo_num, mo_num)]
&BEGIN_PROVIDER [ double precision, overlap_diag_bi_ortho, (mo_num)]
BEGIN_DOC
! Overlap matrix between the RIGHT and LEFT MOs. Should be the identity matrix
END_DOC
implicit none
integer :: i, k, m, n
double precision :: accu_d, accu_nd
double precision, allocatable :: tmp(:,:)
! TODO : re do the DEGEMM
overlap_bi_ortho = 0.d0
do i = 1, mo_num
do k = 1, mo_num
do m = 1, ao_num
do n = 1, ao_num
overlap_bi_ortho(k,i) += ao_overlap(n,m) * mo_l_coef(n,k) * mo_r_coef(m,i)
enddo
enddo
enddo
enddo
! allocate( tmp(mo_num,ao_num) )
!
! ! tmp <-- L.T x S_ao
! call dgemm( "T", "N", mo_num, ao_num, ao_num, 1.d0 &
! , mo_l_coef, size(mo_l_coef, 1), ao_overlap, size(ao_overlap, 1) &
! , 0.d0, tmp, size(tmp, 1) )
!
! ! S <-- tmp x R
! call dgemm( "N", "N", mo_num, mo_num, ao_num, 1.d0 &
! , tmp, size(tmp, 1), mo_r_coef, size(mo_r_coef, 1) &
! , 0.d0, overlap_bi_ortho, size(overlap_bi_ortho, 1) )
!
! deallocate( tmp )
do i = 1, mo_num
overlap_diag_bi_ortho(i) = overlap_bi_ortho(i,i)
enddo
accu_d = 0.d0
accu_nd = 0.d0
do i = 1, mo_num
do k = 1, mo_num
if(i==k) then
accu_d += dabs(overlap_bi_ortho(k,i))
else
accu_nd += dabs(overlap_bi_ortho(k,i))
endif
enddo
enddo
accu_d = accu_d/dble(mo_num)
accu_nd = accu_nd/dble(mo_num**2-mo_num)
if(dabs(accu_d-1.d0).gt.1.d-10.or.dabs(accu_nd).gt.1.d-10)then
print*,'Warning !!!'
print*,'Average trace of overlap_bi_ortho is different from 1 by ', dabs(accu_d-1.d0)
print*,'And bi orthogonality is off by an average of ',accu_nd
print*,'****************'
print*,'Overlap matrix betwee mo_l_coef and mo_r_coef '
do i = 1, mo_num
write(*,'(100(F16.10,X))')overlap_bi_ortho(i,:)
enddo
endif
print*,'Average trace of overlap_bi_ortho (should be 1.)'
print*,'accu_d = ',accu_d
print*,'Sum of off diagonal terms of overlap_bi_ortho (should be zero)'
print*,'accu_nd = ',accu_nd
print*,'****************'
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, overlap_mo_r, (mo_num, mo_num)]
&BEGIN_PROVIDER [ double precision, overlap_mo_l, (mo_num, mo_num)]
BEGIN_DOC
! overlap_mo_r_mo(j,i) = <MO_i|MO_R_j>
END_DOC
implicit none
integer :: i, j, p, q
overlap_mo_r = 0.d0
overlap_mo_l = 0.d0
do i = 1, mo_num
do j = 1, mo_num
do p = 1, ao_num
do q = 1, ao_num
overlap_mo_r(j,i) += mo_r_coef(q,i) * mo_r_coef(p,j) * ao_overlap(q,p)
overlap_mo_l(j,i) += mo_l_coef(q,i) * mo_l_coef(p,j) * ao_overlap(q,p)
enddo
enddo
enddo
enddo
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, overlap_mo_r_mo, (mo_num, mo_num)]
&BEGIN_PROVIDER [ double precision, overlap_mo_l_mo, (mo_num, mo_num)]
BEGIN_DOC
! overlap_mo_r_mo(j,i) = <MO_j|MO_R_i>
END_DOC
implicit none
integer :: i, j, p, q
overlap_mo_r_mo = 0.d0
overlap_mo_l_mo = 0.d0
do i = 1, mo_num
do j = 1, mo_num
do p = 1, ao_num
do q = 1, ao_num
overlap_mo_r_mo(j,i) += mo_coef(p,j) * mo_r_coef(q,i) * ao_overlap(q,p)
overlap_mo_l_mo(j,i) += mo_coef(p,j) * mo_l_coef(q,i) * ao_overlap(q,p)
enddo
enddo
enddo
enddo
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, angle_left_right, (mo_num)]
&BEGIN_PROVIDER [ double precision, max_angle_left_right]
BEGIN_DOC
! angle_left_right(i) = angle between the left-eigenvector chi_i and the right-eigenvector phi_i
END_DOC
implicit none
integer :: i, j
double precision :: left, right, arg
double precision :: angle(mo_num)
do i = 1, mo_num
left = overlap_mo_l(i,i)
right = overlap_mo_r(i,i)
arg = min(overlap_bi_ortho(i,i)/(left*right),1.d0)
arg = max(arg, -1.d0)
angle_left_right(i) = dacos(arg) * 180.d0/dacos(-1.d0)
enddo
angle(1:mo_num) = dabs(angle_left_right(1:mo_num))
max_angle_left_right = maxval(angle)
END_PROVIDER
! ---