mirror of
https://github.com/QuantumPackage/qp2.git
synced 2024-11-03 20:53:54 +01:00
508 lines
15 KiB
Fortran
508 lines
15 KiB
Fortran
|
|
|
|
! ---
|
|
|
|
subroutine LTxSxR(n, m, L, S, R, C)
|
|
|
|
implicit none
|
|
integer, intent(in) :: n, m
|
|
double precision, intent(in) :: L(n,m), S(n,n), R(n,m)
|
|
double precision, intent(out) :: C(m,m)
|
|
integer :: i, j
|
|
double precision :: accu_d, accu_nd
|
|
double precision, allocatable :: tmp(:,:)
|
|
|
|
! L.T x S x R
|
|
allocate(tmp(m,n))
|
|
call dgemm( 'T', 'N', m, n, n, 1.d0 &
|
|
, L, size(L, 1), S, size(S, 1) &
|
|
, 0.d0, tmp, size(tmp, 1) )
|
|
call dgemm( 'N', 'N', m, m, n, 1.d0 &
|
|
, tmp, size(tmp, 1), R, size(R, 1) &
|
|
, 0.d0, C, size(C, 1) )
|
|
deallocate(tmp)
|
|
|
|
accu_d = 0.d0
|
|
accu_nd = 0.d0
|
|
do i = 1, m
|
|
do j = 1, m
|
|
if(j.eq.i) then
|
|
accu_d += dabs(C(j,i))
|
|
else
|
|
accu_nd += C(j,i) * C(j,i)
|
|
endif
|
|
enddo
|
|
enddo
|
|
accu_nd = dsqrt(accu_nd)
|
|
|
|
print*, ' accu_d = ', accu_d
|
|
print*, ' accu_nd = ', accu_nd
|
|
|
|
end subroutine LTxR
|
|
|
|
! ---
|
|
|
|
|
|
! ---
|
|
|
|
subroutine minimize_tc_orb_angles()
|
|
|
|
BEGIN_DOC
|
|
! routine that minimizes the angle between left- and right-orbitals when degeneracies are found
|
|
END_DOC
|
|
|
|
implicit none
|
|
logical :: good_angles
|
|
integer :: i
|
|
double precision :: thr_deg
|
|
|
|
good_angles = .False.
|
|
thr_deg = thr_degen_tc
|
|
|
|
call print_energy_and_mos(good_angles)
|
|
|
|
print *, ' Minimizing the angles between the TC orbitals'
|
|
i = 1
|
|
do while (.not. good_angles)
|
|
print *, ' iteration = ', i
|
|
call routine_save_rotated_mos(thr_deg, good_angles)
|
|
thr_deg *= 10.d0
|
|
i += 1
|
|
if(i .gt. 100) then
|
|
print *, ' minimize_tc_orb_angles does not seem to converge ..'
|
|
print *, ' Something is weird in the tc orbitals ...'
|
|
print *, ' STOPPING'
|
|
stop
|
|
endif
|
|
enddo
|
|
print *, ' Converged ANGLES MINIMIZATION !!'
|
|
|
|
call print_angles_tc()
|
|
call print_energy_and_mos(good_angles)
|
|
|
|
end
|
|
|
|
! ---
|
|
|
|
subroutine routine_save_rotated_mos(thr_deg, good_angles)
|
|
|
|
implicit none
|
|
|
|
double precision, intent(in) :: thr_deg
|
|
logical, intent(out) :: good_angles
|
|
|
|
integer :: i, j, k, n_degen_list, m, n, n_degen, ilast, ifirst
|
|
double precision :: max_angle, norm
|
|
double precision :: E_old, E_new, E_thr
|
|
integer, allocatable :: list_degen(:,:)
|
|
double precision, allocatable :: new_angles(:)
|
|
double precision, allocatable :: mo_r_coef_old(:,:), mo_l_coef_old(:,:)
|
|
double precision, allocatable :: mo_r_coef_good(:,:), mo_l_coef_good(:,:)
|
|
double precision, allocatable :: mo_r_coef_new(:,:)
|
|
double precision, allocatable :: fock_diag(:), s_mat(:,:)
|
|
double precision, allocatable :: stmp(:,:), T(:,:), Snew(:,:), smat2(:,:)
|
|
double precision, allocatable :: mo_l_coef_tmp(:,:), mo_r_coef_tmp(:,:), mo_l_coef_new(:,:)
|
|
|
|
E_thr = 1d-04
|
|
E_old = TC_HF_energy
|
|
allocate(mo_l_coef_old(ao_num,mo_num), mo_r_coef_old(ao_num,mo_num))
|
|
mo_r_coef_old = mo_r_coef
|
|
mo_l_coef_old = mo_l_coef
|
|
|
|
good_angles = .False.
