mirror of
https://github.com/LCPQ/quantum_package
synced 2024-11-03 20:54:00 +01:00
Small changes in MRCC
This commit is contained in:
parent
4694c08979
commit
2f1c7c5ce9
@ -271,7 +271,7 @@ subroutine mrcc_dress(delta_ij_, delta_ii_, Nstates, Ndet_non_ref, Ndet_ref,i_ge
|
||||
!delta_ii_(i_state,i_I) = 0.d0
|
||||
do l_sd=1,idx_alpha(0)
|
||||
k_sd = idx_alpha(l_sd)
|
||||
delta_ij_(i_state,k_sd,i_I) = delta_ij_(i_state,k_sd,i_I) + dIa_hla(i_state,k_sd)
|
||||
delta_ij_(i_state,k_sd,i_I) = delta_ij_(i_state,k_sd,i_I) + 0.5d0 * dIa_hla(i_state,k_sd)
|
||||
enddo
|
||||
endif
|
||||
enddo
|
||||
|
@ -685,7 +685,7 @@ END_PROVIDER
|
||||
|
||||
|
||||
|
||||
do s = 1, N_states
|
||||
do s=1, N_states
|
||||
|
||||
A_val = 0d0
|
||||
A_ind = 0
|
||||
@ -698,7 +698,7 @@ END_PROVIDER
|
||||
|
||||
!$OMP PARALLEL default(none) shared(psi_non_ref, hh_exists, pp_exists, N_int, A_val, A_ind)&
|
||||
!$OMP shared(s, hh_shortcut, psi_ref_coef, N_det_non_ref, psi_non_ref_sorted, psi_non_ref_sorted_idx, psi_ref, N_det_ref)&
|
||||
!$OMP shared(active, active_hh_idx, active_pp_idx, nactive)&
|
||||
!$OMP shared(active, active_hh_idx, active_pp_idx, nactive) &
|
||||
!$OMP private(lref, pp, II, ok, myMask, myDet, ind, phase, wk, ppp, hh)
|
||||
allocate(lref(N_det_non_ref))
|
||||
!$OMP DO schedule(static,10)
|
||||
@ -901,11 +901,11 @@ END_PROVIDER
|
||||
endif
|
||||
resold = res
|
||||
|
||||
if(mod(k, 5) == 0) then
|
||||
if(mod(k, 100) == 0) then
|
||||
print *, "res ", k, res
|
||||
end if
|
||||
|
||||
if(res < 1d-12) exit
|
||||
if(res < 1d-9) exit
|
||||
end do
|
||||
|
||||
|
||||
@ -923,7 +923,119 @@ END_PROVIDER
|
||||
|
||||
print *, k, "res : ", res, "norm : ", sqrt(norm)
|
||||
|
||||
!dIj_unique(:size(X), s) = X(:)
|
||||
!---------------
|
||||
! double precision :: e_0, overlap
|
||||
! double precision, allocatable :: u_0(:)
|
||||
! integer(bit_kind), allocatable :: keys_tmp(:,:,:)
|
||||
! allocate (u_0(N_det), keys_tmp(N_int,2,N_det) )
|
||||
! k=0
|
||||
! overlap = 0.d0
|
||||
! do i=1,N_det_ref
|
||||
! k = k+1
|
||||
! u_0(k) = psi_ref_coef(i,1)
|
||||
! keys_tmp(:,:,k) = psi_ref(:,:,i)
|
||||
! overlap += u_0(k)*psi_ref_coef(i,1)
|
||||
! enddo
|
||||
! norm = 0.d0
|
||||
! do i=1,N_det_non_ref
|
||||
! k = k+1
|
||||
! u_0(k) = psi_non_ref_coef(i,1)
|
||||
! keys_tmp(:,:,k) = psi_non_ref(:,:,i)
|
||||
! overlap += u_0(k)*psi_non_ref_coef(i,1)
|
||||
! enddo
|
||||
!
|
||||
! call u_0_H_u_0(e_0,u_0,N_det,keys_tmp,N_int,1,N_det)
|
||||
! print *, 'Energy of |Psi_CASSD> : ', e_0 + nuclear_repulsion, overlap
|
||||
!
|
||||
! k=0
|
||||
! overlap = 0.d0
|
||||
! do i=1,N_det_ref
|
||||
! k = k+1
|
||||
! u_0(k) = psi_ref_coef(i,1)
|
||||
! keys_tmp(:,:,k) = psi_ref(:,:,i)
|
||||
! overlap += u_0(k)*psi_ref_coef(i,1)
|
||||
! enddo
|
||||
! norm = 0.d0
|
||||
! do i=1,N_det_non_ref
|
||||
! k = k+1
|
||||
! ! f is such that f.\tilde{c_i} = c_i
|
||||
! f = psi_non_ref_coef(i,1) / rho_mrcc(i,1)
|
||||
!
|
||||
! ! Avoid numerical instabilities
|
||||
! f = min(f,2.d0)
|
||||
! f = max(f,-2.d0)
|
||||
!
|
||||
! f = 1.d0
|
||||
!
|
||||
! u_0(k) = rho_mrcc(i,1)*f
|
||||
! keys_tmp(:,:,k) = psi_non_ref(:,:,i)
|
||||
! norm += u_0(k)**2
|
||||
! overlap += u_0(k)*psi_non_ref_coef(i,1)
|
||||
! enddo
|
||||
!
|
||||
! call u_0_H_u_0(e_0,u_0,N_det,keys_tmp,N_int,1,N_det)
|
||||
! print *, 'Energy of |(1+T)Psi_0> : ', e_0 + nuclear_repulsion, overlap
|
||||
!
|
||||
! f = 1.d0/norm
|
||||
! norm = 1.d0
|
||||
! do i=1,N_det_ref
|
||||
! norm = norm - psi_ref_coef(i,s)*psi_ref_coef(i,s)
|
||||
! enddo
|
||||
! f = dsqrt(f*norm)
|
||||
! overlap = norm
|
||||
! do i=1,N_det_non_ref
|
||||
! u_0(k) = rho_mrcc(i,1)*f
|
||||
! overlap += u_0(k)*psi_non_ref_coef(i,1)
|
||||
! enddo
|
||||
!
|
||||
! call u_0_H_u_0(e_0,u_0,N_det,keys_tmp,N_int,1,N_det)
|
||||
! print *, 'Energy of |(1+T)Psi_0> (normalized) : ', e_0 + nuclear_repulsion, overlap
|
||||
!
|
||||
! k=0
|
||||
! overlap = 0.d0
|
||||
! do i=1,N_det_ref
|
||||
! k = k+1
|
||||
! u_0(k) = psi_ref_coef(i,1)
|
||||
! keys_tmp(:,:,k) = psi_ref(:,:,i)
|
||||
! overlap += u_0(k)*psi_ref_coef(i,1)
|
||||
! enddo
|
||||
! norm = 0.d0
|
||||
! do i=1,N_det_non_ref
|
||||
! k = k+1
|
||||
! ! f is such that f.\tilde{c_i} = c_i
|
||||
! f = psi_non_ref_coef(i,1) / rho_mrcc(i,1)
|
||||
!
|
||||
! ! Avoid numerical instabilities
|
||||
! f = min(f,2.d0)
|
||||
! f = max(f,-2.d0)
|
||||
!
|
||||
! u_0(k) = rho_mrcc(i,1)*f
|
||||
! keys_tmp(:,:,k) = psi_non_ref(:,:,i)
|
||||
! norm += u_0(k)**2
|
||||
! overlap += u_0(k)*psi_non_ref_coef(i,1)
|
||||
! enddo
|
||||
!
|
||||
! call u_0_H_u_0(e_0,u_0,N_det,keys_tmp,N_int,1,N_det)
|
||||
! print *, 'Energy of |(1+T)Psi_0> (mu_i): ', e_0 + nuclear_repulsion, overlap
|
||||
!
|
||||
! f = 1.d0/norm
|
||||
! norm = 1.d0
|
||||
! do i=1,N_det_ref
|
||||
! norm = norm - psi_ref_coef(i,s)*psi_ref_coef(i,s)
|
||||
! enddo
|
||||
! overlap = norm
|
||||
! f = dsqrt(f*norm)
|
||||
! do i=1,N_det_non_ref
|
||||
! u_0(k) = rho_mrcc(i,1)*f
|
||||
! overlap += u_0(k)*psi_non_ref_coef(i,1)
|
||||
! enddo
|
||||
!
|
||||
! call u_0_H_u_0(e_0,u_0,N_det,keys_tmp,N_int,1,N_det)
|
||||
! print *, 'Energy of |(1+T)Psi_0> (normalized mu_i) : ', e_0 + nuclear_repulsion, overlap
|
||||
!
|
||||
! deallocate(u_0, keys_tmp)
|
||||
!