|
|
|
|
allocate(mo_l_coef_good(ao_num, mo_num), mo_r_coef_good(ao_num,mo_num))
|
|
|
|
print *, ' ***************************************'
|
|
print *, ' ***************************************'
|
|
print *, ' THRESHOLD FOR DEGENERACIES ::: ', thr_deg
|
|
print *, ' ***************************************'
|
|
print *, ' ***************************************'
|
|
print *, ' Starting with the following TC energy gradient :', grad_non_hermit
|
|
|
|
mo_r_coef_good = mo_r_coef
|
|
mo_l_coef_good = mo_l_coef
|
|
|
|
allocate(mo_r_coef_new(ao_num, mo_num))
|
|
mo_r_coef_new = mo_r_coef
|
|
do i = 1, mo_num
|
|
norm = 1.d0/dsqrt(overlap_mo_r(i,i))
|
|
do j = 1, ao_num
|
|
mo_r_coef_new(j,i) *= norm
|
|
enddo
|
|
enddo
|
|
|
|
allocate(list_degen(mo_num,0:mo_num), s_mat(mo_num,mo_num), fock_diag(mo_num))
|
|
do i = 1, mo_num
|
|
fock_diag(i) = Fock_matrix_tc_mo_tot(i,i)
|
|
enddo
|
|
|
|
! compute the overlap between the left and rescaled right
|
|
call build_s_matrix(ao_num, mo_num, mo_r_coef_new, mo_r_coef_new, ao_overlap, s_mat)
|
|
! call give_degen(fock_diag,mo_num,thr_deg,list_degen,n_degen_list)
|
|
if(n_core_orb.ne.0)then
|
|
call give_degen_full_listcore(fock_diag, mo_num, list_core, n_core_orb, thr_deg, list_degen, n_degen_list)
|
|
else
|
|
call give_degen_full_list(fock_diag, mo_num, thr_deg, list_degen, n_degen_list)
|
|
endif
|
|
print *, ' fock_matrix_mo'
|
|
do i = 1, mo_num
|
|
print *, i, fock_diag(i), angle_left_right(i)
|
|
enddo
|
|
|
|
do i = 1, n_degen_list
|
|
! ifirst = list_degen(1,i)
|
|
! ilast = list_degen(2,i)
|
|
! n_degen = ilast - ifirst +1
|
|
|
|
n_degen = list_degen(i,0)
|
|
if(n_degen .ge. 1000)n_degen = 1 ! convention for core orbitals
|
|
|
|
if(n_degen .eq. 1) cycle
|
|
|
|
allocate(stmp(n_degen,n_degen), smat2(n_degen,n_degen))
|
|
allocate(mo_r_coef_tmp(ao_num,n_degen), mo_l_coef_tmp(ao_num,n_degen), mo_l_coef_new(ao_num,n_degen))
|
|
allocate(T(n_degen,n_degen), Snew(n_degen,n_degen))
|
|
|
|
print*,'Right orbitals before'
|
|
do j = 1, n_degen
|
|
write(*,'(100(F16.10,X))') mo_r_coef_new(1:ao_num,list_degen(i,j))
|
|
enddo
|
|
print*,'Left orbitals before'
|
|
do j = 1, n_degen
|
|
write(*,'(100(F16.10,X))')mo_l_coef(1:ao_num,list_degen(i,j))
|
|
enddo
|
|
if(angle_left_right(list_degen(i,1)).gt.80.d0.and.n_degen==2)then
|
|
integer :: i_list, j_list
|
|
i_list = list_degen(i,1)
|
|
j_list = list_degen(i,2)
|
|
print*,'Huge angle !!! == ',angle_left_right(list_degen(i,1)),angle_left_right(list_degen(i,2))
|
|
print*,'i_list = ',i_list
|
|
print*,'i_list = ',j_list
|
|
print*,'Swapping left/right orbitals'
|
|
call print_strong_overlap(i_list, j_list)
|
|
mo_r_coef_tmp(1:ao_num,1) = mo_r_coef_new(1:ao_num,i_list)
|
|
mo_r_coef_tmp(1:ao_num,2) = mo_l_coef(1:ao_num,i_list)
|
|
mo_l_coef_tmp(1:ao_num,1) = mo_l_coef(1:ao_num,j_list)
|
|
mo_l_coef_tmp(1:ao_num,2) = mo_r_coef_new(1:ao_num,j_list)
|
|
else
|
|
do j = 1, n_degen
|
|
print*,'i_list = ',list_degen(i,j)
|
|
mo_r_coef_tmp(1:ao_num,j) = mo_r_coef_new(1:ao_num,list_degen(i,j))
|
|
mo_l_coef_tmp(1:ao_num,j) = mo_l_coef(1:ao_num,list_degen(i,j))
|
|
enddo
|
|
endif
|
|
print*,'Right orbitals '
|
|
do j = 1, n_degen
|
|
write(*,'(100(F16.10,X))')mo_r_coef_tmp(1:ao_num,j)
|
|
enddo
|
|
print*,'Left orbitals '
|
|
do j = 1, n_degen
|
|
write(*,'(100(F16.10,X))')mo_l_coef_tmp(1:ao_num,j)
|
|
enddo
|
|
! Orthogonalization of right functions
|
|
print *, ' Orthogonalization of RIGHT functions'
|
|
print *, ' ------------------------------------'
|
|
call orthog_functions(ao_num, n_degen, mo_r_coef_tmp, ao_overlap)
|
|
|
|
! Orthogonalization of left functions
|
|
print *, ' Orthogonalization of LEFT functions'
|
|
print *, ' ------------------------------------'
|
|
call orthog_functions(ao_num, n_degen, mo_l_coef_tmp, ao_overlap)
|
|
|
|
print *, ' Overlap left-right '
|
|
call build_s_matrix(ao_num, n_degen, mo_r_coef_tmp, mo_l_coef_tmp, ao_overlap, stmp)
|
|
do j = 1, n_degen
|
|
write(*,'(100(F8.4,X))') stmp(:,j)
|
|
enddo
|
|
call build_s_matrix(ao_num, n_degen, mo_l_coef_tmp, mo_l_coef_tmp, ao_overlap, stmp)
|
|
|
|
!print*,'LEFT/LEFT OVERLAP '
|
|
!do j = 1, n_degen
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
call build_s_matrix(ao_num, n_degen, mo_r_coef_tmp, mo_r_coef_tmp, ao_overlap, stmp)
|
|
!print*,'RIGHT/RIGHT OVERLAP '
|
|
!do j = 1, n_degen
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
|
|
if(maxovl_tc) then
|
|
T = 0.d0
|
|
Snew = 0.d0
|
|
call maxovl(n_degen, n_degen, stmp, T, Snew)
|
|
!print*,'overlap after'
|
|
!do j = 1, n_degen
|
|
! write(*,'(100(F16.10,X))')Snew(:,j)
|
|
!enddo
|
|
call dgemm( 'N', 'N', ao_num, n_degen, n_degen, 1.d0 &
|
|
, mo_l_coef_tmp, size(mo_l_coef_tmp, 1), T(1,1), size(T, 1) &
|
|
, 0.d0, mo_l_coef_new, size(mo_l_coef_new, 1) )
|
|
call build_s_matrix(ao_num, n_degen, mo_l_coef_new, mo_r_coef_tmp, ao_overlap, stmp)
|
|
!print*,'Overlap test'
|
|
!do j = 1, n_degen
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
else
|
|
mo_l_coef_new = mo_l_coef_tmp
|
|
endif
|
|
|
|
call impose_weighted_biorthog_svd(ao_num, n_degen, ao_overlap, mo_l_coef_new, mo_r_coef_tmp)
|
|
|
|
!call build_s_matrix(ao_num, n_degen, mo_l_coef_new, mo_r_coef_tmp, ao_overlap, stmp)
|
|
!print*,'LAST OVERLAP '
|
|
!do j = 1, n_degen
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