|
||||
!---------------
|
||||
|
||||
norm = 0.d0
|
||||
double precision :: f
|
||||
@ -971,6 +1083,7 @@ END_PROVIDER
|
||||
|
||||
dIj_unique(:size(X), s) = X(:)
|
||||
end do
|
||||
|
||||
END_PROVIDER
|
||||
|
||||
|
||||
|
@ -23,7 +23,7 @@ interface: ezfio
|
||||
type: Threshold
|
||||
doc: Threshold on the convergence of the dressed CI energy
|
||||
interface: ezfio,provider,ocaml
|
||||
default: 1.e-4
|
||||
default: 5.e-5
|
||||
|
||||
[n_it_max_dressed_ci]
|
||||
type: Strictly_positive_int
|
||||
|
@ -299,7 +299,7 @@ subroutine mrcc_part_dress(delta_ij_, delta_ii_,i_generator,n_selected,det_buffe
|
||||
delta_ii_(i_state,i_I) = 0.d0
|
||||
do l_sd=1,idx_alpha(0)
|
||||
k_sd = idx_alpha(l_sd)
|
||||
delta_ij_(i_state,k_sd,i_I) = delta_ij_(i_state,k_sd,i_I) + dIa_hla(i_state,k_sd)
|
||||
delta_ij_(i_state,k_sd,i_I) = delta_ij_(i_state,k_sd,i_I) + 0.5d0*dIa_hla(i_state,k_sd)
|
||||
enddo
|
||||
endif
|
||||
enddo
|
||||
@ -554,7 +554,7 @@ END_PROVIDER
|
||||
do k=1,N_det_non_ref
|
||||
|
||||
call i_h_j(psi_ref(1,1,j), psi_non_ref(1,1,k),N_int,Hjk)
|
||||
call i_h_j(psi_non_ref(1,1,k),psi_ref(1,1,i), N_int,Hki)
|
||||
! call i_h_j(psi_non_ref(1,1,k),psi_ref(1,1,i), N_int,Hki)
|
||||
|
||||
delta_cas(i,j,i_state) += Hjk * dij(i, k, i_state) ! * Hki * lambda_mrcc(i_state, k)
|
||||
!print *, Hjk * get_dij(psi_ref(1,1,i), psi_non_ref(1,1,k), N_int), Hki * get_dij(psi_ref(1,1,j), psi_non_ref(1,1,k), N_int)
|
||||
@ -647,7 +647,7 @@ end function
|
||||
integer :: p1,p2,h1,h2,s1,s2, p1_,p2_,h1_,h2_,s1_,s2_, sortRefIdx(N_det_ref)
|
||||
logical :: ok
|
||||
double precision :: phase_iI, phase_Ik, phase_Jl, phase_IJ, phase_al, diI, hIi, hJi, delta_JI, dkI(1), HkI, ci_inv(1), dia_hla(1)
|
||||
double precision :: contrib, HIIi, HJk, wall
|
||||
double precision :: contrib, contrib2, HIIi, HJk, wall
|
||||
integer, dimension(0:2,2,2) :: exc_iI, exc_Ik, exc_IJ
|
||||
integer(bit_kind) :: det_tmp(N_int, 2), made_hole(N_int,2), made_particle(N_int,2), myActive(N_int,2)
|
||||
integer(bit_kind),allocatable :: sortRef(:,:,:)
|
||||
@ -677,7 +677,7 @@ end function
|
||||
delta_mrcepa0_ij(:,:,:) = 0d0
|
||||
|
||||
!$OMP PARALLEL DO default(none) schedule(dynamic) shared(delta_mrcepa0_ij, delta_mrcepa0_ii) &
|
||||
!$OMP private(m,i,II,J,k,degree,myActive,made_hole,made_particle,hjk,contrib) &
|
||||
!$OMP private(m,i,II,J,k,degree,myActive,made_hole,made_particle,hjk,contrib,contrib2) &
|
||||
!$OMP shared(active_sorb, psi_non_ref, psi_non_ref_coef, psi_ref, psi_ref_coef, cepa0_shortcut, det_cepa0_active) &
|
||||
!$OMP shared(N_det_ref, N_det_non_ref,N_int,det_cepa0_idx,lambda_mrcc,det_ref_active, delta_cas) &
|
||||
!$OMP shared(notf,i_state, sortRef, sortRefIdx, dij)
|
||||
@ -720,16 +720,18 @@ end function
|
||||
!$OMP ATOMIC
|
||||
notf = notf+1
|
||||
|
||||
call i_h_j(psi_non_ref(1,1,det_cepa0_idx(k)),psi_ref(1,1,J),N_int,HJk)
|
||||