!call build_s_matrix(ao_num, n_degen, mo_l_coef_new, mo_l_coef_new, ao_overlap, stmp)
|
|
!print*,'LEFT OVERLAP '
|
|
!do j = 1, n_degen
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
!call build_s_matrix(ao_num, n_degen, mo_r_coef_tmp, mo_r_coef_tmp, ao_overlap, stmp)
|
|
!print*,'RIGHT OVERLAP '
|
|
!do j = 1, n_degen
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
do j = 1, n_degen
|
|
!!! mo_l_coef_good(1:ao_num,j+ifirst-1) = mo_l_coef_new(1:ao_num,j)
|
|
!!! mo_r_coef_good(1:ao_num,j+ifirst-1) = mo_r_coef_tmp(1:ao_num,j)
|
|
mo_l_coef_good(1:ao_num,list_degen(i,j)) = mo_l_coef_new(1:ao_num,j)
|
|
mo_r_coef_good(1:ao_num,list_degen(i,j)) = mo_r_coef_tmp(1:ao_num,j)
|
|
enddo
|
|
|
|
deallocate(stmp, smat2)
|
|
deallocate(mo_r_coef_tmp, mo_l_coef_tmp, mo_l_coef_new)
|
|
deallocate(T, Snew)
|
|
enddo
|
|
|
|
!allocate(stmp(mo_num, mo_num))
|
|
!call build_s_matrix(ao_num, mo_num, mo_l_coef_good, mo_r_coef_good, ao_overlap, stmp)
|
|
!print*,'LEFT/RIGHT OVERLAP '
|
|
!do j = 1, mo_num
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
!call build_s_matrix(ao_num, mo_num, mo_l_coef_good, mo_l_coef_good, ao_overlap, stmp)
|
|
!print*,'LEFT/LEFT OVERLAP '
|
|
!do j = 1, mo_num
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
!call build_s_matrix(ao_num, mo_num, mo_r_coef_good, mo_r_coef_good, ao_overlap, stmp)
|
|
!print*,'RIGHT/RIGHT OVERLAP '
|
|
!do j = 1, mo_num
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
|
|
mo_r_coef = mo_r_coef_good
|
|
mo_l_coef = mo_l_coef_good
|
|
call ezfio_set_bi_ortho_mos_mo_l_coef(mo_l_coef)
|
|
call ezfio_set_bi_ortho_mos_mo_r_coef(mo_r_coef)
|
|
TOUCH mo_l_coef mo_r_coef
|
|
|
|
! check if TC energy has changed
|
|
E_new = TC_HF_energy
|
|
if(dabs(E_new - E_old) .gt. E_thr) then
|
|
mo_r_coef = mo_r_coef_old
|
|
mo_l_coef = mo_l_coef_old
|
|
deallocate(mo_l_coef_old, mo_r_coef_old)
|
|
call ezfio_set_bi_ortho_mos_mo_l_coef(mo_l_coef)
|
|
call ezfio_set_bi_ortho_mos_mo_r_coef(mo_r_coef)
|
|
TOUCH mo_l_coef mo_r_coef
|
|
print*, ' TC energy bef rotation = ', E_old
|
|
print*, ' TC energy aft rotation = ', E_new
|
|
print*, ' the rotation is refused'
|
|
stop
|
|
endif
|
|
|
|
allocate(new_angles(mo_num))
|
|
new_angles(1:mo_num) = dabs(angle_left_right(1:mo_num))
|
|
max_angle = maxval(new_angles)
|
|
good_angles = max_angle.lt.thresh_lr_angle
|
|
print *, ' max_angle = ', max_angle
|
|
deallocate(new_angles)
|
|
|
|
|
|
deallocate(mo_l_coef_old, mo_r_coef_old)
|
|
deallocate(mo_l_coef_good, mo_r_coef_good)
|
|
deallocate(mo_r_coef_new)
|
|
|
|
end
|
|
|
|
! ---
|
|
|
|
subroutine build_s_matrix(m, n, C1, C2, overlap, smat)
|
|
|
|
implicit none
|
|
integer, intent(in) :: m, n
|
|
double precision, intent(in) :: C1(m,n), C2(m,n), overlap(m,m)
|
|
double precision, intent(out) :: smat(n,n)
|
|
integer :: i, j, k, l
|
|
double precision, allocatable :: S_tmp(:,:)
|
|
|
|
smat = 0.d0
|
|
|
|
!do i = 1, n
|
|
! do j = 1, n
|
|
! do k = 1, m
|
|
! do l = 1, m
|
|
! smat(i,j) += C1(k,i) * overlap(l,k) * C2(l,j)
|
|
! enddo
|
|
! enddo
|
|
! enddo
|
|
!enddo
|
|
|
|
! C1.T x overlap
|
|
allocate(S_tmp(n,m))
|
|
call dgemm( 'T', 'N', n, m, m, 1.d0 &
|
|
, C1, size(C1, 1), overlap, size(overlap, 1) &
|
|
, 0.d0, S_tmp, size(S_tmp, 1) )
|
|
! C1.T x overlap x C2
|
|
call dgemm( 'N', 'N', n, n, m, 1.d0 &
|
|
, S_tmp, size(S_tmp, 1), C2(1,1), size(C2, 1) &
|
|
, 0.d0, smat, size(smat, 1) )
|
|
deallocate(S_tmp)
|
|
|
|
end
|
|
|
|
! ---
|
|
|
|
subroutine orthog_functions(m, n, coef, overlap)
|
|
|
|
implicit none
|
|
|
|
integer, intent(in) :: m, n
|
|
double precision, intent(in) :: overlap(m,m)
|
|
double precision, intent(inout) :: coef(m,n)
|
|
double precision, allocatable :: stmp(:,:)
|
|
integer :: j, k
|
|
|
|
allocate(stmp(n,n))
|
|
call build_s_matrix(m, n, coef, coef, overlap, stmp)
|
|
! print*,'overlap before'
|
|
! do j = 1, n
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
! enddo
|
|
call impose_orthog_svd_overlap(m, n, coef, overlap)
|
|
call build_s_matrix(m, n, coef, coef, overlap, stmp)
|
|
do j = 1, n
|
|
! ---
|
|
! TODO: MANU check ici
|
|
!coef(1,:m) *= 1.d0/dsqrt(stmp(j,j))
|
|
do k = 1, m
|
|
coef(k,j) *= 1.d0/dsqrt(stmp(j,j))
|
|
enddo
|
|
! ---
|
|
enddo
|
|
call build_s_matrix(m, n, coef, coef, overlap, stmp)
|
|
|
|
!print*,'overlap after'
|
|
!do j = 1, n
|
|
! write(*,'(100(F16.10,X))')stmp(:,j)
|
|
!enddo
|
|
|
|
deallocate(stmp)
|
|
|
|
end
|
|
|
|
! ---
|
|
|
|
subroutine print_angles_tc()
|
|
|
|
implicit none
|
|
integer :: i, j
|
|
double precision :: left, right
|
|
|
|
print *, ' product of norms, angle between vectors'
|
|
do i = 1, mo_num
|
|
left = overlap_mo_l(i,i)
|
|
right = overlap_mo_r(i,i)
|
|
! print*,Fock_matrix_tc_mo_tot(i,i),left*right,angle_left_right(i)
|
|
print *, left*right, angle_left_right(i)
|
|
enddo
|
|
|
|
end
|
|
|
|
! ---
|
|
|
|
subroutine print_energy_and_mos(good_angles)
|
|
|
|
implicit none
|
|
logical, intent(out) :: good_angles
|
|
integer :: i
|
|
|
|
print *, ' '
|
|
print *, ' TC energy = ', TC_HF_energy
|
|
print *, ' TC SCF energy gradient = ', grad_non_hermit
|
|
print *, ' Max angle Left/right = ', max_angle_left_right
|
|
call print_angles_tc()
|
|
|
|
if(max_angle_left_right .lt. thresh_lr_angle) then
|
|
print *, ' Maximum angle BELOW 45 degrees, everthing is OK !'