!contrib = delta_cas(II, J, i_state) * HJk * lambda_mrcc(i_state, det_cepa0_idx(k))
|
||||
! call i_h_j(psi_non_ref(1,1,det_cepa0_idx(k)),psi_ref(1,1,J),N_int,HJk)
|
||||
contrib = delta_cas(II, J, i_state) * dij(J, det_cepa0_idx(k), i_state)
|
||||
!$OMP ATOMIC
|
||||
delta_mrcepa0_ij(J, det_cepa0_idx(i), i_state) += contrib
|
||||
|
||||
if(dabs(psi_ref_coef(J,i_state)).ge.5.d-5) then
|
||||
contrib2 = contrib / psi_ref_coef(J, i_state) * psi_non_ref_coef(det_cepa0_idx(i),i_state)
|
||||
!$OMP ATOMIC
|
||||
delta_mrcepa0_ii(J,i_state) -= contrib / psi_ref_coef(J, i_state) * psi_non_ref_coef(det_cepa0_idx(i),i_state)
|
||||
delta_mrcepa0_ii(J,i_state) -= contrib2
|
||||
else
|
||||
contrib = contrib * 0.5d0
|
||||
end if
|
||||
!$OMP ATOMIC
|
||||
delta_mrcepa0_ij(J, det_cepa0_idx(i), i_state) += contrib
|
||||
|
||||
end do kloop
|
||||
end do
|
||||
@ -753,7 +755,7 @@ END_PROVIDER
|
||||
integer :: p1,p2,h1,h2,s1,s2, p1_,p2_,h1_,h2_,s1_,s2_
|
||||
logical :: ok
|
||||
double precision :: phase_Ji, phase_Ik, phase_Ii
|
||||
double precision :: contrib, delta_IJk, HJk, HIk, HIl
|
||||
double precision :: contrib, contrib2, delta_IJk, HJk, HIk, HIl
|
||||
integer, dimension(0:2,2,2) :: exc_Ik, exc_Ji, exc_Ii
|
||||
integer(bit_kind) :: det_tmp(N_int, 2), det_tmp2(N_int, 2)
|
||||
integer, allocatable :: idx_sorted_bit(:)
|
||||
@ -778,7 +780,7 @@ END_PROVIDER
|
||||
!$OMP PARALLEL DO default(none) schedule(dynamic,10) shared(delta_sub_ij, delta_sub_ii) &
|
||||
!$OMP private(i, J, k, degree, degree2, l, deg, ni) &
|
||||
!$OMP private(p1,p2,h1,h2,s1,s2, p1_,p2_,h1_,h2_,s1_,s2_) &
|
||||
!$OMP private(ok, phase_Ji, phase_Ik, phase_Ii, contrib, delta_IJk, HJk, HIk, HIl, exc_Ik, exc_Ji, exc_Ii) &
|
||||
!$OMP private(ok, phase_Ji, phase_Ik, phase_Ii, contrib2, contrib, delta_IJk, HJk, HIk, HIl, exc_Ik, exc_Ji, exc_Ii) &
|
||||
!$OMP private(det_tmp, det_tmp2, II, blok) &
|
||||
!$OMP shared(idx_sorted_bit, N_det_non_ref, N_det_ref, N_int, psi_non_ref, psi_non_ref_coef, psi_ref, psi_ref_coef) &
|
||||
!$OMP shared(i_state,lambda_mrcc, hf_bitmask, active_sorb)
|
||||
@ -828,12 +830,15 @@ END_PROVIDER
|
||||
call apply_excitation(psi_non_ref(1,1,i),exc_Ik,det_tmp,ok,N_int)
|
||||
if(ok) cycle
|
||||
contrib = delta_IJk * HIl * lambda_mrcc(i_state,l)
|
||||
if(dabs(psi_ref_coef(II,i_state)).ge.5.d-5) then
|
||||
contrib2 = contrib / psi_ref_coef(II, i_state) * psi_non_ref_coef(l,i_state)
|
||||
!$OMP ATOMIC
|
||||
delta_sub_ii(II,i_state) -= contrib2
|
||||
else
|
||||
contrib = contrib * 0.5d0
|
||||
endif
|
||||
!$OMP ATOMIC
|
||||
delta_sub_ij(II, i, i_state) += contrib
|
||||
if(dabs(psi_ref_coef(II,i_state)).ge.5.d-5) then
|
||||
!$OMP ATOMIC
|
||||
delta_sub_ii(II,i_state) -= contrib / psi_ref_coef(II, i_state) * psi_non_ref_coef(l,i_state)
|
||||
endif
|
||||
end do
|
||||
end do
|
||||
end do
|
||||
|
Loading…
Reference in New Issue
Block a user