|
|
good_angles = .true.
|
|
! else if(max_angle_left_right .gt. thresh_lr_angle .and. max_angle_left_right .lt. 75.d0) then
|
|
! print *, ' Maximum angle between thresh_lr_angle and 75 degrees, this is not the best for TC-CI calculations ...'
|
|
! good_angles = .false.
|
|
! else if(max_angle_left_right .gt. 75.d0) then
|
|
! print *, ' Maximum angle between ABOVE 75 degrees, YOU WILL CERTAINLY FIND TROUBLES IN TC-CI calculations ...'
|
|
! good_angles = .false.
|
|
endif
|
|
!
|
|
! print *, ' Diag Fock elem, product of left/right norm, angle left/right '
|
|
! do i = 1, mo_num
|
|
! write(*, '(I3,X,100(F16.10,X))') i, Fock_matrix_tc_mo_tot(i,i), overlap_mo_l(i,i)*overlap_mo_r(i,i), angle_left_right(i)
|
|
! enddo
|
|
|
|
end
|
|
|
|
! ---
|
|
|
|
subroutine sort_by_tc_fock
|
|
implicit none
|
|
integer, allocatable :: iorder(:)
|
|
double precision, allocatable :: mo_l_tmp(:,:), mo_r_tmp(:,:),fock(:)
|
|
allocate(iorder(mo_num),fock(mo_num),mo_l_tmp(ao_num, mo_num),mo_r_tmp(ao_num,mo_num))
|
|
integer :: i
|
|
mo_l_tmp = mo_l_coef
|
|
mo_r_tmp = mo_r_coef
|
|
do i = 1, mo_num
|
|
iorder(i) = i
|
|
fock(i) = Fock_matrix_tc_mo_tot(i,i)
|
|
enddo
|
|
call dsort(fock,iorder,mo_num)
|
|
do i = 1, mo_num
|
|
mo_l_coef(1:ao_num,i) = mo_l_tmp(1:ao_num,iorder(i))
|
|
mo_r_coef(1:ao_num,i) = mo_r_tmp(1:ao_num,iorder(i))
|
|
enddo
|
|
touch mo_l_coef mo_r_coef
|
|
|
|
end
|
|
|
|
|
|
subroutine print_strong_overlap(i_list, j_list)
|
|
implicit none
|
|
integer, intent(in) :: i_list,j_list
|
|
double precision :: o_i, o_j,o_ij
|
|
double precision :: s_mat_r(2,2),s_mat_l(2,2)
|
|
o_i = dsqrt(overlap_mo_r(i_list, i_list))
|
|
o_j = dsqrt(overlap_mo_r(j_list, j_list))
|
|
o_ij = overlap_mo_r(j_list, i_list)
|
|
s_mat_r(1,1) = o_i*o_i
|
|
s_mat_r(2,1) = o_ij/(o_i * o_j)
|
|
s_mat_r(2,2) = o_j*o_j
|
|
s_mat_r(1,2) = s_mat_r(2,1)
|
|
print*,'Right overlap matrix '
|
|
write(*,'(2(F10.5,X))')s_mat_r(1:2,1)
|
|
write(*,'(2(F10.5,X))')s_mat_r(1:2,2)
|
|
o_i = dsqrt(overlap_mo_l(i_list, i_list))
|
|
o_j = dsqrt(overlap_mo_l(j_list, j_list))
|
|
o_ij = overlap_mo_l(j_list, i_list)
|
|
s_mat_l(1,1) = o_i*o_i
|
|
s_mat_l(2,1) = o_ij/(o_i * o_j)
|
|
s_mat_l(2,2) = o_j*o_j
|
|
s_mat_l(1,2) = s_mat_l(2,1)
|
|
print*,'Left overlap matrix '
|
|
write(*,'(2(F10.5,X))')s_mat_l(1:2,1)
|
|
write(*,'(2(F10.5,X))')s_mat_l(1:2,2)
|
|
|
|
end
|